inode.c 173.9 KB
Newer Older
1
/*
2
 *  linux/fs/ext4/inode.c
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
 *
 * Copyright (C) 1992, 1993, 1994, 1995
 * Remy Card (card@masi.ibp.fr)
 * Laboratoire MASI - Institut Blaise Pascal
 * Universite Pierre et Marie Curie (Paris VI)
 *
 *  from
 *
 *  linux/fs/minix/inode.c
 *
 *  Copyright (C) 1991, 1992  Linus Torvalds
 *
 *  Goal-directed block allocation by Stephen Tweedie
 *	(sct@redhat.com), 1993, 1998
 *  Big-endian to little-endian byte-swapping/bitmaps by
 *        David S. Miller (davem@caip.rutgers.edu), 1995
 *  64-bit file support on 64-bit platforms by Jakub Jelinek
 *	(jj@sunsite.ms.mff.cuni.cz)
 *
22
 *  Assorted race fixes, rewrite of ext4_get_block() by Al Viro, 2000
23 24 25 26 27
 */

#include <linux/module.h>
#include <linux/fs.h>
#include <linux/time.h>
28
#include <linux/jbd2.h>
29 30 31 32 33 34
#include <linux/highuid.h>
#include <linux/pagemap.h>
#include <linux/quotaops.h>
#include <linux/string.h>
#include <linux/buffer_head.h>
#include <linux/writeback.h>
35
#include <linux/pagevec.h>
36
#include <linux/mpage.h>
37
#include <linux/namei.h>
38 39
#include <linux/uio.h>
#include <linux/bio.h>
40
#include <linux/workqueue.h>
41
#include <linux/kernel.h>
42
#include <linux/slab.h>
43

44
#include "ext4_jbd2.h"
45 46
#include "xattr.h"
#include "acl.h"
47
#include "ext4_extents.h"
48

49 50
#include <trace/events/ext4.h>

51 52
#define MPAGE_DA_EXTENT_TAIL 0x01

53 54 55
static inline int ext4_begin_ordered_truncate(struct inode *inode,
					      loff_t new_size)
{
56 57 58 59
	return jbd2_journal_begin_ordered_truncate(
					EXT4_SB(inode->i_sb)->s_journal,
					&EXT4_I(inode)->jinode,
					new_size);
60 61
}

62 63
static void ext4_invalidatepage(struct page *page, unsigned long offset);

64 65 66
/*
 * Test whether an inode is a fast symlink.
 */
67
static int ext4_inode_is_fast_symlink(struct inode *inode)
68
{
69
	int ea_blocks = EXT4_I(inode)->i_file_acl ?
70 71 72 73 74 75 76 77 78 79 80
		(inode->i_sb->s_blocksize >> 9) : 0;

	return (S_ISLNK(inode->i_mode) && inode->i_blocks - ea_blocks == 0);
}

/*
 * Work out how many blocks we need to proceed with the next chunk of a
 * truncate transaction.
 */
static unsigned long blocks_for_truncate(struct inode *inode)
{
A
Aneesh Kumar K.V 已提交
81
	ext4_lblk_t needed;
82 83 84 85 86 87

	needed = inode->i_blocks >> (inode->i_sb->s_blocksize_bits - 9);

	/* Give ourselves just enough room to cope with inodes in which
	 * i_blocks is corrupt: we've seen disk corruptions in the past
	 * which resulted in random data in an inode which looked enough
88
	 * like a regular file for ext4 to try to delete it.  Things
89 90 91 92 93 94 95
	 * will go a bit crazy if that happens, but at least we should
	 * try not to panic the whole kernel. */
	if (needed < 2)
		needed = 2;

	/* But we need to bound the transaction so we don't overflow the
	 * journal. */
96 97
	if (needed > EXT4_MAX_TRANS_DATA)
		needed = EXT4_MAX_TRANS_DATA;
98

99
	return EXT4_DATA_TRANS_BLOCKS(inode->i_sb) + needed;
100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115
}

/*
 * Truncate transactions can be complex and absolutely huge.  So we need to
 * be able to restart the transaction at a conventient checkpoint to make
 * sure we don't overflow the journal.
 *
 * start_transaction gets us a new handle for a truncate transaction,
 * and extend_transaction tries to extend the existing one a bit.  If
 * extend fails, we need to propagate the failure up and restart the
 * transaction in the top-level truncate loop. --sct
 */
static handle_t *start_transaction(struct inode *inode)
{
	handle_t *result;

116
	result = ext4_journal_start(inode, blocks_for_truncate(inode));
117 118 119
	if (!IS_ERR(result))
		return result;

120
	ext4_std_error(inode->i_sb, PTR_ERR(result));
121 122 123 124 125 126 127 128 129 130 131
	return result;
}

/*
 * Try to extend this transaction for the purposes of truncation.
 *
 * Returns 0 if we managed to create more room.  If we can't create more
 * room, and the transaction must be restarted we return 1.
 */
static int try_to_extend_transaction(handle_t *handle, struct inode *inode)
{
132 133 134
	if (!ext4_handle_valid(handle))
		return 0;
	if (ext4_handle_has_enough_credits(handle, EXT4_RESERVE_TRANS_BLOCKS+1))
135
		return 0;
136
	if (!ext4_journal_extend(handle, blocks_for_truncate(inode)))
137 138 139 140 141 142 143 144 145
		return 0;
	return 1;
}

/*
 * Restart the transaction associated with *handle.  This does a commit,
 * so before we call here everything must be consistently dirtied against
 * this transaction.
 */
146
int ext4_truncate_restart_trans(handle_t *handle, struct inode *inode,
147
				 int nblocks)
148
{
149 150 151 152 153 154 155 156
	int ret;

	/*
	 * Drop i_data_sem to avoid deadlock with ext4_get_blocks At this
	 * moment, get_block can be called only for blocks inside i_size since
	 * page cache has been already dropped and writes are blocked by
	 * i_mutex. So we can safely drop the i_data_sem here.
	 */
157
	BUG_ON(EXT4_JOURNAL(inode) == NULL);
158
	jbd_debug(2, "restarting handle %p\n", handle);
159 160 161
	up_write(&EXT4_I(inode)->i_data_sem);
	ret = ext4_journal_restart(handle, blocks_for_truncate(inode));
	down_write(&EXT4_I(inode)->i_data_sem);
162
	ext4_discard_preallocations(inode);
163 164

	return ret;
165 166 167 168 169
}

/*
 * Called at the last iput() if i_nlink is zero.
 */
170
void ext4_delete_inode(struct inode *inode)
171 172
{
	handle_t *handle;
173
	int err;
174

175
	if (!is_bad_inode(inode))
176
		dquot_initialize(inode);
177

178 179
	if (ext4_should_order_data(inode))
		ext4_begin_ordered_truncate(inode, 0);
180 181 182 183 184
	truncate_inode_pages(&inode->i_data, 0);

	if (is_bad_inode(inode))
		goto no_delete;

185
	handle = ext4_journal_start(inode, blocks_for_truncate(inode)+3);
186
	if (IS_ERR(handle)) {
187
		ext4_std_error(inode->i_sb, PTR_ERR(handle));
188 189 190 191 192
		/*
		 * If we're going to skip the normal cleanup, we still need to
		 * make sure that the in-core orphan linked list is properly
		 * cleaned up.
		 */
193
		ext4_orphan_del(NULL, inode);
194 195 196 197
		goto no_delete;
	}

	if (IS_SYNC(inode))
198
		ext4_handle_sync(handle);
199
	inode->i_size = 0;
200 201
	err = ext4_mark_inode_dirty(handle, inode);
	if (err) {
202
		ext4_warning(inode->i_sb,
203 204 205
			     "couldn't mark inode dirty (err %d)", err);
		goto stop_handle;
	}
206
	if (inode->i_blocks)
207
		ext4_truncate(inode);
208 209 210 211 212 213 214

	/*
	 * ext4_ext_truncate() doesn't reserve any slop when it
	 * restarts journal transactions; therefore there may not be
	 * enough credits left in the handle to remove the inode from
	 * the orphan list and set the dtime field.
	 */
215
	if (!ext4_handle_has_enough_credits(handle, 3)) {
216 217 218 219
		err = ext4_journal_extend(handle, 3);
		if (err > 0)
			err = ext4_journal_restart(handle, 3);
		if (err != 0) {
220
			ext4_warning(inode->i_sb,
221 222 223 224 225 226 227
				     "couldn't extend journal (err %d)", err);
		stop_handle:
			ext4_journal_stop(handle);
			goto no_delete;
		}
	}

228
	/*
229
	 * Kill off the orphan record which ext4_truncate created.
230
	 * AKPM: I think this can be inside the above `if'.
231
	 * Note that ext4_orphan_del() has to be able to cope with the
232
	 * deletion of a non-existent orphan - this is because we don't
233
	 * know if ext4_truncate() actually created an orphan record.
234 235
	 * (Well, we could do this if we need to, but heck - it works)
	 */
236 237
	ext4_orphan_del(handle, inode);
	EXT4_I(inode)->i_dtime	= get_seconds();
238 239 240 241 242 243 244 245

	/*
	 * One subtle ordering requirement: if anything has gone wrong
	 * (transaction abort, IO errors, whatever), then we can still
	 * do these next steps (the fs will already have been marked as
	 * having errors), but we can't free the inode if the mark_dirty
	 * fails.
	 */
246
	if (ext4_mark_inode_dirty(handle, inode))
247 248 249
		/* If that failed, just do the required in-core inode clear. */
		clear_inode(inode);
	else
250 251
		ext4_free_inode(handle, inode);
	ext4_journal_stop(handle);
252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269
	return;
no_delete:
	clear_inode(inode);	/* We must guarantee clearing of inode... */
}

typedef struct {
	__le32	*p;
	__le32	key;
	struct buffer_head *bh;
} Indirect;

static inline void add_chain(Indirect *p, struct buffer_head *bh, __le32 *v)
{
	p->key = *(p->p = v);
	p->bh = bh;
}

/**
270
 *	ext4_block_to_path - parse the block number into array of offsets
271 272 273
 *	@inode: inode in question (we are only interested in its superblock)
 *	@i_block: block number to be parsed
 *	@offsets: array to store the offsets in
D
Dave Kleikamp 已提交
274 275
 *	@boundary: set this non-zero if the referred-to block is likely to be
 *	       followed (on disk) by an indirect block.
276
 *
277
 *	To store the locations of file's data ext4 uses a data structure common
278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299
 *	for UNIX filesystems - tree of pointers anchored in the inode, with
 *	data blocks at leaves and indirect blocks in intermediate nodes.
 *	This function translates the block number into path in that tree -
 *	return value is the path length and @offsets[n] is the offset of
 *	pointer to (n+1)th node in the nth one. If @block is out of range
 *	(negative or too large) warning is printed and zero returned.
 *
 *	Note: function doesn't find node addresses, so no IO is needed. All
 *	we need to know is the capacity of indirect blocks (taken from the
 *	inode->i_sb).
 */

/*
 * Portability note: the last comparison (check that we fit into triple
 * indirect block) is spelled differently, because otherwise on an
 * architecture with 32-bit longs and 8Kb pages we might get into trouble
 * if our filesystem had 8Kb blocks. We might use long long, but that would
 * kill us on x86. Oh, well, at least the sign propagation does not matter -
 * i_block would have to be negative in the very beginning, so we would not
 * get there at all.
 */

300
static int ext4_block_to_path(struct inode *inode,
301 302
			      ext4_lblk_t i_block,
			      ext4_lblk_t offsets[4], int *boundary)
303
{
304 305 306
	int ptrs = EXT4_ADDR_PER_BLOCK(inode->i_sb);
	int ptrs_bits = EXT4_ADDR_PER_BLOCK_BITS(inode->i_sb);
	const long direct_blocks = EXT4_NDIR_BLOCKS,
307 308 309 310 311
		indirect_blocks = ptrs,
		double_blocks = (1 << (ptrs_bits * 2));
	int n = 0;
	int final = 0;

312
	if (i_block < direct_blocks) {
313 314
		offsets[n++] = i_block;
		final = direct_blocks;
315
	} else if ((i_block -= direct_blocks) < indirect_blocks) {
316
		offsets[n++] = EXT4_IND_BLOCK;
317 318 319
		offsets[n++] = i_block;
		final = ptrs;
	} else if ((i_block -= indirect_blocks) < double_blocks) {
320
		offsets[n++] = EXT4_DIND_BLOCK;
321 322 323 324
		offsets[n++] = i_block >> ptrs_bits;
		offsets[n++] = i_block & (ptrs - 1);
		final = ptrs;
	} else if (((i_block -= double_blocks) >> (ptrs_bits * 2)) < ptrs) {
325
		offsets[n++] = EXT4_TIND_BLOCK;
326 327 328 329 330
		offsets[n++] = i_block >> (ptrs_bits * 2);
		offsets[n++] = (i_block >> ptrs_bits) & (ptrs - 1);
		offsets[n++] = i_block & (ptrs - 1);
		final = ptrs;
	} else {
331
		ext4_warning(inode->i_sb, "block %lu > max in inode %lu",
332 333
			     i_block + direct_blocks +
			     indirect_blocks + double_blocks, inode->i_ino);
334 335 336 337 338 339
	}
	if (boundary)
		*boundary = final - 1 - (i_block & (ptrs - 1));
	return n;
}

340
static int __ext4_check_blockref(const char *function, struct inode *inode,
341 342
				 __le32 *p, unsigned int max)
{
343
	__le32 *bref = p;
344 345
	unsigned int blk;

346
	while (bref < p+max) {
347
		blk = le32_to_cpu(*bref++);
348 349
		if (blk &&
		    unlikely(!ext4_data_block_valid(EXT4_SB(inode->i_sb),
350
						    blk, 1))) {
351
			__ext4_error(inode->i_sb, function,
352 353
				   "invalid block reference %u "
				   "in inode #%lu", blk, inode->i_ino);
354 355 356 357
			return -EIO;
		}
	}
	return 0;
358 359 360 361
}


#define ext4_check_indirect_blockref(inode, bh)                         \
362
	__ext4_check_blockref(__func__, inode, (__le32 *)(bh)->b_data,  \
363 364 365
			      EXT4_ADDR_PER_BLOCK((inode)->i_sb))

#define ext4_check_inode_blockref(inode)                                \
366
	__ext4_check_blockref(__func__, inode, EXT4_I(inode)->i_data,   \
367 368
			      EXT4_NDIR_BLOCKS)

369
/**
370
 *	ext4_get_branch - read the chain of indirect blocks leading to data
371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394
 *	@inode: inode in question
 *	@depth: depth of the chain (1 - direct pointer, etc.)
 *	@offsets: offsets of pointers in inode/indirect blocks
 *	@chain: place to store the result
 *	@err: here we store the error value
 *
 *	Function fills the array of triples <key, p, bh> and returns %NULL
 *	if everything went OK or the pointer to the last filled triple
 *	(incomplete one) otherwise. Upon the return chain[i].key contains
 *	the number of (i+1)-th block in the chain (as it is stored in memory,
 *	i.e. little-endian 32-bit), chain[i].p contains the address of that
 *	number (it points into struct inode for i==0 and into the bh->b_data
 *	for i>0) and chain[i].bh points to the buffer_head of i-th indirect
 *	block for i>0 and NULL for i==0. In other words, it holds the block
 *	numbers of the chain, addresses they were taken from (and where we can
 *	verify that chain did not change) and buffer_heads hosting these
 *	numbers.
 *
 *	Function stops when it stumbles upon zero pointer (absent block)
 *		(pointer to last triple returned, *@err == 0)
 *	or when it gets an IO error reading an indirect block
 *		(ditto, *@err == -EIO)
 *	or when it reads all @depth-1 indirect blocks successfully and finds
 *	the whole chain, all way to the data (returns %NULL, *err == 0).
395 396
 *
 *      Need to be called with
397
 *      down_read(&EXT4_I(inode)->i_data_sem)
398
 */
A
Aneesh Kumar K.V 已提交
399 400
static Indirect *ext4_get_branch(struct inode *inode, int depth,
				 ext4_lblk_t  *offsets,
401 402 403 404 405 406 407 408
				 Indirect chain[4], int *err)
{
	struct super_block *sb = inode->i_sb;
	Indirect *p = chain;
	struct buffer_head *bh;

	*err = 0;
	/* i_data is not going away, no lock needed */
409
	add_chain(chain, NULL, EXT4_I(inode)->i_data + *offsets);
410 411 412
	if (!p->key)
		goto no_block;
	while (--depth) {
413 414
		bh = sb_getblk(sb, le32_to_cpu(p->key));
		if (unlikely(!bh))
415
			goto failure;
416

417 418 419 420 421 422 423 424 425 426 427
		if (!bh_uptodate_or_lock(bh)) {
			if (bh_submit_read(bh) < 0) {
				put_bh(bh);
				goto failure;
			}
			/* validate block references */
			if (ext4_check_indirect_blockref(inode, bh)) {
				put_bh(bh);
				goto failure;
			}
		}
428

429
		add_chain(++p, bh, (__le32 *)bh->b_data + *++offsets);
430 431 432 433 434 435 436 437 438 439 440 441 442
		/* Reader: end */
		if (!p->key)
			goto no_block;
	}
	return NULL;

failure:
	*err = -EIO;
no_block:
	return p;
}

/**
443
 *	ext4_find_near - find a place for allocation with sufficient locality
444 445 446
 *	@inode: owner
 *	@ind: descriptor of indirect block.
 *
447
 *	This function returns the preferred place for block allocation.
448 449 450 451 452 453 454 455 456 457 458 459 460 461
 *	It is used when heuristic for sequential allocation fails.
 *	Rules are:
 *	  + if there is a block to the left of our position - allocate near it.
 *	  + if pointer will live in indirect block - allocate near that block.
 *	  + if pointer will live in inode - allocate in the same
 *	    cylinder group.
 *
 * In the latter case we colour the starting block by the callers PID to
 * prevent it from clashing with concurrent allocations for a different inode
 * in the same block group.   The PID is used here so that functionally related
 * files will be close-by on-disk.
 *
 *	Caller must make sure that @ind is valid and will stay that way.
 */
462
static ext4_fsblk_t ext4_find_near(struct inode *inode, Indirect *ind)
463
{
464
	struct ext4_inode_info *ei = EXT4_I(inode);
465
	__le32 *start = ind->bh ? (__le32 *) ind->bh->b_data : ei->i_data;
466
	__le32 *p;
467
	ext4_fsblk_t bg_start;
468
	ext4_fsblk_t last_block;
469
	ext4_grpblk_t colour;
470 471
	ext4_group_t block_group;
	int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb));
472 473 474 475 476 477 478 479 480 481 482 483 484 485 486

	/* Try to find previous block */
	for (p = ind->p - 1; p >= start; p--) {
		if (*p)
			return le32_to_cpu(*p);
	}

	/* No such thing, so let's try location of indirect block */
	if (ind->bh)
		return ind->bh->b_blocknr;

	/*
	 * It is going to be referred to from the inode itself? OK, just put it
	 * into the same cylinder group then.
	 */
487 488 489 490 491 492 493
	block_group = ei->i_block_group;
	if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) {
		block_group &= ~(flex_size-1);
		if (S_ISREG(inode->i_mode))
			block_group++;
	}
	bg_start = ext4_group_first_block_no(inode->i_sb, block_group);
494 495
	last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1;

496 497 498 499 500 501 502
	/*
	 * If we are doing delayed allocation, we don't need take
	 * colour into account.
	 */
	if (test_opt(inode->i_sb, DELALLOC))
		return bg_start;

503 504
	if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block)
		colour = (current->pid % 16) *
505
			(EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16);
506 507
	else
		colour = (current->pid % 16) * ((last_block - bg_start) / 16);
508 509 510 511
	return bg_start + colour;
}

/**
512
 *	ext4_find_goal - find a preferred place for allocation.
513 514 515 516
 *	@inode: owner
 *	@block:  block we want
 *	@partial: pointer to the last triple within a chain
 *
517
 *	Normally this function find the preferred place for block allocation,
518
 *	returns it.
519 520
 *	Because this is only used for non-extent files, we limit the block nr
 *	to 32 bits.
521
 */
A
Aneesh Kumar K.V 已提交
522
static ext4_fsblk_t ext4_find_goal(struct inode *inode, ext4_lblk_t block,
523
				   Indirect *partial)
524
{
525 526
	ext4_fsblk_t goal;

527
	/*
528
	 * XXX need to get goal block from mballoc's data structures
529 530
	 */

531 532 533
	goal = ext4_find_near(inode, partial);
	goal = goal & EXT4_MAX_BLOCK_FILE_PHYS;
	return goal;
534 535 536
}

/**
537
 *	ext4_blks_to_allocate: Look up the block map and count the number
538 539 540 541 542 543 544 545 546 547
 *	of direct blocks need to be allocated for the given branch.
 *
 *	@branch: chain of indirect blocks
 *	@k: number of blocks need for indirect blocks
 *	@blks: number of data blocks to be mapped.
 *	@blocks_to_boundary:  the offset in the indirect block
 *
 *	return the total number of blocks to be allocate, including the
 *	direct and indirect blocks.
 */
548
static int ext4_blks_to_allocate(Indirect *branch, int k, unsigned int blks,
549
				 int blocks_to_boundary)
550
{
551
	unsigned int count = 0;
552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574

	/*
	 * Simple case, [t,d]Indirect block(s) has not allocated yet
	 * then it's clear blocks on that path have not allocated
	 */
	if (k > 0) {
		/* right now we don't handle cross boundary allocation */
		if (blks < blocks_to_boundary + 1)
			count += blks;
		else
			count += blocks_to_boundary + 1;
		return count;
	}

	count++;
	while (count < blks && count <= blocks_to_boundary &&
		le32_to_cpu(*(branch[0].p + count)) == 0) {
		count++;
	}
	return count;
}

/**
575
 *	ext4_alloc_blocks: multiple allocate blocks needed for a branch
576 577 578 579 580 581 582 583
 *	@indirect_blks: the number of blocks need to allocate for indirect
 *			blocks
 *
 *	@new_blocks: on return it will store the new block numbers for
 *	the indirect blocks(if needed) and the first direct block,
 *	@blks:	on return it will store the total number of allocated
 *		direct blocks
 */
584
static int ext4_alloc_blocks(handle_t *handle, struct inode *inode,
585 586 587
			     ext4_lblk_t iblock, ext4_fsblk_t goal,
			     int indirect_blks, int blks,
			     ext4_fsblk_t new_blocks[4], int *err)
588
{
589
	struct ext4_allocation_request ar;
590
	int target, i;
591
	unsigned long count = 0, blk_allocated = 0;
592
	int index = 0;
593
	ext4_fsblk_t current_block = 0;
594 595 596 597 598 599 600 601 602 603
	int ret = 0;

	/*
	 * Here we try to allocate the requested multiple blocks at once,
	 * on a best-effort basis.
	 * To build a branch, we should allocate blocks for
	 * the indirect blocks(if not allocated yet), and at least
	 * the first direct block of this branch.  That's the
	 * minimum number of blocks need to allocate(required)
	 */
604 605 606
	/* first we try to allocate the indirect blocks */
	target = indirect_blks;
	while (target > 0) {
607 608
		count = target;
		/* allocating blocks for indirect blocks and direct blocks */
609 610
		current_block = ext4_new_meta_blocks(handle, inode,
							goal, &count, err);
611 612 613
		if (*err)
			goto failed_out;

614 615 616 617 618 619 620 621
		if (unlikely(current_block + count > EXT4_MAX_BLOCK_FILE_PHYS)) {
			EXT4_ERROR_INODE(inode,
					 "current_block %llu + count %lu > %d!",
					 current_block, count,
					 EXT4_MAX_BLOCK_FILE_PHYS);
			*err = -EIO;
			goto failed_out;
		}
622

623 624 625 626 627 628
		target -= count;
		/* allocate blocks for indirect blocks */
		while (index < indirect_blks && count) {
			new_blocks[index++] = current_block++;
			count--;
		}
629 630 631 632 633 634 635 636 637
		if (count > 0) {
			/*
			 * save the new block number
			 * for the first direct block
			 */
			new_blocks[index] = current_block;
			printk(KERN_INFO "%s returned more blocks than "
						"requested\n", __func__);
			WARN_ON(1);
638
			break;
639
		}
640 641
	}

642 643 644 645 646
	target = blks - count ;
	blk_allocated = count;
	if (!target)
		goto allocated;
	/* Now allocate data blocks */
647 648 649 650 651 652 653 654 655 656
	memset(&ar, 0, sizeof(ar));
	ar.inode = inode;
	ar.goal = goal;
	ar.len = target;
	ar.logical = iblock;
	if (S_ISREG(inode->i_mode))
		/* enable in-core preallocation only for regular files */
		ar.flags = EXT4_MB_HINT_DATA;

	current_block = ext4_mb_new_blocks(handle, &ar, err);
657 658 659 660 661 662 663 664
	if (unlikely(current_block + ar.len > EXT4_MAX_BLOCK_FILE_PHYS)) {
		EXT4_ERROR_INODE(inode,
				 "current_block %llu + ar.len %d > %d!",
				 current_block, ar.len,
				 EXT4_MAX_BLOCK_FILE_PHYS);
		*err = -EIO;
		goto failed_out;
	}
665

666 667 668 669 670 671 672 673 674
	if (*err && (target == blks)) {
		/*
		 * if the allocation failed and we didn't allocate
		 * any blocks before
		 */
		goto failed_out;
	}
	if (!*err) {
		if (target == blks) {
675 676 677 678
			/*
			 * save the new block number
			 * for the first direct block
			 */
679 680
			new_blocks[index] = current_block;
		}
681
		blk_allocated += ar.len;
682 683
	}
allocated:
684
	/* total number of blocks allocated for direct blocks */
685
	ret = blk_allocated;
686 687 688
	*err = 0;
	return ret;
failed_out:
689
	for (i = 0; i < index; i++)
690
		ext4_free_blocks(handle, inode, 0, new_blocks[i], 1, 0);
691 692 693 694
	return ret;
}

/**
695
 *	ext4_alloc_branch - allocate and set up a chain of blocks.
696 697 698 699 700 701 702 703 704 705
 *	@inode: owner
 *	@indirect_blks: number of allocated indirect blocks
 *	@blks: number of allocated direct blocks
 *	@offsets: offsets (in the blocks) to store the pointers to next.
 *	@branch: place to store the chain in.
 *
 *	This function allocates blocks, zeroes out all but the last one,
 *	links them into chain and (if we are synchronous) writes them to disk.
 *	In other words, it prepares a branch that can be spliced onto the
 *	inode. It stores the information about that chain in the branch[], in
706
 *	the same format as ext4_get_branch() would do. We are calling it after
707 708
 *	we had read the existing part of chain and partial points to the last
 *	triple of that (one with zero ->key). Upon the exit we have the same
709
 *	picture as after the successful ext4_get_block(), except that in one
710 711 712 713 714 715
 *	place chain is disconnected - *branch->p is still zero (we did not
 *	set the last link), but branch->key contains the number that should
 *	be placed into *branch->p to fill that gap.
 *
 *	If allocation fails we free all blocks we've allocated (and forget
 *	their buffer_heads) and return the error value the from failed
716
 *	ext4_alloc_block() (normally -ENOSPC). Otherwise we set the chain
717 718
 *	as described above and return 0.
 */
719
static int ext4_alloc_branch(handle_t *handle, struct inode *inode,
720 721 722
			     ext4_lblk_t iblock, int indirect_blks,
			     int *blks, ext4_fsblk_t goal,
			     ext4_lblk_t *offsets, Indirect *branch)
723 724 725 726 727 728
{
	int blocksize = inode->i_sb->s_blocksize;
	int i, n = 0;
	int err = 0;
	struct buffer_head *bh;
	int num;
729 730
	ext4_fsblk_t new_blocks[4];
	ext4_fsblk_t current_block;
731

732
	num = ext4_alloc_blocks(handle, inode, iblock, goal, indirect_blks,
733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750
				*blks, new_blocks, &err);
	if (err)
		return err;

	branch[0].key = cpu_to_le32(new_blocks[0]);
	/*
	 * metadata blocks and data blocks are allocated.
	 */
	for (n = 1; n <= indirect_blks;  n++) {
		/*
		 * Get buffer_head for parent block, zero it out
		 * and set the pointer to new one, then send
		 * parent to disk.
		 */
		bh = sb_getblk(inode->i_sb, new_blocks[n-1]);
		branch[n].bh = bh;
		lock_buffer(bh);
		BUFFER_TRACE(bh, "call get_create_access");
751
		err = ext4_journal_get_create_access(handle, bh);
752
		if (err) {
753 754
			/* Don't brelse(bh) here; it's done in
			 * ext4_journal_forget() below */
755 756 757 758 759 760 761 762
			unlock_buffer(bh);
			goto failed;
		}

		memset(bh->b_data, 0, blocksize);
		branch[n].p = (__le32 *) bh->b_data + offsets[n];
		branch[n].key = cpu_to_le32(new_blocks[n]);
		*branch[n].p = branch[n].key;
763
		if (n == indirect_blks) {
764 765 766 767 768 769
			current_block = new_blocks[n];
			/*
			 * End of chain, update the last new metablock of
			 * the chain to point to the new allocated
			 * data blocks numbers
			 */
770
			for (i = 1; i < num; i++)
771 772 773 774 775 776
				*(branch[n].p + i) = cpu_to_le32(++current_block);
		}
		BUFFER_TRACE(bh, "marking uptodate");
		set_buffer_uptodate(bh);
		unlock_buffer(bh);

777 778
		BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
		err = ext4_handle_dirty_metadata(handle, inode, bh);
779 780 781 782 783 784 785
		if (err)
			goto failed;
	}
	*blks = num;
	return err;
failed:
	/* Allocation failed, free what we already allocated */
786
	ext4_free_blocks(handle, inode, 0, new_blocks[0], 1, 0);
787
	for (i = 1; i <= n ; i++) {
788
		/* 
789 790 791
		 * branch[i].bh is newly allocated, so there is no
		 * need to revoke the block, which is why we don't
		 * need to set EXT4_FREE_BLOCKS_METADATA.
792
		 */
793 794
		ext4_free_blocks(handle, inode, 0, new_blocks[i], 1,
				 EXT4_FREE_BLOCKS_FORGET);
795
	}
796 797
	for (i = n+1; i < indirect_blks; i++)
		ext4_free_blocks(handle, inode, 0, new_blocks[i], 1, 0);
798

799
	ext4_free_blocks(handle, inode, 0, new_blocks[i], num, 0);
800 801 802 803 804

	return err;
}

/**
805
 * ext4_splice_branch - splice the allocated branch onto inode.
806 807 808
 * @inode: owner
 * @block: (logical) number of block we are adding
 * @chain: chain of indirect blocks (with a missing link - see
809
 *	ext4_alloc_branch)
810 811 812 813 814 815 816 817
 * @where: location of missing link
 * @num:   number of indirect blocks we are adding
 * @blks:  number of direct blocks we are adding
 *
 * This function fills the missing link and does all housekeeping needed in
 * inode (->i_blocks, etc.). In case of success we end up with the full
 * chain to new block and return 0.
 */
818
static int ext4_splice_branch(handle_t *handle, struct inode *inode,
819 820
			      ext4_lblk_t block, Indirect *where, int num,
			      int blks)
821 822 823
{
	int i;
	int err = 0;
824
	ext4_fsblk_t current_block;
825 826 827 828 829 830 831 832

	/*
	 * If we're splicing into a [td]indirect block (as opposed to the
	 * inode) then we need to get write access to the [td]indirect block
	 * before the splice.
	 */
	if (where->bh) {
		BUFFER_TRACE(where->bh, "get_write_access");
833
		err = ext4_journal_get_write_access(handle, where->bh);
834 835 836 837 838 839 840 841 842 843 844 845 846 847
		if (err)
			goto err_out;
	}
	/* That's it */

	*where->p = where->key;

	/*
	 * Update the host buffer_head or inode to point to more just allocated
	 * direct blocks blocks
	 */
	if (num == 0 && blks > 1) {
		current_block = le32_to_cpu(where->key) + 1;
		for (i = 1; i < blks; i++)
848
			*(where->p + i) = cpu_to_le32(current_block++);
849 850 851 852 853 854 855 856 857 858 859
	}

	/* We are done with atomic stuff, now do the rest of housekeeping */
	/* had we spliced it onto indirect block? */
	if (where->bh) {
		/*
		 * If we spliced it onto an indirect block, we haven't
		 * altered the inode.  Note however that if it is being spliced
		 * onto an indirect block at the very end of the file (the
		 * file is growing) then we *will* alter the inode to reflect
		 * the new i_size.  But that is not done here - it is done in
860
		 * generic_commit_write->__mark_inode_dirty->ext4_dirty_inode.
861 862
		 */
		jbd_debug(5, "splicing indirect only\n");
863 864
		BUFFER_TRACE(where->bh, "call ext4_handle_dirty_metadata");
		err = ext4_handle_dirty_metadata(handle, inode, where->bh);
865 866 867 868 869 870
		if (err)
			goto err_out;
	} else {
		/*
		 * OK, we spliced it into the inode itself on a direct block.
		 */
871
		ext4_mark_inode_dirty(handle, inode);
872 873 874 875 876 877
		jbd_debug(5, "splicing direct\n");
	}
	return err;

err_out:
	for (i = 1; i <= num; i++) {
878
		/* 
879 880 881
		 * branch[i].bh is newly allocated, so there is no
		 * need to revoke the block, which is why we don't
		 * need to set EXT4_FREE_BLOCKS_METADATA.
882
		 */
883 884
		ext4_free_blocks(handle, inode, where[i].bh, 0, 1,
				 EXT4_FREE_BLOCKS_FORGET);
885
	}
886 887
	ext4_free_blocks(handle, inode, 0, le32_to_cpu(where[num].key),
			 blks, 0);
888 889 890 891 892

	return err;
}

/*
893 894 895 896
 * The ext4_ind_get_blocks() function handles non-extents inodes
 * (i.e., using the traditional indirect/double-indirect i_blocks
 * scheme) for ext4_get_blocks().
 *
897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912
 * Allocation strategy is simple: if we have to allocate something, we will
 * have to go the whole way to leaf. So let's do it before attaching anything
 * to tree, set linkage between the newborn blocks, write them if sync is
 * required, recheck the path, free and repeat if check fails, otherwise
 * set the last missing link (that will protect us from any truncate-generated
 * removals - all blocks on the path are immune now) and possibly force the
 * write on the parent block.
 * That has a nice additional property: no special recovery from the failed
 * allocations is needed - we simply release blocks and do not touch anything
 * reachable from inode.
 *
 * `handle' can be NULL if create == 0.
 *
 * return > 0, # of blocks mapped or allocated.
 * return = 0, if plain lookup failed.
 * return < 0, error case.
913
 *
914 915 916 917 918
 * The ext4_ind_get_blocks() function should be called with
 * down_write(&EXT4_I(inode)->i_data_sem) if allocating filesystem
 * blocks (i.e., flags has EXT4_GET_BLOCKS_CREATE set) or
 * down_read(&EXT4_I(inode)->i_data_sem) if not allocating file system
 * blocks.
919
 */
920
static int ext4_ind_get_blocks(handle_t *handle, struct inode *inode,
921 922 923
			       ext4_lblk_t iblock, unsigned int maxblocks,
			       struct buffer_head *bh_result,
			       int flags)
924 925
{
	int err = -EIO;
A
Aneesh Kumar K.V 已提交
926
	ext4_lblk_t offsets[4];
927 928
	Indirect chain[4];
	Indirect *partial;
929
	ext4_fsblk_t goal;
930 931 932 933
	int indirect_blks;
	int blocks_to_boundary = 0;
	int depth;
	int count = 0;
934
	ext4_fsblk_t first_block = 0;
935

A
Alex Tomas 已提交
936
	J_ASSERT(!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL));
937
	J_ASSERT(handle != NULL || (flags & EXT4_GET_BLOCKS_CREATE) == 0);
A
Aneesh Kumar K.V 已提交
938
	depth = ext4_block_to_path(inode, iblock, offsets,
939
				   &blocks_to_boundary);
940 941 942 943

	if (depth == 0)
		goto out;

944
	partial = ext4_get_branch(inode, depth, offsets, chain, &err);
945 946 947 948 949 950 951 952

	/* Simplest case - block found, no allocation needed */
	if (!partial) {
		first_block = le32_to_cpu(chain[depth - 1].key);
		clear_buffer_new(bh_result);
		count++;
		/*map more blocks*/
		while (count < maxblocks && count <= blocks_to_boundary) {
953
			ext4_fsblk_t blk;
954 955 956 957 958 959 960 961

			blk = le32_to_cpu(*(chain[depth-1].p + count));

			if (blk == first_block + count)
				count++;
			else
				break;
		}
962
		goto got_it;
963 964 965
	}

	/* Next simple case - plain lookup or failed read of indirect block */
966
	if ((flags & EXT4_GET_BLOCKS_CREATE) == 0 || err == -EIO)
967 968 969
		goto cleanup;

	/*
970
	 * Okay, we need to do block allocation.
971
	*/
972
	goal = ext4_find_goal(inode, iblock, partial);
973 974 975 976 977 978 979 980

	/* the number of blocks need to allocate for [d,t]indirect blocks */
	indirect_blks = (chain + depth) - partial - 1;

	/*
	 * Next look up the indirect map to count the totoal number of
	 * direct blocks to allocate for this branch.
	 */
981
	count = ext4_blks_to_allocate(partial, indirect_blks,
982 983
					maxblocks, blocks_to_boundary);
	/*
984
	 * Block out ext4_truncate while we alter the tree
985
	 */
986
	err = ext4_alloc_branch(handle, inode, iblock, indirect_blks,
987 988
				&count, goal,
				offsets + (partial - chain), partial);
989 990

	/*
991
	 * The ext4_splice_branch call will free and forget any buffers
992 993 994 995 996 997
	 * on the new chain if there is a failure, but that risks using
	 * up transaction credits, especially for bitmaps where the
	 * credits cannot be returned.  Can we handle this somehow?  We
	 * may need to return -EAGAIN upwards in the worst case.  --sct
	 */
	if (!err)
998
		err = ext4_splice_branch(handle, inode, iblock,
999
					 partial, indirect_blks, count);
1000
	if (err)
1001 1002 1003
		goto cleanup;

	set_buffer_new(bh_result);
1004 1005

	ext4_update_inode_fsync_trans(handle, inode, 1);
1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023
got_it:
	map_bh(bh_result, inode->i_sb, le32_to_cpu(chain[depth-1].key));
	if (count > blocks_to_boundary)
		set_buffer_boundary(bh_result);
	err = count;
	/* Clean up and exit */
	partial = chain + depth - 1;	/* the whole chain */
cleanup:
	while (partial > chain) {
		BUFFER_TRACE(partial->bh, "call brelse");
		brelse(partial->bh);
		partial--;
	}
	BUFFER_TRACE(bh_result, "returned");
out:
	return err;
}

1024 1025
#ifdef CONFIG_QUOTA
qsize_t *ext4_get_reserved_space(struct inode *inode)
1026
{
1027
	return &EXT4_I(inode)->i_reserved_quota;
1028
}
1029
#endif
1030

1031 1032
/*
 * Calculate the number of metadata blocks need to reserve
1033
 * to allocate a new block at @lblocks for non extent file based file
1034
 */
1035 1036
static int ext4_indirect_calc_metadata_amount(struct inode *inode,
					      sector_t lblock)
1037
{
1038
	struct ext4_inode_info *ei = EXT4_I(inode);
1039
	sector_t dind_mask = ~((sector_t)EXT4_ADDR_PER_BLOCK(inode->i_sb) - 1);
1040
	int blk_bits;
1041

1042 1043
	if (lblock < EXT4_NDIR_BLOCKS)
		return 0;
1044

1045
	lblock -= EXT4_NDIR_BLOCKS;
1046

1047 1048 1049 1050 1051 1052 1053
	if (ei->i_da_metadata_calc_len &&
	    (lblock & dind_mask) == ei->i_da_metadata_calc_last_lblock) {
		ei->i_da_metadata_calc_len++;
		return 0;
	}
	ei->i_da_metadata_calc_last_lblock = lblock & dind_mask;
	ei->i_da_metadata_calc_len = 1;
1054
	blk_bits = order_base_2(lblock);
1055
	return (blk_bits / EXT4_ADDR_PER_BLOCK_BITS(inode->i_sb)) + 1;
1056 1057 1058 1059
}

/*
 * Calculate the number of metadata blocks need to reserve
1060
 * to allocate a block located at @lblock
1061
 */
1062
static int ext4_calc_metadata_amount(struct inode *inode, sector_t lblock)
1063 1064
{
	if (EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL)
1065
		return ext4_ext_calc_metadata_amount(inode, lblock);
1066

1067
	return ext4_indirect_calc_metadata_amount(inode, lblock);
1068 1069
}

1070 1071 1072 1073
/*
 * Called with i_data_sem down, which is important since we can call
 * ext4_discard_preallocations() from here.
 */
1074 1075
void ext4_da_update_reserve_space(struct inode *inode,
					int used, int quota_claim)
1076 1077
{
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1078
	struct ext4_inode_info *ei = EXT4_I(inode);
1079
	int mdb_free = 0, allocated_meta_blocks = 0;
1080 1081

	spin_lock(&ei->i_block_reservation_lock);
1082
	trace_ext4_da_update_reserve_space(inode, used);
1083 1084 1085 1086 1087 1088 1089 1090
	if (unlikely(used > ei->i_reserved_data_blocks)) {
		ext4_msg(inode->i_sb, KERN_NOTICE, "%s: ino %lu, used %d "
			 "with only %d reserved data blocks\n",
			 __func__, inode->i_ino, used,
			 ei->i_reserved_data_blocks);
		WARN_ON(1);
		used = ei->i_reserved_data_blocks;
	}
1091

1092 1093 1094 1095
	/* Update per-inode reservations */
	ei->i_reserved_data_blocks -= used;
	used += ei->i_allocated_meta_blocks;
	ei->i_reserved_meta_blocks -= ei->i_allocated_meta_blocks;
1096
	allocated_meta_blocks = ei->i_allocated_meta_blocks;
1097 1098
	ei->i_allocated_meta_blocks = 0;
	percpu_counter_sub(&sbi->s_dirtyblocks_counter, used);
1099

1100 1101 1102 1103 1104 1105
	if (ei->i_reserved_data_blocks == 0) {
		/*
		 * We can release all of the reserved metadata blocks
		 * only when we have written all of the delayed
		 * allocation blocks.
		 */
1106 1107
		mdb_free = ei->i_reserved_meta_blocks;
		ei->i_reserved_meta_blocks = 0;
1108
		ei->i_da_metadata_calc_len = 0;
1109 1110
		percpu_counter_sub(&sbi->s_dirtyblocks_counter, mdb_free);
	}
1111
	spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
1112

1113
	/* Update quota subsystem */
1114
	if (quota_claim) {
1115
		dquot_claim_block(inode, used);
1116
		if (mdb_free)
1117
			dquot_release_reservation_block(inode, mdb_free);
1118 1119 1120 1121 1122 1123 1124 1125 1126 1127
	} else {
		/*
		 * We did fallocate with an offset that is already delayed
		 * allocated. So on delayed allocated writeback we should
		 * not update the quota for allocated blocks. But then
		 * converting an fallocate region to initialized region would
		 * have caused a metadata allocation. So claim quota for
		 * that
		 */
		if (allocated_meta_blocks)
1128
			dquot_claim_block(inode, allocated_meta_blocks);
1129 1130
		dquot_release_reservation_block(inode, mdb_free + used -
						allocated_meta_blocks);
1131
	}
1132 1133 1134 1135 1136 1137

	/*
	 * If we have done all the pending block allocations and if
	 * there aren't any writers on the inode, we can discard the
	 * inode's preallocations.
	 */
1138 1139
	if ((ei->i_reserved_data_blocks == 0) &&
	    (atomic_read(&inode->i_writecount) == 0))
1140
		ext4_discard_preallocations(inode);
1141 1142
}

1143 1144
static int check_block_validity(struct inode *inode, const char *msg,
				sector_t logical, sector_t phys, int len)
1145 1146
{
	if (!ext4_data_block_valid(EXT4_SB(inode->i_sb), phys, len)) {
1147
		__ext4_error(inode->i_sb, msg,
1148 1149 1150 1151 1152 1153 1154 1155 1156
			   "inode #%lu logical block %llu mapped to %llu "
			   "(size %d)", inode->i_ino,
			   (unsigned long long) logical,
			   (unsigned long long) phys, len);
		return -EIO;
	}
	return 0;
}

1157
/*
1158 1159
 * Return the number of contiguous dirty pages in a given inode
 * starting at page frame idx.
1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192
 */
static pgoff_t ext4_num_dirty_pages(struct inode *inode, pgoff_t idx,
				    unsigned int max_pages)
{
	struct address_space *mapping = inode->i_mapping;
	pgoff_t	index;
	struct pagevec pvec;
	pgoff_t num = 0;
	int i, nr_pages, done = 0;

	if (max_pages == 0)
		return 0;
	pagevec_init(&pvec, 0);
	while (!done) {
		index = idx;
		nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
					      PAGECACHE_TAG_DIRTY,
					      (pgoff_t)PAGEVEC_SIZE);
		if (nr_pages == 0)
			break;
		for (i = 0; i < nr_pages; i++) {
			struct page *page = pvec.pages[i];
			struct buffer_head *bh, *head;

			lock_page(page);
			if (unlikely(page->mapping != mapping) ||
			    !PageDirty(page) ||
			    PageWriteback(page) ||
			    page->index != idx) {
				done = 1;
				unlock_page(page);
				break;
			}
1193 1194 1195 1196 1197 1198 1199 1200 1201
			if (page_has_buffers(page)) {
				bh = head = page_buffers(page);
				do {
					if (!buffer_delay(bh) &&
					    !buffer_unwritten(bh))
						done = 1;
					bh = bh->b_this_page;
				} while (!done && (bh != head));
			}
1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214
			unlock_page(page);
			if (done)
				break;
			idx++;
			num++;
			if (num >= max_pages)
				break;
		}
		pagevec_release(&pvec);
	}
	return num;
}

1215
/*
1216
 * The ext4_get_blocks() function tries to look up the requested blocks,
1217
 * and returns if the blocks are already mapped.
1218 1219 1220 1221 1222 1223
 *
 * Otherwise it takes the write lock of the i_data_sem and allocate blocks
 * and store the allocated blocks in the result buffer head and mark it
 * mapped.
 *
 * If file type is extents based, it will call ext4_ext_get_blocks(),
1224
 * Otherwise, call with ext4_ind_get_blocks() to handle indirect mapping
1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236
 * based files
 *
 * On success, it returns the number of blocks being mapped or allocate.
 * if create==0 and the blocks are pre-allocated and uninitialized block,
 * the result buffer head is unmapped. If the create ==1, it will make sure
 * the buffer head is mapped.
 *
 * It returns 0 if plain look up failed (blocks have not been allocated), in
 * that casem, buffer head is unmapped
 *
 * It returns the error in case of allocation failure.
 */
1237 1238
int ext4_get_blocks(handle_t *handle, struct inode *inode, sector_t block,
		    unsigned int max_blocks, struct buffer_head *bh,
1239
		    int flags)
1240 1241
{
	int retval;
1242 1243

	clear_buffer_mapped(bh);
1244
	clear_buffer_unwritten(bh);
1245

1246 1247 1248
	ext_debug("ext4_get_blocks(): inode %lu, flag %d, max_blocks %u,"
		  "logical block %lu\n", inode->i_ino, flags, max_blocks,
		  (unsigned long)block);
1249
	/*
1250 1251
	 * Try to see if we can get the block without requesting a new
	 * file system block.
1252 1253 1254 1255
	 */
	down_read((&EXT4_I(inode)->i_data_sem));
	if (EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL) {
		retval =  ext4_ext_get_blocks(handle, inode, block, max_blocks,
1256
				bh, 0);
1257
	} else {
1258
		retval = ext4_ind_get_blocks(handle, inode, block, max_blocks,
1259
					     bh, 0);
1260
	}
1261
	up_read((&EXT4_I(inode)->i_data_sem));
1262

1263
	if (retval > 0 && buffer_mapped(bh)) {
1264 1265
		int ret = check_block_validity(inode, "file system corruption",
					       block, bh->b_blocknr, retval);
1266 1267 1268 1269
		if (ret != 0)
			return ret;
	}

1270
	/* If it is only a block(s) look up */
1271
	if ((flags & EXT4_GET_BLOCKS_CREATE) == 0)
1272 1273 1274 1275 1276 1277 1278 1279 1280 1281
		return retval;

	/*
	 * Returns if the blocks have already allocated
	 *
	 * Note that if blocks have been preallocated
	 * ext4_ext_get_block() returns th create = 0
	 * with buffer head unmapped.
	 */
	if (retval > 0 && buffer_mapped(bh))
1282 1283
		return retval;

1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295
	/*
	 * When we call get_blocks without the create flag, the
	 * BH_Unwritten flag could have gotten set if the blocks
	 * requested were part of a uninitialized extent.  We need to
	 * clear this flag now that we are committed to convert all or
	 * part of the uninitialized extent to be an initialized
	 * extent.  This is because we need to avoid the combination
	 * of BH_Unwritten and BH_Mapped flags being simultaneously
	 * set on the buffer_head.
	 */
	clear_buffer_unwritten(bh);

1296
	/*
1297 1298 1299 1300
	 * New blocks allocate and/or writing to uninitialized extent
	 * will possibly result in updating i_data, so we take
	 * the write lock of i_data_sem, and call get_blocks()
	 * with create == 1 flag.
1301 1302
	 */
	down_write((&EXT4_I(inode)->i_data_sem));
1303 1304 1305 1306 1307 1308 1309

	/*
	 * if the caller is from delayed allocation writeout path
	 * we have already reserved fs blocks for allocation
	 * let the underlying get_block() function know to
	 * avoid double accounting
	 */
1310
	if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
1311
		EXT4_I(inode)->i_delalloc_reserved_flag = 1;
1312 1313 1314 1315
	/*
	 * We need to check for EXT4 here because migrate
	 * could have changed the inode type in between
	 */
1316 1317
	if (EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL) {
		retval =  ext4_ext_get_blocks(handle, inode, block, max_blocks,
1318
					      bh, flags);
1319
	} else {
1320
		retval = ext4_ind_get_blocks(handle, inode, block,
1321
					     max_blocks, bh, flags);
1322 1323 1324 1325 1326 1327 1328

		if (retval > 0 && buffer_new(bh)) {
			/*
			 * We allocated new blocks which will result in
			 * i_data's format changing.  Force the migrate
			 * to fail by clearing migrate flags
			 */
1329
			ext4_clear_inode_state(inode, EXT4_STATE_EXT_MIGRATE);
1330
		}
1331

1332 1333 1334 1335 1336 1337 1338
		/*
		 * Update reserved blocks/metadata blocks after successful
		 * block allocation which had been deferred till now. We don't
		 * support fallocate for non extent files. So we can update
		 * reserve space here.
		 */
		if ((retval > 0) &&
1339
			(flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE))
1340 1341
			ext4_da_update_reserve_space(inode, retval, 1);
	}
1342
	if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
1343
		EXT4_I(inode)->i_delalloc_reserved_flag = 0;
1344

1345
	up_write((&EXT4_I(inode)->i_data_sem));
1346
	if (retval > 0 && buffer_mapped(bh)) {
1347 1348 1349
		int ret = check_block_validity(inode, "file system "
					       "corruption after allocation",
					       block, bh->b_blocknr, retval);
1350 1351 1352
		if (ret != 0)
			return ret;
	}
1353 1354 1355
	return retval;
}

1356 1357 1358
/* Maximum number of blocks we map for direct IO at once. */
#define DIO_MAX_BLOCKS 4096

1359 1360
int ext4_get_block(struct inode *inode, sector_t iblock,
		   struct buffer_head *bh_result, int create)
1361
{
1362
	handle_t *handle = ext4_journal_current_handle();
J
Jan Kara 已提交
1363
	int ret = 0, started = 0;
1364
	unsigned max_blocks = bh_result->b_size >> inode->i_blkbits;
1365
	int dio_credits;
1366

J
Jan Kara 已提交
1367 1368 1369 1370
	if (create && !handle) {
		/* Direct IO write... */
		if (max_blocks > DIO_MAX_BLOCKS)
			max_blocks = DIO_MAX_BLOCKS;
1371 1372
		dio_credits = ext4_chunk_trans_blocks(inode, max_blocks);
		handle = ext4_journal_start(inode, dio_credits);
J
Jan Kara 已提交
1373
		if (IS_ERR(handle)) {
1374
			ret = PTR_ERR(handle);
J
Jan Kara 已提交
1375
			goto out;
1376
		}
J
Jan Kara 已提交
1377
		started = 1;
1378 1379
	}

1380
	ret = ext4_get_blocks(handle, inode, iblock, max_blocks, bh_result,
1381
			      create ? EXT4_GET_BLOCKS_CREATE : 0);
J
Jan Kara 已提交
1382 1383 1384
	if (ret > 0) {
		bh_result->b_size = (ret << inode->i_blkbits);
		ret = 0;
1385
	}
J
Jan Kara 已提交
1386 1387 1388
	if (started)
		ext4_journal_stop(handle);
out:
1389 1390 1391 1392 1393 1394
	return ret;
}

/*
 * `handle' can be NULL if create is zero
 */
1395
struct buffer_head *ext4_getblk(handle_t *handle, struct inode *inode,
A
Aneesh Kumar K.V 已提交
1396
				ext4_lblk_t block, int create, int *errp)
1397 1398 1399
{
	struct buffer_head dummy;
	int fatal = 0, err;
1400
	int flags = 0;
1401 1402 1403 1404 1405 1406

	J_ASSERT(handle != NULL || create == 0);

	dummy.b_state = 0;
	dummy.b_blocknr = -1000;
	buffer_trace_init(&dummy.b_history);
1407 1408 1409
	if (create)
		flags |= EXT4_GET_BLOCKS_CREATE;
	err = ext4_get_blocks(handle, inode, block, 1, &dummy, flags);
1410
	/*
1411 1412
	 * ext4_get_blocks() returns number of blocks mapped. 0 in
	 * case of a HOLE.
1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428
	 */
	if (err > 0) {
		if (err > 1)
			WARN_ON(1);
		err = 0;
	}
	*errp = err;
	if (!err && buffer_mapped(&dummy)) {
		struct buffer_head *bh;
		bh = sb_getblk(inode->i_sb, dummy.b_blocknr);
		if (!bh) {
			*errp = -EIO;
			goto err;
		}
		if (buffer_new(&dummy)) {
			J_ASSERT(create != 0);
A
Aneesh Kumar K.V 已提交
1429
			J_ASSERT(handle != NULL);
1430 1431 1432 1433 1434

			/*
			 * Now that we do not always journal data, we should
			 * keep in mind whether this should always journal the
			 * new buffer as metadata.  For now, regular file
1435
			 * writes use ext4_get_block instead, so it's not a
1436 1437 1438 1439
			 * problem.
			 */
			lock_buffer(bh);
			BUFFER_TRACE(bh, "call get_create_access");
1440
			fatal = ext4_journal_get_create_access(handle, bh);
1441
			if (!fatal && !buffer_uptodate(bh)) {
1442
				memset(bh->b_data, 0, inode->i_sb->s_blocksize);
1443 1444 1445
				set_buffer_uptodate(bh);
			}
			unlock_buffer(bh);
1446 1447
			BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
			err = ext4_handle_dirty_metadata(handle, inode, bh);
1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463
			if (!fatal)
				fatal = err;
		} else {
			BUFFER_TRACE(bh, "not a new buffer");
		}
		if (fatal) {
			*errp = fatal;
			brelse(bh);
			bh = NULL;
		}
		return bh;
	}
err:
	return NULL;
}

1464
struct buffer_head *ext4_bread(handle_t *handle, struct inode *inode,
A
Aneesh Kumar K.V 已提交
1465
			       ext4_lblk_t block, int create, int *err)
1466
{
1467
	struct buffer_head *bh;
1468

1469
	bh = ext4_getblk(handle, inode, block, create, err);
1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482
	if (!bh)
		return bh;
	if (buffer_uptodate(bh))
		return bh;
	ll_rw_block(READ_META, 1, &bh);
	wait_on_buffer(bh);
	if (buffer_uptodate(bh))
		return bh;
	put_bh(bh);
	*err = -EIO;
	return NULL;
}

1483 1484 1485 1486 1487 1488 1489
static int walk_page_buffers(handle_t *handle,
			     struct buffer_head *head,
			     unsigned from,
			     unsigned to,
			     int *partial,
			     int (*fn)(handle_t *handle,
				       struct buffer_head *bh))
1490 1491 1492 1493 1494 1495 1496
{
	struct buffer_head *bh;
	unsigned block_start, block_end;
	unsigned blocksize = head->b_size;
	int err, ret = 0;
	struct buffer_head *next;

1497 1498
	for (bh = head, block_start = 0;
	     ret == 0 && (bh != head || !block_start);
1499
	     block_start = block_end, bh = next) {
1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516
		next = bh->b_this_page;
		block_end = block_start + blocksize;
		if (block_end <= from || block_start >= to) {
			if (partial && !buffer_uptodate(bh))
				*partial = 1;
			continue;
		}
		err = (*fn)(handle, bh);
		if (!ret)
			ret = err;
	}
	return ret;
}

/*
 * To preserve ordering, it is essential that the hole instantiation and
 * the data write be encapsulated in a single transaction.  We cannot
1517
 * close off a transaction and start a new one between the ext4_get_block()
1518
 * and the commit_write().  So doing the jbd2_journal_start at the start of
1519 1520
 * prepare_write() is the right place.
 *
1521 1522
 * Also, this function can nest inside ext4_writepage() ->
 * block_write_full_page(). In that case, we *know* that ext4_writepage()
1523 1524 1525 1526
 * has generated enough buffer credits to do the whole page.  So we won't
 * block on the journal in that case, which is good, because the caller may
 * be PF_MEMALLOC.
 *
1527
 * By accident, ext4 can be reentered when a transaction is open via
1528 1529 1530 1531 1532 1533
 * quota file writes.  If we were to commit the transaction while thus
 * reentered, there can be a deadlock - we would be holding a quota
 * lock, and the commit would never complete if another thread had a
 * transaction open and was blocking on the quota lock - a ranking
 * violation.
 *
1534
 * So what we do is to rely on the fact that jbd2_journal_stop/journal_start
1535 1536 1537 1538 1539
 * will _not_ run commit under these circumstances because handle->h_ref
 * is elevated.  We'll still have enough credits for the tiny quotafile
 * write.
 */
static int do_journal_get_write_access(handle_t *handle,
1540
				       struct buffer_head *bh)
1541 1542 1543
{
	if (!buffer_mapped(bh) || buffer_freed(bh))
		return 0;
1544
	return ext4_journal_get_write_access(handle, bh);
1545 1546
}

1547 1548 1549 1550 1551 1552 1553 1554 1555 1556
/*
 * Truncate blocks that were not used by write. We have to truncate the
 * pagecache as well so that corresponding buffers get properly unmapped.
 */
static void ext4_truncate_failed_write(struct inode *inode)
{
	truncate_inode_pages(inode->i_mapping, inode->i_size);
	ext4_truncate(inode);
}

1557 1558
static int ext4_get_block_write(struct inode *inode, sector_t iblock,
		   struct buffer_head *bh_result, int create);
N
Nick Piggin 已提交
1559
static int ext4_write_begin(struct file *file, struct address_space *mapping,
1560 1561
			    loff_t pos, unsigned len, unsigned flags,
			    struct page **pagep, void **fsdata)
1562
{
1563
	struct inode *inode = mapping->host;
1564
	int ret, needed_blocks;
1565 1566
	handle_t *handle;
	int retries = 0;
1567
	struct page *page;
1568
	pgoff_t index;
1569
	unsigned from, to;
N
Nick Piggin 已提交
1570

1571
	trace_ext4_write_begin(inode, pos, len, flags);
1572 1573 1574 1575 1576
	/*
	 * Reserve one block more for addition to orphan list in case
	 * we allocate blocks but write fails for some reason
	 */
	needed_blocks = ext4_writepage_trans_blocks(inode) + 1;
1577
	index = pos >> PAGE_CACHE_SHIFT;
1578 1579
	from = pos & (PAGE_CACHE_SIZE - 1);
	to = from + len;
1580 1581

retry:
1582 1583 1584 1585
	handle = ext4_journal_start(inode, needed_blocks);
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		goto out;
1586
	}
1587

1588 1589 1590 1591
	/* We cannot recurse into the filesystem as the transaction is already
	 * started */
	flags |= AOP_FLAG_NOFS;

1592
	page = grab_cache_page_write_begin(mapping, index, flags);
1593 1594 1595 1596 1597 1598 1599
	if (!page) {
		ext4_journal_stop(handle);
		ret = -ENOMEM;
		goto out;
	}
	*pagep = page;

1600 1601 1602 1603 1604 1605
	if (ext4_should_dioread_nolock(inode))
		ret = block_write_begin(file, mapping, pos, len, flags, pagep,
				fsdata, ext4_get_block_write);
	else
		ret = block_write_begin(file, mapping, pos, len, flags, pagep,
				fsdata, ext4_get_block);
N
Nick Piggin 已提交
1606 1607

	if (!ret && ext4_should_journal_data(inode)) {
1608 1609 1610
		ret = walk_page_buffers(handle, page_buffers(page),
				from, to, NULL, do_journal_get_write_access);
	}
N
Nick Piggin 已提交
1611 1612

	if (ret) {
1613 1614
		unlock_page(page);
		page_cache_release(page);
1615 1616 1617 1618
		/*
		 * block_write_begin may have instantiated a few blocks
		 * outside i_size.  Trim these off again. Don't need
		 * i_size_read because we hold i_mutex.
1619 1620 1621
		 *
		 * Add inode to orphan list in case we crash before
		 * truncate finishes
1622
		 */
1623
		if (pos + len > inode->i_size && ext4_can_truncate(inode))
1624 1625 1626 1627
			ext4_orphan_add(handle, inode);

		ext4_journal_stop(handle);
		if (pos + len > inode->i_size) {
1628
			ext4_truncate_failed_write(inode);
1629
			/*
1630
			 * If truncate failed early the inode might
1631 1632 1633 1634 1635 1636 1637
			 * still be on the orphan list; we need to
			 * make sure the inode is removed from the
			 * orphan list in that case.
			 */
			if (inode->i_nlink)
				ext4_orphan_del(NULL, inode);
		}
N
Nick Piggin 已提交
1638 1639
	}

1640
	if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1641
		goto retry;
1642
out:
1643 1644 1645
	return ret;
}

N
Nick Piggin 已提交
1646 1647
/* For write_end() in data=journal mode */
static int write_end_fn(handle_t *handle, struct buffer_head *bh)
1648 1649 1650 1651
{
	if (!buffer_mapped(bh) || buffer_freed(bh))
		return 0;
	set_buffer_uptodate(bh);
1652
	return ext4_handle_dirty_metadata(handle, NULL, bh);
1653 1654
}

1655
static int ext4_generic_write_end(struct file *file,
1656 1657 1658
				  struct address_space *mapping,
				  loff_t pos, unsigned len, unsigned copied,
				  struct page *page, void *fsdata)
1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700
{
	int i_size_changed = 0;
	struct inode *inode = mapping->host;
	handle_t *handle = ext4_journal_current_handle();

	copied = block_write_end(file, mapping, pos, len, copied, page, fsdata);

	/*
	 * No need to use i_size_read() here, the i_size
	 * cannot change under us because we hold i_mutex.
	 *
	 * But it's important to update i_size while still holding page lock:
	 * page writeout could otherwise come in and zero beyond i_size.
	 */
	if (pos + copied > inode->i_size) {
		i_size_write(inode, pos + copied);
		i_size_changed = 1;
	}

	if (pos + copied >  EXT4_I(inode)->i_disksize) {
		/* We need to mark inode dirty even if
		 * new_i_size is less that inode->i_size
		 * bu greater than i_disksize.(hint delalloc)
		 */
		ext4_update_i_disksize(inode, (pos + copied));
		i_size_changed = 1;
	}
	unlock_page(page);
	page_cache_release(page);

	/*
	 * Don't mark the inode dirty under page lock. First, it unnecessarily
	 * makes the holding time of page lock longer. Second, it forces lock
	 * ordering of page lock and transaction start for journaling
	 * filesystems.
	 */
	if (i_size_changed)
		ext4_mark_inode_dirty(handle, inode);

	return copied;
}

1701 1702 1703 1704
/*
 * We need to pick up the new inode size which generic_commit_write gave us
 * `file' can be NULL - eg, when called from page_symlink().
 *
1705
 * ext4 never places buffers on inode->i_mapping->private_list.  metadata
1706 1707
 * buffers are managed internally.
 */
N
Nick Piggin 已提交
1708
static int ext4_ordered_write_end(struct file *file,
1709 1710 1711
				  struct address_space *mapping,
				  loff_t pos, unsigned len, unsigned copied,
				  struct page *page, void *fsdata)
1712
{
1713
	handle_t *handle = ext4_journal_current_handle();
1714
	struct inode *inode = mapping->host;
1715 1716
	int ret = 0, ret2;

1717
	trace_ext4_ordered_write_end(inode, pos, len, copied);
1718
	ret = ext4_jbd2_file_inode(handle, inode);
1719 1720

	if (ret == 0) {
1721
		ret2 = ext4_generic_write_end(file, mapping, pos, len, copied,
N
Nick Piggin 已提交
1722
							page, fsdata);
1723
		copied = ret2;
1724
		if (pos + len > inode->i_size && ext4_can_truncate(inode))
1725 1726 1727 1728 1729
			/* if we have allocated more blocks and copied
			 * less. We will have blocks allocated outside
			 * inode->i_size. So truncate them
			 */
			ext4_orphan_add(handle, inode);
1730 1731
		if (ret2 < 0)
			ret = ret2;
1732
	}
1733
	ret2 = ext4_journal_stop(handle);
1734 1735
	if (!ret)
		ret = ret2;
N
Nick Piggin 已提交
1736

1737
	if (pos + len > inode->i_size) {
1738
		ext4_truncate_failed_write(inode);
1739
		/*
1740
		 * If truncate failed early the inode might still be
1741 1742 1743 1744 1745 1746 1747 1748
		 * on the orphan list; we need to make sure the inode
		 * is removed from the orphan list in that case.
		 */
		if (inode->i_nlink)
			ext4_orphan_del(NULL, inode);
	}


N
Nick Piggin 已提交
1749
	return ret ? ret : copied;
1750 1751
}

N
Nick Piggin 已提交
1752
static int ext4_writeback_write_end(struct file *file,
1753 1754 1755
				    struct address_space *mapping,
				    loff_t pos, unsigned len, unsigned copied,
				    struct page *page, void *fsdata)
1756
{
1757
	handle_t *handle = ext4_journal_current_handle();
1758
	struct inode *inode = mapping->host;
1759 1760
	int ret = 0, ret2;

1761
	trace_ext4_writeback_write_end(inode, pos, len, copied);
1762
	ret2 = ext4_generic_write_end(file, mapping, pos, len, copied,
N
Nick Piggin 已提交
1763
							page, fsdata);
1764
	copied = ret2;
1765
	if (pos + len > inode->i_size && ext4_can_truncate(inode))
1766 1767 1768 1769 1770 1771
		/* if we have allocated more blocks and copied
		 * less. We will have blocks allocated outside
		 * inode->i_size. So truncate them
		 */
		ext4_orphan_add(handle, inode);

1772 1773
	if (ret2 < 0)
		ret = ret2;
1774

1775
	ret2 = ext4_journal_stop(handle);
1776 1777
	if (!ret)
		ret = ret2;
N
Nick Piggin 已提交
1778

1779
	if (pos + len > inode->i_size) {
1780
		ext4_truncate_failed_write(inode);
1781
		/*
1782
		 * If truncate failed early the inode might still be
1783 1784 1785 1786 1787 1788 1789
		 * on the orphan list; we need to make sure the inode
		 * is removed from the orphan list in that case.
		 */
		if (inode->i_nlink)
			ext4_orphan_del(NULL, inode);
	}

N
Nick Piggin 已提交
1790
	return ret ? ret : copied;
1791 1792
}

N
Nick Piggin 已提交
1793
static int ext4_journalled_write_end(struct file *file,
1794 1795 1796
				     struct address_space *mapping,
				     loff_t pos, unsigned len, unsigned copied,
				     struct page *page, void *fsdata)
1797
{
1798
	handle_t *handle = ext4_journal_current_handle();
N
Nick Piggin 已提交
1799
	struct inode *inode = mapping->host;
1800 1801
	int ret = 0, ret2;
	int partial = 0;
N
Nick Piggin 已提交
1802
	unsigned from, to;
1803
	loff_t new_i_size;
1804

1805
	trace_ext4_journalled_write_end(inode, pos, len, copied);
N
Nick Piggin 已提交
1806 1807 1808 1809 1810 1811 1812 1813
	from = pos & (PAGE_CACHE_SIZE - 1);
	to = from + len;

	if (copied < len) {
		if (!PageUptodate(page))
			copied = 0;
		page_zero_new_buffers(page, from+copied, to);
	}
1814 1815

	ret = walk_page_buffers(handle, page_buffers(page), from,
N
Nick Piggin 已提交
1816
				to, &partial, write_end_fn);
1817 1818
	if (!partial)
		SetPageUptodate(page);
1819 1820
	new_i_size = pos + copied;
	if (new_i_size > inode->i_size)
N
Nick Piggin 已提交
1821
		i_size_write(inode, pos+copied);
1822
	ext4_set_inode_state(inode, EXT4_STATE_JDATA);
1823 1824
	if (new_i_size > EXT4_I(inode)->i_disksize) {
		ext4_update_i_disksize(inode, new_i_size);
1825
		ret2 = ext4_mark_inode_dirty(handle, inode);
1826 1827 1828
		if (!ret)
			ret = ret2;
	}
N
Nick Piggin 已提交
1829

1830
	unlock_page(page);
1831
	page_cache_release(page);
1832
	if (pos + len > inode->i_size && ext4_can_truncate(inode))
1833 1834 1835 1836 1837 1838
		/* if we have allocated more blocks and copied
		 * less. We will have blocks allocated outside
		 * inode->i_size. So truncate them
		 */
		ext4_orphan_add(handle, inode);

1839
	ret2 = ext4_journal_stop(handle);
1840 1841
	if (!ret)
		ret = ret2;
1842
	if (pos + len > inode->i_size) {
1843
		ext4_truncate_failed_write(inode);
1844
		/*
1845
		 * If truncate failed early the inode might still be
1846 1847 1848 1849 1850 1851
		 * on the orphan list; we need to make sure the inode
		 * is removed from the orphan list in that case.
		 */
		if (inode->i_nlink)
			ext4_orphan_del(NULL, inode);
	}
N
Nick Piggin 已提交
1852 1853

	return ret ? ret : copied;
1854
}
1855

1856 1857 1858 1859
/*
 * Reserve a single block located at lblock
 */
static int ext4_da_reserve_space(struct inode *inode, sector_t lblock)
1860
{
A
Aneesh Kumar K.V 已提交
1861
	int retries = 0;
1862
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1863
	struct ext4_inode_info *ei = EXT4_I(inode);
1864
	unsigned long md_needed, md_reserved;
1865
	int ret;
1866 1867 1868 1869 1870 1871

	/*
	 * recalculate the amount of metadata blocks to reserve
	 * in order to allocate nrblocks
	 * worse case is one extent per block
	 */
A
Aneesh Kumar K.V 已提交
1872
repeat:
1873 1874
	spin_lock(&ei->i_block_reservation_lock);
	md_reserved = ei->i_reserved_meta_blocks;
1875
	md_needed = ext4_calc_metadata_amount(inode, lblock);
1876
	trace_ext4_da_reserve_space(inode, md_needed);
1877
	spin_unlock(&ei->i_block_reservation_lock);
1878

1879 1880 1881 1882 1883
	/*
	 * Make quota reservation here to prevent quota overflow
	 * later. Real quota accounting is done at pages writeout
	 * time.
	 */
1884 1885 1886
	ret = dquot_reserve_block(inode, md_needed + 1);
	if (ret)
		return ret;
1887

1888
	if (ext4_claim_free_blocks(sbi, md_needed + 1)) {
1889
		dquot_release_reservation_block(inode, md_needed + 1);
A
Aneesh Kumar K.V 已提交
1890 1891 1892 1893
		if (ext4_should_retry_alloc(inode->i_sb, &retries)) {
			yield();
			goto repeat;
		}
1894 1895
		return -ENOSPC;
	}
1896
	spin_lock(&ei->i_block_reservation_lock);
1897
	ei->i_reserved_data_blocks++;
1898 1899
	ei->i_reserved_meta_blocks += md_needed;
	spin_unlock(&ei->i_block_reservation_lock);
1900

1901 1902 1903
	return 0;       /* success */
}

1904
static void ext4_da_release_space(struct inode *inode, int to_free)
1905 1906
{
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1907
	struct ext4_inode_info *ei = EXT4_I(inode);
1908

1909 1910 1911
	if (!to_free)
		return;		/* Nothing to release, exit */

1912
	spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
1913

1914
	if (unlikely(to_free > ei->i_reserved_data_blocks)) {
1915
		/*
1916 1917 1918 1919
		 * if there aren't enough reserved blocks, then the
		 * counter is messed up somewhere.  Since this
		 * function is called from invalidate page, it's
		 * harmless to return without any action.
1920
		 */
1921 1922 1923 1924 1925 1926
		ext4_msg(inode->i_sb, KERN_NOTICE, "ext4_da_release_space: "
			 "ino %lu, to_free %d with only %d reserved "
			 "data blocks\n", inode->i_ino, to_free,
			 ei->i_reserved_data_blocks);
		WARN_ON(1);
		to_free = ei->i_reserved_data_blocks;
1927
	}
1928
	ei->i_reserved_data_blocks -= to_free;
1929

1930 1931 1932 1933 1934 1935
	if (ei->i_reserved_data_blocks == 0) {
		/*
		 * We can release all of the reserved metadata blocks
		 * only when we have written all of the delayed
		 * allocation blocks.
		 */
1936 1937
		to_free += ei->i_reserved_meta_blocks;
		ei->i_reserved_meta_blocks = 0;
1938
		ei->i_da_metadata_calc_len = 0;
1939
	}
1940

1941 1942
	/* update fs dirty blocks counter */
	percpu_counter_sub(&sbi->s_dirtyblocks_counter, to_free);
1943 1944

	spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
1945

1946
	dquot_release_reservation_block(inode, to_free);
1947 1948 1949
}

static void ext4_da_page_release_reservation(struct page *page,
1950
					     unsigned long offset)
1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966
{
	int to_release = 0;
	struct buffer_head *head, *bh;
	unsigned int curr_off = 0;

	head = page_buffers(page);
	bh = head;
	do {
		unsigned int next_off = curr_off + bh->b_size;

		if ((offset <= curr_off) && (buffer_delay(bh))) {
			to_release++;
			clear_buffer_delay(bh);
		}
		curr_off = next_off;
	} while ((bh = bh->b_this_page) != head);
1967
	ext4_da_release_space(page->mapping->host, to_release);
1968
}
1969

1970 1971 1972 1973 1974 1975
/*
 * Delayed allocation stuff
 */

/*
 * mpage_da_submit_io - walks through extent of pages and try to write
1976
 * them with writepage() call back
1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988
 *
 * @mpd->inode: inode
 * @mpd->first_page: first page of the extent
 * @mpd->next_page: page after the last page of the extent
 *
 * By the time mpage_da_submit_io() is called we expect all blocks
 * to be allocated. this may be wrong if allocation failed.
 *
 * As pages are already locked by write_cache_pages(), we can't use it
 */
static int mpage_da_submit_io(struct mpage_da_data *mpd)
{
1989
	long pages_skipped;
1990 1991 1992 1993 1994
	struct pagevec pvec;
	unsigned long index, end;
	int ret = 0, err, nr_pages, i;
	struct inode *inode = mpd->inode;
	struct address_space *mapping = inode->i_mapping;
1995 1996

	BUG_ON(mpd->next_page <= mpd->first_page);
1997 1998 1999
	/*
	 * We need to start from the first_page to the next_page - 1
	 * to make sure we also write the mapped dirty buffer_heads.
2000
	 * If we look at mpd->b_blocknr we would only be looking
2001 2002
	 * at the currently mapped buffer_heads.
	 */
2003 2004 2005
	index = mpd->first_page;
	end = mpd->next_page - 1;

2006
	pagevec_init(&pvec, 0);
2007
	while (index <= end) {
2008
		nr_pages = pagevec_lookup(&pvec, mapping, index, PAGEVEC_SIZE);
2009 2010 2011 2012 2013
		if (nr_pages == 0)
			break;
		for (i = 0; i < nr_pages; i++) {
			struct page *page = pvec.pages[i];

2014 2015 2016 2017 2018 2019 2020 2021
			index = page->index;
			if (index > end)
				break;
			index++;

			BUG_ON(!PageLocked(page));
			BUG_ON(PageWriteback(page));

2022
			pages_skipped = mpd->wbc->pages_skipped;
2023
			err = mapping->a_ops->writepage(page, mpd->wbc);
2024 2025 2026 2027 2028
			if (!err && (pages_skipped == mpd->wbc->pages_skipped))
				/*
				 * have successfully written the page
				 * without skipping the same
				 */
2029
				mpd->pages_written++;
2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051
			/*
			 * In error case, we have to continue because
			 * remaining pages are still locked
			 * XXX: unlock and re-dirty them?
			 */
			if (ret == 0)
				ret = err;
		}
		pagevec_release(&pvec);
	}
	return ret;
}

/*
 * mpage_put_bnr_to_bhs - walk blocks and assign them actual numbers
 *
 * @mpd->inode - inode to walk through
 * @exbh->b_blocknr - first block on a disk
 * @exbh->b_size - amount of space in bytes
 * @logical - first logical block to start assignment with
 *
 * the function goes through all passed space and put actual disk
2052
 * block numbers into buffer heads, dropping BH_Delay and BH_Unwritten
2053 2054 2055 2056 2057 2058 2059 2060 2061
 */
static void mpage_put_bnr_to_bhs(struct mpage_da_data *mpd, sector_t logical,
				 struct buffer_head *exbh)
{
	struct inode *inode = mpd->inode;
	struct address_space *mapping = inode->i_mapping;
	int blocks = exbh->b_size >> inode->i_blkbits;
	sector_t pblock = exbh->b_blocknr, cur_logical;
	struct buffer_head *head, *bh;
2062
	pgoff_t index, end;
2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101
	struct pagevec pvec;
	int nr_pages, i;

	index = logical >> (PAGE_CACHE_SHIFT - inode->i_blkbits);
	end = (logical + blocks - 1) >> (PAGE_CACHE_SHIFT - inode->i_blkbits);
	cur_logical = index << (PAGE_CACHE_SHIFT - inode->i_blkbits);

	pagevec_init(&pvec, 0);

	while (index <= end) {
		/* XXX: optimize tail */
		nr_pages = pagevec_lookup(&pvec, mapping, index, PAGEVEC_SIZE);
		if (nr_pages == 0)
			break;
		for (i = 0; i < nr_pages; i++) {
			struct page *page = pvec.pages[i];

			index = page->index;
			if (index > end)
				break;
			index++;

			BUG_ON(!PageLocked(page));
			BUG_ON(PageWriteback(page));
			BUG_ON(!page_has_buffers(page));

			bh = page_buffers(page);
			head = bh;

			/* skip blocks out of the range */
			do {
				if (cur_logical >= logical)
					break;
				cur_logical++;
			} while ((bh = bh->b_this_page) != head);

			do {
				if (cur_logical >= logical + blocks)
					break;
2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119

				if (buffer_delay(bh) ||
						buffer_unwritten(bh)) {

					BUG_ON(bh->b_bdev != inode->i_sb->s_bdev);

					if (buffer_delay(bh)) {
						clear_buffer_delay(bh);
						bh->b_blocknr = pblock;
					} else {
						/*
						 * unwritten already should have
						 * blocknr assigned. Verify that
						 */
						clear_buffer_unwritten(bh);
						BUG_ON(bh->b_blocknr != pblock);
					}

2120
				} else if (buffer_mapped(bh))
2121 2122
					BUG_ON(bh->b_blocknr != pblock);

2123 2124
				if (buffer_uninit(exbh))
					set_buffer_uninit(bh);
2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148
				cur_logical++;
				pblock++;
			} while ((bh = bh->b_this_page) != head);
		}
		pagevec_release(&pvec);
	}
}


/*
 * __unmap_underlying_blocks - just a helper function to unmap
 * set of blocks described by @bh
 */
static inline void __unmap_underlying_blocks(struct inode *inode,
					     struct buffer_head *bh)
{
	struct block_device *bdev = inode->i_sb->s_bdev;
	int blocks, i;

	blocks = bh->b_size >> inode->i_blkbits;
	for (i = 0; i < blocks; i++)
		unmap_underlying_metadata(bdev, bh->b_blocknr + i);
}

2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166
static void ext4_da_block_invalidatepages(struct mpage_da_data *mpd,
					sector_t logical, long blk_cnt)
{
	int nr_pages, i;
	pgoff_t index, end;
	struct pagevec pvec;
	struct inode *inode = mpd->inode;
	struct address_space *mapping = inode->i_mapping;

	index = logical >> (PAGE_CACHE_SHIFT - inode->i_blkbits);
	end   = (logical + blk_cnt - 1) >>
				(PAGE_CACHE_SHIFT - inode->i_blkbits);
	while (index <= end) {
		nr_pages = pagevec_lookup(&pvec, mapping, index, PAGEVEC_SIZE);
		if (nr_pages == 0)
			break;
		for (i = 0; i < nr_pages; i++) {
			struct page *page = pvec.pages[i];
2167
			if (page->index > end)
2168 2169 2170 2171 2172 2173 2174
				break;
			BUG_ON(!PageLocked(page));
			BUG_ON(PageWriteback(page));
			block_invalidatepage(page, 0);
			ClearPageUptodate(page);
			unlock_page(page);
		}
2175 2176
		index = pvec.pages[nr_pages - 1]->index + 1;
		pagevec_release(&pvec);
2177 2178 2179 2180
	}
	return;
}

2181 2182 2183
static void ext4_print_free_blocks(struct inode *inode)
{
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195
	printk(KERN_CRIT "Total free blocks count %lld\n",
	       ext4_count_free_blocks(inode->i_sb));
	printk(KERN_CRIT "Free/Dirty block details\n");
	printk(KERN_CRIT "free_blocks=%lld\n",
	       (long long) percpu_counter_sum(&sbi->s_freeblocks_counter));
	printk(KERN_CRIT "dirty_blocks=%lld\n",
	       (long long) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
	printk(KERN_CRIT "Block reservation details\n");
	printk(KERN_CRIT "i_reserved_data_blocks=%u\n",
	       EXT4_I(inode)->i_reserved_data_blocks);
	printk(KERN_CRIT "i_reserved_meta_blocks=%u\n",
	       EXT4_I(inode)->i_reserved_meta_blocks);
2196 2197 2198
	return;
}

2199 2200 2201
/*
 * mpage_da_map_blocks - go through given space
 *
2202
 * @mpd - bh describing space
2203 2204 2205 2206
 *
 * The function skips space we know is already mapped to disk blocks.
 *
 */
2207
static int mpage_da_map_blocks(struct mpage_da_data *mpd)
2208
{
2209
	int err, blks, get_blocks_flags;
A
Aneesh Kumar K.V 已提交
2210
	struct buffer_head new;
2211 2212 2213 2214
	sector_t next = mpd->b_blocknr;
	unsigned max_blocks = mpd->b_size >> mpd->inode->i_blkbits;
	loff_t disksize = EXT4_I(mpd->inode)->i_disksize;
	handle_t *handle = NULL;
2215 2216 2217 2218

	/*
	 * We consider only non-mapped and non-allocated blocks
	 */
2219
	if ((mpd->b_state  & (1 << BH_Mapped)) &&
2220 2221
		!(mpd->b_state & (1 << BH_Delay)) &&
		!(mpd->b_state & (1 << BH_Unwritten)))
2222
		return 0;
2223 2224 2225 2226 2227 2228 2229 2230 2231 2232

	/*
	 * If we didn't accumulate anything to write simply return
	 */
	if (!mpd->b_size)
		return 0;

	handle = ext4_journal_current_handle();
	BUG_ON(!handle);

2233
	/*
2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249
	 * Call ext4_get_blocks() to allocate any delayed allocation
	 * blocks, or to convert an uninitialized extent to be
	 * initialized (in the case where we have written into
	 * one or more preallocated blocks).
	 *
	 * We pass in the magic EXT4_GET_BLOCKS_DELALLOC_RESERVE to
	 * indicate that we are on the delayed allocation path.  This
	 * affects functions in many different parts of the allocation
	 * call path.  This flag exists primarily because we don't
	 * want to change *many* call functions, so ext4_get_blocks()
	 * will set the magic i_delalloc_reserved_flag once the
	 * inode's allocation semaphore is taken.
	 *
	 * If the blocks in questions were delalloc blocks, set
	 * EXT4_GET_BLOCKS_DELALLOC_RESERVE so the delalloc accounting
	 * variables are updated after the blocks have been allocated.
2250
	 */
2251
	new.b_state = 0;
2252
	get_blocks_flags = EXT4_GET_BLOCKS_CREATE;
2253 2254
	if (ext4_should_dioread_nolock(mpd->inode))
		get_blocks_flags |= EXT4_GET_BLOCKS_IO_CREATE_EXT;
2255
	if (mpd->b_state & (1 << BH_Delay))
2256 2257
		get_blocks_flags |= EXT4_GET_BLOCKS_DELALLOC_RESERVE;

2258
	blks = ext4_get_blocks(handle, mpd->inode, next, max_blocks,
2259
			       &new, get_blocks_flags);
2260 2261
	if (blks < 0) {
		err = blks;
2262 2263 2264 2265
		/*
		 * If get block returns with error we simply
		 * return. Later writepage will redirty the page and
		 * writepages will find the dirty page again
2266 2267 2268
		 */
		if (err == -EAGAIN)
			return 0;
2269 2270

		if (err == -ENOSPC &&
2271
		    ext4_count_free_blocks(mpd->inode->i_sb)) {
2272 2273 2274 2275
			mpd->retval = err;
			return 0;
		}

2276
		/*
2277 2278 2279 2280 2281
		 * get block failure will cause us to loop in
		 * writepages, because a_ops->writepage won't be able
		 * to make progress. The page will be redirtied by
		 * writepage and writepages will again try to write
		 * the same.
2282
		 */
2283 2284 2285 2286 2287 2288 2289 2290
		ext4_msg(mpd->inode->i_sb, KERN_CRIT,
			 "delayed block allocation failed for inode %lu at "
			 "logical offset %llu with max blocks %zd with "
			 "error %d\n", mpd->inode->i_ino,
			 (unsigned long long) next,
			 mpd->b_size >> mpd->inode->i_blkbits, err);
		printk(KERN_CRIT "This should not happen!!  "
		       "Data will be lost\n");
A
Aneesh Kumar K.V 已提交
2291
		if (err == -ENOSPC) {
2292
			ext4_print_free_blocks(mpd->inode);
A
Aneesh Kumar K.V 已提交
2293
		}
2294
		/* invalidate all the pages */
2295
		ext4_da_block_invalidatepages(mpd, next,
2296
				mpd->b_size >> mpd->inode->i_blkbits);
2297 2298
		return err;
	}
2299 2300 2301
	BUG_ON(blks == 0);

	new.b_size = (blks << mpd->inode->i_blkbits);
2302

2303 2304
	if (buffer_new(&new))
		__unmap_underlying_blocks(mpd->inode, &new);
2305

2306 2307 2308 2309
	/*
	 * If blocks are delayed marked, we need to
	 * put actual blocknr and drop delayed bit
	 */
2310 2311
	if ((mpd->b_state & (1 << BH_Delay)) ||
	    (mpd->b_state & (1 << BH_Unwritten)))
2312
		mpage_put_bnr_to_bhs(mpd, next, &new);
2313

2314 2315 2316 2317 2318 2319 2320
	if (ext4_should_order_data(mpd->inode)) {
		err = ext4_jbd2_file_inode(handle, mpd->inode);
		if (err)
			return err;
	}

	/*
2321
	 * Update on-disk size along with block allocation.
2322 2323 2324 2325 2326 2327 2328 2329 2330
	 */
	disksize = ((loff_t) next + blks) << mpd->inode->i_blkbits;
	if (disksize > i_size_read(mpd->inode))
		disksize = i_size_read(mpd->inode);
	if (disksize > EXT4_I(mpd->inode)->i_disksize) {
		ext4_update_i_disksize(mpd->inode, disksize);
		return ext4_mark_inode_dirty(handle, mpd->inode);
	}

2331
	return 0;
2332 2333
}

2334 2335
#define BH_FLAGS ((1 << BH_Uptodate) | (1 << BH_Mapped) | \
		(1 << BH_Delay) | (1 << BH_Unwritten))
2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346

/*
 * mpage_add_bh_to_extent - try to add one more block to extent of blocks
 *
 * @mpd->lbh - extent of blocks
 * @logical - logical number of the block in the file
 * @bh - bh of the block (used to access block's state)
 *
 * the function is used to collect contig. blocks in same state
 */
static void mpage_add_bh_to_extent(struct mpage_da_data *mpd,
2347 2348
				   sector_t logical, size_t b_size,
				   unsigned long b_state)
2349 2350
{
	sector_t next;
2351
	int nrblocks = mpd->b_size >> mpd->inode->i_blkbits;
2352

2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374
	/* check if thereserved journal credits might overflow */
	if (!(EXT4_I(mpd->inode)->i_flags & EXT4_EXTENTS_FL)) {
		if (nrblocks >= EXT4_MAX_TRANS_DATA) {
			/*
			 * With non-extent format we are limited by the journal
			 * credit available.  Total credit needed to insert
			 * nrblocks contiguous blocks is dependent on the
			 * nrblocks.  So limit nrblocks.
			 */
			goto flush_it;
		} else if ((nrblocks + (b_size >> mpd->inode->i_blkbits)) >
				EXT4_MAX_TRANS_DATA) {
			/*
			 * Adding the new buffer_head would make it cross the
			 * allowed limit for which we have journal credit
			 * reserved. So limit the new bh->b_size
			 */
			b_size = (EXT4_MAX_TRANS_DATA - nrblocks) <<
						mpd->inode->i_blkbits;
			/* we will do mpage_da_submit_io in the next loop */
		}
	}
2375 2376 2377
	/*
	 * First block in the extent
	 */
2378 2379 2380 2381
	if (mpd->b_size == 0) {
		mpd->b_blocknr = logical;
		mpd->b_size = b_size;
		mpd->b_state = b_state & BH_FLAGS;
2382 2383 2384
		return;
	}

2385
	next = mpd->b_blocknr + nrblocks;
2386 2387 2388
	/*
	 * Can we merge the block to our big extent?
	 */
2389 2390
	if (logical == next && (b_state & BH_FLAGS) == mpd->b_state) {
		mpd->b_size += b_size;
2391 2392 2393
		return;
	}

2394
flush_it:
2395 2396 2397 2398
	/*
	 * We couldn't merge the block to our extent, so we
	 * need to flush current  extent and start new one
	 */
2399 2400
	if (mpage_da_map_blocks(mpd) == 0)
		mpage_da_submit_io(mpd);
2401 2402
	mpd->io_done = 1;
	return;
2403 2404
}

2405
static int ext4_bh_delay_or_unwritten(handle_t *handle, struct buffer_head *bh)
2406
{
2407
	return (buffer_delay(bh) || buffer_unwritten(bh)) && buffer_dirty(bh);
2408 2409
}

2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423
/*
 * __mpage_da_writepage - finds extent of pages and blocks
 *
 * @page: page to consider
 * @wbc: not used, we just follow rules
 * @data: context
 *
 * The function finds extents of pages and scan them for all blocks.
 */
static int __mpage_da_writepage(struct page *page,
				struct writeback_control *wbc, void *data)
{
	struct mpage_da_data *mpd = data;
	struct inode *inode = mpd->inode;
2424
	struct buffer_head *bh, *head;
2425 2426
	sector_t logical;

2427 2428 2429 2430
	if (mpd->io_done) {
		/*
		 * Rest of the page in the page_vec
		 * redirty then and skip then. We will
2431
		 * try to write them again after
2432 2433 2434 2435 2436 2437
		 * starting a new transaction
		 */
		redirty_page_for_writepage(wbc, page);
		unlock_page(page);
		return MPAGE_DA_EXTENT_TAIL;
	}
2438 2439 2440 2441 2442 2443
	/*
	 * Can we merge this page to current extent?
	 */
	if (mpd->next_page != page->index) {
		/*
		 * Nope, we can't. So, we map non-allocated blocks
2444
		 * and start IO on them using writepage()
2445 2446
		 */
		if (mpd->next_page != mpd->first_page) {
2447 2448
			if (mpage_da_map_blocks(mpd) == 0)
				mpage_da_submit_io(mpd);
2449 2450 2451 2452 2453 2454 2455
			/*
			 * skip rest of the page in the page_vec
			 */
			mpd->io_done = 1;
			redirty_page_for_writepage(wbc, page);
			unlock_page(page);
			return MPAGE_DA_EXTENT_TAIL;
2456 2457 2458 2459 2460 2461 2462 2463 2464 2465
		}

		/*
		 * Start next extent of pages ...
		 */
		mpd->first_page = page->index;

		/*
		 * ... and blocks
		 */
2466 2467 2468
		mpd->b_size = 0;
		mpd->b_state = 0;
		mpd->b_blocknr = 0;
2469 2470 2471 2472 2473 2474 2475
	}

	mpd->next_page = page->index + 1;
	logical = (sector_t) page->index <<
		  (PAGE_CACHE_SHIFT - inode->i_blkbits);

	if (!page_has_buffers(page)) {
2476 2477
		mpage_add_bh_to_extent(mpd, logical, PAGE_CACHE_SIZE,
				       (1 << BH_Dirty) | (1 << BH_Uptodate));
2478 2479
		if (mpd->io_done)
			return MPAGE_DA_EXTENT_TAIL;
2480 2481 2482 2483 2484 2485 2486 2487
	} else {
		/*
		 * Page with regular buffer heads, just add all dirty ones
		 */
		head = page_buffers(page);
		bh = head;
		do {
			BUG_ON(buffer_locked(bh));
2488 2489 2490 2491
			/*
			 * We need to try to allocate
			 * unmapped blocks in the same page.
			 * Otherwise we won't make progress
2492
			 * with the page in ext4_writepage
2493
			 */
2494
			if (ext4_bh_delay_or_unwritten(NULL, bh)) {
2495 2496 2497
				mpage_add_bh_to_extent(mpd, logical,
						       bh->b_size,
						       bh->b_state);
2498 2499
				if (mpd->io_done)
					return MPAGE_DA_EXTENT_TAIL;
2500 2501 2502 2503 2504 2505 2506 2507 2508
			} else if (buffer_dirty(bh) && (buffer_mapped(bh))) {
				/*
				 * mapped dirty buffer. We need to update
				 * the b_state because we look at
				 * b_state in mpage_da_map_blocks. We don't
				 * update b_size because if we find an
				 * unmapped buffer_head later we need to
				 * use the b_state flag of that buffer_head.
				 */
2509 2510
				if (mpd->b_size == 0)
					mpd->b_state = bh->b_state & BH_FLAGS;
2511
			}
2512 2513 2514 2515 2516 2517 2518 2519
			logical++;
		} while ((bh = bh->b_this_page) != head);
	}

	return 0;
}

/*
2520 2521 2522
 * This is a special get_blocks_t callback which is used by
 * ext4_da_write_begin().  It will either return mapped block or
 * reserve space for a single block.
2523 2524 2525 2526 2527 2528 2529
 *
 * For delayed buffer_head we have BH_Mapped, BH_New, BH_Delay set.
 * We also have b_blocknr = -1 and b_bdev initialized properly
 *
 * For unwritten buffer_head we have BH_Mapped, BH_New, BH_Unwritten set.
 * We also have b_blocknr = physicalblock mapping unwritten extent and b_bdev
 * initialized properly.
2530 2531 2532 2533 2534
 */
static int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
				  struct buffer_head *bh_result, int create)
{
	int ret = 0;
2535 2536 2537 2538
	sector_t invalid_block = ~((sector_t) 0xffff);

	if (invalid_block < ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es))
		invalid_block = ~0;
2539 2540 2541 2542 2543 2544 2545 2546 2547

	BUG_ON(create == 0);
	BUG_ON(bh_result->b_size != inode->i_sb->s_blocksize);

	/*
	 * first, we need to know whether the block is allocated already
	 * preallocated blocks are unmapped but should treated
	 * the same as allocated blocks.
	 */
2548
	ret = ext4_get_blocks(NULL, inode, iblock, 1,  bh_result, 0);
2549 2550
	if ((ret == 0) && !buffer_delay(bh_result)) {
		/* the block isn't (pre)allocated yet, let's reserve space */
2551 2552 2553 2554
		/*
		 * XXX: __block_prepare_write() unmaps passed block,
		 * is it OK?
		 */
2555
		ret = ext4_da_reserve_space(inode, iblock);
2556 2557 2558 2559
		if (ret)
			/* not enough space to reserve */
			return ret;

2560
		map_bh(bh_result, inode->i_sb, invalid_block);
2561 2562 2563 2564
		set_buffer_new(bh_result);
		set_buffer_delay(bh_result);
	} else if (ret > 0) {
		bh_result->b_size = (ret << inode->i_blkbits);
2565 2566 2567 2568 2569 2570 2571 2572
		if (buffer_unwritten(bh_result)) {
			/* A delayed write to unwritten bh should
			 * be marked new and mapped.  Mapped ensures
			 * that we don't do get_block multiple times
			 * when we write to the same offset and new
			 * ensures that we do proper zero out for
			 * partial write.
			 */
2573
			set_buffer_new(bh_result);
2574 2575
			set_buffer_mapped(bh_result);
		}
2576 2577 2578 2579 2580
		ret = 0;
	}

	return ret;
}
2581

2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598
/*
 * This function is used as a standard get_block_t calback function
 * when there is no desire to allocate any blocks.  It is used as a
 * callback function for block_prepare_write(), nobh_writepage(), and
 * block_write_full_page().  These functions should only try to map a
 * single block at a time.
 *
 * Since this function doesn't do block allocations even if the caller
 * requests it by passing in create=1, it is critically important that
 * any caller checks to make sure that any buffer heads are returned
 * by this function are either all already mapped or marked for
 * delayed allocation before calling nobh_writepage() or
 * block_write_full_page().  Otherwise, b_blocknr could be left
 * unitialized, and the page write functions will be taken by
 * surprise.
 */
static int noalloc_get_block_write(struct inode *inode, sector_t iblock,
2599 2600 2601 2602 2603
				   struct buffer_head *bh_result, int create)
{
	int ret = 0;
	unsigned max_blocks = bh_result->b_size >> inode->i_blkbits;

2604 2605
	BUG_ON(bh_result->b_size != inode->i_sb->s_blocksize);

2606 2607 2608 2609
	/*
	 * we don't want to do block allocation in writepage
	 * so call get_block_wrap with create = 0
	 */
2610
	ret = ext4_get_blocks(NULL, inode, iblock, max_blocks, bh_result, 0);
2611 2612 2613 2614 2615
	if (ret > 0) {
		bh_result->b_size = (ret << inode->i_blkbits);
		ret = 0;
	}
	return ret;
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 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664
static int bget_one(handle_t *handle, struct buffer_head *bh)
{
	get_bh(bh);
	return 0;
}

static int bput_one(handle_t *handle, struct buffer_head *bh)
{
	put_bh(bh);
	return 0;
}

static int __ext4_journalled_writepage(struct page *page,
				       unsigned int len)
{
	struct address_space *mapping = page->mapping;
	struct inode *inode = mapping->host;
	struct buffer_head *page_bufs;
	handle_t *handle = NULL;
	int ret = 0;
	int err;

	page_bufs = page_buffers(page);
	BUG_ON(!page_bufs);
	walk_page_buffers(handle, page_bufs, 0, len, NULL, bget_one);
	/* As soon as we unlock the page, it can go away, but we have
	 * references to buffers so we are safe */
	unlock_page(page);

	handle = ext4_journal_start(inode, ext4_writepage_trans_blocks(inode));
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		goto out;
	}

	ret = walk_page_buffers(handle, page_bufs, 0, len, NULL,
				do_journal_get_write_access);

	err = walk_page_buffers(handle, page_bufs, 0, len, NULL,
				write_end_fn);
	if (ret == 0)
		ret = err;
	err = ext4_journal_stop(handle);
	if (!ret)
		ret = err;

	walk_page_buffers(handle, page_bufs, 0, len, NULL, bput_one);
2665
	ext4_set_inode_state(inode, EXT4_STATE_JDATA);
2666 2667 2668 2669
out:
	return ret;
}

2670 2671 2672
static int ext4_set_bh_endio(struct buffer_head *bh, struct inode *inode);
static void ext4_end_io_buffer_write(struct buffer_head *bh, int uptodate);

2673
/*
2674 2675 2676 2677 2678 2679 2680 2681 2682
 * Note that we don't need to start a transaction unless we're journaling data
 * because we should have holes filled from ext4_page_mkwrite(). We even don't
 * need to file the inode to the transaction's list in ordered mode because if
 * we are writing back data added by write(), the inode is already there and if
 * we are writing back data modified via mmap(), noone guarantees in which
 * transaction the data will hit the disk. In case we are journaling data, we
 * cannot start transaction directly because transaction start ranks above page
 * lock so we have to do some magic.
 *
2683 2684 2685 2686 2687
 * This function can get called via...
 *   - ext4_da_writepages after taking page lock (have journal handle)
 *   - journal_submit_inode_data_buffers (no journal handle)
 *   - shrink_page_list via pdflush (no journal handle)
 *   - grab_page_cache when doing write_begin (have journal handle)
2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712
 *
 * We don't do any block allocation in this function. If we have page with
 * multiple blocks we need to write those buffer_heads that are mapped. This
 * is important for mmaped based write. So if we do with blocksize 1K
 * truncate(f, 1024);
 * a = mmap(f, 0, 4096);
 * a[0] = 'a';
 * truncate(f, 4096);
 * we have in the page first buffer_head mapped via page_mkwrite call back
 * but other bufer_heads would be unmapped but dirty(dirty done via the
 * do_wp_page). So writepage should write the first block. If we modify
 * the mmap area beyond 1024 we will again get a page_fault and the
 * page_mkwrite callback will do the block allocation and mark the
 * buffer_heads mapped.
 *
 * We redirty the page if we have any buffer_heads that is either delay or
 * unwritten in the page.
 *
 * We can get recursively called as show below.
 *
 *	ext4_writepage() -> kmalloc() -> __alloc_pages() -> page_launder() ->
 *		ext4_writepage()
 *
 * But since we don't do any block allocation we should not deadlock.
 * Page also have the dirty flag cleared so we don't get recurive page_lock.
2713
 */
2714
static int ext4_writepage(struct page *page,
2715
			  struct writeback_control *wbc)
2716 2717
{
	int ret = 0;
2718
	loff_t size;
2719
	unsigned int len;
2720
	struct buffer_head *page_bufs = NULL;
2721 2722
	struct inode *inode = page->mapping->host;

2723
	trace_ext4_writepage(inode, page);
2724 2725 2726 2727 2728
	size = i_size_read(inode);
	if (page->index == size >> PAGE_CACHE_SHIFT)
		len = size & ~PAGE_CACHE_MASK;
	else
		len = PAGE_CACHE_SIZE;
2729

2730
	if (page_has_buffers(page)) {
2731
		page_bufs = page_buffers(page);
2732
		if (walk_page_buffers(NULL, page_bufs, 0, len, NULL,
2733
					ext4_bh_delay_or_unwritten)) {
2734
			/*
2735 2736
			 * We don't want to do  block allocation
			 * So redirty the page and return
2737 2738 2739
			 * We may reach here when we do a journal commit
			 * via journal_submit_inode_data_buffers.
			 * If we don't have mapping block we just ignore
2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759
			 * them. We can also reach here via shrink_page_list
			 */
			redirty_page_for_writepage(wbc, page);
			unlock_page(page);
			return 0;
		}
	} else {
		/*
		 * The test for page_has_buffers() is subtle:
		 * We know the page is dirty but it lost buffers. That means
		 * that at some moment in time after write_begin()/write_end()
		 * has been called all buffers have been clean and thus they
		 * must have been written at least once. So they are all
		 * mapped and we can happily proceed with mapping them
		 * and writing the page.
		 *
		 * Try to initialize the buffer_heads and check whether
		 * all are mapped and non delay. We don't want to
		 * do block allocation here.
		 */
2760
		ret = block_prepare_write(page, 0, len,
2761
					  noalloc_get_block_write);
2762 2763 2764 2765
		if (!ret) {
			page_bufs = page_buffers(page);
			/* check whether all are mapped and non delay */
			if (walk_page_buffers(NULL, page_bufs, 0, len, NULL,
2766
						ext4_bh_delay_or_unwritten)) {
2767 2768 2769 2770 2771 2772 2773 2774 2775
				redirty_page_for_writepage(wbc, page);
				unlock_page(page);
				return 0;
			}
		} else {
			/*
			 * We can't do block allocation here
			 * so just redity the page and unlock
			 * and return
2776 2777 2778 2779 2780
			 */
			redirty_page_for_writepage(wbc, page);
			unlock_page(page);
			return 0;
		}
2781
		/* now mark the buffer_heads as dirty and uptodate */
2782
		block_commit_write(page, 0, len);
2783 2784
	}

2785 2786 2787 2788 2789 2790
	if (PageChecked(page) && ext4_should_journal_data(inode)) {
		/*
		 * It's mmapped pagecache.  Add buffers and journal it.  There
		 * doesn't seem much point in redirtying the page here.
		 */
		ClearPageChecked(page);
2791
		return __ext4_journalled_writepage(page, len);
2792 2793
	}

2794
	if (test_opt(inode->i_sb, NOBH) && ext4_should_writeback_data(inode))
2795
		ret = nobh_writepage(page, noalloc_get_block_write, wbc);
2796 2797 2798 2799 2800
	else if (page_bufs && buffer_uninit(page_bufs)) {
		ext4_set_bh_endio(page_bufs, inode);
		ret = block_write_full_page_endio(page, noalloc_get_block_write,
					    wbc, ext4_end_io_buffer_write);
	} else
2801 2802
		ret = block_write_full_page(page, noalloc_get_block_write,
					    wbc);
2803 2804 2805 2806

	return ret;
}

2807
/*
2808 2809 2810 2811 2812
 * This is called via ext4_da_writepages() to
 * calulate the total number of credits to reserve to fit
 * a single extent allocation into a single transaction,
 * ext4_da_writpeages() will loop calling this before
 * the block allocation.
2813
 */
2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824

static int ext4_da_writepages_trans_blocks(struct inode *inode)
{
	int max_blocks = EXT4_I(inode)->i_reserved_data_blocks;

	/*
	 * With non-extent format the journal credit needed to
	 * insert nrblocks contiguous block is dependent on
	 * number of contiguous block. So we will limit
	 * number of contiguous block to a sane value
	 */
2825
	if (!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL) &&
2826 2827 2828 2829 2830
	    (max_blocks > EXT4_MAX_TRANS_DATA))
		max_blocks = EXT4_MAX_TRANS_DATA;

	return ext4_chunk_trans_blocks(inode, max_blocks);
}
2831

2832
static int ext4_da_writepages(struct address_space *mapping,
2833
			      struct writeback_control *wbc)
2834
{
2835 2836
	pgoff_t	index;
	int range_whole = 0;
2837
	handle_t *handle = NULL;
2838
	struct mpage_da_data mpd;
2839
	struct inode *inode = mapping->host;
2840
	int no_nrwrite_index_update;
2841 2842
	int pages_written = 0;
	long pages_skipped;
2843
	unsigned int max_pages;
2844
	int range_cyclic, cycled = 1, io_done = 0;
2845 2846
	int needed_blocks, ret = 0;
	long desired_nr_to_write, nr_to_writebump = 0;
2847
	loff_t range_start = wbc->range_start;
2848
	struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb);
2849

2850
	trace_ext4_da_writepages(inode, wbc);
2851

2852 2853 2854 2855 2856
	/*
	 * No pages to write? This is mainly a kludge to avoid starting
	 * a transaction for special inodes like journal inode on last iput()
	 * because that could violate lock ordering on umount
	 */
2857
	if (!mapping->nrpages || !mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
2858
		return 0;
2859 2860 2861 2862 2863

	/*
	 * If the filesystem has aborted, it is read-only, so return
	 * right away instead of dumping stack traces later on that
	 * will obscure the real source of the problem.  We test
2864
	 * EXT4_MF_FS_ABORTED instead of sb->s_flag's MS_RDONLY because
2865 2866 2867 2868 2869
	 * the latter could be true if the filesystem is mounted
	 * read-only, and in that case, ext4_da_writepages should
	 * *never* be called, so if that ever happens, we would want
	 * the stack trace.
	 */
2870
	if (unlikely(sbi->s_mount_flags & EXT4_MF_FS_ABORTED))
2871 2872
		return -EROFS;

2873 2874
	if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
		range_whole = 1;
2875

2876 2877
	range_cyclic = wbc->range_cyclic;
	if (wbc->range_cyclic) {
2878
		index = mapping->writeback_index;
2879 2880 2881 2882 2883 2884
		if (index)
			cycled = 0;
		wbc->range_start = index << PAGE_CACHE_SHIFT;
		wbc->range_end  = LLONG_MAX;
		wbc->range_cyclic = 0;
	} else
2885
		index = wbc->range_start >> PAGE_CACHE_SHIFT;
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
	/*
	 * This works around two forms of stupidity.  The first is in
	 * the writeback code, which caps the maximum number of pages
	 * written to be 1024 pages.  This is wrong on multiple
	 * levels; different architectues have a different page size,
	 * which changes the maximum amount of data which gets
	 * written.  Secondly, 4 megabytes is way too small.  XFS
	 * forces this value to be 16 megabytes by multiplying
	 * nr_to_write parameter by four, and then relies on its
	 * allocator to allocate larger extents to make them
	 * contiguous.  Unfortunately this brings us to the second
	 * stupidity, which is that ext4's mballoc code only allocates
	 * at most 2048 blocks.  So we force contiguous writes up to
	 * the number of dirty blocks in the inode, or
	 * sbi->max_writeback_mb_bump whichever is smaller.
	 */
	max_pages = sbi->s_max_writeback_mb_bump << (20 - PAGE_CACHE_SHIFT);
	if (!range_cyclic && range_whole)
		desired_nr_to_write = wbc->nr_to_write * 8;
	else
		desired_nr_to_write = ext4_num_dirty_pages(inode, index,
							   max_pages);
	if (desired_nr_to_write > max_pages)
		desired_nr_to_write = max_pages;

	if (wbc->nr_to_write < desired_nr_to_write) {
		nr_to_writebump = desired_nr_to_write - wbc->nr_to_write;
		wbc->nr_to_write = desired_nr_to_write;
	}

2917 2918 2919
	mpd.wbc = wbc;
	mpd.inode = mapping->host;

2920 2921 2922 2923 2924 2925 2926 2927
	/*
	 * we don't want write_cache_pages to update
	 * nr_to_write and writeback_index
	 */
	no_nrwrite_index_update = wbc->no_nrwrite_index_update;
	wbc->no_nrwrite_index_update = 1;
	pages_skipped = wbc->pages_skipped;

2928
retry:
2929
	while (!ret && wbc->nr_to_write > 0) {
2930 2931 2932 2933 2934 2935 2936 2937

		/*
		 * we  insert one extent at a time. So we need
		 * credit needed for single extent allocation.
		 * journalled mode is currently not supported
		 * by delalloc
		 */
		BUG_ON(ext4_should_journal_data(inode));
2938
		needed_blocks = ext4_da_writepages_trans_blocks(inode);
2939

2940 2941 2942 2943
		/* start a new transaction*/
		handle = ext4_journal_start(inode, needed_blocks);
		if (IS_ERR(handle)) {
			ret = PTR_ERR(handle);
2944
			ext4_msg(inode->i_sb, KERN_CRIT, "%s: jbd2_start: "
2945 2946
			       "%ld pages, ino %lu; err %d\n", __func__,
				wbc->nr_to_write, inode->i_ino, ret);
2947 2948
			goto out_writepages;
		}
2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969

		/*
		 * Now call __mpage_da_writepage to find the next
		 * contiguous region of logical blocks that need
		 * blocks to be allocated by ext4.  We don't actually
		 * submit the blocks for I/O here, even though
		 * write_cache_pages thinks it will, and will set the
		 * pages as clean for write before calling
		 * __mpage_da_writepage().
		 */
		mpd.b_size = 0;
		mpd.b_state = 0;
		mpd.b_blocknr = 0;
		mpd.first_page = 0;
		mpd.next_page = 0;
		mpd.io_done = 0;
		mpd.pages_written = 0;
		mpd.retval = 0;
		ret = write_cache_pages(mapping, wbc, __mpage_da_writepage,
					&mpd);
		/*
2970
		 * If we have a contiguous extent of pages and we
2971 2972 2973 2974 2975 2976 2977 2978 2979
		 * haven't done the I/O yet, map the blocks and submit
		 * them for I/O.
		 */
		if (!mpd.io_done && mpd.next_page != mpd.first_page) {
			if (mpage_da_map_blocks(&mpd) == 0)
				mpage_da_submit_io(&mpd);
			mpd.io_done = 1;
			ret = MPAGE_DA_EXTENT_TAIL;
		}
2980
		trace_ext4_da_write_pages(inode, &mpd);
2981
		wbc->nr_to_write -= mpd.pages_written;
2982

2983
		ext4_journal_stop(handle);
2984

2985
		if ((mpd.retval == -ENOSPC) && sbi->s_journal) {
2986 2987 2988 2989
			/* commit the transaction which would
			 * free blocks released in the transaction
			 * and try again
			 */
2990
			jbd2_journal_force_commit_nested(sbi->s_journal);
2991 2992 2993
			wbc->pages_skipped = pages_skipped;
			ret = 0;
		} else if (ret == MPAGE_DA_EXTENT_TAIL) {
2994 2995 2996 2997
			/*
			 * got one extent now try with
			 * rest of the pages
			 */
2998 2999
			pages_written += mpd.pages_written;
			wbc->pages_skipped = pages_skipped;
3000
			ret = 0;
3001
			io_done = 1;
3002
		} else if (wbc->nr_to_write)
3003 3004 3005 3006 3007 3008
			/*
			 * There is no more writeout needed
			 * or we requested for a noblocking writeout
			 * and we found the device congested
			 */
			break;
3009
	}
3010 3011 3012 3013 3014 3015 3016
	if (!io_done && !cycled) {
		cycled = 1;
		index = 0;
		wbc->range_start = index << PAGE_CACHE_SHIFT;
		wbc->range_end  = mapping->writeback_index - 1;
		goto retry;
	}
3017
	if (pages_skipped != wbc->pages_skipped)
3018 3019 3020 3021
		ext4_msg(inode->i_sb, KERN_CRIT,
			 "This should not happen leaving %s "
			 "with nr_to_write = %ld ret = %d\n",
			 __func__, wbc->nr_to_write, ret);
3022 3023 3024

	/* Update index */
	index += pages_written;
3025
	wbc->range_cyclic = range_cyclic;
3026 3027 3028 3029 3030 3031
	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
		/*
		 * set the writeback_index so that range_cyclic
		 * mode will write it back later
		 */
		mapping->writeback_index = index;
3032

3033
out_writepages:
3034 3035
	if (!no_nrwrite_index_update)
		wbc->no_nrwrite_index_update = 0;
3036
	wbc->nr_to_write -= nr_to_writebump;
3037
	wbc->range_start = range_start;
3038
	trace_ext4_da_writepages_result(inode, wbc, ret, pages_written);
3039
	return ret;
3040 3041
}

3042 3043 3044 3045 3046 3047 3048 3049 3050
#define FALL_BACK_TO_NONDELALLOC 1
static int ext4_nonda_switch(struct super_block *sb)
{
	s64 free_blocks, dirty_blocks;
	struct ext4_sb_info *sbi = EXT4_SB(sb);

	/*
	 * switch to non delalloc mode if we are running low
	 * on free block. The free block accounting via percpu
3051
	 * counters can get slightly wrong with percpu_counter_batch getting
3052 3053 3054 3055 3056 3057 3058 3059 3060
	 * accumulated on each CPU without updating global counters
	 * Delalloc need an accurate free block accounting. So switch
	 * to non delalloc when we are near to error range.
	 */
	free_blocks  = percpu_counter_read_positive(&sbi->s_freeblocks_counter);
	dirty_blocks = percpu_counter_read_positive(&sbi->s_dirtyblocks_counter);
	if (2 * free_blocks < 3 * dirty_blocks ||
		free_blocks < (dirty_blocks + EXT4_FREEBLOCKS_WATERMARK)) {
		/*
3061 3062
		 * free block count is less than 150% of dirty blocks
		 * or free blocks is less than watermark
3063 3064 3065
		 */
		return 1;
	}
3066 3067 3068 3069 3070 3071 3072
	/*
	 * Even if we don't switch but are nearing capacity,
	 * start pushing delalloc when 1/2 of free blocks are dirty.
	 */
	if (free_blocks < 2 * dirty_blocks)
		writeback_inodes_sb_if_idle(sb);

3073 3074 3075
	return 0;
}

3076
static int ext4_da_write_begin(struct file *file, struct address_space *mapping,
3077 3078
			       loff_t pos, unsigned len, unsigned flags,
			       struct page **pagep, void **fsdata)
3079
{
3080
	int ret, retries = 0, quota_retries = 0;
3081 3082 3083 3084 3085 3086 3087 3088 3089
	struct page *page;
	pgoff_t index;
	unsigned from, to;
	struct inode *inode = mapping->host;
	handle_t *handle;

	index = pos >> PAGE_CACHE_SHIFT;
	from = pos & (PAGE_CACHE_SIZE - 1);
	to = from + len;
3090 3091 3092 3093 3094 3095 3096

	if (ext4_nonda_switch(inode->i_sb)) {
		*fsdata = (void *)FALL_BACK_TO_NONDELALLOC;
		return ext4_write_begin(file, mapping, pos,
					len, flags, pagep, fsdata);
	}
	*fsdata = (void *)0;
3097
	trace_ext4_da_write_begin(inode, pos, len, flags);
3098
retry:
3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109
	/*
	 * With delayed allocation, we don't log the i_disksize update
	 * if there is delayed block allocation. But we still need
	 * to journalling the i_disksize update if writes to the end
	 * of file which has an already mapped buffer.
	 */
	handle = ext4_journal_start(inode, 1);
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		goto out;
	}
3110 3111 3112
	/* We cannot recurse into the filesystem as the transaction is already
	 * started */
	flags |= AOP_FLAG_NOFS;
3113

3114
	page = grab_cache_page_write_begin(mapping, index, flags);
3115 3116 3117 3118 3119
	if (!page) {
		ext4_journal_stop(handle);
		ret = -ENOMEM;
		goto out;
	}
3120 3121 3122
	*pagep = page;

	ret = block_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
3123
				ext4_da_get_block_prep);
3124 3125 3126 3127
	if (ret < 0) {
		unlock_page(page);
		ext4_journal_stop(handle);
		page_cache_release(page);
3128 3129 3130 3131 3132 3133
		/*
		 * block_write_begin may have instantiated a few blocks
		 * outside i_size.  Trim these off again. Don't need
		 * i_size_read because we hold i_mutex.
		 */
		if (pos + len > inode->i_size)
3134
			ext4_truncate_failed_write(inode);
3135 3136
	}

3137 3138
	if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
		goto retry;
3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154

	if ((ret == -EDQUOT) &&
	    EXT4_I(inode)->i_reserved_meta_blocks &&
	    (quota_retries++ < 3)) {
		/*
		 * Since we often over-estimate the number of meta
		 * data blocks required, we may sometimes get a
		 * spurios out of quota error even though there would
		 * be enough space once we write the data blocks and
		 * find out how many meta data blocks were _really_
		 * required.  So try forcing the inode write to see if
		 * that helps.
		 */
		write_inode_now(inode, (quota_retries == 3));
		goto retry;
	}
3155 3156 3157 3158
out:
	return ret;
}

3159 3160 3161 3162 3163
/*
 * Check if we should update i_disksize
 * when write to the end of file but not require block allocation
 */
static int ext4_da_should_update_i_disksize(struct page *page,
3164
					    unsigned long offset)
3165 3166 3167 3168 3169 3170 3171 3172 3173
{
	struct buffer_head *bh;
	struct inode *inode = page->mapping->host;
	unsigned int idx;
	int i;

	bh = page_buffers(page);
	idx = offset >> inode->i_blkbits;

3174
	for (i = 0; i < idx; i++)
3175 3176
		bh = bh->b_this_page;

3177
	if (!buffer_mapped(bh) || (buffer_delay(bh)) || buffer_unwritten(bh))
3178 3179 3180 3181
		return 0;
	return 1;
}

3182
static int ext4_da_write_end(struct file *file,
3183 3184 3185
			     struct address_space *mapping,
			     loff_t pos, unsigned len, unsigned copied,
			     struct page *page, void *fsdata)
3186 3187 3188 3189 3190
{
	struct inode *inode = mapping->host;
	int ret = 0, ret2;
	handle_t *handle = ext4_journal_current_handle();
	loff_t new_i_size;
3191
	unsigned long start, end;
3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204
	int write_mode = (int)(unsigned long)fsdata;

	if (write_mode == FALL_BACK_TO_NONDELALLOC) {
		if (ext4_should_order_data(inode)) {
			return ext4_ordered_write_end(file, mapping, pos,
					len, copied, page, fsdata);
		} else if (ext4_should_writeback_data(inode)) {
			return ext4_writeback_write_end(file, mapping, pos,
					len, copied, page, fsdata);
		} else {
			BUG();
		}
	}
3205

3206
	trace_ext4_da_write_end(inode, pos, len, copied);
3207
	start = pos & (PAGE_CACHE_SIZE - 1);
3208
	end = start + copied - 1;
3209 3210 3211 3212 3213 3214 3215 3216

	/*
	 * generic_write_end() will run mark_inode_dirty() if i_size
	 * changes.  So let's piggyback the i_disksize mark_inode_dirty
	 * into that.
	 */

	new_i_size = pos + copied;
3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227
	if (new_i_size > EXT4_I(inode)->i_disksize) {
		if (ext4_da_should_update_i_disksize(page, end)) {
			down_write(&EXT4_I(inode)->i_data_sem);
			if (new_i_size > EXT4_I(inode)->i_disksize) {
				/*
				 * Updating i_disksize when extending file
				 * without needing block allocation
				 */
				if (ext4_should_order_data(inode))
					ret = ext4_jbd2_file_inode(handle,
								   inode);
3228

3229 3230 3231
				EXT4_I(inode)->i_disksize = new_i_size;
			}
			up_write(&EXT4_I(inode)->i_data_sem);
3232 3233 3234 3235 3236
			/* We need to mark inode dirty even if
			 * new_i_size is less that inode->i_size
			 * bu greater than i_disksize.(hint delalloc)
			 */
			ext4_mark_inode_dirty(handle, inode);
3237
		}
3238
	}
3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259
	ret2 = generic_write_end(file, mapping, pos, len, copied,
							page, fsdata);
	copied = ret2;
	if (ret2 < 0)
		ret = ret2;
	ret2 = ext4_journal_stop(handle);
	if (!ret)
		ret = ret2;

	return ret ? ret : copied;
}

static void ext4_da_invalidatepage(struct page *page, unsigned long offset)
{
	/*
	 * Drop reserved blocks
	 */
	BUG_ON(!PageLocked(page));
	if (!page_has_buffers(page))
		goto out;

3260
	ext4_da_page_release_reservation(page, offset);
3261 3262 3263 3264 3265 3266 3267

out:
	ext4_invalidatepage(page, offset);

	return;
}

3268 3269 3270 3271 3272
/*
 * Force all delayed allocation blocks to be allocated for a given inode.
 */
int ext4_alloc_da_blocks(struct inode *inode)
{
3273 3274
	trace_ext4_alloc_da_blocks(inode);

3275 3276 3277 3278 3279 3280 3281 3282 3283 3284
	if (!EXT4_I(inode)->i_reserved_data_blocks &&
	    !EXT4_I(inode)->i_reserved_meta_blocks)
		return 0;

	/*
	 * We do something simple for now.  The filemap_flush() will
	 * also start triggering a write of the data blocks, which is
	 * not strictly speaking necessary (and for users of
	 * laptop_mode, not even desirable).  However, to do otherwise
	 * would require replicating code paths in:
3285
	 *
3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304
	 * ext4_da_writepages() ->
	 *    write_cache_pages() ---> (via passed in callback function)
	 *        __mpage_da_writepage() -->
	 *           mpage_add_bh_to_extent()
	 *           mpage_da_map_blocks()
	 *
	 * The problem is that write_cache_pages(), located in
	 * mm/page-writeback.c, marks pages clean in preparation for
	 * doing I/O, which is not desirable if we're not planning on
	 * doing I/O at all.
	 *
	 * We could call write_cache_pages(), and then redirty all of
	 * the pages by calling redirty_page_for_writeback() but that
	 * would be ugly in the extreme.  So instead we would need to
	 * replicate parts of the code in the above functions,
	 * simplifying them becuase we wouldn't actually intend to
	 * write out the pages, but rather only collect contiguous
	 * logical block extents, call the multi-block allocator, and
	 * then update the buffer heads with the block allocations.
3305
	 *
3306 3307 3308 3309 3310 3311
	 * For now, though, we'll cheat by calling filemap_flush(),
	 * which will map the blocks, and start the I/O, but not
	 * actually wait for the I/O to complete.
	 */
	return filemap_flush(inode->i_mapping);
}
3312

3313 3314 3315 3316 3317
/*
 * bmap() is special.  It gets used by applications such as lilo and by
 * the swapper to find the on-disk block of a specific piece of data.
 *
 * Naturally, this is dangerous if the block concerned is still in the
3318
 * journal.  If somebody makes a swapfile on an ext4 data-journaling
3319 3320 3321 3322 3323 3324 3325 3326
 * filesystem and enables swap, then they may get a nasty shock when the
 * data getting swapped to that swapfile suddenly gets overwritten by
 * the original zero's written out previously to the journal and
 * awaiting writeback in the kernel's buffer cache.
 *
 * So, if we see any bmap calls here on a modified, data-journaled file,
 * take extra steps to flush any blocks which might be in the cache.
 */
3327
static sector_t ext4_bmap(struct address_space *mapping, sector_t block)
3328 3329 3330 3331 3332
{
	struct inode *inode = mapping->host;
	journal_t *journal;
	int err;

3333 3334 3335 3336 3337 3338 3339 3340 3341 3342
	if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
			test_opt(inode->i_sb, DELALLOC)) {
		/*
		 * With delalloc we want to sync the file
		 * so that we can make sure we allocate
		 * blocks for file
		 */
		filemap_write_and_wait(mapping);
	}

3343 3344
	if (EXT4_JOURNAL(inode) &&
	    ext4_test_inode_state(inode, EXT4_STATE_JDATA)) {
3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355
		/*
		 * This is a REALLY heavyweight approach, but the use of
		 * bmap on dirty files is expected to be extremely rare:
		 * only if we run lilo or swapon on a freshly made file
		 * do we expect this to happen.
		 *
		 * (bmap requires CAP_SYS_RAWIO so this does not
		 * represent an unprivileged user DOS attack --- we'd be
		 * in trouble if mortal users could trigger this path at
		 * will.)
		 *
3356
		 * NB. EXT4_STATE_JDATA is not set on files other than
3357 3358 3359 3360 3361 3362
		 * regular files.  If somebody wants to bmap a directory
		 * or symlink and gets confused because the buffer
		 * hasn't yet been flushed to disk, they deserve
		 * everything they get.
		 */

3363
		ext4_clear_inode_state(inode, EXT4_STATE_JDATA);
3364
		journal = EXT4_JOURNAL(inode);
3365 3366 3367
		jbd2_journal_lock_updates(journal);
		err = jbd2_journal_flush(journal);
		jbd2_journal_unlock_updates(journal);
3368 3369 3370 3371 3372

		if (err)
			return 0;
	}

3373
	return generic_block_bmap(mapping, block, ext4_get_block);
3374 3375
}

3376
static int ext4_readpage(struct file *file, struct page *page)
3377
{
3378
	return mpage_readpage(page, ext4_get_block);
3379 3380 3381
}

static int
3382
ext4_readpages(struct file *file, struct address_space *mapping,
3383 3384
		struct list_head *pages, unsigned nr_pages)
{
3385
	return mpage_readpages(mapping, pages, nr_pages, ext4_get_block);
3386 3387
}

3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416
static void ext4_free_io_end(ext4_io_end_t *io)
{
	BUG_ON(!io);
	if (io->page)
		put_page(io->page);
	iput(io->inode);
	kfree(io);
}

static void ext4_invalidatepage_free_endio(struct page *page, unsigned long offset)
{
	struct buffer_head *head, *bh;
	unsigned int curr_off = 0;

	if (!page_has_buffers(page))
		return;
	head = bh = page_buffers(page);
	do {
		if (offset <= curr_off && test_clear_buffer_uninit(bh)
					&& bh->b_private) {
			ext4_free_io_end(bh->b_private);
			bh->b_private = NULL;
			bh->b_end_io = NULL;
		}
		curr_off = curr_off + bh->b_size;
		bh = bh->b_this_page;
	} while (bh != head);
}

3417
static void ext4_invalidatepage(struct page *page, unsigned long offset)
3418
{
3419
	journal_t *journal = EXT4_JOURNAL(page->mapping->host);
3420

3421 3422 3423 3424 3425
	/*
	 * free any io_end structure allocated for buffers to be discarded
	 */
	if (ext4_should_dioread_nolock(page->mapping->host))
		ext4_invalidatepage_free_endio(page, offset);
3426 3427 3428 3429 3430 3431
	/*
	 * If it's a full truncate we just forget about the pending dirtying
	 */
	if (offset == 0)
		ClearPageChecked(page);

3432 3433 3434 3435
	if (journal)
		jbd2_journal_invalidatepage(journal, page, offset);
	else
		block_invalidatepage(page, offset);
3436 3437
}

3438
static int ext4_releasepage(struct page *page, gfp_t wait)
3439
{
3440
	journal_t *journal = EXT4_JOURNAL(page->mapping->host);
3441 3442 3443 3444

	WARN_ON(PageChecked(page));
	if (!page_has_buffers(page))
		return 0;
3445 3446 3447 3448
	if (journal)
		return jbd2_journal_try_to_free_buffers(journal, page, wait);
	else
		return try_to_free_buffers(page);
3449 3450 3451
}

/*
3452 3453
 * O_DIRECT for ext3 (or indirect map) based files
 *
3454 3455 3456 3457 3458
 * If the O_DIRECT write will extend the file then add this inode to the
 * orphan list.  So recovery will truncate it back to the original size
 * if the machine crashes during the write.
 *
 * If the O_DIRECT write is intantiating holes inside i_size and the machine
J
Jan Kara 已提交
3459 3460
 * crashes then stale disk data _may_ be exposed inside the file. But current
 * VFS code falls back into buffered path in that case so we are safe.
3461
 */
3462
static ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
3463 3464
			      const struct iovec *iov, loff_t offset,
			      unsigned long nr_segs)
3465 3466 3467
{
	struct file *file = iocb->ki_filp;
	struct inode *inode = file->f_mapping->host;
3468
	struct ext4_inode_info *ei = EXT4_I(inode);
J
Jan Kara 已提交
3469
	handle_t *handle;
3470 3471 3472
	ssize_t ret;
	int orphan = 0;
	size_t count = iov_length(iov, nr_segs);
3473
	int retries = 0;
3474 3475 3476 3477 3478

	if (rw == WRITE) {
		loff_t final_size = offset + count;

		if (final_size > inode->i_size) {
J
Jan Kara 已提交
3479 3480 3481 3482 3483 3484
			/* Credits for sb + inode write */
			handle = ext4_journal_start(inode, 2);
			if (IS_ERR(handle)) {
				ret = PTR_ERR(handle);
				goto out;
			}
3485
			ret = ext4_orphan_add(handle, inode);
J
Jan Kara 已提交
3486 3487 3488 3489
			if (ret) {
				ext4_journal_stop(handle);
				goto out;
			}
3490 3491
			orphan = 1;
			ei->i_disksize = inode->i_size;
J
Jan Kara 已提交
3492
			ext4_journal_stop(handle);
3493 3494 3495
		}
	}

3496
retry:
3497 3498 3499 3500 3501 3502 3503 3504
	if (rw == READ && ext4_should_dioread_nolock(inode))
		ret = blockdev_direct_IO_no_locking(rw, iocb, inode,
				 inode->i_sb->s_bdev, iov,
				 offset, nr_segs,
				 ext4_get_block, NULL);
	else
		ret = blockdev_direct_IO(rw, iocb, inode,
				 inode->i_sb->s_bdev, iov,
3505
				 offset, nr_segs,
3506
				 ext4_get_block, NULL);
3507 3508
	if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
		goto retry;
3509

J
Jan Kara 已提交
3510
	if (orphan) {
3511 3512
		int err;

J
Jan Kara 已提交
3513 3514 3515 3516 3517 3518 3519
		/* Credits for sb + inode write */
		handle = ext4_journal_start(inode, 2);
		if (IS_ERR(handle)) {
			/* This is really bad luck. We've written the data
			 * but cannot extend i_size. Bail out and pretend
			 * the write failed... */
			ret = PTR_ERR(handle);
3520 3521 3522
			if (inode->i_nlink)
				ext4_orphan_del(NULL, inode);

J
Jan Kara 已提交
3523 3524 3525
			goto out;
		}
		if (inode->i_nlink)
3526
			ext4_orphan_del(handle, inode);
J
Jan Kara 已提交
3527
		if (ret > 0) {
3528 3529 3530 3531 3532 3533 3534 3535
			loff_t end = offset + ret;
			if (end > inode->i_size) {
				ei->i_disksize = end;
				i_size_write(inode, end);
				/*
				 * We're going to return a positive `ret'
				 * here due to non-zero-length I/O, so there's
				 * no way of reporting error returns from
3536
				 * ext4_mark_inode_dirty() to userspace.  So
3537 3538
				 * ignore it.
				 */
3539
				ext4_mark_inode_dirty(handle, inode);
3540 3541
			}
		}
3542
		err = ext4_journal_stop(handle);
3543 3544 3545 3546 3547 3548 3549
		if (ret == 0)
			ret = err;
	}
out:
	return ret;
}

3550
static int ext4_get_block_write(struct inode *inode, sector_t iblock,
3551 3552
		   struct buffer_head *bh_result, int create)
{
3553
	handle_t *handle = ext4_journal_current_handle();
3554 3555 3556
	int ret = 0;
	unsigned max_blocks = bh_result->b_size >> inode->i_blkbits;
	int dio_credits;
3557
	int started = 0;
3558

3559
	ext4_debug("ext4_get_block_write: inode %lu, create flag %d\n",
3560
		   inode->i_ino, create);
3561
	/*
3562 3563 3564
	 * ext4_get_block in prepare for a DIO write or buffer write.
	 * We allocate an uinitialized extent if blocks haven't been allocated.
	 * The extent will be converted to initialized after IO complete.
3565
	 */
3566
	create = EXT4_GET_BLOCKS_IO_CREATE_EXT;
3567

3568 3569 3570 3571 3572 3573 3574 3575 3576 3577
	if (!handle) {
		if (max_blocks > DIO_MAX_BLOCKS)
			max_blocks = DIO_MAX_BLOCKS;
		dio_credits = ext4_chunk_trans_blocks(inode, max_blocks);
		handle = ext4_journal_start(inode, dio_credits);
		if (IS_ERR(handle)) {
			ret = PTR_ERR(handle);
			goto out;
		}
		started = 1;
3578
	}
3579

3580 3581 3582 3583 3584 3585
	ret = ext4_get_blocks(handle, inode, iblock, max_blocks, bh_result,
			      create);
	if (ret > 0) {
		bh_result->b_size = (ret << inode->i_blkbits);
		ret = 0;
	}
3586 3587
	if (started)
		ext4_journal_stop(handle);
3588 3589 3590 3591
out:
	return ret;
}

3592
static void dump_completed_IO(struct inode * inode)
3593 3594 3595 3596
{
#ifdef	EXT4_DEBUG
	struct list_head *cur, *before, *after;
	ext4_io_end_t *io, *io0, *io1;
3597
	unsigned long flags;
3598

3599 3600
	if (list_empty(&EXT4_I(inode)->i_completed_io_list)){
		ext4_debug("inode %lu completed_io list is empty\n", inode->i_ino);
3601 3602 3603
		return;
	}

3604
	ext4_debug("Dump inode %lu completed_io list \n", inode->i_ino);
3605
	spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags);
3606
	list_for_each_entry(io, &EXT4_I(inode)->i_completed_io_list, list){
3607 3608 3609 3610 3611 3612 3613 3614 3615
		cur = &io->list;
		before = cur->prev;
		io0 = container_of(before, ext4_io_end_t, list);
		after = cur->next;
		io1 = container_of(after, ext4_io_end_t, list);

		ext4_debug("io 0x%p from inode %lu,prev 0x%p,next 0x%p\n",
			    io, inode->i_ino, io0, io1);
	}
3616
	spin_unlock_irqrestore(&EXT4_I(inode)->i_completed_io_lock, flags);
3617 3618
#endif
}
3619 3620 3621 3622

/*
 * check a range of space and convert unwritten extents to written.
 */
3623
static int ext4_end_io_nolock(ext4_io_end_t *io)
3624 3625 3626
{
	struct inode *inode = io->inode;
	loff_t offset = io->offset;
3627
	ssize_t size = io->size;
3628 3629
	int ret = 0;

3630
	ext4_debug("ext4_end_io_nolock: io 0x%p from inode %lu,list->next 0x%p,"
3631 3632 3633 3634 3635 3636
		   "list->prev 0x%p\n",
	           io, inode->i_ino, io->list.next, io->list.prev);

	if (list_empty(&io->list))
		return ret;

3637
	if (io->flag != EXT4_IO_UNWRITTEN)
3638 3639
		return ret;

3640
	ret = ext4_convert_unwritten_extents(inode, offset, size);
3641
	if (ret < 0) {
3642
		printk(KERN_EMERG "%s: failed to convert unwritten"
3643 3644 3645 3646 3647
			"extents to written extents, error is %d"
			" io is still on inode %lu aio dio list\n",
                       __func__, ret, inode->i_ino);
		return ret;
	}
3648

3649 3650 3651
	/* clear the DIO AIO unwritten flag */
	io->flag = 0;
	return ret;
3652
}
3653

3654 3655 3656
/*
 * work on completed aio dio IO, to convert unwritten extents to extents
 */
3657
static void ext4_end_io_work(struct work_struct *work)
3658
{
3659 3660 3661 3662 3663
	ext4_io_end_t		*io = container_of(work, ext4_io_end_t, work);
	struct inode		*inode = io->inode;
	struct ext4_inode_info	*ei = EXT4_I(inode);
	unsigned long		flags;
	int			ret;
3664

3665
	mutex_lock(&inode->i_mutex);
3666
	ret = ext4_end_io_nolock(io);
3667 3668 3669
	if (ret < 0) {
		mutex_unlock(&inode->i_mutex);
		return;
3670
	}
3671 3672 3673 3674 3675

	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
	if (!list_empty(&io->list))
		list_del_init(&io->list);
	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
3676
	mutex_unlock(&inode->i_mutex);
3677
	ext4_free_io_end(io);
3678
}
3679

3680 3681 3682
/*
 * This function is called from ext4_sync_file().
 *
3683 3684
 * When IO is completed, the work to convert unwritten extents to
 * written is queued on workqueue but may not get immediately
3685 3686
 * scheduled. When fsync is called, we need to ensure the
 * conversion is complete before fsync returns.
3687 3688 3689 3690 3691
 * The inode keeps track of a list of pending/completed IO that
 * might needs to do the conversion. This function walks through
 * the list and convert the related unwritten extents for completed IO
 * to written.
 * The function return the number of pending IOs on success.
3692
 */
3693
int flush_completed_IO(struct inode *inode)
3694 3695
{
	ext4_io_end_t *io;
3696 3697
	struct ext4_inode_info *ei = EXT4_I(inode);
	unsigned long flags;
3698 3699 3700
	int ret = 0;
	int ret2 = 0;

3701
	if (list_empty(&ei->i_completed_io_list))
3702 3703
		return ret;

3704
	dump_completed_IO(inode);
3705 3706 3707
	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
	while (!list_empty(&ei->i_completed_io_list)){
		io = list_entry(ei->i_completed_io_list.next,
3708 3709
				ext4_io_end_t, list);
		/*
3710
		 * Calling ext4_end_io_nolock() to convert completed
3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722
		 * IO to written.
		 *
		 * When ext4_sync_file() is called, run_queue() may already
		 * about to flush the work corresponding to this io structure.
		 * It will be upset if it founds the io structure related
		 * to the work-to-be schedule is freed.
		 *
		 * Thus we need to keep the io structure still valid here after
		 * convertion finished. The io structure has a flag to
		 * avoid double converting from both fsync and background work
		 * queue work.
		 */
3723
		spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
3724
		ret = ext4_end_io_nolock(io);
3725
		spin_lock_irqsave(&ei->i_completed_io_lock, flags);
3726 3727 3728 3729 3730
		if (ret < 0)
			ret2 = ret;
		else
			list_del_init(&io->list);
	}
3731
	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
3732 3733 3734
	return (ret2 < 0) ? ret2 : 0;
}

3735
static ext4_io_end_t *ext4_init_io_end (struct inode *inode, gfp_t flags)
3736 3737 3738
{
	ext4_io_end_t *io = NULL;

3739
	io = kmalloc(sizeof(*io), flags);
3740 3741

	if (io) {
3742
		igrab(inode);
3743
		io->inode = inode;
3744
		io->flag = 0;
3745 3746
		io->offset = 0;
		io->size = 0;
3747
		io->page = NULL;
3748
		INIT_WORK(&io->work, ext4_end_io_work);
3749
		INIT_LIST_HEAD(&io->list);
3750 3751 3752 3753 3754 3755 3756 3757 3758 3759
	}

	return io;
}

static void ext4_end_io_dio(struct kiocb *iocb, loff_t offset,
			    ssize_t size, void *private)
{
        ext4_io_end_t *io_end = iocb->private;
	struct workqueue_struct *wq;
3760 3761
	unsigned long flags;
	struct ext4_inode_info *ei;
3762

3763 3764 3765 3766
	/* if not async direct IO or dio with 0 bytes write, just return */
	if (!io_end || !size)
		return;

3767 3768 3769 3770 3771 3772
	ext_debug("ext4_end_io_dio(): io_end 0x%p"
		  "for inode %lu, iocb 0x%p, offset %llu, size %llu\n",
 		  iocb->private, io_end->inode->i_ino, iocb, offset,
		  size);

	/* if not aio dio with unwritten extents, just free io and return */
3773
	if (io_end->flag != EXT4_IO_UNWRITTEN){
3774 3775
		ext4_free_io_end(io_end);
		iocb->private = NULL;
3776
		return;
3777 3778
	}

3779 3780
	io_end->offset = offset;
	io_end->size = size;
3781
	io_end->flag = EXT4_IO_UNWRITTEN;
3782 3783
	wq = EXT4_SB(io_end->inode->i_sb)->dio_unwritten_wq;

3784
	/* queue the work to convert unwritten extents to written */
3785 3786
	queue_work(wq, &io_end->work);

3787
	/* Add the io_end to per-inode completed aio dio list*/
3788 3789 3790 3791
	ei = EXT4_I(io_end->inode);
	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
	list_add_tail(&io_end->list, &ei->i_completed_io_list);
	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
3792 3793
	iocb->private = NULL;
}
3794

3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859
static void ext4_end_io_buffer_write(struct buffer_head *bh, int uptodate)
{
	ext4_io_end_t *io_end = bh->b_private;
	struct workqueue_struct *wq;
	struct inode *inode;
	unsigned long flags;

	if (!test_clear_buffer_uninit(bh) || !io_end)
		goto out;

	if (!(io_end->inode->i_sb->s_flags & MS_ACTIVE)) {
		printk("sb umounted, discard end_io request for inode %lu\n",
			io_end->inode->i_ino);
		ext4_free_io_end(io_end);
		goto out;
	}

	io_end->flag = EXT4_IO_UNWRITTEN;
	inode = io_end->inode;

	/* Add the io_end to per-inode completed io list*/
	spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags);
	list_add_tail(&io_end->list, &EXT4_I(inode)->i_completed_io_list);
	spin_unlock_irqrestore(&EXT4_I(inode)->i_completed_io_lock, flags);

	wq = EXT4_SB(inode->i_sb)->dio_unwritten_wq;
	/* queue the work to convert unwritten extents to written */
	queue_work(wq, &io_end->work);
out:
	bh->b_private = NULL;
	bh->b_end_io = NULL;
	clear_buffer_uninit(bh);
	end_buffer_async_write(bh, uptodate);
}

static int ext4_set_bh_endio(struct buffer_head *bh, struct inode *inode)
{
	ext4_io_end_t *io_end;
	struct page *page = bh->b_page;
	loff_t offset = (sector_t)page->index << PAGE_CACHE_SHIFT;
	size_t size = bh->b_size;

retry:
	io_end = ext4_init_io_end(inode, GFP_ATOMIC);
	if (!io_end) {
		if (printk_ratelimit())
			printk(KERN_WARNING "%s: allocation fail\n", __func__);
		schedule();
		goto retry;
	}
	io_end->offset = offset;
	io_end->size = size;
	/*
	 * We need to hold a reference to the page to make sure it
	 * doesn't get evicted before ext4_end_io_work() has a chance
	 * to convert the extent from written to unwritten.
	 */
	io_end->page = page;
	get_page(io_end->page);

	bh->b_private = io_end;
	bh->b_end_io = ext4_end_io_buffer_write;
	return 0;
}

3860 3861 3862 3863 3864 3865 3866 3867 3868
/*
 * For ext4 extent files, ext4 will do direct-io write to holes,
 * preallocated extents, and those write extend the file, no need to
 * fall back to buffered IO.
 *
 * For holes, we fallocate those blocks, mark them as unintialized
 * If those blocks were preallocated, we mark sure they are splited, but
 * still keep the range to write as unintialized.
 *
3869 3870 3871 3872
 * The unwrritten extents will be converted to written when DIO is completed.
 * For async direct IO, since the IO may still pending when return, we
 * set up an end_io call back function, which will do the convertion
 * when async direct IO completed.
3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890
 *
 * If the O_DIRECT write will extend the file then add this inode to the
 * orphan list.  So recovery will truncate it back to the original size
 * if the machine crashes during the write.
 *
 */
static ssize_t ext4_ext_direct_IO(int rw, struct kiocb *iocb,
			      const struct iovec *iov, loff_t offset,
			      unsigned long nr_segs)
{
	struct file *file = iocb->ki_filp;
	struct inode *inode = file->f_mapping->host;
	ssize_t ret;
	size_t count = iov_length(iov, nr_segs);

	loff_t final_size = offset + count;
	if (rw == WRITE && final_size <= inode->i_size) {
		/*
3891 3892 3893
 		 * We could direct write to holes and fallocate.
		 *
 		 * Allocated blocks to fill the hole are marked as uninitialized
3894 3895
 		 * to prevent paralel buffered read to expose the stale data
 		 * before DIO complete the data IO.
3896 3897
		 *
 		 * As to previously fallocated extents, ext4 get_block
3898 3899 3900
 		 * will just simply mark the buffer mapped but still
 		 * keep the extents uninitialized.
 		 *
3901 3902 3903 3904 3905 3906 3907 3908
		 * for non AIO case, we will convert those unwritten extents
		 * to written after return back from blockdev_direct_IO.
		 *
		 * for async DIO, the conversion needs to be defered when
		 * the IO is completed. The ext4 end_io callback function
		 * will be called to take care of the conversion work.
		 * Here for async case, we allocate an io_end structure to
		 * hook to the iocb.
3909
 		 */
3910 3911 3912
		iocb->private = NULL;
		EXT4_I(inode)->cur_aio_dio = NULL;
		if (!is_sync_kiocb(iocb)) {
3913
			iocb->private = ext4_init_io_end(inode, GFP_NOFS);
3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925
			if (!iocb->private)
				return -ENOMEM;
			/*
			 * we save the io structure for current async
			 * direct IO, so that later ext4_get_blocks()
			 * could flag the io structure whether there
			 * is a unwritten extents needs to be converted
			 * when IO is completed.
			 */
			EXT4_I(inode)->cur_aio_dio = iocb->private;
		}

3926 3927 3928
		ret = blockdev_direct_IO(rw, iocb, inode,
					 inode->i_sb->s_bdev, iov,
					 offset, nr_segs,
3929
					 ext4_get_block_write,
3930
					 ext4_end_io_dio);
3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949
		if (iocb->private)
			EXT4_I(inode)->cur_aio_dio = NULL;
		/*
		 * The io_end structure takes a reference to the inode,
		 * that structure needs to be destroyed and the
		 * reference to the inode need to be dropped, when IO is
		 * complete, even with 0 byte write, or failed.
		 *
		 * In the successful AIO DIO case, the io_end structure will be
		 * desctroyed and the reference to the inode will be dropped
		 * after the end_io call back function is called.
		 *
		 * In the case there is 0 byte write, or error case, since
		 * VFS direct IO won't invoke the end_io call back function,
		 * we need to free the end_io structure here.
		 */
		if (ret != -EIOCBQUEUED && ret <= 0 && iocb->private) {
			ext4_free_io_end(iocb->private);
			iocb->private = NULL;
3950 3951
		} else if (ret > 0 && ext4_test_inode_state(inode,
						EXT4_STATE_DIO_UNWRITTEN)) {
3952
			int err;
3953 3954 3955 3956
			/*
			 * for non AIO case, since the IO is already
			 * completed, we could do the convertion right here
			 */
3957 3958 3959 3960
			err = ext4_convert_unwritten_extents(inode,
							     offset, ret);
			if (err < 0)
				ret = err;
3961
			ext4_clear_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);
3962
		}
3963 3964
		return ret;
	}
3965 3966

	/* for write the the end of file case, we fall back to old way */
3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982
	return ext4_ind_direct_IO(rw, iocb, iov, offset, nr_segs);
}

static ssize_t ext4_direct_IO(int rw, struct kiocb *iocb,
			      const struct iovec *iov, loff_t offset,
			      unsigned long nr_segs)
{
	struct file *file = iocb->ki_filp;
	struct inode *inode = file->f_mapping->host;

	if (EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL)
		return ext4_ext_direct_IO(rw, iocb, iov, offset, nr_segs);

	return ext4_ind_direct_IO(rw, iocb, iov, offset, nr_segs);
}

3983
/*
3984
 * Pages can be marked dirty completely asynchronously from ext4's journalling
3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995
 * activity.  By filemap_sync_pte(), try_to_unmap_one(), etc.  We cannot do
 * much here because ->set_page_dirty is called under VFS locks.  The page is
 * not necessarily locked.
 *
 * We cannot just dirty the page and leave attached buffers clean, because the
 * buffers' dirty state is "definitive".  We cannot just set the buffers dirty
 * or jbddirty because all the journalling code will explode.
 *
 * So what we do is to mark the page "pending dirty" and next time writepage
 * is called, propagate that into the buffers appropriately.
 */
3996
static int ext4_journalled_set_page_dirty(struct page *page)
3997 3998 3999 4000 4001
{
	SetPageChecked(page);
	return __set_page_dirty_nobuffers(page);
}

4002
static const struct address_space_operations ext4_ordered_aops = {
4003 4004
	.readpage		= ext4_readpage,
	.readpages		= ext4_readpages,
4005
	.writepage		= ext4_writepage,
4006 4007 4008 4009 4010 4011 4012 4013 4014
	.sync_page		= block_sync_page,
	.write_begin		= ext4_write_begin,
	.write_end		= ext4_ordered_write_end,
	.bmap			= ext4_bmap,
	.invalidatepage		= ext4_invalidatepage,
	.releasepage		= ext4_releasepage,
	.direct_IO		= ext4_direct_IO,
	.migratepage		= buffer_migrate_page,
	.is_partially_uptodate  = block_is_partially_uptodate,
4015
	.error_remove_page	= generic_error_remove_page,
4016 4017
};

4018
static const struct address_space_operations ext4_writeback_aops = {
4019 4020
	.readpage		= ext4_readpage,
	.readpages		= ext4_readpages,
4021
	.writepage		= ext4_writepage,
4022 4023 4024 4025 4026 4027 4028 4029 4030
	.sync_page		= block_sync_page,
	.write_begin		= ext4_write_begin,
	.write_end		= ext4_writeback_write_end,
	.bmap			= ext4_bmap,
	.invalidatepage		= ext4_invalidatepage,
	.releasepage		= ext4_releasepage,
	.direct_IO		= ext4_direct_IO,
	.migratepage		= buffer_migrate_page,
	.is_partially_uptodate  = block_is_partially_uptodate,
4031
	.error_remove_page	= generic_error_remove_page,
4032 4033
};

4034
static const struct address_space_operations ext4_journalled_aops = {
4035 4036
	.readpage		= ext4_readpage,
	.readpages		= ext4_readpages,
4037
	.writepage		= ext4_writepage,
4038 4039 4040 4041 4042 4043 4044 4045
	.sync_page		= block_sync_page,
	.write_begin		= ext4_write_begin,
	.write_end		= ext4_journalled_write_end,
	.set_page_dirty		= ext4_journalled_set_page_dirty,
	.bmap			= ext4_bmap,
	.invalidatepage		= ext4_invalidatepage,
	.releasepage		= ext4_releasepage,
	.is_partially_uptodate  = block_is_partially_uptodate,
4046
	.error_remove_page	= generic_error_remove_page,
4047 4048
};

4049
static const struct address_space_operations ext4_da_aops = {
4050 4051
	.readpage		= ext4_readpage,
	.readpages		= ext4_readpages,
4052
	.writepage		= ext4_writepage,
4053 4054 4055 4056 4057 4058 4059 4060 4061 4062
	.writepages		= ext4_da_writepages,
	.sync_page		= block_sync_page,
	.write_begin		= ext4_da_write_begin,
	.write_end		= ext4_da_write_end,
	.bmap			= ext4_bmap,
	.invalidatepage		= ext4_da_invalidatepage,
	.releasepage		= ext4_releasepage,
	.direct_IO		= ext4_direct_IO,
	.migratepage		= buffer_migrate_page,
	.is_partially_uptodate  = block_is_partially_uptodate,
4063
	.error_remove_page	= generic_error_remove_page,
4064 4065
};

4066
void ext4_set_aops(struct inode *inode)
4067
{
4068 4069 4070 4071
	if (ext4_should_order_data(inode) &&
		test_opt(inode->i_sb, DELALLOC))
		inode->i_mapping->a_ops = &ext4_da_aops;
	else if (ext4_should_order_data(inode))
4072
		inode->i_mapping->a_ops = &ext4_ordered_aops;
4073 4074 4075
	else if (ext4_should_writeback_data(inode) &&
		 test_opt(inode->i_sb, DELALLOC))
		inode->i_mapping->a_ops = &ext4_da_aops;
4076 4077
	else if (ext4_should_writeback_data(inode))
		inode->i_mapping->a_ops = &ext4_writeback_aops;
4078
	else
4079
		inode->i_mapping->a_ops = &ext4_journalled_aops;
4080 4081 4082
}

/*
4083
 * ext4_block_truncate_page() zeroes out a mapping from file offset `from'
4084 4085 4086 4087
 * up to the end of the block which corresponds to `from'.
 * This required during truncate. We need to physically zero the tail end
 * of that block so it doesn't yield old data if the file is later grown.
 */
4088
int ext4_block_truncate_page(handle_t *handle,
4089 4090
		struct address_space *mapping, loff_t from)
{
4091
	ext4_fsblk_t index = from >> PAGE_CACHE_SHIFT;
4092
	unsigned offset = from & (PAGE_CACHE_SIZE-1);
A
Aneesh Kumar K.V 已提交
4093 4094
	unsigned blocksize, length, pos;
	ext4_lblk_t iblock;
4095 4096
	struct inode *inode = mapping->host;
	struct buffer_head *bh;
4097
	struct page *page;
4098 4099
	int err = 0;

4100 4101
	page = find_or_create_page(mapping, from >> PAGE_CACHE_SHIFT,
				   mapping_gfp_mask(mapping) & ~__GFP_FS);
4102 4103 4104
	if (!page)
		return -EINVAL;

4105 4106 4107 4108 4109 4110 4111 4112 4113
	blocksize = inode->i_sb->s_blocksize;
	length = blocksize - (offset & (blocksize - 1));
	iblock = index << (PAGE_CACHE_SHIFT - inode->i_sb->s_blocksize_bits);

	/*
	 * For "nobh" option,  we can only work if we don't need to
	 * read-in the page - otherwise we create buffers to do the IO.
	 */
	if (!page_has_buffers(page) && test_opt(inode->i_sb, NOBH) &&
4114
	     ext4_should_writeback_data(inode) && PageUptodate(page)) {
4115
		zero_user(page, offset, length);
4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139
		set_page_dirty(page);
		goto unlock;
	}

	if (!page_has_buffers(page))
		create_empty_buffers(page, blocksize, 0);

	/* Find the buffer that contains "offset" */
	bh = page_buffers(page);
	pos = blocksize;
	while (offset >= pos) {
		bh = bh->b_this_page;
		iblock++;
		pos += blocksize;
	}

	err = 0;
	if (buffer_freed(bh)) {
		BUFFER_TRACE(bh, "freed: skip");
		goto unlock;
	}

	if (!buffer_mapped(bh)) {
		BUFFER_TRACE(bh, "unmapped");
4140
		ext4_get_block(inode, iblock, bh, 0);
4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160
		/* unmapped? It's a hole - nothing to do */
		if (!buffer_mapped(bh)) {
			BUFFER_TRACE(bh, "still unmapped");
			goto unlock;
		}
	}

	/* Ok, it's mapped. Make sure it's up-to-date */
	if (PageUptodate(page))
		set_buffer_uptodate(bh);

	if (!buffer_uptodate(bh)) {
		err = -EIO;
		ll_rw_block(READ, 1, &bh);
		wait_on_buffer(bh);
		/* Uhhuh. Read error. Complain and punt. */
		if (!buffer_uptodate(bh))
			goto unlock;
	}

4161
	if (ext4_should_journal_data(inode)) {
4162
		BUFFER_TRACE(bh, "get write access");
4163
		err = ext4_journal_get_write_access(handle, bh);
4164 4165 4166 4167
		if (err)
			goto unlock;
	}

4168
	zero_user(page, offset, length);
4169 4170 4171 4172

	BUFFER_TRACE(bh, "zeroed end of block");

	err = 0;
4173
	if (ext4_should_journal_data(inode)) {
4174
		err = ext4_handle_dirty_metadata(handle, inode, bh);
4175
	} else {
4176
		if (ext4_should_order_data(inode))
4177
			err = ext4_jbd2_file_inode(handle, inode);
4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200
		mark_buffer_dirty(bh);
	}

unlock:
	unlock_page(page);
	page_cache_release(page);
	return err;
}

/*
 * Probably it should be a library function... search for first non-zero word
 * or memcmp with zero_page, whatever is better for particular architecture.
 * Linus?
 */
static inline int all_zeroes(__le32 *p, __le32 *q)
{
	while (p < q)
		if (*p++)
			return 0;
	return 1;
}

/**
4201
 *	ext4_find_shared - find the indirect blocks for partial truncation.
4202 4203
 *	@inode:	  inode in question
 *	@depth:	  depth of the affected branch
4204
 *	@offsets: offsets of pointers in that branch (see ext4_block_to_path)
4205 4206 4207
 *	@chain:	  place to store the pointers to partial indirect blocks
 *	@top:	  place to the (detached) top of branch
 *
4208
 *	This is a helper function used by ext4_truncate().
4209 4210 4211 4212 4213 4214 4215
 *
 *	When we do truncate() we may have to clean the ends of several
 *	indirect blocks but leave the blocks themselves alive. Block is
 *	partially truncated if some data below the new i_size is refered
 *	from it (and it is on the path to the first completely truncated
 *	data block, indeed).  We have to free the top of that path along
 *	with everything to the right of the path. Since no allocation
4216
 *	past the truncation point is possible until ext4_truncate()
4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234
 *	finishes, we may safely do the latter, but top of branch may
 *	require special attention - pageout below the truncation point
 *	might try to populate it.
 *
 *	We atomically detach the top of branch from the tree, store the
 *	block number of its root in *@top, pointers to buffer_heads of
 *	partially truncated blocks - in @chain[].bh and pointers to
 *	their last elements that should not be removed - in
 *	@chain[].p. Return value is the pointer to last filled element
 *	of @chain.
 *
 *	The work left to caller to do the actual freeing of subtrees:
 *		a) free the subtree starting from *@top
 *		b) free the subtrees whose roots are stored in
 *			(@chain[i].p+1 .. end of @chain[i].bh->b_data)
 *		c) free the subtrees growing from the inode past the @chain[0].
 *			(no partially truncated stuff there).  */

4235
static Indirect *ext4_find_shared(struct inode *inode, int depth,
4236 4237
				  ext4_lblk_t offsets[4], Indirect chain[4],
				  __le32 *top)
4238 4239 4240 4241 4242
{
	Indirect *partial, *p;
	int k, err;

	*top = 0;
4243
	/* Make k index the deepest non-null offset + 1 */
4244 4245
	for (k = depth; k > 1 && !offsets[k-1]; k--)
		;
4246
	partial = ext4_get_branch(inode, k, offsets, chain, &err);
4247 4248 4249 4250 4251 4252 4253 4254 4255 4256
	/* Writer: pointers */
	if (!partial)
		partial = chain + k-1;
	/*
	 * If the branch acquired continuation since we've looked at it -
	 * fine, it should all survive and (new) top doesn't belong to us.
	 */
	if (!partial->key && *partial->p)
		/* Writer: end */
		goto no_top;
4257
	for (p = partial; (p > chain) && all_zeroes((__le32 *) p->bh->b_data, p->p); p--)
4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268
		;
	/*
	 * OK, we've found the last block that must survive. The rest of our
	 * branch should be detached before unlocking. However, if that rest
	 * of branch is all ours and does not grow immediately from the inode
	 * it's easier to cheat and just decrement partial->p.
	 */
	if (p == chain + k - 1 && p > chain) {
		p->p--;
	} else {
		*top = *p->p;
4269
		/* Nope, don't do this in ext4.  Must leave the tree intact */
4270 4271 4272 4273 4274 4275
#if 0
		*p->p = 0;
#endif
	}
	/* Writer: end */

4276
	while (partial > p) {
4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291
		brelse(partial->bh);
		partial--;
	}
no_top:
	return partial;
}

/*
 * Zero a number of block pointers in either an inode or an indirect block.
 * If we restart the transaction we must again get write access to the
 * indirect block for further modification.
 *
 * We release `count' blocks on disk, but (last - first) may be greater
 * than `count' because there can be holes in there.
 */
4292 4293 4294 4295 4296
static int ext4_clear_blocks(handle_t *handle, struct inode *inode,
			     struct buffer_head *bh,
			     ext4_fsblk_t block_to_free,
			     unsigned long count, __le32 *first,
			     __le32 *last)
4297 4298
{
	__le32 *p;
4299
	int	flags = EXT4_FREE_BLOCKS_FORGET | EXT4_FREE_BLOCKS_VALIDATED;
4300 4301 4302

	if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
		flags |= EXT4_FREE_BLOCKS_METADATA;
4303

4304 4305
	if (!ext4_data_block_valid(EXT4_SB(inode->i_sb), block_to_free,
				   count)) {
4306
		ext4_error(inode->i_sb, "inode #%lu: "
4307 4308 4309 4310 4311 4312
			   "attempt to clear blocks %llu len %lu, invalid",
			   inode->i_ino, (unsigned long long) block_to_free,
			   count);
		return 1;
	}

4313 4314
	if (try_to_extend_transaction(handle, inode)) {
		if (bh) {
4315 4316
			BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
			ext4_handle_dirty_metadata(handle, inode, bh);
4317
		}
4318
		ext4_mark_inode_dirty(handle, inode);
4319 4320
		ext4_truncate_restart_trans(handle, inode,
					    blocks_for_truncate(inode));
4321 4322
		if (bh) {
			BUFFER_TRACE(bh, "retaking write access");
4323
			ext4_journal_get_write_access(handle, bh);
4324 4325 4326
		}
	}

4327 4328
	for (p = first; p < last; p++)
		*p = 0;
4329

4330
	ext4_free_blocks(handle, inode, 0, block_to_free, count, flags);
4331
	return 0;
4332 4333 4334
}

/**
4335
 * ext4_free_data - free a list of data blocks
4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352
 * @handle:	handle for this transaction
 * @inode:	inode we are dealing with
 * @this_bh:	indirect buffer_head which contains *@first and *@last
 * @first:	array of block numbers
 * @last:	points immediately past the end of array
 *
 * We are freeing all blocks refered from that array (numbers are stored as
 * little-endian 32-bit) and updating @inode->i_blocks appropriately.
 *
 * We accumulate contiguous runs of blocks to free.  Conveniently, if these
 * blocks are contiguous then releasing them at one time will only affect one
 * or two bitmap blocks (+ group descriptor(s) and superblock) and we won't
 * actually use a lot of journal space.
 *
 * @this_bh will be %NULL if @first and @last point into the inode's direct
 * block pointers.
 */
4353
static void ext4_free_data(handle_t *handle, struct inode *inode,
4354 4355 4356
			   struct buffer_head *this_bh,
			   __le32 *first, __le32 *last)
{
4357
	ext4_fsblk_t block_to_free = 0;    /* Starting block # of a run */
4358 4359 4360 4361
	unsigned long count = 0;	    /* Number of blocks in the run */
	__le32 *block_to_free_p = NULL;	    /* Pointer into inode/ind
					       corresponding to
					       block_to_free */
4362
	ext4_fsblk_t nr;		    /* Current block # */
4363 4364 4365 4366 4367 4368
	__le32 *p;			    /* Pointer into inode/ind
					       for current block */
	int err;

	if (this_bh) {				/* For indirect block */
		BUFFER_TRACE(this_bh, "get_write_access");
4369
		err = ext4_journal_get_write_access(handle, this_bh);
4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386
		/* Important: if we can't update the indirect pointers
		 * to the blocks, we can't free them. */
		if (err)
			return;
	}

	for (p = first; p < last; p++) {
		nr = le32_to_cpu(*p);
		if (nr) {
			/* accumulate blocks to free if they're contiguous */
			if (count == 0) {
				block_to_free = nr;
				block_to_free_p = p;
				count = 1;
			} else if (nr == block_to_free + count) {
				count++;
			} else {
4387 4388 4389 4390
				if (ext4_clear_blocks(handle, inode, this_bh,
						      block_to_free, count,
						      block_to_free_p, p))
					break;
4391 4392 4393 4394 4395 4396 4397 4398
				block_to_free = nr;
				block_to_free_p = p;
				count = 1;
			}
		}
	}

	if (count > 0)
4399
		ext4_clear_blocks(handle, inode, this_bh, block_to_free,
4400 4401 4402
				  count, block_to_free_p, p);

	if (this_bh) {
4403
		BUFFER_TRACE(this_bh, "call ext4_handle_dirty_metadata");
4404 4405 4406 4407 4408 4409 4410

		/*
		 * The buffer head should have an attached journal head at this
		 * point. However, if the data is corrupted and an indirect
		 * block pointed to itself, it would have been detached when
		 * the block was cleared. Check for this instead of OOPSing.
		 */
4411
		if ((EXT4_JOURNAL(inode) == NULL) || bh2jh(this_bh))
4412
			ext4_handle_dirty_metadata(handle, inode, this_bh);
4413
		else
4414
			ext4_error(inode->i_sb,
4415 4416 4417 4418
				   "circular indirect block detected, "
				   "inode=%lu, block=%llu",
				   inode->i_ino,
				   (unsigned long long) this_bh->b_blocknr);
4419 4420 4421 4422
	}
}

/**
4423
 *	ext4_free_branches - free an array of branches
4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434
 *	@handle: JBD handle for this transaction
 *	@inode:	inode we are dealing with
 *	@parent_bh: the buffer_head which contains *@first and *@last
 *	@first:	array of block numbers
 *	@last:	pointer immediately past the end of array
 *	@depth:	depth of the branches to free
 *
 *	We are freeing all blocks refered from these branches (numbers are
 *	stored as little-endian 32-bit) and updating @inode->i_blocks
 *	appropriately.
 */
4435
static void ext4_free_branches(handle_t *handle, struct inode *inode,
4436 4437 4438
			       struct buffer_head *parent_bh,
			       __le32 *first, __le32 *last, int depth)
{
4439
	ext4_fsblk_t nr;
4440 4441
	__le32 *p;

4442
	if (ext4_handle_is_aborted(handle))
4443 4444 4445 4446
		return;

	if (depth--) {
		struct buffer_head *bh;
4447
		int addr_per_block = EXT4_ADDR_PER_BLOCK(inode->i_sb);
4448 4449 4450 4451 4452 4453
		p = last;
		while (--p >= first) {
			nr = le32_to_cpu(*p);
			if (!nr)
				continue;		/* A hole */

4454 4455
			if (!ext4_data_block_valid(EXT4_SB(inode->i_sb),
						   nr, 1)) {
4456
				ext4_error(inode->i_sb,
4457 4458 4459 4460 4461 4462 4463
					   "indirect mapped block in inode "
					   "#%lu invalid (level %d, blk #%lu)",
					   inode->i_ino, depth,
					   (unsigned long) nr);
				break;
			}

4464 4465 4466 4467 4468 4469 4470 4471
			/* Go read the buffer for the next level down */
			bh = sb_bread(inode->i_sb, nr);

			/*
			 * A read failure? Report error and clear slot
			 * (should be rare).
			 */
			if (!bh) {
4472
				ext4_error(inode->i_sb,
4473
					   "Read failure, inode=%lu, block=%llu",
4474 4475 4476 4477 4478 4479
					   inode->i_ino, nr);
				continue;
			}

			/* This zaps the entire block.  Bottom up. */
			BUFFER_TRACE(bh, "free child branches");
4480
			ext4_free_branches(handle, inode, bh,
4481 4482 4483
					(__le32 *) bh->b_data,
					(__le32 *) bh->b_data + addr_per_block,
					depth);
4484 4485 4486 4487 4488

			/*
			 * We've probably journalled the indirect block several
			 * times during the truncate.  But it's no longer
			 * needed and we now drop it from the transaction via
4489
			 * jbd2_journal_revoke().
4490 4491 4492
			 *
			 * That's easy if it's exclusively part of this
			 * transaction.  But if it's part of the committing
4493
			 * transaction then jbd2_journal_forget() will simply
4494
			 * brelse() it.  That means that if the underlying
4495
			 * block is reallocated in ext4_get_block(),
4496 4497 4498 4499 4500 4501 4502 4503
			 * unmap_underlying_metadata() will find this block
			 * and will try to get rid of it.  damn, damn.
			 *
			 * If this block has already been committed to the
			 * journal, a revoke record will be written.  And
			 * revoke records must be emitted *before* clearing
			 * this block's bit in the bitmaps.
			 */
4504
			ext4_forget(handle, 1, inode, bh, bh->b_blocknr);
4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521

			/*
			 * Everything below this this pointer has been
			 * released.  Now let this top-of-subtree go.
			 *
			 * We want the freeing of this indirect block to be
			 * atomic in the journal with the updating of the
			 * bitmap block which owns it.  So make some room in
			 * the journal.
			 *
			 * We zero the parent pointer *after* freeing its
			 * pointee in the bitmaps, so if extend_transaction()
			 * for some reason fails to put the bitmap changes and
			 * the release into the same transaction, recovery
			 * will merely complain about releasing a free block,
			 * rather than leaking blocks.
			 */
4522
			if (ext4_handle_is_aborted(handle))
4523 4524
				return;
			if (try_to_extend_transaction(handle, inode)) {
4525
				ext4_mark_inode_dirty(handle, inode);
4526 4527
				ext4_truncate_restart_trans(handle, inode,
					    blocks_for_truncate(inode));
4528 4529
			}

4530 4531
			ext4_free_blocks(handle, inode, 0, nr, 1,
					 EXT4_FREE_BLOCKS_METADATA);
4532 4533 4534 4535 4536 4537 4538

			if (parent_bh) {
				/*
				 * The block which we have just freed is
				 * pointed to by an indirect block: journal it
				 */
				BUFFER_TRACE(parent_bh, "get_write_access");
4539
				if (!ext4_journal_get_write_access(handle,
4540 4541 4542
								   parent_bh)){
					*p = 0;
					BUFFER_TRACE(parent_bh,
4543 4544 4545 4546
					"call ext4_handle_dirty_metadata");
					ext4_handle_dirty_metadata(handle,
								   inode,
								   parent_bh);
4547 4548 4549 4550 4551 4552
				}
			}
		}
	} else {
		/* We have reached the bottom of the tree. */
		BUFFER_TRACE(parent_bh, "free data blocks");
4553
		ext4_free_data(handle, inode, parent_bh, first, last);
4554 4555 4556
	}
}

4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569
int ext4_can_truncate(struct inode *inode)
{
	if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
		return 0;
	if (S_ISREG(inode->i_mode))
		return 1;
	if (S_ISDIR(inode->i_mode))
		return 1;
	if (S_ISLNK(inode->i_mode))
		return !ext4_inode_is_fast_symlink(inode);
	return 0;
}

4570
/*
4571
 * ext4_truncate()
4572
 *
4573 4574
 * We block out ext4_get_block() block instantiations across the entire
 * transaction, and VFS/VM ensures that ext4_truncate() cannot run
4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590
 * simultaneously on behalf of the same inode.
 *
 * As we work through the truncate and commmit bits of it to the journal there
 * is one core, guiding principle: the file's tree must always be consistent on
 * disk.  We must be able to restart the truncate after a crash.
 *
 * The file's tree may be transiently inconsistent in memory (although it
 * probably isn't), but whenever we close off and commit a journal transaction,
 * the contents of (the filesystem + the journal) must be consistent and
 * restartable.  It's pretty simple, really: bottom up, right to left (although
 * left-to-right works OK too).
 *
 * Note that at recovery time, journal replay occurs *before* the restart of
 * truncate against the orphan inode list.
 *
 * The committed inode has the new, desired i_size (which is the same as
4591
 * i_disksize in this case).  After a crash, ext4_orphan_cleanup() will see
4592
 * that this inode's truncate did not complete and it will again call
4593 4594
 * ext4_truncate() to have another go.  So there will be instantiated blocks
 * to the right of the truncation point in a crashed ext4 filesystem.  But
4595
 * that's fine - as long as they are linked from the inode, the post-crash
4596
 * ext4_truncate() run will find them and release them.
4597
 */
4598
void ext4_truncate(struct inode *inode)
4599 4600
{
	handle_t *handle;
4601
	struct ext4_inode_info *ei = EXT4_I(inode);
4602
	__le32 *i_data = ei->i_data;
4603
	int addr_per_block = EXT4_ADDR_PER_BLOCK(inode->i_sb);
4604
	struct address_space *mapping = inode->i_mapping;
A
Aneesh Kumar K.V 已提交
4605
	ext4_lblk_t offsets[4];
4606 4607 4608 4609
	Indirect chain[4];
	Indirect *partial;
	__le32 nr = 0;
	int n;
A
Aneesh Kumar K.V 已提交
4610
	ext4_lblk_t last_block;
4611 4612
	unsigned blocksize = inode->i_sb->s_blocksize;

4613
	if (!ext4_can_truncate(inode))
4614 4615
		return;

4616 4617
	EXT4_I(inode)->i_flags &= ~EXT4_EOFBLOCKS_FL;

4618
	if (inode->i_size == 0 && !test_opt(inode->i_sb, NO_AUTO_DA_ALLOC))
4619
		ext4_set_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE);
4620

A
Aneesh Kumar K.V 已提交
4621
	if (EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL) {
4622
		ext4_ext_truncate(inode);
A
Aneesh Kumar K.V 已提交
4623 4624
		return;
	}
A
Alex Tomas 已提交
4625

4626
	handle = start_transaction(inode);
4627
	if (IS_ERR(handle))
4628 4629 4630
		return;		/* AKPM: return what? */

	last_block = (inode->i_size + blocksize-1)
4631
					>> EXT4_BLOCK_SIZE_BITS(inode->i_sb);
4632

4633 4634 4635
	if (inode->i_size & (blocksize - 1))
		if (ext4_block_truncate_page(handle, mapping, inode->i_size))
			goto out_stop;
4636

4637
	n = ext4_block_to_path(inode, last_block, offsets, NULL);
4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649
	if (n == 0)
		goto out_stop;	/* error */

	/*
	 * OK.  This truncate is going to happen.  We add the inode to the
	 * orphan list, so that if this truncate spans multiple transactions,
	 * and we crash, we will resume the truncate when the filesystem
	 * recovers.  It also marks the inode dirty, to catch the new size.
	 *
	 * Implication: the file must always be in a sane, consistent
	 * truncatable state while each transaction commits.
	 */
4650
	if (ext4_orphan_add(handle, inode))
4651 4652
		goto out_stop;

4653 4654 4655 4656 4657
	/*
	 * From here we block out all ext4_get_block() callers who want to
	 * modify the block allocation tree.
	 */
	down_write(&ei->i_data_sem);
4658

4659
	ext4_discard_preallocations(inode);
4660

4661 4662 4663 4664 4665
	/*
	 * The orphan list entry will now protect us from any crash which
	 * occurs before the truncate completes, so it is now safe to propagate
	 * the new, shorter inode size (held for now in i_size) into the
	 * on-disk inode. We do this via i_disksize, which is the value which
4666
	 * ext4 *really* writes onto the disk inode.
4667 4668 4669 4670
	 */
	ei->i_disksize = inode->i_size;

	if (n == 1) {		/* direct blocks */
4671 4672
		ext4_free_data(handle, inode, NULL, i_data+offsets[0],
			       i_data + EXT4_NDIR_BLOCKS);
4673 4674 4675
		goto do_indirects;
	}

4676
	partial = ext4_find_shared(inode, n, offsets, chain, &nr);
4677 4678 4679 4680
	/* Kill the top of shared branch (not detached) */
	if (nr) {
		if (partial == chain) {
			/* Shared branch grows from the inode */
4681
			ext4_free_branches(handle, inode, NULL,
4682 4683 4684 4685 4686 4687 4688 4689 4690
					   &nr, &nr+1, (chain+n-1) - partial);
			*partial->p = 0;
			/*
			 * We mark the inode dirty prior to restart,
			 * and prior to stop.  No need for it here.
			 */
		} else {
			/* Shared branch grows from an indirect block */
			BUFFER_TRACE(partial->bh, "get_write_access");
4691
			ext4_free_branches(handle, inode, partial->bh,
4692 4693 4694 4695 4696 4697
					partial->p,
					partial->p+1, (chain+n-1) - partial);
		}
	}
	/* Clear the ends of indirect blocks on the shared branch */
	while (partial > chain) {
4698
		ext4_free_branches(handle, inode, partial->bh, partial->p + 1,
4699 4700 4701
				   (__le32*)partial->bh->b_data+addr_per_block,
				   (chain+n-1) - partial);
		BUFFER_TRACE(partial->bh, "call brelse");
4702
		brelse(partial->bh);
4703 4704 4705 4706 4707 4708
		partial--;
	}
do_indirects:
	/* Kill the remaining (whole) subtrees */
	switch (offsets[0]) {
	default:
4709
		nr = i_data[EXT4_IND_BLOCK];
4710
		if (nr) {
4711 4712
			ext4_free_branches(handle, inode, NULL, &nr, &nr+1, 1);
			i_data[EXT4_IND_BLOCK] = 0;
4713
		}
4714 4715
	case EXT4_IND_BLOCK:
		nr = i_data[EXT4_DIND_BLOCK];
4716
		if (nr) {
4717 4718
			ext4_free_branches(handle, inode, NULL, &nr, &nr+1, 2);
			i_data[EXT4_DIND_BLOCK] = 0;
4719
		}
4720 4721
	case EXT4_DIND_BLOCK:
		nr = i_data[EXT4_TIND_BLOCK];
4722
		if (nr) {
4723 4724
			ext4_free_branches(handle, inode, NULL, &nr, &nr+1, 3);
			i_data[EXT4_TIND_BLOCK] = 0;
4725
		}
4726
	case EXT4_TIND_BLOCK:
4727 4728 4729
		;
	}

4730
	up_write(&ei->i_data_sem);
K
Kalpak Shah 已提交
4731
	inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
4732
	ext4_mark_inode_dirty(handle, inode);
4733 4734 4735 4736 4737 4738

	/*
	 * In a multi-transaction truncate, we only make the final transaction
	 * synchronous
	 */
	if (IS_SYNC(inode))
4739
		ext4_handle_sync(handle);
4740 4741 4742 4743 4744
out_stop:
	/*
	 * If this was a simple ftruncate(), and the file will remain alive
	 * 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
4745
	 * ext4_delete_inode(), and we allow that function to clean up the
4746 4747 4748
	 * orphan info for us.
	 */
	if (inode->i_nlink)
4749
		ext4_orphan_del(handle, inode);
4750

4751
	ext4_journal_stop(handle);
4752 4753 4754
}

/*
4755
 * ext4_get_inode_loc returns with an extra refcount against the inode's
4756 4757 4758 4759
 * underlying buffer_head on success. If 'in_mem' is true, we have all
 * data in memory that is needed to recreate the on-disk version of this
 * inode.
 */
4760 4761
static int __ext4_get_inode_loc(struct inode *inode,
				struct ext4_iloc *iloc, int in_mem)
4762
{
4763 4764 4765 4766 4767 4768
	struct ext4_group_desc	*gdp;
	struct buffer_head	*bh;
	struct super_block	*sb = inode->i_sb;
	ext4_fsblk_t		block;
	int			inodes_per_block, inode_offset;

A
Aneesh Kumar K.V 已提交
4769
	iloc->bh = NULL;
4770 4771
	if (!ext4_valid_inum(sb, inode->i_ino))
		return -EIO;
4772

4773 4774 4775
	iloc->block_group = (inode->i_ino - 1) / EXT4_INODES_PER_GROUP(sb);
	gdp = ext4_get_group_desc(sb, iloc->block_group, NULL);
	if (!gdp)
4776 4777
		return -EIO;

4778 4779 4780 4781 4782 4783 4784 4785 4786 4787
	/*
	 * Figure out the offset within the block group inode table
	 */
	inodes_per_block = (EXT4_BLOCK_SIZE(sb) / EXT4_INODE_SIZE(sb));
	inode_offset = ((inode->i_ino - 1) %
			EXT4_INODES_PER_GROUP(sb));
	block = ext4_inode_table(sb, gdp) + (inode_offset / inodes_per_block);
	iloc->offset = (inode_offset % inodes_per_block) * EXT4_INODE_SIZE(sb);

	bh = sb_getblk(sb, block);
4788
	if (!bh) {
4789 4790
		ext4_error(sb, "unable to read inode block - "
			   "inode=%lu, block=%llu", inode->i_ino, block);
4791 4792 4793 4794
		return -EIO;
	}
	if (!buffer_uptodate(bh)) {
		lock_buffer(bh);
4795 4796 4797 4798 4799 4800 4801 4802 4803 4804

		/*
		 * If the buffer has the write error flag, we have failed
		 * to write out another inode in the same block.  In this
		 * case, we don't have to read the block because we may
		 * read the old inode data successfully.
		 */
		if (buffer_write_io_error(bh) && !buffer_uptodate(bh))
			set_buffer_uptodate(bh);

4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817
		if (buffer_uptodate(bh)) {
			/* someone brought it uptodate while we waited */
			unlock_buffer(bh);
			goto has_buffer;
		}

		/*
		 * If we have all information of the inode in memory and this
		 * is the only valid inode in the block, we need not read the
		 * block.
		 */
		if (in_mem) {
			struct buffer_head *bitmap_bh;
4818
			int i, start;
4819

4820
			start = inode_offset & ~(inodes_per_block - 1);
4821

4822 4823
			/* Is the inode bitmap in cache? */
			bitmap_bh = sb_getblk(sb, ext4_inode_bitmap(sb, gdp));
4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835
			if (!bitmap_bh)
				goto make_io;

			/*
			 * If the inode bitmap isn't in cache then the
			 * optimisation may end up performing two reads instead
			 * of one, so skip it.
			 */
			if (!buffer_uptodate(bitmap_bh)) {
				brelse(bitmap_bh);
				goto make_io;
			}
4836
			for (i = start; i < start + inodes_per_block; i++) {
4837 4838
				if (i == inode_offset)
					continue;
4839
				if (ext4_test_bit(i, bitmap_bh->b_data))
4840 4841 4842
					break;
			}
			brelse(bitmap_bh);
4843
			if (i == start + inodes_per_block) {
4844 4845 4846 4847 4848 4849 4850 4851 4852
				/* all other inodes are free, so skip I/O */
				memset(bh->b_data, 0, bh->b_size);
				set_buffer_uptodate(bh);
				unlock_buffer(bh);
				goto has_buffer;
			}
		}

make_io:
4853 4854 4855 4856 4857 4858 4859 4860 4861
		/*
		 * If we need to do any I/O, try to pre-readahead extra
		 * blocks from the inode table.
		 */
		if (EXT4_SB(sb)->s_inode_readahead_blks) {
			ext4_fsblk_t b, end, table;
			unsigned num;

			table = ext4_inode_table(sb, gdp);
T
Theodore Ts'o 已提交
4862
			/* s_inode_readahead_blks is always a power of 2 */
4863 4864 4865 4866 4867 4868 4869
			b = block & ~(EXT4_SB(sb)->s_inode_readahead_blks-1);
			if (table > b)
				b = table;
			end = b + EXT4_SB(sb)->s_inode_readahead_blks;
			num = EXT4_INODES_PER_GROUP(sb);
			if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
				       EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
4870
				num -= ext4_itable_unused_count(sb, gdp);
4871 4872 4873 4874 4875 4876 4877
			table += num / inodes_per_block;
			if (end > table)
				end = table;
			while (b <= end)
				sb_breadahead(sb, b++);
		}

4878 4879 4880 4881 4882 4883 4884 4885 4886 4887
		/*
		 * There are other valid inodes in the buffer, this inode
		 * has in-inode xattrs, or we don't have this inode in memory.
		 * Read the block from disk.
		 */
		get_bh(bh);
		bh->b_end_io = end_buffer_read_sync;
		submit_bh(READ_META, bh);
		wait_on_buffer(bh);
		if (!buffer_uptodate(bh)) {
4888 4889
			ext4_error(sb, "unable to read inode block - inode=%lu,"
				   " block=%llu", inode->i_ino, block);
4890 4891 4892 4893 4894 4895 4896 4897 4898
			brelse(bh);
			return -EIO;
		}
	}
has_buffer:
	iloc->bh = bh;
	return 0;
}

4899
int ext4_get_inode_loc(struct inode *inode, struct ext4_iloc *iloc)
4900 4901
{
	/* We have all inode data except xattrs in memory here. */
4902
	return __ext4_get_inode_loc(inode, iloc,
4903
		!ext4_test_inode_state(inode, EXT4_STATE_XATTR));
4904 4905
}

4906
void ext4_set_inode_flags(struct inode *inode)
4907
{
4908
	unsigned int flags = EXT4_I(inode)->i_flags;
4909 4910

	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
4911
	if (flags & EXT4_SYNC_FL)
4912
		inode->i_flags |= S_SYNC;
4913
	if (flags & EXT4_APPEND_FL)
4914
		inode->i_flags |= S_APPEND;
4915
	if (flags & EXT4_IMMUTABLE_FL)
4916
		inode->i_flags |= S_IMMUTABLE;
4917
	if (flags & EXT4_NOATIME_FL)
4918
		inode->i_flags |= S_NOATIME;
4919
	if (flags & EXT4_DIRSYNC_FL)
4920 4921 4922
		inode->i_flags |= S_DIRSYNC;
}

4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940
/* Propagate flags from i_flags to EXT4_I(inode)->i_flags */
void ext4_get_inode_flags(struct ext4_inode_info *ei)
{
	unsigned int flags = ei->vfs_inode.i_flags;

	ei->i_flags &= ~(EXT4_SYNC_FL|EXT4_APPEND_FL|
			EXT4_IMMUTABLE_FL|EXT4_NOATIME_FL|EXT4_DIRSYNC_FL);
	if (flags & S_SYNC)
		ei->i_flags |= EXT4_SYNC_FL;
	if (flags & S_APPEND)
		ei->i_flags |= EXT4_APPEND_FL;
	if (flags & S_IMMUTABLE)
		ei->i_flags |= EXT4_IMMUTABLE_FL;
	if (flags & S_NOATIME)
		ei->i_flags |= EXT4_NOATIME_FL;
	if (flags & S_DIRSYNC)
		ei->i_flags |= EXT4_DIRSYNC_FL;
}
4941

4942
static blkcnt_t ext4_inode_blocks(struct ext4_inode *raw_inode,
4943
				  struct ext4_inode_info *ei)
4944 4945
{
	blkcnt_t i_blocks ;
A
Aneesh Kumar K.V 已提交
4946 4947
	struct inode *inode = &(ei->vfs_inode);
	struct super_block *sb = inode->i_sb;
4948 4949 4950 4951 4952 4953

	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
				EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
		/* we are using combined 48 bit field */
		i_blocks = ((u64)le16_to_cpu(raw_inode->i_blocks_high)) << 32 |
					le32_to_cpu(raw_inode->i_blocks_lo);
A
Aneesh Kumar K.V 已提交
4954 4955 4956 4957 4958 4959
		if (ei->i_flags & EXT4_HUGE_FILE_FL) {
			/* i_blocks represent file system block size */
			return i_blocks  << (inode->i_blkbits - 9);
		} else {
			return i_blocks;
		}
4960 4961 4962 4963
	} else {
		return le32_to_cpu(raw_inode->i_blocks_lo);
	}
}
4964

4965
struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
4966
{
4967 4968
	struct ext4_iloc iloc;
	struct ext4_inode *raw_inode;
4969 4970
	struct ext4_inode_info *ei;
	struct inode *inode;
4971
	journal_t *journal = EXT4_SB(sb)->s_journal;
4972
	long ret;
4973 4974
	int block;

4975 4976 4977 4978 4979 4980 4981
	inode = iget_locked(sb, ino);
	if (!inode)
		return ERR_PTR(-ENOMEM);
	if (!(inode->i_state & I_NEW))
		return inode;

	ei = EXT4_I(inode);
4982
	iloc.bh = 0;
4983

4984 4985
	ret = __ext4_get_inode_loc(inode, &iloc, 0);
	if (ret < 0)
4986
		goto bad_inode;
4987
	raw_inode = ext4_raw_inode(&iloc);
4988 4989 4990
	inode->i_mode = le16_to_cpu(raw_inode->i_mode);
	inode->i_uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low);
	inode->i_gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low);
4991
	if (!(test_opt(inode->i_sb, NO_UID32))) {
4992 4993 4994 4995 4996
		inode->i_uid |= le16_to_cpu(raw_inode->i_uid_high) << 16;
		inode->i_gid |= le16_to_cpu(raw_inode->i_gid_high) << 16;
	}
	inode->i_nlink = le16_to_cpu(raw_inode->i_links_count);

4997
	ei->i_state_flags = 0;
4998 4999 5000 5001 5002 5003 5004 5005 5006
	ei->i_dir_start_lookup = 0;
	ei->i_dtime = le32_to_cpu(raw_inode->i_dtime);
	/* We now have enough fields to check if the inode was active or not.
	 * This is needed because nfsd might try to access dead inodes
	 * the test is that same one that e2fsck uses
	 * NeilBrown 1999oct15
	 */
	if (inode->i_nlink == 0) {
		if (inode->i_mode == 0 ||
5007
		    !(EXT4_SB(inode->i_sb)->s_mount_state & EXT4_ORPHAN_FS)) {
5008
			/* this inode is deleted */
5009
			ret = -ESTALE;
5010 5011 5012 5013 5014 5015 5016 5017
			goto bad_inode;
		}
		/* The only unlinked inodes we let through here have
		 * valid i_mode and are being read by the orphan
		 * recovery code: that's fine, we're about to complete
		 * the process of deleting those. */
	}
	ei->i_flags = le32_to_cpu(raw_inode->i_flags);
5018
	inode->i_blocks = ext4_inode_blocks(raw_inode, ei);
5019
	ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl_lo);
5020
	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT))
B
Badari Pulavarty 已提交
5021 5022
		ei->i_file_acl |=
			((__u64)le16_to_cpu(raw_inode->i_file_acl_high)) << 32;
5023
	inode->i_size = ext4_isize(raw_inode);
5024
	ei->i_disksize = inode->i_size;
5025 5026 5027
#ifdef CONFIG_QUOTA
	ei->i_reserved_quota = 0;
#endif
5028 5029
	inode->i_generation = le32_to_cpu(raw_inode->i_generation);
	ei->i_block_group = iloc.block_group;
5030
	ei->i_last_alloc_group = ~0;
5031 5032 5033 5034
	/*
	 * NOTE! The in-memory inode i_data array is in little-endian order
	 * even on big-endian machines: we do NOT byteswap the block numbers!
	 */
5035
	for (block = 0; block < EXT4_N_BLOCKS; block++)
5036 5037 5038
		ei->i_data[block] = raw_inode->i_block[block];
	INIT_LIST_HEAD(&ei->i_orphan);

5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063
	/*
	 * Set transaction id's of transactions that have to be committed
	 * to finish f[data]sync. We set them to currently running transaction
	 * as we cannot be sure that the inode or some of its metadata isn't
	 * part of the transaction - the inode could have been reclaimed and
	 * now it is reread from disk.
	 */
	if (journal) {
		transaction_t *transaction;
		tid_t tid;

		spin_lock(&journal->j_state_lock);
		if (journal->j_running_transaction)
			transaction = journal->j_running_transaction;
		else
			transaction = journal->j_committing_transaction;
		if (transaction)
			tid = transaction->t_tid;
		else
			tid = journal->j_commit_sequence;
		spin_unlock(&journal->j_state_lock);
		ei->i_sync_tid = tid;
		ei->i_datasync_tid = tid;
	}

5064
	if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE) {
5065
		ei->i_extra_isize = le16_to_cpu(raw_inode->i_extra_isize);
5066
		if (EXT4_GOOD_OLD_INODE_SIZE + ei->i_extra_isize >
5067
		    EXT4_INODE_SIZE(inode->i_sb)) {
5068
			ret = -EIO;
5069
			goto bad_inode;
5070
		}
5071 5072
		if (ei->i_extra_isize == 0) {
			/* The extra space is currently unused. Use it. */
5073 5074
			ei->i_extra_isize = sizeof(struct ext4_inode) -
					    EXT4_GOOD_OLD_INODE_SIZE;
5075 5076
		} else {
			__le32 *magic = (void *)raw_inode +
5077
					EXT4_GOOD_OLD_INODE_SIZE +
5078
					ei->i_extra_isize;
5079
			if (*magic == cpu_to_le32(EXT4_XATTR_MAGIC))
5080
				ext4_set_inode_state(inode, EXT4_STATE_XATTR);
5081 5082 5083 5084
		}
	} else
		ei->i_extra_isize = 0;

K
Kalpak Shah 已提交
5085 5086 5087 5088 5089
	EXT4_INODE_GET_XTIME(i_ctime, inode, raw_inode);
	EXT4_INODE_GET_XTIME(i_mtime, inode, raw_inode);
	EXT4_INODE_GET_XTIME(i_atime, inode, raw_inode);
	EXT4_EINODE_GET_XTIME(i_crtime, ei, raw_inode);

5090 5091 5092 5093 5094 5095 5096
	inode->i_version = le32_to_cpu(raw_inode->i_disk_version);
	if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE) {
		if (EXT4_FITS_IN_INODE(raw_inode, ei, i_version_hi))
			inode->i_version |=
			(__u64)(le32_to_cpu(raw_inode->i_version_hi)) << 32;
	}

5097
	ret = 0;
5098
	if (ei->i_file_acl &&
5099
	    !ext4_data_block_valid(EXT4_SB(sb), ei->i_file_acl, 1)) {
5100
		ext4_error(sb, "bad extended attribute block %llu inode #%lu",
5101 5102 5103 5104
			   ei->i_file_acl, inode->i_ino);
		ret = -EIO;
		goto bad_inode;
	} else if (ei->i_flags & EXT4_EXTENTS_FL) {
5105 5106 5107 5108 5109
		if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
		    (S_ISLNK(inode->i_mode) &&
		     !ext4_inode_is_fast_symlink(inode)))
			/* Validate extent which is part of inode */
			ret = ext4_ext_check_inode(inode);
5110
	} else if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
5111 5112
		   (S_ISLNK(inode->i_mode) &&
		    !ext4_inode_is_fast_symlink(inode))) {
5113
		/* Validate block references which are part of inode */
5114 5115
		ret = ext4_check_inode_blockref(inode);
	}
5116
	if (ret)
5117
		goto bad_inode;
5118

5119
	if (S_ISREG(inode->i_mode)) {
5120 5121 5122
		inode->i_op = &ext4_file_inode_operations;
		inode->i_fop = &ext4_file_operations;
		ext4_set_aops(inode);
5123
	} else if (S_ISDIR(inode->i_mode)) {
5124 5125
		inode->i_op = &ext4_dir_inode_operations;
		inode->i_fop = &ext4_dir_operations;
5126
	} else if (S_ISLNK(inode->i_mode)) {
5127
		if (ext4_inode_is_fast_symlink(inode)) {
5128
			inode->i_op = &ext4_fast_symlink_inode_operations;
5129 5130 5131
			nd_terminate_link(ei->i_data, inode->i_size,
				sizeof(ei->i_data) - 1);
		} else {
5132 5133
			inode->i_op = &ext4_symlink_inode_operations;
			ext4_set_aops(inode);
5134
		}
5135 5136
	} else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
	      S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
5137
		inode->i_op = &ext4_special_inode_operations;
5138 5139 5140 5141 5142 5143
		if (raw_inode->i_block[0])
			init_special_inode(inode, inode->i_mode,
			   old_decode_dev(le32_to_cpu(raw_inode->i_block[0])));
		else
			init_special_inode(inode, inode->i_mode,
			   new_decode_dev(le32_to_cpu(raw_inode->i_block[1])));
5144 5145
	} else {
		ret = -EIO;
5146
		ext4_error(inode->i_sb, "bogus i_mode (%o) for inode=%lu",
5147 5148
			   inode->i_mode, inode->i_ino);
		goto bad_inode;
5149
	}
5150
	brelse(iloc.bh);
5151
	ext4_set_inode_flags(inode);
5152 5153
	unlock_new_inode(inode);
	return inode;
5154 5155

bad_inode:
5156
	brelse(iloc.bh);
5157 5158
	iget_failed(inode);
	return ERR_PTR(ret);
5159 5160
}

5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173
static int ext4_inode_blocks_set(handle_t *handle,
				struct ext4_inode *raw_inode,
				struct ext4_inode_info *ei)
{
	struct inode *inode = &(ei->vfs_inode);
	u64 i_blocks = inode->i_blocks;
	struct super_block *sb = inode->i_sb;

	if (i_blocks <= ~0U) {
		/*
		 * i_blocks can be represnted in a 32 bit variable
		 * as multiple of 512 bytes
		 */
A
Aneesh Kumar K.V 已提交
5174
		raw_inode->i_blocks_lo   = cpu_to_le32(i_blocks);
5175
		raw_inode->i_blocks_high = 0;
A
Aneesh Kumar K.V 已提交
5176
		ei->i_flags &= ~EXT4_HUGE_FILE_FL;
5177 5178 5179 5180 5181 5182
		return 0;
	}
	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE))
		return -EFBIG;

	if (i_blocks <= 0xffffffffffffULL) {
5183 5184 5185 5186
		/*
		 * i_blocks can be represented in a 48 bit variable
		 * as multiple of 512 bytes
		 */
A
Aneesh Kumar K.V 已提交
5187
		raw_inode->i_blocks_lo   = cpu_to_le32(i_blocks);
5188
		raw_inode->i_blocks_high = cpu_to_le16(i_blocks >> 32);
A
Aneesh Kumar K.V 已提交
5189
		ei->i_flags &= ~EXT4_HUGE_FILE_FL;
5190
	} else {
A
Aneesh Kumar K.V 已提交
5191 5192 5193 5194 5195
		ei->i_flags |= EXT4_HUGE_FILE_FL;
		/* i_block is stored in file system block size */
		i_blocks = i_blocks >> (inode->i_blkbits - 9);
		raw_inode->i_blocks_lo   = cpu_to_le32(i_blocks);
		raw_inode->i_blocks_high = cpu_to_le16(i_blocks >> 32);
5196
	}
5197
	return 0;
5198 5199
}

5200 5201 5202 5203 5204 5205 5206
/*
 * Post the struct inode info into an on-disk inode location in the
 * buffer-cache.  This gobbles the caller's reference to the
 * buffer_head in the inode location struct.
 *
 * The caller must have write access to iloc->bh.
 */
5207
static int ext4_do_update_inode(handle_t *handle,
5208
				struct inode *inode,
5209
				struct ext4_iloc *iloc)
5210
{
5211 5212
	struct ext4_inode *raw_inode = ext4_raw_inode(iloc);
	struct ext4_inode_info *ei = EXT4_I(inode);
5213 5214 5215 5216 5217
	struct buffer_head *bh = iloc->bh;
	int err = 0, rc, block;

	/* For fields not not tracking in the in-memory inode,
	 * initialise them to zero for new inodes. */
5218
	if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
5219
		memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
5220

5221
	ext4_get_inode_flags(ei);
5222
	raw_inode->i_mode = cpu_to_le16(inode->i_mode);
5223
	if (!(test_opt(inode->i_sb, NO_UID32))) {
5224 5225 5226 5227 5228 5229
		raw_inode->i_uid_low = cpu_to_le16(low_16_bits(inode->i_uid));
		raw_inode->i_gid_low = cpu_to_le16(low_16_bits(inode->i_gid));
/*
 * Fix up interoperability with old kernels. Otherwise, old inodes get
 * re-used with the upper 16 bits of the uid/gid intact
 */
5230
		if (!ei->i_dtime) {
5231 5232 5233 5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247
			raw_inode->i_uid_high =
				cpu_to_le16(high_16_bits(inode->i_uid));
			raw_inode->i_gid_high =
				cpu_to_le16(high_16_bits(inode->i_gid));
		} else {
			raw_inode->i_uid_high = 0;
			raw_inode->i_gid_high = 0;
		}
	} else {
		raw_inode->i_uid_low =
			cpu_to_le16(fs_high2lowuid(inode->i_uid));
		raw_inode->i_gid_low =
			cpu_to_le16(fs_high2lowgid(inode->i_gid));
		raw_inode->i_uid_high = 0;
		raw_inode->i_gid_high = 0;
	}
	raw_inode->i_links_count = cpu_to_le16(inode->i_nlink);
K
Kalpak Shah 已提交
5248 5249 5250 5251 5252 5253

	EXT4_INODE_SET_XTIME(i_ctime, inode, raw_inode);
	EXT4_INODE_SET_XTIME(i_mtime, inode, raw_inode);
	EXT4_INODE_SET_XTIME(i_atime, inode, raw_inode);
	EXT4_EINODE_SET_XTIME(i_crtime, ei, raw_inode);

5254 5255
	if (ext4_inode_blocks_set(handle, raw_inode, ei))
		goto out_brelse;
5256
	raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
5257
	raw_inode->i_flags = cpu_to_le32(ei->i_flags);
5258 5259
	if (EXT4_SB(inode->i_sb)->s_es->s_creator_os !=
	    cpu_to_le32(EXT4_OS_HURD))
B
Badari Pulavarty 已提交
5260 5261
		raw_inode->i_file_acl_high =
			cpu_to_le16(ei->i_file_acl >> 32);
5262
	raw_inode->i_file_acl_lo = cpu_to_le32(ei->i_file_acl);
5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278
	ext4_isize_set(raw_inode, ei->i_disksize);
	if (ei->i_disksize > 0x7fffffffULL) {
		struct super_block *sb = inode->i_sb;
		if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
				EXT4_FEATURE_RO_COMPAT_LARGE_FILE) ||
				EXT4_SB(sb)->s_es->s_rev_level ==
				cpu_to_le32(EXT4_GOOD_OLD_REV)) {
			/* If this is the first large file
			 * created, add a flag to the superblock.
			 */
			err = ext4_journal_get_write_access(handle,
					EXT4_SB(sb)->s_sbh);
			if (err)
				goto out_brelse;
			ext4_update_dynamic_rev(sb);
			EXT4_SET_RO_COMPAT_FEATURE(sb,
5279
					EXT4_FEATURE_RO_COMPAT_LARGE_FILE);
5280
			sb->s_dirt = 1;
5281
			ext4_handle_sync(handle);
5282
			err = ext4_handle_dirty_metadata(handle, NULL,
5283
					EXT4_SB(sb)->s_sbh);
5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297
		}
	}
	raw_inode->i_generation = cpu_to_le32(inode->i_generation);
	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
		if (old_valid_dev(inode->i_rdev)) {
			raw_inode->i_block[0] =
				cpu_to_le32(old_encode_dev(inode->i_rdev));
			raw_inode->i_block[1] = 0;
		} else {
			raw_inode->i_block[0] = 0;
			raw_inode->i_block[1] =
				cpu_to_le32(new_encode_dev(inode->i_rdev));
			raw_inode->i_block[2] = 0;
		}
5298 5299 5300
	} else
		for (block = 0; block < EXT4_N_BLOCKS; block++)
			raw_inode->i_block[block] = ei->i_data[block];
5301

5302 5303 5304 5305 5306
	raw_inode->i_disk_version = cpu_to_le32(inode->i_version);
	if (ei->i_extra_isize) {
		if (EXT4_FITS_IN_INODE(raw_inode, ei, i_version_hi))
			raw_inode->i_version_hi =
			cpu_to_le32(inode->i_version >> 32);
5307
		raw_inode->i_extra_isize = cpu_to_le16(ei->i_extra_isize);
5308 5309
	}

5310
	BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
5311
	rc = ext4_handle_dirty_metadata(handle, NULL, bh);
5312 5313
	if (!err)
		err = rc;
5314
	ext4_clear_inode_state(inode, EXT4_STATE_NEW);
5315

5316
	ext4_update_inode_fsync_trans(handle, inode, 0);
5317
out_brelse:
5318
	brelse(bh);
5319
	ext4_std_error(inode->i_sb, err);
5320 5321 5322 5323
	return err;
}

/*
5324
 * ext4_write_inode()
5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340
 *
 * We are called from a few places:
 *
 * - Within generic_file_write() for O_SYNC files.
 *   Here, there will be no transaction running. We wait for any running
 *   trasnaction to commit.
 *
 * - Within sys_sync(), kupdate and such.
 *   We wait on commit, if tol to.
 *
 * - Within prune_icache() (PF_MEMALLOC == true)
 *   Here we simply return.  We can't afford to block kswapd on the
 *   journal commit.
 *
 * In all cases it is actually safe for us to return without doing anything,
 * because the inode has been copied into a raw inode buffer in
5341
 * ext4_mark_inode_dirty().  This is a correctness thing for O_SYNC and for
5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357
 * knfsd.
 *
 * Note that we are absolutely dependent upon all inode dirtiers doing the
 * right thing: they *must* call mark_inode_dirty() after dirtying info in
 * which we are interested.
 *
 * It would be a bug for them to not do this.  The code:
 *
 *	mark_inode_dirty(inode)
 *	stuff();
 *	inode->i_size = expr;
 *
 * is in error because a kswapd-driven write_inode() could occur while
 * `stuff()' is running, and the new i_size will be lost.  Plus the inode
 * will no longer be on the superblock's dirty inode list.
 */
5358
int ext4_write_inode(struct inode *inode, struct writeback_control *wbc)
5359
{
5360 5361
	int err;

5362 5363 5364
	if (current->flags & PF_MEMALLOC)
		return 0;

5365 5366 5367 5368 5369 5370
	if (EXT4_SB(inode->i_sb)->s_journal) {
		if (ext4_journal_current_handle()) {
			jbd_debug(1, "called recursively, non-PF_MEMALLOC!\n");
			dump_stack();
			return -EIO;
		}
5371

5372
		if (wbc->sync_mode != WB_SYNC_ALL)
5373 5374 5375 5376 5377
			return 0;

		err = ext4_force_commit(inode->i_sb);
	} else {
		struct ext4_iloc iloc;
5378

5379
		err = __ext4_get_inode_loc(inode, &iloc, 0);
5380 5381
		if (err)
			return err;
5382
		if (wbc->sync_mode == WB_SYNC_ALL)
5383 5384
			sync_dirty_buffer(iloc.bh);
		if (buffer_req(iloc.bh) && !buffer_uptodate(iloc.bh)) {
5385 5386
			ext4_error(inode->i_sb, "IO error syncing inode, "
				   "inode=%lu, block=%llu", inode->i_ino,
5387 5388 5389
				   (unsigned long long)iloc.bh->b_blocknr);
			err = -EIO;
		}
5390
		brelse(iloc.bh);
5391 5392
	}
	return err;
5393 5394 5395
}

/*
5396
 * ext4_setattr()
5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409
 *
 * Called from notify_change.
 *
 * We want to trap VFS attempts to truncate the file as soon as
 * possible.  In particular, we want to make sure that when the VFS
 * shrinks i_size, we put the inode on the orphan list and modify
 * i_disksize immediately, so that during the subsequent flushing of
 * dirty pages and freeing of disk blocks, we can guarantee that any
 * commit will leave the blocks being flushed in an unused state on
 * disk.  (On recovery, the inode will get truncated and the blocks will
 * be freed, so we have a strong guarantee that no future commit will
 * leave these blocks visible to the user.)
 *
5410 5411 5412 5413 5414 5415 5416 5417
 * Another thing we have to assure is that if we are in ordered mode
 * and inode is still attached to the committing transaction, we must
 * we start writeout of all the dirty pages which are being truncated.
 * This way we are sure that all the data written in the previous
 * transaction are already on disk (truncate waits for pages under
 * writeback).
 *
 * Called with inode->i_mutex down.
5418
 */
5419
int ext4_setattr(struct dentry *dentry, struct iattr *attr)
5420 5421 5422 5423 5424 5425 5426 5427 5428
{
	struct inode *inode = dentry->d_inode;
	int error, rc = 0;
	const unsigned int ia_valid = attr->ia_valid;

	error = inode_change_ok(inode, attr);
	if (error)
		return error;

5429
	if (ia_valid & ATTR_SIZE)
5430
		dquot_initialize(inode);
5431 5432 5433 5434 5435 5436
	if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
		(ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid)) {
		handle_t *handle;

		/* (user+group)*(old+new) structure, inode write (sb,
		 * inode block, ? - but truncate inode update has it) */
D
Dmitry Monakhov 已提交
5437
		handle = ext4_journal_start(inode, (EXT4_MAXQUOTAS_INIT_BLOCKS(inode->i_sb)+
5438
					EXT4_MAXQUOTAS_DEL_BLOCKS(inode->i_sb))+3);
5439 5440 5441 5442
		if (IS_ERR(handle)) {
			error = PTR_ERR(handle);
			goto err_out;
		}
5443
		error = dquot_transfer(inode, attr);
5444
		if (error) {
5445
			ext4_journal_stop(handle);
5446 5447 5448 5449 5450 5451 5452 5453
			return error;
		}
		/* Update corresponding info in inode so that everything is in
		 * one transaction */
		if (attr->ia_valid & ATTR_UID)
			inode->i_uid = attr->ia_uid;
		if (attr->ia_valid & ATTR_GID)
			inode->i_gid = attr->ia_gid;
5454 5455
		error = ext4_mark_inode_dirty(handle, inode);
		ext4_journal_stop(handle);
5456 5457
	}

5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468
	if (attr->ia_valid & ATTR_SIZE) {
		if (!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL)) {
			struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);

			if (attr->ia_size > sbi->s_bitmap_maxbytes) {
				error = -EFBIG;
				goto err_out;
			}
		}
	}

5469
	if (S_ISREG(inode->i_mode) &&
5470 5471 5472
	    attr->ia_valid & ATTR_SIZE &&
	    (attr->ia_size < inode->i_size ||
	     (EXT4_I(inode)->i_flags & EXT4_EOFBLOCKS_FL))) {
5473 5474
		handle_t *handle;

5475
		handle = ext4_journal_start(inode, 3);
5476 5477 5478 5479 5480
		if (IS_ERR(handle)) {
			error = PTR_ERR(handle);
			goto err_out;
		}

5481 5482 5483
		error = ext4_orphan_add(handle, inode);
		EXT4_I(inode)->i_disksize = attr->ia_size;
		rc = ext4_mark_inode_dirty(handle, inode);
5484 5485
		if (!error)
			error = rc;
5486
		ext4_journal_stop(handle);
5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502

		if (ext4_should_order_data(inode)) {
			error = ext4_begin_ordered_truncate(inode,
							    attr->ia_size);
			if (error) {
				/* Do as much error cleanup as possible */
				handle = ext4_journal_start(inode, 3);
				if (IS_ERR(handle)) {
					ext4_orphan_del(NULL, inode);
					goto err_out;
				}
				ext4_orphan_del(handle, inode);
				ext4_journal_stop(handle);
				goto err_out;
			}
		}
5503 5504 5505
		/* ext4_truncate will clear the flag */
		if ((EXT4_I(inode)->i_flags & EXT4_EOFBLOCKS_FL))
			ext4_truncate(inode);
5506 5507 5508 5509
	}

	rc = inode_setattr(inode, attr);

5510
	/* If inode_setattr's call to ext4_truncate failed to get a
5511 5512 5513
	 * transaction handle at all, we need to clean up the in-core
	 * orphan list manually. */
	if (inode->i_nlink)
5514
		ext4_orphan_del(NULL, inode);
5515 5516

	if (!rc && (ia_valid & ATTR_MODE))
5517
		rc = ext4_acl_chmod(inode);
5518 5519

err_out:
5520
	ext4_std_error(inode->i_sb, error);
5521 5522 5523 5524 5525
	if (!error)
		error = rc;
	return error;
}

5526 5527 5528 5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551
int ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
		 struct kstat *stat)
{
	struct inode *inode;
	unsigned long delalloc_blocks;

	inode = dentry->d_inode;
	generic_fillattr(inode, stat);

	/*
	 * We can't update i_blocks if the block allocation is delayed
	 * otherwise in the case of system crash before the real block
	 * allocation is done, we will have i_blocks inconsistent with
	 * on-disk file blocks.
	 * We always keep i_blocks updated together with real
	 * allocation. But to not confuse with user, stat
	 * will return the blocks that include the delayed allocation
	 * blocks for this file.
	 */
	spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
	delalloc_blocks = EXT4_I(inode)->i_reserved_data_blocks;
	spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);

	stat->blocks += (delalloc_blocks << inode->i_sb->s_blocksize_bits)>>9;
	return 0;
}
5552

5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580
static int ext4_indirect_trans_blocks(struct inode *inode, int nrblocks,
				      int chunk)
{
	int indirects;

	/* if nrblocks are contiguous */
	if (chunk) {
		/*
		 * With N contiguous data blocks, it need at most
		 * N/EXT4_ADDR_PER_BLOCK(inode->i_sb) indirect blocks
		 * 2 dindirect blocks
		 * 1 tindirect block
		 */
		indirects = nrblocks / EXT4_ADDR_PER_BLOCK(inode->i_sb);
		return indirects + 3;
	}
	/*
	 * if nrblocks are not contiguous, worse case, each block touch
	 * a indirect block, and each indirect block touch a double indirect
	 * block, plus a triple indirect block
	 */
	indirects = nrblocks * 2 + 1;
	return indirects;
}

static int ext4_index_trans_blocks(struct inode *inode, int nrblocks, int chunk)
{
	if (!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL))
5581 5582
		return ext4_indirect_trans_blocks(inode, nrblocks, chunk);
	return ext4_ext_index_trans_blocks(inode, nrblocks, chunk);
5583
}
5584

5585
/*
5586 5587 5588
 * Account for index blocks, block groups bitmaps and block group
 * descriptor blocks if modify datablocks and index blocks
 * worse case, the indexs blocks spread over different block groups
5589
 *
5590
 * If datablocks are discontiguous, they are possible to spread over
5591
 * different block groups too. If they are contiuguous, with flexbg,
5592
 * they could still across block group boundary.
5593
 *
5594 5595 5596 5597
 * Also account for superblock, inode, quota and xattr blocks
 */
int ext4_meta_trans_blocks(struct inode *inode, int nrblocks, int chunk)
{
5598 5599
	ext4_group_t groups, ngroups = ext4_get_groups_count(inode->i_sb);
	int gdpblocks;
5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620 5621 5622 5623 5624 5625
	int idxblocks;
	int ret = 0;

	/*
	 * How many index blocks need to touch to modify nrblocks?
	 * The "Chunk" flag indicating whether the nrblocks is
	 * physically contiguous on disk
	 *
	 * For Direct IO and fallocate, they calls get_block to allocate
	 * one single extent at a time, so they could set the "Chunk" flag
	 */
	idxblocks = ext4_index_trans_blocks(inode, nrblocks, chunk);

	ret = idxblocks;

	/*
	 * Now let's see how many group bitmaps and group descriptors need
	 * to account
	 */
	groups = idxblocks;
	if (chunk)
		groups += 1;
	else
		groups += nrblocks;

	gdpblocks = groups;
5626 5627
	if (groups > ngroups)
		groups = ngroups;
5628 5629 5630 5631 5632 5633 5634 5635 5636 5637 5638 5639 5640 5641
	if (groups > EXT4_SB(inode->i_sb)->s_gdb_count)
		gdpblocks = EXT4_SB(inode->i_sb)->s_gdb_count;

	/* bitmaps and block group descriptor blocks */
	ret += groups + gdpblocks;

	/* Blocks for super block, inode, quota and xattr blocks */
	ret += EXT4_META_TRANS_BLOCKS(inode->i_sb);

	return ret;
}

/*
 * Calulate the total number of credits to reserve to fit
5642 5643
 * the modification of a single pages into a single transaction,
 * which may include multiple chunks of block allocations.
5644
 *
5645
 * This could be called via ext4_write_begin()
5646
 *
5647
 * We need to consider the worse case, when
5648
 * one new block per extent.
5649
 */
A
Alex Tomas 已提交
5650
int ext4_writepage_trans_blocks(struct inode *inode)
5651
{
5652
	int bpp = ext4_journal_blocks_per_page(inode);
5653 5654
	int ret;

5655
	ret = ext4_meta_trans_blocks(inode, bpp, 0);
A
Alex Tomas 已提交
5656

5657
	/* Account for data blocks for journalled mode */
5658
	if (ext4_should_journal_data(inode))
5659
		ret += bpp;
5660 5661
	return ret;
}
5662 5663 5664 5665 5666

/*
 * Calculate the journal credits for a chunk of data modification.
 *
 * This is called from DIO, fallocate or whoever calling
5667
 * ext4_get_blocks() to map/allocate a chunk of contiguous disk blocks.
5668 5669 5670 5671 5672 5673 5674 5675 5676
 *
 * journal buffers for data blocks are not included here, as DIO
 * and fallocate do no need to journal data buffers.
 */
int ext4_chunk_trans_blocks(struct inode *inode, int nrblocks)
{
	return ext4_meta_trans_blocks(inode, nrblocks, 1);
}

5677
/*
5678
 * The caller must have previously called ext4_reserve_inode_write().
5679 5680
 * Give this, we know that the caller already has write access to iloc->bh.
 */
5681
int ext4_mark_iloc_dirty(handle_t *handle,
5682
			 struct inode *inode, struct ext4_iloc *iloc)
5683 5684 5685
{
	int err = 0;

5686 5687 5688
	if (test_opt(inode->i_sb, I_VERSION))
		inode_inc_iversion(inode);

5689 5690 5691
	/* the do_update_inode consumes one bh->b_count */
	get_bh(iloc->bh);

5692
	/* ext4_do_update_inode() does jbd2_journal_dirty_metadata */
5693
	err = ext4_do_update_inode(handle, inode, iloc);
5694 5695 5696 5697 5698 5699 5700 5701 5702 5703
	put_bh(iloc->bh);
	return err;
}

/*
 * On success, We end up with an outstanding reference count against
 * iloc->bh.  This _must_ be cleaned up later.
 */

int
5704 5705
ext4_reserve_inode_write(handle_t *handle, struct inode *inode,
			 struct ext4_iloc *iloc)
5706
{
5707 5708 5709 5710 5711 5712 5713 5714 5715
	int err;

	err = ext4_get_inode_loc(inode, iloc);
	if (!err) {
		BUFFER_TRACE(iloc->bh, "get_write_access");
		err = ext4_journal_get_write_access(handle, iloc->bh);
		if (err) {
			brelse(iloc->bh);
			iloc->bh = NULL;
5716 5717
		}
	}
5718
	ext4_std_error(inode->i_sb, err);
5719 5720 5721
	return err;
}

5722 5723 5724 5725
/*
 * Expand an inode by new_extra_isize bytes.
 * Returns 0 on success or negative error number on failure.
 */
A
Aneesh Kumar K.V 已提交
5726 5727 5728 5729
static int ext4_expand_extra_isize(struct inode *inode,
				   unsigned int new_extra_isize,
				   struct ext4_iloc iloc,
				   handle_t *handle)
5730 5731 5732 5733 5734 5735 5736 5737 5738 5739 5740 5741 5742 5743
{
	struct ext4_inode *raw_inode;
	struct ext4_xattr_ibody_header *header;
	struct ext4_xattr_entry *entry;

	if (EXT4_I(inode)->i_extra_isize >= new_extra_isize)
		return 0;

	raw_inode = ext4_raw_inode(&iloc);

	header = IHDR(inode, raw_inode);
	entry = IFIRST(header);

	/* No extended attributes present */
5744 5745
	if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR) ||
	    header->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC)) {
5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756
		memset((void *)raw_inode + EXT4_GOOD_OLD_INODE_SIZE, 0,
			new_extra_isize);
		EXT4_I(inode)->i_extra_isize = new_extra_isize;
		return 0;
	}

	/* try to expand with EAs present */
	return ext4_expand_extra_isize_ea(inode, new_extra_isize,
					  raw_inode, handle);
}

5757 5758 5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770 5771 5772 5773 5774 5775 5776 5777
/*
 * What we do here is to mark the in-core inode as clean with respect to inode
 * dirtiness (it may still be data-dirty).
 * This means that the in-core inode may be reaped by prune_icache
 * without having to perform any I/O.  This is a very good thing,
 * because *any* task may call prune_icache - even ones which
 * have a transaction open against a different journal.
 *
 * Is this cheating?  Not really.  Sure, we haven't written the
 * inode out, but prune_icache isn't a user-visible syncing function.
 * Whenever the user wants stuff synced (sys_sync, sys_msync, sys_fsync)
 * we start and wait on commits.
 *
 * Is this efficient/effective?  Well, we're being nice to the system
 * by cleaning up our inodes proactively so they can be reaped
 * without I/O.  But we are potentially leaving up to five seconds'
 * worth of inodes floating about which prune_icache wants us to
 * write out.  One way to fix that would be to get prune_icache()
 * to do a write_super() to free up some memory.  It has the desired
 * effect.
 */
5778
int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode)
5779
{
5780
	struct ext4_iloc iloc;
5781 5782 5783
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
	static unsigned int mnt_count;
	int err, ret;
5784 5785

	might_sleep();
5786
	err = ext4_reserve_inode_write(handle, inode, &iloc);
5787 5788
	if (ext4_handle_valid(handle) &&
	    EXT4_I(inode)->i_extra_isize < sbi->s_want_extra_isize &&
5789
	    !ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND)) {
5790 5791 5792 5793 5794 5795 5796 5797 5798 5799 5800 5801 5802
		/*
		 * We need extra buffer credits since we may write into EA block
		 * with this same handle. If journal_extend fails, then it will
		 * only result in a minor loss of functionality for that inode.
		 * If this is felt to be critical, then e2fsck should be run to
		 * force a large enough s_min_extra_isize.
		 */
		if ((jbd2_journal_extend(handle,
			     EXT4_DATA_TRANS_BLOCKS(inode->i_sb))) == 0) {
			ret = ext4_expand_extra_isize(inode,
						      sbi->s_want_extra_isize,
						      iloc, handle);
			if (ret) {
5803 5804
				ext4_set_inode_state(inode,
						     EXT4_STATE_NO_EXPAND);
A
Aneesh Kumar K.V 已提交
5805 5806
				if (mnt_count !=
					le16_to_cpu(sbi->s_es->s_mnt_count)) {
5807
					ext4_warning(inode->i_sb,
5808 5809 5810
					"Unable to expand inode %lu. Delete"
					" some EAs or run e2fsck.",
					inode->i_ino);
A
Aneesh Kumar K.V 已提交
5811 5812
					mnt_count =
					  le16_to_cpu(sbi->s_es->s_mnt_count);
5813 5814 5815 5816
				}
			}
		}
	}
5817
	if (!err)
5818
		err = ext4_mark_iloc_dirty(handle, inode, &iloc);
5819 5820 5821 5822
	return err;
}

/*
5823
 * ext4_dirty_inode() is called from __mark_inode_dirty()
5824 5825 5826 5827 5828
 *
 * We're really interested in the case where a file is being extended.
 * i_size has been changed by generic_commit_write() and we thus need
 * to include the updated inode in the current transaction.
 *
5829
 * Also, dquot_alloc_block() will always dirty the inode when blocks
5830 5831 5832 5833 5834 5835
 * are allocated to the file.
 *
 * If the inode is marked synchronous, we don't honour that here - doing
 * so would cause a commit on atime updates, which we don't bother doing.
 * We handle synchronous inodes at the highest possible level.
 */
5836
void ext4_dirty_inode(struct inode *inode)
5837 5838 5839
{
	handle_t *handle;

5840
	handle = ext4_journal_start(inode, 2);
5841 5842
	if (IS_ERR(handle))
		goto out;
5843 5844 5845

	ext4_mark_inode_dirty(handle, inode);

5846
	ext4_journal_stop(handle);
5847 5848 5849 5850 5851 5852 5853 5854
out:
	return;
}

#if 0
/*
 * Bind an inode's backing buffer_head into this transaction, to prevent
 * it from being flushed to disk early.  Unlike
5855
 * ext4_reserve_inode_write, this leaves behind no bh reference and
5856 5857 5858
 * returns no iloc structure, so the caller needs to repeat the iloc
 * lookup to mark the inode dirty later.
 */
5859
static int ext4_pin_inode(handle_t *handle, struct inode *inode)
5860
{
5861
	struct ext4_iloc iloc;
5862 5863 5864

	int err = 0;
	if (handle) {
5865
		err = ext4_get_inode_loc(inode, &iloc);
5866 5867
		if (!err) {
			BUFFER_TRACE(iloc.bh, "get_write_access");
5868
			err = jbd2_journal_get_write_access(handle, iloc.bh);
5869
			if (!err)
5870
				err = ext4_handle_dirty_metadata(handle,
5871
								 NULL,
5872
								 iloc.bh);
5873 5874 5875
			brelse(iloc.bh);
		}
	}
5876
	ext4_std_error(inode->i_sb, err);
5877 5878 5879 5880
	return err;
}
#endif

5881
int ext4_change_inode_journal_flag(struct inode *inode, int val)
5882 5883 5884 5885 5886 5887 5888 5889 5890 5891 5892 5893 5894 5895 5896
{
	journal_t *journal;
	handle_t *handle;
	int err;

	/*
	 * We have to be very careful here: changing a data block's
	 * journaling status dynamically is dangerous.  If we write a
	 * data block to the journal, change the status and then delete
	 * that block, we risk forgetting to revoke the old log record
	 * from the journal and so a subsequent replay can corrupt data.
	 * So, first we make sure that the journal is empty and that
	 * nobody is changing anything.
	 */

5897
	journal = EXT4_JOURNAL(inode);
5898 5899
	if (!journal)
		return 0;
5900
	if (is_journal_aborted(journal))
5901 5902
		return -EROFS;

5903 5904
	jbd2_journal_lock_updates(journal);
	jbd2_journal_flush(journal);
5905 5906 5907 5908 5909 5910 5911 5912 5913 5914

	/*
	 * OK, there are no updates running now, and all cached data is
	 * synced to disk.  We are now in a completely consistent state
	 * which doesn't have anything in the journal, and we know that
	 * no filesystem updates are running, so it is safe to modify
	 * the inode's in-core data-journaling state flag now.
	 */

	if (val)
5915
		EXT4_I(inode)->i_flags |= EXT4_JOURNAL_DATA_FL;
5916
	else
5917 5918
		EXT4_I(inode)->i_flags &= ~EXT4_JOURNAL_DATA_FL;
	ext4_set_aops(inode);
5919

5920
	jbd2_journal_unlock_updates(journal);
5921 5922 5923

	/* Finally we can mark the inode as dirty. */

5924
	handle = ext4_journal_start(inode, 1);
5925 5926 5927
	if (IS_ERR(handle))
		return PTR_ERR(handle);

5928
	err = ext4_mark_inode_dirty(handle, inode);
5929
	ext4_handle_sync(handle);
5930 5931
	ext4_journal_stop(handle);
	ext4_std_error(inode->i_sb, err);
5932 5933 5934

	return err;
}
5935 5936 5937 5938 5939 5940

static int ext4_bh_unmapped(handle_t *handle, struct buffer_head *bh)
{
	return !buffer_mapped(bh);
}

5941
int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
5942
{
5943
	struct page *page = vmf->page;
5944 5945 5946
	loff_t size;
	unsigned long len;
	int ret = -EINVAL;
5947
	void *fsdata;
5948 5949 5950 5951 5952 5953 5954 5955 5956 5957 5958 5959 5960 5961 5962 5963 5964 5965 5966 5967 5968 5969 5970 5971
	struct file *file = vma->vm_file;
	struct inode *inode = file->f_path.dentry->d_inode;
	struct address_space *mapping = inode->i_mapping;

	/*
	 * Get i_alloc_sem to stop truncates messing with the inode. We cannot
	 * get i_mutex because we are already holding mmap_sem.
	 */
	down_read(&inode->i_alloc_sem);
	size = i_size_read(inode);
	if (page->mapping != mapping || size <= page_offset(page)
	    || !PageUptodate(page)) {
		/* page got truncated from under us? */
		goto out_unlock;
	}
	ret = 0;
	if (PageMappedToDisk(page))
		goto out_unlock;

	if (page->index == size >> PAGE_CACHE_SHIFT)
		len = size & ~PAGE_CACHE_MASK;
	else
		len = PAGE_CACHE_SIZE;

5972 5973 5974 5975 5976 5977 5978
	lock_page(page);
	/*
	 * return if we have all the buffers mapped. This avoid
	 * the need to call write_begin/write_end which does a
	 * journal_start/journal_stop which can block and take
	 * long time
	 */
5979 5980
	if (page_has_buffers(page)) {
		if (!walk_page_buffers(NULL, page_buffers(page), 0, len, NULL,
5981 5982
					ext4_bh_unmapped)) {
			unlock_page(page);
5983
			goto out_unlock;
5984
		}
5985
	}
5986
	unlock_page(page);
5987 5988 5989 5990 5991 5992 5993 5994
	/*
	 * OK, we need to fill the hole... Do write_begin write_end
	 * to do block allocation/reservation.We are not holding
	 * inode.i__mutex here. That allow * parallel write_begin,
	 * write_end call. lock_page prevent this from happening
	 * on the same page though
	 */
	ret = mapping->a_ops->write_begin(file, mapping, page_offset(page),
5995
			len, AOP_FLAG_UNINTERRUPTIBLE, &page, &fsdata);
5996 5997 5998
	if (ret < 0)
		goto out_unlock;
	ret = mapping->a_ops->write_end(file, mapping, page_offset(page),
5999
			len, len, page, fsdata);
6000 6001 6002 6003
	if (ret < 0)
		goto out_unlock;
	ret = 0;
out_unlock:
6004 6005
	if (ret)
		ret = VM_FAULT_SIGBUS;
6006 6007 6008
	up_read(&inode->i_alloc_sem);
	return ret;
}