extent-tree.c 196.6 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>
C
Chris Mason 已提交
26
#include "compat.h"
27
#include "hash.h"
28 29 30
#include "ctree.h"
#include "disk-io.h"
#include "print-tree.h"
31
#include "transaction.h"
32
#include "volumes.h"
33
#include "locking.h"
34
#include "free-space-cache.h"
35

36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54
/* 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,
};

55 56 57 58 59 60 61 62 63 64 65 66 67 68 69
/*
 * 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,
};

70 71
static int update_block_group(struct btrfs_trans_handle *trans,
			      struct btrfs_root *root,
72
			      u64 bytenr, u64 num_bytes, int alloc);
73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91
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
92 93 94
static int do_chunk_alloc(struct btrfs_trans_handle *trans,
			  struct btrfs_root *extent_root, u64 alloc_bytes,
			  u64 flags, int force);
95 96
static int find_next_key(struct btrfs_path *path, int level,
			 struct btrfs_key *key);
J
Josef Bacik 已提交
97 98
static void dump_space_info(struct btrfs_space_info *info, u64 bytes,
			    int dump_block_groups);
99 100
static int btrfs_update_reserved_bytes(struct btrfs_block_group_cache *cache,
				       u64 num_bytes, int reserve);
J
Josef Bacik 已提交
101

J
Josef Bacik 已提交
102 103 104 105 106 107 108
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
109 110 111 112 113
static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
{
	return (cache->flags & bits) == bits;
}

114
static void btrfs_get_block_group(struct btrfs_block_group_cache *cache)
115 116 117 118 119 120
{
	atomic_inc(&cache->count);
}

void btrfs_put_block_group(struct btrfs_block_group_cache *cache)
{
121 122 123
	if (atomic_dec_and_test(&cache->count)) {
		WARN_ON(cache->pinned > 0);
		WARN_ON(cache->reserved > 0);
124
		kfree(cache->free_space_ctl);
125
		kfree(cache);
126
	}
127 128
}

J
Josef Bacik 已提交
129 130 131 132
/*
 * this adds the block group to the fs_info rb tree for the block group
 * cache
 */
133
static int btrfs_add_block_group_cache(struct btrfs_fs_info *info,
J
Josef Bacik 已提交
134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200
				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;
		}
	}
201
	if (ret)
202
		btrfs_get_block_group(ret);
J
Josef Bacik 已提交
203 204 205 206 207
	spin_unlock(&info->block_group_cache_lock);

	return ret;
}

208 209
static int add_excluded_extent(struct btrfs_root *root,
			       u64 start, u64 num_bytes)
J
Josef Bacik 已提交
210
{
211 212 213 214 215 216 217
	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
218

219 220 221 222
static void free_excluded_extents(struct btrfs_root *root,
				  struct btrfs_block_group_cache *cache)
{
	u64 start, end;
J
Josef Bacik 已提交
223

224 225 226 227 228 229 230
	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
231 232
}

233 234
static int exclude_super_stripes(struct btrfs_root *root,
				 struct btrfs_block_group_cache *cache)
J
Josef Bacik 已提交
235 236 237 238 239 240
{
	u64 bytenr;
	u64 *logical;
	int stripe_len;
	int i, nr, ret;

241 242 243 244 245 246 247 248
	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
249 250 251 252 253 254
	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);
255

J
Josef Bacik 已提交
256
		while (nr--) {
257
			cache->bytes_super += stripe_len;
258 259 260
			ret = add_excluded_extent(root, logical[nr],
						  stripe_len);
			BUG_ON(ret);
J
Josef Bacik 已提交
261
		}
262

J
Josef Bacik 已提交
263 264 265 266 267
		kfree(logical);
	}
	return 0;
}

268 269 270 271 272 273 274 275 276 277 278
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;
	}

279 280 281
	/* We're loading it the fast way, so we don't have a caching_ctl. */
	if (!cache->caching_ctl) {
		spin_unlock(&cache->lock);
282 283 284 285 286 287 288 289 290 291 292 293 294 295 296
		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
297 298 299 300 301
/*
 * 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
302
static u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
J
Josef Bacik 已提交
303 304
			      struct btrfs_fs_info *info, u64 start, u64 end)
{
J
Josef Bacik 已提交
305
	u64 extent_start, extent_end, size, total_added = 0;
J
Josef Bacik 已提交
306 307 308
	int ret;

	while (start < end) {
309
		ret = find_first_extent_bit(info->pinned_extents, start,
J
Josef Bacik 已提交
310
					    &extent_start, &extent_end,
311
					    EXTENT_DIRTY | EXTENT_UPTODATE);
J
Josef Bacik 已提交
312 313 314
		if (ret)
			break;

315
		if (extent_start <= start) {
J
Josef Bacik 已提交
316 317 318
			start = extent_end + 1;
		} else if (extent_start > start && extent_start < end) {
			size = extent_start - start;
J
Josef Bacik 已提交
319
			total_added += size;
320 321
			ret = btrfs_add_free_space(block_group, start,
						   size);
J
Josef Bacik 已提交
322 323 324 325 326 327 328 329 330
			BUG_ON(ret);
			start = extent_end + 1;
		} else {
			break;
		}
	}

	if (start < end) {
		size = end - start;
J
Josef Bacik 已提交
331
		total_added += size;
332
		ret = btrfs_add_free_space(block_group, start, size);
J
Josef Bacik 已提交
333 334 335
		BUG_ON(ret);
	}

J
Josef Bacik 已提交
336
	return total_added;
J
Josef Bacik 已提交
337 338
}

339
static noinline void caching_thread(struct btrfs_work *work)
340
{
341 342 343 344
	struct btrfs_block_group_cache *block_group;
	struct btrfs_fs_info *fs_info;
	struct btrfs_caching_control *caching_ctl;
	struct btrfs_root *extent_root;
345
	struct btrfs_path *path;
346
	struct extent_buffer *leaf;
347
	struct btrfs_key key;
J
Josef Bacik 已提交
348
	u64 total_found = 0;
349 350 351
	u64 last = 0;
	u32 nritems;
	int ret = 0;
352

353 354 355 356 357
	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;

358 359
	path = btrfs_alloc_path();
	if (!path)
360
		goto out;
361

J
Josef Bacik 已提交
362
	last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
363

364
	/*
J
Josef Bacik 已提交
365 366 367 368
	 * 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
369 370
	 */
	path->skip_locking = 1;
J
Josef Bacik 已提交
371
	path->search_commit_root = 1;
J
Josef Bacik 已提交
372
	path->reada = 1;
J
Josef Bacik 已提交
373

Y
Yan Zheng 已提交
374
	key.objectid = last;
375
	key.offset = 0;
376
	key.type = BTRFS_EXTENT_ITEM_KEY;
377
again:
378
	mutex_lock(&caching_ctl->mutex);
379 380 381
	/* need to make sure the commit_root doesn't disappear */
	down_read(&fs_info->extent_commit_sem);

382
	ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
383
	if (ret < 0)
384
		goto err;
Y
Yan Zheng 已提交
385

386 387 388
	leaf = path->nodes[0];
	nritems = btrfs_header_nritems(leaf);

C
Chris Mason 已提交
389
	while (1) {
390
		if (btrfs_fs_closing(fs_info) > 1) {
391
			last = (u64)-1;
J
Josef Bacik 已提交
392
			break;
393
		}
J
Josef Bacik 已提交
394

395 396 397 398 399
		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)
400
				break;
J
Josef Bacik 已提交
401

402 403 404
			if (need_resched() ||
			    btrfs_next_leaf(extent_root, path)) {
				caching_ctl->progress = last;
C
Chris Mason 已提交
405
				btrfs_release_path(path);
406 407
				up_read(&fs_info->extent_commit_sem);
				mutex_unlock(&caching_ctl->mutex);
408
				cond_resched();
409 410 411 412 413
				goto again;
			}
			leaf = path->nodes[0];
			nritems = btrfs_header_nritems(leaf);
			continue;
414
		}
J
Josef Bacik 已提交
415

416 417
		if (key.objectid < block_group->key.objectid) {
			path->slots[0]++;
J
Josef Bacik 已提交
418
			continue;
419
		}
J
Josef Bacik 已提交
420

421
		if (key.objectid >= block_group->key.objectid +
J
Josef Bacik 已提交
422
		    block_group->key.offset)
423
			break;
424

425
		if (key.type == BTRFS_EXTENT_ITEM_KEY) {
J
Josef Bacik 已提交
426 427 428
			total_found += add_new_free_space(block_group,
							  fs_info, last,
							  key.objectid);
429
			last = key.objectid + key.offset;
J
Josef Bacik 已提交
430

431 432 433 434
			if (total_found > (1024 * 1024 * 2)) {
				total_found = 0;
				wake_up(&caching_ctl->wait);
			}
J
Josef Bacik 已提交
435
		}
436 437
		path->slots[0]++;
	}
J
Josef Bacik 已提交
438
	ret = 0;
439

J
Josef Bacik 已提交
440 441 442
	total_found += add_new_free_space(block_group, fs_info, last,
					  block_group->key.objectid +
					  block_group->key.offset);
443
	caching_ctl->progress = (u64)-1;
J
Josef Bacik 已提交
444 445

	spin_lock(&block_group->lock);
446
	block_group->caching_ctl = NULL;
J
Josef Bacik 已提交
447 448
	block_group->cached = BTRFS_CACHE_FINISHED;
	spin_unlock(&block_group->lock);
J
Josef Bacik 已提交
449

450
err:
451
	btrfs_free_path(path);
452
	up_read(&fs_info->extent_commit_sem);
J
Josef Bacik 已提交
453

454 455 456
	free_excluded_extents(extent_root, block_group);

	mutex_unlock(&caching_ctl->mutex);
457
out:
458 459 460
	wake_up(&caching_ctl->wait);

	put_caching_control(caching_ctl);
461
	btrfs_put_block_group(block_group);
J
Josef Bacik 已提交
462 463
}

464 465
static int cache_block_group(struct btrfs_block_group_cache *cache,
			     struct btrfs_trans_handle *trans,
466
			     struct btrfs_root *root,
467
			     int load_cache_only)
J
Josef Bacik 已提交
468
{
469 470
	struct btrfs_fs_info *fs_info = cache->fs_info;
	struct btrfs_caching_control *caching_ctl;
J
Josef Bacik 已提交
471 472
	int ret = 0;

473 474 475 476
	smp_mb();
	if (cache->cached != BTRFS_CACHE_NO)
		return 0;

477 478
	/*
	 * We can't do the read from on-disk cache during a commit since we need
479 480 481
	 * 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.
482
	 */
483
	if (trans && (!trans->transaction->in_commit) &&
484 485
	    (root && root != root->fs_info->tree_root) &&
	    btrfs_test_opt(root, SPACE_CACHE)) {
486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503
		spin_lock(&cache->lock);
		if (cache->cached != BTRFS_CACHE_NO) {
			spin_unlock(&cache->lock);
			return 0;
		}
		cache->cached = BTRFS_CACHE_STARTED;
		spin_unlock(&cache->lock);

		ret = load_free_space_cache(fs_info, cache);

		spin_lock(&cache->lock);
		if (ret == 1) {
			cache->cached = BTRFS_CACHE_FINISHED;
			cache->last_byte_to_unpin = (u64)-1;
		} else {
			cache->cached = BTRFS_CACHE_NO;
		}
		spin_unlock(&cache->lock);
504 505
		if (ret == 1) {
			free_excluded_extents(fs_info->extent_root, cache);
506
			return 0;
507
		}
508 509 510 511 512
	}

	if (load_cache_only)
		return 0;

513
	caching_ctl = kzalloc(sizeof(*caching_ctl), GFP_NOFS);
514 515 516 517 518 519 520 521 522
	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;
	/* one for caching kthread, one for caching block group list */
	atomic_set(&caching_ctl->count, 2);
523
	caching_ctl->work.func = caching_thread;
524

J
Josef Bacik 已提交
525 526 527
	spin_lock(&cache->lock);
	if (cache->cached != BTRFS_CACHE_NO) {
		spin_unlock(&cache->lock);
528 529
		kfree(caching_ctl);
		return 0;
J
Josef Bacik 已提交
530
	}
531
	cache->caching_ctl = caching_ctl;
J
Josef Bacik 已提交
532 533 534
	cache->cached = BTRFS_CACHE_STARTED;
	spin_unlock(&cache->lock);

535 536 537 538
	down_write(&fs_info->extent_commit_sem);
	list_add_tail(&caching_ctl->list, &fs_info->caching_block_groups);
	up_write(&fs_info->extent_commit_sem);

539
	btrfs_get_block_group(cache);
540

541
	btrfs_queue_worker(&fs_info->caching_workers, &caching_ctl->work);
J
Josef Bacik 已提交
542

543
	return ret;
544 545
}

J
Josef Bacik 已提交
546 547 548
/*
 * return the block group that starts at or after bytenr
 */
C
Chris Mason 已提交
549 550
static struct btrfs_block_group_cache *
btrfs_lookup_first_block_group(struct btrfs_fs_info *info, u64 bytenr)
C
Chris Mason 已提交
551
{
J
Josef Bacik 已提交
552
	struct btrfs_block_group_cache *cache;
C
Chris Mason 已提交
553

J
Josef Bacik 已提交
554
	cache = block_group_cache_tree_search(info, bytenr, 0);
C
Chris Mason 已提交
555

J
Josef Bacik 已提交
556
	return cache;
C
Chris Mason 已提交
557 558
}

J
Josef Bacik 已提交
559
/*
560
 * return the block group that contains the given bytenr
J
Josef Bacik 已提交
561
 */
C
Chris Mason 已提交
562 563 564
struct btrfs_block_group_cache *btrfs_lookup_block_group(
						 struct btrfs_fs_info *info,
						 u64 bytenr)
C
Chris Mason 已提交
565
{
J
Josef Bacik 已提交
566
	struct btrfs_block_group_cache *cache;
C
Chris Mason 已提交
567

J
Josef Bacik 已提交
568
	cache = block_group_cache_tree_search(info, bytenr, 1);
569

J
Josef Bacik 已提交
570
	return cache;
C
Chris Mason 已提交
571
}
572

J
Josef Bacik 已提交
573 574
static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info,
						  u64 flags)
575
{
J
Josef Bacik 已提交
576 577
	struct list_head *head = &info->space_info;
	struct btrfs_space_info *found;
578

579 580 581
	flags &= BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_SYSTEM |
		 BTRFS_BLOCK_GROUP_METADATA;

582 583
	rcu_read_lock();
	list_for_each_entry_rcu(found, head, list) {
584
		if (found->flags & flags) {
585
			rcu_read_unlock();
J
Josef Bacik 已提交
586
			return found;
587
		}
J
Josef Bacik 已提交
588
	}
589
	rcu_read_unlock();
J
Josef Bacik 已提交
590
	return NULL;
591 592
}

593 594 595 596 597 598 599 600 601 602 603 604 605 606 607
/*
 * 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();
}

608 609 610 611 612 613 614 615 616
static u64 div_factor(u64 num, int factor)
{
	if (factor == 10)
		return num;
	num *= factor;
	do_div(num, 10);
	return num;
}

617 618 619 620 621 622 623 624 625
static u64 div_factor_fine(u64 num, int factor)
{
	if (factor == 100)
		return num;
	num *= factor;
	do_div(num, 100);
	return num;
}

626 627
u64 btrfs_find_block_group(struct btrfs_root *root,
			   u64 search_start, u64 search_hint, int owner)
C
Chris Mason 已提交
628
{
629
	struct btrfs_block_group_cache *cache;
C
Chris Mason 已提交
630
	u64 used;
631 632
	u64 last = max(search_hint, search_start);
	u64 group_start = 0;
633
	int full_search = 0;
634
	int factor = 9;
C
Chris Mason 已提交
635
	int wrapped = 0;
636
again:
637 638
	while (1) {
		cache = btrfs_lookup_first_block_group(root->fs_info, last);
J
Josef Bacik 已提交
639 640
		if (!cache)
			break;
641

642
		spin_lock(&cache->lock);
643 644 645
		last = cache->key.objectid + cache->key.offset;
		used = btrfs_block_group_used(&cache->item);

646 647
		if ((full_search || !cache->ro) &&
		    block_group_bits(cache, BTRFS_BLOCK_GROUP_METADATA)) {
648
			if (used + cache->pinned + cache->reserved <
649 650
			    div_factor(cache->key.offset, factor)) {
				group_start = cache->key.objectid;
651
				spin_unlock(&cache->lock);
652
				btrfs_put_block_group(cache);
653 654
				goto found;
			}
655
		}
656
		spin_unlock(&cache->lock);
657
		btrfs_put_block_group(cache);
658
		cond_resched();
C
Chris Mason 已提交
659
	}
C
Chris Mason 已提交
660 661 662 663 664 665
	if (!wrapped) {
		last = search_start;
		wrapped = 1;
		goto again;
	}
	if (!full_search && factor < 10) {
C
Chris Mason 已提交
666
		last = search_start;
667
		full_search = 1;
C
Chris Mason 已提交
668
		factor = 10;
669 670
		goto again;
	}
C
Chris Mason 已提交
671
found:
672
	return group_start;
673
}
J
Josef Bacik 已提交
674

675
/* simple helper to search for an existing extent at a given offset */
Z
Zheng Yan 已提交
676
int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len)
677 678 679
{
	int ret;
	struct btrfs_key key;
Z
Zheng Yan 已提交
680
	struct btrfs_path *path;
681

Z
Zheng Yan 已提交
682
	path = btrfs_alloc_path();
683 684 685
	if (!path)
		return -ENOMEM;

686 687 688 689 690
	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
691
	btrfs_free_path(path);
692 693 694
	return ret;
}

695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774
/*
 * 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);

775
			btrfs_release_path(path);
776

777 778 779 780
			/*
			 * Mutex was contended, block until it's released and try
			 * again
			 */
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
			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;
}

806 807 808 809 810 811 812 813 814 815 816 817 818 819
/*
 * 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.
 *
820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864
 * 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.
 *
865 866 867
 * File extents can be referenced by:
 *
 * - multiple snapshots, subvolumes, or different generations in one subvol
Z
Zheng Yan 已提交
868
 * - different files inside a single subvolume
869 870
 * - different offsets inside a file (bookend extents in file.c)
 *
871
 * The extent ref structure for the implicit back refs has fields for:
872 873 874
 *
 * - Objectid of the subvolume root
 * - objectid of the file holding the reference
875 876
 * - original offset in the file
 * - how many bookend extents
877
 *
878 879
 * The key offset for the implicit back refs is hash of the first
 * three fields.
880
 *
881
 * The extent ref structure for the full back refs has field for:
882
 *
883
 * - number of pointers in the tree leaf
884
 *
885 886
 * The key offset for the implicit back refs is the first byte of
 * the tree leaf
887
 *
888 889
 * When a file extent is allocated, The implicit back refs is used.
 * the fields are filled in:
890
 *
891
 *     (root_key.objectid, inode objectid, offset in file, 1)
892
 *
893 894
 * When a file extent is removed file truncation, we find the
 * corresponding implicit back refs and check the following fields:
895
 *
896
 *     (btrfs_header_owner(leaf), inode objectid, offset in file)
897
 *
898
 * Btree extents can be referenced by:
899
 *
900
 * - Different subvolumes
901
 *
902 903 904 905
 * 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.
906
 *
907 908 909
 * 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.
910
 */
Z
Zheng Yan 已提交
911

912 913 914 915 916
#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)
917
{
918 919 920 921 922
	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;
923
	struct btrfs_key key;
924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958
	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;
		}
	}
959
	btrfs_release_path(path);
960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000

	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);
1001
	high_crc = crc32c(high_crc, &lenum, sizeof(lenum));
1002
	lenum = cpu_to_le64(owner);
1003
	low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
1004
	lenum = cpu_to_le64(offset);
1005
	low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
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

	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
1038
	struct extent_buffer *leaf;
1039
	u32 nritems;
1040
	int ret;
1041 1042
	int recow;
	int err = -ENOENT;
1043

Z
Zheng Yan 已提交
1044
	key.objectid = bytenr;
1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059
	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
1060

1061 1062 1063 1064 1065
	if (parent) {
		if (!ret)
			return 0;
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
		key.type = BTRFS_EXTENT_REF_V0_KEY;
1066
		btrfs_release_path(path);
1067 1068 1069 1070 1071 1072 1073 1074 1075
		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
1076 1077 1078
	}

	leaf = path->nodes[0];
1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103
	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) {
1104
				btrfs_release_path(path);
1105 1106 1107 1108 1109 1110
				goto again;
			}
			err = 0;
			break;
		}
		path->slots[0]++;
Z
Zheng Yan 已提交
1111
	}
1112 1113
fail:
	return err;
Z
Zheng Yan 已提交
1114 1115
}

1116 1117 1118 1119 1120 1121
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
1122 1123 1124
{
	struct btrfs_key key;
	struct extent_buffer *leaf;
1125
	u32 size;
Z
Zheng Yan 已提交
1126 1127
	u32 num_refs;
	int ret;
1128 1129

	key.objectid = bytenr;
1130 1131 1132 1133 1134 1135 1136 1137 1138 1139
	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);
	}
1140

1141 1142 1143 1144 1145 1146 1147
	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
1148
		ref = btrfs_item_ptr(leaf, path->slots[0],
1149 1150 1151 1152 1153 1154 1155
				     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
1156
		}
1157 1158 1159 1160 1161 1162 1163 1164
	} 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;
1165
			btrfs_release_path(path);
1166 1167 1168 1169 1170
			key.offset++;
			ret = btrfs_insert_empty_item(trans, root, path, &key,
						      size);
			if (ret && ret != -EEXIST)
				goto fail;
Z
Zheng Yan 已提交
1171

1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185
			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
1186 1187
		}
	}
1188 1189 1190
	btrfs_mark_buffer_dirty(leaf);
	ret = 0;
fail:
1191
	btrfs_release_path(path);
1192
	return ret;
1193 1194
}

1195 1196 1197 1198
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
1199
{
1200 1201 1202
	struct btrfs_key key;
	struct btrfs_extent_data_ref *ref1 = NULL;
	struct btrfs_shared_data_ref *ref2 = NULL;
Z
Zheng Yan 已提交
1203
	struct extent_buffer *leaf;
1204
	u32 num_refs = 0;
Z
Zheng Yan 已提交
1205 1206 1207
	int ret = 0;

	leaf = path->nodes[0];
1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228
	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();
	}

1229 1230
	BUG_ON(num_refs < refs_to_drop);
	num_refs -= refs_to_drop;
1231

Z
Zheng Yan 已提交
1232 1233 1234
	if (num_refs == 0) {
		ret = btrfs_del_item(trans, root, path);
	} else {
1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246
		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
1247 1248 1249 1250 1251
		btrfs_mark_buffer_dirty(leaf);
	}
	return ret;
}

1252 1253 1254
static noinline u32 extent_data_ref_count(struct btrfs_root *root,
					  struct btrfs_path *path,
					  struct btrfs_extent_inline_ref *iref)
1255
{
1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286
	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
1287
#endif
1288 1289 1290 1291 1292
	} else {
		WARN_ON(1);
	}
	return num_refs;
}
1293

1294 1295 1296 1297 1298
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)
1299
{
1300
	struct btrfs_key key;
1301 1302
	int ret;

1303 1304 1305 1306 1307 1308 1309
	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;
1310 1311
	}

1312 1313 1314 1315 1316
	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) {
1317
		btrfs_release_path(path);
1318 1319 1320 1321 1322
		key.type = BTRFS_EXTENT_REF_V0_KEY;
		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
		if (ret > 0)
			ret = -ENOENT;
	}
1323
#endif
1324
	return ret;
1325 1326
}

1327 1328 1329 1330 1331
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
1332
{
1333
	struct btrfs_key key;
Z
Zheng Yan 已提交
1334 1335
	int ret;

1336 1337 1338 1339 1340 1341 1342 1343 1344 1345
	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);
1346
	btrfs_release_path(path);
Z
Zheng Yan 已提交
1347 1348 1349
	return ret;
}

1350
static inline int extent_ref_type(u64 parent, u64 owner)
Z
Zheng Yan 已提交
1351
{
1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364
	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
1365
}
1366

1367 1368
static int find_next_key(struct btrfs_path *path, int level,
			 struct btrfs_key *key)
1369

C
Chris Mason 已提交
1370
{
1371
	for (; level < BTRFS_MAX_LEVEL; level++) {
1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386
		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
1387

1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422
/*
 * 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;
1423

1424
	key.objectid = bytenr;
Z
Zheng Yan 已提交
1425
	key.type = BTRFS_EXTENT_ITEM_KEY;
1426
	key.offset = num_bytes;
Z
Zheng Yan 已提交
1427

1428 1429 1430
	want = extent_ref_type(parent, owner);
	if (insert) {
		extra_size = btrfs_extent_inline_ref_size(want);
1431
		path->keep_locks = 1;
1432 1433 1434
	} else
		extra_size = -1;
	ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1);
1435
	if (ret < 0) {
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 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 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
		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
		 */
1533 1534
		if (find_next_key(path, 0, &key) == 0 &&
		    key.objectid == bytenr &&
1535
		    key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) {
1536 1537 1538 1539 1540 1541
			err = -EAGAIN;
			goto out;
		}
	}
	*ref_ret = (struct btrfs_extent_inline_ref *)ptr;
out:
1542
	if (insert) {
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 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 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
		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)
1629
		return ret;
1630

1631
	btrfs_release_path(path);
1632 1633 1634 1635 1636 1637 1638 1639
	*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);
1640
	}
1641 1642
	return ret;
}
Z
Zheng Yan 已提交
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 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686
/*
 * 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);
1687
	}
Z
Zheng Yan 已提交
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 1728 1729 1730 1731 1732 1733 1734 1735
	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);
1736
	}
1737 1738
	return ret;
}
Z
Zheng Yan 已提交
1739

1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757
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;
}
1758

1759 1760 1761 1762 1763 1764 1765
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;
1766

1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778
	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;
}

1779
static int btrfs_issue_discard(struct block_device *bdev,
1780 1781
				u64 start, u64 len)
{
1782
	return blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_NOFS, 0);
1783 1784 1785
}

static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
1786
				u64 num_bytes, u64 *actual_bytes)
1787 1788
{
	int ret;
1789
	u64 discarded_bytes = 0;
1790 1791
	struct btrfs_multi_bio *multi = NULL;

C
Christoph Hellwig 已提交
1792

1793
	/* Tell the block device(s) that the sectors can be discarded */
1794 1795
	ret = btrfs_map_block(&root->fs_info->mapping_tree, REQ_DISCARD,
			      bytenr, &num_bytes, &multi, 0);
1796 1797 1798 1799 1800 1801
	if (!ret) {
		struct btrfs_bio_stripe *stripe = multi->stripes;
		int i;


		for (i = 0; i < multi->num_stripes; i++, stripe++) {
1802 1803 1804
			if (!stripe->dev->can_discard)
				continue;

1805 1806 1807 1808 1809 1810 1811
			ret = btrfs_issue_discard(stripe->dev->bdev,
						  stripe->physical,
						  stripe->length);
			if (!ret)
				discarded_bytes += stripe->length;
			else if (ret != -EOPNOTSUPP)
				break;
1812 1813 1814 1815 1816 1817 1818

			/*
			 * 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;
1819 1820 1821
		}
		kfree(multi);
	}
1822 1823 1824 1825

	if (actual_bytes)
		*actual_bytes = discarded_bytes;

1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889

	return ret;
}

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 ret;
	BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID &&
	       root_objectid == BTRFS_TREE_LOG_OBJECTID);

	if (owner < BTRFS_FIRST_FREE_OBJECTID) {
		ret = btrfs_add_delayed_tree_ref(trans, bytenr, num_bytes,
					parent, root_objectid, (int)owner,
					BTRFS_ADD_DELAYED_REF, NULL);
	} else {
		ret = btrfs_add_delayed_data_ref(trans, bytenr, num_bytes,
					parent, root_objectid, owner, offset,
					BTRFS_ADD_DELAYED_REF, NULL);
	}
	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);
1890

1891
	btrfs_mark_buffer_dirty(leaf);
1892
	btrfs_release_path(path);
1893 1894

	path->reada = 1;
1895 1896
	path->leave_spinning = 1;

1897 1898
	/* now insert the actual backref */
	ret = insert_extent_backref(trans, root->fs_info->extent_root,
1899 1900
				    path, bytenr, parent, root_objectid,
				    owner, offset, refs_to_add);
1901
	BUG_ON(ret);
1902
out:
1903
	btrfs_free_path(path);
1904
	return err;
1905 1906
}

1907 1908 1909 1910 1911
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)
1912
{
1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984
	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;
1985
	int ret;
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
	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);
2026

2027 2028 2029 2030
	btrfs_mark_buffer_dirty(leaf);
out:
	btrfs_free_path(path);
	return err;
2031 2032
}

2033 2034 2035 2036 2037
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)
2038 2039
{
	int ret = 0;
2040 2041 2042 2043
	struct btrfs_delayed_tree_ref *ref;
	struct btrfs_key ins;
	u64 parent = 0;
	u64 ref_root = 0;
2044

2045 2046 2047
	ins.objectid = node->bytenr;
	ins.offset = node->num_bytes;
	ins.type = BTRFS_EXTENT_ITEM_KEY;
2048

2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074
	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();
	}
2075 2076 2077 2078
	return ret;
}

/* helper function to actually process a single delayed ref entry */
2079 2080 2081 2082 2083
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)
2084 2085
{
	int ret;
2086
	if (btrfs_delayed_ref_is_head(node)) {
2087 2088 2089 2090 2091 2092 2093
		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
		 */
2094 2095
		BUG_ON(extent_op);
		head = btrfs_delayed_node_to_head(node);
2096
		if (insert_reserved) {
2097 2098
			btrfs_pin_extent(root, node->bytenr,
					 node->num_bytes, 1);
2099 2100 2101 2102 2103 2104
			if (head->is_data) {
				ret = btrfs_del_csums(trans, root,
						      node->bytenr,
						      node->num_bytes);
				BUG_ON(ret);
			}
2105 2106 2107 2108 2109
		}
		mutex_unlock(&head->mutex);
		return 0;
	}

2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120
	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;
2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142
}

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;
2143
		if (ref->action == action)
2144 2145 2146 2147 2148 2149 2150 2151 2152 2153
			return ref;
		node = rb_prev(node);
	}
	if (action == BTRFS_ADD_DELAYED_REF) {
		action = BTRFS_DROP_DELAYED_REF;
		goto again;
	}
	return NULL;
}

2154 2155 2156
static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
				       struct btrfs_root *root,
				       struct list_head *cluster)
2157 2158 2159 2160
{
	struct btrfs_delayed_ref_root *delayed_refs;
	struct btrfs_delayed_ref_node *ref;
	struct btrfs_delayed_ref_head *locked_ref = NULL;
2161
	struct btrfs_delayed_extent_op *extent_op;
2162
	int ret;
2163
	int count = 0;
2164 2165 2166 2167 2168
	int must_insert_reserved = 0;

	delayed_refs = &trans->transaction->delayed_refs;
	while (1) {
		if (!locked_ref) {
2169 2170
			/* pick a new head ref from the cluster list */
			if (list_empty(cluster))
2171 2172
				break;

2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189
			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;
2190 2191
			}
		}
2192

2193 2194 2195 2196 2197 2198
		/*
		 * 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;
2199

2200 2201 2202
		extent_op = locked_ref->extent_op;
		locked_ref->extent_op = NULL;

2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213
		/*
		 * 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) {
			/* 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;
2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232

			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);

				cond_resched();
				spin_lock(&delayed_refs->lock);
				continue;
			}

2233
			list_del_init(&locked_ref->cluster);
2234 2235
			locked_ref = NULL;
		}
C
Chris Mason 已提交
2236

2237 2238 2239
		ref->in_tree = 0;
		rb_erase(&ref->rb_node, &delayed_refs->root);
		delayed_refs->num_entries--;
2240

2241
		spin_unlock(&delayed_refs->lock);
2242

2243
		ret = run_one_delayed_ref(trans, root, ref, extent_op,
2244 2245
					  must_insert_reserved);
		BUG_ON(ret);
2246

2247 2248
		btrfs_put_delayed_ref(ref);
		kfree(extent_op);
2249
		count++;
2250

2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289
		cond_resched();
		spin_lock(&delayed_refs->lock);
	}
	return count;
}

/*
 * 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;
	int run_all = count == (unsigned long)-1;
	int run_most = 0;

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

	delayed_refs = &trans->transaction->delayed_refs;
	INIT_LIST_HEAD(&cluster);
again:
	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;
2290

2291
		/*
2292 2293 2294 2295
		 * 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
2296
		 */
2297 2298 2299
		ret = btrfs_find_ref_cluster(trans, &cluster,
					     delayed_refs->run_delayed_start);
		if (ret)
2300 2301
			break;

2302 2303 2304 2305 2306 2307 2308
		ret = run_clustered_refs(trans, root, &cluster);
		BUG_ON(ret < 0);

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

		if (count == 0)
			break;
2309
	}
2310

2311 2312
	if (run_all) {
		node = rb_first(&delayed_refs->root);
2313
		if (!node)
2314
			goto out;
2315
		count = (unsigned long)-1;
2316

2317 2318 2319 2320 2321
		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;
2322

2323 2324 2325 2326
				head = btrfs_delayed_node_to_head(ref);
				atomic_inc(&ref->refs);

				spin_unlock(&delayed_refs->lock);
2327 2328 2329 2330
				/*
				 * Mutex was contended, block until it's
				 * released and try again
				 */
2331 2332 2333 2334
				mutex_lock(&head->mutex);
				mutex_unlock(&head->mutex);

				btrfs_put_delayed_ref(ref);
2335
				cond_resched();
2336 2337 2338 2339 2340 2341 2342
				goto again;
			}
			node = rb_next(node);
		}
		spin_unlock(&delayed_refs->lock);
		schedule_timeout(1);
		goto again;
2343
	}
2344
out:
2345
	spin_unlock(&delayed_refs->lock);
2346 2347 2348
	return 0;
}

2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394
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;

	ret = btrfs_add_delayed_extent_op(trans, bytenr, num_bytes, extent_op);
	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);

2395
		btrfs_release_path(path);
2396

2397 2398 2399 2400
		/*
		 * Mutex was contended, block until it's released and let
		 * caller try again
		 */
2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444
		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)
2445 2446
{
	struct btrfs_root *extent_root = root->fs_info->extent_root;
2447
	struct extent_buffer *leaf;
2448 2449 2450
	struct btrfs_extent_data_ref *ref;
	struct btrfs_extent_inline_ref *iref;
	struct btrfs_extent_item *ei;
2451
	struct btrfs_key key;
2452
	u32 item_size;
2453
	int ret;
2454

2455
	key.objectid = bytenr;
Z
Zheng Yan 已提交
2456
	key.offset = (u64)-1;
2457
	key.type = BTRFS_EXTENT_ITEM_KEY;
2458 2459 2460 2461 2462

	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
2463 2464 2465

	ret = -ENOENT;
	if (path->slots[0] == 0)
Z
Zheng Yan 已提交
2466
		goto out;
2467

Z
Zheng Yan 已提交
2468
	path->slots[0]--;
2469
	leaf = path->nodes[0];
2470
	btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
2471

2472
	if (key.objectid != bytenr || key.type != BTRFS_EXTENT_ITEM_KEY)
2473
		goto out;
2474

2475 2476 2477 2478 2479 2480 2481 2482 2483
	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);
2484

2485 2486 2487
	if (item_size != sizeof(*ei) +
	    btrfs_extent_inline_ref_size(BTRFS_EXTENT_DATA_REF_KEY))
		goto out;
2488

2489 2490 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 2519 2520 2521 2522 2523 2524 2525 2526 2527
	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)
2528
			goto out;
Y
Yan Zheng 已提交
2529

2530 2531 2532 2533 2534 2535 2536
		ret2 = check_delayed_ref(trans, root, path, objectid,
					 offset, bytenr);
	} while (ret2 == -EAGAIN);

	if (ret2 && ret2 != -ENOENT) {
		ret = ret2;
		goto out;
2537
	}
2538 2539 2540

	if (ret != -ENOENT || ret2 != -ENOENT)
		ret = 0;
2541
out:
Y
Yan Zheng 已提交
2542
	btrfs_free_path(path);
2543 2544
	if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
		WARN_ON(ret > 0);
2545
	return ret;
2546
}
C
Chris Mason 已提交
2547

2548
static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
2549
			   struct btrfs_root *root,
2550 2551
			   struct extent_buffer *buf,
			   int full_backref, int inc)
Z
Zheng Yan 已提交
2552 2553
{
	u64 bytenr;
2554 2555
	u64 num_bytes;
	u64 parent;
Z
Zheng Yan 已提交
2556 2557 2558 2559 2560 2561 2562 2563
	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 *,
2564
			    u64, u64, u64, u64, u64, u64);
Z
Zheng Yan 已提交
2565 2566 2567 2568 2569

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

2570 2571
	if (!root->ref_cows && level == 0)
		return 0;
Z
Zheng Yan 已提交
2572

2573 2574 2575 2576
	if (inc)
		process_func = btrfs_inc_extent_ref;
	else
		process_func = btrfs_free_extent;
Z
Zheng Yan 已提交
2577

2578 2579 2580 2581 2582 2583
	if (full_backref)
		parent = buf->start;
	else
		parent = 0;

	for (i = 0; i < nritems; i++) {
Z
Zheng Yan 已提交
2584
		if (level == 0) {
2585
			btrfs_item_key_to_cpu(buf, &key, i);
Z
Zheng Yan 已提交
2586 2587
			if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
				continue;
2588
			fi = btrfs_item_ptr(buf, i,
Z
Zheng Yan 已提交
2589 2590 2591 2592 2593 2594 2595
					    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;
2596 2597 2598 2599 2600 2601

			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,
					   key.offset);
Z
Zheng Yan 已提交
2602 2603 2604
			if (ret)
				goto fail;
		} else {
2605 2606 2607 2608
			bytenr = btrfs_node_blockptr(buf, i);
			num_bytes = btrfs_level_size(root, level - 1);
			ret = process_func(trans, root, bytenr, num_bytes,
					   parent, ref_root, level - 1, 0);
Z
Zheng Yan 已提交
2609 2610 2611 2612 2613 2614
			if (ret)
				goto fail;
		}
	}
	return 0;
fail:
2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628
	BUG();
	return ret;
}

int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
		  struct extent_buffer *buf, int full_backref)
{
	return __btrfs_mod_ref(trans, root, buf, full_backref, 1);
}

int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
		  struct extent_buffer *buf, int full_backref)
{
	return __btrfs_mod_ref(trans, root, buf, full_backref, 0);
Z
Zheng Yan 已提交
2629 2630
}

C
Chris Mason 已提交
2631 2632 2633 2634 2635 2636 2637
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;
2638 2639
	unsigned long bi;
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2640 2641

	ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
2642 2643
	if (ret < 0)
		goto fail;
C
Chris Mason 已提交
2644
	BUG_ON(ret);
2645 2646 2647 2648 2649

	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);
2650
	btrfs_release_path(path);
2651
fail:
C
Chris Mason 已提交
2652 2653 2654 2655 2656 2657
	if (ret)
		return ret;
	return 0;

}

2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668
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);
2669
		btrfs_get_block_group(cache);
2670 2671 2672 2673 2674 2675
	} else
		cache = NULL;
	spin_unlock(&root->fs_info->block_group_cache_lock);
	return cache;
}

2676 2677 2678 2679 2680 2681 2682
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;
2683
	int dcs = BTRFS_DC_ERROR;
2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702
	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);
2703
		btrfs_release_path(path);
2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737
		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;
	}

	/*
	 * 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) {
2738 2739
		/* We're not cached, don't bother trying to write stuff out */
		dcs = BTRFS_DC_WRITTEN;
2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758
		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;

2759
	ret = btrfs_delalloc_reserve_space(inode, num_pages);
2760 2761 2762 2763 2764 2765
	if (ret)
		goto out_put;

	ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, num_pages,
					      num_pages, num_pages,
					      &alloc_hint);
2766
	if (!ret) {
2767
		dcs = BTRFS_DC_SETUP;
2768 2769 2770 2771 2772
		btrfs_free_reserved_data_space(inode, num_pages);
	} else {
		btrfs_delalloc_release_space(inode, num_pages);
	}

2773 2774 2775
out_put:
	iput(inode);
out_free:
2776
	btrfs_release_path(path);
2777 2778
out:
	spin_lock(&block_group->lock);
2779
	block_group->disk_cache_state = dcs;
2780 2781 2782 2783 2784
	spin_unlock(&block_group->lock);

	return ret;
}

2785 2786
int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
				   struct btrfs_root *root)
C
Chris Mason 已提交
2787
{
2788
	struct btrfs_block_group_cache *cache;
C
Chris Mason 已提交
2789 2790
	int err = 0;
	struct btrfs_path *path;
2791
	u64 last = 0;
C
Chris Mason 已提交
2792 2793 2794 2795 2796

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

2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815
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
2816
	while (1) {
2817 2818 2819 2820
		if (last == 0) {
			err = btrfs_run_delayed_refs(trans, root,
						     (unsigned long)-1);
			BUG_ON(err);
J
Josef Bacik 已提交
2821
		}
2822

2823 2824
		cache = btrfs_lookup_first_block_group(root->fs_info, last);
		while (cache) {
2825 2826 2827 2828 2829
			if (cache->disk_cache_state == BTRFS_DC_CLEAR) {
				btrfs_put_block_group(cache);
				goto again;
			}

2830 2831 2832 2833 2834 2835 2836 2837 2838 2839
			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
2840

J
Josef Bacik 已提交
2841 2842
		if (cache->disk_cache_state == BTRFS_DC_SETUP)
			cache->disk_cache_state = BTRFS_DC_NEED_WRITE;
2843
		cache->dirty = 0;
2844
		last = cache->key.objectid + cache->key.offset;
J
Josef Bacik 已提交
2845

2846 2847 2848
		err = write_one_cache_group(trans, root, path, cache);
		BUG_ON(err);
		btrfs_put_block_group(cache);
C
Chris Mason 已提交
2849
	}
2850

J
Josef Bacik 已提交
2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 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
	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
2897
	btrfs_free_path(path);
2898
	return 0;
C
Chris Mason 已提交
2899 2900
}

2901 2902 2903 2904 2905 2906 2907 2908 2909
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)
2910
		btrfs_put_block_group(block_group);
2911 2912 2913
	return readonly;
}

2914 2915 2916 2917 2918
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;
2919 2920 2921 2922 2923 2924 2925 2926
	int i;
	int factor;

	if (flags & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1 |
		     BTRFS_BLOCK_GROUP_RAID10))
		factor = 2;
	else
		factor = 1;
2927 2928 2929

	found = __find_space_info(info, flags);
	if (found) {
2930
		spin_lock(&found->lock);
2931
		found->total_bytes += total_bytes;
J
Josef Bacik 已提交
2932
		found->disk_total += total_bytes * factor;
2933
		found->bytes_used += bytes_used;
2934
		found->disk_used += bytes_used * factor;
2935
		found->full = 0;
2936
		spin_unlock(&found->lock);
2937 2938 2939
		*space_info = found;
		return 0;
	}
Y
Yan Zheng 已提交
2940
	found = kzalloc(sizeof(*found), GFP_NOFS);
2941 2942 2943
	if (!found)
		return -ENOMEM;

2944 2945
	for (i = 0; i < BTRFS_NR_RAID_TYPES; i++)
		INIT_LIST_HEAD(&found->block_groups[i]);
2946
	init_rwsem(&found->groups_sem);
J
Josef Bacik 已提交
2947
	spin_lock_init(&found->lock);
2948 2949 2950
	found->flags = flags & (BTRFS_BLOCK_GROUP_DATA |
				BTRFS_BLOCK_GROUP_SYSTEM |
				BTRFS_BLOCK_GROUP_METADATA);
2951
	found->total_bytes = total_bytes;
J
Josef Bacik 已提交
2952
	found->disk_total = total_bytes * factor;
2953
	found->bytes_used = bytes_used;
2954
	found->disk_used = bytes_used * factor;
2955
	found->bytes_pinned = 0;
2956
	found->bytes_reserved = 0;
Y
Yan Zheng 已提交
2957
	found->bytes_readonly = 0;
2958
	found->bytes_may_use = 0;
2959
	found->full = 0;
2960
	found->force_alloc = CHUNK_ALLOC_NO_FORCE;
2961
	found->chunk_alloc = 0;
2962 2963
	found->flush = 0;
	init_waitqueue_head(&found->wait);
2964
	*space_info = found;
2965
	list_add_rcu(&found->list, &info->space_info);
2966 2967 2968
	return 0;
}

2969 2970 2971
static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
{
	u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 |
2972
				   BTRFS_BLOCK_GROUP_RAID1 |
C
Chris Mason 已提交
2973
				   BTRFS_BLOCK_GROUP_RAID10 |
2974
				   BTRFS_BLOCK_GROUP_DUP);
2975 2976 2977 2978 2979 2980 2981 2982 2983
	if (extra_flags) {
		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;
	}
}
2984

Y
Yan Zheng 已提交
2985
u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
2986
{
2987 2988 2989 2990 2991 2992 2993
	/*
	 * 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;
2994 2995 2996 2997 2998 2999

	if (num_devices == 1)
		flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
	if (num_devices < 4)
		flags &= ~BTRFS_BLOCK_GROUP_RAID10;

3000 3001
	if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
	    (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3002
		      BTRFS_BLOCK_GROUP_RAID10))) {
3003
		flags &= ~BTRFS_BLOCK_GROUP_DUP;
3004
	}
3005 3006

	if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
3007
	    (flags & BTRFS_BLOCK_GROUP_RAID10)) {
3008
		flags &= ~BTRFS_BLOCK_GROUP_RAID1;
3009
	}
3010 3011 3012 3013 3014 3015 3016 3017 3018

	if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
	    ((flags & BTRFS_BLOCK_GROUP_RAID1) |
	     (flags & BTRFS_BLOCK_GROUP_RAID10) |
	     (flags & BTRFS_BLOCK_GROUP_DUP)))
		flags &= ~BTRFS_BLOCK_GROUP_RAID0;
	return flags;
}

3019
static u64 get_alloc_profile(struct btrfs_root *root, u64 flags)
J
Josef Bacik 已提交
3020
{
3021 3022 3023 3024 3025 3026 3027 3028 3029 3030
	if (flags & BTRFS_BLOCK_GROUP_DATA)
		flags |= root->fs_info->avail_data_alloc_bits &
			 root->fs_info->data_alloc_profile;
	else if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
		flags |= root->fs_info->avail_system_alloc_bits &
			 root->fs_info->system_alloc_profile;
	else if (flags & BTRFS_BLOCK_GROUP_METADATA)
		flags |= root->fs_info->avail_metadata_alloc_bits &
			 root->fs_info->metadata_alloc_profile;
	return btrfs_reduce_alloc_profile(root, flags);
J
Josef Bacik 已提交
3031 3032
}

3033
u64 btrfs_get_alloc_profile(struct btrfs_root *root, int data)
J
Josef Bacik 已提交
3034
{
3035
	u64 flags;
J
Josef Bacik 已提交
3036

3037 3038 3039 3040
	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
3041
	else
3042
		flags = BTRFS_BLOCK_GROUP_METADATA;
J
Josef Bacik 已提交
3043

3044
	return get_alloc_profile(root, flags);
J
Josef Bacik 已提交
3045
}
J
Josef Bacik 已提交
3046

J
Josef Bacik 已提交
3047 3048 3049
void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *inode)
{
	BTRFS_I(inode)->space_info = __find_space_info(root->fs_info,
3050
						       BTRFS_BLOCK_GROUP_DATA);
J
Josef Bacik 已提交
3051 3052
}

J
Josef Bacik 已提交
3053 3054 3055 3056
/*
 * This will check the space that the inode allocates from to make sure we have
 * enough space for bytes.
 */
3057
int btrfs_check_data_free_space(struct inode *inode, u64 bytes)
J
Josef Bacik 已提交
3058 3059
{
	struct btrfs_space_info *data_sinfo;
3060
	struct btrfs_root *root = BTRFS_I(inode)->root;
3061
	u64 used;
3062
	int ret = 0, committed = 0, alloc_chunk = 1;
J
Josef Bacik 已提交
3063 3064 3065 3066

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

3067 3068
	if (root == root->fs_info->tree_root ||
	    BTRFS_I(inode)->location.objectid == BTRFS_FREE_INO_OBJECTID) {
3069 3070 3071 3072
		alloc_chunk = 0;
		committed = 1;
	}

J
Josef Bacik 已提交
3073
	data_sinfo = BTRFS_I(inode)->space_info;
C
Chris Mason 已提交
3074 3075
	if (!data_sinfo)
		goto alloc;
J
Josef Bacik 已提交
3076

J
Josef Bacik 已提交
3077 3078 3079
again:
	/* make sure we have enough space to handle the data first */
	spin_lock(&data_sinfo->lock);
3080 3081 3082
	used = data_sinfo->bytes_used + data_sinfo->bytes_reserved +
		data_sinfo->bytes_pinned + data_sinfo->bytes_readonly +
		data_sinfo->bytes_may_use;
3083 3084

	if (used + bytes > data_sinfo->total_bytes) {
3085
		struct btrfs_trans_handle *trans;
J
Josef Bacik 已提交
3086

J
Josef Bacik 已提交
3087 3088 3089 3090
		/*
		 * if we don't have enough free bytes in this space then we need
		 * to alloc a new chunk.
		 */
3091
		if (!data_sinfo->full && alloc_chunk) {
J
Josef Bacik 已提交
3092
			u64 alloc_target;
J
Josef Bacik 已提交
3093

3094
			data_sinfo->force_alloc = CHUNK_ALLOC_FORCE;
J
Josef Bacik 已提交
3095
			spin_unlock(&data_sinfo->lock);
C
Chris Mason 已提交
3096
alloc:
J
Josef Bacik 已提交
3097
			alloc_target = btrfs_get_alloc_profile(root, 1);
3098
			trans = btrfs_join_transaction(root);
3099 3100
			if (IS_ERR(trans))
				return PTR_ERR(trans);
J
Josef Bacik 已提交
3101

J
Josef Bacik 已提交
3102 3103
			ret = do_chunk_alloc(trans, root->fs_info->extent_root,
					     bytes + 2 * 1024 * 1024,
3104 3105
					     alloc_target,
					     CHUNK_ALLOC_NO_FORCE);
J
Josef Bacik 已提交
3106
			btrfs_end_transaction(trans, root);
3107 3108 3109 3110 3111 3112
			if (ret < 0) {
				if (ret != -ENOSPC)
					return ret;
				else
					goto commit_trans;
			}
J
Josef Bacik 已提交
3113

C
Chris Mason 已提交
3114 3115 3116 3117
			if (!data_sinfo) {
				btrfs_set_inode_space_info(root, inode);
				data_sinfo = BTRFS_I(inode)->space_info;
			}
J
Josef Bacik 已提交
3118 3119
			goto again;
		}
3120 3121 3122 3123 3124 3125 3126

		/*
		 * 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
3127 3128
		spin_unlock(&data_sinfo->lock);

3129
		/* commit the current transaction and try again */
3130
commit_trans:
J
Josef Bacik 已提交
3131 3132
		if (!committed &&
		    !atomic_read(&root->fs_info->open_ioctl_trans)) {
3133
			committed = 1;
3134
			trans = btrfs_join_transaction(root);
3135 3136
			if (IS_ERR(trans))
				return PTR_ERR(trans);
3137 3138 3139 3140 3141
			ret = btrfs_commit_transaction(trans, root);
			if (ret)
				return ret;
			goto again;
		}
J
Josef Bacik 已提交
3142

J
Josef Bacik 已提交
3143 3144 3145 3146 3147
		return -ENOSPC;
	}
	data_sinfo->bytes_may_use += bytes;
	spin_unlock(&data_sinfo->lock);

J
Josef Bacik 已提交
3148 3149
	return 0;
}
J
Josef Bacik 已提交
3150 3151

/*
3152
 * Called if we need to clear a data reservation for this inode.
J
Josef Bacik 已提交
3153
 */
3154
void btrfs_free_reserved_data_space(struct inode *inode, u64 bytes)
3155
{
3156
	struct btrfs_root *root = BTRFS_I(inode)->root;
J
Josef Bacik 已提交
3157
	struct btrfs_space_info *data_sinfo;
3158

J
Josef Bacik 已提交
3159 3160
	/* make sure bytes are sectorsize aligned */
	bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
3161

J
Josef Bacik 已提交
3162 3163 3164 3165
	data_sinfo = BTRFS_I(inode)->space_info;
	spin_lock(&data_sinfo->lock);
	data_sinfo->bytes_may_use -= bytes;
	spin_unlock(&data_sinfo->lock);
3166 3167
}

3168
static void force_metadata_allocation(struct btrfs_fs_info *info)
3169
{
3170 3171
	struct list_head *head = &info->space_info;
	struct btrfs_space_info *found;
3172

3173 3174 3175
	rcu_read_lock();
	list_for_each_entry_rcu(found, head, list) {
		if (found->flags & BTRFS_BLOCK_GROUP_METADATA)
3176
			found->force_alloc = CHUNK_ALLOC_FORCE;
3177
	}
3178
	rcu_read_unlock();
3179 3180
}

3181
static int should_alloc_chunk(struct btrfs_root *root,
3182 3183
			      struct btrfs_space_info *sinfo, u64 alloc_bytes,
			      int force)
3184
{
3185
	struct btrfs_block_rsv *global_rsv = &root->fs_info->global_block_rsv;
3186
	u64 num_bytes = sinfo->total_bytes - sinfo->bytes_readonly;
3187
	u64 num_allocated = sinfo->bytes_used + sinfo->bytes_reserved;
3188
	u64 thresh;
3189

3190 3191 3192
	if (force == CHUNK_ALLOC_FORCE)
		return 1;

3193 3194 3195 3196 3197 3198 3199
	/*
	 * 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;

3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221
	/*
	 * in limited mode, we want to have some free space up to
	 * about 1% of the FS size.
	 */
	if (force == CHUNK_ALLOC_LIMITED) {
		thresh = btrfs_super_total_bytes(&root->fs_info->super_copy);
		thresh = max_t(u64, 64 * 1024 * 1024,
			       div_factor_fine(thresh, 1));

		if (num_bytes - num_allocated < thresh)
			return 1;
	}

	/*
	 * we have two similar checks here, one based on percentage
	 * and once based on a hard number of 256MB.  The idea
	 * is that if we have a good amount of free
	 * room, don't allocate a chunk.  A good mount is
	 * less than 80% utilized of the chunks we have allocated,
	 * or more than 256MB free
	 */
	if (num_allocated + alloc_bytes + 256 * 1024 * 1024 < num_bytes)
3222
		return 0;
3223

3224
	if (num_allocated + alloc_bytes < div_factor(num_bytes, 8))
3225
		return 0;
3226

3227
	thresh = btrfs_super_total_bytes(&root->fs_info->super_copy);
3228 3229

	/* 256MB or 5% of the FS */
3230 3231 3232
	thresh = max_t(u64, 256 * 1024 * 1024, div_factor_fine(thresh, 5));

	if (num_bytes > thresh && sinfo->bytes_used < div_factor(num_bytes, 3))
3233
		return 0;
3234
	return 1;
3235 3236
}

3237 3238
static int do_chunk_alloc(struct btrfs_trans_handle *trans,
			  struct btrfs_root *extent_root, u64 alloc_bytes,
C
Chris Mason 已提交
3239
			  u64 flags, int force)
J
Josef Bacik 已提交
3240
{
3241
	struct btrfs_space_info *space_info;
3242
	struct btrfs_fs_info *fs_info = extent_root->fs_info;
3243
	int wait_for_alloc = 0;
J
Josef Bacik 已提交
3244 3245
	int ret = 0;

Y
Yan Zheng 已提交
3246
	flags = btrfs_reduce_alloc_profile(extent_root, flags);
3247

3248
	space_info = __find_space_info(extent_root->fs_info, flags);
3249 3250 3251 3252
	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
3253
	}
3254
	BUG_ON(!space_info);
J
Josef Bacik 已提交
3255

3256
again:
3257
	spin_lock(&space_info->lock);
J
Josef Bacik 已提交
3258
	if (space_info->force_alloc)
3259
		force = space_info->force_alloc;
3260 3261
	if (space_info->full) {
		spin_unlock(&space_info->lock);
3262
		return 0;
J
Josef Bacik 已提交
3263 3264
	}

3265
	if (!should_alloc_chunk(extent_root, space_info, alloc_bytes, force)) {
3266
		spin_unlock(&space_info->lock);
3267 3268 3269 3270 3271
		return 0;
	} else if (space_info->chunk_alloc) {
		wait_for_alloc = 1;
	} else {
		space_info->chunk_alloc = 1;
J
Josef Bacik 已提交
3272
	}
3273

3274
	spin_unlock(&space_info->lock);
J
Josef Bacik 已提交
3275

3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289
	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;
	}

3290 3291 3292 3293 3294 3295 3296
	/*
	 * 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);

3297 3298 3299 3300 3301
	/*
	 * 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
3302
	if (flags & BTRFS_BLOCK_GROUP_DATA && fs_info->metadata_ratio) {
3303 3304 3305 3306
		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
3307 3308
	}

Y
Yan Zheng 已提交
3309
	ret = btrfs_alloc_chunk(trans, extent_root, flags);
3310 3311 3312
	if (ret < 0 && ret != -ENOSPC)
		goto out;

J
Josef Bacik 已提交
3313 3314
	spin_lock(&space_info->lock);
	if (ret)
3315
		space_info->full = 1;
3316 3317
	else
		ret = 1;
3318

3319
	space_info->force_alloc = CHUNK_ALLOC_NO_FORCE;
3320
	space_info->chunk_alloc = 0;
J
Josef Bacik 已提交
3321
	spin_unlock(&space_info->lock);
3322
out:
Y
Yan Zheng 已提交
3323
	mutex_unlock(&extent_root->fs_info->chunk_mutex);
J
Josef Bacik 已提交
3324
	return ret;
3325
}
J
Josef Bacik 已提交
3326 3327

/*
3328
 * shrink metadata reservation for delalloc
J
Josef Bacik 已提交
3329
 */
3330
static int shrink_delalloc(struct btrfs_trans_handle *trans,
J
Josef Bacik 已提交
3331
			   struct btrfs_root *root, u64 to_reclaim, int sync)
3332
{
3333
	struct btrfs_block_rsv *block_rsv;
J
Josef Bacik 已提交
3334
	struct btrfs_space_info *space_info;
3335 3336 3337
	u64 reserved;
	u64 max_reclaim;
	u64 reclaimed = 0;
3338
	long time_left;
3339
	int nr_pages = (2 * 1024 * 1024) >> PAGE_CACHE_SHIFT;
3340
	int loops = 0;
3341
	unsigned long progress;
3342

3343
	block_rsv = &root->fs_info->delalloc_block_rsv;
J
Josef Bacik 已提交
3344
	space_info = block_rsv->space_info;
3345 3346

	smp_mb();
3347
	reserved = space_info->bytes_may_use;
3348
	progress = space_info->reservation_progress;
3349 3350 3351

	if (reserved == 0)
		return 0;
3352

3353 3354 3355 3356 3357 3358 3359 3360
	smp_mb();
	if (root->fs_info->delalloc_bytes == 0) {
		if (trans)
			return 0;
		btrfs_wait_ordered_extents(root, 0, 0);
		return 0;
	}

3361 3362
	max_reclaim = min(reserved, to_reclaim);

3363
	while (loops < 1024) {
3364 3365 3366 3367 3368
		/* 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);
		writeback_inodes_sb_nr_if_idle(root->fs_info->sb, nr_pages);
3369

J
Josef Bacik 已提交
3370
		spin_lock(&space_info->lock);
3371 3372 3373
		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
3374
		spin_unlock(&space_info->lock);
3375

3376 3377
		loops++;

3378 3379 3380 3381 3382
		if (reserved == 0 || reclaimed >= max_reclaim)
			break;

		if (trans && trans->transaction->blocked)
			return -EAGAIN;
3383

3384
		time_left = schedule_timeout_interruptible(1);
3385 3386 3387 3388 3389

		/* We were interrupted, exit */
		if (time_left)
			break;

3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400
		/* 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;
		}
3401

3402
	}
3403 3404
	if (reclaimed >= to_reclaim && !trans)
		btrfs_wait_ordered_extents(root, 0, 0);
3405 3406 3407
	return reclaimed >= to_reclaim;
}

3408 3409 3410 3411 3412 3413
/**
 * 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
3414
 * @check - wether this is just to check if we have enough space or not
3415
 *
3416 3417 3418 3419 3420 3421
 * 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.
3422
 */
3423
static int reserve_metadata_bytes(struct btrfs_root *root,
3424
				  struct btrfs_block_rsv *block_rsv,
3425
				  u64 orig_bytes, int flush, int check)
J
Josef Bacik 已提交
3426
{
3427
	struct btrfs_space_info *space_info = block_rsv->space_info;
3428
	struct btrfs_trans_handle *trans;
3429
	u64 used;
3430 3431 3432
	u64 num_bytes = orig_bytes;
	int retries = 0;
	int ret = 0;
3433
	bool committed = false;
3434
	bool flushing = false;
3435 3436

	trans = (struct btrfs_trans_handle *)current->journal_info;
3437
again:
3438
	ret = 0;
3439
	spin_lock(&space_info->lock);
3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463
	/*
	 * 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.
		 */
		if (trans)
			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;
3464 3465 3466
	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
3467

3468 3469 3470 3471 3472 3473 3474
	/*
	 * 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.
	 */
3475 3476
	if (used <= space_info->total_bytes) {
		if (used + orig_bytes <= space_info->total_bytes) {
3477
			space_info->bytes_may_use += orig_bytes;
3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492
			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.
		 */
3493
		num_bytes = used - space_info->total_bytes +
3494 3495
			(orig_bytes * (retries + 1));
	}
J
Josef Bacik 已提交
3496

3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529
	if (ret && !check) {
		u64 profile = btrfs_get_alloc_profile(root, 0);
		u64 avail;

		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);

		if (used + orig_bytes < space_info->total_bytes + avail) {
			space_info->bytes_may_use += orig_bytes;
			ret = 0;
		}
	}

3530 3531 3532 3533 3534
	/*
	 * 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.
	 */
3535 3536 3537
	if (ret && flush) {
		flushing = true;
		space_info->flush = 1;
3538
	}
J
Josef Bacik 已提交
3539

3540
	spin_unlock(&space_info->lock);
J
Josef Bacik 已提交
3541

3542
	if (!ret || !flush)
3543
		goto out;
3544

3545 3546 3547 3548 3549
	/*
	 * We do synchronous shrinking since we don't actually unreserve
	 * metadata until after the IO is completed.
	 */
	ret = shrink_delalloc(trans, root, num_bytes, 1);
3550
	if (ret < 0)
3551
		goto out;
3552

3553 3554
	ret = 0;

3555 3556 3557 3558 3559 3560 3561 3562 3563
	/*
	 * 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) {
		retries++;
		goto again;
	}
3564

3565 3566 3567 3568
	/*
	 * Not enough space to be reclaimed, don't bother committing the
	 * transaction.
	 */
3569
	spin_lock(&space_info->lock);
3570 3571 3572 3573 3574
	if (space_info->bytes_pinned < orig_bytes)
		ret = -ENOSPC;
	spin_unlock(&space_info->lock);
	if (ret)
		goto out;
3575

3576
	ret = -EAGAIN;
3577
	if (trans)
3578
		goto out;
3579

3580
	ret = -ENOSPC;
3581 3582 3583
	if (committed)
		goto out;

3584
	trans = btrfs_join_transaction(root);
3585 3586 3587
	if (IS_ERR(trans))
		goto out;
	ret = btrfs_commit_transaction(trans, root);
3588 3589 3590
	if (!ret) {
		trans = NULL;
		committed = true;
3591
		goto again;
3592
	}
3593 3594

out:
3595
	if (flushing) {
3596
		spin_lock(&space_info->lock);
3597 3598
		space_info->flush = 0;
		wake_up_all(&space_info->wait);
3599
		spin_unlock(&space_info->lock);
3600 3601 3602 3603 3604 3605 3606
	}
	return ret;
}

static struct btrfs_block_rsv *get_block_rsv(struct btrfs_trans_handle *trans,
					     struct btrfs_root *root)
{
3607 3608 3609
	struct btrfs_block_rsv *block_rsv = NULL;

	if (root->ref_cows || root == root->fs_info->csum_root)
3610
		block_rsv = trans->block_rsv;
3611 3612

	if (!block_rsv)
3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647
		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);
}

3648 3649
static void block_rsv_release_bytes(struct btrfs_block_rsv *block_rsv,
				    struct btrfs_block_rsv *dest, u64 num_bytes)
3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667
{
	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) {
3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681
			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) {
3682
			spin_lock(&space_info->lock);
3683
			space_info->bytes_may_use -= num_bytes;
3684
			space_info->reservation_progress++;
3685
			spin_unlock(&space_info->lock);
3686
		}
J
Josef Bacik 已提交
3687
	}
3688
}
3689

3690 3691 3692 3693
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
3694

3695 3696 3697
	ret = block_rsv_use_bytes(src, num_bytes);
	if (ret)
		return ret;
J
Josef Bacik 已提交
3698

3699
	block_rsv_add_bytes(dst, num_bytes, 1);
J
Josef Bacik 已提交
3700 3701 3702
	return 0;
}

3703
void btrfs_init_block_rsv(struct btrfs_block_rsv *rsv)
J
Josef Bacik 已提交
3704
{
3705 3706 3707 3708 3709 3710 3711 3712
	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
3713

3714 3715 3716
	block_rsv = kmalloc(sizeof(*block_rsv), GFP_NOFS);
	if (!block_rsv)
		return NULL;
J
Josef Bacik 已提交
3717

3718 3719 3720 3721 3722
	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
3723

3724 3725 3726
void btrfs_free_block_rsv(struct btrfs_root *root,
			  struct btrfs_block_rsv *rsv)
{
3727 3728
	btrfs_block_rsv_release(root, rsv, (u64)-1);
	kfree(rsv);
J
Josef Bacik 已提交
3729 3730
}

3731
int btrfs_block_rsv_add(struct btrfs_root *root,
3732
			struct btrfs_block_rsv *block_rsv,
3733
			u64 num_bytes)
3734 3735
{
	int ret;
J
Josef Bacik 已提交
3736

3737 3738
	if (num_bytes == 0)
		return 0;
3739

3740
	ret = reserve_metadata_bytes(root, block_rsv, num_bytes, 1, 0);
3741 3742 3743 3744
	if (!ret) {
		block_rsv_add_bytes(block_rsv, num_bytes, 1);
		return 0;
	}
J
Josef Bacik 已提交
3745

3746 3747
	return ret;
}
J
Josef Bacik 已提交
3748

3749
int btrfs_block_rsv_check(struct btrfs_root *root,
3750
			  struct btrfs_block_rsv *block_rsv,
3751
			  u64 min_reserved, int min_factor, int flush)
3752 3753 3754
{
	u64 num_bytes = 0;
	int ret = -ENOSPC;
J
Josef Bacik 已提交
3755

3756 3757
	if (!block_rsv)
		return 0;
J
Josef Bacik 已提交
3758

3759 3760 3761 3762 3763
	spin_lock(&block_rsv->lock);
	if (min_factor > 0)
		num_bytes = div_factor(block_rsv->size, min_factor);
	if (min_reserved > num_bytes)
		num_bytes = min_reserved;
J
Josef Bacik 已提交
3764

3765
	if (block_rsv->reserved >= num_bytes)
3766
		ret = 0;
3767
	else
3768 3769
		num_bytes -= block_rsv->reserved;
	spin_unlock(&block_rsv->lock);
3770

3771 3772 3773
	if (!ret)
		return 0;

3774
	ret = reserve_metadata_bytes(root, block_rsv, num_bytes, flush, !flush);
3775 3776 3777
	if (!ret) {
		block_rsv_add_bytes(block_rsv, num_bytes, 0);
		return 0;
3778
	}
J
Josef Bacik 已提交
3779

3780
	return ret;
3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798
}

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;
	block_rsv_release_bytes(block_rsv, global_rsv, num_bytes);
J
Josef Bacik 已提交
3799 3800 3801
}

/*
3802 3803 3804
 * 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
3805
 */
3806
static u64 calc_global_metadata_size(struct btrfs_fs_info *fs_info)
J
Josef Bacik 已提交
3807
{
3808 3809 3810 3811 3812
	struct btrfs_space_info *sinfo;
	u64 num_bytes;
	u64 meta_used;
	u64 data_used;
	int csum_size = btrfs_super_csum_size(&fs_info->super_copy);
J
Josef Bacik 已提交
3813

3814 3815 3816 3817
	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
3818

3819 3820
	sinfo = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_METADATA);
	spin_lock(&sinfo->lock);
3821 3822
	if (sinfo->flags & BTRFS_BLOCK_GROUP_DATA)
		data_used = 0;
3823 3824
	meta_used = sinfo->bytes_used;
	spin_unlock(&sinfo->lock);
3825

3826 3827 3828
	num_bytes = (data_used >> fs_info->sb->s_blocksize_bits) *
		    csum_size * 2;
	num_bytes += div64_u64(data_used + meta_used, 50);
3829

3830 3831
	if (num_bytes * 3 > meta_used)
		num_bytes = div64_u64(meta_used, 3);
3832

3833 3834
	return ALIGN(num_bytes, fs_info->extent_root->leafsize << 10);
}
J
Josef Bacik 已提交
3835

3836 3837 3838 3839 3840
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
3841

3842
	num_bytes = calc_global_metadata_size(fs_info);
C
Chris Mason 已提交
3843

3844 3845
	spin_lock(&block_rsv->lock);
	spin_lock(&sinfo->lock);
3846

3847
	block_rsv->size = num_bytes;
3848

3849
	num_bytes = sinfo->bytes_used + sinfo->bytes_pinned +
3850 3851
		    sinfo->bytes_reserved + sinfo->bytes_readonly +
		    sinfo->bytes_may_use;
3852 3853 3854 3855

	if (sinfo->total_bytes > num_bytes) {
		num_bytes = sinfo->total_bytes - num_bytes;
		block_rsv->reserved += num_bytes;
3856
		sinfo->bytes_may_use += num_bytes;
J
Josef Bacik 已提交
3857 3858
	}

3859 3860
	if (block_rsv->reserved >= block_rsv->size) {
		num_bytes = block_rsv->reserved - block_rsv->size;
3861
		sinfo->bytes_may_use -= num_bytes;
3862
		sinfo->reservation_progress++;
3863 3864 3865
		block_rsv->reserved = block_rsv->size;
		block_rsv->full = 1;
	}
3866

3867 3868
	spin_unlock(&sinfo->lock);
	spin_unlock(&block_rsv->lock);
J
Josef Bacik 已提交
3869 3870
}

3871
static void init_global_block_rsv(struct btrfs_fs_info *fs_info)
J
Josef Bacik 已提交
3872
{
3873
	struct btrfs_space_info *space_info;
J
Josef Bacik 已提交
3874

3875 3876
	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
3877

3878
	space_info = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_METADATA);
3879 3880
	fs_info->global_block_rsv.space_info = space_info;
	fs_info->delalloc_block_rsv.space_info = space_info;
3881 3882 3883
	fs_info->trans_block_rsv.space_info = space_info;
	fs_info->empty_block_rsv.space_info = space_info;

3884 3885 3886 3887
	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;
3888
	fs_info->chunk_root->block_rsv = &fs_info->chunk_block_rsv;
3889 3890

	update_global_block_rsv(fs_info);
J
Josef Bacik 已提交
3891 3892
}

3893
static void release_global_block_rsv(struct btrfs_fs_info *fs_info)
J
Josef Bacik 已提交
3894
{
3895 3896 3897 3898 3899 3900 3901
	block_rsv_release_bytes(&fs_info->global_block_rsv, NULL, (u64)-1);
	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);
3902
}
J
Josef Bacik 已提交
3903

3904 3905
void btrfs_trans_release_metadata(struct btrfs_trans_handle *trans,
				  struct btrfs_root *root)
J
Josef Bacik 已提交
3906
{
3907 3908
	struct btrfs_block_rsv *block_rsv;

3909 3910
	if (!trans->bytes_reserved)
		return;
J
Josef Bacik 已提交
3911

3912 3913
	block_rsv = &root->fs_info->trans_block_rsv;
	btrfs_block_rsv_release(root, block_rsv, trans->bytes_reserved);
3914 3915
	trans->bytes_reserved = 0;
}
J
Josef Bacik 已提交
3916

3917 3918 3919 3920 3921 3922 3923 3924
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;

	/*
3925 3926 3927
	 * 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.
3928
	 */
C
Chris Mason 已提交
3929
	u64 num_bytes = btrfs_calc_trans_metadata_size(root, 1);
3930
	return block_rsv_migrate_bytes(src_rsv, dst_rsv, num_bytes);
J
Josef Bacik 已提交
3931 3932
}

3933
void btrfs_orphan_release_metadata(struct inode *inode)
3934
{
3935
	struct btrfs_root *root = BTRFS_I(inode)->root;
C
Chris Mason 已提交
3936
	u64 num_bytes = btrfs_calc_trans_metadata_size(root, 1);
3937 3938
	btrfs_block_rsv_release(root, root->orphan_block_rsv, num_bytes);
}
3939

3940 3941 3942 3943 3944 3945 3946 3947 3948 3949
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.
	 */
3950
	u64 num_bytes = btrfs_calc_trans_metadata_size(root, 5);
3951 3952
	dst_rsv->space_info = src_rsv->space_info;
	return block_rsv_migrate_bytes(src_rsv, dst_rsv, num_bytes);
3953 3954
}

3955 3956 3957 3958 3959 3960 3961 3962 3963
/**
 * 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.
 */
3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976
static unsigned drop_outstanding_extent(struct inode *inode)
{
	unsigned dropped_extents = 0;

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

	/*
	 * 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)
3977
		return 0;
3978 3979 3980 3981 3982 3983 3984

	dropped_extents = BTRFS_I(inode)->reserved_extents -
		BTRFS_I(inode)->outstanding_extents;
	BTRFS_I(inode)->reserved_extents -= dropped_extents;
	return dropped_extents;
}

3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004
/**
 * 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)
4005
{
4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040
	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);
4041
}
Y
Yan Zheng 已提交
4042

4043 4044 4045 4046
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;
4047 4048
	u64 to_reserve = 0;
	unsigned nr_extents = 0;
4049
	int flush = 1;
4050
	int ret;
4051

4052 4053 4054 4055
	if (btrfs_is_free_space_inode(root, inode))
		flush = 0;

	if (flush && btrfs_transaction_in_commit(root->fs_info))
4056
		schedule_timeout(1);
4057

4058
	num_bytes = ALIGN(num_bytes, root->sectorsize);
4059

4060 4061 4062 4063 4064 4065 4066 4067
	spin_lock(&BTRFS_I(inode)->lock);
	BTRFS_I(inode)->outstanding_extents++;

	if (BTRFS_I(inode)->outstanding_extents >
	    BTRFS_I(inode)->reserved_extents) {
		nr_extents = BTRFS_I(inode)->outstanding_extents -
			BTRFS_I(inode)->reserved_extents;
		BTRFS_I(inode)->reserved_extents += nr_extents;
4068

4069
		to_reserve = btrfs_calc_trans_metadata_size(root, nr_extents);
4070
	}
4071
	to_reserve += calc_csum_metadata_size(inode, num_bytes, 1);
4072
	spin_unlock(&BTRFS_I(inode)->lock);
4073

4074
	ret = reserve_metadata_bytes(root, block_rsv, to_reserve, flush, 0);
4075
	if (ret) {
4076
		u64 to_free = 0;
4077
		unsigned dropped;
4078

4079
		spin_lock(&BTRFS_I(inode)->lock);
4080
		dropped = drop_outstanding_extent(inode);
4081
		to_free = calc_csum_metadata_size(inode, num_bytes, 0);
4082
		spin_unlock(&BTRFS_I(inode)->lock);
4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093
		to_free += btrfs_calc_trans_metadata_size(root, dropped);

		/*
		 * Somebody could have come in and twiddled with the
		 * reservation, so if we have to free more than we would have
		 * reserved from this reservation go ahead and release those
		 * bytes.
		 */
		to_free -= to_reserve;
		if (to_free)
			btrfs_block_rsv_release(root, block_rsv, to_free);
4094
		return ret;
4095
	}
4096

4097 4098 4099 4100 4101
	block_rsv_add_bytes(block_rsv, to_reserve, 1);

	return 0;
}

4102 4103 4104 4105 4106 4107 4108 4109 4110
/**
 * 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.
 */
4111 4112 4113
void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
4114 4115
	u64 to_free = 0;
	unsigned dropped;
4116 4117

	num_bytes = ALIGN(num_bytes, root->sectorsize);
4118
	spin_lock(&BTRFS_I(inode)->lock);
4119
	dropped = drop_outstanding_extent(inode);
4120

4121 4122
	to_free = calc_csum_metadata_size(inode, num_bytes, 0);
	spin_unlock(&BTRFS_I(inode)->lock);
4123 4124
	if (dropped > 0)
		to_free += btrfs_calc_trans_metadata_size(root, dropped);
4125 4126 4127 4128 4129

	btrfs_block_rsv_release(root, &root->fs_info->delalloc_block_rsv,
				to_free);
}

4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144
/**
 * 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.
 */
4145 4146 4147 4148 4149
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
4150
	if (ret)
4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161
		return ret;

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

	return 0;
}

4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174
/**
 * 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.
 */
4175 4176 4177 4178
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);
4179 4180
}

C
Chris Mason 已提交
4181 4182
static int update_block_group(struct btrfs_trans_handle *trans,
			      struct btrfs_root *root,
4183
			      u64 bytenr, u64 num_bytes, int alloc)
C
Chris Mason 已提交
4184
{
4185
	struct btrfs_block_group_cache *cache = NULL;
C
Chris Mason 已提交
4186
	struct btrfs_fs_info *info = root->fs_info;
4187
	u64 total = num_bytes;
C
Chris Mason 已提交
4188
	u64 old_val;
4189
	u64 byte_in_group;
4190
	int factor;
C
Chris Mason 已提交
4191

4192 4193 4194 4195 4196 4197 4198 4199 4200 4201
	/* block accounting for super block */
	spin_lock(&info->delalloc_lock);
	old_val = btrfs_super_bytes_used(&info->super_copy);
	if (alloc)
		old_val += num_bytes;
	else
		old_val -= num_bytes;
	btrfs_set_super_bytes_used(&info->super_copy, old_val);
	spin_unlock(&info->delalloc_lock);

C
Chris Mason 已提交
4202
	while (total) {
4203
		cache = btrfs_lookup_block_group(info, bytenr);
4204
		if (!cache)
C
Chris Mason 已提交
4205
			return -1;
4206 4207 4208 4209 4210 4211
		if (cache->flags & (BTRFS_BLOCK_GROUP_DUP |
				    BTRFS_BLOCK_GROUP_RAID1 |
				    BTRFS_BLOCK_GROUP_RAID10))
			factor = 2;
		else
			factor = 1;
4212 4213 4214 4215 4216 4217 4218
		/*
		 * 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)
4219
			cache_block_group(cache, trans, NULL, 1);
4220

4221 4222
		byte_in_group = bytenr - cache->key.objectid;
		WARN_ON(byte_in_group > cache->key.offset);
C
Chris Mason 已提交
4223

4224
		spin_lock(&cache->space_info->lock);
4225
		spin_lock(&cache->lock);
4226

4227
		if (btrfs_test_opt(root, SPACE_CACHE) &&
4228 4229 4230
		    cache->disk_cache_state < BTRFS_DC_CLEAR)
			cache->disk_cache_state = BTRFS_DC_CLEAR;

J
Josef Bacik 已提交
4231
		cache->dirty = 1;
C
Chris Mason 已提交
4232
		old_val = btrfs_block_group_used(&cache->item);
4233
		num_bytes = min(total, cache->key.offset - byte_in_group);
C
Chris Mason 已提交
4234
		if (alloc) {
4235
			old_val += num_bytes;
4236 4237 4238
			btrfs_set_block_group_used(&cache->item, old_val);
			cache->reserved -= num_bytes;
			cache->space_info->bytes_reserved -= num_bytes;
4239 4240
			cache->space_info->bytes_used += num_bytes;
			cache->space_info->disk_used += num_bytes * factor;
4241
			spin_unlock(&cache->lock);
4242
			spin_unlock(&cache->space_info->lock);
C
Chris Mason 已提交
4243
		} else {
4244
			old_val -= num_bytes;
4245
			btrfs_set_block_group_used(&cache->item, old_val);
4246 4247
			cache->pinned += num_bytes;
			cache->space_info->bytes_pinned += num_bytes;
4248
			cache->space_info->bytes_used -= num_bytes;
4249
			cache->space_info->disk_used -= num_bytes * factor;
4250
			spin_unlock(&cache->lock);
4251
			spin_unlock(&cache->space_info->lock);
4252

4253 4254 4255
			set_extent_dirty(info->pinned_extents,
					 bytenr, bytenr + num_bytes - 1,
					 GFP_NOFS | __GFP_NOFAIL);
C
Chris Mason 已提交
4256
		}
4257
		btrfs_put_block_group(cache);
4258 4259
		total -= num_bytes;
		bytenr += num_bytes;
C
Chris Mason 已提交
4260 4261 4262
	}
	return 0;
}
4263

4264 4265
static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
{
J
Josef Bacik 已提交
4266
	struct btrfs_block_group_cache *cache;
4267
	u64 bytenr;
J
Josef Bacik 已提交
4268 4269 4270

	cache = btrfs_lookup_first_block_group(root->fs_info, search_start);
	if (!cache)
4271
		return 0;
J
Josef Bacik 已提交
4272

4273
	bytenr = cache->key.objectid;
4274
	btrfs_put_block_group(cache);
4275 4276

	return bytenr;
4277 4278
}

4279 4280 4281
static int pin_down_extent(struct btrfs_root *root,
			   struct btrfs_block_group_cache *cache,
			   u64 bytenr, u64 num_bytes, int reserved)
4282
{
4283 4284 4285 4286 4287 4288 4289 4290 4291 4292
	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
4293

4294 4295 4296 4297
	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
4298

4299 4300 4301 4302 4303 4304 4305
/*
 * 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
4306

4307 4308 4309 4310 4311 4312
	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);
4313 4314 4315
	return 0;
}

4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336
/**
 * 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.
4337
 */
4338 4339
static int btrfs_update_reserved_bytes(struct btrfs_block_group_cache *cache,
				       u64 num_bytes, int reserve)
4340
{
4341
	struct btrfs_space_info *space_info = cache->space_info;
4342
	int ret = 0;
4343 4344 4345
	spin_lock(&space_info->lock);
	spin_lock(&cache->lock);
	if (reserve != RESERVE_FREE) {
4346 4347 4348
		if (cache->ro) {
			ret = -EAGAIN;
		} else {
4349 4350 4351 4352 4353 4354
			cache->reserved += num_bytes;
			space_info->bytes_reserved += num_bytes;
			if (reserve == RESERVE_ALLOC) {
				BUG_ON(space_info->bytes_may_use < num_bytes);
				space_info->bytes_may_use -= num_bytes;
			}
4355
		}
4356 4357 4358 4359 4360 4361
	} else {
		if (cache->ro)
			space_info->bytes_readonly += num_bytes;
		cache->reserved -= num_bytes;
		space_info->bytes_reserved -= num_bytes;
		space_info->reservation_progress++;
4362
	}
4363 4364
	spin_unlock(&cache->lock);
	spin_unlock(&space_info->lock);
4365
	return ret;
4366
}
C
Chris Mason 已提交
4367

4368 4369
int btrfs_prepare_extent_commit(struct btrfs_trans_handle *trans,
				struct btrfs_root *root)
4370 4371
{
	struct btrfs_fs_info *fs_info = root->fs_info;
4372 4373 4374
	struct btrfs_caching_control *next;
	struct btrfs_caching_control *caching_ctl;
	struct btrfs_block_group_cache *cache;
4375

4376
	down_write(&fs_info->extent_commit_sem);
4377

4378 4379 4380 4381 4382 4383 4384
	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);
4385
		} else {
4386
			cache->last_byte_to_unpin = caching_ctl->progress;
4387 4388
		}
	}
4389 4390 4391 4392 4393 4394 4395

	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);
4396 4397

	update_global_block_rsv(fs_info);
4398 4399 4400
	return 0;
}

4401
static int unpin_extent_range(struct btrfs_root *root, u64 start, u64 end)
C
Chris Mason 已提交
4402
{
4403 4404 4405
	struct btrfs_fs_info *fs_info = root->fs_info;
	struct btrfs_block_group_cache *cache = NULL;
	u64 len;
C
Chris Mason 已提交
4406

4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423
	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);
		}

4424 4425
		start += len;

4426 4427 4428 4429
		spin_lock(&cache->space_info->lock);
		spin_lock(&cache->lock);
		cache->pinned -= len;
		cache->space_info->bytes_pinned -= len;
4430
		if (cache->ro)
4431
			cache->space_info->bytes_readonly += len;
4432 4433
		spin_unlock(&cache->lock);
		spin_unlock(&cache->space_info->lock);
C
Chris Mason 已提交
4434
	}
4435 4436 4437

	if (cache)
		btrfs_put_block_group(cache);
C
Chris Mason 已提交
4438 4439 4440 4441
	return 0;
}

int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
4442
			       struct btrfs_root *root)
4443
{
4444 4445
	struct btrfs_fs_info *fs_info = root->fs_info;
	struct extent_io_tree *unpin;
4446 4447
	u64 start;
	u64 end;
4448 4449
	int ret;

4450 4451 4452 4453 4454
	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
4455
	while (1) {
4456 4457 4458
		ret = find_first_extent_bit(unpin, 0, &start, &end,
					    EXTENT_DIRTY);
		if (ret)
4459
			break;
4460

4461 4462 4463
		if (btrfs_test_opt(root, DISCARD))
			ret = btrfs_discard_extent(root, start,
						   end + 1 - start, NULL);
4464

4465
		clear_extent_dirty(unpin, start, end, GFP_NOFS);
4466
		unpin_extent_range(root, start, end);
4467
		cond_resched();
4468
	}
J
Josef Bacik 已提交
4469

C
Chris Mason 已提交
4470 4471 4472
	return 0;
}

4473 4474 4475 4476 4477 4478
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)
4479
{
C
Chris Mason 已提交
4480
	struct btrfs_key key;
4481
	struct btrfs_path *path;
4482 4483
	struct btrfs_fs_info *info = root->fs_info;
	struct btrfs_root *extent_root = info->extent_root;
4484
	struct extent_buffer *leaf;
4485 4486
	struct btrfs_extent_item *ei;
	struct btrfs_extent_inline_ref *iref;
4487
	int ret;
4488
	int is_data;
4489 4490 4491
	int extent_slot = 0;
	int found_extent = 0;
	int num_to_del = 1;
4492 4493
	u32 item_size;
	u64 refs;
C
Chris Mason 已提交
4494

4495
	path = btrfs_alloc_path();
4496 4497
	if (!path)
		return -ENOMEM;
4498

4499
	path->reada = 1;
4500
	path->leave_spinning = 1;
4501 4502 4503 4504 4505 4506 4507 4508

	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);
4509
	if (ret == 0) {
4510
		extent_slot = path->slots[0];
4511 4512
		while (extent_slot >= 0) {
			btrfs_item_key_to_cpu(path->nodes[0], &key,
4513
					      extent_slot);
4514
			if (key.objectid != bytenr)
4515
				break;
4516 4517
			if (key.type == BTRFS_EXTENT_ITEM_KEY &&
			    key.offset == num_bytes) {
4518 4519 4520 4521 4522
				found_extent = 1;
				break;
			}
			if (path->slots[0] - extent_slot > 5)
				break;
4523
			extent_slot--;
4524
		}
4525 4526 4527 4528 4529
#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
4530
		if (!found_extent) {
4531
			BUG_ON(iref);
4532
			ret = remove_extent_backref(trans, extent_root, path,
4533 4534
						    NULL, refs_to_drop,
						    is_data);
Z
Zheng Yan 已提交
4535
			BUG_ON(ret);
4536
			btrfs_release_path(path);
4537
			path->leave_spinning = 1;
4538 4539 4540 4541 4542

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

Z
Zheng Yan 已提交
4543 4544
			ret = btrfs_search_slot(trans, extent_root,
						&key, path, -1, 1);
4545 4546
			if (ret) {
				printk(KERN_ERR "umm, got %d back from search"
C
Chris Mason 已提交
4547 4548
				       ", was looking for %llu\n", ret,
				       (unsigned long long)bytenr);
4549 4550 4551
				if (ret > 0)
					btrfs_print_leaf(extent_root,
							 path->nodes[0]);
4552
			}
Z
Zheng Yan 已提交
4553 4554 4555
			BUG_ON(ret);
			extent_slot = path->slots[0];
		}
4556 4557 4558
	} else {
		btrfs_print_leaf(extent_root, path->nodes[0]);
		WARN_ON(1);
C
Chris Mason 已提交
4559
		printk(KERN_ERR "btrfs unable to find ref byte nr %llu "
4560
		       "parent %llu root %llu  owner %llu offset %llu\n",
C
Chris Mason 已提交
4561
		       (unsigned long long)bytenr,
4562
		       (unsigned long long)parent,
C
Chris Mason 已提交
4563
		       (unsigned long long)root_objectid,
4564 4565
		       (unsigned long long)owner_objectid,
		       (unsigned long long)owner_offset);
4566
	}
4567 4568

	leaf = path->nodes[0];
4569 4570 4571 4572 4573 4574 4575 4576
	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);

4577
		btrfs_release_path(path);
4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598
		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));
4599
	ei = btrfs_item_ptr(leaf, extent_slot,
C
Chris Mason 已提交
4600
			    struct btrfs_extent_item);
4601 4602 4603 4604 4605 4606
	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));
	}
4607

4608
	refs = btrfs_extent_refs(leaf, ei);
4609 4610
	BUG_ON(refs < refs_to_drop);
	refs -= refs_to_drop;
4611

4612 4613 4614 4615 4616 4617
	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
4618
		 */
4619 4620 4621 4622 4623 4624 4625 4626 4627 4628
		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);
4629 4630
			BUG_ON(ret);
		}
4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641
	} 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
4642
		}
4643

4644 4645
		ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
				      num_to_del);
Z
Zheng Yan 已提交
4646
		BUG_ON(ret);
4647
		btrfs_release_path(path);
4648

4649
		if (is_data) {
4650 4651
			ret = btrfs_del_csums(trans, root, bytenr, num_bytes);
			BUG_ON(ret);
4652 4653 4654 4655
		} else {
			invalidate_mapping_pages(info->btree_inode->i_mapping,
			     bytenr >> PAGE_CACHE_SHIFT,
			     (bytenr + num_bytes - 1) >> PAGE_CACHE_SHIFT);
4656 4657
		}

4658
		ret = update_block_group(trans, root, bytenr, num_bytes, 0);
4659
		BUG_ON(ret);
4660
	}
4661
	btrfs_free_path(path);
4662 4663 4664
	return ret;
}

4665
/*
4666
 * when we free an block, it is possible (and likely) that we free the last
4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677
 * 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;
4678
	int ret = 0;
4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694 4695

	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;

4696 4697 4698 4699 4700 4701 4702
	if (head->extent_op) {
		if (!head->must_insert_reserved)
			goto out;
		kfree(head->extent_op);
		head->extent_op = NULL;
	}

4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715
	/*
	 * 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);
4716

4717 4718 4719 4720 4721 4722
	delayed_refs->num_entries--;

	/*
	 * 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.
	 */
4723 4724 4725 4726 4727
	delayed_refs->num_heads--;
	if (list_empty(&head->cluster))
		delayed_refs->num_heads_ready--;

	list_del_init(&head->cluster);
4728 4729
	spin_unlock(&delayed_refs->lock);

4730 4731 4732 4733 4734
	BUG_ON(head->extent_op);
	if (head->must_insert_reserved)
		ret = 1;

	mutex_unlock(&head->mutex);
4735
	btrfs_put_delayed_ref(&head->node);
4736
	return ret;
4737 4738 4739 4740 4741
out:
	spin_unlock(&delayed_refs->lock);
	return 0;
}

4742 4743 4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765 4766
void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
			   struct btrfs_root *root,
			   struct extent_buffer *buf,
			   u64 parent, int last_ref)
{
	struct btrfs_block_group_cache *cache = NULL;
	int ret;

	if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
		ret = btrfs_add_delayed_tree_ref(trans, buf->start, buf->len,
						parent, root->root_key.objectid,
						btrfs_header_level(buf),
						BTRFS_DROP_DELAYED_REF, NULL);
		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)
4767
				goto out;
4768 4769 4770 4771
		}

		if (btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
			pin_down_extent(root, cache, buf->start, buf->len, 1);
4772
			goto out;
4773 4774 4775 4776 4777
		}

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

		btrfs_add_free_space(cache, buf->start, buf->len);
4778
		btrfs_update_reserved_bytes(cache, buf->len, RESERVE_FREE);
4779 4780
	}
out:
4781 4782 4783 4784 4785
	/*
	 * Deleting the buffer, clear the corrupt flag since it doesn't matter
	 * anymore.
	 */
	clear_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags);
4786 4787 4788
	btrfs_put_block_group(cache);
}

4789
int btrfs_free_extent(struct btrfs_trans_handle *trans,
Z
Zheng Yan 已提交
4790 4791
		      struct btrfs_root *root,
		      u64 bytenr, u64 num_bytes, u64 parent,
4792
		      u64 root_objectid, u64 owner, u64 offset)
4793 4794 4795
{
	int ret;

4796 4797 4798 4799
	/*
	 * tree log blocks never actually go into the extent allocation
	 * tree, just update pinning info and exit early.
	 */
4800 4801
	if (root_objectid == BTRFS_TREE_LOG_OBJECTID) {
		WARN_ON(owner >= BTRFS_FIRST_FREE_OBJECTID);
4802
		/* unlocks the pinned mutex */
4803
		btrfs_pin_extent(root, bytenr, num_bytes, 1);
4804
		ret = 0;
4805 4806 4807 4808
	} else if (owner < BTRFS_FIRST_FREE_OBJECTID) {
		ret = btrfs_add_delayed_tree_ref(trans, bytenr, num_bytes,
					parent, root_objectid, (int)owner,
					BTRFS_DROP_DELAYED_REF, NULL);
4809
		BUG_ON(ret);
4810 4811 4812 4813 4814
	} else {
		ret = btrfs_add_delayed_data_ref(trans, bytenr, num_bytes,
					parent, root_objectid, owner,
					offset, BTRFS_DROP_DELAYED_REF, NULL);
		BUG_ON(ret);
4815
	}
4816 4817 4818
	return ret;
}

4819 4820 4821 4822 4823 4824 4825
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
4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840
/*
 * 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)
{
4841
	struct btrfs_caching_control *caching_ctl;
J
Josef Bacik 已提交
4842 4843
	DEFINE_WAIT(wait);

4844 4845
	caching_ctl = get_caching_control(cache);
	if (!caching_ctl)
J
Josef Bacik 已提交
4846 4847
		return 0;

4848
	wait_event(caching_ctl->wait, block_group_cache_done(cache) ||
4849
		   (cache->free_space_ctl->free_space >= num_bytes));
4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867

	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
4868 4869 4870
	return 0;
}

4871 4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886
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
4887
enum btrfs_loop_type {
4888
	LOOP_FIND_IDEAL = 0,
J
Josef Bacik 已提交
4889 4890 4891 4892 4893 4894
	LOOP_CACHING_NOWAIT = 1,
	LOOP_CACHING_WAIT = 2,
	LOOP_ALLOC_CHUNK = 3,
	LOOP_NO_EMPTY_SIZE = 4,
};

4895 4896 4897 4898
/*
 * 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
4899
 * ins->flags = BTRFS_EXTENT_ITEM_KEY
4900 4901 4902
 * ins->offset == number of blocks
 * Any available blocks before search_start are skipped.
 */
C
Chris Mason 已提交
4903
static noinline int find_free_extent(struct btrfs_trans_handle *trans,
4904 4905 4906 4907
				     struct btrfs_root *orig_root,
				     u64 num_bytes, u64 empty_size,
				     u64 search_start, u64 search_end,
				     u64 hint_byte, struct btrfs_key *ins,
4908
				     u64 data)
4909
{
4910
	int ret = 0;
C
Chris Mason 已提交
4911
	struct btrfs_root *root = orig_root->fs_info->extent_root;
4912
	struct btrfs_free_cluster *last_ptr = NULL;
4913
	struct btrfs_block_group_cache *block_group = NULL;
4914
	int empty_cluster = 2 * 1024 * 1024;
C
Chris Mason 已提交
4915
	int allowed_chunk_alloc = 0;
4916
	int done_chunk_alloc = 0;
4917
	struct btrfs_space_info *space_info;
4918 4919
	int last_ptr_loop = 0;
	int loop = 0;
4920
	int index = 0;
4921 4922
	int alloc_type = (data & BTRFS_BLOCK_GROUP_DATA) ?
		RESERVE_ALLOC_NO_ACCOUNT : RESERVE_ALLOC;
J
Josef Bacik 已提交
4923
	bool found_uncached_bg = false;
4924
	bool failed_cluster_refill = false;
4925
	bool failed_alloc = false;
4926
	bool use_cluster = true;
4927 4928
	u64 ideal_cache_percent = 0;
	u64 ideal_cache_offset = 0;
4929

4930
	WARN_ON(num_bytes < root->sectorsize);
4931
	btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
4932 4933
	ins->objectid = 0;
	ins->offset = 0;
4934

J
Josef Bacik 已提交
4935
	space_info = __find_space_info(root->fs_info, data);
4936
	if (!space_info) {
4937
		printk(KERN_ERR "No space info for %llu\n", data);
4938 4939
		return -ENOSPC;
	}
J
Josef Bacik 已提交
4940

4941 4942 4943 4944 4945 4946 4947
	/*
	 * 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
4948 4949 4950
	if (orig_root->ref_cows || empty_size)
		allowed_chunk_alloc = 1;

4951
	if (data & BTRFS_BLOCK_GROUP_METADATA && use_cluster) {
4952
		last_ptr = &root->fs_info->meta_alloc_cluster;
4953 4954
		if (!btrfs_test_opt(root, SSD))
			empty_cluster = 64 * 1024;
4955 4956
	}

4957 4958
	if ((data & BTRFS_BLOCK_GROUP_DATA) && use_cluster &&
	    btrfs_test_opt(root, SSD)) {
4959 4960
		last_ptr = &root->fs_info->data_alloc_cluster;
	}
J
Josef Bacik 已提交
4961

4962
	if (last_ptr) {
4963 4964 4965 4966
		spin_lock(&last_ptr->lock);
		if (last_ptr->block_group)
			hint_byte = last_ptr->window_start;
		spin_unlock(&last_ptr->lock);
4967
	}
4968

4969
	search_start = max(search_start, first_logical_byte(root, 0));
4970
	search_start = max(search_start, hint_byte);
4971

J
Josef Bacik 已提交
4972
	if (!last_ptr)
4973 4974
		empty_cluster = 0;

J
Josef Bacik 已提交
4975
	if (search_start == hint_byte) {
4976
ideal_cache:
J
Josef Bacik 已提交
4977 4978
		block_group = btrfs_lookup_block_group(root->fs_info,
						       search_start);
J
Josef Bacik 已提交
4979 4980 4981
		/*
		 * we don't want to use the block group if it doesn't match our
		 * allocation bits, or if its not cached.
4982 4983 4984
		 *
		 * 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
4985 4986
		 */
		if (block_group && block_group_bits(block_group, data) &&
4987 4988
		    (block_group->cached != BTRFS_CACHE_NO ||
		     search_start == ideal_cache_offset)) {
J
Josef Bacik 已提交
4989
			down_read(&space_info->groups_sem);
4990 4991 4992 4993 4994 4995 4996 4997 4998 4999
			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);
5000
			} else {
5001
				index = get_block_group_index(block_group);
5002
				goto have_block_group;
5003
			}
J
Josef Bacik 已提交
5004
		} else if (block_group) {
5005
			btrfs_put_block_group(block_group);
J
Josef Bacik 已提交
5006
		}
5007
	}
J
Josef Bacik 已提交
5008
search:
5009
	down_read(&space_info->groups_sem);
5010 5011
	list_for_each_entry(block_group, &space_info->block_groups[index],
			    list) {
5012
		u64 offset;
J
Josef Bacik 已提交
5013
		int cached;
5014

5015
		btrfs_get_block_group(block_group);
J
Josef Bacik 已提交
5016
		search_start = block_group->key.objectid;
5017

5018 5019 5020 5021 5022 5023 5024 5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036
		/*
		 * 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
5037
have_block_group:
J
Josef Bacik 已提交
5038
		if (unlikely(block_group->cached == BTRFS_CACHE_NO)) {
5039 5040
			u64 free_percent;

5041 5042
			ret = cache_block_group(block_group, trans,
						orig_root, 1);
5043 5044 5045
			if (block_group->cached == BTRFS_CACHE_FINISHED)
				goto have_block_group;

5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056
			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
5057
			/*
5058 5059
			 * 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
5060
			 */
5061
			if (loop > LOOP_FIND_IDEAL) {
5062 5063
				ret = cache_block_group(block_group, trans,
							orig_root, 0);
J
Josef Bacik 已提交
5064
				BUG_ON(ret);
J
Josef Bacik 已提交
5065
			}
J
Josef Bacik 已提交
5066 5067
			found_uncached_bg = true;

5068 5069 5070 5071 5072
			/*
			 * If loop is set for cached only, try the next block
			 * group.
			 */
			if (loop == LOOP_FIND_IDEAL)
J
Josef Bacik 已提交
5073 5074 5075
				goto loop;
		}

5076 5077 5078 5079
		cached = block_group_cache_done(block_group);
		if (unlikely(!cached))
			found_uncached_bg = true;

5080
		if (unlikely(block_group->ro))
J
Josef Bacik 已提交
5081
			goto loop;
J
Josef Bacik 已提交
5082

C
Chris Mason 已提交
5083
		spin_lock(&block_group->free_space_ctl->tree_lock);
5084
		if (cached &&
C
Chris Mason 已提交
5085 5086 5087
		    block_group->free_space_ctl->free_space <
		    num_bytes + empty_size) {
			spin_unlock(&block_group->free_space_ctl->tree_lock);
5088 5089
			goto loop;
		}
C
Chris Mason 已提交
5090
		spin_unlock(&block_group->free_space_ctl->tree_lock);
5091

5092 5093 5094 5095 5096 5097 5098 5099 5100 5101
		/*
		 * Ok we want to try and use the cluster allocator, so lets look
		 * there, unless we are on LOOP_NO_EMPTY_SIZE, since we will
		 * have tried the cluster allocator plenty of times at this
		 * point and not have found anything, so we are likely way too
		 * fragmented for the clustering stuff to find anything, so lets
		 * just skip it and let the allocator find whatever block it can
		 * find
		 */
		if (last_ptr && loop < LOOP_NO_EMPTY_SIZE) {
5102 5103 5104 5105 5106
			/*
			 * the refill lock keeps out other
			 * people trying to start a new cluster
			 */
			spin_lock(&last_ptr->refill_lock);
5107 5108 5109 5110 5111 5112 5113
			if (last_ptr->block_group &&
			    (last_ptr->block_group->ro ||
			    !block_group_bits(last_ptr->block_group, data))) {
				offset = 0;
				goto refill_cluster;
			}

5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128
			offset = btrfs_alloc_from_cluster(block_group, last_ptr,
						 num_bytes, search_start);
			if (offset) {
				/* we have a block, we're done */
				spin_unlock(&last_ptr->refill_lock);
				goto checks;
			}

			spin_lock(&last_ptr->lock);
			/*
			 * whoops, this cluster doesn't actually point to
			 * this block group.  Get a ref on the block
			 * group is does point to and try again
			 */
			if (!last_ptr_loop && last_ptr->block_group &&
5129 5130 5131
			    last_ptr->block_group != block_group &&
			    index <=
				 get_block_group_index(last_ptr->block_group)) {
5132 5133 5134

				btrfs_put_block_group(block_group);
				block_group = last_ptr->block_group;
5135
				btrfs_get_block_group(block_group);
5136 5137 5138 5139 5140
				spin_unlock(&last_ptr->lock);
				spin_unlock(&last_ptr->refill_lock);

				last_ptr_loop = 1;
				search_start = block_group->key.objectid;
5141 5142 5143 5144 5145 5146 5147
				/*
				 * we know this block group is properly
				 * in the list because
				 * btrfs_remove_block_group, drops the
				 * cluster before it removes the block
				 * group from the list
				 */
5148 5149 5150
				goto have_block_group;
			}
			spin_unlock(&last_ptr->lock);
5151
refill_cluster:
5152 5153 5154 5155 5156 5157 5158 5159 5160
			/*
			 * this cluster didn't work out, free it and
			 * start over
			 */
			btrfs_return_cluster_to_free_space(NULL, last_ptr);

			last_ptr_loop = 0;

			/* allocate a cluster in this block group */
5161
			ret = btrfs_find_space_cluster(trans, root,
5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176 5177
					       block_group, last_ptr,
					       offset, num_bytes,
					       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);
					goto checks;
				}
5178 5179
			} else if (!cached && loop > LOOP_CACHING_NOWAIT
				   && !failed_cluster_refill) {
J
Josef Bacik 已提交
5180 5181
				spin_unlock(&last_ptr->refill_lock);

5182
				failed_cluster_refill = true;
J
Josef Bacik 已提交
5183 5184 5185
				wait_block_group_cache_progress(block_group,
				       num_bytes + empty_cluster + empty_size);
				goto have_block_group;
5186
			}
J
Josef Bacik 已提交
5187

5188 5189 5190 5191 5192 5193
			/*
			 * 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
			 */
5194
			btrfs_return_cluster_to_free_space(NULL, last_ptr);
5195
			spin_unlock(&last_ptr->refill_lock);
5196
			goto loop;
5197 5198
		}

5199 5200
		offset = btrfs_find_space_for_alloc(block_group, search_start,
						    num_bytes, empty_size);
5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211
		/*
		 * 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
5212
			wait_block_group_cache_progress(block_group,
5213 5214
						num_bytes + empty_size);
			failed_alloc = true;
J
Josef Bacik 已提交
5215
			goto have_block_group;
5216 5217
		} else if (!offset) {
			goto loop;
J
Josef Bacik 已提交
5218
		}
5219
checks:
5220
		search_start = stripe_align(root, offset);
J
Josef Bacik 已提交
5221
		/* move on to the next group */
5222 5223
		if (search_start + num_bytes >= search_end) {
			btrfs_add_free_space(block_group, offset, num_bytes);
J
Josef Bacik 已提交
5224
			goto loop;
5225
		}
5226

J
Josef Bacik 已提交
5227 5228
		/* move on to the next group */
		if (search_start + num_bytes >
5229 5230
		    block_group->key.objectid + block_group->key.offset) {
			btrfs_add_free_space(block_group, offset, num_bytes);
J
Josef Bacik 已提交
5231
			goto loop;
5232
		}
5233

5234 5235
		ins->objectid = search_start;
		ins->offset = num_bytes;
J
Josef Bacik 已提交
5236

5237 5238 5239 5240
		if (offset < search_start)
			btrfs_add_free_space(block_group, offset,
					     search_start - offset);
		BUG_ON(offset > search_start);
J
Josef Bacik 已提交
5241

5242 5243
		ret = btrfs_update_reserved_bytes(block_group, num_bytes,
						  alloc_type);
5244
		if (ret == -EAGAIN) {
5245
			btrfs_add_free_space(block_group, offset, num_bytes);
J
Josef Bacik 已提交
5246
			goto loop;
J
Josef Bacik 已提交
5247
		}
5248

5249
		/* we are all good, lets return */
J
Josef Bacik 已提交
5250 5251
		ins->objectid = search_start;
		ins->offset = num_bytes;
5252

5253 5254 5255 5256
		if (offset < search_start)
			btrfs_add_free_space(block_group, offset,
					     search_start - offset);
		BUG_ON(offset > search_start);
5257
		btrfs_put_block_group(block_group);
J
Josef Bacik 已提交
5258 5259
		break;
loop:
5260
		failed_cluster_refill = false;
5261
		failed_alloc = false;
5262
		BUG_ON(index != get_block_group_index(block_group));
5263
		btrfs_put_block_group(block_group);
J
Josef Bacik 已提交
5264 5265 5266
	}
	up_read(&space_info->groups_sem);

5267 5268 5269
	if (!ins->objectid && ++index < BTRFS_NR_RAID_TYPES)
		goto search;

5270 5271 5272 5273 5274
	/* 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
5275 5276 5277 5278
	 * 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
5279
	 */
5280
	if (!ins->objectid && loop < LOOP_NO_EMPTY_SIZE) {
5281
		index = 0;
5282
		if (loop == LOOP_FIND_IDEAL && found_uncached_bg) {
J
Josef Bacik 已提交
5283
			found_uncached_bg = false;
5284
			loop++;
5285
			if (!ideal_cache_percent)
J
Josef Bacik 已提交
5286
				goto search;
5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313 5314 5315 5316 5317 5318

			/*
			 * 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;
		}

5319
		loop++;
J
Josef Bacik 已提交
5320 5321

		if (loop == LOOP_ALLOC_CHUNK) {
5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332 5333
		       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
5334

5335 5336 5337 5338 5339 5340
		       /*
			* 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
5341 5342
		}

5343 5344 5345
		if (loop == LOOP_NO_EMPTY_SIZE) {
			empty_size = 0;
			empty_cluster = 0;
5346
		}
5347 5348

		goto search;
J
Josef Bacik 已提交
5349 5350
	} else if (!ins->objectid) {
		ret = -ENOSPC;
5351
	} else if (ins->objectid) {
5352
		ret = 0;
C
Chris Mason 已提交
5353 5354
	}

C
Chris Mason 已提交
5355
	return ret;
5356
}
5357

J
Josef Bacik 已提交
5358 5359
static void dump_space_info(struct btrfs_space_info *info, u64 bytes,
			    int dump_block_groups)
J
Josef Bacik 已提交
5360 5361
{
	struct btrfs_block_group_cache *cache;
5362
	int index = 0;
J
Josef Bacik 已提交
5363

J
Josef Bacik 已提交
5364
	spin_lock(&info->lock);
5365 5366
	printk(KERN_INFO "space_info %llu has %llu free, is %sfull\n",
	       (unsigned long long)info->flags,
C
Chris Mason 已提交
5367
	       (unsigned long long)(info->total_bytes - info->bytes_used -
J
Josef Bacik 已提交
5368
				    info->bytes_pinned - info->bytes_reserved -
5369
				    info->bytes_readonly),
C
Chris Mason 已提交
5370
	       (info->full) ? "" : "not ");
5371 5372
	printk(KERN_INFO "space_info total=%llu, used=%llu, pinned=%llu, "
	       "reserved=%llu, may_use=%llu, readonly=%llu\n",
5373
	       (unsigned long long)info->total_bytes,
5374
	       (unsigned long long)info->bytes_used,
5375
	       (unsigned long long)info->bytes_pinned,
5376
	       (unsigned long long)info->bytes_reserved,
5377
	       (unsigned long long)info->bytes_may_use,
5378
	       (unsigned long long)info->bytes_readonly);
J
Josef Bacik 已提交
5379 5380 5381 5382
	spin_unlock(&info->lock);

	if (!dump_block_groups)
		return;
J
Josef Bacik 已提交
5383

5384
	down_read(&info->groups_sem);
5385 5386
again:
	list_for_each_entry(cache, &info->block_groups[index], list) {
J
Josef Bacik 已提交
5387
		spin_lock(&cache->lock);
C
Chris Mason 已提交
5388 5389 5390 5391 5392 5393 5394
		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
5395 5396 5397
		btrfs_dump_free_space(cache, bytes);
		spin_unlock(&cache->lock);
	}
5398 5399
	if (++index < BTRFS_NR_RAID_TYPES)
		goto again;
5400
	up_read(&info->groups_sem);
J
Josef Bacik 已提交
5401
}
5402

5403 5404 5405 5406 5407 5408
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)
5409 5410
{
	int ret;
5411
	u64 search_start = 0;
5412

J
Josef Bacik 已提交
5413
	data = btrfs_get_alloc_profile(root, data);
5414
again:
C
Chris Mason 已提交
5415 5416 5417 5418
	/*
	 * the only place that sets empty_size is btrfs_realloc_node, which
	 * is not called recursively on allocations
	 */
J
Josef Bacik 已提交
5419
	if (empty_size || root->ref_cows)
5420
		ret = do_chunk_alloc(trans, root->fs_info->extent_root,
5421 5422
				     num_bytes + 2 * 1024 * 1024, data,
				     CHUNK_ALLOC_NO_FORCE);
5423

5424 5425
	WARN_ON(num_bytes < root->sectorsize);
	ret = find_free_extent(trans, root, num_bytes, empty_size,
5426 5427
			       search_start, search_end, hint_byte,
			       ins, data);
5428

5429 5430
	if (ret == -ENOSPC && num_bytes > min_alloc_size) {
		num_bytes = num_bytes >> 1;
J
Josef Bacik 已提交
5431
		num_bytes = num_bytes & ~(root->sectorsize - 1);
5432
		num_bytes = max(num_bytes, min_alloc_size);
C
Chris Mason 已提交
5433
		do_chunk_alloc(trans, root->fs_info->extent_root,
5434
			       num_bytes, data, CHUNK_ALLOC_FORCE);
5435 5436
		goto again;
	}
5437
	if (ret == -ENOSPC && btrfs_test_opt(root, ENOSPC_DEBUG)) {
J
Josef Bacik 已提交
5438 5439 5440
		struct btrfs_space_info *sinfo;

		sinfo = __find_space_info(root->fs_info, data);
C
Chris Mason 已提交
5441 5442 5443
		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
5444
		dump_space_info(sinfo, num_bytes, 1);
5445
	}
J
Josef Bacik 已提交
5446

5447 5448
	trace_btrfs_reserved_extent_alloc(root, ins->objectid, ins->offset);

J
Josef Bacik 已提交
5449
	return ret;
5450 5451
}

5452 5453
int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len)
{
J
Josef Bacik 已提交
5454
	struct btrfs_block_group_cache *cache;
5455
	int ret = 0;
J
Josef Bacik 已提交
5456 5457 5458

	cache = btrfs_lookup_block_group(root->fs_info, start);
	if (!cache) {
C
Chris Mason 已提交
5459 5460
		printk(KERN_ERR "Unable to find block group for %llu\n",
		       (unsigned long long)start);
J
Josef Bacik 已提交
5461 5462
		return -ENOSPC;
	}
5463

5464 5465
	if (btrfs_test_opt(root, DISCARD))
		ret = btrfs_discard_extent(root, start, len, NULL);
5466

J
Josef Bacik 已提交
5467
	btrfs_add_free_space(cache, start, len);
5468
	btrfs_update_reserved_bytes(cache, len, RESERVE_FREE);
5469
	btrfs_put_block_group(cache);
J
Josef Bacik 已提交
5470

5471 5472
	trace_btrfs_reserved_extent_free(root, start, len);

5473 5474 5475
	return ret;
}

5476 5477 5478 5479 5480
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)
5481 5482
{
	int ret;
5483
	struct btrfs_fs_info *fs_info = root->fs_info;
5484
	struct btrfs_extent_item *extent_item;
5485
	struct btrfs_extent_inline_ref *iref;
5486
	struct btrfs_path *path;
5487 5488 5489
	struct extent_buffer *leaf;
	int type;
	u32 size;
5490

5491 5492 5493 5494
	if (parent > 0)
		type = BTRFS_SHARED_DATA_REF_KEY;
	else
		type = BTRFS_EXTENT_DATA_REF_KEY;
5495

5496
	size = sizeof(*extent_item) + btrfs_extent_inline_ref_size(type);
5497 5498

	path = btrfs_alloc_path();
T
Tsutomu Itoh 已提交
5499 5500
	if (!path)
		return -ENOMEM;
5501

5502
	path->leave_spinning = 1;
5503 5504
	ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
				      ins, size);
C
Chris Mason 已提交
5505
	BUG_ON(ret);
J
Josef Bacik 已提交
5506

5507 5508
	leaf = path->nodes[0];
	extent_item = btrfs_item_ptr(leaf, path->slots[0],
5509
				     struct btrfs_extent_item);
5510 5511 5512 5513 5514 5515 5516 5517 5518 5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529
	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);
	}
5530 5531

	btrfs_mark_buffer_dirty(path->nodes[0]);
5532
	btrfs_free_path(path);
5533

5534
	ret = update_block_group(trans, root, ins->objectid, ins->offset, 1);
5535
	if (ret) {
C
Chris Mason 已提交
5536 5537 5538
		printk(KERN_ERR "btrfs update block group failed for %llu "
		       "%llu\n", (unsigned long long)ins->objectid,
		       (unsigned long long)ins->offset);
5539 5540
		BUG();
	}
5541 5542 5543
	return ret;
}

5544 5545 5546 5547 5548
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)
5549 5550
{
	int ret;
5551 5552 5553 5554 5555 5556 5557
	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);
5558

5559
	path = btrfs_alloc_path();
5560 5561
	if (!path)
		return -ENOMEM;
5562

5563 5564 5565
	path->leave_spinning = 1;
	ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
				      ins, size);
5566
	BUG_ON(ret);
5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594

	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);

5595
	ret = update_block_group(trans, root, ins->objectid, ins->offset, 1);
5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616
	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);

	ret = btrfs_add_delayed_data_ref(trans, ins->objectid, ins->offset,
					 0, root_objectid, owner, offset,
					 BTRFS_ADD_DELAYED_EXTENT, NULL);
5617 5618
	return ret;
}
5619 5620 5621 5622 5623 5624

/*
 * 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
 */
5625 5626 5627 5628
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)
5629 5630 5631
{
	int ret;
	struct btrfs_block_group_cache *block_group;
5632 5633 5634
	struct btrfs_caching_control *caching_ctl;
	u64 start = ins->objectid;
	u64 num_bytes = ins->offset;
5635 5636

	block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
5637
	cache_block_group(block_group, trans, NULL, 0);
5638
	caching_ctl = get_caching_control(block_group);
5639

5640 5641 5642 5643 5644 5645 5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663 5664 5665 5666 5667 5668 5669 5670
	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);
	}

5671 5672
	ret = btrfs_update_reserved_bytes(block_group, ins->offset,
					  RESERVE_ALLOC_NO_ACCOUNT);
5673
	BUG_ON(ret);
5674
	btrfs_put_block_group(block_group);
5675 5676
	ret = alloc_reserved_file_extent(trans, root, 0, root_objectid,
					 0, owner, offset, ins, 1);
5677 5678 5679
	return ret;
}

5680 5681
struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
					    struct btrfs_root *root,
5682 5683
					    u64 bytenr, u32 blocksize,
					    int level)
5684 5685 5686 5687 5688 5689 5690
{
	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);
5691
	btrfs_set_buffer_lockdep_class(root->root_key.objectid, buf, level);
5692 5693
	btrfs_tree_lock(buf);
	clean_tree_block(trans, root, buf);
5694 5695

	btrfs_set_lock_blocking(buf);
5696
	btrfs_set_buffer_uptodate(buf);
5697

5698
	if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
5699 5700 5701 5702 5703 5704 5705 5706 5707 5708
		/*
		 * 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);
5709 5710
	} else {
		set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
5711
			 buf->start + buf->len - 1, GFP_NOFS);
5712
	}
5713
	trans->blocks_used++;
5714
	/* this returns a buffer locked for blocking */
5715 5716 5717
	return buf;
}

5718 5719 5720 5721 5722
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;
5723
	struct btrfs_block_rsv *global_rsv = &root->fs_info->global_block_rsv;
5724 5725 5726 5727 5728
	int ret;

	block_rsv = get_block_rsv(trans, root);

	if (block_rsv->size == 0) {
5729
		ret = reserve_metadata_bytes(root, block_rsv, blocksize, 0, 0);
5730 5731 5732 5733 5734 5735 5736 5737
		/*
		 * 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;
5738
			return ERR_PTR(ret);
5739
		} else if (ret) {
5740
			return ERR_PTR(ret);
5741
		}
5742 5743 5744 5745 5746 5747
		return block_rsv;
	}

	ret = block_rsv_use_bytes(block_rsv, blocksize);
	if (!ret)
		return block_rsv;
5748 5749
	if (ret) {
		WARN_ON(1);
5750
		ret = reserve_metadata_bytes(root, block_rsv, blocksize, 0, 0);
5751 5752 5753 5754 5755 5756 5757 5758
		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;
		}
	}
5759 5760 5761 5762 5763 5764 5765 5766 5767 5768

	return ERR_PTR(-ENOSPC);
}

static void unuse_block_rsv(struct btrfs_block_rsv *block_rsv, u32 blocksize)
{
	block_rsv_add_bytes(block_rsv, blocksize, 0);
	block_rsv_release_bytes(block_rsv, NULL, 0);
}

5769
/*
5770 5771 5772 5773
 * 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.
 *
5774 5775
 * returns the tree buffer or NULL.
 */
5776
struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
5777 5778 5779 5780
					struct btrfs_root *root, u32 blocksize,
					u64 parent, u64 root_objectid,
					struct btrfs_disk_key *key, int level,
					u64 hint, u64 empty_size)
5781
{
C
Chris Mason 已提交
5782
	struct btrfs_key ins;
5783
	struct btrfs_block_rsv *block_rsv;
5784
	struct extent_buffer *buf;
5785 5786 5787
	u64 flags = 0;
	int ret;

5788

5789 5790 5791 5792 5793 5794
	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);
5795
	if (ret) {
5796
		unuse_block_rsv(block_rsv, blocksize);
5797
		return ERR_PTR(ret);
5798
	}
5799

5800 5801
	buf = btrfs_init_new_buffer(trans, root, ins.objectid,
				    blocksize, level);
5802 5803 5804 5805 5806 5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817 5818 5819 5820 5821 5822 5823 5824 5825 5826 5827 5828 5829
	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;

		ret = btrfs_add_delayed_tree_ref(trans, ins.objectid,
					ins.offset, parent, root_objectid,
					level, BTRFS_ADD_DELAYED_EXTENT,
					extent_op);
		BUG_ON(ret);
	}
5830 5831
	return buf;
}
5832

5833 5834 5835 5836 5837 5838 5839 5840 5841
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
5842 5843
	int reada_slot;
	int reada_count;
5844 5845 5846 5847 5848
};

#define DROP_REFERENCE	1
#define UPDATE_BACKREF	2

Y
Yan, Zheng 已提交
5849 5850 5851 5852
static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root,
				     struct walk_control *wc,
				     struct btrfs_path *path)
5853
{
Y
Yan, Zheng 已提交
5854 5855 5856
	u64 bytenr;
	u64 generation;
	u64 refs;
5857
	u64 flags;
5858
	u32 nritems;
Y
Yan, Zheng 已提交
5859 5860 5861
	u32 blocksize;
	struct btrfs_key key;
	struct extent_buffer *eb;
5862
	int ret;
Y
Yan, Zheng 已提交
5863 5864
	int slot;
	int nread = 0;
5865

Y
Yan, Zheng 已提交
5866 5867 5868 5869 5870 5871 5872 5873
	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));
	}
5874

Y
Yan, Zheng 已提交
5875 5876 5877
	eb = path->nodes[wc->level];
	nritems = btrfs_header_nritems(eb);
	blocksize = btrfs_level_size(root, wc->level - 1);
5878

Y
Yan, Zheng 已提交
5879 5880 5881
	for (slot = path->slots[wc->level]; slot < nritems; slot++) {
		if (nread >= wc->reada_count)
			break;
5882

C
Chris Mason 已提交
5883
		cond_resched();
Y
Yan, Zheng 已提交
5884 5885
		bytenr = btrfs_node_blockptr(eb, slot);
		generation = btrfs_node_ptr_generation(eb, slot);
C
Chris Mason 已提交
5886

Y
Yan, Zheng 已提交
5887 5888
		if (slot == path->slots[wc->level])
			goto reada;
5889

Y
Yan, Zheng 已提交
5890 5891
		if (wc->stage == UPDATE_BACKREF &&
		    generation <= root->root_key.offset)
5892 5893
			continue;

5894 5895 5896 5897 5898 5899
		/* 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
5900 5901 5902
		if (wc->stage == DROP_REFERENCE) {
			if (refs == 1)
				goto reada;
5903

5904 5905 5906
			if (wc->level == 1 &&
			    (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
				continue;
Y
Yan, Zheng 已提交
5907 5908 5909 5910 5911 5912 5913 5914
			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;
5915 5916 5917 5918
		} else {
			if (wc->level == 1 &&
			    (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
				continue;
5919
		}
Y
Yan, Zheng 已提交
5920 5921 5922 5923
reada:
		ret = readahead_tree_block(root, bytenr, blocksize,
					   generation);
		if (ret)
5924
			break;
Y
Yan, Zheng 已提交
5925
		nread++;
C
Chris Mason 已提交
5926
	}
Y
Yan, Zheng 已提交
5927
	wc->reada_slot = slot;
C
Chris Mason 已提交
5928
}
5929

Y
Yan Zheng 已提交
5930
/*
5931 5932 5933 5934 5935 5936
 * 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
5937
 */
5938
static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
5939
				   struct btrfs_root *root,
5940
				   struct btrfs_path *path,
5941
				   struct walk_control *wc, int lookup_info)
Y
Yan Zheng 已提交
5942
{
5943 5944 5945
	int level = wc->level;
	struct extent_buffer *eb = path->nodes[level];
	u64 flag = BTRFS_BLOCK_FLAG_FULL_BACKREF;
Y
Yan Zheng 已提交
5946 5947
	int ret;

5948 5949 5950
	if (wc->stage == UPDATE_BACKREF &&
	    btrfs_header_owner(eb) != root->root_key.objectid)
		return 1;
Y
Yan Zheng 已提交
5951

5952 5953 5954 5955
	/*
	 * when reference count of tree block is 1, it won't increase
	 * again. once full backref flag is set, we never clear it.
	 */
5956 5957 5958
	if (lookup_info &&
	    ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
	     (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag)))) {
5959 5960 5961 5962 5963 5964 5965 5966
		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);
	}
5967

5968 5969 5970
	if (wc->stage == DROP_REFERENCE) {
		if (wc->refs[level] > 1)
			return 1;
Y
Yan Zheng 已提交
5971

5972
		if (path->locks[level] && !wc->keep_locks) {
5973
			btrfs_tree_unlock_rw(eb, path->locks[level]);
5974 5975 5976 5977
			path->locks[level] = 0;
		}
		return 0;
	}
Y
Yan Zheng 已提交
5978

5979 5980 5981 5982
	/* wc->stage == UPDATE_BACKREF */
	if (!(wc->flags[level] & flag)) {
		BUG_ON(!path->locks[level]);
		ret = btrfs_inc_ref(trans, root, eb, 1);
Y
Yan Zheng 已提交
5983
		BUG_ON(ret);
5984 5985 5986 5987 5988 5989 5990 5991 5992 5993 5994 5995 5996
		ret = btrfs_dec_ref(trans, root, eb, 0);
		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) {
5997
		btrfs_tree_unlock_rw(eb, path->locks[level]);
5998 5999 6000 6001 6002
		path->locks[level] = 0;
	}
	return 0;
}

Y
Yan, Zheng 已提交
6003 6004 6005 6006 6007 6008 6009 6010 6011 6012 6013 6014 6015 6016 6017 6018
/*
 * 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,
6019
				 struct walk_control *wc, int *lookup_info)
Y
Yan, Zheng 已提交
6020 6021 6022 6023 6024 6025 6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038
{
	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 &&
6039 6040
	    generation <= root->root_key.offset) {
		*lookup_info = 1;
Y
Yan, Zheng 已提交
6041
		return 1;
6042
	}
Y
Yan, Zheng 已提交
6043 6044 6045 6046 6047 6048 6049

	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);
6050 6051
		if (!next)
			return -ENOMEM;
Y
Yan, Zheng 已提交
6052 6053 6054 6055 6056
		reada = 1;
	}
	btrfs_tree_lock(next);
	btrfs_set_lock_blocking(next);

6057 6058 6059 6060 6061 6062
	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
6063

6064
	if (wc->stage == DROP_REFERENCE) {
Y
Yan, Zheng 已提交
6065
		if (wc->refs[level - 1] > 1) {
6066 6067 6068 6069
			if (level == 1 &&
			    (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
				goto skip;

Y
Yan, Zheng 已提交
6070 6071 6072 6073 6074 6075 6076 6077 6078 6079 6080 6081 6082
			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;
		}
6083 6084 6085 6086
	} else {
		if (level == 1 &&
		    (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
			goto skip;
Y
Yan, Zheng 已提交
6087 6088 6089 6090 6091 6092
	}

	if (!btrfs_buffer_uptodate(next, generation)) {
		btrfs_tree_unlock(next);
		free_extent_buffer(next);
		next = NULL;
6093
		*lookup_info = 1;
Y
Yan, Zheng 已提交
6094 6095 6096 6097 6098 6099
	}

	if (!next) {
		if (reada && level == 1)
			reada_walk_down(trans, root, wc, path);
		next = read_tree_block(root, bytenr, blocksize, generation);
6100 6101
		if (!next)
			return -EIO;
Y
Yan, Zheng 已提交
6102 6103 6104 6105 6106 6107 6108 6109
		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;
6110
	path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
Y
Yan, Zheng 已提交
6111 6112 6113 6114 6115 6116 6117
	wc->level = level;
	if (wc->level == 1)
		wc->reada_slot = 0;
	return 0;
skip:
	wc->refs[level - 1] = 0;
	wc->flags[level - 1] = 0;
6118 6119 6120 6121 6122 6123 6124 6125
	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
6126

6127 6128 6129
		ret = btrfs_free_extent(trans, root, bytenr, blocksize, parent,
					root->root_key.objectid, level - 1, 0);
		BUG_ON(ret);
Y
Yan, Zheng 已提交
6130 6131 6132
	}
	btrfs_tree_unlock(next);
	free_extent_buffer(next);
6133
	*lookup_info = 1;
Y
Yan, Zheng 已提交
6134 6135 6136
	return 1;
}

6137 6138 6139 6140 6141 6142 6143 6144 6145 6146 6147 6148 6149 6150 6151 6152 6153
/*
 * 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)
{
6154
	int ret;
6155 6156 6157 6158 6159 6160 6161 6162 6163 6164 6165 6166 6167 6168 6169 6170 6171 6172 6173 6174 6175 6176 6177 6178 6179 6180
	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);
6181
			path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
6182 6183 6184 6185 6186

			ret = btrfs_lookup_extent_info(trans, root,
						       eb->start, eb->len,
						       &wc->refs[level],
						       &wc->flags[level]);
Y
Yan Zheng 已提交
6187
			BUG_ON(ret);
6188 6189
			BUG_ON(wc->refs[level] == 0);
			if (wc->refs[level] == 1) {
6190
				btrfs_tree_unlock_rw(eb, path->locks[level]);
6191 6192
				return 1;
			}
Y
Yan Zheng 已提交
6193
		}
6194
	}
Y
Yan Zheng 已提交
6195

6196 6197
	/* wc->stage == DROP_REFERENCE */
	BUG_ON(wc->refs[level] > 1 && !path->locks[level]);
6198

6199 6200 6201 6202 6203 6204 6205 6206 6207 6208 6209 6210 6211
	if (wc->refs[level] == 1) {
		if (level == 0) {
			if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
				ret = btrfs_dec_ref(trans, root, eb, 1);
			else
				ret = btrfs_dec_ref(trans, root, eb, 0);
			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);
6212
			path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
6213 6214 6215 6216 6217 6218 6219 6220 6221 6222 6223 6224 6225 6226 6227 6228
		}
		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
6229 6230
	}

6231
	btrfs_free_tree_block(trans, root, eb, parent, wc->refs[level] == 1);
6232 6233 6234
out:
	wc->refs[level] = 0;
	wc->flags[level] = 0;
6235
	return 0;
6236 6237 6238 6239 6240 6241 6242 6243
}

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;
6244
	int lookup_info = 1;
6245 6246 6247
	int ret;

	while (level >= 0) {
6248
		ret = walk_down_proc(trans, root, path, wc, lookup_info);
6249 6250 6251 6252 6253 6254
		if (ret > 0)
			break;

		if (level == 0)
			break;

6255 6256 6257 6258
		if (path->slots[level] >=
		    btrfs_header_nritems(path->nodes[level]))
			break;

6259
		ret = do_walk_down(trans, root, path, wc, &lookup_info);
Y
Yan, Zheng 已提交
6260 6261 6262
		if (ret > 0) {
			path->slots[level]++;
			continue;
6263 6264
		} else if (ret < 0)
			return ret;
Y
Yan, Zheng 已提交
6265
		level = wc->level;
Y
Yan Zheng 已提交
6266 6267 6268 6269
	}
	return 0;
}

C
Chris Mason 已提交
6270
static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
6271
				 struct btrfs_root *root,
Y
Yan Zheng 已提交
6272
				 struct btrfs_path *path,
6273
				 struct walk_control *wc, int max_level)
C
Chris Mason 已提交
6274
{
6275
	int level = wc->level;
C
Chris Mason 已提交
6276
	int ret;
6277

6278 6279 6280 6281 6282 6283
	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
6284 6285
			return 0;
		} else {
6286 6287 6288
			ret = walk_up_proc(trans, root, path, wc);
			if (ret > 0)
				return 0;
6289

6290
			if (path->locks[level]) {
6291 6292
				btrfs_tree_unlock_rw(path->nodes[level],
						     path->locks[level]);
6293
				path->locks[level] = 0;
Y
Yan Zheng 已提交
6294
			}
6295 6296 6297
			free_extent_buffer(path->nodes[level]);
			path->nodes[level] = NULL;
			level++;
C
Chris Mason 已提交
6298 6299 6300 6301 6302
		}
	}
	return 1;
}

C
Chris Mason 已提交
6303
/*
6304 6305 6306 6307 6308 6309 6310 6311 6312
 * 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
6313
 */
6314 6315
void btrfs_drop_snapshot(struct btrfs_root *root,
			 struct btrfs_block_rsv *block_rsv, int update_ref)
C
Chris Mason 已提交
6316
{
6317
	struct btrfs_path *path;
6318 6319
	struct btrfs_trans_handle *trans;
	struct btrfs_root *tree_root = root->fs_info->tree_root;
6320
	struct btrfs_root_item *root_item = &root->root_item;
6321 6322 6323 6324 6325
	struct walk_control *wc;
	struct btrfs_key key;
	int err = 0;
	int ret;
	int level;
C
Chris Mason 已提交
6326

6327
	path = btrfs_alloc_path();
6328 6329 6330 6331
	if (!path) {
		err = -ENOMEM;
		goto out;
	}
C
Chris Mason 已提交
6332

6333
	wc = kzalloc(sizeof(*wc), GFP_NOFS);
6334 6335
	if (!wc) {
		btrfs_free_path(path);
6336 6337
		err = -ENOMEM;
		goto out;
6338
	}
6339

6340
	trans = btrfs_start_transaction(tree_root, 0);
6341 6342
	BUG_ON(IS_ERR(trans));

6343 6344
	if (block_rsv)
		trans->block_rsv = block_rsv;
6345

6346
	if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
6347
		level = btrfs_header_level(root->node);
6348 6349
		path->nodes[level] = btrfs_lock_root_node(root);
		btrfs_set_lock_blocking(path->nodes[level]);
6350
		path->slots[level] = 0;
6351
		path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
6352 6353
		memset(&wc->update_progress, 0,
		       sizeof(wc->update_progress));
6354 6355
	} else {
		btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
6356 6357 6358
		memcpy(&wc->update_progress, &key,
		       sizeof(wc->update_progress));

6359
		level = root_item->drop_level;
6360
		BUG_ON(level == 0);
6361
		path->lowest_level = level;
6362 6363 6364 6365
		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
		path->lowest_level = 0;
		if (ret < 0) {
			err = ret;
6366
			goto out_free;
6367
		}
Y
Yan, Zheng 已提交
6368
		WARN_ON(ret > 0);
6369

6370 6371 6372 6373
		/*
		 * unlock our path, this is safe because only this
		 * function is allowed to delete this snapshot
		 */
6374
		btrfs_unlock_up_safe(path, 0);
6375 6376 6377 6378 6379 6380 6381 6382 6383 6384 6385 6386 6387 6388 6389 6390 6391 6392 6393 6394 6395

		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--;
		}
6396
	}
6397 6398 6399 6400 6401 6402

	wc->level = level;
	wc->shared_level = -1;
	wc->stage = DROP_REFERENCE;
	wc->update_ref = update_ref;
	wc->keep_locks = 0;
Y
Yan, Zheng 已提交
6403
	wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(root);
6404

C
Chris Mason 已提交
6405
	while (1) {
6406 6407 6408
		ret = walk_down_tree(trans, root, path, wc);
		if (ret < 0) {
			err = ret;
C
Chris Mason 已提交
6409
			break;
6410
		}
C
Chris Mason 已提交
6411

6412 6413 6414
		ret = walk_up_tree(trans, root, path, wc, BTRFS_MAX_LEVEL);
		if (ret < 0) {
			err = ret;
C
Chris Mason 已提交
6415
			break;
6416 6417 6418 6419
		}

		if (ret > 0) {
			BUG_ON(wc->stage != DROP_REFERENCE);
6420 6421
			break;
		}
6422 6423 6424 6425 6426 6427 6428 6429 6430 6431

		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);
6432
		if (btrfs_should_end_transaction(trans, tree_root)) {
6433 6434 6435 6436 6437
			ret = btrfs_update_root(trans, tree_root,
						&root->root_key,
						root_item);
			BUG_ON(ret);

6438
			btrfs_end_transaction_throttle(trans, tree_root);
6439
			trans = btrfs_start_transaction(tree_root, 0);
6440
			BUG_ON(IS_ERR(trans));
6441 6442
			if (block_rsv)
				trans->block_rsv = block_rsv;
6443
		}
C
Chris Mason 已提交
6444
	}
6445
	btrfs_release_path(path);
6446 6447 6448 6449 6450
	BUG_ON(err);

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

6451 6452 6453 6454 6455
	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) {
6456 6457 6458 6459 6460 6461 6462
			/* 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);
6463 6464 6465 6466 6467 6468 6469 6470 6471 6472
		}
	}

	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);
	}
6473
out_free:
6474
	btrfs_end_transaction_throttle(trans, tree_root);
6475
	kfree(wc);
6476
	btrfs_free_path(path);
6477 6478 6479 6480
out:
	if (err)
		btrfs_std_error(root->fs_info, err);
	return;
C
Chris Mason 已提交
6481
}
C
Chris Mason 已提交
6482

6483 6484 6485 6486 6487
/*
 * drop subtree rooted at tree block 'node'.
 *
 * NOTE: this function will unlock and release tree block 'node'
 */
Y
Yan Zheng 已提交
6488 6489 6490 6491 6492 6493
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;
6494
	struct walk_control *wc;
Y
Yan Zheng 已提交
6495 6496 6497 6498 6499
	int level;
	int parent_level;
	int ret = 0;
	int wret;

6500 6501
	BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);

Y
Yan Zheng 已提交
6502
	path = btrfs_alloc_path();
T
Tsutomu Itoh 已提交
6503 6504
	if (!path)
		return -ENOMEM;
Y
Yan Zheng 已提交
6505

6506
	wc = kzalloc(sizeof(*wc), GFP_NOFS);
T
Tsutomu Itoh 已提交
6507 6508 6509 6510
	if (!wc) {
		btrfs_free_path(path);
		return -ENOMEM;
	}
6511

6512
	btrfs_assert_tree_locked(parent);
Y
Yan Zheng 已提交
6513 6514 6515 6516 6517
	parent_level = btrfs_header_level(parent);
	extent_buffer_get(parent);
	path->nodes[parent_level] = parent;
	path->slots[parent_level] = btrfs_header_nritems(parent);

6518
	btrfs_assert_tree_locked(node);
Y
Yan Zheng 已提交
6519 6520 6521
	level = btrfs_header_level(node);
	path->nodes[level] = node;
	path->slots[level] = 0;
6522
	path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
6523 6524 6525 6526 6527 6528 6529 6530

	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;
Y
Yan, Zheng 已提交
6531
	wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(root);
Y
Yan Zheng 已提交
6532 6533

	while (1) {
6534 6535
		wret = walk_down_tree(trans, root, path, wc);
		if (wret < 0) {
Y
Yan Zheng 已提交
6536 6537
			ret = wret;
			break;
6538
		}
Y
Yan Zheng 已提交
6539

6540
		wret = walk_up_tree(trans, root, path, wc, parent_level);
Y
Yan Zheng 已提交
6541 6542 6543 6544 6545 6546
		if (wret < 0)
			ret = wret;
		if (wret != 0)
			break;
	}

6547
	kfree(wc);
Y
Yan Zheng 已提交
6548 6549 6550 6551
	btrfs_free_path(path);
	return ret;
}

6552 6553 6554 6555 6556 6557
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;

6558 6559 6560 6561 6562 6563 6564 6565
	/*
	 * 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;

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 6591 6592 6593 6594 6595 6596
	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;
}

6597
static int set_block_group_ro(struct btrfs_block_group_cache *cache, int force)
C
Chris Mason 已提交
6598
{
6599 6600
	struct btrfs_space_info *sinfo = cache->space_info;
	u64 num_bytes;
6601
	u64 min_allocable_bytes;
6602
	int ret = -ENOSPC;
C
Chris Mason 已提交
6603

6604

6605 6606 6607 6608 6609 6610 6611 6612 6613 6614 6615 6616
	/*
	 * 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;

6617 6618
	spin_lock(&sinfo->lock);
	spin_lock(&cache->lock);
6619 6620 6621 6622 6623 6624

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

6625 6626 6627 6628
	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 +
6629 6630
	    sinfo->bytes_may_use + sinfo->bytes_readonly + num_bytes +
	    min_allocable_bytes <= sinfo->total_bytes) {
6631 6632 6633 6634
		sinfo->bytes_readonly += num_bytes;
		cache->ro = 1;
		ret = 0;
	}
6635
out:
6636 6637 6638 6639
	spin_unlock(&cache->lock);
	spin_unlock(&sinfo->lock);
	return ret;
}
6640

6641 6642
int btrfs_set_block_group_ro(struct btrfs_root *root,
			     struct btrfs_block_group_cache *cache)
6643

6644 6645 6646 6647
{
	struct btrfs_trans_handle *trans;
	u64 alloc_flags;
	int ret;
6648

6649
	BUG_ON(cache->ro);
C
Chris Mason 已提交
6650

C
Chris Mason 已提交
6651
	trans = btrfs_join_transaction(root);
6652
	BUG_ON(IS_ERR(trans));
6653

6654 6655
	alloc_flags = update_block_group_flags(root, cache->flags);
	if (alloc_flags != cache->flags)
6656 6657
		do_chunk_alloc(trans, root, 2 * 1024 * 1024, alloc_flags,
			       CHUNK_ALLOC_FORCE);
6658

6659
	ret = set_block_group_ro(cache, 0);
6660 6661 6662
	if (!ret)
		goto out;
	alloc_flags = get_alloc_profile(root, cache->space_info->flags);
6663 6664
	ret = do_chunk_alloc(trans, root, 2 * 1024 * 1024, alloc_flags,
			     CHUNK_ALLOC_FORCE);
6665 6666
	if (ret < 0)
		goto out;
6667
	ret = set_block_group_ro(cache, 0);
6668 6669 6670 6671
out:
	btrfs_end_transaction(trans, root);
	return ret;
}
6672

6673 6674 6675 6676
int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans,
			    struct btrfs_root *root, u64 type)
{
	u64 alloc_flags = get_alloc_profile(root, type);
6677 6678
	return do_chunk_alloc(trans, root, 2 * 1024 * 1024, alloc_flags,
			      CHUNK_ALLOC_FORCE);
6679 6680
}

6681 6682 6683 6684 6685 6686 6687 6688 6689 6690 6691 6692 6693 6694 6695 6696 6697 6698 6699 6700 6701 6702 6703 6704 6705 6706 6707 6708 6709 6710 6711 6712 6713 6714 6715 6716 6717 6718 6719 6720 6721 6722 6723 6724 6725 6726 6727 6728 6729 6730 6731 6732 6733 6734 6735 6736
/*
 * 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;
}

6737 6738
int btrfs_set_block_group_rw(struct btrfs_root *root,
			      struct btrfs_block_group_cache *cache)
6739
{
6740 6741 6742 6743 6744 6745 6746 6747 6748 6749 6750 6751 6752
	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);
6753 6754 6755
	return 0;
}

6756 6757 6758 6759 6760 6761 6762
/*
 * 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
6763
{
6764 6765 6766 6767
	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;
6768
	u64 min_free;
J
Josef Bacik 已提交
6769 6770
	u64 dev_min = 1;
	u64 dev_nr = 0;
6771
	int index;
6772 6773
	int full = 0;
	int ret = 0;
Z
Zheng Yan 已提交
6774

6775
	block_group = btrfs_lookup_block_group(root->fs_info, bytenr);
Z
Zheng Yan 已提交
6776

6777 6778 6779
	/* odd, couldn't find the block group, leave it alone */
	if (!block_group)
		return -1;
Z
Zheng Yan 已提交
6780

6781 6782
	min_free = btrfs_block_group_used(&block_group->item);

6783
	/* no bytes used, we're good */
6784
	if (!min_free)
Z
Zheng Yan 已提交
6785 6786
		goto out;

6787 6788
	space_info = block_group->space_info;
	spin_lock(&space_info->lock);
6789

6790
	full = space_info->full;
6791

6792 6793
	/*
	 * if this is the last block group we have in this space, we can't
6794 6795 6796 6797
	 * 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
6798
	 */
6799
	if ((space_info->total_bytes != block_group->key.offset) &&
6800 6801 6802
	    (space_info->bytes_used + space_info->bytes_reserved +
	     space_info->bytes_pinned + space_info->bytes_readonly +
	     min_free < space_info->total_bytes)) {
6803 6804
		spin_unlock(&space_info->lock);
		goto out;
6805
	}
6806
	spin_unlock(&space_info->lock);
6807

6808 6809 6810 6811 6812 6813 6814 6815 6816 6817
	/*
	 * 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;
6818

6819 6820 6821 6822 6823 6824 6825 6826 6827 6828 6829
	/*
	 * 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
6830 6831
		/* Divide by 2 */
		min_free >>= 1;
6832 6833 6834
	} else if (index == 1) {
		dev_min = 2;
	} else if (index == 2) {
J
Josef Bacik 已提交
6835 6836
		/* Multiply by 2 */
		min_free <<= 1;
6837 6838
	} else if (index == 3) {
		dev_min = fs_devices->rw_devices;
J
Josef Bacik 已提交
6839
		do_div(min_free, dev_min);
6840 6841
	}

6842 6843
	mutex_lock(&root->fs_info->chunk_mutex);
	list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) {
6844
		u64 dev_offset;
6845

6846 6847 6848 6849 6850 6851
		/*
		 * 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) {
			ret = find_free_dev_extent(NULL, device, min_free,
6852
						   &dev_offset, NULL);
6853
			if (!ret)
6854 6855 6856
				dev_nr++;

			if (dev_nr >= dev_min)
6857
				break;
6858

6859
			ret = -1;
6860
		}
6861
	}
6862
	mutex_unlock(&root->fs_info->chunk_mutex);
6863
out:
6864
	btrfs_put_block_group(block_group);
6865 6866 6867
	return ret;
}

6868 6869
static int find_first_block_group(struct btrfs_root *root,
		struct btrfs_path *path, struct btrfs_key *key)
6870
{
6871
	int ret = 0;
6872 6873 6874
	struct btrfs_key found_key;
	struct extent_buffer *leaf;
	int slot;
6875

6876 6877
	ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
	if (ret < 0)
6878 6879
		goto out;

C
Chris Mason 已提交
6880
	while (1) {
6881
		slot = path->slots[0];
6882
		leaf = path->nodes[0];
6883 6884 6885 6886 6887
		if (slot >= btrfs_header_nritems(leaf)) {
			ret = btrfs_next_leaf(root, path);
			if (ret == 0)
				continue;
			if (ret < 0)
6888
				goto out;
6889
			break;
6890
		}
6891
		btrfs_item_key_to_cpu(leaf, &found_key, slot);
6892

6893
		if (found_key.objectid >= key->objectid &&
6894 6895 6896 6897
		    found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
			ret = 0;
			goto out;
		}
6898
		path->slots[0]++;
6899
	}
6900
out:
6901
	return ret;
6902 6903
}

6904 6905 6906 6907 6908 6909 6910 6911 6912 6913 6914 6915 6916 6917 6918 6919 6920 6921 6922 6923 6924 6925 6926 6927 6928 6929 6930 6931 6932 6933 6934 6935 6936 6937
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
6938 6939 6940
int btrfs_free_block_groups(struct btrfs_fs_info *info)
{
	struct btrfs_block_group_cache *block_group;
6941
	struct btrfs_space_info *space_info;
6942
	struct btrfs_caching_control *caching_ctl;
Z
Zheng Yan 已提交
6943 6944
	struct rb_node *n;

6945 6946 6947 6948 6949 6950 6951 6952 6953
	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
6954 6955 6956 6957 6958 6959
	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
6960 6961
		spin_unlock(&info->block_group_cache_lock);

6962
		down_write(&block_group->space_info->groups_sem);
Z
Zheng Yan 已提交
6963
		list_del(&block_group->list);
6964
		up_write(&block_group->space_info->groups_sem);
6965

J
Josef Bacik 已提交
6966
		if (block_group->cached == BTRFS_CACHE_STARTED)
6967
			wait_block_group_cache_done(block_group);
J
Josef Bacik 已提交
6968

6969 6970 6971 6972 6973 6974 6975
		/*
		 * 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
6976
		btrfs_remove_free_space_cache(block_group);
6977
		btrfs_put_block_group(block_group);
Y
Yan Zheng 已提交
6978 6979

		spin_lock(&info->block_group_cache_lock);
Z
Zheng Yan 已提交
6980 6981
	}
	spin_unlock(&info->block_group_cache_lock);
6982 6983 6984 6985 6986 6987 6988 6989 6990

	/* 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();

6991 6992
	release_global_block_rsv(info);

6993 6994 6995 6996
	while(!list_empty(&info->space_info)) {
		space_info = list_entry(info->space_info.next,
					struct btrfs_space_info,
					list);
6997
		if (space_info->bytes_pinned > 0 ||
6998 6999
		    space_info->bytes_reserved > 0 ||
		    space_info->bytes_may_use > 0) {
7000 7001 7002
			WARN_ON(1);
			dump_space_info(space_info, 0, 0);
		}
7003 7004 7005
		list_del(&space_info->list);
		kfree(space_info);
	}
Z
Zheng Yan 已提交
7006 7007 7008
	return 0;
}

7009 7010 7011 7012 7013 7014 7015 7016 7017 7018
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
7019 7020 7021 7022 7023
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
7024
	struct btrfs_fs_info *info = root->fs_info;
7025
	struct btrfs_space_info *space_info;
C
Chris Mason 已提交
7026 7027
	struct btrfs_key key;
	struct btrfs_key found_key;
7028
	struct extent_buffer *leaf;
7029 7030
	int need_clear = 0;
	u64 cache_gen;
7031

C
Chris Mason 已提交
7032
	root = info->extent_root;
C
Chris Mason 已提交
7033
	key.objectid = 0;
7034
	key.offset = 0;
C
Chris Mason 已提交
7035 7036 7037 7038
	btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
J
Josef Bacik 已提交
7039
	path->reada = 1;
C
Chris Mason 已提交
7040

7041
	cache_gen = btrfs_super_cache_generation(&root->fs_info->super_copy);
7042
	if (btrfs_test_opt(root, SPACE_CACHE) &&
7043 7044
	    btrfs_super_generation(&root->fs_info->super_copy) != cache_gen)
		need_clear = 1;
7045 7046
	if (btrfs_test_opt(root, CLEAR_CACHE))
		need_clear = 1;
7047

C
Chris Mason 已提交
7048
	while (1) {
7049
		ret = find_first_block_group(root, path, &key);
7050 7051
		if (ret > 0)
			break;
7052 7053
		if (ret != 0)
			goto error;
7054 7055
		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
7056
		cache = kzalloc(sizeof(*cache), GFP_NOFS);
C
Chris Mason 已提交
7057
		if (!cache) {
7058
			ret = -ENOMEM;
7059
			goto error;
C
Chris Mason 已提交
7060
		}
7061 7062 7063 7064 7065 7066 7067
		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
7068

7069
		atomic_set(&cache->count, 1);
7070
		spin_lock_init(&cache->lock);
J
Josef Bacik 已提交
7071
		cache->fs_info = info;
J
Josef Bacik 已提交
7072
		INIT_LIST_HEAD(&cache->list);
7073
		INIT_LIST_HEAD(&cache->cluster_list);
7074

7075 7076 7077
		if (need_clear)
			cache->disk_cache_state = BTRFS_DC_CLEAR;

7078 7079 7080
		read_extent_buffer(leaf, &cache->item,
				   btrfs_item_ptr_offset(leaf, path->slots[0]),
				   sizeof(cache->item));
C
Chris Mason 已提交
7081
		memcpy(&cache->key, &found_key, sizeof(found_key));
7082

C
Chris Mason 已提交
7083
		key.objectid = found_key.objectid + found_key.offset;
7084
		btrfs_release_path(path);
7085
		cache->flags = btrfs_block_group_flags(&cache->item);
J
Josef Bacik 已提交
7086 7087
		cache->sectorsize = root->sectorsize;

7088 7089
		btrfs_init_free_space_ctl(cache);

7090 7091 7092 7093 7094 7095 7096
		/*
		 * 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
7097 7098 7099 7100 7101 7102 7103 7104
		/*
		 * 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)) {
7105
			cache->last_byte_to_unpin = (u64)-1;
J
Josef Bacik 已提交
7106
			cache->cached = BTRFS_CACHE_FINISHED;
7107
			free_excluded_extents(root, cache);
J
Josef Bacik 已提交
7108
		} else if (btrfs_block_group_used(&cache->item) == 0) {
7109
			cache->last_byte_to_unpin = (u64)-1;
J
Josef Bacik 已提交
7110 7111 7112 7113 7114
			cache->cached = BTRFS_CACHE_FINISHED;
			add_new_free_space(cache, root->fs_info,
					   found_key.objectid,
					   found_key.objectid +
					   found_key.offset);
7115
			free_excluded_extents(root, cache);
J
Josef Bacik 已提交
7116
		}
7117

7118 7119 7120 7121 7122
		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;
7123
		spin_lock(&cache->space_info->lock);
7124
		cache->space_info->bytes_readonly += cache->bytes_super;
7125 7126
		spin_unlock(&cache->space_info->lock);

7127
		__link_block_group(space_info, cache);
J
Josef Bacik 已提交
7128 7129 7130

		ret = btrfs_add_block_group_cache(root->fs_info, cache);
		BUG_ON(ret);
7131 7132

		set_avail_alloc_bits(root->fs_info, cache->flags);
Y
Yan Zheng 已提交
7133
		if (btrfs_chunk_readonly(root, cache->key.objectid))
7134
			set_block_group_ro(cache, 1);
C
Chris Mason 已提交
7135
	}
7136 7137 7138 7139 7140 7141 7142 7143 7144 7145 7146 7147

	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)
7148
			set_block_group_ro(cache, 1);
7149
		list_for_each_entry(cache, &space_info->block_groups[4], list)
7150
			set_block_group_ro(cache, 1);
C
Chris Mason 已提交
7151
	}
7152 7153

	init_global_block_rsv(info);
7154 7155
	ret = 0;
error:
C
Chris Mason 已提交
7156
	btrfs_free_path(path);
7157
	return ret;
C
Chris Mason 已提交
7158
}
7159 7160 7161

int btrfs_make_block_group(struct btrfs_trans_handle *trans,
			   struct btrfs_root *root, u64 bytes_used,
7162
			   u64 type, u64 chunk_objectid, u64 chunk_offset,
7163 7164 7165 7166 7167 7168 7169 7170
			   u64 size)
{
	int ret;
	struct btrfs_root *extent_root;
	struct btrfs_block_group_cache *cache;

	extent_root = root->fs_info->extent_root;

7171
	root->fs_info->last_trans_log_full_commit = trans->transid;
7172

7173
	cache = kzalloc(sizeof(*cache), GFP_NOFS);
J
Josef Bacik 已提交
7174 7175
	if (!cache)
		return -ENOMEM;
7176 7177 7178 7179 7180 7181
	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
7182

7183
	cache->key.objectid = chunk_offset;
7184
	cache->key.offset = size;
7185
	cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
7186
	cache->sectorsize = root->sectorsize;
7187
	cache->fs_info = root->fs_info;
7188

7189
	atomic_set(&cache->count, 1);
7190
	spin_lock_init(&cache->lock);
J
Josef Bacik 已提交
7191
	INIT_LIST_HEAD(&cache->list);
7192
	INIT_LIST_HEAD(&cache->cluster_list);
C
Chris Mason 已提交
7193

7194 7195
	btrfs_init_free_space_ctl(cache);

7196 7197 7198 7199 7200
	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);

7201
	cache->last_byte_to_unpin = (u64)-1;
J
Josef Bacik 已提交
7202
	cache->cached = BTRFS_CACHE_FINISHED;
7203
	exclude_super_stripes(root, cache);
7204

J
Josef Bacik 已提交
7205 7206 7207
	add_new_free_space(cache, root->fs_info, chunk_offset,
			   chunk_offset + size);

7208 7209
	free_excluded_extents(root, cache);

7210 7211 7212
	ret = update_space_info(root->fs_info, cache->flags, size, bytes_used,
				&cache->space_info);
	BUG_ON(ret);
7213 7214

	spin_lock(&cache->space_info->lock);
7215
	cache->space_info->bytes_readonly += cache->bytes_super;
7216 7217
	spin_unlock(&cache->space_info->lock);

7218
	__link_block_group(cache->space_info, cache);
7219

J
Josef Bacik 已提交
7220 7221
	ret = btrfs_add_block_group_cache(root->fs_info, cache);
	BUG_ON(ret);
7222

7223 7224 7225 7226
	ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item,
				sizeof(cache->item));
	BUG_ON(ret);

C
Chris Mason 已提交
7227
	set_avail_alloc_bits(extent_root->fs_info, type);
7228

7229 7230
	return 0;
}
Z
Zheng Yan 已提交
7231 7232 7233 7234 7235 7236

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;
7237
	struct btrfs_free_cluster *cluster;
7238
	struct btrfs_root *tree_root = root->fs_info->tree_root;
Z
Zheng Yan 已提交
7239
	struct btrfs_key key;
7240
	struct inode *inode;
Z
Zheng Yan 已提交
7241
	int ret;
J
Josef Bacik 已提交
7242
	int factor;
Z
Zheng Yan 已提交
7243 7244 7245 7246 7247

	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
7248
	BUG_ON(!block_group->ro);
Z
Zheng Yan 已提交
7249

7250 7251 7252 7253 7254 7255
	/*
	 * Free the reserved super bytes from this block group before
	 * remove it.
	 */
	free_excluded_extents(root, block_group);

Z
Zheng Yan 已提交
7256
	memcpy(&key, &block_group->key, sizeof(key));
J
Josef Bacik 已提交
7257 7258 7259 7260 7261 7262
	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
7263

7264 7265 7266 7267 7268 7269 7270 7271 7272 7273 7274 7275 7276 7277 7278
	/* 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
7279
	path = btrfs_alloc_path();
7280 7281 7282 7283
	if (!path) {
		ret = -ENOMEM;
		goto out;
	}
Z
Zheng Yan 已提交
7284

7285 7286
	inode = lookup_free_space_inode(root, block_group, path);
	if (!IS_ERR(inode)) {
7287 7288
		ret = btrfs_orphan_add(trans, inode);
		BUG_ON(ret);
7289 7290 7291 7292 7293 7294 7295 7296 7297 7298 7299 7300
		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 */
7301
		btrfs_add_delayed_iput(inode);
7302 7303 7304 7305 7306 7307 7308 7309 7310 7311
	}

	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)
7312
		btrfs_release_path(path);
7313 7314 7315 7316
	if (ret == 0) {
		ret = btrfs_del_item(trans, tree_root, path);
		if (ret)
			goto out;
7317
		btrfs_release_path(path);
7318 7319
	}

7320
	spin_lock(&root->fs_info->block_group_cache_lock);
Z
Zheng Yan 已提交
7321 7322
	rb_erase(&block_group->cache_node,
		 &root->fs_info->block_group_cache_tree);
7323
	spin_unlock(&root->fs_info->block_group_cache_lock);
J
Josef Bacik 已提交
7324

7325
	down_write(&block_group->space_info->groups_sem);
7326 7327 7328 7329 7330
	/*
	 * 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);
7331
	up_write(&block_group->space_info->groups_sem);
Z
Zheng Yan 已提交
7332

J
Josef Bacik 已提交
7333
	if (block_group->cached == BTRFS_CACHE_STARTED)
7334
		wait_block_group_cache_done(block_group);
J
Josef Bacik 已提交
7335 7336 7337

	btrfs_remove_free_space_cache(block_group);

Y
Yan Zheng 已提交
7338 7339 7340
	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
7341
	block_group->space_info->disk_total -= block_group->key.offset * factor;
Y
Yan Zheng 已提交
7342
	spin_unlock(&block_group->space_info->lock);
7343

7344 7345
	memcpy(&key, &block_group->key, sizeof(key));

7346
	btrfs_clear_space_info_full(root->fs_info);
Y
Yan Zheng 已提交
7347

7348 7349
	btrfs_put_block_group(block_group);
	btrfs_put_block_group(block_group);
Z
Zheng Yan 已提交
7350 7351 7352 7353 7354 7355 7356 7357 7358 7359 7360 7361

	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
7362

7363 7364 7365
int btrfs_init_space_info(struct btrfs_fs_info *fs_info)
{
	struct btrfs_space_info *space_info;
7366 7367 7368 7369
	struct btrfs_super_block *disk_super;
	u64 features;
	u64 flags;
	int mixed = 0;
7370 7371
	int ret;

7372 7373 7374
	disk_super = &fs_info->super_copy;
	if (!btrfs_super_root(disk_super))
		return 1;
7375

7376 7377 7378
	features = btrfs_super_incompat_flags(disk_super);
	if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS)
		mixed = 1;
7379

7380 7381
	flags = BTRFS_BLOCK_GROUP_SYSTEM;
	ret = update_space_info(fs_info, flags, 0, 0, &space_info);
7382
	if (ret)
7383
		goto out;
7384

7385 7386 7387 7388 7389 7390 7391 7392 7393 7394 7395 7396 7397
	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:
7398 7399 7400
	return ret;
}

L
liubo 已提交
7401 7402 7403 7404 7405 7406
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,
7407
			       u64 num_bytes, u64 *actual_bytes)
L
liubo 已提交
7408
{
7409
	return btrfs_discard_extent(root, bytenr, num_bytes, actual_bytes);
L
liubo 已提交
7410
}
7411 7412 7413 7414 7415 7416 7417 7418 7419 7420 7421 7422 7423 7424 7425 7426 7427 7428 7429 7430 7431 7432 7433 7434 7435 7436 7437 7438 7439 7440 7441 7442 7443 7444 7445 7446 7447 7448 7449 7450 7451 7452 7453 7454 7455 7456 7457 7458

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;
}