blob: 2adcde137c3fb8d569cddcdf83dfd443a11f693f [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
Jan Kara53873632014-09-29 15:08:19 +0200100#ifdef CONFIG_QUOTA
101 struct dquot *i_dquot[MAXQUOTAS];
102#endif
103
Al Viro765fd6b2012-03-17 01:19:24 -0400104 struct inode vfs_inode;
105};
106
107typedef enum {
108 reiserfs_attrs_cleared = 0x00000001,
109} reiserfs_super_block_flags;
110
Jeff Mahoney098297b2014-04-23 10:00:36 -0400111/*
112 * struct reiserfs_super_block accessors/mutators since this is a disk
113 * structure, it will always be in little endian format.
114 */
Al Viro765fd6b2012-03-17 01:19:24 -0400115#define sb_block_count(sbp) (le32_to_cpu((sbp)->s_v1.s_block_count))
116#define set_sb_block_count(sbp,v) ((sbp)->s_v1.s_block_count = cpu_to_le32(v))
117#define sb_free_blocks(sbp) (le32_to_cpu((sbp)->s_v1.s_free_blocks))
118#define set_sb_free_blocks(sbp,v) ((sbp)->s_v1.s_free_blocks = cpu_to_le32(v))
119#define sb_root_block(sbp) (le32_to_cpu((sbp)->s_v1.s_root_block))
120#define set_sb_root_block(sbp,v) ((sbp)->s_v1.s_root_block = cpu_to_le32(v))
121
122#define sb_jp_journal_1st_block(sbp) \
123 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_1st_block))
124#define set_sb_jp_journal_1st_block(sbp,v) \
125 ((sbp)->s_v1.s_journal.jp_journal_1st_block = cpu_to_le32(v))
126#define sb_jp_journal_dev(sbp) \
127 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_dev))
128#define set_sb_jp_journal_dev(sbp,v) \
129 ((sbp)->s_v1.s_journal.jp_journal_dev = cpu_to_le32(v))
130#define sb_jp_journal_size(sbp) \
131 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_size))
132#define set_sb_jp_journal_size(sbp,v) \
133 ((sbp)->s_v1.s_journal.jp_journal_size = cpu_to_le32(v))
134#define sb_jp_journal_trans_max(sbp) \
135 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_trans_max))
136#define set_sb_jp_journal_trans_max(sbp,v) \
137 ((sbp)->s_v1.s_journal.jp_journal_trans_max = cpu_to_le32(v))
138#define sb_jp_journal_magic(sbp) \
139 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_magic))
140#define set_sb_jp_journal_magic(sbp,v) \
141 ((sbp)->s_v1.s_journal.jp_journal_magic = cpu_to_le32(v))
142#define sb_jp_journal_max_batch(sbp) \
143 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_batch))
144#define set_sb_jp_journal_max_batch(sbp,v) \
145 ((sbp)->s_v1.s_journal.jp_journal_max_batch = cpu_to_le32(v))
146#define sb_jp_jourmal_max_commit_age(sbp) \
147 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_commit_age))
148#define set_sb_jp_journal_max_commit_age(sbp,v) \
149 ((sbp)->s_v1.s_journal.jp_journal_max_commit_age = cpu_to_le32(v))
150
151#define sb_blocksize(sbp) (le16_to_cpu((sbp)->s_v1.s_blocksize))
152#define set_sb_blocksize(sbp,v) ((sbp)->s_v1.s_blocksize = cpu_to_le16(v))
153#define sb_oid_maxsize(sbp) (le16_to_cpu((sbp)->s_v1.s_oid_maxsize))
154#define set_sb_oid_maxsize(sbp,v) ((sbp)->s_v1.s_oid_maxsize = cpu_to_le16(v))
155#define sb_oid_cursize(sbp) (le16_to_cpu((sbp)->s_v1.s_oid_cursize))
156#define set_sb_oid_cursize(sbp,v) ((sbp)->s_v1.s_oid_cursize = cpu_to_le16(v))
157#define sb_umount_state(sbp) (le16_to_cpu((sbp)->s_v1.s_umount_state))
158#define set_sb_umount_state(sbp,v) ((sbp)->s_v1.s_umount_state = cpu_to_le16(v))
159#define sb_fs_state(sbp) (le16_to_cpu((sbp)->s_v1.s_fs_state))
160#define set_sb_fs_state(sbp,v) ((sbp)->s_v1.s_fs_state = cpu_to_le16(v))
161#define sb_hash_function_code(sbp) \
162 (le32_to_cpu((sbp)->s_v1.s_hash_function_code))
163#define set_sb_hash_function_code(sbp,v) \
164 ((sbp)->s_v1.s_hash_function_code = cpu_to_le32(v))
165#define sb_tree_height(sbp) (le16_to_cpu((sbp)->s_v1.s_tree_height))
166#define set_sb_tree_height(sbp,v) ((sbp)->s_v1.s_tree_height = cpu_to_le16(v))
167#define sb_bmap_nr(sbp) (le16_to_cpu((sbp)->s_v1.s_bmap_nr))
168#define set_sb_bmap_nr(sbp,v) ((sbp)->s_v1.s_bmap_nr = cpu_to_le16(v))
169#define sb_version(sbp) (le16_to_cpu((sbp)->s_v1.s_version))
170#define set_sb_version(sbp,v) ((sbp)->s_v1.s_version = cpu_to_le16(v))
171
172#define sb_mnt_count(sbp) (le16_to_cpu((sbp)->s_mnt_count))
173#define set_sb_mnt_count(sbp, v) ((sbp)->s_mnt_count = cpu_to_le16(v))
174
175#define sb_reserved_for_journal(sbp) \
176 (le16_to_cpu((sbp)->s_v1.s_reserved_for_journal))
177#define set_sb_reserved_for_journal(sbp,v) \
178 ((sbp)->s_v1.s_reserved_for_journal = cpu_to_le16(v))
179
180/* LOGGING -- */
181
Jeff Mahoney098297b2014-04-23 10:00:36 -0400182/*
183 * These all interelate for performance.
184 *
185 * If the journal block count is smaller than n transactions, you lose speed.
186 * I don't know what n is yet, I'm guessing 8-16.
187 *
188 * typical transaction size depends on the application, how often fsync is
189 * called, and how many metadata blocks you dirty in a 30 second period.
190 * The more small files (<16k) you use, the larger your transactions will
191 * be.
192 *
193 * If your journal fills faster than dirty buffers get flushed to disk, it
194 * must flush them before allowing the journal to wrap, which slows things
195 * down. If you need high speed meta data updates, the journal should be
196 * big enough to prevent wrapping before dirty meta blocks get to disk.
197 *
198 * If the batch max is smaller than the transaction max, you'll waste space
199 * at the end of the journal because journal_end sets the next transaction
200 * to start at 0 if the next transaction has any chance of wrapping.
201 *
202 * The large the batch max age, the better the speed, and the more meta
203 * data changes you'll lose after a crash.
204 */
Al Viro765fd6b2012-03-17 01:19:24 -0400205
206/* don't mess with these for a while */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400207/* we have a node size define somewhere in reiserfs_fs.h. -Hans */
Al Viro765fd6b2012-03-17 01:19:24 -0400208#define JOURNAL_BLOCK_SIZE 4096 /* BUG gotta get rid of this */
209#define JOURNAL_MAX_CNODE 1500 /* max cnodes to allocate. */
210#define JOURNAL_HASH_SIZE 8192
Al Viro765fd6b2012-03-17 01:19:24 -0400211
Jeff Mahoney098297b2014-04-23 10:00:36 -0400212/* number of copies of the bitmaps to have floating. Must be >= 2 */
213#define JOURNAL_NUM_BITMAPS 5
214
215/*
216 * One of these for every block in every transaction
217 * Each one is in two hash tables. First, a hash of the current transaction,
218 * and after journal_end, a hash of all the in memory transactions.
219 * next and prev are used by the current transaction (journal_hash).
220 * hnext and hprev are used by journal_list_hash. If a block is in more
221 * than one transaction, the journal_list_hash links it in multiple times.
222 * This allows flush_journal_list to remove just the cnode belonging to a
223 * given transaction.
224 */
Al Viro765fd6b2012-03-17 01:19:24 -0400225struct reiserfs_journal_cnode {
226 struct buffer_head *bh; /* real buffer head */
227 struct super_block *sb; /* dev of real buffer head */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400228
229 /* block number of real buffer head, == 0 when buffer on disk */
230 __u32 blocknr;
231
Al Viro765fd6b2012-03-17 01:19:24 -0400232 unsigned long state;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400233
234 /* journal list this cnode lives in */
235 struct reiserfs_journal_list *jlist;
236
Al Viro765fd6b2012-03-17 01:19:24 -0400237 struct reiserfs_journal_cnode *next; /* next in transaction list */
238 struct reiserfs_journal_cnode *prev; /* prev in transaction list */
239 struct reiserfs_journal_cnode *hprev; /* prev in hash list */
240 struct reiserfs_journal_cnode *hnext; /* next in hash list */
241};
242
243struct reiserfs_bitmap_node {
244 int id;
245 char *data;
246 struct list_head list;
247};
248
249struct reiserfs_list_bitmap {
250 struct reiserfs_journal_list *journal_list;
251 struct reiserfs_bitmap_node **bitmaps;
252};
253
254/*
Jeff Mahoney098297b2014-04-23 10:00:36 -0400255 * one of these for each transaction. The most important part here is the
256 * j_realblock. this list of cnodes is used to hash all the blocks in all
257 * the commits, to mark all the real buffer heads dirty once all the commits
258 * hit the disk, and to make sure every real block in a transaction is on
259 * disk before allowing the log area to be overwritten
260 */
Al Viro765fd6b2012-03-17 01:19:24 -0400261struct reiserfs_journal_list {
262 unsigned long j_start;
263 unsigned long j_state;
264 unsigned long j_len;
265 atomic_t j_nonzerolen;
266 atomic_t j_commit_left;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400267
268 /* all commits older than this on disk */
269 atomic_t j_older_commits_done;
270
Al Viro765fd6b2012-03-17 01:19:24 -0400271 struct mutex j_commit_mutex;
272 unsigned int j_trans_id;
273 time_t j_timestamp;
274 struct reiserfs_list_bitmap *j_list_bitmap;
275 struct buffer_head *j_commit_bh; /* commit buffer head */
276 struct reiserfs_journal_cnode *j_realblock;
277 struct reiserfs_journal_cnode *j_freedlist; /* list of buffers that were freed during this trans. free each of these on flush */
278 /* time ordered list of all active transactions */
279 struct list_head j_list;
280
Jeff Mahoney098297b2014-04-23 10:00:36 -0400281 /*
282 * time ordered list of all transactions we haven't tried
283 * to flush yet
284 */
Al Viro765fd6b2012-03-17 01:19:24 -0400285 struct list_head j_working_list;
286
287 /* list of tail conversion targets in need of flush before commit */
288 struct list_head j_tail_bh_list;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400289
Al Viro765fd6b2012-03-17 01:19:24 -0400290 /* list of data=ordered buffers in need of flush before commit */
291 struct list_head j_bh_list;
292 int j_refcount;
293};
294
295struct reiserfs_journal {
296 struct buffer_head **j_ap_blocks; /* journal blocks on disk */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400297 /* newest journal block */
298 struct reiserfs_journal_cnode *j_last;
299
300 /* oldest journal block. start here for traverse */
301 struct reiserfs_journal_cnode *j_first;
Al Viro765fd6b2012-03-17 01:19:24 -0400302
303 struct block_device *j_dev_bd;
304 fmode_t j_dev_mode;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400305
306 /* first block on s_dev of reserved area journal */
307 int j_1st_reserved_block;
Al Viro765fd6b2012-03-17 01:19:24 -0400308
309 unsigned long j_state;
310 unsigned int j_trans_id;
311 unsigned long j_mount_id;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400312
313 /* start of current waiting commit (index into j_ap_blocks) */
314 unsigned long j_start;
Al Viro765fd6b2012-03-17 01:19:24 -0400315 unsigned long j_len; /* length of current waiting commit */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400316
317 /* number of buffers requested by journal_begin() */
318 unsigned long j_len_alloc;
319
Al Viro765fd6b2012-03-17 01:19:24 -0400320 atomic_t j_wcount; /* count of writers for current commit */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400321
322 /* batch count. allows turning X transactions into 1 */
323 unsigned long j_bcount;
324
325 /* first unflushed transactions offset */
326 unsigned long j_first_unflushed_offset;
327
328 /* last fully flushed journal timestamp */
329 unsigned j_last_flush_trans_id;
330
Al Viro765fd6b2012-03-17 01:19:24 -0400331 struct buffer_head *j_header_bh;
332
333 time_t j_trans_start_time; /* time this transaction started */
334 struct mutex j_mutex;
335 struct mutex j_flush_mutex;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400336
337 /* wait for current transaction to finish before starting new one */
338 wait_queue_head_t j_join_wait;
339
340 atomic_t j_jlock; /* lock for j_join_wait */
Al Viro765fd6b2012-03-17 01:19:24 -0400341 int j_list_bitmap_index; /* number of next list bitmap to use */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400342
343 /* no more journal begins allowed. MUST sleep on j_join_wait */
344 int j_must_wait;
345
346 /* next journal_end will flush all journal list */
347 int j_next_full_flush;
348
349 /* next journal_end will flush all async commits */
350 int j_next_async_flush;
Al Viro765fd6b2012-03-17 01:19:24 -0400351
352 int j_cnode_used; /* number of cnodes on the used list */
353 int j_cnode_free; /* number of cnodes on the free list */
354
Jeff Mahoney098297b2014-04-23 10:00:36 -0400355 /* max number of blocks in a transaction. */
356 unsigned int j_trans_max;
357
358 /* max number of blocks to batch into a trans */
359 unsigned int j_max_batch;
360
361 /* in seconds, how old can an async commit be */
362 unsigned int j_max_commit_age;
363
364 /* in seconds, how old can a transaction be */
365 unsigned int j_max_trans_age;
366
367 /* the default for the max commit age */
368 unsigned int j_default_max_commit_age;
Al Viro765fd6b2012-03-17 01:19:24 -0400369
370 struct reiserfs_journal_cnode *j_cnode_free_list;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400371
372 /* orig pointer returned from vmalloc */
373 struct reiserfs_journal_cnode *j_cnode_free_orig;
Al Viro765fd6b2012-03-17 01:19:24 -0400374
375 struct reiserfs_journal_list *j_current_jl;
376 int j_free_bitmap_nodes;
377 int j_used_bitmap_nodes;
378
379 int j_num_lists; /* total number of active transactions */
380 int j_num_work_lists; /* number that need attention from kreiserfsd */
381
382 /* debugging to make sure things are flushed in order */
383 unsigned int j_last_flush_id;
384
385 /* debugging to make sure things are committed in order */
386 unsigned int j_last_commit_id;
387
388 struct list_head j_bitmap_nodes;
389 struct list_head j_dirty_buffers;
390 spinlock_t j_dirty_buffers_lock; /* protects j_dirty_buffers */
391
392 /* list of all active transactions */
393 struct list_head j_journal_list;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400394
Al Viro765fd6b2012-03-17 01:19:24 -0400395 /* lists that haven't been touched by writeback attempts */
396 struct list_head j_working_list;
397
Jeff Mahoney098297b2014-04-23 10:00:36 -0400398 /* hash table for real buffer heads in current trans */
399 struct reiserfs_journal_cnode *j_hash_table[JOURNAL_HASH_SIZE];
400
401 /* hash table for all the real buffer heads in all the transactions */
402 struct reiserfs_journal_cnode *j_list_hash_table[JOURNAL_HASH_SIZE];
403
404 /* array of bitmaps to record the deleted blocks */
405 struct reiserfs_list_bitmap j_list_bitmap[JOURNAL_NUM_BITMAPS];
406
407 /* list of inodes which have preallocated blocks */
408 struct list_head j_prealloc_list;
Al Viro765fd6b2012-03-17 01:19:24 -0400409 int j_persistent_trans;
410 unsigned long j_max_trans_size;
411 unsigned long j_max_batch_size;
412
413 int j_errno;
414
415 /* when flushing ordered buffers, throttle new ordered writers */
416 struct delayed_work j_work;
417 struct super_block *j_work_sb;
418 atomic_t j_async_throttle;
419};
420
421enum journal_state_bits {
422 J_WRITERS_BLOCKED = 1, /* set when new writers not allowed */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400423 J_WRITERS_QUEUED, /* set when log is full due to too many writers */
424 J_ABORTED, /* set when log is aborted */
Al Viro765fd6b2012-03-17 01:19:24 -0400425};
426
Jeff Mahoney098297b2014-04-23 10:00:36 -0400427/* ick. magic string to find desc blocks in the journal */
428#define JOURNAL_DESC_MAGIC "ReIsErLB"
Al Viro765fd6b2012-03-17 01:19:24 -0400429
430typedef __u32(*hashf_t) (const signed char *, int);
431
432struct reiserfs_bitmap_info {
433 __u32 free_count;
434};
435
436struct proc_dir_entry;
437
438#if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )
439typedef unsigned long int stat_cnt_t;
440typedef struct reiserfs_proc_info_data {
441 spinlock_t lock;
442 int exiting;
443 int max_hash_collisions;
444
445 stat_cnt_t breads;
446 stat_cnt_t bread_miss;
447 stat_cnt_t search_by_key;
448 stat_cnt_t search_by_key_fs_changed;
449 stat_cnt_t search_by_key_restarted;
450
451 stat_cnt_t insert_item_restarted;
452 stat_cnt_t paste_into_item_restarted;
453 stat_cnt_t cut_from_item_restarted;
454 stat_cnt_t delete_solid_item_restarted;
455 stat_cnt_t delete_item_restarted;
456
457 stat_cnt_t leaked_oid;
458 stat_cnt_t leaves_removable;
459
Jeff Mahoney098297b2014-04-23 10:00:36 -0400460 /*
461 * balances per level.
462 * Use explicit 5 as MAX_HEIGHT is not visible yet.
463 */
Al Viro765fd6b2012-03-17 01:19:24 -0400464 stat_cnt_t balance_at[5]; /* XXX */
465 /* sbk == search_by_key */
466 stat_cnt_t sbk_read_at[5]; /* XXX */
467 stat_cnt_t sbk_fs_changed[5];
468 stat_cnt_t sbk_restarted[5];
469 stat_cnt_t items_at[5]; /* XXX */
470 stat_cnt_t free_at[5]; /* XXX */
471 stat_cnt_t can_node_be_removed[5]; /* XXX */
472 long int lnum[5]; /* XXX */
473 long int rnum[5]; /* XXX */
474 long int lbytes[5]; /* XXX */
475 long int rbytes[5]; /* XXX */
476 stat_cnt_t get_neighbors[5];
477 stat_cnt_t get_neighbors_restart[5];
478 stat_cnt_t need_l_neighbor[5];
479 stat_cnt_t need_r_neighbor[5];
480
481 stat_cnt_t free_block;
482 struct __scan_bitmap_stats {
483 stat_cnt_t call;
484 stat_cnt_t wait;
485 stat_cnt_t bmap;
486 stat_cnt_t retry;
487 stat_cnt_t in_journal_hint;
488 stat_cnt_t in_journal_nohint;
489 stat_cnt_t stolen;
490 } scan_bitmap;
491 struct __journal_stats {
492 stat_cnt_t in_journal;
493 stat_cnt_t in_journal_bitmap;
494 stat_cnt_t in_journal_reusable;
495 stat_cnt_t lock_journal;
496 stat_cnt_t lock_journal_wait;
497 stat_cnt_t journal_being;
498 stat_cnt_t journal_relock_writers;
499 stat_cnt_t journal_relock_wcount;
500 stat_cnt_t mark_dirty;
501 stat_cnt_t mark_dirty_already;
502 stat_cnt_t mark_dirty_notjournal;
503 stat_cnt_t restore_prepared;
504 stat_cnt_t prepare;
505 stat_cnt_t prepare_retry;
506 } journal;
507} reiserfs_proc_info_data_t;
508#else
509typedef struct reiserfs_proc_info_data {
510} reiserfs_proc_info_data_t;
511#endif
512
Jan Karaaca60612014-09-10 20:57:45 +0200513/* Number of quota types we support */
514#define REISERFS_MAXQUOTAS 2
515
Al Viro765fd6b2012-03-17 01:19:24 -0400516/* reiserfs union of in-core super block data */
517struct reiserfs_sb_info {
Jeff Mahoney098297b2014-04-23 10:00:36 -0400518 /* Buffer containing the super block */
519 struct buffer_head *s_sbh;
520
521 /* Pointer to the on-disk super block in the buffer */
522 struct reiserfs_super_block *s_rs;
Al Viro765fd6b2012-03-17 01:19:24 -0400523 struct reiserfs_bitmap_info *s_ap_bitmap;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400524
525 /* pointer to journal information */
526 struct reiserfs_journal *s_journal;
527
Al Viro765fd6b2012-03-17 01:19:24 -0400528 unsigned short s_mount_state; /* reiserfs state (valid, invalid) */
529
530 /* Serialize writers access, replace the old bkl */
531 struct mutex lock;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400532
Al Viro765fd6b2012-03-17 01:19:24 -0400533 /* Owner of the lock (can be recursive) */
534 struct task_struct *lock_owner;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400535
Al Viro765fd6b2012-03-17 01:19:24 -0400536 /* Depth of the lock, start from -1 like the bkl */
537 int lock_depth;
538
Jeff Mahoney797d9012014-04-23 10:00:34 -0400539 struct workqueue_struct *commit_wq;
540
Al Viro765fd6b2012-03-17 01:19:24 -0400541 /* Comment? -Hans */
542 void (*end_io_handler) (struct buffer_head *, int);
Al Viro765fd6b2012-03-17 01:19:24 -0400543
Jeff Mahoney098297b2014-04-23 10:00:36 -0400544 /*
545 * pointer to function which is used to sort names in directory.
546 * Set on mount
547 */
548 hashf_t s_hash_function;
549
550 /* reiserfs's mount options are set here */
551 unsigned long s_mount_opt;
552
553 /* This is a structure that describes block allocator options */
554 struct {
555 /* Bitfield for enable/disable kind of options */
556 unsigned long bits;
557
558 /*
559 * size started from which we consider file
560 * to be a large one (in blocks)
561 */
562 unsigned long large_file_size;
563
Al Viro765fd6b2012-03-17 01:19:24 -0400564 int border; /* percentage of disk, border takes */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400565
566 /*
567 * Minimal file size (in blocks) starting
568 * from which we do preallocations
569 */
570 int preallocmin;
571
572 /*
573 * Number of blocks we try to prealloc when file
574 * reaches preallocmin size (in blocks) or prealloc_list
575 is empty.
576 */
577 int preallocsize;
Al Viro765fd6b2012-03-17 01:19:24 -0400578 } s_alloc_options;
579
580 /* Comment? -Hans */
581 wait_queue_head_t s_wait;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400582 /* increased by one every time the tree gets re-balanced */
583 atomic_t s_generation_counter;
584
585 /* File system properties. Currently holds on-disk FS format */
586 unsigned long s_properties;
Al Viro765fd6b2012-03-17 01:19:24 -0400587
588 /* session statistics */
589 int s_disk_reads;
590 int s_disk_writes;
591 int s_fix_nodes;
592 int s_do_balance;
593 int s_unneeded_left_neighbor;
594 int s_good_search_by_key_reada;
595 int s_bmaps;
596 int s_bmaps_without_search;
597 int s_direct2indirect;
598 int s_indirect2direct;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400599
600 /*
601 * set up when it's ok for reiserfs_read_inode2() to read from
602 * disk inode with nlink==0. Currently this is only used during
603 * finish_unfinished() processing at mount time
604 */
Al Viro765fd6b2012-03-17 01:19:24 -0400605 int s_is_unlinked_ok;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400606
Al Viro765fd6b2012-03-17 01:19:24 -0400607 reiserfs_proc_info_data_t s_proc_info_data;
608 struct proc_dir_entry *procdir;
Jeff Mahoney098297b2014-04-23 10:00:36 -0400609
610 /* amount of blocks reserved for further allocations */
611 int reserved_blocks;
612
613
614 /* this lock on now only used to protect reserved_blocks variable */
615 spinlock_t bitmap_lock;
Al Viro765fd6b2012-03-17 01:19:24 -0400616 struct dentry *priv_root; /* root of /.reiserfs_priv */
617 struct dentry *xattr_root; /* root of /.reiserfs_priv/xattrs */
618 int j_errno;
Artem Bityutskiy033369d2012-06-01 17:18:08 +0300619
620 int work_queued; /* non-zero delayed work is queued */
621 struct delayed_work old_work; /* old transactions flush delayed work */
622 spinlock_t old_work_lock; /* protects old_work and work_queued */
623
Al Viro765fd6b2012-03-17 01:19:24 -0400624#ifdef CONFIG_QUOTA
Jan Karaaca60612014-09-10 20:57:45 +0200625 char *s_qf_names[REISERFS_MAXQUOTAS];
Al Viro765fd6b2012-03-17 01:19:24 -0400626 int s_jquota_fmt;
627#endif
628 char *s_jdev; /* Stored jdev for mount option showing */
629#ifdef CONFIG_REISERFS_CHECK
630
Jeff Mahoney098297b2014-04-23 10:00:36 -0400631 /*
632 * Detects whether more than one copy of tb exists per superblock
633 * as a means of checking whether do_balance is executing
634 * concurrently against another tree reader/writer on a same
635 * mount point.
636 */
637 struct tree_balance *cur_tb;
Al Viro765fd6b2012-03-17 01:19:24 -0400638#endif
639};
640
641/* Definitions of reiserfs on-disk properties: */
642#define REISERFS_3_5 0
643#define REISERFS_3_6 1
644#define REISERFS_OLD_FORMAT 2
645
Al Viro765fd6b2012-03-17 01:19:24 -0400646/* Mount options */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400647enum reiserfs_mount_options {
648 /* large tails will be created in a session */
649 REISERFS_LARGETAIL,
650 /*
651 * small (for files less than block size) tails will
652 * be created in a session
653 */
654 REISERFS_SMALLTAIL,
Al Viro765fd6b2012-03-17 01:19:24 -0400655
Jeff Mahoney098297b2014-04-23 10:00:36 -0400656 /* replay journal and return 0. Use by fsck */
657 REPLAYONLY,
658
659 /*
660 * -o conv: causes conversion of old format super block to the
661 * new format. If not specified - old partition will be dealt
662 * with in a manner of 3.5.x
663 */
664 REISERFS_CONVERT,
665
666 /*
667 * -o hash={tea, rupasov, r5, detect} is meant for properly mounting
668 * reiserfs disks from 3.5.19 or earlier. 99% of the time, this
669 * option is not required. If the normal autodection code can't
670 * determine which hash to use (because both hashes had the same
671 * value for a file) use this option to force a specific hash.
672 * It won't allow you to override the existing hash on the FS, so
673 * if you have a tea hash disk, and mount with -o hash=rupasov,
674 * the mount will fail.
675 */
Al Viro765fd6b2012-03-17 01:19:24 -0400676 FORCE_TEA_HASH, /* try to force tea hash on mount */
677 FORCE_RUPASOV_HASH, /* try to force rupasov hash on mount */
678 FORCE_R5_HASH, /* try to force rupasov hash on mount */
679 FORCE_HASH_DETECT, /* try to detect hash function on mount */
680
681 REISERFS_DATA_LOG,
682 REISERFS_DATA_ORDERED,
683 REISERFS_DATA_WRITEBACK,
684
Jeff Mahoney098297b2014-04-23 10:00:36 -0400685 /*
686 * used for testing experimental features, makes benchmarking new
687 * features with and without more convenient, should never be used by
688 * users in any code shipped to users (ideally)
689 */
Al Viro765fd6b2012-03-17 01:19:24 -0400690
691 REISERFS_NO_BORDER,
692 REISERFS_NO_UNHASHED_RELOCATION,
693 REISERFS_HASHED_RELOCATION,
694 REISERFS_ATTRS,
695 REISERFS_XATTRS_USER,
696 REISERFS_POSIXACL,
697 REISERFS_EXPOSE_PRIVROOT,
698 REISERFS_BARRIER_NONE,
699 REISERFS_BARRIER_FLUSH,
700
701 /* Actions on error */
702 REISERFS_ERROR_PANIC,
703 REISERFS_ERROR_RO,
704 REISERFS_ERROR_CONTINUE,
705
706 REISERFS_USRQUOTA, /* User quota option specified */
707 REISERFS_GRPQUOTA, /* Group quota option specified */
708
709 REISERFS_TEST1,
710 REISERFS_TEST2,
711 REISERFS_TEST3,
712 REISERFS_TEST4,
713 REISERFS_UNSUPPORTED_OPT,
714};
715
716#define reiserfs_r5_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_R5_HASH))
717#define reiserfs_rupasov_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_RUPASOV_HASH))
718#define reiserfs_tea_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_TEA_HASH))
719#define reiserfs_hash_detect(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_HASH_DETECT))
720#define reiserfs_no_border(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_BORDER))
721#define reiserfs_no_unhashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_UNHASHED_RELOCATION))
722#define reiserfs_hashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_HASHED_RELOCATION))
723#define reiserfs_test4(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_TEST4))
724
725#define have_large_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_LARGETAIL))
726#define have_small_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_SMALLTAIL))
727#define replay_only(s) (REISERFS_SB(s)->s_mount_opt & (1 << REPLAYONLY))
728#define reiserfs_attrs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ATTRS))
729#define old_format_only(s) (REISERFS_SB(s)->s_properties & (1 << REISERFS_3_5))
730#define convert_reiserfs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_CONVERT))
731#define reiserfs_data_log(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_LOG))
732#define reiserfs_data_ordered(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_ORDERED))
733#define reiserfs_data_writeback(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_WRITEBACK))
734#define reiserfs_xattrs_user(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_XATTRS_USER))
735#define reiserfs_posixacl(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_POSIXACL))
736#define reiserfs_expose_privroot(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_EXPOSE_PRIVROOT))
737#define reiserfs_xattrs_optional(s) (reiserfs_xattrs_user(s) || reiserfs_posixacl(s))
738#define reiserfs_barrier_none(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_NONE))
739#define reiserfs_barrier_flush(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_FLUSH))
740
741#define reiserfs_error_panic(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_PANIC))
742#define reiserfs_error_ro(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_RO))
743
744void reiserfs_file_buffer(struct buffer_head *bh, int list);
745extern struct file_system_type reiserfs_fs_type;
746int reiserfs_resize(struct super_block *, unsigned long);
747
748#define CARRY_ON 0
749#define SCHEDULE_OCCURRED 1
750
751#define SB_BUFFER_WITH_SB(s) (REISERFS_SB(s)->s_sbh)
752#define SB_JOURNAL(s) (REISERFS_SB(s)->s_journal)
753#define SB_JOURNAL_1st_RESERVED_BLOCK(s) (SB_JOURNAL(s)->j_1st_reserved_block)
754#define SB_JOURNAL_LEN_FREE(s) (SB_JOURNAL(s)->j_journal_len_free)
755#define SB_AP_BITMAP(s) (REISERFS_SB(s)->s_ap_bitmap)
756
757#define SB_DISK_JOURNAL_HEAD(s) (SB_JOURNAL(s)->j_header_bh->)
758
Al Viro765fd6b2012-03-17 01:19:24 -0400759#define reiserfs_is_journal_aborted(journal) (unlikely (__reiserfs_is_journal_aborted (journal)))
760static inline int __reiserfs_is_journal_aborted(struct reiserfs_journal
761 *journal)
762{
763 return test_bit(J_ABORTED, &journal->j_state);
764}
765
Al Virof466c6f2012-03-17 01:16:43 -0400766/*
767 * Locking primitives. The write lock is a per superblock
768 * special mutex that has properties close to the Big Kernel Lock
769 * which was used in the previous locking scheme.
770 */
771void reiserfs_write_lock(struct super_block *s);
772void reiserfs_write_unlock(struct super_block *s);
Jeff Mahoney278f6672013-08-08 17:34:46 -0400773int __must_check reiserfs_write_unlock_nested(struct super_block *s);
774void reiserfs_write_lock_nested(struct super_block *s, int depth);
Al Virof466c6f2012-03-17 01:16:43 -0400775
776#ifdef CONFIG_REISERFS_CHECK
777void reiserfs_lock_check_recursive(struct super_block *s);
778#else
779static inline void reiserfs_lock_check_recursive(struct super_block *s) { }
780#endif
781
782/*
783 * Several mutexes depend on the write lock.
784 * However sometimes we want to relax the write lock while we hold
785 * these mutexes, according to the release/reacquire on schedule()
786 * properties of the Bkl that were used.
787 * Reiserfs performances and locking were based on this scheme.
788 * Now that the write lock is a mutex and not the bkl anymore, doing so
789 * may result in a deadlock:
790 *
791 * A acquire write_lock
792 * A acquire j_commit_mutex
793 * A release write_lock and wait for something
794 * B acquire write_lock
795 * B can't acquire j_commit_mutex and sleep
796 * A can't acquire write lock anymore
797 * deadlock
798 *
799 * What we do here is avoiding such deadlock by playing the same game
800 * than the Bkl: if we can't acquire a mutex that depends on the write lock,
801 * we release the write lock, wait a bit and then retry.
802 *
803 * The mutexes concerned by this hack are:
804 * - The commit mutex of a journal list
805 * - The flush mutex
806 * - The journal lock
807 * - The inode mutex
808 */
809static inline void reiserfs_mutex_lock_safe(struct mutex *m,
Jeff Mahoney278f6672013-08-08 17:34:46 -0400810 struct super_block *s)
Al Virof466c6f2012-03-17 01:16:43 -0400811{
Jeff Mahoney278f6672013-08-08 17:34:46 -0400812 int depth;
813
814 depth = reiserfs_write_unlock_nested(s);
Al Virof466c6f2012-03-17 01:16:43 -0400815 mutex_lock(m);
Jeff Mahoney278f6672013-08-08 17:34:46 -0400816 reiserfs_write_lock_nested(s, depth);
Al Virof466c6f2012-03-17 01:16:43 -0400817}
818
819static inline void
820reiserfs_mutex_lock_nested_safe(struct mutex *m, unsigned int subclass,
Jeff Mahoney278f6672013-08-08 17:34:46 -0400821 struct super_block *s)
Al Virof466c6f2012-03-17 01:16:43 -0400822{
Jeff Mahoney278f6672013-08-08 17:34:46 -0400823 int depth;
824
825 depth = reiserfs_write_unlock_nested(s);
Al Virof466c6f2012-03-17 01:16:43 -0400826 mutex_lock_nested(m, subclass);
Jeff Mahoney278f6672013-08-08 17:34:46 -0400827 reiserfs_write_lock_nested(s, depth);
Al Virof466c6f2012-03-17 01:16:43 -0400828}
829
830static inline void
831reiserfs_down_read_safe(struct rw_semaphore *sem, struct super_block *s)
832{
Jeff Mahoney278f6672013-08-08 17:34:46 -0400833 int depth;
834 depth = reiserfs_write_unlock_nested(s);
835 down_read(sem);
836 reiserfs_write_lock_nested(s, depth);
Al Virof466c6f2012-03-17 01:16:43 -0400837}
838
839/*
840 * When we schedule, we usually want to also release the write lock,
841 * according to the previous bkl based locking scheme of reiserfs.
842 */
843static inline void reiserfs_cond_resched(struct super_block *s)
844{
845 if (need_resched()) {
Jeff Mahoney278f6672013-08-08 17:34:46 -0400846 int depth;
847
848 depth = reiserfs_write_unlock_nested(s);
Al Virof466c6f2012-03-17 01:16:43 -0400849 schedule();
Jeff Mahoney278f6672013-08-08 17:34:46 -0400850 reiserfs_write_lock_nested(s, depth);
Al Virof466c6f2012-03-17 01:16:43 -0400851 }
852}
853
854struct fid;
855
Jeff Mahoney098297b2014-04-23 10:00:36 -0400856/*
857 * in reading the #defines, it may help to understand that they employ
858 * the following abbreviations:
859 *
860 * B = Buffer
861 * I = Item header
862 * H = Height within the tree (should be changed to LEV)
863 * N = Number of the item in the node
864 * STAT = stat data
865 * DEH = Directory Entry Header
866 * EC = Entry Count
867 * E = Entry number
868 * UL = Unsigned Long
869 * BLKH = BLocK Header
870 * UNFM = UNForMatted node
871 * DC = Disk Child
872 * P = Path
873 *
874 * These #defines are named by concatenating these abbreviations,
875 * where first comes the arguments, and last comes the return value,
876 * of the macro.
877 */
Al Virof466c6f2012-03-17 01:16:43 -0400878
879#define USE_INODE_GENERATION_COUNTER
880
881#define REISERFS_PREALLOCATE
882#define DISPLACE_NEW_PACKING_LOCALITIES
883#define PREALLOCATION_SIZE 9
884
885/* n must be power of 2 */
886#define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
887
Jeff Mahoney098297b2014-04-23 10:00:36 -0400888/*
889 * to be ok for alpha and others we have to align structures to 8 byte
890 * boundary.
891 * FIXME: do not change 4 by anything else: there is code which relies on that
892 */
Al Virof466c6f2012-03-17 01:16:43 -0400893#define ROUND_UP(x) _ROUND_UP(x,8LL)
894
Jeff Mahoney098297b2014-04-23 10:00:36 -0400895/*
896 * debug levels. Right now, CONFIG_REISERFS_CHECK means print all debug
897 * messages.
898 */
Al Virof466c6f2012-03-17 01:16:43 -0400899#define REISERFS_DEBUG_CODE 5 /* extra messages to help find/debug errors */
900
901void __reiserfs_warning(struct super_block *s, const char *id,
902 const char *func, const char *fmt, ...);
903#define reiserfs_warning(s, id, fmt, args...) \
904 __reiserfs_warning(s, id, __func__, fmt, ##args)
905/* assertions handling */
906
Jeff Mahoney098297b2014-04-23 10:00:36 -0400907/* always check a condition and panic if it's false. */
Al Virof466c6f2012-03-17 01:16:43 -0400908#define __RASSERT(cond, scond, format, args...) \
909do { \
910 if (!(cond)) \
911 reiserfs_panic(NULL, "assertion failure", "(" #cond ") at " \
912 __FILE__ ":%i:%s: " format "\n", \
Al Virof466c6f2012-03-17 01:16:43 -0400913 __LINE__, __func__ , ##args); \
914} while (0)
915
916#define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args)
917
918#if defined( CONFIG_REISERFS_CHECK )
919#define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args)
920#else
921#define RFALSE( cond, format, args... ) do {;} while( 0 )
922#endif
923
924#define CONSTF __attribute_const__
925/*
926 * Disk Data Structures
927 */
928
Jeff Mahoney098297b2014-04-23 10:00:36 -0400929/***************************************************************************
930 * SUPER BLOCK *
931 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -0400932
933/*
Jeff Mahoney098297b2014-04-23 10:00:36 -0400934 * Structure of super block on disk, a version of which in RAM is often
935 * accessed as REISERFS_SB(s)->s_rs. The version in RAM is part of a larger
936 * structure containing fields never written to disk.
Al Virof466c6f2012-03-17 01:16:43 -0400937 */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400938#define UNSET_HASH 0 /* Detect hash on disk */
Al Virof466c6f2012-03-17 01:16:43 -0400939#define TEA_HASH 1
940#define YURA_HASH 2
941#define R5_HASH 3
942#define DEFAULT_HASH R5_HASH
943
944struct journal_params {
Jeff Mahoney098297b2014-04-23 10:00:36 -0400945 /* where does journal start from on its * device */
946 __le32 jp_journal_1st_block;
947
948 /* journal device st_rdev */
949 __le32 jp_journal_dev;
950
951 /* size of the journal */
952 __le32 jp_journal_size;
953
954 /* max number of blocks in a transaction. */
955 __le32 jp_journal_trans_max;
956
957 /*
958 * random value made on fs creation
959 * (this was sb_journal_block_count)
960 */
961 __le32 jp_journal_magic;
962
963 /* max number of blocks to batch into a trans */
964 __le32 jp_journal_max_batch;
965
966 /* in seconds, how old can an async commit be */
967 __le32 jp_journal_max_commit_age;
968
969 /* in seconds, how old can a transaction be */
970 __le32 jp_journal_max_trans_age;
Al Virof466c6f2012-03-17 01:16:43 -0400971};
972
973/* this is the super from 3.5.X, where X >= 10 */
974struct reiserfs_super_block_v1 {
975 __le32 s_block_count; /* blocks count */
976 __le32 s_free_blocks; /* free blocks count */
977 __le32 s_root_block; /* root block number */
978 struct journal_params s_journal;
979 __le16 s_blocksize; /* block size */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400980
981 /* max size of object id array, see get_objectid() commentary */
982 __le16 s_oid_maxsize;
Al Virof466c6f2012-03-17 01:16:43 -0400983 __le16 s_oid_cursize; /* current size of object id array */
Jeff Mahoney098297b2014-04-23 10:00:36 -0400984
985 /* this is set to 1 when filesystem was umounted, to 2 - when not */
986 __le16 s_umount_state;
987
988 /*
989 * reiserfs magic string indicates that file system is reiserfs:
990 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs"
991 */
992 char s_magic[10];
993
994 /*
995 * it is set to used by fsck to mark which
996 * phase of rebuilding is done
997 */
998 __le16 s_fs_state;
999 /*
1000 * indicate, what hash function is being use
1001 * to sort names in a directory
1002 */
1003 __le32 s_hash_function_code;
Al Virof466c6f2012-03-17 01:16:43 -04001004 __le16 s_tree_height; /* height of disk tree */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001005
1006 /*
1007 * amount of bitmap blocks needed to address
1008 * each block of file system
1009 */
1010 __le16 s_bmap_nr;
1011
1012 /*
1013 * this field is only reliable on filesystem with non-standard journal
1014 */
1015 __le16 s_version;
1016
1017 /*
1018 * size in blocks of journal area on main device, we need to
1019 * keep after making fs with non-standard journal
1020 */
1021 __le16 s_reserved_for_journal;
Al Virof466c6f2012-03-17 01:16:43 -04001022} __attribute__ ((__packed__));
1023
1024#define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))
1025
1026/* this is the on disk super block */
1027struct reiserfs_super_block {
1028 struct reiserfs_super_block_v1 s_v1;
1029 __le32 s_inode_generation;
Jeff Mahoney098297b2014-04-23 10:00:36 -04001030
1031 /* Right now used only by inode-attributes, if enabled */
1032 __le32 s_flags;
1033
Al Virof466c6f2012-03-17 01:16:43 -04001034 unsigned char s_uuid[16]; /* filesystem unique identifier */
1035 unsigned char s_label[16]; /* filesystem volume label */
1036 __le16 s_mnt_count; /* Count of mounts since last fsck */
1037 __le16 s_max_mnt_count; /* Maximum mounts before check */
1038 __le32 s_lastcheck; /* Timestamp of last fsck */
1039 __le32 s_check_interval; /* Interval between checks */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001040
1041 /*
1042 * zero filled by mkreiserfs and reiserfs_convert_objectid_map_v1()
1043 * so any additions must be updated there as well. */
1044 char s_unused[76];
Al Virof466c6f2012-03-17 01:16:43 -04001045} __attribute__ ((__packed__));
1046
1047#define SB_SIZE (sizeof(struct reiserfs_super_block))
1048
1049#define REISERFS_VERSION_1 0
1050#define REISERFS_VERSION_2 2
1051
Jeff Mahoney098297b2014-04-23 10:00:36 -04001052/* on-disk super block fields converted to cpu form */
Al Virof466c6f2012-03-17 01:16:43 -04001053#define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)
1054#define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))
1055#define SB_BLOCKSIZE(s) \
1056 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))
1057#define SB_BLOCK_COUNT(s) \
1058 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))
1059#define SB_FREE_BLOCKS(s) \
1060 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))
1061#define SB_REISERFS_MAGIC(s) \
1062 (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)
1063#define SB_ROOT_BLOCK(s) \
1064 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))
1065#define SB_TREE_HEIGHT(s) \
1066 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))
1067#define SB_REISERFS_STATE(s) \
1068 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))
1069#define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))
1070#define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))
1071
1072#define PUT_SB_BLOCK_COUNT(s, val) \
1073 do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)
1074#define PUT_SB_FREE_BLOCKS(s, val) \
1075 do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)
1076#define PUT_SB_ROOT_BLOCK(s, val) \
1077 do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)
1078#define PUT_SB_TREE_HEIGHT(s, val) \
1079 do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)
1080#define PUT_SB_REISERFS_STATE(s, val) \
1081 do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)
1082#define PUT_SB_VERSION(s, val) \
1083 do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)
1084#define PUT_SB_BMAP_NR(s, val) \
1085 do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)
1086
1087#define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)
1088#define SB_ONDISK_JOURNAL_SIZE(s) \
1089 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))
1090#define SB_ONDISK_JOURNAL_1st_BLOCK(s) \
1091 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))
1092#define SB_ONDISK_JOURNAL_DEVICE(s) \
1093 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
1094#define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
1095 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
1096
1097#define is_block_in_log_or_reserved_area(s, block) \
1098 block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
1099 && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) + \
1100 ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \
1101 SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))
1102
1103int is_reiserfs_3_5(struct reiserfs_super_block *rs);
1104int is_reiserfs_3_6(struct reiserfs_super_block *rs);
1105int is_reiserfs_jr(struct reiserfs_super_block *rs);
1106
Jeff Mahoney098297b2014-04-23 10:00:36 -04001107/*
1108 * ReiserFS leaves the first 64k unused, so that partition labels have
1109 * enough space. If someone wants to write a fancy bootloader that
1110 * needs more than 64k, let us know, and this will be increased in size.
1111 * This number must be larger than than the largest block size on any
1112 * platform, or code will break. -Hans
1113 */
Al Virof466c6f2012-03-17 01:16:43 -04001114#define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
1115#define REISERFS_FIRST_BLOCK unused_define
1116#define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES
1117
1118/* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
1119#define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
1120
1121/* reiserfs internal error code (used by search_by_key and fix_nodes)) */
1122#define CARRY_ON 0
1123#define REPEAT_SEARCH -1
1124#define IO_ERROR -2
1125#define NO_DISK_SPACE -3
1126#define NO_BALANCING_NEEDED (-4)
1127#define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
1128#define QUOTA_EXCEEDED -6
1129
1130typedef __u32 b_blocknr_t;
1131typedef __le32 unp_t;
1132
1133struct unfm_nodeinfo {
1134 unp_t unfm_nodenum;
1135 unsigned short unfm_freespace;
1136};
1137
Jeff Mahoney098297b2014-04-23 10:00:36 -04001138/* there are two formats of keys: 3.5 and 3.6 */
Al Virof466c6f2012-03-17 01:16:43 -04001139#define KEY_FORMAT_3_5 0
1140#define KEY_FORMAT_3_6 1
1141
1142/* there are two stat datas */
1143#define STAT_DATA_V1 0
1144#define STAT_DATA_V2 1
1145
1146static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)
1147{
1148 return container_of(inode, struct reiserfs_inode_info, vfs_inode);
1149}
1150
1151static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)
1152{
1153 return sb->s_fs_info;
1154}
1155
Jeff Mahoney098297b2014-04-23 10:00:36 -04001156/*
1157 * Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u16
1158 * which overflows on large file systems.
1159 */
Al Virof466c6f2012-03-17 01:16:43 -04001160static inline __u32 reiserfs_bmap_count(struct super_block *sb)
1161{
1162 return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1;
1163}
1164
1165static inline int bmap_would_wrap(unsigned bmap_nr)
1166{
1167 return bmap_nr > ((1LL << 16) - 1);
1168}
1169
Jeff Mahoney098297b2014-04-23 10:00:36 -04001170/*
1171 * this says about version of key of all items (but stat data) the
1172 * object consists of
1173 */
Al Virof466c6f2012-03-17 01:16:43 -04001174#define get_inode_item_key_version( inode ) \
1175 ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
1176
1177#define set_inode_item_key_version( inode, version ) \
1178 ({ if((version)==KEY_FORMAT_3_6) \
1179 REISERFS_I(inode)->i_flags |= i_item_key_version_mask; \
1180 else \
1181 REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })
1182
1183#define get_inode_sd_version(inode) \
1184 ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
1185
1186#define set_inode_sd_version(inode, version) \
1187 ({ if((version)==STAT_DATA_V2) \
1188 REISERFS_I(inode)->i_flags |= i_stat_data_version_mask; \
1189 else \
1190 REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })
1191
Jeff Mahoney098297b2014-04-23 10:00:36 -04001192/*
1193 * This is an aggressive tail suppression policy, I am hoping it
1194 * improves our benchmarks. The principle behind it is that percentage
1195 * space saving is what matters, not absolute space saving. This is
1196 * non-intuitive, but it helps to understand it if you consider that the
1197 * cost to access 4 blocks is not much more than the cost to access 1
1198 * block, if you have to do a seek and rotate. A tail risks a
1199 * non-linear disk access that is significant as a percentage of total
1200 * time cost for a 4 block file and saves an amount of space that is
1201 * less significant as a percentage of space, or so goes the hypothesis.
1202 * -Hans
1203 */
Al Virof466c6f2012-03-17 01:16:43 -04001204#define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
1205(\
1206 (!(n_tail_size)) || \
1207 (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
1208 ( (n_file_size) >= (n_block_size) * 4 ) || \
1209 ( ( (n_file_size) >= (n_block_size) * 3 ) && \
1210 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
1211 ( ( (n_file_size) >= (n_block_size) * 2 ) && \
1212 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
1213 ( ( (n_file_size) >= (n_block_size) ) && \
1214 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
1215)
1216
Jeff Mahoney098297b2014-04-23 10:00:36 -04001217/*
1218 * Another strategy for tails, this one means only create a tail if all the
1219 * file would fit into one DIRECT item.
1220 * Primary intention for this one is to increase performance by decreasing
1221 * seeking.
Al Virof466c6f2012-03-17 01:16:43 -04001222*/
1223#define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
1224(\
1225 (!(n_tail_size)) || \
1226 (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
1227)
1228
1229/*
1230 * values for s_umount_state field
1231 */
1232#define REISERFS_VALID_FS 1
1233#define REISERFS_ERROR_FS 2
1234
Jeff Mahoney098297b2014-04-23 10:00:36 -04001235/*
1236 * there are 5 item types currently
1237 */
Al Virof466c6f2012-03-17 01:16:43 -04001238#define TYPE_STAT_DATA 0
1239#define TYPE_INDIRECT 1
1240#define TYPE_DIRECT 2
1241#define TYPE_DIRENTRY 3
1242#define TYPE_MAXTYPE 3
Jeff Mahoney098297b2014-04-23 10:00:36 -04001243#define TYPE_ANY 15 /* FIXME: comment is required */
Al Virof466c6f2012-03-17 01:16:43 -04001244
Jeff Mahoney098297b2014-04-23 10:00:36 -04001245/***************************************************************************
1246 * KEY & ITEM HEAD *
1247 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -04001248
Jeff Mahoney098297b2014-04-23 10:00:36 -04001249/* * directories use this key as well as old files */
Al Virof466c6f2012-03-17 01:16:43 -04001250struct offset_v1 {
1251 __le32 k_offset;
1252 __le32 k_uniqueness;
1253} __attribute__ ((__packed__));
1254
1255struct offset_v2 {
1256 __le64 v;
1257} __attribute__ ((__packed__));
1258
1259static inline __u16 offset_v2_k_type(const struct offset_v2 *v2)
1260{
1261 __u8 type = le64_to_cpu(v2->v) >> 60;
1262 return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY;
1263}
1264
1265static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type)
1266{
1267 v2->v =
1268 (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60);
1269}
1270
1271static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2)
1272{
1273 return le64_to_cpu(v2->v) & (~0ULL >> 4);
1274}
1275
1276static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset)
1277{
1278 offset &= (~0ULL >> 4);
1279 v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset);
1280}
1281
Jeff Mahoney098297b2014-04-23 10:00:36 -04001282/*
1283 * Key of an item determines its location in the S+tree, and
1284 * is composed of 4 components
1285 */
Al Virof466c6f2012-03-17 01:16:43 -04001286struct reiserfs_key {
Jeff Mahoney098297b2014-04-23 10:00:36 -04001287 /* packing locality: by default parent directory object id */
1288 __le32 k_dir_id;
1289
Al Virof466c6f2012-03-17 01:16:43 -04001290 __le32 k_objectid; /* object identifier */
1291 union {
1292 struct offset_v1 k_offset_v1;
1293 struct offset_v2 k_offset_v2;
1294 } __attribute__ ((__packed__)) u;
1295} __attribute__ ((__packed__));
1296
1297struct in_core_key {
Jeff Mahoney098297b2014-04-23 10:00:36 -04001298 /* packing locality: by default parent directory object id */
1299 __u32 k_dir_id;
Al Virof466c6f2012-03-17 01:16:43 -04001300 __u32 k_objectid; /* object identifier */
1301 __u64 k_offset;
1302 __u8 k_type;
1303};
1304
1305struct cpu_key {
1306 struct in_core_key on_disk_key;
1307 int version;
Jeff Mahoney098297b2014-04-23 10:00:36 -04001308 /* 3 in all cases but direct2indirect and indirect2direct conversion */
1309 int key_length;
Al Virof466c6f2012-03-17 01:16:43 -04001310};
1311
Jeff Mahoney098297b2014-04-23 10:00:36 -04001312/*
1313 * Our function for comparing keys can compare keys of different
1314 * lengths. It takes as a parameter the length of the keys it is to
1315 * compare. These defines are used in determining what is to be passed
1316 * to it as that parameter.
1317 */
Al Virof466c6f2012-03-17 01:16:43 -04001318#define REISERFS_FULL_KEY_LEN 4
1319#define REISERFS_SHORT_KEY_LEN 2
1320
1321/* The result of the key compare */
1322#define FIRST_GREATER 1
1323#define SECOND_GREATER -1
1324#define KEYS_IDENTICAL 0
1325#define KEY_FOUND 1
1326#define KEY_NOT_FOUND 0
1327
1328#define KEY_SIZE (sizeof(struct reiserfs_key))
1329#define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
1330
1331/* return values for search_by_key and clones */
1332#define ITEM_FOUND 1
1333#define ITEM_NOT_FOUND 0
1334#define ENTRY_FOUND 1
1335#define ENTRY_NOT_FOUND 0
1336#define DIRECTORY_NOT_FOUND -1
1337#define REGULAR_FILE_FOUND -2
1338#define DIRECTORY_FOUND -3
1339#define BYTE_FOUND 1
1340#define BYTE_NOT_FOUND 0
1341#define FILE_NOT_FOUND -1
1342
1343#define POSITION_FOUND 1
1344#define POSITION_NOT_FOUND 0
1345
Jeff Mahoney098297b2014-04-23 10:00:36 -04001346/* return values for reiserfs_find_entry and search_by_entry_key */
Al Virof466c6f2012-03-17 01:16:43 -04001347#define NAME_FOUND 1
1348#define NAME_NOT_FOUND 0
1349#define GOTO_PREVIOUS_ITEM 2
1350#define NAME_FOUND_INVISIBLE 3
1351
Jeff Mahoney098297b2014-04-23 10:00:36 -04001352/*
1353 * Everything in the filesystem is stored as a set of items. The
1354 * item head contains the key of the item, its free space (for
1355 * indirect items) and specifies the location of the item itself
1356 * within the block.
1357 */
Al Virof466c6f2012-03-17 01:16:43 -04001358
1359struct item_head {
Jeff Mahoney098297b2014-04-23 10:00:36 -04001360 /*
1361 * Everything in the tree is found by searching for it based on
1362 * its key.
1363 */
Al Virof466c6f2012-03-17 01:16:43 -04001364 struct reiserfs_key ih_key;
1365 union {
Jeff Mahoney098297b2014-04-23 10:00:36 -04001366 /*
1367 * The free space in the last unformatted node of an
1368 * indirect item if this is an indirect item. This
1369 * equals 0xFFFF iff this is a direct item or stat data
1370 * item. Note that the key, not this field, is used to
1371 * determine the item type, and thus which field this
1372 * union contains.
1373 */
Al Virof466c6f2012-03-17 01:16:43 -04001374 __le16 ih_free_space_reserved;
Jeff Mahoney098297b2014-04-23 10:00:36 -04001375
1376 /*
1377 * Iff this is a directory item, this field equals the
1378 * number of directory entries in the directory item.
1379 */
Al Virof466c6f2012-03-17 01:16:43 -04001380 __le16 ih_entry_count;
1381 } __attribute__ ((__packed__)) u;
1382 __le16 ih_item_len; /* total size of the item body */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001383
1384 /* an offset to the item body within the block */
1385 __le16 ih_item_location;
1386
1387 /*
1388 * 0 for all old items, 2 for new ones. Highest bit is set by fsck
1389 * temporary, cleaned after all done
1390 */
1391 __le16 ih_version;
Al Virof466c6f2012-03-17 01:16:43 -04001392} __attribute__ ((__packed__));
1393/* size of item header */
1394#define IH_SIZE (sizeof(struct item_head))
1395
1396#define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved)
1397#define ih_version(ih) le16_to_cpu((ih)->ih_version)
1398#define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count)
1399#define ih_location(ih) le16_to_cpu((ih)->ih_item_location)
1400#define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len)
1401
1402#define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
1403#define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0)
1404#define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
1405#define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
1406#define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
1407
1408#define unreachable_item(ih) (ih_version(ih) & (1 << 15))
1409
1410#define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
1411#define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
1412
Jeff Mahoney098297b2014-04-23 10:00:36 -04001413/*
1414 * these operate on indirect items, where you've got an array of ints
1415 * at a possibly unaligned location. These are a noop on ia32
1416 *
1417 * p is the array of __u32, i is the index into the array, v is the value
1418 * to store there.
1419 */
Al Virof466c6f2012-03-17 01:16:43 -04001420#define get_block_num(p, i) get_unaligned_le32((p) + (i))
1421#define put_block_num(p, i, v) put_unaligned_le32((v), (p) + (i))
1422
Jeff Mahoney098297b2014-04-23 10:00:36 -04001423/* * in old version uniqueness field shows key type */
Al Virof466c6f2012-03-17 01:16:43 -04001424#define V1_SD_UNIQUENESS 0
1425#define V1_INDIRECT_UNIQUENESS 0xfffffffe
1426#define V1_DIRECT_UNIQUENESS 0xffffffff
1427#define V1_DIRENTRY_UNIQUENESS 500
Jeff Mahoney098297b2014-04-23 10:00:36 -04001428#define V1_ANY_UNIQUENESS 555 /* FIXME: comment is required */
Al Virof466c6f2012-03-17 01:16:43 -04001429
Jeff Mahoney098297b2014-04-23 10:00:36 -04001430/* here are conversion routines */
Al Virof466c6f2012-03-17 01:16:43 -04001431static inline int uniqueness2type(__u32 uniqueness) CONSTF;
1432static inline int uniqueness2type(__u32 uniqueness)
1433{
1434 switch ((int)uniqueness) {
1435 case V1_SD_UNIQUENESS:
1436 return TYPE_STAT_DATA;
1437 case V1_INDIRECT_UNIQUENESS:
1438 return TYPE_INDIRECT;
1439 case V1_DIRECT_UNIQUENESS:
1440 return TYPE_DIRECT;
1441 case V1_DIRENTRY_UNIQUENESS:
1442 return TYPE_DIRENTRY;
1443 case V1_ANY_UNIQUENESS:
1444 default:
1445 return TYPE_ANY;
1446 }
1447}
1448
1449static inline __u32 type2uniqueness(int type) CONSTF;
1450static inline __u32 type2uniqueness(int type)
1451{
1452 switch (type) {
1453 case TYPE_STAT_DATA:
1454 return V1_SD_UNIQUENESS;
1455 case TYPE_INDIRECT:
1456 return V1_INDIRECT_UNIQUENESS;
1457 case TYPE_DIRECT:
1458 return V1_DIRECT_UNIQUENESS;
1459 case TYPE_DIRENTRY:
1460 return V1_DIRENTRY_UNIQUENESS;
1461 case TYPE_ANY:
1462 default:
1463 return V1_ANY_UNIQUENESS;
1464 }
1465}
1466
Jeff Mahoney098297b2014-04-23 10:00:36 -04001467/*
1468 * key is pointer to on disk key which is stored in le, result is cpu,
1469 * there is no way to get version of object from key, so, provide
1470 * version to these defines
1471 */
Al Virof466c6f2012-03-17 01:16:43 -04001472static inline loff_t le_key_k_offset(int version,
1473 const struct reiserfs_key *key)
1474{
1475 return (version == KEY_FORMAT_3_5) ?
1476 le32_to_cpu(key->u.k_offset_v1.k_offset) :
1477 offset_v2_k_offset(&(key->u.k_offset_v2));
1478}
1479
1480static inline loff_t le_ih_k_offset(const struct item_head *ih)
1481{
1482 return le_key_k_offset(ih_version(ih), &(ih->ih_key));
1483}
1484
1485static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key)
1486{
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04001487 if (version == KEY_FORMAT_3_5) {
1488 loff_t val = le32_to_cpu(key->u.k_offset_v1.k_uniqueness);
1489 return uniqueness2type(val);
1490 } else
1491 return offset_v2_k_type(&(key->u.k_offset_v2));
Al Virof466c6f2012-03-17 01:16:43 -04001492}
1493
1494static inline loff_t le_ih_k_type(const struct item_head *ih)
1495{
1496 return le_key_k_type(ih_version(ih), &(ih->ih_key));
1497}
1498
1499static inline void set_le_key_k_offset(int version, struct reiserfs_key *key,
1500 loff_t offset)
1501{
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04001502 if (version == KEY_FORMAT_3_5)
1503 key->u.k_offset_v1.k_offset = cpu_to_le32(offset);
1504 else
1505 set_offset_v2_k_offset(&key->u.k_offset_v2, offset);
1506}
1507
1508static inline void add_le_key_k_offset(int version, struct reiserfs_key *key,
1509 loff_t offset)
1510{
1511 set_le_key_k_offset(version, key,
1512 le_key_k_offset(version, key) + offset);
1513}
1514
1515static inline void add_le_ih_k_offset(struct item_head *ih, loff_t offset)
1516{
1517 add_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
Al Virof466c6f2012-03-17 01:16:43 -04001518}
1519
1520static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset)
1521{
1522 set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
1523}
1524
1525static inline void set_le_key_k_type(int version, struct reiserfs_key *key,
1526 int type)
1527{
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04001528 if (version == KEY_FORMAT_3_5) {
1529 type = type2uniqueness(type);
1530 key->u.k_offset_v1.k_uniqueness = cpu_to_le32(type);
1531 } else
1532 set_offset_v2_k_type(&key->u.k_offset_v2, type);
Al Virof466c6f2012-03-17 01:16:43 -04001533}
1534
1535static inline void set_le_ih_k_type(struct item_head *ih, int type)
1536{
1537 set_le_key_k_type(ih_version(ih), &(ih->ih_key), type);
1538}
1539
1540static inline int is_direntry_le_key(int version, struct reiserfs_key *key)
1541{
1542 return le_key_k_type(version, key) == TYPE_DIRENTRY;
1543}
1544
1545static inline int is_direct_le_key(int version, struct reiserfs_key *key)
1546{
1547 return le_key_k_type(version, key) == TYPE_DIRECT;
1548}
1549
1550static inline int is_indirect_le_key(int version, struct reiserfs_key *key)
1551{
1552 return le_key_k_type(version, key) == TYPE_INDIRECT;
1553}
1554
1555static inline int is_statdata_le_key(int version, struct reiserfs_key *key)
1556{
1557 return le_key_k_type(version, key) == TYPE_STAT_DATA;
1558}
1559
Jeff Mahoney098297b2014-04-23 10:00:36 -04001560/* item header has version. */
Al Virof466c6f2012-03-17 01:16:43 -04001561static inline int is_direntry_le_ih(struct item_head *ih)
1562{
1563 return is_direntry_le_key(ih_version(ih), &ih->ih_key);
1564}
1565
1566static inline int is_direct_le_ih(struct item_head *ih)
1567{
1568 return is_direct_le_key(ih_version(ih), &ih->ih_key);
1569}
1570
1571static inline int is_indirect_le_ih(struct item_head *ih)
1572{
1573 return is_indirect_le_key(ih_version(ih), &ih->ih_key);
1574}
1575
1576static inline int is_statdata_le_ih(struct item_head *ih)
1577{
1578 return is_statdata_le_key(ih_version(ih), &ih->ih_key);
1579}
1580
Jeff Mahoney098297b2014-04-23 10:00:36 -04001581/* key is pointer to cpu key, result is cpu */
Al Virof466c6f2012-03-17 01:16:43 -04001582static inline loff_t cpu_key_k_offset(const struct cpu_key *key)
1583{
1584 return key->on_disk_key.k_offset;
1585}
1586
1587static inline loff_t cpu_key_k_type(const struct cpu_key *key)
1588{
1589 return key->on_disk_key.k_type;
1590}
1591
1592static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset)
1593{
1594 key->on_disk_key.k_offset = offset;
1595}
1596
1597static inline void set_cpu_key_k_type(struct cpu_key *key, int type)
1598{
1599 key->on_disk_key.k_type = type;
1600}
1601
1602static inline void cpu_key_k_offset_dec(struct cpu_key *key)
1603{
1604 key->on_disk_key.k_offset--;
1605}
1606
1607#define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
1608#define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
1609#define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
1610#define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
1611
1612/* are these used ? */
1613#define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
1614#define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
1615#define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
1616#define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
1617
1618#define I_K_KEY_IN_ITEM(ih, key, n_blocksize) \
1619 (!COMP_SHORT_KEYS(ih, key) && \
1620 I_OFF_BYTE_IN_ITEM(ih, k_offset(key), n_blocksize))
1621
1622/* maximal length of item */
1623#define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
1624#define MIN_ITEM_LEN 1
1625
1626/* object identifier for root dir */
1627#define REISERFS_ROOT_OBJECTID 2
1628#define REISERFS_ROOT_PARENT_OBJECTID 1
1629
1630extern struct reiserfs_key root_key;
1631
Jeff Mahoney098297b2014-04-23 10:00:36 -04001632/*
Al Virof466c6f2012-03-17 01:16:43 -04001633 * Picture represents a leaf of the S+tree
1634 * ______________________________________________________
1635 * | | Array of | | |
1636 * |Block | Object-Item | F r e e | Objects- |
1637 * | head | Headers | S p a c e | Items |
1638 * |______|_______________|___________________|___________|
1639 */
1640
Jeff Mahoney098297b2014-04-23 10:00:36 -04001641/*
1642 * Header of a disk block. More precisely, header of a formatted leaf
1643 * or internal node, and not the header of an unformatted node.
1644 */
Al Virof466c6f2012-03-17 01:16:43 -04001645struct block_head {
1646 __le16 blk_level; /* Level of a block in the tree. */
1647 __le16 blk_nr_item; /* Number of keys/items in a block. */
1648 __le16 blk_free_space; /* Block free space in bytes. */
1649 __le16 blk_reserved;
1650 /* dump this in v4/planA */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001651
1652 /* kept only for compatibility */
1653 struct reiserfs_key blk_right_delim_key;
Al Virof466c6f2012-03-17 01:16:43 -04001654};
1655
1656#define BLKH_SIZE (sizeof(struct block_head))
1657#define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level))
1658#define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item))
1659#define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space))
1660#define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved))
1661#define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val))
1662#define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val))
1663#define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
1664#define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
1665#define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key)
1666#define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val)
1667
Jeff Mahoney098297b2014-04-23 10:00:36 -04001668/* values for blk_level field of the struct block_head */
Al Virof466c6f2012-03-17 01:16:43 -04001669
Jeff Mahoney098297b2014-04-23 10:00:36 -04001670/*
1671 * When node gets removed from the tree its blk_level is set to FREE_LEVEL.
1672 * It is then used to see whether the node is still in the tree
1673 */
1674#define FREE_LEVEL 0
Al Virof466c6f2012-03-17 01:16:43 -04001675
1676#define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level. */
1677
Jeff Mahoney098297b2014-04-23 10:00:36 -04001678/*
1679 * Given the buffer head of a formatted node, resolve to the
1680 * block head of that node.
1681 */
Al Virof466c6f2012-03-17 01:16:43 -04001682#define B_BLK_HEAD(bh) ((struct block_head *)((bh)->b_data))
1683/* Number of items that are in buffer. */
1684#define B_NR_ITEMS(bh) (blkh_nr_item(B_BLK_HEAD(bh)))
1685#define B_LEVEL(bh) (blkh_level(B_BLK_HEAD(bh)))
1686#define B_FREE_SPACE(bh) (blkh_free_space(B_BLK_HEAD(bh)))
1687
1688#define PUT_B_NR_ITEMS(bh, val) do { set_blkh_nr_item(B_BLK_HEAD(bh), val); } while (0)
1689#define PUT_B_LEVEL(bh, val) do { set_blkh_level(B_BLK_HEAD(bh), val); } while (0)
1690#define PUT_B_FREE_SPACE(bh, val) do { set_blkh_free_space(B_BLK_HEAD(bh), val); } while (0)
1691
1692/* Get right delimiting key. -- little endian */
1693#define B_PRIGHT_DELIM_KEY(bh) (&(blk_right_delim_key(B_BLK_HEAD(bh))))
1694
1695/* Does the buffer contain a disk leaf. */
1696#define B_IS_ITEMS_LEVEL(bh) (B_LEVEL(bh) == DISK_LEAF_NODE_LEVEL)
1697
1698/* Does the buffer contain a disk internal node */
1699#define B_IS_KEYS_LEVEL(bh) (B_LEVEL(bh) > DISK_LEAF_NODE_LEVEL \
1700 && B_LEVEL(bh) <= MAX_HEIGHT)
1701
Jeff Mahoney098297b2014-04-23 10:00:36 -04001702/***************************************************************************
1703 * STAT DATA *
1704 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -04001705
Jeff Mahoney098297b2014-04-23 10:00:36 -04001706/*
1707 * old stat data is 32 bytes long. We are going to distinguish new one by
1708 * different size
1709*/
Al Virof466c6f2012-03-17 01:16:43 -04001710struct stat_data_v1 {
1711 __le16 sd_mode; /* file type, permissions */
1712 __le16 sd_nlink; /* number of hard links */
1713 __le16 sd_uid; /* owner */
1714 __le16 sd_gid; /* group */
1715 __le32 sd_size; /* file size */
1716 __le32 sd_atime; /* time of last access */
1717 __le32 sd_mtime; /* time file was last modified */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001718
1719 /*
1720 * time inode (stat data) was last changed
1721 * (except changes to sd_atime and sd_mtime)
1722 */
1723 __le32 sd_ctime;
Al Virof466c6f2012-03-17 01:16:43 -04001724 union {
1725 __le32 sd_rdev;
1726 __le32 sd_blocks; /* number of blocks file uses */
1727 } __attribute__ ((__packed__)) u;
Jeff Mahoney098297b2014-04-23 10:00:36 -04001728
1729 /*
1730 * first byte of file which is stored in a direct item: except that if
1731 * it equals 1 it is a symlink and if it equals ~(__u32)0 there is no
1732 * direct item. The existence of this field really grates on me.
1733 * Let's replace it with a macro based on sd_size and our tail
1734 * suppression policy. Someday. -Hans
1735 */
1736 __le32 sd_first_direct_byte;
Al Virof466c6f2012-03-17 01:16:43 -04001737} __attribute__ ((__packed__));
1738
1739#define SD_V1_SIZE (sizeof(struct stat_data_v1))
1740#define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5)
1741#define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
1742#define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
1743#define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink))
1744#define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v))
1745#define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid))
1746#define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v))
1747#define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid))
1748#define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v))
1749#define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size))
1750#define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v))
1751#define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
1752#define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
1753#define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
1754#define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
1755#define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
1756#define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
1757#define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
1758#define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
1759#define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks))
1760#define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
1761#define sd_v1_first_direct_byte(sdp) \
1762 (le32_to_cpu((sdp)->sd_first_direct_byte))
1763#define set_sd_v1_first_direct_byte(sdp,v) \
1764 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
1765
1766/* inode flags stored in sd_attrs (nee sd_reserved) */
1767
Jeff Mahoney098297b2014-04-23 10:00:36 -04001768/*
1769 * we want common flags to have the same values as in ext2,
1770 * so chattr(1) will work without problems
1771 */
Al Virof466c6f2012-03-17 01:16:43 -04001772#define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL
1773#define REISERFS_APPEND_FL FS_APPEND_FL
1774#define REISERFS_SYNC_FL FS_SYNC_FL
1775#define REISERFS_NOATIME_FL FS_NOATIME_FL
1776#define REISERFS_NODUMP_FL FS_NODUMP_FL
1777#define REISERFS_SECRM_FL FS_SECRM_FL
1778#define REISERFS_UNRM_FL FS_UNRM_FL
1779#define REISERFS_COMPR_FL FS_COMPR_FL
1780#define REISERFS_NOTAIL_FL FS_NOTAIL_FL
1781
1782/* persistent flags that file inherits from the parent directory */
1783#define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \
1784 REISERFS_SYNC_FL | \
1785 REISERFS_NOATIME_FL | \
1786 REISERFS_NODUMP_FL | \
1787 REISERFS_SECRM_FL | \
1788 REISERFS_COMPR_FL | \
1789 REISERFS_NOTAIL_FL )
1790
Jeff Mahoney098297b2014-04-23 10:00:36 -04001791/*
1792 * Stat Data on disk (reiserfs version of UFS disk inode minus the
1793 * address blocks)
1794 */
Al Virof466c6f2012-03-17 01:16:43 -04001795struct stat_data {
1796 __le16 sd_mode; /* file type, permissions */
1797 __le16 sd_attrs; /* persistent inode flags */
1798 __le32 sd_nlink; /* number of hard links */
1799 __le64 sd_size; /* file size */
1800 __le32 sd_uid; /* owner */
1801 __le32 sd_gid; /* group */
1802 __le32 sd_atime; /* time of last access */
1803 __le32 sd_mtime; /* time file was last modified */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001804
1805 /*
1806 * time inode (stat data) was last changed
1807 * (except changes to sd_atime and sd_mtime)
1808 */
1809 __le32 sd_ctime;
Al Virof466c6f2012-03-17 01:16:43 -04001810 __le32 sd_blocks;
1811 union {
1812 __le32 sd_rdev;
1813 __le32 sd_generation;
Al Virof466c6f2012-03-17 01:16:43 -04001814 } __attribute__ ((__packed__)) u;
1815} __attribute__ ((__packed__));
Jeff Mahoney098297b2014-04-23 10:00:36 -04001816
1817/* this is 44 bytes long */
Al Virof466c6f2012-03-17 01:16:43 -04001818#define SD_SIZE (sizeof(struct stat_data))
1819#define SD_V2_SIZE SD_SIZE
1820#define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6)
1821#define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
1822#define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
1823/* sd_reserved */
1824/* set_sd_reserved */
1825#define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink))
1826#define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v))
1827#define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size))
1828#define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v))
1829#define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid))
1830#define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v))
1831#define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid))
1832#define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v))
1833#define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
1834#define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
1835#define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
1836#define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
1837#define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
1838#define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
1839#define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks))
1840#define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
1841#define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
1842#define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
1843#define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation))
1844#define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
1845#define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs))
1846#define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v))
1847
Jeff Mahoney098297b2014-04-23 10:00:36 -04001848/***************************************************************************
1849 * DIRECTORY STRUCTURE *
1850 ***************************************************************************/
1851/*
1852 * Picture represents the structure of directory items
1853 * ________________________________________________
1854 * | Array of | | | | | |
1855 * | directory |N-1| N-2 | .... | 1st |0th|
1856 * | entry headers | | | | | |
1857 * |_______________|___|_____|________|_______|___|
1858 * <---- directory entries ------>
1859 *
1860 * First directory item has k_offset component 1. We store "." and ".."
1861 * in one item, always, we never split "." and ".." into differing
1862 * items. This makes, among other things, the code for removing
1863 * directories simpler.
1864 */
Al Virof466c6f2012-03-17 01:16:43 -04001865#define SD_OFFSET 0
1866#define SD_UNIQUENESS 0
1867#define DOT_OFFSET 1
1868#define DOT_DOT_OFFSET 2
1869#define DIRENTRY_UNIQUENESS 500
1870
Al Virof466c6f2012-03-17 01:16:43 -04001871#define FIRST_ITEM_OFFSET 1
1872
1873/*
Jeff Mahoney098297b2014-04-23 10:00:36 -04001874 * Q: How to get key of object pointed to by entry from entry?
1875 *
1876 * A: Each directory entry has its header. This header has deh_dir_id
1877 * and deh_objectid fields, those are key of object, entry points to
1878 */
Al Virof466c6f2012-03-17 01:16:43 -04001879
Jeff Mahoney098297b2014-04-23 10:00:36 -04001880/*
1881 * NOT IMPLEMENTED:
1882 * Directory will someday contain stat data of object
1883 */
Al Virof466c6f2012-03-17 01:16:43 -04001884
1885struct reiserfs_de_head {
1886 __le32 deh_offset; /* third component of the directory entry key */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001887
1888 /*
1889 * objectid of the parent directory of the object, that is referenced
1890 * by directory entry
1891 */
1892 __le32 deh_dir_id;
1893
1894 /* objectid of the object, that is referenced by directory entry */
1895 __le32 deh_objectid;
Al Virof466c6f2012-03-17 01:16:43 -04001896 __le16 deh_location; /* offset of name in the whole item */
Jeff Mahoney098297b2014-04-23 10:00:36 -04001897
1898 /*
1899 * whether 1) entry contains stat data (for future), and
1900 * 2) whether entry is hidden (unlinked)
1901 */
1902 __le16 deh_state;
Al Virof466c6f2012-03-17 01:16:43 -04001903} __attribute__ ((__packed__));
1904#define DEH_SIZE sizeof(struct reiserfs_de_head)
1905#define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset))
1906#define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id))
1907#define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid))
1908#define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location))
1909#define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state))
1910
1911#define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v)))
1912#define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v)))
1913#define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
1914#define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
1915#define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v)))
1916
1917/* empty directory contains two entries "." and ".." and their headers */
1918#define EMPTY_DIR_SIZE \
1919(DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
1920
1921/* old format directories have this size when empty */
1922#define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
1923
1924#define DEH_Statdata 0 /* not used now */
1925#define DEH_Visible 2
1926
1927/* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
1928#if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
1929# define ADDR_UNALIGNED_BITS (3)
1930#endif
1931
Jeff Mahoney098297b2014-04-23 10:00:36 -04001932/*
1933 * These are only used to manipulate deh_state.
Al Virof466c6f2012-03-17 01:16:43 -04001934 * Because of this, we'll use the ext2_ bit routines,
Jeff Mahoney098297b2014-04-23 10:00:36 -04001935 * since they are little endian
1936 */
Al Virof466c6f2012-03-17 01:16:43 -04001937#ifdef ADDR_UNALIGNED_BITS
1938
1939# define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
1940# define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
1941
1942# define set_bit_unaligned(nr, addr) \
1943 __test_and_set_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1944# define clear_bit_unaligned(nr, addr) \
1945 __test_and_clear_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1946# define test_bit_unaligned(nr, addr) \
1947 test_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1948
1949#else
1950
1951# define set_bit_unaligned(nr, addr) __test_and_set_bit_le(nr, addr)
1952# define clear_bit_unaligned(nr, addr) __test_and_clear_bit_le(nr, addr)
1953# define test_bit_unaligned(nr, addr) test_bit_le(nr, addr)
1954
1955#endif
1956
1957#define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1958#define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1959#define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1960#define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1961
1962#define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1963#define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1964#define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1965
1966extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid,
1967 __le32 par_dirid, __le32 par_objid);
1968extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid,
1969 __le32 par_dirid, __le32 par_objid);
1970
Jeff Mahoney098297b2014-04-23 10:00:36 -04001971/* two entries per block (at least) */
Al Virof466c6f2012-03-17 01:16:43 -04001972#define REISERFS_MAX_NAME(block_size) 255
1973
Jeff Mahoney098297b2014-04-23 10:00:36 -04001974/*
1975 * this structure is used for operations on directory entries. It is
1976 * not a disk structure.
1977 *
1978 * When reiserfs_find_entry or search_by_entry_key find directory
1979 * entry, they return filled reiserfs_dir_entry structure
1980 */
Al Virof466c6f2012-03-17 01:16:43 -04001981struct reiserfs_dir_entry {
1982 struct buffer_head *de_bh;
1983 int de_item_num;
1984 struct item_head *de_ih;
1985 int de_entry_num;
1986 struct reiserfs_de_head *de_deh;
1987 int de_entrylen;
1988 int de_namelen;
1989 char *de_name;
1990 unsigned long *de_gen_number_bit_string;
1991
1992 __u32 de_dir_id;
1993 __u32 de_objectid;
1994
1995 struct cpu_key de_entry_key;
1996};
1997
Jeff Mahoney098297b2014-04-23 10:00:36 -04001998/*
1999 * these defines are useful when a particular member of
2000 * a reiserfs_dir_entry is needed
2001 */
Al Virof466c6f2012-03-17 01:16:43 -04002002
2003/* pointer to file name, stored in entry */
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04002004#define B_I_DEH_ENTRY_FILE_NAME(bh, ih, deh) \
2005 (ih_item_body(bh, ih) + deh_location(deh))
Al Virof466c6f2012-03-17 01:16:43 -04002006
2007/* length of name */
2008#define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
2009(I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
2010
2011/* hash value occupies bits from 7 up to 30 */
2012#define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
2013/* generation number occupies 7 bits starting from 0 up to 6 */
2014#define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
2015#define MAX_GENERATION_NUMBER 127
2016
2017#define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
2018
2019/*
2020 * Picture represents an internal node of the reiserfs tree
2021 * ______________________________________________________
2022 * | | Array of | Array of | Free |
2023 * |block | keys | pointers | space |
2024 * | head | N | N+1 | |
2025 * |______|_______________|___________________|___________|
2026 */
2027
Jeff Mahoney098297b2014-04-23 10:00:36 -04002028/***************************************************************************
2029 * DISK CHILD *
2030 ***************************************************************************/
2031/*
2032 * Disk child pointer:
2033 * The pointer from an internal node of the tree to a node that is on disk.
2034 */
Al Virof466c6f2012-03-17 01:16:43 -04002035struct disk_child {
2036 __le32 dc_block_number; /* Disk child's block number. */
2037 __le16 dc_size; /* Disk child's used space. */
2038 __le16 dc_reserved;
2039};
2040
2041#define DC_SIZE (sizeof(struct disk_child))
2042#define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number))
2043#define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size))
2044#define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
2045#define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
2046
2047/* Get disk child by buffer header and position in the tree node. */
2048#define B_N_CHILD(bh, n_pos) ((struct disk_child *)\
2049((bh)->b_data + BLKH_SIZE + B_NR_ITEMS(bh) * KEY_SIZE + DC_SIZE * (n_pos)))
2050
2051/* Get disk child number by buffer header and position in the tree node. */
2052#define B_N_CHILD_NUM(bh, n_pos) (dc_block_number(B_N_CHILD(bh, n_pos)))
2053#define PUT_B_N_CHILD_NUM(bh, n_pos, val) \
2054 (put_dc_block_number(B_N_CHILD(bh, n_pos), val))
2055
2056 /* maximal value of field child_size in structure disk_child */
2057 /* child size is the combined size of all items and their headers */
2058#define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
2059
2060/* amount of used space in buffer (not including block head) */
2061#define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
2062
2063/* max and min number of keys in internal node */
2064#define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
2065#define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2)
2066
Jeff Mahoney098297b2014-04-23 10:00:36 -04002067/***************************************************************************
2068 * PATH STRUCTURES AND DEFINES *
2069 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -04002070
Jeff Mahoney098297b2014-04-23 10:00:36 -04002071/*
2072 * search_by_key fills up the path from the root to the leaf as it descends
2073 * the tree looking for the key. It uses reiserfs_bread to try to find
2074 * buffers in the cache given their block number. If it does not find
2075 * them in the cache it reads them from disk. For each node search_by_key
2076 * finds using reiserfs_bread it then uses bin_search to look through that
2077 * node. bin_search will find the position of the block_number of the next
2078 * node if it is looking through an internal node. If it is looking through
2079 * a leaf node bin_search will find the position of the item which has key
2080 * either equal to given key, or which is the maximal key less than the
2081 * given key.
2082 */
Al Virof466c6f2012-03-17 01:16:43 -04002083
2084struct path_element {
Jeff Mahoney098297b2014-04-23 10:00:36 -04002085 /* Pointer to the buffer at the path in the tree. */
2086 struct buffer_head *pe_buffer;
2087 /* Position in the tree node which is placed in the buffer above. */
2088 int pe_position;
Al Virof466c6f2012-03-17 01:16:43 -04002089};
2090
Jeff Mahoney098297b2014-04-23 10:00:36 -04002091/*
2092 * maximal height of a tree. don't change this without
2093 * changing JOURNAL_PER_BALANCE_CNT
2094 */
2095#define MAX_HEIGHT 5
Al Virof466c6f2012-03-17 01:16:43 -04002096
Jeff Mahoney098297b2014-04-23 10:00:36 -04002097/* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
2098#define EXTENDED_MAX_HEIGHT 7
Al Virof466c6f2012-03-17 01:16:43 -04002099
Jeff Mahoney098297b2014-04-23 10:00:36 -04002100/* Must be equal to at least 2. */
2101#define FIRST_PATH_ELEMENT_OFFSET 2
Al Virof466c6f2012-03-17 01:16:43 -04002102
Jeff Mahoney098297b2014-04-23 10:00:36 -04002103/* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
2104#define ILLEGAL_PATH_ELEMENT_OFFSET 1
2105
2106/* this MUST be MAX_HEIGHT + 1. See about FEB below */
2107#define MAX_FEB_SIZE 6
2108
2109/*
2110 * We need to keep track of who the ancestors of nodes are. When we
2111 * perform a search we record which nodes were visited while
2112 * descending the tree looking for the node we searched for. This list
2113 * of nodes is called the path. This information is used while
2114 * performing balancing. Note that this path information may become
2115 * invalid, and this means we must check it when using it to see if it
2116 * is still valid. You'll need to read search_by_key and the comments
2117 * in it, especially about decrement_counters_in_path(), to understand
2118 * this structure.
2119 *
2120 * Paths make the code so much harder to work with and debug.... An
2121 * enormous number of bugs are due to them, and trying to write or modify
2122 * code that uses them just makes my head hurt. They are based on an
2123 * excessive effort to avoid disturbing the precious VFS code.:-( The
2124 * gods only know how we are going to SMP the code that uses them.
2125 * znodes are the way!
2126 */
Al Virof466c6f2012-03-17 01:16:43 -04002127
2128#define PATH_READA 0x1 /* do read ahead */
2129#define PATH_READA_BACK 0x2 /* read backwards */
2130
2131struct treepath {
2132 int path_length; /* Length of the array above. */
2133 int reada;
Jeff Mahoney098297b2014-04-23 10:00:36 -04002134 /* Array of the path elements. */
2135 struct path_element path_elements[EXTENDED_MAX_HEIGHT];
Al Virof466c6f2012-03-17 01:16:43 -04002136 int pos_in_item;
2137};
2138
2139#define pos_in_item(path) ((path)->pos_in_item)
2140
2141#define INITIALIZE_PATH(var) \
2142struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
2143
2144/* Get path element by path and path position. */
2145#define PATH_OFFSET_PELEMENT(path, n_offset) ((path)->path_elements + (n_offset))
2146
2147/* Get buffer header at the path by path and path position. */
2148#define PATH_OFFSET_PBUFFER(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_buffer)
2149
2150/* Get position in the element at the path by path and path position. */
2151#define PATH_OFFSET_POSITION(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_position)
2152
2153#define PATH_PLAST_BUFFER(path) (PATH_OFFSET_PBUFFER((path), (path)->path_length))
Jeff Mahoney098297b2014-04-23 10:00:36 -04002154
2155/*
2156 * you know, to the person who didn't write this the macro name does not
2157 * at first suggest what it does. Maybe POSITION_FROM_PATH_END? Or
2158 * maybe we should just focus on dumping paths... -Hans
2159 */
Al Virof466c6f2012-03-17 01:16:43 -04002160#define PATH_LAST_POSITION(path) (PATH_OFFSET_POSITION((path), (path)->path_length))
2161
Jeff Mahoney098297b2014-04-23 10:00:36 -04002162/*
2163 * in do_balance leaf has h == 0 in contrast with path structure,
2164 * where root has level == 0. That is why we need these defines
2165 */
2166
2167/* tb->S[h] */
2168#define PATH_H_PBUFFER(path, h) \
2169 PATH_OFFSET_PBUFFER(path, path->path_length - (h))
2170
2171/* tb->F[h] or tb->S[0]->b_parent */
2172#define PATH_H_PPARENT(path, h) PATH_H_PBUFFER(path, (h) + 1)
2173
2174#define PATH_H_POSITION(path, h) \
2175 PATH_OFFSET_POSITION(path, path->path_length - (h))
2176
2177/* tb->S[h]->b_item_order */
2178#define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1)
Al Virof466c6f2012-03-17 01:16:43 -04002179
2180#define PATH_H_PATH_OFFSET(path, n_h) ((path)->path_length - (n_h))
2181
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04002182static inline void *reiserfs_node_data(const struct buffer_head *bh)
2183{
2184 return bh->b_data + sizeof(struct block_head);
2185}
2186
2187/* get key from internal node */
2188static inline struct reiserfs_key *internal_key(struct buffer_head *bh,
2189 int item_num)
2190{
2191 struct reiserfs_key *key = reiserfs_node_data(bh);
2192
2193 return &key[item_num];
2194}
2195
2196/* get the item header from leaf node */
2197static inline struct item_head *item_head(const struct buffer_head *bh,
2198 int item_num)
2199{
2200 struct item_head *ih = reiserfs_node_data(bh);
2201
2202 return &ih[item_num];
2203}
2204
2205/* get the key from leaf node */
2206static inline struct reiserfs_key *leaf_key(const struct buffer_head *bh,
2207 int item_num)
2208{
2209 return &item_head(bh, item_num)->ih_key;
2210}
2211
2212static inline void *ih_item_body(const struct buffer_head *bh,
2213 const struct item_head *ih)
2214{
2215 return bh->b_data + ih_location(ih);
2216}
2217
2218/* get item body from leaf node */
2219static inline void *item_body(const struct buffer_head *bh, int item_num)
2220{
2221 return ih_item_body(bh, item_head(bh, item_num));
2222}
2223
2224static inline struct item_head *tp_item_head(const struct treepath *path)
2225{
2226 return item_head(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path));
2227}
2228
2229static inline void *tp_item_body(const struct treepath *path)
2230{
2231 return item_body(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path));
2232}
2233
Al Virof466c6f2012-03-17 01:16:43 -04002234#define get_last_bh(path) PATH_PLAST_BUFFER(path)
Al Virof466c6f2012-03-17 01:16:43 -04002235#define get_item_pos(path) PATH_LAST_POSITION(path)
Al Virof466c6f2012-03-17 01:16:43 -04002236#define item_moved(ih,path) comp_items(ih, path)
2237#define path_changed(ih,path) comp_items (ih, path)
2238
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04002239/* array of the entry headers */
2240 /* get item body */
2241#define B_I_DEH(bh, ih) ((struct reiserfs_de_head *)(ih_item_body(bh, ih)))
2242
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04002243/*
Jeff Mahoney098297b2014-04-23 10:00:36 -04002244 * length of the directory entry in directory item. This define
2245 * calculates length of i-th directory entry using directory entry
2246 * locations from dir entry head. When it calculates length of 0-th
2247 * directory entry, it uses length of whole item in place of entry
2248 * location of the non-existent following entry in the calculation.
2249 * See picture above.
2250 */
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04002251static inline int entry_length(const struct buffer_head *bh,
2252 const struct item_head *ih, int pos_in_item)
2253{
2254 struct reiserfs_de_head *deh;
2255
2256 deh = B_I_DEH(bh, ih) + pos_in_item;
2257 if (pos_in_item)
2258 return deh_location(deh - 1) - deh_location(deh);
2259
2260 return ih_item_len(ih) - deh_location(deh);
2261}
2262
Jeff Mahoney098297b2014-04-23 10:00:36 -04002263/***************************************************************************
2264 * MISC *
2265 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -04002266
2267/* Size of pointer to the unformatted node. */
2268#define UNFM_P_SIZE (sizeof(unp_t))
2269#define UNFM_P_SHIFT 2
2270
Jeff Mahoney098297b2014-04-23 10:00:36 -04002271/* in in-core inode key is stored on le form */
Al Virof466c6f2012-03-17 01:16:43 -04002272#define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
2273
2274#define MAX_UL_INT 0xffffffff
2275#define MAX_INT 0x7ffffff
2276#define MAX_US_INT 0xffff
2277
2278// reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
Al Virof466c6f2012-03-17 01:16:43 -04002279static inline loff_t max_reiserfs_offset(struct inode *inode)
2280{
2281 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
2282 return (loff_t) U32_MAX;
2283
2284 return (loff_t) ((~(__u64) 0) >> 4);
2285}
2286
Al Virof466c6f2012-03-17 01:16:43 -04002287#define MAX_KEY_OBJECTID MAX_UL_INT
2288
2289#define MAX_B_NUM MAX_UL_INT
2290#define MAX_FC_NUM MAX_US_INT
2291
2292/* the purpose is to detect overflow of an unsigned short */
2293#define REISERFS_LINK_MAX (MAX_US_INT - 1000)
2294
Jeff Mahoney098297b2014-04-23 10:00:36 -04002295/*
2296 * The following defines are used in reiserfs_insert_item
2297 * and reiserfs_append_item
2298 */
2299#define REISERFS_KERNEL_MEM 0 /* kernel memory mode */
2300#define REISERFS_USER_MEM 1 /* user memory mode */
Al Virof466c6f2012-03-17 01:16:43 -04002301
2302#define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
2303#define get_generation(s) atomic_read (&fs_generation(s))
2304#define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen)
2305#define __fs_changed(gen,s) (gen != get_generation (s))
2306#define fs_changed(gen,s) \
2307({ \
2308 reiserfs_cond_resched(s); \
2309 __fs_changed(gen, s); \
2310})
2311
Jeff Mahoney098297b2014-04-23 10:00:36 -04002312/***************************************************************************
2313 * FIXATE NODES *
2314 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -04002315
2316#define VI_TYPE_LEFT_MERGEABLE 1
2317#define VI_TYPE_RIGHT_MERGEABLE 2
2318
Jeff Mahoney098297b2014-04-23 10:00:36 -04002319/*
2320 * To make any changes in the tree we always first find node, that
2321 * contains item to be changed/deleted or place to insert a new
2322 * item. We call this node S. To do balancing we need to decide what
2323 * we will shift to left/right neighbor, or to a new node, where new
2324 * item will be etc. To make this analysis simpler we build virtual
2325 * node. Virtual node is an array of items, that will replace items of
2326 * node S. (For instance if we are going to delete an item, virtual
2327 * node does not contain it). Virtual node keeps information about
2328 * item sizes and types, mergeability of first and last items, sizes
2329 * of all entries in directory item. We use this array of items when
2330 * calculating what we can shift to neighbors and how many nodes we
2331 * have to have if we do not any shiftings, if we shift to left/right
2332 * neighbor or to both.
2333 */
Al Virof466c6f2012-03-17 01:16:43 -04002334struct virtual_item {
Jeff Mahoney098297b2014-04-23 10:00:36 -04002335 int vi_index; /* index in the array of item operations */
2336 unsigned short vi_type; /* left/right mergeability */
2337
2338 /* length of item that it will have after balancing */
2339 unsigned short vi_item_len;
2340
Al Virof466c6f2012-03-17 01:16:43 -04002341 struct item_head *vi_ih;
Jeff Mahoney098297b2014-04-23 10:00:36 -04002342 const char *vi_item; /* body of item (old or new) */
2343 const void *vi_new_data; /* 0 always but paste mode */
2344 void *vi_uarea; /* item specific area */
Al Virof466c6f2012-03-17 01:16:43 -04002345};
2346
2347struct virtual_node {
Jeff Mahoney098297b2014-04-23 10:00:36 -04002348 /* this is a pointer to the free space in the buffer */
2349 char *vn_free_ptr;
2350
Al Virof466c6f2012-03-17 01:16:43 -04002351 unsigned short vn_nr_item; /* number of items in virtual node */
Jeff Mahoney098297b2014-04-23 10:00:36 -04002352
2353 /*
2354 * size of node , that node would have if it has
2355 * unlimited size and no balancing is performed
2356 */
2357 short vn_size;
2358
2359 /* mode of balancing (paste, insert, delete, cut) */
2360 short vn_mode;
2361
Al Virof466c6f2012-03-17 01:16:43 -04002362 short vn_affected_item_num;
2363 short vn_pos_in_item;
Jeff Mahoney098297b2014-04-23 10:00:36 -04002364
2365 /* item header of inserted item, 0 for other modes */
2366 struct item_head *vn_ins_ih;
Al Virof466c6f2012-03-17 01:16:43 -04002367 const void *vn_data;
Jeff Mahoney098297b2014-04-23 10:00:36 -04002368
2369 /* array of items (including a new one, excluding item to be deleted) */
2370 struct virtual_item *vn_vi;
Al Virof466c6f2012-03-17 01:16:43 -04002371};
2372
2373/* used by directory items when creating virtual nodes */
2374struct direntry_uarea {
2375 int flags;
2376 __u16 entry_count;
2377 __u16 entry_sizes[1];
2378} __attribute__ ((__packed__));
2379
Jeff Mahoney098297b2014-04-23 10:00:36 -04002380/***************************************************************************
2381 * TREE BALANCE *
2382 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -04002383
Jeff Mahoney098297b2014-04-23 10:00:36 -04002384/*
2385 * This temporary structure is used in tree balance algorithms, and
2386 * constructed as we go to the extent that its various parts are
2387 * needed. It contains arrays of nodes that can potentially be
2388 * involved in the balancing of node S, and parameters that define how
2389 * each of the nodes must be balanced. Note that in these algorithms
2390 * for balancing the worst case is to need to balance the current node
2391 * S and the left and right neighbors and all of their parents plus
2392 * create a new node. We implement S1 balancing for the leaf nodes
2393 * and S0 balancing for the internal nodes (S1 and S0 are defined in
2394 * our papers.)
2395 */
Al Virof466c6f2012-03-17 01:16:43 -04002396
Jeff Mahoney098297b2014-04-23 10:00:36 -04002397/* size of the array of buffers to free at end of do_balance */
2398#define MAX_FREE_BLOCK 7
Al Virof466c6f2012-03-17 01:16:43 -04002399
2400/* maximum number of FEB blocknrs on a single level */
2401#define MAX_AMOUNT_NEEDED 2
2402
2403/* someday somebody will prefix every field in this struct with tb_ */
2404struct tree_balance {
2405 int tb_mode;
2406 int need_balance_dirty;
2407 struct super_block *tb_sb;
2408 struct reiserfs_transaction_handle *transaction_handle;
2409 struct treepath *tb_path;
Al Virof466c6f2012-03-17 01:16:43 -04002410
Jeff Mahoney098297b2014-04-23 10:00:36 -04002411 /* array of left neighbors of nodes in the path */
2412 struct buffer_head *L[MAX_HEIGHT];
2413
2414 /* array of right neighbors of nodes in the path */
2415 struct buffer_head *R[MAX_HEIGHT];
2416
2417 /* array of fathers of the left neighbors */
2418 struct buffer_head *FL[MAX_HEIGHT];
2419
2420 /* array of fathers of the right neighbors */
2421 struct buffer_head *FR[MAX_HEIGHT];
2422 /* array of common parents of center node and its left neighbor */
2423 struct buffer_head *CFL[MAX_HEIGHT];
2424
2425 /* array of common parents of center node and its right neighbor */
2426 struct buffer_head *CFR[MAX_HEIGHT];
2427
2428 /*
2429 * array of empty buffers. Number of buffers in array equals
2430 * cur_blknum.
2431 */
2432 struct buffer_head *FEB[MAX_FEB_SIZE];
Al Virof466c6f2012-03-17 01:16:43 -04002433 struct buffer_head *used[MAX_FEB_SIZE];
2434 struct buffer_head *thrown[MAX_FEB_SIZE];
Jeff Mahoney098297b2014-04-23 10:00:36 -04002435
2436 /*
2437 * array of number of items which must be shifted to the left in
2438 * order to balance the current node; for leaves includes item that
2439 * will be partially shifted; for internal nodes, it is the number
2440 * of child pointers rather than items. It includes the new item
2441 * being created. The code sometimes subtracts one to get the
2442 * number of wholly shifted items for other purposes.
2443 */
2444 int lnum[MAX_HEIGHT];
2445
2446 /* substitute right for left in comment above */
2447 int rnum[MAX_HEIGHT];
2448
2449 /*
2450 * array indexed by height h mapping the key delimiting L[h] and
2451 * S[h] to its item number within the node CFL[h]
2452 */
2453 int lkey[MAX_HEIGHT];
2454
2455 /* substitute r for l in comment above */
2456 int rkey[MAX_HEIGHT];
2457
2458 /*
2459 * the number of bytes by we are trying to add or remove from
2460 * S[h]. A negative value means removing.
2461 */
2462 int insert_size[MAX_HEIGHT];
2463
2464 /*
2465 * number of nodes that will replace node S[h] after balancing
2466 * on the level h of the tree. If 0 then S is being deleted,
2467 * if 1 then S is remaining and no new nodes are being created,
2468 * if 2 or 3 then 1 or 2 new nodes is being created
2469 */
2470 int blknum[MAX_HEIGHT];
Al Virof466c6f2012-03-17 01:16:43 -04002471
2472 /* fields that are used only for balancing leaves of the tree */
Al Virof466c6f2012-03-17 01:16:43 -04002473
Jeff Mahoney098297b2014-04-23 10:00:36 -04002474 /* number of empty blocks having been already allocated */
2475 int cur_blknum;
2476
2477 /* number of items that fall into left most node when S[0] splits */
2478 int s0num;
2479
Jeff Mahoney098297b2014-04-23 10:00:36 -04002480 /*
2481 * number of bytes which can flow to the left neighbor from the left
2482 * most liquid item that cannot be shifted from S[0] entirely
2483 * if -1 then nothing will be partially shifted
2484 */
2485 int lbytes;
2486
2487 /*
2488 * number of bytes which will flow to the right neighbor from the right
2489 * most liquid item that cannot be shifted from S[0] entirely
2490 * if -1 then nothing will be partially shifted
2491 */
2492 int rbytes;
2493
Jeff Mahoneyb49fb112014-04-23 10:00:46 -04002494
Jeff Mahoney098297b2014-04-23 10:00:36 -04002495 /*
Jeff Mahoneyb49fb112014-04-23 10:00:46 -04002496 * index into the array of item headers in
2497 * S[0] of the affected item
2498 */
2499 int item_pos;
2500
2501 /* new nodes allocated to hold what could not fit into S */
2502 struct buffer_head *S_new[2];
2503
2504 /*
2505 * number of items that will be placed into nodes in S_new
2506 * when S[0] splits
2507 */
2508 int snum[2];
2509
2510 /*
2511 * number of bytes which flow to nodes in S_new when S[0] splits
Jeff Mahoney098297b2014-04-23 10:00:36 -04002512 * note: if S[0] splits into 3 nodes, then items do not need to be cut
2513 */
Jeff Mahoneyb49fb112014-04-23 10:00:46 -04002514 int sbytes[2];
2515
2516 int pos_in_item;
2517 int zeroes_num;
Jeff Mahoney098297b2014-04-23 10:00:36 -04002518
2519 /*
2520 * buffers which are to be freed after do_balance finishes
2521 * by unfix_nodes
2522 */
2523 struct buffer_head *buf_to_free[MAX_FREE_BLOCK];
2524
2525 /*
2526 * kmalloced memory. Used to create virtual node and keep
2527 * map of dirtied bitmap blocks
2528 */
2529 char *vn_buf;
2530
2531 int vn_buf_size; /* size of the vn_buf */
2532
2533 /* VN starts after bitmap of bitmap blocks */
2534 struct virtual_node *tb_vn;
2535
2536 /*
2537 * saved value of `reiserfs_generation' counter see
2538 * FILESYSTEM_CHANGED() macro in reiserfs_fs.h
2539 */
2540 int fs_gen;
2541
Al Virof466c6f2012-03-17 01:16:43 -04002542#ifdef DISPLACE_NEW_PACKING_LOCALITIES
Jeff Mahoney098297b2014-04-23 10:00:36 -04002543 /*
2544 * key pointer, to pass to block allocator or
2545 * another low-level subsystem
2546 */
2547 struct in_core_key key;
Al Virof466c6f2012-03-17 01:16:43 -04002548#endif
2549};
2550
2551/* These are modes of balancing */
2552
2553/* When inserting an item. */
2554#define M_INSERT 'i'
Jeff Mahoney098297b2014-04-23 10:00:36 -04002555/*
2556 * When inserting into (directories only) or appending onto an already
2557 * existent item.
2558 */
Al Virof466c6f2012-03-17 01:16:43 -04002559#define M_PASTE 'p'
2560/* When deleting an item. */
2561#define M_DELETE 'd'
2562/* When truncating an item or removing an entry from a (directory) item. */
Jeff Mahoney098297b2014-04-23 10:00:36 -04002563#define M_CUT 'c'
Al Virof466c6f2012-03-17 01:16:43 -04002564
2565/* used when balancing on leaf level skipped (in reiserfsck) */
2566#define M_INTERNAL 'n'
2567
Jeff Mahoney098297b2014-04-23 10:00:36 -04002568/*
2569 * When further balancing is not needed, then do_balance does not need
2570 * to be called.
2571 */
2572#define M_SKIP_BALANCING 's'
Al Virof466c6f2012-03-17 01:16:43 -04002573#define M_CONVERT 'v'
2574
2575/* modes of leaf_move_items */
2576#define LEAF_FROM_S_TO_L 0
2577#define LEAF_FROM_S_TO_R 1
2578#define LEAF_FROM_R_TO_L 2
2579#define LEAF_FROM_L_TO_R 3
2580#define LEAF_FROM_S_TO_SNEW 4
2581
2582#define FIRST_TO_LAST 0
2583#define LAST_TO_FIRST 1
2584
Jeff Mahoney098297b2014-04-23 10:00:36 -04002585/*
2586 * used in do_balance for passing parent of node information that has
2587 * been gotten from tb struct
2588 */
Al Virof466c6f2012-03-17 01:16:43 -04002589struct buffer_info {
2590 struct tree_balance *tb;
2591 struct buffer_head *bi_bh;
2592 struct buffer_head *bi_parent;
2593 int bi_position;
2594};
2595
2596static inline struct super_block *sb_from_tb(struct tree_balance *tb)
2597{
2598 return tb ? tb->tb_sb : NULL;
2599}
2600
2601static inline struct super_block *sb_from_bi(struct buffer_info *bi)
2602{
2603 return bi ? sb_from_tb(bi->tb) : NULL;
2604}
2605
Jeff Mahoney098297b2014-04-23 10:00:36 -04002606/*
2607 * there are 4 types of items: stat data, directory item, indirect, direct.
2608 * +-------------------+------------+--------------+------------+
2609 * | | k_offset | k_uniqueness | mergeable? |
2610 * +-------------------+------------+--------------+------------+
2611 * | stat data | 0 | 0 | no |
2612 * +-------------------+------------+--------------+------------+
2613 * | 1st directory item| DOT_OFFSET | DIRENTRY_ .. | no |
2614 * | non 1st directory | hash value | UNIQUENESS | yes |
2615 * | item | | | |
2616 * +-------------------+------------+--------------+------------+
2617 * | indirect item | offset + 1 |TYPE_INDIRECT | [1] |
2618 * +-------------------+------------+--------------+------------+
2619 * | direct item | offset + 1 |TYPE_DIRECT | [2] |
2620 * +-------------------+------------+--------------+------------+
2621 *
2622 * [1] if this is not the first indirect item of the object
2623 * [2] if this is not the first direct item of the object
Al Virof466c6f2012-03-17 01:16:43 -04002624*/
2625
2626struct item_operations {
2627 int (*bytes_number) (struct item_head * ih, int block_size);
2628 void (*decrement_key) (struct cpu_key *);
2629 int (*is_left_mergeable) (struct reiserfs_key * ih,
2630 unsigned long bsize);
2631 void (*print_item) (struct item_head *, char *item);
2632 void (*check_item) (struct item_head *, char *item);
2633
2634 int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
2635 int is_affected, int insert_size);
2636 int (*check_left) (struct virtual_item * vi, int free,
2637 int start_skip, int end_skip);
2638 int (*check_right) (struct virtual_item * vi, int free);
2639 int (*part_size) (struct virtual_item * vi, int from, int to);
2640 int (*unit_num) (struct virtual_item * vi);
2641 void (*print_vi) (struct virtual_item * vi);
2642};
2643
2644extern struct item_operations *item_ops[TYPE_ANY + 1];
2645
2646#define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
2647#define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
2648#define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item)
2649#define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item)
2650#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)
2651#define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
2652#define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free)
2653#define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to)
2654#define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi)
2655#define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi)
2656
2657#define COMP_SHORT_KEYS comp_short_keys
2658
2659/* number of blocks pointed to by the indirect item */
2660#define I_UNFM_NUM(ih) (ih_item_len(ih) / UNFM_P_SIZE)
2661
Jeff Mahoney098297b2014-04-23 10:00:36 -04002662/*
2663 * the used space within the unformatted node corresponding
2664 * to pos within the item pointed to by ih
2665 */
Al Virof466c6f2012-03-17 01:16:43 -04002666#define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
2667
Jeff Mahoney098297b2014-04-23 10:00:36 -04002668/*
2669 * number of bytes contained by the direct item or the
2670 * unformatted nodes the indirect item points to
2671 */
Al Virof466c6f2012-03-17 01:16:43 -04002672
Jeff Mahoney098297b2014-04-23 10:00:36 -04002673/* following defines use reiserfs buffer header and item header */
Al Virof466c6f2012-03-17 01:16:43 -04002674
2675/* get stat-data */
2676#define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
2677
Jeff Mahoney098297b2014-04-23 10:00:36 -04002678/* this is 3976 for size==4096 */
Al Virof466c6f2012-03-17 01:16:43 -04002679#define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
2680
Jeff Mahoney098297b2014-04-23 10:00:36 -04002681/*
2682 * indirect items consist of entries which contain blocknrs, pos
2683 * indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
2684 * blocknr contained by the entry pos points to
2685 */
Jeff Mahoney4cf5f7a2014-04-23 10:00:35 -04002686#define B_I_POS_UNFM_POINTER(bh, ih, pos) \
2687 le32_to_cpu(*(((unp_t *)ih_item_body(bh, ih)) + (pos)))
2688#define PUT_B_I_POS_UNFM_POINTER(bh, ih, pos, val) \
2689 (*(((unp_t *)ih_item_body(bh, ih)) + (pos)) = cpu_to_le32(val))
Al Virof466c6f2012-03-17 01:16:43 -04002690
2691struct reiserfs_iget_args {
2692 __u32 objectid;
2693 __u32 dirid;
2694};
2695
Jeff Mahoney098297b2014-04-23 10:00:36 -04002696/***************************************************************************
2697 * FUNCTION DECLARATIONS *
2698 ***************************************************************************/
Al Virof466c6f2012-03-17 01:16:43 -04002699
2700#define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
2701
2702#define journal_trans_half(blocksize) \
2703 ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
2704
2705/* journal.c see journal.c for all the comments here */
2706
2707/* first block written in a commit. */
2708struct reiserfs_journal_desc {
2709 __le32 j_trans_id; /* id of commit */
Jeff Mahoney098297b2014-04-23 10:00:36 -04002710
2711 /* length of commit. len +1 is the commit block */
2712 __le32 j_len;
2713
Al Virof466c6f2012-03-17 01:16:43 -04002714 __le32 j_mount_id; /* mount id of this trans */
2715 __le32 j_realblock[1]; /* real locations for each block */
2716};
2717
2718#define get_desc_trans_id(d) le32_to_cpu((d)->j_trans_id)
2719#define get_desc_trans_len(d) le32_to_cpu((d)->j_len)
2720#define get_desc_mount_id(d) le32_to_cpu((d)->j_mount_id)
2721
2722#define set_desc_trans_id(d,val) do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
2723#define set_desc_trans_len(d,val) do { (d)->j_len = cpu_to_le32 (val); } while (0)
2724#define set_desc_mount_id(d,val) do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
2725
2726/* last block written in a commit */
2727struct reiserfs_journal_commit {
2728 __le32 j_trans_id; /* must match j_trans_id from the desc block */
2729 __le32 j_len; /* ditto */
2730 __le32 j_realblock[1]; /* real locations for each block */
2731};
2732
2733#define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
2734#define get_commit_trans_len(c) le32_to_cpu((c)->j_len)
2735#define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
2736
2737#define set_commit_trans_id(c,val) do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
2738#define set_commit_trans_len(c,val) do { (c)->j_len = cpu_to_le32 (val); } while (0)
2739
Jeff Mahoney098297b2014-04-23 10:00:36 -04002740/*
2741 * this header block gets written whenever a transaction is considered
2742 * fully flushed, and is more recent than the last fully flushed transaction.
2743 * fully flushed means all the log blocks and all the real blocks are on
2744 * disk, and this transaction does not need to be replayed.
2745 */
Al Virof466c6f2012-03-17 01:16:43 -04002746struct reiserfs_journal_header {
Jeff Mahoney098297b2014-04-23 10:00:36 -04002747 /* id of last fully flushed transaction */
2748 __le32 j_last_flush_trans_id;
2749
2750 /* offset in the log of where to start replay after a crash */
2751 __le32 j_first_unflushed_offset;
2752
Al Virof466c6f2012-03-17 01:16:43 -04002753 __le32 j_mount_id;
2754 /* 12 */ struct journal_params jh_journal;
2755};
2756
2757/* biggest tunable defines are right here */
2758#define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */
Jeff Mahoney098297b2014-04-23 10:00:36 -04002759
2760/* biggest possible single transaction, don't change for now (8/3/99) */
2761#define JOURNAL_TRANS_MAX_DEFAULT 1024
Al Virof466c6f2012-03-17 01:16:43 -04002762#define JOURNAL_TRANS_MIN_DEFAULT 256
Jeff Mahoney098297b2014-04-23 10:00:36 -04002763
2764/*
2765 * max blocks to batch into one transaction,
2766 * don't make this any bigger than 900
2767 */
2768#define JOURNAL_MAX_BATCH_DEFAULT 900
Al Virof466c6f2012-03-17 01:16:43 -04002769#define JOURNAL_MIN_RATIO 2
2770#define JOURNAL_MAX_COMMIT_AGE 30
2771#define JOURNAL_MAX_TRANS_AGE 30
2772#define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
2773#define JOURNAL_BLOCKS_PER_OBJECT(sb) (JOURNAL_PER_BALANCE_CNT * 3 + \
2774 2 * (REISERFS_QUOTA_INIT_BLOCKS(sb) + \
2775 REISERFS_QUOTA_TRANS_BLOCKS(sb)))
2776
2777#ifdef CONFIG_QUOTA
2778#define REISERFS_QUOTA_OPTS ((1 << REISERFS_USRQUOTA) | (1 << REISERFS_GRPQUOTA))
2779/* We need to update data and inode (atime) */
2780#define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? 2 : 0)
2781/* 1 balancing, 1 bitmap, 1 data per write + stat data update */
2782#define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
2783(DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0)
2784/* same as with INIT */
2785#define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
2786(DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0)
2787#else
2788#define REISERFS_QUOTA_TRANS_BLOCKS(s) 0
2789#define REISERFS_QUOTA_INIT_BLOCKS(s) 0
2790#define REISERFS_QUOTA_DEL_BLOCKS(s) 0
2791#endif
2792
Jeff Mahoney098297b2014-04-23 10:00:36 -04002793/*
2794 * both of these can be as low as 1, or as high as you want. The min is the
2795 * number of 4k bitmap nodes preallocated on mount. New nodes are allocated
2796 * as needed, and released when transactions are committed. On release, if
2797 * the current number of nodes is > max, the node is freed, otherwise,
2798 * it is put on a free list for faster use later.
Al Virof466c6f2012-03-17 01:16:43 -04002799*/
2800#define REISERFS_MIN_BITMAP_NODES 10
2801#define REISERFS_MAX_BITMAP_NODES 100
2802
Jeff Mahoney098297b2014-04-23 10:00:36 -04002803/* these are based on journal hash size of 8192 */
2804#define JBH_HASH_SHIFT 13
Al Virof466c6f2012-03-17 01:16:43 -04002805#define JBH_HASH_MASK 8191
2806
2807#define _jhashfn(sb,block) \
2808 (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
2809 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
2810#define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
2811
Jeff Mahoney098297b2014-04-23 10:00:36 -04002812/* We need these to make journal.c code more readable */
Al Virof466c6f2012-03-17 01:16:43 -04002813#define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2814#define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2815#define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2816
2817enum reiserfs_bh_state_bits {
2818 BH_JDirty = BH_PrivateStart, /* buffer is in current transaction */
2819 BH_JDirty_wait,
Jeff Mahoney098297b2014-04-23 10:00:36 -04002820 /*
2821 * disk block was taken off free list before being in a
2822 * finished transaction, or written to disk. Can be reused immed.
2823 */
2824 BH_JNew,
Al Virof466c6f2012-03-17 01:16:43 -04002825 BH_JPrepared,
2826 BH_JRestore_dirty,
Jeff Mahoney098297b2014-04-23 10:00:36 -04002827 BH_JTest, /* debugging only will go away */
Al Virof466c6f2012-03-17 01:16:43 -04002828};
2829
2830BUFFER_FNS(JDirty, journaled);
2831TAS_BUFFER_FNS(JDirty, journaled);
2832BUFFER_FNS(JDirty_wait, journal_dirty);
2833TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
2834BUFFER_FNS(JNew, journal_new);
2835TAS_BUFFER_FNS(JNew, journal_new);
2836BUFFER_FNS(JPrepared, journal_prepared);
2837TAS_BUFFER_FNS(JPrepared, journal_prepared);
2838BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
2839TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
2840BUFFER_FNS(JTest, journal_test);
2841TAS_BUFFER_FNS(JTest, journal_test);
2842
Jeff Mahoney098297b2014-04-23 10:00:36 -04002843/* transaction handle which is passed around for all journal calls */
Al Virof466c6f2012-03-17 01:16:43 -04002844struct reiserfs_transaction_handle {
Jeff Mahoney098297b2014-04-23 10:00:36 -04002845 /*
2846 * super for this FS when journal_begin was called. saves calls to
2847 * reiserfs_get_super also used by nested transactions to make
2848 * sure they are nesting on the right FS _must_ be first
2849 * in the handle
2850 */
2851 struct super_block *t_super;
2852
Al Virof466c6f2012-03-17 01:16:43 -04002853 int t_refcount;
2854 int t_blocks_logged; /* number of blocks this writer has logged */
2855 int t_blocks_allocated; /* number of blocks this writer allocated */
Jeff Mahoney098297b2014-04-23 10:00:36 -04002856
2857 /* sanity check, equals the current trans id */
2858 unsigned int t_trans_id;
2859
Al Virof466c6f2012-03-17 01:16:43 -04002860 void *t_handle_save; /* save existing current->journal_info */
Jeff Mahoney098297b2014-04-23 10:00:36 -04002861
2862 /*
2863 * if new block allocation occurres, that block
2864 * should be displaced from others
2865 */
2866 unsigned displace_new_blocks:1;
2867
Al Virof466c6f2012-03-17 01:16:43 -04002868 struct list_head t_list;
2869};
2870
Jeff Mahoney098297b2014-04-23 10:00:36 -04002871/*
2872 * used to keep track of ordered and tail writes, attached to the buffer
Al Virof466c6f2012-03-17 01:16:43 -04002873 * head through b_journal_head.
2874 */
2875struct reiserfs_jh {
2876 struct reiserfs_journal_list *jl;
2877 struct buffer_head *bh;
2878 struct list_head list;
2879};
2880
2881void reiserfs_free_jh(struct buffer_head *bh);
2882int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
2883int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
2884int journal_mark_dirty(struct reiserfs_transaction_handle *,
Jeff Mahoney09f1b802014-04-23 10:00:39 -04002885 struct buffer_head *bh);
Al Virof466c6f2012-03-17 01:16:43 -04002886
2887static inline int reiserfs_file_data_log(struct inode *inode)
2888{
2889 if (reiserfs_data_log(inode->i_sb) ||
2890 (REISERFS_I(inode)->i_flags & i_data_log))
2891 return 1;
2892 return 0;
2893}
2894
2895static inline int reiserfs_transaction_running(struct super_block *s)
2896{
2897 struct reiserfs_transaction_handle *th = current->journal_info;
2898 if (th && th->t_super == s)
2899 return 1;
2900 if (th && th->t_super == NULL)
2901 BUG();
2902 return 0;
2903}
2904
2905static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th)
2906{
2907 return th->t_blocks_allocated - th->t_blocks_logged;
2908}
2909
2910struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
2911 super_block
2912 *,
2913 int count);
2914int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
Marco Stornellicfac4b42012-12-15 11:47:31 +01002915void reiserfs_vfs_truncate_file(struct inode *inode);
Al Virof466c6f2012-03-17 01:16:43 -04002916int reiserfs_commit_page(struct inode *inode, struct page *page,
2917 unsigned from, unsigned to);
Artem Bityutskiy25729b02012-06-01 17:18:05 +03002918void reiserfs_flush_old_commits(struct super_block *);
Al Virof466c6f2012-03-17 01:16:43 -04002919int reiserfs_commit_for_inode(struct inode *);
2920int reiserfs_inode_needs_commit(struct inode *);
2921void reiserfs_update_inode_transaction(struct inode *);
2922void reiserfs_wait_on_write_block(struct super_block *s);
2923void reiserfs_block_writes(struct reiserfs_transaction_handle *th);
2924void reiserfs_allow_writes(struct super_block *s);
2925void reiserfs_check_lock_depth(struct super_block *s, char *caller);
2926int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh,
2927 int wait);
2928void reiserfs_restore_prepared_buffer(struct super_block *,
2929 struct buffer_head *bh);
2930int journal_init(struct super_block *, const char *j_dev_name, int old_format,
2931 unsigned int);
2932int journal_release(struct reiserfs_transaction_handle *, struct super_block *);
2933int journal_release_error(struct reiserfs_transaction_handle *,
2934 struct super_block *);
Jeff Mahoney58d85422014-04-23 10:00:38 -04002935int journal_end(struct reiserfs_transaction_handle *);
2936int journal_end_sync(struct reiserfs_transaction_handle *);
Al Virof466c6f2012-03-17 01:16:43 -04002937int journal_mark_freed(struct reiserfs_transaction_handle *,
2938 struct super_block *, b_blocknr_t blocknr);
2939int journal_transaction_should_end(struct reiserfs_transaction_handle *, int);
2940int reiserfs_in_journal(struct super_block *sb, unsigned int bmap_nr,
2941 int bit_nr, int searchall, b_blocknr_t *next);
2942int journal_begin(struct reiserfs_transaction_handle *,
2943 struct super_block *sb, unsigned long);
2944int journal_join_abort(struct reiserfs_transaction_handle *,
Jeff Mahoneyb491dd12014-04-23 10:00:40 -04002945 struct super_block *sb);
Al Virof466c6f2012-03-17 01:16:43 -04002946void reiserfs_abort_journal(struct super_block *sb, int errno);
2947void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...);
2948int reiserfs_allocate_list_bitmaps(struct super_block *s,
2949 struct reiserfs_list_bitmap *, unsigned int);
2950
Artem Bityutskiy033369d2012-06-01 17:18:08 +03002951void reiserfs_schedule_old_flush(struct super_block *s);
Al Virof466c6f2012-03-17 01:16:43 -04002952void add_save_link(struct reiserfs_transaction_handle *th,
2953 struct inode *inode, int truncate);
2954int remove_save_link(struct inode *inode, int truncate);
2955
2956/* objectid.c */
2957__u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th);
2958void reiserfs_release_objectid(struct reiserfs_transaction_handle *th,
2959 __u32 objectid_to_release);
2960int reiserfs_convert_objectid_map_v1(struct super_block *);
2961
2962/* stree.c */
2963int B_IS_IN_TREE(const struct buffer_head *);
2964extern void copy_item_head(struct item_head *to,
2965 const struct item_head *from);
2966
Jeff Mahoney098297b2014-04-23 10:00:36 -04002967/* first key is in cpu form, second - le */
Al Virof466c6f2012-03-17 01:16:43 -04002968extern int comp_short_keys(const struct reiserfs_key *le_key,
2969 const struct cpu_key *cpu_key);
2970extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from);
2971
Jeff Mahoney098297b2014-04-23 10:00:36 -04002972/* both are in le form */
Al Virof466c6f2012-03-17 01:16:43 -04002973extern int comp_le_keys(const struct reiserfs_key *,
2974 const struct reiserfs_key *);
2975extern int comp_short_le_keys(const struct reiserfs_key *,
2976 const struct reiserfs_key *);
2977
Jeff Mahoney098297b2014-04-23 10:00:36 -04002978/* * get key version from on disk key - kludge */
Al Virof466c6f2012-03-17 01:16:43 -04002979static inline int le_key_version(const struct reiserfs_key *key)
2980{
2981 int type;
2982
2983 type = offset_v2_k_type(&(key->u.k_offset_v2));
2984 if (type != TYPE_DIRECT && type != TYPE_INDIRECT
2985 && type != TYPE_DIRENTRY)
2986 return KEY_FORMAT_3_5;
2987
2988 return KEY_FORMAT_3_6;
2989
2990}
2991
2992static inline void copy_key(struct reiserfs_key *to,
2993 const struct reiserfs_key *from)
2994{
2995 memcpy(to, from, KEY_SIZE);
2996}
2997
2998int comp_items(const struct item_head *stored_ih, const struct treepath *path);
2999const struct reiserfs_key *get_rkey(const struct treepath *chk_path,
3000 const struct super_block *sb);
3001int search_by_key(struct super_block *, const struct cpu_key *,
3002 struct treepath *, int);
3003#define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
3004int search_for_position_by_key(struct super_block *sb,
3005 const struct cpu_key *cpu_key,
3006 struct treepath *search_path);
3007extern void decrement_bcount(struct buffer_head *bh);
3008void decrement_counters_in_path(struct treepath *search_path);
3009void pathrelse(struct treepath *search_path);
3010int reiserfs_check_path(struct treepath *p);
3011void pathrelse_and_restore(struct super_block *s, struct treepath *search_path);
3012
3013int reiserfs_insert_item(struct reiserfs_transaction_handle *th,
3014 struct treepath *path,
3015 const struct cpu_key *key,
3016 struct item_head *ih,
3017 struct inode *inode, const char *body);
3018
3019int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th,
3020 struct treepath *path,
3021 const struct cpu_key *key,
3022 struct inode *inode,
3023 const char *body, int paste_size);
3024
3025int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,
3026 struct treepath *path,
3027 struct cpu_key *key,
3028 struct inode *inode,
3029 struct page *page, loff_t new_file_size);
3030
3031int reiserfs_delete_item(struct reiserfs_transaction_handle *th,
3032 struct treepath *path,
3033 const struct cpu_key *key,
3034 struct inode *inode, struct buffer_head *un_bh);
3035
3036void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
3037 struct inode *inode, struct reiserfs_key *key);
3038int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
3039 struct inode *inode);
3040int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,
3041 struct inode *inode, struct page *,
3042 int update_timestamps);
3043
3044#define i_block_size(inode) ((inode)->i_sb->s_blocksize)
3045#define file_size(inode) ((inode)->i_size)
3046#define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
3047
3048#define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
3049!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 )
3050
3051void padd_item(char *item, int total_length, int length);
3052
3053/* inode.c */
3054/* args for the create parameter of reiserfs_get_block */
Jeff Mahoney098297b2014-04-23 10:00:36 -04003055#define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */
3056#define GET_BLOCK_CREATE 1 /* add anything you need to find block */
3057#define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */
3058#define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
3059#define GET_BLOCK_NO_IMUX 8 /* i_mutex is not held, don't preallocate */
3060#define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */
Al Virof466c6f2012-03-17 01:16:43 -04003061
3062void reiserfs_read_locked_inode(struct inode *inode,
3063 struct reiserfs_iget_args *args);
3064int reiserfs_find_actor(struct inode *inode, void *p);
3065int reiserfs_init_locked_inode(struct inode *inode, void *p);
3066void reiserfs_evict_inode(struct inode *inode);
3067int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc);
3068int reiserfs_get_block(struct inode *inode, sector_t block,
3069 struct buffer_head *bh_result, int create);
3070struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3071 int fh_len, int fh_type);
3072struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
3073 int fh_len, int fh_type);
Al Virob0b03822012-04-02 14:34:06 -04003074int reiserfs_encode_fh(struct inode *inode, __u32 * data, int *lenp,
3075 struct inode *parent);
Al Virof466c6f2012-03-17 01:16:43 -04003076
3077int reiserfs_truncate_file(struct inode *, int update_timestamps);
3078void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset,
3079 int type, int key_length);
3080void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
3081 int version,
3082 loff_t offset, int type, int length, int entry_count);
3083struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key);
3084
3085struct reiserfs_security_handle;
3086int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
3087 struct inode *dir, umode_t mode,
3088 const char *symname, loff_t i_size,
3089 struct dentry *dentry, struct inode *inode,
3090 struct reiserfs_security_handle *security);
3091
3092void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
3093 struct inode *inode, loff_t size);
3094
3095static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
3096 struct inode *inode)
3097{
3098 reiserfs_update_sd_size(th, inode, inode->i_size);
3099}
3100
3101void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode);
3102void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs);
3103int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
3104
3105int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len);
3106
3107/* namei.c */
3108void set_de_name_and_namelen(struct reiserfs_dir_entry *de);
3109int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,
3110 struct treepath *path, struct reiserfs_dir_entry *de);
3111struct dentry *reiserfs_get_parent(struct dentry *);
3112
3113#ifdef CONFIG_REISERFS_PROC_INFO
3114int reiserfs_proc_info_init(struct super_block *sb);
3115int reiserfs_proc_info_done(struct super_block *sb);
3116int reiserfs_proc_info_global_init(void);
3117int reiserfs_proc_info_global_done(void);
3118
3119#define PROC_EXP( e ) e
3120
3121#define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
3122#define PROC_INFO_MAX( sb, field, value ) \
3123 __PINFO( sb ).field = \
3124 max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
3125#define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
3126#define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
3127#define PROC_INFO_BH_STAT( sb, bh, level ) \
3128 PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \
3129 PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \
3130 PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
3131#else
3132static inline int reiserfs_proc_info_init(struct super_block *sb)
3133{
3134 return 0;
3135}
3136
3137static inline int reiserfs_proc_info_done(struct super_block *sb)
3138{
3139 return 0;
3140}
3141
3142static inline int reiserfs_proc_info_global_init(void)
3143{
3144 return 0;
3145}
3146
3147static inline int reiserfs_proc_info_global_done(void)
3148{
3149 return 0;
3150}
3151
3152#define PROC_EXP( e )
3153#define VOID_V ( ( void ) 0 )
3154#define PROC_INFO_MAX( sb, field, value ) VOID_V
3155#define PROC_INFO_INC( sb, field ) VOID_V
3156#define PROC_INFO_ADD( sb, field, val ) VOID_V
3157#define PROC_INFO_BH_STAT(sb, bh, n_node_level) VOID_V
3158#endif
3159
3160/* dir.c */
3161extern const struct inode_operations reiserfs_dir_inode_operations;
3162extern const struct inode_operations reiserfs_symlink_inode_operations;
3163extern const struct inode_operations reiserfs_special_inode_operations;
3164extern const struct file_operations reiserfs_dir_operations;
Al Virocd62cda2013-05-17 22:58:58 -04003165int reiserfs_readdir_inode(struct inode *, struct dir_context *);
Al Virof466c6f2012-03-17 01:16:43 -04003166
3167/* tail_conversion.c */
3168int direct2indirect(struct reiserfs_transaction_handle *, struct inode *,
3169 struct treepath *, struct buffer_head *, loff_t);
3170int indirect2direct(struct reiserfs_transaction_handle *, struct inode *,
3171 struct page *, struct treepath *, const struct cpu_key *,
3172 loff_t, char *);
3173void reiserfs_unmap_buffer(struct buffer_head *);
3174
3175/* file.c */
3176extern const struct inode_operations reiserfs_file_inode_operations;
3177extern const struct file_operations reiserfs_file_operations;
3178extern const struct address_space_operations reiserfs_address_space_operations;
3179
3180/* fix_nodes.c */
3181
3182int fix_nodes(int n_op_mode, struct tree_balance *tb,
3183 struct item_head *ins_ih, const void *);
3184void unfix_nodes(struct tree_balance *);
3185
3186/* prints.c */
3187void __reiserfs_panic(struct super_block *s, const char *id,
3188 const char *function, const char *fmt, ...)
3189 __attribute__ ((noreturn));
3190#define reiserfs_panic(s, id, fmt, args...) \
3191 __reiserfs_panic(s, id, __func__, fmt, ##args)
3192void __reiserfs_error(struct super_block *s, const char *id,
3193 const char *function, const char *fmt, ...);
3194#define reiserfs_error(s, id, fmt, args...) \
3195 __reiserfs_error(s, id, __func__, fmt, ##args)
3196void reiserfs_info(struct super_block *s, const char *fmt, ...);
3197void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...);
3198void print_indirect_item(struct buffer_head *bh, int item_num);
3199void store_print_tb(struct tree_balance *tb);
3200void print_cur_tb(char *mes);
3201void print_de(struct reiserfs_dir_entry *de);
3202void print_bi(struct buffer_info *bi, char *mes);
3203#define PRINT_LEAF_ITEMS 1 /* print all items */
3204#define PRINT_DIRECTORY_ITEMS 2 /* print directory items */
3205#define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */
3206void print_block(struct buffer_head *bh, ...);
3207void print_bmap(struct super_block *s, int silent);
3208void print_bmap_block(int i, char *data, int size, int silent);
3209/*void print_super_block (struct super_block * s, char * mes);*/
3210void print_objectid_map(struct super_block *s);
3211void print_block_head(struct buffer_head *bh, char *mes);
3212void check_leaf(struct buffer_head *bh);
3213void check_internal(struct buffer_head *bh);
3214void print_statistics(struct super_block *s);
3215char *reiserfs_hashname(int code);
3216
3217/* lbalance.c */
3218int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num,
3219 int mov_bytes, struct buffer_head *Snew);
3220int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes);
3221int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes);
3222void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first,
3223 int del_num, int del_bytes);
3224void leaf_insert_into_buf(struct buffer_info *bi, int before,
Jeff Mahoney27d0e5b2014-08-04 19:51:47 -04003225 struct item_head * const inserted_item_ih,
3226 const char * const inserted_item_body,
Al Virof466c6f2012-03-17 01:16:43 -04003227 int zeros_number);
Jeff Mahoney27d0e5b2014-08-04 19:51:47 -04003228void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num,
3229 int pos_in_item, int paste_size,
3230 const char * const body, int zeros_number);
Al Virof466c6f2012-03-17 01:16:43 -04003231void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num,
3232 int pos_in_item, int cut_size);
3233void leaf_paste_entries(struct buffer_info *bi, int item_num, int before,
3234 int new_entry_count, struct reiserfs_de_head *new_dehs,
3235 const char *records, int paste_size);
3236/* ibalance.c */
3237int balance_internal(struct tree_balance *, int, int, struct item_head *,
3238 struct buffer_head **);
3239
3240/* do_balance.c */
3241void do_balance_mark_leaf_dirty(struct tree_balance *tb,
3242 struct buffer_head *bh, int flag);
3243#define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
3244#define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
3245
3246void do_balance(struct tree_balance *tb, struct item_head *ih,
3247 const char *body, int flag);
3248void reiserfs_invalidate_buffer(struct tree_balance *tb,
3249 struct buffer_head *bh);
3250
3251int get_left_neighbor_position(struct tree_balance *tb, int h);
3252int get_right_neighbor_position(struct tree_balance *tb, int h);
3253void replace_key(struct tree_balance *tb, struct buffer_head *, int,
3254 struct buffer_head *, int);
3255void make_empty_node(struct buffer_info *);
3256struct buffer_head *get_FEB(struct tree_balance *);
3257
3258/* bitmap.c */
3259
Jeff Mahoney098297b2014-04-23 10:00:36 -04003260/*
3261 * structure contains hints for block allocator, and it is a container for
3262 * arguments, such as node, search path, transaction_handle, etc.
3263 */
Al Virof466c6f2012-03-17 01:16:43 -04003264struct __reiserfs_blocknr_hint {
Jeff Mahoney098297b2014-04-23 10:00:36 -04003265 /* inode passed to allocator, if we allocate unf. nodes */
3266 struct inode *inode;
3267
Al Virof466c6f2012-03-17 01:16:43 -04003268 sector_t block; /* file offset, in blocks */
3269 struct in_core_key key;
Al Virof466c6f2012-03-17 01:16:43 -04003270
Jeff Mahoney098297b2014-04-23 10:00:36 -04003271 /*
3272 * search path, used by allocator to deternine search_start by
3273 * various ways
3274 */
3275 struct treepath *path;
3276
3277 /*
3278 * transaction handle is needed to log super blocks
3279 * and bitmap blocks changes
3280 */
3281 struct reiserfs_transaction_handle *th;
3282
3283 b_blocknr_t beg, end;
3284
3285 /*
3286 * a field used to transfer search start value (block number)
3287 * between different block allocator procedures
3288 * (determine_search_start() and others)
3289 */
3290 b_blocknr_t search_start;
3291
3292 /*
3293 * is set in determine_prealloc_size() function,
3294 * used by underlayed function that do actual allocation
3295 */
3296 int prealloc_size;
3297
3298 /*
3299 * the allocator uses different polices for getting disk
3300 * space for formatted/unformatted blocks with/without preallocation
3301 */
3302 unsigned formatted_node:1;
Al Virof466c6f2012-03-17 01:16:43 -04003303 unsigned preallocate:1;
3304};
3305
3306typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
3307
3308int reiserfs_parse_alloc_options(struct super_block *, char *);
3309void reiserfs_init_alloc_options(struct super_block *s);
3310
3311/*
3312 * given a directory, this will tell you what packing locality
3313 * to use for a new object underneat it. The locality is returned
3314 * in disk byte order (le).
3315 */
3316__le32 reiserfs_choose_packing(struct inode *dir);
3317
Rashika Kheriaea0856c2014-04-03 14:50:36 -07003318void show_alloc_options(struct seq_file *seq, struct super_block *s);
Al Virof466c6f2012-03-17 01:16:43 -04003319int reiserfs_init_bitmap_cache(struct super_block *sb);
3320void reiserfs_free_bitmap_cache(struct super_block *sb);
3321void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info);
3322struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap);
3323int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value);
3324void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *,
3325 b_blocknr_t, int for_unformatted);
3326int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int,
3327 int);
3328static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb,
3329 b_blocknr_t * new_blocknrs,
3330 int amount_needed)
3331{
3332 reiserfs_blocknr_hint_t hint = {
3333 .th = tb->transaction_handle,
3334 .path = tb->tb_path,
3335 .inode = NULL,
3336 .key = tb->key,
3337 .block = 0,
3338 .formatted_node = 1
3339 };
3340 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed,
3341 0);
3342}
3343
3344static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle
3345 *th, struct inode *inode,
3346 b_blocknr_t * new_blocknrs,
3347 struct treepath *path,
3348 sector_t block)
3349{
3350 reiserfs_blocknr_hint_t hint = {
3351 .th = th,
3352 .path = path,
3353 .inode = inode,
3354 .block = block,
3355 .formatted_node = 0,
3356 .preallocate = 0
3357 };
3358 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
3359}
3360
3361#ifdef REISERFS_PREALLOCATE
3362static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle
3363 *th, struct inode *inode,
3364 b_blocknr_t * new_blocknrs,
3365 struct treepath *path,
3366 sector_t block)
3367{
3368 reiserfs_blocknr_hint_t hint = {
3369 .th = th,
3370 .path = path,
3371 .inode = inode,
3372 .block = block,
3373 .formatted_node = 0,
3374 .preallocate = 1
3375 };
3376 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
3377}
3378
3379void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
3380 struct inode *inode);
3381void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th);
3382#endif
3383
3384/* hashes.c */
3385__u32 keyed_hash(const signed char *msg, int len);
3386__u32 yura_hash(const signed char *msg, int len);
3387__u32 r5_hash(const signed char *msg, int len);
3388
3389#define reiserfs_set_le_bit __set_bit_le
3390#define reiserfs_test_and_set_le_bit __test_and_set_bit_le
3391#define reiserfs_clear_le_bit __clear_bit_le
3392#define reiserfs_test_and_clear_le_bit __test_and_clear_bit_le
3393#define reiserfs_test_le_bit test_bit_le
3394#define reiserfs_find_next_zero_le_bit find_next_zero_bit_le
3395
Jeff Mahoney098297b2014-04-23 10:00:36 -04003396/*
3397 * sometimes reiserfs_truncate may require to allocate few new blocks
3398 * to perform indirect2direct conversion. People probably used to
3399 * think, that truncate should work without problems on a filesystem
3400 * without free disk space. They may complain that they can not
3401 * truncate due to lack of free disk space. This spare space allows us
3402 * to not worry about it. 500 is probably too much, but it should be
3403 * absolutely safe
3404 */
Al Virof466c6f2012-03-17 01:16:43 -04003405#define SPARE_SPACE 500
3406
3407/* prototypes from ioctl.c */
3408long reiserfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg);
3409long reiserfs_compat_ioctl(struct file *filp,
3410 unsigned int cmd, unsigned long arg);
3411int reiserfs_unpack(struct inode *inode, struct file *filp);