extent-tree.c 210.7 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
/*
 * 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.
 */
Z
Zach Brown 已提交
18
#include <linux/sched.h>
19
#include <linux/pagemap.h>
20
#include <linux/writeback.h>
21
#include <linux/blkdev.h>
22
#include <linux/sort.h>
23
#include <linux/rcupdate.h>
J
Josef Bacik 已提交
24
#include <linux/kthread.h>
25
#include <linux/slab.h>
26
#include <linux/ratelimit.h>
C
Chris Mason 已提交
27
#include "compat.h"
28
#include "hash.h"
29 30 31
#include "ctree.h"
#include "disk-io.h"
#include "print-tree.h"
32
#include "transaction.h"
33
#include "volumes.h"
34
#include "locking.h"
35
#include "free-space-cache.h"
36

37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55
/* control flags for do_chunk_alloc's force field
 * CHUNK_ALLOC_NO_FORCE means to only allocate a chunk
 * if we really need one.
 *
 * CHUNK_ALLOC_FORCE means it must try to allocate one
 *
 * CHUNK_ALLOC_LIMITED means to only try and allocate one
 * if we have very few chunks already allocated.  This is
 * used as part of the clustering code to help make sure
 * we have a good pool of storage to cluster in, without
 * filling the FS with empty chunks
 *
 */
enum {
	CHUNK_ALLOC_NO_FORCE = 0,
	CHUNK_ALLOC_FORCE = 1,
	CHUNK_ALLOC_LIMITED = 2,
};

56 57 58 59 60 61 62 63 64 65 66 67 68 69 70
/*
 * Control how reservations are dealt with.
 *
 * RESERVE_FREE - freeing a reservation.
 * RESERVE_ALLOC - allocating space and we need to update bytes_may_use for
 *   ENOSPC accounting
 * RESERVE_ALLOC_NO_ACCOUNT - allocating space and we should not update
 *   bytes_may_use as the ENOSPC accounting is done elsewhere
 */
enum {
	RESERVE_FREE = 0,
	RESERVE_ALLOC = 1,
	RESERVE_ALLOC_NO_ACCOUNT = 2,
};

71 72
static int update_block_group(struct btrfs_trans_handle *trans,
			      struct btrfs_root *root,
73
			      u64 bytenr, u64 num_bytes, int alloc);
74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92
static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
				struct btrfs_root *root,
				u64 bytenr, u64 num_bytes, u64 parent,
				u64 root_objectid, u64 owner_objectid,
				u64 owner_offset, int refs_to_drop,
				struct btrfs_delayed_extent_op *extra_op);
static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
				    struct extent_buffer *leaf,
				    struct btrfs_extent_item *ei);
static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
				      struct btrfs_root *root,
				      u64 parent, u64 root_objectid,
				      u64 flags, u64 owner, u64 offset,
				      struct btrfs_key *ins, int ref_mod);
static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root,
				     u64 parent, u64 root_objectid,
				     u64 flags, struct btrfs_disk_key *key,
				     int level, struct btrfs_key *ins);
J
Josef Bacik 已提交
93 94 95
static int do_chunk_alloc(struct btrfs_trans_handle *trans,
			  struct btrfs_root *extent_root, u64 alloc_bytes,
			  u64 flags, int force);
96 97
static int find_next_key(struct btrfs_path *path, int level,
			 struct btrfs_key *key);
J
Josef Bacik 已提交
98 99
static void dump_space_info(struct btrfs_space_info *info, u64 bytes,
			    int dump_block_groups);
100 101
static int btrfs_update_reserved_bytes(struct btrfs_block_group_cache *cache,
				       u64 num_bytes, int reserve);
J
Josef Bacik 已提交
102

J
Josef Bacik 已提交
103 104 105 106 107 108 109
static noinline int
block_group_cache_done(struct btrfs_block_group_cache *cache)
{
	smp_mb();
	return cache->cached == BTRFS_CACHE_FINISHED;
}

J
Josef Bacik 已提交
110 111 112 113 114
static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
{
	return (cache->flags & bits) == bits;
}

115
static void btrfs_get_block_group(struct btrfs_block_group_cache *cache)
116 117 118 119 120 121
{
	atomic_inc(&cache->count);
}

void btrfs_put_block_group(struct btrfs_block_group_cache *cache)
{
122 123 124
	if (atomic_dec_and_test(&cache->count)) {
		WARN_ON(cache->pinned > 0);
		WARN_ON(cache->reserved > 0);
125
		kfree(cache->free_space_ctl);
126
		kfree(cache);
127
	}
128 129
}

J
Josef Bacik 已提交
130 131 132 133
/*
 * this adds the block group to the fs_info rb tree for the block group
 * cache
 */
134
static int btrfs_add_block_group_cache(struct btrfs_fs_info *info,
J
Josef Bacik 已提交
135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201
				struct btrfs_block_group_cache *block_group)
{
	struct rb_node **p;
	struct rb_node *parent = NULL;
	struct btrfs_block_group_cache *cache;

	spin_lock(&info->block_group_cache_lock);
	p = &info->block_group_cache_tree.rb_node;

	while (*p) {
		parent = *p;
		cache = rb_entry(parent, struct btrfs_block_group_cache,
				 cache_node);
		if (block_group->key.objectid < cache->key.objectid) {
			p = &(*p)->rb_left;
		} else if (block_group->key.objectid > cache->key.objectid) {
			p = &(*p)->rb_right;
		} else {
			spin_unlock(&info->block_group_cache_lock);
			return -EEXIST;
		}
	}

	rb_link_node(&block_group->cache_node, parent, p);
	rb_insert_color(&block_group->cache_node,
			&info->block_group_cache_tree);
	spin_unlock(&info->block_group_cache_lock);

	return 0;
}

/*
 * This will return the block group at or after bytenr if contains is 0, else
 * it will return the block group that contains the bytenr
 */
static struct btrfs_block_group_cache *
block_group_cache_tree_search(struct btrfs_fs_info *info, u64 bytenr,
			      int contains)
{
	struct btrfs_block_group_cache *cache, *ret = NULL;
	struct rb_node *n;
	u64 end, start;

	spin_lock(&info->block_group_cache_lock);
	n = info->block_group_cache_tree.rb_node;

	while (n) {
		cache = rb_entry(n, struct btrfs_block_group_cache,
				 cache_node);
		end = cache->key.objectid + cache->key.offset - 1;
		start = cache->key.objectid;

		if (bytenr < start) {
			if (!contains && (!ret || start < ret->key.objectid))
				ret = cache;
			n = n->rb_left;
		} else if (bytenr > start) {
			if (contains && bytenr <= end) {
				ret = cache;
				break;
			}
			n = n->rb_right;
		} else {
			ret = cache;
			break;
		}
	}
202
	if (ret)
203
		btrfs_get_block_group(ret);
J
Josef Bacik 已提交
204 205 206 207 208
	spin_unlock(&info->block_group_cache_lock);

	return ret;
}

209 210
static int add_excluded_extent(struct btrfs_root *root,
			       u64 start, u64 num_bytes)
J
Josef Bacik 已提交
211
{
212 213 214 215 216 217 218
	u64 end = start + num_bytes - 1;
	set_extent_bits(&root->fs_info->freed_extents[0],
			start, end, EXTENT_UPTODATE, GFP_NOFS);
	set_extent_bits(&root->fs_info->freed_extents[1],
			start, end, EXTENT_UPTODATE, GFP_NOFS);
	return 0;
}
J
Josef Bacik 已提交
219

220 221 222 223
static void free_excluded_extents(struct btrfs_root *root,
				  struct btrfs_block_group_cache *cache)
{
	u64 start, end;
J
Josef Bacik 已提交
224

225 226 227 228 229 230 231
	start = cache->key.objectid;
	end = start + cache->key.offset - 1;

	clear_extent_bits(&root->fs_info->freed_extents[0],
			  start, end, EXTENT_UPTODATE, GFP_NOFS);
	clear_extent_bits(&root->fs_info->freed_extents[1],
			  start, end, EXTENT_UPTODATE, GFP_NOFS);
J
Josef Bacik 已提交
232 233
}

234 235
static int exclude_super_stripes(struct btrfs_root *root,
				 struct btrfs_block_group_cache *cache)
J
Josef Bacik 已提交
236 237 238 239 240 241
{
	u64 bytenr;
	u64 *logical;
	int stripe_len;
	int i, nr, ret;

242 243 244 245 246 247 248 249
	if (cache->key.objectid < BTRFS_SUPER_INFO_OFFSET) {
		stripe_len = BTRFS_SUPER_INFO_OFFSET - cache->key.objectid;
		cache->bytes_super += stripe_len;
		ret = add_excluded_extent(root, cache->key.objectid,
					  stripe_len);
		BUG_ON(ret);
	}

J
Josef Bacik 已提交
250 251 252 253 254 255
	for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
		bytenr = btrfs_sb_offset(i);
		ret = btrfs_rmap_block(&root->fs_info->mapping_tree,
				       cache->key.objectid, bytenr,
				       0, &logical, &nr, &stripe_len);
		BUG_ON(ret);
256

J
Josef Bacik 已提交
257
		while (nr--) {
258
			cache->bytes_super += stripe_len;
259 260 261
			ret = add_excluded_extent(root, logical[nr],
						  stripe_len);
			BUG_ON(ret);
J
Josef Bacik 已提交
262
		}
263

J
Josef Bacik 已提交
264 265 266 267 268
		kfree(logical);
	}
	return 0;
}

269 270 271 272 273 274 275 276 277 278 279
static struct btrfs_caching_control *
get_caching_control(struct btrfs_block_group_cache *cache)
{
	struct btrfs_caching_control *ctl;

	spin_lock(&cache->lock);
	if (cache->cached != BTRFS_CACHE_STARTED) {
		spin_unlock(&cache->lock);
		return NULL;
	}

280 281 282
	/* We're loading it the fast way, so we don't have a caching_ctl. */
	if (!cache->caching_ctl) {
		spin_unlock(&cache->lock);
283 284 285 286 287 288 289 290 291 292 293 294 295 296 297
		return NULL;
	}

	ctl = cache->caching_ctl;
	atomic_inc(&ctl->count);
	spin_unlock(&cache->lock);
	return ctl;
}

static void put_caching_control(struct btrfs_caching_control *ctl)
{
	if (atomic_dec_and_test(&ctl->count))
		kfree(ctl);
}

J
Josef Bacik 已提交
298 299 300 301 302
/*
 * this is only called by cache_block_group, since we could have freed extents
 * we need to check the pinned_extents for any extents that can't be used yet
 * since their free space will be released as soon as the transaction commits.
 */
J
Josef Bacik 已提交
303
static u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
J
Josef Bacik 已提交
304 305
			      struct btrfs_fs_info *info, u64 start, u64 end)
{
J
Josef Bacik 已提交
306
	u64 extent_start, extent_end, size, total_added = 0;
J
Josef Bacik 已提交
307 308 309
	int ret;

	while (start < end) {
310
		ret = find_first_extent_bit(info->pinned_extents, start,
J
Josef Bacik 已提交
311
					    &extent_start, &extent_end,
312
					    EXTENT_DIRTY | EXTENT_UPTODATE);
J
Josef Bacik 已提交
313 314 315
		if (ret)
			break;

316
		if (extent_start <= start) {
J
Josef Bacik 已提交
317 318 319
			start = extent_end + 1;
		} else if (extent_start > start && extent_start < end) {
			size = extent_start - start;
J
Josef Bacik 已提交
320
			total_added += size;
321 322
			ret = btrfs_add_free_space(block_group, start,
						   size);
J
Josef Bacik 已提交
323 324 325 326 327 328 329 330 331
			BUG_ON(ret);
			start = extent_end + 1;
		} else {
			break;
		}
	}

	if (start < end) {
		size = end - start;
J
Josef Bacik 已提交
332
		total_added += size;
333
		ret = btrfs_add_free_space(block_group, start, size);
J
Josef Bacik 已提交
334 335 336
		BUG_ON(ret);
	}

J
Josef Bacik 已提交
337
	return total_added;
J
Josef Bacik 已提交
338 339
}

340
static noinline void caching_thread(struct btrfs_work *work)
341
{
342 343 344 345
	struct btrfs_block_group_cache *block_group;
	struct btrfs_fs_info *fs_info;
	struct btrfs_caching_control *caching_ctl;
	struct btrfs_root *extent_root;
346
	struct btrfs_path *path;
347
	struct extent_buffer *leaf;
348
	struct btrfs_key key;
J
Josef Bacik 已提交
349
	u64 total_found = 0;
350 351 352
	u64 last = 0;
	u32 nritems;
	int ret = 0;
353

354 355 356 357 358
	caching_ctl = container_of(work, struct btrfs_caching_control, work);
	block_group = caching_ctl->block_group;
	fs_info = block_group->fs_info;
	extent_root = fs_info->extent_root;

359 360
	path = btrfs_alloc_path();
	if (!path)
361
		goto out;
362

J
Josef Bacik 已提交
363
	last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
364

365
	/*
J
Josef Bacik 已提交
366 367 368 369
	 * We don't want to deadlock with somebody trying to allocate a new
	 * extent for the extent root while also trying to search the extent
	 * root to add free space.  So we skip locking and search the commit
	 * root, since its read-only
370 371
	 */
	path->skip_locking = 1;
J
Josef Bacik 已提交
372
	path->search_commit_root = 1;
J
Josef Bacik 已提交
373
	path->reada = 1;
J
Josef Bacik 已提交
374

Y
Yan Zheng 已提交
375
	key.objectid = last;
376
	key.offset = 0;
377
	key.type = BTRFS_EXTENT_ITEM_KEY;
378
again:
379
	mutex_lock(&caching_ctl->mutex);
380 381 382
	/* need to make sure the commit_root doesn't disappear */
	down_read(&fs_info->extent_commit_sem);

383
	ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
384
	if (ret < 0)
385
		goto err;
Y
Yan Zheng 已提交
386

387 388 389
	leaf = path->nodes[0];
	nritems = btrfs_header_nritems(leaf);

C
Chris Mason 已提交
390
	while (1) {
391
		if (btrfs_fs_closing(fs_info) > 1) {
392
			last = (u64)-1;
J
Josef Bacik 已提交
393
			break;
394
		}
J
Josef Bacik 已提交
395

396 397 398 399 400
		if (path->slots[0] < nritems) {
			btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
		} else {
			ret = find_next_key(path, 0, &key);
			if (ret)
401
				break;
J
Josef Bacik 已提交
402

403 404 405
			if (need_resched() ||
			    btrfs_next_leaf(extent_root, path)) {
				caching_ctl->progress = last;
C
Chris Mason 已提交
406
				btrfs_release_path(path);
407 408
				up_read(&fs_info->extent_commit_sem);
				mutex_unlock(&caching_ctl->mutex);
409
				cond_resched();
410 411 412 413 414
				goto again;
			}
			leaf = path->nodes[0];
			nritems = btrfs_header_nritems(leaf);
			continue;
415
		}
J
Josef Bacik 已提交
416

417 418
		if (key.objectid < block_group->key.objectid) {
			path->slots[0]++;
J
Josef Bacik 已提交
419
			continue;
420
		}
J
Josef Bacik 已提交
421

422
		if (key.objectid >= block_group->key.objectid +
J
Josef Bacik 已提交
423
		    block_group->key.offset)
424
			break;
425

426
		if (key.type == BTRFS_EXTENT_ITEM_KEY) {
J
Josef Bacik 已提交
427 428 429
			total_found += add_new_free_space(block_group,
							  fs_info, last,
							  key.objectid);
430
			last = key.objectid + key.offset;
J
Josef Bacik 已提交
431

432 433 434 435
			if (total_found > (1024 * 1024 * 2)) {
				total_found = 0;
				wake_up(&caching_ctl->wait);
			}
J
Josef Bacik 已提交
436
		}
437 438
		path->slots[0]++;
	}
J
Josef Bacik 已提交
439
	ret = 0;
440

J
Josef Bacik 已提交
441 442 443
	total_found += add_new_free_space(block_group, fs_info, last,
					  block_group->key.objectid +
					  block_group->key.offset);
444
	caching_ctl->progress = (u64)-1;
J
Josef Bacik 已提交
445 446

	spin_lock(&block_group->lock);
447
	block_group->caching_ctl = NULL;
J
Josef Bacik 已提交
448 449
	block_group->cached = BTRFS_CACHE_FINISHED;
	spin_unlock(&block_group->lock);
J
Josef Bacik 已提交
450

451
err:
452
	btrfs_free_path(path);
453
	up_read(&fs_info->extent_commit_sem);
J
Josef Bacik 已提交
454

455 456 457
	free_excluded_extents(extent_root, block_group);

	mutex_unlock(&caching_ctl->mutex);
458
out:
459 460 461
	wake_up(&caching_ctl->wait);

	put_caching_control(caching_ctl);
462
	btrfs_put_block_group(block_group);
J
Josef Bacik 已提交
463 464
}

465 466
static int cache_block_group(struct btrfs_block_group_cache *cache,
			     struct btrfs_trans_handle *trans,
467
			     struct btrfs_root *root,
468
			     int load_cache_only)
J
Josef Bacik 已提交
469
{
470
	DEFINE_WAIT(wait);
471 472
	struct btrfs_fs_info *fs_info = cache->fs_info;
	struct btrfs_caching_control *caching_ctl;
J
Josef Bacik 已提交
473 474
	int ret = 0;

475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516
	caching_ctl = kzalloc(sizeof(*caching_ctl), GFP_NOFS);
	BUG_ON(!caching_ctl);

	INIT_LIST_HEAD(&caching_ctl->list);
	mutex_init(&caching_ctl->mutex);
	init_waitqueue_head(&caching_ctl->wait);
	caching_ctl->block_group = cache;
	caching_ctl->progress = cache->key.objectid;
	atomic_set(&caching_ctl->count, 1);
	caching_ctl->work.func = caching_thread;

	spin_lock(&cache->lock);
	/*
	 * This should be a rare occasion, but this could happen I think in the
	 * case where one thread starts to load the space cache info, and then
	 * some other thread starts a transaction commit which tries to do an
	 * allocation while the other thread is still loading the space cache
	 * info.  The previous loop should have kept us from choosing this block
	 * group, but if we've moved to the state where we will wait on caching
	 * block groups we need to first check if we're doing a fast load here,
	 * so we can wait for it to finish, otherwise we could end up allocating
	 * from a block group who's cache gets evicted for one reason or
	 * another.
	 */
	while (cache->cached == BTRFS_CACHE_FAST) {
		struct btrfs_caching_control *ctl;

		ctl = cache->caching_ctl;
		atomic_inc(&ctl->count);
		prepare_to_wait(&ctl->wait, &wait, TASK_UNINTERRUPTIBLE);
		spin_unlock(&cache->lock);

		schedule();

		finish_wait(&ctl->wait, &wait);
		put_caching_control(ctl);
		spin_lock(&cache->lock);
	}

	if (cache->cached != BTRFS_CACHE_NO) {
		spin_unlock(&cache->lock);
		kfree(caching_ctl);
517
		return 0;
518 519 520 521 522
	}
	WARN_ON(cache->caching_ctl);
	cache->caching_ctl = caching_ctl;
	cache->cached = BTRFS_CACHE_FAST;
	spin_unlock(&cache->lock);
523

524 525
	/*
	 * We can't do the read from on-disk cache during a commit since we need
526 527 528
	 * to have the normal tree locking.  Also if we are currently trying to
	 * allocate blocks for the tree root we can't do the fast caching since
	 * we likely hold important locks.
529
	 */
530
	if (trans && (!trans->transaction->in_commit) &&
531 532
	    (root && root != root->fs_info->tree_root) &&
	    btrfs_test_opt(root, SPACE_CACHE)) {
533 534 535 536
		ret = load_free_space_cache(fs_info, cache);

		spin_lock(&cache->lock);
		if (ret == 1) {
537
			cache->caching_ctl = NULL;
538 539 540
			cache->cached = BTRFS_CACHE_FINISHED;
			cache->last_byte_to_unpin = (u64)-1;
		} else {
541 542 543 544 545 546
			if (load_cache_only) {
				cache->caching_ctl = NULL;
				cache->cached = BTRFS_CACHE_NO;
			} else {
				cache->cached = BTRFS_CACHE_STARTED;
			}
547 548
		}
		spin_unlock(&cache->lock);
549
		wake_up(&caching_ctl->wait);
550
		if (ret == 1) {
551
			put_caching_control(caching_ctl);
552
			free_excluded_extents(fs_info->extent_root, cache);
553
			return 0;
554
		}
555 556 557 558 559 560 561 562 563 564 565 566 567 568
	} else {
		/*
		 * We are not going to do the fast caching, set cached to the
		 * appropriate value and wakeup any waiters.
		 */
		spin_lock(&cache->lock);
		if (load_cache_only) {
			cache->caching_ctl = NULL;
			cache->cached = BTRFS_CACHE_NO;
		} else {
			cache->cached = BTRFS_CACHE_STARTED;
		}
		spin_unlock(&cache->lock);
		wake_up(&caching_ctl->wait);
569 570
	}

571 572
	if (load_cache_only) {
		put_caching_control(caching_ctl);
573
		return 0;
J
Josef Bacik 已提交
574 575
	}

576
	down_write(&fs_info->extent_commit_sem);
577
	atomic_inc(&caching_ctl->count);
578 579 580
	list_add_tail(&caching_ctl->list, &fs_info->caching_block_groups);
	up_write(&fs_info->extent_commit_sem);

581
	btrfs_get_block_group(cache);
582

583
	btrfs_queue_worker(&fs_info->caching_workers, &caching_ctl->work);
J
Josef Bacik 已提交
584

585
	return ret;
586 587
}

J
Josef Bacik 已提交
588 589 590
/*
 * return the block group that starts at or after bytenr
 */
C
Chris Mason 已提交
591 592
static struct btrfs_block_group_cache *
btrfs_lookup_first_block_group(struct btrfs_fs_info *info, u64 bytenr)
C
Chris Mason 已提交
593
{
J
Josef Bacik 已提交
594
	struct btrfs_block_group_cache *cache;
C
Chris Mason 已提交
595

J
Josef Bacik 已提交
596
	cache = block_group_cache_tree_search(info, bytenr, 0);
C
Chris Mason 已提交
597

J
Josef Bacik 已提交
598
	return cache;
C
Chris Mason 已提交
599 600
}

J
Josef Bacik 已提交
601
/*
602
 * return the block group that contains the given bytenr
J
Josef Bacik 已提交
603
 */
C
Chris Mason 已提交
604 605 606
struct btrfs_block_group_cache *btrfs_lookup_block_group(
						 struct btrfs_fs_info *info,
						 u64 bytenr)
C
Chris Mason 已提交
607
{
J
Josef Bacik 已提交
608
	struct btrfs_block_group_cache *cache;
C
Chris Mason 已提交
609

J
Josef Bacik 已提交
610
	cache = block_group_cache_tree_search(info, bytenr, 1);
611

J
Josef Bacik 已提交
612
	return cache;
C
Chris Mason 已提交
613
}
614

J
Josef Bacik 已提交
615 616
static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info,
						  u64 flags)
617
{
J
Josef Bacik 已提交
618 619
	struct list_head *head = &info->space_info;
	struct btrfs_space_info *found;
620

621
	flags &= BTRFS_BLOCK_GROUP_TYPE_MASK;
622

623 624
	rcu_read_lock();
	list_for_each_entry_rcu(found, head, list) {
625
		if (found->flags & flags) {
626
			rcu_read_unlock();
J
Josef Bacik 已提交
627
			return found;
628
		}
J
Josef Bacik 已提交
629
	}
630
	rcu_read_unlock();
J
Josef Bacik 已提交
631
	return NULL;
632 633
}

634 635 636 637 638 639 640 641 642 643 644 645 646 647 648
/*
 * after adding space to the filesystem, we need to clear the full flags
 * on all the space infos.
 */
void btrfs_clear_space_info_full(struct btrfs_fs_info *info)
{
	struct list_head *head = &info->space_info;
	struct btrfs_space_info *found;

	rcu_read_lock();
	list_for_each_entry_rcu(found, head, list)
		found->full = 0;
	rcu_read_unlock();
}

649 650 651 652 653 654 655 656 657
static u64 div_factor(u64 num, int factor)
{
	if (factor == 10)
		return num;
	num *= factor;
	do_div(num, 10);
	return num;
}

658 659 660 661 662 663 664 665 666
static u64 div_factor_fine(u64 num, int factor)
{
	if (factor == 100)
		return num;
	num *= factor;
	do_div(num, 100);
	return num;
}

667 668
u64 btrfs_find_block_group(struct btrfs_root *root,
			   u64 search_start, u64 search_hint, int owner)
C
Chris Mason 已提交
669
{
670
	struct btrfs_block_group_cache *cache;
C
Chris Mason 已提交
671
	u64 used;
672 673
	u64 last = max(search_hint, search_start);
	u64 group_start = 0;
674
	int full_search = 0;
675
	int factor = 9;
C
Chris Mason 已提交
676
	int wrapped = 0;
677
again:
678 679
	while (1) {
		cache = btrfs_lookup_first_block_group(root->fs_info, last);
J
Josef Bacik 已提交
680 681
		if (!cache)
			break;
682

683
		spin_lock(&cache->lock);
684 685 686
		last = cache->key.objectid + cache->key.offset;
		used = btrfs_block_group_used(&cache->item);

687 688
		if ((full_search || !cache->ro) &&
		    block_group_bits(cache, BTRFS_BLOCK_GROUP_METADATA)) {
689
			if (used + cache->pinned + cache->reserved <
690 691
			    div_factor(cache->key.offset, factor)) {
				group_start = cache->key.objectid;
692
				spin_unlock(&cache->lock);
693
				btrfs_put_block_group(cache);
694 695
				goto found;
			}
696
		}
697
		spin_unlock(&cache->lock);
698
		btrfs_put_block_group(cache);
699
		cond_resched();
C
Chris Mason 已提交
700
	}
C
Chris Mason 已提交
701 702 703 704 705 706
	if (!wrapped) {
		last = search_start;
		wrapped = 1;
		goto again;
	}
	if (!full_search && factor < 10) {
C
Chris Mason 已提交
707
		last = search_start;
708
		full_search = 1;
C
Chris Mason 已提交
709
		factor = 10;
710 711
		goto again;
	}
C
Chris Mason 已提交
712
found:
713
	return group_start;
714
}
J
Josef Bacik 已提交
715

716
/* simple helper to search for an existing extent at a given offset */
Z
Zheng Yan 已提交
717
int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len)
718 719 720
{
	int ret;
	struct btrfs_key key;
Z
Zheng Yan 已提交
721
	struct btrfs_path *path;
722

Z
Zheng Yan 已提交
723
	path = btrfs_alloc_path();
724 725 726
	if (!path)
		return -ENOMEM;

727 728 729 730 731
	key.objectid = start;
	key.offset = len;
	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
	ret = btrfs_search_slot(NULL, root->fs_info->extent_root, &key, path,
				0, 0);
Z
Zheng Yan 已提交
732
	btrfs_free_path(path);
733 734 735
	return ret;
}

736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815
/*
 * helper function to lookup reference count and flags of extent.
 *
 * the head node for delayed ref is used to store the sum of all the
 * reference count modifications queued up in the rbtree. the head
 * node may also store the extent flags to set. This way you can check
 * to see what the reference count and extent flags would be if all of
 * the delayed refs are not processed.
 */
int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
			     struct btrfs_root *root, u64 bytenr,
			     u64 num_bytes, u64 *refs, u64 *flags)
{
	struct btrfs_delayed_ref_head *head;
	struct btrfs_delayed_ref_root *delayed_refs;
	struct btrfs_path *path;
	struct btrfs_extent_item *ei;
	struct extent_buffer *leaf;
	struct btrfs_key key;
	u32 item_size;
	u64 num_refs;
	u64 extent_flags;
	int ret;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

	key.objectid = bytenr;
	key.type = BTRFS_EXTENT_ITEM_KEY;
	key.offset = num_bytes;
	if (!trans) {
		path->skip_locking = 1;
		path->search_commit_root = 1;
	}
again:
	ret = btrfs_search_slot(trans, root->fs_info->extent_root,
				&key, path, 0, 0);
	if (ret < 0)
		goto out_free;

	if (ret == 0) {
		leaf = path->nodes[0];
		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
		if (item_size >= sizeof(*ei)) {
			ei = btrfs_item_ptr(leaf, path->slots[0],
					    struct btrfs_extent_item);
			num_refs = btrfs_extent_refs(leaf, ei);
			extent_flags = btrfs_extent_flags(leaf, ei);
		} else {
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
			struct btrfs_extent_item_v0 *ei0;
			BUG_ON(item_size != sizeof(*ei0));
			ei0 = btrfs_item_ptr(leaf, path->slots[0],
					     struct btrfs_extent_item_v0);
			num_refs = btrfs_extent_refs_v0(leaf, ei0);
			/* FIXME: this isn't correct for data */
			extent_flags = BTRFS_BLOCK_FLAG_FULL_BACKREF;
#else
			BUG();
#endif
		}
		BUG_ON(num_refs == 0);
	} else {
		num_refs = 0;
		extent_flags = 0;
		ret = 0;
	}

	if (!trans)
		goto out;

	delayed_refs = &trans->transaction->delayed_refs;
	spin_lock(&delayed_refs->lock);
	head = btrfs_find_delayed_ref_head(trans, bytenr);
	if (head) {
		if (!mutex_trylock(&head->mutex)) {
			atomic_inc(&head->node.refs);
			spin_unlock(&delayed_refs->lock);

816
			btrfs_release_path(path);
817

818 819 820 821
			/*
			 * Mutex was contended, block until it's released and try
			 * again
			 */
822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846
			mutex_lock(&head->mutex);
			mutex_unlock(&head->mutex);
			btrfs_put_delayed_ref(&head->node);
			goto again;
		}
		if (head->extent_op && head->extent_op->update_flags)
			extent_flags |= head->extent_op->flags_to_set;
		else
			BUG_ON(num_refs == 0);

		num_refs += head->node.ref_mod;
		mutex_unlock(&head->mutex);
	}
	spin_unlock(&delayed_refs->lock);
out:
	WARN_ON(num_refs == 0);
	if (refs)
		*refs = num_refs;
	if (flags)
		*flags = extent_flags;
out_free:
	btrfs_free_path(path);
	return ret;
}

847 848 849 850 851 852 853 854 855 856 857 858 859 860
/*
 * Back reference rules.  Back refs have three main goals:
 *
 * 1) differentiate between all holders of references to an extent so that
 *    when a reference is dropped we can make sure it was a valid reference
 *    before freeing the extent.
 *
 * 2) Provide enough information to quickly find the holders of an extent
 *    if we notice a given block is corrupted or bad.
 *
 * 3) Make it easy to migrate blocks for FS shrinking or storage pool
 *    maintenance.  This is actually the same as #2, but with a slightly
 *    different use case.
 *
861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905
 * There are two kinds of back refs. The implicit back refs is optimized
 * for pointers in non-shared tree blocks. For a given pointer in a block,
 * back refs of this kind provide information about the block's owner tree
 * and the pointer's key. These information allow us to find the block by
 * b-tree searching. The full back refs is for pointers in tree blocks not
 * referenced by their owner trees. The location of tree block is recorded
 * in the back refs. Actually the full back refs is generic, and can be
 * used in all cases the implicit back refs is used. The major shortcoming
 * of the full back refs is its overhead. Every time a tree block gets
 * COWed, we have to update back refs entry for all pointers in it.
 *
 * For a newly allocated tree block, we use implicit back refs for
 * pointers in it. This means most tree related operations only involve
 * implicit back refs. For a tree block created in old transaction, the
 * only way to drop a reference to it is COW it. So we can detect the
 * event that tree block loses its owner tree's reference and do the
 * back refs conversion.
 *
 * When a tree block is COW'd through a tree, there are four cases:
 *
 * The reference count of the block is one and the tree is the block's
 * owner tree. Nothing to do in this case.
 *
 * The reference count of the block is one and the tree is not the
 * block's owner tree. In this case, full back refs is used for pointers
 * in the block. Remove these full back refs, add implicit back refs for
 * every pointers in the new block.
 *
 * The reference count of the block is greater than one and the tree is
 * the block's owner tree. In this case, implicit back refs is used for
 * pointers in the block. Add full back refs for every pointers in the
 * block, increase lower level extents' reference counts. The original
 * implicit back refs are entailed to the new block.
 *
 * The reference count of the block is greater than one and the tree is
 * not the block's owner tree. Add implicit back refs for every pointer in
 * the new block, increase lower level extents' reference count.
 *
 * Back Reference Key composing:
 *
 * The key objectid corresponds to the first byte in the extent,
 * The key type is used to differentiate between types of back refs.
 * There are different meanings of the key offset for different types
 * of back refs.
 *
906 907 908
 * File extents can be referenced by:
 *
 * - multiple snapshots, subvolumes, or different generations in one subvol
Z
Zheng Yan 已提交
909
 * - different files inside a single subvolume
910 911
 * - different offsets inside a file (bookend extents in file.c)
 *
912
 * The extent ref structure for the implicit back refs has fields for:
913 914 915
 *
 * - Objectid of the subvolume root
 * - objectid of the file holding the reference
916 917
 * - original offset in the file
 * - how many bookend extents
918
 *
919 920
 * The key offset for the implicit back refs is hash of the first
 * three fields.
921
 *
922
 * The extent ref structure for the full back refs has field for:
923
 *
924
 * - number of pointers in the tree leaf
925
 *
926 927
 * The key offset for the implicit back refs is the first byte of
 * the tree leaf
928
 *
929 930
 * When a file extent is allocated, The implicit back refs is used.
 * the fields are filled in:
931
 *
932
 *     (root_key.objectid, inode objectid, offset in file, 1)
933
 *
934 935
 * When a file extent is removed file truncation, we find the
 * corresponding implicit back refs and check the following fields:
936
 *
937
 *     (btrfs_header_owner(leaf), inode objectid, offset in file)
938
 *
939
 * Btree extents can be referenced by:
940
 *
941
 * - Different subvolumes
942
 *
943 944 945 946
 * Both the implicit back refs and the full back refs for tree blocks
 * only consist of key. The key offset for the implicit back refs is
 * objectid of block's owner tree. The key offset for the full back refs
 * is the first byte of parent block.
947
 *
948 949 950
 * When implicit back refs is used, information about the lowest key and
 * level of the tree block are required. These information are stored in
 * tree block info structure.
951
 */
Z
Zheng Yan 已提交
952

953 954 955 956 957
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
static int convert_extent_item_v0(struct btrfs_trans_handle *trans,
				  struct btrfs_root *root,
				  struct btrfs_path *path,
				  u64 owner, u32 extra_size)
958
{
959 960 961 962 963
	struct btrfs_extent_item *item;
	struct btrfs_extent_item_v0 *ei0;
	struct btrfs_extent_ref_v0 *ref0;
	struct btrfs_tree_block_info *bi;
	struct extent_buffer *leaf;
964
	struct btrfs_key key;
965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999
	struct btrfs_key found_key;
	u32 new_size = sizeof(*item);
	u64 refs;
	int ret;

	leaf = path->nodes[0];
	BUG_ON(btrfs_item_size_nr(leaf, path->slots[0]) != sizeof(*ei0));

	btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
	ei0 = btrfs_item_ptr(leaf, path->slots[0],
			     struct btrfs_extent_item_v0);
	refs = btrfs_extent_refs_v0(leaf, ei0);

	if (owner == (u64)-1) {
		while (1) {
			if (path->slots[0] >= btrfs_header_nritems(leaf)) {
				ret = btrfs_next_leaf(root, path);
				if (ret < 0)
					return ret;
				BUG_ON(ret > 0);
				leaf = path->nodes[0];
			}
			btrfs_item_key_to_cpu(leaf, &found_key,
					      path->slots[0]);
			BUG_ON(key.objectid != found_key.objectid);
			if (found_key.type != BTRFS_EXTENT_REF_V0_KEY) {
				path->slots[0]++;
				continue;
			}
			ref0 = btrfs_item_ptr(leaf, path->slots[0],
					      struct btrfs_extent_ref_v0);
			owner = btrfs_ref_objectid_v0(leaf, ref0);
			break;
		}
	}
1000
	btrfs_release_path(path);
1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041

	if (owner < BTRFS_FIRST_FREE_OBJECTID)
		new_size += sizeof(*bi);

	new_size -= sizeof(*ei0);
	ret = btrfs_search_slot(trans, root, &key, path,
				new_size + extra_size, 1);
	if (ret < 0)
		return ret;
	BUG_ON(ret);

	ret = btrfs_extend_item(trans, root, path, new_size);

	leaf = path->nodes[0];
	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
	btrfs_set_extent_refs(leaf, item, refs);
	/* FIXME: get real generation */
	btrfs_set_extent_generation(leaf, item, 0);
	if (owner < BTRFS_FIRST_FREE_OBJECTID) {
		btrfs_set_extent_flags(leaf, item,
				       BTRFS_EXTENT_FLAG_TREE_BLOCK |
				       BTRFS_BLOCK_FLAG_FULL_BACKREF);
		bi = (struct btrfs_tree_block_info *)(item + 1);
		/* FIXME: get first key of the block */
		memset_extent_buffer(leaf, 0, (unsigned long)bi, sizeof(*bi));
		btrfs_set_tree_block_level(leaf, bi, (int)owner);
	} else {
		btrfs_set_extent_flags(leaf, item, BTRFS_EXTENT_FLAG_DATA);
	}
	btrfs_mark_buffer_dirty(leaf);
	return 0;
}
#endif

static u64 hash_extent_data_ref(u64 root_objectid, u64 owner, u64 offset)
{
	u32 high_crc = ~(u32)0;
	u32 low_crc = ~(u32)0;
	__le64 lenum;

	lenum = cpu_to_le64(root_objectid);
1042
	high_crc = crc32c(high_crc, &lenum, sizeof(lenum));
1043
	lenum = cpu_to_le64(owner);
1044
	low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
1045
	lenum = cpu_to_le64(offset);
1046
	low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078

	return ((u64)high_crc << 31) ^ (u64)low_crc;
}

static u64 hash_extent_data_ref_item(struct extent_buffer *leaf,
				     struct btrfs_extent_data_ref *ref)
{
	return hash_extent_data_ref(btrfs_extent_data_ref_root(leaf, ref),
				    btrfs_extent_data_ref_objectid(leaf, ref),
				    btrfs_extent_data_ref_offset(leaf, ref));
}

static int match_extent_data_ref(struct extent_buffer *leaf,
				 struct btrfs_extent_data_ref *ref,
				 u64 root_objectid, u64 owner, u64 offset)
{
	if (btrfs_extent_data_ref_root(leaf, ref) != root_objectid ||
	    btrfs_extent_data_ref_objectid(leaf, ref) != owner ||
	    btrfs_extent_data_ref_offset(leaf, ref) != offset)
		return 0;
	return 1;
}

static noinline int lookup_extent_data_ref(struct btrfs_trans_handle *trans,
					   struct btrfs_root *root,
					   struct btrfs_path *path,
					   u64 bytenr, u64 parent,
					   u64 root_objectid,
					   u64 owner, u64 offset)
{
	struct btrfs_key key;
	struct btrfs_extent_data_ref *ref;
Z
Zheng Yan 已提交
1079
	struct extent_buffer *leaf;
1080
	u32 nritems;
1081
	int ret;
1082 1083
	int recow;
	int err = -ENOENT;
1084

Z
Zheng Yan 已提交
1085
	key.objectid = bytenr;
1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100
	if (parent) {
		key.type = BTRFS_SHARED_DATA_REF_KEY;
		key.offset = parent;
	} else {
		key.type = BTRFS_EXTENT_DATA_REF_KEY;
		key.offset = hash_extent_data_ref(root_objectid,
						  owner, offset);
	}
again:
	recow = 0;
	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
	if (ret < 0) {
		err = ret;
		goto fail;
	}
Z
Zheng Yan 已提交
1101

1102 1103 1104 1105 1106
	if (parent) {
		if (!ret)
			return 0;
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
		key.type = BTRFS_EXTENT_REF_V0_KEY;
1107
		btrfs_release_path(path);
1108 1109 1110 1111 1112 1113 1114 1115 1116
		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
		if (ret < 0) {
			err = ret;
			goto fail;
		}
		if (!ret)
			return 0;
#endif
		goto fail;
Z
Zheng Yan 已提交
1117 1118 1119
	}

	leaf = path->nodes[0];
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
	nritems = btrfs_header_nritems(leaf);
	while (1) {
		if (path->slots[0] >= nritems) {
			ret = btrfs_next_leaf(root, path);
			if (ret < 0)
				err = ret;
			if (ret)
				goto fail;

			leaf = path->nodes[0];
			nritems = btrfs_header_nritems(leaf);
			recow = 1;
		}

		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
		if (key.objectid != bytenr ||
		    key.type != BTRFS_EXTENT_DATA_REF_KEY)
			goto fail;

		ref = btrfs_item_ptr(leaf, path->slots[0],
				     struct btrfs_extent_data_ref);

		if (match_extent_data_ref(leaf, ref, root_objectid,
					  owner, offset)) {
			if (recow) {
1145
				btrfs_release_path(path);
1146 1147 1148 1149 1150 1151
				goto again;
			}
			err = 0;
			break;
		}
		path->slots[0]++;
Z
Zheng Yan 已提交
1152
	}
1153 1154
fail:
	return err;
Z
Zheng Yan 已提交
1155 1156
}

1157 1158 1159 1160 1161 1162
static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans,
					   struct btrfs_root *root,
					   struct btrfs_path *path,
					   u64 bytenr, u64 parent,
					   u64 root_objectid, u64 owner,
					   u64 offset, int refs_to_add)
Z
Zheng Yan 已提交
1163 1164 1165
{
	struct btrfs_key key;
	struct extent_buffer *leaf;
1166
	u32 size;
Z
Zheng Yan 已提交
1167 1168
	u32 num_refs;
	int ret;
1169 1170

	key.objectid = bytenr;
1171 1172 1173 1174 1175 1176 1177 1178 1179 1180
	if (parent) {
		key.type = BTRFS_SHARED_DATA_REF_KEY;
		key.offset = parent;
		size = sizeof(struct btrfs_shared_data_ref);
	} else {
		key.type = BTRFS_EXTENT_DATA_REF_KEY;
		key.offset = hash_extent_data_ref(root_objectid,
						  owner, offset);
		size = sizeof(struct btrfs_extent_data_ref);
	}
1181

1182 1183 1184 1185 1186 1187 1188
	ret = btrfs_insert_empty_item(trans, root, path, &key, size);
	if (ret && ret != -EEXIST)
		goto fail;

	leaf = path->nodes[0];
	if (parent) {
		struct btrfs_shared_data_ref *ref;
Z
Zheng Yan 已提交
1189
		ref = btrfs_item_ptr(leaf, path->slots[0],
1190 1191 1192 1193 1194 1195 1196
				     struct btrfs_shared_data_ref);
		if (ret == 0) {
			btrfs_set_shared_data_ref_count(leaf, ref, refs_to_add);
		} else {
			num_refs = btrfs_shared_data_ref_count(leaf, ref);
			num_refs += refs_to_add;
			btrfs_set_shared_data_ref_count(leaf, ref, num_refs);
Z
Zheng Yan 已提交
1197
		}
1198 1199 1200 1201 1202 1203 1204 1205
	} else {
		struct btrfs_extent_data_ref *ref;
		while (ret == -EEXIST) {
			ref = btrfs_item_ptr(leaf, path->slots[0],
					     struct btrfs_extent_data_ref);
			if (match_extent_data_ref(leaf, ref, root_objectid,
						  owner, offset))
				break;
1206
			btrfs_release_path(path);
1207 1208 1209 1210 1211
			key.offset++;
			ret = btrfs_insert_empty_item(trans, root, path, &key,
						      size);
			if (ret && ret != -EEXIST)
				goto fail;
Z
Zheng Yan 已提交
1212

1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226
			leaf = path->nodes[0];
		}
		ref = btrfs_item_ptr(leaf, path->slots[0],
				     struct btrfs_extent_data_ref);
		if (ret == 0) {
			btrfs_set_extent_data_ref_root(leaf, ref,
						       root_objectid);
			btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
			btrfs_set_extent_data_ref_offset(leaf, ref, offset);
			btrfs_set_extent_data_ref_count(leaf, ref, refs_to_add);
		} else {
			num_refs = btrfs_extent_data_ref_count(leaf, ref);
			num_refs += refs_to_add;
			btrfs_set_extent_data_ref_count(leaf, ref, num_refs);
Z
Zheng Yan 已提交
1227 1228
		}
	}
1229 1230 1231
	btrfs_mark_buffer_dirty(leaf);
	ret = 0;
fail:
1232
	btrfs_release_path(path);
1233
	return ret;
1234 1235
}

1236 1237 1238 1239
static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans,
					   struct btrfs_root *root,
					   struct btrfs_path *path,
					   int refs_to_drop)
Z
Zheng Yan 已提交
1240
{
1241 1242 1243
	struct btrfs_key key;
	struct btrfs_extent_data_ref *ref1 = NULL;
	struct btrfs_shared_data_ref *ref2 = NULL;
Z
Zheng Yan 已提交
1244
	struct extent_buffer *leaf;
1245
	u32 num_refs = 0;
Z
Zheng Yan 已提交
1246 1247 1248
	int ret = 0;

	leaf = path->nodes[0];
1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269
	btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);

	if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
		ref1 = btrfs_item_ptr(leaf, path->slots[0],
				      struct btrfs_extent_data_ref);
		num_refs = btrfs_extent_data_ref_count(leaf, ref1);
	} else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
		ref2 = btrfs_item_ptr(leaf, path->slots[0],
				      struct btrfs_shared_data_ref);
		num_refs = btrfs_shared_data_ref_count(leaf, ref2);
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
	} else if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
		struct btrfs_extent_ref_v0 *ref0;
		ref0 = btrfs_item_ptr(leaf, path->slots[0],
				      struct btrfs_extent_ref_v0);
		num_refs = btrfs_ref_count_v0(leaf, ref0);
#endif
	} else {
		BUG();
	}

1270 1271
	BUG_ON(num_refs < refs_to_drop);
	num_refs -= refs_to_drop;
1272

Z
Zheng Yan 已提交
1273 1274 1275
	if (num_refs == 0) {
		ret = btrfs_del_item(trans, root, path);
	} else {
1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287
		if (key.type == BTRFS_EXTENT_DATA_REF_KEY)
			btrfs_set_extent_data_ref_count(leaf, ref1, num_refs);
		else if (key.type == BTRFS_SHARED_DATA_REF_KEY)
			btrfs_set_shared_data_ref_count(leaf, ref2, num_refs);
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
		else {
			struct btrfs_extent_ref_v0 *ref0;
			ref0 = btrfs_item_ptr(leaf, path->slots[0],
					struct btrfs_extent_ref_v0);
			btrfs_set_ref_count_v0(leaf, ref0, num_refs);
		}
#endif
Z
Zheng Yan 已提交
1288 1289 1290 1291 1292
		btrfs_mark_buffer_dirty(leaf);
	}
	return ret;
}

1293 1294 1295
static noinline u32 extent_data_ref_count(struct btrfs_root *root,
					  struct btrfs_path *path,
					  struct btrfs_extent_inline_ref *iref)
1296
{
1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327
	struct btrfs_key key;
	struct extent_buffer *leaf;
	struct btrfs_extent_data_ref *ref1;
	struct btrfs_shared_data_ref *ref2;
	u32 num_refs = 0;

	leaf = path->nodes[0];
	btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
	if (iref) {
		if (btrfs_extent_inline_ref_type(leaf, iref) ==
		    BTRFS_EXTENT_DATA_REF_KEY) {
			ref1 = (struct btrfs_extent_data_ref *)(&iref->offset);
			num_refs = btrfs_extent_data_ref_count(leaf, ref1);
		} else {
			ref2 = (struct btrfs_shared_data_ref *)(iref + 1);
			num_refs = btrfs_shared_data_ref_count(leaf, ref2);
		}
	} else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
		ref1 = btrfs_item_ptr(leaf, path->slots[0],
				      struct btrfs_extent_data_ref);
		num_refs = btrfs_extent_data_ref_count(leaf, ref1);
	} else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
		ref2 = btrfs_item_ptr(leaf, path->slots[0],
				      struct btrfs_shared_data_ref);
		num_refs = btrfs_shared_data_ref_count(leaf, ref2);
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
	} else if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
		struct btrfs_extent_ref_v0 *ref0;
		ref0 = btrfs_item_ptr(leaf, path->slots[0],
				      struct btrfs_extent_ref_v0);
		num_refs = btrfs_ref_count_v0(leaf, ref0);
C
Chris Mason 已提交
1328
#endif
1329 1330 1331 1332 1333
	} else {
		WARN_ON(1);
	}
	return num_refs;
}
1334

1335 1336 1337 1338 1339
static noinline int lookup_tree_block_ref(struct btrfs_trans_handle *trans,
					  struct btrfs_root *root,
					  struct btrfs_path *path,
					  u64 bytenr, u64 parent,
					  u64 root_objectid)
1340
{
1341
	struct btrfs_key key;
1342 1343
	int ret;

1344 1345 1346 1347 1348 1349 1350
	key.objectid = bytenr;
	if (parent) {
		key.type = BTRFS_SHARED_BLOCK_REF_KEY;
		key.offset = parent;
	} else {
		key.type = BTRFS_TREE_BLOCK_REF_KEY;
		key.offset = root_objectid;
1351 1352
	}

1353 1354 1355 1356 1357
	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
	if (ret > 0)
		ret = -ENOENT;
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
	if (ret == -ENOENT && parent) {
1358
		btrfs_release_path(path);
1359 1360 1361 1362 1363
		key.type = BTRFS_EXTENT_REF_V0_KEY;
		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
		if (ret > 0)
			ret = -ENOENT;
	}
1364
#endif
1365
	return ret;
1366 1367
}

1368 1369 1370 1371 1372
static noinline int insert_tree_block_ref(struct btrfs_trans_handle *trans,
					  struct btrfs_root *root,
					  struct btrfs_path *path,
					  u64 bytenr, u64 parent,
					  u64 root_objectid)
Z
Zheng Yan 已提交
1373
{
1374
	struct btrfs_key key;
Z
Zheng Yan 已提交
1375 1376
	int ret;

1377 1378 1379 1380 1381 1382 1383 1384 1385 1386
	key.objectid = bytenr;
	if (parent) {
		key.type = BTRFS_SHARED_BLOCK_REF_KEY;
		key.offset = parent;
	} else {
		key.type = BTRFS_TREE_BLOCK_REF_KEY;
		key.offset = root_objectid;
	}

	ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
1387
	btrfs_release_path(path);
Z
Zheng Yan 已提交
1388 1389 1390
	return ret;
}

1391
static inline int extent_ref_type(u64 parent, u64 owner)
Z
Zheng Yan 已提交
1392
{
1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405
	int type;
	if (owner < BTRFS_FIRST_FREE_OBJECTID) {
		if (parent > 0)
			type = BTRFS_SHARED_BLOCK_REF_KEY;
		else
			type = BTRFS_TREE_BLOCK_REF_KEY;
	} else {
		if (parent > 0)
			type = BTRFS_SHARED_DATA_REF_KEY;
		else
			type = BTRFS_EXTENT_DATA_REF_KEY;
	}
	return type;
Z
Zheng Yan 已提交
1406
}
1407

1408 1409
static int find_next_key(struct btrfs_path *path, int level,
			 struct btrfs_key *key)
1410

C
Chris Mason 已提交
1411
{
1412
	for (; level < BTRFS_MAX_LEVEL; level++) {
1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427
		if (!path->nodes[level])
			break;
		if (path->slots[level] + 1 >=
		    btrfs_header_nritems(path->nodes[level]))
			continue;
		if (level == 0)
			btrfs_item_key_to_cpu(path->nodes[level], key,
					      path->slots[level] + 1);
		else
			btrfs_node_key_to_cpu(path->nodes[level], key,
					      path->slots[level] + 1);
		return 0;
	}
	return 1;
}
C
Chris Mason 已提交
1428

1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463
/*
 * look for inline back ref. if back ref is found, *ref_ret is set
 * to the address of inline back ref, and 0 is returned.
 *
 * if back ref isn't found, *ref_ret is set to the address where it
 * should be inserted, and -ENOENT is returned.
 *
 * if insert is true and there are too many inline back refs, the path
 * points to the extent item, and -EAGAIN is returned.
 *
 * NOTE: inline back refs are ordered in the same way that back ref
 *	 items in the tree are ordered.
 */
static noinline_for_stack
int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
				 struct btrfs_root *root,
				 struct btrfs_path *path,
				 struct btrfs_extent_inline_ref **ref_ret,
				 u64 bytenr, u64 num_bytes,
				 u64 parent, u64 root_objectid,
				 u64 owner, u64 offset, int insert)
{
	struct btrfs_key key;
	struct extent_buffer *leaf;
	struct btrfs_extent_item *ei;
	struct btrfs_extent_inline_ref *iref;
	u64 flags;
	u64 item_size;
	unsigned long ptr;
	unsigned long end;
	int extra_size;
	int type;
	int want;
	int ret;
	int err = 0;
1464

1465
	key.objectid = bytenr;
Z
Zheng Yan 已提交
1466
	key.type = BTRFS_EXTENT_ITEM_KEY;
1467
	key.offset = num_bytes;
Z
Zheng Yan 已提交
1468

1469 1470 1471
	want = extent_ref_type(parent, owner);
	if (insert) {
		extra_size = btrfs_extent_inline_ref_size(want);
1472
		path->keep_locks = 1;
1473 1474 1475
	} else
		extra_size = -1;
	ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1);
1476
	if (ret < 0) {
1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573
		err = ret;
		goto out;
	}
	BUG_ON(ret);

	leaf = path->nodes[0];
	item_size = btrfs_item_size_nr(leaf, path->slots[0]);
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
	if (item_size < sizeof(*ei)) {
		if (!insert) {
			err = -ENOENT;
			goto out;
		}
		ret = convert_extent_item_v0(trans, root, path, owner,
					     extra_size);
		if (ret < 0) {
			err = ret;
			goto out;
		}
		leaf = path->nodes[0];
		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
	}
#endif
	BUG_ON(item_size < sizeof(*ei));

	ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
	flags = btrfs_extent_flags(leaf, ei);

	ptr = (unsigned long)(ei + 1);
	end = (unsigned long)ei + item_size;

	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
		ptr += sizeof(struct btrfs_tree_block_info);
		BUG_ON(ptr > end);
	} else {
		BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
	}

	err = -ENOENT;
	while (1) {
		if (ptr >= end) {
			WARN_ON(ptr > end);
			break;
		}
		iref = (struct btrfs_extent_inline_ref *)ptr;
		type = btrfs_extent_inline_ref_type(leaf, iref);
		if (want < type)
			break;
		if (want > type) {
			ptr += btrfs_extent_inline_ref_size(type);
			continue;
		}

		if (type == BTRFS_EXTENT_DATA_REF_KEY) {
			struct btrfs_extent_data_ref *dref;
			dref = (struct btrfs_extent_data_ref *)(&iref->offset);
			if (match_extent_data_ref(leaf, dref, root_objectid,
						  owner, offset)) {
				err = 0;
				break;
			}
			if (hash_extent_data_ref_item(leaf, dref) <
			    hash_extent_data_ref(root_objectid, owner, offset))
				break;
		} else {
			u64 ref_offset;
			ref_offset = btrfs_extent_inline_ref_offset(leaf, iref);
			if (parent > 0) {
				if (parent == ref_offset) {
					err = 0;
					break;
				}
				if (ref_offset < parent)
					break;
			} else {
				if (root_objectid == ref_offset) {
					err = 0;
					break;
				}
				if (ref_offset < root_objectid)
					break;
			}
		}
		ptr += btrfs_extent_inline_ref_size(type);
	}
	if (err == -ENOENT && insert) {
		if (item_size + extra_size >=
		    BTRFS_MAX_EXTENT_ITEM_SIZE(root)) {
			err = -EAGAIN;
			goto out;
		}
		/*
		 * To add new inline back ref, we have to make sure
		 * there is no corresponding back ref item.
		 * For simplicity, we just do not add new inline back
		 * ref if there is any kind of item for this block
		 */
1574 1575
		if (find_next_key(path, 0, &key) == 0 &&
		    key.objectid == bytenr &&
1576
		    key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) {
1577 1578 1579 1580 1581 1582
			err = -EAGAIN;
			goto out;
		}
	}
	*ref_ret = (struct btrfs_extent_inline_ref *)ptr;
out:
1583
	if (insert) {
1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669
		path->keep_locks = 0;
		btrfs_unlock_up_safe(path, 1);
	}
	return err;
}

/*
 * helper to add new inline back ref
 */
static noinline_for_stack
int setup_inline_extent_backref(struct btrfs_trans_handle *trans,
				struct btrfs_root *root,
				struct btrfs_path *path,
				struct btrfs_extent_inline_ref *iref,
				u64 parent, u64 root_objectid,
				u64 owner, u64 offset, int refs_to_add,
				struct btrfs_delayed_extent_op *extent_op)
{
	struct extent_buffer *leaf;
	struct btrfs_extent_item *ei;
	unsigned long ptr;
	unsigned long end;
	unsigned long item_offset;
	u64 refs;
	int size;
	int type;
	int ret;

	leaf = path->nodes[0];
	ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
	item_offset = (unsigned long)iref - (unsigned long)ei;

	type = extent_ref_type(parent, owner);
	size = btrfs_extent_inline_ref_size(type);

	ret = btrfs_extend_item(trans, root, path, size);

	ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
	refs = btrfs_extent_refs(leaf, ei);
	refs += refs_to_add;
	btrfs_set_extent_refs(leaf, ei, refs);
	if (extent_op)
		__run_delayed_extent_op(extent_op, leaf, ei);

	ptr = (unsigned long)ei + item_offset;
	end = (unsigned long)ei + btrfs_item_size_nr(leaf, path->slots[0]);
	if (ptr < end - size)
		memmove_extent_buffer(leaf, ptr + size, ptr,
				      end - size - ptr);

	iref = (struct btrfs_extent_inline_ref *)ptr;
	btrfs_set_extent_inline_ref_type(leaf, iref, type);
	if (type == BTRFS_EXTENT_DATA_REF_KEY) {
		struct btrfs_extent_data_ref *dref;
		dref = (struct btrfs_extent_data_ref *)(&iref->offset);
		btrfs_set_extent_data_ref_root(leaf, dref, root_objectid);
		btrfs_set_extent_data_ref_objectid(leaf, dref, owner);
		btrfs_set_extent_data_ref_offset(leaf, dref, offset);
		btrfs_set_extent_data_ref_count(leaf, dref, refs_to_add);
	} else if (type == BTRFS_SHARED_DATA_REF_KEY) {
		struct btrfs_shared_data_ref *sref;
		sref = (struct btrfs_shared_data_ref *)(iref + 1);
		btrfs_set_shared_data_ref_count(leaf, sref, refs_to_add);
		btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
	} else if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
		btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
	} else {
		btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
	}
	btrfs_mark_buffer_dirty(leaf);
	return 0;
}

static int lookup_extent_backref(struct btrfs_trans_handle *trans,
				 struct btrfs_root *root,
				 struct btrfs_path *path,
				 struct btrfs_extent_inline_ref **ref_ret,
				 u64 bytenr, u64 num_bytes, u64 parent,
				 u64 root_objectid, u64 owner, u64 offset)
{
	int ret;

	ret = lookup_inline_extent_backref(trans, root, path, ref_ret,
					   bytenr, num_bytes, parent,
					   root_objectid, owner, offset, 0);
	if (ret != -ENOENT)
1670
		return ret;
1671

1672
	btrfs_release_path(path);
1673 1674 1675 1676 1677 1678 1679 1680
	*ref_ret = NULL;

	if (owner < BTRFS_FIRST_FREE_OBJECTID) {
		ret = lookup_tree_block_ref(trans, root, path, bytenr, parent,
					    root_objectid);
	} else {
		ret = lookup_extent_data_ref(trans, root, path, bytenr, parent,
					     root_objectid, owner, offset);
1681
	}
1682 1683
	return ret;
}
Z
Zheng Yan 已提交
1684

1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727
/*
 * helper to update/remove inline back ref
 */
static noinline_for_stack
int update_inline_extent_backref(struct btrfs_trans_handle *trans,
				 struct btrfs_root *root,
				 struct btrfs_path *path,
				 struct btrfs_extent_inline_ref *iref,
				 int refs_to_mod,
				 struct btrfs_delayed_extent_op *extent_op)
{
	struct extent_buffer *leaf;
	struct btrfs_extent_item *ei;
	struct btrfs_extent_data_ref *dref = NULL;
	struct btrfs_shared_data_ref *sref = NULL;
	unsigned long ptr;
	unsigned long end;
	u32 item_size;
	int size;
	int type;
	int ret;
	u64 refs;

	leaf = path->nodes[0];
	ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
	refs = btrfs_extent_refs(leaf, ei);
	WARN_ON(refs_to_mod < 0 && refs + refs_to_mod <= 0);
	refs += refs_to_mod;
	btrfs_set_extent_refs(leaf, ei, refs);
	if (extent_op)
		__run_delayed_extent_op(extent_op, leaf, ei);

	type = btrfs_extent_inline_ref_type(leaf, iref);

	if (type == BTRFS_EXTENT_DATA_REF_KEY) {
		dref = (struct btrfs_extent_data_ref *)(&iref->offset);
		refs = btrfs_extent_data_ref_count(leaf, dref);
	} else if (type == BTRFS_SHARED_DATA_REF_KEY) {
		sref = (struct btrfs_shared_data_ref *)(iref + 1);
		refs = btrfs_shared_data_ref_count(leaf, sref);
	} else {
		refs = 1;
		BUG_ON(refs_to_mod != -1);
1728
	}
Z
Zheng Yan 已提交
1729

1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776
	BUG_ON(refs_to_mod < 0 && refs < -refs_to_mod);
	refs += refs_to_mod;

	if (refs > 0) {
		if (type == BTRFS_EXTENT_DATA_REF_KEY)
			btrfs_set_extent_data_ref_count(leaf, dref, refs);
		else
			btrfs_set_shared_data_ref_count(leaf, sref, refs);
	} else {
		size =  btrfs_extent_inline_ref_size(type);
		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
		ptr = (unsigned long)iref;
		end = (unsigned long)ei + item_size;
		if (ptr + size < end)
			memmove_extent_buffer(leaf, ptr, ptr + size,
					      end - ptr - size);
		item_size -= size;
		ret = btrfs_truncate_item(trans, root, path, item_size, 1);
	}
	btrfs_mark_buffer_dirty(leaf);
	return 0;
}

static noinline_for_stack
int insert_inline_extent_backref(struct btrfs_trans_handle *trans,
				 struct btrfs_root *root,
				 struct btrfs_path *path,
				 u64 bytenr, u64 num_bytes, u64 parent,
				 u64 root_objectid, u64 owner,
				 u64 offset, int refs_to_add,
				 struct btrfs_delayed_extent_op *extent_op)
{
	struct btrfs_extent_inline_ref *iref;
	int ret;

	ret = lookup_inline_extent_backref(trans, root, path, &iref,
					   bytenr, num_bytes, parent,
					   root_objectid, owner, offset, 1);
	if (ret == 0) {
		BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID);
		ret = update_inline_extent_backref(trans, root, path, iref,
						   refs_to_add, extent_op);
	} else if (ret == -ENOENT) {
		ret = setup_inline_extent_backref(trans, root, path, iref,
						  parent, root_objectid,
						  owner, offset, refs_to_add,
						  extent_op);
1777
	}
1778 1779
	return ret;
}
Z
Zheng Yan 已提交
1780

1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798
static int insert_extent_backref(struct btrfs_trans_handle *trans,
				 struct btrfs_root *root,
				 struct btrfs_path *path,
				 u64 bytenr, u64 parent, u64 root_objectid,
				 u64 owner, u64 offset, int refs_to_add)
{
	int ret;
	if (owner < BTRFS_FIRST_FREE_OBJECTID) {
		BUG_ON(refs_to_add != 1);
		ret = insert_tree_block_ref(trans, root, path, bytenr,
					    parent, root_objectid);
	} else {
		ret = insert_extent_data_ref(trans, root, path, bytenr,
					     parent, root_objectid,
					     owner, offset, refs_to_add);
	}
	return ret;
}
1799

1800 1801 1802 1803 1804 1805 1806
static int remove_extent_backref(struct btrfs_trans_handle *trans,
				 struct btrfs_root *root,
				 struct btrfs_path *path,
				 struct btrfs_extent_inline_ref *iref,
				 int refs_to_drop, int is_data)
{
	int ret;
1807

1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819
	BUG_ON(!is_data && refs_to_drop != 1);
	if (iref) {
		ret = update_inline_extent_backref(trans, root, path, iref,
						   -refs_to_drop, NULL);
	} else if (is_data) {
		ret = remove_extent_data_ref(trans, root, path, refs_to_drop);
	} else {
		ret = btrfs_del_item(trans, root, path);
	}
	return ret;
}

1820
static int btrfs_issue_discard(struct block_device *bdev,
1821 1822
				u64 start, u64 len)
{
1823
	return blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_NOFS, 0);
1824 1825 1826
}

static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
1827
				u64 num_bytes, u64 *actual_bytes)
1828 1829
{
	int ret;
1830
	u64 discarded_bytes = 0;
1831
	struct btrfs_bio *bbio = NULL;
1832

C
Christoph Hellwig 已提交
1833

1834
	/* Tell the block device(s) that the sectors can be discarded */
1835
	ret = btrfs_map_block(&root->fs_info->mapping_tree, REQ_DISCARD,
1836
			      bytenr, &num_bytes, &bbio, 0);
1837
	if (!ret) {
1838
		struct btrfs_bio_stripe *stripe = bbio->stripes;
1839 1840 1841
		int i;


1842
		for (i = 0; i < bbio->num_stripes; i++, stripe++) {
1843 1844 1845
			if (!stripe->dev->can_discard)
				continue;

1846 1847 1848 1849 1850 1851 1852
			ret = btrfs_issue_discard(stripe->dev->bdev,
						  stripe->physical,
						  stripe->length);
			if (!ret)
				discarded_bytes += stripe->length;
			else if (ret != -EOPNOTSUPP)
				break;
1853 1854 1855 1856 1857 1858 1859

			/*
			 * Just in case we get back EOPNOTSUPP for some reason,
			 * just ignore the return value so we don't screw up
			 * people calling discard_extent.
			 */
			ret = 0;
1860
		}
1861
		kfree(bbio);
1862
	}
1863 1864 1865 1866

	if (actual_bytes)
		*actual_bytes = discarded_bytes;

1867 1868 1869 1870 1871 1872 1873

	return ret;
}

int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
			 struct btrfs_root *root,
			 u64 bytenr, u64 num_bytes, u64 parent,
A
Arne Jansen 已提交
1874
			 u64 root_objectid, u64 owner, u64 offset, int for_cow)
1875 1876
{
	int ret;
A
Arne Jansen 已提交
1877 1878
	struct btrfs_fs_info *fs_info = root->fs_info;

1879 1880 1881 1882
	BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID &&
	       root_objectid == BTRFS_TREE_LOG_OBJECTID);

	if (owner < BTRFS_FIRST_FREE_OBJECTID) {
A
Arne Jansen 已提交
1883 1884
		ret = btrfs_add_delayed_tree_ref(fs_info, trans, bytenr,
					num_bytes,
1885
					parent, root_objectid, (int)owner,
A
Arne Jansen 已提交
1886
					BTRFS_ADD_DELAYED_REF, NULL, for_cow);
1887
	} else {
A
Arne Jansen 已提交
1888 1889
		ret = btrfs_add_delayed_data_ref(fs_info, trans, bytenr,
					num_bytes,
1890
					parent, root_objectid, owner, offset,
A
Arne Jansen 已提交
1891
					BTRFS_ADD_DELAYED_REF, NULL, for_cow);
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 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934
	}
	return ret;
}

static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
				  struct btrfs_root *root,
				  u64 bytenr, u64 num_bytes,
				  u64 parent, u64 root_objectid,
				  u64 owner, u64 offset, int refs_to_add,
				  struct btrfs_delayed_extent_op *extent_op)
{
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct btrfs_extent_item *item;
	u64 refs;
	int ret;
	int err = 0;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

	path->reada = 1;
	path->leave_spinning = 1;
	/* this will setup the path even if it fails to insert the back ref */
	ret = insert_inline_extent_backref(trans, root->fs_info->extent_root,
					   path, bytenr, num_bytes, parent,
					   root_objectid, owner, offset,
					   refs_to_add, extent_op);
	if (ret == 0)
		goto out;

	if (ret != -EAGAIN) {
		err = ret;
		goto out;
	}

	leaf = path->nodes[0];
	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
	refs = btrfs_extent_refs(leaf, item);
	btrfs_set_extent_refs(leaf, item, refs + refs_to_add);
	if (extent_op)
		__run_delayed_extent_op(extent_op, leaf, item);
1935

1936
	btrfs_mark_buffer_dirty(leaf);
1937
	btrfs_release_path(path);
1938 1939

	path->reada = 1;
1940 1941
	path->leave_spinning = 1;

1942 1943
	/* now insert the actual backref */
	ret = insert_extent_backref(trans, root->fs_info->extent_root,
1944 1945
				    path, bytenr, parent, root_objectid,
				    owner, offset, refs_to_add);
1946
	BUG_ON(ret);
1947
out:
1948
	btrfs_free_path(path);
1949
	return err;
1950 1951
}

1952 1953 1954 1955 1956
static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
				struct btrfs_root *root,
				struct btrfs_delayed_ref_node *node,
				struct btrfs_delayed_extent_op *extent_op,
				int insert_reserved)
1957
{
1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029
	int ret = 0;
	struct btrfs_delayed_data_ref *ref;
	struct btrfs_key ins;
	u64 parent = 0;
	u64 ref_root = 0;
	u64 flags = 0;

	ins.objectid = node->bytenr;
	ins.offset = node->num_bytes;
	ins.type = BTRFS_EXTENT_ITEM_KEY;

	ref = btrfs_delayed_node_to_data_ref(node);
	if (node->type == BTRFS_SHARED_DATA_REF_KEY)
		parent = ref->parent;
	else
		ref_root = ref->root;

	if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
		if (extent_op) {
			BUG_ON(extent_op->update_key);
			flags |= extent_op->flags_to_set;
		}
		ret = alloc_reserved_file_extent(trans, root,
						 parent, ref_root, flags,
						 ref->objectid, ref->offset,
						 &ins, node->ref_mod);
	} else if (node->action == BTRFS_ADD_DELAYED_REF) {
		ret = __btrfs_inc_extent_ref(trans, root, node->bytenr,
					     node->num_bytes, parent,
					     ref_root, ref->objectid,
					     ref->offset, node->ref_mod,
					     extent_op);
	} else if (node->action == BTRFS_DROP_DELAYED_REF) {
		ret = __btrfs_free_extent(trans, root, node->bytenr,
					  node->num_bytes, parent,
					  ref_root, ref->objectid,
					  ref->offset, node->ref_mod,
					  extent_op);
	} else {
		BUG();
	}
	return ret;
}

static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
				    struct extent_buffer *leaf,
				    struct btrfs_extent_item *ei)
{
	u64 flags = btrfs_extent_flags(leaf, ei);
	if (extent_op->update_flags) {
		flags |= extent_op->flags_to_set;
		btrfs_set_extent_flags(leaf, ei, flags);
	}

	if (extent_op->update_key) {
		struct btrfs_tree_block_info *bi;
		BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK));
		bi = (struct btrfs_tree_block_info *)(ei + 1);
		btrfs_set_tree_block_key(leaf, bi, &extent_op->key);
	}
}

static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
				 struct btrfs_root *root,
				 struct btrfs_delayed_ref_node *node,
				 struct btrfs_delayed_extent_op *extent_op)
{
	struct btrfs_key key;
	struct btrfs_path *path;
	struct btrfs_extent_item *ei;
	struct extent_buffer *leaf;
	u32 item_size;
2030
	int ret;
2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070
	int err = 0;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

	key.objectid = node->bytenr;
	key.type = BTRFS_EXTENT_ITEM_KEY;
	key.offset = node->num_bytes;

	path->reada = 1;
	path->leave_spinning = 1;
	ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key,
				path, 0, 1);
	if (ret < 0) {
		err = ret;
		goto out;
	}
	if (ret > 0) {
		err = -EIO;
		goto out;
	}

	leaf = path->nodes[0];
	item_size = btrfs_item_size_nr(leaf, path->slots[0]);
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
	if (item_size < sizeof(*ei)) {
		ret = convert_extent_item_v0(trans, root->fs_info->extent_root,
					     path, (u64)-1, 0);
		if (ret < 0) {
			err = ret;
			goto out;
		}
		leaf = path->nodes[0];
		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
	}
#endif
	BUG_ON(item_size < sizeof(*ei));
	ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
	__run_delayed_extent_op(extent_op, leaf, ei);
2071

2072 2073 2074 2075
	btrfs_mark_buffer_dirty(leaf);
out:
	btrfs_free_path(path);
	return err;
2076 2077
}

2078 2079 2080 2081 2082
static int run_delayed_tree_ref(struct btrfs_trans_handle *trans,
				struct btrfs_root *root,
				struct btrfs_delayed_ref_node *node,
				struct btrfs_delayed_extent_op *extent_op,
				int insert_reserved)
2083 2084
{
	int ret = 0;
2085 2086 2087 2088
	struct btrfs_delayed_tree_ref *ref;
	struct btrfs_key ins;
	u64 parent = 0;
	u64 ref_root = 0;
2089

2090 2091 2092
	ins.objectid = node->bytenr;
	ins.offset = node->num_bytes;
	ins.type = BTRFS_EXTENT_ITEM_KEY;
2093

2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119
	ref = btrfs_delayed_node_to_tree_ref(node);
	if (node->type == BTRFS_SHARED_BLOCK_REF_KEY)
		parent = ref->parent;
	else
		ref_root = ref->root;

	BUG_ON(node->ref_mod != 1);
	if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
		BUG_ON(!extent_op || !extent_op->update_flags ||
		       !extent_op->update_key);
		ret = alloc_reserved_tree_block(trans, root,
						parent, ref_root,
						extent_op->flags_to_set,
						&extent_op->key,
						ref->level, &ins);
	} else if (node->action == BTRFS_ADD_DELAYED_REF) {
		ret = __btrfs_inc_extent_ref(trans, root, node->bytenr,
					     node->num_bytes, parent, ref_root,
					     ref->level, 0, 1, extent_op);
	} else if (node->action == BTRFS_DROP_DELAYED_REF) {
		ret = __btrfs_free_extent(trans, root, node->bytenr,
					  node->num_bytes, parent, ref_root,
					  ref->level, 0, 1, extent_op);
	} else {
		BUG();
	}
2120 2121 2122 2123
	return ret;
}

/* helper function to actually process a single delayed ref entry */
2124 2125 2126 2127 2128
static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
			       struct btrfs_root *root,
			       struct btrfs_delayed_ref_node *node,
			       struct btrfs_delayed_extent_op *extent_op,
			       int insert_reserved)
2129 2130
{
	int ret;
2131
	if (btrfs_delayed_ref_is_head(node)) {
2132 2133 2134 2135 2136 2137 2138
		struct btrfs_delayed_ref_head *head;
		/*
		 * we've hit the end of the chain and we were supposed
		 * to insert this extent into the tree.  But, it got
		 * deleted before we ever needed to insert it, so all
		 * we have to do is clean up the accounting
		 */
2139 2140
		BUG_ON(extent_op);
		head = btrfs_delayed_node_to_head(node);
2141
		if (insert_reserved) {
2142 2143
			btrfs_pin_extent(root, node->bytenr,
					 node->num_bytes, 1);
2144 2145 2146 2147 2148 2149
			if (head->is_data) {
				ret = btrfs_del_csums(trans, root,
						      node->bytenr,
						      node->num_bytes);
				BUG_ON(ret);
			}
2150 2151 2152 2153 2154
		}
		mutex_unlock(&head->mutex);
		return 0;
	}

2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165
	if (node->type == BTRFS_TREE_BLOCK_REF_KEY ||
	    node->type == BTRFS_SHARED_BLOCK_REF_KEY)
		ret = run_delayed_tree_ref(trans, root, node, extent_op,
					   insert_reserved);
	else if (node->type == BTRFS_EXTENT_DATA_REF_KEY ||
		 node->type == BTRFS_SHARED_DATA_REF_KEY)
		ret = run_delayed_data_ref(trans, root, node, extent_op,
					   insert_reserved);
	else
		BUG();
	return ret;
2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187
}

static noinline struct btrfs_delayed_ref_node *
select_delayed_ref(struct btrfs_delayed_ref_head *head)
{
	struct rb_node *node;
	struct btrfs_delayed_ref_node *ref;
	int action = BTRFS_ADD_DELAYED_REF;
again:
	/*
	 * select delayed ref of type BTRFS_ADD_DELAYED_REF first.
	 * this prevents ref count from going down to zero when
	 * there still are pending delayed ref.
	 */
	node = rb_prev(&head->node.rb_node);
	while (1) {
		if (!node)
			break;
		ref = rb_entry(node, struct btrfs_delayed_ref_node,
				rb_node);
		if (ref->bytenr != head->node.bytenr)
			break;
2188
		if (ref->action == action)
2189 2190 2191 2192 2193 2194 2195 2196 2197 2198
			return ref;
		node = rb_prev(node);
	}
	if (action == BTRFS_ADD_DELAYED_REF) {
		action = BTRFS_DROP_DELAYED_REF;
		goto again;
	}
	return NULL;
}

2199 2200 2201
static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
				       struct btrfs_root *root,
				       struct list_head *cluster)
2202 2203 2204 2205
{
	struct btrfs_delayed_ref_root *delayed_refs;
	struct btrfs_delayed_ref_node *ref;
	struct btrfs_delayed_ref_head *locked_ref = NULL;
2206
	struct btrfs_delayed_extent_op *extent_op;
2207
	int ret;
2208
	int count = 0;
2209 2210 2211 2212 2213
	int must_insert_reserved = 0;

	delayed_refs = &trans->transaction->delayed_refs;
	while (1) {
		if (!locked_ref) {
2214 2215
			/* pick a new head ref from the cluster list */
			if (list_empty(cluster))
2216 2217
				break;

2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234
			locked_ref = list_entry(cluster->next,
				     struct btrfs_delayed_ref_head, cluster);

			/* grab the lock that says we are going to process
			 * all the refs for this head */
			ret = btrfs_delayed_ref_lock(trans, locked_ref);

			/*
			 * we may have dropped the spin lock to get the head
			 * mutex lock, and that might have given someone else
			 * time to free the head.  If that's true, it has been
			 * removed from our list and we can move on.
			 */
			if (ret == -EAGAIN) {
				locked_ref = NULL;
				count++;
				continue;
2235 2236
			}
		}
2237

2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259
		/*
		 * locked_ref is the head node, so we have to go one
		 * node back for any delayed ref updates
		 */
		ref = select_delayed_ref(locked_ref);

		if (ref && ref->seq &&
		    btrfs_check_delayed_seq(delayed_refs, ref->seq)) {
			/*
			 * there are still refs with lower seq numbers in the
			 * process of being added. Don't run this ref yet.
			 */
			list_del_init(&locked_ref->cluster);
			mutex_unlock(&locked_ref->mutex);
			locked_ref = NULL;
			delayed_refs->num_heads_ready++;
			spin_unlock(&delayed_refs->lock);
			cond_resched();
			spin_lock(&delayed_refs->lock);
			continue;
		}

2260 2261 2262 2263 2264 2265
		/*
		 * record the must insert reserved flag before we
		 * drop the spin lock.
		 */
		must_insert_reserved = locked_ref->must_insert_reserved;
		locked_ref->must_insert_reserved = 0;
2266

2267 2268 2269
		extent_op = locked_ref->extent_op;
		locked_ref->extent_op = NULL;

2270 2271 2272 2273 2274 2275
		if (!ref) {
			/* All delayed refs have been processed, Go ahead
			 * and send the head node to run_one_delayed_ref,
			 * so that any accounting fixes can happen
			 */
			ref = &locked_ref->node;
2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289

			if (extent_op && must_insert_reserved) {
				kfree(extent_op);
				extent_op = NULL;
			}

			if (extent_op) {
				spin_unlock(&delayed_refs->lock);

				ret = run_delayed_extent_op(trans, root,
							    ref, extent_op);
				BUG_ON(ret);
				kfree(extent_op);

2290
				goto next;
2291 2292
			}

2293
			list_del_init(&locked_ref->cluster);
2294 2295
			locked_ref = NULL;
		}
C
Chris Mason 已提交
2296

2297 2298 2299
		ref->in_tree = 0;
		rb_erase(&ref->rb_node, &delayed_refs->root);
		delayed_refs->num_entries--;
2300 2301 2302 2303 2304 2305
		/*
		 * we modified num_entries, but as we're currently running
		 * delayed refs, skip
		 *     wake_up(&delayed_refs->seq_wait);
		 * here.
		 */
2306
		spin_unlock(&delayed_refs->lock);
2307

2308
		ret = run_one_delayed_ref(trans, root, ref, extent_op,
2309 2310
					  must_insert_reserved);
		BUG_ON(ret);
2311

2312 2313
		btrfs_put_delayed_ref(ref);
		kfree(extent_op);
2314
		count++;
2315 2316 2317 2318 2319
next:
		do_chunk_alloc(trans, root->fs_info->extent_root,
			       2 * 1024 * 1024,
			       btrfs_get_alloc_profile(root, 0),
			       CHUNK_ALLOC_NO_FORCE);
2320 2321 2322 2323 2324 2325
		cond_resched();
		spin_lock(&delayed_refs->lock);
	}
	return count;
}

2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342

static void wait_for_more_refs(struct btrfs_delayed_ref_root *delayed_refs,
			unsigned long num_refs)
{
	struct list_head *first_seq = delayed_refs->seq_head.next;

	spin_unlock(&delayed_refs->lock);
	pr_debug("waiting for more refs (num %ld, first %p)\n",
		 num_refs, first_seq);
	wait_event(delayed_refs->seq_wait,
		   num_refs != delayed_refs->num_entries ||
		   delayed_refs->seq_head.next != first_seq);
	pr_debug("done waiting for more refs (num %ld, first %p)\n",
		 delayed_refs->num_entries, delayed_refs->seq_head.next);
	spin_lock(&delayed_refs->lock);
}

2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357
/*
 * this starts processing the delayed reference count updates and
 * extent insertions we have queued up so far.  count can be
 * 0, which means to process everything in the tree at the start
 * of the run (but not newly added entries), or it can be some target
 * number you'd like to process.
 */
int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
			   struct btrfs_root *root, unsigned long count)
{
	struct rb_node *node;
	struct btrfs_delayed_ref_root *delayed_refs;
	struct btrfs_delayed_ref_node *ref;
	struct list_head cluster;
	int ret;
2358
	u64 delayed_start;
2359 2360
	int run_all = count == (unsigned long)-1;
	int run_most = 0;
2361 2362
	unsigned long num_refs = 0;
	int consider_waiting;
2363 2364 2365 2366

	if (root == root->fs_info->extent_root)
		root = root->fs_info->tree_root;

2367 2368 2369 2370
	do_chunk_alloc(trans, root->fs_info->extent_root,
		       2 * 1024 * 1024, btrfs_get_alloc_profile(root, 0),
		       CHUNK_ALLOC_NO_FORCE);

2371 2372 2373
	delayed_refs = &trans->transaction->delayed_refs;
	INIT_LIST_HEAD(&cluster);
again:
2374
	consider_waiting = 0;
2375 2376 2377 2378 2379 2380 2381 2382 2383
	spin_lock(&delayed_refs->lock);
	if (count == 0) {
		count = delayed_refs->num_entries * 2;
		run_most = 1;
	}
	while (1) {
		if (!(run_all || run_most) &&
		    delayed_refs->num_heads_ready < 64)
			break;
2384

2385
		/*
2386 2387 2388 2389
		 * go find something we can process in the rbtree.  We start at
		 * the beginning of the tree, and then build a cluster
		 * of refs to process starting at the first one we are able to
		 * lock
2390
		 */
2391
		delayed_start = delayed_refs->run_delayed_start;
2392 2393 2394
		ret = btrfs_find_ref_cluster(trans, &cluster,
					     delayed_refs->run_delayed_start);
		if (ret)
2395 2396
			break;

2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419
		if (delayed_start >= delayed_refs->run_delayed_start) {
			if (consider_waiting == 0) {
				/*
				 * btrfs_find_ref_cluster looped. let's do one
				 * more cycle. if we don't run any delayed ref
				 * during that cycle (because we can't because
				 * all of them are blocked) and if the number of
				 * refs doesn't change, we avoid busy waiting.
				 */
				consider_waiting = 1;
				num_refs = delayed_refs->num_entries;
			} else {
				wait_for_more_refs(delayed_refs, num_refs);
				/*
				 * after waiting, things have changed. we
				 * dropped the lock and someone else might have
				 * run some refs, built new clusters and so on.
				 * therefore, we restart staleness detection.
				 */
				consider_waiting = 0;
			}
		}

2420 2421 2422 2423 2424 2425 2426
		ret = run_clustered_refs(trans, root, &cluster);
		BUG_ON(ret < 0);

		count -= min_t(unsigned long, ret, count);

		if (count == 0)
			break;
2427 2428 2429 2430 2431

		if (ret || delayed_refs->run_delayed_start == 0) {
			/* refs were run, let's reset staleness detection */
			consider_waiting = 0;
		}
2432
	}
2433

2434 2435
	if (run_all) {
		node = rb_first(&delayed_refs->root);
2436
		if (!node)
2437
			goto out;
2438
		count = (unsigned long)-1;
2439

2440 2441 2442 2443 2444
		while (node) {
			ref = rb_entry(node, struct btrfs_delayed_ref_node,
				       rb_node);
			if (btrfs_delayed_ref_is_head(ref)) {
				struct btrfs_delayed_ref_head *head;
2445

2446 2447 2448 2449
				head = btrfs_delayed_node_to_head(ref);
				atomic_inc(&ref->refs);

				spin_unlock(&delayed_refs->lock);
2450 2451 2452 2453
				/*
				 * Mutex was contended, block until it's
				 * released and try again
				 */
2454 2455 2456 2457
				mutex_lock(&head->mutex);
				mutex_unlock(&head->mutex);

				btrfs_put_delayed_ref(ref);
2458
				cond_resched();
2459 2460 2461 2462 2463 2464 2465
				goto again;
			}
			node = rb_next(node);
		}
		spin_unlock(&delayed_refs->lock);
		schedule_timeout(1);
		goto again;
2466
	}
2467
out:
2468
	spin_unlock(&delayed_refs->lock);
2469 2470 2471
	return 0;
}

2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488
int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
				struct btrfs_root *root,
				u64 bytenr, u64 num_bytes, u64 flags,
				int is_data)
{
	struct btrfs_delayed_extent_op *extent_op;
	int ret;

	extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
	if (!extent_op)
		return -ENOMEM;

	extent_op->flags_to_set = flags;
	extent_op->update_flags = 1;
	extent_op->update_key = 0;
	extent_op->is_data = is_data ? 1 : 0;

A
Arne Jansen 已提交
2489 2490
	ret = btrfs_add_delayed_extent_op(root->fs_info, trans, bytenr,
					  num_bytes, extent_op);
2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518
	if (ret)
		kfree(extent_op);
	return ret;
}

static noinline int check_delayed_ref(struct btrfs_trans_handle *trans,
				      struct btrfs_root *root,
				      struct btrfs_path *path,
				      u64 objectid, u64 offset, u64 bytenr)
{
	struct btrfs_delayed_ref_head *head;
	struct btrfs_delayed_ref_node *ref;
	struct btrfs_delayed_data_ref *data_ref;
	struct btrfs_delayed_ref_root *delayed_refs;
	struct rb_node *node;
	int ret = 0;

	ret = -ENOENT;
	delayed_refs = &trans->transaction->delayed_refs;
	spin_lock(&delayed_refs->lock);
	head = btrfs_find_delayed_ref_head(trans, bytenr);
	if (!head)
		goto out;

	if (!mutex_trylock(&head->mutex)) {
		atomic_inc(&head->node.refs);
		spin_unlock(&delayed_refs->lock);

2519
		btrfs_release_path(path);
2520

2521 2522 2523 2524
		/*
		 * Mutex was contended, block until it's released and let
		 * caller try again
		 */
2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568
		mutex_lock(&head->mutex);
		mutex_unlock(&head->mutex);
		btrfs_put_delayed_ref(&head->node);
		return -EAGAIN;
	}

	node = rb_prev(&head->node.rb_node);
	if (!node)
		goto out_unlock;

	ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);

	if (ref->bytenr != bytenr)
		goto out_unlock;

	ret = 1;
	if (ref->type != BTRFS_EXTENT_DATA_REF_KEY)
		goto out_unlock;

	data_ref = btrfs_delayed_node_to_data_ref(ref);

	node = rb_prev(node);
	if (node) {
		ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
		if (ref->bytenr == bytenr)
			goto out_unlock;
	}

	if (data_ref->root != root->root_key.objectid ||
	    data_ref->objectid != objectid || data_ref->offset != offset)
		goto out_unlock;

	ret = 0;
out_unlock:
	mutex_unlock(&head->mutex);
out:
	spin_unlock(&delayed_refs->lock);
	return ret;
}

static noinline int check_committed_ref(struct btrfs_trans_handle *trans,
					struct btrfs_root *root,
					struct btrfs_path *path,
					u64 objectid, u64 offset, u64 bytenr)
2569 2570
{
	struct btrfs_root *extent_root = root->fs_info->extent_root;
2571
	struct extent_buffer *leaf;
2572 2573 2574
	struct btrfs_extent_data_ref *ref;
	struct btrfs_extent_inline_ref *iref;
	struct btrfs_extent_item *ei;
2575
	struct btrfs_key key;
2576
	u32 item_size;
2577
	int ret;
2578

2579
	key.objectid = bytenr;
Z
Zheng Yan 已提交
2580
	key.offset = (u64)-1;
2581
	key.type = BTRFS_EXTENT_ITEM_KEY;
2582 2583 2584 2585 2586

	ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
	if (ret < 0)
		goto out;
	BUG_ON(ret == 0);
Y
Yan Zheng 已提交
2587 2588 2589

	ret = -ENOENT;
	if (path->slots[0] == 0)
Z
Zheng Yan 已提交
2590
		goto out;
2591

Z
Zheng Yan 已提交
2592
	path->slots[0]--;
2593
	leaf = path->nodes[0];
2594
	btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
2595

2596
	if (key.objectid != bytenr || key.type != BTRFS_EXTENT_ITEM_KEY)
2597
		goto out;
2598

2599 2600 2601 2602 2603 2604 2605 2606 2607
	ret = 1;
	item_size = btrfs_item_size_nr(leaf, path->slots[0]);
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
	if (item_size < sizeof(*ei)) {
		WARN_ON(item_size != sizeof(struct btrfs_extent_item_v0));
		goto out;
	}
#endif
	ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
2608

2609 2610 2611
	if (item_size != sizeof(*ei) +
	    btrfs_extent_inline_ref_size(BTRFS_EXTENT_DATA_REF_KEY))
		goto out;
2612

2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651
	if (btrfs_extent_generation(leaf, ei) <=
	    btrfs_root_last_snapshot(&root->root_item))
		goto out;

	iref = (struct btrfs_extent_inline_ref *)(ei + 1);
	if (btrfs_extent_inline_ref_type(leaf, iref) !=
	    BTRFS_EXTENT_DATA_REF_KEY)
		goto out;

	ref = (struct btrfs_extent_data_ref *)(&iref->offset);
	if (btrfs_extent_refs(leaf, ei) !=
	    btrfs_extent_data_ref_count(leaf, ref) ||
	    btrfs_extent_data_ref_root(leaf, ref) !=
	    root->root_key.objectid ||
	    btrfs_extent_data_ref_objectid(leaf, ref) != objectid ||
	    btrfs_extent_data_ref_offset(leaf, ref) != offset)
		goto out;

	ret = 0;
out:
	return ret;
}

int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
			  struct btrfs_root *root,
			  u64 objectid, u64 offset, u64 bytenr)
{
	struct btrfs_path *path;
	int ret;
	int ret2;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOENT;

	do {
		ret = check_committed_ref(trans, root, path, objectid,
					  offset, bytenr);
		if (ret && ret != -ENOENT)
2652
			goto out;
Y
Yan Zheng 已提交
2653

2654 2655 2656 2657 2658 2659 2660
		ret2 = check_delayed_ref(trans, root, path, objectid,
					 offset, bytenr);
	} while (ret2 == -EAGAIN);

	if (ret2 && ret2 != -ENOENT) {
		ret = ret2;
		goto out;
2661
	}
2662 2663 2664

	if (ret != -ENOENT || ret2 != -ENOENT)
		ret = 0;
2665
out:
Y
Yan Zheng 已提交
2666
	btrfs_free_path(path);
2667 2668
	if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
		WARN_ON(ret > 0);
2669
	return ret;
2670
}
C
Chris Mason 已提交
2671

2672
static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
2673
			   struct btrfs_root *root,
2674
			   struct extent_buffer *buf,
A
Arne Jansen 已提交
2675
			   int full_backref, int inc, int for_cow)
Z
Zheng Yan 已提交
2676 2677
{
	u64 bytenr;
2678 2679
	u64 num_bytes;
	u64 parent;
Z
Zheng Yan 已提交
2680 2681 2682 2683 2684 2685 2686 2687
	u64 ref_root;
	u32 nritems;
	struct btrfs_key key;
	struct btrfs_file_extent_item *fi;
	int i;
	int level;
	int ret = 0;
	int (*process_func)(struct btrfs_trans_handle *, struct btrfs_root *,
A
Arne Jansen 已提交
2688
			    u64, u64, u64, u64, u64, u64, int);
Z
Zheng Yan 已提交
2689 2690 2691 2692 2693

	ref_root = btrfs_header_owner(buf);
	nritems = btrfs_header_nritems(buf);
	level = btrfs_header_level(buf);

2694 2695
	if (!root->ref_cows && level == 0)
		return 0;
Z
Zheng Yan 已提交
2696

2697 2698 2699 2700
	if (inc)
		process_func = btrfs_inc_extent_ref;
	else
		process_func = btrfs_free_extent;
Z
Zheng Yan 已提交
2701

2702 2703 2704 2705 2706 2707
	if (full_backref)
		parent = buf->start;
	else
		parent = 0;

	for (i = 0; i < nritems; i++) {
Z
Zheng Yan 已提交
2708
		if (level == 0) {
2709
			btrfs_item_key_to_cpu(buf, &key, i);
Z
Zheng Yan 已提交
2710 2711
			if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
				continue;
2712
			fi = btrfs_item_ptr(buf, i,
Z
Zheng Yan 已提交
2713 2714 2715 2716 2717 2718 2719
					    struct btrfs_file_extent_item);
			if (btrfs_file_extent_type(buf, fi) ==
			    BTRFS_FILE_EXTENT_INLINE)
				continue;
			bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
			if (bytenr == 0)
				continue;
2720 2721 2722 2723 2724

			num_bytes = btrfs_file_extent_disk_num_bytes(buf, fi);
			key.offset -= btrfs_file_extent_offset(buf, fi);
			ret = process_func(trans, root, bytenr, num_bytes,
					   parent, ref_root, key.objectid,
A
Arne Jansen 已提交
2725
					   key.offset, for_cow);
Z
Zheng Yan 已提交
2726 2727 2728
			if (ret)
				goto fail;
		} else {
2729 2730 2731
			bytenr = btrfs_node_blockptr(buf, i);
			num_bytes = btrfs_level_size(root, level - 1);
			ret = process_func(trans, root, bytenr, num_bytes,
A
Arne Jansen 已提交
2732 2733
					   parent, ref_root, level - 1, 0,
					   for_cow);
Z
Zheng Yan 已提交
2734 2735 2736 2737 2738 2739
			if (ret)
				goto fail;
		}
	}
	return 0;
fail:
2740 2741 2742 2743 2744
	BUG();
	return ret;
}

int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
A
Arne Jansen 已提交
2745
		  struct extent_buffer *buf, int full_backref, int for_cow)
2746
{
A
Arne Jansen 已提交
2747
	return __btrfs_mod_ref(trans, root, buf, full_backref, 1, for_cow);
2748 2749 2750
}

int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
A
Arne Jansen 已提交
2751
		  struct extent_buffer *buf, int full_backref, int for_cow)
2752
{
A
Arne Jansen 已提交
2753
	return __btrfs_mod_ref(trans, root, buf, full_backref, 0, for_cow);
Z
Zheng Yan 已提交
2754 2755
}

C
Chris Mason 已提交
2756 2757 2758 2759 2760 2761 2762
static int write_one_cache_group(struct btrfs_trans_handle *trans,
				 struct btrfs_root *root,
				 struct btrfs_path *path,
				 struct btrfs_block_group_cache *cache)
{
	int ret;
	struct btrfs_root *extent_root = root->fs_info->extent_root;
2763 2764
	unsigned long bi;
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2765 2766

	ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
2767 2768
	if (ret < 0)
		goto fail;
C
Chris Mason 已提交
2769
	BUG_ON(ret);
2770 2771 2772 2773 2774

	leaf = path->nodes[0];
	bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
	write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
	btrfs_mark_buffer_dirty(leaf);
2775
	btrfs_release_path(path);
2776
fail:
C
Chris Mason 已提交
2777 2778 2779 2780 2781 2782
	if (ret)
		return ret;
	return 0;

}

2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793
static struct btrfs_block_group_cache *
next_block_group(struct btrfs_root *root,
		 struct btrfs_block_group_cache *cache)
{
	struct rb_node *node;
	spin_lock(&root->fs_info->block_group_cache_lock);
	node = rb_next(&cache->cache_node);
	btrfs_put_block_group(cache);
	if (node) {
		cache = rb_entry(node, struct btrfs_block_group_cache,
				 cache_node);
2794
		btrfs_get_block_group(cache);
2795 2796 2797 2798 2799 2800
	} else
		cache = NULL;
	spin_unlock(&root->fs_info->block_group_cache_lock);
	return cache;
}

2801 2802 2803 2804 2805 2806 2807
static int cache_save_setup(struct btrfs_block_group_cache *block_group,
			    struct btrfs_trans_handle *trans,
			    struct btrfs_path *path)
{
	struct btrfs_root *root = block_group->fs_info->tree_root;
	struct inode *inode = NULL;
	u64 alloc_hint = 0;
2808
	int dcs = BTRFS_DC_ERROR;
2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827
	int num_pages = 0;
	int retries = 0;
	int ret = 0;

	/*
	 * If this block group is smaller than 100 megs don't bother caching the
	 * block group.
	 */
	if (block_group->key.offset < (100 * 1024 * 1024)) {
		spin_lock(&block_group->lock);
		block_group->disk_cache_state = BTRFS_DC_WRITTEN;
		spin_unlock(&block_group->lock);
		return 0;
	}

again:
	inode = lookup_free_space_inode(root, block_group, path);
	if (IS_ERR(inode) && PTR_ERR(inode) != -ENOENT) {
		ret = PTR_ERR(inode);
2828
		btrfs_release_path(path);
2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844
		goto out;
	}

	if (IS_ERR(inode)) {
		BUG_ON(retries);
		retries++;

		if (block_group->ro)
			goto out_free;

		ret = create_free_space_inode(root, trans, block_group, path);
		if (ret)
			goto out_free;
		goto again;
	}

2845 2846 2847 2848 2849 2850 2851
	/* We've already setup this transaction, go ahead and exit */
	if (block_group->cache_generation == trans->transid &&
	    i_size_read(inode)) {
		dcs = BTRFS_DC_SETUP;
		goto out_put;
	}

2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869
	/*
	 * We want to set the generation to 0, that way if anything goes wrong
	 * from here on out we know not to trust this cache when we load up next
	 * time.
	 */
	BTRFS_I(inode)->generation = 0;
	ret = btrfs_update_inode(trans, root, inode);
	WARN_ON(ret);

	if (i_size_read(inode) > 0) {
		ret = btrfs_truncate_free_space_cache(root, trans, path,
						      inode);
		if (ret)
			goto out_put;
	}

	spin_lock(&block_group->lock);
	if (block_group->cached != BTRFS_CACHE_FINISHED) {
2870 2871
		/* We're not cached, don't bother trying to write stuff out */
		dcs = BTRFS_DC_WRITTEN;
2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897
		spin_unlock(&block_group->lock);
		goto out_put;
	}
	spin_unlock(&block_group->lock);

	num_pages = (int)div64_u64(block_group->key.offset, 1024 * 1024 * 1024);
	if (!num_pages)
		num_pages = 1;

	/*
	 * Just to make absolutely sure we have enough space, we're going to
	 * preallocate 12 pages worth of space for each block group.  In
	 * practice we ought to use at most 8, but we need extra space so we can
	 * add our header and have a terminator between the extents and the
	 * bitmaps.
	 */
	num_pages *= 16;
	num_pages *= PAGE_CACHE_SIZE;

	ret = btrfs_check_data_free_space(inode, num_pages);
	if (ret)
		goto out_put;

	ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, num_pages,
					      num_pages, num_pages,
					      &alloc_hint);
2898 2899
	if (!ret)
		dcs = BTRFS_DC_SETUP;
2900
	btrfs_free_reserved_data_space(inode, num_pages);
2901

2902 2903 2904
out_put:
	iput(inode);
out_free:
2905
	btrfs_release_path(path);
2906 2907
out:
	spin_lock(&block_group->lock);
2908
	if (!ret && dcs == BTRFS_DC_SETUP)
2909
		block_group->cache_generation = trans->transid;
2910
	block_group->disk_cache_state = dcs;
2911 2912 2913 2914 2915
	spin_unlock(&block_group->lock);

	return ret;
}

2916 2917
int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
				   struct btrfs_root *root)
C
Chris Mason 已提交
2918
{
2919
	struct btrfs_block_group_cache *cache;
C
Chris Mason 已提交
2920 2921
	int err = 0;
	struct btrfs_path *path;
2922
	u64 last = 0;
C
Chris Mason 已提交
2923 2924 2925 2926 2927

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946
again:
	while (1) {
		cache = btrfs_lookup_first_block_group(root->fs_info, last);
		while (cache) {
			if (cache->disk_cache_state == BTRFS_DC_CLEAR)
				break;
			cache = next_block_group(root, cache);
		}
		if (!cache) {
			if (last == 0)
				break;
			last = 0;
			continue;
		}
		err = cache_save_setup(cache, trans, path);
		last = cache->key.objectid + cache->key.offset;
		btrfs_put_block_group(cache);
	}

C
Chris Mason 已提交
2947
	while (1) {
2948 2949 2950 2951
		if (last == 0) {
			err = btrfs_run_delayed_refs(trans, root,
						     (unsigned long)-1);
			BUG_ON(err);
J
Josef Bacik 已提交
2952
		}
2953

2954 2955
		cache = btrfs_lookup_first_block_group(root->fs_info, last);
		while (cache) {
2956 2957 2958 2959 2960
			if (cache->disk_cache_state == BTRFS_DC_CLEAR) {
				btrfs_put_block_group(cache);
				goto again;
			}

2961 2962 2963 2964 2965 2966 2967 2968 2969 2970
			if (cache->dirty)
				break;
			cache = next_block_group(root, cache);
		}
		if (!cache) {
			if (last == 0)
				break;
			last = 0;
			continue;
		}
J
Josef Bacik 已提交
2971

J
Josef Bacik 已提交
2972 2973
		if (cache->disk_cache_state == BTRFS_DC_SETUP)
			cache->disk_cache_state = BTRFS_DC_NEED_WRITE;
2974
		cache->dirty = 0;
2975
		last = cache->key.objectid + cache->key.offset;
J
Josef Bacik 已提交
2976

2977 2978 2979
		err = write_one_cache_group(trans, root, path, cache);
		BUG_ON(err);
		btrfs_put_block_group(cache);
C
Chris Mason 已提交
2980
	}
2981

J
Josef Bacik 已提交
2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027
	while (1) {
		/*
		 * I don't think this is needed since we're just marking our
		 * preallocated extent as written, but just in case it can't
		 * hurt.
		 */
		if (last == 0) {
			err = btrfs_run_delayed_refs(trans, root,
						     (unsigned long)-1);
			BUG_ON(err);
		}

		cache = btrfs_lookup_first_block_group(root->fs_info, last);
		while (cache) {
			/*
			 * Really this shouldn't happen, but it could if we
			 * couldn't write the entire preallocated extent and
			 * splitting the extent resulted in a new block.
			 */
			if (cache->dirty) {
				btrfs_put_block_group(cache);
				goto again;
			}
			if (cache->disk_cache_state == BTRFS_DC_NEED_WRITE)
				break;
			cache = next_block_group(root, cache);
		}
		if (!cache) {
			if (last == 0)
				break;
			last = 0;
			continue;
		}

		btrfs_write_out_cache(root, trans, cache, path);

		/*
		 * If we didn't have an error then the cache state is still
		 * NEED_WRITE, so we can set it to WRITTEN.
		 */
		if (cache->disk_cache_state == BTRFS_DC_NEED_WRITE)
			cache->disk_cache_state = BTRFS_DC_WRITTEN;
		last = cache->key.objectid + cache->key.offset;
		btrfs_put_block_group(cache);
	}

C
Chris Mason 已提交
3028
	btrfs_free_path(path);
3029
	return 0;
C
Chris Mason 已提交
3030 3031
}

3032 3033 3034 3035 3036 3037 3038 3039 3040
int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr)
{
	struct btrfs_block_group_cache *block_group;
	int readonly = 0;

	block_group = btrfs_lookup_block_group(root->fs_info, bytenr);
	if (!block_group || block_group->ro)
		readonly = 1;
	if (block_group)
3041
		btrfs_put_block_group(block_group);
3042 3043 3044
	return readonly;
}

3045 3046 3047 3048 3049
static int update_space_info(struct btrfs_fs_info *info, u64 flags,
			     u64 total_bytes, u64 bytes_used,
			     struct btrfs_space_info **space_info)
{
	struct btrfs_space_info *found;
3050 3051 3052 3053 3054 3055 3056 3057
	int i;
	int factor;

	if (flags & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1 |
		     BTRFS_BLOCK_GROUP_RAID10))
		factor = 2;
	else
		factor = 1;
3058 3059 3060

	found = __find_space_info(info, flags);
	if (found) {
3061
		spin_lock(&found->lock);
3062
		found->total_bytes += total_bytes;
J
Josef Bacik 已提交
3063
		found->disk_total += total_bytes * factor;
3064
		found->bytes_used += bytes_used;
3065
		found->disk_used += bytes_used * factor;
3066
		found->full = 0;
3067
		spin_unlock(&found->lock);
3068 3069 3070
		*space_info = found;
		return 0;
	}
Y
Yan Zheng 已提交
3071
	found = kzalloc(sizeof(*found), GFP_NOFS);
3072 3073 3074
	if (!found)
		return -ENOMEM;

3075 3076
	for (i = 0; i < BTRFS_NR_RAID_TYPES; i++)
		INIT_LIST_HEAD(&found->block_groups[i]);
3077
	init_rwsem(&found->groups_sem);
J
Josef Bacik 已提交
3078
	spin_lock_init(&found->lock);
3079
	found->flags = flags & BTRFS_BLOCK_GROUP_TYPE_MASK;
3080
	found->total_bytes = total_bytes;
J
Josef Bacik 已提交
3081
	found->disk_total = total_bytes * factor;
3082
	found->bytes_used = bytes_used;
3083
	found->disk_used = bytes_used * factor;
3084
	found->bytes_pinned = 0;
3085
	found->bytes_reserved = 0;
Y
Yan Zheng 已提交
3086
	found->bytes_readonly = 0;
3087
	found->bytes_may_use = 0;
3088
	found->full = 0;
3089
	found->force_alloc = CHUNK_ALLOC_NO_FORCE;
3090
	found->chunk_alloc = 0;
3091 3092
	found->flush = 0;
	init_waitqueue_head(&found->wait);
3093
	*space_info = found;
3094
	list_add_rcu(&found->list, &info->space_info);
3095 3096 3097
	return 0;
}

3098 3099
static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
{
3100
	u64 extra_flags = flags & BTRFS_BLOCK_GROUP_PROFILE_MASK;
3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111

	/* chunk -> extended profile */
	if (extra_flags == 0)
		extra_flags = BTRFS_AVAIL_ALLOC_BIT_SINGLE;

	if (flags & BTRFS_BLOCK_GROUP_DATA)
		fs_info->avail_data_alloc_bits |= extra_flags;
	if (flags & BTRFS_BLOCK_GROUP_METADATA)
		fs_info->avail_metadata_alloc_bits |= extra_flags;
	if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
		fs_info->avail_system_alloc_bits |= extra_flags;
3112
}
3113

3114 3115 3116
/*
 * @flags: available profiles in extended format (see ctree.h)
 *
3117 3118 3119
 * Returns reduced profile in chunk format.  If profile changing is in
 * progress (either running or paused) picks the target profile (if it's
 * already available), otherwise falls back to plain reducing.
3120
 */
Y
Yan Zheng 已提交
3121
u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
3122
{
3123 3124 3125 3126 3127 3128 3129
	/*
	 * we add in the count of missing devices because we want
	 * to make sure that any RAID levels on a degraded FS
	 * continue to be honored.
	 */
	u64 num_devices = root->fs_info->fs_devices->rw_devices +
		root->fs_info->fs_devices->missing_devices;
3130

3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158
	/* pick restriper's target profile if it's available */
	spin_lock(&root->fs_info->balance_lock);
	if (root->fs_info->balance_ctl) {
		struct btrfs_balance_control *bctl = root->fs_info->balance_ctl;
		u64 tgt = 0;

		if ((flags & BTRFS_BLOCK_GROUP_DATA) &&
		    (bctl->data.flags & BTRFS_BALANCE_ARGS_CONVERT) &&
		    (flags & bctl->data.target)) {
			tgt = BTRFS_BLOCK_GROUP_DATA | bctl->data.target;
		} else if ((flags & BTRFS_BLOCK_GROUP_SYSTEM) &&
			   (bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) &&
			   (flags & bctl->sys.target)) {
			tgt = BTRFS_BLOCK_GROUP_SYSTEM | bctl->sys.target;
		} else if ((flags & BTRFS_BLOCK_GROUP_METADATA) &&
			   (bctl->meta.flags & BTRFS_BALANCE_ARGS_CONVERT) &&
			   (flags & bctl->meta.target)) {
			tgt = BTRFS_BLOCK_GROUP_METADATA | bctl->meta.target;
		}

		if (tgt) {
			spin_unlock(&root->fs_info->balance_lock);
			flags = tgt;
			goto out;
		}
	}
	spin_unlock(&root->fs_info->balance_lock);

3159 3160 3161 3162 3163
	if (num_devices == 1)
		flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
	if (num_devices < 4)
		flags &= ~BTRFS_BLOCK_GROUP_RAID10;

3164 3165
	if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
	    (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3166
		      BTRFS_BLOCK_GROUP_RAID10))) {
3167
		flags &= ~BTRFS_BLOCK_GROUP_DUP;
3168
	}
3169 3170

	if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
3171
	    (flags & BTRFS_BLOCK_GROUP_RAID10)) {
3172
		flags &= ~BTRFS_BLOCK_GROUP_RAID1;
3173
	}
3174 3175 3176 3177

	if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
	    ((flags & BTRFS_BLOCK_GROUP_RAID1) |
	     (flags & BTRFS_BLOCK_GROUP_RAID10) |
3178
	     (flags & BTRFS_BLOCK_GROUP_DUP))) {
3179
		flags &= ~BTRFS_BLOCK_GROUP_RAID0;
3180 3181
	}

3182
out:
3183 3184
	/* extended -> chunk profile */
	flags &= ~BTRFS_AVAIL_ALLOC_BIT_SINGLE;
3185 3186 3187
	return flags;
}

3188
static u64 get_alloc_profile(struct btrfs_root *root, u64 flags)
J
Josef Bacik 已提交
3189
{
3190
	if (flags & BTRFS_BLOCK_GROUP_DATA)
3191
		flags |= root->fs_info->avail_data_alloc_bits;
3192
	else if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
3193
		flags |= root->fs_info->avail_system_alloc_bits;
3194
	else if (flags & BTRFS_BLOCK_GROUP_METADATA)
3195 3196
		flags |= root->fs_info->avail_metadata_alloc_bits;

3197
	return btrfs_reduce_alloc_profile(root, flags);
J
Josef Bacik 已提交
3198 3199
}

3200
u64 btrfs_get_alloc_profile(struct btrfs_root *root, int data)
J
Josef Bacik 已提交
3201
{
3202
	u64 flags;
J
Josef Bacik 已提交
3203

3204 3205 3206 3207
	if (data)
		flags = BTRFS_BLOCK_GROUP_DATA;
	else if (root == root->fs_info->chunk_root)
		flags = BTRFS_BLOCK_GROUP_SYSTEM;
J
Josef Bacik 已提交
3208
	else
3209
		flags = BTRFS_BLOCK_GROUP_METADATA;
J
Josef Bacik 已提交
3210

3211
	return get_alloc_profile(root, flags);
J
Josef Bacik 已提交
3212
}
J
Josef Bacik 已提交
3213

J
Josef Bacik 已提交
3214 3215 3216
void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *inode)
{
	BTRFS_I(inode)->space_info = __find_space_info(root->fs_info,
3217
						       BTRFS_BLOCK_GROUP_DATA);
J
Josef Bacik 已提交
3218 3219
}

J
Josef Bacik 已提交
3220 3221 3222 3223
/*
 * This will check the space that the inode allocates from to make sure we have
 * enough space for bytes.
 */
3224
int btrfs_check_data_free_space(struct inode *inode, u64 bytes)
J
Josef Bacik 已提交
3225 3226
{
	struct btrfs_space_info *data_sinfo;
3227
	struct btrfs_root *root = BTRFS_I(inode)->root;
3228
	u64 used;
3229
	int ret = 0, committed = 0, alloc_chunk = 1;
J
Josef Bacik 已提交
3230 3231 3232 3233

	/* make sure bytes are sectorsize aligned */
	bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);

3234 3235
	if (root == root->fs_info->tree_root ||
	    BTRFS_I(inode)->location.objectid == BTRFS_FREE_INO_OBJECTID) {
3236 3237 3238 3239
		alloc_chunk = 0;
		committed = 1;
	}

J
Josef Bacik 已提交
3240
	data_sinfo = BTRFS_I(inode)->space_info;
C
Chris Mason 已提交
3241 3242
	if (!data_sinfo)
		goto alloc;
J
Josef Bacik 已提交
3243

J
Josef Bacik 已提交
3244 3245 3246
again:
	/* make sure we have enough space to handle the data first */
	spin_lock(&data_sinfo->lock);
3247 3248 3249
	used = data_sinfo->bytes_used + data_sinfo->bytes_reserved +
		data_sinfo->bytes_pinned + data_sinfo->bytes_readonly +
		data_sinfo->bytes_may_use;
3250 3251

	if (used + bytes > data_sinfo->total_bytes) {
3252
		struct btrfs_trans_handle *trans;
J
Josef Bacik 已提交
3253

J
Josef Bacik 已提交
3254 3255 3256 3257
		/*
		 * if we don't have enough free bytes in this space then we need
		 * to alloc a new chunk.
		 */
3258
		if (!data_sinfo->full && alloc_chunk) {
J
Josef Bacik 已提交
3259
			u64 alloc_target;
J
Josef Bacik 已提交
3260

3261
			data_sinfo->force_alloc = CHUNK_ALLOC_FORCE;
J
Josef Bacik 已提交
3262
			spin_unlock(&data_sinfo->lock);
C
Chris Mason 已提交
3263
alloc:
J
Josef Bacik 已提交
3264
			alloc_target = btrfs_get_alloc_profile(root, 1);
3265
			trans = btrfs_join_transaction(root);
3266 3267
			if (IS_ERR(trans))
				return PTR_ERR(trans);
J
Josef Bacik 已提交
3268

J
Josef Bacik 已提交
3269 3270
			ret = do_chunk_alloc(trans, root->fs_info->extent_root,
					     bytes + 2 * 1024 * 1024,
3271 3272
					     alloc_target,
					     CHUNK_ALLOC_NO_FORCE);
J
Josef Bacik 已提交
3273
			btrfs_end_transaction(trans, root);
3274 3275 3276 3277 3278 3279
			if (ret < 0) {
				if (ret != -ENOSPC)
					return ret;
				else
					goto commit_trans;
			}
J
Josef Bacik 已提交
3280

C
Chris Mason 已提交
3281 3282 3283 3284
			if (!data_sinfo) {
				btrfs_set_inode_space_info(root, inode);
				data_sinfo = BTRFS_I(inode)->space_info;
			}
J
Josef Bacik 已提交
3285 3286
			goto again;
		}
3287 3288 3289 3290 3291 3292 3293

		/*
		 * If we have less pinned bytes than we want to allocate then
		 * don't bother committing the transaction, it won't help us.
		 */
		if (data_sinfo->bytes_pinned < bytes)
			committed = 1;
J
Josef Bacik 已提交
3294 3295
		spin_unlock(&data_sinfo->lock);

3296
		/* commit the current transaction and try again */
3297
commit_trans:
J
Josef Bacik 已提交
3298 3299
		if (!committed &&
		    !atomic_read(&root->fs_info->open_ioctl_trans)) {
3300
			committed = 1;
3301
			trans = btrfs_join_transaction(root);
3302 3303
			if (IS_ERR(trans))
				return PTR_ERR(trans);
3304 3305 3306 3307 3308
			ret = btrfs_commit_transaction(trans, root);
			if (ret)
				return ret;
			goto again;
		}
J
Josef Bacik 已提交
3309

J
Josef Bacik 已提交
3310 3311 3312
		return -ENOSPC;
	}
	data_sinfo->bytes_may_use += bytes;
J
Josef Bacik 已提交
3313 3314
	trace_btrfs_space_reservation(root->fs_info, "space_info",
				      (u64)data_sinfo, bytes, 1);
J
Josef Bacik 已提交
3315 3316
	spin_unlock(&data_sinfo->lock);

J
Josef Bacik 已提交
3317 3318
	return 0;
}
J
Josef Bacik 已提交
3319 3320

/*
3321
 * Called if we need to clear a data reservation for this inode.
J
Josef Bacik 已提交
3322
 */
3323
void btrfs_free_reserved_data_space(struct inode *inode, u64 bytes)
3324
{
3325
	struct btrfs_root *root = BTRFS_I(inode)->root;
J
Josef Bacik 已提交
3326
	struct btrfs_space_info *data_sinfo;
3327

J
Josef Bacik 已提交
3328 3329
	/* make sure bytes are sectorsize aligned */
	bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
3330

J
Josef Bacik 已提交
3331 3332 3333
	data_sinfo = BTRFS_I(inode)->space_info;
	spin_lock(&data_sinfo->lock);
	data_sinfo->bytes_may_use -= bytes;
J
Josef Bacik 已提交
3334 3335
	trace_btrfs_space_reservation(root->fs_info, "space_info",
				      (u64)data_sinfo, bytes, 0);
J
Josef Bacik 已提交
3336
	spin_unlock(&data_sinfo->lock);
3337 3338
}

3339
static void force_metadata_allocation(struct btrfs_fs_info *info)
3340
{
3341 3342
	struct list_head *head = &info->space_info;
	struct btrfs_space_info *found;
3343

3344 3345 3346
	rcu_read_lock();
	list_for_each_entry_rcu(found, head, list) {
		if (found->flags & BTRFS_BLOCK_GROUP_METADATA)
3347
			found->force_alloc = CHUNK_ALLOC_FORCE;
3348
	}
3349
	rcu_read_unlock();
3350 3351
}

3352
static int should_alloc_chunk(struct btrfs_root *root,
3353 3354
			      struct btrfs_space_info *sinfo, u64 alloc_bytes,
			      int force)
3355
{
3356
	struct btrfs_block_rsv *global_rsv = &root->fs_info->global_block_rsv;
3357
	u64 num_bytes = sinfo->total_bytes - sinfo->bytes_readonly;
3358
	u64 num_allocated = sinfo->bytes_used + sinfo->bytes_reserved;
3359
	u64 thresh;
3360

3361 3362 3363
	if (force == CHUNK_ALLOC_FORCE)
		return 1;

3364 3365 3366 3367 3368 3369 3370
	/*
	 * We need to take into account the global rsv because for all intents
	 * and purposes it's used space.  Don't worry about locking the
	 * global_rsv, it doesn't change except when the transaction commits.
	 */
	num_allocated += global_rsv->size;

3371 3372 3373 3374 3375
	/*
	 * in limited mode, we want to have some free space up to
	 * about 1% of the FS size.
	 */
	if (force == CHUNK_ALLOC_LIMITED) {
3376
		thresh = btrfs_super_total_bytes(root->fs_info->super_copy);
3377 3378 3379 3380 3381 3382
		thresh = max_t(u64, 64 * 1024 * 1024,
			       div_factor_fine(thresh, 1));

		if (num_bytes - num_allocated < thresh)
			return 1;
	}
3383
	thresh = btrfs_super_total_bytes(root->fs_info->super_copy);
3384

3385 3386
	/* 256MB or 2% of the FS */
	thresh = max_t(u64, 256 * 1024 * 1024, div_factor_fine(thresh, 2));
3387

3388
	if (num_bytes > thresh && sinfo->bytes_used < div_factor(num_bytes, 8))
3389
		return 0;
3390
	return 1;
3391 3392
}

3393 3394
static int do_chunk_alloc(struct btrfs_trans_handle *trans,
			  struct btrfs_root *extent_root, u64 alloc_bytes,
C
Chris Mason 已提交
3395
			  u64 flags, int force)
J
Josef Bacik 已提交
3396
{
3397
	struct btrfs_space_info *space_info;
3398
	struct btrfs_fs_info *fs_info = extent_root->fs_info;
3399
	int wait_for_alloc = 0;
J
Josef Bacik 已提交
3400 3401
	int ret = 0;

3402
	BUG_ON(!profile_is_valid(flags, 0));
3403

3404
	space_info = __find_space_info(extent_root->fs_info, flags);
3405 3406 3407 3408
	if (!space_info) {
		ret = update_space_info(extent_root->fs_info, flags,
					0, 0, &space_info);
		BUG_ON(ret);
J
Josef Bacik 已提交
3409
	}
3410
	BUG_ON(!space_info);
J
Josef Bacik 已提交
3411

3412
again:
3413
	spin_lock(&space_info->lock);
J
Josef Bacik 已提交
3414
	if (space_info->force_alloc)
3415
		force = space_info->force_alloc;
3416 3417
	if (space_info->full) {
		spin_unlock(&space_info->lock);
3418
		return 0;
J
Josef Bacik 已提交
3419 3420
	}

3421
	if (!should_alloc_chunk(extent_root, space_info, alloc_bytes, force)) {
3422
		spin_unlock(&space_info->lock);
3423 3424 3425 3426 3427
		return 0;
	} else if (space_info->chunk_alloc) {
		wait_for_alloc = 1;
	} else {
		space_info->chunk_alloc = 1;
J
Josef Bacik 已提交
3428
	}
3429

3430
	spin_unlock(&space_info->lock);
J
Josef Bacik 已提交
3431

3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445
	mutex_lock(&fs_info->chunk_mutex);

	/*
	 * The chunk_mutex is held throughout the entirety of a chunk
	 * allocation, so once we've acquired the chunk_mutex we know that the
	 * other guy is done and we need to recheck and see if we should
	 * allocate.
	 */
	if (wait_for_alloc) {
		mutex_unlock(&fs_info->chunk_mutex);
		wait_for_alloc = 0;
		goto again;
	}

3446 3447 3448 3449 3450 3451 3452
	/*
	 * If we have mixed data/metadata chunks we want to make sure we keep
	 * allocating mixed chunks instead of individual chunks.
	 */
	if (btrfs_mixed_space_info(space_info))
		flags |= (BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA);

3453 3454 3455 3456 3457
	/*
	 * if we're doing a data chunk, go ahead and make sure that
	 * we keep a reasonable number of metadata chunks allocated in the
	 * FS as well.
	 */
J
Josef Bacik 已提交
3458
	if (flags & BTRFS_BLOCK_GROUP_DATA && fs_info->metadata_ratio) {
3459 3460 3461 3462
		fs_info->data_chunk_allocations++;
		if (!(fs_info->data_chunk_allocations %
		      fs_info->metadata_ratio))
			force_metadata_allocation(fs_info);
J
Josef Bacik 已提交
3463 3464
	}

Y
Yan Zheng 已提交
3465
	ret = btrfs_alloc_chunk(trans, extent_root, flags);
3466 3467 3468
	if (ret < 0 && ret != -ENOSPC)
		goto out;

J
Josef Bacik 已提交
3469 3470
	spin_lock(&space_info->lock);
	if (ret)
3471
		space_info->full = 1;
3472 3473
	else
		ret = 1;
3474

3475
	space_info->force_alloc = CHUNK_ALLOC_NO_FORCE;
3476
	space_info->chunk_alloc = 0;
J
Josef Bacik 已提交
3477
	spin_unlock(&space_info->lock);
3478
out:
Y
Yan Zheng 已提交
3479
	mutex_unlock(&extent_root->fs_info->chunk_mutex);
J
Josef Bacik 已提交
3480
	return ret;
3481
}
J
Josef Bacik 已提交
3482 3483

/*
3484
 * shrink metadata reservation for delalloc
J
Josef Bacik 已提交
3485
 */
3486
static int shrink_delalloc(struct btrfs_root *root, u64 to_reclaim,
3487
			   bool wait_ordered)
3488
{
3489
	struct btrfs_block_rsv *block_rsv;
J
Josef Bacik 已提交
3490
	struct btrfs_space_info *space_info;
3491
	struct btrfs_trans_handle *trans;
3492 3493 3494
	u64 reserved;
	u64 max_reclaim;
	u64 reclaimed = 0;
3495
	long time_left;
3496
	unsigned long nr_pages = (2 * 1024 * 1024) >> PAGE_CACHE_SHIFT;
3497
	int loops = 0;
3498
	unsigned long progress;
3499

3500
	trans = (struct btrfs_trans_handle *)current->journal_info;
3501
	block_rsv = &root->fs_info->delalloc_block_rsv;
J
Josef Bacik 已提交
3502
	space_info = block_rsv->space_info;
3503 3504

	smp_mb();
3505
	reserved = space_info->bytes_may_use;
3506
	progress = space_info->reservation_progress;
3507 3508 3509

	if (reserved == 0)
		return 0;
3510

3511 3512 3513 3514 3515 3516 3517 3518
	smp_mb();
	if (root->fs_info->delalloc_bytes == 0) {
		if (trans)
			return 0;
		btrfs_wait_ordered_extents(root, 0, 0);
		return 0;
	}

3519
	max_reclaim = min(reserved, to_reclaim);
3520 3521
	nr_pages = max_t(unsigned long, nr_pages,
			 max_reclaim >> PAGE_CACHE_SHIFT);
3522
	while (loops < 1024) {
3523 3524 3525 3526
		/* have the flusher threads jump in and do some IO */
		smp_mb();
		nr_pages = min_t(unsigned long, nr_pages,
		       root->fs_info->delalloc_bytes >> PAGE_CACHE_SHIFT);
3527 3528
		writeback_inodes_sb_nr_if_idle(root->fs_info->sb, nr_pages,
						WB_REASON_FS_FREE_SPACE);
3529

J
Josef Bacik 已提交
3530
		spin_lock(&space_info->lock);
3531 3532 3533
		if (reserved > space_info->bytes_may_use)
			reclaimed += reserved - space_info->bytes_may_use;
		reserved = space_info->bytes_may_use;
J
Josef Bacik 已提交
3534
		spin_unlock(&space_info->lock);
3535

3536 3537
		loops++;

3538 3539 3540 3541 3542
		if (reserved == 0 || reclaimed >= max_reclaim)
			break;

		if (trans && trans->transaction->blocked)
			return -EAGAIN;
3543

3544 3545 3546 3547
		if (wait_ordered && !trans) {
			btrfs_wait_ordered_extents(root, 0, 0);
		} else {
			time_left = schedule_timeout_interruptible(1);
3548

3549 3550 3551 3552
			/* We were interrupted, exit */
			if (time_left)
				break;
		}
3553

3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564
		/* we've kicked the IO a few times, if anything has been freed,
		 * exit.  There is no sense in looping here for a long time
		 * when we really need to commit the transaction, or there are
		 * just too many writers without enough free space
		 */

		if (loops > 3) {
			smp_mb();
			if (progress != space_info->reservation_progress)
				break;
		}
3565

3566
	}
3567

3568 3569 3570
	return reclaimed >= to_reclaim;
}

3571 3572 3573 3574 3575
/**
 * maybe_commit_transaction - possibly commit the transaction if its ok to
 * @root - the root we're allocating for
 * @bytes - the number of bytes we want to reserve
 * @force - force the commit
3576
 *
3577 3578 3579
 * This will check to make sure that committing the transaction will actually
 * get us somewhere and then commit the transaction if it does.  Otherwise it
 * will return -ENOSPC.
3580
 */
3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624
static int may_commit_transaction(struct btrfs_root *root,
				  struct btrfs_space_info *space_info,
				  u64 bytes, int force)
{
	struct btrfs_block_rsv *delayed_rsv = &root->fs_info->delayed_block_rsv;
	struct btrfs_trans_handle *trans;

	trans = (struct btrfs_trans_handle *)current->journal_info;
	if (trans)
		return -EAGAIN;

	if (force)
		goto commit;

	/* See if there is enough pinned space to make this reservation */
	spin_lock(&space_info->lock);
	if (space_info->bytes_pinned >= bytes) {
		spin_unlock(&space_info->lock);
		goto commit;
	}
	spin_unlock(&space_info->lock);

	/*
	 * See if there is some space in the delayed insertion reservation for
	 * this reservation.
	 */
	if (space_info != delayed_rsv->space_info)
		return -ENOSPC;

	spin_lock(&delayed_rsv->lock);
	if (delayed_rsv->size < bytes) {
		spin_unlock(&delayed_rsv->lock);
		return -ENOSPC;
	}
	spin_unlock(&delayed_rsv->lock);

commit:
	trans = btrfs_join_transaction(root);
	if (IS_ERR(trans))
		return -ENOSPC;

	return btrfs_commit_transaction(trans, root);
}

3625 3626 3627 3628 3629 3630
/**
 * reserve_metadata_bytes - try to reserve bytes from the block_rsv's space
 * @root - the root we're allocating for
 * @block_rsv - the block_rsv we're allocating for
 * @orig_bytes - the number of bytes we want
 * @flush - wether or not we can flush to make our reservation
3631
 *
3632 3633 3634 3635 3636 3637
 * This will reserve orgi_bytes number of bytes from the space info associated
 * with the block_rsv.  If there is not enough space it will make an attempt to
 * flush out space to make room.  It will do this by flushing delalloc if
 * possible or committing the transaction.  If flush is 0 then no attempts to
 * regain reservations will be made and this will fail if there is not enough
 * space already.
3638
 */
3639
static int reserve_metadata_bytes(struct btrfs_root *root,
3640 3641
				  struct btrfs_block_rsv *block_rsv,
				  u64 orig_bytes, int flush)
J
Josef Bacik 已提交
3642
{
3643
	struct btrfs_space_info *space_info = block_rsv->space_info;
3644
	u64 used;
3645 3646 3647
	u64 num_bytes = orig_bytes;
	int retries = 0;
	int ret = 0;
3648
	bool committed = false;
3649
	bool flushing = false;
3650
	bool wait_ordered = false;
J
Josef Bacik 已提交
3651

3652
again:
3653
	ret = 0;
3654
	spin_lock(&space_info->lock);
3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666
	/*
	 * We only want to wait if somebody other than us is flushing and we are
	 * actually alloed to flush.
	 */
	while (flush && !flushing && space_info->flush) {
		spin_unlock(&space_info->lock);
		/*
		 * If we have a trans handle we can't wait because the flusher
		 * may have to commit the transaction, which would mean we would
		 * deadlock since we are waiting for the flusher to finish, but
		 * hold the current transaction open.
		 */
3667
		if (current->journal_info)
3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678
			return -EAGAIN;
		ret = wait_event_interruptible(space_info->wait,
					       !space_info->flush);
		/* Must have been interrupted, return */
		if (ret)
			return -EINTR;

		spin_lock(&space_info->lock);
	}

	ret = -ENOSPC;
3679 3680 3681
	used = space_info->bytes_used + space_info->bytes_reserved +
		space_info->bytes_pinned + space_info->bytes_readonly +
		space_info->bytes_may_use;
J
Josef Bacik 已提交
3682

3683 3684 3685 3686 3687 3688 3689
	/*
	 * The idea here is that we've not already over-reserved the block group
	 * then we can go ahead and save our reservation first and then start
	 * flushing if we need to.  Otherwise if we've already overcommitted
	 * lets start flushing stuff first and then come back and try to make
	 * our reservation.
	 */
3690 3691
	if (used <= space_info->total_bytes) {
		if (used + orig_bytes <= space_info->total_bytes) {
3692
			space_info->bytes_may_use += orig_bytes;
J
Josef Bacik 已提交
3693 3694 3695 3696
			trace_btrfs_space_reservation(root->fs_info,
						      "space_info",
						      (u64)space_info,
						      orig_bytes, 1);
3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711
			ret = 0;
		} else {
			/*
			 * Ok set num_bytes to orig_bytes since we aren't
			 * overocmmitted, this way we only try and reclaim what
			 * we need.
			 */
			num_bytes = orig_bytes;
		}
	} else {
		/*
		 * Ok we're over committed, set num_bytes to the overcommitted
		 * amount plus the amount of bytes that we need for this
		 * reservation.
		 */
3712
		wait_ordered = true;
3713
		num_bytes = used - space_info->total_bytes +
3714 3715
			(orig_bytes * (retries + 1));
	}
J
Josef Bacik 已提交
3716

3717
	if (ret) {
3718 3719 3720
		u64 profile = btrfs_get_alloc_profile(root, 0);
		u64 avail;

3721 3722 3723 3724 3725 3726
		/*
		 * If we have a lot of space that's pinned, don't bother doing
		 * the overcommit dance yet and just commit the transaction.
		 */
		avail = (space_info->total_bytes - space_info->bytes_used) * 8;
		do_div(avail, 10);
3727
		if (space_info->bytes_pinned >= avail && flush && !committed) {
3728 3729 3730
			space_info->flush = 1;
			flushing = true;
			spin_unlock(&space_info->lock);
3731 3732 3733 3734 3735 3736
			ret = may_commit_transaction(root, space_info,
						     orig_bytes, 1);
			if (ret)
				goto out;
			committed = true;
			goto again;
3737 3738
		}

3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761
		spin_lock(&root->fs_info->free_chunk_lock);
		avail = root->fs_info->free_chunk_space;

		/*
		 * If we have dup, raid1 or raid10 then only half of the free
		 * space is actually useable.
		 */
		if (profile & (BTRFS_BLOCK_GROUP_DUP |
			       BTRFS_BLOCK_GROUP_RAID1 |
			       BTRFS_BLOCK_GROUP_RAID10))
			avail >>= 1;

		/*
		 * If we aren't flushing don't let us overcommit too much, say
		 * 1/8th of the space.  If we can flush, let it overcommit up to
		 * 1/2 of the space.
		 */
		if (flush)
			avail >>= 3;
		else
			avail >>= 1;
		 spin_unlock(&root->fs_info->free_chunk_lock);

3762
		if (used + num_bytes < space_info->total_bytes + avail) {
3763
			space_info->bytes_may_use += orig_bytes;
J
Josef Bacik 已提交
3764 3765 3766 3767
			trace_btrfs_space_reservation(root->fs_info,
						      "space_info",
						      (u64)space_info,
						      orig_bytes, 1);
3768
			ret = 0;
3769 3770
		} else {
			wait_ordered = true;
3771 3772 3773
		}
	}

3774 3775 3776 3777 3778
	/*
	 * Couldn't make our reservation, save our place so while we're trying
	 * to reclaim space we can actually use it instead of somebody else
	 * stealing it from us.
	 */
3779 3780 3781
	if (ret && flush) {
		flushing = true;
		space_info->flush = 1;
3782
	}
J
Josef Bacik 已提交
3783

3784
	spin_unlock(&space_info->lock);
J
Josef Bacik 已提交
3785

3786
	if (!ret || !flush)
3787
		goto out;
3788

3789 3790 3791 3792
	/*
	 * We do synchronous shrinking since we don't actually unreserve
	 * metadata until after the IO is completed.
	 */
3793
	ret = shrink_delalloc(root, num_bytes, wait_ordered);
3794
	if (ret < 0)
3795
		goto out;
3796

3797 3798
	ret = 0;

3799 3800 3801 3802 3803 3804
	/*
	 * So if we were overcommitted it's possible that somebody else flushed
	 * out enough space and we simply didn't have enough space to reclaim,
	 * so go back around and try again.
	 */
	if (retries < 2) {
3805
		wait_ordered = true;
3806 3807 3808
		retries++;
		goto again;
	}
3809

3810
	ret = -ENOSPC;
3811 3812 3813
	if (committed)
		goto out;

3814
	ret = may_commit_transaction(root, space_info, orig_bytes, 0);
3815 3816
	if (!ret) {
		committed = true;
3817
		goto again;
3818
	}
3819 3820

out:
3821
	if (flushing) {
3822
		spin_lock(&space_info->lock);
3823 3824
		space_info->flush = 0;
		wake_up_all(&space_info->wait);
3825
		spin_unlock(&space_info->lock);
3826 3827 3828 3829 3830 3831 3832
	}
	return ret;
}

static struct btrfs_block_rsv *get_block_rsv(struct btrfs_trans_handle *trans,
					     struct btrfs_root *root)
{
3833 3834 3835
	struct btrfs_block_rsv *block_rsv = NULL;

	if (root->ref_cows || root == root->fs_info->csum_root)
3836
		block_rsv = trans->block_rsv;
3837 3838

	if (!block_rsv)
3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873
		block_rsv = root->block_rsv;

	if (!block_rsv)
		block_rsv = &root->fs_info->empty_block_rsv;

	return block_rsv;
}

static int block_rsv_use_bytes(struct btrfs_block_rsv *block_rsv,
			       u64 num_bytes)
{
	int ret = -ENOSPC;
	spin_lock(&block_rsv->lock);
	if (block_rsv->reserved >= num_bytes) {
		block_rsv->reserved -= num_bytes;
		if (block_rsv->reserved < block_rsv->size)
			block_rsv->full = 0;
		ret = 0;
	}
	spin_unlock(&block_rsv->lock);
	return ret;
}

static void block_rsv_add_bytes(struct btrfs_block_rsv *block_rsv,
				u64 num_bytes, int update_size)
{
	spin_lock(&block_rsv->lock);
	block_rsv->reserved += num_bytes;
	if (update_size)
		block_rsv->size += num_bytes;
	else if (block_rsv->reserved >= block_rsv->size)
		block_rsv->full = 1;
	spin_unlock(&block_rsv->lock);
}

J
Josef Bacik 已提交
3874 3875
static void block_rsv_release_bytes(struct btrfs_fs_info *fs_info,
				    struct btrfs_block_rsv *block_rsv,
3876
				    struct btrfs_block_rsv *dest, u64 num_bytes)
3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894
{
	struct btrfs_space_info *space_info = block_rsv->space_info;

	spin_lock(&block_rsv->lock);
	if (num_bytes == (u64)-1)
		num_bytes = block_rsv->size;
	block_rsv->size -= num_bytes;
	if (block_rsv->reserved >= block_rsv->size) {
		num_bytes = block_rsv->reserved - block_rsv->size;
		block_rsv->reserved = block_rsv->size;
		block_rsv->full = 1;
	} else {
		num_bytes = 0;
	}
	spin_unlock(&block_rsv->lock);

	if (num_bytes > 0) {
		if (dest) {
3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908
			spin_lock(&dest->lock);
			if (!dest->full) {
				u64 bytes_to_add;

				bytes_to_add = dest->size - dest->reserved;
				bytes_to_add = min(num_bytes, bytes_to_add);
				dest->reserved += bytes_to_add;
				if (dest->reserved >= dest->size)
					dest->full = 1;
				num_bytes -= bytes_to_add;
			}
			spin_unlock(&dest->lock);
		}
		if (num_bytes) {
3909
			spin_lock(&space_info->lock);
3910
			space_info->bytes_may_use -= num_bytes;
J
Josef Bacik 已提交
3911 3912 3913
			trace_btrfs_space_reservation(fs_info, "space_info",
						      (u64)space_info,
						      num_bytes, 0);
3914
			space_info->reservation_progress++;
3915
			spin_unlock(&space_info->lock);
3916
		}
J
Josef Bacik 已提交
3917
	}
3918
}
3919

3920 3921 3922 3923
static int block_rsv_migrate_bytes(struct btrfs_block_rsv *src,
				   struct btrfs_block_rsv *dst, u64 num_bytes)
{
	int ret;
J
Josef Bacik 已提交
3924

3925 3926 3927
	ret = block_rsv_use_bytes(src, num_bytes);
	if (ret)
		return ret;
J
Josef Bacik 已提交
3928

3929
	block_rsv_add_bytes(dst, num_bytes, 1);
J
Josef Bacik 已提交
3930 3931 3932
	return 0;
}

3933
void btrfs_init_block_rsv(struct btrfs_block_rsv *rsv)
J
Josef Bacik 已提交
3934
{
3935 3936 3937 3938 3939 3940 3941 3942
	memset(rsv, 0, sizeof(*rsv));
	spin_lock_init(&rsv->lock);
}

struct btrfs_block_rsv *btrfs_alloc_block_rsv(struct btrfs_root *root)
{
	struct btrfs_block_rsv *block_rsv;
	struct btrfs_fs_info *fs_info = root->fs_info;
J
Josef Bacik 已提交
3943

3944 3945 3946
	block_rsv = kmalloc(sizeof(*block_rsv), GFP_NOFS);
	if (!block_rsv)
		return NULL;
J
Josef Bacik 已提交
3947

3948 3949 3950 3951 3952
	btrfs_init_block_rsv(block_rsv);
	block_rsv->space_info = __find_space_info(fs_info,
						  BTRFS_BLOCK_GROUP_METADATA);
	return block_rsv;
}
J
Josef Bacik 已提交
3953

3954 3955 3956
void btrfs_free_block_rsv(struct btrfs_root *root,
			  struct btrfs_block_rsv *rsv)
{
3957 3958
	btrfs_block_rsv_release(root, rsv, (u64)-1);
	kfree(rsv);
J
Josef Bacik 已提交
3959 3960
}

3961 3962 3963
static inline int __block_rsv_add(struct btrfs_root *root,
				  struct btrfs_block_rsv *block_rsv,
				  u64 num_bytes, int flush)
J
Josef Bacik 已提交
3964
{
3965
	int ret;
J
Josef Bacik 已提交
3966

3967 3968
	if (num_bytes == 0)
		return 0;
3969

3970
	ret = reserve_metadata_bytes(root, block_rsv, num_bytes, flush);
3971 3972 3973 3974
	if (!ret) {
		block_rsv_add_bytes(block_rsv, num_bytes, 1);
		return 0;
	}
J
Josef Bacik 已提交
3975

3976 3977
	return ret;
}
J
Josef Bacik 已提交
3978

3979 3980 3981 3982 3983 3984 3985
int btrfs_block_rsv_add(struct btrfs_root *root,
			struct btrfs_block_rsv *block_rsv,
			u64 num_bytes)
{
	return __block_rsv_add(root, block_rsv, num_bytes, 1);
}

3986 3987 3988
int btrfs_block_rsv_add_noflush(struct btrfs_root *root,
				struct btrfs_block_rsv *block_rsv,
				u64 num_bytes)
3989
{
3990
	return __block_rsv_add(root, block_rsv, num_bytes, 0);
3991
}
J
Josef Bacik 已提交
3992

3993
int btrfs_block_rsv_check(struct btrfs_root *root,
3994
			  struct btrfs_block_rsv *block_rsv, int min_factor)
3995 3996 3997
{
	u64 num_bytes = 0;
	int ret = -ENOSPC;
J
Josef Bacik 已提交
3998

3999 4000
	if (!block_rsv)
		return 0;
J
Josef Bacik 已提交
4001

4002
	spin_lock(&block_rsv->lock);
4003 4004 4005 4006
	num_bytes = div_factor(block_rsv->size, min_factor);
	if (block_rsv->reserved >= num_bytes)
		ret = 0;
	spin_unlock(&block_rsv->lock);
J
Josef Bacik 已提交
4007

4008 4009 4010
	return ret;
}

4011 4012 4013
static inline int __btrfs_block_rsv_refill(struct btrfs_root *root,
					   struct btrfs_block_rsv *block_rsv,
					   u64 min_reserved, int flush)
4014 4015 4016 4017 4018 4019 4020 4021 4022
{
	u64 num_bytes = 0;
	int ret = -ENOSPC;

	if (!block_rsv)
		return 0;

	spin_lock(&block_rsv->lock);
	num_bytes = min_reserved;
4023
	if (block_rsv->reserved >= num_bytes)
4024
		ret = 0;
4025
	else
4026 4027
		num_bytes -= block_rsv->reserved;
	spin_unlock(&block_rsv->lock);
4028

4029 4030 4031
	if (!ret)
		return 0;

4032
	ret = reserve_metadata_bytes(root, block_rsv, num_bytes, flush);
4033 4034
	if (!ret) {
		block_rsv_add_bytes(block_rsv, num_bytes, 0);
4035
		return 0;
J
Josef Bacik 已提交
4036
	}
J
Josef Bacik 已提交
4037

4038
	return ret;
4039 4040
}

4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054
int btrfs_block_rsv_refill(struct btrfs_root *root,
			   struct btrfs_block_rsv *block_rsv,
			   u64 min_reserved)
{
	return __btrfs_block_rsv_refill(root, block_rsv, min_reserved, 1);
}

int btrfs_block_rsv_refill_noflush(struct btrfs_root *root,
				   struct btrfs_block_rsv *block_rsv,
				   u64 min_reserved)
{
	return __btrfs_block_rsv_refill(root, block_rsv, min_reserved, 0);
}

4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069
int btrfs_block_rsv_migrate(struct btrfs_block_rsv *src_rsv,
			    struct btrfs_block_rsv *dst_rsv,
			    u64 num_bytes)
{
	return block_rsv_migrate_bytes(src_rsv, dst_rsv, num_bytes);
}

void btrfs_block_rsv_release(struct btrfs_root *root,
			     struct btrfs_block_rsv *block_rsv,
			     u64 num_bytes)
{
	struct btrfs_block_rsv *global_rsv = &root->fs_info->global_block_rsv;
	if (global_rsv->full || global_rsv == block_rsv ||
	    block_rsv->space_info != global_rsv->space_info)
		global_rsv = NULL;
J
Josef Bacik 已提交
4070 4071
	block_rsv_release_bytes(root->fs_info, block_rsv, global_rsv,
				num_bytes);
J
Josef Bacik 已提交
4072 4073 4074
}

/*
4075 4076 4077
 * helper to calculate size of global block reservation.
 * the desired value is sum of space used by extent tree,
 * checksum tree and root tree
J
Josef Bacik 已提交
4078
 */
4079
static u64 calc_global_metadata_size(struct btrfs_fs_info *fs_info)
J
Josef Bacik 已提交
4080
{
4081 4082 4083 4084
	struct btrfs_space_info *sinfo;
	u64 num_bytes;
	u64 meta_used;
	u64 data_used;
4085
	int csum_size = btrfs_super_csum_size(fs_info->super_copy);
J
Josef Bacik 已提交
4086

4087 4088 4089 4090
	sinfo = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_DATA);
	spin_lock(&sinfo->lock);
	data_used = sinfo->bytes_used;
	spin_unlock(&sinfo->lock);
C
Chris Mason 已提交
4091

4092 4093
	sinfo = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_METADATA);
	spin_lock(&sinfo->lock);
4094 4095
	if (sinfo->flags & BTRFS_BLOCK_GROUP_DATA)
		data_used = 0;
4096 4097
	meta_used = sinfo->bytes_used;
	spin_unlock(&sinfo->lock);
4098

4099 4100 4101
	num_bytes = (data_used >> fs_info->sb->s_blocksize_bits) *
		    csum_size * 2;
	num_bytes += div64_u64(data_used + meta_used, 50);
4102

4103 4104
	if (num_bytes * 3 > meta_used)
		num_bytes = div64_u64(meta_used, 3);
4105

4106 4107
	return ALIGN(num_bytes, fs_info->extent_root->leafsize << 10);
}
J
Josef Bacik 已提交
4108

4109 4110 4111 4112 4113
static void update_global_block_rsv(struct btrfs_fs_info *fs_info)
{
	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
	struct btrfs_space_info *sinfo = block_rsv->space_info;
	u64 num_bytes;
J
Josef Bacik 已提交
4114

4115
	num_bytes = calc_global_metadata_size(fs_info);
C
Chris Mason 已提交
4116

4117 4118
	spin_lock(&block_rsv->lock);
	spin_lock(&sinfo->lock);
4119

4120
	block_rsv->size = num_bytes;
4121

4122
	num_bytes = sinfo->bytes_used + sinfo->bytes_pinned +
4123 4124
		    sinfo->bytes_reserved + sinfo->bytes_readonly +
		    sinfo->bytes_may_use;
4125 4126 4127 4128

	if (sinfo->total_bytes > num_bytes) {
		num_bytes = sinfo->total_bytes - num_bytes;
		block_rsv->reserved += num_bytes;
4129
		sinfo->bytes_may_use += num_bytes;
J
Josef Bacik 已提交
4130 4131
		trace_btrfs_space_reservation(fs_info, "space_info",
					      (u64)sinfo, num_bytes, 1);
J
Josef Bacik 已提交
4132 4133
	}

4134 4135
	if (block_rsv->reserved >= block_rsv->size) {
		num_bytes = block_rsv->reserved - block_rsv->size;
4136
		sinfo->bytes_may_use -= num_bytes;
J
Josef Bacik 已提交
4137 4138
		trace_btrfs_space_reservation(fs_info, "space_info",
					      (u64)sinfo, num_bytes, 0);
4139
		sinfo->reservation_progress++;
4140 4141 4142
		block_rsv->reserved = block_rsv->size;
		block_rsv->full = 1;
	}
4143

4144 4145
	spin_unlock(&sinfo->lock);
	spin_unlock(&block_rsv->lock);
J
Josef Bacik 已提交
4146 4147
}

4148
static void init_global_block_rsv(struct btrfs_fs_info *fs_info)
J
Josef Bacik 已提交
4149
{
4150
	struct btrfs_space_info *space_info;
J
Josef Bacik 已提交
4151

4152 4153
	space_info = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
	fs_info->chunk_block_rsv.space_info = space_info;
J
Josef Bacik 已提交
4154

4155
	space_info = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_METADATA);
4156 4157
	fs_info->global_block_rsv.space_info = space_info;
	fs_info->delalloc_block_rsv.space_info = space_info;
4158 4159
	fs_info->trans_block_rsv.space_info = space_info;
	fs_info->empty_block_rsv.space_info = space_info;
4160
	fs_info->delayed_block_rsv.space_info = space_info;
4161

4162 4163 4164 4165
	fs_info->extent_root->block_rsv = &fs_info->global_block_rsv;
	fs_info->csum_root->block_rsv = &fs_info->global_block_rsv;
	fs_info->dev_root->block_rsv = &fs_info->global_block_rsv;
	fs_info->tree_root->block_rsv = &fs_info->global_block_rsv;
4166
	fs_info->chunk_root->block_rsv = &fs_info->chunk_block_rsv;
4167 4168

	update_global_block_rsv(fs_info);
J
Josef Bacik 已提交
4169 4170
}

4171
static void release_global_block_rsv(struct btrfs_fs_info *fs_info)
J
Josef Bacik 已提交
4172
{
J
Josef Bacik 已提交
4173 4174
	block_rsv_release_bytes(fs_info, &fs_info->global_block_rsv, NULL,
				(u64)-1);
4175 4176 4177 4178 4179 4180
	WARN_ON(fs_info->delalloc_block_rsv.size > 0);
	WARN_ON(fs_info->delalloc_block_rsv.reserved > 0);
	WARN_ON(fs_info->trans_block_rsv.size > 0);
	WARN_ON(fs_info->trans_block_rsv.reserved > 0);
	WARN_ON(fs_info->chunk_block_rsv.size > 0);
	WARN_ON(fs_info->chunk_block_rsv.reserved > 0);
4181 4182
	WARN_ON(fs_info->delayed_block_rsv.size > 0);
	WARN_ON(fs_info->delayed_block_rsv.reserved > 0);
4183 4184
}

4185 4186
void btrfs_trans_release_metadata(struct btrfs_trans_handle *trans,
				  struct btrfs_root *root)
J
Josef Bacik 已提交
4187
{
4188 4189
	if (!trans->bytes_reserved)
		return;
J
Josef Bacik 已提交
4190

J
Josef Bacik 已提交
4191 4192
	trace_btrfs_space_reservation(root->fs_info, "transaction", (u64)trans,
				      trans->bytes_reserved, 0);
4193
	btrfs_block_rsv_release(root, trans->block_rsv, trans->bytes_reserved);
4194 4195
	trans->bytes_reserved = 0;
}
J
Josef Bacik 已提交
4196

4197 4198 4199 4200 4201 4202 4203 4204
int btrfs_orphan_reserve_metadata(struct btrfs_trans_handle *trans,
				  struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_block_rsv *src_rsv = get_block_rsv(trans, root);
	struct btrfs_block_rsv *dst_rsv = root->orphan_block_rsv;

	/*
4205 4206 4207
	 * We need to hold space in order to delete our orphan item once we've
	 * added it, so this takes the reservation so we can release it later
	 * when we are truly done with the orphan item.
4208
	 */
C
Chris Mason 已提交
4209
	u64 num_bytes = btrfs_calc_trans_metadata_size(root, 1);
J
Josef Bacik 已提交
4210 4211
	trace_btrfs_space_reservation(root->fs_info, "orphan",
				      btrfs_ino(inode), num_bytes, 1);
4212
	return block_rsv_migrate_bytes(src_rsv, dst_rsv, num_bytes);
J
Josef Bacik 已提交
4213 4214
}

4215
void btrfs_orphan_release_metadata(struct inode *inode)
4216
{
4217
	struct btrfs_root *root = BTRFS_I(inode)->root;
C
Chris Mason 已提交
4218
	u64 num_bytes = btrfs_calc_trans_metadata_size(root, 1);
J
Josef Bacik 已提交
4219 4220
	trace_btrfs_space_reservation(root->fs_info, "orphan",
				      btrfs_ino(inode), num_bytes, 0);
4221 4222
	btrfs_block_rsv_release(root, root->orphan_block_rsv, num_bytes);
}
4223

4224 4225 4226 4227 4228 4229 4230 4231 4232 4233
int btrfs_snap_reserve_metadata(struct btrfs_trans_handle *trans,
				struct btrfs_pending_snapshot *pending)
{
	struct btrfs_root *root = pending->root;
	struct btrfs_block_rsv *src_rsv = get_block_rsv(trans, root);
	struct btrfs_block_rsv *dst_rsv = &pending->block_rsv;
	/*
	 * two for root back/forward refs, two for directory entries
	 * and one for root of the snapshot.
	 */
4234
	u64 num_bytes = btrfs_calc_trans_metadata_size(root, 5);
4235 4236
	dst_rsv->space_info = src_rsv->space_info;
	return block_rsv_migrate_bytes(src_rsv, dst_rsv, num_bytes);
4237 4238
}

4239 4240 4241 4242 4243 4244 4245 4246 4247
/**
 * drop_outstanding_extent - drop an outstanding extent
 * @inode: the inode we're dropping the extent for
 *
 * This is called when we are freeing up an outstanding extent, either called
 * after an error or after an extent is written.  This will return the number of
 * reserved extents that need to be freed.  This must be called with
 * BTRFS_I(inode)->lock held.
 */
4248 4249
static unsigned drop_outstanding_extent(struct inode *inode)
{
4250
	unsigned drop_inode_space = 0;
4251 4252 4253 4254 4255
	unsigned dropped_extents = 0;

	BUG_ON(!BTRFS_I(inode)->outstanding_extents);
	BTRFS_I(inode)->outstanding_extents--;

4256 4257 4258 4259 4260 4261
	if (BTRFS_I(inode)->outstanding_extents == 0 &&
	    BTRFS_I(inode)->delalloc_meta_reserved) {
		drop_inode_space = 1;
		BTRFS_I(inode)->delalloc_meta_reserved = 0;
	}

4262 4263 4264 4265 4266 4267
	/*
	 * If we have more or the same amount of outsanding extents than we have
	 * reserved then we need to leave the reserved extents count alone.
	 */
	if (BTRFS_I(inode)->outstanding_extents >=
	    BTRFS_I(inode)->reserved_extents)
4268
		return drop_inode_space;
4269 4270 4271 4272

	dropped_extents = BTRFS_I(inode)->reserved_extents -
		BTRFS_I(inode)->outstanding_extents;
	BTRFS_I(inode)->reserved_extents -= dropped_extents;
4273
	return dropped_extents + drop_inode_space;
4274 4275
}

4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295
/**
 * calc_csum_metadata_size - return the amount of metada space that must be
 *	reserved/free'd for the given bytes.
 * @inode: the inode we're manipulating
 * @num_bytes: the number of bytes in question
 * @reserve: 1 if we are reserving space, 0 if we are freeing space
 *
 * This adjusts the number of csum_bytes in the inode and then returns the
 * correct amount of metadata that must either be reserved or freed.  We
 * calculate how many checksums we can fit into one leaf and then divide the
 * number of bytes that will need to be checksumed by this value to figure out
 * how many checksums will be required.  If we are adding bytes then the number
 * may go up and we will return the number of additional bytes that must be
 * reserved.  If it is going down we will return the number of bytes that must
 * be freed.
 *
 * This must be called with BTRFS_I(inode)->lock held.
 */
static u64 calc_csum_metadata_size(struct inode *inode, u64 num_bytes,
				   int reserve)
4296
{
4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u64 csum_size;
	int num_csums_per_leaf;
	int num_csums;
	int old_csums;

	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM &&
	    BTRFS_I(inode)->csum_bytes == 0)
		return 0;

	old_csums = (int)div64_u64(BTRFS_I(inode)->csum_bytes, root->sectorsize);
	if (reserve)
		BTRFS_I(inode)->csum_bytes += num_bytes;
	else
		BTRFS_I(inode)->csum_bytes -= num_bytes;
	csum_size = BTRFS_LEAF_DATA_SIZE(root) - sizeof(struct btrfs_item);
	num_csums_per_leaf = (int)div64_u64(csum_size,
					    sizeof(struct btrfs_csum_item) +
					    sizeof(struct btrfs_disk_key));
	num_csums = (int)div64_u64(BTRFS_I(inode)->csum_bytes, root->sectorsize);
	num_csums = num_csums + num_csums_per_leaf - 1;
	num_csums = num_csums / num_csums_per_leaf;

	old_csums = old_csums + num_csums_per_leaf - 1;
	old_csums = old_csums / num_csums_per_leaf;

	/* No change, no need to reserve more */
	if (old_csums == num_csums)
		return 0;

	if (reserve)
		return btrfs_calc_trans_metadata_size(root,
						      num_csums - old_csums);

	return btrfs_calc_trans_metadata_size(root, old_csums - num_csums);
4332
}
Y
Yan Zheng 已提交
4333

4334 4335 4336 4337
int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_block_rsv *block_rsv = &root->fs_info->delalloc_block_rsv;
4338
	u64 to_reserve = 0;
4339
	u64 csum_bytes;
4340
	unsigned nr_extents = 0;
4341
	int extra_reserve = 0;
4342
	int flush = 1;
4343
	int ret;
4344

4345
	/* Need to be holding the i_mutex here if we aren't free space cache */
4346 4347
	if (btrfs_is_free_space_inode(root, inode))
		flush = 0;
4348 4349
	else
		WARN_ON(!mutex_is_locked(&inode->i_mutex));
4350 4351

	if (flush && btrfs_transaction_in_commit(root->fs_info))
4352
		schedule_timeout(1);
4353

4354
	num_bytes = ALIGN(num_bytes, root->sectorsize);
4355

4356 4357 4358 4359
	spin_lock(&BTRFS_I(inode)->lock);
	BTRFS_I(inode)->outstanding_extents++;

	if (BTRFS_I(inode)->outstanding_extents >
4360
	    BTRFS_I(inode)->reserved_extents)
4361 4362
		nr_extents = BTRFS_I(inode)->outstanding_extents -
			BTRFS_I(inode)->reserved_extents;
4363

4364 4365 4366 4367 4368 4369
	/*
	 * Add an item to reserve for updating the inode when we complete the
	 * delalloc io.
	 */
	if (!BTRFS_I(inode)->delalloc_meta_reserved) {
		nr_extents++;
4370
		extra_reserve = 1;
4371
	}
4372 4373

	to_reserve = btrfs_calc_trans_metadata_size(root, nr_extents);
4374
	to_reserve += calc_csum_metadata_size(inode, num_bytes, 1);
4375
	csum_bytes = BTRFS_I(inode)->csum_bytes;
4376
	spin_unlock(&BTRFS_I(inode)->lock);
4377

4378
	ret = reserve_metadata_bytes(root, block_rsv, to_reserve, flush);
4379
	if (ret) {
4380
		u64 to_free = 0;
4381
		unsigned dropped;
4382

4383
		spin_lock(&BTRFS_I(inode)->lock);
4384 4385
		dropped = drop_outstanding_extent(inode);
		/*
4386 4387 4388 4389 4390 4391
		 * If the inodes csum_bytes is the same as the original
		 * csum_bytes then we know we haven't raced with any free()ers
		 * so we can just reduce our inodes csum bytes and carry on.
		 * Otherwise we have to do the normal free thing to account for
		 * the case that the free side didn't free up its reserve
		 * because of this outstanding reservation.
4392
		 */
4393 4394 4395 4396 4397 4398 4399 4400
		if (BTRFS_I(inode)->csum_bytes == csum_bytes)
			calc_csum_metadata_size(inode, num_bytes, 0);
		else
			to_free = calc_csum_metadata_size(inode, num_bytes, 0);
		spin_unlock(&BTRFS_I(inode)->lock);
		if (dropped)
			to_free += btrfs_calc_trans_metadata_size(root, dropped);

J
Josef Bacik 已提交
4401
		if (to_free) {
4402
			btrfs_block_rsv_release(root, block_rsv, to_free);
J
Josef Bacik 已提交
4403 4404 4405 4406 4407
			trace_btrfs_space_reservation(root->fs_info,
						      "delalloc",
						      btrfs_ino(inode),
						      to_free, 0);
		}
4408
		return ret;
4409
	}
4410

4411 4412 4413 4414 4415 4416 4417 4418
	spin_lock(&BTRFS_I(inode)->lock);
	if (extra_reserve) {
		BTRFS_I(inode)->delalloc_meta_reserved = 1;
		nr_extents--;
	}
	BTRFS_I(inode)->reserved_extents += nr_extents;
	spin_unlock(&BTRFS_I(inode)->lock);

J
Josef Bacik 已提交
4419 4420 4421
	if (to_reserve)
		trace_btrfs_space_reservation(root->fs_info,"delalloc",
					      btrfs_ino(inode), to_reserve, 1);
4422 4423 4424 4425 4426
	block_rsv_add_bytes(block_rsv, to_reserve, 1);

	return 0;
}

4427 4428 4429 4430 4431 4432 4433 4434 4435
/**
 * btrfs_delalloc_release_metadata - release a metadata reservation for an inode
 * @inode: the inode to release the reservation for
 * @num_bytes: the number of bytes we're releasing
 *
 * This will release the metadata reservation for an inode.  This can be called
 * once we complete IO for a given set of bytes to release their metadata
 * reservations.
 */
4436 4437 4438
void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
4439 4440
	u64 to_free = 0;
	unsigned dropped;
4441 4442

	num_bytes = ALIGN(num_bytes, root->sectorsize);
4443
	spin_lock(&BTRFS_I(inode)->lock);
4444
	dropped = drop_outstanding_extent(inode);
4445

4446 4447
	to_free = calc_csum_metadata_size(inode, num_bytes, 0);
	spin_unlock(&BTRFS_I(inode)->lock);
4448 4449
	if (dropped > 0)
		to_free += btrfs_calc_trans_metadata_size(root, dropped);
4450

J
Josef Bacik 已提交
4451 4452
	trace_btrfs_space_reservation(root->fs_info, "delalloc",
				      btrfs_ino(inode), to_free, 0);
4453 4454 4455 4456
	btrfs_block_rsv_release(root, &root->fs_info->delalloc_block_rsv,
				to_free);
}

4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471
/**
 * btrfs_delalloc_reserve_space - reserve data and metadata space for delalloc
 * @inode: inode we're writing to
 * @num_bytes: the number of bytes we want to allocate
 *
 * This will do the following things
 *
 * o reserve space in the data space info for num_bytes
 * o reserve space in the metadata space info based on number of outstanding
 *   extents and how much csums will be needed
 * o add to the inodes ->delalloc_bytes
 * o add it to the fs_info's delalloc inodes list.
 *
 * This will return 0 for success and -ENOSPC if there is no space left.
 */
4472 4473 4474 4475 4476
int btrfs_delalloc_reserve_space(struct inode *inode, u64 num_bytes)
{
	int ret;

	ret = btrfs_check_data_free_space(inode, num_bytes);
C
Chris Mason 已提交
4477
	if (ret)
4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488
		return ret;

	ret = btrfs_delalloc_reserve_metadata(inode, num_bytes);
	if (ret) {
		btrfs_free_reserved_data_space(inode, num_bytes);
		return ret;
	}

	return 0;
}

4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501
/**
 * btrfs_delalloc_release_space - release data and metadata space for delalloc
 * @inode: inode we're releasing space for
 * @num_bytes: the number of bytes we want to free up
 *
 * This must be matched with a call to btrfs_delalloc_reserve_space.  This is
 * called in the case that we don't need the metadata AND data reservations
 * anymore.  So if there is an error or we insert an inline extent.
 *
 * This function will release the metadata space that was not used and will
 * decrement ->delalloc_bytes and remove it from the fs_info delalloc_inodes
 * list if there are no delalloc bytes left.
 */
4502 4503 4504 4505
void btrfs_delalloc_release_space(struct inode *inode, u64 num_bytes)
{
	btrfs_delalloc_release_metadata(inode, num_bytes);
	btrfs_free_reserved_data_space(inode, num_bytes);
4506 4507
}

C
Chris Mason 已提交
4508 4509
static int update_block_group(struct btrfs_trans_handle *trans,
			      struct btrfs_root *root,
4510
			      u64 bytenr, u64 num_bytes, int alloc)
C
Chris Mason 已提交
4511
{
4512
	struct btrfs_block_group_cache *cache = NULL;
C
Chris Mason 已提交
4513
	struct btrfs_fs_info *info = root->fs_info;
4514
	u64 total = num_bytes;
C
Chris Mason 已提交
4515
	u64 old_val;
4516
	u64 byte_in_group;
4517
	int factor;
C
Chris Mason 已提交
4518

4519 4520
	/* block accounting for super block */
	spin_lock(&info->delalloc_lock);
4521
	old_val = btrfs_super_bytes_used(info->super_copy);
4522 4523 4524 4525
	if (alloc)
		old_val += num_bytes;
	else
		old_val -= num_bytes;
4526
	btrfs_set_super_bytes_used(info->super_copy, old_val);
4527 4528
	spin_unlock(&info->delalloc_lock);

C
Chris Mason 已提交
4529
	while (total) {
4530
		cache = btrfs_lookup_block_group(info, bytenr);
4531
		if (!cache)
C
Chris Mason 已提交
4532
			return -1;
4533 4534 4535 4536 4537 4538
		if (cache->flags & (BTRFS_BLOCK_GROUP_DUP |
				    BTRFS_BLOCK_GROUP_RAID1 |
				    BTRFS_BLOCK_GROUP_RAID10))
			factor = 2;
		else
			factor = 1;
4539 4540 4541 4542 4543 4544 4545
		/*
		 * If this block group has free space cache written out, we
		 * need to make sure to load it if we are removing space.  This
		 * is because we need the unpinning stage to actually add the
		 * space back to the block group, otherwise we will leak space.
		 */
		if (!alloc && cache->cached == BTRFS_CACHE_NO)
4546
			cache_block_group(cache, trans, NULL, 1);
4547

4548 4549
		byte_in_group = bytenr - cache->key.objectid;
		WARN_ON(byte_in_group > cache->key.offset);
C
Chris Mason 已提交
4550

4551
		spin_lock(&cache->space_info->lock);
4552
		spin_lock(&cache->lock);
4553

4554
		if (btrfs_test_opt(root, SPACE_CACHE) &&
4555 4556 4557
		    cache->disk_cache_state < BTRFS_DC_CLEAR)
			cache->disk_cache_state = BTRFS_DC_CLEAR;

J
Josef Bacik 已提交
4558
		cache->dirty = 1;
C
Chris Mason 已提交
4559
		old_val = btrfs_block_group_used(&cache->item);
4560
		num_bytes = min(total, cache->key.offset - byte_in_group);
C
Chris Mason 已提交
4561
		if (alloc) {
4562
			old_val += num_bytes;
4563 4564 4565
			btrfs_set_block_group_used(&cache->item, old_val);
			cache->reserved -= num_bytes;
			cache->space_info->bytes_reserved -= num_bytes;
4566 4567
			cache->space_info->bytes_used += num_bytes;
			cache->space_info->disk_used += num_bytes * factor;
4568
			spin_unlock(&cache->lock);
4569
			spin_unlock(&cache->space_info->lock);
C
Chris Mason 已提交
4570
		} else {
4571
			old_val -= num_bytes;
4572
			btrfs_set_block_group_used(&cache->item, old_val);
4573 4574
			cache->pinned += num_bytes;
			cache->space_info->bytes_pinned += num_bytes;
4575
			cache->space_info->bytes_used -= num_bytes;
4576
			cache->space_info->disk_used -= num_bytes * factor;
4577
			spin_unlock(&cache->lock);
4578
			spin_unlock(&cache->space_info->lock);
4579

4580 4581 4582
			set_extent_dirty(info->pinned_extents,
					 bytenr, bytenr + num_bytes - 1,
					 GFP_NOFS | __GFP_NOFAIL);
C
Chris Mason 已提交
4583
		}
4584
		btrfs_put_block_group(cache);
4585 4586
		total -= num_bytes;
		bytenr += num_bytes;
C
Chris Mason 已提交
4587 4588 4589
	}
	return 0;
}
4590

4591 4592
static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
{
J
Josef Bacik 已提交
4593
	struct btrfs_block_group_cache *cache;
4594
	u64 bytenr;
J
Josef Bacik 已提交
4595 4596 4597

	cache = btrfs_lookup_first_block_group(root->fs_info, search_start);
	if (!cache)
4598
		return 0;
J
Josef Bacik 已提交
4599

4600
	bytenr = cache->key.objectid;
4601
	btrfs_put_block_group(cache);
4602 4603

	return bytenr;
4604 4605
}

4606 4607 4608
static int pin_down_extent(struct btrfs_root *root,
			   struct btrfs_block_group_cache *cache,
			   u64 bytenr, u64 num_bytes, int reserved)
4609
{
4610 4611 4612 4613 4614 4615 4616 4617 4618 4619
	spin_lock(&cache->space_info->lock);
	spin_lock(&cache->lock);
	cache->pinned += num_bytes;
	cache->space_info->bytes_pinned += num_bytes;
	if (reserved) {
		cache->reserved -= num_bytes;
		cache->space_info->bytes_reserved -= num_bytes;
	}
	spin_unlock(&cache->lock);
	spin_unlock(&cache->space_info->lock);
J
Josef Bacik 已提交
4620

4621 4622 4623 4624
	set_extent_dirty(root->fs_info->pinned_extents, bytenr,
			 bytenr + num_bytes - 1, GFP_NOFS | __GFP_NOFAIL);
	return 0;
}
J
Josef Bacik 已提交
4625

4626 4627 4628 4629 4630 4631 4632
/*
 * this function must be called within transaction
 */
int btrfs_pin_extent(struct btrfs_root *root,
		     u64 bytenr, u64 num_bytes, int reserved)
{
	struct btrfs_block_group_cache *cache;
J
Josef Bacik 已提交
4633

4634 4635 4636 4637 4638 4639
	cache = btrfs_lookup_block_group(root->fs_info, bytenr);
	BUG_ON(!cache);

	pin_down_extent(root, cache, bytenr, num_bytes, reserved);

	btrfs_put_block_group(cache);
4640 4641 4642
	return 0;
}

4643
/*
4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670
 * this function must be called within transaction
 */
int btrfs_pin_extent_for_log_replay(struct btrfs_trans_handle *trans,
				    struct btrfs_root *root,
				    u64 bytenr, u64 num_bytes)
{
	struct btrfs_block_group_cache *cache;

	cache = btrfs_lookup_block_group(root->fs_info, bytenr);
	BUG_ON(!cache);

	/*
	 * pull in the free space cache (if any) so that our pin
	 * removes the free space from the cache.  We have load_only set
	 * to one because the slow code to read in the free extents does check
	 * the pinned extents.
	 */
	cache_block_group(cache, trans, root, 1);

	pin_down_extent(root, cache, bytenr, num_bytes, 0);

	/* remove us from the free space cache (if we're there at all) */
	btrfs_remove_free_space(cache, bytenr, num_bytes);
	btrfs_put_block_group(cache);
	return 0;
}

4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691
/**
 * btrfs_update_reserved_bytes - update the block_group and space info counters
 * @cache:	The cache we are manipulating
 * @num_bytes:	The number of bytes in question
 * @reserve:	One of the reservation enums
 *
 * This is called by the allocator when it reserves space, or by somebody who is
 * freeing space that was never actually used on disk.  For example if you
 * reserve some space for a new leaf in transaction A and before transaction A
 * commits you free that leaf, you call this with reserve set to 0 in order to
 * clear the reservation.
 *
 * Metadata reservations should be called with RESERVE_ALLOC so we do the proper
 * ENOSPC accounting.  For data we handle the reservation through clearing the
 * delalloc bits in the io_tree.  We have to do this since we could end up
 * allocating less disk space for the amount of data we have reserved in the
 * case of compression.
 *
 * If this is a reservation and the block group has become read only we cannot
 * make the reservation and return -EAGAIN, otherwise this function always
 * succeeds.
4692
 */
4693 4694
static int btrfs_update_reserved_bytes(struct btrfs_block_group_cache *cache,
				       u64 num_bytes, int reserve)
4695
{
4696
	struct btrfs_space_info *space_info = cache->space_info;
4697
	int ret = 0;
4698 4699 4700
	spin_lock(&space_info->lock);
	spin_lock(&cache->lock);
	if (reserve != RESERVE_FREE) {
4701 4702 4703
		if (cache->ro) {
			ret = -EAGAIN;
		} else {
4704 4705 4706
			cache->reserved += num_bytes;
			space_info->bytes_reserved += num_bytes;
			if (reserve == RESERVE_ALLOC) {
J
Josef Bacik 已提交
4707 4708 4709 4710
				trace_btrfs_space_reservation(cache->fs_info,
							      "space_info",
							      (u64)space_info,
							      num_bytes, 0);
4711 4712
				space_info->bytes_may_use -= num_bytes;
			}
4713
		}
4714 4715 4716 4717 4718 4719
	} else {
		if (cache->ro)
			space_info->bytes_readonly += num_bytes;
		cache->reserved -= num_bytes;
		space_info->bytes_reserved -= num_bytes;
		space_info->reservation_progress++;
4720
	}
4721 4722
	spin_unlock(&cache->lock);
	spin_unlock(&space_info->lock);
4723
	return ret;
4724
}
C
Chris Mason 已提交
4725

4726 4727
int btrfs_prepare_extent_commit(struct btrfs_trans_handle *trans,
				struct btrfs_root *root)
4728 4729
{
	struct btrfs_fs_info *fs_info = root->fs_info;
4730 4731 4732
	struct btrfs_caching_control *next;
	struct btrfs_caching_control *caching_ctl;
	struct btrfs_block_group_cache *cache;
4733

4734
	down_write(&fs_info->extent_commit_sem);
4735

4736 4737 4738 4739 4740 4741 4742
	list_for_each_entry_safe(caching_ctl, next,
				 &fs_info->caching_block_groups, list) {
		cache = caching_ctl->block_group;
		if (block_group_cache_done(cache)) {
			cache->last_byte_to_unpin = (u64)-1;
			list_del_init(&caching_ctl->list);
			put_caching_control(caching_ctl);
4743
		} else {
4744
			cache->last_byte_to_unpin = caching_ctl->progress;
4745 4746
		}
	}
4747 4748 4749 4750 4751 4752 4753

	if (fs_info->pinned_extents == &fs_info->freed_extents[0])
		fs_info->pinned_extents = &fs_info->freed_extents[1];
	else
		fs_info->pinned_extents = &fs_info->freed_extents[0];

	up_write(&fs_info->extent_commit_sem);
4754 4755

	update_global_block_rsv(fs_info);
4756 4757 4758
	return 0;
}

4759
static int unpin_extent_range(struct btrfs_root *root, u64 start, u64 end)
C
Chris Mason 已提交
4760
{
4761 4762 4763
	struct btrfs_fs_info *fs_info = root->fs_info;
	struct btrfs_block_group_cache *cache = NULL;
	u64 len;
C
Chris Mason 已提交
4764

4765 4766 4767 4768 4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779 4780 4781
	while (start <= end) {
		if (!cache ||
		    start >= cache->key.objectid + cache->key.offset) {
			if (cache)
				btrfs_put_block_group(cache);
			cache = btrfs_lookup_block_group(fs_info, start);
			BUG_ON(!cache);
		}

		len = cache->key.objectid + cache->key.offset - start;
		len = min(len, end + 1 - start);

		if (start < cache->last_byte_to_unpin) {
			len = min(len, cache->last_byte_to_unpin - start);
			btrfs_add_free_space(cache, start, len);
		}

4782 4783
		start += len;

4784 4785 4786 4787
		spin_lock(&cache->space_info->lock);
		spin_lock(&cache->lock);
		cache->pinned -= len;
		cache->space_info->bytes_pinned -= len;
4788
		if (cache->ro)
4789
			cache->space_info->bytes_readonly += len;
4790 4791
		spin_unlock(&cache->lock);
		spin_unlock(&cache->space_info->lock);
C
Chris Mason 已提交
4792
	}
4793 4794 4795

	if (cache)
		btrfs_put_block_group(cache);
C
Chris Mason 已提交
4796 4797 4798 4799
	return 0;
}

int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
4800
			       struct btrfs_root *root)
4801
{
4802 4803
	struct btrfs_fs_info *fs_info = root->fs_info;
	struct extent_io_tree *unpin;
4804 4805
	u64 start;
	u64 end;
4806 4807
	int ret;

4808 4809 4810 4811 4812
	if (fs_info->pinned_extents == &fs_info->freed_extents[0])
		unpin = &fs_info->freed_extents[1];
	else
		unpin = &fs_info->freed_extents[0];

C
Chris Mason 已提交
4813
	while (1) {
4814 4815 4816
		ret = find_first_extent_bit(unpin, 0, &start, &end,
					    EXTENT_DIRTY);
		if (ret)
4817
			break;
4818

4819 4820 4821
		if (btrfs_test_opt(root, DISCARD))
			ret = btrfs_discard_extent(root, start,
						   end + 1 - start, NULL);
4822

4823
		clear_extent_dirty(unpin, start, end, GFP_NOFS);
4824
		unpin_extent_range(root, start, end);
4825
		cond_resched();
4826
	}
J
Josef Bacik 已提交
4827

C
Chris Mason 已提交
4828 4829 4830
	return 0;
}

4831 4832 4833 4834 4835 4836
static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
				struct btrfs_root *root,
				u64 bytenr, u64 num_bytes, u64 parent,
				u64 root_objectid, u64 owner_objectid,
				u64 owner_offset, int refs_to_drop,
				struct btrfs_delayed_extent_op *extent_op)
4837
{
C
Chris Mason 已提交
4838
	struct btrfs_key key;
4839
	struct btrfs_path *path;
4840 4841
	struct btrfs_fs_info *info = root->fs_info;
	struct btrfs_root *extent_root = info->extent_root;
4842
	struct extent_buffer *leaf;
4843 4844
	struct btrfs_extent_item *ei;
	struct btrfs_extent_inline_ref *iref;
4845
	int ret;
4846
	int is_data;
4847 4848 4849
	int extent_slot = 0;
	int found_extent = 0;
	int num_to_del = 1;
4850 4851
	u32 item_size;
	u64 refs;
C
Chris Mason 已提交
4852

4853
	path = btrfs_alloc_path();
4854 4855
	if (!path)
		return -ENOMEM;
4856

4857
	path->reada = 1;
4858
	path->leave_spinning = 1;
4859 4860 4861 4862 4863 4864 4865 4866

	is_data = owner_objectid >= BTRFS_FIRST_FREE_OBJECTID;
	BUG_ON(!is_data && refs_to_drop != 1);

	ret = lookup_extent_backref(trans, extent_root, path, &iref,
				    bytenr, num_bytes, parent,
				    root_objectid, owner_objectid,
				    owner_offset);
4867
	if (ret == 0) {
4868
		extent_slot = path->slots[0];
4869 4870
		while (extent_slot >= 0) {
			btrfs_item_key_to_cpu(path->nodes[0], &key,
4871
					      extent_slot);
4872
			if (key.objectid != bytenr)
4873
				break;
4874 4875
			if (key.type == BTRFS_EXTENT_ITEM_KEY &&
			    key.offset == num_bytes) {
4876 4877 4878 4879 4880
				found_extent = 1;
				break;
			}
			if (path->slots[0] - extent_slot > 5)
				break;
4881
			extent_slot--;
4882
		}
4883 4884 4885 4886 4887
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
		item_size = btrfs_item_size_nr(path->nodes[0], extent_slot);
		if (found_extent && item_size < sizeof(*ei))
			found_extent = 0;
#endif
Z
Zheng Yan 已提交
4888
		if (!found_extent) {
4889
			BUG_ON(iref);
4890
			ret = remove_extent_backref(trans, extent_root, path,
4891 4892
						    NULL, refs_to_drop,
						    is_data);
Z
Zheng Yan 已提交
4893
			BUG_ON(ret);
4894
			btrfs_release_path(path);
4895
			path->leave_spinning = 1;
4896 4897 4898 4899 4900

			key.objectid = bytenr;
			key.type = BTRFS_EXTENT_ITEM_KEY;
			key.offset = num_bytes;

Z
Zheng Yan 已提交
4901 4902
			ret = btrfs_search_slot(trans, extent_root,
						&key, path, -1, 1);
4903 4904
			if (ret) {
				printk(KERN_ERR "umm, got %d back from search"
C
Chris Mason 已提交
4905 4906
				       ", was looking for %llu\n", ret,
				       (unsigned long long)bytenr);
4907 4908 4909
				if (ret > 0)
					btrfs_print_leaf(extent_root,
							 path->nodes[0]);
4910
			}
Z
Zheng Yan 已提交
4911 4912 4913
			BUG_ON(ret);
			extent_slot = path->slots[0];
		}
4914 4915 4916
	} else {
		btrfs_print_leaf(extent_root, path->nodes[0]);
		WARN_ON(1);
C
Chris Mason 已提交
4917
		printk(KERN_ERR "btrfs unable to find ref byte nr %llu "
4918
		       "parent %llu root %llu  owner %llu offset %llu\n",
C
Chris Mason 已提交
4919
		       (unsigned long long)bytenr,
4920
		       (unsigned long long)parent,
C
Chris Mason 已提交
4921
		       (unsigned long long)root_objectid,
4922 4923
		       (unsigned long long)owner_objectid,
		       (unsigned long long)owner_offset);
4924
	}
4925 4926

	leaf = path->nodes[0];
4927 4928 4929 4930 4931 4932 4933 4934
	item_size = btrfs_item_size_nr(leaf, extent_slot);
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
	if (item_size < sizeof(*ei)) {
		BUG_ON(found_extent || extent_slot != path->slots[0]);
		ret = convert_extent_item_v0(trans, extent_root, path,
					     owner_objectid, 0);
		BUG_ON(ret < 0);

4935
		btrfs_release_path(path);
4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953 4954 4955 4956
		path->leave_spinning = 1;

		key.objectid = bytenr;
		key.type = BTRFS_EXTENT_ITEM_KEY;
		key.offset = num_bytes;

		ret = btrfs_search_slot(trans, extent_root, &key, path,
					-1, 1);
		if (ret) {
			printk(KERN_ERR "umm, got %d back from search"
			       ", was looking for %llu\n", ret,
			       (unsigned long long)bytenr);
			btrfs_print_leaf(extent_root, path->nodes[0]);
		}
		BUG_ON(ret);
		extent_slot = path->slots[0];
		leaf = path->nodes[0];
		item_size = btrfs_item_size_nr(leaf, extent_slot);
	}
#endif
	BUG_ON(item_size < sizeof(*ei));
4957
	ei = btrfs_item_ptr(leaf, extent_slot,
C
Chris Mason 已提交
4958
			    struct btrfs_extent_item);
4959 4960 4961 4962 4963 4964
	if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID) {
		struct btrfs_tree_block_info *bi;
		BUG_ON(item_size < sizeof(*ei) + sizeof(*bi));
		bi = (struct btrfs_tree_block_info *)(ei + 1);
		WARN_ON(owner_objectid != btrfs_tree_block_level(leaf, bi));
	}
4965

4966
	refs = btrfs_extent_refs(leaf, ei);
4967 4968
	BUG_ON(refs < refs_to_drop);
	refs -= refs_to_drop;
4969

4970 4971 4972 4973 4974 4975
	if (refs > 0) {
		if (extent_op)
			__run_delayed_extent_op(extent_op, leaf, ei);
		/*
		 * In the case of inline back ref, reference count will
		 * be updated by remove_extent_backref
4976
		 */
4977 4978 4979 4980 4981 4982 4983 4984 4985 4986
		if (iref) {
			BUG_ON(!found_extent);
		} else {
			btrfs_set_extent_refs(leaf, ei, refs);
			btrfs_mark_buffer_dirty(leaf);
		}
		if (found_extent) {
			ret = remove_extent_backref(trans, extent_root, path,
						    iref, refs_to_drop,
						    is_data);
4987 4988
			BUG_ON(ret);
		}
4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999
	} else {
		if (found_extent) {
			BUG_ON(is_data && refs_to_drop !=
			       extent_data_ref_count(root, path, iref));
			if (iref) {
				BUG_ON(path->slots[0] != extent_slot);
			} else {
				BUG_ON(path->slots[0] != extent_slot + 1);
				path->slots[0] = extent_slot;
				num_to_del = 2;
			}
C
Chris Mason 已提交
5000
		}
5001

5002 5003
		ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
				      num_to_del);
Z
Zheng Yan 已提交
5004
		BUG_ON(ret);
5005
		btrfs_release_path(path);
5006

5007
		if (is_data) {
5008 5009
			ret = btrfs_del_csums(trans, root, bytenr, num_bytes);
			BUG_ON(ret);
5010 5011 5012 5013
		} else {
			invalidate_mapping_pages(info->btree_inode->i_mapping,
			     bytenr >> PAGE_CACHE_SHIFT,
			     (bytenr + num_bytes - 1) >> PAGE_CACHE_SHIFT);
5014 5015
		}

5016
		ret = update_block_group(trans, root, bytenr, num_bytes, 0);
5017
		BUG_ON(ret);
5018
	}
5019
	btrfs_free_path(path);
5020 5021 5022
	return ret;
}

5023
/*
5024
 * when we free an block, it is possible (and likely) that we free the last
5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035
 * delayed ref for that extent as well.  This searches the delayed ref tree for
 * a given extent, and if there are no other delayed refs to be processed, it
 * removes it from the tree.
 */
static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
				      struct btrfs_root *root, u64 bytenr)
{
	struct btrfs_delayed_ref_head *head;
	struct btrfs_delayed_ref_root *delayed_refs;
	struct btrfs_delayed_ref_node *ref;
	struct rb_node *node;
5036
	int ret = 0;
5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051 5052 5053

	delayed_refs = &trans->transaction->delayed_refs;
	spin_lock(&delayed_refs->lock);
	head = btrfs_find_delayed_ref_head(trans, bytenr);
	if (!head)
		goto out;

	node = rb_prev(&head->node.rb_node);
	if (!node)
		goto out;

	ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);

	/* there are still entries for this ref, we can't drop it */
	if (ref->bytenr == bytenr)
		goto out;

5054 5055 5056 5057 5058 5059 5060
	if (head->extent_op) {
		if (!head->must_insert_reserved)
			goto out;
		kfree(head->extent_op);
		head->extent_op = NULL;
	}

5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 5072 5073
	/*
	 * waiting for the lock here would deadlock.  If someone else has it
	 * locked they are already in the process of dropping it anyway
	 */
	if (!mutex_trylock(&head->mutex))
		goto out;

	/*
	 * at this point we have a head with no other entries.  Go
	 * ahead and process it.
	 */
	head->node.in_tree = 0;
	rb_erase(&head->node.rb_node, &delayed_refs->root);
5074

5075
	delayed_refs->num_entries--;
5076 5077
	if (waitqueue_active(&delayed_refs->seq_wait))
		wake_up(&delayed_refs->seq_wait);
5078 5079 5080 5081 5082

	/*
	 * we don't take a ref on the node because we're removing it from the
	 * tree, so we just steal the ref the tree was holding.
	 */
5083 5084 5085 5086 5087
	delayed_refs->num_heads--;
	if (list_empty(&head->cluster))
		delayed_refs->num_heads_ready--;

	list_del_init(&head->cluster);
5088 5089
	spin_unlock(&delayed_refs->lock);

5090 5091 5092 5093 5094
	BUG_ON(head->extent_op);
	if (head->must_insert_reserved)
		ret = 1;

	mutex_unlock(&head->mutex);
5095
	btrfs_put_delayed_ref(&head->node);
5096
	return ret;
5097 5098 5099 5100 5101
out:
	spin_unlock(&delayed_refs->lock);
	return 0;
}

5102 5103 5104
void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
			   struct btrfs_root *root,
			   struct extent_buffer *buf,
A
Arne Jansen 已提交
5105
			   u64 parent, int last_ref, int for_cow)
5106 5107 5108 5109 5110
{
	struct btrfs_block_group_cache *cache = NULL;
	int ret;

	if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
A
Arne Jansen 已提交
5111 5112 5113 5114 5115
		ret = btrfs_add_delayed_tree_ref(root->fs_info, trans,
					buf->start, buf->len,
					parent, root->root_key.objectid,
					btrfs_header_level(buf),
					BTRFS_DROP_DELAYED_REF, NULL, for_cow);
5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127
		BUG_ON(ret);
	}

	if (!last_ref)
		return;

	cache = btrfs_lookup_block_group(root->fs_info, buf->start);

	if (btrfs_header_generation(buf) == trans->transid) {
		if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
			ret = check_ref_cleanup(trans, root, buf->start);
			if (!ret)
5128
				goto out;
5129 5130 5131 5132
		}

		if (btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
			pin_down_extent(root, cache, buf->start, buf->len, 1);
5133
			goto out;
5134 5135 5136 5137 5138
		}

		WARN_ON(test_bit(EXTENT_BUFFER_DIRTY, &buf->bflags));

		btrfs_add_free_space(cache, buf->start, buf->len);
5139
		btrfs_update_reserved_bytes(cache, buf->len, RESERVE_FREE);
5140 5141
	}
out:
5142 5143 5144 5145 5146
	/*
	 * Deleting the buffer, clear the corrupt flag since it doesn't matter
	 * anymore.
	 */
	clear_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags);
5147 5148 5149
	btrfs_put_block_group(cache);
}

A
Arne Jansen 已提交
5150 5151 5152
int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root,
		      u64 bytenr, u64 num_bytes, u64 parent, u64 root_objectid,
		      u64 owner, u64 offset, int for_cow)
5153 5154
{
	int ret;
A
Arne Jansen 已提交
5155
	struct btrfs_fs_info *fs_info = root->fs_info;
5156

5157 5158 5159 5160
	/*
	 * tree log blocks never actually go into the extent allocation
	 * tree, just update pinning info and exit early.
	 */
5161 5162
	if (root_objectid == BTRFS_TREE_LOG_OBJECTID) {
		WARN_ON(owner >= BTRFS_FIRST_FREE_OBJECTID);
5163
		/* unlocks the pinned mutex */
5164
		btrfs_pin_extent(root, bytenr, num_bytes, 1);
5165
		ret = 0;
5166
	} else if (owner < BTRFS_FIRST_FREE_OBJECTID) {
A
Arne Jansen 已提交
5167 5168
		ret = btrfs_add_delayed_tree_ref(fs_info, trans, bytenr,
					num_bytes,
5169
					parent, root_objectid, (int)owner,
A
Arne Jansen 已提交
5170
					BTRFS_DROP_DELAYED_REF, NULL, for_cow);
5171
		BUG_ON(ret);
5172
	} else {
A
Arne Jansen 已提交
5173 5174 5175 5176 5177
		ret = btrfs_add_delayed_data_ref(fs_info, trans, bytenr,
						num_bytes,
						parent, root_objectid, owner,
						offset, BTRFS_DROP_DELAYED_REF,
						NULL, for_cow);
5178
		BUG_ON(ret);
5179
	}
5180 5181 5182
	return ret;
}

5183 5184 5185 5186 5187 5188 5189
static u64 stripe_align(struct btrfs_root *root, u64 val)
{
	u64 mask = ((u64)root->stripesize - 1);
	u64 ret = (val + mask) & ~mask;
	return ret;
}

J
Josef Bacik 已提交
5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204
/*
 * when we wait for progress in the block group caching, its because
 * our allocation attempt failed at least once.  So, we must sleep
 * and let some progress happen before we try again.
 *
 * This function will sleep at least once waiting for new free space to
 * show up, and then it will check the block group free space numbers
 * for our min num_bytes.  Another option is to have it go ahead
 * and look in the rbtree for a free extent of a given size, but this
 * is a good start.
 */
static noinline int
wait_block_group_cache_progress(struct btrfs_block_group_cache *cache,
				u64 num_bytes)
{
5205
	struct btrfs_caching_control *caching_ctl;
J
Josef Bacik 已提交
5206 5207
	DEFINE_WAIT(wait);

5208 5209
	caching_ctl = get_caching_control(cache);
	if (!caching_ctl)
J
Josef Bacik 已提交
5210 5211
		return 0;

5212
	wait_event(caching_ctl->wait, block_group_cache_done(cache) ||
5213
		   (cache->free_space_ctl->free_space >= num_bytes));
5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231

	put_caching_control(caching_ctl);
	return 0;
}

static noinline int
wait_block_group_cache_done(struct btrfs_block_group_cache *cache)
{
	struct btrfs_caching_control *caching_ctl;
	DEFINE_WAIT(wait);

	caching_ctl = get_caching_control(cache);
	if (!caching_ctl)
		return 0;

	wait_event(caching_ctl->wait, block_group_cache_done(cache));

	put_caching_control(caching_ctl);
J
Josef Bacik 已提交
5232 5233 5234
	return 0;
}

5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250
static int get_block_group_index(struct btrfs_block_group_cache *cache)
{
	int index;
	if (cache->flags & BTRFS_BLOCK_GROUP_RAID10)
		index = 0;
	else if (cache->flags & BTRFS_BLOCK_GROUP_RAID1)
		index = 1;
	else if (cache->flags & BTRFS_BLOCK_GROUP_DUP)
		index = 2;
	else if (cache->flags & BTRFS_BLOCK_GROUP_RAID0)
		index = 3;
	else
		index = 4;
	return index;
}

J
Josef Bacik 已提交
5251
enum btrfs_loop_type {
5252
	LOOP_FIND_IDEAL = 0,
J
Josef Bacik 已提交
5253 5254 5255 5256 5257 5258
	LOOP_CACHING_NOWAIT = 1,
	LOOP_CACHING_WAIT = 2,
	LOOP_ALLOC_CHUNK = 3,
	LOOP_NO_EMPTY_SIZE = 4,
};

5259 5260 5261 5262
/*
 * walks the btree of allocated extents and find a hole of a given size.
 * The key ins is changed to record the hole:
 * ins->objectid == block start
5263
 * ins->flags = BTRFS_EXTENT_ITEM_KEY
5264 5265 5266
 * ins->offset == number of blocks
 * Any available blocks before search_start are skipped.
 */
C
Chris Mason 已提交
5267
static noinline int find_free_extent(struct btrfs_trans_handle *trans,
5268 5269 5270 5271
				     struct btrfs_root *orig_root,
				     u64 num_bytes, u64 empty_size,
				     u64 search_start, u64 search_end,
				     u64 hint_byte, struct btrfs_key *ins,
5272
				     u64 data)
5273
{
5274
	int ret = 0;
C
Chris Mason 已提交
5275
	struct btrfs_root *root = orig_root->fs_info->extent_root;
5276
	struct btrfs_free_cluster *last_ptr = NULL;
5277
	struct btrfs_block_group_cache *block_group = NULL;
5278
	struct btrfs_block_group_cache *used_block_group;
5279
	int empty_cluster = 2 * 1024 * 1024;
C
Chris Mason 已提交
5280
	int allowed_chunk_alloc = 0;
5281
	int done_chunk_alloc = 0;
5282
	struct btrfs_space_info *space_info;
5283
	int loop = 0;
5284
	int index = 0;
5285 5286
	int alloc_type = (data & BTRFS_BLOCK_GROUP_DATA) ?
		RESERVE_ALLOC_NO_ACCOUNT : RESERVE_ALLOC;
J
Josef Bacik 已提交
5287
	bool found_uncached_bg = false;
5288
	bool failed_cluster_refill = false;
5289
	bool failed_alloc = false;
5290
	bool use_cluster = true;
5291
	bool have_caching_bg = false;
5292 5293
	u64 ideal_cache_percent = 0;
	u64 ideal_cache_offset = 0;
5294

5295
	WARN_ON(num_bytes < root->sectorsize);
5296
	btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
5297 5298
	ins->objectid = 0;
	ins->offset = 0;
5299

J
Josef Bacik 已提交
5300 5301
	trace_find_free_extent(orig_root, num_bytes, empty_size, data);

J
Josef Bacik 已提交
5302
	space_info = __find_space_info(root->fs_info, data);
5303
	if (!space_info) {
5304
		printk(KERN_ERR "No space info for %llu\n", data);
5305 5306
		return -ENOSPC;
	}
J
Josef Bacik 已提交
5307

5308 5309 5310 5311 5312 5313 5314
	/*
	 * If the space info is for both data and metadata it means we have a
	 * small filesystem and we can't use the clustering stuff.
	 */
	if (btrfs_mixed_space_info(space_info))
		use_cluster = false;

C
Chris Mason 已提交
5315 5316 5317
	if (orig_root->ref_cows || empty_size)
		allowed_chunk_alloc = 1;

5318
	if (data & BTRFS_BLOCK_GROUP_METADATA && use_cluster) {
5319
		last_ptr = &root->fs_info->meta_alloc_cluster;
5320 5321
		if (!btrfs_test_opt(root, SSD))
			empty_cluster = 64 * 1024;
5322 5323
	}

5324 5325
	if ((data & BTRFS_BLOCK_GROUP_DATA) && use_cluster &&
	    btrfs_test_opt(root, SSD)) {
5326 5327
		last_ptr = &root->fs_info->data_alloc_cluster;
	}
J
Josef Bacik 已提交
5328

5329
	if (last_ptr) {
5330 5331 5332 5333
		spin_lock(&last_ptr->lock);
		if (last_ptr->block_group)
			hint_byte = last_ptr->window_start;
		spin_unlock(&last_ptr->lock);
5334
	}
5335

5336
	search_start = max(search_start, first_logical_byte(root, 0));
5337
	search_start = max(search_start, hint_byte);
5338

J
Josef Bacik 已提交
5339
	if (!last_ptr)
5340 5341
		empty_cluster = 0;

J
Josef Bacik 已提交
5342
	if (search_start == hint_byte) {
5343
ideal_cache:
J
Josef Bacik 已提交
5344 5345
		block_group = btrfs_lookup_block_group(root->fs_info,
						       search_start);
5346
		used_block_group = block_group;
J
Josef Bacik 已提交
5347 5348 5349
		/*
		 * we don't want to use the block group if it doesn't match our
		 * allocation bits, or if its not cached.
5350 5351 5352
		 *
		 * However if we are re-searching with an ideal block group
		 * picked out then we don't care that the block group is cached.
J
Josef Bacik 已提交
5353 5354
		 */
		if (block_group && block_group_bits(block_group, data) &&
5355 5356
		    (block_group->cached != BTRFS_CACHE_NO ||
		     search_start == ideal_cache_offset)) {
J
Josef Bacik 已提交
5357
			down_read(&space_info->groups_sem);
5358 5359 5360 5361 5362 5363 5364 5365 5366 5367
			if (list_empty(&block_group->list) ||
			    block_group->ro) {
				/*
				 * someone is removing this block group,
				 * we can't jump into the have_block_group
				 * target because our list pointers are not
				 * valid
				 */
				btrfs_put_block_group(block_group);
				up_read(&space_info->groups_sem);
5368
			} else {
5369
				index = get_block_group_index(block_group);
5370
				goto have_block_group;
5371
			}
J
Josef Bacik 已提交
5372
		} else if (block_group) {
5373
			btrfs_put_block_group(block_group);
J
Josef Bacik 已提交
5374
		}
5375
	}
J
Josef Bacik 已提交
5376
search:
5377
	have_caching_bg = false;
5378
	down_read(&space_info->groups_sem);
5379 5380
	list_for_each_entry(block_group, &space_info->block_groups[index],
			    list) {
5381
		u64 offset;
J
Josef Bacik 已提交
5382
		int cached;
5383

5384
		used_block_group = block_group;
5385
		btrfs_get_block_group(block_group);
J
Josef Bacik 已提交
5386
		search_start = block_group->key.objectid;
5387

5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406
		/*
		 * this can happen if we end up cycling through all the
		 * raid types, but we want to make sure we only allocate
		 * for the proper type.
		 */
		if (!block_group_bits(block_group, data)) {
		    u64 extra = BTRFS_BLOCK_GROUP_DUP |
				BTRFS_BLOCK_GROUP_RAID1 |
				BTRFS_BLOCK_GROUP_RAID10;

			/*
			 * if they asked for extra copies and this block group
			 * doesn't provide them, bail.  This does allow us to
			 * fill raid0 from raid1.
			 */
			if ((data & extra) && !(block_group->flags & extra))
				goto loop;
		}

J
Josef Bacik 已提交
5407
have_block_group:
5408 5409
		cached = block_group_cache_done(block_group);
		if (unlikely(!cached)) {
5410 5411
			u64 free_percent;

5412
			found_uncached_bg = true;
5413 5414
			ret = cache_block_group(block_group, trans,
						orig_root, 1);
5415
			if (block_group->cached == BTRFS_CACHE_FINISHED)
5416
				goto alloc;
5417

5418 5419 5420 5421 5422 5423 5424 5425 5426 5427 5428
			free_percent = btrfs_block_group_used(&block_group->item);
			free_percent *= 100;
			free_percent = div64_u64(free_percent,
						 block_group->key.offset);
			free_percent = 100 - free_percent;
			if (free_percent > ideal_cache_percent &&
			    likely(!block_group->ro)) {
				ideal_cache_offset = block_group->key.objectid;
				ideal_cache_percent = free_percent;
			}

J
Josef Bacik 已提交
5429
			/*
5430 5431
			 * The caching workers are limited to 2 threads, so we
			 * can queue as much work as we care to.
J
Josef Bacik 已提交
5432
			 */
5433
			if (loop > LOOP_FIND_IDEAL) {
5434 5435
				ret = cache_block_group(block_group, trans,
							orig_root, 0);
J
Josef Bacik 已提交
5436
				BUG_ON(ret);
J
Josef Bacik 已提交
5437
			}
J
Josef Bacik 已提交
5438

5439 5440 5441 5442 5443
			/*
			 * If loop is set for cached only, try the next block
			 * group.
			 */
			if (loop == LOOP_FIND_IDEAL)
J
Josef Bacik 已提交
5444 5445 5446
				goto loop;
		}

5447
alloc:
5448
		if (unlikely(block_group->ro))
J
Josef Bacik 已提交
5449
			goto loop;
J
Josef Bacik 已提交
5450

5451
		/*
5452 5453
		 * Ok we want to try and use the cluster allocator, so
		 * lets look there
5454
		 */
5455
		if (last_ptr) {
5456 5457 5458 5459 5460
			/*
			 * the refill lock keeps out other
			 * people trying to start a new cluster
			 */
			spin_lock(&last_ptr->refill_lock);
5461 5462 5463 5464 5465 5466
			used_block_group = last_ptr->block_group;
			if (used_block_group != block_group &&
			    (!used_block_group ||
			     used_block_group->ro ||
			     !block_group_bits(used_block_group, data))) {
				used_block_group = block_group;
5467
				goto refill_cluster;
5468 5469 5470 5471
			}

			if (used_block_group != block_group)
				btrfs_get_block_group(used_block_group);
5472

5473 5474
			offset = btrfs_alloc_from_cluster(used_block_group,
			  last_ptr, num_bytes, used_block_group->key.objectid);
5475 5476 5477
			if (offset) {
				/* we have a block, we're done */
				spin_unlock(&last_ptr->refill_lock);
J
Josef Bacik 已提交
5478 5479
				trace_btrfs_reserve_extent_cluster(root,
					block_group, search_start, num_bytes);
5480 5481 5482
				goto checks;
			}

5483 5484 5485 5486
			WARN_ON(last_ptr->block_group != used_block_group);
			if (used_block_group != block_group) {
				btrfs_put_block_group(used_block_group);
				used_block_group = block_group;
5487
			}
5488
refill_cluster:
5489
			BUG_ON(used_block_group != block_group);
5490 5491 5492 5493 5494 5495 5496 5497
			/* If we are on LOOP_NO_EMPTY_SIZE, we can't
			 * set up a new clusters, so lets just skip it
			 * and let the allocator find whatever block
			 * it can find.  If we reach this point, we
			 * will have tried the cluster allocator
			 * plenty of times and not have found
			 * anything, so we are likely way too
			 * fragmented for the clustering stuff to find
5498 5499 5500 5501 5502 5503 5504 5505 5506
			 * anything.
			 *
			 * However, if the cluster is taken from the
			 * current block group, release the cluster
			 * first, so that we stand a better chance of
			 * succeeding in the unclustered
			 * allocation.  */
			if (loop >= LOOP_NO_EMPTY_SIZE &&
			    last_ptr->block_group != block_group) {
5507 5508 5509 5510
				spin_unlock(&last_ptr->refill_lock);
				goto unclustered_alloc;
			}

5511 5512 5513 5514 5515 5516
			/*
			 * this cluster didn't work out, free it and
			 * start over
			 */
			btrfs_return_cluster_to_free_space(NULL, last_ptr);

5517 5518 5519 5520 5521
			if (loop >= LOOP_NO_EMPTY_SIZE) {
				spin_unlock(&last_ptr->refill_lock);
				goto unclustered_alloc;
			}

5522
			/* allocate a cluster in this block group */
5523
			ret = btrfs_find_space_cluster(trans, root,
5524
					       block_group, last_ptr,
5525
					       search_start, num_bytes,
5526 5527 5528 5529 5530 5531 5532 5533 5534 5535 5536 5537
					       empty_cluster + empty_size);
			if (ret == 0) {
				/*
				 * now pull our allocation out of this
				 * cluster
				 */
				offset = btrfs_alloc_from_cluster(block_group,
						  last_ptr, num_bytes,
						  search_start);
				if (offset) {
					/* we found one, proceed */
					spin_unlock(&last_ptr->refill_lock);
J
Josef Bacik 已提交
5538 5539 5540
					trace_btrfs_reserve_extent_cluster(root,
						block_group, search_start,
						num_bytes);
5541 5542
					goto checks;
				}
5543 5544
			} else if (!cached && loop > LOOP_CACHING_NOWAIT
				   && !failed_cluster_refill) {
J
Josef Bacik 已提交
5545 5546
				spin_unlock(&last_ptr->refill_lock);

5547
				failed_cluster_refill = true;
J
Josef Bacik 已提交
5548 5549 5550
				wait_block_group_cache_progress(block_group,
				       num_bytes + empty_cluster + empty_size);
				goto have_block_group;
5551
			}
J
Josef Bacik 已提交
5552

5553 5554 5555 5556 5557 5558
			/*
			 * at this point we either didn't find a cluster
			 * or we weren't able to allocate a block from our
			 * cluster.  Free the cluster we've been trying
			 * to use, and go to the next block group
			 */
5559
			btrfs_return_cluster_to_free_space(NULL, last_ptr);
5560
			spin_unlock(&last_ptr->refill_lock);
5561
			goto loop;
5562 5563
		}

5564
unclustered_alloc:
5565 5566 5567 5568 5569 5570 5571 5572 5573
		spin_lock(&block_group->free_space_ctl->tree_lock);
		if (cached &&
		    block_group->free_space_ctl->free_space <
		    num_bytes + empty_cluster + empty_size) {
			spin_unlock(&block_group->free_space_ctl->tree_lock);
			goto loop;
		}
		spin_unlock(&block_group->free_space_ctl->tree_lock);

5574 5575
		offset = btrfs_find_space_for_alloc(block_group, search_start,
						    num_bytes, empty_size);
5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586
		/*
		 * If we didn't find a chunk, and we haven't failed on this
		 * block group before, and this block group is in the middle of
		 * caching and we are ok with waiting, then go ahead and wait
		 * for progress to be made, and set failed_alloc to true.
		 *
		 * If failed_alloc is true then we've already waited on this
		 * block group once and should move on to the next block group.
		 */
		if (!offset && !failed_alloc && !cached &&
		    loop > LOOP_CACHING_NOWAIT) {
J
Josef Bacik 已提交
5587
			wait_block_group_cache_progress(block_group,
5588 5589
						num_bytes + empty_size);
			failed_alloc = true;
J
Josef Bacik 已提交
5590
			goto have_block_group;
5591
		} else if (!offset) {
5592 5593
			if (!cached)
				have_caching_bg = true;
5594
			goto loop;
J
Josef Bacik 已提交
5595
		}
5596
checks:
5597
		search_start = stripe_align(root, offset);
J
Josef Bacik 已提交
5598
		/* move on to the next group */
5599
		if (search_start + num_bytes >= search_end) {
5600
			btrfs_add_free_space(used_block_group, offset, num_bytes);
J
Josef Bacik 已提交
5601
			goto loop;
5602
		}
5603

J
Josef Bacik 已提交
5604 5605
		/* move on to the next group */
		if (search_start + num_bytes >
5606 5607
		    used_block_group->key.objectid + used_block_group->key.offset) {
			btrfs_add_free_space(used_block_group, offset, num_bytes);
J
Josef Bacik 已提交
5608
			goto loop;
5609
		}
5610

5611
		if (offset < search_start)
5612
			btrfs_add_free_space(used_block_group, offset,
5613 5614
					     search_start - offset);
		BUG_ON(offset > search_start);
J
Josef Bacik 已提交
5615

5616
		ret = btrfs_update_reserved_bytes(used_block_group, num_bytes,
5617
						  alloc_type);
5618
		if (ret == -EAGAIN) {
5619
			btrfs_add_free_space(used_block_group, offset, num_bytes);
J
Josef Bacik 已提交
5620
			goto loop;
J
Josef Bacik 已提交
5621
		}
5622

5623
		/* we are all good, lets return */
J
Josef Bacik 已提交
5624 5625
		ins->objectid = search_start;
		ins->offset = num_bytes;
5626

J
Josef Bacik 已提交
5627 5628
		trace_btrfs_reserve_extent(orig_root, block_group,
					   search_start, num_bytes);
5629
		if (offset < search_start)
5630
			btrfs_add_free_space(used_block_group, offset,
5631 5632
					     search_start - offset);
		BUG_ON(offset > search_start);
5633 5634
		if (used_block_group != block_group)
			btrfs_put_block_group(used_block_group);
5635
		btrfs_put_block_group(block_group);
J
Josef Bacik 已提交
5636 5637
		break;
loop:
5638
		failed_cluster_refill = false;
5639
		failed_alloc = false;
5640
		BUG_ON(index != get_block_group_index(block_group));
5641 5642
		if (used_block_group != block_group)
			btrfs_put_block_group(used_block_group);
5643
		btrfs_put_block_group(block_group);
J
Josef Bacik 已提交
5644 5645 5646
	}
	up_read(&space_info->groups_sem);

5647 5648 5649
	if (!ins->objectid && loop >= LOOP_CACHING_WAIT && have_caching_bg)
		goto search;

5650 5651 5652
	if (!ins->objectid && ++index < BTRFS_NR_RAID_TYPES)
		goto search;

5653 5654 5655 5656 5657
	/* LOOP_FIND_IDEAL, only search caching/cached bg's, and don't wait for
	 *			for them to make caching progress.  Also
	 *			determine the best possible bg to cache
	 * LOOP_CACHING_NOWAIT, search partially cached block groups, kicking
	 *			caching kthreads as we move along
J
Josef Bacik 已提交
5658 5659 5660 5661
	 * LOOP_CACHING_WAIT, search everything, and wait if our bg is caching
	 * LOOP_ALLOC_CHUNK, force a chunk allocation and try again
	 * LOOP_NO_EMPTY_SIZE, set empty_size and empty_cluster to 0 and try
	 *			again
5662
	 */
5663
	if (!ins->objectid && loop < LOOP_NO_EMPTY_SIZE) {
5664
		index = 0;
5665
		if (loop == LOOP_FIND_IDEAL && found_uncached_bg) {
J
Josef Bacik 已提交
5666
			found_uncached_bg = false;
5667
			loop++;
5668
			if (!ideal_cache_percent)
J
Josef Bacik 已提交
5669
				goto search;
5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692 5693 5694 5695 5696 5697 5698 5699 5700 5701

			/*
			 * 1 of the following 2 things have happened so far
			 *
			 * 1) We found an ideal block group for caching that
			 * is mostly full and will cache quickly, so we might
			 * as well wait for it.
			 *
			 * 2) We searched for cached only and we didn't find
			 * anything, and we didn't start any caching kthreads
			 * either, so chances are we will loop through and
			 * start a couple caching kthreads, and then come back
			 * around and just wait for them.  This will be slower
			 * because we will have 2 caching kthreads reading at
			 * the same time when we could have just started one
			 * and waited for it to get far enough to give us an
			 * allocation, so go ahead and go to the wait caching
			 * loop.
			 */
			loop = LOOP_CACHING_WAIT;
			search_start = ideal_cache_offset;
			ideal_cache_percent = 0;
			goto ideal_cache;
		} else if (loop == LOOP_FIND_IDEAL) {
			/*
			 * Didn't find a uncached bg, wait on anything we find
			 * next.
			 */
			loop = LOOP_CACHING_WAIT;
			goto search;
		}

5702
		loop++;
J
Josef Bacik 已提交
5703 5704

		if (loop == LOOP_ALLOC_CHUNK) {
5705 5706 5707 5708 5709 5710 5711 5712 5713 5714 5715 5716
		       if (allowed_chunk_alloc) {
				ret = do_chunk_alloc(trans, root, num_bytes +
						     2 * 1024 * 1024, data,
						     CHUNK_ALLOC_LIMITED);
				allowed_chunk_alloc = 0;
				if (ret == 1)
					done_chunk_alloc = 1;
			} else if (!done_chunk_alloc &&
				   space_info->force_alloc ==
				   CHUNK_ALLOC_NO_FORCE) {
				space_info->force_alloc = CHUNK_ALLOC_LIMITED;
			}
J
Josef Bacik 已提交
5717

5718 5719 5720 5721 5722 5723
		       /*
			* We didn't allocate a chunk, go ahead and drop the
			* empty size and loop again.
			*/
		       if (!done_chunk_alloc)
			       loop = LOOP_NO_EMPTY_SIZE;
J
Josef Bacik 已提交
5724 5725
		}

5726 5727 5728
		if (loop == LOOP_NO_EMPTY_SIZE) {
			empty_size = 0;
			empty_cluster = 0;
5729
		}
5730 5731

		goto search;
J
Josef Bacik 已提交
5732 5733
	} else if (!ins->objectid) {
		ret = -ENOSPC;
5734
	} else if (ins->objectid) {
5735
		ret = 0;
C
Chris Mason 已提交
5736 5737
	}

C
Chris Mason 已提交
5738
	return ret;
5739
}
5740

J
Josef Bacik 已提交
5741 5742
static void dump_space_info(struct btrfs_space_info *info, u64 bytes,
			    int dump_block_groups)
J
Josef Bacik 已提交
5743 5744
{
	struct btrfs_block_group_cache *cache;
5745
	int index = 0;
J
Josef Bacik 已提交
5746

J
Josef Bacik 已提交
5747
	spin_lock(&info->lock);
5748 5749
	printk(KERN_INFO "space_info %llu has %llu free, is %sfull\n",
	       (unsigned long long)info->flags,
C
Chris Mason 已提交
5750
	       (unsigned long long)(info->total_bytes - info->bytes_used -
J
Josef Bacik 已提交
5751
				    info->bytes_pinned - info->bytes_reserved -
5752
				    info->bytes_readonly),
C
Chris Mason 已提交
5753
	       (info->full) ? "" : "not ");
5754 5755
	printk(KERN_INFO "space_info total=%llu, used=%llu, pinned=%llu, "
	       "reserved=%llu, may_use=%llu, readonly=%llu\n",
5756
	       (unsigned long long)info->total_bytes,
5757
	       (unsigned long long)info->bytes_used,
5758
	       (unsigned long long)info->bytes_pinned,
5759
	       (unsigned long long)info->bytes_reserved,
5760
	       (unsigned long long)info->bytes_may_use,
5761
	       (unsigned long long)info->bytes_readonly);
J
Josef Bacik 已提交
5762 5763 5764 5765
	spin_unlock(&info->lock);

	if (!dump_block_groups)
		return;
J
Josef Bacik 已提交
5766

5767
	down_read(&info->groups_sem);
5768 5769
again:
	list_for_each_entry(cache, &info->block_groups[index], list) {
J
Josef Bacik 已提交
5770
		spin_lock(&cache->lock);
C
Chris Mason 已提交
5771 5772 5773 5774 5775 5776 5777
		printk(KERN_INFO "block group %llu has %llu bytes, %llu used "
		       "%llu pinned %llu reserved\n",
		       (unsigned long long)cache->key.objectid,
		       (unsigned long long)cache->key.offset,
		       (unsigned long long)btrfs_block_group_used(&cache->item),
		       (unsigned long long)cache->pinned,
		       (unsigned long long)cache->reserved);
J
Josef Bacik 已提交
5778 5779 5780
		btrfs_dump_free_space(cache, bytes);
		spin_unlock(&cache->lock);
	}
5781 5782
	if (++index < BTRFS_NR_RAID_TYPES)
		goto again;
5783
	up_read(&info->groups_sem);
J
Josef Bacik 已提交
5784
}
5785

5786 5787 5788 5789 5790 5791
int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
			 struct btrfs_root *root,
			 u64 num_bytes, u64 min_alloc_size,
			 u64 empty_size, u64 hint_byte,
			 u64 search_end, struct btrfs_key *ins,
			 u64 data)
5792 5793
{
	int ret;
5794
	u64 search_start = 0;
5795

J
Josef Bacik 已提交
5796
	data = btrfs_get_alloc_profile(root, data);
5797
again:
C
Chris Mason 已提交
5798 5799 5800 5801
	/*
	 * the only place that sets empty_size is btrfs_realloc_node, which
	 * is not called recursively on allocations
	 */
J
Josef Bacik 已提交
5802
	if (empty_size || root->ref_cows)
5803
		ret = do_chunk_alloc(trans, root->fs_info->extent_root,
5804 5805
				     num_bytes + 2 * 1024 * 1024, data,
				     CHUNK_ALLOC_NO_FORCE);
5806

5807 5808
	WARN_ON(num_bytes < root->sectorsize);
	ret = find_free_extent(trans, root, num_bytes, empty_size,
5809 5810
			       search_start, search_end, hint_byte,
			       ins, data);
5811

5812 5813
	if (ret == -ENOSPC && num_bytes > min_alloc_size) {
		num_bytes = num_bytes >> 1;
J
Josef Bacik 已提交
5814
		num_bytes = num_bytes & ~(root->sectorsize - 1);
5815
		num_bytes = max(num_bytes, min_alloc_size);
C
Chris Mason 已提交
5816
		do_chunk_alloc(trans, root->fs_info->extent_root,
5817
			       num_bytes, data, CHUNK_ALLOC_FORCE);
5818 5819
		goto again;
	}
5820
	if (ret == -ENOSPC && btrfs_test_opt(root, ENOSPC_DEBUG)) {
J
Josef Bacik 已提交
5821 5822 5823
		struct btrfs_space_info *sinfo;

		sinfo = __find_space_info(root->fs_info, data);
C
Chris Mason 已提交
5824 5825 5826
		printk(KERN_ERR "btrfs allocation failed flags %llu, "
		       "wanted %llu\n", (unsigned long long)data,
		       (unsigned long long)num_bytes);
J
Josef Bacik 已提交
5827
		dump_space_info(sinfo, num_bytes, 1);
5828
	}
J
Josef Bacik 已提交
5829

5830 5831
	trace_btrfs_reserved_extent_alloc(root, ins->objectid, ins->offset);

J
Josef Bacik 已提交
5832
	return ret;
5833 5834
}

5835 5836
static int __btrfs_free_reserved_extent(struct btrfs_root *root,
					u64 start, u64 len, int pin)
5837
{
J
Josef Bacik 已提交
5838
	struct btrfs_block_group_cache *cache;
5839
	int ret = 0;
J
Josef Bacik 已提交
5840 5841 5842

	cache = btrfs_lookup_block_group(root->fs_info, start);
	if (!cache) {
C
Chris Mason 已提交
5843 5844
		printk(KERN_ERR "Unable to find block group for %llu\n",
		       (unsigned long long)start);
J
Josef Bacik 已提交
5845 5846
		return -ENOSPC;
	}
5847

5848 5849
	if (btrfs_test_opt(root, DISCARD))
		ret = btrfs_discard_extent(root, start, len, NULL);
5850

5851 5852 5853 5854 5855 5856
	if (pin)
		pin_down_extent(root, cache, start, len, 1);
	else {
		btrfs_add_free_space(cache, start, len);
		btrfs_update_reserved_bytes(cache, len, RESERVE_FREE);
	}
5857
	btrfs_put_block_group(cache);
J
Josef Bacik 已提交
5858

5859 5860
	trace_btrfs_reserved_extent_free(root, start, len);

5861 5862 5863
	return ret;
}

5864 5865 5866 5867 5868 5869 5870 5871 5872 5873 5874 5875
int btrfs_free_reserved_extent(struct btrfs_root *root,
					u64 start, u64 len)
{
	return __btrfs_free_reserved_extent(root, start, len, 0);
}

int btrfs_free_and_pin_reserved_extent(struct btrfs_root *root,
				       u64 start, u64 len)
{
	return __btrfs_free_reserved_extent(root, start, len, 1);
}

5876 5877 5878 5879 5880
static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
				      struct btrfs_root *root,
				      u64 parent, u64 root_objectid,
				      u64 flags, u64 owner, u64 offset,
				      struct btrfs_key *ins, int ref_mod)
5881 5882
{
	int ret;
5883
	struct btrfs_fs_info *fs_info = root->fs_info;
5884
	struct btrfs_extent_item *extent_item;
5885
	struct btrfs_extent_inline_ref *iref;
5886
	struct btrfs_path *path;
5887 5888 5889
	struct extent_buffer *leaf;
	int type;
	u32 size;
5890

5891 5892 5893 5894
	if (parent > 0)
		type = BTRFS_SHARED_DATA_REF_KEY;
	else
		type = BTRFS_EXTENT_DATA_REF_KEY;
5895

5896
	size = sizeof(*extent_item) + btrfs_extent_inline_ref_size(type);
5897 5898

	path = btrfs_alloc_path();
T
Tsutomu Itoh 已提交
5899 5900
	if (!path)
		return -ENOMEM;
5901

5902
	path->leave_spinning = 1;
5903 5904
	ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
				      ins, size);
C
Chris Mason 已提交
5905
	BUG_ON(ret);
J
Josef Bacik 已提交
5906

5907 5908
	leaf = path->nodes[0];
	extent_item = btrfs_item_ptr(leaf, path->slots[0],
5909
				     struct btrfs_extent_item);
5910 5911 5912 5913 5914 5915 5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929
	btrfs_set_extent_refs(leaf, extent_item, ref_mod);
	btrfs_set_extent_generation(leaf, extent_item, trans->transid);
	btrfs_set_extent_flags(leaf, extent_item,
			       flags | BTRFS_EXTENT_FLAG_DATA);

	iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
	btrfs_set_extent_inline_ref_type(leaf, iref, type);
	if (parent > 0) {
		struct btrfs_shared_data_ref *ref;
		ref = (struct btrfs_shared_data_ref *)(iref + 1);
		btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
		btrfs_set_shared_data_ref_count(leaf, ref, ref_mod);
	} else {
		struct btrfs_extent_data_ref *ref;
		ref = (struct btrfs_extent_data_ref *)(&iref->offset);
		btrfs_set_extent_data_ref_root(leaf, ref, root_objectid);
		btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
		btrfs_set_extent_data_ref_offset(leaf, ref, offset);
		btrfs_set_extent_data_ref_count(leaf, ref, ref_mod);
	}
5930 5931

	btrfs_mark_buffer_dirty(path->nodes[0]);
5932
	btrfs_free_path(path);
5933

5934
	ret = update_block_group(trans, root, ins->objectid, ins->offset, 1);
5935
	if (ret) {
C
Chris Mason 已提交
5936 5937 5938
		printk(KERN_ERR "btrfs update block group failed for %llu "
		       "%llu\n", (unsigned long long)ins->objectid,
		       (unsigned long long)ins->offset);
5939 5940
		BUG();
	}
5941 5942 5943
	return ret;
}

5944 5945 5946 5947 5948
static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root,
				     u64 parent, u64 root_objectid,
				     u64 flags, struct btrfs_disk_key *key,
				     int level, struct btrfs_key *ins)
5949 5950
{
	int ret;
5951 5952 5953 5954 5955 5956 5957
	struct btrfs_fs_info *fs_info = root->fs_info;
	struct btrfs_extent_item *extent_item;
	struct btrfs_tree_block_info *block_info;
	struct btrfs_extent_inline_ref *iref;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	u32 size = sizeof(*extent_item) + sizeof(*block_info) + sizeof(*iref);
5958

5959
	path = btrfs_alloc_path();
5960 5961
	if (!path)
		return -ENOMEM;
5962

5963 5964 5965
	path->leave_spinning = 1;
	ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
				      ins, size);
5966
	BUG_ON(ret);
5967 5968 5969 5970 5971 5972 5973 5974 5975 5976 5977 5978 5979 5980 5981 5982 5983 5984 5985 5986 5987 5988 5989 5990 5991 5992 5993 5994

	leaf = path->nodes[0];
	extent_item = btrfs_item_ptr(leaf, path->slots[0],
				     struct btrfs_extent_item);
	btrfs_set_extent_refs(leaf, extent_item, 1);
	btrfs_set_extent_generation(leaf, extent_item, trans->transid);
	btrfs_set_extent_flags(leaf, extent_item,
			       flags | BTRFS_EXTENT_FLAG_TREE_BLOCK);
	block_info = (struct btrfs_tree_block_info *)(extent_item + 1);

	btrfs_set_tree_block_key(leaf, block_info, key);
	btrfs_set_tree_block_level(leaf, block_info, level);

	iref = (struct btrfs_extent_inline_ref *)(block_info + 1);
	if (parent > 0) {
		BUG_ON(!(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
		btrfs_set_extent_inline_ref_type(leaf, iref,
						 BTRFS_SHARED_BLOCK_REF_KEY);
		btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
	} else {
		btrfs_set_extent_inline_ref_type(leaf, iref,
						 BTRFS_TREE_BLOCK_REF_KEY);
		btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
	}

	btrfs_mark_buffer_dirty(leaf);
	btrfs_free_path(path);

5995
	ret = update_block_group(trans, root, ins->objectid, ins->offset, 1);
5996 5997 5998 5999 6000 6001 6002 6003 6004 6005 6006 6007 6008 6009 6010 6011 6012 6013
	if (ret) {
		printk(KERN_ERR "btrfs update block group failed for %llu "
		       "%llu\n", (unsigned long long)ins->objectid,
		       (unsigned long long)ins->offset);
		BUG();
	}
	return ret;
}

int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root,
				     u64 root_objectid, u64 owner,
				     u64 offset, struct btrfs_key *ins)
{
	int ret;

	BUG_ON(root_objectid == BTRFS_TREE_LOG_OBJECTID);

A
Arne Jansen 已提交
6014 6015 6016 6017
	ret = btrfs_add_delayed_data_ref(root->fs_info, trans, ins->objectid,
					 ins->offset, 0,
					 root_objectid, owner, offset,
					 BTRFS_ADD_DELAYED_EXTENT, NULL, 0);
6018 6019
	return ret;
}
6020 6021 6022 6023 6024 6025

/*
 * this is used by the tree logging recovery code.  It records that
 * an extent has been allocated and makes sure to clear the free
 * space cache bits as well
 */
6026 6027 6028 6029
int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
				   struct btrfs_root *root,
				   u64 root_objectid, u64 owner, u64 offset,
				   struct btrfs_key *ins)
6030 6031 6032
{
	int ret;
	struct btrfs_block_group_cache *block_group;
6033 6034 6035
	struct btrfs_caching_control *caching_ctl;
	u64 start = ins->objectid;
	u64 num_bytes = ins->offset;
6036 6037

	block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
6038
	cache_block_group(block_group, trans, NULL, 0);
6039
	caching_ctl = get_caching_control(block_group);
6040

6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071
	if (!caching_ctl) {
		BUG_ON(!block_group_cache_done(block_group));
		ret = btrfs_remove_free_space(block_group, start, num_bytes);
		BUG_ON(ret);
	} else {
		mutex_lock(&caching_ctl->mutex);

		if (start >= caching_ctl->progress) {
			ret = add_excluded_extent(root, start, num_bytes);
			BUG_ON(ret);
		} else if (start + num_bytes <= caching_ctl->progress) {
			ret = btrfs_remove_free_space(block_group,
						      start, num_bytes);
			BUG_ON(ret);
		} else {
			num_bytes = caching_ctl->progress - start;
			ret = btrfs_remove_free_space(block_group,
						      start, num_bytes);
			BUG_ON(ret);

			start = caching_ctl->progress;
			num_bytes = ins->objectid + ins->offset -
				    caching_ctl->progress;
			ret = add_excluded_extent(root, start, num_bytes);
			BUG_ON(ret);
		}

		mutex_unlock(&caching_ctl->mutex);
		put_caching_control(caching_ctl);
	}

6072 6073
	ret = btrfs_update_reserved_bytes(block_group, ins->offset,
					  RESERVE_ALLOC_NO_ACCOUNT);
6074
	BUG_ON(ret);
6075
	btrfs_put_block_group(block_group);
6076 6077
	ret = alloc_reserved_file_extent(trans, root, 0, root_objectid,
					 0, owner, offset, ins, 1);
6078 6079 6080
	return ret;
}

6081 6082
struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
					    struct btrfs_root *root,
6083 6084
					    u64 bytenr, u32 blocksize,
					    int level)
6085 6086 6087 6088 6089 6090 6091
{
	struct extent_buffer *buf;

	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
	if (!buf)
		return ERR_PTR(-ENOMEM);
	btrfs_set_header_generation(buf, trans->transid);
6092
	btrfs_set_buffer_lockdep_class(root->root_key.objectid, buf, level);
6093 6094
	btrfs_tree_lock(buf);
	clean_tree_block(trans, root, buf);
6095 6096

	btrfs_set_lock_blocking(buf);
6097
	btrfs_set_buffer_uptodate(buf);
6098

6099
	if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
6100 6101 6102 6103 6104 6105 6106 6107 6108 6109
		/*
		 * we allow two log transactions at a time, use different
		 * EXENT bit to differentiate dirty pages.
		 */
		if (root->log_transid % 2 == 0)
			set_extent_dirty(&root->dirty_log_pages, buf->start,
					buf->start + buf->len - 1, GFP_NOFS);
		else
			set_extent_new(&root->dirty_log_pages, buf->start,
					buf->start + buf->len - 1, GFP_NOFS);
6110 6111
	} else {
		set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
6112
			 buf->start + buf->len - 1, GFP_NOFS);
6113
	}
6114
	trans->blocks_used++;
6115
	/* this returns a buffer locked for blocking */
6116 6117 6118
	return buf;
}

6119 6120 6121 6122 6123
static struct btrfs_block_rsv *
use_block_rsv(struct btrfs_trans_handle *trans,
	      struct btrfs_root *root, u32 blocksize)
{
	struct btrfs_block_rsv *block_rsv;
6124
	struct btrfs_block_rsv *global_rsv = &root->fs_info->global_block_rsv;
6125 6126 6127 6128 6129
	int ret;

	block_rsv = get_block_rsv(trans, root);

	if (block_rsv->size == 0) {
6130
		ret = reserve_metadata_bytes(root, block_rsv, blocksize, 0);
6131 6132 6133 6134 6135 6136 6137 6138
		/*
		 * If we couldn't reserve metadata bytes try and use some from
		 * the global reserve.
		 */
		if (ret && block_rsv != global_rsv) {
			ret = block_rsv_use_bytes(global_rsv, blocksize);
			if (!ret)
				return global_rsv;
6139
			return ERR_PTR(ret);
6140
		} else if (ret) {
6141
			return ERR_PTR(ret);
6142
		}
6143 6144 6145 6146 6147 6148
		return block_rsv;
	}

	ret = block_rsv_use_bytes(block_rsv, blocksize);
	if (!ret)
		return block_rsv;
6149
	if (ret) {
6150 6151 6152 6153 6154 6155 6156
		static DEFINE_RATELIMIT_STATE(_rs,
				DEFAULT_RATELIMIT_INTERVAL,
				/*DEFAULT_RATELIMIT_BURST*/ 2);
		if (__ratelimit(&_rs)) {
			printk(KERN_DEBUG "btrfs: block rsv returned %d\n", ret);
			WARN_ON(1);
		}
6157
		ret = reserve_metadata_bytes(root, block_rsv, blocksize, 0);
6158 6159 6160 6161 6162 6163 6164 6165
		if (!ret) {
			return block_rsv;
		} else if (ret && block_rsv != global_rsv) {
			ret = block_rsv_use_bytes(global_rsv, blocksize);
			if (!ret)
				return global_rsv;
		}
	}
6166 6167 6168 6169

	return ERR_PTR(-ENOSPC);
}

J
Josef Bacik 已提交
6170 6171
static void unuse_block_rsv(struct btrfs_fs_info *fs_info,
			    struct btrfs_block_rsv *block_rsv, u32 blocksize)
6172 6173
{
	block_rsv_add_bytes(block_rsv, blocksize, 0);
J
Josef Bacik 已提交
6174
	block_rsv_release_bytes(fs_info, block_rsv, NULL, 0);
6175 6176
}

6177
/*
6178 6179 6180 6181
 * finds a free extent and does all the dirty work required for allocation
 * returns the key for the extent through ins, and a tree buffer for
 * the first block of the extent through buf.
 *
6182 6183
 * returns the tree buffer or NULL.
 */
6184
struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
6185 6186 6187
					struct btrfs_root *root, u32 blocksize,
					u64 parent, u64 root_objectid,
					struct btrfs_disk_key *key, int level,
A
Arne Jansen 已提交
6188
					u64 hint, u64 empty_size, int for_cow)
6189
{
C
Chris Mason 已提交
6190
	struct btrfs_key ins;
6191
	struct btrfs_block_rsv *block_rsv;
6192
	struct extent_buffer *buf;
6193 6194 6195
	u64 flags = 0;
	int ret;

6196

6197 6198 6199 6200 6201 6202
	block_rsv = use_block_rsv(trans, root, blocksize);
	if (IS_ERR(block_rsv))
		return ERR_CAST(block_rsv);

	ret = btrfs_reserve_extent(trans, root, blocksize, blocksize,
				   empty_size, hint, (u64)-1, &ins, 0);
6203
	if (ret) {
J
Josef Bacik 已提交
6204
		unuse_block_rsv(root->fs_info, block_rsv, blocksize);
6205
		return ERR_PTR(ret);
6206
	}
6207

6208 6209
	buf = btrfs_init_new_buffer(trans, root, ins.objectid,
				    blocksize, level);
6210 6211 6212 6213 6214 6215 6216 6217 6218 6219 6220 6221 6222 6223 6224 6225 6226 6227 6228 6229 6230 6231
	BUG_ON(IS_ERR(buf));

	if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
		if (parent == 0)
			parent = ins.objectid;
		flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
	} else
		BUG_ON(parent > 0);

	if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
		struct btrfs_delayed_extent_op *extent_op;
		extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
		BUG_ON(!extent_op);
		if (key)
			memcpy(&extent_op->key, key, sizeof(extent_op->key));
		else
			memset(&extent_op->key, 0, sizeof(extent_op->key));
		extent_op->flags_to_set = flags;
		extent_op->update_key = 1;
		extent_op->update_flags = 1;
		extent_op->is_data = 0;

A
Arne Jansen 已提交
6232 6233
		ret = btrfs_add_delayed_tree_ref(root->fs_info, trans,
					ins.objectid,
6234 6235
					ins.offset, parent, root_objectid,
					level, BTRFS_ADD_DELAYED_EXTENT,
A
Arne Jansen 已提交
6236
					extent_op, for_cow);
6237 6238
		BUG_ON(ret);
	}
6239 6240
	return buf;
}
6241

6242 6243 6244 6245 6246 6247 6248 6249 6250
struct walk_control {
	u64 refs[BTRFS_MAX_LEVEL];
	u64 flags[BTRFS_MAX_LEVEL];
	struct btrfs_key update_progress;
	int stage;
	int level;
	int shared_level;
	int update_ref;
	int keep_locks;
Y
Yan, Zheng 已提交
6251 6252
	int reada_slot;
	int reada_count;
A
Arne Jansen 已提交
6253
	int for_reloc;
6254 6255 6256 6257 6258
};

#define DROP_REFERENCE	1
#define UPDATE_BACKREF	2

Y
Yan, Zheng 已提交
6259 6260 6261 6262
static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root,
				     struct walk_control *wc,
				     struct btrfs_path *path)
6263
{
Y
Yan, Zheng 已提交
6264 6265 6266
	u64 bytenr;
	u64 generation;
	u64 refs;
6267
	u64 flags;
6268
	u32 nritems;
Y
Yan, Zheng 已提交
6269 6270 6271
	u32 blocksize;
	struct btrfs_key key;
	struct extent_buffer *eb;
6272
	int ret;
Y
Yan, Zheng 已提交
6273 6274
	int slot;
	int nread = 0;
6275

Y
Yan, Zheng 已提交
6276 6277 6278 6279 6280 6281 6282 6283
	if (path->slots[wc->level] < wc->reada_slot) {
		wc->reada_count = wc->reada_count * 2 / 3;
		wc->reada_count = max(wc->reada_count, 2);
	} else {
		wc->reada_count = wc->reada_count * 3 / 2;
		wc->reada_count = min_t(int, wc->reada_count,
					BTRFS_NODEPTRS_PER_BLOCK(root));
	}
6284

Y
Yan, Zheng 已提交
6285 6286 6287
	eb = path->nodes[wc->level];
	nritems = btrfs_header_nritems(eb);
	blocksize = btrfs_level_size(root, wc->level - 1);
6288

Y
Yan, Zheng 已提交
6289 6290 6291
	for (slot = path->slots[wc->level]; slot < nritems; slot++) {
		if (nread >= wc->reada_count)
			break;
6292

C
Chris Mason 已提交
6293
		cond_resched();
Y
Yan, Zheng 已提交
6294 6295
		bytenr = btrfs_node_blockptr(eb, slot);
		generation = btrfs_node_ptr_generation(eb, slot);
C
Chris Mason 已提交
6296

Y
Yan, Zheng 已提交
6297 6298
		if (slot == path->slots[wc->level])
			goto reada;
6299

Y
Yan, Zheng 已提交
6300 6301
		if (wc->stage == UPDATE_BACKREF &&
		    generation <= root->root_key.offset)
6302 6303
			continue;

6304 6305 6306 6307 6308 6309
		/* We don't lock the tree block, it's OK to be racy here */
		ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
					       &refs, &flags);
		BUG_ON(ret);
		BUG_ON(refs == 0);

Y
Yan, Zheng 已提交
6310 6311 6312
		if (wc->stage == DROP_REFERENCE) {
			if (refs == 1)
				goto reada;
6313

6314 6315 6316
			if (wc->level == 1 &&
			    (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
				continue;
Y
Yan, Zheng 已提交
6317 6318 6319 6320 6321 6322 6323 6324
			if (!wc->update_ref ||
			    generation <= root->root_key.offset)
				continue;
			btrfs_node_key_to_cpu(eb, &key, slot);
			ret = btrfs_comp_cpu_keys(&key,
						  &wc->update_progress);
			if (ret < 0)
				continue;
6325 6326 6327 6328
		} else {
			if (wc->level == 1 &&
			    (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
				continue;
6329
		}
Y
Yan, Zheng 已提交
6330 6331 6332 6333
reada:
		ret = readahead_tree_block(root, bytenr, blocksize,
					   generation);
		if (ret)
6334
			break;
Y
Yan, Zheng 已提交
6335
		nread++;
C
Chris Mason 已提交
6336
	}
Y
Yan, Zheng 已提交
6337
	wc->reada_slot = slot;
C
Chris Mason 已提交
6338
}
6339

Y
Yan Zheng 已提交
6340
/*
6341 6342 6343 6344 6345 6346
 * hepler to process tree block while walking down the tree.
 *
 * when wc->stage == UPDATE_BACKREF, this function updates
 * back refs for pointers in the block.
 *
 * NOTE: return value 1 means we should stop walking down.
Y
Yan Zheng 已提交
6347
 */
6348
static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
6349
				   struct btrfs_root *root,
6350
				   struct btrfs_path *path,
6351
				   struct walk_control *wc, int lookup_info)
Y
Yan Zheng 已提交
6352
{
6353 6354 6355
	int level = wc->level;
	struct extent_buffer *eb = path->nodes[level];
	u64 flag = BTRFS_BLOCK_FLAG_FULL_BACKREF;
Y
Yan Zheng 已提交
6356 6357
	int ret;

6358 6359 6360
	if (wc->stage == UPDATE_BACKREF &&
	    btrfs_header_owner(eb) != root->root_key.objectid)
		return 1;
Y
Yan Zheng 已提交
6361

6362 6363 6364 6365
	/*
	 * when reference count of tree block is 1, it won't increase
	 * again. once full backref flag is set, we never clear it.
	 */
6366 6367 6368
	if (lookup_info &&
	    ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
	     (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag)))) {
6369 6370 6371 6372 6373 6374 6375 6376
		BUG_ON(!path->locks[level]);
		ret = btrfs_lookup_extent_info(trans, root,
					       eb->start, eb->len,
					       &wc->refs[level],
					       &wc->flags[level]);
		BUG_ON(ret);
		BUG_ON(wc->refs[level] == 0);
	}
6377

6378 6379 6380
	if (wc->stage == DROP_REFERENCE) {
		if (wc->refs[level] > 1)
			return 1;
Y
Yan Zheng 已提交
6381

6382
		if (path->locks[level] && !wc->keep_locks) {
6383
			btrfs_tree_unlock_rw(eb, path->locks[level]);
6384 6385 6386 6387
			path->locks[level] = 0;
		}
		return 0;
	}
Y
Yan Zheng 已提交
6388

6389 6390 6391
	/* wc->stage == UPDATE_BACKREF */
	if (!(wc->flags[level] & flag)) {
		BUG_ON(!path->locks[level]);
A
Arne Jansen 已提交
6392
		ret = btrfs_inc_ref(trans, root, eb, 1, wc->for_reloc);
Y
Yan Zheng 已提交
6393
		BUG_ON(ret);
A
Arne Jansen 已提交
6394
		ret = btrfs_dec_ref(trans, root, eb, 0, wc->for_reloc);
6395 6396 6397 6398 6399 6400 6401 6402 6403 6404 6405 6406
		BUG_ON(ret);
		ret = btrfs_set_disk_extent_flags(trans, root, eb->start,
						  eb->len, flag, 0);
		BUG_ON(ret);
		wc->flags[level] |= flag;
	}

	/*
	 * the block is shared by multiple trees, so it's not good to
	 * keep the tree lock
	 */
	if (path->locks[level] && level > 0) {
6407
		btrfs_tree_unlock_rw(eb, path->locks[level]);
6408 6409 6410 6411 6412
		path->locks[level] = 0;
	}
	return 0;
}

Y
Yan, Zheng 已提交
6413 6414 6415 6416 6417 6418 6419 6420 6421 6422 6423 6424 6425 6426 6427 6428
/*
 * hepler to process tree block pointer.
 *
 * when wc->stage == DROP_REFERENCE, this function checks
 * reference count of the block pointed to. if the block
 * is shared and we need update back refs for the subtree
 * rooted at the block, this function changes wc->stage to
 * UPDATE_BACKREF. if the block is shared and there is no
 * need to update back, this function drops the reference
 * to the block.
 *
 * NOTE: return value 1 means we should stop walking down.
 */
static noinline int do_walk_down(struct btrfs_trans_handle *trans,
				 struct btrfs_root *root,
				 struct btrfs_path *path,
6429
				 struct walk_control *wc, int *lookup_info)
Y
Yan, Zheng 已提交
6430 6431 6432 6433 6434 6435 6436 6437 6438 6439 6440 6441 6442 6443 6444 6445 6446 6447 6448
{
	u64 bytenr;
	u64 generation;
	u64 parent;
	u32 blocksize;
	struct btrfs_key key;
	struct extent_buffer *next;
	int level = wc->level;
	int reada = 0;
	int ret = 0;

	generation = btrfs_node_ptr_generation(path->nodes[level],
					       path->slots[level]);
	/*
	 * if the lower level block was created before the snapshot
	 * was created, we know there is no need to update back refs
	 * for the subtree
	 */
	if (wc->stage == UPDATE_BACKREF &&
6449 6450
	    generation <= root->root_key.offset) {
		*lookup_info = 1;
Y
Yan, Zheng 已提交
6451
		return 1;
6452
	}
Y
Yan, Zheng 已提交
6453 6454 6455 6456 6457 6458 6459

	bytenr = btrfs_node_blockptr(path->nodes[level], path->slots[level]);
	blocksize = btrfs_level_size(root, level - 1);

	next = btrfs_find_tree_block(root, bytenr, blocksize);
	if (!next) {
		next = btrfs_find_create_tree_block(root, bytenr, blocksize);
6460 6461
		if (!next)
			return -ENOMEM;
Y
Yan, Zheng 已提交
6462 6463 6464 6465 6466
		reada = 1;
	}
	btrfs_tree_lock(next);
	btrfs_set_lock_blocking(next);

6467 6468 6469 6470 6471 6472
	ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
				       &wc->refs[level - 1],
				       &wc->flags[level - 1]);
	BUG_ON(ret);
	BUG_ON(wc->refs[level - 1] == 0);
	*lookup_info = 0;
Y
Yan, Zheng 已提交
6473

6474
	if (wc->stage == DROP_REFERENCE) {
Y
Yan, Zheng 已提交
6475
		if (wc->refs[level - 1] > 1) {
6476 6477 6478 6479
			if (level == 1 &&
			    (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
				goto skip;

Y
Yan, Zheng 已提交
6480 6481 6482 6483 6484 6485 6486 6487 6488 6489 6490 6491 6492
			if (!wc->update_ref ||
			    generation <= root->root_key.offset)
				goto skip;

			btrfs_node_key_to_cpu(path->nodes[level], &key,
					      path->slots[level]);
			ret = btrfs_comp_cpu_keys(&key, &wc->update_progress);
			if (ret < 0)
				goto skip;

			wc->stage = UPDATE_BACKREF;
			wc->shared_level = level - 1;
		}
6493 6494 6495 6496
	} else {
		if (level == 1 &&
		    (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
			goto skip;
Y
Yan, Zheng 已提交
6497 6498 6499 6500 6501 6502
	}

	if (!btrfs_buffer_uptodate(next, generation)) {
		btrfs_tree_unlock(next);
		free_extent_buffer(next);
		next = NULL;
6503
		*lookup_info = 1;
Y
Yan, Zheng 已提交
6504 6505 6506 6507 6508 6509
	}

	if (!next) {
		if (reada && level == 1)
			reada_walk_down(trans, root, wc, path);
		next = read_tree_block(root, bytenr, blocksize, generation);
6510 6511
		if (!next)
			return -EIO;
Y
Yan, Zheng 已提交
6512 6513 6514 6515 6516 6517 6518 6519
		btrfs_tree_lock(next);
		btrfs_set_lock_blocking(next);
	}

	level--;
	BUG_ON(level != btrfs_header_level(next));
	path->nodes[level] = next;
	path->slots[level] = 0;
6520
	path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
Y
Yan, Zheng 已提交
6521 6522 6523 6524 6525 6526 6527
	wc->level = level;
	if (wc->level == 1)
		wc->reada_slot = 0;
	return 0;
skip:
	wc->refs[level - 1] = 0;
	wc->flags[level - 1] = 0;
6528 6529 6530 6531 6532 6533 6534 6535
	if (wc->stage == DROP_REFERENCE) {
		if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
			parent = path->nodes[level]->start;
		} else {
			BUG_ON(root->root_key.objectid !=
			       btrfs_header_owner(path->nodes[level]));
			parent = 0;
		}
Y
Yan, Zheng 已提交
6536

6537
		ret = btrfs_free_extent(trans, root, bytenr, blocksize, parent,
A
Arne Jansen 已提交
6538
				root->root_key.objectid, level - 1, 0, 0);
6539
		BUG_ON(ret);
Y
Yan, Zheng 已提交
6540 6541 6542
	}
	btrfs_tree_unlock(next);
	free_extent_buffer(next);
6543
	*lookup_info = 1;
Y
Yan, Zheng 已提交
6544 6545 6546
	return 1;
}

6547 6548 6549 6550 6551 6552 6553 6554 6555 6556 6557 6558 6559 6560 6561 6562 6563
/*
 * hepler to process tree block while walking up the tree.
 *
 * when wc->stage == DROP_REFERENCE, this function drops
 * reference count on the block.
 *
 * when wc->stage == UPDATE_BACKREF, this function changes
 * wc->stage back to DROP_REFERENCE if we changed wc->stage
 * to UPDATE_BACKREF previously while processing the block.
 *
 * NOTE: return value 1 means we should stop walking up.
 */
static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
				 struct btrfs_root *root,
				 struct btrfs_path *path,
				 struct walk_control *wc)
{
6564
	int ret;
6565 6566 6567 6568 6569 6570 6571 6572 6573 6574 6575 6576 6577 6578 6579 6580 6581 6582 6583 6584 6585 6586 6587 6588 6589 6590
	int level = wc->level;
	struct extent_buffer *eb = path->nodes[level];
	u64 parent = 0;

	if (wc->stage == UPDATE_BACKREF) {
		BUG_ON(wc->shared_level < level);
		if (level < wc->shared_level)
			goto out;

		ret = find_next_key(path, level + 1, &wc->update_progress);
		if (ret > 0)
			wc->update_ref = 0;

		wc->stage = DROP_REFERENCE;
		wc->shared_level = -1;
		path->slots[level] = 0;

		/*
		 * check reference count again if the block isn't locked.
		 * we should start walking down the tree again if reference
		 * count is one.
		 */
		if (!path->locks[level]) {
			BUG_ON(level == 0);
			btrfs_tree_lock(eb);
			btrfs_set_lock_blocking(eb);
6591
			path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
6592 6593 6594 6595 6596

			ret = btrfs_lookup_extent_info(trans, root,
						       eb->start, eb->len,
						       &wc->refs[level],
						       &wc->flags[level]);
Y
Yan Zheng 已提交
6597
			BUG_ON(ret);
6598 6599
			BUG_ON(wc->refs[level] == 0);
			if (wc->refs[level] == 1) {
6600
				btrfs_tree_unlock_rw(eb, path->locks[level]);
6601 6602
				return 1;
			}
Y
Yan Zheng 已提交
6603
		}
6604
	}
Y
Yan Zheng 已提交
6605

6606 6607
	/* wc->stage == DROP_REFERENCE */
	BUG_ON(wc->refs[level] > 1 && !path->locks[level]);
6608

6609 6610 6611
	if (wc->refs[level] == 1) {
		if (level == 0) {
			if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
A
Arne Jansen 已提交
6612 6613
				ret = btrfs_dec_ref(trans, root, eb, 1,
						    wc->for_reloc);
6614
			else
A
Arne Jansen 已提交
6615 6616
				ret = btrfs_dec_ref(trans, root, eb, 0,
						    wc->for_reloc);
6617 6618 6619 6620 6621 6622 6623
			BUG_ON(ret);
		}
		/* make block locked assertion in clean_tree_block happy */
		if (!path->locks[level] &&
		    btrfs_header_generation(eb) == trans->transid) {
			btrfs_tree_lock(eb);
			btrfs_set_lock_blocking(eb);
6624
			path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
6625 6626 6627 6628 6629 6630 6631 6632 6633 6634 6635 6636 6637 6638 6639 6640
		}
		clean_tree_block(trans, root, eb);
	}

	if (eb == root->node) {
		if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
			parent = eb->start;
		else
			BUG_ON(root->root_key.objectid !=
			       btrfs_header_owner(eb));
	} else {
		if (wc->flags[level + 1] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
			parent = path->nodes[level + 1]->start;
		else
			BUG_ON(root->root_key.objectid !=
			       btrfs_header_owner(path->nodes[level + 1]));
Y
Yan Zheng 已提交
6641 6642
	}

A
Arne Jansen 已提交
6643
	btrfs_free_tree_block(trans, root, eb, parent, wc->refs[level] == 1, 0);
6644 6645 6646
out:
	wc->refs[level] = 0;
	wc->flags[level] = 0;
6647
	return 0;
6648 6649 6650 6651 6652 6653 6654 6655
}

static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
				   struct btrfs_root *root,
				   struct btrfs_path *path,
				   struct walk_control *wc)
{
	int level = wc->level;
6656
	int lookup_info = 1;
6657 6658 6659
	int ret;

	while (level >= 0) {
6660
		ret = walk_down_proc(trans, root, path, wc, lookup_info);
6661 6662 6663 6664 6665 6666
		if (ret > 0)
			break;

		if (level == 0)
			break;

6667 6668 6669 6670
		if (path->slots[level] >=
		    btrfs_header_nritems(path->nodes[level]))
			break;

6671
		ret = do_walk_down(trans, root, path, wc, &lookup_info);
Y
Yan, Zheng 已提交
6672 6673 6674
		if (ret > 0) {
			path->slots[level]++;
			continue;
6675 6676
		} else if (ret < 0)
			return ret;
Y
Yan, Zheng 已提交
6677
		level = wc->level;
Y
Yan Zheng 已提交
6678 6679 6680 6681
	}
	return 0;
}

C
Chris Mason 已提交
6682
static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
6683
				 struct btrfs_root *root,
Y
Yan Zheng 已提交
6684
				 struct btrfs_path *path,
6685
				 struct walk_control *wc, int max_level)
C
Chris Mason 已提交
6686
{
6687
	int level = wc->level;
C
Chris Mason 已提交
6688
	int ret;
6689

6690 6691 6692 6693 6694 6695
	path->slots[level] = btrfs_header_nritems(path->nodes[level]);
	while (level < max_level && path->nodes[level]) {
		wc->level = level;
		if (path->slots[level] + 1 <
		    btrfs_header_nritems(path->nodes[level])) {
			path->slots[level]++;
C
Chris Mason 已提交
6696 6697
			return 0;
		} else {
6698 6699 6700
			ret = walk_up_proc(trans, root, path, wc);
			if (ret > 0)
				return 0;
6701

6702
			if (path->locks[level]) {
6703 6704
				btrfs_tree_unlock_rw(path->nodes[level],
						     path->locks[level]);
6705
				path->locks[level] = 0;
Y
Yan Zheng 已提交
6706
			}
6707 6708 6709
			free_extent_buffer(path->nodes[level]);
			path->nodes[level] = NULL;
			level++;
C
Chris Mason 已提交
6710 6711 6712 6713 6714
		}
	}
	return 1;
}

C
Chris Mason 已提交
6715
/*
6716 6717 6718 6719 6720 6721 6722 6723 6724
 * drop a subvolume tree.
 *
 * this function traverses the tree freeing any blocks that only
 * referenced by the tree.
 *
 * when a shared tree block is found. this function decreases its
 * reference count by one. if update_ref is true, this function
 * also make sure backrefs for the shared block and all lower level
 * blocks are properly updated.
C
Chris Mason 已提交
6725
 */
6726
void btrfs_drop_snapshot(struct btrfs_root *root,
A
Arne Jansen 已提交
6727 6728
			 struct btrfs_block_rsv *block_rsv, int update_ref,
			 int for_reloc)
C
Chris Mason 已提交
6729
{
6730
	struct btrfs_path *path;
6731 6732
	struct btrfs_trans_handle *trans;
	struct btrfs_root *tree_root = root->fs_info->tree_root;
6733
	struct btrfs_root_item *root_item = &root->root_item;
6734 6735 6736 6737 6738
	struct walk_control *wc;
	struct btrfs_key key;
	int err = 0;
	int ret;
	int level;
C
Chris Mason 已提交
6739

6740
	path = btrfs_alloc_path();
6741 6742 6743 6744
	if (!path) {
		err = -ENOMEM;
		goto out;
	}
C
Chris Mason 已提交
6745

6746
	wc = kzalloc(sizeof(*wc), GFP_NOFS);
6747 6748
	if (!wc) {
		btrfs_free_path(path);
6749 6750
		err = -ENOMEM;
		goto out;
6751
	}
6752

6753
	trans = btrfs_start_transaction(tree_root, 0);
6754 6755
	BUG_ON(IS_ERR(trans));

6756 6757
	if (block_rsv)
		trans->block_rsv = block_rsv;
6758

6759
	if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
6760
		level = btrfs_header_level(root->node);
6761 6762
		path->nodes[level] = btrfs_lock_root_node(root);
		btrfs_set_lock_blocking(path->nodes[level]);
6763
		path->slots[level] = 0;
6764
		path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
6765 6766
		memset(&wc->update_progress, 0,
		       sizeof(wc->update_progress));
6767 6768
	} else {
		btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
6769 6770 6771
		memcpy(&wc->update_progress, &key,
		       sizeof(wc->update_progress));

6772
		level = root_item->drop_level;
6773
		BUG_ON(level == 0);
6774
		path->lowest_level = level;
6775 6776 6777 6778
		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
		path->lowest_level = 0;
		if (ret < 0) {
			err = ret;
6779
			goto out_free;
6780
		}
Y
Yan, Zheng 已提交
6781
		WARN_ON(ret > 0);
6782

6783 6784 6785 6786
		/*
		 * unlock our path, this is safe because only this
		 * function is allowed to delete this snapshot
		 */
6787
		btrfs_unlock_up_safe(path, 0);
6788 6789 6790 6791 6792 6793 6794 6795 6796 6797 6798 6799 6800 6801 6802 6803 6804 6805 6806 6807 6808

		level = btrfs_header_level(root->node);
		while (1) {
			btrfs_tree_lock(path->nodes[level]);
			btrfs_set_lock_blocking(path->nodes[level]);

			ret = btrfs_lookup_extent_info(trans, root,
						path->nodes[level]->start,
						path->nodes[level]->len,
						&wc->refs[level],
						&wc->flags[level]);
			BUG_ON(ret);
			BUG_ON(wc->refs[level] == 0);

			if (level == root_item->drop_level)
				break;

			btrfs_tree_unlock(path->nodes[level]);
			WARN_ON(wc->refs[level] != 1);
			level--;
		}
6809
	}
6810 6811 6812 6813 6814 6815

	wc->level = level;
	wc->shared_level = -1;
	wc->stage = DROP_REFERENCE;
	wc->update_ref = update_ref;
	wc->keep_locks = 0;
A
Arne Jansen 已提交
6816
	wc->for_reloc = for_reloc;
Y
Yan, Zheng 已提交
6817
	wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(root);
6818

C
Chris Mason 已提交
6819
	while (1) {
6820 6821 6822
		ret = walk_down_tree(trans, root, path, wc);
		if (ret < 0) {
			err = ret;
C
Chris Mason 已提交
6823
			break;
6824
		}
C
Chris Mason 已提交
6825

6826 6827 6828
		ret = walk_up_tree(trans, root, path, wc, BTRFS_MAX_LEVEL);
		if (ret < 0) {
			err = ret;
C
Chris Mason 已提交
6829
			break;
6830 6831 6832 6833
		}

		if (ret > 0) {
			BUG_ON(wc->stage != DROP_REFERENCE);
6834 6835
			break;
		}
6836 6837 6838 6839 6840 6841 6842 6843 6844 6845

		if (wc->stage == DROP_REFERENCE) {
			level = wc->level;
			btrfs_node_key(path->nodes[level],
				       &root_item->drop_progress,
				       path->slots[level]);
			root_item->drop_level = level;
		}

		BUG_ON(wc->level == 0);
6846
		if (btrfs_should_end_transaction(trans, tree_root)) {
6847 6848 6849 6850 6851
			ret = btrfs_update_root(trans, tree_root,
						&root->root_key,
						root_item);
			BUG_ON(ret);

6852
			btrfs_end_transaction_throttle(trans, tree_root);
6853
			trans = btrfs_start_transaction(tree_root, 0);
6854
			BUG_ON(IS_ERR(trans));
6855 6856
			if (block_rsv)
				trans->block_rsv = block_rsv;
6857
		}
C
Chris Mason 已提交
6858
	}
6859
	btrfs_release_path(path);
6860 6861 6862 6863 6864
	BUG_ON(err);

	ret = btrfs_del_root(trans, tree_root, &root->root_key);
	BUG_ON(ret);

6865 6866 6867 6868 6869
	if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
		ret = btrfs_find_last_root(tree_root, root->root_key.objectid,
					   NULL, NULL);
		BUG_ON(ret < 0);
		if (ret > 0) {
6870 6871 6872 6873 6874 6875 6876
			/* if we fail to delete the orphan item this time
			 * around, it'll get picked up the next time.
			 *
			 * The most common failure here is just -ENOENT.
			 */
			btrfs_del_orphan_item(trans, tree_root,
					      root->root_key.objectid);
6877 6878 6879 6880 6881 6882 6883 6884 6885 6886
		}
	}

	if (root->in_radix) {
		btrfs_free_fs_root(tree_root->fs_info, root);
	} else {
		free_extent_buffer(root->node);
		free_extent_buffer(root->commit_root);
		kfree(root);
	}
6887
out_free:
6888
	btrfs_end_transaction_throttle(trans, tree_root);
6889
	kfree(wc);
6890
	btrfs_free_path(path);
6891 6892 6893 6894
out:
	if (err)
		btrfs_std_error(root->fs_info, err);
	return;
C
Chris Mason 已提交
6895
}
C
Chris Mason 已提交
6896

6897 6898 6899 6900
/*
 * drop subtree rooted at tree block 'node'.
 *
 * NOTE: this function will unlock and release tree block 'node'
A
Arne Jansen 已提交
6901
 * only used by relocation code
6902
 */
Y
Yan Zheng 已提交
6903 6904 6905 6906 6907 6908
int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
			struct btrfs_root *root,
			struct extent_buffer *node,
			struct extent_buffer *parent)
{
	struct btrfs_path *path;
6909
	struct walk_control *wc;
Y
Yan Zheng 已提交
6910 6911 6912 6913 6914
	int level;
	int parent_level;
	int ret = 0;
	int wret;

6915 6916
	BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);

Y
Yan Zheng 已提交
6917
	path = btrfs_alloc_path();
T
Tsutomu Itoh 已提交
6918 6919
	if (!path)
		return -ENOMEM;
Y
Yan Zheng 已提交
6920

6921
	wc = kzalloc(sizeof(*wc), GFP_NOFS);
T
Tsutomu Itoh 已提交
6922 6923 6924 6925
	if (!wc) {
		btrfs_free_path(path);
		return -ENOMEM;
	}
6926

6927
	btrfs_assert_tree_locked(parent);
Y
Yan Zheng 已提交
6928 6929 6930 6931 6932
	parent_level = btrfs_header_level(parent);
	extent_buffer_get(parent);
	path->nodes[parent_level] = parent;
	path->slots[parent_level] = btrfs_header_nritems(parent);

6933
	btrfs_assert_tree_locked(node);
Y
Yan Zheng 已提交
6934 6935 6936
	level = btrfs_header_level(node);
	path->nodes[level] = node;
	path->slots[level] = 0;
6937
	path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
6938 6939 6940 6941 6942 6943 6944 6945

	wc->refs[parent_level] = 1;
	wc->flags[parent_level] = BTRFS_BLOCK_FLAG_FULL_BACKREF;
	wc->level = level;
	wc->shared_level = -1;
	wc->stage = DROP_REFERENCE;
	wc->update_ref = 0;
	wc->keep_locks = 1;
A
Arne Jansen 已提交
6946
	wc->for_reloc = 1;
Y
Yan, Zheng 已提交
6947
	wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(root);
Y
Yan Zheng 已提交
6948 6949

	while (1) {
6950 6951
		wret = walk_down_tree(trans, root, path, wc);
		if (wret < 0) {
Y
Yan Zheng 已提交
6952 6953
			ret = wret;
			break;
6954
		}
Y
Yan Zheng 已提交
6955

6956
		wret = walk_up_tree(trans, root, path, wc, parent_level);
Y
Yan Zheng 已提交
6957 6958 6959 6960 6961 6962
		if (wret < 0)
			ret = wret;
		if (wret != 0)
			break;
	}

6963
	kfree(wc);
Y
Yan Zheng 已提交
6964 6965 6966 6967
	btrfs_free_path(path);
	return ret;
}

6968 6969 6970 6971 6972 6973
static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
{
	u64 num_devices;
	u64 stripped = BTRFS_BLOCK_GROUP_RAID0 |
		BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;

6974 6975 6976 6977 6978 6979 6980 6981 6982 6983 6984 6985 6986 6987 6988 6989 6990 6991 6992 6993 6994 6995 6996
	if (root->fs_info->balance_ctl) {
		struct btrfs_balance_control *bctl = root->fs_info->balance_ctl;
		u64 tgt = 0;

		/* pick restriper's target profile and return */
		if (flags & BTRFS_BLOCK_GROUP_DATA &&
		    bctl->data.flags & BTRFS_BALANCE_ARGS_CONVERT) {
			tgt = BTRFS_BLOCK_GROUP_DATA | bctl->data.target;
		} else if (flags & BTRFS_BLOCK_GROUP_SYSTEM &&
			   bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) {
			tgt = BTRFS_BLOCK_GROUP_SYSTEM | bctl->sys.target;
		} else if (flags & BTRFS_BLOCK_GROUP_METADATA &&
			   bctl->meta.flags & BTRFS_BALANCE_ARGS_CONVERT) {
			tgt = BTRFS_BLOCK_GROUP_METADATA | bctl->meta.target;
		}

		if (tgt) {
			/* extended -> chunk profile */
			tgt &= ~BTRFS_AVAIL_ALLOC_BIT_SINGLE;
			return tgt;
		}
	}

6997 6998 6999 7000 7001 7002 7003 7004
	/*
	 * we add in the count of missing devices because we want
	 * to make sure that any RAID levels on a degraded FS
	 * continue to be honored.
	 */
	num_devices = root->fs_info->fs_devices->rw_devices +
		root->fs_info->fs_devices->missing_devices;

7005 7006 7007 7008 7009 7010 7011 7012 7013 7014 7015 7016 7017 7018 7019 7020 7021 7022 7023 7024 7025 7026 7027 7028 7029 7030 7031 7032 7033 7034 7035
	if (num_devices == 1) {
		stripped |= BTRFS_BLOCK_GROUP_DUP;
		stripped = flags & ~stripped;

		/* turn raid0 into single device chunks */
		if (flags & BTRFS_BLOCK_GROUP_RAID0)
			return stripped;

		/* turn mirroring into duplication */
		if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
			     BTRFS_BLOCK_GROUP_RAID10))
			return stripped | BTRFS_BLOCK_GROUP_DUP;
		return flags;
	} else {
		/* they already had raid on here, just return */
		if (flags & stripped)
			return flags;

		stripped |= BTRFS_BLOCK_GROUP_DUP;
		stripped = flags & ~stripped;

		/* switch duplicated blocks with raid1 */
		if (flags & BTRFS_BLOCK_GROUP_DUP)
			return stripped | BTRFS_BLOCK_GROUP_RAID1;

		/* turn single device chunks into raid0 */
		return stripped | BTRFS_BLOCK_GROUP_RAID0;
	}
	return flags;
}

7036
static int set_block_group_ro(struct btrfs_block_group_cache *cache, int force)
C
Chris Mason 已提交
7037
{
7038 7039
	struct btrfs_space_info *sinfo = cache->space_info;
	u64 num_bytes;
7040
	u64 min_allocable_bytes;
7041
	int ret = -ENOSPC;
C
Chris Mason 已提交
7042

7043

7044 7045 7046 7047 7048 7049 7050 7051 7052 7053 7054 7055
	/*
	 * We need some metadata space and system metadata space for
	 * allocating chunks in some corner cases until we force to set
	 * it to be readonly.
	 */
	if ((sinfo->flags &
	     (BTRFS_BLOCK_GROUP_SYSTEM | BTRFS_BLOCK_GROUP_METADATA)) &&
	    !force)
		min_allocable_bytes = 1 * 1024 * 1024;
	else
		min_allocable_bytes = 0;

7056 7057
	spin_lock(&sinfo->lock);
	spin_lock(&cache->lock);
7058 7059 7060 7061 7062 7063

	if (cache->ro) {
		ret = 0;
		goto out;
	}

7064 7065 7066 7067
	num_bytes = cache->key.offset - cache->reserved - cache->pinned -
		    cache->bytes_super - btrfs_block_group_used(&cache->item);

	if (sinfo->bytes_used + sinfo->bytes_reserved + sinfo->bytes_pinned +
7068 7069
	    sinfo->bytes_may_use + sinfo->bytes_readonly + num_bytes +
	    min_allocable_bytes <= sinfo->total_bytes) {
7070 7071 7072 7073
		sinfo->bytes_readonly += num_bytes;
		cache->ro = 1;
		ret = 0;
	}
7074
out:
7075 7076 7077 7078
	spin_unlock(&cache->lock);
	spin_unlock(&sinfo->lock);
	return ret;
}
7079

7080 7081
int btrfs_set_block_group_ro(struct btrfs_root *root,
			     struct btrfs_block_group_cache *cache)
7082

7083 7084 7085 7086
{
	struct btrfs_trans_handle *trans;
	u64 alloc_flags;
	int ret;
7087

7088
	BUG_ON(cache->ro);
C
Chris Mason 已提交
7089

C
Chris Mason 已提交
7090
	trans = btrfs_join_transaction(root);
7091
	BUG_ON(IS_ERR(trans));
7092

7093 7094
	alloc_flags = update_block_group_flags(root, cache->flags);
	if (alloc_flags != cache->flags)
7095 7096
		do_chunk_alloc(trans, root, 2 * 1024 * 1024, alloc_flags,
			       CHUNK_ALLOC_FORCE);
7097

7098
	ret = set_block_group_ro(cache, 0);
7099 7100 7101
	if (!ret)
		goto out;
	alloc_flags = get_alloc_profile(root, cache->space_info->flags);
7102 7103
	ret = do_chunk_alloc(trans, root, 2 * 1024 * 1024, alloc_flags,
			     CHUNK_ALLOC_FORCE);
7104 7105
	if (ret < 0)
		goto out;
7106
	ret = set_block_group_ro(cache, 0);
7107 7108 7109 7110
out:
	btrfs_end_transaction(trans, root);
	return ret;
}
7111

7112 7113 7114 7115
int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans,
			    struct btrfs_root *root, u64 type)
{
	u64 alloc_flags = get_alloc_profile(root, type);
7116 7117
	return do_chunk_alloc(trans, root, 2 * 1024 * 1024, alloc_flags,
			      CHUNK_ALLOC_FORCE);
7118 7119
}

7120 7121 7122 7123 7124 7125 7126 7127 7128 7129 7130 7131 7132 7133 7134 7135 7136 7137 7138 7139 7140 7141 7142 7143 7144 7145 7146 7147 7148 7149 7150 7151 7152 7153 7154 7155 7156 7157 7158 7159 7160 7161 7162 7163 7164 7165 7166 7167 7168 7169 7170 7171 7172 7173 7174 7175
/*
 * helper to account the unused space of all the readonly block group in the
 * list. takes mirrors into account.
 */
static u64 __btrfs_get_ro_block_group_free_space(struct list_head *groups_list)
{
	struct btrfs_block_group_cache *block_group;
	u64 free_bytes = 0;
	int factor;

	list_for_each_entry(block_group, groups_list, list) {
		spin_lock(&block_group->lock);

		if (!block_group->ro) {
			spin_unlock(&block_group->lock);
			continue;
		}

		if (block_group->flags & (BTRFS_BLOCK_GROUP_RAID1 |
					  BTRFS_BLOCK_GROUP_RAID10 |
					  BTRFS_BLOCK_GROUP_DUP))
			factor = 2;
		else
			factor = 1;

		free_bytes += (block_group->key.offset -
			       btrfs_block_group_used(&block_group->item)) *
			       factor;

		spin_unlock(&block_group->lock);
	}

	return free_bytes;
}

/*
 * helper to account the unused space of all the readonly block group in the
 * space_info. takes mirrors into account.
 */
u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo)
{
	int i;
	u64 free_bytes = 0;

	spin_lock(&sinfo->lock);

	for(i = 0; i < BTRFS_NR_RAID_TYPES; i++)
		if (!list_empty(&sinfo->block_groups[i]))
			free_bytes += __btrfs_get_ro_block_group_free_space(
						&sinfo->block_groups[i]);

	spin_unlock(&sinfo->lock);

	return free_bytes;
}

7176 7177
int btrfs_set_block_group_rw(struct btrfs_root *root,
			      struct btrfs_block_group_cache *cache)
7178
{
7179 7180 7181 7182 7183 7184 7185 7186 7187 7188 7189 7190 7191
	struct btrfs_space_info *sinfo = cache->space_info;
	u64 num_bytes;

	BUG_ON(!cache->ro);

	spin_lock(&sinfo->lock);
	spin_lock(&cache->lock);
	num_bytes = cache->key.offset - cache->reserved - cache->pinned -
		    cache->bytes_super - btrfs_block_group_used(&cache->item);
	sinfo->bytes_readonly -= num_bytes;
	cache->ro = 0;
	spin_unlock(&cache->lock);
	spin_unlock(&sinfo->lock);
7192 7193 7194
	return 0;
}

7195 7196 7197 7198 7199 7200 7201
/*
 * checks to see if its even possible to relocate this block group.
 *
 * @return - -1 if it's not a good idea to relocate this block group, 0 if its
 * ok to go ahead and try.
 */
int btrfs_can_relocate(struct btrfs_root *root, u64 bytenr)
Z
Zheng Yan 已提交
7202
{
7203 7204 7205 7206
	struct btrfs_block_group_cache *block_group;
	struct btrfs_space_info *space_info;
	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
	struct btrfs_device *device;
7207
	u64 min_free;
J
Josef Bacik 已提交
7208 7209
	u64 dev_min = 1;
	u64 dev_nr = 0;
7210
	int index;
7211 7212
	int full = 0;
	int ret = 0;
Z
Zheng Yan 已提交
7213

7214
	block_group = btrfs_lookup_block_group(root->fs_info, bytenr);
Z
Zheng Yan 已提交
7215

7216 7217 7218
	/* odd, couldn't find the block group, leave it alone */
	if (!block_group)
		return -1;
Z
Zheng Yan 已提交
7219

7220 7221
	min_free = btrfs_block_group_used(&block_group->item);

7222
	/* no bytes used, we're good */
7223
	if (!min_free)
Z
Zheng Yan 已提交
7224 7225
		goto out;

7226 7227
	space_info = block_group->space_info;
	spin_lock(&space_info->lock);
7228

7229
	full = space_info->full;
7230

7231 7232
	/*
	 * if this is the last block group we have in this space, we can't
7233 7234 7235 7236
	 * relocate it unless we're able to allocate a new chunk below.
	 *
	 * Otherwise, we need to make sure we have room in the space to handle
	 * all of the extents from this block group.  If we can, we're good
7237
	 */
7238
	if ((space_info->total_bytes != block_group->key.offset) &&
7239 7240 7241
	    (space_info->bytes_used + space_info->bytes_reserved +
	     space_info->bytes_pinned + space_info->bytes_readonly +
	     min_free < space_info->total_bytes)) {
7242 7243
		spin_unlock(&space_info->lock);
		goto out;
7244
	}
7245
	spin_unlock(&space_info->lock);
7246

7247 7248 7249 7250 7251 7252 7253 7254 7255 7256
	/*
	 * ok we don't have enough space, but maybe we have free space on our
	 * devices to allocate new chunks for relocation, so loop through our
	 * alloc devices and guess if we have enough space.  However, if we
	 * were marked as full, then we know there aren't enough chunks, and we
	 * can just return.
	 */
	ret = -1;
	if (full)
		goto out;
7257

7258 7259 7260 7261 7262 7263 7264 7265 7266 7267 7268
	/*
	 * index:
	 *      0: raid10
	 *      1: raid1
	 *      2: dup
	 *      3: raid0
	 *      4: single
	 */
	index = get_block_group_index(block_group);
	if (index == 0) {
		dev_min = 4;
J
Josef Bacik 已提交
7269 7270
		/* Divide by 2 */
		min_free >>= 1;
7271 7272 7273
	} else if (index == 1) {
		dev_min = 2;
	} else if (index == 2) {
J
Josef Bacik 已提交
7274 7275
		/* Multiply by 2 */
		min_free <<= 1;
7276 7277
	} else if (index == 3) {
		dev_min = fs_devices->rw_devices;
J
Josef Bacik 已提交
7278
		do_div(min_free, dev_min);
7279 7280
	}

7281 7282
	mutex_lock(&root->fs_info->chunk_mutex);
	list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) {
7283
		u64 dev_offset;
7284

7285 7286 7287 7288 7289
		/*
		 * check to make sure we can actually find a chunk with enough
		 * space to fit our block group in.
		 */
		if (device->total_bytes > device->bytes_used + min_free) {
7290
			ret = find_free_dev_extent(device, min_free,
7291
						   &dev_offset, NULL);
7292
			if (!ret)
7293 7294 7295
				dev_nr++;

			if (dev_nr >= dev_min)
7296
				break;
7297

7298
			ret = -1;
7299
		}
7300
	}
7301
	mutex_unlock(&root->fs_info->chunk_mutex);
7302
out:
7303
	btrfs_put_block_group(block_group);
7304 7305 7306
	return ret;
}

7307 7308
static int find_first_block_group(struct btrfs_root *root,
		struct btrfs_path *path, struct btrfs_key *key)
7309
{
7310
	int ret = 0;
7311 7312 7313
	struct btrfs_key found_key;
	struct extent_buffer *leaf;
	int slot;
7314

7315 7316
	ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
	if (ret < 0)
7317 7318
		goto out;

C
Chris Mason 已提交
7319
	while (1) {
7320
		slot = path->slots[0];
7321
		leaf = path->nodes[0];
7322 7323 7324 7325 7326
		if (slot >= btrfs_header_nritems(leaf)) {
			ret = btrfs_next_leaf(root, path);
			if (ret == 0)
				continue;
			if (ret < 0)
7327
				goto out;
7328
			break;
7329
		}
7330
		btrfs_item_key_to_cpu(leaf, &found_key, slot);
7331

7332
		if (found_key.objectid >= key->objectid &&
7333 7334 7335 7336
		    found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
			ret = 0;
			goto out;
		}
7337
		path->slots[0]++;
7338
	}
7339
out:
7340
	return ret;
7341 7342
}

7343 7344 7345 7346 7347 7348 7349 7350 7351 7352 7353 7354 7355 7356 7357 7358 7359 7360 7361 7362 7363 7364 7365 7366 7367 7368 7369 7370 7371 7372 7373 7374 7375 7376
void btrfs_put_block_group_cache(struct btrfs_fs_info *info)
{
	struct btrfs_block_group_cache *block_group;
	u64 last = 0;

	while (1) {
		struct inode *inode;

		block_group = btrfs_lookup_first_block_group(info, last);
		while (block_group) {
			spin_lock(&block_group->lock);
			if (block_group->iref)
				break;
			spin_unlock(&block_group->lock);
			block_group = next_block_group(info->tree_root,
						       block_group);
		}
		if (!block_group) {
			if (last == 0)
				break;
			last = 0;
			continue;
		}

		inode = block_group->inode;
		block_group->iref = 0;
		block_group->inode = NULL;
		spin_unlock(&block_group->lock);
		iput(inode);
		last = block_group->key.objectid + block_group->key.offset;
		btrfs_put_block_group(block_group);
	}
}

Z
Zheng Yan 已提交
7377 7378 7379
int btrfs_free_block_groups(struct btrfs_fs_info *info)
{
	struct btrfs_block_group_cache *block_group;
7380
	struct btrfs_space_info *space_info;
7381
	struct btrfs_caching_control *caching_ctl;
Z
Zheng Yan 已提交
7382 7383
	struct rb_node *n;

7384 7385 7386 7387 7388 7389 7390 7391 7392
	down_write(&info->extent_commit_sem);
	while (!list_empty(&info->caching_block_groups)) {
		caching_ctl = list_entry(info->caching_block_groups.next,
					 struct btrfs_caching_control, list);
		list_del(&caching_ctl->list);
		put_caching_control(caching_ctl);
	}
	up_write(&info->extent_commit_sem);

Z
Zheng Yan 已提交
7393 7394 7395 7396 7397 7398
	spin_lock(&info->block_group_cache_lock);
	while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {
		block_group = rb_entry(n, struct btrfs_block_group_cache,
				       cache_node);
		rb_erase(&block_group->cache_node,
			 &info->block_group_cache_tree);
Y
Yan Zheng 已提交
7399 7400
		spin_unlock(&info->block_group_cache_lock);

7401
		down_write(&block_group->space_info->groups_sem);
Z
Zheng Yan 已提交
7402
		list_del(&block_group->list);
7403
		up_write(&block_group->space_info->groups_sem);
7404

J
Josef Bacik 已提交
7405
		if (block_group->cached == BTRFS_CACHE_STARTED)
7406
			wait_block_group_cache_done(block_group);
J
Josef Bacik 已提交
7407

7408 7409 7410 7411 7412 7413 7414
		/*
		 * We haven't cached this block group, which means we could
		 * possibly have excluded extents on this block group.
		 */
		if (block_group->cached == BTRFS_CACHE_NO)
			free_excluded_extents(info->extent_root, block_group);

J
Josef Bacik 已提交
7415
		btrfs_remove_free_space_cache(block_group);
7416
		btrfs_put_block_group(block_group);
Y
Yan Zheng 已提交
7417 7418

		spin_lock(&info->block_group_cache_lock);
Z
Zheng Yan 已提交
7419 7420
	}
	spin_unlock(&info->block_group_cache_lock);
7421 7422 7423 7424 7425 7426 7427 7428 7429

	/* now that all the block groups are freed, go through and
	 * free all the space_info structs.  This is only called during
	 * the final stages of unmount, and so we know nobody is
	 * using them.  We call synchronize_rcu() once before we start,
	 * just to be on the safe side.
	 */
	synchronize_rcu();

7430 7431
	release_global_block_rsv(info);

7432 7433 7434 7435
	while(!list_empty(&info->space_info)) {
		space_info = list_entry(info->space_info.next,
					struct btrfs_space_info,
					list);
7436
		if (space_info->bytes_pinned > 0 ||
7437 7438
		    space_info->bytes_reserved > 0 ||
		    space_info->bytes_may_use > 0) {
7439 7440 7441
			WARN_ON(1);
			dump_space_info(space_info, 0, 0);
		}
7442 7443 7444
		list_del(&space_info->list);
		kfree(space_info);
	}
Z
Zheng Yan 已提交
7445 7446 7447
	return 0;
}

7448 7449 7450 7451 7452 7453 7454 7455 7456 7457
static void __link_block_group(struct btrfs_space_info *space_info,
			       struct btrfs_block_group_cache *cache)
{
	int index = get_block_group_index(cache);

	down_write(&space_info->groups_sem);
	list_add_tail(&cache->list, &space_info->block_groups[index]);
	up_write(&space_info->groups_sem);
}

C
Chris Mason 已提交
7458 7459 7460 7461 7462
int btrfs_read_block_groups(struct btrfs_root *root)
{
	struct btrfs_path *path;
	int ret;
	struct btrfs_block_group_cache *cache;
C
Chris Mason 已提交
7463
	struct btrfs_fs_info *info = root->fs_info;
7464
	struct btrfs_space_info *space_info;
C
Chris Mason 已提交
7465 7466
	struct btrfs_key key;
	struct btrfs_key found_key;
7467
	struct extent_buffer *leaf;
7468 7469
	int need_clear = 0;
	u64 cache_gen;
7470

C
Chris Mason 已提交
7471
	root = info->extent_root;
C
Chris Mason 已提交
7472
	key.objectid = 0;
7473
	key.offset = 0;
C
Chris Mason 已提交
7474 7475 7476 7477
	btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
J
Josef Bacik 已提交
7478
	path->reada = 1;
C
Chris Mason 已提交
7479

7480
	cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
7481
	if (btrfs_test_opt(root, SPACE_CACHE) &&
7482
	    btrfs_super_generation(root->fs_info->super_copy) != cache_gen)
7483
		need_clear = 1;
7484 7485
	if (btrfs_test_opt(root, CLEAR_CACHE))
		need_clear = 1;
7486

C
Chris Mason 已提交
7487
	while (1) {
7488
		ret = find_first_block_group(root, path, &key);
7489 7490
		if (ret > 0)
			break;
7491 7492
		if (ret != 0)
			goto error;
7493 7494
		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
7495
		cache = kzalloc(sizeof(*cache), GFP_NOFS);
C
Chris Mason 已提交
7496
		if (!cache) {
7497
			ret = -ENOMEM;
7498
			goto error;
C
Chris Mason 已提交
7499
		}
7500 7501 7502 7503 7504 7505 7506
		cache->free_space_ctl = kzalloc(sizeof(*cache->free_space_ctl),
						GFP_NOFS);
		if (!cache->free_space_ctl) {
			kfree(cache);
			ret = -ENOMEM;
			goto error;
		}
C
Chris Mason 已提交
7507

7508
		atomic_set(&cache->count, 1);
7509
		spin_lock_init(&cache->lock);
J
Josef Bacik 已提交
7510
		cache->fs_info = info;
J
Josef Bacik 已提交
7511
		INIT_LIST_HEAD(&cache->list);
7512
		INIT_LIST_HEAD(&cache->cluster_list);
7513

7514 7515 7516
		if (need_clear)
			cache->disk_cache_state = BTRFS_DC_CLEAR;

7517 7518 7519
		read_extent_buffer(leaf, &cache->item,
				   btrfs_item_ptr_offset(leaf, path->slots[0]),
				   sizeof(cache->item));
C
Chris Mason 已提交
7520
		memcpy(&cache->key, &found_key, sizeof(found_key));
7521

C
Chris Mason 已提交
7522
		key.objectid = found_key.objectid + found_key.offset;
7523
		btrfs_release_path(path);
7524
		cache->flags = btrfs_block_group_flags(&cache->item);
J
Josef Bacik 已提交
7525 7526
		cache->sectorsize = root->sectorsize;

7527 7528
		btrfs_init_free_space_ctl(cache);

7529 7530 7531 7532 7533 7534 7535
		/*
		 * We need to exclude the super stripes now so that the space
		 * info has super bytes accounted for, otherwise we'll think
		 * we have more space than we actually do.
		 */
		exclude_super_stripes(root, cache);

J
Josef Bacik 已提交
7536 7537 7538 7539 7540 7541 7542 7543
		/*
		 * check for two cases, either we are full, and therefore
		 * don't need to bother with the caching work since we won't
		 * find any space, or we are empty, and we can just add all
		 * the space in and be done with it.  This saves us _alot_ of
		 * time, particularly in the full case.
		 */
		if (found_key.offset == btrfs_block_group_used(&cache->item)) {
7544
			cache->last_byte_to_unpin = (u64)-1;
J
Josef Bacik 已提交
7545
			cache->cached = BTRFS_CACHE_FINISHED;
7546
			free_excluded_extents(root, cache);
J
Josef Bacik 已提交
7547
		} else if (btrfs_block_group_used(&cache->item) == 0) {
7548
			cache->last_byte_to_unpin = (u64)-1;
J
Josef Bacik 已提交
7549 7550 7551 7552 7553
			cache->cached = BTRFS_CACHE_FINISHED;
			add_new_free_space(cache, root->fs_info,
					   found_key.objectid,
					   found_key.objectid +
					   found_key.offset);
7554
			free_excluded_extents(root, cache);
J
Josef Bacik 已提交
7555
		}
7556

7557 7558 7559 7560 7561
		ret = update_space_info(info, cache->flags, found_key.offset,
					btrfs_block_group_used(&cache->item),
					&space_info);
		BUG_ON(ret);
		cache->space_info = space_info;
7562
		spin_lock(&cache->space_info->lock);
7563
		cache->space_info->bytes_readonly += cache->bytes_super;
7564 7565
		spin_unlock(&cache->space_info->lock);

7566
		__link_block_group(space_info, cache);
J
Josef Bacik 已提交
7567 7568 7569

		ret = btrfs_add_block_group_cache(root->fs_info, cache);
		BUG_ON(ret);
7570 7571

		set_avail_alloc_bits(root->fs_info, cache->flags);
Y
Yan Zheng 已提交
7572
		if (btrfs_chunk_readonly(root, cache->key.objectid))
7573
			set_block_group_ro(cache, 1);
C
Chris Mason 已提交
7574
	}
7575 7576 7577 7578 7579 7580 7581 7582 7583 7584 7585 7586

	list_for_each_entry_rcu(space_info, &root->fs_info->space_info, list) {
		if (!(get_alloc_profile(root, space_info->flags) &
		      (BTRFS_BLOCK_GROUP_RAID10 |
		       BTRFS_BLOCK_GROUP_RAID1 |
		       BTRFS_BLOCK_GROUP_DUP)))
			continue;
		/*
		 * avoid allocating from un-mirrored block group if there are
		 * mirrored block groups.
		 */
		list_for_each_entry(cache, &space_info->block_groups[3], list)
7587
			set_block_group_ro(cache, 1);
7588
		list_for_each_entry(cache, &space_info->block_groups[4], list)
7589
			set_block_group_ro(cache, 1);
C
Chris Mason 已提交
7590
	}
7591 7592

	init_global_block_rsv(info);
7593 7594
	ret = 0;
error:
C
Chris Mason 已提交
7595
	btrfs_free_path(path);
7596
	return ret;
C
Chris Mason 已提交
7597
}
7598 7599 7600

int btrfs_make_block_group(struct btrfs_trans_handle *trans,
			   struct btrfs_root *root, u64 bytes_used,
7601
			   u64 type, u64 chunk_objectid, u64 chunk_offset,
7602 7603 7604 7605 7606 7607 7608 7609
			   u64 size)
{
	int ret;
	struct btrfs_root *extent_root;
	struct btrfs_block_group_cache *cache;

	extent_root = root->fs_info->extent_root;

7610
	root->fs_info->last_trans_log_full_commit = trans->transid;
7611

7612
	cache = kzalloc(sizeof(*cache), GFP_NOFS);
J
Josef Bacik 已提交
7613 7614
	if (!cache)
		return -ENOMEM;
7615 7616 7617 7618 7619 7620
	cache->free_space_ctl = kzalloc(sizeof(*cache->free_space_ctl),
					GFP_NOFS);
	if (!cache->free_space_ctl) {
		kfree(cache);
		return -ENOMEM;
	}
J
Josef Bacik 已提交
7621

7622
	cache->key.objectid = chunk_offset;
7623
	cache->key.offset = size;
7624
	cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
7625
	cache->sectorsize = root->sectorsize;
7626
	cache->fs_info = root->fs_info;
7627

7628
	atomic_set(&cache->count, 1);
7629
	spin_lock_init(&cache->lock);
J
Josef Bacik 已提交
7630
	INIT_LIST_HEAD(&cache->list);
7631
	INIT_LIST_HEAD(&cache->cluster_list);
C
Chris Mason 已提交
7632

7633 7634
	btrfs_init_free_space_ctl(cache);

7635 7636 7637 7638 7639
	btrfs_set_block_group_used(&cache->item, bytes_used);
	btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid);
	cache->flags = type;
	btrfs_set_block_group_flags(&cache->item, type);

7640
	cache->last_byte_to_unpin = (u64)-1;
J
Josef Bacik 已提交
7641
	cache->cached = BTRFS_CACHE_FINISHED;
7642
	exclude_super_stripes(root, cache);
7643

J
Josef Bacik 已提交
7644 7645 7646
	add_new_free_space(cache, root->fs_info, chunk_offset,
			   chunk_offset + size);

7647 7648
	free_excluded_extents(root, cache);

7649 7650 7651
	ret = update_space_info(root->fs_info, cache->flags, size, bytes_used,
				&cache->space_info);
	BUG_ON(ret);
7652
	update_global_block_rsv(root->fs_info);
7653 7654

	spin_lock(&cache->space_info->lock);
7655
	cache->space_info->bytes_readonly += cache->bytes_super;
7656 7657
	spin_unlock(&cache->space_info->lock);

7658
	__link_block_group(cache->space_info, cache);
7659

J
Josef Bacik 已提交
7660 7661
	ret = btrfs_add_block_group_cache(root->fs_info, cache);
	BUG_ON(ret);
7662

7663 7664 7665 7666
	ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item,
				sizeof(cache->item));
	BUG_ON(ret);

C
Chris Mason 已提交
7667
	set_avail_alloc_bits(extent_root->fs_info, type);
7668

7669 7670
	return 0;
}
Z
Zheng Yan 已提交
7671

7672 7673 7674 7675 7676 7677 7678 7679 7680 7681 7682 7683 7684 7685 7686 7687
static void clear_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
{
	u64 extra_flags = flags & BTRFS_BLOCK_GROUP_PROFILE_MASK;

	/* chunk -> extended profile */
	if (extra_flags == 0)
		extra_flags = BTRFS_AVAIL_ALLOC_BIT_SINGLE;

	if (flags & BTRFS_BLOCK_GROUP_DATA)
		fs_info->avail_data_alloc_bits &= ~extra_flags;
	if (flags & BTRFS_BLOCK_GROUP_METADATA)
		fs_info->avail_metadata_alloc_bits &= ~extra_flags;
	if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
		fs_info->avail_system_alloc_bits &= ~extra_flags;
}

Z
Zheng Yan 已提交
7688 7689 7690 7691 7692
int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
			     struct btrfs_root *root, u64 group_start)
{
	struct btrfs_path *path;
	struct btrfs_block_group_cache *block_group;
7693
	struct btrfs_free_cluster *cluster;
7694
	struct btrfs_root *tree_root = root->fs_info->tree_root;
Z
Zheng Yan 已提交
7695
	struct btrfs_key key;
7696
	struct inode *inode;
Z
Zheng Yan 已提交
7697
	int ret;
7698
	int index;
J
Josef Bacik 已提交
7699
	int factor;
Z
Zheng Yan 已提交
7700 7701 7702 7703 7704

	root = root->fs_info->extent_root;

	block_group = btrfs_lookup_block_group(root->fs_info, group_start);
	BUG_ON(!block_group);
Y
Yan Zheng 已提交
7705
	BUG_ON(!block_group->ro);
Z
Zheng Yan 已提交
7706

7707 7708 7709 7710 7711 7712
	/*
	 * Free the reserved super bytes from this block group before
	 * remove it.
	 */
	free_excluded_extents(root, block_group);

Z
Zheng Yan 已提交
7713
	memcpy(&key, &block_group->key, sizeof(key));
7714
	index = get_block_group_index(block_group);
J
Josef Bacik 已提交
7715 7716 7717 7718 7719 7720
	if (block_group->flags & (BTRFS_BLOCK_GROUP_DUP |
				  BTRFS_BLOCK_GROUP_RAID1 |
				  BTRFS_BLOCK_GROUP_RAID10))
		factor = 2;
	else
		factor = 1;
Z
Zheng Yan 已提交
7721

7722 7723 7724 7725 7726 7727 7728 7729 7730 7731 7732 7733 7734 7735 7736
	/* make sure this block group isn't part of an allocation cluster */
	cluster = &root->fs_info->data_alloc_cluster;
	spin_lock(&cluster->refill_lock);
	btrfs_return_cluster_to_free_space(block_group, cluster);
	spin_unlock(&cluster->refill_lock);

	/*
	 * make sure this block group isn't part of a metadata
	 * allocation cluster
	 */
	cluster = &root->fs_info->meta_alloc_cluster;
	spin_lock(&cluster->refill_lock);
	btrfs_return_cluster_to_free_space(block_group, cluster);
	spin_unlock(&cluster->refill_lock);

Z
Zheng Yan 已提交
7737
	path = btrfs_alloc_path();
7738 7739 7740 7741
	if (!path) {
		ret = -ENOMEM;
		goto out;
	}
Z
Zheng Yan 已提交
7742

7743
	inode = lookup_free_space_inode(tree_root, block_group, path);
7744
	if (!IS_ERR(inode)) {
7745 7746
		ret = btrfs_orphan_add(trans, inode);
		BUG_ON(ret);
7747 7748 7749 7750 7751 7752 7753 7754 7755 7756 7757 7758
		clear_nlink(inode);
		/* One for the block groups ref */
		spin_lock(&block_group->lock);
		if (block_group->iref) {
			block_group->iref = 0;
			block_group->inode = NULL;
			spin_unlock(&block_group->lock);
			iput(inode);
		} else {
			spin_unlock(&block_group->lock);
		}
		/* One for our lookup ref */
7759
		btrfs_add_delayed_iput(inode);
7760 7761 7762 7763 7764 7765 7766 7767 7768 7769
	}

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

	ret = btrfs_search_slot(trans, tree_root, &key, path, -1, 1);
	if (ret < 0)
		goto out;
	if (ret > 0)
7770
		btrfs_release_path(path);
7771 7772 7773 7774
	if (ret == 0) {
		ret = btrfs_del_item(trans, tree_root, path);
		if (ret)
			goto out;
7775
		btrfs_release_path(path);
7776 7777
	}

7778
	spin_lock(&root->fs_info->block_group_cache_lock);
Z
Zheng Yan 已提交
7779 7780
	rb_erase(&block_group->cache_node,
		 &root->fs_info->block_group_cache_tree);
7781
	spin_unlock(&root->fs_info->block_group_cache_lock);
J
Josef Bacik 已提交
7782

7783
	down_write(&block_group->space_info->groups_sem);
7784 7785 7786 7787 7788
	/*
	 * we must use list_del_init so people can check to see if they
	 * are still on the list after taking the semaphore
	 */
	list_del_init(&block_group->list);
7789 7790
	if (list_empty(&block_group->space_info->block_groups[index]))
		clear_avail_alloc_bits(root->fs_info, block_group->flags);
7791
	up_write(&block_group->space_info->groups_sem);
Z
Zheng Yan 已提交
7792

J
Josef Bacik 已提交
7793
	if (block_group->cached == BTRFS_CACHE_STARTED)
7794
		wait_block_group_cache_done(block_group);
J
Josef Bacik 已提交
7795 7796 7797

	btrfs_remove_free_space_cache(block_group);

Y
Yan Zheng 已提交
7798 7799 7800
	spin_lock(&block_group->space_info->lock);
	block_group->space_info->total_bytes -= block_group->key.offset;
	block_group->space_info->bytes_readonly -= block_group->key.offset;
J
Josef Bacik 已提交
7801
	block_group->space_info->disk_total -= block_group->key.offset * factor;
Y
Yan Zheng 已提交
7802
	spin_unlock(&block_group->space_info->lock);
7803

7804 7805
	memcpy(&key, &block_group->key, sizeof(key));

7806
	btrfs_clear_space_info_full(root->fs_info);
Y
Yan Zheng 已提交
7807

7808 7809
	btrfs_put_block_group(block_group);
	btrfs_put_block_group(block_group);
Z
Zheng Yan 已提交
7810 7811 7812 7813 7814 7815 7816 7817 7818 7819 7820 7821

	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
	if (ret > 0)
		ret = -EIO;
	if (ret < 0)
		goto out;

	ret = btrfs_del_item(trans, root, path);
out:
	btrfs_free_path(path);
	return ret;
}
L
liubo 已提交
7822

7823 7824 7825
int btrfs_init_space_info(struct btrfs_fs_info *fs_info)
{
	struct btrfs_space_info *space_info;
7826 7827 7828 7829
	struct btrfs_super_block *disk_super;
	u64 features;
	u64 flags;
	int mixed = 0;
7830 7831
	int ret;

7832
	disk_super = fs_info->super_copy;
7833 7834
	if (!btrfs_super_root(disk_super))
		return 1;
7835

7836 7837 7838
	features = btrfs_super_incompat_flags(disk_super);
	if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS)
		mixed = 1;
7839

7840 7841
	flags = BTRFS_BLOCK_GROUP_SYSTEM;
	ret = update_space_info(fs_info, flags, 0, 0, &space_info);
7842
	if (ret)
7843
		goto out;
7844

7845 7846 7847 7848 7849 7850 7851 7852 7853 7854 7855 7856 7857
	if (mixed) {
		flags = BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA;
		ret = update_space_info(fs_info, flags, 0, 0, &space_info);
	} else {
		flags = BTRFS_BLOCK_GROUP_METADATA;
		ret = update_space_info(fs_info, flags, 0, 0, &space_info);
		if (ret)
			goto out;

		flags = BTRFS_BLOCK_GROUP_DATA;
		ret = update_space_info(fs_info, flags, 0, 0, &space_info);
	}
out:
7858 7859 7860
	return ret;
}

L
liubo 已提交
7861 7862 7863 7864 7865 7866
int btrfs_error_unpin_extent_range(struct btrfs_root *root, u64 start, u64 end)
{
	return unpin_extent_range(root, start, end);
}

int btrfs_error_discard_extent(struct btrfs_root *root, u64 bytenr,
7867
			       u64 num_bytes, u64 *actual_bytes)
L
liubo 已提交
7868
{
7869
	return btrfs_discard_extent(root, bytenr, num_bytes, actual_bytes);
L
liubo 已提交
7870
}
7871 7872 7873 7874 7875 7876 7877 7878 7879 7880 7881 7882 7883 7884 7885 7886 7887 7888 7889 7890 7891 7892 7893 7894 7895 7896 7897 7898 7899 7900 7901 7902 7903 7904 7905 7906 7907 7908 7909 7910 7911 7912 7913 7914 7915 7916 7917 7918

int btrfs_trim_fs(struct btrfs_root *root, struct fstrim_range *range)
{
	struct btrfs_fs_info *fs_info = root->fs_info;
	struct btrfs_block_group_cache *cache = NULL;
	u64 group_trimmed;
	u64 start;
	u64 end;
	u64 trimmed = 0;
	int ret = 0;

	cache = btrfs_lookup_block_group(fs_info, range->start);

	while (cache) {
		if (cache->key.objectid >= (range->start + range->len)) {
			btrfs_put_block_group(cache);
			break;
		}

		start = max(range->start, cache->key.objectid);
		end = min(range->start + range->len,
				cache->key.objectid + cache->key.offset);

		if (end - start >= range->minlen) {
			if (!block_group_cache_done(cache)) {
				ret = cache_block_group(cache, NULL, root, 0);
				if (!ret)
					wait_block_group_cache_done(cache);
			}
			ret = btrfs_trim_block_group(cache,
						     &group_trimmed,
						     start,
						     end,
						     range->minlen);

			trimmed += group_trimmed;
			if (ret) {
				btrfs_put_block_group(cache);
				break;
			}
		}

		cache = next_block_group(fs_info->tree_root, cache);
	}

	range->len = trimmed;
	return ret;
}