refcounttree.c 122 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * refcounttree.c
 *
 * Copyright (C) 2009 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 version 2 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.
 */

18
#include <linux/sort.h>
19 20 21 22 23 24 25 26 27 28
#include <cluster/masklog.h>
#include "ocfs2.h"
#include "inode.h"
#include "alloc.h"
#include "suballoc.h"
#include "journal.h"
#include "uptodate.h"
#include "super.h"
#include "buffer_head_io.h"
#include "blockcheck.h"
29
#include "refcounttree.h"
Tao Ma's avatar
Tao Ma committed
30
#include "sysfile.h"
31
#include "dlmglue.h"
32
#include "extent_map.h"
Tao Ma's avatar
Tao Ma committed
33
#include "aops.h"
Tao Ma's avatar
Tao Ma committed
34
#include "xattr.h"
Tao Ma's avatar
Tao Ma committed
35
#include "namei.h"
Tao Ma's avatar
Tao Ma committed
36
#include "ocfs2_trace.h"
37
#include "file.h"
Tao Ma's avatar
Tao Ma committed
38 39 40 41 42 43 44

#include <linux/bio.h>
#include <linux/blkdev.h>
#include <linux/slab.h>
#include <linux/writeback.h>
#include <linux/pagevec.h>
#include <linux/swap.h>
Tao Ma's avatar
Tao Ma committed
45 46 47 48 49
#include <linux/security.h>
#include <linux/fsnotify.h>
#include <linux/quotaops.h>
#include <linux/namei.h>
#include <linux/mount.h>
50
#include <linux/posix_acl.h>
Tao Ma's avatar
Tao Ma committed
51 52 53 54 55

struct ocfs2_cow_context {
	struct inode *inode;
	u32 cow_start;
	u32 cow_len;
56 57
	struct ocfs2_extent_tree data_et;
	struct ocfs2_refcount_tree *ref_tree;
Tao Ma's avatar
Tao Ma committed
58 59 60 61
	struct buffer_head *ref_root_bh;
	struct ocfs2_alloc_context *meta_ac;
	struct ocfs2_alloc_context *data_ac;
	struct ocfs2_cached_dealloc_ctxt dealloc;
Tao Ma's avatar
Tao Ma committed
62 63 64
	void *cow_object;
	struct ocfs2_post_refcount *post_refcount;
	int extra_credits;
65 66 67 68 69
	int (*get_clusters)(struct ocfs2_cow_context *context,
			    u32 v_cluster, u32 *p_cluster,
			    u32 *num_clusters,
			    unsigned int *extent_flags);
	int (*cow_duplicate_clusters)(handle_t *handle,
70
				      struct inode *inode,
71 72
				      u32 cpos, u32 old_cluster,
				      u32 new_cluster, u32 new_len);
Tao Ma's avatar
Tao Ma committed
73
};
74 75 76 77 78 79

static inline struct ocfs2_refcount_tree *
cache_info_to_refcount(struct ocfs2_caching_info *ci)
{
	return container_of(ci, struct ocfs2_refcount_tree, rf_ci);
}
80 81 82 83 84 85 86 87

static int ocfs2_validate_refcount_block(struct super_block *sb,
					 struct buffer_head *bh)
{
	int rc;
	struct ocfs2_refcount_block *rb =
		(struct ocfs2_refcount_block *)bh->b_data;

Tao Ma's avatar
Tao Ma committed
88
	trace_ocfs2_validate_refcount_block((unsigned long long)bh->b_blocknr);
89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105

	BUG_ON(!buffer_uptodate(bh));

	/*
	 * If the ecc fails, we return the error but otherwise
	 * leave the filesystem running.  We know any error is
	 * local to this block.
	 */
	rc = ocfs2_validate_meta_ecc(sb, bh->b_data, &rb->rf_check);
	if (rc) {
		mlog(ML_ERROR, "Checksum failed for refcount block %llu\n",
		     (unsigned long long)bh->b_blocknr);
		return rc;
	}


	if (!OCFS2_IS_VALID_REFCOUNT_BLOCK(rb)) {
106
		rc = ocfs2_error(sb,
107 108 109
				 "Refcount block #%llu has bad signature %.*s\n",
				 (unsigned long long)bh->b_blocknr, 7,
				 rb->rf_signature);
110
		goto out;
111 112 113
	}

	if (le64_to_cpu(rb->rf_blkno) != bh->b_blocknr) {
114
		rc = ocfs2_error(sb,
115 116 117
				 "Refcount block #%llu has an invalid rf_blkno of %llu\n",
				 (unsigned long long)bh->b_blocknr,
				 (unsigned long long)le64_to_cpu(rb->rf_blkno));
118
		goto out;
119 120 121
	}

	if (le32_to_cpu(rb->rf_fs_generation) != OCFS2_SB(sb)->fs_generation) {
122
		rc = ocfs2_error(sb,
123 124 125
				 "Refcount block #%llu has an invalid rf_fs_generation of #%u\n",
				 (unsigned long long)bh->b_blocknr,
				 le32_to_cpu(rb->rf_fs_generation));
126
		goto out;
127
	}
128 129
out:
	return rc;
130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147
}

static int ocfs2_read_refcount_block(struct ocfs2_caching_info *ci,
				     u64 rb_blkno,
				     struct buffer_head **bh)
{
	int rc;
	struct buffer_head *tmp = *bh;

	rc = ocfs2_read_block(ci, rb_blkno, &tmp,
			      ocfs2_validate_refcount_block);

	/* If ocfs2_read_block() got us a new bh, pass it up. */
	if (!rc && !*bh)
		*bh = tmp;

	return rc;
}
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

static u64 ocfs2_refcount_cache_owner(struct ocfs2_caching_info *ci)
{
	struct ocfs2_refcount_tree *rf = cache_info_to_refcount(ci);

	return rf->rf_blkno;
}

static struct super_block *
ocfs2_refcount_cache_get_super(struct ocfs2_caching_info *ci)
{
	struct ocfs2_refcount_tree *rf = cache_info_to_refcount(ci);

	return rf->rf_sb;
}

static void ocfs2_refcount_cache_lock(struct ocfs2_caching_info *ci)
{
	struct ocfs2_refcount_tree *rf = cache_info_to_refcount(ci);

	spin_lock(&rf->rf_lock);
}

static void ocfs2_refcount_cache_unlock(struct ocfs2_caching_info *ci)
{
	struct ocfs2_refcount_tree *rf = cache_info_to_refcount(ci);

	spin_unlock(&rf->rf_lock);
}

static void ocfs2_refcount_cache_io_lock(struct ocfs2_caching_info *ci)
{
	struct ocfs2_refcount_tree *rf = cache_info_to_refcount(ci);

	mutex_lock(&rf->rf_io_mutex);
}

static void ocfs2_refcount_cache_io_unlock(struct ocfs2_caching_info *ci)
{
	struct ocfs2_refcount_tree *rf = cache_info_to_refcount(ci);

	mutex_unlock(&rf->rf_io_mutex);
}

static const struct ocfs2_caching_operations ocfs2_refcount_caching_ops = {
	.co_owner		= ocfs2_refcount_cache_owner,
	.co_get_super		= ocfs2_refcount_cache_get_super,
	.co_cache_lock		= ocfs2_refcount_cache_lock,
	.co_cache_unlock	= ocfs2_refcount_cache_unlock,
	.co_io_lock		= ocfs2_refcount_cache_io_lock,
	.co_io_unlock		= ocfs2_refcount_cache_io_unlock,
};
200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276

static struct ocfs2_refcount_tree *
ocfs2_find_refcount_tree(struct ocfs2_super *osb, u64 blkno)
{
	struct rb_node *n = osb->osb_rf_lock_tree.rb_node;
	struct ocfs2_refcount_tree *tree = NULL;

	while (n) {
		tree = rb_entry(n, struct ocfs2_refcount_tree, rf_node);

		if (blkno < tree->rf_blkno)
			n = n->rb_left;
		else if (blkno > tree->rf_blkno)
			n = n->rb_right;
		else
			return tree;
	}

	return NULL;
}

/* osb_lock is already locked. */
static void ocfs2_insert_refcount_tree(struct ocfs2_super *osb,
				       struct ocfs2_refcount_tree *new)
{
	u64 rf_blkno = new->rf_blkno;
	struct rb_node *parent = NULL;
	struct rb_node **p = &osb->osb_rf_lock_tree.rb_node;
	struct ocfs2_refcount_tree *tmp;

	while (*p) {
		parent = *p;

		tmp = rb_entry(parent, struct ocfs2_refcount_tree,
			       rf_node);

		if (rf_blkno < tmp->rf_blkno)
			p = &(*p)->rb_left;
		else if (rf_blkno > tmp->rf_blkno)
			p = &(*p)->rb_right;
		else {
			/* This should never happen! */
			mlog(ML_ERROR, "Duplicate refcount block %llu found!\n",
			     (unsigned long long)rf_blkno);
			BUG();
		}
	}

	rb_link_node(&new->rf_node, parent, p);
	rb_insert_color(&new->rf_node, &osb->osb_rf_lock_tree);
}

static void ocfs2_free_refcount_tree(struct ocfs2_refcount_tree *tree)
{
	ocfs2_metadata_cache_exit(&tree->rf_ci);
	ocfs2_simple_drop_lockres(OCFS2_SB(tree->rf_sb), &tree->rf_lockres);
	ocfs2_lock_res_free(&tree->rf_lockres);
	kfree(tree);
}

static inline void
ocfs2_erase_refcount_tree_from_list_no_lock(struct ocfs2_super *osb,
					struct ocfs2_refcount_tree *tree)
{
	rb_erase(&tree->rf_node, &osb->osb_rf_lock_tree);
	if (osb->osb_ref_tree_lru && osb->osb_ref_tree_lru == tree)
		osb->osb_ref_tree_lru = NULL;
}

static void ocfs2_erase_refcount_tree_from_list(struct ocfs2_super *osb,
					struct ocfs2_refcount_tree *tree)
{
	spin_lock(&osb->osb_lock);
	ocfs2_erase_refcount_tree_from_list_no_lock(osb, tree);
	spin_unlock(&osb->osb_lock);
}

Tao Ma's avatar
Tao Ma committed
277
static void ocfs2_kref_remove_refcount_tree(struct kref *kref)
278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314
{
	struct ocfs2_refcount_tree *tree =
		container_of(kref, struct ocfs2_refcount_tree, rf_getcnt);

	ocfs2_free_refcount_tree(tree);
}

static inline void
ocfs2_refcount_tree_get(struct ocfs2_refcount_tree *tree)
{
	kref_get(&tree->rf_getcnt);
}

static inline void
ocfs2_refcount_tree_put(struct ocfs2_refcount_tree *tree)
{
	kref_put(&tree->rf_getcnt, ocfs2_kref_remove_refcount_tree);
}

static inline void ocfs2_init_refcount_tree_ci(struct ocfs2_refcount_tree *new,
					       struct super_block *sb)
{
	ocfs2_metadata_cache_init(&new->rf_ci, &ocfs2_refcount_caching_ops);
	mutex_init(&new->rf_io_mutex);
	new->rf_sb = sb;
	spin_lock_init(&new->rf_lock);
}

static inline void ocfs2_init_refcount_tree_lock(struct ocfs2_super *osb,
					struct ocfs2_refcount_tree *new,
					u64 rf_blkno, u32 generation)
{
	init_rwsem(&new->rf_sem);
	ocfs2_refcount_lock_res_init(&new->rf_lockres, osb,
				     rf_blkno, generation);
}

Tao Ma's avatar
Tao Ma committed
315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330
static struct ocfs2_refcount_tree*
ocfs2_allocate_refcount_tree(struct ocfs2_super *osb, u64 rf_blkno)
{
	struct ocfs2_refcount_tree *new;

	new = kzalloc(sizeof(struct ocfs2_refcount_tree), GFP_NOFS);
	if (!new)
		return NULL;

	new->rf_blkno = rf_blkno;
	kref_init(&new->rf_getcnt);
	ocfs2_init_refcount_tree_ci(new, osb->sb);

	return new;
}

331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349
static int ocfs2_get_refcount_tree(struct ocfs2_super *osb, u64 rf_blkno,
				   struct ocfs2_refcount_tree **ret_tree)
{
	int ret = 0;
	struct ocfs2_refcount_tree *tree, *new = NULL;
	struct buffer_head *ref_root_bh = NULL;
	struct ocfs2_refcount_block *ref_rb;

	spin_lock(&osb->osb_lock);
	if (osb->osb_ref_tree_lru &&
	    osb->osb_ref_tree_lru->rf_blkno == rf_blkno)
		tree = osb->osb_ref_tree_lru;
	else
		tree = ocfs2_find_refcount_tree(osb, rf_blkno);
	if (tree)
		goto out;

	spin_unlock(&osb->osb_lock);

Tao Ma's avatar
Tao Ma committed
350
	new = ocfs2_allocate_refcount_tree(osb, rf_blkno);
351 352
	if (!new) {
		ret = -ENOMEM;
Tao Ma's avatar
Tao Ma committed
353
		mlog_errno(ret);
354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413
		return ret;
	}
	/*
	 * We need the generation to create the refcount tree lock and since
	 * it isn't changed during the tree modification, we are safe here to
	 * read without protection.
	 * We also have to purge the cache after we create the lock since the
	 * refcount block may have the stale data. It can only be trusted when
	 * we hold the refcount lock.
	 */
	ret = ocfs2_read_refcount_block(&new->rf_ci, rf_blkno, &ref_root_bh);
	if (ret) {
		mlog_errno(ret);
		ocfs2_metadata_cache_exit(&new->rf_ci);
		kfree(new);
		return ret;
	}

	ref_rb = (struct ocfs2_refcount_block *)ref_root_bh->b_data;
	new->rf_generation = le32_to_cpu(ref_rb->rf_generation);
	ocfs2_init_refcount_tree_lock(osb, new, rf_blkno,
				      new->rf_generation);
	ocfs2_metadata_cache_purge(&new->rf_ci);

	spin_lock(&osb->osb_lock);
	tree = ocfs2_find_refcount_tree(osb, rf_blkno);
	if (tree)
		goto out;

	ocfs2_insert_refcount_tree(osb, new);

	tree = new;
	new = NULL;

out:
	*ret_tree = tree;

	osb->osb_ref_tree_lru = tree;

	spin_unlock(&osb->osb_lock);

	if (new)
		ocfs2_free_refcount_tree(new);

	brelse(ref_root_bh);
	return ret;
}

static int ocfs2_get_refcount_block(struct inode *inode, u64 *ref_blkno)
{
	int ret;
	struct buffer_head *di_bh = NULL;
	struct ocfs2_dinode *di;

	ret = ocfs2_read_inode_block(inode, &di_bh);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

414
	BUG_ON(!ocfs2_is_refcount_inode(inode));
415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544

	di = (struct ocfs2_dinode *)di_bh->b_data;
	*ref_blkno = le64_to_cpu(di->i_refcount_loc);
	brelse(di_bh);
out:
	return ret;
}

static int __ocfs2_lock_refcount_tree(struct ocfs2_super *osb,
				      struct ocfs2_refcount_tree *tree, int rw)
{
	int ret;

	ret = ocfs2_refcount_lock(tree, rw);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	if (rw)
		down_write(&tree->rf_sem);
	else
		down_read(&tree->rf_sem);

out:
	return ret;
}

/*
 * Lock the refcount tree pointed by ref_blkno and return the tree.
 * In most case, we lock the tree and read the refcount block.
 * So read it here if the caller really needs it.
 *
 * If the tree has been re-created by other node, it will free the
 * old one and re-create it.
 */
int ocfs2_lock_refcount_tree(struct ocfs2_super *osb,
			     u64 ref_blkno, int rw,
			     struct ocfs2_refcount_tree **ret_tree,
			     struct buffer_head **ref_bh)
{
	int ret, delete_tree = 0;
	struct ocfs2_refcount_tree *tree = NULL;
	struct buffer_head *ref_root_bh = NULL;
	struct ocfs2_refcount_block *rb;

again:
	ret = ocfs2_get_refcount_tree(osb, ref_blkno, &tree);
	if (ret) {
		mlog_errno(ret);
		return ret;
	}

	ocfs2_refcount_tree_get(tree);

	ret = __ocfs2_lock_refcount_tree(osb, tree, rw);
	if (ret) {
		mlog_errno(ret);
		ocfs2_refcount_tree_put(tree);
		goto out;
	}

	ret = ocfs2_read_refcount_block(&tree->rf_ci, tree->rf_blkno,
					&ref_root_bh);
	if (ret) {
		mlog_errno(ret);
		ocfs2_unlock_refcount_tree(osb, tree, rw);
		goto out;
	}

	rb = (struct ocfs2_refcount_block *)ref_root_bh->b_data;
	/*
	 * If the refcount block has been freed and re-created, we may need
	 * to recreate the refcount tree also.
	 *
	 * Here we just remove the tree from the rb-tree, and the last
	 * kref holder will unlock and delete this refcount_tree.
	 * Then we goto "again" and ocfs2_get_refcount_tree will create
	 * the new refcount tree for us.
	 */
	if (tree->rf_generation != le32_to_cpu(rb->rf_generation)) {
		if (!tree->rf_removed) {
			ocfs2_erase_refcount_tree_from_list(osb, tree);
			tree->rf_removed = 1;
			delete_tree = 1;
		}

		ocfs2_unlock_refcount_tree(osb, tree, rw);
		/*
		 * We get an extra reference when we create the refcount
		 * tree, so another put will destroy it.
		 */
		if (delete_tree)
			ocfs2_refcount_tree_put(tree);
		brelse(ref_root_bh);
		ref_root_bh = NULL;
		goto again;
	}

	*ret_tree = tree;
	if (ref_bh) {
		*ref_bh = ref_root_bh;
		ref_root_bh = NULL;
	}
out:
	brelse(ref_root_bh);
	return ret;
}

void ocfs2_unlock_refcount_tree(struct ocfs2_super *osb,
				struct ocfs2_refcount_tree *tree, int rw)
{
	if (rw)
		up_write(&tree->rf_sem);
	else
		up_read(&tree->rf_sem);

	ocfs2_refcount_unlock(tree, rw);
	ocfs2_refcount_tree_put(tree);
}

void ocfs2_purge_refcount_trees(struct ocfs2_super *osb)
{
	struct rb_node *node;
	struct ocfs2_refcount_tree *tree;
	struct rb_root *root = &osb->osb_rf_lock_tree;

	while ((node = rb_last(root)) != NULL) {
		tree = rb_entry(node, struct ocfs2_refcount_tree, rf_node);

Tao Ma's avatar
Tao Ma committed
545 546
		trace_ocfs2_purge_refcount_trees(
				(unsigned long long) tree->rf_blkno);
547 548 549 550 551

		rb_erase(&tree->rf_node, root);
		ocfs2_free_refcount_tree(tree);
	}
}
Tao Ma's avatar
Tao Ma committed
552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570

/*
 * Create a refcount tree for an inode.
 * We take for granted that the inode is already locked.
 */
static int ocfs2_create_refcount_tree(struct inode *inode,
				      struct buffer_head *di_bh)
{
	int ret;
	handle_t *handle = NULL;
	struct ocfs2_alloc_context *meta_ac = NULL;
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	struct buffer_head *new_bh = NULL;
	struct ocfs2_refcount_block *rb;
	struct ocfs2_refcount_tree *new_tree = NULL, *tree = NULL;
	u16 suballoc_bit_start;
	u32 num_got;
571
	u64 suballoc_loc, first_blkno;
Tao Ma's avatar
Tao Ma committed
572

573
	BUG_ON(ocfs2_is_refcount_inode(inode));
Tao Ma's avatar
Tao Ma committed
574

Tao Ma's avatar
Tao Ma committed
575
	trace_ocfs2_create_refcount_tree(
576
		(unsigned long long)oi->ip_blkno);
Tao Ma's avatar
Tao Ma committed
577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597

	ret = ocfs2_reserve_new_metadata_blocks(osb, 1, &meta_ac);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	handle = ocfs2_start_trans(osb, OCFS2_REFCOUNT_TREE_CREATE_CREDITS);
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		mlog_errno(ret);
		goto out;
	}

	ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out_commit;
	}

598
	ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc,
Tao Ma's avatar
Tao Ma committed
599 600 601 602 603 604 605 606 607 608 609 610 611 612 613
				   &suballoc_bit_start, &num_got,
				   &first_blkno);
	if (ret) {
		mlog_errno(ret);
		goto out_commit;
	}

	new_tree = ocfs2_allocate_refcount_tree(osb, first_blkno);
	if (!new_tree) {
		ret = -ENOMEM;
		mlog_errno(ret);
		goto out_commit;
	}

	new_bh = sb_getblk(inode->i_sb, first_blkno);
614 615 616 617 618
	if (!new_bh) {
		ret = -ENOMEM;
		mlog_errno(ret);
		goto out_commit;
	}
Tao Ma's avatar
Tao Ma committed
619 620 621 622 623 624 625 626 627 628 629 630 631
	ocfs2_set_new_buffer_uptodate(&new_tree->rf_ci, new_bh);

	ret = ocfs2_journal_access_rb(handle, &new_tree->rf_ci, new_bh,
				      OCFS2_JOURNAL_ACCESS_CREATE);
	if (ret) {
		mlog_errno(ret);
		goto out_commit;
	}

	/* Initialize ocfs2_refcount_block. */
	rb = (struct ocfs2_refcount_block *)new_bh->b_data;
	memset(rb, 0, inode->i_sb->s_blocksize);
	strcpy((void *)rb, OCFS2_REFCOUNT_BLOCK_SIGNATURE);
632
	rb->rf_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot);
633
	rb->rf_suballoc_loc = cpu_to_le64(suballoc_loc);
Tao Ma's avatar
Tao Ma committed
634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651
	rb->rf_suballoc_bit = cpu_to_le16(suballoc_bit_start);
	rb->rf_fs_generation = cpu_to_le32(osb->fs_generation);
	rb->rf_blkno = cpu_to_le64(first_blkno);
	rb->rf_count = cpu_to_le32(1);
	rb->rf_records.rl_count =
			cpu_to_le16(ocfs2_refcount_recs_per_rb(osb->sb));
	spin_lock(&osb->osb_lock);
	rb->rf_generation = osb->s_next_generation++;
	spin_unlock(&osb->osb_lock);

	ocfs2_journal_dirty(handle, new_bh);

	spin_lock(&oi->ip_lock);
	oi->ip_dyn_features |= OCFS2_HAS_REFCOUNT_FL;
	di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
	di->i_refcount_loc = cpu_to_le64(first_blkno);
	spin_unlock(&oi->ip_lock);

Tao Ma's avatar
Tao Ma committed
652
	trace_ocfs2_create_refcount_tree_blkno((unsigned long long)first_blkno);
Tao Ma's avatar
Tao Ma committed
653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710

	ocfs2_journal_dirty(handle, di_bh);

	/*
	 * We have to init the tree lock here since it will use
	 * the generation number to create it.
	 */
	new_tree->rf_generation = le32_to_cpu(rb->rf_generation);
	ocfs2_init_refcount_tree_lock(osb, new_tree, first_blkno,
				      new_tree->rf_generation);

	spin_lock(&osb->osb_lock);
	tree = ocfs2_find_refcount_tree(osb, first_blkno);

	/*
	 * We've just created a new refcount tree in this block.  If
	 * we found a refcount tree on the ocfs2_super, it must be
	 * one we just deleted.  We free the old tree before
	 * inserting the new tree.
	 */
	BUG_ON(tree && tree->rf_generation == new_tree->rf_generation);
	if (tree)
		ocfs2_erase_refcount_tree_from_list_no_lock(osb, tree);
	ocfs2_insert_refcount_tree(osb, new_tree);
	spin_unlock(&osb->osb_lock);
	new_tree = NULL;
	if (tree)
		ocfs2_refcount_tree_put(tree);

out_commit:
	ocfs2_commit_trans(osb, handle);

out:
	if (new_tree) {
		ocfs2_metadata_cache_exit(&new_tree->rf_ci);
		kfree(new_tree);
	}

	brelse(new_bh);
	if (meta_ac)
		ocfs2_free_alloc_context(meta_ac);

	return ret;
}

static int ocfs2_set_refcount_tree(struct inode *inode,
				   struct buffer_head *di_bh,
				   u64 refcount_loc)
{
	int ret;
	handle_t *handle = NULL;
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	struct buffer_head *ref_root_bh = NULL;
	struct ocfs2_refcount_block *rb;
	struct ocfs2_refcount_tree *ref_tree;

711
	BUG_ON(ocfs2_is_refcount_inode(inode));
Tao Ma's avatar
Tao Ma committed
712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777

	ret = ocfs2_lock_refcount_tree(osb, refcount_loc, 1,
				       &ref_tree, &ref_root_bh);
	if (ret) {
		mlog_errno(ret);
		return ret;
	}

	handle = ocfs2_start_trans(osb, OCFS2_REFCOUNT_TREE_SET_CREDITS);
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		mlog_errno(ret);
		goto out;
	}

	ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out_commit;
	}

	ret = ocfs2_journal_access_rb(handle, &ref_tree->rf_ci, ref_root_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out_commit;
	}

	rb = (struct ocfs2_refcount_block *)ref_root_bh->b_data;
	le32_add_cpu(&rb->rf_count, 1);

	ocfs2_journal_dirty(handle, ref_root_bh);

	spin_lock(&oi->ip_lock);
	oi->ip_dyn_features |= OCFS2_HAS_REFCOUNT_FL;
	di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
	di->i_refcount_loc = cpu_to_le64(refcount_loc);
	spin_unlock(&oi->ip_lock);
	ocfs2_journal_dirty(handle, di_bh);

out_commit:
	ocfs2_commit_trans(osb, handle);
out:
	ocfs2_unlock_refcount_tree(osb, ref_tree, 1);
	brelse(ref_root_bh);

	return ret;
}

int ocfs2_remove_refcount_tree(struct inode *inode, struct buffer_head *di_bh)
{
	int ret, delete_tree = 0;
	handle_t *handle = NULL;
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	struct ocfs2_refcount_block *rb;
	struct inode *alloc_inode = NULL;
	struct buffer_head *alloc_bh = NULL;
	struct buffer_head *blk_bh = NULL;
	struct ocfs2_refcount_tree *ref_tree;
	int credits = OCFS2_REFCOUNT_TREE_REMOVE_CREDITS;
	u64 blk = 0, bg_blkno = 0, ref_blkno = le64_to_cpu(di->i_refcount_loc);
	u16 bit = 0;

778
	if (!ocfs2_is_refcount_inode(inode))
Tao Ma's avatar
Tao Ma committed
779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796
		return 0;

	BUG_ON(!ref_blkno);
	ret = ocfs2_lock_refcount_tree(osb, ref_blkno, 1, &ref_tree, &blk_bh);
	if (ret) {
		mlog_errno(ret);
		return ret;
	}

	rb = (struct ocfs2_refcount_block *)blk_bh->b_data;

	/*
	 * If we are the last user, we need to free the block.
	 * So lock the allocator ahead.
	 */
	if (le32_to_cpu(rb->rf_count) == 1) {
		blk = le64_to_cpu(rb->rf_blkno);
		bit = le16_to_cpu(rb->rf_suballoc_bit);
797 798 799 800
		if (rb->rf_suballoc_loc)
			bg_blkno = le64_to_cpu(rb->rf_suballoc_loc);
		else
			bg_blkno = ocfs2_which_suballoc_group(blk, bit);
Tao Ma's avatar
Tao Ma committed
801 802 803 804 805 806 807 808 809

		alloc_inode = ocfs2_get_system_file_inode(osb,
					EXTENT_ALLOC_SYSTEM_INODE,
					le16_to_cpu(rb->rf_suballoc_slot));
		if (!alloc_inode) {
			ret = -ENOMEM;
			mlog_errno(ret);
			goto out;
		}
Al Viro's avatar
Al Viro committed
810
		inode_lock(alloc_inode);
Tao Ma's avatar
Tao Ma committed
811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869

		ret = ocfs2_inode_lock(alloc_inode, &alloc_bh, 1);
		if (ret) {
			mlog_errno(ret);
			goto out_mutex;
		}

		credits += OCFS2_SUBALLOC_FREE;
	}

	handle = ocfs2_start_trans(osb, credits);
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		mlog_errno(ret);
		goto out_unlock;
	}

	ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out_commit;
	}

	ret = ocfs2_journal_access_rb(handle, &ref_tree->rf_ci, blk_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out_commit;
	}

	spin_lock(&oi->ip_lock);
	oi->ip_dyn_features &= ~OCFS2_HAS_REFCOUNT_FL;
	di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
	di->i_refcount_loc = 0;
	spin_unlock(&oi->ip_lock);
	ocfs2_journal_dirty(handle, di_bh);

	le32_add_cpu(&rb->rf_count , -1);
	ocfs2_journal_dirty(handle, blk_bh);

	if (!rb->rf_count) {
		delete_tree = 1;
		ocfs2_erase_refcount_tree_from_list(osb, ref_tree);
		ret = ocfs2_free_suballoc_bits(handle, alloc_inode,
					       alloc_bh, bit, bg_blkno, 1);
		if (ret)
			mlog_errno(ret);
	}

out_commit:
	ocfs2_commit_trans(osb, handle);
out_unlock:
	if (alloc_inode) {
		ocfs2_inode_unlock(alloc_inode, 1);
		brelse(alloc_bh);
	}
out_mutex:
	if (alloc_inode) {
Al Viro's avatar
Al Viro committed
870
		inode_unlock(alloc_inode);
Tao Ma's avatar
Tao Ma committed
871 872 873 874 875 876 877 878 879 880
		iput(alloc_inode);
	}
out:
	ocfs2_unlock_refcount_tree(osb, ref_tree, 1);
	if (delete_tree)
		ocfs2_refcount_tree_put(ref_tree);
	brelse(blk_bh);

	return ret;
}
881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923

static void ocfs2_find_refcount_rec_in_rl(struct ocfs2_caching_info *ci,
					  struct buffer_head *ref_leaf_bh,
					  u64 cpos, unsigned int len,
					  struct ocfs2_refcount_rec *ret_rec,
					  int *index)
{
	int i = 0;
	struct ocfs2_refcount_block *rb =
		(struct ocfs2_refcount_block *)ref_leaf_bh->b_data;
	struct ocfs2_refcount_rec *rec = NULL;

	for (; i < le16_to_cpu(rb->rf_records.rl_used); i++) {
		rec = &rb->rf_records.rl_recs[i];

		if (le64_to_cpu(rec->r_cpos) +
		    le32_to_cpu(rec->r_clusters) <= cpos)
			continue;
		else if (le64_to_cpu(rec->r_cpos) > cpos)
			break;

		/* ok, cpos fail in this rec. Just return. */
		if (ret_rec)
			*ret_rec = *rec;
		goto out;
	}

	if (ret_rec) {
		/* We meet with a hole here, so fake the rec. */
		ret_rec->r_cpos = cpu_to_le64(cpos);
		ret_rec->r_refcount = 0;
		if (i < le16_to_cpu(rb->rf_records.rl_used) &&
		    le64_to_cpu(rec->r_cpos) < cpos + len)
			ret_rec->r_clusters =
				cpu_to_le32(le64_to_cpu(rec->r_cpos) - cpos);
		else
			ret_rec->r_clusters = cpu_to_le32(len);
	}

out:
	*index = i;
}

924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959
/*
 * Try to remove refcount tree. The mechanism is:
 * 1) Check whether i_clusters == 0, if no, exit.
 * 2) check whether we have i_xattr_loc in dinode. if yes, exit.
 * 3) Check whether we have inline xattr stored outside, if yes, exit.
 * 4) Remove the tree.
 */
int ocfs2_try_remove_refcount_tree(struct inode *inode,
				   struct buffer_head *di_bh)
{
	int ret;
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;

	down_write(&oi->ip_xattr_sem);
	down_write(&oi->ip_alloc_sem);

	if (oi->ip_clusters)
		goto out;

	if ((oi->ip_dyn_features & OCFS2_HAS_XATTR_FL) && di->i_xattr_loc)
		goto out;

	if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL &&
	    ocfs2_has_inline_xattr_value_outside(inode, di))
		goto out;

	ret = ocfs2_remove_refcount_tree(inode, di_bh);
	if (ret)
		mlog_errno(ret);
out:
	up_write(&oi->ip_alloc_sem);
	up_write(&oi->ip_xattr_sem);
	return 0;
}

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 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041
/*
 * Find the end range for a leaf refcount block indicated by
 * el->l_recs[index].e_blkno.
 */
static int ocfs2_get_refcount_cpos_end(struct ocfs2_caching_info *ci,
				       struct buffer_head *ref_root_bh,
				       struct ocfs2_extent_block *eb,
				       struct ocfs2_extent_list *el,
				       int index,  u32 *cpos_end)
{
	int ret, i, subtree_root;
	u32 cpos;
	u64 blkno;
	struct super_block *sb = ocfs2_metadata_cache_get_super(ci);
	struct ocfs2_path *left_path = NULL, *right_path = NULL;
	struct ocfs2_extent_tree et;
	struct ocfs2_extent_list *tmp_el;

	if (index < le16_to_cpu(el->l_next_free_rec) - 1) {
		/*
		 * We have a extent rec after index, so just use the e_cpos
		 * of the next extent rec.
		 */
		*cpos_end = le32_to_cpu(el->l_recs[index+1].e_cpos);
		return 0;
	}

	if (!eb || (eb && !eb->h_next_leaf_blk)) {
		/*
		 * We are the last extent rec, so any high cpos should
		 * be stored in this leaf refcount block.
		 */
		*cpos_end = UINT_MAX;
		return 0;
	}

	/*
	 * If the extent block isn't the last one, we have to find
	 * the subtree root between this extent block and the next
	 * leaf extent block and get the corresponding e_cpos from
	 * the subroot. Otherwise we may corrupt the b-tree.
	 */
	ocfs2_init_refcount_extent_tree(&et, ci, ref_root_bh);

	left_path = ocfs2_new_path_from_et(&et);
	if (!left_path) {
		ret = -ENOMEM;
		mlog_errno(ret);
		goto out;
	}

	cpos = le32_to_cpu(eb->h_list.l_recs[index].e_cpos);
	ret = ocfs2_find_path(ci, left_path, cpos);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	right_path = ocfs2_new_path_from_path(left_path);
	if (!right_path) {
		ret = -ENOMEM;
		mlog_errno(ret);
		goto out;
	}

	ret = ocfs2_find_cpos_for_right_leaf(sb, left_path, &cpos);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	ret = ocfs2_find_path(ci, right_path, cpos);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	subtree_root = ocfs2_find_subtree_root(&et, left_path,
					       right_path);

	tmp_el = left_path->p_node[subtree_root].el;
	blkno = left_path->p_node[subtree_root+1].bh->b_blocknr;
1042
	for (i = 0; i < le16_to_cpu(tmp_el->l_next_free_rec); i++) {
1043 1044 1045 1046 1047 1048
		if (le64_to_cpu(tmp_el->l_recs[i].e_blkno) == blkno) {
			*cpos_end = le32_to_cpu(tmp_el->l_recs[i+1].e_cpos);
			break;
		}
	}

1049
	BUG_ON(i == le16_to_cpu(tmp_el->l_next_free_rec));
1050 1051 1052 1053 1054 1055 1056

out:
	ocfs2_free_path(left_path);
	ocfs2_free_path(right_path);
	return ret;
}

1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071
/*
 * Given a cpos and len, try to find the refcount record which contains cpos.
 * 1. If cpos can be found in one refcount record, return the record.
 * 2. If cpos can't be found, return a fake record which start from cpos
 *    and end at a small value between cpos+len and start of the next record.
 *    This fake record has r_refcount = 0.
 */
static int ocfs2_get_refcount_rec(struct ocfs2_caching_info *ci,
				  struct buffer_head *ref_root_bh,
				  u64 cpos, unsigned int len,
				  struct ocfs2_refcount_rec *ret_rec,
				  int *index,
				  struct buffer_head **ret_bh)
{
	int ret = 0, i, found;
1072
	u32 low_cpos, uninitialized_var(cpos_end);
1073
	struct ocfs2_extent_list *el;
1074 1075
	struct ocfs2_extent_rec *rec = NULL;
	struct ocfs2_extent_block *eb = NULL;
1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102
	struct buffer_head *eb_bh = NULL, *ref_leaf_bh = NULL;
	struct super_block *sb = ocfs2_metadata_cache_get_super(ci);
	struct ocfs2_refcount_block *rb =
			(struct ocfs2_refcount_block *)ref_root_bh->b_data;

	if (!(le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL)) {
		ocfs2_find_refcount_rec_in_rl(ci, ref_root_bh, cpos, len,
					      ret_rec, index);
		*ret_bh = ref_root_bh;
		get_bh(ref_root_bh);
		return 0;
	}

	el = &rb->rf_list;
	low_cpos = cpos & OCFS2_32BIT_POS_MASK;

	if (el->l_tree_depth) {
		ret = ocfs2_find_leaf(ci, el, low_cpos, &eb_bh);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}

		eb = (struct ocfs2_extent_block *) eb_bh->b_data;
		el = &eb->h_list;

		if (el->l_tree_depth) {
1103
			ret = ocfs2_error(sb,
1104 1105 1106
					  "refcount tree %llu has non zero tree depth in leaf btree tree block %llu\n",
					  (unsigned long long)ocfs2_metadata_cache_owner(ci),
					  (unsigned long long)eb_bh->b_blocknr);
1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120
			goto out;
		}
	}

	found = 0;
	for (i = le16_to_cpu(el->l_next_free_rec) - 1; i >= 0; i--) {
		rec = &el->l_recs[i];

		if (le32_to_cpu(rec->e_cpos) <= low_cpos) {
			found = 1;
			break;
		}
	}

1121 1122 1123 1124 1125 1126 1127
	if (found) {
		ret = ocfs2_get_refcount_cpos_end(ci, ref_root_bh,
						  eb, el, i, &cpos_end);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}
1128

1129 1130
		if (cpos_end < low_cpos + len)
			len = cpos_end - low_cpos;
1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236
	}

	ret = ocfs2_read_refcount_block(ci, le64_to_cpu(rec->e_blkno),
					&ref_leaf_bh);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	ocfs2_find_refcount_rec_in_rl(ci, ref_leaf_bh, cpos, len,
				      ret_rec, index);
	*ret_bh = ref_leaf_bh;
out:
	brelse(eb_bh);
	return ret;
}

enum ocfs2_ref_rec_contig {
	REF_CONTIG_NONE = 0,
	REF_CONTIG_LEFT,
	REF_CONTIG_RIGHT,
	REF_CONTIG_LEFTRIGHT,
};

static enum ocfs2_ref_rec_contig
	ocfs2_refcount_rec_adjacent(struct ocfs2_refcount_block *rb,
				    int index)
{
	if ((rb->rf_records.rl_recs[index].r_refcount ==
	    rb->rf_records.rl_recs[index + 1].r_refcount) &&
	    (le64_to_cpu(rb->rf_records.rl_recs[index].r_cpos) +
	    le32_to_cpu(rb->rf_records.rl_recs[index].r_clusters) ==
	    le64_to_cpu(rb->rf_records.rl_recs[index + 1].r_cpos)))
		return REF_CONTIG_RIGHT;

	return REF_CONTIG_NONE;
}

static enum ocfs2_ref_rec_contig
	ocfs2_refcount_rec_contig(struct ocfs2_refcount_block *rb,
				  int index)
{
	enum ocfs2_ref_rec_contig ret = REF_CONTIG_NONE;

	if (index < le16_to_cpu(rb->rf_records.rl_used) - 1)
		ret = ocfs2_refcount_rec_adjacent(rb, index);

	if (index > 0) {
		enum ocfs2_ref_rec_contig tmp;

		tmp = ocfs2_refcount_rec_adjacent(rb, index - 1);

		if (tmp == REF_CONTIG_RIGHT) {
			if (ret == REF_CONTIG_RIGHT)
				ret = REF_CONTIG_LEFTRIGHT;
			else
				ret = REF_CONTIG_LEFT;
		}
	}

	return ret;
}

static void ocfs2_rotate_refcount_rec_left(struct ocfs2_refcount_block *rb,
					   int index)
{
	BUG_ON(rb->rf_records.rl_recs[index].r_refcount !=
	       rb->rf_records.rl_recs[index+1].r_refcount);

	le32_add_cpu(&rb->rf_records.rl_recs[index].r_clusters,
		     le32_to_cpu(rb->rf_records.rl_recs[index+1].r_clusters));

	if (index < le16_to_cpu(rb->rf_records.rl_used) - 2)
		memmove(&rb->rf_records.rl_recs[index + 1],
			&rb->rf_records.rl_recs[index + 2],
			sizeof(struct ocfs2_refcount_rec) *
			(le16_to_cpu(rb->rf_records.rl_used) - index - 2));

	memset(&rb->rf_records.rl_recs[le16_to_cpu(rb->rf_records.rl_used) - 1],
	       0, sizeof(struct ocfs2_refcount_rec));
	le16_add_cpu(&rb->rf_records.rl_used, -1);
}

/*
 * Merge the refcount rec if we are contiguous with the adjacent recs.
 */
static void ocfs2_refcount_rec_merge(struct ocfs2_refcount_block *rb,
				     int index)
{
	enum ocfs2_ref_rec_contig contig =
				ocfs2_refcount_rec_contig(rb, index);

	if (contig == REF_CONTIG_NONE)
		return;

	if (contig == REF_CONTIG_LEFT || contig == REF_CONTIG_LEFTRIGHT) {
		BUG_ON(index == 0);
		index--;
	}

	ocfs2_rotate_refcount_rec_left(rb, index);

	if (contig == REF_CONTIG_LEFTRIGHT)
		ocfs2_rotate_refcount_rec_left(rb, index);
}

1237 1238 1239 1240
/*
 * Change the refcount indexed by "index" in ref_bh.
 * If refcount reaches 0, remove it.
 */
1241 1242 1243
static int ocfs2_change_refcount_rec(handle_t *handle,
				     struct ocfs2_caching_info *ci,
				     struct buffer_head *ref_leaf_bh,
1244
				     int index, int merge, int change)
1245 1246 1247 1248
{
	int ret;
	struct ocfs2_refcount_block *rb =
			(struct ocfs2_refcount_block *)ref_leaf_bh->b_data;
1249 1250
	struct ocfs2_refcount_list *rl = &rb->rf_records;
	struct ocfs2_refcount_rec *rec = &rl->rl_recs[index];
1251 1252 1253 1254 1255 1256 1257 1258

	ret = ocfs2_journal_access_rb(handle, ci, ref_leaf_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

Tao Ma's avatar
Tao Ma committed
1259 1260 1261
	trace_ocfs2_change_refcount_rec(
		(unsigned long long)ocfs2_metadata_cache_owner(ci),
		index, le32_to_cpu(rec->r_refcount), change);
1262 1263
	le32_add_cpu(&rec->r_refcount, change);

1264 1265 1266 1267 1268 1269 1270 1271 1272 1273
	if (!rec->r_refcount) {
		if (index != le16_to_cpu(rl->rl_used) - 1) {
			memmove(rec, rec + 1,
				(le16_to_cpu(rl->rl_used) - index - 1) *
				sizeof(struct ocfs2_refcount_rec));
			memset(&rl->rl_recs[le16_to_cpu(rl->rl_used) - 1],
			       0, sizeof(struct ocfs2_refcount_rec));
		}

		le16_add_cpu(&rl->rl_used, -1);
1274
	} else if (merge)
1275
		ocfs2_refcount_rec_merge(rb, index);
1276

1277
	ocfs2_journal_dirty(handle, ref_leaf_bh);
1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290
out:
	return ret;
}

static int ocfs2_expand_inline_ref_root(handle_t *handle,
					struct ocfs2_caching_info *ci,
					struct buffer_head *ref_root_bh,
					struct buffer_head **ref_leaf_bh,
					struct ocfs2_alloc_context *meta_ac)
{
	int ret;
	u16 suballoc_bit_start;
	u32 num_got;
1291
	u64 suballoc_loc, blkno;
1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304
	struct super_block *sb = ocfs2_metadata_cache_get_super(ci);
	struct buffer_head *new_bh = NULL;
	struct ocfs2_refcount_block *new_rb;
	struct ocfs2_refcount_block *root_rb =
			(struct ocfs2_refcount_block *)ref_root_bh->b_data;

	ret = ocfs2_journal_access_rb(handle, ci, ref_root_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

1305
	ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc,
1306 1307 1308 1309 1310 1311 1312 1313 1314
				   &suballoc_bit_start, &num_got,
				   &blkno);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	new_bh = sb_getblk(sb, blkno);
	if (new_bh == NULL) {
1315
		ret = -ENOMEM;
1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335
		mlog_errno(ret);
		goto out;
	}
	ocfs2_set_new_buffer_uptodate(ci, new_bh);

	ret = ocfs2_journal_access_rb(handle, ci, new_bh,
				      OCFS2_JOURNAL_ACCESS_CREATE);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	/*
	 * Initialize ocfs2_refcount_block.
	 * It should contain the same information as the old root.
	 * so just memcpy it and change the corresponding field.
	 */
	memcpy(new_bh->b_data, ref_root_bh->b_data, sb->s_blocksize);

	new_rb = (struct ocfs2_refcount_block *)new_bh->b_data;
1336
	new_rb->rf_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot);
1337
	new_rb->rf_suballoc_loc = cpu_to_le64(suballoc_loc);
1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356
	new_rb->rf_suballoc_bit = cpu_to_le16(suballoc_bit_start);
	new_rb->rf_blkno = cpu_to_le64(blkno);
	new_rb->rf_cpos = cpu_to_le32(0);
	new_rb->rf_parent = cpu_to_le64(ref_root_bh->b_blocknr);
	new_rb->rf_flags = cpu_to_le32(OCFS2_REFCOUNT_LEAF_FL);
	ocfs2_journal_dirty(handle, new_bh);

	/* Now change the root. */
	memset(&root_rb->rf_list, 0, sb->s_blocksize -
	       offsetof(struct ocfs2_refcount_block, rf_list));
	root_rb->rf_list.l_count = cpu_to_le16(ocfs2_extent_recs_per_rb(sb));
	root_rb->rf_clusters = cpu_to_le32(1);
	root_rb->rf_list.l_next_free_rec = cpu_to_le16(1);
	root_rb->rf_list.l_recs[0].e_blkno = cpu_to_le64(blkno);
	root_rb->rf_list.l_recs[0].e_leaf_clusters = cpu_to_le16(1);
	root_rb->rf_flags = cpu_to_le32(OCFS2_REFCOUNT_TREE_FL);

	ocfs2_journal_dirty(handle, ref_root_bh);

Tao Ma's avatar
Tao Ma committed
1357 1358
	trace_ocfs2_expand_inline_ref_root((unsigned long long)blkno,
		le16_to_cpu(new_rb->rf_records.rl_used));
1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404

	*ref_leaf_bh = new_bh;
	new_bh = NULL;
out:
	brelse(new_bh);
	return ret;
}

static int ocfs2_refcount_rec_no_intersect(struct ocfs2_refcount_rec *prev,
					   struct ocfs2_refcount_rec *next)
{
	if (ocfs2_get_ref_rec_low_cpos(prev) + le32_to_cpu(prev->r_clusters) <=
		ocfs2_get_ref_rec_low_cpos(next))
		return 1;

	return 0;
}

static int cmp_refcount_rec_by_low_cpos(const void *a, const void *b)
{
	const struct ocfs2_refcount_rec *l = a, *r = b;
	u32 l_cpos = ocfs2_get_ref_rec_low_cpos(l);
	u32 r_cpos = ocfs2_get_ref_rec_low_cpos(r);

	if (l_cpos > r_cpos)
		return 1;
	if (l_cpos < r_cpos)
		return -1;
	return 0;
}

static int cmp_refcount_rec_by_cpos(const void *a, const void *b)
{
	const struct ocfs2_refcount_rec *l = a, *r = b;
	u64 l_cpos = le64_to_cpu(l->r_cpos);
	u64 r_cpos = le64_to_cpu(r->r_cpos);

	if (l_cpos > r_cpos)
		return 1;
	if (l_cpos < r_cpos)
		return -1;
	return 0;
}

static void swap_refcount_rec(void *a, void *b, int size)
{
1405
	struct ocfs2_refcount_rec *l = a, *r = b;
1406

1407
	swap(*l, *r);
1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466
}

/*
 * The refcount cpos are ordered by their 64bit cpos,
 * But we will use the low 32 bit to be the e_cpos in the b-tree.
 * So we need to make sure that this pos isn't intersected with others.
 *
 * Note: The refcount block is already sorted by their low 32 bit cpos,
 *       So just try the middle pos first, and we will exit when we find
 *       the good position.
 */
static int ocfs2_find_refcount_split_pos(struct ocfs2_refcount_list *rl,
					 u32 *split_pos, int *split_index)
{
	int num_used = le16_to_cpu(rl->rl_used);
	int delta, middle = num_used / 2;

	for (delta = 0; delta < middle; delta++) {
		/* Let's check delta earlier than middle */
		if (ocfs2_refcount_rec_no_intersect(
					&rl->rl_recs[middle - delta - 1],
					&rl->rl_recs[middle - delta])) {
			*split_index = middle - delta;
			break;
		}

		/* For even counts, don't walk off the end */
		if ((middle + delta + 1) == num_used)
			continue;

		/* Now try delta past middle */
		if (ocfs2_refcount_rec_no_intersect(
					&rl->rl_recs[middle + delta],
					&rl->rl_recs[middle + delta + 1])) {
			*split_index = middle + delta + 1;
			break;
		}
	}

	if (delta >= middle)
		return -ENOSPC;

	*split_pos = ocfs2_get_ref_rec_low_cpos(&rl->rl_recs[*split_index]);
	return 0;
}

static int ocfs2_divide_leaf_refcount_block(struct buffer_head *ref_leaf_bh,
					    struct buffer_head *new_bh,
					    u32 *split_cpos)
{
	int split_index = 0, num_moved, ret;
	u32 cpos = 0;
	struct ocfs2_refcount_block *rb =
			(struct ocfs2_refcount_block *)ref_leaf_bh->b_data;
	struct ocfs2_refcount_list *rl = &rb->rf_records;
	struct ocfs2_refcount_block *new_rb =
			(struct ocfs2_refcount_block *)new_bh->b_data;
	struct ocfs2_refcount_list *new_rl = &new_rb->rf_records;

Tao Ma's avatar
Tao Ma committed
1467 1468
	trace_ocfs2_divide_leaf_refcount_block(
		(unsigned long long)ref_leaf_bh->b_blocknr,
1469
		le16_to_cpu(rl->rl_count), le16_to_cpu(rl->rl_used));
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

	/*
	 * XXX: Improvement later.
	 * If we know all the high 32 bit cpos is the same, no need to sort.
	 *
	 * In order to make the whole process safe, we do:
	 * 1. sort the entries by their low 32 bit cpos first so that we can
	 *    find the split cpos easily.
	 * 2. call ocfs2_insert_extent to insert the new refcount block.
	 * 3. move the refcount rec to the new block.
	 * 4. sort the entries by their 64 bit cpos.
	 * 5. dirty the new_rb and rb.
	 */
	sort(&rl->rl_recs, le16_to_cpu(rl->rl_used),
	     sizeof(struct ocfs2_refcount_rec),
	     cmp_refcount_rec_by_low_cpos, swap_refcount_rec);

	ret = ocfs2_find_refcount_split_pos(rl, &cpos, &split_index);
	if (ret) {
		mlog_errno(ret);
		return ret;
	}

	new_rb->rf_cpos = cpu_to_le32(cpos);

	/* move refcount records starting from split_index to the new block. */
	num_moved = le16_to_cpu(rl->rl_used) - split_index;
	memcpy(new_rl->rl_recs, &rl->rl_recs[split_index],
	       num_moved * sizeof(struct ocfs2_refcount_rec));

	/*ok, remove the entries we just moved over to the other block. */
	memset(&rl->rl_recs[split_index], 0,
	       num_moved * sizeof(struct ocfs2_refcount_rec));

	/* change old and new rl_used accordingly. */
	le16_add_cpu(&rl->rl_used, -num_moved);
Tao Ma's avatar
Tao Ma committed
1506
	new_rl->rl_used = cpu_to_le16(num_moved);
1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528

	sort(&rl->rl_recs, le16_to_cpu(rl->rl_used),
	     sizeof(struct ocfs2_refcount_rec),
	     cmp_refcount_rec_by_cpos, swap_refcount_rec);

	sort(&new_rl->rl_recs, le16_to_cpu(new_rl->rl_used),
	     sizeof(struct ocfs2_refcount_rec),
	     cmp_refcount_rec_by_cpos, swap_refcount_rec);

	*split_cpos = cpos;
	return 0;
}

static int ocfs2_new_leaf_refcount_block(handle_t *handle,
					 struct ocfs2_caching_info *ci,
					 struct buffer_head *ref_root_bh,
					 struct buffer_head *ref_leaf_bh,
					 struct ocfs2_alloc_context *meta_ac)
{
	int ret;
	u16 suballoc_bit_start;
	u32 num_got, new_cpos;
1529
	u64 suballoc_loc, blkno;
1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552
	struct super_block *sb = ocfs2_metadata_cache_get_super(ci);
	struct ocfs2_refcount_block *root_rb =
			(struct ocfs2_refcount_block *)ref_root_bh->b_data;
	struct buffer_head *new_bh = NULL;
	struct ocfs2_refcount_block *new_rb;
	struct ocfs2_extent_tree ref_et;

	BUG_ON(!(le32_to_cpu(root_rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL));

	ret = ocfs2_journal_access_rb(handle, ci, ref_root_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	ret = ocfs2_journal_access_rb(handle, ci, ref_leaf_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

1553
	ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc,
1554 1555 1556 1557 1558 1559 1560 1561 1562
				   &suballoc_bit_start, &num_got,
				   &blkno);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	new_bh = sb_getblk(sb, blkno);
	if (new_bh == NULL) {
1563
		ret = -ENOMEM;
1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579
		mlog_errno(ret);
		goto out;
	}
	ocfs2_set_new_buffer_uptodate(ci, new_bh);

	ret = ocfs2_journal_access_rb(handle, ci, new_bh,
				      OCFS2_JOURNAL_ACCESS_CREATE);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	/* Initialize ocfs2_refcount_block. */
	new_rb = (struct ocfs2_refcount_block *)new_bh->b_data;
	memset(new_rb, 0, sb->s_blocksize);
	strcpy((void *)new_rb, OCFS2_REFCOUNT_BLOCK_SIGNATURE);
1580
	new_rb->rf_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot);
1581
	new_rb->rf_suballoc_loc = cpu_to_le64(suballoc_loc);
1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601
	new_rb->rf_suballoc_bit = cpu_to_le16(suballoc_bit_start);
	new_rb->rf_fs_generation = cpu_to_le32(OCFS2_SB(sb)->fs_generation);
	new_rb->rf_blkno = cpu_to_le64(blkno);
	new_rb->rf_parent = cpu_to_le64(ref_root_bh->b_blocknr);
	new_rb->rf_flags = cpu_to_le32(OCFS2_REFCOUNT_LEAF_FL);
	new_rb->rf_records.rl_count =
				cpu_to_le16(ocfs2_refcount_recs_per_rb(sb));
	new_rb->rf_generation = root_rb->rf_generation;

	ret = ocfs2_divide_leaf_refcount_block(ref_leaf_bh, new_bh, &new_cpos);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	ocfs2_journal_dirty(handle, ref_leaf_bh);
	ocfs2_journal_dirty(handle, new_bh);

	ocfs2_init_refcount_extent_tree(&ref_et, ci, ref_root_bh);

Tao Ma's avatar
Tao Ma committed
1602 1603
	trace_ocfs2_new_leaf_refcount_block(
			(unsigned long long)new_bh->b_blocknr, new_cpos);
1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699

	/* Insert the new leaf block with the specific offset cpos. */
	ret = ocfs2_insert_extent(handle, &ref_et, new_cpos, new_bh->b_blocknr,
				  1, 0, meta_ac);
	if (ret)
		mlog_errno(ret);

out:
	brelse(new_bh);
	return ret;
}

static int ocfs2_expand_refcount_tree(handle_t *handle,
				      struct ocfs2_caching_info *ci,
				      struct buffer_head *ref_root_bh,
				      struct buffer_head *ref_leaf_bh,
				      struct ocfs2_alloc_context *meta_ac)
{
	int ret;
	struct buffer_head *expand_bh = NULL;

	if (ref_root_bh == ref_leaf_bh) {
		/*
		 * the old root bh hasn't been expanded to a b-tree,
		 * so expand it first.
		 */
		ret = ocfs2_expand_inline_ref_root(handle, ci, ref_root_bh,
						   &expand_bh, meta_ac);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}
	} else {
		expand_bh = ref_leaf_bh;
		get_bh(expand_bh);
	}


	/* Now add a new refcount block into the tree.*/
	ret = ocfs2_new_leaf_refcount_block(handle, ci, ref_root_bh,
					    expand_bh, meta_ac);
	if (ret)
		mlog_errno(ret);
out:
	brelse(expand_bh);
	return ret;
}

/*
 * Adjust the extent rec in b-tree representing ref_leaf_bh.
 *
 * Only called when we have inserted a new refcount rec at index 0
 * which means ocfs2_extent_rec.e_cpos may need some change.
 */
static int ocfs2_adjust_refcount_rec(handle_t *handle,
				     struct ocfs2_caching_info *ci,
				     struct buffer_head *ref_root_bh,
				     struct buffer_head *ref_leaf_bh,
				     struct ocfs2_refcount_rec *rec)
{
	int ret = 0, i;
	u32 new_cpos, old_cpos;
	struct ocfs2_path *path = NULL;
	struct ocfs2_extent_tree et;
	struct ocfs2_refcount_block *rb =
		(struct ocfs2_refcount_block *)ref_root_bh->b_data;
	struct ocfs2_extent_list *el;

	if (!(le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL))
		goto out;

	rb = (struct ocfs2_refcount_block *)ref_leaf_bh->b_data;
	old_cpos = le32_to_cpu(rb->rf_cpos);
	new_cpos = le64_to_cpu(rec->r_cpos) & OCFS2_32BIT_POS_MASK;
	if (old_cpos <= new_cpos)
		goto out;

	ocfs2_init_refcount_extent_tree(&et, ci, ref_root_bh);

	path = ocfs2_new_path_from_et(&et);
	if (!path) {
		ret = -ENOMEM;
		mlog_errno(ret);
		goto out;
	}

	ret = ocfs2_find_path(ci, path, old_cpos);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	/*
	 * 2 more credits, one for the leaf refcount block, one for
	 * the extent block contains the extent rec.
	 */
1700
	ret = ocfs2_extend_trans(handle, 2);
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 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746
	if (ret < 0) {
		mlog_errno(ret);
		goto out;
	}

	ret = ocfs2_journal_access_rb(handle, ci, ref_leaf_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret < 0) {
		mlog_errno(ret);
		goto out;
	}

	ret = ocfs2_journal_access_eb(handle, ci, path_leaf_bh(path),
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret < 0) {
		mlog_errno(ret);
		goto out;
	}

	/* change the leaf extent block first. */
	el = path_leaf_el(path);

	for (i = 0; i < le16_to_cpu(el->l_next_free_rec); i++)
		if (le32_to_cpu(el->l_recs[i].e_cpos) == old_cpos)
			break;

	BUG_ON(i == le16_to_cpu(el->l_next_free_rec));

	el->l_recs[i].e_cpos = cpu_to_le32(new_cpos);

	/* change the r_cpos in the leaf block. */
	rb->rf_cpos = cpu_to_le32(new_cpos);

	ocfs2_journal_dirty(handle, path_leaf_bh(path));
	ocfs2_journal_dirty(handle, ref_leaf_bh);

out:
	ocfs2_free_path(path);
	return ret;
}

static int ocfs2_insert_refcount_rec(handle_t *handle,
				     struct ocfs2_caching_info *ci,
				     struct buffer_head *ref_root_bh,
				     struct buffer_head *ref_leaf_bh,
				     struct ocfs2_refcount_rec *rec,
1747
				     int index, int merge,
1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794
				     struct ocfs2_alloc_context *meta_ac)
{
	int ret;
	struct ocfs2_refcount_block *rb =
			(struct ocfs2_refcount_block *)ref_leaf_bh->b_data;
	struct ocfs2_refcount_list *rf_list = &rb->rf_records;
	struct buffer_head *new_bh = NULL;

	BUG_ON(le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL);

	if (rf_list->rl_used == rf_list->rl_count) {
		u64 cpos = le64_to_cpu(rec->r_cpos);
		u32 len = le32_to_cpu(rec->r_clusters);

		ret = ocfs2_expand_refcount_tree(handle, ci, ref_root_bh,
						 ref_leaf_bh, meta_ac);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}

		ret = ocfs2_get_refcount_rec(ci, ref_root_bh,
					     cpos, len, NULL, &index,
					     &new_bh);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}

		ref_leaf_bh = new_bh;
		rb = (struct ocfs2_refcount_block *)ref_leaf_bh->b_data;
		rf_list = &rb->rf_records;
	}

	ret = ocfs2_journal_access_rb(handle, ci, ref_leaf_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	if (index < le16_to_cpu(rf_list->rl_used))
		memmove(&rf_list->rl_recs[index + 1],
			&rf_list->rl_recs[index],
			(le16_to_cpu(rf_list->rl_used) - index) *
			 sizeof(struct ocfs2_refcount_rec));

Tao Ma's avatar
Tao Ma committed
1795 1796 1797 1798
	trace_ocfs2_insert_refcount_rec(
		(unsigned long long)ref_leaf_bh->b_blocknr, index,
		(unsigned long long)le64_to_cpu(rec->r_cpos),
		le32_to_cpu(rec->r_clusters), le32_to_cpu(rec->r_refcount));
1799 1800 1801 1802 1803

	rf_list->rl_recs[index] = *rec;

	le16_add_cpu(&rf_list->rl_used, 1);

1804 1805
	if (merge)
		ocfs2_refcount_rec_merge(rb, index);
1806

1807
	ocfs2_journal_dirty(handle, ref_leaf_bh);
1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834

	if (index == 0) {
		ret = ocfs2_adjust_refcount_rec(handle, ci,
						ref_root_bh,
						ref_leaf_bh, rec);
		if (ret)
			mlog_errno(ret);
	}
out:
	brelse(new_bh);
	return ret;
}

/*
 * Split the refcount_rec indexed by "index" in ref_leaf_bh.
 * This is much simple than our b-tree code.
 * split_rec is the new refcount rec we want to insert.
 * If split_rec->r_refcount > 0, we are changing the refcount(in case we
 * increase refcount or decrease a refcount to non-zero).
 * If split_rec->r_refcount == 0, we are punching a hole in current refcount
 * rec( in case we decrease a refcount to zero).
 */
static int ocfs2_split_refcount_rec(handle_t *handle,
				    struct ocfs2_caching_info *ci,
				    struct buffer_head *ref_root_bh,
				    struct buffer_head *ref_leaf_bh,
				    struct ocfs2_refcount_rec *split_rec,
1835
				    int index, int merge,
1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849
				    struct ocfs2_alloc_context *meta_ac,
				    struct ocfs2_cached_dealloc_ctxt *dealloc)
{
	int ret, recs_need;
	u32 len;
	struct ocfs2_refcount_block *rb =
			(struct ocfs2_refcount_block *)ref_leaf_bh->b_data;
	struct ocfs2_refcount_list *rf_list = &rb->rf_records;
	struct ocfs2_refcount_rec *orig_rec = &rf_list->rl_recs[index];
	struct ocfs2_refcount_rec *tail_rec = NULL;
	struct buffer_head *new_bh = NULL;

	BUG_ON(le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL);

Tao Ma's avatar
Tao Ma committed
1850 1851 1852 1853 1854 1855
	trace_ocfs2_split_refcount_rec(le64_to_cpu(orig_rec->r_cpos),
		le32_to_cpu(orig_rec->r_clusters),
		le32_to_cpu(orig_rec->r_refcount),
		le64_to_cpu(split_rec->r_cpos),
		le32_to_cpu(split_rec->r_clusters),
		le32_to_cpu(split_rec->r_refcount));
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

	/*
	 * If we just need to split the header or tail clusters,
	 * no more recs are needed, just split is OK.
	 * Otherwise we at least need one new recs.
	 */
	if (!split_rec->r_refcount &&
	    (split_rec->r_cpos == orig_rec->r_cpos ||
	     le64_to_cpu(split_rec->r_cpos) +
	     le32_to_cpu(split_rec->r_clusters) ==
	     le64_to_cpu(orig_rec->r_cpos) + le32_to_cpu(orig_rec->r_clusters)))
		recs_need = 0;
	else
		recs_need = 1;

	/*
	 * We need one more rec if we split in the middle and the new rec have
	 * some refcount in it.
	 */
	if (split_rec->r_refcount &&
	    (split_rec->r_cpos != orig_rec->r_cpos &&
	     le64_to_cpu(split_rec->r_cpos) +
	     le32_to_cpu(split_rec->r_clusters) !=
	     le64_to_cpu(orig_rec->r_cpos) + le32_to_cpu(orig_rec->r_clusters)))
		recs_need++;

	/* If the leaf block don't have enough record, expand it. */
Tao Ma's avatar
Tao Ma committed
1883 1884
	if (le16_to_cpu(rf_list->rl_used) + recs_need >
					 le16_to_cpu(rf_list->rl_count)) {
1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 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
		struct ocfs2_refcount_rec tmp_rec;
		u64 cpos = le64_to_cpu(orig_rec->r_cpos);
		len = le32_to_cpu(orig_rec->r_clusters);
		ret = ocfs2_expand_refcount_tree(handle, ci, ref_root_bh,
						 ref_leaf_bh, meta_ac);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}

		/*
		 * We have to re-get it since now cpos may be moved to
		 * another leaf block.
		 */
		ret = ocfs2_get_refcount_rec(ci, ref_root_bh,
					     cpos, len, &tmp_rec, &index,
					     &new_bh);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}

		ref_leaf_bh = new_bh;
		rb = (struct ocfs2_refcount_block *)ref_leaf_bh->b_data;
		rf_list = &rb->rf_records;
		orig_rec = &rf_list->rl_recs[index];
	}

	ret = ocfs2_journal_access_rb(handle, ci, ref_leaf_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	/*
	 * We have calculated out how many new records we need and store
	 * in recs_need, so spare enough space first by moving the records
	 * after "index" to the end.
	 */
	if (index != le16_to_cpu(rf_list->rl_used) - 1)
		memmove(&rf_list->rl_recs[index + 1 + recs_need],
			&rf_list->rl_recs[index + 1],
			(le16_to_cpu(rf_list->rl_used) - index - 1) *
			 sizeof(struct ocfs2_refcount_rec));

	len = (le64_to_cpu(orig_rec->r_cpos) +
	      le32_to_cpu(orig_rec->r_clusters)) -
	      (le64_to_cpu(split_rec->r_cpos) +
	      le32_to_cpu(split_rec->r_clusters));

	/*
	 * If we have "len", the we will split in the tail and move it
	 * to the end of the space we have just spared.
	 */
	if (len) {
		tail_rec = &rf_list->rl_recs[index + recs_need];

		memcpy(tail_rec, orig_rec, sizeof(struct ocfs2_refcount_rec));
		le64_add_cpu(&tail_rec->r_cpos,
			     le32_to_cpu(tail_rec->r_clusters) - len);
Tao Ma's avatar
Tao Ma committed
1946
		tail_rec->r_clusters = cpu_to_le32(len);
1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968
	}

	/*
	 * If the split pos isn't the same as the original one, we need to
	 * split in the head.
	 *
	 * Note: We have the chance that split_rec.r_refcount = 0,
	 * recs_need = 0 and len > 0, which means we just cut the head from
	 * the orig_rec and in that case we have done some modification in
	 * orig_rec above, so the check for r_cpos is faked.
	 */
	if (split_rec->r_cpos != orig_rec->r_cpos && tail_rec != orig_rec) {
		len = le64_to_cpu(split_rec->r_cpos) -
		      le64_to_cpu(orig_rec->r_cpos);
		orig_rec->r_clusters = cpu_to_le32(len);
		index++;
	}

	le16_add_cpu(&rf_list->rl_used, recs_need);

	if (split_rec->r_refcount) {
		rf_list->rl_recs[index] = *split_rec;
Tao Ma's avatar
Tao Ma committed
1969 1970 1971 1972 1973
		trace_ocfs2_split_refcount_rec_insert(
			(unsigned long long)ref_leaf_bh->b_blocknr, index,
			(unsigned long long)le64_to_cpu(split_rec->r_cpos),
			le32_to_cpu(split_rec->r_clusters),
			le32_to_cpu(split_rec->r_refcount));
1974

1975 1976
		if (merge)
			ocfs2_refcount_rec_merge(rb, index);
1977 1978
	}

1979
	ocfs2_journal_dirty(handle, ref_leaf_bh);
1980 1981 1982 1983 1984 1985

out:
	brelse(new_bh);
	return ret;
}

1986 1987 1988 1989 1990 1991
static int __ocfs2_increase_refcount(handle_t *handle,
				     struct ocfs2_caching_info *ci,
				     struct buffer_head *ref_root_bh,
				     u64 cpos, u32 len, int merge,
				     struct ocfs2_alloc_context *meta_ac,
				     struct ocfs2_cached_dealloc_ctxt *dealloc)
1992 1993 1994 1995 1996 1997
{
	int ret = 0, index;
	struct buffer_head *ref_leaf_bh = NULL;
	struct ocfs2_refcount_rec rec;
	unsigned int set_len = 0;

Tao Ma's avatar
Tao Ma committed
1998
	trace_ocfs2_increase_refcount_begin(
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
	     (unsigned long long)ocfs2_metadata_cache_owner(ci),
	     (unsigned long long)cpos, len);

	while (len) {
		ret = ocfs2_get_refcount_rec(ci, ref_root_bh,
					     cpos, len, &rec, &index,
					     &ref_leaf_bh);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}

		set_len = le32_to_cpu(rec.r_clusters);

		/*
		 * Here we may meet with 3 situations:
		 *
		 * 1. If we find an already existing record, and the length
		 *    is the same, cool, we just need to increase the r_refcount
		 *    and it is OK.
		 * 2. If we find a hole, just insert it with r_refcount = 1.
		 * 3. If we are in the middle of one extent record, split
		 *    it.
		 */
		if (rec.r_refcount && le64_to_cpu(rec.r_cpos) == cpos &&
		    set_len <= len) {
Tao Ma's avatar
Tao Ma committed
2025 2026 2027
			trace_ocfs2_increase_refcount_change(
				(unsigned long long)cpos, set_len,
				le32_to_cpu(rec.r_refcount));
2028
			ret = ocfs2_change_refcount_rec(handle, ci,
2029 2030
							ref_leaf_bh, index,
							merge, 1);
2031 2032 2033 2034 2035 2036 2037
			if (ret) {
				mlog_errno(ret);
				goto out;
			}
		} else if (!rec.r_refcount) {
			rec.r_refcount = cpu_to_le32(1);

Tao Ma's avatar
Tao Ma committed
2038
			trace_ocfs2_increase_refcount_insert(
2039 2040 2041 2042
			     (unsigned long long)le64_to_cpu(rec.r_cpos),
			     set_len);
			ret = ocfs2_insert_refcount_rec(handle, ci, ref_root_bh,
							ref_leaf_bh,
2043 2044
							&rec, index,
							merge, meta_ac);
2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055
			if (ret) {
				mlog_errno(ret);
				goto out;
			}
		} else  {
			set_len = min((u64)(cpos + len),
				      le64_to_cpu(rec.r_cpos) + set_len) - cpos;
			rec.r_cpos = cpu_to_le64(cpos);
			rec.r_clusters = cpu_to_le32(set_len);
			le32_add_cpu(&rec.r_refcount, 1);

Tao Ma's avatar
Tao Ma committed
2056
			trace_ocfs2_increase_refcount_split(
2057 2058 2059 2060
			     (unsigned long long)le64_to_cpu(rec.r_cpos),
			     set_len, le32_to_cpu(rec.r_refcount));
			ret = ocfs2_split_refcount_rec(handle, ci,
						       ref_root_bh, ref_leaf_bh,
2061
						       &rec, index, merge,
2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078
						       meta_ac, dealloc);
			if (ret) {
				mlog_errno(ret);
				goto out;
			}
		}

		cpos += set_len;
		len -= set_len;
		brelse(ref_leaf_bh);
		ref_leaf_bh = NULL;
	}

out:
	brelse(ref_leaf_bh);
	return ret;
}
2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094

static int ocfs2_remove_refcount_extent(handle_t *handle,
				struct ocfs2_caching_info *ci,
				struct buffer_head *ref_root_bh,
				struct buffer_head *ref_leaf_bh,
				struct ocfs2_alloc_context *meta_ac,
				struct ocfs2_cached_dealloc_ctxt *dealloc)
{
	int ret;
	struct super_block *sb = ocfs2_metadata_cache_get_super(ci);
	struct ocfs2_refcount_block *rb =
			(struct ocfs2_refcount_block *)ref_leaf_bh->b_data;
	struct ocfs2_extent_tree et;

	BUG_ON(rb->rf_records.rl_used);

Tao Ma's avatar
Tao Ma committed
2095 2096 2097 2098 2099
	trace_ocfs2_remove_refcount_extent(
		(unsigned long long)ocfs2_metadata_cache_owner(ci),
		(unsigned long long)ref_leaf_bh->b_blocknr,
		le32_to_cpu(rb->rf_cpos));

2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115
	ocfs2_init_refcount_extent_tree(&et, ci, ref_root_bh);
	ret = ocfs2_remove_extent(handle, &et, le32_to_cpu(rb->rf_cpos),
				  1, meta_ac, dealloc);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	ocfs2_remove_from_cache(ci, ref_leaf_bh);

	/*
	 * add the freed block to the dealloc so that it will be freed
	 * when we run dealloc.
	 */
	ret = ocfs2_cache_block_dealloc(dealloc, EXTENT_ALLOC_SYSTEM_INODE,
					le16_to_cpu(rb->rf_suballoc_slot),
2116
					le64_to_cpu(rb->rf_suballoc_loc),
2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141
					le64_to_cpu(rb->rf_blkno),
					le16_to_cpu(rb->rf_suballoc_bit));
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	ret = ocfs2_journal_access_rb(handle, ci, ref_root_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	rb = (struct ocfs2_refcount_block *)ref_root_bh->b_data;

	le32_add_cpu(&rb->rf_clusters, -1);

	/*
	 * check whether we need to restore the root refcount block if
	 * there is no leaf extent block at atll.
	 */
	if (!rb->rf_list.l_next_free_rec) {
		BUG_ON(rb->rf_clusters);

Tao Ma's avatar
Tao Ma committed
2142
		trace_ocfs2_restore_refcount_block(
2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159
		     (unsigned long long)ref_root_bh->b_blocknr);

		rb->rf_flags = 0;
		rb->rf_parent = 0;
		rb->rf_cpos = 0;
		memset(&rb->rf_records, 0, sb->s_blocksize -
		       offsetof(struct ocfs2_refcount_block, rf_records));
		rb->rf_records.rl_count =
				cpu_to_le16(ocfs2_refcount_recs_per_rb(sb));
	}

	ocfs2_journal_dirty(handle, ref_root_bh);

out:
	return ret;
}

2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171
int ocfs2_increase_refcount(handle_t *handle,
			    struct ocfs2_caching_info *ci,
			    struct buffer_head *ref_root_bh,
			    u64 cpos, u32 len,
			    struct ocfs2_alloc_context *meta_ac,
			    struct ocfs2_cached_dealloc_ctxt *dealloc)
{
	return __ocfs2_increase_refcount(handle, ci, ref_root_bh,
					 cpos, len, 1,
					 meta_ac, dealloc);
}

2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188
static int ocfs2_decrease_refcount_rec(handle_t *handle,
				struct ocfs2_caching_info *ci,
				struct buffer_head *ref_root_bh,
				struct buffer_head *ref_leaf_bh,
				int index, u64 cpos, unsigned int len,
				struct ocfs2_alloc_context *meta_ac,
				struct ocfs2_cached_dealloc_ctxt *dealloc)
{
	int ret;
	struct ocfs2_refcount_block *rb =
			(struct ocfs2_refcount_block *)ref_leaf_bh->b_data;
	struct ocfs2_refcount_rec *rec = &rb->rf_records.rl_recs[index];

	BUG_ON(cpos < le64_to_cpu(rec->r_cpos));
	BUG_ON(cpos + len >
	       le64_to_cpu(rec->r_cpos) + le32_to_cpu(rec->r_clusters));

Tao Ma's avatar
Tao Ma committed
2189 2190 2191 2192
	trace_ocfs2_decrease_refcount_rec(
		(unsigned long long)ocfs2_metadata_cache_owner(ci),
		(unsigned long long)cpos, len);

2193 2194 2195
	if (cpos == le64_to_cpu(rec->r_cpos) &&
	    len == le32_to_cpu(rec->r_clusters))
		ret = ocfs2_change_refcount_rec(handle, ci,