namei.c 84.2 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

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

43
#include <trace/events/ext4.h>
44 45 46 47 48
/*
 * 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
{
	struct buffer_head *bh;

58 59 60 61 62 63 64
	if (unlikely(EXT4_SB(inode->i_sb)->s_max_dir_size_kb &&
		     ((inode->i_size >> 10) >=
		      EXT4_SB(inode->i_sb)->s_max_dir_size_kb))) {
		*err = -ENOSPC;
		return NULL;
	}

65 66
	*block = inode->i_size >> inode->i_sb->s_blocksize_bits;

67 68
	bh = ext4_bread(handle, inode, *block, 1, err);
	if (bh) {
69
		inode->i_size += inode->i_sb->s_blocksize;
70
		EXT4_I(inode)->i_disksize = inode->i_size;
71 72 73 74 75
		*err = ext4_journal_get_write_access(handle, bh);
		if (*err) {
			brelse(bh);
			bh = NULL;
		}
76
	}
C
Carlos Maiolino 已提交
77 78 79 80 81 82
	if (!bh && !(*err)) {
		*err = -EIO;
		ext4_error(inode->i_sb,
			   "Directory hole detected on inode %lu\n",
			   inode->i_ino);
	}
83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156
	return bh;
}

#ifndef assert
#define assert(test) J_ASSERT(test)
#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;
157 158
	u16 offs;
	u16 size;
159 160
};

161 162 163 164 165 166 167 168
/*
 * This goes at the end of each htree block.
 */
struct dx_tail {
	u32 dt_reserved;
	__le32 dt_checksum;	/* crc32c(uuid+inum+dirblock) */
};

A
Aneesh Kumar K.V 已提交
169 170
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);
171 172 173 174 175 176 177 178
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);
179
static struct dx_frame *dx_probe(const struct qstr *d_name,
180 181 182 183
				 struct inode *dir,
				 struct dx_hash_info *hinfo,
				 struct dx_frame *frame,
				 int *err);
184
static void dx_release(struct dx_frame *frames);
185
static int dx_make_map(struct ext4_dir_entry_2 *de, unsigned blocksize,
186
		       struct dx_hash_info *hinfo, struct dx_map_entry map[]);
187
static void dx_sort_map(struct dx_map_entry *map, unsigned count);
188
static struct ext4_dir_entry_2 *dx_move_dirents(char *from, char *to,
189
		struct dx_map_entry *offsets, int count, unsigned blocksize);
190
static struct ext4_dir_entry_2* dx_pack_dirents(char *base, unsigned blocksize);
A
Aneesh Kumar K.V 已提交
191 192
static void dx_insert_block(struct dx_frame *frame,
					u32 hash, ext4_lblk_t block);
193
static int ext4_htree_next_block(struct inode *dir, __u32 hash,
194 195 196
				 struct dx_frame *frame,
				 struct dx_frame *frames,
				 __u32 *start_hash);
197 198 199 200
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);
201
static int ext4_dx_add_entry(handle_t *handle, struct dentry *dentry,
202 203
			     struct inode *inode);

204
/* checksumming functions */
205 206
void initialize_dirent_tail(struct ext4_dir_entry_tail *t,
			    unsigned int blocksize)
207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258
{
	memset(t, 0, sizeof(struct ext4_dir_entry_tail));
	t->det_rec_len = ext4_rec_len_to_disk(
			sizeof(struct ext4_dir_entry_tail), blocksize);
	t->det_reserved_ft = EXT4_FT_DIR_CSUM;
}

/* Walk through a dirent block to find a checksum "dirent" at the tail */
static struct ext4_dir_entry_tail *get_dirent_tail(struct inode *inode,
						   struct ext4_dir_entry *de)
{
	struct ext4_dir_entry_tail *t;

#ifdef PARANOID
	struct ext4_dir_entry *d, *top;

	d = de;
	top = (struct ext4_dir_entry *)(((void *)de) +
		(EXT4_BLOCK_SIZE(inode->i_sb) -
		sizeof(struct ext4_dir_entry_tail)));
	while (d < top && d->rec_len)
		d = (struct ext4_dir_entry *)(((void *)d) +
		    le16_to_cpu(d->rec_len));

	if (d != top)
		return NULL;

	t = (struct ext4_dir_entry_tail *)d;
#else
	t = EXT4_DIRENT_TAIL(de, EXT4_BLOCK_SIZE(inode->i_sb));
#endif

	if (t->det_reserved_zero1 ||
	    le16_to_cpu(t->det_rec_len) != sizeof(struct ext4_dir_entry_tail) ||
	    t->det_reserved_zero2 ||
	    t->det_reserved_ft != EXT4_FT_DIR_CSUM)
		return NULL;

	return t;
}

static __le32 ext4_dirent_csum(struct inode *inode,
			       struct ext4_dir_entry *dirent, int size)
{
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
	struct ext4_inode_info *ei = EXT4_I(inode);
	__u32 csum;

	csum = ext4_chksum(sbi, ei->i_csum_seed, (__u8 *)dirent, size);
	return cpu_to_le32(csum);
}

259 260 261 262 263 264
static void warn_no_space_for_csum(struct inode *inode)
{
	ext4_warning(inode->i_sb, "no space in directory inode %lu leaf for "
		     "checksum.  Please run e2fsck -D.", inode->i_ino);
}

265 266 267 268 269 270 271 272 273 274
int ext4_dirent_csum_verify(struct inode *inode, struct ext4_dir_entry *dirent)
{
	struct ext4_dir_entry_tail *t;

	if (!EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
					EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
		return 1;

	t = get_dirent_tail(inode, dirent);
	if (!t) {
275
		warn_no_space_for_csum(inode);
276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296
		return 0;
	}

	if (t->det_checksum != ext4_dirent_csum(inode, dirent,
						(void *)t - (void *)dirent))
		return 0;

	return 1;
}

static void ext4_dirent_csum_set(struct inode *inode,
				 struct ext4_dir_entry *dirent)
{
	struct ext4_dir_entry_tail *t;

	if (!EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
					EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
		return;

	t = get_dirent_tail(inode, dirent);
	if (!t) {
297
		warn_no_space_for_csum(inode);
298 299 300 301 302 303 304
		return;
	}

	t->det_checksum = ext4_dirent_csum(inode, dirent,
					   (void *)t - (void *)dirent);
}

305 306 307
int ext4_handle_dirty_dirent_node(handle_t *handle,
				  struct inode *inode,
				  struct buffer_head *bh)
308 309 310 311 312
{
	ext4_dirent_csum_set(inode, (struct ext4_dir_entry *)bh->b_data);
	return ext4_handle_dirty_metadata(handle, inode, bh);
}

313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378
static struct dx_countlimit *get_dx_countlimit(struct inode *inode,
					       struct ext4_dir_entry *dirent,
					       int *offset)
{
	struct ext4_dir_entry *dp;
	struct dx_root_info *root;
	int count_offset;

	if (le16_to_cpu(dirent->rec_len) == EXT4_BLOCK_SIZE(inode->i_sb))
		count_offset = 8;
	else if (le16_to_cpu(dirent->rec_len) == 12) {
		dp = (struct ext4_dir_entry *)(((void *)dirent) + 12);
		if (le16_to_cpu(dp->rec_len) !=
		    EXT4_BLOCK_SIZE(inode->i_sb) - 12)
			return NULL;
		root = (struct dx_root_info *)(((void *)dp + 12));
		if (root->reserved_zero ||
		    root->info_length != sizeof(struct dx_root_info))
			return NULL;
		count_offset = 32;
	} else
		return NULL;

	if (offset)
		*offset = count_offset;
	return (struct dx_countlimit *)(((void *)dirent) + count_offset);
}

static __le32 ext4_dx_csum(struct inode *inode, struct ext4_dir_entry *dirent,
			   int count_offset, int count, struct dx_tail *t)
{
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
	struct ext4_inode_info *ei = EXT4_I(inode);
	__u32 csum, old_csum;
	int size;

	size = count_offset + (count * sizeof(struct dx_entry));
	old_csum = t->dt_checksum;
	t->dt_checksum = 0;
	csum = ext4_chksum(sbi, ei->i_csum_seed, (__u8 *)dirent, size);
	csum = ext4_chksum(sbi, csum, (__u8 *)t, sizeof(struct dx_tail));
	t->dt_checksum = old_csum;

	return cpu_to_le32(csum);
}

static int ext4_dx_csum_verify(struct inode *inode,
			       struct ext4_dir_entry *dirent)
{
	struct dx_countlimit *c;
	struct dx_tail *t;
	int count_offset, limit, count;

	if (!EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
					EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
		return 1;

	c = get_dx_countlimit(inode, dirent, &count_offset);
	if (!c) {
		EXT4_ERROR_INODE(inode, "dir seems corrupt?  Run e2fsck -D.");
		return 1;
	}
	limit = le16_to_cpu(c->limit);
	count = le16_to_cpu(c->count);
	if (count_offset + (limit * sizeof(struct dx_entry)) >
	    EXT4_BLOCK_SIZE(inode->i_sb) - sizeof(struct dx_tail)) {
379
		warn_no_space_for_csum(inode);
380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408
		return 1;
	}
	t = (struct dx_tail *)(((struct dx_entry *)c) + limit);

	if (t->dt_checksum != ext4_dx_csum(inode, dirent, count_offset,
					    count, t))
		return 0;
	return 1;
}

static void ext4_dx_csum_set(struct inode *inode, struct ext4_dir_entry *dirent)
{
	struct dx_countlimit *c;
	struct dx_tail *t;
	int count_offset, limit, count;

	if (!EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
					EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
		return;

	c = get_dx_countlimit(inode, dirent, &count_offset);
	if (!c) {
		EXT4_ERROR_INODE(inode, "dir seems corrupt?  Run e2fsck -D.");
		return;
	}
	limit = le16_to_cpu(c->limit);
	count = le16_to_cpu(c->count);
	if (count_offset + (limit * sizeof(struct dx_entry)) >
	    EXT4_BLOCK_SIZE(inode->i_sb) - sizeof(struct dx_tail)) {
409
		warn_no_space_for_csum(inode);
410 411 412 413 414 415 416 417 418 419 420 421 422 423 424
		return;
	}
	t = (struct dx_tail *)(((struct dx_entry *)c) + limit);

	t->dt_checksum = ext4_dx_csum(inode, dirent, count_offset, count, t);
}

static inline int ext4_handle_dirty_dx_node(handle_t *handle,
					    struct inode *inode,
					    struct buffer_head *bh)
{
	ext4_dx_csum_set(inode, (struct ext4_dir_entry *)bh->b_data);
	return ext4_handle_dirty_metadata(handle, inode, bh);
}

425 426 427 428
/*
 * p is at least 6 bytes before the end of page
 */
static inline struct ext4_dir_entry_2 *
429
ext4_next_entry(struct ext4_dir_entry_2 *p, unsigned long blocksize)
430 431
{
	return (struct ext4_dir_entry_2 *)((char *)p +
432
		ext4_rec_len_from_disk(p->rec_len, blocksize));
433 434
}

435 436 437 438 439
/*
 * 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
440
static inline ext4_lblk_t dx_get_block(struct dx_entry *entry)
441 442 443 444
{
	return le32_to_cpu(entry->block) & 0x00ffffff;
}

A
Aneesh Kumar K.V 已提交
445
static inline void dx_set_block(struct dx_entry *entry, ext4_lblk_t value)
446 447 448 449
{
	entry->block = cpu_to_le32(value);
}

450
static inline unsigned dx_get_hash(struct dx_entry *entry)
451 452 453 454
{
	return le32_to_cpu(entry->hash);
}

455
static inline void dx_set_hash(struct dx_entry *entry, unsigned value)
456 457 458 459
{
	entry->hash = cpu_to_le32(value);
}

460
static inline unsigned dx_get_count(struct dx_entry *entries)
461 462 463 464
{
	return le16_to_cpu(((struct dx_countlimit *) entries)->count);
}

465
static inline unsigned dx_get_limit(struct dx_entry *entries)
466 467 468 469
{
	return le16_to_cpu(((struct dx_countlimit *) entries)->limit);
}

470
static inline void dx_set_count(struct dx_entry *entries, unsigned value)
471 472 473 474
{
	((struct dx_countlimit *) entries)->count = cpu_to_le16(value);
}

475
static inline void dx_set_limit(struct dx_entry *entries, unsigned value)
476 477 478 479
{
	((struct dx_countlimit *) entries)->limit = cpu_to_le16(value);
}

480
static inline unsigned dx_root_limit(struct inode *dir, unsigned infosize)
481
{
482 483
	unsigned entry_space = dir->i_sb->s_blocksize - EXT4_DIR_REC_LEN(1) -
		EXT4_DIR_REC_LEN(2) - infosize;
484 485 486 487

	if (EXT4_HAS_RO_COMPAT_FEATURE(dir->i_sb,
				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
		entry_space -= sizeof(struct dx_tail);
488
	return entry_space / sizeof(struct dx_entry);
489 490
}

491
static inline unsigned dx_node_limit(struct inode *dir)
492
{
493
	unsigned entry_space = dir->i_sb->s_blocksize - EXT4_DIR_REC_LEN(0);
494 495 496 497

	if (EXT4_HAS_RO_COMPAT_FEATURE(dir->i_sb,
				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
		entry_space -= sizeof(struct dx_tail);
498
	return entry_space / sizeof(struct dx_entry);
499 500 501 502 503 504
}

/*
 * Debug
 */
#ifdef DX_DEBUG
505
static void dx_show_index(char * label, struct dx_entry *entries)
506
{
A
Andrew Morton 已提交
507
	int i, n = dx_get_count (entries);
508
	printk(KERN_DEBUG "%s index ", label);
A
Andrew Morton 已提交
509
	for (i = 0; i < n; i++) {
510
		printk("%x->%lu ", i ? dx_get_hash(entries + i) :
A
Aneesh Kumar K.V 已提交
511
				0, (unsigned long)dx_get_block(entries + i));
A
Andrew Morton 已提交
512 513
	}
	printk("\n");
514 515 516 517 518 519 520 521 522
}

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

523
static struct stats dx_show_leaf(struct dx_hash_info *hinfo, struct ext4_dir_entry_2 *de,
524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539
				 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++);
540
				ext4fs_dirhash(de->name, de->name_len, &h);
541
				printk(":%x.%u ", h.hash,
542
				       (unsigned) ((char *) de - base));
543
			}
544
			space += EXT4_DIR_REC_LEN(de->name_len);
545 546
			names++;
		}
547
		de = ext4_next_entry(de, size);
548 549 550 551 552 553 554 555 556
	}
	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;
557
	unsigned count = dx_get_count(entries), names = 0, space = 0, i;
558 559 560 561 562 563
	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
564 565
		ext4_lblk_t block = dx_get_block(entries);
		ext4_lblk_t hash  = i ? dx_get_hash(entries): 0;
566 567 568
		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);
569
		if (!(bh = ext4_bread (NULL,dir, block, 0,&err))) continue;
570 571
		stats = levels?
		   dx_show_entries(hinfo, dir, ((struct dx_node *) bh->b_data)->entries, levels - 1):
572
		   dx_show_leaf(hinfo, (struct ext4_dir_entry_2 *) bh->b_data, blocksize, 0);
573 574 575
		names += stats.names;
		space += stats.space;
		bcount += stats.bcount;
576
		brelse(bh);
577 578
	}
	if (bcount)
579
		printk(KERN_DEBUG "%snames %u, fullness %u (%u%%)\n",
580 581
		       levels ? "" : "   ", names, space/bcount,
		       (space/bcount)*100/blocksize);
582 583 584 585 586 587 588 589 590 591 592 593 594 595
	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 *
596
dx_probe(const struct qstr *d_name, struct inode *dir,
597 598 599 600 601 602 603 604 605 606
	 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;
C
Carlos Maiolino 已提交
607 608 609
	if (!(bh = ext4_bread(NULL, dir, 0, 0, err))) {
		if (*err == 0)
			*err = ERR_BAD_DX_DIR;
610
		goto fail;
C
Carlos Maiolino 已提交
611
	}
612 613 614 615
	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) {
616
		ext4_warning(dir->i_sb, "Unrecognised inode hash code %d",
617 618 619 620 621 622
			     root->info.hash_version);
		brelse(bh);
		*err = ERR_BAD_DX_DIR;
		goto fail;
	}
	hinfo->hash_version = root->info.hash_version;
623 624
	if (hinfo->hash_version <= DX_HASH_TEA)
		hinfo->hash_version += EXT4_SB(dir->i_sb)->s_hash_unsigned;
625
	hinfo->seed = EXT4_SB(dir->i_sb)->s_hash_seed;
626 627
	if (d_name)
		ext4fs_dirhash(d_name->name, d_name->len, hinfo);
628 629 630
	hash = hinfo->hash;

	if (root->info.unused_flags & 1) {
631
		ext4_warning(dir->i_sb, "Unimplemented inode hash flags: %#06x",
632 633 634 635 636 637 638
			     root->info.unused_flags);
		brelse(bh);
		*err = ERR_BAD_DX_DIR;
		goto fail;
	}

	if ((indirect = root->info.indirect_levels) > 1) {
639
		ext4_warning(dir->i_sb, "Unimplemented inode hash depth: %#06x",
640 641 642 643 644 645
			     root->info.indirect_levels);
		brelse(bh);
		*err = ERR_BAD_DX_DIR;
		goto fail;
	}

646 647 648 649 650 651 652 653 654
	if (!buffer_verified(bh) &&
	    !ext4_dx_csum_verify(dir, (struct ext4_dir_entry *)bh->b_data)) {
		ext4_warning(dir->i_sb, "Root failed checksum");
		brelse(bh);
		*err = ERR_BAD_DX_DIR;
		goto fail;
	}
	set_buffer_verified(bh);

655 656
	entries = (struct dx_entry *) (((char *)&root->info) +
				       root->info.info_length);
657 658 659

	if (dx_get_limit(entries) != dx_root_limit(dir,
						   root->info.info_length)) {
660
		ext4_warning(dir->i_sb, "dx entry: limit != root limit");
661 662 663 664 665
		brelse(bh);
		*err = ERR_BAD_DX_DIR;
		goto fail;
	}

666
	dxtrace(printk("Look up %x", hash));
667 668 669
	while (1)
	{
		count = dx_get_count(entries);
670
		if (!count || count > dx_get_limit(entries)) {
671
			ext4_warning(dir->i_sb,
672 673 674 675 676 677
				     "dx entry: no count or count > limit");
			brelse(bh);
			*err = ERR_BAD_DX_DIR;
			goto fail2;
		}

678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711
		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;
C
Carlos Maiolino 已提交
712 713 714
		if (!(bh = ext4_bread(NULL, dir, dx_get_block(at), 0, err))) {
			if (!(*err))
				*err = ERR_BAD_DX_DIR;
715
			goto fail2;
C
Carlos Maiolino 已提交
716
		}
717
		at = entries = ((struct dx_node *) bh->b_data)->entries;
718 719 720 721 722 723 724 725 726 727 728

		if (!buffer_verified(bh) &&
		    !ext4_dx_csum_verify(dir,
					 (struct ext4_dir_entry *)bh->b_data)) {
			ext4_warning(dir->i_sb, "Node failed checksum");
			brelse(bh);
			*err = ERR_BAD_DX_DIR;
			goto fail;
		}
		set_buffer_verified(bh);

729
		if (dx_get_limit(entries) != dx_node_limit (dir)) {
730
			ext4_warning(dir->i_sb,
731 732 733 734 735
				     "dx entry: limit != node limit");
			brelse(bh);
			*err = ERR_BAD_DX_DIR;
			goto fail2;
		}
736
		frame++;
737
		frame->bh = NULL;
738 739 740 741 742 743 744
	}
fail2:
	while (frame >= frame_in) {
		brelse(frame->bh);
		frame--;
	}
fail:
745
	if (*err == ERR_BAD_DX_DIR)
746
		ext4_warning(dir->i_sb,
Z
Zheng Liu 已提交
747
			     "Corrupt dir inode %lu, running e2fsck is "
748
			     "recommended.", dir->i_ino);
749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778
	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.
 */
779
static int ext4_htree_next_block(struct inode *dir, __u32 hash,
780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824
				 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--) {
825
		if (!(bh = ext4_bread(NULL, dir, dx_get_block(p->at),
C
Carlos Maiolino 已提交
826 827 828 829 830 831 832
				      0, &err))) {
			if (!err) {
				ext4_error(dir->i_sb,
					   "Directory hole detected on inode %lu\n",
					   dir->i_ino);
				return -EIO;
			}
833
			return err; /* Failure */
C
Carlos Maiolino 已提交
834
		}
835 836 837 838 839 840 841 842 843

		if (!buffer_verified(bh) &&
		    !ext4_dx_csum_verify(dir,
					 (struct ext4_dir_entry *)bh->b_data)) {
			ext4_warning(dir->i_sb, "Node failed checksum");
			return -EIO;
		}
		set_buffer_verified(bh);

844
		p++;
845
		brelse(p->bh);
846 847 848 849 850 851 852 853 854 855 856 857 858
		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
859
				  struct inode *dir, ext4_lblk_t block,
860 861 862 863
				  struct dx_hash_info *hinfo,
				  __u32 start_hash, __u32 start_minor_hash)
{
	struct buffer_head *bh;
864
	struct ext4_dir_entry_2 *de, *top;
865
	int err = 0, count = 0;
866

A
Aneesh Kumar K.V 已提交
867 868
	dxtrace(printk(KERN_INFO "In htree dirblock_to_tree: block %lu\n",
							(unsigned long)block));
C
Carlos Maiolino 已提交
869 870 871 872 873 874 875
	if (!(bh = ext4_bread(NULL, dir, block, 0, &err))) {
		if (!err) {
			err = -EIO;
			ext4_error(dir->i_sb,
				   "Directory hole detected on inode %lu\n",
				   dir->i_ino);
		}
876
		return err;
C
Carlos Maiolino 已提交
877
	}
878

879 880 881 882 883
	if (!buffer_verified(bh) &&
	    !ext4_dirent_csum_verify(dir, (struct ext4_dir_entry *)bh->b_data))
		return -EIO;
	set_buffer_verified(bh);

884 885
	de = (struct ext4_dir_entry_2 *) bh->b_data;
	top = (struct ext4_dir_entry_2 *) ((char *) de +
886
					   dir->i_sb->s_blocksize -
887
					   EXT4_DIR_REC_LEN(0));
888
	for (; de < top; de = ext4_next_entry(de, dir->i_sb->s_blocksize)) {
889
		if (ext4_check_dir_entry(dir, NULL, de, bh,
890
				bh->b_data, bh->b_size,
891 892
				(block<<EXT4_BLOCK_SIZE_BITS(dir->i_sb))
					 + ((char *)de - bh->b_data))) {
893 894 895
			/* 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;
896
			brelse(bh);
897 898
			return count;
		}
899
		ext4fs_dirhash(de->name, de->name_len, hinfo);
900 901 902 903 904 905
		if ((hinfo->hash < start_hash) ||
		    ((hinfo->hash == start_hash) &&
		     (hinfo->minor_hash < start_minor_hash)))
			continue;
		if (de->inode == 0)
			continue;
906
		if ((err = ext4_htree_store_dirent(dir_file,
907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925
				   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.
 */
926
int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
927 928 929
			 __u32 start_minor_hash, __u32 *next_hash)
{
	struct dx_hash_info hinfo;
930
	struct ext4_dir_entry_2 *de;
931 932
	struct dx_frame frames[2], *frame;
	struct inode *dir;
A
Aneesh Kumar K.V 已提交
933
	ext4_lblk_t block;
934
	int count = 0;
A
Aneesh Kumar K.V 已提交
935
	int ret, err;
936 937
	__u32 hashval;

938
	dxtrace(printk(KERN_DEBUG "In htree_fill_tree, start hash: %x:%x\n",
939
		       start_hash, start_minor_hash));
940
	dir = dir_file->f_path.dentry->d_inode;
941
	if (!(ext4_test_inode_flag(dir, EXT4_INODE_INDEX))) {
942
		hinfo.hash_version = EXT4_SB(dir->i_sb)->s_def_hash_version;
943 944 945
		if (hinfo.hash_version <= DX_HASH_TEA)
			hinfo.hash_version +=
				EXT4_SB(dir->i_sb)->s_hash_unsigned;
946
		hinfo.seed = EXT4_SB(dir->i_sb)->s_hash_seed;
947 948 949 950 951 952 953
		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;
954
	frame = dx_probe(NULL, dir, &hinfo, frames, &err);
955 956 957 958 959
	if (!frame)
		return err;

	/* Add '.' and '..' from the htree header */
	if (!start_hash && !start_minor_hash) {
960 961
		de = (struct ext4_dir_entry_2 *) frames[0].bh->b_data;
		if ((err = ext4_htree_store_dirent(dir_file, 0, 0, de)) != 0)
962 963 964 965
			goto errout;
		count++;
	}
	if (start_hash < 2 || (start_hash ==2 && start_minor_hash==0)) {
966
		de = (struct ext4_dir_entry_2 *) frames[0].bh->b_data;
967
		de = ext4_next_entry(de, dir->i_sb->s_blocksize);
968
		if ((err = ext4_htree_store_dirent(dir_file, 2, 0, de)) != 0)
969 970 971 972 973 974 975 976 977 978 979 980 981 982
			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;
983
		ret = ext4_htree_next_block(dir, HASH_NB_ALWAYS,
984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999
					    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);
1000 1001
	dxtrace(printk(KERN_DEBUG "Fill tree: returned %d entries, "
		       "next hash: %x\n", count, *next_hash));
1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012
	return count;
errout:
	dx_release(frames);
	return (err);
}


/*
 * Directory block splitting, compacting
 */

1013 1014 1015 1016
/*
 * Create map of hash values, offsets, and sizes, stored at end of block.
 * Returns number of entries mapped.
 */
1017 1018 1019
static int dx_make_map(struct ext4_dir_entry_2 *de, unsigned blocksize,
		       struct dx_hash_info *hinfo,
		       struct dx_map_entry *map_tail)
1020 1021 1022 1023 1024
{
	int count = 0;
	char *base = (char *) de;
	struct dx_hash_info h = *hinfo;

1025
	while ((char *) de < base + blocksize) {
1026
		if (de->name_len && de->inode) {
1027
			ext4fs_dirhash(de->name, de->name_len, &h);
1028 1029
			map_tail--;
			map_tail->hash = h.hash;
1030
			map_tail->offs = ((char *) de - base)>>2;
1031
			map_tail->size = le16_to_cpu(de->rec_len);
1032 1033 1034 1035
			count++;
			cond_resched();
		}
		/* XXX: do we need to check rec_len == 0 case? -Chris */
1036
		de = ext4_next_entry(de, blocksize);
1037 1038 1039 1040
	}
	return count;
}

1041
/* Sort map by hash value */
1042 1043
static void dx_sort_map (struct dx_map_entry *map, unsigned count)
{
A
Andrew Morton 已提交
1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
	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)
1061
				continue;
A
Andrew Morton 已提交
1062 1063
			swap(*(q+1), *q);
			more = 1;
1064 1065 1066 1067
		}
	} while(more);
}

A
Aneesh Kumar K.V 已提交
1068
static void dx_insert_block(struct dx_frame *frame, u32 hash, ext4_lblk_t block)
1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082
{
	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);
}

/*
1083
 * NOTE! unlike strncmp, ext4_match returns 1 for success, 0 for failure.
1084
 *
1085
 * `len <= EXT4_NAME_LEN' is guaranteed by caller.
1086 1087
 * `de != NULL' is guaranteed by caller.
 */
1088 1089
static inline int ext4_match (int len, const char * const name,
			      struct ext4_dir_entry_2 * de)
1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100
{
	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
 */
1101
static inline int search_dirblock(struct buffer_head *bh,
1102
				  struct inode *dir,
1103
				  const struct qstr *d_name,
1104
				  unsigned int offset,
1105
				  struct ext4_dir_entry_2 ** res_dir)
1106
{
1107
	struct ext4_dir_entry_2 * de;
1108 1109
	char * dlimit;
	int de_len;
1110 1111
	const char *name = d_name->name;
	int namelen = d_name->len;
1112

1113
	de = (struct ext4_dir_entry_2 *) bh->b_data;
1114 1115 1116 1117 1118 1119
	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 &&
1120
		    ext4_match (namelen, name, de)) {
1121
			/* found a match - just to be sure, do a full check */
1122 1123
			if (ext4_check_dir_entry(dir, NULL, de, bh, bh->b_data,
						 bh->b_size, offset))
1124 1125 1126 1127 1128
				return -1;
			*res_dir = de;
			return 1;
		}
		/* prevent looping on a bad block */
1129 1130
		de_len = ext4_rec_len_from_disk(de->rec_len,
						dir->i_sb->s_blocksize);
1131 1132 1133
		if (de_len <= 0)
			return -1;
		offset += de_len;
1134
		de = (struct ext4_dir_entry_2 *) ((char *) de + de_len);
1135 1136 1137 1138
	}
	return 0;
}

1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153
static int is_dx_internal_node(struct inode *dir, ext4_lblk_t block,
			       struct ext4_dir_entry *de)
{
	struct super_block *sb = dir->i_sb;

	if (!is_dx(dir))
		return 0;
	if (block == 0)
		return 1;
	if (de->inode == 0 &&
	    ext4_rec_len_from_disk(de->rec_len, sb->s_blocksize) ==
			sb->s_blocksize)
		return 1;
	return 0;
}
1154 1155

/*
1156
 *	ext4_find_entry()
1157 1158 1159 1160 1161 1162 1163 1164 1165
 *
 * 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.
 */
1166 1167
static struct buffer_head * ext4_find_entry (struct inode *dir,
					const struct qstr *d_name,
1168
					struct ext4_dir_entry_2 ** res_dir)
1169
{
1170 1171 1172
	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
1173
	ext4_lblk_t start, block, b;
1174
	const u8 *name = d_name->name;
1175 1176 1177 1178 1179
	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
1180 1181
	ext4_lblk_t  nblocks;
	int i, err;
1182 1183 1184 1185
	int namelen;

	*res_dir = NULL;
	sb = dir->i_sb;
1186
	namelen = d_name->len;
1187
	if (namelen > EXT4_NAME_LEN)
1188
		return NULL;
1189
	if ((namelen <= 2) && (name[0] == '.') &&
1190
	    (name[1] == '.' || name[1] == '\0')) {
1191 1192 1193 1194 1195 1196 1197 1198
		/*
		 * "." or ".." will only be in the first block
		 * NFS may look up ".."; "." should be handled by the VFS
		 */
		block = start = 0;
		nblocks = 1;
		goto restart;
	}
1199
	if (is_dx(dir)) {
1200
		bh = ext4_dx_find_entry(dir, d_name, res_dir, &err);
1201 1202 1203 1204 1205 1206 1207
		/*
		 * 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;
1208 1209
		dxtrace(printk(KERN_DEBUG "ext4_find_entry: dx failed, "
			       "falling back\n"));
1210
	}
1211 1212
	nblocks = dir->i_size >> EXT4_BLOCK_SIZE_BITS(sb);
	start = EXT4_I(dir)->i_dir_start_lookup;
1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235
	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++;
1236
				bh = ext4_getblk(NULL, dir, b++, 0, &err);
1237 1238
				bh_use[ra_max] = bh;
				if (bh)
1239 1240
					ll_rw_block(READ | REQ_META | REQ_PRIO,
						    1, &bh);
1241 1242 1243 1244 1245 1246 1247
			}
		}
		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 */
1248 1249
			EXT4_ERROR_INODE(dir, "reading directory lblock %lu",
					 (unsigned long) block);
1250 1251 1252
			brelse(bh);
			goto next;
		}
1253
		if (!buffer_verified(bh) &&
1254 1255
		    !is_dx_internal_node(dir, block,
					 (struct ext4_dir_entry *)bh->b_data) &&
1256 1257 1258 1259 1260 1261 1262 1263
		    !ext4_dirent_csum_verify(dir,
				(struct ext4_dir_entry *)bh->b_data)) {
			EXT4_ERROR_INODE(dir, "checksumming directory "
					 "block %lu", (unsigned long)block);
			brelse(bh);
			goto next;
		}
		set_buffer_verified(bh);
1264
		i = search_dirblock(bh, dir, d_name,
1265
			    block << EXT4_BLOCK_SIZE_BITS(sb), res_dir);
1266
		if (i == 1) {
1267
			EXT4_I(dir)->i_dir_start_lookup = block;
1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284
			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;
1285
	nblocks = dir->i_size >> EXT4_BLOCK_SIZE_BITS(sb);
1286 1287 1288 1289 1290 1291 1292 1293
	if (block < nblocks) {
		start = 0;
		goto restart;
	}

cleanup_and_exit:
	/* Clean up the read-ahead blocks */
	for (; ra_ptr < ra_max; ra_ptr++)
1294
		brelse(bh_use[ra_ptr]);
1295 1296 1297
	return ret;
}

1298
static struct buffer_head * ext4_dx_find_entry(struct inode *dir, const struct qstr *d_name,
1299
		       struct ext4_dir_entry_2 **res_dir, int *err)
1300
{
1301
	struct super_block * sb = dir->i_sb;
1302 1303 1304
	struct dx_hash_info	hinfo;
	struct dx_frame frames[2], *frame;
	struct buffer_head *bh;
A
Aneesh Kumar K.V 已提交
1305
	ext4_lblk_t block;
1306 1307
	int retval;

1308 1309
	if (!(frame = dx_probe(d_name, dir, &hinfo, frames, err)))
		return NULL;
1310 1311
	do {
		block = dx_get_block(frame->at);
C
Carlos Maiolino 已提交
1312 1313 1314 1315 1316 1317 1318
		if (!(bh = ext4_bread(NULL, dir, block, 0, err))) {
			if (!(*err)) {
				*err = -EIO;
				ext4_error(dir->i_sb,
					   "Directory hole detected on inode %lu\n",
					   dir->i_ino);
			}
1319
			goto errout;
C
Carlos Maiolino 已提交
1320
		}
1321

1322 1323 1324 1325 1326 1327 1328 1329 1330 1331
		if (!buffer_verified(bh) &&
		    !ext4_dirent_csum_verify(dir,
				(struct ext4_dir_entry *)bh->b_data)) {
			EXT4_ERROR_INODE(dir, "checksumming directory "
					 "block %lu", (unsigned long)block);
			brelse(bh);
			*err = -EIO;
			goto errout;
		}
		set_buffer_verified(bh);
1332 1333 1334 1335 1336 1337
		retval = search_dirblock(bh, dir, d_name,
					 block << EXT4_BLOCK_SIZE_BITS(sb),
					 res_dir);
		if (retval == 1) { 	/* Success! */
			dx_release(frames);
			return bh;
1338
		}
1339
		brelse(bh);
1340 1341 1342 1343 1344
		if (retval == -1) {
			*err = ERR_BAD_DX_DIR;
			goto errout;
		}

1345
		/* Check to see if we should continue to search */
1346
		retval = ext4_htree_next_block(dir, hinfo.hash, frame,
1347 1348
					       frames, NULL);
		if (retval < 0) {
1349
			ext4_warning(sb,
1350 1351 1352 1353 1354 1355 1356 1357 1358
			     "error reading index page in directory #%lu",
			     dir->i_ino);
			*err = retval;
			goto errout;
		}
	} while (retval == 1);

	*err = -ENOENT;
errout:
1359
	dxtrace(printk(KERN_DEBUG "%s not found\n", d_name->name));
1360 1361 1362 1363
	dx_release (frames);
	return NULL;
}

A
Al Viro 已提交
1364
static struct dentry *ext4_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1365
{
1366 1367 1368
	struct inode *inode;
	struct ext4_dir_entry_2 *de;
	struct buffer_head *bh;
1369

1370
	if (dentry->d_name.len > EXT4_NAME_LEN)
1371 1372
		return ERR_PTR(-ENAMETOOLONG);

1373
	bh = ext4_find_entry(dir, &dentry->d_name, &de);
1374 1375
	inode = NULL;
	if (bh) {
1376
		__u32 ino = le32_to_cpu(de->inode);
1377
		brelse(bh);
1378
		if (!ext4_valid_inum(dir->i_sb, ino)) {
1379
			EXT4_ERROR_INODE(dir, "bad inode number: %u", ino);
1380
			return ERR_PTR(-EIO);
1381
		}
1382 1383 1384 1385 1386 1387
		if (unlikely(ino == dir->i_ino)) {
			EXT4_ERROR_INODE(dir, "'%.*s' linked to parent dir",
					 dentry->d_name.len,
					 dentry->d_name.name);
			return ERR_PTR(-EIO);
		}
1388
		inode = ext4_iget(dir->i_sb, ino);
1389 1390 1391 1392 1393
		if (inode == ERR_PTR(-ESTALE)) {
			EXT4_ERROR_INODE(dir,
					 "deleted inode referenced: %u",
					 ino);
			return ERR_PTR(-EIO);
1394
		}
1395 1396 1397 1398 1399
	}
	return d_splice_alias(inode, dentry);
}


1400
struct dentry *ext4_get_parent(struct dentry *child)
1401
{
1402
	__u32 ino;
1403
	static const struct qstr dotdot = QSTR_INIT("..", 2);
1404
	struct ext4_dir_entry_2 * de;
1405 1406
	struct buffer_head *bh;

1407
	bh = ext4_find_entry(child->d_inode, &dotdot, &de);
1408 1409 1410 1411 1412
	if (!bh)
		return ERR_PTR(-ENOENT);
	ino = le32_to_cpu(de->inode);
	brelse(bh);

1413
	if (!ext4_valid_inum(child->d_inode->i_sb, ino)) {
1414 1415
		EXT4_ERROR_INODE(child->d_inode,
				 "bad parent inode number: %u", ino);
1416
		return ERR_PTR(-EIO);
1417 1418
	}

1419
	return d_obtain_alias(ext4_iget(child->d_inode->i_sb, ino));
1420 1421 1422
}

#define S_SHIFT 12
1423 1424 1425 1426 1427 1428 1429 1430
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,
1431 1432
};

1433 1434
static inline void ext4_set_de_type(struct super_block *sb,
				struct ext4_dir_entry_2 *de,
1435
				umode_t mode) {
1436 1437
	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE))
		de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
1438 1439
}

1440 1441 1442 1443
/*
 * Move count entries from end of map between two memory locations.
 * Returns pointer to last entry moved.
 */
1444
static struct ext4_dir_entry_2 *
1445 1446
dx_move_dirents(char *from, char *to, struct dx_map_entry *map, int count,
		unsigned blocksize)
1447 1448 1449 1450
{
	unsigned rec_len = 0;

	while (count--) {
1451
		struct ext4_dir_entry_2 *de = (struct ext4_dir_entry_2 *)
1452
						(from + (map->offs<<2));
1453
		rec_len = EXT4_DIR_REC_LEN(de->name_len);
1454
		memcpy (to, de, rec_len);
1455
		((struct ext4_dir_entry_2 *) to)->rec_len =
1456
				ext4_rec_len_to_disk(rec_len, blocksize);
1457 1458 1459 1460
		de->inode = 0;
		map++;
		to += rec_len;
	}
1461
	return (struct ext4_dir_entry_2 *) (to - rec_len);
1462 1463
}

1464 1465 1466 1467
/*
 * Compact each dir entry in the range to the minimal rec_len.
 * Returns pointer to last entry in range.
 */
1468
static struct ext4_dir_entry_2* dx_pack_dirents(char *base, unsigned blocksize)
1469
{
1470
	struct ext4_dir_entry_2 *next, *to, *prev, *de = (struct ext4_dir_entry_2 *) base;
1471 1472 1473
	unsigned rec_len = 0;

	prev = to = de;
1474
	while ((char*)de < base + blocksize) {
1475
		next = ext4_next_entry(de, blocksize);
1476
		if (de->inode && de->name_len) {
1477
			rec_len = EXT4_DIR_REC_LEN(de->name_len);
1478 1479
			if (de > to)
				memmove(to, de, rec_len);
1480
			to->rec_len = ext4_rec_len_to_disk(rec_len, blocksize);
1481
			prev = to;
1482
			to = (struct ext4_dir_entry_2 *) (((char *) to) + rec_len);
1483 1484 1485 1486 1487 1488
		}
		de = next;
	}
	return prev;
}

1489 1490 1491 1492 1493
/*
 * 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.
 */
1494
static struct ext4_dir_entry_2 *do_split(handle_t *handle, struct inode *dir,
1495 1496 1497 1498 1499 1500
			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
1501
	ext4_lblk_t newblock;
1502 1503 1504
	u32 hash2;
	struct dx_map_entry *map;
	char *data1 = (*bh)->b_data, *data2;
1505
	unsigned split, move, size;
1506
	struct ext4_dir_entry_2 *de = NULL, *de2;
1507 1508
	struct ext4_dir_entry_tail *t;
	int	csum_size = 0;
1509
	int	err = 0, i;
1510

1511 1512 1513 1514
	if (EXT4_HAS_RO_COMPAT_FEATURE(dir->i_sb,
				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
		csum_size = sizeof(struct ext4_dir_entry_tail);

1515
	bh2 = ext4_append (handle, dir, &newblock, &err);
1516 1517 1518 1519 1520 1521 1522
	if (!(bh2)) {
		brelse(*bh);
		*bh = NULL;
		goto errout;
	}

	BUFFER_TRACE(*bh, "get_write_access");
1523
	err = ext4_journal_get_write_access(handle, *bh);
1524 1525 1526
	if (err)
		goto journal_error;

1527
	BUFFER_TRACE(frame->bh, "get_write_access");
1528
	err = ext4_journal_get_write_access(handle, frame->bh);
1529 1530 1531 1532 1533 1534 1535
	if (err)
		goto journal_error;

	data2 = bh2->b_data;

	/* create map in the end of data2 block */
	map = (struct dx_map_entry *) (data2 + blocksize);
1536
	count = dx_make_map((struct ext4_dir_entry_2 *) data1,
1537 1538
			     blocksize, hinfo, map);
	map -= count;
1539
	dx_sort_map(map, count);
1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551
	/* 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;
1552 1553
	hash2 = map[split].hash;
	continued = hash2 == map[split - 1].hash;
A
Aneesh Kumar K.V 已提交
1554 1555 1556
	dxtrace(printk(KERN_INFO "Split block %lu at %x, %i/%i\n",
			(unsigned long)dx_get_block(frame->at),
					hash2, split, count-split));
1557 1558

	/* Fancy dance to stay within two buffers */
1559
	de2 = dx_move_dirents(data1, data2, map + split, count - split, blocksize);
1560
	de = dx_pack_dirents(data1, blocksize);
1561 1562
	de->rec_len = ext4_rec_len_to_disk(data1 + (blocksize - csum_size) -
					   (char *) de,
1563
					   blocksize);
1564 1565
	de2->rec_len = ext4_rec_len_to_disk(data2 + (blocksize - csum_size) -
					    (char *) de2,
1566
					    blocksize);
1567 1568 1569 1570 1571 1572 1573 1574
	if (csum_size) {
		t = EXT4_DIRENT_TAIL(data2, blocksize);
		initialize_dirent_tail(t, blocksize);

		t = EXT4_DIRENT_TAIL(data1, blocksize);
		initialize_dirent_tail(t, blocksize);
	}

1575 1576
	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));
1577 1578 1579 1580 1581 1582 1583

	/* Which block gets the new entry? */
	if (hinfo->hash >= hash2)
	{
		swap(*bh, bh2);
		de = de2;
	}
1584
	dx_insert_block(frame, hash2 + continued, newblock);
1585
	err = ext4_handle_dirty_dirent_node(handle, dir, bh2);
1586 1587
	if (err)
		goto journal_error;
1588
	err = ext4_handle_dirty_dx_node(handle, dir, frame->bh);
1589 1590
	if (err)
		goto journal_error;
1591 1592
	brelse(bh2);
	dxtrace(dx_show_index("frame", frame->entries));
1593
	return de;
1594 1595 1596 1597 1598 1599 1600 1601 1602

journal_error:
	brelse(*bh);
	brelse(bh2);
	*bh = NULL;
	ext4_std_error(dir->i_sb, err);
errout:
	*error = err;
	return NULL;
1603 1604
}

1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661
int ext4_find_dest_de(struct inode *dir, struct inode *inode,
		      struct buffer_head *bh,
		      void *buf, int buf_size,
		      const char *name, int namelen,
		      struct ext4_dir_entry_2 **dest_de)
{
	struct ext4_dir_entry_2 *de;
	unsigned short reclen = EXT4_DIR_REC_LEN(namelen);
	int nlen, rlen;
	unsigned int offset = 0;
	char *top;

	de = (struct ext4_dir_entry_2 *)buf;
	top = buf + buf_size - reclen;
	while ((char *) de <= top) {
		if (ext4_check_dir_entry(dir, NULL, de, bh,
					 buf, buf_size, offset))
			return -EIO;
		if (ext4_match(namelen, name, de))
			return -EEXIST;
		nlen = EXT4_DIR_REC_LEN(de->name_len);
		rlen = ext4_rec_len_from_disk(de->rec_len, buf_size);
		if ((de->inode ? rlen - nlen : rlen) >= reclen)
			break;
		de = (struct ext4_dir_entry_2 *)((char *)de + rlen);
		offset += rlen;
	}
	if ((char *) de > top)
		return -ENOSPC;

	*dest_de = de;
	return 0;
}

void ext4_insert_dentry(struct inode *inode,
			struct ext4_dir_entry_2 *de,
			int buf_size,
			const char *name, int namelen)
{

	int nlen, rlen;

	nlen = EXT4_DIR_REC_LEN(de->name_len);
	rlen = ext4_rec_len_from_disk(de->rec_len, buf_size);
	if (de->inode) {
		struct ext4_dir_entry_2 *de1 =
				(struct ext4_dir_entry_2 *)((char *)de + nlen);
		de1->rec_len = ext4_rec_len_to_disk(rlen - nlen, buf_size);
		de->rec_len = ext4_rec_len_to_disk(nlen, buf_size);
		de = de1;
	}
	de->file_type = EXT4_FT_UNKNOWN;
	de->inode = cpu_to_le32(inode->i_ino);
	ext4_set_de_type(inode->i_sb, de, inode->i_mode);
	de->name_len = namelen;
	memcpy(de->name, name, namelen);
}
1662 1663 1664 1665 1666 1667 1668 1669 1670
/*
 * 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.
 */
static int add_dirent_to_buf(handle_t *handle, struct dentry *dentry,
1671
			     struct inode *inode, struct ext4_dir_entry_2 *de,
1672
			     struct buffer_head *bh)
1673 1674 1675 1676
{
	struct inode	*dir = dentry->d_parent->d_inode;
	const char	*name = dentry->d_name.name;
	int		namelen = dentry->d_name.len;
1677
	unsigned int	blocksize = dir->i_sb->s_blocksize;
1678
	unsigned short	reclen;
1679
	int		csum_size = 0;
1680
	int		err;
1681 1682 1683 1684

	if (EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
		csum_size = sizeof(struct ext4_dir_entry_tail);
1685

1686
	reclen = EXT4_DIR_REC_LEN(namelen);
1687
	if (!de) {
1688 1689 1690 1691 1692
		err = ext4_find_dest_de(dir, inode,
					bh, bh->b_data, blocksize - csum_size,
					name, namelen, &de);
		if (err)
			return err;
1693 1694
	}
	BUFFER_TRACE(bh, "get_write_access");
1695
	err = ext4_journal_get_write_access(handle, bh);
1696
	if (err) {
1697
		ext4_std_error(dir->i_sb, err);
1698 1699 1700 1701
		return err;
	}

	/* By now the buffer is marked for journaling */
1702 1703
	ext4_insert_dentry(inode, de, blocksize, name, namelen);

1704 1705 1706 1707 1708 1709
	/*
	 * 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
1710
	 * ext4_new_inode() setting the times, but error
1711 1712 1713 1714
	 * 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
1715
	dir->i_mtime = dir->i_ctime = ext4_current_time(dir);
1716
	ext4_update_dx_flag(dir);
1717
	dir->i_version++;
1718
	ext4_mark_inode_dirty(handle, dir);
1719
	BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
1720
	err = ext4_handle_dirty_dirent_node(handle, dir, bh);
1721
	if (err)
1722
		ext4_std_error(dir->i_sb, err);
1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739
	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;
1740
	struct ext4_dir_entry_2	*de, *de2;
1741
	struct ext4_dir_entry_tail *t;
1742 1743 1744 1745 1746
	char		*data1, *top;
	unsigned	len;
	int		retval;
	unsigned	blocksize;
	struct dx_hash_info hinfo;
A
Aneesh Kumar K.V 已提交
1747
	ext4_lblk_t  block;
1748
	struct fake_dirent *fde;
1749 1750 1751 1752 1753
	int		csum_size = 0;

	if (EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
		csum_size = sizeof(struct ext4_dir_entry_tail);
1754 1755

	blocksize =  dir->i_sb->s_blocksize;
1756
	dxtrace(printk(KERN_DEBUG "Creating index: inode %lu\n", dir->i_ino));
1757
	retval = ext4_journal_get_write_access(handle, bh);
1758
	if (retval) {
1759
		ext4_std_error(dir->i_sb, retval);
1760 1761 1762 1763 1764
		brelse(bh);
		return retval;
	}
	root = (struct dx_root *) bh->b_data;

1765 1766 1767
	/* The 0th block becomes the root, move the dirents out */
	fde = &root->dotdot;
	de = (struct ext4_dir_entry_2 *)((char *)fde +
1768
		ext4_rec_len_from_disk(fde->rec_len, blocksize));
1769
	if ((char *) de >= (((char *) root) + blocksize)) {
1770
		EXT4_ERROR_INODE(dir, "invalid rec_len for '..'");
1771 1772 1773
		brelse(bh);
		return -EIO;
	}
1774
	len = ((char *) root) + (blocksize - csum_size) - (char *) de;
1775 1776

	/* Allocate new block for the 0th block's dirents */
1777
	bh2 = ext4_append(handle, dir, &block, &retval);
1778 1779 1780 1781
	if (!(bh2)) {
		brelse(bh);
		return retval;
	}
1782
	ext4_set_inode_flag(dir, EXT4_INODE_INDEX);
1783 1784 1785
	data1 = bh2->b_data;

	memcpy (data1, de, len);
1786
	de = (struct ext4_dir_entry_2 *) data1;
1787
	top = data1 + len;
1788
	while ((char *)(de2 = ext4_next_entry(de, blocksize)) < top)
1789
		de = de2;
1790 1791
	de->rec_len = ext4_rec_len_to_disk(data1 + (blocksize - csum_size) -
					   (char *) de,
1792
					   blocksize);
1793 1794 1795 1796 1797 1798

	if (csum_size) {
		t = EXT4_DIRENT_TAIL(data1, blocksize);
		initialize_dirent_tail(t, blocksize);
	}

1799
	/* Initialize the root; the dot dirents already exist */
1800
	de = (struct ext4_dir_entry_2 *) (&root->dotdot);
1801 1802
	de->rec_len = ext4_rec_len_to_disk(blocksize - EXT4_DIR_REC_LEN(2),
					   blocksize);
1803 1804
	memset (&root->info, 0, sizeof(root->info));
	root->info.info_length = sizeof(root->info);
1805
	root->info.hash_version = EXT4_SB(dir->i_sb)->s_def_hash_version;
1806
	entries = root->entries;
1807 1808 1809
	dx_set_block(entries, 1);
	dx_set_count(entries, 1);
	dx_set_limit(entries, dx_root_limit(dir, sizeof(root->info)));
1810 1811 1812

	/* Initialize as for dx_probe */
	hinfo.hash_version = root->info.hash_version;
1813 1814
	if (hinfo.hash_version <= DX_HASH_TEA)
		hinfo.hash_version += EXT4_SB(dir->i_sb)->s_hash_unsigned;
1815 1816
	hinfo.seed = EXT4_SB(dir->i_sb)->s_hash_seed;
	ext4fs_dirhash(name, namelen, &hinfo);
1817 1818 1819 1820 1821
	frame = frames;
	frame->entries = entries;
	frame->at = entries;
	frame->bh = bh;
	bh = bh2;
1822

1823
	ext4_handle_dirty_dx_node(handle, dir, frame->bh);
1824
	ext4_handle_dirty_dirent_node(handle, dir, bh);
1825

1826
	de = do_split(handle,dir, &bh, frame, &hinfo, &retval);
1827 1828 1829 1830 1831 1832 1833 1834
	if (!de) {
		/*
		 * Even if the block split failed, we have to properly write
		 * out all the changes we did so far. Otherwise we can end up
		 * with corrupted filesystem.
		 */
		ext4_mark_inode_dirty(handle, dir);
		dx_release(frames);
1835
		return retval;
1836 1837
	}
	dx_release(frames);
1838

1839 1840 1841
	retval = add_dirent_to_buf(handle, dentry, inode, de, bh);
	brelse(bh);
	return retval;
1842 1843 1844
}

/*
1845
 *	ext4_add_entry()
1846 1847
 *
 * adds a file entry to the specified directory, using the same
1848
 * semantics as ext4_find_entry(). It returns NULL if it failed.
1849 1850 1851 1852 1853
 *
 * 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.
 */
1854 1855
static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
			  struct inode *inode)
1856 1857
{
	struct inode *dir = dentry->d_parent->d_inode;
1858
	struct buffer_head *bh;
1859
	struct ext4_dir_entry_2 *de;
1860
	struct ext4_dir_entry_tail *t;
1861
	struct super_block *sb;
1862 1863 1864
	int	retval;
	int	dx_fallback=0;
	unsigned blocksize;
A
Aneesh Kumar K.V 已提交
1865
	ext4_lblk_t block, blocks;
1866 1867 1868 1869 1870
	int	csum_size = 0;

	if (EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
		csum_size = sizeof(struct ext4_dir_entry_tail);
1871 1872 1873 1874 1875

	sb = dir->i_sb;
	blocksize = sb->s_blocksize;
	if (!dentry->d_name.len)
		return -EINVAL;
1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886

	if (ext4_has_inline_data(dir)) {
		retval = ext4_try_add_inline_entry(handle, dentry, inode);
		if (retval < 0)
			return retval;
		if (retval == 1) {
			retval = 0;
			return retval;
		}
	}

1887
	if (is_dx(dir)) {
1888
		retval = ext4_dx_add_entry(handle, dentry, inode);
1889 1890
		if (!retval || (retval != ERR_BAD_DX_DIR))
			return retval;
1891
		ext4_clear_inode_flag(dir, EXT4_INODE_INDEX);
1892
		dx_fallback++;
1893
		ext4_mark_inode_dirty(handle, dir);
1894 1895
	}
	blocks = dir->i_size >> sb->s_blocksize_bits;
1896
	for (block = 0; block < blocks; block++) {
C
Carlos Maiolino 已提交
1897 1898 1899 1900 1901 1902 1903
		if (!(bh = ext4_bread(handle, dir, block, 0, &retval))) {
			if (!retval) {
				retval = -EIO;
				ext4_error(inode->i_sb,
					   "Directory hole detected on inode %lu\n",
					   inode->i_ino);
			}
1904
			return retval;
C
Carlos Maiolino 已提交
1905
		}
1906 1907 1908 1909 1910
		if (!buffer_verified(bh) &&
		    !ext4_dirent_csum_verify(dir,
				(struct ext4_dir_entry *)bh->b_data))
			return -EIO;
		set_buffer_verified(bh);
1911
		retval = add_dirent_to_buf(handle, dentry, inode, NULL, bh);
1912 1913
		if (retval != -ENOSPC) {
			brelse(bh);
1914
			return retval;
1915
		}
1916 1917

		if (blocks == 1 && !dx_fallback &&
1918 1919
		    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX))
			return make_indexed_dir(handle, dentry, inode, bh);
1920 1921
		brelse(bh);
	}
1922
	bh = ext4_append(handle, dir, &block, &retval);
1923 1924
	if (!bh)
		return retval;
1925
	de = (struct ext4_dir_entry_2 *) bh->b_data;
1926
	de->inode = 0;
1927 1928 1929 1930 1931 1932 1933
	de->rec_len = ext4_rec_len_to_disk(blocksize - csum_size, blocksize);

	if (csum_size) {
		t = EXT4_DIRENT_TAIL(bh->b_data, blocksize);
		initialize_dirent_tail(t, blocksize);
	}

1934 1935
	retval = add_dirent_to_buf(handle, dentry, inode, de, bh);
	brelse(bh);
1936 1937
	if (retval == 0)
		ext4_set_inode_state(inode, EXT4_STATE_NEWENTRY);
1938
	return retval;
1939 1940 1941 1942 1943
}

/*
 * Returns 0 for success, or a negative error value
 */
1944
static int ext4_dx_add_entry(handle_t *handle, struct dentry *dentry,
1945 1946 1947 1948 1949
			     struct inode *inode)
{
	struct dx_frame frames[2], *frame;
	struct dx_entry *entries, *at;
	struct dx_hash_info hinfo;
1950
	struct buffer_head *bh;
1951
	struct inode *dir = dentry->d_parent->d_inode;
1952
	struct super_block *sb = dir->i_sb;
1953
	struct ext4_dir_entry_2 *de;
1954 1955
	int err;

1956
	frame = dx_probe(&dentry->d_name, dir, &hinfo, frames, &err);
1957 1958 1959 1960 1961
	if (!frame)
		return err;
	entries = frame->entries;
	at = frame->at;

C
Carlos Maiolino 已提交
1962 1963 1964 1965 1966 1967 1968
	if (!(bh = ext4_bread(handle, dir, dx_get_block(frame->at), 0, &err))) {
		if (!err) {
			err = -EIO;
			ext4_error(dir->i_sb,
				   "Directory hole detected on inode %lu\n",
				   dir->i_ino);
		}
1969
		goto cleanup;
C
Carlos Maiolino 已提交
1970
	}
1971

1972 1973 1974 1975 1976
	if (!buffer_verified(bh) &&
	    !ext4_dirent_csum_verify(dir, (struct ext4_dir_entry *)bh->b_data))
		goto journal_error;
	set_buffer_verified(bh);

1977
	BUFFER_TRACE(bh, "get_write_access");
1978
	err = ext4_journal_get_write_access(handle, bh);
1979 1980 1981 1982
	if (err)
		goto journal_error;

	err = add_dirent_to_buf(handle, dentry, inode, NULL, bh);
1983
	if (err != -ENOSPC)
1984 1985 1986
		goto cleanup;

	/* Block full, should compress but for now just split */
1987
	dxtrace(printk(KERN_DEBUG "using %u of %u node entries\n",
1988 1989 1990
		       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
1991
		ext4_lblk_t newblock;
1992 1993 1994 1995 1996 1997 1998 1999
		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))) {
2000
			ext4_warning(sb, "Directory index full!");
2001 2002 2003
			err = -ENOSPC;
			goto cleanup;
		}
2004
		bh2 = ext4_append (handle, dir, &newblock, &err);
2005 2006 2007 2008
		if (!(bh2))
			goto cleanup;
		node2 = (struct dx_node *)(bh2->b_data);
		entries2 = node2->entries;
2009
		memset(&node2->fake, 0, sizeof(struct fake_dirent));
2010 2011
		node2->fake.rec_len = ext4_rec_len_to_disk(sb->s_blocksize,
							   sb->s_blocksize);
2012
		BUFFER_TRACE(frame->bh, "get_write_access");
2013
		err = ext4_journal_get_write_access(handle, frame->bh);
2014 2015 2016 2017 2018
		if (err)
			goto journal_error;
		if (levels) {
			unsigned icount1 = icount/2, icount2 = icount - icount1;
			unsigned hash2 = dx_get_hash(entries + icount1);
2019 2020
			dxtrace(printk(KERN_DEBUG "Split index %i/%i\n",
				       icount1, icount2));
2021 2022

			BUFFER_TRACE(frame->bh, "get_write_access"); /* index root */
2023
			err = ext4_journal_get_write_access(handle,
2024 2025 2026 2027
							     frames[0].bh);
			if (err)
				goto journal_error;

2028 2029 2030 2031 2032
			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));
2033 2034 2035 2036 2037 2038 2039

			/* 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);
			}
2040 2041 2042
			dx_insert_block(frames + 0, hash2, newblock);
			dxtrace(dx_show_index("node", frames[1].entries));
			dxtrace(dx_show_index("node",
2043
			       ((struct dx_node *) bh2->b_data)->entries));
2044
			err = ext4_handle_dirty_dx_node(handle, dir, bh2);
2045 2046 2047 2048
			if (err)
				goto journal_error;
			brelse (bh2);
		} else {
2049 2050
			dxtrace(printk(KERN_DEBUG
				       "Creating second level index...\n"));
2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064
			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;
2065
			err = ext4_journal_get_write_access(handle,
2066 2067 2068 2069
							     frame->bh);
			if (err)
				goto journal_error;
		}
2070
		err = ext4_handle_dirty_dx_node(handle, dir, frames[0].bh);
2071 2072 2073 2074
		if (err) {
			ext4_std_error(inode->i_sb, err);
			goto cleanup;
		}
2075 2076 2077 2078 2079 2080 2081 2082
	}
	de = do_split(handle, dir, &bh, frame, &hinfo, &err);
	if (!de)
		goto cleanup;
	err = add_dirent_to_buf(handle, dentry, inode, de, bh);
	goto cleanup;

journal_error:
2083
	ext4_std_error(dir->i_sb, err);
2084 2085 2086 2087 2088 2089 2090 2091
cleanup:
	if (bh)
		brelse(bh);
	dx_release(frames);
	return err;
}

/*
2092
 * ext4_delete_entry deletes a directory entry by merging it with the
2093 2094
 * previous entry
 */
2095 2096 2097 2098
static int ext4_delete_entry(handle_t *handle,
			     struct inode *dir,
			     struct ext4_dir_entry_2 *de_del,
			     struct buffer_head *bh)
2099
{
2100
	struct ext4_dir_entry_2 *de, *pde;
2101
	unsigned int blocksize = dir->i_sb->s_blocksize;
2102
	int csum_size = 0;
2103
	int i, err;
2104

2105 2106 2107 2108
	if (EXT4_HAS_RO_COMPAT_FEATURE(dir->i_sb,
				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
		csum_size = sizeof(struct ext4_dir_entry_tail);

2109 2110
	i = 0;
	pde = NULL;
2111
	de = (struct ext4_dir_entry_2 *) bh->b_data;
2112
	while (i < bh->b_size - csum_size) {
2113 2114
		if (ext4_check_dir_entry(dir, NULL, de, bh,
					 bh->b_data, bh->b_size, i))
2115 2116 2117
			return -EIO;
		if (de == de_del)  {
			BUFFER_TRACE(bh, "get_write_access");
2118 2119 2120 2121 2122
			err = ext4_journal_get_write_access(handle, bh);
			if (unlikely(err)) {
				ext4_std_error(dir->i_sb, err);
				return err;
			}
2123
			if (pde)
2124
				pde->rec_len = ext4_rec_len_to_disk(
2125 2126 2127 2128 2129
					ext4_rec_len_from_disk(pde->rec_len,
							       blocksize) +
					ext4_rec_len_from_disk(de->rec_len,
							       blocksize),
					blocksize);
2130 2131 2132
			else
				de->inode = 0;
			dir->i_version++;
2133
			BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
2134
			err = ext4_handle_dirty_dirent_node(handle, dir, bh);
2135 2136 2137 2138
			if (unlikely(err)) {
				ext4_std_error(dir->i_sb, err);
				return err;
			}
2139 2140
			return 0;
		}
2141
		i += ext4_rec_len_from_disk(de->rec_len, blocksize);
2142
		pde = de;
2143
		de = ext4_next_entry(de, blocksize);
2144 2145 2146 2147
	}
	return -ENOENT;
}

2148 2149 2150 2151 2152 2153 2154 2155 2156 2157
/*
 * 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) {
M
Miklos Szeredi 已提交
2158
			set_nlink(inode, 1);
2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170
			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)
{
2171 2172
	if (!S_ISDIR(inode->i_mode) || inode->i_nlink > 2)
		drop_nlink(inode);
2173 2174 2175
}


2176
static int ext4_add_nondir(handle_t *handle,
2177 2178
		struct dentry *dentry, struct inode *inode)
{
2179
	int err = ext4_add_entry(handle, dentry, inode);
2180
	if (!err) {
2181
		ext4_mark_inode_dirty(handle, inode);
A
Al Viro 已提交
2182
		unlock_new_inode(inode);
2183
		d_instantiate(dentry, inode);
2184 2185
		return 0;
	}
2186
	drop_nlink(inode);
A
Al Viro 已提交
2187
	unlock_new_inode(inode);
2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199
	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().
 */
A
Al Viro 已提交
2200
static int ext4_create(struct inode *dir, struct dentry *dentry, umode_t mode,
A
Al Viro 已提交
2201
		       bool excl)
2202 2203
{
	handle_t *handle;
2204
	struct inode *inode;
2205 2206
	int err, retries = 0;

2207
	dquot_initialize(dir);
2208

2209
retry:
2210 2211
	handle = ext4_journal_start(dir, EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
					EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 +
D
Dmitry Monakhov 已提交
2212
					EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb));
2213 2214 2215 2216
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	if (IS_DIRSYNC(dir))
2217
		ext4_handle_sync(handle);
2218

2219
	inode = ext4_new_inode(handle, dir, mode, &dentry->d_name, 0, NULL);
2220 2221
	err = PTR_ERR(inode);
	if (!IS_ERR(inode)) {
2222 2223 2224 2225
		inode->i_op = &ext4_file_inode_operations;
		inode->i_fop = &ext4_file_operations;
		ext4_set_aops(inode);
		err = ext4_add_nondir(handle, dentry, inode);
2226
	}
2227 2228
	ext4_journal_stop(handle);
	if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
2229 2230 2231 2232
		goto retry;
	return err;
}

2233
static int ext4_mknod(struct inode *dir, struct dentry *dentry,
A
Al Viro 已提交
2234
		      umode_t mode, dev_t rdev)
2235 2236 2237 2238 2239 2240 2241 2242
{
	handle_t *handle;
	struct inode *inode;
	int err, retries = 0;

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

2243
	dquot_initialize(dir);
2244

2245
retry:
2246 2247
	handle = ext4_journal_start(dir, EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
					EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 +
D
Dmitry Monakhov 已提交
2248
					EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb));
2249 2250 2251 2252
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	if (IS_DIRSYNC(dir))
2253
		ext4_handle_sync(handle);
2254

2255
	inode = ext4_new_inode(handle, dir, mode, &dentry->d_name, 0, NULL);
2256 2257 2258
	err = PTR_ERR(inode);
	if (!IS_ERR(inode)) {
		init_special_inode(inode, inode->i_mode, rdev);
2259 2260
		inode->i_op = &ext4_special_inode_operations;
		err = ext4_add_nondir(handle, dentry, inode);
2261
	}
2262 2263
	ext4_journal_stop(handle);
	if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
2264 2265 2266 2267
		goto retry;
	return err;
}

2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297
struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
			  struct ext4_dir_entry_2 *de,
			  int blocksize, int csum_size,
			  unsigned int parent_ino, int dotdot_real_len)
{
	de->inode = cpu_to_le32(inode->i_ino);
	de->name_len = 1;
	de->rec_len = ext4_rec_len_to_disk(EXT4_DIR_REC_LEN(de->name_len),
					   blocksize);
	strcpy(de->name, ".");
	ext4_set_de_type(inode->i_sb, de, S_IFDIR);

	de = ext4_next_entry(de, blocksize);
	de->inode = cpu_to_le32(parent_ino);
	de->name_len = 2;
	if (!dotdot_real_len)
		de->rec_len = ext4_rec_len_to_disk(blocksize -
					(csum_size + EXT4_DIR_REC_LEN(1)),
					blocksize);
	else
		de->rec_len = ext4_rec_len_to_disk(
				EXT4_DIR_REC_LEN(de->name_len), blocksize);
	strcpy(de->name, "..");
	ext4_set_de_type(inode->i_sb, de, S_IFDIR);

	return ext4_next_entry(de, blocksize);
}

static int ext4_init_new_dir(handle_t *handle, struct inode *dir,
			     struct inode *inode)
2298
{
2299
	struct buffer_head *dir_block = NULL;
2300
	struct ext4_dir_entry_2 *de;
2301
	struct ext4_dir_entry_tail *t;
2302
	unsigned int blocksize = dir->i_sb->s_blocksize;
2303
	int csum_size = 0;
2304
	int err;
2305

2306 2307 2308 2309
	if (EXT4_HAS_RO_COMPAT_FEATURE(dir->i_sb,
				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
		csum_size = sizeof(struct ext4_dir_entry_tail);

2310 2311 2312 2313 2314 2315 2316 2317
	if (ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA)) {
		err = ext4_try_create_inline_dir(handle, dir, inode);
		if (err < 0 && err != -ENOSPC)
			goto out;
		if (!err)
			goto out;
	}

2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356
	inode->i_size = EXT4_I(inode)->i_disksize = blocksize;
	dir_block = ext4_bread(handle, inode, 0, 1, &err);
	if (!(dir_block = ext4_bread(handle, inode, 0, 1, &err))) {
		if (!err) {
			err = -EIO;
			ext4_error(inode->i_sb,
				   "Directory hole detected on inode %lu\n",
				   inode->i_ino);
		}
		goto out;
	}
	BUFFER_TRACE(dir_block, "get_write_access");
	err = ext4_journal_get_write_access(handle, dir_block);
	if (err)
		goto out;
	de = (struct ext4_dir_entry_2 *)dir_block->b_data;
	ext4_init_dot_dotdot(inode, de, blocksize, csum_size, dir->i_ino, 0);
	set_nlink(inode, 2);
	if (csum_size) {
		t = EXT4_DIRENT_TAIL(dir_block->b_data, blocksize);
		initialize_dirent_tail(t, blocksize);
	}

	BUFFER_TRACE(dir_block, "call ext4_handle_dirty_metadata");
	err = ext4_handle_dirty_dirent_node(handle, inode, dir_block);
	if (err)
		goto out;
	set_buffer_verified(dir_block);
out:
	brelse(dir_block);
	return err;
}

static int ext4_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
{
	handle_t *handle;
	struct inode *inode;
	int err, retries = 0;

2357
	if (EXT4_DIR_LINK_MAX(dir))
2358 2359
		return -EMLINK;

2360
	dquot_initialize(dir);
2361

2362
retry:
2363 2364
	handle = ext4_journal_start(dir, EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
					EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 +
D
Dmitry Monakhov 已提交
2365
					EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb));
2366 2367 2368 2369
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	if (IS_DIRSYNC(dir))
2370
		ext4_handle_sync(handle);
2371

2372
	inode = ext4_new_inode(handle, dir, S_IFDIR | mode,
2373
			       &dentry->d_name, 0, NULL);
2374 2375 2376 2377
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
		goto out_stop;

2378 2379
	inode->i_op = &ext4_dir_inode_operations;
	inode->i_fop = &ext4_dir_operations;
2380
	err = ext4_init_new_dir(handle, dir, inode);
2381 2382 2383 2384 2385
	if (err)
		goto out_clear_inode;
	err = ext4_mark_inode_dirty(handle, inode);
	if (!err)
		err = ext4_add_entry(handle, dentry, inode);
2386
	if (err) {
2387 2388
out_clear_inode:
		clear_nlink(inode);
A
Al Viro 已提交
2389
		unlock_new_inode(inode);
2390
		ext4_mark_inode_dirty(handle, inode);
2391
		iput(inode);
2392 2393
		goto out_stop;
	}
2394
	ext4_inc_count(handle, dir);
2395
	ext4_update_dx_flag(dir);
2396 2397 2398
	err = ext4_mark_inode_dirty(handle, dir);
	if (err)
		goto out_clear_inode;
A
Al Viro 已提交
2399
	unlock_new_inode(inode);
2400
	d_instantiate(dentry, inode);
2401
out_stop:
2402 2403
	ext4_journal_stop(handle);
	if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
2404 2405 2406 2407 2408 2409 2410
		goto retry;
	return err;
}

/*
 * routine to check that the specified directory is empty (for rmdir)
 */
2411
static int empty_dir(struct inode *inode)
2412
{
2413
	unsigned int offset;
2414 2415 2416
	struct buffer_head *bh;
	struct ext4_dir_entry_2 *de, *de1;
	struct super_block *sb;
2417 2418 2419
	int err = 0;

	sb = inode->i_sb;
2420
	if (inode->i_size < EXT4_DIR_REC_LEN(1) + EXT4_DIR_REC_LEN(2) ||
2421
	    !(bh = ext4_bread(NULL, inode, 0, 0, &err))) {
2422
		if (err)
2423 2424
			EXT4_ERROR_INODE(inode,
				"error %d reading directory lblock 0", err);
2425
		else
2426
			ext4_warning(inode->i_sb,
2427 2428 2429 2430
				     "bad directory (dir #%lu) - no data block",
				     inode->i_ino);
		return 1;
	}
2431 2432 2433 2434 2435 2436 2437 2438
	if (!buffer_verified(bh) &&
	    !ext4_dirent_csum_verify(inode,
			(struct ext4_dir_entry *)bh->b_data)) {
		EXT4_ERROR_INODE(inode, "checksum error reading directory "
				 "lblock 0");
		return -EIO;
	}
	set_buffer_verified(bh);
2439
	de = (struct ext4_dir_entry_2 *) bh->b_data;
2440
	de1 = ext4_next_entry(de, sb->s_blocksize);
2441 2442
	if (le32_to_cpu(de->inode) != inode->i_ino ||
			!le32_to_cpu(de1->inode) ||
2443 2444
			strcmp(".", de->name) ||
			strcmp("..", de1->name)) {
2445
		ext4_warning(inode->i_sb,
2446 2447 2448
			     "bad directory (dir #%lu) - no `.' or `..'",
			     inode->i_ino);
		brelse(bh);
2449 2450
		return 1;
	}
2451 2452 2453
	offset = ext4_rec_len_from_disk(de->rec_len, sb->s_blocksize) +
		 ext4_rec_len_from_disk(de1->rec_len, sb->s_blocksize);
	de = ext4_next_entry(de1, sb->s_blocksize);
2454
	while (offset < inode->i_size) {
2455
		if (!bh ||
2456 2457
		    (void *) de >= (void *) (bh->b_data+sb->s_blocksize)) {
			unsigned int lblock;
2458
			err = 0;
2459
			brelse(bh);
2460 2461
			lblock = offset >> EXT4_BLOCK_SIZE_BITS(sb);
			bh = ext4_bread(NULL, inode, lblock, 0, &err);
2462 2463
			if (!bh) {
				if (err)
2464 2465 2466
					EXT4_ERROR_INODE(inode,
						"error %d reading directory "
						"lblock %u", err, lblock);
C
Carlos Maiolino 已提交
2467 2468 2469 2470 2471
				else
					ext4_warning(inode->i_sb,
						"bad directory (dir #%lu) - no data block",
						inode->i_ino);

2472 2473 2474
				offset += sb->s_blocksize;
				continue;
			}
2475 2476 2477 2478 2479 2480 2481 2482
			if (!buffer_verified(bh) &&
			    !ext4_dirent_csum_verify(inode,
					(struct ext4_dir_entry *)bh->b_data)) {
				EXT4_ERROR_INODE(inode, "checksum error "
						 "reading directory lblock 0");
				return -EIO;
			}
			set_buffer_verified(bh);
2483
			de = (struct ext4_dir_entry_2 *) bh->b_data;
2484
		}
2485 2486
		if (ext4_check_dir_entry(inode, NULL, de, bh,
					 bh->b_data, bh->b_size, offset)) {
2487
			de = (struct ext4_dir_entry_2 *)(bh->b_data +
2488 2489 2490 2491 2492
							 sb->s_blocksize);
			offset = (offset | (sb->s_blocksize - 1)) + 1;
			continue;
		}
		if (le32_to_cpu(de->inode)) {
2493
			brelse(bh);
2494 2495
			return 0;
		}
2496 2497
		offset += ext4_rec_len_from_disk(de->rec_len, sb->s_blocksize);
		de = ext4_next_entry(de, sb->s_blocksize);
2498
	}
2499
	brelse(bh);
2500 2501 2502
	return 1;
}

2503
/* ext4_orphan_add() links an unlinked or truncated inode into a list of
2504 2505 2506 2507 2508
 * 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
2509
 * inodes and truncating linked inodes in ext4_orphan_cleanup().
2510
 */
2511
int ext4_orphan_add(handle_t *handle, struct inode *inode)
2512 2513
{
	struct super_block *sb = inode->i_sb;
2514
	struct ext4_iloc iloc;
2515 2516
	int err = 0, rc;

2517
	if (!EXT4_SB(sb)->s_journal)
2518 2519
		return 0;

2520
	mutex_lock(&EXT4_SB(sb)->s_orphan_lock);
2521
	if (!list_empty(&EXT4_I(inode)->i_orphan))
2522 2523
		goto out_unlock;

2524 2525 2526 2527 2528
	/*
	 * Orphan handling is only valid for files with data blocks
	 * being truncated, or files being unlinked. Note that we either
	 * hold i_mutex, or the inode can not be referenced from outside,
	 * so i_nlink should not be bumped due to race
2529
	 */
2530 2531
	J_ASSERT((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
		  S_ISLNK(inode->i_mode)) || inode->i_nlink == 0);
2532

2533 2534
	BUFFER_TRACE(EXT4_SB(sb)->s_sbh, "get_write_access");
	err = ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh);
2535 2536 2537
	if (err)
		goto out_unlock;

2538
	err = ext4_reserve_inode_write(handle, inode, &iloc);
2539 2540
	if (err)
		goto out_unlock;
2541 2542 2543 2544 2545 2546 2547
	/*
	 * Due to previous errors inode may be already a part of on-disk
	 * orphan list. If so skip on-disk list modification.
	 */
	if (NEXT_ORPHAN(inode) && NEXT_ORPHAN(inode) <=
		(le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)))
			goto mem_insert;
2548 2549

	/* Insert this inode at the head of the on-disk orphan list... */
2550 2551
	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);
2552
	err = ext4_handle_dirty_super(handle, sb);
2553
	rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564
	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. */
2565
mem_insert:
2566
	if (!err)
2567
		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2568 2569 2570 2571 2572

	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:
2573
	mutex_unlock(&EXT4_SB(sb)->s_orphan_lock);
2574
	ext4_std_error(inode->i_sb, err);
2575 2576 2577 2578
	return err;
}

/*
2579
 * ext4_orphan_del() removes an unlinked or truncated inode from the list
2580 2581
 * of such inodes stored on disk, because it is finally being cleaned up.
 */
2582
int ext4_orphan_del(handle_t *handle, struct inode *inode)
2583 2584
{
	struct list_head *prev;
2585 2586
	struct ext4_inode_info *ei = EXT4_I(inode);
	struct ext4_sb_info *sbi;
2587
	__u32 ino_next;
2588
	struct ext4_iloc iloc;
2589 2590
	int err = 0;

2591
	if (!EXT4_SB(inode->i_sb)->s_journal)
2592 2593
		return 0;

2594 2595 2596
	mutex_lock(&EXT4_SB(inode->i_sb)->s_orphan_lock);
	if (list_empty(&ei->i_orphan))
		goto out;
2597 2598 2599

	ino_next = NEXT_ORPHAN(inode);
	prev = ei->i_orphan.prev;
2600
	sbi = EXT4_SB(inode->i_sb);
2601 2602 2603 2604 2605 2606 2607 2608 2609

	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. */
2610
	if (!handle)
2611 2612
		goto out;

2613
	err = ext4_reserve_inode_write(handle, inode, &iloc);
2614 2615 2616 2617
	if (err)
		goto out_err;

	if (prev == &sbi->s_orphan) {
2618
		jbd_debug(4, "superblock will point to %u\n", ino_next);
2619
		BUFFER_TRACE(sbi->s_sbh, "get_write_access");
2620
		err = ext4_journal_get_write_access(handle, sbi->s_sbh);
2621 2622 2623
		if (err)
			goto out_brelse;
		sbi->s_es->s_last_orphan = cpu_to_le32(ino_next);
2624
		err = ext4_handle_dirty_super(handle, inode->i_sb);
2625
	} else {
2626
		struct ext4_iloc iloc2;
2627
		struct inode *i_prev =
2628
			&list_entry(prev, struct ext4_inode_info, i_orphan)->vfs_inode;
2629

2630
		jbd_debug(4, "orphan inode %lu will point to %u\n",
2631
			  i_prev->i_ino, ino_next);
2632
		err = ext4_reserve_inode_write(handle, i_prev, &iloc2);
2633 2634 2635
		if (err)
			goto out_brelse;
		NEXT_ORPHAN(i_prev) = ino_next;
2636
		err = ext4_mark_iloc_dirty(handle, i_prev, &iloc2);
2637 2638 2639 2640
	}
	if (err)
		goto out_brelse;
	NEXT_ORPHAN(inode) = 0;
2641
	err = ext4_mark_iloc_dirty(handle, inode, &iloc);
2642 2643

out_err:
2644
	ext4_std_error(inode->i_sb, err);
2645
out:
2646
	mutex_unlock(&EXT4_SB(inode->i_sb)->s_orphan_lock);
2647 2648 2649 2650 2651 2652 2653
	return err;

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

2654
static int ext4_rmdir(struct inode *dir, struct dentry *dentry)
2655 2656
{
	int retval;
2657 2658 2659
	struct inode *inode;
	struct buffer_head *bh;
	struct ext4_dir_entry_2 *de;
2660 2661 2662 2663
	handle_t *handle;

	/* Initialize quotas before so that eventual writes go in
	 * separate transaction */
2664 2665
	dquot_initialize(dir);
	dquot_initialize(dentry->d_inode);
2666

2667
	handle = ext4_journal_start(dir, EXT4_DELETE_TRANS_BLOCKS(dir->i_sb));
2668 2669 2670 2671
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	retval = -ENOENT;
2672
	bh = ext4_find_entry(dir, &dentry->d_name, &de);
2673 2674 2675 2676
	if (!bh)
		goto end_rmdir;

	if (IS_DIRSYNC(dir))
2677
		ext4_handle_sync(handle);
2678 2679 2680 2681 2682 2683 2684 2685

	inode = dentry->d_inode;

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

	retval = -ENOTEMPTY;
2686
	if (!empty_dir(inode))
2687 2688
		goto end_rmdir;

2689
	retval = ext4_delete_entry(handle, dir, de, bh);
2690 2691
	if (retval)
		goto end_rmdir;
2692
	if (!EXT4_DIR_LINK_EMPTY(inode))
2693
		ext4_warning(inode->i_sb,
2694 2695
			     "empty directory has too many links (%d)",
			     inode->i_nlink);
2696 2697 2698 2699 2700 2701
	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;
2702
	ext4_orphan_add(handle, inode);
K
Kalpak Shah 已提交
2703
	inode->i_ctime = dir->i_ctime = dir->i_mtime = ext4_current_time(inode);
2704
	ext4_mark_inode_dirty(handle, inode);
2705
	ext4_dec_count(handle, dir);
2706 2707
	ext4_update_dx_flag(dir);
	ext4_mark_inode_dirty(handle, dir);
2708 2709

end_rmdir:
2710
	ext4_journal_stop(handle);
2711
	brelse(bh);
2712 2713 2714
	return retval;
}

2715
static int ext4_unlink(struct inode *dir, struct dentry *dentry)
2716 2717
{
	int retval;
2718 2719 2720
	struct inode *inode;
	struct buffer_head *bh;
	struct ext4_dir_entry_2 *de;
2721 2722
	handle_t *handle;

2723
	trace_ext4_unlink_enter(dir, dentry);
2724 2725
	/* Initialize quotas before so that eventual writes go
	 * in separate transaction */
2726 2727
	dquot_initialize(dir);
	dquot_initialize(dentry->d_inode);
2728

2729
	handle = ext4_journal_start(dir, EXT4_DELETE_TRANS_BLOCKS(dir->i_sb));
2730 2731 2732 2733
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	if (IS_DIRSYNC(dir))
2734
		ext4_handle_sync(handle);
2735 2736

	retval = -ENOENT;
2737
	bh = ext4_find_entry(dir, &dentry->d_name, &de);
2738 2739 2740 2741 2742 2743 2744 2745 2746 2747
	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) {
2748
		ext4_warning(inode->i_sb,
2749 2750
			     "Deleting nonexistent file (%lu), %d",
			     inode->i_ino, inode->i_nlink);
M
Miklos Szeredi 已提交
2751
		set_nlink(inode, 1);
2752
	}
2753
	retval = ext4_delete_entry(handle, dir, de, bh);
2754 2755
	if (retval)
		goto end_unlink;
K
Kalpak Shah 已提交
2756
	dir->i_ctime = dir->i_mtime = ext4_current_time(dir);
2757 2758
	ext4_update_dx_flag(dir);
	ext4_mark_inode_dirty(handle, dir);
2759
	drop_nlink(inode);
2760
	if (!inode->i_nlink)
2761
		ext4_orphan_add(handle, inode);
K
Kalpak Shah 已提交
2762
	inode->i_ctime = ext4_current_time(inode);
2763
	ext4_mark_inode_dirty(handle, inode);
2764 2765 2766
	retval = 0;

end_unlink:
2767
	ext4_journal_stop(handle);
2768
	brelse(bh);
2769
	trace_ext4_unlink_exit(dentry, retval);
2770 2771 2772
	return retval;
}

2773 2774
static int ext4_symlink(struct inode *dir,
			struct dentry *dentry, const char *symname)
2775 2776
{
	handle_t *handle;
2777
	struct inode *inode;
2778
	int l, err, retries = 0;
2779
	int credits;
2780 2781 2782 2783 2784

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

2785
	dquot_initialize(dir);
2786

2787 2788 2789 2790
	if (l > EXT4_N_BLOCKS * 4) {
		/*
		 * For non-fast symlinks, we just allocate inode and put it on
		 * orphan list in the first transaction => we need bitmap,
2791 2792
		 * group descriptor, sb, inode block, quota blocks, and
		 * possibly selinux xattr blocks.
2793
		 */
2794 2795
		credits = 4 + EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb) +
			  EXT4_XATTR_TRANS_BLOCKS;
2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806
	} else {
		/*
		 * Fast symlink. We have to add entry to directory
		 * (EXT4_DATA_TRANS_BLOCKS + EXT4_INDEX_EXTRA_TRANS_BLOCKS),
		 * allocate new inode (bitmap, group descriptor, inode block,
		 * quota blocks, sb is already counted in previous macros).
		 */
		credits = EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
			  EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 +
			  EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb);
	}
2807
retry:
2808
	handle = ext4_journal_start(dir, credits);
2809 2810 2811 2812
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	if (IS_DIRSYNC(dir))
2813
		ext4_handle_sync(handle);
2814

2815
	inode = ext4_new_inode(handle, dir, S_IFLNK|S_IRWXUGO,
2816
			       &dentry->d_name, 0, NULL);
2817 2818 2819 2820
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
		goto out_stop;

2821
	if (l > EXT4_N_BLOCKS * 4) {
2822 2823
		inode->i_op = &ext4_symlink_inode_operations;
		ext4_set_aops(inode);
2824
		/*
2825 2826 2827 2828 2829 2830 2831 2832
		 * We cannot call page_symlink() with transaction started
		 * because it calls into ext4_write_begin() which can wait
		 * for transaction commit if we are running out of space
		 * and thus we deadlock. So we have to stop transaction now
		 * and restart it when symlink contents is written.
		 * 
		 * To keep fs consistent in case of crash, we have to put inode
		 * to orphan list in the mean time.
2833
		 */
2834 2835 2836 2837 2838
		drop_nlink(inode);
		err = ext4_orphan_add(handle, inode);
		ext4_journal_stop(handle);
		if (err)
			goto err_drop_inode;
2839
		err = __page_symlink(inode, symname, l, 1);
2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852
		if (err)
			goto err_drop_inode;
		/*
		 * Now inode is being linked into dir (EXT4_DATA_TRANS_BLOCKS
		 * + EXT4_INDEX_EXTRA_TRANS_BLOCKS), inode is also modified
		 */
		handle = ext4_journal_start(dir,
				EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
				EXT4_INDEX_EXTRA_TRANS_BLOCKS + 1);
		if (IS_ERR(handle)) {
			err = PTR_ERR(handle);
			goto err_drop_inode;
		}
2853
		set_nlink(inode, 1);
2854
		err = ext4_orphan_del(handle, inode);
2855
		if (err) {
2856
			ext4_journal_stop(handle);
2857
			clear_nlink(inode);
2858
			goto err_drop_inode;
2859 2860
		}
	} else {
2861
		/* clear the extent format for fast symlink */
2862
		ext4_clear_inode_flag(inode, EXT4_INODE_EXTENTS);
2863
		inode->i_op = &ext4_fast_symlink_inode_operations;
2864
		memcpy((char *)&EXT4_I(inode)->i_data, symname, l);
2865 2866
		inode->i_size = l-1;
	}
2867 2868
	EXT4_I(inode)->i_disksize = inode->i_size;
	err = ext4_add_nondir(handle, dentry, inode);
2869
out_stop:
2870 2871
	ext4_journal_stop(handle);
	if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
2872 2873
		goto retry;
	return err;
2874 2875 2876 2877
err_drop_inode:
	unlock_new_inode(inode);
	iput(inode);
	return err;
2878 2879
}

2880 2881
static int ext4_link(struct dentry *old_dentry,
		     struct inode *dir, struct dentry *dentry)
2882 2883 2884 2885 2886
{
	handle_t *handle;
	struct inode *inode = old_dentry->d_inode;
	int err, retries = 0;

2887
	if (inode->i_nlink >= EXT4_LINK_MAX)
2888
		return -EMLINK;
2889

2890
	dquot_initialize(dir);
2891

2892
retry:
2893 2894
	handle = ext4_journal_start(dir, EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
					EXT4_INDEX_EXTRA_TRANS_BLOCKS);
2895 2896 2897 2898
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	if (IS_DIRSYNC(dir))
2899
		ext4_handle_sync(handle);
2900

K
Kalpak Shah 已提交
2901
	inode->i_ctime = ext4_current_time(inode);
2902
	ext4_inc_count(handle, inode);
A
Al Viro 已提交
2903
	ihold(inode);
2904

A
Al Viro 已提交
2905 2906 2907 2908 2909 2910 2911 2912
	err = ext4_add_entry(handle, dentry, inode);
	if (!err) {
		ext4_mark_inode_dirty(handle, inode);
		d_instantiate(dentry, inode);
	} else {
		drop_nlink(inode);
		iput(inode);
	}
2913 2914
	ext4_journal_stop(handle);
	if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
2915 2916 2917 2918
		goto retry;
	return err;
}

2919 2920
#define PARENT_INO(buffer, size) \
	(ext4_next_entry((struct ext4_dir_entry_2 *)(buffer), size)->inode)
2921 2922 2923 2924 2925

/*
 * Anybody can rename anything with this: the permission checks are left to the
 * higher-level routines.
 */
2926 2927
static int ext4_rename(struct inode *old_dir, struct dentry *old_dentry,
		       struct inode *new_dir, struct dentry *new_dentry)
2928 2929
{
	handle_t *handle;
2930 2931 2932
	struct inode *old_inode, *new_inode;
	struct buffer_head *old_bh, *new_bh, *dir_bh;
	struct ext4_dir_entry_2 *old_de, *new_de;
2933
	int retval, force_da_alloc = 0;
2934

2935 2936
	dquot_initialize(old_dir);
	dquot_initialize(new_dir);
2937

2938 2939 2940 2941 2942
	old_bh = new_bh = dir_bh = NULL;

	/* Initialize quotas before so that eventual writes go
	 * in separate transaction */
	if (new_dentry->d_inode)
2943
		dquot_initialize(new_dentry->d_inode);
2944 2945 2946
	handle = ext4_journal_start(old_dir, 2 *
					EXT4_DATA_TRANS_BLOCKS(old_dir->i_sb) +
					EXT4_INDEX_EXTRA_TRANS_BLOCKS + 2);
2947 2948 2949 2950
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	if (IS_DIRSYNC(old_dir) || IS_DIRSYNC(new_dir))
2951
		ext4_handle_sync(handle);
2952

2953
	old_bh = ext4_find_entry(old_dir, &old_dentry->d_name, &old_de);
2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965
	/*
	 *  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;
2966
	new_bh = ext4_find_entry(new_dir, &new_dentry->d_name, &new_de);
2967 2968
	if (new_bh) {
		if (!new_inode) {
2969
			brelse(new_bh);
2970 2971 2972 2973 2974 2975
			new_bh = NULL;
		}
	}
	if (S_ISDIR(old_inode->i_mode)) {
		if (new_inode) {
			retval = -ENOTEMPTY;
2976
			if (!empty_dir(new_inode))
2977 2978 2979
				goto end_rename;
		}
		retval = -EIO;
C
Carlos Maiolino 已提交
2980 2981 2982 2983 2984 2985 2986
		if (!(dir_bh = ext4_bread(handle, old_inode, 0, 0, &retval))) {
			if (!retval) {
				retval = -EIO;
				ext4_error(old_inode->i_sb,
					   "Directory hole detected on inode %lu\n",
					   old_inode->i_ino);
			}
2987
			goto end_rename;
C
Carlos Maiolino 已提交
2988
		}
2989 2990 2991 2992 2993
		if (!buffer_verified(dir_bh) &&
		    !ext4_dirent_csum_verify(old_inode,
				(struct ext4_dir_entry *)dir_bh->b_data))
			goto end_rename;
		set_buffer_verified(dir_bh);
2994 2995
		if (le32_to_cpu(PARENT_INO(dir_bh->b_data,
				old_dir->i_sb->s_blocksize)) != old_dir->i_ino)
2996 2997
			goto end_rename;
		retval = -EMLINK;
2998
		if (!new_inode && new_dir != old_dir &&
2999
		    EXT4_DIR_LINK_MAX(new_dir))
3000
			goto end_rename;
3001 3002 3003 3004
		BUFFER_TRACE(dir_bh, "get_write_access");
		retval = ext4_journal_get_write_access(handle, dir_bh);
		if (retval)
			goto end_rename;
3005 3006
	}
	if (!new_bh) {
3007
		retval = ext4_add_entry(handle, new_dentry, old_inode);
3008 3009 3010 3011
		if (retval)
			goto end_rename;
	} else {
		BUFFER_TRACE(new_bh, "get write access");
3012 3013 3014
		retval = ext4_journal_get_write_access(handle, new_bh);
		if (retval)
			goto end_rename;
3015
		new_de->inode = cpu_to_le32(old_inode->i_ino);
3016 3017
		if (EXT4_HAS_INCOMPAT_FEATURE(new_dir->i_sb,
					      EXT4_FEATURE_INCOMPAT_FILETYPE))
3018 3019
			new_de->file_type = old_de->file_type;
		new_dir->i_version++;
3020 3021 3022
		new_dir->i_ctime = new_dir->i_mtime =
					ext4_current_time(new_dir);
		ext4_mark_inode_dirty(handle, new_dir);
3023
		BUFFER_TRACE(new_bh, "call ext4_handle_dirty_metadata");
3024
		retval = ext4_handle_dirty_dirent_node(handle, new_dir, new_bh);
3025 3026 3027 3028
		if (unlikely(retval)) {
			ext4_std_error(new_dir->i_sb, retval);
			goto end_rename;
		}
3029 3030 3031 3032 3033 3034 3035 3036
		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
3037
	old_inode->i_ctime = ext4_current_time(old_inode);
3038
	ext4_mark_inode_dirty(handle, old_inode);
3039 3040 3041 3042 3043 3044 3045

	/*
	 * 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) ||
3046
	    (retval = ext4_delete_entry(handle, old_dir,
3047 3048 3049 3050 3051 3052
					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;
3053
		struct ext4_dir_entry_2 *old_de2;
3054

3055
		old_bh2 = ext4_find_entry(old_dir, &old_dentry->d_name, &old_de2);
3056
		if (old_bh2) {
3057
			retval = ext4_delete_entry(handle, old_dir,
3058 3059 3060 3061 3062
						   old_de2, old_bh2);
			brelse(old_bh2);
		}
	}
	if (retval) {
3063
		ext4_warning(old_dir->i_sb,
3064 3065 3066 3067 3068
				"Deleting old file (%lu), %d, error=%d",
				old_dir->i_ino, old_dir->i_nlink, retval);
	}

	if (new_inode) {
3069
		ext4_dec_count(handle, new_inode);
K
Kalpak Shah 已提交
3070
		new_inode->i_ctime = ext4_current_time(new_inode);
3071
	}
K
Kalpak Shah 已提交
3072
	old_dir->i_ctime = old_dir->i_mtime = ext4_current_time(old_dir);
3073
	ext4_update_dx_flag(old_dir);
3074
	if (dir_bh) {
3075 3076
		PARENT_INO(dir_bh->b_data, new_dir->i_sb->s_blocksize) =
						cpu_to_le32(new_dir->i_ino);
3077
		BUFFER_TRACE(dir_bh, "call ext4_handle_dirty_metadata");
3078 3079 3080 3081 3082 3083 3084 3085 3086
		if (is_dx(old_inode)) {
			retval = ext4_handle_dirty_dx_node(handle,
							   old_inode,
							   dir_bh);
		} else {
			retval = ext4_handle_dirty_dirent_node(handle,
							       old_inode,
							       dir_bh);
		}
3087 3088 3089 3090
		if (retval) {
			ext4_std_error(old_dir->i_sb, retval);
			goto end_rename;
		}
3091
		ext4_dec_count(handle, old_dir);
3092
		if (new_inode) {
3093
			/* checked empty_dir above, can't have another parent,
3094
			 * ext4_dec_count() won't work for many-linked dirs */
3095
			clear_nlink(new_inode);
3096
		} else {
3097
			ext4_inc_count(handle, new_dir);
3098 3099
			ext4_update_dx_flag(new_dir);
			ext4_mark_inode_dirty(handle, new_dir);
3100 3101
		}
	}
3102
	ext4_mark_inode_dirty(handle, old_dir);
3103
	if (new_inode) {
3104
		ext4_mark_inode_dirty(handle, new_inode);
3105
		if (!new_inode->i_nlink)
3106
			ext4_orphan_add(handle, new_inode);
3107 3108
		if (!test_opt(new_dir->i_sb, NO_AUTO_DA_ALLOC))
			force_da_alloc = 1;
3109 3110 3111 3112
	}
	retval = 0;

end_rename:
3113 3114 3115
	brelse(dir_bh);
	brelse(old_bh);
	brelse(new_bh);
3116
	ext4_journal_stop(handle);
3117 3118
	if (retval == 0 && force_da_alloc)
		ext4_alloc_da_blocks(old_inode);
3119 3120 3121 3122 3123 3124
	return retval;
}

/*
 * directories can handle most operations...
 */
3125
const struct inode_operations ext4_dir_inode_operations = {
3126 3127 3128 3129 3130 3131 3132 3133 3134 3135
	.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
3136
#ifdef CONFIG_EXT4_FS_XATTR
3137 3138
	.setxattr	= generic_setxattr,
	.getxattr	= generic_getxattr,
3139
	.listxattr	= ext4_listxattr,
3140 3141
	.removexattr	= generic_removexattr,
#endif
3142
	.get_acl	= ext4_get_acl,
3143
	.fiemap         = ext4_fiemap,
3144 3145
};

3146
const struct inode_operations ext4_special_inode_operations = {
3147
	.setattr	= ext4_setattr,
T
Theodore Ts'o 已提交
3148
#ifdef CONFIG_EXT4_FS_XATTR
3149 3150
	.setxattr	= generic_setxattr,
	.getxattr	= generic_getxattr,
3151
	.listxattr	= ext4_listxattr,
3152 3153
	.removexattr	= generic_removexattr,
#endif
3154
	.get_acl	= ext4_get_acl,
3155
};