suballoc.c 55.3 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37
/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * suballoc.c
 *
 * metadata alloc and free
 * Inspired by ext3 block groups.
 *
 * Copyright (C) 2002, 2004 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 as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

#include <linux/fs.h>
#include <linux/types.h>
#include <linux/slab.h>
#include <linux/highmem.h>

#define MLOG_MASK_PREFIX ML_DISK_ALLOC
#include <cluster/masklog.h>

#include "ocfs2.h"

#include "alloc.h"
J
Joel Becker 已提交
38
#include "blockcheck.h"
39 40 41 42 43 44 45 46 47 48 49
#include "dlmglue.h"
#include "inode.h"
#include "journal.h"
#include "localalloc.h"
#include "suballoc.h"
#include "super.h"
#include "sysfile.h"
#include "uptodate.h"

#include "buffer_head_io.h"

50 51 52
#define NOT_ALLOC_NEW_GROUP		0
#define ALLOC_NEW_GROUP			1

53 54
#define OCFS2_MAX_INODES_TO_STEAL	1024

55 56 57
static inline void ocfs2_debug_bg(struct ocfs2_group_desc *bg);
static inline void ocfs2_debug_suballoc_inode(struct ocfs2_dinode *fe);
static inline u16 ocfs2_find_victim_chain(struct ocfs2_chain_list *cl);
58
static int ocfs2_block_group_fill(handle_t *handle,
59 60 61 62 63 64 65
				  struct inode *alloc_inode,
				  struct buffer_head *bg_bh,
				  u64 group_blkno,
				  u16 my_chain,
				  struct ocfs2_chain_list *cl);
static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
				   struct inode *alloc_inode,
66 67
				   struct buffer_head *bh,
				   u64 max_block);
68 69 70 71

static int ocfs2_cluster_group_search(struct inode *inode,
				      struct buffer_head *group_bh,
				      u32 bits_wanted, u32 min_bits,
72
				      u64 max_block,
73 74 75 76
				      u16 *bit_off, u16 *bits_found);
static int ocfs2_block_group_search(struct inode *inode,
				    struct buffer_head *group_bh,
				    u32 bits_wanted, u32 min_bits,
77
				    u64 max_block,
78 79 80
				    u16 *bit_off, u16 *bits_found);
static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
				     struct ocfs2_alloc_context *ac,
81
				     handle_t *handle,
82 83 84 85 86 87 88
				     u32 bits_wanted,
				     u32 min_bits,
				     u16 *bit_off,
				     unsigned int *num_bits,
				     u64 *bg_blkno);
static int ocfs2_test_bg_bit_allocatable(struct buffer_head *bg_bh,
					 int nr);
89
static inline int ocfs2_block_group_set_bits(handle_t *handle,
90 91 92 93 94
					     struct inode *alloc_inode,
					     struct ocfs2_group_desc *bg,
					     struct buffer_head *group_bh,
					     unsigned int bit_off,
					     unsigned int num_bits);
95
static inline int ocfs2_block_group_clear_bits(handle_t *handle,
96 97 98 99 100 101
					       struct inode *alloc_inode,
					       struct ocfs2_group_desc *bg,
					       struct buffer_head *group_bh,
					       unsigned int bit_off,
					       unsigned int num_bits);

102
static int ocfs2_relink_block_group(handle_t *handle,
103 104 105 106 107 108 109 110 111 112 113 114 115 116
				    struct inode *alloc_inode,
				    struct buffer_head *fe_bh,
				    struct buffer_head *bg_bh,
				    struct buffer_head *prev_bg_bh,
				    u16 chain);
static inline int ocfs2_block_group_reasonably_empty(struct ocfs2_group_desc *bg,
						     u32 wanted);
static inline u32 ocfs2_desc_bitmap_to_cluster_off(struct inode *inode,
						   u64 bg_blkno,
						   u16 bg_bit_off);
static inline void ocfs2_block_to_cluster_group(struct inode *inode,
						u64 data_blkno,
						u64 *bg_blkno,
						u16 *bg_bit_off);
117 118 119
static int ocfs2_reserve_clusters_with_limit(struct ocfs2_super *osb,
					     u32 bits_wanted, u64 max_block,
					     struct ocfs2_alloc_context **ac);
120

121
void ocfs2_free_ac_resource(struct ocfs2_alloc_context *ac)
122
{
123 124 125 126
	struct inode *inode = ac->ac_inode;

	if (inode) {
		if (ac->ac_which != OCFS2_AC_USE_LOCAL)
M
Mark Fasheh 已提交
127
			ocfs2_inode_unlock(inode, 1);
128 129 130 131

		mutex_unlock(&inode->i_mutex);

		iput(inode);
132
		ac->ac_inode = NULL;
133
	}
134 135
	brelse(ac->ac_bh);
	ac->ac_bh = NULL;
136 137 138 139 140
}

void ocfs2_free_alloc_context(struct ocfs2_alloc_context *ac)
{
	ocfs2_free_ac_resource(ac);
141 142 143 144 145 146 147 148
	kfree(ac);
}

static u32 ocfs2_bits_per_group(struct ocfs2_chain_list *cl)
{
	return (u32)le16_to_cpu(cl->cl_cpg) * (u32)le16_to_cpu(cl->cl_bpc);
}

149 150 151 152 153 154 155 156
#define do_error(fmt, ...)						\
	do{								\
		if (clean_error)					\
			mlog(ML_ERROR, fmt "\n", ##__VA_ARGS__);	\
		else							\
			ocfs2_error(sb, fmt, ##__VA_ARGS__);		\
	} while (0)

157 158 159 160 161 162
static int ocfs2_validate_gd_self(struct super_block *sb,
				  struct buffer_head *bh,
				  int clean_error)
{
	struct ocfs2_group_desc *gd = (struct ocfs2_group_desc *)bh->b_data;

163
	if (!OCFS2_IS_VALID_GROUP_DESC(gd)) {
164 165
		do_error("Group descriptor #%llu has bad signature %.*s",
			 (unsigned long long)bh->b_blocknr, 7,
166 167
			 gd->bg_signature);
		return -EINVAL;
168 169
	}

170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185
	if (le64_to_cpu(gd->bg_blkno) != bh->b_blocknr) {
		do_error("Group descriptor #%llu has an invalid bg_blkno "
			 "of %llu",
			 (unsigned long long)bh->b_blocknr,
			 (unsigned long long)le64_to_cpu(gd->bg_blkno));
		return -EINVAL;
	}

	if (le32_to_cpu(gd->bg_generation) != OCFS2_SB(sb)->fs_generation) {
		do_error("Group descriptor #%llu has an invalid "
			 "fs_generation of #%u",
			 (unsigned long long)bh->b_blocknr,
			 le32_to_cpu(gd->bg_generation));
		return -EINVAL;
	}

186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214
	if (le16_to_cpu(gd->bg_free_bits_count) > le16_to_cpu(gd->bg_bits)) {
		do_error("Group descriptor #%llu has bit count %u but "
			 "claims that %u are free",
			 (unsigned long long)bh->b_blocknr,
			 le16_to_cpu(gd->bg_bits),
			 le16_to_cpu(gd->bg_free_bits_count));
		return -EINVAL;
	}

	if (le16_to_cpu(gd->bg_bits) > (8 * le16_to_cpu(gd->bg_size))) {
		do_error("Group descriptor #%llu has bit count %u but "
			 "max bitmap bits of %u",
			 (unsigned long long)bh->b_blocknr,
			 le16_to_cpu(gd->bg_bits),
			 8 * le16_to_cpu(gd->bg_size));
		return -EINVAL;
	}

	return 0;
}

static int ocfs2_validate_gd_parent(struct super_block *sb,
				    struct ocfs2_dinode *di,
				    struct buffer_head *bh,
				    int clean_error)
{
	unsigned int max_bits;
	struct ocfs2_group_desc *gd = (struct ocfs2_group_desc *)bh->b_data;

215
	if (di->i_blkno != gd->bg_parent_dinode) {
216
		do_error("Group descriptor #%llu has bad parent "
217
			 "pointer (%llu, expected %llu)",
218
			 (unsigned long long)bh->b_blocknr,
219 220 221
			 (unsigned long long)le64_to_cpu(gd->bg_parent_dinode),
			 (unsigned long long)le64_to_cpu(di->i_blkno));
		return -EINVAL;
222 223 224 225
	}

	max_bits = le16_to_cpu(di->id2.i_chain.cl_cpg) * le16_to_cpu(di->id2.i_chain.cl_bpc);
	if (le16_to_cpu(gd->bg_bits) > max_bits) {
226 227
		do_error("Group descriptor #%llu has bit count of %u",
			 (unsigned long long)bh->b_blocknr,
228 229
			 le16_to_cpu(gd->bg_bits));
		return -EINVAL;
230 231 232 233
	}

	if (le16_to_cpu(gd->bg_chain) >=
	    le16_to_cpu(di->id2.i_chain.cl_next_free_rec)) {
234 235
		do_error("Group descriptor #%llu has bad chain %u",
			 (unsigned long long)bh->b_blocknr,
236 237
			 le16_to_cpu(gd->bg_chain));
		return -EINVAL;
238 239
	}

240 241
	return 0;
}
242

243
#undef do_error
244

245 246 247 248 249 250 251 252 253
/*
 * This version only prints errors.  It does not fail the filesystem, and
 * exists only for resize.
 */
int ocfs2_check_group_descriptor(struct super_block *sb,
				 struct ocfs2_dinode *di,
				 struct buffer_head *bh)
{
	int rc;
J
Joel Becker 已提交
254 255 256
	struct ocfs2_group_desc *gd = (struct ocfs2_group_desc *)bh->b_data;

	BUG_ON(!buffer_uptodate(bh));
257

J
Joel Becker 已提交
258 259 260 261 262 263
	/*
	 * 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, &gd->bg_check);
264 265 266 267 268
	if (rc) {
		mlog(ML_ERROR,
		     "Checksum failed for group descriptor %llu\n",
		     (unsigned long long)bh->b_blocknr);
	} else
J
Joel Becker 已提交
269
		rc = ocfs2_validate_gd_self(sb, bh, 1);
270 271 272 273 274 275 276 277 278
	if (!rc)
		rc = ocfs2_validate_gd_parent(sb, di, bh, 1);

	return rc;
}

static int ocfs2_validate_group_descriptor(struct super_block *sb,
					   struct buffer_head *bh)
{
J
Joel Becker 已提交
279 280 281
	int rc;
	struct ocfs2_group_desc *gd = (struct ocfs2_group_desc *)bh->b_data;

282 283 284
	mlog(0, "Validating group descriptor %llu\n",
	     (unsigned long long)bh->b_blocknr);

J
Joel Becker 已提交
285 286 287 288 289 290 291 292 293 294 295 296 297 298 299
	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, &gd->bg_check);
	if (rc)
		return rc;

	/*
	 * Errors after here are fatal.
	 */

300
	return ocfs2_validate_gd_self(sb, bh, 0);
301 302
}

303 304 305 306 307 308
int ocfs2_read_group_descriptor(struct inode *inode, struct ocfs2_dinode *di,
				u64 gd_blkno, struct buffer_head **bh)
{
	int rc;
	struct buffer_head *tmp = *bh;

309 310
	rc = ocfs2_read_block(inode, gd_blkno, &tmp,
			      ocfs2_validate_group_descriptor);
311 312 313
	if (rc)
		goto out;

314
	rc = ocfs2_validate_gd_parent(inode->i_sb, di, tmp, 0);
315 316 317 318 319 320 321 322 323 324 325 326 327
	if (rc) {
		brelse(tmp);
		goto out;
	}

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

out:
	return rc;
}

328
static int ocfs2_block_group_fill(handle_t *handle,
329 330 331 332 333 334 335 336 337 338 339 340 341
				  struct inode *alloc_inode,
				  struct buffer_head *bg_bh,
				  u64 group_blkno,
				  u16 my_chain,
				  struct ocfs2_chain_list *cl)
{
	int status = 0;
	struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data;
	struct super_block * sb = alloc_inode->i_sb;

	mlog_entry_void();

	if (((unsigned long long) bg_bh->b_blocknr) != group_blkno) {
342 343 344
		ocfs2_error(alloc_inode->i_sb, "group block (%llu) != "
			    "b_blocknr (%llu)",
			    (unsigned long long)group_blkno,
345 346 347 348 349
			    (unsigned long long) bg_bh->b_blocknr);
		status = -EIO;
		goto bail;
	}

350 351 352 353
	status = ocfs2_journal_access_gd(handle,
					 alloc_inode,
					 bg_bh,
					 OCFS2_JOURNAL_ACCESS_CREATE);
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
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	memset(bg, 0, sb->s_blocksize);
	strcpy(bg->bg_signature, OCFS2_GROUP_DESC_SIGNATURE);
	bg->bg_generation = cpu_to_le32(OCFS2_SB(sb)->fs_generation);
	bg->bg_size = cpu_to_le16(ocfs2_group_bitmap_size(sb));
	bg->bg_bits = cpu_to_le16(ocfs2_bits_per_group(cl));
	bg->bg_chain = cpu_to_le16(my_chain);
	bg->bg_next_group = cl->cl_recs[my_chain].c_blkno;
	bg->bg_parent_dinode = cpu_to_le64(OCFS2_I(alloc_inode)->ip_blkno);
	bg->bg_blkno = cpu_to_le64(group_blkno);
	/* set the 1st bit in the bitmap to account for the descriptor block */
	ocfs2_set_bit(0, (unsigned long *)bg->bg_bitmap);
	bg->bg_free_bits_count = cpu_to_le16(le16_to_cpu(bg->bg_bits) - 1);

	status = ocfs2_journal_dirty(handle, bg_bh);
	if (status < 0)
		mlog_errno(status);

	/* There is no need to zero out or otherwise initialize the
	 * other blocks in a group - All valid FS metadata in a block
	 * group stores the superblock fs_generation value at
	 * allocation time. */

bail:
	mlog_exit(status);
	return status;
}

static inline u16 ocfs2_find_smallest_chain(struct ocfs2_chain_list *cl)
{
	u16 curr, best;

	best = curr = 0;
	while (curr < le16_to_cpu(cl->cl_count)) {
		if (le32_to_cpu(cl->cl_recs[best].c_total) >
		    le32_to_cpu(cl->cl_recs[curr].c_total))
			best = curr;
		curr++;
	}
	return best;
}

/*
 * We expect the block group allocator to already be locked.
 */
static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
				   struct inode *alloc_inode,
405 406
				   struct buffer_head *bh,
				   u64 max_block)
407 408 409 410 411
{
	int status, credits;
	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) bh->b_data;
	struct ocfs2_chain_list *cl;
	struct ocfs2_alloc_context *ac = NULL;
412
	handle_t *handle = NULL;
413 414 415 416 417 418 419 420 421 422 423
	u32 bit_off, num_bits;
	u16 alloc_rec;
	u64 bg_blkno;
	struct buffer_head *bg_bh = NULL;
	struct ocfs2_group_desc *bg;

	BUG_ON(ocfs2_is_cluster_bitmap(alloc_inode));

	mlog_entry_void();

	cl = &fe->id2.i_chain;
424 425 426
	status = ocfs2_reserve_clusters_with_limit(osb,
						   le16_to_cpu(cl->cl_cpg),
						   max_block, &ac);
427 428 429 430 431 432 433 434
	if (status < 0) {
		if (status != -ENOSPC)
			mlog_errno(status);
		goto bail;
	}

	credits = ocfs2_calc_group_alloc_credits(osb->sb,
						 le16_to_cpu(cl->cl_cpg));
435
	handle = ocfs2_start_trans(osb, credits);
436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458
	if (IS_ERR(handle)) {
		status = PTR_ERR(handle);
		handle = NULL;
		mlog_errno(status);
		goto bail;
	}

	status = ocfs2_claim_clusters(osb,
				      handle,
				      ac,
				      le16_to_cpu(cl->cl_cpg),
				      &bit_off,
				      &num_bits);
	if (status < 0) {
		if (status != -ENOSPC)
			mlog_errno(status);
		goto bail;
	}

	alloc_rec = ocfs2_find_smallest_chain(cl);

	/* setup the group */
	bg_blkno = ocfs2_clusters_to_blocks(osb->sb, bit_off);
459 460
	mlog(0, "new descriptor, record %u, at block %llu\n",
	     alloc_rec, (unsigned long long)bg_blkno);
461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482

	bg_bh = sb_getblk(osb->sb, bg_blkno);
	if (!bg_bh) {
		status = -EIO;
		mlog_errno(status);
		goto bail;
	}
	ocfs2_set_new_buffer_uptodate(alloc_inode, bg_bh);

	status = ocfs2_block_group_fill(handle,
					alloc_inode,
					bg_bh,
					bg_blkno,
					alloc_rec,
					cl);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	bg = (struct ocfs2_group_desc *) bg_bh->b_data;

483 484
	status = ocfs2_journal_access_di(handle, alloc_inode,
					 bh, OCFS2_JOURNAL_ACCESS_WRITE);
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
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	le32_add_cpu(&cl->cl_recs[alloc_rec].c_free,
		     le16_to_cpu(bg->bg_free_bits_count));
	le32_add_cpu(&cl->cl_recs[alloc_rec].c_total, le16_to_cpu(bg->bg_bits));
	cl->cl_recs[alloc_rec].c_blkno  = cpu_to_le64(bg_blkno);
	if (le16_to_cpu(cl->cl_next_free_rec) < le16_to_cpu(cl->cl_count))
		le16_add_cpu(&cl->cl_next_free_rec, 1);

	le32_add_cpu(&fe->id1.bitmap1.i_used, le16_to_cpu(bg->bg_bits) -
					le16_to_cpu(bg->bg_free_bits_count));
	le32_add_cpu(&fe->id1.bitmap1.i_total, le16_to_cpu(bg->bg_bits));
	le32_add_cpu(&fe->i_clusters, le16_to_cpu(cl->cl_cpg));

	status = ocfs2_journal_dirty(handle, bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	spin_lock(&OCFS2_I(alloc_inode)->ip_lock);
	OCFS2_I(alloc_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
	fe->i_size = cpu_to_le64(ocfs2_clusters_to_bytes(alloc_inode->i_sb,
					     le32_to_cpu(fe->i_clusters)));
	spin_unlock(&OCFS2_I(alloc_inode)->ip_lock);
	i_size_write(alloc_inode, le64_to_cpu(fe->i_size));
514
	alloc_inode->i_blocks = ocfs2_inode_sector_count(alloc_inode);
515 516 517 518

	status = 0;
bail:
	if (handle)
519
		ocfs2_commit_trans(osb, handle);
520 521 522 523

	if (ac)
		ocfs2_free_alloc_context(ac);

524
	brelse(bg_bh);
525 526 527 528 529 530

	mlog_exit(status);
	return status;
}

static int ocfs2_reserve_suballoc_bits(struct ocfs2_super *osb,
531 532
				       struct ocfs2_alloc_context *ac,
				       int type,
533 534
				       u32 slot,
				       int alloc_new_group)
535 536 537
{
	int status;
	u32 bits_wanted = ac->ac_bits_wanted;
538
	struct inode *alloc_inode;
539 540 541 542 543 544
	struct buffer_head *bh = NULL;
	struct ocfs2_dinode *fe;
	u32 free_bits;

	mlog_entry_void();

545 546 547 548 549
	alloc_inode = ocfs2_get_system_file_inode(osb, type, slot);
	if (!alloc_inode) {
		mlog_errno(-EINVAL);
		return -EINVAL;
	}
550

551 552
	mutex_lock(&alloc_inode->i_mutex);

M
Mark Fasheh 已提交
553
	status = ocfs2_inode_lock(alloc_inode, &bh, 1);
554
	if (status < 0) {
555 556 557
		mutex_unlock(&alloc_inode->i_mutex);
		iput(alloc_inode);

558
		mlog_errno(status);
559
		return status;
560 561
	}

562
	ac->ac_inode = alloc_inode;
563
	ac->ac_alloc_slot = slot;
564

565
	fe = (struct ocfs2_dinode *) bh->b_data;
566 567 568 569 570

	/* The bh was validated by the inode read inside
	 * ocfs2_inode_lock().  Any corruption is a code bug. */
	BUG_ON(!OCFS2_IS_VALID_DINODE(fe));

571
	if (!(fe->i_flags & cpu_to_le32(OCFS2_CHAIN_FL))) {
572 573
		ocfs2_error(alloc_inode->i_sb, "Invalid chain allocator %llu",
			    (unsigned long long)le64_to_cpu(fe->i_blkno));
574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589
		status = -EIO;
		goto bail;
	}

	free_bits = le32_to_cpu(fe->id1.bitmap1.i_total) -
		le32_to_cpu(fe->id1.bitmap1.i_used);

	if (bits_wanted > free_bits) {
		/* cluster bitmap never grows */
		if (ocfs2_is_cluster_bitmap(alloc_inode)) {
			mlog(0, "Disk Full: wanted=%u, free_bits=%u\n",
			     bits_wanted, free_bits);
			status = -ENOSPC;
			goto bail;
		}

590 591 592 593 594 595 596 597
		if (alloc_new_group != ALLOC_NEW_GROUP) {
			mlog(0, "Alloc File %u Full: wanted=%u, free_bits=%u, "
			     "and we don't alloc a new group for it.\n",
			     slot, bits_wanted, free_bits);
			status = -ENOSPC;
			goto bail;
		}

598 599
		status = ocfs2_block_group_alloc(osb, alloc_inode, bh,
						 ac->ac_max_block);
600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615
		if (status < 0) {
			if (status != -ENOSPC)
				mlog_errno(status);
			goto bail;
		}
		atomic_inc(&osb->alloc_stats.bg_extends);

		/* You should never ask for this much metadata */
		BUG_ON(bits_wanted >
		       (le32_to_cpu(fe->id1.bitmap1.i_total)
			- le32_to_cpu(fe->id1.bitmap1.i_used)));
	}

	get_bh(bh);
	ac->ac_bh = bh;
bail:
616
	brelse(bh);
617 618 619 620 621

	mlog_exit(status);
	return status;
}

T
Tiger Yang 已提交
622 623 624
int ocfs2_reserve_new_metadata_blocks(struct ocfs2_super *osb,
				      int blocks,
				      struct ocfs2_alloc_context **ac)
625 626
{
	int status;
627
	u32 slot;
628

629
	*ac = kzalloc(sizeof(struct ocfs2_alloc_context), GFP_KERNEL);
630 631 632 633 634 635
	if (!(*ac)) {
		status = -ENOMEM;
		mlog_errno(status);
		goto bail;
	}

T
Tiger Yang 已提交
636
	(*ac)->ac_bits_wanted = blocks;
637
	(*ac)->ac_which = OCFS2_AC_USE_META;
638
	slot = osb->slot_num;
639 640
	(*ac)->ac_group_search = ocfs2_block_group_search;

641
	status = ocfs2_reserve_suballoc_bits(osb, (*ac),
642 643
					     EXTENT_ALLOC_SYSTEM_INODE,
					     slot, ALLOC_NEW_GROUP);
644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660
	if (status < 0) {
		if (status != -ENOSPC)
			mlog_errno(status);
		goto bail;
	}

	status = 0;
bail:
	if ((status < 0) && *ac) {
		ocfs2_free_alloc_context(*ac);
		*ac = NULL;
	}

	mlog_exit(status);
	return status;
}

T
Tiger Yang 已提交
661 662 663 664 665 666 667 668 669
int ocfs2_reserve_new_metadata(struct ocfs2_super *osb,
			       struct ocfs2_extent_list *root_el,
			       struct ocfs2_alloc_context **ac)
{
	return ocfs2_reserve_new_metadata_blocks(osb,
					ocfs2_extend_meta_needed(root_el),
					ac);
}

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
static int ocfs2_steal_inode_from_other_nodes(struct ocfs2_super *osb,
					      struct ocfs2_alloc_context *ac)
{
	int i, status = -ENOSPC;
	s16 slot = ocfs2_get_inode_steal_slot(osb);

	/* Start to steal inodes from the first slot after ours. */
	if (slot == OCFS2_INVALID_SLOT)
		slot = osb->slot_num + 1;

	for (i = 0; i < osb->max_slots; i++, slot++) {
		if (slot == osb->max_slots)
			slot = 0;

		if (slot == osb->slot_num)
			continue;

		status = ocfs2_reserve_suballoc_bits(osb, ac,
						     INODE_ALLOC_SYSTEM_INODE,
						     slot, NOT_ALLOC_NEW_GROUP);
		if (status >= 0) {
			ocfs2_set_inode_steal_slot(osb, slot);
			break;
		}

		ocfs2_free_ac_resource(ac);
	}

	return status;
}

701 702 703 704
int ocfs2_reserve_new_inode(struct ocfs2_super *osb,
			    struct ocfs2_alloc_context **ac)
{
	int status;
705
	s16 slot = ocfs2_get_inode_steal_slot(osb);
706

707
	*ac = kzalloc(sizeof(struct ocfs2_alloc_context), GFP_KERNEL);
708 709 710 711 712 713 714 715 716 717 718
	if (!(*ac)) {
		status = -ENOMEM;
		mlog_errno(status);
		goto bail;
	}

	(*ac)->ac_bits_wanted = 1;
	(*ac)->ac_which = OCFS2_AC_USE_INODE;

	(*ac)->ac_group_search = ocfs2_block_group_search;

719 720 721
	/*
	 * stat(2) can't handle i_ino > 32bits, so we tell the
	 * lower levels not to allocate us a block group past that
722
	 * limit.  The 'inode64' mount option avoids this behavior.
723
	 */
724 725
	if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64))
		(*ac)->ac_max_block = (u32)~0U;
726

727 728 729 730 731 732 733 734 735 736 737 738 739 740
	/*
	 * slot is set when we successfully steal inode from other nodes.
	 * It is reset in 3 places:
	 * 1. when we flush the truncate log
	 * 2. when we complete local alloc recovery.
	 * 3. when we successfully allocate from our own slot.
	 * After it is set, we will go on stealing inodes until we find the
	 * need to check our slots to see whether there is some space for us.
	 */
	if (slot != OCFS2_INVALID_SLOT &&
	    atomic_read(&osb->s_num_inodes_stolen) < OCFS2_MAX_INODES_TO_STEAL)
		goto inode_steal;

	atomic_set(&osb->s_num_inodes_stolen, 0);
741 742
	status = ocfs2_reserve_suballoc_bits(osb, *ac,
					     INODE_ALLOC_SYSTEM_INODE,
743
					     osb->slot_num, ALLOC_NEW_GROUP);
744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763
	if (status >= 0) {
		status = 0;

		/*
		 * Some inodes must be freed by us, so try to allocate
		 * from our own next time.
		 */
		if (slot != OCFS2_INVALID_SLOT)
			ocfs2_init_inode_steal_slot(osb);
		goto bail;
	} else if (status < 0 && status != -ENOSPC) {
		mlog_errno(status);
		goto bail;
	}

	ocfs2_free_ac_resource(*ac);

inode_steal:
	status = ocfs2_steal_inode_from_other_nodes(osb, *ac);
	atomic_inc(&osb->s_num_inodes_stolen);
764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790
	if (status < 0) {
		if (status != -ENOSPC)
			mlog_errno(status);
		goto bail;
	}

	status = 0;
bail:
	if ((status < 0) && *ac) {
		ocfs2_free_alloc_context(*ac);
		*ac = NULL;
	}

	mlog_exit(status);
	return status;
}

/* local alloc code has to do the same thing, so rather than do this
 * twice.. */
int ocfs2_reserve_cluster_bitmap_bits(struct ocfs2_super *osb,
				      struct ocfs2_alloc_context *ac)
{
	int status;

	ac->ac_which = OCFS2_AC_USE_MAIN;
	ac->ac_group_search = ocfs2_cluster_group_search;

791 792
	status = ocfs2_reserve_suballoc_bits(osb, ac,
					     GLOBAL_BITMAP_SYSTEM_INODE,
793 794
					     OCFS2_INVALID_SLOT,
					     ALLOC_NEW_GROUP);
795
	if (status < 0 && status != -ENOSPC) {
796
		mlog_errno(status);
797 798 799
		goto bail;
	}

800 801 802 803 804 805 806
bail:
	return status;
}

/* Callers don't need to care which bitmap (local alloc or main) to
 * use so we figure it out for them, but unfortunately this clutters
 * things a bit. */
807 808 809
static int ocfs2_reserve_clusters_with_limit(struct ocfs2_super *osb,
					     u32 bits_wanted, u64 max_block,
					     struct ocfs2_alloc_context **ac)
810 811 812 813 814
{
	int status;

	mlog_entry_void();

815
	*ac = kzalloc(sizeof(struct ocfs2_alloc_context), GFP_KERNEL);
816 817 818 819 820 821 822
	if (!(*ac)) {
		status = -ENOMEM;
		mlog_errno(status);
		goto bail;
	}

	(*ac)->ac_bits_wanted = bits_wanted;
823
	(*ac)->ac_max_block = max_block;
824 825 826 827 828 829

	status = -ENOSPC;
	if (ocfs2_alloc_should_use_local(osb, bits_wanted)) {
		status = ocfs2_reserve_local_alloc_bits(osb,
							bits_wanted,
							*ac);
830 831 832 833 834
		if (status == -EFBIG) {
			/* The local alloc window is outside ac_max_block.
			 * use the main bitmap. */
			status = -ENOSPC;
		} else if ((status < 0) && (status != -ENOSPC)) {
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
			mlog_errno(status);
			goto bail;
		}
	}

	if (status == -ENOSPC) {
		status = ocfs2_reserve_cluster_bitmap_bits(osb, *ac);
		if (status < 0) {
			if (status != -ENOSPC)
				mlog_errno(status);
			goto bail;
		}
	}

	status = 0;
bail:
	if ((status < 0) && *ac) {
		ocfs2_free_alloc_context(*ac);
		*ac = NULL;
	}

	mlog_exit(status);
	return status;
}

860 861 862 863 864 865 866
int ocfs2_reserve_clusters(struct ocfs2_super *osb,
			   u32 bits_wanted,
			   struct ocfs2_alloc_context **ac)
{
	return ocfs2_reserve_clusters_with_limit(osb, bits_wanted, 0, ac);
}

867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885
/*
 * More or less lifted from ext3. I'll leave their description below:
 *
 * "For ext3 allocations, we must not reuse any blocks which are
 * allocated in the bitmap buffer's "last committed data" copy.  This
 * prevents deletes from freeing up the page for reuse until we have
 * committed the delete transaction.
 *
 * If we didn't do this, then deleting something and reallocating it as
 * data would allow the old block to be overwritten before the
 * transaction committed (because we force data to disk before commit).
 * This would lead to corruption if we crashed between overwriting the
 * data and committing the delete.
 *
 * @@@ We may want to make this allocation behaviour conditional on
 * data-writes at some point, and disable it for metadata allocations or
 * sync-data inodes."
 *
 * Note: OCFS2 already does this differently for metadata vs data
J
Joe Perches 已提交
886
 * allocations, as those bitmaps are separate and undo access is never
887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905
 * called on a metadata group descriptor.
 */
static int ocfs2_test_bg_bit_allocatable(struct buffer_head *bg_bh,
					 int nr)
{
	struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data;

	if (ocfs2_test_bit(nr, (unsigned long *)bg->bg_bitmap))
		return 0;
	if (!buffer_jbd(bg_bh) || !bh2jh(bg_bh)->b_committed_data)
		return 1;

	bg = (struct ocfs2_group_desc *) bh2jh(bg_bh)->b_committed_data;
	return !ocfs2_test_bit(nr, (unsigned long *)bg->bg_bitmap);
}

static int ocfs2_block_group_find_clear_bits(struct ocfs2_super *osb,
					     struct buffer_head *bg_bh,
					     unsigned int bits_wanted,
906
					     unsigned int total_bits,
907 908 909 910 911 912 913 914
					     u16 *bit_off,
					     u16 *bits_found)
{
	void *bitmap;
	u16 best_offset, best_size;
	int offset, start, found, status = 0;
	struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data;

915 916 917
	/* Callers got this descriptor from
	 * ocfs2_read_group_descriptor().  Any corruption is a code bug. */
	BUG_ON(!OCFS2_IS_VALID_GROUP_DESC(bg));
918 919 920 921

	found = start = best_offset = best_size = 0;
	bitmap = bg->bg_bitmap;

922 923
	while((offset = ocfs2_find_next_zero_bit(bitmap, total_bits, start)) != -1) {
		if (offset == total_bits)
924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968
			break;

		if (!ocfs2_test_bg_bit_allocatable(bg_bh, offset)) {
			/* We found a zero, but we can't use it as it
			 * hasn't been put to disk yet! */
			found = 0;
			start = offset + 1;
		} else if (offset == start) {
			/* we found a zero */
			found++;
			/* move start to the next bit to test */
			start++;
		} else {
			/* got a zero after some ones */
			found = 1;
			start = offset + 1;
		}
		if (found > best_size) {
			best_size = found;
			best_offset = start - found;
		}
		/* we got everything we needed */
		if (found == bits_wanted) {
			/* mlog(0, "Found it all!\n"); */
			break;
		}
	}

	/* XXX: I think the first clause is equivalent to the second
	 * 	- jlbec */
	if (found == bits_wanted) {
		*bit_off = start - found;
		*bits_found = found;
	} else if (best_size) {
		*bit_off = best_offset;
		*bits_found = best_size;
	} else {
		status = -ENOSPC;
		/* No error log here -- see the comment above
		 * ocfs2_test_bg_bit_allocatable */
	}

	return status;
}

969
static inline int ocfs2_block_group_set_bits(handle_t *handle,
970 971 972 973 974 975 976 977 978 979 980 981
					     struct inode *alloc_inode,
					     struct ocfs2_group_desc *bg,
					     struct buffer_head *group_bh,
					     unsigned int bit_off,
					     unsigned int num_bits)
{
	int status;
	void *bitmap = bg->bg_bitmap;
	int journal_type = OCFS2_JOURNAL_ACCESS_WRITE;

	mlog_entry_void();

982 983 984
	/* All callers get the descriptor via
	 * ocfs2_read_group_descriptor().  Any corruption is a code bug. */
	BUG_ON(!OCFS2_IS_VALID_GROUP_DESC(bg));
985 986 987 988 989 990 991 992
	BUG_ON(le16_to_cpu(bg->bg_free_bits_count) < num_bits);

	mlog(0, "block_group_set_bits: off = %u, num = %u\n", bit_off,
	     num_bits);

	if (ocfs2_is_cluster_bitmap(alloc_inode))
		journal_type = OCFS2_JOURNAL_ACCESS_UNDO;

993 994 995 996
	status = ocfs2_journal_access_gd(handle,
					 alloc_inode,
					 group_bh,
					 journal_type);
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
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	le16_add_cpu(&bg->bg_free_bits_count, -num_bits);

	while(num_bits--)
		ocfs2_set_bit(bit_off++, bitmap);

	status = ocfs2_journal_dirty(handle,
				     group_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

bail:
	mlog_exit(status);
	return status;
}

/* find the one with the most empty bits */
static inline u16 ocfs2_find_victim_chain(struct ocfs2_chain_list *cl)
{
	u16 curr, best;

	BUG_ON(!cl->cl_next_free_rec);

	best = curr = 0;
	while (curr < le16_to_cpu(cl->cl_next_free_rec)) {
		if (le32_to_cpu(cl->cl_recs[curr].c_free) >
		    le32_to_cpu(cl->cl_recs[best].c_free))
			best = curr;
		curr++;
	}

	BUG_ON(best >= le16_to_cpu(cl->cl_next_free_rec));
	return best;
}

1038
static int ocfs2_relink_block_group(handle_t *handle,
1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
				    struct inode *alloc_inode,
				    struct buffer_head *fe_bh,
				    struct buffer_head *bg_bh,
				    struct buffer_head *prev_bg_bh,
				    u16 chain)
{
	int status;
	/* there is a really tiny chance the journal calls could fail,
	 * but we wouldn't want inconsistent blocks in *any* case. */
	u64 fe_ptr, bg_ptr, prev_bg_ptr;
	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) fe_bh->b_data;
	struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data;
	struct ocfs2_group_desc *prev_bg = (struct ocfs2_group_desc *) prev_bg_bh->b_data;

1053 1054 1055 1056
	/* The caller got these descriptors from
	 * ocfs2_read_group_descriptor().  Any corruption is a code bug. */
	BUG_ON(!OCFS2_IS_VALID_GROUP_DESC(bg));
	BUG_ON(!OCFS2_IS_VALID_GROUP_DESC(prev_bg));
1057

1058
	mlog(0, "Suballoc %llu, chain %u, move group %llu to top, prev = %llu\n",
1059 1060 1061
	     (unsigned long long)le64_to_cpu(fe->i_blkno), chain,
	     (unsigned long long)le64_to_cpu(bg->bg_blkno),
	     (unsigned long long)le64_to_cpu(prev_bg->bg_blkno));
1062 1063 1064 1065 1066

	fe_ptr = le64_to_cpu(fe->id2.i_chain.cl_recs[chain].c_blkno);
	bg_ptr = le64_to_cpu(bg->bg_next_group);
	prev_bg_ptr = le64_to_cpu(prev_bg->bg_next_group);

1067 1068
	status = ocfs2_journal_access_gd(handle, alloc_inode, prev_bg_bh,
					 OCFS2_JOURNAL_ACCESS_WRITE);
1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081
	if (status < 0) {
		mlog_errno(status);
		goto out_rollback;
	}

	prev_bg->bg_next_group = bg->bg_next_group;

	status = ocfs2_journal_dirty(handle, prev_bg_bh);
	if (status < 0) {
		mlog_errno(status);
		goto out_rollback;
	}

1082 1083
	status = ocfs2_journal_access_gd(handle, alloc_inode, bg_bh,
					 OCFS2_JOURNAL_ACCESS_WRITE);
1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096
	if (status < 0) {
		mlog_errno(status);
		goto out_rollback;
	}

	bg->bg_next_group = fe->id2.i_chain.cl_recs[chain].c_blkno;

	status = ocfs2_journal_dirty(handle, bg_bh);
	if (status < 0) {
		mlog_errno(status);
		goto out_rollback;
	}

1097 1098
	status = ocfs2_journal_access_di(handle, alloc_inode, fe_bh,
					 OCFS2_JOURNAL_ACCESS_WRITE);
1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118
	if (status < 0) {
		mlog_errno(status);
		goto out_rollback;
	}

	fe->id2.i_chain.cl_recs[chain].c_blkno = bg->bg_blkno;

	status = ocfs2_journal_dirty(handle, fe_bh);
	if (status < 0) {
		mlog_errno(status);
		goto out_rollback;
	}

	status = 0;
out_rollback:
	if (status < 0) {
		fe->id2.i_chain.cl_recs[chain].c_blkno = cpu_to_le64(fe_ptr);
		bg->bg_next_group = cpu_to_le64(bg_ptr);
		prev_bg->bg_next_group = cpu_to_le64(prev_bg_ptr);
	}
1119

1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134
	mlog_exit(status);
	return status;
}

static inline int ocfs2_block_group_reasonably_empty(struct ocfs2_group_desc *bg,
						     u32 wanted)
{
	return le16_to_cpu(bg->bg_free_bits_count) > wanted;
}

/* return 0 on success, -ENOSPC to keep searching and any other < 0
 * value on error. */
static int ocfs2_cluster_group_search(struct inode *inode,
				      struct buffer_head *group_bh,
				      u32 bits_wanted, u32 min_bits,
1135
				      u64 max_block,
1136 1137 1138 1139
				      u16 *bit_off, u16 *bits_found)
{
	int search = -ENOSPC;
	int ret;
1140
	u64 blkoff;
1141
	struct ocfs2_group_desc *gd = (struct ocfs2_group_desc *) group_bh->b_data;
1142
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1143
	u16 tmp_off, tmp_found;
1144
	unsigned int max_bits, gd_cluster_off;
1145 1146 1147

	BUG_ON(!ocfs2_is_cluster_bitmap(inode));

1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167
	if (gd->bg_free_bits_count) {
		max_bits = le16_to_cpu(gd->bg_bits);

		/* Tail groups in cluster bitmaps which aren't cpg
		 * aligned are prone to partial extention by a failed
		 * fs resize. If the file system resize never got to
		 * update the dinode cluster count, then we don't want
		 * to trust any clusters past it, regardless of what
		 * the group descriptor says. */
		gd_cluster_off = ocfs2_blocks_to_clusters(inode->i_sb,
							  le64_to_cpu(gd->bg_blkno));
		if ((gd_cluster_off + max_bits) >
		    OCFS2_I(inode)->ip_clusters) {
			max_bits = OCFS2_I(inode)->ip_clusters - gd_cluster_off;
			mlog(0, "Desc %llu, bg_bits %u, clusters %u, use %u\n",
			     (unsigned long long)le64_to_cpu(gd->bg_blkno),
			     le16_to_cpu(gd->bg_bits),
			     OCFS2_I(inode)->ip_clusters, max_bits);
		}

1168 1169
		ret = ocfs2_block_group_find_clear_bits(OCFS2_SB(inode->i_sb),
							group_bh, bits_wanted,
1170
							max_bits,
1171 1172 1173 1174
							&tmp_off, &tmp_found);
		if (ret)
			return ret;

1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185
		if (max_block) {
			blkoff = ocfs2_clusters_to_blocks(inode->i_sb,
							  gd_cluster_off +
							  tmp_off + tmp_found);
			mlog(0, "Checking %llu against %llu\n",
			     (unsigned long long)blkoff,
			     (unsigned long long)max_block);
			if (blkoff > max_block)
				return -ENOSPC;
		}

1186 1187 1188 1189 1190 1191 1192 1193
		/* ocfs2_block_group_find_clear_bits() might
		 * return success, but we still want to return
		 * -ENOSPC unless it found the minimum number
		 * of bits. */
		if (min_bits <= tmp_found) {
			*bit_off = tmp_off;
			*bits_found = tmp_found;
			search = 0; /* success */
1194 1195 1196 1197 1198 1199
		} else if (tmp_found) {
			/*
			 * Don't show bits which we'll be returning
			 * for allocation to the local alloc bitmap.
			 */
			ocfs2_local_alloc_seen_free_bits(osb, tmp_found);
1200 1201 1202 1203 1204 1205 1206 1207 1208
		}
	}

	return search;
}

static int ocfs2_block_group_search(struct inode *inode,
				    struct buffer_head *group_bh,
				    u32 bits_wanted, u32 min_bits,
1209
				    u64 max_block,
1210 1211 1212
				    u16 *bit_off, u16 *bits_found)
{
	int ret = -ENOSPC;
1213
	u64 blkoff;
1214 1215 1216 1217 1218
	struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) group_bh->b_data;

	BUG_ON(min_bits != 1);
	BUG_ON(ocfs2_is_cluster_bitmap(inode));

1219
	if (bg->bg_free_bits_count) {
1220 1221
		ret = ocfs2_block_group_find_clear_bits(OCFS2_SB(inode->i_sb),
							group_bh, bits_wanted,
1222
							le16_to_cpu(bg->bg_bits),
1223
							bit_off, bits_found);
1224 1225 1226 1227 1228 1229 1230 1231 1232 1233
		if (!ret && max_block) {
			blkoff = le64_to_cpu(bg->bg_blkno) + *bit_off +
				*bits_found;
			mlog(0, "Checking %llu against %llu\n",
			     (unsigned long long)blkoff,
			     (unsigned long long)max_block);
			if (blkoff > max_block)
				ret = -ENOSPC;
		}
	}
1234 1235 1236 1237

	return ret;
}

M
Mark Fasheh 已提交
1238
static int ocfs2_alloc_dinode_update_counts(struct inode *inode,
1239
				       handle_t *handle,
M
Mark Fasheh 已提交
1240 1241 1242 1243 1244 1245 1246 1247 1248
				       struct buffer_head *di_bh,
				       u32 num_bits,
				       u16 chain)
{
	int ret;
	u32 tmp_used;
	struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data;
	struct ocfs2_chain_list *cl = (struct ocfs2_chain_list *) &di->id2.i_chain;

1249 1250
	ret = ocfs2_journal_access_di(handle, inode, di_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
M
Mark Fasheh 已提交
1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268
	if (ret < 0) {
		mlog_errno(ret);
		goto out;
	}

	tmp_used = le32_to_cpu(di->id1.bitmap1.i_used);
	di->id1.bitmap1.i_used = cpu_to_le32(num_bits + tmp_used);
	le32_add_cpu(&cl->cl_recs[chain].c_free, -num_bits);

	ret = ocfs2_journal_dirty(handle, di_bh);
	if (ret < 0)
		mlog_errno(ret);

out:
	return ret;
}

static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
1269
				  handle_t *handle,
M
Mark Fasheh 已提交
1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280
				  u32 bits_wanted,
				  u32 min_bits,
				  u16 *bit_off,
				  unsigned int *num_bits,
				  u64 gd_blkno,
				  u16 *bits_left)
{
	int ret;
	u16 found;
	struct buffer_head *group_bh = NULL;
	struct ocfs2_group_desc *gd;
1281
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)ac->ac_bh->b_data;
M
Mark Fasheh 已提交
1282 1283
	struct inode *alloc_inode = ac->ac_inode;

1284 1285
	ret = ocfs2_read_group_descriptor(alloc_inode, di, gd_blkno,
					  &group_bh);
M
Mark Fasheh 已提交
1286 1287 1288 1289 1290 1291 1292
	if (ret < 0) {
		mlog_errno(ret);
		return ret;
	}

	gd = (struct ocfs2_group_desc *) group_bh->b_data;
	ret = ac->ac_group_search(alloc_inode, group_bh, bits_wanted, min_bits,
1293
				  ac->ac_max_block, bit_off, &found);
M
Mark Fasheh 已提交
1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322
	if (ret < 0) {
		if (ret != -ENOSPC)
			mlog_errno(ret);
		goto out;
	}

	*num_bits = found;

	ret = ocfs2_alloc_dinode_update_counts(alloc_inode, handle, ac->ac_bh,
					       *num_bits,
					       le16_to_cpu(gd->bg_chain));
	if (ret < 0) {
		mlog_errno(ret);
		goto out;
	}

	ret = ocfs2_block_group_set_bits(handle, alloc_inode, gd, group_bh,
					 *bit_off, *num_bits);
	if (ret < 0)
		mlog_errno(ret);

	*bits_left = le16_to_cpu(gd->bg_free_bits_count);

out:
	brelse(group_bh);

	return ret;
}

1323
static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
1324
			      handle_t *handle,
1325 1326 1327 1328
			      u32 bits_wanted,
			      u32 min_bits,
			      u16 *bit_off,
			      unsigned int *num_bits,
M
Mark Fasheh 已提交
1329 1330
			      u64 *bg_blkno,
			      u16 *bits_left)
1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343
{
	int status;
	u16 chain, tmp_bits;
	u32 tmp_used;
	u64 next_group;
	struct inode *alloc_inode = ac->ac_inode;
	struct buffer_head *group_bh = NULL;
	struct buffer_head *prev_group_bh = NULL;
	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) ac->ac_bh->b_data;
	struct ocfs2_chain_list *cl = (struct ocfs2_chain_list *) &fe->id2.i_chain;
	struct ocfs2_group_desc *bg;

	chain = ac->ac_chain;
1344 1345 1346
	mlog(0, "trying to alloc %u bits from chain %u, inode %llu\n",
	     bits_wanted, chain,
	     (unsigned long long)OCFS2_I(alloc_inode)->ip_blkno);
1347

1348 1349 1350
	status = ocfs2_read_group_descriptor(alloc_inode, fe,
					     le64_to_cpu(cl->cl_recs[chain].c_blkno),
					     &group_bh);
1351 1352 1353 1354 1355 1356 1357 1358 1359
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}
	bg = (struct ocfs2_group_desc *) group_bh->b_data;

	status = -ENOSPC;
	/* for now, the chain search is a bit simplistic. We just use
	 * the 1st group with any empty bits. */
1360 1361 1362 1363
	while ((status = ac->ac_group_search(alloc_inode, group_bh,
					     bits_wanted, min_bits,
					     ac->ac_max_block, bit_off,
					     &tmp_bits)) == -ENOSPC) {
1364 1365
		if (!bg->bg_next_group)
			break;
1366 1367 1368 1369

		brelse(prev_group_bh);
		prev_group_bh = NULL;

1370 1371 1372
		next_group = le64_to_cpu(bg->bg_next_group);
		prev_group_bh = group_bh;
		group_bh = NULL;
1373 1374
		status = ocfs2_read_group_descriptor(alloc_inode, fe,
						     next_group, &group_bh);
1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}
		bg = (struct ocfs2_group_desc *) group_bh->b_data;
	}
	if (status < 0) {
		if (status != -ENOSPC)
			mlog_errno(status);
		goto bail;
	}

1387
	mlog(0, "alloc succeeds: we give %u bits from block group %llu\n",
1388
	     tmp_bits, (unsigned long long)le64_to_cpu(bg->bg_blkno));
1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420

	*num_bits = tmp_bits;

	BUG_ON(*num_bits == 0);

	/*
	 * Keep track of previous block descriptor read. When
	 * we find a target, if we have read more than X
	 * number of descriptors, and the target is reasonably
	 * empty, relink him to top of his chain.
	 *
	 * We've read 0 extra blocks and only send one more to
	 * the transaction, yet the next guy to search has a
	 * much easier time.
	 *
	 * Do this *after* figuring out how many bits we're taking out
	 * of our target group.
	 */
	if (ac->ac_allow_chain_relink &&
	    (prev_group_bh) &&
	    (ocfs2_block_group_reasonably_empty(bg, *num_bits))) {
		status = ocfs2_relink_block_group(handle, alloc_inode,
						  ac->ac_bh, group_bh,
						  prev_group_bh, chain);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}
	}

	/* Ok, claim our bits now: set the info on dinode, chainlist
	 * and then the group */
1421 1422 1423 1424
	status = ocfs2_journal_access_di(handle,
					 alloc_inode,
					 ac->ac_bh,
					 OCFS2_JOURNAL_ACCESS_WRITE);
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
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	tmp_used = le32_to_cpu(fe->id1.bitmap1.i_used);
	fe->id1.bitmap1.i_used = cpu_to_le32(*num_bits + tmp_used);
	le32_add_cpu(&cl->cl_recs[chain].c_free, -(*num_bits));

	status = ocfs2_journal_dirty(handle,
				     ac->ac_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	status = ocfs2_block_group_set_bits(handle,
					    alloc_inode,
					    bg,
					    group_bh,
					    *bit_off,
					    *num_bits);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

1452
	mlog(0, "Allocated %u bits from suballocator %llu\n", *num_bits,
1453
	     (unsigned long long)le64_to_cpu(fe->i_blkno));
1454 1455

	*bg_blkno = le64_to_cpu(bg->bg_blkno);
M
Mark Fasheh 已提交
1456
	*bits_left = le16_to_cpu(bg->bg_free_bits_count);
1457
bail:
1458 1459
	brelse(group_bh);
	brelse(prev_group_bh);
1460 1461 1462 1463 1464 1465 1466 1467

	mlog_exit(status);
	return status;
}

/* will give out up to bits_wanted contiguous bits. */
static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
				     struct ocfs2_alloc_context *ac,
1468
				     handle_t *handle,
1469 1470 1471 1472 1473 1474 1475 1476
				     u32 bits_wanted,
				     u32 min_bits,
				     u16 *bit_off,
				     unsigned int *num_bits,
				     u64 *bg_blkno)
{
	int status;
	u16 victim, i;
M
Mark Fasheh 已提交
1477 1478
	u16 bits_left = 0;
	u64 hint_blkno = ac->ac_last_group;
1479 1480 1481 1482 1483 1484 1485 1486 1487 1488
	struct ocfs2_chain_list *cl;
	struct ocfs2_dinode *fe;

	mlog_entry_void();

	BUG_ON(ac->ac_bits_given >= ac->ac_bits_wanted);
	BUG_ON(bits_wanted > (ac->ac_bits_wanted - ac->ac_bits_given));
	BUG_ON(!ac->ac_bh);

	fe = (struct ocfs2_dinode *) ac->ac_bh->b_data;
1489 1490 1491 1492 1493

	/* The bh was validated by the inode read during
	 * ocfs2_reserve_suballoc_bits().  Any corruption is a code bug. */
	BUG_ON(!OCFS2_IS_VALID_DINODE(fe));

1494 1495
	if (le32_to_cpu(fe->id1.bitmap1.i_used) >=
	    le32_to_cpu(fe->id1.bitmap1.i_total)) {
1496 1497 1498
		ocfs2_error(osb->sb, "Chain allocator dinode %llu has %u used "
			    "bits but only %u total.",
			    (unsigned long long)le64_to_cpu(fe->i_blkno),
1499 1500 1501 1502 1503 1504
			    le32_to_cpu(fe->id1.bitmap1.i_used),
			    le32_to_cpu(fe->id1.bitmap1.i_total));
		status = -EIO;
		goto bail;
	}

M
Mark Fasheh 已提交
1505 1506 1507 1508 1509
	if (hint_blkno) {
		/* Attempt to short-circuit the usual search mechanism
		 * by jumping straight to the most recently used
		 * allocation group. This helps us mantain some
		 * contiguousness across allocations. */
1510 1511
		status = ocfs2_search_one_group(ac, handle, bits_wanted,
						min_bits, bit_off, num_bits,
M
Mark Fasheh 已提交
1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526
						hint_blkno, &bits_left);
		if (!status) {
			/* Be careful to update *bg_blkno here as the
			 * caller is expecting it to be filled in, and
			 * ocfs2_search_one_group() won't do that for
			 * us. */
			*bg_blkno = hint_blkno;
			goto set_hint;
		}
		if (status < 0 && status != -ENOSPC) {
			mlog_errno(status);
			goto bail;
		}
	}

1527 1528 1529 1530 1531 1532
	cl = (struct ocfs2_chain_list *) &fe->id2.i_chain;

	victim = ocfs2_find_victim_chain(cl);
	ac->ac_chain = victim;
	ac->ac_allow_chain_relink = 1;

1533
	status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits, bit_off,
M
Mark Fasheh 已提交
1534
				    num_bits, bg_blkno, &bits_left);
1535
	if (!status)
M
Mark Fasheh 已提交
1536
		goto set_hint;
1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556
	if (status < 0 && status != -ENOSPC) {
		mlog_errno(status);
		goto bail;
	}

	mlog(0, "Search of victim chain %u came up with nothing, "
	     "trying all chains now.\n", victim);

	/* If we didn't pick a good victim, then just default to
	 * searching each chain in order. Don't allow chain relinking
	 * because we only calculate enough journal credits for one
	 * relink per alloc. */
	ac->ac_allow_chain_relink = 0;
	for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i ++) {
		if (i == victim)
			continue;
		if (!cl->cl_recs[i].c_free)
			continue;

		ac->ac_chain = i;
1557
		status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits,
M
Mark Fasheh 已提交
1558 1559
					    bit_off, num_bits, bg_blkno,
					    &bits_left);
1560 1561 1562 1563 1564 1565 1566 1567
		if (!status)
			break;
		if (status < 0 && status != -ENOSPC) {
			mlog_errno(status);
			goto bail;
		}
	}

M
Mark Fasheh 已提交
1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579
set_hint:
	if (status != -ENOSPC) {
		/* If the next search of this group is not likely to
		 * yield a suitable extent, then we reset the last
		 * group hint so as to not waste a disk read */
		if (bits_left < min_bits)
			ac->ac_last_group = 0;
		else
			ac->ac_last_group = *bg_blkno;
	}

bail:
1580 1581 1582 1583 1584
	mlog_exit(status);
	return status;
}

int ocfs2_claim_metadata(struct ocfs2_super *osb,
1585
			 handle_t *handle,
1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600
			 struct ocfs2_alloc_context *ac,
			 u32 bits_wanted,
			 u16 *suballoc_bit_start,
			 unsigned int *num_bits,
			 u64 *blkno_start)
{
	int status;
	u64 bg_blkno;

	BUG_ON(!ac);
	BUG_ON(ac->ac_bits_wanted < (ac->ac_bits_given + bits_wanted));
	BUG_ON(ac->ac_which != OCFS2_AC_USE_META);

	status = ocfs2_claim_suballoc_bits(osb,
					   ac,
1601
					   handle,
1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621
					   bits_wanted,
					   1,
					   suballoc_bit_start,
					   num_bits,
					   &bg_blkno);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}
	atomic_inc(&osb->alloc_stats.bg_allocs);

	*blkno_start = bg_blkno + (u64) *suballoc_bit_start;
	ac->ac_bits_given += (*num_bits);
	status = 0;
bail:
	mlog_exit(status);
	return status;
}

int ocfs2_claim_new_inode(struct ocfs2_super *osb,
1622
			  handle_t *handle,
1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639
			  struct ocfs2_alloc_context *ac,
			  u16 *suballoc_bit,
			  u64 *fe_blkno)
{
	int status;
	unsigned int num_bits;
	u64 bg_blkno;

	mlog_entry_void();

	BUG_ON(!ac);
	BUG_ON(ac->ac_bits_given != 0);
	BUG_ON(ac->ac_bits_wanted != 1);
	BUG_ON(ac->ac_which != OCFS2_AC_USE_INODE);

	status = ocfs2_claim_suballoc_bits(osb,
					   ac,
1640
					   handle,
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
					   1,
					   1,
					   suballoc_bit,
					   &num_bits,
					   &bg_blkno);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}
	atomic_inc(&osb->alloc_stats.bg_allocs);

	BUG_ON(num_bits != 1);

	*fe_blkno = bg_blkno + (u64) (*suballoc_bit);
	ac->ac_bits_given++;
	status = 0;
bail:
	mlog_exit(status);
	return status;
}

/* translate a group desc. blkno and it's bitmap offset into
 * disk cluster offset. */
static inline u32 ocfs2_desc_bitmap_to_cluster_off(struct inode *inode,
						   u64 bg_blkno,
						   u16 bg_bit_off)
{
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	u32 cluster = 0;

	BUG_ON(!ocfs2_is_cluster_bitmap(inode));

	if (bg_blkno != osb->first_cluster_group_blkno)
		cluster = ocfs2_blocks_to_clusters(inode->i_sb, bg_blkno);
	cluster += (u32) bg_bit_off;
	return cluster;
}

/* given a cluster offset, calculate which block group it belongs to
 * and return that block offset. */
1681
u64 ocfs2_which_cluster_group(struct inode *inode, u32 cluster)
1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722
{
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	u32 group_no;

	BUG_ON(!ocfs2_is_cluster_bitmap(inode));

	group_no = cluster / osb->bitmap_cpg;
	if (!group_no)
		return osb->first_cluster_group_blkno;
	return ocfs2_clusters_to_blocks(inode->i_sb,
					group_no * osb->bitmap_cpg);
}

/* given the block number of a cluster start, calculate which cluster
 * group and descriptor bitmap offset that corresponds to. */
static inline void ocfs2_block_to_cluster_group(struct inode *inode,
						u64 data_blkno,
						u64 *bg_blkno,
						u16 *bg_bit_off)
{
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	u32 data_cluster = ocfs2_blocks_to_clusters(osb->sb, data_blkno);

	BUG_ON(!ocfs2_is_cluster_bitmap(inode));

	*bg_blkno = ocfs2_which_cluster_group(inode,
					      data_cluster);

	if (*bg_blkno == osb->first_cluster_group_blkno)
		*bg_bit_off = (u16) data_cluster;
	else
		*bg_bit_off = (u16) ocfs2_blocks_to_clusters(osb->sb,
							     data_blkno - *bg_blkno);
}

/*
 * min_bits - minimum contiguous chunk from this total allocation we
 * can handle. set to what we asked for originally for a full
 * contig. allocation, set to '1' to indicate we can deal with extents
 * of any size.
 */
1723 1724 1725 1726 1727 1728 1729
int __ocfs2_claim_clusters(struct ocfs2_super *osb,
			   handle_t *handle,
			   struct ocfs2_alloc_context *ac,
			   u32 min_clusters,
			   u32 max_clusters,
			   u32 *cluster_start,
			   u32 *num_clusters)
1730 1731
{
	int status;
1732
	unsigned int bits_wanted = max_clusters;
M
Mark Fasheh 已提交
1733
	u64 bg_blkno = 0;
1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755
	u16 bg_bit_off;

	mlog_entry_void();

	BUG_ON(ac->ac_bits_given >= ac->ac_bits_wanted);

	BUG_ON(ac->ac_which != OCFS2_AC_USE_LOCAL
	       && ac->ac_which != OCFS2_AC_USE_MAIN);

	if (ac->ac_which == OCFS2_AC_USE_LOCAL) {
		status = ocfs2_claim_local_alloc_bits(osb,
						      handle,
						      ac,
						      bits_wanted,
						      cluster_start,
						      num_clusters);
		if (!status)
			atomic_inc(&osb->alloc_stats.local_data);
	} else {
		if (min_clusters > (osb->bitmap_cpg - 1)) {
			/* The only paths asking for contiguousness
			 * should know about this already. */
1756 1757 1758
			mlog(ML_ERROR, "minimum allocation requested %u exceeds "
			     "group bitmap size %u!\n", min_clusters,
			     osb->bitmap_cpg);
1759 1760 1761 1762 1763 1764 1765 1766 1767
			status = -ENOSPC;
			goto bail;
		}
		/* clamp the current request down to a realistic size. */
		if (bits_wanted > (osb->bitmap_cpg - 1))
			bits_wanted = osb->bitmap_cpg - 1;

		status = ocfs2_claim_suballoc_bits(osb,
						   ac,
1768
						   handle,
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
						   bits_wanted,
						   min_clusters,
						   &bg_bit_off,
						   num_clusters,
						   &bg_blkno);
		if (!status) {
			*cluster_start =
				ocfs2_desc_bitmap_to_cluster_off(ac->ac_inode,
								 bg_blkno,
								 bg_bit_off);
			atomic_inc(&osb->alloc_stats.bitmap_data);
		}
	}
	if (status < 0) {
		if (status != -ENOSPC)
			mlog_errno(status);
		goto bail;
	}

	ac->ac_bits_given += *num_clusters;

bail:
	mlog_exit(status);
	return status;
}

1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807
int ocfs2_claim_clusters(struct ocfs2_super *osb,
			 handle_t *handle,
			 struct ocfs2_alloc_context *ac,
			 u32 min_clusters,
			 u32 *cluster_start,
			 u32 *num_clusters)
{
	unsigned int bits_wanted = ac->ac_bits_wanted - ac->ac_bits_given;

	return __ocfs2_claim_clusters(osb, handle, ac, min_clusters,
				      bits_wanted, cluster_start, num_clusters);
}

1808
static inline int ocfs2_block_group_clear_bits(handle_t *handle,
1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821
					       struct inode *alloc_inode,
					       struct ocfs2_group_desc *bg,
					       struct buffer_head *group_bh,
					       unsigned int bit_off,
					       unsigned int num_bits)
{
	int status;
	unsigned int tmp;
	int journal_type = OCFS2_JOURNAL_ACCESS_WRITE;
	struct ocfs2_group_desc *undo_bg = NULL;

	mlog_entry_void();

1822 1823 1824
	/* The caller got this descriptor from
	 * ocfs2_read_group_descriptor().  Any corruption is a code bug. */
	BUG_ON(!OCFS2_IS_VALID_GROUP_DESC(bg));
1825 1826 1827 1828 1829 1830

	mlog(0, "off = %u, num = %u\n", bit_off, num_bits);

	if (ocfs2_is_cluster_bitmap(alloc_inode))
		journal_type = OCFS2_JOURNAL_ACCESS_UNDO;

1831 1832
	status = ocfs2_journal_access_gd(handle, alloc_inode, group_bh,
					 journal_type);
1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	if (ocfs2_is_cluster_bitmap(alloc_inode))
		undo_bg = (struct ocfs2_group_desc *) bh2jh(group_bh)->b_committed_data;

	tmp = num_bits;
	while(tmp--) {
		ocfs2_clear_bit((bit_off + tmp),
				(unsigned long *) bg->bg_bitmap);
		if (ocfs2_is_cluster_bitmap(alloc_inode))
			ocfs2_set_bit(bit_off + tmp,
				      (unsigned long *) undo_bg->bg_bitmap);
	}
	le16_add_cpu(&bg->bg_free_bits_count, num_bits);

	status = ocfs2_journal_dirty(handle, group_bh);
	if (status < 0)
		mlog_errno(status);
bail:
	return status;
}

/*
 * expects the suballoc inode to already be locked.
 */
1861 1862 1863 1864 1865 1866
int ocfs2_free_suballoc_bits(handle_t *handle,
			     struct inode *alloc_inode,
			     struct buffer_head *alloc_bh,
			     unsigned int start_bit,
			     u64 bg_blkno,
			     unsigned int count)
1867 1868 1869 1870 1871 1872 1873 1874 1875 1876
{
	int status = 0;
	u32 tmp_used;
	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) alloc_bh->b_data;
	struct ocfs2_chain_list *cl = &fe->id2.i_chain;
	struct buffer_head *group_bh = NULL;
	struct ocfs2_group_desc *group;

	mlog_entry_void();

1877 1878 1879 1880 1881 1882
	/* The alloc_bh comes from ocfs2_free_dinode() or
	 * ocfs2_free_clusters().  The callers have all locked the
	 * allocator and gotten alloc_bh from the lock call.  This
	 * validates the dinode buffer.  Any corruption that has happended
	 * is a code bug. */
	BUG_ON(!OCFS2_IS_VALID_DINODE(fe));
1883 1884
	BUG_ON((count + start_bit) > ocfs2_bits_per_group(cl));

1885 1886 1887
	mlog(0, "%llu: freeing %u bits from group %llu, starting at %u\n",
	     (unsigned long long)OCFS2_I(alloc_inode)->ip_blkno, count,
	     (unsigned long long)bg_blkno, start_bit);
1888

1889 1890
	status = ocfs2_read_group_descriptor(alloc_inode, fe, bg_blkno,
					     &group_bh);
1891 1892 1893 1894 1895
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}
	group = (struct ocfs2_group_desc *) group_bh->b_data;
1896

1897 1898 1899 1900 1901 1902 1903 1904 1905 1906
	BUG_ON((count + start_bit) > le16_to_cpu(group->bg_bits));

	status = ocfs2_block_group_clear_bits(handle, alloc_inode,
					      group, group_bh,
					      start_bit, count);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

1907 1908
	status = ocfs2_journal_access_di(handle, alloc_inode, alloc_bh,
					 OCFS2_JOURNAL_ACCESS_WRITE);
1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	le32_add_cpu(&cl->cl_recs[le16_to_cpu(group->bg_chain)].c_free,
		     count);
	tmp_used = le32_to_cpu(fe->id1.bitmap1.i_used);
	fe->id1.bitmap1.i_used = cpu_to_le32(tmp_used - count);

	status = ocfs2_journal_dirty(handle, alloc_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

bail:
1926
	brelse(group_bh);
1927 1928 1929 1930 1931

	mlog_exit(status);
	return status;
}

1932
int ocfs2_free_dinode(handle_t *handle,
1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944
		      struct inode *inode_alloc_inode,
		      struct buffer_head *inode_alloc_bh,
		      struct ocfs2_dinode *di)
{
	u64 blk = le64_to_cpu(di->i_blkno);
	u16 bit = le16_to_cpu(di->i_suballoc_bit);
	u64 bg_blkno = ocfs2_which_suballoc_group(blk, bit);

	return ocfs2_free_suballoc_bits(handle, inode_alloc_inode,
					inode_alloc_bh, bit, bg_blkno, 1);
}

1945
int ocfs2_free_clusters(handle_t *handle,
1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970
		       struct inode *bitmap_inode,
		       struct buffer_head *bitmap_bh,
		       u64 start_blk,
		       unsigned int num_clusters)
{
	int status;
	u16 bg_start_bit;
	u64 bg_blkno;
	struct ocfs2_dinode *fe;

	/* You can't ever have a contiguous set of clusters
	 * bigger than a block group bitmap so we never have to worry
	 * about looping on them. */

	mlog_entry_void();

	/* This is expensive. We can safely remove once this stuff has
	 * gotten tested really well. */
	BUG_ON(start_blk != ocfs2_clusters_to_blocks(bitmap_inode->i_sb, ocfs2_blocks_to_clusters(bitmap_inode->i_sb, start_blk)));

	fe = (struct ocfs2_dinode *) bitmap_bh->b_data;

	ocfs2_block_to_cluster_group(bitmap_inode, start_blk, &bg_blkno,
				     &bg_start_bit);

1971 1972 1973 1974
	mlog(0, "want to free %u clusters starting at block %llu\n",
	     num_clusters, (unsigned long long)start_blk);
	mlog(0, "bg_blkno = %llu, bg_start_bit = %u\n",
	     (unsigned long long)bg_blkno, bg_start_bit);
1975 1976 1977 1978

	status = ocfs2_free_suballoc_bits(handle, bitmap_inode, bitmap_bh,
					  bg_start_bit, bg_blkno,
					  num_clusters);
1979
	if (status < 0) {
1980
		mlog_errno(status);
1981 1982 1983 1984 1985
		goto out;
	}

	ocfs2_local_alloc_seen_free_bits(OCFS2_SB(bitmap_inode->i_sb),
					 num_clusters);
1986

1987
out:
1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
	mlog_exit(status);
	return status;
}

static inline void ocfs2_debug_bg(struct ocfs2_group_desc *bg)
{
	printk("Block Group:\n");
	printk("bg_signature:       %s\n", bg->bg_signature);
	printk("bg_size:            %u\n", bg->bg_size);
	printk("bg_bits:            %u\n", bg->bg_bits);
	printk("bg_free_bits_count: %u\n", bg->bg_free_bits_count);
	printk("bg_chain:           %u\n", bg->bg_chain);
	printk("bg_generation:      %u\n", le32_to_cpu(bg->bg_generation));
2001 2002 2003 2004 2005 2006
	printk("bg_next_group:      %llu\n",
	       (unsigned long long)bg->bg_next_group);
	printk("bg_parent_dinode:   %llu\n",
	       (unsigned long long)bg->bg_parent_dinode);
	printk("bg_blkno:           %llu\n",
	       (unsigned long long)bg->bg_blkno);
2007 2008 2009 2010 2011 2012
}

static inline void ocfs2_debug_suballoc_inode(struct ocfs2_dinode *fe)
{
	int i;

2013
	printk("Suballoc Inode %llu:\n", (unsigned long long)fe->i_blkno);
2014
	printk("i_signature:                  %s\n", fe->i_signature);
2015 2016
	printk("i_size:                       %llu\n",
	       (unsigned long long)fe->i_size);
2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033
	printk("i_clusters:                   %u\n", fe->i_clusters);
	printk("i_generation:                 %u\n",
	       le32_to_cpu(fe->i_generation));
	printk("id1.bitmap1.i_used:           %u\n",
	       le32_to_cpu(fe->id1.bitmap1.i_used));
	printk("id1.bitmap1.i_total:          %u\n",
	       le32_to_cpu(fe->id1.bitmap1.i_total));
	printk("id2.i_chain.cl_cpg:           %u\n", fe->id2.i_chain.cl_cpg);
	printk("id2.i_chain.cl_bpc:           %u\n", fe->id2.i_chain.cl_bpc);
	printk("id2.i_chain.cl_count:         %u\n", fe->id2.i_chain.cl_count);
	printk("id2.i_chain.cl_next_free_rec: %u\n",
	       fe->id2.i_chain.cl_next_free_rec);
	for(i = 0; i < fe->id2.i_chain.cl_next_free_rec; i++) {
		printk("fe->id2.i_chain.cl_recs[%d].c_free:  %u\n", i,
		       fe->id2.i_chain.cl_recs[i].c_free);
		printk("fe->id2.i_chain.cl_recs[%d].c_total: %u\n", i,
		       fe->id2.i_chain.cl_recs[i].c_total);
2034 2035
		printk("fe->id2.i_chain.cl_recs[%d].c_blkno: %llu\n", i,
		       (unsigned long long)fe->id2.i_chain.cl_recs[i].c_blkno);
2036 2037
	}
}
2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048

/*
 * For a given allocation, determine which allocators will need to be
 * accessed, and lock them, reserving the appropriate number of bits.
 *
 * Sparse file systems call this from ocfs2_write_begin_nolock()
 * and ocfs2_allocate_unwritten_extents().
 *
 * File systems which don't support holes call this from
 * ocfs2_extend_allocation().
 */
2049 2050
int ocfs2_lock_allocators(struct inode *inode,
			  struct ocfs2_extent_tree *et,
2051 2052
			  u32 clusters_to_add, u32 extents_to_split,
			  struct ocfs2_alloc_context **data_ac,
2053
			  struct ocfs2_alloc_context **meta_ac)
2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064
{
	int ret = 0, num_free_extents;
	unsigned int max_recs_needed = clusters_to_add + 2 * extents_to_split;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);

	*meta_ac = NULL;
	if (data_ac)
		*data_ac = NULL;

	BUG_ON(clusters_to_add != 0 && data_ac == NULL);

2065
	num_free_extents = ocfs2_num_free_extents(osb, inode, et);
2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086
	if (num_free_extents < 0) {
		ret = num_free_extents;
		mlog_errno(ret);
		goto out;
	}

	/*
	 * Sparse allocation file systems need to be more conservative
	 * with reserving room for expansion - the actual allocation
	 * happens while we've got a journal handle open so re-taking
	 * a cluster lock (because we ran out of room for another
	 * extent) will violate ordering rules.
	 *
	 * Most of the time we'll only be seeing this 1 cluster at a time
	 * anyway.
	 *
	 * Always lock for any unwritten extents - we might want to
	 * add blocks during a split.
	 */
	if (!num_free_extents ||
	    (ocfs2_sparse_alloc(osb) && num_free_extents < max_recs_needed)) {
2087
		ret = ocfs2_reserve_new_metadata(osb, et->et_root_el, meta_ac);
2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118
		if (ret < 0) {
			if (ret != -ENOSPC)
				mlog_errno(ret);
			goto out;
		}
	}

	if (clusters_to_add == 0)
		goto out;

	ret = ocfs2_reserve_clusters(osb, clusters_to_add, data_ac);
	if (ret < 0) {
		if (ret != -ENOSPC)
			mlog_errno(ret);
		goto out;
	}

out:
	if (ret) {
		if (*meta_ac) {
			ocfs2_free_alloc_context(*meta_ac);
			*meta_ac = NULL;
		}

		/*
		 * We cannot have an error and a non null *data_ac.
		 */
	}

	return ret;
}