blob: 1894d96ccb7c936b4cbe753564c01d3589a125a7 [file] [log] [blame]
Al Virof466c6f2012-03-17 01:16:43 -04001/*
Jeff Mahoney098297b2014-04-23 10:00:36 -04002 * Copyright 1996, 1997, 1998 Hans Reiser, see reiserfs/README for
3 * licensing and copyright details
Al Virof466c6f2012-03-17 01:16:43 -04004 */
5
6#include <linux/reiserfs_fs.h>
7
8#include <linux/slab.h>
9#include <linux/interrupt.h>
10#include <linux/sched.h>
Linus Torvaldsed2d2652012-03-24 10:08:39 -070011#include <linux/bug.h>
Al Virof466c6f2012-03-17 01:16:43 -040012#include <linux/workqueue.h>
13#include <asm/unaligned.h>
14#include <linux/bitops.h>
15#include <linux/proc_fs.h>
16#include <linux/buffer_head.h>
Al Virof466c6f2012-03-17 01:16:43 -040017
18/* the 32 bit compat definitions with int argument */
19#define REISERFS_IOC32_UNPACK _IOW(0xCD, 1, int)
20#define REISERFS_IOC32_GETFLAGS FS_IOC32_GETFLAGS
21#define REISERFS_IOC32_SETFLAGS FS_IOC32_SETFLAGS
22#define REISERFS_IOC32_GETVERSION FS_IOC32_GETVERSION
23#define REISERFS_IOC32_SETVERSION FS_IOC32_SETVERSION
24
Al Viro765fd6b2012-03-17 01:19:24 -040025struct reiserfs_journal_list;
26
Jeff Mahoney098297b2014-04-23 10:00:36 -040027/* bitmasks for i_flags field in reiserfs-specific part of inode */
Al Viro765fd6b2012-03-17 01:19:24 -040028typedef enum {
Jeff Mahoney098297b2014-04-23 10:00:36 -040029 /*
30 * this says what format of key do all items (but stat data) of
31 * an object have. If this is set, that format is 3.6 otherwise - 3.5
32 */
Al Viro765fd6b2012-03-17 01:19:24 -040033 i_item_key_version_mask = 0x0001,
Jeff Mahoney098297b2014-04-23 10:00:36 -040034
35 /*
36 * If this is unset, object has 3.5 stat data, otherwise,
37 * it has 3.6 stat data with 64bit size, 32bit nlink etc.
38 */
Al Viro765fd6b2012-03-17 01:19:24 -040039 i_stat_data_version_mask = 0x0002,
Jeff Mahoney098297b2014-04-23 10:00:36 -040040
41 /* file might need tail packing on close */
Al Viro765fd6b2012-03-17 01:19:24 -040042 i_pack_on_close_mask = 0x0004,
Jeff Mahoney098297b2014-04-23 10:00:36 -040043
44 /* don't pack tail of file */
Al Viro765fd6b2012-03-17 01:19:24 -040045 i_nopack_mask = 0x0008,
Jeff Mahoney098297b2014-04-23 10:00:36 -040046
47 /*
48 * If either of these are set, "safe link" was created for this
49 * file during truncate or unlink. Safe link is used to avoid
50 * leakage of disk space on crash with some files open, but unlinked.
51 */
Al Viro765fd6b2012-03-17 01:19:24 -040052 i_link_saved_unlink_mask = 0x0010,
53 i_link_saved_truncate_mask = 0x0020,
Jeff Mahoney098297b2014-04-23 10:00:36 -040054
Al Viro765fd6b2012-03-17 01:19:24 -040055 i_has_xattr_dir = 0x0040,
56 i_data_log = 0x0080,
57} reiserfs_inode_flags;
58
59struct reiserfs_inode_info {
60 __u32 i_key[4]; /* key is still 4 32 bit integers */
Jeff Mahoney098297b2014-04-23 10:00:36 -040061
62 /*
63 * transient inode flags that are never stored on disk. Bitmasks
64 * for this field are defined above.
65 */
Al Viro765fd6b2012-03-17 01:19:24 -040066 __u32 i_flags;
67
Jeff Mahoney098297b2014-04-23 10:00:36 -040068 /* offset of first byte stored in direct item. */
69 __u32 i_first_direct_byte;
Al Viro765fd6b2012-03-17 01:19:24 -040070
71 /* copy of persistent inode flags read from sd_attrs. */
72 __u32 i_attrs;
73
Jeff Mahoney098297b2014-04-23 10:00:36 -040074 /* first unused block of a sequence of unused blocks */
75 int i_prealloc_block;
Al Viro765fd6b2012-03-17 01:19:24 -040076 int i_prealloc_count; /* length of that sequence */
Al Viro765fd6b2012-03-17 01:19:24 -040077
Jeff Mahoney098297b2014-04-23 10:00:36 -040078 /* per-transaction list of inodes which have preallocated blocks */
79 struct list_head i_prealloc_list;
Al Viro765fd6b2012-03-17 01:19:24 -040080
Jeff Mahoney098297b2014-04-23 10:00:36 -040081 /*
82 * new_packing_locality is created; new blocks for the contents
83 * of this directory should be displaced
84 */
85 unsigned new_packing_locality:1;
86
87 /*
88 * we use these for fsync or O_SYNC to decide which transaction
89 * needs to be committed in order for this inode to be properly
90 * flushed
91 */
Al Viro765fd6b2012-03-17 01:19:24 -040092 unsigned int i_trans_id;
Jeff Mahoney098297b2014-04-23 10:00:36 -040093
Al Viro765fd6b2012-03-17 01:19:24 -040094 struct reiserfs_journal_list *i_jl;
95 atomic_t openers;
96 struct mutex tailpack;
97#ifdef CONFIG_REISERFS_FS_XATTR
98 struct rw_semaphore i_xattr_sem;
99#endif
100 struct inode vfs_inode;
101};
102
103typedef enum {
104 reiserfs_attrs_cleared = 0x00000001,
105} reiserfs_super_block_flags;
106
Jeff Mahoney098297b2014-04-23 10:00:36 -0400107/*
108 * struct reiserfs_super_block accessors/mutators since this is a disk
109 * structure, it will always be in little endian format.
110 */
Al Viro765fd6b2012-03-17 01:19:24 -0400111#define sb_block_count(sbp) (le32_to_cpu((sbp)->s_v1.s_block_count))
112#define set_sb_block_count(sbp,v) ((sbp)->s_v1.s_block_count = cpu_to_le32(v))
113#define sb_free_blocks(sbp) (le32_to_cpu((sbp)->s_v1.s_free_blocks))
114#define set_sb_free_blocks(sbp,v) ((sbp)->s_v1.s_free_blocks = cpu_to_le32(v))
115#define sb_root_block(sbp) (le32_to_cpu((sbp)->s_v1.s_root_block))
116#define set_sb_root_block(sbp,v) ((sbp)->s_v1.s_root_block = cpu_to_le32(v))
117
118#define sb_jp_journal_1st_block(sbp) \
119 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_1st_block))
120#define set_sb_jp_journal_1st_block(sbp,v) \
121 ((sbp)->s_v1.s_journal.jp_journal_1st_block = cpu_to_le32(v))
122#define sb_jp_journal_dev(sbp) \
123 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_dev))
124#define set_sb_jp_journal_dev(sbp,v) \
125 ((sbp)->s_v1.s_journal.jp_journal_dev = cpu_to_le32(v))
126#define sb_jp_journal_size(sbp) \
127 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_size))
128#define set_sb_jp_journal_size(sbp,v) \
129 ((sbp)->s_v1.s_journal.jp_journal_size = cpu_to_le32(v))
130#define sb_jp_journal_trans_max(sbp) \
131 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_trans_max))
132#define set_sb_jp_journal_trans_max(sbp,v) \
133 ((sbp)->s_v1.s_journal.jp_journal_trans_max = cpu_to_le32(v))
134#define sb_jp_journal_magic(sbp) \
135 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_magic))
136#define set_sb_jp_journal_magic(sbp,v) \
137 ((sbp)->s_v1.s_journal.jp_journal_magic = cpu_to_le32(v))
138#define sb_jp_journal_max_batch(sbp) \
139 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_batch))
140#define set_sb_jp_journal_max_batch(sbp,v) \
141 ((sbp)->s_v1.s_journal.jp_journal_max_batch = cpu_to_le32(v))
142#define sb_jp_jourmal_max_commit_age(sbp) \
143 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_commit_age))
144#define set_sb_jp_journal_max_commit_age(sbp,v) \
145 ((sbp)->s_v1.s_journal.jp_journal_max_commit_age = cpu_to_le32(v))
146
147#define sb_blocksize(sbp) (le16_to_cpu((sbp)->s_v1.s_blocksize))
148#define set_sb_blocksize(sbp,v) ((sbp)->s_v1.s_blocksize = cpu_to_le16(v))
149#define sb_oid_maxsize(sbp) (le16_to_cpu((sbp)->s_v1.s_oid_maxsize))
150#define set_sb_oid_maxsize(sbp,v) ((sbp)->s_v1.s_oid_maxsize = cpu_to_le16(v))
151#define sb_oid_cursize(sbp) (le16_to_cpu((sbp)->s_v1.s_oid_cursize))
152#define set_sb_oid_cursize(sbp,v) ((sbp)->s_v1.s_oid_cursize = cpu_to_le16(v))
153#define sb_umount_state(sbp) (le16_to_cpu((sbp)->s_v1.s_umount_state))
154#define set_sb_umount_state(sbp,v) ((sbp)->s_v1.s_umount_state = cpu_to_le16(v))
155#define sb_fs_state(sbp) (le16_to_cpu((sbp)->s_v1.s_fs_state))
156#define set_sb_fs_state(sbp,v) ((sbp)->s_v1.s_fs_state = cpu_to_le16(v))
157#define sb_hash_function_code(sbp) \
158 (le32_to_cpu((sbp)->s_v1.s_hash_function_code))
159#define set_sb_hash_function_code(sbp,v) \
160 ((sbp)->s_v1.s_hash_function_code = cpu_to_le32(v))
161#define sb_tree_height(sbp) (le16_to_cpu((sbp)->s_v1.s_tree_height))
162#define set_sb_tree_height(sbp,v) ((sbp)->s_v1.s_tree_height = cpu_to_le16(v))
163#define sb_bmap_nr(sbp) (le16_to_cpu((sbp)->s_v1.s_bmap_nr))
164#define set_sb_bmap_nr(sbp,v) ((sbp)->s_v1.s_bmap_nr = cpu_to_le16(v))
165#define sb_version(sbp) (le16_to_cpu((sbp)->s_v1.s_version))
166#define set_sb_version(sbp,v) ((sbp)->s_v1.s_version = cpu_to_le16(v))
167
168#define sb_mnt_count(sbp) (le16_to_cpu((sbp)->s_mnt_count))
169#define set_sb_mnt_count(sbp, v) ((sbp)->s_mnt_count = cpu_to_le16(v))
170
171#define sb_reserved_for_journal(sbp) \
172 (le16_to_cpu((sbp)->s_v1.s_reserved_for_journal))
173#define set_sb_reserved_for_journal(sbp,v) \
174 ((sbp)->s_v1.s_reserved_for_journal = cpu_to_le16(v))
175
176/* LOGGING -- */
177
Jeff Mahoney098297b2014-04-23 10:00:36 -0400178/*
179 * These all interelate for performance.
180 *
181 * If the journal block count is smaller than n transactions, you lose speed.
182 * I don't know what n is yet, I'm guessing 8-16.
183 *
184 * typical transaction size depends on the application, how often fsync is
185 * called, and how many metadata blocks you dirty in a 30 second period.
186 * The more small files (<16k) you use, the larger your transactions will
187 * be.
188 *
189 * If your journal fills faster than dirty buffers get flushed to disk, it
190 * must flush them before allowing the journal to wrap, which slows things
191 * down. If you need high speed meta data updates, the journal should be
192 * big enough to prevent wrapping before dirty meta blocks get to disk.
193 *
194 * If the batch max is smaller than the transaction max, you'll waste space
195 * at the end of the journal because journal_end sets the next transaction
196 * to start at 0 if the next transaction has any chance of wrapping.
197 *
198 * The large the batch max age, the better the speed, and the more meta
199 * data changes you'll lose after a crash.
200 */
Al Viro765fd6b2012-03-17 01:19:24 -0400201
202/* don't mess with these for a while */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400203/* we have a node size define somewhere in reiserfs_fs.h. -Hans */
Al Viro765fd6b2012-03-17 01:19:24 -0400204#define JOURNAL_BLOCK_SIZE 4096 /* BUG gotta get rid of this */
205#define JOURNAL_MAX_CNODE 1500 /* max cnodes to allocate. */
206#define JOURNAL_HASH_SIZE 8192
Al Viro765fd6b2012-03-17 01:19:24 -0400207
Jeff Mahoney098297b2014-04-23 10:00:36 -0400208/* number of copies of the bitmaps to have floating. Must be >= 2 */
209#define JOURNAL_NUM_BITMAPS 5
210
211/*
212 * One of these for every block in every transaction
213 * Each one is in two hash tables. First, a hash of the current transaction,
214 * and after journal_end, a hash of all the in memory transactions.
215 * next and prev are used by the current transaction (journal_hash).
216 * hnext and hprev are used by journal_list_hash. If a block is in more
217 * than one transaction, the journal_list_hash links it in multiple times.
218 * This allows flush_journal_list to remove just the cnode belonging to a
219 * given transaction.
220 */
Al Viro765fd6b2012-03-17 01:19:24 -0400221struct reiserfs_journal_cnode {
222 struct buffer_head *bh; /* real buffer head */
223 struct super_block *sb; /* dev of real buffer head */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400224
225 /* block number of real buffer head, == 0 when buffer on disk */
226 __u32 blocknr;
227
Al Viro765fd6b2012-03-17 01:19:24 -0400228 unsigned long state;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400229
230 /* journal list this cnode lives in */
231 struct reiserfs_journal_list *jlist;
232
Al Viro765fd6b2012-03-17 01:19:24 -0400233 struct reiserfs_journal_cnode *next; /* next in transaction list */
234 struct reiserfs_journal_cnode *prev; /* prev in transaction list */
235 struct reiserfs_journal_cnode *hprev; /* prev in hash list */
236 struct reiserfs_journal_cnode *hnext; /* next in hash list */
237};
238
239struct reiserfs_bitmap_node {
240 int id;
241 char *data;
242 struct list_head list;
243};
244
245struct reiserfs_list_bitmap {
246 struct reiserfs_journal_list *journal_list;
247 struct reiserfs_bitmap_node **bitmaps;
248};
249
250/*
Jeff Mahoney098297b2014-04-23 10:00:36 -0400251 * one of these for each transaction. The most important part here is the
252 * j_realblock. this list of cnodes is used to hash all the blocks in all
253 * the commits, to mark all the real buffer heads dirty once all the commits
254 * hit the disk, and to make sure every real block in a transaction is on
255 * disk before allowing the log area to be overwritten
256 */
Al Viro765fd6b2012-03-17 01:19:24 -0400257struct reiserfs_journal_list {
258 unsigned long j_start;
259 unsigned long j_state;
260 unsigned long j_len;
261 atomic_t j_nonzerolen;
262 atomic_t j_commit_left;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400263
264 /* all commits older than this on disk */
265 atomic_t j_older_commits_done;
266
Al Viro765fd6b2012-03-17 01:19:24 -0400267 struct mutex j_commit_mutex;
268 unsigned int j_trans_id;
269 time_t j_timestamp;
270 struct reiserfs_list_bitmap *j_list_bitmap;
271 struct buffer_head *j_commit_bh; /* commit buffer head */
272 struct reiserfs_journal_cnode *j_realblock;
273 struct reiserfs_journal_cnode *j_freedlist; /* list of buffers that were freed during this trans. free each of these on flush */
274 /* time ordered list of all active transactions */
275 struct list_head j_list;
276
Jeff Mahoney098297b2014-04-23 10:00:36 -0400277 /*
278 * time ordered list of all transactions we haven't tried
279 * to flush yet
280 */
Al Viro765fd6b2012-03-17 01:19:24 -0400281 struct list_head j_working_list;
282
283 /* list of tail conversion targets in need of flush before commit */
284 struct list_head j_tail_bh_list;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400285
Al Viro765fd6b2012-03-17 01:19:24 -0400286 /* list of data=ordered buffers in need of flush before commit */
287 struct list_head j_bh_list;
288 int j_refcount;
289};
290
291struct reiserfs_journal {
292 struct buffer_head **j_ap_blocks; /* journal blocks on disk */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400293 /* newest journal block */
294 struct reiserfs_journal_cnode *j_last;
295
296 /* oldest journal block. start here for traverse */
297 struct reiserfs_journal_cnode *j_first;
Al Viro765fd6b2012-03-17 01:19:24 -0400298
299 struct block_device *j_dev_bd;
300 fmode_t j_dev_mode;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400301
302 /* first block on s_dev of reserved area journal */
303 int j_1st_reserved_block;
Al Viro765fd6b2012-03-17 01:19:24 -0400304
305 unsigned long j_state;
306 unsigned int j_trans_id;
307 unsigned long j_mount_id;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400308
309 /* start of current waiting commit (index into j_ap_blocks) */
310 unsigned long j_start;
Al Viro765fd6b2012-03-17 01:19:24 -0400311 unsigned long j_len; /* length of current waiting commit */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400312
313 /* number of buffers requested by journal_begin() */
314 unsigned long j_len_alloc;
315
Al Viro765fd6b2012-03-17 01:19:24 -0400316 atomic_t j_wcount; /* count of writers for current commit */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400317
318 /* batch count. allows turning X transactions into 1 */
319 unsigned long j_bcount;
320
321 /* first unflushed transactions offset */
322 unsigned long j_first_unflushed_offset;
323
324 /* last fully flushed journal timestamp */
325 unsigned j_last_flush_trans_id;
326
Al Viro765fd6b2012-03-17 01:19:24 -0400327 struct buffer_head *j_header_bh;
328
329 time_t j_trans_start_time; /* time this transaction started */
330 struct mutex j_mutex;
331 struct mutex j_flush_mutex;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400332
333 /* wait for current transaction to finish before starting new one */
334 wait_queue_head_t j_join_wait;
335
336 atomic_t j_jlock; /* lock for j_join_wait */
Al Viro765fd6b2012-03-17 01:19:24 -0400337 int j_list_bitmap_index; /* number of next list bitmap to use */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400338
339 /* no more journal begins allowed. MUST sleep on j_join_wait */
340 int j_must_wait;
341
342 /* next journal_end will flush all journal list */
343 int j_next_full_flush;
344
345 /* next journal_end will flush all async commits */
346 int j_next_async_flush;
Al Viro765fd6b2012-03-17 01:19:24 -0400347
348 int j_cnode_used; /* number of cnodes on the used list */
349 int j_cnode_free; /* number of cnodes on the free list */
350
Jeff Mahoney098297b2014-04-23 10:00:36 -0400351 /* max number of blocks in a transaction. */
352 unsigned int j_trans_max;
353
354 /* max number of blocks to batch into a trans */
355 unsigned int j_max_batch;
356
357 /* in seconds, how old can an async commit be */
358 unsigned int j_max_commit_age;
359
360 /* in seconds, how old can a transaction be */
361 unsigned int j_max_trans_age;
362
363 /* the default for the max commit age */
364 unsigned int j_default_max_commit_age;
Al Viro765fd6b2012-03-17 01:19:24 -0400365
366 struct reiserfs_journal_cnode *j_cnode_free_list;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400367
368 /* orig pointer returned from vmalloc */
369 struct reiserfs_journal_cnode *j_cnode_free_orig;
Al Viro765fd6b2012-03-17 01:19:24 -0400370
371 struct reiserfs_journal_list *j_current_jl;
372 int j_free_bitmap_nodes;
373 int j_used_bitmap_nodes;
374
375 int j_num_lists; /* total number of active transactions */
376 int j_num_work_lists; /* number that need attention from kreiserfsd */
377
378 /* debugging to make sure things are flushed in order */
379 unsigned int j_last_flush_id;
380
381 /* debugging to make sure things are committed in order */
382 unsigned int j_last_commit_id;
383
384 struct list_head j_bitmap_nodes;
385 struct list_head j_dirty_buffers;
386 spinlock_t j_dirty_buffers_lock; /* protects j_dirty_buffers */
387
388 /* list of all active transactions */
389 struct list_head j_journal_list;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400390
Al Viro765fd6b2012-03-17 01:19:24 -0400391 /* lists that haven't been touched by writeback attempts */
392 struct list_head j_working_list;
393
Jeff Mahoney098297b2014-04-23 10:00:36 -0400394 /* hash table for real buffer heads in current trans */
395 struct reiserfs_journal_cnode *j_hash_table[JOURNAL_HASH_SIZE];
396
397 /* hash table for all the real buffer heads in all the transactions */
398 struct reiserfs_journal_cnode *j_list_hash_table[JOURNAL_HASH_SIZE];
399
400 /* array of bitmaps to record the deleted blocks */
401 struct reiserfs_list_bitmap j_list_bitmap[JOURNAL_NUM_BITMAPS];
402
403 /* list of inodes which have preallocated blocks */
404 struct list_head j_prealloc_list;
Al Viro765fd6b2012-03-17 01:19:24 -0400405 int j_persistent_trans;
406 unsigned long j_max_trans_size;
407 unsigned long j_max_batch_size;
408
409 int j_errno;
410
411 /* when flushing ordered buffers, throttle new ordered writers */
412 struct delayed_work j_work;
413 struct super_block *j_work_sb;
414 atomic_t j_async_throttle;
415};
416
417enum journal_state_bits {
418 J_WRITERS_BLOCKED = 1, /* set when new writers not allowed */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400419 J_WRITERS_QUEUED, /* set when log is full due to too many writers */
420 J_ABORTED, /* set when log is aborted */
Al Viro765fd6b2012-03-17 01:19:24 -0400421};
422
Jeff Mahoney098297b2014-04-23 10:00:36 -0400423/* ick. magic string to find desc blocks in the journal */
424#define JOURNAL_DESC_MAGIC "ReIsErLB"
Al Viro765fd6b2012-03-17 01:19:24 -0400425
426typedef __u32(*hashf_t) (const signed char *, int);
427
428struct reiserfs_bitmap_info {
429 __u32 free_count;
430};
431
432struct proc_dir_entry;
433
434#if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )
435typedef unsigned long int stat_cnt_t;
436typedef struct reiserfs_proc_info_data {
437 spinlock_t lock;
438 int exiting;
439 int max_hash_collisions;
440
441 stat_cnt_t breads;
442 stat_cnt_t bread_miss;
443 stat_cnt_t search_by_key;
444 stat_cnt_t search_by_key_fs_changed;
445 stat_cnt_t search_by_key_restarted;
446
447 stat_cnt_t insert_item_restarted;
448 stat_cnt_t paste_into_item_restarted;
449 stat_cnt_t cut_from_item_restarted;
450 stat_cnt_t delete_solid_item_restarted;
451 stat_cnt_t delete_item_restarted;
452
453 stat_cnt_t leaked_oid;
454 stat_cnt_t leaves_removable;
455
Jeff Mahoney098297b2014-04-23 10:00:36 -0400456 /*
457 * balances per level.
458 * Use explicit 5 as MAX_HEIGHT is not visible yet.
459 */
Al Viro765fd6b2012-03-17 01:19:24 -0400460 stat_cnt_t balance_at[5]; /* XXX */
461 /* sbk == search_by_key */
462 stat_cnt_t sbk_read_at[5]; /* XXX */
463 stat_cnt_t sbk_fs_changed[5];
464 stat_cnt_t sbk_restarted[5];
465 stat_cnt_t items_at[5]; /* XXX */
466 stat_cnt_t free_at[5]; /* XXX */
467 stat_cnt_t can_node_be_removed[5]; /* XXX */
468 long int lnum[5]; /* XXX */
469 long int rnum[5]; /* XXX */
470 long int lbytes[5]; /* XXX */
471 long int rbytes[5]; /* XXX */
472 stat_cnt_t get_neighbors[5];
473 stat_cnt_t get_neighbors_restart[5];
474 stat_cnt_t need_l_neighbor[5];
475 stat_cnt_t need_r_neighbor[5];
476
477 stat_cnt_t free_block;
478 struct __scan_bitmap_stats {
479 stat_cnt_t call;
480 stat_cnt_t wait;
481 stat_cnt_t bmap;
482 stat_cnt_t retry;
483 stat_cnt_t in_journal_hint;
484 stat_cnt_t in_journal_nohint;
485 stat_cnt_t stolen;
486 } scan_bitmap;
487 struct __journal_stats {
488 stat_cnt_t in_journal;
489 stat_cnt_t in_journal_bitmap;
490 stat_cnt_t in_journal_reusable;
491 stat_cnt_t lock_journal;
492 stat_cnt_t lock_journal_wait;
493 stat_cnt_t journal_being;
494 stat_cnt_t journal_relock_writers;
495 stat_cnt_t journal_relock_wcount;
496 stat_cnt_t mark_dirty;
497 stat_cnt_t mark_dirty_already;
498 stat_cnt_t mark_dirty_notjournal;
499 stat_cnt_t restore_prepared;
500 stat_cnt_t prepare;
501 stat_cnt_t prepare_retry;
502 } journal;
503} reiserfs_proc_info_data_t;
504#else
505typedef struct reiserfs_proc_info_data {
506} reiserfs_proc_info_data_t;
507#endif
508
Jan Karaaca60612014-09-10 20:57:45 +0200509/* Number of quota types we support */
510#define REISERFS_MAXQUOTAS 2
511
Al Viro765fd6b2012-03-17 01:19:24 -0400512/* reiserfs union of in-core super block data */
513struct reiserfs_sb_info {
Jeff Mahoney098297b2014-04-23 10:00:36 -0400514 /* Buffer containing the super block */
515 struct buffer_head *s_sbh;
516
517 /* Pointer to the on-disk super block in the buffer */
518 struct reiserfs_super_block *s_rs;
Al Viro765fd6b2012-03-17 01:19:24 -0400519 struct reiserfs_bitmap_info *s_ap_bitmap;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400520
521 /* pointer to journal information */
522 struct reiserfs_journal *s_journal;
523
Al Viro765fd6b2012-03-17 01:19:24 -0400524 unsigned short s_mount_state; /* reiserfs state (valid, invalid) */
525
526 /* Serialize writers access, replace the old bkl */
527 struct mutex lock;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400528
Al Viro765fd6b2012-03-17 01:19:24 -0400529 /* Owner of the lock (can be recursive) */
530 struct task_struct *lock_owner;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400531
Al Viro765fd6b2012-03-17 01:19:24 -0400532 /* Depth of the lock, start from -1 like the bkl */
533 int lock_depth;
534
Jeff Mahoney797d9012014-04-23 10:00:34 -0400535 struct workqueue_struct *commit_wq;
536
Al Viro765fd6b2012-03-17 01:19:24 -0400537 /* Comment? -Hans */
538 void (*end_io_handler) (struct buffer_head *, int);
Al Viro765fd6b2012-03-17 01:19:24 -0400539
Jeff Mahoney098297b2014-04-23 10:00:36 -0400540 /*
541 * pointer to function which is used to sort names in directory.
542 * Set on mount
543 */
544 hashf_t s_hash_function;
545
546 /* reiserfs's mount options are set here */
547 unsigned long s_mount_opt;
548
549 /* This is a structure that describes block allocator options */
550 struct {
551 /* Bitfield for enable/disable kind of options */
552 unsigned long bits;
553
554 /*
555 * size started from which we consider file
556 * to be a large one (in blocks)
557 */
558 unsigned long large_file_size;
559
Al Viro765fd6b2012-03-17 01:19:24 -0400560 int border; /* percentage of disk, border takes */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400561
562 /*
563 * Minimal file size (in blocks) starting
564 * from which we do preallocations
565 */
566 int preallocmin;
567
568 /*
569 * Number of blocks we try to prealloc when file
570 * reaches preallocmin size (in blocks) or prealloc_list
571 is empty.
572 */
573 int preallocsize;
Al Viro765fd6b2012-03-17 01:19:24 -0400574 } s_alloc_options;
575
576 /* Comment? -Hans */
577 wait_queue_head_t s_wait;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400578 /* increased by one every time the tree gets re-balanced */
579 atomic_t s_generation_counter;
580
581 /* File system properties. Currently holds on-disk FS format */
582 unsigned long s_properties;
Al Viro765fd6b2012-03-17 01:19:24 -0400583
584 /* session statistics */
585 int s_disk_reads;
586 int s_disk_writes;
587 int s_fix_nodes;
588 int s_do_balance;
589 int s_unneeded_left_neighbor;
590 int s_good_search_by_key_reada;
591 int s_bmaps;
592 int s_bmaps_without_search;
593 int s_direct2indirect;
594 int s_indirect2direct;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400595
596 /*
597 * set up when it's ok for reiserfs_read_inode2() to read from
598 * disk inode with nlink==0. Currently this is only used during
599 * finish_unfinished() processing at mount time
600 */
Al Viro765fd6b2012-03-17 01:19:24 -0400601 int s_is_unlinked_ok;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400602
Al Viro765fd6b2012-03-17 01:19:24 -0400603 reiserfs_proc_info_data_t s_proc_info_data;
604 struct proc_dir_entry *procdir;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400605
606 /* amount of blocks reserved for further allocations */
607 int reserved_blocks;
608
609
610 /* this lock on now only used to protect reserved_blocks variable */
611 spinlock_t bitmap_lock;
Al Viro765fd6b2012-03-17 01:19:24 -0400612 struct dentry *priv_root; /* root of /.reiserfs_priv */
613 struct dentry *xattr_root; /* root of /.reiserfs_priv/xattrs */
614 int j_errno;
Artem Bityutskiy033369d2012-06-01 17:18:08 +0300615
616 int work_queued; /* non-zero delayed work is queued */
617 struct delayed_work old_work; /* old transactions flush delayed work */
618 spinlock_t old_work_lock; /* protects old_work and work_queued */
619
Al Viro765fd6b2012-03-17 01:19:24 -0400620#ifdef CONFIG_QUOTA
Jan Karaaca60612014-09-10 20:57:45 +0200621 char *s_qf_names[REISERFS_MAXQUOTAS];
Al Viro765fd6b2012-03-17 01:19:24 -0400622 int s_jquota_fmt;
623#endif
624 char *s_jdev; /* Stored jdev for mount option showing */
625#ifdef CONFIG_REISERFS_CHECK
626
Jeff Mahoney098297b2014-04-23 10:00:36 -0400627 /*
628 * Detects whether more than one copy of tb exists per superblock
629 * as a means of checking whether do_balance is executing
630 * concurrently against another tree reader/writer on a same
631 * mount point.
632 */
633 struct tree_balance *cur_tb;
Al Viro765fd6b2012-03-17 01:19:24 -0400634#endif
635};
636
637/* Definitions of reiserfs on-disk properties: */
638#define REISERFS_3_5 0
639#define REISERFS_3_6 1
640#define REISERFS_OLD_FORMAT 2
641
Al Viro765fd6b2012-03-17 01:19:24 -0400642/* Mount options */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400643enum reiserfs_mount_options {
644 /* large tails will be created in a session */
645 REISERFS_LARGETAIL,
646 /*
647 * small (for files less than block size) tails will
648 * be created in a session
649 */
650 REISERFS_SMALLTAIL,
Al Viro765fd6b2012-03-17 01:19:24 -0400651
Jeff Mahoney098297b2014-04-23 10:00:36 -0400652 /* replay journal and return 0. Use by fsck */
653 REPLAYONLY,
654
655 /*
656 * -o conv: causes conversion of old format super block to the
657 * new format. If not specified - old partition will be dealt
658 * with in a manner of 3.5.x
659 */
660 REISERFS_CONVERT,
661
662 /*
663 * -o hash={tea, rupasov, r5, detect} is meant for properly mounting
664 * reiserfs disks from 3.5.19 or earlier. 99% of the time, this
665 * option is not required. If the normal autodection code can't
666 * determine which hash to use (because both hashes had the same
667 * value for a file) use this option to force a specific hash.
668 * It won't allow you to override the existing hash on the FS, so
669 * if you have a tea hash disk, and mount with -o hash=rupasov,
670 * the mount will fail.
671 */
Al Viro765fd6b2012-03-17 01:19:24 -0400672 FORCE_TEA_HASH, /* try to force tea hash on mount */
673 FORCE_RUPASOV_HASH, /* try to force rupasov hash on mount */
674 FORCE_R5_HASH, /* try to force rupasov hash on mount */
675 FORCE_HASH_DETECT, /* try to detect hash function on mount */
676
677 REISERFS_DATA_LOG,
678 REISERFS_DATA_ORDERED,
679 REISERFS_DATA_WRITEBACK,
680
Jeff Mahoney098297b2014-04-23 10:00:36 -0400681 /*
682 * used for testing experimental features, makes benchmarking new
683 * features with and without more convenient, should never be used by
684 * users in any code shipped to users (ideally)
685 */
Al Viro765fd6b2012-03-17 01:19:24 -0400686
687 REISERFS_NO_BORDER,
688 REISERFS_NO_UNHASHED_RELOCATION,
689 REISERFS_HASHED_RELOCATION,
690 REISERFS_ATTRS,
691 REISERFS_XATTRS_USER,
692 REISERFS_POSIXACL,
693 REISERFS_EXPOSE_PRIVROOT,
694 REISERFS_BARRIER_NONE,
695 REISERFS_BARRIER_FLUSH,
696
697 /* Actions on error */
698 REISERFS_ERROR_PANIC,
699 REISERFS_ERROR_RO,
700 REISERFS_ERROR_CONTINUE,
701
702 REISERFS_USRQUOTA, /* User quota option specified */
703 REISERFS_GRPQUOTA, /* Group quota option specified */
704
705 REISERFS_TEST1,
706 REISERFS_TEST2,
707 REISERFS_TEST3,
708 REISERFS_TEST4,
709 REISERFS_UNSUPPORTED_OPT,
710};
711
712#define reiserfs_r5_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_R5_HASH))
713#define reiserfs_rupasov_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_RUPASOV_HASH))
714#define reiserfs_tea_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_TEA_HASH))
715#define reiserfs_hash_detect(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_HASH_DETECT))
716#define reiserfs_no_border(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_BORDER))
717#define reiserfs_no_unhashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_UNHASHED_RELOCATION))
718#define reiserfs_hashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_HASHED_RELOCATION))
719#define reiserfs_test4(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_TEST4))
720
721#define have_large_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_LARGETAIL))
722#define have_small_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_SMALLTAIL))
723#define replay_only(s) (REISERFS_SB(s)->s_mount_opt & (1 << REPLAYONLY))
724#define reiserfs_attrs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ATTRS))
725#define old_format_only(s) (REISERFS_SB(s)->s_properties & (1 << REISERFS_3_5))
726#define convert_reiserfs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_CONVERT))
727#define reiserfs_data_log(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_LOG))
728#define reiserfs_data_ordered(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_ORDERED))
729#define reiserfs_data_writeback(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_WRITEBACK))
730#define reiserfs_xattrs_user(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_XATTRS_USER))
731#define reiserfs_posixacl(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_POSIXACL))
732#define reiserfs_expose_privroot(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_EXPOSE_PRIVROOT))
733#define reiserfs_xattrs_optional(s) (reiserfs_xattrs_user(s) || reiserfs_posixacl(s))
734#define reiserfs_barrier_none(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_NONE))
735#define reiserfs_barrier_flush(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_FLUSH))
736
737#define reiserfs_error_panic(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_PANIC))
738#define reiserfs_error_ro(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_RO))
739
740void reiserfs_file_buffer(struct buffer_head *bh, int list);
741extern struct file_system_type reiserfs_fs_type;
742int reiserfs_resize(struct super_block *, unsigned long);
743
744#define CARRY_ON 0
745#define SCHEDULE_OCCURRED 1
746
747#define SB_BUFFER_WITH_SB(s) (REISERFS_SB(s)->s_sbh)
748#define SB_JOURNAL(s) (REISERFS_SB(s)->s_journal)
749#define SB_JOURNAL_1st_RESERVED_BLOCK(s) (SB_JOURNAL(s)->j_1st_reserved_block)
750#define SB_JOURNAL_LEN_FREE(s) (SB_JOURNAL(s)->j_journal_len_free)
751#define SB_AP_BITMAP(s) (REISERFS_SB(s)->s_ap_bitmap)
752
753#define SB_DISK_JOURNAL_HEAD(s) (SB_JOURNAL(s)->j_header_bh->)
754
Al Viro765fd6b2012-03-17 01:19:24 -0400755#define reiserfs_is_journal_aborted(journal) (unlikely (__reiserfs_is_journal_aborted (journal)))
756static inline int __reiserfs_is_journal_aborted(struct reiserfs_journal
757 *journal)
758{
759 return test_bit(J_ABORTED, &journal->j_state);
760}
761
Al Virof466c6f2012-03-17 01:16:43 -0400762/*
763 * Locking primitives. The write lock is a per superblock
764 * special mutex that has properties close to the Big Kernel Lock
765 * which was used in the previous locking scheme.
766 */
767void reiserfs_write_lock(struct super_block *s);
768void reiserfs_write_unlock(struct super_block *s);
Jeff Mahoney278f6672013-08-08 17:34:46 -0400769int __must_check reiserfs_write_unlock_nested(struct super_block *s);
770void reiserfs_write_lock_nested(struct super_block *s, int depth);
Al Virof466c6f2012-03-17 01:16:43 -0400771
772#ifdef CONFIG_REISERFS_CHECK
773void reiserfs_lock_check_recursive(struct super_block *s);
774#else
775static inline void reiserfs_lock_check_recursive(struct super_block *s) { }
776#endif
777
778/*
779 * Several mutexes depend on the write lock.
780 * However sometimes we want to relax the write lock while we hold
781 * these mutexes, according to the release/reacquire on schedule()
782 * properties of the Bkl that were used.
783 * Reiserfs performances and locking were based on this scheme.
784 * Now that the write lock is a mutex and not the bkl anymore, doing so
785 * may result in a deadlock:
786 *
787 * A acquire write_lock
788 * A acquire j_commit_mutex
789 * A release write_lock and wait for something
790 * B acquire write_lock
791 * B can't acquire j_commit_mutex and sleep
792 * A can't acquire write lock anymore
793 * deadlock
794 *
795 * What we do here is avoiding such deadlock by playing the same game
796 * than the Bkl: if we can't acquire a mutex that depends on the write lock,
797 * we release the write lock, wait a bit and then retry.
798 *
799 * The mutexes concerned by this hack are:
800 * - The commit mutex of a journal list
801 * - The flush mutex
802 * - The journal lock
803 * - The inode mutex
804 */
805static inline void reiserfs_mutex_lock_safe(struct mutex *m,
Jeff Mahoney278f6672013-08-08 17:34:46 -0400806 struct super_block *s)
Al Virof466c6f2012-03-17 01:16:43 -0400807{
Jeff Mahoney278f6672013-08-08 17:34:46 -0400808 int depth;
809
810 depth = reiserfs_write_unlock_nested(s);
Al Virof466c6f2012-03-17 01:16:43 -0400811 mutex_lock(m);
Jeff Mahoney278f6672013-08-08 17:34:46 -0400812 reiserfs_write_lock_nested(s, depth);
Al Virof466c6f2012-03-17 01:16:43 -0400813}
814
815static inline void
816reiserfs_mutex_lock_nested_safe(struct mutex *m, unsigned int subclass,
Jeff Mahoney278f6672013-08-08 17:34:46 -0400817 struct super_block *s)
Al Virof466c6f2012-03-17 01:16:43 -0400818{
Jeff Mahoney278f6672013-08-08 17:34:46 -0400819 int depth;
820
821 depth = reiserfs_write_unlock_nested(s);
Al Virof466c6f2012-03-17 01:16:43 -0400822 mutex_lock_nested(m, subclass);
Jeff Mahoney278f6672013-08-08 17:34:46 -0400823 reiserfs_write_lock_nested(s, depth);
Al Virof466c6f2012-03-17 01:16:43 -0400824}
825
826static inline void
827reiserfs_down_read_safe(struct rw_semaphore *sem, struct super_block *s)
828{
Jeff Mahoney278f6672013-08-08 17:34:46 -0400829 int depth;
830 depth = reiserfs_write_unlock_nested(s);
831 down_read(sem);
832 reiserfs_write_lock_nested(s, depth);
Al Virof466c6f2012-03-17 01:16:43 -0400833}
834
835/*
836 * When we schedule, we usually want to also release the write lock,
837 * according to the previous bkl based locking scheme of reiserfs.
838 */
839static inline void reiserfs_cond_resched(struct super_block *s)
840{
841 if (need_resched()) {
Jeff Mahoney278f6672013-08-08 17:34:46 -0400842 int depth;
843
844 depth = reiserfs_write_unlock_nested(s);
Al Virof466c6f2012-03-17 01:16:43 -0400845 schedule();
Jeff Mahoney278f6672013-08-08 17:34:46 -0400846 reiserfs_write_lock_nested(s, depth);
Al Virof466c6f2012-03-17 01:16:43 -0400847 }
848}
849
850struct fid;
851
Jeff Mahoney098297b2014-04-23 10:00:36 -0400852/*
853 * in reading the #defines, it may help to understand that they employ
854 * the following abbreviations:
855 *
856 * B = Buffer
857 * I = Item header
858 * H = Height within the tree (should be changed to LEV)
859 * N = Number of the item in the node
860 * STAT = stat data
861 * DEH = Directory Entry Header
862 * EC = Entry Count
863 * E = Entry number
864 * UL = Unsigned Long
865 * BLKH = BLocK Header
866 * UNFM = UNForMatted node
867 * DC = Disk Child
868 * P = Path
869 *
870 * These #defines are named by concatenating these abbreviations,
871 * where first comes the arguments, and last comes the return value,
872 * of the macro.
873 */
Al Virof466c6f2012-03-17 01:16:43 -0400874
875#define USE_INODE_GENERATION_COUNTER
876
877#define REISERFS_PREALLOCATE
878#define DISPLACE_NEW_PACKING_LOCALITIES
879#define PREALLOCATION_SIZE 9
880
881/* n must be power of 2 */
882#define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
883
Jeff Mahoney098297b2014-04-23 10:00:36 -0400884/*
885 * to be ok for alpha and others we have to align structures to 8 byte
886 * boundary.
887 * FIXME: do not change 4 by anything else: there is code which relies on that
888 */
Al Virof466c6f2012-03-17 01:16:43 -0400889#define ROUND_UP(x) _ROUND_UP(x,8LL)
890
Jeff Mahoney098297b2014-04-23 10:00:36 -0400891/*
892 * debug levels. Right now, CONFIG_REISERFS_CHECK means print all debug
893 * messages.
894 */
Al Virof466c6f2012-03-17 01:16:43 -0400895#define REISERFS_DEBUG_CODE 5 /* extra messages to help find/debug errors */
896
897void __reiserfs_warning(struct super_block *s, const char *id,
898 const char *func, const char *fmt, ...);
899#define reiserfs_warning(s, id, fmt, args...) \
900 __reiserfs_warning(s, id, __func__, fmt, ##args)
901/* assertions handling */
902
Jeff Mahoney098297b2014-04-23 10:00:36 -0400903/* always check a condition and panic if it's false. */
Al Virof466c6f2012-03-17 01:16:43 -0400904#define __RASSERT(cond, scond, format, args...) \
905do { \
906 if (!(cond)) \
907 reiserfs_panic(NULL, "assertion failure", "(" #cond ") at " \
908 __FILE__ ":%i:%s: " format "\n", \
909 in_interrupt() ? -1 : task_pid_nr(current), \
910 __LINE__, __func__ , ##args); \
911} while (0)
912
913#define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args)
914
915#if defined( CONFIG_REISERFS_CHECK )
916#define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args)
917#else
918#define RFALSE( cond, format, args... ) do {;} while( 0 )
919#endif
920
921#define CONSTF __attribute_const__
922/*
923 * Disk Data Structures
924 */
925
Jeff Mahoney098297b2014-04-23 10:00:36 -0400926/***************************************************************************
927 * SUPER BLOCK *
928 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -0400929
930/*
Jeff Mahoney098297b2014-04-23 10:00:36 -0400931 * Structure of super block on disk, a version of which in RAM is often
932 * accessed as REISERFS_SB(s)->s_rs. The version in RAM is part of a larger
933 * structure containing fields never written to disk.
Al Virof466c6f2012-03-17 01:16:43 -0400934 */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400935#define UNSET_HASH 0 /* Detect hash on disk */
Al Virof466c6f2012-03-17 01:16:43 -0400936#define TEA_HASH 1
937#define YURA_HASH 2
938#define R5_HASH 3
939#define DEFAULT_HASH R5_HASH
940
941struct journal_params {
Jeff Mahoney098297b2014-04-23 10:00:36 -0400942 /* where does journal start from on its * device */
943 __le32 jp_journal_1st_block;
944
945 /* journal device st_rdev */
946 __le32 jp_journal_dev;
947
948 /* size of the journal */
949 __le32 jp_journal_size;
950
951 /* max number of blocks in a transaction. */
952 __le32 jp_journal_trans_max;
953
954 /*
955 * random value made on fs creation
956 * (this was sb_journal_block_count)
957 */
958 __le32 jp_journal_magic;
959
960 /* max number of blocks to batch into a trans */
961 __le32 jp_journal_max_batch;
962
963 /* in seconds, how old can an async commit be */
964 __le32 jp_journal_max_commit_age;
965
966 /* in seconds, how old can a transaction be */
967 __le32 jp_journal_max_trans_age;
Al Virof466c6f2012-03-17 01:16:43 -0400968};
969
970/* this is the super from 3.5.X, where X >= 10 */
971struct reiserfs_super_block_v1 {
972 __le32 s_block_count; /* blocks count */
973 __le32 s_free_blocks; /* free blocks count */
974 __le32 s_root_block; /* root block number */
975 struct journal_params s_journal;
976 __le16 s_blocksize; /* block size */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400977
978 /* max size of object id array, see get_objectid() commentary */
979 __le16 s_oid_maxsize;
Al Virof466c6f2012-03-17 01:16:43 -0400980 __le16 s_oid_cursize; /* current size of object id array */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400981
982 /* this is set to 1 when filesystem was umounted, to 2 - when not */
983 __le16 s_umount_state;
984
985 /*
986 * reiserfs magic string indicates that file system is reiserfs:
987 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs"
988 */
989 char s_magic[10];
990
991 /*
992 * it is set to used by fsck to mark which
993 * phase of rebuilding is done
994 */
995 __le16 s_fs_state;
996 /*
997 * indicate, what hash function is being use
998 * to sort names in a directory
999 */
1000 __le32 s_hash_function_code;
Al Virof466c6f2012-03-17 01:16:43 -04001001 __le16 s_tree_height; /* height of disk tree */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001002
1003 /*
1004 * amount of bitmap blocks needed to address
1005 * each block of file system
1006 */
1007 __le16 s_bmap_nr;
1008
1009 /*
1010 * this field is only reliable on filesystem with non-standard journal
1011 */
1012 __le16 s_version;
1013
1014 /*
1015 * size in blocks of journal area on main device, we need to
1016 * keep after making fs with non-standard journal
1017 */
1018 __le16 s_reserved_for_journal;
Al Virof466c6f2012-03-17 01:16:43 -04001019} __attribute__ ((__packed__));
1020
1021#define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))
1022
1023/* this is the on disk super block */
1024struct reiserfs_super_block {
1025 struct reiserfs_super_block_v1 s_v1;
1026 __le32 s_inode_generation;
Jeff Mahoney098297b2014-04-23 10:00:36 -04001027
1028 /* Right now used only by inode-attributes, if enabled */
1029 __le32 s_flags;
1030
Al Virof466c6f2012-03-17 01:16:43 -04001031 unsigned char s_uuid[16]; /* filesystem unique identifier */
1032 unsigned char s_label[16]; /* filesystem volume label */
1033 __le16 s_mnt_count; /* Count of mounts since last fsck */
1034 __le16 s_max_mnt_count; /* Maximum mounts before check */
1035 __le32 s_lastcheck; /* Timestamp of last fsck */
1036 __le32 s_check_interval; /* Interval between checks */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001037
1038 /*
1039 * zero filled by mkreiserfs and reiserfs_convert_objectid_map_v1()
1040 * so any additions must be updated there as well. */
1041 char s_unused[76];
Al Virof466c6f2012-03-17 01:16:43 -04001042} __attribute__ ((__packed__));
1043
1044#define SB_SIZE (sizeof(struct reiserfs_super_block))
1045
1046#define REISERFS_VERSION_1 0
1047#define REISERFS_VERSION_2 2
1048
Jeff Mahoney098297b2014-04-23 10:00:36 -04001049/* on-disk super block fields converted to cpu form */
Al Virof466c6f2012-03-17 01:16:43 -04001050#define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)
1051#define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))
1052#define SB_BLOCKSIZE(s) \
1053 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))
1054#define SB_BLOCK_COUNT(s) \
1055 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))
1056#define SB_FREE_BLOCKS(s) \
1057 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))
1058#define SB_REISERFS_MAGIC(s) \
1059 (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)
1060#define SB_ROOT_BLOCK(s) \
1061 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))
1062#define SB_TREE_HEIGHT(s) \
1063 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))
1064#define SB_REISERFS_STATE(s) \
1065 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))
1066#define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))
1067#define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))
1068
1069#define PUT_SB_BLOCK_COUNT(s, val) \
1070 do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)
1071#define PUT_SB_FREE_BLOCKS(s, val) \
1072 do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)
1073#define PUT_SB_ROOT_BLOCK(s, val) \
1074 do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)
1075#define PUT_SB_TREE_HEIGHT(s, val) \
1076 do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)
1077#define PUT_SB_REISERFS_STATE(s, val) \
1078 do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)
1079#define PUT_SB_VERSION(s, val) \
1080 do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)
1081#define PUT_SB_BMAP_NR(s, val) \
1082 do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)
1083
1084#define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)
1085#define SB_ONDISK_JOURNAL_SIZE(s) \
1086 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))
1087#define SB_ONDISK_JOURNAL_1st_BLOCK(s) \
1088 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))
1089#define SB_ONDISK_JOURNAL_DEVICE(s) \
1090 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
1091#define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
1092 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
1093
1094#define is_block_in_log_or_reserved_area(s, block) \
1095 block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
1096 && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) + \
1097 ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \
1098 SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))
1099
1100int is_reiserfs_3_5(struct reiserfs_super_block *rs);
1101int is_reiserfs_3_6(struct reiserfs_super_block *rs);
1102int is_reiserfs_jr(struct reiserfs_super_block *rs);
1103
Jeff Mahoney098297b2014-04-23 10:00:36 -04001104/*
1105 * ReiserFS leaves the first 64k unused, so that partition labels have
1106 * enough space. If someone wants to write a fancy bootloader that
1107 * needs more than 64k, let us know, and this will be increased in size.
1108 * This number must be larger than than the largest block size on any
1109 * platform, or code will break. -Hans
1110 */
Al Virof466c6f2012-03-17 01:16:43 -04001111#define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
1112#define REISERFS_FIRST_BLOCK unused_define
1113#define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES
1114
1115/* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
1116#define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
1117
1118/* reiserfs internal error code (used by search_by_key and fix_nodes)) */
1119#define CARRY_ON 0
1120#define REPEAT_SEARCH -1
1121#define IO_ERROR -2
1122#define NO_DISK_SPACE -3
1123#define NO_BALANCING_NEEDED (-4)
1124#define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
1125#define QUOTA_EXCEEDED -6
1126
1127typedef __u32 b_blocknr_t;
1128typedef __le32 unp_t;
1129
1130struct unfm_nodeinfo {
1131 unp_t unfm_nodenum;
1132 unsigned short unfm_freespace;
1133};
1134
Jeff Mahoney098297b2014-04-23 10:00:36 -04001135/* there are two formats of keys: 3.5 and 3.6 */
Al Virof466c6f2012-03-17 01:16:43 -04001136#define KEY_FORMAT_3_5 0
1137#define KEY_FORMAT_3_6 1
1138
1139/* there are two stat datas */
1140#define STAT_DATA_V1 0
1141#define STAT_DATA_V2 1
1142
1143static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)
1144{
1145 return container_of(inode, struct reiserfs_inode_info, vfs_inode);
1146}
1147
1148static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)
1149{
1150 return sb->s_fs_info;
1151}
1152
Jeff Mahoney098297b2014-04-23 10:00:36 -04001153/*
1154 * Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u16
1155 * which overflows on large file systems.
1156 */
Al Virof466c6f2012-03-17 01:16:43 -04001157static inline __u32 reiserfs_bmap_count(struct super_block *sb)
1158{
1159 return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1;
1160}
1161
1162static inline int bmap_would_wrap(unsigned bmap_nr)
1163{
1164 return bmap_nr > ((1LL << 16) - 1);
1165}
1166
Jeff Mahoney098297b2014-04-23 10:00:36 -04001167/*
1168 * this says about version of key of all items (but stat data) the
1169 * object consists of
1170 */
Al Virof466c6f2012-03-17 01:16:43 -04001171#define get_inode_item_key_version( inode ) \
1172 ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
1173
1174#define set_inode_item_key_version( inode, version ) \
1175 ({ if((version)==KEY_FORMAT_3_6) \
1176 REISERFS_I(inode)->i_flags |= i_item_key_version_mask; \
1177 else \
1178 REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })
1179
1180#define get_inode_sd_version(inode) \
1181 ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
1182
1183#define set_inode_sd_version(inode, version) \
1184 ({ if((version)==STAT_DATA_V2) \
1185 REISERFS_I(inode)->i_flags |= i_stat_data_version_mask; \
1186 else \
1187 REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })
1188
Jeff Mahoney098297b2014-04-23 10:00:36 -04001189/*
1190 * This is an aggressive tail suppression policy, I am hoping it
1191 * improves our benchmarks. The principle behind it is that percentage
1192 * space saving is what matters, not absolute space saving. This is
1193 * non-intuitive, but it helps to understand it if you consider that the
1194 * cost to access 4 blocks is not much more than the cost to access 1
1195 * block, if you have to do a seek and rotate. A tail risks a
1196 * non-linear disk access that is significant as a percentage of total
1197 * time cost for a 4 block file and saves an amount of space that is
1198 * less significant as a percentage of space, or so goes the hypothesis.
1199 * -Hans
1200 */
Al Virof466c6f2012-03-17 01:16:43 -04001201#define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
1202(\
1203 (!(n_tail_size)) || \
1204 (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
1205 ( (n_file_size) >= (n_block_size) * 4 ) || \
1206 ( ( (n_file_size) >= (n_block_size) * 3 ) && \
1207 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
1208 ( ( (n_file_size) >= (n_block_size) * 2 ) && \
1209 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
1210 ( ( (n_file_size) >= (n_block_size) ) && \
1211 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
1212)
1213
Jeff Mahoney098297b2014-04-23 10:00:36 -04001214/*
1215 * Another strategy for tails, this one means only create a tail if all the
1216 * file would fit into one DIRECT item.
1217 * Primary intention for this one is to increase performance by decreasing
1218 * seeking.
Al Virof466c6f2012-03-17 01:16:43 -04001219*/
1220#define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
1221(\
1222 (!(n_tail_size)) || \
1223 (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
1224)
1225
1226/*
1227 * values for s_umount_state field
1228 */
1229#define REISERFS_VALID_FS 1
1230#define REISERFS_ERROR_FS 2
1231
Jeff Mahoney098297b2014-04-23 10:00:36 -04001232/*
1233 * there are 5 item types currently
1234 */
Al Virof466c6f2012-03-17 01:16:43 -04001235#define TYPE_STAT_DATA 0
1236#define TYPE_INDIRECT 1
1237#define TYPE_DIRECT 2
1238#define TYPE_DIRENTRY 3
1239#define TYPE_MAXTYPE 3
Jeff Mahoney098297b2014-04-23 10:00:36 -04001240#define TYPE_ANY 15 /* FIXME: comment is required */
Al Virof466c6f2012-03-17 01:16:43 -04001241
Jeff Mahoney098297b2014-04-23 10:00:36 -04001242/***************************************************************************
1243 * KEY & ITEM HEAD *
1244 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -04001245
Jeff Mahoney098297b2014-04-23 10:00:36 -04001246/* * directories use this key as well as old files */
Al Virof466c6f2012-03-17 01:16:43 -04001247struct offset_v1 {
1248 __le32 k_offset;
1249 __le32 k_uniqueness;
1250} __attribute__ ((__packed__));
1251
1252struct offset_v2 {
1253 __le64 v;
1254} __attribute__ ((__packed__));
1255
1256static inline __u16 offset_v2_k_type(const struct offset_v2 *v2)
1257{
1258 __u8 type = le64_to_cpu(v2->v) >> 60;
1259 return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY;
1260}
1261
1262static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type)
1263{
1264 v2->v =
1265 (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60);
1266}
1267
1268static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2)
1269{
1270 return le64_to_cpu(v2->v) & (~0ULL >> 4);
1271}
1272
1273static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset)
1274{
1275 offset &= (~0ULL >> 4);
1276 v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset);
1277}
1278
Jeff Mahoney098297b2014-04-23 10:00:36 -04001279/*
1280 * Key of an item determines its location in the S+tree, and
1281 * is composed of 4 components
1282 */
Al Virof466c6f2012-03-17 01:16:43 -04001283struct reiserfs_key {
Jeff Mahoney098297b2014-04-23 10:00:36 -04001284 /* packing locality: by default parent directory object id */
1285 __le32 k_dir_id;
1286
Al Virof466c6f2012-03-17 01:16:43 -04001287 __le32 k_objectid; /* object identifier */
1288 union {
1289 struct offset_v1 k_offset_v1;
1290 struct offset_v2 k_offset_v2;
1291 } __attribute__ ((__packed__)) u;
1292} __attribute__ ((__packed__));
1293
1294struct in_core_key {
Jeff Mahoney098297b2014-04-23 10:00:36 -04001295 /* packing locality: by default parent directory object id */
1296 __u32 k_dir_id;
Al Virof466c6f2012-03-17 01:16:43 -04001297 __u32 k_objectid; /* object identifier */
1298 __u64 k_offset;
1299 __u8 k_type;
1300};
1301
1302struct cpu_key {
1303 struct in_core_key on_disk_key;
1304 int version;
Jeff Mahoney098297b2014-04-23 10:00:36 -04001305 /* 3 in all cases but direct2indirect and indirect2direct conversion */
1306 int key_length;
Al Virof466c6f2012-03-17 01:16:43 -04001307};
1308
Jeff Mahoney098297b2014-04-23 10:00:36 -04001309/*
1310 * Our function for comparing keys can compare keys of different
1311 * lengths. It takes as a parameter the length of the keys it is to
1312 * compare. These defines are used in determining what is to be passed
1313 * to it as that parameter.
1314 */
Al Virof466c6f2012-03-17 01:16:43 -04001315#define REISERFS_FULL_KEY_LEN 4
1316#define REISERFS_SHORT_KEY_LEN 2
1317
1318/* The result of the key compare */
1319#define FIRST_GREATER 1
1320#define SECOND_GREATER -1
1321#define KEYS_IDENTICAL 0
1322#define KEY_FOUND 1
1323#define KEY_NOT_FOUND 0
1324
1325#define KEY_SIZE (sizeof(struct reiserfs_key))
1326#define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
1327
1328/* return values for search_by_key and clones */
1329#define ITEM_FOUND 1
1330#define ITEM_NOT_FOUND 0
1331#define ENTRY_FOUND 1
1332#define ENTRY_NOT_FOUND 0
1333#define DIRECTORY_NOT_FOUND -1
1334#define REGULAR_FILE_FOUND -2
1335#define DIRECTORY_FOUND -3
1336#define BYTE_FOUND 1
1337#define BYTE_NOT_FOUND 0
1338#define FILE_NOT_FOUND -1
1339
1340#define POSITION_FOUND 1
1341#define POSITION_NOT_FOUND 0
1342
Jeff Mahoney098297b2014-04-23 10:00:36 -04001343/* return values for reiserfs_find_entry and search_by_entry_key */
Al Virof466c6f2012-03-17 01:16:43 -04001344#define NAME_FOUND 1
1345#define NAME_NOT_FOUND 0
1346#define GOTO_PREVIOUS_ITEM 2
1347#define NAME_FOUND_INVISIBLE 3
1348
Jeff Mahoney098297b2014-04-23 10:00:36 -04001349/*
1350 * Everything in the filesystem is stored as a set of items. The
1351 * item head contains the key of the item, its free space (for
1352 * indirect items) and specifies the location of the item itself
1353 * within the block.
1354 */
Al Virof466c6f2012-03-17 01:16:43 -04001355
1356struct item_head {
Jeff Mahoney098297b2014-04-23 10:00:36 -04001357 /*
1358 * Everything in the tree is found by searching for it based on
1359 * its key.
1360 */
Al Virof466c6f2012-03-17 01:16:43 -04001361 struct reiserfs_key ih_key;
1362 union {
Jeff Mahoney098297b2014-04-23 10:00:36 -04001363 /*
1364 * The free space in the last unformatted node of an
1365 * indirect item if this is an indirect item. This
1366 * equals 0xFFFF iff this is a direct item or stat data
1367 * item. Note that the key, not this field, is used to
1368 * determine the item type, and thus which field this
1369 * union contains.
1370 */
Al Virof466c6f2012-03-17 01:16:43 -04001371 __le16 ih_free_space_reserved;
Jeff Mahoney098297b2014-04-23 10:00:36 -04001372
1373 /*
1374 * Iff this is a directory item, this field equals the
1375 * number of directory entries in the directory item.
1376 */
Al Virof466c6f2012-03-17 01:16:43 -04001377 __le16 ih_entry_count;
1378 } __attribute__ ((__packed__)) u;
1379 __le16 ih_item_len; /* total size of the item body */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001380
1381 /* an offset to the item body within the block */
1382 __le16 ih_item_location;
1383
1384 /*
1385 * 0 for all old items, 2 for new ones. Highest bit is set by fsck
1386 * temporary, cleaned after all done
1387 */
1388 __le16 ih_version;
Al Virof466c6f2012-03-17 01:16:43 -04001389} __attribute__ ((__packed__));
1390/* size of item header */
1391#define IH_SIZE (sizeof(struct item_head))
1392
1393#define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved)
1394#define ih_version(ih) le16_to_cpu((ih)->ih_version)
1395#define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count)
1396#define ih_location(ih) le16_to_cpu((ih)->ih_item_location)
1397#define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len)
1398
1399#define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
1400#define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0)
1401#define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
1402#define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
1403#define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
1404
1405#define unreachable_item(ih) (ih_version(ih) & (1 << 15))
1406
1407#define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
1408#define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
1409
Jeff Mahoney098297b2014-04-23 10:00:36 -04001410/*
1411 * these operate on indirect items, where you've got an array of ints
1412 * at a possibly unaligned location. These are a noop on ia32
1413 *
1414 * p is the array of __u32, i is the index into the array, v is the value
1415 * to store there.
1416 */
Al Virof466c6f2012-03-17 01:16:43 -04001417#define get_block_num(p, i) get_unaligned_le32((p) + (i))
1418#define put_block_num(p, i, v) put_unaligned_le32((v), (p) + (i))
1419
Jeff Mahoney098297b2014-04-23 10:00:36 -04001420/* * in old version uniqueness field shows key type */
Al Virof466c6f2012-03-17 01:16:43 -04001421#define V1_SD_UNIQUENESS 0
1422#define V1_INDIRECT_UNIQUENESS 0xfffffffe
1423#define V1_DIRECT_UNIQUENESS 0xffffffff
1424#define V1_DIRENTRY_UNIQUENESS 500
Jeff Mahoney098297b2014-04-23 10:00:36 -04001425#define V1_ANY_UNIQUENESS 555 /* FIXME: comment is required */
Al Virof466c6f2012-03-17 01:16:43 -04001426
Jeff Mahoney098297b2014-04-23 10:00:36 -04001427/* here are conversion routines */
Al Virof466c6f2012-03-17 01:16:43 -04001428static inline int uniqueness2type(__u32 uniqueness) CONSTF;
1429static inline int uniqueness2type(__u32 uniqueness)
1430{
1431 switch ((int)uniqueness) {
1432 case V1_SD_UNIQUENESS:
1433 return TYPE_STAT_DATA;
1434 case V1_INDIRECT_UNIQUENESS:
1435 return TYPE_INDIRECT;
1436 case V1_DIRECT_UNIQUENESS:
1437 return TYPE_DIRECT;
1438 case V1_DIRENTRY_UNIQUENESS:
1439 return TYPE_DIRENTRY;
1440 case V1_ANY_UNIQUENESS:
1441 default:
1442 return TYPE_ANY;
1443 }
1444}
1445
1446static inline __u32 type2uniqueness(int type) CONSTF;
1447static inline __u32 type2uniqueness(int type)
1448{
1449 switch (type) {
1450 case TYPE_STAT_DATA:
1451 return V1_SD_UNIQUENESS;
1452 case TYPE_INDIRECT:
1453 return V1_INDIRECT_UNIQUENESS;
1454 case TYPE_DIRECT:
1455 return V1_DIRECT_UNIQUENESS;
1456 case TYPE_DIRENTRY:
1457 return V1_DIRENTRY_UNIQUENESS;
1458 case TYPE_ANY:
1459 default:
1460 return V1_ANY_UNIQUENESS;
1461 }
1462}
1463
Jeff Mahoney098297b2014-04-23 10:00:36 -04001464/*
1465 * key is pointer to on disk key which is stored in le, result is cpu,
1466 * there is no way to get version of object from key, so, provide
1467 * version to these defines
1468 */
Al Virof466c6f2012-03-17 01:16:43 -04001469static inline loff_t le_key_k_offset(int version,
1470 const struct reiserfs_key *key)
1471{
1472 return (version == KEY_FORMAT_3_5) ?
1473 le32_to_cpu(key->u.k_offset_v1.k_offset) :
1474 offset_v2_k_offset(&(key->u.k_offset_v2));
1475}
1476
1477static inline loff_t le_ih_k_offset(const struct item_head *ih)
1478{
1479 return le_key_k_offset(ih_version(ih), &(ih->ih_key));
1480}
1481
1482static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key)
1483{
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04001484 if (version == KEY_FORMAT_3_5) {
1485 loff_t val = le32_to_cpu(key->u.k_offset_v1.k_uniqueness);
1486 return uniqueness2type(val);
1487 } else
1488 return offset_v2_k_type(&(key->u.k_offset_v2));
Al Virof466c6f2012-03-17 01:16:43 -04001489}
1490
1491static inline loff_t le_ih_k_type(const struct item_head *ih)
1492{
1493 return le_key_k_type(ih_version(ih), &(ih->ih_key));
1494}
1495
1496static inline void set_le_key_k_offset(int version, struct reiserfs_key *key,
1497 loff_t offset)
1498{
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04001499 if (version == KEY_FORMAT_3_5)
1500 key->u.k_offset_v1.k_offset = cpu_to_le32(offset);
1501 else
1502 set_offset_v2_k_offset(&key->u.k_offset_v2, offset);
1503}
1504
1505static inline void add_le_key_k_offset(int version, struct reiserfs_key *key,
1506 loff_t offset)
1507{
1508 set_le_key_k_offset(version, key,
1509 le_key_k_offset(version, key) + offset);
1510}
1511
1512static inline void add_le_ih_k_offset(struct item_head *ih, loff_t offset)
1513{
1514 add_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
Al Virof466c6f2012-03-17 01:16:43 -04001515}
1516
1517static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset)
1518{
1519 set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
1520}
1521
1522static inline void set_le_key_k_type(int version, struct reiserfs_key *key,
1523 int type)
1524{
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04001525 if (version == KEY_FORMAT_3_5) {
1526 type = type2uniqueness(type);
1527 key->u.k_offset_v1.k_uniqueness = cpu_to_le32(type);
1528 } else
1529 set_offset_v2_k_type(&key->u.k_offset_v2, type);
Al Virof466c6f2012-03-17 01:16:43 -04001530}
1531
1532static inline void set_le_ih_k_type(struct item_head *ih, int type)
1533{
1534 set_le_key_k_type(ih_version(ih), &(ih->ih_key), type);
1535}
1536
1537static inline int is_direntry_le_key(int version, struct reiserfs_key *key)
1538{
1539 return le_key_k_type(version, key) == TYPE_DIRENTRY;
1540}
1541
1542static inline int is_direct_le_key(int version, struct reiserfs_key *key)
1543{
1544 return le_key_k_type(version, key) == TYPE_DIRECT;
1545}
1546
1547static inline int is_indirect_le_key(int version, struct reiserfs_key *key)
1548{
1549 return le_key_k_type(version, key) == TYPE_INDIRECT;
1550}
1551
1552static inline int is_statdata_le_key(int version, struct reiserfs_key *key)
1553{
1554 return le_key_k_type(version, key) == TYPE_STAT_DATA;
1555}
1556
Jeff Mahoney098297b2014-04-23 10:00:36 -04001557/* item header has version. */
Al Virof466c6f2012-03-17 01:16:43 -04001558static inline int is_direntry_le_ih(struct item_head *ih)
1559{
1560 return is_direntry_le_key(ih_version(ih), &ih->ih_key);
1561}
1562
1563static inline int is_direct_le_ih(struct item_head *ih)
1564{
1565 return is_direct_le_key(ih_version(ih), &ih->ih_key);
1566}
1567
1568static inline int is_indirect_le_ih(struct item_head *ih)
1569{
1570 return is_indirect_le_key(ih_version(ih), &ih->ih_key);
1571}
1572
1573static inline int is_statdata_le_ih(struct item_head *ih)
1574{
1575 return is_statdata_le_key(ih_version(ih), &ih->ih_key);
1576}
1577
Jeff Mahoney098297b2014-04-23 10:00:36 -04001578/* key is pointer to cpu key, result is cpu */
Al Virof466c6f2012-03-17 01:16:43 -04001579static inline loff_t cpu_key_k_offset(const struct cpu_key *key)
1580{
1581 return key->on_disk_key.k_offset;
1582}
1583
1584static inline loff_t cpu_key_k_type(const struct cpu_key *key)
1585{
1586 return key->on_disk_key.k_type;
1587}
1588
1589static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset)
1590{
1591 key->on_disk_key.k_offset = offset;
1592}
1593
1594static inline void set_cpu_key_k_type(struct cpu_key *key, int type)
1595{
1596 key->on_disk_key.k_type = type;
1597}
1598
1599static inline void cpu_key_k_offset_dec(struct cpu_key *key)
1600{
1601 key->on_disk_key.k_offset--;
1602}
1603
1604#define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
1605#define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
1606#define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
1607#define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
1608
1609/* are these used ? */
1610#define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
1611#define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
1612#define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
1613#define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
1614
1615#define I_K_KEY_IN_ITEM(ih, key, n_blocksize) \
1616 (!COMP_SHORT_KEYS(ih, key) && \
1617 I_OFF_BYTE_IN_ITEM(ih, k_offset(key), n_blocksize))
1618
1619/* maximal length of item */
1620#define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
1621#define MIN_ITEM_LEN 1
1622
1623/* object identifier for root dir */
1624#define REISERFS_ROOT_OBJECTID 2
1625#define REISERFS_ROOT_PARENT_OBJECTID 1
1626
1627extern struct reiserfs_key root_key;
1628
Jeff Mahoney098297b2014-04-23 10:00:36 -04001629/*
Al Virof466c6f2012-03-17 01:16:43 -04001630 * Picture represents a leaf of the S+tree
1631 * ______________________________________________________
1632 * | | Array of | | |
1633 * |Block | Object-Item | F r e e | Objects- |
1634 * | head | Headers | S p a c e | Items |
1635 * |______|_______________|___________________|___________|
1636 */
1637
Jeff Mahoney098297b2014-04-23 10:00:36 -04001638/*
1639 * Header of a disk block. More precisely, header of a formatted leaf
1640 * or internal node, and not the header of an unformatted node.
1641 */
Al Virof466c6f2012-03-17 01:16:43 -04001642struct block_head {
1643 __le16 blk_level; /* Level of a block in the tree. */
1644 __le16 blk_nr_item; /* Number of keys/items in a block. */
1645 __le16 blk_free_space; /* Block free space in bytes. */
1646 __le16 blk_reserved;
1647 /* dump this in v4/planA */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001648
1649 /* kept only for compatibility */
1650 struct reiserfs_key blk_right_delim_key;
Al Virof466c6f2012-03-17 01:16:43 -04001651};
1652
1653#define BLKH_SIZE (sizeof(struct block_head))
1654#define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level))
1655#define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item))
1656#define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space))
1657#define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved))
1658#define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val))
1659#define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val))
1660#define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
1661#define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
1662#define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key)
1663#define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val)
1664
Jeff Mahoney098297b2014-04-23 10:00:36 -04001665/* values for blk_level field of the struct block_head */
Al Virof466c6f2012-03-17 01:16:43 -04001666
Jeff Mahoney098297b2014-04-23 10:00:36 -04001667/*
1668 * When node gets removed from the tree its blk_level is set to FREE_LEVEL.
1669 * It is then used to see whether the node is still in the tree
1670 */
1671#define FREE_LEVEL 0
Al Virof466c6f2012-03-17 01:16:43 -04001672
1673#define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level. */
1674
Jeff Mahoney098297b2014-04-23 10:00:36 -04001675/*
1676 * Given the buffer head of a formatted node, resolve to the
1677 * block head of that node.
1678 */
Al Virof466c6f2012-03-17 01:16:43 -04001679#define B_BLK_HEAD(bh) ((struct block_head *)((bh)->b_data))
1680/* Number of items that are in buffer. */
1681#define B_NR_ITEMS(bh) (blkh_nr_item(B_BLK_HEAD(bh)))
1682#define B_LEVEL(bh) (blkh_level(B_BLK_HEAD(bh)))
1683#define B_FREE_SPACE(bh) (blkh_free_space(B_BLK_HEAD(bh)))
1684
1685#define PUT_B_NR_ITEMS(bh, val) do { set_blkh_nr_item(B_BLK_HEAD(bh), val); } while (0)
1686#define PUT_B_LEVEL(bh, val) do { set_blkh_level(B_BLK_HEAD(bh), val); } while (0)
1687#define PUT_B_FREE_SPACE(bh, val) do { set_blkh_free_space(B_BLK_HEAD(bh), val); } while (0)
1688
1689/* Get right delimiting key. -- little endian */
1690#define B_PRIGHT_DELIM_KEY(bh) (&(blk_right_delim_key(B_BLK_HEAD(bh))))
1691
1692/* Does the buffer contain a disk leaf. */
1693#define B_IS_ITEMS_LEVEL(bh) (B_LEVEL(bh) == DISK_LEAF_NODE_LEVEL)
1694
1695/* Does the buffer contain a disk internal node */
1696#define B_IS_KEYS_LEVEL(bh) (B_LEVEL(bh) > DISK_LEAF_NODE_LEVEL \
1697 && B_LEVEL(bh) <= MAX_HEIGHT)
1698
Jeff Mahoney098297b2014-04-23 10:00:36 -04001699/***************************************************************************
1700 * STAT DATA *
1701 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -04001702
Jeff Mahoney098297b2014-04-23 10:00:36 -04001703/*
1704 * old stat data is 32 bytes long. We are going to distinguish new one by
1705 * different size
1706*/
Al Virof466c6f2012-03-17 01:16:43 -04001707struct stat_data_v1 {
1708 __le16 sd_mode; /* file type, permissions */
1709 __le16 sd_nlink; /* number of hard links */
1710 __le16 sd_uid; /* owner */
1711 __le16 sd_gid; /* group */
1712 __le32 sd_size; /* file size */
1713 __le32 sd_atime; /* time of last access */
1714 __le32 sd_mtime; /* time file was last modified */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001715
1716 /*
1717 * time inode (stat data) was last changed
1718 * (except changes to sd_atime and sd_mtime)
1719 */
1720 __le32 sd_ctime;
Al Virof466c6f2012-03-17 01:16:43 -04001721 union {
1722 __le32 sd_rdev;
1723 __le32 sd_blocks; /* number of blocks file uses */
1724 } __attribute__ ((__packed__)) u;
Jeff Mahoney098297b2014-04-23 10:00:36 -04001725
1726 /*
1727 * first byte of file which is stored in a direct item: except that if
1728 * it equals 1 it is a symlink and if it equals ~(__u32)0 there is no
1729 * direct item. The existence of this field really grates on me.
1730 * Let's replace it with a macro based on sd_size and our tail
1731 * suppression policy. Someday. -Hans
1732 */
1733 __le32 sd_first_direct_byte;
Al Virof466c6f2012-03-17 01:16:43 -04001734} __attribute__ ((__packed__));
1735
1736#define SD_V1_SIZE (sizeof(struct stat_data_v1))
1737#define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5)
1738#define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
1739#define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
1740#define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink))
1741#define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v))
1742#define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid))
1743#define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v))
1744#define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid))
1745#define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v))
1746#define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size))
1747#define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v))
1748#define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
1749#define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
1750#define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
1751#define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
1752#define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
1753#define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
1754#define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
1755#define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
1756#define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks))
1757#define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
1758#define sd_v1_first_direct_byte(sdp) \
1759 (le32_to_cpu((sdp)->sd_first_direct_byte))
1760#define set_sd_v1_first_direct_byte(sdp,v) \
1761 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
1762
1763/* inode flags stored in sd_attrs (nee sd_reserved) */
1764
Jeff Mahoney098297b2014-04-23 10:00:36 -04001765/*
1766 * we want common flags to have the same values as in ext2,
1767 * so chattr(1) will work without problems
1768 */
Al Virof466c6f2012-03-17 01:16:43 -04001769#define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL
1770#define REISERFS_APPEND_FL FS_APPEND_FL
1771#define REISERFS_SYNC_FL FS_SYNC_FL
1772#define REISERFS_NOATIME_FL FS_NOATIME_FL
1773#define REISERFS_NODUMP_FL FS_NODUMP_FL
1774#define REISERFS_SECRM_FL FS_SECRM_FL
1775#define REISERFS_UNRM_FL FS_UNRM_FL
1776#define REISERFS_COMPR_FL FS_COMPR_FL
1777#define REISERFS_NOTAIL_FL FS_NOTAIL_FL
1778
1779/* persistent flags that file inherits from the parent directory */
1780#define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \
1781 REISERFS_SYNC_FL | \
1782 REISERFS_NOATIME_FL | \
1783 REISERFS_NODUMP_FL | \
1784 REISERFS_SECRM_FL | \
1785 REISERFS_COMPR_FL | \
1786 REISERFS_NOTAIL_FL )
1787
Jeff Mahoney098297b2014-04-23 10:00:36 -04001788/*
1789 * Stat Data on disk (reiserfs version of UFS disk inode minus the
1790 * address blocks)
1791 */
Al Virof466c6f2012-03-17 01:16:43 -04001792struct stat_data {
1793 __le16 sd_mode; /* file type, permissions */
1794 __le16 sd_attrs; /* persistent inode flags */
1795 __le32 sd_nlink; /* number of hard links */
1796 __le64 sd_size; /* file size */
1797 __le32 sd_uid; /* owner */
1798 __le32 sd_gid; /* group */
1799 __le32 sd_atime; /* time of last access */
1800 __le32 sd_mtime; /* time file was last modified */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001801
1802 /*
1803 * time inode (stat data) was last changed
1804 * (except changes to sd_atime and sd_mtime)
1805 */
1806 __le32 sd_ctime;
Al Virof466c6f2012-03-17 01:16:43 -04001807 __le32 sd_blocks;
1808 union {
1809 __le32 sd_rdev;
1810 __le32 sd_generation;
Al Virof466c6f2012-03-17 01:16:43 -04001811 } __attribute__ ((__packed__)) u;
1812} __attribute__ ((__packed__));
Jeff Mahoney098297b2014-04-23 10:00:36 -04001813
1814/* this is 44 bytes long */
Al Virof466c6f2012-03-17 01:16:43 -04001815#define SD_SIZE (sizeof(struct stat_data))
1816#define SD_V2_SIZE SD_SIZE
1817#define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6)
1818#define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
1819#define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
1820/* sd_reserved */
1821/* set_sd_reserved */
1822#define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink))
1823#define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v))
1824#define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size))
1825#define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v))
1826#define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid))
1827#define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v))
1828#define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid))
1829#define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v))
1830#define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
1831#define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
1832#define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
1833#define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
1834#define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
1835#define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
1836#define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks))
1837#define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
1838#define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
1839#define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
1840#define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation))
1841#define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
1842#define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs))
1843#define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v))
1844
Jeff Mahoney098297b2014-04-23 10:00:36 -04001845/***************************************************************************
1846 * DIRECTORY STRUCTURE *
1847 ***************************************************************************/
1848/*
1849 * Picture represents the structure of directory items
1850 * ________________________________________________
1851 * | Array of | | | | | |
1852 * | directory |N-1| N-2 | .... | 1st |0th|
1853 * | entry headers | | | | | |
1854 * |_______________|___|_____|________|_______|___|
1855 * <---- directory entries ------>
1856 *
1857 * First directory item has k_offset component 1. We store "." and ".."
1858 * in one item, always, we never split "." and ".." into differing
1859 * items. This makes, among other things, the code for removing
1860 * directories simpler.
1861 */
Al Virof466c6f2012-03-17 01:16:43 -04001862#define SD_OFFSET 0
1863#define SD_UNIQUENESS 0
1864#define DOT_OFFSET 1
1865#define DOT_DOT_OFFSET 2
1866#define DIRENTRY_UNIQUENESS 500
1867
Al Virof466c6f2012-03-17 01:16:43 -04001868#define FIRST_ITEM_OFFSET 1
1869
1870/*
Jeff Mahoney098297b2014-04-23 10:00:36 -04001871 * Q: How to get key of object pointed to by entry from entry?
1872 *
1873 * A: Each directory entry has its header. This header has deh_dir_id
1874 * and deh_objectid fields, those are key of object, entry points to
1875 */
Al Virof466c6f2012-03-17 01:16:43 -04001876
Jeff Mahoney098297b2014-04-23 10:00:36 -04001877/*
1878 * NOT IMPLEMENTED:
1879 * Directory will someday contain stat data of object
1880 */
Al Virof466c6f2012-03-17 01:16:43 -04001881
1882struct reiserfs_de_head {
1883 __le32 deh_offset; /* third component of the directory entry key */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001884
1885 /*
1886 * objectid of the parent directory of the object, that is referenced
1887 * by directory entry
1888 */
1889 __le32 deh_dir_id;
1890
1891 /* objectid of the object, that is referenced by directory entry */
1892 __le32 deh_objectid;
Al Virof466c6f2012-03-17 01:16:43 -04001893 __le16 deh_location; /* offset of name in the whole item */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001894
1895 /*
1896 * whether 1) entry contains stat data (for future), and
1897 * 2) whether entry is hidden (unlinked)
1898 */
1899 __le16 deh_state;
Al Virof466c6f2012-03-17 01:16:43 -04001900} __attribute__ ((__packed__));
1901#define DEH_SIZE sizeof(struct reiserfs_de_head)
1902#define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset))
1903#define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id))
1904#define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid))
1905#define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location))
1906#define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state))
1907
1908#define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v)))
1909#define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v)))
1910#define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
1911#define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
1912#define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v)))
1913
1914/* empty directory contains two entries "." and ".." and their headers */
1915#define EMPTY_DIR_SIZE \
1916(DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
1917
1918/* old format directories have this size when empty */
1919#define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
1920
1921#define DEH_Statdata 0 /* not used now */
1922#define DEH_Visible 2
1923
1924/* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
1925#if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
1926# define ADDR_UNALIGNED_BITS (3)
1927#endif
1928
Jeff Mahoney098297b2014-04-23 10:00:36 -04001929/*
1930 * These are only used to manipulate deh_state.
Al Virof466c6f2012-03-17 01:16:43 -04001931 * Because of this, we'll use the ext2_ bit routines,
Jeff Mahoney098297b2014-04-23 10:00:36 -04001932 * since they are little endian
1933 */
Al Virof466c6f2012-03-17 01:16:43 -04001934#ifdef ADDR_UNALIGNED_BITS
1935
1936# define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
1937# define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
1938
1939# define set_bit_unaligned(nr, addr) \
1940 __test_and_set_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1941# define clear_bit_unaligned(nr, addr) \
1942 __test_and_clear_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1943# define test_bit_unaligned(nr, addr) \
1944 test_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1945
1946#else
1947
1948# define set_bit_unaligned(nr, addr) __test_and_set_bit_le(nr, addr)
1949# define clear_bit_unaligned(nr, addr) __test_and_clear_bit_le(nr, addr)
1950# define test_bit_unaligned(nr, addr) test_bit_le(nr, addr)
1951
1952#endif
1953
1954#define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1955#define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1956#define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1957#define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1958
1959#define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1960#define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1961#define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1962
1963extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid,
1964 __le32 par_dirid, __le32 par_objid);
1965extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid,
1966 __le32 par_dirid, __le32 par_objid);
1967
Jeff Mahoney098297b2014-04-23 10:00:36 -04001968/* two entries per block (at least) */
Al Virof466c6f2012-03-17 01:16:43 -04001969#define REISERFS_MAX_NAME(block_size) 255
1970
Jeff Mahoney098297b2014-04-23 10:00:36 -04001971/*
1972 * this structure is used for operations on directory entries. It is
1973 * not a disk structure.
1974 *
1975 * When reiserfs_find_entry or search_by_entry_key find directory
1976 * entry, they return filled reiserfs_dir_entry structure
1977 */
Al Virof466c6f2012-03-17 01:16:43 -04001978struct reiserfs_dir_entry {
1979 struct buffer_head *de_bh;
1980 int de_item_num;
1981 struct item_head *de_ih;
1982 int de_entry_num;
1983 struct reiserfs_de_head *de_deh;
1984 int de_entrylen;
1985 int de_namelen;
1986 char *de_name;
1987 unsigned long *de_gen_number_bit_string;
1988
1989 __u32 de_dir_id;
1990 __u32 de_objectid;
1991
1992 struct cpu_key de_entry_key;
1993};
1994
Jeff Mahoney098297b2014-04-23 10:00:36 -04001995/*
1996 * these defines are useful when a particular member of
1997 * a reiserfs_dir_entry is needed
1998 */
Al Virof466c6f2012-03-17 01:16:43 -04001999
2000/* pointer to file name, stored in entry */
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04002001#define B_I_DEH_ENTRY_FILE_NAME(bh, ih, deh) \
2002 (ih_item_body(bh, ih) + deh_location(deh))
Al Virof466c6f2012-03-17 01:16:43 -04002003
2004/* length of name */
2005#define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
2006(I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
2007
2008/* hash value occupies bits from 7 up to 30 */
2009#define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
2010/* generation number occupies 7 bits starting from 0 up to 6 */
2011#define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
2012#define MAX_GENERATION_NUMBER 127
2013
2014#define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
2015
2016/*
2017 * Picture represents an internal node of the reiserfs tree
2018 * ______________________________________________________
2019 * | | Array of | Array of | Free |
2020 * |block | keys | pointers | space |
2021 * | head | N | N+1 | |
2022 * |______|_______________|___________________|___________|
2023 */
2024
Jeff Mahoney098297b2014-04-23 10:00:36 -04002025/***************************************************************************
2026 * DISK CHILD *
2027 ***************************************************************************/
2028/*
2029 * Disk child pointer:
2030 * The pointer from an internal node of the tree to a node that is on disk.
2031 */
Al Virof466c6f2012-03-17 01:16:43 -04002032struct disk_child {
2033 __le32 dc_block_number; /* Disk child's block number. */
2034 __le16 dc_size; /* Disk child's used space. */
2035 __le16 dc_reserved;
2036};
2037
2038#define DC_SIZE (sizeof(struct disk_child))
2039#define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number))
2040#define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size))
2041#define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
2042#define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
2043
2044/* Get disk child by buffer header and position in the tree node. */
2045#define B_N_CHILD(bh, n_pos) ((struct disk_child *)\
2046((bh)->b_data + BLKH_SIZE + B_NR_ITEMS(bh) * KEY_SIZE + DC_SIZE * (n_pos)))
2047
2048/* Get disk child number by buffer header and position in the tree node. */
2049#define B_N_CHILD_NUM(bh, n_pos) (dc_block_number(B_N_CHILD(bh, n_pos)))
2050#define PUT_B_N_CHILD_NUM(bh, n_pos, val) \
2051 (put_dc_block_number(B_N_CHILD(bh, n_pos), val))
2052
2053 /* maximal value of field child_size in structure disk_child */
2054 /* child size is the combined size of all items and their headers */
2055#define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
2056
2057/* amount of used space in buffer (not including block head) */
2058#define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
2059
2060/* max and min number of keys in internal node */
2061#define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
2062#define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2)
2063
Jeff Mahoney098297b2014-04-23 10:00:36 -04002064/***************************************************************************
2065 * PATH STRUCTURES AND DEFINES *
2066 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -04002067
Jeff Mahoney098297b2014-04-23 10:00:36 -04002068/*
2069 * search_by_key fills up the path from the root to the leaf as it descends
2070 * the tree looking for the key. It uses reiserfs_bread to try to find
2071 * buffers in the cache given their block number. If it does not find
2072 * them in the cache it reads them from disk. For each node search_by_key
2073 * finds using reiserfs_bread it then uses bin_search to look through that
2074 * node. bin_search will find the position of the block_number of the next
2075 * node if it is looking through an internal node. If it is looking through
2076 * a leaf node bin_search will find the position of the item which has key
2077 * either equal to given key, or which is the maximal key less than the
2078 * given key.
2079 */
Al Virof466c6f2012-03-17 01:16:43 -04002080
2081struct path_element {
Jeff Mahoney098297b2014-04-23 10:00:36 -04002082 /* Pointer to the buffer at the path in the tree. */
2083 struct buffer_head *pe_buffer;
2084 /* Position in the tree node which is placed in the buffer above. */
2085 int pe_position;
Al Virof466c6f2012-03-17 01:16:43 -04002086};
2087
Jeff Mahoney098297b2014-04-23 10:00:36 -04002088/*
2089 * maximal height of a tree. don't change this without
2090 * changing JOURNAL_PER_BALANCE_CNT
2091 */
2092#define MAX_HEIGHT 5
Al Virof466c6f2012-03-17 01:16:43 -04002093
Jeff Mahoney098297b2014-04-23 10:00:36 -04002094/* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
2095#define EXTENDED_MAX_HEIGHT 7
Al Virof466c6f2012-03-17 01:16:43 -04002096
Jeff Mahoney098297b2014-04-23 10:00:36 -04002097/* Must be equal to at least 2. */
2098#define FIRST_PATH_ELEMENT_OFFSET 2
Al Virof466c6f2012-03-17 01:16:43 -04002099
Jeff Mahoney098297b2014-04-23 10:00:36 -04002100/* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
2101#define ILLEGAL_PATH_ELEMENT_OFFSET 1
2102
2103/* this MUST be MAX_HEIGHT + 1. See about FEB below */
2104#define MAX_FEB_SIZE 6
2105
2106/*
2107 * We need to keep track of who the ancestors of nodes are. When we
2108 * perform a search we record which nodes were visited while
2109 * descending the tree looking for the node we searched for. This list
2110 * of nodes is called the path. This information is used while
2111 * performing balancing. Note that this path information may become
2112 * invalid, and this means we must check it when using it to see if it
2113 * is still valid. You'll need to read search_by_key and the comments
2114 * in it, especially about decrement_counters_in_path(), to understand
2115 * this structure.
2116 *
2117 * Paths make the code so much harder to work with and debug.... An
2118 * enormous number of bugs are due to them, and trying to write or modify
2119 * code that uses them just makes my head hurt. They are based on an
2120 * excessive effort to avoid disturbing the precious VFS code.:-( The
2121 * gods only know how we are going to SMP the code that uses them.
2122 * znodes are the way!
2123 */
Al Virof466c6f2012-03-17 01:16:43 -04002124
2125#define PATH_READA 0x1 /* do read ahead */
2126#define PATH_READA_BACK 0x2 /* read backwards */
2127
2128struct treepath {
2129 int path_length; /* Length of the array above. */
2130 int reada;
Jeff Mahoney098297b2014-04-23 10:00:36 -04002131 /* Array of the path elements. */
2132 struct path_element path_elements[EXTENDED_MAX_HEIGHT];
Al Virof466c6f2012-03-17 01:16:43 -04002133 int pos_in_item;
2134};
2135
2136#define pos_in_item(path) ((path)->pos_in_item)
2137
2138#define INITIALIZE_PATH(var) \
2139struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
2140
2141/* Get path element by path and path position. */
2142#define PATH_OFFSET_PELEMENT(path, n_offset) ((path)->path_elements + (n_offset))
2143
2144/* Get buffer header at the path by path and path position. */
2145#define PATH_OFFSET_PBUFFER(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_buffer)
2146
2147/* Get position in the element at the path by path and path position. */
2148#define PATH_OFFSET_POSITION(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_position)
2149
2150#define PATH_PLAST_BUFFER(path) (PATH_OFFSET_PBUFFER((path), (path)->path_length))
Jeff Mahoney098297b2014-04-23 10:00:36 -04002151
2152/*
2153 * you know, to the person who didn't write this the macro name does not
2154 * at first suggest what it does. Maybe POSITION_FROM_PATH_END? Or
2155 * maybe we should just focus on dumping paths... -Hans
2156 */
Al Virof466c6f2012-03-17 01:16:43 -04002157#define PATH_LAST_POSITION(path) (PATH_OFFSET_POSITION((path), (path)->path_length))
2158
Jeff Mahoney098297b2014-04-23 10:00:36 -04002159/*
2160 * in do_balance leaf has h == 0 in contrast with path structure,
2161 * where root has level == 0. That is why we need these defines
2162 */
2163
2164/* tb->S[h] */
2165#define PATH_H_PBUFFER(path, h) \
2166 PATH_OFFSET_PBUFFER(path, path->path_length - (h))
2167
2168/* tb->F[h] or tb->S[0]->b_parent */
2169#define PATH_H_PPARENT(path, h) PATH_H_PBUFFER(path, (h) + 1)
2170
2171#define PATH_H_POSITION(path, h) \
2172 PATH_OFFSET_POSITION(path, path->path_length - (h))
2173
2174/* tb->S[h]->b_item_order */
2175#define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1)
Al Virof466c6f2012-03-17 01:16:43 -04002176
2177#define PATH_H_PATH_OFFSET(path, n_h) ((path)->path_length - (n_h))
2178
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04002179static inline void *reiserfs_node_data(const struct buffer_head *bh)
2180{
2181 return bh->b_data + sizeof(struct block_head);
2182}
2183
2184/* get key from internal node */
2185static inline struct reiserfs_key *internal_key(struct buffer_head *bh,
2186 int item_num)
2187{
2188 struct reiserfs_key *key = reiserfs_node_data(bh);
2189
2190 return &key[item_num];
2191}
2192
2193/* get the item header from leaf node */
2194static inline struct item_head *item_head(const struct buffer_head *bh,
2195 int item_num)
2196{
2197 struct item_head *ih = reiserfs_node_data(bh);
2198
2199 return &ih[item_num];
2200}
2201
2202/* get the key from leaf node */
2203static inline struct reiserfs_key *leaf_key(const struct buffer_head *bh,
2204 int item_num)
2205{
2206 return &item_head(bh, item_num)->ih_key;
2207}
2208
2209static inline void *ih_item_body(const struct buffer_head *bh,
2210 const struct item_head *ih)
2211{
2212 return bh->b_data + ih_location(ih);
2213}
2214
2215/* get item body from leaf node */
2216static inline void *item_body(const struct buffer_head *bh, int item_num)
2217{
2218 return ih_item_body(bh, item_head(bh, item_num));
2219}
2220
2221static inline struct item_head *tp_item_head(const struct treepath *path)
2222{
2223 return item_head(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path));
2224}
2225
2226static inline void *tp_item_body(const struct treepath *path)
2227{
2228 return item_body(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path));
2229}
2230
Al Virof466c6f2012-03-17 01:16:43 -04002231#define get_last_bh(path) PATH_PLAST_BUFFER(path)
Al Virof466c6f2012-03-17 01:16:43 -04002232#define get_item_pos(path) PATH_LAST_POSITION(path)
Al Virof466c6f2012-03-17 01:16:43 -04002233#define item_moved(ih,path) comp_items(ih, path)
2234#define path_changed(ih,path) comp_items (ih, path)
2235
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04002236/* array of the entry headers */
2237 /* get item body */
2238#define B_I_DEH(bh, ih) ((struct reiserfs_de_head *)(ih_item_body(bh, ih)))
2239
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04002240/*
Jeff Mahoney098297b2014-04-23 10:00:36 -04002241 * length of the directory entry in directory item. This define
2242 * calculates length of i-th directory entry using directory entry
2243 * locations from dir entry head. When it calculates length of 0-th
2244 * directory entry, it uses length of whole item in place of entry
2245 * location of the non-existent following entry in the calculation.
2246 * See picture above.
2247 */
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04002248static inline int entry_length(const struct buffer_head *bh,
2249 const struct item_head *ih, int pos_in_item)
2250{
2251 struct reiserfs_de_head *deh;
2252
2253 deh = B_I_DEH(bh, ih) + pos_in_item;
2254 if (pos_in_item)
2255 return deh_location(deh - 1) - deh_location(deh);
2256
2257 return ih_item_len(ih) - deh_location(deh);
2258}
2259
Jeff Mahoney098297b2014-04-23 10:00:36 -04002260/***************************************************************************
2261 * MISC *
2262 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -04002263
2264/* Size of pointer to the unformatted node. */
2265#define UNFM_P_SIZE (sizeof(unp_t))
2266#define UNFM_P_SHIFT 2
2267
Jeff Mahoney098297b2014-04-23 10:00:36 -04002268/* in in-core inode key is stored on le form */
Al Virof466c6f2012-03-17 01:16:43 -04002269#define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
2270
2271#define MAX_UL_INT 0xffffffff
2272#define MAX_INT 0x7ffffff
2273#define MAX_US_INT 0xffff
2274
2275// reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
Al Virof466c6f2012-03-17 01:16:43 -04002276static inline loff_t max_reiserfs_offset(struct inode *inode)
2277{
2278 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
2279 return (loff_t) U32_MAX;
2280
2281 return (loff_t) ((~(__u64) 0) >> 4);
2282}
2283
Al Virof466c6f2012-03-17 01:16:43 -04002284#define MAX_KEY_OBJECTID MAX_UL_INT
2285
2286#define MAX_B_NUM MAX_UL_INT
2287#define MAX_FC_NUM MAX_US_INT
2288
2289/* the purpose is to detect overflow of an unsigned short */
2290#define REISERFS_LINK_MAX (MAX_US_INT - 1000)
2291
Jeff Mahoney098297b2014-04-23 10:00:36 -04002292/*
2293 * The following defines are used in reiserfs_insert_item
2294 * and reiserfs_append_item
2295 */
2296#define REISERFS_KERNEL_MEM 0 /* kernel memory mode */
2297#define REISERFS_USER_MEM 1 /* user memory mode */
Al Virof466c6f2012-03-17 01:16:43 -04002298
2299#define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
2300#define get_generation(s) atomic_read (&fs_generation(s))
2301#define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen)
2302#define __fs_changed(gen,s) (gen != get_generation (s))
2303#define fs_changed(gen,s) \
2304({ \
2305 reiserfs_cond_resched(s); \
2306 __fs_changed(gen, s); \
2307})
2308
Jeff Mahoney098297b2014-04-23 10:00:36 -04002309/***************************************************************************
2310 * FIXATE NODES *
2311 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -04002312
2313#define VI_TYPE_LEFT_MERGEABLE 1
2314#define VI_TYPE_RIGHT_MERGEABLE 2
2315
Jeff Mahoney098297b2014-04-23 10:00:36 -04002316/*
2317 * To make any changes in the tree we always first find node, that
2318 * contains item to be changed/deleted or place to insert a new
2319 * item. We call this node S. To do balancing we need to decide what
2320 * we will shift to left/right neighbor, or to a new node, where new
2321 * item will be etc. To make this analysis simpler we build virtual
2322 * node. Virtual node is an array of items, that will replace items of
2323 * node S. (For instance if we are going to delete an item, virtual
2324 * node does not contain it). Virtual node keeps information about
2325 * item sizes and types, mergeability of first and last items, sizes
2326 * of all entries in directory item. We use this array of items when
2327 * calculating what we can shift to neighbors and how many nodes we
2328 * have to have if we do not any shiftings, if we shift to left/right
2329 * neighbor or to both.
2330 */
Al Virof466c6f2012-03-17 01:16:43 -04002331struct virtual_item {
Jeff Mahoney098297b2014-04-23 10:00:36 -04002332 int vi_index; /* index in the array of item operations */
2333 unsigned short vi_type; /* left/right mergeability */
2334
2335 /* length of item that it will have after balancing */
2336 unsigned short vi_item_len;
2337
Al Virof466c6f2012-03-17 01:16:43 -04002338 struct item_head *vi_ih;
Jeff Mahoney098297b2014-04-23 10:00:36 -04002339 const char *vi_item; /* body of item (old or new) */
2340 const void *vi_new_data; /* 0 always but paste mode */
2341 void *vi_uarea; /* item specific area */
Al Virof466c6f2012-03-17 01:16:43 -04002342};
2343
2344struct virtual_node {
Jeff Mahoney098297b2014-04-23 10:00:36 -04002345 /* this is a pointer to the free space in the buffer */
2346 char *vn_free_ptr;
2347
Al Virof466c6f2012-03-17 01:16:43 -04002348 unsigned short vn_nr_item; /* number of items in virtual node */
Jeff Mahoney098297b2014-04-23 10:00:36 -04002349
2350 /*
2351 * size of node , that node would have if it has
2352 * unlimited size and no balancing is performed
2353 */
2354 short vn_size;
2355
2356 /* mode of balancing (paste, insert, delete, cut) */
2357 short vn_mode;
2358
Al Virof466c6f2012-03-17 01:16:43 -04002359 short vn_affected_item_num;
2360 short vn_pos_in_item;
Jeff Mahoney098297b2014-04-23 10:00:36 -04002361
2362 /* item header of inserted item, 0 for other modes */
2363 struct item_head *vn_ins_ih;
Al Virof466c6f2012-03-17 01:16:43 -04002364 const void *vn_data;
Jeff Mahoney098297b2014-04-23 10:00:36 -04002365
2366 /* array of items (including a new one, excluding item to be deleted) */
2367 struct virtual_item *vn_vi;
Al Virof466c6f2012-03-17 01:16:43 -04002368};
2369
2370/* used by directory items when creating virtual nodes */
2371struct direntry_uarea {
2372 int flags;
2373 __u16 entry_count;
2374 __u16 entry_sizes[1];
2375} __attribute__ ((__packed__));
2376
Jeff Mahoney098297b2014-04-23 10:00:36 -04002377/***************************************************************************
2378 * TREE BALANCE *
2379 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -04002380
Jeff Mahoney098297b2014-04-23 10:00:36 -04002381/*
2382 * This temporary structure is used in tree balance algorithms, and
2383 * constructed as we go to the extent that its various parts are
2384 * needed. It contains arrays of nodes that can potentially be
2385 * involved in the balancing of node S, and parameters that define how
2386 * each of the nodes must be balanced. Note that in these algorithms
2387 * for balancing the worst case is to need to balance the current node
2388 * S and the left and right neighbors and all of their parents plus
2389 * create a new node. We implement S1 balancing for the leaf nodes
2390 * and S0 balancing for the internal nodes (S1 and S0 are defined in
2391 * our papers.)
2392 */
Al Virof466c6f2012-03-17 01:16:43 -04002393
Jeff Mahoney098297b2014-04-23 10:00:36 -04002394/* size of the array of buffers to free at end of do_balance */
2395#define MAX_FREE_BLOCK 7
Al Virof466c6f2012-03-17 01:16:43 -04002396
2397/* maximum number of FEB blocknrs on a single level */
2398#define MAX_AMOUNT_NEEDED 2
2399
2400/* someday somebody will prefix every field in this struct with tb_ */
2401struct tree_balance {
2402 int tb_mode;
2403 int need_balance_dirty;
2404 struct super_block *tb_sb;
2405 struct reiserfs_transaction_handle *transaction_handle;
2406 struct treepath *tb_path;
Al Virof466c6f2012-03-17 01:16:43 -04002407
Jeff Mahoney098297b2014-04-23 10:00:36 -04002408 /* array of left neighbors of nodes in the path */
2409 struct buffer_head *L[MAX_HEIGHT];
2410
2411 /* array of right neighbors of nodes in the path */
2412 struct buffer_head *R[MAX_HEIGHT];
2413
2414 /* array of fathers of the left neighbors */
2415 struct buffer_head *FL[MAX_HEIGHT];
2416
2417 /* array of fathers of the right neighbors */
2418 struct buffer_head *FR[MAX_HEIGHT];
2419 /* array of common parents of center node and its left neighbor */
2420 struct buffer_head *CFL[MAX_HEIGHT];
2421
2422 /* array of common parents of center node and its right neighbor */
2423 struct buffer_head *CFR[MAX_HEIGHT];
2424
2425 /*
2426 * array of empty buffers. Number of buffers in array equals
2427 * cur_blknum.
2428 */
2429 struct buffer_head *FEB[MAX_FEB_SIZE];
Al Virof466c6f2012-03-17 01:16:43 -04002430 struct buffer_head *used[MAX_FEB_SIZE];
2431 struct buffer_head *thrown[MAX_FEB_SIZE];
Jeff Mahoney098297b2014-04-23 10:00:36 -04002432
2433 /*
2434 * array of number of items which must be shifted to the left in
2435 * order to balance the current node; for leaves includes item that
2436 * will be partially shifted; for internal nodes, it is the number
2437 * of child pointers rather than items. It includes the new item
2438 * being created. The code sometimes subtracts one to get the
2439 * number of wholly shifted items for other purposes.
2440 */
2441 int lnum[MAX_HEIGHT];
2442
2443 /* substitute right for left in comment above */
2444 int rnum[MAX_HEIGHT];
2445
2446 /*
2447 * array indexed by height h mapping the key delimiting L[h] and
2448 * S[h] to its item number within the node CFL[h]
2449 */
2450 int lkey[MAX_HEIGHT];
2451
2452 /* substitute r for l in comment above */
2453 int rkey[MAX_HEIGHT];
2454
2455 /*
2456 * the number of bytes by we are trying to add or remove from
2457 * S[h]. A negative value means removing.
2458 */
2459 int insert_size[MAX_HEIGHT];
2460
2461 /*
2462 * number of nodes that will replace node S[h] after balancing
2463 * on the level h of the tree. If 0 then S is being deleted,
2464 * if 1 then S is remaining and no new nodes are being created,
2465 * if 2 or 3 then 1 or 2 new nodes is being created
2466 */
2467 int blknum[MAX_HEIGHT];
Al Virof466c6f2012-03-17 01:16:43 -04002468
2469 /* fields that are used only for balancing leaves of the tree */
Al Virof466c6f2012-03-17 01:16:43 -04002470
Jeff Mahoney098297b2014-04-23 10:00:36 -04002471 /* number of empty blocks having been already allocated */
2472 int cur_blknum;
2473
2474 /* number of items that fall into left most node when S[0] splits */
2475 int s0num;
2476
Jeff Mahoney098297b2014-04-23 10:00:36 -04002477 /*
2478 * number of bytes which can flow to the left neighbor from the left
2479 * most liquid item that cannot be shifted from S[0] entirely
2480 * if -1 then nothing will be partially shifted
2481 */
2482 int lbytes;
2483
2484 /*
2485 * number of bytes which will flow to the right neighbor from the right
2486 * most liquid item that cannot be shifted from S[0] entirely
2487 * if -1 then nothing will be partially shifted
2488 */
2489 int rbytes;
2490
Jeff Mahoneyb49fb112014-04-23 10:00:46 -04002491
Jeff Mahoney098297b2014-04-23 10:00:36 -04002492 /*
Jeff Mahoneyb49fb112014-04-23 10:00:46 -04002493 * index into the array of item headers in
2494 * S[0] of the affected item
2495 */
2496 int item_pos;
2497
2498 /* new nodes allocated to hold what could not fit into S */
2499 struct buffer_head *S_new[2];
2500
2501 /*
2502 * number of items that will be placed into nodes in S_new
2503 * when S[0] splits
2504 */
2505 int snum[2];
2506
2507 /*
2508 * number of bytes which flow to nodes in S_new when S[0] splits
Jeff Mahoney098297b2014-04-23 10:00:36 -04002509 * note: if S[0] splits into 3 nodes, then items do not need to be cut
2510 */
Jeff Mahoneyb49fb112014-04-23 10:00:46 -04002511 int sbytes[2];
2512
2513 int pos_in_item;
2514 int zeroes_num;
Jeff Mahoney098297b2014-04-23 10:00:36 -04002515
2516 /*
2517 * buffers which are to be freed after do_balance finishes
2518 * by unfix_nodes
2519 */
2520 struct buffer_head *buf_to_free[MAX_FREE_BLOCK];
2521
2522 /*
2523 * kmalloced memory. Used to create virtual node and keep
2524 * map of dirtied bitmap blocks
2525 */
2526 char *vn_buf;
2527
2528 int vn_buf_size; /* size of the vn_buf */
2529
2530 /* VN starts after bitmap of bitmap blocks */
2531 struct virtual_node *tb_vn;
2532
2533 /*
2534 * saved value of `reiserfs_generation' counter see
2535 * FILESYSTEM_CHANGED() macro in reiserfs_fs.h
2536 */
2537 int fs_gen;
2538
Al Virof466c6f2012-03-17 01:16:43 -04002539#ifdef DISPLACE_NEW_PACKING_LOCALITIES
Jeff Mahoney098297b2014-04-23 10:00:36 -04002540 /*
2541 * key pointer, to pass to block allocator or
2542 * another low-level subsystem
2543 */
2544 struct in_core_key key;
Al Virof466c6f2012-03-17 01:16:43 -04002545#endif
2546};
2547
2548/* These are modes of balancing */
2549
2550/* When inserting an item. */
2551#define M_INSERT 'i'
Jeff Mahoney098297b2014-04-23 10:00:36 -04002552/*
2553 * When inserting into (directories only) or appending onto an already
2554 * existent item.
2555 */
Al Virof466c6f2012-03-17 01:16:43 -04002556#define M_PASTE 'p'
2557/* When deleting an item. */
2558#define M_DELETE 'd'
2559/* When truncating an item or removing an entry from a (directory) item. */
Jeff Mahoney098297b2014-04-23 10:00:36 -04002560#define M_CUT 'c'
Al Virof466c6f2012-03-17 01:16:43 -04002561
2562/* used when balancing on leaf level skipped (in reiserfsck) */
2563#define M_INTERNAL 'n'
2564
Jeff Mahoney098297b2014-04-23 10:00:36 -04002565/*
2566 * When further balancing is not needed, then do_balance does not need
2567 * to be called.
2568 */
2569#define M_SKIP_BALANCING 's'
Al Virof466c6f2012-03-17 01:16:43 -04002570#define M_CONVERT 'v'
2571
2572/* modes of leaf_move_items */
2573#define LEAF_FROM_S_TO_L 0
2574#define LEAF_FROM_S_TO_R 1
2575#define LEAF_FROM_R_TO_L 2
2576#define LEAF_FROM_L_TO_R 3
2577#define LEAF_FROM_S_TO_SNEW 4
2578
2579#define FIRST_TO_LAST 0
2580#define LAST_TO_FIRST 1
2581
Jeff Mahoney098297b2014-04-23 10:00:36 -04002582/*
2583 * used in do_balance for passing parent of node information that has
2584 * been gotten from tb struct
2585 */
Al Virof466c6f2012-03-17 01:16:43 -04002586struct buffer_info {
2587 struct tree_balance *tb;
2588 struct buffer_head *bi_bh;
2589 struct buffer_head *bi_parent;
2590 int bi_position;
2591};
2592
2593static inline struct super_block *sb_from_tb(struct tree_balance *tb)
2594{
2595 return tb ? tb->tb_sb : NULL;
2596}
2597
2598static inline struct super_block *sb_from_bi(struct buffer_info *bi)
2599{
2600 return bi ? sb_from_tb(bi->tb) : NULL;
2601}
2602
Jeff Mahoney098297b2014-04-23 10:00:36 -04002603/*
2604 * there are 4 types of items: stat data, directory item, indirect, direct.
2605 * +-------------------+------------+--------------+------------+
2606 * | | k_offset | k_uniqueness | mergeable? |
2607 * +-------------------+------------+--------------+------------+
2608 * | stat data | 0 | 0 | no |
2609 * +-------------------+------------+--------------+------------+
2610 * | 1st directory item| DOT_OFFSET | DIRENTRY_ .. | no |
2611 * | non 1st directory | hash value | UNIQUENESS | yes |
2612 * | item | | | |
2613 * +-------------------+------------+--------------+------------+
2614 * | indirect item | offset + 1 |TYPE_INDIRECT | [1] |
2615 * +-------------------+------------+--------------+------------+
2616 * | direct item | offset + 1 |TYPE_DIRECT | [2] |
2617 * +-------------------+------------+--------------+------------+
2618 *
2619 * [1] if this is not the first indirect item of the object
2620 * [2] if this is not the first direct item of the object
Al Virof466c6f2012-03-17 01:16:43 -04002621*/
2622
2623struct item_operations {
2624 int (*bytes_number) (struct item_head * ih, int block_size);
2625 void (*decrement_key) (struct cpu_key *);
2626 int (*is_left_mergeable) (struct reiserfs_key * ih,
2627 unsigned long bsize);
2628 void (*print_item) (struct item_head *, char *item);
2629 void (*check_item) (struct item_head *, char *item);
2630
2631 int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
2632 int is_affected, int insert_size);
2633 int (*check_left) (struct virtual_item * vi, int free,
2634 int start_skip, int end_skip);
2635 int (*check_right) (struct virtual_item * vi, int free);
2636 int (*part_size) (struct virtual_item * vi, int from, int to);
2637 int (*unit_num) (struct virtual_item * vi);
2638 void (*print_vi) (struct virtual_item * vi);
2639};
2640
2641extern struct item_operations *item_ops[TYPE_ANY + 1];
2642
2643#define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
2644#define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
2645#define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item)
2646#define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item)
2647#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)
2648#define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
2649#define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free)
2650#define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to)
2651#define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi)
2652#define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi)
2653
2654#define COMP_SHORT_KEYS comp_short_keys
2655
2656/* number of blocks pointed to by the indirect item */
2657#define I_UNFM_NUM(ih) (ih_item_len(ih) / UNFM_P_SIZE)
2658
Jeff Mahoney098297b2014-04-23 10:00:36 -04002659/*
2660 * the used space within the unformatted node corresponding
2661 * to pos within the item pointed to by ih
2662 */
Al Virof466c6f2012-03-17 01:16:43 -04002663#define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
2664
Jeff Mahoney098297b2014-04-23 10:00:36 -04002665/*
2666 * number of bytes contained by the direct item or the
2667 * unformatted nodes the indirect item points to
2668 */
Al Virof466c6f2012-03-17 01:16:43 -04002669
Jeff Mahoney098297b2014-04-23 10:00:36 -04002670/* following defines use reiserfs buffer header and item header */
Al Virof466c6f2012-03-17 01:16:43 -04002671
2672/* get stat-data */
2673#define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
2674
Jeff Mahoney098297b2014-04-23 10:00:36 -04002675/* this is 3976 for size==4096 */
Al Virof466c6f2012-03-17 01:16:43 -04002676#define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
2677
Jeff Mahoney098297b2014-04-23 10:00:36 -04002678/*
2679 * indirect items consist of entries which contain blocknrs, pos
2680 * indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
2681 * blocknr contained by the entry pos points to
2682 */
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04002683#define B_I_POS_UNFM_POINTER(bh, ih, pos) \
2684 le32_to_cpu(*(((unp_t *)ih_item_body(bh, ih)) + (pos)))
2685#define PUT_B_I_POS_UNFM_POINTER(bh, ih, pos, val) \
2686 (*(((unp_t *)ih_item_body(bh, ih)) + (pos)) = cpu_to_le32(val))
Al Virof466c6f2012-03-17 01:16:43 -04002687
2688struct reiserfs_iget_args {
2689 __u32 objectid;
2690 __u32 dirid;
2691};
2692
Jeff Mahoney098297b2014-04-23 10:00:36 -04002693/***************************************************************************
2694 * FUNCTION DECLARATIONS *
2695 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -04002696
2697#define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
2698
2699#define journal_trans_half(blocksize) \
2700 ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
2701
2702/* journal.c see journal.c for all the comments here */
2703
2704/* first block written in a commit. */
2705struct reiserfs_journal_desc {
2706 __le32 j_trans_id; /* id of commit */
Jeff Mahoney098297b2014-04-23 10:00:36 -04002707
2708 /* length of commit. len +1 is the commit block */
2709 __le32 j_len;
2710
Al Virof466c6f2012-03-17 01:16:43 -04002711 __le32 j_mount_id; /* mount id of this trans */
2712 __le32 j_realblock[1]; /* real locations for each block */
2713};
2714
2715#define get_desc_trans_id(d) le32_to_cpu((d)->j_trans_id)
2716#define get_desc_trans_len(d) le32_to_cpu((d)->j_len)
2717#define get_desc_mount_id(d) le32_to_cpu((d)->j_mount_id)
2718
2719#define set_desc_trans_id(d,val) do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
2720#define set_desc_trans_len(d,val) do { (d)->j_len = cpu_to_le32 (val); } while (0)
2721#define set_desc_mount_id(d,val) do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
2722
2723/* last block written in a commit */
2724struct reiserfs_journal_commit {
2725 __le32 j_trans_id; /* must match j_trans_id from the desc block */
2726 __le32 j_len; /* ditto */
2727 __le32 j_realblock[1]; /* real locations for each block */
2728};
2729
2730#define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
2731#define get_commit_trans_len(c) le32_to_cpu((c)->j_len)
2732#define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
2733
2734#define set_commit_trans_id(c,val) do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
2735#define set_commit_trans_len(c,val) do { (c)->j_len = cpu_to_le32 (val); } while (0)
2736
Jeff Mahoney098297b2014-04-23 10:00:36 -04002737/*
2738 * this header block gets written whenever a transaction is considered
2739 * fully flushed, and is more recent than the last fully flushed transaction.
2740 * fully flushed means all the log blocks and all the real blocks are on
2741 * disk, and this transaction does not need to be replayed.
2742 */
Al Virof466c6f2012-03-17 01:16:43 -04002743struct reiserfs_journal_header {
Jeff Mahoney098297b2014-04-23 10:00:36 -04002744 /* id of last fully flushed transaction */
2745 __le32 j_last_flush_trans_id;
2746
2747 /* offset in the log of where to start replay after a crash */
2748 __le32 j_first_unflushed_offset;
2749
Al Virof466c6f2012-03-17 01:16:43 -04002750 __le32 j_mount_id;
2751 /* 12 */ struct journal_params jh_journal;
2752};
2753
2754/* biggest tunable defines are right here */
2755#define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */
Jeff Mahoney098297b2014-04-23 10:00:36 -04002756
2757/* biggest possible single transaction, don't change for now (8/3/99) */
2758#define JOURNAL_TRANS_MAX_DEFAULT 1024
Al Virof466c6f2012-03-17 01:16:43 -04002759#define JOURNAL_TRANS_MIN_DEFAULT 256
Jeff Mahoney098297b2014-04-23 10:00:36 -04002760
2761/*
2762 * max blocks to batch into one transaction,
2763 * don't make this any bigger than 900
2764 */
2765#define JOURNAL_MAX_BATCH_DEFAULT 900
Al Virof466c6f2012-03-17 01:16:43 -04002766#define JOURNAL_MIN_RATIO 2
2767#define JOURNAL_MAX_COMMIT_AGE 30
2768#define JOURNAL_MAX_TRANS_AGE 30
2769#define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
2770#define JOURNAL_BLOCKS_PER_OBJECT(sb) (JOURNAL_PER_BALANCE_CNT * 3 + \
2771 2 * (REISERFS_QUOTA_INIT_BLOCKS(sb) + \
2772 REISERFS_QUOTA_TRANS_BLOCKS(sb)))
2773
2774#ifdef CONFIG_QUOTA
2775#define REISERFS_QUOTA_OPTS ((1 << REISERFS_USRQUOTA) | (1 << REISERFS_GRPQUOTA))
2776/* We need to update data and inode (atime) */
2777#define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? 2 : 0)
2778/* 1 balancing, 1 bitmap, 1 data per write + stat data update */
2779#define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
2780(DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0)
2781/* same as with INIT */
2782#define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
2783(DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0)
2784#else
2785#define REISERFS_QUOTA_TRANS_BLOCKS(s) 0
2786#define REISERFS_QUOTA_INIT_BLOCKS(s) 0
2787#define REISERFS_QUOTA_DEL_BLOCKS(s) 0
2788#endif
2789
Jeff Mahoney098297b2014-04-23 10:00:36 -04002790/*
2791 * both of these can be as low as 1, or as high as you want. The min is the
2792 * number of 4k bitmap nodes preallocated on mount. New nodes are allocated
2793 * as needed, and released when transactions are committed. On release, if
2794 * the current number of nodes is > max, the node is freed, otherwise,
2795 * it is put on a free list for faster use later.
Al Virof466c6f2012-03-17 01:16:43 -04002796*/
2797#define REISERFS_MIN_BITMAP_NODES 10
2798#define REISERFS_MAX_BITMAP_NODES 100
2799
Jeff Mahoney098297b2014-04-23 10:00:36 -04002800/* these are based on journal hash size of 8192 */
2801#define JBH_HASH_SHIFT 13
Al Virof466c6f2012-03-17 01:16:43 -04002802#define JBH_HASH_MASK 8191
2803
2804#define _jhashfn(sb,block) \
2805 (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
2806 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
2807#define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
2808
Jeff Mahoney098297b2014-04-23 10:00:36 -04002809/* We need these to make journal.c code more readable */
Al Virof466c6f2012-03-17 01:16:43 -04002810#define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2811#define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2812#define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2813
2814enum reiserfs_bh_state_bits {
2815 BH_JDirty = BH_PrivateStart, /* buffer is in current transaction */
2816 BH_JDirty_wait,
Jeff Mahoney098297b2014-04-23 10:00:36 -04002817 /*
2818 * disk block was taken off free list before being in a
2819 * finished transaction, or written to disk. Can be reused immed.
2820 */
2821 BH_JNew,
Al Virof466c6f2012-03-17 01:16:43 -04002822 BH_JPrepared,
2823 BH_JRestore_dirty,
Jeff Mahoney098297b2014-04-23 10:00:36 -04002824 BH_JTest, /* debugging only will go away */
Al Virof466c6f2012-03-17 01:16:43 -04002825};
2826
2827BUFFER_FNS(JDirty, journaled);
2828TAS_BUFFER_FNS(JDirty, journaled);
2829BUFFER_FNS(JDirty_wait, journal_dirty);
2830TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
2831BUFFER_FNS(JNew, journal_new);
2832TAS_BUFFER_FNS(JNew, journal_new);
2833BUFFER_FNS(JPrepared, journal_prepared);
2834TAS_BUFFER_FNS(JPrepared, journal_prepared);
2835BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
2836TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
2837BUFFER_FNS(JTest, journal_test);
2838TAS_BUFFER_FNS(JTest, journal_test);
2839
Jeff Mahoney098297b2014-04-23 10:00:36 -04002840/* transaction handle which is passed around for all journal calls */
Al Virof466c6f2012-03-17 01:16:43 -04002841struct reiserfs_transaction_handle {
Jeff Mahoney098297b2014-04-23 10:00:36 -04002842 /*
2843 * super for this FS when journal_begin was called. saves calls to
2844 * reiserfs_get_super also used by nested transactions to make
2845 * sure they are nesting on the right FS _must_ be first
2846 * in the handle
2847 */
2848 struct super_block *t_super;
2849
Al Virof466c6f2012-03-17 01:16:43 -04002850 int t_refcount;
2851 int t_blocks_logged; /* number of blocks this writer has logged */
2852 int t_blocks_allocated; /* number of blocks this writer allocated */
Jeff Mahoney098297b2014-04-23 10:00:36 -04002853
2854 /* sanity check, equals the current trans id */
2855 unsigned int t_trans_id;
2856
Al Virof466c6f2012-03-17 01:16:43 -04002857 void *t_handle_save; /* save existing current->journal_info */
Jeff Mahoney098297b2014-04-23 10:00:36 -04002858
2859 /*
2860 * if new block allocation occurres, that block
2861 * should be displaced from others
2862 */
2863 unsigned displace_new_blocks:1;
2864
Al Virof466c6f2012-03-17 01:16:43 -04002865 struct list_head t_list;
2866};
2867
Jeff Mahoney098297b2014-04-23 10:00:36 -04002868/*
2869 * used to keep track of ordered and tail writes, attached to the buffer
Al Virof466c6f2012-03-17 01:16:43 -04002870 * head through b_journal_head.
2871 */
2872struct reiserfs_jh {
2873 struct reiserfs_journal_list *jl;
2874 struct buffer_head *bh;
2875 struct list_head list;
2876};
2877
2878void reiserfs_free_jh(struct buffer_head *bh);
2879int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
2880int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
2881int journal_mark_dirty(struct reiserfs_transaction_handle *,
Jeff Mahoney09f1b802014-04-23 10:00:39 -04002882 struct buffer_head *bh);
Al Virof466c6f2012-03-17 01:16:43 -04002883
2884static inline int reiserfs_file_data_log(struct inode *inode)
2885{
2886 if (reiserfs_data_log(inode->i_sb) ||
2887 (REISERFS_I(inode)->i_flags & i_data_log))
2888 return 1;
2889 return 0;
2890}
2891
2892static inline int reiserfs_transaction_running(struct super_block *s)
2893{
2894 struct reiserfs_transaction_handle *th = current->journal_info;
2895 if (th && th->t_super == s)
2896 return 1;
2897 if (th && th->t_super == NULL)
2898 BUG();
2899 return 0;
2900}
2901
2902static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th)
2903{
2904 return th->t_blocks_allocated - th->t_blocks_logged;
2905}
2906
2907struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
2908 super_block
2909 *,
2910 int count);
2911int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
Marco Stornellicfac4b42012-12-15 11:47:31 +01002912void reiserfs_vfs_truncate_file(struct inode *inode);
Al Virof466c6f2012-03-17 01:16:43 -04002913int reiserfs_commit_page(struct inode *inode, struct page *page,
2914 unsigned from, unsigned to);
Artem Bityutskiy25729b02012-06-01 17:18:05 +03002915void reiserfs_flush_old_commits(struct super_block *);
Al Virof466c6f2012-03-17 01:16:43 -04002916int reiserfs_commit_for_inode(struct inode *);
2917int reiserfs_inode_needs_commit(struct inode *);
2918void reiserfs_update_inode_transaction(struct inode *);
2919void reiserfs_wait_on_write_block(struct super_block *s);
2920void reiserfs_block_writes(struct reiserfs_transaction_handle *th);
2921void reiserfs_allow_writes(struct super_block *s);
2922void reiserfs_check_lock_depth(struct super_block *s, char *caller);
2923int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh,
2924 int wait);
2925void reiserfs_restore_prepared_buffer(struct super_block *,
2926 struct buffer_head *bh);
2927int journal_init(struct super_block *, const char *j_dev_name, int old_format,
2928 unsigned int);
2929int journal_release(struct reiserfs_transaction_handle *, struct super_block *);
2930int journal_release_error(struct reiserfs_transaction_handle *,
2931 struct super_block *);
Jeff Mahoney58d85422014-04-23 10:00:38 -04002932int journal_end(struct reiserfs_transaction_handle *);
2933int journal_end_sync(struct reiserfs_transaction_handle *);
Al Virof466c6f2012-03-17 01:16:43 -04002934int journal_mark_freed(struct reiserfs_transaction_handle *,
2935 struct super_block *, b_blocknr_t blocknr);
2936int journal_transaction_should_end(struct reiserfs_transaction_handle *, int);
2937int reiserfs_in_journal(struct super_block *sb, unsigned int bmap_nr,
2938 int bit_nr, int searchall, b_blocknr_t *next);
2939int journal_begin(struct reiserfs_transaction_handle *,
2940 struct super_block *sb, unsigned long);
2941int journal_join_abort(struct reiserfs_transaction_handle *,
Jeff Mahoneyb491dd172014-04-23 10:00:40 -04002942 struct super_block *sb);
Al Virof466c6f2012-03-17 01:16:43 -04002943void reiserfs_abort_journal(struct super_block *sb, int errno);
2944void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...);
2945int reiserfs_allocate_list_bitmaps(struct super_block *s,
2946 struct reiserfs_list_bitmap *, unsigned int);
2947
Artem Bityutskiy033369d2012-06-01 17:18:08 +03002948void reiserfs_schedule_old_flush(struct super_block *s);
Al Virof466c6f2012-03-17 01:16:43 -04002949void add_save_link(struct reiserfs_transaction_handle *th,
2950 struct inode *inode, int truncate);
2951int remove_save_link(struct inode *inode, int truncate);
2952
2953/* objectid.c */
2954__u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th);
2955void reiserfs_release_objectid(struct reiserfs_transaction_handle *th,
2956 __u32 objectid_to_release);
2957int reiserfs_convert_objectid_map_v1(struct super_block *);
2958
2959/* stree.c */
2960int B_IS_IN_TREE(const struct buffer_head *);
2961extern void copy_item_head(struct item_head *to,
2962 const struct item_head *from);
2963
Jeff Mahoney098297b2014-04-23 10:00:36 -04002964/* first key is in cpu form, second - le */
Al Virof466c6f2012-03-17 01:16:43 -04002965extern int comp_short_keys(const struct reiserfs_key *le_key,
2966 const struct cpu_key *cpu_key);
2967extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from);
2968
Jeff Mahoney098297b2014-04-23 10:00:36 -04002969/* both are in le form */
Al Virof466c6f2012-03-17 01:16:43 -04002970extern int comp_le_keys(const struct reiserfs_key *,
2971 const struct reiserfs_key *);
2972extern int comp_short_le_keys(const struct reiserfs_key *,
2973 const struct reiserfs_key *);
2974
Jeff Mahoney098297b2014-04-23 10:00:36 -04002975/* * get key version from on disk key - kludge */
Al Virof466c6f2012-03-17 01:16:43 -04002976static inline int le_key_version(const struct reiserfs_key *key)
2977{
2978 int type;
2979
2980 type = offset_v2_k_type(&(key->u.k_offset_v2));
2981 if (type != TYPE_DIRECT && type != TYPE_INDIRECT
2982 && type != TYPE_DIRENTRY)
2983 return KEY_FORMAT_3_5;
2984
2985 return KEY_FORMAT_3_6;
2986
2987}
2988
2989static inline void copy_key(struct reiserfs_key *to,
2990 const struct reiserfs_key *from)
2991{
2992 memcpy(to, from, KEY_SIZE);
2993}
2994
2995int comp_items(const struct item_head *stored_ih, const struct treepath *path);
2996const struct reiserfs_key *get_rkey(const struct treepath *chk_path,
2997 const struct super_block *sb);
2998int search_by_key(struct super_block *, const struct cpu_key *,
2999 struct treepath *, int);
3000#define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
3001int search_for_position_by_key(struct super_block *sb,
3002 const struct cpu_key *cpu_key,
3003 struct treepath *search_path);
3004extern void decrement_bcount(struct buffer_head *bh);
3005void decrement_counters_in_path(struct treepath *search_path);
3006void pathrelse(struct treepath *search_path);
3007int reiserfs_check_path(struct treepath *p);
3008void pathrelse_and_restore(struct super_block *s, struct treepath *search_path);
3009
3010int reiserfs_insert_item(struct reiserfs_transaction_handle *th,
3011 struct treepath *path,
3012 const struct cpu_key *key,
3013 struct item_head *ih,
3014 struct inode *inode, const char *body);
3015
3016int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th,
3017 struct treepath *path,
3018 const struct cpu_key *key,
3019 struct inode *inode,
3020 const char *body, int paste_size);
3021
3022int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,
3023 struct treepath *path,
3024 struct cpu_key *key,
3025 struct inode *inode,
3026 struct page *page, loff_t new_file_size);
3027
3028int reiserfs_delete_item(struct reiserfs_transaction_handle *th,
3029 struct treepath *path,
3030 const struct cpu_key *key,
3031 struct inode *inode, struct buffer_head *un_bh);
3032
3033void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
3034 struct inode *inode, struct reiserfs_key *key);
3035int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
3036 struct inode *inode);
3037int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,
3038 struct inode *inode, struct page *,
3039 int update_timestamps);
3040
3041#define i_block_size(inode) ((inode)->i_sb->s_blocksize)
3042#define file_size(inode) ((inode)->i_size)
3043#define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
3044
3045#define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
3046!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 )
3047
3048void padd_item(char *item, int total_length, int length);
3049
3050/* inode.c */
3051/* args for the create parameter of reiserfs_get_block */
Jeff Mahoney098297b2014-04-23 10:00:36 -04003052#define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */
3053#define GET_BLOCK_CREATE 1 /* add anything you need to find block */
3054#define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */
3055#define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
3056#define GET_BLOCK_NO_IMUX 8 /* i_mutex is not held, don't preallocate */
3057#define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */
Al Virof466c6f2012-03-17 01:16:43 -04003058
3059void reiserfs_read_locked_inode(struct inode *inode,
3060 struct reiserfs_iget_args *args);
3061int reiserfs_find_actor(struct inode *inode, void *p);
3062int reiserfs_init_locked_inode(struct inode *inode, void *p);
3063void reiserfs_evict_inode(struct inode *inode);
3064int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc);
3065int reiserfs_get_block(struct inode *inode, sector_t block,
3066 struct buffer_head *bh_result, int create);
3067struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3068 int fh_len, int fh_type);
3069struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
3070 int fh_len, int fh_type);
Al Virob0b03822012-04-02 14:34:06 -04003071int reiserfs_encode_fh(struct inode *inode, __u32 * data, int *lenp,
3072 struct inode *parent);
Al Virof466c6f2012-03-17 01:16:43 -04003073
3074int reiserfs_truncate_file(struct inode *, int update_timestamps);
3075void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset,
3076 int type, int key_length);
3077void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
3078 int version,
3079 loff_t offset, int type, int length, int entry_count);
3080struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key);
3081
3082struct reiserfs_security_handle;
3083int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
3084 struct inode *dir, umode_t mode,
3085 const char *symname, loff_t i_size,
3086 struct dentry *dentry, struct inode *inode,
3087 struct reiserfs_security_handle *security);
3088
3089void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
3090 struct inode *inode, loff_t size);
3091
3092static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
3093 struct inode *inode)
3094{
3095 reiserfs_update_sd_size(th, inode, inode->i_size);
3096}
3097
3098void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode);
3099void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs);
3100int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
3101
3102int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len);
3103
3104/* namei.c */
3105void set_de_name_and_namelen(struct reiserfs_dir_entry *de);
3106int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,
3107 struct treepath *path, struct reiserfs_dir_entry *de);
3108struct dentry *reiserfs_get_parent(struct dentry *);
3109
3110#ifdef CONFIG_REISERFS_PROC_INFO
3111int reiserfs_proc_info_init(struct super_block *sb);
3112int reiserfs_proc_info_done(struct super_block *sb);
3113int reiserfs_proc_info_global_init(void);
3114int reiserfs_proc_info_global_done(void);
3115
3116#define PROC_EXP( e ) e
3117
3118#define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
3119#define PROC_INFO_MAX( sb, field, value ) \
3120 __PINFO( sb ).field = \
3121 max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
3122#define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
3123#define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
3124#define PROC_INFO_BH_STAT( sb, bh, level ) \
3125 PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \
3126 PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \
3127 PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
3128#else
3129static inline int reiserfs_proc_info_init(struct super_block *sb)
3130{
3131 return 0;
3132}
3133
3134static inline int reiserfs_proc_info_done(struct super_block *sb)
3135{
3136 return 0;
3137}
3138
3139static inline int reiserfs_proc_info_global_init(void)
3140{
3141 return 0;
3142}
3143
3144static inline int reiserfs_proc_info_global_done(void)
3145{
3146 return 0;
3147}
3148
3149#define PROC_EXP( e )
3150#define VOID_V ( ( void ) 0 )
3151#define PROC_INFO_MAX( sb, field, value ) VOID_V
3152#define PROC_INFO_INC( sb, field ) VOID_V
3153#define PROC_INFO_ADD( sb, field, val ) VOID_V
3154#define PROC_INFO_BH_STAT(sb, bh, n_node_level) VOID_V
3155#endif
3156
3157/* dir.c */
3158extern const struct inode_operations reiserfs_dir_inode_operations;
3159extern const struct inode_operations reiserfs_symlink_inode_operations;
3160extern const struct inode_operations reiserfs_special_inode_operations;
3161extern const struct file_operations reiserfs_dir_operations;
Al Virocd62cda2013-05-17 22:58:58 -04003162int reiserfs_readdir_inode(struct inode *, struct dir_context *);
Al Virof466c6f2012-03-17 01:16:43 -04003163
3164/* tail_conversion.c */
3165int direct2indirect(struct reiserfs_transaction_handle *, struct inode *,
3166 struct treepath *, struct buffer_head *, loff_t);
3167int indirect2direct(struct reiserfs_transaction_handle *, struct inode *,
3168 struct page *, struct treepath *, const struct cpu_key *,
3169 loff_t, char *);
3170void reiserfs_unmap_buffer(struct buffer_head *);
3171
3172/* file.c */
3173extern const struct inode_operations reiserfs_file_inode_operations;
3174extern const struct file_operations reiserfs_file_operations;
3175extern const struct address_space_operations reiserfs_address_space_operations;
3176
3177/* fix_nodes.c */
3178
3179int fix_nodes(int n_op_mode, struct tree_balance *tb,
3180 struct item_head *ins_ih, const void *);
3181void unfix_nodes(struct tree_balance *);
3182
3183/* prints.c */
3184void __reiserfs_panic(struct super_block *s, const char *id,
3185 const char *function, const char *fmt, ...)
3186 __attribute__ ((noreturn));
3187#define reiserfs_panic(s, id, fmt, args...) \
3188 __reiserfs_panic(s, id, __func__, fmt, ##args)
3189void __reiserfs_error(struct super_block *s, const char *id,
3190 const char *function, const char *fmt, ...);
3191#define reiserfs_error(s, id, fmt, args...) \
3192 __reiserfs_error(s, id, __func__, fmt, ##args)
3193void reiserfs_info(struct super_block *s, const char *fmt, ...);
3194void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...);
3195void print_indirect_item(struct buffer_head *bh, int item_num);
3196void store_print_tb(struct tree_balance *tb);
3197void print_cur_tb(char *mes);
3198void print_de(struct reiserfs_dir_entry *de);
3199void print_bi(struct buffer_info *bi, char *mes);
3200#define PRINT_LEAF_ITEMS 1 /* print all items */
3201#define PRINT_DIRECTORY_ITEMS 2 /* print directory items */
3202#define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */
3203void print_block(struct buffer_head *bh, ...);
3204void print_bmap(struct super_block *s, int silent);
3205void print_bmap_block(int i, char *data, int size, int silent);
3206/*void print_super_block (struct super_block * s, char * mes);*/
3207void print_objectid_map(struct super_block *s);
3208void print_block_head(struct buffer_head *bh, char *mes);
3209void check_leaf(struct buffer_head *bh);
3210void check_internal(struct buffer_head *bh);
3211void print_statistics(struct super_block *s);
3212char *reiserfs_hashname(int code);
3213
3214/* lbalance.c */
3215int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num,
3216 int mov_bytes, struct buffer_head *Snew);
3217int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes);
3218int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes);
3219void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first,
3220 int del_num, int del_bytes);
3221void leaf_insert_into_buf(struct buffer_info *bi, int before,
Jeff Mahoney27d0e5b2014-08-04 19:51:47 -04003222 struct item_head * const inserted_item_ih,
3223 const char * const inserted_item_body,
Al Virof466c6f2012-03-17 01:16:43 -04003224 int zeros_number);
Jeff Mahoney27d0e5b2014-08-04 19:51:47 -04003225void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num,
3226 int pos_in_item, int paste_size,
3227 const char * const body, int zeros_number);
Al Virof466c6f2012-03-17 01:16:43 -04003228void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num,
3229 int pos_in_item, int cut_size);
3230void leaf_paste_entries(struct buffer_info *bi, int item_num, int before,
3231 int new_entry_count, struct reiserfs_de_head *new_dehs,
3232 const char *records, int paste_size);
3233/* ibalance.c */
3234int balance_internal(struct tree_balance *, int, int, struct item_head *,
3235 struct buffer_head **);
3236
3237/* do_balance.c */
3238void do_balance_mark_leaf_dirty(struct tree_balance *tb,
3239 struct buffer_head *bh, int flag);
3240#define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
3241#define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
3242
3243void do_balance(struct tree_balance *tb, struct item_head *ih,
3244 const char *body, int flag);
3245void reiserfs_invalidate_buffer(struct tree_balance *tb,
3246 struct buffer_head *bh);
3247
3248int get_left_neighbor_position(struct tree_balance *tb, int h);
3249int get_right_neighbor_position(struct tree_balance *tb, int h);
3250void replace_key(struct tree_balance *tb, struct buffer_head *, int,
3251 struct buffer_head *, int);
3252void make_empty_node(struct buffer_info *);
3253struct buffer_head *get_FEB(struct tree_balance *);
3254
3255/* bitmap.c */
3256
Jeff Mahoney098297b2014-04-23 10:00:36 -04003257/*
3258 * structure contains hints for block allocator, and it is a container for
3259 * arguments, such as node, search path, transaction_handle, etc.
3260 */
Al Virof466c6f2012-03-17 01:16:43 -04003261struct __reiserfs_blocknr_hint {
Jeff Mahoney098297b2014-04-23 10:00:36 -04003262 /* inode passed to allocator, if we allocate unf. nodes */
3263 struct inode *inode;
3264
Al Virof466c6f2012-03-17 01:16:43 -04003265 sector_t block; /* file offset, in blocks */
3266 struct in_core_key key;
Al Virof466c6f2012-03-17 01:16:43 -04003267
Jeff Mahoney098297b2014-04-23 10:00:36 -04003268 /*
3269 * search path, used by allocator to deternine search_start by
3270 * various ways
3271 */
3272 struct treepath *path;
3273
3274 /*
3275 * transaction handle is needed to log super blocks
3276 * and bitmap blocks changes
3277 */
3278 struct reiserfs_transaction_handle *th;
3279
3280 b_blocknr_t beg, end;
3281
3282 /*
3283 * a field used to transfer search start value (block number)
3284 * between different block allocator procedures
3285 * (determine_search_start() and others)
3286 */
3287 b_blocknr_t search_start;
3288
3289 /*
3290 * is set in determine_prealloc_size() function,
3291 * used by underlayed function that do actual allocation
3292 */
3293 int prealloc_size;
3294
3295 /*
3296 * the allocator uses different polices for getting disk
3297 * space for formatted/unformatted blocks with/without preallocation
3298 */
3299 unsigned formatted_node:1;
Al Virof466c6f2012-03-17 01:16:43 -04003300 unsigned preallocate:1;
3301};
3302
3303typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
3304
3305int reiserfs_parse_alloc_options(struct super_block *, char *);
3306void reiserfs_init_alloc_options(struct super_block *s);
3307
3308/*
3309 * given a directory, this will tell you what packing locality
3310 * to use for a new object underneat it. The locality is returned
3311 * in disk byte order (le).
3312 */
3313__le32 reiserfs_choose_packing(struct inode *dir);
3314
Rashika Kheriaea0856c2014-04-03 14:50:36 -07003315void show_alloc_options(struct seq_file *seq, struct super_block *s);
Al Virof466c6f2012-03-17 01:16:43 -04003316int reiserfs_init_bitmap_cache(struct super_block *sb);
3317void reiserfs_free_bitmap_cache(struct super_block *sb);
3318void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info);
3319struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap);
3320int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value);
3321void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *,
3322 b_blocknr_t, int for_unformatted);
3323int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int,
3324 int);
3325static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb,
3326 b_blocknr_t * new_blocknrs,
3327 int amount_needed)
3328{
3329 reiserfs_blocknr_hint_t hint = {
3330 .th = tb->transaction_handle,
3331 .path = tb->tb_path,
3332 .inode = NULL,
3333 .key = tb->key,
3334 .block = 0,
3335 .formatted_node = 1
3336 };
3337 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed,
3338 0);
3339}
3340
3341static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle
3342 *th, struct inode *inode,
3343 b_blocknr_t * new_blocknrs,
3344 struct treepath *path,
3345 sector_t block)
3346{
3347 reiserfs_blocknr_hint_t hint = {
3348 .th = th,
3349 .path = path,
3350 .inode = inode,
3351 .block = block,
3352 .formatted_node = 0,
3353 .preallocate = 0
3354 };
3355 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
3356}
3357
3358#ifdef REISERFS_PREALLOCATE
3359static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle
3360 *th, struct inode *inode,
3361 b_blocknr_t * new_blocknrs,
3362 struct treepath *path,
3363 sector_t block)
3364{
3365 reiserfs_blocknr_hint_t hint = {
3366 .th = th,
3367 .path = path,
3368 .inode = inode,
3369 .block = block,
3370 .formatted_node = 0,
3371 .preallocate = 1
3372 };
3373 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
3374}
3375
3376void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
3377 struct inode *inode);
3378void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th);
3379#endif
3380
3381/* hashes.c */
3382__u32 keyed_hash(const signed char *msg, int len);
3383__u32 yura_hash(const signed char *msg, int len);
3384__u32 r5_hash(const signed char *msg, int len);
3385
3386#define reiserfs_set_le_bit __set_bit_le
3387#define reiserfs_test_and_set_le_bit __test_and_set_bit_le
3388#define reiserfs_clear_le_bit __clear_bit_le
3389#define reiserfs_test_and_clear_le_bit __test_and_clear_bit_le
3390#define reiserfs_test_le_bit test_bit_le
3391#define reiserfs_find_next_zero_le_bit find_next_zero_bit_le
3392
Jeff Mahoney098297b2014-04-23 10:00:36 -04003393/*
3394 * sometimes reiserfs_truncate may require to allocate few new blocks
3395 * to perform indirect2direct conversion. People probably used to
3396 * think, that truncate should work without problems on a filesystem
3397 * without free disk space. They may complain that they can not
3398 * truncate due to lack of free disk space. This spare space allows us
3399 * to not worry about it. 500 is probably too much, but it should be
3400 * absolutely safe
3401 */
Al Virof466c6f2012-03-17 01:16:43 -04003402#define SPARE_SPACE 500
3403
3404/* prototypes from ioctl.c */
3405long reiserfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg);
3406long reiserfs_compat_ioctl(struct file *filp,
3407 unsigned int cmd, unsigned long arg);
3408int reiserfs_unpack(struct inode *inode, struct file *filp);