blob: 157e474ab30347cde97360a4c94de73b5aff2164 [file] [log] [blame]
Al Virof466c6f2012-03-17 01:16:43 -04001/*
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>
Linus Torvaldsed2d2652012-03-24 10:08:39 -070010#include <linux/bug.h>
Al Virof466c6f2012-03-17 01:16:43 -040011#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>
Al Virof466c6f2012-03-17 01:16:43 -040016
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
Al Viro765fd6b2012-03-17 01:19:24 -040024struct reiserfs_journal_list;
25
26/** bitmasks for i_flags field in reiserfs-specific part of inode */
27typedef 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
48struct 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
81typedef 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*/
191struct 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
203struct reiserfs_bitmap_node {
204 int id;
205 char *data;
206 struct list_head list;
207};
208
209struct 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 */
220struct 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
247struct 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
329enum 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
337typedef __u32(*hashf_t) (const signed char *, int);
338
339struct reiserfs_bitmap_info {
340 __u32 free_count;
341};
342
343struct proc_dir_entry;
344
345#if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )
346typedef unsigned long int stat_cnt_t;
347typedef 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
413typedef struct reiserfs_proc_info_data {
414} reiserfs_proc_info_data_t;
415#endif
416
417/* reiserfs union of in-core super block data */
418struct 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;
Artem Bityutskiy033369d2012-06-01 17:18:08 +0300483
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
Al Viro765fd6b2012-03-17 01:19:24 -0400488#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
511enum 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
596void reiserfs_file_buffer(struct buffer_head *bh, int list);
597extern struct file_system_type reiserfs_fs_type;
598int 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/* A safe version of the "bdevname", which returns the "s_id" field of
612 * a superblock or else "Null superblock" if the super block is NULL.
613 */
614static inline char *reiserfs_bdevname(struct super_block *s)
615{
616 return (s == NULL) ? "Null superblock" : s->s_id;
617}
618
619#define reiserfs_is_journal_aborted(journal) (unlikely (__reiserfs_is_journal_aborted (journal)))
620static inline int __reiserfs_is_journal_aborted(struct reiserfs_journal
621 *journal)
622{
623 return test_bit(J_ABORTED, &journal->j_state);
624}
625
Al Virof466c6f2012-03-17 01:16:43 -0400626/*
627 * Locking primitives. The write lock is a per superblock
628 * special mutex that has properties close to the Big Kernel Lock
629 * which was used in the previous locking scheme.
630 */
631void reiserfs_write_lock(struct super_block *s);
632void reiserfs_write_unlock(struct super_block *s);
633int reiserfs_write_lock_once(struct super_block *s);
634void reiserfs_write_unlock_once(struct super_block *s, int lock_depth);
635
636#ifdef CONFIG_REISERFS_CHECK
637void reiserfs_lock_check_recursive(struct super_block *s);
638#else
639static inline void reiserfs_lock_check_recursive(struct super_block *s) { }
640#endif
641
642/*
643 * Several mutexes depend on the write lock.
644 * However sometimes we want to relax the write lock while we hold
645 * these mutexes, according to the release/reacquire on schedule()
646 * properties of the Bkl that were used.
647 * Reiserfs performances and locking were based on this scheme.
648 * Now that the write lock is a mutex and not the bkl anymore, doing so
649 * may result in a deadlock:
650 *
651 * A acquire write_lock
652 * A acquire j_commit_mutex
653 * A release write_lock and wait for something
654 * B acquire write_lock
655 * B can't acquire j_commit_mutex and sleep
656 * A can't acquire write lock anymore
657 * deadlock
658 *
659 * What we do here is avoiding such deadlock by playing the same game
660 * than the Bkl: if we can't acquire a mutex that depends on the write lock,
661 * we release the write lock, wait a bit and then retry.
662 *
663 * The mutexes concerned by this hack are:
664 * - The commit mutex of a journal list
665 * - The flush mutex
666 * - The journal lock
667 * - The inode mutex
668 */
669static inline void reiserfs_mutex_lock_safe(struct mutex *m,
670 struct super_block *s)
671{
672 reiserfs_lock_check_recursive(s);
673 reiserfs_write_unlock(s);
674 mutex_lock(m);
675 reiserfs_write_lock(s);
676}
677
678static inline void
679reiserfs_mutex_lock_nested_safe(struct mutex *m, unsigned int subclass,
680 struct super_block *s)
681{
682 reiserfs_lock_check_recursive(s);
683 reiserfs_write_unlock(s);
684 mutex_lock_nested(m, subclass);
685 reiserfs_write_lock(s);
686}
687
688static inline void
689reiserfs_down_read_safe(struct rw_semaphore *sem, struct super_block *s)
690{
691 reiserfs_lock_check_recursive(s);
692 reiserfs_write_unlock(s);
693 down_read(sem);
694 reiserfs_write_lock(s);
695}
696
697/*
698 * When we schedule, we usually want to also release the write lock,
699 * according to the previous bkl based locking scheme of reiserfs.
700 */
701static inline void reiserfs_cond_resched(struct super_block *s)
702{
703 if (need_resched()) {
704 reiserfs_write_unlock(s);
705 schedule();
706 reiserfs_write_lock(s);
707 }
708}
709
710struct fid;
711
712/* in reading the #defines, it may help to understand that they employ
713 the following abbreviations:
714
715 B = Buffer
716 I = Item header
717 H = Height within the tree (should be changed to LEV)
718 N = Number of the item in the node
719 STAT = stat data
720 DEH = Directory Entry Header
721 EC = Entry Count
722 E = Entry number
723 UL = Unsigned Long
724 BLKH = BLocK Header
725 UNFM = UNForMatted node
726 DC = Disk Child
727 P = Path
728
729 These #defines are named by concatenating these abbreviations,
730 where first comes the arguments, and last comes the return value,
731 of the macro.
732
733*/
734
735#define USE_INODE_GENERATION_COUNTER
736
737#define REISERFS_PREALLOCATE
738#define DISPLACE_NEW_PACKING_LOCALITIES
739#define PREALLOCATION_SIZE 9
740
741/* n must be power of 2 */
742#define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
743
744// to be ok for alpha and others we have to align structures to 8 byte
745// boundary.
746// FIXME: do not change 4 by anything else: there is code which relies on that
747#define ROUND_UP(x) _ROUND_UP(x,8LL)
748
749/* debug levels. Right now, CONFIG_REISERFS_CHECK means print all debug
750** messages.
751*/
752#define REISERFS_DEBUG_CODE 5 /* extra messages to help find/debug errors */
753
754void __reiserfs_warning(struct super_block *s, const char *id,
755 const char *func, const char *fmt, ...);
756#define reiserfs_warning(s, id, fmt, args...) \
757 __reiserfs_warning(s, id, __func__, fmt, ##args)
758/* assertions handling */
759
760/** always check a condition and panic if it's false. */
761#define __RASSERT(cond, scond, format, args...) \
762do { \
763 if (!(cond)) \
764 reiserfs_panic(NULL, "assertion failure", "(" #cond ") at " \
765 __FILE__ ":%i:%s: " format "\n", \
766 in_interrupt() ? -1 : task_pid_nr(current), \
767 __LINE__, __func__ , ##args); \
768} while (0)
769
770#define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args)
771
772#if defined( CONFIG_REISERFS_CHECK )
773#define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args)
774#else
775#define RFALSE( cond, format, args... ) do {;} while( 0 )
776#endif
777
778#define CONSTF __attribute_const__
779/*
780 * Disk Data Structures
781 */
782
783/***************************************************************************/
784/* SUPER BLOCK */
785/***************************************************************************/
786
787/*
788 * Structure of super block on disk, a version of which in RAM is often accessed as REISERFS_SB(s)->s_rs
789 * the version in RAM is part of a larger structure containing fields never written to disk.
790 */
791#define UNSET_HASH 0 // read_super will guess about, what hash names
792 // in directories were sorted with
793#define TEA_HASH 1
794#define YURA_HASH 2
795#define R5_HASH 3
796#define DEFAULT_HASH R5_HASH
797
798struct journal_params {
799 __le32 jp_journal_1st_block; /* where does journal start from on its
800 * device */
801 __le32 jp_journal_dev; /* journal device st_rdev */
802 __le32 jp_journal_size; /* size of the journal */
803 __le32 jp_journal_trans_max; /* max number of blocks in a transaction. */
804 __le32 jp_journal_magic; /* random value made on fs creation (this
805 * was sb_journal_block_count) */
806 __le32 jp_journal_max_batch; /* max number of blocks to batch into a
807 * trans */
808 __le32 jp_journal_max_commit_age; /* in seconds, how old can an async
809 * commit be */
810 __le32 jp_journal_max_trans_age; /* in seconds, how old can a transaction
811 * be */
812};
813
814/* this is the super from 3.5.X, where X >= 10 */
815struct reiserfs_super_block_v1 {
816 __le32 s_block_count; /* blocks count */
817 __le32 s_free_blocks; /* free blocks count */
818 __le32 s_root_block; /* root block number */
819 struct journal_params s_journal;
820 __le16 s_blocksize; /* block size */
821 __le16 s_oid_maxsize; /* max size of object id array, see
822 * get_objectid() commentary */
823 __le16 s_oid_cursize; /* current size of object id array */
824 __le16 s_umount_state; /* this is set to 1 when filesystem was
825 * umounted, to 2 - when not */
826 char s_magic[10]; /* reiserfs magic string indicates that
827 * file system is reiserfs:
828 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs" */
829 __le16 s_fs_state; /* it is set to used by fsck to mark which
830 * phase of rebuilding is done */
831 __le32 s_hash_function_code; /* indicate, what hash function is being use
832 * to sort names in a directory*/
833 __le16 s_tree_height; /* height of disk tree */
834 __le16 s_bmap_nr; /* amount of bitmap blocks needed to address
835 * each block of file system */
836 __le16 s_version; /* this field is only reliable on filesystem
837 * with non-standard journal */
838 __le16 s_reserved_for_journal; /* size in blocks of journal area on main
839 * device, we need to keep after
840 * making fs with non-standard journal */
841} __attribute__ ((__packed__));
842
843#define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))
844
845/* this is the on disk super block */
846struct reiserfs_super_block {
847 struct reiserfs_super_block_v1 s_v1;
848 __le32 s_inode_generation;
849 __le32 s_flags; /* Right now used only by inode-attributes, if enabled */
850 unsigned char s_uuid[16]; /* filesystem unique identifier */
851 unsigned char s_label[16]; /* filesystem volume label */
852 __le16 s_mnt_count; /* Count of mounts since last fsck */
853 __le16 s_max_mnt_count; /* Maximum mounts before check */
854 __le32 s_lastcheck; /* Timestamp of last fsck */
855 __le32 s_check_interval; /* Interval between checks */
856 char s_unused[76]; /* zero filled by mkreiserfs and
857 * reiserfs_convert_objectid_map_v1()
858 * so any additions must be updated
859 * there as well. */
860} __attribute__ ((__packed__));
861
862#define SB_SIZE (sizeof(struct reiserfs_super_block))
863
864#define REISERFS_VERSION_1 0
865#define REISERFS_VERSION_2 2
866
867// on-disk super block fields converted to cpu form
868#define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)
869#define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))
870#define SB_BLOCKSIZE(s) \
871 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))
872#define SB_BLOCK_COUNT(s) \
873 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))
874#define SB_FREE_BLOCKS(s) \
875 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))
876#define SB_REISERFS_MAGIC(s) \
877 (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)
878#define SB_ROOT_BLOCK(s) \
879 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))
880#define SB_TREE_HEIGHT(s) \
881 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))
882#define SB_REISERFS_STATE(s) \
883 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))
884#define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))
885#define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))
886
887#define PUT_SB_BLOCK_COUNT(s, val) \
888 do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)
889#define PUT_SB_FREE_BLOCKS(s, val) \
890 do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)
891#define PUT_SB_ROOT_BLOCK(s, val) \
892 do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)
893#define PUT_SB_TREE_HEIGHT(s, val) \
894 do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)
895#define PUT_SB_REISERFS_STATE(s, val) \
896 do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)
897#define PUT_SB_VERSION(s, val) \
898 do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)
899#define PUT_SB_BMAP_NR(s, val) \
900 do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)
901
902#define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)
903#define SB_ONDISK_JOURNAL_SIZE(s) \
904 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))
905#define SB_ONDISK_JOURNAL_1st_BLOCK(s) \
906 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))
907#define SB_ONDISK_JOURNAL_DEVICE(s) \
908 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
909#define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
910 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
911
912#define is_block_in_log_or_reserved_area(s, block) \
913 block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
914 && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) + \
915 ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \
916 SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))
917
918int is_reiserfs_3_5(struct reiserfs_super_block *rs);
919int is_reiserfs_3_6(struct reiserfs_super_block *rs);
920int is_reiserfs_jr(struct reiserfs_super_block *rs);
921
922/* ReiserFS leaves the first 64k unused, so that partition labels have
923 enough space. If someone wants to write a fancy bootloader that
924 needs more than 64k, let us know, and this will be increased in size.
925 This number must be larger than than the largest block size on any
926 platform, or code will break. -Hans */
927#define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
928#define REISERFS_FIRST_BLOCK unused_define
929#define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES
930
931/* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
932#define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
933
934/* reiserfs internal error code (used by search_by_key and fix_nodes)) */
935#define CARRY_ON 0
936#define REPEAT_SEARCH -1
937#define IO_ERROR -2
938#define NO_DISK_SPACE -3
939#define NO_BALANCING_NEEDED (-4)
940#define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
941#define QUOTA_EXCEEDED -6
942
943typedef __u32 b_blocknr_t;
944typedef __le32 unp_t;
945
946struct unfm_nodeinfo {
947 unp_t unfm_nodenum;
948 unsigned short unfm_freespace;
949};
950
951/* there are two formats of keys: 3.5 and 3.6
952 */
953#define KEY_FORMAT_3_5 0
954#define KEY_FORMAT_3_6 1
955
956/* there are two stat datas */
957#define STAT_DATA_V1 0
958#define STAT_DATA_V2 1
959
960static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)
961{
962 return container_of(inode, struct reiserfs_inode_info, vfs_inode);
963}
964
965static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)
966{
967 return sb->s_fs_info;
968}
969
970/* Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u16
971 * which overflows on large file systems. */
972static inline __u32 reiserfs_bmap_count(struct super_block *sb)
973{
974 return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1;
975}
976
977static inline int bmap_would_wrap(unsigned bmap_nr)
978{
979 return bmap_nr > ((1LL << 16) - 1);
980}
981
982/** this says about version of key of all items (but stat data) the
983 object consists of */
984#define get_inode_item_key_version( inode ) \
985 ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
986
987#define set_inode_item_key_version( inode, version ) \
988 ({ if((version)==KEY_FORMAT_3_6) \
989 REISERFS_I(inode)->i_flags |= i_item_key_version_mask; \
990 else \
991 REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })
992
993#define get_inode_sd_version(inode) \
994 ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
995
996#define set_inode_sd_version(inode, version) \
997 ({ if((version)==STAT_DATA_V2) \
998 REISERFS_I(inode)->i_flags |= i_stat_data_version_mask; \
999 else \
1000 REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })
1001
1002/* This is an aggressive tail suppression policy, I am hoping it
1003 improves our benchmarks. The principle behind it is that percentage
1004 space saving is what matters, not absolute space saving. This is
1005 non-intuitive, but it helps to understand it if you consider that the
1006 cost to access 4 blocks is not much more than the cost to access 1
1007 block, if you have to do a seek and rotate. A tail risks a
1008 non-linear disk access that is significant as a percentage of total
1009 time cost for a 4 block file and saves an amount of space that is
1010 less significant as a percentage of space, or so goes the hypothesis.
1011 -Hans */
1012#define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
1013(\
1014 (!(n_tail_size)) || \
1015 (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
1016 ( (n_file_size) >= (n_block_size) * 4 ) || \
1017 ( ( (n_file_size) >= (n_block_size) * 3 ) && \
1018 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
1019 ( ( (n_file_size) >= (n_block_size) * 2 ) && \
1020 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
1021 ( ( (n_file_size) >= (n_block_size) ) && \
1022 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
1023)
1024
1025/* Another strategy for tails, this one means only create a tail if all the
1026 file would fit into one DIRECT item.
1027 Primary intention for this one is to increase performance by decreasing
1028 seeking.
1029*/
1030#define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
1031(\
1032 (!(n_tail_size)) || \
1033 (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
1034)
1035
1036/*
1037 * values for s_umount_state field
1038 */
1039#define REISERFS_VALID_FS 1
1040#define REISERFS_ERROR_FS 2
1041
1042//
1043// there are 5 item types currently
1044//
1045#define TYPE_STAT_DATA 0
1046#define TYPE_INDIRECT 1
1047#define TYPE_DIRECT 2
1048#define TYPE_DIRENTRY 3
1049#define TYPE_MAXTYPE 3
1050#define TYPE_ANY 15 // FIXME: comment is required
1051
1052/***************************************************************************/
1053/* KEY & ITEM HEAD */
1054/***************************************************************************/
1055
1056//
1057// directories use this key as well as old files
1058//
1059struct offset_v1 {
1060 __le32 k_offset;
1061 __le32 k_uniqueness;
1062} __attribute__ ((__packed__));
1063
1064struct offset_v2 {
1065 __le64 v;
1066} __attribute__ ((__packed__));
1067
1068static inline __u16 offset_v2_k_type(const struct offset_v2 *v2)
1069{
1070 __u8 type = le64_to_cpu(v2->v) >> 60;
1071 return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY;
1072}
1073
1074static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type)
1075{
1076 v2->v =
1077 (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60);
1078}
1079
1080static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2)
1081{
1082 return le64_to_cpu(v2->v) & (~0ULL >> 4);
1083}
1084
1085static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset)
1086{
1087 offset &= (~0ULL >> 4);
1088 v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset);
1089}
1090
1091/* Key of an item determines its location in the S+tree, and
1092 is composed of 4 components */
1093struct reiserfs_key {
1094 __le32 k_dir_id; /* packing locality: by default parent
1095 directory object id */
1096 __le32 k_objectid; /* object identifier */
1097 union {
1098 struct offset_v1 k_offset_v1;
1099 struct offset_v2 k_offset_v2;
1100 } __attribute__ ((__packed__)) u;
1101} __attribute__ ((__packed__));
1102
1103struct in_core_key {
1104 __u32 k_dir_id; /* packing locality: by default parent
1105 directory object id */
1106 __u32 k_objectid; /* object identifier */
1107 __u64 k_offset;
1108 __u8 k_type;
1109};
1110
1111struct cpu_key {
1112 struct in_core_key on_disk_key;
1113 int version;
1114 int key_length; /* 3 in all cases but direct2indirect and
1115 indirect2direct conversion */
1116};
1117
1118/* Our function for comparing keys can compare keys of different
1119 lengths. It takes as a parameter the length of the keys it is to
1120 compare. These defines are used in determining what is to be passed
1121 to it as that parameter. */
1122#define REISERFS_FULL_KEY_LEN 4
1123#define REISERFS_SHORT_KEY_LEN 2
1124
1125/* The result of the key compare */
1126#define FIRST_GREATER 1
1127#define SECOND_GREATER -1
1128#define KEYS_IDENTICAL 0
1129#define KEY_FOUND 1
1130#define KEY_NOT_FOUND 0
1131
1132#define KEY_SIZE (sizeof(struct reiserfs_key))
1133#define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
1134
1135/* return values for search_by_key and clones */
1136#define ITEM_FOUND 1
1137#define ITEM_NOT_FOUND 0
1138#define ENTRY_FOUND 1
1139#define ENTRY_NOT_FOUND 0
1140#define DIRECTORY_NOT_FOUND -1
1141#define REGULAR_FILE_FOUND -2
1142#define DIRECTORY_FOUND -3
1143#define BYTE_FOUND 1
1144#define BYTE_NOT_FOUND 0
1145#define FILE_NOT_FOUND -1
1146
1147#define POSITION_FOUND 1
1148#define POSITION_NOT_FOUND 0
1149
1150// return values for reiserfs_find_entry and search_by_entry_key
1151#define NAME_FOUND 1
1152#define NAME_NOT_FOUND 0
1153#define GOTO_PREVIOUS_ITEM 2
1154#define NAME_FOUND_INVISIBLE 3
1155
1156/* Everything in the filesystem is stored as a set of items. The
1157 item head contains the key of the item, its free space (for
1158 indirect items) and specifies the location of the item itself
1159 within the block. */
1160
1161struct item_head {
1162 /* Everything in the tree is found by searching for it based on
1163 * its key.*/
1164 struct reiserfs_key ih_key;
1165 union {
1166 /* The free space in the last unformatted node of an
1167 indirect item if this is an indirect item. This
1168 equals 0xFFFF iff this is a direct item or stat data
1169 item. Note that the key, not this field, is used to
1170 determine the item type, and thus which field this
1171 union contains. */
1172 __le16 ih_free_space_reserved;
1173 /* Iff this is a directory item, this field equals the
1174 number of directory entries in the directory item. */
1175 __le16 ih_entry_count;
1176 } __attribute__ ((__packed__)) u;
1177 __le16 ih_item_len; /* total size of the item body */
1178 __le16 ih_item_location; /* an offset to the item body
1179 * within the block */
1180 __le16 ih_version; /* 0 for all old items, 2 for new
1181 ones. Highest bit is set by fsck
1182 temporary, cleaned after all
1183 done */
1184} __attribute__ ((__packed__));
1185/* size of item header */
1186#define IH_SIZE (sizeof(struct item_head))
1187
1188#define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved)
1189#define ih_version(ih) le16_to_cpu((ih)->ih_version)
1190#define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count)
1191#define ih_location(ih) le16_to_cpu((ih)->ih_item_location)
1192#define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len)
1193
1194#define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
1195#define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0)
1196#define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
1197#define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
1198#define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
1199
1200#define unreachable_item(ih) (ih_version(ih) & (1 << 15))
1201
1202#define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
1203#define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
1204
1205/* these operate on indirect items, where you've got an array of ints
1206** at a possibly unaligned location. These are a noop on ia32
1207**
1208** p is the array of __u32, i is the index into the array, v is the value
1209** to store there.
1210*/
1211#define get_block_num(p, i) get_unaligned_le32((p) + (i))
1212#define put_block_num(p, i, v) put_unaligned_le32((v), (p) + (i))
1213
1214//
1215// in old version uniqueness field shows key type
1216//
1217#define V1_SD_UNIQUENESS 0
1218#define V1_INDIRECT_UNIQUENESS 0xfffffffe
1219#define V1_DIRECT_UNIQUENESS 0xffffffff
1220#define V1_DIRENTRY_UNIQUENESS 500
1221#define V1_ANY_UNIQUENESS 555 // FIXME: comment is required
1222
1223//
1224// here are conversion routines
1225//
1226static inline int uniqueness2type(__u32 uniqueness) CONSTF;
1227static inline int uniqueness2type(__u32 uniqueness)
1228{
1229 switch ((int)uniqueness) {
1230 case V1_SD_UNIQUENESS:
1231 return TYPE_STAT_DATA;
1232 case V1_INDIRECT_UNIQUENESS:
1233 return TYPE_INDIRECT;
1234 case V1_DIRECT_UNIQUENESS:
1235 return TYPE_DIRECT;
1236 case V1_DIRENTRY_UNIQUENESS:
1237 return TYPE_DIRENTRY;
1238 case V1_ANY_UNIQUENESS:
1239 default:
1240 return TYPE_ANY;
1241 }
1242}
1243
1244static inline __u32 type2uniqueness(int type) CONSTF;
1245static inline __u32 type2uniqueness(int type)
1246{
1247 switch (type) {
1248 case TYPE_STAT_DATA:
1249 return V1_SD_UNIQUENESS;
1250 case TYPE_INDIRECT:
1251 return V1_INDIRECT_UNIQUENESS;
1252 case TYPE_DIRECT:
1253 return V1_DIRECT_UNIQUENESS;
1254 case TYPE_DIRENTRY:
1255 return V1_DIRENTRY_UNIQUENESS;
1256 case TYPE_ANY:
1257 default:
1258 return V1_ANY_UNIQUENESS;
1259 }
1260}
1261
1262//
1263// key is pointer to on disk key which is stored in le, result is cpu,
1264// there is no way to get version of object from key, so, provide
1265// version to these defines
1266//
1267static inline loff_t le_key_k_offset(int version,
1268 const struct reiserfs_key *key)
1269{
1270 return (version == KEY_FORMAT_3_5) ?
1271 le32_to_cpu(key->u.k_offset_v1.k_offset) :
1272 offset_v2_k_offset(&(key->u.k_offset_v2));
1273}
1274
1275static inline loff_t le_ih_k_offset(const struct item_head *ih)
1276{
1277 return le_key_k_offset(ih_version(ih), &(ih->ih_key));
1278}
1279
1280static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key)
1281{
1282 return (version == KEY_FORMAT_3_5) ?
1283 uniqueness2type(le32_to_cpu(key->u.k_offset_v1.k_uniqueness)) :
1284 offset_v2_k_type(&(key->u.k_offset_v2));
1285}
1286
1287static inline loff_t le_ih_k_type(const struct item_head *ih)
1288{
1289 return le_key_k_type(ih_version(ih), &(ih->ih_key));
1290}
1291
1292static inline void set_le_key_k_offset(int version, struct reiserfs_key *key,
1293 loff_t offset)
1294{
1295 (version == KEY_FORMAT_3_5) ? (void)(key->u.k_offset_v1.k_offset = cpu_to_le32(offset)) : /* jdm check */
1296 (void)(set_offset_v2_k_offset(&(key->u.k_offset_v2), offset));
1297}
1298
1299static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset)
1300{
1301 set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
1302}
1303
1304static inline void set_le_key_k_type(int version, struct reiserfs_key *key,
1305 int type)
1306{
1307 (version == KEY_FORMAT_3_5) ?
1308 (void)(key->u.k_offset_v1.k_uniqueness =
1309 cpu_to_le32(type2uniqueness(type)))
1310 : (void)(set_offset_v2_k_type(&(key->u.k_offset_v2), type));
1311}
1312
1313static inline void set_le_ih_k_type(struct item_head *ih, int type)
1314{
1315 set_le_key_k_type(ih_version(ih), &(ih->ih_key), type);
1316}
1317
1318static inline int is_direntry_le_key(int version, struct reiserfs_key *key)
1319{
1320 return le_key_k_type(version, key) == TYPE_DIRENTRY;
1321}
1322
1323static inline int is_direct_le_key(int version, struct reiserfs_key *key)
1324{
1325 return le_key_k_type(version, key) == TYPE_DIRECT;
1326}
1327
1328static inline int is_indirect_le_key(int version, struct reiserfs_key *key)
1329{
1330 return le_key_k_type(version, key) == TYPE_INDIRECT;
1331}
1332
1333static inline int is_statdata_le_key(int version, struct reiserfs_key *key)
1334{
1335 return le_key_k_type(version, key) == TYPE_STAT_DATA;
1336}
1337
1338//
1339// item header has version.
1340//
1341static inline int is_direntry_le_ih(struct item_head *ih)
1342{
1343 return is_direntry_le_key(ih_version(ih), &ih->ih_key);
1344}
1345
1346static inline int is_direct_le_ih(struct item_head *ih)
1347{
1348 return is_direct_le_key(ih_version(ih), &ih->ih_key);
1349}
1350
1351static inline int is_indirect_le_ih(struct item_head *ih)
1352{
1353 return is_indirect_le_key(ih_version(ih), &ih->ih_key);
1354}
1355
1356static inline int is_statdata_le_ih(struct item_head *ih)
1357{
1358 return is_statdata_le_key(ih_version(ih), &ih->ih_key);
1359}
1360
1361//
1362// key is pointer to cpu key, result is cpu
1363//
1364static inline loff_t cpu_key_k_offset(const struct cpu_key *key)
1365{
1366 return key->on_disk_key.k_offset;
1367}
1368
1369static inline loff_t cpu_key_k_type(const struct cpu_key *key)
1370{
1371 return key->on_disk_key.k_type;
1372}
1373
1374static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset)
1375{
1376 key->on_disk_key.k_offset = offset;
1377}
1378
1379static inline void set_cpu_key_k_type(struct cpu_key *key, int type)
1380{
1381 key->on_disk_key.k_type = type;
1382}
1383
1384static inline void cpu_key_k_offset_dec(struct cpu_key *key)
1385{
1386 key->on_disk_key.k_offset--;
1387}
1388
1389#define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
1390#define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
1391#define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
1392#define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
1393
1394/* are these used ? */
1395#define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
1396#define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
1397#define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
1398#define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
1399
1400#define I_K_KEY_IN_ITEM(ih, key, n_blocksize) \
1401 (!COMP_SHORT_KEYS(ih, key) && \
1402 I_OFF_BYTE_IN_ITEM(ih, k_offset(key), n_blocksize))
1403
1404/* maximal length of item */
1405#define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
1406#define MIN_ITEM_LEN 1
1407
1408/* object identifier for root dir */
1409#define REISERFS_ROOT_OBJECTID 2
1410#define REISERFS_ROOT_PARENT_OBJECTID 1
1411
1412extern struct reiserfs_key root_key;
1413
1414/*
1415 * Picture represents a leaf of the S+tree
1416 * ______________________________________________________
1417 * | | Array of | | |
1418 * |Block | Object-Item | F r e e | Objects- |
1419 * | head | Headers | S p a c e | Items |
1420 * |______|_______________|___________________|___________|
1421 */
1422
1423/* Header of a disk block. More precisely, header of a formatted leaf
1424 or internal node, and not the header of an unformatted node. */
1425struct block_head {
1426 __le16 blk_level; /* Level of a block in the tree. */
1427 __le16 blk_nr_item; /* Number of keys/items in a block. */
1428 __le16 blk_free_space; /* Block free space in bytes. */
1429 __le16 blk_reserved;
1430 /* dump this in v4/planA */
1431 struct reiserfs_key blk_right_delim_key; /* kept only for compatibility */
1432};
1433
1434#define BLKH_SIZE (sizeof(struct block_head))
1435#define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level))
1436#define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item))
1437#define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space))
1438#define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved))
1439#define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val))
1440#define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val))
1441#define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
1442#define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
1443#define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key)
1444#define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val)
1445
1446/*
1447 * values for blk_level field of the struct block_head
1448 */
1449
1450#define FREE_LEVEL 0 /* when node gets removed from the tree its
1451 blk_level is set to FREE_LEVEL. It is then
1452 used to see whether the node is still in the
1453 tree */
1454
1455#define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level. */
1456
1457/* Given the buffer head of a formatted node, resolve to the block head of that node. */
1458#define B_BLK_HEAD(bh) ((struct block_head *)((bh)->b_data))
1459/* Number of items that are in buffer. */
1460#define B_NR_ITEMS(bh) (blkh_nr_item(B_BLK_HEAD(bh)))
1461#define B_LEVEL(bh) (blkh_level(B_BLK_HEAD(bh)))
1462#define B_FREE_SPACE(bh) (blkh_free_space(B_BLK_HEAD(bh)))
1463
1464#define PUT_B_NR_ITEMS(bh, val) do { set_blkh_nr_item(B_BLK_HEAD(bh), val); } while (0)
1465#define PUT_B_LEVEL(bh, val) do { set_blkh_level(B_BLK_HEAD(bh), val); } while (0)
1466#define PUT_B_FREE_SPACE(bh, val) do { set_blkh_free_space(B_BLK_HEAD(bh), val); } while (0)
1467
1468/* Get right delimiting key. -- little endian */
1469#define B_PRIGHT_DELIM_KEY(bh) (&(blk_right_delim_key(B_BLK_HEAD(bh))))
1470
1471/* Does the buffer contain a disk leaf. */
1472#define B_IS_ITEMS_LEVEL(bh) (B_LEVEL(bh) == DISK_LEAF_NODE_LEVEL)
1473
1474/* Does the buffer contain a disk internal node */
1475#define B_IS_KEYS_LEVEL(bh) (B_LEVEL(bh) > DISK_LEAF_NODE_LEVEL \
1476 && B_LEVEL(bh) <= MAX_HEIGHT)
1477
1478/***************************************************************************/
1479/* STAT DATA */
1480/***************************************************************************/
1481
1482//
1483// old stat data is 32 bytes long. We are going to distinguish new one by
1484// different size
1485//
1486struct stat_data_v1 {
1487 __le16 sd_mode; /* file type, permissions */
1488 __le16 sd_nlink; /* number of hard links */
1489 __le16 sd_uid; /* owner */
1490 __le16 sd_gid; /* group */
1491 __le32 sd_size; /* file size */
1492 __le32 sd_atime; /* time of last access */
1493 __le32 sd_mtime; /* time file was last modified */
1494 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
1495 union {
1496 __le32 sd_rdev;
1497 __le32 sd_blocks; /* number of blocks file uses */
1498 } __attribute__ ((__packed__)) u;
1499 __le32 sd_first_direct_byte; /* first byte of file which is stored
1500 in a direct item: except that if it
1501 equals 1 it is a symlink and if it
1502 equals ~(__u32)0 there is no
1503 direct item. The existence of this
1504 field really grates on me. Let's
1505 replace it with a macro based on
1506 sd_size and our tail suppression
1507 policy. Someday. -Hans */
1508} __attribute__ ((__packed__));
1509
1510#define SD_V1_SIZE (sizeof(struct stat_data_v1))
1511#define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5)
1512#define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
1513#define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
1514#define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink))
1515#define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v))
1516#define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid))
1517#define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v))
1518#define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid))
1519#define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v))
1520#define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size))
1521#define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v))
1522#define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
1523#define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
1524#define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
1525#define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
1526#define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
1527#define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
1528#define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
1529#define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
1530#define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks))
1531#define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
1532#define sd_v1_first_direct_byte(sdp) \
1533 (le32_to_cpu((sdp)->sd_first_direct_byte))
1534#define set_sd_v1_first_direct_byte(sdp,v) \
1535 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
1536
1537/* inode flags stored in sd_attrs (nee sd_reserved) */
1538
1539/* we want common flags to have the same values as in ext2,
1540 so chattr(1) will work without problems */
1541#define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL
1542#define REISERFS_APPEND_FL FS_APPEND_FL
1543#define REISERFS_SYNC_FL FS_SYNC_FL
1544#define REISERFS_NOATIME_FL FS_NOATIME_FL
1545#define REISERFS_NODUMP_FL FS_NODUMP_FL
1546#define REISERFS_SECRM_FL FS_SECRM_FL
1547#define REISERFS_UNRM_FL FS_UNRM_FL
1548#define REISERFS_COMPR_FL FS_COMPR_FL
1549#define REISERFS_NOTAIL_FL FS_NOTAIL_FL
1550
1551/* persistent flags that file inherits from the parent directory */
1552#define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \
1553 REISERFS_SYNC_FL | \
1554 REISERFS_NOATIME_FL | \
1555 REISERFS_NODUMP_FL | \
1556 REISERFS_SECRM_FL | \
1557 REISERFS_COMPR_FL | \
1558 REISERFS_NOTAIL_FL )
1559
1560/* Stat Data on disk (reiserfs version of UFS disk inode minus the
1561 address blocks) */
1562struct stat_data {
1563 __le16 sd_mode; /* file type, permissions */
1564 __le16 sd_attrs; /* persistent inode flags */
1565 __le32 sd_nlink; /* number of hard links */
1566 __le64 sd_size; /* file size */
1567 __le32 sd_uid; /* owner */
1568 __le32 sd_gid; /* group */
1569 __le32 sd_atime; /* time of last access */
1570 __le32 sd_mtime; /* time file was last modified */
1571 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
1572 __le32 sd_blocks;
1573 union {
1574 __le32 sd_rdev;
1575 __le32 sd_generation;
1576 //__le32 sd_first_direct_byte;
1577 /* first byte of file which is stored in a
1578 direct item: except that if it equals 1
1579 it is a symlink and if it equals
1580 ~(__u32)0 there is no direct item. The
1581 existence of this field really grates
1582 on me. Let's replace it with a macro
1583 based on sd_size and our tail
1584 suppression policy? */
1585 } __attribute__ ((__packed__)) u;
1586} __attribute__ ((__packed__));
1587//
1588// this is 44 bytes long
1589//
1590#define SD_SIZE (sizeof(struct stat_data))
1591#define SD_V2_SIZE SD_SIZE
1592#define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6)
1593#define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
1594#define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
1595/* sd_reserved */
1596/* set_sd_reserved */
1597#define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink))
1598#define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v))
1599#define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size))
1600#define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v))
1601#define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid))
1602#define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v))
1603#define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid))
1604#define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v))
1605#define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
1606#define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
1607#define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
1608#define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
1609#define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
1610#define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
1611#define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks))
1612#define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
1613#define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
1614#define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
1615#define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation))
1616#define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
1617#define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs))
1618#define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v))
1619
1620/***************************************************************************/
1621/* DIRECTORY STRUCTURE */
1622/***************************************************************************/
1623/*
1624 Picture represents the structure of directory items
1625 ________________________________________________
1626 | Array of | | | | | |
1627 | directory |N-1| N-2 | .... | 1st |0th|
1628 | entry headers | | | | | |
1629 |_______________|___|_____|________|_______|___|
1630 <---- directory entries ------>
1631
1632 First directory item has k_offset component 1. We store "." and ".."
1633 in one item, always, we never split "." and ".." into differing
1634 items. This makes, among other things, the code for removing
1635 directories simpler. */
1636#define SD_OFFSET 0
1637#define SD_UNIQUENESS 0
1638#define DOT_OFFSET 1
1639#define DOT_DOT_OFFSET 2
1640#define DIRENTRY_UNIQUENESS 500
1641
1642/* */
1643#define FIRST_ITEM_OFFSET 1
1644
1645/*
1646 Q: How to get key of object pointed to by entry from entry?
1647
1648 A: Each directory entry has its header. This header has deh_dir_id and deh_objectid fields, those are key
1649 of object, entry points to */
1650
1651/* NOT IMPLEMENTED:
1652 Directory will someday contain stat data of object */
1653
1654struct reiserfs_de_head {
1655 __le32 deh_offset; /* third component of the directory entry key */
1656 __le32 deh_dir_id; /* objectid of the parent directory of the object, that is referenced
1657 by directory entry */
1658 __le32 deh_objectid; /* objectid of the object, that is referenced by directory entry */
1659 __le16 deh_location; /* offset of name in the whole item */
1660 __le16 deh_state; /* whether 1) entry contains stat data (for future), and 2) whether
1661 entry is hidden (unlinked) */
1662} __attribute__ ((__packed__));
1663#define DEH_SIZE sizeof(struct reiserfs_de_head)
1664#define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset))
1665#define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id))
1666#define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid))
1667#define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location))
1668#define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state))
1669
1670#define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v)))
1671#define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v)))
1672#define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
1673#define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
1674#define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v)))
1675
1676/* empty directory contains two entries "." and ".." and their headers */
1677#define EMPTY_DIR_SIZE \
1678(DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
1679
1680/* old format directories have this size when empty */
1681#define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
1682
1683#define DEH_Statdata 0 /* not used now */
1684#define DEH_Visible 2
1685
1686/* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
1687#if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
1688# define ADDR_UNALIGNED_BITS (3)
1689#endif
1690
1691/* These are only used to manipulate deh_state.
1692 * Because of this, we'll use the ext2_ bit routines,
1693 * since they are little endian */
1694#ifdef ADDR_UNALIGNED_BITS
1695
1696# define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
1697# define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
1698
1699# define set_bit_unaligned(nr, addr) \
1700 __test_and_set_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1701# define clear_bit_unaligned(nr, addr) \
1702 __test_and_clear_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1703# define test_bit_unaligned(nr, addr) \
1704 test_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1705
1706#else
1707
1708# define set_bit_unaligned(nr, addr) __test_and_set_bit_le(nr, addr)
1709# define clear_bit_unaligned(nr, addr) __test_and_clear_bit_le(nr, addr)
1710# define test_bit_unaligned(nr, addr) test_bit_le(nr, addr)
1711
1712#endif
1713
1714#define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1715#define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1716#define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1717#define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1718
1719#define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1720#define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1721#define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1722
1723extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid,
1724 __le32 par_dirid, __le32 par_objid);
1725extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid,
1726 __le32 par_dirid, __le32 par_objid);
1727
1728/* array of the entry headers */
1729 /* get item body */
1730#define B_I_PITEM(bh,ih) ( (bh)->b_data + ih_location(ih) )
1731#define B_I_DEH(bh,ih) ((struct reiserfs_de_head *)(B_I_PITEM(bh,ih)))
1732
1733/* length of the directory entry in directory item. This define
1734 calculates length of i-th directory entry using directory entry
1735 locations from dir entry head. When it calculates length of 0-th
1736 directory entry, it uses length of whole item in place of entry
1737 location of the non-existent following entry in the calculation.
1738 See picture above.*/
1739/*
1740#define I_DEH_N_ENTRY_LENGTH(ih,deh,i) \
1741((i) ? (deh_location((deh)-1) - deh_location((deh))) : (ih_item_len((ih)) - deh_location((deh))))
1742*/
1743static inline int entry_length(const struct buffer_head *bh,
1744 const struct item_head *ih, int pos_in_item)
1745{
1746 struct reiserfs_de_head *deh;
1747
1748 deh = B_I_DEH(bh, ih) + pos_in_item;
1749 if (pos_in_item)
1750 return deh_location(deh - 1) - deh_location(deh);
1751
1752 return ih_item_len(ih) - deh_location(deh);
1753}
1754
1755/* number of entries in the directory item, depends on ENTRY_COUNT being at the start of directory dynamic data. */
1756#define I_ENTRY_COUNT(ih) (ih_entry_count((ih)))
1757
1758/* name by bh, ih and entry_num */
1759#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))))
1760
1761// two entries per block (at least)
1762#define REISERFS_MAX_NAME(block_size) 255
1763
1764/* this structure is used for operations on directory entries. It is
1765 not a disk structure. */
1766/* When reiserfs_find_entry or search_by_entry_key find directory
1767 entry, they return filled reiserfs_dir_entry structure */
1768struct reiserfs_dir_entry {
1769 struct buffer_head *de_bh;
1770 int de_item_num;
1771 struct item_head *de_ih;
1772 int de_entry_num;
1773 struct reiserfs_de_head *de_deh;
1774 int de_entrylen;
1775 int de_namelen;
1776 char *de_name;
1777 unsigned long *de_gen_number_bit_string;
1778
1779 __u32 de_dir_id;
1780 __u32 de_objectid;
1781
1782 struct cpu_key de_entry_key;
1783};
1784
1785/* these defines are useful when a particular member of a reiserfs_dir_entry is needed */
1786
1787/* pointer to file name, stored in entry */
1788#define B_I_DEH_ENTRY_FILE_NAME(bh,ih,deh) (B_I_PITEM (bh, ih) + deh_location(deh))
1789
1790/* length of name */
1791#define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
1792(I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
1793
1794/* hash value occupies bits from 7 up to 30 */
1795#define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
1796/* generation number occupies 7 bits starting from 0 up to 6 */
1797#define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
1798#define MAX_GENERATION_NUMBER 127
1799
1800#define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
1801
1802/*
1803 * Picture represents an internal node of the reiserfs tree
1804 * ______________________________________________________
1805 * | | Array of | Array of | Free |
1806 * |block | keys | pointers | space |
1807 * | head | N | N+1 | |
1808 * |______|_______________|___________________|___________|
1809 */
1810
1811/***************************************************************************/
1812/* DISK CHILD */
1813/***************************************************************************/
1814/* Disk child pointer: The pointer from an internal node of the tree
1815 to a node that is on disk. */
1816struct disk_child {
1817 __le32 dc_block_number; /* Disk child's block number. */
1818 __le16 dc_size; /* Disk child's used space. */
1819 __le16 dc_reserved;
1820};
1821
1822#define DC_SIZE (sizeof(struct disk_child))
1823#define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number))
1824#define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size))
1825#define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
1826#define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
1827
1828/* Get disk child by buffer header and position in the tree node. */
1829#define B_N_CHILD(bh, n_pos) ((struct disk_child *)\
1830((bh)->b_data + BLKH_SIZE + B_NR_ITEMS(bh) * KEY_SIZE + DC_SIZE * (n_pos)))
1831
1832/* Get disk child number by buffer header and position in the tree node. */
1833#define B_N_CHILD_NUM(bh, n_pos) (dc_block_number(B_N_CHILD(bh, n_pos)))
1834#define PUT_B_N_CHILD_NUM(bh, n_pos, val) \
1835 (put_dc_block_number(B_N_CHILD(bh, n_pos), val))
1836
1837 /* maximal value of field child_size in structure disk_child */
1838 /* child size is the combined size of all items and their headers */
1839#define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
1840
1841/* amount of used space in buffer (not including block head) */
1842#define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
1843
1844/* max and min number of keys in internal node */
1845#define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
1846#define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2)
1847
1848/***************************************************************************/
1849/* PATH STRUCTURES AND DEFINES */
1850/***************************************************************************/
1851
1852/* Search_by_key fills up the path from the root to the leaf as it descends the tree looking for the
1853 key. It uses reiserfs_bread to try to find buffers in the cache given their block number. If it
1854 does not find them in the cache it reads them from disk. For each node search_by_key finds using
1855 reiserfs_bread it then uses bin_search to look through that node. bin_search will find the
1856 position of the block_number of the next node if it is looking through an internal node. If it
1857 is looking through a leaf node bin_search will find the position of the item which has key either
1858 equal to given key, or which is the maximal key less than the given key. */
1859
1860struct path_element {
1861 struct buffer_head *pe_buffer; /* Pointer to the buffer at the path in the tree. */
1862 int pe_position; /* Position in the tree node which is placed in the */
1863 /* buffer above. */
1864};
1865
1866#define MAX_HEIGHT 5 /* maximal height of a tree. don't change this without changing JOURNAL_PER_BALANCE_CNT */
1867#define EXTENDED_MAX_HEIGHT 7 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
1868#define FIRST_PATH_ELEMENT_OFFSET 2 /* Must be equal to at least 2. */
1869
1870#define ILLEGAL_PATH_ELEMENT_OFFSET 1 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
1871#define MAX_FEB_SIZE 6 /* this MUST be MAX_HEIGHT + 1. See about FEB below */
1872
1873/* We need to keep track of who the ancestors of nodes are. When we
1874 perform a search we record which nodes were visited while
1875 descending the tree looking for the node we searched for. This list
1876 of nodes is called the path. This information is used while
1877 performing balancing. Note that this path information may become
1878 invalid, and this means we must check it when using it to see if it
1879 is still valid. You'll need to read search_by_key and the comments
1880 in it, especially about decrement_counters_in_path(), to understand
1881 this structure.
1882
1883Paths make the code so much harder to work with and debug.... An
1884enormous number of bugs are due to them, and trying to write or modify
1885code that uses them just makes my head hurt. They are based on an
1886excessive effort to avoid disturbing the precious VFS code.:-( The
1887gods only know how we are going to SMP the code that uses them.
1888znodes are the way! */
1889
1890#define PATH_READA 0x1 /* do read ahead */
1891#define PATH_READA_BACK 0x2 /* read backwards */
1892
1893struct treepath {
1894 int path_length; /* Length of the array above. */
1895 int reada;
1896 struct path_element path_elements[EXTENDED_MAX_HEIGHT]; /* Array of the path elements. */
1897 int pos_in_item;
1898};
1899
1900#define pos_in_item(path) ((path)->pos_in_item)
1901
1902#define INITIALIZE_PATH(var) \
1903struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
1904
1905/* Get path element by path and path position. */
1906#define PATH_OFFSET_PELEMENT(path, n_offset) ((path)->path_elements + (n_offset))
1907
1908/* Get buffer header at the path by path and path position. */
1909#define PATH_OFFSET_PBUFFER(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_buffer)
1910
1911/* Get position in the element at the path by path and path position. */
1912#define PATH_OFFSET_POSITION(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_position)
1913
1914#define PATH_PLAST_BUFFER(path) (PATH_OFFSET_PBUFFER((path), (path)->path_length))
1915 /* you know, to the person who didn't
1916 write this the macro name does not
1917 at first suggest what it does.
1918 Maybe POSITION_FROM_PATH_END? Or
1919 maybe we should just focus on
1920 dumping paths... -Hans */
1921#define PATH_LAST_POSITION(path) (PATH_OFFSET_POSITION((path), (path)->path_length))
1922
1923#define PATH_PITEM_HEAD(path) B_N_PITEM_HEAD(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path))
1924
1925/* in do_balance leaf has h == 0 in contrast with path structure,
1926 where root has level == 0. That is why we need these defines */
1927#define PATH_H_PBUFFER(path, h) PATH_OFFSET_PBUFFER (path, path->path_length - (h)) /* tb->S[h] */
1928#define PATH_H_PPARENT(path, h) PATH_H_PBUFFER (path, (h) + 1) /* tb->F[h] or tb->S[0]->b_parent */
1929#define PATH_H_POSITION(path, h) PATH_OFFSET_POSITION (path, path->path_length - (h))
1930#define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1) /* tb->S[h]->b_item_order */
1931
1932#define PATH_H_PATH_OFFSET(path, n_h) ((path)->path_length - (n_h))
1933
1934#define get_last_bh(path) PATH_PLAST_BUFFER(path)
1935#define get_ih(path) PATH_PITEM_HEAD(path)
1936#define get_item_pos(path) PATH_LAST_POSITION(path)
1937#define get_item(path) ((void *)B_N_PITEM(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION (path)))
1938#define item_moved(ih,path) comp_items(ih, path)
1939#define path_changed(ih,path) comp_items (ih, path)
1940
1941/***************************************************************************/
1942/* MISC */
1943/***************************************************************************/
1944
1945/* Size of pointer to the unformatted node. */
1946#define UNFM_P_SIZE (sizeof(unp_t))
1947#define UNFM_P_SHIFT 2
1948
1949// in in-core inode key is stored on le form
1950#define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
1951
1952#define MAX_UL_INT 0xffffffff
1953#define MAX_INT 0x7ffffff
1954#define MAX_US_INT 0xffff
1955
1956// reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
1957#define U32_MAX (~(__u32)0)
1958
1959static inline loff_t max_reiserfs_offset(struct inode *inode)
1960{
1961 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
1962 return (loff_t) U32_MAX;
1963
1964 return (loff_t) ((~(__u64) 0) >> 4);
1965}
1966
1967/*#define MAX_KEY_UNIQUENESS MAX_UL_INT*/
1968#define MAX_KEY_OBJECTID MAX_UL_INT
1969
1970#define MAX_B_NUM MAX_UL_INT
1971#define MAX_FC_NUM MAX_US_INT
1972
1973/* the purpose is to detect overflow of an unsigned short */
1974#define REISERFS_LINK_MAX (MAX_US_INT - 1000)
1975
1976/* The following defines are used in reiserfs_insert_item and reiserfs_append_item */
1977#define REISERFS_KERNEL_MEM 0 /* reiserfs kernel memory mode */
1978#define REISERFS_USER_MEM 1 /* reiserfs user memory mode */
1979
1980#define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
1981#define get_generation(s) atomic_read (&fs_generation(s))
1982#define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen)
1983#define __fs_changed(gen,s) (gen != get_generation (s))
1984#define fs_changed(gen,s) \
1985({ \
1986 reiserfs_cond_resched(s); \
1987 __fs_changed(gen, s); \
1988})
1989
1990/***************************************************************************/
1991/* FIXATE NODES */
1992/***************************************************************************/
1993
1994#define VI_TYPE_LEFT_MERGEABLE 1
1995#define VI_TYPE_RIGHT_MERGEABLE 2
1996
1997/* To make any changes in the tree we always first find node, that
1998 contains item to be changed/deleted or place to insert a new
1999 item. We call this node S. To do balancing we need to decide what
2000 we will shift to left/right neighbor, or to a new node, where new
2001 item will be etc. To make this analysis simpler we build virtual
2002 node. Virtual node is an array of items, that will replace items of
2003 node S. (For instance if we are going to delete an item, virtual
2004 node does not contain it). Virtual node keeps information about
2005 item sizes and types, mergeability of first and last items, sizes
2006 of all entries in directory item. We use this array of items when
2007 calculating what we can shift to neighbors and how many nodes we
2008 have to have if we do not any shiftings, if we shift to left/right
2009 neighbor or to both. */
2010struct virtual_item {
2011 int vi_index; // index in the array of item operations
2012 unsigned short vi_type; // left/right mergeability
2013 unsigned short vi_item_len; /* length of item that it will have after balancing */
2014 struct item_head *vi_ih;
2015 const char *vi_item; // body of item (old or new)
2016 const void *vi_new_data; // 0 always but paste mode
2017 void *vi_uarea; // item specific area
2018};
2019
2020struct virtual_node {
2021 char *vn_free_ptr; /* this is a pointer to the free space in the buffer */
2022 unsigned short vn_nr_item; /* number of items in virtual node */
2023 short vn_size; /* size of node , that node would have if it has unlimited size and no balancing is performed */
2024 short vn_mode; /* mode of balancing (paste, insert, delete, cut) */
2025 short vn_affected_item_num;
2026 short vn_pos_in_item;
2027 struct item_head *vn_ins_ih; /* item header of inserted item, 0 for other modes */
2028 const void *vn_data;
2029 struct virtual_item *vn_vi; /* array of items (including a new one, excluding item to be deleted) */
2030};
2031
2032/* used by directory items when creating virtual nodes */
2033struct direntry_uarea {
2034 int flags;
2035 __u16 entry_count;
2036 __u16 entry_sizes[1];
2037} __attribute__ ((__packed__));
2038
2039/***************************************************************************/
2040/* TREE BALANCE */
2041/***************************************************************************/
2042
2043/* This temporary structure is used in tree balance algorithms, and
2044 constructed as we go to the extent that its various parts are
2045 needed. It contains arrays of nodes that can potentially be
2046 involved in the balancing of node S, and parameters that define how
2047 each of the nodes must be balanced. Note that in these algorithms
2048 for balancing the worst case is to need to balance the current node
2049 S and the left and right neighbors and all of their parents plus
2050 create a new node. We implement S1 balancing for the leaf nodes
2051 and S0 balancing for the internal nodes (S1 and S0 are defined in
2052 our papers.)*/
2053
2054#define MAX_FREE_BLOCK 7 /* size of the array of buffers to free at end of do_balance */
2055
2056/* maximum number of FEB blocknrs on a single level */
2057#define MAX_AMOUNT_NEEDED 2
2058
2059/* someday somebody will prefix every field in this struct with tb_ */
2060struct tree_balance {
2061 int tb_mode;
2062 int need_balance_dirty;
2063 struct super_block *tb_sb;
2064 struct reiserfs_transaction_handle *transaction_handle;
2065 struct treepath *tb_path;
2066 struct buffer_head *L[MAX_HEIGHT]; /* array of left neighbors of nodes in the path */
2067 struct buffer_head *R[MAX_HEIGHT]; /* array of right neighbors of nodes in the path */
2068 struct buffer_head *FL[MAX_HEIGHT]; /* array of fathers of the left neighbors */
2069 struct buffer_head *FR[MAX_HEIGHT]; /* array of fathers of the right neighbors */
2070 struct buffer_head *CFL[MAX_HEIGHT]; /* array of common parents of center node and its left neighbor */
2071 struct buffer_head *CFR[MAX_HEIGHT]; /* array of common parents of center node and its right neighbor */
2072
2073 struct buffer_head *FEB[MAX_FEB_SIZE]; /* array of empty buffers. Number of buffers in array equals
2074 cur_blknum. */
2075 struct buffer_head *used[MAX_FEB_SIZE];
2076 struct buffer_head *thrown[MAX_FEB_SIZE];
2077 int lnum[MAX_HEIGHT]; /* array of number of items which must be
2078 shifted to the left in order to balance the
2079 current node; for leaves includes item that
2080 will be partially shifted; for internal
2081 nodes, it is the number of child pointers
2082 rather than items. It includes the new item
2083 being created. The code sometimes subtracts
2084 one to get the number of wholly shifted
2085 items for other purposes. */
2086 int rnum[MAX_HEIGHT]; /* substitute right for left in comment above */
2087 int lkey[MAX_HEIGHT]; /* array indexed by height h mapping the key delimiting L[h] and
2088 S[h] to its item number within the node CFL[h] */
2089 int rkey[MAX_HEIGHT]; /* substitute r for l in comment above */
2090 int insert_size[MAX_HEIGHT]; /* the number of bytes by we are trying to add or remove from
2091 S[h]. A negative value means removing. */
2092 int blknum[MAX_HEIGHT]; /* number of nodes that will replace node S[h] after
2093 balancing on the level h of the tree. If 0 then S is
2094 being deleted, if 1 then S is remaining and no new nodes
2095 are being created, if 2 or 3 then 1 or 2 new nodes is
2096 being created */
2097
2098 /* fields that are used only for balancing leaves of the tree */
2099 int cur_blknum; /* number of empty blocks having been already allocated */
2100 int s0num; /* number of items that fall into left most node when S[0] splits */
2101 int s1num; /* number of items that fall into first new node when S[0] splits */
2102 int s2num; /* number of items that fall into second new node when S[0] splits */
2103 int lbytes; /* number of bytes which can flow to the left neighbor from the left */
2104 /* most liquid item that cannot be shifted from S[0] entirely */
2105 /* if -1 then nothing will be partially shifted */
2106 int rbytes; /* number of bytes which will flow to the right neighbor from the right */
2107 /* most liquid item that cannot be shifted from S[0] entirely */
2108 /* if -1 then nothing will be partially shifted */
2109 int s1bytes; /* number of bytes which flow to the first new node when S[0] splits */
2110 /* note: if S[0] splits into 3 nodes, then items do not need to be cut */
2111 int s2bytes;
2112 struct buffer_head *buf_to_free[MAX_FREE_BLOCK]; /* buffers which are to be freed after do_balance finishes by unfix_nodes */
2113 char *vn_buf; /* kmalloced memory. Used to create
2114 virtual node and keep map of
2115 dirtied bitmap blocks */
2116 int vn_buf_size; /* size of the vn_buf */
2117 struct virtual_node *tb_vn; /* VN starts after bitmap of bitmap blocks */
2118
2119 int fs_gen; /* saved value of `reiserfs_generation' counter
2120 see FILESYSTEM_CHANGED() macro in reiserfs_fs.h */
2121#ifdef DISPLACE_NEW_PACKING_LOCALITIES
2122 struct in_core_key key; /* key pointer, to pass to block allocator or
2123 another low-level subsystem */
2124#endif
2125};
2126
2127/* These are modes of balancing */
2128
2129/* When inserting an item. */
2130#define M_INSERT 'i'
2131/* When inserting into (directories only) or appending onto an already
2132 existent item. */
2133#define M_PASTE 'p'
2134/* When deleting an item. */
2135#define M_DELETE 'd'
2136/* When truncating an item or removing an entry from a (directory) item. */
2137#define M_CUT 'c'
2138
2139/* used when balancing on leaf level skipped (in reiserfsck) */
2140#define M_INTERNAL 'n'
2141
2142/* When further balancing is not needed, then do_balance does not need
2143 to be called. */
2144#define M_SKIP_BALANCING 's'
2145#define M_CONVERT 'v'
2146
2147/* modes of leaf_move_items */
2148#define LEAF_FROM_S_TO_L 0
2149#define LEAF_FROM_S_TO_R 1
2150#define LEAF_FROM_R_TO_L 2
2151#define LEAF_FROM_L_TO_R 3
2152#define LEAF_FROM_S_TO_SNEW 4
2153
2154#define FIRST_TO_LAST 0
2155#define LAST_TO_FIRST 1
2156
2157/* used in do_balance for passing parent of node information that has
2158 been gotten from tb struct */
2159struct buffer_info {
2160 struct tree_balance *tb;
2161 struct buffer_head *bi_bh;
2162 struct buffer_head *bi_parent;
2163 int bi_position;
2164};
2165
2166static inline struct super_block *sb_from_tb(struct tree_balance *tb)
2167{
2168 return tb ? tb->tb_sb : NULL;
2169}
2170
2171static inline struct super_block *sb_from_bi(struct buffer_info *bi)
2172{
2173 return bi ? sb_from_tb(bi->tb) : NULL;
2174}
2175
2176/* there are 4 types of items: stat data, directory item, indirect, direct.
2177+-------------------+------------+--------------+------------+
2178| | k_offset | k_uniqueness | mergeable? |
2179+-------------------+------------+--------------+------------+
2180| stat data | 0 | 0 | no |
2181+-------------------+------------+--------------+------------+
2182| 1st directory item| DOT_OFFSET |DIRENTRY_UNIQUENESS| no |
2183| non 1st directory | hash value | | yes |
2184| item | | | |
2185+-------------------+------------+--------------+------------+
2186| indirect item | offset + 1 |TYPE_INDIRECT | if this is not the first indirect item of the object
2187+-------------------+------------+--------------+------------+
2188| direct item | offset + 1 |TYPE_DIRECT | if not this is not the first direct item of the object
2189+-------------------+------------+--------------+------------+
2190*/
2191
2192struct item_operations {
2193 int (*bytes_number) (struct item_head * ih, int block_size);
2194 void (*decrement_key) (struct cpu_key *);
2195 int (*is_left_mergeable) (struct reiserfs_key * ih,
2196 unsigned long bsize);
2197 void (*print_item) (struct item_head *, char *item);
2198 void (*check_item) (struct item_head *, char *item);
2199
2200 int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
2201 int is_affected, int insert_size);
2202 int (*check_left) (struct virtual_item * vi, int free,
2203 int start_skip, int end_skip);
2204 int (*check_right) (struct virtual_item * vi, int free);
2205 int (*part_size) (struct virtual_item * vi, int from, int to);
2206 int (*unit_num) (struct virtual_item * vi);
2207 void (*print_vi) (struct virtual_item * vi);
2208};
2209
2210extern struct item_operations *item_ops[TYPE_ANY + 1];
2211
2212#define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
2213#define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
2214#define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item)
2215#define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item)
2216#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)
2217#define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
2218#define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free)
2219#define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to)
2220#define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi)
2221#define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi)
2222
2223#define COMP_SHORT_KEYS comp_short_keys
2224
2225/* number of blocks pointed to by the indirect item */
2226#define I_UNFM_NUM(ih) (ih_item_len(ih) / UNFM_P_SIZE)
2227
2228/* the used space within the unformatted node corresponding to pos within the item pointed to by ih */
2229#define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
2230
2231/* number of bytes contained by the direct item or the unformatted nodes the indirect item points to */
2232
2233/* get the item header */
2234#define B_N_PITEM_HEAD(bh,item_num) ( (struct item_head * )((bh)->b_data + BLKH_SIZE) + (item_num) )
2235
2236/* get key */
2237#define B_N_PDELIM_KEY(bh,item_num) ( (struct reiserfs_key * )((bh)->b_data + BLKH_SIZE) + (item_num) )
2238
2239/* get the key */
2240#define B_N_PKEY(bh,item_num) ( &(B_N_PITEM_HEAD(bh,item_num)->ih_key) )
2241
2242/* get item body */
2243#define B_N_PITEM(bh,item_num) ( (bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(item_num))))
2244
2245/* get the stat data by the buffer header and the item order */
2246#define B_N_STAT_DATA(bh,nr) \
2247( (struct stat_data *)((bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(nr))) ) )
2248
2249 /* following defines use reiserfs buffer header and item header */
2250
2251/* get stat-data */
2252#define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
2253
2254// this is 3976 for size==4096
2255#define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
2256
2257/* indirect items consist of entries which contain blocknrs, pos
2258 indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
2259 blocknr contained by the entry pos points to */
2260#define B_I_POS_UNFM_POINTER(bh,ih,pos) le32_to_cpu(*(((unp_t *)B_I_PITEM(bh,ih)) + (pos)))
2261#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)
2262
2263struct reiserfs_iget_args {
2264 __u32 objectid;
2265 __u32 dirid;
2266};
2267
2268/***************************************************************************/
2269/* FUNCTION DECLARATIONS */
2270/***************************************************************************/
2271
2272#define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
2273
2274#define journal_trans_half(blocksize) \
2275 ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
2276
2277/* journal.c see journal.c for all the comments here */
2278
2279/* first block written in a commit. */
2280struct reiserfs_journal_desc {
2281 __le32 j_trans_id; /* id of commit */
2282 __le32 j_len; /* length of commit. len +1 is the commit block */
2283 __le32 j_mount_id; /* mount id of this trans */
2284 __le32 j_realblock[1]; /* real locations for each block */
2285};
2286
2287#define get_desc_trans_id(d) le32_to_cpu((d)->j_trans_id)
2288#define get_desc_trans_len(d) le32_to_cpu((d)->j_len)
2289#define get_desc_mount_id(d) le32_to_cpu((d)->j_mount_id)
2290
2291#define set_desc_trans_id(d,val) do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
2292#define set_desc_trans_len(d,val) do { (d)->j_len = cpu_to_le32 (val); } while (0)
2293#define set_desc_mount_id(d,val) do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
2294
2295/* last block written in a commit */
2296struct reiserfs_journal_commit {
2297 __le32 j_trans_id; /* must match j_trans_id from the desc block */
2298 __le32 j_len; /* ditto */
2299 __le32 j_realblock[1]; /* real locations for each block */
2300};
2301
2302#define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
2303#define get_commit_trans_len(c) le32_to_cpu((c)->j_len)
2304#define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
2305
2306#define set_commit_trans_id(c,val) do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
2307#define set_commit_trans_len(c,val) do { (c)->j_len = cpu_to_le32 (val); } while (0)
2308
2309/* this header block gets written whenever a transaction is considered fully flushed, and is more recent than the
2310** last fully flushed transaction. fully flushed means all the log blocks and all the real blocks are on disk,
2311** and this transaction does not need to be replayed.
2312*/
2313struct reiserfs_journal_header {
2314 __le32 j_last_flush_trans_id; /* id of last fully flushed transaction */
2315 __le32 j_first_unflushed_offset; /* offset in the log of where to start replay after a crash */
2316 __le32 j_mount_id;
2317 /* 12 */ struct journal_params jh_journal;
2318};
2319
2320/* biggest tunable defines are right here */
2321#define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */
2322#define JOURNAL_TRANS_MAX_DEFAULT 1024 /* biggest possible single transaction, don't change for now (8/3/99) */
2323#define JOURNAL_TRANS_MIN_DEFAULT 256
2324#define JOURNAL_MAX_BATCH_DEFAULT 900 /* max blocks to batch into one transaction, don't make this any bigger than 900 */
2325#define JOURNAL_MIN_RATIO 2
2326#define JOURNAL_MAX_COMMIT_AGE 30
2327#define JOURNAL_MAX_TRANS_AGE 30
2328#define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
2329#define JOURNAL_BLOCKS_PER_OBJECT(sb) (JOURNAL_PER_BALANCE_CNT * 3 + \
2330 2 * (REISERFS_QUOTA_INIT_BLOCKS(sb) + \
2331 REISERFS_QUOTA_TRANS_BLOCKS(sb)))
2332
2333#ifdef CONFIG_QUOTA
2334#define REISERFS_QUOTA_OPTS ((1 << REISERFS_USRQUOTA) | (1 << REISERFS_GRPQUOTA))
2335/* We need to update data and inode (atime) */
2336#define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? 2 : 0)
2337/* 1 balancing, 1 bitmap, 1 data per write + stat data update */
2338#define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
2339(DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0)
2340/* same as with INIT */
2341#define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
2342(DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0)
2343#else
2344#define REISERFS_QUOTA_TRANS_BLOCKS(s) 0
2345#define REISERFS_QUOTA_INIT_BLOCKS(s) 0
2346#define REISERFS_QUOTA_DEL_BLOCKS(s) 0
2347#endif
2348
2349/* both of these can be as low as 1, or as high as you want. The min is the
2350** number of 4k bitmap nodes preallocated on mount. New nodes are allocated
2351** as needed, and released when transactions are committed. On release, if
2352** the current number of nodes is > max, the node is freed, otherwise,
2353** it is put on a free list for faster use later.
2354*/
2355#define REISERFS_MIN_BITMAP_NODES 10
2356#define REISERFS_MAX_BITMAP_NODES 100
2357
2358#define JBH_HASH_SHIFT 13 /* these are based on journal hash size of 8192 */
2359#define JBH_HASH_MASK 8191
2360
2361#define _jhashfn(sb,block) \
2362 (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
2363 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
2364#define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
2365
2366// We need these to make journal.c code more readable
2367#define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2368#define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2369#define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2370
2371enum reiserfs_bh_state_bits {
2372 BH_JDirty = BH_PrivateStart, /* buffer is in current transaction */
2373 BH_JDirty_wait,
2374 BH_JNew, /* disk block was taken off free list before
2375 * being in a finished transaction, or
2376 * written to disk. Can be reused immed. */
2377 BH_JPrepared,
2378 BH_JRestore_dirty,
2379 BH_JTest, // debugging only will go away
2380};
2381
2382BUFFER_FNS(JDirty, journaled);
2383TAS_BUFFER_FNS(JDirty, journaled);
2384BUFFER_FNS(JDirty_wait, journal_dirty);
2385TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
2386BUFFER_FNS(JNew, journal_new);
2387TAS_BUFFER_FNS(JNew, journal_new);
2388BUFFER_FNS(JPrepared, journal_prepared);
2389TAS_BUFFER_FNS(JPrepared, journal_prepared);
2390BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
2391TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
2392BUFFER_FNS(JTest, journal_test);
2393TAS_BUFFER_FNS(JTest, journal_test);
2394
2395/*
2396** transaction handle which is passed around for all journal calls
2397*/
2398struct reiserfs_transaction_handle {
2399 struct super_block *t_super; /* super for this FS when journal_begin was
2400 called. saves calls to reiserfs_get_super
2401 also used by nested transactions to make
2402 sure they are nesting on the right FS
2403 _must_ be first in the handle
2404 */
2405 int t_refcount;
2406 int t_blocks_logged; /* number of blocks this writer has logged */
2407 int t_blocks_allocated; /* number of blocks this writer allocated */
2408 unsigned int t_trans_id; /* sanity check, equals the current trans id */
2409 void *t_handle_save; /* save existing current->journal_info */
2410 unsigned displace_new_blocks:1; /* if new block allocation occurres, that block
2411 should be displaced from others */
2412 struct list_head t_list;
2413};
2414
2415/* used to keep track of ordered and tail writes, attached to the buffer
2416 * head through b_journal_head.
2417 */
2418struct reiserfs_jh {
2419 struct reiserfs_journal_list *jl;
2420 struct buffer_head *bh;
2421 struct list_head list;
2422};
2423
2424void reiserfs_free_jh(struct buffer_head *bh);
2425int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
2426int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
2427int journal_mark_dirty(struct reiserfs_transaction_handle *,
2428 struct super_block *, struct buffer_head *bh);
2429
2430static inline int reiserfs_file_data_log(struct inode *inode)
2431{
2432 if (reiserfs_data_log(inode->i_sb) ||
2433 (REISERFS_I(inode)->i_flags & i_data_log))
2434 return 1;
2435 return 0;
2436}
2437
2438static inline int reiserfs_transaction_running(struct super_block *s)
2439{
2440 struct reiserfs_transaction_handle *th = current->journal_info;
2441 if (th && th->t_super == s)
2442 return 1;
2443 if (th && th->t_super == NULL)
2444 BUG();
2445 return 0;
2446}
2447
2448static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th)
2449{
2450 return th->t_blocks_allocated - th->t_blocks_logged;
2451}
2452
2453struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
2454 super_block
2455 *,
2456 int count);
2457int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
Marco Stornellicfac4b42012-12-15 11:47:31 +01002458void reiserfs_vfs_truncate_file(struct inode *inode);
Al Virof466c6f2012-03-17 01:16:43 -04002459int reiserfs_commit_page(struct inode *inode, struct page *page,
2460 unsigned from, unsigned to);
Artem Bityutskiy25729b02012-06-01 17:18:05 +03002461void reiserfs_flush_old_commits(struct super_block *);
Al Virof466c6f2012-03-17 01:16:43 -04002462int reiserfs_commit_for_inode(struct inode *);
2463int reiserfs_inode_needs_commit(struct inode *);
2464void reiserfs_update_inode_transaction(struct inode *);
2465void reiserfs_wait_on_write_block(struct super_block *s);
2466void reiserfs_block_writes(struct reiserfs_transaction_handle *th);
2467void reiserfs_allow_writes(struct super_block *s);
2468void reiserfs_check_lock_depth(struct super_block *s, char *caller);
2469int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh,
2470 int wait);
2471void reiserfs_restore_prepared_buffer(struct super_block *,
2472 struct buffer_head *bh);
2473int journal_init(struct super_block *, const char *j_dev_name, int old_format,
2474 unsigned int);
2475int journal_release(struct reiserfs_transaction_handle *, struct super_block *);
2476int journal_release_error(struct reiserfs_transaction_handle *,
2477 struct super_block *);
2478int journal_end(struct reiserfs_transaction_handle *, struct super_block *,
2479 unsigned long);
2480int journal_end_sync(struct reiserfs_transaction_handle *, struct super_block *,
2481 unsigned long);
2482int journal_mark_freed(struct reiserfs_transaction_handle *,
2483 struct super_block *, b_blocknr_t blocknr);
2484int journal_transaction_should_end(struct reiserfs_transaction_handle *, int);
2485int reiserfs_in_journal(struct super_block *sb, unsigned int bmap_nr,
2486 int bit_nr, int searchall, b_blocknr_t *next);
2487int journal_begin(struct reiserfs_transaction_handle *,
2488 struct super_block *sb, unsigned long);
2489int journal_join_abort(struct reiserfs_transaction_handle *,
2490 struct super_block *sb, unsigned long);
2491void reiserfs_abort_journal(struct super_block *sb, int errno);
2492void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...);
2493int reiserfs_allocate_list_bitmaps(struct super_block *s,
2494 struct reiserfs_list_bitmap *, unsigned int);
2495
Artem Bityutskiy033369d2012-06-01 17:18:08 +03002496void reiserfs_schedule_old_flush(struct super_block *s);
Al Virof466c6f2012-03-17 01:16:43 -04002497void add_save_link(struct reiserfs_transaction_handle *th,
2498 struct inode *inode, int truncate);
2499int remove_save_link(struct inode *inode, int truncate);
2500
2501/* objectid.c */
2502__u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th);
2503void reiserfs_release_objectid(struct reiserfs_transaction_handle *th,
2504 __u32 objectid_to_release);
2505int reiserfs_convert_objectid_map_v1(struct super_block *);
2506
2507/* stree.c */
2508int B_IS_IN_TREE(const struct buffer_head *);
2509extern void copy_item_head(struct item_head *to,
2510 const struct item_head *from);
2511
2512// first key is in cpu form, second - le
2513extern int comp_short_keys(const struct reiserfs_key *le_key,
2514 const struct cpu_key *cpu_key);
2515extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from);
2516
2517// both are in le form
2518extern int comp_le_keys(const struct reiserfs_key *,
2519 const struct reiserfs_key *);
2520extern int comp_short_le_keys(const struct reiserfs_key *,
2521 const struct reiserfs_key *);
2522
2523//
2524// get key version from on disk key - kludge
2525//
2526static inline int le_key_version(const struct reiserfs_key *key)
2527{
2528 int type;
2529
2530 type = offset_v2_k_type(&(key->u.k_offset_v2));
2531 if (type != TYPE_DIRECT && type != TYPE_INDIRECT
2532 && type != TYPE_DIRENTRY)
2533 return KEY_FORMAT_3_5;
2534
2535 return KEY_FORMAT_3_6;
2536
2537}
2538
2539static inline void copy_key(struct reiserfs_key *to,
2540 const struct reiserfs_key *from)
2541{
2542 memcpy(to, from, KEY_SIZE);
2543}
2544
2545int comp_items(const struct item_head *stored_ih, const struct treepath *path);
2546const struct reiserfs_key *get_rkey(const struct treepath *chk_path,
2547 const struct super_block *sb);
2548int search_by_key(struct super_block *, const struct cpu_key *,
2549 struct treepath *, int);
2550#define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
2551int search_for_position_by_key(struct super_block *sb,
2552 const struct cpu_key *cpu_key,
2553 struct treepath *search_path);
2554extern void decrement_bcount(struct buffer_head *bh);
2555void decrement_counters_in_path(struct treepath *search_path);
2556void pathrelse(struct treepath *search_path);
2557int reiserfs_check_path(struct treepath *p);
2558void pathrelse_and_restore(struct super_block *s, struct treepath *search_path);
2559
2560int reiserfs_insert_item(struct reiserfs_transaction_handle *th,
2561 struct treepath *path,
2562 const struct cpu_key *key,
2563 struct item_head *ih,
2564 struct inode *inode, const char *body);
2565
2566int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th,
2567 struct treepath *path,
2568 const struct cpu_key *key,
2569 struct inode *inode,
2570 const char *body, int paste_size);
2571
2572int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,
2573 struct treepath *path,
2574 struct cpu_key *key,
2575 struct inode *inode,
2576 struct page *page, loff_t new_file_size);
2577
2578int reiserfs_delete_item(struct reiserfs_transaction_handle *th,
2579 struct treepath *path,
2580 const struct cpu_key *key,
2581 struct inode *inode, struct buffer_head *un_bh);
2582
2583void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
2584 struct inode *inode, struct reiserfs_key *key);
2585int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
2586 struct inode *inode);
2587int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,
2588 struct inode *inode, struct page *,
2589 int update_timestamps);
2590
2591#define i_block_size(inode) ((inode)->i_sb->s_blocksize)
2592#define file_size(inode) ((inode)->i_size)
2593#define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
2594
2595#define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
2596!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 )
2597
2598void padd_item(char *item, int total_length, int length);
2599
2600/* inode.c */
2601/* args for the create parameter of reiserfs_get_block */
2602#define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */
2603#define GET_BLOCK_CREATE 1 /* add anything you need to find block */
2604#define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */
2605#define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
2606#define GET_BLOCK_NO_IMUX 8 /* i_mutex is not held, don't preallocate */
2607#define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */
2608
2609void reiserfs_read_locked_inode(struct inode *inode,
2610 struct reiserfs_iget_args *args);
2611int reiserfs_find_actor(struct inode *inode, void *p);
2612int reiserfs_init_locked_inode(struct inode *inode, void *p);
2613void reiserfs_evict_inode(struct inode *inode);
2614int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc);
2615int reiserfs_get_block(struct inode *inode, sector_t block,
2616 struct buffer_head *bh_result, int create);
2617struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2618 int fh_len, int fh_type);
2619struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
2620 int fh_len, int fh_type);
Al Virob0b03822012-04-02 14:34:06 -04002621int reiserfs_encode_fh(struct inode *inode, __u32 * data, int *lenp,
2622 struct inode *parent);
Al Virof466c6f2012-03-17 01:16:43 -04002623
2624int reiserfs_truncate_file(struct inode *, int update_timestamps);
2625void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset,
2626 int type, int key_length);
2627void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
2628 int version,
2629 loff_t offset, int type, int length, int entry_count);
2630struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key);
2631
2632struct reiserfs_security_handle;
2633int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
2634 struct inode *dir, umode_t mode,
2635 const char *symname, loff_t i_size,
2636 struct dentry *dentry, struct inode *inode,
2637 struct reiserfs_security_handle *security);
2638
2639void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
2640 struct inode *inode, loff_t size);
2641
2642static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
2643 struct inode *inode)
2644{
2645 reiserfs_update_sd_size(th, inode, inode->i_size);
2646}
2647
2648void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode);
2649void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs);
2650int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
2651
2652int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len);
2653
2654/* namei.c */
2655void set_de_name_and_namelen(struct reiserfs_dir_entry *de);
2656int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,
2657 struct treepath *path, struct reiserfs_dir_entry *de);
2658struct dentry *reiserfs_get_parent(struct dentry *);
2659
2660#ifdef CONFIG_REISERFS_PROC_INFO
2661int reiserfs_proc_info_init(struct super_block *sb);
2662int reiserfs_proc_info_done(struct super_block *sb);
2663int reiserfs_proc_info_global_init(void);
2664int reiserfs_proc_info_global_done(void);
2665
2666#define PROC_EXP( e ) e
2667
2668#define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
2669#define PROC_INFO_MAX( sb, field, value ) \
2670 __PINFO( sb ).field = \
2671 max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
2672#define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
2673#define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
2674#define PROC_INFO_BH_STAT( sb, bh, level ) \
2675 PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \
2676 PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \
2677 PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
2678#else
2679static inline int reiserfs_proc_info_init(struct super_block *sb)
2680{
2681 return 0;
2682}
2683
2684static inline int reiserfs_proc_info_done(struct super_block *sb)
2685{
2686 return 0;
2687}
2688
2689static inline int reiserfs_proc_info_global_init(void)
2690{
2691 return 0;
2692}
2693
2694static inline int reiserfs_proc_info_global_done(void)
2695{
2696 return 0;
2697}
2698
2699#define PROC_EXP( e )
2700#define VOID_V ( ( void ) 0 )
2701#define PROC_INFO_MAX( sb, field, value ) VOID_V
2702#define PROC_INFO_INC( sb, field ) VOID_V
2703#define PROC_INFO_ADD( sb, field, val ) VOID_V
2704#define PROC_INFO_BH_STAT(sb, bh, n_node_level) VOID_V
2705#endif
2706
2707/* dir.c */
2708extern const struct inode_operations reiserfs_dir_inode_operations;
2709extern const struct inode_operations reiserfs_symlink_inode_operations;
2710extern const struct inode_operations reiserfs_special_inode_operations;
2711extern const struct file_operations reiserfs_dir_operations;
2712int reiserfs_readdir_dentry(struct dentry *, void *, filldir_t, loff_t *);
2713
2714/* tail_conversion.c */
2715int direct2indirect(struct reiserfs_transaction_handle *, struct inode *,
2716 struct treepath *, struct buffer_head *, loff_t);
2717int indirect2direct(struct reiserfs_transaction_handle *, struct inode *,
2718 struct page *, struct treepath *, const struct cpu_key *,
2719 loff_t, char *);
2720void reiserfs_unmap_buffer(struct buffer_head *);
2721
2722/* file.c */
2723extern const struct inode_operations reiserfs_file_inode_operations;
2724extern const struct file_operations reiserfs_file_operations;
2725extern const struct address_space_operations reiserfs_address_space_operations;
2726
2727/* fix_nodes.c */
2728
2729int fix_nodes(int n_op_mode, struct tree_balance *tb,
2730 struct item_head *ins_ih, const void *);
2731void unfix_nodes(struct tree_balance *);
2732
2733/* prints.c */
2734void __reiserfs_panic(struct super_block *s, const char *id,
2735 const char *function, const char *fmt, ...)
2736 __attribute__ ((noreturn));
2737#define reiserfs_panic(s, id, fmt, args...) \
2738 __reiserfs_panic(s, id, __func__, fmt, ##args)
2739void __reiserfs_error(struct super_block *s, const char *id,
2740 const char *function, const char *fmt, ...);
2741#define reiserfs_error(s, id, fmt, args...) \
2742 __reiserfs_error(s, id, __func__, fmt, ##args)
2743void reiserfs_info(struct super_block *s, const char *fmt, ...);
2744void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...);
2745void print_indirect_item(struct buffer_head *bh, int item_num);
2746void store_print_tb(struct tree_balance *tb);
2747void print_cur_tb(char *mes);
2748void print_de(struct reiserfs_dir_entry *de);
2749void print_bi(struct buffer_info *bi, char *mes);
2750#define PRINT_LEAF_ITEMS 1 /* print all items */
2751#define PRINT_DIRECTORY_ITEMS 2 /* print directory items */
2752#define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */
2753void print_block(struct buffer_head *bh, ...);
2754void print_bmap(struct super_block *s, int silent);
2755void print_bmap_block(int i, char *data, int size, int silent);
2756/*void print_super_block (struct super_block * s, char * mes);*/
2757void print_objectid_map(struct super_block *s);
2758void print_block_head(struct buffer_head *bh, char *mes);
2759void check_leaf(struct buffer_head *bh);
2760void check_internal(struct buffer_head *bh);
2761void print_statistics(struct super_block *s);
2762char *reiserfs_hashname(int code);
2763
2764/* lbalance.c */
2765int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num,
2766 int mov_bytes, struct buffer_head *Snew);
2767int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes);
2768int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes);
2769void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first,
2770 int del_num, int del_bytes);
2771void leaf_insert_into_buf(struct buffer_info *bi, int before,
2772 struct item_head *inserted_item_ih,
2773 const char *inserted_item_body, int zeros_number);
2774void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num,
2775 int pos_in_item, int paste_size, const char *body,
2776 int zeros_number);
2777void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num,
2778 int pos_in_item, int cut_size);
2779void leaf_paste_entries(struct buffer_info *bi, int item_num, int before,
2780 int new_entry_count, struct reiserfs_de_head *new_dehs,
2781 const char *records, int paste_size);
2782/* ibalance.c */
2783int balance_internal(struct tree_balance *, int, int, struct item_head *,
2784 struct buffer_head **);
2785
2786/* do_balance.c */
2787void do_balance_mark_leaf_dirty(struct tree_balance *tb,
2788 struct buffer_head *bh, int flag);
2789#define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
2790#define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
2791
2792void do_balance(struct tree_balance *tb, struct item_head *ih,
2793 const char *body, int flag);
2794void reiserfs_invalidate_buffer(struct tree_balance *tb,
2795 struct buffer_head *bh);
2796
2797int get_left_neighbor_position(struct tree_balance *tb, int h);
2798int get_right_neighbor_position(struct tree_balance *tb, int h);
2799void replace_key(struct tree_balance *tb, struct buffer_head *, int,
2800 struct buffer_head *, int);
2801void make_empty_node(struct buffer_info *);
2802struct buffer_head *get_FEB(struct tree_balance *);
2803
2804/* bitmap.c */
2805
2806/* structure contains hints for block allocator, and it is a container for
2807 * arguments, such as node, search path, transaction_handle, etc. */
2808struct __reiserfs_blocknr_hint {
2809 struct inode *inode; /* inode passed to allocator, if we allocate unf. nodes */
2810 sector_t block; /* file offset, in blocks */
2811 struct in_core_key key;
2812 struct treepath *path; /* search path, used by allocator to deternine search_start by
2813 * various ways */
2814 struct reiserfs_transaction_handle *th; /* transaction handle is needed to log super blocks and
2815 * bitmap blocks changes */
2816 b_blocknr_t beg, end;
2817 b_blocknr_t search_start; /* a field used to transfer search start value (block number)
2818 * between different block allocator procedures
2819 * (determine_search_start() and others) */
2820 int prealloc_size; /* is set in determine_prealloc_size() function, used by underlayed
2821 * function that do actual allocation */
2822
2823 unsigned formatted_node:1; /* the allocator uses different polices for getting disk space for
2824 * formatted/unformatted blocks with/without preallocation */
2825 unsigned preallocate:1;
2826};
2827
2828typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
2829
2830int reiserfs_parse_alloc_options(struct super_block *, char *);
2831void reiserfs_init_alloc_options(struct super_block *s);
2832
2833/*
2834 * given a directory, this will tell you what packing locality
2835 * to use for a new object underneat it. The locality is returned
2836 * in disk byte order (le).
2837 */
2838__le32 reiserfs_choose_packing(struct inode *dir);
2839
2840int reiserfs_init_bitmap_cache(struct super_block *sb);
2841void reiserfs_free_bitmap_cache(struct super_block *sb);
2842void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info);
2843struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap);
2844int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value);
2845void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *,
2846 b_blocknr_t, int for_unformatted);
2847int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int,
2848 int);
2849static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb,
2850 b_blocknr_t * new_blocknrs,
2851 int amount_needed)
2852{
2853 reiserfs_blocknr_hint_t hint = {
2854 .th = tb->transaction_handle,
2855 .path = tb->tb_path,
2856 .inode = NULL,
2857 .key = tb->key,
2858 .block = 0,
2859 .formatted_node = 1
2860 };
2861 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed,
2862 0);
2863}
2864
2865static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle
2866 *th, struct inode *inode,
2867 b_blocknr_t * new_blocknrs,
2868 struct treepath *path,
2869 sector_t block)
2870{
2871 reiserfs_blocknr_hint_t hint = {
2872 .th = th,
2873 .path = path,
2874 .inode = inode,
2875 .block = block,
2876 .formatted_node = 0,
2877 .preallocate = 0
2878 };
2879 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
2880}
2881
2882#ifdef REISERFS_PREALLOCATE
2883static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle
2884 *th, struct inode *inode,
2885 b_blocknr_t * new_blocknrs,
2886 struct treepath *path,
2887 sector_t block)
2888{
2889 reiserfs_blocknr_hint_t hint = {
2890 .th = th,
2891 .path = path,
2892 .inode = inode,
2893 .block = block,
2894 .formatted_node = 0,
2895 .preallocate = 1
2896 };
2897 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
2898}
2899
2900void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
2901 struct inode *inode);
2902void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th);
2903#endif
2904
2905/* hashes.c */
2906__u32 keyed_hash(const signed char *msg, int len);
2907__u32 yura_hash(const signed char *msg, int len);
2908__u32 r5_hash(const signed char *msg, int len);
2909
2910#define reiserfs_set_le_bit __set_bit_le
2911#define reiserfs_test_and_set_le_bit __test_and_set_bit_le
2912#define reiserfs_clear_le_bit __clear_bit_le
2913#define reiserfs_test_and_clear_le_bit __test_and_clear_bit_le
2914#define reiserfs_test_le_bit test_bit_le
2915#define reiserfs_find_next_zero_le_bit find_next_zero_bit_le
2916
2917/* sometimes reiserfs_truncate may require to allocate few new blocks
2918 to perform indirect2direct conversion. People probably used to
2919 think, that truncate should work without problems on a filesystem
2920 without free disk space. They may complain that they can not
2921 truncate due to lack of free disk space. This spare space allows us
2922 to not worry about it. 500 is probably too much, but it should be
2923 absolutely safe */
2924#define SPARE_SPACE 500
2925
2926/* prototypes from ioctl.c */
2927long reiserfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg);
2928long reiserfs_compat_ioctl(struct file *filp,
2929 unsigned int cmd, unsigned long arg);
2930int reiserfs_unpack(struct inode *inode, struct file *filp);