namei.c 66.4 KB
Newer Older
1
/*
2
 *  linux/fs/ext4/namei.c
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
 *
 * Copyright (C) 1992, 1993, 1994, 1995
 * Remy Card (card@masi.ibp.fr)
 * Laboratoire MASI - Institut Blaise Pascal
 * Universite Pierre et Marie Curie (Paris VI)
 *
 *  from
 *
 *  linux/fs/minix/namei.c
 *
 *  Copyright (C) 1991, 1992  Linus Torvalds
 *
 *  Big-endian to little-endian byte-swapping/bitmaps by
 *        David S. Miller (davem@caip.rutgers.edu), 1995
 *  Directory entry file type support and forward compatibility hooks
 *	for B-tree directories by Theodore Ts'o (tytso@mit.edu), 1998
 *  Hash Tree Directory indexing (c)
 *	Daniel Phillips, 2001
 *  Hash Tree Directory indexing porting
 *	Christopher Li, 2002
 *  Hash Tree Directory indexing cleanup
 *	Theodore Ts'o, 2002
 */

#include <linux/fs.h>
#include <linux/pagemap.h>
29
#include <linux/jbd2.h>
30 31 32 33 34 35 36
#include <linux/time.h>
#include <linux/fcntl.h>
#include <linux/stat.h>
#include <linux/string.h>
#include <linux/quotaops.h>
#include <linux/buffer_head.h>
#include <linux/bio.h>
37 38
#include "ext4.h"
#include "ext4_jbd2.h"
39 40 41 42 43 44 45 46 47 48

#include "namei.h"
#include "xattr.h"
#include "acl.h"

/*
 * define how far ahead to read directories while searching them.
 */
#define NAMEI_RA_CHUNKS  2
#define NAMEI_RA_BLOCKS  4
D
Dave Kleikamp 已提交
49
#define NAMEI_RA_SIZE	     (NAMEI_RA_CHUNKS * NAMEI_RA_BLOCKS)
50 51
#define NAMEI_RA_INDEX(c,b)  (((c) * NAMEI_RA_BLOCKS) + (b))

52
static struct buffer_head *ext4_append(handle_t *handle,
53
					struct inode *inode,
A
Aneesh Kumar K.V 已提交
54
					ext4_lblk_t *block, int *err)
55 56 57 58 59
{
	struct buffer_head *bh;

	*block = inode->i_size >> inode->i_sb->s_blocksize_bits;

60 61
	bh = ext4_bread(handle, inode, *block, 1, err);
	if (bh) {
62
		inode->i_size += inode->i_sb->s_blocksize;
63
		EXT4_I(inode)->i_disksize = inode->i_size;
64 65 66 67 68
		*err = ext4_journal_get_write_access(handle, bh);
		if (*err) {
			brelse(bh);
			bh = NULL;
		}
69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147
	}
	return bh;
}

#ifndef assert
#define assert(test) J_ASSERT(test)
#endif

#ifndef swap
#define swap(x, y) do { typeof(x) z = x; x = y; y = z; } while (0)
#endif

#ifdef DX_DEBUG
#define dxtrace(command) command
#else
#define dxtrace(command)
#endif

struct fake_dirent
{
	__le32 inode;
	__le16 rec_len;
	u8 name_len;
	u8 file_type;
};

struct dx_countlimit
{
	__le16 limit;
	__le16 count;
};

struct dx_entry
{
	__le32 hash;
	__le32 block;
};

/*
 * dx_root_info is laid out so that if it should somehow get overlaid by a
 * dirent the two low bits of the hash version will be zero.  Therefore, the
 * hash version mod 4 should never be 0.  Sincerely, the paranoia department.
 */

struct dx_root
{
	struct fake_dirent dot;
	char dot_name[4];
	struct fake_dirent dotdot;
	char dotdot_name[4];
	struct dx_root_info
	{
		__le32 reserved_zero;
		u8 hash_version;
		u8 info_length; /* 8 */
		u8 indirect_levels;
		u8 unused_flags;
	}
	info;
	struct dx_entry	entries[0];
};

struct dx_node
{
	struct fake_dirent fake;
	struct dx_entry	entries[0];
};


struct dx_frame
{
	struct buffer_head *bh;
	struct dx_entry *entries;
	struct dx_entry *at;
};

struct dx_map_entry
{
	u32 hash;
148 149
	u16 offs;
	u16 size;
150 151
};

A
Aneesh Kumar K.V 已提交
152 153
static inline ext4_lblk_t dx_get_block(struct dx_entry *entry);
static void dx_set_block(struct dx_entry *entry, ext4_lblk_t value);
154 155 156 157 158 159 160 161
static inline unsigned dx_get_hash(struct dx_entry *entry);
static void dx_set_hash(struct dx_entry *entry, unsigned value);
static unsigned dx_get_count(struct dx_entry *entries);
static unsigned dx_get_limit(struct dx_entry *entries);
static void dx_set_count(struct dx_entry *entries, unsigned value);
static void dx_set_limit(struct dx_entry *entries, unsigned value);
static unsigned dx_root_limit(struct inode *dir, unsigned infosize);
static unsigned dx_node_limit(struct inode *dir);
162
static struct dx_frame *dx_probe(const struct qstr *d_name,
163 164 165 166
				 struct inode *dir,
				 struct dx_hash_info *hinfo,
				 struct dx_frame *frame,
				 int *err);
167 168 169
static void dx_release(struct dx_frame *frames);
static int dx_make_map(struct ext4_dir_entry_2 *de, int size,
		       struct dx_hash_info *hinfo, struct dx_map_entry map[]);
170
static void dx_sort_map(struct dx_map_entry *map, unsigned count);
171
static struct ext4_dir_entry_2 *dx_move_dirents(char *from, char *to,
172
		struct dx_map_entry *offsets, int count);
173
static struct ext4_dir_entry_2* dx_pack_dirents(char *base, int size);
A
Aneesh Kumar K.V 已提交
174 175
static void dx_insert_block(struct dx_frame *frame,
					u32 hash, ext4_lblk_t block);
176
static int ext4_htree_next_block(struct inode *dir, __u32 hash,
177 178 179
				 struct dx_frame *frame,
				 struct dx_frame *frames,
				 __u32 *start_hash);
180 181 182 183
static struct buffer_head * ext4_dx_find_entry(struct inode *dir,
		const struct qstr *d_name,
		struct ext4_dir_entry_2 **res_dir,
		int *err);
184
static int ext4_dx_add_entry(handle_t *handle, struct dentry *dentry,
185 186
			     struct inode *inode);

187 188 189 190 191 192 193 194 195 196
/*
 * p is at least 6 bytes before the end of page
 */
static inline struct ext4_dir_entry_2 *
ext4_next_entry(struct ext4_dir_entry_2 *p)
{
	return (struct ext4_dir_entry_2 *)((char *)p +
		ext4_rec_len_from_disk(p->rec_len));
}

197 198 199 200 201
/*
 * Future: use high four bits of block for coalesce-on-delete flags
 * Mask them off for now.
 */

A
Aneesh Kumar K.V 已提交
202
static inline ext4_lblk_t dx_get_block(struct dx_entry *entry)
203 204 205 206
{
	return le32_to_cpu(entry->block) & 0x00ffffff;
}

A
Aneesh Kumar K.V 已提交
207
static inline void dx_set_block(struct dx_entry *entry, ext4_lblk_t value)
208 209 210 211
{
	entry->block = cpu_to_le32(value);
}

212
static inline unsigned dx_get_hash(struct dx_entry *entry)
213 214 215 216
{
	return le32_to_cpu(entry->hash);
}

217
static inline void dx_set_hash(struct dx_entry *entry, unsigned value)
218 219 220 221
{
	entry->hash = cpu_to_le32(value);
}

222
static inline unsigned dx_get_count(struct dx_entry *entries)
223 224 225 226
{
	return le16_to_cpu(((struct dx_countlimit *) entries)->count);
}

227
static inline unsigned dx_get_limit(struct dx_entry *entries)
228 229 230 231
{
	return le16_to_cpu(((struct dx_countlimit *) entries)->limit);
}

232
static inline void dx_set_count(struct dx_entry *entries, unsigned value)
233 234 235 236
{
	((struct dx_countlimit *) entries)->count = cpu_to_le16(value);
}

237
static inline void dx_set_limit(struct dx_entry *entries, unsigned value)
238 239 240 241
{
	((struct dx_countlimit *) entries)->limit = cpu_to_le16(value);
}

242
static inline unsigned dx_root_limit(struct inode *dir, unsigned infosize)
243
{
244 245
	unsigned entry_space = dir->i_sb->s_blocksize - EXT4_DIR_REC_LEN(1) -
		EXT4_DIR_REC_LEN(2) - infosize;
246
	return entry_space / sizeof(struct dx_entry);
247 248
}

249
static inline unsigned dx_node_limit(struct inode *dir)
250
{
251
	unsigned entry_space = dir->i_sb->s_blocksize - EXT4_DIR_REC_LEN(0);
252
	return entry_space / sizeof(struct dx_entry);
253 254 255 256 257 258
}

/*
 * Debug
 */
#ifdef DX_DEBUG
259
static void dx_show_index(char * label, struct dx_entry *entries)
260
{
A
Andrew Morton 已提交
261
	int i, n = dx_get_count (entries);
262
	printk(KERN_DEBUG "%s index ", label);
A
Andrew Morton 已提交
263
	for (i = 0; i < n; i++) {
264
		printk("%x->%lu ", i ? dx_get_hash(entries + i) :
A
Aneesh Kumar K.V 已提交
265
				0, (unsigned long)dx_get_block(entries + i));
A
Andrew Morton 已提交
266 267
	}
	printk("\n");
268 269 270 271 272 273 274 275 276
}

struct stats
{
	unsigned names;
	unsigned space;
	unsigned bcount;
};

277
static struct stats dx_show_leaf(struct dx_hash_info *hinfo, struct ext4_dir_entry_2 *de,
278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293
				 int size, int show_names)
{
	unsigned names = 0, space = 0;
	char *base = (char *) de;
	struct dx_hash_info h = *hinfo;

	printk("names: ");
	while ((char *) de < base + size)
	{
		if (de->inode)
		{
			if (show_names)
			{
				int len = de->name_len;
				char *name = de->name;
				while (len--) printk("%c", *name++);
294
				ext4fs_dirhash(de->name, de->name_len, &h);
295 296 297
				printk(":%x.%u ", h.hash,
				       ((char *) de - base));
			}
298
			space += EXT4_DIR_REC_LEN(de->name_len);
299 300
			names++;
		}
301
		de = ext4_next_entry(de);
302 303 304 305 306 307 308 309 310
	}
	printk("(%i)\n", names);
	return (struct stats) { names, space, 1 };
}

struct stats dx_show_entries(struct dx_hash_info *hinfo, struct inode *dir,
			     struct dx_entry *entries, int levels)
{
	unsigned blocksize = dir->i_sb->s_blocksize;
311
	unsigned count = dx_get_count(entries), names = 0, space = 0, i;
312 313 314 315 316 317
	unsigned bcount = 0;
	struct buffer_head *bh;
	int err;
	printk("%i indexed blocks...\n", count);
	for (i = 0; i < count; i++, entries++)
	{
A
Aneesh Kumar K.V 已提交
318 319
		ext4_lblk_t block = dx_get_block(entries);
		ext4_lblk_t hash  = i ? dx_get_hash(entries): 0;
320 321 322
		u32 range = i < count - 1? (dx_get_hash(entries + 1) - hash): ~hash;
		struct stats stats;
		printk("%s%3u:%03u hash %8x/%8x ",levels?"":"   ", i, block, hash, range);
323
		if (!(bh = ext4_bread (NULL,dir, block, 0,&err))) continue;
324 325
		stats = levels?
		   dx_show_entries(hinfo, dir, ((struct dx_node *) bh->b_data)->entries, levels - 1):
326
		   dx_show_leaf(hinfo, (struct ext4_dir_entry_2 *) bh->b_data, blocksize, 0);
327 328 329
		names += stats.names;
		space += stats.space;
		bcount += stats.bcount;
330
		brelse(bh);
331 332
	}
	if (bcount)
333 334 335
		printk(KERN_DEBUG "%snames %u, fullness %u (%u%%)\n", 
		       levels ? "" : "   ", names, space/bcount,
		       (space/bcount)*100/blocksize);
336 337 338 339 340 341 342 343 344 345 346 347 348 349
	return (struct stats) { names, space, bcount};
}
#endif /* DX_DEBUG */

/*
 * Probe for a directory leaf block to search.
 *
 * dx_probe can return ERR_BAD_DX_DIR, which means there was a format
 * error in the directory index, and the caller should fall back to
 * searching the directory normally.  The callers of dx_probe **MUST**
 * check for this error code, and make sure it never gets reflected
 * back to userspace.
 */
static struct dx_frame *
350
dx_probe(const struct qstr *d_name, struct inode *dir,
351 352 353 354 355 356 357 358 359 360
	 struct dx_hash_info *hinfo, struct dx_frame *frame_in, int *err)
{
	unsigned count, indirect;
	struct dx_entry *at, *entries, *p, *q, *m;
	struct dx_root *root;
	struct buffer_head *bh;
	struct dx_frame *frame = frame_in;
	u32 hash;

	frame->bh = NULL;
361
	if (!(bh = ext4_bread (NULL,dir, 0, 0, err)))
362 363 364 365 366
		goto fail;
	root = (struct dx_root *) bh->b_data;
	if (root->info.hash_version != DX_HASH_TEA &&
	    root->info.hash_version != DX_HASH_HALF_MD4 &&
	    root->info.hash_version != DX_HASH_LEGACY) {
367
		ext4_warning(dir->i_sb, __func__,
368 369 370 371 372 373 374
			     "Unrecognised inode hash code %d",
			     root->info.hash_version);
		brelse(bh);
		*err = ERR_BAD_DX_DIR;
		goto fail;
	}
	hinfo->hash_version = root->info.hash_version;
375
	hinfo->seed = EXT4_SB(dir->i_sb)->s_hash_seed;
376 377
	if (d_name)
		ext4fs_dirhash(d_name->name, d_name->len, hinfo);
378 379 380
	hash = hinfo->hash;

	if (root->info.unused_flags & 1) {
381
		ext4_warning(dir->i_sb, __func__,
382 383 384 385 386 387 388 389
			     "Unimplemented inode hash flags: %#06x",
			     root->info.unused_flags);
		brelse(bh);
		*err = ERR_BAD_DX_DIR;
		goto fail;
	}

	if ((indirect = root->info.indirect_levels) > 1) {
390
		ext4_warning(dir->i_sb, __func__,
391 392 393 394 395 396 397 398 399
			     "Unimplemented inode hash depth: %#06x",
			     root->info.indirect_levels);
		brelse(bh);
		*err = ERR_BAD_DX_DIR;
		goto fail;
	}

	entries = (struct dx_entry *) (((char *)&root->info) +
				       root->info.info_length);
400 401 402

	if (dx_get_limit(entries) != dx_root_limit(dir,
						   root->info.info_length)) {
403
		ext4_warning(dir->i_sb, __func__,
404 405 406 407 408 409
			     "dx entry: limit != root limit");
		brelse(bh);
		*err = ERR_BAD_DX_DIR;
		goto fail;
	}

410
	dxtrace(printk("Look up %x", hash));
411 412 413
	while (1)
	{
		count = dx_get_count(entries);
414
		if (!count || count > dx_get_limit(entries)) {
415
			ext4_warning(dir->i_sb, __func__,
416 417 418 419 420 421
				     "dx entry: no count or count > limit");
			brelse(bh);
			*err = ERR_BAD_DX_DIR;
			goto fail2;
		}

422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455
		p = entries + 1;
		q = entries + count - 1;
		while (p <= q)
		{
			m = p + (q - p)/2;
			dxtrace(printk("."));
			if (dx_get_hash(m) > hash)
				q = m - 1;
			else
				p = m + 1;
		}

		if (0) // linear search cross check
		{
			unsigned n = count - 1;
			at = entries;
			while (n--)
			{
				dxtrace(printk(","));
				if (dx_get_hash(++at) > hash)
				{
					at--;
					break;
				}
			}
			assert (at == p - 1);
		}

		at = p - 1;
		dxtrace(printk(" %x->%u\n", at == entries? 0: dx_get_hash(at), dx_get_block(at)));
		frame->bh = bh;
		frame->entries = entries;
		frame->at = at;
		if (!indirect--) return frame;
456
		if (!(bh = ext4_bread (NULL,dir, dx_get_block(at), 0, err)))
457 458
			goto fail2;
		at = entries = ((struct dx_node *) bh->b_data)->entries;
459
		if (dx_get_limit(entries) != dx_node_limit (dir)) {
460
			ext4_warning(dir->i_sb, __func__,
461 462 463 464 465
				     "dx entry: limit != node limit");
			brelse(bh);
			*err = ERR_BAD_DX_DIR;
			goto fail2;
		}
466
		frame++;
467
		frame->bh = NULL;
468 469 470 471 472 473 474
	}
fail2:
	while (frame >= frame_in) {
		brelse(frame->bh);
		frame--;
	}
fail:
475
	if (*err == ERR_BAD_DX_DIR)
476
		ext4_warning(dir->i_sb, __func__,
477 478
			     "Corrupt dir inode %ld, running e2fsck is "
			     "recommended.", dir->i_ino);
479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508
	return NULL;
}

static void dx_release (struct dx_frame *frames)
{
	if (frames[0].bh == NULL)
		return;

	if (((struct dx_root *) frames[0].bh->b_data)->info.indirect_levels)
		brelse(frames[1].bh);
	brelse(frames[0].bh);
}

/*
 * This function increments the frame pointer to search the next leaf
 * block, and reads in the necessary intervening nodes if the search
 * should be necessary.  Whether or not the search is necessary is
 * controlled by the hash parameter.  If the hash value is even, then
 * the search is only continued if the next block starts with that
 * hash value.  This is used if we are searching for a specific file.
 *
 * If the hash value is HASH_NB_ALWAYS, then always go to the next block.
 *
 * This function returns 1 if the caller should continue to search,
 * or 0 if it should not.  If there is an error reading one of the
 * index blocks, it will a negative error code.
 *
 * If start_hash is non-null, it will be filled in with the starting
 * hash of the next page.
 */
509
static int ext4_htree_next_block(struct inode *dir, __u32 hash,
510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554
				 struct dx_frame *frame,
				 struct dx_frame *frames,
				 __u32 *start_hash)
{
	struct dx_frame *p;
	struct buffer_head *bh;
	int err, num_frames = 0;
	__u32 bhash;

	p = frame;
	/*
	 * Find the next leaf page by incrementing the frame pointer.
	 * If we run out of entries in the interior node, loop around and
	 * increment pointer in the parent node.  When we break out of
	 * this loop, num_frames indicates the number of interior
	 * nodes need to be read.
	 */
	while (1) {
		if (++(p->at) < p->entries + dx_get_count(p->entries))
			break;
		if (p == frames)
			return 0;
		num_frames++;
		p--;
	}

	/*
	 * If the hash is 1, then continue only if the next page has a
	 * continuation hash of any value.  This is used for readdir
	 * handling.  Otherwise, check to see if the hash matches the
	 * desired contiuation hash.  If it doesn't, return since
	 * there's no point to read in the successive index pages.
	 */
	bhash = dx_get_hash(p->at);
	if (start_hash)
		*start_hash = bhash;
	if ((hash & 1) == 0) {
		if ((bhash & ~1) != hash)
			return 0;
	}
	/*
	 * If the hash is HASH_NB_ALWAYS, we always go to the next
	 * block so no check is necessary
	 */
	while (num_frames--) {
555
		if (!(bh = ext4_bread(NULL, dir, dx_get_block(p->at),
556 557 558
				      0, &err)))
			return err; /* Failure */
		p++;
559
		brelse(p->bh);
560 561 562 563 564 565 566 567 568 569 570 571 572
		p->bh = bh;
		p->at = p->entries = ((struct dx_node *) bh->b_data)->entries;
	}
	return 1;
}


/*
 * This function fills a red-black tree with information from a
 * directory block.  It returns the number directory entries loaded
 * into the tree.  If there is an error it is returned in err.
 */
static int htree_dirblock_to_tree(struct file *dir_file,
A
Aneesh Kumar K.V 已提交
573
				  struct inode *dir, ext4_lblk_t block,
574 575 576 577
				  struct dx_hash_info *hinfo,
				  __u32 start_hash, __u32 start_minor_hash)
{
	struct buffer_head *bh;
578
	struct ext4_dir_entry_2 *de, *top;
579 580
	int err, count = 0;

A
Aneesh Kumar K.V 已提交
581 582
	dxtrace(printk(KERN_INFO "In htree dirblock_to_tree: block %lu\n",
							(unsigned long)block));
583
	if (!(bh = ext4_bread (NULL, dir, block, 0, &err)))
584 585
		return err;

586 587
	de = (struct ext4_dir_entry_2 *) bh->b_data;
	top = (struct ext4_dir_entry_2 *) ((char *) de +
588
					   dir->i_sb->s_blocksize -
589 590
					   EXT4_DIR_REC_LEN(0));
	for (; de < top; de = ext4_next_entry(de)) {
591 592 593 594 595 596
		if (!ext4_check_dir_entry("htree_dirblock_to_tree", dir, de, bh,
					(block<<EXT4_BLOCK_SIZE_BITS(dir->i_sb))
						+((char *)de - bh->b_data))) {
			/* On error, skip the f_pos to the next block. */
			dir_file->f_pos = (dir_file->f_pos |
					(dir->i_sb->s_blocksize - 1)) + 1;
597
			brelse(bh);
598 599
			return count;
		}
600
		ext4fs_dirhash(de->name, de->name_len, hinfo);
601 602 603 604 605 606
		if ((hinfo->hash < start_hash) ||
		    ((hinfo->hash == start_hash) &&
		     (hinfo->minor_hash < start_minor_hash)))
			continue;
		if (de->inode == 0)
			continue;
607
		if ((err = ext4_htree_store_dirent(dir_file,
608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626
				   hinfo->hash, hinfo->minor_hash, de)) != 0) {
			brelse(bh);
			return err;
		}
		count++;
	}
	brelse(bh);
	return count;
}


/*
 * This function fills a red-black tree with information from a
 * directory.  We start scanning the directory in hash order, starting
 * at start_hash and start_minor_hash.
 *
 * This function returns the number of entries inserted into the tree,
 * or a negative error code.
 */
627
int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
628 629 630
			 __u32 start_minor_hash, __u32 *next_hash)
{
	struct dx_hash_info hinfo;
631
	struct ext4_dir_entry_2 *de;
632 633
	struct dx_frame frames[2], *frame;
	struct inode *dir;
A
Aneesh Kumar K.V 已提交
634
	ext4_lblk_t block;
635
	int count = 0;
A
Aneesh Kumar K.V 已提交
636
	int ret, err;
637 638
	__u32 hashval;

639 640
	dxtrace(printk(KERN_DEBUG "In htree_fill_tree, start hash: %x:%x\n", 
		       start_hash, start_minor_hash));
641
	dir = dir_file->f_path.dentry->d_inode;
642 643 644
	if (!(EXT4_I(dir)->i_flags & EXT4_INDEX_FL)) {
		hinfo.hash_version = EXT4_SB(dir->i_sb)->s_def_hash_version;
		hinfo.seed = EXT4_SB(dir->i_sb)->s_hash_seed;
645 646 647 648 649 650 651
		count = htree_dirblock_to_tree(dir_file, dir, 0, &hinfo,
					       start_hash, start_minor_hash);
		*next_hash = ~0;
		return count;
	}
	hinfo.hash = start_hash;
	hinfo.minor_hash = 0;
652
	frame = dx_probe(NULL, dir, &hinfo, frames, &err);
653 654 655 656 657
	if (!frame)
		return err;

	/* Add '.' and '..' from the htree header */
	if (!start_hash && !start_minor_hash) {
658 659
		de = (struct ext4_dir_entry_2 *) frames[0].bh->b_data;
		if ((err = ext4_htree_store_dirent(dir_file, 0, 0, de)) != 0)
660 661 662 663
			goto errout;
		count++;
	}
	if (start_hash < 2 || (start_hash ==2 && start_minor_hash==0)) {
664 665 666
		de = (struct ext4_dir_entry_2 *) frames[0].bh->b_data;
		de = ext4_next_entry(de);
		if ((err = ext4_htree_store_dirent(dir_file, 2, 0, de)) != 0)
667 668 669 670 671 672 673 674 675 676 677 678 679 680
			goto errout;
		count++;
	}

	while (1) {
		block = dx_get_block(frame->at);
		ret = htree_dirblock_to_tree(dir_file, dir, block, &hinfo,
					     start_hash, start_minor_hash);
		if (ret < 0) {
			err = ret;
			goto errout;
		}
		count += ret;
		hashval = ~0;
681
		ret = ext4_htree_next_block(dir, HASH_NB_ALWAYS,
682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697
					    frame, frames, &hashval);
		*next_hash = hashval;
		if (ret < 0) {
			err = ret;
			goto errout;
		}
		/*
		 * Stop if:  (a) there are no more entries, or
		 * (b) we have inserted at least one entry and the
		 * next hash value is not a continuation
		 */
		if ((ret == 0) ||
		    (count && ((hashval & 1) == 0)))
			break;
	}
	dx_release(frames);
698 699
	dxtrace(printk(KERN_DEBUG "Fill tree: returned %d entries, "
		       "next hash: %x\n", count, *next_hash));
700 701 702 703 704 705 706 707 708 709 710
	return count;
errout:
	dx_release(frames);
	return (err);
}


/*
 * Directory block splitting, compacting
 */

711 712 713 714
/*
 * Create map of hash values, offsets, and sizes, stored at end of block.
 * Returns number of entries mapped.
 */
715
static int dx_make_map (struct ext4_dir_entry_2 *de, int size,
716 717 718 719 720 721 722 723 724
			struct dx_hash_info *hinfo, struct dx_map_entry *map_tail)
{
	int count = 0;
	char *base = (char *) de;
	struct dx_hash_info h = *hinfo;

	while ((char *) de < base + size)
	{
		if (de->name_len && de->inode) {
725
			ext4fs_dirhash(de->name, de->name_len, &h);
726 727
			map_tail--;
			map_tail->hash = h.hash;
728 729
			map_tail->offs = (u16) ((char *) de - base);
			map_tail->size = le16_to_cpu(de->rec_len);
730 731 732 733
			count++;
			cond_resched();
		}
		/* XXX: do we need to check rec_len == 0 case? -Chris */
734
		de = ext4_next_entry(de);
735 736 737 738
	}
	return count;
}

739
/* Sort map by hash value */
740 741
static void dx_sort_map (struct dx_map_entry *map, unsigned count)
{
A
Andrew Morton 已提交
742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758
	struct dx_map_entry *p, *q, *top = map + count - 1;
	int more;
	/* Combsort until bubble sort doesn't suck */
	while (count > 2) {
		count = count*10/13;
		if (count - 9 < 2) /* 9, 10 -> 11 */
			count = 11;
		for (p = top, q = p - count; q >= map; p--, q--)
			if (p->hash < q->hash)
				swap(*p, *q);
	}
	/* Garden variety bubble sort */
	do {
		more = 0;
		q = top;
		while (q-- > map) {
			if (q[1].hash >= q[0].hash)
759
				continue;
A
Andrew Morton 已提交
760 761
			swap(*(q+1), *q);
			more = 1;
762 763 764 765
		}
	} while(more);
}

A
Aneesh Kumar K.V 已提交
766
static void dx_insert_block(struct dx_frame *frame, u32 hash, ext4_lblk_t block)
767 768 769 770 771 772 773 774 775 776 777 778 779
{
	struct dx_entry *entries = frame->entries;
	struct dx_entry *old = frame->at, *new = old + 1;
	int count = dx_get_count(entries);

	assert(count < dx_get_limit(entries));
	assert(old < entries + count);
	memmove(new + 1, new, (char *)(entries + count) - (char *)(new));
	dx_set_hash(new, hash);
	dx_set_block(new, block);
	dx_set_count(entries, count + 1);
}

780
static void ext4_update_dx_flag(struct inode *inode)
781
{
782 783 784
	if (!EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
				     EXT4_FEATURE_COMPAT_DIR_INDEX))
		EXT4_I(inode)->i_flags &= ~EXT4_INDEX_FL;
785 786 787
}

/*
788
 * NOTE! unlike strncmp, ext4_match returns 1 for success, 0 for failure.
789
 *
790
 * `len <= EXT4_NAME_LEN' is guaranteed by caller.
791 792
 * `de != NULL' is guaranteed by caller.
 */
793 794
static inline int ext4_match (int len, const char * const name,
			      struct ext4_dir_entry_2 * de)
795 796 797 798 799 800 801 802 803 804 805
{
	if (len != de->name_len)
		return 0;
	if (!de->inode)
		return 0;
	return !memcmp(name, de->name, len);
}

/*
 * Returns 0 if not found, -1 on failure, and 1 on success
 */
806
static inline int search_dirblock(struct buffer_head *bh,
807
				  struct inode *dir,
808
				  const struct qstr *d_name,
809
				  unsigned long offset,
810
				  struct ext4_dir_entry_2 ** res_dir)
811
{
812
	struct ext4_dir_entry_2 * de;
813 814
	char * dlimit;
	int de_len;
815 816
	const char *name = d_name->name;
	int namelen = d_name->len;
817

818
	de = (struct ext4_dir_entry_2 *) bh->b_data;
819 820 821 822 823 824
	dlimit = bh->b_data + dir->i_sb->s_blocksize;
	while ((char *) de < dlimit) {
		/* this code is executed quadratically often */
		/* do minimal checking `by hand' */

		if ((char *) de + namelen <= dlimit &&
825
		    ext4_match (namelen, name, de)) {
826
			/* found a match - just to be sure, do a full check */
827
			if (!ext4_check_dir_entry("ext4_find_entry",
828 829 830 831 832 833
						  dir, de, bh, offset))
				return -1;
			*res_dir = de;
			return 1;
		}
		/* prevent looping on a bad block */
834
		de_len = ext4_rec_len_from_disk(de->rec_len);
835 836 837
		if (de_len <= 0)
			return -1;
		offset += de_len;
838
		de = (struct ext4_dir_entry_2 *) ((char *) de + de_len);
839 840 841 842 843 844
	}
	return 0;
}


/*
845
 *	ext4_find_entry()
846 847 848 849 850 851 852 853 854
 *
 * finds an entry in the specified directory with the wanted name. It
 * returns the cache buffer in which the entry was found, and the entry
 * itself (as a parameter - res_dir). It does NOT read the inode of the
 * entry - you'll have to do that yourself if you want to.
 *
 * The returned buffer_head has ->b_count elevated.  The caller is expected
 * to brelse() it when appropriate.
 */
855 856
static struct buffer_head * ext4_find_entry (struct inode *dir,
					const struct qstr *d_name,
857
					struct ext4_dir_entry_2 ** res_dir)
858
{
859 860 861
	struct super_block *sb;
	struct buffer_head *bh_use[NAMEI_RA_SIZE];
	struct buffer_head *bh, *ret = NULL;
A
Aneesh Kumar K.V 已提交
862
	ext4_lblk_t start, block, b;
863 864 865 866 867
	int ra_max = 0;		/* Number of bh's in the readahead
				   buffer, bh_use[] */
	int ra_ptr = 0;		/* Current index into readahead
				   buffer */
	int num = 0;
A
Aneesh Kumar K.V 已提交
868 869
	ext4_lblk_t  nblocks;
	int i, err;
870 871 872 873
	int namelen;

	*res_dir = NULL;
	sb = dir->i_sb;
874
	namelen = d_name->len;
875
	if (namelen > EXT4_NAME_LEN)
876 877
		return NULL;
	if (is_dx(dir)) {
878
		bh = ext4_dx_find_entry(dir, d_name, res_dir, &err);
879 880 881 882 883 884 885
		/*
		 * On success, or if the error was file not found,
		 * return.  Otherwise, fall back to doing a search the
		 * old fashioned way.
		 */
		if (bh || (err != ERR_BAD_DX_DIR))
			return bh;
886 887
		dxtrace(printk(KERN_DEBUG "ext4_find_entry: dx failed, "
			       "falling back\n"));
888
	}
889 890
	nblocks = dir->i_size >> EXT4_BLOCK_SIZE_BITS(sb);
	start = EXT4_I(dir)->i_dir_start_lookup;
891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913
	if (start >= nblocks)
		start = 0;
	block = start;
restart:
	do {
		/*
		 * We deal with the read-ahead logic here.
		 */
		if (ra_ptr >= ra_max) {
			/* Refill the readahead buffer */
			ra_ptr = 0;
			b = block;
			for (ra_max = 0; ra_max < NAMEI_RA_SIZE; ra_max++) {
				/*
				 * Terminate if we reach the end of the
				 * directory and must wrap, or if our
				 * search has finished at this block.
				 */
				if (b >= nblocks || (num && block == start)) {
					bh_use[ra_max] = NULL;
					break;
				}
				num++;
914
				bh = ext4_getblk(NULL, dir, b++, 0, &err);
915 916 917 918 919 920 921 922 923 924
				bh_use[ra_max] = bh;
				if (bh)
					ll_rw_block(READ_META, 1, &bh);
			}
		}
		if ((bh = bh_use[ra_ptr++]) == NULL)
			goto next;
		wait_on_buffer(bh);
		if (!buffer_uptodate(bh)) {
			/* read error, skip block & hope for the best */
925
			ext4_error(sb, __func__, "reading directory #%lu "
A
Aneesh Kumar K.V 已提交
926 927
				   "offset %lu", dir->i_ino,
				   (unsigned long)block);
928 929 930
			brelse(bh);
			goto next;
		}
931
		i = search_dirblock(bh, dir, d_name,
932
			    block << EXT4_BLOCK_SIZE_BITS(sb), res_dir);
933
		if (i == 1) {
934
			EXT4_I(dir)->i_dir_start_lookup = block;
935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951
			ret = bh;
			goto cleanup_and_exit;
		} else {
			brelse(bh);
			if (i < 0)
				goto cleanup_and_exit;
		}
	next:
		if (++block >= nblocks)
			block = 0;
	} while (block != start);

	/*
	 * If the directory has grown while we were searching, then
	 * search the last part of the directory before giving up.
	 */
	block = nblocks;
952
	nblocks = dir->i_size >> EXT4_BLOCK_SIZE_BITS(sb);
953 954 955 956 957 958 959 960
	if (block < nblocks) {
		start = 0;
		goto restart;
	}

cleanup_and_exit:
	/* Clean up the read-ahead blocks */
	for (; ra_ptr < ra_max; ra_ptr++)
961
		brelse(bh_use[ra_ptr]);
962 963 964
	return ret;
}

965
static struct buffer_head * ext4_dx_find_entry(struct inode *dir, const struct qstr *d_name,
966
		       struct ext4_dir_entry_2 **res_dir, int *err)
967 968 969 970 971
{
	struct super_block * sb;
	struct dx_hash_info	hinfo;
	u32 hash;
	struct dx_frame frames[2], *frame;
972
	struct ext4_dir_entry_2 *de, *top;
973
	struct buffer_head *bh;
A
Aneesh Kumar K.V 已提交
974
	ext4_lblk_t block;
975
	int retval;
976 977
	int namelen = d_name->len;
	const u8 *name = d_name->name;
978 979 980 981

	sb = dir->i_sb;
	/* NFS may look up ".." - look at dx_root directory block */
	if (namelen > 2 || name[0] != '.'||(name[1] != '.' && name[1] != '\0')){
982
		if (!(frame = dx_probe(d_name, dir, &hinfo, frames, err)))
983 984 985 986 987 988 989 990 991 992
			return NULL;
	} else {
		frame = frames;
		frame->bh = NULL;			/* for dx_release() */
		frame->at = (struct dx_entry *)frames;	/* hack for zero entry*/
		dx_set_block(frame->at, 0);		/* dx_root block is 0 */
	}
	hash = hinfo.hash;
	do {
		block = dx_get_block(frame->at);
993
		if (!(bh = ext4_bread (NULL,dir, block, 0, err)))
994
			goto errout;
995 996 997
		de = (struct ext4_dir_entry_2 *) bh->b_data;
		top = (struct ext4_dir_entry_2 *) ((char *) de + sb->s_blocksize -
				       EXT4_DIR_REC_LEN(0));
998 999 1000 1001 1002 1003
		for (; de < top; de = ext4_next_entry(de)) {
			int off = (block << EXT4_BLOCK_SIZE_BITS(sb))
				  + ((char *) de - bh->b_data);

			if (!ext4_check_dir_entry(__func__, dir, de, bh, off)) {
				brelse(bh);
1004
				*err = ERR_BAD_DX_DIR;
1005 1006
				goto errout;
			}
1007 1008 1009 1010 1011 1012

			if (ext4_match(namelen, name, de)) {
				*res_dir = de;
				dx_release(frames);
				return bh;
			}
1013
		}
1014
		brelse(bh);
1015
		/* Check to see if we should continue to search */
1016
		retval = ext4_htree_next_block(dir, hash, frame,
1017 1018
					       frames, NULL);
		if (retval < 0) {
1019
			ext4_warning(sb, __func__,
1020 1021 1022 1023 1024 1025 1026 1027 1028
			     "error reading index page in directory #%lu",
			     dir->i_ino);
			*err = retval;
			goto errout;
		}
	} while (retval == 1);

	*err = -ENOENT;
errout:
1029
	dxtrace(printk(KERN_DEBUG "%s not found\n", name));
1030 1031 1032 1033
	dx_release (frames);
	return NULL;
}

1034
static struct dentry *ext4_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
1035
{
1036 1037 1038
	struct inode *inode;
	struct ext4_dir_entry_2 *de;
	struct buffer_head *bh;
1039

1040
	if (dentry->d_name.len > EXT4_NAME_LEN)
1041 1042
		return ERR_PTR(-ENAMETOOLONG);

1043
	bh = ext4_find_entry(dir, &dentry->d_name, &de);
1044 1045 1046
	inode = NULL;
	if (bh) {
		unsigned long ino = le32_to_cpu(de->inode);
1047
		brelse(bh);
1048 1049
		if (!ext4_valid_inum(dir->i_sb, ino)) {
			ext4_error(dir->i_sb, "ext4_lookup",
1050
				   "bad inode number: %lu", ino);
1051
			return ERR_PTR(-EIO);
1052
		}
1053 1054 1055
		inode = ext4_iget(dir->i_sb, ino);
		if (IS_ERR(inode))
			return ERR_CAST(inode);
1056 1057 1058 1059 1060
	}
	return d_splice_alias(inode, dentry);
}


1061
struct dentry *ext4_get_parent(struct dentry *child)
1062 1063 1064
{
	unsigned long ino;
	struct inode *inode;
1065 1066 1067 1068
	static const struct qstr dotdot = {
		.name = "..",
		.len = 2,
	};
1069
	struct ext4_dir_entry_2 * de;
1070 1071
	struct buffer_head *bh;

1072
	bh = ext4_find_entry(child->d_inode, &dotdot, &de);
1073 1074 1075 1076 1077 1078
	inode = NULL;
	if (!bh)
		return ERR_PTR(-ENOENT);
	ino = le32_to_cpu(de->inode);
	brelse(bh);

1079 1080
	if (!ext4_valid_inum(child->d_inode->i_sb, ino)) {
		ext4_error(child->d_inode->i_sb, "ext4_get_parent",
1081
			   "bad inode number: %lu", ino);
1082
		return ERR_PTR(-EIO);
1083 1084
	}

1085
	return d_obtain_alias(ext4_iget(child->d_inode->i_sb, ino));
1086 1087 1088
}

#define S_SHIFT 12
1089 1090 1091 1092 1093 1094 1095 1096
static unsigned char ext4_type_by_mode[S_IFMT >> S_SHIFT] = {
	[S_IFREG >> S_SHIFT]	= EXT4_FT_REG_FILE,
	[S_IFDIR >> S_SHIFT]	= EXT4_FT_DIR,
	[S_IFCHR >> S_SHIFT]	= EXT4_FT_CHRDEV,
	[S_IFBLK >> S_SHIFT]	= EXT4_FT_BLKDEV,
	[S_IFIFO >> S_SHIFT]	= EXT4_FT_FIFO,
	[S_IFSOCK >> S_SHIFT]	= EXT4_FT_SOCK,
	[S_IFLNK >> S_SHIFT]	= EXT4_FT_SYMLINK,
1097 1098
};

1099 1100
static inline void ext4_set_de_type(struct super_block *sb,
				struct ext4_dir_entry_2 *de,
1101
				umode_t mode) {
1102 1103
	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE))
		de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
1104 1105
}

1106 1107 1108 1109
/*
 * Move count entries from end of map between two memory locations.
 * Returns pointer to last entry moved.
 */
1110
static struct ext4_dir_entry_2 *
1111 1112 1113 1114 1115
dx_move_dirents(char *from, char *to, struct dx_map_entry *map, int count)
{
	unsigned rec_len = 0;

	while (count--) {
1116 1117
		struct ext4_dir_entry_2 *de = (struct ext4_dir_entry_2 *) (from + map->offs);
		rec_len = EXT4_DIR_REC_LEN(de->name_len);
1118
		memcpy (to, de, rec_len);
1119
		((struct ext4_dir_entry_2 *) to)->rec_len =
1120
				ext4_rec_len_to_disk(rec_len);
1121 1122 1123 1124
		de->inode = 0;
		map++;
		to += rec_len;
	}
1125
	return (struct ext4_dir_entry_2 *) (to - rec_len);
1126 1127
}

1128 1129 1130 1131
/*
 * Compact each dir entry in the range to the minimal rec_len.
 * Returns pointer to last entry in range.
 */
1132
static struct ext4_dir_entry_2* dx_pack_dirents(char *base, int size)
1133
{
1134
	struct ext4_dir_entry_2 *next, *to, *prev, *de = (struct ext4_dir_entry_2 *) base;
1135 1136 1137 1138
	unsigned rec_len = 0;

	prev = to = de;
	while ((char*)de < base + size) {
1139
		next = ext4_next_entry(de);
1140
		if (de->inode && de->name_len) {
1141
			rec_len = EXT4_DIR_REC_LEN(de->name_len);
1142 1143
			if (de > to)
				memmove(to, de, rec_len);
1144
			to->rec_len = ext4_rec_len_to_disk(rec_len);
1145
			prev = to;
1146
			to = (struct ext4_dir_entry_2 *) (((char *) to) + rec_len);
1147 1148 1149 1150 1151 1152
		}
		de = next;
	}
	return prev;
}

1153 1154 1155 1156 1157
/*
 * Split a full leaf block to make room for a new dir entry.
 * Allocate a new block, and move entries so that they are approx. equally full.
 * Returns pointer to de in block into which the new entry will be inserted.
 */
1158
static struct ext4_dir_entry_2 *do_split(handle_t *handle, struct inode *dir,
1159 1160 1161 1162 1163 1164
			struct buffer_head **bh,struct dx_frame *frame,
			struct dx_hash_info *hinfo, int *error)
{
	unsigned blocksize = dir->i_sb->s_blocksize;
	unsigned count, continued;
	struct buffer_head *bh2;
A
Aneesh Kumar K.V 已提交
1165
	ext4_lblk_t newblock;
1166 1167 1168
	u32 hash2;
	struct dx_map_entry *map;
	char *data1 = (*bh)->b_data, *data2;
1169
	unsigned split, move, size, i;
1170
	struct ext4_dir_entry_2 *de = NULL, *de2;
1171
	int	err = 0;
1172

1173
	bh2 = ext4_append (handle, dir, &newblock, &err);
1174 1175 1176 1177 1178 1179 1180
	if (!(bh2)) {
		brelse(*bh);
		*bh = NULL;
		goto errout;
	}

	BUFFER_TRACE(*bh, "get_write_access");
1181
	err = ext4_journal_get_write_access(handle, *bh);
1182 1183 1184
	if (err)
		goto journal_error;

1185
	BUFFER_TRACE(frame->bh, "get_write_access");
1186
	err = ext4_journal_get_write_access(handle, frame->bh);
1187 1188 1189 1190 1191 1192 1193
	if (err)
		goto journal_error;

	data2 = bh2->b_data;

	/* create map in the end of data2 block */
	map = (struct dx_map_entry *) (data2 + blocksize);
1194
	count = dx_make_map((struct ext4_dir_entry_2 *) data1,
1195 1196
			     blocksize, hinfo, map);
	map -= count;
1197
	dx_sort_map(map, count);
1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209
	/* Split the existing block in the middle, size-wise */
	size = 0;
	move = 0;
	for (i = count-1; i >= 0; i--) {
		/* is more than half of this entry in 2nd half of the block? */
		if (size + map[i].size/2 > blocksize/2)
			break;
		size += map[i].size;
		move++;
	}
	/* map index at which we will split */
	split = count - move;
1210 1211
	hash2 = map[split].hash;
	continued = hash2 == map[split - 1].hash;
A
Aneesh Kumar K.V 已提交
1212 1213 1214
	dxtrace(printk(KERN_INFO "Split block %lu at %x, %i/%i\n",
			(unsigned long)dx_get_block(frame->at),
					hash2, split, count-split));
1215 1216 1217

	/* Fancy dance to stay within two buffers */
	de2 = dx_move_dirents(data1, data2, map + split, count - split);
1218
	de = dx_pack_dirents(data1, blocksize);
1219 1220
	de->rec_len = ext4_rec_len_to_disk(data1 + blocksize - (char *) de);
	de2->rec_len = ext4_rec_len_to_disk(data2 + blocksize - (char *) de2);
1221 1222
	dxtrace(dx_show_leaf (hinfo, (struct ext4_dir_entry_2 *) data1, blocksize, 1));
	dxtrace(dx_show_leaf (hinfo, (struct ext4_dir_entry_2 *) data2, blocksize, 1));
1223 1224 1225 1226 1227 1228 1229

	/* Which block gets the new entry? */
	if (hinfo->hash >= hash2)
	{
		swap(*bh, bh2);
		de = de2;
	}
1230 1231
	dx_insert_block(frame, hash2 + continued, newblock);
	err = ext4_journal_dirty_metadata(handle, bh2);
1232 1233
	if (err)
		goto journal_error;
1234
	err = ext4_journal_dirty_metadata(handle, frame->bh);
1235 1236
	if (err)
		goto journal_error;
1237 1238
	brelse(bh2);
	dxtrace(dx_show_index("frame", frame->entries));
1239
	return de;
1240 1241 1242 1243 1244 1245 1246 1247 1248

journal_error:
	brelse(*bh);
	brelse(bh2);
	*bh = NULL;
	ext4_std_error(dir->i_sb, err);
errout:
	*error = err;
	return NULL;
1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262
}

/*
 * Add a new entry into a directory (leaf) block.  If de is non-NULL,
 * it points to a directory entry which is guaranteed to be large
 * enough for new directory entry.  If de is NULL, then
 * add_dirent_to_buf will attempt search the directory block for
 * space.  It will return -ENOSPC if no space is available, and -EIO
 * and -EEXIST if directory entry already exists.
 *
 * NOTE!  bh is NOT released in the case where ENOSPC is returned.  In
 * all other cases bh is released.
 */
static int add_dirent_to_buf(handle_t *handle, struct dentry *dentry,
1263
			     struct inode *inode, struct ext4_dir_entry_2 *de,
1264
			     struct buffer_head *bh)
1265 1266 1267 1268 1269 1270 1271 1272 1273
{
	struct inode	*dir = dentry->d_parent->d_inode;
	const char	*name = dentry->d_name.name;
	int		namelen = dentry->d_name.len;
	unsigned long	offset = 0;
	unsigned short	reclen;
	int		nlen, rlen, err;
	char		*top;

1274
	reclen = EXT4_DIR_REC_LEN(namelen);
1275
	if (!de) {
1276
		de = (struct ext4_dir_entry_2 *)bh->b_data;
1277 1278
		top = bh->b_data + dir->i_sb->s_blocksize - reclen;
		while ((char *) de <= top) {
1279
			if (!ext4_check_dir_entry("ext4_add_entry", dir, de,
1280
						  bh, offset)) {
1281
				brelse(bh);
1282 1283
				return -EIO;
			}
1284 1285
			if (ext4_match(namelen, name, de)) {
				brelse(bh);
1286 1287
				return -EEXIST;
			}
1288
			nlen = EXT4_DIR_REC_LEN(de->name_len);
1289
			rlen = ext4_rec_len_from_disk(de->rec_len);
1290 1291
			if ((de->inode? rlen - nlen: rlen) >= reclen)
				break;
1292
			de = (struct ext4_dir_entry_2 *)((char *)de + rlen);
1293 1294 1295 1296 1297 1298
			offset += rlen;
		}
		if ((char *) de > top)
			return -ENOSPC;
	}
	BUFFER_TRACE(bh, "get_write_access");
1299
	err = ext4_journal_get_write_access(handle, bh);
1300
	if (err) {
1301
		ext4_std_error(dir->i_sb, err);
1302 1303 1304 1305 1306
		brelse(bh);
		return err;
	}

	/* By now the buffer is marked for journaling */
1307
	nlen = EXT4_DIR_REC_LEN(de->name_len);
1308
	rlen = ext4_rec_len_from_disk(de->rec_len);
1309
	if (de->inode) {
1310
		struct ext4_dir_entry_2 *de1 = (struct ext4_dir_entry_2 *)((char *)de + nlen);
1311 1312
		de1->rec_len = ext4_rec_len_to_disk(rlen - nlen);
		de->rec_len = ext4_rec_len_to_disk(nlen);
1313 1314
		de = de1;
	}
1315
	de->file_type = EXT4_FT_UNKNOWN;
1316 1317
	if (inode) {
		de->inode = cpu_to_le32(inode->i_ino);
1318
		ext4_set_de_type(dir->i_sb, de, inode->i_mode);
1319 1320 1321
	} else
		de->inode = 0;
	de->name_len = namelen;
1322
	memcpy(de->name, name, namelen);
1323 1324 1325 1326 1327 1328
	/*
	 * XXX shouldn't update any times until successful
	 * completion of syscall, but too many callers depend
	 * on this.
	 *
	 * XXX similarly, too many callers depend on
1329
	 * ext4_new_inode() setting the times, but error
1330 1331 1332 1333
	 * recovery deletes the inode, so the worst that can
	 * happen is that the times are slightly out of date
	 * and/or different from the directory change time.
	 */
K
Kalpak Shah 已提交
1334
	dir->i_mtime = dir->i_ctime = ext4_current_time(dir);
1335
	ext4_update_dx_flag(dir);
1336
	dir->i_version++;
1337 1338 1339
	ext4_mark_inode_dirty(handle, dir);
	BUFFER_TRACE(bh, "call ext4_journal_dirty_metadata");
	err = ext4_journal_dirty_metadata(handle, bh);
1340
	if (err)
1341
		ext4_std_error(dir->i_sb, err);
1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359
	brelse(bh);
	return 0;
}

/*
 * This converts a one block unindexed directory to a 3 block indexed
 * directory, and adds the dentry to the indexed directory.
 */
static int make_indexed_dir(handle_t *handle, struct dentry *dentry,
			    struct inode *inode, struct buffer_head *bh)
{
	struct inode	*dir = dentry->d_parent->d_inode;
	const char	*name = dentry->d_name.name;
	int		namelen = dentry->d_name.len;
	struct buffer_head *bh2;
	struct dx_root	*root;
	struct dx_frame	frames[2], *frame;
	struct dx_entry *entries;
1360
	struct ext4_dir_entry_2	*de, *de2;
1361 1362 1363 1364 1365
	char		*data1, *top;
	unsigned	len;
	int		retval;
	unsigned	blocksize;
	struct dx_hash_info hinfo;
A
Aneesh Kumar K.V 已提交
1366
	ext4_lblk_t  block;
1367 1368 1369
	struct fake_dirent *fde;

	blocksize =  dir->i_sb->s_blocksize;
1370
	dxtrace(printk(KERN_DEBUG "Creating index\n"));
1371
	retval = ext4_journal_get_write_access(handle, bh);
1372
	if (retval) {
1373
		ext4_std_error(dir->i_sb, retval);
1374 1375 1376 1377 1378
		brelse(bh);
		return retval;
	}
	root = (struct dx_root *) bh->b_data;

1379
	bh2 = ext4_append(handle, dir, &block, &retval);
1380 1381 1382 1383
	if (!(bh2)) {
		brelse(bh);
		return retval;
	}
1384
	EXT4_I(dir)->i_flags |= EXT4_INDEX_FL;
1385 1386 1387 1388
	data1 = bh2->b_data;

	/* The 0th block becomes the root, move the dirents out */
	fde = &root->dotdot;
1389 1390
	de = (struct ext4_dir_entry_2 *)((char *)fde +
		ext4_rec_len_from_disk(fde->rec_len));
1391 1392
	len = ((char *) root) + blocksize - (char *) de;
	memcpy (data1, de, len);
1393
	de = (struct ext4_dir_entry_2 *) data1;
1394
	top = data1 + len;
1395
	while ((char *)(de2 = ext4_next_entry(de)) < top)
1396
		de = de2;
1397
	de->rec_len = ext4_rec_len_to_disk(data1 + blocksize - (char *) de);
1398
	/* Initialize the root; the dot dirents already exist */
1399
	de = (struct ext4_dir_entry_2 *) (&root->dotdot);
1400
	de->rec_len = ext4_rec_len_to_disk(blocksize - EXT4_DIR_REC_LEN(2));
1401 1402
	memset (&root->info, 0, sizeof(root->info));
	root->info.info_length = sizeof(root->info);
1403
	root->info.hash_version = EXT4_SB(dir->i_sb)->s_def_hash_version;
1404
	entries = root->entries;
1405 1406 1407
	dx_set_block(entries, 1);
	dx_set_count(entries, 1);
	dx_set_limit(entries, dx_root_limit(dir, sizeof(root->info)));
1408 1409 1410

	/* Initialize as for dx_probe */
	hinfo.hash_version = root->info.hash_version;
1411 1412
	hinfo.seed = EXT4_SB(dir->i_sb)->s_hash_seed;
	ext4fs_dirhash(name, namelen, &hinfo);
1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426
	frame = frames;
	frame->entries = entries;
	frame->at = entries;
	frame->bh = bh;
	bh = bh2;
	de = do_split(handle,dir, &bh, frame, &hinfo, &retval);
	dx_release (frames);
	if (!(de))
		return retval;

	return add_dirent_to_buf(handle, dentry, inode, de, bh);
}

/*
1427
 *	ext4_add_entry()
1428 1429
 *
 * adds a file entry to the specified directory, using the same
1430
 * semantics as ext4_find_entry(). It returns NULL if it failed.
1431 1432 1433 1434 1435
 *
 * NOTE!! The inode part of 'de' is left at 0 - which means you
 * may not sleep between calling this and putting something into
 * the entry, as someone else might have used it while you slept.
 */
1436 1437
static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
			  struct inode *inode)
1438 1439 1440
{
	struct inode *dir = dentry->d_parent->d_inode;
	unsigned long offset;
1441
	struct buffer_head *bh;
1442
	struct ext4_dir_entry_2 *de;
1443
	struct super_block *sb;
1444 1445 1446
	int	retval;
	int	dx_fallback=0;
	unsigned blocksize;
A
Aneesh Kumar K.V 已提交
1447
	ext4_lblk_t block, blocks;
1448 1449 1450 1451 1452 1453

	sb = dir->i_sb;
	blocksize = sb->s_blocksize;
	if (!dentry->d_name.len)
		return -EINVAL;
	if (is_dx(dir)) {
1454
		retval = ext4_dx_add_entry(handle, dentry, inode);
1455 1456
		if (!retval || (retval != ERR_BAD_DX_DIR))
			return retval;
1457
		EXT4_I(dir)->i_flags &= ~EXT4_INDEX_FL;
1458
		dx_fallback++;
1459
		ext4_mark_inode_dirty(handle, dir);
1460 1461 1462
	}
	blocks = dir->i_size >> sb->s_blocksize_bits;
	for (block = 0, offset = 0; block < blocks; block++) {
1463
		bh = ext4_bread(handle, dir, block, 0, &retval);
1464 1465 1466 1467 1468 1469 1470
		if(!bh)
			return retval;
		retval = add_dirent_to_buf(handle, dentry, inode, NULL, bh);
		if (retval != -ENOSPC)
			return retval;

		if (blocks == 1 && !dx_fallback &&
1471
		    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX))
1472 1473 1474
			return make_indexed_dir(handle, dentry, inode, bh);
		brelse(bh);
	}
1475
	bh = ext4_append(handle, dir, &block, &retval);
1476 1477
	if (!bh)
		return retval;
1478
	de = (struct ext4_dir_entry_2 *) bh->b_data;
1479
	de->inode = 0;
1480
	de->rec_len = ext4_rec_len_to_disk(blocksize);
1481 1482 1483 1484 1485 1486
	return add_dirent_to_buf(handle, dentry, inode, de, bh);
}

/*
 * Returns 0 for success, or a negative error value
 */
1487
static int ext4_dx_add_entry(handle_t *handle, struct dentry *dentry,
1488 1489 1490 1491 1492
			     struct inode *inode)
{
	struct dx_frame frames[2], *frame;
	struct dx_entry *entries, *at;
	struct dx_hash_info hinfo;
1493
	struct buffer_head *bh;
1494
	struct inode *dir = dentry->d_parent->d_inode;
1495
	struct super_block *sb = dir->i_sb;
1496
	struct ext4_dir_entry_2 *de;
1497 1498
	int err;

1499
	frame = dx_probe(&dentry->d_name, dir, &hinfo, frames, &err);
1500 1501 1502 1503 1504
	if (!frame)
		return err;
	entries = frame->entries;
	at = frame->at;

1505
	if (!(bh = ext4_bread(handle,dir, dx_get_block(frame->at), 0, &err)))
1506 1507 1508
		goto cleanup;

	BUFFER_TRACE(bh, "get_write_access");
1509
	err = ext4_journal_get_write_access(handle, bh);
1510 1511 1512 1513 1514 1515 1516 1517 1518 1519
	if (err)
		goto journal_error;

	err = add_dirent_to_buf(handle, dentry, inode, NULL, bh);
	if (err != -ENOSPC) {
		bh = NULL;
		goto cleanup;
	}

	/* Block full, should compress but for now just split */
1520
	dxtrace(printk(KERN_DEBUG "using %u of %u node entries\n",
1521 1522 1523
		       dx_get_count(entries), dx_get_limit(entries)));
	/* Need to split index? */
	if (dx_get_count(entries) == dx_get_limit(entries)) {
A
Aneesh Kumar K.V 已提交
1524
		ext4_lblk_t newblock;
1525 1526 1527 1528 1529 1530 1531 1532
		unsigned icount = dx_get_count(entries);
		int levels = frame - frames;
		struct dx_entry *entries2;
		struct dx_node *node2;
		struct buffer_head *bh2;

		if (levels && (dx_get_count(frames->entries) ==
			       dx_get_limit(frames->entries))) {
1533
			ext4_warning(sb, __func__,
1534 1535 1536 1537
				     "Directory index full!");
			err = -ENOSPC;
			goto cleanup;
		}
1538
		bh2 = ext4_append (handle, dir, &newblock, &err);
1539 1540 1541 1542
		if (!(bh2))
			goto cleanup;
		node2 = (struct dx_node *)(bh2->b_data);
		entries2 = node2->entries;
1543
		node2->fake.rec_len = ext4_rec_len_to_disk(sb->s_blocksize);
1544 1545
		node2->fake.inode = 0;
		BUFFER_TRACE(frame->bh, "get_write_access");
1546
		err = ext4_journal_get_write_access(handle, frame->bh);
1547 1548 1549 1550 1551
		if (err)
			goto journal_error;
		if (levels) {
			unsigned icount1 = icount/2, icount2 = icount - icount1;
			unsigned hash2 = dx_get_hash(entries + icount1);
1552 1553
			dxtrace(printk(KERN_DEBUG "Split index %i/%i\n",
				       icount1, icount2));
1554 1555

			BUFFER_TRACE(frame->bh, "get_write_access"); /* index root */
1556
			err = ext4_journal_get_write_access(handle,
1557 1558 1559 1560
							     frames[0].bh);
			if (err)
				goto journal_error;

1561 1562 1563 1564 1565
			memcpy((char *) entries2, (char *) (entries + icount1),
			       icount2 * sizeof(struct dx_entry));
			dx_set_count(entries, icount1);
			dx_set_count(entries2, icount2);
			dx_set_limit(entries2, dx_node_limit(dir));
1566 1567 1568 1569 1570 1571 1572

			/* Which index block gets the new entry? */
			if (at - entries >= icount1) {
				frame->at = at = at - entries - icount1 + entries2;
				frame->entries = entries = entries2;
				swap(frame->bh, bh2);
			}
1573 1574 1575
			dx_insert_block(frames + 0, hash2, newblock);
			dxtrace(dx_show_index("node", frames[1].entries));
			dxtrace(dx_show_index("node",
1576
			       ((struct dx_node *) bh2->b_data)->entries));
1577
			err = ext4_journal_dirty_metadata(handle, bh2);
1578 1579 1580 1581
			if (err)
				goto journal_error;
			brelse (bh2);
		} else {
1582 1583
			dxtrace(printk(KERN_DEBUG
				       "Creating second level index...\n"));
1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597
			memcpy((char *) entries2, (char *) entries,
			       icount * sizeof(struct dx_entry));
			dx_set_limit(entries2, dx_node_limit(dir));

			/* Set up root */
			dx_set_count(entries, 1);
			dx_set_block(entries + 0, newblock);
			((struct dx_root *) frames[0].bh->b_data)->info.indirect_levels = 1;

			/* Add new access path frame */
			frame = frames + 1;
			frame->at = at = at - entries + entries2;
			frame->entries = entries = entries2;
			frame->bh = bh2;
1598
			err = ext4_journal_get_write_access(handle,
1599 1600 1601 1602
							     frame->bh);
			if (err)
				goto journal_error;
		}
1603
		ext4_journal_dirty_metadata(handle, frames[0].bh);
1604 1605 1606 1607 1608 1609 1610 1611 1612
	}
	de = do_split(handle, dir, &bh, frame, &hinfo, &err);
	if (!de)
		goto cleanup;
	err = add_dirent_to_buf(handle, dentry, inode, de, bh);
	bh = NULL;
	goto cleanup;

journal_error:
1613
	ext4_std_error(dir->i_sb, err);
1614 1615 1616 1617 1618 1619 1620 1621
cleanup:
	if (bh)
		brelse(bh);
	dx_release(frames);
	return err;
}

/*
1622
 * ext4_delete_entry deletes a directory entry by merging it with the
1623 1624
 * previous entry
 */
1625 1626 1627 1628
static int ext4_delete_entry(handle_t *handle,
			     struct inode *dir,
			     struct ext4_dir_entry_2 *de_del,
			     struct buffer_head *bh)
1629
{
1630
	struct ext4_dir_entry_2 *de, *pde;
1631 1632 1633 1634
	int i;

	i = 0;
	pde = NULL;
1635
	de = (struct ext4_dir_entry_2 *) bh->b_data;
1636
	while (i < bh->b_size) {
1637
		if (!ext4_check_dir_entry("ext4_delete_entry", dir, de, bh, i))
1638 1639 1640
			return -EIO;
		if (de == de_del)  {
			BUFFER_TRACE(bh, "get_write_access");
1641
			ext4_journal_get_write_access(handle, bh);
1642
			if (pde)
1643 1644 1645
				pde->rec_len = ext4_rec_len_to_disk(
					ext4_rec_len_from_disk(pde->rec_len) +
					ext4_rec_len_from_disk(de->rec_len));
1646 1647 1648
			else
				de->inode = 0;
			dir->i_version++;
1649 1650
			BUFFER_TRACE(bh, "call ext4_journal_dirty_metadata");
			ext4_journal_dirty_metadata(handle, bh);
1651 1652
			return 0;
		}
1653
		i += ext4_rec_len_from_disk(de->rec_len);
1654
		pde = de;
1655
		de = ext4_next_entry(de);
1656 1657 1658 1659
	}
	return -ENOENT;
}

1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688
/*
 * DIR_NLINK feature is set if 1) nlinks > EXT4_LINK_MAX or 2) nlinks == 2,
 * since this indicates that nlinks count was previously 1.
 */
static void ext4_inc_count(handle_t *handle, struct inode *inode)
{
	inc_nlink(inode);
	if (is_dx(inode) && inode->i_nlink > 1) {
		/* limit is 16-bit i_links_count */
		if (inode->i_nlink >= EXT4_LINK_MAX || inode->i_nlink == 2) {
			inode->i_nlink = 1;
			EXT4_SET_RO_COMPAT_FEATURE(inode->i_sb,
					      EXT4_FEATURE_RO_COMPAT_DIR_NLINK);
		}
	}
}

/*
 * If a directory had nlink == 1, then we should let it be 1. This indicates
 * directory has >EXT4_LINK_MAX subdirs.
 */
static void ext4_dec_count(handle_t *handle, struct inode *inode)
{
	drop_nlink(inode);
	if (S_ISDIR(inode->i_mode) && inode->i_nlink == 0)
		inc_nlink(inode);
}


1689
static int ext4_add_nondir(handle_t *handle,
1690 1691
		struct dentry *dentry, struct inode *inode)
{
1692
	int err = ext4_add_entry(handle, dentry, inode);
1693
	if (!err) {
1694
		ext4_mark_inode_dirty(handle, inode);
1695 1696 1697
		d_instantiate(dentry, inode);
		return 0;
	}
1698
	drop_nlink(inode);
1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710
	iput(inode);
	return err;
}

/*
 * By the time this is called, we already have created
 * the directory cache entry for the new file, but it
 * is so far negative - it has no inode.
 *
 * If the create succeeds, we fill in the inode information
 * with d_instantiate().
 */
1711 1712
static int ext4_create(struct inode *dir, struct dentry *dentry, int mode,
		       struct nameidata *nd)
1713 1714
{
	handle_t *handle;
1715
	struct inode *inode;
1716 1717 1718
	int err, retries = 0;

retry:
1719 1720 1721
	handle = ext4_journal_start(dir, EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
					EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 +
					2*EXT4_QUOTA_INIT_BLOCKS(dir->i_sb));
1722 1723 1724 1725 1726 1727
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	if (IS_DIRSYNC(dir))
		handle->h_sync = 1;

1728
	inode = ext4_new_inode (handle, dir, mode);
1729 1730
	err = PTR_ERR(inode);
	if (!IS_ERR(inode)) {
1731 1732 1733 1734
		inode->i_op = &ext4_file_inode_operations;
		inode->i_fop = &ext4_file_operations;
		ext4_set_aops(inode);
		err = ext4_add_nondir(handle, dentry, inode);
1735
	}
1736 1737
	ext4_journal_stop(handle);
	if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
1738 1739 1740 1741
		goto retry;
	return err;
}

1742 1743
static int ext4_mknod(struct inode *dir, struct dentry *dentry,
		      int mode, dev_t rdev)
1744 1745 1746 1747 1748 1749 1750 1751 1752
{
	handle_t *handle;
	struct inode *inode;
	int err, retries = 0;

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

retry:
1753 1754 1755
	handle = ext4_journal_start(dir, EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
					EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 +
					2*EXT4_QUOTA_INIT_BLOCKS(dir->i_sb));
1756 1757 1758 1759 1760 1761
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	if (IS_DIRSYNC(dir))
		handle->h_sync = 1;

1762
	inode = ext4_new_inode(handle, dir, mode);
1763 1764 1765
	err = PTR_ERR(inode);
	if (!IS_ERR(inode)) {
		init_special_inode(inode, inode->i_mode, rdev);
T
Theodore Ts'o 已提交
1766
#ifdef CONFIG_EXT4_FS_XATTR
1767
		inode->i_op = &ext4_special_inode_operations;
1768
#endif
1769
		err = ext4_add_nondir(handle, dentry, inode);
1770
	}
1771 1772
	ext4_journal_stop(handle);
	if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
1773 1774 1775 1776
		goto retry;
	return err;
}

1777
static int ext4_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1778 1779
{
	handle_t *handle;
1780 1781 1782
	struct inode *inode;
	struct buffer_head *dir_block;
	struct ext4_dir_entry_2 *de;
1783 1784
	int err, retries = 0;

1785
	if (EXT4_DIR_LINK_MAX(dir))
1786 1787 1788
		return -EMLINK;

retry:
1789 1790 1791
	handle = ext4_journal_start(dir, EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
					EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 +
					2*EXT4_QUOTA_INIT_BLOCKS(dir->i_sb));
1792 1793 1794 1795 1796 1797
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	if (IS_DIRSYNC(dir))
		handle->h_sync = 1;

1798
	inode = ext4_new_inode(handle, dir, S_IFDIR | mode);
1799 1800 1801 1802
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
		goto out_stop;

1803 1804 1805
	inode->i_op = &ext4_dir_inode_operations;
	inode->i_fop = &ext4_dir_operations;
	inode->i_size = EXT4_I(inode)->i_disksize = inode->i_sb->s_blocksize;
1806
	dir_block = ext4_bread(handle, inode, 0, 1, &err);
1807 1808
	if (!dir_block)
		goto out_clear_inode;
1809
	BUFFER_TRACE(dir_block, "get_write_access");
1810 1811
	ext4_journal_get_write_access(handle, dir_block);
	de = (struct ext4_dir_entry_2 *) dir_block->b_data;
1812 1813
	de->inode = cpu_to_le32(inode->i_ino);
	de->name_len = 1;
1814
	de->rec_len = ext4_rec_len_to_disk(EXT4_DIR_REC_LEN(de->name_len));
1815
	strcpy(de->name, ".");
1816
	ext4_set_de_type(dir->i_sb, de, S_IFDIR);
1817
	de = ext4_next_entry(de);
1818
	de->inode = cpu_to_le32(dir->i_ino);
1819 1820
	de->rec_len = ext4_rec_len_to_disk(inode->i_sb->s_blocksize -
						EXT4_DIR_REC_LEN(1));
1821
	de->name_len = 2;
1822
	strcpy(de->name, "..");
1823
	ext4_set_de_type(dir->i_sb, de, S_IFDIR);
1824
	inode->i_nlink = 2;
1825 1826
	BUFFER_TRACE(dir_block, "call ext4_journal_dirty_metadata");
	ext4_journal_dirty_metadata(handle, dir_block);
1827
	brelse(dir_block);
1828
	ext4_mark_inode_dirty(handle, inode);
1829
	err = ext4_add_entry(handle, dentry, inode);
1830
	if (err) {
1831 1832
out_clear_inode:
		clear_nlink(inode);
1833
		ext4_mark_inode_dirty(handle, inode);
1834
		iput(inode);
1835 1836
		goto out_stop;
	}
1837
	ext4_inc_count(handle, dir);
1838 1839
	ext4_update_dx_flag(dir);
	ext4_mark_inode_dirty(handle, dir);
1840 1841
	d_instantiate(dentry, inode);
out_stop:
1842 1843
	ext4_journal_stop(handle);
	if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
1844 1845 1846 1847 1848 1849 1850
		goto retry;
	return err;
}

/*
 * routine to check that the specified directory is empty (for rmdir)
 */
1851
static int empty_dir(struct inode *inode)
1852 1853
{
	unsigned long offset;
1854 1855 1856
	struct buffer_head *bh;
	struct ext4_dir_entry_2 *de, *de1;
	struct super_block *sb;
1857 1858 1859
	int err = 0;

	sb = inode->i_sb;
1860
	if (inode->i_size < EXT4_DIR_REC_LEN(1) + EXT4_DIR_REC_LEN(2) ||
1861
	    !(bh = ext4_bread(NULL, inode, 0, 0, &err))) {
1862
		if (err)
1863
			ext4_error(inode->i_sb, __func__,
1864 1865 1866
				   "error %d reading directory #%lu offset 0",
				   err, inode->i_ino);
		else
1867
			ext4_warning(inode->i_sb, __func__,
1868 1869 1870 1871
				     "bad directory (dir #%lu) - no data block",
				     inode->i_ino);
		return 1;
	}
1872
	de = (struct ext4_dir_entry_2 *) bh->b_data;
1873
	de1 = ext4_next_entry(de);
1874 1875
	if (le32_to_cpu(de->inode) != inode->i_ino ||
			!le32_to_cpu(de1->inode) ||
1876 1877 1878 1879 1880 1881
			strcmp(".", de->name) ||
			strcmp("..", de1->name)) {
		ext4_warning(inode->i_sb, "empty_dir",
			     "bad directory (dir #%lu) - no `.' or `..'",
			     inode->i_ino);
		brelse(bh);
1882 1883
		return 1;
	}
1884 1885 1886
	offset = ext4_rec_len_from_disk(de->rec_len) +
		 ext4_rec_len_from_disk(de1->rec_len);
	de = ext4_next_entry(de1);
1887
	while (offset < inode->i_size) {
1888 1889 1890
		if (!bh ||
			(void *) de >= (void *) (bh->b_data+sb->s_blocksize)) {
			err = 0;
1891 1892
			brelse(bh);
			bh = ext4_bread(NULL, inode,
1893
				offset >> EXT4_BLOCK_SIZE_BITS(sb), 0, &err);
1894 1895
			if (!bh) {
				if (err)
1896
					ext4_error(sb, __func__,
1897 1898 1899 1900 1901 1902
						   "error %d reading directory"
						   " #%lu offset %lu",
						   err, inode->i_ino, offset);
				offset += sb->s_blocksize;
				continue;
			}
1903
			de = (struct ext4_dir_entry_2 *) bh->b_data;
1904
		}
1905 1906
		if (!ext4_check_dir_entry("empty_dir", inode, de, bh, offset)) {
			de = (struct ext4_dir_entry_2 *)(bh->b_data +
1907 1908 1909 1910 1911
							 sb->s_blocksize);
			offset = (offset | (sb->s_blocksize - 1)) + 1;
			continue;
		}
		if (le32_to_cpu(de->inode)) {
1912
			brelse(bh);
1913 1914
			return 0;
		}
1915 1916
		offset += ext4_rec_len_from_disk(de->rec_len);
		de = ext4_next_entry(de);
1917
	}
1918
	brelse(bh);
1919 1920 1921
	return 1;
}

1922
/* ext4_orphan_add() links an unlinked or truncated inode into a list of
1923 1924 1925 1926 1927
 * such inodes, starting at the superblock, in case we crash before the
 * file is closed/deleted, or in case the inode truncate spans multiple
 * transactions and the last transaction is not recovered after a crash.
 *
 * At filesystem recovery time, we walk this list deleting unlinked
1928
 * inodes and truncating linked inodes in ext4_orphan_cleanup().
1929
 */
1930
int ext4_orphan_add(handle_t *handle, struct inode *inode)
1931 1932
{
	struct super_block *sb = inode->i_sb;
1933
	struct ext4_iloc iloc;
1934 1935 1936
	int err = 0, rc;

	lock_super(sb);
1937
	if (!list_empty(&EXT4_I(inode)->i_orphan))
1938 1939 1940 1941 1942 1943
		goto out_unlock;

	/* Orphan handling is only valid for files with data blocks
	 * being truncated, or files being unlinked. */

	/* @@@ FIXME: Observation from aviro:
1944 1945
	 * I think I can trigger J_ASSERT in ext4_orphan_add().  We block
	 * here (on lock_super()), so race with ext4_link() which might bump
1946 1947 1948
	 * ->i_nlink. For, say it, character device. Not a regular file,
	 * not a directory, not a symlink and ->i_nlink > 0.
	 */
1949 1950
	J_ASSERT((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
		  S_ISLNK(inode->i_mode)) || inode->i_nlink == 0);
1951

1952 1953
	BUFFER_TRACE(EXT4_SB(sb)->s_sbh, "get_write_access");
	err = ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh);
1954 1955 1956
	if (err)
		goto out_unlock;

1957
	err = ext4_reserve_inode_write(handle, inode, &iloc);
1958 1959 1960 1961
	if (err)
		goto out_unlock;

	/* Insert this inode at the head of the on-disk orphan list... */
1962 1963 1964 1965
	NEXT_ORPHAN(inode) = le32_to_cpu(EXT4_SB(sb)->s_es->s_last_orphan);
	EXT4_SB(sb)->s_es->s_last_orphan = cpu_to_le32(inode->i_ino);
	err = ext4_journal_dirty_metadata(handle, EXT4_SB(sb)->s_sbh);
	rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977
	if (!err)
		err = rc;

	/* Only add to the head of the in-memory list if all the
	 * previous operations succeeded.  If the orphan_add is going to
	 * fail (possibly taking the journal offline), we can't risk
	 * leaving the inode on the orphan list: stray orphan-list
	 * entries can cause panics at unmount time.
	 *
	 * This is safe: on error we're going to ignore the orphan list
	 * anyway on the next recovery. */
	if (!err)
1978
		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1979 1980 1981 1982 1983 1984

	jbd_debug(4, "superblock will point to %lu\n", inode->i_ino);
	jbd_debug(4, "orphan inode %lu will point to %d\n",
			inode->i_ino, NEXT_ORPHAN(inode));
out_unlock:
	unlock_super(sb);
1985
	ext4_std_error(inode->i_sb, err);
1986 1987 1988 1989
	return err;
}

/*
1990
 * ext4_orphan_del() removes an unlinked or truncated inode from the list
1991 1992
 * of such inodes stored on disk, because it is finally being cleaned up.
 */
1993
int ext4_orphan_del(handle_t *handle, struct inode *inode)
1994 1995
{
	struct list_head *prev;
1996 1997
	struct ext4_inode_info *ei = EXT4_I(inode);
	struct ext4_sb_info *sbi;
1998
	unsigned long ino_next;
1999
	struct ext4_iloc iloc;
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
	int err = 0;

	lock_super(inode->i_sb);
	if (list_empty(&ei->i_orphan)) {
		unlock_super(inode->i_sb);
		return 0;
	}

	ino_next = NEXT_ORPHAN(inode);
	prev = ei->i_orphan.prev;
2010
	sbi = EXT4_SB(inode->i_sb);
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022

	jbd_debug(4, "remove inode %lu from orphan list\n", inode->i_ino);

	list_del_init(&ei->i_orphan);

	/* If we're on an error path, we may not have a valid
	 * transaction handle with which to update the orphan list on
	 * disk, but we still need to remove the inode from the linked
	 * list in memory. */
	if (!handle)
		goto out;

2023
	err = ext4_reserve_inode_write(handle, inode, &iloc);
2024 2025 2026 2027 2028 2029
	if (err)
		goto out_err;

	if (prev == &sbi->s_orphan) {
		jbd_debug(4, "superblock will point to %lu\n", ino_next);
		BUFFER_TRACE(sbi->s_sbh, "get_write_access");
2030
		err = ext4_journal_get_write_access(handle, sbi->s_sbh);
2031 2032 2033
		if (err)
			goto out_brelse;
		sbi->s_es->s_last_orphan = cpu_to_le32(ino_next);
2034
		err = ext4_journal_dirty_metadata(handle, sbi->s_sbh);
2035
	} else {
2036
		struct ext4_iloc iloc2;
2037
		struct inode *i_prev =
2038
			&list_entry(prev, struct ext4_inode_info, i_orphan)->vfs_inode;
2039 2040 2041

		jbd_debug(4, "orphan inode %lu will point to %lu\n",
			  i_prev->i_ino, ino_next);
2042
		err = ext4_reserve_inode_write(handle, i_prev, &iloc2);
2043 2044 2045
		if (err)
			goto out_brelse;
		NEXT_ORPHAN(i_prev) = ino_next;
2046
		err = ext4_mark_iloc_dirty(handle, i_prev, &iloc2);
2047 2048 2049 2050
	}
	if (err)
		goto out_brelse;
	NEXT_ORPHAN(inode) = 0;
2051
	err = ext4_mark_iloc_dirty(handle, inode, &iloc);
2052 2053

out_err:
2054
	ext4_std_error(inode->i_sb, err);
2055 2056 2057 2058 2059 2060 2061 2062 2063
out:
	unlock_super(inode->i_sb);
	return err;

out_brelse:
	brelse(iloc.bh);
	goto out_err;
}

2064
static int ext4_rmdir(struct inode *dir, struct dentry *dentry)
2065 2066
{
	int retval;
2067 2068 2069
	struct inode *inode;
	struct buffer_head *bh;
	struct ext4_dir_entry_2 *de;
2070 2071 2072 2073 2074
	handle_t *handle;

	/* Initialize quotas before so that eventual writes go in
	 * separate transaction */
	DQUOT_INIT(dentry->d_inode);
2075
	handle = ext4_journal_start(dir, EXT4_DELETE_TRANS_BLOCKS(dir->i_sb));
2076 2077 2078 2079
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	retval = -ENOENT;
2080
	bh = ext4_find_entry(dir, &dentry->d_name, &de);
2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093
	if (!bh)
		goto end_rmdir;

	if (IS_DIRSYNC(dir))
		handle->h_sync = 1;

	inode = dentry->d_inode;

	retval = -EIO;
	if (le32_to_cpu(de->inode) != inode->i_ino)
		goto end_rmdir;

	retval = -ENOTEMPTY;
2094
	if (!empty_dir(inode))
2095 2096
		goto end_rmdir;

2097
	retval = ext4_delete_entry(handle, dir, de, bh);
2098 2099
	if (retval)
		goto end_rmdir;
2100
	if (!EXT4_DIR_LINK_EMPTY(inode))
2101 2102 2103
		ext4_warning(inode->i_sb, "ext4_rmdir",
			     "empty directory has too many links (%d)",
			     inode->i_nlink);
2104 2105 2106 2107 2108 2109
	inode->i_version++;
	clear_nlink(inode);
	/* There's no need to set i_disksize: the fact that i_nlink is
	 * zero will ensure that the right thing happens during any
	 * recovery. */
	inode->i_size = 0;
2110
	ext4_orphan_add(handle, inode);
K
Kalpak Shah 已提交
2111
	inode->i_ctime = dir->i_ctime = dir->i_mtime = ext4_current_time(inode);
2112
	ext4_mark_inode_dirty(handle, inode);
2113
	ext4_dec_count(handle, dir);
2114 2115
	ext4_update_dx_flag(dir);
	ext4_mark_inode_dirty(handle, dir);
2116 2117

end_rmdir:
2118
	ext4_journal_stop(handle);
2119
	brelse(bh);
2120 2121 2122
	return retval;
}

2123
static int ext4_unlink(struct inode *dir, struct dentry *dentry)
2124 2125
{
	int retval;
2126 2127 2128
	struct inode *inode;
	struct buffer_head *bh;
	struct ext4_dir_entry_2 *de;
2129 2130 2131 2132 2133
	handle_t *handle;

	/* Initialize quotas before so that eventual writes go
	 * in separate transaction */
	DQUOT_INIT(dentry->d_inode);
2134
	handle = ext4_journal_start(dir, EXT4_DELETE_TRANS_BLOCKS(dir->i_sb));
2135 2136 2137 2138 2139 2140 2141
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	if (IS_DIRSYNC(dir))
		handle->h_sync = 1;

	retval = -ENOENT;
2142
	bh = ext4_find_entry(dir, &dentry->d_name, &de);
2143 2144 2145 2146 2147 2148 2149 2150 2151 2152
	if (!bh)
		goto end_unlink;

	inode = dentry->d_inode;

	retval = -EIO;
	if (le32_to_cpu(de->inode) != inode->i_ino)
		goto end_unlink;

	if (!inode->i_nlink) {
2153 2154 2155
		ext4_warning(inode->i_sb, "ext4_unlink",
			     "Deleting nonexistent file (%lu), %d",
			     inode->i_ino, inode->i_nlink);
2156 2157
		inode->i_nlink = 1;
	}
2158
	retval = ext4_delete_entry(handle, dir, de, bh);
2159 2160
	if (retval)
		goto end_unlink;
K
Kalpak Shah 已提交
2161
	dir->i_ctime = dir->i_mtime = ext4_current_time(dir);
2162 2163
	ext4_update_dx_flag(dir);
	ext4_mark_inode_dirty(handle, dir);
2164
	drop_nlink(inode);
2165
	if (!inode->i_nlink)
2166
		ext4_orphan_add(handle, inode);
K
Kalpak Shah 已提交
2167
	inode->i_ctime = ext4_current_time(inode);
2168
	ext4_mark_inode_dirty(handle, inode);
2169 2170 2171
	retval = 0;

end_unlink:
2172
	ext4_journal_stop(handle);
2173
	brelse(bh);
2174 2175 2176
	return retval;
}

2177 2178
static int ext4_symlink(struct inode *dir,
			struct dentry *dentry, const char *symname)
2179 2180
{
	handle_t *handle;
2181
	struct inode *inode;
2182 2183 2184 2185 2186 2187 2188
	int l, err, retries = 0;

	l = strlen(symname)+1;
	if (l > dir->i_sb->s_blocksize)
		return -ENAMETOOLONG;

retry:
2189 2190 2191
	handle = ext4_journal_start(dir, EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
					EXT4_INDEX_EXTRA_TRANS_BLOCKS + 5 +
					2*EXT4_QUOTA_INIT_BLOCKS(dir->i_sb));
2192 2193 2194 2195 2196 2197
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	if (IS_DIRSYNC(dir))
		handle->h_sync = 1;

2198
	inode = ext4_new_inode(handle, dir, S_IFLNK|S_IRWXUGO);
2199 2200 2201 2202
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
		goto out_stop;

2203
	if (l > sizeof(EXT4_I(inode)->i_data)) {
2204 2205
		inode->i_op = &ext4_symlink_inode_operations;
		ext4_set_aops(inode);
2206
		/*
2207
		 * page_symlink() calls into ext4_prepare/commit_write.
2208 2209 2210 2211 2212 2213
		 * We have a transaction open.  All is sweetness.  It also sets
		 * i_size in generic_commit_write().
		 */
		err = __page_symlink(inode, symname, l,
				mapping_gfp_mask(inode->i_mapping) & ~__GFP_FS);
		if (err) {
2214
			clear_nlink(inode);
2215
			ext4_mark_inode_dirty(handle, inode);
2216
			iput(inode);
2217 2218 2219
			goto out_stop;
		}
	} else {
2220 2221
		/* clear the extent format for fast symlink */
		EXT4_I(inode)->i_flags &= ~EXT4_EXTENTS_FL;
2222
		inode->i_op = &ext4_fast_symlink_inode_operations;
2223
		memcpy((char *)&EXT4_I(inode)->i_data, symname, l);
2224 2225
		inode->i_size = l-1;
	}
2226 2227
	EXT4_I(inode)->i_disksize = inode->i_size;
	err = ext4_add_nondir(handle, dentry, inode);
2228
out_stop:
2229 2230
	ext4_journal_stop(handle);
	if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
2231 2232 2233 2234
		goto retry;
	return err;
}

2235 2236
static int ext4_link(struct dentry *old_dentry,
		     struct inode *dir, struct dentry *dentry)
2237 2238 2239 2240 2241
{
	handle_t *handle;
	struct inode *inode = old_dentry->d_inode;
	int err, retries = 0;

2242
	if (EXT4_DIR_LINK_MAX(inode))
2243
		return -EMLINK;
2244

2245 2246 2247 2248 2249 2250
	/*
	 * Return -ENOENT if we've raced with unlink and i_nlink is 0.  Doing
	 * otherwise has the potential to corrupt the orphan inode list.
	 */
	if (inode->i_nlink == 0)
		return -ENOENT;
2251 2252

retry:
2253 2254
	handle = ext4_journal_start(dir, EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
					EXT4_INDEX_EXTRA_TRANS_BLOCKS);
2255 2256 2257 2258 2259 2260
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	if (IS_DIRSYNC(dir))
		handle->h_sync = 1;

K
Kalpak Shah 已提交
2261
	inode->i_ctime = ext4_current_time(inode);
2262
	ext4_inc_count(handle, inode);
2263 2264
	atomic_inc(&inode->i_count);

2265 2266 2267
	err = ext4_add_nondir(handle, dentry, inode);
	ext4_journal_stop(handle);
	if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
2268 2269 2270 2271 2272
		goto retry;
	return err;
}

#define PARENT_INO(buffer) \
2273
	(ext4_next_entry((struct ext4_dir_entry_2 *)(buffer))->inode)
2274 2275 2276 2277 2278

/*
 * Anybody can rename anything with this: the permission checks are left to the
 * higher-level routines.
 */
2279 2280
static int ext4_rename(struct inode *old_dir, struct dentry *old_dentry,
		       struct inode *new_dir, struct dentry *new_dentry)
2281 2282
{
	handle_t *handle;
2283 2284 2285
	struct inode *old_inode, *new_inode;
	struct buffer_head *old_bh, *new_bh, *dir_bh;
	struct ext4_dir_entry_2 *old_de, *new_de;
2286 2287 2288 2289 2290 2291 2292 2293
	int retval;

	old_bh = new_bh = dir_bh = NULL;

	/* Initialize quotas before so that eventual writes go
	 * in separate transaction */
	if (new_dentry->d_inode)
		DQUOT_INIT(new_dentry->d_inode);
2294 2295 2296
	handle = ext4_journal_start(old_dir, 2 *
					EXT4_DATA_TRANS_BLOCKS(old_dir->i_sb) +
					EXT4_INDEX_EXTRA_TRANS_BLOCKS + 2);
2297 2298 2299 2300 2301 2302
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	if (IS_DIRSYNC(old_dir) || IS_DIRSYNC(new_dir))
		handle->h_sync = 1;

2303
	old_bh = ext4_find_entry(old_dir, &old_dentry->d_name, &old_de);
2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315
	/*
	 *  Check for inode number is _not_ due to possible IO errors.
	 *  We might rmdir the source, keep it as pwd of some process
	 *  and merrily kill the link to whatever was created under the
	 *  same name. Goodbye sticky bit ;-<
	 */
	old_inode = old_dentry->d_inode;
	retval = -ENOENT;
	if (!old_bh || le32_to_cpu(old_de->inode) != old_inode->i_ino)
		goto end_rename;

	new_inode = new_dentry->d_inode;
2316
	new_bh = ext4_find_entry(new_dir, &new_dentry->d_name, &new_de);
2317 2318
	if (new_bh) {
		if (!new_inode) {
2319
			brelse(new_bh);
2320 2321 2322 2323 2324 2325
			new_bh = NULL;
		}
	}
	if (S_ISDIR(old_inode->i_mode)) {
		if (new_inode) {
			retval = -ENOTEMPTY;
2326
			if (!empty_dir(new_inode))
2327 2328 2329
				goto end_rename;
		}
		retval = -EIO;
2330
		dir_bh = ext4_bread(handle, old_inode, 0, 0, &retval);
2331 2332 2333 2334 2335
		if (!dir_bh)
			goto end_rename;
		if (le32_to_cpu(PARENT_INO(dir_bh->b_data)) != old_dir->i_ino)
			goto end_rename;
		retval = -EMLINK;
2336
		if (!new_inode && new_dir != old_dir &&
2337
				new_dir->i_nlink >= EXT4_LINK_MAX)
2338 2339 2340
			goto end_rename;
	}
	if (!new_bh) {
2341
		retval = ext4_add_entry(handle, new_dentry, old_inode);
2342 2343 2344 2345
		if (retval)
			goto end_rename;
	} else {
		BUFFER_TRACE(new_bh, "get write access");
2346
		ext4_journal_get_write_access(handle, new_bh);
2347
		new_de->inode = cpu_to_le32(old_inode->i_ino);
2348 2349
		if (EXT4_HAS_INCOMPAT_FEATURE(new_dir->i_sb,
					      EXT4_FEATURE_INCOMPAT_FILETYPE))
2350 2351
			new_de->file_type = old_de->file_type;
		new_dir->i_version++;
2352 2353 2354
		new_dir->i_ctime = new_dir->i_mtime =
					ext4_current_time(new_dir);
		ext4_mark_inode_dirty(handle, new_dir);
2355 2356
		BUFFER_TRACE(new_bh, "call ext4_journal_dirty_metadata");
		ext4_journal_dirty_metadata(handle, new_bh);
2357 2358 2359 2360 2361 2362 2363 2364
		brelse(new_bh);
		new_bh = NULL;
	}

	/*
	 * Like most other Unix systems, set the ctime for inodes on a
	 * rename.
	 */
K
Kalpak Shah 已提交
2365
	old_inode->i_ctime = ext4_current_time(old_inode);
2366
	ext4_mark_inode_dirty(handle, old_inode);
2367 2368 2369 2370 2371 2372 2373

	/*
	 * ok, that's it
	 */
	if (le32_to_cpu(old_de->inode) != old_inode->i_ino ||
	    old_de->name_len != old_dentry->d_name.len ||
	    strncmp(old_de->name, old_dentry->d_name.name, old_de->name_len) ||
2374
	    (retval = ext4_delete_entry(handle, old_dir,
2375 2376 2377 2378 2379 2380
					old_de, old_bh)) == -ENOENT) {
		/* old_de could have moved from under us during htree split, so
		 * make sure that we are deleting the right entry.  We might
		 * also be pointing to a stale entry in the unused part of
		 * old_bh so just checking inum and the name isn't enough. */
		struct buffer_head *old_bh2;
2381
		struct ext4_dir_entry_2 *old_de2;
2382

2383
		old_bh2 = ext4_find_entry(old_dir, &old_dentry->d_name, &old_de2);
2384
		if (old_bh2) {
2385
			retval = ext4_delete_entry(handle, old_dir,
2386 2387 2388 2389 2390
						   old_de2, old_bh2);
			brelse(old_bh2);
		}
	}
	if (retval) {
2391
		ext4_warning(old_dir->i_sb, "ext4_rename",
2392 2393 2394 2395 2396
				"Deleting old file (%lu), %d, error=%d",
				old_dir->i_ino, old_dir->i_nlink, retval);
	}

	if (new_inode) {
2397
		ext4_dec_count(handle, new_inode);
K
Kalpak Shah 已提交
2398
		new_inode->i_ctime = ext4_current_time(new_inode);
2399
	}
K
Kalpak Shah 已提交
2400
	old_dir->i_ctime = old_dir->i_mtime = ext4_current_time(old_dir);
2401
	ext4_update_dx_flag(old_dir);
2402 2403
	if (dir_bh) {
		BUFFER_TRACE(dir_bh, "get_write_access");
2404
		ext4_journal_get_write_access(handle, dir_bh);
2405
		PARENT_INO(dir_bh->b_data) = cpu_to_le32(new_dir->i_ino);
2406 2407
		BUFFER_TRACE(dir_bh, "call ext4_journal_dirty_metadata");
		ext4_journal_dirty_metadata(handle, dir_bh);
2408
		ext4_dec_count(handle, old_dir);
2409
		if (new_inode) {
2410
			/* checked empty_dir above, can't have another parent,
2411
			 * ext4_dec_count() won't work for many-linked dirs */
2412
			new_inode->i_nlink = 0;
2413
		} else {
2414
			ext4_inc_count(handle, new_dir);
2415 2416
			ext4_update_dx_flag(new_dir);
			ext4_mark_inode_dirty(handle, new_dir);
2417 2418
		}
	}
2419
	ext4_mark_inode_dirty(handle, old_dir);
2420
	if (new_inode) {
2421
		ext4_mark_inode_dirty(handle, new_inode);
2422
		if (!new_inode->i_nlink)
2423
			ext4_orphan_add(handle, new_inode);
2424 2425 2426 2427
	}
	retval = 0;

end_rename:
2428 2429 2430
	brelse(dir_bh);
	brelse(old_bh);
	brelse(new_bh);
2431
	ext4_journal_stop(handle);
2432 2433 2434 2435 2436 2437
	return retval;
}

/*
 * directories can handle most operations...
 */
2438
const struct inode_operations ext4_dir_inode_operations = {
2439 2440 2441 2442 2443 2444 2445 2446 2447 2448
	.create		= ext4_create,
	.lookup		= ext4_lookup,
	.link		= ext4_link,
	.unlink		= ext4_unlink,
	.symlink	= ext4_symlink,
	.mkdir		= ext4_mkdir,
	.rmdir		= ext4_rmdir,
	.mknod		= ext4_mknod,
	.rename		= ext4_rename,
	.setattr	= ext4_setattr,
T
Theodore Ts'o 已提交
2449
#ifdef CONFIG_EXT4_FS_XATTR
2450 2451
	.setxattr	= generic_setxattr,
	.getxattr	= generic_getxattr,
2452
	.listxattr	= ext4_listxattr,
2453 2454
	.removexattr	= generic_removexattr,
#endif
2455
	.permission	= ext4_permission,
2456 2457
};

2458
const struct inode_operations ext4_special_inode_operations = {
2459
	.setattr	= ext4_setattr,
T
Theodore Ts'o 已提交
2460
#ifdef CONFIG_EXT4_FS_XATTR
2461 2462
	.setxattr	= generic_setxattr,
	.getxattr	= generic_getxattr,
2463
	.listxattr	= ext4_listxattr,
2464 2465
	.removexattr	= generic_removexattr,
#endif
2466
	.permission	= ext4_permission,
2467
};