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