1 /* 2 * Copyright 1996, 1997, 1998 Hans Reiser, see reiserfs/README for licensing and copyright details 3 */ 4 5 #include <linux/reiserfs_fs.h> 6 7 #include <linux/slab.h> 8 #include <linux/interrupt.h> 9 #include <linux/sched.h> 10 #include <linux/bug.h> 11 #include <linux/workqueue.h> 12 #include <asm/unaligned.h> 13 #include <linux/bitops.h> 14 #include <linux/proc_fs.h> 15 #include <linux/buffer_head.h> 16 17 /* the 32 bit compat definitions with int argument */ 18 #define REISERFS_IOC32_UNPACK _IOW(0xCD, 1, int) 19 #define REISERFS_IOC32_GETFLAGS FS_IOC32_GETFLAGS 20 #define REISERFS_IOC32_SETFLAGS FS_IOC32_SETFLAGS 21 #define REISERFS_IOC32_GETVERSION FS_IOC32_GETVERSION 22 #define REISERFS_IOC32_SETVERSION FS_IOC32_SETVERSION 23 24 struct reiserfs_journal_list; 25 26 /** bitmasks for i_flags field in reiserfs-specific part of inode */ 27 typedef enum { 28 /** this says what format of key do all items (but stat data) of 29 an object have. If this is set, that format is 3.6 otherwise 30 - 3.5 */ 31 i_item_key_version_mask = 0x0001, 32 /** If this is unset, object has 3.5 stat data, otherwise, it has 33 3.6 stat data with 64bit size, 32bit nlink etc. */ 34 i_stat_data_version_mask = 0x0002, 35 /** file might need tail packing on close */ 36 i_pack_on_close_mask = 0x0004, 37 /** don't pack tail of file */ 38 i_nopack_mask = 0x0008, 39 /** If those is set, "safe link" was created for this file during 40 truncate or unlink. Safe link is used to avoid leakage of disk 41 space on crash with some files open, but unlinked. */ 42 i_link_saved_unlink_mask = 0x0010, 43 i_link_saved_truncate_mask = 0x0020, 44 i_has_xattr_dir = 0x0040, 45 i_data_log = 0x0080, 46 } reiserfs_inode_flags; 47 48 struct reiserfs_inode_info { 49 __u32 i_key[4]; /* key is still 4 32 bit integers */ 50 /** transient inode flags that are never stored on disk. Bitmasks 51 for this field are defined above. */ 52 __u32 i_flags; 53 54 __u32 i_first_direct_byte; // offset of first byte stored in direct item. 55 56 /* copy of persistent inode flags read from sd_attrs. */ 57 __u32 i_attrs; 58 59 int i_prealloc_block; /* first unused block of a sequence of unused blocks */ 60 int i_prealloc_count; /* length of that sequence */ 61 struct list_head i_prealloc_list; /* per-transaction list of inodes which 62 * have preallocated blocks */ 63 64 unsigned new_packing_locality:1; /* new_packig_locality is created; new blocks 65 * for the contents of this directory should be 66 * displaced */ 67 68 /* we use these for fsync or O_SYNC to decide which transaction 69 ** needs to be committed in order for this inode to be properly 70 ** flushed */ 71 unsigned int i_trans_id; 72 struct reiserfs_journal_list *i_jl; 73 atomic_t openers; 74 struct mutex tailpack; 75 #ifdef CONFIG_REISERFS_FS_XATTR 76 struct rw_semaphore i_xattr_sem; 77 #endif 78 struct inode vfs_inode; 79 }; 80 81 typedef enum { 82 reiserfs_attrs_cleared = 0x00000001, 83 } reiserfs_super_block_flags; 84 85 /* struct reiserfs_super_block accessors/mutators 86 * since this is a disk structure, it will always be in 87 * little endian format. */ 88 #define sb_block_count(sbp) (le32_to_cpu((sbp)->s_v1.s_block_count)) 89 #define set_sb_block_count(sbp,v) ((sbp)->s_v1.s_block_count = cpu_to_le32(v)) 90 #define sb_free_blocks(sbp) (le32_to_cpu((sbp)->s_v1.s_free_blocks)) 91 #define set_sb_free_blocks(sbp,v) ((sbp)->s_v1.s_free_blocks = cpu_to_le32(v)) 92 #define sb_root_block(sbp) (le32_to_cpu((sbp)->s_v1.s_root_block)) 93 #define set_sb_root_block(sbp,v) ((sbp)->s_v1.s_root_block = cpu_to_le32(v)) 94 95 #define sb_jp_journal_1st_block(sbp) \ 96 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_1st_block)) 97 #define set_sb_jp_journal_1st_block(sbp,v) \ 98 ((sbp)->s_v1.s_journal.jp_journal_1st_block = cpu_to_le32(v)) 99 #define sb_jp_journal_dev(sbp) \ 100 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_dev)) 101 #define set_sb_jp_journal_dev(sbp,v) \ 102 ((sbp)->s_v1.s_journal.jp_journal_dev = cpu_to_le32(v)) 103 #define sb_jp_journal_size(sbp) \ 104 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_size)) 105 #define set_sb_jp_journal_size(sbp,v) \ 106 ((sbp)->s_v1.s_journal.jp_journal_size = cpu_to_le32(v)) 107 #define sb_jp_journal_trans_max(sbp) \ 108 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_trans_max)) 109 #define set_sb_jp_journal_trans_max(sbp,v) \ 110 ((sbp)->s_v1.s_journal.jp_journal_trans_max = cpu_to_le32(v)) 111 #define sb_jp_journal_magic(sbp) \ 112 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_magic)) 113 #define set_sb_jp_journal_magic(sbp,v) \ 114 ((sbp)->s_v1.s_journal.jp_journal_magic = cpu_to_le32(v)) 115 #define sb_jp_journal_max_batch(sbp) \ 116 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_batch)) 117 #define set_sb_jp_journal_max_batch(sbp,v) \ 118 ((sbp)->s_v1.s_journal.jp_journal_max_batch = cpu_to_le32(v)) 119 #define sb_jp_jourmal_max_commit_age(sbp) \ 120 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_commit_age)) 121 #define set_sb_jp_journal_max_commit_age(sbp,v) \ 122 ((sbp)->s_v1.s_journal.jp_journal_max_commit_age = cpu_to_le32(v)) 123 124 #define sb_blocksize(sbp) (le16_to_cpu((sbp)->s_v1.s_blocksize)) 125 #define set_sb_blocksize(sbp,v) ((sbp)->s_v1.s_blocksize = cpu_to_le16(v)) 126 #define sb_oid_maxsize(sbp) (le16_to_cpu((sbp)->s_v1.s_oid_maxsize)) 127 #define set_sb_oid_maxsize(sbp,v) ((sbp)->s_v1.s_oid_maxsize = cpu_to_le16(v)) 128 #define sb_oid_cursize(sbp) (le16_to_cpu((sbp)->s_v1.s_oid_cursize)) 129 #define set_sb_oid_cursize(sbp,v) ((sbp)->s_v1.s_oid_cursize = cpu_to_le16(v)) 130 #define sb_umount_state(sbp) (le16_to_cpu((sbp)->s_v1.s_umount_state)) 131 #define set_sb_umount_state(sbp,v) ((sbp)->s_v1.s_umount_state = cpu_to_le16(v)) 132 #define sb_fs_state(sbp) (le16_to_cpu((sbp)->s_v1.s_fs_state)) 133 #define set_sb_fs_state(sbp,v) ((sbp)->s_v1.s_fs_state = cpu_to_le16(v)) 134 #define sb_hash_function_code(sbp) \ 135 (le32_to_cpu((sbp)->s_v1.s_hash_function_code)) 136 #define set_sb_hash_function_code(sbp,v) \ 137 ((sbp)->s_v1.s_hash_function_code = cpu_to_le32(v)) 138 #define sb_tree_height(sbp) (le16_to_cpu((sbp)->s_v1.s_tree_height)) 139 #define set_sb_tree_height(sbp,v) ((sbp)->s_v1.s_tree_height = cpu_to_le16(v)) 140 #define sb_bmap_nr(sbp) (le16_to_cpu((sbp)->s_v1.s_bmap_nr)) 141 #define set_sb_bmap_nr(sbp,v) ((sbp)->s_v1.s_bmap_nr = cpu_to_le16(v)) 142 #define sb_version(sbp) (le16_to_cpu((sbp)->s_v1.s_version)) 143 #define set_sb_version(sbp,v) ((sbp)->s_v1.s_version = cpu_to_le16(v)) 144 145 #define sb_mnt_count(sbp) (le16_to_cpu((sbp)->s_mnt_count)) 146 #define set_sb_mnt_count(sbp, v) ((sbp)->s_mnt_count = cpu_to_le16(v)) 147 148 #define sb_reserved_for_journal(sbp) \ 149 (le16_to_cpu((sbp)->s_v1.s_reserved_for_journal)) 150 #define set_sb_reserved_for_journal(sbp,v) \ 151 ((sbp)->s_v1.s_reserved_for_journal = cpu_to_le16(v)) 152 153 /* LOGGING -- */ 154 155 /* These all interelate for performance. 156 ** 157 ** If the journal block count is smaller than n transactions, you lose speed. 158 ** I don't know what n is yet, I'm guessing 8-16. 159 ** 160 ** typical transaction size depends on the application, how often fsync is 161 ** called, and how many metadata blocks you dirty in a 30 second period. 162 ** The more small files (<16k) you use, the larger your transactions will 163 ** be. 164 ** 165 ** If your journal fills faster than dirty buffers get flushed to disk, it must flush them before allowing the journal 166 ** to wrap, which slows things down. If you need high speed meta data updates, the journal should be big enough 167 ** to prevent wrapping before dirty meta blocks get to disk. 168 ** 169 ** If the batch max is smaller than the transaction max, you'll waste space at the end of the journal 170 ** because journal_end sets the next transaction to start at 0 if the next transaction has any chance of wrapping. 171 ** 172 ** The large the batch max age, the better the speed, and the more meta data changes you'll lose after a crash. 173 ** 174 */ 175 176 /* don't mess with these for a while */ 177 /* we have a node size define somewhere in reiserfs_fs.h. -Hans */ 178 #define JOURNAL_BLOCK_SIZE 4096 /* BUG gotta get rid of this */ 179 #define JOURNAL_MAX_CNODE 1500 /* max cnodes to allocate. */ 180 #define JOURNAL_HASH_SIZE 8192 181 #define JOURNAL_NUM_BITMAPS 5 /* number of copies of the bitmaps to have floating. Must be >= 2 */ 182 183 /* One of these for every block in every transaction 184 ** Each one is in two hash tables. First, a hash of the current transaction, and after journal_end, a 185 ** hash of all the in memory transactions. 186 ** next and prev are used by the current transaction (journal_hash). 187 ** hnext and hprev are used by journal_list_hash. If a block is in more than one transaction, the journal_list_hash 188 ** links it in multiple times. This allows flush_journal_list to remove just the cnode belonging 189 ** to a given transaction. 190 */ 191 struct reiserfs_journal_cnode { 192 struct buffer_head *bh; /* real buffer head */ 193 struct super_block *sb; /* dev of real buffer head */ 194 __u32 blocknr; /* block number of real buffer head, == 0 when buffer on disk */ 195 unsigned long state; 196 struct reiserfs_journal_list *jlist; /* journal list this cnode lives in */ 197 struct reiserfs_journal_cnode *next; /* next in transaction list */ 198 struct reiserfs_journal_cnode *prev; /* prev in transaction list */ 199 struct reiserfs_journal_cnode *hprev; /* prev in hash list */ 200 struct reiserfs_journal_cnode *hnext; /* next in hash list */ 201 }; 202 203 struct reiserfs_bitmap_node { 204 int id; 205 char *data; 206 struct list_head list; 207 }; 208 209 struct reiserfs_list_bitmap { 210 struct reiserfs_journal_list *journal_list; 211 struct reiserfs_bitmap_node **bitmaps; 212 }; 213 214 /* 215 ** one of these for each transaction. The most important part here is the j_realblock. 216 ** this list of cnodes is used to hash all the blocks in all the commits, to mark all the 217 ** real buffer heads dirty once all the commits hit the disk, 218 ** and to make sure every real block in a transaction is on disk before allowing the log area 219 ** to be overwritten */ 220 struct reiserfs_journal_list { 221 unsigned long j_start; 222 unsigned long j_state; 223 unsigned long j_len; 224 atomic_t j_nonzerolen; 225 atomic_t j_commit_left; 226 atomic_t j_older_commits_done; /* all commits older than this on disk */ 227 struct mutex j_commit_mutex; 228 unsigned int j_trans_id; 229 time_t j_timestamp; 230 struct reiserfs_list_bitmap *j_list_bitmap; 231 struct buffer_head *j_commit_bh; /* commit buffer head */ 232 struct reiserfs_journal_cnode *j_realblock; 233 struct reiserfs_journal_cnode *j_freedlist; /* list of buffers that were freed during this trans. free each of these on flush */ 234 /* time ordered list of all active transactions */ 235 struct list_head j_list; 236 237 /* time ordered list of all transactions we haven't tried to flush yet */ 238 struct list_head j_working_list; 239 240 /* list of tail conversion targets in need of flush before commit */ 241 struct list_head j_tail_bh_list; 242 /* list of data=ordered buffers in need of flush before commit */ 243 struct list_head j_bh_list; 244 int j_refcount; 245 }; 246 247 struct reiserfs_journal { 248 struct buffer_head **j_ap_blocks; /* journal blocks on disk */ 249 struct reiserfs_journal_cnode *j_last; /* newest journal block */ 250 struct reiserfs_journal_cnode *j_first; /* oldest journal block. start here for traverse */ 251 252 struct block_device *j_dev_bd; 253 fmode_t j_dev_mode; 254 int j_1st_reserved_block; /* first block on s_dev of reserved area journal */ 255 256 unsigned long j_state; 257 unsigned int j_trans_id; 258 unsigned long j_mount_id; 259 unsigned long j_start; /* start of current waiting commit (index into j_ap_blocks) */ 260 unsigned long j_len; /* length of current waiting commit */ 261 unsigned long j_len_alloc; /* number of buffers requested by journal_begin() */ 262 atomic_t j_wcount; /* count of writers for current commit */ 263 unsigned long j_bcount; /* batch count. allows turning X transactions into 1 */ 264 unsigned long j_first_unflushed_offset; /* first unflushed transactions offset */ 265 unsigned j_last_flush_trans_id; /* last fully flushed journal timestamp */ 266 struct buffer_head *j_header_bh; 267 268 time_t j_trans_start_time; /* time this transaction started */ 269 struct mutex j_mutex; 270 struct mutex j_flush_mutex; 271 wait_queue_head_t j_join_wait; /* wait for current transaction to finish before starting new one */ 272 atomic_t j_jlock; /* lock for j_join_wait */ 273 int j_list_bitmap_index; /* number of next list bitmap to use */ 274 int j_must_wait; /* no more journal begins allowed. MUST sleep on j_join_wait */ 275 int j_next_full_flush; /* next journal_end will flush all journal list */ 276 int j_next_async_flush; /* next journal_end will flush all async commits */ 277 278 int j_cnode_used; /* number of cnodes on the used list */ 279 int j_cnode_free; /* number of cnodes on the free list */ 280 281 unsigned int j_trans_max; /* max number of blocks in a transaction. */ 282 unsigned int j_max_batch; /* max number of blocks to batch into a trans */ 283 unsigned int j_max_commit_age; /* in seconds, how old can an async commit be */ 284 unsigned int j_max_trans_age; /* in seconds, how old can a transaction be */ 285 unsigned int j_default_max_commit_age; /* the default for the max commit age */ 286 287 struct reiserfs_journal_cnode *j_cnode_free_list; 288 struct reiserfs_journal_cnode *j_cnode_free_orig; /* orig pointer returned from vmalloc */ 289 290 struct reiserfs_journal_list *j_current_jl; 291 int j_free_bitmap_nodes; 292 int j_used_bitmap_nodes; 293 294 int j_num_lists; /* total number of active transactions */ 295 int j_num_work_lists; /* number that need attention from kreiserfsd */ 296 297 /* debugging to make sure things are flushed in order */ 298 unsigned int j_last_flush_id; 299 300 /* debugging to make sure things are committed in order */ 301 unsigned int j_last_commit_id; 302 303 struct list_head j_bitmap_nodes; 304 struct list_head j_dirty_buffers; 305 spinlock_t j_dirty_buffers_lock; /* protects j_dirty_buffers */ 306 307 /* list of all active transactions */ 308 struct list_head j_journal_list; 309 /* lists that haven't been touched by writeback attempts */ 310 struct list_head j_working_list; 311 312 struct reiserfs_list_bitmap j_list_bitmap[JOURNAL_NUM_BITMAPS]; /* array of bitmaps to record the deleted blocks */ 313 struct reiserfs_journal_cnode *j_hash_table[JOURNAL_HASH_SIZE]; /* hash table for real buffer heads in current trans */ 314 struct reiserfs_journal_cnode *j_list_hash_table[JOURNAL_HASH_SIZE]; /* hash table for all the real buffer heads in all 315 the transactions */ 316 struct list_head j_prealloc_list; /* list of inodes which have preallocated blocks */ 317 int j_persistent_trans; 318 unsigned long j_max_trans_size; 319 unsigned long j_max_batch_size; 320 321 int j_errno; 322 323 /* when flushing ordered buffers, throttle new ordered writers */ 324 struct delayed_work j_work; 325 struct super_block *j_work_sb; 326 atomic_t j_async_throttle; 327 }; 328 329 enum journal_state_bits { 330 J_WRITERS_BLOCKED = 1, /* set when new writers not allowed */ 331 J_WRITERS_QUEUED, /* set when log is full due to too many writers */ 332 J_ABORTED, /* set when log is aborted */ 333 }; 334 335 #define JOURNAL_DESC_MAGIC "ReIsErLB" /* ick. magic string to find desc blocks in the journal */ 336 337 typedef __u32(*hashf_t) (const signed char *, int); 338 339 struct reiserfs_bitmap_info { 340 __u32 free_count; 341 }; 342 343 struct proc_dir_entry; 344 345 #if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO ) 346 typedef unsigned long int stat_cnt_t; 347 typedef struct reiserfs_proc_info_data { 348 spinlock_t lock; 349 int exiting; 350 int max_hash_collisions; 351 352 stat_cnt_t breads; 353 stat_cnt_t bread_miss; 354 stat_cnt_t search_by_key; 355 stat_cnt_t search_by_key_fs_changed; 356 stat_cnt_t search_by_key_restarted; 357 358 stat_cnt_t insert_item_restarted; 359 stat_cnt_t paste_into_item_restarted; 360 stat_cnt_t cut_from_item_restarted; 361 stat_cnt_t delete_solid_item_restarted; 362 stat_cnt_t delete_item_restarted; 363 364 stat_cnt_t leaked_oid; 365 stat_cnt_t leaves_removable; 366 367 /* balances per level. Use explicit 5 as MAX_HEIGHT is not visible yet. */ 368 stat_cnt_t balance_at[5]; /* XXX */ 369 /* sbk == search_by_key */ 370 stat_cnt_t sbk_read_at[5]; /* XXX */ 371 stat_cnt_t sbk_fs_changed[5]; 372 stat_cnt_t sbk_restarted[5]; 373 stat_cnt_t items_at[5]; /* XXX */ 374 stat_cnt_t free_at[5]; /* XXX */ 375 stat_cnt_t can_node_be_removed[5]; /* XXX */ 376 long int lnum[5]; /* XXX */ 377 long int rnum[5]; /* XXX */ 378 long int lbytes[5]; /* XXX */ 379 long int rbytes[5]; /* XXX */ 380 stat_cnt_t get_neighbors[5]; 381 stat_cnt_t get_neighbors_restart[5]; 382 stat_cnt_t need_l_neighbor[5]; 383 stat_cnt_t need_r_neighbor[5]; 384 385 stat_cnt_t free_block; 386 struct __scan_bitmap_stats { 387 stat_cnt_t call; 388 stat_cnt_t wait; 389 stat_cnt_t bmap; 390 stat_cnt_t retry; 391 stat_cnt_t in_journal_hint; 392 stat_cnt_t in_journal_nohint; 393 stat_cnt_t stolen; 394 } scan_bitmap; 395 struct __journal_stats { 396 stat_cnt_t in_journal; 397 stat_cnt_t in_journal_bitmap; 398 stat_cnt_t in_journal_reusable; 399 stat_cnt_t lock_journal; 400 stat_cnt_t lock_journal_wait; 401 stat_cnt_t journal_being; 402 stat_cnt_t journal_relock_writers; 403 stat_cnt_t journal_relock_wcount; 404 stat_cnt_t mark_dirty; 405 stat_cnt_t mark_dirty_already; 406 stat_cnt_t mark_dirty_notjournal; 407 stat_cnt_t restore_prepared; 408 stat_cnt_t prepare; 409 stat_cnt_t prepare_retry; 410 } journal; 411 } reiserfs_proc_info_data_t; 412 #else 413 typedef struct reiserfs_proc_info_data { 414 } reiserfs_proc_info_data_t; 415 #endif 416 417 /* reiserfs union of in-core super block data */ 418 struct reiserfs_sb_info { 419 struct buffer_head *s_sbh; /* Buffer containing the super block */ 420 /* both the comment and the choice of 421 name are unclear for s_rs -Hans */ 422 struct reiserfs_super_block *s_rs; /* Pointer to the super block in the buffer */ 423 struct reiserfs_bitmap_info *s_ap_bitmap; 424 struct reiserfs_journal *s_journal; /* pointer to journal information */ 425 unsigned short s_mount_state; /* reiserfs state (valid, invalid) */ 426 427 /* Serialize writers access, replace the old bkl */ 428 struct mutex lock; 429 /* Owner of the lock (can be recursive) */ 430 struct task_struct *lock_owner; 431 /* Depth of the lock, start from -1 like the bkl */ 432 int lock_depth; 433 434 /* Comment? -Hans */ 435 void (*end_io_handler) (struct buffer_head *, int); 436 hashf_t s_hash_function; /* pointer to function which is used 437 to sort names in directory. Set on 438 mount */ 439 unsigned long s_mount_opt; /* reiserfs's mount options are set 440 here (currently - NOTAIL, NOLOG, 441 REPLAYONLY) */ 442 443 struct { /* This is a structure that describes block allocator options */ 444 unsigned long bits; /* Bitfield for enable/disable kind of options */ 445 unsigned long large_file_size; /* size started from which we consider file to be a large one(in blocks) */ 446 int border; /* percentage of disk, border takes */ 447 int preallocmin; /* Minimal file size (in blocks) starting from which we do preallocations */ 448 int preallocsize; /* Number of blocks we try to prealloc when file 449 reaches preallocmin size (in blocks) or 450 prealloc_list is empty. */ 451 } s_alloc_options; 452 453 /* Comment? -Hans */ 454 wait_queue_head_t s_wait; 455 /* To be obsoleted soon by per buffer seals.. -Hans */ 456 atomic_t s_generation_counter; // increased by one every time the 457 // tree gets re-balanced 458 unsigned long s_properties; /* File system properties. Currently holds 459 on-disk FS format */ 460 461 /* session statistics */ 462 int s_disk_reads; 463 int s_disk_writes; 464 int s_fix_nodes; 465 int s_do_balance; 466 int s_unneeded_left_neighbor; 467 int s_good_search_by_key_reada; 468 int s_bmaps; 469 int s_bmaps_without_search; 470 int s_direct2indirect; 471 int s_indirect2direct; 472 /* set up when it's ok for reiserfs_read_inode2() to read from 473 disk inode with nlink==0. Currently this is only used during 474 finish_unfinished() processing at mount time */ 475 int s_is_unlinked_ok; 476 reiserfs_proc_info_data_t s_proc_info_data; 477 struct proc_dir_entry *procdir; 478 int reserved_blocks; /* amount of blocks reserved for further allocations */ 479 spinlock_t bitmap_lock; /* this lock on now only used to protect reserved_blocks variable */ 480 struct dentry *priv_root; /* root of /.reiserfs_priv */ 481 struct dentry *xattr_root; /* root of /.reiserfs_priv/xattrs */ 482 int j_errno; 483 484 int work_queued; /* non-zero delayed work is queued */ 485 struct delayed_work old_work; /* old transactions flush delayed work */ 486 spinlock_t old_work_lock; /* protects old_work and work_queued */ 487 488 #ifdef CONFIG_QUOTA 489 char *s_qf_names[MAXQUOTAS]; 490 int s_jquota_fmt; 491 #endif 492 char *s_jdev; /* Stored jdev for mount option showing */ 493 #ifdef CONFIG_REISERFS_CHECK 494 495 struct tree_balance *cur_tb; /* 496 * Detects whether more than one 497 * copy of tb exists per superblock 498 * as a means of checking whether 499 * do_balance is executing concurrently 500 * against another tree reader/writer 501 * on a same mount point. 502 */ 503 #endif 504 }; 505 506 /* Definitions of reiserfs on-disk properties: */ 507 #define REISERFS_3_5 0 508 #define REISERFS_3_6 1 509 #define REISERFS_OLD_FORMAT 2 510 511 enum reiserfs_mount_options { 512 /* Mount options */ 513 REISERFS_LARGETAIL, /* large tails will be created in a session */ 514 REISERFS_SMALLTAIL, /* small (for files less than block size) tails will be created in a session */ 515 REPLAYONLY, /* replay journal and return 0. Use by fsck */ 516 REISERFS_CONVERT, /* -o conv: causes conversion of old 517 format super block to the new 518 format. If not specified - old 519 partition will be dealt with in a 520 manner of 3.5.x */ 521 522 /* -o hash={tea, rupasov, r5, detect} is meant for properly mounting 523 ** reiserfs disks from 3.5.19 or earlier. 99% of the time, this option 524 ** is not required. If the normal autodection code can't determine which 525 ** hash to use (because both hashes had the same value for a file) 526 ** use this option to force a specific hash. It won't allow you to override 527 ** the existing hash on the FS, so if you have a tea hash disk, and mount 528 ** with -o hash=rupasov, the mount will fail. 529 */ 530 FORCE_TEA_HASH, /* try to force tea hash on mount */ 531 FORCE_RUPASOV_HASH, /* try to force rupasov hash on mount */ 532 FORCE_R5_HASH, /* try to force rupasov hash on mount */ 533 FORCE_HASH_DETECT, /* try to detect hash function on mount */ 534 535 REISERFS_DATA_LOG, 536 REISERFS_DATA_ORDERED, 537 REISERFS_DATA_WRITEBACK, 538 539 /* used for testing experimental features, makes benchmarking new 540 features with and without more convenient, should never be used by 541 users in any code shipped to users (ideally) */ 542 543 REISERFS_NO_BORDER, 544 REISERFS_NO_UNHASHED_RELOCATION, 545 REISERFS_HASHED_RELOCATION, 546 REISERFS_ATTRS, 547 REISERFS_XATTRS_USER, 548 REISERFS_POSIXACL, 549 REISERFS_EXPOSE_PRIVROOT, 550 REISERFS_BARRIER_NONE, 551 REISERFS_BARRIER_FLUSH, 552 553 /* Actions on error */ 554 REISERFS_ERROR_PANIC, 555 REISERFS_ERROR_RO, 556 REISERFS_ERROR_CONTINUE, 557 558 REISERFS_USRQUOTA, /* User quota option specified */ 559 REISERFS_GRPQUOTA, /* Group quota option specified */ 560 561 REISERFS_TEST1, 562 REISERFS_TEST2, 563 REISERFS_TEST3, 564 REISERFS_TEST4, 565 REISERFS_UNSUPPORTED_OPT, 566 }; 567 568 #define reiserfs_r5_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_R5_HASH)) 569 #define reiserfs_rupasov_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_RUPASOV_HASH)) 570 #define reiserfs_tea_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_TEA_HASH)) 571 #define reiserfs_hash_detect(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_HASH_DETECT)) 572 #define reiserfs_no_border(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_BORDER)) 573 #define reiserfs_no_unhashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_UNHASHED_RELOCATION)) 574 #define reiserfs_hashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_HASHED_RELOCATION)) 575 #define reiserfs_test4(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_TEST4)) 576 577 #define have_large_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_LARGETAIL)) 578 #define have_small_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_SMALLTAIL)) 579 #define replay_only(s) (REISERFS_SB(s)->s_mount_opt & (1 << REPLAYONLY)) 580 #define reiserfs_attrs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ATTRS)) 581 #define old_format_only(s) (REISERFS_SB(s)->s_properties & (1 << REISERFS_3_5)) 582 #define convert_reiserfs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_CONVERT)) 583 #define reiserfs_data_log(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_LOG)) 584 #define reiserfs_data_ordered(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_ORDERED)) 585 #define reiserfs_data_writeback(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_WRITEBACK)) 586 #define reiserfs_xattrs_user(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_XATTRS_USER)) 587 #define reiserfs_posixacl(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_POSIXACL)) 588 #define reiserfs_expose_privroot(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_EXPOSE_PRIVROOT)) 589 #define reiserfs_xattrs_optional(s) (reiserfs_xattrs_user(s) || reiserfs_posixacl(s)) 590 #define reiserfs_barrier_none(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_NONE)) 591 #define reiserfs_barrier_flush(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_FLUSH)) 592 593 #define reiserfs_error_panic(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_PANIC)) 594 #define reiserfs_error_ro(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_RO)) 595 596 void reiserfs_file_buffer(struct buffer_head *bh, int list); 597 extern struct file_system_type reiserfs_fs_type; 598 int reiserfs_resize(struct super_block *, unsigned long); 599 600 #define CARRY_ON 0 601 #define SCHEDULE_OCCURRED 1 602 603 #define SB_BUFFER_WITH_SB(s) (REISERFS_SB(s)->s_sbh) 604 #define SB_JOURNAL(s) (REISERFS_SB(s)->s_journal) 605 #define SB_JOURNAL_1st_RESERVED_BLOCK(s) (SB_JOURNAL(s)->j_1st_reserved_block) 606 #define SB_JOURNAL_LEN_FREE(s) (SB_JOURNAL(s)->j_journal_len_free) 607 #define SB_AP_BITMAP(s) (REISERFS_SB(s)->s_ap_bitmap) 608 609 #define SB_DISK_JOURNAL_HEAD(s) (SB_JOURNAL(s)->j_header_bh->) 610 611 #define reiserfs_is_journal_aborted(journal) (unlikely (__reiserfs_is_journal_aborted (journal))) 612 static inline int __reiserfs_is_journal_aborted(struct reiserfs_journal 613 *journal) 614 { 615 return test_bit(J_ABORTED, &journal->j_state); 616 } 617 618 /* 619 * Locking primitives. The write lock is a per superblock 620 * special mutex that has properties close to the Big Kernel Lock 621 * which was used in the previous locking scheme. 622 */ 623 void reiserfs_write_lock(struct super_block *s); 624 void reiserfs_write_unlock(struct super_block *s); 625 int __must_check reiserfs_write_unlock_nested(struct super_block *s); 626 void reiserfs_write_lock_nested(struct super_block *s, int depth); 627 628 #ifdef CONFIG_REISERFS_CHECK 629 void reiserfs_lock_check_recursive(struct super_block *s); 630 #else 631 static inline void reiserfs_lock_check_recursive(struct super_block *s) { } 632 #endif 633 634 /* 635 * Several mutexes depend on the write lock. 636 * However sometimes we want to relax the write lock while we hold 637 * these mutexes, according to the release/reacquire on schedule() 638 * properties of the Bkl that were used. 639 * Reiserfs performances and locking were based on this scheme. 640 * Now that the write lock is a mutex and not the bkl anymore, doing so 641 * may result in a deadlock: 642 * 643 * A acquire write_lock 644 * A acquire j_commit_mutex 645 * A release write_lock and wait for something 646 * B acquire write_lock 647 * B can't acquire j_commit_mutex and sleep 648 * A can't acquire write lock anymore 649 * deadlock 650 * 651 * What we do here is avoiding such deadlock by playing the same game 652 * than the Bkl: if we can't acquire a mutex that depends on the write lock, 653 * we release the write lock, wait a bit and then retry. 654 * 655 * The mutexes concerned by this hack are: 656 * - The commit mutex of a journal list 657 * - The flush mutex 658 * - The journal lock 659 * - The inode mutex 660 */ 661 static inline void reiserfs_mutex_lock_safe(struct mutex *m, 662 struct super_block *s) 663 { 664 int depth; 665 666 depth = reiserfs_write_unlock_nested(s); 667 mutex_lock(m); 668 reiserfs_write_lock_nested(s, depth); 669 } 670 671 static inline void 672 reiserfs_mutex_lock_nested_safe(struct mutex *m, unsigned int subclass, 673 struct super_block *s) 674 { 675 int depth; 676 677 depth = reiserfs_write_unlock_nested(s); 678 mutex_lock_nested(m, subclass); 679 reiserfs_write_lock_nested(s, depth); 680 } 681 682 static inline void 683 reiserfs_down_read_safe(struct rw_semaphore *sem, struct super_block *s) 684 { 685 int depth; 686 depth = reiserfs_write_unlock_nested(s); 687 down_read(sem); 688 reiserfs_write_lock_nested(s, depth); 689 } 690 691 /* 692 * When we schedule, we usually want to also release the write lock, 693 * according to the previous bkl based locking scheme of reiserfs. 694 */ 695 static inline void reiserfs_cond_resched(struct super_block *s) 696 { 697 if (need_resched()) { 698 int depth; 699 700 depth = reiserfs_write_unlock_nested(s); 701 schedule(); 702 reiserfs_write_lock_nested(s, depth); 703 } 704 } 705 706 struct fid; 707 708 /* in reading the #defines, it may help to understand that they employ 709 the following abbreviations: 710 711 B = Buffer 712 I = Item header 713 H = Height within the tree (should be changed to LEV) 714 N = Number of the item in the node 715 STAT = stat data 716 DEH = Directory Entry Header 717 EC = Entry Count 718 E = Entry number 719 UL = Unsigned Long 720 BLKH = BLocK Header 721 UNFM = UNForMatted node 722 DC = Disk Child 723 P = Path 724 725 These #defines are named by concatenating these abbreviations, 726 where first comes the arguments, and last comes the return value, 727 of the macro. 728 729 */ 730 731 #define USE_INODE_GENERATION_COUNTER 732 733 #define REISERFS_PREALLOCATE 734 #define DISPLACE_NEW_PACKING_LOCALITIES 735 #define PREALLOCATION_SIZE 9 736 737 /* n must be power of 2 */ 738 #define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u)) 739 740 // to be ok for alpha and others we have to align structures to 8 byte 741 // boundary. 742 // FIXME: do not change 4 by anything else: there is code which relies on that 743 #define ROUND_UP(x) _ROUND_UP(x,8LL) 744 745 /* debug levels. Right now, CONFIG_REISERFS_CHECK means print all debug 746 ** messages. 747 */ 748 #define REISERFS_DEBUG_CODE 5 /* extra messages to help find/debug errors */ 749 750 void __reiserfs_warning(struct super_block *s, const char *id, 751 const char *func, const char *fmt, ...); 752 #define reiserfs_warning(s, id, fmt, args...) \ 753 __reiserfs_warning(s, id, __func__, fmt, ##args) 754 /* assertions handling */ 755 756 /** always check a condition and panic if it's false. */ 757 #define __RASSERT(cond, scond, format, args...) \ 758 do { \ 759 if (!(cond)) \ 760 reiserfs_panic(NULL, "assertion failure", "(" #cond ") at " \ 761 __FILE__ ":%i:%s: " format "\n", \ 762 in_interrupt() ? -1 : task_pid_nr(current), \ 763 __LINE__, __func__ , ##args); \ 764 } while (0) 765 766 #define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args) 767 768 #if defined( CONFIG_REISERFS_CHECK ) 769 #define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args) 770 #else 771 #define RFALSE( cond, format, args... ) do {;} while( 0 ) 772 #endif 773 774 #define CONSTF __attribute_const__ 775 /* 776 * Disk Data Structures 777 */ 778 779 /***************************************************************************/ 780 /* SUPER BLOCK */ 781 /***************************************************************************/ 782 783 /* 784 * Structure of super block on disk, a version of which in RAM is often accessed as REISERFS_SB(s)->s_rs 785 * the version in RAM is part of a larger structure containing fields never written to disk. 786 */ 787 #define UNSET_HASH 0 // read_super will guess about, what hash names 788 // in directories were sorted with 789 #define TEA_HASH 1 790 #define YURA_HASH 2 791 #define R5_HASH 3 792 #define DEFAULT_HASH R5_HASH 793 794 struct journal_params { 795 __le32 jp_journal_1st_block; /* where does journal start from on its 796 * device */ 797 __le32 jp_journal_dev; /* journal device st_rdev */ 798 __le32 jp_journal_size; /* size of the journal */ 799 __le32 jp_journal_trans_max; /* max number of blocks in a transaction. */ 800 __le32 jp_journal_magic; /* random value made on fs creation (this 801 * was sb_journal_block_count) */ 802 __le32 jp_journal_max_batch; /* max number of blocks to batch into a 803 * trans */ 804 __le32 jp_journal_max_commit_age; /* in seconds, how old can an async 805 * commit be */ 806 __le32 jp_journal_max_trans_age; /* in seconds, how old can a transaction 807 * be */ 808 }; 809 810 /* this is the super from 3.5.X, where X >= 10 */ 811 struct reiserfs_super_block_v1 { 812 __le32 s_block_count; /* blocks count */ 813 __le32 s_free_blocks; /* free blocks count */ 814 __le32 s_root_block; /* root block number */ 815 struct journal_params s_journal; 816 __le16 s_blocksize; /* block size */ 817 __le16 s_oid_maxsize; /* max size of object id array, see 818 * get_objectid() commentary */ 819 __le16 s_oid_cursize; /* current size of object id array */ 820 __le16 s_umount_state; /* this is set to 1 when filesystem was 821 * umounted, to 2 - when not */ 822 char s_magic[10]; /* reiserfs magic string indicates that 823 * file system is reiserfs: 824 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs" */ 825 __le16 s_fs_state; /* it is set to used by fsck to mark which 826 * phase of rebuilding is done */ 827 __le32 s_hash_function_code; /* indicate, what hash function is being use 828 * to sort names in a directory*/ 829 __le16 s_tree_height; /* height of disk tree */ 830 __le16 s_bmap_nr; /* amount of bitmap blocks needed to address 831 * each block of file system */ 832 __le16 s_version; /* this field is only reliable on filesystem 833 * with non-standard journal */ 834 __le16 s_reserved_for_journal; /* size in blocks of journal area on main 835 * device, we need to keep after 836 * making fs with non-standard journal */ 837 } __attribute__ ((__packed__)); 838 839 #define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1)) 840 841 /* this is the on disk super block */ 842 struct reiserfs_super_block { 843 struct reiserfs_super_block_v1 s_v1; 844 __le32 s_inode_generation; 845 __le32 s_flags; /* Right now used only by inode-attributes, if enabled */ 846 unsigned char s_uuid[16]; /* filesystem unique identifier */ 847 unsigned char s_label[16]; /* filesystem volume label */ 848 __le16 s_mnt_count; /* Count of mounts since last fsck */ 849 __le16 s_max_mnt_count; /* Maximum mounts before check */ 850 __le32 s_lastcheck; /* Timestamp of last fsck */ 851 __le32 s_check_interval; /* Interval between checks */ 852 char s_unused[76]; /* zero filled by mkreiserfs and 853 * reiserfs_convert_objectid_map_v1() 854 * so any additions must be updated 855 * there as well. */ 856 } __attribute__ ((__packed__)); 857 858 #define SB_SIZE (sizeof(struct reiserfs_super_block)) 859 860 #define REISERFS_VERSION_1 0 861 #define REISERFS_VERSION_2 2 862 863 // on-disk super block fields converted to cpu form 864 #define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs) 865 #define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1)) 866 #define SB_BLOCKSIZE(s) \ 867 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize)) 868 #define SB_BLOCK_COUNT(s) \ 869 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count)) 870 #define SB_FREE_BLOCKS(s) \ 871 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks)) 872 #define SB_REISERFS_MAGIC(s) \ 873 (SB_V1_DISK_SUPER_BLOCK(s)->s_magic) 874 #define SB_ROOT_BLOCK(s) \ 875 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block)) 876 #define SB_TREE_HEIGHT(s) \ 877 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height)) 878 #define SB_REISERFS_STATE(s) \ 879 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state)) 880 #define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version)) 881 #define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr)) 882 883 #define PUT_SB_BLOCK_COUNT(s, val) \ 884 do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0) 885 #define PUT_SB_FREE_BLOCKS(s, val) \ 886 do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0) 887 #define PUT_SB_ROOT_BLOCK(s, val) \ 888 do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0) 889 #define PUT_SB_TREE_HEIGHT(s, val) \ 890 do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0) 891 #define PUT_SB_REISERFS_STATE(s, val) \ 892 do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0) 893 #define PUT_SB_VERSION(s, val) \ 894 do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0) 895 #define PUT_SB_BMAP_NR(s, val) \ 896 do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0) 897 898 #define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal) 899 #define SB_ONDISK_JOURNAL_SIZE(s) \ 900 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size)) 901 #define SB_ONDISK_JOURNAL_1st_BLOCK(s) \ 902 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block)) 903 #define SB_ONDISK_JOURNAL_DEVICE(s) \ 904 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev)) 905 #define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \ 906 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal)) 907 908 #define is_block_in_log_or_reserved_area(s, block) \ 909 block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \ 910 && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) + \ 911 ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \ 912 SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s))) 913 914 int is_reiserfs_3_5(struct reiserfs_super_block *rs); 915 int is_reiserfs_3_6(struct reiserfs_super_block *rs); 916 int is_reiserfs_jr(struct reiserfs_super_block *rs); 917 918 /* ReiserFS leaves the first 64k unused, so that partition labels have 919 enough space. If someone wants to write a fancy bootloader that 920 needs more than 64k, let us know, and this will be increased in size. 921 This number must be larger than than the largest block size on any 922 platform, or code will break. -Hans */ 923 #define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024) 924 #define REISERFS_FIRST_BLOCK unused_define 925 #define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES 926 927 /* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */ 928 #define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024) 929 930 /* reiserfs internal error code (used by search_by_key and fix_nodes)) */ 931 #define CARRY_ON 0 932 #define REPEAT_SEARCH -1 933 #define IO_ERROR -2 934 #define NO_DISK_SPACE -3 935 #define NO_BALANCING_NEEDED (-4) 936 #define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5) 937 #define QUOTA_EXCEEDED -6 938 939 typedef __u32 b_blocknr_t; 940 typedef __le32 unp_t; 941 942 struct unfm_nodeinfo { 943 unp_t unfm_nodenum; 944 unsigned short unfm_freespace; 945 }; 946 947 /* there are two formats of keys: 3.5 and 3.6 948 */ 949 #define KEY_FORMAT_3_5 0 950 #define KEY_FORMAT_3_6 1 951 952 /* there are two stat datas */ 953 #define STAT_DATA_V1 0 954 #define STAT_DATA_V2 1 955 956 static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode) 957 { 958 return container_of(inode, struct reiserfs_inode_info, vfs_inode); 959 } 960 961 static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb) 962 { 963 return sb->s_fs_info; 964 } 965 966 /* Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u16 967 * which overflows on large file systems. */ 968 static inline __u32 reiserfs_bmap_count(struct super_block *sb) 969 { 970 return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1; 971 } 972 973 static inline int bmap_would_wrap(unsigned bmap_nr) 974 { 975 return bmap_nr > ((1LL << 16) - 1); 976 } 977 978 /** this says about version of key of all items (but stat data) the 979 object consists of */ 980 #define get_inode_item_key_version( inode ) \ 981 ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5) 982 983 #define set_inode_item_key_version( inode, version ) \ 984 ({ if((version)==KEY_FORMAT_3_6) \ 985 REISERFS_I(inode)->i_flags |= i_item_key_version_mask; \ 986 else \ 987 REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; }) 988 989 #define get_inode_sd_version(inode) \ 990 ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1) 991 992 #define set_inode_sd_version(inode, version) \ 993 ({ if((version)==STAT_DATA_V2) \ 994 REISERFS_I(inode)->i_flags |= i_stat_data_version_mask; \ 995 else \ 996 REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; }) 997 998 /* This is an aggressive tail suppression policy, I am hoping it 999 improves our benchmarks. The principle behind it is that percentage 1000 space saving is what matters, not absolute space saving. This is 1001 non-intuitive, but it helps to understand it if you consider that the 1002 cost to access 4 blocks is not much more than the cost to access 1 1003 block, if you have to do a seek and rotate. A tail risks a 1004 non-linear disk access that is significant as a percentage of total 1005 time cost for a 4 block file and saves an amount of space that is 1006 less significant as a percentage of space, or so goes the hypothesis. 1007 -Hans */ 1008 #define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \ 1009 (\ 1010 (!(n_tail_size)) || \ 1011 (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \ 1012 ( (n_file_size) >= (n_block_size) * 4 ) || \ 1013 ( ( (n_file_size) >= (n_block_size) * 3 ) && \ 1014 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \ 1015 ( ( (n_file_size) >= (n_block_size) * 2 ) && \ 1016 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \ 1017 ( ( (n_file_size) >= (n_block_size) ) && \ 1018 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \ 1019 ) 1020 1021 /* Another strategy for tails, this one means only create a tail if all the 1022 file would fit into one DIRECT item. 1023 Primary intention for this one is to increase performance by decreasing 1024 seeking. 1025 */ 1026 #define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \ 1027 (\ 1028 (!(n_tail_size)) || \ 1029 (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \ 1030 ) 1031 1032 /* 1033 * values for s_umount_state field 1034 */ 1035 #define REISERFS_VALID_FS 1 1036 #define REISERFS_ERROR_FS 2 1037 1038 // 1039 // there are 5 item types currently 1040 // 1041 #define TYPE_STAT_DATA 0 1042 #define TYPE_INDIRECT 1 1043 #define TYPE_DIRECT 2 1044 #define TYPE_DIRENTRY 3 1045 #define TYPE_MAXTYPE 3 1046 #define TYPE_ANY 15 // FIXME: comment is required 1047 1048 /***************************************************************************/ 1049 /* KEY & ITEM HEAD */ 1050 /***************************************************************************/ 1051 1052 // 1053 // directories use this key as well as old files 1054 // 1055 struct offset_v1 { 1056 __le32 k_offset; 1057 __le32 k_uniqueness; 1058 } __attribute__ ((__packed__)); 1059 1060 struct offset_v2 { 1061 __le64 v; 1062 } __attribute__ ((__packed__)); 1063 1064 static inline __u16 offset_v2_k_type(const struct offset_v2 *v2) 1065 { 1066 __u8 type = le64_to_cpu(v2->v) >> 60; 1067 return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY; 1068 } 1069 1070 static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type) 1071 { 1072 v2->v = 1073 (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60); 1074 } 1075 1076 static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2) 1077 { 1078 return le64_to_cpu(v2->v) & (~0ULL >> 4); 1079 } 1080 1081 static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset) 1082 { 1083 offset &= (~0ULL >> 4); 1084 v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset); 1085 } 1086 1087 /* Key of an item determines its location in the S+tree, and 1088 is composed of 4 components */ 1089 struct reiserfs_key { 1090 __le32 k_dir_id; /* packing locality: by default parent 1091 directory object id */ 1092 __le32 k_objectid; /* object identifier */ 1093 union { 1094 struct offset_v1 k_offset_v1; 1095 struct offset_v2 k_offset_v2; 1096 } __attribute__ ((__packed__)) u; 1097 } __attribute__ ((__packed__)); 1098 1099 struct in_core_key { 1100 __u32 k_dir_id; /* packing locality: by default parent 1101 directory object id */ 1102 __u32 k_objectid; /* object identifier */ 1103 __u64 k_offset; 1104 __u8 k_type; 1105 }; 1106 1107 struct cpu_key { 1108 struct in_core_key on_disk_key; 1109 int version; 1110 int key_length; /* 3 in all cases but direct2indirect and 1111 indirect2direct conversion */ 1112 }; 1113 1114 /* Our function for comparing keys can compare keys of different 1115 lengths. It takes as a parameter the length of the keys it is to 1116 compare. These defines are used in determining what is to be passed 1117 to it as that parameter. */ 1118 #define REISERFS_FULL_KEY_LEN 4 1119 #define REISERFS_SHORT_KEY_LEN 2 1120 1121 /* The result of the key compare */ 1122 #define FIRST_GREATER 1 1123 #define SECOND_GREATER -1 1124 #define KEYS_IDENTICAL 0 1125 #define KEY_FOUND 1 1126 #define KEY_NOT_FOUND 0 1127 1128 #define KEY_SIZE (sizeof(struct reiserfs_key)) 1129 #define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32)) 1130 1131 /* return values for search_by_key and clones */ 1132 #define ITEM_FOUND 1 1133 #define ITEM_NOT_FOUND 0 1134 #define ENTRY_FOUND 1 1135 #define ENTRY_NOT_FOUND 0 1136 #define DIRECTORY_NOT_FOUND -1 1137 #define REGULAR_FILE_FOUND -2 1138 #define DIRECTORY_FOUND -3 1139 #define BYTE_FOUND 1 1140 #define BYTE_NOT_FOUND 0 1141 #define FILE_NOT_FOUND -1 1142 1143 #define POSITION_FOUND 1 1144 #define POSITION_NOT_FOUND 0 1145 1146 // return values for reiserfs_find_entry and search_by_entry_key 1147 #define NAME_FOUND 1 1148 #define NAME_NOT_FOUND 0 1149 #define GOTO_PREVIOUS_ITEM 2 1150 #define NAME_FOUND_INVISIBLE 3 1151 1152 /* Everything in the filesystem is stored as a set of items. The 1153 item head contains the key of the item, its free space (for 1154 indirect items) and specifies the location of the item itself 1155 within the block. */ 1156 1157 struct item_head { 1158 /* Everything in the tree is found by searching for it based on 1159 * its key.*/ 1160 struct reiserfs_key ih_key; 1161 union { 1162 /* The free space in the last unformatted node of an 1163 indirect item if this is an indirect item. This 1164 equals 0xFFFF iff this is a direct item or stat data 1165 item. Note that the key, not this field, is used to 1166 determine the item type, and thus which field this 1167 union contains. */ 1168 __le16 ih_free_space_reserved; 1169 /* Iff this is a directory item, this field equals the 1170 number of directory entries in the directory item. */ 1171 __le16 ih_entry_count; 1172 } __attribute__ ((__packed__)) u; 1173 __le16 ih_item_len; /* total size of the item body */ 1174 __le16 ih_item_location; /* an offset to the item body 1175 * within the block */ 1176 __le16 ih_version; /* 0 for all old items, 2 for new 1177 ones. Highest bit is set by fsck 1178 temporary, cleaned after all 1179 done */ 1180 } __attribute__ ((__packed__)); 1181 /* size of item header */ 1182 #define IH_SIZE (sizeof(struct item_head)) 1183 1184 #define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved) 1185 #define ih_version(ih) le16_to_cpu((ih)->ih_version) 1186 #define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count) 1187 #define ih_location(ih) le16_to_cpu((ih)->ih_item_location) 1188 #define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len) 1189 1190 #define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0) 1191 #define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0) 1192 #define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0) 1193 #define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0) 1194 #define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0) 1195 1196 #define unreachable_item(ih) (ih_version(ih) & (1 << 15)) 1197 1198 #define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih)) 1199 #define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val))) 1200 1201 /* these operate on indirect items, where you've got an array of ints 1202 ** at a possibly unaligned location. These are a noop on ia32 1203 ** 1204 ** p is the array of __u32, i is the index into the array, v is the value 1205 ** to store there. 1206 */ 1207 #define get_block_num(p, i) get_unaligned_le32((p) + (i)) 1208 #define put_block_num(p, i, v) put_unaligned_le32((v), (p) + (i)) 1209 1210 // 1211 // in old version uniqueness field shows key type 1212 // 1213 #define V1_SD_UNIQUENESS 0 1214 #define V1_INDIRECT_UNIQUENESS 0xfffffffe 1215 #define V1_DIRECT_UNIQUENESS 0xffffffff 1216 #define V1_DIRENTRY_UNIQUENESS 500 1217 #define V1_ANY_UNIQUENESS 555 // FIXME: comment is required 1218 1219 // 1220 // here are conversion routines 1221 // 1222 static inline int uniqueness2type(__u32 uniqueness) CONSTF; 1223 static inline int uniqueness2type(__u32 uniqueness) 1224 { 1225 switch ((int)uniqueness) { 1226 case V1_SD_UNIQUENESS: 1227 return TYPE_STAT_DATA; 1228 case V1_INDIRECT_UNIQUENESS: 1229 return TYPE_INDIRECT; 1230 case V1_DIRECT_UNIQUENESS: 1231 return TYPE_DIRECT; 1232 case V1_DIRENTRY_UNIQUENESS: 1233 return TYPE_DIRENTRY; 1234 case V1_ANY_UNIQUENESS: 1235 default: 1236 return TYPE_ANY; 1237 } 1238 } 1239 1240 static inline __u32 type2uniqueness(int type) CONSTF; 1241 static inline __u32 type2uniqueness(int type) 1242 { 1243 switch (type) { 1244 case TYPE_STAT_DATA: 1245 return V1_SD_UNIQUENESS; 1246 case TYPE_INDIRECT: 1247 return V1_INDIRECT_UNIQUENESS; 1248 case TYPE_DIRECT: 1249 return V1_DIRECT_UNIQUENESS; 1250 case TYPE_DIRENTRY: 1251 return V1_DIRENTRY_UNIQUENESS; 1252 case TYPE_ANY: 1253 default: 1254 return V1_ANY_UNIQUENESS; 1255 } 1256 } 1257 1258 // 1259 // key is pointer to on disk key which is stored in le, result is cpu, 1260 // there is no way to get version of object from key, so, provide 1261 // version to these defines 1262 // 1263 static inline loff_t le_key_k_offset(int version, 1264 const struct reiserfs_key *key) 1265 { 1266 return (version == KEY_FORMAT_3_5) ? 1267 le32_to_cpu(key->u.k_offset_v1.k_offset) : 1268 offset_v2_k_offset(&(key->u.k_offset_v2)); 1269 } 1270 1271 static inline loff_t le_ih_k_offset(const struct item_head *ih) 1272 { 1273 return le_key_k_offset(ih_version(ih), &(ih->ih_key)); 1274 } 1275 1276 static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key) 1277 { 1278 return (version == KEY_FORMAT_3_5) ? 1279 uniqueness2type(le32_to_cpu(key->u.k_offset_v1.k_uniqueness)) : 1280 offset_v2_k_type(&(key->u.k_offset_v2)); 1281 } 1282 1283 static inline loff_t le_ih_k_type(const struct item_head *ih) 1284 { 1285 return le_key_k_type(ih_version(ih), &(ih->ih_key)); 1286 } 1287 1288 static inline void set_le_key_k_offset(int version, struct reiserfs_key *key, 1289 loff_t offset) 1290 { 1291 (version == KEY_FORMAT_3_5) ? (void)(key->u.k_offset_v1.k_offset = cpu_to_le32(offset)) : /* jdm check */ 1292 (void)(set_offset_v2_k_offset(&(key->u.k_offset_v2), offset)); 1293 } 1294 1295 static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset) 1296 { 1297 set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset); 1298 } 1299 1300 static inline void set_le_key_k_type(int version, struct reiserfs_key *key, 1301 int type) 1302 { 1303 (version == KEY_FORMAT_3_5) ? 1304 (void)(key->u.k_offset_v1.k_uniqueness = 1305 cpu_to_le32(type2uniqueness(type))) 1306 : (void)(set_offset_v2_k_type(&(key->u.k_offset_v2), type)); 1307 } 1308 1309 static inline void set_le_ih_k_type(struct item_head *ih, int type) 1310 { 1311 set_le_key_k_type(ih_version(ih), &(ih->ih_key), type); 1312 } 1313 1314 static inline int is_direntry_le_key(int version, struct reiserfs_key *key) 1315 { 1316 return le_key_k_type(version, key) == TYPE_DIRENTRY; 1317 } 1318 1319 static inline int is_direct_le_key(int version, struct reiserfs_key *key) 1320 { 1321 return le_key_k_type(version, key) == TYPE_DIRECT; 1322 } 1323 1324 static inline int is_indirect_le_key(int version, struct reiserfs_key *key) 1325 { 1326 return le_key_k_type(version, key) == TYPE_INDIRECT; 1327 } 1328 1329 static inline int is_statdata_le_key(int version, struct reiserfs_key *key) 1330 { 1331 return le_key_k_type(version, key) == TYPE_STAT_DATA; 1332 } 1333 1334 // 1335 // item header has version. 1336 // 1337 static inline int is_direntry_le_ih(struct item_head *ih) 1338 { 1339 return is_direntry_le_key(ih_version(ih), &ih->ih_key); 1340 } 1341 1342 static inline int is_direct_le_ih(struct item_head *ih) 1343 { 1344 return is_direct_le_key(ih_version(ih), &ih->ih_key); 1345 } 1346 1347 static inline int is_indirect_le_ih(struct item_head *ih) 1348 { 1349 return is_indirect_le_key(ih_version(ih), &ih->ih_key); 1350 } 1351 1352 static inline int is_statdata_le_ih(struct item_head *ih) 1353 { 1354 return is_statdata_le_key(ih_version(ih), &ih->ih_key); 1355 } 1356 1357 // 1358 // key is pointer to cpu key, result is cpu 1359 // 1360 static inline loff_t cpu_key_k_offset(const struct cpu_key *key) 1361 { 1362 return key->on_disk_key.k_offset; 1363 } 1364 1365 static inline loff_t cpu_key_k_type(const struct cpu_key *key) 1366 { 1367 return key->on_disk_key.k_type; 1368 } 1369 1370 static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset) 1371 { 1372 key->on_disk_key.k_offset = offset; 1373 } 1374 1375 static inline void set_cpu_key_k_type(struct cpu_key *key, int type) 1376 { 1377 key->on_disk_key.k_type = type; 1378 } 1379 1380 static inline void cpu_key_k_offset_dec(struct cpu_key *key) 1381 { 1382 key->on_disk_key.k_offset--; 1383 } 1384 1385 #define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY) 1386 #define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT) 1387 #define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT) 1388 #define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA) 1389 1390 /* are these used ? */ 1391 #define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key))) 1392 #define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key))) 1393 #define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key))) 1394 #define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key))) 1395 1396 #define I_K_KEY_IN_ITEM(ih, key, n_blocksize) \ 1397 (!COMP_SHORT_KEYS(ih, key) && \ 1398 I_OFF_BYTE_IN_ITEM(ih, k_offset(key), n_blocksize)) 1399 1400 /* maximal length of item */ 1401 #define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE) 1402 #define MIN_ITEM_LEN 1 1403 1404 /* object identifier for root dir */ 1405 #define REISERFS_ROOT_OBJECTID 2 1406 #define REISERFS_ROOT_PARENT_OBJECTID 1 1407 1408 extern struct reiserfs_key root_key; 1409 1410 /* 1411 * Picture represents a leaf of the S+tree 1412 * ______________________________________________________ 1413 * | | Array of | | | 1414 * |Block | Object-Item | F r e e | Objects- | 1415 * | head | Headers | S p a c e | Items | 1416 * |______|_______________|___________________|___________| 1417 */ 1418 1419 /* Header of a disk block. More precisely, header of a formatted leaf 1420 or internal node, and not the header of an unformatted node. */ 1421 struct block_head { 1422 __le16 blk_level; /* Level of a block in the tree. */ 1423 __le16 blk_nr_item; /* Number of keys/items in a block. */ 1424 __le16 blk_free_space; /* Block free space in bytes. */ 1425 __le16 blk_reserved; 1426 /* dump this in v4/planA */ 1427 struct reiserfs_key blk_right_delim_key; /* kept only for compatibility */ 1428 }; 1429 1430 #define BLKH_SIZE (sizeof(struct block_head)) 1431 #define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level)) 1432 #define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item)) 1433 #define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space)) 1434 #define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved)) 1435 #define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val)) 1436 #define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val)) 1437 #define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val)) 1438 #define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val)) 1439 #define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key) 1440 #define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val) 1441 1442 /* 1443 * values for blk_level field of the struct block_head 1444 */ 1445 1446 #define FREE_LEVEL 0 /* when node gets removed from the tree its 1447 blk_level is set to FREE_LEVEL. It is then 1448 used to see whether the node is still in the 1449 tree */ 1450 1451 #define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level. */ 1452 1453 /* Given the buffer head of a formatted node, resolve to the block head of that node. */ 1454 #define B_BLK_HEAD(bh) ((struct block_head *)((bh)->b_data)) 1455 /* Number of items that are in buffer. */ 1456 #define B_NR_ITEMS(bh) (blkh_nr_item(B_BLK_HEAD(bh))) 1457 #define B_LEVEL(bh) (blkh_level(B_BLK_HEAD(bh))) 1458 #define B_FREE_SPACE(bh) (blkh_free_space(B_BLK_HEAD(bh))) 1459 1460 #define PUT_B_NR_ITEMS(bh, val) do { set_blkh_nr_item(B_BLK_HEAD(bh), val); } while (0) 1461 #define PUT_B_LEVEL(bh, val) do { set_blkh_level(B_BLK_HEAD(bh), val); } while (0) 1462 #define PUT_B_FREE_SPACE(bh, val) do { set_blkh_free_space(B_BLK_HEAD(bh), val); } while (0) 1463 1464 /* Get right delimiting key. -- little endian */ 1465 #define B_PRIGHT_DELIM_KEY(bh) (&(blk_right_delim_key(B_BLK_HEAD(bh)))) 1466 1467 /* Does the buffer contain a disk leaf. */ 1468 #define B_IS_ITEMS_LEVEL(bh) (B_LEVEL(bh) == DISK_LEAF_NODE_LEVEL) 1469 1470 /* Does the buffer contain a disk internal node */ 1471 #define B_IS_KEYS_LEVEL(bh) (B_LEVEL(bh) > DISK_LEAF_NODE_LEVEL \ 1472 && B_LEVEL(bh) <= MAX_HEIGHT) 1473 1474 /***************************************************************************/ 1475 /* STAT DATA */ 1476 /***************************************************************************/ 1477 1478 // 1479 // old stat data is 32 bytes long. We are going to distinguish new one by 1480 // different size 1481 // 1482 struct stat_data_v1 { 1483 __le16 sd_mode; /* file type, permissions */ 1484 __le16 sd_nlink; /* number of hard links */ 1485 __le16 sd_uid; /* owner */ 1486 __le16 sd_gid; /* group */ 1487 __le32 sd_size; /* file size */ 1488 __le32 sd_atime; /* time of last access */ 1489 __le32 sd_mtime; /* time file was last modified */ 1490 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */ 1491 union { 1492 __le32 sd_rdev; 1493 __le32 sd_blocks; /* number of blocks file uses */ 1494 } __attribute__ ((__packed__)) u; 1495 __le32 sd_first_direct_byte; /* first byte of file which is stored 1496 in a direct item: except that if it 1497 equals 1 it is a symlink and if it 1498 equals ~(__u32)0 there is no 1499 direct item. The existence of this 1500 field really grates on me. Let's 1501 replace it with a macro based on 1502 sd_size and our tail suppression 1503 policy. Someday. -Hans */ 1504 } __attribute__ ((__packed__)); 1505 1506 #define SD_V1_SIZE (sizeof(struct stat_data_v1)) 1507 #define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5) 1508 #define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode)) 1509 #define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v)) 1510 #define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink)) 1511 #define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v)) 1512 #define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid)) 1513 #define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v)) 1514 #define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid)) 1515 #define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v)) 1516 #define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size)) 1517 #define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v)) 1518 #define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime)) 1519 #define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v)) 1520 #define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime)) 1521 #define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v)) 1522 #define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime)) 1523 #define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v)) 1524 #define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev)) 1525 #define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v)) 1526 #define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks)) 1527 #define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v)) 1528 #define sd_v1_first_direct_byte(sdp) \ 1529 (le32_to_cpu((sdp)->sd_first_direct_byte)) 1530 #define set_sd_v1_first_direct_byte(sdp,v) \ 1531 ((sdp)->sd_first_direct_byte = cpu_to_le32(v)) 1532 1533 /* inode flags stored in sd_attrs (nee sd_reserved) */ 1534 1535 /* we want common flags to have the same values as in ext2, 1536 so chattr(1) will work without problems */ 1537 #define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL 1538 #define REISERFS_APPEND_FL FS_APPEND_FL 1539 #define REISERFS_SYNC_FL FS_SYNC_FL 1540 #define REISERFS_NOATIME_FL FS_NOATIME_FL 1541 #define REISERFS_NODUMP_FL FS_NODUMP_FL 1542 #define REISERFS_SECRM_FL FS_SECRM_FL 1543 #define REISERFS_UNRM_FL FS_UNRM_FL 1544 #define REISERFS_COMPR_FL FS_COMPR_FL 1545 #define REISERFS_NOTAIL_FL FS_NOTAIL_FL 1546 1547 /* persistent flags that file inherits from the parent directory */ 1548 #define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \ 1549 REISERFS_SYNC_FL | \ 1550 REISERFS_NOATIME_FL | \ 1551 REISERFS_NODUMP_FL | \ 1552 REISERFS_SECRM_FL | \ 1553 REISERFS_COMPR_FL | \ 1554 REISERFS_NOTAIL_FL ) 1555 1556 /* Stat Data on disk (reiserfs version of UFS disk inode minus the 1557 address blocks) */ 1558 struct stat_data { 1559 __le16 sd_mode; /* file type, permissions */ 1560 __le16 sd_attrs; /* persistent inode flags */ 1561 __le32 sd_nlink; /* number of hard links */ 1562 __le64 sd_size; /* file size */ 1563 __le32 sd_uid; /* owner */ 1564 __le32 sd_gid; /* group */ 1565 __le32 sd_atime; /* time of last access */ 1566 __le32 sd_mtime; /* time file was last modified */ 1567 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */ 1568 __le32 sd_blocks; 1569 union { 1570 __le32 sd_rdev; 1571 __le32 sd_generation; 1572 //__le32 sd_first_direct_byte; 1573 /* first byte of file which is stored in a 1574 direct item: except that if it equals 1 1575 it is a symlink and if it equals 1576 ~(__u32)0 there is no direct item. The 1577 existence of this field really grates 1578 on me. Let's replace it with a macro 1579 based on sd_size and our tail 1580 suppression policy? */ 1581 } __attribute__ ((__packed__)) u; 1582 } __attribute__ ((__packed__)); 1583 // 1584 // this is 44 bytes long 1585 // 1586 #define SD_SIZE (sizeof(struct stat_data)) 1587 #define SD_V2_SIZE SD_SIZE 1588 #define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6) 1589 #define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode)) 1590 #define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v)) 1591 /* sd_reserved */ 1592 /* set_sd_reserved */ 1593 #define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink)) 1594 #define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v)) 1595 #define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size)) 1596 #define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v)) 1597 #define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid)) 1598 #define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v)) 1599 #define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid)) 1600 #define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v)) 1601 #define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime)) 1602 #define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v)) 1603 #define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime)) 1604 #define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v)) 1605 #define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime)) 1606 #define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v)) 1607 #define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks)) 1608 #define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v)) 1609 #define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev)) 1610 #define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v)) 1611 #define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation)) 1612 #define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v)) 1613 #define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs)) 1614 #define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v)) 1615 1616 /***************************************************************************/ 1617 /* DIRECTORY STRUCTURE */ 1618 /***************************************************************************/ 1619 /* 1620 Picture represents the structure of directory items 1621 ________________________________________________ 1622 | Array of | | | | | | 1623 | directory |N-1| N-2 | .... | 1st |0th| 1624 | entry headers | | | | | | 1625 |_______________|___|_____|________|_______|___| 1626 <---- directory entries ------> 1627 1628 First directory item has k_offset component 1. We store "." and ".." 1629 in one item, always, we never split "." and ".." into differing 1630 items. This makes, among other things, the code for removing 1631 directories simpler. */ 1632 #define SD_OFFSET 0 1633 #define SD_UNIQUENESS 0 1634 #define DOT_OFFSET 1 1635 #define DOT_DOT_OFFSET 2 1636 #define DIRENTRY_UNIQUENESS 500 1637 1638 /* */ 1639 #define FIRST_ITEM_OFFSET 1 1640 1641 /* 1642 Q: How to get key of object pointed to by entry from entry? 1643 1644 A: Each directory entry has its header. This header has deh_dir_id and deh_objectid fields, those are key 1645 of object, entry points to */ 1646 1647 /* NOT IMPLEMENTED: 1648 Directory will someday contain stat data of object */ 1649 1650 struct reiserfs_de_head { 1651 __le32 deh_offset; /* third component of the directory entry key */ 1652 __le32 deh_dir_id; /* objectid of the parent directory of the object, that is referenced 1653 by directory entry */ 1654 __le32 deh_objectid; /* objectid of the object, that is referenced by directory entry */ 1655 __le16 deh_location; /* offset of name in the whole item */ 1656 __le16 deh_state; /* whether 1) entry contains stat data (for future), and 2) whether 1657 entry is hidden (unlinked) */ 1658 } __attribute__ ((__packed__)); 1659 #define DEH_SIZE sizeof(struct reiserfs_de_head) 1660 #define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset)) 1661 #define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id)) 1662 #define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid)) 1663 #define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location)) 1664 #define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state)) 1665 1666 #define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v))) 1667 #define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v))) 1668 #define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v))) 1669 #define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v))) 1670 #define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v))) 1671 1672 /* empty directory contains two entries "." and ".." and their headers */ 1673 #define EMPTY_DIR_SIZE \ 1674 (DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen (".."))) 1675 1676 /* old format directories have this size when empty */ 1677 #define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3) 1678 1679 #define DEH_Statdata 0 /* not used now */ 1680 #define DEH_Visible 2 1681 1682 /* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */ 1683 #if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__) 1684 # define ADDR_UNALIGNED_BITS (3) 1685 #endif 1686 1687 /* These are only used to manipulate deh_state. 1688 * Because of this, we'll use the ext2_ bit routines, 1689 * since they are little endian */ 1690 #ifdef ADDR_UNALIGNED_BITS 1691 1692 # define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1))) 1693 # define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3) 1694 1695 # define set_bit_unaligned(nr, addr) \ 1696 __test_and_set_bit_le((nr) + unaligned_offset(addr), aligned_address(addr)) 1697 # define clear_bit_unaligned(nr, addr) \ 1698 __test_and_clear_bit_le((nr) + unaligned_offset(addr), aligned_address(addr)) 1699 # define test_bit_unaligned(nr, addr) \ 1700 test_bit_le((nr) + unaligned_offset(addr), aligned_address(addr)) 1701 1702 #else 1703 1704 # define set_bit_unaligned(nr, addr) __test_and_set_bit_le(nr, addr) 1705 # define clear_bit_unaligned(nr, addr) __test_and_clear_bit_le(nr, addr) 1706 # define test_bit_unaligned(nr, addr) test_bit_le(nr, addr) 1707 1708 #endif 1709 1710 #define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state)) 1711 #define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state)) 1712 #define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state)) 1713 #define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state)) 1714 1715 #define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state)) 1716 #define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state)) 1717 #define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state)) 1718 1719 extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid, 1720 __le32 par_dirid, __le32 par_objid); 1721 extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid, 1722 __le32 par_dirid, __le32 par_objid); 1723 1724 /* array of the entry headers */ 1725 /* get item body */ 1726 #define B_I_PITEM(bh,ih) ( (bh)->b_data + ih_location(ih) ) 1727 #define B_I_DEH(bh,ih) ((struct reiserfs_de_head *)(B_I_PITEM(bh,ih))) 1728 1729 /* length of the directory entry in directory item. This define 1730 calculates length of i-th directory entry using directory entry 1731 locations from dir entry head. When it calculates length of 0-th 1732 directory entry, it uses length of whole item in place of entry 1733 location of the non-existent following entry in the calculation. 1734 See picture above.*/ 1735 /* 1736 #define I_DEH_N_ENTRY_LENGTH(ih,deh,i) \ 1737 ((i) ? (deh_location((deh)-1) - deh_location((deh))) : (ih_item_len((ih)) - deh_location((deh)))) 1738 */ 1739 static inline int entry_length(const struct buffer_head *bh, 1740 const struct item_head *ih, int pos_in_item) 1741 { 1742 struct reiserfs_de_head *deh; 1743 1744 deh = B_I_DEH(bh, ih) + pos_in_item; 1745 if (pos_in_item) 1746 return deh_location(deh - 1) - deh_location(deh); 1747 1748 return ih_item_len(ih) - deh_location(deh); 1749 } 1750 1751 /* number of entries in the directory item, depends on ENTRY_COUNT being at the start of directory dynamic data. */ 1752 #define I_ENTRY_COUNT(ih) (ih_entry_count((ih))) 1753 1754 /* name by bh, ih and entry_num */ 1755 #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)))) 1756 1757 // two entries per block (at least) 1758 #define REISERFS_MAX_NAME(block_size) 255 1759 1760 /* this structure is used for operations on directory entries. It is 1761 not a disk structure. */ 1762 /* When reiserfs_find_entry or search_by_entry_key find directory 1763 entry, they return filled reiserfs_dir_entry structure */ 1764 struct reiserfs_dir_entry { 1765 struct buffer_head *de_bh; 1766 int de_item_num; 1767 struct item_head *de_ih; 1768 int de_entry_num; 1769 struct reiserfs_de_head *de_deh; 1770 int de_entrylen; 1771 int de_namelen; 1772 char *de_name; 1773 unsigned long *de_gen_number_bit_string; 1774 1775 __u32 de_dir_id; 1776 __u32 de_objectid; 1777 1778 struct cpu_key de_entry_key; 1779 }; 1780 1781 /* these defines are useful when a particular member of a reiserfs_dir_entry is needed */ 1782 1783 /* pointer to file name, stored in entry */ 1784 #define B_I_DEH_ENTRY_FILE_NAME(bh,ih,deh) (B_I_PITEM (bh, ih) + deh_location(deh)) 1785 1786 /* length of name */ 1787 #define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \ 1788 (I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0)) 1789 1790 /* hash value occupies bits from 7 up to 30 */ 1791 #define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL) 1792 /* generation number occupies 7 bits starting from 0 up to 6 */ 1793 #define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL) 1794 #define MAX_GENERATION_NUMBER 127 1795 1796 #define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number)) 1797 1798 /* 1799 * Picture represents an internal node of the reiserfs tree 1800 * ______________________________________________________ 1801 * | | Array of | Array of | Free | 1802 * |block | keys | pointers | space | 1803 * | head | N | N+1 | | 1804 * |______|_______________|___________________|___________| 1805 */ 1806 1807 /***************************************************************************/ 1808 /* DISK CHILD */ 1809 /***************************************************************************/ 1810 /* Disk child pointer: The pointer from an internal node of the tree 1811 to a node that is on disk. */ 1812 struct disk_child { 1813 __le32 dc_block_number; /* Disk child's block number. */ 1814 __le16 dc_size; /* Disk child's used space. */ 1815 __le16 dc_reserved; 1816 }; 1817 1818 #define DC_SIZE (sizeof(struct disk_child)) 1819 #define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number)) 1820 #define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size)) 1821 #define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0) 1822 #define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0) 1823 1824 /* Get disk child by buffer header and position in the tree node. */ 1825 #define B_N_CHILD(bh, n_pos) ((struct disk_child *)\ 1826 ((bh)->b_data + BLKH_SIZE + B_NR_ITEMS(bh) * KEY_SIZE + DC_SIZE * (n_pos))) 1827 1828 /* Get disk child number by buffer header and position in the tree node. */ 1829 #define B_N_CHILD_NUM(bh, n_pos) (dc_block_number(B_N_CHILD(bh, n_pos))) 1830 #define PUT_B_N_CHILD_NUM(bh, n_pos, val) \ 1831 (put_dc_block_number(B_N_CHILD(bh, n_pos), val)) 1832 1833 /* maximal value of field child_size in structure disk_child */ 1834 /* child size is the combined size of all items and their headers */ 1835 #define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE )) 1836 1837 /* amount of used space in buffer (not including block head) */ 1838 #define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur))) 1839 1840 /* max and min number of keys in internal node */ 1841 #define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) ) 1842 #define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2) 1843 1844 /***************************************************************************/ 1845 /* PATH STRUCTURES AND DEFINES */ 1846 /***************************************************************************/ 1847 1848 /* Search_by_key fills up the path from the root to the leaf as it descends the tree looking for the 1849 key. It uses reiserfs_bread to try to find buffers in the cache given their block number. If it 1850 does not find them in the cache it reads them from disk. For each node search_by_key finds using 1851 reiserfs_bread it then uses bin_search to look through that node. bin_search will find the 1852 position of the block_number of the next node if it is looking through an internal node. If it 1853 is looking through a leaf node bin_search will find the position of the item which has key either 1854 equal to given key, or which is the maximal key less than the given key. */ 1855 1856 struct path_element { 1857 struct buffer_head *pe_buffer; /* Pointer to the buffer at the path in the tree. */ 1858 int pe_position; /* Position in the tree node which is placed in the */ 1859 /* buffer above. */ 1860 }; 1861 1862 #define MAX_HEIGHT 5 /* maximal height of a tree. don't change this without changing JOURNAL_PER_BALANCE_CNT */ 1863 #define EXTENDED_MAX_HEIGHT 7 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */ 1864 #define FIRST_PATH_ELEMENT_OFFSET 2 /* Must be equal to at least 2. */ 1865 1866 #define ILLEGAL_PATH_ELEMENT_OFFSET 1 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */ 1867 #define MAX_FEB_SIZE 6 /* this MUST be MAX_HEIGHT + 1. See about FEB below */ 1868 1869 /* We need to keep track of who the ancestors of nodes are. When we 1870 perform a search we record which nodes were visited while 1871 descending the tree looking for the node we searched for. This list 1872 of nodes is called the path. This information is used while 1873 performing balancing. Note that this path information may become 1874 invalid, and this means we must check it when using it to see if it 1875 is still valid. You'll need to read search_by_key and the comments 1876 in it, especially about decrement_counters_in_path(), to understand 1877 this structure. 1878 1879 Paths make the code so much harder to work with and debug.... An 1880 enormous number of bugs are due to them, and trying to write or modify 1881 code that uses them just makes my head hurt. They are based on an 1882 excessive effort to avoid disturbing the precious VFS code.:-( The 1883 gods only know how we are going to SMP the code that uses them. 1884 znodes are the way! */ 1885 1886 #define PATH_READA 0x1 /* do read ahead */ 1887 #define PATH_READA_BACK 0x2 /* read backwards */ 1888 1889 struct treepath { 1890 int path_length; /* Length of the array above. */ 1891 int reada; 1892 struct path_element path_elements[EXTENDED_MAX_HEIGHT]; /* Array of the path elements. */ 1893 int pos_in_item; 1894 }; 1895 1896 #define pos_in_item(path) ((path)->pos_in_item) 1897 1898 #define INITIALIZE_PATH(var) \ 1899 struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,} 1900 1901 /* Get path element by path and path position. */ 1902 #define PATH_OFFSET_PELEMENT(path, n_offset) ((path)->path_elements + (n_offset)) 1903 1904 /* Get buffer header at the path by path and path position. */ 1905 #define PATH_OFFSET_PBUFFER(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_buffer) 1906 1907 /* Get position in the element at the path by path and path position. */ 1908 #define PATH_OFFSET_POSITION(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_position) 1909 1910 #define PATH_PLAST_BUFFER(path) (PATH_OFFSET_PBUFFER((path), (path)->path_length)) 1911 /* you know, to the person who didn't 1912 write this the macro name does not 1913 at first suggest what it does. 1914 Maybe POSITION_FROM_PATH_END? Or 1915 maybe we should just focus on 1916 dumping paths... -Hans */ 1917 #define PATH_LAST_POSITION(path) (PATH_OFFSET_POSITION((path), (path)->path_length)) 1918 1919 #define PATH_PITEM_HEAD(path) B_N_PITEM_HEAD(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path)) 1920 1921 /* in do_balance leaf has h == 0 in contrast with path structure, 1922 where root has level == 0. That is why we need these defines */ 1923 #define PATH_H_PBUFFER(path, h) PATH_OFFSET_PBUFFER (path, path->path_length - (h)) /* tb->S[h] */ 1924 #define PATH_H_PPARENT(path, h) PATH_H_PBUFFER (path, (h) + 1) /* tb->F[h] or tb->S[0]->b_parent */ 1925 #define PATH_H_POSITION(path, h) PATH_OFFSET_POSITION (path, path->path_length - (h)) 1926 #define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1) /* tb->S[h]->b_item_order */ 1927 1928 #define PATH_H_PATH_OFFSET(path, n_h) ((path)->path_length - (n_h)) 1929 1930 #define get_last_bh(path) PATH_PLAST_BUFFER(path) 1931 #define get_ih(path) PATH_PITEM_HEAD(path) 1932 #define get_item_pos(path) PATH_LAST_POSITION(path) 1933 #define get_item(path) ((void *)B_N_PITEM(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION (path))) 1934 #define item_moved(ih,path) comp_items(ih, path) 1935 #define path_changed(ih,path) comp_items (ih, path) 1936 1937 /***************************************************************************/ 1938 /* MISC */ 1939 /***************************************************************************/ 1940 1941 /* Size of pointer to the unformatted node. */ 1942 #define UNFM_P_SIZE (sizeof(unp_t)) 1943 #define UNFM_P_SHIFT 2 1944 1945 // in in-core inode key is stored on le form 1946 #define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key)) 1947 1948 #define MAX_UL_INT 0xffffffff 1949 #define MAX_INT 0x7ffffff 1950 #define MAX_US_INT 0xffff 1951 1952 // reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset 1953 static inline loff_t max_reiserfs_offset(struct inode *inode) 1954 { 1955 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5) 1956 return (loff_t) U32_MAX; 1957 1958 return (loff_t) ((~(__u64) 0) >> 4); 1959 } 1960 1961 /*#define MAX_KEY_UNIQUENESS MAX_UL_INT*/ 1962 #define MAX_KEY_OBJECTID MAX_UL_INT 1963 1964 #define MAX_B_NUM MAX_UL_INT 1965 #define MAX_FC_NUM MAX_US_INT 1966 1967 /* the purpose is to detect overflow of an unsigned short */ 1968 #define REISERFS_LINK_MAX (MAX_US_INT - 1000) 1969 1970 /* The following defines are used in reiserfs_insert_item and reiserfs_append_item */ 1971 #define REISERFS_KERNEL_MEM 0 /* reiserfs kernel memory mode */ 1972 #define REISERFS_USER_MEM 1 /* reiserfs user memory mode */ 1973 1974 #define fs_generation(s) (REISERFS_SB(s)->s_generation_counter) 1975 #define get_generation(s) atomic_read (&fs_generation(s)) 1976 #define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen) 1977 #define __fs_changed(gen,s) (gen != get_generation (s)) 1978 #define fs_changed(gen,s) \ 1979 ({ \ 1980 reiserfs_cond_resched(s); \ 1981 __fs_changed(gen, s); \ 1982 }) 1983 1984 /***************************************************************************/ 1985 /* FIXATE NODES */ 1986 /***************************************************************************/ 1987 1988 #define VI_TYPE_LEFT_MERGEABLE 1 1989 #define VI_TYPE_RIGHT_MERGEABLE 2 1990 1991 /* To make any changes in the tree we always first find node, that 1992 contains item to be changed/deleted or place to insert a new 1993 item. We call this node S. To do balancing we need to decide what 1994 we will shift to left/right neighbor, or to a new node, where new 1995 item will be etc. To make this analysis simpler we build virtual 1996 node. Virtual node is an array of items, that will replace items of 1997 node S. (For instance if we are going to delete an item, virtual 1998 node does not contain it). Virtual node keeps information about 1999 item sizes and types, mergeability of first and last items, sizes 2000 of all entries in directory item. We use this array of items when 2001 calculating what we can shift to neighbors and how many nodes we 2002 have to have if we do not any shiftings, if we shift to left/right 2003 neighbor or to both. */ 2004 struct virtual_item { 2005 int vi_index; // index in the array of item operations 2006 unsigned short vi_type; // left/right mergeability 2007 unsigned short vi_item_len; /* length of item that it will have after balancing */ 2008 struct item_head *vi_ih; 2009 const char *vi_item; // body of item (old or new) 2010 const void *vi_new_data; // 0 always but paste mode 2011 void *vi_uarea; // item specific area 2012 }; 2013 2014 struct virtual_node { 2015 char *vn_free_ptr; /* this is a pointer to the free space in the buffer */ 2016 unsigned short vn_nr_item; /* number of items in virtual node */ 2017 short vn_size; /* size of node , that node would have if it has unlimited size and no balancing is performed */ 2018 short vn_mode; /* mode of balancing (paste, insert, delete, cut) */ 2019 short vn_affected_item_num; 2020 short vn_pos_in_item; 2021 struct item_head *vn_ins_ih; /* item header of inserted item, 0 for other modes */ 2022 const void *vn_data; 2023 struct virtual_item *vn_vi; /* array of items (including a new one, excluding item to be deleted) */ 2024 }; 2025 2026 /* used by directory items when creating virtual nodes */ 2027 struct direntry_uarea { 2028 int flags; 2029 __u16 entry_count; 2030 __u16 entry_sizes[1]; 2031 } __attribute__ ((__packed__)); 2032 2033 /***************************************************************************/ 2034 /* TREE BALANCE */ 2035 /***************************************************************************/ 2036 2037 /* This temporary structure is used in tree balance algorithms, and 2038 constructed as we go to the extent that its various parts are 2039 needed. It contains arrays of nodes that can potentially be 2040 involved in the balancing of node S, and parameters that define how 2041 each of the nodes must be balanced. Note that in these algorithms 2042 for balancing the worst case is to need to balance the current node 2043 S and the left and right neighbors and all of their parents plus 2044 create a new node. We implement S1 balancing for the leaf nodes 2045 and S0 balancing for the internal nodes (S1 and S0 are defined in 2046 our papers.)*/ 2047 2048 #define MAX_FREE_BLOCK 7 /* size of the array of buffers to free at end of do_balance */ 2049 2050 /* maximum number of FEB blocknrs on a single level */ 2051 #define MAX_AMOUNT_NEEDED 2 2052 2053 /* someday somebody will prefix every field in this struct with tb_ */ 2054 struct tree_balance { 2055 int tb_mode; 2056 int need_balance_dirty; 2057 struct super_block *tb_sb; 2058 struct reiserfs_transaction_handle *transaction_handle; 2059 struct treepath *tb_path; 2060 struct buffer_head *L[MAX_HEIGHT]; /* array of left neighbors of nodes in the path */ 2061 struct buffer_head *R[MAX_HEIGHT]; /* array of right neighbors of nodes in the path */ 2062 struct buffer_head *FL[MAX_HEIGHT]; /* array of fathers of the left neighbors */ 2063 struct buffer_head *FR[MAX_HEIGHT]; /* array of fathers of the right neighbors */ 2064 struct buffer_head *CFL[MAX_HEIGHT]; /* array of common parents of center node and its left neighbor */ 2065 struct buffer_head *CFR[MAX_HEIGHT]; /* array of common parents of center node and its right neighbor */ 2066 2067 struct buffer_head *FEB[MAX_FEB_SIZE]; /* array of empty buffers. Number of buffers in array equals 2068 cur_blknum. */ 2069 struct buffer_head *used[MAX_FEB_SIZE]; 2070 struct buffer_head *thrown[MAX_FEB_SIZE]; 2071 int lnum[MAX_HEIGHT]; /* array of number of items which must be 2072 shifted to the left in order to balance the 2073 current node; for leaves includes item that 2074 will be partially shifted; for internal 2075 nodes, it is the number of child pointers 2076 rather than items. It includes the new item 2077 being created. The code sometimes subtracts 2078 one to get the number of wholly shifted 2079 items for other purposes. */ 2080 int rnum[MAX_HEIGHT]; /* substitute right for left in comment above */ 2081 int lkey[MAX_HEIGHT]; /* array indexed by height h mapping the key delimiting L[h] and 2082 S[h] to its item number within the node CFL[h] */ 2083 int rkey[MAX_HEIGHT]; /* substitute r for l in comment above */ 2084 int insert_size[MAX_HEIGHT]; /* the number of bytes by we are trying to add or remove from 2085 S[h]. A negative value means removing. */ 2086 int blknum[MAX_HEIGHT]; /* number of nodes that will replace node S[h] after 2087 balancing on the level h of the tree. If 0 then S is 2088 being deleted, if 1 then S is remaining and no new nodes 2089 are being created, if 2 or 3 then 1 or 2 new nodes is 2090 being created */ 2091 2092 /* fields that are used only for balancing leaves of the tree */ 2093 int cur_blknum; /* number of empty blocks having been already allocated */ 2094 int s0num; /* number of items that fall into left most node when S[0] splits */ 2095 int s1num; /* number of items that fall into first new node when S[0] splits */ 2096 int s2num; /* number of items that fall into second new node when S[0] splits */ 2097 int lbytes; /* number of bytes which can flow to the left neighbor from the left */ 2098 /* most liquid item that cannot be shifted from S[0] entirely */ 2099 /* if -1 then nothing will be partially shifted */ 2100 int rbytes; /* number of bytes which will flow to the right neighbor from the right */ 2101 /* most liquid item that cannot be shifted from S[0] entirely */ 2102 /* if -1 then nothing will be partially shifted */ 2103 int s1bytes; /* number of bytes which flow to the first new node when S[0] splits */ 2104 /* note: if S[0] splits into 3 nodes, then items do not need to be cut */ 2105 int s2bytes; 2106 struct buffer_head *buf_to_free[MAX_FREE_BLOCK]; /* buffers which are to be freed after do_balance finishes by unfix_nodes */ 2107 char *vn_buf; /* kmalloced memory. Used to create 2108 virtual node and keep map of 2109 dirtied bitmap blocks */ 2110 int vn_buf_size; /* size of the vn_buf */ 2111 struct virtual_node *tb_vn; /* VN starts after bitmap of bitmap blocks */ 2112 2113 int fs_gen; /* saved value of `reiserfs_generation' counter 2114 see FILESYSTEM_CHANGED() macro in reiserfs_fs.h */ 2115 #ifdef DISPLACE_NEW_PACKING_LOCALITIES 2116 struct in_core_key key; /* key pointer, to pass to block allocator or 2117 another low-level subsystem */ 2118 #endif 2119 }; 2120 2121 /* These are modes of balancing */ 2122 2123 /* When inserting an item. */ 2124 #define M_INSERT 'i' 2125 /* When inserting into (directories only) or appending onto an already 2126 existent item. */ 2127 #define M_PASTE 'p' 2128 /* When deleting an item. */ 2129 #define M_DELETE 'd' 2130 /* When truncating an item or removing an entry from a (directory) item. */ 2131 #define M_CUT 'c' 2132 2133 /* used when balancing on leaf level skipped (in reiserfsck) */ 2134 #define M_INTERNAL 'n' 2135 2136 /* When further balancing is not needed, then do_balance does not need 2137 to be called. */ 2138 #define M_SKIP_BALANCING 's' 2139 #define M_CONVERT 'v' 2140 2141 /* modes of leaf_move_items */ 2142 #define LEAF_FROM_S_TO_L 0 2143 #define LEAF_FROM_S_TO_R 1 2144 #define LEAF_FROM_R_TO_L 2 2145 #define LEAF_FROM_L_TO_R 3 2146 #define LEAF_FROM_S_TO_SNEW 4 2147 2148 #define FIRST_TO_LAST 0 2149 #define LAST_TO_FIRST 1 2150 2151 /* used in do_balance for passing parent of node information that has 2152 been gotten from tb struct */ 2153 struct buffer_info { 2154 struct tree_balance *tb; 2155 struct buffer_head *bi_bh; 2156 struct buffer_head *bi_parent; 2157 int bi_position; 2158 }; 2159 2160 static inline struct super_block *sb_from_tb(struct tree_balance *tb) 2161 { 2162 return tb ? tb->tb_sb : NULL; 2163 } 2164 2165 static inline struct super_block *sb_from_bi(struct buffer_info *bi) 2166 { 2167 return bi ? sb_from_tb(bi->tb) : NULL; 2168 } 2169 2170 /* there are 4 types of items: stat data, directory item, indirect, direct. 2171 +-------------------+------------+--------------+------------+ 2172 | | k_offset | k_uniqueness | mergeable? | 2173 +-------------------+------------+--------------+------------+ 2174 | stat data | 0 | 0 | no | 2175 +-------------------+------------+--------------+------------+ 2176 | 1st directory item| DOT_OFFSET |DIRENTRY_UNIQUENESS| no | 2177 | non 1st directory | hash value | | yes | 2178 | item | | | | 2179 +-------------------+------------+--------------+------------+ 2180 | indirect item | offset + 1 |TYPE_INDIRECT | if this is not the first indirect item of the object 2181 +-------------------+------------+--------------+------------+ 2182 | direct item | offset + 1 |TYPE_DIRECT | if not this is not the first direct item of the object 2183 +-------------------+------------+--------------+------------+ 2184 */ 2185 2186 struct item_operations { 2187 int (*bytes_number) (struct item_head * ih, int block_size); 2188 void (*decrement_key) (struct cpu_key *); 2189 int (*is_left_mergeable) (struct reiserfs_key * ih, 2190 unsigned long bsize); 2191 void (*print_item) (struct item_head *, char *item); 2192 void (*check_item) (struct item_head *, char *item); 2193 2194 int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi, 2195 int is_affected, int insert_size); 2196 int (*check_left) (struct virtual_item * vi, int free, 2197 int start_skip, int end_skip); 2198 int (*check_right) (struct virtual_item * vi, int free); 2199 int (*part_size) (struct virtual_item * vi, int from, int to); 2200 int (*unit_num) (struct virtual_item * vi); 2201 void (*print_vi) (struct virtual_item * vi); 2202 }; 2203 2204 extern struct item_operations *item_ops[TYPE_ANY + 1]; 2205 2206 #define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize) 2207 #define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize) 2208 #define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item) 2209 #define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item) 2210 #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) 2211 #define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip) 2212 #define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free) 2213 #define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to) 2214 #define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi) 2215 #define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi) 2216 2217 #define COMP_SHORT_KEYS comp_short_keys 2218 2219 /* number of blocks pointed to by the indirect item */ 2220 #define I_UNFM_NUM(ih) (ih_item_len(ih) / UNFM_P_SIZE) 2221 2222 /* the used space within the unformatted node corresponding to pos within the item pointed to by ih */ 2223 #define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size)) 2224 2225 /* number of bytes contained by the direct item or the unformatted nodes the indirect item points to */ 2226 2227 /* get the item header */ 2228 #define B_N_PITEM_HEAD(bh,item_num) ( (struct item_head * )((bh)->b_data + BLKH_SIZE) + (item_num) ) 2229 2230 /* get key */ 2231 #define B_N_PDELIM_KEY(bh,item_num) ( (struct reiserfs_key * )((bh)->b_data + BLKH_SIZE) + (item_num) ) 2232 2233 /* get the key */ 2234 #define B_N_PKEY(bh,item_num) ( &(B_N_PITEM_HEAD(bh,item_num)->ih_key) ) 2235 2236 /* get item body */ 2237 #define B_N_PITEM(bh,item_num) ( (bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(item_num)))) 2238 2239 /* get the stat data by the buffer header and the item order */ 2240 #define B_N_STAT_DATA(bh,nr) \ 2241 ( (struct stat_data *)((bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(nr))) ) ) 2242 2243 /* following defines use reiserfs buffer header and item header */ 2244 2245 /* get stat-data */ 2246 #define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) ) 2247 2248 // this is 3976 for size==4096 2249 #define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE) 2250 2251 /* indirect items consist of entries which contain blocknrs, pos 2252 indicates which entry, and B_I_POS_UNFM_POINTER resolves to the 2253 blocknr contained by the entry pos points to */ 2254 #define B_I_POS_UNFM_POINTER(bh,ih,pos) le32_to_cpu(*(((unp_t *)B_I_PITEM(bh,ih)) + (pos))) 2255 #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) 2256 2257 struct reiserfs_iget_args { 2258 __u32 objectid; 2259 __u32 dirid; 2260 }; 2261 2262 /***************************************************************************/ 2263 /* FUNCTION DECLARATIONS */ 2264 /***************************************************************************/ 2265 2266 #define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12) 2267 2268 #define journal_trans_half(blocksize) \ 2269 ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32)) 2270 2271 /* journal.c see journal.c for all the comments here */ 2272 2273 /* first block written in a commit. */ 2274 struct reiserfs_journal_desc { 2275 __le32 j_trans_id; /* id of commit */ 2276 __le32 j_len; /* length of commit. len +1 is the commit block */ 2277 __le32 j_mount_id; /* mount id of this trans */ 2278 __le32 j_realblock[1]; /* real locations for each block */ 2279 }; 2280 2281 #define get_desc_trans_id(d) le32_to_cpu((d)->j_trans_id) 2282 #define get_desc_trans_len(d) le32_to_cpu((d)->j_len) 2283 #define get_desc_mount_id(d) le32_to_cpu((d)->j_mount_id) 2284 2285 #define set_desc_trans_id(d,val) do { (d)->j_trans_id = cpu_to_le32 (val); } while (0) 2286 #define set_desc_trans_len(d,val) do { (d)->j_len = cpu_to_le32 (val); } while (0) 2287 #define set_desc_mount_id(d,val) do { (d)->j_mount_id = cpu_to_le32 (val); } while (0) 2288 2289 /* last block written in a commit */ 2290 struct reiserfs_journal_commit { 2291 __le32 j_trans_id; /* must match j_trans_id from the desc block */ 2292 __le32 j_len; /* ditto */ 2293 __le32 j_realblock[1]; /* real locations for each block */ 2294 }; 2295 2296 #define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id) 2297 #define get_commit_trans_len(c) le32_to_cpu((c)->j_len) 2298 #define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id) 2299 2300 #define set_commit_trans_id(c,val) do { (c)->j_trans_id = cpu_to_le32 (val); } while (0) 2301 #define set_commit_trans_len(c,val) do { (c)->j_len = cpu_to_le32 (val); } while (0) 2302 2303 /* this header block gets written whenever a transaction is considered fully flushed, and is more recent than the 2304 ** last fully flushed transaction. fully flushed means all the log blocks and all the real blocks are on disk, 2305 ** and this transaction does not need to be replayed. 2306 */ 2307 struct reiserfs_journal_header { 2308 __le32 j_last_flush_trans_id; /* id of last fully flushed transaction */ 2309 __le32 j_first_unflushed_offset; /* offset in the log of where to start replay after a crash */ 2310 __le32 j_mount_id; 2311 /* 12 */ struct journal_params jh_journal; 2312 }; 2313 2314 /* biggest tunable defines are right here */ 2315 #define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */ 2316 #define JOURNAL_TRANS_MAX_DEFAULT 1024 /* biggest possible single transaction, don't change for now (8/3/99) */ 2317 #define JOURNAL_TRANS_MIN_DEFAULT 256 2318 #define JOURNAL_MAX_BATCH_DEFAULT 900 /* max blocks to batch into one transaction, don't make this any bigger than 900 */ 2319 #define JOURNAL_MIN_RATIO 2 2320 #define JOURNAL_MAX_COMMIT_AGE 30 2321 #define JOURNAL_MAX_TRANS_AGE 30 2322 #define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9) 2323 #define JOURNAL_BLOCKS_PER_OBJECT(sb) (JOURNAL_PER_BALANCE_CNT * 3 + \ 2324 2 * (REISERFS_QUOTA_INIT_BLOCKS(sb) + \ 2325 REISERFS_QUOTA_TRANS_BLOCKS(sb))) 2326 2327 #ifdef CONFIG_QUOTA 2328 #define REISERFS_QUOTA_OPTS ((1 << REISERFS_USRQUOTA) | (1 << REISERFS_GRPQUOTA)) 2329 /* We need to update data and inode (atime) */ 2330 #define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? 2 : 0) 2331 /* 1 balancing, 1 bitmap, 1 data per write + stat data update */ 2332 #define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \ 2333 (DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0) 2334 /* same as with INIT */ 2335 #define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \ 2336 (DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0) 2337 #else 2338 #define REISERFS_QUOTA_TRANS_BLOCKS(s) 0 2339 #define REISERFS_QUOTA_INIT_BLOCKS(s) 0 2340 #define REISERFS_QUOTA_DEL_BLOCKS(s) 0 2341 #endif 2342 2343 /* both of these can be as low as 1, or as high as you want. The min is the 2344 ** number of 4k bitmap nodes preallocated on mount. New nodes are allocated 2345 ** as needed, and released when transactions are committed. On release, if 2346 ** the current number of nodes is > max, the node is freed, otherwise, 2347 ** it is put on a free list for faster use later. 2348 */ 2349 #define REISERFS_MIN_BITMAP_NODES 10 2350 #define REISERFS_MAX_BITMAP_NODES 100 2351 2352 #define JBH_HASH_SHIFT 13 /* these are based on journal hash size of 8192 */ 2353 #define JBH_HASH_MASK 8191 2354 2355 #define _jhashfn(sb,block) \ 2356 (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \ 2357 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12)))) 2358 #define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK]) 2359 2360 // We need these to make journal.c code more readable 2361 #define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize) 2362 #define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize) 2363 #define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize) 2364 2365 enum reiserfs_bh_state_bits { 2366 BH_JDirty = BH_PrivateStart, /* buffer is in current transaction */ 2367 BH_JDirty_wait, 2368 BH_JNew, /* disk block was taken off free list before 2369 * being in a finished transaction, or 2370 * written to disk. Can be reused immed. */ 2371 BH_JPrepared, 2372 BH_JRestore_dirty, 2373 BH_JTest, // debugging only will go away 2374 }; 2375 2376 BUFFER_FNS(JDirty, journaled); 2377 TAS_BUFFER_FNS(JDirty, journaled); 2378 BUFFER_FNS(JDirty_wait, journal_dirty); 2379 TAS_BUFFER_FNS(JDirty_wait, journal_dirty); 2380 BUFFER_FNS(JNew, journal_new); 2381 TAS_BUFFER_FNS(JNew, journal_new); 2382 BUFFER_FNS(JPrepared, journal_prepared); 2383 TAS_BUFFER_FNS(JPrepared, journal_prepared); 2384 BUFFER_FNS(JRestore_dirty, journal_restore_dirty); 2385 TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty); 2386 BUFFER_FNS(JTest, journal_test); 2387 TAS_BUFFER_FNS(JTest, journal_test); 2388 2389 /* 2390 ** transaction handle which is passed around for all journal calls 2391 */ 2392 struct reiserfs_transaction_handle { 2393 struct super_block *t_super; /* super for this FS when journal_begin was 2394 called. saves calls to reiserfs_get_super 2395 also used by nested transactions to make 2396 sure they are nesting on the right FS 2397 _must_ be first in the handle 2398 */ 2399 int t_refcount; 2400 int t_blocks_logged; /* number of blocks this writer has logged */ 2401 int t_blocks_allocated; /* number of blocks this writer allocated */ 2402 unsigned int t_trans_id; /* sanity check, equals the current trans id */ 2403 void *t_handle_save; /* save existing current->journal_info */ 2404 unsigned displace_new_blocks:1; /* if new block allocation occurres, that block 2405 should be displaced from others */ 2406 struct list_head t_list; 2407 }; 2408 2409 /* used to keep track of ordered and tail writes, attached to the buffer 2410 * head through b_journal_head. 2411 */ 2412 struct reiserfs_jh { 2413 struct reiserfs_journal_list *jl; 2414 struct buffer_head *bh; 2415 struct list_head list; 2416 }; 2417 2418 void reiserfs_free_jh(struct buffer_head *bh); 2419 int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh); 2420 int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh); 2421 int journal_mark_dirty(struct reiserfs_transaction_handle *, 2422 struct super_block *, struct buffer_head *bh); 2423 2424 static inline int reiserfs_file_data_log(struct inode *inode) 2425 { 2426 if (reiserfs_data_log(inode->i_sb) || 2427 (REISERFS_I(inode)->i_flags & i_data_log)) 2428 return 1; 2429 return 0; 2430 } 2431 2432 static inline int reiserfs_transaction_running(struct super_block *s) 2433 { 2434 struct reiserfs_transaction_handle *th = current->journal_info; 2435 if (th && th->t_super == s) 2436 return 1; 2437 if (th && th->t_super == NULL) 2438 BUG(); 2439 return 0; 2440 } 2441 2442 static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th) 2443 { 2444 return th->t_blocks_allocated - th->t_blocks_logged; 2445 } 2446 2447 struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct 2448 super_block 2449 *, 2450 int count); 2451 int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *); 2452 void reiserfs_vfs_truncate_file(struct inode *inode); 2453 int reiserfs_commit_page(struct inode *inode, struct page *page, 2454 unsigned from, unsigned to); 2455 void reiserfs_flush_old_commits(struct super_block *); 2456 int reiserfs_commit_for_inode(struct inode *); 2457 int reiserfs_inode_needs_commit(struct inode *); 2458 void reiserfs_update_inode_transaction(struct inode *); 2459 void reiserfs_wait_on_write_block(struct super_block *s); 2460 void reiserfs_block_writes(struct reiserfs_transaction_handle *th); 2461 void reiserfs_allow_writes(struct super_block *s); 2462 void reiserfs_check_lock_depth(struct super_block *s, char *caller); 2463 int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh, 2464 int wait); 2465 void reiserfs_restore_prepared_buffer(struct super_block *, 2466 struct buffer_head *bh); 2467 int journal_init(struct super_block *, const char *j_dev_name, int old_format, 2468 unsigned int); 2469 int journal_release(struct reiserfs_transaction_handle *, struct super_block *); 2470 int journal_release_error(struct reiserfs_transaction_handle *, 2471 struct super_block *); 2472 int journal_end(struct reiserfs_transaction_handle *, struct super_block *, 2473 unsigned long); 2474 int journal_end_sync(struct reiserfs_transaction_handle *, struct super_block *, 2475 unsigned long); 2476 int journal_mark_freed(struct reiserfs_transaction_handle *, 2477 struct super_block *, b_blocknr_t blocknr); 2478 int journal_transaction_should_end(struct reiserfs_transaction_handle *, int); 2479 int reiserfs_in_journal(struct super_block *sb, unsigned int bmap_nr, 2480 int bit_nr, int searchall, b_blocknr_t *next); 2481 int journal_begin(struct reiserfs_transaction_handle *, 2482 struct super_block *sb, unsigned long); 2483 int journal_join_abort(struct reiserfs_transaction_handle *, 2484 struct super_block *sb, unsigned long); 2485 void reiserfs_abort_journal(struct super_block *sb, int errno); 2486 void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...); 2487 int reiserfs_allocate_list_bitmaps(struct super_block *s, 2488 struct reiserfs_list_bitmap *, unsigned int); 2489 2490 void reiserfs_schedule_old_flush(struct super_block *s); 2491 void add_save_link(struct reiserfs_transaction_handle *th, 2492 struct inode *inode, int truncate); 2493 int remove_save_link(struct inode *inode, int truncate); 2494 2495 /* objectid.c */ 2496 __u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th); 2497 void reiserfs_release_objectid(struct reiserfs_transaction_handle *th, 2498 __u32 objectid_to_release); 2499 int reiserfs_convert_objectid_map_v1(struct super_block *); 2500 2501 /* stree.c */ 2502 int B_IS_IN_TREE(const struct buffer_head *); 2503 extern void copy_item_head(struct item_head *to, 2504 const struct item_head *from); 2505 2506 // first key is in cpu form, second - le 2507 extern int comp_short_keys(const struct reiserfs_key *le_key, 2508 const struct cpu_key *cpu_key); 2509 extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from); 2510 2511 // both are in le form 2512 extern int comp_le_keys(const struct reiserfs_key *, 2513 const struct reiserfs_key *); 2514 extern int comp_short_le_keys(const struct reiserfs_key *, 2515 const struct reiserfs_key *); 2516 2517 // 2518 // get key version from on disk key - kludge 2519 // 2520 static inline int le_key_version(const struct reiserfs_key *key) 2521 { 2522 int type; 2523 2524 type = offset_v2_k_type(&(key->u.k_offset_v2)); 2525 if (type != TYPE_DIRECT && type != TYPE_INDIRECT 2526 && type != TYPE_DIRENTRY) 2527 return KEY_FORMAT_3_5; 2528 2529 return KEY_FORMAT_3_6; 2530 2531 } 2532 2533 static inline void copy_key(struct reiserfs_key *to, 2534 const struct reiserfs_key *from) 2535 { 2536 memcpy(to, from, KEY_SIZE); 2537 } 2538 2539 int comp_items(const struct item_head *stored_ih, const struct treepath *path); 2540 const struct reiserfs_key *get_rkey(const struct treepath *chk_path, 2541 const struct super_block *sb); 2542 int search_by_key(struct super_block *, const struct cpu_key *, 2543 struct treepath *, int); 2544 #define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL) 2545 int search_for_position_by_key(struct super_block *sb, 2546 const struct cpu_key *cpu_key, 2547 struct treepath *search_path); 2548 extern void decrement_bcount(struct buffer_head *bh); 2549 void decrement_counters_in_path(struct treepath *search_path); 2550 void pathrelse(struct treepath *search_path); 2551 int reiserfs_check_path(struct treepath *p); 2552 void pathrelse_and_restore(struct super_block *s, struct treepath *search_path); 2553 2554 int reiserfs_insert_item(struct reiserfs_transaction_handle *th, 2555 struct treepath *path, 2556 const struct cpu_key *key, 2557 struct item_head *ih, 2558 struct inode *inode, const char *body); 2559 2560 int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th, 2561 struct treepath *path, 2562 const struct cpu_key *key, 2563 struct inode *inode, 2564 const char *body, int paste_size); 2565 2566 int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th, 2567 struct treepath *path, 2568 struct cpu_key *key, 2569 struct inode *inode, 2570 struct page *page, loff_t new_file_size); 2571 2572 int reiserfs_delete_item(struct reiserfs_transaction_handle *th, 2573 struct treepath *path, 2574 const struct cpu_key *key, 2575 struct inode *inode, struct buffer_head *un_bh); 2576 2577 void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th, 2578 struct inode *inode, struct reiserfs_key *key); 2579 int reiserfs_delete_object(struct reiserfs_transaction_handle *th, 2580 struct inode *inode); 2581 int reiserfs_do_truncate(struct reiserfs_transaction_handle *th, 2582 struct inode *inode, struct page *, 2583 int update_timestamps); 2584 2585 #define i_block_size(inode) ((inode)->i_sb->s_blocksize) 2586 #define file_size(inode) ((inode)->i_size) 2587 #define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1)) 2588 2589 #define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\ 2590 !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 ) 2591 2592 void padd_item(char *item, int total_length, int length); 2593 2594 /* inode.c */ 2595 /* args for the create parameter of reiserfs_get_block */ 2596 #define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */ 2597 #define GET_BLOCK_CREATE 1 /* add anything you need to find block */ 2598 #define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */ 2599 #define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */ 2600 #define GET_BLOCK_NO_IMUX 8 /* i_mutex is not held, don't preallocate */ 2601 #define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */ 2602 2603 void reiserfs_read_locked_inode(struct inode *inode, 2604 struct reiserfs_iget_args *args); 2605 int reiserfs_find_actor(struct inode *inode, void *p); 2606 int reiserfs_init_locked_inode(struct inode *inode, void *p); 2607 void reiserfs_evict_inode(struct inode *inode); 2608 int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc); 2609 int reiserfs_get_block(struct inode *inode, sector_t block, 2610 struct buffer_head *bh_result, int create); 2611 struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid, 2612 int fh_len, int fh_type); 2613 struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid, 2614 int fh_len, int fh_type); 2615 int reiserfs_encode_fh(struct inode *inode, __u32 * data, int *lenp, 2616 struct inode *parent); 2617 2618 int reiserfs_truncate_file(struct inode *, int update_timestamps); 2619 void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset, 2620 int type, int key_length); 2621 void make_le_item_head(struct item_head *ih, const struct cpu_key *key, 2622 int version, 2623 loff_t offset, int type, int length, int entry_count); 2624 struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key); 2625 2626 struct reiserfs_security_handle; 2627 int reiserfs_new_inode(struct reiserfs_transaction_handle *th, 2628 struct inode *dir, umode_t mode, 2629 const char *symname, loff_t i_size, 2630 struct dentry *dentry, struct inode *inode, 2631 struct reiserfs_security_handle *security); 2632 2633 void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th, 2634 struct inode *inode, loff_t size); 2635 2636 static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th, 2637 struct inode *inode) 2638 { 2639 reiserfs_update_sd_size(th, inode, inode->i_size); 2640 } 2641 2642 void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode); 2643 void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs); 2644 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr); 2645 2646 int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len); 2647 2648 /* namei.c */ 2649 void set_de_name_and_namelen(struct reiserfs_dir_entry *de); 2650 int search_by_entry_key(struct super_block *sb, const struct cpu_key *key, 2651 struct treepath *path, struct reiserfs_dir_entry *de); 2652 struct dentry *reiserfs_get_parent(struct dentry *); 2653 2654 #ifdef CONFIG_REISERFS_PROC_INFO 2655 int reiserfs_proc_info_init(struct super_block *sb); 2656 int reiserfs_proc_info_done(struct super_block *sb); 2657 int reiserfs_proc_info_global_init(void); 2658 int reiserfs_proc_info_global_done(void); 2659 2660 #define PROC_EXP( e ) e 2661 2662 #define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data 2663 #define PROC_INFO_MAX( sb, field, value ) \ 2664 __PINFO( sb ).field = \ 2665 max( REISERFS_SB( sb ) -> s_proc_info_data.field, value ) 2666 #define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) ) 2667 #define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) ) 2668 #define PROC_INFO_BH_STAT( sb, bh, level ) \ 2669 PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \ 2670 PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \ 2671 PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) ) 2672 #else 2673 static inline int reiserfs_proc_info_init(struct super_block *sb) 2674 { 2675 return 0; 2676 } 2677 2678 static inline int reiserfs_proc_info_done(struct super_block *sb) 2679 { 2680 return 0; 2681 } 2682 2683 static inline int reiserfs_proc_info_global_init(void) 2684 { 2685 return 0; 2686 } 2687 2688 static inline int reiserfs_proc_info_global_done(void) 2689 { 2690 return 0; 2691 } 2692 2693 #define PROC_EXP( e ) 2694 #define VOID_V ( ( void ) 0 ) 2695 #define PROC_INFO_MAX( sb, field, value ) VOID_V 2696 #define PROC_INFO_INC( sb, field ) VOID_V 2697 #define PROC_INFO_ADD( sb, field, val ) VOID_V 2698 #define PROC_INFO_BH_STAT(sb, bh, n_node_level) VOID_V 2699 #endif 2700 2701 /* dir.c */ 2702 extern const struct inode_operations reiserfs_dir_inode_operations; 2703 extern const struct inode_operations reiserfs_symlink_inode_operations; 2704 extern const struct inode_operations reiserfs_special_inode_operations; 2705 extern const struct file_operations reiserfs_dir_operations; 2706 int reiserfs_readdir_inode(struct inode *, struct dir_context *); 2707 2708 /* tail_conversion.c */ 2709 int direct2indirect(struct reiserfs_transaction_handle *, struct inode *, 2710 struct treepath *, struct buffer_head *, loff_t); 2711 int indirect2direct(struct reiserfs_transaction_handle *, struct inode *, 2712 struct page *, struct treepath *, const struct cpu_key *, 2713 loff_t, char *); 2714 void reiserfs_unmap_buffer(struct buffer_head *); 2715 2716 /* file.c */ 2717 extern const struct inode_operations reiserfs_file_inode_operations; 2718 extern const struct file_operations reiserfs_file_operations; 2719 extern const struct address_space_operations reiserfs_address_space_operations; 2720 2721 /* fix_nodes.c */ 2722 2723 int fix_nodes(int n_op_mode, struct tree_balance *tb, 2724 struct item_head *ins_ih, const void *); 2725 void unfix_nodes(struct tree_balance *); 2726 2727 /* prints.c */ 2728 void __reiserfs_panic(struct super_block *s, const char *id, 2729 const char *function, const char *fmt, ...) 2730 __attribute__ ((noreturn)); 2731 #define reiserfs_panic(s, id, fmt, args...) \ 2732 __reiserfs_panic(s, id, __func__, fmt, ##args) 2733 void __reiserfs_error(struct super_block *s, const char *id, 2734 const char *function, const char *fmt, ...); 2735 #define reiserfs_error(s, id, fmt, args...) \ 2736 __reiserfs_error(s, id, __func__, fmt, ##args) 2737 void reiserfs_info(struct super_block *s, const char *fmt, ...); 2738 void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...); 2739 void print_indirect_item(struct buffer_head *bh, int item_num); 2740 void store_print_tb(struct tree_balance *tb); 2741 void print_cur_tb(char *mes); 2742 void print_de(struct reiserfs_dir_entry *de); 2743 void print_bi(struct buffer_info *bi, char *mes); 2744 #define PRINT_LEAF_ITEMS 1 /* print all items */ 2745 #define PRINT_DIRECTORY_ITEMS 2 /* print directory items */ 2746 #define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */ 2747 void print_block(struct buffer_head *bh, ...); 2748 void print_bmap(struct super_block *s, int silent); 2749 void print_bmap_block(int i, char *data, int size, int silent); 2750 /*void print_super_block (struct super_block * s, char * mes);*/ 2751 void print_objectid_map(struct super_block *s); 2752 void print_block_head(struct buffer_head *bh, char *mes); 2753 void check_leaf(struct buffer_head *bh); 2754 void check_internal(struct buffer_head *bh); 2755 void print_statistics(struct super_block *s); 2756 char *reiserfs_hashname(int code); 2757 2758 /* lbalance.c */ 2759 int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num, 2760 int mov_bytes, struct buffer_head *Snew); 2761 int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes); 2762 int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes); 2763 void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first, 2764 int del_num, int del_bytes); 2765 void leaf_insert_into_buf(struct buffer_info *bi, int before, 2766 struct item_head *inserted_item_ih, 2767 const char *inserted_item_body, int zeros_number); 2768 void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num, 2769 int pos_in_item, int paste_size, const char *body, 2770 int zeros_number); 2771 void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num, 2772 int pos_in_item, int cut_size); 2773 void leaf_paste_entries(struct buffer_info *bi, int item_num, int before, 2774 int new_entry_count, struct reiserfs_de_head *new_dehs, 2775 const char *records, int paste_size); 2776 /* ibalance.c */ 2777 int balance_internal(struct tree_balance *, int, int, struct item_head *, 2778 struct buffer_head **); 2779 2780 /* do_balance.c */ 2781 void do_balance_mark_leaf_dirty(struct tree_balance *tb, 2782 struct buffer_head *bh, int flag); 2783 #define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty 2784 #define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty 2785 2786 void do_balance(struct tree_balance *tb, struct item_head *ih, 2787 const char *body, int flag); 2788 void reiserfs_invalidate_buffer(struct tree_balance *tb, 2789 struct buffer_head *bh); 2790 2791 int get_left_neighbor_position(struct tree_balance *tb, int h); 2792 int get_right_neighbor_position(struct tree_balance *tb, int h); 2793 void replace_key(struct tree_balance *tb, struct buffer_head *, int, 2794 struct buffer_head *, int); 2795 void make_empty_node(struct buffer_info *); 2796 struct buffer_head *get_FEB(struct tree_balance *); 2797 2798 /* bitmap.c */ 2799 2800 /* structure contains hints for block allocator, and it is a container for 2801 * arguments, such as node, search path, transaction_handle, etc. */ 2802 struct __reiserfs_blocknr_hint { 2803 struct inode *inode; /* inode passed to allocator, if we allocate unf. nodes */ 2804 sector_t block; /* file offset, in blocks */ 2805 struct in_core_key key; 2806 struct treepath *path; /* search path, used by allocator to deternine search_start by 2807 * various ways */ 2808 struct reiserfs_transaction_handle *th; /* transaction handle is needed to log super blocks and 2809 * bitmap blocks changes */ 2810 b_blocknr_t beg, end; 2811 b_blocknr_t search_start; /* a field used to transfer search start value (block number) 2812 * between different block allocator procedures 2813 * (determine_search_start() and others) */ 2814 int prealloc_size; /* is set in determine_prealloc_size() function, used by underlayed 2815 * function that do actual allocation */ 2816 2817 unsigned formatted_node:1; /* the allocator uses different polices for getting disk space for 2818 * formatted/unformatted blocks with/without preallocation */ 2819 unsigned preallocate:1; 2820 }; 2821 2822 typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t; 2823 2824 int reiserfs_parse_alloc_options(struct super_block *, char *); 2825 void reiserfs_init_alloc_options(struct super_block *s); 2826 2827 /* 2828 * given a directory, this will tell you what packing locality 2829 * to use for a new object underneat it. The locality is returned 2830 * in disk byte order (le). 2831 */ 2832 __le32 reiserfs_choose_packing(struct inode *dir); 2833 2834 int reiserfs_init_bitmap_cache(struct super_block *sb); 2835 void reiserfs_free_bitmap_cache(struct super_block *sb); 2836 void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info); 2837 struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap); 2838 int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value); 2839 void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *, 2840 b_blocknr_t, int for_unformatted); 2841 int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int, 2842 int); 2843 static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb, 2844 b_blocknr_t * new_blocknrs, 2845 int amount_needed) 2846 { 2847 reiserfs_blocknr_hint_t hint = { 2848 .th = tb->transaction_handle, 2849 .path = tb->tb_path, 2850 .inode = NULL, 2851 .key = tb->key, 2852 .block = 0, 2853 .formatted_node = 1 2854 }; 2855 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed, 2856 0); 2857 } 2858 2859 static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle 2860 *th, struct inode *inode, 2861 b_blocknr_t * new_blocknrs, 2862 struct treepath *path, 2863 sector_t block) 2864 { 2865 reiserfs_blocknr_hint_t hint = { 2866 .th = th, 2867 .path = path, 2868 .inode = inode, 2869 .block = block, 2870 .formatted_node = 0, 2871 .preallocate = 0 2872 }; 2873 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0); 2874 } 2875 2876 #ifdef REISERFS_PREALLOCATE 2877 static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle 2878 *th, struct inode *inode, 2879 b_blocknr_t * new_blocknrs, 2880 struct treepath *path, 2881 sector_t block) 2882 { 2883 reiserfs_blocknr_hint_t hint = { 2884 .th = th, 2885 .path = path, 2886 .inode = inode, 2887 .block = block, 2888 .formatted_node = 0, 2889 .preallocate = 1 2890 }; 2891 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0); 2892 } 2893 2894 void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th, 2895 struct inode *inode); 2896 void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th); 2897 #endif 2898 2899 /* hashes.c */ 2900 __u32 keyed_hash(const signed char *msg, int len); 2901 __u32 yura_hash(const signed char *msg, int len); 2902 __u32 r5_hash(const signed char *msg, int len); 2903 2904 #define reiserfs_set_le_bit __set_bit_le 2905 #define reiserfs_test_and_set_le_bit __test_and_set_bit_le 2906 #define reiserfs_clear_le_bit __clear_bit_le 2907 #define reiserfs_test_and_clear_le_bit __test_and_clear_bit_le 2908 #define reiserfs_test_le_bit test_bit_le 2909 #define reiserfs_find_next_zero_le_bit find_next_zero_bit_le 2910 2911 /* sometimes reiserfs_truncate may require to allocate few new blocks 2912 to perform indirect2direct conversion. People probably used to 2913 think, that truncate should work without problems on a filesystem 2914 without free disk space. They may complain that they can not 2915 truncate due to lack of free disk space. This spare space allows us 2916 to not worry about it. 500 is probably too much, but it should be 2917 absolutely safe */ 2918 #define SPARE_SPACE 500 2919 2920 /* prototypes from ioctl.c */ 2921 long reiserfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg); 2922 long reiserfs_compat_ioctl(struct file *filp, 2923 unsigned int cmd, unsigned long arg); 2924 int reiserfs_unpack(struct inode *inode, struct file *filp); 2925