checkpoint.c 35.7 KB
Newer Older
J
Jaegeuk Kim 已提交
1
/*
J
Jaegeuk Kim 已提交
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
 * fs/f2fs/checkpoint.c
 *
 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
 *             http://www.samsung.com/
 *
 * 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.
 */
#include <linux/fs.h>
#include <linux/bio.h>
#include <linux/mpage.h>
#include <linux/writeback.h>
#include <linux/blkdev.h>
#include <linux/f2fs_fs.h>
#include <linux/pagevec.h>
#include <linux/swap.h>

#include "f2fs.h"
#include "node.h"
#include "segment.h"
J
Jaegeuk Kim 已提交
23
#include "trace.h"
24
#include <trace/events/f2fs.h>
J
Jaegeuk Kim 已提交
25

J
Jaegeuk Kim 已提交
26
static struct kmem_cache *ino_entry_slab;
27
struct kmem_cache *inode_entry_slab;
J
Jaegeuk Kim 已提交
28

29 30
void f2fs_stop_checkpoint(struct f2fs_sb_info *sbi, bool end_io)
{
31
	set_ckpt_flags(sbi, CP_ERROR_FLAG);
32
	if (!end_io)
33
		f2fs_flush_merged_writes(sbi);
34 35
}

J
Jaegeuk Kim 已提交
36
/*
J
Jaegeuk Kim 已提交
37 38 39 40
 * We guarantee no failure on the returned page.
 */
struct page *grab_meta_page(struct f2fs_sb_info *sbi, pgoff_t index)
{
G
Gu Zheng 已提交
41
	struct address_space *mapping = META_MAPPING(sbi);
J
Jaegeuk Kim 已提交
42 43
	struct page *page = NULL;
repeat:
44
	page = f2fs_grab_cache_page(mapping, index, false);
J
Jaegeuk Kim 已提交
45 46 47 48
	if (!page) {
		cond_resched();
		goto repeat;
	}
49
	f2fs_wait_on_page_writeback(page, META, true);
50 51
	if (!PageUptodate(page))
		SetPageUptodate(page);
J
Jaegeuk Kim 已提交
52 53 54
	return page;
}

J
Jaegeuk Kim 已提交
55
/*
J
Jaegeuk Kim 已提交
56 57
 * We guarantee no failure on the returned page.
 */
58 59
static struct page *__get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index,
							bool is_meta)
J
Jaegeuk Kim 已提交
60
{
G
Gu Zheng 已提交
61
	struct address_space *mapping = META_MAPPING(sbi);
J
Jaegeuk Kim 已提交
62
	struct page *page;
63
	struct f2fs_io_info fio = {
64
		.sbi = sbi,
65
		.type = META,
M
Mike Christie 已提交
66
		.op = REQ_OP_READ,
67
		.op_flags = REQ_META | REQ_PRIO,
68 69
		.old_blkaddr = index,
		.new_blkaddr = index,
70
		.encrypted_page = NULL,
71
	};
72 73

	if (unlikely(!is_meta))
M
Mike Christie 已提交
74
		fio.op_flags &= ~REQ_META;
J
Jaegeuk Kim 已提交
75
repeat:
76
	page = f2fs_grab_cache_page(mapping, index, false);
J
Jaegeuk Kim 已提交
77 78 79 80
	if (!page) {
		cond_resched();
		goto repeat;
	}
81 82 83
	if (PageUptodate(page))
		goto out;

84 85
	fio.page = page;

86 87
	if (f2fs_submit_page_bio(&fio)) {
		f2fs_put_page(page, 1);
J
Jaegeuk Kim 已提交
88
		goto repeat;
89
	}
J
Jaegeuk Kim 已提交
90

91
	lock_page(page);
92
	if (unlikely(page->mapping != mapping)) {
93 94 95
		f2fs_put_page(page, 1);
		goto repeat;
	}
96 97 98 99 100 101 102

	/*
	 * if there is any IO error when accessing device, make our filesystem
	 * readonly and make sure do not write checkpoint with non-uptodate
	 * meta page.
	 */
	if (unlikely(!PageUptodate(page)))
103
		f2fs_stop_checkpoint(sbi, false);
104
out:
J
Jaegeuk Kim 已提交
105 106 107
	return page;
}

108 109 110 111 112 113 114 115 116 117 118
struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index)
{
	return __get_meta_page(sbi, index, true);
}

/* for POR only */
struct page *get_tmp_page(struct f2fs_sb_info *sbi, pgoff_t index)
{
	return __get_meta_page(sbi, index, false);
}

119
bool is_valid_blkaddr(struct f2fs_sb_info *sbi, block_t blkaddr, int type)
120 121 122
{
	switch (type) {
	case META_NAT:
123
		break;
124
	case META_SIT:
125 126 127
		if (unlikely(blkaddr >= SIT_BLK_CNT(sbi)))
			return false;
		break;
128
	case META_SSA:
129 130 131 132
		if (unlikely(blkaddr >= MAIN_BLKADDR(sbi) ||
			blkaddr < SM_I(sbi)->ssa_blkaddr))
			return false;
		break;
133
	case META_CP:
134 135 136 137
		if (unlikely(blkaddr >= SIT_I(sbi)->sit_base_addr ||
			blkaddr < __start_cp_addr(sbi)))
			return false;
		break;
138
	case META_POR:
139 140 141 142
		if (unlikely(blkaddr >= MAX_BLKADDR(sbi) ||
			blkaddr < MAIN_BLKADDR(sbi)))
			return false;
		break;
143 144 145
	default:
		BUG();
	}
146 147

	return true;
148 149 150
}

/*
151
 * Readahead CP/NAT/SIT/SSA pages
152
 */
153 154
int ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages,
							int type, bool sync)
155 156
{
	struct page *page;
157
	block_t blkno = start;
158
	struct f2fs_io_info fio = {
159
		.sbi = sbi,
160
		.type = META,
M
Mike Christie 已提交
161
		.op = REQ_OP_READ,
162
		.op_flags = sync ? (REQ_META | REQ_PRIO) : REQ_RAHEAD,
163
		.encrypted_page = NULL,
164
		.in_list = false,
165
	};
C
Chao Yu 已提交
166
	struct blk_plug plug;
167

168
	if (unlikely(type == META_POR))
M
Mike Christie 已提交
169
		fio.op_flags &= ~REQ_META;
170

C
Chao Yu 已提交
171
	blk_start_plug(&plug);
172 173
	for (; nrpages-- > 0; blkno++) {

174 175 176
		if (!is_valid_blkaddr(sbi, blkno, type))
			goto out;

177 178
		switch (type) {
		case META_NAT:
179 180
			if (unlikely(blkno >=
					NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid)))
181
				blkno = 0;
182
			/* get nat block addr */
183
			fio.new_blkaddr = current_nat_addr(sbi,
184 185 186 187
					blkno * NAT_ENTRY_PER_BLOCK);
			break;
		case META_SIT:
			/* get sit block addr */
188
			fio.new_blkaddr = current_sit_addr(sbi,
189 190
					blkno * SIT_ENTRY_PER_BLOCK);
			break;
191
		case META_SSA:
192
		case META_CP:
193
		case META_POR:
194
			fio.new_blkaddr = blkno;
195 196 197 198 199
			break;
		default:
			BUG();
		}

200 201
		page = f2fs_grab_cache_page(META_MAPPING(sbi),
						fio.new_blkaddr, false);
202 203 204 205 206 207 208
		if (!page)
			continue;
		if (PageUptodate(page)) {
			f2fs_put_page(page, 1);
			continue;
		}

209
		fio.page = page;
210
		f2fs_submit_page_bio(&fio);
211 212 213
		f2fs_put_page(page, 0);
	}
out:
C
Chao Yu 已提交
214
	blk_finish_plug(&plug);
215 216 217
	return blkno - start;
}

218 219 220 221 222 223
void ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index)
{
	struct page *page;
	bool readahead = false;

	page = find_get_page(META_MAPPING(sbi), index);
224
	if (!page || !PageUptodate(page))
225 226 227 228
		readahead = true;
	f2fs_put_page(page, 0);

	if (readahead)
229
		ra_meta_pages(sbi, index, BIO_MAX_PAGES, META_POR, true);
230 231
}

C
Chao Yu 已提交
232 233 234
static int __f2fs_write_meta_page(struct page *page,
				struct writeback_control *wbc,
				enum iostat_type io_type)
J
Jaegeuk Kim 已提交
235
{
236
	struct f2fs_sb_info *sbi = F2FS_P_SB(page);
J
Jaegeuk Kim 已提交
237

238 239
	trace_f2fs_writepage(page, META);

240 241 242 243 244
	if (unlikely(f2fs_cp_error(sbi))) {
		dec_page_count(sbi, F2FS_DIRTY_META);
		unlock_page(page);
		return 0;
	}
245
	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
246
		goto redirty_out;
247
	if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0))
248
		goto redirty_out;
J
Jaegeuk Kim 已提交
249

C
Chao Yu 已提交
250
	write_meta_page(sbi, page, io_type);
251
	dec_page_count(sbi, F2FS_DIRTY_META);
252 253

	if (wbc->for_reclaim)
254 255
		f2fs_submit_merged_write_cond(sbi, page->mapping->host,
						0, page->index, META);
256

257
	unlock_page(page);
258

259
	if (unlikely(f2fs_cp_error(sbi)))
260
		f2fs_submit_merged_write(sbi, META);
261

262
	return 0;
263 264

redirty_out:
265
	redirty_page_for_writepage(wbc, page);
266
	return AOP_WRITEPAGE_ACTIVATE;
J
Jaegeuk Kim 已提交
267 268
}

C
Chao Yu 已提交
269 270 271 272 273 274
static int f2fs_write_meta_page(struct page *page,
				struct writeback_control *wbc)
{
	return __f2fs_write_meta_page(page, wbc, FS_META_IO);
}

J
Jaegeuk Kim 已提交
275 276 277
static int f2fs_write_meta_pages(struct address_space *mapping,
				struct writeback_control *wbc)
{
278
	struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
279
	long diff, written;
J
Jaegeuk Kim 已提交
280

281 282 283
	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
		goto skip_write;

284
	/* collect a number of dirty meta pages and write together */
285 286
	if (wbc->for_kupdate ||
		get_pages(sbi, F2FS_DIRTY_META) < nr_pages_to_skip(sbi, META))
287
		goto skip_write;
J
Jaegeuk Kim 已提交
288

289 290 291
	/* if locked failed, cp will flush dirty pages instead */
	if (!mutex_trylock(&sbi->cp_mutex))
		goto skip_write;
Y
Yunlei He 已提交
292

293
	trace_f2fs_writepages(mapping->host, wbc, META);
294
	diff = nr_pages_to_write(sbi, META, wbc);
C
Chao Yu 已提交
295
	written = sync_meta_pages(sbi, META, wbc->nr_to_write, FS_META_IO);
J
Jaegeuk Kim 已提交
296
	mutex_unlock(&sbi->cp_mutex);
297
	wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff);
J
Jaegeuk Kim 已提交
298
	return 0;
299 300 301

skip_write:
	wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META);
Y
Yunlei He 已提交
302
	trace_f2fs_writepages(mapping->host, wbc, META);
303
	return 0;
J
Jaegeuk Kim 已提交
304 305 306
}

long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type,
C
Chao Yu 已提交
307
				long nr_to_write, enum iostat_type io_type)
J
Jaegeuk Kim 已提交
308
{
G
Gu Zheng 已提交
309
	struct address_space *mapping = META_MAPPING(sbi);
J
Jan Kara 已提交
310
	pgoff_t index = 0, prev = ULONG_MAX;
J
Jaegeuk Kim 已提交
311 312
	struct pagevec pvec;
	long nwritten = 0;
J
Jan Kara 已提交
313
	int nr_pages;
J
Jaegeuk Kim 已提交
314 315 316
	struct writeback_control wbc = {
		.for_reclaim = 0,
	};
C
Chao Yu 已提交
317
	struct blk_plug plug;
J
Jaegeuk Kim 已提交
318

319
	pagevec_init(&pvec);
J
Jaegeuk Kim 已提交
320

C
Chao Yu 已提交
321 322
	blk_start_plug(&plug);

J
Jan Kara 已提交
323
	while ((nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
324
				PAGECACHE_TAG_DIRTY))) {
J
Jan Kara 已提交
325
		int i;
J
Jaegeuk Kim 已提交
326 327 328

		for (i = 0; i < nr_pages; i++) {
			struct page *page = pvec.pages[i];
329

330
			if (prev == ULONG_MAX)
331 332 333 334 335 336
				prev = page->index - 1;
			if (nr_to_write != LONG_MAX && page->index != prev + 1) {
				pagevec_release(&pvec);
				goto stop;
			}

J
Jaegeuk Kim 已提交
337
			lock_page(page);
338 339 340 341 342 343 344 345 346 347 348

			if (unlikely(page->mapping != mapping)) {
continue_unlock:
				unlock_page(page);
				continue;
			}
			if (!PageDirty(page)) {
				/* someone wrote it for us */
				goto continue_unlock;
			}

349 350 351
			f2fs_wait_on_page_writeback(page, META, true);

			BUG_ON(PageWriteback(page));
352 353 354
			if (!clear_page_dirty_for_io(page))
				goto continue_unlock;

C
Chao Yu 已提交
355
			if (__f2fs_write_meta_page(page, &wbc, io_type)) {
356 357 358
				unlock_page(page);
				break;
			}
359
			nwritten++;
360
			prev = page->index;
361
			if (unlikely(nwritten >= nr_to_write))
J
Jaegeuk Kim 已提交
362 363 364 365 366
				break;
		}
		pagevec_release(&pvec);
		cond_resched();
	}
367
stop:
J
Jaegeuk Kim 已提交
368
	if (nwritten)
369
		f2fs_submit_merged_write(sbi, type);
J
Jaegeuk Kim 已提交
370

C
Chao Yu 已提交
371 372
	blk_finish_plug(&plug);

J
Jaegeuk Kim 已提交
373 374 375 376 377
	return nwritten;
}

static int f2fs_set_meta_page_dirty(struct page *page)
{
378 379
	trace_f2fs_set_page_dirty(page, META);

380 381
	if (!PageUptodate(page))
		SetPageUptodate(page);
J
Jaegeuk Kim 已提交
382
	if (!PageDirty(page)) {
383
		f2fs_set_page_dirty_nobuffers(page);
384
		inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META);
385
		SetPagePrivate(page);
J
Jaegeuk Kim 已提交
386
		f2fs_trace_pid(page);
J
Jaegeuk Kim 已提交
387 388 389 390 391 392 393 394 395
		return 1;
	}
	return 0;
}

const struct address_space_operations f2fs_meta_aops = {
	.writepage	= f2fs_write_meta_page,
	.writepages	= f2fs_write_meta_pages,
	.set_page_dirty	= f2fs_set_meta_page_dirty,
396 397
	.invalidatepage = f2fs_invalidate_page,
	.releasepage	= f2fs_release_page,
398 399 400
#ifdef CONFIG_MIGRATION
	.migratepage    = f2fs_migrate_page,
#endif
J
Jaegeuk Kim 已提交
401 402
};

C
Chao Yu 已提交
403 404
static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino,
						unsigned int devidx, int type)
405
{
406
	struct inode_management *im = &sbi->im[type];
407 408 409
	struct ino_entry *e, *tmp;

	tmp = f2fs_kmem_cache_alloc(ino_entry_slab, GFP_NOFS);
410

411
	radix_tree_preload(GFP_NOFS | __GFP_NOFAIL);
412

413 414
	spin_lock(&im->ino_lock);
	e = radix_tree_lookup(&im->ino_root, ino);
415
	if (!e) {
416
		e = tmp;
417 418 419
		if (unlikely(radix_tree_insert(&im->ino_root, ino, e)))
			f2fs_bug_on(sbi, 1);

420 421
		memset(e, 0, sizeof(struct ino_entry));
		e->ino = ino;
422

423
		list_add_tail(&e->list, &im->ino_list);
424
		if (type != ORPHAN_INO)
425
			im->ino_num++;
426
	}
C
Chao Yu 已提交
427 428 429 430

	if (type == FLUSH_INO)
		f2fs_set_bit(devidx, (char *)&e->dirty_device);

431
	spin_unlock(&im->ino_lock);
432
	radix_tree_preload_end();
433 434 435

	if (e != tmp)
		kmem_cache_free(ino_entry_slab, tmp);
436 437
}

J
Jaegeuk Kim 已提交
438
static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
439
{
440
	struct inode_management *im = &sbi->im[type];
J
Jaegeuk Kim 已提交
441
	struct ino_entry *e;
442

443 444
	spin_lock(&im->ino_lock);
	e = radix_tree_lookup(&im->ino_root, ino);
445 446
	if (e) {
		list_del(&e->list);
447 448 449
		radix_tree_delete(&im->ino_root, ino);
		im->ino_num--;
		spin_unlock(&im->ino_lock);
450 451
		kmem_cache_free(ino_entry_slab, e);
		return;
452
	}
453
	spin_unlock(&im->ino_lock);
454 455
}

456
void add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
457 458
{
	/* add new dirty ino entry into list */
C
Chao Yu 已提交
459
	__add_ino_entry(sbi, ino, 0, type);
460 461
}

462
void remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
463 464 465 466 467 468 469 470
{
	/* remove dirty ino entry from list */
	__remove_ino_entry(sbi, ino, type);
}

/* mode should be APPEND_INO or UPDATE_INO */
bool exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode)
{
471
	struct inode_management *im = &sbi->im[mode];
472
	struct ino_entry *e;
473 474 475 476

	spin_lock(&im->ino_lock);
	e = radix_tree_lookup(&im->ino_root, ino);
	spin_unlock(&im->ino_lock);
477 478 479
	return e ? true : false;
}

480
void release_ino_entry(struct f2fs_sb_info *sbi, bool all)
481 482 483 484
{
	struct ino_entry *e, *tmp;
	int i;

C
Chao Yu 已提交
485
	for (i = all ? ORPHAN_INO : APPEND_INO; i < MAX_INO_ENTRY; i++) {
486 487 488 489
		struct inode_management *im = &sbi->im[i];

		spin_lock(&im->ino_lock);
		list_for_each_entry_safe(e, tmp, &im->ino_list, list) {
490
			list_del(&e->list);
491
			radix_tree_delete(&im->ino_root, e->ino);
492
			kmem_cache_free(ino_entry_slab, e);
493
			im->ino_num--;
494
		}
495
		spin_unlock(&im->ino_lock);
496 497 498
	}
}

C
Chao Yu 已提交
499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519
void set_dirty_device(struct f2fs_sb_info *sbi, nid_t ino,
					unsigned int devidx, int type)
{
	__add_ino_entry(sbi, ino, devidx, type);
}

bool is_dirty_device(struct f2fs_sb_info *sbi, nid_t ino,
					unsigned int devidx, int type)
{
	struct inode_management *im = &sbi->im[type];
	struct ino_entry *e;
	bool is_dirty = false;

	spin_lock(&im->ino_lock);
	e = radix_tree_lookup(&im->ino_root, ino);
	if (e && f2fs_test_bit(devidx, (char *)&e->dirty_device))
		is_dirty = true;
	spin_unlock(&im->ino_lock);
	return is_dirty;
}

J
Jaegeuk Kim 已提交
520
int acquire_orphan_inode(struct f2fs_sb_info *sbi)
J
Jaegeuk Kim 已提交
521
{
522
	struct inode_management *im = &sbi->im[ORPHAN_INO];
J
Jaegeuk Kim 已提交
523 524
	int err = 0;

525
	spin_lock(&im->ino_lock);
J
Jaegeuk Kim 已提交
526 527

#ifdef CONFIG_F2FS_FAULT_INJECTION
528
	if (time_to_inject(sbi, FAULT_ORPHAN)) {
J
Jaegeuk Kim 已提交
529
		spin_unlock(&im->ino_lock);
530
		f2fs_show_injection_info(FAULT_ORPHAN);
J
Jaegeuk Kim 已提交
531 532 533
		return -ENOSPC;
	}
#endif
534
	if (unlikely(im->ino_num >= sbi->max_orphans))
J
Jaegeuk Kim 已提交
535
		err = -ENOSPC;
J
Jaegeuk Kim 已提交
536
	else
537 538
		im->ino_num++;
	spin_unlock(&im->ino_lock);
539

J
Jaegeuk Kim 已提交
540 541 542
	return err;
}

J
Jaegeuk Kim 已提交
543 544
void release_orphan_inode(struct f2fs_sb_info *sbi)
{
545 546 547 548 549 550
	struct inode_management *im = &sbi->im[ORPHAN_INO];

	spin_lock(&im->ino_lock);
	f2fs_bug_on(sbi, im->ino_num == 0);
	im->ino_num--;
	spin_unlock(&im->ino_lock);
J
Jaegeuk Kim 已提交
551 552
}

553
void add_orphan_inode(struct inode *inode)
J
Jaegeuk Kim 已提交
554
{
555
	/* add new orphan ino entry into list */
C
Chao Yu 已提交
556
	__add_ino_entry(F2FS_I_SB(inode), inode->i_ino, 0, ORPHAN_INO);
557
	update_inode_page(inode);
J
Jaegeuk Kim 已提交
558 559 560 561
}

void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino)
{
562
	/* remove orphan entry from orphan list */
J
Jaegeuk Kim 已提交
563
	__remove_ino_entry(sbi, ino, ORPHAN_INO);
J
Jaegeuk Kim 已提交
564 565
}

566
static int recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino)
J
Jaegeuk Kim 已提交
567
{
568
	struct inode *inode;
569
	struct node_info ni;
570 571 572 573 574 575 576 577 578 579
	int err = acquire_orphan_inode(sbi);

	if (err) {
		set_sbi_flag(sbi, SBI_NEED_FSCK);
		f2fs_msg(sbi->sb, KERN_WARNING,
				"%s: orphan failed (ino=%x), run fsck to fix.",
				__func__, ino);
		return err;
	}

C
Chao Yu 已提交
580
	__add_ino_entry(sbi, ino, 0, ORPHAN_INO);
581

582
	inode = f2fs_iget_retry(sbi->sb, ino);
583 584 585 586 587 588 589 590 591
	if (IS_ERR(inode)) {
		/*
		 * there should be a bug that we can't find the entry
		 * to orphan inode.
		 */
		f2fs_bug_on(sbi, PTR_ERR(inode) == -ENOENT);
		return PTR_ERR(inode);
	}

J
Jaegeuk Kim 已提交
592 593 594 595
	clear_nlink(inode);

	/* truncate all the data during iput */
	iput(inode);
596 597 598 599 600

	get_node_info(sbi, ino, &ni);

	/* ENOMEM was fully retried in f2fs_evict_inode. */
	if (ni.blk_addr != NULL_ADDR) {
601 602
		set_sbi_flag(sbi, SBI_NEED_FSCK);
		f2fs_msg(sbi->sb, KERN_WARNING,
603
			"%s: orphan failed (ino=%x) by kernel, retry mount.",
604 605
				__func__, ino);
		return -EIO;
606
	}
607
	__remove_ino_entry(sbi, ino, ORPHAN_INO);
608
	return 0;
J
Jaegeuk Kim 已提交
609 610
}

611
int recover_orphan_inodes(struct f2fs_sb_info *sbi)
J
Jaegeuk Kim 已提交
612
{
613
	block_t start_blk, orphan_blocks, i, j;
C
Chao Yu 已提交
614 615
	unsigned int s_flags = sbi->sb->s_flags;
	int err = 0;
J
Jaegeuk Kim 已提交
616 617 618
#ifdef CONFIG_QUOTA
	int quota_enabled;
#endif
J
Jaegeuk Kim 已提交
619

620
	if (!is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG))
621
		return 0;
J
Jaegeuk Kim 已提交
622

623
	if (s_flags & SB_RDONLY) {
C
Chao Yu 已提交
624
		f2fs_msg(sbi->sb, KERN_INFO, "orphan cleanup on readonly fs");
625
		sbi->sb->s_flags &= ~SB_RDONLY;
C
Chao Yu 已提交
626 627 628 629
	}

#ifdef CONFIG_QUOTA
	/* Needed for iput() to work correctly and not trash data */
630
	sbi->sb->s_flags |= SB_ACTIVE;
J
Jaegeuk Kim 已提交
631

C
Chao Yu 已提交
632
	/* Turn on quotas so that they are updated correctly */
633
	quota_enabled = f2fs_enable_quota_files(sbi, s_flags & SB_RDONLY);
C
Chao Yu 已提交
634 635
#endif

W
Wanpeng Li 已提交
636
	start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi);
637
	orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi);
J
Jaegeuk Kim 已提交
638

639
	ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP, true);
640

641
	for (i = 0; i < orphan_blocks; i++) {
J
Jaegeuk Kim 已提交
642 643 644 645 646 647
		struct page *page = get_meta_page(sbi, start_blk + i);
		struct f2fs_orphan_block *orphan_blk;

		orphan_blk = (struct f2fs_orphan_block *)page_address(page);
		for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) {
			nid_t ino = le32_to_cpu(orphan_blk->ino[j]);
648 649 650
			err = recover_orphan_inode(sbi, ino);
			if (err) {
				f2fs_put_page(page, 1);
C
Chao Yu 已提交
651
				goto out;
652
			}
J
Jaegeuk Kim 已提交
653 654 655 656
		}
		f2fs_put_page(page, 1);
	}
	/* clear Orphan Flag */
657
	clear_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG);
C
Chao Yu 已提交
658 659 660
out:
#ifdef CONFIG_QUOTA
	/* Turn quotas off */
J
Jaegeuk Kim 已提交
661 662
	if (quota_enabled)
		f2fs_quota_off_umount(sbi->sb);
C
Chao Yu 已提交
663
#endif
664
	sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
C
Chao Yu 已提交
665 666

	return err;
J
Jaegeuk Kim 已提交
667 668 669 670
}

static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk)
{
671
	struct list_head *head;
J
Jaegeuk Kim 已提交
672 673
	struct f2fs_orphan_block *orphan_blk = NULL;
	unsigned int nentries = 0;
C
Chao Yu 已提交
674
	unsigned short index = 1;
675
	unsigned short orphan_blocks;
676
	struct page *page = NULL;
J
Jaegeuk Kim 已提交
677
	struct ino_entry *orphan = NULL;
678
	struct inode_management *im = &sbi->im[ORPHAN_INO];
J
Jaegeuk Kim 已提交
679

680
	orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num);
681

682 683 684 685 686
	/*
	 * we don't need to do spin_lock(&im->ino_lock) here, since all the
	 * orphan inode operations are covered under f2fs_lock_op().
	 * And, spin_lock should be avoided due to page operations below.
	 */
687
	head = &im->ino_list;
J
Jaegeuk Kim 已提交
688 689

	/* loop for each orphan inode entry and write them in Jornal block */
690 691
	list_for_each_entry(orphan, head, list) {
		if (!page) {
C
Chao Yu 已提交
692
			page = grab_meta_page(sbi, start_blk++);
693 694 695 696
			orphan_blk =
				(struct f2fs_orphan_block *)page_address(page);
			memset(orphan_blk, 0, sizeof(*orphan_blk));
		}
J
Jaegeuk Kim 已提交
697

698
		orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino);
J
Jaegeuk Kim 已提交
699

700
		if (nentries == F2FS_ORPHANS_PER_BLOCK) {
J
Jaegeuk Kim 已提交
701 702 703 704 705 706 707 708 709 710 711 712 713 714
			/*
			 * an orphan block is full of 1020 entries,
			 * then we need to flush current orphan blocks
			 * and bring another one in memory
			 */
			orphan_blk->blk_addr = cpu_to_le16(index);
			orphan_blk->blk_count = cpu_to_le16(orphan_blocks);
			orphan_blk->entry_count = cpu_to_le32(nentries);
			set_page_dirty(page);
			f2fs_put_page(page, 1);
			index++;
			nentries = 0;
			page = NULL;
		}
715
	}
J
Jaegeuk Kim 已提交
716

717 718 719 720 721 722
	if (page) {
		orphan_blk->blk_addr = cpu_to_le16(index);
		orphan_blk->blk_count = cpu_to_le16(orphan_blocks);
		orphan_blk->entry_count = cpu_to_le32(nentries);
		set_page_dirty(page);
		f2fs_put_page(page, 1);
J
Jaegeuk Kim 已提交
723 724 725
	}
}

726 727 728
static int get_checkpoint_version(struct f2fs_sb_info *sbi, block_t cp_addr,
		struct f2fs_checkpoint **cp_block, struct page **cp_page,
		unsigned long long *version)
J
Jaegeuk Kim 已提交
729 730
{
	unsigned long blk_size = sbi->blocksize;
731
	size_t crc_offset = 0;
J
Jaegeuk Kim 已提交
732
	__u32 crc = 0;
J
Jaegeuk Kim 已提交
733

734 735
	*cp_page = get_meta_page(sbi, cp_addr);
	*cp_block = (struct f2fs_checkpoint *)page_address(*cp_page);
J
Jaegeuk Kim 已提交
736

737
	crc_offset = le32_to_cpu((*cp_block)->checksum_offset);
738
	if (crc_offset > (blk_size - sizeof(__le32))) {
739 740 741 742
		f2fs_msg(sbi->sb, KERN_WARNING,
			"invalid crc_offset: %zu", crc_offset);
		return -EINVAL;
	}
J
Jaegeuk Kim 已提交
743

744
	crc = cur_cp_crc(*cp_block);
745 746 747 748
	if (!f2fs_crc_valid(sbi, crc, *cp_block, crc_offset)) {
		f2fs_msg(sbi->sb, KERN_WARNING, "invalid crc value");
		return -EINVAL;
	}
J
Jaegeuk Kim 已提交
749

750 751 752
	*version = cur_cp_version(*cp_block);
	return 0;
}
J
Jaegeuk Kim 已提交
753

754 755 756 757 758 759 760
static struct page *validate_checkpoint(struct f2fs_sb_info *sbi,
				block_t cp_addr, unsigned long long *version)
{
	struct page *cp_page_1 = NULL, *cp_page_2 = NULL;
	struct f2fs_checkpoint *cp_block = NULL;
	unsigned long long cur_version = 0, pre_version = 0;
	int err;
J
Jaegeuk Kim 已提交
761

762 763 764 765 766
	err = get_checkpoint_version(sbi, cp_addr, &cp_block,
					&cp_page_1, version);
	if (err)
		goto invalid_cp1;
	pre_version = *version;
J
Jaegeuk Kim 已提交
767

768 769 770 771
	cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1;
	err = get_checkpoint_version(sbi, cp_addr, &cp_block,
					&cp_page_2, version);
	if (err)
J
Jaegeuk Kim 已提交
772
		goto invalid_cp2;
773
	cur_version = *version;
J
Jaegeuk Kim 已提交
774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794

	if (cur_version == pre_version) {
		*version = cur_version;
		f2fs_put_page(cp_page_2, 1);
		return cp_page_1;
	}
invalid_cp2:
	f2fs_put_page(cp_page_2, 1);
invalid_cp1:
	f2fs_put_page(cp_page_1, 1);
	return NULL;
}

int get_valid_checkpoint(struct f2fs_sb_info *sbi)
{
	struct f2fs_checkpoint *cp_block;
	struct f2fs_super_block *fsb = sbi->raw_super;
	struct page *cp1, *cp2, *cur_page;
	unsigned long blk_size = sbi->blocksize;
	unsigned long long cp1_version = 0, cp2_version = 0;
	unsigned long long cp_start_blk_no;
W
Wanpeng Li 已提交
795
	unsigned int cp_blks = 1 + __cp_payload(sbi);
C
Changman Lee 已提交
796 797
	block_t cp_blk_no;
	int i;
J
Jaegeuk Kim 已提交
798

C
Chao Yu 已提交
799
	sbi->ckpt = f2fs_kzalloc(sbi, cp_blks * blk_size, GFP_KERNEL);
J
Jaegeuk Kim 已提交
800 801 802 803 804 805 806 807 808 809
	if (!sbi->ckpt)
		return -ENOMEM;
	/*
	 * Finding out valid cp block involves read both
	 * sets( cp pack1 and cp pack 2)
	 */
	cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr);
	cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version);

	/* The second checkpoint pack should start at the next segment */
810 811
	cp_start_blk_no += ((unsigned long long)1) <<
				le32_to_cpu(fsb->log_blocks_per_seg);
J
Jaegeuk Kim 已提交
812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829
	cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version);

	if (cp1 && cp2) {
		if (ver_after(cp2_version, cp1_version))
			cur_page = cp2;
		else
			cur_page = cp1;
	} else if (cp1) {
		cur_page = cp1;
	} else if (cp2) {
		cur_page = cp2;
	} else {
		goto fail_no_cp;
	}

	cp_block = (struct f2fs_checkpoint *)page_address(cur_page);
	memcpy(sbi->ckpt, cp_block, blk_size);

830 831
	/* Sanity checking of checkpoint */
	if (sanity_check_ckpt(sbi))
832
		goto free_fail_no_cp;
833

834 835 836 837
	if (cur_page == cp1)
		sbi->cur_cp_pack = 1;
	else
		sbi->cur_cp_pack = 2;
838

C
Changman Lee 已提交
839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855
	if (cp_blks <= 1)
		goto done;

	cp_blk_no = le32_to_cpu(fsb->cp_blkaddr);
	if (cur_page == cp2)
		cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg);

	for (i = 1; i < cp_blks; i++) {
		void *sit_bitmap_ptr;
		unsigned char *ckpt = (unsigned char *)sbi->ckpt;

		cur_page = get_meta_page(sbi, cp_blk_no + i);
		sit_bitmap_ptr = page_address(cur_page);
		memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size);
		f2fs_put_page(cur_page, 1);
	}
done:
J
Jaegeuk Kim 已提交
856 857 858 859
	f2fs_put_page(cp1, 1);
	f2fs_put_page(cp2, 1);
	return 0;

860 861 862
free_fail_no_cp:
	f2fs_put_page(cp1, 1);
	f2fs_put_page(cp2, 1);
J
Jaegeuk Kim 已提交
863 864 865 866 867
fail_no_cp:
	kfree(sbi->ckpt);
	return -EINVAL;
}

868
static void __add_dirty_inode(struct inode *inode, enum inode_type type)
J
Jaegeuk Kim 已提交
869
{
870
	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
871
	int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE;
J
Jaegeuk Kim 已提交
872

873
	if (is_inode_flag_set(inode, flag))
874
		return;
875

876
	set_inode_flag(inode, flag);
877 878 879
	if (!f2fs_is_volatile_file(inode))
		list_add_tail(&F2FS_I(inode)->dirty_list,
						&sbi->inode_list[type]);
C
Chao Yu 已提交
880
	stat_inc_dirty_inode(sbi, type);
881 882
}

883
static void __remove_dirty_inode(struct inode *inode, enum inode_type type)
C
Chao Yu 已提交
884
{
885
	int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE;
C
Chao Yu 已提交
886

887
	if (get_dirty_pages(inode) || !is_inode_flag_set(inode, flag))
C
Chao Yu 已提交
888 889
		return;

890 891
	list_del_init(&F2FS_I(inode)->dirty_list);
	clear_inode_flag(inode, flag);
C
Chao Yu 已提交
892
	stat_dec_dirty_inode(F2FS_I_SB(inode), type);
C
Chao Yu 已提交
893 894
}

895
void update_dirty_page(struct inode *inode, struct page *page)
896
{
897
	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
898
	enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE;
899

900 901
	if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) &&
			!S_ISLNK(inode->i_mode))
J
Jaegeuk Kim 已提交
902
		return;
903

904 905
	spin_lock(&sbi->inode_lock[type]);
	if (type != FILE_INODE || test_opt(sbi, DATA_FLUSH))
906
		__add_dirty_inode(inode, type);
907
	inode_inc_dirty_pages(inode);
908 909
	spin_unlock(&sbi->inode_lock[type]);

910
	SetPagePrivate(page);
J
Jaegeuk Kim 已提交
911
	f2fs_trace_pid(page);
912 913
}

914
void remove_dirty_inode(struct inode *inode)
J
Jaegeuk Kim 已提交
915
{
916
	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
917
	enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE;
J
Jaegeuk Kim 已提交
918

919 920
	if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) &&
			!S_ISLNK(inode->i_mode))
J
Jaegeuk Kim 已提交
921 922
		return;

923 924 925
	if (type == FILE_INODE && !test_opt(sbi, DATA_FLUSH))
		return;

926 927 928
	spin_lock(&sbi->inode_lock[type]);
	__remove_dirty_inode(inode, type);
	spin_unlock(&sbi->inode_lock[type]);
929 930
}

C
Chao Yu 已提交
931
int sync_dirty_inodes(struct f2fs_sb_info *sbi, enum inode_type type)
J
Jaegeuk Kim 已提交
932
{
933
	struct list_head *head;
J
Jaegeuk Kim 已提交
934
	struct inode *inode;
935
	struct f2fs_inode_info *fi;
936
	bool is_dir = (type == DIR_INODE);
J
Jaegeuk Kim 已提交
937
	unsigned long ino = 0;
938 939 940 941

	trace_f2fs_sync_dirty_inodes_enter(sbi->sb, is_dir,
				get_pages(sbi, is_dir ?
				F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA));
J
Jaegeuk Kim 已提交
942
retry:
943
	if (unlikely(f2fs_cp_error(sbi)))
C
Chao Yu 已提交
944
		return -EIO;
945

946
	spin_lock(&sbi->inode_lock[type]);
947

948
	head = &sbi->inode_list[type];
J
Jaegeuk Kim 已提交
949
	if (list_empty(head)) {
950
		spin_unlock(&sbi->inode_lock[type]);
951 952 953
		trace_f2fs_sync_dirty_inodes_exit(sbi->sb, is_dir,
				get_pages(sbi, is_dir ?
				F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA));
C
Chao Yu 已提交
954
		return 0;
J
Jaegeuk Kim 已提交
955
	}
956
	fi = list_first_entry(head, struct f2fs_inode_info, dirty_list);
957
	inode = igrab(&fi->vfs_inode);
958
	spin_unlock(&sbi->inode_lock[type]);
J
Jaegeuk Kim 已提交
959
	if (inode) {
J
Jaegeuk Kim 已提交
960 961
		unsigned long cur_ino = inode->i_ino;

C
Chao Yu 已提交
962 963 964
		if (is_dir)
			F2FS_I(inode)->cp_task = current;

965
		filemap_fdatawrite(inode->i_mapping);
C
Chao Yu 已提交
966 967 968 969

		if (is_dir)
			F2FS_I(inode)->cp_task = NULL;

J
Jaegeuk Kim 已提交
970
		iput(inode);
J
Jaegeuk Kim 已提交
971 972 973 974 975 976 977
		/* We need to give cpu to another writers. */
		if (ino == cur_ino) {
			congestion_wait(BLK_RW_ASYNC, HZ/50);
			cond_resched();
		} else {
			ino = cur_ino;
		}
J
Jaegeuk Kim 已提交
978 979 980 981 982
	} else {
		/*
		 * We should submit bio, since it exists several
		 * wribacking dentry pages in the freeing inode.
		 */
983
		f2fs_submit_merged_write(sbi, DATA);
984
		cond_resched();
J
Jaegeuk Kim 已提交
985 986 987 988
	}
	goto retry;
}

989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004
int f2fs_sync_inode_meta(struct f2fs_sb_info *sbi)
{
	struct list_head *head = &sbi->inode_list[DIRTY_META];
	struct inode *inode;
	struct f2fs_inode_info *fi;
	s64 total = get_pages(sbi, F2FS_DIRTY_IMETA);

	while (total--) {
		if (unlikely(f2fs_cp_error(sbi)))
			return -EIO;

		spin_lock(&sbi->inode_lock[DIRTY_META]);
		if (list_empty(head)) {
			spin_unlock(&sbi->inode_lock[DIRTY_META]);
			return 0;
		}
1005
		fi = list_first_entry(head, struct f2fs_inode_info,
1006 1007 1008 1009
							gdirty_list);
		inode = igrab(&fi->vfs_inode);
		spin_unlock(&sbi->inode_lock[DIRTY_META]);
		if (inode) {
1010 1011 1012 1013 1014
			sync_inode_metadata(inode, 0);

			/* it's on eviction */
			if (is_inode_flag_set(inode, FI_DIRTY_INODE))
				update_inode_page(inode);
1015 1016
			iput(inode);
		}
C
Chao Yu 已提交
1017
	}
1018 1019 1020
	return 0;
}

1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033
static void __prepare_cp_block(struct f2fs_sb_info *sbi)
{
	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
	struct f2fs_nm_info *nm_i = NM_I(sbi);
	nid_t last_nid = nm_i->next_scan_nid;

	next_free_nid(sbi, &last_nid);
	ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi));
	ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi));
	ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi));
	ckpt->next_free_nid = cpu_to_le32(last_nid);
}

J
Jaegeuk Kim 已提交
1034
/*
J
Jaegeuk Kim 已提交
1035 1036
 * Freeze all the FS-operations for checkpoint.
 */
1037
static int block_operations(struct f2fs_sb_info *sbi)
J
Jaegeuk Kim 已提交
1038 1039 1040 1041 1042 1043
{
	struct writeback_control wbc = {
		.sync_mode = WB_SYNC_ALL,
		.nr_to_write = LONG_MAX,
		.for_reclaim = 0,
	};
1044
	struct blk_plug plug;
1045
	int err = 0;
1046 1047 1048

	blk_start_plug(&plug);

1049
retry_flush_dents:
1050
	f2fs_lock_all(sbi);
J
Jaegeuk Kim 已提交
1051 1052
	/* write all the dirty dentry pages */
	if (get_pages(sbi, F2FS_DIRTY_DENTS)) {
1053
		f2fs_unlock_all(sbi);
C
Chao Yu 已提交
1054 1055
		err = sync_dirty_inodes(sbi, DIR_INODE);
		if (err)
1056
			goto out;
1057
		cond_resched();
1058
		goto retry_flush_dents;
J
Jaegeuk Kim 已提交
1059 1060
	}

1061 1062 1063 1064 1065 1066
	/*
	 * POR: we should ensure that there are no dirty node pages
	 * until finishing nat/sit flush. inode->i_blocks can be updated.
	 */
	down_write(&sbi->node_change);

1067
	if (get_pages(sbi, F2FS_DIRTY_IMETA)) {
1068
		up_write(&sbi->node_change);
1069 1070 1071 1072
		f2fs_unlock_all(sbi);
		err = f2fs_sync_inode_meta(sbi);
		if (err)
			goto out;
1073
		cond_resched();
1074 1075 1076
		goto retry_flush_dents;
	}

1077
retry_flush_nodes:
1078
	down_write(&sbi->node_write);
J
Jaegeuk Kim 已提交
1079 1080

	if (get_pages(sbi, F2FS_DIRTY_NODES)) {
1081
		up_write(&sbi->node_write);
C
Chao Yu 已提交
1082
		err = sync_node_pages(sbi, &wbc, false, FS_CP_NODE_IO);
C
Chao Yu 已提交
1083
		if (err) {
1084
			up_write(&sbi->node_change);
1085 1086 1087
			f2fs_unlock_all(sbi);
			goto out;
		}
1088
		cond_resched();
1089
		goto retry_flush_nodes;
J
Jaegeuk Kim 已提交
1090
	}
1091 1092 1093 1094 1095 1096 1097

	/*
	 * sbi->node_change is used only for AIO write_begin path which produces
	 * dirty node blocks and some checkpoint values by block allocation.
	 */
	__prepare_cp_block(sbi);
	up_write(&sbi->node_change);
1098
out:
1099
	blk_finish_plug(&plug);
1100
	return err;
J
Jaegeuk Kim 已提交
1101 1102 1103 1104
}

static void unblock_operations(struct f2fs_sb_info *sbi)
{
1105
	up_write(&sbi->node_write);
1106
	f2fs_unlock_all(sbi);
J
Jaegeuk Kim 已提交
1107 1108
}

1109 1110 1111 1112 1113 1114 1115
static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi)
{
	DEFINE_WAIT(wait);

	for (;;) {
		prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE);

1116
		if (!get_pages(sbi, F2FS_WB_CP_DATA))
1117 1118
			break;

1119
		io_schedule_timeout(5*HZ);
1120 1121 1122 1123
	}
	finish_wait(&sbi->cp_wait, &wait);
}

1124 1125 1126 1127
static void update_ckpt_flags(struct f2fs_sb_info *sbi, struct cp_control *cpc)
{
	unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num;
	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1128
	unsigned long flags;
1129

1130
	spin_lock_irqsave(&sbi->cp_lock, flags);
1131

1132
	if ((cpc->reason & CP_UMOUNT) &&
1133
			le32_to_cpu(ckpt->cp_pack_total_block_count) >
1134 1135 1136
			sbi->blocks_per_seg - NM_I(sbi)->nat_bits_blocks)
		disable_nat_bits(sbi, false);

1137 1138
	if (cpc->reason & CP_TRIMMED)
		__set_ckpt_flags(ckpt, CP_TRIMMED_FLAG);
C
Chao Yu 已提交
1139 1140
	else
		__clear_ckpt_flags(ckpt, CP_TRIMMED_FLAG);
1141

1142
	if (cpc->reason & CP_UMOUNT)
1143 1144 1145 1146
		__set_ckpt_flags(ckpt, CP_UMOUNT_FLAG);
	else
		__clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG);

1147
	if (cpc->reason & CP_FASTBOOT)
1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161
		__set_ckpt_flags(ckpt, CP_FASTBOOT_FLAG);
	else
		__clear_ckpt_flags(ckpt, CP_FASTBOOT_FLAG);

	if (orphan_num)
		__set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG);
	else
		__clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG);

	if (is_sbi_flag_set(sbi, SBI_NEED_FSCK))
		__set_ckpt_flags(ckpt, CP_FSCK_FLAG);

	/* set this flag to activate crc|cp_ver for recovery */
	__set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG);
1162
	__clear_ckpt_flags(ckpt, CP_NOCRC_RECOVERY_FLAG);
1163

1164
	spin_unlock_irqrestore(&sbi->cp_lock, flags);
1165 1166
}

C
Chao Yu 已提交
1167
static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
J
Jaegeuk Kim 已提交
1168 1169
{
	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1170
	struct f2fs_nm_info *nm_i = NM_I(sbi);
1171
	unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num, flags;
J
Jaegeuk Kim 已提交
1172 1173
	block_t start_blk;
	unsigned int data_sum_blocks, orphan_blocks;
J
Jaegeuk Kim 已提交
1174
	__u32 crc32 = 0;
J
Jaegeuk Kim 已提交
1175
	int i;
W
Wanpeng Li 已提交
1176
	int cp_payload_blks = __cp_payload(sbi);
1177 1178 1179
	struct super_block *sb = sbi->sb;
	struct curseg_info *seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
	u64 kbytes_written;
1180
	int err;
J
Jaegeuk Kim 已提交
1181 1182

	/* Flush all the NAT/SIT pages */
1183
	while (get_pages(sbi, F2FS_DIRTY_META)) {
C
Chao Yu 已提交
1184
		sync_meta_pages(sbi, META, LONG_MAX, FS_CP_META_IO);
1185
		if (unlikely(f2fs_cp_error(sbi)))
C
Chao Yu 已提交
1186
			return -EIO;
1187
	}
J
Jaegeuk Kim 已提交
1188 1189 1190 1191 1192 1193 1194

	/*
	 * modify checkpoint
	 * version number is already updated
	 */
	ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi));
	ckpt->free_segment_count = cpu_to_le32(free_segments(sbi));
C
Chao Yu 已提交
1195
	for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
J
Jaegeuk Kim 已提交
1196 1197 1198 1199 1200 1201 1202
		ckpt->cur_node_segno[i] =
			cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE));
		ckpt->cur_node_blkoff[i] =
			cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE));
		ckpt->alloc_type[i + CURSEG_HOT_NODE] =
				curseg_alloc_type(sbi, i + CURSEG_HOT_NODE);
	}
C
Chao Yu 已提交
1203
	for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
J
Jaegeuk Kim 已提交
1204 1205 1206 1207 1208 1209 1210 1211 1212
		ckpt->cur_data_segno[i] =
			cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA));
		ckpt->cur_data_blkoff[i] =
			cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA));
		ckpt->alloc_type[i + CURSEG_HOT_DATA] =
				curseg_alloc_type(sbi, i + CURSEG_HOT_DATA);
	}

	/* 2 cp  + n data seg summary + orphan inode blocks */
1213
	data_sum_blocks = npages_for_summary_flush(sbi, false);
1214
	spin_lock_irqsave(&sbi->cp_lock, flags);
C
Chao Yu 已提交
1215
	if (data_sum_blocks < NR_CURSEG_DATA_TYPE)
1216
		__set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG);
J
Jaegeuk Kim 已提交
1217
	else
1218
		__clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG);
1219
	spin_unlock_irqrestore(&sbi->cp_lock, flags);
J
Jaegeuk Kim 已提交
1220

1221
	orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num);
C
Changman Lee 已提交
1222 1223
	ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks +
			orphan_blocks);
J
Jaegeuk Kim 已提交
1224

1225
	if (__remain_node_summaries(cpc->reason))
C
Chao Yu 已提交
1226
		ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+
C
Changman Lee 已提交
1227 1228
				cp_payload_blks + data_sum_blocks +
				orphan_blocks + NR_CURSEG_NODE_TYPE);
1229
	else
C
Chao Yu 已提交
1230
		ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS +
C
Changman Lee 已提交
1231 1232
				cp_payload_blks + data_sum_blocks +
				orphan_blocks);
1233

1234 1235
	/* update ckpt flag for checkpoint */
	update_ckpt_flags(sbi, cpc);
1236

J
Jaegeuk Kim 已提交
1237 1238 1239 1240
	/* update SIT/NAT bitmap */
	get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP));
	get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP));

K
Keith Mok 已提交
1241
	crc32 = f2fs_crc32(sbi, ckpt, le32_to_cpu(ckpt->checksum_offset));
J
Jaegeuk Kim 已提交
1242 1243
	*((__le32 *)((unsigned char *)ckpt +
				le32_to_cpu(ckpt->checksum_offset)))
J
Jaegeuk Kim 已提交
1244 1245
				= cpu_to_le32(crc32);

1246
	start_blk = __start_cp_next_addr(sbi);
J
Jaegeuk Kim 已提交
1247

1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262
	/* write nat bits */
	if (enabled_nat_bits(sbi, cpc)) {
		__u64 cp_ver = cur_cp_version(ckpt);
		block_t blk;

		cp_ver |= ((__u64)crc32 << 32);
		*(__le64 *)nm_i->nat_bits = cpu_to_le64(cp_ver);

		blk = start_blk + sbi->blocks_per_seg - nm_i->nat_bits_blocks;
		for (i = 0; i < nm_i->nat_bits_blocks; i++)
			update_meta_page(sbi, nm_i->nat_bits +
					(i << F2FS_BLKSIZE_BITS), blk + i);

		/* Flush all the NAT BITS pages */
		while (get_pages(sbi, F2FS_DIRTY_META)) {
C
Chao Yu 已提交
1263
			sync_meta_pages(sbi, META, LONG_MAX, FS_CP_META_IO);
1264 1265 1266 1267 1268
			if (unlikely(f2fs_cp_error(sbi)))
				return -EIO;
		}
	}

1269 1270 1271
	/* need to wait for end_io results */
	wait_on_all_pages_writeback(sbi);
	if (unlikely(f2fs_cp_error(sbi)))
C
Chao Yu 已提交
1272
		return -EIO;
1273

1274 1275 1276 1277 1278
	/* flush all device cache */
	err = f2fs_flush_device_cache(sbi);
	if (err)
		return err;

J
Jaegeuk Kim 已提交
1279
	/* write out checkpoint buffer at block 0 */
C
Chao Yu 已提交
1280 1281 1282 1283 1284
	update_meta_page(sbi, ckpt, start_blk++);

	for (i = 1; i < 1 + cp_payload_blks; i++)
		update_meta_page(sbi, (char *)ckpt + i * F2FS_BLKSIZE,
							start_blk++);
C
Changman Lee 已提交
1285

1286
	if (orphan_num) {
J
Jaegeuk Kim 已提交
1287 1288 1289 1290 1291 1292
		write_orphan_inodes(sbi, start_blk);
		start_blk += orphan_blocks;
	}

	write_data_summaries(sbi, start_blk);
	start_blk += data_sum_blocks;
1293 1294 1295 1296 1297 1298

	/* Record write statistics in the hot node summary */
	kbytes_written = sbi->kbytes_written;
	if (sb->s_bdev->bd_part)
		kbytes_written += BD_PART_WRITTEN(sbi);

1299
	seg_i->journal->info.kbytes_written = cpu_to_le64(kbytes_written);
1300

1301
	if (__remain_node_summaries(cpc->reason)) {
J
Jaegeuk Kim 已提交
1302 1303 1304 1305 1306
		write_node_summaries(sbi, start_blk);
		start_blk += NR_CURSEG_NODE_TYPE;
	}

	/* writeout checkpoint block */
C
Chao Yu 已提交
1307
	update_meta_page(sbi, ckpt, start_blk);
J
Jaegeuk Kim 已提交
1308 1309

	/* wait for previous submitted node/meta pages writeback */
1310
	wait_on_all_pages_writeback(sbi);
J
Jaegeuk Kim 已提交
1311

1312
	if (unlikely(f2fs_cp_error(sbi)))
C
Chao Yu 已提交
1313
		return -EIO;
1314

1315 1316
	filemap_fdatawait_range(NODE_MAPPING(sbi), 0, LLONG_MAX);
	filemap_fdatawait_range(META_MAPPING(sbi), 0, LLONG_MAX);
J
Jaegeuk Kim 已提交
1317 1318 1319

	/* update user_block_counts */
	sbi->last_valid_block_count = sbi->total_valid_block_count;
1320
	percpu_counter_set(&sbi->alloc_valid_block_count, 0);
J
Jaegeuk Kim 已提交
1321 1322

	/* Here, we only have one bio having CP pack */
C
Chao Yu 已提交
1323
	sync_meta_pages(sbi, META_FLUSH, LONG_MAX, FS_CP_META_IO);
J
Jaegeuk Kim 已提交
1324

1325 1326 1327
	/* wait for previous submitted meta pages writeback */
	wait_on_all_pages_writeback(sbi);

1328
	release_ino_entry(sbi, false);
1329 1330

	if (unlikely(f2fs_cp_error(sbi)))
C
Chao Yu 已提交
1331
		return -EIO;
1332

1333
	clear_sbi_flag(sbi, SBI_IS_DIRTY);
1334
	clear_sbi_flag(sbi, SBI_NEED_CP);
1335
	__set_cp_next_pack(sbi);
C
Chao Yu 已提交
1336

1337 1338 1339 1340 1341 1342 1343 1344 1345 1346
	/*
	 * redirty superblock if metadata like node page or inode cache is
	 * updated during writing checkpoint.
	 */
	if (get_pages(sbi, F2FS_DIRTY_NODES) ||
			get_pages(sbi, F2FS_DIRTY_IMETA))
		set_sbi_flag(sbi, SBI_IS_DIRTY);

	f2fs_bug_on(sbi, get_pages(sbi, F2FS_DIRTY_DENTS));

C
Chao Yu 已提交
1347
	return 0;
J
Jaegeuk Kim 已提交
1348 1349
}

J
Jaegeuk Kim 已提交
1350
/*
A
arter97 已提交
1351
 * We guarantee that this checkpoint procedure will not fail.
J
Jaegeuk Kim 已提交
1352
 */
C
Chao Yu 已提交
1353
int write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
J
Jaegeuk Kim 已提交
1354 1355 1356
{
	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
	unsigned long long ckpt_ver;
C
Chao Yu 已提交
1357
	int err = 0;
J
Jaegeuk Kim 已提交
1358

1359
	mutex_lock(&sbi->cp_mutex);
J
Jaegeuk Kim 已提交
1360

1361
	if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) &&
1362 1363
		((cpc->reason & CP_FASTBOOT) || (cpc->reason & CP_SYNC) ||
		((cpc->reason & CP_DISCARD) && !sbi->discard_blks)))
J
Jaegeuk Kim 已提交
1364
		goto out;
C
Chao Yu 已提交
1365 1366
	if (unlikely(f2fs_cp_error(sbi))) {
		err = -EIO;
1367
		goto out;
C
Chao Yu 已提交
1368 1369 1370
	}
	if (f2fs_readonly(sbi->sb)) {
		err = -EROFS;
1371
		goto out;
C
Chao Yu 已提交
1372
	}
W
Wanpeng Li 已提交
1373 1374 1375

	trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops");

C
Chao Yu 已提交
1376 1377
	err = block_operations(sbi);
	if (err)
1378
		goto out;
J
Jaegeuk Kim 已提交
1379

1380
	trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops");
1381

1382
	f2fs_flush_merged_writes(sbi);
J
Jaegeuk Kim 已提交
1383

1384
	/* this is the case of multiple fstrims without any changes */
1385
	if (cpc->reason & CP_DISCARD) {
1386 1387 1388 1389 1390
		if (!exist_trim_candidates(sbi, cpc)) {
			unblock_operations(sbi);
			goto out;
		}

1391 1392 1393 1394 1395 1396 1397 1398
		if (NM_I(sbi)->dirty_nat_cnt == 0 &&
				SIT_I(sbi)->dirty_sentries == 0 &&
				prefree_segments(sbi) == 0) {
			flush_sit_entries(sbi, cpc);
			clear_prefree_segments(sbi, cpc);
			unblock_operations(sbi);
			goto out;
		}
1399 1400
	}

J
Jaegeuk Kim 已提交
1401 1402 1403 1404 1405
	/*
	 * update checkpoint pack index
	 * Increase the version number so that
	 * SIT entries and seg summaries are written at correct place
	 */
1406
	ckpt_ver = cur_cp_version(ckpt);
J
Jaegeuk Kim 已提交
1407 1408 1409
	ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver);

	/* write cached NAT/SIT entries to NAT/SIT area */
1410
	flush_nat_entries(sbi, cpc);
1411
	flush_sit_entries(sbi, cpc);
J
Jaegeuk Kim 已提交
1412 1413

	/* unlock all the fs_lock[] in do_checkpoint() */
C
Chao Yu 已提交
1414
	err = do_checkpoint(sbi, cpc);
1415
	if (err)
1416
		release_discard_addrs(sbi);
1417
	else
1418
		clear_prefree_segments(sbi, cpc);
C
Chao Yu 已提交
1419

J
Jaegeuk Kim 已提交
1420
	unblock_operations(sbi);
1421
	stat_inc_cp_count(sbi->stat_info);
1422

1423
	if (cpc->reason & CP_RECOVERY)
1424 1425
		f2fs_msg(sbi->sb, KERN_NOTICE,
			"checkpoint: version = %llx", ckpt_ver);
1426 1427

	/* do checkpoint periodically */
1428
	f2fs_update_time(sbi, CP_TIME);
1429
	trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint");
J
Jaegeuk Kim 已提交
1430 1431
out:
	mutex_unlock(&sbi->cp_mutex);
C
Chao Yu 已提交
1432
	return err;
J
Jaegeuk Kim 已提交
1433 1434
}

J
Jaegeuk Kim 已提交
1435
void init_ino_entry_info(struct f2fs_sb_info *sbi)
J
Jaegeuk Kim 已提交
1436
{
J
Jaegeuk Kim 已提交
1437 1438 1439
	int i;

	for (i = 0; i < MAX_INO_ENTRY; i++) {
1440 1441 1442 1443 1444 1445
		struct inode_management *im = &sbi->im[i];

		INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC);
		spin_lock_init(&im->ino_lock);
		INIT_LIST_HEAD(&im->ino_list);
		im->ino_num = 0;
J
Jaegeuk Kim 已提交
1446 1447
	}

C
Chao Yu 已提交
1448
	sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS -
1449 1450
			NR_CURSEG_TYPE - __cp_payload(sbi)) *
				F2FS_ORPHANS_PER_BLOCK;
J
Jaegeuk Kim 已提交
1451 1452
}

1453
int __init create_checkpoint_caches(void)
J
Jaegeuk Kim 已提交
1454
{
J
Jaegeuk Kim 已提交
1455 1456 1457
	ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry",
			sizeof(struct ino_entry));
	if (!ino_entry_slab)
J
Jaegeuk Kim 已提交
1458
		return -ENOMEM;
1459 1460
	inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry",
			sizeof(struct inode_entry));
1461
	if (!inode_entry_slab) {
J
Jaegeuk Kim 已提交
1462
		kmem_cache_destroy(ino_entry_slab);
J
Jaegeuk Kim 已提交
1463 1464 1465 1466 1467 1468 1469
		return -ENOMEM;
	}
	return 0;
}

void destroy_checkpoint_caches(void)
{
J
Jaegeuk Kim 已提交
1470
	kmem_cache_destroy(ino_entry_slab);
J
Jaegeuk Kim 已提交
1471 1472
	kmem_cache_destroy(inode_entry_slab);
}