free-space-cache.c 53.0 KB
Newer Older
J
Josef Bacik 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/*
 * Copyright (C) 2008 Red Hat.  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.
 */

19
#include <linux/pagemap.h>
J
Josef Bacik 已提交
20
#include <linux/sched.h>
21
#include <linux/slab.h>
22
#include <linux/math64.h>
J
Josef Bacik 已提交
23
#include "ctree.h"
24 25
#include "free-space-cache.h"
#include "transaction.h"
26
#include "disk-io.h"
27

28 29
#define BITS_PER_BITMAP		(PAGE_CACHE_SIZE * 8)
#define MAX_CACHE_BYTES_PER_GIG	(32 * 1024)
J
Josef Bacik 已提交
30

J
Josef Bacik 已提交
31 32 33 34 35
static void recalculate_thresholds(struct btrfs_block_group_cache
				   *block_group);
static int link_free_space(struct btrfs_block_group_cache *block_group,
			   struct btrfs_free_space *info);

36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189
struct inode *lookup_free_space_inode(struct btrfs_root *root,
				      struct btrfs_block_group_cache
				      *block_group, struct btrfs_path *path)
{
	struct btrfs_key key;
	struct btrfs_key location;
	struct btrfs_disk_key disk_key;
	struct btrfs_free_space_header *header;
	struct extent_buffer *leaf;
	struct inode *inode = NULL;
	int ret;

	spin_lock(&block_group->lock);
	if (block_group->inode)
		inode = igrab(block_group->inode);
	spin_unlock(&block_group->lock);
	if (inode)
		return inode;

	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
	key.offset = block_group->key.objectid;
	key.type = 0;

	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0)
		return ERR_PTR(ret);
	if (ret > 0) {
		btrfs_release_path(root, path);
		return ERR_PTR(-ENOENT);
	}

	leaf = path->nodes[0];
	header = btrfs_item_ptr(leaf, path->slots[0],
				struct btrfs_free_space_header);
	btrfs_free_space_key(leaf, header, &disk_key);
	btrfs_disk_key_to_cpu(&location, &disk_key);
	btrfs_release_path(root, path);

	inode = btrfs_iget(root->fs_info->sb, &location, root, NULL);
	if (!inode)
		return ERR_PTR(-ENOENT);
	if (IS_ERR(inode))
		return inode;
	if (is_bad_inode(inode)) {
		iput(inode);
		return ERR_PTR(-ENOENT);
	}

	spin_lock(&block_group->lock);
	if (!root->fs_info->closing) {
		block_group->inode = igrab(inode);
		block_group->iref = 1;
	}
	spin_unlock(&block_group->lock);

	return inode;
}

int create_free_space_inode(struct btrfs_root *root,
			    struct btrfs_trans_handle *trans,
			    struct btrfs_block_group_cache *block_group,
			    struct btrfs_path *path)
{
	struct btrfs_key key;
	struct btrfs_disk_key disk_key;
	struct btrfs_free_space_header *header;
	struct btrfs_inode_item *inode_item;
	struct extent_buffer *leaf;
	u64 objectid;
	int ret;

	ret = btrfs_find_free_objectid(trans, root, 0, &objectid);
	if (ret < 0)
		return ret;

	ret = btrfs_insert_empty_inode(trans, root, path, objectid);
	if (ret)
		return ret;

	leaf = path->nodes[0];
	inode_item = btrfs_item_ptr(leaf, path->slots[0],
				    struct btrfs_inode_item);
	btrfs_item_key(leaf, &disk_key, path->slots[0]);
	memset_extent_buffer(leaf, 0, (unsigned long)inode_item,
			     sizeof(*inode_item));
	btrfs_set_inode_generation(leaf, inode_item, trans->transid);
	btrfs_set_inode_size(leaf, inode_item, 0);
	btrfs_set_inode_nbytes(leaf, inode_item, 0);
	btrfs_set_inode_uid(leaf, inode_item, 0);
	btrfs_set_inode_gid(leaf, inode_item, 0);
	btrfs_set_inode_mode(leaf, inode_item, S_IFREG | 0600);
	btrfs_set_inode_flags(leaf, inode_item, BTRFS_INODE_NOCOMPRESS |
			      BTRFS_INODE_PREALLOC | BTRFS_INODE_NODATASUM);
	btrfs_set_inode_nlink(leaf, inode_item, 1);
	btrfs_set_inode_transid(leaf, inode_item, trans->transid);
	btrfs_set_inode_block_group(leaf, inode_item,
				    block_group->key.objectid);
	btrfs_mark_buffer_dirty(leaf);
	btrfs_release_path(root, path);

	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
	key.offset = block_group->key.objectid;
	key.type = 0;

	ret = btrfs_insert_empty_item(trans, root, path, &key,
				      sizeof(struct btrfs_free_space_header));
	if (ret < 0) {
		btrfs_release_path(root, path);
		return ret;
	}
	leaf = path->nodes[0];
	header = btrfs_item_ptr(leaf, path->slots[0],
				struct btrfs_free_space_header);
	memset_extent_buffer(leaf, 0, (unsigned long)header, sizeof(*header));
	btrfs_set_free_space_key(leaf, header, &disk_key);
	btrfs_mark_buffer_dirty(leaf);
	btrfs_release_path(root, path);

	return 0;
}

int btrfs_truncate_free_space_cache(struct btrfs_root *root,
				    struct btrfs_trans_handle *trans,
				    struct btrfs_path *path,
				    struct inode *inode)
{
	loff_t oldsize;
	int ret = 0;

	trans->block_rsv = root->orphan_block_rsv;
	ret = btrfs_block_rsv_check(trans, root,
				    root->orphan_block_rsv,
				    0, 5);
	if (ret)
		return ret;

	oldsize = i_size_read(inode);
	btrfs_i_size_write(inode, 0);
	truncate_pagecache(inode, oldsize, 0);

	/*
	 * We don't need an orphan item because truncating the free space cache
	 * will never be split across transactions.
	 */
	ret = btrfs_truncate_inode_items(trans, root, inode,
					 0, BTRFS_EXTENT_DATA_KEY);
	if (ret) {
		WARN_ON(1);
		return ret;
	}

	return btrfs_update_inode(trans, root, inode);
}

190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292
static int readahead_cache(struct inode *inode)
{
	struct file_ra_state *ra;
	unsigned long last_index;

	ra = kzalloc(sizeof(*ra), GFP_NOFS);
	if (!ra)
		return -ENOMEM;

	file_ra_state_init(ra, inode->i_mapping);
	last_index = (i_size_read(inode) - 1) >> PAGE_CACHE_SHIFT;

	page_cache_sync_readahead(inode->i_mapping, ra, NULL, 0, last_index);

	kfree(ra);

	return 0;
}

int load_free_space_cache(struct btrfs_fs_info *fs_info,
			  struct btrfs_block_group_cache *block_group)
{
	struct btrfs_root *root = fs_info->tree_root;
	struct inode *inode;
	struct btrfs_free_space_header *header;
	struct extent_buffer *leaf;
	struct page *page;
	struct btrfs_path *path;
	u32 *checksums = NULL, *crc;
	char *disk_crcs = NULL;
	struct btrfs_key key;
	struct list_head bitmaps;
	u64 num_entries;
	u64 num_bitmaps;
	u64 generation;
	u32 cur_crc = ~(u32)0;
	pgoff_t index = 0;
	unsigned long first_page_offset;
	int num_checksums;
	int ret = 0;

	/*
	 * If we're unmounting then just return, since this does a search on the
	 * normal root and not the commit root and we could deadlock.
	 */
	smp_mb();
	if (fs_info->closing)
		return 0;

	/*
	 * If this block group has been marked to be cleared for one reason or
	 * another then we can't trust the on disk cache, so just return.
	 */
	spin_lock(&block_group->lock);
	if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
		spin_unlock(&block_group->lock);
		return 0;
	}
	spin_unlock(&block_group->lock);

	INIT_LIST_HEAD(&bitmaps);

	path = btrfs_alloc_path();
	if (!path)
		return 0;

	inode = lookup_free_space_inode(root, block_group, path);
	if (IS_ERR(inode)) {
		btrfs_free_path(path);
		return 0;
	}

	/* Nothing in the space cache, goodbye */
	if (!i_size_read(inode)) {
		btrfs_free_path(path);
		goto out;
	}

	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
	key.offset = block_group->key.objectid;
	key.type = 0;

	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret) {
		btrfs_free_path(path);
		goto out;
	}

	leaf = path->nodes[0];
	header = btrfs_item_ptr(leaf, path->slots[0],
				struct btrfs_free_space_header);
	num_entries = btrfs_free_space_entries(leaf, header);
	num_bitmaps = btrfs_free_space_bitmaps(leaf, header);
	generation = btrfs_free_space_generation(leaf, header);
	btrfs_free_path(path);

	if (BTRFS_I(inode)->generation != generation) {
		printk(KERN_ERR "btrfs: free space inode generation (%llu) did"
		       " not match free space cache generation (%llu) for "
		       "block group %llu\n",
		       (unsigned long long)BTRFS_I(inode)->generation,
		       (unsigned long long)generation,
		       (unsigned long long)block_group->key.objectid);
293
		goto free_cache;
294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483
	}

	if (!num_entries)
		goto out;

	/* Setup everything for doing checksumming */
	num_checksums = i_size_read(inode) / PAGE_CACHE_SIZE;
	checksums = crc = kzalloc(sizeof(u32) * num_checksums, GFP_NOFS);
	if (!checksums)
		goto out;
	first_page_offset = (sizeof(u32) * num_checksums) + sizeof(u64);
	disk_crcs = kzalloc(first_page_offset, GFP_NOFS);
	if (!disk_crcs)
		goto out;

	ret = readahead_cache(inode);
	if (ret) {
		ret = 0;
		goto out;
	}

	while (1) {
		struct btrfs_free_space_entry *entry;
		struct btrfs_free_space *e;
		void *addr;
		unsigned long offset = 0;
		unsigned long start_offset = 0;
		int need_loop = 0;

		if (!num_entries && !num_bitmaps)
			break;

		if (index == 0) {
			start_offset = first_page_offset;
			offset = start_offset;
		}

		page = grab_cache_page(inode->i_mapping, index);
		if (!page) {
			ret = 0;
			goto free_cache;
		}

		if (!PageUptodate(page)) {
			btrfs_readpage(NULL, page);
			lock_page(page);
			if (!PageUptodate(page)) {
				unlock_page(page);
				page_cache_release(page);
				printk(KERN_ERR "btrfs: error reading free "
				       "space cache: %llu\n",
				       (unsigned long long)
				       block_group->key.objectid);
				goto free_cache;
			}
		}
		addr = kmap(page);

		if (index == 0) {
			u64 *gen;

			memcpy(disk_crcs, addr, first_page_offset);
			gen = addr + (sizeof(u32) * num_checksums);
			if (*gen != BTRFS_I(inode)->generation) {
				printk(KERN_ERR "btrfs: space cache generation"
				       " (%llu) does not match inode (%llu) "
				       "for block group %llu\n",
				       (unsigned long long)*gen,
				       (unsigned long long)
				       BTRFS_I(inode)->generation,
				       (unsigned long long)
				       block_group->key.objectid);
				kunmap(page);
				unlock_page(page);
				page_cache_release(page);
				goto free_cache;
			}
			crc = (u32 *)disk_crcs;
		}
		entry = addr + start_offset;

		/* First lets check our crc before we do anything fun */
		cur_crc = ~(u32)0;
		cur_crc = btrfs_csum_data(root, addr + start_offset, cur_crc,
					  PAGE_CACHE_SIZE - start_offset);
		btrfs_csum_final(cur_crc, (char *)&cur_crc);
		if (cur_crc != *crc) {
			printk(KERN_ERR "btrfs: crc mismatch for page %lu in "
			       "block group %llu\n", index,
			       (unsigned long long)block_group->key.objectid);
			kunmap(page);
			unlock_page(page);
			page_cache_release(page);
			goto free_cache;
		}
		crc++;

		while (1) {
			if (!num_entries)
				break;

			need_loop = 1;
			e = kzalloc(sizeof(struct btrfs_free_space), GFP_NOFS);
			if (!e) {
				kunmap(page);
				unlock_page(page);
				page_cache_release(page);
				goto free_cache;
			}

			e->offset = le64_to_cpu(entry->offset);
			e->bytes = le64_to_cpu(entry->bytes);
			if (!e->bytes) {
				kunmap(page);
				kfree(e);
				unlock_page(page);
				page_cache_release(page);
				goto free_cache;
			}

			if (entry->type == BTRFS_FREE_SPACE_EXTENT) {
				spin_lock(&block_group->tree_lock);
				ret = link_free_space(block_group, e);
				spin_unlock(&block_group->tree_lock);
				BUG_ON(ret);
			} else {
				e->bitmap = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS);
				if (!e->bitmap) {
					kunmap(page);
					kfree(e);
					unlock_page(page);
					page_cache_release(page);
					goto free_cache;
				}
				spin_lock(&block_group->tree_lock);
				ret = link_free_space(block_group, e);
				block_group->total_bitmaps++;
				recalculate_thresholds(block_group);
				spin_unlock(&block_group->tree_lock);
				list_add_tail(&e->list, &bitmaps);
			}

			num_entries--;
			offset += sizeof(struct btrfs_free_space_entry);
			if (offset + sizeof(struct btrfs_free_space_entry) >=
			    PAGE_CACHE_SIZE)
				break;
			entry++;
		}

		/*
		 * We read an entry out of this page, we need to move on to the
		 * next page.
		 */
		if (need_loop) {
			kunmap(page);
			goto next;
		}

		/*
		 * We add the bitmaps at the end of the entries in order that
		 * the bitmap entries are added to the cache.
		 */
		e = list_entry(bitmaps.next, struct btrfs_free_space, list);
		list_del_init(&e->list);
		memcpy(e->bitmap, addr, PAGE_CACHE_SIZE);
		kunmap(page);
		num_bitmaps--;
next:
		unlock_page(page);
		page_cache_release(page);
		index++;
	}

	ret = 1;
out:
	kfree(checksums);
	kfree(disk_crcs);
	iput(inode);
	return ret;

free_cache:
	/* This cache is bogus, make sure it gets cleared */
	spin_lock(&block_group->lock);
	block_group->disk_cache_state = BTRFS_DC_CLEAR;
	spin_unlock(&block_group->lock);
	btrfs_remove_free_space_cache(block_group);
	goto out;
}

J
Josef Bacik 已提交
484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526
int btrfs_write_out_cache(struct btrfs_root *root,
			  struct btrfs_trans_handle *trans,
			  struct btrfs_block_group_cache *block_group,
			  struct btrfs_path *path)
{
	struct btrfs_free_space_header *header;
	struct extent_buffer *leaf;
	struct inode *inode;
	struct rb_node *node;
	struct list_head *pos, *n;
	struct page *page;
	struct extent_state *cached_state = NULL;
	struct list_head bitmap_list;
	struct btrfs_key key;
	u64 bytes = 0;
	u32 *crc, *checksums;
	pgoff_t index = 0, last_index = 0;
	unsigned long first_page_offset;
	int num_checksums;
	int entries = 0;
	int bitmaps = 0;
	int ret = 0;

	root = root->fs_info->tree_root;

	INIT_LIST_HEAD(&bitmap_list);

	spin_lock(&block_group->lock);
	if (block_group->disk_cache_state < BTRFS_DC_SETUP) {
		spin_unlock(&block_group->lock);
		return 0;
	}
	spin_unlock(&block_group->lock);

	inode = lookup_free_space_inode(root, block_group, path);
	if (IS_ERR(inode))
		return 0;

	if (!i_size_read(inode)) {
		iput(inode);
		return 0;
	}

527 528 529 530 531 532
	node = rb_first(&block_group->free_space_offset);
	if (!node) {
		iput(inode);
		return 0;
	}

J
Josef Bacik 已提交
533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782
	last_index = (i_size_read(inode) - 1) >> PAGE_CACHE_SHIFT;
	filemap_write_and_wait(inode->i_mapping);
	btrfs_wait_ordered_range(inode, inode->i_size &
				 ~(root->sectorsize - 1), (u64)-1);

	/* We need a checksum per page. */
	num_checksums = i_size_read(inode) / PAGE_CACHE_SIZE;
	crc = checksums  = kzalloc(sizeof(u32) * num_checksums, GFP_NOFS);
	if (!crc) {
		iput(inode);
		return 0;
	}

	/* Since the first page has all of our checksums and our generation we
	 * need to calculate the offset into the page that we can start writing
	 * our entries.
	 */
	first_page_offset = (sizeof(u32) * num_checksums) + sizeof(u64);

	/*
	 * Lock all pages first so we can lock the extent safely.
	 *
	 * NOTE: Because we hold the ref the entire time we're going to write to
	 * the page find_get_page should never fail, so we don't do a check
	 * after find_get_page at this point.  Just putting this here so people
	 * know and don't freak out.
	 */
	while (index <= last_index) {
		page = grab_cache_page(inode->i_mapping, index);
		if (!page) {
			pgoff_t i = 0;

			while (i < index) {
				page = find_get_page(inode->i_mapping, i);
				unlock_page(page);
				page_cache_release(page);
				page_cache_release(page);
				i++;
			}
			goto out_free;
		}
		index++;
	}

	index = 0;
	lock_extent_bits(&BTRFS_I(inode)->io_tree, 0, i_size_read(inode) - 1,
			 0, &cached_state, GFP_NOFS);

	/* Write out the extent entries */
	do {
		struct btrfs_free_space_entry *entry;
		void *addr;
		unsigned long offset = 0;
		unsigned long start_offset = 0;

		if (index == 0) {
			start_offset = first_page_offset;
			offset = start_offset;
		}

		page = find_get_page(inode->i_mapping, index);

		addr = kmap(page);
		entry = addr + start_offset;

		memset(addr, 0, PAGE_CACHE_SIZE);
		while (1) {
			struct btrfs_free_space *e;

			e = rb_entry(node, struct btrfs_free_space, offset_index);
			entries++;

			entry->offset = cpu_to_le64(e->offset);
			entry->bytes = cpu_to_le64(e->bytes);
			if (e->bitmap) {
				entry->type = BTRFS_FREE_SPACE_BITMAP;
				list_add_tail(&e->list, &bitmap_list);
				bitmaps++;
			} else {
				entry->type = BTRFS_FREE_SPACE_EXTENT;
			}
			node = rb_next(node);
			if (!node)
				break;
			offset += sizeof(struct btrfs_free_space_entry);
			if (offset + sizeof(struct btrfs_free_space_entry) >=
			    PAGE_CACHE_SIZE)
				break;
			entry++;
		}
		*crc = ~(u32)0;
		*crc = btrfs_csum_data(root, addr + start_offset, *crc,
				       PAGE_CACHE_SIZE - start_offset);
		kunmap(page);

		btrfs_csum_final(*crc, (char *)crc);
		crc++;

		bytes += PAGE_CACHE_SIZE;

		ClearPageChecked(page);
		set_page_extent_mapped(page);
		SetPageUptodate(page);
		set_page_dirty(page);

		/*
		 * We need to release our reference we got for grab_cache_page,
		 * except for the first page which will hold our checksums, we
		 * do that below.
		 */
		if (index != 0) {
			unlock_page(page);
			page_cache_release(page);
		}

		page_cache_release(page);

		index++;
	} while (node);

	/* Write out the bitmaps */
	list_for_each_safe(pos, n, &bitmap_list) {
		void *addr;
		struct btrfs_free_space *entry =
			list_entry(pos, struct btrfs_free_space, list);

		page = find_get_page(inode->i_mapping, index);

		addr = kmap(page);
		memcpy(addr, entry->bitmap, PAGE_CACHE_SIZE);
		*crc = ~(u32)0;
		*crc = btrfs_csum_data(root, addr, *crc, PAGE_CACHE_SIZE);
		kunmap(page);
		btrfs_csum_final(*crc, (char *)crc);
		crc++;
		bytes += PAGE_CACHE_SIZE;

		ClearPageChecked(page);
		set_page_extent_mapped(page);
		SetPageUptodate(page);
		set_page_dirty(page);
		unlock_page(page);
		page_cache_release(page);
		page_cache_release(page);
		list_del_init(&entry->list);
		index++;
	}

	/* Zero out the rest of the pages just to make sure */
	while (index <= last_index) {
		void *addr;

		page = find_get_page(inode->i_mapping, index);

		addr = kmap(page);
		memset(addr, 0, PAGE_CACHE_SIZE);
		kunmap(page);
		ClearPageChecked(page);
		set_page_extent_mapped(page);
		SetPageUptodate(page);
		set_page_dirty(page);
		unlock_page(page);
		page_cache_release(page);
		page_cache_release(page);
		bytes += PAGE_CACHE_SIZE;
		index++;
	}

	btrfs_set_extent_delalloc(inode, 0, bytes - 1, &cached_state);

	/* Write the checksums and trans id to the first page */
	{
		void *addr;
		u64 *gen;

		page = find_get_page(inode->i_mapping, 0);

		addr = kmap(page);
		memcpy(addr, checksums, sizeof(u32) * num_checksums);
		gen = addr + (sizeof(u32) * num_checksums);
		*gen = trans->transid;
		kunmap(page);
		ClearPageChecked(page);
		set_page_extent_mapped(page);
		SetPageUptodate(page);
		set_page_dirty(page);
		unlock_page(page);
		page_cache_release(page);
		page_cache_release(page);
	}
	BTRFS_I(inode)->generation = trans->transid;

	unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
			     i_size_read(inode) - 1, &cached_state, GFP_NOFS);

	filemap_write_and_wait(inode->i_mapping);

	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
	key.offset = block_group->key.objectid;
	key.type = 0;

	ret = btrfs_search_slot(trans, root, &key, path, 1, 1);
	if (ret < 0) {
		ret = 0;
		clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, bytes - 1,
				 EXTENT_DIRTY | EXTENT_DELALLOC |
				 EXTENT_DO_ACCOUNTING, 0, 0, NULL, GFP_NOFS);
		goto out_free;
	}
	leaf = path->nodes[0];
	if (ret > 0) {
		struct btrfs_key found_key;
		BUG_ON(!path->slots[0]);
		path->slots[0]--;
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
		if (found_key.objectid != BTRFS_FREE_SPACE_OBJECTID ||
		    found_key.offset != block_group->key.objectid) {
			ret = 0;
			clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, bytes - 1,
					 EXTENT_DIRTY | EXTENT_DELALLOC |
					 EXTENT_DO_ACCOUNTING, 0, 0, NULL,
					 GFP_NOFS);
			btrfs_release_path(root, path);
			goto out_free;
		}
	}
	header = btrfs_item_ptr(leaf, path->slots[0],
				struct btrfs_free_space_header);
	btrfs_set_free_space_entries(leaf, header, entries);
	btrfs_set_free_space_bitmaps(leaf, header, bitmaps);
	btrfs_set_free_space_generation(leaf, header, trans->transid);
	btrfs_mark_buffer_dirty(leaf);
	btrfs_release_path(root, path);

	ret = 1;

out_free:
	if (ret == 0) {
		invalidate_inode_pages2_range(inode->i_mapping, 0, index);
		spin_lock(&block_group->lock);
		block_group->disk_cache_state = BTRFS_DC_ERROR;
		spin_unlock(&block_group->lock);
		BTRFS_I(inode)->generation = 0;
	}
	kfree(checksums);
	btrfs_update_inode(trans, root, inode);
	iput(inode);
	return ret;
}

783 784
static inline unsigned long offset_to_bit(u64 bitmap_start, u64 sectorsize,
					  u64 offset)
J
Josef Bacik 已提交
785
{
786 787 788 789
	BUG_ON(offset < bitmap_start);
	offset -= bitmap_start;
	return (unsigned long)(div64_u64(offset, sectorsize));
}
J
Josef Bacik 已提交
790

791 792 793 794
static inline unsigned long bytes_to_bits(u64 bytes, u64 sectorsize)
{
	return (unsigned long)(div64_u64(bytes, sectorsize));
}
J
Josef Bacik 已提交
795

796 797 798 799 800
static inline u64 offset_to_bitmap(struct btrfs_block_group_cache *block_group,
				   u64 offset)
{
	u64 bitmap_start;
	u64 bytes_per_bitmap;
J
Josef Bacik 已提交
801

802 803 804 805 806
	bytes_per_bitmap = BITS_PER_BITMAP * block_group->sectorsize;
	bitmap_start = offset - block_group->key.objectid;
	bitmap_start = div64_u64(bitmap_start, bytes_per_bitmap);
	bitmap_start *= bytes_per_bitmap;
	bitmap_start += block_group->key.objectid;
J
Josef Bacik 已提交
807

808
	return bitmap_start;
J
Josef Bacik 已提交
809 810
}

811 812
static int tree_insert_offset(struct rb_root *root, u64 offset,
			      struct rb_node *node, int bitmap)
J
Josef Bacik 已提交
813 814 815 816 817 818 819
{
	struct rb_node **p = &root->rb_node;
	struct rb_node *parent = NULL;
	struct btrfs_free_space *info;

	while (*p) {
		parent = *p;
820
		info = rb_entry(parent, struct btrfs_free_space, offset_index);
J
Josef Bacik 已提交
821

822
		if (offset < info->offset) {
J
Josef Bacik 已提交
823
			p = &(*p)->rb_left;
824
		} else if (offset > info->offset) {
J
Josef Bacik 已提交
825
			p = &(*p)->rb_right;
826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847
		} else {
			/*
			 * we could have a bitmap entry and an extent entry
			 * share the same offset.  If this is the case, we want
			 * the extent entry to always be found first if we do a
			 * linear search through the tree, since we want to have
			 * the quickest allocation time, and allocating from an
			 * extent is faster than allocating from a bitmap.  So
			 * if we're inserting a bitmap and we find an entry at
			 * this offset, we want to go right, or after this entry
			 * logically.  If we are inserting an extent and we've
			 * found a bitmap, we want to go left, or before
			 * logically.
			 */
			if (bitmap) {
				WARN_ON(info->bitmap);
				p = &(*p)->rb_right;
			} else {
				WARN_ON(!info->bitmap);
				p = &(*p)->rb_left;
			}
		}
J
Josef Bacik 已提交
848 849 850 851 852 853 854 855 856
	}

	rb_link_node(node, parent, p);
	rb_insert_color(node, root);

	return 0;
}

/*
J
Josef Bacik 已提交
857 858
 * searches the tree for the given offset.
 *
859 860 861
 * fuzzy - If this is set, then we are trying to make an allocation, and we just
 * want a section that has at least bytes size and comes at or after the given
 * offset.
J
Josef Bacik 已提交
862
 */
863 864 865
static struct btrfs_free_space *
tree_search_offset(struct btrfs_block_group_cache *block_group,
		   u64 offset, int bitmap_only, int fuzzy)
J
Josef Bacik 已提交
866
{
867 868 869 870 871 872 873 874 875
	struct rb_node *n = block_group->free_space_offset.rb_node;
	struct btrfs_free_space *entry, *prev = NULL;

	/* find entry that is closest to the 'offset' */
	while (1) {
		if (!n) {
			entry = NULL;
			break;
		}
J
Josef Bacik 已提交
876 877

		entry = rb_entry(n, struct btrfs_free_space, offset_index);
878
		prev = entry;
J
Josef Bacik 已提交
879

880
		if (offset < entry->offset)
J
Josef Bacik 已提交
881
			n = n->rb_left;
882
		else if (offset > entry->offset)
J
Josef Bacik 已提交
883
			n = n->rb_right;
884
		else
J
Josef Bacik 已提交
885 886 887
			break;
	}

888 889 890 891 892
	if (bitmap_only) {
		if (!entry)
			return NULL;
		if (entry->bitmap)
			return entry;
J
Josef Bacik 已提交
893

894 895 896 897 898 899 900 901 902 903
		/*
		 * bitmap entry and extent entry may share same offset,
		 * in that case, bitmap entry comes after extent entry.
		 */
		n = rb_next(n);
		if (!n)
			return NULL;
		entry = rb_entry(n, struct btrfs_free_space, offset_index);
		if (entry->offset != offset)
			return NULL;
J
Josef Bacik 已提交
904

905 906 907 908
		WARN_ON(!entry->bitmap);
		return entry;
	} else if (entry) {
		if (entry->bitmap) {
J
Josef Bacik 已提交
909
			/*
910 911
			 * if previous extent entry covers the offset,
			 * we should return it instead of the bitmap entry
J
Josef Bacik 已提交
912
			 */
913 914 915 916 917 918 919 920 921 922 923 924
			n = &entry->offset_index;
			while (1) {
				n = rb_prev(n);
				if (!n)
					break;
				prev = rb_entry(n, struct btrfs_free_space,
						offset_index);
				if (!prev->bitmap) {
					if (prev->offset + prev->bytes > offset)
						entry = prev;
					break;
				}
J
Josef Bacik 已提交
925
			}
926 927 928 929 930 931 932 933 934 935 936 937 938 939 940
		}
		return entry;
	}

	if (!prev)
		return NULL;

	/* find last entry before the 'offset' */
	entry = prev;
	if (entry->offset > offset) {
		n = rb_prev(&entry->offset_index);
		if (n) {
			entry = rb_entry(n, struct btrfs_free_space,
					offset_index);
			BUG_ON(entry->offset > offset);
J
Josef Bacik 已提交
941
		} else {
942 943 944 945
			if (fuzzy)
				return entry;
			else
				return NULL;
J
Josef Bacik 已提交
946 947 948
		}
	}

949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987
	if (entry->bitmap) {
		n = &entry->offset_index;
		while (1) {
			n = rb_prev(n);
			if (!n)
				break;
			prev = rb_entry(n, struct btrfs_free_space,
					offset_index);
			if (!prev->bitmap) {
				if (prev->offset + prev->bytes > offset)
					return prev;
				break;
			}
		}
		if (entry->offset + BITS_PER_BITMAP *
		    block_group->sectorsize > offset)
			return entry;
	} else if (entry->offset + entry->bytes > offset)
		return entry;

	if (!fuzzy)
		return NULL;

	while (1) {
		if (entry->bitmap) {
			if (entry->offset + BITS_PER_BITMAP *
			    block_group->sectorsize > offset)
				break;
		} else {
			if (entry->offset + entry->bytes > offset)
				break;
		}

		n = rb_next(&entry->offset_index);
		if (!n)
			return NULL;
		entry = rb_entry(n, struct btrfs_free_space, offset_index);
	}
	return entry;
J
Josef Bacik 已提交
988 989 990 991 992 993
}

static void unlink_free_space(struct btrfs_block_group_cache *block_group,
			      struct btrfs_free_space *info)
{
	rb_erase(&info->offset_index, &block_group->free_space_offset);
994
	block_group->free_extents--;
J
Josef Bacik 已提交
995
	block_group->free_space -= info->bytes;
J
Josef Bacik 已提交
996 997 998 999 1000 1001 1002
}

static int link_free_space(struct btrfs_block_group_cache *block_group,
			   struct btrfs_free_space *info)
{
	int ret = 0;

1003
	BUG_ON(!info->bitmap && !info->bytes);
J
Josef Bacik 已提交
1004
	ret = tree_insert_offset(&block_group->free_space_offset, info->offset,
1005
				 &info->offset_index, (info->bitmap != NULL));
J
Josef Bacik 已提交
1006 1007 1008
	if (ret)
		return ret;

J
Josef Bacik 已提交
1009
	block_group->free_space += info->bytes;
1010 1011 1012 1013 1014 1015
	block_group->free_extents++;
	return ret;
}

static void recalculate_thresholds(struct btrfs_block_group_cache *block_group)
{
1016 1017 1018
	u64 max_bytes;
	u64 bitmap_bytes;
	u64 extent_bytes;
1019 1020 1021 1022 1023 1024 1025 1026 1027

	/*
	 * The goal is to keep the total amount of memory used per 1gb of space
	 * at or below 32k, so we need to adjust how much memory we allow to be
	 * used by extent based free space tracking
	 */
	max_bytes = MAX_CACHE_BYTES_PER_GIG *
		(div64_u64(block_group->key.offset, 1024 * 1024 * 1024));

1028 1029 1030 1031 1032 1033
	/*
	 * we want to account for 1 more bitmap than what we have so we can make
	 * sure we don't go over our overall goal of MAX_CACHE_BYTES_PER_GIG as
	 * we add more bitmaps.
	 */
	bitmap_bytes = (block_group->total_bitmaps + 1) * PAGE_CACHE_SIZE;
1034

1035 1036 1037 1038
	if (bitmap_bytes >= max_bytes) {
		block_group->extents_thresh = 0;
		return;
	}
1039

1040 1041 1042 1043 1044 1045
	/*
	 * we want the extent entry threshold to always be at most 1/2 the maxw
	 * bytes we can have, or whatever is less than that.
	 */
	extent_bytes = max_bytes - bitmap_bytes;
	extent_bytes = min_t(u64, extent_bytes, div64_u64(max_bytes, 2));
1046

1047 1048
	block_group->extents_thresh =
		div64_u64(extent_bytes, (sizeof(struct btrfs_free_space)));
1049 1050
}

J
Josef Bacik 已提交
1051 1052 1053
static void bitmap_clear_bits(struct btrfs_block_group_cache *block_group,
			      struct btrfs_free_space *info, u64 offset,
			      u64 bytes)
1054 1055 1056 1057
{
	unsigned long start, end;
	unsigned long i;

J
Josef Bacik 已提交
1058 1059
	start = offset_to_bit(info->offset, block_group->sectorsize, offset);
	end = start + bytes_to_bits(bytes, block_group->sectorsize);
1060 1061 1062 1063 1064 1065
	BUG_ON(end > BITS_PER_BITMAP);

	for (i = start; i < end; i++)
		clear_bit(i, info->bitmap);

	info->bytes -= bytes;
J
Josef Bacik 已提交
1066
	block_group->free_space -= bytes;
1067 1068
}

J
Josef Bacik 已提交
1069 1070 1071
static void bitmap_set_bits(struct btrfs_block_group_cache *block_group,
			    struct btrfs_free_space *info, u64 offset,
			    u64 bytes)
1072 1073 1074 1075
{
	unsigned long start, end;
	unsigned long i;

J
Josef Bacik 已提交
1076 1077
	start = offset_to_bit(info->offset, block_group->sectorsize, offset);
	end = start + bytes_to_bits(bytes, block_group->sectorsize);
1078 1079 1080 1081 1082 1083
	BUG_ON(end > BITS_PER_BITMAP);

	for (i = start; i < end; i++)
		set_bit(i, info->bitmap);

	info->bytes += bytes;
J
Josef Bacik 已提交
1084
	block_group->free_space += bytes;
1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166
}

static int search_bitmap(struct btrfs_block_group_cache *block_group,
			 struct btrfs_free_space *bitmap_info, u64 *offset,
			 u64 *bytes)
{
	unsigned long found_bits = 0;
	unsigned long bits, i;
	unsigned long next_zero;

	i = offset_to_bit(bitmap_info->offset, block_group->sectorsize,
			  max_t(u64, *offset, bitmap_info->offset));
	bits = bytes_to_bits(*bytes, block_group->sectorsize);

	for (i = find_next_bit(bitmap_info->bitmap, BITS_PER_BITMAP, i);
	     i < BITS_PER_BITMAP;
	     i = find_next_bit(bitmap_info->bitmap, BITS_PER_BITMAP, i + 1)) {
		next_zero = find_next_zero_bit(bitmap_info->bitmap,
					       BITS_PER_BITMAP, i);
		if ((next_zero - i) >= bits) {
			found_bits = next_zero - i;
			break;
		}
		i = next_zero;
	}

	if (found_bits) {
		*offset = (u64)(i * block_group->sectorsize) +
			bitmap_info->offset;
		*bytes = (u64)(found_bits) * block_group->sectorsize;
		return 0;
	}

	return -1;
}

static struct btrfs_free_space *find_free_space(struct btrfs_block_group_cache
						*block_group, u64 *offset,
						u64 *bytes, int debug)
{
	struct btrfs_free_space *entry;
	struct rb_node *node;
	int ret;

	if (!block_group->free_space_offset.rb_node)
		return NULL;

	entry = tree_search_offset(block_group,
				   offset_to_bitmap(block_group, *offset),
				   0, 1);
	if (!entry)
		return NULL;

	for (node = &entry->offset_index; node; node = rb_next(node)) {
		entry = rb_entry(node, struct btrfs_free_space, offset_index);
		if (entry->bytes < *bytes)
			continue;

		if (entry->bitmap) {
			ret = search_bitmap(block_group, entry, offset, bytes);
			if (!ret)
				return entry;
			continue;
		}

		*offset = entry->offset;
		*bytes = entry->bytes;
		return entry;
	}

	return NULL;
}

static void add_new_bitmap(struct btrfs_block_group_cache *block_group,
			   struct btrfs_free_space *info, u64 offset)
{
	u64 bytes_per_bg = BITS_PER_BITMAP * block_group->sectorsize;
	int max_bitmaps = (int)div64_u64(block_group->key.offset +
					 bytes_per_bg - 1, bytes_per_bg);
	BUG_ON(block_group->total_bitmaps >= max_bitmaps);

	info->offset = offset_to_bitmap(block_group, offset);
J
Josef Bacik 已提交
1167
	info->bytes = 0;
1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178
	link_free_space(block_group, info);
	block_group->total_bitmaps++;

	recalculate_thresholds(block_group);
}

static noinline int remove_from_bitmap(struct btrfs_block_group_cache *block_group,
			      struct btrfs_free_space *bitmap_info,
			      u64 *offset, u64 *bytes)
{
	u64 end;
1179 1180
	u64 search_start, search_bytes;
	int ret;
1181 1182 1183 1184 1185

again:
	end = bitmap_info->offset +
		(u64)(BITS_PER_BITMAP * block_group->sectorsize) - 1;

1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201
	/*
	 * XXX - this can go away after a few releases.
	 *
	 * since the only user of btrfs_remove_free_space is the tree logging
	 * stuff, and the only way to test that is under crash conditions, we
	 * want to have this debug stuff here just in case somethings not
	 * working.  Search the bitmap for the space we are trying to use to
	 * make sure its actually there.  If its not there then we need to stop
	 * because something has gone wrong.
	 */
	search_start = *offset;
	search_bytes = *bytes;
	ret = search_bitmap(block_group, bitmap_info, &search_start,
			    &search_bytes);
	BUG_ON(ret < 0 || search_start != *offset);

1202
	if (*offset > bitmap_info->offset && *offset + *bytes > end) {
J
Josef Bacik 已提交
1203 1204
		bitmap_clear_bits(block_group, bitmap_info, *offset,
				  end - *offset + 1);
1205 1206 1207
		*bytes -= end - *offset + 1;
		*offset = end + 1;
	} else if (*offset >= bitmap_info->offset && *offset + *bytes <= end) {
J
Josef Bacik 已提交
1208
		bitmap_clear_bits(block_group, bitmap_info, *offset, *bytes);
1209 1210 1211 1212
		*bytes = 0;
	}

	if (*bytes) {
1213
		struct rb_node *next = rb_next(&bitmap_info->offset_index);
1214 1215 1216 1217 1218 1219 1220 1221
		if (!bitmap_info->bytes) {
			unlink_free_space(block_group, bitmap_info);
			kfree(bitmap_info->bitmap);
			kfree(bitmap_info);
			block_group->total_bitmaps--;
			recalculate_thresholds(block_group);
		}

1222 1223 1224 1225 1226
		/*
		 * no entry after this bitmap, but we still have bytes to
		 * remove, so something has gone wrong.
		 */
		if (!next)
1227 1228
			return -EINVAL;

1229 1230 1231 1232 1233 1234 1235
		bitmap_info = rb_entry(next, struct btrfs_free_space,
				       offset_index);

		/*
		 * if the next entry isn't a bitmap we need to return to let the
		 * extent stuff do its work.
		 */
1236 1237 1238
		if (!bitmap_info->bitmap)
			return -EAGAIN;

1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251
		/*
		 * Ok the next item is a bitmap, but it may not actually hold
		 * the information for the rest of this free space stuff, so
		 * look for it, and if we don't find it return so we can try
		 * everything over again.
		 */
		search_start = *offset;
		search_bytes = *bytes;
		ret = search_bitmap(block_group, bitmap_info, &search_start,
				    &search_bytes);
		if (ret < 0 || search_start != *offset)
			return -EAGAIN;

1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303
		goto again;
	} else if (!bitmap_info->bytes) {
		unlink_free_space(block_group, bitmap_info);
		kfree(bitmap_info->bitmap);
		kfree(bitmap_info);
		block_group->total_bitmaps--;
		recalculate_thresholds(block_group);
	}

	return 0;
}

static int insert_into_bitmap(struct btrfs_block_group_cache *block_group,
			      struct btrfs_free_space *info)
{
	struct btrfs_free_space *bitmap_info;
	int added = 0;
	u64 bytes, offset, end;
	int ret;

	/*
	 * If we are below the extents threshold then we can add this as an
	 * extent, and don't have to deal with the bitmap
	 */
	if (block_group->free_extents < block_group->extents_thresh &&
	    info->bytes > block_group->sectorsize * 4)
		return 0;

	/*
	 * some block groups are so tiny they can't be enveloped by a bitmap, so
	 * don't even bother to create a bitmap for this
	 */
	if (BITS_PER_BITMAP * block_group->sectorsize >
	    block_group->key.offset)
		return 0;

	bytes = info->bytes;
	offset = info->offset;

again:
	bitmap_info = tree_search_offset(block_group,
					 offset_to_bitmap(block_group, offset),
					 1, 0);
	if (!bitmap_info) {
		BUG_ON(added);
		goto new_bitmap;
	}

	end = bitmap_info->offset +
		(u64)(BITS_PER_BITMAP * block_group->sectorsize);

	if (offset >= bitmap_info->offset && offset + bytes > end) {
J
Josef Bacik 已提交
1304 1305
		bitmap_set_bits(block_group, bitmap_info, offset,
				end - offset);
1306 1307 1308 1309
		bytes -= end - offset;
		offset = end;
		added = 0;
	} else if (offset >= bitmap_info->offset && offset + bytes <= end) {
J
Josef Bacik 已提交
1310
		bitmap_set_bits(block_group, bitmap_info, offset, bytes);
1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357
		bytes = 0;
	} else {
		BUG();
	}

	if (!bytes) {
		ret = 1;
		goto out;
	} else
		goto again;

new_bitmap:
	if (info && info->bitmap) {
		add_new_bitmap(block_group, info, offset);
		added = 1;
		info = NULL;
		goto again;
	} else {
		spin_unlock(&block_group->tree_lock);

		/* no pre-allocated info, allocate a new one */
		if (!info) {
			info = kzalloc(sizeof(struct btrfs_free_space),
				       GFP_NOFS);
			if (!info) {
				spin_lock(&block_group->tree_lock);
				ret = -ENOMEM;
				goto out;
			}
		}

		/* allocate the bitmap */
		info->bitmap = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS);
		spin_lock(&block_group->tree_lock);
		if (!info->bitmap) {
			ret = -ENOMEM;
			goto out;
		}
		goto again;
	}

out:
	if (info) {
		if (info->bitmap)
			kfree(info->bitmap);
		kfree(info);
	}
J
Josef Bacik 已提交
1358 1359 1360 1361

	return ret;
}

1362 1363
int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
			 u64 offset, u64 bytes)
J
Josef Bacik 已提交
1364
{
1365 1366
	struct btrfs_free_space *right_info = NULL;
	struct btrfs_free_space *left_info = NULL;
J
Josef Bacik 已提交
1367 1368 1369
	struct btrfs_free_space *info = NULL;
	int ret = 0;

1370 1371 1372 1373 1374 1375 1376 1377 1378
	info = kzalloc(sizeof(struct btrfs_free_space), GFP_NOFS);
	if (!info)
		return -ENOMEM;

	info->offset = offset;
	info->bytes = bytes;

	spin_lock(&block_group->tree_lock);

J
Josef Bacik 已提交
1379 1380 1381 1382 1383
	/*
	 * first we want to see if there is free space adjacent to the range we
	 * are adding, if there is remove that struct and add a new one to
	 * cover the entire range
	 */
1384 1385 1386 1387 1388 1389
	right_info = tree_search_offset(block_group, offset + bytes, 0, 0);
	if (right_info && rb_prev(&right_info->offset_index))
		left_info = rb_entry(rb_prev(&right_info->offset_index),
				     struct btrfs_free_space, offset_index);
	else
		left_info = tree_search_offset(block_group, offset - 1, 0, 0);
J
Josef Bacik 已提交
1390

1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408
	/*
	 * If there was no extent directly to the left or right of this new
	 * extent then we know we're going to have to allocate a new extent, so
	 * before we do that see if we need to drop this into a bitmap
	 */
	if ((!left_info || left_info->bitmap) &&
	    (!right_info || right_info->bitmap)) {
		ret = insert_into_bitmap(block_group, info);

		if (ret < 0) {
			goto out;
		} else if (ret) {
			ret = 0;
			goto out;
		}
	}

	if (right_info && !right_info->bitmap) {
J
Josef Bacik 已提交
1409
		unlink_free_space(block_group, right_info);
1410 1411
		info->bytes += right_info->bytes;
		kfree(right_info);
J
Josef Bacik 已提交
1412 1413
	}

1414 1415
	if (left_info && !left_info->bitmap &&
	    left_info->offset + left_info->bytes == offset) {
J
Josef Bacik 已提交
1416
		unlink_free_space(block_group, left_info);
1417 1418 1419
		info->offset = left_info->offset;
		info->bytes += left_info->bytes;
		kfree(left_info);
J
Josef Bacik 已提交
1420 1421 1422 1423 1424
	}

	ret = link_free_space(block_group, info);
	if (ret)
		kfree(info);
1425
out:
1426 1427
	spin_unlock(&block_group->tree_lock);

J
Josef Bacik 已提交
1428
	if (ret) {
1429
		printk(KERN_CRIT "btrfs: unable to add free space :%d\n", ret);
S
Stoyan Gaydarov 已提交
1430
		BUG_ON(ret == -EEXIST);
J
Josef Bacik 已提交
1431 1432 1433 1434 1435
	}

	return ret;
}

1436 1437
int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
			    u64 offset, u64 bytes)
J
Josef Bacik 已提交
1438 1439
{
	struct btrfs_free_space *info;
1440
	struct btrfs_free_space *next_info = NULL;
J
Josef Bacik 已提交
1441 1442
	int ret = 0;

1443 1444
	spin_lock(&block_group->tree_lock);

1445 1446 1447
again:
	info = tree_search_offset(block_group, offset, 0, 0);
	if (!info) {
1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458
		/*
		 * oops didn't find an extent that matched the space we wanted
		 * to remove, look for a bitmap instead
		 */
		info = tree_search_offset(block_group,
					  offset_to_bitmap(block_group, offset),
					  1, 0);
		if (!info) {
			WARN_ON(1);
			goto out_lock;
		}
1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479
	}

	if (info->bytes < bytes && rb_next(&info->offset_index)) {
		u64 end;
		next_info = rb_entry(rb_next(&info->offset_index),
					     struct btrfs_free_space,
					     offset_index);

		if (next_info->bitmap)
			end = next_info->offset + BITS_PER_BITMAP *
				block_group->sectorsize - 1;
		else
			end = next_info->offset + next_info->bytes;

		if (next_info->bytes < bytes ||
		    next_info->offset > offset || offset > end) {
			printk(KERN_CRIT "Found free space at %llu, size %llu,"
			      " trying to use %llu\n",
			      (unsigned long long)info->offset,
			      (unsigned long long)info->bytes,
			      (unsigned long long)bytes);
J
Josef Bacik 已提交
1480 1481
			WARN_ON(1);
			ret = -EINVAL;
1482
			goto out_lock;
J
Josef Bacik 已提交
1483 1484
		}

1485 1486 1487 1488 1489 1490 1491 1492
		info = next_info;
	}

	if (info->bytes == bytes) {
		unlink_free_space(block_group, info);
		if (info->bitmap) {
			kfree(info->bitmap);
			block_group->total_bitmaps--;
J
Josef Bacik 已提交
1493
		}
1494 1495 1496
		kfree(info);
		goto out_lock;
	}
J
Josef Bacik 已提交
1497

1498 1499
	if (!info->bitmap && info->offset == offset) {
		unlink_free_space(block_group, info);
J
Josef Bacik 已提交
1500 1501
		info->offset += bytes;
		info->bytes -= bytes;
1502 1503 1504
		link_free_space(block_group, info);
		goto out_lock;
	}
J
Josef Bacik 已提交
1505

1506 1507
	if (!info->bitmap && info->offset <= offset &&
	    info->offset + info->bytes >= offset + bytes) {
1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522
		u64 old_start = info->offset;
		/*
		 * we're freeing space in the middle of the info,
		 * this can happen during tree log replay
		 *
		 * first unlink the old info and then
		 * insert it again after the hole we're creating
		 */
		unlink_free_space(block_group, info);
		if (offset + bytes < info->offset + info->bytes) {
			u64 old_end = info->offset + info->bytes;

			info->offset = offset + bytes;
			info->bytes = old_end - info->offset;
			ret = link_free_space(block_group, info);
1523 1524 1525
			WARN_ON(ret);
			if (ret)
				goto out_lock;
1526 1527 1528 1529 1530 1531
		} else {
			/* the hole we're creating ends at the end
			 * of the info struct, just free the info
			 */
			kfree(info);
		}
1532
		spin_unlock(&block_group->tree_lock);
1533 1534 1535

		/* step two, insert a new info struct to cover
		 * anything before the hole
1536
		 */
1537 1538
		ret = btrfs_add_free_space(block_group, old_start,
					   offset - old_start);
1539 1540
		WARN_ON(ret);
		goto out;
J
Josef Bacik 已提交
1541
	}
1542 1543 1544 1545 1546 1547 1548

	ret = remove_from_bitmap(block_group, info, &offset, &bytes);
	if (ret == -EAGAIN)
		goto again;
	BUG_ON(ret);
out_lock:
	spin_unlock(&block_group->tree_lock);
J
Josef Bacik 已提交
1549
out:
1550 1551 1552
	return ret;
}

J
Josef Bacik 已提交
1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563
void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
			   u64 bytes)
{
	struct btrfs_free_space *info;
	struct rb_node *n;
	int count = 0;

	for (n = rb_first(&block_group->free_space_offset); n; n = rb_next(n)) {
		info = rb_entry(n, struct btrfs_free_space, offset_index);
		if (info->bytes >= bytes)
			count++;
1564
		printk(KERN_CRIT "entry offset %llu, bytes %llu, bitmap %s\n",
1565
		       (unsigned long long)info->offset,
1566 1567
		       (unsigned long long)info->bytes,
		       (info->bitmap) ? "yes" : "no");
J
Josef Bacik 已提交
1568
	}
1569 1570
	printk(KERN_INFO "block group has cluster?: %s\n",
	       list_empty(&block_group->cluster_list) ? "no" : "yes");
J
Josef Bacik 已提交
1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589
	printk(KERN_INFO "%d blocks of free space at or bigger than bytes is"
	       "\n", count);
}

u64 btrfs_block_group_free_space(struct btrfs_block_group_cache *block_group)
{
	struct btrfs_free_space *info;
	struct rb_node *n;
	u64 ret = 0;

	for (n = rb_first(&block_group->free_space_offset); n;
	     n = rb_next(n)) {
		info = rb_entry(n, struct btrfs_free_space, offset_index);
		ret += info->bytes;
	}

	return ret;
}

1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602
/*
 * for a given cluster, put all of its extents back into the free
 * space cache.  If the block group passed doesn't match the block group
 * pointed to by the cluster, someone else raced in and freed the
 * cluster already.  In that case, we just return without changing anything
 */
static int
__btrfs_return_cluster_to_free_space(
			     struct btrfs_block_group_cache *block_group,
			     struct btrfs_free_cluster *cluster)
{
	struct btrfs_free_space *entry;
	struct rb_node *node;
1603
	bool bitmap;
1604 1605 1606 1607 1608

	spin_lock(&cluster->lock);
	if (cluster->block_group != block_group)
		goto out;

1609 1610
	bitmap = cluster->points_to_bitmap;
	cluster->block_group = NULL;
1611
	cluster->window_start = 0;
1612 1613 1614 1615 1616 1617
	list_del_init(&cluster->block_group_list);
	cluster->points_to_bitmap = false;

	if (bitmap)
		goto out;

1618
	node = rb_first(&cluster->root);
1619
	while (node) {
1620 1621 1622
		entry = rb_entry(node, struct btrfs_free_space, offset_index);
		node = rb_next(&entry->offset_index);
		rb_erase(&entry->offset_index, &cluster->root);
1623 1624 1625
		BUG_ON(entry->bitmap);
		tree_insert_offset(&block_group->free_space_offset,
				   entry->offset, &entry->offset_index, 0);
1626
	}
1627
	cluster->root = RB_ROOT;
1628

1629 1630
out:
	spin_unlock(&cluster->lock);
1631
	btrfs_put_block_group(block_group);
1632 1633 1634
	return 0;
}

J
Josef Bacik 已提交
1635 1636 1637 1638
void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group)
{
	struct btrfs_free_space *info;
	struct rb_node *node;
1639
	struct btrfs_free_cluster *cluster;
1640
	struct list_head *head;
J
Josef Bacik 已提交
1641

1642
	spin_lock(&block_group->tree_lock);
1643 1644 1645 1646
	while ((head = block_group->cluster_list.next) !=
	       &block_group->cluster_list) {
		cluster = list_entry(head, struct btrfs_free_cluster,
				     block_group_list);
1647 1648 1649

		WARN_ON(cluster->block_group != block_group);
		__btrfs_return_cluster_to_free_space(block_group, cluster);
1650 1651 1652 1653 1654
		if (need_resched()) {
			spin_unlock(&block_group->tree_lock);
			cond_resched();
			spin_lock(&block_group->tree_lock);
		}
1655 1656
	}

1657 1658
	while ((node = rb_last(&block_group->free_space_offset)) != NULL) {
		info = rb_entry(node, struct btrfs_free_space, offset_index);
J
Josef Bacik 已提交
1659
		unlink_free_space(block_group, info);
1660 1661
		if (info->bitmap)
			kfree(info->bitmap);
J
Josef Bacik 已提交
1662 1663
		kfree(info);
		if (need_resched()) {
1664
			spin_unlock(&block_group->tree_lock);
J
Josef Bacik 已提交
1665
			cond_resched();
1666
			spin_lock(&block_group->tree_lock);
J
Josef Bacik 已提交
1667 1668
		}
	}
1669

1670
	spin_unlock(&block_group->tree_lock);
J
Josef Bacik 已提交
1671 1672
}

1673 1674
u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
			       u64 offset, u64 bytes, u64 empty_size)
J
Josef Bacik 已提交
1675
{
1676
	struct btrfs_free_space *entry = NULL;
1677
	u64 bytes_search = bytes + empty_size;
1678
	u64 ret = 0;
J
Josef Bacik 已提交
1679

1680
	spin_lock(&block_group->tree_lock);
1681
	entry = find_free_space(block_group, &offset, &bytes_search, 0);
1682
	if (!entry)
1683 1684 1685 1686
		goto out;

	ret = offset;
	if (entry->bitmap) {
J
Josef Bacik 已提交
1687
		bitmap_clear_bits(block_group, entry, offset, bytes);
1688 1689 1690 1691 1692 1693 1694 1695
		if (!entry->bytes) {
			unlink_free_space(block_group, entry);
			kfree(entry->bitmap);
			kfree(entry);
			block_group->total_bitmaps--;
			recalculate_thresholds(block_group);
		}
	} else {
1696 1697 1698 1699 1700 1701 1702 1703
		unlink_free_space(block_group, entry);
		entry->offset += bytes;
		entry->bytes -= bytes;
		if (!entry->bytes)
			kfree(entry);
		else
			link_free_space(block_group, entry);
	}
J
Josef Bacik 已提交
1704

1705 1706
out:
	spin_unlock(&block_group->tree_lock);
J
Josef Bacik 已提交
1707

J
Josef Bacik 已提交
1708 1709
	return ret;
}
1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750

/*
 * given a cluster, put all of its extents back into the free space
 * cache.  If a block group is passed, this function will only free
 * a cluster that belongs to the passed block group.
 *
 * Otherwise, it'll get a reference on the block group pointed to by the
 * cluster and remove the cluster from it.
 */
int btrfs_return_cluster_to_free_space(
			       struct btrfs_block_group_cache *block_group,
			       struct btrfs_free_cluster *cluster)
{
	int ret;

	/* first, get a safe pointer to the block group */
	spin_lock(&cluster->lock);
	if (!block_group) {
		block_group = cluster->block_group;
		if (!block_group) {
			spin_unlock(&cluster->lock);
			return 0;
		}
	} else if (cluster->block_group != block_group) {
		/* someone else has already freed it don't redo their work */
		spin_unlock(&cluster->lock);
		return 0;
	}
	atomic_inc(&block_group->count);
	spin_unlock(&cluster->lock);

	/* now return any extents the cluster had on it */
	spin_lock(&block_group->tree_lock);
	ret = __btrfs_return_cluster_to_free_space(block_group, cluster);
	spin_unlock(&block_group->tree_lock);

	/* finally drop our ref */
	btrfs_put_block_group(block_group);
	return ret;
}

1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769
static u64 btrfs_alloc_from_bitmap(struct btrfs_block_group_cache *block_group,
				   struct btrfs_free_cluster *cluster,
				   u64 bytes, u64 min_start)
{
	struct btrfs_free_space *entry;
	int err;
	u64 search_start = cluster->window_start;
	u64 search_bytes = bytes;
	u64 ret = 0;

	spin_lock(&block_group->tree_lock);
	spin_lock(&cluster->lock);

	if (!cluster->points_to_bitmap)
		goto out;

	if (cluster->block_group != block_group)
		goto out;

1770 1771 1772 1773 1774 1775 1776 1777 1778
	/*
	 * search_start is the beginning of the bitmap, but at some point it may
	 * be a good idea to point to the actual start of the free area in the
	 * bitmap, so do the offset_to_bitmap trick anyway, and set bitmap_only
	 * to 1 to make sure we get the bitmap entry
	 */
	entry = tree_search_offset(block_group,
				   offset_to_bitmap(block_group, search_start),
				   1, 0);
1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790
	if (!entry || !entry->bitmap)
		goto out;

	search_start = min_start;
	search_bytes = bytes;

	err = search_bitmap(block_group, entry, &search_start,
			    &search_bytes);
	if (err)
		goto out;

	ret = search_start;
J
Josef Bacik 已提交
1791
	bitmap_clear_bits(block_group, entry, ret, bytes);
1792 1793 1794 1795 1796 1797 1798
out:
	spin_unlock(&cluster->lock);
	spin_unlock(&block_group->tree_lock);

	return ret;
}

1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811
/*
 * given a cluster, try to allocate 'bytes' from it, returns 0
 * if it couldn't find anything suitably large, or a logical disk offset
 * if things worked out
 */
u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
			     struct btrfs_free_cluster *cluster, u64 bytes,
			     u64 min_start)
{
	struct btrfs_free_space *entry = NULL;
	struct rb_node *node;
	u64 ret = 0;

1812 1813 1814 1815
	if (cluster->points_to_bitmap)
		return btrfs_alloc_from_bitmap(block_group, cluster, bytes,
					       min_start);

1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852
	spin_lock(&cluster->lock);
	if (bytes > cluster->max_size)
		goto out;

	if (cluster->block_group != block_group)
		goto out;

	node = rb_first(&cluster->root);
	if (!node)
		goto out;

	entry = rb_entry(node, struct btrfs_free_space, offset_index);

	while(1) {
		if (entry->bytes < bytes || entry->offset < min_start) {
			struct rb_node *node;

			node = rb_next(&entry->offset_index);
			if (!node)
				break;
			entry = rb_entry(node, struct btrfs_free_space,
					 offset_index);
			continue;
		}
		ret = entry->offset;

		entry->offset += bytes;
		entry->bytes -= bytes;

		if (entry->bytes == 0) {
			rb_erase(&entry->offset_index, &cluster->root);
			kfree(entry);
		}
		break;
	}
out:
	spin_unlock(&cluster->lock);
1853

1854 1855 1856
	return ret;
}

1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919
static int btrfs_bitmap_cluster(struct btrfs_block_group_cache *block_group,
				struct btrfs_free_space *entry,
				struct btrfs_free_cluster *cluster,
				u64 offset, u64 bytes, u64 min_bytes)
{
	unsigned long next_zero;
	unsigned long i;
	unsigned long search_bits;
	unsigned long total_bits;
	unsigned long found_bits;
	unsigned long start = 0;
	unsigned long total_found = 0;
	bool found = false;

	i = offset_to_bit(entry->offset, block_group->sectorsize,
			  max_t(u64, offset, entry->offset));
	search_bits = bytes_to_bits(min_bytes, block_group->sectorsize);
	total_bits = bytes_to_bits(bytes, block_group->sectorsize);

again:
	found_bits = 0;
	for (i = find_next_bit(entry->bitmap, BITS_PER_BITMAP, i);
	     i < BITS_PER_BITMAP;
	     i = find_next_bit(entry->bitmap, BITS_PER_BITMAP, i + 1)) {
		next_zero = find_next_zero_bit(entry->bitmap,
					       BITS_PER_BITMAP, i);
		if (next_zero - i >= search_bits) {
			found_bits = next_zero - i;
			break;
		}
		i = next_zero;
	}

	if (!found_bits)
		return -1;

	if (!found) {
		start = i;
		found = true;
	}

	total_found += found_bits;

	if (cluster->max_size < found_bits * block_group->sectorsize)
		cluster->max_size = found_bits * block_group->sectorsize;

	if (total_found < total_bits) {
		i = find_next_bit(entry->bitmap, BITS_PER_BITMAP, next_zero);
		if (i - start > total_bits * 2) {
			total_found = 0;
			cluster->max_size = 0;
			found = false;
		}
		goto again;
	}

	cluster->window_start = start * block_group->sectorsize +
		entry->offset;
	cluster->points_to_bitmap = true;

	return 0;
}

1920 1921 1922 1923 1924 1925 1926 1927 1928
/*
 * here we try to find a cluster of blocks in a block group.  The goal
 * is to find at least bytes free and up to empty_size + bytes free.
 * We might not find them all in one contiguous area.
 *
 * returns zero and sets up cluster if things worked out, otherwise
 * it returns -enospc
 */
int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
1929
			     struct btrfs_root *root,
1930 1931 1932 1933 1934 1935 1936
			     struct btrfs_block_group_cache *block_group,
			     struct btrfs_free_cluster *cluster,
			     u64 offset, u64 bytes, u64 empty_size)
{
	struct btrfs_free_space *entry = NULL;
	struct rb_node *node;
	struct btrfs_free_space *next;
1937
	struct btrfs_free_space *last = NULL;
1938 1939 1940 1941
	u64 min_bytes;
	u64 window_start;
	u64 window_free;
	u64 max_extent = 0;
1942
	bool found_bitmap = false;
1943 1944 1945
	int ret;

	/* for metadata, allow allocates with more holes */
1946 1947 1948
	if (btrfs_test_opt(root, SSD_SPREAD)) {
		min_bytes = bytes + empty_size;
	} else if (block_group->flags & BTRFS_BLOCK_GROUP_METADATA) {
1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969
		/*
		 * we want to do larger allocations when we are
		 * flushing out the delayed refs, it helps prevent
		 * making more work as we go along.
		 */
		if (trans->transaction->delayed_refs.flushing)
			min_bytes = max(bytes, (bytes + empty_size) >> 1);
		else
			min_bytes = max(bytes, (bytes + empty_size) >> 4);
	} else
		min_bytes = max(bytes, (bytes + empty_size) >> 2);

	spin_lock(&block_group->tree_lock);
	spin_lock(&cluster->lock);

	/* someone already found a cluster, hooray */
	if (cluster->block_group) {
		ret = 0;
		goto out;
	}
again:
1970
	entry = tree_search_offset(block_group, offset, found_bitmap, 1);
1971 1972 1973 1974
	if (!entry) {
		ret = -ENOSPC;
		goto out;
	}
1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

	/*
	 * If found_bitmap is true, we exhausted our search for extent entries,
	 * and we just want to search all of the bitmaps that we can find, and
	 * ignore any extent entries we find.
	 */
	while (entry->bitmap || found_bitmap ||
	       (!entry->bitmap && entry->bytes < min_bytes)) {
		struct rb_node *node = rb_next(&entry->offset_index);

		if (entry->bitmap && entry->bytes > bytes + empty_size) {
			ret = btrfs_bitmap_cluster(block_group, entry, cluster,
						   offset, bytes + empty_size,
						   min_bytes);
			if (!ret)
				goto got_it;
		}

		if (!node) {
			ret = -ENOSPC;
			goto out;
		}
		entry = rb_entry(node, struct btrfs_free_space, offset_index);
	}

	/*
	 * We already searched all the extent entries from the passed in offset
	 * to the end and didn't find enough space for the cluster, and we also
	 * didn't find any bitmaps that met our criteria, just go ahead and exit
	 */
	if (found_bitmap) {
		ret = -ENOSPC;
		goto out;
	}

	cluster->points_to_bitmap = false;
2011 2012 2013 2014 2015
	window_start = entry->offset;
	window_free = entry->bytes;
	last = entry;
	max_extent = entry->bytes;

2016
	while (1) {
2017 2018 2019 2020 2021 2022
		/* out window is just right, lets fill it */
		if (window_free >= bytes + empty_size)
			break;

		node = rb_next(&last->offset_index);
		if (!node) {
2023 2024
			if (found_bitmap)
				goto again;
2025 2026 2027 2028 2029
			ret = -ENOSPC;
			goto out;
		}
		next = rb_entry(node, struct btrfs_free_space, offset_index);

2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042
		/*
		 * we found a bitmap, so if this search doesn't result in a
		 * cluster, we know to go and search again for the bitmaps and
		 * start looking for space there
		 */
		if (next->bitmap) {
			if (!found_bitmap)
				offset = next->offset;
			found_bitmap = true;
			last = next;
			continue;
		}

2043 2044 2045 2046
		/*
		 * we haven't filled the empty size and the window is
		 * very large.  reset and try again
		 */
2047 2048
		if (next->offset - (last->offset + last->bytes) > 128 * 1024 ||
		    next->offset - window_start > (bytes + empty_size) * 2) {
2049 2050 2051 2052
			entry = next;
			window_start = entry->offset;
			window_free = entry->bytes;
			last = entry;
2053
			max_extent = entry->bytes;
2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070
		} else {
			last = next;
			window_free += next->bytes;
			if (entry->bytes > max_extent)
				max_extent = entry->bytes;
		}
	}

	cluster->window_start = entry->offset;

	/*
	 * now we've found our entries, pull them out of the free space
	 * cache and put them into the cluster rbtree
	 *
	 * The cluster includes an rbtree, but only uses the offset index
	 * of each free space cache entry.
	 */
2071
	while (1) {
2072
		node = rb_next(&entry->offset_index);
2073 2074 2075 2076 2077 2078 2079 2080 2081
		if (entry->bitmap && node) {
			entry = rb_entry(node, struct btrfs_free_space,
					 offset_index);
			continue;
		} else if (entry->bitmap && !node) {
			break;
		}

		rb_erase(&entry->offset_index, &block_group->free_space_offset);
2082
		ret = tree_insert_offset(&cluster->root, entry->offset,
2083
					 &entry->offset_index, 0);
2084 2085 2086 2087 2088 2089 2090
		BUG_ON(ret);

		if (!node || entry == last)
			break;

		entry = rb_entry(node, struct btrfs_free_space, offset_index);
	}
2091

2092
	cluster->max_size = max_extent;
2093 2094
got_it:
	ret = 0;
2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111
	atomic_inc(&block_group->count);
	list_add_tail(&cluster->block_group_list, &block_group->cluster_list);
	cluster->block_group = block_group;
out:
	spin_unlock(&cluster->lock);
	spin_unlock(&block_group->tree_lock);

	return ret;
}

/*
 * simple code to zero out a cluster
 */
void btrfs_init_free_cluster(struct btrfs_free_cluster *cluster)
{
	spin_lock_init(&cluster->lock);
	spin_lock_init(&cluster->refill_lock);
2112
	cluster->root = RB_ROOT;
2113
	cluster->max_size = 0;
2114
	cluster->points_to_bitmap = false;
2115 2116 2117 2118
	INIT_LIST_HEAD(&cluster->block_group_list);
	cluster->block_group = NULL;
}