blob: e356c99f06590c0d48e1aa687f5c9465813608bb [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Copyright 1996, 1997, 1998 Hans Reiser, see reiserfs/README for licensing and copyright details
3 */
4
5 /* this file has an amazingly stupid
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07006 name, yura please fix it to be
7 reiserfs.h, and merge all the rest
8 of our .h files that are in this
9 directory into it. */
Linus Torvalds1da177e2005-04-16 15:20:36 -070010
11#ifndef _LINUX_REISER_FS_H
12#define _LINUX_REISER_FS_H
13
14#include <linux/types.h>
Jeff Garzike18fa702006-09-24 11:13:19 -040015#include <linux/magic.h>
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#ifdef __KERNEL__
18#include <linux/slab.h>
19#include <linux/interrupt.h>
20#include <linux/sched.h>
21#include <linux/workqueue.h>
22#include <asm/unaligned.h>
23#include <linux/bitops.h>
24#include <linux/proc_fs.h>
25#include <linux/smp_lock.h>
26#include <linux/buffer_head.h>
27#include <linux/reiserfs_fs_i.h>
28#include <linux/reiserfs_fs_sb.h>
29#endif
30
31/*
32 * include/linux/reiser_fs.h
33 *
34 * Reiser File System constants and structures
35 *
36 */
37
Jaswinder Singh Rajput750e1c12009-02-03 19:40:03 +053038/* ioctl's command */
39#define REISERFS_IOC_UNPACK _IOW(0xCD,1,long)
40/* define following flags to be the same as in ext2, so that chattr(1),
41 lsattr(1) will work with us. */
42#define REISERFS_IOC_GETFLAGS FS_IOC_GETFLAGS
43#define REISERFS_IOC_SETFLAGS FS_IOC_SETFLAGS
44#define REISERFS_IOC_GETVERSION FS_IOC_GETVERSION
45#define REISERFS_IOC_SETVERSION FS_IOC_SETVERSION
46
47#ifdef __KERNEL__
48/* the 32 bit compat definitions with int argument */
49#define REISERFS_IOC32_UNPACK _IOW(0xCD, 1, int)
50#define REISERFS_IOC32_GETFLAGS FS_IOC32_GETFLAGS
51#define REISERFS_IOC32_SETFLAGS FS_IOC32_SETFLAGS
52#define REISERFS_IOC32_GETVERSION FS_IOC32_GETVERSION
53#define REISERFS_IOC32_SETVERSION FS_IOC32_SETVERSION
54
55/* Locking primitives */
56/* Right now we are still falling back to (un)lock_kernel, but eventually that
57 would evolve into real per-fs locks */
58#define reiserfs_write_lock( sb ) lock_kernel()
59#define reiserfs_write_unlock( sb ) unlock_kernel()
60
61/* xattr stuff */
62#define REISERFS_XATTR_DIR_SEM(s) (REISERFS_SB(s)->xattr_dir_sem)
63struct fid;
64
Linus Torvalds1da177e2005-04-16 15:20:36 -070065/* in reading the #defines, it may help to understand that they employ
66 the following abbreviations:
67
68 B = Buffer
69 I = Item header
70 H = Height within the tree (should be changed to LEV)
71 N = Number of the item in the node
72 STAT = stat data
73 DEH = Directory Entry Header
74 EC = Entry Count
75 E = Entry number
76 UL = Unsigned Long
77 BLKH = BLocK Header
78 UNFM = UNForMatted node
79 DC = Disk Child
80 P = Path
81
82 These #defines are named by concatenating these abbreviations,
83 where first comes the arguments, and last comes the return value,
84 of the macro.
85
86*/
87
88#define USE_INODE_GENERATION_COUNTER
89
90#define REISERFS_PREALLOCATE
91#define DISPLACE_NEW_PACKING_LOCALITIES
92#define PREALLOCATION_SIZE 9
93
94/* n must be power of 2 */
95#define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
96
97// to be ok for alpha and others we have to align structures to 8 byte
98// boundary.
99// FIXME: do not change 4 by anything else: there is code which relies on that
100#define ROUND_UP(x) _ROUND_UP(x,8LL)
101
102/* debug levels. Right now, CONFIG_REISERFS_CHECK means print all debug
103** messages.
104*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700105#define REISERFS_DEBUG_CODE 5 /* extra messages to help find/debug errors */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700107void reiserfs_warning(struct super_block *s, const char *fmt, ...);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108/* assertions handling */
109
110/** always check a condition and panic if it's false. */
Al Viro2d954d02007-07-27 14:23:44 +0100111#define __RASSERT( cond, scond, format, args... ) \
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112if( !( cond ) ) \
Al Viro2d954d02007-07-27 14:23:44 +0100113 reiserfs_panic( NULL, "reiserfs[%i]: assertion " scond " failed at " \
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114 __FILE__ ":%i:%s: " format "\n", \
Harvey Harrisond5c003b2008-10-15 22:01:24 -0700115 in_interrupt() ? -1 : task_pid_nr(current), __LINE__ , __func__ , ##args )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116
Al Viro2d954d02007-07-27 14:23:44 +0100117#define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args)
118
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119#if defined( CONFIG_REISERFS_CHECK )
Al Viro2d954d02007-07-27 14:23:44 +0100120#define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121#else
122#define RFALSE( cond, format, args... ) do {;} while( 0 )
123#endif
124
125#define CONSTF __attribute_const__
126/*
127 * Disk Data Structures
128 */
129
130/***************************************************************************/
131/* SUPER BLOCK */
132/***************************************************************************/
133
134/*
135 * Structure of super block on disk, a version of which in RAM is often accessed as REISERFS_SB(s)->s_rs
136 * the version in RAM is part of a larger structure containing fields never written to disk.
137 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700138#define UNSET_HASH 0 // read_super will guess about, what hash names
139 // in directories were sorted with
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140#define TEA_HASH 1
141#define YURA_HASH 2
142#define R5_HASH 3
143#define DEFAULT_HASH R5_HASH
144
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145struct journal_params {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700146 __le32 jp_journal_1st_block; /* where does journal start from on its
147 * device */
148 __le32 jp_journal_dev; /* journal device st_rdev */
149 __le32 jp_journal_size; /* size of the journal */
150 __le32 jp_journal_trans_max; /* max number of blocks in a transaction. */
151 __le32 jp_journal_magic; /* random value made on fs creation (this
152 * was sb_journal_block_count) */
153 __le32 jp_journal_max_batch; /* max number of blocks to batch into a
154 * trans */
155 __le32 jp_journal_max_commit_age; /* in seconds, how old can an async
156 * commit be */
157 __le32 jp_journal_max_trans_age; /* in seconds, how old can a transaction
158 * be */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159};
160
161/* this is the super from 3.5.X, where X >= 10 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700162struct reiserfs_super_block_v1 {
163 __le32 s_block_count; /* blocks count */
164 __le32 s_free_blocks; /* free blocks count */
165 __le32 s_root_block; /* root block number */
166 struct journal_params s_journal;
167 __le16 s_blocksize; /* block size */
168 __le16 s_oid_maxsize; /* max size of object id array, see
169 * get_objectid() commentary */
170 __le16 s_oid_cursize; /* current size of object id array */
171 __le16 s_umount_state; /* this is set to 1 when filesystem was
172 * umounted, to 2 - when not */
173 char s_magic[10]; /* reiserfs magic string indicates that
174 * file system is reiserfs:
175 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs" */
176 __le16 s_fs_state; /* it is set to used by fsck to mark which
177 * phase of rebuilding is done */
178 __le32 s_hash_function_code; /* indicate, what hash function is being use
179 * to sort names in a directory*/
180 __le16 s_tree_height; /* height of disk tree */
181 __le16 s_bmap_nr; /* amount of bitmap blocks needed to address
182 * each block of file system */
183 __le16 s_version; /* this field is only reliable on filesystem
184 * with non-standard journal */
185 __le16 s_reserved_for_journal; /* size in blocks of journal area on main
186 * device, we need to keep after
187 * making fs with non-standard journal */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188} __attribute__ ((__packed__));
189
190#define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))
191
192/* this is the on disk super block */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700193struct reiserfs_super_block {
194 struct reiserfs_super_block_v1 s_v1;
195 __le32 s_inode_generation;
196 __le32 s_flags; /* Right now used only by inode-attributes, if enabled */
197 unsigned char s_uuid[16]; /* filesystem unique identifier */
198 unsigned char s_label[16]; /* filesystem volume label */
199 char s_unused[88]; /* zero filled by mkreiserfs and
200 * reiserfs_convert_objectid_map_v1()
201 * so any additions must be updated
202 * there as well. */
203} __attribute__ ((__packed__));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204
205#define SB_SIZE (sizeof(struct reiserfs_super_block))
206
207#define REISERFS_VERSION_1 0
208#define REISERFS_VERSION_2 2
209
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210// on-disk super block fields converted to cpu form
211#define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)
212#define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))
213#define SB_BLOCKSIZE(s) \
214 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))
215#define SB_BLOCK_COUNT(s) \
216 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))
217#define SB_FREE_BLOCKS(s) \
218 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))
219#define SB_REISERFS_MAGIC(s) \
220 (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)
221#define SB_ROOT_BLOCK(s) \
222 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))
223#define SB_TREE_HEIGHT(s) \
224 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))
225#define SB_REISERFS_STATE(s) \
226 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))
227#define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))
228#define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))
229
230#define PUT_SB_BLOCK_COUNT(s, val) \
231 do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)
232#define PUT_SB_FREE_BLOCKS(s, val) \
233 do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)
234#define PUT_SB_ROOT_BLOCK(s, val) \
235 do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)
236#define PUT_SB_TREE_HEIGHT(s, val) \
237 do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)
238#define PUT_SB_REISERFS_STATE(s, val) \
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700239 do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240#define PUT_SB_VERSION(s, val) \
241 do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)
242#define PUT_SB_BMAP_NR(s, val) \
243 do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)
244
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245#define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)
246#define SB_ONDISK_JOURNAL_SIZE(s) \
247 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))
248#define SB_ONDISK_JOURNAL_1st_BLOCK(s) \
249 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))
250#define SB_ONDISK_JOURNAL_DEVICE(s) \
251 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
252#define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
Al Virob8cc9362005-05-01 08:59:18 -0700253 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254
255#define is_block_in_log_or_reserved_area(s, block) \
256 block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
257 && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) + \
258 ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700259 SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700261int is_reiserfs_3_5(struct reiserfs_super_block *rs);
262int is_reiserfs_3_6(struct reiserfs_super_block *rs);
263int is_reiserfs_jr(struct reiserfs_super_block *rs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264
265/* ReiserFS leaves the first 64k unused, so that partition labels have
266 enough space. If someone wants to write a fancy bootloader that
267 needs more than 64k, let us know, and this will be increased in size.
268 This number must be larger than than the largest block size on any
269 platform, or code will break. -Hans */
270#define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
271#define REISERFS_FIRST_BLOCK unused_define
272#define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES
273
274/* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
275#define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
276
277// reiserfs internal error code (used by search_by_key adn fix_nodes))
278#define CARRY_ON 0
279#define REPEAT_SEARCH -1
280#define IO_ERROR -2
281#define NO_DISK_SPACE -3
282#define NO_BALANCING_NEEDED (-4)
283#define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
284#define QUOTA_EXCEEDED -6
285
286typedef __u32 b_blocknr_t;
Al Viro3e8962b2005-05-01 08:59:18 -0700287typedef __le32 unp_t;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288
289struct unfm_nodeinfo {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700290 unp_t unfm_nodenum;
291 unsigned short unfm_freespace;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292};
293
294/* there are two formats of keys: 3.5 and 3.6
295 */
296#define KEY_FORMAT_3_5 0
297#define KEY_FORMAT_3_6 1
298
299/* there are two stat datas */
300#define STAT_DATA_V1 0
301#define STAT_DATA_V2 1
302
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)
304{
305 return container_of(inode, struct reiserfs_inode_info, vfs_inode);
306}
307
308static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)
309{
310 return sb->s_fs_info;
311}
312
Jeff Mahoneycb680c12007-10-18 23:39:27 -0700313/* Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u16
314 * which overflows on large file systems. */
Mike Frysinger13d8bcd262008-02-08 04:21:54 -0800315static inline __u32 reiserfs_bmap_count(struct super_block *sb)
Jeff Mahoneycb680c12007-10-18 23:39:27 -0700316{
317 return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1;
318}
319
320static inline int bmap_would_wrap(unsigned bmap_nr)
321{
322 return bmap_nr > ((1LL << 16) - 1);
323}
324
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325/** this says about version of key of all items (but stat data) the
326 object consists of */
327#define get_inode_item_key_version( inode ) \
328 ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
329
330#define set_inode_item_key_version( inode, version ) \
331 ({ if((version)==KEY_FORMAT_3_6) \
332 REISERFS_I(inode)->i_flags |= i_item_key_version_mask; \
333 else \
334 REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })
335
336#define get_inode_sd_version(inode) \
337 ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
338
339#define set_inode_sd_version(inode, version) \
340 ({ if((version)==STAT_DATA_V2) \
341 REISERFS_I(inode)->i_flags |= i_stat_data_version_mask; \
342 else \
343 REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })
344
345/* This is an aggressive tail suppression policy, I am hoping it
346 improves our benchmarks. The principle behind it is that percentage
347 space saving is what matters, not absolute space saving. This is
348 non-intuitive, but it helps to understand it if you consider that the
349 cost to access 4 blocks is not much more than the cost to access 1
350 block, if you have to do a seek and rotate. A tail risks a
351 non-linear disk access that is significant as a percentage of total
352 time cost for a 4 block file and saves an amount of space that is
353 less significant as a percentage of space, or so goes the hypothesis.
354 -Hans */
355#define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
356(\
357 (!(n_tail_size)) || \
358 (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
359 ( (n_file_size) >= (n_block_size) * 4 ) || \
360 ( ( (n_file_size) >= (n_block_size) * 3 ) && \
361 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
362 ( ( (n_file_size) >= (n_block_size) * 2 ) && \
363 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
364 ( ( (n_file_size) >= (n_block_size) ) && \
365 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
366)
367
368/* Another strategy for tails, this one means only create a tail if all the
369 file would fit into one DIRECT item.
370 Primary intention for this one is to increase performance by decreasing
371 seeking.
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700372*/
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373#define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
374(\
375 (!(n_tail_size)) || \
376 (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
377)
378
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379/*
380 * values for s_umount_state field
381 */
382#define REISERFS_VALID_FS 1
383#define REISERFS_ERROR_FS 2
384
385//
386// there are 5 item types currently
387//
388#define TYPE_STAT_DATA 0
389#define TYPE_INDIRECT 1
390#define TYPE_DIRECT 2
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700391#define TYPE_DIRENTRY 3
392#define TYPE_MAXTYPE 3
393#define TYPE_ANY 15 // FIXME: comment is required
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394
395/***************************************************************************/
396/* KEY & ITEM HEAD */
397/***************************************************************************/
398
399//
400// directories use this key as well as old files
401//
402struct offset_v1 {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700403 __le32 k_offset;
404 __le32 k_uniqueness;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405} __attribute__ ((__packed__));
406
407struct offset_v2 {
Al Virof8e08a82005-05-01 08:59:19 -0700408 __le64 v;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409} __attribute__ ((__packed__));
410
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700411static inline __u16 offset_v2_k_type(const struct offset_v2 *v2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412{
Al Virof8e08a82005-05-01 08:59:19 -0700413 __u8 type = le64_to_cpu(v2->v) >> 60;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700414 return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415}
416
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700417static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type)
418{
419 v2->v =
420 (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60);
421}
422
423static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2)
424{
425 return le64_to_cpu(v2->v) & (~0ULL >> 4);
426}
427
428static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset)
429{
430 offset &= (~0ULL >> 4);
431 v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433
434/* Key of an item determines its location in the S+tree, and
435 is composed of 4 components */
436struct reiserfs_key {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700437 __le32 k_dir_id; /* packing locality: by default parent
438 directory object id */
439 __le32 k_objectid; /* object identifier */
440 union {
441 struct offset_v1 k_offset_v1;
442 struct offset_v2 k_offset_v2;
443 } __attribute__ ((__packed__)) u;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444} __attribute__ ((__packed__));
445
Al Viro6a3a16f2005-05-01 08:59:17 -0700446struct in_core_key {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700447 __u32 k_dir_id; /* packing locality: by default parent
448 directory object id */
449 __u32 k_objectid; /* object identifier */
450 __u64 k_offset;
451 __u8 k_type;
Al Viro6b9f5822005-05-01 08:59:19 -0700452};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453
454struct cpu_key {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700455 struct in_core_key on_disk_key;
456 int version;
457 int key_length; /* 3 in all cases but direct2indirect and
458 indirect2direct conversion */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459};
460
461/* Our function for comparing keys can compare keys of different
462 lengths. It takes as a parameter the length of the keys it is to
463 compare. These defines are used in determining what is to be passed
464 to it as that parameter. */
465#define REISERFS_FULL_KEY_LEN 4
466#define REISERFS_SHORT_KEY_LEN 2
467
468/* The result of the key compare */
469#define FIRST_GREATER 1
470#define SECOND_GREATER -1
471#define KEYS_IDENTICAL 0
472#define KEY_FOUND 1
473#define KEY_NOT_FOUND 0
474
475#define KEY_SIZE (sizeof(struct reiserfs_key))
476#define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
477
478/* return values for search_by_key and clones */
479#define ITEM_FOUND 1
480#define ITEM_NOT_FOUND 0
481#define ENTRY_FOUND 1
482#define ENTRY_NOT_FOUND 0
483#define DIRECTORY_NOT_FOUND -1
484#define REGULAR_FILE_FOUND -2
485#define DIRECTORY_FOUND -3
486#define BYTE_FOUND 1
487#define BYTE_NOT_FOUND 0
488#define FILE_NOT_FOUND -1
489
490#define POSITION_FOUND 1
491#define POSITION_NOT_FOUND 0
492
493// return values for reiserfs_find_entry and search_by_entry_key
494#define NAME_FOUND 1
495#define NAME_NOT_FOUND 0
496#define GOTO_PREVIOUS_ITEM 2
497#define NAME_FOUND_INVISIBLE 3
498
499/* Everything in the filesystem is stored as a set of items. The
500 item head contains the key of the item, its free space (for
501 indirect items) and specifies the location of the item itself
502 within the block. */
503
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700504struct item_head {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 /* Everything in the tree is found by searching for it based on
506 * its key.*/
507 struct reiserfs_key ih_key;
508 union {
509 /* The free space in the last unformatted node of an
510 indirect item if this is an indirect item. This
511 equals 0xFFFF iff this is a direct item or stat data
512 item. Note that the key, not this field, is used to
513 determine the item type, and thus which field this
514 union contains. */
Al Viro3e8962b2005-05-01 08:59:18 -0700515 __le16 ih_free_space_reserved;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 /* Iff this is a directory item, this field equals the
517 number of directory entries in the directory item. */
Al Viro3e8962b2005-05-01 08:59:18 -0700518 __le16 ih_entry_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 } __attribute__ ((__packed__)) u;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700520 __le16 ih_item_len; /* total size of the item body */
521 __le16 ih_item_location; /* an offset to the item body
522 * within the block */
523 __le16 ih_version; /* 0 for all old items, 2 for new
524 ones. Highest bit is set by fsck
525 temporary, cleaned after all
526 done */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527} __attribute__ ((__packed__));
528/* size of item header */
529#define IH_SIZE (sizeof(struct item_head))
530
531#define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved)
532#define ih_version(ih) le16_to_cpu((ih)->ih_version)
533#define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count)
534#define ih_location(ih) le16_to_cpu((ih)->ih_item_location)
535#define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len)
536
537#define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
538#define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0)
539#define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
540#define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
541#define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
542
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543#define unreachable_item(ih) (ih_version(ih) & (1 << 15))
544
545#define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
546#define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
547
548/* these operate on indirect items, where you've got an array of ints
549** at a possibly unaligned location. These are a noop on ia32
550**
551** p is the array of __u32, i is the index into the array, v is the value
552** to store there.
553*/
Harvey Harrison8b5ac312008-07-25 01:45:26 -0700554#define get_block_num(p, i) get_unaligned_le32((p) + (i))
555#define put_block_num(p, i, v) put_unaligned_le32((v), (p) + (i))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556
557//
558// in old version uniqueness field shows key type
559//
560#define V1_SD_UNIQUENESS 0
561#define V1_INDIRECT_UNIQUENESS 0xfffffffe
562#define V1_DIRECT_UNIQUENESS 0xffffffff
563#define V1_DIRENTRY_UNIQUENESS 500
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700564#define V1_ANY_UNIQUENESS 555 // FIXME: comment is required
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565
566//
567// here are conversion routines
568//
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700569static inline int uniqueness2type(__u32 uniqueness) CONSTF;
570static inline int uniqueness2type(__u32 uniqueness)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700572 switch ((int)uniqueness) {
573 case V1_SD_UNIQUENESS:
574 return TYPE_STAT_DATA;
575 case V1_INDIRECT_UNIQUENESS:
576 return TYPE_INDIRECT;
577 case V1_DIRECT_UNIQUENESS:
578 return TYPE_DIRECT;
579 case V1_DIRENTRY_UNIQUENESS:
580 return TYPE_DIRENTRY;
581 default:
582 reiserfs_warning(NULL, "vs-500: unknown uniqueness %d",
583 uniqueness);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 case V1_ANY_UNIQUENESS:
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700585 return TYPE_ANY;
586 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587}
588
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700589static inline __u32 type2uniqueness(int type) CONSTF;
590static inline __u32 type2uniqueness(int type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700592 switch (type) {
593 case TYPE_STAT_DATA:
594 return V1_SD_UNIQUENESS;
595 case TYPE_INDIRECT:
596 return V1_INDIRECT_UNIQUENESS;
597 case TYPE_DIRECT:
598 return V1_DIRECT_UNIQUENESS;
599 case TYPE_DIRENTRY:
600 return V1_DIRENTRY_UNIQUENESS;
601 default:
602 reiserfs_warning(NULL, "vs-501: unknown type %d", type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 case TYPE_ANY:
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700604 return V1_ANY_UNIQUENESS;
605 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606}
607
608//
609// key is pointer to on disk key which is stored in le, result is cpu,
610// there is no way to get version of object from key, so, provide
611// version to these defines
612//
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700613static inline loff_t le_key_k_offset(int version,
614 const struct reiserfs_key *key)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700616 return (version == KEY_FORMAT_3_5) ?
617 le32_to_cpu(key->u.k_offset_v1.k_offset) :
618 offset_v2_k_offset(&(key->u.k_offset_v2));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619}
620
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700621static inline loff_t le_ih_k_offset(const struct item_head *ih)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700623 return le_key_k_offset(ih_version(ih), &(ih->ih_key));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624}
625
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700626static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700628 return (version == KEY_FORMAT_3_5) ?
629 uniqueness2type(le32_to_cpu(key->u.k_offset_v1.k_uniqueness)) :
630 offset_v2_k_type(&(key->u.k_offset_v2));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631}
632
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700633static inline loff_t le_ih_k_type(const struct item_head *ih)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700635 return le_key_k_type(ih_version(ih), &(ih->ih_key));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636}
637
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700638static inline void set_le_key_k_offset(int version, struct reiserfs_key *key,
639 loff_t offset)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700641 (version == KEY_FORMAT_3_5) ? (void)(key->u.k_offset_v1.k_offset = cpu_to_le32(offset)) : /* jdm check */
642 (void)(set_offset_v2_k_offset(&(key->u.k_offset_v2), offset));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643}
644
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700645static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700647 set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648}
649
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700650static inline void set_le_key_k_type(int version, struct reiserfs_key *key,
651 int type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700653 (version == KEY_FORMAT_3_5) ?
654 (void)(key->u.k_offset_v1.k_uniqueness =
655 cpu_to_le32(type2uniqueness(type)))
656 : (void)(set_offset_v2_k_type(&(key->u.k_offset_v2), type));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657}
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700658static inline void set_le_ih_k_type(struct item_head *ih, int type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700660 set_le_key_k_type(ih_version(ih), &(ih->ih_key), type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661}
662
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663#define is_direntry_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRENTRY)
664#define is_direct_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRECT)
665#define is_indirect_le_key(version,key) (le_key_k_type (version, key) == TYPE_INDIRECT)
666#define is_statdata_le_key(version,key) (le_key_k_type (version, key) == TYPE_STAT_DATA)
667
668//
669// item header has version.
670//
671#define is_direntry_le_ih(ih) is_direntry_le_key (ih_version (ih), &((ih)->ih_key))
672#define is_direct_le_ih(ih) is_direct_le_key (ih_version (ih), &((ih)->ih_key))
673#define is_indirect_le_ih(ih) is_indirect_le_key (ih_version(ih), &((ih)->ih_key))
674#define is_statdata_le_ih(ih) is_statdata_le_key (ih_version (ih), &((ih)->ih_key))
675
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676//
677// key is pointer to cpu key, result is cpu
678//
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700679static inline loff_t cpu_key_k_offset(const struct cpu_key *key)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700681 return key->on_disk_key.k_offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682}
683
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700684static inline loff_t cpu_key_k_type(const struct cpu_key *key)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700686 return key->on_disk_key.k_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687}
688
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700689static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690{
Al Viro6b9f5822005-05-01 08:59:19 -0700691 key->on_disk_key.k_offset = offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692}
693
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700694static inline void set_cpu_key_k_type(struct cpu_key *key, int type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695{
Al Viro6b9f5822005-05-01 08:59:19 -0700696 key->on_disk_key.k_type = type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697}
698
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700699static inline void cpu_key_k_offset_dec(struct cpu_key *key)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700701 key->on_disk_key.k_offset--;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702}
703
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704#define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
705#define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
706#define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
707#define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
708
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709/* are these used ? */
710#define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
711#define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
712#define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
713#define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
714
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715#define I_K_KEY_IN_ITEM(p_s_ih, p_s_key, n_blocksize) \
716 ( ! COMP_SHORT_KEYS(p_s_ih, p_s_key) && \
717 I_OFF_BYTE_IN_ITEM(p_s_ih, k_offset (p_s_key), n_blocksize) )
718
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700719/* maximal length of item */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720#define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
721#define MIN_ITEM_LEN 1
722
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723/* object identifier for root dir */
724#define REISERFS_ROOT_OBJECTID 2
725#define REISERFS_ROOT_PARENT_OBJECTID 1
Jaswinder Singh Rajput750e1c12009-02-03 19:40:03 +0530726
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727extern struct reiserfs_key root_key;
728
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729/*
730 * Picture represents a leaf of the S+tree
731 * ______________________________________________________
732 * | | Array of | | |
733 * |Block | Object-Item | F r e e | Objects- |
734 * | head | Headers | S p a c e | Items |
735 * |______|_______________|___________________|___________|
736 */
737
738/* Header of a disk block. More precisely, header of a formatted leaf
739 or internal node, and not the header of an unformatted node. */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700740struct block_head {
741 __le16 blk_level; /* Level of a block in the tree. */
742 __le16 blk_nr_item; /* Number of keys/items in a block. */
743 __le16 blk_free_space; /* Block free space in bytes. */
744 __le16 blk_reserved;
745 /* dump this in v4/planA */
746 struct reiserfs_key blk_right_delim_key; /* kept only for compatibility */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747};
748
749#define BLKH_SIZE (sizeof(struct block_head))
750#define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level))
751#define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item))
752#define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space))
753#define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved))
754#define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val))
755#define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val))
756#define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
757#define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
758#define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key)
759#define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val)
760
761/*
762 * values for blk_level field of the struct block_head
763 */
764
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700765#define FREE_LEVEL 0 /* when node gets removed from the tree its
766 blk_level is set to FREE_LEVEL. It is then
767 used to see whether the node is still in the
768 tree */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700770#define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771
772/* Given the buffer head of a formatted node, resolve to the block head of that node. */
773#define B_BLK_HEAD(p_s_bh) ((struct block_head *)((p_s_bh)->b_data))
774/* Number of items that are in buffer. */
775#define B_NR_ITEMS(p_s_bh) (blkh_nr_item(B_BLK_HEAD(p_s_bh)))
776#define B_LEVEL(p_s_bh) (blkh_level(B_BLK_HEAD(p_s_bh)))
777#define B_FREE_SPACE(p_s_bh) (blkh_free_space(B_BLK_HEAD(p_s_bh)))
778
779#define PUT_B_NR_ITEMS(p_s_bh,val) do { set_blkh_nr_item(B_BLK_HEAD(p_s_bh),val); } while (0)
780#define PUT_B_LEVEL(p_s_bh,val) do { set_blkh_level(B_BLK_HEAD(p_s_bh),val); } while (0)
781#define PUT_B_FREE_SPACE(p_s_bh,val) do { set_blkh_free_space(B_BLK_HEAD(p_s_bh),val); } while (0)
782
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783/* Get right delimiting key. -- little endian */
Mariusz Kozlowski5296c7b2006-12-06 20:40:10 -0800784#define B_PRIGHT_DELIM_KEY(p_s_bh) (&(blk_right_delim_key(B_BLK_HEAD(p_s_bh))))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785
786/* Does the buffer contain a disk leaf. */
787#define B_IS_ITEMS_LEVEL(p_s_bh) (B_LEVEL(p_s_bh) == DISK_LEAF_NODE_LEVEL)
788
789/* Does the buffer contain a disk internal node */
790#define B_IS_KEYS_LEVEL(p_s_bh) (B_LEVEL(p_s_bh) > DISK_LEAF_NODE_LEVEL \
791 && B_LEVEL(p_s_bh) <= MAX_HEIGHT)
792
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793/***************************************************************************/
794/* STAT DATA */
795/***************************************************************************/
796
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797//
798// old stat data is 32 bytes long. We are going to distinguish new one by
799// different size
800//
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700801struct stat_data_v1 {
802 __le16 sd_mode; /* file type, permissions */
803 __le16 sd_nlink; /* number of hard links */
804 __le16 sd_uid; /* owner */
805 __le16 sd_gid; /* group */
806 __le32 sd_size; /* file size */
807 __le32 sd_atime; /* time of last access */
808 __le32 sd_mtime; /* time file was last modified */
809 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
810 union {
811 __le32 sd_rdev;
812 __le32 sd_blocks; /* number of blocks file uses */
813 } __attribute__ ((__packed__)) u;
814 __le32 sd_first_direct_byte; /* first byte of file which is stored
815 in a direct item: except that if it
816 equals 1 it is a symlink and if it
817 equals ~(__u32)0 there is no
818 direct item. The existence of this
819 field really grates on me. Let's
820 replace it with a macro based on
821 sd_size and our tail suppression
822 policy. Someday. -Hans */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823} __attribute__ ((__packed__));
824
825#define SD_V1_SIZE (sizeof(struct stat_data_v1))
826#define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5)
827#define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
828#define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
829#define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink))
830#define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v))
831#define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid))
832#define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v))
833#define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid))
834#define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v))
835#define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size))
836#define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v))
837#define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
838#define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
839#define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
840#define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
841#define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
842#define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
843#define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
844#define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
845#define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks))
846#define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
847#define sd_v1_first_direct_byte(sdp) \
848 (le32_to_cpu((sdp)->sd_first_direct_byte))
849#define set_sd_v1_first_direct_byte(sdp,v) \
850 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
851
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852/* inode flags stored in sd_attrs (nee sd_reserved) */
853
854/* we want common flags to have the same values as in ext2,
855 so chattr(1) will work without problems */
David Howells36695672006-08-29 19:06:16 +0100856#define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL
857#define REISERFS_APPEND_FL FS_APPEND_FL
858#define REISERFS_SYNC_FL FS_SYNC_FL
859#define REISERFS_NOATIME_FL FS_NOATIME_FL
860#define REISERFS_NODUMP_FL FS_NODUMP_FL
861#define REISERFS_SECRM_FL FS_SECRM_FL
862#define REISERFS_UNRM_FL FS_UNRM_FL
863#define REISERFS_COMPR_FL FS_COMPR_FL
864#define REISERFS_NOTAIL_FL FS_NOTAIL_FL
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865
866/* persistent flags that file inherits from the parent directory */
867#define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \
868 REISERFS_SYNC_FL | \
869 REISERFS_NOATIME_FL | \
870 REISERFS_NODUMP_FL | \
871 REISERFS_SECRM_FL | \
872 REISERFS_COMPR_FL | \
873 REISERFS_NOTAIL_FL )
874
875/* Stat Data on disk (reiserfs version of UFS disk inode minus the
876 address blocks) */
877struct stat_data {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700878 __le16 sd_mode; /* file type, permissions */
879 __le16 sd_attrs; /* persistent inode flags */
880 __le32 sd_nlink; /* number of hard links */
881 __le64 sd_size; /* file size */
882 __le32 sd_uid; /* owner */
883 __le32 sd_gid; /* group */
884 __le32 sd_atime; /* time of last access */
885 __le32 sd_mtime; /* time file was last modified */
886 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
887 __le32 sd_blocks;
888 union {
889 __le32 sd_rdev;
890 __le32 sd_generation;
891 //__le32 sd_first_direct_byte;
892 /* first byte of file which is stored in a
893 direct item: except that if it equals 1
894 it is a symlink and if it equals
895 ~(__u32)0 there is no direct item. The
896 existence of this field really grates
897 on me. Let's replace it with a macro
898 based on sd_size and our tail
899 suppression policy? */
900 } __attribute__ ((__packed__)) u;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901} __attribute__ ((__packed__));
902//
903// this is 44 bytes long
904//
905#define SD_SIZE (sizeof(struct stat_data))
906#define SD_V2_SIZE SD_SIZE
907#define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6)
908#define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
909#define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
910/* sd_reserved */
911/* set_sd_reserved */
912#define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink))
913#define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v))
914#define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size))
915#define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v))
916#define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid))
917#define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v))
918#define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid))
919#define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v))
920#define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
921#define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
922#define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
923#define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
924#define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
925#define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
926#define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks))
927#define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
928#define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
929#define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
930#define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation))
931#define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
932#define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs))
933#define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v))
934
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935/***************************************************************************/
936/* DIRECTORY STRUCTURE */
937/***************************************************************************/
938/*
939 Picture represents the structure of directory items
940 ________________________________________________
941 | Array of | | | | | |
942 | directory |N-1| N-2 | .... | 1st |0th|
943 | entry headers | | | | | |
944 |_______________|___|_____|________|_______|___|
945 <---- directory entries ------>
946
947 First directory item has k_offset component 1. We store "." and ".."
948 in one item, always, we never split "." and ".." into differing
949 items. This makes, among other things, the code for removing
950 directories simpler. */
951#define SD_OFFSET 0
952#define SD_UNIQUENESS 0
953#define DOT_OFFSET 1
954#define DOT_DOT_OFFSET 2
955#define DIRENTRY_UNIQUENESS 500
956
957/* */
958#define FIRST_ITEM_OFFSET 1
959
960/*
961 Q: How to get key of object pointed to by entry from entry?
962
963 A: Each directory entry has its header. This header has deh_dir_id and deh_objectid fields, those are key
964 of object, entry points to */
965
966/* NOT IMPLEMENTED:
967 Directory will someday contain stat data of object */
968
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700969struct reiserfs_de_head {
970 __le32 deh_offset; /* third component of the directory entry key */
971 __le32 deh_dir_id; /* objectid of the parent directory of the object, that is referenced
972 by directory entry */
973 __le32 deh_objectid; /* objectid of the object, that is referenced by directory entry */
974 __le16 deh_location; /* offset of name in the whole item */
975 __le16 deh_state; /* whether 1) entry contains stat data (for future), and 2) whether
976 entry is hidden (unlinked) */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977} __attribute__ ((__packed__));
978#define DEH_SIZE sizeof(struct reiserfs_de_head)
979#define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset))
980#define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id))
981#define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid))
982#define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location))
983#define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state))
984
985#define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v)))
986#define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v)))
987#define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
988#define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
989#define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v)))
990
991/* empty directory contains two entries "." and ".." and their headers */
992#define EMPTY_DIR_SIZE \
993(DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
994
995/* old format directories have this size when empty */
996#define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
997
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700998#define DEH_Statdata 0 /* not used now */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999#define DEH_Visible 2
1000
1001/* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
1002#if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
1003# define ADDR_UNALIGNED_BITS (3)
1004#endif
1005
1006/* These are only used to manipulate deh_state.
1007 * Because of this, we'll use the ext2_ bit routines,
1008 * since they are little endian */
1009#ifdef ADDR_UNALIGNED_BITS
1010
1011# define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
1012# define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
1013
1014# define set_bit_unaligned(nr, addr) ext2_set_bit((nr) + unaligned_offset(addr), aligned_address(addr))
1015# define clear_bit_unaligned(nr, addr) ext2_clear_bit((nr) + unaligned_offset(addr), aligned_address(addr))
1016# define test_bit_unaligned(nr, addr) ext2_test_bit((nr) + unaligned_offset(addr), aligned_address(addr))
1017
1018#else
1019
1020# define set_bit_unaligned(nr, addr) ext2_set_bit(nr, addr)
1021# define clear_bit_unaligned(nr, addr) ext2_clear_bit(nr, addr)
1022# define test_bit_unaligned(nr, addr) ext2_test_bit(nr, addr)
1023
1024#endif
1025
1026#define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1027#define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1028#define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1029#define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1030
1031#define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1032#define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1033#define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1034
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001035extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid,
1036 __le32 par_dirid, __le32 par_objid);
1037extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid,
1038 __le32 par_dirid, __le32 par_objid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039
1040/* array of the entry headers */
1041 /* get item body */
1042#define B_I_PITEM(bh,ih) ( (bh)->b_data + ih_location(ih) )
1043#define B_I_DEH(bh,ih) ((struct reiserfs_de_head *)(B_I_PITEM(bh,ih)))
1044
1045/* length of the directory entry in directory item. This define
1046 calculates length of i-th directory entry using directory entry
1047 locations from dir entry head. When it calculates length of 0-th
1048 directory entry, it uses length of whole item in place of entry
1049 location of the non-existent following entry in the calculation.
1050 See picture above.*/
1051/*
1052#define I_DEH_N_ENTRY_LENGTH(ih,deh,i) \
1053((i) ? (deh_location((deh)-1) - deh_location((deh))) : (ih_item_len((ih)) - deh_location((deh))))
1054*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001055static inline int entry_length(const struct buffer_head *bh,
1056 const struct item_head *ih, int pos_in_item)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001058 struct reiserfs_de_head *deh;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001060 deh = B_I_DEH(bh, ih) + pos_in_item;
1061 if (pos_in_item)
1062 return deh_location(deh - 1) - deh_location(deh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001064 return ih_item_len(ih) - deh_location(deh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065}
1066
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067/* number of entries in the directory item, depends on ENTRY_COUNT being at the start of directory dynamic data. */
1068#define I_ENTRY_COUNT(ih) (ih_entry_count((ih)))
1069
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070/* name by bh, ih and entry_num */
1071#define B_I_E_NAME(bh,ih,entry_num) ((char *)(bh->b_data + ih_location(ih) + deh_location(B_I_DEH(bh,ih)+(entry_num))))
1072
1073// two entries per block (at least)
1074#define REISERFS_MAX_NAME(block_size) 255
1075
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076/* this structure is used for operations on directory entries. It is
1077 not a disk structure. */
1078/* When reiserfs_find_entry or search_by_entry_key find directory
1079 entry, they return filled reiserfs_dir_entry structure */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001080struct reiserfs_dir_entry {
1081 struct buffer_head *de_bh;
1082 int de_item_num;
1083 struct item_head *de_ih;
1084 int de_entry_num;
1085 struct reiserfs_de_head *de_deh;
1086 int de_entrylen;
1087 int de_namelen;
1088 char *de_name;
Jeff Mahoney3af1efe2006-03-02 13:25:26 -05001089 unsigned long *de_gen_number_bit_string;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001091 __u32 de_dir_id;
1092 __u32 de_objectid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001094 struct cpu_key de_entry_key;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095};
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001096
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097/* these defines are useful when a particular member of a reiserfs_dir_entry is needed */
1098
1099/* pointer to file name, stored in entry */
1100#define B_I_DEH_ENTRY_FILE_NAME(bh,ih,deh) (B_I_PITEM (bh, ih) + deh_location(deh))
1101
1102/* length of name */
1103#define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
1104(I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
1105
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106/* hash value occupies bits from 7 up to 30 */
1107#define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
1108/* generation number occupies 7 bits starting from 0 up to 6 */
1109#define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
1110#define MAX_GENERATION_NUMBER 127
1111
1112#define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
1113
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114/*
1115 * Picture represents an internal node of the reiserfs tree
1116 * ______________________________________________________
1117 * | | Array of | Array of | Free |
1118 * |block | keys | pointers | space |
1119 * | head | N | N+1 | |
1120 * |______|_______________|___________________|___________|
1121 */
1122
1123/***************************************************************************/
1124/* DISK CHILD */
1125/***************************************************************************/
1126/* Disk child pointer: The pointer from an internal node of the tree
1127 to a node that is on disk. */
1128struct disk_child {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001129 __le32 dc_block_number; /* Disk child's block number. */
1130 __le16 dc_size; /* Disk child's used space. */
1131 __le16 dc_reserved;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132};
1133
1134#define DC_SIZE (sizeof(struct disk_child))
1135#define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number))
1136#define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size))
1137#define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
1138#define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
1139
1140/* Get disk child by buffer header and position in the tree node. */
1141#define B_N_CHILD(p_s_bh,n_pos) ((struct disk_child *)\
1142((p_s_bh)->b_data+BLKH_SIZE+B_NR_ITEMS(p_s_bh)*KEY_SIZE+DC_SIZE*(n_pos)))
1143
1144/* Get disk child number by buffer header and position in the tree node. */
1145#define B_N_CHILD_NUM(p_s_bh,n_pos) (dc_block_number(B_N_CHILD(p_s_bh,n_pos)))
1146#define PUT_B_N_CHILD_NUM(p_s_bh,n_pos, val) (put_dc_block_number(B_N_CHILD(p_s_bh,n_pos), val ))
1147
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001148 /* maximal value of field child_size in structure disk_child */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 /* child size is the combined size of all items and their headers */
1150#define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
1151
1152/* amount of used space in buffer (not including block head) */
1153#define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
1154
1155/* max and min number of keys in internal node */
1156#define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
1157#define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2)
1158
1159/***************************************************************************/
1160/* PATH STRUCTURES AND DEFINES */
1161/***************************************************************************/
1162
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163/* Search_by_key fills up the path from the root to the leaf as it descends the tree looking for the
1164 key. It uses reiserfs_bread to try to find buffers in the cache given their block number. If it
1165 does not find them in the cache it reads them from disk. For each node search_by_key finds using
1166 reiserfs_bread it then uses bin_search to look through that node. bin_search will find the
1167 position of the block_number of the next node if it is looking through an internal node. If it
1168 is looking through a leaf node bin_search will find the position of the item which has key either
1169 equal to given key, or which is the maximal key less than the given key. */
1170
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001171struct path_element {
1172 struct buffer_head *pe_buffer; /* Pointer to the buffer at the path in the tree. */
1173 int pe_position; /* Position in the tree node which is placed in the */
1174 /* buffer above. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175};
1176
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001177#define MAX_HEIGHT 5 /* maximal height of a tree. don't change this without changing JOURNAL_PER_BALANCE_CNT */
1178#define EXTENDED_MAX_HEIGHT 7 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
1179#define FIRST_PATH_ELEMENT_OFFSET 2 /* Must be equal to at least 2. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001181#define ILLEGAL_PATH_ELEMENT_OFFSET 1 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
1182#define MAX_FEB_SIZE 6 /* this MUST be MAX_HEIGHT + 1. See about FEB below */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183
1184/* We need to keep track of who the ancestors of nodes are. When we
1185 perform a search we record which nodes were visited while
1186 descending the tree looking for the node we searched for. This list
1187 of nodes is called the path. This information is used while
1188 performing balancing. Note that this path information may become
1189 invalid, and this means we must check it when using it to see if it
1190 is still valid. You'll need to read search_by_key and the comments
1191 in it, especially about decrement_counters_in_path(), to understand
1192 this structure.
1193
1194Paths make the code so much harder to work with and debug.... An
1195enormous number of bugs are due to them, and trying to write or modify
1196code that uses them just makes my head hurt. They are based on an
1197excessive effort to avoid disturbing the precious VFS code.:-( The
1198gods only know how we are going to SMP the code that uses them.
1199znodes are the way! */
1200
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001201#define PATH_READA 0x1 /* do read ahead */
1202#define PATH_READA_BACK 0x2 /* read backwards */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001204struct treepath {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001205 int path_length; /* Length of the array above. */
1206 int reada;
1207 struct path_element path_elements[EXTENDED_MAX_HEIGHT]; /* Array of the path elements. */
1208 int pos_in_item;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209};
1210
1211#define pos_in_item(path) ((path)->pos_in_item)
1212
1213#define INITIALIZE_PATH(var) \
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001214struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215
1216/* Get path element by path and path position. */
1217#define PATH_OFFSET_PELEMENT(p_s_path,n_offset) ((p_s_path)->path_elements +(n_offset))
1218
1219/* Get buffer header at the path by path and path position. */
1220#define PATH_OFFSET_PBUFFER(p_s_path,n_offset) (PATH_OFFSET_PELEMENT(p_s_path,n_offset)->pe_buffer)
1221
1222/* Get position in the element at the path by path and path position. */
1223#define PATH_OFFSET_POSITION(p_s_path,n_offset) (PATH_OFFSET_PELEMENT(p_s_path,n_offset)->pe_position)
1224
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225#define PATH_PLAST_BUFFER(p_s_path) (PATH_OFFSET_PBUFFER((p_s_path), (p_s_path)->path_length))
1226 /* you know, to the person who didn't
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001227 write this the macro name does not
1228 at first suggest what it does.
1229 Maybe POSITION_FROM_PATH_END? Or
1230 maybe we should just focus on
1231 dumping paths... -Hans */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232#define PATH_LAST_POSITION(p_s_path) (PATH_OFFSET_POSITION((p_s_path), (p_s_path)->path_length))
1233
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234#define PATH_PITEM_HEAD(p_s_path) B_N_PITEM_HEAD(PATH_PLAST_BUFFER(p_s_path),PATH_LAST_POSITION(p_s_path))
1235
1236/* in do_balance leaf has h == 0 in contrast with path structure,
1237 where root has level == 0. That is why we need these defines */
1238#define PATH_H_PBUFFER(p_s_path, h) PATH_OFFSET_PBUFFER (p_s_path, p_s_path->path_length - (h)) /* tb->S[h] */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001239#define PATH_H_PPARENT(path, h) PATH_H_PBUFFER (path, (h) + 1) /* tb->F[h] or tb->S[0]->b_parent */
1240#define PATH_H_POSITION(path, h) PATH_OFFSET_POSITION (path, path->path_length - (h))
1241#define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1) /* tb->S[h]->b_item_order */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242
1243#define PATH_H_PATH_OFFSET(p_s_path, n_h) ((p_s_path)->path_length - (n_h))
1244
1245#define get_last_bh(path) PATH_PLAST_BUFFER(path)
1246#define get_ih(path) PATH_PITEM_HEAD(path)
1247#define get_item_pos(path) PATH_LAST_POSITION(path)
1248#define get_item(path) ((void *)B_N_PITEM(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION (path)))
1249#define item_moved(ih,path) comp_items(ih, path)
1250#define path_changed(ih,path) comp_items (ih, path)
1251
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252/***************************************************************************/
1253/* MISC */
1254/***************************************************************************/
1255
1256/* Size of pointer to the unformatted node. */
1257#define UNFM_P_SIZE (sizeof(unp_t))
1258#define UNFM_P_SHIFT 2
1259
1260// in in-core inode key is stored on le form
1261#define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
1262
1263#define MAX_UL_INT 0xffffffff
1264#define MAX_INT 0x7ffffff
1265#define MAX_US_INT 0xffff
1266
1267// reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
1268#define U32_MAX (~(__u32)0)
1269
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001270static inline loff_t max_reiserfs_offset(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001272 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
1273 return (loff_t) U32_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001275 return (loff_t) ((~(__u64) 0) >> 4);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276}
1277
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278/*#define MAX_KEY_UNIQUENESS MAX_UL_INT*/
1279#define MAX_KEY_OBJECTID MAX_UL_INT
1280
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281#define MAX_B_NUM MAX_UL_INT
1282#define MAX_FC_NUM MAX_US_INT
1283
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284/* the purpose is to detect overflow of an unsigned short */
1285#define REISERFS_LINK_MAX (MAX_US_INT - 1000)
1286
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287/* The following defines are used in reiserfs_insert_item and reiserfs_append_item */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001288#define REISERFS_KERNEL_MEM 0 /* reiserfs kernel memory mode */
1289#define REISERFS_USER_MEM 1 /* reiserfs user memory mode */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290
1291#define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
1292#define get_generation(s) atomic_read (&fs_generation(s))
1293#define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen)
1294#define __fs_changed(gen,s) (gen != get_generation (s))
1295#define fs_changed(gen,s) ({cond_resched(); __fs_changed(gen, s);})
1296
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297/***************************************************************************/
1298/* FIXATE NODES */
1299/***************************************************************************/
1300
1301#define VI_TYPE_LEFT_MERGEABLE 1
1302#define VI_TYPE_RIGHT_MERGEABLE 2
1303
1304/* To make any changes in the tree we always first find node, that
1305 contains item to be changed/deleted or place to insert a new
1306 item. We call this node S. To do balancing we need to decide what
1307 we will shift to left/right neighbor, or to a new node, where new
1308 item will be etc. To make this analysis simpler we build virtual
1309 node. Virtual node is an array of items, that will replace items of
1310 node S. (For instance if we are going to delete an item, virtual
1311 node does not contain it). Virtual node keeps information about
1312 item sizes and types, mergeability of first and last items, sizes
1313 of all entries in directory item. We use this array of items when
1314 calculating what we can shift to neighbors and how many nodes we
1315 have to have if we do not any shiftings, if we shift to left/right
1316 neighbor or to both. */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001317struct virtual_item {
1318 int vi_index; // index in the array of item operations
1319 unsigned short vi_type; // left/right mergeability
1320 unsigned short vi_item_len; /* length of item that it will have after balancing */
1321 struct item_head *vi_ih;
1322 const char *vi_item; // body of item (old or new)
1323 const void *vi_new_data; // 0 always but paste mode
1324 void *vi_uarea; // item specific area
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325};
1326
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001327struct virtual_node {
1328 char *vn_free_ptr; /* this is a pointer to the free space in the buffer */
1329 unsigned short vn_nr_item; /* number of items in virtual node */
1330 short vn_size; /* size of node , that node would have if it has unlimited size and no balancing is performed */
1331 short vn_mode; /* mode of balancing (paste, insert, delete, cut) */
1332 short vn_affected_item_num;
1333 short vn_pos_in_item;
1334 struct item_head *vn_ins_ih; /* item header of inserted item, 0 for other modes */
1335 const void *vn_data;
1336 struct virtual_item *vn_vi; /* array of items (including a new one, excluding item to be deleted) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337};
1338
1339/* used by directory items when creating virtual nodes */
1340struct direntry_uarea {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001341 int flags;
1342 __u16 entry_count;
1343 __u16 entry_sizes[1];
1344} __attribute__ ((__packed__));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345
1346/***************************************************************************/
1347/* TREE BALANCE */
1348/***************************************************************************/
1349
1350/* This temporary structure is used in tree balance algorithms, and
1351 constructed as we go to the extent that its various parts are
1352 needed. It contains arrays of nodes that can potentially be
1353 involved in the balancing of node S, and parameters that define how
1354 each of the nodes must be balanced. Note that in these algorithms
1355 for balancing the worst case is to need to balance the current node
1356 S and the left and right neighbors and all of their parents plus
1357 create a new node. We implement S1 balancing for the leaf nodes
1358 and S0 balancing for the internal nodes (S1 and S0 are defined in
1359 our papers.)*/
1360
1361#define MAX_FREE_BLOCK 7 /* size of the array of buffers to free at end of do_balance */
1362
1363/* maximum number of FEB blocknrs on a single level */
1364#define MAX_AMOUNT_NEEDED 2
1365
1366/* someday somebody will prefix every field in this struct with tb_ */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001367struct tree_balance {
1368 int tb_mode;
1369 int need_balance_dirty;
1370 struct super_block *tb_sb;
1371 struct reiserfs_transaction_handle *transaction_handle;
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001372 struct treepath *tb_path;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001373 struct buffer_head *L[MAX_HEIGHT]; /* array of left neighbors of nodes in the path */
1374 struct buffer_head *R[MAX_HEIGHT]; /* array of right neighbors of nodes in the path */
1375 struct buffer_head *FL[MAX_HEIGHT]; /* array of fathers of the left neighbors */
1376 struct buffer_head *FR[MAX_HEIGHT]; /* array of fathers of the right neighbors */
1377 struct buffer_head *CFL[MAX_HEIGHT]; /* array of common parents of center node and its left neighbor */
1378 struct buffer_head *CFR[MAX_HEIGHT]; /* array of common parents of center node and its right neighbor */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001380 struct buffer_head *FEB[MAX_FEB_SIZE]; /* array of empty buffers. Number of buffers in array equals
1381 cur_blknum. */
1382 struct buffer_head *used[MAX_FEB_SIZE];
1383 struct buffer_head *thrown[MAX_FEB_SIZE];
1384 int lnum[MAX_HEIGHT]; /* array of number of items which must be
1385 shifted to the left in order to balance the
1386 current node; for leaves includes item that
1387 will be partially shifted; for internal
1388 nodes, it is the number of child pointers
1389 rather than items. It includes the new item
1390 being created. The code sometimes subtracts
1391 one to get the number of wholly shifted
1392 items for other purposes. */
1393 int rnum[MAX_HEIGHT]; /* substitute right for left in comment above */
1394 int lkey[MAX_HEIGHT]; /* array indexed by height h mapping the key delimiting L[h] and
1395 S[h] to its item number within the node CFL[h] */
1396 int rkey[MAX_HEIGHT]; /* substitute r for l in comment above */
1397 int insert_size[MAX_HEIGHT]; /* the number of bytes by we are trying to add or remove from
1398 S[h]. A negative value means removing. */
1399 int blknum[MAX_HEIGHT]; /* number of nodes that will replace node S[h] after
1400 balancing on the level h of the tree. If 0 then S is
1401 being deleted, if 1 then S is remaining and no new nodes
1402 are being created, if 2 or 3 then 1 or 2 new nodes is
1403 being created */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001405 /* fields that are used only for balancing leaves of the tree */
1406 int cur_blknum; /* number of empty blocks having been already allocated */
1407 int s0num; /* number of items that fall into left most node when S[0] splits */
1408 int s1num; /* number of items that fall into first new node when S[0] splits */
1409 int s2num; /* number of items that fall into second new node when S[0] splits */
1410 int lbytes; /* number of bytes which can flow to the left neighbor from the left */
1411 /* most liquid item that cannot be shifted from S[0] entirely */
1412 /* if -1 then nothing will be partially shifted */
1413 int rbytes; /* number of bytes which will flow to the right neighbor from the right */
1414 /* most liquid item that cannot be shifted from S[0] entirely */
1415 /* if -1 then nothing will be partially shifted */
1416 int s1bytes; /* number of bytes which flow to the first new node when S[0] splits */
1417 /* note: if S[0] splits into 3 nodes, then items do not need to be cut */
1418 int s2bytes;
1419 struct buffer_head *buf_to_free[MAX_FREE_BLOCK]; /* buffers which are to be freed after do_balance finishes by unfix_nodes */
1420 char *vn_buf; /* kmalloced memory. Used to create
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 virtual node and keep map of
1422 dirtied bitmap blocks */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001423 int vn_buf_size; /* size of the vn_buf */
1424 struct virtual_node *tb_vn; /* VN starts after bitmap of bitmap blocks */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001426 int fs_gen; /* saved value of `reiserfs_generation' counter
1427 see FILESYSTEM_CHANGED() macro in reiserfs_fs.h */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428#ifdef DISPLACE_NEW_PACKING_LOCALITIES
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001429 struct in_core_key key; /* key pointer, to pass to block allocator or
1430 another low-level subsystem */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431#endif
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001432};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433
1434/* These are modes of balancing */
1435
1436/* When inserting an item. */
1437#define M_INSERT 'i'
1438/* When inserting into (directories only) or appending onto an already
1439 existant item. */
1440#define M_PASTE 'p'
1441/* When deleting an item. */
1442#define M_DELETE 'd'
1443/* When truncating an item or removing an entry from a (directory) item. */
1444#define M_CUT 'c'
1445
1446/* used when balancing on leaf level skipped (in reiserfsck) */
1447#define M_INTERNAL 'n'
1448
1449/* When further balancing is not needed, then do_balance does not need
1450 to be called. */
1451#define M_SKIP_BALANCING 's'
1452#define M_CONVERT 'v'
1453
1454/* modes of leaf_move_items */
1455#define LEAF_FROM_S_TO_L 0
1456#define LEAF_FROM_S_TO_R 1
1457#define LEAF_FROM_R_TO_L 2
1458#define LEAF_FROM_L_TO_R 3
1459#define LEAF_FROM_S_TO_SNEW 4
1460
1461#define FIRST_TO_LAST 0
1462#define LAST_TO_FIRST 1
1463
1464/* used in do_balance for passing parent of node information that has
1465 been gotten from tb struct */
1466struct buffer_info {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001467 struct tree_balance *tb;
1468 struct buffer_head *bi_bh;
1469 struct buffer_head *bi_parent;
1470 int bi_position;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471};
1472
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473/* there are 4 types of items: stat data, directory item, indirect, direct.
1474+-------------------+------------+--------------+------------+
1475| | k_offset | k_uniqueness | mergeable? |
1476+-------------------+------------+--------------+------------+
1477| stat data | 0 | 0 | no |
1478+-------------------+------------+--------------+------------+
1479| 1st directory item| DOT_OFFSET |DIRENTRY_UNIQUENESS| no |
1480| non 1st directory | hash value | | yes |
1481| item | | | |
1482+-------------------+------------+--------------+------------+
1483| indirect item | offset + 1 |TYPE_INDIRECT | if this is not the first indirect item of the object
1484+-------------------+------------+--------------+------------+
1485| direct item | offset + 1 |TYPE_DIRECT | if not this is not the first direct item of the object
1486+-------------------+------------+--------------+------------+
1487*/
1488
1489struct item_operations {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001490 int (*bytes_number) (struct item_head * ih, int block_size);
1491 void (*decrement_key) (struct cpu_key *);
1492 int (*is_left_mergeable) (struct reiserfs_key * ih,
1493 unsigned long bsize);
1494 void (*print_item) (struct item_head *, char *item);
1495 void (*check_item) (struct item_head *, char *item);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001497 int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
1498 int is_affected, int insert_size);
1499 int (*check_left) (struct virtual_item * vi, int free,
1500 int start_skip, int end_skip);
1501 int (*check_right) (struct virtual_item * vi, int free);
1502 int (*part_size) (struct virtual_item * vi, int from, int to);
1503 int (*unit_num) (struct virtual_item * vi);
1504 void (*print_vi) (struct virtual_item * vi);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505};
1506
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001507extern struct item_operations *item_ops[TYPE_ANY + 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508
1509#define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
1510#define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
1511#define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item)
1512#define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item)
1513#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)
1514#define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
1515#define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free)
1516#define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to)
1517#define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi)
1518#define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi)
1519
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520#define COMP_SHORT_KEYS comp_short_keys
1521
1522/* number of blocks pointed to by the indirect item */
1523#define I_UNFM_NUM(p_s_ih) ( ih_item_len(p_s_ih) / UNFM_P_SIZE )
1524
1525/* the used space within the unformatted node corresponding to pos within the item pointed to by ih */
1526#define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
1527
1528/* number of bytes contained by the direct item or the unformatted nodes the indirect item points to */
1529
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001530/* get the item header */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531#define B_N_PITEM_HEAD(bh,item_num) ( (struct item_head * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1532
1533/* get key */
1534#define B_N_PDELIM_KEY(bh,item_num) ( (struct reiserfs_key * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1535
1536/* get the key */
1537#define B_N_PKEY(bh,item_num) ( &(B_N_PITEM_HEAD(bh,item_num)->ih_key) )
1538
1539/* get item body */
1540#define B_N_PITEM(bh,item_num) ( (bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(item_num))))
1541
1542/* get the stat data by the buffer header and the item order */
1543#define B_N_STAT_DATA(bh,nr) \
1544( (struct stat_data *)((bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(nr))) ) )
1545
1546 /* following defines use reiserfs buffer header and item header */
1547
1548/* get stat-data */
1549#define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
1550
1551// this is 3976 for size==4096
1552#define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
1553
1554/* indirect items consist of entries which contain blocknrs, pos
1555 indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
1556 blocknr contained by the entry pos points to */
1557#define B_I_POS_UNFM_POINTER(bh,ih,pos) le32_to_cpu(*(((unp_t *)B_I_PITEM(bh,ih)) + (pos)))
1558#define PUT_B_I_POS_UNFM_POINTER(bh,ih,pos, val) do {*(((unp_t *)B_I_PITEM(bh,ih)) + (pos)) = cpu_to_le32(val); } while (0)
1559
1560struct reiserfs_iget_args {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001561 __u32 objectid;
1562 __u32 dirid;
1563};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564
1565/***************************************************************************/
1566/* FUNCTION DECLARATIONS */
1567/***************************************************************************/
1568
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569#define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
1570
1571#define journal_trans_half(blocksize) \
1572 ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
1573
1574/* journal.c see journal.c for all the comments here */
1575
1576/* first block written in a commit. */
1577struct reiserfs_journal_desc {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001578 __le32 j_trans_id; /* id of commit */
1579 __le32 j_len; /* length of commit. len +1 is the commit block */
1580 __le32 j_mount_id; /* mount id of this trans */
1581 __le32 j_realblock[1]; /* real locations for each block */
1582};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583
1584#define get_desc_trans_id(d) le32_to_cpu((d)->j_trans_id)
1585#define get_desc_trans_len(d) le32_to_cpu((d)->j_len)
1586#define get_desc_mount_id(d) le32_to_cpu((d)->j_mount_id)
1587
1588#define set_desc_trans_id(d,val) do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
1589#define set_desc_trans_len(d,val) do { (d)->j_len = cpu_to_le32 (val); } while (0)
1590#define set_desc_mount_id(d,val) do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
1591
1592/* last block written in a commit */
1593struct reiserfs_journal_commit {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001594 __le32 j_trans_id; /* must match j_trans_id from the desc block */
1595 __le32 j_len; /* ditto */
1596 __le32 j_realblock[1]; /* real locations for each block */
1597};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598
1599#define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
1600#define get_commit_trans_len(c) le32_to_cpu((c)->j_len)
1601#define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
1602
1603#define set_commit_trans_id(c,val) do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
1604#define set_commit_trans_len(c,val) do { (c)->j_len = cpu_to_le32 (val); } while (0)
1605
1606/* this header block gets written whenever a transaction is considered fully flushed, and is more recent than the
1607** last fully flushed transaction. fully flushed means all the log blocks and all the real blocks are on disk,
1608** and this transaction does not need to be replayed.
1609*/
1610struct reiserfs_journal_header {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001611 __le32 j_last_flush_trans_id; /* id of last fully flushed transaction */
1612 __le32 j_first_unflushed_offset; /* offset in the log of where to start replay after a crash */
1613 __le32 j_mount_id;
1614 /* 12 */ struct journal_params jh_journal;
1615};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616
1617/* biggest tunable defines are right here */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001618#define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */
1619#define JOURNAL_TRANS_MAX_DEFAULT 1024 /* biggest possible single transaction, don't change for now (8/3/99) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620#define JOURNAL_TRANS_MIN_DEFAULT 256
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001621#define JOURNAL_MAX_BATCH_DEFAULT 900 /* max blocks to batch into one transaction, don't make this any bigger than 900 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622#define JOURNAL_MIN_RATIO 2
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001623#define JOURNAL_MAX_COMMIT_AGE 30
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624#define JOURNAL_MAX_TRANS_AGE 30
1625#define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
1626#ifdef CONFIG_QUOTA
Jan Kara556a2a42005-06-23 22:01:06 -07001627/* We need to update data and inode (atime) */
1628#define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? 2 : 0)
1629/* 1 balancing, 1 bitmap, 1 data per write + stat data update */
1630#define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? \
1631(DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0)
1632/* same as with INIT */
1633#define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? \
1634(DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635#else
Jan Kara556a2a42005-06-23 22:01:06 -07001636#define REISERFS_QUOTA_TRANS_BLOCKS(s) 0
1637#define REISERFS_QUOTA_INIT_BLOCKS(s) 0
1638#define REISERFS_QUOTA_DEL_BLOCKS(s) 0
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639#endif
1640
1641/* both of these can be as low as 1, or as high as you want. The min is the
1642** number of 4k bitmap nodes preallocated on mount. New nodes are allocated
1643** as needed, and released when transactions are committed. On release, if
1644** the current number of nodes is > max, the node is freed, otherwise,
1645** it is put on a free list for faster use later.
1646*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001647#define REISERFS_MIN_BITMAP_NODES 10
1648#define REISERFS_MAX_BITMAP_NODES 100
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001650#define JBH_HASH_SHIFT 13 /* these are based on journal hash size of 8192 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651#define JBH_HASH_MASK 8191
1652
1653#define _jhashfn(sb,block) \
1654 (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
1655 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
1656#define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
1657
1658// We need these to make journal.c code more readable
1659#define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1660#define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1661#define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1662
1663enum reiserfs_bh_state_bits {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001664 BH_JDirty = BH_PrivateStart, /* buffer is in current transaction */
1665 BH_JDirty_wait,
1666 BH_JNew, /* disk block was taken off free list before
1667 * being in a finished transaction, or
1668 * written to disk. Can be reused immed. */
1669 BH_JPrepared,
1670 BH_JRestore_dirty,
1671 BH_JTest, // debugging only will go away
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672};
1673
1674BUFFER_FNS(JDirty, journaled);
1675TAS_BUFFER_FNS(JDirty, journaled);
1676BUFFER_FNS(JDirty_wait, journal_dirty);
1677TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
1678BUFFER_FNS(JNew, journal_new);
1679TAS_BUFFER_FNS(JNew, journal_new);
1680BUFFER_FNS(JPrepared, journal_prepared);
1681TAS_BUFFER_FNS(JPrepared, journal_prepared);
1682BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1683TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1684BUFFER_FNS(JTest, journal_test);
1685TAS_BUFFER_FNS(JTest, journal_test);
1686
1687/*
1688** transaction handle which is passed around for all journal calls
1689*/
1690struct reiserfs_transaction_handle {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001691 struct super_block *t_super; /* super for this FS when journal_begin was
1692 called. saves calls to reiserfs_get_super
1693 also used by nested transactions to make
1694 sure they are nesting on the right FS
1695 _must_ be first in the handle
1696 */
1697 int t_refcount;
1698 int t_blocks_logged; /* number of blocks this writer has logged */
1699 int t_blocks_allocated; /* number of blocks this writer allocated */
1700 unsigned long t_trans_id; /* sanity check, equals the current trans id */
1701 void *t_handle_save; /* save existing current->journal_info */
1702 unsigned displace_new_blocks:1; /* if new block allocation occurres, that block
1703 should be displaced from others */
1704 struct list_head t_list;
1705};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706
1707/* used to keep track of ordered and tail writes, attached to the buffer
1708 * head through b_journal_head.
1709 */
1710struct reiserfs_jh {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001711 struct reiserfs_journal_list *jl;
1712 struct buffer_head *bh;
1713 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714};
1715
1716void reiserfs_free_jh(struct buffer_head *bh);
1717int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
1718int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001719int journal_mark_dirty(struct reiserfs_transaction_handle *,
1720 struct super_block *, struct buffer_head *bh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001722static inline int reiserfs_file_data_log(struct inode *inode)
1723{
1724 if (reiserfs_data_log(inode->i_sb) ||
1725 (REISERFS_I(inode)->i_flags & i_data_log))
1726 return 1;
1727 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728}
1729
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001730static inline int reiserfs_transaction_running(struct super_block *s)
1731{
1732 struct reiserfs_transaction_handle *th = current->journal_info;
1733 if (th && th->t_super == s)
1734 return 1;
1735 if (th && th->t_super == NULL)
1736 BUG();
1737 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738}
1739
Alexander Zarochentzev23f9e0f2006-03-25 03:06:57 -08001740static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th)
1741{
1742 return th->t_blocks_allocated - th->t_blocks_logged;
1743}
1744
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001745struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
1746 super_block
1747 *,
1748 int count);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
1750int reiserfs_commit_page(struct inode *inode, struct page *page,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001751 unsigned from, unsigned to);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752int reiserfs_flush_old_commits(struct super_block *);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001753int reiserfs_commit_for_inode(struct inode *);
1754int reiserfs_inode_needs_commit(struct inode *);
1755void reiserfs_update_inode_transaction(struct inode *);
1756void reiserfs_wait_on_write_block(struct super_block *s);
1757void reiserfs_block_writes(struct reiserfs_transaction_handle *th);
1758void reiserfs_allow_writes(struct super_block *s);
1759void reiserfs_check_lock_depth(struct super_block *s, char *caller);
1760int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh,
1761 int wait);
1762void reiserfs_restore_prepared_buffer(struct super_block *,
1763 struct buffer_head *bh);
1764int journal_init(struct super_block *, const char *j_dev_name, int old_format,
1765 unsigned int);
1766int journal_release(struct reiserfs_transaction_handle *, struct super_block *);
1767int journal_release_error(struct reiserfs_transaction_handle *,
1768 struct super_block *);
1769int journal_end(struct reiserfs_transaction_handle *, struct super_block *,
1770 unsigned long);
1771int journal_end_sync(struct reiserfs_transaction_handle *, struct super_block *,
1772 unsigned long);
1773int journal_mark_freed(struct reiserfs_transaction_handle *,
1774 struct super_block *, b_blocknr_t blocknr);
1775int journal_transaction_should_end(struct reiserfs_transaction_handle *, int);
Jeff Mahoney3ee16672007-10-18 23:39:25 -07001776int reiserfs_in_journal(struct super_block *p_s_sb, unsigned int bmap_nr,
1777 int bit_nr, int searchall, b_blocknr_t *next);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001778int journal_begin(struct reiserfs_transaction_handle *,
1779 struct super_block *p_s_sb, unsigned long);
1780int journal_join_abort(struct reiserfs_transaction_handle *,
1781 struct super_block *p_s_sb, unsigned long);
1782void reiserfs_journal_abort(struct super_block *sb, int errno);
1783void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...);
1784int reiserfs_allocate_list_bitmaps(struct super_block *s,
Jeff Mahoney3ee16672007-10-18 23:39:25 -07001785 struct reiserfs_list_bitmap *, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001787void add_save_link(struct reiserfs_transaction_handle *th,
1788 struct inode *inode, int truncate);
1789int remove_save_link(struct inode *inode, int truncate);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790
1791/* objectid.c */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001792__u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th);
1793void reiserfs_release_objectid(struct reiserfs_transaction_handle *th,
1794 __u32 objectid_to_release);
1795int reiserfs_convert_objectid_map_v1(struct super_block *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796
1797/* stree.c */
1798int B_IS_IN_TREE(const struct buffer_head *);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001799extern void copy_item_head(struct item_head *p_v_to,
1800 const struct item_head *p_v_from);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801
1802// first key is in cpu form, second - le
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001803extern int comp_short_keys(const struct reiserfs_key *le_key,
1804 const struct cpu_key *cpu_key);
1805extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806
1807// both are in le form
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001808extern int comp_le_keys(const struct reiserfs_key *,
1809 const struct reiserfs_key *);
1810extern int comp_short_le_keys(const struct reiserfs_key *,
1811 const struct reiserfs_key *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812
1813//
1814// get key version from on disk key - kludge
1815//
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001816static inline int le_key_version(const struct reiserfs_key *key)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001818 int type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001820 type = offset_v2_k_type(&(key->u.k_offset_v2));
1821 if (type != TYPE_DIRECT && type != TYPE_INDIRECT
1822 && type != TYPE_DIRENTRY)
1823 return KEY_FORMAT_3_5;
1824
1825 return KEY_FORMAT_3_6;
1826
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827}
1828
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001829static inline void copy_key(struct reiserfs_key *to,
1830 const struct reiserfs_key *from)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001832 memcpy(to, from, KEY_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833}
1834
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001835int comp_items(const struct item_head *stored_ih, const struct treepath *p_s_path);
1836const struct reiserfs_key *get_rkey(const struct treepath *p_s_chk_path,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001837 const struct super_block *p_s_sb);
1838int search_by_key(struct super_block *, const struct cpu_key *,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001839 struct treepath *, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840#define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001841int search_for_position_by_key(struct super_block *p_s_sb,
1842 const struct cpu_key *p_s_cpu_key,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001843 struct treepath *p_s_search_path);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001844extern void decrement_bcount(struct buffer_head *p_s_bh);
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001845void decrement_counters_in_path(struct treepath *p_s_search_path);
1846void pathrelse(struct treepath *p_s_search_path);
1847int reiserfs_check_path(struct treepath *p);
1848void pathrelse_and_restore(struct super_block *s, struct treepath *p_s_search_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001850int reiserfs_insert_item(struct reiserfs_transaction_handle *th,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001851 struct treepath *path,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001852 const struct cpu_key *key,
1853 struct item_head *ih,
1854 struct inode *inode, const char *body);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001856int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001857 struct treepath *path,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001858 const struct cpu_key *key,
1859 struct inode *inode,
1860 const char *body, int paste_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001862int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001863 struct treepath *path,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001864 struct cpu_key *key,
1865 struct inode *inode,
1866 struct page *page, loff_t new_file_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001868int reiserfs_delete_item(struct reiserfs_transaction_handle *th,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001869 struct treepath *path,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001870 const struct cpu_key *key,
1871 struct inode *inode, struct buffer_head *p_s_un_bh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001873void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
1874 struct inode *inode, struct reiserfs_key *key);
1875int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
1876 struct inode *p_s_inode);
1877int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,
1878 struct inode *p_s_inode, struct page *,
1879 int update_timestamps);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880
1881#define i_block_size(inode) ((inode)->i_sb->s_blocksize)
1882#define file_size(inode) ((inode)->i_size)
1883#define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
1884
1885#define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
1886!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 )
1887
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001888void padd_item(char *item, int total_length, int length);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889
1890/* inode.c */
1891/* args for the create parameter of reiserfs_get_block */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001892#define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */
1893#define GET_BLOCK_CREATE 1 /* add anything you need to find block */
1894#define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */
1895#define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08001896#define GET_BLOCK_NO_IMUX 8 /* i_mutex is not held, don't preallocate */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001897#define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001899void reiserfs_read_locked_inode(struct inode *inode,
1900 struct reiserfs_iget_args *args);
1901int reiserfs_find_actor(struct inode *inode, void *p);
1902int reiserfs_init_locked_inode(struct inode *inode, void *p);
1903void reiserfs_delete_inode(struct inode *inode);
1904int reiserfs_write_inode(struct inode *inode, int);
1905int reiserfs_get_block(struct inode *inode, sector_t block,
1906 struct buffer_head *bh_result, int create);
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001907struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1908 int fh_len, int fh_type);
1909struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
1910 int fh_len, int fh_type);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001911int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
1912 int connectable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001914int reiserfs_truncate_file(struct inode *, int update_timestamps);
1915void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset,
1916 int type, int key_length);
1917void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
1918 int version,
1919 loff_t offset, int type, int length, int entry_count);
1920struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001922int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1923 struct inode *dir, int mode,
1924 const char *symname, loff_t i_size,
1925 struct dentry *dentry, struct inode *inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001927void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1928 struct inode *inode, loff_t size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929
1930static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001931 struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001933 reiserfs_update_sd_size(th, inode, inode->i_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934}
1935
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001936void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode);
1937void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
1939
1940/* namei.c */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001941void set_de_name_and_namelen(struct reiserfs_dir_entry *de);
1942int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001943 struct treepath *path, struct reiserfs_dir_entry *de);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001944struct dentry *reiserfs_get_parent(struct dentry *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945/* procfs.c */
1946
1947#if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )
1948#define REISERFS_PROC_INFO
1949#else
1950#undef REISERFS_PROC_INFO
1951#endif
1952
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001953int reiserfs_proc_info_init(struct super_block *sb);
1954int reiserfs_proc_info_done(struct super_block *sb);
1955struct proc_dir_entry *reiserfs_proc_register_global(char *name,
1956 read_proc_t * func);
1957void reiserfs_proc_unregister_global(const char *name);
1958int reiserfs_proc_info_global_init(void);
1959int reiserfs_proc_info_global_done(void);
1960int reiserfs_global_version_in_proc(char *buffer, char **start, off_t offset,
1961 int count, int *eof, void *data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001962
1963#if defined( REISERFS_PROC_INFO )
1964
1965#define PROC_EXP( e ) e
1966
1967#define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
1968#define PROC_INFO_MAX( sb, field, value ) \
1969 __PINFO( sb ).field = \
1970 max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
1971#define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
1972#define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
1973#define PROC_INFO_BH_STAT( sb, bh, level ) \
1974 PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \
1975 PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \
1976 PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
1977#else
1978#define PROC_EXP( e )
1979#define VOID_V ( ( void ) 0 )
1980#define PROC_INFO_MAX( sb, field, value ) VOID_V
1981#define PROC_INFO_INC( sb, field ) VOID_V
1982#define PROC_INFO_ADD( sb, field, val ) VOID_V
1983#define PROC_INFO_BH_STAT( p_s_sb, p_s_bh, n_node_level ) VOID_V
1984#endif
1985
1986/* dir.c */
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08001987extern const struct inode_operations reiserfs_dir_inode_operations;
1988extern const struct inode_operations reiserfs_symlink_inode_operations;
1989extern const struct inode_operations reiserfs_special_inode_operations;
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -08001990extern const struct file_operations reiserfs_dir_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991
1992/* tail_conversion.c */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001993int direct2indirect(struct reiserfs_transaction_handle *, struct inode *,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001994 struct treepath *, struct buffer_head *, loff_t);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001995int indirect2direct(struct reiserfs_transaction_handle *, struct inode *,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001996 struct page *, struct treepath *, const struct cpu_key *,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001997 loff_t, char *);
1998void reiserfs_unmap_buffer(struct buffer_head *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999
2000/* file.c */
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002001extern const struct inode_operations reiserfs_file_inode_operations;
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -08002002extern const struct file_operations reiserfs_file_operations;
Christoph Hellwigf5e54d62006-06-28 04:26:44 -07002003extern const struct address_space_operations reiserfs_address_space_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004
2005/* fix_nodes.c */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002007int fix_nodes(int n_op_mode, struct tree_balance *p_s_tb,
2008 struct item_head *p_s_ins_ih, const void *);
2009void unfix_nodes(struct tree_balance *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002010
2011/* prints.c */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002012void reiserfs_panic(struct super_block *s, const char *fmt, ...)
2013 __attribute__ ((noreturn));
2014void reiserfs_info(struct super_block *s, const char *fmt, ...);
2015void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...);
2016void print_indirect_item(struct buffer_head *bh, int item_num);
2017void store_print_tb(struct tree_balance *tb);
2018void print_cur_tb(char *mes);
2019void print_de(struct reiserfs_dir_entry *de);
2020void print_bi(struct buffer_info *bi, char *mes);
2021#define PRINT_LEAF_ITEMS 1 /* print all items */
2022#define PRINT_DIRECTORY_ITEMS 2 /* print directory items */
2023#define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */
2024void print_block(struct buffer_head *bh, ...);
2025void print_bmap(struct super_block *s, int silent);
2026void print_bmap_block(int i, char *data, int size, int silent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027/*void print_super_block (struct super_block * s, char * mes);*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002028void print_objectid_map(struct super_block *s);
2029void print_block_head(struct buffer_head *bh, char *mes);
2030void check_leaf(struct buffer_head *bh);
2031void check_internal(struct buffer_head *bh);
2032void print_statistics(struct super_block *s);
2033char *reiserfs_hashname(int code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034
2035/* lbalance.c */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002036int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num,
2037 int mov_bytes, struct buffer_head *Snew);
2038int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes);
2039int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes);
2040void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first,
2041 int del_num, int del_bytes);
2042void leaf_insert_into_buf(struct buffer_info *bi, int before,
2043 struct item_head *inserted_item_ih,
2044 const char *inserted_item_body, int zeros_number);
2045void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num,
2046 int pos_in_item, int paste_size, const char *body,
2047 int zeros_number);
2048void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num,
2049 int pos_in_item, int cut_size);
2050void leaf_paste_entries(struct buffer_head *bh, int item_num, int before,
2051 int new_entry_count, struct reiserfs_de_head *new_dehs,
2052 const char *records, int paste_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053/* ibalance.c */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002054int balance_internal(struct tree_balance *, int, int, struct item_head *,
2055 struct buffer_head **);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056
2057/* do_balance.c */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002058void do_balance_mark_leaf_dirty(struct tree_balance *tb,
2059 struct buffer_head *bh, int flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002060#define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
2061#define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
2062
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002063void do_balance(struct tree_balance *tb, struct item_head *ih,
2064 const char *body, int flag);
2065void reiserfs_invalidate_buffer(struct tree_balance *tb,
2066 struct buffer_head *bh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002067
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002068int get_left_neighbor_position(struct tree_balance *tb, int h);
2069int get_right_neighbor_position(struct tree_balance *tb, int h);
2070void replace_key(struct tree_balance *tb, struct buffer_head *, int,
2071 struct buffer_head *, int);
2072void make_empty_node(struct buffer_info *);
2073struct buffer_head *get_FEB(struct tree_balance *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074
2075/* bitmap.c */
2076
2077/* structure contains hints for block allocator, and it is a container for
2078 * arguments, such as node, search path, transaction_handle, etc. */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002079struct __reiserfs_blocknr_hint {
2080 struct inode *inode; /* inode passed to allocator, if we allocate unf. nodes */
Jeff Mahoney3ee16672007-10-18 23:39:25 -07002081 sector_t block; /* file offset, in blocks */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002082 struct in_core_key key;
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08002083 struct treepath *path; /* search path, used by allocator to deternine search_start by
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002084 * various ways */
2085 struct reiserfs_transaction_handle *th; /* transaction handle is needed to log super blocks and
2086 * bitmap blocks changes */
2087 b_blocknr_t beg, end;
2088 b_blocknr_t search_start; /* a field used to transfer search start value (block number)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002089 * between different block allocator procedures
2090 * (determine_search_start() and others) */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002091 int prealloc_size; /* is set in determine_prealloc_size() function, used by underlayed
2092 * function that do actual allocation */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002094 unsigned formatted_node:1; /* the allocator uses different polices for getting disk space for
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095 * formatted/unformatted blocks with/without preallocation */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002096 unsigned preallocate:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097};
2098
2099typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
2100
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002101int reiserfs_parse_alloc_options(struct super_block *, char *);
2102void reiserfs_init_alloc_options(struct super_block *s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103
2104/*
2105 * given a directory, this will tell you what packing locality
2106 * to use for a new object underneat it. The locality is returned
2107 * in disk byte order (le).
2108 */
Al Viro3e8962b2005-05-01 08:59:18 -07002109__le32 reiserfs_choose_packing(struct inode *dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110
Jeff Mahoney6f010462006-09-30 23:28:43 -07002111int reiserfs_init_bitmap_cache(struct super_block *sb);
2112void reiserfs_free_bitmap_cache(struct super_block *sb);
2113void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info);
2114struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002115int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value);
2116void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *,
2117 b_blocknr_t, int for_unformatted);
2118int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int,
2119 int);
Adrian Bunk9adeb1b2005-09-10 00:27:18 -07002120static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002121 b_blocknr_t * new_blocknrs,
2122 int amount_needed)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002124 reiserfs_blocknr_hint_t hint = {
2125 .th = tb->transaction_handle,
2126 .path = tb->tb_path,
2127 .inode = NULL,
2128 .key = tb->key,
2129 .block = 0,
2130 .formatted_node = 1
2131 };
2132 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed,
2133 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134}
2135
Adrian Bunk9adeb1b2005-09-10 00:27:18 -07002136static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002137 *th, struct inode *inode,
2138 b_blocknr_t * new_blocknrs,
Jeff Mahoney3ee16672007-10-18 23:39:25 -07002139 struct treepath *path,
2140 sector_t block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002142 reiserfs_blocknr_hint_t hint = {
2143 .th = th,
2144 .path = path,
2145 .inode = inode,
2146 .block = block,
2147 .formatted_node = 0,
2148 .preallocate = 0
2149 };
2150 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151}
2152
2153#ifdef REISERFS_PREALLOCATE
Adrian Bunk9adeb1b2005-09-10 00:27:18 -07002154static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002155 *th, struct inode *inode,
2156 b_blocknr_t * new_blocknrs,
Jeff Mahoney3ee16672007-10-18 23:39:25 -07002157 struct treepath *path,
2158 sector_t block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002159{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002160 reiserfs_blocknr_hint_t hint = {
2161 .th = th,
2162 .path = path,
2163 .inode = inode,
2164 .block = block,
2165 .formatted_node = 0,
2166 .preallocate = 1
2167 };
2168 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002169}
2170
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002171void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
2172 struct inode *inode);
2173void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002174#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002175
2176/* hashes.c */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002177__u32 keyed_hash(const signed char *msg, int len);
2178__u32 yura_hash(const signed char *msg, int len);
2179__u32 r5_hash(const signed char *msg, int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002180
2181/* the ext2 bit routines adjust for big or little endian as
2182** appropriate for the arch, so in our laziness we use them rather
2183** than using the bit routines they call more directly. These
2184** routines must be used when changing on disk bitmaps. */
2185#define reiserfs_test_and_set_le_bit ext2_set_bit
2186#define reiserfs_test_and_clear_le_bit ext2_clear_bit
2187#define reiserfs_test_le_bit ext2_test_bit
2188#define reiserfs_find_next_zero_le_bit ext2_find_next_zero_bit
2189
2190/* sometimes reiserfs_truncate may require to allocate few new blocks
2191 to perform indirect2direct conversion. People probably used to
2192 think, that truncate should work without problems on a filesystem
2193 without free disk space. They may complain that they can not
2194 truncate due to lack of free disk space. This spare space allows us
2195 to not worry about it. 500 is probably too much, but it should be
2196 absolutely safe */
2197#define SPARE_SPACE 500
2198
Linus Torvalds1da177e2005-04-16 15:20:36 -07002199/* prototypes from ioctl.c */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002200int reiserfs_ioctl(struct inode *inode, struct file *filp,
2201 unsigned int cmd, unsigned long arg);
David Howells52b499c2006-08-29 19:06:18 +01002202long reiserfs_compat_ioctl(struct file *filp,
2203 unsigned int cmd, unsigned long arg);
Jan Karad5dee5c2008-04-28 02:16:23 -07002204int reiserfs_unpack(struct inode *inode, struct file *filp);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002205
Linus Torvalds1da177e2005-04-16 15:20:36 -07002206
Jaswinder Singh Rajput11d9f652009-02-02 21:45:41 +05302207#endif /* __KERNEL__ */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002208#endif /* _LINUX_REISER_FS_H */