linux/include/linux/reiserfs_fs.h
<<
>>
Prefs
   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
   6                                   name, yura please fix it to be
   7                                   reiserfs.h, and merge all the rest
   8                                   of our .h files that are in this
   9                                   directory into it.  */
  10
  11#ifndef _LINUX_REISER_FS_H
  12#define _LINUX_REISER_FS_H
  13
  14#include <linux/types.h>
  15#include <linux/magic.h>
  16
  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/buffer_head.h>
  26#include <linux/reiserfs_fs_i.h>
  27#include <linux/reiserfs_fs_sb.h>
  28#endif
  29
  30/*
  31 *  include/linux/reiser_fs.h
  32 *
  33 *  Reiser File System constants and structures
  34 *
  35 */
  36
  37/* ioctl's command */
  38#define REISERFS_IOC_UNPACK             _IOW(0xCD,1,long)
  39/* define following flags to be the same as in ext2, so that chattr(1),
  40   lsattr(1) will work with us. */
  41#define REISERFS_IOC_GETFLAGS           FS_IOC_GETFLAGS
  42#define REISERFS_IOC_SETFLAGS           FS_IOC_SETFLAGS
  43#define REISERFS_IOC_GETVERSION         FS_IOC_GETVERSION
  44#define REISERFS_IOC_SETVERSION         FS_IOC_SETVERSION
  45
  46#ifdef __KERNEL__
  47/* the 32 bit compat definitions with int argument */
  48#define REISERFS_IOC32_UNPACK           _IOW(0xCD, 1, int)
  49#define REISERFS_IOC32_GETFLAGS         FS_IOC32_GETFLAGS
  50#define REISERFS_IOC32_SETFLAGS         FS_IOC32_SETFLAGS
  51#define REISERFS_IOC32_GETVERSION       FS_IOC32_GETVERSION
  52#define REISERFS_IOC32_SETVERSION       FS_IOC32_SETVERSION
  53
  54/*
  55 * Locking primitives. The write lock is a per superblock
  56 * special mutex that has properties close to the Big Kernel Lock
  57 * which was used in the previous locking scheme.
  58 */
  59void reiserfs_write_lock(struct super_block *s);
  60void reiserfs_write_unlock(struct super_block *s);
  61int reiserfs_write_lock_once(struct super_block *s);
  62void reiserfs_write_unlock_once(struct super_block *s, int lock_depth);
  63
  64#ifdef CONFIG_REISERFS_CHECK
  65void reiserfs_lock_check_recursive(struct super_block *s);
  66#else
  67static inline void reiserfs_lock_check_recursive(struct super_block *s) { }
  68#endif
  69
  70/*
  71 * Several mutexes depend on the write lock.
  72 * However sometimes we want to relax the write lock while we hold
  73 * these mutexes, according to the release/reacquire on schedule()
  74 * properties of the Bkl that were used.
  75 * Reiserfs performances and locking were based on this scheme.
  76 * Now that the write lock is a mutex and not the bkl anymore, doing so
  77 * may result in a deadlock:
  78 *
  79 * A acquire write_lock
  80 * A acquire j_commit_mutex
  81 * A release write_lock and wait for something
  82 * B acquire write_lock
  83 * B can't acquire j_commit_mutex and sleep
  84 * A can't acquire write lock anymore
  85 * deadlock
  86 *
  87 * What we do here is avoiding such deadlock by playing the same game
  88 * than the Bkl: if we can't acquire a mutex that depends on the write lock,
  89 * we release the write lock, wait a bit and then retry.
  90 *
  91 * The mutexes concerned by this hack are:
  92 * - The commit mutex of a journal list
  93 * - The flush mutex
  94 * - The journal lock
  95 * - The inode mutex
  96 */
  97static inline void reiserfs_mutex_lock_safe(struct mutex *m,
  98                               struct super_block *s)
  99{
 100        reiserfs_lock_check_recursive(s);
 101        reiserfs_write_unlock(s);
 102        mutex_lock(m);
 103        reiserfs_write_lock(s);
 104}
 105
 106static inline void
 107reiserfs_mutex_lock_nested_safe(struct mutex *m, unsigned int subclass,
 108                               struct super_block *s)
 109{
 110        reiserfs_lock_check_recursive(s);
 111        reiserfs_write_unlock(s);
 112        mutex_lock_nested(m, subclass);
 113        reiserfs_write_lock(s);
 114}
 115
 116static inline void
 117reiserfs_down_read_safe(struct rw_semaphore *sem, struct super_block *s)
 118{
 119        reiserfs_lock_check_recursive(s);
 120        reiserfs_write_unlock(s);
 121        down_read(sem);
 122        reiserfs_write_lock(s);
 123}
 124
 125/*
 126 * When we schedule, we usually want to also release the write lock,
 127 * according to the previous bkl based locking scheme of reiserfs.
 128 */
 129static inline void reiserfs_cond_resched(struct super_block *s)
 130{
 131        if (need_resched()) {
 132                reiserfs_write_unlock(s);
 133                schedule();
 134                reiserfs_write_lock(s);
 135        }
 136}
 137
 138struct fid;
 139
 140/* in reading the #defines, it may help to understand that they employ
 141   the following abbreviations:
 142
 143   B = Buffer
 144   I = Item header
 145   H = Height within the tree (should be changed to LEV)
 146   N = Number of the item in the node
 147   STAT = stat data
 148   DEH = Directory Entry Header
 149   EC = Entry Count
 150   E = Entry number
 151   UL = Unsigned Long
 152   BLKH = BLocK Header
 153   UNFM = UNForMatted node
 154   DC = Disk Child
 155   P = Path
 156
 157   These #defines are named by concatenating these abbreviations,
 158   where first comes the arguments, and last comes the return value,
 159   of the macro.
 160
 161*/
 162
 163#define USE_INODE_GENERATION_COUNTER
 164
 165#define REISERFS_PREALLOCATE
 166#define DISPLACE_NEW_PACKING_LOCALITIES
 167#define PREALLOCATION_SIZE 9
 168
 169/* n must be power of 2 */
 170#define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
 171
 172// to be ok for alpha and others we have to align structures to 8 byte
 173// boundary.
 174// FIXME: do not change 4 by anything else: there is code which relies on that
 175#define ROUND_UP(x) _ROUND_UP(x,8LL)
 176
 177/* debug levels.  Right now, CONFIG_REISERFS_CHECK means print all debug
 178** messages.
 179*/
 180#define REISERFS_DEBUG_CODE 5   /* extra messages to help find/debug errors */
 181
 182void __reiserfs_warning(struct super_block *s, const char *id,
 183                         const char *func, const char *fmt, ...);
 184#define reiserfs_warning(s, id, fmt, args...) \
 185         __reiserfs_warning(s, id, __func__, fmt, ##args)
 186/* assertions handling */
 187
 188/** always check a condition and panic if it's false. */
 189#define __RASSERT(cond, scond, format, args...)                 \
 190do {                                                                    \
 191        if (!(cond))                                                    \
 192                reiserfs_panic(NULL, "assertion failure", "(" #cond ") at " \
 193                               __FILE__ ":%i:%s: " format "\n",         \
 194                               in_interrupt() ? -1 : task_pid_nr(current), \
 195                               __LINE__, __func__ , ##args);            \
 196} while (0)
 197
 198#define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args)
 199
 200#if defined( CONFIG_REISERFS_CHECK )
 201#define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args)
 202#else
 203#define RFALSE( cond, format, args... ) do {;} while( 0 )
 204#endif
 205
 206#define CONSTF __attribute_const__
 207/*
 208 * Disk Data Structures
 209 */
 210
 211/***************************************************************************/
 212/*                             SUPER BLOCK                                 */
 213/***************************************************************************/
 214
 215/*
 216 * Structure of super block on disk, a version of which in RAM is often accessed as REISERFS_SB(s)->s_rs
 217 * the version in RAM is part of a larger structure containing fields never written to disk.
 218 */
 219#define UNSET_HASH 0            // read_super will guess about, what hash names
 220                     // in directories were sorted with
 221#define TEA_HASH  1
 222#define YURA_HASH 2
 223#define R5_HASH   3
 224#define DEFAULT_HASH R5_HASH
 225
 226struct journal_params {
 227        __le32 jp_journal_1st_block;    /* where does journal start from on its
 228                                         * device */
 229        __le32 jp_journal_dev;  /* journal device st_rdev */
 230        __le32 jp_journal_size; /* size of the journal */
 231        __le32 jp_journal_trans_max;    /* max number of blocks in a transaction. */
 232        __le32 jp_journal_magic;        /* random value made on fs creation (this
 233                                         * was sb_journal_block_count) */
 234        __le32 jp_journal_max_batch;    /* max number of blocks to batch into a
 235                                         * trans */
 236        __le32 jp_journal_max_commit_age;       /* in seconds, how old can an async
 237                                                 * commit be */
 238        __le32 jp_journal_max_trans_age;        /* in seconds, how old can a transaction
 239                                                 * be */
 240};
 241
 242/* this is the super from 3.5.X, where X >= 10 */
 243struct reiserfs_super_block_v1 {
 244        __le32 s_block_count;   /* blocks count         */
 245        __le32 s_free_blocks;   /* free blocks count    */
 246        __le32 s_root_block;    /* root block number    */
 247        struct journal_params s_journal;
 248        __le16 s_blocksize;     /* block size */
 249        __le16 s_oid_maxsize;   /* max size of object id array, see
 250                                 * get_objectid() commentary  */
 251        __le16 s_oid_cursize;   /* current size of object id array */
 252        __le16 s_umount_state;  /* this is set to 1 when filesystem was
 253                                 * umounted, to 2 - when not */
 254        char s_magic[10];       /* reiserfs magic string indicates that
 255                                 * file system is reiserfs:
 256                                 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs" */
 257        __le16 s_fs_state;      /* it is set to used by fsck to mark which
 258                                 * phase of rebuilding is done */
 259        __le32 s_hash_function_code;    /* indicate, what hash function is being use
 260                                         * to sort names in a directory*/
 261        __le16 s_tree_height;   /* height of disk tree */
 262        __le16 s_bmap_nr;       /* amount of bitmap blocks needed to address
 263                                 * each block of file system */
 264        __le16 s_version;       /* this field is only reliable on filesystem
 265                                 * with non-standard journal */
 266        __le16 s_reserved_for_journal;  /* size in blocks of journal area on main
 267                                         * device, we need to keep after
 268                                         * making fs with non-standard journal */
 269} __attribute__ ((__packed__));
 270
 271#define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))
 272
 273/* this is the on disk super block */
 274struct reiserfs_super_block {
 275        struct reiserfs_super_block_v1 s_v1;
 276        __le32 s_inode_generation;
 277        __le32 s_flags;         /* Right now used only by inode-attributes, if enabled */
 278        unsigned char s_uuid[16];       /* filesystem unique identifier */
 279        unsigned char s_label[16];      /* filesystem volume label */
 280        __le16 s_mnt_count;             /* Count of mounts since last fsck */
 281        __le16 s_max_mnt_count;         /* Maximum mounts before check */
 282        __le32 s_lastcheck;             /* Timestamp of last fsck */
 283        __le32 s_check_interval;        /* Interval between checks */
 284        char s_unused[76];      /* zero filled by mkreiserfs and
 285                                 * reiserfs_convert_objectid_map_v1()
 286                                 * so any additions must be updated
 287                                 * there as well. */
 288} __attribute__ ((__packed__));
 289
 290#define SB_SIZE (sizeof(struct reiserfs_super_block))
 291
 292#define REISERFS_VERSION_1 0
 293#define REISERFS_VERSION_2 2
 294
 295// on-disk super block fields converted to cpu form
 296#define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)
 297#define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))
 298#define SB_BLOCKSIZE(s) \
 299        le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))
 300#define SB_BLOCK_COUNT(s) \
 301        le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))
 302#define SB_FREE_BLOCKS(s) \
 303        le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))
 304#define SB_REISERFS_MAGIC(s) \
 305        (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)
 306#define SB_ROOT_BLOCK(s) \
 307        le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))
 308#define SB_TREE_HEIGHT(s) \
 309        le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))
 310#define SB_REISERFS_STATE(s) \
 311        le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))
 312#define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))
 313#define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))
 314
 315#define PUT_SB_BLOCK_COUNT(s, val) \
 316   do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)
 317#define PUT_SB_FREE_BLOCKS(s, val) \
 318   do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)
 319#define PUT_SB_ROOT_BLOCK(s, val) \
 320   do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)
 321#define PUT_SB_TREE_HEIGHT(s, val) \
 322   do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)
 323#define PUT_SB_REISERFS_STATE(s, val) \
 324   do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)
 325#define PUT_SB_VERSION(s, val) \
 326   do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)
 327#define PUT_SB_BMAP_NR(s, val) \
 328   do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)
 329
 330#define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)
 331#define SB_ONDISK_JOURNAL_SIZE(s) \
 332         le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))
 333#define SB_ONDISK_JOURNAL_1st_BLOCK(s) \
 334         le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))
 335#define SB_ONDISK_JOURNAL_DEVICE(s) \
 336         le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
 337#define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
 338         le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
 339
 340#define is_block_in_log_or_reserved_area(s, block) \
 341         block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
 342         && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) +  \
 343         ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \
 344         SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))
 345
 346int is_reiserfs_3_5(struct reiserfs_super_block *rs);
 347int is_reiserfs_3_6(struct reiserfs_super_block *rs);
 348int is_reiserfs_jr(struct reiserfs_super_block *rs);
 349
 350/* ReiserFS leaves the first 64k unused, so that partition labels have
 351   enough space.  If someone wants to write a fancy bootloader that
 352   needs more than 64k, let us know, and this will be increased in size.
 353   This number must be larger than than the largest block size on any
 354   platform, or code will break.  -Hans */
 355#define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
 356#define REISERFS_FIRST_BLOCK unused_define
 357#define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES
 358
 359/* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
 360#define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
 361
 362/* reiserfs internal error code (used by search_by_key and fix_nodes)) */
 363#define CARRY_ON      0
 364#define REPEAT_SEARCH -1
 365#define IO_ERROR      -2
 366#define NO_DISK_SPACE -3
 367#define NO_BALANCING_NEEDED  (-4)
 368#define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
 369#define QUOTA_EXCEEDED -6
 370
 371typedef __u32 b_blocknr_t;
 372typedef __le32 unp_t;
 373
 374struct unfm_nodeinfo {
 375        unp_t unfm_nodenum;
 376        unsigned short unfm_freespace;
 377};
 378
 379/* there are two formats of keys: 3.5 and 3.6
 380 */
 381#define KEY_FORMAT_3_5 0
 382#define KEY_FORMAT_3_6 1
 383
 384/* there are two stat datas */
 385#define STAT_DATA_V1 0
 386#define STAT_DATA_V2 1
 387
 388static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)
 389{
 390        return container_of(inode, struct reiserfs_inode_info, vfs_inode);
 391}
 392
 393static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)
 394{
 395        return sb->s_fs_info;
 396}
 397
 398/* Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u16
 399 * which overflows on large file systems. */
 400static inline __u32 reiserfs_bmap_count(struct super_block *sb)
 401{
 402        return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1;
 403}
 404
 405static inline int bmap_would_wrap(unsigned bmap_nr)
 406{
 407        return bmap_nr > ((1LL << 16) - 1);
 408}
 409
 410/** this says about version of key of all items (but stat data) the
 411    object consists of */
 412#define get_inode_item_key_version( inode )                                    \
 413    ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
 414
 415#define set_inode_item_key_version( inode, version )                           \
 416         ({ if((version)==KEY_FORMAT_3_6)                                      \
 417                REISERFS_I(inode)->i_flags |= i_item_key_version_mask;      \
 418            else                                                               \
 419                REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })
 420
 421#define get_inode_sd_version(inode)                                            \
 422    ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
 423
 424#define set_inode_sd_version(inode, version)                                   \
 425         ({ if((version)==STAT_DATA_V2)                                        \
 426                REISERFS_I(inode)->i_flags |= i_stat_data_version_mask;     \
 427            else                                                               \
 428                REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })
 429
 430/* This is an aggressive tail suppression policy, I am hoping it
 431   improves our benchmarks. The principle behind it is that percentage
 432   space saving is what matters, not absolute space saving.  This is
 433   non-intuitive, but it helps to understand it if you consider that the
 434   cost to access 4 blocks is not much more than the cost to access 1
 435   block, if you have to do a seek and rotate.  A tail risks a
 436   non-linear disk access that is significant as a percentage of total
 437   time cost for a 4 block file and saves an amount of space that is
 438   less significant as a percentage of space, or so goes the hypothesis.
 439   -Hans */
 440#define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
 441(\
 442  (!(n_tail_size)) || \
 443  (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
 444   ( (n_file_size) >= (n_block_size) * 4 ) || \
 445   ( ( (n_file_size) >= (n_block_size) * 3 ) && \
 446     ( (n_tail_size) >=   (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
 447   ( ( (n_file_size) >= (n_block_size) * 2 ) && \
 448     ( (n_tail_size) >=   (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
 449   ( ( (n_file_size) >= (n_block_size) ) && \
 450     ( (n_tail_size) >=   (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
 451)
 452
 453/* Another strategy for tails, this one means only create a tail if all the
 454   file would fit into one DIRECT item.
 455   Primary intention for this one is to increase performance by decreasing
 456   seeking.
 457*/
 458#define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
 459(\
 460  (!(n_tail_size)) || \
 461  (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
 462)
 463
 464/*
 465 * values for s_umount_state field
 466 */
 467#define REISERFS_VALID_FS    1
 468#define REISERFS_ERROR_FS    2
 469
 470//
 471// there are 5 item types currently
 472//
 473#define TYPE_STAT_DATA 0
 474#define TYPE_INDIRECT 1
 475#define TYPE_DIRECT 2
 476#define TYPE_DIRENTRY 3
 477#define TYPE_MAXTYPE 3
 478#define TYPE_ANY 15             // FIXME: comment is required
 479
 480/***************************************************************************/
 481/*                       KEY & ITEM HEAD                                   */
 482/***************************************************************************/
 483
 484//
 485// directories use this key as well as old files
 486//
 487struct offset_v1 {
 488        __le32 k_offset;
 489        __le32 k_uniqueness;
 490} __attribute__ ((__packed__));
 491
 492struct offset_v2 {
 493        __le64 v;
 494} __attribute__ ((__packed__));
 495
 496static inline __u16 offset_v2_k_type(const struct offset_v2 *v2)
 497{
 498        __u8 type = le64_to_cpu(v2->v) >> 60;
 499        return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY;
 500}
 501
 502static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type)
 503{
 504        v2->v =
 505            (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60);
 506}
 507
 508static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2)
 509{
 510        return le64_to_cpu(v2->v) & (~0ULL >> 4);
 511}
 512
 513static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset)
 514{
 515        offset &= (~0ULL >> 4);
 516        v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset);
 517}
 518
 519/* Key of an item determines its location in the S+tree, and
 520   is composed of 4 components */
 521struct reiserfs_key {
 522        __le32 k_dir_id;        /* packing locality: by default parent
 523                                   directory object id */
 524        __le32 k_objectid;      /* object identifier */
 525        union {
 526                struct offset_v1 k_offset_v1;
 527                struct offset_v2 k_offset_v2;
 528        } __attribute__ ((__packed__)) u;
 529} __attribute__ ((__packed__));
 530
 531struct in_core_key {
 532        __u32 k_dir_id;         /* packing locality: by default parent
 533                                   directory object id */
 534        __u32 k_objectid;       /* object identifier */
 535        __u64 k_offset;
 536        __u8 k_type;
 537};
 538
 539struct cpu_key {
 540        struct in_core_key on_disk_key;
 541        int version;
 542        int key_length;         /* 3 in all cases but direct2indirect and
 543                                   indirect2direct conversion */
 544};
 545
 546/* Our function for comparing keys can compare keys of different
 547   lengths.  It takes as a parameter the length of the keys it is to
 548   compare.  These defines are used in determining what is to be passed
 549   to it as that parameter. */
 550#define REISERFS_FULL_KEY_LEN     4
 551#define REISERFS_SHORT_KEY_LEN    2
 552
 553/* The result of the key compare */
 554#define FIRST_GREATER 1
 555#define SECOND_GREATER -1
 556#define KEYS_IDENTICAL 0
 557#define KEY_FOUND 1
 558#define KEY_NOT_FOUND 0
 559
 560#define KEY_SIZE (sizeof(struct reiserfs_key))
 561#define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
 562
 563/* return values for search_by_key and clones */
 564#define ITEM_FOUND 1
 565#define ITEM_NOT_FOUND 0
 566#define ENTRY_FOUND 1
 567#define ENTRY_NOT_FOUND 0
 568#define DIRECTORY_NOT_FOUND -1
 569#define REGULAR_FILE_FOUND -2
 570#define DIRECTORY_FOUND -3
 571#define BYTE_FOUND 1
 572#define BYTE_NOT_FOUND 0
 573#define FILE_NOT_FOUND -1
 574
 575#define POSITION_FOUND 1
 576#define POSITION_NOT_FOUND 0
 577
 578// return values for reiserfs_find_entry and search_by_entry_key
 579#define NAME_FOUND 1
 580#define NAME_NOT_FOUND 0
 581#define GOTO_PREVIOUS_ITEM 2
 582#define NAME_FOUND_INVISIBLE 3
 583
 584/*  Everything in the filesystem is stored as a set of items.  The
 585    item head contains the key of the item, its free space (for
 586    indirect items) and specifies the location of the item itself
 587    within the block.  */
 588
 589struct item_head {
 590        /* Everything in the tree is found by searching for it based on
 591         * its key.*/
 592        struct reiserfs_key ih_key;
 593        union {
 594                /* The free space in the last unformatted node of an
 595                   indirect item if this is an indirect item.  This
 596                   equals 0xFFFF iff this is a direct item or stat data
 597                   item. Note that the key, not this field, is used to
 598                   determine the item type, and thus which field this
 599                   union contains. */
 600                __le16 ih_free_space_reserved;
 601                /* Iff this is a directory item, this field equals the
 602                   number of directory entries in the directory item. */
 603                __le16 ih_entry_count;
 604        } __attribute__ ((__packed__)) u;
 605        __le16 ih_item_len;     /* total size of the item body */
 606        __le16 ih_item_location;        /* an offset to the item body
 607                                         * within the block */
 608        __le16 ih_version;      /* 0 for all old items, 2 for new
 609                                   ones. Highest bit is set by fsck
 610                                   temporary, cleaned after all
 611                                   done */
 612} __attribute__ ((__packed__));
 613/* size of item header     */
 614#define IH_SIZE (sizeof(struct item_head))
 615
 616#define ih_free_space(ih)            le16_to_cpu((ih)->u.ih_free_space_reserved)
 617#define ih_version(ih)               le16_to_cpu((ih)->ih_version)
 618#define ih_entry_count(ih)           le16_to_cpu((ih)->u.ih_entry_count)
 619#define ih_location(ih)              le16_to_cpu((ih)->ih_item_location)
 620#define ih_item_len(ih)              le16_to_cpu((ih)->ih_item_len)
 621
 622#define put_ih_free_space(ih, val)   do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
 623#define put_ih_version(ih, val)      do { (ih)->ih_version = cpu_to_le16(val); } while (0)
 624#define put_ih_entry_count(ih, val)  do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
 625#define put_ih_location(ih, val)     do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
 626#define put_ih_item_len(ih, val)     do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
 627
 628#define unreachable_item(ih) (ih_version(ih) & (1 << 15))
 629
 630#define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
 631#define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
 632
 633/* these operate on indirect items, where you've got an array of ints
 634** at a possibly unaligned location.  These are a noop on ia32
 635** 
 636** p is the array of __u32, i is the index into the array, v is the value
 637** to store there.
 638*/
 639#define get_block_num(p, i) get_unaligned_le32((p) + (i))
 640#define put_block_num(p, i, v) put_unaligned_le32((v), (p) + (i))
 641
 642//
 643// in old version uniqueness field shows key type
 644//
 645#define V1_SD_UNIQUENESS 0
 646#define V1_INDIRECT_UNIQUENESS 0xfffffffe
 647#define V1_DIRECT_UNIQUENESS 0xffffffff
 648#define V1_DIRENTRY_UNIQUENESS 500
 649#define V1_ANY_UNIQUENESS 555   // FIXME: comment is required
 650
 651//
 652// here are conversion routines
 653//
 654static inline int uniqueness2type(__u32 uniqueness) CONSTF;
 655static inline int uniqueness2type(__u32 uniqueness)
 656{
 657        switch ((int)uniqueness) {
 658        case V1_SD_UNIQUENESS:
 659                return TYPE_STAT_DATA;
 660        case V1_INDIRECT_UNIQUENESS:
 661                return TYPE_INDIRECT;
 662        case V1_DIRECT_UNIQUENESS:
 663                return TYPE_DIRECT;
 664        case V1_DIRENTRY_UNIQUENESS:
 665                return TYPE_DIRENTRY;
 666        case V1_ANY_UNIQUENESS:
 667        default:
 668                return TYPE_ANY;
 669        }
 670}
 671
 672static inline __u32 type2uniqueness(int type) CONSTF;
 673static inline __u32 type2uniqueness(int type)
 674{
 675        switch (type) {
 676        case TYPE_STAT_DATA:
 677                return V1_SD_UNIQUENESS;
 678        case TYPE_INDIRECT:
 679                return V1_INDIRECT_UNIQUENESS;
 680        case TYPE_DIRECT:
 681                return V1_DIRECT_UNIQUENESS;
 682        case TYPE_DIRENTRY:
 683                return V1_DIRENTRY_UNIQUENESS;
 684        case TYPE_ANY:
 685        default:
 686                return V1_ANY_UNIQUENESS;
 687        }
 688}
 689
 690//
 691// key is pointer to on disk key which is stored in le, result is cpu,
 692// there is no way to get version of object from key, so, provide
 693// version to these defines
 694//
 695static inline loff_t le_key_k_offset(int version,
 696                                     const struct reiserfs_key *key)
 697{
 698        return (version == KEY_FORMAT_3_5) ?
 699            le32_to_cpu(key->u.k_offset_v1.k_offset) :
 700            offset_v2_k_offset(&(key->u.k_offset_v2));
 701}
 702
 703static inline loff_t le_ih_k_offset(const struct item_head *ih)
 704{
 705        return le_key_k_offset(ih_version(ih), &(ih->ih_key));
 706}
 707
 708static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key)
 709{
 710        return (version == KEY_FORMAT_3_5) ?
 711            uniqueness2type(le32_to_cpu(key->u.k_offset_v1.k_uniqueness)) :
 712            offset_v2_k_type(&(key->u.k_offset_v2));
 713}
 714
 715static inline loff_t le_ih_k_type(const struct item_head *ih)
 716{
 717        return le_key_k_type(ih_version(ih), &(ih->ih_key));
 718}
 719
 720static inline void set_le_key_k_offset(int version, struct reiserfs_key *key,
 721                                       loff_t offset)
 722{
 723        (version == KEY_FORMAT_3_5) ? (void)(key->u.k_offset_v1.k_offset = cpu_to_le32(offset)) :       /* jdm check */
 724            (void)(set_offset_v2_k_offset(&(key->u.k_offset_v2), offset));
 725}
 726
 727static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset)
 728{
 729        set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
 730}
 731
 732static inline void set_le_key_k_type(int version, struct reiserfs_key *key,
 733                                     int type)
 734{
 735        (version == KEY_FORMAT_3_5) ?
 736            (void)(key->u.k_offset_v1.k_uniqueness =
 737                   cpu_to_le32(type2uniqueness(type)))
 738            : (void)(set_offset_v2_k_type(&(key->u.k_offset_v2), type));
 739}
 740
 741static inline void set_le_ih_k_type(struct item_head *ih, int type)
 742{
 743        set_le_key_k_type(ih_version(ih), &(ih->ih_key), type);
 744}
 745
 746static inline int is_direntry_le_key(int version, struct reiserfs_key *key)
 747{
 748        return le_key_k_type(version, key) == TYPE_DIRENTRY;
 749}
 750
 751static inline int is_direct_le_key(int version, struct reiserfs_key *key)
 752{
 753        return le_key_k_type(version, key) == TYPE_DIRECT;
 754}
 755
 756static inline int is_indirect_le_key(int version, struct reiserfs_key *key)
 757{
 758        return le_key_k_type(version, key) == TYPE_INDIRECT;
 759}
 760
 761static inline int is_statdata_le_key(int version, struct reiserfs_key *key)
 762{
 763        return le_key_k_type(version, key) == TYPE_STAT_DATA;
 764}
 765
 766//
 767// item header has version.
 768//
 769static inline int is_direntry_le_ih(struct item_head *ih)
 770{
 771        return is_direntry_le_key(ih_version(ih), &ih->ih_key);
 772}
 773
 774static inline int is_direct_le_ih(struct item_head *ih)
 775{
 776        return is_direct_le_key(ih_version(ih), &ih->ih_key);
 777}
 778
 779static inline int is_indirect_le_ih(struct item_head *ih)
 780{
 781        return is_indirect_le_key(ih_version(ih), &ih->ih_key);
 782}
 783
 784static inline int is_statdata_le_ih(struct item_head *ih)
 785{
 786        return is_statdata_le_key(ih_version(ih), &ih->ih_key);
 787}
 788
 789//
 790// key is pointer to cpu key, result is cpu
 791//
 792static inline loff_t cpu_key_k_offset(const struct cpu_key *key)
 793{
 794        return key->on_disk_key.k_offset;
 795}
 796
 797static inline loff_t cpu_key_k_type(const struct cpu_key *key)
 798{
 799        return key->on_disk_key.k_type;
 800}
 801
 802static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset)
 803{
 804        key->on_disk_key.k_offset = offset;
 805}
 806
 807static inline void set_cpu_key_k_type(struct cpu_key *key, int type)
 808{
 809        key->on_disk_key.k_type = type;
 810}
 811
 812static inline void cpu_key_k_offset_dec(struct cpu_key *key)
 813{
 814        key->on_disk_key.k_offset--;
 815}
 816
 817#define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
 818#define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
 819#define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
 820#define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
 821
 822/* are these used ? */
 823#define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
 824#define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
 825#define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
 826#define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
 827
 828#define I_K_KEY_IN_ITEM(ih, key, n_blocksize) \
 829    (!COMP_SHORT_KEYS(ih, key) && \
 830          I_OFF_BYTE_IN_ITEM(ih, k_offset(key), n_blocksize))
 831
 832/* maximal length of item */
 833#define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
 834#define MIN_ITEM_LEN 1
 835
 836/* object identifier for root dir */
 837#define REISERFS_ROOT_OBJECTID 2
 838#define REISERFS_ROOT_PARENT_OBJECTID 1
 839
 840extern struct reiserfs_key root_key;
 841
 842/* 
 843 * Picture represents a leaf of the S+tree
 844 *  ______________________________________________________
 845 * |      |  Array of     |                   |           |
 846 * |Block |  Object-Item  |      F r e e      |  Objects- |
 847 * | head |  Headers      |     S p a c e     |   Items   |
 848 * |______|_______________|___________________|___________|
 849 */
 850
 851/* Header of a disk block.  More precisely, header of a formatted leaf
 852   or internal node, and not the header of an unformatted node. */
 853struct block_head {
 854        __le16 blk_level;       /* Level of a block in the tree. */
 855        __le16 blk_nr_item;     /* Number of keys/items in a block. */
 856        __le16 blk_free_space;  /* Block free space in bytes. */
 857        __le16 blk_reserved;
 858        /* dump this in v4/planA */
 859        struct reiserfs_key blk_right_delim_key;        /* kept only for compatibility */
 860};
 861
 862#define BLKH_SIZE                     (sizeof(struct block_head))
 863#define blkh_level(p_blkh)            (le16_to_cpu((p_blkh)->blk_level))
 864#define blkh_nr_item(p_blkh)          (le16_to_cpu((p_blkh)->blk_nr_item))
 865#define blkh_free_space(p_blkh)       (le16_to_cpu((p_blkh)->blk_free_space))
 866#define blkh_reserved(p_blkh)         (le16_to_cpu((p_blkh)->blk_reserved))
 867#define set_blkh_level(p_blkh,val)    ((p_blkh)->blk_level = cpu_to_le16(val))
 868#define set_blkh_nr_item(p_blkh,val)  ((p_blkh)->blk_nr_item = cpu_to_le16(val))
 869#define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
 870#define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
 871#define blkh_right_delim_key(p_blkh)  ((p_blkh)->blk_right_delim_key)
 872#define set_blkh_right_delim_key(p_blkh,val)  ((p_blkh)->blk_right_delim_key = val)
 873
 874/*
 875 * values for blk_level field of the struct block_head
 876 */
 877
 878#define FREE_LEVEL 0            /* when node gets removed from the tree its
 879                                   blk_level is set to FREE_LEVEL. It is then
 880                                   used to see whether the node is still in the
 881                                   tree */
 882
 883#define DISK_LEAF_NODE_LEVEL  1 /* Leaf node level. */
 884
 885/* Given the buffer head of a formatted node, resolve to the block head of that node. */
 886#define B_BLK_HEAD(bh)                  ((struct block_head *)((bh)->b_data))
 887/* Number of items that are in buffer. */
 888#define B_NR_ITEMS(bh)                  (blkh_nr_item(B_BLK_HEAD(bh)))
 889#define B_LEVEL(bh)                     (blkh_level(B_BLK_HEAD(bh)))
 890#define B_FREE_SPACE(bh)                (blkh_free_space(B_BLK_HEAD(bh)))
 891
 892#define PUT_B_NR_ITEMS(bh, val)         do { set_blkh_nr_item(B_BLK_HEAD(bh), val); } while (0)
 893#define PUT_B_LEVEL(bh, val)            do { set_blkh_level(B_BLK_HEAD(bh), val); } while (0)
 894#define PUT_B_FREE_SPACE(bh, val)       do { set_blkh_free_space(B_BLK_HEAD(bh), val); } while (0)
 895
 896/* Get right delimiting key. -- little endian */
 897#define B_PRIGHT_DELIM_KEY(bh)          (&(blk_right_delim_key(B_BLK_HEAD(bh))))
 898
 899/* Does the buffer contain a disk leaf. */
 900#define B_IS_ITEMS_LEVEL(bh)            (B_LEVEL(bh) == DISK_LEAF_NODE_LEVEL)
 901
 902/* Does the buffer contain a disk internal node */
 903#define B_IS_KEYS_LEVEL(bh)      (B_LEVEL(bh) > DISK_LEAF_NODE_LEVEL \
 904                                            && B_LEVEL(bh) <= MAX_HEIGHT)
 905
 906/***************************************************************************/
 907/*                             STAT DATA                                   */
 908/***************************************************************************/
 909
 910//
 911// old stat data is 32 bytes long. We are going to distinguish new one by
 912// different size
 913//
 914struct stat_data_v1 {
 915        __le16 sd_mode;         /* file type, permissions */
 916        __le16 sd_nlink;        /* number of hard links */
 917        __le16 sd_uid;          /* owner */
 918        __le16 sd_gid;          /* group */
 919        __le32 sd_size;         /* file size */
 920        __le32 sd_atime;        /* time of last access */
 921        __le32 sd_mtime;        /* time file was last modified  */
 922        __le32 sd_ctime;        /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
 923        union {
 924                __le32 sd_rdev;
 925                __le32 sd_blocks;       /* number of blocks file uses */
 926        } __attribute__ ((__packed__)) u;
 927        __le32 sd_first_direct_byte;    /* first byte of file which is stored
 928                                           in a direct item: except that if it
 929                                           equals 1 it is a symlink and if it
 930                                           equals ~(__u32)0 there is no
 931                                           direct item.  The existence of this
 932                                           field really grates on me. Let's
 933                                           replace it with a macro based on
 934                                           sd_size and our tail suppression
 935                                           policy.  Someday.  -Hans */
 936} __attribute__ ((__packed__));
 937
 938#define SD_V1_SIZE              (sizeof(struct stat_data_v1))
 939#define stat_data_v1(ih)        (ih_version (ih) == KEY_FORMAT_3_5)
 940#define sd_v1_mode(sdp)         (le16_to_cpu((sdp)->sd_mode))
 941#define set_sd_v1_mode(sdp,v)   ((sdp)->sd_mode = cpu_to_le16(v))
 942#define sd_v1_nlink(sdp)        (le16_to_cpu((sdp)->sd_nlink))
 943#define set_sd_v1_nlink(sdp,v)  ((sdp)->sd_nlink = cpu_to_le16(v))
 944#define sd_v1_uid(sdp)          (le16_to_cpu((sdp)->sd_uid))
 945#define set_sd_v1_uid(sdp,v)    ((sdp)->sd_uid = cpu_to_le16(v))
 946#define sd_v1_gid(sdp)          (le16_to_cpu((sdp)->sd_gid))
 947#define set_sd_v1_gid(sdp,v)    ((sdp)->sd_gid = cpu_to_le16(v))
 948#define sd_v1_size(sdp)         (le32_to_cpu((sdp)->sd_size))
 949#define set_sd_v1_size(sdp,v)   ((sdp)->sd_size = cpu_to_le32(v))
 950#define sd_v1_atime(sdp)        (le32_to_cpu((sdp)->sd_atime))
 951#define set_sd_v1_atime(sdp,v)  ((sdp)->sd_atime = cpu_to_le32(v))
 952#define sd_v1_mtime(sdp)        (le32_to_cpu((sdp)->sd_mtime))
 953#define set_sd_v1_mtime(sdp,v)  ((sdp)->sd_mtime = cpu_to_le32(v))
 954#define sd_v1_ctime(sdp)        (le32_to_cpu((sdp)->sd_ctime))
 955#define set_sd_v1_ctime(sdp,v)  ((sdp)->sd_ctime = cpu_to_le32(v))
 956#define sd_v1_rdev(sdp)         (le32_to_cpu((sdp)->u.sd_rdev))
 957#define set_sd_v1_rdev(sdp,v)   ((sdp)->u.sd_rdev = cpu_to_le32(v))
 958#define sd_v1_blocks(sdp)       (le32_to_cpu((sdp)->u.sd_blocks))
 959#define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
 960#define sd_v1_first_direct_byte(sdp) \
 961                                (le32_to_cpu((sdp)->sd_first_direct_byte))
 962#define set_sd_v1_first_direct_byte(sdp,v) \
 963                                ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
 964
 965/* inode flags stored in sd_attrs (nee sd_reserved) */
 966
 967/* we want common flags to have the same values as in ext2,
 968   so chattr(1) will work without problems */
 969#define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL
 970#define REISERFS_APPEND_FL    FS_APPEND_FL
 971#define REISERFS_SYNC_FL      FS_SYNC_FL
 972#define REISERFS_NOATIME_FL   FS_NOATIME_FL
 973#define REISERFS_NODUMP_FL    FS_NODUMP_FL
 974#define REISERFS_SECRM_FL     FS_SECRM_FL
 975#define REISERFS_UNRM_FL      FS_UNRM_FL
 976#define REISERFS_COMPR_FL     FS_COMPR_FL
 977#define REISERFS_NOTAIL_FL    FS_NOTAIL_FL
 978
 979/* persistent flags that file inherits from the parent directory */
 980#define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \
 981                                REISERFS_SYNC_FL |      \
 982                                REISERFS_NOATIME_FL |   \
 983                                REISERFS_NODUMP_FL |    \
 984                                REISERFS_SECRM_FL |     \
 985                                REISERFS_COMPR_FL |     \
 986                                REISERFS_NOTAIL_FL )
 987
 988/* Stat Data on disk (reiserfs version of UFS disk inode minus the
 989   address blocks) */
 990struct stat_data {
 991        __le16 sd_mode;         /* file type, permissions */
 992        __le16 sd_attrs;        /* persistent inode flags */
 993        __le32 sd_nlink;        /* number of hard links */
 994        __le64 sd_size;         /* file size */
 995        __le32 sd_uid;          /* owner */
 996        __le32 sd_gid;          /* group */
 997        __le32 sd_atime;        /* time of last access */
 998        __le32 sd_mtime;        /* time file was last modified  */
 999        __le32 sd_ctime;        /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
1000        __le32 sd_blocks;
1001        union {
1002                __le32 sd_rdev;
1003                __le32 sd_generation;
1004                //__le32 sd_first_direct_byte;
1005                /* first byte of file which is stored in a
1006                   direct item: except that if it equals 1
1007                   it is a symlink and if it equals
1008                   ~(__u32)0 there is no direct item.  The
1009                   existence of this field really grates
1010                   on me. Let's replace it with a macro
1011                   based on sd_size and our tail
1012                   suppression policy? */
1013        } __attribute__ ((__packed__)) u;
1014} __attribute__ ((__packed__));
1015//
1016// this is 44 bytes long
1017//
1018#define SD_SIZE (sizeof(struct stat_data))
1019#define SD_V2_SIZE              SD_SIZE
1020#define stat_data_v2(ih)        (ih_version (ih) == KEY_FORMAT_3_6)
1021#define sd_v2_mode(sdp)         (le16_to_cpu((sdp)->sd_mode))
1022#define set_sd_v2_mode(sdp,v)   ((sdp)->sd_mode = cpu_to_le16(v))
1023/* sd_reserved */
1024/* set_sd_reserved */
1025#define sd_v2_nlink(sdp)        (le32_to_cpu((sdp)->sd_nlink))
1026#define set_sd_v2_nlink(sdp,v)  ((sdp)->sd_nlink = cpu_to_le32(v))
1027#define sd_v2_size(sdp)         (le64_to_cpu((sdp)->sd_size))
1028#define set_sd_v2_size(sdp,v)   ((sdp)->sd_size = cpu_to_le64(v))
1029#define sd_v2_uid(sdp)          (le32_to_cpu((sdp)->sd_uid))
1030#define set_sd_v2_uid(sdp,v)    ((sdp)->sd_uid = cpu_to_le32(v))
1031#define sd_v2_gid(sdp)          (le32_to_cpu((sdp)->sd_gid))
1032#define set_sd_v2_gid(sdp,v)    ((sdp)->sd_gid = cpu_to_le32(v))
1033#define sd_v2_atime(sdp)        (le32_to_cpu((sdp)->sd_atime))
1034#define set_sd_v2_atime(sdp,v)  ((sdp)->sd_atime = cpu_to_le32(v))
1035#define sd_v2_mtime(sdp)        (le32_to_cpu((sdp)->sd_mtime))
1036#define set_sd_v2_mtime(sdp,v)  ((sdp)->sd_mtime = cpu_to_le32(v))
1037#define sd_v2_ctime(sdp)        (le32_to_cpu((sdp)->sd_ctime))
1038#define set_sd_v2_ctime(sdp,v)  ((sdp)->sd_ctime = cpu_to_le32(v))
1039#define sd_v2_blocks(sdp)       (le32_to_cpu((sdp)->sd_blocks))
1040#define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
1041#define sd_v2_rdev(sdp)         (le32_to_cpu((sdp)->u.sd_rdev))
1042#define set_sd_v2_rdev(sdp,v)   ((sdp)->u.sd_rdev = cpu_to_le32(v))
1043#define sd_v2_generation(sdp)   (le32_to_cpu((sdp)->u.sd_generation))
1044#define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
1045#define sd_v2_attrs(sdp)         (le16_to_cpu((sdp)->sd_attrs))
1046#define set_sd_v2_attrs(sdp,v)   ((sdp)->sd_attrs = cpu_to_le16(v))
1047
1048/***************************************************************************/
1049/*                      DIRECTORY STRUCTURE                                */
1050/***************************************************************************/
1051/* 
1052   Picture represents the structure of directory items
1053   ________________________________________________
1054   |  Array of     |   |     |        |       |   |
1055   | directory     |N-1| N-2 | ....   |   1st |0th|
1056   | entry headers |   |     |        |       |   |
1057   |_______________|___|_____|________|_______|___|
1058                    <----   directory entries         ------>
1059
1060 First directory item has k_offset component 1. We store "." and ".."
1061 in one item, always, we never split "." and ".." into differing
1062 items.  This makes, among other things, the code for removing
1063 directories simpler. */
1064#define SD_OFFSET  0
1065#define SD_UNIQUENESS 0
1066#define DOT_OFFSET 1
1067#define DOT_DOT_OFFSET 2
1068#define DIRENTRY_UNIQUENESS 500
1069
1070/* */
1071#define FIRST_ITEM_OFFSET 1
1072
1073/*
1074   Q: How to get key of object pointed to by entry from entry?  
1075
1076   A: Each directory entry has its header. This header has deh_dir_id and deh_objectid fields, those are key
1077      of object, entry points to */
1078
1079/* NOT IMPLEMENTED:   
1080   Directory will someday contain stat data of object */
1081
1082struct reiserfs_de_head {
1083        __le32 deh_offset;      /* third component of the directory entry key */
1084        __le32 deh_dir_id;      /* objectid of the parent directory of the object, that is referenced
1085                                   by directory entry */
1086        __le32 deh_objectid;    /* objectid of the object, that is referenced by directory entry */
1087        __le16 deh_location;    /* offset of name in the whole item */
1088        __le16 deh_state;       /* whether 1) entry contains stat data (for future), and 2) whether
1089                                   entry is hidden (unlinked) */
1090} __attribute__ ((__packed__));
1091#define DEH_SIZE                  sizeof(struct reiserfs_de_head)
1092#define deh_offset(p_deh)         (le32_to_cpu((p_deh)->deh_offset))
1093#define deh_dir_id(p_deh)         (le32_to_cpu((p_deh)->deh_dir_id))
1094#define deh_objectid(p_deh)       (le32_to_cpu((p_deh)->deh_objectid))
1095#define deh_location(p_deh)       (le16_to_cpu((p_deh)->deh_location))
1096#define deh_state(p_deh)          (le16_to_cpu((p_deh)->deh_state))
1097
1098#define put_deh_offset(p_deh,v)   ((p_deh)->deh_offset = cpu_to_le32((v)))
1099#define put_deh_dir_id(p_deh,v)   ((p_deh)->deh_dir_id = cpu_to_le32((v)))
1100#define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
1101#define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
1102#define put_deh_state(p_deh,v)    ((p_deh)->deh_state = cpu_to_le16((v)))
1103
1104/* empty directory contains two entries "." and ".." and their headers */
1105#define EMPTY_DIR_SIZE \
1106(DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
1107
1108/* old format directories have this size when empty */
1109#define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
1110
1111#define DEH_Statdata 0          /* not used now */
1112#define DEH_Visible 2
1113
1114/* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
1115#if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
1116#   define ADDR_UNALIGNED_BITS  (3)
1117#endif
1118
1119/* These are only used to manipulate deh_state.
1120 * Because of this, we'll use the ext2_ bit routines,
1121 * since they are little endian */
1122#ifdef ADDR_UNALIGNED_BITS
1123
1124#   define aligned_address(addr)           ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
1125#   define unaligned_offset(addr)          (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
1126
1127#   define set_bit_unaligned(nr, addr)     ext2_set_bit((nr) + unaligned_offset(addr), aligned_address(addr))
1128#   define clear_bit_unaligned(nr, addr)   ext2_clear_bit((nr) + unaligned_offset(addr), aligned_address(addr))
1129#   define test_bit_unaligned(nr, addr)    ext2_test_bit((nr) + unaligned_offset(addr), aligned_address(addr))
1130
1131#else
1132
1133#   define set_bit_unaligned(nr, addr)     ext2_set_bit(nr, addr)
1134#   define clear_bit_unaligned(nr, addr)   ext2_clear_bit(nr, addr)
1135#   define test_bit_unaligned(nr, addr)    ext2_test_bit(nr, addr)
1136
1137#endif
1138
1139#define mark_de_with_sd(deh)        set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1140#define mark_de_without_sd(deh)     clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1141#define mark_de_visible(deh)        set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1142#define mark_de_hidden(deh)         clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1143
1144#define de_with_sd(deh)             test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1145#define de_visible(deh)             test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1146#define de_hidden(deh)              !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1147
1148extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid,
1149                                   __le32 par_dirid, __le32 par_objid);
1150extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid,
1151                                __le32 par_dirid, __le32 par_objid);
1152
1153/* array of the entry headers */
1154 /* get item body */
1155#define B_I_PITEM(bh,ih) ( (bh)->b_data + ih_location(ih) )
1156#define B_I_DEH(bh,ih) ((struct reiserfs_de_head *)(B_I_PITEM(bh,ih)))
1157
1158/* length of the directory entry in directory item. This define
1159   calculates length of i-th directory entry using directory entry
1160   locations from dir entry head. When it calculates length of 0-th
1161   directory entry, it uses length of whole item in place of entry
1162   location of the non-existent following entry in the calculation.
1163   See picture above.*/
1164/*
1165#define I_DEH_N_ENTRY_LENGTH(ih,deh,i) \
1166((i) ? (deh_location((deh)-1) - deh_location((deh))) : (ih_item_len((ih)) - deh_location((deh))))
1167*/
1168static inline int entry_length(const struct buffer_head *bh,
1169                               const struct item_head *ih, int pos_in_item)
1170{
1171        struct reiserfs_de_head *deh;
1172
1173        deh = B_I_DEH(bh, ih) + pos_in_item;
1174        if (pos_in_item)
1175                return deh_location(deh - 1) - deh_location(deh);
1176
1177        return ih_item_len(ih) - deh_location(deh);
1178}
1179
1180/* number of entries in the directory item, depends on ENTRY_COUNT being at the start of directory dynamic data. */
1181#define I_ENTRY_COUNT(ih) (ih_entry_count((ih)))
1182
1183/* name by bh, ih and entry_num */
1184#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))))
1185
1186// two entries per block (at least)
1187#define REISERFS_MAX_NAME(block_size) 255
1188
1189/* this structure is used for operations on directory entries. It is
1190   not a disk structure. */
1191/* When reiserfs_find_entry or search_by_entry_key find directory
1192   entry, they return filled reiserfs_dir_entry structure */
1193struct reiserfs_dir_entry {
1194        struct buffer_head *de_bh;
1195        int de_item_num;
1196        struct item_head *de_ih;
1197        int de_entry_num;
1198        struct reiserfs_de_head *de_deh;
1199        int de_entrylen;
1200        int de_namelen;
1201        char *de_name;
1202        unsigned long *de_gen_number_bit_string;
1203
1204        __u32 de_dir_id;
1205        __u32 de_objectid;
1206
1207        struct cpu_key de_entry_key;
1208};
1209
1210/* these defines are useful when a particular member of a reiserfs_dir_entry is needed */
1211
1212/* pointer to file name, stored in entry */
1213#define B_I_DEH_ENTRY_FILE_NAME(bh,ih,deh) (B_I_PITEM (bh, ih) + deh_location(deh))
1214
1215/* length of name */
1216#define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
1217(I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
1218
1219/* hash value occupies bits from 7 up to 30 */
1220#define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
1221/* generation number occupies 7 bits starting from 0 up to 6 */
1222#define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
1223#define MAX_GENERATION_NUMBER  127
1224
1225#define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
1226
1227/*
1228 * Picture represents an internal node of the reiserfs tree
1229 *  ______________________________________________________
1230 * |      |  Array of     |  Array of         |  Free     |
1231 * |block |    keys       |  pointers         | space     |
1232 * | head |      N        |      N+1          |           |
1233 * |______|_______________|___________________|___________|
1234 */
1235
1236/***************************************************************************/
1237/*                      DISK CHILD                                         */
1238/***************************************************************************/
1239/* Disk child pointer: The pointer from an internal node of the tree
1240   to a node that is on disk. */
1241struct disk_child {
1242        __le32 dc_block_number; /* Disk child's block number. */
1243        __le16 dc_size;         /* Disk child's used space.   */
1244        __le16 dc_reserved;
1245};
1246
1247#define DC_SIZE (sizeof(struct disk_child))
1248#define dc_block_number(dc_p)   (le32_to_cpu((dc_p)->dc_block_number))
1249#define dc_size(dc_p)           (le16_to_cpu((dc_p)->dc_size))
1250#define put_dc_block_number(dc_p, val)   do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
1251#define put_dc_size(dc_p, val)   do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
1252
1253/* Get disk child by buffer header and position in the tree node. */
1254#define B_N_CHILD(bh, n_pos)  ((struct disk_child *)\
1255((bh)->b_data + BLKH_SIZE + B_NR_ITEMS(bh) * KEY_SIZE + DC_SIZE * (n_pos)))
1256
1257/* Get disk child number by buffer header and position in the tree node. */
1258#define B_N_CHILD_NUM(bh, n_pos) (dc_block_number(B_N_CHILD(bh, n_pos)))
1259#define PUT_B_N_CHILD_NUM(bh, n_pos, val) \
1260                                (put_dc_block_number(B_N_CHILD(bh, n_pos), val))
1261
1262 /* maximal value of field child_size in structure disk_child */
1263 /* child size is the combined size of all items and their headers */
1264#define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
1265
1266/* amount of used space in buffer (not including block head) */
1267#define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
1268
1269/* max and min number of keys in internal node */
1270#define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
1271#define MIN_NR_KEY(bh)    (MAX_NR_KEY(bh)/2)
1272
1273/***************************************************************************/
1274/*                      PATH STRUCTURES AND DEFINES                        */
1275/***************************************************************************/
1276
1277/* Search_by_key fills up the path from the root to the leaf as it descends the tree looking for the
1278   key.  It uses reiserfs_bread to try to find buffers in the cache given their block number.  If it
1279   does not find them in the cache it reads them from disk.  For each node search_by_key finds using
1280   reiserfs_bread it then uses bin_search to look through that node.  bin_search will find the
1281   position of the block_number of the next node if it is looking through an internal node.  If it
1282   is looking through a leaf node bin_search will find the position of the item which has key either
1283   equal to given key, or which is the maximal key less than the given key. */
1284
1285struct path_element {
1286        struct buffer_head *pe_buffer;  /* Pointer to the buffer at the path in the tree. */
1287        int pe_position;        /* Position in the tree node which is placed in the */
1288        /* buffer above.                                  */
1289};
1290
1291#define MAX_HEIGHT 5            /* maximal height of a tree. don't change this without changing JOURNAL_PER_BALANCE_CNT */
1292#define EXTENDED_MAX_HEIGHT         7   /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
1293#define FIRST_PATH_ELEMENT_OFFSET   2   /* Must be equal to at least 2. */
1294
1295#define ILLEGAL_PATH_ELEMENT_OFFSET 1   /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
1296#define MAX_FEB_SIZE 6          /* this MUST be MAX_HEIGHT + 1. See about FEB below */
1297
1298/* We need to keep track of who the ancestors of nodes are.  When we
1299   perform a search we record which nodes were visited while
1300   descending the tree looking for the node we searched for. This list
1301   of nodes is called the path.  This information is used while
1302   performing balancing.  Note that this path information may become
1303   invalid, and this means we must check it when using it to see if it
1304   is still valid. You'll need to read search_by_key and the comments
1305   in it, especially about decrement_counters_in_path(), to understand
1306   this structure.  
1307
1308Paths make the code so much harder to work with and debug.... An
1309enormous number of bugs are due to them, and trying to write or modify
1310code that uses them just makes my head hurt.  They are based on an
1311excessive effort to avoid disturbing the precious VFS code.:-( The
1312gods only know how we are going to SMP the code that uses them.
1313znodes are the way! */
1314
1315#define PATH_READA      0x1     /* do read ahead */
1316#define PATH_READA_BACK 0x2     /* read backwards */
1317
1318struct treepath {
1319        int path_length;        /* Length of the array above.   */
1320        int reada;
1321        struct path_element path_elements[EXTENDED_MAX_HEIGHT]; /* Array of the path elements.  */
1322        int pos_in_item;
1323};
1324
1325#define pos_in_item(path) ((path)->pos_in_item)
1326
1327#define INITIALIZE_PATH(var) \
1328struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
1329
1330/* Get path element by path and path position. */
1331#define PATH_OFFSET_PELEMENT(path, n_offset)  ((path)->path_elements + (n_offset))
1332
1333/* Get buffer header at the path by path and path position. */
1334#define PATH_OFFSET_PBUFFER(path, n_offset)   (PATH_OFFSET_PELEMENT(path, n_offset)->pe_buffer)
1335
1336/* Get position in the element at the path by path and path position. */
1337#define PATH_OFFSET_POSITION(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_position)
1338
1339#define PATH_PLAST_BUFFER(path) (PATH_OFFSET_PBUFFER((path), (path)->path_length))
1340                                /* you know, to the person who didn't
1341                                   write this the macro name does not
1342                                   at first suggest what it does.
1343                                   Maybe POSITION_FROM_PATH_END? Or
1344                                   maybe we should just focus on
1345                                   dumping paths... -Hans */
1346#define PATH_LAST_POSITION(path) (PATH_OFFSET_POSITION((path), (path)->path_length))
1347
1348#define PATH_PITEM_HEAD(path)    B_N_PITEM_HEAD(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path))
1349
1350/* in do_balance leaf has h == 0 in contrast with path structure,
1351   where root has level == 0. That is why we need these defines */
1352#define PATH_H_PBUFFER(path, h) PATH_OFFSET_PBUFFER (path, path->path_length - (h))     /* tb->S[h] */
1353#define PATH_H_PPARENT(path, h) PATH_H_PBUFFER (path, (h) + 1)  /* tb->F[h] or tb->S[0]->b_parent */
1354#define PATH_H_POSITION(path, h) PATH_OFFSET_POSITION (path, path->path_length - (h))
1355#define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1)       /* tb->S[h]->b_item_order */
1356
1357#define PATH_H_PATH_OFFSET(path, n_h) ((path)->path_length - (n_h))
1358
1359#define get_last_bh(path) PATH_PLAST_BUFFER(path)
1360#define get_ih(path) PATH_PITEM_HEAD(path)
1361#define get_item_pos(path) PATH_LAST_POSITION(path)
1362#define get_item(path) ((void *)B_N_PITEM(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION (path)))
1363#define item_moved(ih,path) comp_items(ih, path)
1364#define path_changed(ih,path) comp_items (ih, path)
1365
1366/***************************************************************************/
1367/*                       MISC                                              */
1368/***************************************************************************/
1369
1370/* Size of pointer to the unformatted node. */
1371#define UNFM_P_SIZE (sizeof(unp_t))
1372#define UNFM_P_SHIFT 2
1373
1374// in in-core inode key is stored on le form
1375#define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
1376
1377#define MAX_UL_INT 0xffffffff
1378#define MAX_INT    0x7ffffff
1379#define MAX_US_INT 0xffff
1380
1381// reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
1382#define U32_MAX (~(__u32)0)
1383
1384static inline loff_t max_reiserfs_offset(struct inode *inode)
1385{
1386        if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
1387                return (loff_t) U32_MAX;
1388
1389        return (loff_t) ((~(__u64) 0) >> 4);
1390}
1391
1392/*#define MAX_KEY_UNIQUENESS    MAX_UL_INT*/
1393#define MAX_KEY_OBJECTID        MAX_UL_INT
1394
1395#define MAX_B_NUM  MAX_UL_INT
1396#define MAX_FC_NUM MAX_US_INT
1397
1398/* the purpose is to detect overflow of an unsigned short */
1399#define REISERFS_LINK_MAX (MAX_US_INT - 1000)
1400
1401/* The following defines are used in reiserfs_insert_item and reiserfs_append_item  */
1402#define REISERFS_KERNEL_MEM             0       /* reiserfs kernel memory mode  */
1403#define REISERFS_USER_MEM               1       /* reiserfs user memory mode            */
1404
1405#define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
1406#define get_generation(s) atomic_read (&fs_generation(s))
1407#define FILESYSTEM_CHANGED_TB(tb)  (get_generation((tb)->tb_sb) != (tb)->fs_gen)
1408#define __fs_changed(gen,s) (gen != get_generation (s))
1409#define fs_changed(gen,s)               \
1410({                                      \
1411        reiserfs_cond_resched(s);       \
1412        __fs_changed(gen, s);           \
1413})
1414
1415/***************************************************************************/
1416/*                  FIXATE NODES                                           */
1417/***************************************************************************/
1418
1419#define VI_TYPE_LEFT_MERGEABLE 1
1420#define VI_TYPE_RIGHT_MERGEABLE 2
1421
1422/* To make any changes in the tree we always first find node, that
1423   contains item to be changed/deleted or place to insert a new
1424   item. We call this node S. To do balancing we need to decide what
1425   we will shift to left/right neighbor, or to a new node, where new
1426   item will be etc. To make this analysis simpler we build virtual
1427   node. Virtual node is an array of items, that will replace items of
1428   node S. (For instance if we are going to delete an item, virtual
1429   node does not contain it). Virtual node keeps information about
1430   item sizes and types, mergeability of first and last items, sizes
1431   of all entries in directory item. We use this array of items when
1432   calculating what we can shift to neighbors and how many nodes we
1433   have to have if we do not any shiftings, if we shift to left/right
1434   neighbor or to both. */
1435struct virtual_item {
1436        int vi_index;           // index in the array of item operations
1437        unsigned short vi_type; // left/right mergeability
1438        unsigned short vi_item_len;     /* length of item that it will have after balancing */
1439        struct item_head *vi_ih;
1440        const char *vi_item;    // body of item (old or new)
1441        const void *vi_new_data;        // 0 always but paste mode
1442        void *vi_uarea;         // item specific area
1443};
1444
1445struct virtual_node {
1446        char *vn_free_ptr;      /* this is a pointer to the free space in the buffer */
1447        unsigned short vn_nr_item;      /* number of items in virtual node */
1448        short vn_size;          /* size of node , that node would have if it has unlimited size and no balancing is performed */
1449        short vn_mode;          /* mode of balancing (paste, insert, delete, cut) */
1450        short vn_affected_item_num;
1451        short vn_pos_in_item;
1452        struct item_head *vn_ins_ih;    /* item header of inserted item, 0 for other modes */
1453        const void *vn_data;
1454        struct virtual_item *vn_vi;     /* array of items (including a new one, excluding item to be deleted) */
1455};
1456
1457/* used by directory items when creating virtual nodes */
1458struct direntry_uarea {
1459        int flags;
1460        __u16 entry_count;
1461        __u16 entry_sizes[1];
1462} __attribute__ ((__packed__));
1463
1464/***************************************************************************/
1465/*                  TREE BALANCE                                           */
1466/***************************************************************************/
1467
1468/* This temporary structure is used in tree balance algorithms, and
1469   constructed as we go to the extent that its various parts are
1470   needed.  It contains arrays of nodes that can potentially be
1471   involved in the balancing of node S, and parameters that define how
1472   each of the nodes must be balanced.  Note that in these algorithms
1473   for balancing the worst case is to need to balance the current node
1474   S and the left and right neighbors and all of their parents plus
1475   create a new node.  We implement S1 balancing for the leaf nodes
1476   and S0 balancing for the internal nodes (S1 and S0 are defined in
1477   our papers.)*/
1478
1479#define MAX_FREE_BLOCK 7        /* size of the array of buffers to free at end of do_balance */
1480
1481/* maximum number of FEB blocknrs on a single level */
1482#define MAX_AMOUNT_NEEDED 2
1483
1484/* someday somebody will prefix every field in this struct with tb_ */
1485struct tree_balance {
1486        int tb_mode;
1487        int need_balance_dirty;
1488        struct super_block *tb_sb;
1489        struct reiserfs_transaction_handle *transaction_handle;
1490        struct treepath *tb_path;
1491        struct buffer_head *L[MAX_HEIGHT];      /* array of left neighbors of nodes in the path */
1492        struct buffer_head *R[MAX_HEIGHT];      /* array of right neighbors of nodes in the path */
1493        struct buffer_head *FL[MAX_HEIGHT];     /* array of fathers of the left  neighbors      */
1494        struct buffer_head *FR[MAX_HEIGHT];     /* array of fathers of the right neighbors      */
1495        struct buffer_head *CFL[MAX_HEIGHT];    /* array of common parents of center node and its left neighbor  */
1496        struct buffer_head *CFR[MAX_HEIGHT];    /* array of common parents of center node and its right neighbor */
1497
1498        struct buffer_head *FEB[MAX_FEB_SIZE];  /* array of empty buffers. Number of buffers in array equals
1499                                                   cur_blknum. */
1500        struct buffer_head *used[MAX_FEB_SIZE];
1501        struct buffer_head *thrown[MAX_FEB_SIZE];
1502        int lnum[MAX_HEIGHT];   /* array of number of items which must be
1503                                   shifted to the left in order to balance the
1504                                   current node; for leaves includes item that
1505                                   will be partially shifted; for internal
1506                                   nodes, it is the number of child pointers
1507                                   rather than items. It includes the new item
1508                                   being created. The code sometimes subtracts
1509                                   one to get the number of wholly shifted
1510                                   items for other purposes. */
1511        int rnum[MAX_HEIGHT];   /* substitute right for left in comment above */
1512        int lkey[MAX_HEIGHT];   /* array indexed by height h mapping the key delimiting L[h] and
1513                                   S[h] to its item number within the node CFL[h] */
1514        int rkey[MAX_HEIGHT];   /* substitute r for l in comment above */
1515        int insert_size[MAX_HEIGHT];    /* the number of bytes by we are trying to add or remove from
1516                                           S[h]. A negative value means removing.  */
1517        int blknum[MAX_HEIGHT]; /* number of nodes that will replace node S[h] after
1518                                   balancing on the level h of the tree.  If 0 then S is
1519                                   being deleted, if 1 then S is remaining and no new nodes
1520                                   are being created, if 2 or 3 then 1 or 2 new nodes is
1521                                   being created */
1522
1523        /* fields that are used only for balancing leaves of the tree */
1524        int cur_blknum;         /* number of empty blocks having been already allocated                 */
1525        int s0num;              /* number of items that fall into left most  node when S[0] splits     */
1526        int s1num;              /* number of items that fall into first  new node when S[0] splits     */
1527        int s2num;              /* number of items that fall into second new node when S[0] splits     */
1528        int lbytes;             /* number of bytes which can flow to the left neighbor from the        left    */
1529        /* most liquid item that cannot be shifted from S[0] entirely         */
1530        /* if -1 then nothing will be partially shifted */
1531        int rbytes;             /* number of bytes which will flow to the right neighbor from the right        */
1532        /* most liquid item that cannot be shifted from S[0] entirely         */
1533        /* if -1 then nothing will be partially shifted                           */
1534        int s1bytes;            /* number of bytes which flow to the first  new node when S[0] splits   */
1535        /* note: if S[0] splits into 3 nodes, then items do not need to be cut  */
1536        int s2bytes;
1537        struct buffer_head *buf_to_free[MAX_FREE_BLOCK];        /* buffers which are to be freed after do_balance finishes by unfix_nodes */
1538        char *vn_buf;           /* kmalloced memory. Used to create
1539                                   virtual node and keep map of
1540                                   dirtied bitmap blocks */
1541        int vn_buf_size;        /* size of the vn_buf */
1542        struct virtual_node *tb_vn;     /* VN starts after bitmap of bitmap blocks */
1543
1544        int fs_gen;             /* saved value of `reiserfs_generation' counter
1545                                   see FILESYSTEM_CHANGED() macro in reiserfs_fs.h */
1546#ifdef DISPLACE_NEW_PACKING_LOCALITIES
1547        struct in_core_key key; /* key pointer, to pass to block allocator or
1548                                   another low-level subsystem */
1549#endif
1550};
1551
1552/* These are modes of balancing */
1553
1554/* When inserting an item. */
1555#define M_INSERT        'i'
1556/* When inserting into (directories only) or appending onto an already
1557   existant item. */
1558#define M_PASTE         'p'
1559/* When deleting an item. */
1560#define M_DELETE        'd'
1561/* When truncating an item or removing an entry from a (directory) item. */
1562#define M_CUT           'c'
1563
1564/* used when balancing on leaf level skipped (in reiserfsck) */
1565#define M_INTERNAL      'n'
1566
1567/* When further balancing is not needed, then do_balance does not need
1568   to be called. */
1569#define M_SKIP_BALANCING                's'
1570#define M_CONVERT       'v'
1571
1572/* modes of leaf_move_items */
1573#define LEAF_FROM_S_TO_L 0
1574#define LEAF_FROM_S_TO_R 1
1575#define LEAF_FROM_R_TO_L 2
1576#define LEAF_FROM_L_TO_R 3
1577#define LEAF_FROM_S_TO_SNEW 4
1578
1579#define FIRST_TO_LAST 0
1580#define LAST_TO_FIRST 1
1581
1582/* used in do_balance for passing parent of node information that has
1583   been gotten from tb struct */
1584struct buffer_info {
1585        struct tree_balance *tb;
1586        struct buffer_head *bi_bh;
1587        struct buffer_head *bi_parent;
1588        int bi_position;
1589};
1590
1591static inline struct super_block *sb_from_tb(struct tree_balance *tb)
1592{
1593        return tb ? tb->tb_sb : NULL;
1594}
1595
1596static inline struct super_block *sb_from_bi(struct buffer_info *bi)
1597{
1598        return bi ? sb_from_tb(bi->tb) : NULL;
1599}
1600
1601/* there are 4 types of items: stat data, directory item, indirect, direct.
1602+-------------------+------------+--------------+------------+
1603|                   |  k_offset  | k_uniqueness | mergeable? |
1604+-------------------+------------+--------------+------------+
1605|     stat data     |   0        |      0       |   no       |
1606+-------------------+------------+--------------+------------+
1607| 1st directory item| DOT_OFFSET |DIRENTRY_UNIQUENESS|   no       | 
1608| non 1st directory | hash value |              |   yes      |
1609|     item          |            |              |            |
1610+-------------------+------------+--------------+------------+
1611| indirect item     | offset + 1 |TYPE_INDIRECT |   if this is not the first indirect item of the object
1612+-------------------+------------+--------------+------------+
1613| direct item       | offset + 1 |TYPE_DIRECT   | if not this is not the first direct item of the object
1614+-------------------+------------+--------------+------------+
1615*/
1616
1617struct item_operations {
1618        int (*bytes_number) (struct item_head * ih, int block_size);
1619        void (*decrement_key) (struct cpu_key *);
1620        int (*is_left_mergeable) (struct reiserfs_key * ih,
1621                                  unsigned long bsize);
1622        void (*print_item) (struct item_head *, char *item);
1623        void (*check_item) (struct item_head *, char *item);
1624
1625        int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
1626                          int is_affected, int insert_size);
1627        int (*check_left) (struct virtual_item * vi, int free,
1628                           int start_skip, int end_skip);
1629        int (*check_right) (struct virtual_item * vi, int free);
1630        int (*part_size) (struct virtual_item * vi, int from, int to);
1631        int (*unit_num) (struct virtual_item * vi);
1632        void (*print_vi) (struct virtual_item * vi);
1633};
1634
1635extern struct item_operations *item_ops[TYPE_ANY + 1];
1636
1637#define op_bytes_number(ih,bsize)                    item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
1638#define op_is_left_mergeable(key,bsize)              item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
1639#define op_print_item(ih,item)                       item_ops[le_ih_k_type (ih)]->print_item (ih, item)
1640#define op_check_item(ih,item)                       item_ops[le_ih_k_type (ih)]->check_item (ih, item)
1641#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)
1642#define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
1643#define op_check_right(vi,free)                      item_ops[(vi)->vi_index]->check_right (vi, free)
1644#define op_part_size(vi,from,to)                     item_ops[(vi)->vi_index]->part_size (vi, from, to)
1645#define op_unit_num(vi)                              item_ops[(vi)->vi_index]->unit_num (vi)
1646#define op_print_vi(vi)                              item_ops[(vi)->vi_index]->print_vi (vi)
1647
1648#define COMP_SHORT_KEYS comp_short_keys
1649
1650/* number of blocks pointed to by the indirect item */
1651#define I_UNFM_NUM(ih)  (ih_item_len(ih) / UNFM_P_SIZE)
1652
1653/* the used space within the unformatted node corresponding to pos within the item pointed to by ih */
1654#define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
1655
1656/* number of bytes contained by the direct item or the unformatted nodes the indirect item points to */
1657
1658/* get the item header */
1659#define B_N_PITEM_HEAD(bh,item_num) ( (struct item_head * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1660
1661/* get key */
1662#define B_N_PDELIM_KEY(bh,item_num) ( (struct reiserfs_key * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1663
1664/* get the key */
1665#define B_N_PKEY(bh,item_num) ( &(B_N_PITEM_HEAD(bh,item_num)->ih_key) )
1666
1667/* get item body */
1668#define B_N_PITEM(bh,item_num) ( (bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(item_num))))
1669
1670/* get the stat data by the buffer header and the item order */
1671#define B_N_STAT_DATA(bh,nr) \
1672( (struct stat_data *)((bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(nr))) ) )
1673
1674    /* following defines use reiserfs buffer header and item header */
1675
1676/* get stat-data */
1677#define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
1678
1679// this is 3976 for size==4096
1680#define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
1681
1682/* indirect items consist of entries which contain blocknrs, pos
1683   indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
1684   blocknr contained by the entry pos points to */
1685#define B_I_POS_UNFM_POINTER(bh,ih,pos) le32_to_cpu(*(((unp_t *)B_I_PITEM(bh,ih)) + (pos)))
1686#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)
1687
1688struct reiserfs_iget_args {
1689        __u32 objectid;
1690        __u32 dirid;
1691};
1692
1693/***************************************************************************/
1694/*                    FUNCTION DECLARATIONS                                */
1695/***************************************************************************/
1696
1697#define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
1698
1699#define journal_trans_half(blocksize) \
1700        ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
1701
1702/* journal.c see journal.c for all the comments here */
1703
1704/* first block written in a commit.  */
1705struct reiserfs_journal_desc {
1706        __le32 j_trans_id;      /* id of commit */
1707        __le32 j_len;           /* length of commit. len +1 is the commit block */
1708        __le32 j_mount_id;      /* mount id of this trans */
1709        __le32 j_realblock[1];  /* real locations for each block */
1710};
1711
1712#define get_desc_trans_id(d)   le32_to_cpu((d)->j_trans_id)
1713#define get_desc_trans_len(d)  le32_to_cpu((d)->j_len)
1714#define get_desc_mount_id(d)   le32_to_cpu((d)->j_mount_id)
1715
1716#define set_desc_trans_id(d,val)       do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
1717#define set_desc_trans_len(d,val)      do { (d)->j_len = cpu_to_le32 (val); } while (0)
1718#define set_desc_mount_id(d,val)       do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
1719
1720/* last block written in a commit */
1721struct reiserfs_journal_commit {
1722        __le32 j_trans_id;      /* must match j_trans_id from the desc block */
1723        __le32 j_len;           /* ditto */
1724        __le32 j_realblock[1];  /* real locations for each block */
1725};
1726
1727#define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
1728#define get_commit_trans_len(c)        le32_to_cpu((c)->j_len)
1729#define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
1730
1731#define set_commit_trans_id(c,val)     do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
1732#define set_commit_trans_len(c,val)    do { (c)->j_len = cpu_to_le32 (val); } while (0)
1733
1734/* this header block gets written whenever a transaction is considered fully flushed, and is more recent than the
1735** last fully flushed transaction.  fully flushed means all the log blocks and all the real blocks are on disk,
1736** and this transaction does not need to be replayed.
1737*/
1738struct reiserfs_journal_header {
1739        __le32 j_last_flush_trans_id;   /* id of last fully flushed transaction */
1740        __le32 j_first_unflushed_offset;        /* offset in the log of where to start replay after a crash */
1741        __le32 j_mount_id;
1742        /* 12 */ struct journal_params jh_journal;
1743};
1744
1745/* biggest tunable defines are right here */
1746#define JOURNAL_BLOCK_COUNT 8192        /* number of blocks in the journal */
1747#define JOURNAL_TRANS_MAX_DEFAULT 1024  /* biggest possible single transaction, don't change for now (8/3/99) */
1748#define JOURNAL_TRANS_MIN_DEFAULT 256
1749#define JOURNAL_MAX_BATCH_DEFAULT   900 /* max blocks to batch into one transaction, don't make this any bigger than 900 */
1750#define JOURNAL_MIN_RATIO 2
1751#define JOURNAL_MAX_COMMIT_AGE 30
1752#define JOURNAL_MAX_TRANS_AGE 30
1753#define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
1754#define JOURNAL_BLOCKS_PER_OBJECT(sb)  (JOURNAL_PER_BALANCE_CNT * 3 + \
1755                                         2 * (REISERFS_QUOTA_INIT_BLOCKS(sb) + \
1756                                              REISERFS_QUOTA_TRANS_BLOCKS(sb)))
1757
1758#ifdef CONFIG_QUOTA
1759/* We need to update data and inode (atime) */
1760#define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? 2 : 0)
1761/* 1 balancing, 1 bitmap, 1 data per write + stat data update */
1762#define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? \
1763(DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0)
1764/* same as with INIT */
1765#define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? \
1766(DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0)
1767#else
1768#define REISERFS_QUOTA_TRANS_BLOCKS(s) 0
1769#define REISERFS_QUOTA_INIT_BLOCKS(s) 0
1770#define REISERFS_QUOTA_DEL_BLOCKS(s) 0
1771#endif
1772
1773/* both of these can be as low as 1, or as high as you want.  The min is the
1774** number of 4k bitmap nodes preallocated on mount. New nodes are allocated
1775** as needed, and released when transactions are committed.  On release, if 
1776** the current number of nodes is > max, the node is freed, otherwise, 
1777** it is put on a free list for faster use later.
1778*/
1779#define REISERFS_MIN_BITMAP_NODES 10
1780#define REISERFS_MAX_BITMAP_NODES 100
1781
1782#define JBH_HASH_SHIFT 13       /* these are based on journal hash size of 8192 */
1783#define JBH_HASH_MASK 8191
1784
1785#define _jhashfn(sb,block)      \
1786        (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
1787         (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
1788#define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
1789
1790// We need these to make journal.c code more readable
1791#define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1792#define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1793#define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1794
1795enum reiserfs_bh_state_bits {
1796        BH_JDirty = BH_PrivateStart,    /* buffer is in current transaction */
1797        BH_JDirty_wait,
1798        BH_JNew,                /* disk block was taken off free list before
1799                                 * being in a finished transaction, or
1800                                 * written to disk. Can be reused immed. */
1801        BH_JPrepared,
1802        BH_JRestore_dirty,
1803        BH_JTest,               // debugging only will go away
1804};
1805
1806BUFFER_FNS(JDirty, journaled);
1807TAS_BUFFER_FNS(JDirty, journaled);
1808BUFFER_FNS(JDirty_wait, journal_dirty);
1809TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
1810BUFFER_FNS(JNew, journal_new);
1811TAS_BUFFER_FNS(JNew, journal_new);
1812BUFFER_FNS(JPrepared, journal_prepared);
1813TAS_BUFFER_FNS(JPrepared, journal_prepared);
1814BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1815TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1816BUFFER_FNS(JTest, journal_test);
1817TAS_BUFFER_FNS(JTest, journal_test);
1818
1819/*
1820** transaction handle which is passed around for all journal calls
1821*/
1822struct reiserfs_transaction_handle {
1823        struct super_block *t_super;    /* super for this FS when journal_begin was
1824                                           called. saves calls to reiserfs_get_super
1825                                           also used by nested transactions to make
1826                                           sure they are nesting on the right FS
1827                                           _must_ be first in the handle
1828                                         */
1829        int t_refcount;
1830        int t_blocks_logged;    /* number of blocks this writer has logged */
1831        int t_blocks_allocated; /* number of blocks this writer allocated */
1832        unsigned int t_trans_id;        /* sanity check, equals the current trans id */
1833        void *t_handle_save;    /* save existing current->journal_info */
1834        unsigned displace_new_blocks:1; /* if new block allocation occurres, that block
1835                                           should be displaced from others */
1836        struct list_head t_list;
1837};
1838
1839/* used to keep track of ordered and tail writes, attached to the buffer
1840 * head through b_journal_head.
1841 */
1842struct reiserfs_jh {
1843        struct reiserfs_journal_list *jl;
1844        struct buffer_head *bh;
1845        struct list_head list;
1846};
1847
1848void reiserfs_free_jh(struct buffer_head *bh);
1849int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
1850int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
1851int journal_mark_dirty(struct reiserfs_transaction_handle *,
1852                       struct super_block *, struct buffer_head *bh);
1853
1854static inline int reiserfs_file_data_log(struct inode *inode)
1855{
1856        if (reiserfs_data_log(inode->i_sb) ||
1857            (REISERFS_I(inode)->i_flags & i_data_log))
1858                return 1;
1859        return 0;
1860}
1861
1862static inline int reiserfs_transaction_running(struct super_block *s)
1863{
1864        struct reiserfs_transaction_handle *th = current->journal_info;
1865        if (th && th->t_super == s)
1866                return 1;
1867        if (th && th->t_super == NULL)
1868                BUG();
1869        return 0;
1870}
1871
1872static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th)
1873{
1874        return th->t_blocks_allocated - th->t_blocks_logged;
1875}
1876
1877struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
1878                                                                    super_block
1879                                                                    *,
1880                                                                    int count);
1881int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
1882int reiserfs_commit_page(struct inode *inode, struct page *page,
1883                         unsigned from, unsigned to);
1884int reiserfs_flush_old_commits(struct super_block *);
1885int reiserfs_commit_for_inode(struct inode *);
1886int reiserfs_inode_needs_commit(struct inode *);
1887void reiserfs_update_inode_transaction(struct inode *);
1888void reiserfs_wait_on_write_block(struct super_block *s);
1889void reiserfs_block_writes(struct reiserfs_transaction_handle *th);
1890void reiserfs_allow_writes(struct super_block *s);
1891void reiserfs_check_lock_depth(struct super_block *s, char *caller);
1892int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh,
1893                                 int wait);
1894void reiserfs_restore_prepared_buffer(struct super_block *,
1895                                      struct buffer_head *bh);
1896int journal_init(struct super_block *, const char *j_dev_name, int old_format,
1897                 unsigned int);
1898int journal_release(struct reiserfs_transaction_handle *, struct super_block *);
1899int journal_release_error(struct reiserfs_transaction_handle *,
1900                          struct super_block *);
1901int journal_end(struct reiserfs_transaction_handle *, struct super_block *,
1902                unsigned long);
1903int journal_end_sync(struct reiserfs_transaction_handle *, struct super_block *,
1904                     unsigned long);
1905int journal_mark_freed(struct reiserfs_transaction_handle *,
1906                       struct super_block *, b_blocknr_t blocknr);
1907int journal_transaction_should_end(struct reiserfs_transaction_handle *, int);
1908int reiserfs_in_journal(struct super_block *sb, unsigned int bmap_nr,
1909                         int bit_nr, int searchall, b_blocknr_t *next);
1910int journal_begin(struct reiserfs_transaction_handle *,
1911                  struct super_block *sb, unsigned long);
1912int journal_join_abort(struct reiserfs_transaction_handle *,
1913                       struct super_block *sb, unsigned long);
1914void reiserfs_abort_journal(struct super_block *sb, int errno);
1915void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...);
1916int reiserfs_allocate_list_bitmaps(struct super_block *s,
1917                                   struct reiserfs_list_bitmap *, unsigned int);
1918
1919void add_save_link(struct reiserfs_transaction_handle *th,
1920                   struct inode *inode, int truncate);
1921int remove_save_link(struct inode *inode, int truncate);
1922
1923/* objectid.c */
1924__u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th);
1925void reiserfs_release_objectid(struct reiserfs_transaction_handle *th,
1926                               __u32 objectid_to_release);
1927int reiserfs_convert_objectid_map_v1(struct super_block *);
1928
1929/* stree.c */
1930int B_IS_IN_TREE(const struct buffer_head *);
1931extern void copy_item_head(struct item_head *to,
1932                           const struct item_head *from);
1933
1934// first key is in cpu form, second - le
1935extern int comp_short_keys(const struct reiserfs_key *le_key,
1936                           const struct cpu_key *cpu_key);
1937extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from);
1938
1939// both are in le form
1940extern int comp_le_keys(const struct reiserfs_key *,
1941                        const struct reiserfs_key *);
1942extern int comp_short_le_keys(const struct reiserfs_key *,
1943                              const struct reiserfs_key *);
1944
1945//
1946// get key version from on disk key - kludge
1947//
1948static inline int le_key_version(const struct reiserfs_key *key)
1949{
1950        int type;
1951
1952        type = offset_v2_k_type(&(key->u.k_offset_v2));
1953        if (type != TYPE_DIRECT && type != TYPE_INDIRECT
1954            && type != TYPE_DIRENTRY)
1955                return KEY_FORMAT_3_5;
1956
1957        return KEY_FORMAT_3_6;
1958
1959}
1960
1961static inline void copy_key(struct reiserfs_key *to,
1962                            const struct reiserfs_key *from)
1963{
1964        memcpy(to, from, KEY_SIZE);
1965}
1966
1967int comp_items(const struct item_head *stored_ih, const struct treepath *path);
1968const struct reiserfs_key *get_rkey(const struct treepath *chk_path,
1969                                    const struct super_block *sb);
1970int search_by_key(struct super_block *, const struct cpu_key *,
1971                  struct treepath *, int);
1972#define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
1973int search_for_position_by_key(struct super_block *sb,
1974                               const struct cpu_key *cpu_key,
1975                               struct treepath *search_path);
1976extern void decrement_bcount(struct buffer_head *bh);
1977void decrement_counters_in_path(struct treepath *search_path);
1978void pathrelse(struct treepath *search_path);
1979int reiserfs_check_path(struct treepath *p);
1980void pathrelse_and_restore(struct super_block *s, struct treepath *search_path);
1981
1982int reiserfs_insert_item(struct reiserfs_transaction_handle *th,
1983                         struct treepath *path,
1984                         const struct cpu_key *key,
1985                         struct item_head *ih,
1986                         struct inode *inode, const char *body);
1987
1988int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th,
1989                             struct treepath *path,
1990                             const struct cpu_key *key,
1991                             struct inode *inode,
1992                             const char *body, int paste_size);
1993
1994int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,
1995                           struct treepath *path,
1996                           struct cpu_key *key,
1997                           struct inode *inode,
1998                           struct page *page, loff_t new_file_size);
1999
2000int reiserfs_delete_item(struct reiserfs_transaction_handle *th,
2001                         struct treepath *path,
2002                         const struct cpu_key *key,
2003                         struct inode *inode, struct buffer_head *un_bh);
2004
2005void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
2006                                struct inode *inode, struct reiserfs_key *key);
2007int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
2008                           struct inode *inode);
2009int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,
2010                         struct inode *inode, struct page *,
2011                         int update_timestamps);
2012
2013#define i_block_size(inode) ((inode)->i_sb->s_blocksize)
2014#define file_size(inode) ((inode)->i_size)
2015#define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
2016
2017#define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
2018!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 )
2019
2020void padd_item(char *item, int total_length, int length);
2021
2022/* inode.c */
2023/* args for the create parameter of reiserfs_get_block */
2024#define GET_BLOCK_NO_CREATE 0   /* don't create new blocks or convert tails */
2025#define GET_BLOCK_CREATE 1      /* add anything you need to find block */
2026#define GET_BLOCK_NO_HOLE 2     /* return -ENOENT for file holes */
2027#define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
2028#define GET_BLOCK_NO_IMUX     8 /* i_mutex is not held, don't preallocate */
2029#define GET_BLOCK_NO_DANGLE   16        /* don't leave any transactions running */
2030
2031void reiserfs_read_locked_inode(struct inode *inode,
2032                                struct reiserfs_iget_args *args);
2033int reiserfs_find_actor(struct inode *inode, void *p);
2034int reiserfs_init_locked_inode(struct inode *inode, void *p);
2035void reiserfs_evict_inode(struct inode *inode);
2036int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc);
2037int reiserfs_get_block(struct inode *inode, sector_t block,
2038                       struct buffer_head *bh_result, int create);
2039struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2040                                     int fh_len, int fh_type);
2041struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
2042                                     int fh_len, int fh_type);
2043int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
2044                       int connectable);
2045
2046int reiserfs_truncate_file(struct inode *, int update_timestamps);
2047void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset,
2048                  int type, int key_length);
2049void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
2050                       int version,
2051                       loff_t offset, int type, int length, int entry_count);
2052struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key);
2053
2054struct reiserfs_security_handle;
2055int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
2056                       struct inode *dir, int mode,
2057                       const char *symname, loff_t i_size,
2058                       struct dentry *dentry, struct inode *inode,
2059                       struct reiserfs_security_handle *security);
2060
2061void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
2062                             struct inode *inode, loff_t size);
2063
2064static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
2065                                      struct inode *inode)
2066{
2067        reiserfs_update_sd_size(th, inode, inode->i_size);
2068}
2069
2070void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode);
2071void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs);
2072int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
2073
2074int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len);
2075
2076/* namei.c */
2077void set_de_name_and_namelen(struct reiserfs_dir_entry *de);
2078int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,
2079                        struct treepath *path, struct reiserfs_dir_entry *de);
2080struct dentry *reiserfs_get_parent(struct dentry *);
2081
2082#ifdef CONFIG_REISERFS_PROC_INFO
2083int reiserfs_proc_info_init(struct super_block *sb);
2084int reiserfs_proc_info_done(struct super_block *sb);
2085int reiserfs_proc_info_global_init(void);
2086int reiserfs_proc_info_global_done(void);
2087
2088#define PROC_EXP( e )   e
2089
2090#define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
2091#define PROC_INFO_MAX( sb, field, value )                                                               \
2092    __PINFO( sb ).field =                                                                                               \
2093        max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
2094#define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
2095#define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
2096#define PROC_INFO_BH_STAT( sb, bh, level )                                                      \
2097    PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] );                                              \
2098    PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) );      \
2099    PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
2100#else
2101static inline int reiserfs_proc_info_init(struct super_block *sb)
2102{
2103        return 0;
2104}
2105
2106static inline int reiserfs_proc_info_done(struct super_block *sb)
2107{
2108        return 0;
2109}
2110
2111static inline int reiserfs_proc_info_global_init(void)
2112{
2113        return 0;
2114}
2115
2116static inline int reiserfs_proc_info_global_done(void)
2117{
2118        return 0;
2119}
2120
2121#define PROC_EXP( e )
2122#define VOID_V ( ( void ) 0 )
2123#define PROC_INFO_MAX( sb, field, value ) VOID_V
2124#define PROC_INFO_INC( sb, field ) VOID_V
2125#define PROC_INFO_ADD( sb, field, val ) VOID_V
2126#define PROC_INFO_BH_STAT(sb, bh, n_node_level) VOID_V
2127#endif
2128
2129/* dir.c */
2130extern const struct inode_operations reiserfs_dir_inode_operations;
2131extern const struct inode_operations reiserfs_symlink_inode_operations;
2132extern const struct inode_operations reiserfs_special_inode_operations;
2133extern const struct file_operations reiserfs_dir_operations;
2134int reiserfs_readdir_dentry(struct dentry *, void *, filldir_t, loff_t *);
2135
2136/* tail_conversion.c */
2137int direct2indirect(struct reiserfs_transaction_handle *, struct inode *,
2138                    struct treepath *, struct buffer_head *, loff_t);
2139int indirect2direct(struct reiserfs_transaction_handle *, struct inode *,
2140                    struct page *, struct treepath *, const struct cpu_key *,
2141                    loff_t, char *);
2142void reiserfs_unmap_buffer(struct buffer_head *);
2143
2144/* file.c */
2145extern const struct inode_operations reiserfs_file_inode_operations;
2146extern const struct file_operations reiserfs_file_operations;
2147extern const struct address_space_operations reiserfs_address_space_operations;
2148
2149/* fix_nodes.c */
2150
2151int fix_nodes(int n_op_mode, struct tree_balance *tb,
2152              struct item_head *ins_ih, const void *);
2153void unfix_nodes(struct tree_balance *);
2154
2155/* prints.c */
2156void __reiserfs_panic(struct super_block *s, const char *id,
2157                      const char *function, const char *fmt, ...)
2158    __attribute__ ((noreturn));
2159#define reiserfs_panic(s, id, fmt, args...) \
2160        __reiserfs_panic(s, id, __func__, fmt, ##args)
2161void __reiserfs_error(struct super_block *s, const char *id,
2162                      const char *function, const char *fmt, ...);
2163#define reiserfs_error(s, id, fmt, args...) \
2164         __reiserfs_error(s, id, __func__, fmt, ##args)
2165void reiserfs_info(struct super_block *s, const char *fmt, ...);
2166void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...);
2167void print_indirect_item(struct buffer_head *bh, int item_num);
2168void store_print_tb(struct tree_balance *tb);
2169void print_cur_tb(char *mes);
2170void print_de(struct reiserfs_dir_entry *de);
2171void print_bi(struct buffer_info *bi, char *mes);
2172#define PRINT_LEAF_ITEMS 1      /* print all items */
2173#define PRINT_DIRECTORY_ITEMS 2 /* print directory items */
2174#define PRINT_DIRECT_ITEMS 4    /* print contents of direct items */
2175void print_block(struct buffer_head *bh, ...);
2176void print_bmap(struct super_block *s, int silent);
2177void print_bmap_block(int i, char *data, int size, int silent);
2178/*void print_super_block (struct super_block * s, char * mes);*/
2179void print_objectid_map(struct super_block *s);
2180void print_block_head(struct buffer_head *bh, char *mes);
2181void check_leaf(struct buffer_head *bh);
2182void check_internal(struct buffer_head *bh);
2183void print_statistics(struct super_block *s);
2184char *reiserfs_hashname(int code);
2185
2186/* lbalance.c */
2187int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num,
2188                    int mov_bytes, struct buffer_head *Snew);
2189int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes);
2190int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes);
2191void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first,
2192                       int del_num, int del_bytes);
2193void leaf_insert_into_buf(struct buffer_info *bi, int before,
2194                          struct item_head *inserted_item_ih,
2195                          const char *inserted_item_body, int zeros_number);
2196void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num,
2197                          int pos_in_item, int paste_size, const char *body,
2198                          int zeros_number);
2199void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num,
2200                          int pos_in_item, int cut_size);
2201void leaf_paste_entries(struct buffer_info *bi, int item_num, int before,
2202                        int new_entry_count, struct reiserfs_de_head *new_dehs,
2203                        const char *records, int paste_size);
2204/* ibalance.c */
2205int balance_internal(struct tree_balance *, int, int, struct item_head *,
2206                     struct buffer_head **);
2207
2208/* do_balance.c */
2209void do_balance_mark_leaf_dirty(struct tree_balance *tb,
2210                                struct buffer_head *bh, int flag);
2211#define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
2212#define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
2213
2214void do_balance(struct tree_balance *tb, struct item_head *ih,
2215                const char *body, int flag);
2216void reiserfs_invalidate_buffer(struct tree_balance *tb,
2217                                struct buffer_head *bh);
2218
2219int get_left_neighbor_position(struct tree_balance *tb, int h);
2220int get_right_neighbor_position(struct tree_balance *tb, int h);
2221void replace_key(struct tree_balance *tb, struct buffer_head *, int,
2222                 struct buffer_head *, int);
2223void make_empty_node(struct buffer_info *);
2224struct buffer_head *get_FEB(struct tree_balance *);
2225
2226/* bitmap.c */
2227
2228/* structure contains hints for block allocator, and it is a container for
2229 * arguments, such as node, search path, transaction_handle, etc. */
2230struct __reiserfs_blocknr_hint {
2231        struct inode *inode;    /* inode passed to allocator, if we allocate unf. nodes */
2232        sector_t block;         /* file offset, in blocks */
2233        struct in_core_key key;
2234        struct treepath *path;  /* search path, used by allocator to deternine search_start by
2235                                 * various ways */
2236        struct reiserfs_transaction_handle *th; /* transaction handle is needed to log super blocks and
2237                                                 * bitmap blocks changes  */
2238        b_blocknr_t beg, end;
2239        b_blocknr_t search_start;       /* a field used to transfer search start value (block number)
2240                                         * between different block allocator procedures
2241                                         * (determine_search_start() and others) */
2242        int prealloc_size;      /* is set in determine_prealloc_size() function, used by underlayed
2243                                 * function that do actual allocation */
2244
2245        unsigned formatted_node:1;      /* the allocator uses different polices for getting disk space for
2246                                         * formatted/unformatted blocks with/without preallocation */
2247        unsigned preallocate:1;
2248};
2249
2250typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
2251
2252int reiserfs_parse_alloc_options(struct super_block *, char *);
2253void reiserfs_init_alloc_options(struct super_block *s);
2254
2255/*
2256 * given a directory, this will tell you what packing locality
2257 * to use for a new object underneat it.  The locality is returned
2258 * in disk byte order (le).
2259 */
2260__le32 reiserfs_choose_packing(struct inode *dir);
2261
2262int reiserfs_init_bitmap_cache(struct super_block *sb);
2263void reiserfs_free_bitmap_cache(struct super_block *sb);
2264void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info);
2265struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap);
2266int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value);
2267void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *,
2268                         b_blocknr_t, int for_unformatted);
2269int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int,
2270                               int);
2271static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb,
2272                                             b_blocknr_t * new_blocknrs,
2273                                             int amount_needed)
2274{
2275        reiserfs_blocknr_hint_t hint = {
2276                .th = tb->transaction_handle,
2277                .path = tb->tb_path,
2278                .inode = NULL,
2279                .key = tb->key,
2280                .block = 0,
2281                .formatted_node = 1
2282        };
2283        return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed,
2284                                          0);
2285}
2286
2287static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle
2288                                            *th, struct inode *inode,
2289                                            b_blocknr_t * new_blocknrs,
2290                                            struct treepath *path,
2291                                            sector_t block)
2292{
2293        reiserfs_blocknr_hint_t hint = {
2294                .th = th,
2295                .path = path,
2296                .inode = inode,
2297                .block = block,
2298                .formatted_node = 0,
2299                .preallocate = 0
2300        };
2301        return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
2302}
2303
2304#ifdef REISERFS_PREALLOCATE
2305static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle
2306                                             *th, struct inode *inode,
2307                                             b_blocknr_t * new_blocknrs,
2308                                             struct treepath *path,
2309                                             sector_t block)
2310{
2311        reiserfs_blocknr_hint_t hint = {
2312                .th = th,
2313                .path = path,
2314                .inode = inode,
2315                .block = block,
2316                .formatted_node = 0,
2317                .preallocate = 1
2318        };
2319        return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
2320}
2321
2322void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
2323                               struct inode *inode);
2324void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th);
2325#endif
2326
2327/* hashes.c */
2328__u32 keyed_hash(const signed char *msg, int len);
2329__u32 yura_hash(const signed char *msg, int len);
2330__u32 r5_hash(const signed char *msg, int len);
2331
2332/* the ext2 bit routines adjust for big or little endian as
2333** appropriate for the arch, so in our laziness we use them rather
2334** than using the bit routines they call more directly.  These
2335** routines must be used when changing on disk bitmaps.  */
2336#define reiserfs_test_and_set_le_bit   ext2_set_bit
2337#define reiserfs_test_and_clear_le_bit ext2_clear_bit
2338#define reiserfs_test_le_bit           ext2_test_bit
2339#define reiserfs_find_next_zero_le_bit ext2_find_next_zero_bit
2340
2341/* sometimes reiserfs_truncate may require to allocate few new blocks
2342   to perform indirect2direct conversion. People probably used to
2343   think, that truncate should work without problems on a filesystem
2344   without free disk space. They may complain that they can not
2345   truncate due to lack of free disk space. This spare space allows us
2346   to not worry about it. 500 is probably too much, but it should be
2347   absolutely safe */
2348#define SPARE_SPACE 500
2349
2350/* prototypes from ioctl.c */
2351long reiserfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg);
2352long reiserfs_compat_ioctl(struct file *filp,
2353                   unsigned int cmd, unsigned long arg);
2354int reiserfs_unpack(struct inode *inode, struct file *filp);
2355
2356#endif /* __KERNEL__ */
2357
2358#endif                          /* _LINUX_REISER_FS_H */
2359