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