extents.c 130 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
/*
 * Copyright (c) 2003-2006, Cluster File Systems, Inc, info@clusterfs.com
 * Written by Alex Tomas <alex@clusterfs.com>
 *
 * Architecture independence:
 *   Copyright (c) 2005, Bull S.A.
 *   Written by Pierre Peiffer <pierre.peiffer@bull.net>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public Licens
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-
 */

/*
 * Extents support for EXT4
 *
 * TODO:
 *   - ext4*_error() should be used in some situations
 *   - analyze all BUG()/BUG_ON(), use -EIO where appropriate
 *   - smart tree reduction
 */

#include <linux/fs.h>
#include <linux/time.h>
34
#include <linux/jbd2.h>
35 36 37 38 39
#include <linux/highuid.h>
#include <linux/pagemap.h>
#include <linux/quotaops.h>
#include <linux/string.h>
#include <linux/slab.h>
Amit Arora's avatar
Amit Arora committed
40
#include <linux/falloc.h>
41
#include <asm/uaccess.h>
42
#include <linux/fiemap.h>
43
#include "ext4_jbd2.h"
44
#include "ext4_extents.h"
45
#include "xattr.h"
46

47 48
#include <trace/events/ext4.h>

49 50 51 52 53 54 55 56
/*
 * used by extent splitting.
 */
#define EXT4_EXT_MAY_ZEROOUT	0x1  /* safe to zeroout if split fails \
					due to ENOSPC */
#define EXT4_EXT_MARK_UNINIT1	0x2  /* mark first half uninitialized */
#define EXT4_EXT_MARK_UNINIT2	0x4  /* mark second half uninitialized */

57 58 59
#define EXT4_EXT_DATA_VALID1	0x8  /* first half contains valid data */
#define EXT4_EXT_DATA_VALID2	0x10 /* second half contains valid data */

60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
static __le32 ext4_extent_block_csum(struct inode *inode,
				     struct ext4_extent_header *eh)
{
	struct ext4_inode_info *ei = EXT4_I(inode);
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
	__u32 csum;

	csum = ext4_chksum(sbi, ei->i_csum_seed, (__u8 *)eh,
			   EXT4_EXTENT_TAIL_OFFSET(eh));
	return cpu_to_le32(csum);
}

static int ext4_extent_block_csum_verify(struct inode *inode,
					 struct ext4_extent_header *eh)
{
	struct ext4_extent_tail *et;

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

	et = find_ext4_extent_tail(eh);
	if (et->et_checksum != ext4_extent_block_csum(inode, eh))
		return 0;
	return 1;
}

static void ext4_extent_block_csum_set(struct inode *inode,
				       struct ext4_extent_header *eh)
{
	struct ext4_extent_tail *et;

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

	et = find_ext4_extent_tail(eh);
	et->et_checksum = ext4_extent_block_csum(inode, eh);
}

Allison Henderson's avatar
Allison Henderson committed
100 101 102 103 104 105 106
static int ext4_split_extent(handle_t *handle,
				struct inode *inode,
				struct ext4_ext_path *path,
				struct ext4_map_blocks *map,
				int split_flag,
				int flags);

107 108 109 110 111 112 113
static int ext4_split_extent_at(handle_t *handle,
			     struct inode *inode,
			     struct ext4_ext_path *path,
			     ext4_lblk_t split,
			     int split_flag,
			     int flags);

114
static int ext4_find_delayed_extent(struct inode *inode,
Zheng Liu's avatar
Zheng Liu committed
115
				    struct extent_status *newes);
116

117 118 119
static int ext4_ext_truncate_extend_restart(handle_t *handle,
					    struct inode *inode,
					    int needed)
120 121 122
{
	int err;

123 124
	if (!ext4_handle_valid(handle))
		return 0;
125
	if (handle->h_buffer_credits > needed)
126 127
		return 0;
	err = ext4_journal_extend(handle, needed);
128
	if (err <= 0)
129
		return err;
130
	err = ext4_truncate_restart_trans(handle, inode, needed);
131 132
	if (err == 0)
		err = -EAGAIN;
133 134

	return err;
135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159
}

/*
 * could return:
 *  - EROFS
 *  - ENOMEM
 */
static int ext4_ext_get_access(handle_t *handle, struct inode *inode,
				struct ext4_ext_path *path)
{
	if (path->p_bh) {
		/* path points to block */
		return ext4_journal_get_write_access(handle, path->p_bh);
	}
	/* path points to leaf/index in inode body */
	/* we use in-core data, no need to protect them */
	return 0;
}

/*
 * could return:
 *  - EROFS
 *  - ENOMEM
 *  - EIO
 */
160 161 162 163 164
#define ext4_ext_dirty(handle, inode, path) \
		__ext4_ext_dirty(__func__, __LINE__, (handle), (inode), (path))
static int __ext4_ext_dirty(const char *where, unsigned int line,
			    handle_t *handle, struct inode *inode,
			    struct ext4_ext_path *path)
165 166 167
{
	int err;
	if (path->p_bh) {
168
		ext4_extent_block_csum_set(inode, ext_block_hdr(path->p_bh));
169
		/* path points to block */
170 171
		err = __ext4_handle_dirty_metadata(where, line, handle,
						   inode, path->p_bh);
172 173 174 175 176 177 178
	} else {
		/* path points to leaf/index in inode body */
		err = ext4_mark_inode_dirty(handle, inode);
	}
	return err;
}

179
static ext4_fsblk_t ext4_ext_find_goal(struct inode *inode,
180
			      struct ext4_ext_path *path,
181
			      ext4_lblk_t block)
182 183
{
	if (path) {
184
		int depth = path->p_depth;
185 186
		struct ext4_extent *ex;

187 188 189 190 191 192 193 194 195 196 197 198 199 200
		/*
		 * Try to predict block placement assuming that we are
		 * filling in a file which will eventually be
		 * non-sparse --- i.e., in the case of libbfd writing
		 * an ELF object sections out-of-order but in a way
		 * the eventually results in a contiguous object or
		 * executable file, or some database extending a table
		 * space file.  However, this is actually somewhat
		 * non-ideal if we are writing a sparse file such as
		 * qemu or KVM writing a raw image file that is going
		 * to stay fairly sparse, since it will end up
		 * fragmenting the file system's free space.  Maybe we
		 * should have some hueristics or some way to allow
		 * userspace to pass a hint to file system,
201
		 * especially if the latter case turns out to be
202 203
		 * common.
		 */
204
		ex = path[depth].p_ext;
205 206 207 208 209 210 211 212 213
		if (ex) {
			ext4_fsblk_t ext_pblk = ext4_ext_pblock(ex);
			ext4_lblk_t ext_block = le32_to_cpu(ex->ee_block);

			if (block > ext_block)
				return ext_pblk + (block - ext_block);
			else
				return ext_pblk - (ext_block - block);
		}
214

215 216
		/* it looks like index is empty;
		 * try to find starting block from index itself */
217 218 219 220 221
		if (path[depth].p_bh)
			return path[depth].p_bh->b_blocknr;
	}

	/* OK. use inode's group */
222
	return ext4_inode_to_goal_block(inode);
223 224
}

225 226 227
/*
 * Allocation for a meta data block
 */
228
static ext4_fsblk_t
229
ext4_ext_new_meta_block(handle_t *handle, struct inode *inode,
230
			struct ext4_ext_path *path,
231
			struct ext4_extent *ex, int *err, unsigned int flags)
232
{
233
	ext4_fsblk_t goal, newblock;
234 235

	goal = ext4_ext_find_goal(inode, path, le32_to_cpu(ex->ee_block));
236 237
	newblock = ext4_new_meta_blocks(handle, inode, goal, flags,
					NULL, err);
238 239 240
	return newblock;
}

241
static inline int ext4_ext_space_block(struct inode *inode, int check)
242 243 244 245 246
{
	int size;

	size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
			/ sizeof(struct ext4_extent);
247
#ifdef AGGRESSIVE_TEST
248 249
	if (!check && size > 6)
		size = 6;
250 251 252 253
#endif
	return size;
}

254
static inline int ext4_ext_space_block_idx(struct inode *inode, int check)
255 256 257 258 259
{
	int size;

	size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
			/ sizeof(struct ext4_extent_idx);
260
#ifdef AGGRESSIVE_TEST
261 262
	if (!check && size > 5)
		size = 5;
263 264 265 266
#endif
	return size;
}

267
static inline int ext4_ext_space_root(struct inode *inode, int check)
268 269 270 271 272 273
{
	int size;

	size = sizeof(EXT4_I(inode)->i_data);
	size -= sizeof(struct ext4_extent_header);
	size /= sizeof(struct ext4_extent);
274
#ifdef AGGRESSIVE_TEST
275 276
	if (!check && size > 3)
		size = 3;
277 278 279 280
#endif
	return size;
}

281
static inline int ext4_ext_space_root_idx(struct inode *inode, int check)
282 283 284 285 286 287
{
	int size;

	size = sizeof(EXT4_I(inode)->i_data);
	size -= sizeof(struct ext4_extent_header);
	size /= sizeof(struct ext4_extent_idx);
288
#ifdef AGGRESSIVE_TEST
289 290
	if (!check && size > 4)
		size = 4;
291 292 293 294
#endif
	return size;
}

295 296 297 298 299
/*
 * Calculate the number of metadata blocks needed
 * to allocate @blocks
 * Worse case is one block per extent
 */
300
int ext4_ext_calc_metadata_amount(struct inode *inode, ext4_lblk_t lblock)
301
{
302
	struct ext4_inode_info *ei = EXT4_I(inode);
303
	int idxs;
304

305 306
	idxs = ((inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
		/ sizeof(struct ext4_extent_idx));
307 308

	/*
309 310 311 312 313 314
	 * If the new delayed allocation block is contiguous with the
	 * previous da block, it can share index blocks with the
	 * previous block, so we only need to allocate a new index
	 * block every idxs leaf blocks.  At ldxs**2 blocks, we need
	 * an additional index block, and at ldxs**3 blocks, yet
	 * another index blocks.
315
	 */
316 317
	if (ei->i_da_metadata_calc_len &&
	    ei->i_da_metadata_calc_last_lblock+1 == lblock) {
318 319
		int num = 0;

320 321 322 323 324 325 326 327 328 329 330 331
		if ((ei->i_da_metadata_calc_len % idxs) == 0)
			num++;
		if ((ei->i_da_metadata_calc_len % (idxs*idxs)) == 0)
			num++;
		if ((ei->i_da_metadata_calc_len % (idxs*idxs*idxs)) == 0) {
			num++;
			ei->i_da_metadata_calc_len = 0;
		} else
			ei->i_da_metadata_calc_len++;
		ei->i_da_metadata_calc_last_lblock++;
		return num;
	}
332

333 334 335 336 337 338 339
	/*
	 * In the worst case we need a new set of index blocks at
	 * every level of the inode's extent tree.
	 */
	ei->i_da_metadata_calc_len = 1;
	ei->i_da_metadata_calc_last_lblock = lblock;
	return ext_depth(inode) + 1;
340 341
}

342 343 344 345 346 347 348
static int
ext4_ext_max_entries(struct inode *inode, int depth)
{
	int max;

	if (depth == ext_depth(inode)) {
		if (depth == 0)
349
			max = ext4_ext_space_root(inode, 1);
350
		else
351
			max = ext4_ext_space_root_idx(inode, 1);
352 353
	} else {
		if (depth == 0)
354
			max = ext4_ext_space_block(inode, 1);
355
		else
356
			max = ext4_ext_space_block_idx(inode, 1);
357 358 359 360 361
	}

	return max;
}

362 363
static int ext4_valid_extent(struct inode *inode, struct ext4_extent *ext)
{
364
	ext4_fsblk_t block = ext4_ext_pblock(ext);
365
	int len = ext4_ext_get_actual_len(ext);
366

367 368
	if (len == 0)
		return 0;
369
	return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, len);
370 371 372 373 374
}

static int ext4_valid_extent_idx(struct inode *inode,
				struct ext4_extent_idx *ext_idx)
{
375
	ext4_fsblk_t block = ext4_idx_pblock(ext_idx);
376

377
	return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, 1);
378 379 380 381 382 383 384 385 386 387 388 389 390 391
}

static int ext4_valid_extent_entries(struct inode *inode,
				struct ext4_extent_header *eh,
				int depth)
{
	unsigned short entries;
	if (eh->eh_entries == 0)
		return 1;

	entries = le16_to_cpu(eh->eh_entries);

	if (depth == 0) {
		/* leaf entries */
392
		struct ext4_extent *ext = EXT_FIRST_EXTENT(eh);
393 394 395 396 397 398 399
		while (entries) {
			if (!ext4_valid_extent(inode, ext))
				return 0;
			ext++;
			entries--;
		}
	} else {
400
		struct ext4_extent_idx *ext_idx = EXT_FIRST_INDEX(eh);
401 402 403 404 405 406 407 408 409 410
		while (entries) {
			if (!ext4_valid_extent_idx(inode, ext_idx))
				return 0;
			ext_idx++;
			entries--;
		}
	}
	return 1;
}

411 412 413
static int __ext4_ext_check(const char *function, unsigned int line,
			    struct inode *inode, struct ext4_extent_header *eh,
			    int depth)
414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438
{
	const char *error_msg;
	int max = 0;

	if (unlikely(eh->eh_magic != EXT4_EXT_MAGIC)) {
		error_msg = "invalid magic";
		goto corrupted;
	}
	if (unlikely(le16_to_cpu(eh->eh_depth) != depth)) {
		error_msg = "unexpected eh_depth";
		goto corrupted;
	}
	if (unlikely(eh->eh_max == 0)) {
		error_msg = "invalid eh_max";
		goto corrupted;
	}
	max = ext4_ext_max_entries(inode, depth);
	if (unlikely(le16_to_cpu(eh->eh_max) > max)) {
		error_msg = "too large eh_max";
		goto corrupted;
	}
	if (unlikely(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max))) {
		error_msg = "invalid eh_entries";
		goto corrupted;
	}
439 440 441 442
	if (!ext4_valid_extent_entries(inode, eh, depth)) {
		error_msg = "invalid extent entries";
		goto corrupted;
	}
443 444 445 446 447 448
	/* Verify checksum on non-root extent tree nodes */
	if (ext_depth(inode) != depth &&
	    !ext4_extent_block_csum_verify(inode, eh)) {
		error_msg = "extent tree corrupted";
		goto corrupted;
	}
449 450 451
	return 0;

corrupted:
452
	ext4_error_inode(inode, function, line, 0,
453
			"bad header/extent: %s - magic %x, "
454
			"entries %u, max %u(%u), depth %u(%u)",
455
			error_msg, le16_to_cpu(eh->eh_magic),
456 457 458 459 460 461
			le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max),
			max, le16_to_cpu(eh->eh_depth), depth);

	return -EIO;
}

462
#define ext4_ext_check(inode, eh, depth)	\
463
	__ext4_ext_check(__func__, __LINE__, inode, eh, depth)
464

465 466 467 468 469
int ext4_ext_check_inode(struct inode *inode)
{
	return ext4_ext_check(inode, ext_inode_hdr(inode), ext_depth(inode));
}

470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489
static int __ext4_ext_check_block(const char *function, unsigned int line,
				  struct inode *inode,
				  struct ext4_extent_header *eh,
				  int depth,
				  struct buffer_head *bh)
{
	int ret;

	if (buffer_verified(bh))
		return 0;
	ret = ext4_ext_check(inode, eh, depth);
	if (ret)
		return ret;
	set_buffer_verified(bh);
	return ret;
}

#define ext4_ext_check_block(inode, eh, depth, bh)	\
	__ext4_ext_check_block(__func__, __LINE__, inode, eh, depth, bh)

490 491 492 493 494 495 496 497
#ifdef EXT_DEBUG
static void ext4_ext_show_path(struct inode *inode, struct ext4_ext_path *path)
{
	int k, l = path->p_depth;

	ext_debug("path:");
	for (k = 0; k <= l; k++, path++) {
		if (path->p_idx) {
498
		  ext_debug("  %d->%llu", le32_to_cpu(path->p_idx->ei_block),
499
			    ext4_idx_pblock(path->p_idx));
500
		} else if (path->p_ext) {
501
			ext_debug("  %d:[%d]%d:%llu ",
502
				  le32_to_cpu(path->p_ext->ee_block),
503
				  ext4_ext_is_uninitialized(path->p_ext),
Amit Arora's avatar
Amit Arora committed
504
				  ext4_ext_get_actual_len(path->p_ext),
505
				  ext4_ext_pblock(path->p_ext));
506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524
		} else
			ext_debug("  []");
	}
	ext_debug("\n");
}

static void ext4_ext_show_leaf(struct inode *inode, struct ext4_ext_path *path)
{
	int depth = ext_depth(inode);
	struct ext4_extent_header *eh;
	struct ext4_extent *ex;
	int i;

	if (!path)
		return;

	eh = path[depth].p_hdr;
	ex = EXT_FIRST_EXTENT(eh);

525 526
	ext_debug("Displaying leaf extents for inode %lu\n", inode->i_ino);

527
	for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ex++) {
528 529
		ext_debug("%d:[%d]%d:%llu ", le32_to_cpu(ex->ee_block),
			  ext4_ext_is_uninitialized(ex),
530
			  ext4_ext_get_actual_len(ex), ext4_ext_pblock(ex));
531 532 533
	}
	ext_debug("\n");
}
534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566

static void ext4_ext_show_move(struct inode *inode, struct ext4_ext_path *path,
			ext4_fsblk_t newblock, int level)
{
	int depth = ext_depth(inode);
	struct ext4_extent *ex;

	if (depth != level) {
		struct ext4_extent_idx *idx;
		idx = path[level].p_idx;
		while (idx <= EXT_MAX_INDEX(path[level].p_hdr)) {
			ext_debug("%d: move %d:%llu in new index %llu\n", level,
					le32_to_cpu(idx->ei_block),
					ext4_idx_pblock(idx),
					newblock);
			idx++;
		}

		return;
	}

	ex = path[depth].p_ext;
	while (ex <= EXT_MAX_EXTENT(path[depth].p_hdr)) {
		ext_debug("move %d:%llu:[%d]%d in new leaf %llu\n",
				le32_to_cpu(ex->ee_block),
				ext4_ext_pblock(ex),
				ext4_ext_is_uninitialized(ex),
				ext4_ext_get_actual_len(ex),
				newblock);
		ex++;
	}
}

567
#else
568 569
#define ext4_ext_show_path(inode, path)
#define ext4_ext_show_leaf(inode, path)
570
#define ext4_ext_show_move(inode, path, newblock, level)
571 572
#endif

573
void ext4_ext_drop_refs(struct ext4_ext_path *path)
574 575 576 577 578 579 580 581 582 583 584 585
{
	int depth = path->p_depth;
	int i;

	for (i = 0; i <= depth; i++, path++)
		if (path->p_bh) {
			brelse(path->p_bh);
			path->p_bh = NULL;
		}
}

/*
586 587
 * ext4_ext_binsearch_idx:
 * binary search for the closest index of the given block
588
 * the header must be checked before calling this
589 590
 */
static void
591 592
ext4_ext_binsearch_idx(struct inode *inode,
			struct ext4_ext_path *path, ext4_lblk_t block)
593 594 595 596 597
{
	struct ext4_extent_header *eh = path->p_hdr;
	struct ext4_extent_idx *r, *l, *m;


598
	ext_debug("binsearch for %u(idx):  ", block);
599 600

	l = EXT_FIRST_INDEX(eh) + 1;
601
	r = EXT_LAST_INDEX(eh);
602 603 604 605 606 607
	while (l <= r) {
		m = l + (r - l) / 2;
		if (block < le32_to_cpu(m->ei_block))
			r = m - 1;
		else
			l = m + 1;
608 609 610
		ext_debug("%p(%u):%p(%u):%p(%u) ", l, le32_to_cpu(l->ei_block),
				m, le32_to_cpu(m->ei_block),
				r, le32_to_cpu(r->ei_block));
611 612 613
	}

	path->p_idx = l - 1;
614
	ext_debug("  -> %u->%lld ", le32_to_cpu(path->p_idx->ei_block),
615
		  ext4_idx_pblock(path->p_idx));
616 617 618 619 620 621 622 623 624 625

#ifdef CHECK_BINSEARCH
	{
		struct ext4_extent_idx *chix, *ix;
		int k;

		chix = ix = EXT_FIRST_INDEX(eh);
		for (k = 0; k < le16_to_cpu(eh->eh_entries); k++, ix++) {
		  if (k != 0 &&
		      le32_to_cpu(ix->ei_block) <= le32_to_cpu(ix[-1].ei_block)) {
626 627 628 629
				printk(KERN_DEBUG "k=%d, ix=0x%p, "
				       "first=0x%p\n", k,
				       ix, EXT_FIRST_INDEX(eh));
				printk(KERN_DEBUG "%u <= %u\n",
630 631 632 633
				       le32_to_cpu(ix->ei_block),
				       le32_to_cpu(ix[-1].ei_block));
			}
			BUG_ON(k && le32_to_cpu(ix->ei_block)
Dave Kleikamp's avatar
Dave Kleikamp committed
634
					   <= le32_to_cpu(ix[-1].ei_block));
635 636 637 638 639 640 641 642 643 644 645
			if (block < le32_to_cpu(ix->ei_block))
				break;
			chix = ix;
		}
		BUG_ON(chix != path->p_idx);
	}
#endif

}

/*
646 647
 * ext4_ext_binsearch:
 * binary search for closest extent of the given block
648
 * the header must be checked before calling this
649 650
 */
static void
651 652
ext4_ext_binsearch(struct inode *inode,
		struct ext4_ext_path *path, ext4_lblk_t block)
653 654 655 656 657 658
{
	struct ext4_extent_header *eh = path->p_hdr;
	struct ext4_extent *r, *l, *m;

	if (eh->eh_entries == 0) {
		/*
659 660
		 * this leaf is empty:
		 * we get such a leaf in split/add case
661 662 663 664
		 */
		return;
	}

665
	ext_debug("binsearch for %u:  ", block);
666 667

	l = EXT_FIRST_EXTENT(eh) + 1;
668
	r = EXT_LAST_EXTENT(eh);
669 670 671 672 673 674 675

	while (l <= r) {
		m = l + (r - l) / 2;
		if (block < le32_to_cpu(m->ee_block))
			r = m - 1;
		else
			l = m + 1;
676 677 678
		ext_debug("%p(%u):%p(%u):%p(%u) ", l, le32_to_cpu(l->ee_block),
				m, le32_to_cpu(m->ee_block),
				r, le32_to_cpu(r->ee_block));
679 680 681
	}

	path->p_ext = l - 1;
682
	ext_debug("  -> %d:%llu:[%d]%d ",
Dave Kleikamp's avatar
Dave Kleikamp committed
683
			le32_to_cpu(path->p_ext->ee_block),
684
			ext4_ext_pblock(path->p_ext),
685
			ext4_ext_is_uninitialized(path->p_ext),
Amit Arora's avatar
Amit Arora committed
686
			ext4_ext_get_actual_len(path->p_ext));
687 688 689 690 691 692 693 694 695

#ifdef CHECK_BINSEARCH
	{
		struct ext4_extent *chex, *ex;
		int k;

		chex = ex = EXT_FIRST_EXTENT(eh);
		for (k = 0; k < le16_to_cpu(eh->eh_entries); k++, ex++) {
			BUG_ON(k && le32_to_cpu(ex->ee_block)
Dave Kleikamp's avatar
Dave Kleikamp committed
696
					  <= le32_to_cpu(ex[-1].ee_block));
697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714
			if (block < le32_to_cpu(ex->ee_block))
				break;
			chex = ex;
		}
		BUG_ON(chex != path->p_ext);
	}
#endif

}

int ext4_ext_tree_init(handle_t *handle, struct inode *inode)
{
	struct ext4_extent_header *eh;

	eh = ext_inode_hdr(inode);
	eh->eh_depth = 0;
	eh->eh_entries = 0;
	eh->eh_magic = EXT4_EXT_MAGIC;
715
	eh->eh_max = cpu_to_le16(ext4_ext_space_root(inode, 0));
716 717 718 719 720
	ext4_mark_inode_dirty(handle, inode);
	return 0;
}

struct ext4_ext_path *
721 722
ext4_ext_find_extent(struct inode *inode, ext4_lblk_t block,
					struct ext4_ext_path *path)
723 724 725 726
{
	struct ext4_extent_header *eh;
	struct buffer_head *bh;
	short int depth, i, ppos = 0, alloc = 0;
727
	int ret;
728 729

	eh = ext_inode_hdr(inode);
730
	depth = ext_depth(inode);
731 732 733

	/* account possible depth increase */
	if (!path) {
734
		path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 2),
735 736 737 738 739 740
				GFP_NOFS);
		if (!path)
			return ERR_PTR(-ENOMEM);
		alloc = 1;
	}
	path[0].p_hdr = eh;
741
	path[0].p_bh = NULL;
742

743
	i = depth;
744 745 746 747
	/* walk through the tree */
	while (i) {
		ext_debug("depth %d: num %d, max %d\n",
			  ppos, le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max));
748

749
		ext4_ext_binsearch_idx(inode, path + ppos, block);
750
		path[ppos].p_block = ext4_idx_pblock(path[ppos].p_idx);
751 752 753
		path[ppos].p_depth = i;
		path[ppos].p_ext = NULL;

754
		bh = sb_getblk(inode->i_sb, path[ppos].p_block);
755 756
		if (unlikely(!bh)) {
			ret = -ENOMEM;
757
			goto err;
758
		}
759
		if (!bh_uptodate_or_lock(bh)) {
760 761
			trace_ext4_ext_load_extent(inode, block,
						path[ppos].p_block);
762 763
			ret = bh_submit_read(bh);
			if (ret < 0) {
764 765 766 767
				put_bh(bh);
				goto err;
			}
		}
768 769
		eh = ext_block_hdr(bh);
		ppos++;
770 771 772 773
		if (unlikely(ppos > depth)) {
			put_bh(bh);
			EXT4_ERROR_INODE(inode,
					 "ppos %d > depth %d", ppos, depth);
774
			ret = -EIO;
775 776
			goto err;
		}
777 778 779 780
		path[ppos].p_bh = bh;
		path[ppos].p_hdr = eh;
		i--;

781 782
		ret = ext4_ext_check_block(inode, eh, i, bh);
		if (ret < 0)
783 784 785 786 787 788 789 790 791
			goto err;
	}

	path[ppos].p_depth = i;
	path[ppos].p_ext = NULL;
	path[ppos].p_idx = NULL;

	/* find extent */
	ext4_ext_binsearch(inode, path + ppos, block);
792 793
	/* if not an empty leaf */
	if (path[ppos].p_ext)
794
		path[ppos].p_block = ext4_ext_pblock(path[ppos].p_ext);
795 796 797 798 799 800 801 802 803

	ext4_ext_show_path(inode, path);

	return path;

err:
	ext4_ext_drop_refs(path);
	if (alloc)
		kfree(path);
804
	return ERR_PTR(ret);
805 806 807
}

/*
808 809 810
 * ext4_ext_insert_index:
 * insert new index [@logical;@ptr] into the block at @curp;
 * check where to insert: before @curp or after @curp
811
 */
812 813 814
static int ext4_ext_insert_index(handle_t *handle, struct inode *inode,
				 struct ext4_ext_path *curp,
				 int logical, ext4_fsblk_t ptr)
815 816 817 818
{
	struct ext4_extent_idx *ix;
	int len, err;

819 820
	err = ext4_ext_get_access(handle, inode, curp);
	if (err)
821 822
		return err;

823 824 825 826 827 828
	if (unlikely(logical == le32_to_cpu(curp->p_idx->ei_block))) {
		EXT4_ERROR_INODE(inode,
				 "logical %d == ei_block %d!",
				 logical, le32_to_cpu(curp->p_idx->ei_block));
		return -EIO;
	}
829 830 831 832 833 834 835 836 837 838

	if (unlikely(le16_to_cpu(curp->p_hdr->eh_entries)
			     >= le16_to_cpu(curp->p_hdr->eh_max))) {
		EXT4_ERROR_INODE(inode,
				 "eh_entries %d >= eh_max %d!",
				 le16_to_cpu(curp->p_hdr->eh_entries),
				 le16_to_cpu(curp->p_hdr->eh_max));
		return -EIO;
	}

839 840
	if (logical > le32_to_cpu(curp->p_idx->ei_block)) {
		/* insert after */
841
		ext_debug("insert new index %d after: %llu\n", logical, ptr);
842 843 844
		ix = curp->p_idx + 1;
	} else {
		/* insert before */
845
		ext_debug("insert new index %d before: %llu\n", logical, ptr);
846 847 848
		ix = curp->p_idx;
	}

849 850 851 852 853 854 855 856 857
	len = EXT_LAST_INDEX(curp->p_hdr) - ix + 1;
	BUG_ON(len < 0);
	if (len > 0) {
		ext_debug("insert new index %d: "
				"move %d indices from 0x%p to 0x%p\n",
				logical, len, ix, ix + 1);
		memmove(ix + 1, ix, len * sizeof(struct ext4_extent_idx));
	}

858 859 860 861 862
	if (unlikely(ix > EXT_MAX_INDEX(curp->p_hdr))) {
		EXT4_ERROR_INODE(inode, "ix > EXT_MAX_INDEX!");
		return -EIO;
	}

863
	ix->ei_block = cpu_to_le32(logical);
864
	ext4_idx_store_pblock(ix, ptr);
865
	le16_add_cpu(&curp->p_hdr->eh_entries, 1);
866

867 868 869 870
	if (unlikely(ix > EXT_LAST_INDEX(curp->p_hdr))) {
		EXT4_ERROR_INODE(inode, "ix > EXT_LAST_INDEX!");
		return -EIO;
	}
871 872 873 874 875 876 877 878

	err = ext4_ext_dirty(handle, inode, curp);
	ext4_std_error(inode->i_sb, err);

	return err;
}

/*
879 880 881 882 883 884 885 886
 * ext4_ext_split:
 * inserts new subtree into the path, using free index entry
 * at depth @at:
 * - allocates all needed blocks (new leaf and all intermediate index blocks)
 * - makes decision where to split
 * - moves remaining extents and index entries (right to the split point)
 *   into the newly allocated blocks
 * - initializes subtree
887 888
 */
static int ext4_ext_split(handle_t *handle, struct inode *inode,
889 890 891
			  unsigned int flags,
			  struct ext4_ext_path *path,
			  struct ext4_extent *newext, int at)
892 893 894 895 896 897
{
	struct buffer_head *bh = NULL;
	int depth = ext_depth(inode);
	struct ext4_extent_header *neh;
	struct ext4_extent_idx *fidx;
	int i = at, k, m, a;
898
	ext4_fsblk_t newblock, oldblock;
899
	__le32 border;
900
	ext4_fsblk_t *ablocks = NULL; /* array of allocated blocks */
901 902 903
	int err = 0;

	/* make decision: where to split? */
904
	/* FIXME: now decision is simplest: at current extent */
905

906
	/* if current leaf will be split, then we should use
907
	 * border from split point */
908 909 910 911
	if (unlikely(path[depth].p_ext > EXT_MAX_EXTENT(path[depth].p_hdr))) {
		EXT4_ERROR_INODE(inode, "p_ext > EXT_MAX_EXTENT!");
		return -EIO;
	}
912 913
	if (path[depth].p_ext != EXT_MAX_EXTENT(path[depth].p_hdr)) {
		border = path[depth].p_ext[1].ee_block;
914
		ext_debug("leaf will be split."
915
				" next leaf starts at %d\n",
Dave Kleikamp's avatar
Dave Kleikamp committed
916
				  le32_to_cpu(border));
917 918 919 920
	} else {
		border = newext->ee_block;
		ext_debug("leaf will be added."
				" next leaf starts at %d\n",
Dave Kleikamp's avatar
Dave Kleikamp committed
921
				le32_to_cpu(border));
922 923 924
	}

	/*
925 926
	 * If error occurs, then we break processing
	 * and mark filesystem read-only. index won't
927
	 * be inserted and tree will be in consistent
928
	 * state. Next mount will repair buffers too.
929 930 931
	 */

	/*
932 933 934
	 * Get array to track all allocated blocks.
	 * We need this to handle errors and free blocks
	 * upon them.
935
	 */
936
	ablocks = kzalloc(sizeof(ext4_fsblk_t) * depth, GFP_NOFS);
937 938 939 940 941 942
	if (!ablocks)
		return -ENOMEM;

	/* allocate all needed blocks */
	ext_debug("allocate %d blocks for indexes/leaf\n", depth - at);
	for (a = 0; a < depth - at; a++) {
943
		newblock = ext4_ext_new_meta_block(handle, inode, path,
944
						   newext, &err, flags);
945 946 947 948 949 950 951
		if (newblock == 0)
			goto cleanup;
		ablocks[a] = newblock;
	}

	/* initialize new leaf */
	newblock = ablocks[--a];
952 953 954 955 956
	if (unlikely(newblock == 0)) {
		EXT4_ERROR_INODE(inode, "newblock == 0!");
		err = -EIO;
		goto cleanup;
	}
957
	bh = sb_getblk(inode->i_sb, newblock);
958
	if (unlikely(!bh)) {
959
		err = -ENOMEM;
960 961 962 963
		goto cleanup;
	}
	lock_buffer(bh);

964 965
	err = ext4_journal_get_create_access(handle, bh);
	if (err)
966 967 968 969
		goto cleanup;

	neh = ext_block_hdr(bh);
	neh->eh_entries = 0;
970
	neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode, 0));
971 972 973
	neh->eh_magic = EXT4_EXT_MAGIC;
	neh->eh_depth = 0;

974
	/* move remainder of path[depth] to the new leaf */
975 976 977 978 979 980 981 982
	if (unlikely(path[depth].p_hdr->eh_entries !=
		     path[depth].p_hdr->eh_max)) {
		EXT4_ERROR_INODE(inode, "eh_entries %d != eh_max %d!",
				 path[depth].p_hdr->eh_entries,
				 path[depth].p_hdr->eh_max);
		err = -EIO;
		goto cleanup;
	}
983
	/* start copy from next extent */
984 985
	m = EXT_MAX_EXTENT(path[depth].p_hdr) - path[depth].p_ext++;
	ext4_ext_show_move(inode, path, newblock, depth);
986
	if (m) {
987 988 989
		struct ext4_extent *ex;
		ex = EXT_FIRST_EXTENT(neh);
		memmove(ex, path[depth].p_ext, sizeof(struct ext4_extent) * m);
990
		le16_add_cpu(&neh->eh_entries, m);
991 992
	}

993
	ext4_extent_block_csum_set(inode, neh);
994 995 996
	set_buffer_uptodate(bh);
	unlock_buffer(bh);

997
	err = ext4_handle_dirty_metadata(handle, inode, bh);
998
	if (err)
999 1000 1001 1002 1003 1004
		goto cleanup;
	brelse(bh);
	bh = NULL;

	/* correct old leaf */
	if (m) {
1005 1006
		err = ext4_ext_get_access(handle, inode, path + depth);
		if (err)
1007
			goto cleanup;
1008
		le16_add_cpu(&path[depth].p_hdr->eh_entries, -m);
1009 1010
		err = ext4_ext_dirty(handle, inode, path + depth);
		if (err)
1011 1012 1013 1014 1015 1016
			goto cleanup;

	}

	/* create intermediate indexes */
	k = depth - at - 1;
1017 1018 1019 1020 1021
	if (unlikely(k < 0)) {
		EXT4_ERROR_INODE(inode, "k %d < 0!", k);
		err = -EIO;
		goto cleanup;
	}
1022 1023 1024 1025 1026 1027 1028 1029
	if (k)
		ext_debug("create %d intermediate indices\n", k);
	/* insert new index into current index block */
	/* current depth stored in i var */
	i = depth - 1;
	while (k--) {
		oldblock = newblock;
		newblock = ablocks[--a];
1030
		bh = sb_getblk(inode->i_sb, newblock);
1031
		if (unlikely(!bh)) {
1032
			err = -ENOMEM;
1033 1034 1035 1036
			goto cleanup;
		}
		lock_buffer(bh);

1037 1038
		err = ext4_journal_get_create_access(handle, bh);
		if (err)
1039 1040 1041 1042 1043
			goto cleanup;

		neh = ext_block_hdr(bh);
		neh->eh_entries = cpu_to_le16(1);
		neh->eh_magic = EXT4_EXT_MAGIC;
1044
		neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode, 0));
1045 1046 1047
		neh->eh_depth = cpu_to_le16(depth - i);
		fidx = EXT_FIRST_INDEX(neh);
		fidx->ei_block = border;
1048
		ext4_idx_store_pblock(fidx, oldblock);
1049

1050 1051
		ext_debug("int.index at %d (block %llu): %u -> %llu\n",
				i, newblock, le32_to_cpu(border), oldblock);
1052

1053
		/* move remainder of path[i] to the new index block */
1054 1055 1056 1057 1058 1059 1060 1061
		if (unlikely(EXT_MAX_INDEX(path[i].p_hdr) !=
					EXT_LAST_INDEX(path[i].p_hdr))) {
			EXT4_ERROR_INODE(inode,
					 "EXT_MAX_INDEX != EXT_LAST_INDEX ee_block %d!",
					 le32_to_cpu(path[i].p_ext->ee_block));
			err = -EIO;
			goto cleanup;
		}
1062 1063 1064 1065 1066
		/* start copy indexes */
		m = EXT_MAX_INDEX(path[i].p_hdr) - path[i].p_idx++;
		ext_debug("cur 0x%p, last 0x%p\n", path[i].p_idx,
				EXT_MAX_INDEX(path[i].p_hdr));
		ext4_ext_show_move(inode, path, newblock, i);
1067
		if (m) {
1068
			memmove(++fidx, path[i].p_idx,
1069
				sizeof(struct ext4_extent_idx) * m);
1070
			le16_add_cpu(&neh->eh_entries, m);
1071
		}
1072
		ext4_extent_block_csum_set(inode, neh);
1073 1074 1075
		set_buffer_uptodate(bh);
		unlock_buffer(bh);

1076
		err = ext4_handle_dirty_metadata(handle, inode, bh);
1077
		if (err)
1078 1079 1080 1081 1082 1083 1084 1085 1086
			goto cleanup;
		brelse(bh);
		bh = NULL;

		/* correct old index */
		if (m) {
			err = ext4_ext_get_access(handle, inode, path + i);
			if (err)
				goto cleanup;
1087
			le16_add_cpu(&path[i].p_hdr->eh_entries, -m);
1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111
			err = ext4_ext_dirty(handle, inode, path + i);
			if (err)
				goto cleanup;
		}

		i--;
	}

	/* insert new index */
	err = ext4_ext_insert_index(handle, inode, path + at,
				    le32_to_cpu(border), newblock);

cleanup:
	if (bh) {
		if (buffer_locked(bh))
			unlock_buffer(bh);
		brelse(bh);
	}

	if (err) {
		/* free all allocated blocks in error case */
		for (i = 0; i < depth; i++) {
			if (!ablocks[i])
				continue;
1112
			ext4_free_blocks(handle, inode, NULL, ablocks[i], 1,
1113
					 EXT4_FREE_BLOCKS_METADATA);
1114 1115 1116 1117 1118 1119 1120 1121
		}
	}
	kfree(ablocks);

	return err;
}

/*
1122 1123 1124 1125 1126 1127
 * ext4_ext_grow_indepth:
 * implements tree growing procedure:
 * - allocates new block
 * - moves top-level data (index block or leaf) into the new block
 * - initializes new top-level, creating index that points to the
 *   just created block
1128 1129
 */
static int ext4_ext_grow_indepth(handle_t *handle, struct inode *inode,
1130 1131
				 unsigned int flags,
				 struct ext4_extent *newext)
1132 1133 1134
{
	struct ext4_extent_header *neh;
	struct buffer_head *bh;
1135
	ext4_fsblk_t newblock;
1136 1137
	int err = 0;

1138
	newblock = ext4_ext_new_meta_block(handle, inode, NULL,
1139
		newext, &err, flags);
1140 1141 1142 1143
	if (newblock == 0)
		return err;

	bh = sb_getblk(inode->i_sb, newblock);
1144
	if (unlikely(!bh))
1145
		return -ENOMEM;
1146 1147
	lock_buffer(bh);

1148 1149
	err = ext4_journal_get_create_access(handle, bh);
	if (err) {
1150 1151 1152 1153 1154
		unlock_buffer(bh);
		goto out;
	}

	/* move top-level index/leaf into new block */
1155 1156
	memmove(bh->b_data, EXT4_I(inode)->i_data,
		sizeof(EXT4_I(inode)->i_data));
1157 1158 1159 1160 1161 1162

	/* set size of new block */
	neh = ext_block_hdr(bh);
	/* old root could have indexes or leaves
	 * so calculate e_max right way */
	if (ext_depth(inode))
1163
		neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode, 0));
1164
	else
1165
		neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode, 0));
1166
	neh->eh_magic = EXT4_EXT_MAGIC;
1167
	ext4_extent_block_csum_set(inode, neh);
1168 1169 1170
	set_buffer_uptodate(bh);
	unlock_buffer(bh);

1171
	err = ext4_handle_dirty_metadata(handle, inode, bh);
1172
	if (err)
1173 1174
		goto out;

1175
	/* Update top-level index: num,max,pointer */
1176
	neh = ext_inode_hdr(inode);
1177 1178 1179 1180 1181 1182 1183 1184
	neh->eh_entries = cpu_to_le16(1);
	ext4_idx_store_pblock(EXT_FIRST_INDEX(neh), newblock);
	if (neh->eh_depth == 0) {
		/* Root extent block becomes index block */
		neh->eh_max = cpu_to_le16(ext4_ext_space_root_idx(inode, 0));
		EXT_FIRST_INDEX(neh)->ei_block =
			EXT_FIRST_EXTENT(neh)->ee_block;
	}
1185
	ext_debug("new root: num %d(%d), lblock %d, ptr %llu\n",
1186
		  le16_to_cpu(neh->eh_entries), le16_to_cpu(neh->eh_max),
1187
		  le32_to_cpu(EXT_FIRST_INDEX(neh)->ei_block),
1188
		  ext4_idx_pblock(EXT_FIRST_INDEX(neh)));
1189

1190
	le16_add_cpu(&neh->eh_depth, 1);
1191
	ext4_mark_inode_dirty(handle, inode);
1192 1193 1194 1195 1196 1197 1198
out:
	brelse(bh);

	return err;
}

/*
1199 1200 1201
 * ext4_ext_create_new_leaf:
 * finds empty index and adds new leaf.
 * if no free index is found, then it requests in-depth growing.
1202 1203
 */
static int ext4_ext_create_new_leaf(handle_t *handle, struct inode *inode,
1204 1205 1206
				    unsigned int flags,
				    struct ext4_ext_path *path,
				    struct ext4_extent *newext)
1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220
{
	struct ext4_ext_path *curp;
	int depth, i, err = 0;

repeat:
	i = depth = ext_depth(inode);

	/* walk up to the tree and look for free index entry */
	curp = path + depth;
	while (i > 0 && !EXT_HAS_FREE_INDEX(curp)) {
		i--;
		curp--;
	}

1221 1222
	/* we use already allocated block for index block,
	 * so subsequent data blocks should be contiguous */
1223 1224 1225
	if (EXT_HAS_FREE_INDEX(curp)) {
		/* if we found index with free entry, then use that
		 * entry: create all needed subtree and add new leaf */
1226
		err = ext4_ext_split(handle, inode, flags, path, newext, i);
1227 1228
		if (err)
			goto out;
1229 1230 1231 1232

		/* refill path */
		ext4_ext_drop_refs(path);
		path = ext4_ext_find_extent(inode,
1233 1234
				    (ext4_lblk_t)le32_to_cpu(newext->ee_block),
				    path);
1235 1236 1237 1238
		if (IS_ERR(path))
			err = PTR_ERR(path);
	} else {
		/* tree is full, time to grow in depth */
1239
		err = ext4_ext_grow_indepth(handle, inode, flags, newext);
1240 1241 1242 1243 1244 1245
		if (err)
			goto out;

		/* refill path */
		ext4_ext_drop_refs(path);
		path = ext4_ext_find_extent(inode,
1246 1247
				   (ext4_lblk_t)le32_to_cpu(newext->ee_block),
				    path);
1248 1249 1250 1251 1252 1253
		if (IS_ERR(path)) {
			err = PTR_ERR(path);
			goto out;
		}

		/*
1254 1255
		 * only first (depth 0 -> 1) produces free space;
		 * in all other cases we have to split the grown tree
1256 1257 1258
		 */
		depth = ext_depth(inode);
		if (path[depth].p_hdr->eh_entries == path[depth].p_hdr->eh_max) {
1259
			/* now we need to split */
1260 1261 1262 1263 1264 1265 1266 1267
			goto repeat;
		}
	}

out:
	return err;
}

1268 1269 1270 1271 1272 1273 1274
/*
 * search the closest allocated block to the left for *logical
 * and returns it at @logical + it's physical address at @phys
 * if *logical is the smallest allocated block, the function
 * returns 0 at @phys
 * return value contains 0 (success) or error code
 */
1275 1276 1277
static int ext4_ext_search_left(struct inode *inode,
				struct ext4_ext_path *path,
				ext4_lblk_t *logical, ext4_fsblk_t *phys)
1278 1279 1280
{
	struct ext4_extent_idx *ix;
	struct ext4_extent *ex;
1281
	int depth, ee_len;
1282

1283 1284 1285 1286
	if (unlikely(path == NULL)) {
		EXT4_ERROR_INODE(inode, "path == NULL *logical %d!", *logical);
		return -EIO;
	}
1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297
	depth = path->p_depth;
	*phys = 0;

	if (depth == 0 && path->p_ext == NULL)
		return 0;

	/* usually extent in the path covers blocks smaller
	 * then *logical, but it can be that extent is the
	 * first one in the file */

	ex = path[depth].p_ext;
1298
	ee_len = ext4_ext_get_actual_len(ex);
1299
	if (*logical < le32_to_cpu(ex->ee_block)) {
1300 1301 1302 1303 1304 1305
		if (unlikely(EXT_FIRST_EXTENT(path[depth].p_hdr) != ex)) {
			EXT4_ERROR_INODE(inode,
					 "EXT_FIRST_EXTENT != ex *logical %d ee_block %d!",
					 *logical, le32_to_cpu(ex->ee_block));
			return -EIO;
		}
1306 1307
		while (--depth >= 0) {
			ix = path[depth].p_idx;
1308 1309 1310
			if (unlikely(ix != EXT_FIRST_INDEX(path[depth].p_hdr))) {
				EXT4_ERROR_INODE(inode,
				  "ix (%d) != EXT_FIRST_INDEX (%d) (depth %d)!",
1311
				  ix != NULL ? le32_to_cpu(ix->ei_block) : 0,
1312
				  EXT_FIRST_INDEX(path[depth].p_hdr) != NULL ?
1313
		le32_to_cpu(EXT_FIRST_INDEX(path[depth].p_hdr)->ei_block) : 0,
1314 1315 1316
				  depth);
				return -EIO;
			}
1317 1318 1319 1320
		}
		return 0;
	}

1321 1322 1323 1324 1325 1326
	if (unlikely(*logical < (le32_to_cpu(ex->ee_block) + ee_len))) {
		EXT4_ERROR_INODE(inode,
				 "logical %d < ee_block %d + ee_len %d!",
				 *logical, le32_to_cpu(ex->ee_block), ee_len);
		return -EIO;
	}
1327

1328
	*logical = le32_to_cpu(ex->ee_block) + ee_len - 1;
1329
	*phys = ext4_ext_pblock(ex) + ee_len - 1;
1330 1331 1332 1333 1334 1335
	return 0;
}

/*
 * search the closest allocated block to the right for *logical
 * and returns it at @logical + it's physical address at @phys
1336
 * if *logical is the largest allocated block, the function
1337 1338 1339
 * returns 0 at @phys
 * return value contains 0 (success) or error code
 */
1340 1341
static int ext4_ext_search_right(struct inode *inode,
				 struct ext4_ext_path *path,
1342 1343
				 ext4_lblk_t *logical, ext4_fsblk_t *phys,
				 struct ext4_extent **ret_ex)
1344 1345 1346 1347 1348 1349
{
	struct buffer_head *bh = NULL;
	struct ext4_extent_header *eh;
	struct ext4_extent_idx *ix;
	struct ext4_extent *ex;
	ext4_fsblk_t block;
1350 1351
	int depth;	/* Note, NOT eh_depth; depth from top of tree */
	int ee_len;
1352

1353 1354 1355 1356
	if (unlikely(path == NULL)) {
		EXT4_ERROR_INODE(inode, "path == NULL *logical %d!", *logical);
		return -EIO;
	}
1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367
	depth = path->p_depth;
	*phys = 0;

	if (depth == 0 && path->p_ext == NULL)
		return 0;

	/* usually extent in the path covers blocks smaller
	 * then *logical, but it can be that extent is the
	 * first one in the file */

	ex = path[depth].p_ext;
1368
	ee_len = ext4_ext_get_actual_len(ex);
1369
	if (*logical < le32_to_cpu(ex->ee_block)) {
1370 1371 1372 1373 1374 1375
		if (unlikely(EXT_FIRST_EXTENT(path[depth].p_hdr) != ex)) {
			EXT4_ERROR_INODE(inode,
					 "first_extent(path[%d].p_hdr) != ex",
					 depth);
			return -EIO;
		}
1376 1377
		while (--depth >= 0) {
			ix = path[depth].p_idx;
1378 1379 1380 1381 1382 1383
			if (unlikely(ix != EXT_FIRST_INDEX(path[depth].p_hdr))) {
				EXT4_ERROR_INODE(inode,
						 "ix != EXT_FIRST_INDEX *logical %d!",
						 *logical);
				return -EIO;
			}
1384
		}
1385
		goto found_extent;
1386 1387
	}

1388 1389 1390 1391 1392 1393
	if (unlikely(*logical < (le32_to_cpu(ex->ee_block) + ee_len))) {
		EXT4_ERROR_INODE(inode,
				 "logical %d < ee_block %d + ee_len %d!",
				 *logical, le32_to_cpu(ex->ee_block), ee_len);
		return -EIO;
	}
1394 1395 1396 1397

	if (ex != EXT_LAST_EXTENT(path[depth].p_hdr)) {
		/* next allocated block in this leaf */
		ex++;
1398
		goto found_extent;
1399 1400 1401 1402 1403 1404
	}

	/* go up and search for index to the right */
	while (--depth >= 0) {
		ix = path[depth].p_idx;
		if (ix != EXT_LAST_INDEX(path[depth].p_hdr))
Wu Fengguang's avatar
Wu Fengguang committed
1405
			goto got_index;
1406 1407
	}

Wu Fengguang's avatar
Wu Fengguang committed
1408 1409
	/* we've gone up to the root and found no index to the right */
	return 0;
1410

Wu Fengguang's avatar
Wu Fengguang committed
1411
got_index:
1412 1413 1414 1415
	/* we've found index to the right, let's
	 * follow it and find the closest allocated
	 * block to the right */
	ix++;
1416
	block = ext4_idx_pblock(ix);
1417 1418 1419 1420 1421
	while (++depth < path->p_depth) {
		bh = sb_bread(inode->i_sb, block);
		if (bh == NULL)
			return -EIO;
		eh = ext_block_hdr(bh);
1422
		/* subtract from p_depth to get proper eh_depth */
1423 1424
		if (ext4_ext_check_block(inode, eh,
					 path->p_depth - depth, bh)) {
1425 1426 1427 1428
			put_bh(bh);
			return -EIO;
		}
		ix = EXT_FIRST_INDEX(eh);
1429
		block = ext4_idx_pblock(ix);
1430 1431 1432 1433 1434 1435 1436
		put_bh(bh);
	}

	bh = sb_bread(inode->i_sb, block);
	if (bh == NULL)
		return -EIO;
	eh = ext_block_hdr(bh);
1437
	if (ext4_ext_check_block(inode, eh, path->p_depth - depth, bh)) {
1438 1439 1440 1441
		put_bh(bh);
		return -EIO;
	}
	ex = EXT_FIRST_EXTENT(eh);
1442
found_extent:
1443
	*logical = le32_to_cpu(ex->ee_block);
1444
	*phys = ext4_ext_pblock(ex);
1445 1446 1447
	*ret_ex = ex;
	if (bh)
		put_bh(bh);
1448 1449 1450
	return 0;
}

1451
/*
1452
 * ext4_ext_next_allocated_block:
1453
 * returns allocated block in subsequent extent or EXT_MAX_BLOCKS.
1454 1455 1456
 * NOTE: it considers block number from index entry as
 * allocated block. Thus, index entries have to be consistent
 * with leaves.
1457
 */
1458
static ext4_lblk_t
1459 1460 1461 1462 1463 1464 1465 1466
ext4_ext_next_allocated_block(struct ext4_ext_path *path)
{
	int depth;

	BUG_ON(path == NULL);
	depth = path->p_depth;

	if (depth == 0 && path->p_ext == NULL)
1467
		return EXT_MAX_BLOCKS;
1468 1469 1470 1471

	while (depth >= 0) {
		if (depth == path->p_depth) {
			/* leaf */
1472 1473
			if (path[depth].p_ext &&
				path[depth].p_ext !=
1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484
					EXT_LAST_EXTENT(path[depth].p_hdr))
			  return le32_to_cpu(path[depth].p_ext[1].ee_block);
		} else {
			/* index */
			if (path[depth].p_idx !=
					EXT_LAST_INDEX(path[depth].p_hdr))
			  return le32_to_cpu(path[depth].p_idx[1].ei_block);
		}
		depth--;
	}

1485
	return EXT_MAX_BLOCKS;
1486 1487 1488
}

/*
1489
 * ext4_ext_next_leaf_block:
1490
 * returns first allocated block from next leaf or EXT_MAX_BLOCKS
1491
 */
1492
static ext4_lblk_t ext4_ext_next_leaf_block(struct ext4_ext_path *path)
1493 1494 1495 1496 1497 1498 1499 1500
{
	int depth;

	BUG_ON(path == NULL);
	depth = path->p_depth;

	/* zero-tree has no leaf blocks at all */
	if (depth == 0)
1501
		return EXT_MAX_BLOCKS;
1502 1503 1504 1505 1506 1507 1508

	/* go to index block */
	depth--;

	while (depth >= 0) {
		if (path[depth].p_idx !=
				EXT_LAST_INDEX(path[depth].p_hdr))
1509 1510
			return (ext4_lblk_t)
				le32_to_cpu(path[depth].p_idx[1].ei_block);
1511 1512 1513
		depth--;
	}

1514
	return EXT_MAX_BLOCKS;
1515 1516 1517
}

/*
1518 1519 1520
 * ext4_ext_correct_indexes:
 * if leaf gets modified and modified extent is first in the leaf,
 * then we have to correct all indexes above.
1521 1522
 * TODO: do we need to correct tree in all cases?
 */
1523
static int ext4_ext_correct_indexes(handle_t *handle, struct inode *inode,
1524 1525 1526 1527 1528 1529 1530 1531 1532 1533
				struct ext4_ext_path *path)
{
	struct ext4_extent_header *eh;
	int depth = ext_depth(inode);
	struct ext4_extent *ex;
	__le32 border;
	int k, err = 0;

	eh = path[depth].p_hdr;
	ex = path[depth].p_ext;
1534 1535 1536 1537 1538 1539

	if (unlikely(ex == NULL || eh == NULL)) {
		EXT4_ERROR_INODE(inode,
				 "ex %p == NULL or eh %p == NULL", ex, eh);
		return -EIO;
	}
1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551

	if (depth == 0) {
		/* there is no tree at all */
		return 0;
	}

	if (ex != EXT_FIRST_EXTENT(eh)) {
		/* we correct tree if first leaf got modified only */
		return 0;
	}

	/*
1552
	 * TODO: we need correction if border is smaller than current one
1553 1554 1555
	 */
	k = depth - 1;
	border = path[depth].p_ext->ee_block;
1556 1557
	err = ext4_ext_get_access(handle, inode, path + k);
	if (err)
1558 1559
		return err;
	path[k].p_idx->ei_block = border;
1560 1561
	err = ext4_ext_dirty(handle, inode, path + k);
	if (err)
1562 1563 1564 1565 1566 1567
		return err;

	while (k--) {
		/* change all left-side indexes */
		if (path[k+1].p_idx != EXT_FIRST_INDEX(path[k+1].p_hdr))
			break;
1568 1569
		err = ext4_ext_get_access(handle, inode, path + k);
		if (err)
1570 1571
			break;
		path[k].p_idx->ei_block = border;
1572 1573
		err = ext4_ext_dirty(handle, inode, path + k);
		if (err)
1574 1575 1576 1577 1578 1579
			break;
	}

	return err;
}

1580
int
1581 1582 1583
ext4_can_extents_be_merged(struct inode *inode, struct ext4_extent *ex1,
				struct ext4_extent *ex2)
{
1584
	unsigned short ext1_ee_len, ext2_ee_len, max_len;
Amit Arora's avatar
Amit Arora committed
1585 1586

	/*
1587 1588 1589 1590
	 * Make sure that both extents are initialized. We don't merge
	 * uninitialized extents so that we can be sure that end_io code has
	 * the extent that was written properly split out and conversion to
	 * initialized is trivial.
Amit Arora's avatar
Amit Arora committed
1591
	 */
1592
	if (ext4_ext_is_uninitialized(ex1) || ext4_ext_is_uninitialized(ex2))
Amit Arora's avatar
Amit Arora committed
1593 1594
		return 0;

1595 1596 1597 1598 1599
	if (ext4_ext_is_uninitialized(ex1))
		max_len = EXT_UNINIT_MAX_LEN;
	else
		max_len = EXT_INIT_MAX_LEN;

Amit Arora's avatar
Amit Arora committed
1600 1601 1602 1603
	ext1_ee_len = ext4_ext_get_actual_len(ex1);
	ext2_ee_len = ext4_ext_get_actual_len(ex2);

	if (le32_to_cpu(ex1->ee_block) + ext1_ee_len !=
1604
			le32_to_cpu(ex2->ee_block))
1605 1606
		return 0;

1607 1608 1609
	/*
	 * To allow future support for preallocated extents to be added
	 * as an RO_COMPAT feature, refuse to merge to extents if
1610
	 * this can result in the top bit of ee_len being set.
1611
	 */
1612
	if (ext1_ee_len + ext2_ee_len > max_len)
1613
		return 0;
1614
#ifdef AGGRESSIVE_TEST
1615
	if (ext1_ee_len >= 4)
1616 1617 1618
		return 0;
#endif

1619
	if (ext4_ext_pblock(ex1) + ext1_ee_len == ext4_ext_pblock(ex2))
1620 1621 1622 1623
		return 1;
	return 0;
}

1624 1625 1626 1627 1628 1629 1630
/*
 * This function tries to merge the "ex" extent to the next extent in the tree.
 * It always tries to merge towards right. If you want to merge towards
 * left, pass "ex - 1" as argument instead of "ex".
 * Returns 0 if the extents (ex and ex+1) were _not_ merged and returns
 * 1 if they got merged.
 */
1631
static int ext4_ext_try_to_merge_right(struct inode *inode,
1632 1633
				 struct ext4_ext_path *path,
				 struct ext4_extent *ex)
1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659
{
	struct ext4_extent_header *eh;
	unsigned int depth, len;
	int merge_done = 0;
	int uninitialized = 0;

	depth = ext_depth(inode);
	BUG_ON(path[depth].p_hdr == NULL);
	eh = path[depth].p_hdr;

	while (ex < EXT_LAST_EXTENT(eh)) {
		if (!ext4_can_extents_be_merged(inode, ex, ex + 1))
			break;
		/* merge with next extent! */
		if (ext4_ext_is_uninitialized(ex))
			uninitialized = 1;
		ex->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex)
				+ ext4_ext_get_actual_len(ex + 1));
		if (uninitialized)
			ext4_ext_mark_uninitialized(ex);

		if (ex + 1 < EXT_LAST_EXTENT(eh)) {
			len = (EXT_LAST_EXTENT(eh) - ex - 1)
				* sizeof(struct ext4_extent);
			memmove(ex + 1, ex + 2, len);
		}
1660
		le16_add_cpu(&eh->eh_entries, -1);
1661 1662 1663
		merge_done = 1;
		WARN_ON(eh->eh_entries == 0);
		if (!eh->eh_entries)
1664
			EXT4_ERROR_INODE(inode, "eh->eh_entries = 0!");
1665 1666 1667 1668 1669
	}

	return merge_done;
}

1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713
/*
 * This function does a very simple check to see if we can collapse
 * an extent tree with a single extent tree leaf block into the inode.
 */
static void ext4_ext_try_to_merge_up(handle_t *handle,
				     struct inode *inode,
				     struct ext4_ext_path *path)
{
	size_t s;
	unsigned max_root = ext4_ext_space_root(inode, 0);
	ext4_fsblk_t blk;

	if ((path[0].p_depth != 1) ||
	    (le16_to_cpu(path[0].p_hdr->eh_entries) != 1) ||
	    (le16_to_cpu(path[1].p_hdr->eh_entries) > max_root))
		return;

	/*
	 * We need to modify the block allocation bitmap and the block
	 * group descriptor to release the extent tree block.  If we
	 * can't get the journal credits, give up.
	 */
	if (ext4_journal_extend(handle, 2))
		return;

	/*
	 * Copy the extent data up to the inode
	 */
	blk = ext4_idx_pblock(path[0].p_idx);
	s = le16_to_cpu(path[1].p_hdr->eh_entries) *
		sizeof(struct ext4_extent_idx);
	s += sizeof(struct ext4_extent_header);

	memcpy(path[0].p_hdr, path[1].p_hdr, s);
	path[0].p_depth = 0;
	path[0].p_ext = EXT_FIRST_EXTENT(path[0].p_hdr) +
		(path[1].p_ext - EXT_FIRST_EXTENT(path[1].p_hdr));
	path[0].p_hdr->eh_max = cpu_to_le16(max_root);

	brelse(path[1].p_bh);
	ext4_free_blocks(handle, inode, NULL, blk, 1,
			 EXT4_FREE_BLOCKS_METADATA | EXT4_FREE_BLOCKS_FORGET);
}

1714 1715 1716 1717
/*
 * This function tries to merge the @ex extent to neighbours in the tree.
 * return 1 if merge left else 0.
 */
1718 1719
static void ext4_ext_try_to_merge(handle_t *handle,
				  struct inode *inode,
1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733
				  struct ext4_ext_path *path,
				  struct ext4_extent *ex) {
	struct ext4_extent_header *eh;
	unsigned int depth;
	int merge_done = 0;

	depth = ext_depth(inode);
	BUG_ON(path[depth].p_hdr == NULL);
	eh = path[depth].p_hdr;

	if (ex > EXT_FIRST_EXTENT(eh))
		merge_done = ext4_ext_try_to_merge_right(inode, path, ex - 1);

	if (!merge_done)
1734
		(void) ext4_ext_try_to_merge_right(inode, path, ex);
1735

1736
	ext4_ext_try_to_merge_up(handle, inode, path);
1737 1738
}

Amit Arora's avatar
Amit Arora committed
1739 1740 1741 1742 1743 1744 1745 1746
/*
 * check if a portion of the "newext" extent overlaps with an
 * existing extent.
 *
 * If there is an overlap discovered, it updates the length of the newext
 * such that there will be no overlap, and then returns 1.
 * If there is no overlap found, it returns 0.
 */
1747 1748
static unsigned int ext4_ext_check_overlap(struct ext4_sb_info *sbi,
					   struct inode *inode,
1749 1750
					   struct ext4_extent *newext,
					   struct ext4_ext_path *path)
Amit Arora's avatar
Amit Arora committed
1751
{
1752
	ext4_lblk_t b1, b2;
Amit Arora's avatar
Amit Arora committed
1753 1754 1755 1756
	unsigned int depth, len1;
	unsigned int ret = 0;

	b1 = le32_to_cpu(newext->ee_block);
Amit Arora's avatar
Amit Arora committed
1757
	len1 = ext4_ext_get_actual_len(newext);
Amit Arora's avatar
Amit Arora committed
1758 1759 1760 1761
	depth = ext_depth(inode);
	if (!path[depth].p_ext)
		goto out;
	b2 = le32_to_cpu(path[depth].p_ext->ee_block);
1762
	b2 &= ~(sbi->s_cluster_ratio - 1);
Amit Arora's avatar
Amit Arora committed
1763 1764 1765

	/*
	 * get the next allocated block if the extent in the path
1766
	 * is before the requested block(s)
Amit Arora's avatar
Amit Arora committed
1767 1768 1769
	 */
	if (b2 < b1) {
		b2 = ext4_ext_next_allocated_block(path);
1770
		if (b2 == EXT_MAX_BLOCKS)
Amit Arora's avatar
Amit Arora committed
1771
			goto out;
1772
		b2 &= ~(sbi->s_cluster_ratio - 1);
Amit Arora's avatar
Amit Arora committed
1773 1774
	}

1775
	/* check for wrap through zero on extent logical start block*/
Amit Arora's avatar
Amit Arora committed
1776
	if (b1 + len1 < b1) {
1777
		len1 = EXT_MAX_BLOCKS - b1;
Amit Arora's avatar
Amit Arora committed
1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790
		newext->ee_len = cpu_to_le16(len1);
		ret = 1;
	}

	/* check for overlap */
	if (b1 + len1 > b2) {
		newext->ee_len = cpu_to_le16(b2 - b1);
		ret = 1;
	}
out:
	return ret;
}

1791
/*
1792 1793 1794 1795
 * ext4_ext_insert_extent:
 * tries to merge requsted extent into the existing extent or
 * inserts requested extent as new one into the tree,
 * creating new leaf in the no-space case.
1796 1797 1798
 */
int ext4_ext_insert_extent(handle_t *handle, struct inode *inode,
				struct ext4_ext_path *path,
1799
				struct ext4_extent *newext, int flag)
1800
{
1801
	struct ext4_extent_header *eh;
1802 1803 1804
	struct ext4_extent *ex, *fex;
	struct ext4_extent *nearex; /* nearest extent */
	struct ext4_ext_path *npath = NULL;
1805 1806
	int depth, len, err;
	ext4_lblk_t next;
Amit Arora's avatar
Amit Arora committed
1807
	unsigned uninitialized = 0;
1808
	int flags = 0;
1809

1810 1811 1812 1813
	if (unlikely(ext4_ext_get_actual_len(newext) == 0)) {
		EXT4_ERROR_INODE(inode, "ext4_ext_get_actual_len(newext) == 0");
		return -EIO;
	}
1814 1815
	depth = ext_depth(inode);
	ex = path[depth].p_ext;
1816
	eh = path[depth].p_hdr;
1817 1818 1819 1820
	if (unlikely(path[depth].p_hdr == NULL)) {
		EXT4_ERROR_INODE(inode, "path[%d].p_hdr == NULL", depth);
		return -EIO;
	}
1821 1822

	/* try to insert block into found extent and return */
1823
	if (ex && !(flag & EXT4_GET_BLOCKS_PRE_IO)) {
Amit Arora's avatar
Amit Arora committed
1824 1825

		/*
1826 1827 1828 1829 1830
		 * Try to see whether we should rather test the extent on
		 * right from ex, or from the left of ex. This is because
		 * ext4_ext_find_extent() can return either extent on the
		 * left, or on the right from the searched position. This
		 * will make merging more effective.
Amit Arora's avatar
Amit Arora committed
1831
		 */
1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867
		if (ex < EXT_LAST_EXTENT(eh) &&
		    (le32_to_cpu(ex->ee_block) +
		    ext4_ext_get_actual_len(ex) <
		    le32_to_cpu(newext->ee_block))) {
			ex += 1;
			goto prepend;
		} else if ((ex > EXT_FIRST_EXTENT(eh)) &&
			   (le32_to_cpu(newext->ee_block) +
			   ext4_ext_get_actual_len(newext) <
			   le32_to_cpu(ex->ee_block)))
			ex -= 1;

		/* Try to append newex to the ex */
		if (ext4_can_extents_be_merged(inode, ex, newext)) {
			ext_debug("append [%d]%d block to %u:[%d]%d"
				  "(from %llu)\n",
				  ext4_ext_is_uninitialized(newext),
				  ext4_ext_get_actual_len(newext),
				  le32_to_cpu(ex->ee_block),
				  ext4_ext_is_uninitialized(ex),
				  ext4_ext_get_actual_len(ex),
				  ext4_ext_pblock(ex));
			err = ext4_ext_get_access(handle, inode,
						  path + depth);
			if (err)
				return err;

			/*
			 * ext4_can_extents_be_merged should have checked
			 * that either both extents are uninitialized, or
			 * both aren't. Thus we need to check only one of
			 * them here.
			 */
			if (ext4_ext_is_uninitialized(ex))
				uninitialized = 1;
			ex->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex)
Amit Arora's avatar
Amit Arora committed
1868
					+ ext4_ext_get_actual_len(newext));
1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910
			if (uninitialized)
				ext4_ext_mark_uninitialized(ex);
			eh = path[depth].p_hdr;
			nearex = ex;
			goto merge;
		}

prepend:
		/* Try to prepend newex to the ex */
		if (ext4_can_extents_be_merged(inode, newext, ex)) {
			ext_debug("prepend %u[%d]%d block to %u:[%d]%d"
				  "(from %llu)\n",
				  le32_to_cpu(newext->ee_block),
				  ext4_ext_is_uninitialized(newext),
				  ext4_ext_get_actual_len(newext),
				  le32_to_cpu(ex->ee_block),
				  ext4_ext_is_uninitialized(ex),
				  ext4_ext_get_actual_len(ex),
				  ext4_ext_pblock(ex));
			err = ext4_ext_get_access(handle, inode,
						  path + depth);
			if (err)
				return err;

			/*
			 * ext4_can_extents_be_merged should have checked
			 * that either both extents are uninitialized, or
			 * both aren't. Thus we need to check only one of
			 * them here.
			 */
			if (ext4_ext_is_uninitialized(ex))
				uninitialized = 1;
			ex->ee_block = newext->ee_block;
			ext4_ext_store_pblock(ex, ext4_ext_pblock(newext));
			ex->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex)
					+ ext4_ext_get_actual_len(newext));
			if (uninitialized)
				ext4_ext_mark_uninitialized(ex);
			eh = path[depth].p_hdr;
			nearex = ex;
			goto merge;
		}
1911 1912 1913 1914 1915 1916 1917 1918 1919
	}

	depth = ext_depth(inode);
	eh = path[depth].p_hdr;
	if (le16_to_cpu(eh->eh_entries) < le16_to_cpu(eh->eh_max))
		goto has_space;

	/* probably next leaf has space for us? */
	fex = EXT_LAST_EXTENT(eh);
1920 1921
	next = EXT_MAX_BLOCKS;
	if (le32_to_cpu(newext->ee_block) > le32_to_cpu(fex->ee_block))
1922
		next = ext4_ext_next_leaf_block(path);
1923
	if (next != EXT_MAX_BLOCKS) {
1924
		ext_debug("next leaf block - %u\n", next);
1925 1926 1927 1928 1929 1930 1931
		BUG_ON(npath != NULL);
		npath = ext4_ext_find_extent(inode, next, NULL);
		if (IS_ERR(npath))
			return PTR_ERR(npath);
		BUG_ON(npath->p_depth != path->p_depth);
		eh = npath[depth].p_hdr;
		if (le16_to_cpu(eh->eh_entries) < le16_to_cpu(eh->eh_max)) {
Lucas De Marchi's avatar
Lucas De Marchi committed
1932
			ext_debug("next leaf isn't full(%d)\n",
1933 1934
				  le16_to_cpu(eh->eh_entries));
			path = npath;
1935
			goto has_space;
1936 1937 1938 1939 1940 1941
		}
		ext_debug("next leaf has no free space(%d,%d)\n",
			  le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max));
	}

	/*
1942 1943
	 * There is no free space in the found leaf.
	 * We're gonna add a new leaf in the tree.
1944
	 */
1945 1946
	if (flag & EXT4_GET_BLOCKS_METADATA_NOFAIL)
		flags = EXT4_MB_USE_RESERVED;
1947
	err = ext4_ext_create_new_leaf(handle, inode, flags, path, newext);
1948 1949 1950 1951 1952 1953 1954 1955
	if (err)
		goto cleanup;
	depth = ext_depth(inode);
	eh = path[depth].p_hdr;

has_space:
	nearex = path[depth].p_ext;

1956 1957
	err = ext4_ext_get_access(handle, inode, path + depth);
	if (err)
1958 1959 1960 1961
		goto cleanup;

	if (!nearex) {
		/* there is no extent in this leaf, create first one */
1962
		ext_debug("first extent in the leaf: %u:%llu:[%d]%d\n",
Dave Kleikamp's avatar
Dave Kleikamp committed
1963
				le32_to_cpu(newext->ee_block),
1964
				ext4_ext_pblock(newext),
1965
				ext4_ext_is_uninitialized(newext),
Amit Arora's avatar
Amit Arora committed
1966
				ext4_ext_get_actual_len(newext));
1967 1968 1969
		nearex = EXT_FIRST_EXTENT(eh);
	} else {
		if (le32_to_cpu(newext->ee_block)
Dave Kleikamp's avatar
Dave Kleikamp committed
1970
			   > le32_to_cpu(nearex->ee_block)) {
1971
			/* Insert after */
1972 1973
			ext_debug("insert %u:%llu:[%d]%d before: "
					"nearest %p\n",
1974 1975 1976 1977 1978 1979 1980 1981 1982
					le32_to_cpu(newext->ee_block),
					ext4_ext_pblock(newext),
					ext4_ext_is_uninitialized(newext),
					ext4_ext_get_actual_len(newext),
					nearex);
			nearex++;
		} else {
			/* Insert before */
			BUG_ON(newext->ee_block == nearex->ee_block);
1983 1984
			ext_debug("insert %u:%llu:[%d]%d after: "
					"nearest %p\n",
Dave Kleikamp's avatar
Dave Kleikamp committed
1985
					le32_to_cpu(newext->ee_block),
1986
					ext4_ext_pblock(newext),
1987
					ext4_ext_is_uninitialized(newext),
Amit Arora's avatar
Amit Arora committed
1988
					ext4_ext_get_actual_len(newext),
1989 1990 1991 1992
					nearex);
		}
		len = EXT_LAST_EXTENT(eh) - nearex + 1;
		if (len > 0) {
1993
			ext_debug("insert %u:%llu:[%d]%d: "
1994 1995 1996 1997 1998 1999 2000 2001
					"move %d extents from 0x%p to 0x%p\n",
					le32_to_cpu(newext->ee_block),
					ext4_ext_pblock(newext),
					ext4_ext_is_uninitialized(newext),
					ext4_ext_get_actual_len(newext),
					len, nearex, nearex + 1);
			memmove(nearex + 1, nearex,
				len * sizeof(struct ext4_extent));
2002 2003 2004
		}
	}

2005
	le16_add_cpu(&eh->eh_entries, 1);
2006
	path[depth].p_ext = nearex;
2007
	nearex->ee_block = newext->ee_block;
2008
	ext4_ext_store_pblock(nearex, ext4_ext_pblock(newext));
2009 2010 2011
	nearex->ee_len = newext->ee_len;

merge:
2012
	/* try to merge extents */
2013
	if (!(flag & EXT4_GET_BLOCKS_PRE_IO))
2014
		ext4_ext_try_to_merge(handle, inode, path, nearex);
2015 2016 2017 2018 2019 2020 2021


	/* time to correct all indexes above */
	err = ext4_ext_correct_indexes(handle, inode, path);
	if (err)
		goto cleanup;

2022
	err = ext4_ext_dirty(handle, inode, path + path->p_depth);
2023 2024 2025 2026 2027 2028 2029 2030 2031

cleanup:
	if (npath) {
		ext4_ext_drop_refs(npath);
		kfree(npath);
	}
	return err;
}

2032 2033 2034
static int ext4_fill_fiemap_extents(struct inode *inode,
				    ext4_lblk_t block, ext4_lblk_t num,
				    struct fiemap_extent_info *fieinfo)
2035 2036 2037
{
	struct ext4_ext_path *path = NULL;
	struct ext4_extent *ex;
Zheng Liu's avatar
Zheng Liu committed
2038
	struct extent_status es;
2039
	ext4_lblk_t next, next_del, start = 0, end = 0;
2040
	ext4_lblk_t last = block + num;
2041 2042 2043
	int exists, depth = 0, err = 0;
	unsigned int flags = 0;
	unsigned char blksize_bits = inode->i_sb->s_blocksize_bits;
2044

2045
	while (block < last && block != EXT_MAX_BLOCKS) {
2046 2047
		num = last - block;
		/* find extent for this block */
2048
		down_read(&EXT4_I(inode)->i_data_sem);
2049 2050 2051 2052 2053 2054 2055

		if (path && ext_depth(inode) != depth) {
			/* depth was changed. we have to realloc path */
			kfree(path);
			path = NULL;
		}

2056 2057
		path = ext4_ext_find_extent(inode, block, path);
		if (IS_ERR(path)) {
2058
			up_read(&EXT4_I(inode)->i_data_sem);
2059 2060 2061 2062 2063 2064
			err = PTR_ERR(path);
			path = NULL;
			break;
		}

		depth = ext_depth(inode);
2065
		if (unlikely(path[depth].p_hdr == NULL)) {
2066
			up_read(&EXT4_I(inode)->i_data_sem);
2067 2068 2069 2070
			EXT4_ERROR_INODE(inode, "path[%d].p_hdr == NULL", depth);
			err = -EIO;
			break;
		}
2071 2072
		ex = path[depth].p_ext;
		next = ext4_ext_next_allocated_block(path);
2073
		ext4_ext_drop_refs(path);
2074

2075
		flags = 0;
2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111
		exists = 0;
		if (!ex) {
			/* there is no extent yet, so try to allocate
			 * all requested space */
			start = block;
			end = block + num;
		} else if (le32_to_cpu(ex->ee_block) > block) {
			/* need to allocate space before found extent */
			start = block;
			end = le32_to_cpu(ex->ee_block);
			if (block + num < end)
				end = block + num;
		} else if (block >= le32_to_cpu(ex->ee_block)
					+ ext4_ext_get_actual_len(ex)) {
			/* need to allocate space after found extent */
			start = block;
			end = block + num;
			if (end >= next)
				end = next;
		} else if (block >= le32_to_cpu(ex->ee_block)) {
			/*
			 * some part of requested space is covered
			 * by found extent
			 */
			start = block;
			end = le32_to_cpu(ex->ee_block)
				+ ext4_ext_get_actual_len(ex);
			if (block + num < end)
				end = block + num;
			exists = 1;
		} else {
			BUG();
		}
		BUG_ON(end <= start);

		if (!exists) {
Zheng Liu's avatar
Zheng Liu committed
2112 2113 2114
			es.es_lblk = start;
			es.es_len = end - start;
			es.es_pblk = 0;
2115
		} else {
Zheng Liu's avatar
Zheng Liu committed
2116 2117 2118
			es.es_lblk = le32_to_cpu(ex->ee_block);
			es.es_len = ext4_ext_get_actual_len(ex);
			es.es_pblk = ext4_ext_pblock(ex);
2119 2120
			if (ext4_ext_is_uninitialized(ex))
				flags |= FIEMAP_EXTENT_UNWRITTEN;
2121 2122
		}

2123
		/*
Zheng Liu's avatar
Zheng Liu committed
2124 2125
		 * Find delayed extent and update es accordingly. We call
		 * it even in !exists case to find out whether es is the
2126 2127
		 * last existing extent or not.
		 */
Zheng Liu's avatar
Zheng Liu committed
2128
		next_del = ext4_find_delayed_extent(inode, &es);
2129 2130 2131 2132 2133 2134
		if (!exists && next_del) {
			exists = 1;
			flags |= FIEMAP_EXTENT_DELALLOC;
		}
		up_read(&EXT4_I(inode)->i_data_sem);

Zheng Liu's avatar
Zheng Liu committed
2135 2136
		if (unlikely(es.es_len == 0)) {
			EXT4_ERROR_INODE(inode, "es.es_len == 0");
2137 2138 2139
			err = -EIO;
			break;
		}
2140

2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152
		/*
		 * This is possible iff next == next_del == EXT_MAX_BLOCKS.
		 * we need to check next == EXT_MAX_BLOCKS because it is
		 * possible that an extent is with unwritten and delayed
		 * status due to when an extent is delayed allocated and
		 * is allocated by fallocate status tree will track both of
		 * them in a extent.
		 *
		 * So we could return a unwritten and delayed extent, and
		 * its block is equal to 'next'.
		 */
		if (next == next_del && next == EXT_MAX_BLOCKS) {
2153 2154 2155 2156 2157 2158 2159 2160 2161 2162
			flags |= FIEMAP_EXTENT_LAST;
			if (unlikely(next_del != EXT_MAX_BLOCKS ||
				     next != EXT_MAX_BLOCKS)) {
				EXT4_ERROR_INODE(inode,
						 "next extent == %u, next "
						 "delalloc extent = %u",
						 next, next_del);
				err = -EIO;
				break;
			}
2163 2164
		}

2165 2166
		if (exists) {
			err = fiemap_fill_next_extent(fieinfo,
Zheng Liu's avatar
Zheng Liu committed
2167 2168 2169
				(__u64)es.es_lblk << blksize_bits,
				(__u64)es.es_pblk << blksize_bits,
				(__u64)es.es_len << blksize_bits,
2170 2171 2172 2173 2174 2175 2176
				flags);
			if (err < 0)
				break;
			if (err == 1) {
				err = 0;
				break;
			}
2177 2178
		}

Zheng Liu's avatar
Zheng Liu committed
2179
		block = es.es_lblk + es.es_len;
2180 2181 2182 2183 2184 2185 2186 2187 2188 2189
	}

	if (path) {
		ext4_ext_drop_refs(path);
		kfree(path);
	}

	return err;
}

2190
/*
2191 2192
 * ext4_ext_put_gap_in_cache:
 * calculate boundaries of the gap that the requested block fits into
2193 2194
 * and cache this gap
 */
2195
static void
2196
ext4_ext_put_gap_in_cache(struct inode *inode, struct ext4_ext_path *path,
2197
				ext4_lblk_t block)
2198 2199
{
	int depth = ext_depth(inode);
2200 2201
	unsigned long len;
	ext4_lblk_t lblock;
2202 2203 2204 2205
	struct ext4_extent *ex;

	ex = path[depth].p_ext;
	if (ex == NULL) {
Zheng Liu's avatar
Zheng Liu committed
2206 2207 2208 2209
		/*
		 * there is no extent yet, so gap is [0;-] and we
		 * don't cache it
		 */
2210 2211 2212 2213
		ext_debug("cache gap(whole file):");
	} else if (block < le32_to_cpu(ex->ee_block)) {
		lblock = block;
		len = le32_to_cpu(ex->ee_block) - block;
2214 2215 2216 2217
		ext_debug("cache gap(before): %u [%u:%u]",
				block,
				le32_to_cpu(ex->ee_block),
				 ext4_ext_get_actual_len(ex));
2218 2219 2220
		if (!ext4_find_delalloc_range(inode, lblock, lblock + len - 1))
			ext4_es_insert_extent(inode, lblock, len, ~0,
					      EXTENT_STATUS_HOLE);
2221
	} else if (block >= le32_to_cpu(ex->ee_block)
Amit Arora's avatar
Amit Arora committed
2222
			+ ext4_ext_get_actual_len(ex)) {
2223
		ext4_lblk_t next;
Dave Kleikamp's avatar
Dave Kleikamp committed
2224
		lblock = le32_to_cpu(ex->ee_block)
Amit Arora's avatar
Amit Arora committed
2225
			+ ext4_ext_get_actual_len(ex);
2226 2227

		next = ext4_ext_next_allocated_block(path);
2228 2229 2230 2231
		ext_debug("cache gap(after): [%u:%u] %u",
				le32_to_cpu(ex->ee_block),
				ext4_ext_get_actual_len(ex),
				block);
2232 2233
		BUG_ON(next == lblock);
		len = next - lblock;
2234 2235 2236
		if (!ext4_find_delalloc_range(inode, lblock, lblock + len - 1))
			ext4_es_insert_extent(inode, lblock, len, ~0,
					      EXTENT_STATUS_HOLE);
2237 2238 2239 2240 2241
	} else {
		lblock = len = 0;
		BUG();
	}

2242
	ext_debug(" -> %u:%lu\n", lblock, len);
2243 2244 2245
}

/*
2246 2247
 * ext4_ext_rm_idx:
 * removes index from the index block.
2248
 */
2249
static int ext4_ext_rm_idx(handle_t *handle, struct inode *inode,
2250
			struct ext4_ext_path *path, int depth)
2251 2252
{
	int err;
2253
	ext4_fsblk_t leaf;
2254 2255

	/* free index block */
2256 2257
	depth--;
	path = path + depth;
2258
	leaf = ext4_idx_pblock(path->p_idx);
2259 2260 2261 2262
	if (unlikely(path->p_hdr->eh_entries == 0)) {
		EXT4_ERROR_INODE(inode, "path->p_hdr->eh_entries == 0");
		return -EIO;
	}
2263 2264
	err = ext4_ext_get_access(handle, inode, path);
	if (err)
2265
		return err;
2266 2267 2268 2269 2270 2271 2272

	if (path->p_idx != EXT_LAST_INDEX(path->p_hdr)) {
		int len = EXT_LAST_INDEX(path->p_hdr) - path->p_idx;
		len *= sizeof(struct ext4_extent_idx);
		memmove(path->p_idx, path->p_idx + 1, len);
	}

2273
	le16_add_cpu(&path->p_hdr->eh_entries, -1);
2274 2275
	err = ext4_ext_dirty(handle, inode, path);
	if (err)
2276
		return err;
2277
	ext_debug("index is empty, remove it, free block %llu\n", leaf);
2278 2279
	trace_ext4_ext_rm_idx(inode, leaf);

2280
	ext4_free_blocks(handle, inode, NULL, leaf, 1,
2281
			 EXT4_FREE_BLOCKS_METADATA | EXT4_FREE_BLOCKS_FORGET);
2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294

	while (--depth >= 0) {
		if (path->p_idx != EXT_FIRST_INDEX(path->p_hdr))
			break;
		path--;
		err = ext4_ext_get_access(handle, inode, path);
		if (err)
			break;
		path->p_idx->ei_block = (path+1)->p_idx->ei_block;
		err = ext4_ext_dirty(handle, inode, path);
		if (err)
			break;
	}
2295 2296 2297 2298
	return err;
}

/*
2299 2300 2301 2302 2303
 * ext4_ext_calc_credits_for_single_extent:
 * This routine returns max. credits that needed to insert an extent
 * to the extent tree.
 * When pass the actual path, the caller should calculate credits
 * under i_data_sem.
2304
 */
2305
int ext4_ext_calc_credits_for_single_extent(struct inode *inode, int nrblocks,
2306 2307 2308
						struct ext4_ext_path *path)
{
	if (path) {
2309
		int depth = ext_depth(inode);
2310
		int ret = 0;
2311

2312 2313
		/* probably there is space in leaf? */
		if (le16_to_cpu(path[depth].p_hdr->eh_entries)
2314
				< le16_to_cpu(path[depth].p_hdr->eh_max)) {
2315

2316 2317 2318 2319 2320
			/*
			 *  There are some space in the leaf tree, no
			 *  need to account for leaf block credit
			 *
			 *  bitmaps and block group descriptor blocks
2321
			 *  and other metadata blocks still need to be
2322 2323
			 *  accounted.
			 */
2324
			/* 1 bitmap, 1 block group descriptor */
2325
			ret = 2 + EXT4_META_TRANS_BLOCKS(inode->i_sb);
2326
			return ret;
2327 2328
		}
	}
2329

2330
	return ext4_chunk_trans_blocks(inode, nrblocks);
2331
}
2332

2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343
/*
 * How many index/leaf blocks need to change/allocate to modify nrblocks?
 *
 * if nrblocks are fit in a single extent (chunk flag is 1), then
 * in the worse case, each tree level index/leaf need to be changed
 * if the tree split due to insert a new extent, then the old tree
 * index/leaf need to be updated too
 *
 * If the nrblocks are discontiguous, they could cause
 * the whole tree split more than once, but this is really rare.
 */
2344
int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks, int chunk)
2345 2346
{
	int index;
2347 2348 2349 2350 2351 2352 2353
	int depth;

	/* If we are converting the inline data, only one is needed here. */
	if (ext4_has_inline_data(inode))
		return 1;

	depth = ext_depth(inode);
2354

2355 2356 2357 2358
	if (chunk)
		index = depth * 2;
	else
		index = depth * 3;
2359

2360
	return index;
2361 2362 2363
}

static int ext4_remove_blocks(handle_t *handle, struct inode *inode,
2364 2365 2366
			      struct ext4_extent *ex,
			      ext4_fsblk_t *partial_cluster,
			      ext4_lblk_t from, ext4_lblk_t to)
2367
{
2368
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
Amit Arora's avatar
Amit Arora committed
2369
	unsigned short ee_len =  ext4_ext_get_actual_len(ex);
2370
	ext4_fsblk_t pblk;
2371
	int flags = 0;
2372

2373
	if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
2374 2375 2376 2377
		flags |= EXT4_FREE_BLOCKS_METADATA | EXT4_FREE_BLOCKS_FORGET;
	else if (ext4_should_journal_data(inode))
		flags |= EXT4_FREE_BLOCKS_FORGET;

2378 2379 2380 2381 2382 2383 2384 2385 2386
	/*
	 * For bigalloc file systems, we never free a partial cluster
	 * at the beginning of the extent.  Instead, we make a note
	 * that we tried freeing the cluster, and check to see if we
	 * need to free it on a subsequent call to ext4_remove_blocks,
	 * or at the end of the ext4_truncate() operation.
	 */
	flags |= EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER;

2387
	trace_ext4_remove_blocks(inode, ex, from, to, *partial_cluster);
2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400
	/*
	 * If we have a partial cluster, and it's different from the
	 * cluster of the last block, we need to explicitly free the
	 * partial cluster here.
	 */
	pblk = ext4_ext_pblock(ex) + ee_len - 1;
	if (*partial_cluster && (EXT4_B2C(sbi, pblk) != *partial_cluster)) {
		ext4_free_blocks(handle, inode, NULL,
				 EXT4_C2B(sbi, *partial_cluster),
				 sbi->s_cluster_ratio, flags);
		*partial_cluster = 0;
	}

2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416
#ifdef EXTENTS_STATS
	{
		struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
		spin_lock(&sbi->s_ext_stats_lock);
		sbi->s_ext_blocks += ee_len;
		sbi->s_ext_extents++;
		if (ee_len < sbi->s_ext_min)
			sbi->s_ext_min = ee_len;
		if (ee_len > sbi->s_ext_max)
			sbi->s_ext_max = ee_len;
		if (ext_depth(inode) > sbi->s_depth_max)
			sbi->s_depth_max = ext_depth(inode);
		spin_unlock(&sbi->s_ext_stats_lock);
	}
#endif
	if (from >= le32_to_cpu(ex->ee_block)
Amit Arora's avatar
Amit Arora committed
2417
	    && to == le32_to_cpu(ex->ee_block) + ee_len - 1) {
2418
		/* tail removal */
2419 2420
		ext4_lblk_t num;

Amit Arora's avatar
Amit Arora committed
2421
		num = le32_to_cpu(ex->ee_block) + ee_len - from;
2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437
		pblk = ext4_ext_pblock(ex) + ee_len - num;
		ext_debug("free last %u blocks starting %llu\n", num, pblk);
		ext4_free_blocks(handle, inode, NULL, pblk, num, flags);
		/*
		 * If the block range to be freed didn't start at the
		 * beginning of a cluster, and we removed the entire
		 * extent, save the partial cluster here, since we
		 * might need to delete if we determine that the
		 * truncate operation has removed all of the blocks in
		 * the cluster.
		 */
		if (pblk & (sbi->s_cluster_ratio - 1) &&
		    (ee_len == num))
			*partial_cluster = EXT4_B2C(sbi, pblk);
		else
			*partial_cluster = 0;
2438
	} else if (from == le32_to_cpu(ex->ee_block)
Amit Arora's avatar
Amit Arora committed
2439
		   && to <= le32_to_cpu(ex->ee_block) + ee_len - 1) {
Allison Henderson's avatar
Allison Henderson committed
2440 2441 2442 2443 2444 2445 2446 2447
		/* head removal */
		ext4_lblk_t num;
		ext4_fsblk_t start;

		num = to - from;
		start = ext4_ext_pblock(ex);

		ext_debug("free first %u blocks starting %llu\n", num, start);
2448
		ext4_free_blocks(handle, inode, NULL, start, num, flags);
Allison Henderson's avatar
Allison Henderson committed
2449

2450
	} else {
2451 2452 2453
		printk(KERN_INFO "strange request: removal(2) "
				"%u-%u from %u:%u\n",
				from, to, le32_to_cpu(ex->ee_block), ee_len);
2454 2455 2456 2457
	}
	return 0;
}

Allison Henderson's avatar
Allison Henderson committed
2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469

/*
 * ext4_ext_rm_leaf() Removes the extents associated with the
 * blocks appearing between "start" and "end", and splits the extents
 * if "start" and "end" appear in the same extent
 *
 * @handle: The journal handle
 * @inode:  The files inode
 * @path:   The path to the leaf
 * @start:  The first block to remove
 * @end:   The last block to remove
 */
2470 2471
static int
ext4_ext_rm_leaf(handle_t *handle, struct inode *inode,
2472 2473
		 struct ext4_ext_path *path, ext4_fsblk_t *partial_cluster,
		 ext4_lblk_t start, ext4_lblk_t end)
2474
{
2475
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
2476 2477 2478
	int err = 0, correct_index = 0;
	int depth = ext_depth(inode), credits;
	struct ext4_extent_header *eh;
2479
	ext4_lblk_t a, b;
2480 2481
	unsigned num;
	ext4_lblk_t ex_ee_block;
2482
	unsigned short ex_ee_len;
Amit Arora's avatar
Amit Arora committed
2483
	unsigned uninitialized = 0;
2484 2485
	struct ext4_extent *ex;

2486
	/* the header must be checked already in ext4_ext_remove_space() */
2487
	ext_debug("truncate since %u in leaf to %u\n", start, end);
2488 2489 2490
	if (!path[depth].p_hdr)
		path[depth].p_hdr = ext_block_hdr(path[depth].p_bh);
	eh = path[depth].p_hdr;
2491 2492 2493 2494
	if (unlikely(path[depth].p_hdr == NULL)) {
		EXT4_ERROR_INODE(inode, "path[%d].p_hdr == NULL", depth);
		return -EIO;
	}
2495 2496 2497 2498
	/* find where to start removing */
	ex = EXT_LAST_EXTENT(eh);

	ex_ee_block = le32_to_cpu(ex->ee_block);
Amit Arora's avatar
Amit Arora committed
2499
	ex_ee_len = ext4_ext_get_actual_len(ex);
2500

2501 2502
	trace_ext4_ext_rm_leaf(inode, start, ex, *partial_cluster);

2503 2504
	while (ex >= EXT_FIRST_EXTENT(eh) &&
			ex_ee_block + ex_ee_len > start) {
2505 2506 2507 2508 2509 2510

		if (ext4_ext_is_uninitialized(ex))
			uninitialized = 1;
		else
			uninitialized = 0;

2511 2512
		ext_debug("remove ext %u:[%d]%d\n", ex_ee_block,
			 uninitialized, ex_ee_len);
2513 2514 2515
		path[depth].p_ext = ex;

		a = ex_ee_block > start ? ex_ee_block : start;
Allison Henderson's avatar
Allison Henderson committed
2516 2517
		b = ex_ee_block+ex_ee_len - 1 < end ?
			ex_ee_block+ex_ee_len - 1 : end;
2518 2519 2520

		ext_debug("  border %u:%u\n", a, b);

Allison Henderson's avatar
Allison Henderson committed
2521
		/* If this extent is beyond the end of the hole, skip it */
2522
		if (end < ex_ee_block) {
Allison Henderson's avatar
Allison Henderson committed
2523 2524 2525 2526
			ex--;
			ex_ee_block = le32_to_cpu(ex->ee_block);
			ex_ee_len = ext4_ext_get_actual_len(ex);
			continue;
2527
		} else if (b != ex_ee_block + ex_ee_len - 1) {
2528 2529 2530 2531 2532
			EXT4_ERROR_INODE(inode,
					 "can not handle truncate %u:%u "
					 "on extent %u:%u",
					 start, end, ex_ee_block,
					 ex_ee_block + ex_ee_len - 1);
2533 2534
			err = -EIO;
			goto out;
2535 2536
		} else if (a != ex_ee_block) {
			/* remove tail of the extent */
2537
			num = a - ex_ee_block;
2538 2539 2540 2541
		} else {
			/* remove whole extent: excellent! */
			num = 0;
		}
2542 2543 2544 2545 2546 2547 2548
		/*
		 * 3 for leaf, sb, and inode plus 2 (bmap and group
		 * descriptor) for each block group; assume two block
		 * groups plus ex_ee_len/blocks_per_block_group for
		 * the worst case
		 */
		credits = 7 + 2*(ex_ee_len/EXT4_BLOCKS_PER_GROUP(inode->i_sb));
2549 2550 2551 2552
		if (ex == EXT_FIRST_EXTENT(eh)) {
			correct_index = 1;
			credits += (ext_depth(inode)) + 1;
		}
Dmitry Monakhov's avatar
Dmitry Monakhov committed
2553
		credits += EXT4_MAXQUOTAS_TRANS_BLOCKS(inode->i_sb);
2554

2555
		err = ext4_ext_truncate_extend_restart(handle, inode, credits);
2556
		if (err)
2557 2558 2559 2560 2561 2562
			goto out;

		err = ext4_ext_get_access(handle, inode, path + depth);
		if (err)
			goto out;

2563 2564
		err = ext4_remove_blocks(handle, inode, ex, partial_cluster,
					 a, b);
2565 2566 2567
		if (err)
			goto out;

2568
		if (num == 0)
2569
			/* this extent is removed; mark slot entirely unused */
2570
			ext4_ext_store_pblock(ex, 0);
2571 2572

		ex->ee_len = cpu_to_le16(num);
2573 2574 2575 2576 2577
		/*
		 * Do not mark uninitialized if all the blocks in the
		 * extent have been removed.
		 */
		if (uninitialized && num)
Amit Arora's avatar
Amit Arora committed
2578
			ext4_ext_mark_uninitialized(ex);
Allison Henderson's avatar
Allison Henderson committed
2579 2580 2581 2582 2583
		/*
		 * If the extent was completely released,
		 * we need to remove it from the leaf
		 */
		if (num == 0) {
2584
			if (end != EXT_MAX_BLOCKS - 1) {
Allison Henderson's avatar
Allison Henderson committed
2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597
				/*
				 * For hole punching, we need to scoot all the
				 * extents up when an extent is removed so that
				 * we dont have blank extents in the middle
				 */
				memmove(ex, ex+1, (EXT_LAST_EXTENT(eh) - ex) *
					sizeof(struct ext4_extent));

				/* Now get rid of the one at the end */
				memset(EXT_LAST_EXTENT(eh), 0,
					sizeof(struct ext4_extent));
			}
			le16_add_cpu(&eh->eh_entries, -1);
2598 2599
		} else
			*partial_cluster = 0;
Allison Henderson's avatar
Allison Henderson committed
2600

2601 2602 2603 2604
		err = ext4_ext_dirty(handle, inode, path + depth);
		if (err)
			goto out;

2605
		ext_debug("new extent: %u:%u:%llu\n", ex_ee_block, num,
2606
				ext4_ext_pblock(ex));
2607 2608
		ex--;
		ex_ee_block = le32_to_cpu(ex->ee_block);
Amit Arora's avatar
Amit Arora committed
2609
		ex_ee_len = ext4_ext_get_actual_len(ex);
2610 2611 2612 2613 2614
	}

	if (correct_index && eh->eh_entries)
		err = ext4_ext_correct_indexes(handle, inode, path);

2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633
	/*
	 * If there is still a entry in the leaf node, check to see if
	 * it references the partial cluster.  This is the only place
	 * where it could; if it doesn't, we can free the cluster.
	 */
	if (*partial_cluster && ex >= EXT_FIRST_EXTENT(eh) &&
	    (EXT4_B2C(sbi, ext4_ext_pblock(ex) + ex_ee_len - 1) !=
	     *partial_cluster)) {
		int flags = EXT4_FREE_BLOCKS_FORGET;

		if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
			flags |= EXT4_FREE_BLOCKS_METADATA;

		ext4_free_blocks(handle, inode, NULL,
				 EXT4_C2B(sbi, *partial_cluster),
				 sbi->s_cluster_ratio, flags);
		*partial_cluster = 0;
	}

2634 2635 2636
	/* if this leaf is free, then we should
	 * remove it from index block above */
	if (err == 0 && eh->eh_entries == 0 && path[depth].p_bh != NULL)
2637
		err = ext4_ext_rm_idx(handle, inode, path, depth);
2638 2639 2640 2641 2642 2643

out:
	return err;
}

/*
2644 2645
 * ext4_ext_more_to_rm:
 * returns 1 if current index has to be freed (even partial)
2646
 */
2647
static int
2648 2649 2650 2651 2652 2653 2654 2655
ext4_ext_more_to_rm(struct ext4_ext_path *path)
{
	BUG_ON(path->p_idx == NULL);

	if (path->p_idx < EXT_FIRST_INDEX(path->p_hdr))
		return 0;

	/*
2656
	 * if truncate on deeper level happened, it wasn't partial,
2657 2658 2659 2660 2661 2662 2663
	 * so we have to consider current index for truncation
	 */
	if (le16_to_cpu(path->p_hdr->eh_entries) == path->p_block)
		return 0;
	return 1;
}

2664 2665
int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
			  ext4_lblk_t end)
2666 2667 2668
{
	struct super_block *sb = inode->i_sb;
	int depth = ext_depth(inode);
2669
	struct ext4_ext_path *path = NULL;
2670
	ext4_fsblk_t partial_cluster = 0;
2671
	handle_t *handle;
2672
	int i = 0, err = 0;
2673

2674
	ext_debug("truncate since %u to %u\n", start, end);
2675 2676

	/* probably first extent we're gonna free will be last in block */
2677
	handle = ext4_journal_start(inode, EXT4_HT_TRUNCATE, depth + 1);
2678 2679 2680
	if (IS_ERR(handle))
		return PTR_ERR(handle);

2681
again:
2682 2683
	trace_ext4_ext_remove_space(inode, start, depth);

2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701
	/*
	 * Check if we are removing extents inside the extent tree. If that
	 * is the case, we are going to punch a hole inside the extent tree
	 * so we have to check whether we need to split the extent covering
	 * the last block to remove so we can easily remove the part of it
	 * in ext4_ext_rm_leaf().
	 */
	if (end < EXT_MAX_BLOCKS - 1) {
		struct ext4_extent *ex;
		ext4_lblk_t ee_block;

		/* find extent for this block */
		path = ext4_ext_find_extent(inode, end, NULL);
		if (IS_ERR(path)) {
			ext4_journal_stop(handle);
			return PTR_ERR(path);
		}
		depth = ext_depth(inode);
2702
		/* Leaf not may not exist only if inode has no blocks at all */
2703
		ex = path[depth].p_ext;
2704
		if (!ex) {
2705 2706 2707 2708 2709 2710 2711
			if (depth) {
				EXT4_ERROR_INODE(inode,
						 "path[%d].p_hdr == NULL",
						 depth);
				err = -EIO;
			}
			goto out;
2712
		}
2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731

		ee_block = le32_to_cpu(ex->ee_block);

		/*
		 * See if the last block is inside the extent, if so split
		 * the extent at 'end' block so we can easily remove the
		 * tail of the first part of the split extent in
		 * ext4_ext_rm_leaf().
		 */
		if (end >= ee_block &&
		    end < ee_block + ext4_ext_get_actual_len(ex) - 1) {
			int split_flag = 0;

			if (ext4_ext_is_uninitialized(ex))
				split_flag = EXT4_EXT_MARK_UNINIT1 |
					     EXT4_EXT_MARK_UNINIT2;

			/*
			 * Split the extent in two so that 'end' is the last
2732 2733 2734
			 * block in the first new extent. Also we should not
			 * fail removing space due to ENOSPC so try to use
			 * reserved block if that happens.
2735 2736
			 */
			err = ext4_split_extent_at(handle, inode, path,
2737 2738 2739
					end + 1, split_flag,
					EXT4_GET_BLOCKS_PRE_IO |
					EXT4_GET_BLOCKS_METADATA_NOFAIL);
2740 2741 2742 2743 2744

			if (err < 0)
				goto out;
		}
	}
2745
	/*
2746 2747
	 * We start scanning from right side, freeing all the blocks
	 * after i_size and walking into the tree depth-wise.
2748
	 */
2749
	depth = ext_depth(inode);
2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763
	if (path) {
		int k = i = depth;
		while (--k > 0)
			path[k].p_block =
				le16_to_cpu(path[k].p_hdr->eh_entries)+1;
	} else {
		path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 1),
			       GFP_NOFS);
		if (path == NULL) {
			ext4_journal_stop(handle);
			return -ENOMEM;
		}
		path[0].p_depth = depth;
		path[0].p_hdr = ext_inode_hdr(inode);
2764
		i = 0;
2765

2766 2767 2768 2769
		if (ext4_ext_check(inode, path[0].p_hdr, depth)) {
			err = -EIO;
			goto out;
		}
2770
	}
2771
	err = 0;
2772 2773 2774 2775

	while (i >= 0 && err == 0) {
		if (i == depth) {
			/* this is leaf block */
Allison Henderson's avatar
Allison Henderson committed
2776
			err = ext4_ext_rm_leaf(handle, inode, path,
2777
					       &partial_cluster, start,
2778
					       end);
2779
			/* root level has p_bh == NULL, brelse() eats this */
2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792
			brelse(path[i].p_bh);
			path[i].p_bh = NULL;
			i--;
			continue;
		}

		/* this is index block */
		if (!path[i].p_hdr) {
			ext_debug("initialize header\n");
			path[i].p_hdr = ext_block_hdr(path[i].p_bh);
		}

		if (!path[i].p_idx) {
2793
			/* this level hasn't been touched yet */
2794 2795 2796 2797 2798 2799
			path[i].p_idx = EXT_LAST_INDEX(path[i].p_hdr);
			path[i].p_block = le16_to_cpu(path[i].p_hdr->eh_entries)+1;
			ext_debug("init index ptr: hdr 0x%p, num %d\n",
				  path[i].p_hdr,
				  le16_to_cpu(path[i].p_hdr->eh_entries));
		} else {
2800
			/* we were already here, see at next index */
2801 2802 2803 2804 2805 2806 2807
			path[i].p_idx--;
		}

		ext_debug("level %d - index, first 0x%p, cur 0x%p\n",
				i, EXT_FIRST_INDEX(path[i].p_hdr),
				path[i].p_idx);
		if (ext4_ext_more_to_rm(path + i)) {
2808
			struct buffer_head *bh;
2809
			/* go to the next level */
2810
			ext_debug("move to level %d (block %llu)\n",
2811
				  i + 1, ext4_idx_pblock(path[i].p_idx));
2812
			memset(path + i + 1, 0, sizeof(*path));
2813
			bh = sb_bread(sb, ext4_idx_pblock(path[i].p_idx));
2814
			if (!bh) {
2815 2816 2817 2818
				/* should we reset i_size? */
				err = -EIO;
				break;
			}
2819 2820 2821 2822
			if (WARN_ON(i + 1 > depth)) {
				err = -EIO;
				break;
			}
2823 2824
			if (ext4_ext_check_block(inode, ext_block_hdr(bh),
							depth - i - 1, bh)) {
2825 2826 2827 2828
				err = -EIO;
				break;
			}
			path[i + 1].p_bh = bh;
2829

2830 2831
			/* save actual number of indexes since this
			 * number is changed at the next iteration */
2832 2833 2834
			path[i].p_block = le16_to_cpu(path[i].p_hdr->eh_entries);
			i++;
		} else {
2835
			/* we finished processing this index, go up */
2836
			if (path[i].p_hdr->eh_entries == 0 && i > 0) {
2837
				/* index is empty, remove it;
2838 2839
				 * handle must be already prepared by the
				 * truncatei_leaf() */
2840
				err = ext4_ext_rm_idx(handle, inode, path, i);
2841
			}
2842
			/* root level has p_bh == NULL, brelse() eats this */
2843 2844 2845 2846 2847 2848 2849
			brelse(path[i].p_bh);
			path[i].p_bh = NULL;
			i--;
			ext_debug("return to level %d\n", i);
		}
	}

2850 2851 2852
	trace_ext4_ext_remove_space_done(inode, start, depth, partial_cluster,
			path->p_hdr->eh_entries);

2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867
	/* If we still have something in the partial cluster and we have removed
	 * even the first extent, then we should free the blocks in the partial
	 * cluster as well. */
	if (partial_cluster && path->p_hdr->eh_entries == 0) {
		int flags = EXT4_FREE_BLOCKS_FORGET;

		if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
			flags |= EXT4_FREE_BLOCKS_METADATA;

		ext4_free_blocks(handle, inode, NULL,
				 EXT4_C2B(EXT4_SB(sb), partial_cluster),
				 EXT4_SB(sb)->s_cluster_ratio, flags);
		partial_cluster = 0;
	}

2868 2869 2870
	/* TODO: flexible tree reduction should be here */
	if (path->p_hdr->eh_entries == 0) {
		/*
2871 2872
		 * truncate to zero freed all the tree,
		 * so we need to correct eh_depth
2873 2874 2875 2876 2877
		 */
		err = ext4_ext_get_access(handle, inode, path);
		if (err == 0) {
			ext_inode_hdr(inode)->eh_depth = 0;
			ext_inode_hdr(inode)->eh_max =
2878
				cpu_to_le16(ext4_ext_space_root(inode, 0));
2879 2880 2881 2882 2883 2884
			err = ext4_ext_dirty(handle, inode, path);
		}
	}
out:
	ext4_ext_drop_refs(path);
	kfree(path);
2885 2886
	if (err == -EAGAIN) {
		path = NULL;
2887
		goto again;
2888
	}
2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902
	ext4_journal_stop(handle);

	return err;
}

/*
 * called at mount time
 */
void ext4_ext_init(struct super_block *sb)
{
	/*
	 * possible initialization would be here
	 */

2903
	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2904
#if defined(AGGRESSIVE_TEST) || defined(CHECK_BINSEARCH) || defined(EXTENTS_STATS)
2905
		printk(KERN_INFO "EXT4-fs: file extents enabled"
2906
#ifdef AGGRESSIVE_TEST
2907
		       ", aggressive tests"
2908 2909
#endif
#ifdef CHECK_BINSEARCH
2910
		       ", check binsearch"
2911 2912
#endif
#ifdef EXTENTS_STATS
2913
		       ", stats"
2914
#endif
2915
		       "\n");
2916
#endif
2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929
#ifdef EXTENTS_STATS
		spin_lock_init(&EXT4_SB(sb)->s_ext_stats_lock);
		EXT4_SB(sb)->s_ext_min = 1 << 30;
		EXT4_SB(sb)->s_ext_max = 0;
#endif
	}
}

/*
 * called at umount time
 */
void ext4_ext_release(struct super_block *sb)
{
2930
	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944
		return;

#ifdef EXTENTS_STATS
	if (EXT4_SB(sb)->s_ext_blocks && EXT4_SB(sb)->s_ext_extents) {
		struct ext4_sb_info *sbi = EXT4_SB(sb);
		printk(KERN_ERR "EXT4-fs: %lu blocks in %lu extents (%lu ave)\n",
			sbi->s_ext_blocks, sbi->s_ext_extents,
			sbi->s_ext_blocks / sbi->s_ext_extents);
		printk(KERN_ERR "EXT4-fs: extents: %lu min, %lu max, max depth %lu\n",
			sbi->s_ext_min, sbi->s_ext_max, sbi->s_depth_max);
	}
#endif
}

2945 2946 2947
/* FIXME!! we need to try to merge to left or right after zero-out  */
static int ext4_ext_zeroout(struct inode *inode, struct ext4_extent *ex)
{
2948 2949
	ext4_fsblk_t ee_pblock;
	unsigned int ee_len;
2950
	int ret;
2951 2952

	ee_len    = ext4_ext_get_actual_len(ex);
2953
	ee_pblock = ext4_ext_pblock(ex);
2954

2955
	ret = sb_issue_zeroout(inode->i_sb, ee_pblock, ee_len, GFP_NOFS);
2956 2957
	if (ret > 0)
		ret = 0;
2958

2959
	return ret;
2960 2961
}

2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991
/*
 * ext4_split_extent_at() splits an extent at given block.
 *
 * @handle: the journal handle
 * @inode: the file inode
 * @path: the path to the extent
 * @split: the logical block where the extent is splitted.
 * @split_flags: indicates if the extent could be zeroout if split fails, and
 *		 the states(init or uninit) of new extents.
 * @flags: flags used to insert new extent to extent tree.
 *
 *
 * Splits extent [a, b] into two extents [a, @split) and [@split, b], states
 * of which are deterimined by split_flag.
 *
 * There are two cases:
 *  a> the extent are splitted into two extent.
 *  b> split is not needed, and just mark the extent.
 *
 * return 0 on success.
 */
static int ext4_split_extent_at(handle_t *handle,
			     struct inode *inode,
			     struct ext4_ext_path *path,
			     ext4_lblk_t split,
			     int split_flag,
			     int flags)
{
	ext4_fsblk_t newblock;
	ext4_lblk_t ee_block;
2992
	struct ext4_extent *ex, newex, orig_ex, zero_ex;
2993 2994 2995 2996
	struct ext4_extent *ex2 = NULL;
	unsigned int ee_len, depth;
	int err = 0;

2997 2998 2999
	BUG_ON((split_flag & (EXT4_EXT_DATA_VALID1 | EXT4_EXT_DATA_VALID2)) ==
	       (EXT4_EXT_DATA_VALID1 | EXT4_EXT_DATA_VALID2));

3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011
	ext_debug("ext4_split_extents_at: inode %lu, logical"
		"block %llu\n", inode->i_ino, (unsigned long long)split);

	ext4_ext_show_leaf(inode, path);

	depth = ext_depth(inode);
	ex = path[depth].p_ext;
	ee_block = le32_to_cpu(ex->ee_block);
	ee_len = ext4_ext_get_actual_len(ex);
	newblock = split - ee_block + ext4_ext_pblock(ex);

	BUG_ON(split < ee_block || split >= (ee_block + ee_len));
3012 3013 3014 3015
	BUG_ON(!ext4_ext_is_uninitialized(ex) &&
	       split_flag & (EXT4_EXT_MAY_ZEROOUT |
			     EXT4_EXT_MARK_UNINIT1 |
			     EXT4_EXT_MARK_UNINIT2));
3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032

	err = ext4_ext_get_access(handle, inode, path + depth);
	if (err)
		goto out;

	if (split == ee_block) {
		/*
		 * case b: block @split is the block that the extent begins with
		 * then we just change the state of the extent, and splitting
		 * is not needed.
		 */
		if (split_flag & EXT4_EXT_MARK_UNINIT2)
			ext4_ext_mark_uninitialized(ex);
		else
			ext4_ext_mark_initialized(ex);

		if (!(flags & EXT4_GET_BLOCKS_PRE_IO))
3033
			ext4_ext_try_to_merge(handle, inode, path, ex);
3034

3035
		err = ext4_ext_dirty(handle, inode, path + path->p_depth);
3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061
		goto out;
	}

	/* case a */
	memcpy(&orig_ex, ex, sizeof(orig_ex));
	ex->ee_len = cpu_to_le16(split - ee_block);
	if (split_flag & EXT4_EXT_MARK_UNINIT1)
		ext4_ext_mark_uninitialized(ex);

	/*
	 * path may lead to new leaf, not to original leaf any more
	 * after ext4_ext_insert_extent() returns,
	 */
	err = ext4_ext_dirty(handle, inode, path + depth);
	if (err)
		goto fix_extent_len;

	ex2 = &newex;
	ex2->ee_block = cpu_to_le32(split);
	ex2->ee_len   = cpu_to_le16(ee_len - (split - ee_block));
	ext4_ext_store_pblock(ex2, newblock);
	if (split_flag & EXT4_EXT_MARK_UNINIT2)
		ext4_ext_mark_uninitialized(ex2);

	err = ext4_ext_insert_extent(handle, inode, path, &newex, flags);
	if (err == -ENOSPC && (EXT4_EXT_MAY_ZEROOUT & split_flag)) {
3062
		if (split_flag & (EXT4_EXT_DATA_VALID1|EXT4_EXT_DATA_VALID2)) {
3063
			if (split_flag & EXT4_EXT_DATA_VALID1) {
3064
				err = ext4_ext_zeroout(inode, ex2);
3065
				zero_ex.ee_block = ex2->ee_block;
3066 3067
				zero_ex.ee_len = cpu_to_le16(
						ext4_ext_get_actual_len(ex2));
3068 3069 3070
				ext4_ext_store_pblock(&zero_ex,
						      ext4_ext_pblock(ex2));
			} else {
3071
				err = ext4_ext_zeroout(inode, ex);
3072
				zero_ex.ee_block = ex->ee_block;
3073 3074
				zero_ex.ee_len = cpu_to_le16(
						ext4_ext_get_actual_len(ex));
3075 3076 3077 3078
				ext4_ext_store_pblock(&zero_ex,
						      ext4_ext_pblock(ex));
			}
		} else {
3079
			err = ext4_ext_zeroout(inode, &orig_ex);
3080
			zero_ex.ee_block = orig_ex.ee_block;
3081 3082
			zero_ex.ee_len = cpu_to_le16(
						ext4_ext_get_actual_len(&orig_ex));
3083 3084 3085
			ext4_ext_store_pblock(&zero_ex,
					      ext4_ext_pblock(&orig_ex));
		}
3086

3087 3088 3089
		if (err)
			goto fix_extent_len;
		/* update the extent length and mark as initialized */
3090
		ex->ee_len = cpu_to_le16(ee_len);
3091 3092
		ext4_ext_try_to_merge(handle, inode, path, ex);
		err = ext4_ext_dirty(handle, inode, path + path->p_depth);
3093 3094 3095 3096 3097 3098
		if (err)
			goto fix_extent_len;

		/* update extent status tree */
		err = ext4_es_zeroout(inode, &zero_ex);

3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136
		goto out;
	} else if (err)
		goto fix_extent_len;

out:
	ext4_ext_show_leaf(inode, path);
	return err;

fix_extent_len:
	ex->ee_len = orig_ex.ee_len;
	ext4_ext_dirty(handle, inode, path + depth);
	return err;
}

/*
 * ext4_split_extents() splits an extent and mark extent which is covered
 * by @map as split_flags indicates
 *
 * It may result in splitting the extent into multiple extents (upto three)
 * There are three possibilities:
 *   a> There is no split required
 *   b> Splits in two extents: Split is happening at either end of the extent
 *   c> Splits in three extents: Somone is splitting in middle of the extent
 *
 */
static int ext4_split_extent(handle_t *handle,
			      struct inode *inode,
			      struct ext4_ext_path *path,
			      struct ext4_map_blocks *map,
			      int split_flag,
			      int flags)
{
	ext4_lblk_t ee_block;
	struct ext4_extent *ex;
	unsigned int ee_len, depth;
	int err = 0;
	int uninitialized;
	int split_flag1, flags1;
3137
	int allocated = map->m_len;
3138 3139 3140 3141 3142 3143 3144 3145

	depth = ext_depth(inode);
	ex = path[depth].p_ext;
	ee_block = le32_to_cpu(ex->ee_block);
	ee_len = ext4_ext_get_actual_len(ex);
	uninitialized = ext4_ext_is_uninitialized(ex);

	if (map->m_lblk + map->m_len < ee_block + ee_len) {
3146
		split_flag1 = split_flag & EXT4_EXT_MAY_ZEROOUT;
3147 3148 3149 3150
		flags1 = flags | EXT4_GET_BLOCKS_PRE_IO;
		if (uninitialized)
			split_flag1 |= EXT4_EXT_MARK_UNINIT1 |
				       EXT4_EXT_MARK_UNINIT2;
3151 3152
		if (split_flag & EXT4_EXT_DATA_VALID2)
			split_flag1 |= EXT4_EXT_DATA_VALID1;
3153 3154
		err = ext4_split_extent_at(handle, inode, path,
				map->m_lblk + map->m_len, split_flag1, flags1);
3155 3156
		if (err)
			goto out;
3157 3158
	} else {
		allocated = ee_len - (map->m_lblk - ee_block);
3159
	}
3160 3161 3162 3163
	/*
	 * Update path is required because previous ext4_split_extent_at() may
	 * result in split of original leaf or extent zeroout.
	 */
3164 3165 3166 3167
	ext4_ext_drop_refs(path);
	path = ext4_ext_find_extent(inode, map->m_lblk, path);
	if (IS_ERR(path))
		return PTR_ERR(path);
3168 3169 3170 3171
	depth = ext_depth(inode);
	ex = path[depth].p_ext;
	uninitialized = ext4_ext_is_uninitialized(ex);
	split_flag1 = 0;
3172 3173

	if (map->m_lblk >= ee_block) {
3174 3175
		split_flag1 = split_flag & EXT4_EXT_DATA_VALID2;
		if (uninitialized) {
3176
			split_flag1 |= EXT4_EXT_MARK_UNINIT1;
3177 3178 3179
			split_flag1 |= split_flag & (EXT4_EXT_MAY_ZEROOUT |
						     EXT4_EXT_MARK_UNINIT2);
		}
3180 3181 3182 3183 3184 3185 3186 3187
		err = ext4_split_extent_at(handle, inode, path,
				map->m_lblk, split_flag1, flags);
		if (err)
			goto out;
	}

	ext4_ext_show_leaf(inode, path);
out:
3188
	return err ? err : allocated;
3189 3190
}

3191
/*
3192
 * This function is called by ext4_ext_map_blocks() if someone tries to write
3193
 * to an uninitialized extent. It may result in splitting the uninitialized
Lucas De Marchi's avatar
Lucas De Marchi committed
3194
 * extent into multiple extents (up to three - one initialized and two
3195 3196 3197 3198 3199
 * uninitialized).
 * There are three possibilities:
 *   a> There is no split required: Entire extent should be initialized
 *   b> Splits in two extents: Write is happening at either end of the extent
 *   c> Splits in three extents: Somone is writing in middle of the extent
3200 3201 3202 3203 3204 3205 3206 3207 3208 3209
 *
 * Pre-conditions:
 *  - The extent pointed to by 'path' is uninitialized.
 *  - The extent pointed to by 'path' contains a superset
 *    of the logical span [map->m_lblk, map->m_lblk + map->m_len).
 *
 * Post-conditions on success:
 *  - the returned value is the number of blocks beyond map->l_lblk
 *    that are allocated and initialized.
 *    It is guaranteed to be >= map->m_len.
3210
 */
3211
static int ext4_ext_convert_to_initialized(handle_t *handle,
3212 3213
					   struct inode *inode,
					   struct ext4_map_blocks *map,
3214 3215
					   struct ext4_ext_path *path,
					   int flags)
3216
{
3217
	struct ext4_sb_info *sbi;
3218
	struct ext4_extent_header *eh;
3219 3220
	struct ext4_map_blocks split_map;
	struct ext4_extent zero_ex;
3221
	struct ext4_extent *ex, *abut_ex;
3222
	ext4_lblk_t ee_block, eof_block;
3223 3224
	unsigned int ee_len, depth, map_len = map->m_len;
	int allocated = 0, max_zeroout = 0;
3225
	int err = 0;
3226
	int split_flag = 0;
3227 3228 3229

	ext_debug("ext4_ext_convert_to_initialized: inode %lu, logical"
		"block %llu, max_blocks %u\n", inode->i_ino,
3230
		(unsigned long long)map->m_lblk, map_len);
3231

3232
	sbi = EXT4_SB(inode->i_sb);
3233 3234
	eof_block = (inode->i_size + inode->i_sb->s_blocksize - 1) >>
		inode->i_sb->s_blocksize_bits;
3235 3236
	if (eof_block < map->m_lblk + map_len)
		eof_block = map->m_lblk + map_len;
3237 3238

	depth = ext_depth(inode);
3239
	eh = path[depth].p_hdr;
3240 3241 3242
	ex = path[depth].p_ext;
	ee_block = le32_to_cpu(ex->ee_block);
	ee_len = ext4_ext_get_actual_len(ex);
3243
	zero_ex.ee_len = 0;
3244

3245 3246 3247 3248 3249 3250 3251 3252
	trace_ext4_ext_convert_to_initialized_enter(inode, map, ex);

	/* Pre-conditions */
	BUG_ON(!ext4_ext_is_uninitialized(ex));
	BUG_ON(!in_range(map->m_lblk, ee_block, ee_len));

	/*
	 * Attempt to transfer newly initialized blocks from the currently
3253
	 * uninitialized extent to its neighbor. This is much cheaper
3254
	 * than an insertion followed by a merge as those involve costly
3255 3256 3257
	 * memmove() calls. Transferring to the left is the common case in
	 * steady state for workloads doing fallocate(FALLOC_FL_KEEP_SIZE)
	 * followed by append writes.
3258 3259
	 *
	 * Limitations of the current logic:
3260
	 *  - L1: we do not deal with writes covering the whole extent.
3261 3262
	 *    This would require removing the extent if the transfer
	 *    is possible.
3263
	 *  - L2: we only attempt to merge with an extent stored in the
3264 3265
	 *    same extent tree node.
	 */
3266 3267 3268 3269
	if ((map->m_lblk == ee_block) &&
		/* See if we can merge left */
		(map_len < ee_len) &&		/*L1*/
		(ex > EXT_FIRST_EXTENT(eh))) {	/*L2*/
3270 3271
		ext4_lblk_t prev_lblk;
		ext4_fsblk_t prev_pblk, ee_pblk;
3272
		unsigned int prev_len;
3273

3274 3275 3276 3277
		abut_ex = ex - 1;
		prev_lblk = le32_to_cpu(abut_ex->ee_block);
		prev_len = ext4_ext_get_actual_len(abut_ex);
		prev_pblk = ext4_ext_pblock(abut_ex);
3278 3279 3280
		ee_pblk = ext4_ext_pblock(ex);

		/*
3281
		 * A transfer of blocks from 'ex' to 'abut_ex' is allowed
3282
		 * upon those conditions:
3283 3284 3285 3286
		 * - C1: abut_ex is initialized,
		 * - C2: abut_ex is logically abutting ex,
		 * - C3: abut_ex is physically abutting ex,
		 * - C4: abut_ex can receive the additional blocks without
3287 3288
		 *   overflowing the (initialized) length limit.
		 */
3289
		if ((!ext4_ext_is_uninitialized(abut_ex)) &&		/*C1*/
3290 3291
			((prev_lblk + prev_len) == ee_block) &&		/*C2*/
			((prev_pblk + prev_len) == ee_pblk) &&		/*C3*/
3292
			(prev_len < (EXT_INIT_MAX_LEN - map_len))) {	/*C4*/
3293 3294 3295 3296 3297
			err = ext4_ext_get_access(handle, inode, path + depth);
			if (err)
				goto out;

			trace_ext4_ext_convert_to_initialized_fastpath(inode,
3298
				map, ex, abut_ex);
3299

3300 3301 3302 3303
			/* Shift the start of ex by 'map_len' blocks */
			ex->ee_block = cpu_to_le32(ee_block + map_len);
			ext4_ext_store_pblock(ex, ee_pblk + map_len);
			ex->ee_len = cpu_to_le16(ee_len - map_len);
3304 3305
			ext4_ext_mark_uninitialized(ex); /* Restore the flag */

3306 3307
			/* Extend abut_ex by 'map_len' blocks */
			abut_ex->ee_len = cpu_to_le16(prev_len + map_len);
3308

3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324
			/* Result: number of initialized blocks past m_lblk */
			allocated = map_len;
		}
	} else if (((map->m_lblk + map_len) == (ee_block + ee_len)) &&
		   (map_len < ee_len) &&	/*L1*/
		   ex < EXT_LAST_EXTENT(eh)) {	/*L2*/
		/* See if we can merge right */
		ext4_lblk_t next_lblk;
		ext4_fsblk_t next_pblk, ee_pblk;
		unsigned int next_len;

		abut_ex = ex + 1;
		next_lblk = le32_to_cpu(abut_ex->ee_block);
		next_len = ext4_ext_get_actual_len(abut_ex);
		next_pblk = ext4_ext_pblock(abut_ex);
		ee_pblk = ext4_ext_pblock(ex);
3325

3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353
		/*
		 * A transfer of blocks from 'ex' to 'abut_ex' is allowed
		 * upon those conditions:
		 * - C1: abut_ex is initialized,
		 * - C2: abut_ex is logically abutting ex,
		 * - C3: abut_ex is physically abutting ex,
		 * - C4: abut_ex can receive the additional blocks without
		 *   overflowing the (initialized) length limit.
		 */
		if ((!ext4_ext_is_uninitialized(abut_ex)) &&		/*C1*/
		    ((map->m_lblk + map_len) == next_lblk) &&		/*C2*/
		    ((ee_pblk + ee_len) == next_pblk) &&		/*C3*/
		    (next_len < (EXT_INIT_MAX_LEN - map_len))) {	/*C4*/
			err = ext4_ext_get_access(handle, inode, path + depth);
			if (err)
				goto out;

			trace_ext4_ext_convert_to_initialized_fastpath(inode,
				map, ex, abut_ex);

			/* Shift the start of abut_ex by 'map_len' blocks */
			abut_ex->ee_block = cpu_to_le32(next_lblk - map_len);
			ext4_ext_store_pblock(abut_ex, next_pblk - map_len);
			ex->ee_len = cpu_to_le16(ee_len - map_len);
			ext4_ext_mark_uninitialized(ex); /* Restore the flag */

			/* Extend abut_ex by 'map_len' blocks */
			abut_ex->ee_len = cpu_to_le16(next_len + map_len);
3354 3355

			/* Result: number of initialized blocks past m_lblk */
3356
			allocated = map_len;
3357 3358
		}
	}
3359 3360 3361 3362 3363 3364 3365 3366 3367
	if (allocated) {
		/* Mark the block containing both extents as dirty */
		ext4_ext_dirty(handle, inode, path + depth);

		/* Update path to point to the right extent */
		path[depth].p_ext = abut_ex;
		goto out;
	} else
		allocated = ee_len - (map->m_lblk - ee_block);
3368

3369
	WARN_ON(map->m_lblk < ee_block);
3370 3371 3372 3373
	/*
	 * It is safe to convert extent to initialized via explicit
	 * zeroout only if extent is fully insde i_size or new_size.
	 */
3374
	split_flag |= ee_block + ee_len <= eof_block ? EXT4_EXT_MAY_ZEROOUT : 0;
3375

3376 3377
	if (EXT4_EXT_MAY_ZEROOUT & split_flag)
		max_zeroout = sbi->s_extent_max_zeroout_kb >>
3378
			(inode->i_sb->s_blocksize_bits - 10);
3379 3380 3381

	/* If extent is less than s_max_zeroout_kb, zeroout directly */
	if (max_zeroout && (ee_len <= max_zeroout)) {
3382
		err = ext4_ext_zeroout(inode, ex);
3383
		if (err)
3384
			goto out;
3385
		zero_ex.ee_block = ex->ee_block;
3386
		zero_ex.ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex));
3387
		ext4_ext_store_pblock(&zero_ex, ext4_ext_pblock(ex));
3388 3389 3390 3391

		err = ext4_ext_get_access(handle, inode, path + depth);
		if (err)
			goto out;
3392
		ext4_ext_mark_initialized(ex);
3393 3394
		ext4_ext_try_to_merge(handle, inode, path, ex);
		err = ext4_ext_dirty(handle, inode, path + path->p_depth);
3395
		goto out;
3396
	}
3397

3398
	/*
3399 3400 3401 3402 3403
	 * four cases:
	 * 1. split the extent into three extents.
	 * 2. split the extent into two extents, zeroout the first half.
	 * 3. split the extent into two extents, zeroout the second half.
	 * 4. split the extent into two extents with out zeroout.
3404
	 */
3405 3406 3407
	split_map.m_lblk = map->m_lblk;
	split_map.m_len = map->m_len;

3408 3409
	if (max_zeroout && (allocated > map->m_len)) {
		if (allocated <= max_zeroout) {
3410 3411
			/* case 3 */
			zero_ex.ee_block =
3412 3413
					 cpu_to_le32(map->m_lblk);
			zero_ex.ee_len = cpu_to_le16(allocated);
3414 3415 3416
			ext4_ext_store_pblock(&zero_ex,
				ext4_ext_pblock(ex) + map->m_lblk - ee_block);
			err = ext4_ext_zeroout(inode, &zero_ex);
3417 3418
			if (err)
				goto out;
3419 3420
			split_map.m_lblk = map->m_lblk;
			split_map.m_len = allocated;
3421
		} else if (map->m_lblk - ee_block + map->m_len < max_zeroout) {
3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434
			/* case 2 */
			if (map->m_lblk != ee_block) {
				zero_ex.ee_block = ex->ee_block;
				zero_ex.ee_len = cpu_to_le16(map->m_lblk -
							ee_block);
				ext4_ext_store_pblock(&zero_ex,
						      ext4_ext_pblock(ex));
				err = ext4_ext_zeroout(inode, &zero_ex);
				if (err)
					goto out;
			}

			split_map.m_lblk = ee_block;
3435 3436
			split_map.m_len = map->m_lblk - ee_block + map->m_len;
			allocated = map->m_len;
3437 3438
		}
	}
3439 3440

	allocated = ext4_split_extent(handle, inode, path,
3441
				      &split_map, split_flag, flags);
3442 3443 3444
	if (allocated < 0)
		err = allocated;

3445
out:
3446 3447 3448
	/* If we have gotten a failure, don't zero out status tree */
	if (!err)
		err = ext4_es_zeroout(inode, &zero_ex);
3449 3450 3451
	return err ? err : allocated;
}

3452
/*
3453
 * This function is called by ext4_ext_map_blocks() from
3454 3455 3456
 * ext4_get_blocks_dio_write() when DIO to write
 * to an uninitialized extent.
 *
3457
 * Writing to an uninitialized extent may result in splitting the uninitialized
3458
 * extent into multiple initialized/uninitialized extents (up to three)
3459 3460 3461 3462 3463 3464
 * There are three possibilities:
 *   a> There is no split required: Entire extent should be uninitialized
 *   b> Splits in two extents: Write is happening at either end of the extent
 *   c> Splits in three extents: Somone is writing in middle of the extent
 *
 * One of more index blocks maybe needed if the extent tree grow after
3465
 * the uninitialized extent split. To prevent ENOSPC occur at the IO
3466
 * complete, we need to split the uninitialized extent before DIO submit
3467
 * the IO. The uninitialized extent called at this time will be split
3468 3469 3470
 * into three uninitialized extent(at most). After IO complete, the part
 * being filled will be convert to initialized by the end_io callback function
 * via ext4_convert_unwritten_extents().
3471 3472
 *
 * Returns the size of uninitialized extent to be written on success.
3473 3474 3475
 */
static int ext4_split_unwritten_extents(handle_t *handle,
					struct inode *inode,
3476
					struct ext4_map_blocks *map,
3477 3478 3479
					struct ext4_ext_path *path,
					int flags)
{
3480 3481 3482 3483 3484
	ext4_lblk_t eof_block;
	ext4_lblk_t ee_block;
	struct ext4_extent *ex;
	unsigned int ee_len;
	int split_flag = 0, depth;
3485 3486 3487

	ext_debug("ext4_split_unwritten_extents: inode %lu, logical"
		"block %llu, max_blocks %u\n", inode->i_ino,
3488
		(unsigned long long)map->m_lblk, map->m_len);
3489 3490 3491

	eof_block = (inode->i_size + inode->i_sb->s_blocksize - 1) >>
		inode->i_sb->s_blocksize_bits;
3492 3493
	if (eof_block < map->m_lblk + map->m_len)
		eof_block = map->m_lblk + map->m_len;
3494 3495 3496 3497
	/*
	 * It is safe to convert extent to initialized via explicit
	 * zeroout only if extent is fully insde i_size or new_size.
	 */
3498 3499 3500 3501
	depth = ext_depth(inode);
	ex = path[depth].p_ext;
	ee_block = le32_to_cpu(ex->ee_block);
	ee_len = ext4_ext_get_actual_len(ex);
3502

3503 3504
	split_flag |= ee_block + ee_len <= eof_block ? EXT4_EXT_MAY_ZEROOUT : 0;
	split_flag |= EXT4_EXT_MARK_UNINIT2;
3505 3506
	if (flags & EXT4_GET_BLOCKS_CONVERT)
		split_flag |= EXT4_EXT_DATA_VALID2;
3507 3508
	flags |= EXT4_GET_BLOCKS_PRE_IO;
	return ext4_split_extent(handle, inode, path, map, split_flag, flags);
3509
}
3510

3511
static int ext4_convert_unwritten_extents_endio(handle_t *handle,
3512 3513 3514
						struct inode *inode,
						struct ext4_map_blocks *map,
						struct ext4_ext_path *path)
3515 3516
{
	struct ext4_extent *ex;
3517 3518
	ext4_lblk_t ee_block;
	unsigned int ee_len;
3519 3520 3521 3522 3523
	int depth;
	int err = 0;

	depth = ext_depth(inode);
	ex = path[depth].p_ext;
3524 3525
	ee_block = le32_to_cpu(ex->ee_block);
	ee_len = ext4_ext_get_actual_len(ex);
3526

3527 3528
	ext_debug("ext4_convert_unwritten_extents_endio: inode %lu, logical"
		"block %llu, max_blocks %u\n", inode->i_ino,
3529 3530
		  (unsigned long long)ee_block, ee_len);

3531 3532 3533 3534 3535 3536
	/* If extent is larger than requested it is a clear sign that we still
	 * have some extent state machine issues left. So extent_split is still
	 * required.
	 * TODO: Once all related issues will be fixed this situation should be
	 * illegal.
	 */
3537
	if (ee_block != map->m_lblk || ee_len > map->m_len) {
3538 3539 3540 3541 3542 3543
#ifdef EXT4_DEBUG
		ext4_warning("Inode (%ld) finished: extent logical block %llu,"
			     " len %u; IO logical block %llu, len %u\n",
			     inode->i_ino, (unsigned long long)ee_block, ee_len,
			     (unsigned long long)map->m_lblk, map->m_len);
#endif
3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556
		err = ext4_split_unwritten_extents(handle, inode, map, path,
						   EXT4_GET_BLOCKS_CONVERT);
		if (err < 0)
			goto out;
		ext4_ext_drop_refs(path);
		path = ext4_ext_find_extent(inode, map->m_lblk, path);
		if (IS_ERR(path)) {
			err = PTR_ERR(path);
			goto out;
		}
		depth = ext_depth(inode);
		ex = path[depth].p_ext;
	}
3557

3558 3559 3560 3561 3562 3563
	err = ext4_ext_get_access(handle, inode, path + depth);
	if (err)
		goto out;
	/* first mark the extent as initialized */
	ext4_ext_mark_initialized(ex);

3564 3565
	/* note: ext4_ext_correct_indexes() isn't needed here because
	 * borders are not changed
3566
	 */
3567
	ext4_ext_try_to_merge(handle, inode, path, ex);
3568

3569
	/* Mark modified extent as dirty */
3570
	err = ext4_ext_dirty(handle, inode, path + path->p_depth);
3571 3572 3573 3574 3575
out:
	ext4_ext_show_leaf(inode, path);
	return err;
}

3576 3577 3578 3579 3580 3581 3582 3583
static void unmap_underlying_metadata_blocks(struct block_device *bdev,
			sector_t block, int count)
{
	int i;
	for (i = 0; i < count; i++)
                unmap_underlying_metadata(bdev, block + i);
}

3584 3585 3586 3587
/*
 * Handle EOFBLOCKS_FL flag, clearing it if necessary
 */
static int check_eofblocks_fl(handle_t *handle, struct inode *inode,
3588
			      ext4_lblk_t lblk,
3589 3590 3591 3592 3593
			      struct ext4_ext_path *path,
			      unsigned int len)
{
	int i, depth;
	struct ext4_extent_header *eh;
3594
	struct ext4_extent *last_ex;
3595 3596 3597 3598 3599 3600 3601

	if (!ext4_test_inode_flag(inode, EXT4_INODE_EOFBLOCKS))
		return 0;

	depth = ext_depth(inode);
	eh = path[depth].p_hdr;

3602 3603 3604 3605 3606 3607 3608
	/*
	 * We're going to remove EOFBLOCKS_FL entirely in future so we
	 * do not care for this case anymore. Simply remove the flag
	 * if there are no extents.
	 */
	if (unlikely(!eh->eh_entries))
		goto out;
3609 3610 3611 3612 3613 3614 3615 3616 3617 3618
	last_ex = EXT_LAST_EXTENT(eh);
	/*
	 * We should clear the EOFBLOCKS_FL flag if we are writing the
	 * last block in the last extent in the file.  We test this by
	 * first checking to see if the caller to
	 * ext4_ext_get_blocks() was interested in the last block (or
	 * a block beyond the last block) in the current extent.  If
	 * this turns out to be false, we can bail out from this
	 * function immediately.
	 */
3619
	if (lblk + len < le32_to_cpu(last_ex->ee_block) +
3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631
	    ext4_ext_get_actual_len(last_ex))
		return 0;
	/*
	 * If the caller does appear to be planning to write at or
	 * beyond the end of the current extent, we then test to see
	 * if the current extent is the last extent in the file, by
	 * checking to make sure it was reached via the rightmost node
	 * at each level of the tree.
	 */
	for (i = depth-1; i >= 0; i--)
		if (path[i].p_idx != EXT_LAST_INDEX(path[i].p_hdr))
			return 0;
3632
out:
3633 3634 3635 3636
	ext4_clear_inode_flag(inode, EXT4_INODE_EOFBLOCKS);
	return ext4_mark_inode_dirty(handle, inode);
}

3637 3638 3639
/**
 * ext4_find_delalloc_range: find delayed allocated block in the given range.
 *
3640
 * Return 1 if there is a delalloc block in the range, otherwise 0.
3641
 */
3642 3643 3644
int ext4_find_delalloc_range(struct inode *inode,
			     ext4_lblk_t lblk_start,
			     ext4_lblk_t lblk_end)
3645
{
3646
	struct extent_status es;
3647

3648
	ext4_es_find_delayed_extent(inode, lblk_start, &es);
Zheng Liu's avatar
Zheng Liu committed
3649
	if (es.es_len == 0)
3650
		return 0; /* there is no delay extent in this tree */
Zheng Liu's avatar
Zheng Liu committed
3651 3652
	else if (es.es_lblk <= lblk_start &&
		 lblk_start < es.es_lblk + es.es_len)
3653
		return 1;
Zheng Liu's avatar
Zheng Liu committed
3654
	else if (lblk_start <= es.es_lblk && es.es_lblk <= lblk_end)
3655
		return 1;
3656
	else
3657
		return 0;
3658 3659
}

3660
int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk)
3661 3662 3663 3664 3665 3666
{
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
	ext4_lblk_t lblk_start, lblk_end;
	lblk_start = lblk & (~(sbi->s_cluster_ratio - 1));
	lblk_end = lblk_start + sbi->s_cluster_ratio - 1;

3667
	return ext4_find_delalloc_range(inode, lblk_start, lblk_end);
3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719
}

/**
 * Determines how many complete clusters (out of those specified by the 'map')
 * are under delalloc and were reserved quota for.
 * This function is called when we are writing out the blocks that were
 * originally written with their allocation delayed, but then the space was
 * allocated using fallocate() before the delayed allocation could be resolved.
 * The cases to look for are:
 * ('=' indicated delayed allocated blocks
 *  '-' indicates non-delayed allocated blocks)
 * (a) partial clusters towards beginning and/or end outside of allocated range
 *     are not delalloc'ed.
 *	Ex:
 *	|----c---=|====c====|====c====|===-c----|
 *	         |++++++ allocated ++++++|
 *	==> 4 complete clusters in above example
 *
 * (b) partial cluster (outside of allocated range) towards either end is
 *     marked for delayed allocation. In this case, we will exclude that
 *     cluster.
 *	Ex:
 *	|----====c========|========c========|
 *	     |++++++ allocated ++++++|
 *	==> 1 complete clusters in above example
 *
 *	Ex:
 *	|================c================|
 *            |++++++ allocated ++++++|
 *	==> 0 complete clusters in above example
 *
 * The ext4_da_update_reserve_space will be called only if we
 * determine here that there were some "entire" clusters that span
 * this 'allocated' range.
 * In the non-bigalloc case, this function will just end up returning num_blks
 * without ever calling ext4_find_delalloc_range.
 */
static unsigned int
get_reserved_cluster_alloc(struct inode *inode, ext4_lblk_t lblk_start,
			   unsigned int num_blks)
{
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
	ext4_lblk_t alloc_cluster_start, alloc_cluster_end;
	ext4_lblk_t lblk_from, lblk_to, c_offset;
	unsigned int allocated_clusters = 0;

	alloc_cluster_start = EXT4_B2C(sbi, lblk_start);
	alloc_cluster_end = EXT4_B2C(sbi, lblk_start + num_blks - 1);

	/* max possible clusters for this allocation */
	allocated_clusters = alloc_cluster_end - alloc_cluster_start + 1;

3720 3721
	trace_ext4_get_reserved_cluster_alloc(inode, lblk_start, num_blks);

3722 3723 3724 3725 3726 3727
	/* Check towards left side */
	c_offset = lblk_start & (sbi->s_cluster_ratio - 1);
	if (c_offset) {
		lblk_from = lblk_start & (~(sbi->s_cluster_ratio - 1));
		lblk_to = lblk_from + c_offset - 1;

3728
		if (ext4_find_delalloc_range(inode, lblk_from, lblk_to))
3729 3730 3731 3732 3733 3734 3735 3736 3737
			allocated_clusters--;
	}

	/* Now check towards right. */
	c_offset = (lblk_start + num_blks) & (sbi->s_cluster_ratio - 1);
	if (allocated_clusters && c_offset) {
		lblk_from = lblk_start + num_blks;
		lblk_to = lblk_from + (sbi->s_cluster_ratio - c_offset) - 1;

3738
		if (ext4_find_delalloc_range(inode, lblk_from, lblk_to))
3739 3740 3741 3742 3743 3744
			allocated_clusters--;
	}

	return allocated_clusters;
}

3745 3746
static int
ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode,
3747
			struct ext4_map_blocks *map,
3748
			struct ext4_ext_path *path, int flags,
3749
			unsigned int allocated, ext4_fsblk_t newblock)
3750 3751 3752
{
	int ret = 0;
	int err = 0;
Dmitry Monakhov's avatar
Dmitry Monakhov committed
3753
	ext4_io_end_t *io = ext4_inode_aio(inode);
3754

3755 3756
	ext_debug("ext4_ext_handle_uninitialized_extents: inode %lu, logical "
		  "block %llu, max_blocks %u, flags %x, allocated %u\n",
3757
		  inode->i_ino, (unsigned long long)map->m_lblk, map->m_len,
3758 3759 3760
		  flags, allocated);
	ext4_ext_show_leaf(inode, path);

3761 3762 3763 3764 3765 3766
	/*
	 * When writing into uninitialized space, we should not fail to
	 * allocate metadata blocks for the new extent block if needed.
	 */
	flags |= EXT4_GET_BLOCKS_METADATA_NOFAIL;

3767 3768
	trace_ext4_ext_handle_uninitialized_extents(inode, map, flags,
						    allocated, newblock);
3769

3770
	/* get_block() before submit the IO, split the extent */
3771
	if ((flags & EXT4_GET_BLOCKS_PRE_IO)) {
3772 3773
		ret = ext4_split_unwritten_extents(handle, inode, map,
						   path, flags);
3774 3775
		if (ret <= 0)
			goto out;
3776 3777
		/*
		 * Flag the inode(non aio case) or end_io struct (aio case)
Lucas De Marchi's avatar
Lucas De Marchi committed
3778
		 * that this IO needs to conversion to written when IO is
3779 3780
		 * completed
		 */
3781 3782 3783
		if (io)
			ext4_set_io_unwritten_flag(inode, io);
		else
3784
			ext4_set_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);
3785
		map->m_flags |= EXT4_MAP_UNWRITTEN;
3786
		if (ext4_should_dioread_nolock(inode))
3787
			map->m_flags |= EXT4_MAP_UNINIT;
3788 3789
		goto out;
	}
3790
	/* IO end_io complete, convert the filled extent to written */
3791
	if ((flags & EXT4_GET_BLOCKS_CONVERT)) {
3792
		ret = ext4_convert_unwritten_extents_endio(handle, inode, map,
3793
							path);
3794
		if (ret >= 0) {
3795
			ext4_update_inode_fsync_trans(handle, inode, 1);
3796 3797
			err = check_eofblocks_fl(handle, inode, map->m_lblk,
						 path, map->m_len);
3798 3799
		} else
			err = ret;
3800 3801 3802 3803
		map->m_flags |= EXT4_MAP_MAPPED;
		if (allocated > map->m_len)
			allocated = map->m_len;
		map->m_len = allocated;
3804 3805 3806 3807 3808 3809 3810
		goto out2;
	}
	/* buffered IO case */
	/*
	 * repeat fallocate creation request
	 * we already have an unwritten extent
	 */
3811 3812
	if (flags & EXT4_GET_BLOCKS_UNINIT_EXT) {
		map->m_flags |= EXT4_MAP_UNWRITTEN;
3813
		goto map_out;
3814
	}
3815 3816 3817 3818 3819 3820 3821 3822 3823 3824

	/* buffered READ or buffered write_begin() lookup */
	if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) {
		/*
		 * We have blocks reserved already.  We
		 * return allocated blocks so that delalloc
		 * won't do block reservation for us.  But
		 * the buffer head will be unmapped so that
		 * a read from the block returns 0s.
		 */
3825
		map->m_flags |= EXT4_MAP_UNWRITTEN;
3826 3827 3828 3829
		goto out1;
	}

	/* buffered write, writepage time, convert*/
3830
	ret = ext4_ext_convert_to_initialized(handle, inode, map, path, flags);
3831
	if (ret >= 0)
3832
		ext4_update_inode_fsync_trans(handle, inode, 1);
3833 3834 3835 3836 3837 3838
out:
	if (ret <= 0) {
		err = ret;
		goto out2;
	} else
		allocated = ret;
3839
	map->m_flags |= EXT4_MAP_NEW;
3840 3841 3842 3843 3844 3845 3846
	/*
	 * if we allocated more blocks than requested
	 * we need to make sure we unmap the extra block
	 * allocated. The actual needed block will get
	 * unmapped later when we find the buffer_head marked
	 * new.
	 */
3847
	if (allocated > map->m_len) {
3848
		unmap_underlying_metadata_blocks(inode->i_sb->s_bdev,
3849 3850 3851
					newblock + map->m_len,
					allocated - map->m_len);
		allocated = map->m_len;
3852
	}
3853
	map->m_len = allocated;
3854 3855 3856 3857 3858 3859 3860 3861

	/*
	 * If we have done fallocate with the offset that is already
	 * delayed allocated, we would have block reservation
	 * and quota reservation done in the delayed write path.
	 * But fallocate would have already updated quota and block
	 * count for this offset. So cancel these reservation
	 */
3862 3863 3864 3865 3866 3867 3868 3869 3870
	if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE) {
		unsigned int reserved_clusters;
		reserved_clusters = get_reserved_cluster_alloc(inode,
				map->m_lblk, map->m_len);
		if (reserved_clusters)
			ext4_da_update_reserve_space(inode,
						     reserved_clusters,
						     0);
	}
3871

3872
map_out:
3873
	map->m_flags |= EXT4_MAP_MAPPED;
3874 3875 3876 3877 3878 3879
	if ((flags & EXT4_GET_BLOCKS_KEEP_SIZE) == 0) {
		err = check_eofblocks_fl(handle, inode, map->m_lblk, path,
					 map->m_len);
		if (err < 0)
			goto out2;
	}
3880
out1:
3881 3882
	if (allocated > map->m_len)
		allocated = map->m_len;
3883
	ext4_ext_show_leaf(inode, path);
3884 3885
	map->m_pblk = newblock;
	map->m_len = allocated;
3886 3887 3888 3889 3890 3891 3892
out2:
	if (path) {
		ext4_ext_drop_refs(path);
		kfree(path);
	}
	return err ? err : allocated;
}
3893

3894 3895 3896 3897
/*
 * get_implied_cluster_alloc - check to see if the requested
 * allocation (in the map structure) overlaps with a cluster already
 * allocated in an extent.
3898
 *	@sb	The filesystem superblock structure
3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934
 *	@map	The requested lblk->pblk mapping
 *	@ex	The extent structure which might contain an implied
 *			cluster allocation
 *
 * This function is called by ext4_ext_map_blocks() after we failed to
 * find blocks that were already in the inode's extent tree.  Hence,
 * we know that the beginning of the requested region cannot overlap
 * the extent from the inode's extent tree.  There are three cases we
 * want to catch.  The first is this case:
 *
 *		 |--- cluster # N--|
 *    |--- extent ---|	|---- requested region ---|
 *			|==========|
 *
 * The second case that we need to test for is this one:
 *
 *   |--------- cluster # N ----------------|
 *	   |--- requested region --|   |------- extent ----|
 *	   |=======================|
 *
 * The third case is when the requested region lies between two extents
 * within the same cluster:
 *          |------------- cluster # N-------------|
 * |----- ex -----|                  |---- ex_right ----|
 *                  |------ requested region ------|
 *                  |================|
 *
 * In each of the above cases, we need to set the map->m_pblk and
 * map->m_len so it corresponds to the return the extent labelled as
 * "|====|" from cluster #N, since it is already in use for data in
 * cluster EXT4_B2C(sbi, map->m_lblk).	We will then return 1 to
 * signal to ext4_ext_map_blocks() that map->m_pblk should be treated
 * as a new "allocated" block region.  Otherwise, we will return 0 and
 * ext4_ext_map_blocks() will then allocate one or more new clusters
 * by calling ext4_mb_new_blocks().
 */
3935
static int get_implied_cluster_alloc(struct super_block *sb,
3936 3937 3938 3939
				     struct ext4_map_blocks *map,
				     struct ext4_extent *ex,
				     struct ext4_ext_path *path)
{
3940
	struct ext4_sb_info *sbi = EXT4_SB(sb);
3941 3942
	ext4_lblk_t c_offset = map->m_lblk & (sbi->s_cluster_ratio-1);
	ext4_lblk_t ex_cluster_start, ex_cluster_end;
3943
	ext4_lblk_t rr_cluster_start;
3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987
	ext4_lblk_t ee_block = le32_to_cpu(ex->ee_block);
	ext4_fsblk_t ee_start = ext4_ext_pblock(ex);
	unsigned short ee_len = ext4_ext_get_actual_len(ex);

	/* The extent passed in that we are trying to match */
	ex_cluster_start = EXT4_B2C(sbi, ee_block);
	ex_cluster_end = EXT4_B2C(sbi, ee_block + ee_len - 1);

	/* The requested region passed into ext4_map_blocks() */
	rr_cluster_start = EXT4_B2C(sbi, map->m_lblk);

	if ((rr_cluster_start == ex_cluster_end) ||
	    (rr_cluster_start == ex_cluster_start)) {
		if (rr_cluster_start == ex_cluster_end)
			ee_start += ee_len - 1;
		map->m_pblk = (ee_start & ~(sbi->s_cluster_ratio - 1)) +
			c_offset;
		map->m_len = min(map->m_len,
				 (unsigned) sbi->s_cluster_ratio - c_offset);
		/*
		 * Check for and handle this case:
		 *
		 *   |--------- cluster # N-------------|
		 *		       |------- extent ----|
		 *	   |--- requested region ---|
		 *	   |===========|
		 */

		if (map->m_lblk < ee_block)
			map->m_len = min(map->m_len, ee_block - map->m_lblk);

		/*
		 * Check for the case where there is already another allocated
		 * block to the right of 'ex' but before the end of the cluster.
		 *
		 *          |------------- cluster # N-------------|
		 * |----- ex -----|                  |---- ex_right ----|
		 *                  |------ requested region ------|
		 *                  |================|
		 */
		if (map->m_lblk > ee_block) {
			ext4_lblk_t next = ext4_ext_next_allocated_block(path);
			map->m_len = min(map->m_len, next - map->m_lblk);
		}
3988 3989

		trace_ext4_get_implied_cluster_alloc_exit(sb, map, 1);
3990 3991
		return 1;
	}
3992 3993

	trace_ext4_get_implied_cluster_alloc_exit(sb, map, 0);
3994 3995 3996 3997
	return 0;
}


3998
/*
3999 4000 4001
 * Block allocation/map/preallocation routine for extents based files
 *
 *
4002
 * Need to be called with
4003 4004
 * down_read(&EXT4_I(inode)->i_data_sem) if not allocating file system block
 * (ie, create is zero). Otherwise down_write(&EXT4_I(inode)->i_data_sem)
4005 4006 4007 4008 4009 4010 4011 4012 4013 4014
 *
 * return > 0, number of of blocks already mapped/allocated
 *          if create == 0 and these are pre-allocated blocks
 *          	buffer head is unmapped
 *          otherwise blocks are mapped
 *
 * return = 0, if plain look up failed (blocks have not been allocated)
 *          buffer head is unmapped
 *
 * return < 0, error case.
4015
 */
4016 4017
int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
			struct ext4_map_blocks *map, int flags)
4018 4019
{
	struct ext4_ext_path *path = NULL;
4020 4021
	struct ext4_extent newex, *ex, *ex2;
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
4022
	ext4_fsblk_t newblock = 0;
4023
	int free_on_err = 0, err = 0, depth;
4024
	unsigned int allocated = 0, offset = 0;
4025
	unsigned int allocated_clusters = 0;
4026
	struct ext4_allocation_request ar;
Dmitry Monakhov's avatar
Dmitry Monakhov committed
4027
	ext4_io_end_t *io = ext4_inode_aio(inode);
4028
	ext4_lblk_t cluster_offset;
4029
	int set_unwritten = 0;
4030

4031
	ext_debug("blocks %u/%u requested for inode %lu\n",
4032
		  map->m_lblk, map->m_len, inode->i_ino);
4033
	trace_ext4_ext_map_blocks_enter(inode, map->m_lblk, map->m_len, flags);
4034 4035

	/* find extent for this block */
4036
	path = ext4_ext_find_extent(inode, map->m_lblk, NULL);
4037 4038 4039 4040 4041 4042 4043 4044 4045
	if (IS_ERR(path)) {
		err = PTR_ERR(path);
		path = NULL;
		goto out2;
	}

	depth = ext_depth(inode);

	/*
4046 4047
	 * consistent leaf must not be empty;
	 * this situation is possible, though, _during_ tree modification;
4048 4049
	 * this is why assert can't be put in ext4_ext_find_extent()
	 */
4050 4051
	if (unlikely(path[depth].p_ext == NULL && depth != 0)) {
		EXT4_ERROR_INODE(inode, "bad extent address "
4052 4053 4054
				 "lblock: %lu, depth: %d pblock %lld",
				 (unsigned long) map->m_lblk, depth,
				 path[depth].p_block);
4055 4056 4057
		err = -EIO;
		goto out2;
	}
4058

4059 4060
	ex = path[depth].p_ext;
	if (ex) {
4061
		ext4_lblk_t ee_block = le32_to_cpu(ex->ee_block);
4062
		ext4_fsblk_t ee_start = ext4_ext_pblock(ex);
Amit Arora's avatar
Amit Arora committed
4063
		unsigned short ee_len;
4064 4065 4066

		/*
		 * Uninitialized extents are treated as holes, except that
4067
		 * we split out initialized portions during a write.
4068
		 */
Amit Arora's avatar
Amit Arora committed
4069
		ee_len = ext4_ext_get_actual_len(ex);
4070 4071 4072

		trace_ext4_ext_show_extent(inode, ee_block, ee_start, ee_len);

4073
		/* if found extent covers block, simply return it */
4074 4075
		if (in_range(map->m_lblk, ee_block, ee_len)) {
			newblock = map->m_lblk - ee_block + ee_start;
4076
			/* number of remaining blocks in the extent */
4077 4078 4079
			allocated = ee_len - (map->m_lblk - ee_block);
			ext_debug("%u fit into %u:%d -> %llu\n", map->m_lblk,
				  ee_block, ee_len, newblock);
4080

Zheng Liu's avatar
Zheng Liu committed
4081
			if (!ext4_ext_is_uninitialized(ex))
4082
				goto out;
Zheng Liu's avatar
Zheng Liu committed
4083

4084
			allocated = ext4_ext_handle_uninitialized_extents(
4085 4086
				handle, inode, map, path, flags,
				allocated, newblock);
4087
			goto out3;
4088 4089 4090
		}
	}

4091
	if ((sbi->s_cluster_ratio > 1) &&
4092
	    ext4_find_delalloc_cluster(inode, map->m_lblk))
4093 4094
		map->m_flags |= EXT4_MAP_FROM_CLUSTER;

4095
	/*
4096
	 * requested block isn't allocated yet;
4097 4098
	 * we couldn't try to create block if create flag is zero
	 */
4099
	if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) {
4100 4101 4102 4103
		/*
		 * put just found gap into cache to speed up
		 * subsequent requests
		 */
4104 4105
		if ((flags & EXT4_GET_BLOCKS_NO_PUT_HOLE) == 0)
			ext4_ext_put_gap_in_cache(inode, path, map->m_lblk);
4106 4107
		goto out2;
	}
4108

4109
	/*
4110
	 * Okay, we need to do block allocation.
4111
	 */
4112
	map->m_flags &= ~EXT4_MAP_FROM_CLUSTER;
4113 4114 4115 4116 4117 4118 4119 4120
	newex.ee_block = cpu_to_le32(map->m_lblk);
	cluster_offset = map->m_lblk & (sbi->s_cluster_ratio-1);

	/*
	 * If we are doing bigalloc, check to see if the extent returned
	 * by ext4_ext_find_extent() implies a cluster we can use.
	 */
	if (cluster_offset && ex &&
4121
	    get_implied_cluster_alloc(inode->i_sb, map, ex, path)) {
4122 4123
		ar.len = allocated = map->m_len;
		newblock = map->m_pblk;
4124
		map->m_flags |= EXT4_MAP_FROM_CLUSTER;
4125 4126
		goto got_allocated_blocks;
	}
4127

4128
	/* find neighbour allocated blocks */
4129
	ar.lleft = map->m_lblk;
4130 4131 4132
	err = ext4_ext_search_left(inode, path, &ar.lleft, &ar.pleft);
	if (err)
		goto out2;
4133
	ar.lright = map->m_lblk;
4134 4135
	ex2 = NULL;
	err = ext4_ext_search_right(inode, path, &ar.lright, &ar.pright, &ex2);
4136 4137
	if (err)
		goto out2;
Amit Arora's avatar
Amit Arora committed
4138

4139 4140 4141
	/* Check if the extent after searching to the right implies a
	 * cluster we can use. */
	if ((sbi->s_cluster_ratio > 1) && ex2 &&
4142
	    get_implied_cluster_alloc(inode->i_sb, map, ex2, path)) {
4143 4144
		ar.len = allocated = map->m_len;
		newblock = map->m_pblk;
4145
		map->m_flags |= EXT4_MAP_FROM_CLUSTER;
4146 4147 4148
		goto got_allocated_blocks;
	}

4149 4150 4151 4152 4153 4154
	/*
	 * See if request is beyond maximum number of blocks we can have in
	 * a single extent. For an initialized extent this limit is
	 * EXT_INIT_MAX_LEN and for an uninitialized extent this limit is
	 * EXT_UNINIT_MAX_LEN.
	 */
4155
	if (map->m_len > EXT_INIT_MAX_LEN &&
4156
	    !(flags & EXT4_GET_BLOCKS_UNINIT_EXT))
4157 4158
		map->m_len = EXT_INIT_MAX_LEN;
	else if (map->m_len > EXT_UNINIT_MAX_LEN &&
4159
		 (flags & EXT4_GET_BLOCKS_UNINIT_EXT))
4160
		map->m_len = EXT_UNINIT_MAX_LEN;
4161

4162 4163
	/* Check if we can really insert (m_lblk)::(m_lblk + m_len) extent */
	newex.ee_len = cpu_to_le16(map->m_len);
4164
	err = ext4_ext_check_overlap(sbi, inode, &newex, path);
Amit Arora's avatar
Amit Arora committed
4165
	if (err)
4166
		allocated = ext4_ext_get_actual_len(&newex);
Amit Arora's avatar
Amit Arora committed
4167
	else
4168
		allocated = map->m_len;
4169 4170 4171

	/* allocate new block */
	ar.inode = inode;
4172 4173
	ar.goal = ext4_ext_find_goal(inode, path, map->m_lblk);
	ar.logical = map->m_lblk;
4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185
	/*
	 * We calculate the offset from the beginning of the cluster
	 * for the logical block number, since when we allocate a
	 * physical cluster, the physical block should start at the
	 * same offset from the beginning of the cluster.  This is
	 * needed so that future calls to get_implied_cluster_alloc()
	 * work correctly.
	 */
	offset = map->m_lblk & (sbi->s_cluster_ratio - 1);
	ar.len = EXT4_NUM_B2C(sbi, offset+allocated);
	ar.goal -= offset;
	ar.logical -= offset;
4186 4187 4188 4189 4190
	if (S_ISREG(inode->i_mode))
		ar.flags = EXT4_MB_HINT_DATA;
	else
		/* disable in-core preallocation for non-regular files */
		ar.flags = 0;
4191 4192
	if (flags & EXT4_GET_BLOCKS_NO_NORMALIZE)
		ar.flags |= EXT4_MB_HINT_NOPREALLOC;
4193
	newblock = ext4_mb_new_blocks(handle, &ar, &err);
4194 4195
	if (!newblock)
		goto out2;
4196
	ext_debug("allocate new block: goal %llu, found %llu/%u\n",
4197
		  ar.goal, newblock, allocated);
4198
	free_on_err = 1;
4199
	allocated_clusters = ar.len;
4200 4201 4202
	ar.len = EXT4_C2B(sbi, ar.len) - offset;
	if (ar.len > allocated)
		ar.len = allocated;
4203

4204
got_allocated_blocks:
4205
	/* try to insert new extent into found leaf and return */
4206
	ext4_ext_store_pblock(&newex, newblock + offset);
4207
	newex.ee_len = cpu_to_le16(ar.len);
4208 4209
	/* Mark uninitialized */
	if (flags & EXT4_GET_BLOCKS_UNINIT_EXT){
Amit Arora's avatar
Amit Arora committed
4210
		ext4_ext_mark_uninitialized(&newex);
4211
		map->m_flags |= EXT4_MAP_UNWRITTEN;
4212
		/*
4213
		 * io_end structure was created for every IO write to an
Lucas De Marchi's avatar
Lucas De Marchi committed
4214
		 * uninitialized extent. To avoid unnecessary conversion,
4215
		 * here we flag the IO that really needs the conversion.
4216
		 * For non asycn direct IO case, flag the inode state
Lucas De Marchi's avatar
Lucas De Marchi committed
4217
		 * that we need to perform conversion when IO is done.
4218
		 */
4219 4220
		if ((flags & EXT4_GET_BLOCKS_PRE_IO))
			set_unwritten = 1;
4221
		if (ext4_should_dioread_nolock(inode))
4222
			map->m_flags |= EXT4_MAP_UNINIT;
4223
	}
4224

4225 4226 4227 4228
	err = 0;
	if ((flags & EXT4_GET_BLOCKS_KEEP_SIZE) == 0)
		err = check_eofblocks_fl(handle, inode, map->m_lblk,
					 path, ar.len);
4229 4230 4231
	if (!err)
		err = ext4_ext_insert_extent(handle, inode, path,
					     &newex, flags);
4232 4233 4234 4235 4236 4237 4238 4239 4240

	if (!err && set_unwritten) {
		if (io)
			ext4_set_io_unwritten_flag(inode, io);
		else
			ext4_set_inode_state(inode,
					     EXT4_STATE_DIO_UNWRITTEN);
	}

4241
	if (err && free_on_err) {
4242 4243
		int fb_flags = flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE ?
			EXT4_FREE_BLOCKS_NO_QUOT_UPDATE : 0;
4244
		/* free data blocks we just allocated */
4245 4246
		/* not a good idea to call discard here directly,
		 * but otherwise we'd need to call it every free() */
4247
		ext4_discard_preallocations(inode);
4248
		ext4_free_blocks(handle, inode, NULL, ext4_ext_pblock(&newex),
4249
				 ext4_ext_get_actual_len(&newex), fb_flags);
4250
		goto out2;
4251
	}
4252 4253

	/* previous routine could use block we allocated */
4254
	newblock = ext4_ext_pblock(&newex);
4255
	allocated = ext4_ext_get_actual_len(&newex);
4256 4257 4258
	if (allocated > map->m_len)
		allocated = map->m_len;
	map->m_flags |= EXT4_MAP_NEW;
4259

4260 4261 4262 4263
	/*
	 * Update reserved blocks/metadata blocks after successful
	 * block allocation which had been deferred till now.
	 */
4264
	if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE) {
4265
		unsigned int reserved_clusters;
4266
		/*
4267
		 * Check how many clusters we had reserved this allocated range
4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285
		 */
		reserved_clusters = get_reserved_cluster_alloc(inode,
						map->m_lblk, allocated);
		if (map->m_flags & EXT4_MAP_FROM_CLUSTER) {
			if (reserved_clusters) {
				/*
				 * We have clusters reserved for this range.
				 * But since we are not doing actual allocation
				 * and are simply using blocks from previously
				 * allocated cluster, we should release the
				 * reservation and not claim quota.
				 */
				ext4_da_update_reserve_space(inode,
						reserved_clusters, 0);
			}
		} else {
			BUG_ON(allocated_clusters < reserved_clusters);
			if (reserved_clusters < allocated_clusters) {
4286
				struct ext4_inode_info *ei = EXT4_I(inode);
4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328
				int reservation = allocated_clusters -
						  reserved_clusters;
				/*
				 * It seems we claimed few clusters outside of
				 * the range of this allocation. We should give
				 * it back to the reservation pool. This can
				 * happen in the following case:
				 *
				 * * Suppose s_cluster_ratio is 4 (i.e., each
				 *   cluster has 4 blocks. Thus, the clusters
				 *   are [0-3],[4-7],[8-11]...
				 * * First comes delayed allocation write for
				 *   logical blocks 10 & 11. Since there were no
				 *   previous delayed allocated blocks in the
				 *   range [8-11], we would reserve 1 cluster
				 *   for this write.
				 * * Next comes write for logical blocks 3 to 8.
				 *   In this case, we will reserve 2 clusters
				 *   (for [0-3] and [4-7]; and not for [8-11] as
				 *   that range has a delayed allocated blocks.
				 *   Thus total reserved clusters now becomes 3.
				 * * Now, during the delayed allocation writeout
				 *   time, we will first write blocks [3-8] and
				 *   allocate 3 clusters for writing these
				 *   blocks. Also, we would claim all these
				 *   three clusters above.
				 * * Now when we come here to writeout the
				 *   blocks [10-11], we would expect to claim
				 *   the reservation of 1 cluster we had made
				 *   (and we would claim it since there are no
				 *   more delayed allocated blocks in the range
				 *   [8-11]. But our reserved cluster count had
				 *   already gone to 0.
				 *
				 *   Thus, at the step 4 above when we determine
				 *   that there are still some unwritten delayed
				 *   allocated blocks outside of our current
				 *   block range, we should increment the
				 *   reserved clusters count so that when the
				 *   remaining blocks finally gets written, we
				 *   could claim them.
				 */
4329 4330 4331 4332 4333
				dquot_reserve_block(inode,
						EXT4_C2B(sbi, reservation));
				spin_lock(&ei->i_block_reservation_lock);
				ei->i_reserved_data_blocks += reservation;
				spin_unlock(&ei->i_block_reservation_lock);
4334
			}
4335 4336 4337 4338 4339 4340 4341 4342 4343
			/*
			 * We will claim quota for all newly allocated blocks.
			 * We're updating the reserved space *after* the
			 * correction above so we do not accidentally free
			 * all the metadata reservation because we might
			 * actually need it later on.
			 */
			ext4_da_update_reserve_space(inode, allocated_clusters,
							1);
4344 4345
		}
	}
4346

4347 4348 4349 4350
	/*
	 * Cache the extent and update transaction to commit on fdatasync only
	 * when it is _not_ an uninitialized extent.
	 */
Zheng Liu's avatar
Zheng Liu committed
4351
	if ((flags & EXT4_GET_BLOCKS_UNINIT_EXT) == 0)
4352
		ext4_update_inode_fsync_trans(handle, inode, 1);
Zheng Liu's avatar
Zheng Liu committed
4353
	else
4354
		ext4_update_inode_fsync_trans(handle, inode, 0);
4355
out:
4356 4357
	if (allocated > map->m_len)
		allocated = map->m_len;
4358
	ext4_ext_show_leaf(inode, path);
4359 4360 4361
	map->m_flags |= EXT4_MAP_MAPPED;
	map->m_pblk = newblock;
	map->m_len = allocated;
4362 4363 4364 4365 4366
out2:
	if (path) {
		ext4_ext_drop_refs(path);
		kfree(path);
	}
4367

4368
out3:
4369
	trace_ext4_ext_map_blocks_exit(inode, map, err ? err : allocated);
4370

4371
	return err ? err : allocated;
4372 4373
}

Theodore Ts'o's avatar
Theodore Ts'o committed
4374
void ext4_ext_truncate(handle_t *handle, struct inode *inode)
4375 4376
{
	struct super_block *sb = inode->i_sb;
4377
	ext4_lblk_t last_block;
4378 4379 4380
	int err = 0;

	/*
4381 4382 4383
	 * TODO: optimization is possible here.
	 * Probably we need not scan at all,
	 * because page truncation is enough.
4384 4385 4386 4387 4388 4389 4390 4391
	 */

	/* we have to know where to truncate from in crash case */
	EXT4_I(inode)->i_disksize = inode->i_size;
	ext4_mark_inode_dirty(handle, inode);

	last_block = (inode->i_size + sb->s_blocksize - 1)
			>> EXT4_BLOCK_SIZE_BITS(sb);
4392 4393
	err = ext4_es_remove_extent(inode, last_block,
				    EXT_MAX_BLOCKS - last_block);
4394
	err = ext4_ext_remove_space(inode, last_block, EXT_MAX_BLOCKS - 1);
4395 4396
}

4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410
static void ext4_falloc_update_inode(struct inode *inode,
				int mode, loff_t new_size, int update_ctime)
{
	struct timespec now;

	if (update_ctime) {
		now = current_fs_time(inode->i_sb);
		if (!timespec_equal(&inode->i_ctime, &now))
			inode->i_ctime = now;
	}
	/*
	 * Update only when preallocation was requested beyond
	 * the file size.
	 */
4411 4412 4413 4414 4415
	if (!(mode & FALLOC_FL_KEEP_SIZE)) {
		if (new_size > i_size_read(inode))
			i_size_write(inode, new_size);
		if (new_size > EXT4_I(inode)->i_disksize)
			ext4_update_i_disksize(inode, new_size);
4416 4417 4418 4419 4420 4421
	} else {
		/*
		 * Mark that we allocate beyond EOF so the subsequent truncate
		 * can proceed even if the new size is the same as i_size.
		 */
		if (new_size > i_size_read(inode))
4422
			ext4_set_inode_flag(inode, EXT4_INODE_EOFBLOCKS);
4423 4424 4425 4426
	}

}

Amit Arora's avatar
Amit Arora committed
4427
/*
4428
 * preallocate space for a file. This implements ext4's fallocate file
Amit Arora's avatar
Amit Arora committed
4429 4430 4431 4432 4433
 * operation, which gets called from sys_fallocate system call.
 * For block-mapped files, posix_fallocate should fall back to the method
 * of writing zeroes to the required new blocks (the same behavior which is
 * expected for file systems which do not support fallocate() system call).
 */
4434
long ext4_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
Amit Arora's avatar
Amit Arora committed
4435
{
Al Viro's avatar
Al Viro committed
4436
	struct inode *inode = file_inode(file);
Amit Arora's avatar
Amit Arora committed
4437
	handle_t *handle;
4438
	loff_t new_size;
4439
	unsigned int max_blocks;
Amit Arora's avatar
Amit Arora committed
4440 4441 4442
	int ret = 0;
	int ret2 = 0;
	int retries = 0;
4443
	int flags;
4444
	struct ext4_map_blocks map;
Amit Arora's avatar
Amit Arora committed
4445 4446
	unsigned int credits, blkbits = inode->i_blkbits;

4447 4448 4449 4450 4451 4452 4453
	/* Return error if mode is not supported */
	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
		return -EOPNOTSUPP;

	if (mode & FALLOC_FL_PUNCH_HOLE)
		return ext4_punch_hole(file, offset, len);

4454 4455 4456 4457
	ret = ext4_convert_inline_data(inode);
	if (ret)
		return ret;

4458 4459 4460 4461 4462 4463 4464
	/*
	 * currently supporting (pre)allocate mode for extent-based
	 * files _only_
	 */
	if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
		return -EOPNOTSUPP;

4465
	trace_ext4_fallocate_enter(inode, offset, len, mode);
4466
	map.m_lblk = offset >> blkbits;
4467 4468 4469 4470
	/*
	 * We can't just convert len to max_blocks because
	 * If blocksize = 4096 offset = 3072 and len = 2048
	 */
Amit Arora's avatar
Amit Arora committed
4471
	max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits)
4472
		- map.m_lblk;
Amit Arora's avatar
Amit Arora committed
4473
	/*
4474
	 * credits to insert 1 extent into extent tree
Amit Arora's avatar
Amit Arora committed
4475
	 */
4476
	credits = ext4_chunk_trans_blocks(inode, max_blocks);
4477
	mutex_lock(&inode->i_mutex);
4478 4479 4480
	ret = inode_newsize_ok(inode, (len + offset));
	if (ret) {
		mutex_unlock(&inode->i_mutex);
4481
		trace_ext4_fallocate_exit(inode, offset, max_blocks, ret);
4482 4483
		return ret;
	}
4484
	flags = EXT4_GET_BLOCKS_CREATE_UNINIT_EXT;
4485 4486
	if (mode & FALLOC_FL_KEEP_SIZE)
		flags |= EXT4_GET_BLOCKS_KEEP_SIZE;
4487 4488 4489 4490 4491 4492 4493
	/*
	 * Don't normalize the request if it can fit in one extent so
	 * that it doesn't get unnecessarily split into multiple
	 * extents.
	 */
	if (len <= EXT_UNINIT_MAX_LEN << blkbits)
		flags |= EXT4_GET_BLOCKS_NO_NORMALIZE;
4494

Amit Arora's avatar
Amit Arora committed
4495 4496
retry:
	while (ret >= 0 && ret < max_blocks) {
4497 4498
		map.m_lblk = map.m_lblk + ret;
		map.m_len = max_blocks = max_blocks - ret;
4499 4500
		handle = ext4_journal_start(inode, EXT4_HT_MAP_BLOCKS,
					    credits);
Amit Arora's avatar
Amit Arora committed
4501 4502 4503 4504
		if (IS_ERR(handle)) {
			ret = PTR_ERR(handle);
			break;
		}
4505
		ret = ext4_map_blocks(handle, inode, &map, flags);
4506
		if (ret <= 0) {
4507
#ifdef EXT4FS_DEBUG
4508 4509 4510 4511 4512
			ext4_warning(inode->i_sb,
				     "inode #%lu: block %u: len %u: "
				     "ext4_ext_map_blocks returned %d",
				     inode->i_ino, map.m_lblk,
				     map.m_len, ret);
4513
#endif
Amit Arora's avatar
Amit Arora committed
4514 4515 4516 4517
			ext4_mark_inode_dirty(handle, inode);
			ret2 = ext4_journal_stop(handle);
			break;
		}
4518
		if ((map.m_lblk + ret) >= (EXT4_BLOCK_ALIGN(offset + len,
4519 4520 4521
						blkbits) >> blkbits))
			new_size = offset + len;
		else
4522
			new_size = ((loff_t) map.m_lblk + ret) << blkbits;
Amit Arora's avatar
Amit Arora committed
4523

4524
		ext4_falloc_update_inode(inode, mode, new_size,
4525
					 (map.m_flags & EXT4_MAP_NEW));
Amit Arora's avatar
Amit Arora committed
4526
		ext4_mark_inode_dirty(handle, inode);
4527 4528
		if ((file->f_flags & O_SYNC) && ret >= max_blocks)
			ext4_handle_sync(handle);
Amit Arora's avatar
Amit Arora committed
4529 4530 4531 4532
		ret2 = ext4_journal_stop(handle);
		if (ret2)
			break;
	}
4533 4534 4535
	if (ret == -ENOSPC &&
			ext4_should_retry_alloc(inode->i_sb, &retries)) {
		ret = 0;
Amit Arora's avatar
Amit Arora committed
4536 4537
		goto retry;
	}
4538
	mutex_unlock(&inode->i_mutex);
4539 4540
	trace_ext4_fallocate_exit(inode, offset, max_blocks,
				ret > 0 ? ret2 : ret);
Amit Arora's avatar
Amit Arora committed
4541 4542
	return ret > 0 ? ret2 : ret;
}
4543

4544 4545 4546 4547 4548 4549 4550 4551
/*
 * This function convert a range of blocks to written extents
 * The caller of this function will pass the start offset and the size.
 * all unwritten extents within this range will be converted to
 * written extents.
 *
 * This function is called from the direct IO end io call back
 * function, to convert the fallocated extents after IO is completed.
4552
 * Returns 0 on success.
4553 4554
 */
int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
4555
				    ssize_t len)
4556 4557 4558 4559 4560
{
	handle_t *handle;
	unsigned int max_blocks;
	int ret = 0;
	int ret2 = 0;
4561
	struct ext4_map_blocks map;
4562 4563
	unsigned int credits, blkbits = inode->i_blkbits;

4564
	map.m_lblk = offset >> blkbits;
4565 4566 4567 4568
	/*
	 * We can't just convert len to max_blocks because
	 * If blocksize = 4096 offset = 3072 and len = 2048
	 */
4569 4570
	max_blocks = ((EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits) -
		      map.m_lblk);
4571 4572 4573 4574 4575
	/*
	 * credits to insert 1 extent into extent tree
	 */
	credits = ext4_chunk_trans_blocks(inode, max_blocks);
	while (ret >= 0 && ret < max_blocks) {
4576 4577
		map.m_lblk += ret;
		map.m_len = (max_blocks -= ret);
4578
		handle = ext4_journal_start(inode, EXT4_HT_MAP_BLOCKS, credits);
4579 4580 4581 4582
		if (IS_ERR(handle)) {
			ret = PTR_ERR(handle);
			break;
		}
4583
		ret = ext4_map_blocks(handle, inode, &map,
4584
				      EXT4_GET_BLOCKS_IO_CONVERT_EXT);
4585 4586 4587 4588 4589 4590
		if (ret <= 0)
			ext4_warning(inode->i_sb,
				     "inode #%lu: block %u: len %u: "
				     "ext4_ext_map_blocks returned %d",
				     inode->i_ino, map.m_lblk,
				     map.m_len, ret);
4591 4592 4593 4594 4595 4596 4597
		ext4_mark_inode_dirty(handle, inode);
		ret2 = ext4_journal_stop(handle);
		if (ret <= 0 || ret2 )
			break;
	}
	return ret > 0 ? ret2 : ret;
}
4598

4599
/*
Zheng Liu's avatar
Zheng Liu committed
4600 4601
 * If newes is not existing extent (newes->ec_pblk equals zero) find
 * delayed extent at start of newes and update newes accordingly and
4602 4603
 * return start of the next delayed extent.
 *
Zheng Liu's avatar
Zheng Liu committed
4604
 * If newes is existing extent (newes->ec_pblk is not equal zero)
4605
 * return start of next delayed extent or EXT_MAX_BLOCKS if no delayed
Zheng Liu's avatar
Zheng Liu committed
4606
 * extent found. Leave newes unmodified.
4607
 */
4608
static int ext4_find_delayed_extent(struct inode *inode,
Zheng Liu's avatar
Zheng Liu committed
4609
				    struct extent_status *newes)
4610
{
4611
	struct extent_status es;
4612
	ext4_lblk_t block, next_del;
4613

Zheng Liu's avatar
Zheng Liu committed
4614
	ext4_es_find_delayed_extent(inode, newes->es_lblk, &es);
4615

Zheng Liu's avatar
Zheng Liu committed
4616
	if (newes->es_pblk == 0) {
4617
		/*
Zheng Liu's avatar
Zheng Liu committed
4618
		 * No extent in extent-tree contains block @newes->es_pblk,
4619 4620
		 * then the block may stay in 1)a hole or 2)delayed-extent.
		 */
Zheng Liu's avatar
Zheng Liu committed
4621
		if (es.es_len == 0)
4622
			/* A hole found. */
4623
			return 0;
4624

Zheng Liu's avatar
Zheng Liu committed
4625
		if (es.es_lblk > newes->es_lblk) {
4626
			/* A hole found. */
Zheng Liu's avatar
Zheng Liu committed
4627 4628
			newes->es_len = min(es.es_lblk - newes->es_lblk,
					    newes->es_len);
4629
			return 0;
4630
		}
4631

Zheng Liu's avatar
Zheng Liu committed
4632
		newes->es_len = es.es_lblk + es.es_len - newes->es_lblk;
4633 4634
	}

Zheng Liu's avatar
Zheng Liu committed
4635
	block = newes->es_lblk + newes->es_len;
4636 4637 4638 4639 4640 4641
	ext4_es_find_delayed_extent(inode, block, &es);
	if (es.es_len == 0)
		next_del = EXT_MAX_BLOCKS;
	else
		next_del = es.es_lblk;

4642
	return next_del;
4643 4644 4645 4646
}
/* fiemap flags we can handle specified here */
#define EXT4_FIEMAP_FLAGS	(FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR)

Aneesh Kumar K.V's avatar
Aneesh Kumar K.V committed
4647 4648
static int ext4_xattr_fiemap(struct inode *inode,
				struct fiemap_extent_info *fieinfo)
4649 4650 4651 4652 4653 4654 4655 4656
{
	__u64 physical = 0;
	__u64 length;
	__u32 flags = FIEMAP_EXTENT_LAST;
	int blockbits = inode->i_sb->s_blocksize_bits;
	int error = 0;

	/* in-inode? */
4657
	if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669
		struct ext4_iloc iloc;
		int offset;	/* offset of xattr in inode */

		error = ext4_get_inode_loc(inode, &iloc);
		if (error)
			return error;
		physical = iloc.bh->b_blocknr << blockbits;
		offset = EXT4_GOOD_OLD_INODE_SIZE +
				EXT4_I(inode)->i_extra_isize;
		physical += offset;
		length = EXT4_SB(inode->i_sb)->s_inode_size - offset;
		flags |= FIEMAP_EXTENT_DATA_INLINE;
4670
		brelse(iloc.bh);
4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687
	} else { /* external block */
		physical = EXT4_I(inode)->i_file_acl << blockbits;
		length = inode->i_sb->s_blocksize;
	}

	if (physical)
		error = fiemap_fill_next_extent(fieinfo, 0, physical,
						length, flags);
	return (error < 0 ? error : 0);
}

int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
		__u64 start, __u64 len)
{
	ext4_lblk_t start_blk;
	int error = 0;

4688 4689 4690 4691 4692 4693 4694 4695 4696
	if (ext4_has_inline_data(inode)) {
		int has_inline = 1;

		error = ext4_inline_data_fiemap(inode, fieinfo, &has_inline);

		if (has_inline)
			return error;
	}

4697
	/* fallback to generic here if not in extents fmt */
4698
	if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
4699 4700 4701 4702 4703 4704 4705 4706 4707
		return generic_block_fiemap(inode, fieinfo, start, len,
			ext4_get_block);

	if (fiemap_check_flags(fieinfo, EXT4_FIEMAP_FLAGS))
		return -EBADR;

	if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
		error = ext4_xattr_fiemap(inode, fieinfo);
	} else {
4708 4709 4710
		ext4_lblk_t len_blks;
		__u64 last_blk;

4711
		start_blk = start >> inode->i_sb->s_blocksize_bits;
4712
		last_blk = (start + len - 1) >> inode->i_sb->s_blocksize_bits;
4713 4714
		if (last_blk >= EXT_MAX_BLOCKS)
			last_blk = EXT_MAX_BLOCKS-1;
4715
		len_blks = ((ext4_lblk_t) last_blk) - start_blk + 1;
4716 4717

		/*
4718 4719
		 * Walk the extent tree gathering extent information
		 * and pushing extents back to the user.
4720
		 */
4721 4722
		error = ext4_fill_fiemap_extents(inode, start_blk,
						 len_blks, fieinfo);
4723 4724 4725 4726
	}

	return error;
}