file.c 76.1 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/falloc.h>
C
Chris Mason 已提交
28 29 30 31
#include <linux/swap.h>
#include <linux/writeback.h>
#include <linux/statfs.h>
#include <linux/compat.h>
32
#include <linux/slab.h>
33
#include <linux/btrfs.h>
34
#include <linux/uio.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"
J
Josef Bacik 已提交
43
#include "qgroup.h"
C
Chris Mason 已提交
44

45
static struct kmem_cache *btrfs_inode_defrag_cachep;
C
Chris Mason 已提交
46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70
/*
 * 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;
};

71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
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
86 87 88 89 90 91 92 93 94
/* 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
 */
95
static int __btrfs_add_inode_defrag(struct inode *inode,
C
Chris Mason 已提交
96 97 98 99 100 101
				    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;
102
	int ret;
C
Chris Mason 已提交
103 104 105 106 107 108

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

109 110
		ret = __compare_inode_defrag(defrag, entry);
		if (ret < 0)
C
Chris Mason 已提交
111
			p = &parent->rb_left;
112
		else if (ret > 0)
C
Chris Mason 已提交
113 114 115 116 117 118 119 120 121 122
			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;
123
			return -EEXIST;
C
Chris Mason 已提交
124 125
		}
	}
126
	set_bit(BTRFS_INODE_IN_DEFRAG, &BTRFS_I(inode)->runtime_flags);
C
Chris Mason 已提交
127 128
	rb_link_node(&defrag->rb_node, parent, p);
	rb_insert_color(&defrag->rb_node, &root->fs_info->defrag_inodes);
129 130
	return 0;
}
C
Chris Mason 已提交
131

132 133 134 135 136 137 138
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
139

140
	return 1;
C
Chris Mason 已提交
141 142 143 144 145 146 147 148 149 150 151 152
}

/*
 * 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;
153
	int ret;
C
Chris Mason 已提交
154

155
	if (!__need_auto_defrag(root))
C
Chris Mason 已提交
156 157
		return 0;

158
	if (test_bit(BTRFS_INODE_IN_DEFRAG, &BTRFS_I(inode)->runtime_flags))
C
Chris Mason 已提交
159 160 161 162 163 164 165
		return 0;

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

166
	defrag = kmem_cache_zalloc(btrfs_inode_defrag_cachep, GFP_NOFS);
C
Chris Mason 已提交
167 168 169
	if (!defrag)
		return -ENOMEM;

170
	defrag->ino = btrfs_ino(inode);
C
Chris Mason 已提交
171 172 173 174
	defrag->transid = transid;
	defrag->root = root->root_key.objectid;

	spin_lock(&root->fs_info->defrag_inodes_lock);
175 176 177 178 179 180 181 182 183 184
	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 {
185
		kmem_cache_free(btrfs_inode_defrag_cachep, defrag);
186
	}
C
Chris Mason 已提交
187
	spin_unlock(&root->fs_info->defrag_inodes_lock);
188
	return 0;
C
Chris Mason 已提交
189 190 191
}

/*
192 193 194
 * 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
195
 */
196 197
static void btrfs_requeue_inode_defrag(struct inode *inode,
				       struct inode_defrag *defrag)
198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218
{
	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
219
/*
220 221
 * pick the defragable inode that we want, if it doesn't exist, we will get
 * the next one.
C
Chris Mason 已提交
222
 */
223 224
static struct inode_defrag *
btrfs_pick_defrag_inode(struct btrfs_fs_info *fs_info, u64 root, u64 ino)
C
Chris Mason 已提交
225 226
{
	struct inode_defrag *entry = NULL;
227
	struct inode_defrag tmp;
C
Chris Mason 已提交
228 229
	struct rb_node *p;
	struct rb_node *parent = NULL;
230 231 232 233
	int ret;

	tmp.ino = ino;
	tmp.root = root;
C
Chris Mason 已提交
234

235 236
	spin_lock(&fs_info->defrag_inodes_lock);
	p = fs_info->defrag_inodes.rb_node;
C
Chris Mason 已提交
237 238 239 240
	while (p) {
		parent = p;
		entry = rb_entry(parent, struct inode_defrag, rb_node);

241 242
		ret = __compare_inode_defrag(&tmp, entry);
		if (ret < 0)
C
Chris Mason 已提交
243
			p = parent->rb_left;
244
		else if (ret > 0)
C
Chris Mason 已提交
245 246
			p = parent->rb_right;
		else
247
			goto out;
C
Chris Mason 已提交
248 249
	}

250 251 252
	if (parent && __compare_inode_defrag(&tmp, entry) > 0) {
		parent = rb_next(parent);
		if (parent)
C
Chris Mason 已提交
253
			entry = rb_entry(parent, struct inode_defrag, rb_node);
254 255
		else
			entry = NULL;
C
Chris Mason 已提交
256
	}
257 258 259 260 261
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
262 263
}

264
void btrfs_cleanup_defrag_inodes(struct btrfs_fs_info *fs_info)
C
Chris Mason 已提交
265 266
{
	struct inode_defrag *defrag;
267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291
	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
292 293 294 295 296
	struct btrfs_root *inode_root;
	struct inode *inode;
	struct btrfs_key key;
	struct btrfs_ioctl_defrag_range_args range;
	int num_defrag;
297 298
	int index;
	int ret;
C
Chris Mason 已提交
299

300 301
	/* get the inode */
	key.objectid = defrag->root;
302
	key.type = BTRFS_ROOT_ITEM_KEY;
303
	key.offset = (u64)-1;
304 305 306

	index = srcu_read_lock(&fs_info->subvol_srcu);

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

	key.objectid = defrag->ino;
314
	key.type = BTRFS_INODE_ITEM_KEY;
315 316 317
	key.offset = 0;
	inode = btrfs_iget(fs_info->sb, &key, inode_root, NULL);
	if (IS_ERR(inode)) {
318 319
		ret = PTR_ERR(inode);
		goto cleanup;
320
	}
321
	srcu_read_unlock(&fs_info->subvol_srcu, index);
322 323 324

	/* do a chunk of defrag */
	clear_bit(BTRFS_INODE_IN_DEFRAG, &BTRFS_I(inode)->runtime_flags);
C
Chris Mason 已提交
325 326
	memset(&range, 0, sizeof(range));
	range.len = (u64)-1;
327
	range.start = defrag->last_offset;
M
Miao Xie 已提交
328 329

	sb_start_write(fs_info->sb);
330 331
	num_defrag = btrfs_defrag_file(inode, NULL, &range, defrag->transid,
				       BTRFS_DEFRAG_BATCH);
M
Miao Xie 已提交
332
	sb_end_write(fs_info->sb);
333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355
	/*
	 * 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;
356 357 358 359
cleanup:
	srcu_read_unlock(&fs_info->subvol_srcu, index);
	kmem_cache_free(btrfs_inode_defrag_cachep, defrag);
	return ret;
360 361 362 363 364 365 366 367 368 369 370
}

/*
 * 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
371 372

	atomic_inc(&fs_info->defrag_running);
373
	while (1) {
M
Miao Xie 已提交
374 375 376 377 378
		/* Pause the auto defragger. */
		if (test_bit(BTRFS_FS_STATE_REMOUNTING,
			     &fs_info->fs_state))
			break;

379 380
		if (!__need_auto_defrag(fs_info->tree_root))
			break;
C
Chris Mason 已提交
381 382

		/* find an inode to defrag */
383 384
		defrag = btrfs_pick_defrag_inode(fs_info, root_objectid,
						 first_ino);
C
Chris Mason 已提交
385
		if (!defrag) {
386
			if (root_objectid || first_ino) {
387
				root_objectid = 0;
C
Chris Mason 已提交
388 389 390 391 392 393 394 395
				first_ino = 0;
				continue;
			} else {
				break;
			}
		}

		first_ino = defrag->ino + 1;
396
		root_objectid = defrag->root;
C
Chris Mason 已提交
397

398
		__btrfs_run_defrag_inode(fs_info, defrag);
C
Chris Mason 已提交
399 400 401 402 403 404 405 406 407 408
	}
	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
409

C
Chris Mason 已提交
410 411 412
/* 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
413
static noinline int btrfs_copy_from_user(loff_t pos, int num_pages,
J
Josef Bacik 已提交
414
					 size_t write_bytes,
415
					 struct page **prepared_pages,
416
					 struct iov_iter *i)
C
Chris Mason 已提交
417
{
418
	size_t copied = 0;
J
Josef Bacik 已提交
419
	size_t total_copied = 0;
420
	int pg = 0;
C
Chris Mason 已提交
421 422
	int offset = pos & (PAGE_CACHE_SIZE - 1);

423
	while (write_bytes > 0) {
C
Chris Mason 已提交
424 425
		size_t count = min_t(size_t,
				     PAGE_CACHE_SIZE - offset, write_bytes);
426
		struct page *page = prepared_pages[pg];
427 428 429 430
		/*
		 * Copy data from userspace to the current page
		 */
		copied = iov_iter_copy_from_user_atomic(page, i, offset, count);
431

C
Chris Mason 已提交
432 433
		/* Flush processor's dcache for this page */
		flush_dcache_page(page);
434 435 436 437 438 439 440 441 442 443 444 445 446

		/*
		 * 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;

447 448
		iov_iter_advance(i, copied);
		write_bytes -= copied;
449
		total_copied += copied;
C
Chris Mason 已提交
450

A
Al Viro 已提交
451
		/* Return to btrfs_file_write_iter to fault page */
J
Josef Bacik 已提交
452
		if (unlikely(copied == 0))
453
			break;
454

455
		if (copied < PAGE_CACHE_SIZE - offset) {
456 457 458 459 460
			offset += copied;
		} else {
			pg++;
			offset = 0;
		}
C
Chris Mason 已提交
461
	}
462
	return total_copied;
C
Chris Mason 已提交
463 464
}

C
Chris Mason 已提交
465 466 467
/*
 * unlocks pages after btrfs_file_write is done with them
 */
468
static void btrfs_drop_pages(struct page **pages, size_t num_pages)
C
Chris Mason 已提交
469 470 471
{
	size_t i;
	for (i = 0; i < num_pages; i++) {
C
Chris Mason 已提交
472 473
		/* page checked is some magic around finding pages that
		 * have been modified without going through btrfs_set_page_dirty
474 475 476
		 * clear it here. There should be no need to mark the pages
		 * accessed as prepare_pages should have marked them accessed
		 * in prepare_pages via find_or_create_page()
C
Chris Mason 已提交
477
		 */
C
Chris Mason 已提交
478
		ClearPageChecked(pages[i]);
C
Chris Mason 已提交
479 480 481 482 483
		unlock_page(pages[i]);
		page_cache_release(pages[i]);
	}
}

C
Chris Mason 已提交
484 485 486 487 488 489 490 491
/*
 * 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.
 */
492
int btrfs_dirty_pages(struct btrfs_root *root, struct inode *inode,
493 494 495
			     struct page **pages, size_t num_pages,
			     loff_t pos, size_t write_bytes,
			     struct extent_state **cached)
C
Chris Mason 已提交
496 497
{
	int err = 0;
498
	int i;
499
	u64 num_bytes;
500 501 502 503
	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
504

505
	start_pos = pos & ~((u64)root->sectorsize - 1);
506
	num_bytes = ALIGN(write_bytes + pos - start_pos, root->sectorsize);
C
Chris Mason 已提交
507

508
	end_of_last_block = start_pos + num_bytes - 1;
509
	err = btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block,
510
					cached);
J
Josef Bacik 已提交
511 512
	if (err)
		return err;
J
Josef Bacik 已提交
513

C
Chris Mason 已提交
514 515 516 517 518
	for (i = 0; i < num_pages; i++) {
		struct page *p = pages[i];
		SetPageUptodate(p);
		ClearPageChecked(p);
		set_page_dirty(p);
519
	}
J
Josef Bacik 已提交
520 521 522 523 524 525 526

	/*
	 * 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)
527
		i_size_write(inode, end_pos);
528
	return 0;
C
Chris Mason 已提交
529 530
}

C
Chris Mason 已提交
531 532 533 534
/*
 * this drops all the extents in the cache that intersect the range
 * [start, end].  Existing extents are split as required.
 */
535 536
void btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
			     int skip_pinned)
537 538
{
	struct extent_map *em;
539 540
	struct extent_map *split = NULL;
	struct extent_map *split2 = NULL;
541
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
542
	u64 len = end - start + 1;
J
Josef Bacik 已提交
543
	u64 gen;
544 545
	int ret;
	int testend = 1;
546
	unsigned long flags;
C
Chris Mason 已提交
547
	int compressed = 0;
J
Josef Bacik 已提交
548
	bool modified;
549

550
	WARN_ON(end < start);
551
	if (end == (u64)-1) {
552
		len = (u64)-1;
553 554
		testend = 0;
	}
C
Chris Mason 已提交
555
	while (1) {
556 557
		int no_splits = 0;

J
Josef Bacik 已提交
558
		modified = false;
559
		if (!split)
560
			split = alloc_extent_map();
561
		if (!split2)
562
			split2 = alloc_extent_map();
563 564
		if (!split || !split2)
			no_splits = 1;
565

566
		write_lock(&em_tree->lock);
567
		em = lookup_extent_mapping(em_tree, start, len);
568
		if (!em) {
569
			write_unlock(&em_tree->lock);
570
			break;
571
		}
572
		flags = em->flags;
J
Josef Bacik 已提交
573
		gen = em->generation;
574
		if (skip_pinned && test_bit(EXTENT_FLAG_PINNED, &em->flags)) {
575
			if (testend && em->start + em->len >= start + len) {
576
				free_extent_map(em);
C
Chris Mason 已提交
577
				write_unlock(&em_tree->lock);
578 579
				break;
			}
580 581
			start = em->start + em->len;
			if (testend)
582 583
				len = start + len - (em->start + em->len);
			free_extent_map(em);
C
Chris Mason 已提交
584
			write_unlock(&em_tree->lock);
585 586
			continue;
		}
C
Chris Mason 已提交
587
		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
588
		clear_bit(EXTENT_FLAG_PINNED, &em->flags);
L
Liu Bo 已提交
589
		clear_bit(EXTENT_FLAG_LOGGING, &flags);
J
Josef Bacik 已提交
590
		modified = !list_empty(&em->list);
591 592
		if (no_splits)
			goto next;
593

594
		if (em->start < start) {
595 596
			split->start = em->start;
			split->len = start - em->start;
597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616

			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
617
			split->generation = gen;
618
			split->bdev = em->bdev;
619
			split->flags = flags;
620
			split->compress_type = em->compress_type;
621
			replace_extent_mapping(em_tree, em, split, modified);
622 623 624 625
			free_extent_map(split);
			split = split2;
			split2 = NULL;
		}
626
		if (testend && em->start + em->len > start + len) {
627 628 629 630 631
			u64 diff = start + len - em->start;

			split->start = start + len;
			split->len = em->start + em->len - (start + len);
			split->bdev = em->bdev;
632
			split->flags = flags;
633
			split->compress_type = em->compress_type;
J
Josef Bacik 已提交
634
			split->generation = gen;
635 636 637

			if (em->block_start < EXTENT_MAP_LAST_BYTE) {
				split->orig_block_len = max(em->block_len,
638
						    em->orig_block_len);
639

640 641 642 643 644 645 646 647 648 649 650
				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
651
			} else {
652 653 654 655 656
				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
657
			}
658

659 660 661 662 663 664 665 666
			if (extent_map_in_tree(em)) {
				replace_extent_mapping(em_tree, em, split,
						       modified);
			} else {
				ret = add_extent_mapping(em_tree, split,
							 modified);
				ASSERT(ret == 0); /* Logic error */
			}
667 668 669
			free_extent_map(split);
			split = NULL;
		}
670
next:
671 672
		if (extent_map_in_tree(em))
			remove_extent_mapping(em_tree, em);
673
		write_unlock(&em_tree->lock);
674

675 676 677 678 679
		/* once for us */
		free_extent_map(em);
		/* once for the tree*/
		free_extent_map(em);
	}
680 681 682 683
	if (split)
		free_extent_map(split);
	if (split2)
		free_extent_map(split2);
684 685
}

C
Chris Mason 已提交
686 687 688 689 690 691 692 693 694
/*
 * 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
695 696 697
int __btrfs_drop_extents(struct btrfs_trans_handle *trans,
			 struct btrfs_root *root, struct inode *inode,
			 struct btrfs_path *path, u64 start, u64 end,
698 699 700 701
			 u64 *drop_end, int drop_cache,
			 int replace_extent,
			 u32 extent_item_size,
			 int *key_inserted)
C
Chris Mason 已提交
702
{
703
	struct extent_buffer *leaf;
Y
Yan, Zheng 已提交
704
	struct btrfs_file_extent_item *fi;
705
	struct btrfs_key key;
Y
Yan, Zheng 已提交
706
	struct btrfs_key new_key;
L
Li Zefan 已提交
707
	u64 ino = btrfs_ino(inode);
Y
Yan, Zheng 已提交
708 709 710 711 712 713 714 715
	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
716
	int recow;
717
	int ret;
718
	int modify_tree = -1;
719
	int update_refs;
720
	int found = 0;
721
	int leafs_visited = 0;
C
Chris Mason 已提交
722

C
Chris Mason 已提交
723 724
	if (drop_cache)
		btrfs_drop_extent_cache(inode, start, end - 1, 0);
725

726
	if (start >= BTRFS_I(inode)->disk_i_size && !replace_extent)
727 728
		modify_tree = 0;

729 730
	update_refs = (test_bit(BTRFS_ROOT_REF_COWS, &root->state) ||
		       root == root->fs_info->tree_root);
C
Chris Mason 已提交
731
	while (1) {
C
Chris Mason 已提交
732
		recow = 0;
L
Li Zefan 已提交
733
		ret = btrfs_lookup_file_extent(trans, root, path, ino,
734
					       search_start, modify_tree);
C
Chris Mason 已提交
735
		if (ret < 0)
Y
Yan, Zheng 已提交
736 737 738 739
			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
740
			if (key.objectid == ino &&
Y
Yan, Zheng 已提交
741 742
			    key.type == BTRFS_EXTENT_DATA_KEY)
				path->slots[0]--;
C
Chris Mason 已提交
743
		}
Y
Yan, Zheng 已提交
744
		ret = 0;
745
		leafs_visited++;
746
next_slot:
747
		leaf = path->nodes[0];
Y
Yan, Zheng 已提交
748 749 750 751 752 753 754 755
		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;
756
			}
757
			leafs_visited++;
Y
Yan, Zheng 已提交
758 759 760 761 762
			leaf = path->nodes[0];
			recow = 1;
		}

		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
L
Li Zefan 已提交
763
		if (key.objectid > ino ||
Y
Yan, Zheng 已提交
764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779
		    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 +
780 781
				btrfs_file_extent_inline_len(leaf,
						     path->slots[0], fi);
782
		} else {
Y
Yan, Zheng 已提交
783
			WARN_ON(1);
784
			extent_end = search_start;
C
Chris Mason 已提交
785 786
		}

787 788 789 790 791 792 793 794 795 796 797 798
		/*
		 * Don't skip extent items representing 0 byte lengths. They
		 * used to be created (bug) if while punching holes we hit
		 * -ENOSPC condition. So if we find one here, just ensure we
		 * delete it, otherwise we would insert a new file extent item
		 * with the same key (offset) as that 0 bytes length file
		 * extent item in the call to setup_items_for_insert() later
		 * in this function.
		 */
		if (extent_end == key.offset && extent_end >= search_start)
			goto delete_extent_item;

Y
Yan, Zheng 已提交
799 800
		if (extent_end <= search_start) {
			path->slots[0]++;
801
			goto next_slot;
C
Chris Mason 已提交
802 803
		}

804
		found = 1;
Y
Yan, Zheng 已提交
805
		search_start = max(key.offset, start);
806 807
		if (recow || !modify_tree) {
			modify_tree = -1;
808
			btrfs_release_path(path);
Y
Yan, Zheng 已提交
809
			continue;
C
Chris Mason 已提交
810
		}
Y
Yan Zheng 已提交
811

Y
Yan, Zheng 已提交
812 813 814 815 816 817
		/*
		 *     | - range to drop - |
		 *  | -------- extent -------- |
		 */
		if (start > key.offset && end < extent_end) {
			BUG_ON(del_nr > 0);
818
			if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
819
				ret = -EOPNOTSUPP;
820 821
				break;
			}
Y
Yan, Zheng 已提交
822 823 824 825 826 827

			memcpy(&new_key, &key, sizeof(new_key));
			new_key.offset = start;
			ret = btrfs_duplicate_item(trans, root, path,
						   &new_key);
			if (ret == -EAGAIN) {
828
				btrfs_release_path(path);
Y
Yan, Zheng 已提交
829
				continue;
Y
Yan Zheng 已提交
830
			}
Y
Yan, Zheng 已提交
831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848
			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
849
			if (update_refs && disk_bytenr > 0) {
850
				ret = btrfs_inc_extent_ref(trans, root,
Y
Yan, Zheng 已提交
851 852 853
						disk_bytenr, num_bytes, 0,
						root->root_key.objectid,
						new_key.objectid,
J
Josef Bacik 已提交
854
						start - extent_offset, 1);
855
				BUG_ON(ret); /* -ENOMEM */
856
			}
Y
Yan, Zheng 已提交
857
			key.offset = start;
Y
Yan Zheng 已提交
858
		}
Y
Yan, Zheng 已提交
859 860 861 862 863
		/*
		 *  | ---- range to drop ----- |
		 *      | -------- extent -------- |
		 */
		if (start <= key.offset && end < extent_end) {
864
			if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
865
				ret = -EOPNOTSUPP;
866 867
				break;
			}
Y
Yan Zheng 已提交
868

Y
Yan, Zheng 已提交
869 870
			memcpy(&new_key, &key, sizeof(new_key));
			new_key.offset = end;
871
			btrfs_set_item_key_safe(root, path, &new_key);
Y
Yan Zheng 已提交
872

Y
Yan, Zheng 已提交
873 874 875 876 877
			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);
878
			if (update_refs && disk_bytenr > 0)
Y
Yan, Zheng 已提交
879 880
				inode_sub_bytes(inode, end - key.offset);
			break;
C
Chris Mason 已提交
881
		}
882

Y
Yan, Zheng 已提交
883 884 885 886 887 888 889
		search_start = extent_end;
		/*
		 *       | ---- range to drop ----- |
		 *  | -------- extent -------- |
		 */
		if (start > key.offset && end >= extent_end) {
			BUG_ON(del_nr > 0);
890
			if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
891
				ret = -EOPNOTSUPP;
892 893
				break;
			}
894

Y
Yan, Zheng 已提交
895 896 897
			btrfs_set_file_extent_num_bytes(leaf, fi,
							start - key.offset);
			btrfs_mark_buffer_dirty(leaf);
898
			if (update_refs && disk_bytenr > 0)
Y
Yan, Zheng 已提交
899 900 901
				inode_sub_bytes(inode, extent_end - start);
			if (end == extent_end)
				break;
C
Chris Mason 已提交
902

Y
Yan, Zheng 已提交
903 904
			path->slots[0]++;
			goto next_slot;
Z
Zheng Yan 已提交
905 906
		}

Y
Yan, Zheng 已提交
907 908 909 910 911
		/*
		 *  | ---- range to drop ----- |
		 *    | ------ extent ------ |
		 */
		if (start <= key.offset && end >= extent_end) {
912
delete_extent_item:
Y
Yan, Zheng 已提交
913 914 915 916 917 918 919
			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
920

J
Josef Bacik 已提交
921 922
			if (update_refs &&
			    extent_type == BTRFS_FILE_EXTENT_INLINE) {
923
				inode_sub_bytes(inode,
Y
Yan, Zheng 已提交
924 925 926
						extent_end - key.offset);
				extent_end = ALIGN(extent_end,
						   root->sectorsize);
J
Josef Bacik 已提交
927
			} else if (update_refs && disk_bytenr > 0) {
Z
Zheng Yan 已提交
928
				ret = btrfs_free_extent(trans, root,
Y
Yan, Zheng 已提交
929 930
						disk_bytenr, num_bytes, 0,
						root->root_key.objectid,
931
						key.objectid, key.offset -
A
Arne Jansen 已提交
932
						extent_offset, 0);
933
				BUG_ON(ret); /* -ENOMEM */
Y
Yan, Zheng 已提交
934 935
				inode_sub_bytes(inode,
						extent_end - key.offset);
Z
Zheng Yan 已提交
936 937
			}

Y
Yan, Zheng 已提交
938 939 940 941 942 943 944 945 946 947
			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);
948 949
			if (ret) {
				btrfs_abort_transaction(trans, root, ret);
J
Josef Bacik 已提交
950
				break;
951
			}
Y
Yan, Zheng 已提交
952 953 954 955

			del_nr = 0;
			del_slot = 0;

956
			btrfs_release_path(path);
Y
Yan, Zheng 已提交
957
			continue;
C
Chris Mason 已提交
958
		}
Y
Yan, Zheng 已提交
959 960

		BUG_ON(1);
C
Chris Mason 已提交
961
	}
Y
Yan, Zheng 已提交
962

963
	if (!ret && del_nr > 0) {
964 965 966 967
		/*
		 * Set path->slots[0] to first slot, so that after the delete
		 * if items are move off from our leaf to its immediate left or
		 * right neighbor leafs, we end up with a correct and adjusted
968
		 * path->slots[0] for our insertion (if replace_extent != 0).
969 970
		 */
		path->slots[0] = del_slot;
Y
Yan, Zheng 已提交
971
		ret = btrfs_del_items(trans, root, path, del_slot, del_nr);
972 973
		if (ret)
			btrfs_abort_transaction(trans, root, ret);
974
	}
975

976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996
	leaf = path->nodes[0];
	/*
	 * If btrfs_del_items() was called, it might have deleted a leaf, in
	 * which case it unlocked our path, so check path->locks[0] matches a
	 * write lock.
	 */
	if (!ret && replace_extent && leafs_visited == 1 &&
	    (path->locks[0] == BTRFS_WRITE_LOCK_BLOCKING ||
	     path->locks[0] == BTRFS_WRITE_LOCK) &&
	    btrfs_leaf_free_space(root, leaf) >=
	    sizeof(struct btrfs_item) + extent_item_size) {

		key.objectid = ino;
		key.type = BTRFS_EXTENT_DATA_KEY;
		key.offset = start;
		if (!del_nr && path->slots[0] < btrfs_header_nritems(leaf)) {
			struct btrfs_key slot_key;

			btrfs_item_key_to_cpu(leaf, &slot_key, path->slots[0]);
			if (btrfs_comp_cpu_keys(&key, &slot_key) > 0)
				path->slots[0]++;
997
		}
998 999 1000 1001 1002 1003
		setup_items_for_insert(root, path, &key,
				       &extent_item_size,
				       extent_item_size,
				       sizeof(struct btrfs_item) +
				       extent_item_size, 1);
		*key_inserted = 1;
Y
Yan Zheng 已提交
1004
	}
Y
Yan, Zheng 已提交
1005

1006 1007
	if (!replace_extent || !(*key_inserted))
		btrfs_release_path(path);
J
Josef Bacik 已提交
1008
	if (drop_end)
1009
		*drop_end = found ? min(end, extent_end) : end;
J
Josef Bacik 已提交
1010 1011 1012 1013 1014
	return ret;
}

int btrfs_drop_extents(struct btrfs_trans_handle *trans,
		       struct btrfs_root *root, struct inode *inode, u64 start,
1015
		       u64 end, int drop_cache)
J
Josef Bacik 已提交
1016 1017 1018 1019 1020 1021 1022
{
	struct btrfs_path *path;
	int ret;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
J
Josef Bacik 已提交
1023
	ret = __btrfs_drop_extents(trans, root, inode, path, start, end, NULL,
1024
				   drop_cache, 0, 0, NULL);
Y
Yan, Zheng 已提交
1025
	btrfs_free_path(path);
C
Chris Mason 已提交
1026 1027 1028
	return ret;
}

Y
Yan Zheng 已提交
1029
static int extent_mergeable(struct extent_buffer *leaf, int slot,
1030 1031
			    u64 objectid, u64 bytenr, u64 orig_offset,
			    u64 *start, u64 *end)
Y
Yan Zheng 已提交
1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
{
	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 ||
1047
	    btrfs_file_extent_offset(leaf, fi) != key.offset - orig_offset ||
Y
Yan Zheng 已提交
1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071
	    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
1072
	struct btrfs_root *root = BTRFS_I(inode)->root;
Y
Yan Zheng 已提交
1073 1074 1075 1076
	struct extent_buffer *leaf;
	struct btrfs_path *path;
	struct btrfs_file_extent_item *fi;
	struct btrfs_key key;
Y
Yan, Zheng 已提交
1077
	struct btrfs_key new_key;
Y
Yan Zheng 已提交
1078 1079 1080
	u64 bytenr;
	u64 num_bytes;
	u64 extent_end;
1081
	u64 orig_offset;
Y
Yan Zheng 已提交
1082 1083
	u64 other_start;
	u64 other_end;
Y
Yan, Zheng 已提交
1084 1085 1086
	u64 split;
	int del_nr = 0;
	int del_slot = 0;
1087
	int recow;
Y
Yan Zheng 已提交
1088
	int ret;
L
Li Zefan 已提交
1089
	u64 ino = btrfs_ino(inode);
Y
Yan Zheng 已提交
1090 1091

	path = btrfs_alloc_path();
1092 1093
	if (!path)
		return -ENOMEM;
Y
Yan Zheng 已提交
1094
again:
1095
	recow = 0;
Y
Yan, Zheng 已提交
1096
	split = start;
L
Li Zefan 已提交
1097
	key.objectid = ino;
Y
Yan Zheng 已提交
1098
	key.type = BTRFS_EXTENT_DATA_KEY;
Y
Yan, Zheng 已提交
1099
	key.offset = split;
Y
Yan Zheng 已提交
1100 1101

	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1102 1103
	if (ret < 0)
		goto out;
Y
Yan Zheng 已提交
1104 1105 1106 1107 1108
	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
1109
	BUG_ON(key.objectid != ino || key.type != BTRFS_EXTENT_DATA_KEY);
Y
Yan Zheng 已提交
1110 1111
	fi = btrfs_item_ptr(leaf, path->slots[0],
			    struct btrfs_file_extent_item);
Y
Yan, Zheng 已提交
1112 1113
	BUG_ON(btrfs_file_extent_type(leaf, fi) !=
	       BTRFS_FILE_EXTENT_PREALLOC);
Y
Yan Zheng 已提交
1114 1115 1116 1117 1118
	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);
1119
	orig_offset = key.offset - btrfs_file_extent_offset(leaf, fi);
1120 1121 1122 1123 1124 1125
	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
1126
				     ino, bytenr, orig_offset,
1127 1128
				     &other_start, &other_end)) {
			new_key.offset = end;
1129
			btrfs_set_item_key_safe(root, path, &new_key);
1130 1131
			fi = btrfs_item_ptr(leaf, path->slots[0],
					    struct btrfs_file_extent_item);
1132 1133
			btrfs_set_file_extent_generation(leaf, fi,
							 trans->transid);
1134 1135 1136 1137 1138 1139
			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);
1140 1141
			btrfs_set_file_extent_generation(leaf, fi,
							 trans->transid);
1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152
			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
1153
				     ino, bytenr, orig_offset,
1154 1155 1156 1157 1158
				     &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);
1159 1160
			btrfs_set_file_extent_generation(leaf, fi,
							 trans->transid);
1161 1162
			path->slots[0]++;
			new_key.offset = start;
1163
			btrfs_set_item_key_safe(root, path, &new_key);
1164 1165 1166

			fi = btrfs_item_ptr(leaf, path->slots[0],
					    struct btrfs_file_extent_item);
1167 1168
			btrfs_set_file_extent_generation(leaf, fi,
							 trans->transid);
1169 1170 1171 1172 1173 1174 1175 1176
			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
1177

Y
Yan, Zheng 已提交
1178 1179 1180 1181 1182 1183 1184
	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) {
1185
			btrfs_release_path(path);
Y
Yan, Zheng 已提交
1186
			goto again;
Y
Yan Zheng 已提交
1187
		}
1188 1189 1190 1191
		if (ret < 0) {
			btrfs_abort_transaction(trans, root, ret);
			goto out;
		}
Y
Yan Zheng 已提交
1192

Y
Yan, Zheng 已提交
1193 1194
		leaf = path->nodes[0];
		fi = btrfs_item_ptr(leaf, path->slots[0] - 1,
Y
Yan Zheng 已提交
1195
				    struct btrfs_file_extent_item);
1196
		btrfs_set_file_extent_generation(leaf, fi, trans->transid);
Y
Yan Zheng 已提交
1197
		btrfs_set_file_extent_num_bytes(leaf, fi,
Y
Yan, Zheng 已提交
1198 1199 1200 1201 1202
						split - key.offset);

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

1203
		btrfs_set_file_extent_generation(leaf, fi, trans->transid);
Y
Yan, Zheng 已提交
1204 1205 1206
		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
1207 1208
		btrfs_mark_buffer_dirty(leaf);

Y
Yan, Zheng 已提交
1209 1210
		ret = btrfs_inc_extent_ref(trans, root, bytenr, num_bytes, 0,
					   root->root_key.objectid,
J
Josef Bacik 已提交
1211
					   ino, orig_offset, 1);
1212
		BUG_ON(ret); /* -ENOMEM */
Y
Yan Zheng 已提交
1213

Y
Yan, Zheng 已提交
1214 1215 1216 1217
		if (split == start) {
			key.offset = start;
		} else {
			BUG_ON(start != key.offset);
Y
Yan Zheng 已提交
1218
			path->slots[0]--;
Y
Yan, Zheng 已提交
1219
			extent_end = end;
Y
Yan Zheng 已提交
1220
		}
1221
		recow = 1;
Y
Yan Zheng 已提交
1222 1223
	}

Y
Yan, Zheng 已提交
1224 1225
	other_start = end;
	other_end = 0;
1226
	if (extent_mergeable(leaf, path->slots[0] + 1,
L
Li Zefan 已提交
1227
			     ino, bytenr, orig_offset,
1228 1229
			     &other_start, &other_end)) {
		if (recow) {
1230
			btrfs_release_path(path);
1231 1232
			goto again;
		}
Y
Yan, Zheng 已提交
1233 1234 1235 1236 1237
		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
1238
					ino, orig_offset, 0);
1239
		BUG_ON(ret); /* -ENOMEM */
Y
Yan Zheng 已提交
1240
	}
Y
Yan, Zheng 已提交
1241 1242
	other_start = 0;
	other_end = start;
1243
	if (extent_mergeable(leaf, path->slots[0] - 1,
L
Li Zefan 已提交
1244
			     ino, bytenr, orig_offset,
1245 1246
			     &other_start, &other_end)) {
		if (recow) {
1247
			btrfs_release_path(path);
1248 1249
			goto again;
		}
Y
Yan, Zheng 已提交
1250 1251 1252 1253 1254
		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
1255
					ino, orig_offset, 0);
1256
		BUG_ON(ret); /* -ENOMEM */
Y
Yan, Zheng 已提交
1257 1258
	}
	if (del_nr == 0) {
1259 1260
		fi = btrfs_item_ptr(leaf, path->slots[0],
			   struct btrfs_file_extent_item);
Y
Yan, Zheng 已提交
1261 1262
		btrfs_set_file_extent_type(leaf, fi,
					   BTRFS_FILE_EXTENT_REG);
1263
		btrfs_set_file_extent_generation(leaf, fi, trans->transid);
Y
Yan, Zheng 已提交
1264
		btrfs_mark_buffer_dirty(leaf);
1265
	} else {
1266 1267
		fi = btrfs_item_ptr(leaf, del_slot - 1,
			   struct btrfs_file_extent_item);
1268 1269
		btrfs_set_file_extent_type(leaf, fi,
					   BTRFS_FILE_EXTENT_REG);
1270
		btrfs_set_file_extent_generation(leaf, fi, trans->transid);
1271 1272 1273
		btrfs_set_file_extent_num_bytes(leaf, fi,
						extent_end - key.offset);
		btrfs_mark_buffer_dirty(leaf);
Y
Yan, Zheng 已提交
1274

1275
		ret = btrfs_del_items(trans, root, path, del_slot, del_nr);
1276 1277 1278 1279
		if (ret < 0) {
			btrfs_abort_transaction(trans, root, ret);
			goto out;
		}
1280
	}
Y
Yan, Zheng 已提交
1281
out:
Y
Yan Zheng 已提交
1282 1283 1284 1285
	btrfs_free_path(path);
	return 0;
}

1286 1287 1288 1289
/*
 * on error we return an unlocked page and the error value
 * on success we return a locked page and 0
 */
1290 1291
static int prepare_uptodate_page(struct page *page, u64 pos,
				 bool force_uptodate)
1292 1293 1294
{
	int ret = 0;

1295 1296
	if (((pos & (PAGE_CACHE_SIZE - 1)) || force_uptodate) &&
	    !PageUptodate(page)) {
1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308
		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
1309
/*
1310
 * this just gets pages into the page cache and locks them down.
C
Chris Mason 已提交
1311
 */
1312 1313 1314
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
1315 1316 1317
{
	int i;
	unsigned long index = pos >> PAGE_CACHE_SHIFT;
1318
	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1319
	int err = 0;
1320
	int faili;
1321

C
Chris Mason 已提交
1322
	for (i = 0; i < num_pages; i++) {
1323
		pages[i] = find_or_create_page(inode->i_mapping, index + i,
1324
					       mask | __GFP_WRITE);
C
Chris Mason 已提交
1325
		if (!pages[i]) {
1326 1327 1328 1329 1330 1331
			faili = i - 1;
			err = -ENOMEM;
			goto fail;
		}

		if (i == 0)
1332 1333
			err = prepare_uptodate_page(pages[i], pos,
						    force_uptodate);
1334 1335
		if (i == num_pages - 1)
			err = prepare_uptodate_page(pages[i],
1336
						    pos + write_bytes, false);
1337 1338 1339 1340
		if (err) {
			page_cache_release(pages[i]);
			faili = i - 1;
			goto fail;
C
Chris Mason 已提交
1341
		}
C
Chris Mason 已提交
1342
		wait_on_page_writeback(pages[i]);
C
Chris Mason 已提交
1343
	}
1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379

	return 0;
fail:
	while (faili >= 0) {
		unlock_page(pages[faili]);
		page_cache_release(pages[faili]);
		faili--;
	}
	return err;

}

/*
 * This function locks the extent and properly waits for data=ordered extents
 * to finish before allowing the pages to be modified if need.
 *
 * The return value:
 * 1 - the extent is locked
 * 0 - the extent is not locked, and everything is OK
 * -EAGAIN - need re-prepare the pages
 * the other < 0 number - Something wrong happens
 */
static noinline int
lock_and_cleanup_extent_if_need(struct inode *inode, struct page **pages,
				size_t num_pages, loff_t pos,
				u64 *lockstart, u64 *lockend,
				struct extent_state **cached_state)
{
	u64 start_pos;
	u64 last_pos;
	int i;
	int ret = 0;

	start_pos = pos & ~((u64)PAGE_CACHE_SIZE - 1);
	last_pos = start_pos + ((u64)num_pages << PAGE_CACHE_SHIFT) - 1;

1380
	if (start_pos < inode->i_size) {
1381
		struct btrfs_ordered_extent *ordered;
1382
		lock_extent_bits(&BTRFS_I(inode)->io_tree,
1383
				 start_pos, last_pos, 0, cached_state);
1384 1385
		ordered = btrfs_lookup_ordered_range(inode, start_pos,
						     last_pos - start_pos + 1);
1386 1387
		if (ordered &&
		    ordered->file_offset + ordered->len > start_pos &&
1388
		    ordered->file_offset <= last_pos) {
1389
			unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1390 1391
					     start_pos, last_pos,
					     cached_state, GFP_NOFS);
1392 1393 1394 1395
			for (i = 0; i < num_pages; i++) {
				unlock_page(pages[i]);
				page_cache_release(pages[i]);
			}
1396 1397 1398
			btrfs_start_ordered_extent(inode, ordered, 1);
			btrfs_put_ordered_extent(ordered);
			return -EAGAIN;
1399 1400 1401 1402
		}
		if (ordered)
			btrfs_put_ordered_extent(ordered);

1403
		clear_extent_bit(&BTRFS_I(inode)->io_tree, start_pos,
1404
				  last_pos, EXTENT_DIRTY | EXTENT_DELALLOC |
1405
				  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
1406 1407 1408 1409
				  0, 0, cached_state, GFP_NOFS);
		*lockstart = start_pos;
		*lockend = last_pos;
		ret = 1;
1410
	}
1411

1412
	for (i = 0; i < num_pages; i++) {
1413 1414
		if (clear_page_dirty_for_io(pages[i]))
			account_page_redirty(pages[i]);
1415 1416 1417
		set_page_extent_mapped(pages[i]);
		WARN_ON(!PageLocked(pages[i]));
	}
1418

1419
	return ret;
C
Chris Mason 已提交
1420 1421
}

1422 1423 1424 1425 1426 1427 1428 1429 1430
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;

1431
	ret = btrfs_start_write_no_snapshoting(root);
1432 1433 1434
	if (!ret)
		return -ENOSPC;

1435
	lockstart = round_down(pos, root->sectorsize);
1436
	lockend = round_up(pos + *write_bytes, root->sectorsize) - 1;
1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450

	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;
1451
	ret = can_nocow_extent(inode, lockstart, &num_bytes, NULL, NULL, NULL);
1452 1453
	if (ret <= 0) {
		ret = 0;
1454
		btrfs_end_write_no_snapshoting(root);
1455
	} else {
1456 1457
		*write_bytes = min_t(size_t, *write_bytes ,
				     num_bytes - pos + lockstart);
1458 1459 1460 1461 1462 1463 1464
	}

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

	return ret;
}

J
Josef Bacik 已提交
1465 1466 1467
static noinline ssize_t __btrfs_buffered_write(struct file *file,
					       struct iov_iter *i,
					       loff_t pos)
1468
{
A
Al Viro 已提交
1469
	struct inode *inode = file_inode(file);
1470 1471
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct page **pages = NULL;
1472
	struct extent_state *cached_state = NULL;
1473
	u64 release_bytes = 0;
1474 1475
	u64 lockstart;
	u64 lockend;
C
Chris Mason 已提交
1476
	unsigned long first_index;
J
Josef Bacik 已提交
1477 1478
	size_t num_written = 0;
	int nrptrs;
1479
	int ret = 0;
1480
	bool only_release_metadata = false;
1481
	bool force_page_uptodate = false;
1482
	bool need_unlock;
1483

1484 1485
	nrptrs = min(DIV_ROUND_UP(iov_iter_count(i), PAGE_CACHE_SIZE),
			PAGE_CACHE_SIZE / (sizeof(struct page *)));
1486 1487
	nrptrs = min(nrptrs, current->nr_dirtied_pause - current->nr_dirtied);
	nrptrs = max(nrptrs, 8);
1488
	pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL);
J
Josef Bacik 已提交
1489 1490
	if (!pages)
		return -ENOMEM;
1491

C
Chris Mason 已提交
1492 1493
	first_index = pos >> PAGE_CACHE_SHIFT;

J
Josef Bacik 已提交
1494
	while (iov_iter_count(i) > 0) {
C
Chris Mason 已提交
1495
		size_t offset = pos & (PAGE_CACHE_SIZE - 1);
J
Josef Bacik 已提交
1496
		size_t write_bytes = min(iov_iter_count(i),
1497
					 nrptrs * (size_t)PAGE_CACHE_SIZE -
1498
					 offset);
1499 1500
		size_t num_pages = DIV_ROUND_UP(write_bytes + offset,
						PAGE_CACHE_SIZE);
1501
		size_t reserve_bytes;
J
Josef Bacik 已提交
1502 1503
		size_t dirty_pages;
		size_t copied;
C
Chris Mason 已提交
1504

1505
		WARN_ON(num_pages > nrptrs);
1506

1507 1508 1509 1510
		/*
		 * Fault pages before locking them in prepare_pages
		 * to avoid recursive lock
		 */
J
Josef Bacik 已提交
1511
		if (unlikely(iov_iter_fault_in_readable(i, write_bytes))) {
1512
			ret = -EFAULT;
J
Josef Bacik 已提交
1513
			break;
1514 1515
		}

1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527
		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.
				 */
1528 1529
				num_pages = DIV_ROUND_UP(write_bytes + offset,
							 PAGE_CACHE_SIZE);
1530 1531 1532 1533 1534 1535 1536
				reserve_bytes = num_pages << PAGE_CACHE_SHIFT;
				ret = 0;
			} else {
				ret = -ENOSPC;
			}
		}

1537
		if (ret)
J
Josef Bacik 已提交
1538
			break;
1539

1540 1541 1542 1543 1544
		ret = btrfs_delalloc_reserve_metadata(inode, reserve_bytes);
		if (ret) {
			if (!only_release_metadata)
				btrfs_free_reserved_data_space(inode,
							       reserve_bytes);
1545
			else
1546
				btrfs_end_write_no_snapshoting(root);
1547 1548 1549 1550
			break;
		}

		release_bytes = reserve_bytes;
1551 1552
		need_unlock = false;
again:
1553 1554 1555 1556 1557
		/*
		 * 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
		 */
1558 1559
		ret = prepare_pages(inode, pages, num_pages,
				    pos, write_bytes,
1560
				    force_page_uptodate);
1561
		if (ret)
J
Josef Bacik 已提交
1562
			break;
C
Chris Mason 已提交
1563

1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575
		ret = lock_and_cleanup_extent_if_need(inode, pages, num_pages,
						      pos, &lockstart, &lockend,
						      &cached_state);
		if (ret < 0) {
			if (ret == -EAGAIN)
				goto again;
			break;
		} else if (ret > 0) {
			need_unlock = true;
			ret = 0;
		}

1576
		copied = btrfs_copy_from_user(pos, num_pages,
J
Josef Bacik 已提交
1577
					   write_bytes, pages, i);
1578 1579 1580 1581 1582 1583 1584 1585

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

1586 1587
		if (copied == 0) {
			force_page_uptodate = true;
1588
			dirty_pages = 0;
1589 1590
		} else {
			force_page_uptodate = false;
1591 1592
			dirty_pages = DIV_ROUND_UP(copied + offset,
						   PAGE_CACHE_SIZE);
1593
		}
1594

J
Josef Bacik 已提交
1595 1596 1597 1598 1599 1600 1601
		/*
		 * 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.
		 */
1602
		if (num_pages > dirty_pages) {
1603 1604
			release_bytes = (num_pages - dirty_pages) <<
				PAGE_CACHE_SHIFT;
1605 1606 1607 1608 1609
			if (copied > 0) {
				spin_lock(&BTRFS_I(inode)->lock);
				BTRFS_I(inode)->outstanding_extents++;
				spin_unlock(&BTRFS_I(inode)->lock);
			}
1610 1611 1612 1613 1614 1615
			if (only_release_metadata)
				btrfs_delalloc_release_metadata(inode,
								release_bytes);
			else
				btrfs_delalloc_release_space(inode,
							     release_bytes);
1616 1617
		}

1618
		release_bytes = dirty_pages << PAGE_CACHE_SHIFT;
1619 1620

		if (copied > 0)
1621 1622 1623
			ret = btrfs_dirty_pages(root, inode, pages,
						dirty_pages, pos, copied,
						NULL);
1624 1625 1626 1627
		if (need_unlock)
			unlock_extent_cached(&BTRFS_I(inode)->io_tree,
					     lockstart, lockend, &cached_state,
					     GFP_NOFS);
1628 1629
		if (ret) {
			btrfs_drop_pages(pages, num_pages);
1630
			break;
1631
		}
C
Chris Mason 已提交
1632

1633
		release_bytes = 0;
1634
		if (only_release_metadata)
1635
			btrfs_end_write_no_snapshoting(root);
1636

1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647
		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;
		}

1648 1649
		btrfs_drop_pages(pages, num_pages);

J
Josef Bacik 已提交
1650 1651
		cond_resched();

1652
		balance_dirty_pages_ratelimited(inode->i_mapping);
1653
		if (dirty_pages < (root->nodesize >> PAGE_CACHE_SHIFT) + 1)
1654
			btrfs_btree_balance_dirty(root);
1655

1656 1657
		pos += copied;
		num_written += copied;
J
Josef Bacik 已提交
1658
	}
C
Chris Mason 已提交
1659

J
Josef Bacik 已提交
1660 1661
	kfree(pages);

1662
	if (release_bytes) {
1663
		if (only_release_metadata) {
1664
			btrfs_end_write_no_snapshoting(root);
1665
			btrfs_delalloc_release_metadata(inode, release_bytes);
1666
		} else {
1667
			btrfs_delalloc_release_space(inode, release_bytes);
1668
		}
1669 1670
	}

J
Josef Bacik 已提交
1671 1672 1673 1674
	return num_written ? num_written : ret;
}

static ssize_t __btrfs_direct_write(struct kiocb *iocb,
1675
				    struct iov_iter *from,
A
Al Viro 已提交
1676
				    loff_t pos)
J
Josef Bacik 已提交
1677 1678
{
	struct file *file = iocb->ki_filp;
1679
	struct inode *inode = file_inode(file);
J
Josef Bacik 已提交
1680 1681 1682 1683 1684
	ssize_t written;
	ssize_t written_buffered;
	loff_t endbyte;
	int err;

A
Al Viro 已提交
1685
	written = generic_file_direct_write(iocb, from, pos);
J
Josef Bacik 已提交
1686

A
Al Viro 已提交
1687
	if (written < 0 || !iov_iter_count(from))
J
Josef Bacik 已提交
1688 1689 1690
		return written;

	pos += written;
1691
	written_buffered = __btrfs_buffered_write(file, from, pos);
J
Josef Bacik 已提交
1692 1693 1694
	if (written_buffered < 0) {
		err = written_buffered;
		goto out;
C
Chris Mason 已提交
1695
	}
1696 1697 1698 1699
	/*
	 * Ensure all data is persisted. We want the next direct IO read to be
	 * able to read what was just written.
	 */
J
Josef Bacik 已提交
1700
	endbyte = pos + written_buffered - 1;
1701
	err = btrfs_fdatawrite_range(inode, pos, endbyte);
1702 1703
	if (err)
		goto out;
1704
	err = filemap_fdatawait_range(inode->i_mapping, pos, endbyte);
J
Josef Bacik 已提交
1705 1706 1707
	if (err)
		goto out;
	written += written_buffered;
1708
	iocb->ki_pos = pos + written_buffered;
J
Josef Bacik 已提交
1709 1710
	invalidate_mapping_pages(file->f_mapping, pos >> PAGE_CACHE_SHIFT,
				 endbyte >> PAGE_CACHE_SHIFT);
C
Chris Mason 已提交
1711
out:
J
Josef Bacik 已提交
1712 1713
	return written ? written : err;
}
1714

1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732
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);
}

A
Al Viro 已提交
1733 1734
static ssize_t btrfs_file_write_iter(struct kiocb *iocb,
				    struct iov_iter *from)
J
Josef Bacik 已提交
1735 1736
{
	struct file *file = iocb->ki_filp;
A
Al Viro 已提交
1737
	struct inode *inode = file_inode(file);
J
Josef Bacik 已提交
1738
	struct btrfs_root *root = BTRFS_I(inode)->root;
1739
	u64 start_pos;
1740
	u64 end_pos;
J
Josef Bacik 已提交
1741 1742
	ssize_t num_written = 0;
	ssize_t err = 0;
A
Al Viro 已提交
1743
	size_t count = iov_iter_count(from);
1744
	bool sync = (file->f_flags & O_DSYNC) || IS_SYNC(file->f_mapping->host);
A
Al Viro 已提交
1745
	loff_t pos = iocb->ki_pos;
J
Josef Bacik 已提交
1746 1747 1748

	mutex_lock(&inode->i_mutex);

1749
	current->backing_dev_info = inode_to_bdi(inode);
1750
	err = generic_write_checks(file, &pos, &count);
J
Josef Bacik 已提交
1751 1752 1753 1754 1755 1756 1757 1758 1759 1760
	if (err) {
		mutex_unlock(&inode->i_mutex);
		goto out;
	}

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

A
Al Viro 已提交
1761
	iov_iter_truncate(from, count);
1762

J
Josef Bacik 已提交
1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774
	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.
	 */
1775
	if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
J
Josef Bacik 已提交
1776 1777 1778 1779 1780
		mutex_unlock(&inode->i_mutex);
		err = -EROFS;
		goto out;
	}

1781 1782 1783 1784 1785 1786 1787
	/*
	 * 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
1788

1789 1790
	start_pos = round_down(pos, root->sectorsize);
	if (start_pos > i_size_read(inode)) {
1791
		/* Expand hole size to cover write data, preventing empty gap */
1792
		end_pos = round_up(pos + count, root->sectorsize);
1793
		err = btrfs_cont_expand(inode, i_size_read(inode), end_pos);
1794 1795 1796 1797 1798 1799
		if (err) {
			mutex_unlock(&inode->i_mutex);
			goto out;
		}
	}

1800 1801 1802
	if (sync)
		atomic_inc(&BTRFS_I(inode)->sync_writers);

1803
	if (file->f_flags & O_DIRECT) {
A
Al Viro 已提交
1804
		num_written = __btrfs_direct_write(iocb, from, pos);
J
Josef Bacik 已提交
1805
	} else {
A
Al Viro 已提交
1806
		num_written = __btrfs_buffered_write(file, from, pos);
J
Josef Bacik 已提交
1807
		if (num_written > 0)
1808
			iocb->ki_pos = pos + num_written;
J
Josef Bacik 已提交
1809 1810 1811
	}

	mutex_unlock(&inode->i_mutex);
1812

1813
	/*
1814 1815 1816
	 * 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.
1817
	 */
1818
	BTRFS_I(inode)->last_sub_trans = root->log_transid;
1819
	if (num_written > 0) {
J
Josef Bacik 已提交
1820
		err = generic_write_sync(file, pos, num_written);
1821
		if (err < 0)
1822 1823
			num_written = err;
	}
1824

1825 1826
	if (sync)
		atomic_dec(&BTRFS_I(inode)->sync_writers);
1827
out:
C
Chris Mason 已提交
1828 1829 1830 1831
	current->backing_dev_info = NULL;
	return num_written ? num_written : err;
}

C
Chris Mason 已提交
1832
int btrfs_release_file(struct inode *inode, struct file *filp)
1833
{
S
Sage Weil 已提交
1834 1835
	if (filp->private_data)
		btrfs_ioctl_trans_end(filp);
1836 1837 1838 1839 1840 1841 1842 1843 1844
	/*
	 * 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.
	 */
	if (test_and_clear_bit(BTRFS_INODE_ORDERED_DATA_CLOSE,
			       &BTRFS_I(inode)->runtime_flags))
			filemap_flush(inode->i_mapping);
1845 1846 1847
	return 0;
}

1848 1849 1850 1851 1852
static int start_ordered_ops(struct inode *inode, loff_t start, loff_t end)
{
	int ret;

	atomic_inc(&BTRFS_I(inode)->sync_writers);
1853
	ret = btrfs_fdatawrite_range(inode, start, end);
1854 1855 1856 1857 1858
	atomic_dec(&BTRFS_I(inode)->sync_writers);

	return ret;
}

C
Chris Mason 已提交
1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869
/*
 * 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.
 */
1870
int btrfs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
C
Chris Mason 已提交
1871
{
1872
	struct dentry *dentry = file->f_path.dentry;
C
Chris Mason 已提交
1873 1874 1875
	struct inode *inode = dentry->d_inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
1876 1877
	struct btrfs_log_ctx ctx;
	int ret = 0;
1878
	bool full_sync = 0;
C
Chris Mason 已提交
1879

1880
	trace_btrfs_sync_file(file, datasync);
1881

1882 1883 1884
	/*
	 * 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
1885 1886
	 * multi-task, and make the performance up.  See
	 * btrfs_wait_ordered_range for an explanation of the ASYNC check.
1887
	 */
1888
	ret = start_ordered_ops(inode, start, end);
1889 1890 1891
	if (ret)
		return ret;

1892
	mutex_lock(&inode->i_mutex);
M
Miao Xie 已提交
1893
	atomic_inc(&root->log_batch);
1894 1895
	full_sync = test_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
			     &BTRFS_I(inode)->runtime_flags);
1896 1897 1898 1899
	/*
	 * We might have have had more pages made dirty after calling
	 * start_ordered_ops and before acquiring the inode's i_mutex.
	 */
1900
	if (full_sync) {
1901 1902 1903 1904 1905 1906
		/*
		 * For a full sync, we need to make sure any ordered operations
		 * start and finish before we start logging the inode, so that
		 * all extents are persisted and the respective file extent
		 * items are in the fs/subvol btree.
		 */
1907
		ret = btrfs_wait_ordered_range(inode, start, end - start + 1);
1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945
	} else {
		/*
		 * Start any new ordered operations before starting to log the
		 * inode. We will wait for them to finish in btrfs_sync_log().
		 *
		 * Right before acquiring the inode's mutex, we might have new
		 * writes dirtying pages, which won't immediately start the
		 * respective ordered operations - that is done through the
		 * fill_delalloc callbacks invoked from the writepage and
		 * writepages address space operations. So make sure we start
		 * all ordered operations before starting to log our inode. Not
		 * doing this means that while logging the inode, writeback
		 * could start and invoke writepage/writepages, which would call
		 * the fill_delalloc callbacks (cow_file_range,
		 * submit_compressed_extents). These callbacks add first an
		 * extent map to the modified list of extents and then create
		 * the respective ordered operation, which means in
		 * tree-log.c:btrfs_log_inode() we might capture all existing
		 * ordered operations (with btrfs_get_logged_extents()) before
		 * the fill_delalloc callback adds its ordered operation, and by
		 * the time we visit the modified list of extent maps (with
		 * btrfs_log_changed_extents()), we see and process the extent
		 * map they created. We then use the extent map to construct a
		 * file extent item for logging without waiting for the
		 * respective ordered operation to finish - this file extent
		 * item points to a disk location that might not have yet been
		 * written to, containing random data - so after a crash a log
		 * replay will make our inode have file extent items that point
		 * to disk locations containing invalid data, as we returned
		 * success to userspace without waiting for the respective
		 * ordered operation to finish, because it wasn't captured by
		 * btrfs_get_logged_extents().
		 */
		ret = start_ordered_ops(inode, start, end);
	}
	if (ret) {
		mutex_unlock(&inode->i_mutex);
		goto out;
1946
	}
M
Miao Xie 已提交
1947
	atomic_inc(&root->log_batch);
1948

C
Chris Mason 已提交
1949
	/*
1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975
	 * If the last transaction that changed this file was before the current
	 * transaction and we have the full sync flag set in our inode, we can
	 * bail out now without any syncing.
	 *
	 * Note that we can't bail out if the full sync flag isn't set. This is
	 * because when the full sync flag is set we start all ordered extents
	 * and wait for them to fully complete - when they complete they update
	 * the inode's last_trans field through:
	 *
	 *     btrfs_finish_ordered_io() ->
	 *         btrfs_update_inode_fallback() ->
	 *             btrfs_update_inode() ->
	 *                 btrfs_set_inode_last_trans()
	 *
	 * So we are sure that last_trans is up to date and can do this check to
	 * bail out safely. For the fast path, when the full sync flag is not
	 * set in our inode, we can not do it because we start only our ordered
	 * extents and don't wait for them to complete (that is when
	 * btrfs_finish_ordered_io runs), so here at this point their last_trans
	 * value might be less than or equals to fs_info->last_trans_committed,
	 * and setting a speculative last_trans for an inode when a buffered
	 * write is made (such as fs_info->generation + 1 for example) would not
	 * be reliable since after setting the value and before fsync is called
	 * any number of transactions can start and commit (transaction kthread
	 * commits the current transaction periodically), and a transaction
	 * commit does not start nor waits for ordered extents to complete.
1976
	 */
J
Josef Bacik 已提交
1977
	smp_mb();
1978
	if (btrfs_inode_in_log(inode, root->fs_info->generation) ||
1979 1980
	    (full_sync && BTRFS_I(inode)->last_trans <=
	     root->fs_info->last_trans_committed)) {
J
Josef Bacik 已提交
1981 1982 1983 1984 1985 1986 1987
		/*
		 * 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);
1988
		mutex_unlock(&inode->i_mutex);
1989 1990 1991 1992
		goto out;
	}

	/*
1993 1994
	 * ok we haven't committed the transaction yet, lets do a commit
	 */
1995
	if (file->private_data)
S
Sage Weil 已提交
1996 1997
		btrfs_ioctl_trans_end(file);

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
	/*
	 * We use start here because we will need to wait on the IO to complete
	 * in btrfs_sync_log, which could require joining a transaction (for
	 * example checking cross references in the nocow path).  If we use join
	 * here we could get into a situation where we're waiting on IO to
	 * happen that is blocked on a transaction trying to commit.  With start
	 * we inc the extwriter counter, so we wait for all extwriters to exit
	 * before we start blocking join'ers.  This comment is to keep somebody
	 * from thinking they are super smart and changing this to
	 * btrfs_join_transaction *cough*Josef*cough*.
	 */
2009 2010 2011
	trans = btrfs_start_transaction(root, 0);
	if (IS_ERR(trans)) {
		ret = PTR_ERR(trans);
2012
		mutex_unlock(&inode->i_mutex);
C
Chris Mason 已提交
2013 2014
		goto out;
	}
2015
	trans->sync = true;
2016

2017 2018
	btrfs_init_log_ctx(&ctx);

2019
	ret = btrfs_log_dentry_safe(trans, root, dentry, start, end, &ctx);
2020
	if (ret < 0) {
2021 2022
		/* Fallthrough and commit/free transaction. */
		ret = 1;
2023
	}
C
Chris Mason 已提交
2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034

	/* 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.
	 */
2035
	mutex_unlock(&inode->i_mutex);
C
Chris Mason 已提交
2036

2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055
	/*
	 * If any of the ordered extents had an error, just return it to user
	 * space, so that the application knows some writes didn't succeed and
	 * can take proper action (retry for e.g.). Blindly committing the
	 * transaction in this case, would fool userspace that everything was
	 * successful. And we also want to make sure our log doesn't contain
	 * file extent items pointing to extents that weren't fully written to -
	 * just like in the non fast fsync path, where we check for the ordered
	 * operation's error flag before writing to the log tree and return -EIO
	 * if any of them had this flag set (btrfs_wait_ordered_range) -
	 * therefore we need to check for errors in the ordered operations,
	 * which are indicated by ctx.io_err.
	 */
	if (ctx.io_err) {
		btrfs_end_transaction(trans, root);
		ret = ctx.io_err;
		goto out;
	}

2056
	if (ret != BTRFS_NO_LOG_SYNC) {
2057
		if (!ret) {
2058
			ret = btrfs_sync_log(trans, root, &ctx);
2059
			if (!ret) {
2060
				ret = btrfs_end_transaction(trans, root);
2061
				goto out;
2062
			}
2063
		}
2064 2065 2066
		if (!full_sync) {
			ret = btrfs_wait_ordered_range(inode, start,
						       end - start + 1);
2067 2068
			if (ret) {
				btrfs_end_transaction(trans, root);
2069
				goto out;
2070
			}
2071 2072
		}
		ret = btrfs_commit_transaction(trans, root);
2073 2074
	} else {
		ret = btrfs_end_transaction(trans, root);
2075
	}
C
Chris Mason 已提交
2076
out:
2077
	return ret > 0 ? -EIO : ret;
C
Chris Mason 已提交
2078 2079
}

2080
static const struct vm_operations_struct btrfs_file_vm_ops = {
2081
	.fault		= filemap_fault,
2082
	.map_pages	= filemap_map_pages,
C
Chris Mason 已提交
2083 2084 2085 2086 2087
	.page_mkwrite	= btrfs_page_mkwrite,
};

static int btrfs_file_mmap(struct file	*filp, struct vm_area_struct *vma)
{
M
Miao Xie 已提交
2088 2089 2090 2091 2092
	struct address_space *mapping = filp->f_mapping;

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

C
Chris Mason 已提交
2093
	file_accessed(filp);
M
Miao Xie 已提交
2094 2095
	vma->vm_ops = &btrfs_file_vm_ops;

C
Chris Mason 已提交
2096 2097 2098
	return 0;
}

J
Josef Bacik 已提交
2099 2100 2101 2102 2103 2104 2105 2106 2107 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
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;

2139 2140 2141
	if (btrfs_fs_incompat(root->fs_info, NO_HOLES))
		goto out;

J
Josef Bacik 已提交
2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166
	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;
	}

2167
	if (hole_mergeable(inode, leaf, path->slots[0], offset, end)) {
J
Josef Bacik 已提交
2168 2169 2170
		u64 num_bytes;

		key.offset = offset;
2171
		btrfs_set_item_key_safe(root, path, &key);
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
		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
2201
		hole_em->ram_bytes = hole_em->len;
J
Josef Bacik 已提交
2202 2203 2204 2205
		hole_em->orig_start = offset;

		hole_em->block_start = EXTENT_MAP_HOLE;
		hole_em->block_len = 0;
2206
		hole_em->orig_block_len = 0;
J
Josef Bacik 已提交
2207 2208 2209 2210 2211 2212 2213
		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
2214
			ret = add_extent_mapping(em_tree, hole_em, 1);
J
Josef Bacik 已提交
2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225
			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;
}

2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256
/*
 * Find a hole extent on given inode and change start/len to the end of hole
 * extent.(hole/vacuum extent whose em->start <= start &&
 *	   em->start + em->len > start)
 * When a hole extent is found, return 1 and modify start/len.
 */
static int find_first_non_hole(struct inode *inode, u64 *start, u64 *len)
{
	struct extent_map *em;
	int ret = 0;

	em = btrfs_get_extent(inode, NULL, 0, *start, *len, 0);
	if (IS_ERR_OR_NULL(em)) {
		if (!em)
			ret = -ENOMEM;
		else
			ret = PTR_ERR(em);
		return ret;
	}

	/* Hole or vacuum extent(only exists in no-hole mode) */
	if (em->block_start == EXTENT_MAP_HOLE) {
		ret = 1;
		*len = em->start + em->len > *start + *len ?
		       0 : *start + *len - em->start - em->len;
		*start = em->start + em->len;
	}
	free_extent_map(em);
	return ret;
}

J
Josef Bacik 已提交
2257 2258 2259 2260 2261 2262 2263
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;
2264 2265 2266 2267 2268 2269
	u64 lockstart;
	u64 lockend;
	u64 tail_start;
	u64 tail_len;
	u64 orig_start = offset;
	u64 cur_offset;
J
Josef Bacik 已提交
2270 2271 2272 2273
	u64 min_size = btrfs_calc_trunc_metadata_size(root, 1);
	u64 drop_end;
	int ret = 0;
	int err = 0;
2274
	int rsv_count;
2275
	bool same_page;
2276
	bool no_holes = btrfs_fs_incompat(root->fs_info, NO_HOLES);
2277
	u64 ino_size;
2278 2279
	bool truncated_page = false;
	bool updated_inode = false;
J
Josef Bacik 已提交
2280

2281 2282 2283
	ret = btrfs_wait_ordered_range(inode, offset, len);
	if (ret)
		return ret;
J
Josef Bacik 已提交
2284 2285

	mutex_lock(&inode->i_mutex);
2286
	ino_size = round_up(inode->i_size, PAGE_CACHE_SIZE);
2287 2288 2289 2290 2291 2292 2293 2294 2295
	ret = find_first_non_hole(inode, &offset, &len);
	if (ret < 0)
		goto out_only_mutex;
	if (ret && !len) {
		/* Already in a large hole */
		ret = 0;
		goto out_only_mutex;
	}

2296
	lockstart = round_up(offset, BTRFS_I(inode)->root->sectorsize);
2297 2298 2299 2300 2301
	lockend = round_down(offset + len,
			     BTRFS_I(inode)->root->sectorsize) - 1;
	same_page = ((offset >> PAGE_CACHE_SHIFT) ==
		    ((offset + len - 1) >> PAGE_CACHE_SHIFT));

2302 2303 2304 2305
	/*
	 * 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
2306 2307 2308 2309 2310
	/*
	 * 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) {
2311 2312
		if (offset < ino_size) {
			truncated_page = true;
2313
			ret = btrfs_truncate_page(inode, offset, len, 0);
2314 2315 2316
		} else {
			ret = 0;
		}
2317
		goto out_only_mutex;
J
Josef Bacik 已提交
2318 2319 2320
	}

	/* zero back part of the first page */
2321
	if (offset < ino_size) {
2322
		truncated_page = true;
2323 2324 2325 2326 2327
		ret = btrfs_truncate_page(inode, offset, 0, 0);
		if (ret) {
			mutex_unlock(&inode->i_mutex);
			return ret;
		}
J
Josef Bacik 已提交
2328 2329
	}

2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357
	/* Check the aligned pages after the first unaligned page,
	 * if offset != orig_start, which means the first unaligned page
	 * including serveral following pages are already in holes,
	 * the extra check can be skipped */
	if (offset == orig_start) {
		/* after truncate page, check hole again */
		len = offset + len - lockstart;
		offset = lockstart;
		ret = find_first_non_hole(inode, &offset, &len);
		if (ret < 0)
			goto out_only_mutex;
		if (ret && !len) {
			ret = 0;
			goto out_only_mutex;
		}
		lockstart = offset;
	}

	/* Check the tail unaligned part is in a hole */
	tail_start = lockend + 1;
	tail_len = offset + len - tail_start;
	if (tail_len) {
		ret = find_first_non_hole(inode, &tail_start, &tail_len);
		if (unlikely(ret < 0))
			goto out_only_mutex;
		if (!ret) {
			/* zero the front end of the last page */
			if (tail_start + tail_len < ino_size) {
2358
				truncated_page = true;
2359 2360 2361 2362
				ret = btrfs_truncate_page(inode,
						tail_start + tail_len, 0, 1);
				if (ret)
					goto out_only_mutex;
2363
			}
M
Miao Xie 已提交
2364
		}
J
Josef Bacik 已提交
2365 2366 2367
	}

	if (lockend < lockstart) {
2368 2369
		ret = 0;
		goto out_only_mutex;
J
Josef Bacik 已提交
2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386
	}

	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 ||
2387
		    (ordered->file_offset + ordered->len <= lockstart ||
J
Josef Bacik 已提交
2388
		     ordered->file_offset > lockend)) &&
2389
		     !btrfs_page_exists_in_range(inode, lockstart, lockend)) {
J
Josef Bacik 已提交
2390 2391 2392 2393 2394 2395 2396 2397
			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);
2398 2399 2400 2401 2402 2403
		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
2404 2405 2406 2407 2408 2409 2410 2411
	}

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

2412
	rsv = btrfs_alloc_block_rsv(root, BTRFS_BLOCK_RSV_TEMP);
J
Josef Bacik 已提交
2413 2414 2415 2416 2417 2418 2419 2420 2421 2422
	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
2423
	 * 1 - adding the hole extent if no_holes isn't set
J
Josef Bacik 已提交
2424
	 */
2425 2426
	rsv_count = no_holes ? 2 : 3;
	trans = btrfs_start_transaction(root, rsv_count);
J
Josef Bacik 已提交
2427 2428 2429 2430 2431 2432 2433 2434 2435 2436
	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;

2437 2438
	cur_offset = lockstart;
	len = lockend - cur_offset;
J
Josef Bacik 已提交
2439 2440 2441
	while (cur_offset < lockend) {
		ret = __btrfs_drop_extents(trans, root, inode, path,
					   cur_offset, lockend + 1,
2442
					   &drop_end, 1, 0, 0, NULL);
J
Josef Bacik 已提交
2443 2444 2445 2446 2447
		if (ret != -ENOSPC)
			break;

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

2448 2449 2450 2451 2452 2453 2454
		if (cur_offset < ino_size) {
			ret = fill_holes(trans, inode, path, cur_offset,
					 drop_end);
			if (ret) {
				err = ret;
				break;
			}
J
Josef Bacik 已提交
2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465
		}

		cur_offset = drop_end;

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

		btrfs_end_transaction(trans, root);
2466
		btrfs_btree_balance_dirty(root);
J
Josef Bacik 已提交
2467

2468
		trans = btrfs_start_transaction(root, rsv_count);
J
Josef Bacik 已提交
2469 2470 2471 2472 2473 2474 2475 2476 2477 2478
		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;
2479 2480 2481 2482 2483 2484 2485 2486

		ret = find_first_non_hole(inode, &cur_offset, &len);
		if (unlikely(ret < 0))
			break;
		if (ret && !len) {
			ret = 0;
			break;
		}
J
Josef Bacik 已提交
2487 2488 2489 2490 2491 2492 2493 2494
	}

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

	trans->block_rsv = &root->fs_info->trans_block_rsv;
2495 2496 2497 2498 2499 2500
	/*
	 * Don't insert file hole extent item if it's for a range beyond eof
	 * (because it's useless) or if it represents a 0 bytes range (when
	 * cur_offset == drop_end).
	 */
	if (cur_offset < ino_size && cur_offset < drop_end) {
2501 2502 2503 2504 2505
		ret = fill_holes(trans, inode, path, cur_offset, drop_end);
		if (ret) {
			err = ret;
			goto out_trans;
		}
J
Josef Bacik 已提交
2506 2507 2508 2509 2510 2511
	}

out_trans:
	if (!trans)
		goto out_free;

2512 2513 2514
	inode_inc_iversion(inode);
	inode->i_mtime = inode->i_ctime = CURRENT_TIME;

J
Josef Bacik 已提交
2515 2516
	trans->block_rsv = &root->fs_info->trans_block_rsv;
	ret = btrfs_update_inode(trans, root, inode);
2517
	updated_inode = true;
J
Josef Bacik 已提交
2518
	btrfs_end_transaction(trans, root);
2519
	btrfs_btree_balance_dirty(root);
J
Josef Bacik 已提交
2520 2521 2522 2523 2524 2525
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);
2526
out_only_mutex:
2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542
	if (!updated_inode && truncated_page && !ret && !err) {
		/*
		 * If we only end up zeroing part of a page, we still need to
		 * update the inode item, so that all the time fields are
		 * updated as well as the necessary btrfs inode in memory fields
		 * for detecting, at fsync time, if the inode isn't yet in the
		 * log tree or it's there but not up to date.
		 */
		trans = btrfs_start_transaction(root, 1);
		if (IS_ERR(trans)) {
			err = PTR_ERR(trans);
		} else {
			err = btrfs_update_inode(trans, root, inode);
			ret = btrfs_end_transaction(trans, root);
		}
	}
J
Josef Bacik 已提交
2543 2544 2545 2546 2547 2548
	mutex_unlock(&inode->i_mutex);
	if (ret && !err)
		err = ret;
	return err;
}

2549 2550 2551
static long btrfs_fallocate(struct file *file, int mode,
			    loff_t offset, loff_t len)
{
A
Al Viro 已提交
2552
	struct inode *inode = file_inode(file);
2553
	struct extent_state *cached_state = NULL;
2554
	struct btrfs_root *root = BTRFS_I(inode)->root;
2555 2556 2557 2558 2559 2560 2561
	u64 cur_offset;
	u64 last_byte;
	u64 alloc_start;
	u64 alloc_end;
	u64 alloc_hint = 0;
	u64 locked_end;
	struct extent_map *em;
2562
	int blocksize = BTRFS_I(inode)->root->sectorsize;
2563 2564
	int ret;

2565 2566
	alloc_start = round_down(offset, blocksize);
	alloc_end = round_up(offset + len, blocksize);
2567

J
Josef Bacik 已提交
2568 2569
	/* Make sure we aren't being give some crap mode */
	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
2570 2571
		return -EOPNOTSUPP;

J
Josef Bacik 已提交
2572 2573 2574
	if (mode & FALLOC_FL_PUNCH_HOLE)
		return btrfs_punch_hole(inode, offset, len);

2575 2576 2577 2578
	/*
	 * Make sure we have enough space before we do the
	 * allocation.
	 */
2579
	ret = btrfs_check_data_free_space(inode, alloc_end - alloc_start);
2580 2581
	if (ret)
		return ret;
2582 2583 2584 2585 2586
	if (root->fs_info->quota_enabled) {
		ret = btrfs_qgroup_reserve(root, alloc_end - alloc_start);
		if (ret)
			goto out_reserve_fail;
	}
2587

2588 2589 2590 2591 2592 2593
	mutex_lock(&inode->i_mutex);
	ret = inode_newsize_ok(inode, alloc_end);
	if (ret)
		goto out;

	if (alloc_start > inode->i_size) {
2594 2595
		ret = btrfs_cont_expand(inode, i_size_read(inode),
					alloc_start);
2596 2597
		if (ret)
			goto out;
2598 2599 2600 2601 2602 2603 2604 2605 2606
	} 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;
2607 2608
	}

2609 2610 2611 2612
	/*
	 * wait for ordered IO before we have any locks.  We'll loop again
	 * below with the locks held.
	 */
2613 2614 2615 2616
	ret = btrfs_wait_ordered_range(inode, alloc_start,
				       alloc_end - alloc_start);
	if (ret)
		goto out;
2617

2618 2619 2620 2621 2622 2623 2624 2625
	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,
2626
				 locked_end, 0, &cached_state);
2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639
		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
			 */
2640 2641 2642 2643
			ret = btrfs_wait_ordered_range(inode, alloc_start,
						       alloc_end - alloc_start);
			if (ret)
				goto out;
2644 2645 2646 2647 2648 2649 2650 2651 2652
		} else {
			if (ordered)
				btrfs_put_ordered_extent(ordered);
			break;
		}
	}

	cur_offset = alloc_start;
	while (1) {
2653 2654
		u64 actual_end;

2655 2656
		em = btrfs_get_extent(inode, NULL, 0, cur_offset,
				      alloc_end - cur_offset, 0);
2657 2658 2659 2660 2661 2662 2663
		if (IS_ERR_OR_NULL(em)) {
			if (!em)
				ret = -ENOMEM;
			else
				ret = PTR_ERR(em);
			break;
		}
2664
		last_byte = min(extent_map_end(em), alloc_end);
2665
		actual_end = min_t(u64, extent_map_end(em), offset + len);
2666
		last_byte = ALIGN(last_byte, blocksize);
2667

2668 2669 2670 2671 2672 2673 2674 2675
		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);
2676

2677 2678 2679 2680
			if (ret < 0) {
				free_extent_map(em);
				break;
			}
2681 2682 2683 2684 2685 2686 2687 2688 2689 2690
		} 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);
2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703
		}
		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);
2704 2705 2706
	if (root->fs_info->quota_enabled)
		btrfs_qgroup_free(root, alloc_end - alloc_start);
out_reserve_fail:
2707
	/* Let go of our reservation. */
2708
	btrfs_free_reserved_data_space(inode, alloc_end - alloc_start);
2709 2710 2711
	return ret;
}

2712
static int find_desired_extent(struct inode *inode, loff_t *offset, int whence)
J
Josef Bacik 已提交
2713 2714
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
2715
	struct extent_map *em = NULL;
J
Josef Bacik 已提交
2716
	struct extent_state *cached_state = NULL;
L
Liu Bo 已提交
2717 2718 2719 2720
	u64 lockstart;
	u64 lockend;
	u64 start;
	u64 len;
J
Josef Bacik 已提交
2721 2722
	int ret = 0;

L
Liu Bo 已提交
2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733
	if (inode->i_size == 0)
		return -ENXIO;

	/*
	 * *offset can be negative, in this case we start finding DATA/HOLE from
	 * the very start of the file.
	 */
	start = max_t(loff_t, 0, *offset);

	lockstart = round_down(start, root->sectorsize);
	lockend = round_up(i_size_read(inode), root->sectorsize);
J
Josef Bacik 已提交
2734 2735
	if (lockend <= lockstart)
		lockend = lockstart + root->sectorsize;
L
Liu Bo 已提交
2736
	lockend--;
J
Josef Bacik 已提交
2737 2738 2739
	len = lockend - lockstart + 1;

	lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend, 0,
2740
			 &cached_state);
J
Josef Bacik 已提交
2741

2742
	while (start < inode->i_size) {
J
Josef Bacik 已提交
2743 2744
		em = btrfs_get_extent_fiemap(inode, NULL, 0, start, len, 0);
		if (IS_ERR(em)) {
2745
			ret = PTR_ERR(em);
2746
			em = NULL;
J
Josef Bacik 已提交
2747 2748 2749
			break;
		}

2750 2751 2752 2753 2754 2755 2756 2757
		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
2758 2759 2760

		start = em->start + em->len;
		free_extent_map(em);
2761
		em = NULL;
J
Josef Bacik 已提交
2762 2763
		cond_resched();
	}
2764 2765 2766 2767 2768 2769 2770
	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
2771 2772 2773 2774 2775
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
			     &cached_state, GFP_NOFS);
	return ret;
}

2776
static loff_t btrfs_file_llseek(struct file *file, loff_t offset, int whence)
J
Josef Bacik 已提交
2777 2778 2779 2780 2781
{
	struct inode *inode = file->f_mapping->host;
	int ret;

	mutex_lock(&inode->i_mutex);
2782
	switch (whence) {
J
Josef Bacik 已提交
2783 2784
	case SEEK_END:
	case SEEK_CUR:
2785
		offset = generic_file_llseek(file, offset, whence);
J
Josef Bacik 已提交
2786 2787 2788
		goto out;
	case SEEK_DATA:
	case SEEK_HOLE:
2789 2790 2791 2792 2793
		if (offset >= i_size_read(inode)) {
			mutex_unlock(&inode->i_mutex);
			return -ENXIO;
		}

2794
		ret = find_desired_extent(inode, &offset, whence);
J
Josef Bacik 已提交
2795 2796 2797 2798 2799 2800
		if (ret) {
			mutex_unlock(&inode->i_mutex);
			return ret;
		}
	}

J
Jie Liu 已提交
2801
	offset = vfs_setpos(file, offset, inode->i_sb->s_maxbytes);
J
Josef Bacik 已提交
2802 2803 2804 2805 2806
out:
	mutex_unlock(&inode->i_mutex);
	return offset;
}

2807
const struct file_operations btrfs_file_operations = {
J
Josef Bacik 已提交
2808
	.llseek		= btrfs_file_llseek,
2809
	.read_iter      = generic_file_read_iter,
C
Chris Mason 已提交
2810
	.splice_read	= generic_file_splice_read,
A
Al Viro 已提交
2811
	.write_iter	= btrfs_file_write_iter,
C
Chris Mason 已提交
2812
	.mmap		= btrfs_file_mmap,
C
Chris Mason 已提交
2813
	.open		= generic_file_open,
2814
	.release	= btrfs_release_file,
C
Chris Mason 已提交
2815
	.fsync		= btrfs_sync_file,
2816
	.fallocate	= btrfs_fallocate,
C
Christoph Hellwig 已提交
2817
	.unlocked_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
2818
#ifdef CONFIG_COMPAT
C
Christoph Hellwig 已提交
2819
	.compat_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
2820 2821
#endif
};
2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839

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;
}
2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865

int btrfs_fdatawrite_range(struct inode *inode, loff_t start, loff_t end)
{
	int ret;

	/*
	 * So with compression we will find and lock a dirty page and clear the
	 * first one as dirty, setup an async extent, and immediately return
	 * with the entire range locked but with nobody actually marked with
	 * writeback.  So we can't just filemap_write_and_wait_range() and
	 * expect it to work since it will just kick off a thread to do the
	 * actual work.  So we need to call filemap_fdatawrite_range _again_
	 * since it will wait on the page lock, which won't be unlocked until
	 * after the pages have been marked as writeback and so we're good to go
	 * from there.  We have to do this otherwise we'll miss the ordered
	 * extents and that results in badness.  Please Josef, do not think you
	 * know better and pull this out at some point in the future, it is
	 * right and you are wrong.
	 */
	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);

	return ret;
}