dir.c 47.6 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

#include <linux/slab.h>
#include <linux/spinlock.h>
#include <linux/buffer_head.h>
#include <linux/sort.h>
60
#include <linux/gfs2_ondisk.h>
61
#include <linux/crc32.h>
62
#include <linux/vmalloc.h>
63
#include <linux/lm_interface.h>
D
David Teigland 已提交
64 65

#include "gfs2.h"
66
#include "incore.h"
D
David Teigland 已提交
67 68 69 70 71 72 73
#include "dir.h"
#include "glock.h"
#include "inode.h"
#include "meta_io.h"
#include "quota.h"
#include "rgrp.h"
#include "trans.h"
74
#include "bmap.h"
75
#include "util.h"
D
David Teigland 已提交
76 77 78 79

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

80 81
#define gfs2_disk_hash2offset(h) (((u64)(h)) >> 1)
#define gfs2_dir_offset2hash(p) ((u32)(((u64)(p)) << 1))
D
David Teigland 已提交
82

83
typedef int (*leaf_call_t) (struct gfs2_inode *dip, u32 index, u32 len,
84 85 86
			    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
87

88

89
int gfs2_dir_get_new_buffer(struct gfs2_inode *ip, u64 block,
90
			    struct buffer_head **bhp)
91 92 93
{
	struct buffer_head *bh;

94 95 96 97
	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));
98 99 100 101
	*bhp = bh;
	return 0;
}

102
static int gfs2_dir_get_existing_buffer(struct gfs2_inode *ip, u64 block,
103 104 105 106
					struct buffer_head **bhp)
{
	struct buffer_head *bh;
	int error;
107

S
Steven Whitehouse 已提交
108
	error = gfs2_meta_read(ip->i_gl, block, DIO_WAIT, &bh);
109 110
	if (error)
		return error;
111
	if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), bh, GFS2_METATYPE_JD)) {
112 113 114 115 116 117
		brelse(bh);
		return -EIO;
	}
	*bhp = bh;
	return 0;
}
118 119 120 121 122 123 124 125 126 127 128 129

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);
130
	memcpy(dibh->b_data + offset + sizeof(struct gfs2_dinode), buf, size);
131 132
	if (ip->i_di.di_size < offset + size)
		ip->i_di.di_size = offset + size;
133
	ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME;
134
	gfs2_dinode_out(ip, dibh->b_data);
135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152

	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,
153
			       u64 offset, unsigned int size)
154
{
155
	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
156
	struct buffer_head *dibh;
157 158
	u64 lblock, dblock;
	u32 extlen = 0;
159 160 161 162 163 164 165 166 167
	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))
168 169
		return gfs2_dir_write_stuffed(ip, buf, (unsigned int)offset,
					      size);
170 171 172 173 174

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

	if (gfs2_is_stuffed(ip)) {
175
		error = gfs2_unstuff_dinode(ip, NULL);
176
		if (error)
177
			return error;
178 179 180 181 182 183 184 185
	}

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

	while (copied < size) {
		unsigned int amount;
		struct buffer_head *bh;
186
		int new = 0;
187 188 189 190 191 192 193

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

		if (!extlen) {
			new = 1;
194
			error = gfs2_extent_map(&ip->i_inode, lblock, &new,
S
Steven Whitehouse 已提交
195
						&dblock, &extlen);
196 197 198 199 200 201 202
			if (error)
				goto fail;
			error = -EIO;
			if (gfs2_assert_withdraw(sdp, dblock))
				goto fail;
		}

203 204 205 206 207
		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);

208 209 210 211 212 213 214
		if (error)
			goto fail;

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

215
		buf += amount;
216 217 218 219 220 221 222 223 224 225 226 227 228 229 230
		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;
231
	ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME;
232 233

	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
234
	gfs2_dinode_out(ip, dibh->b_data);
235 236 237 238 239 240 241 242 243 244
	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
245
				 u64 offset, unsigned int size)
246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269
{
	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
270 271
static int gfs2_dir_read_data(struct gfs2_inode *ip, char *buf, u64 offset,
			      unsigned int size, unsigned ra)
272
{
273
	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
274 275
	u64 lblock, dblock;
	u32 extlen = 0;
276 277 278 279 280 281 282
	unsigned int o;
	int copied = 0;
	int error = 0;

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

J
Jan Engelhardt 已提交
283
	if (offset + size > ip->i_di.di_size)
284 285 286 287 288 289
		size = ip->i_di.di_size - offset;

	if (!size)
		return 0;

	if (gfs2_is_stuffed(ip))
S
Steven Whitehouse 已提交
290
		return gfs2_dir_read_stuffed(ip, buf, offset, size);
291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308

	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;
309
			error = gfs2_extent_map(&ip->i_inode, lblock, &new,
S
Steven Whitehouse 已提交
310
						&dblock, &extlen);
S
Steven Whitehouse 已提交
311
			if (error || !dblock)
312
				goto fail;
S
Steven Whitehouse 已提交
313 314 315 316
			BUG_ON(extlen < 1);
			if (!ra)
				extlen = 1;
			bh = gfs2_meta_ra(ip->i_gl, dblock, extlen);
317
		} else {
S
Steven Whitehouse 已提交
318
			error = gfs2_meta_read(ip->i_gl, dblock, DIO_WAIT, &bh);
319 320
			if (error)
				goto fail;
S
Steven Whitehouse 已提交
321 322 323 324 325 326 327 328
		}
		error = gfs2_metatype_check(sdp, bh, GFS2_METATYPE_JD);
		if (error) {
			brelse(bh);
			goto fail;
		}
		dblock++;
		extlen--;
329 330
		memcpy(buf, bh->b_data + o, amount);
		brelse(bh);
331
		buf += amount;
332 333 334 335 336 337 338 339 340 341
		copied += amount;
		lblock++;
		o = sizeof(struct gfs2_meta_header);
	}

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

342 343 344 345 346
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;
}

347 348 349
static inline int __gfs2_dirent_find(const struct gfs2_dirent *dent,
				     const struct qstr *name, int ret)
{
350
	if (!gfs2_dirent_sentinel(dent) &&
351 352
	    be32_to_cpu(dent->de_hash) == name->hash &&
	    be16_to_cpu(dent->de_name_len) == name->len &&
353
	    memcmp(dent+1, name->name, name->len) == 0)
354 355 356 357 358
		return ret;
	return 0;
}

static int gfs2_dirent_find(const struct gfs2_dirent *dent,
359 360
			    const struct qstr *name,
			    void *opaque)
361 362 363 364 365
{
	return __gfs2_dirent_find(dent, name, 1);
}

static int gfs2_dirent_prev(const struct gfs2_dirent *dent,
366 367
			    const struct qstr *name,
			    void *opaque)
368 369 370 371 372 373 374
{
	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
375
 */
376
static int gfs2_dirent_last(const struct gfs2_dirent *dent,
377 378
			    const struct qstr *name,
			    void *opaque)
379 380 381 382 383 384 385
{
	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
386

387
static int gfs2_dirent_find_space(const struct gfs2_dirent *dent,
388 389
				  const struct qstr *name,
				  void *opaque)
D
David Teigland 已提交
390
{
391 392 393 394
	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);

395
	if (gfs2_dirent_sentinel(dent))
396
		actual = GFS2_DIRENT_SIZE(0);
J
Jan Engelhardt 已提交
397
	if (totlen - actual >= required)
398 399 400 401
		return 1;
	return 0;
}

402 403 404 405 406 407 408 409 410 411
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;
412
	if (!gfs2_dirent_sentinel(dent)) {
413 414 415 416 417
		g->pdent[g->offset++] = dent;
	}
	return 0;
}

418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439
/*
 * 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";
440
	if (unlikely(!first && gfs2_dirent_sentinel(dent)))
441 442
		goto error;
	msg = "name length is greater than space in dirent";
443
	if (!gfs2_dirent_sentinel(dent) &&
444 445 446 447 448 449 450 451
	    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
452 453
}

454
static int gfs2_dirent_offset(const void *buf)
455
{
456 457
	const struct gfs2_meta_header *h = buf;
	int offset;
458 459 460

	BUG_ON(buf == NULL);

461
	switch(be32_to_cpu(h->mh_type)) {
462 463 464 465 466 467 468 469 470
	case GFS2_METATYPE_LF:
		offset = sizeof(struct gfs2_leaf);
		break;
	case GFS2_METATYPE_DI:
		offset = sizeof(struct gfs2_dinode);
		break;
	default:
		goto wrong_type;
	}
471 472 473
	return offset;
wrong_type:
	printk(KERN_WARNING "gfs2_scan_dirent: wrong block type %u\n",
474
	       be32_to_cpu(h->mh_type));
475 476 477
	return -1;
}

478
static struct gfs2_dirent *gfs2_dirent_scan(struct inode *inode, void *buf,
479 480 481 482 483 484 485 486
					    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;
487

488 489 490 491 492
	ret = gfs2_dirent_offset(buf);
	if (ret < 0)
		goto consist_inode;

	offset = ret;
493
	prev = NULL;
494
	dent = buf + offset;
495 496 497 498
	size = be16_to_cpu(dent->de_rec_len);
	if (gfs2_check_dirent(dent, offset, size, len, 1))
		goto consist_inode;
	do {
499
		ret = scan(dent, name, opaque);
500 501 502 503 504 505
		if (ret)
			break;
		offset += size;
		if (offset == len)
			break;
		prev = dent;
506
		dent = buf + offset;
507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524
		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:
525
	gfs2_consist_inode(GFS2_I(inode));
526 527 528 529
	return ERR_PTR(-EIO);
}


D
David Teigland 已提交
530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545
/**
 * 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;

546
	if (be32_to_cpu(h->mh_type) == GFS2_METATYPE_LF) {
547
		if (gfs2_meta_check(GFS2_SB(&dip->i_inode), bh))
D
David Teigland 已提交
548 549 550 551 552
			return -EIO;
		*dent = (struct gfs2_dirent *)(bh->b_data +
					       sizeof(struct gfs2_leaf));
		return IS_LEAF;
	} else {
553
		if (gfs2_metatype_check(GFS2_SB(&dip->i_inode), bh, GFS2_METATYPE_DI))
D
David Teigland 已提交
554 555 556 557 558 559 560
			return -EIO;
		*dent = (struct gfs2_dirent *)(bh->b_data +
					       sizeof(struct gfs2_dinode));
		return IS_DINODE;
	}
}

561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578
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
579 580 581 582 583 584 585 586 587 588 589 590
/**
 * 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)
{
591 592 593
	struct gfs2_dirent *cur = *dent, *tmp;
	char *bh_end = bh->b_data + bh->b_size;
	int ret;
D
David Teigland 已提交
594

595 596 597
	ret = dirent_check_reclen(dip, cur, bh_end);
	if (ret < 0)
		return ret;
598

599 600 601 602
	tmp = (void *)cur + ret;
	ret = dirent_check_reclen(dip, tmp, bh_end);
	if (ret == -EIO)
		return ret;
603

D
David Teigland 已提交
604
        /* Only the first dent could ever have de_inum.no_addr == 0 */
605
	if (gfs2_dirent_sentinel(tmp)) {
D
David Teigland 已提交
606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625
		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)
{
626
	u16 cur_rec_len, prev_rec_len;
D
David Teigland 已提交
627

628
	if (gfs2_dirent_sentinel(cur)) {
D
David Teigland 已提交
629 630 631 632
		gfs2_consist_inode(dip);
		return;
	}

633
	gfs2_trans_add_bh(dip->i_gl, bh, 1);
D
David Teigland 已提交
634 635 636 637 638 639

	/* 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) {
640 641
		cur->de_inum.no_addr = 0;
		cur->de_inum.no_formal_ino = 0;
D
David Teigland 已提交
642 643 644 645 646
		return;
	}

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

647 648
	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
649 650 651 652 653 654 655

	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;
656
	prev->de_rec_len = cpu_to_be16(prev_rec_len);
D
David Teigland 已提交
657 658
}

659 660 661
/*
 * Takes a dent from which to grab space as an argument. Returns the
 * newly created dent.
D
David Teigland 已提交
662
 */
663 664 665 666
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
667
{
668
	struct gfs2_inode *ip = GFS2_I(inode);
669 670 671
	struct gfs2_dirent *ndent;
	unsigned offset = 0, totlen;

672
	if (!gfs2_dirent_sentinel(dent))
673 674 675 676 677 678 679 680
		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
681 682
}

683 684 685
static struct gfs2_dirent *gfs2_dirent_alloc(struct inode *inode,
					     struct buffer_head *bh,
					     const struct qstr *name)
D
David Teigland 已提交
686 687
{
	struct gfs2_dirent *dent;
688
	dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size,
689
				gfs2_dirent_find_space, name, NULL);
690 691 692
	if (!dent || IS_ERR(dent))
		return dent;
	return gfs2_init_dirent(inode, dent, name, bh);
D
David Teigland 已提交
693 694
}

695
static int get_leaf(struct gfs2_inode *dip, u64 leaf_no,
D
David Teigland 已提交
696 697 698 699
		    struct buffer_head **bhp)
{
	int error;

S
Steven Whitehouse 已提交
700
	error = gfs2_meta_read(dip->i_gl, leaf_no, DIO_WAIT, bhp);
701 702
	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
703
		error = -EIO;
704
	}
D
David Teigland 已提交
705 706 707 708 709 710 711 712 713 714 715 716 717

	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
 */

718 719
static int get_leaf_nr(struct gfs2_inode *dip, u32 index,
		       u64 *leaf_out)
D
David Teigland 已提交
720
{
A
Al Viro 已提交
721
	__be64 leaf_no;
D
David Teigland 已提交
722 723
	int error;

724
	error = gfs2_dir_read_data(dip, (char *)&leaf_no,
A
Al Viro 已提交
725 726
				    index * sizeof(__be64),
				    sizeof(__be64), 0);
727
	if (error != sizeof(u64))
D
David Teigland 已提交
728 729 730 731 732 733 734
		return (error < 0) ? error : -EIO;

	*leaf_out = be64_to_cpu(leaf_no);

	return 0;
}

735
static int get_first_leaf(struct gfs2_inode *dip, u32 index,
D
David Teigland 已提交
736 737
			  struct buffer_head **bh_out)
{
738
	u64 leaf_no;
D
David Teigland 已提交
739 740 741 742 743 744 745 746 747
	int error;

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

	return error;
}

748 749 750 751
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
752
{
753 754
	struct buffer_head *bh;
	struct gfs2_dirent *dent;
755
	struct gfs2_inode *ip = GFS2_I(inode);
D
David Teigland 已提交
756 757
	int error;

758 759 760 761 762 763 764 765 766
	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);
		}
767

768 769 770 771 772 773
		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,
774
						scan, name, NULL);
775 776 777 778
			if (dent)
				goto got_dent;
			leaf = (struct gfs2_leaf *)bh->b_data;
			ln = be64_to_cpu(leaf->lf_next);
779
			brelse(bh);
780 781
			if (!ln)
				break;
782

783 784
			error = get_leaf(ip, ln, &bh);
		} while(!error);
D
David Teigland 已提交
785

786
		return error ? ERR_PTR(error) : NULL;
D
David Teigland 已提交
787 788
	}

789

790 791 792
	error = gfs2_meta_inode_buffer(ip, &bh);
	if (error)
		return ERR_PTR(error);
793
	dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size, scan, name, NULL);
794
got_dent:
795
	if (unlikely(dent == NULL || IS_ERR(dent))) {
796 797 798
		brelse(bh);
		bh = NULL;
	}
799 800 801
	*pbh = bh;
	return dent;
}
D
David Teigland 已提交
802

803 804
static struct gfs2_leaf *new_leaf(struct inode *inode, struct buffer_head **pbh, u16 depth)
{
805
	struct gfs2_inode *ip = GFS2_I(inode);
806 807 808 809
	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;
810
	struct qstr name = { .name = "", .len = 0, .hash = 0 };
811 812
	if (!bh)
		return NULL;
813

814 815 816 817
	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);
818
	leaf->lf_entries = 0;
819
	leaf->lf_dirent_format = cpu_to_be32(GFS2_FORMAT_DE);
820
	leaf->lf_next = 0;
821 822
	memset(leaf->lf_reserved, 0, sizeof(leaf->lf_reserved));
	dent = (struct gfs2_dirent *)(leaf+1);
823
	gfs2_qstr2dirent(&name, bh->b_size - sizeof(struct gfs2_leaf), dent);
824 825
	*pbh = bh;
	return leaf;
D
David Teigland 已提交
826 827 828 829 830 831 832 833 834
}

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

835
static int dir_make_exhash(struct inode *inode)
D
David Teigland 已提交
836
{
837 838
	struct gfs2_inode *dip = GFS2_I(inode);
	struct gfs2_sbd *sdp = GFS2_SB(inode);
D
David Teigland 已提交
839
	struct gfs2_dirent *dent;
840
	struct qstr args;
D
David Teigland 已提交
841 842 843
	struct buffer_head *bh, *dibh;
	struct gfs2_leaf *leaf;
	int y;
844
	u32 x;
A
Al Viro 已提交
845 846
	__be64 *lp;
	u64 bn;
D
David Teigland 已提交
847 848 849 850 851 852 853 854
	int error;

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

	/*  Turn over a new leaf  */

855 856 857 858
	leaf = new_leaf(inode, &bh, 0);
	if (!leaf)
		return -ENOSPC;
	bn = bh->b_blocknr;
D
David Teigland 已提交
859 860 861 862 863 864 865 866 867 868 869 870

	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;
871 872 873
	args.len = bh->b_size - sizeof(struct gfs2_dinode) +
		   sizeof(struct gfs2_leaf);
	args.name = bh->b_data;
874
	dent = gfs2_dirent_scan(&dip->i_inode, bh->b_data, bh->b_size,
875
				gfs2_dirent_last, &args, NULL);
876 877 878 879 880 881 882 883 884
	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
885 886 887 888 889
	}

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

890
	dent->de_rec_len = cpu_to_be16(be16_to_cpu(dent->de_rec_len) +
D
David Teigland 已提交
891
		sizeof(struct gfs2_dinode) -
892
		sizeof(struct gfs2_leaf));
D
David Teigland 已提交
893 894 895 896 897 898

	brelse(bh);

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

899
	gfs2_trans_add_bh(dip->i_gl, dibh, 1);
D
David Teigland 已提交
900 901
	gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));

A
Al Viro 已提交
902
	lp = (__be64 *)(dibh->b_data + sizeof(struct gfs2_dinode));
D
David Teigland 已提交
903 904 905 906 907 908

	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++;
909
	gfs2_set_inode_blocks(&dip->i_inode);
D
David Teigland 已提交
910 911 912 913 914
	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;

915
	gfs2_dinode_out(dip, dibh->b_data);
D
David Teigland 已提交
916 917 918 919 920 921 922 923 924 925 926 927 928 929 930

	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
 */

931
static int dir_split_leaf(struct inode *inode, const struct qstr *name)
D
David Teigland 已提交
932
{
933
	struct gfs2_inode *dip = GFS2_I(inode);
D
David Teigland 已提交
934 935
	struct buffer_head *nbh, *obh, *dibh;
	struct gfs2_leaf *nleaf, *oleaf;
936
	struct gfs2_dirent *dent = NULL, *prev = NULL, *next = NULL, *new;
937
	u32 start, len, half_len, divider;
A
Al Viro 已提交
938 939
	u64 bn, leaf_no;
	__be64 *lp;
940
	u32 index;
D
David Teigland 已提交
941 942 943
	int x, moved = 0;
	int error;

944 945 946 947
	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
948 949 950 951

	/*  Get the old leaf block  */
	error = get_leaf(dip, leaf_no, &obh);
	if (error)
952
		return error;
D
David Teigland 已提交
953 954

	oleaf = (struct gfs2_leaf *)obh->b_data;
955 956 957 958 959 960
	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
961

962 963 964 965 966 967
	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
968

969
	/*  Compute the start and len of leaf pointers in the hash table.  */
D
David Teigland 已提交
970 971 972
	len = 1 << (dip->i_di.di_depth - be16_to_cpu(oleaf->lf_depth));
	half_len = len >> 1;
	if (!half_len) {
973
		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
974 975 976 977 978 979 980 981 982 983
		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
984
	lp = kmalloc(half_len * sizeof(__be64), GFP_NOFS | __GFP_NOFAIL);
D
David Teigland 已提交
985 986 987 988
	/*  Change the pointers  */
	for (x = 0; x < half_len; x++)
		lp[x] = cpu_to_be64(bn);

989 990 991
	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
992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009
		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;

1010
		if (!gfs2_dirent_sentinel(dent) &&
D
David Teigland 已提交
1011
		    be32_to_cpu(dent->de_hash) < divider) {
1012 1013 1014 1015
			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);
1016
			new = gfs2_dirent_alloc(inode, nbh, &str);
1017 1018 1019 1020
			if (IS_ERR(new)) {
				error = PTR_ERR(new);
				break;
			}
D
David Teigland 已提交
1021 1022 1023

			new->de_inum = dent->de_inum; /* No endian worries */
			new->de_type = dent->de_type; /* No endian worries */
1024
			nleaf->lf_entries = cpu_to_be16(be16_to_cpu(nleaf->lf_entries)+1);
D
David Teigland 已提交
1025 1026 1027 1028 1029

			dirent_del(dip, obh, prev, dent);

			if (!oleaf->lf_entries)
				gfs2_consist_inode(dip);
1030
			oleaf->lf_entries = cpu_to_be16(be16_to_cpu(oleaf->lf_entries)-1);
D
David Teigland 已提交
1031 1032 1033 1034 1035

			if (!prev)
				prev = dent;

			moved = 1;
1036
		} else {
D
David Teigland 已提交
1037
			prev = dent;
1038
		}
D
David Teigland 已提交
1039
		dent = next;
1040
	} while (dent);
D
David Teigland 已提交
1041

1042
	oleaf->lf_depth = nleaf->lf_depth;
D
David Teigland 已提交
1043 1044

	error = gfs2_meta_inode_buffer(dip, &dibh);
1045
	if (!gfs2_assert_withdraw(GFS2_SB(&dip->i_inode), !error)) {
1046
		gfs2_trans_add_bh(dip->i_gl, dibh, 1);
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
	struct gfs2_sbd *sdp = GFS2_SB(inode);
1267 1268
	struct buffer_head *bh;
	struct gfs2_leaf *lf;
1269
	unsigned entries = 0, entries2 = 0;
1270 1271 1272 1273 1274 1275 1276
	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
1277 1278

	do {
1279
		error = get_leaf(ip, lfn, &bh);
D
David Teigland 已提交
1280
		if (error)
1281 1282 1283 1284 1285 1286 1287 1288 1289
			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
1290 1291 1292 1293

	if (!entries)
		return 0;

1294
	error = -ENOMEM;
1295 1296 1297 1298 1299 1300 1301
	/*
	 * The extra 99 entries are not normally used, but are a buffer
	 * zone in case the number of entries in the leaf is corrupt.
	 * 99 is the maximum number of entries that can fit in a single
	 * leaf block.
	 */
	larr = vmalloc((leaves + entries + 99) * sizeof(void *));
1302 1303 1304 1305 1306 1307
	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
1308

1309 1310
	do {
		error = get_leaf(ip, lfn, &bh);
D
David Teigland 已提交
1311
		if (error)
1312 1313 1314 1315
			goto out_kfree;
		lf = (struct gfs2_leaf *)bh->b_data;
		lfn = be64_to_cpu(lf->lf_next);
		if (lf->lf_entries) {
1316
			entries2 += be16_to_cpu(lf->lf_entries);
1317 1318 1319
			dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size,
						gfs2_dirent_gather, NULL, &g);
			error = PTR_ERR(dent);
1320 1321 1322
			if (IS_ERR(dent))
				goto out_kfree;
			if (entries2 != g.offset) {
1323 1324 1325 1326 1327
				fs_warn(sdp, "Number of entries corrupt in dir "
						"leaf %llu, entries2 (%u) != "
						"g.offset (%u)\n",
					(unsigned long long)bh->b_blocknr,
					entries2, g.offset);
1328 1329
					
				error = -EIO;
1330 1331 1332 1333
				goto out_kfree;
			}
			error = 0;
			larr[leaf++] = bh;
D
David Teigland 已提交
1334
		} else {
1335
			brelse(bh);
D
David Teigland 已提交
1336
		}
1337
	} while(lfn);
D
David Teigland 已提交
1338

1339
	BUG_ON(entries2 != entries);
1340
	error = do_filldir_main(ip, offset, opaque, filldir, darr,
D
David Teigland 已提交
1341
				entries, copied);
1342 1343 1344
out_kfree:
	for(i = 0; i < leaf; i++)
		brelse(larr[i]);
1345
	vfree(larr);
1346
out:
D
David Teigland 已提交
1347 1348 1349 1350
	return error;
}

/**
1351 1352 1353 1354 1355
 * 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
1356
 *
1357
 * Returns: errno
D
David Teigland 已提交
1358 1359
 */

1360
static int dir_e_read(struct inode *inode, u64 *offset, void *opaque,
1361
		      filldir_t filldir)
D
David Teigland 已提交
1362
{
1363 1364
	struct gfs2_inode *dip = GFS2_I(inode);
	struct gfs2_sbd *sdp = GFS2_SB(inode);
1365 1366 1367
	u32 hsize, len = 0;
	u32 ht_offset, lp_offset, ht_offset_cur = -1;
	u32 hash, index;
A
Al Viro 已提交
1368
	__be64 *lp;
1369 1370
	int copied = 0;
	int error = 0;
1371
	unsigned depth = 0;
D
David Teigland 已提交
1372 1373

	hsize = 1 << dip->i_di.di_depth;
1374
	if (hsize * sizeof(u64) != dip->i_di.di_size) {
D
David Teigland 已提交
1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390
		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) {
1391
			error = gfs2_dir_read_data(dip, (char *)lp,
A
Al Viro 已提交
1392
						ht_offset * sizeof(__be64),
S
Steven Whitehouse 已提交
1393
						sdp->sd_hash_bsize, 1);
D
David Teigland 已提交
1394 1395 1396 1397 1398 1399 1400 1401
			if (error != sdp->sd_hash_bsize) {
				if (error >= 0)
					error = -EIO;
				goto out;
			}
			ht_offset_cur = ht_offset;
		}

1402 1403 1404
		error = gfs2_dir_read_leaf(inode, offset, opaque, filldir,
					   &copied, &depth,
					   be64_to_cpu(lp[lp_offset]));
D
David Teigland 已提交
1405
		if (error)
1406
			break;
D
David Teigland 已提交
1407

1408
		len = 1 << (dip->i_di.di_depth - depth);
D
David Teigland 已提交
1409 1410 1411
		index = (index & ~(len - 1)) + len;
	}

1412
out:
D
David Teigland 已提交
1413
	kfree(lp);
1414 1415
	if (error > 0)
		error = 0;
D
David Teigland 已提交
1416 1417 1418
	return error;
}

1419
int gfs2_dir_read(struct inode *inode, u64 *offset, void *opaque,
1420
		  filldir_t filldir)
D
David Teigland 已提交
1421
{
1422
	struct gfs2_inode *dip = GFS2_I(inode);
1423
	struct gfs2_sbd *sdp = GFS2_SB(inode);
1424 1425
	struct dirent_gather g;
	const struct gfs2_dirent **darr, *dent;
D
David Teigland 已提交
1426 1427 1428 1429
	struct buffer_head *dibh;
	int copied = 0;
	int error;

1430 1431 1432 1433 1434 1435
	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
1436 1437 1438 1439 1440 1441 1442 1443 1444
	if (!gfs2_is_stuffed(dip)) {
		gfs2_consist_inode(dip);
		return -EIO;
	}

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

1445
	error = -ENOMEM;
1446 1447
	/* 96 is max number of dirents which can be stuffed into an inode */
	darr = kmalloc(96 * sizeof(struct gfs2_dirent *), GFP_KERNEL);
1448 1449 1450 1451 1452 1453 1454 1455 1456
	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;
		}
1457 1458 1459
		if (dip->i_di.di_entries != g.offset) {
			fs_warn(sdp, "Number of entries corrupt in dir %llu, "
				"ip->i_di.di_entries (%u) != g.offset (%u)\n",
1460
				(unsigned long long)dip->i_no_addr,
1461
				dip->i_di.di_entries,
1462 1463 1464 1465
				g.offset);
			error = -EIO;
			goto out;
		}
1466 1467 1468 1469 1470 1471
		error = do_filldir_main(dip, offset, opaque, filldir, darr,
					dip->i_di.di_entries, &copied);
out:
		kfree(darr);
	}

D
David Teigland 已提交
1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491
	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
 */

1492
struct inode *gfs2_dir_search(struct inode *dir, const struct qstr *name)
D
David Teigland 已提交
1493
{
1494 1495
	struct buffer_head *bh;
	struct gfs2_dirent *dent;
1496 1497 1498 1499 1500
	struct inode *inode;

	dent = gfs2_dirent_search(dir, name, gfs2_dirent_find, &bh);
	if (dent) {
		if (IS_ERR(dent))
1501
			return ERR_CAST(dent);
1502 1503 1504
		inode = gfs2_inode_lookup(dir->i_sb, 
				be16_to_cpu(dent->de_type),
				be64_to_cpu(dent->de_inum.no_addr),
1505
				be64_to_cpu(dent->de_inum.no_formal_ino), 0);
1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517
		brelse(bh);
		return inode;
	}
	return ERR_PTR(-ENOENT);
}

int gfs2_dir_check(struct inode *dir, const struct qstr *name,
		   const struct gfs2_inode *ip)
{
	struct buffer_head *bh;
	struct gfs2_dirent *dent;
	int ret = -ENOENT;
1518 1519 1520 1521 1522

	dent = gfs2_dirent_search(dir, name, gfs2_dirent_find, &bh);
	if (dent) {
		if (IS_ERR(dent))
			return PTR_ERR(dent);
1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537
		if (ip) {
			if (be64_to_cpu(dent->de_inum.no_addr) != ip->i_no_addr)
				goto out;
			if (be64_to_cpu(dent->de_inum.no_formal_ino) !=
			    ip->i_no_formal_ino)
				goto out;
			if (unlikely(IF2DT(ip->i_inode.i_mode) !=
			    be16_to_cpu(dent->de_type))) {
				gfs2_consist_inode(GFS2_I(dir));
				ret = -EIO;
				goto out;
			}
		}
		ret = 0;
out:
1538 1539
		brelse(bh);
	}
1540
	return ret;
1541 1542 1543 1544 1545
}

static int dir_new_leaf(struct inode *inode, const struct qstr *name)
{
	struct buffer_head *bh, *obh;
1546
	struct gfs2_inode *ip = GFS2_I(inode);
1547
	struct gfs2_leaf *leaf, *oleaf;
D
David Teigland 已提交
1548
	int error;
1549 1550
	u32 index;
	u64 bn;
D
David Teigland 已提交
1551

1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565
	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
1566

1567 1568 1569 1570 1571 1572 1573
	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;
	}
1574
	oleaf->lf_next = cpu_to_be64(bh->b_blocknr);
1575 1576 1577 1578 1579 1580 1581 1582
	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++;
1583
	gfs2_set_inode_blocks(&ip->i_inode);
1584
	gfs2_dinode_out(ip, bh->b_data);
1585 1586
	brelse(bh);
	return 0;
D
David Teigland 已提交
1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598
}

/**
 * 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
 */

1599
int gfs2_dir_add(struct inode *inode, const struct qstr *name,
1600
		 const struct gfs2_inode *nip, unsigned type)
D
David Teigland 已提交
1601
{
1602
	struct gfs2_inode *ip = GFS2_I(inode);
1603 1604 1605
	struct buffer_head *bh;
	struct gfs2_dirent *dent;
	struct gfs2_leaf *leaf;
D
David Teigland 已提交
1606 1607
	int error;

1608 1609 1610 1611 1612 1613 1614
	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);
1615
			gfs2_inum_out(nip, dent);
1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626
			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++;
1627
			ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME;
1628
			gfs2_dinode_out(ip, bh->b_data);
1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641
			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;
1642
		if (error < 0)
1643 1644 1645 1646 1647 1648
			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);
1649
			if (error < 0)
1650
				break;
1651 1652
			if (error == 0)
				continue;
1653 1654 1655 1656 1657 1658 1659
		}
		error = dir_new_leaf(inode, name);
		if (!error)
			continue;
		error = -ENOSPC;
		break;
	}
D
David Teigland 已提交
1660 1661 1662
	return error;
}

1663

D
David Teigland 已提交
1664 1665 1666 1667 1668 1669 1670 1671
/**
 * gfs2_dir_del - Delete a directory entry
 * @dip: The GFS2 inode
 * @filename: The filename
 *
 * Returns: 0 on success, error code on failure
 */

1672
int gfs2_dir_del(struct gfs2_inode *dip, const struct qstr *name)
D
David Teigland 已提交
1673
{
1674 1675
	struct gfs2_dirent *dent, *prev = NULL;
	struct buffer_head *bh;
D
David Teigland 已提交
1676 1677
	int error;

1678 1679
	/* Returns _either_ the entry (if its first in block) or the
	   previous entry otherwise */
1680
	dent = gfs2_dirent_search(&dip->i_inode, name, gfs2_dirent_prev, &bh);
1681 1682 1683 1684 1685 1686 1687 1688 1689
	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 */
1690
	if (gfs2_dirent_find(dent, name, NULL) == 0) {
1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702
		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);
	}
1703
	brelse(bh);
1704 1705 1706 1707 1708 1709 1710 1711 1712

	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--;
1713
	dip->i_inode.i_mtime = dip->i_inode.i_ctime = CURRENT_TIME;
1714
	gfs2_dinode_out(dip, bh->b_data);
1715
	brelse(bh);
1716
	mark_inode_dirty(&dip->i_inode);
D
David Teigland 已提交
1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733

	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
 */

1734
int gfs2_dir_mvino(struct gfs2_inode *dip, const struct qstr *filename,
1735
		   const struct gfs2_inode *nip, unsigned int new_type)
D
David Teigland 已提交
1736
{
1737 1738
	struct buffer_head *bh;
	struct gfs2_dirent *dent;
D
David Teigland 已提交
1739 1740
	int error;

1741
	dent = gfs2_dirent_search(&dip->i_inode, filename, gfs2_dirent_find, &bh);
1742 1743 1744 1745 1746 1747
	if (!dent) {
		gfs2_consist_inode(dip);
		return -EIO;
	}
	if (IS_ERR(dent))
		return PTR_ERR(dent);
D
David Teigland 已提交
1748

1749
	gfs2_trans_add_bh(dip->i_gl, bh, 1);
1750
	gfs2_inum_out(nip, dent);
1751 1752 1753 1754 1755 1756 1757 1758 1759 1760
	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);
	}

1761
	dip->i_inode.i_mtime = dip->i_inode.i_ctime = CURRENT_TIME;
1762
	gfs2_dinode_out(dip, bh->b_data);
1763 1764
	brelse(bh);
	return 0;
D
David Teigland 已提交
1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777
}

/**
 * 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)
{
1778
	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
D
David Teigland 已提交
1779
	struct buffer_head *bh;
1780
	struct gfs2_leaf *leaf;
1781 1782 1783
	u32 hsize, len;
	u32 ht_offset, lp_offset, ht_offset_cur = -1;
	u32 index = 0;
A
Al Viro 已提交
1784
	__be64 *lp;
1785
	u64 leaf_no;
D
David Teigland 已提交
1786 1787 1788
	int error = 0;

	hsize = 1 << dip->i_di.di_depth;
1789
	if (hsize * sizeof(u64) != dip->i_di.di_size) {
D
David Teigland 已提交
1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802
		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) {
1803
			error = gfs2_dir_read_data(dip, (char *)lp,
A
Al Viro 已提交
1804
						ht_offset * sizeof(__be64),
S
Steven Whitehouse 已提交
1805
						sdp->sd_hash_bsize, 1);
D
David Teigland 已提交
1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818
			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;
1819 1820
			leaf = (struct gfs2_leaf *)bh->b_data;
			len = 1 << (dip->i_di.di_depth - be16_to_cpu(leaf->lf_depth));
1821
			brelse(bh);
D
David Teigland 已提交
1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836

			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;
	}

1837
out:
D
David Teigland 已提交
1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853
	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
 */

1854 1855
static int leaf_dealloc(struct gfs2_inode *dip, u32 index, u32 len,
			u64 leaf_no, void *data)
D
David Teigland 已提交
1856
{
1857
	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
1858
	struct gfs2_leaf *tmp_leaf;
D
David Teigland 已提交
1859 1860
	struct gfs2_rgrp_list rlist;
	struct buffer_head *bh, *dibh;
1861
	u64 blk, nblk;
D
David Teigland 已提交
1862 1863
	unsigned int rg_blocks = 0, l_blocks = 0;
	char *ht;
1864
	unsigned int x, size = len * sizeof(u64);
D
David Teigland 已提交
1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878
	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;

1879
	error = gfs2_rindex_hold(sdp, &dip->i_alloc->al_ri_gh);
D
David Teigland 已提交
1880 1881 1882 1883 1884
	if (error)
		goto out_qs;

	/*  Count the number of leaves  */

1885
	for (blk = leaf_no; blk; blk = nblk) {
D
David Teigland 已提交
1886 1887 1888
		error = get_leaf(dip, blk, &bh);
		if (error)
			goto out_rlist;
1889 1890
		tmp_leaf = (struct gfs2_leaf *)bh->b_data;
		nblk = be64_to_cpu(tmp_leaf->lf_next);
D
David Teigland 已提交
1891 1892 1893 1894 1895 1896
		brelse(bh);

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

1897
	gfs2_rlist_alloc(&rlist, LM_ST_EXCLUSIVE);
D
David Teigland 已提交
1898 1899 1900

	for (x = 0; x < rlist.rl_rgrps; x++) {
		struct gfs2_rgrpd *rgd;
1901
		rgd = rlist.rl_ghs[x].gh_gl->gl_object;
1902
		rg_blocks += rgd->rd_length;
D
David Teigland 已提交
1903 1904 1905 1906 1907 1908 1909
	}

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

	error = gfs2_trans_begin(sdp,
1910
			rg_blocks + (DIV_ROUND_UP(size, sdp->sd_jbsize) + 1) +
D
David Teigland 已提交
1911 1912 1913 1914
			RES_DINODE + RES_STATFS + RES_QUOTA, l_blocks);
	if (error)
		goto out_rg_gunlock;

1915
	for (blk = leaf_no; blk; blk = nblk) {
D
David Teigland 已提交
1916 1917 1918
		error = get_leaf(dip, blk, &bh);
		if (error)
			goto out_end_trans;
1919 1920
		tmp_leaf = (struct gfs2_leaf *)bh->b_data;
		nblk = be64_to_cpu(tmp_leaf->lf_next);
D
David Teigland 已提交
1921 1922 1923 1924 1925 1926 1927
		brelse(bh);

		gfs2_free_meta(dip, blk, 1);

		if (!dip->i_di.di_blocks)
			gfs2_consist_inode(dip);
		dip->i_di.di_blocks--;
1928
		gfs2_set_inode_blocks(&dip->i_inode);
D
David Teigland 已提交
1929 1930
	}

1931
	error = gfs2_dir_write_data(dip, ht, index * sizeof(u64), size);
D
David Teigland 已提交
1932 1933 1934 1935 1936 1937 1938 1939 1940 1941
	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;

1942
	gfs2_trans_add_bh(dip->i_gl, dibh, 1);
1943
	gfs2_dinode_out(dip, dibh->b_data);
D
David Teigland 已提交
1944 1945
	brelse(dibh);

1946
out_end_trans:
D
David Teigland 已提交
1947
	gfs2_trans_end(sdp);
1948
out_rg_gunlock:
D
David Teigland 已提交
1949
	gfs2_glock_dq_m(rlist.rl_rgrps, rlist.rl_ghs);
1950
out_rlist:
D
David Teigland 已提交
1951
	gfs2_rlist_free(&rlist);
1952
	gfs2_glock_dq_uninit(&dip->i_alloc->al_ri_gh);
1953
out_qs:
D
David Teigland 已提交
1954
	gfs2_quota_unhold(dip);
1955
out:
D
David Teigland 已提交
1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972
	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)
{
1973
	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
D
David Teigland 已提交
1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990
	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) {
1991
		gfs2_trans_add_bh(dip->i_gl, bh, 1);
1992 1993
		((struct gfs2_dinode *)bh->b_data)->di_mode =
						cpu_to_be32(S_IFREG);
D
David Teigland 已提交
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
		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
 *
2007
 * Returns: 1 if alloc required, 0 if not, -ve on error
D
David Teigland 已提交
2008 2009
 */

2010
int gfs2_diradd_alloc_required(struct inode *inode, const struct qstr *name)
D
David Teigland 已提交
2011
{
2012 2013
	struct gfs2_dirent *dent;
	struct buffer_head *bh;
D
David Teigland 已提交
2014

2015
	dent = gfs2_dirent_search(inode, name, gfs2_dirent_find_space, &bh);
2016
	if (!dent) {
2017
		return 1;
2018
	}
2019 2020 2021 2022
	if (IS_ERR(dent))
		return PTR_ERR(dent);
	brelse(bh);
	return 0;
D
David Teigland 已提交
2023 2024
}