inode.c 128 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
 *
 * 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
 *
 *  64-bit file support on 64-bit platforms by Jakub Jelinek
 *	(jj@sunsite.ms.mff.cuni.cz)
 *
18
 *  Assorted race fixes, rewrite of ext4_get_block() by Al Viro, 2000
19
20
21
22
23
 */

#include <linux/module.h>
#include <linux/fs.h>
#include <linux/time.h>
24
#include <linux/jbd2.h>
25
26
27
28
29
30
#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>
31
#include <linux/pagevec.h>
32
#include <linux/mpage.h>
33
#include <linux/namei.h>
34
35
#include <linux/uio.h>
#include <linux/bio.h>
36
#include <linux/workqueue.h>
37
#include <linux/kernel.h>
38
#include <linux/printk.h>
39
#include <linux/slab.h>
40
#include <linux/ratelimit.h>
41

42
#include "ext4_jbd2.h"
43
44
#include "xattr.h"
#include "acl.h"
45
#include "ext4_extents.h"
46
#include "truncate.h"
47

48
49
#include <trace/events/ext4.h>

50
51
#define MPAGE_DA_EXTENT_TAIL 0x01

52
53
54
static inline int ext4_begin_ordered_truncate(struct inode *inode,
					      loff_t new_size)
{
55
	trace_ext4_begin_ordered_truncate(inode, new_size);
56
57
58
59
60
61
62
63
64
65
66
	/*
	 * If jinode is zero, then we never opened the file for
	 * writing, so there's no need to call
	 * jbd2_journal_begin_ordered_truncate() since there's no
	 * outstanding writes we need to flush.
	 */
	if (!EXT4_I(inode)->jinode)
		return 0;
	return jbd2_journal_begin_ordered_truncate(EXT4_JOURNAL(inode),
						   EXT4_I(inode)->jinode,
						   new_size);
67
68
}

69
static void ext4_invalidatepage(struct page *page, unsigned long offset);
70
71
72
73
74
75
static int noalloc_get_block_write(struct inode *inode, sector_t iblock,
				   struct buffer_head *bh_result, int create);
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);
static int __ext4_journalled_writepage(struct page *page, unsigned int len);
static int ext4_bh_delay_or_unwritten(handle_t *handle, struct buffer_head *bh);
76

77
78
79
/*
 * Test whether an inode is a fast symlink.
 */
80
static int ext4_inode_is_fast_symlink(struct inode *inode)
81
{
82
	int ea_blocks = EXT4_I(inode)->i_file_acl ?
83
84
85
86
87
88
89
90
91
92
		(inode->i_sb->s_blocksize >> 9) : 0;

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

/*
 * Restart the transaction associated with *handle.  This does a commit,
 * so before we call here everything must be consistently dirtied against
 * this transaction.
 */
93
int ext4_truncate_restart_trans(handle_t *handle, struct inode *inode,
94
				 int nblocks)
95
{
96
97
98
	int ret;

	/*
99
	 * Drop i_data_sem to avoid deadlock with ext4_map_blocks.  At this
100
101
102
103
	 * 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.
	 */
104
	BUG_ON(EXT4_JOURNAL(inode) == NULL);
105
	jbd_debug(2, "restarting handle %p\n", handle);
106
	up_write(&EXT4_I(inode)->i_data_sem);
107
	ret = ext4_journal_restart(handle, nblocks);
108
	down_write(&EXT4_I(inode)->i_data_sem);
109
	ext4_discard_preallocations(inode);
110
111

	return ret;
112
113
114
115
116
}

/*
 * Called at the last iput() if i_nlink is zero.
 */
Al Viro's avatar
Al Viro committed
117
void ext4_evict_inode(struct inode *inode)
118
119
{
	handle_t *handle;
120
	int err;
121

122
	trace_ext4_evict_inode(inode);
Al Viro's avatar
Al Viro committed
123
	if (inode->i_nlink) {
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
		/*
		 * When journalling data dirty buffers are tracked only in the
		 * journal. So although mm thinks everything is clean and
		 * ready for reaping the inode might still have some pages to
		 * write in the running transaction or waiting to be
		 * checkpointed. Thus calling jbd2_journal_invalidatepage()
		 * (via truncate_inode_pages()) to discard these buffers can
		 * cause data loss. Also even if we did not discard these
		 * buffers, we would have no way to find them after the inode
		 * is reaped and thus user could see stale data if he tries to
		 * read them before the transaction is checkpointed. So be
		 * careful and force everything to disk here... We use
		 * ei->i_datasync_tid to store the newest transaction
		 * containing inode's data.
		 *
		 * Note that directories do not have this problem because they
		 * don't use page cache.
		 */
		if (ext4_should_journal_data(inode) &&
		    (S_ISLNK(inode->i_mode) || S_ISREG(inode->i_mode))) {
			journal_t *journal = EXT4_SB(inode->i_sb)->s_journal;
			tid_t commit_tid = EXT4_I(inode)->i_datasync_tid;

			jbd2_log_start_commit(journal, commit_tid);
			jbd2_log_wait_commit(journal, commit_tid);
			filemap_write_and_wait(&inode->i_data);
		}
Al Viro's avatar
Al Viro committed
151
152
153
154
		truncate_inode_pages(&inode->i_data, 0);
		goto no_delete;
	}

155
	if (!is_bad_inode(inode))
156
		dquot_initialize(inode);
157

158
159
	if (ext4_should_order_data(inode))
		ext4_begin_ordered_truncate(inode, 0);
160
161
162
163
164
	truncate_inode_pages(&inode->i_data, 0);

	if (is_bad_inode(inode))
		goto no_delete;

165
	handle = ext4_journal_start(inode, ext4_blocks_for_truncate(inode)+3);
166
	if (IS_ERR(handle)) {
167
		ext4_std_error(inode->i_sb, PTR_ERR(handle));
168
169
170
171
172
		/*
		 * 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.
		 */
173
		ext4_orphan_del(NULL, inode);
174
175
176
177
		goto no_delete;
	}

	if (IS_SYNC(inode))
178
		ext4_handle_sync(handle);
179
	inode->i_size = 0;
180
181
	err = ext4_mark_inode_dirty(handle, inode);
	if (err) {
182
		ext4_warning(inode->i_sb,
183
184
185
			     "couldn't mark inode dirty (err %d)", err);
		goto stop_handle;
	}
186
	if (inode->i_blocks)
187
		ext4_truncate(inode);
188
189
190
191
192
193
194

	/*
	 * 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.
	 */
195
	if (!ext4_handle_has_enough_credits(handle, 3)) {
196
197
198
199
		err = ext4_journal_extend(handle, 3);
		if (err > 0)
			err = ext4_journal_restart(handle, 3);
		if (err != 0) {
200
			ext4_warning(inode->i_sb,
201
202
203
				     "couldn't extend journal (err %d)", err);
		stop_handle:
			ext4_journal_stop(handle);
204
			ext4_orphan_del(NULL, inode);
205
206
207
208
			goto no_delete;
		}
	}

209
	/*
210
	 * Kill off the orphan record which ext4_truncate created.
211
	 * AKPM: I think this can be inside the above `if'.
212
	 * Note that ext4_orphan_del() has to be able to cope with the
213
	 * deletion of a non-existent orphan - this is because we don't
214
	 * know if ext4_truncate() actually created an orphan record.
215
216
	 * (Well, we could do this if we need to, but heck - it works)
	 */
217
218
	ext4_orphan_del(handle, inode);
	EXT4_I(inode)->i_dtime	= get_seconds();
219
220
221
222
223
224
225
226

	/*
	 * 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.
	 */
227
	if (ext4_mark_inode_dirty(handle, inode))
228
		/* If that failed, just do the required in-core inode clear. */
Al Viro's avatar
Al Viro committed
229
		ext4_clear_inode(inode);
230
	else
231
232
		ext4_free_inode(handle, inode);
	ext4_journal_stop(handle);
233
234
	return;
no_delete:
Al Viro's avatar
Al Viro committed
235
	ext4_clear_inode(inode);	/* We must guarantee clearing of inode... */
236
237
}

238
239
#ifdef CONFIG_QUOTA
qsize_t *ext4_get_reserved_space(struct inode *inode)
240
{
241
	return &EXT4_I(inode)->i_reserved_quota;
242
}
243
#endif
244

245
246
/*
 * Calculate the number of metadata blocks need to reserve
247
 * to allocate a block located at @lblock
248
 */
249
static int ext4_calc_metadata_amount(struct inode *inode, ext4_lblk_t lblock)
250
{
251
	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
252
		return ext4_ext_calc_metadata_amount(inode, lblock);
253

254
	return ext4_ind_calc_metadata_amount(inode, lblock);
255
256
}

257
258
259
260
/*
 * Called with i_data_sem down, which is important since we can call
 * ext4_discard_preallocations() from here.
 */
261
262
void ext4_da_update_reserve_space(struct inode *inode,
					int used, int quota_claim)
263
264
{
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
265
266
267
	struct ext4_inode_info *ei = EXT4_I(inode);

	spin_lock(&ei->i_block_reservation_lock);
268
	trace_ext4_da_update_reserve_space(inode, used);
269
270
271
272
273
274
275
276
	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;
	}
277

278
279
280
	/* Update per-inode reservations */
	ei->i_reserved_data_blocks -= used;
	ei->i_reserved_meta_blocks -= ei->i_allocated_meta_blocks;
281
282
	percpu_counter_sub(&sbi->s_dirtyblocks_counter,
			   used + ei->i_allocated_meta_blocks);
283
	ei->i_allocated_meta_blocks = 0;
284

285
286
287
288
289
290
	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.
		 */
291
292
		percpu_counter_sub(&sbi->s_dirtyblocks_counter,
				   ei->i_reserved_meta_blocks);
293
		ei->i_reserved_meta_blocks = 0;
294
		ei->i_da_metadata_calc_len = 0;
295
	}
296
	spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
297

298
299
	/* Update quota subsystem for data blocks */
	if (quota_claim)
300
		dquot_claim_block(inode, used);
301
	else {
302
303
304
		/*
		 * We did fallocate with an offset that is already delayed
		 * allocated. So on delayed allocated writeback we should
305
		 * not re-claim the quota for fallocated blocks.
306
		 */
307
		dquot_release_reservation_block(inode, used);
308
	}
309
310
311
312
313
314

	/*
	 * 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.
	 */
315
316
	if ((ei->i_reserved_data_blocks == 0) &&
	    (atomic_read(&inode->i_writecount) == 0))
317
		ext4_discard_preallocations(inode);
318
319
}

320
static int __check_block_validity(struct inode *inode, const char *func,
321
322
				unsigned int line,
				struct ext4_map_blocks *map)
323
{
324
325
	if (!ext4_data_block_valid(EXT4_SB(inode->i_sb), map->m_pblk,
				   map->m_len)) {
326
327
328
329
		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);
330
331
332
333
334
		return -EIO;
	}
	return 0;
}

335
#define check_block_validity(inode, map)	\
336
	__check_block_validity((inode), __func__, __LINE__, (map))
337

338
/*
339
340
 * Return the number of contiguous dirty pages in a given inode
 * starting at page frame idx.
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
 */
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;
			}
374
375
376
377
378
379
380
381
382
			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));
			}
383
384
385
386
387
			unlock_page(page);
			if (done)
				break;
			idx++;
			num++;
388
389
			if (num >= max_pages) {
				done = 1;
390
				break;
391
			}
392
393
394
395
396
397
		}
		pagevec_release(&pvec);
	}
	return num;
}

398
/*
399
 * The ext4_map_blocks() function tries to look up the requested blocks,
400
 * and returns if the blocks are already mapped.
401
402
403
404
405
 *
 * 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.
 *
406
407
 * If file type is extents based, it will call ext4_ext_map_blocks(),
 * Otherwise, call with ext4_ind_map_blocks() to handle indirect mapping
408
409
410
411
412
413
414
415
416
417
418
419
 * 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.
 */
420
421
int ext4_map_blocks(handle_t *handle, struct inode *inode,
		    struct ext4_map_blocks *map, int flags)
422
423
{
	int retval;
424

425
426
427
428
	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);
429
	/*
430
431
	 * Try to see if we can get the block without requesting a new
	 * file system block.
432
433
	 */
	down_read((&EXT4_I(inode)->i_data_sem));
434
	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
435
		retval = ext4_ext_map_blocks(handle, inode, map, 0);
436
	} else {
437
		retval = ext4_ind_map_blocks(handle, inode, map, 0);
438
	}
439
	up_read((&EXT4_I(inode)->i_data_sem));
440

441
	if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED) {
442
		int ret = check_block_validity(inode, map);
443
444
445
446
		if (ret != 0)
			return ret;
	}

447
	/* If it is only a block(s) look up */
448
	if ((flags & EXT4_GET_BLOCKS_CREATE) == 0)
449
450
451
452
453
454
455
456
457
		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.
	 */
458
	if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED)
459
460
		return retval;

461
462
463
464
465
466
467
468
469
470
	/*
	 * 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.
	 */
471
	map->m_flags &= ~EXT4_MAP_UNWRITTEN;
472

473
	/*
474
475
476
477
	 * 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.
478
479
	 */
	down_write((&EXT4_I(inode)->i_data_sem));
480
481
482
483
484
485
486

	/*
	 * 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
	 */
487
	if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
488
		ext4_set_inode_state(inode, EXT4_STATE_DELALLOC_RESERVED);
489
490
491
492
	/*
	 * We need to check for EXT4 here because migrate
	 * could have changed the inode type in between
	 */
493
	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
494
		retval = ext4_ext_map_blocks(handle, inode, map, flags);
495
	} else {
496
		retval = ext4_ind_map_blocks(handle, inode, map, flags);
497

498
		if (retval > 0 && map->m_flags & EXT4_MAP_NEW) {
499
500
501
502
503
			/*
			 * We allocated new blocks which will result in
			 * i_data's format changing.  Force the migrate
			 * to fail by clearing migrate flags
			 */
504
			ext4_clear_inode_state(inode, EXT4_STATE_EXT_MIGRATE);
505
		}
506

507
508
509
510
511
512
513
		/*
		 * 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) &&
514
			(flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE))
515
516
			ext4_da_update_reserve_space(inode, retval, 1);
	}
517
	if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
518
		ext4_clear_inode_state(inode, EXT4_STATE_DELALLOC_RESERVED);
519

520
	up_write((&EXT4_I(inode)->i_data_sem));
521
	if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED) {
522
		int ret = check_block_validity(inode, map);
523
524
525
		if (ret != 0)
			return ret;
	}
526
527
528
	return retval;
}

529
530
531
/* Maximum number of blocks we map for direct IO at once. */
#define DIO_MAX_BLOCKS 4096

532
533
static int _ext4_get_block(struct inode *inode, sector_t iblock,
			   struct buffer_head *bh, int flags)
534
{
535
	handle_t *handle = ext4_journal_current_handle();
536
	struct ext4_map_blocks map;
Jan Kara's avatar
Jan Kara committed
537
	int ret = 0, started = 0;
538
	int dio_credits;
539

540
541
542
543
	map.m_lblk = iblock;
	map.m_len = bh->b_size >> inode->i_blkbits;

	if (flags && !handle) {
Jan Kara's avatar
Jan Kara committed
544
		/* Direct IO write... */
545
546
547
		if (map.m_len > DIO_MAX_BLOCKS)
			map.m_len = DIO_MAX_BLOCKS;
		dio_credits = ext4_chunk_trans_blocks(inode, map.m_len);
548
		handle = ext4_journal_start(inode, dio_credits);
Jan Kara's avatar
Jan Kara committed
549
		if (IS_ERR(handle)) {
550
			ret = PTR_ERR(handle);
551
			return ret;
552
		}
Jan Kara's avatar
Jan Kara committed
553
		started = 1;
554
555
	}

556
	ret = ext4_map_blocks(handle, inode, &map, flags);
Jan Kara's avatar
Jan Kara committed
557
	if (ret > 0) {
558
559
560
		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;
Jan Kara's avatar
Jan Kara committed
561
		ret = 0;
562
	}
Jan Kara's avatar
Jan Kara committed
563
564
	if (started)
		ext4_journal_stop(handle);
565
566
567
	return ret;
}

568
569
570
571
572
573
574
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);
}

575
576
577
/*
 * `handle' can be NULL if create is zero
 */
578
struct buffer_head *ext4_getblk(handle_t *handle, struct inode *inode,
Aneesh Kumar K.V's avatar
Aneesh Kumar K.V committed
579
				ext4_lblk_t block, int create, int *errp)
580
{
581
582
	struct ext4_map_blocks map;
	struct buffer_head *bh;
583
584
585
586
	int fatal = 0, err;

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

587
588
589
590
	map.m_lblk = block;
	map.m_len = 1;
	err = ext4_map_blocks(handle, inode, &map,
			      create ? EXT4_GET_BLOCKS_CREATE : 0);
591

592
593
594
595
596
597
598
599
600
601
	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;
602
	}
603
604
605
	if (map.m_flags & EXT4_MAP_NEW) {
		J_ASSERT(create != 0);
		J_ASSERT(handle != NULL);
606

607
608
609
610
611
612
613
614
615
616
617
618
619
		/*
		 * 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);
620
		}
621
622
623
624
625
626
627
		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");
628
	}
629
630
631
632
633
634
	if (fatal) {
		*errp = fatal;
		brelse(bh);
		bh = NULL;
	}
	return bh;
635
636
}

637
struct buffer_head *ext4_bread(handle_t *handle, struct inode *inode,
Aneesh Kumar K.V's avatar
Aneesh Kumar K.V committed
638
			       ext4_lblk_t block, int create, int *err)
639
{
640
	struct buffer_head *bh;
641

642
	bh = ext4_getblk(handle, inode, block, create, err);
643
644
645
646
647
648
649
650
651
652
653
654
655
	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;
}

656
657
658
659
660
661
662
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))
663
664
665
666
667
668
669
{
	struct buffer_head *bh;
	unsigned block_start, block_end;
	unsigned blocksize = head->b_size;
	int err, ret = 0;
	struct buffer_head *next;

670
671
	for (bh = head, block_start = 0;
	     ret == 0 && (bh != head || !block_start);
672
	     block_start = block_end, bh = next) {
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
		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
690
 * close off a transaction and start a new one between the ext4_get_block()
691
 * and the commit_write().  So doing the jbd2_journal_start at the start of
692
693
 * prepare_write() is the right place.
 *
694
695
 * Also, this function can nest inside ext4_writepage() ->
 * block_write_full_page(). In that case, we *know* that ext4_writepage()
696
697
698
699
 * 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.
 *
700
 * By accident, ext4 can be reentered when a transaction is open via
701
702
703
704
705
706
 * 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.
 *
707
 * So what we do is to rely on the fact that jbd2_journal_stop/journal_start
708
709
710
711
712
 * 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,
713
				       struct buffer_head *bh)
714
{
715
716
717
	int dirty = buffer_dirty(bh);
	int ret;

718
719
	if (!buffer_mapped(bh) || buffer_freed(bh))
		return 0;
720
	/*
721
	 * __block_write_begin() could have dirtied some buffers. Clean
722
723
	 * the dirty bit as jbd2_journal_get_write_access() could complain
	 * otherwise about fs integrity issues. Setting of the dirty bit
724
	 * by __block_write_begin() isn't a real problem here as we clear
725
726
727
728
729
730
731
732
733
	 * 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;
734
735
}

736
737
static int ext4_get_block_write(struct inode *inode, sector_t iblock,
		   struct buffer_head *bh_result, int create);
Nick Piggin's avatar
Nick Piggin committed
738
static int ext4_write_begin(struct file *file, struct address_space *mapping,
739
740
			    loff_t pos, unsigned len, unsigned flags,
			    struct page **pagep, void **fsdata)
741
{
742
	struct inode *inode = mapping->host;
743
	int ret, needed_blocks;
744
745
	handle_t *handle;
	int retries = 0;
746
	struct page *page;
747
	pgoff_t index;
748
	unsigned from, to;
Nick Piggin's avatar
Nick Piggin committed
749

750
	trace_ext4_write_begin(inode, pos, len, flags);
751
752
753
754
755
	/*
	 * 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;
756
	index = pos >> PAGE_CACHE_SHIFT;
757
758
	from = pos & (PAGE_CACHE_SIZE - 1);
	to = from + len;
759
760

retry:
761
762
763
764
	handle = ext4_journal_start(inode, needed_blocks);
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		goto out;
765
	}
766

767
768
769
770
	/* We cannot recurse into the filesystem as the transaction is already
	 * started */
	flags |= AOP_FLAG_NOFS;

771
	page = grab_cache_page_write_begin(mapping, index, flags);
772
773
774
775
776
777
778
	if (!page) {
		ext4_journal_stop(handle);
		ret = -ENOMEM;
		goto out;
	}
	*pagep = page;

779
	if (ext4_should_dioread_nolock(inode))
780
		ret = __block_write_begin(page, pos, len, ext4_get_block_write);
781
	else
782
		ret = __block_write_begin(page, pos, len, ext4_get_block);
Nick Piggin's avatar
Nick Piggin committed
783
784

	if (!ret && ext4_should_journal_data(inode)) {
785
786
787
		ret = walk_page_buffers(handle, page_buffers(page),
				from, to, NULL, do_journal_get_write_access);
	}
Nick Piggin's avatar
Nick Piggin committed
788
789

	if (ret) {
790
791
		unlock_page(page);
		page_cache_release(page);
792
		/*
793
		 * __block_write_begin may have instantiated a few blocks
794
795
		 * outside i_size.  Trim these off again. Don't need
		 * i_size_read because we hold i_mutex.
796
797
798
		 *
		 * Add inode to orphan list in case we crash before
		 * truncate finishes
799
		 */
800
		if (pos + len > inode->i_size && ext4_can_truncate(inode))
801
802
803
804
			ext4_orphan_add(handle, inode);

		ext4_journal_stop(handle);
		if (pos + len > inode->i_size) {
805
			ext4_truncate_failed_write(inode);
806
			/*
807
			 * If truncate failed early the inode might
808
809
810
811
812
813
814
			 * 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);
		}
Nick Piggin's avatar
Nick Piggin committed
815
816
	}

817
	if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
818
		goto retry;
819
out:
820
821
822
	return ret;
}

Nick Piggin's avatar
Nick Piggin committed
823
824
/* For write_end() in data=journal mode */
static int write_end_fn(handle_t *handle, struct buffer_head *bh)
825
826
827
828
{
	if (!buffer_mapped(bh) || buffer_freed(bh))
		return 0;
	set_buffer_uptodate(bh);
829
	return ext4_handle_dirty_metadata(handle, NULL, bh);
830
831
}

832
static int ext4_generic_write_end(struct file *file,
833
834
835
				  struct address_space *mapping,
				  loff_t pos, unsigned len, unsigned copied,
				  struct page *page, void *fsdata)
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
{
	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;
}

878
879
880
881
/*
 * 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().
 *
882
 * ext4 never places buffers on inode->i_mapping->private_list.  metadata
883
884
 * buffers are managed internally.
 */
Nick Piggin's avatar
Nick Piggin committed
885
static int ext4_ordered_write_end(struct file *file,
886
887
888
				  struct address_space *mapping,
				  loff_t pos, unsigned len, unsigned copied,
				  struct page *page, void *fsdata)
889
{
890
	handle_t *handle = ext4_journal_current_handle();
891
	struct inode *inode = mapping->host;
892
893
	int ret = 0, ret2;

894
	trace_ext4_ordered_write_end(inode, pos, len, copied);
895
	ret = ext4_jbd2_file_inode(handle, inode);
896
897

	if (ret == 0) {
898
		ret2 = ext4_generic_write_end(file, mapping, pos, len, copied,
Nick Piggin's avatar
Nick Piggin committed
899
							page, fsdata);
900
		copied = ret2;
901
		if (pos + len > inode->i_size && ext4_can_truncate(inode))
902
903
904
905
906
			/* 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);
907
908
		if (ret2 < 0)
			ret = ret2;
909
	}
910
	ret2 = ext4_journal_stop(handle);
911
912
	if (!ret)
		ret = ret2;
Nick Piggin's avatar
Nick Piggin committed
913

914
	if (pos + len > inode->i_size) {
915
		ext4_truncate_failed_write(inode);
916
		/*
917
		 * If truncate failed early the inode might still be
918
919
920
921
922
923
924
925
		 * 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);
	}


Nick Piggin's avatar
Nick Piggin committed
926
	return ret ? ret : copied;
927
928
}

Nick Piggin's avatar
Nick Piggin committed
929
static int ext4_writeback_write_end(struct file *file,
930
931
932
				    struct address_space *mapping,
				    loff_t pos, unsigned len, unsigned copied,
				    struct page *page, void *fsdata)
933
{
934
	handle_t *handle = ext4_journal_current_handle();
935
	struct inode *inode = mapping->host;
936
937
	int ret = 0, ret2;

938
	trace_ext4_writeback_write_end(inode, pos, len, copied);
939
	ret2 = ext4_generic_write_end(file, mapping, pos, len, copied,
Nick Piggin's avatar
Nick Piggin committed
940
							page, fsdata);
941
	copied = ret2;
942
	if (pos + len > inode->i_size && ext4_can_truncate(inode))
943
944
945
946
947
948
		/* 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);

949
950
	if (ret2 < 0)
		ret = ret2;
951

952
	ret2 = ext4_journal_stop(handle);
953
954
	if (!ret)
		ret = ret2;
Nick Piggin's avatar
Nick Piggin committed
955

956
	if (pos + len > inode->i_size) {
957
		ext4_truncate_failed_write(inode);
958
		/*
959
		 * If truncate failed early the inode might still be
960
961
962
963
964
965
966
		 * 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);
	}

Nick Piggin's avatar
Nick Piggin committed
967
	return ret ? ret : copied;
968
969
}

Nick Piggin's avatar
Nick Piggin committed
970
static int ext4_journalled_write_end(struct file *file,
971
972
973
				     struct address_space *mapping,
				     loff_t pos, unsigned len, unsigned copied,
				     struct page *page, void *fsdata)
974
{
975
	handle_t *handle = ext4_journal_current_handle();
Nick Piggin's avatar
Nick Piggin committed
976
	struct inode *inode = mapping->host;
977
978
	int ret = 0, ret2;
	int partial = 0;
Nick Piggin's avatar
Nick Piggin committed
979
	unsigned from, to;
980
	loff_t new_i_size;
981

982
	trace_ext4_journalled_write_end(inode, pos, len, copied);
Nick Piggin's avatar
Nick Piggin committed
983
984
985
	from = pos & (PAGE_CACHE_SIZE - 1);
	to = from + len;

986
987
	BUG_ON(!ext4_handle_valid(handle));

Nick Piggin's avatar
Nick Piggin committed
988
989
990
991
992
	if (copied < len) {
		if (!PageUptodate(page))
			copied = 0;
		page_zero_new_buffers(page, from+copied, to);
	}
993
994

	ret = walk_page_buffers(handle, page_buffers(page), from,
Nick Piggin's avatar
Nick Piggin committed
995
				to, &partial, write_end_fn);
996
997
	if (!partial)
		SetPageUptodate(page);
998
999
	new_i_size = pos + copied;
	if (new_i_size > inode->i_size)
Nick Piggin's avatar
Nick Piggin committed
1000
		i_size_write(inode, pos+copied);
For faster browsing, not all history is shown. View entire blame