dir.c 46.1 KB
Newer Older
D
David Teigland 已提交
1 2
/*
 * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
3
 * Copyright (C) 2004-2006 Red Hat, Inc.  All rights reserved.
D
David Teigland 已提交
4 5 6
 *
 * This copyrighted material is made available to anyone wishing to use,
 * modify, copy, or redistribute it subject to the terms and conditions
7
 * of the GNU General Public License version 2.
D
David Teigland 已提交
8 9 10
 */

/*
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54
 * Implements Extendible Hashing as described in:
 *   "Extendible Hashing" by Fagin, et al in
 *     __ACM Trans. on Database Systems__, Sept 1979.
 *
 *
 * Here's the layout of dirents which is essentially the same as that of ext2
 * within a single block. The field de_name_len is the number of bytes
 * actually required for the name (no null terminator). The field de_rec_len
 * is the number of bytes allocated to the dirent. The offset of the next
 * dirent in the block is (dirent + dirent->de_rec_len). When a dirent is
 * deleted, the preceding dirent inherits its allocated space, ie
 * prev->de_rec_len += deleted->de_rec_len. Since the next dirent is obtained
 * by adding de_rec_len to the current dirent, this essentially causes the
 * deleted dirent to get jumped over when iterating through all the dirents.
 *
 * When deleting the first dirent in a block, there is no previous dirent so
 * the field de_ino is set to zero to designate it as deleted. When allocating
 * a dirent, gfs2_dirent_alloc iterates through the dirents in a block. If the
 * first dirent has (de_ino == 0) and de_rec_len is large enough, this first
 * dirent is allocated. Otherwise it must go through all the 'used' dirents
 * searching for one in which the amount of total space minus the amount of
 * used space will provide enough space for the new dirent.
 *
 * There are two types of blocks in which dirents reside. In a stuffed dinode,
 * the dirents begin at offset sizeof(struct gfs2_dinode) from the beginning of
 * the block.  In leaves, they begin at offset sizeof(struct gfs2_leaf) from the
 * beginning of the leaf block. The dirents reside in leaves when
 *
 * dip->i_di.di_flags & GFS2_DIF_EXHASH is true
 *
 * Otherwise, the dirents are "linear", within a single stuffed dinode block.
 *
 * When the dirents are in leaves, the actual contents of the directory file are
 * used as an array of 64-bit block pointers pointing to the leaf blocks. The
 * dirents are NOT in the directory file itself. There can be more than one
 * block pointer in the array that points to the same leaf. In fact, when a
 * directory is first converted from linear to exhash, all of the pointers
 * point to the same leaf.
 *
 * When a leaf is completely full, the size of the hash table can be
 * doubled unless it is already at the maximum size which is hard coded into
 * GFS2_DIR_MAX_DEPTH. After that, leaves are chained together in a linked list,
 * but never before the maximum hash table size has been reached.
 */
D
David Teigland 已提交
55 56 57 58 59 60

#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/spinlock.h>
#include <linux/buffer_head.h>
#include <linux/sort.h>
61
#include <linux/gfs2_ondisk.h>
62
#include <linux/crc32.h>
63
#include <linux/vmalloc.h>
64
#include <linux/lm_interface.h>
D
David Teigland 已提交
65 66

#include "gfs2.h"
67
#include "incore.h"
D
David Teigland 已提交
68 69 70 71 72 73 74
#include "dir.h"
#include "glock.h"
#include "inode.h"
#include "meta_io.h"
#include "quota.h"
#include "rgrp.h"
#include "trans.h"
75
#include "bmap.h"
76
#include "util.h"
D
David Teigland 已提交
77 78 79 80

#define IS_LEAF     1 /* Hashed (leaf) directory */
#define IS_DINODE   2 /* Linear (stuffed dinode block) directory */

81 82
#define gfs2_disk_hash2offset(h) (((u64)(h)) >> 1)
#define gfs2_dir_offset2hash(p) ((u32)(((u64)(p)) << 1))
D
David Teigland 已提交
83

84
typedef int (*leaf_call_t) (struct gfs2_inode *dip, u32 index, u32 len,
85 86 87
			    u64 leaf_no, void *data);
typedef int (*gfs2_dscan_t)(const struct gfs2_dirent *dent,
			    const struct qstr *name, void *opaque);
D
David Teigland 已提交
88

89

90
int gfs2_dir_get_new_buffer(struct gfs2_inode *ip, u64 block,
91
			    struct buffer_head **bhp)
92 93 94
{
	struct buffer_head *bh;

95 96 97 98
	bh = gfs2_meta_new(ip->i_gl, block);
	gfs2_trans_add_bh(ip->i_gl, bh, 1);
	gfs2_metatype_set(bh, GFS2_METATYPE_JD, GFS2_FORMAT_JD);
	gfs2_buffer_clear_tail(bh, sizeof(struct gfs2_meta_header));
99 100 101 102
	*bhp = bh;
	return 0;
}

103
static int gfs2_dir_get_existing_buffer(struct gfs2_inode *ip, u64 block,
104 105 106 107
					struct buffer_head **bhp)
{
	struct buffer_head *bh;
	int error;
108

S
Steven Whitehouse 已提交
109
	error = gfs2_meta_read(ip->i_gl, block, DIO_WAIT, &bh);
110 111
	if (error)
		return error;
112
	if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), bh, GFS2_METATYPE_JD)) {
113 114 115 116 117 118
		brelse(bh);
		return -EIO;
	}
	*bhp = bh;
	return 0;
}
119 120 121 122 123 124 125 126 127 128 129 130

static int gfs2_dir_write_stuffed(struct gfs2_inode *ip, const char *buf,
				  unsigned int offset, unsigned int size)
{
	struct buffer_head *dibh;
	int error;

	error = gfs2_meta_inode_buffer(ip, &dibh);
	if (error)
		return error;

	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
131
	memcpy(dibh->b_data + offset + sizeof(struct gfs2_dinode), buf, size);
132 133
	if (ip->i_di.di_size < offset + size)
		ip->i_di.di_size = offset + size;
134
	ip->i_inode.i_mtime.tv_sec = ip->i_inode.i_ctime.tv_sec = get_seconds();
135
	gfs2_dinode_out(ip, dibh->b_data);
136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153

	brelse(dibh);

	return size;
}



/**
 * gfs2_dir_write_data - Write directory information to the inode
 * @ip: The GFS2 inode
 * @buf: The buffer containing information to be written
 * @offset: The file offset to start writing at
 * @size: The amount of data to write
 *
 * Returns: The number of bytes correctly written or error code
 */
static int gfs2_dir_write_data(struct gfs2_inode *ip, const char *buf,
154
			       u64 offset, unsigned int size)
155
{
156
	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
157
	struct buffer_head *dibh;
158 159
	u64 lblock, dblock;
	u32 extlen = 0;
160 161 162 163 164 165 166 167 168
	unsigned int o;
	int copied = 0;
	int error = 0;

	if (!size)
		return 0;

	if (gfs2_is_stuffed(ip) &&
	    offset + size <= sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode))
169 170
		return gfs2_dir_write_stuffed(ip, buf, (unsigned int)offset,
					      size);
171 172 173 174 175

	if (gfs2_assert_warn(sdp, gfs2_is_jdata(ip)))
		return -EINVAL;

	if (gfs2_is_stuffed(ip)) {
176
		error = gfs2_unstuff_dinode(ip, NULL);
177
		if (error)
178
			return error;
179 180 181 182 183 184 185 186
	}

	lblock = offset;
	o = do_div(lblock, sdp->sd_jbsize) + sizeof(struct gfs2_meta_header);

	while (copied < size) {
		unsigned int amount;
		struct buffer_head *bh;
187
		int new = 0;
188 189 190 191 192 193 194

		amount = size - copied;
		if (amount > sdp->sd_sb.sb_bsize - o)
			amount = sdp->sd_sb.sb_bsize - o;

		if (!extlen) {
			new = 1;
195
			error = gfs2_extent_map(&ip->i_inode, lblock, &new,
S
Steven Whitehouse 已提交
196
						&dblock, &extlen);
197 198 199 200 201 202 203
			if (error)
				goto fail;
			error = -EIO;
			if (gfs2_assert_withdraw(sdp, dblock))
				goto fail;
		}

204 205 206 207 208
		if (amount == sdp->sd_jbsize || new)
			error = gfs2_dir_get_new_buffer(ip, dblock, &bh);
		else
			error = gfs2_dir_get_existing_buffer(ip, dblock, &bh);

209 210 211 212 213 214 215
		if (error)
			goto fail;

		gfs2_trans_add_bh(ip->i_gl, bh, 1);
		memcpy(bh->b_data + o, buf, amount);
		brelse(bh);

216
		buf += amount;
217 218 219 220 221 222 223 224 225 226 227 228 229 230 231
		copied += amount;
		lblock++;
		dblock++;
		extlen--;

		o = sizeof(struct gfs2_meta_header);
	}

out:
	error = gfs2_meta_inode_buffer(ip, &dibh);
	if (error)
		return error;

	if (ip->i_di.di_size < offset + copied)
		ip->i_di.di_size = offset + copied;
232
	ip->i_inode.i_mtime.tv_sec = ip->i_inode.i_ctime.tv_sec = get_seconds();
233 234

	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
235
	gfs2_dinode_out(ip, dibh->b_data);
236 237 238 239 240 241 242 243 244 245
	brelse(dibh);

	return copied;
fail:
	if (copied)
		goto out;
	return error;
}

static int gfs2_dir_read_stuffed(struct gfs2_inode *ip, char *buf,
S
Steven Whitehouse 已提交
246
				 u64 offset, unsigned int size)
247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270
{
	struct buffer_head *dibh;
	int error;

	error = gfs2_meta_inode_buffer(ip, &dibh);
	if (!error) {
		offset += sizeof(struct gfs2_dinode);
		memcpy(buf, dibh->b_data + offset, size);
		brelse(dibh);
	}

	return (error) ? error : size;
}


/**
 * gfs2_dir_read_data - Read a data from a directory inode
 * @ip: The GFS2 Inode
 * @buf: The buffer to place result into
 * @offset: File offset to begin jdata_readng from
 * @size: Amount of data to transfer
 *
 * Returns: The amount of data actually copied or the error
 */
S
Steven Whitehouse 已提交
271 272
static int gfs2_dir_read_data(struct gfs2_inode *ip, char *buf, u64 offset,
			      unsigned int size, unsigned ra)
273
{
274
	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
275 276
	u64 lblock, dblock;
	u32 extlen = 0;
277 278 279 280 281 282 283
	unsigned int o;
	int copied = 0;
	int error = 0;

	if (offset >= ip->i_di.di_size)
		return 0;

J
Jan Engelhardt 已提交
284
	if (offset + size > ip->i_di.di_size)
285 286 287 288 289 290
		size = ip->i_di.di_size - offset;

	if (!size)
		return 0;

	if (gfs2_is_stuffed(ip))
S
Steven Whitehouse 已提交
291
		return gfs2_dir_read_stuffed(ip, buf, offset, size);
292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309

	if (gfs2_assert_warn(sdp, gfs2_is_jdata(ip)))
		return -EINVAL;

	lblock = offset;
	o = do_div(lblock, sdp->sd_jbsize) + sizeof(struct gfs2_meta_header);

	while (copied < size) {
		unsigned int amount;
		struct buffer_head *bh;
		int new;

		amount = size - copied;
		if (amount > sdp->sd_sb.sb_bsize - o)
			amount = sdp->sd_sb.sb_bsize - o;

		if (!extlen) {
			new = 0;
310
			error = gfs2_extent_map(&ip->i_inode, lblock, &new,
S
Steven Whitehouse 已提交
311
						&dblock, &extlen);
S
Steven Whitehouse 已提交
312
			if (error || !dblock)
313
				goto fail;
S
Steven Whitehouse 已提交
314 315 316 317
			BUG_ON(extlen < 1);
			if (!ra)
				extlen = 1;
			bh = gfs2_meta_ra(ip->i_gl, dblock, extlen);
318
		} else {
S
Steven Whitehouse 已提交
319
			error = gfs2_meta_read(ip->i_gl, dblock, DIO_WAIT, &bh);
320 321
			if (error)
				goto fail;
S
Steven Whitehouse 已提交
322 323 324 325 326 327 328 329
		}
		error = gfs2_metatype_check(sdp, bh, GFS2_METATYPE_JD);
		if (error) {
			brelse(bh);
			goto fail;
		}
		dblock++;
		extlen--;
330 331
		memcpy(buf, bh->b_data + o, amount);
		brelse(bh);
332
		buf += amount;
333 334 335 336 337 338 339 340 341 342
		copied += amount;
		lblock++;
		o = sizeof(struct gfs2_meta_header);
	}

	return copied;
fail:
	return (copied) ? copied : error;
}

343 344 345 346 347
static inline int gfs2_dirent_sentinel(const struct gfs2_dirent *dent)
{
	return dent->de_inum.no_addr == 0 || dent->de_inum.no_formal_ino == 0;
}

348 349 350
static inline int __gfs2_dirent_find(const struct gfs2_dirent *dent,
				     const struct qstr *name, int ret)
{
351
	if (!gfs2_dirent_sentinel(dent) &&
352 353
	    be32_to_cpu(dent->de_hash) == name->hash &&
	    be16_to_cpu(dent->de_name_len) == name->len &&
354
	    memcmp(dent+1, name->name, name->len) == 0)
355 356 357 358 359
		return ret;
	return 0;
}

static int gfs2_dirent_find(const struct gfs2_dirent *dent,
360 361
			    const struct qstr *name,
			    void *opaque)
362 363 364 365 366
{
	return __gfs2_dirent_find(dent, name, 1);
}

static int gfs2_dirent_prev(const struct gfs2_dirent *dent,
367 368
			    const struct qstr *name,
			    void *opaque)
369 370 371 372 373 374 375
{
	return __gfs2_dirent_find(dent, name, 2);
}

/*
 * name->name holds ptr to start of block.
 * name->len holds size of block.
D
David Teigland 已提交
376
 */
377
static int gfs2_dirent_last(const struct gfs2_dirent *dent,
378 379
			    const struct qstr *name,
			    void *opaque)
380 381 382 383 384 385 386
{
	const char *start = name->name;
	const char *end = (const char *)dent + be16_to_cpu(dent->de_rec_len);
	if (name->len == (end - start))
		return 1;
	return 0;
}
D
David Teigland 已提交
387

388
static int gfs2_dirent_find_space(const struct gfs2_dirent *dent,
389 390
				  const struct qstr *name,
				  void *opaque)
D
David Teigland 已提交
391
{
392 393 394 395
	unsigned required = GFS2_DIRENT_SIZE(name->len);
	unsigned actual = GFS2_DIRENT_SIZE(be16_to_cpu(dent->de_name_len));
	unsigned totlen = be16_to_cpu(dent->de_rec_len);

396
	if (gfs2_dirent_sentinel(dent))
397
		actual = GFS2_DIRENT_SIZE(0);
J
Jan Engelhardt 已提交
398
	if (totlen - actual >= required)
399 400 401 402
		return 1;
	return 0;
}

403 404 405 406 407 408 409 410 411 412
struct dirent_gather {
	const struct gfs2_dirent **pdent;
	unsigned offset;
};

static int gfs2_dirent_gather(const struct gfs2_dirent *dent,
			      const struct qstr *name,
			      void *opaque)
{
	struct dirent_gather *g = opaque;
413
	if (!gfs2_dirent_sentinel(dent)) {
414 415 416 417 418
		g->pdent[g->offset++] = dent;
	}
	return 0;
}

419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440
/*
 * Other possible things to check:
 * - Inode located within filesystem size (and on valid block)
 * - Valid directory entry type
 * Not sure how heavy-weight we want to make this... could also check
 * hash is correct for example, but that would take a lot of extra time.
 * For now the most important thing is to check that the various sizes
 * are correct.
 */
static int gfs2_check_dirent(struct gfs2_dirent *dent, unsigned int offset,
			     unsigned int size, unsigned int len, int first)
{
	const char *msg = "gfs2_dirent too small";
	if (unlikely(size < sizeof(struct gfs2_dirent)))
		goto error;
	msg = "gfs2_dirent misaligned";
	if (unlikely(offset & 0x7))
		goto error;
	msg = "gfs2_dirent points beyond end of block";
	if (unlikely(offset + size > len))
		goto error;
	msg = "zero inode number";
441
	if (unlikely(!first && gfs2_dirent_sentinel(dent)))
442 443
		goto error;
	msg = "name length is greater than space in dirent";
444
	if (!gfs2_dirent_sentinel(dent) &&
445 446 447 448 449 450 451 452
	    unlikely(sizeof(struct gfs2_dirent)+be16_to_cpu(dent->de_name_len) >
		     size))
		goto error;
	return 0;
error:
	printk(KERN_WARNING "gfs2_check_dirent: %s (%s)\n", msg,
	       first ? "first in block" : "not first in block");
	return -EIO;
D
David Teigland 已提交
453 454
}

455
static int gfs2_dirent_offset(const void *buf)
456
{
457 458
	const struct gfs2_meta_header *h = buf;
	int offset;
459 460 461

	BUG_ON(buf == NULL);

462
	switch(be32_to_cpu(h->mh_type)) {
463 464 465 466 467 468 469 470 471
	case GFS2_METATYPE_LF:
		offset = sizeof(struct gfs2_leaf);
		break;
	case GFS2_METATYPE_DI:
		offset = sizeof(struct gfs2_dinode);
		break;
	default:
		goto wrong_type;
	}
472 473 474
	return offset;
wrong_type:
	printk(KERN_WARNING "gfs2_scan_dirent: wrong block type %u\n",
475
	       be32_to_cpu(h->mh_type));
476 477 478
	return -1;
}

479
static struct gfs2_dirent *gfs2_dirent_scan(struct inode *inode, void *buf,
480 481 482 483 484 485 486 487
					    unsigned int len, gfs2_dscan_t scan,
					    const struct qstr *name,
					    void *opaque)
{
	struct gfs2_dirent *dent, *prev;
	unsigned offset;
	unsigned size;
	int ret = 0;
488

489 490 491 492 493
	ret = gfs2_dirent_offset(buf);
	if (ret < 0)
		goto consist_inode;

	offset = ret;
494
	prev = NULL;
495
	dent = buf + offset;
496 497 498 499
	size = be16_to_cpu(dent->de_rec_len);
	if (gfs2_check_dirent(dent, offset, size, len, 1))
		goto consist_inode;
	do {
500
		ret = scan(dent, name, opaque);
501 502 503 504 505 506
		if (ret)
			break;
		offset += size;
		if (offset == len)
			break;
		prev = dent;
507
		dent = buf + offset;
508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525
		size = be16_to_cpu(dent->de_rec_len);
		if (gfs2_check_dirent(dent, offset, size, len, 0))
			goto consist_inode;
	} while(1);

	switch(ret) {
	case 0:
		return NULL;
	case 1:
		return dent;
	case 2:
		return prev ? prev : dent;
	default:
		BUG_ON(ret > 0);
		return ERR_PTR(ret);
	}

consist_inode:
526
	gfs2_consist_inode(GFS2_I(inode));
527 528 529 530
	return ERR_PTR(-EIO);
}


D
David Teigland 已提交
531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546
/**
 * dirent_first - Return the first dirent
 * @dip: the directory
 * @bh: The buffer
 * @dent: Pointer to list of dirents
 *
 * return first dirent whether bh points to leaf or stuffed dinode
 *
 * Returns: IS_LEAF, IS_DINODE, or -errno
 */

static int dirent_first(struct gfs2_inode *dip, struct buffer_head *bh,
			struct gfs2_dirent **dent)
{
	struct gfs2_meta_header *h = (struct gfs2_meta_header *)bh->b_data;

547
	if (be32_to_cpu(h->mh_type) == GFS2_METATYPE_LF) {
548
		if (gfs2_meta_check(GFS2_SB(&dip->i_inode), bh))
D
David Teigland 已提交
549 550 551 552 553
			return -EIO;
		*dent = (struct gfs2_dirent *)(bh->b_data +
					       sizeof(struct gfs2_leaf));
		return IS_LEAF;
	} else {
554
		if (gfs2_metatype_check(GFS2_SB(&dip->i_inode), bh, GFS2_METATYPE_DI))
D
David Teigland 已提交
555 556 557 558 559 560 561
			return -EIO;
		*dent = (struct gfs2_dirent *)(bh->b_data +
					       sizeof(struct gfs2_dinode));
		return IS_DINODE;
	}
}

562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579
static int dirent_check_reclen(struct gfs2_inode *dip,
			       const struct gfs2_dirent *d, const void *end_p)
{
	const void *ptr = d;
	u16 rec_len = be16_to_cpu(d->de_rec_len);

	if (unlikely(rec_len < sizeof(struct gfs2_dirent)))
		goto broken;
	ptr += rec_len;
	if (ptr < end_p)
		return rec_len;
	if (ptr == end_p)
		return -ENOENT;
broken:
	gfs2_consist_inode(dip);
	return -EIO;
}

D
David Teigland 已提交
580 581 582 583 584 585 586 587 588 589 590 591
/**
 * dirent_next - Next dirent
 * @dip: the directory
 * @bh: The buffer
 * @dent: Pointer to list of dirents
 *
 * Returns: 0 on success, error code otherwise
 */

static int dirent_next(struct gfs2_inode *dip, struct buffer_head *bh,
		       struct gfs2_dirent **dent)
{
592 593 594
	struct gfs2_dirent *cur = *dent, *tmp;
	char *bh_end = bh->b_data + bh->b_size;
	int ret;
D
David Teigland 已提交
595

596 597 598
	ret = dirent_check_reclen(dip, cur, bh_end);
	if (ret < 0)
		return ret;
599

600 601 602 603
	tmp = (void *)cur + ret;
	ret = dirent_check_reclen(dip, tmp, bh_end);
	if (ret == -EIO)
		return ret;
604

D
David Teigland 已提交
605
        /* Only the first dent could ever have de_inum.no_addr == 0 */
606
	if (gfs2_dirent_sentinel(tmp)) {
D
David Teigland 已提交
607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626
		gfs2_consist_inode(dip);
		return -EIO;
	}

	*dent = tmp;
	return 0;
}

/**
 * dirent_del - Delete a dirent
 * @dip: The GFS2 inode
 * @bh: The buffer
 * @prev: The previous dirent
 * @cur: The current dirent
 *
 */

static void dirent_del(struct gfs2_inode *dip, struct buffer_head *bh,
		       struct gfs2_dirent *prev, struct gfs2_dirent *cur)
{
627
	u16 cur_rec_len, prev_rec_len;
D
David Teigland 已提交
628

629
	if (gfs2_dirent_sentinel(cur)) {
D
David Teigland 已提交
630 631 632 633
		gfs2_consist_inode(dip);
		return;
	}

634
	gfs2_trans_add_bh(dip->i_gl, bh, 1);
D
David Teigland 已提交
635 636 637 638 639 640

	/* If there is no prev entry, this is the first entry in the block.
	   The de_rec_len is already as big as it needs to be.  Just zero
	   out the inode number and return.  */

	if (!prev) {
641 642
		cur->de_inum.no_addr = 0;
		cur->de_inum.no_formal_ino = 0;
D
David Teigland 已提交
643 644 645 646 647
		return;
	}

	/*  Combine this dentry with the previous one.  */

648 649
	prev_rec_len = be16_to_cpu(prev->de_rec_len);
	cur_rec_len = be16_to_cpu(cur->de_rec_len);
D
David Teigland 已提交
650 651 652 653 654 655 656

	if ((char *)prev + prev_rec_len != (char *)cur)
		gfs2_consist_inode(dip);
	if ((char *)cur + cur_rec_len > bh->b_data + bh->b_size)
		gfs2_consist_inode(dip);

	prev_rec_len += cur_rec_len;
657
	prev->de_rec_len = cpu_to_be16(prev_rec_len);
D
David Teigland 已提交
658 659
}

660 661 662
/*
 * Takes a dent from which to grab space as an argument. Returns the
 * newly created dent.
D
David Teigland 已提交
663
 */
664 665 666 667
static struct gfs2_dirent *gfs2_init_dirent(struct inode *inode,
					    struct gfs2_dirent *dent,
					    const struct qstr *name,
					    struct buffer_head *bh)
D
David Teigland 已提交
668
{
669
	struct gfs2_inode *ip = GFS2_I(inode);
670 671 672
	struct gfs2_dirent *ndent;
	unsigned offset = 0, totlen;

673
	if (!gfs2_dirent_sentinel(dent))
674 675 676 677 678 679 680 681
		offset = GFS2_DIRENT_SIZE(be16_to_cpu(dent->de_name_len));
	totlen = be16_to_cpu(dent->de_rec_len);
	BUG_ON(offset + name->len > totlen);
	gfs2_trans_add_bh(ip->i_gl, bh, 1);
	ndent = (struct gfs2_dirent *)((char *)dent + offset);
	dent->de_rec_len = cpu_to_be16(offset);
	gfs2_qstr2dirent(name, totlen - offset, ndent);
	return ndent;
D
David Teigland 已提交
682 683
}

684 685 686
static struct gfs2_dirent *gfs2_dirent_alloc(struct inode *inode,
					     struct buffer_head *bh,
					     const struct qstr *name)
D
David Teigland 已提交
687 688
{
	struct gfs2_dirent *dent;
689
	dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size,
690
				gfs2_dirent_find_space, name, NULL);
691 692 693
	if (!dent || IS_ERR(dent))
		return dent;
	return gfs2_init_dirent(inode, dent, name, bh);
D
David Teigland 已提交
694 695
}

696
static int get_leaf(struct gfs2_inode *dip, u64 leaf_no,
D
David Teigland 已提交
697 698 699 700
		    struct buffer_head **bhp)
{
	int error;

S
Steven Whitehouse 已提交
701
	error = gfs2_meta_read(dip->i_gl, leaf_no, DIO_WAIT, bhp);
702 703
	if (!error && gfs2_metatype_check(GFS2_SB(&dip->i_inode), *bhp, GFS2_METATYPE_LF)) {
		/* printk(KERN_INFO "block num=%llu\n", leaf_no); */
D
David Teigland 已提交
704
		error = -EIO;
705
	}
D
David Teigland 已提交
706 707 708 709 710 711 712 713 714 715 716 717 718

	return error;
}

/**
 * get_leaf_nr - Get a leaf number associated with the index
 * @dip: The GFS2 inode
 * @index:
 * @leaf_out:
 *
 * Returns: 0 on success, error code otherwise
 */

719 720
static int get_leaf_nr(struct gfs2_inode *dip, u32 index,
		       u64 *leaf_out)
D
David Teigland 已提交
721
{
A
Al Viro 已提交
722
	__be64 leaf_no;
D
David Teigland 已提交
723 724
	int error;

725
	error = gfs2_dir_read_data(dip, (char *)&leaf_no,
A
Al Viro 已提交
726 727
				    index * sizeof(__be64),
				    sizeof(__be64), 0);
728
	if (error != sizeof(u64))
D
David Teigland 已提交
729 730 731 732 733 734 735
		return (error < 0) ? error : -EIO;

	*leaf_out = be64_to_cpu(leaf_no);

	return 0;
}

736
static int get_first_leaf(struct gfs2_inode *dip, u32 index,
D
David Teigland 已提交
737 738
			  struct buffer_head **bh_out)
{
739
	u64 leaf_no;
D
David Teigland 已提交
740 741 742 743 744 745 746 747 748
	int error;

	error = get_leaf_nr(dip, index, &leaf_no);
	if (!error)
		error = get_leaf(dip, leaf_no, bh_out);

	return error;
}

749 750 751 752
static struct gfs2_dirent *gfs2_dirent_search(struct inode *inode,
					      const struct qstr *name,
					      gfs2_dscan_t scan,
					      struct buffer_head **pbh)
D
David Teigland 已提交
753
{
754 755
	struct buffer_head *bh;
	struct gfs2_dirent *dent;
756
	struct gfs2_inode *ip = GFS2_I(inode);
D
David Teigland 已提交
757 758
	int error;

759 760 761 762 763 764 765 766 767
	if (ip->i_di.di_flags & GFS2_DIF_EXHASH) {
		struct gfs2_leaf *leaf;
		unsigned hsize = 1 << ip->i_di.di_depth;
		unsigned index;
		u64 ln;
		if (hsize * sizeof(u64) != ip->i_di.di_size) {
			gfs2_consist_inode(ip);
			return ERR_PTR(-EIO);
		}
768

769 770 771 772 773 774
		index = name->hash >> (32 - ip->i_di.di_depth);
		error = get_first_leaf(ip, index, &bh);
		if (error)
			return ERR_PTR(error);
		do {
			dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size,
775
						scan, name, NULL);
776 777 778 779
			if (dent)
				goto got_dent;
			leaf = (struct gfs2_leaf *)bh->b_data;
			ln = be64_to_cpu(leaf->lf_next);
780
			brelse(bh);
781 782
			if (!ln)
				break;
783

784 785
			error = get_leaf(ip, ln, &bh);
		} while(!error);
D
David Teigland 已提交
786

787
		return error ? ERR_PTR(error) : NULL;
D
David Teigland 已提交
788 789
	}

790

791 792 793
	error = gfs2_meta_inode_buffer(ip, &bh);
	if (error)
		return ERR_PTR(error);
794
	dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size, scan, name, NULL);
795
got_dent:
796
	if (unlikely(dent == NULL || IS_ERR(dent))) {
797 798 799
		brelse(bh);
		bh = NULL;
	}
800 801 802
	*pbh = bh;
	return dent;
}
D
David Teigland 已提交
803

804 805
static struct gfs2_leaf *new_leaf(struct inode *inode, struct buffer_head **pbh, u16 depth)
{
806
	struct gfs2_inode *ip = GFS2_I(inode);
807 808 809 810
	u64 bn = gfs2_alloc_meta(ip);
	struct buffer_head *bh = gfs2_meta_new(ip->i_gl, bn);
	struct gfs2_leaf *leaf;
	struct gfs2_dirent *dent;
811
	struct qstr name = { .name = "", .len = 0, .hash = 0 };
812 813
	if (!bh)
		return NULL;
814

815 816 817 818
	gfs2_trans_add_bh(ip->i_gl, bh, 1);
	gfs2_metatype_set(bh, GFS2_METATYPE_LF, GFS2_FORMAT_LF);
	leaf = (struct gfs2_leaf *)bh->b_data;
	leaf->lf_depth = cpu_to_be16(depth);
819
	leaf->lf_entries = 0;
820
	leaf->lf_dirent_format = cpu_to_be32(GFS2_FORMAT_DE);
821
	leaf->lf_next = 0;
822 823
	memset(leaf->lf_reserved, 0, sizeof(leaf->lf_reserved));
	dent = (struct gfs2_dirent *)(leaf+1);
824
	gfs2_qstr2dirent(&name, bh->b_size - sizeof(struct gfs2_leaf), dent);
825 826
	*pbh = bh;
	return leaf;
D
David Teigland 已提交
827 828 829 830 831 832 833 834 835
}

/**
 * dir_make_exhash - Convert a stuffed directory into an ExHash directory
 * @dip: The GFS2 inode
 *
 * Returns: 0 on success, error code otherwise
 */

836
static int dir_make_exhash(struct inode *inode)
D
David Teigland 已提交
837
{
838 839
	struct gfs2_inode *dip = GFS2_I(inode);
	struct gfs2_sbd *sdp = GFS2_SB(inode);
D
David Teigland 已提交
840
	struct gfs2_dirent *dent;
841
	struct qstr args;
D
David Teigland 已提交
842 843 844
	struct buffer_head *bh, *dibh;
	struct gfs2_leaf *leaf;
	int y;
845
	u32 x;
A
Al Viro 已提交
846 847
	__be64 *lp;
	u64 bn;
D
David Teigland 已提交
848 849 850 851 852 853 854 855
	int error;

	error = gfs2_meta_inode_buffer(dip, &dibh);
	if (error)
		return error;

	/*  Turn over a new leaf  */

856 857 858 859
	leaf = new_leaf(inode, &bh, 0);
	if (!leaf)
		return -ENOSPC;
	bn = bh->b_blocknr;
D
David Teigland 已提交
860 861 862 863 864 865 866 867 868 869 870 871

	gfs2_assert(sdp, dip->i_di.di_entries < (1 << 16));
	leaf->lf_entries = cpu_to_be16(dip->i_di.di_entries);

	/*  Copy dirents  */

	gfs2_buffer_copy_tail(bh, sizeof(struct gfs2_leaf), dibh,
			     sizeof(struct gfs2_dinode));

	/*  Find last entry  */

	x = 0;
872 873 874
	args.len = bh->b_size - sizeof(struct gfs2_dinode) +
		   sizeof(struct gfs2_leaf);
	args.name = bh->b_data;
875
	dent = gfs2_dirent_scan(&dip->i_inode, bh->b_data, bh->b_size,
876
				gfs2_dirent_last, &args, NULL);
877 878 879 880 881 882 883 884 885
	if (!dent) {
		brelse(bh);
		brelse(dibh);
		return -EIO;
	}
	if (IS_ERR(dent)) {
		brelse(bh);
		brelse(dibh);
		return PTR_ERR(dent);
D
David Teigland 已提交
886 887 888 889 890
	}

	/*  Adjust the last dirent's record length
	   (Remember that dent still points to the last entry.)  */

891
	dent->de_rec_len = cpu_to_be16(be16_to_cpu(dent->de_rec_len) +
D
David Teigland 已提交
892
		sizeof(struct gfs2_dinode) -
893
		sizeof(struct gfs2_leaf));
D
David Teigland 已提交
894 895 896 897 898 899

	brelse(bh);

	/*  We're done with the new leaf block, now setup the new
	    hash table.  */

900
	gfs2_trans_add_bh(dip->i_gl, dibh, 1);
D
David Teigland 已提交
901 902
	gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));

A
Al Viro 已提交
903
	lp = (__be64 *)(dibh->b_data + sizeof(struct gfs2_dinode));
D
David Teigland 已提交
904 905 906 907 908 909

	for (x = sdp->sd_hash_ptrs; x--; lp++)
		*lp = cpu_to_be64(bn);

	dip->i_di.di_size = sdp->sd_sb.sb_bsize / 2;
	dip->i_di.di_blocks++;
910
	gfs2_set_inode_blocks(&dip->i_inode);
D
David Teigland 已提交
911 912 913 914 915
	dip->i_di.di_flags |= GFS2_DIF_EXHASH;

	for (x = sdp->sd_hash_ptrs, y = -1; x; x >>= 1, y++) ;
	dip->i_di.di_depth = y;

916
	gfs2_dinode_out(dip, dibh->b_data);
D
David Teigland 已提交
917 918 919 920 921 922 923 924 925 926 927 928 929 930 931

	brelse(dibh);

	return 0;
}

/**
 * dir_split_leaf - Split a leaf block into two
 * @dip: The GFS2 inode
 * @index:
 * @leaf_no:
 *
 * Returns: 0 on success, error code on failure
 */

932
static int dir_split_leaf(struct inode *inode, const struct qstr *name)
D
David Teigland 已提交
933
{
934
	struct gfs2_inode *dip = GFS2_I(inode);
D
David Teigland 已提交
935 936
	struct buffer_head *nbh, *obh, *dibh;
	struct gfs2_leaf *nleaf, *oleaf;
937
	struct gfs2_dirent *dent = NULL, *prev = NULL, *next = NULL, *new;
938
	u32 start, len, half_len, divider;
A
Al Viro 已提交
939 940
	u64 bn, leaf_no;
	__be64 *lp;
941
	u32 index;
D
David Teigland 已提交
942 943 944
	int x, moved = 0;
	int error;

945 946 947 948
	index = name->hash >> (32 - dip->i_di.di_depth);
	error = get_leaf_nr(dip, index, &leaf_no);
	if (error)
		return error;
D
David Teigland 已提交
949 950 951 952

	/*  Get the old leaf block  */
	error = get_leaf(dip, leaf_no, &obh);
	if (error)
953
		return error;
D
David Teigland 已提交
954 955

	oleaf = (struct gfs2_leaf *)obh->b_data;
956 957 958 959 960 961
	if (dip->i_di.di_depth == be16_to_cpu(oleaf->lf_depth)) {
		brelse(obh);
		return 1; /* can't split */
	}

	gfs2_trans_add_bh(dip->i_gl, obh, 1);
D
David Teigland 已提交
962

963 964 965 966 967 968
	nleaf = new_leaf(inode, &nbh, be16_to_cpu(oleaf->lf_depth) + 1);
	if (!nleaf) {
		brelse(obh);
		return -ENOSPC;
	}
	bn = nbh->b_blocknr;
D
David Teigland 已提交
969

970
	/*  Compute the start and len of leaf pointers in the hash table.  */
D
David Teigland 已提交
971 972 973
	len = 1 << (dip->i_di.di_depth - be16_to_cpu(oleaf->lf_depth));
	half_len = len >> 1;
	if (!half_len) {
974
		printk(KERN_WARNING "di_depth %u lf_depth %u index %u\n", dip->i_di.di_depth, be16_to_cpu(oleaf->lf_depth), index);
D
David Teigland 已提交
975 976 977 978 979 980 981 982 983 984
		gfs2_consist_inode(dip);
		error = -EIO;
		goto fail_brelse;
	}

	start = (index & ~(len - 1));

	/* Change the pointers.
	   Don't bother distinguishing stuffed from non-stuffed.
	   This code is complicated enough already. */
A
Al Viro 已提交
985
	lp = kmalloc(half_len * sizeof(__be64), GFP_NOFS | __GFP_NOFAIL);
D
David Teigland 已提交
986 987 988 989
	/*  Change the pointers  */
	for (x = 0; x < half_len; x++)
		lp[x] = cpu_to_be64(bn);

990 991 992
	error = gfs2_dir_write_data(dip, (char *)lp, start * sizeof(u64),
				    half_len * sizeof(u64));
	if (error != half_len * sizeof(u64)) {
D
David Teigland 已提交
993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010
		if (error >= 0)
			error = -EIO;
		goto fail_lpfree;
	}

	kfree(lp);

	/*  Compute the divider  */
	divider = (start + half_len) << (32 - dip->i_di.di_depth);

	/*  Copy the entries  */
	dirent_first(dip, obh, &dent);

	do {
		next = dent;
		if (dirent_next(dip, obh, &next))
			next = NULL;

1011
		if (!gfs2_dirent_sentinel(dent) &&
D
David Teigland 已提交
1012
		    be32_to_cpu(dent->de_hash) < divider) {
1013 1014 1015 1016
			struct qstr str;
			str.name = (char*)(dent+1);
			str.len = be16_to_cpu(dent->de_name_len);
			str.hash = be32_to_cpu(dent->de_hash);
1017
			new = gfs2_dirent_alloc(inode, nbh, &str);
1018 1019 1020 1021
			if (IS_ERR(new)) {
				error = PTR_ERR(new);
				break;
			}
D
David Teigland 已提交
1022 1023 1024

			new->de_inum = dent->de_inum; /* No endian worries */
			new->de_type = dent->de_type; /* No endian worries */
1025
			nleaf->lf_entries = cpu_to_be16(be16_to_cpu(nleaf->lf_entries)+1);
D
David Teigland 已提交
1026 1027 1028 1029 1030

			dirent_del(dip, obh, prev, dent);

			if (!oleaf->lf_entries)
				gfs2_consist_inode(dip);
1031
			oleaf->lf_entries = cpu_to_be16(be16_to_cpu(oleaf->lf_entries)-1);
D
David Teigland 已提交
1032 1033 1034 1035 1036

			if (!prev)
				prev = dent;

			moved = 1;
1037
		} else {
D
David Teigland 已提交
1038
			prev = dent;
1039
		}
D
David Teigland 已提交
1040
		dent = next;
1041
	} while (dent);
D
David Teigland 已提交
1042

1043
	oleaf->lf_depth = nleaf->lf_depth;
D
David Teigland 已提交
1044 1045

	error = gfs2_meta_inode_buffer(dip, &dibh);
1046
	if (!gfs2_assert_withdraw(GFS2_SB(&dip->i_inode), !error)) {
D
David Teigland 已提交
1047
		dip->i_di.di_blocks++;
1048
		gfs2_set_inode_blocks(&dip->i_inode);
1049
		gfs2_dinode_out(dip, dibh->b_data);
D
David Teigland 已提交
1050 1051 1052 1053 1054 1055 1056 1057
		brelse(dibh);
	}

	brelse(obh);
	brelse(nbh);

	return error;

1058
fail_lpfree:
D
David Teigland 已提交
1059 1060
	kfree(lp);

1061
fail_brelse:
D
David Teigland 已提交
1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075
	brelse(obh);
	brelse(nbh);
	return error;
}

/**
 * dir_double_exhash - Double size of ExHash table
 * @dip: The GFS2 dinode
 *
 * Returns: 0 on success, error code on failure
 */

static int dir_double_exhash(struct gfs2_inode *dip)
{
1076
	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
D
David Teigland 已提交
1077
	struct buffer_head *dibh;
1078 1079 1080 1081
	u32 hsize;
	u64 *buf;
	u64 *from, *to;
	u64 block;
D
David Teigland 已提交
1082 1083 1084 1085
	int x;
	int error = 0;

	hsize = 1 << dip->i_di.di_depth;
1086
	if (hsize * sizeof(u64) != dip->i_di.di_size) {
D
David Teigland 已提交
1087 1088 1089 1090 1091 1092 1093 1094 1095
		gfs2_consist_inode(dip);
		return -EIO;
	}

	/*  Allocate both the "from" and "to" buffers in one big chunk  */

	buf = kcalloc(3, sdp->sd_hash_bsize, GFP_KERNEL | __GFP_NOFAIL);

	for (block = dip->i_di.di_size >> sdp->sd_hash_bsize_shift; block--;) {
1096
		error = gfs2_dir_read_data(dip, (char *)buf,
D
David Teigland 已提交
1097
					    block * sdp->sd_hash_bsize,
S
Steven Whitehouse 已提交
1098
					    sdp->sd_hash_bsize, 1);
D
David Teigland 已提交
1099 1100 1101 1102 1103 1104 1105
		if (error != sdp->sd_hash_bsize) {
			if (error >= 0)
				error = -EIO;
			goto fail;
		}

		from = buf;
1106
		to = (u64 *)((char *)buf + sdp->sd_hash_bsize);
D
David Teigland 已提交
1107 1108 1109 1110 1111 1112

		for (x = sdp->sd_hash_ptrs; x--; from++) {
			*to++ = *from;	/*  No endianess worries  */
			*to++ = *from;
		}

1113
		error = gfs2_dir_write_data(dip,
D
David Teigland 已提交
1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128
					     (char *)buf + sdp->sd_hash_bsize,
					     block * sdp->sd_sb.sb_bsize,
					     sdp->sd_sb.sb_bsize);
		if (error != sdp->sd_sb.sb_bsize) {
			if (error >= 0)
				error = -EIO;
			goto fail;
		}
	}

	kfree(buf);

	error = gfs2_meta_inode_buffer(dip, &dibh);
	if (!gfs2_assert_withdraw(sdp, !error)) {
		dip->i_di.di_depth++;
1129
		gfs2_dinode_out(dip, dibh->b_data);
D
David Teigland 已提交
1130 1131 1132 1133 1134
		brelse(dibh);
	}

	return error;

1135
fail:
D
David Teigland 已提交
1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152
	kfree(buf);
	return error;
}

/**
 * compare_dents - compare directory entries by hash value
 * @a: first dent
 * @b: second dent
 *
 * When comparing the hash entries of @a to @b:
 *   gt: returns 1
 *   lt: returns -1
 *   eq: returns 0
 */

static int compare_dents(const void *a, const void *b)
{
1153
	const struct gfs2_dirent *dent_a, *dent_b;
1154
	u32 hash_a, hash_b;
D
David Teigland 已提交
1155 1156
	int ret = 0;

1157
	dent_a = *(const struct gfs2_dirent **)a;
1158
	hash_a = be32_to_cpu(dent_a->de_hash);
D
David Teigland 已提交
1159

1160
	dent_b = *(const struct gfs2_dirent **)b;
1161
	hash_b = be32_to_cpu(dent_b->de_hash);
D
David Teigland 已提交
1162 1163 1164 1165 1166 1167

	if (hash_a > hash_b)
		ret = 1;
	else if (hash_a < hash_b)
		ret = -1;
	else {
1168 1169
		unsigned int len_a = be16_to_cpu(dent_a->de_name_len);
		unsigned int len_b = be16_to_cpu(dent_b->de_name_len);
D
David Teigland 已提交
1170 1171 1172 1173 1174 1175

		if (len_a > len_b)
			ret = 1;
		else if (len_a < len_b)
			ret = -1;
		else
1176
			ret = memcmp(dent_a + 1, dent_b + 1, len_a);
D
David Teigland 已提交
1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199
	}

	return ret;
}

/**
 * do_filldir_main - read out directory entries
 * @dip: The GFS2 inode
 * @offset: The offset in the file to read from
 * @opaque: opaque data to pass to filldir
 * @filldir: The function to pass entries to
 * @darr: an array of struct gfs2_dirent pointers to read
 * @entries: the number of entries in darr
 * @copied: pointer to int that's non-zero if a entry has been copied out
 *
 * Jump through some hoops to make sure that if there are hash collsions,
 * they are read out at the beginning of a buffer.  We want to minimize
 * the possibility that they will fall into different readdir buffers or
 * that someone will want to seek to that location.
 *
 * Returns: errno, >0 on exception from filldir
 */

1200
static int do_filldir_main(struct gfs2_inode *dip, u64 *offset,
1201
			   void *opaque, filldir_t filldir,
1202
			   const struct gfs2_dirent **darr, u32 entries,
D
David Teigland 已提交
1203 1204
			   int *copied)
{
1205
	const struct gfs2_dirent *dent, *dent_next;
1206
	u64 off, off_next;
D
David Teigland 已提交
1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241
	unsigned int x, y;
	int run = 0;
	int error = 0;

	sort(darr, entries, sizeof(struct gfs2_dirent *), compare_dents, NULL);

	dent_next = darr[0];
	off_next = be32_to_cpu(dent_next->de_hash);
	off_next = gfs2_disk_hash2offset(off_next);

	for (x = 0, y = 1; x < entries; x++, y++) {
		dent = dent_next;
		off = off_next;

		if (y < entries) {
			dent_next = darr[y];
			off_next = be32_to_cpu(dent_next->de_hash);
			off_next = gfs2_disk_hash2offset(off_next);

			if (off < *offset)
				continue;
			*offset = off;

			if (off_next == off) {
				if (*copied && !run)
					return 1;
				run = 1;
			} else
				run = 0;
		} else {
			if (off < *offset)
				continue;
			*offset = off;
		}

1242
		error = filldir(opaque, (const char *)(dent + 1),
1243
				be16_to_cpu(dent->de_name_len),
1244
				off, be64_to_cpu(dent->de_inum.no_addr),
1245
				be16_to_cpu(dent->de_type));
D
David Teigland 已提交
1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260
		if (error)
			return 1;

		*copied = 1;
	}

	/* Increment the *offset by one, so the next time we come into the
	   do_filldir fxn, we get the next entry instead of the last one in the
	   current leaf */

	(*offset)++;

	return 0;
}

1261
static int gfs2_dir_read_leaf(struct inode *inode, u64 *offset, void *opaque,
1262 1263
			      filldir_t filldir, int *copied, unsigned *depth,
			      u64 leaf_no)
D
David Teigland 已提交
1264
{
1265
	struct gfs2_inode *ip = GFS2_I(inode);
1266 1267 1268 1269 1270 1271 1272 1273 1274 1275
	struct buffer_head *bh;
	struct gfs2_leaf *lf;
	unsigned entries = 0;
	unsigned leaves = 0;
	const struct gfs2_dirent **darr, *dent;
	struct dirent_gather g;
	struct buffer_head **larr;
	int leaf = 0;
	int error, i;
	u64 lfn = leaf_no;
D
David Teigland 已提交
1276 1277

	do {
1278
		error = get_leaf(ip, lfn, &bh);
D
David Teigland 已提交
1279
		if (error)
1280 1281 1282 1283 1284 1285 1286 1287 1288
			goto out;
		lf = (struct gfs2_leaf *)bh->b_data;
		if (leaves == 0)
			*depth = be16_to_cpu(lf->lf_depth);
		entries += be16_to_cpu(lf->lf_entries);
		leaves++;
		lfn = be64_to_cpu(lf->lf_next);
		brelse(bh);
	} while(lfn);
D
David Teigland 已提交
1289 1290 1291 1292

	if (!entries)
		return 0;

1293
	error = -ENOMEM;
1294
	larr = vmalloc((leaves + entries) * sizeof(void *));
1295 1296 1297 1298 1299 1300
	if (!larr)
		goto out;
	darr = (const struct gfs2_dirent **)(larr + leaves);
	g.pdent = darr;
	g.offset = 0;
	lfn = leaf_no;
D
David Teigland 已提交
1301

1302 1303
	do {
		error = get_leaf(ip, lfn, &bh);
D
David Teigland 已提交
1304
		if (error)
1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316
			goto out_kfree;
		lf = (struct gfs2_leaf *)bh->b_data;
		lfn = be64_to_cpu(lf->lf_next);
		if (lf->lf_entries) {
			dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size,
						gfs2_dirent_gather, NULL, &g);
			error = PTR_ERR(dent);
			if (IS_ERR(dent)) {
				goto out_kfree;
			}
			error = 0;
			larr[leaf++] = bh;
D
David Teigland 已提交
1317
		} else {
1318
			brelse(bh);
D
David Teigland 已提交
1319
		}
1320
	} while(lfn);
D
David Teigland 已提交
1321

1322
	error = do_filldir_main(ip, offset, opaque, filldir, darr,
D
David Teigland 已提交
1323
				entries, copied);
1324 1325 1326
out_kfree:
	for(i = 0; i < leaf; i++)
		brelse(larr[i]);
1327
	vfree(larr);
1328
out:
D
David Teigland 已提交
1329 1330 1331 1332
	return error;
}

/**
1333 1334 1335 1336 1337
 * dir_e_read - Reads the entries from a directory into a filldir buffer
 * @dip: dinode pointer
 * @offset: the hash of the last entry read shifted to the right once
 * @opaque: buffer for the filldir function to fill
 * @filldir: points to the filldir function to use
D
David Teigland 已提交
1338
 *
1339
 * Returns: errno
D
David Teigland 已提交
1340 1341
 */

1342
static int dir_e_read(struct inode *inode, u64 *offset, void *opaque,
1343
		      filldir_t filldir)
D
David Teigland 已提交
1344
{
1345 1346
	struct gfs2_inode *dip = GFS2_I(inode);
	struct gfs2_sbd *sdp = GFS2_SB(inode);
1347 1348 1349
	u32 hsize, len = 0;
	u32 ht_offset, lp_offset, ht_offset_cur = -1;
	u32 hash, index;
A
Al Viro 已提交
1350
	__be64 *lp;
1351 1352
	int copied = 0;
	int error = 0;
1353
	unsigned depth = 0;
D
David Teigland 已提交
1354 1355

	hsize = 1 << dip->i_di.di_depth;
1356
	if (hsize * sizeof(u64) != dip->i_di.di_size) {
D
David Teigland 已提交
1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372
		gfs2_consist_inode(dip);
		return -EIO;
	}

	hash = gfs2_dir_offset2hash(*offset);
	index = hash >> (32 - dip->i_di.di_depth);

	lp = kmalloc(sdp->sd_hash_bsize, GFP_KERNEL);
	if (!lp)
		return -ENOMEM;

	while (index < hsize) {
		lp_offset = index & (sdp->sd_hash_ptrs - 1);
		ht_offset = index - lp_offset;

		if (ht_offset_cur != ht_offset) {
1373
			error = gfs2_dir_read_data(dip, (char *)lp,
A
Al Viro 已提交
1374
						ht_offset * sizeof(__be64),
S
Steven Whitehouse 已提交
1375
						sdp->sd_hash_bsize, 1);
D
David Teigland 已提交
1376 1377 1378 1379 1380 1381 1382 1383
			if (error != sdp->sd_hash_bsize) {
				if (error >= 0)
					error = -EIO;
				goto out;
			}
			ht_offset_cur = ht_offset;
		}

1384 1385 1386
		error = gfs2_dir_read_leaf(inode, offset, opaque, filldir,
					   &copied, &depth,
					   be64_to_cpu(lp[lp_offset]));
D
David Teigland 已提交
1387
		if (error)
1388
			break;
D
David Teigland 已提交
1389

1390
		len = 1 << (dip->i_di.di_depth - depth);
D
David Teigland 已提交
1391 1392 1393
		index = (index & ~(len - 1)) + len;
	}

1394
out:
D
David Teigland 已提交
1395
	kfree(lp);
1396 1397
	if (error > 0)
		error = 0;
D
David Teigland 已提交
1398 1399 1400
	return error;
}

1401
int gfs2_dir_read(struct inode *inode, u64 *offset, void *opaque,
1402
		  filldir_t filldir)
D
David Teigland 已提交
1403
{
1404
	struct gfs2_inode *dip = GFS2_I(inode);
1405 1406
	struct dirent_gather g;
	const struct gfs2_dirent **darr, *dent;
D
David Teigland 已提交
1407 1408 1409 1410
	struct buffer_head *dibh;
	int copied = 0;
	int error;

1411 1412 1413 1414 1415 1416
	if (!dip->i_di.di_entries)
		return 0;

	if (dip->i_di.di_flags & GFS2_DIF_EXHASH)
		return dir_e_read(inode, offset, opaque, filldir);

D
David Teigland 已提交
1417 1418 1419 1420 1421 1422 1423 1424 1425
	if (!gfs2_is_stuffed(dip)) {
		gfs2_consist_inode(dip);
		return -EIO;
	}

	error = gfs2_meta_inode_buffer(dip, &dibh);
	if (error)
		return error;

1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443
	error = -ENOMEM;
	darr = kmalloc(dip->i_di.di_entries * sizeof(struct gfs2_dirent *),
		       GFP_KERNEL);
	if (darr) {
		g.pdent = darr;
		g.offset = 0;
		dent = gfs2_dirent_scan(inode, dibh->b_data, dibh->b_size,
					gfs2_dirent_gather, NULL, &g);
		if (IS_ERR(dent)) {
			error = PTR_ERR(dent);
			goto out;
		}
		error = do_filldir_main(dip, offset, opaque, filldir, darr,
					dip->i_di.di_entries, &copied);
out:
		kfree(darr);
	}

D
David Teigland 已提交
1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463
	if (error > 0)
		error = 0;

	brelse(dibh);

	return error;
}

/**
 * gfs2_dir_search - Search a directory
 * @dip: The GFS2 inode
 * @filename:
 * @inode:
 *
 * This routine searches a directory for a file or another directory.
 * Assumes a glock is held on dip.
 *
 * Returns: errno
 */

1464
int gfs2_dir_search(struct inode *dir, const struct qstr *name,
A
Al Viro 已提交
1465
		    struct gfs2_inum_host *inum, unsigned int *type)
D
David Teigland 已提交
1466
{
1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486
	struct buffer_head *bh;
	struct gfs2_dirent *dent;

	dent = gfs2_dirent_search(dir, name, gfs2_dirent_find, &bh);
	if (dent) {
		if (IS_ERR(dent))
			return PTR_ERR(dent);
		if (inum)
			gfs2_inum_in(inum, (char *)&dent->de_inum);
		if (type)
			*type = be16_to_cpu(dent->de_type);
		brelse(bh);
		return 0;
	}
	return -ENOENT;
}

static int dir_new_leaf(struct inode *inode, const struct qstr *name)
{
	struct buffer_head *bh, *obh;
1487
	struct gfs2_inode *ip = GFS2_I(inode);
1488
	struct gfs2_leaf *leaf, *oleaf;
D
David Teigland 已提交
1489
	int error;
1490 1491
	u32 index;
	u64 bn;
D
David Teigland 已提交
1492

1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506
	index = name->hash >> (32 - ip->i_di.di_depth);
	error = get_first_leaf(ip, index, &obh);
	if (error)
		return error;
	do {
		oleaf = (struct gfs2_leaf *)obh->b_data;
		bn = be64_to_cpu(oleaf->lf_next);
		if (!bn)
			break;
		brelse(obh);
		error = get_leaf(ip, bn, &obh);
		if (error)
			return error;
	} while(1);
D
David Teigland 已提交
1507

1508 1509 1510 1511 1512 1513 1514
	gfs2_trans_add_bh(ip->i_gl, obh, 1);

	leaf = new_leaf(inode, &bh, be16_to_cpu(oleaf->lf_depth));
	if (!leaf) {
		brelse(obh);
		return -ENOSPC;
	}
1515
	oleaf->lf_next = cpu_to_be64(bh->b_blocknr);
1516 1517 1518 1519 1520 1521 1522 1523
	brelse(bh);
	brelse(obh);

	error = gfs2_meta_inode_buffer(ip, &bh);
	if (error)
		return error;
	gfs2_trans_add_bh(ip->i_gl, bh, 1);
	ip->i_di.di_blocks++;
1524
	gfs2_set_inode_blocks(&ip->i_inode);
1525
	gfs2_dinode_out(ip, bh->b_data);
1526 1527
	brelse(bh);
	return 0;
D
David Teigland 已提交
1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539
}

/**
 * gfs2_dir_add - Add new filename into directory
 * @dip: The GFS2 inode
 * @filename: The new name
 * @inode: The inode number of the entry
 * @type: The type of the entry
 *
 * Returns: 0 on success, error code on failure
 */

1540
int gfs2_dir_add(struct inode *inode, const struct qstr *name,
A
Al Viro 已提交
1541
		 const struct gfs2_inum_host *inum, unsigned type)
D
David Teigland 已提交
1542
{
1543
	struct gfs2_inode *ip = GFS2_I(inode);
1544 1545 1546
	struct buffer_head *bh;
	struct gfs2_dirent *dent;
	struct gfs2_leaf *leaf;
D
David Teigland 已提交
1547 1548
	int error;

1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567
	while(1) {
		dent = gfs2_dirent_search(inode, name, gfs2_dirent_find_space,
					  &bh);
		if (dent) {
			if (IS_ERR(dent))
				return PTR_ERR(dent);
			dent = gfs2_init_dirent(inode, dent, name, bh);
			gfs2_inum_out(inum, (char *)&dent->de_inum);
			dent->de_type = cpu_to_be16(type);
			if (ip->i_di.di_flags & GFS2_DIF_EXHASH) {
				leaf = (struct gfs2_leaf *)bh->b_data;
				leaf->lf_entries = cpu_to_be16(be16_to_cpu(leaf->lf_entries) + 1);
			}
			brelse(bh);
			error = gfs2_meta_inode_buffer(ip, &bh);
			if (error)
				break;
			gfs2_trans_add_bh(ip->i_gl, bh, 1);
			ip->i_di.di_entries++;
1568
			ip->i_inode.i_mtime.tv_sec = ip->i_inode.i_ctime.tv_sec = get_seconds();
1569
			gfs2_dinode_out(ip, bh->b_data);
1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582
			brelse(bh);
			error = 0;
			break;
		}
		if (!(ip->i_di.di_flags & GFS2_DIF_EXHASH)) {
			error = dir_make_exhash(inode);
			if (error)
				break;
			continue;
		}
		error = dir_split_leaf(inode, name);
		if (error == 0)
			continue;
1583
		if (error < 0)
1584 1585 1586 1587 1588 1589
			break;
		if (ip->i_di.di_depth < GFS2_DIR_MAX_DEPTH) {
			error = dir_double_exhash(ip);
			if (error)
				break;
			error = dir_split_leaf(inode, name);
1590
			if (error < 0)
1591
				break;
1592 1593
			if (error == 0)
				continue;
1594 1595 1596 1597 1598 1599 1600
		}
		error = dir_new_leaf(inode, name);
		if (!error)
			continue;
		error = -ENOSPC;
		break;
	}
D
David Teigland 已提交
1601 1602 1603
	return error;
}

1604

D
David Teigland 已提交
1605 1606 1607 1608 1609 1610 1611 1612
/**
 * gfs2_dir_del - Delete a directory entry
 * @dip: The GFS2 inode
 * @filename: The filename
 *
 * Returns: 0 on success, error code on failure
 */

1613
int gfs2_dir_del(struct gfs2_inode *dip, const struct qstr *name)
D
David Teigland 已提交
1614
{
1615 1616
	struct gfs2_dirent *dent, *prev = NULL;
	struct buffer_head *bh;
D
David Teigland 已提交
1617 1618
	int error;

1619 1620
	/* Returns _either_ the entry (if its first in block) or the
	   previous entry otherwise */
1621
	dent = gfs2_dirent_search(&dip->i_inode, name, gfs2_dirent_prev, &bh);
1622 1623 1624 1625 1626 1627 1628 1629 1630
	if (!dent) {
		gfs2_consist_inode(dip);
		return -EIO;
	}
	if (IS_ERR(dent)) {
		gfs2_consist_inode(dip);
		return PTR_ERR(dent);
	}
	/* If not first in block, adjust pointers accordingly */
1631
	if (gfs2_dirent_find(dent, name, NULL) == 0) {
1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643
		prev = dent;
		dent = (struct gfs2_dirent *)((char *)dent + be16_to_cpu(prev->de_rec_len));
	}

	dirent_del(dip, bh, prev, dent);
	if (dip->i_di.di_flags & GFS2_DIF_EXHASH) {
		struct gfs2_leaf *leaf = (struct gfs2_leaf *)bh->b_data;
		u16 entries = be16_to_cpu(leaf->lf_entries);
		if (!entries)
			gfs2_consist_inode(dip);
		leaf->lf_entries = cpu_to_be16(--entries);
	}
1644
	brelse(bh);
1645 1646 1647 1648 1649 1650 1651 1652 1653

	error = gfs2_meta_inode_buffer(dip, &bh);
	if (error)
		return error;

	if (!dip->i_di.di_entries)
		gfs2_consist_inode(dip);
	gfs2_trans_add_bh(dip->i_gl, bh, 1);
	dip->i_di.di_entries--;
1654
	dip->i_inode.i_mtime.tv_sec = dip->i_inode.i_ctime.tv_sec = get_seconds();
1655
	gfs2_dinode_out(dip, bh->b_data);
1656
	brelse(bh);
1657
	mark_inode_dirty(&dip->i_inode);
D
David Teigland 已提交
1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674

	return error;
}

/**
 * gfs2_dir_mvino - Change inode number of directory entry
 * @dip: The GFS2 inode
 * @filename:
 * @new_inode:
 *
 * This routine changes the inode number of a directory entry.  It's used
 * by rename to change ".." when a directory is moved.
 * Assumes a glock is held on dvp.
 *
 * Returns: errno
 */

1675
int gfs2_dir_mvino(struct gfs2_inode *dip, const struct qstr *filename,
A
Al Viro 已提交
1676
		   struct gfs2_inum_host *inum, unsigned int new_type)
D
David Teigland 已提交
1677
{
1678 1679
	struct buffer_head *bh;
	struct gfs2_dirent *dent;
D
David Teigland 已提交
1680 1681
	int error;

1682
	dent = gfs2_dirent_search(&dip->i_inode, filename, gfs2_dirent_find, &bh);
1683 1684 1685 1686 1687 1688
	if (!dent) {
		gfs2_consist_inode(dip);
		return -EIO;
	}
	if (IS_ERR(dent))
		return PTR_ERR(dent);
D
David Teigland 已提交
1689

1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701
	gfs2_trans_add_bh(dip->i_gl, bh, 1);
	gfs2_inum_out(inum, (char *)&dent->de_inum);
	dent->de_type = cpu_to_be16(new_type);

	if (dip->i_di.di_flags & GFS2_DIF_EXHASH) {
		brelse(bh);
		error = gfs2_meta_inode_buffer(dip, &bh);
		if (error)
			return error;
		gfs2_trans_add_bh(dip->i_gl, bh, 1);
	}

1702
	dip->i_inode.i_mtime.tv_sec = dip->i_inode.i_ctime.tv_sec = get_seconds();
1703
	gfs2_dinode_out(dip, bh->b_data);
1704 1705
	brelse(bh);
	return 0;
D
David Teigland 已提交
1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718
}

/**
 * foreach_leaf - call a function for each leaf in a directory
 * @dip: the directory
 * @lc: the function to call for each each
 * @data: private data to pass to it
 *
 * Returns: errno
 */

static int foreach_leaf(struct gfs2_inode *dip, leaf_call_t lc, void *data)
{
1719
	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
D
David Teigland 已提交
1720
	struct buffer_head *bh;
1721
	struct gfs2_leaf *leaf;
1722 1723 1724
	u32 hsize, len;
	u32 ht_offset, lp_offset, ht_offset_cur = -1;
	u32 index = 0;
A
Al Viro 已提交
1725
	__be64 *lp;
1726
	u64 leaf_no;
D
David Teigland 已提交
1727 1728 1729
	int error = 0;

	hsize = 1 << dip->i_di.di_depth;
1730
	if (hsize * sizeof(u64) != dip->i_di.di_size) {
D
David Teigland 已提交
1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743
		gfs2_consist_inode(dip);
		return -EIO;
	}

	lp = kmalloc(sdp->sd_hash_bsize, GFP_KERNEL);
	if (!lp)
		return -ENOMEM;

	while (index < hsize) {
		lp_offset = index & (sdp->sd_hash_ptrs - 1);
		ht_offset = index - lp_offset;

		if (ht_offset_cur != ht_offset) {
1744
			error = gfs2_dir_read_data(dip, (char *)lp,
A
Al Viro 已提交
1745
						ht_offset * sizeof(__be64),
S
Steven Whitehouse 已提交
1746
						sdp->sd_hash_bsize, 1);
D
David Teigland 已提交
1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759
			if (error != sdp->sd_hash_bsize) {
				if (error >= 0)
					error = -EIO;
				goto out;
			}
			ht_offset_cur = ht_offset;
		}

		leaf_no = be64_to_cpu(lp[lp_offset]);
		if (leaf_no) {
			error = get_leaf(dip, leaf_no, &bh);
			if (error)
				goto out;
1760 1761
			leaf = (struct gfs2_leaf *)bh->b_data;
			len = 1 << (dip->i_di.di_depth - be16_to_cpu(leaf->lf_depth));
1762
			brelse(bh);
D
David Teigland 已提交
1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777

			error = lc(dip, index, len, leaf_no, data);
			if (error)
				goto out;

			index = (index & ~(len - 1)) + len;
		} else
			index++;
	}

	if (index != hsize) {
		gfs2_consist_inode(dip);
		error = -EIO;
	}

1778
out:
D
David Teigland 已提交
1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794
	kfree(lp);

	return error;
}

/**
 * leaf_dealloc - Deallocate a directory leaf
 * @dip: the directory
 * @index: the hash table offset in the directory
 * @len: the number of pointers to this leaf
 * @leaf_no: the leaf number
 * @data: not used
 *
 * Returns: errno
 */

1795 1796
static int leaf_dealloc(struct gfs2_inode *dip, u32 index, u32 len,
			u64 leaf_no, void *data)
D
David Teigland 已提交
1797
{
1798
	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
1799
	struct gfs2_leaf *tmp_leaf;
D
David Teigland 已提交
1800 1801
	struct gfs2_rgrp_list rlist;
	struct buffer_head *bh, *dibh;
1802
	u64 blk, nblk;
D
David Teigland 已提交
1803 1804
	unsigned int rg_blocks = 0, l_blocks = 0;
	char *ht;
1805
	unsigned int x, size = len * sizeof(u64);
D
David Teigland 已提交
1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825
	int error;

	memset(&rlist, 0, sizeof(struct gfs2_rgrp_list));

	ht = kzalloc(size, GFP_KERNEL);
	if (!ht)
		return -ENOMEM;

	gfs2_alloc_get(dip);

	error = gfs2_quota_hold(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
	if (error)
		goto out;

	error = gfs2_rindex_hold(sdp, &dip->i_alloc.al_ri_gh);
	if (error)
		goto out_qs;

	/*  Count the number of leaves  */

1826
	for (blk = leaf_no; blk; blk = nblk) {
D
David Teigland 已提交
1827 1828 1829
		error = get_leaf(dip, blk, &bh);
		if (error)
			goto out_rlist;
1830 1831
		tmp_leaf = (struct gfs2_leaf *)bh->b_data;
		nblk = be64_to_cpu(tmp_leaf->lf_next);
D
David Teigland 已提交
1832 1833 1834 1835 1836 1837 1838 1839 1840 1841
		brelse(bh);

		gfs2_rlist_add(sdp, &rlist, blk);
		l_blocks++;
	}

	gfs2_rlist_alloc(&rlist, LM_ST_EXCLUSIVE, 0);

	for (x = 0; x < rlist.rl_rgrps; x++) {
		struct gfs2_rgrpd *rgd;
1842
		rgd = rlist.rl_ghs[x].gh_gl->gl_object;
D
David Teigland 已提交
1843 1844 1845 1846 1847 1848 1849 1850
		rg_blocks += rgd->rd_ri.ri_length;
	}

	error = gfs2_glock_nq_m(rlist.rl_rgrps, rlist.rl_ghs);
	if (error)
		goto out_rlist;

	error = gfs2_trans_begin(sdp,
1851
			rg_blocks + (DIV_ROUND_UP(size, sdp->sd_jbsize) + 1) +
D
David Teigland 已提交
1852 1853 1854 1855
			RES_DINODE + RES_STATFS + RES_QUOTA, l_blocks);
	if (error)
		goto out_rg_gunlock;

1856
	for (blk = leaf_no; blk; blk = nblk) {
D
David Teigland 已提交
1857 1858 1859
		error = get_leaf(dip, blk, &bh);
		if (error)
			goto out_end_trans;
1860 1861
		tmp_leaf = (struct gfs2_leaf *)bh->b_data;
		nblk = be64_to_cpu(tmp_leaf->lf_next);
D
David Teigland 已提交
1862 1863 1864 1865 1866 1867 1868
		brelse(bh);

		gfs2_free_meta(dip, blk, 1);

		if (!dip->i_di.di_blocks)
			gfs2_consist_inode(dip);
		dip->i_di.di_blocks--;
1869
		gfs2_set_inode_blocks(&dip->i_inode);
D
David Teigland 已提交
1870 1871
	}

1872
	error = gfs2_dir_write_data(dip, ht, index * sizeof(u64), size);
D
David Teigland 已提交
1873 1874 1875 1876 1877 1878 1879 1880 1881 1882
	if (error != size) {
		if (error >= 0)
			error = -EIO;
		goto out_end_trans;
	}

	error = gfs2_meta_inode_buffer(dip, &dibh);
	if (error)
		goto out_end_trans;

1883
	gfs2_trans_add_bh(dip->i_gl, dibh, 1);
1884
	gfs2_dinode_out(dip, dibh->b_data);
D
David Teigland 已提交
1885 1886
	brelse(dibh);

1887
out_end_trans:
D
David Teigland 已提交
1888
	gfs2_trans_end(sdp);
1889
out_rg_gunlock:
D
David Teigland 已提交
1890
	gfs2_glock_dq_m(rlist.rl_rgrps, rlist.rl_ghs);
1891
out_rlist:
D
David Teigland 已提交
1892 1893
	gfs2_rlist_free(&rlist);
	gfs2_glock_dq_uninit(&dip->i_alloc.al_ri_gh);
1894
out_qs:
D
David Teigland 已提交
1895
	gfs2_quota_unhold(dip);
1896
out:
D
David Teigland 已提交
1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913
	gfs2_alloc_put(dip);
	kfree(ht);
	return error;
}

/**
 * gfs2_dir_exhash_dealloc - free all the leaf blocks in a directory
 * @dip: the directory
 *
 * Dealloc all on-disk directory leaves to FREEMETA state
 * Change on-disk inode type to "regular file"
 *
 * Returns: errno
 */

int gfs2_dir_exhash_dealloc(struct gfs2_inode *dip)
{
1914
	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
D
David Teigland 已提交
1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931
	struct buffer_head *bh;
	int error;

	/* Dealloc on-disk leaves to FREEMETA state */
	error = foreach_leaf(dip, leaf_dealloc, NULL);
	if (error)
		return error;

	/* Make this a regular file in case we crash.
	   (We don't want to free these blocks a second time.)  */

	error = gfs2_trans_begin(sdp, RES_DINODE, 0);
	if (error)
		return error;

	error = gfs2_meta_inode_buffer(dip, &bh);
	if (!error) {
1932
		gfs2_trans_add_bh(dip->i_gl, bh, 1);
1933 1934
		((struct gfs2_dinode *)bh->b_data)->di_mode =
						cpu_to_be32(S_IFREG);
D
David Teigland 已提交
1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947
		brelse(bh);
	}

	gfs2_trans_end(sdp);

	return error;
}

/**
 * gfs2_diradd_alloc_required - find if adding entry will require an allocation
 * @ip: the file being written to
 * @filname: the filename that's going to be added
 *
1948
 * Returns: 1 if alloc required, 0 if not, -ve on error
D
David Teigland 已提交
1949 1950
 */

1951
int gfs2_diradd_alloc_required(struct inode *inode, const struct qstr *name)
D
David Teigland 已提交
1952
{
1953 1954
	struct gfs2_dirent *dent;
	struct buffer_head *bh;
D
David Teigland 已提交
1955

1956
	dent = gfs2_dirent_search(inode, name, gfs2_dirent_find_space, &bh);
1957
	if (!dent) {
1958
		return 1;
1959
	}
1960 1961 1962 1963
	if (IS_ERR(dent))
		return PTR_ERR(dent);
	brelse(bh);
	return 0;
D
David Teigland 已提交
1964 1965
}