extents.c 133.3 KB
Newer Older
A
Alex Tomas 已提交
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>
A
Alex Tomas 已提交
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>
A
Amit Arora 已提交
40
#include <linux/falloc.h>
A
Alex Tomas 已提交
41
#include <asm/uaccess.h>
42
#include <linux/fiemap.h>
43
#include "ext4_jbd2.h"
A
Alex Tomas 已提交
44

45 46
#include <trace/events/ext4.h>

47 48 49 50 51 52 53 54
/*
 * 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 */

55 56 57 58 59 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
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);
}

A
Allison Henderson 已提交
95 96 97 98 99 100 101
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);

102 103 104 105 106 107 108
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);

109 110 111
static int ext4_ext_truncate_extend_restart(handle_t *handle,
					    struct inode *inode,
					    int needed)
A
Alex Tomas 已提交
112 113 114
{
	int err;

115 116
	if (!ext4_handle_valid(handle))
		return 0;
A
Alex Tomas 已提交
117
	if (handle->h_buffer_credits > needed)
118 119
		return 0;
	err = ext4_journal_extend(handle, needed);
120
	if (err <= 0)
121
		return err;
122
	err = ext4_truncate_restart_trans(handle, inode, needed);
123 124
	if (err == 0)
		err = -EAGAIN;
125 126

	return err;
A
Alex Tomas 已提交
127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151
}

/*
 * 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
 */
152 153 154 155 156
#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)
A
Alex Tomas 已提交
157 158 159
{
	int err;
	if (path->p_bh) {
160
		ext4_extent_block_csum_set(inode, ext_block_hdr(path->p_bh));
A
Alex Tomas 已提交
161
		/* path points to block */
162 163
		err = __ext4_handle_dirty_metadata(where, line, handle,
						   inode, path->p_bh);
A
Alex Tomas 已提交
164 165 166 167 168 169 170
	} else {
		/* path points to leaf/index in inode body */
		err = ext4_mark_inode_dirty(handle, inode);
	}
	return err;
}

171
static ext4_fsblk_t ext4_ext_find_goal(struct inode *inode,
A
Alex Tomas 已提交
172
			      struct ext4_ext_path *path,
A
Aneesh Kumar K.V 已提交
173
			      ext4_lblk_t block)
A
Alex Tomas 已提交
174 175
{
	if (path) {
176
		int depth = path->p_depth;
A
Alex Tomas 已提交
177 178
		struct ext4_extent *ex;

179 180 181 182 183 184 185 186 187 188 189 190 191 192
		/*
		 * 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,
T
Tao Ma 已提交
193
		 * especially if the latter case turns out to be
194 195
		 * common.
		 */
196
		ex = path[depth].p_ext;
197 198 199 200 201 202 203 204 205
		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);
		}
A
Alex Tomas 已提交
206

207 208
		/* it looks like index is empty;
		 * try to find starting block from index itself */
A
Alex Tomas 已提交
209 210 211 212 213
		if (path[depth].p_bh)
			return path[depth].p_bh->b_blocknr;
	}

	/* OK. use inode's group */
214
	return ext4_inode_to_goal_block(inode);
A
Alex Tomas 已提交
215 216
}

A
Aneesh Kumar K.V 已提交
217 218 219
/*
 * Allocation for a meta data block
 */
220
static ext4_fsblk_t
A
Aneesh Kumar K.V 已提交
221
ext4_ext_new_meta_block(handle_t *handle, struct inode *inode,
A
Alex Tomas 已提交
222
			struct ext4_ext_path *path,
223
			struct ext4_extent *ex, int *err, unsigned int flags)
A
Alex Tomas 已提交
224
{
225
	ext4_fsblk_t goal, newblock;
A
Alex Tomas 已提交
226 227

	goal = ext4_ext_find_goal(inode, path, le32_to_cpu(ex->ee_block));
228 229
	newblock = ext4_new_meta_blocks(handle, inode, goal, flags,
					NULL, err);
A
Alex Tomas 已提交
230 231 232
	return newblock;
}

233
static inline int ext4_ext_space_block(struct inode *inode, int check)
A
Alex Tomas 已提交
234 235 236 237 238
{
	int size;

	size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
			/ sizeof(struct ext4_extent);
239
#ifdef AGGRESSIVE_TEST
240 241
	if (!check && size > 6)
		size = 6;
A
Alex Tomas 已提交
242 243 244 245
#endif
	return size;
}

246
static inline int ext4_ext_space_block_idx(struct inode *inode, int check)
A
Alex Tomas 已提交
247 248 249 250 251
{
	int size;

	size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
			/ sizeof(struct ext4_extent_idx);
252
#ifdef AGGRESSIVE_TEST
253 254
	if (!check && size > 5)
		size = 5;
A
Alex Tomas 已提交
255 256 257 258
#endif
	return size;
}

259
static inline int ext4_ext_space_root(struct inode *inode, int check)
A
Alex Tomas 已提交
260 261 262 263 264 265
{
	int size;

	size = sizeof(EXT4_I(inode)->i_data);
	size -= sizeof(struct ext4_extent_header);
	size /= sizeof(struct ext4_extent);
266
#ifdef AGGRESSIVE_TEST
267 268
	if (!check && size > 3)
		size = 3;
A
Alex Tomas 已提交
269 270 271 272
#endif
	return size;
}

273
static inline int ext4_ext_space_root_idx(struct inode *inode, int check)
A
Alex Tomas 已提交
274 275 276 277 278 279
{
	int size;

	size = sizeof(EXT4_I(inode)->i_data);
	size -= sizeof(struct ext4_extent_header);
	size /= sizeof(struct ext4_extent_idx);
280
#ifdef AGGRESSIVE_TEST
281 282
	if (!check && size > 4)
		size = 4;
A
Alex Tomas 已提交
283 284 285 286
#endif
	return size;
}

287 288 289 290 291
/*
 * Calculate the number of metadata blocks needed
 * to allocate @blocks
 * Worse case is one block per extent
 */
292
int ext4_ext_calc_metadata_amount(struct inode *inode, ext4_lblk_t lblock)
293
{
294
	struct ext4_inode_info *ei = EXT4_I(inode);
295
	int idxs;
296

297 298
	idxs = ((inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
		/ sizeof(struct ext4_extent_idx));
299 300

	/*
301 302 303 304 305 306
	 * 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.
307
	 */
308 309
	if (ei->i_da_metadata_calc_len &&
	    ei->i_da_metadata_calc_last_lblock+1 == lblock) {
310 311
		int num = 0;

312 313 314 315 316 317 318 319 320 321 322 323
		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;
	}
324

325 326 327 328 329 330 331
	/*
	 * 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;
332 333
}

334 335 336 337 338 339 340
static int
ext4_ext_max_entries(struct inode *inode, int depth)
{
	int max;

	if (depth == ext_depth(inode)) {
		if (depth == 0)
341
			max = ext4_ext_space_root(inode, 1);
342
		else
343
			max = ext4_ext_space_root_idx(inode, 1);
344 345
	} else {
		if (depth == 0)
346
			max = ext4_ext_space_block(inode, 1);
347
		else
348
			max = ext4_ext_space_block_idx(inode, 1);
349 350 351 352 353
	}

	return max;
}

354 355
static int ext4_valid_extent(struct inode *inode, struct ext4_extent *ext)
{
356
	ext4_fsblk_t block = ext4_ext_pblock(ext);
357
	int len = ext4_ext_get_actual_len(ext);
358

359 360
	if (len == 0)
		return 0;
361
	return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, len);
362 363 364 365 366
}

static int ext4_valid_extent_idx(struct inode *inode,
				struct ext4_extent_idx *ext_idx)
{
367
	ext4_fsblk_t block = ext4_idx_pblock(ext_idx);
368

369
	return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, 1);
370 371 372 373 374 375 376 377 378 379 380 381 382 383
}

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 */
384
		struct ext4_extent *ext = EXT_FIRST_EXTENT(eh);
385 386 387 388 389 390 391
		while (entries) {
			if (!ext4_valid_extent(inode, ext))
				return 0;
			ext++;
			entries--;
		}
	} else {
392
		struct ext4_extent_idx *ext_idx = EXT_FIRST_INDEX(eh);
393 394 395 396 397 398 399 400 401 402
		while (entries) {
			if (!ext4_valid_extent_idx(inode, ext_idx))
				return 0;
			ext_idx++;
			entries--;
		}
	}
	return 1;
}

403 404 405
static int __ext4_ext_check(const char *function, unsigned int line,
			    struct inode *inode, struct ext4_extent_header *eh,
			    int depth)
406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430
{
	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;
	}
431 432 433 434
	if (!ext4_valid_extent_entries(inode, eh, depth)) {
		error_msg = "invalid extent entries";
		goto corrupted;
	}
435 436 437 438 439 440
	/* 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;
	}
441 442 443
	return 0;

corrupted:
444
	ext4_error_inode(inode, function, line, 0,
445
			"bad header/extent: %s - magic %x, "
446
			"entries %u, max %u(%u), depth %u(%u)",
447
			error_msg, le16_to_cpu(eh->eh_magic),
448 449 450 451 452 453
			le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max),
			max, le16_to_cpu(eh->eh_depth), depth);

	return -EIO;
}

454
#define ext4_ext_check(inode, eh, depth)	\
455
	__ext4_ext_check(__func__, __LINE__, inode, eh, depth)
456

457 458 459 460 461
int ext4_ext_check_inode(struct inode *inode)
{
	return ext4_ext_check(inode, ext_inode_hdr(inode), ext_depth(inode));
}

462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481
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)

A
Alex Tomas 已提交
482 483 484 485 486 487 488 489
#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) {
490
		  ext_debug("  %d->%llu", le32_to_cpu(path->p_idx->ei_block),
491
			    ext4_idx_pblock(path->p_idx));
A
Alex Tomas 已提交
492
		} else if (path->p_ext) {
493
			ext_debug("  %d:[%d]%d:%llu ",
A
Alex Tomas 已提交
494
				  le32_to_cpu(path->p_ext->ee_block),
495
				  ext4_ext_is_uninitialized(path->p_ext),
A
Amit Arora 已提交
496
				  ext4_ext_get_actual_len(path->p_ext),
497
				  ext4_ext_pblock(path->p_ext));
A
Alex Tomas 已提交
498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516
		} 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);

517 518
	ext_debug("Displaying leaf extents for inode %lu\n", inode->i_ino);

A
Alex Tomas 已提交
519
	for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ex++) {
520 521
		ext_debug("%d:[%d]%d:%llu ", le32_to_cpu(ex->ee_block),
			  ext4_ext_is_uninitialized(ex),
522
			  ext4_ext_get_actual_len(ex), ext4_ext_pblock(ex));
A
Alex Tomas 已提交
523 524 525
	}
	ext_debug("\n");
}
526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558

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++;
	}
}

A
Alex Tomas 已提交
559
#else
560 561
#define ext4_ext_show_path(inode, path)
#define ext4_ext_show_leaf(inode, path)
562
#define ext4_ext_show_move(inode, path, newblock, level)
A
Alex Tomas 已提交
563 564
#endif

565
void ext4_ext_drop_refs(struct ext4_ext_path *path)
A
Alex Tomas 已提交
566 567 568 569 570 571 572 573 574 575 576 577
{
	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;
		}
}

/*
578 579
 * ext4_ext_binsearch_idx:
 * binary search for the closest index of the given block
580
 * the header must be checked before calling this
A
Alex Tomas 已提交
581 582
 */
static void
A
Aneesh Kumar K.V 已提交
583 584
ext4_ext_binsearch_idx(struct inode *inode,
			struct ext4_ext_path *path, ext4_lblk_t block)
A
Alex Tomas 已提交
585 586 587 588 589
{
	struct ext4_extent_header *eh = path->p_hdr;
	struct ext4_extent_idx *r, *l, *m;


590
	ext_debug("binsearch for %u(idx):  ", block);
A
Alex Tomas 已提交
591 592

	l = EXT_FIRST_INDEX(eh) + 1;
D
Dmitry Monakhov 已提交
593
	r = EXT_LAST_INDEX(eh);
A
Alex Tomas 已提交
594 595 596 597 598 599
	while (l <= r) {
		m = l + (r - l) / 2;
		if (block < le32_to_cpu(m->ei_block))
			r = m - 1;
		else
			l = m + 1;
600 601 602
		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));
A
Alex Tomas 已提交
603 604 605
	}

	path->p_idx = l - 1;
606
	ext_debug("  -> %u->%lld ", le32_to_cpu(path->p_idx->ei_block),
607
		  ext4_idx_pblock(path->p_idx));
A
Alex Tomas 已提交
608 609 610 611 612 613 614 615 616 617

#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)) {
618 619 620 621
				printk(KERN_DEBUG "k=%d, ix=0x%p, "
				       "first=0x%p\n", k,
				       ix, EXT_FIRST_INDEX(eh));
				printk(KERN_DEBUG "%u <= %u\n",
A
Alex Tomas 已提交
622 623 624 625
				       le32_to_cpu(ix->ei_block),
				       le32_to_cpu(ix[-1].ei_block));
			}
			BUG_ON(k && le32_to_cpu(ix->ei_block)
D
Dave Kleikamp 已提交
626
					   <= le32_to_cpu(ix[-1].ei_block));
A
Alex Tomas 已提交
627 628 629 630 631 632 633 634 635 636 637
			if (block < le32_to_cpu(ix->ei_block))
				break;
			chix = ix;
		}
		BUG_ON(chix != path->p_idx);
	}
#endif

}

/*
638 639
 * ext4_ext_binsearch:
 * binary search for closest extent of the given block
640
 * the header must be checked before calling this
A
Alex Tomas 已提交
641 642
 */
static void
A
Aneesh Kumar K.V 已提交
643 644
ext4_ext_binsearch(struct inode *inode,
		struct ext4_ext_path *path, ext4_lblk_t block)
A
Alex Tomas 已提交
645 646 647 648 649 650
{
	struct ext4_extent_header *eh = path->p_hdr;
	struct ext4_extent *r, *l, *m;

	if (eh->eh_entries == 0) {
		/*
651 652
		 * this leaf is empty:
		 * we get such a leaf in split/add case
A
Alex Tomas 已提交
653 654 655 656
		 */
		return;
	}

657
	ext_debug("binsearch for %u:  ", block);
A
Alex Tomas 已提交
658 659

	l = EXT_FIRST_EXTENT(eh) + 1;
D
Dmitry Monakhov 已提交
660
	r = EXT_LAST_EXTENT(eh);
A
Alex Tomas 已提交
661 662 663 664 665 666 667

	while (l <= r) {
		m = l + (r - l) / 2;
		if (block < le32_to_cpu(m->ee_block))
			r = m - 1;
		else
			l = m + 1;
668 669 670
		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));
A
Alex Tomas 已提交
671 672 673
	}

	path->p_ext = l - 1;
674
	ext_debug("  -> %d:%llu:[%d]%d ",
D
Dave Kleikamp 已提交
675
			le32_to_cpu(path->p_ext->ee_block),
676
			ext4_ext_pblock(path->p_ext),
677
			ext4_ext_is_uninitialized(path->p_ext),
A
Amit Arora 已提交
678
			ext4_ext_get_actual_len(path->p_ext));
A
Alex Tomas 已提交
679 680 681 682 683 684 685 686 687

#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)
D
Dave Kleikamp 已提交
688
					  <= le32_to_cpu(ex[-1].ee_block));
A
Alex Tomas 已提交
689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706
			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;
707
	eh->eh_max = cpu_to_le16(ext4_ext_space_root(inode, 0));
A
Alex Tomas 已提交
708 709 710 711 712 713
	ext4_mark_inode_dirty(handle, inode);
	ext4_ext_invalidate_cache(inode);
	return 0;
}

struct ext4_ext_path *
A
Aneesh Kumar K.V 已提交
714 715
ext4_ext_find_extent(struct inode *inode, ext4_lblk_t block,
					struct ext4_ext_path *path)
A
Alex Tomas 已提交
716 717 718 719 720 721
{
	struct ext4_extent_header *eh;
	struct buffer_head *bh;
	short int depth, i, ppos = 0, alloc = 0;

	eh = ext_inode_hdr(inode);
722
	depth = ext_depth(inode);
A
Alex Tomas 已提交
723 724 725

	/* account possible depth increase */
	if (!path) {
A
Avantika Mathur 已提交
726
		path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 2),
A
Alex Tomas 已提交
727 728 729 730 731 732
				GFP_NOFS);
		if (!path)
			return ERR_PTR(-ENOMEM);
		alloc = 1;
	}
	path[0].p_hdr = eh;
733
	path[0].p_bh = NULL;
A
Alex Tomas 已提交
734

735
	i = depth;
A
Alex Tomas 已提交
736 737 738 739
	/* 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));
740

A
Alex Tomas 已提交
741
		ext4_ext_binsearch_idx(inode, path + ppos, block);
742
		path[ppos].p_block = ext4_idx_pblock(path[ppos].p_idx);
A
Alex Tomas 已提交
743 744 745
		path[ppos].p_depth = i;
		path[ppos].p_ext = NULL;

746 747
		bh = sb_getblk(inode->i_sb, path[ppos].p_block);
		if (unlikely(!bh))
A
Alex Tomas 已提交
748
			goto err;
749
		if (!bh_uptodate_or_lock(bh)) {
750 751
			trace_ext4_ext_load_extent(inode, block,
						path[ppos].p_block);
752 753 754 755 756
			if (bh_submit_read(bh) < 0) {
				put_bh(bh);
				goto err;
			}
		}
A
Alex Tomas 已提交
757 758
		eh = ext_block_hdr(bh);
		ppos++;
759 760 761 762 763 764
		if (unlikely(ppos > depth)) {
			put_bh(bh);
			EXT4_ERROR_INODE(inode,
					 "ppos %d > depth %d", ppos, depth);
			goto err;
		}
A
Alex Tomas 已提交
765 766 767 768
		path[ppos].p_bh = bh;
		path[ppos].p_hdr = eh;
		i--;

769
		if (ext4_ext_check_block(inode, eh, i, bh))
A
Alex Tomas 已提交
770 771 772 773 774 775 776 777 778
			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);
779 780
	/* if not an empty leaf */
	if (path[ppos].p_ext)
781
		path[ppos].p_block = ext4_ext_pblock(path[ppos].p_ext);
A
Alex Tomas 已提交
782 783 784 785 786 787 788 789 790 791 792 793 794

	ext4_ext_show_path(inode, path);

	return path;

err:
	ext4_ext_drop_refs(path);
	if (alloc)
		kfree(path);
	return ERR_PTR(-EIO);
}

/*
795 796 797
 * ext4_ext_insert_index:
 * insert new index [@logical;@ptr] into the block at @curp;
 * check where to insert: before @curp or after @curp
A
Alex Tomas 已提交
798
 */
799 800 801
static int ext4_ext_insert_index(handle_t *handle, struct inode *inode,
				 struct ext4_ext_path *curp,
				 int logical, ext4_fsblk_t ptr)
A
Alex Tomas 已提交
802 803 804 805
{
	struct ext4_extent_idx *ix;
	int len, err;

806 807
	err = ext4_ext_get_access(handle, inode, curp);
	if (err)
A
Alex Tomas 已提交
808 809
		return err;

810 811 812 813 814 815
	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;
	}
816 817 818 819 820 821 822 823 824 825

	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;
	}

A
Alex Tomas 已提交
826 827
	if (logical > le32_to_cpu(curp->p_idx->ei_block)) {
		/* insert after */
828
		ext_debug("insert new index %d after: %llu\n", logical, ptr);
A
Alex Tomas 已提交
829 830 831
		ix = curp->p_idx + 1;
	} else {
		/* insert before */
832
		ext_debug("insert new index %d before: %llu\n", logical, ptr);
A
Alex Tomas 已提交
833 834 835
		ix = curp->p_idx;
	}

836 837 838 839 840 841 842 843 844
	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));
	}

845 846 847 848 849
	if (unlikely(ix > EXT_MAX_INDEX(curp->p_hdr))) {
		EXT4_ERROR_INODE(inode, "ix > EXT_MAX_INDEX!");
		return -EIO;
	}

A
Alex Tomas 已提交
850
	ix->ei_block = cpu_to_le32(logical);
851
	ext4_idx_store_pblock(ix, ptr);
M
Marcin Slusarz 已提交
852
	le16_add_cpu(&curp->p_hdr->eh_entries, 1);
A
Alex Tomas 已提交
853

854 855 856 857
	if (unlikely(ix > EXT_LAST_INDEX(curp->p_hdr))) {
		EXT4_ERROR_INODE(inode, "ix > EXT_LAST_INDEX!");
		return -EIO;
	}
A
Alex Tomas 已提交
858 859 860 861 862 863 864 865

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

	return err;
}

/*
866 867 868 869 870 871 872 873
 * 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
A
Alex Tomas 已提交
874 875
 */
static int ext4_ext_split(handle_t *handle, struct inode *inode,
876 877 878
			  unsigned int flags,
			  struct ext4_ext_path *path,
			  struct ext4_extent *newext, int at)
A
Alex Tomas 已提交
879 880 881 882 883 884
{
	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;
885
	ext4_fsblk_t newblock, oldblock;
A
Alex Tomas 已提交
886
	__le32 border;
887
	ext4_fsblk_t *ablocks = NULL; /* array of allocated blocks */
A
Alex Tomas 已提交
888 889 890
	int err = 0;

	/* make decision: where to split? */
891
	/* FIXME: now decision is simplest: at current extent */
A
Alex Tomas 已提交
892

893
	/* if current leaf will be split, then we should use
A
Alex Tomas 已提交
894
	 * border from split point */
895 896 897 898
	if (unlikely(path[depth].p_ext > EXT_MAX_EXTENT(path[depth].p_hdr))) {
		EXT4_ERROR_INODE(inode, "p_ext > EXT_MAX_EXTENT!");
		return -EIO;
	}
A
Alex Tomas 已提交
899 900
	if (path[depth].p_ext != EXT_MAX_EXTENT(path[depth].p_hdr)) {
		border = path[depth].p_ext[1].ee_block;
901
		ext_debug("leaf will be split."
A
Alex Tomas 已提交
902
				" next leaf starts at %d\n",
D
Dave Kleikamp 已提交
903
				  le32_to_cpu(border));
A
Alex Tomas 已提交
904 905 906 907
	} else {
		border = newext->ee_block;
		ext_debug("leaf will be added."
				" next leaf starts at %d\n",
D
Dave Kleikamp 已提交
908
				le32_to_cpu(border));
A
Alex Tomas 已提交
909 910 911
	}

	/*
912 913
	 * If error occurs, then we break processing
	 * and mark filesystem read-only. index won't
A
Alex Tomas 已提交
914
	 * be inserted and tree will be in consistent
915
	 * state. Next mount will repair buffers too.
A
Alex Tomas 已提交
916 917 918
	 */

	/*
919 920 921
	 * Get array to track all allocated blocks.
	 * We need this to handle errors and free blocks
	 * upon them.
A
Alex Tomas 已提交
922
	 */
A
Avantika Mathur 已提交
923
	ablocks = kzalloc(sizeof(ext4_fsblk_t) * depth, GFP_NOFS);
A
Alex Tomas 已提交
924 925 926 927 928 929
	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++) {
A
Aneesh Kumar K.V 已提交
930
		newblock = ext4_ext_new_meta_block(handle, inode, path,
931
						   newext, &err, flags);
A
Alex Tomas 已提交
932 933 934 935 936 937 938
		if (newblock == 0)
			goto cleanup;
		ablocks[a] = newblock;
	}

	/* initialize new leaf */
	newblock = ablocks[--a];
939 940 941 942 943
	if (unlikely(newblock == 0)) {
		EXT4_ERROR_INODE(inode, "newblock == 0!");
		err = -EIO;
		goto cleanup;
	}
A
Alex Tomas 已提交
944 945 946 947 948 949 950
	bh = sb_getblk(inode->i_sb, newblock);
	if (!bh) {
		err = -EIO;
		goto cleanup;
	}
	lock_buffer(bh);

951 952
	err = ext4_journal_get_create_access(handle, bh);
	if (err)
A
Alex Tomas 已提交
953 954 955 956
		goto cleanup;

	neh = ext_block_hdr(bh);
	neh->eh_entries = 0;
957
	neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode, 0));
A
Alex Tomas 已提交
958 959 960
	neh->eh_magic = EXT4_EXT_MAGIC;
	neh->eh_depth = 0;

961
	/* move remainder of path[depth] to the new leaf */
962 963 964 965 966 967 968 969
	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;
	}
A
Alex Tomas 已提交
970
	/* start copy from next extent */
971 972
	m = EXT_MAX_EXTENT(path[depth].p_hdr) - path[depth].p_ext++;
	ext4_ext_show_move(inode, path, newblock, depth);
A
Alex Tomas 已提交
973
	if (m) {
974 975 976
		struct ext4_extent *ex;
		ex = EXT_FIRST_EXTENT(neh);
		memmove(ex, path[depth].p_ext, sizeof(struct ext4_extent) * m);
M
Marcin Slusarz 已提交
977
		le16_add_cpu(&neh->eh_entries, m);
A
Alex Tomas 已提交
978 979
	}

980
	ext4_extent_block_csum_set(inode, neh);
A
Alex Tomas 已提交
981 982 983
	set_buffer_uptodate(bh);
	unlock_buffer(bh);

984
	err = ext4_handle_dirty_metadata(handle, inode, bh);
985
	if (err)
A
Alex Tomas 已提交
986 987 988 989 990 991
		goto cleanup;
	brelse(bh);
	bh = NULL;

	/* correct old leaf */
	if (m) {
992 993
		err = ext4_ext_get_access(handle, inode, path + depth);
		if (err)
A
Alex Tomas 已提交
994
			goto cleanup;
M
Marcin Slusarz 已提交
995
		le16_add_cpu(&path[depth].p_hdr->eh_entries, -m);
996 997
		err = ext4_ext_dirty(handle, inode, path + depth);
		if (err)
A
Alex Tomas 已提交
998 999 1000 1001 1002 1003
			goto cleanup;

	}

	/* create intermediate indexes */
	k = depth - at - 1;
1004 1005 1006 1007 1008
	if (unlikely(k < 0)) {
		EXT4_ERROR_INODE(inode, "k %d < 0!", k);
		err = -EIO;
		goto cleanup;
	}
A
Alex Tomas 已提交
1009 1010 1011 1012 1013 1014 1015 1016
	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];
1017
		bh = sb_getblk(inode->i_sb, newblock);
A
Alex Tomas 已提交
1018 1019 1020 1021 1022 1023
		if (!bh) {
			err = -EIO;
			goto cleanup;
		}
		lock_buffer(bh);

1024 1025
		err = ext4_journal_get_create_access(handle, bh);
		if (err)
A
Alex Tomas 已提交
1026 1027 1028 1029 1030
			goto cleanup;

		neh = ext_block_hdr(bh);
		neh->eh_entries = cpu_to_le16(1);
		neh->eh_magic = EXT4_EXT_MAGIC;
1031
		neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode, 0));
A
Alex Tomas 已提交
1032 1033 1034
		neh->eh_depth = cpu_to_le16(depth - i);
		fidx = EXT_FIRST_INDEX(neh);
		fidx->ei_block = border;
1035
		ext4_idx_store_pblock(fidx, oldblock);
A
Alex Tomas 已提交
1036

1037 1038
		ext_debug("int.index at %d (block %llu): %u -> %llu\n",
				i, newblock, le32_to_cpu(border), oldblock);
A
Alex Tomas 已提交
1039

1040
		/* move remainder of path[i] to the new index block */
1041 1042 1043 1044 1045 1046 1047 1048
		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;
		}
1049 1050 1051 1052 1053
		/* 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);
A
Alex Tomas 已提交
1054
		if (m) {
1055
			memmove(++fidx, path[i].p_idx,
A
Alex Tomas 已提交
1056
				sizeof(struct ext4_extent_idx) * m);
M
Marcin Slusarz 已提交
1057
			le16_add_cpu(&neh->eh_entries, m);
A
Alex Tomas 已提交
1058
		}
1059
		ext4_extent_block_csum_set(inode, neh);
A
Alex Tomas 已提交
1060 1061 1062
		set_buffer_uptodate(bh);
		unlock_buffer(bh);

1063
		err = ext4_handle_dirty_metadata(handle, inode, bh);
1064
		if (err)
A
Alex Tomas 已提交
1065 1066 1067 1068 1069 1070 1071 1072 1073
			goto cleanup;
		brelse(bh);
		bh = NULL;

		/* correct old index */
		if (m) {
			err = ext4_ext_get_access(handle, inode, path + i);
			if (err)
				goto cleanup;
M
Marcin Slusarz 已提交
1074
			le16_add_cpu(&path[i].p_hdr->eh_entries, -m);
A
Alex Tomas 已提交
1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098
			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;
1099
			ext4_free_blocks(handle, inode, NULL, ablocks[i], 1,
1100
					 EXT4_FREE_BLOCKS_METADATA);
A
Alex Tomas 已提交
1101 1102 1103 1104 1105 1106 1107 1108
		}
	}
	kfree(ablocks);

	return err;
}

/*
1109 1110 1111 1112 1113 1114
 * 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
A
Alex Tomas 已提交
1115 1116
 */
static int ext4_ext_grow_indepth(handle_t *handle, struct inode *inode,
1117 1118
				 unsigned int flags,
				 struct ext4_extent *newext)
A
Alex Tomas 已提交
1119 1120 1121
{
	struct ext4_extent_header *neh;
	struct buffer_head *bh;
1122
	ext4_fsblk_t newblock;
A
Alex Tomas 已提交
1123 1124
	int err = 0;

1125
	newblock = ext4_ext_new_meta_block(handle, inode, NULL,
1126
		newext, &err, flags);
A
Alex Tomas 已提交
1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137
	if (newblock == 0)
		return err;

	bh = sb_getblk(inode->i_sb, newblock);
	if (!bh) {
		err = -EIO;
		ext4_std_error(inode->i_sb, err);
		return err;
	}
	lock_buffer(bh);

1138 1139
	err = ext4_journal_get_create_access(handle, bh);
	if (err) {
A
Alex Tomas 已提交
1140 1141 1142 1143 1144
		unlock_buffer(bh);
		goto out;
	}

	/* move top-level index/leaf into new block */
1145 1146
	memmove(bh->b_data, EXT4_I(inode)->i_data,
		sizeof(EXT4_I(inode)->i_data));
A
Alex Tomas 已提交
1147 1148 1149 1150 1151 1152

	/* 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))
1153
		neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode, 0));
A
Alex Tomas 已提交
1154
	else
1155
		neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode, 0));
A
Alex Tomas 已提交
1156
	neh->eh_magic = EXT4_EXT_MAGIC;
1157
	ext4_extent_block_csum_set(inode, neh);
A
Alex Tomas 已提交
1158 1159 1160
	set_buffer_uptodate(bh);
	unlock_buffer(bh);

1161
	err = ext4_handle_dirty_metadata(handle, inode, bh);
1162
	if (err)
A
Alex Tomas 已提交
1163 1164
		goto out;

1165
	/* Update top-level index: num,max,pointer */
A
Alex Tomas 已提交
1166
	neh = ext_inode_hdr(inode);
1167 1168 1169 1170 1171 1172 1173 1174
	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;
	}
1175
	ext_debug("new root: num %d(%d), lblock %d, ptr %llu\n",
A
Alex Tomas 已提交
1176
		  le16_to_cpu(neh->eh_entries), le16_to_cpu(neh->eh_max),
1177
		  le32_to_cpu(EXT_FIRST_INDEX(neh)->ei_block),
1178
		  ext4_idx_pblock(EXT_FIRST_INDEX(neh)));
A
Alex Tomas 已提交
1179

1180
	neh->eh_depth = cpu_to_le16(le16_to_cpu(neh->eh_depth) + 1);
1181
	ext4_mark_inode_dirty(handle, inode);
A
Alex Tomas 已提交
1182 1183 1184 1185 1186 1187 1188
out:
	brelse(bh);

	return err;
}

/*
1189 1190 1191
 * ext4_ext_create_new_leaf:
 * finds empty index and adds new leaf.
 * if no free index is found, then it requests in-depth growing.
A
Alex Tomas 已提交
1192 1193
 */
static int ext4_ext_create_new_leaf(handle_t *handle, struct inode *inode,
1194 1195 1196
				    unsigned int flags,
				    struct ext4_ext_path *path,
				    struct ext4_extent *newext)
A
Alex Tomas 已提交
1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210
{
	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--;
	}

1211 1212
	/* we use already allocated block for index block,
	 * so subsequent data blocks should be contiguous */
A
Alex Tomas 已提交
1213 1214 1215
	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 */
1216
		err = ext4_ext_split(handle, inode, flags, path, newext, i);
1217 1218
		if (err)
			goto out;
A
Alex Tomas 已提交
1219 1220 1221 1222

		/* refill path */
		ext4_ext_drop_refs(path);
		path = ext4_ext_find_extent(inode,
A
Aneesh Kumar K.V 已提交
1223 1224
				    (ext4_lblk_t)le32_to_cpu(newext->ee_block),
				    path);
A
Alex Tomas 已提交
1225 1226 1227 1228
		if (IS_ERR(path))
			err = PTR_ERR(path);
	} else {
		/* tree is full, time to grow in depth */
1229
		err = ext4_ext_grow_indepth(handle, inode, flags, newext);
A
Alex Tomas 已提交
1230 1231 1232 1233 1234 1235
		if (err)
			goto out;

		/* refill path */
		ext4_ext_drop_refs(path);
		path = ext4_ext_find_extent(inode,
A
Aneesh Kumar K.V 已提交
1236 1237
				   (ext4_lblk_t)le32_to_cpu(newext->ee_block),
				    path);
A
Alex Tomas 已提交
1238 1239 1240 1241 1242 1243
		if (IS_ERR(path)) {
			err = PTR_ERR(path);
			goto out;
		}

		/*
1244 1245
		 * only first (depth 0 -> 1) produces free space;
		 * in all other cases we have to split the grown tree
A
Alex Tomas 已提交
1246 1247 1248
		 */
		depth = ext_depth(inode);
		if (path[depth].p_hdr->eh_entries == path[depth].p_hdr->eh_max) {
1249
			/* now we need to split */
A
Alex Tomas 已提交
1250 1251 1252 1253 1254 1255 1256 1257
			goto repeat;
		}
	}

out:
	return err;
}

1258 1259 1260 1261 1262 1263 1264
/*
 * 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
 */
1265 1266 1267
static int ext4_ext_search_left(struct inode *inode,
				struct ext4_ext_path *path,
				ext4_lblk_t *logical, ext4_fsblk_t *phys)
1268 1269 1270
{
	struct ext4_extent_idx *ix;
	struct ext4_extent *ex;
1271
	int depth, ee_len;
1272

1273 1274 1275 1276
	if (unlikely(path == NULL)) {
		EXT4_ERROR_INODE(inode, "path == NULL *logical %d!", *logical);
		return -EIO;
	}
1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287
	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;
1288
	ee_len = ext4_ext_get_actual_len(ex);
1289
	if (*logical < le32_to_cpu(ex->ee_block)) {
1290 1291 1292 1293 1294 1295
		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;
		}
1296 1297
		while (--depth >= 0) {
			ix = path[depth].p_idx;
1298 1299 1300
			if (unlikely(ix != EXT_FIRST_INDEX(path[depth].p_hdr))) {
				EXT4_ERROR_INODE(inode,
				  "ix (%d) != EXT_FIRST_INDEX (%d) (depth %d)!",
1301
				  ix != NULL ? le32_to_cpu(ix->ei_block) : 0,
1302
				  EXT_FIRST_INDEX(path[depth].p_hdr) != NULL ?
1303
		le32_to_cpu(EXT_FIRST_INDEX(path[depth].p_hdr)->ei_block) : 0,
1304 1305 1306
				  depth);
				return -EIO;
			}
1307 1308 1309 1310
		}
		return 0;
	}

1311 1312 1313 1314 1315 1316
	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;
	}
1317

1318
	*logical = le32_to_cpu(ex->ee_block) + ee_len - 1;
1319
	*phys = ext4_ext_pblock(ex) + ee_len - 1;
1320 1321 1322 1323 1324 1325
	return 0;
}

/*
 * search the closest allocated block to the right for *logical
 * and returns it at @logical + it's physical address at @phys
1326
 * if *logical is the largest allocated block, the function
1327 1328 1329
 * returns 0 at @phys
 * return value contains 0 (success) or error code
 */
1330 1331
static int ext4_ext_search_right(struct inode *inode,
				 struct ext4_ext_path *path,
1332 1333
				 ext4_lblk_t *logical, ext4_fsblk_t *phys,
				 struct ext4_extent **ret_ex)
1334 1335 1336 1337 1338 1339
{
	struct buffer_head *bh = NULL;
	struct ext4_extent_header *eh;
	struct ext4_extent_idx *ix;
	struct ext4_extent *ex;
	ext4_fsblk_t block;
1340 1341
	int depth;	/* Note, NOT eh_depth; depth from top of tree */
	int ee_len;
1342

1343 1344 1345 1346
	if (unlikely(path == NULL)) {
		EXT4_ERROR_INODE(inode, "path == NULL *logical %d!", *logical);
		return -EIO;
	}
1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357
	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;
1358
	ee_len = ext4_ext_get_actual_len(ex);
1359
	if (*logical < le32_to_cpu(ex->ee_block)) {
1360 1361 1362 1363 1364 1365
		if (unlikely(EXT_FIRST_EXTENT(path[depth].p_hdr) != ex)) {
			EXT4_ERROR_INODE(inode,
					 "first_extent(path[%d].p_hdr) != ex",
					 depth);
			return -EIO;
		}
1366 1367
		while (--depth >= 0) {
			ix = path[depth].p_idx;
1368 1369 1370 1371 1372 1373
			if (unlikely(ix != EXT_FIRST_INDEX(path[depth].p_hdr))) {
				EXT4_ERROR_INODE(inode,
						 "ix != EXT_FIRST_INDEX *logical %d!",
						 *logical);
				return -EIO;
			}
1374
		}
1375
		goto found_extent;
1376 1377
	}

1378 1379 1380 1381 1382 1383
	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;
	}
1384 1385 1386 1387

	if (ex != EXT_LAST_EXTENT(path[depth].p_hdr)) {
		/* next allocated block in this leaf */
		ex++;
1388
		goto found_extent;
1389 1390 1391 1392 1393 1394
	}

	/* 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))
W
Wu Fengguang 已提交
1395
			goto got_index;
1396 1397
	}

W
Wu Fengguang 已提交
1398 1399
	/* we've gone up to the root and found no index to the right */
	return 0;
1400

W
Wu Fengguang 已提交
1401
got_index:
1402 1403 1404 1405
	/* we've found index to the right, let's
	 * follow it and find the closest allocated
	 * block to the right */
	ix++;
1406
	block = ext4_idx_pblock(ix);
1407 1408 1409 1410 1411
	while (++depth < path->p_depth) {
		bh = sb_bread(inode->i_sb, block);
		if (bh == NULL)
			return -EIO;
		eh = ext_block_hdr(bh);
1412
		/* subtract from p_depth to get proper eh_depth */
1413 1414
		if (ext4_ext_check_block(inode, eh,
					 path->p_depth - depth, bh)) {
1415 1416 1417 1418
			put_bh(bh);
			return -EIO;
		}
		ix = EXT_FIRST_INDEX(eh);
1419
		block = ext4_idx_pblock(ix);
1420 1421 1422 1423 1424 1425 1426
		put_bh(bh);
	}

	bh = sb_bread(inode->i_sb, block);
	if (bh == NULL)
		return -EIO;
	eh = ext_block_hdr(bh);
1427
	if (ext4_ext_check_block(inode, eh, path->p_depth - depth, bh)) {
1428 1429 1430 1431
		put_bh(bh);
		return -EIO;
	}
	ex = EXT_FIRST_EXTENT(eh);
1432
found_extent:
1433
	*logical = le32_to_cpu(ex->ee_block);
1434
	*phys = ext4_ext_pblock(ex);
1435 1436 1437
	*ret_ex = ex;
	if (bh)
		put_bh(bh);
1438 1439 1440
	return 0;
}

A
Alex Tomas 已提交
1441
/*
1442
 * ext4_ext_next_allocated_block:
1443
 * returns allocated block in subsequent extent or EXT_MAX_BLOCKS.
1444 1445 1446
 * NOTE: it considers block number from index entry as
 * allocated block. Thus, index entries have to be consistent
 * with leaves.
A
Alex Tomas 已提交
1447
 */
A
Aneesh Kumar K.V 已提交
1448
static ext4_lblk_t
A
Alex Tomas 已提交
1449 1450 1451 1452 1453 1454 1455 1456
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)
1457
		return EXT_MAX_BLOCKS;
A
Alex Tomas 已提交
1458 1459 1460 1461

	while (depth >= 0) {
		if (depth == path->p_depth) {
			/* leaf */
1462 1463
			if (path[depth].p_ext &&
				path[depth].p_ext !=
A
Alex Tomas 已提交
1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474
					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--;
	}

1475
	return EXT_MAX_BLOCKS;
A
Alex Tomas 已提交
1476 1477 1478
}

/*
1479
 * ext4_ext_next_leaf_block:
1480
 * returns first allocated block from next leaf or EXT_MAX_BLOCKS
A
Alex Tomas 已提交
1481
 */
1482
static ext4_lblk_t ext4_ext_next_leaf_block(struct ext4_ext_path *path)
A
Alex Tomas 已提交
1483 1484 1485 1486 1487 1488 1489 1490
{
	int depth;

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

	/* zero-tree has no leaf blocks at all */
	if (depth == 0)
1491
		return EXT_MAX_BLOCKS;
A
Alex Tomas 已提交
1492 1493 1494 1495 1496 1497 1498

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

	while (depth >= 0) {
		if (path[depth].p_idx !=
				EXT_LAST_INDEX(path[depth].p_hdr))
A
Aneesh Kumar K.V 已提交
1499 1500
			return (ext4_lblk_t)
				le32_to_cpu(path[depth].p_idx[1].ei_block);
A
Alex Tomas 已提交
1501 1502 1503
		depth--;
	}

1504
	return EXT_MAX_BLOCKS;
A
Alex Tomas 已提交
1505 1506 1507
}

/*
1508 1509 1510
 * ext4_ext_correct_indexes:
 * if leaf gets modified and modified extent is first in the leaf,
 * then we have to correct all indexes above.
A
Alex Tomas 已提交
1511 1512
 * TODO: do we need to correct tree in all cases?
 */
A
Aneesh Kumar K.V 已提交
1513
static int ext4_ext_correct_indexes(handle_t *handle, struct inode *inode,
A
Alex Tomas 已提交
1514 1515 1516 1517 1518 1519 1520 1521 1522 1523
				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;
1524 1525 1526 1527 1528 1529

	if (unlikely(ex == NULL || eh == NULL)) {
		EXT4_ERROR_INODE(inode,
				 "ex %p == NULL or eh %p == NULL", ex, eh);
		return -EIO;
	}
A
Alex Tomas 已提交
1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541

	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;
	}

	/*
1542
	 * TODO: we need correction if border is smaller than current one
A
Alex Tomas 已提交
1543 1544 1545
	 */
	k = depth - 1;
	border = path[depth].p_ext->ee_block;
1546 1547
	err = ext4_ext_get_access(handle, inode, path + k);
	if (err)
A
Alex Tomas 已提交
1548 1549
		return err;
	path[k].p_idx->ei_block = border;
1550 1551
	err = ext4_ext_dirty(handle, inode, path + k);
	if (err)
A
Alex Tomas 已提交
1552 1553 1554 1555 1556 1557
		return err;

	while (k--) {
		/* change all left-side indexes */
		if (path[k+1].p_idx != EXT_FIRST_INDEX(path[k+1].p_hdr))
			break;
1558 1559
		err = ext4_ext_get_access(handle, inode, path + k);
		if (err)
A
Alex Tomas 已提交
1560 1561
			break;
		path[k].p_idx->ei_block = border;
1562 1563
		err = ext4_ext_dirty(handle, inode, path + k);
		if (err)
A
Alex Tomas 已提交
1564 1565 1566 1567 1568 1569
			break;
	}

	return err;
}

1570
int
A
Alex Tomas 已提交
1571 1572 1573
ext4_can_extents_be_merged(struct inode *inode, struct ext4_extent *ex1,
				struct ext4_extent *ex2)
{
1574
	unsigned short ext1_ee_len, ext2_ee_len, max_len;
A
Amit Arora 已提交
1575 1576 1577 1578 1579 1580 1581 1582

	/*
	 * Make sure that either both extents are uninitialized, or
	 * both are _not_.
	 */
	if (ext4_ext_is_uninitialized(ex1) ^ ext4_ext_is_uninitialized(ex2))
		return 0;

1583 1584 1585 1586 1587
	if (ext4_ext_is_uninitialized(ex1))
		max_len = EXT_UNINIT_MAX_LEN;
	else
		max_len = EXT_INIT_MAX_LEN;

A
Amit Arora 已提交
1588 1589 1590 1591
	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 !=
A
Andrew Morton 已提交
1592
			le32_to_cpu(ex2->ee_block))
A
Alex Tomas 已提交
1593 1594
		return 0;

1595 1596 1597
	/*
	 * To allow future support for preallocated extents to be added
	 * as an RO_COMPAT feature, refuse to merge to extents if
1598
	 * this can result in the top bit of ee_len being set.
1599
	 */
1600
	if (ext1_ee_len + ext2_ee_len > max_len)
1601
		return 0;
1602
#ifdef AGGRESSIVE_TEST
1603
	if (ext1_ee_len >= 4)
A
Alex Tomas 已提交
1604 1605 1606
		return 0;
#endif

1607
	if (ext4_ext_pblock(ex1) + ext1_ee_len == ext4_ext_pblock(ex2))
A
Alex Tomas 已提交
1608 1609 1610 1611
		return 1;
	return 0;
}

1612 1613 1614 1615 1616 1617 1618
/*
 * 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.
 */
1619
static int ext4_ext_try_to_merge_right(struct inode *inode,
1620 1621
				 struct ext4_ext_path *path,
				 struct ext4_extent *ex)
1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647
{
	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);
		}
M
Marcin Slusarz 已提交
1648
		le16_add_cpu(&eh->eh_entries, -1);
1649 1650 1651
		merge_done = 1;
		WARN_ON(eh->eh_entries == 0);
		if (!eh->eh_entries)
1652
			EXT4_ERROR_INODE(inode, "eh->eh_entries = 0!");
1653 1654 1655 1656 1657
	}

	return merge_done;
}

1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682
/*
 * This function tries to merge the @ex extent to neighbours in the tree.
 * return 1 if merge left else 0.
 */
static int ext4_ext_try_to_merge(struct inode *inode,
				  struct ext4_ext_path *path,
				  struct ext4_extent *ex) {
	struct ext4_extent_header *eh;
	unsigned int depth;
	int merge_done = 0;
	int ret = 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)
		ret = ext4_ext_try_to_merge_right(inode, path, ex);

	return ret;
}

A
Amit Arora 已提交
1683 1684 1685 1686 1687 1688 1689 1690
/*
 * 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.
 */
1691 1692
static unsigned int ext4_ext_check_overlap(struct ext4_sb_info *sbi,
					   struct inode *inode,
1693 1694
					   struct ext4_extent *newext,
					   struct ext4_ext_path *path)
A
Amit Arora 已提交
1695
{
A
Aneesh Kumar K.V 已提交
1696
	ext4_lblk_t b1, b2;
A
Amit Arora 已提交
1697 1698 1699 1700
	unsigned int depth, len1;
	unsigned int ret = 0;

	b1 = le32_to_cpu(newext->ee_block);
A
Amit Arora 已提交
1701
	len1 = ext4_ext_get_actual_len(newext);
A
Amit Arora 已提交
1702 1703 1704 1705
	depth = ext_depth(inode);
	if (!path[depth].p_ext)
		goto out;
	b2 = le32_to_cpu(path[depth].p_ext->ee_block);
1706
	b2 &= ~(sbi->s_cluster_ratio - 1);
A
Amit Arora 已提交
1707 1708 1709

	/*
	 * get the next allocated block if the extent in the path
1710
	 * is before the requested block(s)
A
Amit Arora 已提交
1711 1712 1713
	 */
	if (b2 < b1) {
		b2 = ext4_ext_next_allocated_block(path);
1714
		if (b2 == EXT_MAX_BLOCKS)
A
Amit Arora 已提交
1715
			goto out;
1716
		b2 &= ~(sbi->s_cluster_ratio - 1);
A
Amit Arora 已提交
1717 1718
	}

A
Aneesh Kumar K.V 已提交
1719
	/* check for wrap through zero on extent logical start block*/
A
Amit Arora 已提交
1720
	if (b1 + len1 < b1) {
1721
		len1 = EXT_MAX_BLOCKS - b1;
A
Amit Arora 已提交
1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734
		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;
}

A
Alex Tomas 已提交
1735
/*
1736 1737 1738 1739
 * 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.
A
Alex Tomas 已提交
1740 1741 1742
 */
int ext4_ext_insert_extent(handle_t *handle, struct inode *inode,
				struct ext4_ext_path *path,
1743
				struct ext4_extent *newext, int flag)
A
Alex Tomas 已提交
1744
{
1745
	struct ext4_extent_header *eh;
A
Alex Tomas 已提交
1746 1747 1748
	struct ext4_extent *ex, *fex;
	struct ext4_extent *nearex; /* nearest extent */
	struct ext4_ext_path *npath = NULL;
A
Aneesh Kumar K.V 已提交
1749 1750
	int depth, len, err;
	ext4_lblk_t next;
A
Amit Arora 已提交
1751
	unsigned uninitialized = 0;
1752
	int flags = 0;
A
Alex Tomas 已提交
1753

1754 1755 1756 1757
	if (unlikely(ext4_ext_get_actual_len(newext) == 0)) {
		EXT4_ERROR_INODE(inode, "ext4_ext_get_actual_len(newext) == 0");
		return -EIO;
	}
A
Alex Tomas 已提交
1758 1759
	depth = ext_depth(inode);
	ex = path[depth].p_ext;
1760 1761 1762 1763
	if (unlikely(path[depth].p_hdr == NULL)) {
		EXT4_ERROR_INODE(inode, "path[%d].p_hdr == NULL", depth);
		return -EIO;
	}
A
Alex Tomas 已提交
1764 1765

	/* try to insert block into found extent and return */
1766
	if (ex && !(flag & EXT4_GET_BLOCKS_PRE_IO)
1767
		&& ext4_can_extents_be_merged(inode, ex, newext)) {
1768
		ext_debug("append [%d]%d block to %u:[%d]%d (from %llu)\n",
1769 1770 1771 1772 1773 1774
			  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));
1775 1776
		err = ext4_ext_get_access(handle, inode, path + depth);
		if (err)
A
Alex Tomas 已提交
1777
			return err;
A
Amit Arora 已提交
1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789

		/*
		 * 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)
					+ ext4_ext_get_actual_len(newext));
		if (uninitialized)
			ext4_ext_mark_uninitialized(ex);
A
Alex Tomas 已提交
1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801
		eh = path[depth].p_hdr;
		nearex = ex;
		goto merge;
	}

	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);
1802 1803
	next = EXT_MAX_BLOCKS;
	if (le32_to_cpu(newext->ee_block) > le32_to_cpu(fex->ee_block))
1804
		next = ext4_ext_next_leaf_block(path);
1805
	if (next != EXT_MAX_BLOCKS) {
1806
		ext_debug("next leaf block - %u\n", next);
A
Alex Tomas 已提交
1807 1808 1809 1810 1811 1812 1813
		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)) {
L
Lucas De Marchi 已提交
1814
			ext_debug("next leaf isn't full(%d)\n",
A
Alex Tomas 已提交
1815 1816
				  le16_to_cpu(eh->eh_entries));
			path = npath;
1817
			goto has_space;
A
Alex Tomas 已提交
1818 1819 1820 1821 1822 1823
		}
		ext_debug("next leaf has no free space(%d,%d)\n",
			  le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max));
	}

	/*
1824 1825
	 * There is no free space in the found leaf.
	 * We're gonna add a new leaf in the tree.
A
Alex Tomas 已提交
1826
	 */
1827 1828 1829
	if (flag & EXT4_GET_BLOCKS_PUNCH_OUT_EXT)
		flags = EXT4_MB_USE_ROOT_BLOCKS;
	err = ext4_ext_create_new_leaf(handle, inode, flags, path, newext);
A
Alex Tomas 已提交
1830 1831 1832 1833 1834 1835 1836 1837
	if (err)
		goto cleanup;
	depth = ext_depth(inode);
	eh = path[depth].p_hdr;

has_space:
	nearex = path[depth].p_ext;

1838 1839
	err = ext4_ext_get_access(handle, inode, path + depth);
	if (err)
A
Alex Tomas 已提交
1840 1841 1842 1843
		goto cleanup;

	if (!nearex) {
		/* there is no extent in this leaf, create first one */
1844
		ext_debug("first extent in the leaf: %u:%llu:[%d]%d\n",
D
Dave Kleikamp 已提交
1845
				le32_to_cpu(newext->ee_block),
1846
				ext4_ext_pblock(newext),
1847
				ext4_ext_is_uninitialized(newext),
A
Amit Arora 已提交
1848
				ext4_ext_get_actual_len(newext));
1849 1850 1851
		nearex = EXT_FIRST_EXTENT(eh);
	} else {
		if (le32_to_cpu(newext->ee_block)
D
Dave Kleikamp 已提交
1852
			   > le32_to_cpu(nearex->ee_block)) {
1853
			/* Insert after */
1854 1855
			ext_debug("insert %u:%llu:[%d]%d before: "
					"nearest %p\n",
1856 1857 1858 1859 1860 1861 1862 1863 1864
					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);
1865 1866
			ext_debug("insert %u:%llu:[%d]%d after: "
					"nearest %p\n",
D
Dave Kleikamp 已提交
1867
					le32_to_cpu(newext->ee_block),
1868
					ext4_ext_pblock(newext),
1869
					ext4_ext_is_uninitialized(newext),
A
Amit Arora 已提交
1870
					ext4_ext_get_actual_len(newext),
1871 1872 1873 1874
					nearex);
		}
		len = EXT_LAST_EXTENT(eh) - nearex + 1;
		if (len > 0) {
1875
			ext_debug("insert %u:%llu:[%d]%d: "
1876 1877 1878 1879 1880 1881 1882 1883
					"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));
A
Alex Tomas 已提交
1884 1885 1886
		}
	}

M
Marcin Slusarz 已提交
1887
	le16_add_cpu(&eh->eh_entries, 1);
1888
	path[depth].p_ext = nearex;
A
Alex Tomas 已提交
1889
	nearex->ee_block = newext->ee_block;
1890
	ext4_ext_store_pblock(nearex, ext4_ext_pblock(newext));
A
Alex Tomas 已提交
1891 1892 1893
	nearex->ee_len = newext->ee_len;

merge:
1894
	/* try to merge extents */
1895
	if (!(flag & EXT4_GET_BLOCKS_PRE_IO))
1896
		ext4_ext_try_to_merge(inode, path, nearex);
A
Alex Tomas 已提交
1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914


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

	err = ext4_ext_dirty(handle, inode, path + depth);

cleanup:
	if (npath) {
		ext4_ext_drop_refs(npath);
		kfree(npath);
	}
	ext4_ext_invalidate_cache(inode);
	return err;
}

1915 1916 1917
static int ext4_ext_walk_space(struct inode *inode, ext4_lblk_t block,
			       ext4_lblk_t num, ext_prepare_callback func,
			       void *cbdata)
1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928
{
	struct ext4_ext_path *path = NULL;
	struct ext4_ext_cache cbex;
	struct ext4_extent *ex;
	ext4_lblk_t next, start = 0, end = 0;
	ext4_lblk_t last = block + num;
	int depth, exists, err = 0;

	BUG_ON(func == NULL);
	BUG_ON(inode == NULL);

1929
	while (block < last && block != EXT_MAX_BLOCKS) {
1930 1931
		num = last - block;
		/* find extent for this block */
1932
		down_read(&EXT4_I(inode)->i_data_sem);
1933
		path = ext4_ext_find_extent(inode, block, path);
1934
		up_read(&EXT4_I(inode)->i_data_sem);
1935 1936 1937 1938 1939 1940 1941
		if (IS_ERR(path)) {
			err = PTR_ERR(path);
			path = NULL;
			break;
		}

		depth = ext_depth(inode);
1942 1943 1944 1945 1946
		if (unlikely(path[depth].p_hdr == NULL)) {
			EXT4_ERROR_INODE(inode, "path[%d].p_hdr == NULL", depth);
			err = -EIO;
			break;
		}
1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991
		ex = path[depth].p_ext;
		next = ext4_ext_next_allocated_block(path);

		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) {
			cbex.ec_block = start;
			cbex.ec_len = end - start;
			cbex.ec_start = 0;
		} else {
			cbex.ec_block = le32_to_cpu(ex->ee_block);
			cbex.ec_len = ext4_ext_get_actual_len(ex);
1992
			cbex.ec_start = ext4_ext_pblock(ex);
1993 1994
		}

1995 1996 1997 1998 1999
		if (unlikely(cbex.ec_len == 0)) {
			EXT4_ERROR_INODE(inode, "cbex.ec_len == 0");
			err = -EIO;
			break;
		}
2000
		err = func(inode, next, &cbex, ex, cbdata);
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029
		ext4_ext_drop_refs(path);

		if (err < 0)
			break;

		if (err == EXT_REPEAT)
			continue;
		else if (err == EXT_BREAK) {
			err = 0;
			break;
		}

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

		block = cbex.ec_block + cbex.ec_len;
	}

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

	return err;
}

2030
static void
A
Aneesh Kumar K.V 已提交
2031
ext4_ext_put_in_cache(struct inode *inode, ext4_lblk_t block,
2032
			__u32 len, ext4_fsblk_t start)
A
Alex Tomas 已提交
2033 2034 2035
{
	struct ext4_ext_cache *cex;
	BUG_ON(len == 0);
2036
	spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
2037
	trace_ext4_ext_put_in_cache(inode, block, len, start);
A
Alex Tomas 已提交
2038 2039 2040 2041
	cex = &EXT4_I(inode)->i_cached_extent;
	cex->ec_block = block;
	cex->ec_len = len;
	cex->ec_start = start;
2042
	spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
A
Alex Tomas 已提交
2043 2044 2045
}

/*
2046 2047
 * ext4_ext_put_gap_in_cache:
 * calculate boundaries of the gap that the requested block fits into
A
Alex Tomas 已提交
2048 2049
 * and cache this gap
 */
2050
static void
A
Alex Tomas 已提交
2051
ext4_ext_put_gap_in_cache(struct inode *inode, struct ext4_ext_path *path,
A
Aneesh Kumar K.V 已提交
2052
				ext4_lblk_t block)
A
Alex Tomas 已提交
2053 2054
{
	int depth = ext_depth(inode);
A
Aneesh Kumar K.V 已提交
2055 2056
	unsigned long len;
	ext4_lblk_t lblock;
A
Alex Tomas 已提交
2057 2058 2059 2060 2061 2062
	struct ext4_extent *ex;

	ex = path[depth].p_ext;
	if (ex == NULL) {
		/* there is no extent yet, so gap is [0;-] */
		lblock = 0;
2063
		len = EXT_MAX_BLOCKS;
A
Alex Tomas 已提交
2064 2065 2066 2067
		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;
2068 2069 2070 2071
		ext_debug("cache gap(before): %u [%u:%u]",
				block,
				le32_to_cpu(ex->ee_block),
				 ext4_ext_get_actual_len(ex));
A
Alex Tomas 已提交
2072
	} else if (block >= le32_to_cpu(ex->ee_block)
A
Amit Arora 已提交
2073
			+ ext4_ext_get_actual_len(ex)) {
A
Aneesh Kumar K.V 已提交
2074
		ext4_lblk_t next;
D
Dave Kleikamp 已提交
2075
		lblock = le32_to_cpu(ex->ee_block)
A
Amit Arora 已提交
2076
			+ ext4_ext_get_actual_len(ex);
A
Aneesh Kumar K.V 已提交
2077 2078

		next = ext4_ext_next_allocated_block(path);
2079 2080 2081 2082
		ext_debug("cache gap(after): [%u:%u] %u",
				le32_to_cpu(ex->ee_block),
				ext4_ext_get_actual_len(ex),
				block);
A
Aneesh Kumar K.V 已提交
2083 2084
		BUG_ON(next == lblock);
		len = next - lblock;
A
Alex Tomas 已提交
2085 2086 2087 2088 2089
	} else {
		lblock = len = 0;
		BUG();
	}

2090
	ext_debug(" -> %u:%lu\n", lblock, len);
2091
	ext4_ext_put_in_cache(inode, lblock, len, 0);
A
Alex Tomas 已提交
2092 2093
}

2094
/*
2095
 * ext4_ext_check_cache()
2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107
 * Checks to see if the given block is in the cache.
 * If it is, the cached extent is stored in the given
 * cache extent pointer.  If the cached extent is a hole,
 * this routine should be used instead of
 * ext4_ext_in_cache if the calling function needs to
 * know the size of the hole.
 *
 * @inode: The files inode
 * @block: The block to look for in the cache
 * @ex:    Pointer where the cached extent will be stored
 *         if it contains block
 *
2108 2109
 * Return 0 if cache is invalid; 1 if the cache is valid
 */
2110 2111
static int ext4_ext_check_cache(struct inode *inode, ext4_lblk_t block,
	struct ext4_ext_cache *ex){
A
Alex Tomas 已提交
2112
	struct ext4_ext_cache *cex;
2113
	struct ext4_sb_info *sbi;
2114
	int ret = 0;
A
Alex Tomas 已提交
2115

2116
	/*
2117 2118 2119
	 * We borrow i_block_reservation_lock to protect i_cached_extent
	 */
	spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
A
Alex Tomas 已提交
2120
	cex = &EXT4_I(inode)->i_cached_extent;
2121
	sbi = EXT4_SB(inode->i_sb);
A
Alex Tomas 已提交
2122 2123

	/* has cache valid data? */
2124
	if (cex->ec_len == 0)
2125
		goto errout;
A
Alex Tomas 已提交
2126

2127
	if (in_range(block, cex->ec_block, cex->ec_len)) {
2128
		memcpy(ex, cex, sizeof(struct ext4_ext_cache));
2129 2130 2131
		ext_debug("%u cached by %u:%u:%llu\n",
				block,
				cex->ec_block, cex->ec_len, cex->ec_start);
2132
		ret = 1;
A
Alex Tomas 已提交
2133
	}
2134
errout:
2135
	trace_ext4_ext_in_cache(inode, block, ret);
2136 2137
	spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
	return ret;
A
Alex Tomas 已提交
2138 2139
}

2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170
/*
 * ext4_ext_in_cache()
 * Checks to see if the given block is in the cache.
 * If it is, the cached extent is stored in the given
 * extent pointer.
 *
 * @inode: The files inode
 * @block: The block to look for in the cache
 * @ex:    Pointer where the cached extent will be stored
 *         if it contains block
 *
 * Return 0 if cache is invalid; 1 if the cache is valid
 */
static int
ext4_ext_in_cache(struct inode *inode, ext4_lblk_t block,
			struct ext4_extent *ex)
{
	struct ext4_ext_cache cex;
	int ret = 0;

	if (ext4_ext_check_cache(inode, block, &cex)) {
		ex->ee_block = cpu_to_le32(cex.ec_block);
		ext4_ext_store_pblock(ex, cex.ec_start);
		ex->ee_len = cpu_to_le16(cex.ec_len);
		ret = 1;
	}

	return ret;
}


A
Alex Tomas 已提交
2171
/*
2172 2173
 * ext4_ext_rm_idx:
 * removes index from the index block.
A
Alex Tomas 已提交
2174
 */
A
Aneesh Kumar K.V 已提交
2175
static int ext4_ext_rm_idx(handle_t *handle, struct inode *inode,
A
Alex Tomas 已提交
2176 2177 2178
			struct ext4_ext_path *path)
{
	int err;
2179
	ext4_fsblk_t leaf;
A
Alex Tomas 已提交
2180 2181 2182

	/* free index block */
	path--;
2183
	leaf = ext4_idx_pblock(path->p_idx);
2184 2185 2186 2187
	if (unlikely(path->p_hdr->eh_entries == 0)) {
		EXT4_ERROR_INODE(inode, "path->p_hdr->eh_entries == 0");
		return -EIO;
	}
2188 2189
	err = ext4_ext_get_access(handle, inode, path);
	if (err)
A
Alex Tomas 已提交
2190
		return err;
2191 2192 2193 2194 2195 2196 2197

	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);
	}

M
Marcin Slusarz 已提交
2198
	le16_add_cpu(&path->p_hdr->eh_entries, -1);
2199 2200
	err = ext4_ext_dirty(handle, inode, path);
	if (err)
A
Alex Tomas 已提交
2201
		return err;
2202
	ext_debug("index is empty, remove it, free block %llu\n", leaf);
2203 2204
	trace_ext4_ext_rm_idx(inode, leaf);

2205
	ext4_free_blocks(handle, inode, NULL, leaf, 1,
2206
			 EXT4_FREE_BLOCKS_METADATA | EXT4_FREE_BLOCKS_FORGET);
A
Alex Tomas 已提交
2207 2208 2209 2210
	return err;
}

/*
2211 2212 2213 2214 2215
 * 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.
A
Alex Tomas 已提交
2216
 */
2217
int ext4_ext_calc_credits_for_single_extent(struct inode *inode, int nrblocks,
A
Alex Tomas 已提交
2218 2219 2220
						struct ext4_ext_path *path)
{
	if (path) {
2221
		int depth = ext_depth(inode);
2222
		int ret = 0;
2223

A
Alex Tomas 已提交
2224 2225
		/* probably there is space in leaf? */
		if (le16_to_cpu(path[depth].p_hdr->eh_entries)
2226
				< le16_to_cpu(path[depth].p_hdr->eh_max)) {
A
Alex Tomas 已提交
2227

2228 2229 2230 2231 2232
			/*
			 *  There are some space in the leaf tree, no
			 *  need to account for leaf block credit
			 *
			 *  bitmaps and block group descriptor blocks
2233
			 *  and other metadata blocks still need to be
2234 2235
			 *  accounted.
			 */
2236
			/* 1 bitmap, 1 block group descriptor */
2237
			ret = 2 + EXT4_META_TRANS_BLOCKS(inode->i_sb);
2238
			return ret;
2239 2240
		}
	}
A
Alex Tomas 已提交
2241

2242
	return ext4_chunk_trans_blocks(inode, nrblocks);
2243
}
A
Alex Tomas 已提交
2244

2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255
/*
 * 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.
 */
2256
int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks, int chunk)
2257 2258 2259
{
	int index;
	int depth = ext_depth(inode);
A
Alex Tomas 已提交
2260

2261 2262 2263 2264
	if (chunk)
		index = depth * 2;
	else
		index = depth * 3;
A
Alex Tomas 已提交
2265

2266
	return index;
A
Alex Tomas 已提交
2267 2268 2269
}

static int ext4_remove_blocks(handle_t *handle, struct inode *inode,
2270 2271 2272
			      struct ext4_extent *ex,
			      ext4_fsblk_t *partial_cluster,
			      ext4_lblk_t from, ext4_lblk_t to)
A
Alex Tomas 已提交
2273
{
2274
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
A
Amit Arora 已提交
2275
	unsigned short ee_len =  ext4_ext_get_actual_len(ex);
2276
	ext4_fsblk_t pblk;
2277
	int flags = EXT4_FREE_BLOCKS_FORGET;
A
Alex Tomas 已提交
2278

2279
	if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
2280
		flags |= EXT4_FREE_BLOCKS_METADATA;
2281 2282 2283 2284 2285 2286 2287 2288 2289
	/*
	 * 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;

2290
	trace_ext4_remove_blocks(inode, ex, from, to, *partial_cluster);
2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303
	/*
	 * 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;
	}

A
Alex Tomas 已提交
2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319
#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)
A
Amit Arora 已提交
2320
	    && to == le32_to_cpu(ex->ee_block) + ee_len - 1) {
A
Alex Tomas 已提交
2321
		/* tail removal */
A
Aneesh Kumar K.V 已提交
2322 2323
		ext4_lblk_t num;

A
Amit Arora 已提交
2324
		num = le32_to_cpu(ex->ee_block) + ee_len - from;
2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340
		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;
A
Alex Tomas 已提交
2341
	} else if (from == le32_to_cpu(ex->ee_block)
A
Amit Arora 已提交
2342
		   && to <= le32_to_cpu(ex->ee_block) + ee_len - 1) {
A
Allison Henderson 已提交
2343 2344 2345 2346 2347 2348 2349 2350
		/* 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);
2351
		ext4_free_blocks(handle, inode, NULL, start, num, flags);
A
Allison Henderson 已提交
2352

A
Alex Tomas 已提交
2353
	} else {
A
Aneesh Kumar K.V 已提交
2354 2355 2356
		printk(KERN_INFO "strange request: removal(2) "
				"%u-%u from %u:%u\n",
				from, to, le32_to_cpu(ex->ee_block), ee_len);
A
Alex Tomas 已提交
2357 2358 2359 2360
	}
	return 0;
}

A
Allison Henderson 已提交
2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372

/*
 * 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
 */
A
Alex Tomas 已提交
2373 2374
static int
ext4_ext_rm_leaf(handle_t *handle, struct inode *inode,
2375 2376
		 struct ext4_ext_path *path, ext4_fsblk_t *partial_cluster,
		 ext4_lblk_t start, ext4_lblk_t end)
A
Alex Tomas 已提交
2377
{
2378
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
A
Alex Tomas 已提交
2379 2380 2381
	int err = 0, correct_index = 0;
	int depth = ext_depth(inode), credits;
	struct ext4_extent_header *eh;
2382
	ext4_lblk_t a, b;
A
Aneesh Kumar K.V 已提交
2383 2384
	unsigned num;
	ext4_lblk_t ex_ee_block;
A
Alex Tomas 已提交
2385
	unsigned short ex_ee_len;
A
Amit Arora 已提交
2386
	unsigned uninitialized = 0;
A
Alex Tomas 已提交
2387 2388
	struct ext4_extent *ex;

2389
	/* the header must be checked already in ext4_ext_remove_space() */
2390
	ext_debug("truncate since %u in leaf to %u\n", start, end);
A
Alex Tomas 已提交
2391 2392 2393
	if (!path[depth].p_hdr)
		path[depth].p_hdr = ext_block_hdr(path[depth].p_bh);
	eh = path[depth].p_hdr;
2394 2395 2396 2397
	if (unlikely(path[depth].p_hdr == NULL)) {
		EXT4_ERROR_INODE(inode, "path[%d].p_hdr == NULL", depth);
		return -EIO;
	}
A
Alex Tomas 已提交
2398 2399 2400 2401
	/* find where to start removing */
	ex = EXT_LAST_EXTENT(eh);

	ex_ee_block = le32_to_cpu(ex->ee_block);
A
Amit Arora 已提交
2402
	ex_ee_len = ext4_ext_get_actual_len(ex);
A
Alex Tomas 已提交
2403

2404 2405
	trace_ext4_ext_rm_leaf(inode, start, ex, *partial_cluster);

A
Alex Tomas 已提交
2406 2407
	while (ex >= EXT_FIRST_EXTENT(eh) &&
			ex_ee_block + ex_ee_len > start) {
2408 2409 2410 2411 2412 2413

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

2414 2415
		ext_debug("remove ext %u:[%d]%d\n", ex_ee_block,
			 uninitialized, ex_ee_len);
A
Alex Tomas 已提交
2416 2417 2418
		path[depth].p_ext = ex;

		a = ex_ee_block > start ? ex_ee_block : start;
A
Allison Henderson 已提交
2419 2420
		b = ex_ee_block+ex_ee_len - 1 < end ?
			ex_ee_block+ex_ee_len - 1 : end;
A
Alex Tomas 已提交
2421 2422 2423

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

A
Allison Henderson 已提交
2424
		/* If this extent is beyond the end of the hole, skip it */
2425
		if (end < ex_ee_block) {
A
Allison Henderson 已提交
2426 2427 2428 2429
			ex--;
			ex_ee_block = le32_to_cpu(ex->ee_block);
			ex_ee_len = ext4_ext_get_actual_len(ex);
			continue;
2430
		} else if (b != ex_ee_block + ex_ee_len - 1) {
2431 2432 2433 2434 2435
			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);
2436 2437
			err = -EIO;
			goto out;
A
Alex Tomas 已提交
2438 2439
		} else if (a != ex_ee_block) {
			/* remove tail of the extent */
2440
			num = a - ex_ee_block;
A
Alex Tomas 已提交
2441 2442 2443 2444
		} else {
			/* remove whole extent: excellent! */
			num = 0;
		}
2445 2446 2447 2448 2449 2450 2451
		/*
		 * 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));
A
Alex Tomas 已提交
2452 2453 2454 2455
		if (ex == EXT_FIRST_EXTENT(eh)) {
			correct_index = 1;
			credits += (ext_depth(inode)) + 1;
		}
D
Dmitry Monakhov 已提交
2456
		credits += EXT4_MAXQUOTAS_TRANS_BLOCKS(inode->i_sb);
A
Alex Tomas 已提交
2457

2458
		err = ext4_ext_truncate_extend_restart(handle, inode, credits);
2459
		if (err)
A
Alex Tomas 已提交
2460 2461 2462 2463 2464 2465
			goto out;

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

2466 2467
		err = ext4_remove_blocks(handle, inode, ex, partial_cluster,
					 a, b);
A
Alex Tomas 已提交
2468 2469 2470
		if (err)
			goto out;

2471
		if (num == 0)
2472
			/* this extent is removed; mark slot entirely unused */
2473
			ext4_ext_store_pblock(ex, 0);
A
Alex Tomas 已提交
2474 2475

		ex->ee_len = cpu_to_le16(num);
2476 2477 2478 2479 2480
		/*
		 * Do not mark uninitialized if all the blocks in the
		 * extent have been removed.
		 */
		if (uninitialized && num)
A
Amit Arora 已提交
2481
			ext4_ext_mark_uninitialized(ex);
A
Allison Henderson 已提交
2482 2483 2484 2485 2486
		/*
		 * If the extent was completely released,
		 * we need to remove it from the leaf
		 */
		if (num == 0) {
2487
			if (end != EXT_MAX_BLOCKS - 1) {
A
Allison Henderson 已提交
2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500
				/*
				 * 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);
2501 2502
		} else
			*partial_cluster = 0;
A
Allison Henderson 已提交
2503

2504 2505 2506 2507
		err = ext4_ext_dirty(handle, inode, path + depth);
		if (err)
			goto out;

2508
		ext_debug("new extent: %u:%u:%llu\n", ex_ee_block, num,
2509
				ext4_ext_pblock(ex));
A
Alex Tomas 已提交
2510 2511
		ex--;
		ex_ee_block = le32_to_cpu(ex->ee_block);
A
Amit Arora 已提交
2512
		ex_ee_len = ext4_ext_get_actual_len(ex);
A
Alex Tomas 已提交
2513 2514 2515 2516 2517
	}

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

2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536
	/*
	 * 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;
	}

A
Alex Tomas 已提交
2537 2538 2539 2540 2541 2542 2543 2544 2545 2546
	/* 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)
		err = ext4_ext_rm_idx(handle, inode, path + depth);

out:
	return err;
}

/*
2547 2548
 * ext4_ext_more_to_rm:
 * returns 1 if current index has to be freed (even partial)
A
Alex Tomas 已提交
2549
 */
2550
static int
A
Alex Tomas 已提交
2551 2552 2553 2554 2555 2556 2557 2558
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;

	/*
2559
	 * if truncate on deeper level happened, it wasn't partial,
A
Alex Tomas 已提交
2560 2561 2562 2563 2564 2565 2566
	 * 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;
}

2567 2568
static int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
				 ext4_lblk_t end)
A
Alex Tomas 已提交
2569 2570 2571 2572
{
	struct super_block *sb = inode->i_sb;
	int depth = ext_depth(inode);
	struct ext4_ext_path *path;
2573
	ext4_fsblk_t partial_cluster = 0;
A
Alex Tomas 已提交
2574
	handle_t *handle;
2575
	int i, err;
A
Alex Tomas 已提交
2576

2577
	ext_debug("truncate since %u to %u\n", start, end);
A
Alex Tomas 已提交
2578 2579 2580 2581 2582 2583

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

2584
again:
A
Alex Tomas 已提交
2585 2586
	ext4_ext_invalidate_cache(inode);

2587 2588
	trace_ext4_ext_remove_space(inode, start, depth);

2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643
	/*
	 * 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);
		ex = path[depth].p_ext;
		if (!ex)
			goto cont;

		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
			 * block in the first new extent
			 */
			err = ext4_split_extent_at(handle, inode, path,
						end + 1, split_flag,
						EXT4_GET_BLOCKS_PRE_IO |
						EXT4_GET_BLOCKS_PUNCH_OUT_EXT);

			if (err < 0)
				goto out;
		}
		ext4_ext_drop_refs(path);
		kfree(path);
	}
cont:

A
Alex Tomas 已提交
2644
	/*
2645 2646
	 * We start scanning from right side, freeing all the blocks
	 * after i_size and walking into the tree depth-wise.
A
Alex Tomas 已提交
2647
	 */
2648
	depth = ext_depth(inode);
2649
	path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 1), GFP_NOFS);
A
Alex Tomas 已提交
2650 2651 2652 2653
	if (path == NULL) {
		ext4_journal_stop(handle);
		return -ENOMEM;
	}
2654
	path[0].p_depth = depth;
A
Alex Tomas 已提交
2655
	path[0].p_hdr = ext_inode_hdr(inode);
2656

2657
	if (ext4_ext_check(inode, path[0].p_hdr, depth)) {
A
Alex Tomas 已提交
2658 2659 2660
		err = -EIO;
		goto out;
	}
2661
	i = err = 0;
A
Alex Tomas 已提交
2662 2663 2664 2665

	while (i >= 0 && err == 0) {
		if (i == depth) {
			/* this is leaf block */
A
Allison Henderson 已提交
2666
			err = ext4_ext_rm_leaf(handle, inode, path,
2667
					       &partial_cluster, start,
2668
					       end);
2669
			/* root level has p_bh == NULL, brelse() eats this */
A
Alex Tomas 已提交
2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682
			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) {
2683
			/* this level hasn't been touched yet */
A
Alex Tomas 已提交
2684 2685 2686 2687 2688 2689
			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 {
2690
			/* we were already here, see at next index */
A
Alex Tomas 已提交
2691 2692 2693 2694 2695 2696 2697
			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)) {
2698
			struct buffer_head *bh;
A
Alex Tomas 已提交
2699
			/* go to the next level */
2700
			ext_debug("move to level %d (block %llu)\n",
2701
				  i + 1, ext4_idx_pblock(path[i].p_idx));
A
Alex Tomas 已提交
2702
			memset(path + i + 1, 0, sizeof(*path));
2703
			bh = sb_bread(sb, ext4_idx_pblock(path[i].p_idx));
2704
			if (!bh) {
A
Alex Tomas 已提交
2705 2706 2707 2708
				/* should we reset i_size? */
				err = -EIO;
				break;
			}
2709 2710 2711 2712
			if (WARN_ON(i + 1 > depth)) {
				err = -EIO;
				break;
			}
2713 2714
			if (ext4_ext_check_block(inode, ext_block_hdr(bh),
							depth - i - 1, bh)) {
2715 2716 2717 2718
				err = -EIO;
				break;
			}
			path[i + 1].p_bh = bh;
A
Alex Tomas 已提交
2719

2720 2721
			/* save actual number of indexes since this
			 * number is changed at the next iteration */
A
Alex Tomas 已提交
2722 2723 2724
			path[i].p_block = le16_to_cpu(path[i].p_hdr->eh_entries);
			i++;
		} else {
2725
			/* we finished processing this index, go up */
A
Alex Tomas 已提交
2726
			if (path[i].p_hdr->eh_entries == 0 && i > 0) {
2727
				/* index is empty, remove it;
A
Alex Tomas 已提交
2728 2729 2730 2731
				 * handle must be already prepared by the
				 * truncatei_leaf() */
				err = ext4_ext_rm_idx(handle, inode, path + i);
			}
2732
			/* root level has p_bh == NULL, brelse() eats this */
A
Alex Tomas 已提交
2733 2734 2735 2736 2737 2738 2739
			brelse(path[i].p_bh);
			path[i].p_bh = NULL;
			i--;
			ext_debug("return to level %d\n", i);
		}
	}

2740 2741 2742
	trace_ext4_ext_remove_space_done(inode, start, depth, partial_cluster,
			path->p_hdr->eh_entries);

2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757
	/* 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;
	}

A
Alex Tomas 已提交
2758 2759 2760
	/* TODO: flexible tree reduction should be here */
	if (path->p_hdr->eh_entries == 0) {
		/*
2761 2762
		 * truncate to zero freed all the tree,
		 * so we need to correct eh_depth
A
Alex Tomas 已提交
2763 2764 2765 2766 2767
		 */
		err = ext4_ext_get_access(handle, inode, path);
		if (err == 0) {
			ext_inode_hdr(inode)->eh_depth = 0;
			ext_inode_hdr(inode)->eh_max =
2768
				cpu_to_le16(ext4_ext_space_root(inode, 0));
A
Alex Tomas 已提交
2769 2770 2771 2772 2773 2774
			err = ext4_ext_dirty(handle, inode, path);
		}
	}
out:
	ext4_ext_drop_refs(path);
	kfree(path);
2775 2776
	if (err == -EAGAIN)
		goto again;
A
Alex Tomas 已提交
2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790
	ext4_journal_stop(handle);

	return err;
}

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

2791
	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2792
#if defined(AGGRESSIVE_TEST) || defined(CHECK_BINSEARCH) || defined(EXTENTS_STATS)
2793
		printk(KERN_INFO "EXT4-fs: file extents enabled"
2794
#ifdef AGGRESSIVE_TEST
2795
		       ", aggressive tests"
A
Alex Tomas 已提交
2796 2797
#endif
#ifdef CHECK_BINSEARCH
2798
		       ", check binsearch"
A
Alex Tomas 已提交
2799 2800
#endif
#ifdef EXTENTS_STATS
2801
		       ", stats"
A
Alex Tomas 已提交
2802
#endif
2803
		       "\n");
2804
#endif
A
Alex Tomas 已提交
2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817
#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)
{
2818
	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
A
Alex Tomas 已提交
2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832
		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
}

2833 2834 2835
/* 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)
{
2836 2837
	ext4_fsblk_t ee_pblock;
	unsigned int ee_len;
2838
	int ret;
2839 2840

	ee_len    = ext4_ext_get_actual_len(ex);
2841
	ee_pblock = ext4_ext_pblock(ex);
2842

2843
	ret = sb_issue_zeroout(inode->i_sb, ee_pblock, ee_len, GFP_NOFS);
2844 2845
	if (ret > 0)
		ret = 0;
2846

2847
	return ret;
2848 2849
}

2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946
/*
 * 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;
	struct ext4_extent *ex, newex, orig_ex;
	struct ext4_extent *ex2 = NULL;
	unsigned int ee_len, depth;
	int err = 0;

	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));

	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))
			ext4_ext_try_to_merge(inode, path, ex);

		err = ext4_ext_dirty(handle, inode, path + depth);
		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)) {
		err = ext4_ext_zeroout(inode, &orig_ex);
		if (err)
			goto fix_extent_len;
		/* update the extent length and mark as initialized */
2947
		ex->ee_len = cpu_to_le16(ee_len);
2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 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 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003
		ext4_ext_try_to_merge(inode, path, ex);
		err = ext4_ext_dirty(handle, inode, path + depth);
		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;

	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) {
		split_flag1 = split_flag & EXT4_EXT_MAY_ZEROOUT ?
			      EXT4_EXT_MAY_ZEROOUT : 0;
		flags1 = flags | EXT4_GET_BLOCKS_PRE_IO;
		if (uninitialized)
			split_flag1 |= EXT4_EXT_MARK_UNINIT1 |
				       EXT4_EXT_MARK_UNINIT2;
		err = ext4_split_extent_at(handle, inode, path,
				map->m_lblk + map->m_len, split_flag1, flags1);
3004 3005
		if (err)
			goto out;
3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030
	}

	ext4_ext_drop_refs(path);
	path = ext4_ext_find_extent(inode, map->m_lblk, path);
	if (IS_ERR(path))
		return PTR_ERR(path);

	if (map->m_lblk >= ee_block) {
		split_flag1 = split_flag & EXT4_EXT_MAY_ZEROOUT ?
			      EXT4_EXT_MAY_ZEROOUT : 0;
		if (uninitialized)
			split_flag1 |= EXT4_EXT_MARK_UNINIT1;
		if (split_flag & EXT4_EXT_MARK_UNINIT2)
			split_flag1 |= EXT4_EXT_MARK_UNINIT2;
		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:
	return err ? err : map->m_len;
}

3031
#define EXT4_EXT_ZERO_LEN 7
3032
/*
3033
 * This function is called by ext4_ext_map_blocks() if someone tries to write
3034
 * to an uninitialized extent. It may result in splitting the uninitialized
L
Lucas De Marchi 已提交
3035
 * extent into multiple extents (up to three - one initialized and two
3036 3037 3038 3039 3040
 * 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
3041 3042 3043 3044 3045 3046 3047 3048 3049 3050
 *
 * 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.
3051
 */
A
Aneesh Kumar K.V 已提交
3052
static int ext4_ext_convert_to_initialized(handle_t *handle,
3053 3054 3055
					   struct inode *inode,
					   struct ext4_map_blocks *map,
					   struct ext4_ext_path *path)
3056
{
3057
	struct ext4_extent_header *eh;
3058 3059 3060
	struct ext4_map_blocks split_map;
	struct ext4_extent zero_ex;
	struct ext4_extent *ex;
3061
	ext4_lblk_t ee_block, eof_block;
3062 3063
	unsigned int ee_len, depth;
	int allocated;
3064
	int err = 0;
3065
	int split_flag = 0;
3066 3067 3068

	ext_debug("ext4_ext_convert_to_initialized: inode %lu, logical"
		"block %llu, max_blocks %u\n", inode->i_ino,
3069
		(unsigned long long)map->m_lblk, map->m_len);
3070 3071 3072

	eof_block = (inode->i_size + inode->i_sb->s_blocksize - 1) >>
		inode->i_sb->s_blocksize_bits;
3073 3074
	if (eof_block < map->m_lblk + map->m_len)
		eof_block = map->m_lblk + map->m_len;
3075 3076

	depth = ext_depth(inode);
3077
	eh = path[depth].p_hdr;
3078 3079 3080
	ex = path[depth].p_ext;
	ee_block = le32_to_cpu(ex->ee_block);
	ee_len = ext4_ext_get_actual_len(ex);
3081
	allocated = ee_len - (map->m_lblk - ee_block);
3082

3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 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 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162
	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
	 * uninitialized extent to its left neighbor. This is much cheaper
	 * than an insertion followed by a merge as those involve costly
	 * memmove() calls. This is the common case in steady state for
	 * workloads doing fallocate(FALLOC_FL_KEEP_SIZE) followed by append
	 * writes.
	 *
	 * Limitations of the current logic:
	 *  - L1: we only deal with writes at the start of the extent.
	 *    The approach could be extended to writes at the end
	 *    of the extent but this scenario was deemed less common.
	 *  - L2: we do not deal with writes covering the whole extent.
	 *    This would require removing the extent if the transfer
	 *    is possible.
	 *  - L3: we only attempt to merge with an extent stored in the
	 *    same extent tree node.
	 */
	if ((map->m_lblk == ee_block) &&	/*L1*/
		(map->m_len < ee_len) &&	/*L2*/
		(ex > EXT_FIRST_EXTENT(eh))) {	/*L3*/
		struct ext4_extent *prev_ex;
		ext4_lblk_t prev_lblk;
		ext4_fsblk_t prev_pblk, ee_pblk;
		unsigned int prev_len, write_len;

		prev_ex = ex - 1;
		prev_lblk = le32_to_cpu(prev_ex->ee_block);
		prev_len = ext4_ext_get_actual_len(prev_ex);
		prev_pblk = ext4_ext_pblock(prev_ex);
		ee_pblk = ext4_ext_pblock(ex);
		write_len = map->m_len;

		/*
		 * A transfer of blocks from 'ex' to 'prev_ex' is allowed
		 * upon those conditions:
		 * - C1: prev_ex is initialized,
		 * - C2: prev_ex is logically abutting ex,
		 * - C3: prev_ex is physically abutting ex,
		 * - C4: prev_ex can receive the additional blocks without
		 *   overflowing the (initialized) length limit.
		 */
		if ((!ext4_ext_is_uninitialized(prev_ex)) &&		/*C1*/
			((prev_lblk + prev_len) == ee_block) &&		/*C2*/
			((prev_pblk + prev_len) == ee_pblk) &&		/*C3*/
			(prev_len < (EXT_INIT_MAX_LEN - write_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, prev_ex);

			/* Shift the start of ex by 'write_len' blocks */
			ex->ee_block = cpu_to_le32(ee_block + write_len);
			ext4_ext_store_pblock(ex, ee_pblk + write_len);
			ex->ee_len = cpu_to_le16(ee_len - write_len);
			ext4_ext_mark_uninitialized(ex); /* Restore the flag */

			/* Extend prev_ex by 'write_len' blocks */
			prev_ex->ee_len = cpu_to_le16(prev_len + write_len);

			/* 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 = prev_ex;

			/* Result: number of initialized blocks past m_lblk */
			allocated = write_len;
			goto out;
		}
	}

3163
	WARN_ON(map->m_lblk < ee_block);
3164 3165 3166 3167
	/*
	 * It is safe to convert extent to initialized via explicit
	 * zeroout only if extent is fully insde i_size or new_size.
	 */
3168
	split_flag |= ee_block + ee_len <= eof_block ? EXT4_EXT_MAY_ZEROOUT : 0;
3169

3170
	/* If extent has less than 2*EXT4_EXT_ZERO_LEN zerout directly */
3171 3172 3173
	if (ee_len <= 2*EXT4_EXT_ZERO_LEN &&
	    (EXT4_EXT_MAY_ZEROOUT & split_flag)) {
		err = ext4_ext_zeroout(inode, ex);
3174
		if (err)
3175 3176 3177 3178 3179
			goto out;

		err = ext4_ext_get_access(handle, inode, path + depth);
		if (err)
			goto out;
3180 3181 3182 3183
		ext4_ext_mark_initialized(ex);
		ext4_ext_try_to_merge(inode, path, ex);
		err = ext4_ext_dirty(handle, inode, path + depth);
		goto out;
3184
	}
3185

3186
	/*
3187 3188 3189 3190 3191
	 * 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.
3192
	 */
3193 3194 3195 3196 3197 3198 3199 3200
	split_map.m_lblk = map->m_lblk;
	split_map.m_len = map->m_len;

	if (allocated > map->m_len) {
		if (allocated <= EXT4_EXT_ZERO_LEN &&
		    (EXT4_EXT_MAY_ZEROOUT & split_flag)) {
			/* case 3 */
			zero_ex.ee_block =
3201 3202
					 cpu_to_le32(map->m_lblk);
			zero_ex.ee_len = cpu_to_le16(allocated);
3203 3204 3205
			ext4_ext_store_pblock(&zero_ex,
				ext4_ext_pblock(ex) + map->m_lblk - ee_block);
			err = ext4_ext_zeroout(inode, &zero_ex);
3206 3207
			if (err)
				goto out;
3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225
			split_map.m_lblk = map->m_lblk;
			split_map.m_len = allocated;
		} else if ((map->m_lblk - ee_block + map->m_len <
			   EXT4_EXT_ZERO_LEN) &&
			   (EXT4_EXT_MAY_ZEROOUT & split_flag)) {
			/* 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;
3226 3227
			split_map.m_len = map->m_lblk - ee_block + map->m_len;
			allocated = map->m_len;
3228 3229
		}
	}
3230 3231 3232 3233 3234 3235

	allocated = ext4_split_extent(handle, inode, path,
				       &split_map, split_flag, 0);
	if (allocated < 0)
		err = allocated;

3236 3237 3238 3239
out:
	return err ? err : allocated;
}

3240
/*
3241
 * This function is called by ext4_ext_map_blocks() from
3242 3243 3244
 * ext4_get_blocks_dio_write() when DIO to write
 * to an uninitialized extent.
 *
P
Paul Bolle 已提交
3245
 * Writing to an uninitialized extent may result in splitting the uninitialized
3246
 * extent into multiple /initialized uninitialized extents (up to three)
3247 3248 3249 3250 3251 3252
 * 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
3253
 * the uninitialized extent split. To prevent ENOSPC occur at the IO
3254
 * complete, we need to split the uninitialized extent before DIO submit
3255
 * the IO. The uninitialized extent called at this time will be split
3256 3257 3258
 * 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().
3259 3260
 *
 * Returns the size of uninitialized extent to be written on success.
3261 3262 3263
 */
static int ext4_split_unwritten_extents(handle_t *handle,
					struct inode *inode,
3264
					struct ext4_map_blocks *map,
3265 3266 3267
					struct ext4_ext_path *path,
					int flags)
{
3268 3269 3270 3271 3272
	ext4_lblk_t eof_block;
	ext4_lblk_t ee_block;
	struct ext4_extent *ex;
	unsigned int ee_len;
	int split_flag = 0, depth;
3273 3274 3275

	ext_debug("ext4_split_unwritten_extents: inode %lu, logical"
		"block %llu, max_blocks %u\n", inode->i_ino,
3276
		(unsigned long long)map->m_lblk, map->m_len);
3277 3278 3279

	eof_block = (inode->i_size + inode->i_sb->s_blocksize - 1) >>
		inode->i_sb->s_blocksize_bits;
3280 3281
	if (eof_block < map->m_lblk + map->m_len)
		eof_block = map->m_lblk + map->m_len;
3282 3283 3284 3285
	/*
	 * It is safe to convert extent to initialized via explicit
	 * zeroout only if extent is fully insde i_size or new_size.
	 */
3286 3287 3288 3289
	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);
3290

3291 3292
	split_flag |= ee_block + ee_len <= eof_block ? EXT4_EXT_MAY_ZEROOUT : 0;
	split_flag |= EXT4_EXT_MARK_UNINIT2;
3293

3294 3295
	flags |= EXT4_GET_BLOCKS_PRE_IO;
	return ext4_split_extent(handle, inode, path, map, split_flag, flags);
3296
}
3297

3298
static int ext4_convert_unwritten_extents_endio(handle_t *handle,
3299 3300 3301 3302 3303 3304 3305 3306 3307 3308
					      struct inode *inode,
					      struct ext4_ext_path *path)
{
	struct ext4_extent *ex;
	int depth;
	int err = 0;

	depth = ext_depth(inode);
	ex = path[depth].p_ext;

3309 3310 3311 3312 3313
	ext_debug("ext4_convert_unwritten_extents_endio: inode %lu, logical"
		"block %llu, max_blocks %u\n", inode->i_ino,
		(unsigned long long)le32_to_cpu(ex->ee_block),
		ext4_ext_get_actual_len(ex));

3314 3315 3316 3317 3318 3319
	err = ext4_ext_get_access(handle, inode, path + depth);
	if (err)
		goto out;
	/* first mark the extent as initialized */
	ext4_ext_mark_initialized(ex);

3320 3321
	/* note: ext4_ext_correct_indexes() isn't needed here because
	 * borders are not changed
3322
	 */
3323 3324
	ext4_ext_try_to_merge(inode, path, ex);

3325 3326 3327 3328 3329 3330 3331
	/* Mark modified extent as dirty */
	err = ext4_ext_dirty(handle, inode, path + depth);
out:
	ext4_ext_show_leaf(inode, path);
	return err;
}

3332 3333 3334 3335 3336 3337 3338 3339
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);
}

T
Theodore Ts'o 已提交
3340 3341 3342 3343
/*
 * Handle EOFBLOCKS_FL flag, clearing it if necessary
 */
static int check_eofblocks_fl(handle_t *handle, struct inode *inode,
3344
			      ext4_lblk_t lblk,
T
Theodore Ts'o 已提交
3345 3346 3347 3348 3349
			      struct ext4_ext_path *path,
			      unsigned int len)
{
	int i, depth;
	struct ext4_extent_header *eh;
3350
	struct ext4_extent *last_ex;
T
Theodore Ts'o 已提交
3351 3352 3353 3354 3355 3356 3357

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

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

3358 3359 3360 3361 3362 3363 3364
	/*
	 * 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;
T
Theodore Ts'o 已提交
3365 3366 3367 3368 3369 3370 3371 3372 3373 3374
	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.
	 */
3375
	if (lblk + len < le32_to_cpu(last_ex->ee_block) +
T
Theodore Ts'o 已提交
3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387
	    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;
3388
out:
T
Theodore Ts'o 已提交
3389 3390 3391 3392
	ext4_clear_inode_flag(inode, EXT4_INODE_EOFBLOCKS);
	return ext4_mark_inode_dirty(handle, inode);
}

3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416
/**
 * ext4_find_delalloc_range: find delayed allocated block in the given range.
 *
 * Goes through the buffer heads in the range [lblk_start, lblk_end] and returns
 * whether there are any buffers marked for delayed allocation. It returns '1'
 * on the first delalloc'ed buffer head found. If no buffer head in the given
 * range is marked for delalloc, it returns 0.
 * lblk_start should always be <= lblk_end.
 * search_hint_reverse is to indicate that searching in reverse from lblk_end to
 * lblk_start might be more efficient (i.e., we will likely hit the delalloc'ed
 * block sooner). This is useful when blocks are truncated sequentially from
 * lblk_start towards lblk_end.
 */
static int ext4_find_delalloc_range(struct inode *inode,
				    ext4_lblk_t lblk_start,
				    ext4_lblk_t lblk_end,
				    int search_hint_reverse)
{
	struct address_space *mapping = inode->i_mapping;
	struct buffer_head *head, *bh = NULL;
	struct page *page;
	ext4_lblk_t i, pg_lblk;
	pgoff_t index;

3417 3418 3419
	if (!test_opt(inode->i_sb, DELALLOC))
		return 0;

3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432
	/* reverse search wont work if fs block size is less than page size */
	if (inode->i_blkbits < PAGE_CACHE_SHIFT)
		search_hint_reverse = 0;

	if (search_hint_reverse)
		i = lblk_end;
	else
		i = lblk_start;

	index = i >> (PAGE_CACHE_SHIFT - inode->i_blkbits);

	while ((i >= lblk_start) && (i <= lblk_end)) {
		page = find_get_page(mapping, index);
3433
		if (!page)
3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457
			goto nextpage;

		if (!page_has_buffers(page))
			goto nextpage;

		head = page_buffers(page);
		if (!head)
			goto nextpage;

		bh = head;
		pg_lblk = index << (PAGE_CACHE_SHIFT -
						inode->i_blkbits);
		do {
			if (unlikely(pg_lblk < lblk_start)) {
				/*
				 * This is possible when fs block size is less
				 * than page size and our cluster starts/ends in
				 * middle of the page. So we need to skip the
				 * initial few blocks till we reach the 'lblk'
				 */
				pg_lblk++;
				continue;
			}

3458 3459 3460 3461 3462
			/* Check if the buffer is delayed allocated and that it
			 * is not yet mapped. (when da-buffers are mapped during
			 * their writeout, their da_mapped bit is set.)
			 */
			if (buffer_delay(bh) && !buffer_da_mapped(bh)) {
3463
				page_cache_release(page);
3464 3465 3466 3467
				trace_ext4_find_delalloc_range(inode,
						lblk_start, lblk_end,
						search_hint_reverse,
						1, i);
3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489
				return 1;
			}
			if (search_hint_reverse)
				i--;
			else
				i++;
		} while ((i >= lblk_start) && (i <= lblk_end) &&
				((bh = bh->b_this_page) != head));
nextpage:
		if (page)
			page_cache_release(page);
		/*
		 * Move to next page. 'i' will be the first lblk in the next
		 * page.
		 */
		if (search_hint_reverse)
			index--;
		else
			index++;
		i = index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
	}

3490 3491
	trace_ext4_find_delalloc_range(inode, lblk_start, lblk_end,
					search_hint_reverse, 0, 0);
3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556
	return 0;
}

int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk,
			       int search_hint_reverse)
{
	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;

	return ext4_find_delalloc_range(inode, lblk_start, lblk_end,
					search_hint_reverse);
}

/**
 * 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;

3557 3558
	trace_ext4_get_reserved_cluster_alloc(inode, lblk_start, num_blks);

3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581
	/* 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;

		if (ext4_find_delalloc_range(inode, lblk_from, lblk_to, 0))
			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;

		if (ext4_find_delalloc_range(inode, lblk_from, lblk_to, 0))
			allocated_clusters--;
	}

	return allocated_clusters;
}

3582 3583
static int
ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode,
3584
			struct ext4_map_blocks *map,
3585
			struct ext4_ext_path *path, int flags,
3586
			unsigned int allocated, ext4_fsblk_t newblock)
3587 3588 3589
{
	int ret = 0;
	int err = 0;
3590
	ext4_io_end_t *io = EXT4_I(inode)->cur_aio_dio;
3591

3592 3593
	ext_debug("ext4_ext_handle_uninitialized_extents: inode %lu, logical "
		  "block %llu, max_blocks %u, flags %x, allocated %u\n",
3594
		  inode->i_ino, (unsigned long long)map->m_lblk, map->m_len,
3595 3596 3597
		  flags, allocated);
	ext4_ext_show_leaf(inode, path);

3598 3599 3600
	trace_ext4_ext_handle_uninitialized_extents(inode, map, allocated,
						    newblock);

3601
	/* get_block() before submit the IO, split the extent */
3602
	if ((flags & EXT4_GET_BLOCKS_PRE_IO)) {
3603 3604
		ret = ext4_split_unwritten_extents(handle, inode, map,
						   path, flags);
3605 3606
		/*
		 * Flag the inode(non aio case) or end_io struct (aio case)
L
Lucas De Marchi 已提交
3607
		 * that this IO needs to conversion to written when IO is
3608 3609
		 * completed
		 */
3610 3611 3612
		if (io)
			ext4_set_io_unwritten_flag(inode, io);
		else
3613
			ext4_set_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);
3614
		if (ext4_should_dioread_nolock(inode))
3615
			map->m_flags |= EXT4_MAP_UNINIT;
3616 3617
		goto out;
	}
3618
	/* IO end_io complete, convert the filled extent to written */
3619
	if ((flags & EXT4_GET_BLOCKS_CONVERT)) {
3620
		ret = ext4_convert_unwritten_extents_endio(handle, inode,
3621
							path);
T
Theodore Ts'o 已提交
3622
		if (ret >= 0) {
3623
			ext4_update_inode_fsync_trans(handle, inode, 1);
3624 3625
			err = check_eofblocks_fl(handle, inode, map->m_lblk,
						 path, map->m_len);
T
Theodore Ts'o 已提交
3626 3627
		} else
			err = ret;
3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646
		goto out2;
	}
	/* buffered IO case */
	/*
	 * repeat fallocate creation request
	 * we already have an unwritten extent
	 */
	if (flags & EXT4_GET_BLOCKS_UNINIT_EXT)
		goto map_out;

	/* 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.
		 */
3647
		map->m_flags |= EXT4_MAP_UNWRITTEN;
3648 3649 3650 3651
		goto out1;
	}

	/* buffered write, writepage time, convert*/
3652
	ret = ext4_ext_convert_to_initialized(handle, inode, map, path);
3653
	if (ret >= 0)
3654
		ext4_update_inode_fsync_trans(handle, inode, 1);
3655 3656 3657 3658 3659 3660
out:
	if (ret <= 0) {
		err = ret;
		goto out2;
	} else
		allocated = ret;
3661
	map->m_flags |= EXT4_MAP_NEW;
3662 3663 3664 3665 3666 3667 3668
	/*
	 * 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.
	 */
3669
	if (allocated > map->m_len) {
3670
		unmap_underlying_metadata_blocks(inode->i_sb->s_bdev,
3671 3672 3673
					newblock + map->m_len,
					allocated - map->m_len);
		allocated = map->m_len;
3674
	}
3675 3676 3677 3678 3679 3680 3681 3682

	/*
	 * 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
	 */
3683 3684 3685 3686 3687 3688 3689 3690 3691
	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);
	}
3692

3693
map_out:
3694
	map->m_flags |= EXT4_MAP_MAPPED;
3695 3696 3697 3698 3699 3700
	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;
	}
3701
out1:
3702 3703
	if (allocated > map->m_len)
		allocated = map->m_len;
3704
	ext4_ext_show_leaf(inode, path);
3705 3706
	map->m_pblk = newblock;
	map->m_len = allocated;
3707 3708 3709 3710 3711 3712 3713
out2:
	if (path) {
		ext4_ext_drop_refs(path);
		kfree(path);
	}
	return err ? err : allocated;
}
T
Theodore Ts'o 已提交
3714

3715 3716 3717 3718
/*
 * get_implied_cluster_alloc - check to see if the requested
 * allocation (in the map structure) overlaps with a cluster already
 * allocated in an extent.
3719
 *	@sb	The filesystem superblock structure
3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755
 *	@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().
 */
3756
static int get_implied_cluster_alloc(struct super_block *sb,
3757 3758 3759 3760
				     struct ext4_map_blocks *map,
				     struct ext4_extent *ex,
				     struct ext4_ext_path *path)
{
3761
	struct ext4_sb_info *sbi = EXT4_SB(sb);
3762 3763
	ext4_lblk_t c_offset = map->m_lblk & (sbi->s_cluster_ratio-1);
	ext4_lblk_t ex_cluster_start, ex_cluster_end;
3764
	ext4_lblk_t rr_cluster_start;
3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808
	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);
		}
3809 3810

		trace_ext4_get_implied_cluster_alloc_exit(sb, map, 1);
3811 3812
		return 1;
	}
3813 3814

	trace_ext4_get_implied_cluster_alloc_exit(sb, map, 0);
3815 3816 3817 3818
	return 0;
}


3819
/*
3820 3821 3822
 * Block allocation/map/preallocation routine for extents based files
 *
 *
3823
 * Need to be called with
3824 3825
 * 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)
3826 3827 3828 3829 3830 3831 3832 3833 3834 3835
 *
 * 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.
3836
 */
3837 3838
int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
			struct ext4_map_blocks *map, int flags)
A
Alex Tomas 已提交
3839 3840
{
	struct ext4_ext_path *path = NULL;
3841 3842
	struct ext4_extent newex, *ex, *ex2;
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
3843
	ext4_fsblk_t newblock = 0;
3844 3845
	int free_on_err = 0, err = 0, depth, ret;
	unsigned int allocated = 0, offset = 0;
3846
	unsigned int allocated_clusters = 0;
3847
	struct ext4_allocation_request ar;
3848
	ext4_io_end_t *io = EXT4_I(inode)->cur_aio_dio;
3849
	ext4_lblk_t cluster_offset;
A
Alex Tomas 已提交
3850

3851
	ext_debug("blocks %u/%u requested for inode %lu\n",
3852
		  map->m_lblk, map->m_len, inode->i_ino);
3853
	trace_ext4_ext_map_blocks_enter(inode, map->m_lblk, map->m_len, flags);
A
Alex Tomas 已提交
3854 3855

	/* check in cache */
3856
	if (ext4_ext_in_cache(inode, map->m_lblk, &newex)) {
3857
		if (!newex.ee_start_lo && !newex.ee_start_hi) {
3858 3859 3860 3861
			if ((sbi->s_cluster_ratio > 1) &&
			    ext4_find_delalloc_cluster(inode, map->m_lblk, 0))
				map->m_flags |= EXT4_MAP_FROM_CLUSTER;

3862
			if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) {
3863 3864 3865 3866
				/*
				 * block isn't allocated yet and
				 * user doesn't want to allocate it
				 */
A
Alex Tomas 已提交
3867 3868 3869
				goto out2;
			}
			/* we should allocate requested block */
3870
		} else {
A
Alex Tomas 已提交
3871
			/* block is already allocated */
3872 3873
			if (sbi->s_cluster_ratio > 1)
				map->m_flags |= EXT4_MAP_FROM_CLUSTER;
3874
			newblock = map->m_lblk
D
Dave Kleikamp 已提交
3875
				   - le32_to_cpu(newex.ee_block)
3876
				   + ext4_ext_pblock(&newex);
3877
			/* number of remaining blocks in the extent */
3878
			allocated = ext4_ext_get_actual_len(&newex) -
3879
				(map->m_lblk - le32_to_cpu(newex.ee_block));
A
Alex Tomas 已提交
3880 3881 3882 3883 3884
			goto out;
		}
	}

	/* find extent for this block */
3885
	path = ext4_ext_find_extent(inode, map->m_lblk, NULL);
A
Alex Tomas 已提交
3886 3887 3888 3889 3890 3891 3892 3893 3894
	if (IS_ERR(path)) {
		err = PTR_ERR(path);
		path = NULL;
		goto out2;
	}

	depth = ext_depth(inode);

	/*
3895 3896
	 * consistent leaf must not be empty;
	 * this situation is possible, though, _during_ tree modification;
A
Alex Tomas 已提交
3897 3898
	 * this is why assert can't be put in ext4_ext_find_extent()
	 */
3899 3900
	if (unlikely(path[depth].p_ext == NULL && depth != 0)) {
		EXT4_ERROR_INODE(inode, "bad extent address "
3901 3902 3903
				 "lblock: %lu, depth: %d pblock %lld",
				 (unsigned long) map->m_lblk, depth,
				 path[depth].p_block);
3904 3905 3906
		err = -EIO;
		goto out2;
	}
A
Alex Tomas 已提交
3907

3908 3909
	ex = path[depth].p_ext;
	if (ex) {
A
Aneesh Kumar K.V 已提交
3910
		ext4_lblk_t ee_block = le32_to_cpu(ex->ee_block);
3911
		ext4_fsblk_t ee_start = ext4_ext_pblock(ex);
A
Amit Arora 已提交
3912
		unsigned short ee_len;
3913 3914 3915

		/*
		 * Uninitialized extents are treated as holes, except that
3916
		 * we split out initialized portions during a write.
3917
		 */
A
Amit Arora 已提交
3918
		ee_len = ext4_ext_get_actual_len(ex);
3919 3920 3921

		trace_ext4_ext_show_extent(inode, ee_block, ee_start, ee_len);

3922
		/* if found extent covers block, simply return it */
3923 3924
		if (in_range(map->m_lblk, ee_block, ee_len)) {
			newblock = map->m_lblk - ee_block + ee_start;
3925
			/* number of remaining blocks in the extent */
3926 3927 3928
			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);
3929

3930
			/*
3931 3932
			 * Do not put uninitialized extent
			 * in the cache
3933
			 */
3934 3935 3936 3937
			if (!ext4_ext_is_uninitialized(ex)) {
				ext4_ext_put_in_cache(inode, ee_block,
					ee_len, ee_start);
				goto out;
3938
			}
3939 3940 3941 3942
			ret = ext4_ext_handle_uninitialized_extents(
				handle, inode, map, path, flags,
				allocated, newblock);
			return ret;
A
Alex Tomas 已提交
3943 3944 3945
		}
	}

3946 3947 3948 3949
	if ((sbi->s_cluster_ratio > 1) &&
	    ext4_find_delalloc_cluster(inode, map->m_lblk, 0))
		map->m_flags |= EXT4_MAP_FROM_CLUSTER;

A
Alex Tomas 已提交
3950
	/*
3951
	 * requested block isn't allocated yet;
A
Alex Tomas 已提交
3952 3953
	 * we couldn't try to create block if create flag is zero
	 */
3954
	if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) {
3955 3956 3957 3958
		/*
		 * put just found gap into cache to speed up
		 * subsequent requests
		 */
3959
		ext4_ext_put_gap_in_cache(inode, path, map->m_lblk);
A
Alex Tomas 已提交
3960 3961
		goto out2;
	}
3962

A
Alex Tomas 已提交
3963
	/*
3964
	 * Okay, we need to do block allocation.
A
Andrew Morton 已提交
3965
	 */
3966
	map->m_flags &= ~EXT4_MAP_FROM_CLUSTER;
3967 3968 3969 3970 3971 3972 3973 3974
	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 &&
3975
	    get_implied_cluster_alloc(inode->i_sb, map, ex, path)) {
3976 3977
		ar.len = allocated = map->m_len;
		newblock = map->m_pblk;
3978
		map->m_flags |= EXT4_MAP_FROM_CLUSTER;
3979 3980
		goto got_allocated_blocks;
	}
A
Alex Tomas 已提交
3981

3982
	/* find neighbour allocated blocks */
3983
	ar.lleft = map->m_lblk;
3984 3985 3986
	err = ext4_ext_search_left(inode, path, &ar.lleft, &ar.pleft);
	if (err)
		goto out2;
3987
	ar.lright = map->m_lblk;
3988 3989
	ex2 = NULL;
	err = ext4_ext_search_right(inode, path, &ar.lright, &ar.pright, &ex2);
3990 3991
	if (err)
		goto out2;
A
Amit Arora 已提交
3992

3993 3994 3995
	/* Check if the extent after searching to the right implies a
	 * cluster we can use. */
	if ((sbi->s_cluster_ratio > 1) && ex2 &&
3996
	    get_implied_cluster_alloc(inode->i_sb, map, ex2, path)) {
3997 3998
		ar.len = allocated = map->m_len;
		newblock = map->m_pblk;
3999
		map->m_flags |= EXT4_MAP_FROM_CLUSTER;
4000 4001 4002
		goto got_allocated_blocks;
	}

4003 4004 4005 4006 4007 4008
	/*
	 * 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.
	 */
4009
	if (map->m_len > EXT_INIT_MAX_LEN &&
4010
	    !(flags & EXT4_GET_BLOCKS_UNINIT_EXT))
4011 4012
		map->m_len = EXT_INIT_MAX_LEN;
	else if (map->m_len > EXT_UNINIT_MAX_LEN &&
4013
		 (flags & EXT4_GET_BLOCKS_UNINIT_EXT))
4014
		map->m_len = EXT_UNINIT_MAX_LEN;
4015

4016 4017
	/* Check if we can really insert (m_lblk)::(m_lblk + m_len) extent */
	newex.ee_len = cpu_to_le16(map->m_len);
4018
	err = ext4_ext_check_overlap(sbi, inode, &newex, path);
A
Amit Arora 已提交
4019
	if (err)
4020
		allocated = ext4_ext_get_actual_len(&newex);
A
Amit Arora 已提交
4021
	else
4022
		allocated = map->m_len;
4023 4024 4025

	/* allocate new block */
	ar.inode = inode;
4026 4027
	ar.goal = ext4_ext_find_goal(inode, path, map->m_lblk);
	ar.logical = map->m_lblk;
4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039
	/*
	 * 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;
4040 4041 4042 4043 4044
	if (S_ISREG(inode->i_mode))
		ar.flags = EXT4_MB_HINT_DATA;
	else
		/* disable in-core preallocation for non-regular files */
		ar.flags = 0;
4045 4046
	if (flags & EXT4_GET_BLOCKS_NO_NORMALIZE)
		ar.flags |= EXT4_MB_HINT_NOPREALLOC;
4047
	newblock = ext4_mb_new_blocks(handle, &ar, &err);
A
Alex Tomas 已提交
4048 4049
	if (!newblock)
		goto out2;
4050
	ext_debug("allocate new block: goal %llu, found %llu/%u\n",
4051
		  ar.goal, newblock, allocated);
4052
	free_on_err = 1;
4053
	allocated_clusters = ar.len;
4054 4055 4056
	ar.len = EXT4_C2B(sbi, ar.len) - offset;
	if (ar.len > allocated)
		ar.len = allocated;
A
Alex Tomas 已提交
4057

4058
got_allocated_blocks:
A
Alex Tomas 已提交
4059
	/* try to insert new extent into found leaf and return */
4060
	ext4_ext_store_pblock(&newex, newblock + offset);
4061
	newex.ee_len = cpu_to_le16(ar.len);
4062 4063
	/* Mark uninitialized */
	if (flags & EXT4_GET_BLOCKS_UNINIT_EXT){
A
Amit Arora 已提交
4064
		ext4_ext_mark_uninitialized(&newex);
4065
		/*
4066
		 * io_end structure was created for every IO write to an
L
Lucas De Marchi 已提交
4067
		 * uninitialized extent. To avoid unnecessary conversion,
4068
		 * here we flag the IO that really needs the conversion.
4069
		 * For non asycn direct IO case, flag the inode state
L
Lucas De Marchi 已提交
4070
		 * that we need to perform conversion when IO is done.
4071
		 */
4072
		if ((flags & EXT4_GET_BLOCKS_PRE_IO)) {
4073 4074 4075
			if (io)
				ext4_set_io_unwritten_flag(inode, io);
			else
4076 4077
				ext4_set_inode_state(inode,
						     EXT4_STATE_DIO_UNWRITTEN);
4078
		}
4079
		if (ext4_should_dioread_nolock(inode))
4080
			map->m_flags |= EXT4_MAP_UNINIT;
4081
	}
4082

4083 4084 4085 4086
	err = 0;
	if ((flags & EXT4_GET_BLOCKS_KEEP_SIZE) == 0)
		err = check_eofblocks_fl(handle, inode, map->m_lblk,
					 path, ar.len);
4087 4088 4089
	if (!err)
		err = ext4_ext_insert_extent(handle, inode, path,
					     &newex, flags);
4090
	if (err && free_on_err) {
4091 4092
		int fb_flags = flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE ?
			EXT4_FREE_BLOCKS_NO_QUOT_UPDATE : 0;
4093
		/* free data blocks we just allocated */
4094 4095
		/* not a good idea to call discard here directly,
		 * but otherwise we'd need to call it every free() */
4096
		ext4_discard_preallocations(inode);
4097
		ext4_free_blocks(handle, inode, NULL, ext4_ext_pblock(&newex),
4098
				 ext4_ext_get_actual_len(&newex), fb_flags);
A
Alex Tomas 已提交
4099
		goto out2;
4100
	}
A
Alex Tomas 已提交
4101 4102

	/* previous routine could use block we allocated */
4103
	newblock = ext4_ext_pblock(&newex);
4104
	allocated = ext4_ext_get_actual_len(&newex);
4105 4106 4107
	if (allocated > map->m_len)
		allocated = map->m_len;
	map->m_flags |= EXT4_MAP_NEW;
A
Alex Tomas 已提交
4108

4109 4110 4111 4112
	/*
	 * Update reserved blocks/metadata blocks after successful
	 * block allocation which had been deferred till now.
	 */
4113
	if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE) {
4114
		unsigned int reserved_clusters;
4115
		/*
4116
		 * Check how many clusters we had reserved this allocated range
4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137
		 */
		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);
			/* We will claim quota for all newly allocated blocks.*/
			ext4_da_update_reserve_space(inode, allocated_clusters,
							1);
			if (reserved_clusters < allocated_clusters) {
4138
				struct ext4_inode_info *ei = EXT4_I(inode);
4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180
				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.
				 */
4181 4182 4183 4184 4185
				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);
4186 4187 4188
			}
		}
	}
4189

4190 4191 4192 4193 4194
	/*
	 * Cache the extent and update transaction to commit on fdatasync only
	 * when it is _not_ an uninitialized extent.
	 */
	if ((flags & EXT4_GET_BLOCKS_UNINIT_EXT) == 0) {
4195
		ext4_ext_put_in_cache(inode, map->m_lblk, allocated, newblock);
4196 4197 4198
		ext4_update_inode_fsync_trans(handle, inode, 1);
	} else
		ext4_update_inode_fsync_trans(handle, inode, 0);
A
Alex Tomas 已提交
4199
out:
4200 4201
	if (allocated > map->m_len)
		allocated = map->m_len;
A
Alex Tomas 已提交
4202
	ext4_ext_show_leaf(inode, path);
4203 4204 4205
	map->m_flags |= EXT4_MAP_MAPPED;
	map->m_pblk = newblock;
	map->m_len = allocated;
A
Alex Tomas 已提交
4206 4207 4208 4209 4210
out2:
	if (path) {
		ext4_ext_drop_refs(path);
		kfree(path);
	}
4211

4212
	trace_ext4_ext_map_blocks_exit(inode, map->m_lblk,
4213
		newblock, map->m_len, err ? err : allocated);
4214

4215
	return err ? err : allocated;
A
Alex Tomas 已提交
4216 4217
}

4218
void ext4_ext_truncate(struct inode *inode)
A
Alex Tomas 已提交
4219 4220 4221
{
	struct address_space *mapping = inode->i_mapping;
	struct super_block *sb = inode->i_sb;
A
Aneesh Kumar K.V 已提交
4222
	ext4_lblk_t last_block;
A
Alex Tomas 已提交
4223
	handle_t *handle;
4224
	loff_t page_len;
A
Alex Tomas 已提交
4225 4226
	int err = 0;

4227 4228 4229 4230 4231 4232
	/*
	 * finish any pending end_io work so we won't run the risk of
	 * converting any truncated blocks to initialized later
	 */
	ext4_flush_completed_IO(inode);

A
Alex Tomas 已提交
4233 4234 4235
	/*
	 * probably first extent we're gonna free will be last in block
	 */
4236
	err = ext4_writepage_trans_blocks(inode);
A
Alex Tomas 已提交
4237
	handle = ext4_journal_start(inode, err);
4238
	if (IS_ERR(handle))
A
Alex Tomas 已提交
4239 4240
		return;

4241 4242 4243 4244 4245 4246 4247 4248 4249 4250
	if (inode->i_size % PAGE_CACHE_SIZE != 0) {
		page_len = PAGE_CACHE_SIZE -
			(inode->i_size & (PAGE_CACHE_SIZE - 1));

		err = ext4_discard_partial_page_buffers(handle,
			mapping, inode->i_size, page_len, 0);

		if (err)
			goto out_stop;
	}
A
Alex Tomas 已提交
4251

4252 4253 4254
	if (ext4_orphan_add(handle, inode))
		goto out_stop;

4255
	down_write(&EXT4_I(inode)->i_data_sem);
A
Alex Tomas 已提交
4256 4257
	ext4_ext_invalidate_cache(inode);

4258
	ext4_discard_preallocations(inode);
4259

A
Alex Tomas 已提交
4260
	/*
4261 4262 4263
	 * TODO: optimization is possible here.
	 * Probably we need not scan at all,
	 * because page truncation is enough.
A
Alex Tomas 已提交
4264 4265 4266 4267 4268 4269 4270 4271
	 */

	/* 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);
4272
	err = ext4_ext_remove_space(inode, last_block, EXT_MAX_BLOCKS - 1);
A
Alex Tomas 已提交
4273 4274

	/* In a multi-transaction truncate, we only make the final
4275 4276
	 * transaction synchronous.
	 */
A
Alex Tomas 已提交
4277
	if (IS_SYNC(inode))
4278
		ext4_handle_sync(handle);
A
Alex Tomas 已提交
4279

4280
	up_write(&EXT4_I(inode)->i_data_sem);
4281 4282

out_stop:
A
Alex Tomas 已提交
4283
	/*
4284
	 * If this was a simple ftruncate() and the file will remain alive,
A
Alex Tomas 已提交
4285 4286 4287 4288 4289 4290 4291 4292
	 * then we need to clear up the orphan record which we created above.
	 * However, if this was a real unlink then we were called by
	 * ext4_delete_inode(), and we allow that function to clean up the
	 * orphan info for us.
	 */
	if (inode->i_nlink)
		ext4_orphan_del(handle, inode);

4293 4294
	inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
	ext4_mark_inode_dirty(handle, inode);
A
Alex Tomas 已提交
4295 4296 4297
	ext4_journal_stop(handle);
}

4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311
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.
	 */
4312 4313 4314 4315 4316
	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);
4317 4318 4319 4320 4321 4322
	} 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))
4323
			ext4_set_inode_flag(inode, EXT4_INODE_EOFBLOCKS);
4324 4325 4326 4327
	}

}

A
Amit Arora 已提交
4328
/*
4329
 * preallocate space for a file. This implements ext4's fallocate file
A
Amit Arora 已提交
4330 4331 4332 4333 4334
 * 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).
 */
4335
long ext4_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
A
Amit Arora 已提交
4336
{
4337
	struct inode *inode = file->f_path.dentry->d_inode;
A
Amit Arora 已提交
4338
	handle_t *handle;
4339
	loff_t new_size;
4340
	unsigned int max_blocks;
A
Amit Arora 已提交
4341 4342 4343
	int ret = 0;
	int ret2 = 0;
	int retries = 0;
4344
	int flags;
4345
	struct ext4_map_blocks map;
A
Amit Arora 已提交
4346 4347 4348 4349 4350 4351
	unsigned int credits, blkbits = inode->i_blkbits;

	/*
	 * currently supporting (pre)allocate mode for extent-based
	 * files _only_
	 */
4352
	if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
A
Amit Arora 已提交
4353 4354
		return -EOPNOTSUPP;

4355 4356 4357 4358 4359 4360 4361
	/* 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);

4362
	trace_ext4_fallocate_enter(inode, offset, len, mode);
4363
	map.m_lblk = offset >> blkbits;
4364 4365 4366 4367
	/*
	 * We can't just convert len to max_blocks because
	 * If blocksize = 4096 offset = 3072 and len = 2048
	 */
A
Amit Arora 已提交
4368
	max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits)
4369
		- map.m_lblk;
A
Amit Arora 已提交
4370
	/*
4371
	 * credits to insert 1 extent into extent tree
A
Amit Arora 已提交
4372
	 */
4373
	credits = ext4_chunk_trans_blocks(inode, max_blocks);
4374
	mutex_lock(&inode->i_mutex);
4375 4376 4377
	ret = inode_newsize_ok(inode, (len + offset));
	if (ret) {
		mutex_unlock(&inode->i_mutex);
4378
		trace_ext4_fallocate_exit(inode, offset, max_blocks, ret);
4379 4380
		return ret;
	}
4381
	flags = EXT4_GET_BLOCKS_CREATE_UNINIT_EXT;
4382 4383
	if (mode & FALLOC_FL_KEEP_SIZE)
		flags |= EXT4_GET_BLOCKS_KEEP_SIZE;
4384 4385 4386 4387 4388 4389 4390
	/*
	 * 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;
A
Amit Arora 已提交
4391 4392
retry:
	while (ret >= 0 && ret < max_blocks) {
4393 4394
		map.m_lblk = map.m_lblk + ret;
		map.m_len = max_blocks = max_blocks - ret;
A
Amit Arora 已提交
4395 4396 4397 4398 4399
		handle = ext4_journal_start(inode, credits);
		if (IS_ERR(handle)) {
			ret = PTR_ERR(handle);
			break;
		}
4400
		ret = ext4_map_blocks(handle, inode, &map, flags);
4401
		if (ret <= 0) {
4402 4403
#ifdef EXT4FS_DEBUG
			WARN_ON(ret <= 0);
4404
			printk(KERN_ERR "%s: ext4_ext_map_blocks "
4405
				    "returned error inode#%lu, block=%u, "
4406
				    "max_blocks=%u", __func__,
4407
				    inode->i_ino, map.m_lblk, max_blocks);
4408
#endif
A
Amit Arora 已提交
4409 4410 4411 4412
			ext4_mark_inode_dirty(handle, inode);
			ret2 = ext4_journal_stop(handle);
			break;
		}
4413
		if ((map.m_lblk + ret) >= (EXT4_BLOCK_ALIGN(offset + len,
4414 4415 4416
						blkbits) >> blkbits))
			new_size = offset + len;
		else
4417
			new_size = ((loff_t) map.m_lblk + ret) << blkbits;
A
Amit Arora 已提交
4418

4419
		ext4_falloc_update_inode(inode, mode, new_size,
4420
					 (map.m_flags & EXT4_MAP_NEW));
A
Amit Arora 已提交
4421
		ext4_mark_inode_dirty(handle, inode);
4422 4423
		if ((file->f_flags & O_SYNC) && ret >= max_blocks)
			ext4_handle_sync(handle);
A
Amit Arora 已提交
4424 4425 4426 4427
		ret2 = ext4_journal_stop(handle);
		if (ret2)
			break;
	}
4428 4429 4430
	if (ret == -ENOSPC &&
			ext4_should_retry_alloc(inode->i_sb, &retries)) {
		ret = 0;
A
Amit Arora 已提交
4431 4432
		goto retry;
	}
4433
	mutex_unlock(&inode->i_mutex);
4434 4435
	trace_ext4_fallocate_exit(inode, offset, max_blocks,
				ret > 0 ? ret2 : ret);
A
Amit Arora 已提交
4436 4437
	return ret > 0 ? ret2 : ret;
}
4438

4439 4440 4441 4442 4443 4444 4445 4446
/*
 * 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.
4447
 * Returns 0 on success.
4448 4449
 */
int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
4450
				    ssize_t len)
4451 4452 4453 4454 4455
{
	handle_t *handle;
	unsigned int max_blocks;
	int ret = 0;
	int ret2 = 0;
4456
	struct ext4_map_blocks map;
4457 4458
	unsigned int credits, blkbits = inode->i_blkbits;

4459
	map.m_lblk = offset >> blkbits;
4460 4461 4462 4463
	/*
	 * We can't just convert len to max_blocks because
	 * If blocksize = 4096 offset = 3072 and len = 2048
	 */
4464 4465
	max_blocks = ((EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits) -
		      map.m_lblk);
4466 4467 4468 4469 4470
	/*
	 * credits to insert 1 extent into extent tree
	 */
	credits = ext4_chunk_trans_blocks(inode, max_blocks);
	while (ret >= 0 && ret < max_blocks) {
4471 4472
		map.m_lblk += ret;
		map.m_len = (max_blocks -= ret);
4473 4474 4475 4476 4477
		handle = ext4_journal_start(inode, credits);
		if (IS_ERR(handle)) {
			ret = PTR_ERR(handle);
			break;
		}
4478
		ret = ext4_map_blocks(handle, inode, &map,
4479
				      EXT4_GET_BLOCKS_IO_CONVERT_EXT);
4480 4481
		if (ret <= 0) {
			WARN_ON(ret <= 0);
4482 4483 4484 4485 4486
			ext4_msg(inode->i_sb, KERN_ERR,
				 "%s:%d: inode #%lu: block %u: len %u: "
				 "ext4_ext_map_blocks returned %d",
				 __func__, __LINE__, inode->i_ino, map.m_lblk,
				 map.m_len, ret);
4487 4488 4489 4490 4491 4492 4493 4494
		}
		ext4_mark_inode_dirty(handle, inode);
		ret2 = ext4_journal_stop(handle);
		if (ret <= 0 || ret2 )
			break;
	}
	return ret > 0 ? ret2 : ret;
}
4495

4496 4497 4498
/*
 * Callback function called for each extent to gather FIEMAP information.
 */
4499
static int ext4_ext_fiemap_cb(struct inode *inode, ext4_lblk_t next,
4500 4501 4502 4503 4504 4505 4506
		       struct ext4_ext_cache *newex, struct ext4_extent *ex,
		       void *data)
{
	__u64	logical;
	__u64	physical;
	__u64	length;
	__u32	flags = 0;
4507 4508 4509
	int		ret = 0;
	struct fiemap_extent_info *fieinfo = data;
	unsigned char blksize_bits;
4510

4511 4512
	blksize_bits = inode->i_sb->s_blocksize_bits;
	logical = (__u64)newex->ec_block << blksize_bits;
4513

4514
	if (newex->ec_start == 0) {
4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532
		/*
		 * No extent in extent-tree contains block @newex->ec_start,
		 * then the block may stay in 1)a hole or 2)delayed-extent.
		 *
		 * Holes or delayed-extents are processed as follows.
		 * 1. lookup dirty pages with specified range in pagecache.
		 *    If no page is got, then there is no delayed-extent and
		 *    return with EXT_CONTINUE.
		 * 2. find the 1st mapped buffer,
		 * 3. check if the mapped buffer is both in the request range
		 *    and a delayed buffer. If not, there is no delayed-extent,
		 *    then return.
		 * 4. a delayed-extent is found, the extent will be collected.
		 */
		ext4_lblk_t	end = 0;
		pgoff_t		last_offset;
		pgoff_t		offset;
		pgoff_t		index;
4533
		pgoff_t		start_index = 0;
4534
		struct page	**pages = NULL;
4535
		struct buffer_head *bh = NULL;
4536 4537 4538 4539 4540 4541
		struct buffer_head *head = NULL;
		unsigned int nr_pages = PAGE_SIZE / sizeof(struct page *);

		pages = kmalloc(PAGE_SIZE, GFP_KERNEL);
		if (pages == NULL)
			return -ENOMEM;
4542 4543

		offset = logical >> PAGE_SHIFT;
4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559
repeat:
		last_offset = offset;
		head = NULL;
		ret = find_get_pages_tag(inode->i_mapping, &offset,
					PAGECACHE_TAG_DIRTY, nr_pages, pages);

		if (!(flags & FIEMAP_EXTENT_DELALLOC)) {
			/* First time, try to find a mapped buffer. */
			if (ret == 0) {
out:
				for (index = 0; index < ret; index++)
					page_cache_release(pages[index]);
				/* just a hole. */
				kfree(pages);
				return EXT_CONTINUE;
			}
4560
			index = 0;
4561

4562
next_page:
4563
			/* Try to find the 1st mapped buffer. */
4564
			end = ((__u64)pages[index]->index << PAGE_SHIFT) >>
4565
				  blksize_bits;
4566
			if (!page_has_buffers(pages[index]))
4567
				goto out;
4568
			head = page_buffers(pages[index]);
4569 4570
			if (!head)
				goto out;
4571

4572
			index++;
4573 4574
			bh = head;
			do {
4575 4576 4577 4578 4579 4580 4581 4582 4583 4584
				if (end >= newex->ec_block +
					newex->ec_len)
					/* The buffer is out of
					 * the request range.
					 */
					goto out;

				if (buffer_mapped(bh) &&
				    end >= newex->ec_block) {
					start_index = index - 1;
4585 4586 4587
					/* get the 1st mapped buffer. */
					goto found_mapped_buffer;
				}
4588

4589 4590 4591
				bh = bh->b_this_page;
				end++;
			} while (bh != head);
4592

4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603
			/* No mapped buffer in the range found in this page,
			 * We need to look up next page.
			 */
			if (index >= ret) {
				/* There is no page left, but we need to limit
				 * newex->ec_len.
				 */
				newex->ec_len = end - newex->ec_block;
				goto out;
			}
			goto next_page;
4604
		} else {
4605 4606 4607 4608
			/*Find contiguous delayed buffers. */
			if (ret > 0 && pages[0]->index == last_offset)
				head = page_buffers(pages[0]);
			bh = head;
4609 4610
			index = 1;
			start_index = 0;
4611
		}
4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632

found_mapped_buffer:
		if (bh != NULL && buffer_delay(bh)) {
			/* 1st or contiguous delayed buffer found. */
			if (!(flags & FIEMAP_EXTENT_DELALLOC)) {
				/*
				 * 1st delayed buffer found, record
				 * the start of extent.
				 */
				flags |= FIEMAP_EXTENT_DELALLOC;
				newex->ec_block = end;
				logical = (__u64)end << blksize_bits;
			}
			/* Find contiguous delayed buffers. */
			do {
				if (!buffer_delay(bh))
					goto found_delayed_extent;
				bh = bh->b_this_page;
				end++;
			} while (bh != head);

4633
			for (; index < ret; index++) {
4634 4635 4636 4637 4638 4639 4640 4641 4642
				if (!page_has_buffers(pages[index])) {
					bh = NULL;
					break;
				}
				head = page_buffers(pages[index]);
				if (!head) {
					bh = NULL;
					break;
				}
4643

4644
				if (pages[index]->index !=
4645 4646
				    pages[start_index]->index + index
				    - start_index) {
4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673
					/* Blocks are not contiguous. */
					bh = NULL;
					break;
				}
				bh = head;
				do {
					if (!buffer_delay(bh))
						/* Delayed-extent ends. */
						goto found_delayed_extent;
					bh = bh->b_this_page;
					end++;
				} while (bh != head);
			}
		} else if (!(flags & FIEMAP_EXTENT_DELALLOC))
			/* a hole found. */
			goto out;

found_delayed_extent:
		newex->ec_len = min(end - newex->ec_block,
						(ext4_lblk_t)EXT_INIT_MAX_LEN);
		if (ret == nr_pages && bh != NULL &&
			newex->ec_len < EXT_INIT_MAX_LEN &&
			buffer_delay(bh)) {
			/* Have not collected an extent and continue. */
			for (index = 0; index < ret; index++)
				page_cache_release(pages[index]);
			goto repeat;
4674
		}
4675 4676 4677 4678

		for (index = 0; index < ret; index++)
			page_cache_release(pages[index]);
		kfree(pages);
4679 4680 4681 4682 4683 4684 4685 4686
	}

	physical = (__u64)newex->ec_start << blksize_bits;
	length =   (__u64)newex->ec_len << blksize_bits;

	if (ex && ext4_ext_is_uninitialized(ex))
		flags |= FIEMAP_EXTENT_UNWRITTEN;

4687
	if (next == EXT_MAX_BLOCKS)
4688 4689
		flags |= FIEMAP_EXTENT_LAST;

4690
	ret = fiemap_fill_next_extent(fieinfo, logical, physical,
4691
					length, flags);
4692 4693 4694
	if (ret < 0)
		return ret;
	if (ret == 1)
4695 4696 4697 4698 4699 4700
		return EXT_BREAK;
	return EXT_CONTINUE;
}
/* fiemap flags we can handle specified here */
#define EXT4_FIEMAP_FLAGS	(FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR)

A
Aneesh Kumar K.V 已提交
4701 4702
static int ext4_xattr_fiemap(struct inode *inode,
				struct fiemap_extent_info *fieinfo)
4703 4704 4705 4706 4707 4708 4709 4710
{
	__u64 physical = 0;
	__u64 length;
	__u32 flags = FIEMAP_EXTENT_LAST;
	int blockbits = inode->i_sb->s_blocksize_bits;
	int error = 0;

	/* in-inode? */
4711
	if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
4712 4713 4714 4715 4716 4717 4718 4719 4720 4721 4722 4723
		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;
4724
		brelse(iloc.bh);
4725 4726 4727 4728 4729 4730 4731 4732 4733 4734 4735
	} 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);
}

4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747 4748 4749 4750 4751
/*
 * ext4_ext_punch_hole
 *
 * Punches a hole of "length" bytes in a file starting
 * at byte "offset"
 *
 * @inode:  The inode of the file to punch a hole in
 * @offset: The starting byte offset of the hole
 * @length: The length of the hole
 *
 * Returns the number of blocks removed or negative on err
 */
int ext4_ext_punch_hole(struct file *file, loff_t offset, loff_t length)
{
	struct inode *inode = file->f_path.dentry->d_inode;
	struct super_block *sb = inode->i_sb;
4752
	ext4_lblk_t first_block, stop_block;
4753 4754
	struct address_space *mapping = inode->i_mapping;
	handle_t *handle;
4755 4756
	loff_t first_page, last_page, page_len;
	loff_t first_page_offset, last_page_offset;
4757
	int credits, err = 0;
4758

4759 4760 4761 4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772
	/* No need to punch hole beyond i_size */
	if (offset >= inode->i_size)
		return 0;

	/*
	 * If the hole extends beyond i_size, set the hole
	 * to end after the page that contains i_size
	 */
	if (offset + length > inode->i_size) {
		length = inode->i_size +
		   PAGE_CACHE_SIZE - (inode->i_size & (PAGE_CACHE_SIZE - 1)) -
		   offset;
	}

4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 4783 4784
	first_page = (offset + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
	last_page = (offset + length) >> PAGE_CACHE_SHIFT;

	first_page_offset = first_page << PAGE_CACHE_SHIFT;
	last_page_offset = last_page << PAGE_CACHE_SHIFT;

	/*
	 * Write out all dirty pages to avoid race conditions
	 * Then release them.
	 */
	if (mapping->nrpages && mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) {
		err = filemap_write_and_wait_range(mapping,
4785
			offset, offset + length - 1);
4786

4787 4788
		if (err)
			return err;
4789 4790 4791 4792
	}

	/* Now release the pages */
	if (last_page_offset > first_page_offset) {
4793 4794
		truncate_pagecache_range(inode, first_page_offset,
					 last_page_offset - 1);
4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809
	}

	/* finish any pending end_io work */
	ext4_flush_completed_IO(inode);

	credits = ext4_writepage_trans_blocks(inode);
	handle = ext4_journal_start(inode, credits);
	if (IS_ERR(handle))
		return PTR_ERR(handle);

	err = ext4_orphan_add(handle, inode);
	if (err)
		goto out;

	/*
4810 4811 4812 4813
	 * Now we need to zero out the non-page-aligned data in the
	 * pages at the start and tail of the hole, and unmap the buffer
	 * heads for the block aligned regions of the page that were
	 * completely zeroed.
4814
	 */
4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847
	if (first_page > last_page) {
		/*
		 * If the file space being truncated is contained within a page
		 * just zero out and unmap the middle of that page
		 */
		err = ext4_discard_partial_page_buffers(handle,
			mapping, offset, length, 0);

		if (err)
			goto out;
	} else {
		/*
		 * zero out and unmap the partial page that contains
		 * the start of the hole
		 */
		page_len  = first_page_offset - offset;
		if (page_len > 0) {
			err = ext4_discard_partial_page_buffers(handle, mapping,
						   offset, page_len, 0);
			if (err)
				goto out;
		}

		/*
		 * zero out and unmap the partial page that contains
		 * the end of the hole
		 */
		page_len = offset + length - last_page_offset;
		if (page_len > 0) {
			err = ext4_discard_partial_page_buffers(handle, mapping,
					last_page_offset, page_len, 0);
			if (err)
				goto out;
4848 4849 4850
		}
	}

4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869
	/*
	 * If i_size is contained in the last page, we need to
	 * unmap and zero the partial page after i_size
	 */
	if (inode->i_size >> PAGE_CACHE_SHIFT == last_page &&
	   inode->i_size % PAGE_CACHE_SIZE != 0) {

		page_len = PAGE_CACHE_SIZE -
			(inode->i_size & (PAGE_CACHE_SIZE - 1));

		if (page_len > 0) {
			err = ext4_discard_partial_page_buffers(handle,
			  mapping, inode->i_size, page_len, 0);

			if (err)
				goto out;
		}
	}

4870 4871 4872 4873
	first_block = (offset + sb->s_blocksize - 1) >>
		EXT4_BLOCK_SIZE_BITS(sb);
	stop_block = (offset + length) >> EXT4_BLOCK_SIZE_BITS(sb);

4874
	/* If there are no blocks to remove, return now */
4875
	if (first_block >= stop_block)
4876 4877 4878 4879 4880 4881
		goto out;

	down_write(&EXT4_I(inode)->i_data_sem);
	ext4_ext_invalidate_cache(inode);
	ext4_discard_preallocations(inode);

4882
	err = ext4_ext_remove_space(inode, first_block, stop_block - 1);
4883

4884 4885
	ext4_ext_invalidate_cache(inode);
	ext4_discard_preallocations(inode);
4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896 4897 4898

	if (IS_SYNC(inode))
		ext4_handle_sync(handle);

	up_write(&EXT4_I(inode)->i_data_sem);

out:
	ext4_orphan_del(handle, inode);
	inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
	ext4_mark_inode_dirty(handle, inode);
	ext4_journal_stop(handle);
	return err;
}
4899 4900 4901 4902 4903 4904 4905
int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
		__u64 start, __u64 len)
{
	ext4_lblk_t start_blk;
	int error = 0;

	/* fallback to generic here if not in extents fmt */
4906
	if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
4907 4908 4909 4910 4911 4912 4913 4914 4915
		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 {
4916 4917 4918
		ext4_lblk_t len_blks;
		__u64 last_blk;

4919
		start_blk = start >> inode->i_sb->s_blocksize_bits;
4920
		last_blk = (start + len - 1) >> inode->i_sb->s_blocksize_bits;
4921 4922
		if (last_blk >= EXT_MAX_BLOCKS)
			last_blk = EXT_MAX_BLOCKS-1;
4923
		len_blks = ((ext4_lblk_t) last_blk) - start_blk + 1;
4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934

		/*
		 * Walk the extent tree gathering extent information.
		 * ext4_ext_fiemap_cb will push extents back to user.
		 */
		error = ext4_ext_walk_space(inode, start_blk, len_blks,
					  ext4_ext_fiemap_cb, fieinfo);
	}

	return error;
}