2 * Copyright 1996-2002 Hans Reiser, see reiserfs/README for licensing and copyright details
5 /* this file has an amazingly stupid
6 name, yura please fix it to be
7 reiserfs.h, and merge all the rest
8 of our .h files that are in this
12 #ifndef _LINUX_REISER_FS_H
13 #define _LINUX_REISER_FS_H
15 #include <linux/types.h>
17 #include <linux/slab.h>
18 #include <linux/tqueue.h>
19 #include <asm/unaligned.h>
20 #include <linux/bitops.h>
21 #include <asm/hardirq.h>
22 #include <linux/proc_fs.h>
26 * include/linux/reiser_fs.h
28 * Reiser File System constants and structures
32 /* in reading the #defines, it may help to understand that they employ
33 the following abbreviations:
37 H = Height within the tree (should be changed to LEV)
38 N = Number of the item in the node
40 DEH = Directory Entry Header
45 UNFM = UNForMatted node
49 These #defines are named by concatenating these abbreviations,
50 where first comes the arguments, and last comes the return value,
55 #define USE_INODE_GENERATION_COUNTER
57 #define REISERFS_PREALLOCATE
58 #define DISPLACE_NEW_PACKING_LOCALITIES
59 #define PREALLOCATION_SIZE 9
61 /* n must be power of 2 */
62 #define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
64 // to be ok for alpha and others we have to align structures to 8 byte
66 // FIXME: do not change 4 by anything else: there is code which relies on that
67 #define ROUND_UP(x) _ROUND_UP(x,8LL)
69 /* debug levels. Right now, CONFIG_REISERFS_CHECK means print all debug
72 #define REISERFS_DEBUG_CODE 5 /* extra messages to help find/debug errors */
74 /* assertions handling */
76 /** always check a condition and panic if it's false. */
77 #define RASSERT( cond, format, args... ) \
79 reiserfs_panic( 0, "reiserfs[%i]: assertion " #cond " failed at " \
80 __FILE__ ":%i:%s: " format "\n", \
81 in_interrupt() ? -1 : current -> pid, __LINE__ , __FUNCTION__ , ##args )
83 #if defined( CONFIG_REISERFS_CHECK )
84 #define RFALSE( cond, format, args... ) RASSERT( !( cond ), format, ##args )
86 #define RFALSE( cond, format, args... ) do {;} while( 0 )
89 #define CONSTF __attribute__( ( const ) )
91 * Disk Data Structures
94 /***************************************************************************/
96 /***************************************************************************/
99 * Structure of super block on disk, a version of which in RAM is often accessed as s->u.reiserfs_sb.s_rs
100 * the version in RAM is part of a larger structure containing fields never written to disk.
104 #define REISERFS_SUPER_MAGIC 0x52654973
105 /* used by file system utilities that
106 look at the superblock, etc. */
107 #define REISERFS_SUPER_MAGIC_STRING "ReIsErFs"
108 #define REISER2FS_SUPER_MAGIC_STRING "ReIsEr2Fs"
110 extern char reiserfs_super_magic_string[];
111 extern char reiser2fs_super_magic_string[];
113 static inline int is_reiserfs_magic_string (const struct reiserfs_super_block * rs)
115 return (!strncmp (rs->s_magic, reiserfs_super_magic_string,
116 strlen ( reiserfs_super_magic_string)) ||
117 !strncmp (rs->s_magic, reiser2fs_super_magic_string,
118 strlen ( reiser2fs_super_magic_string)));
121 /* ReiserFS leaves the first 64k unused, so that partition labels have
122 enough space. If someone wants to write a fancy bootloader that
123 needs more than 64k, let us know, and this will be increased in size.
124 This number must be larger than than the largest block size on any
125 platform, or code will break. -Hans */
126 #define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
127 #define REISERFS_FIRST_BLOCK unused_define
129 /* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
130 #define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
132 // reiserfs internal error code (used by search_by_key adn fix_nodes))
134 #define REPEAT_SEARCH -1
136 #define NO_DISK_SPACE -3
137 #define NO_BALANCING_NEEDED (-4)
138 #define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
140 typedef unsigned long b_blocknr_t;
143 struct unfm_nodeinfo {
145 unsigned short unfm_freespace;
149 /* there are two formats of keys: 3.5 and 3.6
151 #define KEY_FORMAT_3_5 0
152 #define KEY_FORMAT_3_6 1
154 /* there are two stat datas */
155 #define STAT_DATA_V1 0
156 #define STAT_DATA_V2 1
158 /** this says about version of key of all items (but stat data) the
159 object consists of */
160 #define get_inode_item_key_version( inode ) \
161 (((inode)->u.reiserfs_i.i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
163 #define set_inode_item_key_version( inode, version ) \
164 ({ if((version)==KEY_FORMAT_3_6) \
165 (inode)->u.reiserfs_i.i_flags |= i_item_key_version_mask; \
167 (inode)->u.reiserfs_i.i_flags &= ~i_item_key_version_mask; })
169 #define get_inode_sd_version(inode) \
170 (((inode)->u.reiserfs_i.i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
172 #define set_inode_sd_version(inode, version) \
173 ({ if((version)==STAT_DATA_V2) \
174 (inode)->u.reiserfs_i.i_flags |= i_stat_data_version_mask; \
176 (inode)->u.reiserfs_i.i_flags &= ~i_stat_data_version_mask; })
178 /* This is an aggressive tail suppression policy, I am hoping it
179 improves our benchmarks. The principle behind it is that percentage
180 space saving is what matters, not absolute space saving. This is
181 non-intuitive, but it helps to understand it if you consider that the
182 cost to access 4 blocks is not much more than the cost to access 1
183 block, if you have to do a seek and rotate. A tail risks a
184 non-linear disk access that is significant as a percentage of total
185 time cost for a 4 block file and saves an amount of space that is
186 less significant as a percentage of space, or so goes the hypothesis.
188 #define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
190 (!(n_tail_size)) || \
191 (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
192 ( (n_file_size) >= (n_block_size) * 4 ) || \
193 ( ( (n_file_size) >= (n_block_size) * 3 ) && \
194 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
195 ( ( (n_file_size) >= (n_block_size) * 2 ) && \
196 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
197 ( ( (n_file_size) >= (n_block_size) ) && \
198 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
201 /* Another strategy for tails, this one means only create a tail if all the
202 file would fit into one DIRECT item.
203 Primary intention for this one is to increase performance by decreasing
206 #define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
208 (!(n_tail_size)) || \
209 (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
215 * values for s_state field
217 #define REISERFS_VALID_FS 1
218 #define REISERFS_ERROR_FS 2
221 // there are 5 item types currently
223 #define TYPE_STAT_DATA 0
224 #define TYPE_INDIRECT 1
225 #define TYPE_DIRECT 2
226 #define TYPE_DIRENTRY 3
227 #define TYPE_MAXTYPE 3
228 #define TYPE_ANY 15 // FIXME: comment is required
230 /***************************************************************************/
231 /* KEY & ITEM HEAD */
232 /***************************************************************************/
235 // directories use this key as well as old files
240 } __attribute__ ((__packed__));
243 #ifdef __LITTLE_ENDIAN
244 /* little endian version */
248 /* big endian version */
252 } __attribute__ ((__packed__));
254 #ifndef __LITTLE_ENDIAN
256 struct offset_v2 offset_v2;
258 } __attribute__ ((__packed__)) offset_v2_esafe_overlay;
260 static inline __u16 offset_v2_k_type( const struct offset_v2 *v2 )
262 offset_v2_esafe_overlay tmp = *(const offset_v2_esafe_overlay *)v2;
263 tmp.linear = le64_to_cpu( tmp.linear );
264 return (tmp.offset_v2.k_type <= TYPE_MAXTYPE)?tmp.offset_v2.k_type:TYPE_ANY;
267 static inline void set_offset_v2_k_type( struct offset_v2 *v2, int type )
269 offset_v2_esafe_overlay *tmp = (offset_v2_esafe_overlay *)v2;
270 tmp->linear = le64_to_cpu(tmp->linear);
271 tmp->offset_v2.k_type = type;
272 tmp->linear = cpu_to_le64(tmp->linear);
275 static inline loff_t offset_v2_k_offset( const struct offset_v2 *v2 )
277 offset_v2_esafe_overlay tmp = *(const offset_v2_esafe_overlay *)v2;
278 tmp.linear = le64_to_cpu( tmp.linear );
279 return tmp.offset_v2.k_offset;
282 static inline void set_offset_v2_k_offset( struct offset_v2 *v2, loff_t offset ){
283 offset_v2_esafe_overlay *tmp = (offset_v2_esafe_overlay *)v2;
284 tmp->linear = le64_to_cpu(tmp->linear);
285 tmp->offset_v2.k_offset = offset;
286 tmp->linear = cpu_to_le64(tmp->linear);
289 # define offset_v2_k_type(v2) ((v2)->k_type)
290 # define set_offset_v2_k_type(v2,val) (offset_v2_k_type(v2) = (val))
291 # define offset_v2_k_offset(v2) ((v2)->k_offset)
292 # define set_offset_v2_k_offset(v2,val) (offset_v2_k_offset(v2) = (val))
295 /* Key of an item determines its location in the S+tree, and
296 is composed of 4 components */
298 __u32 k_dir_id; /* packing locality: by default parent
299 directory object id */
300 __u32 k_objectid; /* object identifier */
302 struct offset_v1 k_offset_v1;
303 struct offset_v2 k_offset_v2;
304 } __attribute__ ((__packed__)) u;
305 } __attribute__ ((__packed__));
309 struct key on_disk_key;
311 int key_length; /* 3 in all cases but direct2indirect and
312 indirect2direct conversion */
315 /* Our function for comparing keys can compare keys of different
316 lengths. It takes as a parameter the length of the keys it is to
317 compare. These defines are used in determining what is to be passed
318 to it as that parameter. */
319 #define REISERFS_FULL_KEY_LEN 4
320 #define REISERFS_SHORT_KEY_LEN 2
322 /* The result of the key compare */
323 #define FIRST_GREATER 1
324 #define SECOND_GREATER -1
325 #define KEYS_IDENTICAL 0
327 #define KEY_NOT_FOUND 0
329 #define KEY_SIZE (sizeof(struct key))
330 #define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
332 /* return values for search_by_key and clones */
334 #define ITEM_NOT_FOUND 0
335 #define ENTRY_FOUND 1
336 #define ENTRY_NOT_FOUND 0
337 #define DIRECTORY_NOT_FOUND -1
338 #define REGULAR_FILE_FOUND -2
339 #define DIRECTORY_FOUND -3
341 #define BYTE_NOT_FOUND 0
342 #define FILE_NOT_FOUND -1
344 #define POSITION_FOUND 1
345 #define POSITION_NOT_FOUND 0
347 // return values for reiserfs_find_entry and search_by_entry_key
349 #define NAME_NOT_FOUND 0
350 #define GOTO_PREVIOUS_ITEM 2
351 #define NAME_FOUND_INVISIBLE 3
353 /* Everything in the filesystem is stored as a set of items. The
354 item head contains the key of the item, its free space (for
355 indirect items) and specifies the location of the item itself
360 /* Everything in the tree is found by searching for it based on
364 /* The free space in the last unformatted node of an
365 indirect item if this is an indirect item. This
366 equals 0xFFFF iff this is a direct item or stat data
367 item. Note that the key, not this field, is used to
368 determine the item type, and thus which field this
370 __u16 ih_free_space_reserved;
371 /* Iff this is a directory item, this field equals the
372 number of directory entries in the directory item. */
373 __u16 ih_entry_count;
374 } __attribute__ ((__packed__)) u;
375 __u16 ih_item_len; /* total size of the item body */
376 __u16 ih_item_location; /* an offset to the item body
377 * within the block */
378 __u16 ih_version; /* 0 for all old items, 2 for new
379 ones. Highest bit is set by fsck
380 temporary, cleaned after all
382 } __attribute__ ((__packed__));
383 /* size of item header */
384 #define IH_SIZE (sizeof(struct item_head))
386 #define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved)
387 #define ih_version(ih) le16_to_cpu((ih)->ih_version)
388 #define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count)
389 #define ih_location(ih) le16_to_cpu((ih)->ih_item_location)
390 #define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len)
392 #define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
393 #define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0)
394 #define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
395 #define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
396 #define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
399 #define unreachable_item(ih) (ih_version(ih) & (1 << 15))
401 #define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
402 #define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
404 /* these operate on indirect items, where you've got an array of ints
405 ** at a possibly unaligned location. These are a noop on ia32
407 ** p is the array of __u32, i is the index into the array, v is the value
410 #define get_block_num(p, i) le32_to_cpu(get_unaligned((p) + (i)))
411 #define put_block_num(p, i, v) put_unaligned(cpu_to_le32(v), (p) + (i))
414 // in old version uniqueness field shows key type
416 #define V1_SD_UNIQUENESS 0
417 #define V1_INDIRECT_UNIQUENESS 0xfffffffe
418 #define V1_DIRECT_UNIQUENESS 0xffffffff
419 #define V1_DIRENTRY_UNIQUENESS 500
420 #define V1_ANY_UNIQUENESS 555 // FIXME: comment is required
422 extern void reiserfs_warning (const char * fmt, ...);
423 /* __attribute__( ( format ( printf, 1, 2 ) ) ); */
426 // here are conversion routines
428 static inline int uniqueness2type (__u32 uniqueness) CONSTF;
429 static inline int uniqueness2type (__u32 uniqueness)
431 switch (uniqueness) {
432 case V1_SD_UNIQUENESS: return TYPE_STAT_DATA;
433 case V1_INDIRECT_UNIQUENESS: return TYPE_INDIRECT;
434 case V1_DIRECT_UNIQUENESS: return TYPE_DIRECT;
435 case V1_DIRENTRY_UNIQUENESS: return TYPE_DIRENTRY;
437 reiserfs_warning( "vs-500: unknown uniqueness %d\n", uniqueness);
438 case V1_ANY_UNIQUENESS:
443 static inline __u32 type2uniqueness (int type) CONSTF;
444 static inline __u32 type2uniqueness (int type)
447 case TYPE_STAT_DATA: return V1_SD_UNIQUENESS;
448 case TYPE_INDIRECT: return V1_INDIRECT_UNIQUENESS;
449 case TYPE_DIRECT: return V1_DIRECT_UNIQUENESS;
450 case TYPE_DIRENTRY: return V1_DIRENTRY_UNIQUENESS;
452 reiserfs_warning( "vs-501: unknown type %d\n", type);
454 return V1_ANY_UNIQUENESS;
459 // key is pointer to on disk key which is stored in le, result is cpu,
460 // there is no way to get version of object from key, so, provide
461 // version to these defines
463 static inline loff_t le_key_k_offset (int version, const struct key * key)
465 return (version == KEY_FORMAT_3_5) ?
466 le32_to_cpu( key->u.k_offset_v1.k_offset ) :
467 offset_v2_k_offset( &(key->u.k_offset_v2) );
470 static inline loff_t le_ih_k_offset (const struct item_head * ih)
472 return le_key_k_offset (ih_version (ih), &(ih->ih_key));
475 static inline loff_t le_key_k_type (int version, const struct key * key)
477 return (version == KEY_FORMAT_3_5) ?
478 uniqueness2type( le32_to_cpu( key->u.k_offset_v1.k_uniqueness)) :
479 offset_v2_k_type( &(key->u.k_offset_v2) );
482 static inline loff_t le_ih_k_type (const struct item_head * ih)
484 return le_key_k_type (ih_version (ih), &(ih->ih_key));
488 static inline void set_le_key_k_offset (int version, struct key * key, loff_t offset)
490 (version == KEY_FORMAT_3_5) ?
491 (key->u.k_offset_v1.k_offset = cpu_to_le32 (offset)) : /* jdm check */
492 (set_offset_v2_k_offset( &(key->u.k_offset_v2), offset ));
496 static inline void set_le_ih_k_offset (struct item_head * ih, loff_t offset)
498 set_le_key_k_offset (ih_version (ih), &(ih->ih_key), offset);
502 static inline void set_le_key_k_type (int version, struct key * key, int type)
504 (version == KEY_FORMAT_3_5) ?
505 (key->u.k_offset_v1.k_uniqueness = cpu_to_le32(type2uniqueness(type))):
506 (set_offset_v2_k_type( &(key->u.k_offset_v2), type ));
508 static inline void set_le_ih_k_type (struct item_head * ih, int type)
510 set_le_key_k_type (ih_version (ih), &(ih->ih_key), type);
514 #define is_direntry_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRENTRY)
515 #define is_direct_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRECT)
516 #define is_indirect_le_key(version,key) (le_key_k_type (version, key) == TYPE_INDIRECT)
517 #define is_statdata_le_key(version,key) (le_key_k_type (version, key) == TYPE_STAT_DATA)
520 // item header has version.
522 #define is_direntry_le_ih(ih) is_direntry_le_key (ih_version (ih), &((ih)->ih_key))
523 #define is_direct_le_ih(ih) is_direct_le_key (ih_version (ih), &((ih)->ih_key))
524 #define is_indirect_le_ih(ih) is_indirect_le_key (ih_version(ih), &((ih)->ih_key))
525 #define is_statdata_le_ih(ih) is_statdata_le_key (ih_version (ih), &((ih)->ih_key))
530 // key is pointer to cpu key, result is cpu
532 static inline loff_t cpu_key_k_offset (const struct cpu_key * key)
534 return (key->version == KEY_FORMAT_3_5) ?
535 key->on_disk_key.u.k_offset_v1.k_offset :
536 key->on_disk_key.u.k_offset_v2.k_offset;
539 static inline loff_t cpu_key_k_type (const struct cpu_key * key)
541 return (key->version == KEY_FORMAT_3_5) ?
542 uniqueness2type (key->on_disk_key.u.k_offset_v1.k_uniqueness) :
543 key->on_disk_key.u.k_offset_v2.k_type;
546 static inline void set_cpu_key_k_offset (struct cpu_key * key, loff_t offset)
548 (key->version == KEY_FORMAT_3_5) ?
549 (key->on_disk_key.u.k_offset_v1.k_offset = offset) :
550 (key->on_disk_key.u.k_offset_v2.k_offset = offset);
554 static inline void set_cpu_key_k_type (struct cpu_key * key, int type)
556 (key->version == KEY_FORMAT_3_5) ?
557 (key->on_disk_key.u.k_offset_v1.k_uniqueness = type2uniqueness (type)):
558 (key->on_disk_key.u.k_offset_v2.k_type = type);
562 static inline void cpu_key_k_offset_dec (struct cpu_key * key)
564 if (key->version == KEY_FORMAT_3_5)
565 key->on_disk_key.u.k_offset_v1.k_offset --;
567 key->on_disk_key.u.k_offset_v2.k_offset --;
571 #define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
572 #define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
573 #define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
574 #define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
577 /* are these used ? */
578 #define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
579 #define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
580 #define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
581 #define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
587 #define I_K_KEY_IN_ITEM(p_s_ih, p_s_key, n_blocksize) \
588 ( ! COMP_SHORT_KEYS(p_s_ih, p_s_key) && \
589 I_OFF_BYTE_IN_ITEM(p_s_ih, k_offset (p_s_key), n_blocksize) )
591 /* maximal length of item */
592 #define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
593 #define MIN_ITEM_LEN 1
596 /* object identifier for root dir */
597 #define REISERFS_ROOT_OBJECTID 2
598 #define REISERFS_ROOT_PARENT_OBJECTID 1
599 extern struct key root_key;
605 * Picture represents a leaf of the S+tree
606 * ______________________________________________________
608 * |Block | Object-Item | F r e e | Objects- |
609 * | head | Headers | S p a c e | Items |
610 * |______|_______________|___________________|___________|
613 /* Header of a disk block. More precisely, header of a formatted leaf
614 or internal node, and not the header of an unformatted node. */
616 __u16 blk_level; /* Level of a block in the tree. */
617 __u16 blk_nr_item; /* Number of keys/items in a block. */
618 __u16 blk_free_space; /* Block free space in bytes. */
620 /* dump this in v4/planA */
621 struct key blk_right_delim_key; /* kept only for compatibility */
624 #define BLKH_SIZE (sizeof(struct block_head))
625 #define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level))
626 #define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item))
627 #define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space))
628 #define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved))
629 #define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val))
630 #define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val))
631 #define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
632 #define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
633 #define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key)
634 #define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val)
637 * values for blk_level field of the struct block_head
640 #define FREE_LEVEL 0 /* when node gets removed from the tree its
641 blk_level is set to FREE_LEVEL. It is then
642 used to see whether the node is still in the
645 #define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level.*/
647 /* Given the buffer head of a formatted node, resolve to the block head of that node. */
648 #define B_BLK_HEAD(p_s_bh) ((struct block_head *)((p_s_bh)->b_data))
649 /* Number of items that are in buffer. */
650 #define B_NR_ITEMS(p_s_bh) (blkh_nr_item(B_BLK_HEAD(p_s_bh)))
651 #define B_LEVEL(p_s_bh) (blkh_level(B_BLK_HEAD(p_s_bh)))
652 #define B_FREE_SPACE(p_s_bh) (blkh_free_space(B_BLK_HEAD(p_s_bh)))
654 #define PUT_B_NR_ITEMS(p_s_bh,val) do { set_blkh_nr_item(B_BLK_HEAD(p_s_bh),val); } while (0)
655 #define PUT_B_LEVEL(p_s_bh,val) do { set_blkh_level(B_BLK_HEAD(p_s_bh),val); } while (0)
656 #define PUT_B_FREE_SPACE(p_s_bh,val) do { set_blkh_free_space(B_BLK_HEAD(p_s_bh),val); } while (0)
659 /* Get right delimiting key. -- little endian */
660 #define B_PRIGHT_DELIM_KEY(p_s_bh) (&(blk_right_delim_key(B_BLK_HEAD(p_s_bh))
662 /* Does the buffer contain a disk leaf. */
663 #define B_IS_ITEMS_LEVEL(p_s_bh) (B_LEVEL(p_s_bh) == DISK_LEAF_NODE_LEVEL)
665 /* Does the buffer contain a disk internal node */
666 #define B_IS_KEYS_LEVEL(p_s_bh) (B_LEVEL(p_s_bh) > DISK_LEAF_NODE_LEVEL \
667 && B_LEVEL(p_s_bh) <= MAX_HEIGHT)
672 /***************************************************************************/
674 /***************************************************************************/
678 // old stat data is 32 bytes long. We are going to distinguish new one by
683 __u16 sd_mode; /* file type, permissions */
684 __u16 sd_nlink; /* number of hard links */
685 __u16 sd_uid; /* owner */
686 __u16 sd_gid; /* group */
687 __u32 sd_size; /* file size */
688 __u32 sd_atime; /* time of last access */
689 __u32 sd_mtime; /* time file was last modified */
690 __u32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
693 __u32 sd_blocks; /* number of blocks file uses */
694 } __attribute__ ((__packed__)) u;
695 __u32 sd_first_direct_byte; /* first byte of file which is stored
696 in a direct item: except that if it
697 equals 1 it is a symlink and if it
698 equals ~(__u32)0 there is no
699 direct item. The existence of this
700 field really grates on me. Let's
701 replace it with a macro based on
702 sd_size and our tail suppression
703 policy. Someday. -Hans */
704 } __attribute__ ((__packed__));
706 #define SD_V1_SIZE (sizeof(struct stat_data_v1))
707 #define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5)
708 #define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
709 #define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
710 #define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink))
711 #define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v))
712 #define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid))
713 #define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v))
714 #define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid))
715 #define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v))
716 #define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size))
717 #define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v))
718 #define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
719 #define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
720 #define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
721 #define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
722 #define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
723 #define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
724 #define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
725 #define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
726 #define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks))
727 #define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
728 #define sd_v1_first_direct_byte(sdp) \
729 (le32_to_cpu((sdp)->sd_first_direct_byte))
730 #define set_sd_v1_first_direct_byte(sdp,v) \
731 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
733 #include <linux/ext2_fs.h>
735 /* inode flags stored in sd_attrs (nee sd_reserved) */
737 /* we want common flags to have the same values as in ext2,
738 so chattr(1) will work without problems */
739 #define REISERFS_IMMUTABLE_FL EXT2_IMMUTABLE_FL
740 #define REISERFS_SYNC_FL EXT2_SYNC_FL
741 #define REISERFS_NOATIME_FL EXT2_NOATIME_FL
742 #define REISERFS_NODUMP_FL EXT2_NODUMP_FL
743 #define REISERFS_SECRM_FL EXT2_SECRM_FL
744 #define REISERFS_UNRM_FL EXT2_UNRM_FL
745 #define REISERFS_COMPR_FL EXT2_COMPR_FL
746 /* persistent flag to disable tails on per-file basic.
747 Note, that is inheritable: mark directory with this and
748 all new files inside will not have tails.
750 Teodore Tso allocated EXT2_NODUMP_FL (0x00008000) for this. Change
751 numeric constant to ext2 macro when available. */
752 #define REISERFS_NOTAIL_FL (0x00008000) /* EXT2_NOTAIL_FL */
754 /* persistent flags that file inherits from the parent directory */
755 #define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \
757 REISERFS_NOATIME_FL | \
758 REISERFS_NODUMP_FL | \
759 REISERFS_SECRM_FL | \
760 REISERFS_COMPR_FL | \
763 /* Stat Data on disk (reiserfs version of UFS disk inode minus the
766 __u16 sd_mode; /* file type, permissions */
767 __u16 sd_attrs; /* persistent inode flags */
768 __u32 sd_nlink; /* number of hard links */
769 __u64 sd_size; /* file size */
770 __u32 sd_uid; /* owner */
771 __u32 sd_gid; /* group */
772 __u32 sd_atime; /* time of last access */
773 __u32 sd_mtime; /* time file was last modified */
774 __u32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
779 //__u32 sd_first_direct_byte;
780 /* first byte of file which is stored in a
781 direct item: except that if it equals 1
782 it is a symlink and if it equals
783 ~(__u32)0 there is no direct item. The
784 existence of this field really grates
785 on me. Let's replace it with a macro
786 based on sd_size and our tail
787 suppression policy? */
788 } __attribute__ ((__packed__)) u;
789 } __attribute__ ((__packed__));
791 // this is 44 bytes long
793 #define SD_SIZE (sizeof(struct stat_data))
794 #define SD_V2_SIZE SD_SIZE
795 #define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6)
796 #define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
797 #define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
799 /* set_sd_reserved */
800 #define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink))
801 #define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v))
802 #define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size))
803 #define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v))
804 #define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid))
805 #define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v))
806 #define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid))
807 #define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v))
808 #define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
809 #define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
810 #define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
811 #define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
812 #define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
813 #define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
814 #define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks))
815 #define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
816 #define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
817 #define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
818 #define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation))
819 #define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
820 #define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs))
821 #define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v))
824 /***************************************************************************/
825 /* DIRECTORY STRUCTURE */
826 /***************************************************************************/
828 Picture represents the structure of directory items
829 ________________________________________________
830 | Array of | | | | | |
831 | directory |N-1| N-2 | .... | 1st |0th|
832 | entry headers | | | | | |
833 |_______________|___|_____|________|_______|___|
834 <---- directory entries ------>
836 First directory item has k_offset component 1. We store "." and ".."
837 in one item, always, we never split "." and ".." into differing
838 items. This makes, among other things, the code for removing
839 directories simpler. */
841 #define SD_UNIQUENESS 0
843 #define DOT_DOT_OFFSET 2
844 #define DIRENTRY_UNIQUENESS 500
847 #define FIRST_ITEM_OFFSET 1
850 Q: How to get key of object pointed to by entry from entry?
852 A: Each directory entry has its header. This header has deh_dir_id and deh_objectid fields, those are key
853 of object, entry points to */
856 Directory will someday contain stat data of object */
860 struct reiserfs_de_head
862 __u32 deh_offset; /* third component of the directory entry key */
863 __u32 deh_dir_id; /* objectid of the parent directory of the object, that is referenced
864 by directory entry */
865 __u32 deh_objectid; /* objectid of the object, that is referenced by directory entry */
866 __u16 deh_location; /* offset of name in the whole item */
867 __u16 deh_state; /* whether 1) entry contains stat data (for future), and 2) whether
868 entry is hidden (unlinked) */
869 } __attribute__ ((__packed__));
870 #define DEH_SIZE sizeof(struct reiserfs_de_head)
871 #define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset))
872 #define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id))
873 #define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid))
874 #define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location))
875 #define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state))
877 #define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v)))
878 #define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v)))
879 #define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
880 #define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
881 #define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v)))
883 /* empty directory contains two entries "." and ".." and their headers */
884 #define EMPTY_DIR_SIZE \
885 (DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
887 /* old format directories have this size when empty */
888 #define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
890 #define DEH_Statdata 0 /* not used now */
891 #define DEH_Visible 2
893 /* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
894 #if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
895 # define ADDR_UNALIGNED_BITS (3)
898 /* These are only used to manipulate deh_state.
899 * Because of this, we'll use the ext2_ bit routines,
900 * since they are little endian */
901 #ifdef ADDR_UNALIGNED_BITS
903 # define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
904 # define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
906 # define set_bit_unaligned(nr, addr) ext2_set_bit((nr) + unaligned_offset(addr), aligned_address(addr))
907 # define clear_bit_unaligned(nr, addr) ext2_clear_bit((nr) + unaligned_offset(addr), aligned_address(addr))
908 # define test_bit_unaligned(nr, addr) ext2_test_bit((nr) + unaligned_offset(addr), aligned_address(addr))
912 # define set_bit_unaligned(nr, addr) ext2_set_bit(nr, addr)
913 # define clear_bit_unaligned(nr, addr) ext2_clear_bit(nr, addr)
914 # define test_bit_unaligned(nr, addr) ext2_test_bit(nr, addr)
918 #define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
919 #define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
920 #define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
921 #define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
923 #define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
924 #define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
925 #define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
927 extern void make_empty_dir_item_v1 (char * body, __u32 dirid, __u32 objid,
928 __u32 par_dirid, __u32 par_objid);
929 extern void make_empty_dir_item (char * body, __u32 dirid, __u32 objid,
930 __u32 par_dirid, __u32 par_objid);
932 /* array of the entry headers */
934 #define B_I_PITEM(bh,ih) ( (bh)->b_data + ih_location(ih) )
935 #define B_I_DEH(bh,ih) ((struct reiserfs_de_head *)(B_I_PITEM(bh,ih)))
937 /* length of the directory entry in directory item. This define
938 calculates length of i-th directory entry using directory entry
939 locations from dir entry head. When it calculates length of 0-th
940 directory entry, it uses length of whole item in place of entry
941 location of the non-existent following entry in the calculation.
944 #define I_DEH_N_ENTRY_LENGTH(ih,deh,i) \
945 ((i) ? (deh_location((deh)-1) - deh_location((deh))) : (ih_item_len((ih)) - deh_location((deh))))
947 static inline int entry_length (const struct buffer_head * bh,
948 const struct item_head * ih, int pos_in_item)
950 struct reiserfs_de_head * deh;
952 deh = B_I_DEH (bh, ih) + pos_in_item;
954 return deh_location(deh-1) - deh_location(deh);
956 return ih_item_len(ih) - deh_location(deh);
961 /* number of entries in the directory item, depends on ENTRY_COUNT being at the start of directory dynamic data. */
962 #define I_ENTRY_COUNT(ih) (ih_entry_count((ih)))
965 /* name by bh, ih and entry_num */
966 #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))))
968 // two entries per block (at least)
969 #define REISERFS_MAX_NAME(block_size) 255
972 /* this structure is used for operations on directory entries. It is
973 not a disk structure. */
974 /* When reiserfs_find_entry or search_by_entry_key find directory
975 entry, they return filled reiserfs_dir_entry structure */
976 struct reiserfs_dir_entry
978 struct buffer_head * de_bh;
980 struct item_head * de_ih;
982 struct reiserfs_de_head * de_deh;
986 char * de_gen_number_bit_string;
991 struct cpu_key de_entry_key;
994 /* these defines are useful when a particular member of a reiserfs_dir_entry is needed */
996 /* pointer to file name, stored in entry */
997 #define B_I_DEH_ENTRY_FILE_NAME(bh,ih,deh) (B_I_PITEM (bh, ih) + deh_location(deh))
1000 #define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
1001 (I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
1005 /* hash value occupies bits from 7 up to 30 */
1006 #define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
1007 /* generation number occupies 7 bits starting from 0 up to 6 */
1008 #define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
1009 #define MAX_GENERATION_NUMBER 127
1011 #define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
1015 * Picture represents an internal node of the reiserfs tree
1016 * ______________________________________________________
1017 * | | Array of | Array of | Free |
1018 * |block | keys | pointers | space |
1019 * | head | N | N+1 | |
1020 * |______|_______________|___________________|___________|
1023 /***************************************************************************/
1025 /***************************************************************************/
1026 /* Disk child pointer: The pointer from an internal node of the tree
1027 to a node that is on disk. */
1029 __u32 dc_block_number; /* Disk child's block number. */
1030 __u16 dc_size; /* Disk child's used space. */
1034 #define DC_SIZE (sizeof(struct disk_child))
1035 #define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number))
1036 #define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size))
1037 #define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
1038 #define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
1040 /* Get disk child by buffer header and position in the tree node. */
1041 #define B_N_CHILD(p_s_bh,n_pos) ((struct disk_child *)\
1042 ((p_s_bh)->b_data+BLKH_SIZE+B_NR_ITEMS(p_s_bh)*KEY_SIZE+DC_SIZE*(n_pos)))
1044 /* Get disk child number by buffer header and position in the tree node. */
1045 #define B_N_CHILD_NUM(p_s_bh,n_pos) (dc_block_number(B_N_CHILD(p_s_bh,n_pos)))
1046 #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 ))
1048 /* maximal value of field child_size in structure disk_child */
1049 /* child size is the combined size of all items and their headers */
1050 #define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
1052 /* amount of used space in buffer (not including block head) */
1053 #define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
1055 /* max and min number of keys in internal node */
1056 #define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
1057 #define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2)
1059 /***************************************************************************/
1060 /* PATH STRUCTURES AND DEFINES */
1061 /***************************************************************************/
1064 /* Search_by_key fills up the path from the root to the leaf as it descends the tree looking for the
1065 key. It uses reiserfs_bread to try to find buffers in the cache given their block number. If it
1066 does not find them in the cache it reads them from disk. For each node search_by_key finds using
1067 reiserfs_bread it then uses bin_search to look through that node. bin_search will find the
1068 position of the block_number of the next node if it is looking through an internal node. If it
1069 is looking through a leaf node bin_search will find the position of the item which has key either
1070 equal to given key, or which is the maximal key less than the given key. */
1072 struct path_element {
1073 struct buffer_head * pe_buffer; /* Pointer to the buffer at the path in the tree. */
1074 int pe_position; /* Position in the tree node which is placed in the */
1078 #define MAX_HEIGHT 5 /* maximal height of a tree. don't change this without changing JOURNAL_PER_BALANCE_CNT */
1079 #define EXTENDED_MAX_HEIGHT 7 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
1080 #define FIRST_PATH_ELEMENT_OFFSET 2 /* Must be equal to at least 2. */
1082 #define ILLEGAL_PATH_ELEMENT_OFFSET 1 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
1083 #define MAX_FEB_SIZE 6 /* this MUST be MAX_HEIGHT + 1. See about FEB below */
1087 /* We need to keep track of who the ancestors of nodes are. When we
1088 perform a search we record which nodes were visited while
1089 descending the tree looking for the node we searched for. This list
1090 of nodes is called the path. This information is used while
1091 performing balancing. Note that this path information may become
1092 invalid, and this means we must check it when using it to see if it
1093 is still valid. You'll need to read search_by_key and the comments
1094 in it, especially about decrement_counters_in_path(), to understand
1097 Paths make the code so much harder to work with and debug.... An
1098 enormous number of bugs are due to them, and trying to write or modify
1099 code that uses them just makes my head hurt. They are based on an
1100 excessive effort to avoid disturbing the precious VFS code.:-( The
1101 gods only know how we are going to SMP the code that uses them.
1102 znodes are the way! */
1106 int path_length; /* Length of the array above. */
1107 struct path_element path_elements[EXTENDED_MAX_HEIGHT]; /* Array of the path elements. */
1111 #define pos_in_item(path) ((path)->pos_in_item)
1113 #define INITIALIZE_PATH(var) \
1114 struct path var = {ILLEGAL_PATH_ELEMENT_OFFSET, }
1116 /* Get path element by path and path position. */
1117 #define PATH_OFFSET_PELEMENT(p_s_path,n_offset) ((p_s_path)->path_elements +(n_offset))
1119 /* Get buffer header at the path by path and path position. */
1120 #define PATH_OFFSET_PBUFFER(p_s_path,n_offset) (PATH_OFFSET_PELEMENT(p_s_path,n_offset)->pe_buffer)
1122 /* Get position in the element at the path by path and path position. */
1123 #define PATH_OFFSET_POSITION(p_s_path,n_offset) (PATH_OFFSET_PELEMENT(p_s_path,n_offset)->pe_position)
1126 #define PATH_PLAST_BUFFER(p_s_path) (PATH_OFFSET_PBUFFER((p_s_path), (p_s_path)->path_length))
1127 /* you know, to the person who didn't
1128 write this the macro name does not
1129 at first suggest what it does.
1130 Maybe POSITION_FROM_PATH_END? Or
1131 maybe we should just focus on
1132 dumping paths... -Hans */
1133 #define PATH_LAST_POSITION(p_s_path) (PATH_OFFSET_POSITION((p_s_path), (p_s_path)->path_length))
1136 #define PATH_PITEM_HEAD(p_s_path) B_N_PITEM_HEAD(PATH_PLAST_BUFFER(p_s_path),PATH_LAST_POSITION(p_s_path))
1138 /* in do_balance leaf has h == 0 in contrast with path structure,
1139 where root has level == 0. That is why we need these defines */
1140 #define PATH_H_PBUFFER(p_s_path, h) PATH_OFFSET_PBUFFER (p_s_path, p_s_path->path_length - (h)) /* tb->S[h] */
1141 #define PATH_H_PPARENT(path, h) PATH_H_PBUFFER (path, (h) + 1) /* tb->F[h] or tb->S[0]->b_parent */
1142 #define PATH_H_POSITION(path, h) PATH_OFFSET_POSITION (path, path->path_length - (h))
1143 #define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1) /* tb->S[h]->b_item_order */
1145 #define PATH_H_PATH_OFFSET(p_s_path, n_h) ((p_s_path)->path_length - (n_h))
1147 #define get_last_bh(path) PATH_PLAST_BUFFER(path)
1148 #define get_ih(path) PATH_PITEM_HEAD(path)
1149 #define get_item_pos(path) PATH_LAST_POSITION(path)
1150 #define get_item(path) ((void *)B_N_PITEM(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION (path)))
1151 #define item_moved(ih,path) comp_items(ih, path)
1152 #define path_changed(ih,path) comp_items (ih, path)
1155 /***************************************************************************/
1157 /***************************************************************************/
1159 /* Size of pointer to the unformatted node. */
1160 #define UNFM_P_SIZE (sizeof(unp_t))
1162 // in in-core inode key is stored on le form
1163 #define INODE_PKEY(inode) ((struct key *)((inode)->u.reiserfs_i.i_key))
1165 #define MAX_UL_INT 0xffffffff
1166 #define MAX_INT 0x7ffffff
1167 #define MAX_US_INT 0xffff
1169 // reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
1170 #define U32_MAX (~(__u32)0)
1172 static inline loff_t max_reiserfs_offset (const struct inode * inode)
1174 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
1175 return (loff_t)U32_MAX;
1177 return (loff_t)((~(__u64)0) >> 4);
1181 /*#define MAX_KEY_UNIQUENESS MAX_UL_INT*/
1182 #define MAX_KEY_OBJECTID MAX_UL_INT
1185 #define MAX_B_NUM MAX_UL_INT
1186 #define MAX_FC_NUM MAX_US_INT
1189 /* the purpose is to detect overflow of an unsigned short */
1190 #define REISERFS_LINK_MAX (MAX_US_INT - 1000)
1193 /* The following defines are used in reiserfs_insert_item and reiserfs_append_item */
1194 #define REISERFS_KERNEL_MEM 0 /* reiserfs kernel memory mode */
1195 #define REISERFS_USER_MEM 1 /* reiserfs user memory mode */
1197 #define fs_generation(s) ((s)->u.reiserfs_sb.s_generation_counter)
1198 #define get_generation(s) atomic_read (&fs_generation(s))
1199 #define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen)
1200 #define fs_changed(gen,s) (gen != get_generation (s))
1203 /***************************************************************************/
1205 /***************************************************************************/
1207 #define VI_TYPE_LEFT_MERGEABLE 1
1208 #define VI_TYPE_RIGHT_MERGEABLE 2
1210 /* To make any changes in the tree we always first find node, that
1211 contains item to be changed/deleted or place to insert a new
1212 item. We call this node S. To do balancing we need to decide what
1213 we will shift to left/right neighbor, or to a new node, where new
1214 item will be etc. To make this analysis simpler we build virtual
1215 node. Virtual node is an array of items, that will replace items of
1216 node S. (For instance if we are going to delete an item, virtual
1217 node does not contain it). Virtual node keeps information about
1218 item sizes and types, mergeability of first and last items, sizes
1219 of all entries in directory item. We use this array of items when
1220 calculating what we can shift to neighbors and how many nodes we
1221 have to have if we do not any shiftings, if we shift to left/right
1222 neighbor or to both. */
1225 int vi_index; // index in the array of item operations
1226 unsigned short vi_type; // left/right mergeability
1227 unsigned short vi_item_len; /* length of item that it will have after balancing */
1228 struct item_head * vi_ih;
1229 const char * vi_item; // body of item (old or new)
1230 const void * vi_new_data; // 0 always but paste mode
1231 void * vi_uarea; // item specific area
1237 char * vn_free_ptr; /* this is a pointer to the free space in the buffer */
1238 unsigned short vn_nr_item; /* number of items in virtual node */
1239 short vn_size; /* size of node , that node would have if it has unlimited size and no balancing is performed */
1240 short vn_mode; /* mode of balancing (paste, insert, delete, cut) */
1241 short vn_affected_item_num;
1242 short vn_pos_in_item;
1243 struct item_head * vn_ins_ih; /* item header of inserted item, 0 for other modes */
1244 const void * vn_data;
1245 struct virtual_item * vn_vi; /* array of items (including a new one, excluding item to be deleted) */
1248 /* used by directory items when creating virtual nodes */
1249 struct direntry_uarea {
1252 __u16 entry_sizes[1];
1253 } __attribute__ ((__packed__)) ;
1256 /***************************************************************************/
1258 /***************************************************************************/
1260 /* This temporary structure is used in tree balance algorithms, and
1261 constructed as we go to the extent that its various parts are
1262 needed. It contains arrays of nodes that can potentially be
1263 involved in the balancing of node S, and parameters that define how
1264 each of the nodes must be balanced. Note that in these algorithms
1265 for balancing the worst case is to need to balance the current node
1266 S and the left and right neighbors and all of their parents plus
1267 create a new node. We implement S1 balancing for the leaf nodes
1268 and S0 balancing for the internal nodes (S1 and S0 are defined in
1271 #define MAX_FREE_BLOCK 7 /* size of the array of buffers to free at end of do_balance */
1273 /* maximum number of FEB blocknrs on a single level */
1274 #define MAX_AMOUNT_NEEDED 2
1276 /* someday somebody will prefix every field in this struct with tb_ */
1280 int need_balance_dirty;
1281 struct super_block * tb_sb;
1282 struct reiserfs_transaction_handle *transaction_handle ;
1283 struct path * tb_path;
1284 struct buffer_head * L[MAX_HEIGHT]; /* array of left neighbors of nodes in the path */
1285 struct buffer_head * R[MAX_HEIGHT]; /* array of right neighbors of nodes in the path*/
1286 struct buffer_head * FL[MAX_HEIGHT]; /* array of fathers of the left neighbors */
1287 struct buffer_head * FR[MAX_HEIGHT]; /* array of fathers of the right neighbors */
1288 struct buffer_head * CFL[MAX_HEIGHT]; /* array of common parents of center node and its left neighbor */
1289 struct buffer_head * CFR[MAX_HEIGHT]; /* array of common parents of center node and its right neighbor */
1291 struct buffer_head * FEB[MAX_FEB_SIZE]; /* array of empty buffers. Number of buffers in array equals
1293 struct buffer_head * used[MAX_FEB_SIZE];
1294 struct buffer_head * thrown[MAX_FEB_SIZE];
1295 int lnum[MAX_HEIGHT]; /* array of number of items which must be
1296 shifted to the left in order to balance the
1297 current node; for leaves includes item that
1298 will be partially shifted; for internal
1299 nodes, it is the number of child pointers
1300 rather than items. It includes the new item
1301 being created. The code sometimes subtracts
1302 one to get the number of wholly shifted
1303 items for other purposes. */
1304 int rnum[MAX_HEIGHT]; /* substitute right for left in comment above */
1305 int lkey[MAX_HEIGHT]; /* array indexed by height h mapping the key delimiting L[h] and
1306 S[h] to its item number within the node CFL[h] */
1307 int rkey[MAX_HEIGHT]; /* substitute r for l in comment above */
1308 int insert_size[MAX_HEIGHT]; /* the number of bytes by we are trying to add or remove from
1309 S[h]. A negative value means removing. */
1310 int blknum[MAX_HEIGHT]; /* number of nodes that will replace node S[h] after
1311 balancing on the level h of the tree. If 0 then S is
1312 being deleted, if 1 then S is remaining and no new nodes
1313 are being created, if 2 or 3 then 1 or 2 new nodes is
1316 /* fields that are used only for balancing leaves of the tree */
1317 int cur_blknum; /* number of empty blocks having been already allocated */
1318 int s0num; /* number of items that fall into left most node when S[0] splits */
1319 int s1num; /* number of items that fall into first new node when S[0] splits */
1320 int s2num; /* number of items that fall into second new node when S[0] splits */
1321 int lbytes; /* number of bytes which can flow to the left neighbor from the left */
1322 /* most liquid item that cannot be shifted from S[0] entirely */
1323 /* if -1 then nothing will be partially shifted */
1324 int rbytes; /* number of bytes which will flow to the right neighbor from the right */
1325 /* most liquid item that cannot be shifted from S[0] entirely */
1326 /* if -1 then nothing will be partially shifted */
1327 int s1bytes; /* number of bytes which flow to the first new node when S[0] splits */
1328 /* note: if S[0] splits into 3 nodes, then items do not need to be cut */
1330 struct buffer_head * buf_to_free[MAX_FREE_BLOCK]; /* buffers which are to be freed after do_balance finishes by unfix_nodes */
1331 char * vn_buf; /* kmalloced memory. Used to create
1332 virtual node and keep map of
1333 dirtied bitmap blocks */
1334 int vn_buf_size; /* size of the vn_buf */
1335 struct virtual_node * tb_vn; /* VN starts after bitmap of bitmap blocks */
1337 int fs_gen; /* saved value of `reiserfs_generation' counter
1338 see FILESYSTEM_CHANGED() macro in reiserfs_fs.h */
1339 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1340 struct key key; /* key pointer, to pass to block allocator or
1341 another low-level subsystem */
1345 /* These are modes of balancing */
1347 /* When inserting an item. */
1348 #define M_INSERT 'i'
1349 /* When inserting into (directories only) or appending onto an already
1352 /* When deleting an item. */
1353 #define M_DELETE 'd'
1354 /* When truncating an item or removing an entry from a (directory) item. */
1357 /* used when balancing on leaf level skipped (in reiserfsck) */
1358 #define M_INTERNAL 'n'
1360 /* When further balancing is not needed, then do_balance does not need
1362 #define M_SKIP_BALANCING 's'
1363 #define M_CONVERT 'v'
1365 /* modes of leaf_move_items */
1366 #define LEAF_FROM_S_TO_L 0
1367 #define LEAF_FROM_S_TO_R 1
1368 #define LEAF_FROM_R_TO_L 2
1369 #define LEAF_FROM_L_TO_R 3
1370 #define LEAF_FROM_S_TO_SNEW 4
1372 #define FIRST_TO_LAST 0
1373 #define LAST_TO_FIRST 1
1375 /* used in do_balance for passing parent of node information that has
1376 been gotten from tb struct */
1377 struct buffer_info {
1378 struct tree_balance * tb;
1379 struct buffer_head * bi_bh;
1380 struct buffer_head * bi_parent;
1385 /* there are 4 types of items: stat data, directory item, indirect, direct.
1386 +-------------------+------------+--------------+------------+
1387 | | k_offset | k_uniqueness | mergeable? |
1388 +-------------------+------------+--------------+------------+
1389 | stat data | 0 | 0 | no |
1390 +-------------------+------------+--------------+------------+
1391 | 1st directory item| DOT_OFFSET |DIRENTRY_UNIQUENESS| no |
1392 | non 1st directory | hash value | | yes |
1394 +-------------------+------------+--------------+------------+
1395 | indirect item | offset + 1 |TYPE_INDIRECT | if this is not the first indirect item of the object
1396 +-------------------+------------+--------------+------------+
1397 | direct item | offset + 1 |TYPE_DIRECT | if not this is not the first direct item of the object
1398 +-------------------+------------+--------------+------------+
1401 struct item_operations {
1402 int (*bytes_number) (struct item_head * ih, int block_size);
1403 void (*decrement_key) (struct cpu_key *);
1404 int (*is_left_mergeable) (struct key * ih, unsigned long bsize);
1405 void (*print_item) (struct item_head *, char * item);
1406 void (*check_item) (struct item_head *, char * item);
1408 int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
1409 int is_affected, int insert_size);
1410 int (*check_left) (struct virtual_item * vi, int free,
1411 int start_skip, int end_skip);
1412 int (*check_right) (struct virtual_item * vi, int free);
1413 int (*part_size) (struct virtual_item * vi, int from, int to);
1414 int (*unit_num) (struct virtual_item * vi);
1415 void (*print_vi) (struct virtual_item * vi);
1419 extern struct item_operations stat_data_ops, indirect_ops, direct_ops,
1421 extern struct item_operations * item_ops [TYPE_ANY + 1];
1423 #define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
1424 #define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
1425 #define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item)
1426 #define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item)
1427 #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)
1428 #define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
1429 #define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free)
1430 #define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to)
1431 #define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi)
1432 #define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi)
1438 #define COMP_KEYS comp_keys
1439 #define COMP_SHORT_KEYS comp_short_keys
1440 /*#define keys_of_same_object comp_short_keys*/
1442 /* number of blocks pointed to by the indirect item */
1443 #define I_UNFM_NUM(p_s_ih) ( ih_item_len(p_s_ih) / UNFM_P_SIZE )
1445 /* the used space within the unformatted node corresponding to pos within the item pointed to by ih */
1446 #define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
1448 /* number of bytes contained by the direct item or the unformatted nodes the indirect item points to */
1451 /* get the item header */
1452 #define B_N_PITEM_HEAD(bh,item_num) ( (struct item_head * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1455 #define B_N_PDELIM_KEY(bh,item_num) ( (struct key * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1458 #define B_N_PKEY(bh,item_num) ( &(B_N_PITEM_HEAD(bh,item_num)->ih_key) )
1461 #define B_N_PITEM(bh,item_num) ( (bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(item_num))))
1463 /* get the stat data by the buffer header and the item order */
1464 #define B_N_STAT_DATA(bh,nr) \
1465 ( (struct stat_data *)((bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(nr))) ) )
1467 /* following defines use reiserfs buffer header and item header */
1470 #define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
1472 // this is 3976 for size==4096
1473 #define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
1475 /* indirect items consist of entries which contain blocknrs, pos
1476 indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
1477 blocknr contained by the entry pos points to */
1478 #define B_I_POS_UNFM_POINTER(bh,ih,pos) le32_to_cpu(*(((unp_t *)B_I_PITEM(bh,ih)) + (pos)))
1479 #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)
1481 struct reiserfs_iget4_args {
1485 /***************************************************************************/
1486 /* FUNCTION DECLARATIONS */
1487 /***************************************************************************/
1489 /*#ifdef __KERNEL__*/
1491 /* journal.c see journal.c for all the comments here */
1493 #define JOURNAL_TRANS_HALF 1018 /* must be correct to keep the desc and commit structs at 4k */
1496 /* first block written in a commit. */
1497 struct reiserfs_journal_desc {
1498 __u32 j_trans_id ; /* id of commit */
1499 __u32 j_len ; /* length of commit. len +1 is the commit block */
1500 __u32 j_mount_id ; /* mount id of this trans*/
1501 __u32 j_realblock[JOURNAL_TRANS_HALF] ; /* real locations for each block */
1505 /* last block written in a commit */
1506 struct reiserfs_journal_commit {
1507 __u32 j_trans_id ; /* must match j_trans_id from the desc block */
1508 __u32 j_len ; /* ditto */
1509 __u32 j_realblock[JOURNAL_TRANS_HALF] ; /* real locations for each block */
1510 char j_digest[16] ; /* md5 sum of all the blocks involved, including desc and commit. not used, kill it */
1513 /* this header block gets written whenever a transaction is considered fully flushed, and is more recent than the
1514 ** last fully flushed transaction. fully flushed means all the log blocks and all the real blocks are on disk,
1515 ** and this transaction does not need to be replayed.
1517 struct reiserfs_journal_header {
1518 __u32 j_last_flush_trans_id ; /* id of last fully flushed transaction */
1519 __u32 j_first_unflushed_offset ; /* offset in the log of where to start replay after a crash */
1523 extern task_queue reiserfs_commit_thread_tq ;
1524 extern wait_queue_head_t reiserfs_commit_thread_wait ;
1526 /* biggest tunable defines are right here */
1527 #define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */
1528 #define JOURNAL_MAX_BATCH 900 /* max blocks to batch into one transaction, don't make this any bigger than 900 */
1529 #define JOURNAL_MAX_COMMIT_AGE 30
1530 #define JOURNAL_MAX_TRANS_AGE 30
1531 #define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
1533 /* both of these can be as low as 1, or as high as you want. The min is the
1534 ** number of 4k bitmap nodes preallocated on mount. New nodes are allocated
1535 ** as needed, and released when transactions are committed. On release, if
1536 ** the current number of nodes is > max, the node is freed, otherwise,
1537 ** it is put on a free list for faster use later.
1539 #define REISERFS_MIN_BITMAP_NODES 10
1540 #define REISERFS_MAX_BITMAP_NODES 100
1542 #define JBH_HASH_SHIFT 13 /* these are based on journal hash size of 8192 */
1543 #define JBH_HASH_MASK 8191
1545 /* After several hours of tedious analysis, the following hash
1546 * function won. Do not mess with it... -DaveM
1548 /* The two definitions below were borrowed from fs/buffer.c file of Linux kernel
1549 * sources and are not licensed by Namesys to its non-GPL license customers */
1550 #define _jhashfn(dev,block) \
1551 ((((dev)<<(JBH_HASH_SHIFT - 6)) ^ ((dev)<<(JBH_HASH_SHIFT - 9))) ^ \
1552 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
1553 #define journal_hash(t,dev,block) ((t)[_jhashfn((dev),(block)) & JBH_HASH_MASK])
1555 /* finds n'th buffer with 0 being the start of this commit. Needs to go away, j_ap_blocks has changed
1556 ** since I created this. One chunk of code in journal.c needs changing before deleting it
1558 #define JOURNAL_BUFFER(j,n) ((j)->j_ap_blocks[((j)->j_start + (n)) % JOURNAL_BLOCK_COUNT])
1560 void reiserfs_commit_for_inode(struct inode *) ;
1561 void reiserfs_update_inode_transaction(struct inode *) ;
1562 void reiserfs_wait_on_write_block(struct super_block *s) ;
1563 void reiserfs_block_writes(struct reiserfs_transaction_handle *th) ;
1564 void reiserfs_allow_writes(struct super_block *s) ;
1565 void reiserfs_check_lock_depth(char *caller) ;
1566 void reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh, int wait) ;
1567 void reiserfs_restore_prepared_buffer(struct super_block *, struct buffer_head *bh) ;
1568 int journal_init(struct super_block *) ;
1569 int journal_release(struct reiserfs_transaction_handle*, struct super_block *) ;
1570 int journal_release_error(struct reiserfs_transaction_handle*, struct super_block *) ;
1571 int journal_end(struct reiserfs_transaction_handle *, struct super_block *, unsigned long) ;
1572 int journal_end_sync(struct reiserfs_transaction_handle *, struct super_block *, unsigned long) ;
1573 int journal_mark_dirty_nolog(struct reiserfs_transaction_handle *, struct super_block *, struct buffer_head *bh) ;
1574 int journal_mark_freed(struct reiserfs_transaction_handle *, struct super_block *, unsigned long blocknr) ;
1575 int push_journal_writer(char *w) ;
1576 int pop_journal_writer(int windex) ;
1577 int journal_transaction_should_end(struct reiserfs_transaction_handle *, int) ;
1578 int reiserfs_in_journal(struct super_block *p_s_sb, kdev_t dev, int bmap_nr, int bit_nr, int size, int searchall, unsigned int *next) ;
1579 int journal_begin(struct reiserfs_transaction_handle *, struct super_block *p_s_sb, unsigned long) ;
1580 struct super_block *reiserfs_get_super(kdev_t dev) ;
1581 void flush_async_commits(struct super_block *p_s_sb) ;
1583 int buffer_journaled(const struct buffer_head *bh) ;
1584 int mark_buffer_journal_new(struct buffer_head *bh) ;
1585 int reiserfs_sync_all_buffers(kdev_t dev, int wait) ;
1586 int reiserfs_sync_buffers(kdev_t dev, int wait) ;
1587 int reiserfs_add_page_to_flush_list(struct reiserfs_transaction_handle *,
1588 struct inode *, struct buffer_head *) ;
1589 int reiserfs_remove_page_from_flush_list(struct reiserfs_transaction_handle *,
1592 int reiserfs_allocate_list_bitmaps(struct super_block *s, struct reiserfs_list_bitmap *, int) ;
1594 /* why is this kerplunked right here? */
1595 static inline int reiserfs_buffer_prepared(const struct buffer_head *bh) {
1596 if (bh && test_bit(BH_JPrepared, &( (struct buffer_head *)bh)->b_state))
1602 /* buffer was journaled, waiting to get to disk */
1603 static inline int buffer_journal_dirty(const struct buffer_head *bh) {
1605 return test_bit(BH_JDirty_wait, &( (struct buffer_head *)bh)->b_state) ;
1609 static inline int mark_buffer_notjournal_dirty(struct buffer_head *bh) {
1611 clear_bit(BH_JDirty_wait, &bh->b_state) ;
1614 static inline int mark_buffer_notjournal_new(struct buffer_head *bh) {
1616 clear_bit(BH_JNew, &bh->b_state) ;
1621 void add_save_link (struct reiserfs_transaction_handle * th,
1622 struct inode * inode, int truncate);
1623 void remove_save_link (struct inode * inode, int truncate);
1626 __u32 reiserfs_get_unused_objectid (struct reiserfs_transaction_handle *th);
1627 void reiserfs_release_objectid (struct reiserfs_transaction_handle *th, __u32 objectid_to_release);
1628 int reiserfs_convert_objectid_map_v1(struct super_block *) ;
1631 int B_IS_IN_TREE(const struct buffer_head *);
1632 extern inline void copy_short_key (void * to, const void * from);
1633 extern inline void copy_item_head(struct item_head * p_v_to,
1634 const struct item_head * p_v_from);
1636 // first key is in cpu form, second - le
1637 extern inline int comp_keys (const struct key * le_key,
1638 const struct cpu_key * cpu_key);
1639 extern inline int comp_short_keys (const struct key * le_key,
1640 const struct cpu_key * cpu_key);
1641 extern inline void le_key2cpu_key (struct cpu_key * to, const struct key * from);
1643 // both are cpu keys
1644 extern inline int comp_cpu_keys (const struct cpu_key *, const struct cpu_key *);
1645 extern inline int comp_short_cpu_keys (const struct cpu_key *,
1646 const struct cpu_key *);
1647 extern inline void cpu_key2cpu_key (struct cpu_key *, const struct cpu_key *);
1649 // both are in le form
1650 extern inline int comp_le_keys (const struct key *, const struct key *);
1651 extern inline int comp_short_le_keys (const struct key *, const struct key *);
1654 // get key version from on disk key - kludge
1656 static inline int le_key_version (const struct key * key)
1660 type = offset_v2_k_type( &(key->u.k_offset_v2));
1661 if (type != TYPE_DIRECT && type != TYPE_INDIRECT && type != TYPE_DIRENTRY)
1662 return KEY_FORMAT_3_5;
1664 return KEY_FORMAT_3_6;
1669 static inline void copy_key (struct key *to, const struct key *from)
1671 memcpy (to, from, KEY_SIZE);
1675 int comp_items (const struct item_head * stored_ih, const struct path * p_s_path);
1676 const struct key * get_rkey (const struct path * p_s_chk_path,
1677 const struct super_block * p_s_sb);
1678 inline int bin_search (const void * p_v_key, const void * p_v_base,
1679 int p_n_num, int p_n_width, int * p_n_pos);
1680 int search_by_key (struct super_block *, const struct cpu_key *,
1681 struct path *, int);
1682 #define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
1683 int search_for_position_by_key (struct super_block * p_s_sb,
1684 const struct cpu_key * p_s_cpu_key,
1685 struct path * p_s_search_path);
1686 extern inline void decrement_bcount (struct buffer_head * p_s_bh);
1687 void decrement_counters_in_path (struct path * p_s_search_path);
1688 void pathrelse (struct path * p_s_search_path);
1689 int reiserfs_check_path(struct path *p) ;
1690 void pathrelse_and_restore (struct super_block *s, struct path * p_s_search_path);
1692 int reiserfs_insert_item (struct reiserfs_transaction_handle *th,
1694 const struct cpu_key * key,
1695 struct item_head * ih, const char * body);
1697 int reiserfs_paste_into_item (struct reiserfs_transaction_handle *th,
1699 const struct cpu_key * key,
1700 const char * body, int paste_size);
1702 int reiserfs_cut_from_item (struct reiserfs_transaction_handle *th,
1704 struct cpu_key * key,
1705 struct inode * inode,
1707 loff_t new_file_size);
1709 int reiserfs_delete_item (struct reiserfs_transaction_handle *th,
1711 const struct cpu_key * key,
1712 struct inode * inode,
1713 struct buffer_head * p_s_un_bh);
1715 void reiserfs_delete_solid_item (struct reiserfs_transaction_handle *th,
1717 void reiserfs_delete_object (struct reiserfs_transaction_handle *th, struct inode * p_s_inode);
1718 void reiserfs_do_truncate (struct reiserfs_transaction_handle *th,
1719 struct inode * p_s_inode, struct page *,
1720 int update_timestamps);
1722 #define block_size(inode) ((inode)->i_sb->s_blocksize)
1723 #define file_size(inode) ((inode)->i_size)
1724 #define tail_size(inode) (file_size (inode) & (block_size (inode) - 1))
1726 #define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
1727 !STORE_TAIL_IN_UNFM_S1(file_size (inode), tail_size(inode), block_size (inode)):have_small_tails ((inode)->i_sb)?!STORE_TAIL_IN_UNFM_S2(file_size (inode), tail_size(inode), block_size (inode)):0 )
1729 void padd_item (char * item, int total_length, int length);
1733 void reiserfs_read_inode (struct inode * inode) ;
1734 void reiserfs_read_inode2(struct inode * inode, void *p) ;
1735 void reiserfs_delete_inode (struct inode * inode);
1736 void reiserfs_write_inode (struct inode * inode, int) ;
1737 struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, __u32 *data,
1738 int len, int fhtype, int parent);
1739 int reiserfs_dentry_to_fh(struct dentry *dentry, __u32 *data, int *lenp, int need_parent);
1741 int reiserfs_prepare_write(struct file *, struct page *, unsigned, unsigned) ;
1742 void reiserfs_truncate_file(struct inode *, int update_timestamps) ;
1743 void make_cpu_key (struct cpu_key * cpu_key, const struct inode * inode, loff_t offset,
1744 int type, int key_length);
1745 void make_le_item_head (struct item_head * ih, const struct cpu_key * key,
1747 loff_t offset, int type, int length, int entry_count);
1748 struct inode * reiserfs_iget (struct super_block * s,
1749 const struct cpu_key * key);
1751 int reiserfs_new_inode (struct reiserfs_transaction_handle *th,
1752 struct inode * dir, int mode,
1753 const char * symname,
1755 struct dentry *dentry,
1756 struct inode *inode);
1757 int reiserfs_sync_inode (struct reiserfs_transaction_handle *th, struct inode * inode);
1758 void reiserfs_update_sd (struct reiserfs_transaction_handle *th, struct inode * inode);
1760 void sd_attrs_to_i_attrs( __u16 sd_attrs, struct inode *inode );
1761 void i_attrs_to_sd_attrs( struct inode *inode, __u16 *sd_attrs );
1764 inline void set_de_name_and_namelen (struct reiserfs_dir_entry * de);
1765 int search_by_entry_key (struct super_block * sb, const struct cpu_key * key,
1767 struct reiserfs_dir_entry * de);
1770 #if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )
1771 #define REISERFS_PROC_INFO
1773 #undef REISERFS_PROC_INFO
1776 int reiserfs_proc_info_init( struct super_block *sb );
1777 int reiserfs_proc_info_done( struct super_block *sb );
1778 struct proc_dir_entry *reiserfs_proc_register( struct super_block *sb,
1779 char *name, read_proc_t *func );
1780 void reiserfs_proc_unregister( struct super_block *sb, const char *name );
1781 struct proc_dir_entry *reiserfs_proc_register_global( char *name,
1782 read_proc_t *func );
1783 void reiserfs_proc_unregister_global( const char *name );
1784 int reiserfs_proc_info_global_init( void );
1785 int reiserfs_proc_info_global_done( void );
1786 int reiserfs_proc_tail( int len, char *buffer, char **start,
1787 off_t offset, int count, int *eof );
1788 int reiserfs_global_version_in_proc( char *buffer, char **start, off_t offset,
1789 int count, int *eof, void *data );
1790 int reiserfs_version_in_proc( char *buffer, char **start, off_t offset,
1791 int count, int *eof, void *data );
1792 int reiserfs_super_in_proc( char *buffer, char **start, off_t offset,
1793 int count, int *eof, void *data );
1794 int reiserfs_per_level_in_proc( char *buffer, char **start, off_t offset,
1795 int count, int *eof, void *data );
1796 int reiserfs_bitmap_in_proc( char *buffer, char **start, off_t offset,
1797 int count, int *eof, void *data );
1798 int reiserfs_on_disk_super_in_proc( char *buffer, char **start, off_t offset,
1799 int count, int *eof, void *data );
1800 int reiserfs_oidmap_in_proc( char *buffer, char **start, off_t offset,
1801 int count, int *eof, void *data );
1802 int reiserfs_journal_in_proc( char *buffer, char **start, off_t offset,
1803 int count, int *eof, void *data );
1805 #if defined( REISERFS_PROC_INFO )
1807 #define PROC_EXP( e ) e
1809 #define MAX( a, b ) ( ( ( a ) > ( b ) ) ? ( a ) : ( b ) )
1810 #define __PINFO( sb ) ( sb ) -> u.reiserfs_sb.s_proc_info_data
1811 #define PROC_INFO_MAX( sb, field, value ) \
1812 __PINFO( sb ).field = \
1813 MAX( ( sb ) -> u.reiserfs_sb.s_proc_info_data.field, value )
1814 #define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
1815 #define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
1816 #define PROC_INFO_BH_STAT( sb, bh, level ) \
1817 PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \
1818 PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \
1819 PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
1821 #define PROC_EXP( e )
1822 #define VOID_V ( ( void ) 0 )
1823 #define PROC_INFO_MAX( sb, field, value ) VOID_V
1824 #define PROC_INFO_INC( sb, field ) VOID_V
1825 #define PROC_INFO_ADD( sb, field, val ) VOID_V
1826 #define PROC_INFO_BH_STAT( p_s_sb, p_s_bh, n_node_level ) VOID_V
1830 extern struct inode_operations reiserfs_dir_inode_operations;
1831 extern struct file_operations reiserfs_dir_operations;
1833 /* tail_conversion.c */
1834 int direct2indirect (struct reiserfs_transaction_handle *, struct inode *, struct path *, struct buffer_head *, loff_t);
1835 int indirect2direct (struct reiserfs_transaction_handle *, struct inode *, struct page *, struct path *, const struct cpu_key *, loff_t, char *);
1836 void reiserfs_unmap_buffer(struct buffer_head *) ;
1840 extern struct inode_operations reiserfs_file_inode_operations;
1841 extern struct file_operations reiserfs_file_operations;
1842 extern struct address_space_operations reiserfs_address_space_operations ;
1843 int get_new_buffer (struct reiserfs_transaction_handle *th, struct buffer_head *,
1844 struct buffer_head **, struct path *);
1848 struct buffer_head * reiserfs_getblk (kdev_t n_dev, int n_block, int n_size);
1849 void wait_buffer_until_released (const struct buffer_head * bh);
1850 struct buffer_head * reiserfs_bread (struct super_block *super, int n_block,
1854 #ifdef CONFIG_REISERFS_CHECK
1855 void * reiserfs_kmalloc (size_t size, int flags, struct super_block * s);
1856 void reiserfs_kfree (const void * vp, size_t size, struct super_block * s);
1858 #define reiserfs_kmalloc(x, y, z) kmalloc(x, y)
1859 #define reiserfs_kfree(x, y, z) kfree(x)
1862 int fix_nodes (int n_op_mode, struct tree_balance * p_s_tb,
1863 struct item_head * p_s_ins_ih, const void *);
1864 void unfix_nodes (struct tree_balance *);
1865 void free_buffers_in_tb (struct tree_balance * p_s_tb);
1869 void reiserfs_panic (struct super_block * s, const char * fmt, ...)
1870 __attribute__ ( ( noreturn ) );/* __attribute__( ( format ( printf, 2, 3 ) ) ) */
1871 void reiserfs_debug (struct super_block *s, int level, const char * fmt, ...);
1872 /* __attribute__( ( format ( printf, 3, 4 ) ) ); */
1873 void print_virtual_node (struct virtual_node * vn);
1874 void print_indirect_item (struct buffer_head * bh, int item_num);
1875 void store_print_tb (struct tree_balance * tb);
1876 void print_cur_tb (char * mes);
1877 void print_de (struct reiserfs_dir_entry * de);
1878 void print_bi (struct buffer_info * bi, char * mes);
1879 #define PRINT_LEAF_ITEMS 1 /* print all items */
1880 #define PRINT_DIRECTORY_ITEMS 2 /* print directory items */
1881 #define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */
1882 void print_block (struct buffer_head * bh, ...);
1883 void print_path (struct tree_balance * tb, struct path * path);
1884 void print_bmap (struct super_block * s, int silent);
1885 void print_bmap_block (int i, char * data, int size, int silent);
1886 /*void print_super_block (struct super_block * s, char * mes);*/
1887 void print_objectid_map (struct super_block * s);
1888 void print_block_head (struct buffer_head * bh, char * mes);
1889 void check_leaf (struct buffer_head * bh);
1890 void check_internal (struct buffer_head * bh);
1891 void print_statistics (struct super_block * s);
1892 char * reiserfs_hashname(int code);
1895 int leaf_move_items (int shift_mode, struct tree_balance * tb, int mov_num, int mov_bytes, struct buffer_head * Snew);
1896 int leaf_shift_left (struct tree_balance * tb, int shift_num, int shift_bytes);
1897 int leaf_shift_right (struct tree_balance * tb, int shift_num, int shift_bytes);
1898 void leaf_delete_items (struct buffer_info * cur_bi, int last_first, int first, int del_num, int del_bytes);
1899 void leaf_insert_into_buf (struct buffer_info * bi, int before,
1900 struct item_head * inserted_item_ih, const char * inserted_item_body, int zeros_number);
1901 void leaf_paste_in_buffer (struct buffer_info * bi, int pasted_item_num,
1902 int pos_in_item, int paste_size, const char * body, int zeros_number);
1903 void leaf_cut_from_buffer (struct buffer_info * bi, int cut_item_num, int pos_in_item,
1905 void leaf_paste_entries (struct buffer_head * bh, int item_num, int before,
1906 int new_entry_count, struct reiserfs_de_head * new_dehs, const char * records, int paste_size);
1908 int balance_internal (struct tree_balance * , int, int, struct item_head * ,
1909 struct buffer_head **);
1912 inline void do_balance_mark_leaf_dirty (struct tree_balance * tb,
1913 struct buffer_head * bh, int flag);
1914 #define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
1915 #define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
1917 void do_balance (struct tree_balance * tb, struct item_head * ih,
1918 const char * body, int flag);
1919 void reiserfs_invalidate_buffer (struct tree_balance * tb, struct buffer_head * bh);
1921 int get_left_neighbor_position (struct tree_balance * tb, int h);
1922 int get_right_neighbor_position (struct tree_balance * tb, int h);
1923 void replace_key (struct tree_balance * tb, struct buffer_head *, int, struct buffer_head *, int);
1924 void replace_lkey (struct tree_balance *, int, struct item_head *);
1925 void replace_rkey (struct tree_balance *, int, struct item_head *);
1926 void make_empty_node (struct buffer_info *);
1927 struct buffer_head * get_FEB (struct tree_balance *);
1931 /* structure contains hints for block allocator, and it is a container for
1932 * arguments, such as node, search path, transaction_handle, etc. */
1933 struct __reiserfs_blocknr_hint {
1934 struct inode * inode; /* inode passed to allocator, if we allocate unf. nodes */
1935 long block; /* file offset, in blocks */
1937 struct path * path; /* search path, used by allocator to deternine search_start by
1939 struct reiserfs_transaction_handle * th; /* transaction handle is needed to log super blocks and
1940 * bitmap blocks changes */
1941 b_blocknr_t beg, end;
1942 b_blocknr_t search_start; /* a field used to transfer search start value (block number)
1943 * between different block allocator procedures
1944 * (determine_search_start() and others) */
1945 int prealloc_size; /* is set in determine_prealloc_size() function, used by underlayed
1946 * function that do actual allocation */
1948 int formatted_node:1; /* the allocator uses different polices for getting disk space for
1949 * formatted/unformatted blocks with/without preallocation */
1953 typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
1955 int reiserfs_parse_alloc_options (struct super_block *, char *);
1956 int is_reusable (struct super_block * s, unsigned long block, int bit_value);
1957 void reiserfs_free_block (struct reiserfs_transaction_handle *th, unsigned long);
1958 int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t * , int, int);
1959 extern inline int reiserfs_new_form_blocknrs (struct tree_balance * tb,
1960 b_blocknr_t *new_blocknrs, int amount_needed)
1962 reiserfs_blocknr_hint_t hint = {
1963 th:tb->transaction_handle,
1970 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed, 0);
1973 extern inline int reiserfs_new_unf_blocknrs (struct reiserfs_transaction_handle *th,
1974 struct inode *inode,
1975 b_blocknr_t *new_blocknrs,
1976 struct path * path, long block)
1978 reiserfs_blocknr_hint_t hint = {
1986 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
1989 #ifdef REISERFS_PREALLOCATE
1990 extern inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle *th,
1991 struct inode * inode,
1992 b_blocknr_t *new_blocknrs,
1993 struct path * path, long block)
1995 reiserfs_blocknr_hint_t hint = {
2003 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
2006 void reiserfs_discard_prealloc (struct reiserfs_transaction_handle *th,
2007 struct inode * inode);
2008 void reiserfs_discard_all_prealloc (struct reiserfs_transaction_handle *th);
2010 void reiserfs_claim_blocks_to_be_allocated( struct super_block *sb, int blocks);
2011 void reiserfs_release_claimed_blocks( struct super_block *sb, int blocks);
2014 __u32 keyed_hash (const signed char *msg, int len);
2015 __u32 yura_hash (const signed char *msg, int len);
2016 __u32 r5_hash (const signed char *msg, int len);
2019 const char *reiserfs_get_version_string(void) CONSTF;
2021 /* the ext2 bit routines adjust for big or little endian as
2022 ** appropriate for the arch, so in our laziness we use them rather
2023 ** than using the bit routines they call more directly. These
2024 ** routines must be used when changing on disk bitmaps. */
2025 #define reiserfs_test_and_set_le_bit ext2_set_bit
2026 #define reiserfs_test_and_clear_le_bit ext2_clear_bit
2027 #define reiserfs_test_le_bit ext2_test_bit
2028 #define reiserfs_find_next_zero_le_bit ext2_find_next_zero_bit
2030 /* sometimes reiserfs_truncate may require to allocate few new blocks
2031 to perform indirect2direct conversion. People probably used to
2032 think, that truncate should work without problems on a filesystem
2033 without free disk space. They may complain that they can not
2034 truncate due to lack of free disk space. This spare space allows us
2035 to not worry about it. 500 is probably too much, but it should be
2037 #define SPARE_SPACE 500
2039 static inline unsigned long reiserfs_get_journal_block(const struct super_block *s) {
2040 return le32_to_cpu(SB_DISK_SUPER_BLOCK(s)->s_journal_block) ;
2042 static inline unsigned long reiserfs_get_journal_orig_size(const struct super_block *s) {
2043 return le32_to_cpu(SB_DISK_SUPER_BLOCK(s)->s_orig_journal_size) ;
2046 /* prototypes from ioctl.c */
2047 int reiserfs_ioctl (struct inode * inode, struct file * filp,
2048 unsigned int cmd, unsigned long arg);
2049 int reiserfs_unpack (struct inode * inode, struct file * filp);
2051 /* ioctl's command */
2052 #define REISERFS_IOC_UNPACK _IOW(0xCD,1,long)
2053 /* define following flags to be the same as in ext2, so that chattr(1),
2054 lsattr(1) will work with us. */
2055 #define REISERFS_IOC_GETFLAGS EXT2_IOC_GETFLAGS
2056 #define REISERFS_IOC_SETFLAGS EXT2_IOC_SETFLAGS
2057 #define REISERFS_IOC_GETVERSION EXT2_IOC_GETVERSION
2058 #define REISERFS_IOC_SETVERSION EXT2_IOC_SETVERSION
2060 #endif /* _LINUX_REISER_FS_H */