inode.c 174.4 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
	int ret;

	/*
152
	 * Drop i_data_sem to avoid deadlock with ext4_map_blocks.  At this
153 154 155 156
	 * 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.
 */
A
Al Viro 已提交
170
void ext4_evict_inode(struct inode *inode)
171 172
{
	handle_t *handle;
173
	int err;
174

A
Al Viro 已提交
175 176 177 178 179
	if (inode->i_nlink) {
		truncate_inode_pages(&inode->i_data, 0);
		goto no_delete;
	}

180
	if (!is_bad_inode(inode))
181
		dquot_initialize(inode);
182

183 184
	if (ext4_should_order_data(inode))
		ext4_begin_ordered_truncate(inode, 0);
185 186 187 188 189
	truncate_inode_pages(&inode->i_data, 0);

	if (is_bad_inode(inode))
		goto no_delete;

190
	handle = ext4_journal_start(inode, blocks_for_truncate(inode)+3);
191
	if (IS_ERR(handle)) {
192
		ext4_std_error(inode->i_sb, PTR_ERR(handle));
193 194 195 196 197
		/*
		 * 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.
		 */
198
		ext4_orphan_del(NULL, inode);
199 200 201 202
		goto no_delete;
	}

	if (IS_SYNC(inode))
203
		ext4_handle_sync(handle);
204
	inode->i_size = 0;
205 206
	err = ext4_mark_inode_dirty(handle, inode);
	if (err) {
207
		ext4_warning(inode->i_sb,
208 209 210
			     "couldn't mark inode dirty (err %d)", err);
		goto stop_handle;
	}
211
	if (inode->i_blocks)
212
		ext4_truncate(inode);
213 214 215 216 217 218 219

	/*
	 * 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.
	 */
220
	if (!ext4_handle_has_enough_credits(handle, 3)) {
221 222 223 224
		err = ext4_journal_extend(handle, 3);
		if (err > 0)
			err = ext4_journal_restart(handle, 3);
		if (err != 0) {
225
			ext4_warning(inode->i_sb,
226 227 228
				     "couldn't extend journal (err %d)", err);
		stop_handle:
			ext4_journal_stop(handle);
229
			ext4_orphan_del(NULL, inode);
230 231 232 233
			goto no_delete;
		}
	}

234
	/*
235
	 * Kill off the orphan record which ext4_truncate created.
236
	 * AKPM: I think this can be inside the above `if'.
237
	 * Note that ext4_orphan_del() has to be able to cope with the
238
	 * deletion of a non-existent orphan - this is because we don't
239
	 * know if ext4_truncate() actually created an orphan record.
240 241
	 * (Well, we could do this if we need to, but heck - it works)
	 */
242 243
	ext4_orphan_del(handle, inode);
	EXT4_I(inode)->i_dtime	= get_seconds();
244 245 246 247 248 249 250 251

	/*
	 * 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.
	 */
252
	if (ext4_mark_inode_dirty(handle, inode))
253
		/* If that failed, just do the required in-core inode clear. */
A
Al Viro 已提交
254
		ext4_clear_inode(inode);
255
	else
256 257
		ext4_free_inode(handle, inode);
	ext4_journal_stop(handle);
258 259
	return;
no_delete:
A
Al Viro 已提交
260
	ext4_clear_inode(inode);	/* We must guarantee clearing of inode... */
261 262 263 264 265 266 267 268 269 270 271 272 273 274 275
}

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;
}

/**
276
 *	ext4_block_to_path - parse the block number into array of offsets
277 278 279
 *	@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
280 281
 *	@boundary: set this non-zero if the referred-to block is likely to be
 *	       followed (on disk) by an indirect block.
282
 *
283
 *	To store the locations of file's data ext4 uses a data structure common
284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305
 *	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.
 */

306
static int ext4_block_to_path(struct inode *inode,
307 308
			      ext4_lblk_t i_block,
			      ext4_lblk_t offsets[4], int *boundary)
309
{
310 311 312
	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,
313 314 315 316 317
		indirect_blocks = ptrs,
		double_blocks = (1 << (ptrs_bits * 2));
	int n = 0;
	int final = 0;

318
	if (i_block < direct_blocks) {
319 320
		offsets[n++] = i_block;
		final = direct_blocks;
321
	} else if ((i_block -= direct_blocks) < indirect_blocks) {
322
		offsets[n++] = EXT4_IND_BLOCK;
323 324 325
		offsets[n++] = i_block;
		final = ptrs;
	} else if ((i_block -= indirect_blocks) < double_blocks) {
326
		offsets[n++] = EXT4_DIND_BLOCK;
327 328 329 330
		offsets[n++] = i_block >> ptrs_bits;
		offsets[n++] = i_block & (ptrs - 1);
		final = ptrs;
	} else if (((i_block -= double_blocks) >> (ptrs_bits * 2)) < ptrs) {
331
		offsets[n++] = EXT4_TIND_BLOCK;
332 333 334 335 336
		offsets[n++] = i_block >> (ptrs_bits * 2);
		offsets[n++] = (i_block >> ptrs_bits) & (ptrs - 1);
		offsets[n++] = i_block & (ptrs - 1);
		final = ptrs;
	} else {
337
		ext4_warning(inode->i_sb, "block %lu > max in inode %lu",
338 339
			     i_block + direct_blocks +
			     indirect_blocks + double_blocks, inode->i_ino);
340 341 342 343 344 345
	}
	if (boundary)
		*boundary = final - 1 - (i_block & (ptrs - 1));
	return n;
}

346 347
static int __ext4_check_blockref(const char *function, unsigned int line,
				 struct inode *inode,
348 349
				 __le32 *p, unsigned int max)
{
350
	struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
351
	__le32 *bref = p;
352 353
	unsigned int blk;

354
	while (bref < p+max) {
355
		blk = le32_to_cpu(*bref++);
356 357
		if (blk &&
		    unlikely(!ext4_data_block_valid(EXT4_SB(inode->i_sb),
358
						    blk, 1))) {
359
			es->s_last_error_block = cpu_to_le64(blk);
360 361
			ext4_error_inode(inode, function, line, blk,
					 "invalid block");
362 363 364 365
			return -EIO;
		}
	}
	return 0;
366 367 368 369
}


#define ext4_check_indirect_blockref(inode, bh)                         \
370 371
	__ext4_check_blockref(__func__, __LINE__, inode,		\
			      (__le32 *)(bh)->b_data,			\
372 373 374
			      EXT4_ADDR_PER_BLOCK((inode)->i_sb))

#define ext4_check_inode_blockref(inode)                                \
375 376
	__ext4_check_blockref(__func__, __LINE__, inode,		\
			      EXT4_I(inode)->i_data,			\
377 378
			      EXT4_NDIR_BLOCKS)

379
/**
380
 *	ext4_get_branch - read the chain of indirect blocks leading to data
381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404
 *	@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).
405 406
 *
 *      Need to be called with
407
 *      down_read(&EXT4_I(inode)->i_data_sem)
408
 */
A
Aneesh Kumar K.V 已提交
409 410
static Indirect *ext4_get_branch(struct inode *inode, int depth,
				 ext4_lblk_t  *offsets,
411 412 413 414 415 416 417 418
				 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 */
419
	add_chain(chain, NULL, EXT4_I(inode)->i_data + *offsets);
420 421 422
	if (!p->key)
		goto no_block;
	while (--depth) {
423 424
		bh = sb_getblk(sb, le32_to_cpu(p->key));
		if (unlikely(!bh))
425
			goto failure;
426

427 428 429 430 431 432 433 434 435 436 437
		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;
			}
		}
438

439
		add_chain(++p, bh, (__le32 *)bh->b_data + *++offsets);
440 441 442 443 444 445 446 447 448 449 450 451 452
		/* Reader: end */
		if (!p->key)
			goto no_block;
	}
	return NULL;

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

/**
453
 *	ext4_find_near - find a place for allocation with sufficient locality
454 455 456
 *	@inode: owner
 *	@ind: descriptor of indirect block.
 *
457
 *	This function returns the preferred place for block allocation.
458 459 460 461 462 463 464 465 466 467 468 469 470 471
 *	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.
 */
472
static ext4_fsblk_t ext4_find_near(struct inode *inode, Indirect *ind)
473
{
474
	struct ext4_inode_info *ei = EXT4_I(inode);
475
	__le32 *start = ind->bh ? (__le32 *) ind->bh->b_data : ei->i_data;
476
	__le32 *p;
477
	ext4_fsblk_t bg_start;
478
	ext4_fsblk_t last_block;
479
	ext4_grpblk_t colour;
480 481
	ext4_group_t block_group;
	int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb));
482 483 484 485 486 487 488 489 490 491 492 493 494 495 496

	/* 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.
	 */
497 498 499 500 501 502 503
	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);
504 505
	last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1;

506 507 508 509 510 511 512
	/*
	 * If we are doing delayed allocation, we don't need take
	 * colour into account.
	 */
	if (test_opt(inode->i_sb, DELALLOC))
		return bg_start;

513 514
	if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block)
		colour = (current->pid % 16) *
515
			(EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16);
516 517
	else
		colour = (current->pid % 16) * ((last_block - bg_start) / 16);
518 519 520 521
	return bg_start + colour;
}

/**
522
 *	ext4_find_goal - find a preferred place for allocation.
523 524 525 526
 *	@inode: owner
 *	@block:  block we want
 *	@partial: pointer to the last triple within a chain
 *
527
 *	Normally this function find the preferred place for block allocation,
528
 *	returns it.
529 530
 *	Because this is only used for non-extent files, we limit the block nr
 *	to 32 bits.
531
 */
A
Aneesh Kumar K.V 已提交
532
static ext4_fsblk_t ext4_find_goal(struct inode *inode, ext4_lblk_t block,
533
				   Indirect *partial)
534
{
535 536
	ext4_fsblk_t goal;

537
	/*
538
	 * XXX need to get goal block from mballoc's data structures
539 540
	 */

541 542 543
	goal = ext4_find_near(inode, partial);
	goal = goal & EXT4_MAX_BLOCK_FILE_PHYS;
	return goal;
544 545 546
}

/**
547
 *	ext4_blks_to_allocate: Look up the block map and count the number
548 549 550 551 552 553 554 555 556 557
 *	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.
 */
558
static int ext4_blks_to_allocate(Indirect *branch, int k, unsigned int blks,
559
				 int blocks_to_boundary)
560
{
561
	unsigned int count = 0;
562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584

	/*
	 * 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;
}

/**
585
 *	ext4_alloc_blocks: multiple allocate blocks needed for a branch
586 587 588 589 590 591 592 593
 *	@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
 */
594
static int ext4_alloc_blocks(handle_t *handle, struct inode *inode,
595 596 597
			     ext4_lblk_t iblock, ext4_fsblk_t goal,
			     int indirect_blks, int blks,
			     ext4_fsblk_t new_blocks[4], int *err)
598
{
599
	struct ext4_allocation_request ar;
600
	int target, i;
601
	unsigned long count = 0, blk_allocated = 0;
602
	int index = 0;
603
	ext4_fsblk_t current_block = 0;
604 605 606 607 608 609 610 611 612 613
	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)
	 */
614 615 616
	/* first we try to allocate the indirect blocks */
	target = indirect_blks;
	while (target > 0) {
617 618
		count = target;
		/* allocating blocks for indirect blocks and direct blocks */
619 620
		current_block = ext4_new_meta_blocks(handle, inode,
							goal, &count, err);
621 622 623
		if (*err)
			goto failed_out;

624 625 626 627 628 629 630 631
		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;
		}
632

633 634 635 636 637 638
		target -= count;
		/* allocate blocks for indirect blocks */
		while (index < indirect_blks && count) {
			new_blocks[index++] = current_block++;
			count--;
		}
639 640 641 642 643 644 645 646 647
		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);
648
			break;
649
		}
650 651
	}

652 653 654 655 656
	target = blks - count ;
	blk_allocated = count;
	if (!target)
		goto allocated;
	/* Now allocate data blocks */
657 658 659 660 661 662 663 664 665 666
	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);
667 668 669 670 671 672 673 674
	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;
	}
675

676 677 678 679 680 681 682 683 684
	if (*err && (target == blks)) {
		/*
		 * if the allocation failed and we didn't allocate
		 * any blocks before
		 */
		goto failed_out;
	}
	if (!*err) {
		if (target == blks) {
685 686 687 688
			/*
			 * save the new block number
			 * for the first direct block
			 */
689 690
			new_blocks[index] = current_block;
		}
691
		blk_allocated += ar.len;
692 693
	}
allocated:
694
	/* total number of blocks allocated for direct blocks */
695
	ret = blk_allocated;
696 697 698
	*err = 0;
	return ret;
failed_out:
699
	for (i = 0; i < index; i++)
700
		ext4_free_blocks(handle, inode, 0, new_blocks[i], 1, 0);
701 702 703 704
	return ret;
}

/**
705
 *	ext4_alloc_branch - allocate and set up a chain of blocks.
706 707 708 709 710 711 712 713 714 715
 *	@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
716
 *	the same format as ext4_get_branch() would do. We are calling it after
717 718
 *	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
719
 *	picture as after the successful ext4_get_block(), except that in one
720 721 722 723 724 725
 *	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
726
 *	ext4_alloc_block() (normally -ENOSPC). Otherwise we set the chain
727 728
 *	as described above and return 0.
 */
729
static int ext4_alloc_branch(handle_t *handle, struct inode *inode,
730 731 732
			     ext4_lblk_t iblock, int indirect_blks,
			     int *blks, ext4_fsblk_t goal,
			     ext4_lblk_t *offsets, Indirect *branch)
733 734 735 736 737 738
{
	int blocksize = inode->i_sb->s_blocksize;
	int i, n = 0;
	int err = 0;
	struct buffer_head *bh;
	int num;
739 740
	ext4_fsblk_t new_blocks[4];
	ext4_fsblk_t current_block;
741

742
	num = ext4_alloc_blocks(handle, inode, iblock, goal, indirect_blks,
743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760
				*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");
761
		err = ext4_journal_get_create_access(handle, bh);
762
		if (err) {
763 764
			/* Don't brelse(bh) here; it's done in
			 * ext4_journal_forget() below */
765 766 767 768 769 770 771 772
			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;
773
		if (n == indirect_blks) {
774 775 776 777 778 779
			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
			 */
780
			for (i = 1; i < num; i++)
781 782 783 784 785 786
				*(branch[n].p + i) = cpu_to_le32(++current_block);
		}
		BUFFER_TRACE(bh, "marking uptodate");
		set_buffer_uptodate(bh);
		unlock_buffer(bh);

787 788
		BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
		err = ext4_handle_dirty_metadata(handle, inode, bh);
789 790 791 792 793 794 795
		if (err)
			goto failed;
	}
	*blks = num;
	return err;
failed:
	/* Allocation failed, free what we already allocated */
796
	ext4_free_blocks(handle, inode, 0, new_blocks[0], 1, 0);
797
	for (i = 1; i <= n ; i++) {
798
		/*
799 800 801
		 * 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.
802
		 */
803 804
		ext4_free_blocks(handle, inode, 0, new_blocks[i], 1,
				 EXT4_FREE_BLOCKS_FORGET);
805
	}
806 807
	for (i = n+1; i < indirect_blks; i++)
		ext4_free_blocks(handle, inode, 0, new_blocks[i], 1, 0);
808

809
	ext4_free_blocks(handle, inode, 0, new_blocks[i], num, 0);
810 811 812 813 814

	return err;
}

/**
815
 * ext4_splice_branch - splice the allocated branch onto inode.
816 817 818
 * @inode: owner
 * @block: (logical) number of block we are adding
 * @chain: chain of indirect blocks (with a missing link - see
819
 *	ext4_alloc_branch)
820 821 822 823 824 825 826 827
 * @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.
 */
828
static int ext4_splice_branch(handle_t *handle, struct inode *inode,
829 830
			      ext4_lblk_t block, Indirect *where, int num,
			      int blks)
831 832 833
{
	int i;
	int err = 0;
834
	ext4_fsblk_t current_block;
835 836 837 838 839 840 841 842

	/*
	 * 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");
843
		err = ext4_journal_get_write_access(handle, where->bh);
844 845 846 847 848 849 850 851 852 853 854 855 856 857
		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++)
858
			*(where->p + i) = cpu_to_le32(current_block++);
859 860 861 862 863 864 865 866 867 868 869
	}

	/* 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
870
		 * generic_commit_write->__mark_inode_dirty->ext4_dirty_inode.
871 872
		 */
		jbd_debug(5, "splicing indirect only\n");
873 874
		BUFFER_TRACE(where->bh, "call ext4_handle_dirty_metadata");
		err = ext4_handle_dirty_metadata(handle, inode, where->bh);
875 876 877 878 879 880
		if (err)
			goto err_out;
	} else {
		/*
		 * OK, we spliced it into the inode itself on a direct block.
		 */
881
		ext4_mark_inode_dirty(handle, inode);
882 883 884 885 886 887
		jbd_debug(5, "splicing direct\n");
	}
	return err;

err_out:
	for (i = 1; i <= num; i++) {
888
		/*
889 890 891
		 * 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.
892
		 */
893 894
		ext4_free_blocks(handle, inode, where[i].bh, 0, 1,
				 EXT4_FREE_BLOCKS_FORGET);
895
	}
896 897
	ext4_free_blocks(handle, inode, 0, le32_to_cpu(where[num].key),
			 blks, 0);
898 899 900 901 902

	return err;
}

/*
903
 * The ext4_ind_map_blocks() function handles non-extents inodes
904
 * (i.e., using the traditional indirect/double-indirect i_blocks
905
 * scheme) for ext4_map_blocks().
906
 *
907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922
 * 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.
923
 *
924 925 926 927 928
 * 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.
929
 */
930 931
static int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
			       struct ext4_map_blocks *map,
932
			       int flags)
933 934
{
	int err = -EIO;
A
Aneesh Kumar K.V 已提交
935
	ext4_lblk_t offsets[4];
936 937
	Indirect chain[4];
	Indirect *partial;
938
	ext4_fsblk_t goal;
939 940 941 942
	int indirect_blks;
	int blocks_to_boundary = 0;
	int depth;
	int count = 0;
943
	ext4_fsblk_t first_block = 0;
944

945
	J_ASSERT(!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)));
946
	J_ASSERT(handle != NULL || (flags & EXT4_GET_BLOCKS_CREATE) == 0);
947
	depth = ext4_block_to_path(inode, map->m_lblk, offsets,
948
				   &blocks_to_boundary);
949 950 951 952

	if (depth == 0)
		goto out;

953
	partial = ext4_get_branch(inode, depth, offsets, chain, &err);
954 955 956 957 958 959

	/* Simplest case - block found, no allocation needed */
	if (!partial) {
		first_block = le32_to_cpu(chain[depth - 1].key);
		count++;
		/*map more blocks*/
960
		while (count < map->m_len && count <= blocks_to_boundary) {
961
			ext4_fsblk_t blk;
962 963 964 965 966 967 968 969

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

			if (blk == first_block + count)
				count++;
			else
				break;
		}
970
		goto got_it;
971 972 973
	}

	/* Next simple case - plain lookup or failed read of indirect block */
974
	if ((flags & EXT4_GET_BLOCKS_CREATE) == 0 || err == -EIO)
975 976 977
		goto cleanup;

	/*
978
	 * Okay, we need to do block allocation.
979
	*/
980
	goal = ext4_find_goal(inode, map->m_lblk, partial);
981 982 983 984 985 986 987 988

	/* 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.
	 */
989
	count = ext4_blks_to_allocate(partial, indirect_blks,
990
				      map->m_len, blocks_to_boundary);
991
	/*
992
	 * Block out ext4_truncate while we alter the tree
993
	 */
994
	err = ext4_alloc_branch(handle, inode, map->m_lblk, indirect_blks,
995 996
				&count, goal,
				offsets + (partial - chain), partial);
997 998

	/*
999
	 * The ext4_splice_branch call will free and forget any buffers
1000 1001 1002 1003 1004 1005
	 * 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)
1006
		err = ext4_splice_branch(handle, inode, map->m_lblk,
1007
					 partial, indirect_blks, count);
1008
	if (err)
1009 1010
		goto cleanup;

1011
	map->m_flags |= EXT4_MAP_NEW;
1012 1013

	ext4_update_inode_fsync_trans(handle, inode, 1);
1014
got_it:
1015 1016 1017
	map->m_flags |= EXT4_MAP_MAPPED;
	map->m_pblk = le32_to_cpu(chain[depth-1].key);
	map->m_len = count;
1018
	if (count > blocks_to_boundary)
1019
		map->m_flags |= EXT4_MAP_BOUNDARY;
1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032
	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--;
	}
out:
	return err;
}

1033 1034
#ifdef CONFIG_QUOTA
qsize_t *ext4_get_reserved_space(struct inode *inode)
1035
{
1036
	return &EXT4_I(inode)->i_reserved_quota;
1037
}
1038
#endif
1039

1040 1041
/*
 * Calculate the number of metadata blocks need to reserve
1042
 * to allocate a new block at @lblocks for non extent file based file
1043
 */
1044 1045
static int ext4_indirect_calc_metadata_amount(struct inode *inode,
					      sector_t lblock)
1046
{
1047
	struct ext4_inode_info *ei = EXT4_I(inode);
1048
	sector_t dind_mask = ~((sector_t)EXT4_ADDR_PER_BLOCK(inode->i_sb) - 1);
1049
	int blk_bits;
1050

1051 1052
	if (lblock < EXT4_NDIR_BLOCKS)
		return 0;
1053

1054
	lblock -= EXT4_NDIR_BLOCKS;
1055

1056 1057 1058 1059 1060 1061 1062
	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;
1063
	blk_bits = order_base_2(lblock);
1064
	return (blk_bits / EXT4_ADDR_PER_BLOCK_BITS(inode->i_sb)) + 1;
1065 1066 1067 1068
}

/*
 * Calculate the number of metadata blocks need to reserve
1069
 * to allocate a block located at @lblock
1070
 */
1071
static int ext4_calc_metadata_amount(struct inode *inode, sector_t lblock)
1072
{
1073
	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
1074
		return ext4_ext_calc_metadata_amount(inode, lblock);
1075

1076
	return ext4_indirect_calc_metadata_amount(inode, lblock);
1077 1078
}

1079 1080 1081 1082
/*
 * Called with i_data_sem down, which is important since we can call
 * ext4_discard_preallocations() from here.
 */
1083 1084
void ext4_da_update_reserve_space(struct inode *inode,
					int used, int quota_claim)
1085 1086
{
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1087 1088 1089
	struct ext4_inode_info *ei = EXT4_I(inode);

	spin_lock(&ei->i_block_reservation_lock);
1090
	trace_ext4_da_update_reserve_space(inode, used);
1091 1092 1093 1094 1095 1096 1097 1098
	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;
	}
1099

1100 1101 1102
	/* Update per-inode reservations */
	ei->i_reserved_data_blocks -= used;
	ei->i_reserved_meta_blocks -= ei->i_allocated_meta_blocks;
1103 1104
	percpu_counter_sub(&sbi->s_dirtyblocks_counter,
			   used + ei->i_allocated_meta_blocks);
1105
	ei->i_allocated_meta_blocks = 0;
1106

1107 1108 1109 1110 1111 1112
	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.
		 */
1113 1114
		percpu_counter_sub(&sbi->s_dirtyblocks_counter,
				   ei->i_reserved_meta_blocks);
1115
		ei->i_reserved_meta_blocks = 0;
1116
		ei->i_da_metadata_calc_len = 0;
1117
	}
1118
	spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
1119

1120 1121
	/* Update quota subsystem for data blocks */
	if (quota_claim)
1122
		dquot_claim_block(inode, used);
1123
	else {
1124 1125 1126
		/*
		 * We did fallocate with an offset that is already delayed
		 * allocated. So on delayed allocated writeback we should
1127
		 * not re-claim the quota for fallocated blocks.
1128
		 */
1129
		dquot_release_reservation_block(inode, used);
1130
	}
1131 1132 1133 1134 1135 1136

	/*
	 * 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.
	 */
1137 1138
	if ((ei->i_reserved_data_blocks == 0) &&
	    (atomic_read(&inode->i_writecount) == 0))
1139
		ext4_discard_preallocations(inode);
1140 1141
}

1142
static int __check_block_validity(struct inode *inode, const char *func,
1143 1144
				unsigned int line,
				struct ext4_map_blocks *map)
1145
{
1146 1147
	if (!ext4_data_block_valid(EXT4_SB(inode->i_sb), map->m_pblk,
				   map->m_len)) {
1148 1149 1150 1151
		ext4_error_inode(inode, func, line, map->m_pblk,
				 "lblock %lu mapped to illegal pblock "
				 "(length %d)", (unsigned long) map->m_lblk,
				 map->m_len);
1152 1153 1154 1155 1156
		return -EIO;
	}
	return 0;
}

1157
#define check_block_validity(inode, map)	\
1158
	__check_block_validity((inode), __func__, __LINE__, (map))
1159

1160
/*
1161 1162
 * Return the number of contiguous dirty pages in a given inode
 * starting at page frame idx.
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 1193 1194 1195
 */
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;
			}
1196 1197 1198 1199 1200 1201 1202 1203 1204
			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));
			}
1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217
			unlock_page(page);
			if (done)
				break;
			idx++;
			num++;
			if (num >= max_pages)
				break;
		}
		pagevec_release(&pvec);
	}
	return num;
}

1218
/*
1219
 * The ext4_map_blocks() function tries to look up the requested blocks,
1220
 * and returns if the blocks are already mapped.
1221 1222 1223 1224 1225
 *
 * 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.
 *
1226 1227
 * If file type is extents based, it will call ext4_ext_map_blocks(),
 * Otherwise, call with ext4_ind_map_blocks() to handle indirect mapping
1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239
 * 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.
 */
1240 1241
int ext4_map_blocks(handle_t *handle, struct inode *inode,
		    struct ext4_map_blocks *map, int flags)
1242 1243
{
	int retval;
1244

1245 1246 1247 1248
	map->m_flags = 0;
	ext_debug("ext4_map_blocks(): inode %lu, flag %d, max_blocks %u,"
		  "logical block %lu\n", inode->i_ino, flags, map->m_len,
		  (unsigned long) map->m_lblk);
1249
	/*
1250 1251
	 * Try to see if we can get the block without requesting a new
	 * file system block.
1252 1253
	 */
	down_read((&EXT4_I(inode)->i_data_sem));
1254
	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
1255
		retval = ext4_ext_map_blocks(handle, inode, map, 0);
1256
	} else {
1257
		retval = ext4_ind_map_blocks(handle, inode, map, 0);
1258
	}
1259
	up_read((&EXT4_I(inode)->i_data_sem));
1260

1261
	if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED) {
1262
		int ret = check_block_validity(inode, map);
1263 1264 1265 1266
		if (ret != 0)
			return ret;
	}

1267
	/* If it is only a block(s) look up */
1268
	if ((flags & EXT4_GET_BLOCKS_CREATE) == 0)
1269 1270 1271 1272 1273 1274 1275 1276 1277
		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.
	 */
1278
	if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED)
1279 1280
		return retval;

1281 1282 1283 1284 1285 1286 1287 1288 1289 1290
	/*
	 * 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.
	 */
1291
	map->m_flags &= ~EXT4_MAP_UNWRITTEN;
1292

1293
	/*
1294 1295 1296 1297
	 * 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.
1298 1299
	 */
	down_write((&EXT4_I(inode)->i_data_sem));
1300 1301 1302 1303 1304 1305 1306

	/*
	 * 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
	 */
1307
	if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
1308
		EXT4_I(inode)->i_delalloc_reserved_flag = 1;
1309 1310 1311 1312
	/*
	 * We need to check for EXT4 here because migrate
	 * could have changed the inode type in between
	 */
1313
	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
1314
		retval = ext4_ext_map_blocks(handle, inode, map, flags);
1315
	} else {
1316
		retval = ext4_ind_map_blocks(handle, inode, map, flags);
1317

1318
		if (retval > 0 && map->m_flags & EXT4_MAP_NEW) {
1319 1320 1321 1322 1323
			/*
			 * We allocated new blocks which will result in
			 * i_data's format changing.  Force the migrate
			 * to fail by clearing migrate flags
			 */
1324
			ext4_clear_inode_state(inode, EXT4_STATE_EXT_MIGRATE);
1325
		}
1326

1327 1328 1329 1330 1331 1332 1333
		/*
		 * 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) &&
1334
			(flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE))
1335 1336
			ext4_da_update_reserve_space(inode, retval, 1);
	}
1337
	if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
1338
		EXT4_I(inode)->i_delalloc_reserved_flag = 0;
1339

1340
	up_write((&EXT4_I(inode)->i_data_sem));
1341
	if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED) {
1342
		int ret = check_block_validity(inode, map);
1343 1344 1345
		if (ret != 0)
			return ret;
	}
1346 1347 1348
	return retval;
}

1349 1350 1351
/* Maximum number of blocks we map for direct IO at once. */
#define DIO_MAX_BLOCKS 4096

1352 1353
static int _ext4_get_block(struct inode *inode, sector_t iblock,
			   struct buffer_head *bh, int flags)
1354
{
1355
	handle_t *handle = ext4_journal_current_handle();
1356
	struct ext4_map_blocks map;
J
Jan Kara 已提交
1357
	int ret = 0, started = 0;
1358
	int dio_credits;
1359

1360 1361 1362 1363
	map.m_lblk = iblock;
	map.m_len = bh->b_size >> inode->i_blkbits;

	if (flags && !handle) {
J
Jan Kara 已提交
1364
		/* Direct IO write... */
1365 1366 1367
		if (map.m_len > DIO_MAX_BLOCKS)
			map.m_len = DIO_MAX_BLOCKS;
		dio_credits = ext4_chunk_trans_blocks(inode, map.m_len);
1368
		handle = ext4_journal_start(inode, dio_credits);
J
Jan Kara 已提交
1369
		if (IS_ERR(handle)) {
1370
			ret = PTR_ERR(handle);
1371
			return ret;
1372
		}
J
Jan Kara 已提交
1373
		started = 1;
1374 1375
	}

1376
	ret = ext4_map_blocks(handle, inode, &map, flags);
J
Jan Kara 已提交
1377
	if (ret > 0) {
1378 1379 1380
		map_bh(bh, inode->i_sb, map.m_pblk);
		bh->b_state = (bh->b_state & ~EXT4_MAP_FLAGS) | map.m_flags;
		bh->b_size = inode->i_sb->s_blocksize * map.m_len;
J
Jan Kara 已提交
1381
		ret = 0;
1382
	}
J
Jan Kara 已提交
1383 1384
	if (started)
		ext4_journal_stop(handle);
1385 1386 1387
	return ret;
}

1388 1389 1390 1391 1392 1393 1394
int ext4_get_block(struct inode *inode, sector_t iblock,
		   struct buffer_head *bh, int create)
{
	return _ext4_get_block(inode, iblock, bh,
			       create ? EXT4_GET_BLOCKS_CREATE : 0);
}

1395 1396 1397
/*
 * `handle' can be NULL if create is zero
 */
1398
struct buffer_head *ext4_getblk(handle_t *handle, struct inode *inode,
A
Aneesh Kumar K.V 已提交
1399
				ext4_lblk_t block, int create, int *errp)
1400
{
1401 1402
	struct ext4_map_blocks map;
	struct buffer_head *bh;
1403 1404 1405 1406
	int fatal = 0, err;

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

1407 1408 1409 1410
	map.m_lblk = block;
	map.m_len = 1;
	err = ext4_map_blocks(handle, inode, &map,
			      create ? EXT4_GET_BLOCKS_CREATE : 0);
1411

1412 1413 1414 1415 1416 1417 1418 1419 1420 1421
	if (err < 0)
		*errp = err;
	if (err <= 0)
		return NULL;
	*errp = 0;

	bh = sb_getblk(inode->i_sb, map.m_pblk);
	if (!bh) {
		*errp = -EIO;
		return NULL;
1422
	}
1423 1424 1425
	if (map.m_flags & EXT4_MAP_NEW) {
		J_ASSERT(create != 0);
		J_ASSERT(handle != NULL);
1426

1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439
		/*
		 * 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
		 * writes use ext4_get_block instead, so it's not a
		 * problem.
		 */
		lock_buffer(bh);
		BUFFER_TRACE(bh, "call get_create_access");
		fatal = ext4_journal_get_create_access(handle, bh);
		if (!fatal && !buffer_uptodate(bh)) {
			memset(bh->b_data, 0, inode->i_sb->s_blocksize);
			set_buffer_uptodate(bh);
1440
		}
1441 1442 1443 1444 1445 1446 1447
		unlock_buffer(bh);
		BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
		err = ext4_handle_dirty_metadata(handle, inode, bh);
		if (!fatal)
			fatal = err;
	} else {
		BUFFER_TRACE(bh, "not a new buffer");
1448
	}
1449 1450 1451 1452 1453 1454
	if (fatal) {
		*errp = fatal;
		brelse(bh);
		bh = NULL;
	}
	return bh;
1455 1456
}

1457
struct buffer_head *ext4_bread(handle_t *handle, struct inode *inode,
A
Aneesh Kumar K.V 已提交
1458
			       ext4_lblk_t block, int create, int *err)
1459
{
1460
	struct buffer_head *bh;
1461

1462
	bh = ext4_getblk(handle, inode, block, create, err);
1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475
	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;
}

1476 1477 1478 1479 1480 1481 1482
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))
1483 1484 1485 1486 1487 1488 1489
{
	struct buffer_head *bh;
	unsigned block_start, block_end;
	unsigned blocksize = head->b_size;
	int err, ret = 0;
	struct buffer_head *next;

1490 1491
	for (bh = head, block_start = 0;
	     ret == 0 && (bh != head || !block_start);
1492
	     block_start = block_end, bh = next) {
1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509
		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
1510
 * close off a transaction and start a new one between the ext4_get_block()
1511
 * and the commit_write().  So doing the jbd2_journal_start at the start of
1512 1513
 * prepare_write() is the right place.
 *
1514 1515
 * Also, this function can nest inside ext4_writepage() ->
 * block_write_full_page(). In that case, we *know* that ext4_writepage()
1516 1517 1518 1519
 * 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.
 *
1520
 * By accident, ext4 can be reentered when a transaction is open via
1521 1522 1523 1524 1525 1526
 * 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.
 *
1527
 * So what we do is to rely on the fact that jbd2_journal_stop/journal_start
1528 1529 1530 1531 1532
 * 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,
1533
				       struct buffer_head *bh)
1534
{
1535 1536 1537
	int dirty = buffer_dirty(bh);
	int ret;

1538 1539
	if (!buffer_mapped(bh) || buffer_freed(bh))
		return 0;
1540
	/*
C
Christoph Hellwig 已提交
1541
	 * __block_write_begin() could have dirtied some buffers. Clean
1542 1543
	 * the dirty bit as jbd2_journal_get_write_access() could complain
	 * otherwise about fs integrity issues. Setting of the dirty bit
C
Christoph Hellwig 已提交
1544
	 * by __block_write_begin() isn't a real problem here as we clear
1545 1546 1547 1548 1549 1550 1551 1552 1553
	 * the bit before releasing a page lock and thus writeback cannot
	 * ever write the buffer.
	 */
	if (dirty)
		clear_buffer_dirty(bh);
	ret = ext4_journal_get_write_access(handle, bh);
	if (!ret && dirty)
		ret = ext4_handle_dirty_metadata(handle, NULL, bh);
	return ret;
1554 1555
}

1556 1557 1558 1559 1560 1561 1562 1563 1564 1565
/*
 * 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);
}

1566 1567
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
1568
static int ext4_write_begin(struct file *file, struct address_space *mapping,
1569 1570
			    loff_t pos, unsigned len, unsigned flags,
			    struct page **pagep, void **fsdata)
1571
{
1572
	struct inode *inode = mapping->host;
1573
	int ret, needed_blocks;
1574 1575
	handle_t *handle;
	int retries = 0;
1576
	struct page *page;
1577
	pgoff_t index;
1578
	unsigned from, to;
N
Nick Piggin 已提交
1579

1580
	trace_ext4_write_begin(inode, pos, len, flags);
1581 1582 1583 1584 1585
	/*
	 * 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;
1586
	index = pos >> PAGE_CACHE_SHIFT;
1587 1588
	from = pos & (PAGE_CACHE_SIZE - 1);
	to = from + len;
1589 1590

retry:
1591 1592 1593 1594
	handle = ext4_journal_start(inode, needed_blocks);
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		goto out;
1595
	}
1596

1597 1598 1599 1600
	/* We cannot recurse into the filesystem as the transaction is already
	 * started */
	flags |= AOP_FLAG_NOFS;

1601
	page = grab_cache_page_write_begin(mapping, index, flags);
1602 1603 1604 1605 1606 1607 1608
	if (!page) {
		ext4_journal_stop(handle);
		ret = -ENOMEM;
		goto out;
	}
	*pagep = page;

1609
	if (ext4_should_dioread_nolock(inode))
1610
		ret = __block_write_begin(page, pos, len, ext4_get_block_write);
1611
	else
1612
		ret = __block_write_begin(page, pos, len, ext4_get_block);
N
Nick Piggin 已提交
1613 1614

	if (!ret && ext4_should_journal_data(inode)) {
1615 1616 1617
		ret = walk_page_buffers(handle, page_buffers(page),
				from, to, NULL, do_journal_get_write_access);
	}
N
Nick Piggin 已提交
1618 1619

	if (ret) {
1620 1621
		unlock_page(page);
		page_cache_release(page);
1622
		/*
1623
		 * __block_write_begin may have instantiated a few blocks
1624 1625
		 * outside i_size.  Trim these off again. Don't need
		 * i_size_read because we hold i_mutex.
1626 1627 1628
		 *
		 * Add inode to orphan list in case we crash before
		 * truncate finishes
1629
		 */
1630
		if (pos + len > inode->i_size && ext4_can_truncate(inode))
1631 1632 1633 1634
			ext4_orphan_add(handle, inode);

		ext4_journal_stop(handle);
		if (pos + len > inode->i_size) {
1635
			ext4_truncate_failed_write(inode);
1636
			/*
1637
			 * If truncate failed early the inode might
1638 1639 1640 1641 1642 1643 1644
			 * 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
1645 1646
	}

1647
	if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1648
		goto retry;
1649
out:
1650 1651 1652
	return ret;
}

N
Nick Piggin 已提交
1653 1654
/* For write_end() in data=journal mode */
static int write_end_fn(handle_t *handle, struct buffer_head *bh)
1655 1656 1657 1658
{
	if (!buffer_mapped(bh) || buffer_freed(bh))
		return 0;
	set_buffer_uptodate(bh);
1659
	return ext4_handle_dirty_metadata(handle, NULL, bh);
1660 1661
}

1662
static int ext4_generic_write_end(struct file *file,
1663 1664 1665
				  struct address_space *mapping,
				  loff_t pos, unsigned len, unsigned copied,
				  struct page *page, void *fsdata)
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 1701 1702 1703 1704 1705 1706 1707
{
	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;
}

1708 1709 1710 1711
/*
 * 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().
 *
1712
 * ext4 never places buffers on inode->i_mapping->private_list.  metadata
1713 1714
 * buffers are managed internally.
 */
N
Nick Piggin 已提交
1715
static int ext4_ordered_write_end(struct file *file,
1716 1717 1718
				  struct address_space *mapping,
				  loff_t pos, unsigned len, unsigned copied,
				  struct page *page, void *fsdata)
1719
{
1720
	handle_t *handle = ext4_journal_current_handle();
1721
	struct inode *inode = mapping->host;
1722 1723
	int ret = 0, ret2;

1724
	trace_ext4_ordered_write_end(inode, pos, len, copied);
1725
	ret = ext4_jbd2_file_inode(handle, inode);
1726 1727

	if (ret == 0) {
1728
		ret2 = ext4_generic_write_end(file, mapping, pos, len, copied,
N
Nick Piggin 已提交
1729
							page, fsdata);
1730
		copied = ret2;
1731
		if (pos + len > inode->i_size && ext4_can_truncate(inode))
1732 1733 1734 1735 1736
			/* 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);
1737 1738
		if (ret2 < 0)
			ret = ret2;
1739
	}
1740
	ret2 = ext4_journal_stop(handle);
1741 1742
	if (!ret)
		ret = ret2;
N
Nick Piggin 已提交
1743

1744
	if (pos + len > inode->i_size) {
1745
		ext4_truncate_failed_write(inode);
1746
		/*
1747
		 * If truncate failed early the inode might still be
1748 1749 1750 1751 1752 1753 1754 1755
		 * 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
1756
	return ret ? ret : copied;
1757 1758
}

N
Nick Piggin 已提交
1759
static int ext4_writeback_write_end(struct file *file,
1760 1761 1762
				    struct address_space *mapping,
				    loff_t pos, unsigned len, unsigned copied,
				    struct page *page, void *fsdata)
1763
{
1764
	handle_t *handle = ext4_journal_current_handle();
1765
	struct inode *inode = mapping->host;
1766 1767
	int ret = 0, ret2;

1768
	trace_ext4_writeback_write_end(inode, pos, len, copied);
1769
	ret2 = ext4_generic_write_end(file, mapping, pos, len, copied,
N
Nick Piggin 已提交
1770
							page, fsdata);
1771
	copied = ret2;
1772
	if (pos + len > inode->i_size && ext4_can_truncate(inode))
1773 1774 1775 1776 1777 1778
		/* 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);

1779 1780
	if (ret2 < 0)
		ret = ret2;
1781

1782
	ret2 = ext4_journal_stop(handle);
1783 1784
	if (!ret)
		ret = ret2;
N
Nick Piggin 已提交
1785

1786
	if (pos + len > inode->i_size) {
1787
		ext4_truncate_failed_write(inode);
1788
		/*
1789
		 * If truncate failed early the inode might still be
1790 1791 1792 1793 1794 1795 1796
		 * 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
1797
	return ret ? ret : copied;
1798 1799
}

N
Nick Piggin 已提交
1800
static int ext4_journalled_write_end(struct file *file,
1801 1802 1803
				     struct address_space *mapping,
				     loff_t pos, unsigned len, unsigned copied,
				     struct page *page, void *fsdata)
1804
{
1805
	handle_t *handle = ext4_journal_current_handle();
N
Nick Piggin 已提交
1806
	struct inode *inode = mapping->host;
1807 1808
	int ret = 0, ret2;
	int partial = 0;
N
Nick Piggin 已提交
1809
	unsigned from, to;
1810
	loff_t new_i_size;
1811

1812
	trace_ext4_journalled_write_end(inode, pos, len, copied);
N
Nick Piggin 已提交
1813 1814 1815 1816 1817 1818 1819 1820
	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);
	}
1821 1822

	ret = walk_page_buffers(handle, page_buffers(page), from,
N
Nick Piggin 已提交
1823
				to, &partial, write_end_fn);
1824 1825
	if (!partial)
		SetPageUptodate(page);
1826 1827
	new_i_size = pos + copied;
	if (new_i_size > inode->i_size)
N
Nick Piggin 已提交
1828
		i_size_write(inode, pos+copied);
1829
	ext4_set_inode_state(inode, EXT4_STATE_JDATA);
1830 1831
	if (new_i_size > EXT4_I(inode)->i_disksize) {
		ext4_update_i_disksize(inode, new_i_size);
1832
		ret2 = ext4_mark_inode_dirty(handle, inode);
1833 1834 1835
		if (!ret)
			ret = ret2;
	}
N
Nick Piggin 已提交
1836

1837
	unlock_page(page);
1838
	page_cache_release(page);
1839
	if (pos + len > inode->i_size && ext4_can_truncate(inode))
1840 1841 1842 1843 1844 1845
		/* 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);

1846
	ret2 = ext4_journal_stop(handle);
1847 1848
	if (!ret)
		ret = ret2;
1849
	if (pos + len > inode->i_size) {
1850
		ext4_truncate_failed_write(inode);
1851
		/*
1852
		 * If truncate failed early the inode might still be
1853 1854 1855 1856 1857 1858
		 * 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
1859 1860

	return ret ? ret : copied;
1861
}
1862

1863 1864 1865 1866
/*
 * Reserve a single block located at lblock
 */
static int ext4_da_reserve_space(struct inode *inode, sector_t lblock)
1867
{
A
Aneesh Kumar K.V 已提交
1868
	int retries = 0;
1869
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1870
	struct ext4_inode_info *ei = EXT4_I(inode);
1871
	unsigned long md_needed;
1872
	int ret;
1873 1874 1875 1876 1877 1878

	/*
	 * 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
1879
repeat:
1880
	spin_lock(&ei->i_block_reservation_lock);
1881
	md_needed = ext4_calc_metadata_amount(inode, lblock);
1882
	trace_ext4_da_reserve_space(inode, md_needed);
1883
	spin_unlock(&ei->i_block_reservation_lock);
1884

1885
	/*
1886 1887 1888
	 * We will charge metadata quota at writeout time; this saves
	 * us from metadata over-estimation, though we may go over by
	 * a small amount in the end.  Here we just reserve for data.
1889
	 */
1890
	ret = dquot_reserve_block(inode, 1);
1891 1892
	if (ret)
		return ret;
1893 1894 1895 1896
	/*
	 * We do still charge estimated metadata to the sb though;
	 * we cannot afford to run out of free blocks.
	 */
1897
	if (ext4_claim_free_blocks(sbi, md_needed + 1)) {
1898
		dquot_release_reservation_block(inode, 1);
A
Aneesh Kumar K.V 已提交
1899 1900 1901 1902
		if (ext4_should_retry_alloc(inode->i_sb, &retries)) {
			yield();
			goto repeat;
		}
1903 1904
		return -ENOSPC;
	}
1905
	spin_lock(&ei->i_block_reservation_lock);
1906
	ei->i_reserved_data_blocks++;
1907 1908
	ei->i_reserved_meta_blocks += md_needed;
	spin_unlock(&ei->i_block_reservation_lock);
1909

1910 1911 1912
	return 0;       /* success */
}

1913
static void ext4_da_release_space(struct inode *inode, int to_free)
1914 1915
{
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1916
	struct ext4_inode_info *ei = EXT4_I(inode);
1917

1918 1919 1920
	if (!to_free)
		return;		/* Nothing to release, exit */

1921
	spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
1922

L
Li Zefan 已提交
1923
	trace_ext4_da_release_space(inode, to_free);
1924
	if (unlikely(to_free > ei->i_reserved_data_blocks)) {
1925
		/*
1926 1927 1928 1929
		 * 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.
1930
		 */
1931 1932 1933 1934 1935 1936
		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;
1937
	}
1938
	ei->i_reserved_data_blocks -= to_free;
1939

1940 1941 1942 1943 1944 1945
	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.
		 */
1946 1947
		percpu_counter_sub(&sbi->s_dirtyblocks_counter,
				   ei->i_reserved_meta_blocks);
1948
		ei->i_reserved_meta_blocks = 0;
1949
		ei->i_da_metadata_calc_len = 0;
1950
	}
1951

1952
	/* update fs dirty data blocks counter */
1953
	percpu_counter_sub(&sbi->s_dirtyblocks_counter, to_free);
1954 1955

	spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
1956

1957
	dquot_release_reservation_block(inode, to_free);
1958 1959 1960
}

static void ext4_da_page_release_reservation(struct page *page,
1961
					     unsigned long offset)
1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977
{
	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);
1978
	ext4_da_release_space(page->mapping->host, to_release);
1979
}
1980

1981 1982 1983 1984 1985 1986
/*
 * Delayed allocation stuff
 */

/*
 * mpage_da_submit_io - walks through extent of pages and try to write
1987
 * them with writepage() call back
1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
 *
 * @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)
{
2000
	long pages_skipped;
2001 2002 2003 2004 2005
	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;
2006 2007

	BUG_ON(mpd->next_page <= mpd->first_page);
2008 2009 2010
	/*
	 * We need to start from the first_page to the next_page - 1
	 * to make sure we also write the mapped dirty buffer_heads.
2011
	 * If we look at mpd->b_blocknr we would only be looking
2012 2013
	 * at the currently mapped buffer_heads.
	 */
2014 2015 2016
	index = mpd->first_page;
	end = mpd->next_page - 1;

2017
	pagevec_init(&pvec, 0);
2018
	while (index <= end) {
2019
		nr_pages = pagevec_lookup(&pvec, mapping, index, PAGEVEC_SIZE);
2020 2021 2022 2023 2024
		if (nr_pages == 0)
			break;
		for (i = 0; i < nr_pages; i++) {
			struct page *page = pvec.pages[i];

2025 2026 2027 2028 2029 2030 2031 2032
			index = page->index;
			if (index > end)
				break;
			index++;

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

2033
			pages_skipped = mpd->wbc->pages_skipped;
2034
			err = mapping->a_ops->writepage(page, mpd->wbc);
2035 2036 2037 2038 2039
			if (!err && (pages_skipped == mpd->wbc->pages_skipped))
				/*
				 * have successfully written the page
				 * without skipping the same
				 */
2040
				mpd->pages_written++;
2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057
			/*
			 * 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
 *
 * the function goes through all passed space and put actual disk
2058
 * block numbers into buffer heads, dropping BH_Delay and BH_Unwritten
2059
 */
2060 2061
static void mpage_put_bnr_to_bhs(struct mpage_da_data *mpd,
				 struct ext4_map_blocks *map)
2062 2063 2064
{
	struct inode *inode = mpd->inode;
	struct address_space *mapping = inode->i_mapping;
2065 2066
	int blocks = map->m_len;
	sector_t pblock = map->m_pblk, cur_logical;
2067
	struct buffer_head *head, *bh;
2068
	pgoff_t index, end;
2069 2070 2071
	struct pagevec pvec;
	int nr_pages, i;

2072 2073
	index = map->m_lblk >> (PAGE_CACHE_SHIFT - inode->i_blkbits);
	end = (map->m_lblk + blocks - 1) >> (PAGE_CACHE_SHIFT - inode->i_blkbits);
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
	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 {
2100
				if (cur_logical >= map->m_lblk)
2101 2102 2103 2104 2105
					break;
				cur_logical++;
			} while ((bh = bh->b_this_page) != head);

			do {
2106
				if (cur_logical >= map->m_lblk + blocks)
2107
					break;
2108

2109
				if (buffer_delay(bh) || buffer_unwritten(bh)) {
2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124

					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);
					}

2125
				} else if (buffer_mapped(bh))
2126 2127
					BUG_ON(bh->b_blocknr != pblock);

2128
				if (map->m_flags & EXT4_MAP_UNINIT)
2129
					set_buffer_uninit(bh);
2130 2131 2132 2133 2134 2135 2136 2137 2138
				cur_logical++;
				pblock++;
			} while ((bh = bh->b_this_page) != head);
		}
		pagevec_release(&pvec);
	}
}


2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156
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];
2157
			if (page->index > end)
2158 2159 2160 2161 2162 2163 2164
				break;
			BUG_ON(!PageLocked(page));
			BUG_ON(PageWriteback(page));
			block_invalidatepage(page, 0);
			ClearPageUptodate(page);
			unlock_page(page);
		}
2165 2166
		index = pvec.pages[nr_pages - 1]->index + 1;
		pagevec_release(&pvec);
2167 2168 2169 2170
	}
	return;
}

2171 2172 2173
static void ext4_print_free_blocks(struct inode *inode)
{
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185
	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);
2186 2187 2188
	return;
}

2189 2190 2191
/*
 * mpage_da_map_blocks - go through given space
 *
2192
 * @mpd - bh describing space
2193 2194 2195 2196
 *
 * The function skips space we know is already mapped to disk blocks.
 *
 */
2197
static int mpage_da_map_blocks(struct mpage_da_data *mpd)
2198
{
2199
	int err, blks, get_blocks_flags;
2200
	struct ext4_map_blocks map;
2201 2202 2203 2204
	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;
2205 2206 2207 2208

	/*
	 * We consider only non-mapped and non-allocated blocks
	 */
2209
	if ((mpd->b_state  & (1 << BH_Mapped)) &&
2210 2211
		!(mpd->b_state & (1 << BH_Delay)) &&
		!(mpd->b_state & (1 << BH_Unwritten)))
2212
		return 0;
2213 2214 2215 2216 2217 2218 2219 2220 2221 2222

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

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

2223
	/*
2224
	 * Call ext4_map_blocks() to allocate any delayed allocation
2225 2226 2227 2228 2229 2230 2231 2232
	 * 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
2233
	 * want to change *many* call functions, so ext4_map_blocks()
2234 2235 2236 2237 2238 2239
	 * 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.
2240
	 */
2241 2242
	map.m_lblk = next;
	map.m_len = max_blocks;
2243
	get_blocks_flags = EXT4_GET_BLOCKS_CREATE;
2244 2245
	if (ext4_should_dioread_nolock(mpd->inode))
		get_blocks_flags |= EXT4_GET_BLOCKS_IO_CREATE_EXT;
2246
	if (mpd->b_state & (1 << BH_Delay))
2247 2248
		get_blocks_flags |= EXT4_GET_BLOCKS_DELALLOC_RESERVE;

2249
	blks = ext4_map_blocks(handle, mpd->inode, &map, get_blocks_flags);
2250
	if (blks < 0) {
2251 2252
		struct super_block *sb = mpd->inode->i_sb;

2253
		err = blks;
2254 2255 2256 2257
		/*
		 * If get block returns with error we simply
		 * return. Later writepage will redirty the page and
		 * writepages will find the dirty page again
2258 2259 2260
		 */
		if (err == -EAGAIN)
			return 0;
2261 2262

		if (err == -ENOSPC &&
2263
		    ext4_count_free_blocks(sb)) {
2264 2265 2266 2267
			mpd->retval = err;
			return 0;
		}

2268
		/*
2269 2270 2271 2272 2273
		 * 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.
2274
		 */
2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285
		if (!(EXT4_SB(sb)->s_mount_flags & EXT4_MF_FS_ABORTED)) {
			ext4_msg(sb, KERN_CRIT,
				 "delayed block allocation failed for inode %lu "
				 "at logical offset %llu with max blocks %zd "
				 "with error %d", mpd->inode->i_ino,
				 (unsigned long long) next,
				 mpd->b_size >> mpd->inode->i_blkbits, err);
			ext4_msg(sb, KERN_CRIT,
				"This should not happen!! Data will be lost\n");
			if (err == -ENOSPC)
				ext4_print_free_blocks(mpd->inode);
A
Aneesh Kumar K.V 已提交
2286
		}
2287
		/* invalidate all the pages */
2288
		ext4_da_block_invalidatepages(mpd, next,
2289
				mpd->b_size >> mpd->inode->i_blkbits);
2290 2291
		return err;
	}
2292 2293
	BUG_ON(blks == 0);

2294 2295 2296
	if (map.m_flags & EXT4_MAP_NEW) {
		struct block_device *bdev = mpd->inode->i_sb->s_bdev;
		int i;
2297

2298 2299 2300
		for (i = 0; i < map.m_len; i++)
			unmap_underlying_metadata(bdev, map.m_pblk + i);
	}
2301

2302 2303 2304 2305
	/*
	 * If blocks are delayed marked, we need to
	 * put actual blocknr and drop delayed bit
	 */
2306 2307
	if ((mpd->b_state & (1 << BH_Delay)) ||
	    (mpd->b_state & (1 << BH_Unwritten)))
2308
		mpage_put_bnr_to_bhs(mpd, &map);
2309

2310 2311 2312 2313 2314 2315 2316
	if (ext4_should_order_data(mpd->inode)) {
		err = ext4_jbd2_file_inode(handle, mpd->inode);
		if (err)
			return err;
	}

	/*
2317
	 * Update on-disk size along with block allocation.
2318 2319 2320 2321 2322 2323 2324 2325 2326
	 */
	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);
	}

2327
	return 0;
2328 2329
}

2330 2331
#define BH_FLAGS ((1 << BH_Uptodate) | (1 << BH_Mapped) | \
		(1 << BH_Delay) | (1 << BH_Unwritten))
2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342

/*
 * 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,
2343 2344
				   sector_t logical, size_t b_size,
				   unsigned long b_state)
2345 2346
{
	sector_t next;
2347
	int nrblocks = mpd->b_size >> mpd->inode->i_blkbits;
2348

2349 2350 2351 2352
	/*
	 * XXX Don't go larger than mballoc is willing to allocate
	 * This is a stopgap solution.  We eventually need to fold
	 * mpage_da_submit_io() into this function and then call
2353
	 * ext4_map_blocks() multiple times in a loop
2354 2355 2356 2357
	 */
	if (nrblocks >= 8*1024*1024/mpd->inode->i_sb->s_blocksize)
		goto flush_it;

2358
	/* check if thereserved journal credits might overflow */
2359
	if (!(ext4_test_inode_flag(mpd->inode, EXT4_INODE_EXTENTS))) {
2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379
		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 */
		}
	}
2380 2381 2382
	/*
	 * First block in the extent
	 */
2383 2384 2385 2386
	if (mpd->b_size == 0) {
		mpd->b_blocknr = logical;
		mpd->b_size = b_size;
		mpd->b_state = b_state & BH_FLAGS;
2387 2388 2389
		return;
	}

2390
	next = mpd->b_blocknr + nrblocks;
2391 2392 2393
	/*
	 * Can we merge the block to our big extent?
	 */
2394 2395
	if (logical == next && (b_state & BH_FLAGS) == mpd->b_state) {
		mpd->b_size += b_size;
2396 2397 2398
		return;
	}

2399
flush_it:
2400 2401 2402 2403
	/*
	 * We couldn't merge the block to our extent, so we
	 * need to flush current  extent and start new one
	 */
2404 2405
	if (mpage_da_map_blocks(mpd) == 0)
		mpage_da_submit_io(mpd);
2406 2407
	mpd->io_done = 1;
	return;
2408 2409
}

2410
static int ext4_bh_delay_or_unwritten(handle_t *handle, struct buffer_head *bh)
2411
{
2412
	return (buffer_delay(bh) || buffer_unwritten(bh)) && buffer_dirty(bh);
2413 2414
}

2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428
/*
 * __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;
2429
	struct buffer_head *bh, *head;
2430 2431 2432 2433 2434 2435 2436 2437
	sector_t logical;

	/*
	 * Can we merge this page to current extent?
	 */
	if (mpd->next_page != page->index) {
		/*
		 * Nope, we can't. So, we map non-allocated blocks
2438
		 * and start IO on them using writepage()
2439 2440
		 */
		if (mpd->next_page != mpd->first_page) {
2441 2442
			if (mpage_da_map_blocks(mpd) == 0)
				mpage_da_submit_io(mpd);
2443 2444 2445 2446 2447 2448 2449
			/*
			 * 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;
2450 2451 2452 2453 2454 2455 2456 2457 2458 2459
		}

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

		/*
		 * ... and blocks
		 */
2460 2461 2462
		mpd->b_size = 0;
		mpd->b_state = 0;
		mpd->b_blocknr = 0;
2463 2464 2465 2466 2467 2468 2469
	}

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

	if (!page_has_buffers(page)) {
2470 2471
		mpage_add_bh_to_extent(mpd, logical, PAGE_CACHE_SIZE,
				       (1 << BH_Dirty) | (1 << BH_Uptodate));
2472 2473
		if (mpd->io_done)
			return MPAGE_DA_EXTENT_TAIL;
2474 2475 2476 2477 2478 2479 2480 2481
	} else {
		/*
		 * Page with regular buffer heads, just add all dirty ones
		 */
		head = page_buffers(page);
		bh = head;
		do {
			BUG_ON(buffer_locked(bh));
2482 2483 2484 2485
			/*
			 * We need to try to allocate
			 * unmapped blocks in the same page.
			 * Otherwise we won't make progress
2486
			 * with the page in ext4_writepage
2487
			 */
2488
			if (ext4_bh_delay_or_unwritten(NULL, bh)) {
2489 2490 2491
				mpage_add_bh_to_extent(mpd, logical,
						       bh->b_size,
						       bh->b_state);
2492 2493
				if (mpd->io_done)
					return MPAGE_DA_EXTENT_TAIL;
2494 2495 2496 2497 2498 2499 2500 2501 2502
			} 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.
				 */
2503 2504
				if (mpd->b_size == 0)
					mpd->b_state = bh->b_state & BH_FLAGS;
2505
			}
2506 2507 2508 2509 2510 2511 2512 2513
			logical++;
		} while ((bh = bh->b_this_page) != head);
	}

	return 0;
}

/*
2514 2515 2516
 * 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.
2517 2518 2519 2520 2521 2522 2523
 *
 * 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.
2524 2525
 */
static int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2526
				  struct buffer_head *bh, int create)
2527
{
2528
	struct ext4_map_blocks map;
2529
	int ret = 0;
2530 2531 2532 2533
	sector_t invalid_block = ~((sector_t) 0xffff);

	if (invalid_block < ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es))
		invalid_block = ~0;
2534 2535

	BUG_ON(create == 0);
2536 2537 2538 2539
	BUG_ON(bh->b_size != inode->i_sb->s_blocksize);

	map.m_lblk = iblock;
	map.m_len = 1;
2540 2541 2542 2543 2544 2545

	/*
	 * first, we need to know whether the block is allocated already
	 * preallocated blocks are unmapped but should treated
	 * the same as allocated blocks.
	 */
2546 2547 2548 2549 2550 2551
	ret = ext4_map_blocks(NULL, inode, &map, 0);
	if (ret < 0)
		return ret;
	if (ret == 0) {
		if (buffer_delay(bh))
			return 0; /* Not sure this could or should happen */
2552
		/*
C
Christoph Hellwig 已提交
2553
		 * XXX: __block_write_begin() unmaps passed block, is it OK?
2554
		 */
2555
		ret = ext4_da_reserve_space(inode, iblock);
2556 2557 2558 2559
		if (ret)
			/* not enough space to reserve */
			return ret;

2560 2561 2562 2563
		map_bh(bh, inode->i_sb, invalid_block);
		set_buffer_new(bh);
		set_buffer_delay(bh);
		return 0;
2564 2565
	}

2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579
	map_bh(bh, inode->i_sb, map.m_pblk);
	bh->b_state = (bh->b_state & ~EXT4_MAP_FLAGS) | map.m_flags;

	if (buffer_unwritten(bh)) {
		/* 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.
		 */
		set_buffer_new(bh);
		set_buffer_mapped(bh);
	}
	return 0;
2580
}
2581

2582 2583 2584
/*
 * 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
C
Christoph Hellwig 已提交
2585
 * callback function for block_write_begin() and block_write_full_page().
2586
 * These functions should only try to map a single block at a time.
2587 2588 2589 2590 2591
 *
 * 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
2592 2593 2594
 * delayed allocation before calling  block_write_full_page().  Otherwise,
 * b_blocknr could be left unitialized, and the page write functions will
 * be taken by surprise.
2595 2596
 */
static int noalloc_get_block_write(struct inode *inode, sector_t iblock,
2597 2598
				   struct buffer_head *bh_result, int create)
{
2599
	BUG_ON(bh_result->b_size != inode->i_sb->s_blocksize);
2600
	return _ext4_get_block(inode, iblock, bh_result, 0);
2601 2602
}

2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649
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);
2650
	ext4_set_inode_state(inode, EXT4_STATE_JDATA);
2651 2652 2653 2654
out:
	return ret;
}

2655 2656 2657
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);

2658
/*
2659 2660 2661 2662 2663 2664 2665 2666 2667
 * 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.
 *
2668 2669 2670 2671 2672
 * 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)
2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697
 *
 * 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.
2698
 */
2699
static int ext4_writepage(struct page *page,
2700
			  struct writeback_control *wbc)
2701 2702
{
	int ret = 0;
2703
	loff_t size;
2704
	unsigned int len;
2705
	struct buffer_head *page_bufs = NULL;
2706 2707
	struct inode *inode = page->mapping->host;

2708
	trace_ext4_writepage(inode, page);
2709 2710 2711 2712 2713
	size = i_size_read(inode);
	if (page->index == size >> PAGE_CACHE_SHIFT)
		len = size & ~PAGE_CACHE_MASK;
	else
		len = PAGE_CACHE_SIZE;
2714

2715
	if (page_has_buffers(page)) {
2716
		page_bufs = page_buffers(page);
2717
		if (walk_page_buffers(NULL, page_bufs, 0, len, NULL,
2718
					ext4_bh_delay_or_unwritten)) {
2719
			/*
2720 2721
			 * We don't want to do  block allocation
			 * So redirty the page and return
2722 2723 2724
			 * 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
2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744
			 * 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.
		 */
C
Christoph Hellwig 已提交
2745
		ret = __block_write_begin(page, 0, len,
2746
					  noalloc_get_block_write);
2747 2748 2749 2750
		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,
2751
						ext4_bh_delay_or_unwritten)) {
2752 2753 2754 2755 2756 2757 2758 2759 2760
				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
2761 2762 2763 2764 2765
			 */
			redirty_page_for_writepage(wbc, page);
			unlock_page(page);
			return 0;
		}
2766
		/* now mark the buffer_heads as dirty and uptodate */
2767
		block_commit_write(page, 0, len);
2768 2769
	}

2770 2771 2772 2773 2774 2775
	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);
2776
		return __ext4_journalled_writepage(page, len);
2777 2778
	}

2779
	if (page_bufs && buffer_uninit(page_bufs)) {
2780 2781 2782 2783
		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
2784 2785
		ret = block_write_full_page(page, noalloc_get_block_write,
					    wbc);
2786 2787 2788 2789

	return ret;
}

2790
/*
2791 2792 2793 2794 2795
 * 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.
2796
 */
2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807

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
	 */
2808
	if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) &&
2809 2810 2811 2812 2813
	    (max_blocks > EXT4_MAX_TRANS_DATA))
		max_blocks = EXT4_MAX_TRANS_DATA;

	return ext4_chunk_trans_blocks(inode, max_blocks);
}
2814

2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932
/*
 * write_cache_pages_da - walk the list of dirty pages of the given
 * address space and call the callback function (which usually writes
 * the pages).
 *
 * This is a forked version of write_cache_pages().  Differences:
 *	Range cyclic is ignored.
 *	no_nrwrite_index_update is always presumed true
 */
static int write_cache_pages_da(struct address_space *mapping,
				struct writeback_control *wbc,
				struct mpage_da_data *mpd)
{
	int ret = 0;
	int done = 0;
	struct pagevec pvec;
	int nr_pages;
	pgoff_t index;
	pgoff_t end;		/* Inclusive */
	long nr_to_write = wbc->nr_to_write;

	pagevec_init(&pvec, 0);
	index = wbc->range_start >> PAGE_CACHE_SHIFT;
	end = wbc->range_end >> PAGE_CACHE_SHIFT;

	while (!done && (index <= end)) {
		int i;

		nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
			      PAGECACHE_TAG_DIRTY,
			      min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1);
		if (nr_pages == 0)
			break;

		for (i = 0; i < nr_pages; i++) {
			struct page *page = pvec.pages[i];

			/*
			 * At this point, the page may be truncated or
			 * invalidated (changing page->mapping to NULL), or
			 * even swizzled back from swapper_space to tmpfs file
			 * mapping. However, page->index will not change
			 * because we have a reference on the page.
			 */
			if (page->index > end) {
				done = 1;
				break;
			}

			lock_page(page);

			/*
			 * Page truncated or invalidated. We can freely skip it
			 * then, even for data integrity operations: the page
			 * has disappeared concurrently, so there could be no
			 * real expectation of this data interity operation
			 * even if there is now a new, dirty page at the same
			 * pagecache address.
			 */
			if (unlikely(page->mapping != mapping)) {
continue_unlock:
				unlock_page(page);
				continue;
			}

			if (!PageDirty(page)) {
				/* someone wrote it for us */
				goto continue_unlock;
			}

			if (PageWriteback(page)) {
				if (wbc->sync_mode != WB_SYNC_NONE)
					wait_on_page_writeback(page);
				else
					goto continue_unlock;
			}

			BUG_ON(PageWriteback(page));
			if (!clear_page_dirty_for_io(page))
				goto continue_unlock;

			ret = __mpage_da_writepage(page, wbc, mpd);
			if (unlikely(ret)) {
				if (ret == AOP_WRITEPAGE_ACTIVATE) {
					unlock_page(page);
					ret = 0;
				} else {
					done = 1;
					break;
				}
			}

			if (nr_to_write > 0) {
				nr_to_write--;
				if (nr_to_write == 0 &&
				    wbc->sync_mode == WB_SYNC_NONE) {
					/*
					 * We stop writing back only if we are
					 * not doing integrity sync. In case of
					 * integrity sync we have to keep going
					 * because someone may be concurrently
					 * dirtying pages, and we might have
					 * synced a lot of newly appeared dirty
					 * pages, but have not synced all of the
					 * old dirty pages.
					 */
					done = 1;
					break;
				}
			}
		}
		pagevec_release(&pvec);
		cond_resched();
	}
	return ret;
}


2933
static int ext4_da_writepages(struct address_space *mapping,
2934
			      struct writeback_control *wbc)
2935
{
2936 2937
	pgoff_t	index;
	int range_whole = 0;
2938
	handle_t *handle = NULL;
2939
	struct mpage_da_data mpd;
2940
	struct inode *inode = mapping->host;
2941 2942
	int pages_written = 0;
	long pages_skipped;
2943
	unsigned int max_pages;
2944
	int range_cyclic, cycled = 1, io_done = 0;
2945 2946
	int needed_blocks, ret = 0;
	long desired_nr_to_write, nr_to_writebump = 0;
2947
	loff_t range_start = wbc->range_start;
2948
	struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb);
2949

2950
	trace_ext4_da_writepages(inode, wbc);
2951

2952 2953 2954 2955 2956
	/*
	 * 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
	 */
2957
	if (!mapping->nrpages || !mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
2958
		return 0;
2959 2960 2961 2962 2963

	/*
	 * 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
2964
	 * EXT4_MF_FS_ABORTED instead of sb->s_flag's MS_RDONLY because
2965 2966 2967 2968 2969
	 * 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.
	 */
2970
	if (unlikely(sbi->s_mount_flags & EXT4_MF_FS_ABORTED))
2971 2972
		return -EROFS;

2973 2974
	if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
		range_whole = 1;
2975

2976 2977
	range_cyclic = wbc->range_cyclic;
	if (wbc->range_cyclic) {
2978
		index = mapping->writeback_index;
2979 2980 2981 2982 2983 2984
		if (index)
			cycled = 0;
		wbc->range_start = index << PAGE_CACHE_SHIFT;
		wbc->range_end  = LLONG_MAX;
		wbc->range_cyclic = 0;
	} else
2985
		index = wbc->range_start >> PAGE_CACHE_SHIFT;
2986

2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016
	/*
	 * 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;
	}

3017 3018 3019
	mpd.wbc = wbc;
	mpd.inode = mapping->host;

3020 3021
	pages_skipped = wbc->pages_skipped;

3022
retry:
3023
	while (!ret && wbc->nr_to_write > 0) {
3024 3025 3026 3027 3028 3029 3030 3031

		/*
		 * 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));
3032
		needed_blocks = ext4_da_writepages_trans_blocks(inode);
3033

3034 3035 3036 3037
		/* start a new transaction*/
		handle = ext4_journal_start(inode, needed_blocks);
		if (IS_ERR(handle)) {
			ret = PTR_ERR(handle);
3038
			ext4_msg(inode->i_sb, KERN_CRIT, "%s: jbd2_start: "
3039
			       "%ld pages, ino %lu; err %d", __func__,
3040
				wbc->nr_to_write, inode->i_ino, ret);
3041 3042
			goto out_writepages;
		}
3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060

		/*
		 * 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;
3061
		ret = write_cache_pages_da(mapping, wbc, &mpd);
3062
		/*
3063
		 * If we have a contiguous extent of pages and we
3064 3065 3066 3067 3068 3069 3070 3071 3072
		 * 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;
		}
3073
		trace_ext4_da_write_pages(inode, &mpd);
3074
		wbc->nr_to_write -= mpd.pages_written;
3075

3076
		ext4_journal_stop(handle);
3077

3078
		if ((mpd.retval == -ENOSPC) && sbi->s_journal) {
3079 3080 3081 3082
			/* commit the transaction which would
			 * free blocks released in the transaction
			 * and try again
			 */
3083
			jbd2_journal_force_commit_nested(sbi->s_journal);
3084 3085 3086
			wbc->pages_skipped = pages_skipped;
			ret = 0;
		} else if (ret == MPAGE_DA_EXTENT_TAIL) {
3087 3088 3089 3090
			/*
			 * got one extent now try with
			 * rest of the pages
			 */
3091 3092
			pages_written += mpd.pages_written;
			wbc->pages_skipped = pages_skipped;
3093
			ret = 0;
3094
			io_done = 1;
3095
		} else if (wbc->nr_to_write)
3096 3097 3098 3099 3100 3101
			/*
			 * There is no more writeout needed
			 * or we requested for a noblocking writeout
			 * and we found the device congested
			 */
			break;
3102
	}
3103 3104 3105 3106 3107 3108 3109
	if (!io_done && !cycled) {
		cycled = 1;
		index = 0;
		wbc->range_start = index << PAGE_CACHE_SHIFT;
		wbc->range_end  = mapping->writeback_index - 1;
		goto retry;
	}
3110
	if (pages_skipped != wbc->pages_skipped)
3111 3112
		ext4_msg(inode->i_sb, KERN_CRIT,
			 "This should not happen leaving %s "
3113
			 "with nr_to_write = %ld ret = %d",
3114
			 __func__, wbc->nr_to_write, ret);
3115 3116 3117

	/* Update index */
	index += pages_written;
3118
	wbc->range_cyclic = range_cyclic;
3119 3120 3121 3122 3123 3124
	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;
3125

3126
out_writepages:
3127
	wbc->nr_to_write -= nr_to_writebump;
3128
	wbc->range_start = range_start;
3129
	trace_ext4_da_writepages_result(inode, wbc, ret, pages_written);
3130
	return ret;
3131 3132
}

3133 3134 3135 3136 3137 3138 3139 3140 3141
#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
3142
	 * counters can get slightly wrong with percpu_counter_batch getting
3143 3144 3145 3146 3147 3148 3149 3150 3151
	 * 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)) {
		/*
3152 3153
		 * free block count is less than 150% of dirty blocks
		 * or free blocks is less than watermark
3154 3155 3156
		 */
		return 1;
	}
3157 3158 3159 3160 3161 3162 3163
	/*
	 * 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);

3164 3165 3166
	return 0;
}

3167
static int ext4_da_write_begin(struct file *file, struct address_space *mapping,
3168 3169
			       loff_t pos, unsigned len, unsigned flags,
			       struct page **pagep, void **fsdata)
3170
{
3171
	int ret, retries = 0;
3172 3173 3174 3175 3176 3177
	struct page *page;
	pgoff_t index;
	struct inode *inode = mapping->host;
	handle_t *handle;

	index = pos >> PAGE_CACHE_SHIFT;
3178 3179 3180 3181 3182 3183 3184

	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;
3185
	trace_ext4_da_write_begin(inode, pos, len, flags);
3186
retry:
3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197
	/*
	 * 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;
	}
3198 3199 3200
	/* We cannot recurse into the filesystem as the transaction is already
	 * started */
	flags |= AOP_FLAG_NOFS;
3201

3202
	page = grab_cache_page_write_begin(mapping, index, flags);
3203 3204 3205 3206 3207
	if (!page) {
		ext4_journal_stop(handle);
		ret = -ENOMEM;
		goto out;
	}
3208 3209
	*pagep = page;

3210
	ret = __block_write_begin(page, pos, len, ext4_da_get_block_prep);
3211 3212 3213 3214
	if (ret < 0) {
		unlock_page(page);
		ext4_journal_stop(handle);
		page_cache_release(page);
3215 3216 3217 3218 3219 3220
		/*
		 * 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)
3221
			ext4_truncate_failed_write(inode);
3222 3223
	}

3224 3225
	if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
		goto retry;
3226 3227 3228 3229
out:
	return ret;
}

3230 3231 3232 3233 3234
/*
 * 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,
3235
					    unsigned long offset)
3236 3237 3238 3239 3240 3241 3242 3243 3244
{
	struct buffer_head *bh;
	struct inode *inode = page->mapping->host;
	unsigned int idx;
	int i;

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

3245
	for (i = 0; i < idx; i++)
3246 3247
		bh = bh->b_this_page;

3248
	if (!buffer_mapped(bh) || (buffer_delay(bh)) || buffer_unwritten(bh))
3249 3250 3251 3252
		return 0;
	return 1;
}

3253
static int ext4_da_write_end(struct file *file,
3254 3255 3256
			     struct address_space *mapping,
			     loff_t pos, unsigned len, unsigned copied,
			     struct page *page, void *fsdata)
3257 3258 3259 3260 3261
{
	struct inode *inode = mapping->host;
	int ret = 0, ret2;
	handle_t *handle = ext4_journal_current_handle();
	loff_t new_i_size;
3262
	unsigned long start, end;
3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275
	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();
		}
	}
3276

3277
	trace_ext4_da_write_end(inode, pos, len, copied);
3278
	start = pos & (PAGE_CACHE_SIZE - 1);
3279
	end = start + copied - 1;
3280 3281 3282 3283 3284 3285 3286 3287

	/*
	 * 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;
3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298
	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);
3299

3300 3301 3302
				EXT4_I(inode)->i_disksize = new_i_size;
			}
			up_write(&EXT4_I(inode)->i_data_sem);
3303 3304 3305 3306 3307
			/* 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);
3308
		}
3309
	}
3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330
	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;

3331
	ext4_da_page_release_reservation(page, offset);
3332 3333 3334 3335 3336 3337 3338

out:
	ext4_invalidatepage(page, offset);

	return;
}

3339 3340 3341 3342 3343
/*
 * Force all delayed allocation blocks to be allocated for a given inode.
 */
int ext4_alloc_da_blocks(struct inode *inode)
{
3344 3345
	trace_ext4_alloc_da_blocks(inode);

3346 3347 3348 3349 3350 3351 3352 3353 3354 3355
	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:
3356
	 *
3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375
	 * 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.
3376
	 *
3377 3378 3379 3380 3381 3382
	 * 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);
}
3383

3384 3385 3386 3387 3388
/*
 * 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
3389
 * journal.  If somebody makes a swapfile on an ext4 data-journaling
3390 3391 3392 3393 3394 3395 3396 3397
 * 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.
 */
3398
static sector_t ext4_bmap(struct address_space *mapping, sector_t block)
3399 3400 3401 3402 3403
{
	struct inode *inode = mapping->host;
	journal_t *journal;
	int err;

3404 3405 3406 3407 3408 3409 3410 3411 3412 3413
	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);
	}

3414 3415
	if (EXT4_JOURNAL(inode) &&
	    ext4_test_inode_state(inode, EXT4_STATE_JDATA)) {
3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426
		/*
		 * 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.)
		 *
3427
		 * NB. EXT4_STATE_JDATA is not set on files other than
3428 3429 3430 3431 3432 3433
		 * 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.
		 */

3434
		ext4_clear_inode_state(inode, EXT4_STATE_JDATA);
3435
		journal = EXT4_JOURNAL(inode);
3436 3437 3438
		jbd2_journal_lock_updates(journal);
		err = jbd2_journal_flush(journal);
		jbd2_journal_unlock_updates(journal);
3439 3440 3441 3442 3443

		if (err)
			return 0;
	}

3444
	return generic_block_bmap(mapping, block, ext4_get_block);
3445 3446
}

3447
static int ext4_readpage(struct file *file, struct page *page)
3448
{
3449
	return mpage_readpage(page, ext4_get_block);
3450 3451 3452
}

static int
3453
ext4_readpages(struct file *file, struct address_space *mapping,
3454 3455
		struct list_head *pages, unsigned nr_pages)
{
3456
	return mpage_readpages(mapping, pages, nr_pages, ext4_get_block);
3457 3458
}

3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487
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);
}

3488
static void ext4_invalidatepage(struct page *page, unsigned long offset)
3489
{
3490
	journal_t *journal = EXT4_JOURNAL(page->mapping->host);
3491

3492 3493 3494 3495 3496
	/*
	 * 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);
3497 3498 3499 3500 3501 3502
	/*
	 * If it's a full truncate we just forget about the pending dirtying
	 */
	if (offset == 0)
		ClearPageChecked(page);

3503 3504 3505 3506
	if (journal)
		jbd2_journal_invalidatepage(journal, page, offset);
	else
		block_invalidatepage(page, offset);
3507 3508
}

3509
static int ext4_releasepage(struct page *page, gfp_t wait)
3510
{
3511
	journal_t *journal = EXT4_JOURNAL(page->mapping->host);
3512 3513 3514 3515

	WARN_ON(PageChecked(page));
	if (!page_has_buffers(page))
		return 0;
3516 3517 3518 3519
	if (journal)
		return jbd2_journal_try_to_free_buffers(journal, page, wait);
	else
		return try_to_free_buffers(page);
3520 3521 3522
}

/*
3523 3524
 * O_DIRECT for ext3 (or indirect map) based files
 *
3525 3526 3527 3528 3529
 * 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
3530 3531
 * 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.
3532
 */
3533
static ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
3534 3535
			      const struct iovec *iov, loff_t offset,
			      unsigned long nr_segs)
3536 3537 3538
{
	struct file *file = iocb->ki_filp;
	struct inode *inode = file->f_mapping->host;
3539
	struct ext4_inode_info *ei = EXT4_I(inode);
J
Jan Kara 已提交
3540
	handle_t *handle;
3541 3542 3543
	ssize_t ret;
	int orphan = 0;
	size_t count = iov_length(iov, nr_segs);
3544
	int retries = 0;
3545 3546 3547 3548 3549

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

		if (final_size > inode->i_size) {
J
Jan Kara 已提交
3550 3551 3552 3553 3554 3555
			/* Credits for sb + inode write */
			handle = ext4_journal_start(inode, 2);
			if (IS_ERR(handle)) {
				ret = PTR_ERR(handle);
				goto out;
			}
3556
			ret = ext4_orphan_add(handle, inode);
J
Jan Kara 已提交
3557 3558 3559 3560
			if (ret) {
				ext4_journal_stop(handle);
				goto out;
			}
3561 3562
			orphan = 1;
			ei->i_disksize = inode->i_size;
J
Jan Kara 已提交
3563
			ext4_journal_stop(handle);
3564 3565 3566
		}
	}

3567
retry:
3568
	if (rw == READ && ext4_should_dioread_nolock(inode))
3569
		ret = __blockdev_direct_IO(rw, iocb, inode,
3570 3571
				 inode->i_sb->s_bdev, iov,
				 offset, nr_segs,
3572 3573
				 ext4_get_block, NULL, NULL, 0);
	else {
3574 3575
		ret = blockdev_direct_IO(rw, iocb, inode,
				 inode->i_sb->s_bdev, iov,
3576
				 offset, nr_segs,
3577
				 ext4_get_block, NULL);
3578 3579 3580 3581 3582 3583 3584 3585 3586

		if (unlikely((rw & WRITE) && ret < 0)) {
			loff_t isize = i_size_read(inode);
			loff_t end = offset + iov_length(iov, nr_segs);

			if (end > isize)
				vmtruncate(inode, isize);
		}
	}
3587 3588
	if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
		goto retry;
3589

J
Jan Kara 已提交
3590
	if (orphan) {
3591 3592
		int err;

J
Jan Kara 已提交
3593 3594 3595 3596 3597 3598 3599
		/* 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);
3600 3601 3602
			if (inode->i_nlink)
				ext4_orphan_del(NULL, inode);

J
Jan Kara 已提交
3603 3604 3605
			goto out;
		}
		if (inode->i_nlink)
3606
			ext4_orphan_del(handle, inode);
J
Jan Kara 已提交
3607
		if (ret > 0) {
3608 3609 3610 3611 3612 3613 3614 3615
			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
3616
				 * ext4_mark_inode_dirty() to userspace.  So
3617 3618
				 * ignore it.
				 */
3619
				ext4_mark_inode_dirty(handle, inode);
3620 3621
			}
		}
3622
		err = ext4_journal_stop(handle);
3623 3624 3625 3626 3627 3628 3629
		if (ret == 0)
			ret = err;
	}
out:
	return ret;
}

3630 3631 3632 3633 3634
/*
 * ext4_get_block used when preparing 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 the IO is complete.
 */
3635
static int ext4_get_block_write(struct inode *inode, sector_t iblock,
3636 3637
		   struct buffer_head *bh_result, int create)
{
3638
	ext4_debug("ext4_get_block_write: inode %lu, create flag %d\n",
3639
		   inode->i_ino, create);
3640 3641
	return _ext4_get_block(inode, iblock, bh_result,
			       EXT4_GET_BLOCKS_IO_CREATE_EXT);
3642 3643
}

3644
static void dump_completed_IO(struct inode * inode)
3645 3646 3647 3648
{
#ifdef	EXT4_DEBUG
	struct list_head *cur, *before, *after;
	ext4_io_end_t *io, *io0, *io1;
3649
	unsigned long flags;
3650

3651 3652
	if (list_empty(&EXT4_I(inode)->i_completed_io_list)){
		ext4_debug("inode %lu completed_io list is empty\n", inode->i_ino);
3653 3654 3655
		return;
	}

3656
	ext4_debug("Dump inode %lu completed_io list \n", inode->i_ino);
3657
	spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags);
3658
	list_for_each_entry(io, &EXT4_I(inode)->i_completed_io_list, list){
3659 3660 3661 3662 3663 3664 3665 3666 3667
		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);
	}
3668
	spin_unlock_irqrestore(&EXT4_I(inode)->i_completed_io_lock, flags);
3669 3670
#endif
}
3671 3672 3673 3674

/*
 * check a range of space and convert unwritten extents to written.
 */
3675
static int ext4_end_io_nolock(ext4_io_end_t *io)
3676 3677 3678
{
	struct inode *inode = io->inode;
	loff_t offset = io->offset;
3679
	ssize_t size = io->size;
3680 3681
	int ret = 0;

3682
	ext4_debug("ext4_end_io_nolock: io 0x%p from inode %lu,list->next 0x%p,"
3683 3684 3685 3686 3687 3688
		   "list->prev 0x%p\n",
	           io, inode->i_ino, io->list.next, io->list.prev);

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

3689
	if (io->flag != EXT4_IO_UNWRITTEN)
3690 3691
		return ret;

3692
	ret = ext4_convert_unwritten_extents(inode, offset, size);
3693
	if (ret < 0) {
3694
		printk(KERN_EMERG "%s: failed to convert unwritten"
3695 3696 3697 3698 3699
			"extents to written extents, error is %d"
			" io is still on inode %lu aio dio list\n",
                       __func__, ret, inode->i_ino);
		return ret;
	}
3700

3701 3702
	if (io->iocb)
		aio_complete(io->iocb, io->result, 0);
3703 3704 3705
	/* clear the DIO AIO unwritten flag */
	io->flag = 0;
	return ret;
3706
}
3707

3708 3709 3710
/*
 * work on completed aio dio IO, to convert unwritten extents to extents
 */
3711
static void ext4_end_io_work(struct work_struct *work)
3712
{
3713 3714 3715 3716 3717
	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;
3718

3719
	mutex_lock(&inode->i_mutex);
3720
	ret = ext4_end_io_nolock(io);
3721 3722 3723
	if (ret < 0) {
		mutex_unlock(&inode->i_mutex);
		return;
3724
	}
3725 3726 3727 3728 3729

	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);
3730
	mutex_unlock(&inode->i_mutex);
3731
	ext4_free_io_end(io);
3732
}
3733

3734 3735 3736
/*
 * This function is called from ext4_sync_file().
 *
3737 3738
 * When IO is completed, the work to convert unwritten extents to
 * written is queued on workqueue but may not get immediately
3739 3740
 * scheduled. When fsync is called, we need to ensure the
 * conversion is complete before fsync returns.
3741 3742 3743 3744 3745
 * 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.
3746
 */
3747
int flush_completed_IO(struct inode *inode)
3748 3749
{
	ext4_io_end_t *io;
3750 3751
	struct ext4_inode_info *ei = EXT4_I(inode);
	unsigned long flags;
3752 3753 3754
	int ret = 0;
	int ret2 = 0;

3755
	if (list_empty(&ei->i_completed_io_list))
3756 3757
		return ret;

3758
	dump_completed_IO(inode);
3759 3760 3761
	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,
3762 3763
				ext4_io_end_t, list);
		/*
3764
		 * Calling ext4_end_io_nolock() to convert completed
3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776
		 * 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.
		 */
3777
		spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
3778
		ret = ext4_end_io_nolock(io);
3779
		spin_lock_irqsave(&ei->i_completed_io_lock, flags);
3780 3781 3782 3783 3784
		if (ret < 0)
			ret2 = ret;
		else
			list_del_init(&io->list);
	}
3785
	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
3786 3787 3788
	return (ret2 < 0) ? ret2 : 0;
}

3789
static ext4_io_end_t *ext4_init_io_end (struct inode *inode, gfp_t flags)
3790 3791 3792
{
	ext4_io_end_t *io = NULL;

3793
	io = kmalloc(sizeof(*io), flags);
3794 3795

	if (io) {
3796
		igrab(inode);
3797
		io->inode = inode;
3798
		io->flag = 0;
3799 3800
		io->offset = 0;
		io->size = 0;
3801
		io->page = NULL;
3802 3803
		io->iocb = NULL;
		io->result = 0;
3804
		INIT_WORK(&io->work, ext4_end_io_work);
3805
		INIT_LIST_HEAD(&io->list);
3806 3807 3808 3809 3810 3811
	}

	return io;
}

static void ext4_end_io_dio(struct kiocb *iocb, loff_t offset,
3812 3813
			    ssize_t size, void *private, int ret,
			    bool is_async)
3814 3815 3816
{
        ext4_io_end_t *io_end = iocb->private;
	struct workqueue_struct *wq;
3817 3818
	unsigned long flags;
	struct ext4_inode_info *ei;
3819

3820 3821
	/* if not async direct IO or dio with 0 bytes write, just return */
	if (!io_end || !size)
3822
		goto out;
3823

3824 3825 3826 3827 3828 3829
	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 */
3830
	if (io_end->flag != EXT4_IO_UNWRITTEN){
3831 3832
		ext4_free_io_end(io_end);
		iocb->private = NULL;
3833 3834 3835 3836
out:
		if (is_async)
			aio_complete(iocb, ret, 0);
		return;
3837 3838
	}

3839 3840
	io_end->offset = offset;
	io_end->size = size;
3841 3842 3843 3844
	if (is_async) {
		io_end->iocb = iocb;
		io_end->result = ret;
	}
3845 3846
	wq = EXT4_SB(io_end->inode->i_sb)->dio_unwritten_wq;

3847
	/* queue the work to convert unwritten extents to written */
3848 3849
	queue_work(wq, &io_end->work);

3850
	/* Add the io_end to per-inode completed aio dio list*/
3851 3852 3853 3854
	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);
3855 3856
	iocb->private = NULL;
}
3857

3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922
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;
}

3923 3924 3925 3926 3927 3928 3929 3930 3931
/*
 * 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.
 *
3932 3933 3934 3935
 * 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.
3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953
 *
 * 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) {
		/*
3954 3955 3956
 		 * We could direct write to holes and fallocate.
		 *
 		 * Allocated blocks to fill the hole are marked as uninitialized
3957 3958
 		 * to prevent paralel buffered read to expose the stale data
 		 * before DIO complete the data IO.
3959 3960
		 *
 		 * As to previously fallocated extents, ext4 get_block
3961 3962 3963
 		 * will just simply mark the buffer mapped but still
 		 * keep the extents uninitialized.
 		 *
3964 3965 3966 3967 3968 3969 3970 3971
		 * 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.
3972
 		 */
3973 3974 3975
		iocb->private = NULL;
		EXT4_I(inode)->cur_aio_dio = NULL;
		if (!is_sync_kiocb(iocb)) {
3976
			iocb->private = ext4_init_io_end(inode, GFP_NOFS);
3977 3978 3979 3980
			if (!iocb->private)
				return -ENOMEM;
			/*
			 * we save the io structure for current async
3981
			 * direct IO, so that later ext4_map_blocks()
3982 3983 3984 3985 3986 3987 3988
			 * 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;
		}

3989 3990 3991
		ret = blockdev_direct_IO(rw, iocb, inode,
					 inode->i_sb->s_bdev, iov,
					 offset, nr_segs,
3992
					 ext4_get_block_write,
3993
					 ext4_end_io_dio);
3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012
		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;
4013 4014
		} else if (ret > 0 && ext4_test_inode_state(inode,
						EXT4_STATE_DIO_UNWRITTEN)) {
4015
			int err;
4016 4017 4018 4019
			/*
			 * for non AIO case, since the IO is already
			 * completed, we could do the convertion right here
			 */
4020 4021 4022 4023
			err = ext4_convert_unwritten_extents(inode,
							     offset, ret);
			if (err < 0)
				ret = err;
4024
			ext4_clear_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);
4025
		}
4026 4027
		return ret;
	}
4028 4029

	/* for write the the end of file case, we fall back to old way */
4030 4031 4032 4033 4034 4035 4036 4037 4038 4039
	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;

4040
	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
4041 4042 4043 4044 4045
		return ext4_ext_direct_IO(rw, iocb, iov, offset, nr_segs);

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

4046
/*
4047
 * Pages can be marked dirty completely asynchronously from ext4's journalling
4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058
 * 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.
 */
4059
static int ext4_journalled_set_page_dirty(struct page *page)
4060 4061 4062 4063 4064
{
	SetPageChecked(page);
	return __set_page_dirty_nobuffers(page);
}

4065
static const struct address_space_operations ext4_ordered_aops = {
4066 4067
	.readpage		= ext4_readpage,
	.readpages		= ext4_readpages,
4068
	.writepage		= ext4_writepage,
4069 4070 4071 4072 4073 4074 4075 4076 4077
	.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,
4078
	.error_remove_page	= generic_error_remove_page,
4079 4080
};

4081
static const struct address_space_operations ext4_writeback_aops = {
4082 4083
	.readpage		= ext4_readpage,
	.readpages		= ext4_readpages,
4084
	.writepage		= ext4_writepage,
4085 4086 4087 4088 4089 4090 4091 4092 4093
	.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,
4094
	.error_remove_page	= generic_error_remove_page,
4095 4096
};

4097
static const struct address_space_operations ext4_journalled_aops = {
4098 4099
	.readpage		= ext4_readpage,
	.readpages		= ext4_readpages,
4100
	.writepage		= ext4_writepage,
4101 4102 4103 4104 4105 4106 4107 4108
	.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,
4109
	.error_remove_page	= generic_error_remove_page,
4110 4111
};

4112
static const struct address_space_operations ext4_da_aops = {
4113 4114
	.readpage		= ext4_readpage,
	.readpages		= ext4_readpages,
4115
	.writepage		= ext4_writepage,
4116 4117 4118 4119 4120 4121 4122 4123 4124 4125
	.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,
4126
	.error_remove_page	= generic_error_remove_page,
4127 4128
};

4129
void ext4_set_aops(struct inode *inode)
4130
{
4131 4132 4133 4134
	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))
4135
		inode->i_mapping->a_ops = &ext4_ordered_aops;
4136 4137 4138
	else if (ext4_should_writeback_data(inode) &&
		 test_opt(inode->i_sb, DELALLOC))
		inode->i_mapping->a_ops = &ext4_da_aops;
4139 4140
	else if (ext4_should_writeback_data(inode))
		inode->i_mapping->a_ops = &ext4_writeback_aops;
4141
	else
4142
		inode->i_mapping->a_ops = &ext4_journalled_aops;
4143 4144 4145
}

/*
4146
 * ext4_block_truncate_page() zeroes out a mapping from file offset `from'
4147 4148 4149 4150
 * 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.
 */
4151
int ext4_block_truncate_page(handle_t *handle,
4152 4153
		struct address_space *mapping, loff_t from)
{
4154
	ext4_fsblk_t index = from >> PAGE_CACHE_SHIFT;
4155
	unsigned offset = from & (PAGE_CACHE_SIZE-1);
A
Aneesh Kumar K.V 已提交
4156 4157
	unsigned blocksize, length, pos;
	ext4_lblk_t iblock;
4158 4159
	struct inode *inode = mapping->host;
	struct buffer_head *bh;
4160
	struct page *page;
4161 4162
	int err = 0;

4163 4164
	page = find_or_create_page(mapping, from >> PAGE_CACHE_SHIFT,
				   mapping_gfp_mask(mapping) & ~__GFP_FS);
4165 4166 4167
	if (!page)
		return -EINVAL;

4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191
	blocksize = inode->i_sb->s_blocksize;
	length = blocksize - (offset & (blocksize - 1));
	iblock = index << (PAGE_CACHE_SHIFT - inode->i_sb->s_blocksize_bits);

	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");
4192
		ext4_get_block(inode, iblock, bh, 0);
4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212
		/* 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;
	}

4213
	if (ext4_should_journal_data(inode)) {
4214
		BUFFER_TRACE(bh, "get write access");
4215
		err = ext4_journal_get_write_access(handle, bh);
4216 4217 4218 4219
		if (err)
			goto unlock;
	}

4220
	zero_user(page, offset, length);
4221 4222 4223 4224

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

	err = 0;
4225
	if (ext4_should_journal_data(inode)) {
4226
		err = ext4_handle_dirty_metadata(handle, inode, bh);
4227
	} else {
4228
		if (ext4_should_order_data(inode))
4229
			err = ext4_jbd2_file_inode(handle, inode);
4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252
		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;
}

/**
4253
 *	ext4_find_shared - find the indirect blocks for partial truncation.
4254 4255
 *	@inode:	  inode in question
 *	@depth:	  depth of the affected branch
4256
 *	@offsets: offsets of pointers in that branch (see ext4_block_to_path)
4257 4258 4259
 *	@chain:	  place to store the pointers to partial indirect blocks
 *	@top:	  place to the (detached) top of branch
 *
4260
 *	This is a helper function used by ext4_truncate().
4261 4262 4263 4264 4265 4266 4267
 *
 *	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
4268
 *	past the truncation point is possible until ext4_truncate()
4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286
 *	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).  */

4287
static Indirect *ext4_find_shared(struct inode *inode, int depth,
4288 4289
				  ext4_lblk_t offsets[4], Indirect chain[4],
				  __le32 *top)
4290 4291 4292 4293 4294
{
	Indirect *partial, *p;
	int k, err;

	*top = 0;
4295
	/* Make k index the deepest non-null offset + 1 */
4296 4297
	for (k = depth; k > 1 && !offsets[k-1]; k--)
		;
4298
	partial = ext4_get_branch(inode, k, offsets, chain, &err);
4299 4300 4301 4302 4303 4304 4305 4306 4307 4308
	/* 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;
4309
	for (p = partial; (p > chain) && all_zeroes((__le32 *) p->bh->b_data, p->p); p--)
4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320
		;
	/*
	 * 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;
4321
		/* Nope, don't do this in ext4.  Must leave the tree intact */
4322 4323 4324 4325 4326 4327
#if 0
		*p->p = 0;
#endif
	}
	/* Writer: end */

4328
	while (partial > p) {
4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343
		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.
 */
4344 4345 4346 4347 4348
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)
4349 4350
{
	__le32 *p;
4351
	int	flags = EXT4_FREE_BLOCKS_FORGET | EXT4_FREE_BLOCKS_VALIDATED;
4352 4353 4354

	if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
		flags |= EXT4_FREE_BLOCKS_METADATA;
4355

4356 4357
	if (!ext4_data_block_valid(EXT4_SB(inode->i_sb), block_to_free,
				   count)) {
4358 4359 4360
		EXT4_ERROR_INODE(inode, "attempt to clear invalid "
				 "blocks %llu len %lu",
				 (unsigned long long) block_to_free, count);
4361 4362 4363
		return 1;
	}

4364 4365
	if (try_to_extend_transaction(handle, inode)) {
		if (bh) {
4366 4367
			BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
			ext4_handle_dirty_metadata(handle, inode, bh);
4368
		}
4369
		ext4_mark_inode_dirty(handle, inode);
4370 4371
		ext4_truncate_restart_trans(handle, inode,
					    blocks_for_truncate(inode));
4372 4373
		if (bh) {
			BUFFER_TRACE(bh, "retaking write access");
4374
			ext4_journal_get_write_access(handle, bh);
4375 4376 4377
		}
	}

4378 4379
	for (p = first; p < last; p++)
		*p = 0;
4380

4381
	ext4_free_blocks(handle, inode, 0, block_to_free, count, flags);
4382
	return 0;
4383 4384 4385
}

/**
4386
 * ext4_free_data - free a list of data blocks
4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403
 * @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.
 */
4404
static void ext4_free_data(handle_t *handle, struct inode *inode,
4405 4406 4407
			   struct buffer_head *this_bh,
			   __le32 *first, __le32 *last)
{
4408
	ext4_fsblk_t block_to_free = 0;    /* Starting block # of a run */
4409 4410 4411 4412
	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 */
4413
	ext4_fsblk_t nr;		    /* Current block # */
4414 4415 4416 4417 4418 4419
	__le32 *p;			    /* Pointer into inode/ind
					       for current block */
	int err;

	if (this_bh) {				/* For indirect block */
		BUFFER_TRACE(this_bh, "get_write_access");
4420
		err = ext4_journal_get_write_access(handle, this_bh);
4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437
		/* 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 {
4438 4439 4440 4441
				if (ext4_clear_blocks(handle, inode, this_bh,
						      block_to_free, count,
						      block_to_free_p, p))
					break;
4442 4443 4444 4445 4446 4447 4448 4449
				block_to_free = nr;
				block_to_free_p = p;
				count = 1;
			}
		}
	}

	if (count > 0)
4450
		ext4_clear_blocks(handle, inode, this_bh, block_to_free,
4451 4452 4453
				  count, block_to_free_p, p);

	if (this_bh) {
4454
		BUFFER_TRACE(this_bh, "call ext4_handle_dirty_metadata");
4455 4456 4457 4458 4459 4460 4461

		/*
		 * 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.
		 */
4462
		if ((EXT4_JOURNAL(inode) == NULL) || bh2jh(this_bh))
4463
			ext4_handle_dirty_metadata(handle, inode, this_bh);
4464
		else
4465 4466 4467 4468
			EXT4_ERROR_INODE(inode,
					 "circular indirect block detected at "
					 "block %llu",
				(unsigned long long) this_bh->b_blocknr);
4469 4470 4471 4472
	}
}

/**
4473
 *	ext4_free_branches - free an array of branches
4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484
 *	@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.
 */
4485
static void ext4_free_branches(handle_t *handle, struct inode *inode,
4486 4487 4488
			       struct buffer_head *parent_bh,
			       __le32 *first, __le32 *last, int depth)
{
4489
	ext4_fsblk_t nr;
4490 4491
	__le32 *p;

4492
	if (ext4_handle_is_aborted(handle))
4493 4494 4495 4496
		return;

	if (depth--) {
		struct buffer_head *bh;
4497
		int addr_per_block = EXT4_ADDR_PER_BLOCK(inode->i_sb);
4498 4499 4500 4501 4502 4503
		p = last;
		while (--p >= first) {
			nr = le32_to_cpu(*p);
			if (!nr)
				continue;		/* A hole */

4504 4505
			if (!ext4_data_block_valid(EXT4_SB(inode->i_sb),
						   nr, 1)) {
4506 4507 4508 4509
				EXT4_ERROR_INODE(inode,
						 "invalid indirect mapped "
						 "block %lu (level %d)",
						 (unsigned long) nr, depth);
4510 4511 4512
				break;
			}

4513 4514 4515 4516 4517 4518 4519 4520
			/* 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) {
4521 4522
				EXT4_ERROR_INODE_BLOCK(inode, nr,
						       "Read failure");
4523 4524 4525 4526 4527
				continue;
			}

			/* This zaps the entire block.  Bottom up. */
			BUFFER_TRACE(bh, "free child branches");
4528
			ext4_free_branches(handle, inode, bh,
4529 4530 4531
					(__le32 *) bh->b_data,
					(__le32 *) bh->b_data + addr_per_block,
					depth);
4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548

			/*
			 * 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.
			 */
4549
			if (ext4_handle_is_aborted(handle))
4550 4551
				return;
			if (try_to_extend_transaction(handle, inode)) {
4552
				ext4_mark_inode_dirty(handle, inode);
4553 4554
				ext4_truncate_restart_trans(handle, inode,
					    blocks_for_truncate(inode));
4555 4556
			}

4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567
			/*
			 * The forget flag here is critical because if
			 * we are journaling (and not doing data
			 * journaling), we have to make sure a revoke
			 * record is written to prevent the journal
			 * replay from overwriting the (former)
			 * indirect block if it gets reallocated as a
			 * data block.  This must happen in the same
			 * transaction where the data blocks are
			 * actually freed.
			 */
4568
			ext4_free_blocks(handle, inode, 0, nr, 1,
4569 4570
					 EXT4_FREE_BLOCKS_METADATA|
					 EXT4_FREE_BLOCKS_FORGET);
4571 4572 4573 4574 4575 4576 4577

			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");
4578
				if (!ext4_journal_get_write_access(handle,
4579 4580 4581
								   parent_bh)){
					*p = 0;
					BUFFER_TRACE(parent_bh,
4582 4583 4584 4585
					"call ext4_handle_dirty_metadata");
					ext4_handle_dirty_metadata(handle,
								   inode,
								   parent_bh);
4586 4587 4588 4589 4590 4591
				}
			}
		}
	} else {
		/* We have reached the bottom of the tree. */
		BUFFER_TRACE(parent_bh, "free data blocks");
4592
		ext4_free_data(handle, inode, parent_bh, first, last);
4593 4594 4595
	}
}

4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608
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;
}

4609
/*
4610
 * ext4_truncate()
4611
 *
4612 4613
 * We block out ext4_get_block() block instantiations across the entire
 * transaction, and VFS/VM ensures that ext4_truncate() cannot run
4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629
 * 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
4630
 * i_disksize in this case).  After a crash, ext4_orphan_cleanup() will see
4631
 * that this inode's truncate did not complete and it will again call
4632 4633
 * 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
4634
 * that's fine - as long as they are linked from the inode, the post-crash
4635
 * ext4_truncate() run will find them and release them.
4636
 */
4637
void ext4_truncate(struct inode *inode)
4638 4639
{
	handle_t *handle;
4640
	struct ext4_inode_info *ei = EXT4_I(inode);
4641
	__le32 *i_data = ei->i_data;
4642
	int addr_per_block = EXT4_ADDR_PER_BLOCK(inode->i_sb);
4643
	struct address_space *mapping = inode->i_mapping;
A
Aneesh Kumar K.V 已提交
4644
	ext4_lblk_t offsets[4];
4645 4646 4647 4648
	Indirect chain[4];
	Indirect *partial;
	__le32 nr = 0;
	int n;
A
Aneesh Kumar K.V 已提交
4649
	ext4_lblk_t last_block;
4650 4651
	unsigned blocksize = inode->i_sb->s_blocksize;

4652
	if (!ext4_can_truncate(inode))
4653 4654
		return;

4655
	ext4_clear_inode_flag(inode, EXT4_INODE_EOFBLOCKS);
4656

4657
	if (inode->i_size == 0 && !test_opt(inode->i_sb, NO_AUTO_DA_ALLOC))
4658
		ext4_set_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE);
4659

4660
	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
4661
		ext4_ext_truncate(inode);
A
Aneesh Kumar K.V 已提交
4662 4663
		return;
	}
A
Alex Tomas 已提交
4664

4665
	handle = start_transaction(inode);
4666
	if (IS_ERR(handle))
4667 4668 4669
		return;		/* AKPM: return what? */

	last_block = (inode->i_size + blocksize-1)
4670
					>> EXT4_BLOCK_SIZE_BITS(inode->i_sb);
4671

4672 4673 4674
	if (inode->i_size & (blocksize - 1))
		if (ext4_block_truncate_page(handle, mapping, inode->i_size))
			goto out_stop;
4675

4676
	n = ext4_block_to_path(inode, last_block, offsets, NULL);
4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688
	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.
	 */
4689
	if (ext4_orphan_add(handle, inode))
4690 4691
		goto out_stop;

4692 4693 4694 4695 4696
	/*
	 * From here we block out all ext4_get_block() callers who want to
	 * modify the block allocation tree.
	 */
	down_write(&ei->i_data_sem);
4697

4698
	ext4_discard_preallocations(inode);
4699

4700 4701 4702 4703 4704
	/*
	 * 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
4705
	 * ext4 *really* writes onto the disk inode.
4706 4707 4708 4709
	 */
	ei->i_disksize = inode->i_size;

	if (n == 1) {		/* direct blocks */
4710 4711
		ext4_free_data(handle, inode, NULL, i_data+offsets[0],
			       i_data + EXT4_NDIR_BLOCKS);
4712 4713 4714
		goto do_indirects;
	}

4715
	partial = ext4_find_shared(inode, n, offsets, chain, &nr);
4716 4717 4718 4719
	/* Kill the top of shared branch (not detached) */
	if (nr) {
		if (partial == chain) {
			/* Shared branch grows from the inode */
4720
			ext4_free_branches(handle, inode, NULL,
4721 4722 4723 4724 4725 4726 4727 4728 4729
					   &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");
4730
			ext4_free_branches(handle, inode, partial->bh,
4731 4732 4733 4734 4735 4736
					partial->p,
					partial->p+1, (chain+n-1) - partial);
		}
	}
	/* Clear the ends of indirect blocks on the shared branch */
	while (partial > chain) {
4737
		ext4_free_branches(handle, inode, partial->bh, partial->p + 1,
4738 4739 4740
				   (__le32*)partial->bh->b_data+addr_per_block,
				   (chain+n-1) - partial);
		BUFFER_TRACE(partial->bh, "call brelse");
4741
		brelse(partial->bh);
4742 4743 4744 4745 4746 4747
		partial--;
	}
do_indirects:
	/* Kill the remaining (whole) subtrees */
	switch (offsets[0]) {
	default:
4748
		nr = i_data[EXT4_IND_BLOCK];
4749
		if (nr) {
4750 4751
			ext4_free_branches(handle, inode, NULL, &nr, &nr+1, 1);
			i_data[EXT4_IND_BLOCK] = 0;
4752
		}
4753 4754
	case EXT4_IND_BLOCK:
		nr = i_data[EXT4_DIND_BLOCK];
4755
		if (nr) {
4756 4757
			ext4_free_branches(handle, inode, NULL, &nr, &nr+1, 2);
			i_data[EXT4_DIND_BLOCK] = 0;
4758
		}
4759 4760
	case EXT4_DIND_BLOCK:
		nr = i_data[EXT4_TIND_BLOCK];
4761
		if (nr) {
4762 4763
			ext4_free_branches(handle, inode, NULL, &nr, &nr+1, 3);
			i_data[EXT4_TIND_BLOCK] = 0;
4764
		}
4765
	case EXT4_TIND_BLOCK:
4766 4767 4768
		;
	}

4769
	up_write(&ei->i_data_sem);
K
Kalpak Shah 已提交
4770
	inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
4771
	ext4_mark_inode_dirty(handle, inode);
4772 4773 4774 4775 4776 4777

	/*
	 * In a multi-transaction truncate, we only make the final transaction
	 * synchronous
	 */
	if (IS_SYNC(inode))
4778
		ext4_handle_sync(handle);
4779 4780 4781 4782 4783
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
4784
	 * ext4_delete_inode(), and we allow that function to clean up the
4785 4786 4787
	 * orphan info for us.
	 */
	if (inode->i_nlink)
4788
		ext4_orphan_del(handle, inode);
4789

4790
	ext4_journal_stop(handle);
4791 4792 4793
}

/*
4794
 * ext4_get_inode_loc returns with an extra refcount against the inode's
4795 4796 4797 4798
 * 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.
 */
4799 4800
static int __ext4_get_inode_loc(struct inode *inode,
				struct ext4_iloc *iloc, int in_mem)
4801
{
4802 4803 4804 4805 4806 4807
	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
4808
	iloc->bh = NULL;
4809 4810
	if (!ext4_valid_inum(sb, inode->i_ino))
		return -EIO;
4811

4812 4813 4814
	iloc->block_group = (inode->i_ino - 1) / EXT4_INODES_PER_GROUP(sb);
	gdp = ext4_get_group_desc(sb, iloc->block_group, NULL);
	if (!gdp)
4815 4816
		return -EIO;

4817 4818 4819 4820 4821 4822 4823 4824 4825 4826
	/*
	 * 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);
4827
	if (!bh) {
4828 4829
		EXT4_ERROR_INODE_BLOCK(inode, block,
				       "unable to read itable block");
4830 4831 4832 4833
		return -EIO;
	}
	if (!buffer_uptodate(bh)) {
		lock_buffer(bh);
4834 4835 4836 4837 4838 4839 4840 4841 4842 4843

		/*
		 * 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);

4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856
		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;
4857
			int i, start;
4858

4859
			start = inode_offset & ~(inodes_per_block - 1);
4860

4861 4862
			/* Is the inode bitmap in cache? */
			bitmap_bh = sb_getblk(sb, ext4_inode_bitmap(sb, gdp));
4863 4864 4865 4866 4867 4868 4869 4870 4871 4872 4873 4874
			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;
			}
4875
			for (i = start; i < start + inodes_per_block; i++) {
4876 4877
				if (i == inode_offset)
					continue;
4878
				if (ext4_test_bit(i, bitmap_bh->b_data))
4879 4880 4881
					break;
			}
			brelse(bitmap_bh);
4882
			if (i == start + inodes_per_block) {
4883 4884 4885 4886 4887 4888 4889 4890 4891
				/* 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:
4892 4893 4894 4895 4896 4897 4898 4899 4900
		/*
		 * 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
4901
			/* s_inode_readahead_blks is always a power of 2 */
4902 4903 4904 4905 4906 4907 4908
			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))
4909
				num -= ext4_itable_unused_count(sb, gdp);
4910 4911 4912 4913 4914 4915 4916
			table += num / inodes_per_block;
			if (end > table)
				end = table;
			while (b <= end)
				sb_breadahead(sb, b++);
		}

4917 4918 4919 4920 4921 4922 4923 4924 4925 4926
		/*
		 * 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)) {
4927 4928
			EXT4_ERROR_INODE_BLOCK(inode, block,
					       "unable to read itable block");
4929 4930 4931 4932 4933 4934 4935 4936 4937
			brelse(bh);
			return -EIO;
		}
	}
has_buffer:
	iloc->bh = bh;
	return 0;
}

4938
int ext4_get_inode_loc(struct inode *inode, struct ext4_iloc *iloc)
4939 4940
{
	/* We have all inode data except xattrs in memory here. */
4941
	return __ext4_get_inode_loc(inode, iloc,
4942
		!ext4_test_inode_state(inode, EXT4_STATE_XATTR));
4943 4944
}

4945
void ext4_set_inode_flags(struct inode *inode)
4946
{
4947
	unsigned int flags = EXT4_I(inode)->i_flags;
4948 4949

	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
4950
	if (flags & EXT4_SYNC_FL)
4951
		inode->i_flags |= S_SYNC;
4952
	if (flags & EXT4_APPEND_FL)
4953
		inode->i_flags |= S_APPEND;
4954
	if (flags & EXT4_IMMUTABLE_FL)
4955
		inode->i_flags |= S_IMMUTABLE;
4956
	if (flags & EXT4_NOATIME_FL)
4957
		inode->i_flags |= S_NOATIME;
4958
	if (flags & EXT4_DIRSYNC_FL)
4959 4960 4961
		inode->i_flags |= S_DIRSYNC;
}

4962 4963 4964
/* Propagate flags from i_flags to EXT4_I(inode)->i_flags */
void ext4_get_inode_flags(struct ext4_inode_info *ei)
{
4965 4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984
	unsigned int vfs_fl;
	unsigned long old_fl, new_fl;

	do {
		vfs_fl = ei->vfs_inode.i_flags;
		old_fl = ei->i_flags;
		new_fl = old_fl & ~(EXT4_SYNC_FL|EXT4_APPEND_FL|
				EXT4_IMMUTABLE_FL|EXT4_NOATIME_FL|
				EXT4_DIRSYNC_FL);
		if (vfs_fl & S_SYNC)
			new_fl |= EXT4_SYNC_FL;
		if (vfs_fl & S_APPEND)
			new_fl |= EXT4_APPEND_FL;
		if (vfs_fl & S_IMMUTABLE)
			new_fl |= EXT4_IMMUTABLE_FL;
		if (vfs_fl & S_NOATIME)
			new_fl |= EXT4_NOATIME_FL;
		if (vfs_fl & S_DIRSYNC)
			new_fl |= EXT4_DIRSYNC_FL;
	} while (cmpxchg(&ei->i_flags, old_fl, new_fl) != old_fl);
4985
}
4986

4987
static blkcnt_t ext4_inode_blocks(struct ext4_inode *raw_inode,
4988
				  struct ext4_inode_info *ei)
4989 4990
{
	blkcnt_t i_blocks ;
A
Aneesh Kumar K.V 已提交
4991 4992
	struct inode *inode = &(ei->vfs_inode);
	struct super_block *sb = inode->i_sb;
4993 4994 4995 4996 4997 4998

	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);
4999
		if (ext4_test_inode_flag(inode, EXT4_INODE_HUGE_FILE)) {
A
Aneesh Kumar K.V 已提交
5000 5001 5002 5003 5004
			/* i_blocks represent file system block size */
			return i_blocks  << (inode->i_blkbits - 9);
		} else {
			return i_blocks;
		}
5005 5006 5007 5008
	} else {
		return le32_to_cpu(raw_inode->i_blocks_lo);
	}
}
5009

5010
struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
5011
{
5012 5013
	struct ext4_iloc iloc;
	struct ext4_inode *raw_inode;
5014 5015
	struct ext4_inode_info *ei;
	struct inode *inode;
5016
	journal_t *journal = EXT4_SB(sb)->s_journal;
5017
	long ret;
5018 5019
	int block;

5020 5021 5022 5023 5024 5025 5026
	inode = iget_locked(sb, ino);
	if (!inode)
		return ERR_PTR(-ENOMEM);
	if (!(inode->i_state & I_NEW))
		return inode;

	ei = EXT4_I(inode);
5027
	iloc.bh = 0;
5028

5029 5030
	ret = __ext4_get_inode_loc(inode, &iloc, 0);
	if (ret < 0)
5031
		goto bad_inode;
5032
	raw_inode = ext4_raw_inode(&iloc);
5033 5034 5035
	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);
5036
	if (!(test_opt(inode->i_sb, NO_UID32))) {
5037 5038 5039 5040 5041
		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);

5042
	ei->i_state_flags = 0;
5043 5044 5045 5046 5047 5048 5049 5050 5051
	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 ||
5052
		    !(EXT4_SB(inode->i_sb)->s_mount_state & EXT4_ORPHAN_FS)) {
5053
			/* this inode is deleted */
5054
			ret = -ESTALE;
5055 5056 5057 5058 5059 5060 5061 5062
			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);
5063
	inode->i_blocks = ext4_inode_blocks(raw_inode, ei);
5064
	ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl_lo);
5065
	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT))
B
Badari Pulavarty 已提交
5066 5067
		ei->i_file_acl |=
			((__u64)le16_to_cpu(raw_inode->i_file_acl_high)) << 32;
5068
	inode->i_size = ext4_isize(raw_inode);
5069
	ei->i_disksize = inode->i_size;
5070 5071 5072
#ifdef CONFIG_QUOTA
	ei->i_reserved_quota = 0;
#endif
5073 5074
	inode->i_generation = le32_to_cpu(raw_inode->i_generation);
	ei->i_block_group = iloc.block_group;
5075
	ei->i_last_alloc_group = ~0;
5076 5077 5078 5079
	/*
	 * 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!
	 */
5080
	for (block = 0; block < EXT4_N_BLOCKS; block++)
5081 5082 5083
		ei->i_data[block] = raw_inode->i_block[block];
	INIT_LIST_HEAD(&ei->i_orphan);

5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094
	/*
	 * 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;

5095
		read_lock(&journal->j_state_lock);
5096 5097 5098 5099 5100 5101 5102 5103
		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;
5104
		read_unlock(&journal->j_state_lock);
5105 5106 5107 5108
		ei->i_sync_tid = tid;
		ei->i_datasync_tid = tid;
	}

5109
	if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE) {
5110
		ei->i_extra_isize = le16_to_cpu(raw_inode->i_extra_isize);
5111
		if (EXT4_GOOD_OLD_INODE_SIZE + ei->i_extra_isize >
5112
		    EXT4_INODE_SIZE(inode->i_sb)) {
5113
			ret = -EIO;
5114
			goto bad_inode;
5115
		}
5116 5117
		if (ei->i_extra_isize == 0) {
			/* The extra space is currently unused. Use it. */
5118 5119
			ei->i_extra_isize = sizeof(struct ext4_inode) -
					    EXT4_GOOD_OLD_INODE_SIZE;
5120 5121
		} else {
			__le32 *magic = (void *)raw_inode +
5122
					EXT4_GOOD_OLD_INODE_SIZE +
5123
					ei->i_extra_isize;
5124
			if (*magic == cpu_to_le32(EXT4_XATTR_MAGIC))
5125
				ext4_set_inode_state(inode, EXT4_STATE_XATTR);
5126 5127 5128 5129
		}
	} else
		ei->i_extra_isize = 0;

K
Kalpak Shah 已提交
5130 5131 5132 5133 5134
	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);

5135 5136 5137 5138 5139 5140 5141
	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;
	}

5142
	ret = 0;
5143
	if (ei->i_file_acl &&
5144
	    !ext4_data_block_valid(EXT4_SB(sb), ei->i_file_acl, 1)) {
5145 5146
		EXT4_ERROR_INODE(inode, "bad extended attribute block %llu",
				 ei->i_file_acl);
5147 5148
		ret = -EIO;
		goto bad_inode;
5149
	} else if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
5150 5151 5152 5153 5154
		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);
5155
	} else if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
5156 5157
		   (S_ISLNK(inode->i_mode) &&
		    !ext4_inode_is_fast_symlink(inode))) {
5158
		/* Validate block references which are part of inode */
5159 5160
		ret = ext4_check_inode_blockref(inode);
	}
5161
	if (ret)
5162
		goto bad_inode;
5163

5164
	if (S_ISREG(inode->i_mode)) {
5165 5166 5167
		inode->i_op = &ext4_file_inode_operations;
		inode->i_fop = &ext4_file_operations;
		ext4_set_aops(inode);
5168
	} else if (S_ISDIR(inode->i_mode)) {
5169 5170
		inode->i_op = &ext4_dir_inode_operations;
		inode->i_fop = &ext4_dir_operations;
5171
	} else if (S_ISLNK(inode->i_mode)) {
5172
		if (ext4_inode_is_fast_symlink(inode)) {
5173
			inode->i_op = &ext4_fast_symlink_inode_operations;
5174 5175 5176
			nd_terminate_link(ei->i_data, inode->i_size,
				sizeof(ei->i_data) - 1);
		} else {
5177 5178
			inode->i_op = &ext4_symlink_inode_operations;
			ext4_set_aops(inode);
5179
		}
5180 5181
	} else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
	      S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
5182
		inode->i_op = &ext4_special_inode_operations;
5183 5184 5185 5186 5187 5188
		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])));
5189 5190
	} else {
		ret = -EIO;
5191
		EXT4_ERROR_INODE(inode, "bogus i_mode (%o)", inode->i_mode);
5192
		goto bad_inode;
5193
	}
5194
	brelse(iloc.bh);
5195
	ext4_set_inode_flags(inode);
5196 5197
	unlock_new_inode(inode);
	return inode;
5198 5199

bad_inode:
5200
	brelse(iloc.bh);
5201 5202
	iget_failed(inode);
	return ERR_PTR(ret);
5203 5204
}

5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217
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
5218
		raw_inode->i_blocks_lo   = cpu_to_le32(i_blocks);
5219
		raw_inode->i_blocks_high = 0;
5220
		ext4_clear_inode_flag(inode, EXT4_INODE_HUGE_FILE);
5221 5222 5223 5224 5225 5226
		return 0;
	}
	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE))
		return -EFBIG;

	if (i_blocks <= 0xffffffffffffULL) {
5227 5228 5229 5230
		/*
		 * i_blocks can be represented in a 48 bit variable
		 * as multiple of 512 bytes
		 */
A
Aneesh Kumar K.V 已提交
5231
		raw_inode->i_blocks_lo   = cpu_to_le32(i_blocks);
5232
		raw_inode->i_blocks_high = cpu_to_le16(i_blocks >> 32);
5233
		ext4_clear_inode_flag(inode, EXT4_INODE_HUGE_FILE);
5234
	} else {
5235
		ext4_set_inode_flag(inode, EXT4_INODE_HUGE_FILE);
A
Aneesh Kumar K.V 已提交
5236 5237 5238 5239
		/* 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);
5240
	}
5241
	return 0;
5242 5243
}

5244 5245 5246 5247 5248 5249 5250
/*
 * 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.
 */
5251
static int ext4_do_update_inode(handle_t *handle,
5252
				struct inode *inode,
5253
				struct ext4_iloc *iloc)
5254
{
5255 5256
	struct ext4_inode *raw_inode = ext4_raw_inode(iloc);
	struct ext4_inode_info *ei = EXT4_I(inode);
5257 5258 5259 5260 5261
	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. */
5262
	if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
5263
		memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
5264

5265
	ext4_get_inode_flags(ei);
5266
	raw_inode->i_mode = cpu_to_le16(inode->i_mode);
5267
	if (!(test_opt(inode->i_sb, NO_UID32))) {
5268 5269 5270 5271 5272 5273
		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
 */
5274
		if (!ei->i_dtime) {
5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291
			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
5292 5293 5294 5295 5296 5297

	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);

5298 5299
	if (ext4_inode_blocks_set(handle, raw_inode, ei))
		goto out_brelse;
5300
	raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
5301
	raw_inode->i_flags = cpu_to_le32(ei->i_flags);
5302 5303
	if (EXT4_SB(inode->i_sb)->s_es->s_creator_os !=
	    cpu_to_le32(EXT4_OS_HURD))
B
Badari Pulavarty 已提交
5304 5305
		raw_inode->i_file_acl_high =
			cpu_to_le16(ei->i_file_acl >> 32);
5306
	raw_inode->i_file_acl_lo = cpu_to_le32(ei->i_file_acl);
5307 5308 5309 5310 5311 5312 5313 5314 5315 5316 5317 5318 5319 5320 5321 5322
	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,
5323
					EXT4_FEATURE_RO_COMPAT_LARGE_FILE);
5324
			sb->s_dirt = 1;
5325
			ext4_handle_sync(handle);
5326
			err = ext4_handle_dirty_metadata(handle, NULL,
5327
					EXT4_SB(sb)->s_sbh);
5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341
		}
	}
	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;
		}
5342 5343 5344
	} else
		for (block = 0; block < EXT4_N_BLOCKS; block++)
			raw_inode->i_block[block] = ei->i_data[block];
5345

5346 5347 5348 5349 5350
	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);
5351
		raw_inode->i_extra_isize = cpu_to_le16(ei->i_extra_isize);
5352 5353
	}

5354
	BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
5355
	rc = ext4_handle_dirty_metadata(handle, NULL, bh);
5356 5357
	if (!err)
		err = rc;
5358
	ext4_clear_inode_state(inode, EXT4_STATE_NEW);
5359

5360
	ext4_update_inode_fsync_trans(handle, inode, 0);
5361
out_brelse:
5362
	brelse(bh);
5363
	ext4_std_error(inode->i_sb, err);
5364 5365 5366 5367
	return err;
}

/*
5368
 * ext4_write_inode()
5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384
 *
 * 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
5385
 * ext4_mark_inode_dirty().  This is a correctness thing for O_SYNC and for
5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401
 * 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.
 */
5402
int ext4_write_inode(struct inode *inode, struct writeback_control *wbc)
5403
{
5404 5405
	int err;

5406 5407 5408
	if (current->flags & PF_MEMALLOC)
		return 0;

5409 5410 5411 5412 5413 5414
	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;
		}
5415

5416
		if (wbc->sync_mode != WB_SYNC_ALL)
5417 5418 5419 5420 5421
			return 0;

		err = ext4_force_commit(inode->i_sb);
	} else {
		struct ext4_iloc iloc;
5422

5423
		err = __ext4_get_inode_loc(inode, &iloc, 0);
5424 5425
		if (err)
			return err;
5426
		if (wbc->sync_mode == WB_SYNC_ALL)
5427 5428
			sync_dirty_buffer(iloc.bh);
		if (buffer_req(iloc.bh) && !buffer_uptodate(iloc.bh)) {
5429 5430
			EXT4_ERROR_INODE_BLOCK(inode, iloc.bh->b_blocknr,
					 "IO error syncing inode");
5431 5432
			err = -EIO;
		}
5433
		brelse(iloc.bh);
5434 5435
	}
	return err;
5436 5437 5438
}

/*
5439
 * ext4_setattr()
5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452
 *
 * 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.)
 *
5453 5454 5455 5456 5457 5458 5459 5460
 * 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.
5461
 */
5462
int ext4_setattr(struct dentry *dentry, struct iattr *attr)
5463 5464 5465 5466 5467 5468 5469 5470 5471
{
	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;

5472
	if (is_quota_modification(inode, attr))
5473
		dquot_initialize(inode);
5474 5475 5476 5477 5478 5479
	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
5480
		handle = ext4_journal_start(inode, (EXT4_MAXQUOTAS_INIT_BLOCKS(inode->i_sb)+
5481
					EXT4_MAXQUOTAS_DEL_BLOCKS(inode->i_sb))+3);
5482 5483 5484 5485
		if (IS_ERR(handle)) {
			error = PTR_ERR(handle);
			goto err_out;
		}
5486
		error = dquot_transfer(inode, attr);
5487
		if (error) {
5488
			ext4_journal_stop(handle);
5489 5490 5491 5492 5493 5494 5495 5496
			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;
5497 5498
		error = ext4_mark_inode_dirty(handle, inode);
		ext4_journal_stop(handle);
5499 5500
	}

5501
	if (attr->ia_valid & ATTR_SIZE) {
5502
		if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) {
5503 5504
			struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);

5505 5506
			if (attr->ia_size > sbi->s_bitmap_maxbytes)
				return -EFBIG;
5507 5508 5509
		}
	}

5510
	if (S_ISREG(inode->i_mode) &&
5511 5512
	    attr->ia_valid & ATTR_SIZE &&
	    (attr->ia_size < inode->i_size ||
5513
	     (ext4_test_inode_flag(inode, EXT4_INODE_EOFBLOCKS)))) {
5514 5515
		handle_t *handle;

5516
		handle = ext4_journal_start(inode, 3);
5517 5518 5519 5520 5521
		if (IS_ERR(handle)) {
			error = PTR_ERR(handle);
			goto err_out;
		}

5522 5523 5524
		error = ext4_orphan_add(handle, inode);
		EXT4_I(inode)->i_disksize = attr->ia_size;
		rc = ext4_mark_inode_dirty(handle, inode);
5525 5526
		if (!error)
			error = rc;
5527
		ext4_journal_stop(handle);
5528 5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543

		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;
			}
		}
5544
		/* ext4_truncate will clear the flag */
5545
		if ((ext4_test_inode_flag(inode, EXT4_INODE_EOFBLOCKS)))
5546
			ext4_truncate(inode);
5547 5548
	}

C
Christoph Hellwig 已提交
5549 5550 5551
	if ((attr->ia_valid & ATTR_SIZE) &&
	    attr->ia_size != i_size_read(inode))
		rc = vmtruncate(inode, attr->ia_size);
5552

C
Christoph Hellwig 已提交
5553 5554 5555 5556 5557 5558 5559 5560 5561
	if (!rc) {
		setattr_copy(inode, attr);
		mark_inode_dirty(inode);
	}

	/*
	 * If the call to ext4_truncate failed to get a transaction handle at
	 * all, we need to clean up the in-core orphan list manually.
	 */
5562
	if (inode->i_nlink)
5563
		ext4_orphan_del(NULL, inode);
5564 5565

	if (!rc && (ia_valid & ATTR_MODE))
5566
		rc = ext4_acl_chmod(inode);
5567 5568

err_out:
5569
	ext4_std_error(inode->i_sb, error);
5570 5571 5572 5573 5574
	if (!error)
		error = rc;
	return error;
}

5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600
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;
}
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 5626 5627 5628
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)
{
5629
	if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
5630 5631
		return ext4_indirect_trans_blocks(inode, nrblocks, chunk);
	return ext4_ext_index_trans_blocks(inode, nrblocks, chunk);
5632
}
5633

5634
/*
5635 5636 5637
 * 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
5638
 *
5639
 * If datablocks are discontiguous, they are possible to spread over
5640
 * different block groups too. If they are contiuguous, with flexbg,
5641
 * they could still across block group boundary.
5642
 *
5643 5644 5645 5646
 * Also account for superblock, inode, quota and xattr blocks
 */
int ext4_meta_trans_blocks(struct inode *inode, int nrblocks, int chunk)
{
5647 5648
	ext4_group_t groups, ngroups = ext4_get_groups_count(inode->i_sb);
	int gdpblocks;
5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663 5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674
	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;
5675 5676
	if (groups > ngroups)
		groups = ngroups;
5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690
	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
5691 5692
 * the modification of a single pages into a single transaction,
 * which may include multiple chunks of block allocations.
5693
 *
5694
 * This could be called via ext4_write_begin()
5695
 *
5696
 * We need to consider the worse case, when
5697
 * one new block per extent.
5698
 */
A
Alex Tomas 已提交
5699
int ext4_writepage_trans_blocks(struct inode *inode)
5700
{
5701
	int bpp = ext4_journal_blocks_per_page(inode);
5702 5703
	int ret;

5704
	ret = ext4_meta_trans_blocks(inode, bpp, 0);
A
Alex Tomas 已提交
5705

5706
	/* Account for data blocks for journalled mode */
5707
	if (ext4_should_journal_data(inode))
5708
		ret += bpp;
5709 5710
	return ret;
}
5711 5712 5713 5714 5715

/*
 * Calculate the journal credits for a chunk of data modification.
 *
 * This is called from DIO, fallocate or whoever calling
5716
 * ext4_map_blocks() to map/allocate a chunk of contiguous disk blocks.
5717 5718 5719 5720 5721 5722 5723 5724 5725
 *
 * 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);
}

5726
/*
5727
 * The caller must have previously called ext4_reserve_inode_write().
5728 5729
 * Give this, we know that the caller already has write access to iloc->bh.
 */
5730
int ext4_mark_iloc_dirty(handle_t *handle,
5731
			 struct inode *inode, struct ext4_iloc *iloc)
5732 5733 5734
{
	int err = 0;

5735 5736 5737
	if (test_opt(inode->i_sb, I_VERSION))
		inode_inc_iversion(inode);

5738 5739 5740
	/* the do_update_inode consumes one bh->b_count */
	get_bh(iloc->bh);

5741
	/* ext4_do_update_inode() does jbd2_journal_dirty_metadata */
5742
	err = ext4_do_update_inode(handle, inode, iloc);
5743 5744 5745 5746 5747 5748 5749 5750 5751 5752
	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
5753 5754
ext4_reserve_inode_write(handle_t *handle, struct inode *inode,
			 struct ext4_iloc *iloc)
5755
{
5756 5757 5758 5759 5760 5761 5762 5763 5764
	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;
5765 5766
		}
	}
5767
	ext4_std_error(inode->i_sb, err);
5768 5769 5770
	return err;
}

5771 5772 5773 5774
/*
 * 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
5775 5776 5777 5778
static int ext4_expand_extra_isize(struct inode *inode,
				   unsigned int new_extra_isize,
				   struct ext4_iloc iloc,
				   handle_t *handle)
5779 5780 5781 5782 5783 5784 5785 5786 5787 5788 5789 5790
{
	struct ext4_inode *raw_inode;
	struct ext4_xattr_ibody_header *header;

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

	raw_inode = ext4_raw_inode(&iloc);

	header = IHDR(inode, raw_inode);

	/* No extended attributes present */
5791 5792
	if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR) ||
	    header->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC)) {
5793 5794 5795 5796 5797 5798 5799 5800 5801 5802 5803
		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);
}

5804 5805 5806 5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817 5818 5819 5820 5821 5822 5823 5824
/*
 * 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.
 */
5825
int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode)
5826
{
5827
	struct ext4_iloc iloc;
5828 5829 5830
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
	static unsigned int mnt_count;
	int err, ret;
5831 5832

	might_sleep();
5833
	err = ext4_reserve_inode_write(handle, inode, &iloc);
5834 5835
	if (ext4_handle_valid(handle) &&
	    EXT4_I(inode)->i_extra_isize < sbi->s_want_extra_isize &&
5836
	    !ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND)) {
5837 5838 5839 5840 5841 5842 5843 5844 5845 5846 5847 5848 5849
		/*
		 * 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) {
5850 5851
				ext4_set_inode_state(inode,
						     EXT4_STATE_NO_EXPAND);
A
Aneesh Kumar K.V 已提交
5852 5853
				if (mnt_count !=
					le16_to_cpu(sbi->s_es->s_mnt_count)) {
5854
					ext4_warning(inode->i_sb,
5855 5856 5857
					"Unable to expand inode %lu. Delete"
					" some EAs or run e2fsck.",
					inode->i_ino);
A
Aneesh Kumar K.V 已提交
5858 5859
					mnt_count =
					  le16_to_cpu(sbi->s_es->s_mnt_count);
5860 5861 5862 5863
				}
			}
		}
	}
5864
	if (!err)
5865
		err = ext4_mark_iloc_dirty(handle, inode, &iloc);
5866 5867 5868 5869
	return err;
}

/*
5870
 * ext4_dirty_inode() is called from __mark_inode_dirty()
5871 5872 5873 5874 5875
 *
 * 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.
 *
5876
 * Also, dquot_alloc_block() will always dirty the inode when blocks
5877 5878 5879 5880 5881 5882
 * 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.
 */
5883
void ext4_dirty_inode(struct inode *inode)
5884 5885 5886
{
	handle_t *handle;

5887
	handle = ext4_journal_start(inode, 2);
5888 5889
	if (IS_ERR(handle))
		goto out;
5890 5891 5892

	ext4_mark_inode_dirty(handle, inode);

5893
	ext4_journal_stop(handle);
5894 5895 5896 5897 5898 5899 5900 5901
out:
	return;
}

#if 0
/*
 * Bind an inode's backing buffer_head into this transaction, to prevent
 * it from being flushed to disk early.  Unlike
5902
 * ext4_reserve_inode_write, this leaves behind no bh reference and
5903 5904 5905
 * returns no iloc structure, so the caller needs to repeat the iloc
 * lookup to mark the inode dirty later.
 */
5906
static int ext4_pin_inode(handle_t *handle, struct inode *inode)
5907
{
5908
	struct ext4_iloc iloc;
5909 5910 5911

	int err = 0;
	if (handle) {
5912
		err = ext4_get_inode_loc(inode, &iloc);
5913 5914
		if (!err) {
			BUFFER_TRACE(iloc.bh, "get_write_access");
5915
			err = jbd2_journal_get_write_access(handle, iloc.bh);
5916
			if (!err)
5917
				err = ext4_handle_dirty_metadata(handle,
5918
								 NULL,
5919
								 iloc.bh);
5920 5921 5922
			brelse(iloc.bh);
		}
	}
5923
	ext4_std_error(inode->i_sb, err);
5924 5925 5926 5927
	return err;
}
#endif

5928
int ext4_change_inode_journal_flag(struct inode *inode, int val)
5929 5930 5931 5932 5933 5934 5935 5936 5937 5938 5939 5940 5941 5942 5943
{
	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.
	 */

5944
	journal = EXT4_JOURNAL(inode);
5945 5946
	if (!journal)
		return 0;
5947
	if (is_journal_aborted(journal))
5948 5949
		return -EROFS;

5950 5951
	jbd2_journal_lock_updates(journal);
	jbd2_journal_flush(journal);
5952 5953 5954 5955 5956 5957 5958 5959 5960 5961

	/*
	 * 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)
5962
		ext4_set_inode_flag(inode, EXT4_INODE_JOURNAL_DATA);
5963
	else
5964
		ext4_clear_inode_flag(inode, EXT4_INODE_JOURNAL_DATA);
5965
	ext4_set_aops(inode);
5966

5967
	jbd2_journal_unlock_updates(journal);
5968 5969 5970

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

5971
	handle = ext4_journal_start(inode, 1);
5972 5973 5974
	if (IS_ERR(handle))
		return PTR_ERR(handle);

5975
	err = ext4_mark_inode_dirty(handle, inode);
5976
	ext4_handle_sync(handle);
5977 5978
	ext4_journal_stop(handle);
	ext4_std_error(inode->i_sb, err);
5979 5980 5981

	return err;
}
5982 5983 5984 5985 5986 5987

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

5988
int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
5989
{
5990
	struct page *page = vmf->page;
5991 5992 5993
	loff_t size;
	unsigned long len;
	int ret = -EINVAL;
5994
	void *fsdata;
5995 5996 5997 5998 5999 6000 6001 6002 6003 6004 6005 6006 6007 6008 6009 6010 6011 6012 6013 6014 6015 6016 6017 6018
	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;

6019 6020 6021 6022 6023 6024 6025
	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
	 */
6026 6027
	if (page_has_buffers(page)) {
		if (!walk_page_buffers(NULL, page_buffers(page), 0, len, NULL,
6028 6029
					ext4_bh_unmapped)) {
			unlock_page(page);
6030
			goto out_unlock;
6031
		}
6032
	}
6033
	unlock_page(page);
6034 6035 6036 6037 6038 6039 6040 6041
	/*
	 * 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),
6042
			len, AOP_FLAG_UNINTERRUPTIBLE, &page, &fsdata);
6043 6044 6045
	if (ret < 0)
		goto out_unlock;
	ret = mapping->a_ops->write_end(file, mapping, page_offset(page),
6046
			len, len, page, fsdata);
6047 6048 6049 6050
	if (ret < 0)
		goto out_unlock;
	ret = 0;
out_unlock:
6051 6052
	if (ret)
		ret = VM_FAULT_SIGBUS;
6053 6054 6055
	up_read(&inode->i_alloc_sem);
	return ret;
}