dir.c 13.6 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

31
static unsigned char ext4_filetype_table[] = {
32 33 34
	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
};

35
static int ext4_readdir(struct file *, void *, filldir_t);
36 37 38 39
static int ext4_dx_readdir(struct file *filp,
			   void *dirent, filldir_t filldir);
static int ext4_release_dir(struct inode *inode,
				struct file *filp);
40

41
const struct file_operations ext4_dir_operations = {
42 43
	.llseek		= generic_file_llseek,
	.read		= generic_read_dir,
44
	.readdir	= ext4_readdir,		/* we take BKL. needed?*/
45
	.unlocked_ioctl = ext4_ioctl,
46
#ifdef CONFIG_COMPAT
47
	.compat_ioctl	= ext4_compat_ioctl,
48
#endif
49
	.fsync		= ext4_sync_file,
50
	.release	= ext4_release_dir,
51 52 53 54 55
};


static unsigned char get_dtype(struct super_block *sb, int filetype)
{
56 57
	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE) ||
	    (filetype >= EXT4_FT_MAX))
58 59
		return DT_UNKNOWN;

60
	return (ext4_filetype_table[filetype]);
61 62 63
}


64 65 66
int ext4_check_dir_entry(const char *function, struct inode *dir,
			 struct ext4_dir_entry_2 *de,
			 struct buffer_head *bh,
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 (rlen < EXT4_DIR_REC_LEN(1))
74 75 76
		error_msg = "rec_len is smaller than minimal";
	else if (rlen % 4 != 0)
		error_msg = "rec_len % 4 != 0";
77
	else if (rlen < EXT4_DIR_REC_LEN(de->name_len))
78 79 80 81
		error_msg = "rec_len is too small for name_len";
	else if (((char *) de - bh->b_data) + rlen > dir->i_sb->s_blocksize)
		error_msg = "directory entry across blocks";
	else if (le32_to_cpu(de->inode) >
82
			le32_to_cpu(EXT4_SB(dir->i_sb)->s_es->s_inodes_count))
83 84 85
		error_msg = "inode out of bounds";

	if (error_msg != NULL)
86 87
		ext4_error_inode(function, dir,
			"bad entry in directory: %s - block=%llu"
88
			"offset=%u(%u), inode=%u, rec_len=%d, name_len=%d",
89
			error_msg, (unsigned long long) bh->b_blocknr,
90
			(unsigned) (offset%bh->b_size), offset,
91
			le32_to_cpu(de->inode),
92 93 94 95
			rlen, de->name_len);
	return error_msg == NULL ? 1 : 0;
}

96 97
static int ext4_readdir(struct file *filp,
			 void *dirent, filldir_t filldir)
98 99
{
	int error = 0;
100
	unsigned int offset;
101
	int i, stored;
102
	struct ext4_dir_entry_2 *de;
103 104
	struct super_block *sb;
	int err;
105
	struct inode *inode = filp->f_path.dentry->d_inode;
106
	int ret = 0;
107
	int dir_has_error = 0;
108 109 110

	sb = inode->i_sb;

111 112
	if (EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
				    EXT4_FEATURE_COMPAT_DIR_INDEX) &&
113
	    ((ext4_test_inode_flag(inode, EXT4_INODE_INDEX)) ||
114
	     ((inode->i_size >> sb->s_blocksize_bits) == 1))) {
115
		err = ext4_dx_readdir(filp, dirent, filldir);
116 117 118 119 120 121 122 123
		if (err != ERR_BAD_DX_DIR) {
			ret = err;
			goto out;
		}
		/*
		 * We don't set the inode dirty flag since it's not
		 * critical that it get flushed back to the disk.
		 */
124
		ext4_clear_inode_flag(filp->f_path.dentry->d_inode, EXT4_INODE_INDEX);
125 126 127 128 129
	}
	stored = 0;
	offset = filp->f_pos & (sb->s_blocksize - 1);

	while (!error && !stored && filp->f_pos < inode->i_size) {
130
		struct ext4_map_blocks map;
131 132
		struct buffer_head *bh = NULL;

133 134 135
		map.m_lblk = filp->f_pos >> EXT4_BLOCK_SIZE_BITS(sb);
		map.m_len = 1;
		err = ext4_map_blocks(NULL, inode, &map, 0);
136
		if (err > 0) {
137
			pgoff_t index = map.m_pblk >>
138 139
					(PAGE_CACHE_SHIFT - inode->i_blkbits);
			if (!ra_has_index(&filp->f_ra, index))
140
				page_cache_sync_readahead(
141 142
					sb->s_bdev->bd_inode->i_mapping,
					&filp->f_ra, filp,
143
					index, 1);
144
			filp->f_ra.prev_pos = (loff_t)index << PAGE_CACHE_SHIFT;
145
			bh = ext4_bread(NULL, inode, map.m_lblk, 0, &err);
146 147 148 149 150 151 152
		}

		/*
		 * We ignore I/O errors on directories so users have a chance
		 * of recovering data when there's a bad sector
		 */
		if (!bh) {
153
			if (!dir_has_error) {
154
				EXT4_ERROR_INODE(inode, "directory "
155 156 157 158
					   "contains a hole at offset %Lu",
					   (unsigned long long) filp->f_pos);
				dir_has_error = 1;
			}
159 160 161
			/* corrupt size?  Maybe no more blocks to read */
			if (filp->f_pos > inode->i_blocks << 9)
				break;
162 163 164 165 166 167 168 169 170 171 172
			filp->f_pos += sb->s_blocksize - offset;
			continue;
		}

revalidate:
		/* 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. */
		if (filp->f_version != inode->i_version) {
			for (i = 0; i < sb->s_blocksize && i < offset; ) {
173
				de = (struct ext4_dir_entry_2 *)
174 175 176 177 178 179 180
					(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. */
181 182
				if (ext4_rec_len_from_disk(de->rec_len,
					sb->s_blocksize) < EXT4_DIR_REC_LEN(1))
183
					break;
184 185
				i += ext4_rec_len_from_disk(de->rec_len,
							    sb->s_blocksize);
186 187 188 189 190 191 192 193 194
			}
			offset = i;
			filp->f_pos = (filp->f_pos & ~(sb->s_blocksize - 1))
				| offset;
			filp->f_version = inode->i_version;
		}

		while (!error && filp->f_pos < inode->i_size
		       && offset < sb->s_blocksize) {
195
			de = (struct ext4_dir_entry_2 *) (bh->b_data + offset);
196 197
			if (!ext4_check_dir_entry("ext4_readdir", inode, de,
						  bh, offset)) {
198 199 200
				/*
				 * On error, skip the f_pos to the next block
				 */
201 202
				filp->f_pos = (filp->f_pos |
						(sb->s_blocksize - 1)) + 1;
203
				brelse(bh);
204 205 206
				ret = stored;
				goto out;
			}
207 208
			offset += ext4_rec_len_from_disk(de->rec_len,
					sb->s_blocksize);
209 210 211 212 213 214 215 216
			if (le32_to_cpu(de->inode)) {
				/* We might block in the next section
				 * if the data destination is
				 * currently swapped out.  So, use a
				 * version stamp to detect whether or
				 * not the directory has been modified
				 * during the copy operation.
				 */
217
				u64 version = filp->f_version;
218 219 220 221 222 223 224 225 226 227

				error = filldir(dirent, de->name,
						de->name_len,
						filp->f_pos,
						le32_to_cpu(de->inode),
						get_dtype(sb, de->file_type));
				if (error)
					break;
				if (version != filp->f_version)
					goto revalidate;
228
				stored++;
229
			}
230 231
			filp->f_pos += ext4_rec_len_from_disk(de->rec_len,
						sb->s_blocksize);
232 233
		}
		offset = 0;
234
		brelse(bh);
235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283
	}
out:
	return ret;
}

/*
 * These functions convert from the major/minor hash to an f_pos
 * value.
 *
 * Currently we only use major hash numer.  This is unfortunate, but
 * on 32-bit machines, the same VFS interface is used for lseek and
 * llseek, so if we use the 64 bit offset, then the 32-bit versions of
 * lseek/telldir/seekdir will blow out spectacularly, and from within
 * the ext2 low-level routine, we don't know if we're being called by
 * a 64-bit version of the system call or the 32-bit version of the
 * system call.  Worse yet, NFSv2 only allows for a 32-bit readdir
 * cookie.  Sigh.
 */
#define hash2pos(major, minor)	(major >> 1)
#define pos2maj_hash(pos)	((pos << 1) & 0xffffffff)
#define pos2min_hash(pos)	(0)

/*
 * 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)
{
	struct rb_node	*n = root->rb_node;
	struct rb_node	*parent;
	struct fname	*fname;

	while (n) {
		/* Do the node's children first */
284
		if (n->rb_left) {
285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300
			n = n->rb_left;
			continue;
		}
		if (n->rb_right) {
			n = n->rb_right;
			continue;
		}
		/*
		 * The node has no children; free it, and then zero
		 * out parent's link to it.  Finally go to the
		 * beginning of the loop and try to free the parent
		 * node.
		 */
		parent = rb_parent(n);
		fname = rb_entry(n, struct fname, rb_hash);
		while (fname) {
301
			struct fname *old = fname;
302
			fname = fname->next;
303
			kfree(old);
304 305
		}
		if (!parent)
306
			*root = RB_ROOT;
307 308 309 310 311 312 313 314 315
		else if (parent->rb_left == n)
			parent->rb_left = NULL;
		else if (parent->rb_right == n)
			parent->rb_right = NULL;
		n = parent;
	}
}


316
static struct dir_private_info *ext4_htree_create_dir_info(loff_t pos)
317 318 319
{
	struct dir_private_info *p;

320
	p = kzalloc(sizeof(struct dir_private_info), GFP_KERNEL);
321 322 323 324 325 326 327
	if (!p)
		return NULL;
	p->curr_hash = pos2maj_hash(pos);
	p->curr_minor_hash = pos2min_hash(pos);
	return p;
}

328
void ext4_htree_free_dir_info(struct dir_private_info *p)
329 330 331 332 333 334 335 336
{
	free_rb_tree_fname(&p->root);
	kfree(p);
}

/*
 * Given a directory entry, enter it into the fname rb tree.
 */
337
int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
338
			     __u32 minor_hash,
339
			     struct ext4_dir_entry_2 *dirent)
340 341
{
	struct rb_node **p, *parent = NULL;
342
	struct fname *fname, *new_fn;
343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394
	struct dir_private_info *info;
	int len;

	info = (struct dir_private_info *) dir_file->private_data;
	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;
}



/*
395
 * This is a helper function for ext4_dx_readdir.  It calls filldir
396 397 398
 * 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.)
 */
399
static int call_filldir(struct file *filp, void *dirent,
400 401 402 403
			filldir_t filldir, struct fname *fname)
{
	struct dir_private_info *info = filp->private_data;
	loff_t	curr_pos;
404
	struct inode *inode = filp->f_path.dentry->d_inode;
405
	struct super_block *sb;
406 407 408 409 410
	int error;

	sb = inode->i_sb;

	if (!fname) {
411
		printk(KERN_ERR "EXT4-fs: call_filldir: called with "
412
		       "null fname?!?\n");
413 414 415 416 417 418 419 420 421 422
		return 0;
	}
	curr_pos = hash2pos(fname->hash, fname->minor_hash);
	while (fname) {
		error = filldir(dirent, fname->name,
				fname->name_len, curr_pos,
				fname->inode,
				get_dtype(sb, fname->file_type));
		if (error) {
			filp->f_pos = curr_pos;
423
			info->extra_fname = fname;
424 425 426 427 428 429 430
			return error;
		}
		fname = fname->next;
	}
	return 0;
}

431 432
static int ext4_dx_readdir(struct file *filp,
			 void *dirent, filldir_t filldir)
433 434
{
	struct dir_private_info *info = filp->private_data;
435
	struct inode *inode = filp->f_path.dentry->d_inode;
436 437 438 439
	struct fname *fname;
	int	ret;

	if (!info) {
440
		info = ext4_htree_create_dir_info(filp->f_pos);
441 442 443 444 445
		if (!info)
			return -ENOMEM;
		filp->private_data = info;
	}

446
	if (filp->f_pos == EXT4_HTREE_EOF)
447 448 449 450 451 452 453 454 455 456 457 458 459 460 461
		return 0;	/* EOF */

	/* Some one has messed with f_pos; reset the world */
	if (info->last_pos != filp->f_pos) {
		free_rb_tree_fname(&info->root);
		info->curr_node = NULL;
		info->extra_fname = NULL;
		info->curr_hash = pos2maj_hash(filp->f_pos);
		info->curr_minor_hash = pos2min_hash(filp->f_pos);
	}

	/*
	 * If there are any leftover names on the hash collision
	 * chain, return them first.
	 */
462 463 464 465
	if (info->extra_fname) {
		if (call_filldir(filp, dirent, filldir, info->extra_fname))
			goto finished;
		info->extra_fname = NULL;
466
		goto next_node;
467
	} else if (!info->curr_node)
468 469 470 471 472 473 474 475 476 477 478 479 480
		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) ||
		    (filp->f_version != inode->i_version)) {
			info->curr_node = NULL;
			free_rb_tree_fname(&info->root);
			filp->f_version = inode->i_version;
481
			ret = ext4_htree_fill_tree(filp, info->curr_hash,
482 483 484 485 486
						   info->curr_minor_hash,
						   &info->next_hash);
			if (ret < 0)
				return ret;
			if (ret == 0) {
487
				filp->f_pos = EXT4_HTREE_EOF;
488 489 490 491 492 493 494 495 496 497
				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;
		if (call_filldir(filp, dirent, filldir, fname))
			break;
498
	next_node:
499
		info->curr_node = rb_next(info->curr_node);
500 501 502 503 504 505
		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 {
506
			if (info->next_hash == ~0) {
507
				filp->f_pos = EXT4_HTREE_EOF;
508 509 510 511 512 513 514 515 516 517 518
				break;
			}
			info->curr_hash = info->next_hash;
			info->curr_minor_hash = 0;
		}
	}
finished:
	info->last_pos = filp->f_pos;
	return 0;
}

519
static int ext4_release_dir(struct inode *inode, struct file *filp)
520
{
521
	if (filp->private_data)
522
		ext4_htree_free_dir_info(filp->private_data);
523 524 525

	return 0;
}