dir.c 15 KB
Newer Older
1
/*
2
 *  linux/fs/ext4/dir.c
3 4 5 6 7 8 9 10 11 12 13 14
 *
 * 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/dir.c
 *
 *  Copyright (C) 1991, 1992  Linus Torvalds
 *
15
 *  ext4 directory handling functions
16 17 18 19 20 21 22 23 24
 *
 *  Big-endian to little-endian byte-swapping/bitmaps by
 *        David S. Miller (davem@caip.rutgers.edu), 1995
 *
 * Hash Tree Directory indexing (c) 2001  Daniel Phillips
 *
 */

#include <linux/fs.h>
25
#include <linux/jbd2.h>
26 27 28
#include <linux/buffer_head.h>
#include <linux/slab.h>
#include <linux/rbtree.h>
29
#include "ext4.h"
30
#include "xattr.h"
31

Al Viro's avatar
Al Viro committed
32
static int ext4_dx_readdir(struct file *, struct dir_context *);
33

34 35
/**
 * Check if the given dir-inode refers to an htree-indexed directory
36
 * (or a directory which could potentially get converted to use htree
37 38 39 40 41 42 43 44 45 46 47
 * indexing).
 *
 * Return 1 if it is a dx dir, 0 if not
 */
static int is_dx_dir(struct inode *inode)
{
	struct super_block *sb = inode->i_sb;

	if (EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
		     EXT4_FEATURE_COMPAT_DIR_INDEX) &&
	    ((ext4_test_inode_flag(inode, EXT4_INODE_INDEX)) ||
48 49
	     ((inode->i_size >> sb->s_blocksize_bits) == 1) ||
	     ext4_has_inline_data(inode)))
50 51 52 53 54
		return 1;

	return 0;
}

55 56 57 58
/*
 * Return 0 if the directory entry is OK, and 1 if there is a problem
 *
 * Note: this is the opposite of what ext2 and ext3 historically returned...
59 60 61
 *
 * bh passed here can be an inode block or a dir data block, depending
 * on the inode inline data flag.
62
 */
63
int __ext4_check_dir_entry(const char *function, unsigned int line,
64
			   struct inode *dir, struct file *filp,
65
			   struct ext4_dir_entry_2 *de,
66
			   struct buffer_head *bh, char *buf, int size,
67
			   unsigned int offset)
68
{
69
	const char *error_msg = NULL;
70 71
	const int rlen = ext4_rec_len_from_disk(de->rec_len,
						dir->i_sb->s_blocksize);
72

73
	if (unlikely(rlen < EXT4_DIR_REC_LEN(1)))
74
		error_msg = "rec_len is smaller than minimal";
75
	else if (unlikely(rlen % 4 != 0))
76
		error_msg = "rec_len % 4 != 0";
77
	else if (unlikely(rlen < EXT4_DIR_REC_LEN(de->name_len)))
78
		error_msg = "rec_len is too small for name_len";
79 80
	else if (unlikely(((char *) de - buf) + rlen > size))
		error_msg = "directory entry across range";
81 82
	else if (unlikely(le32_to_cpu(de->inode) >
			le32_to_cpu(EXT4_SB(dir->i_sb)->s_es->s_inodes_count)))
83
		error_msg = "inode out of bounds";
84 85 86
	else
		return 0;

87
	if (filp)
88
		ext4_error_file(filp, function, line, bh->b_blocknr,
89 90
				"bad entry in directory: %s - offset=%u(%u), "
				"inode=%u, rec_len=%d, name_len=%d",
91
				error_msg, (unsigned) (offset % size),
92 93 94
				offset, le32_to_cpu(de->inode),
				rlen, de->name_len);
	else
95
		ext4_error_inode(dir, function, line, bh->b_blocknr,
96 97
				"bad entry in directory: %s - offset=%u(%u), "
				"inode=%u, rec_len=%d, name_len=%d",
98
				error_msg, (unsigned) (offset % size),
99 100 101
				offset, le32_to_cpu(de->inode),
				rlen, de->name_len);

102
	return 1;
103 104
}

Al Viro's avatar
Al Viro committed
105
static int ext4_readdir(struct file *file, struct dir_context *ctx)
106
{
107
	unsigned int offset;
108
	int i;
109
	struct ext4_dir_entry_2 *de;
110
	int err;
Al Viro's avatar
Al Viro committed
111
	struct inode *inode = file_inode(file);
112
	struct super_block *sb = inode->i_sb;
113
	int dir_has_error = 0;
114

115
	if (is_dx_dir(inode)) {
Al Viro's avatar
Al Viro committed
116
		err = ext4_dx_readdir(file, ctx);
117
		if (err != ERR_BAD_DX_DIR) {
Al Viro's avatar
Al Viro committed
118
			return err;
119 120 121 122 123
		}
		/*
		 * We don't set the inode dirty flag since it's not
		 * critical that it get flushed back to the disk.
		 */
Al Viro's avatar
Al Viro committed
124
		ext4_clear_inode_flag(file_inode(file),
125
				      EXT4_INODE_INDEX);
126
	}
127 128 129

	if (ext4_has_inline_data(inode)) {
		int has_inline_data = 1;
Al Viro's avatar
Al Viro committed
130
		int ret = ext4_read_inline_dir(file, ctx,
131 132 133 134 135
					   &has_inline_data);
		if (has_inline_data)
			return ret;
	}

Al Viro's avatar
Al Viro committed
136
	offset = ctx->pos & (sb->s_blocksize - 1);
137

Al Viro's avatar
Al Viro committed
138
	while (ctx->pos < inode->i_size) {
139
		struct ext4_map_blocks map;
140 141
		struct buffer_head *bh = NULL;

Al Viro's avatar
Al Viro committed
142
		map.m_lblk = ctx->pos >> EXT4_BLOCK_SIZE_BITS(sb);
143 144
		map.m_len = 1;
		err = ext4_map_blocks(NULL, inode, &map, 0);
145
		if (err > 0) {
146
			pgoff_t index = map.m_pblk >>
147
					(PAGE_CACHE_SHIFT - inode->i_blkbits);
Al Viro's avatar
Al Viro committed
148
			if (!ra_has_index(&file->f_ra, index))
149
				page_cache_sync_readahead(
150
					sb->s_bdev->bd_inode->i_mapping,
Al Viro's avatar
Al Viro committed
151
					&file->f_ra, file,
152
					index, 1);
Al Viro's avatar
Al Viro committed
153
			file->f_ra.prev_pos = (loff_t)index << PAGE_CACHE_SHIFT;
154
			bh = ext4_bread(NULL, inode, map.m_lblk, 0, &err);
155 156 157 158 159 160 161
		}

		/*
		 * We ignore I/O errors on directories so users have a chance
		 * of recovering data when there's a bad sector
		 */
		if (!bh) {
162
			if (!dir_has_error) {
Al Viro's avatar
Al Viro committed
163
				EXT4_ERROR_FILE(file, 0,
164 165
						"directory contains a "
						"hole at offset %llu",
Al Viro's avatar
Al Viro committed
166
					   (unsigned long long) ctx->pos);
167 168
				dir_has_error = 1;
			}
169
			/* corrupt size?  Maybe no more blocks to read */
Al Viro's avatar
Al Viro committed
170
			if (ctx->pos > inode->i_blocks << 9)
171
				break;
Al Viro's avatar
Al Viro committed
172
			ctx->pos += sb->s_blocksize - offset;
173 174 175
			continue;
		}

176 177 178 179
		/* Check the checksum */
		if (!buffer_verified(bh) &&
		    !ext4_dirent_csum_verify(inode,
				(struct ext4_dir_entry *)bh->b_data)) {
Al Viro's avatar
Al Viro committed
180
			EXT4_ERROR_FILE(file, 0, "directory fails checksum "
181
					"at offset %llu",
Al Viro's avatar
Al Viro committed
182 183
					(unsigned long long)ctx->pos);
			ctx->pos += sb->s_blocksize - offset;
184
			brelse(bh);
185 186 187 188
			continue;
		}
		set_buffer_verified(bh);

189 190 191 192
		/* If the dir block has changed since the last call to
		 * readdir(2), then we might be pointing to an invalid
		 * dirent right now.  Scan from the start of the block
		 * to make sure. */
Al Viro's avatar
Al Viro committed
193
		if (file->f_version != inode->i_version) {
194
			for (i = 0; i < sb->s_blocksize && i < offset; ) {
195
				de = (struct ext4_dir_entry_2 *)
196 197 198 199 200 201 202
					(bh->b_data + i);
				/* It's too expensive to do a full
				 * dirent test each time round this
				 * loop, but we do have to test at
				 * least that it is non-zero.  A
				 * failure will be detected in the
				 * dirent test below. */
203 204
				if (ext4_rec_len_from_disk(de->rec_len,
					sb->s_blocksize) < EXT4_DIR_REC_LEN(1))
205
					break;
206 207
				i += ext4_rec_len_from_disk(de->rec_len,
							    sb->s_blocksize);
208 209
			}
			offset = i;
Al Viro's avatar
Al Viro committed
210
			ctx->pos = (ctx->pos & ~(sb->s_blocksize - 1))
211
				| offset;
Al Viro's avatar
Al Viro committed
212
			file->f_version = inode->i_version;
213 214
		}

Al Viro's avatar
Al Viro committed
215
		while (ctx->pos < inode->i_size
216
		       && offset < sb->s_blocksize) {
217
			de = (struct ext4_dir_entry_2 *) (bh->b_data + offset);
Al Viro's avatar
Al Viro committed
218
			if (ext4_check_dir_entry(inode, file, de, bh,
219 220
						 bh->b_data, bh->b_size,
						 offset)) {
221
				/*
Al Viro's avatar
Al Viro committed
222
				 * On error, skip to the next block
223
				 */
Al Viro's avatar
Al Viro committed
224
				ctx->pos = (ctx->pos |
225
						(sb->s_blocksize - 1)) + 1;
Al Viro's avatar
Al Viro committed
226
				break;
227
			}
228 229
			offset += ext4_rec_len_from_disk(de->rec_len,
					sb->s_blocksize);
230
			if (le32_to_cpu(de->inode)) {
Al Viro's avatar
Al Viro committed
231
				if (!dir_emit(ctx, de->name,
232 233
						de->name_len,
						le32_to_cpu(de->inode),
Al Viro's avatar
Al Viro committed
234 235 236 237
						get_dtype(sb, de->file_type))) {
					brelse(bh);
					return 0;
				}
238
			}
Al Viro's avatar
Al Viro committed
239
			ctx->pos += ext4_rec_len_from_disk(de->rec_len,
240
						sb->s_blocksize);
241 242
		}
		offset = 0;
243
		brelse(bh);
Al Viro's avatar
Al Viro committed
244 245 246 247
		if (ctx->pos < inode->i_size) {
			if (!dir_relax(inode))
				return 0;
		}
248
	}
Al Viro's avatar
Al Viro committed
249
	return 0;
250 251
}

252 253 254 255 256 257 258 259 260
static inline int is_32bit_api(void)
{
#ifdef CONFIG_COMPAT
	return is_compat_task();
#else
	return (BITS_PER_LONG == 32);
#endif
}

261 262
/*
 * These functions convert from the major/minor hash to an f_pos
263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310
 * value for dx directories
 *
 * Upper layer (for example NFS) should specify FMODE_32BITHASH or
 * FMODE_64BITHASH explicitly. On the other hand, we allow ext4 to be mounted
 * directly on both 32-bit and 64-bit nodes, under such case, neither
 * FMODE_32BITHASH nor FMODE_64BITHASH is specified.
 */
static inline loff_t hash2pos(struct file *filp, __u32 major, __u32 minor)
{
	if ((filp->f_mode & FMODE_32BITHASH) ||
	    (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
		return major >> 1;
	else
		return ((__u64)(major >> 1) << 32) | (__u64)minor;
}

static inline __u32 pos2maj_hash(struct file *filp, loff_t pos)
{
	if ((filp->f_mode & FMODE_32BITHASH) ||
	    (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
		return (pos << 1) & 0xffffffff;
	else
		return ((pos >> 32) << 1) & 0xffffffff;
}

static inline __u32 pos2min_hash(struct file *filp, loff_t pos)
{
	if ((filp->f_mode & FMODE_32BITHASH) ||
	    (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
		return 0;
	else
		return pos & 0xffffffff;
}

/*
 * Return 32- or 64-bit end-of-file for dx directories
 */
static inline loff_t ext4_get_htree_eof(struct file *filp)
{
	if ((filp->f_mode & FMODE_32BITHASH) ||
	    (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
		return EXT4_HTREE_EOF_32BIT;
	else
		return EXT4_HTREE_EOF_64BIT;
}


/*
311 312 313
 * ext4_dir_llseek() calls generic_file_llseek_size to handle htree
 * directories, where the "offset" is in terms of the filename hash
 * value instead of the byte offset.
314
 *
315 316 317 318 319
 * Because we may return a 64-bit hash that is well beyond offset limits,
 * we need to pass the max hash as the maximum allowable offset in
 * the htree directory case.
 *
 * For non-htree, ext4_llseek already chooses the proper max offset.
320
 */
321
static loff_t ext4_dir_llseek(struct file *file, loff_t offset, int whence)
322 323 324
{
	struct inode *inode = file->f_mapping->host;
	int dx_dir = is_dx_dir(inode);
325
	loff_t htree_max = ext4_get_htree_eof(file);
326

327
	if (likely(dx_dir))
328
		return generic_file_llseek_size(file, offset, whence,
329 330
						    htree_max, htree_max);
	else
331
		return ext4_llseek(file, offset, whence);
332
}
333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354

/*
 * This structure holds the nodes of the red-black tree used to store
 * the directory entry in hash order.
 */
struct fname {
	__u32		hash;
	__u32		minor_hash;
	struct rb_node	rb_hash;
	struct fname	*next;
	__u32		inode;
	__u8		name_len;
	__u8		file_type;
	char		name[0];
};

/*
 * This functoin implements a non-recursive way of freeing all of the
 * nodes in the red-black tree.
 */
static void free_rb_tree_fname(struct rb_root *root)
{
355 356 357
	struct fname *fname, *next;

	rbtree_postorder_for_each_entry_safe(fname, next, root, rb_hash)
358
		while (fname) {
359
			struct fname *old = fname;
360
			fname = fname->next;
361
			kfree(old);
362
		}
363 364

	*root = RB_ROOT;
365 366 367
}


368 369
static struct dir_private_info *ext4_htree_create_dir_info(struct file *filp,
							   loff_t pos)
370 371 372
{
	struct dir_private_info *p;

373
	p = kzalloc(sizeof(struct dir_private_info), GFP_KERNEL);
374 375
	if (!p)
		return NULL;
376 377
	p->curr_hash = pos2maj_hash(filp, pos);
	p->curr_minor_hash = pos2min_hash(filp, pos);
378 379 380
	return p;
}

381
void ext4_htree_free_dir_info(struct dir_private_info *p)
382 383 384 385 386 387 388 389
{
	free_rb_tree_fname(&p->root);
	kfree(p);
}

/*
 * Given a directory entry, enter it into the fname rb tree.
 */
390
int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
391
			     __u32 minor_hash,
392
			     struct ext4_dir_entry_2 *dirent)
393 394
{
	struct rb_node **p, *parent = NULL;
395
	struct fname *fname, *new_fn;
396 397 398
	struct dir_private_info *info;
	int len;

399
	info = dir_file->private_data;
400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447
	p = &info->root.rb_node;

	/* Create and allocate the fname structure */
	len = sizeof(struct fname) + dirent->name_len + 1;
	new_fn = kzalloc(len, GFP_KERNEL);
	if (!new_fn)
		return -ENOMEM;
	new_fn->hash = hash;
	new_fn->minor_hash = minor_hash;
	new_fn->inode = le32_to_cpu(dirent->inode);
	new_fn->name_len = dirent->name_len;
	new_fn->file_type = dirent->file_type;
	memcpy(new_fn->name, dirent->name, dirent->name_len);
	new_fn->name[dirent->name_len] = 0;

	while (*p) {
		parent = *p;
		fname = rb_entry(parent, struct fname, rb_hash);

		/*
		 * If the hash and minor hash match up, then we put
		 * them on a linked list.  This rarely happens...
		 */
		if ((new_fn->hash == fname->hash) &&
		    (new_fn->minor_hash == fname->minor_hash)) {
			new_fn->next = fname->next;
			fname->next = new_fn;
			return 0;
		}

		if (new_fn->hash < fname->hash)
			p = &(*p)->rb_left;
		else if (new_fn->hash > fname->hash)
			p = &(*p)->rb_right;
		else if (new_fn->minor_hash < fname->minor_hash)
			p = &(*p)->rb_left;
		else /* if (new_fn->minor_hash > fname->minor_hash) */
			p = &(*p)->rb_right;
	}

	rb_link_node(&new_fn->rb_hash, parent, p);
	rb_insert_color(&new_fn->rb_hash, &info->root);
	return 0;
}



/*
448
 * This is a helper function for ext4_dx_readdir.  It calls filldir
449 450 451
 * for all entres on the fname linked list.  (Normally there is only
 * one entry on the linked list, unless there are 62 bit hash collisions.)
 */
Al Viro's avatar
Al Viro committed
452 453
static int call_filldir(struct file *file, struct dir_context *ctx,
			struct fname *fname)
454
{
Al Viro's avatar
Al Viro committed
455 456 457
	struct dir_private_info *info = file->private_data;
	struct inode *inode = file_inode(file);
	struct super_block *sb = inode->i_sb;
458 459

	if (!fname) {
460 461 462
		ext4_msg(sb, KERN_ERR, "%s:%d: inode #%lu: comm %s: "
			 "called with null fname?!?", __func__, __LINE__,
			 inode->i_ino, current->comm);
463 464
		return 0;
	}
Al Viro's avatar
Al Viro committed
465
	ctx->pos = hash2pos(file, fname->hash, fname->minor_hash);
466
	while (fname) {
Al Viro's avatar
Al Viro committed
467 468
		if (!dir_emit(ctx, fname->name,
				fname->name_len,
469
				fname->inode,
Al Viro's avatar
Al Viro committed
470
				get_dtype(sb, fname->file_type))) {
471
			info->extra_fname = fname;
Al Viro's avatar
Al Viro committed
472
			return 1;
473 474 475 476 477 478
		}
		fname = fname->next;
	}
	return 0;
}

Al Viro's avatar
Al Viro committed
479
static int ext4_dx_readdir(struct file *file, struct dir_context *ctx)
480
{
Al Viro's avatar
Al Viro committed
481 482
	struct dir_private_info *info = file->private_data;
	struct inode *inode = file_inode(file);
483 484 485 486
	struct fname *fname;
	int	ret;

	if (!info) {
Al Viro's avatar
Al Viro committed
487
		info = ext4_htree_create_dir_info(file, ctx->pos);
488 489
		if (!info)
			return -ENOMEM;
Al Viro's avatar
Al Viro committed
490
		file->private_data = info;
491 492
	}

Al Viro's avatar
Al Viro committed
493
	if (ctx->pos == ext4_get_htree_eof(file))
494 495 496
		return 0;	/* EOF */

	/* Some one has messed with f_pos; reset the world */
Al Viro's avatar
Al Viro committed
497
	if (info->last_pos != ctx->pos) {
498 499 500
		free_rb_tree_fname(&info->root);
		info->curr_node = NULL;
		info->extra_fname = NULL;
Al Viro's avatar
Al Viro committed
501 502
		info->curr_hash = pos2maj_hash(file, ctx->pos);
		info->curr_minor_hash = pos2min_hash(file, ctx->pos);
503 504 505 506 507 508
	}

	/*
	 * If there are any leftover names on the hash collision
	 * chain, return them first.
	 */
509
	if (info->extra_fname) {
Al Viro's avatar
Al Viro committed
510
		if (call_filldir(file, ctx, info->extra_fname))
511 512
			goto finished;
		info->extra_fname = NULL;
513
		goto next_node;
514
	} else if (!info->curr_node)
515 516 517 518 519 520 521 522 523
		info->curr_node = rb_first(&info->root);

	while (1) {
		/*
		 * Fill the rbtree if we have no more entries,
		 * or the inode has changed since we last read in the
		 * cached entries.
		 */
		if ((!info->curr_node) ||
Al Viro's avatar
Al Viro committed
524
		    (file->f_version != inode->i_version)) {
525 526
			info->curr_node = NULL;
			free_rb_tree_fname(&info->root);
Al Viro's avatar
Al Viro committed
527 528
			file->f_version = inode->i_version;
			ret = ext4_htree_fill_tree(file, info->curr_hash,
529 530 531 532 533
						   info->curr_minor_hash,
						   &info->next_hash);
			if (ret < 0)
				return ret;
			if (ret == 0) {
Al Viro's avatar
Al Viro committed
534
				ctx->pos = ext4_get_htree_eof(file);
535 536 537 538 539 540 541 542
				break;
			}
			info->curr_node = rb_first(&info->root);
		}

		fname = rb_entry(info->curr_node, struct fname, rb_hash);
		info->curr_hash = fname->hash;
		info->curr_minor_hash = fname->minor_hash;
Al Viro's avatar
Al Viro committed
543
		if (call_filldir(file, ctx, fname))
544
			break;
545
	next_node:
546
		info->curr_node = rb_next(info->curr_node);
547 548 549 550 551 552
		if (info->curr_node) {
			fname = rb_entry(info->curr_node, struct fname,
					 rb_hash);
			info->curr_hash = fname->hash;
			info->curr_minor_hash = fname->minor_hash;
		} else {
553
			if (info->next_hash == ~0) {
Al Viro's avatar
Al Viro committed
554
				ctx->pos = ext4_get_htree_eof(file);
555 556 557 558 559 560 561
				break;
			}
			info->curr_hash = info->next_hash;
			info->curr_minor_hash = 0;
		}
	}
finished:
Al Viro's avatar
Al Viro committed
562
	info->last_pos = ctx->pos;
563 564 565
	return 0;
}

566
static int ext4_release_dir(struct inode *inode, struct file *filp)
567
{
568
	if (filp->private_data)
569
		ext4_htree_free_dir_info(filp->private_data);
570 571 572

	return 0;
}
573 574 575 576

const struct file_operations ext4_dir_operations = {
	.llseek		= ext4_dir_llseek,
	.read		= generic_read_dir,
Al Viro's avatar
Al Viro committed
577
	.iterate	= ext4_readdir,
578 579 580 581 582 583 584
	.unlocked_ioctl = ext4_ioctl,
#ifdef CONFIG_COMPAT
	.compat_ioctl	= ext4_compat_ioctl,
#endif
	.fsync		= ext4_sync_file,
	.release	= ext4_release_dir,
};