file.c 65.3 KB
Newer Older
C
Chris Mason 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/*
 * Copyright (C) 2007 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

C
Chris Mason 已提交
19 20 21 22 23 24 25 26
#include <linux/fs.h>
#include <linux/pagemap.h>
#include <linux/highmem.h>
#include <linux/time.h>
#include <linux/init.h>
#include <linux/string.h>
#include <linux/backing-dev.h>
#include <linux/mpage.h>
27
#include <linux/aio.h>
28
#include <linux/falloc.h>
C
Chris Mason 已提交
29 30 31 32
#include <linux/swap.h>
#include <linux/writeback.h>
#include <linux/statfs.h>
#include <linux/compat.h>
33
#include <linux/slab.h>
34
#include <linux/btrfs.h>
C
Chris Mason 已提交
35 36 37 38 39
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
#include "btrfs_inode.h"
#include "print-tree.h"
40 41
#include "tree-log.h"
#include "locking.h"
J
Josef Bacik 已提交
42
#include "volumes.h"
C
Chris Mason 已提交
43

44
static struct kmem_cache *btrfs_inode_defrag_cachep;
C
Chris Mason 已提交
45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69
/*
 * when auto defrag is enabled we
 * queue up these defrag structs to remember which
 * inodes need defragging passes
 */
struct inode_defrag {
	struct rb_node rb_node;
	/* objectid */
	u64 ino;
	/*
	 * transid where the defrag was added, we search for
	 * extents newer than this
	 */
	u64 transid;

	/* root objectid */
	u64 root;

	/* last offset we were able to defrag */
	u64 last_offset;

	/* if we've wrapped around back to zero once already */
	int cycled;
};

70 71 72 73 74 75 76 77 78 79 80 81 82 83 84
static int __compare_inode_defrag(struct inode_defrag *defrag1,
				  struct inode_defrag *defrag2)
{
	if (defrag1->root > defrag2->root)
		return 1;
	else if (defrag1->root < defrag2->root)
		return -1;
	else if (defrag1->ino > defrag2->ino)
		return 1;
	else if (defrag1->ino < defrag2->ino)
		return -1;
	else
		return 0;
}

C
Chris Mason 已提交
85 86 87 88 89 90 91 92 93
/* pop a record for an inode into the defrag tree.  The lock
 * must be held already
 *
 * If you're inserting a record for an older transid than an
 * existing record, the transid already in the tree is lowered
 *
 * If an existing record is found the defrag item you
 * pass in is freed
 */
94
static int __btrfs_add_inode_defrag(struct inode *inode,
C
Chris Mason 已提交
95 96 97 98 99 100
				    struct inode_defrag *defrag)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct inode_defrag *entry;
	struct rb_node **p;
	struct rb_node *parent = NULL;
101
	int ret;
C
Chris Mason 已提交
102 103 104 105 106 107

	p = &root->fs_info->defrag_inodes.rb_node;
	while (*p) {
		parent = *p;
		entry = rb_entry(parent, struct inode_defrag, rb_node);

108 109
		ret = __compare_inode_defrag(defrag, entry);
		if (ret < 0)
C
Chris Mason 已提交
110
			p = &parent->rb_left;
111
		else if (ret > 0)
C
Chris Mason 已提交
112 113 114 115 116 117 118 119 120 121
			p = &parent->rb_right;
		else {
			/* if we're reinserting an entry for
			 * an old defrag run, make sure to
			 * lower the transid of our existing record
			 */
			if (defrag->transid < entry->transid)
				entry->transid = defrag->transid;
			if (defrag->last_offset > entry->last_offset)
				entry->last_offset = defrag->last_offset;
122
			return -EEXIST;
C
Chris Mason 已提交
123 124
		}
	}
125
	set_bit(BTRFS_INODE_IN_DEFRAG, &BTRFS_I(inode)->runtime_flags);
C
Chris Mason 已提交
126 127
	rb_link_node(&defrag->rb_node, parent, p);
	rb_insert_color(&defrag->rb_node, &root->fs_info->defrag_inodes);
128 129
	return 0;
}
C
Chris Mason 已提交
130

131 132 133 134 135 136 137
static inline int __need_auto_defrag(struct btrfs_root *root)
{
	if (!btrfs_test_opt(root, AUTO_DEFRAG))
		return 0;

	if (btrfs_fs_closing(root->fs_info))
		return 0;
C
Chris Mason 已提交
138

139
	return 1;
C
Chris Mason 已提交
140 141 142 143 144 145 146 147 148 149 150 151
}

/*
 * insert a defrag record for this inode if auto defrag is
 * enabled
 */
int btrfs_add_inode_defrag(struct btrfs_trans_handle *trans,
			   struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct inode_defrag *defrag;
	u64 transid;
152
	int ret;
C
Chris Mason 已提交
153

154
	if (!__need_auto_defrag(root))
C
Chris Mason 已提交
155 156
		return 0;

157
	if (test_bit(BTRFS_INODE_IN_DEFRAG, &BTRFS_I(inode)->runtime_flags))
C
Chris Mason 已提交
158 159 160 161 162 163 164
		return 0;

	if (trans)
		transid = trans->transid;
	else
		transid = BTRFS_I(inode)->root->last_trans;

165
	defrag = kmem_cache_zalloc(btrfs_inode_defrag_cachep, GFP_NOFS);
C
Chris Mason 已提交
166 167 168
	if (!defrag)
		return -ENOMEM;

169
	defrag->ino = btrfs_ino(inode);
C
Chris Mason 已提交
170 171 172 173
	defrag->transid = transid;
	defrag->root = root->root_key.objectid;

	spin_lock(&root->fs_info->defrag_inodes_lock);
174 175 176 177 178 179 180 181 182 183
	if (!test_bit(BTRFS_INODE_IN_DEFRAG, &BTRFS_I(inode)->runtime_flags)) {
		/*
		 * If we set IN_DEFRAG flag and evict the inode from memory,
		 * and then re-read this inode, this new inode doesn't have
		 * IN_DEFRAG flag. At the case, we may find the existed defrag.
		 */
		ret = __btrfs_add_inode_defrag(inode, defrag);
		if (ret)
			kmem_cache_free(btrfs_inode_defrag_cachep, defrag);
	} else {
184
		kmem_cache_free(btrfs_inode_defrag_cachep, defrag);
185
	}
C
Chris Mason 已提交
186
	spin_unlock(&root->fs_info->defrag_inodes_lock);
187
	return 0;
C
Chris Mason 已提交
188 189 190
}

/*
191 192 193
 * Requeue the defrag object. If there is a defrag object that points to
 * the same inode in the tree, we will merge them together (by
 * __btrfs_add_inode_defrag()) and free the one that we want to requeue.
C
Chris Mason 已提交
194
 */
195 196
static void btrfs_requeue_inode_defrag(struct inode *inode,
				       struct inode_defrag *defrag)
197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret;

	if (!__need_auto_defrag(root))
		goto out;

	/*
	 * Here we don't check the IN_DEFRAG flag, because we need merge
	 * them together.
	 */
	spin_lock(&root->fs_info->defrag_inodes_lock);
	ret = __btrfs_add_inode_defrag(inode, defrag);
	spin_unlock(&root->fs_info->defrag_inodes_lock);
	if (ret)
		goto out;
	return;
out:
	kmem_cache_free(btrfs_inode_defrag_cachep, defrag);
}

C
Chris Mason 已提交
218
/*
219 220
 * pick the defragable inode that we want, if it doesn't exist, we will get
 * the next one.
C
Chris Mason 已提交
221
 */
222 223
static struct inode_defrag *
btrfs_pick_defrag_inode(struct btrfs_fs_info *fs_info, u64 root, u64 ino)
C
Chris Mason 已提交
224 225
{
	struct inode_defrag *entry = NULL;
226
	struct inode_defrag tmp;
C
Chris Mason 已提交
227 228
	struct rb_node *p;
	struct rb_node *parent = NULL;
229 230 231 232
	int ret;

	tmp.ino = ino;
	tmp.root = root;
C
Chris Mason 已提交
233

234 235
	spin_lock(&fs_info->defrag_inodes_lock);
	p = fs_info->defrag_inodes.rb_node;
C
Chris Mason 已提交
236 237 238 239
	while (p) {
		parent = p;
		entry = rb_entry(parent, struct inode_defrag, rb_node);

240 241
		ret = __compare_inode_defrag(&tmp, entry);
		if (ret < 0)
C
Chris Mason 已提交
242
			p = parent->rb_left;
243
		else if (ret > 0)
C
Chris Mason 已提交
244 245
			p = parent->rb_right;
		else
246
			goto out;
C
Chris Mason 已提交
247 248
	}

249 250 251
	if (parent && __compare_inode_defrag(&tmp, entry) > 0) {
		parent = rb_next(parent);
		if (parent)
C
Chris Mason 已提交
252
			entry = rb_entry(parent, struct inode_defrag, rb_node);
253 254
		else
			entry = NULL;
C
Chris Mason 已提交
255
	}
256 257 258 259 260
out:
	if (entry)
		rb_erase(parent, &fs_info->defrag_inodes);
	spin_unlock(&fs_info->defrag_inodes_lock);
	return entry;
C
Chris Mason 已提交
261 262
}

263
void btrfs_cleanup_defrag_inodes(struct btrfs_fs_info *fs_info)
C
Chris Mason 已提交
264 265
{
	struct inode_defrag *defrag;
266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290
	struct rb_node *node;

	spin_lock(&fs_info->defrag_inodes_lock);
	node = rb_first(&fs_info->defrag_inodes);
	while (node) {
		rb_erase(node, &fs_info->defrag_inodes);
		defrag = rb_entry(node, struct inode_defrag, rb_node);
		kmem_cache_free(btrfs_inode_defrag_cachep, defrag);

		if (need_resched()) {
			spin_unlock(&fs_info->defrag_inodes_lock);
			cond_resched();
			spin_lock(&fs_info->defrag_inodes_lock);
		}

		node = rb_first(&fs_info->defrag_inodes);
	}
	spin_unlock(&fs_info->defrag_inodes_lock);
}

#define BTRFS_DEFRAG_BATCH	1024

static int __btrfs_run_defrag_inode(struct btrfs_fs_info *fs_info,
				    struct inode_defrag *defrag)
{
C
Chris Mason 已提交
291 292 293 294 295
	struct btrfs_root *inode_root;
	struct inode *inode;
	struct btrfs_key key;
	struct btrfs_ioctl_defrag_range_args range;
	int num_defrag;
296 297
	int index;
	int ret;
C
Chris Mason 已提交
298

299 300 301 302
	/* get the inode */
	key.objectid = defrag->root;
	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
	key.offset = (u64)-1;
303 304 305

	index = srcu_read_lock(&fs_info->subvol_srcu);

306 307
	inode_root = btrfs_read_fs_root_no_name(fs_info, &key);
	if (IS_ERR(inode_root)) {
308 309 310
		ret = PTR_ERR(inode_root);
		goto cleanup;
	}
311 312 313 314 315 316

	key.objectid = defrag->ino;
	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
	key.offset = 0;
	inode = btrfs_iget(fs_info->sb, &key, inode_root, NULL);
	if (IS_ERR(inode)) {
317 318
		ret = PTR_ERR(inode);
		goto cleanup;
319
	}
320
	srcu_read_unlock(&fs_info->subvol_srcu, index);
321 322 323

	/* do a chunk of defrag */
	clear_bit(BTRFS_INODE_IN_DEFRAG, &BTRFS_I(inode)->runtime_flags);
C
Chris Mason 已提交
324 325
	memset(&range, 0, sizeof(range));
	range.len = (u64)-1;
326
	range.start = defrag->last_offset;
M
Miao Xie 已提交
327 328

	sb_start_write(fs_info->sb);
329 330
	num_defrag = btrfs_defrag_file(inode, NULL, &range, defrag->transid,
				       BTRFS_DEFRAG_BATCH);
M
Miao Xie 已提交
331
	sb_end_write(fs_info->sb);
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354
	/*
	 * if we filled the whole defrag batch, there
	 * must be more work to do.  Queue this defrag
	 * again
	 */
	if (num_defrag == BTRFS_DEFRAG_BATCH) {
		defrag->last_offset = range.start;
		btrfs_requeue_inode_defrag(inode, defrag);
	} else if (defrag->last_offset && !defrag->cycled) {
		/*
		 * we didn't fill our defrag batch, but
		 * we didn't start at zero.  Make sure we loop
		 * around to the start of the file.
		 */
		defrag->last_offset = 0;
		defrag->cycled = 1;
		btrfs_requeue_inode_defrag(inode, defrag);
	} else {
		kmem_cache_free(btrfs_inode_defrag_cachep, defrag);
	}

	iput(inode);
	return 0;
355 356 357 358
cleanup:
	srcu_read_unlock(&fs_info->subvol_srcu, index);
	kmem_cache_free(btrfs_inode_defrag_cachep, defrag);
	return ret;
359 360 361 362 363 364 365 366 367 368 369
}

/*
 * run through the list of inodes in the FS that need
 * defragging
 */
int btrfs_run_defrag_inodes(struct btrfs_fs_info *fs_info)
{
	struct inode_defrag *defrag;
	u64 first_ino = 0;
	u64 root_objectid = 0;
C
Chris Mason 已提交
370 371

	atomic_inc(&fs_info->defrag_running);
372
	while (1) {
M
Miao Xie 已提交
373 374 375 376 377
		/* Pause the auto defragger. */
		if (test_bit(BTRFS_FS_STATE_REMOUNTING,
			     &fs_info->fs_state))
			break;

378 379
		if (!__need_auto_defrag(fs_info->tree_root))
			break;
C
Chris Mason 已提交
380 381

		/* find an inode to defrag */
382 383
		defrag = btrfs_pick_defrag_inode(fs_info, root_objectid,
						 first_ino);
C
Chris Mason 已提交
384
		if (!defrag) {
385
			if (root_objectid || first_ino) {
386
				root_objectid = 0;
C
Chris Mason 已提交
387 388 389 390 391 392 393 394
				first_ino = 0;
				continue;
			} else {
				break;
			}
		}

		first_ino = defrag->ino + 1;
395
		root_objectid = defrag->root;
C
Chris Mason 已提交
396

397
		__btrfs_run_defrag_inode(fs_info, defrag);
C
Chris Mason 已提交
398 399 400 401 402 403 404 405 406 407
	}
	atomic_dec(&fs_info->defrag_running);

	/*
	 * during unmount, we use the transaction_wait queue to
	 * wait for the defragger to stop
	 */
	wake_up(&fs_info->transaction_wait);
	return 0;
}
C
Chris Mason 已提交
408

C
Chris Mason 已提交
409 410 411
/* simple helper to fault in pages and copy.  This should go away
 * and be replaced with calls into generic code.
 */
C
Chris Mason 已提交
412
static noinline int btrfs_copy_from_user(loff_t pos, int num_pages,
J
Josef Bacik 已提交
413
					 size_t write_bytes,
414
					 struct page **prepared_pages,
415
					 struct iov_iter *i)
C
Chris Mason 已提交
416
{
417
	size_t copied = 0;
J
Josef Bacik 已提交
418
	size_t total_copied = 0;
419
	int pg = 0;
C
Chris Mason 已提交
420 421
	int offset = pos & (PAGE_CACHE_SIZE - 1);

422
	while (write_bytes > 0) {
C
Chris Mason 已提交
423 424
		size_t count = min_t(size_t,
				     PAGE_CACHE_SIZE - offset, write_bytes);
425
		struct page *page = prepared_pages[pg];
426 427 428 429 430 431 432 433 434
		/*
		 * Copy data from userspace to the current page
		 *
		 * Disable pagefault to avoid recursive lock since
		 * the pages are already locked
		 */
		pagefault_disable();
		copied = iov_iter_copy_from_user_atomic(page, i, offset, count);
		pagefault_enable();
435

C
Chris Mason 已提交
436 437
		/* Flush processor's dcache for this page */
		flush_dcache_page(page);
438 439 440 441 442 443 444 445 446 447 448 449 450

		/*
		 * if we get a partial write, we can end up with
		 * partially up to date pages.  These add
		 * a lot of complexity, so make sure they don't
		 * happen by forcing this copy to be retried.
		 *
		 * The rest of the btrfs_file_write code will fall
		 * back to page at a time copies after we return 0.
		 */
		if (!PageUptodate(page) && copied < count)
			copied = 0;

451 452
		iov_iter_advance(i, copied);
		write_bytes -= copied;
453
		total_copied += copied;
C
Chris Mason 已提交
454

455
		/* Return to btrfs_file_aio_write to fault page */
J
Josef Bacik 已提交
456
		if (unlikely(copied == 0))
457
			break;
458 459 460 461 462 463 464

		if (unlikely(copied < PAGE_CACHE_SIZE - offset)) {
			offset += copied;
		} else {
			pg++;
			offset = 0;
		}
C
Chris Mason 已提交
465
	}
466
	return total_copied;
C
Chris Mason 已提交
467 468
}

C
Chris Mason 已提交
469 470 471
/*
 * unlocks pages after btrfs_file_write is done with them
 */
472
static void btrfs_drop_pages(struct page **pages, size_t num_pages)
C
Chris Mason 已提交
473 474 475
{
	size_t i;
	for (i = 0; i < num_pages; i++) {
C
Chris Mason 已提交
476 477 478 479
		/* page checked is some magic around finding pages that
		 * have been modified without going through btrfs_set_page_dirty
		 * clear it here
		 */
C
Chris Mason 已提交
480
		ClearPageChecked(pages[i]);
C
Chris Mason 已提交
481 482 483 484 485 486
		unlock_page(pages[i]);
		mark_page_accessed(pages[i]);
		page_cache_release(pages[i]);
	}
}

C
Chris Mason 已提交
487 488 489 490 491 492 493 494
/*
 * after copy_from_user, pages need to be dirtied and we need to make
 * sure holes are created between the current EOF and the start of
 * any next extents (if required).
 *
 * this also makes the decision about creating an inline extent vs
 * doing real data extents, marking pages dirty and delalloc as required.
 */
495
int btrfs_dirty_pages(struct btrfs_root *root, struct inode *inode,
496 497 498
			     struct page **pages, size_t num_pages,
			     loff_t pos, size_t write_bytes,
			     struct extent_state **cached)
C
Chris Mason 已提交
499 500
{
	int err = 0;
501
	int i;
502
	u64 num_bytes;
503 504 505 506
	u64 start_pos;
	u64 end_of_last_block;
	u64 end_pos = pos + write_bytes;
	loff_t isize = i_size_read(inode);
C
Chris Mason 已提交
507

508
	start_pos = pos & ~((u64)root->sectorsize - 1);
509
	num_bytes = ALIGN(write_bytes + pos - start_pos, root->sectorsize);
C
Chris Mason 已提交
510

511
	end_of_last_block = start_pos + num_bytes - 1;
512
	err = btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block,
513
					cached);
J
Josef Bacik 已提交
514 515
	if (err)
		return err;
J
Josef Bacik 已提交
516

C
Chris Mason 已提交
517 518 519 520 521
	for (i = 0; i < num_pages; i++) {
		struct page *p = pages[i];
		SetPageUptodate(p);
		ClearPageChecked(p);
		set_page_dirty(p);
522
	}
J
Josef Bacik 已提交
523 524 525 526 527 528 529

	/*
	 * we've only changed i_size in ram, and we haven't updated
	 * the disk i_size.  There is no need to log the inode
	 * at this time.
	 */
	if (end_pos > isize)
530
		i_size_write(inode, end_pos);
531
	return 0;
C
Chris Mason 已提交
532 533
}

C
Chris Mason 已提交
534 535 536 537
/*
 * this drops all the extents in the cache that intersect the range
 * [start, end].  Existing extents are split as required.
 */
538 539
void btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
			     int skip_pinned)
540 541
{
	struct extent_map *em;
542 543
	struct extent_map *split = NULL;
	struct extent_map *split2 = NULL;
544
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
545
	u64 len = end - start + 1;
J
Josef Bacik 已提交
546
	u64 gen;
547 548
	int ret;
	int testend = 1;
549
	unsigned long flags;
C
Chris Mason 已提交
550
	int compressed = 0;
J
Josef Bacik 已提交
551
	bool modified;
552

553
	WARN_ON(end < start);
554
	if (end == (u64)-1) {
555
		len = (u64)-1;
556 557
		testend = 0;
	}
C
Chris Mason 已提交
558
	while (1) {
559 560
		int no_splits = 0;

J
Josef Bacik 已提交
561
		modified = false;
562
		if (!split)
563
			split = alloc_extent_map();
564
		if (!split2)
565
			split2 = alloc_extent_map();
566 567
		if (!split || !split2)
			no_splits = 1;
568

569
		write_lock(&em_tree->lock);
570
		em = lookup_extent_mapping(em_tree, start, len);
571
		if (!em) {
572
			write_unlock(&em_tree->lock);
573
			break;
574
		}
575
		flags = em->flags;
J
Josef Bacik 已提交
576
		gen = em->generation;
577
		if (skip_pinned && test_bit(EXTENT_FLAG_PINNED, &em->flags)) {
578
			if (testend && em->start + em->len >= start + len) {
579
				free_extent_map(em);
C
Chris Mason 已提交
580
				write_unlock(&em_tree->lock);
581 582
				break;
			}
583 584
			start = em->start + em->len;
			if (testend)
585 586
				len = start + len - (em->start + em->len);
			free_extent_map(em);
C
Chris Mason 已提交
587
			write_unlock(&em_tree->lock);
588 589
			continue;
		}
C
Chris Mason 已提交
590
		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
591
		clear_bit(EXTENT_FLAG_PINNED, &em->flags);
L
Liu Bo 已提交
592
		clear_bit(EXTENT_FLAG_LOGGING, &flags);
J
Josef Bacik 已提交
593
		modified = !list_empty(&em->list);
594
		remove_extent_mapping(em_tree, em);
595 596
		if (no_splits)
			goto next;
597

598
		if (em->start < start) {
599 600
			split->start = em->start;
			split->len = start - em->start;
601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620

			if (em->block_start < EXTENT_MAP_LAST_BYTE) {
				split->orig_start = em->orig_start;
				split->block_start = em->block_start;

				if (compressed)
					split->block_len = em->block_len;
				else
					split->block_len = split->len;
				split->orig_block_len = max(split->block_len,
						em->orig_block_len);
				split->ram_bytes = em->ram_bytes;
			} else {
				split->orig_start = split->start;
				split->block_len = 0;
				split->block_start = em->block_start;
				split->orig_block_len = 0;
				split->ram_bytes = split->len;
			}

J
Josef Bacik 已提交
621
			split->generation = gen;
622
			split->bdev = em->bdev;
623
			split->flags = flags;
624
			split->compress_type = em->compress_type;
J
Josef Bacik 已提交
625
			ret = add_extent_mapping(em_tree, split, modified);
626
			BUG_ON(ret); /* Logic error */
627 628 629 630
			free_extent_map(split);
			split = split2;
			split2 = NULL;
		}
631
		if (testend && em->start + em->len > start + len) {
632 633 634 635 636
			u64 diff = start + len - em->start;

			split->start = start + len;
			split->len = em->start + em->len - (start + len);
			split->bdev = em->bdev;
637
			split->flags = flags;
638
			split->compress_type = em->compress_type;
J
Josef Bacik 已提交
639
			split->generation = gen;
640 641 642

			if (em->block_start < EXTENT_MAP_LAST_BYTE) {
				split->orig_block_len = max(em->block_len,
643
						    em->orig_block_len);
644

645 646 647 648 649 650 651 652 653 654 655
				split->ram_bytes = em->ram_bytes;
				if (compressed) {
					split->block_len = em->block_len;
					split->block_start = em->block_start;
					split->orig_start = em->orig_start;
				} else {
					split->block_len = split->len;
					split->block_start = em->block_start
						+ diff;
					split->orig_start = em->orig_start;
				}
C
Chris Mason 已提交
656
			} else {
657 658 659 660 661
				split->ram_bytes = split->len;
				split->orig_start = split->start;
				split->block_len = 0;
				split->block_start = em->block_start;
				split->orig_block_len = 0;
C
Chris Mason 已提交
662
			}
663

J
Josef Bacik 已提交
664
			ret = add_extent_mapping(em_tree, split, modified);
665
			BUG_ON(ret); /* Logic error */
666 667 668
			free_extent_map(split);
			split = NULL;
		}
669
next:
670
		write_unlock(&em_tree->lock);
671

672 673 674 675 676
		/* once for us */
		free_extent_map(em);
		/* once for the tree*/
		free_extent_map(em);
	}
677 678 679 680
	if (split)
		free_extent_map(split);
	if (split2)
		free_extent_map(split2);
681 682
}

C
Chris Mason 已提交
683 684 685 686 687 688 689 690 691
/*
 * this is very complex, but the basic idea is to drop all extents
 * in the range start - end.  hint_block is filled in with a block number
 * that would be a good hint to the block allocator for this file.
 *
 * If an extent intersects the range but is not entirely inside the range
 * it is either truncated or split.  Anything entirely inside the range
 * is deleted from the tree.
 */
J
Josef Bacik 已提交
692 693 694
int __btrfs_drop_extents(struct btrfs_trans_handle *trans,
			 struct btrfs_root *root, struct inode *inode,
			 struct btrfs_path *path, u64 start, u64 end,
J
Josef Bacik 已提交
695
			 u64 *drop_end, int drop_cache)
C
Chris Mason 已提交
696
{
697
	struct extent_buffer *leaf;
Y
Yan, Zheng 已提交
698
	struct btrfs_file_extent_item *fi;
699
	struct btrfs_key key;
Y
Yan, Zheng 已提交
700
	struct btrfs_key new_key;
L
Li Zefan 已提交
701
	u64 ino = btrfs_ino(inode);
Y
Yan, Zheng 已提交
702 703 704 705 706 707 708 709
	u64 search_start = start;
	u64 disk_bytenr = 0;
	u64 num_bytes = 0;
	u64 extent_offset = 0;
	u64 extent_end = 0;
	int del_nr = 0;
	int del_slot = 0;
	int extent_type;
C
Chris Mason 已提交
710
	int recow;
711
	int ret;
712
	int modify_tree = -1;
J
Josef Bacik 已提交
713
	int update_refs = (root->ref_cows || root == root->fs_info->tree_root);
714
	int found = 0;
C
Chris Mason 已提交
715

C
Chris Mason 已提交
716 717
	if (drop_cache)
		btrfs_drop_extent_cache(inode, start, end - 1, 0);
718

719 720 721
	if (start >= BTRFS_I(inode)->disk_i_size)
		modify_tree = 0;

C
Chris Mason 已提交
722
	while (1) {
C
Chris Mason 已提交
723
		recow = 0;
L
Li Zefan 已提交
724
		ret = btrfs_lookup_file_extent(trans, root, path, ino,
725
					       search_start, modify_tree);
C
Chris Mason 已提交
726
		if (ret < 0)
Y
Yan, Zheng 已提交
727 728 729 730
			break;
		if (ret > 0 && path->slots[0] > 0 && search_start == start) {
			leaf = path->nodes[0];
			btrfs_item_key_to_cpu(leaf, &key, path->slots[0] - 1);
L
Li Zefan 已提交
731
			if (key.objectid == ino &&
Y
Yan, Zheng 已提交
732 733
			    key.type == BTRFS_EXTENT_DATA_KEY)
				path->slots[0]--;
C
Chris Mason 已提交
734
		}
Y
Yan, Zheng 已提交
735
		ret = 0;
736
next_slot:
737
		leaf = path->nodes[0];
Y
Yan, Zheng 已提交
738 739 740 741 742 743 744 745
		if (path->slots[0] >= btrfs_header_nritems(leaf)) {
			BUG_ON(del_nr > 0);
			ret = btrfs_next_leaf(root, path);
			if (ret < 0)
				break;
			if (ret > 0) {
				ret = 0;
				break;
746
			}
Y
Yan, Zheng 已提交
747 748 749 750 751
			leaf = path->nodes[0];
			recow = 1;
		}

		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
L
Li Zefan 已提交
752
		if (key.objectid > ino ||
Y
Yan, Zheng 已提交
753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769
		    key.type > BTRFS_EXTENT_DATA_KEY || key.offset >= end)
			break;

		fi = btrfs_item_ptr(leaf, path->slots[0],
				    struct btrfs_file_extent_item);
		extent_type = btrfs_file_extent_type(leaf, fi);

		if (extent_type == BTRFS_FILE_EXTENT_REG ||
		    extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
			disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
			num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
			extent_offset = btrfs_file_extent_offset(leaf, fi);
			extent_end = key.offset +
				btrfs_file_extent_num_bytes(leaf, fi);
		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
			extent_end = key.offset +
				btrfs_file_extent_inline_len(leaf, fi);
770
		} else {
Y
Yan, Zheng 已提交
771
			WARN_ON(1);
772
			extent_end = search_start;
C
Chris Mason 已提交
773 774
		}

Y
Yan, Zheng 已提交
775 776
		if (extent_end <= search_start) {
			path->slots[0]++;
777
			goto next_slot;
C
Chris Mason 已提交
778 779
		}

780
		found = 1;
Y
Yan, Zheng 已提交
781
		search_start = max(key.offset, start);
782 783
		if (recow || !modify_tree) {
			modify_tree = -1;
784
			btrfs_release_path(path);
Y
Yan, Zheng 已提交
785
			continue;
C
Chris Mason 已提交
786
		}
Y
Yan Zheng 已提交
787

Y
Yan, Zheng 已提交
788 789 790 791 792 793 794 795 796 797 798 799 800
		/*
		 *     | - range to drop - |
		 *  | -------- extent -------- |
		 */
		if (start > key.offset && end < extent_end) {
			BUG_ON(del_nr > 0);
			BUG_ON(extent_type == BTRFS_FILE_EXTENT_INLINE);

			memcpy(&new_key, &key, sizeof(new_key));
			new_key.offset = start;
			ret = btrfs_duplicate_item(trans, root, path,
						   &new_key);
			if (ret == -EAGAIN) {
801
				btrfs_release_path(path);
Y
Yan, Zheng 已提交
802
				continue;
Y
Yan Zheng 已提交
803
			}
Y
Yan, Zheng 已提交
804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821
			if (ret < 0)
				break;

			leaf = path->nodes[0];
			fi = btrfs_item_ptr(leaf, path->slots[0] - 1,
					    struct btrfs_file_extent_item);
			btrfs_set_file_extent_num_bytes(leaf, fi,
							start - key.offset);

			fi = btrfs_item_ptr(leaf, path->slots[0],
					    struct btrfs_file_extent_item);

			extent_offset += start - key.offset;
			btrfs_set_file_extent_offset(leaf, fi, extent_offset);
			btrfs_set_file_extent_num_bytes(leaf, fi,
							extent_end - start);
			btrfs_mark_buffer_dirty(leaf);

J
Josef Bacik 已提交
822
			if (update_refs && disk_bytenr > 0) {
823
				ret = btrfs_inc_extent_ref(trans, root,
Y
Yan, Zheng 已提交
824 825 826
						disk_bytenr, num_bytes, 0,
						root->root_key.objectid,
						new_key.objectid,
A
Arne Jansen 已提交
827
						start - extent_offset, 0);
828
				BUG_ON(ret); /* -ENOMEM */
829
			}
Y
Yan, Zheng 已提交
830
			key.offset = start;
Y
Yan Zheng 已提交
831
		}
Y
Yan, Zheng 已提交
832 833 834 835 836 837
		/*
		 *  | ---- range to drop ----- |
		 *      | -------- extent -------- |
		 */
		if (start <= key.offset && end < extent_end) {
			BUG_ON(extent_type == BTRFS_FILE_EXTENT_INLINE);
Y
Yan Zheng 已提交
838

Y
Yan, Zheng 已提交
839 840
			memcpy(&new_key, &key, sizeof(new_key));
			new_key.offset = end;
841
			btrfs_set_item_key_safe(root, path, &new_key);
Y
Yan Zheng 已提交
842

Y
Yan, Zheng 已提交
843 844 845 846 847
			extent_offset += end - key.offset;
			btrfs_set_file_extent_offset(leaf, fi, extent_offset);
			btrfs_set_file_extent_num_bytes(leaf, fi,
							extent_end - end);
			btrfs_mark_buffer_dirty(leaf);
848
			if (update_refs && disk_bytenr > 0)
Y
Yan, Zheng 已提交
849 850
				inode_sub_bytes(inode, end - key.offset);
			break;
C
Chris Mason 已提交
851
		}
852

Y
Yan, Zheng 已提交
853 854 855 856 857 858 859 860
		search_start = extent_end;
		/*
		 *       | ---- range to drop ----- |
		 *  | -------- extent -------- |
		 */
		if (start > key.offset && end >= extent_end) {
			BUG_ON(del_nr > 0);
			BUG_ON(extent_type == BTRFS_FILE_EXTENT_INLINE);
861

Y
Yan, Zheng 已提交
862 863 864
			btrfs_set_file_extent_num_bytes(leaf, fi,
							start - key.offset);
			btrfs_mark_buffer_dirty(leaf);
865
			if (update_refs && disk_bytenr > 0)
Y
Yan, Zheng 已提交
866 867 868
				inode_sub_bytes(inode, extent_end - start);
			if (end == extent_end)
				break;
C
Chris Mason 已提交
869

Y
Yan, Zheng 已提交
870 871
			path->slots[0]++;
			goto next_slot;
Z
Zheng Yan 已提交
872 873
		}

Y
Yan, Zheng 已提交
874 875 876 877 878 879 880 881 882 883 884 885
		/*
		 *  | ---- range to drop ----- |
		 *    | ------ extent ------ |
		 */
		if (start <= key.offset && end >= extent_end) {
			if (del_nr == 0) {
				del_slot = path->slots[0];
				del_nr = 1;
			} else {
				BUG_ON(del_slot + del_nr != path->slots[0]);
				del_nr++;
			}
Z
Zheng Yan 已提交
886

J
Josef Bacik 已提交
887 888
			if (update_refs &&
			    extent_type == BTRFS_FILE_EXTENT_INLINE) {
889
				inode_sub_bytes(inode,
Y
Yan, Zheng 已提交
890 891 892
						extent_end - key.offset);
				extent_end = ALIGN(extent_end,
						   root->sectorsize);
J
Josef Bacik 已提交
893
			} else if (update_refs && disk_bytenr > 0) {
Z
Zheng Yan 已提交
894
				ret = btrfs_free_extent(trans, root,
Y
Yan, Zheng 已提交
895 896
						disk_bytenr, num_bytes, 0,
						root->root_key.objectid,
897
						key.objectid, key.offset -
A
Arne Jansen 已提交
898
						extent_offset, 0);
899
				BUG_ON(ret); /* -ENOMEM */
Y
Yan, Zheng 已提交
900 901
				inode_sub_bytes(inode,
						extent_end - key.offset);
Z
Zheng Yan 已提交
902 903
			}

Y
Yan, Zheng 已提交
904 905 906 907 908 909 910 911 912 913
			if (end == extent_end)
				break;

			if (path->slots[0] + 1 < btrfs_header_nritems(leaf)) {
				path->slots[0]++;
				goto next_slot;
			}

			ret = btrfs_del_items(trans, root, path, del_slot,
					      del_nr);
914 915
			if (ret) {
				btrfs_abort_transaction(trans, root, ret);
J
Josef Bacik 已提交
916
				break;
917
			}
Y
Yan, Zheng 已提交
918 919 920 921

			del_nr = 0;
			del_slot = 0;

922
			btrfs_release_path(path);
Y
Yan, Zheng 已提交
923
			continue;
C
Chris Mason 已提交
924
		}
Y
Yan, Zheng 已提交
925 926

		BUG_ON(1);
C
Chris Mason 已提交
927
	}
Y
Yan, Zheng 已提交
928

929
	if (!ret && del_nr > 0) {
Y
Yan, Zheng 已提交
930
		ret = btrfs_del_items(trans, root, path, del_slot, del_nr);
931 932
		if (ret)
			btrfs_abort_transaction(trans, root, ret);
Y
Yan Zheng 已提交
933
	}
Y
Yan, Zheng 已提交
934

J
Josef Bacik 已提交
935
	if (drop_end)
936
		*drop_end = found ? min(end, extent_end) : end;
J
Josef Bacik 已提交
937 938 939 940 941 942
	btrfs_release_path(path);
	return ret;
}

int btrfs_drop_extents(struct btrfs_trans_handle *trans,
		       struct btrfs_root *root, struct inode *inode, u64 start,
943
		       u64 end, int drop_cache)
J
Josef Bacik 已提交
944 945 946 947 948 949 950
{
	struct btrfs_path *path;
	int ret;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
J
Josef Bacik 已提交
951
	ret = __btrfs_drop_extents(trans, root, inode, path, start, end, NULL,
952
				   drop_cache);
Y
Yan, Zheng 已提交
953
	btrfs_free_path(path);
C
Chris Mason 已提交
954 955 956
	return ret;
}

Y
Yan Zheng 已提交
957
static int extent_mergeable(struct extent_buffer *leaf, int slot,
958 959
			    u64 objectid, u64 bytenr, u64 orig_offset,
			    u64 *start, u64 *end)
Y
Yan Zheng 已提交
960 961 962 963 964 965 966 967 968 969 970 971 972 973 974
{
	struct btrfs_file_extent_item *fi;
	struct btrfs_key key;
	u64 extent_end;

	if (slot < 0 || slot >= btrfs_header_nritems(leaf))
		return 0;

	btrfs_item_key_to_cpu(leaf, &key, slot);
	if (key.objectid != objectid || key.type != BTRFS_EXTENT_DATA_KEY)
		return 0;

	fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
	if (btrfs_file_extent_type(leaf, fi) != BTRFS_FILE_EXTENT_REG ||
	    btrfs_file_extent_disk_bytenr(leaf, fi) != bytenr ||
975
	    btrfs_file_extent_offset(leaf, fi) != key.offset - orig_offset ||
Y
Yan Zheng 已提交
976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999
	    btrfs_file_extent_compression(leaf, fi) ||
	    btrfs_file_extent_encryption(leaf, fi) ||
	    btrfs_file_extent_other_encoding(leaf, fi))
		return 0;

	extent_end = key.offset + btrfs_file_extent_num_bytes(leaf, fi);
	if ((*start && *start != key.offset) || (*end && *end != extent_end))
		return 0;

	*start = key.offset;
	*end = extent_end;
	return 1;
}

/*
 * Mark extent in the range start - end as written.
 *
 * This changes extent type from 'pre-allocated' to 'regular'. If only
 * part of extent is marked as written, the extent will be split into
 * two or three.
 */
int btrfs_mark_extent_written(struct btrfs_trans_handle *trans,
			      struct inode *inode, u64 start, u64 end)
{
Y
Yan, Zheng 已提交
1000
	struct btrfs_root *root = BTRFS_I(inode)->root;
Y
Yan Zheng 已提交
1001 1002 1003 1004
	struct extent_buffer *leaf;
	struct btrfs_path *path;
	struct btrfs_file_extent_item *fi;
	struct btrfs_key key;
Y
Yan, Zheng 已提交
1005
	struct btrfs_key new_key;
Y
Yan Zheng 已提交
1006 1007 1008
	u64 bytenr;
	u64 num_bytes;
	u64 extent_end;
1009
	u64 orig_offset;
Y
Yan Zheng 已提交
1010 1011
	u64 other_start;
	u64 other_end;
Y
Yan, Zheng 已提交
1012 1013 1014
	u64 split;
	int del_nr = 0;
	int del_slot = 0;
1015
	int recow;
Y
Yan Zheng 已提交
1016
	int ret;
L
Li Zefan 已提交
1017
	u64 ino = btrfs_ino(inode);
Y
Yan Zheng 已提交
1018 1019

	path = btrfs_alloc_path();
1020 1021
	if (!path)
		return -ENOMEM;
Y
Yan Zheng 已提交
1022
again:
1023
	recow = 0;
Y
Yan, Zheng 已提交
1024
	split = start;
L
Li Zefan 已提交
1025
	key.objectid = ino;
Y
Yan Zheng 已提交
1026
	key.type = BTRFS_EXTENT_DATA_KEY;
Y
Yan, Zheng 已提交
1027
	key.offset = split;
Y
Yan Zheng 已提交
1028 1029

	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1030 1031
	if (ret < 0)
		goto out;
Y
Yan Zheng 已提交
1032 1033 1034 1035 1036
	if (ret > 0 && path->slots[0] > 0)
		path->slots[0]--;

	leaf = path->nodes[0];
	btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
L
Li Zefan 已提交
1037
	BUG_ON(key.objectid != ino || key.type != BTRFS_EXTENT_DATA_KEY);
Y
Yan Zheng 已提交
1038 1039
	fi = btrfs_item_ptr(leaf, path->slots[0],
			    struct btrfs_file_extent_item);
Y
Yan, Zheng 已提交
1040 1041
	BUG_ON(btrfs_file_extent_type(leaf, fi) !=
	       BTRFS_FILE_EXTENT_PREALLOC);
Y
Yan Zheng 已提交
1042 1043 1044 1045 1046
	extent_end = key.offset + btrfs_file_extent_num_bytes(leaf, fi);
	BUG_ON(key.offset > start || extent_end < end);

	bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
	num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
1047
	orig_offset = key.offset - btrfs_file_extent_offset(leaf, fi);
1048 1049 1050 1051 1052 1053
	memcpy(&new_key, &key, sizeof(new_key));

	if (start == key.offset && end < extent_end) {
		other_start = 0;
		other_end = start;
		if (extent_mergeable(leaf, path->slots[0] - 1,
L
Li Zefan 已提交
1054
				     ino, bytenr, orig_offset,
1055 1056
				     &other_start, &other_end)) {
			new_key.offset = end;
1057
			btrfs_set_item_key_safe(root, path, &new_key);
1058 1059
			fi = btrfs_item_ptr(leaf, path->slots[0],
					    struct btrfs_file_extent_item);
1060 1061
			btrfs_set_file_extent_generation(leaf, fi,
							 trans->transid);
1062 1063 1064 1065 1066 1067
			btrfs_set_file_extent_num_bytes(leaf, fi,
							extent_end - end);
			btrfs_set_file_extent_offset(leaf, fi,
						     end - orig_offset);
			fi = btrfs_item_ptr(leaf, path->slots[0] - 1,
					    struct btrfs_file_extent_item);
1068 1069
			btrfs_set_file_extent_generation(leaf, fi,
							 trans->transid);
1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080
			btrfs_set_file_extent_num_bytes(leaf, fi,
							end - other_start);
			btrfs_mark_buffer_dirty(leaf);
			goto out;
		}
	}

	if (start > key.offset && end == extent_end) {
		other_start = end;
		other_end = 0;
		if (extent_mergeable(leaf, path->slots[0] + 1,
L
Li Zefan 已提交
1081
				     ino, bytenr, orig_offset,
1082 1083 1084 1085 1086
				     &other_start, &other_end)) {
			fi = btrfs_item_ptr(leaf, path->slots[0],
					    struct btrfs_file_extent_item);
			btrfs_set_file_extent_num_bytes(leaf, fi,
							start - key.offset);
1087 1088
			btrfs_set_file_extent_generation(leaf, fi,
							 trans->transid);
1089 1090
			path->slots[0]++;
			new_key.offset = start;
1091
			btrfs_set_item_key_safe(root, path, &new_key);
1092 1093 1094

			fi = btrfs_item_ptr(leaf, path->slots[0],
					    struct btrfs_file_extent_item);
1095 1096
			btrfs_set_file_extent_generation(leaf, fi,
							 trans->transid);
1097 1098 1099 1100 1101 1102 1103 1104
			btrfs_set_file_extent_num_bytes(leaf, fi,
							other_end - start);
			btrfs_set_file_extent_offset(leaf, fi,
						     start - orig_offset);
			btrfs_mark_buffer_dirty(leaf);
			goto out;
		}
	}
Y
Yan Zheng 已提交
1105

Y
Yan, Zheng 已提交
1106 1107 1108 1109 1110 1111 1112
	while (start > key.offset || end < extent_end) {
		if (key.offset == start)
			split = end;

		new_key.offset = split;
		ret = btrfs_duplicate_item(trans, root, path, &new_key);
		if (ret == -EAGAIN) {
1113
			btrfs_release_path(path);
Y
Yan, Zheng 已提交
1114
			goto again;
Y
Yan Zheng 已提交
1115
		}
1116 1117 1118 1119
		if (ret < 0) {
			btrfs_abort_transaction(trans, root, ret);
			goto out;
		}
Y
Yan Zheng 已提交
1120

Y
Yan, Zheng 已提交
1121 1122
		leaf = path->nodes[0];
		fi = btrfs_item_ptr(leaf, path->slots[0] - 1,
Y
Yan Zheng 已提交
1123
				    struct btrfs_file_extent_item);
1124
		btrfs_set_file_extent_generation(leaf, fi, trans->transid);
Y
Yan Zheng 已提交
1125
		btrfs_set_file_extent_num_bytes(leaf, fi,
Y
Yan, Zheng 已提交
1126 1127 1128 1129 1130
						split - key.offset);

		fi = btrfs_item_ptr(leaf, path->slots[0],
				    struct btrfs_file_extent_item);

1131
		btrfs_set_file_extent_generation(leaf, fi, trans->transid);
Y
Yan, Zheng 已提交
1132 1133 1134
		btrfs_set_file_extent_offset(leaf, fi, split - orig_offset);
		btrfs_set_file_extent_num_bytes(leaf, fi,
						extent_end - split);
Y
Yan Zheng 已提交
1135 1136
		btrfs_mark_buffer_dirty(leaf);

Y
Yan, Zheng 已提交
1137 1138
		ret = btrfs_inc_extent_ref(trans, root, bytenr, num_bytes, 0,
					   root->root_key.objectid,
A
Arne Jansen 已提交
1139
					   ino, orig_offset, 0);
1140
		BUG_ON(ret); /* -ENOMEM */
Y
Yan Zheng 已提交
1141

Y
Yan, Zheng 已提交
1142 1143 1144 1145
		if (split == start) {
			key.offset = start;
		} else {
			BUG_ON(start != key.offset);
Y
Yan Zheng 已提交
1146
			path->slots[0]--;
Y
Yan, Zheng 已提交
1147
			extent_end = end;
Y
Yan Zheng 已提交
1148
		}
1149
		recow = 1;
Y
Yan Zheng 已提交
1150 1151
	}

Y
Yan, Zheng 已提交
1152 1153
	other_start = end;
	other_end = 0;
1154
	if (extent_mergeable(leaf, path->slots[0] + 1,
L
Li Zefan 已提交
1155
			     ino, bytenr, orig_offset,
1156 1157
			     &other_start, &other_end)) {
		if (recow) {
1158
			btrfs_release_path(path);
1159 1160
			goto again;
		}
Y
Yan, Zheng 已提交
1161 1162 1163 1164 1165
		extent_end = other_end;
		del_slot = path->slots[0] + 1;
		del_nr++;
		ret = btrfs_free_extent(trans, root, bytenr, num_bytes,
					0, root->root_key.objectid,
A
Arne Jansen 已提交
1166
					ino, orig_offset, 0);
1167
		BUG_ON(ret); /* -ENOMEM */
Y
Yan Zheng 已提交
1168
	}
Y
Yan, Zheng 已提交
1169 1170
	other_start = 0;
	other_end = start;
1171
	if (extent_mergeable(leaf, path->slots[0] - 1,
L
Li Zefan 已提交
1172
			     ino, bytenr, orig_offset,
1173 1174
			     &other_start, &other_end)) {
		if (recow) {
1175
			btrfs_release_path(path);
1176 1177
			goto again;
		}
Y
Yan, Zheng 已提交
1178 1179 1180 1181 1182
		key.offset = other_start;
		del_slot = path->slots[0];
		del_nr++;
		ret = btrfs_free_extent(trans, root, bytenr, num_bytes,
					0, root->root_key.objectid,
A
Arne Jansen 已提交
1183
					ino, orig_offset, 0);
1184
		BUG_ON(ret); /* -ENOMEM */
Y
Yan, Zheng 已提交
1185 1186
	}
	if (del_nr == 0) {
1187 1188
		fi = btrfs_item_ptr(leaf, path->slots[0],
			   struct btrfs_file_extent_item);
Y
Yan, Zheng 已提交
1189 1190
		btrfs_set_file_extent_type(leaf, fi,
					   BTRFS_FILE_EXTENT_REG);
1191
		btrfs_set_file_extent_generation(leaf, fi, trans->transid);
Y
Yan, Zheng 已提交
1192
		btrfs_mark_buffer_dirty(leaf);
1193
	} else {
1194 1195
		fi = btrfs_item_ptr(leaf, del_slot - 1,
			   struct btrfs_file_extent_item);
1196 1197
		btrfs_set_file_extent_type(leaf, fi,
					   BTRFS_FILE_EXTENT_REG);
1198
		btrfs_set_file_extent_generation(leaf, fi, trans->transid);
1199 1200 1201
		btrfs_set_file_extent_num_bytes(leaf, fi,
						extent_end - key.offset);
		btrfs_mark_buffer_dirty(leaf);
Y
Yan, Zheng 已提交
1202

1203
		ret = btrfs_del_items(trans, root, path, del_slot, del_nr);
1204 1205 1206 1207
		if (ret < 0) {
			btrfs_abort_transaction(trans, root, ret);
			goto out;
		}
1208
	}
Y
Yan, Zheng 已提交
1209
out:
Y
Yan Zheng 已提交
1210 1211 1212 1213
	btrfs_free_path(path);
	return 0;
}

1214 1215 1216 1217
/*
 * on error we return an unlocked page and the error value
 * on success we return a locked page and 0
 */
1218 1219
static int prepare_uptodate_page(struct page *page, u64 pos,
				 bool force_uptodate)
1220 1221 1222
{
	int ret = 0;

1223 1224
	if (((pos & (PAGE_CACHE_SIZE - 1)) || force_uptodate) &&
	    !PageUptodate(page)) {
1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236
		ret = btrfs_readpage(NULL, page);
		if (ret)
			return ret;
		lock_page(page);
		if (!PageUptodate(page)) {
			unlock_page(page);
			return -EIO;
		}
	}
	return 0;
}

C
Chris Mason 已提交
1237
/*
C
Chris Mason 已提交
1238 1239 1240
 * this gets pages into the page cache and locks them down, it also properly
 * waits for data=ordered extents to finish before allowing the pages to be
 * modified.
C
Chris Mason 已提交
1241
 */
1242 1243 1244
static noinline int prepare_pages(struct inode *inode, struct page **pages,
				  size_t num_pages, loff_t pos,
				  size_t write_bytes, bool force_uptodate)
C
Chris Mason 已提交
1245
{
1246
	struct extent_state *cached_state = NULL;
C
Chris Mason 已提交
1247 1248
	int i;
	unsigned long index = pos >> PAGE_CACHE_SHIFT;
1249
	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
C
Chris Mason 已提交
1250
	int err = 0;
1251
	int faili = 0;
1252
	u64 start_pos;
1253
	u64 last_pos;
1254

1255
	start_pos = pos & ~((u64)PAGE_CACHE_SIZE - 1);
1256
	last_pos = ((u64)index + num_pages) << PAGE_CACHE_SHIFT;
C
Chris Mason 已提交
1257

1258
again:
C
Chris Mason 已提交
1259
	for (i = 0; i < num_pages; i++) {
1260
		pages[i] = find_or_create_page(inode->i_mapping, index + i,
1261
					       mask | __GFP_WRITE);
C
Chris Mason 已提交
1262
		if (!pages[i]) {
1263 1264 1265 1266 1267 1268
			faili = i - 1;
			err = -ENOMEM;
			goto fail;
		}

		if (i == 0)
1269 1270
			err = prepare_uptodate_page(pages[i], pos,
						    force_uptodate);
1271 1272
		if (i == num_pages - 1)
			err = prepare_uptodate_page(pages[i],
1273
						    pos + write_bytes, false);
1274 1275 1276 1277
		if (err) {
			page_cache_release(pages[i]);
			faili = i - 1;
			goto fail;
C
Chris Mason 已提交
1278
		}
C
Chris Mason 已提交
1279
		wait_on_page_writeback(pages[i]);
C
Chris Mason 已提交
1280
	}
1281
	faili = num_pages - 1;
1282
	err = 0;
1283
	if (start_pos < inode->i_size) {
1284
		struct btrfs_ordered_extent *ordered;
1285
		lock_extent_bits(&BTRFS_I(inode)->io_tree,
1286
				 start_pos, last_pos - 1, 0, &cached_state);
C
Chris Mason 已提交
1287 1288
		ordered = btrfs_lookup_first_ordered_extent(inode,
							    last_pos - 1);
1289 1290 1291 1292
		if (ordered &&
		    ordered->file_offset + ordered->len > start_pos &&
		    ordered->file_offset < last_pos) {
			btrfs_put_ordered_extent(ordered);
1293 1294 1295
			unlock_extent_cached(&BTRFS_I(inode)->io_tree,
					     start_pos, last_pos - 1,
					     &cached_state, GFP_NOFS);
1296 1297 1298 1299
			for (i = 0; i < num_pages; i++) {
				unlock_page(pages[i]);
				page_cache_release(pages[i]);
			}
1300 1301 1302 1303
			err = btrfs_wait_ordered_range(inode, start_pos,
						       last_pos - start_pos);
			if (err)
				goto fail;
1304 1305 1306 1307 1308
			goto again;
		}
		if (ordered)
			btrfs_put_ordered_extent(ordered);

1309
		clear_extent_bit(&BTRFS_I(inode)->io_tree, start_pos,
1310
				  last_pos - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1311 1312
				  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
				  0, 0, &cached_state, GFP_NOFS);
1313 1314 1315
		unlock_extent_cached(&BTRFS_I(inode)->io_tree,
				     start_pos, last_pos - 1, &cached_state,
				     GFP_NOFS);
1316
	}
1317
	for (i = 0; i < num_pages; i++) {
1318 1319
		if (clear_page_dirty_for_io(pages[i]))
			account_page_redirty(pages[i]);
1320 1321 1322
		set_page_extent_mapped(pages[i]);
		WARN_ON(!PageLocked(pages[i]));
	}
C
Chris Mason 已提交
1323
	return 0;
1324 1325 1326 1327 1328 1329 1330 1331
fail:
	while (faili >= 0) {
		unlock_page(pages[faili]);
		page_cache_release(pages[faili]);
		faili--;
	}
	return err;

C
Chris Mason 已提交
1332 1333
}

1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358
static noinline int check_can_nocow(struct inode *inode, loff_t pos,
				    size_t *write_bytes)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_ordered_extent *ordered;
	u64 lockstart, lockend;
	u64 num_bytes;
	int ret;

	lockstart = round_down(pos, root->sectorsize);
	lockend = lockstart + round_up(*write_bytes, root->sectorsize) - 1;

	while (1) {
		lock_extent(&BTRFS_I(inode)->io_tree, lockstart, lockend);
		ordered = btrfs_lookup_ordered_range(inode, lockstart,
						     lockend - lockstart + 1);
		if (!ordered) {
			break;
		}
		unlock_extent(&BTRFS_I(inode)->io_tree, lockstart, lockend);
		btrfs_start_ordered_extent(inode, ordered, 1);
		btrfs_put_ordered_extent(ordered);
	}

	num_bytes = lockend - lockstart + 1;
1359
	ret = can_nocow_extent(inode, lockstart, &num_bytes, NULL, NULL, NULL);
1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374
	if (ret <= 0) {
		ret = 0;
	} else {
		clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, lockend,
				 EXTENT_DIRTY | EXTENT_DELALLOC |
				 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
				 NULL, GFP_NOFS);
		*write_bytes = min_t(size_t, *write_bytes, num_bytes);
	}

	unlock_extent(&BTRFS_I(inode)->io_tree, lockstart, lockend);

	return ret;
}

J
Josef Bacik 已提交
1375 1376 1377
static noinline ssize_t __btrfs_buffered_write(struct file *file,
					       struct iov_iter *i,
					       loff_t pos)
1378
{
A
Al Viro 已提交
1379
	struct inode *inode = file_inode(file);
1380 1381
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct page **pages = NULL;
1382
	u64 release_bytes = 0;
C
Chris Mason 已提交
1383
	unsigned long first_index;
J
Josef Bacik 已提交
1384 1385
	size_t num_written = 0;
	int nrptrs;
1386
	int ret = 0;
1387
	bool only_release_metadata = false;
1388
	bool force_page_uptodate = false;
1389

J
Josef Bacik 已提交
1390
	nrptrs = min((iov_iter_count(i) + PAGE_CACHE_SIZE - 1) /
1391 1392
		     PAGE_CACHE_SIZE, PAGE_CACHE_SIZE /
		     (sizeof(struct page *)));
1393 1394
	nrptrs = min(nrptrs, current->nr_dirtied_pause - current->nr_dirtied);
	nrptrs = max(nrptrs, 8);
1395
	pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL);
J
Josef Bacik 已提交
1396 1397
	if (!pages)
		return -ENOMEM;
1398

C
Chris Mason 已提交
1399 1400
	first_index = pos >> PAGE_CACHE_SHIFT;

J
Josef Bacik 已提交
1401
	while (iov_iter_count(i) > 0) {
C
Chris Mason 已提交
1402
		size_t offset = pos & (PAGE_CACHE_SIZE - 1);
J
Josef Bacik 已提交
1403
		size_t write_bytes = min(iov_iter_count(i),
1404
					 nrptrs * (size_t)PAGE_CACHE_SIZE -
1405
					 offset);
Y
Yan, Zheng 已提交
1406 1407
		size_t num_pages = (write_bytes + offset +
				    PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1408
		size_t reserve_bytes;
J
Josef Bacik 已提交
1409 1410
		size_t dirty_pages;
		size_t copied;
C
Chris Mason 已提交
1411

1412
		WARN_ON(num_pages > nrptrs);
1413

1414 1415 1416 1417
		/*
		 * Fault pages before locking them in prepare_pages
		 * to avoid recursive lock
		 */
J
Josef Bacik 已提交
1418
		if (unlikely(iov_iter_fault_in_readable(i, write_bytes))) {
1419
			ret = -EFAULT;
J
Josef Bacik 已提交
1420
			break;
1421 1422
		}

1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444
		reserve_bytes = num_pages << PAGE_CACHE_SHIFT;
		ret = btrfs_check_data_free_space(inode, reserve_bytes);
		if (ret == -ENOSPC &&
		    (BTRFS_I(inode)->flags & (BTRFS_INODE_NODATACOW |
					      BTRFS_INODE_PREALLOC))) {
			ret = check_can_nocow(inode, pos, &write_bytes);
			if (ret > 0) {
				only_release_metadata = true;
				/*
				 * our prealloc extent may be smaller than
				 * write_bytes, so scale down.
				 */
				num_pages = (write_bytes + offset +
					     PAGE_CACHE_SIZE - 1) >>
					PAGE_CACHE_SHIFT;
				reserve_bytes = num_pages << PAGE_CACHE_SHIFT;
				ret = 0;
			} else {
				ret = -ENOSPC;
			}
		}

1445
		if (ret)
J
Josef Bacik 已提交
1446
			break;
1447

1448 1449 1450 1451 1452 1453 1454 1455 1456 1457
		ret = btrfs_delalloc_reserve_metadata(inode, reserve_bytes);
		if (ret) {
			if (!only_release_metadata)
				btrfs_free_reserved_data_space(inode,
							       reserve_bytes);
			break;
		}

		release_bytes = reserve_bytes;

1458 1459 1460 1461 1462
		/*
		 * This is going to setup the pages array with the number of
		 * pages we want, so we don't really need to worry about the
		 * contents of pages from loop to loop
		 */
1463 1464
		ret = prepare_pages(inode, pages, num_pages,
				    pos, write_bytes,
1465
				    force_page_uptodate);
1466
		if (ret)
J
Josef Bacik 已提交
1467
			break;
C
Chris Mason 已提交
1468

1469
		copied = btrfs_copy_from_user(pos, num_pages,
J
Josef Bacik 已提交
1470
					   write_bytes, pages, i);
1471 1472 1473 1474 1475 1476 1477 1478

		/*
		 * if we have trouble faulting in the pages, fall
		 * back to one page at a time
		 */
		if (copied < write_bytes)
			nrptrs = 1;

1479 1480
		if (copied == 0) {
			force_page_uptodate = true;
1481
			dirty_pages = 0;
1482 1483
		} else {
			force_page_uptodate = false;
1484 1485 1486
			dirty_pages = (copied + offset +
				       PAGE_CACHE_SIZE - 1) >>
				       PAGE_CACHE_SHIFT;
1487
		}
1488

J
Josef Bacik 已提交
1489 1490 1491 1492 1493 1494 1495
		/*
		 * If we had a short copy we need to release the excess delaloc
		 * bytes we reserved.  We need to increment outstanding_extents
		 * because btrfs_delalloc_release_space will decrement it, but
		 * we still have an outstanding extent for the chunk we actually
		 * managed to copy.
		 */
1496
		if (num_pages > dirty_pages) {
1497 1498
			release_bytes = (num_pages - dirty_pages) <<
				PAGE_CACHE_SHIFT;
1499 1500 1501 1502 1503
			if (copied > 0) {
				spin_lock(&BTRFS_I(inode)->lock);
				BTRFS_I(inode)->outstanding_extents++;
				spin_unlock(&BTRFS_I(inode)->lock);
			}
1504 1505 1506 1507 1508 1509
			if (only_release_metadata)
				btrfs_delalloc_release_metadata(inode,
								release_bytes);
			else
				btrfs_delalloc_release_space(inode,
							     release_bytes);
1510 1511
		}

1512
		release_bytes = dirty_pages << PAGE_CACHE_SHIFT;
1513
		if (copied > 0) {
1514 1515 1516
			ret = btrfs_dirty_pages(root, inode, pages,
						dirty_pages, pos, copied,
						NULL);
J
Josef Bacik 已提交
1517 1518 1519 1520
			if (ret) {
				btrfs_drop_pages(pages, num_pages);
				break;
			}
1521
		}
C
Chris Mason 已提交
1522

1523
		release_bytes = 0;
C
Chris Mason 已提交
1524 1525
		btrfs_drop_pages(pages, num_pages);

1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536
		if (only_release_metadata && copied > 0) {
			u64 lockstart = round_down(pos, root->sectorsize);
			u64 lockend = lockstart +
				(dirty_pages << PAGE_CACHE_SHIFT) - 1;

			set_extent_bit(&BTRFS_I(inode)->io_tree, lockstart,
				       lockend, EXTENT_NORESERVE, NULL,
				       NULL, GFP_NOFS);
			only_release_metadata = false;
		}

J
Josef Bacik 已提交
1537 1538
		cond_resched();

1539
		balance_dirty_pages_ratelimited(inode->i_mapping);
J
Josef Bacik 已提交
1540
		if (dirty_pages < (root->leafsize >> PAGE_CACHE_SHIFT) + 1)
1541
			btrfs_btree_balance_dirty(root);
1542

1543 1544
		pos += copied;
		num_written += copied;
J
Josef Bacik 已提交
1545
	}
C
Chris Mason 已提交
1546

J
Josef Bacik 已提交
1547 1548
	kfree(pages);

1549 1550 1551 1552 1553 1554 1555
	if (release_bytes) {
		if (only_release_metadata)
			btrfs_delalloc_release_metadata(inode, release_bytes);
		else
			btrfs_delalloc_release_space(inode, release_bytes);
	}

J
Josef Bacik 已提交
1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583
	return num_written ? num_written : ret;
}

static ssize_t __btrfs_direct_write(struct kiocb *iocb,
				    const struct iovec *iov,
				    unsigned long nr_segs, loff_t pos,
				    loff_t *ppos, size_t count, size_t ocount)
{
	struct file *file = iocb->ki_filp;
	struct iov_iter i;
	ssize_t written;
	ssize_t written_buffered;
	loff_t endbyte;
	int err;

	written = generic_file_direct_write(iocb, iov, &nr_segs, pos, ppos,
					    count, ocount);

	if (written < 0 || written == count)
		return written;

	pos += written;
	count -= written;
	iov_iter_init(&i, iov, nr_segs, count, written);
	written_buffered = __btrfs_buffered_write(file, &i, pos);
	if (written_buffered < 0) {
		err = written_buffered;
		goto out;
C
Chris Mason 已提交
1584
	}
J
Josef Bacik 已提交
1585 1586 1587 1588 1589 1590 1591 1592
	endbyte = pos + written_buffered - 1;
	err = filemap_write_and_wait_range(file->f_mapping, pos, endbyte);
	if (err)
		goto out;
	written += written_buffered;
	*ppos = pos + written_buffered;
	invalidate_mapping_pages(file->f_mapping, pos >> PAGE_CACHE_SHIFT,
				 endbyte >> PAGE_CACHE_SHIFT);
C
Chris Mason 已提交
1593
out:
J
Josef Bacik 已提交
1594 1595
	return written ? written : err;
}
1596

1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614
static void update_time_for_write(struct inode *inode)
{
	struct timespec now;

	if (IS_NOCMTIME(inode))
		return;

	now = current_fs_time(inode->i_sb);
	if (!timespec_equal(&inode->i_mtime, &now))
		inode->i_mtime = now;

	if (!timespec_equal(&inode->i_ctime, &now))
		inode->i_ctime = now;

	if (IS_I_VERSION(inode))
		inode_inc_iversion(inode);
}

J
Josef Bacik 已提交
1615 1616 1617 1618 1619
static ssize_t btrfs_file_aio_write(struct kiocb *iocb,
				    const struct iovec *iov,
				    unsigned long nr_segs, loff_t pos)
{
	struct file *file = iocb->ki_filp;
A
Al Viro 已提交
1620
	struct inode *inode = file_inode(file);
J
Josef Bacik 已提交
1621 1622
	struct btrfs_root *root = BTRFS_I(inode)->root;
	loff_t *ppos = &iocb->ki_pos;
1623
	u64 start_pos;
J
Josef Bacik 已提交
1624 1625 1626
	ssize_t num_written = 0;
	ssize_t err = 0;
	size_t count, ocount;
1627
	bool sync = (file->f_flags & O_DSYNC) || IS_SYNC(file->f_mapping->host);
J
Josef Bacik 已提交
1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661

	mutex_lock(&inode->i_mutex);

	err = generic_segment_checks(iov, &nr_segs, &ocount, VERIFY_READ);
	if (err) {
		mutex_unlock(&inode->i_mutex);
		goto out;
	}
	count = ocount;

	current->backing_dev_info = inode->i_mapping->backing_dev_info;
	err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
	if (err) {
		mutex_unlock(&inode->i_mutex);
		goto out;
	}

	if (count == 0) {
		mutex_unlock(&inode->i_mutex);
		goto out;
	}

	err = file_remove_suid(file);
	if (err) {
		mutex_unlock(&inode->i_mutex);
		goto out;
	}

	/*
	 * If BTRFS flips readonly due to some impossible error
	 * (fs_info->fs_state now has BTRFS_SUPER_FLAG_ERROR),
	 * although we have opened a file as writable, we have
	 * to stop this write operation to ensure FS consistency.
	 */
1662
	if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
J
Josef Bacik 已提交
1663 1664 1665 1666 1667
		mutex_unlock(&inode->i_mutex);
		err = -EROFS;
		goto out;
	}

1668 1669 1670 1671 1672 1673 1674
	/*
	 * We reserve space for updating the inode when we reserve space for the
	 * extent we are going to write, so we will enospc out there.  We don't
	 * need to start yet another transaction to update the inode as we will
	 * update the inode when we finish writing whatever data we write.
	 */
	update_time_for_write(inode);
J
Josef Bacik 已提交
1675

1676 1677 1678 1679 1680 1681 1682 1683 1684
	start_pos = round_down(pos, root->sectorsize);
	if (start_pos > i_size_read(inode)) {
		err = btrfs_cont_expand(inode, i_size_read(inode), start_pos);
		if (err) {
			mutex_unlock(&inode->i_mutex);
			goto out;
		}
	}

1685 1686 1687
	if (sync)
		atomic_inc(&BTRFS_I(inode)->sync_writers);

J
Josef Bacik 已提交
1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701
	if (unlikely(file->f_flags & O_DIRECT)) {
		num_written = __btrfs_direct_write(iocb, iov, nr_segs,
						   pos, ppos, count, ocount);
	} else {
		struct iov_iter i;

		iov_iter_init(&i, iov, nr_segs, count, num_written);

		num_written = __btrfs_buffered_write(file, &i, pos);
		if (num_written > 0)
			*ppos = pos + num_written;
	}

	mutex_unlock(&inode->i_mutex);
1702

1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713
	/*
	 * we want to make sure fsync finds this change
	 * but we haven't joined a transaction running right now.
	 *
	 * Later on, someone is sure to update the inode and get the
	 * real transid recorded.
	 *
	 * We set last_trans now to the fs_info generation + 1,
	 * this will either be one more than the running transaction
	 * or the generation used for the next transaction if there isn't
	 * one running right now.
1714 1715 1716 1717
	 *
	 * We also have to set last_sub_trans to the current log transid,
	 * otherwise subsequent syncs to a file that's been synced in this
	 * transaction will appear to have already occured.
1718 1719
	 */
	BTRFS_I(inode)->last_trans = root->fs_info->generation + 1;
1720
	BTRFS_I(inode)->last_sub_trans = root->log_transid;
1721
	if (num_written > 0) {
J
Josef Bacik 已提交
1722 1723
		err = generic_write_sync(file, pos, num_written);
		if (err < 0 && num_written > 0)
1724 1725
			num_written = err;
	}
1726

1727 1728
	if (sync)
		atomic_dec(&BTRFS_I(inode)->sync_writers);
1729
out:
C
Chris Mason 已提交
1730 1731 1732 1733
	current->backing_dev_info = NULL;
	return num_written ? num_written : err;
}

C
Chris Mason 已提交
1734
int btrfs_release_file(struct inode *inode, struct file *filp)
1735
{
1736 1737 1738 1739 1740 1741
	/*
	 * ordered_data_close is set by settattr when we are about to truncate
	 * a file from a non-zero size to a zero size.  This tries to
	 * flush down new bytes that may have been written if the
	 * application were using truncate to replace a file in place.
	 */
1742 1743
	if (test_and_clear_bit(BTRFS_INODE_ORDERED_DATA_CLOSE,
			       &BTRFS_I(inode)->runtime_flags)) {
1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757
		struct btrfs_trans_handle *trans;
		struct btrfs_root *root = BTRFS_I(inode)->root;

		/*
		 * We need to block on a committing transaction to keep us from
		 * throwing a ordered operation on to the list and causing
		 * something like sync to deadlock trying to flush out this
		 * inode.
		 */
		trans = btrfs_start_transaction(root, 0);
		if (IS_ERR(trans))
			return PTR_ERR(trans);
		btrfs_add_ordered_operation(trans, BTRFS_I(inode)->root, inode);
		btrfs_end_transaction(trans, root);
1758 1759 1760
		if (inode->i_size > BTRFS_ORDERED_OPERATIONS_FLUSH_LIMIT)
			filemap_flush(inode->i_mapping);
	}
S
Sage Weil 已提交
1761 1762
	if (filp->private_data)
		btrfs_ioctl_trans_end(filp);
1763 1764 1765
	return 0;
}

C
Chris Mason 已提交
1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776
/*
 * fsync call for both files and directories.  This logs the inode into
 * the tree log instead of forcing full commits whenever possible.
 *
 * It needs to call filemap_fdatawait so that all ordered extent updates are
 * in the metadata btree are up to date for copying to the log.
 *
 * It drops the inode mutex before doing the tree log commit.  This is an
 * important optimization for directories because holding the mutex prevents
 * new operations on the dir while we write to disk.
 */
1777
int btrfs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
C
Chris Mason 已提交
1778
{
1779
	struct dentry *dentry = file->f_path.dentry;
C
Chris Mason 已提交
1780 1781
	struct inode *inode = dentry->d_inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;
1782
	int ret = 0;
C
Chris Mason 已提交
1783
	struct btrfs_trans_handle *trans;
1784
	bool full_sync = 0;
C
Chris Mason 已提交
1785

1786
	trace_btrfs_sync_file(file, datasync);
1787

1788 1789 1790
	/*
	 * We write the dirty pages in the range and wait until they complete
	 * out of the ->i_mutex. If so, we can flush the dirty pages by
1791 1792
	 * multi-task, and make the performance up.  See
	 * btrfs_wait_ordered_range for an explanation of the ASYNC check.
1793
	 */
1794
	atomic_inc(&BTRFS_I(inode)->sync_writers);
1795 1796 1797 1798
	ret = filemap_fdatawrite_range(inode->i_mapping, start, end);
	if (!ret && test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
			     &BTRFS_I(inode)->runtime_flags))
		ret = filemap_fdatawrite_range(inode->i_mapping, start, end);
1799
	atomic_dec(&BTRFS_I(inode)->sync_writers);
1800 1801 1802
	if (ret)
		return ret;

1803 1804
	mutex_lock(&inode->i_mutex);

1805
	/*
1806 1807
	 * We flush the dirty pages again to avoid some dirty pages in the
	 * range being left.
1808
	 */
M
Miao Xie 已提交
1809
	atomic_inc(&root->log_batch);
1810 1811
	full_sync = test_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
			     &BTRFS_I(inode)->runtime_flags);
1812 1813 1814 1815 1816 1817 1818
	if (full_sync) {
		ret = btrfs_wait_ordered_range(inode, start, end - start + 1);
		if (ret) {
			mutex_unlock(&inode->i_mutex);
			goto out;
		}
	}
M
Miao Xie 已提交
1819
	atomic_inc(&root->log_batch);
1820

C
Chris Mason 已提交
1821
	/*
1822 1823
	 * check the transaction that last modified this inode
	 * and see if its already been committed
C
Chris Mason 已提交
1824
	 */
1825 1826
	if (!BTRFS_I(inode)->last_trans) {
		mutex_unlock(&inode->i_mutex);
1827
		goto out;
1828
	}
1829

1830 1831 1832 1833 1834
	/*
	 * if the last transaction that changed this file was before
	 * the current transaction, we can bail out now without any
	 * syncing
	 */
J
Josef Bacik 已提交
1835
	smp_mb();
1836 1837
	if (btrfs_inode_in_log(inode, root->fs_info->generation) ||
	    BTRFS_I(inode)->last_trans <=
1838 1839
	    root->fs_info->last_trans_committed) {
		BTRFS_I(inode)->last_trans = 0;
J
Josef Bacik 已提交
1840 1841 1842 1843 1844 1845 1846 1847

		/*
		 * We'v had everything committed since the last time we were
		 * modified so clear this flag in case it was set for whatever
		 * reason, it's no longer relevant.
		 */
		clear_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
			  &BTRFS_I(inode)->runtime_flags);
1848
		mutex_unlock(&inode->i_mutex);
1849 1850 1851 1852
		goto out;
	}

	/*
1853 1854
	 * ok we haven't committed the transaction yet, lets do a commit
	 */
1855
	if (file->private_data)
S
Sage Weil 已提交
1856 1857
		btrfs_ioctl_trans_end(file);

1858 1859 1860
	trans = btrfs_start_transaction(root, 0);
	if (IS_ERR(trans)) {
		ret = PTR_ERR(trans);
1861
		mutex_unlock(&inode->i_mutex);
C
Chris Mason 已提交
1862 1863
		goto out;
	}
1864

1865
	ret = btrfs_log_dentry_safe(trans, root, dentry);
1866
	if (ret < 0) {
1867 1868
		/* Fallthrough and commit/free transaction. */
		ret = 1;
1869
	}
C
Chris Mason 已提交
1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880

	/* we've logged all the items and now have a consistent
	 * version of the file in the log.  It is possible that
	 * someone will come in and modify the file, but that's
	 * fine because the log is consistent on disk, and we
	 * have references to all of the file's extents
	 *
	 * It is possible that someone will come in and log the
	 * file again, but that will end up using the synchronization
	 * inside btrfs_sync_log to keep things safe.
	 */
1881
	mutex_unlock(&inode->i_mutex);
C
Chris Mason 已提交
1882

1883
	if (ret != BTRFS_NO_LOG_SYNC) {
1884
		if (!ret) {
1885
			ret = btrfs_sync_log(trans, root);
1886
			if (!ret) {
1887
				ret = btrfs_end_transaction(trans, root);
1888
				goto out;
1889
			}
1890
		}
1891 1892 1893 1894 1895 1896 1897
		if (!full_sync) {
			ret = btrfs_wait_ordered_range(inode, start,
						       end - start + 1);
			if (ret)
				goto out;
		}
		ret = btrfs_commit_transaction(trans, root);
1898 1899
	} else {
		ret = btrfs_end_transaction(trans, root);
1900
	}
C
Chris Mason 已提交
1901
out:
1902
	return ret > 0 ? -EIO : ret;
C
Chris Mason 已提交
1903 1904
}

1905
static const struct vm_operations_struct btrfs_file_vm_ops = {
1906
	.fault		= filemap_fault,
C
Chris Mason 已提交
1907
	.page_mkwrite	= btrfs_page_mkwrite,
1908
	.remap_pages	= generic_file_remap_pages,
C
Chris Mason 已提交
1909 1910 1911 1912
};

static int btrfs_file_mmap(struct file	*filp, struct vm_area_struct *vma)
{
M
Miao Xie 已提交
1913 1914 1915 1916 1917
	struct address_space *mapping = filp->f_mapping;

	if (!mapping->a_ops->readpage)
		return -ENOEXEC;

C
Chris Mason 已提交
1918
	file_accessed(filp);
M
Miao Xie 已提交
1919 1920
	vma->vm_ops = &btrfs_file_vm_ops;

C
Chris Mason 已提交
1921 1922 1923
	return 0;
}

J
Josef Bacik 已提交
1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963
static int hole_mergeable(struct inode *inode, struct extent_buffer *leaf,
			  int slot, u64 start, u64 end)
{
	struct btrfs_file_extent_item *fi;
	struct btrfs_key key;

	if (slot < 0 || slot >= btrfs_header_nritems(leaf))
		return 0;

	btrfs_item_key_to_cpu(leaf, &key, slot);
	if (key.objectid != btrfs_ino(inode) ||
	    key.type != BTRFS_EXTENT_DATA_KEY)
		return 0;

	fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);

	if (btrfs_file_extent_type(leaf, fi) != BTRFS_FILE_EXTENT_REG)
		return 0;

	if (btrfs_file_extent_disk_bytenr(leaf, fi))
		return 0;

	if (key.offset == end)
		return 1;
	if (key.offset + btrfs_file_extent_num_bytes(leaf, fi) == start)
		return 1;
	return 0;
}

static int fill_holes(struct btrfs_trans_handle *trans, struct inode *inode,
		      struct btrfs_path *path, u64 offset, u64 end)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct extent_buffer *leaf;
	struct btrfs_file_extent_item *fi;
	struct extent_map *hole_em;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct btrfs_key key;
	int ret;

1964 1965 1966
	if (btrfs_fs_incompat(root->fs_info, NO_HOLES))
		goto out;

J
Josef Bacik 已提交
1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996
	key.objectid = btrfs_ino(inode);
	key.type = BTRFS_EXTENT_DATA_KEY;
	key.offset = offset;

	ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
	if (ret < 0)
		return ret;
	BUG_ON(!ret);

	leaf = path->nodes[0];
	if (hole_mergeable(inode, leaf, path->slots[0]-1, offset, end)) {
		u64 num_bytes;

		path->slots[0]--;
		fi = btrfs_item_ptr(leaf, path->slots[0],
				    struct btrfs_file_extent_item);
		num_bytes = btrfs_file_extent_num_bytes(leaf, fi) +
			end - offset;
		btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
		btrfs_set_file_extent_ram_bytes(leaf, fi, num_bytes);
		btrfs_set_file_extent_offset(leaf, fi, 0);
		btrfs_mark_buffer_dirty(leaf);
		goto out;
	}

	if (hole_mergeable(inode, leaf, path->slots[0]+1, offset, end)) {
		u64 num_bytes;

		path->slots[0]++;
		key.offset = offset;
1997
		btrfs_set_item_key_safe(root, path, &key);
J
Josef Bacik 已提交
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026
		fi = btrfs_item_ptr(leaf, path->slots[0],
				    struct btrfs_file_extent_item);
		num_bytes = btrfs_file_extent_num_bytes(leaf, fi) + end -
			offset;
		btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
		btrfs_set_file_extent_ram_bytes(leaf, fi, num_bytes);
		btrfs_set_file_extent_offset(leaf, fi, 0);
		btrfs_mark_buffer_dirty(leaf);
		goto out;
	}
	btrfs_release_path(path);

	ret = btrfs_insert_file_extent(trans, root, btrfs_ino(inode), offset,
				       0, 0, end - offset, 0, end - offset,
				       0, 0, 0);
	if (ret)
		return ret;

out:
	btrfs_release_path(path);

	hole_em = alloc_extent_map();
	if (!hole_em) {
		btrfs_drop_extent_cache(inode, offset, end - 1, 0);
		set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
			&BTRFS_I(inode)->runtime_flags);
	} else {
		hole_em->start = offset;
		hole_em->len = end - offset;
J
Josef Bacik 已提交
2027
		hole_em->ram_bytes = hole_em->len;
J
Josef Bacik 已提交
2028 2029 2030 2031
		hole_em->orig_start = offset;

		hole_em->block_start = EXTENT_MAP_HOLE;
		hole_em->block_len = 0;
2032
		hole_em->orig_block_len = 0;
J
Josef Bacik 已提交
2033 2034 2035 2036 2037 2038 2039
		hole_em->bdev = root->fs_info->fs_devices->latest_bdev;
		hole_em->compress_type = BTRFS_COMPRESS_NONE;
		hole_em->generation = trans->transid;

		do {
			btrfs_drop_extent_cache(inode, offset, end - 1, 0);
			write_lock(&em_tree->lock);
J
Josef Bacik 已提交
2040
			ret = add_extent_mapping(em_tree, hole_em, 1);
J
Josef Bacik 已提交
2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058
			write_unlock(&em_tree->lock);
		} while (ret == -EEXIST);
		free_extent_map(hole_em);
		if (ret)
			set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
				&BTRFS_I(inode)->runtime_flags);
	}

	return 0;
}

static int btrfs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct extent_state *cached_state = NULL;
	struct btrfs_path *path;
	struct btrfs_block_rsv *rsv;
	struct btrfs_trans_handle *trans;
M
Miao Xie 已提交
2059 2060 2061
	u64 lockstart = round_up(offset, BTRFS_I(inode)->root->sectorsize);
	u64 lockend = round_down(offset + len,
				 BTRFS_I(inode)->root->sectorsize) - 1;
J
Josef Bacik 已提交
2062 2063 2064 2065 2066
	u64 cur_offset = lockstart;
	u64 min_size = btrfs_calc_trunc_metadata_size(root, 1);
	u64 drop_end;
	int ret = 0;
	int err = 0;
2067
	int rsv_count;
2068 2069
	bool same_page = ((offset >> PAGE_CACHE_SHIFT) ==
			  ((offset + len - 1) >> PAGE_CACHE_SHIFT));
2070
	bool no_holes = btrfs_fs_incompat(root->fs_info, NO_HOLES);
J
Josef Bacik 已提交
2071

2072 2073 2074
	ret = btrfs_wait_ordered_range(inode, offset, len);
	if (ret)
		return ret;
J
Josef Bacik 已提交
2075 2076

	mutex_lock(&inode->i_mutex);
2077 2078 2079 2080
	/*
	 * We needn't truncate any page which is beyond the end of the file
	 * because we are sure there is no data there.
	 */
J
Josef Bacik 已提交
2081 2082 2083 2084 2085
	/*
	 * Only do this if we are in the same page and we aren't doing the
	 * entire page.
	 */
	if (same_page && len < PAGE_CACHE_SIZE) {
2086 2087
		if (offset < round_up(inode->i_size, PAGE_CACHE_SIZE))
			ret = btrfs_truncate_page(inode, offset, len, 0);
J
Josef Bacik 已提交
2088 2089 2090 2091 2092
		mutex_unlock(&inode->i_mutex);
		return ret;
	}

	/* zero back part of the first page */
2093 2094 2095 2096 2097 2098
	if (offset < round_up(inode->i_size, PAGE_CACHE_SIZE)) {
		ret = btrfs_truncate_page(inode, offset, 0, 0);
		if (ret) {
			mutex_unlock(&inode->i_mutex);
			return ret;
		}
J
Josef Bacik 已提交
2099 2100 2101
	}

	/* zero the front end of the last page */
M
Miao Xie 已提交
2102 2103 2104 2105 2106 2107
	if (offset + len < round_up(inode->i_size, PAGE_CACHE_SIZE)) {
		ret = btrfs_truncate_page(inode, offset + len, 0, 1);
		if (ret) {
			mutex_unlock(&inode->i_mutex);
			return ret;
		}
J
Josef Bacik 已提交
2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142
	}

	if (lockend < lockstart) {
		mutex_unlock(&inode->i_mutex);
		return 0;
	}

	while (1) {
		struct btrfs_ordered_extent *ordered;

		truncate_pagecache_range(inode, lockstart, lockend);

		lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend,
				 0, &cached_state);
		ordered = btrfs_lookup_first_ordered_extent(inode, lockend);

		/*
		 * We need to make sure we have no ordered extents in this range
		 * and nobody raced in and read a page in this range, if we did
		 * we need to try again.
		 */
		if ((!ordered ||
		    (ordered->file_offset + ordered->len < lockstart ||
		     ordered->file_offset > lockend)) &&
		     !test_range_bit(&BTRFS_I(inode)->io_tree, lockstart,
				     lockend, EXTENT_UPTODATE, 0,
				     cached_state)) {
			if (ordered)
				btrfs_put_ordered_extent(ordered);
			break;
		}
		if (ordered)
			btrfs_put_ordered_extent(ordered);
		unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart,
				     lockend, &cached_state, GFP_NOFS);
2143 2144 2145 2146 2147 2148
		ret = btrfs_wait_ordered_range(inode, lockstart,
					       lockend - lockstart + 1);
		if (ret) {
			mutex_unlock(&inode->i_mutex);
			return ret;
		}
J
Josef Bacik 已提交
2149 2150 2151 2152 2153 2154 2155 2156
	}

	path = btrfs_alloc_path();
	if (!path) {
		ret = -ENOMEM;
		goto out;
	}

2157
	rsv = btrfs_alloc_block_rsv(root, BTRFS_BLOCK_RSV_TEMP);
J
Josef Bacik 已提交
2158 2159 2160 2161 2162 2163 2164 2165 2166 2167
	if (!rsv) {
		ret = -ENOMEM;
		goto out_free;
	}
	rsv->size = btrfs_calc_trunc_metadata_size(root, 1);
	rsv->failfast = 1;

	/*
	 * 1 - update the inode
	 * 1 - removing the extents in the range
2168
	 * 1 - adding the hole extent if no_holes isn't set
J
Josef Bacik 已提交
2169
	 */
2170 2171
	rsv_count = no_holes ? 2 : 3;
	trans = btrfs_start_transaction(root, rsv_count);
J
Josef Bacik 已提交
2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205
	if (IS_ERR(trans)) {
		err = PTR_ERR(trans);
		goto out_free;
	}

	ret = btrfs_block_rsv_migrate(&root->fs_info->trans_block_rsv, rsv,
				      min_size);
	BUG_ON(ret);
	trans->block_rsv = rsv;

	while (cur_offset < lockend) {
		ret = __btrfs_drop_extents(trans, root, inode, path,
					   cur_offset, lockend + 1,
					   &drop_end, 1);
		if (ret != -ENOSPC)
			break;

		trans->block_rsv = &root->fs_info->trans_block_rsv;

		ret = fill_holes(trans, inode, path, cur_offset, drop_end);
		if (ret) {
			err = ret;
			break;
		}

		cur_offset = drop_end;

		ret = btrfs_update_inode(trans, root, inode);
		if (ret) {
			err = ret;
			break;
		}

		btrfs_end_transaction(trans, root);
2206
		btrfs_btree_balance_dirty(root);
J
Josef Bacik 已提交
2207

2208
		trans = btrfs_start_transaction(root, rsv_count);
J
Josef Bacik 已提交
2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236
		if (IS_ERR(trans)) {
			ret = PTR_ERR(trans);
			trans = NULL;
			break;
		}

		ret = btrfs_block_rsv_migrate(&root->fs_info->trans_block_rsv,
					      rsv, min_size);
		BUG_ON(ret);	/* shouldn't happen */
		trans->block_rsv = rsv;
	}

	if (ret) {
		err = ret;
		goto out_trans;
	}

	trans->block_rsv = &root->fs_info->trans_block_rsv;
	ret = fill_holes(trans, inode, path, cur_offset, drop_end);
	if (ret) {
		err = ret;
		goto out_trans;
	}

out_trans:
	if (!trans)
		goto out_free;

2237 2238 2239
	inode_inc_iversion(inode);
	inode->i_mtime = inode->i_ctime = CURRENT_TIME;

J
Josef Bacik 已提交
2240 2241 2242
	trans->block_rsv = &root->fs_info->trans_block_rsv;
	ret = btrfs_update_inode(trans, root, inode);
	btrfs_end_transaction(trans, root);
2243
	btrfs_btree_balance_dirty(root);
J
Josef Bacik 已提交
2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255
out_free:
	btrfs_free_path(path);
	btrfs_free_block_rsv(root, rsv);
out:
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
			     &cached_state, GFP_NOFS);
	mutex_unlock(&inode->i_mutex);
	if (ret && !err)
		err = ret;
	return err;
}

2256 2257 2258
static long btrfs_fallocate(struct file *file, int mode,
			    loff_t offset, loff_t len)
{
A
Al Viro 已提交
2259
	struct inode *inode = file_inode(file);
2260
	struct extent_state *cached_state = NULL;
2261
	struct btrfs_root *root = BTRFS_I(inode)->root;
2262 2263 2264 2265 2266 2267 2268
	u64 cur_offset;
	u64 last_byte;
	u64 alloc_start;
	u64 alloc_end;
	u64 alloc_hint = 0;
	u64 locked_end;
	struct extent_map *em;
2269
	int blocksize = BTRFS_I(inode)->root->sectorsize;
2270 2271
	int ret;

2272 2273
	alloc_start = round_down(offset, blocksize);
	alloc_end = round_up(offset + len, blocksize);
2274

J
Josef Bacik 已提交
2275 2276
	/* Make sure we aren't being give some crap mode */
	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
2277 2278
		return -EOPNOTSUPP;

J
Josef Bacik 已提交
2279 2280 2281
	if (mode & FALLOC_FL_PUNCH_HOLE)
		return btrfs_punch_hole(inode, offset, len);

2282 2283 2284 2285
	/*
	 * Make sure we have enough space before we do the
	 * allocation.
	 */
2286
	ret = btrfs_check_data_free_space(inode, alloc_end - alloc_start);
2287 2288
	if (ret)
		return ret;
2289 2290 2291 2292 2293
	if (root->fs_info->quota_enabled) {
		ret = btrfs_qgroup_reserve(root, alloc_end - alloc_start);
		if (ret)
			goto out_reserve_fail;
	}
2294

2295 2296 2297 2298 2299 2300
	mutex_lock(&inode->i_mutex);
	ret = inode_newsize_ok(inode, alloc_end);
	if (ret)
		goto out;

	if (alloc_start > inode->i_size) {
2301 2302
		ret = btrfs_cont_expand(inode, i_size_read(inode),
					alloc_start);
2303 2304
		if (ret)
			goto out;
2305 2306 2307 2308 2309 2310 2311 2312 2313
	} else {
		/*
		 * If we are fallocating from the end of the file onward we
		 * need to zero out the end of the page if i_size lands in the
		 * middle of a page.
		 */
		ret = btrfs_truncate_page(inode, inode->i_size, 0, 0);
		if (ret)
			goto out;
2314 2315
	}

2316 2317 2318 2319
	/*
	 * wait for ordered IO before we have any locks.  We'll loop again
	 * below with the locks held.
	 */
2320 2321 2322 2323
	ret = btrfs_wait_ordered_range(inode, alloc_start,
				       alloc_end - alloc_start);
	if (ret)
		goto out;
2324

2325 2326 2327 2328 2329 2330 2331 2332
	locked_end = alloc_end - 1;
	while (1) {
		struct btrfs_ordered_extent *ordered;

		/* the extent lock is ordered inside the running
		 * transaction
		 */
		lock_extent_bits(&BTRFS_I(inode)->io_tree, alloc_start,
2333
				 locked_end, 0, &cached_state);
2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346
		ordered = btrfs_lookup_first_ordered_extent(inode,
							    alloc_end - 1);
		if (ordered &&
		    ordered->file_offset + ordered->len > alloc_start &&
		    ordered->file_offset < alloc_end) {
			btrfs_put_ordered_extent(ordered);
			unlock_extent_cached(&BTRFS_I(inode)->io_tree,
					     alloc_start, locked_end,
					     &cached_state, GFP_NOFS);
			/*
			 * we can't wait on the range with the transaction
			 * running or with the extent lock held
			 */
2347 2348 2349 2350
			ret = btrfs_wait_ordered_range(inode, alloc_start,
						       alloc_end - alloc_start);
			if (ret)
				goto out;
2351 2352 2353 2354 2355 2356 2357 2358 2359
		} else {
			if (ordered)
				btrfs_put_ordered_extent(ordered);
			break;
		}
	}

	cur_offset = alloc_start;
	while (1) {
2360 2361
		u64 actual_end;

2362 2363
		em = btrfs_get_extent(inode, NULL, 0, cur_offset,
				      alloc_end - cur_offset, 0);
2364 2365 2366 2367 2368 2369 2370
		if (IS_ERR_OR_NULL(em)) {
			if (!em)
				ret = -ENOMEM;
			else
				ret = PTR_ERR(em);
			break;
		}
2371
		last_byte = min(extent_map_end(em), alloc_end);
2372
		actual_end = min_t(u64, extent_map_end(em), offset + len);
2373
		last_byte = ALIGN(last_byte, blocksize);
2374

2375 2376 2377 2378 2379 2380 2381 2382
		if (em->block_start == EXTENT_MAP_HOLE ||
		    (cur_offset >= inode->i_size &&
		     !test_bit(EXTENT_FLAG_PREALLOC, &em->flags))) {
			ret = btrfs_prealloc_file_range(inode, mode, cur_offset,
							last_byte - cur_offset,
							1 << inode->i_blkbits,
							offset + len,
							&alloc_hint);
2383

2384 2385 2386 2387
			if (ret < 0) {
				free_extent_map(em);
				break;
			}
2388 2389 2390 2391 2392 2393 2394 2395 2396 2397
		} else if (actual_end > inode->i_size &&
			   !(mode & FALLOC_FL_KEEP_SIZE)) {
			/*
			 * We didn't need to allocate any more space, but we
			 * still extended the size of the file so we need to
			 * update i_size.
			 */
			inode->i_ctime = CURRENT_TIME;
			i_size_write(inode, actual_end);
			btrfs_ordered_update_i_size(inode, actual_end, NULL);
2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410
		}
		free_extent_map(em);

		cur_offset = last_byte;
		if (cur_offset >= alloc_end) {
			ret = 0;
			break;
		}
	}
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, alloc_start, locked_end,
			     &cached_state, GFP_NOFS);
out:
	mutex_unlock(&inode->i_mutex);
2411 2412 2413
	if (root->fs_info->quota_enabled)
		btrfs_qgroup_free(root, alloc_end - alloc_start);
out_reserve_fail:
2414
	/* Let go of our reservation. */
2415
	btrfs_free_reserved_data_space(inode, alloc_end - alloc_start);
2416 2417 2418
	return ret;
}

2419
static int find_desired_extent(struct inode *inode, loff_t *offset, int whence)
J
Josef Bacik 已提交
2420 2421
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
2422
	struct extent_map *em = NULL;
J
Josef Bacik 已提交
2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433
	struct extent_state *cached_state = NULL;
	u64 lockstart = *offset;
	u64 lockend = i_size_read(inode);
	u64 start = *offset;
	u64 len = i_size_read(inode);
	int ret = 0;

	lockend = max_t(u64, root->sectorsize, lockend);
	if (lockend <= lockstart)
		lockend = lockstart + root->sectorsize;

L
Liu Bo 已提交
2434
	lockend--;
J
Josef Bacik 已提交
2435 2436 2437 2438 2439 2440 2441
	len = lockend - lockstart + 1;

	len = max_t(u64, len, root->sectorsize);
	if (inode->i_size == 0)
		return -ENXIO;

	lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend, 0,
2442
			 &cached_state);
J
Josef Bacik 已提交
2443

2444
	while (start < inode->i_size) {
J
Josef Bacik 已提交
2445 2446
		em = btrfs_get_extent_fiemap(inode, NULL, 0, start, len, 0);
		if (IS_ERR(em)) {
2447
			ret = PTR_ERR(em);
2448
			em = NULL;
J
Josef Bacik 已提交
2449 2450 2451
			break;
		}

2452 2453 2454 2455 2456 2457 2458 2459
		if (whence == SEEK_HOLE &&
		    (em->block_start == EXTENT_MAP_HOLE ||
		     test_bit(EXTENT_FLAG_PREALLOC, &em->flags)))
			break;
		else if (whence == SEEK_DATA &&
			   (em->block_start != EXTENT_MAP_HOLE &&
			    !test_bit(EXTENT_FLAG_PREALLOC, &em->flags)))
			break;
J
Josef Bacik 已提交
2460 2461 2462

		start = em->start + em->len;
		free_extent_map(em);
2463
		em = NULL;
J
Josef Bacik 已提交
2464 2465
		cond_resched();
	}
2466 2467 2468 2469 2470 2471 2472
	free_extent_map(em);
	if (!ret) {
		if (whence == SEEK_DATA && start >= inode->i_size)
			ret = -ENXIO;
		else
			*offset = min_t(loff_t, start, inode->i_size);
	}
J
Josef Bacik 已提交
2473 2474 2475 2476 2477
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
			     &cached_state, GFP_NOFS);
	return ret;
}

2478
static loff_t btrfs_file_llseek(struct file *file, loff_t offset, int whence)
J
Josef Bacik 已提交
2479 2480 2481 2482 2483
{
	struct inode *inode = file->f_mapping->host;
	int ret;

	mutex_lock(&inode->i_mutex);
2484
	switch (whence) {
J
Josef Bacik 已提交
2485 2486
	case SEEK_END:
	case SEEK_CUR:
2487
		offset = generic_file_llseek(file, offset, whence);
J
Josef Bacik 已提交
2488 2489 2490
		goto out;
	case SEEK_DATA:
	case SEEK_HOLE:
2491 2492 2493 2494 2495
		if (offset >= i_size_read(inode)) {
			mutex_unlock(&inode->i_mutex);
			return -ENXIO;
		}

2496
		ret = find_desired_extent(inode, &offset, whence);
J
Josef Bacik 已提交
2497 2498 2499 2500 2501 2502
		if (ret) {
			mutex_unlock(&inode->i_mutex);
			return ret;
		}
	}

J
Jie Liu 已提交
2503
	offset = vfs_setpos(file, offset, inode->i_sb->s_maxbytes);
J
Josef Bacik 已提交
2504 2505 2506 2507 2508
out:
	mutex_unlock(&inode->i_mutex);
	return offset;
}

2509
const struct file_operations btrfs_file_operations = {
J
Josef Bacik 已提交
2510
	.llseek		= btrfs_file_llseek,
C
Chris Mason 已提交
2511
	.read		= do_sync_read,
2512
	.write		= do_sync_write,
C
Chris Mason 已提交
2513
	.aio_read       = generic_file_aio_read,
C
Chris Mason 已提交
2514
	.splice_read	= generic_file_splice_read,
2515
	.aio_write	= btrfs_file_aio_write,
C
Chris Mason 已提交
2516
	.mmap		= btrfs_file_mmap,
C
Chris Mason 已提交
2517
	.open		= generic_file_open,
2518
	.release	= btrfs_release_file,
C
Chris Mason 已提交
2519
	.fsync		= btrfs_sync_file,
2520
	.fallocate	= btrfs_fallocate,
C
Christoph Hellwig 已提交
2521
	.unlocked_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
2522
#ifdef CONFIG_COMPAT
C
Christoph Hellwig 已提交
2523
	.compat_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
2524 2525
#endif
};
2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543

void btrfs_auto_defrag_exit(void)
{
	if (btrfs_inode_defrag_cachep)
		kmem_cache_destroy(btrfs_inode_defrag_cachep);
}

int btrfs_auto_defrag_init(void)
{
	btrfs_inode_defrag_cachep = kmem_cache_create("btrfs_inode_defrag",
					sizeof(struct inode_defrag), 0,
					SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
					NULL);
	if (!btrfs_inode_defrag_cachep)
		return -ENOMEM;

	return 0;
}