namei.c 44.2 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
/*
 * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
 *
 * Trivial changes by Alan Cox to remove EHASHCOLLISION for compatibility
 *
 * Trivial Changes:
 * Rights granted to Hans Reiser to redistribute under other terms providing
 * he accepts all liability including but not limited to patent, fitness
 * for purpose, and direct or indirect claims arising from failure to perform.
 *
 * NO WARRANTY
 */

#include <linux/time.h>
#include <linux/bitops.h>
#include <linux/reiserfs_fs.h>
#include <linux/reiserfs_acl.h>
#include <linux/reiserfs_xattr.h>
#include <linux/quotaops.h>

21
#define INC_DIR_INODE_NLINK(i) if (i->i_nlink != 1) { inc_nlink(i); if (i->i_nlink >= REISERFS_LINK_MAX) i->i_nlink=1; }
22
#define DEC_DIR_INODE_NLINK(i) if (i->i_nlink != 1) drop_nlink(i);
L
Linus Torvalds 已提交
23 24 25

// directory item contains array of entry headers. This performs
// binary search through that array
26
static int bin_search_in_dir_item(struct reiserfs_dir_entry *de, loff_t off)
L
Linus Torvalds 已提交
27
{
28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
	struct item_head *ih = de->de_ih;
	struct reiserfs_de_head *deh = de->de_deh;
	int rbound, lbound, j;

	lbound = 0;
	rbound = I_ENTRY_COUNT(ih) - 1;

	for (j = (rbound + lbound) / 2; lbound <= rbound;
	     j = (rbound + lbound) / 2) {
		if (off < deh_offset(deh + j)) {
			rbound = j - 1;
			continue;
		}
		if (off > deh_offset(deh + j)) {
			lbound = j + 1;
			continue;
		}
		// this is not name found, but matched third key component
		de->de_entry_num = j;
		return NAME_FOUND;
L
Linus Torvalds 已提交
48 49
	}

50 51
	de->de_entry_num = lbound;
	return NAME_NOT_FOUND;
L
Linus Torvalds 已提交
52 53 54
}

// comment?  maybe something like set de to point to what the path points to?
55
static inline void set_de_item_location(struct reiserfs_dir_entry *de,
56
					struct treepath *path)
L
Linus Torvalds 已提交
57
{
58 59 60 61 62
	de->de_bh = get_last_bh(path);
	de->de_ih = get_ih(path);
	de->de_deh = B_I_DEH(de->de_bh, de->de_ih);
	de->de_item_num = PATH_LAST_POSITION(path);
}
L
Linus Torvalds 已提交
63 64

// de_bh, de_ih, de_deh (points to first element of array), de_item_num is set
65
inline void set_de_name_and_namelen(struct reiserfs_dir_entry *de)
L
Linus Torvalds 已提交
66
{
67
	struct reiserfs_de_head *deh = de->de_deh + de->de_entry_num;
L
Linus Torvalds 已提交
68

69
	BUG_ON(de->de_entry_num >= ih_entry_count(de->de_ih));
L
Linus Torvalds 已提交
70

71 72 73 74 75
	de->de_entrylen = entry_length(de->de_bh, de->de_ih, de->de_entry_num);
	de->de_namelen = de->de_entrylen - (de_with_sd(deh) ? SD_SIZE : 0);
	de->de_name = B_I_PITEM(de->de_bh, de->de_ih) + deh_location(deh);
	if (de->de_name[de->de_namelen - 1] == 0)
		de->de_namelen = strlen(de->de_name);
L
Linus Torvalds 已提交
76 77 78
}

// what entry points to
79
static inline void set_de_object_key(struct reiserfs_dir_entry *de)
L
Linus Torvalds 已提交
80
{
81
	BUG_ON(de->de_entry_num >= ih_entry_count(de->de_ih));
82 83
	de->de_dir_id = deh_dir_id(&(de->de_deh[de->de_entry_num]));
	de->de_objectid = deh_objectid(&(de->de_deh[de->de_entry_num]));
L
Linus Torvalds 已提交
84 85
}

86
static inline void store_de_entry_key(struct reiserfs_dir_entry *de)
L
Linus Torvalds 已提交
87
{
88 89
	struct reiserfs_de_head *deh = de->de_deh + de->de_entry_num;

90
	BUG_ON(de->de_entry_num >= ih_entry_count(de->de_ih));
91 92 93 94 95 96 97 98 99

	/* store key of the found entry */
	de->de_entry_key.version = KEY_FORMAT_3_5;
	de->de_entry_key.on_disk_key.k_dir_id =
	    le32_to_cpu(de->de_ih->ih_key.k_dir_id);
	de->de_entry_key.on_disk_key.k_objectid =
	    le32_to_cpu(de->de_ih->ih_key.k_objectid);
	set_cpu_key_k_offset(&(de->de_entry_key), deh_offset(deh));
	set_cpu_key_k_type(&(de->de_entry_key), TYPE_DIRENTRY);
L
Linus Torvalds 已提交
100 101 102 103 104 105 106 107 108
}

/* We assign a key to each directory item, and place multiple entries
in a single directory item.  A directory item has a key equal to the
key of the first directory entry in it.

This function first calls search_by_key, then, if item whose first
entry matches is not found it looks for the entry inside directory
item found by search_by_key. Fills the path to the entry, and to the
109
entry position in the item
L
Linus Torvalds 已提交
110 111 112 113

*/

/* The function is NOT SCHEDULE-SAFE! */
114
int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,
115
			struct treepath *path, struct reiserfs_dir_entry *de)
L
Linus Torvalds 已提交
116
{
117 118 119 120 121 122
	int retval;

	retval = search_item(sb, key, path);
	switch (retval) {
	case ITEM_NOT_FOUND:
		if (!PATH_LAST_POSITION(path)) {
J
Jeff Mahoney 已提交
123 124
			reiserfs_error(sb, "vs-7000", "search_by_key "
				       "returned item position == 0");
125 126 127 128 129 130 131 132 133 134 135 136 137
			pathrelse(path);
			return IO_ERROR;
		}
		PATH_LAST_POSITION(path)--;

	case ITEM_FOUND:
		break;

	case IO_ERROR:
		return retval;

	default:
		pathrelse(path);
J
Jeff Mahoney 已提交
138
		reiserfs_error(sb, "vs-7002", "no path to here");
139
		return IO_ERROR;
L
Linus Torvalds 已提交
140 141
	}

142
	set_de_item_location(de, path);
L
Linus Torvalds 已提交
143 144

#ifdef CONFIG_REISERFS_CHECK
145 146 147
	if (!is_direntry_le_ih(de->de_ih) ||
	    COMP_SHORT_KEYS(&(de->de_ih->ih_key), key)) {
		print_block(de->de_bh, 0, -1, -1);
J
Jeff Mahoney 已提交
148 149 150
		reiserfs_panic(sb, "vs-7005", "found item %h is not directory "
			       "item or does not belong to the same directory "
			       "as key %K", de->de_ih, key);
151 152 153 154 155 156 157 158 159 160 161 162 163
	}
#endif				/* CONFIG_REISERFS_CHECK */

	/* binary search in directory item by third componen t of the
	   key. sets de->de_entry_num of de */
	retval = bin_search_in_dir_item(de, cpu_key_k_offset(key));
	path->pos_in_item = de->de_entry_num;
	if (retval != NAME_NOT_FOUND) {
		// ugly, but rename needs de_bh, de_deh, de_name, de_namelen, de_objectid set
		set_de_name_and_namelen(de);
		set_de_object_key(de);
	}
	return retval;
L
Linus Torvalds 已提交
164 165 166 167 168 169 170 171 172
}

/* Keyed 32-bit hash function using TEA in a Davis-Meyer function */

/* The third component is hashed, and you can choose from more than
   one hash function.  Per directory hashes are not yet implemented
   but are thought about. This function should be moved to hashes.c
   Jedi, please do so.  -Hans */

173 174
static __u32 get_third_component(struct super_block *s,
				 const char *name, int len)
L
Linus Torvalds 已提交
175
{
176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
	__u32 res;

	if (!len || (len == 1 && name[0] == '.'))
		return DOT_OFFSET;
	if (len == 2 && name[0] == '.' && name[1] == '.')
		return DOT_DOT_OFFSET;

	res = REISERFS_SB(s)->s_hash_function(name, len);

	// take bits from 7-th to 30-th including both bounds
	res = GET_HASH_VALUE(res);
	if (res == 0)
		// needed to have no names before "." and ".." those have hash
		// value == 0 and generation conters 1 and 2 accordingly
		res = 128;
	return res + MAX_GENERATION_NUMBER;
L
Linus Torvalds 已提交
192 193
}

194 195
static int reiserfs_match(struct reiserfs_dir_entry *de,
			  const char *name, int namelen)
L
Linus Torvalds 已提交
196
{
197
	int retval = NAME_NOT_FOUND;
L
Linus Torvalds 已提交
198

199 200 201 202 203
	if ((namelen == de->de_namelen) &&
	    !memcmp(de->de_name, name, de->de_namelen))
		retval =
		    (de_visible(de->de_deh + de->de_entry_num) ? NAME_FOUND :
		     NAME_FOUND_INVISIBLE);
L
Linus Torvalds 已提交
204

205
	return retval;
L
Linus Torvalds 已提交
206 207 208 209 210 211
}

/* de's de_bh, de_ih, de_deh, de_item_num, de_entry_num are set already */

				/* used when hash collisions exist */

212 213 214
static int linear_search_in_dir_item(struct cpu_key *key,
				     struct reiserfs_dir_entry *de,
				     const char *name, int namelen)
L
Linus Torvalds 已提交
215
{
216 217 218
	struct reiserfs_de_head *deh = de->de_deh;
	int retval;
	int i;
L
Linus Torvalds 已提交
219

220
	i = de->de_entry_num;
L
Linus Torvalds 已提交
221

222 223 224 225 226
	if (i == I_ENTRY_COUNT(de->de_ih) ||
	    GET_HASH_VALUE(deh_offset(deh + i)) !=
	    GET_HASH_VALUE(cpu_key_k_offset(key))) {
		i--;
	}
L
Linus Torvalds 已提交
227

228 229
	RFALSE(de->de_deh != B_I_DEH(de->de_bh, de->de_ih),
	       "vs-7010: array of entry headers not found");
L
Linus Torvalds 已提交
230

231
	deh += i;
L
Linus Torvalds 已提交
232

233 234 235 236 237 238 239 240 241 242
	for (; i >= 0; i--, deh--) {
		if (GET_HASH_VALUE(deh_offset(deh)) !=
		    GET_HASH_VALUE(cpu_key_k_offset(key))) {
			// hash value does not match, no need to check whole name
			return NAME_NOT_FOUND;
		}

		/* mark, that this generation number is used */
		if (de->de_gen_number_bit_string)
			set_bit(GET_GENERATION_NUMBER(deh_offset(deh)),
243
				de->de_gen_number_bit_string);
L
Linus Torvalds 已提交
244

245 246 247
		// calculate pointer to name and namelen
		de->de_entry_num = i;
		set_de_name_and_namelen(de);
L
Linus Torvalds 已提交
248

249 250 251
		if ((retval =
		     reiserfs_match(de, name, namelen)) != NAME_NOT_FOUND) {
			// de's de_name, de_namelen, de_recordlen are set. Fill the rest:
L
Linus Torvalds 已提交
252

253 254
			// key of pointed object
			set_de_object_key(de);
L
Linus Torvalds 已提交
255

256
			store_de_entry_key(de);
L
Linus Torvalds 已提交
257

258 259 260
			// retval can be NAME_FOUND or NAME_FOUND_INVISIBLE
			return retval;
		}
L
Linus Torvalds 已提交
261 262
	}

263 264 265 266 267 268 269 270 271
	if (GET_GENERATION_NUMBER(le_ih_k_offset(de->de_ih)) == 0)
		/* we have reached left most entry in the node. In common we
		   have to go to the left neighbor, but if generation counter
		   is 0 already, we know for sure, that there is no name with
		   the same hash value */
		// FIXME: this work correctly only because hash value can not
		// be 0. Btw, in case of Yura's hash it is probably possible,
		// so, this is a bug
		return NAME_NOT_FOUND;
L
Linus Torvalds 已提交
272

273 274
	RFALSE(de->de_item_num,
	       "vs-7015: two diritems of the same directory in one node?");
L
Linus Torvalds 已提交
275

276 277
	return GOTO_PREVIOUS_ITEM;
}
L
Linus Torvalds 已提交
278 279 280

// may return NAME_FOUND, NAME_FOUND_INVISIBLE, NAME_NOT_FOUND
// FIXME: should add something like IOERROR
281
static int reiserfs_find_entry(struct inode *dir, const char *name, int namelen,
282
			       struct treepath *path_to_entry,
283
			       struct reiserfs_dir_entry *de)
L
Linus Torvalds 已提交
284
{
285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300
	struct cpu_key key_to_search;
	int retval;

	if (namelen > REISERFS_MAX_NAME(dir->i_sb->s_blocksize))
		return NAME_NOT_FOUND;

	/* we will search for this key in the tree */
	make_cpu_key(&key_to_search, dir,
		     get_third_component(dir->i_sb, name, namelen),
		     TYPE_DIRENTRY, 3);

	while (1) {
		retval =
		    search_by_entry_key(dir->i_sb, &key_to_search,
					path_to_entry, de);
		if (retval == IO_ERROR) {
J
Jeff Mahoney 已提交
301
			reiserfs_error(dir->i_sb, "zam-7001", "io error");
302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320
			return IO_ERROR;
		}

		/* compare names for all entries having given hash value */
		retval =
		    linear_search_in_dir_item(&key_to_search, de, name,
					      namelen);
		if (retval != GOTO_PREVIOUS_ITEM) {
			/* there is no need to scan directory anymore. Given entry found or does not exist */
			path_to_entry->pos_in_item = de->de_entry_num;
			return retval;
		}

		/* there is left neighboring item of this directory and given entry can be there */
		set_cpu_key_k_offset(&key_to_search,
				     le_ih_k_offset(de->de_ih) - 1);
		pathrelse(path_to_entry);

	}			/* while (1) */
L
Linus Torvalds 已提交
321 322
}

323 324
static struct dentry *reiserfs_lookup(struct inode *dir, struct dentry *dentry,
				      struct nameidata *nd)
L
Linus Torvalds 已提交
325
{
326
	int retval;
327
	int lock_depth;
328 329 330 331 332 333 334
	struct inode *inode = NULL;
	struct reiserfs_dir_entry de;
	INITIALIZE_PATH(path_to_entry);

	if (REISERFS_MAX_NAME(dir->i_sb->s_blocksize) < dentry->d_name.len)
		return ERR_PTR(-ENAMETOOLONG);

335 336 337 338 339 340 341
	/*
	 * Might be called with or without the write lock, must be careful
	 * to not recursively hold it in case we want to release the lock
	 * before rescheduling.
	 */
	lock_depth = reiserfs_write_lock_once(dir->i_sb);

342 343 344 345 346 347
	de.de_gen_number_bit_string = NULL;
	retval =
	    reiserfs_find_entry(dir, dentry->d_name.name, dentry->d_name.len,
				&path_to_entry, &de);
	pathrelse(&path_to_entry);
	if (retval == NAME_FOUND) {
348 349
		inode = reiserfs_iget(dir->i_sb,
				      (struct cpu_key *)&(de.de_dir_id));
350
		if (!inode || IS_ERR(inode)) {
351
			reiserfs_write_unlock_once(dir->i_sb, lock_depth);
352 353 354
			return ERR_PTR(-EACCES);
		}

355 356 357 358
		/* Propagate the private flag so we know we're
		 * in the priv tree */
		if (IS_PRIVATE(dir))
			inode->i_flags |= S_PRIVATE;
359
	}
360
	reiserfs_write_unlock_once(dir->i_sb, lock_depth);
361 362
	if (retval == IO_ERROR) {
		return ERR_PTR(-EIO);
L
Linus Torvalds 已提交
363 364
	}

365
	return d_splice_alias(inode, dentry);
366
}
L
Linus Torvalds 已提交
367

368
/*
L
Linus Torvalds 已提交
369 370 371 372 373
** looks up the dentry of the parent directory for child.
** taken from ext2_get_parent
*/
struct dentry *reiserfs_get_parent(struct dentry *child)
{
374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
	int retval;
	struct inode *inode = NULL;
	struct reiserfs_dir_entry de;
	INITIALIZE_PATH(path_to_entry);
	struct inode *dir = child->d_inode;

	if (dir->i_nlink == 0) {
		return ERR_PTR(-ENOENT);
	}
	de.de_gen_number_bit_string = NULL;

	reiserfs_write_lock(dir->i_sb);
	retval = reiserfs_find_entry(dir, "..", 2, &path_to_entry, &de);
	pathrelse(&path_to_entry);
	if (retval != NAME_FOUND) {
		reiserfs_write_unlock(dir->i_sb);
		return ERR_PTR(-ENOENT);
	}
	inode = reiserfs_iget(dir->i_sb, (struct cpu_key *)&(de.de_dir_id));
L
Linus Torvalds 已提交
393 394
	reiserfs_write_unlock(dir->i_sb);

395
	return d_obtain_alias(inode);
396
}
L
Linus Torvalds 已提交
397

398
/* add entry to the directory (entry can be hidden).
L
Linus Torvalds 已提交
399 400 401 402 403

insert definition of when hidden directories are used here -Hans

 Does not mark dir   inode dirty, do it after successesfull call to it */

404 405 406
static int reiserfs_add_entry(struct reiserfs_transaction_handle *th,
			      struct inode *dir, const char *name, int namelen,
			      struct inode *inode, int visible)
L
Linus Torvalds 已提交
407
{
408 409 410 411
	struct cpu_key entry_key;
	struct reiserfs_de_head *deh;
	INITIALIZE_PATH(path);
	struct reiserfs_dir_entry de;
412
	DECLARE_BITMAP(bit_string, MAX_GENERATION_NUMBER + 1);
413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436
	int gen_number;
	char small_buf[32 + DEH_SIZE];	/* 48 bytes now and we avoid kmalloc
					   if we create file with short name */
	char *buffer;
	int buflen, paste_size;
	int retval;

	BUG_ON(!th->t_trans_id);

	/* cannot allow items to be added into a busy deleted directory */
	if (!namelen)
		return -EINVAL;

	if (namelen > REISERFS_MAX_NAME(dir->i_sb->s_blocksize))
		return -ENAMETOOLONG;

	/* each entry has unique key. compose it */
	make_cpu_key(&entry_key, dir,
		     get_third_component(dir->i_sb, name, namelen),
		     TYPE_DIRENTRY, 3);

	/* get memory for composing the entry */
	buflen = DEH_SIZE + ROUND_UP(namelen);
	if (buflen > sizeof(small_buf)) {
437
		buffer = kmalloc(buflen, GFP_NOFS);
A
Al Viro 已提交
438
		if (!buffer)
439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466
			return -ENOMEM;
	} else
		buffer = small_buf;

	paste_size =
	    (get_inode_sd_version(dir) ==
	     STAT_DATA_V1) ? (DEH_SIZE + namelen) : buflen;

	/* fill buffer : directory entry head, name[, dir objectid | , stat data | ,stat data, dir objectid ] */
	deh = (struct reiserfs_de_head *)buffer;
	deh->deh_location = 0;	/* JDM Endian safe if 0 */
	put_deh_offset(deh, cpu_key_k_offset(&entry_key));
	deh->deh_state = 0;	/* JDM Endian safe if 0 */
	/* put key (ino analog) to de */
	deh->deh_dir_id = INODE_PKEY(inode)->k_dir_id;	/* safe: k_dir_id is le */
	deh->deh_objectid = INODE_PKEY(inode)->k_objectid;	/* safe: k_objectid is le */

	/* copy name */
	memcpy((char *)(deh + 1), name, namelen);
	/* padd by 0s to the 4 byte boundary */
	padd_item((char *)(deh + 1), ROUND_UP(namelen), namelen);

	/* entry is ready to be pasted into tree, set 'visibility' and 'stat data in entry' attributes */
	mark_de_without_sd(deh);
	visible ? mark_de_visible(deh) : mark_de_hidden(deh);

	/* find the proper place for the new entry */
	memset(bit_string, 0, sizeof(bit_string));
467
	de.de_gen_number_bit_string = bit_string;
468 469 470
	retval = reiserfs_find_entry(dir, name, namelen, &path, &de);
	if (retval != NAME_NOT_FOUND) {
		if (buffer != small_buf)
471
			kfree(buffer);
472 473 474 475 476 477 478
		pathrelse(&path);

		if (retval == IO_ERROR) {
			return -EIO;
		}

		if (retval != NAME_FOUND) {
J
Jeff Mahoney 已提交
479 480 481
			reiserfs_error(dir->i_sb, "zam-7002",
				       "reiserfs_find_entry() returned "
				       "unexpected value (%d)", retval);
482 483 484 485
		}

		return -EEXIST;
	}
L
Linus Torvalds 已提交
486

487
	gen_number =
488
	    find_first_zero_bit(bit_string,
489 490 491
				MAX_GENERATION_NUMBER + 1);
	if (gen_number > MAX_GENERATION_NUMBER) {
		/* there is no free generation number */
J
Jeff Mahoney 已提交
492 493 494
		reiserfs_warning(dir->i_sb, "reiserfs-7010",
				 "Congratulations! we have got hash function "
				 "screwed up");
495
		if (buffer != small_buf)
496
			kfree(buffer);
497 498 499 500 501 502 503 504 505 506 507 508 509
		pathrelse(&path);
		return -EBUSY;
	}
	/* adjust offset of directory enrty */
	put_deh_offset(deh, SET_GENERATION_NUMBER(deh_offset(deh), gen_number));
	set_cpu_key_k_offset(&entry_key, deh_offset(deh));

	/* update max-hash-collisions counter in reiserfs_sb_info */
	PROC_INFO_MAX(th->t_super, max_hash_collisions, gen_number);

	if (gen_number != 0) {	/* we need to re-search for the insertion point */
		if (search_by_entry_key(dir->i_sb, &entry_key, &path, &de) !=
		    NAME_NOT_FOUND) {
J
Jeff Mahoney 已提交
510 511 512
			reiserfs_warning(dir->i_sb, "vs-7032",
					 "entry with this key (%K) already "
					 "exists", &entry_key);
513 514

			if (buffer != small_buf)
515
				kfree(buffer);
516 517 518
			pathrelse(&path);
			return -EBUSY;
		}
L
Linus Torvalds 已提交
519 520
	}

521 522 523 524 525
	/* perform the insertion of the entry that we have prepared */
	retval =
	    reiserfs_paste_into_item(th, &path, &entry_key, dir, buffer,
				     paste_size);
	if (buffer != small_buf)
526
		kfree(buffer);
527 528 529
	if (retval) {
		reiserfs_check_path(&path);
		return retval;
L
Linus Torvalds 已提交
530 531
	}

532 533 534 535 536
	dir->i_size += paste_size;
	dir->i_mtime = dir->i_ctime = CURRENT_TIME_SEC;
	if (!S_ISDIR(inode->i_mode) && visible)
		// reiserfs_mkdir or reiserfs_rename will do that by itself
		reiserfs_update_sd(th, dir);
L
Linus Torvalds 已提交
537

538 539
	reiserfs_check_path(&path);
	return 0;
L
Linus Torvalds 已提交
540 541 542 543 544 545 546
}

/* quota utility function, call if you've had to abort after calling
** new_inode_init, and have not called reiserfs_new_inode yet.
** This should only be called on inodes that do not have stat data
** inserted into the tree yet.
*/
547 548
static int drop_new_inode(struct inode *inode)
{
549
	dquot_drop(inode);
550 551 552 553
	make_bad_inode(inode);
	inode->i_flags |= S_NOQUOTA;
	iput(inode);
	return 0;
L
Linus Torvalds 已提交
554 555
}

556
/* utility function that does setup for reiserfs_new_inode.
557
** dquot_initialize needs lots of credits so it's better to have it
L
Linus Torvalds 已提交
558 559 560
** outside of a transaction, so we had to pull some bits of
** reiserfs_new_inode out into this func.
*/
561 562 563 564 565 566
static int new_inode_init(struct inode *inode, struct inode *dir, int mode)
{

	/* the quota init calls have to know who to charge the quota to, so
	 ** we have to set uid and gid here
	 */
567
	inode->i_uid = current_fsuid();
568
	inode->i_mode = mode;
569 570 571
	/* Make inode invalid - just in case we are going to drop it before
	 * the initialization happens */
	INODE_PKEY(inode)->k_objectid = 0;
572 573 574 575 576 577

	if (dir->i_mode & S_ISGID) {
		inode->i_gid = dir->i_gid;
		if (S_ISDIR(mode))
			inode->i_mode |= S_ISGID;
	} else {
578
		inode->i_gid = current_fsgid();
579
	}
580
	dquot_initialize(inode);
581
	return 0;
L
Linus Torvalds 已提交
582 583
}

584 585
static int reiserfs_create(struct inode *dir, struct dentry *dentry, int mode,
			   struct nameidata *nd)
L
Linus Torvalds 已提交
586
{
587 588 589 590 591 592 593 594
	int retval;
	struct inode *inode;
	/* We need blocks for transaction + (user+group)*(quotas for new inode + update of quota for directory owner) */
	int jbegin_count =
	    JOURNAL_PER_BALANCE_CNT * 2 +
	    2 * (REISERFS_QUOTA_INIT_BLOCKS(dir->i_sb) +
		 REISERFS_QUOTA_TRANS_BLOCKS(dir->i_sb));
	struct reiserfs_transaction_handle th;
595
	struct reiserfs_security_handle security;
596

597
	dquot_initialize(dir);
598

599 600 601 602
	if (!(inode = new_inode(dir->i_sb))) {
		return -ENOMEM;
	}
	new_inode_init(inode, dir, mode);
L
Linus Torvalds 已提交
603

J
Jeff Mahoney 已提交
604
	jbegin_count += reiserfs_cache_default_acl(dir);
605 606 607 608 609 610
	retval = reiserfs_security_init(dir, inode, &security);
	if (retval < 0) {
		drop_new_inode(inode);
		return retval;
	}
	jbegin_count += retval;
611
	reiserfs_write_lock(dir->i_sb);
L
Linus Torvalds 已提交
612

613 614 615 616 617 618 619 620
	retval = journal_begin(&th, dir->i_sb, jbegin_count);
	if (retval) {
		drop_new_inode(inode);
		goto out_failed;
	}

	retval =
	    reiserfs_new_inode(&th, dir, mode, NULL, 0 /*i_size */ , dentry,
621
			       inode, &security);
622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638
	if (retval)
		goto out_failed;

	inode->i_op = &reiserfs_file_inode_operations;
	inode->i_fop = &reiserfs_file_operations;
	inode->i_mapping->a_ops = &reiserfs_address_space_operations;

	retval =
	    reiserfs_add_entry(&th, dir, dentry->d_name.name,
			       dentry->d_name.len, inode, 1 /*visible */ );
	if (retval) {
		int err;
		inode->i_nlink--;
		reiserfs_update_sd(&th, inode);
		err = journal_end(&th, dir->i_sb, jbegin_count);
		if (err)
			retval = err;
A
Al Viro 已提交
639
		unlock_new_inode(inode);
640 641 642 643 644
		iput(inode);
		goto out_failed;
	}
	reiserfs_update_inode_transaction(inode);
	reiserfs_update_inode_transaction(dir);
L
Linus Torvalds 已提交
645

646
	d_instantiate(dentry, inode);
A
Al Viro 已提交
647
	unlock_new_inode(inode);
648
	retval = journal_end(&th, dir->i_sb, jbegin_count);
L
Linus Torvalds 已提交
649

650 651 652 653
      out_failed:
	reiserfs_write_unlock(dir->i_sb);
	return retval;
}
L
Linus Torvalds 已提交
654

655 656 657 658 659 660
static int reiserfs_mknod(struct inode *dir, struct dentry *dentry, int mode,
			  dev_t rdev)
{
	int retval;
	struct inode *inode;
	struct reiserfs_transaction_handle th;
661
	struct reiserfs_security_handle security;
662 663 664 665 666 667 668 669 670
	/* We need blocks for transaction + (user+group)*(quotas for new inode + update of quota for directory owner) */
	int jbegin_count =
	    JOURNAL_PER_BALANCE_CNT * 3 +
	    2 * (REISERFS_QUOTA_INIT_BLOCKS(dir->i_sb) +
		 REISERFS_QUOTA_TRANS_BLOCKS(dir->i_sb));

	if (!new_valid_dev(rdev))
		return -EINVAL;

671
	dquot_initialize(dir);
672

673 674 675 676
	if (!(inode = new_inode(dir->i_sb))) {
		return -ENOMEM;
	}
	new_inode_init(inode, dir, mode);
L
Linus Torvalds 已提交
677

J
Jeff Mahoney 已提交
678
	jbegin_count += reiserfs_cache_default_acl(dir);
679 680 681 682 683 684
	retval = reiserfs_security_init(dir, inode, &security);
	if (retval < 0) {
		drop_new_inode(inode);
		return retval;
	}
	jbegin_count += retval;
685
	reiserfs_write_lock(dir->i_sb);
L
Linus Torvalds 已提交
686

687 688 689 690 691
	retval = journal_begin(&th, dir->i_sb, jbegin_count);
	if (retval) {
		drop_new_inode(inode);
		goto out_failed;
	}
L
Linus Torvalds 已提交
692

693 694
	retval =
	    reiserfs_new_inode(&th, dir, mode, NULL, 0 /*i_size */ , dentry,
695
			       inode, &security);
696 697 698
	if (retval) {
		goto out_failed;
	}
L
Linus Torvalds 已提交
699

700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718
	inode->i_op = &reiserfs_special_inode_operations;
	init_special_inode(inode, inode->i_mode, rdev);

	//FIXME: needed for block and char devices only
	reiserfs_update_sd(&th, inode);

	reiserfs_update_inode_transaction(inode);
	reiserfs_update_inode_transaction(dir);

	retval =
	    reiserfs_add_entry(&th, dir, dentry->d_name.name,
			       dentry->d_name.len, inode, 1 /*visible */ );
	if (retval) {
		int err;
		inode->i_nlink--;
		reiserfs_update_sd(&th, inode);
		err = journal_end(&th, dir->i_sb, jbegin_count);
		if (err)
			retval = err;
A
Al Viro 已提交
719
		unlock_new_inode(inode);
720 721 722
		iput(inode);
		goto out_failed;
	}
L
Linus Torvalds 已提交
723

724
	d_instantiate(dentry, inode);
A
Al Viro 已提交
725
	unlock_new_inode(inode);
726
	retval = journal_end(&th, dir->i_sb, jbegin_count);
L
Linus Torvalds 已提交
727

728 729 730
      out_failed:
	reiserfs_write_unlock(dir->i_sb);
	return retval;
L
Linus Torvalds 已提交
731 732
}

733
static int reiserfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
L
Linus Torvalds 已提交
734
{
735 736 737
	int retval;
	struct inode *inode;
	struct reiserfs_transaction_handle th;
738
	struct reiserfs_security_handle security;
739
	int lock_depth;
740 741 742 743 744
	/* We need blocks for transaction + (user+group)*(quotas for new inode + update of quota for directory owner) */
	int jbegin_count =
	    JOURNAL_PER_BALANCE_CNT * 3 +
	    2 * (REISERFS_QUOTA_INIT_BLOCKS(dir->i_sb) +
		 REISERFS_QUOTA_TRANS_BLOCKS(dir->i_sb));
L
Linus Torvalds 已提交
745

746
	dquot_initialize(dir);
747

L
Linus Torvalds 已提交
748
#ifdef DISPLACE_NEW_PACKING_LOCALITIES
749 750
	/* set flag that new packing locality created and new blocks for the content     * of that directory are not displaced yet */
	REISERFS_I(dir)->new_packing_locality = 1;
L
Linus Torvalds 已提交
751
#endif
752 753 754 755 756 757
	mode = S_IFDIR | mode;
	if (!(inode = new_inode(dir->i_sb))) {
		return -ENOMEM;
	}
	new_inode_init(inode, dir, mode);

J
Jeff Mahoney 已提交
758
	jbegin_count += reiserfs_cache_default_acl(dir);
759 760 761 762 763 764
	retval = reiserfs_security_init(dir, inode, &security);
	if (retval < 0) {
		drop_new_inode(inode);
		return retval;
	}
	jbegin_count += retval;
765
	lock_depth = reiserfs_write_lock_once(dir->i_sb);
766 767 768 769 770 771

	retval = journal_begin(&th, dir->i_sb, jbegin_count);
	if (retval) {
		drop_new_inode(inode);
		goto out_failed;
	}
L
Linus Torvalds 已提交
772

773 774 775 776 777 778 779 780
	/* inc the link count now, so another writer doesn't overflow it while
	 ** we sleep later on.
	 */
	INC_DIR_INODE_NLINK(dir)

	    retval = reiserfs_new_inode(&th, dir, mode, NULL /*symlink */ ,
					old_format_only(dir->i_sb) ?
					EMPTY_DIR_SIZE_V1 : EMPTY_DIR_SIZE,
781
					dentry, inode, &security);
782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804
	if (retval) {
		dir->i_nlink--;
		goto out_failed;
	}

	reiserfs_update_inode_transaction(inode);
	reiserfs_update_inode_transaction(dir);

	inode->i_op = &reiserfs_dir_inode_operations;
	inode->i_fop = &reiserfs_dir_operations;

	// note, _this_ add_entry will not update dir's stat data
	retval =
	    reiserfs_add_entry(&th, dir, dentry->d_name.name,
			       dentry->d_name.len, inode, 1 /*visible */ );
	if (retval) {
		int err;
		inode->i_nlink = 0;
		DEC_DIR_INODE_NLINK(dir);
		reiserfs_update_sd(&th, inode);
		err = journal_end(&th, dir->i_sb, jbegin_count);
		if (err)
			retval = err;
A
Al Viro 已提交
805
		unlock_new_inode(inode);
806 807 808 809 810 811 812
		iput(inode);
		goto out_failed;
	}
	// the above add_entry did not update dir's stat data
	reiserfs_update_sd(&th, dir);

	d_instantiate(dentry, inode);
A
Al Viro 已提交
813
	unlock_new_inode(inode);
814
	retval = journal_end(&th, dir->i_sb, jbegin_count);
815 816
out_failed:
	reiserfs_write_unlock_once(dir->i_sb, lock_depth);
817
	return retval;
L
Linus Torvalds 已提交
818 819
}

820
static inline int reiserfs_empty_dir(struct inode *inode)
L
Linus Torvalds 已提交
821
{
822 823
	/* we can cheat because an old format dir cannot have
	 ** EMPTY_DIR_SIZE, and a new format dir cannot have
824
	 ** EMPTY_DIR_SIZE_V1.  So, if the inode is either size,
825 826 827 828 829 830 831
	 ** regardless of disk format version, the directory is empty.
	 */
	if (inode->i_size != EMPTY_DIR_SIZE &&
	    inode->i_size != EMPTY_DIR_SIZE_V1) {
		return 0;
	}
	return 1;
L
Linus Torvalds 已提交
832 833
}

834
static int reiserfs_rmdir(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
835
{
836 837 838 839 840 841 842 843 844 845 846 847 848 849 850
	int retval, err;
	struct inode *inode;
	struct reiserfs_transaction_handle th;
	int jbegin_count;
	INITIALIZE_PATH(path);
	struct reiserfs_dir_entry de;

	/* we will be doing 2 balancings and update 2 stat data, we change quotas
	 * of the owner of the directory and of the owner of the parent directory.
	 * The quota structure is possibly deleted only on last iput => outside
	 * of this transaction */
	jbegin_count =
	    JOURNAL_PER_BALANCE_CNT * 2 + 2 +
	    4 * REISERFS_QUOTA_TRANS_BLOCKS(dir->i_sb);

851
	dquot_initialize(dir);
852

853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891
	reiserfs_write_lock(dir->i_sb);
	retval = journal_begin(&th, dir->i_sb, jbegin_count);
	if (retval)
		goto out_rmdir;

	de.de_gen_number_bit_string = NULL;
	if ((retval =
	     reiserfs_find_entry(dir, dentry->d_name.name, dentry->d_name.len,
				 &path, &de)) == NAME_NOT_FOUND) {
		retval = -ENOENT;
		goto end_rmdir;
	} else if (retval == IO_ERROR) {
		retval = -EIO;
		goto end_rmdir;
	}

	inode = dentry->d_inode;

	reiserfs_update_inode_transaction(inode);
	reiserfs_update_inode_transaction(dir);

	if (de.de_objectid != inode->i_ino) {
		// FIXME: compare key of an object and a key found in the
		// entry
		retval = -EIO;
		goto end_rmdir;
	}
	if (!reiserfs_empty_dir(inode)) {
		retval = -ENOTEMPTY;
		goto end_rmdir;
	}

	/* cut entry from dir directory */
	retval = reiserfs_cut_from_item(&th, &path, &(de.de_entry_key), dir, NULL,	/* page */
					0 /*new file size - not used here */ );
	if (retval < 0)
		goto end_rmdir;

	if (inode->i_nlink != 2 && inode->i_nlink != 1)
J
Jeff Mahoney 已提交
892 893 894
		reiserfs_error(inode->i_sb, "reiserfs-7040",
			       "empty directory has nlink != 2 (%d)",
			       inode->i_nlink);
895

896
	clear_nlink(inode);
897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920
	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME_SEC;
	reiserfs_update_sd(&th, inode);

	DEC_DIR_INODE_NLINK(dir)
	    dir->i_size -= (DEH_SIZE + de.de_entrylen);
	reiserfs_update_sd(&th, dir);

	/* prevent empty directory from getting lost */
	add_save_link(&th, inode, 0 /* not truncate */ );

	retval = journal_end(&th, dir->i_sb, jbegin_count);
	reiserfs_check_path(&path);
      out_rmdir:
	reiserfs_write_unlock(dir->i_sb);
	return retval;

      end_rmdir:
	/* we must release path, because we did not call
	   reiserfs_cut_from_item, or reiserfs_cut_from_item does not
	   release path if operation was not complete */
	pathrelse(&path);
	err = journal_end(&th, dir->i_sb, jbegin_count);
	reiserfs_write_unlock(dir->i_sb);
	return err ? err : retval;
L
Linus Torvalds 已提交
921 922
}

923
static int reiserfs_unlink(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
924
{
925 926 927 928 929 930 931
	int retval, err;
	struct inode *inode;
	struct reiserfs_dir_entry de;
	INITIALIZE_PATH(path);
	struct reiserfs_transaction_handle th;
	int jbegin_count;
	unsigned long savelink;
932
	int depth;
933

934
	dquot_initialize(dir);
935

936 937 938 939 940 941 942 943 944 945
	inode = dentry->d_inode;

	/* in this transaction we can be doing at max two balancings and update
	 * two stat datas, we change quotas of the owner of the directory and of
	 * the owner of the parent directory. The quota structure is possibly
	 * deleted only on iput => outside of this transaction */
	jbegin_count =
	    JOURNAL_PER_BALANCE_CNT * 2 + 2 +
	    4 * REISERFS_QUOTA_TRANS_BLOCKS(dir->i_sb);

946
	depth = reiserfs_write_lock_once(dir->i_sb);
947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972
	retval = journal_begin(&th, dir->i_sb, jbegin_count);
	if (retval)
		goto out_unlink;

	de.de_gen_number_bit_string = NULL;
	if ((retval =
	     reiserfs_find_entry(dir, dentry->d_name.name, dentry->d_name.len,
				 &path, &de)) == NAME_NOT_FOUND) {
		retval = -ENOENT;
		goto end_unlink;
	} else if (retval == IO_ERROR) {
		retval = -EIO;
		goto end_unlink;
	}

	reiserfs_update_inode_transaction(inode);
	reiserfs_update_inode_transaction(dir);

	if (de.de_objectid != inode->i_ino) {
		// FIXME: compare key of an object and a key found in the
		// entry
		retval = -EIO;
		goto end_unlink;
	}

	if (!inode->i_nlink) {
J
Jeff Mahoney 已提交
973 974 975
		reiserfs_warning(inode->i_sb, "reiserfs-7042",
				 "deleting nonexistent file (%lu), %d",
				 inode->i_ino, inode->i_nlink);
976 977 978
		inode->i_nlink = 1;
	}

979
	drop_nlink(inode);
980 981 982 983 984 985 986 987 988 989 990

	/*
	 * we schedule before doing the add_save_link call, save the link
	 * count so we don't race
	 */
	savelink = inode->i_nlink;

	retval =
	    reiserfs_cut_from_item(&th, &path, &(de.de_entry_key), dir, NULL,
				   0);
	if (retval < 0) {
991
		inc_nlink(inode);
992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006
		goto end_unlink;
	}
	inode->i_ctime = CURRENT_TIME_SEC;
	reiserfs_update_sd(&th, inode);

	dir->i_size -= (de.de_entrylen + DEH_SIZE);
	dir->i_ctime = dir->i_mtime = CURRENT_TIME_SEC;
	reiserfs_update_sd(&th, dir);

	if (!savelink)
		/* prevent file from getting lost */
		add_save_link(&th, inode, 0 /* not truncate */ );

	retval = journal_end(&th, dir->i_sb, jbegin_count);
	reiserfs_check_path(&path);
1007
	reiserfs_write_unlock_once(dir->i_sb, depth);
1008 1009 1010 1011 1012 1013
	return retval;

      end_unlink:
	pathrelse(&path);
	err = journal_end(&th, dir->i_sb, jbegin_count);
	reiserfs_check_path(&path);
L
Linus Torvalds 已提交
1014
	if (err)
1015 1016
		retval = err;
      out_unlink:
1017
	reiserfs_write_unlock_once(dir->i_sb, depth);
1018
	return retval;
L
Linus Torvalds 已提交
1019 1020
}

1021 1022
static int reiserfs_symlink(struct inode *parent_dir,
			    struct dentry *dentry, const char *symname)
L
Linus Torvalds 已提交
1023
{
1024 1025 1026 1027 1028
	int retval;
	struct inode *inode;
	char *name;
	int item_len;
	struct reiserfs_transaction_handle th;
1029
	struct reiserfs_security_handle security;
1030 1031 1032 1033 1034 1035 1036
	int mode = S_IFLNK | S_IRWXUGO;
	/* We need blocks for transaction + (user+group)*(quotas for new inode + update of quota for directory owner) */
	int jbegin_count =
	    JOURNAL_PER_BALANCE_CNT * 3 +
	    2 * (REISERFS_QUOTA_INIT_BLOCKS(parent_dir->i_sb) +
		 REISERFS_QUOTA_TRANS_BLOCKS(parent_dir->i_sb));

1037
	dquot_initialize(parent_dir);
1038

1039 1040 1041 1042 1043
	if (!(inode = new_inode(parent_dir->i_sb))) {
		return -ENOMEM;
	}
	new_inode_init(inode, parent_dir, mode);

1044 1045 1046 1047 1048 1049 1050
	retval = reiserfs_security_init(parent_dir, inode, &security);
	if (retval < 0) {
		drop_new_inode(inode);
		return retval;
	}
	jbegin_count += retval;

1051 1052 1053 1054 1055 1056 1057 1058
	reiserfs_write_lock(parent_dir->i_sb);
	item_len = ROUND_UP(strlen(symname));
	if (item_len > MAX_DIRECT_ITEM_LEN(parent_dir->i_sb->s_blocksize)) {
		retval = -ENAMETOOLONG;
		drop_new_inode(inode);
		goto out_failed;
	}

1059
	name = kmalloc(item_len, GFP_NOFS);
1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070
	if (!name) {
		drop_new_inode(inode);
		retval = -ENOMEM;
		goto out_failed;
	}
	memcpy(name, symname, strlen(symname));
	padd_item(name, item_len, strlen(symname));

	retval = journal_begin(&th, parent_dir->i_sb, jbegin_count);
	if (retval) {
		drop_new_inode(inode);
1071
		kfree(name);
1072 1073 1074 1075 1076
		goto out_failed;
	}

	retval =
	    reiserfs_new_inode(&th, parent_dir, mode, name, strlen(symname),
1077
			       dentry, inode, &security);
1078
	kfree(name);
1079 1080 1081
	if (retval) {		/* reiserfs_new_inode iputs for us */
		goto out_failed;
	}
L
Linus Torvalds 已提交
1082

1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101
	reiserfs_update_inode_transaction(inode);
	reiserfs_update_inode_transaction(parent_dir);

	inode->i_op = &reiserfs_symlink_inode_operations;
	inode->i_mapping->a_ops = &reiserfs_address_space_operations;

	// must be sure this inode is written with this transaction
	//
	//reiserfs_update_sd (&th, inode, READ_BLOCKS);

	retval = reiserfs_add_entry(&th, parent_dir, dentry->d_name.name,
				    dentry->d_name.len, inode, 1 /*visible */ );
	if (retval) {
		int err;
		inode->i_nlink--;
		reiserfs_update_sd(&th, inode);
		err = journal_end(&th, parent_dir->i_sb, jbegin_count);
		if (err)
			retval = err;
A
Al Viro 已提交
1102
		unlock_new_inode(inode);
1103 1104 1105
		iput(inode);
		goto out_failed;
	}
L
Linus Torvalds 已提交
1106

1107
	d_instantiate(dentry, inode);
A
Al Viro 已提交
1108
	unlock_new_inode(inode);
1109 1110 1111 1112
	retval = journal_end(&th, parent_dir->i_sb, jbegin_count);
      out_failed:
	reiserfs_write_unlock(parent_dir->i_sb);
	return retval;
L
Linus Torvalds 已提交
1113 1114
}

1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125
static int reiserfs_link(struct dentry *old_dentry, struct inode *dir,
			 struct dentry *dentry)
{
	int retval;
	struct inode *inode = old_dentry->d_inode;
	struct reiserfs_transaction_handle th;
	/* We need blocks for transaction + update of quotas for the owners of the directory */
	int jbegin_count =
	    JOURNAL_PER_BALANCE_CNT * 3 +
	    2 * REISERFS_QUOTA_TRANS_BLOCKS(dir->i_sb);

1126
	dquot_initialize(dir);
1127

1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139
	reiserfs_write_lock(dir->i_sb);
	if (inode->i_nlink >= REISERFS_LINK_MAX) {
		//FIXME: sd_nlink is 32 bit for new files
		reiserfs_write_unlock(dir->i_sb);
		return -EMLINK;
	}
	if (inode->i_nlink == 0) {
		reiserfs_write_unlock(dir->i_sb);
		return -ENOENT;
	}

	/* inc before scheduling so reiserfs_unlink knows we are here */
1140
	inc_nlink(inode);
1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173

	retval = journal_begin(&th, dir->i_sb, jbegin_count);
	if (retval) {
		inode->i_nlink--;
		reiserfs_write_unlock(dir->i_sb);
		return retval;
	}

	/* create new entry */
	retval =
	    reiserfs_add_entry(&th, dir, dentry->d_name.name,
			       dentry->d_name.len, inode, 1 /*visible */ );

	reiserfs_update_inode_transaction(inode);
	reiserfs_update_inode_transaction(dir);

	if (retval) {
		int err;
		inode->i_nlink--;
		err = journal_end(&th, dir->i_sb, jbegin_count);
		reiserfs_write_unlock(dir->i_sb);
		return err ? err : retval;
	}

	inode->i_ctime = CURRENT_TIME_SEC;
	reiserfs_update_sd(&th, inode);

	atomic_inc(&inode->i_count);
	d_instantiate(dentry, inode);
	retval = journal_end(&th, dir->i_sb, jbegin_count);
	reiserfs_write_unlock(dir->i_sb);
	return retval;
}
L
Linus Torvalds 已提交
1174

1175
/* de contains information pointing to an entry which */
1176 1177
static int de_still_valid(const char *name, int len,
			  struct reiserfs_dir_entry *de)
L
Linus Torvalds 已提交
1178
{
1179 1180 1181 1182 1183 1184 1185 1186
	struct reiserfs_dir_entry tmp = *de;

	// recalculate pointer to name and name length
	set_de_name_and_namelen(&tmp);
	// FIXME: could check more
	if (tmp.de_namelen != len || memcmp(name, de->de_name, len))
		return 0;
	return 1;
L
Linus Torvalds 已提交
1187 1188
}

1189 1190 1191
static int entry_points_to_object(const char *name, int len,
				  struct reiserfs_dir_entry *de,
				  struct inode *inode)
L
Linus Torvalds 已提交
1192
{
1193 1194 1195 1196 1197
	if (!de_still_valid(name, len, de))
		return 0;

	if (inode) {
		if (!de_visible(de->de_deh + de->de_entry_num))
J
Jeff Mahoney 已提交
1198 1199
			reiserfs_panic(inode->i_sb, "vs-7042",
				       "entry must be visible");
1200 1201
		return (de->de_objectid == inode->i_ino) ? 1 : 0;
	}
L
Linus Torvalds 已提交
1202

1203 1204
	/* this must be added hidden entry */
	if (de_visible(de->de_deh + de->de_entry_num))
J
Jeff Mahoney 已提交
1205
		reiserfs_panic(NULL, "vs-7043", "entry must be visible");
L
Linus Torvalds 已提交
1206

1207
	return 1;
L
Linus Torvalds 已提交
1208 1209 1210
}

/* sets key of objectid the entry has to point to */
1211 1212
static void set_ino_in_dir_entry(struct reiserfs_dir_entry *de,
				 struct reiserfs_key *key)
L
Linus Torvalds 已提交
1213
{
1214 1215 1216
	/* JDM These operations are endian safe - both are le */
	de->de_deh[de->de_entry_num].deh_dir_id = key->k_dir_id;
	de->de_deh[de->de_entry_num].deh_objectid = key->k_objectid;
L
Linus Torvalds 已提交
1217 1218
}

1219
/*
L
Linus Torvalds 已提交
1220 1221
 * process, that is going to call fix_nodes/do_balance must hold only
 * one path. If it holds 2 or more, it can get into endless waiting in
1222
 * get_empty_nodes or its clones
L
Linus Torvalds 已提交
1223
 */
1224 1225
static int reiserfs_rename(struct inode *old_dir, struct dentry *old_dentry,
			   struct inode *new_dir, struct dentry *new_dentry)
L
Linus Torvalds 已提交
1226
{
1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251
	int retval;
	INITIALIZE_PATH(old_entry_path);
	INITIALIZE_PATH(new_entry_path);
	INITIALIZE_PATH(dot_dot_entry_path);
	struct item_head new_entry_ih, old_entry_ih, dot_dot_ih;
	struct reiserfs_dir_entry old_de, new_de, dot_dot_de;
	struct inode *old_inode, *new_dentry_inode;
	struct reiserfs_transaction_handle th;
	int jbegin_count;
	umode_t old_inode_mode;
	unsigned long savelink = 1;
	struct timespec ctime;

	/* three balancings: (1) old name removal, (2) new name insertion
	   and (3) maybe "save" link insertion
	   stat data updates: (1) old directory,
	   (2) new directory and (3) maybe old object stat data (when it is
	   directory) and (4) maybe stat data of object to which new entry
	   pointed initially and (5) maybe block containing ".." of
	   renamed directory
	   quota updates: two parent directories */
	jbegin_count =
	    JOURNAL_PER_BALANCE_CNT * 3 + 5 +
	    4 * REISERFS_QUOTA_TRANS_BLOCKS(old_dir->i_sb);

1252 1253
	dquot_initialize(old_dir);
	dquot_initialize(new_dir);
1254

1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267
	old_inode = old_dentry->d_inode;
	new_dentry_inode = new_dentry->d_inode;

	// make sure, that oldname still exists and points to an object we
	// are going to rename
	old_de.de_gen_number_bit_string = NULL;
	reiserfs_write_lock(old_dir->i_sb);
	retval =
	    reiserfs_find_entry(old_dir, old_dentry->d_name.name,
				old_dentry->d_name.len, &old_entry_path,
				&old_de);
	pathrelse(&old_entry_path);
	if (retval == IO_ERROR) {
L
Linus Torvalds 已提交
1268
		reiserfs_write_unlock(old_dir->i_sb);
1269
		return -EIO;
L
Linus Torvalds 已提交
1270
	}
1271 1272 1273 1274

	if (retval != NAME_FOUND || old_de.de_objectid != old_inode->i_ino) {
		reiserfs_write_unlock(old_dir->i_sb);
		return -ENOENT;
L
Linus Torvalds 已提交
1275 1276
	}

1277 1278
	old_inode_mode = old_inode->i_mode;
	if (S_ISDIR(old_inode_mode)) {
1279
		// make sure, that directory being renamed has correct ".."
1280 1281 1282 1283 1284 1285 1286 1287 1288 1289
		// and that its new parent directory has not too many links
		// already

		if (new_dentry_inode) {
			if (!reiserfs_empty_dir(new_dentry_inode)) {
				reiserfs_write_unlock(old_dir->i_sb);
				return -ENOTEMPTY;
			}
		}

1290 1291
		/* directory is renamed, its parent directory will be changed,
		 ** so find ".." entry
1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307
		 */
		dot_dot_de.de_gen_number_bit_string = NULL;
		retval =
		    reiserfs_find_entry(old_inode, "..", 2, &dot_dot_entry_path,
					&dot_dot_de);
		pathrelse(&dot_dot_entry_path);
		if (retval != NAME_FOUND) {
			reiserfs_write_unlock(old_dir->i_sb);
			return -EIO;
		}

		/* inode number of .. must equal old_dir->i_ino */
		if (dot_dot_de.de_objectid != old_dir->i_ino) {
			reiserfs_write_unlock(old_dir->i_sb);
			return -EIO;
		}
L
Linus Torvalds 已提交
1308
	}
1309 1310 1311 1312 1313

	retval = journal_begin(&th, old_dir->i_sb, jbegin_count);
	if (retval) {
		reiserfs_write_unlock(old_dir->i_sb);
		return retval;
L
Linus Torvalds 已提交
1314
	}
1315 1316 1317 1318 1319 1320 1321

	/* add new entry (or find the existing one) */
	retval =
	    reiserfs_add_entry(&th, new_dir, new_dentry->d_name.name,
			       new_dentry->d_name.len, old_inode, 0);
	if (retval == -EEXIST) {
		if (!new_dentry_inode) {
J
Jeff Mahoney 已提交
1322 1323
			reiserfs_panic(old_dir->i_sb, "vs-7050",
				       "new entry is found, new inode == 0");
1324 1325 1326 1327 1328
		}
	} else if (retval) {
		int err = journal_end(&th, old_dir->i_sb, jbegin_count);
		reiserfs_write_unlock(old_dir->i_sb);
		return err ? err : retval;
L
Linus Torvalds 已提交
1329 1330
	}

1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400
	reiserfs_update_inode_transaction(old_dir);
	reiserfs_update_inode_transaction(new_dir);

	/* this makes it so an fsync on an open fd for the old name will
	 ** commit the rename operation
	 */
	reiserfs_update_inode_transaction(old_inode);

	if (new_dentry_inode)
		reiserfs_update_inode_transaction(new_dentry_inode);

	while (1) {
		// look for old name using corresponding entry key (found by reiserfs_find_entry)
		if ((retval =
		     search_by_entry_key(new_dir->i_sb, &old_de.de_entry_key,
					 &old_entry_path,
					 &old_de)) != NAME_FOUND) {
			pathrelse(&old_entry_path);
			journal_end(&th, old_dir->i_sb, jbegin_count);
			reiserfs_write_unlock(old_dir->i_sb);
			return -EIO;
		}

		copy_item_head(&old_entry_ih, get_ih(&old_entry_path));

		reiserfs_prepare_for_journal(old_inode->i_sb, old_de.de_bh, 1);

		// look for new name by reiserfs_find_entry
		new_de.de_gen_number_bit_string = NULL;
		retval =
		    reiserfs_find_entry(new_dir, new_dentry->d_name.name,
					new_dentry->d_name.len, &new_entry_path,
					&new_de);
		// reiserfs_add_entry should not return IO_ERROR, because it is called with essentially same parameters from
		// reiserfs_add_entry above, and we'll catch any i/o errors before we get here.
		if (retval != NAME_FOUND_INVISIBLE && retval != NAME_FOUND) {
			pathrelse(&new_entry_path);
			pathrelse(&old_entry_path);
			journal_end(&th, old_dir->i_sb, jbegin_count);
			reiserfs_write_unlock(old_dir->i_sb);
			return -EIO;
		}

		copy_item_head(&new_entry_ih, get_ih(&new_entry_path));

		reiserfs_prepare_for_journal(old_inode->i_sb, new_de.de_bh, 1);

		if (S_ISDIR(old_inode->i_mode)) {
			if ((retval =
			     search_by_entry_key(new_dir->i_sb,
						 &dot_dot_de.de_entry_key,
						 &dot_dot_entry_path,
						 &dot_dot_de)) != NAME_FOUND) {
				pathrelse(&dot_dot_entry_path);
				pathrelse(&new_entry_path);
				pathrelse(&old_entry_path);
				journal_end(&th, old_dir->i_sb, jbegin_count);
				reiserfs_write_unlock(old_dir->i_sb);
				return -EIO;
			}
			copy_item_head(&dot_dot_ih,
				       get_ih(&dot_dot_entry_path));
			// node containing ".." gets into transaction
			reiserfs_prepare_for_journal(old_inode->i_sb,
						     dot_dot_de.de_bh, 1);
		}
		/* we should check seals here, not do
		   this stuff, yes? Then, having
		   gathered everything into RAM we
		   should lock the buffers, yes?  -Hans */
1401 1402 1403
		/* probably.  our rename needs to hold more
		 ** than one path at once.  The seals would
		 ** have to be written to deal with multi-path
1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450
		 ** issues -chris
		 */
		/* sanity checking before doing the rename - avoid races many
		 ** of the above checks could have scheduled.  We have to be
		 ** sure our items haven't been shifted by another process.
		 */
		if (item_moved(&new_entry_ih, &new_entry_path) ||
		    !entry_points_to_object(new_dentry->d_name.name,
					    new_dentry->d_name.len,
					    &new_de, new_dentry_inode) ||
		    item_moved(&old_entry_ih, &old_entry_path) ||
		    !entry_points_to_object(old_dentry->d_name.name,
					    old_dentry->d_name.len,
					    &old_de, old_inode)) {
			reiserfs_restore_prepared_buffer(old_inode->i_sb,
							 new_de.de_bh);
			reiserfs_restore_prepared_buffer(old_inode->i_sb,
							 old_de.de_bh);
			if (S_ISDIR(old_inode_mode))
				reiserfs_restore_prepared_buffer(old_inode->
								 i_sb,
								 dot_dot_de.
								 de_bh);
			continue;
		}
		if (S_ISDIR(old_inode_mode)) {
			if (item_moved(&dot_dot_ih, &dot_dot_entry_path) ||
			    !entry_points_to_object("..", 2, &dot_dot_de,
						    old_dir)) {
				reiserfs_restore_prepared_buffer(old_inode->
								 i_sb,
								 old_de.de_bh);
				reiserfs_restore_prepared_buffer(old_inode->
								 i_sb,
								 new_de.de_bh);
				reiserfs_restore_prepared_buffer(old_inode->
								 i_sb,
								 dot_dot_de.
								 de_bh);
				continue;
			}
		}

		RFALSE(S_ISDIR(old_inode_mode) &&
		       !buffer_journal_prepared(dot_dot_de.de_bh), "");

		break;
L
Linus Torvalds 已提交
1451 1452
	}

1453 1454
	/* ok, all the changes can be done in one fell swoop when we
	   have claimed all the buffers needed. */
L
Linus Torvalds 已提交
1455

1456 1457 1458
	mark_de_visible(new_de.de_deh + new_de.de_entry_num);
	set_ino_in_dir_entry(&new_de, INODE_PKEY(old_inode));
	journal_mark_dirty(&th, old_dir->i_sb, new_de.de_bh);
L
Linus Torvalds 已提交
1459

1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471
	mark_de_hidden(old_de.de_deh + old_de.de_entry_num);
	journal_mark_dirty(&th, old_dir->i_sb, old_de.de_bh);
	ctime = CURRENT_TIME_SEC;
	old_dir->i_ctime = old_dir->i_mtime = ctime;
	new_dir->i_ctime = new_dir->i_mtime = ctime;
	/* thanks to Alex Adriaanse <alex_a@caltech.edu> for patch which adds ctime update of
	   renamed object */
	old_inode->i_ctime = ctime;

	if (new_dentry_inode) {
		// adjust link number of the victim
		if (S_ISDIR(new_dentry_inode->i_mode)) {
1472
			clear_nlink(new_dentry_inode);
1473
		} else {
1474
			drop_nlink(new_dentry_inode);
1475 1476 1477
		}
		new_dentry_inode->i_ctime = ctime;
		savelink = new_dentry_inode->i_nlink;
L
Linus Torvalds 已提交
1478
	}
1479

L
Linus Torvalds 已提交
1480
	if (S_ISDIR(old_inode_mode)) {
1481
		/* adjust ".." of renamed directory */
1482 1483 1484 1485 1486 1487 1488 1489 1490 1491
		set_ino_in_dir_entry(&dot_dot_de, INODE_PKEY(new_dir));
		journal_mark_dirty(&th, new_dir->i_sb, dot_dot_de.de_bh);

		if (!new_dentry_inode)
			/* there (in new_dir) was no directory, so it got new link
			   (".."  of renamed directory) */
			INC_DIR_INODE_NLINK(new_dir);

		/* old directory lost one link - ".. " of renamed directory */
		DEC_DIR_INODE_NLINK(old_dir);
L
Linus Torvalds 已提交
1492
	}
1493 1494 1495
	// looks like in 2.3.99pre3 brelse is atomic. so we can use pathrelse
	pathrelse(&new_entry_path);
	pathrelse(&dot_dot_entry_path);
L
Linus Torvalds 已提交
1496

1497 1498 1499 1500 1501 1502
	// FIXME: this reiserfs_cut_from_item's return value may screw up
	// anybody, but it will panic if will not be able to find the
	// entry. This needs one more clean up
	if (reiserfs_cut_from_item
	    (&th, &old_entry_path, &(old_de.de_entry_key), old_dir, NULL,
	     0) < 0)
J
Jeff Mahoney 已提交
1503 1504
		reiserfs_error(old_dir->i_sb, "vs-7060",
			       "couldn't not cut old name. Fsck later?");
1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516

	old_dir->i_size -= DEH_SIZE + old_de.de_entrylen;

	reiserfs_update_sd(&th, old_dir);
	reiserfs_update_sd(&th, new_dir);
	reiserfs_update_sd(&th, old_inode);

	if (new_dentry_inode) {
		if (savelink == 0)
			add_save_link(&th, new_dentry_inode,
				      0 /* not truncate */ );
		reiserfs_update_sd(&th, new_dentry_inode);
L
Linus Torvalds 已提交
1517
	}
1518 1519 1520 1521

	retval = journal_end(&th, old_dir->i_sb, jbegin_count);
	reiserfs_write_unlock(old_dir->i_sb);
	return retval;
L
Linus Torvalds 已提交
1522 1523 1524 1525 1526
}

/*
 * directories can handle most operations...
 */
1527
const struct inode_operations reiserfs_dir_inode_operations = {
1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543
	//&reiserfs_dir_operations,   /* default_file_ops */
	.create = reiserfs_create,
	.lookup = reiserfs_lookup,
	.link = reiserfs_link,
	.unlink = reiserfs_unlink,
	.symlink = reiserfs_symlink,
	.mkdir = reiserfs_mkdir,
	.rmdir = reiserfs_rmdir,
	.mknod = reiserfs_mknod,
	.rename = reiserfs_rename,
	.setattr = reiserfs_setattr,
	.setxattr = reiserfs_setxattr,
	.getxattr = reiserfs_getxattr,
	.listxattr = reiserfs_listxattr,
	.removexattr = reiserfs_removexattr,
	.permission = reiserfs_permission,
L
Linus Torvalds 已提交
1544 1545 1546 1547 1548 1549
};

/*
 * symlink operations.. same as page_symlink_inode_operations, with xattr
 * stuff added
 */
1550
const struct inode_operations reiserfs_symlink_inode_operations = {
1551 1552 1553 1554 1555 1556 1557 1558 1559
	.readlink = generic_readlink,
	.follow_link = page_follow_link_light,
	.put_link = page_put_link,
	.setattr = reiserfs_setattr,
	.setxattr = reiserfs_setxattr,
	.getxattr = reiserfs_getxattr,
	.listxattr = reiserfs_listxattr,
	.removexattr = reiserfs_removexattr,
	.permission = reiserfs_permission,
L
Linus Torvalds 已提交
1560 1561 1562 1563 1564 1565

};

/*
 * special file operations.. just xattr/acl stuff
 */
1566
const struct inode_operations reiserfs_special_inode_operations = {
1567 1568 1569 1570 1571 1572
	.setattr = reiserfs_setattr,
	.setxattr = reiserfs_setxattr,
	.getxattr = reiserfs_getxattr,
	.listxattr = reiserfs_listxattr,
	.removexattr = reiserfs_removexattr,
	.permission = reiserfs_permission,
L
Linus Torvalds 已提交
1573 1574

};