delayed-ref.c 25.4 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
2 3 4 5 6
/*
 * Copyright (C) 2009 Oracle.  All rights reserved.
 */

#include <linux/sched.h>
7
#include <linux/slab.h>
8 9 10 11
#include <linux/sort.h>
#include "ctree.h"
#include "delayed-ref.h"
#include "transaction.h"
12
#include "qgroup.h"
13

14 15 16 17
struct kmem_cache *btrfs_delayed_ref_head_cachep;
struct kmem_cache *btrfs_delayed_tree_ref_cachep;
struct kmem_cache *btrfs_delayed_data_ref_cachep;
struct kmem_cache *btrfs_delayed_extent_op_cachep;
18 19 20 21 22 23 24 25 26 27
/*
 * delayed back reference update tracking.  For subvolume trees
 * we queue up extent allocations and backref maintenance for
 * delayed processing.   This avoids deep call chains where we
 * add extents in the middle of btrfs_search_slot, and it allows
 * us to buffer up frequently modified backrefs in an rb tree instead
 * of hammering updates on the extent allocation tree.
 */

/*
28 29
 * compare two delayed tree backrefs with same bytenr and type
 */
J
Josef Bacik 已提交
30 31
static int comp_tree_refs(struct btrfs_delayed_tree_ref *ref1,
			  struct btrfs_delayed_tree_ref *ref2)
32
{
33
	if (ref1->node.type == BTRFS_TREE_BLOCK_REF_KEY) {
34 35 36 37 38 39 40 41 42 43
		if (ref1->root < ref2->root)
			return -1;
		if (ref1->root > ref2->root)
			return 1;
	} else {
		if (ref1->parent < ref2->parent)
			return -1;
		if (ref1->parent > ref2->parent)
			return 1;
	}
44 45 46 47 48
	return 0;
}

/*
 * compare two delayed data backrefs with same bytenr and type
49
 */
J
Josef Bacik 已提交
50 51
static int comp_data_refs(struct btrfs_delayed_data_ref *ref1,
			  struct btrfs_delayed_data_ref *ref2)
52
{
53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74
	if (ref1->node.type == BTRFS_EXTENT_DATA_REF_KEY) {
		if (ref1->root < ref2->root)
			return -1;
		if (ref1->root > ref2->root)
			return 1;
		if (ref1->objectid < ref2->objectid)
			return -1;
		if (ref1->objectid > ref2->objectid)
			return 1;
		if (ref1->offset < ref2->offset)
			return -1;
		if (ref1->offset > ref2->offset)
			return 1;
	} else {
		if (ref1->parent < ref2->parent)
			return -1;
		if (ref1->parent > ref2->parent)
			return 1;
	}
	return 0;
}

J
Josef Bacik 已提交
75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102
static int comp_refs(struct btrfs_delayed_ref_node *ref1,
		     struct btrfs_delayed_ref_node *ref2,
		     bool check_seq)
{
	int ret = 0;

	if (ref1->type < ref2->type)
		return -1;
	if (ref1->type > ref2->type)
		return 1;
	if (ref1->type == BTRFS_TREE_BLOCK_REF_KEY ||
	    ref1->type == BTRFS_SHARED_BLOCK_REF_KEY)
		ret = comp_tree_refs(btrfs_delayed_node_to_tree_ref(ref1),
				     btrfs_delayed_node_to_tree_ref(ref2));
	else
		ret = comp_data_refs(btrfs_delayed_node_to_data_ref(ref1),
				     btrfs_delayed_node_to_data_ref(ref2));
	if (ret)
		return ret;
	if (check_seq) {
		if (ref1->seq < ref2->seq)
			return -1;
		if (ref1->seq > ref2->seq)
			return 1;
	}
	return 0;
}

L
Liu Bo 已提交
103 104 105 106 107 108 109 110 111 112 113
/* insert a new ref to head ref rbtree */
static struct btrfs_delayed_ref_head *htree_insert(struct rb_root *root,
						   struct rb_node *node)
{
	struct rb_node **p = &root->rb_node;
	struct rb_node *parent_node = NULL;
	struct btrfs_delayed_ref_head *entry;
	struct btrfs_delayed_ref_head *ins;
	u64 bytenr;

	ins = rb_entry(node, struct btrfs_delayed_ref_head, href_node);
114
	bytenr = ins->bytenr;
L
Liu Bo 已提交
115 116 117 118 119
	while (*p) {
		parent_node = *p;
		entry = rb_entry(parent_node, struct btrfs_delayed_ref_head,
				 href_node);

120
		if (bytenr < entry->bytenr)
L
Liu Bo 已提交
121
			p = &(*p)->rb_left;
122
		else if (bytenr > entry->bytenr)
L
Liu Bo 已提交
123 124 125 126 127 128 129 130 131 132
			p = &(*p)->rb_right;
		else
			return entry;
	}

	rb_link_node(node, parent_node, p);
	rb_insert_color(node, root);
	return NULL;
}

133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160
static struct btrfs_delayed_ref_node* tree_insert(struct rb_root *root,
		struct btrfs_delayed_ref_node *ins)
{
	struct rb_node **p = &root->rb_node;
	struct rb_node *node = &ins->ref_node;
	struct rb_node *parent_node = NULL;
	struct btrfs_delayed_ref_node *entry;

	while (*p) {
		int comp;

		parent_node = *p;
		entry = rb_entry(parent_node, struct btrfs_delayed_ref_node,
				 ref_node);
		comp = comp_refs(ins, entry, true);
		if (comp < 0)
			p = &(*p)->rb_left;
		else if (comp > 0)
			p = &(*p)->rb_right;
		else
			return entry;
	}

	rb_link_node(node, parent_node, p);
	rb_insert_color(node, root);
	return NULL;
}

161
/*
162
 * find an head entry based on bytenr. This returns the delayed ref
163 164 165
 * head if it was able to find one, or NULL if nothing was in that spot.
 * If return_bigger is given, the next bigger entry is returned if no exact
 * match is found.
166
 */
L
Liu Bo 已提交
167 168
static struct btrfs_delayed_ref_head *
find_ref_head(struct rb_root *root, u64 bytenr,
169
	      int return_bigger)
170
{
171
	struct rb_node *n;
L
Liu Bo 已提交
172
	struct btrfs_delayed_ref_head *entry;
173

174 175
	n = root->rb_node;
	entry = NULL;
176
	while (n) {
L
Liu Bo 已提交
177
		entry = rb_entry(n, struct btrfs_delayed_ref_head, href_node);
178

179
		if (bytenr < entry->bytenr)
180
			n = n->rb_left;
181
		else if (bytenr > entry->bytenr)
182 183 184 185
			n = n->rb_right;
		else
			return entry;
	}
186
	if (entry && return_bigger) {
187
		if (bytenr > entry->bytenr) {
L
Liu Bo 已提交
188
			n = rb_next(&entry->href_node);
189 190
			if (!n)
				n = rb_first(root);
L
Liu Bo 已提交
191 192
			entry = rb_entry(n, struct btrfs_delayed_ref_head,
					 href_node);
193
			return entry;
194 195 196
		}
		return entry;
	}
197 198 199
	return NULL;
}

200 201
int btrfs_delayed_ref_lock(struct btrfs_trans_handle *trans,
			   struct btrfs_delayed_ref_head *head)
202
{
203 204 205
	struct btrfs_delayed_ref_root *delayed_refs;

	delayed_refs = &trans->transaction->delayed_refs;
206
	lockdep_assert_held(&delayed_refs->lock);
207 208 209
	if (mutex_trylock(&head->mutex))
		return 0;

210
	refcount_inc(&head->refs);
211 212 213 214
	spin_unlock(&delayed_refs->lock);

	mutex_lock(&head->mutex);
	spin_lock(&delayed_refs->lock);
215
	if (RB_EMPTY_NODE(&head->href_node)) {
216
		mutex_unlock(&head->mutex);
217
		btrfs_put_delayed_ref_head(head);
218 219
		return -EAGAIN;
	}
220
	btrfs_put_delayed_ref_head(head);
221 222 223
	return 0;
}

224
static inline void drop_delayed_ref(struct btrfs_trans_handle *trans,
225
				    struct btrfs_delayed_ref_root *delayed_refs,
226
				    struct btrfs_delayed_ref_head *head,
227 228
				    struct btrfs_delayed_ref_node *ref)
{
229
	lockdep_assert_held(&head->lock);
230 231
	rb_erase(&ref->ref_node, &head->ref_tree);
	RB_CLEAR_NODE(&ref->ref_node);
232 233
	if (!list_empty(&ref->add_list))
		list_del(&ref->add_list);
234 235
	ref->in_tree = 0;
	btrfs_put_delayed_ref(ref);
236
	atomic_dec(&delayed_refs->num_entries);
237 238 239 240
	if (trans->delayed_ref_updates)
		trans->delayed_ref_updates--;
}

241 242 243 244 245 246 247
static bool merge_ref(struct btrfs_trans_handle *trans,
		      struct btrfs_delayed_ref_root *delayed_refs,
		      struct btrfs_delayed_ref_head *head,
		      struct btrfs_delayed_ref_node *ref,
		      u64 seq)
{
	struct btrfs_delayed_ref_node *next;
248
	struct rb_node *node = rb_next(&ref->ref_node);
249 250
	bool done = false;

251
	while (!done && node) {
252 253
		int mod;

254 255
		next = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
		node = rb_next(node);
256
		if (seq && next->seq >= seq)
257
			break;
J
Josef Bacik 已提交
258
		if (comp_refs(ref, next, false))
259
			break;
260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293

		if (ref->action == next->action) {
			mod = next->ref_mod;
		} else {
			if (ref->ref_mod < next->ref_mod) {
				swap(ref, next);
				done = true;
			}
			mod = -next->ref_mod;
		}

		drop_delayed_ref(trans, delayed_refs, head, next);
		ref->ref_mod += mod;
		if (ref->ref_mod == 0) {
			drop_delayed_ref(trans, delayed_refs, head, ref);
			done = true;
		} else {
			/*
			 * Can't have multiples of the same ref on a tree block.
			 */
			WARN_ON(ref->type == BTRFS_TREE_BLOCK_REF_KEY ||
				ref->type == BTRFS_SHARED_BLOCK_REF_KEY);
		}
	}

	return done;
}

void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans,
			      struct btrfs_fs_info *fs_info,
			      struct btrfs_delayed_ref_root *delayed_refs,
			      struct btrfs_delayed_ref_head *head)
{
	struct btrfs_delayed_ref_node *ref;
294
	struct rb_node *node;
295 296
	u64 seq = 0;

297
	lockdep_assert_held(&head->lock);
298

299
	if (RB_EMPTY_ROOT(&head->ref_tree))
300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315
		return;

	/* We don't have too many refs to merge for data. */
	if (head->is_data)
		return;

	spin_lock(&fs_info->tree_mod_seq_lock);
	if (!list_empty(&fs_info->tree_mod_seq_list)) {
		struct seq_list *elem;

		elem = list_first_entry(&fs_info->tree_mod_seq_list,
					struct seq_list, list);
		seq = elem->seq;
	}
	spin_unlock(&fs_info->tree_mod_seq_lock);

316 317 318
again:
	for (node = rb_first(&head->ref_tree); node; node = rb_next(node)) {
		ref = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
319 320
		if (seq && ref->seq >= seq)
			continue;
321 322
		if (merge_ref(trans, delayed_refs, head, ref, seq))
			goto again;
323 324 325
	}
}

326 327
int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info,
			    struct btrfs_delayed_ref_root *delayed_refs,
328 329 330
			    u64 seq)
{
	struct seq_list *elem;
331 332 333 334 335 336 337
	int ret = 0;

	spin_lock(&fs_info->tree_mod_seq_lock);
	if (!list_empty(&fs_info->tree_mod_seq_list)) {
		elem = list_first_entry(&fs_info->tree_mod_seq_list,
					struct seq_list, list);
		if (seq >= elem->seq) {
338 339 340 341 342
			btrfs_debug(fs_info,
				"holding back delayed_ref %#x.%x, lowest is %#x.%x (%p)",
				(u32)(seq >> 32), (u32)seq,
				(u32)(elem->seq >> 32), (u32)elem->seq,
				delayed_refs);
343 344
			ret = 1;
		}
345
	}
346 347 348

	spin_unlock(&fs_info->tree_mod_seq_lock);
	return ret;
349 350
}

351 352
struct btrfs_delayed_ref_head *
btrfs_select_ref_head(struct btrfs_trans_handle *trans)
353 354
{
	struct btrfs_delayed_ref_root *delayed_refs;
355 356 357
	struct btrfs_delayed_ref_head *head;
	u64 start;
	bool loop = false;
358

359
	delayed_refs = &trans->transaction->delayed_refs;
L
Liu Bo 已提交
360

361
again:
362
	start = delayed_refs->run_delayed_start;
363
	head = find_ref_head(&delayed_refs->href_root, start, 1);
364 365
	if (!head && !loop) {
		delayed_refs->run_delayed_start = 0;
366
		start = 0;
367
		loop = true;
368
		head = find_ref_head(&delayed_refs->href_root, start, 1);
369 370 371 372
		if (!head)
			return NULL;
	} else if (!head && loop) {
		return NULL;
373
	}
374

375 376 377 378 379 380 381 382 383 384 385 386 387 388 389
	while (head->processing) {
		struct rb_node *node;

		node = rb_next(&head->href_node);
		if (!node) {
			if (loop)
				return NULL;
			delayed_refs->run_delayed_start = 0;
			start = 0;
			loop = true;
			goto again;
		}
		head = rb_entry(node, struct btrfs_delayed_ref_head,
				href_node);
	}
390

391 392 393
	head->processing = 1;
	WARN_ON(delayed_refs->num_heads_ready == 0);
	delayed_refs->num_heads_ready--;
394 395
	delayed_refs->run_delayed_start = head->bytenr +
		head->num_bytes;
396
	return head;
397 398
}

399 400 401 402 403 404
/*
 * Helper to insert the ref_node to the tail or merge with tail.
 *
 * Return 0 for insert.
 * Return >0 for merge.
 */
405 406 407 408
static int insert_delayed_ref(struct btrfs_trans_handle *trans,
			      struct btrfs_delayed_ref_root *root,
			      struct btrfs_delayed_ref_head *href,
			      struct btrfs_delayed_ref_node *ref)
409 410 411 412 413 414
{
	struct btrfs_delayed_ref_node *exist;
	int mod;
	int ret = 0;

	spin_lock(&href->lock);
415 416 417
	exist = tree_insert(&href->ref_tree, ref);
	if (!exist)
		goto inserted;
418 419 420 421 422 423 424 425 426 427 428

	/* Now we are sure we can merge */
	ret = 1;
	if (exist->action == ref->action) {
		mod = ref->ref_mod;
	} else {
		/* Need to change action */
		if (exist->ref_mod < ref->ref_mod) {
			exist->action = ref->action;
			mod = -exist->ref_mod;
			exist->ref_mod = ref->ref_mod;
429 430 431 432 433 434 435 436 437
			if (ref->action == BTRFS_ADD_DELAYED_REF)
				list_add_tail(&exist->add_list,
					      &href->ref_add_list);
			else if (ref->action == BTRFS_DROP_DELAYED_REF) {
				ASSERT(!list_empty(&exist->add_list));
				list_del(&exist->add_list);
			} else {
				ASSERT(0);
			}
438 439 440 441 442 443 444 445 446 447
		} else
			mod = -ref->ref_mod;
	}
	exist->ref_mod += mod;

	/* remove existing tail if its ref_mod is zero */
	if (exist->ref_mod == 0)
		drop_delayed_ref(trans, root, href, exist);
	spin_unlock(&href->lock);
	return ret;
448
inserted:
449 450
	if (ref->action == BTRFS_ADD_DELAYED_REF)
		list_add_tail(&ref->add_list, &href->ref_add_list);
451 452 453 454 455 456
	atomic_inc(&root->num_entries);
	trans->delayed_ref_updates++;
	spin_unlock(&href->lock);
	return ret;
}

457 458 459 460 461
/*
 * helper function to update the accounting in the head ref
 * existing and update must have the same bytenr
 */
static noinline void
462
update_existing_head_ref(struct btrfs_delayed_ref_root *delayed_refs,
463 464
			 struct btrfs_delayed_ref_head *existing,
			 struct btrfs_delayed_ref_head *update,
465
			 int *old_ref_mod_ret)
466
{
467
	int old_ref_mod;
468

469
	BUG_ON(existing->is_data != update->is_data);
470

471 472
	spin_lock(&existing->lock);
	if (update->must_insert_reserved) {
473 474 475 476 477 478 479
		/* if the extent was freed and then
		 * reallocated before the delayed ref
		 * entries were processed, we can end up
		 * with an existing head ref without
		 * the must_insert_reserved flag set.
		 * Set it again here
		 */
480
		existing->must_insert_reserved = update->must_insert_reserved;
481 482 483 484 485 486 487 488 489

		/*
		 * update the num_bytes so we make sure the accounting
		 * is done correctly
		 */
		existing->num_bytes = update->num_bytes;

	}

490 491 492
	if (update->extent_op) {
		if (!existing->extent_op) {
			existing->extent_op = update->extent_op;
493
		} else {
494 495 496 497 498
			if (update->extent_op->update_key) {
				memcpy(&existing->extent_op->key,
				       &update->extent_op->key,
				       sizeof(update->extent_op->key));
				existing->extent_op->update_key = true;
499
			}
500 501 502 503
			if (update->extent_op->update_flags) {
				existing->extent_op->flags_to_set |=
					update->extent_op->flags_to_set;
				existing->extent_op->update_flags = true;
504
			}
505
			btrfs_free_delayed_extent_op(update->extent_op);
506 507
		}
	}
508
	/*
509 510 511
	 * update the reference mod on the head to reflect this new operation,
	 * only need the lock for this case cause we could be processing it
	 * currently, for refs we just added we know we're a-ok.
512
	 */
513
	old_ref_mod = existing->total_ref_mod;
514 515
	if (old_ref_mod_ret)
		*old_ref_mod_ret = old_ref_mod;
516
	existing->ref_mod += update->ref_mod;
517
	existing->total_ref_mod += update->ref_mod;
518 519 520 521 522

	/*
	 * If we are going to from a positive ref mod to a negative or vice
	 * versa we need to make sure to adjust pending_csums accordingly.
	 */
523 524
	if (existing->is_data) {
		if (existing->total_ref_mod >= 0 && old_ref_mod < 0)
525
			delayed_refs->pending_csums -= existing->num_bytes;
526
		if (existing->total_ref_mod < 0 && old_ref_mod >= 0)
527 528
			delayed_refs->pending_csums += existing->num_bytes;
	}
529
	spin_unlock(&existing->lock);
530 531 532
}

/*
533
 * helper function to actually insert a head node into the rbtree.
534
 * this does all the dirty work in terms of maintaining the correct
535
 * overall modification count.
536
 */
537 538 539
static noinline struct btrfs_delayed_ref_head *
add_delayed_ref_head(struct btrfs_fs_info *fs_info,
		     struct btrfs_trans_handle *trans,
540
		     struct btrfs_delayed_ref_head *head_ref,
541
		     struct btrfs_qgroup_extent_record *qrecord,
542
		     u64 bytenr, u64 num_bytes, u64 ref_root, u64 reserved,
543 544
		     int action, int is_data, int *qrecord_inserted_ret,
		     int *old_ref_mod, int *new_ref_mod)
545
{
546
	struct btrfs_delayed_ref_head *existing;
547 548 549
	struct btrfs_delayed_ref_root *delayed_refs;
	int count_mod = 1;
	int must_insert_reserved = 0;
550
	int qrecord_inserted = 0;
551

552 553 554
	/* If reserved is provided, it must be a data extent. */
	BUG_ON(!is_data && reserved);

555 556 557 558
	/*
	 * the head node stores the sum of all the mods, so dropping a ref
	 * should drop the sum in the head node by one.
	 */
559 560 561 562
	if (action == BTRFS_UPDATE_DELAYED_HEAD)
		count_mod = 0;
	else if (action == BTRFS_DROP_DELAYED_REF)
		count_mod = -1;
563 564 565 566 567 568 569 570 571 572 573 574

	/*
	 * BTRFS_ADD_DELAYED_EXTENT means that we need to update
	 * the reserved accounting when the extent is finally added, or
	 * if a later modification deletes the delayed ref without ever
	 * inserting the extent into the extent allocation tree.
	 * ref->must_insert_reserved is the flag used to record
	 * that accounting mods are required.
	 *
	 * Once we record must_insert_reserved, switch the action to
	 * BTRFS_ADD_DELAYED_REF because other special casing is not required.
	 */
575
	if (action == BTRFS_ADD_DELAYED_EXTENT)
576
		must_insert_reserved = 1;
577
	else
578 579 580 581
		must_insert_reserved = 0;

	delayed_refs = &trans->transaction->delayed_refs;

582 583 584 585
	refcount_set(&head_ref->refs, 1);
	head_ref->bytenr = bytenr;
	head_ref->num_bytes = num_bytes;
	head_ref->ref_mod = count_mod;
586 587
	head_ref->must_insert_reserved = must_insert_reserved;
	head_ref->is_data = is_data;
588
	head_ref->ref_tree = RB_ROOT;
589
	INIT_LIST_HEAD(&head_ref->ref_add_list);
590
	RB_CLEAR_NODE(&head_ref->href_node);
591
	head_ref->processing = 0;
592
	head_ref->total_ref_mod = count_mod;
593 594
	head_ref->qgroup_reserved = 0;
	head_ref->qgroup_ref_root = 0;
595 596
	spin_lock_init(&head_ref->lock);
	mutex_init(&head_ref->mutex);
597

598 599
	/* Record qgroup extent info if provided */
	if (qrecord) {
600 601 602 603 604
		if (ref_root && reserved) {
			head_ref->qgroup_ref_root = ref_root;
			head_ref->qgroup_reserved = reserved;
		}

605 606 607 608
		qrecord->bytenr = bytenr;
		qrecord->num_bytes = num_bytes;
		qrecord->old_roots = NULL;

609
		if(btrfs_qgroup_trace_extent_nolock(fs_info,
610
					delayed_refs, qrecord))
611
			kfree(qrecord);
612 613
		else
			qrecord_inserted = 1;
614 615
	}

616
	trace_add_delayed_ref_head(fs_info, head_ref, action);
617

618 619
	existing = htree_insert(&delayed_refs->href_root,
				&head_ref->href_node);
620
	if (existing) {
621 622
		WARN_ON(ref_root && reserved && existing->qgroup_ref_root
			&& existing->qgroup_reserved);
623
		update_existing_head_ref(delayed_refs, existing, head_ref,
624
					 old_ref_mod);
625 626 627 628
		/*
		 * we've updated the existing ref, free the newly
		 * allocated ref
		 */
629
		kmem_cache_free(btrfs_delayed_ref_head_cachep, head_ref);
630
		head_ref = existing;
631
	} else {
632 633
		if (old_ref_mod)
			*old_ref_mod = 0;
634 635
		if (is_data && count_mod < 0)
			delayed_refs->pending_csums += num_bytes;
636 637
		delayed_refs->num_heads++;
		delayed_refs->num_heads_ready++;
638
		atomic_inc(&delayed_refs->num_entries);
639 640
		trans->delayed_ref_updates++;
	}
641 642
	if (qrecord_inserted_ret)
		*qrecord_inserted_ret = qrecord_inserted;
643 644
	if (new_ref_mod)
		*new_ref_mod = head_ref->total_ref_mod;
645
	return head_ref;
646 647 648 649 650
}

/*
 * helper to insert a delayed tree ref into the rbtree.
 */
651 652 653 654 655 656
static noinline void
add_delayed_tree_ref(struct btrfs_fs_info *fs_info,
		     struct btrfs_trans_handle *trans,
		     struct btrfs_delayed_ref_head *head_ref,
		     struct btrfs_delayed_ref_node *ref, u64 bytenr,
		     u64 num_bytes, u64 parent, u64 ref_root, int level,
657
		     int action)
658 659 660
{
	struct btrfs_delayed_tree_ref *full_ref;
	struct btrfs_delayed_ref_root *delayed_refs;
661
	u64 seq = 0;
662
	int ret;
663 664 665 666

	if (action == BTRFS_ADD_DELAYED_EXTENT)
		action = BTRFS_ADD_DELAYED_REF;

J
Josef Bacik 已提交
667 668
	if (is_fstree(ref_root))
		seq = atomic64_read(&fs_info->tree_mod_seq);
669 670 671
	delayed_refs = &trans->transaction->delayed_refs;

	/* first set the basic ref node struct up */
672
	refcount_set(&ref->refs, 1);
673
	ref->bytenr = bytenr;
674
	ref->num_bytes = num_bytes;
675 676 677 678
	ref->ref_mod = 1;
	ref->action = action;
	ref->is_head = 0;
	ref->in_tree = 1;
679
	ref->seq = seq;
680
	RB_CLEAR_NODE(&ref->ref_node);
681
	INIT_LIST_HEAD(&ref->add_list);
682

683
	full_ref = btrfs_delayed_node_to_tree_ref(ref);
684 685 686
	full_ref->parent = parent;
	full_ref->root = ref_root;
	if (parent)
687
		ref->type = BTRFS_SHARED_BLOCK_REF_KEY;
688
	else
689 690
		ref->type = BTRFS_TREE_BLOCK_REF_KEY;
	full_ref->level = level;
691

692
	trace_add_delayed_tree_ref(fs_info, ref, full_ref, action);
693

694
	ret = insert_delayed_ref(trans, delayed_refs, head_ref, ref);
695 696 697 698 699 700

	/*
	 * XXX: memory should be freed at the same level allocated.
	 * But bad practice is anywhere... Follow it now. Need cleanup.
	 */
	if (ret > 0)
701
		kmem_cache_free(btrfs_delayed_tree_ref_cachep, full_ref);
702 703 704 705 706
}

/*
 * helper to insert a delayed data ref into the rbtree.
 */
707 708 709 710 711 712
static noinline void
add_delayed_data_ref(struct btrfs_fs_info *fs_info,
		     struct btrfs_trans_handle *trans,
		     struct btrfs_delayed_ref_head *head_ref,
		     struct btrfs_delayed_ref_node *ref, u64 bytenr,
		     u64 num_bytes, u64 parent, u64 ref_root, u64 owner,
713
		     u64 offset, int action)
714 715 716
{
	struct btrfs_delayed_data_ref *full_ref;
	struct btrfs_delayed_ref_root *delayed_refs;
717
	u64 seq = 0;
718
	int ret;
719 720 721 722 723 724

	if (action == BTRFS_ADD_DELAYED_EXTENT)
		action = BTRFS_ADD_DELAYED_REF;

	delayed_refs = &trans->transaction->delayed_refs;

J
Josef Bacik 已提交
725 726 727
	if (is_fstree(ref_root))
		seq = atomic64_read(&fs_info->tree_mod_seq);

728
	/* first set the basic ref node struct up */
729
	refcount_set(&ref->refs, 1);
730 731 732 733 734 735
	ref->bytenr = bytenr;
	ref->num_bytes = num_bytes;
	ref->ref_mod = 1;
	ref->action = action;
	ref->is_head = 0;
	ref->in_tree = 1;
736
	ref->seq = seq;
737
	RB_CLEAR_NODE(&ref->ref_node);
738
	INIT_LIST_HEAD(&ref->add_list);
739

740
	full_ref = btrfs_delayed_node_to_data_ref(ref);
741 742 743
	full_ref->parent = parent;
	full_ref->root = ref_root;
	if (parent)
744
		ref->type = BTRFS_SHARED_DATA_REF_KEY;
745
	else
746
		ref->type = BTRFS_EXTENT_DATA_REF_KEY;
A
Arne Jansen 已提交
747

748 749
	full_ref->objectid = owner;
	full_ref->offset = offset;
750

751
	trace_add_delayed_data_ref(fs_info, ref, full_ref, action);
752

753
	ret = insert_delayed_ref(trans, delayed_refs, head_ref, ref);
754
	if (ret > 0)
755
		kmem_cache_free(btrfs_delayed_data_ref_cachep, full_ref);
756 757 758
}

/*
759
 * add a delayed tree ref.  This does all of the accounting required
760 761 762
 * to make sure the delayed ref is eventually processed before this
 * transaction commits.
 */
A
Arne Jansen 已提交
763 764
int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info,
			       struct btrfs_trans_handle *trans,
765 766
			       u64 bytenr, u64 num_bytes, u64 parent,
			       u64 ref_root,  int level, int action,
767 768
			       struct btrfs_delayed_extent_op *extent_op,
			       int *old_ref_mod, int *new_ref_mod)
769
{
770
	struct btrfs_delayed_tree_ref *ref;
771 772
	struct btrfs_delayed_ref_head *head_ref;
	struct btrfs_delayed_ref_root *delayed_refs;
773
	struct btrfs_qgroup_extent_record *record = NULL;
774
	int qrecord_inserted;
775

776
	BUG_ON(extent_op && extent_op->is_data);
777
	ref = kmem_cache_alloc(btrfs_delayed_tree_ref_cachep, GFP_NOFS);
778 779 780
	if (!ref)
		return -ENOMEM;

781
	head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS);
782 783
	if (!head_ref)
		goto free_ref;
784

785 786
	if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) &&
	    is_fstree(ref_root)) {
787
		record = kmalloc(sizeof(*record), GFP_NOFS);
788 789
		if (!record)
			goto free_head_ref;
790 791
	}

792 793 794 795 796
	head_ref->extent_op = extent_op;

	delayed_refs = &trans->transaction->delayed_refs;
	spin_lock(&delayed_refs->lock);

797
	/*
798 799
	 * insert both the head node and the new ref without dropping
	 * the spin lock
800
	 */
801
	head_ref = add_delayed_ref_head(fs_info, trans, head_ref, record,
802
					bytenr, num_bytes, 0, 0, action, 0,
803 804
					&qrecord_inserted, old_ref_mod,
					new_ref_mod);
805

806
	add_delayed_tree_ref(fs_info, trans, head_ref, &ref->node, bytenr,
807
			     num_bytes, parent, ref_root, level, action);
808
	spin_unlock(&delayed_refs->lock);
809

810
	if (qrecord_inserted)
811 812
		btrfs_qgroup_trace_extent_post(fs_info, record);

813
	return 0;
814 815 816 817 818 819 820

free_head_ref:
	kmem_cache_free(btrfs_delayed_ref_head_cachep, head_ref);
free_ref:
	kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref);

	return -ENOMEM;
821 822 823 824 825
}

/*
 * add a delayed data ref. it's similar to btrfs_add_delayed_tree_ref.
 */
A
Arne Jansen 已提交
826 827
int btrfs_add_delayed_data_ref(struct btrfs_fs_info *fs_info,
			       struct btrfs_trans_handle *trans,
828 829
			       u64 bytenr, u64 num_bytes,
			       u64 parent, u64 ref_root,
830 831
			       u64 owner, u64 offset, u64 reserved, int action,
			       int *old_ref_mod, int *new_ref_mod)
832 833 834 835
{
	struct btrfs_delayed_data_ref *ref;
	struct btrfs_delayed_ref_head *head_ref;
	struct btrfs_delayed_ref_root *delayed_refs;
836
	struct btrfs_qgroup_extent_record *record = NULL;
837
	int qrecord_inserted;
838

839
	ref = kmem_cache_alloc(btrfs_delayed_data_ref_cachep, GFP_NOFS);
840 841
	if (!ref)
		return -ENOMEM;
842

843
	head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS);
844
	if (!head_ref) {
845
		kmem_cache_free(btrfs_delayed_data_ref_cachep, ref);
846 847
		return -ENOMEM;
	}
848

849 850
	if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) &&
	    is_fstree(ref_root)) {
851 852 853 854 855 856 857 858 859
		record = kmalloc(sizeof(*record), GFP_NOFS);
		if (!record) {
			kmem_cache_free(btrfs_delayed_data_ref_cachep, ref);
			kmem_cache_free(btrfs_delayed_ref_head_cachep,
					head_ref);
			return -ENOMEM;
		}
	}

860
	head_ref->extent_op = NULL;
861

862 863 864 865 866 867 868
	delayed_refs = &trans->transaction->delayed_refs;
	spin_lock(&delayed_refs->lock);

	/*
	 * insert both the head node and the new ref without dropping
	 * the spin lock
	 */
869
	head_ref = add_delayed_ref_head(fs_info, trans, head_ref, record,
870
					bytenr, num_bytes, ref_root, reserved,
871 872
					action, 1, &qrecord_inserted,
					old_ref_mod, new_ref_mod);
873

874
	add_delayed_data_ref(fs_info, trans, head_ref, &ref->node, bytenr,
A
Arne Jansen 已提交
875
				   num_bytes, parent, ref_root, owner, offset,
876
				   action);
877
	spin_unlock(&delayed_refs->lock);
878

879 880
	if (qrecord_inserted)
		return btrfs_qgroup_trace_extent_post(fs_info, record);
881 882 883
	return 0;
}

A
Arne Jansen 已提交
884 885
int btrfs_add_delayed_extent_op(struct btrfs_fs_info *fs_info,
				struct btrfs_trans_handle *trans,
886 887 888 889 890 891
				u64 bytenr, u64 num_bytes,
				struct btrfs_delayed_extent_op *extent_op)
{
	struct btrfs_delayed_ref_head *head_ref;
	struct btrfs_delayed_ref_root *delayed_refs;

892
	head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS);
893 894 895 896 897 898 899 900
	if (!head_ref)
		return -ENOMEM;

	head_ref->extent_op = extent_op;

	delayed_refs = &trans->transaction->delayed_refs;
	spin_lock(&delayed_refs->lock);

901
	add_delayed_ref_head(fs_info, trans, head_ref, NULL, bytenr,
902
			     num_bytes, 0, 0, BTRFS_UPDATE_DELAYED_HEAD,
903
			     extent_op->is_data, NULL, NULL, NULL);
904

905 906 907 908
	spin_unlock(&delayed_refs->lock);
	return 0;
}

909 910 911 912 913 914
/*
 * this does a simple search for the head node for a given extent.
 * It must be called with the delayed ref spinlock held, and it returns
 * the head node if any where found, or NULL if not.
 */
struct btrfs_delayed_ref_head *
915
btrfs_find_delayed_ref_head(struct btrfs_delayed_ref_root *delayed_refs, u64 bytenr)
916
{
917
	return find_ref_head(&delayed_refs->href_root, bytenr, 0);
918
}
919

920
void __cold btrfs_delayed_ref_exit(void)
921
{
922 923 924 925
	kmem_cache_destroy(btrfs_delayed_ref_head_cachep);
	kmem_cache_destroy(btrfs_delayed_tree_ref_cachep);
	kmem_cache_destroy(btrfs_delayed_data_ref_cachep);
	kmem_cache_destroy(btrfs_delayed_extent_op_cachep);
926 927
}

928
int __init btrfs_delayed_ref_init(void)
929 930 931 932
{
	btrfs_delayed_ref_head_cachep = kmem_cache_create(
				"btrfs_delayed_ref_head",
				sizeof(struct btrfs_delayed_ref_head), 0,
933
				SLAB_MEM_SPREAD, NULL);
934 935 936 937 938 939
	if (!btrfs_delayed_ref_head_cachep)
		goto fail;

	btrfs_delayed_tree_ref_cachep = kmem_cache_create(
				"btrfs_delayed_tree_ref",
				sizeof(struct btrfs_delayed_tree_ref), 0,
940
				SLAB_MEM_SPREAD, NULL);
941 942 943 944 945 946
	if (!btrfs_delayed_tree_ref_cachep)
		goto fail;

	btrfs_delayed_data_ref_cachep = kmem_cache_create(
				"btrfs_delayed_data_ref",
				sizeof(struct btrfs_delayed_data_ref), 0,
947
				SLAB_MEM_SPREAD, NULL);
948 949 950 951 952 953
	if (!btrfs_delayed_data_ref_cachep)
		goto fail;

	btrfs_delayed_extent_op_cachep = kmem_cache_create(
				"btrfs_delayed_extent_op",
				sizeof(struct btrfs_delayed_extent_op), 0,
954
				SLAB_MEM_SPREAD, NULL);
955 956 957 958 959 960 961 962
	if (!btrfs_delayed_extent_op_cachep)
		goto fail;

	return 0;
fail:
	btrfs_delayed_ref_exit();
	return -ENOMEM;
}