data.c 27.6 KB
Newer Older
J
Jaegeuk Kim 已提交
1
/*
2 3 4 5 6 7 8 9 10 11 12 13 14
 * fs/f2fs/data.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/f2fs_fs.h>
#include <linux/buffer_head.h>
#include <linux/mpage.h>
15
#include <linux/aio.h>
16 17 18 19
#include <linux/writeback.h>
#include <linux/backing-dev.h>
#include <linux/blkdev.h>
#include <linux/bio.h>
20
#include <linux/prefetch.h>
21 22 23 24

#include "f2fs.h"
#include "node.h"
#include "segment.h"
25
#include <trace/events/f2fs.h>
26

27 28
static void f2fs_read_end_io(struct bio *bio, int err)
{
29 30
	struct bio_vec *bvec;
	int i;
31

32
	bio_for_each_segment_all(bvec, bio, i) {
33 34
		struct page *page = bvec->bv_page;

35 36 37
		if (!err) {
			SetPageUptodate(page);
		} else {
38 39 40 41
			ClearPageUptodate(page);
			SetPageError(page);
		}
		unlock_page(page);
42
	}
43 44 45 46 47
	bio_put(bio);
}

static void f2fs_write_end_io(struct bio *bio, int err)
{
48
	struct f2fs_sb_info *sbi = bio->bi_private;
49 50
	struct bio_vec *bvec;
	int i;
51

52
	bio_for_each_segment_all(bvec, bio, i) {
53 54
		struct page *page = bvec->bv_page;

55
		if (unlikely(err)) {
56
			set_page_dirty(page);
57
			set_bit(AS_EIO, &page->mapping->flags);
58
			f2fs_stop_checkpoint(sbi);
59 60 61
		}
		end_page_writeback(page);
		dec_page_count(sbi, F2FS_WRITEBACK);
62
	}
63 64 65 66 67 68 69 70

	if (!get_pages(sbi, F2FS_WRITEBACK) &&
			!list_empty(&sbi->cp_wait.task_list))
		wake_up(&sbi->cp_wait);

	bio_put(bio);
}

71 72 73 74 75 76 77 78 79 80 81 82
/*
 * Low-level block read/write IO operations.
 */
static struct bio *__bio_alloc(struct f2fs_sb_info *sbi, block_t blk_addr,
				int npages, bool is_read)
{
	struct bio *bio;

	/* No failure on bio allocation */
	bio = bio_alloc(GFP_NOIO, npages);

	bio->bi_bdev = sbi->sb->s_bdev;
C
Chao Yu 已提交
83
	bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr);
84
	bio->bi_end_io = is_read ? f2fs_read_end_io : f2fs_write_end_io;
85
	bio->bi_private = sbi;
86 87 88 89

	return bio;
}

J
Jaegeuk Kim 已提交
90
static void __submit_merged_bio(struct f2fs_bio_info *io)
91
{
J
Jaegeuk Kim 已提交
92
	struct f2fs_io_info *fio = &io->fio;
93 94 95 96

	if (!io->bio)
		return;

97 98 99 100 101 102
	if (is_read_io(fio->rw))
		trace_f2fs_submit_read_bio(io->sbi->sb, fio->rw,
							fio->type, io->bio);
	else
		trace_f2fs_submit_write_bio(io->sbi->sb, fio->rw,
							fio->type, io->bio);
103

104
	submit_bio(fio->rw, io->bio);
105 106 107 108
	io->bio = NULL;
}

void f2fs_submit_merged_bio(struct f2fs_sb_info *sbi,
J
Jaegeuk Kim 已提交
109
				enum page_type type, int rw)
110 111 112 113 114 115
{
	enum page_type btype = PAGE_TYPE_OF_BIO(type);
	struct f2fs_bio_info *io;

	io = is_read_io(rw) ? &sbi->read_io : &sbi->write_io[btype];

116
	down_write(&io->io_rwsem);
J
Jaegeuk Kim 已提交
117 118 119 120

	/* change META to META_FLUSH in the checkpoint procedure */
	if (type >= META_FLUSH) {
		io->fio.type = META_FLUSH;
J
Jaegeuk Kim 已提交
121 122 123 124
		if (test_opt(sbi, NOBARRIER))
			io->fio.rw = WRITE_FLUSH | REQ_META | REQ_PRIO;
		else
			io->fio.rw = WRITE_FLUSH_FUA | REQ_META | REQ_PRIO;
J
Jaegeuk Kim 已提交
125 126
	}
	__submit_merged_bio(io);
127
	up_write(&io->io_rwsem);
128 129 130 131 132 133 134 135 136 137 138 139 140 141
}

/*
 * Fill the locked page with data located in the block address.
 * Return unlocked page.
 */
int f2fs_submit_page_bio(struct f2fs_sb_info *sbi, struct page *page,
					block_t blk_addr, int rw)
{
	struct bio *bio;

	trace_f2fs_submit_page_bio(page, blk_addr, rw);

	/* Allocate a new bio */
142
	bio = __bio_alloc(sbi, blk_addr, 1, is_read_io(rw));
143 144 145 146 147 148 149 150 151 152 153 154

	if (bio_add_page(bio, page, PAGE_CACHE_SIZE, 0) < PAGE_CACHE_SIZE) {
		bio_put(bio);
		f2fs_put_page(page, 1);
		return -EFAULT;
	}

	submit_bio(rw, bio);
	return 0;
}

void f2fs_submit_page_mbio(struct f2fs_sb_info *sbi, struct page *page,
J
Jaegeuk Kim 已提交
155
			block_t blk_addr, struct f2fs_io_info *fio)
156
{
J
Jaegeuk Kim 已提交
157
	enum page_type btype = PAGE_TYPE_OF_BIO(fio->type);
158
	struct f2fs_bio_info *io;
159
	bool is_read = is_read_io(fio->rw);
160

161
	io = is_read ? &sbi->read_io : &sbi->write_io[btype];
162 163 164

	verify_block_addr(sbi, blk_addr);

165
	down_write(&io->io_rwsem);
166

167
	if (!is_read)
168 169
		inc_page_count(sbi, F2FS_WRITEBACK);

170
	if (io->bio && (io->last_block_in_bio != blk_addr - 1 ||
J
Jaegeuk Kim 已提交
171 172
						io->fio.rw != fio->rw))
		__submit_merged_bio(io);
173 174
alloc_new:
	if (io->bio == NULL) {
J
Jaegeuk Kim 已提交
175
		int bio_blocks = MAX_BIO_BLOCKS(sbi);
176 177

		io->bio = __bio_alloc(sbi, blk_addr, bio_blocks, is_read);
J
Jaegeuk Kim 已提交
178
		io->fio = *fio;
179 180 181 182
	}

	if (bio_add_page(io->bio, page, PAGE_CACHE_SIZE, 0) <
							PAGE_CACHE_SIZE) {
J
Jaegeuk Kim 已提交
183
		__submit_merged_bio(io);
184 185 186 187 188
		goto alloc_new;
	}

	io->last_block_in_bio = blk_addr;

189
	up_write(&io->io_rwsem);
J
Jaegeuk Kim 已提交
190
	trace_f2fs_submit_page_mbio(page, fio->rw, fio->type, blk_addr);
191 192
}

J
Jaegeuk Kim 已提交
193
/*
194 195 196 197 198 199 200 201 202 203 204 205
 * Lock ordering for the change of data block address:
 * ->data_page
 *  ->node_page
 *    update block addresses in the node page
 */
static void __set_data_blkaddr(struct dnode_of_data *dn, block_t new_addr)
{
	struct f2fs_node *rn;
	__le32 *addr_array;
	struct page *node_page = dn->node_page;
	unsigned int ofs_in_node = dn->ofs_in_node;

206
	f2fs_wait_on_page_writeback(node_page, NODE);
207

208
	rn = F2FS_NODE(node_page);
209 210 211 212 213 214 215 216 217

	/* Get physical address of data block */
	addr_array = blkaddr_in_node(rn);
	addr_array[ofs_in_node] = cpu_to_le32(new_addr);
	set_page_dirty(node_page);
}

int reserve_new_block(struct dnode_of_data *dn)
{
218
	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
219

220
	if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC)))
221
		return -EPERM;
222
	if (unlikely(!inc_valid_block_count(sbi, dn->inode, 1)))
223 224
		return -ENOSPC;

225 226
	trace_f2fs_reserve_new_block(dn->inode, dn->nid, dn->ofs_in_node);

227 228
	__set_data_blkaddr(dn, NEW_ADDR);
	dn->data_blkaddr = NEW_ADDR;
229
	mark_inode_dirty(dn->inode);
230 231 232 233
	sync_inode_page(dn);
	return 0;
}

234 235 236 237 238 239 240 241
int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
{
	bool need_put = dn->inode_page ? false : true;
	int err;

	err = get_dnode_of_data(dn, index, ALLOC_NODE);
	if (err)
		return err;
242

243 244
	if (dn->data_blkaddr == NULL_ADDR)
		err = reserve_new_block(dn);
245
	if (err || need_put)
246 247 248 249
		f2fs_put_dnode(dn);
	return err;
}

250 251 252 253 254 255 256
static int check_extent_cache(struct inode *inode, pgoff_t pgofs,
					struct buffer_head *bh_result)
{
	struct f2fs_inode_info *fi = F2FS_I(inode);
	pgoff_t start_fofs, end_fofs;
	block_t start_blkaddr;

257 258 259
	if (is_inode_flag_set(fi, FI_NO_EXTENT))
		return 0;

260 261 262 263 264 265
	read_lock(&fi->ext.ext_lock);
	if (fi->ext.len == 0) {
		read_unlock(&fi->ext.ext_lock);
		return 0;
	}

266 267
	stat_inc_total_hit(inode->i_sb);

268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284
	start_fofs = fi->ext.fofs;
	end_fofs = fi->ext.fofs + fi->ext.len - 1;
	start_blkaddr = fi->ext.blk_addr;

	if (pgofs >= start_fofs && pgofs <= end_fofs) {
		unsigned int blkbits = inode->i_sb->s_blocksize_bits;
		size_t count;

		clear_buffer_new(bh_result);
		map_bh(bh_result, inode->i_sb,
				start_blkaddr + pgofs - start_fofs);
		count = end_fofs - pgofs + 1;
		if (count < (UINT_MAX >> blkbits))
			bh_result->b_size = (count << blkbits);
		else
			bh_result->b_size = UINT_MAX;

285
		stat_inc_read_hit(inode->i_sb);
286 287 288 289 290 291 292 293 294 295 296 297
		read_unlock(&fi->ext.ext_lock);
		return 1;
	}
	read_unlock(&fi->ext.ext_lock);
	return 0;
}

void update_extent_cache(block_t blk_addr, struct dnode_of_data *dn)
{
	struct f2fs_inode_info *fi = F2FS_I(dn->inode);
	pgoff_t fofs, start_fofs, end_fofs;
	block_t start_blkaddr, end_blkaddr;
298
	int need_update = true;
299

300
	f2fs_bug_on(F2FS_I_SB(dn->inode), blk_addr == NEW_ADDR);
301 302
	fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) +
							dn->ofs_in_node;
303 304 305 306

	/* Update the page address in the parent node */
	__set_data_blkaddr(dn, blk_addr);

307 308 309
	if (is_inode_flag_set(fi, FI_NO_EXTENT))
		return;

310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330
	write_lock(&fi->ext.ext_lock);

	start_fofs = fi->ext.fofs;
	end_fofs = fi->ext.fofs + fi->ext.len - 1;
	start_blkaddr = fi->ext.blk_addr;
	end_blkaddr = fi->ext.blk_addr + fi->ext.len - 1;

	/* Drop and initialize the matched extent */
	if (fi->ext.len == 1 && fofs == start_fofs)
		fi->ext.len = 0;

	/* Initial extent */
	if (fi->ext.len == 0) {
		if (blk_addr != NULL_ADDR) {
			fi->ext.fofs = fofs;
			fi->ext.blk_addr = blk_addr;
			fi->ext.len = 1;
		}
		goto end_update;
	}

N
Namjae Jeon 已提交
331
	/* Front merge */
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355
	if (fofs == start_fofs - 1 && blk_addr == start_blkaddr - 1) {
		fi->ext.fofs--;
		fi->ext.blk_addr--;
		fi->ext.len++;
		goto end_update;
	}

	/* Back merge */
	if (fofs == end_fofs + 1 && blk_addr == end_blkaddr + 1) {
		fi->ext.len++;
		goto end_update;
	}

	/* Split the existing extent */
	if (fi->ext.len > 1 &&
		fofs >= start_fofs && fofs <= end_fofs) {
		if ((end_fofs - fofs) < (fi->ext.len >> 1)) {
			fi->ext.len = fofs - start_fofs;
		} else {
			fi->ext.fofs = fofs + 1;
			fi->ext.blk_addr = start_blkaddr +
					fofs - start_fofs + 1;
			fi->ext.len -= fofs - start_fofs + 1;
		}
356 357
	} else {
		need_update = false;
358 359
	}

360 361 362 363 364 365
	/* Finally, if the extent is very fragmented, let's drop the cache. */
	if (fi->ext.len < F2FS_MIN_EXTENT_LEN) {
		fi->ext.len = 0;
		set_inode_flag(fi, FI_NO_EXTENT);
		need_update = true;
	}
366 367
end_update:
	write_unlock(&fi->ext.ext_lock);
368 369 370
	if (need_update)
		sync_inode_page(dn);
	return;
371 372
}

373
struct page *find_data_page(struct inode *inode, pgoff_t index, bool sync)
374 375 376 377 378 379 380 381 382 383 384 385
{
	struct address_space *mapping = inode->i_mapping;
	struct dnode_of_data dn;
	struct page *page;
	int err;

	page = find_get_page(mapping, index);
	if (page && PageUptodate(page))
		return page;
	f2fs_put_page(page, 0);

	set_new_dnode(&dn, inode, NULL, NULL, 0);
386
	err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
387 388 389 390 391 392 393 394
	if (err)
		return ERR_PTR(err);
	f2fs_put_dnode(&dn);

	if (dn.data_blkaddr == NULL_ADDR)
		return ERR_PTR(-ENOENT);

	/* By fallocate(), there is no cached page, but with NEW_ADDR */
395
	if (unlikely(dn.data_blkaddr == NEW_ADDR))
396 397
		return ERR_PTR(-EINVAL);

398
	page = grab_cache_page(mapping, index);
399 400 401
	if (!page)
		return ERR_PTR(-ENOMEM);

402 403 404 405 406
	if (PageUptodate(page)) {
		unlock_page(page);
		return page;
	}

407
	err = f2fs_submit_page_bio(F2FS_I_SB(inode), page, dn.data_blkaddr,
408
					sync ? READ_SYNC : READA);
409 410 411
	if (err)
		return ERR_PTR(err);

412 413
	if (sync) {
		wait_on_page_locked(page);
414
		if (unlikely(!PageUptodate(page))) {
415 416 417
			f2fs_put_page(page, 0);
			return ERR_PTR(-EIO);
		}
418 419 420 421
	}
	return page;
}

J
Jaegeuk Kim 已提交
422
/*
423 424 425 426 427 428 429 430 431 432 433
 * If it tries to access a hole, return an error.
 * Because, the callers, functions in dir.c and GC, should be able to know
 * whether this page exists or not.
 */
struct page *get_lock_data_page(struct inode *inode, pgoff_t index)
{
	struct address_space *mapping = inode->i_mapping;
	struct dnode_of_data dn;
	struct page *page;
	int err;

434
repeat:
435
	page = grab_cache_page(mapping, index);
436 437 438
	if (!page)
		return ERR_PTR(-ENOMEM);

439
	set_new_dnode(&dn, inode, NULL, NULL, 0);
440
	err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
441 442
	if (err) {
		f2fs_put_page(page, 1);
443
		return ERR_PTR(err);
444
	}
445 446
	f2fs_put_dnode(&dn);

447
	if (unlikely(dn.data_blkaddr == NULL_ADDR)) {
448
		f2fs_put_page(page, 1);
449
		return ERR_PTR(-ENOENT);
450
	}
451 452 453 454

	if (PageUptodate(page))
		return page;

J
Jaegeuk Kim 已提交
455 456 457 458 459 460 461 462 463 464 465
	/*
	 * A new dentry page is allocated but not able to be written, since its
	 * new inode page couldn't be allocated due to -ENOSPC.
	 * In such the case, its blkaddr can be remained as NEW_ADDR.
	 * see, f2fs_add_link -> get_new_data_page -> init_inode_metadata.
	 */
	if (dn.data_blkaddr == NEW_ADDR) {
		zero_user_segment(page, 0, PAGE_CACHE_SIZE);
		SetPageUptodate(page);
		return page;
	}
466

467 468
	err = f2fs_submit_page_bio(F2FS_I_SB(inode), page,
					dn.data_blkaddr, READ_SYNC);
469
	if (err)
470
		return ERR_PTR(err);
471 472

	lock_page(page);
473
	if (unlikely(!PageUptodate(page))) {
474 475
		f2fs_put_page(page, 1);
		return ERR_PTR(-EIO);
476
	}
477
	if (unlikely(page->mapping != mapping)) {
478 479
		f2fs_put_page(page, 1);
		goto repeat;
480 481 482 483
	}
	return page;
}

J
Jaegeuk Kim 已提交
484
/*
485 486
 * Caller ensures that this data page is never allocated.
 * A new zero-filled data page is allocated in the page cache.
487
 *
C
Chao Yu 已提交
488 489
 * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and
 * f2fs_unlock_op().
490
 * Note that, ipage is set only by make_empty_dir.
491
 */
492
struct page *get_new_data_page(struct inode *inode,
493
		struct page *ipage, pgoff_t index, bool new_i_size)
494 495 496 497 498 499
{
	struct address_space *mapping = inode->i_mapping;
	struct page *page;
	struct dnode_of_data dn;
	int err;

500
	set_new_dnode(&dn, inode, ipage, NULL, 0);
501
	err = f2fs_reserve_block(&dn, index);
502 503
	if (err)
		return ERR_PTR(err);
504
repeat:
505
	page = grab_cache_page(mapping, index);
506 507 508 509
	if (!page) {
		err = -ENOMEM;
		goto put_err;
	}
510 511 512 513 514 515

	if (PageUptodate(page))
		return page;

	if (dn.data_blkaddr == NEW_ADDR) {
		zero_user_segment(page, 0, PAGE_CACHE_SIZE);
516
		SetPageUptodate(page);
517
	} else {
518 519
		err = f2fs_submit_page_bio(F2FS_I_SB(inode), page,
						dn.data_blkaddr, READ_SYNC);
520
		if (err)
521 522
			goto put_err;

523
		lock_page(page);
524
		if (unlikely(!PageUptodate(page))) {
525
			f2fs_put_page(page, 1);
526 527
			err = -EIO;
			goto put_err;
528
		}
529
		if (unlikely(page->mapping != mapping)) {
530 531
			f2fs_put_page(page, 1);
			goto repeat;
532 533 534 535 536 537
		}
	}

	if (new_i_size &&
		i_size_read(inode) < ((index + 1) << PAGE_CACHE_SHIFT)) {
		i_size_write(inode, ((index + 1) << PAGE_CACHE_SHIFT));
538 539
		/* Only the directory inode sets new_i_size */
		set_inode_flag(F2FS_I(inode), FI_UPDATE_DIR);
540 541
	}
	return page;
542 543 544 545

put_err:
	f2fs_put_dnode(&dn);
	return ERR_PTR(err);
546 547
}

548 549
static int __allocate_data_block(struct dnode_of_data *dn)
{
550
	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
551
	struct f2fs_inode_info *fi = F2FS_I(dn->inode);
552 553 554
	struct f2fs_summary sum;
	block_t new_blkaddr;
	struct node_info ni;
555
	pgoff_t fofs;
556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577
	int type;

	if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC)))
		return -EPERM;
	if (unlikely(!inc_valid_block_count(sbi, dn->inode, 1)))
		return -ENOSPC;

	__set_data_blkaddr(dn, NEW_ADDR);
	dn->data_blkaddr = NEW_ADDR;

	get_node_info(sbi, dn->nid, &ni);
	set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);

	type = CURSEG_WARM_DATA;

	allocate_data_block(sbi, NULL, NULL_ADDR, &new_blkaddr, &sum, type);

	/* direct IO doesn't use extent cache to maximize the performance */
	set_inode_flag(F2FS_I(dn->inode), FI_NO_EXTENT);
	update_extent_cache(new_blkaddr, dn);
	clear_inode_flag(F2FS_I(dn->inode), FI_NO_EXTENT);

578 579 580 581 582 583
	/* update i_size */
	fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) +
							dn->ofs_in_node;
	if (i_size_read(dn->inode) < ((fofs + 1) << PAGE_CACHE_SHIFT))
		i_size_write(dn->inode, ((fofs + 1) << PAGE_CACHE_SHIFT));

584 585 586 587
	dn->data_blkaddr = new_blkaddr;
	return 0;
}

J
Jaegeuk Kim 已提交
588
/*
C
Chao Yu 已提交
589 590 591 592 593 594
 * get_data_block() now supported readahead/bmap/rw direct_IO with mapped bh.
 * If original data blocks are allocated, then give them to blockdev.
 * Otherwise,
 *     a. preallocate requested block addresses
 *     b. do not use extent cache for better performance
 *     c. give the block addresses to blockdev
595
 */
596 597
static int __get_data_block(struct inode *inode, sector_t iblock,
			struct buffer_head *bh_result, int create, bool fiemap)
598 599 600 601
{
	unsigned int blkbits = inode->i_sb->s_blocksize_bits;
	unsigned maxblocks = bh_result->b_size >> blkbits;
	struct dnode_of_data dn;
602 603 604 605
	int mode = create ? ALLOC_NODE : LOOKUP_NODE_RA;
	pgoff_t pgofs, end_offset;
	int err = 0, ofs = 1;
	bool allocated = false;
606 607 608 609

	/* Get the page offset from the block offset(iblock) */
	pgofs =	(pgoff_t)(iblock >> (PAGE_CACHE_SHIFT - blkbits));

610 611 612
	if (check_extent_cache(inode, pgofs, bh_result))
		goto out;

613
	if (create) {
614 615
		f2fs_balance_fs(F2FS_I_SB(inode));
		f2fs_lock_op(F2FS_I_SB(inode));
616
	}
617 618 619

	/* When reading holes, we need its node page */
	set_new_dnode(&dn, inode, NULL, NULL, 0);
620
	err = get_dnode_of_data(&dn, pgofs, mode);
621
	if (err) {
622 623 624
		if (err == -ENOENT)
			err = 0;
		goto unlock_out;
625
	}
626
	if (dn.data_blkaddr == NEW_ADDR && !fiemap)
627
		goto put_out;
628

629 630 631 632 633 634 635 636 637 638 639 640
	if (dn.data_blkaddr != NULL_ADDR) {
		map_bh(bh_result, inode->i_sb, dn.data_blkaddr);
	} else if (create) {
		err = __allocate_data_block(&dn);
		if (err)
			goto put_out;
		allocated = true;
		map_bh(bh_result, inode->i_sb, dn.data_blkaddr);
	} else {
		goto put_out;
	}

641
	end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
642 643 644 645 646 647 648 649 650 651 652 653 654
	bh_result->b_size = (((size_t)1) << blkbits);
	dn.ofs_in_node++;
	pgofs++;

get_next:
	if (dn.ofs_in_node >= end_offset) {
		if (allocated)
			sync_inode_page(&dn);
		allocated = false;
		f2fs_put_dnode(&dn);

		set_new_dnode(&dn, inode, NULL, NULL, 0);
		err = get_dnode_of_data(&dn, pgofs, mode);
655
		if (err) {
656 657 658 659
			if (err == -ENOENT)
				err = 0;
			goto unlock_out;
		}
660
		if (dn.data_blkaddr == NEW_ADDR && !fiemap)
661 662
			goto put_out;

663
		end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
664
	}
665

666 667 668 669 670 671 672 673 674
	if (maxblocks > (bh_result->b_size >> blkbits)) {
		block_t blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
		if (blkaddr == NULL_ADDR && create) {
			err = __allocate_data_block(&dn);
			if (err)
				goto sync_out;
			allocated = true;
			blkaddr = dn.data_blkaddr;
		}
A
arter97 已提交
675
		/* Give more consecutive addresses for the readahead */
676 677 678 679 680 681 682
		if (blkaddr == (bh_result->b_blocknr + ofs)) {
			ofs++;
			dn.ofs_in_node++;
			pgofs++;
			bh_result->b_size += (((size_t)1) << blkbits);
			goto get_next;
		}
683
	}
684 685 686 687
sync_out:
	if (allocated)
		sync_inode_page(&dn);
put_out:
688
	f2fs_put_dnode(&dn);
689 690
unlock_out:
	if (create)
691
		f2fs_unlock_op(F2FS_I_SB(inode));
692 693 694
out:
	trace_f2fs_get_data_block(inode, iblock, bh_result, err);
	return err;
695 696
}

697 698 699 700 701 702 703 704 705 706 707 708
static int get_data_block(struct inode *inode, sector_t iblock,
			struct buffer_head *bh_result, int create)
{
	return __get_data_block(inode, iblock, bh_result, create, false);
}

static int get_data_block_fiemap(struct inode *inode, sector_t iblock,
			struct buffer_head *bh_result, int create)
{
	return __get_data_block(inode, iblock, bh_result, create, true);
}

J
Jaegeuk Kim 已提交
709 710 711
int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
		u64 start, u64 len)
{
712 713
	return generic_block_fiemap(inode, fieinfo,
				start, len, get_data_block_fiemap);
J
Jaegeuk Kim 已提交
714 715
}

716 717
static int f2fs_read_data_page(struct file *file, struct page *page)
{
H
Huajun Li 已提交
718
	struct inode *inode = page->mapping->host;
719
	int ret = -EAGAIN;
H
Huajun Li 已提交
720

721 722
	trace_f2fs_readpage(page, DATA);

A
arter97 已提交
723
	/* If the file has inline data, try to read it directly */
H
Huajun Li 已提交
724 725
	if (f2fs_has_inline_data(inode))
		ret = f2fs_read_inline_data(inode, page);
726
	if (ret == -EAGAIN)
H
Huajun Li 已提交
727 728 729
		ret = mpage_readpage(page, get_data_block);

	return ret;
730 731 732 733 734 735
}

static int f2fs_read_data_pages(struct file *file,
			struct address_space *mapping,
			struct list_head *pages, unsigned nr_pages)
{
H
Huajun Li 已提交
736 737 738 739 740 741
	struct inode *inode = file->f_mapping->host;

	/* If the file has inline data, skip readpages */
	if (f2fs_has_inline_data(inode))
		return 0;

742
	return mpage_readpages(mapping, pages, nr_pages, get_data_block);
743 744
}

J
Jaegeuk Kim 已提交
745
int do_write_data_page(struct page *page, struct f2fs_io_info *fio)
746 747
{
	struct inode *inode = page->mapping->host;
J
Jaegeuk Kim 已提交
748
	block_t old_blkaddr, new_blkaddr;
749 750 751 752
	struct dnode_of_data dn;
	int err = 0;

	set_new_dnode(&dn, inode, NULL, NULL, 0);
753
	err = get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
754 755 756
	if (err)
		return err;

J
Jaegeuk Kim 已提交
757
	old_blkaddr = dn.data_blkaddr;
758 759

	/* This page is already truncated */
J
Jaegeuk Kim 已提交
760
	if (old_blkaddr == NULL_ADDR)
761 762 763 764 765 766 767 768
		goto out_writepage;

	set_page_writeback(page);

	/*
	 * If current allocation needs SSR,
	 * it had better in-place writes for updated data.
	 */
J
Jaegeuk Kim 已提交
769
	if (unlikely(old_blkaddr != NEW_ADDR &&
770 771
			!is_cold_data(page) &&
			need_inplace_update(inode))) {
J
Jaegeuk Kim 已提交
772
		rewrite_data_page(page, old_blkaddr, fio);
773
		set_inode_flag(F2FS_I(inode), FI_UPDATE_WRITE);
774
	} else {
J
Jaegeuk Kim 已提交
775 776
		write_data_page(page, &dn, &new_blkaddr, fio);
		update_extent_cache(new_blkaddr, &dn);
777
		set_inode_flag(F2FS_I(inode), FI_APPEND_WRITE);
778 779 780 781 782 783 784 785 786 787
	}
out_writepage:
	f2fs_put_dnode(&dn);
	return err;
}

static int f2fs_write_data_page(struct page *page,
					struct writeback_control *wbc)
{
	struct inode *inode = page->mapping->host;
788
	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
789 790 791
	loff_t i_size = i_size_read(inode);
	const pgoff_t end_index = ((unsigned long long) i_size)
							>> PAGE_CACHE_SHIFT;
H
Huajun Li 已提交
792
	unsigned offset = 0;
793
	bool need_balance_fs = false;
794
	int err = 0;
J
Jaegeuk Kim 已提交
795 796
	struct f2fs_io_info fio = {
		.type = DATA,
C
Chris Fries 已提交
797
		.rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE,
J
Jaegeuk Kim 已提交
798
	};
799

800 801
	trace_f2fs_writepage(page, DATA);

802
	if (page->index < end_index)
803
		goto write;
804 805 806 807 808 809

	/*
	 * If the offset is out-of-range of file size,
	 * this page does not have to be written to disk.
	 */
	offset = i_size & (PAGE_CACHE_SIZE - 1);
810
	if ((page->index >= end_index + 1) || !offset)
811
		goto out;
812 813

	zero_user_segment(page, offset, PAGE_CACHE_SIZE);
814
write:
815
	if (unlikely(sbi->por_doing))
816 817
		goto redirty_out;

818
	/* Dentry blocks are controlled by checkpoint */
819
	if (S_ISDIR(inode->i_mode)) {
820 821
		if (unlikely(f2fs_cp_error(sbi)))
			goto redirty_out;
J
Jaegeuk Kim 已提交
822
		err = do_write_data_page(page, &fio);
823 824
		goto done;
	}
H
Huajun Li 已提交
825

826 827 828 829
	/* we should bypass data pages to proceed the kworkder jobs */
	if (unlikely(f2fs_cp_error(sbi))) {
		SetPageError(page);
		unlock_page(page);
830
		goto out;
831 832
	}

833
	if (!wbc->for_reclaim)
834
		need_balance_fs = true;
835
	else if (has_not_enough_free_secs(sbi, 0))
836
		goto redirty_out;
837

838
	err = -EAGAIN;
839
	f2fs_lock_op(sbi);
840 841 842
	if (f2fs_has_inline_data(inode))
		err = f2fs_write_inline_data(inode, page);
	if (err == -EAGAIN)
843 844 845 846 847
		err = do_write_data_page(page, &fio);
	f2fs_unlock_op(sbi);
done:
	if (err && err != -ENOENT)
		goto redirty_out;
848 849

	clear_cold_data(page);
850
out:
851
	inode_dec_dirty_pages(inode);
852
	unlock_page(page);
853
	if (need_balance_fs)
854
		f2fs_balance_fs(sbi);
855 856
	if (wbc->for_reclaim)
		f2fs_submit_merged_bio(sbi, DATA, WRITE);
857 858 859
	return 0;

redirty_out:
860
	redirty_page_for_writepage(wbc, page);
861
	return AOP_WRITEPAGE_ACTIVATE;
862 863
}

864 865 866 867 868 869 870 871 872
static int __f2fs_writepage(struct page *page, struct writeback_control *wbc,
			void *data)
{
	struct address_space *mapping = data;
	int ret = mapping->a_ops->writepage(page, wbc);
	mapping_set_error(mapping, ret);
	return ret;
}

873
static int f2fs_write_data_pages(struct address_space *mapping,
874 875 876
			    struct writeback_control *wbc)
{
	struct inode *inode = mapping->host;
877
	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
878
	bool locked = false;
879
	int ret;
880
	long diff;
881

882 883
	trace_f2fs_writepages(mapping->host, wbc, DATA);

P
P J P 已提交
884 885 886 887
	/* deal with chardevs and other special file */
	if (!mapping->a_ops->writepage)
		return 0;

888
	if (S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_NONE &&
889
			get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) &&
890
			available_free_memory(sbi, DIRTY_DENTS))
891
		goto skip_write;
892

893
	diff = nr_pages_to_write(sbi, DATA, wbc);
894

895
	if (!S_ISDIR(inode->i_mode)) {
896
		mutex_lock(&sbi->writepages);
897 898
		locked = true;
	}
899
	ret = write_cache_pages(mapping, wbc, __f2fs_writepage, mapping);
900
	if (locked)
901
		mutex_unlock(&sbi->writepages);
J
Jaegeuk Kim 已提交
902 903

	f2fs_submit_merged_bio(sbi, DATA, WRITE);
904 905 906

	remove_dirty_dir_inode(inode);

907
	wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff);
908
	return ret;
909 910

skip_write:
911
	wbc->pages_skipped += get_dirty_pages(inode);
912
	return 0;
913 914
}

915 916 917 918 919 920
static void f2fs_write_failed(struct address_space *mapping, loff_t to)
{
	struct inode *inode = mapping->host;

	if (to > inode->i_size) {
		truncate_pagecache(inode, inode->i_size);
921
		truncate_blocks(inode, inode->i_size, true);
922 923 924
	}
}

925 926 927 928 929
static int f2fs_write_begin(struct file *file, struct address_space *mapping,
		loff_t pos, unsigned len, unsigned flags,
		struct page **pagep, void **fsdata)
{
	struct inode *inode = mapping->host;
930
	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
931
	struct page *page, *ipage;
932 933 934 935
	pgoff_t index = ((unsigned long long) pos) >> PAGE_CACHE_SHIFT;
	struct dnode_of_data dn;
	int err = 0;

936 937
	trace_f2fs_write_begin(inode, pos, len, flags);

938
	f2fs_balance_fs(sbi);
939 940 941 942 943 944 945 946 947 948 949

	/*
	 * We should check this at this moment to avoid deadlock on inode page
	 * and #0 page. The locking rule for inline_data conversion should be:
	 * lock_page(page #0) -> lock_page(inode_page)
	 */
	if (index != 0) {
		err = f2fs_convert_inline_inode(inode);
		if (err)
			goto fail;
	}
950
repeat:
951
	page = grab_cache_page_write_begin(mapping, index, flags);
952 953 954 955
	if (!page) {
		err = -ENOMEM;
		goto fail;
	}
956

957 958
	*pagep = page;

959
	f2fs_lock_op(sbi);
960 961 962

	/* check inline_data */
	ipage = get_node_page(sbi, inode->i_ino);
963 964
	if (IS_ERR(ipage)) {
		err = PTR_ERR(ipage);
965
		goto unlock_fail;
966
	}
967

968 969
	set_new_dnode(&dn, inode, ipage, ipage, 0);

970
	if (f2fs_has_inline_data(inode)) {
971 972 973 974 975 976
		if (pos + len <= MAX_INLINE_DATA) {
			read_inline_data(page, ipage);
			set_inode_flag(F2FS_I(inode), FI_DATA_EXIST);
			sync_inode_page(&dn);
			goto put_next;
		}
977 978 979
		err = f2fs_convert_inline_page(&dn, page);
		if (err)
			goto put_fail;
980
	}
981 982
	err = f2fs_reserve_block(&dn, index);
	if (err)
983
		goto put_fail;
984
put_next:
985 986 987
	f2fs_put_dnode(&dn);
	f2fs_unlock_op(sbi);

988 989 990
	if ((len == PAGE_CACHE_SIZE) || PageUptodate(page))
		return 0;

991 992
	f2fs_wait_on_page_writeback(page, DATA);

993 994 995 996 997 998
	if ((pos & PAGE_CACHE_MASK) >= i_size_read(inode)) {
		unsigned start = pos & (PAGE_CACHE_SIZE - 1);
		unsigned end = start + len;

		/* Reading beyond i_size is simple: memset to zero */
		zero_user_segments(page, 0, start, end, PAGE_CACHE_SIZE);
999
		goto out;
1000 1001
	}

1002
	if (dn.data_blkaddr == NEW_ADDR) {
1003 1004
		zero_user_segment(page, 0, PAGE_CACHE_SIZE);
	} else {
1005 1006 1007 1008
		err = f2fs_submit_page_bio(sbi, page, dn.data_blkaddr,
					   READ_SYNC);
		if (err)
			goto fail;
1009

1010
		lock_page(page);
1011
		if (unlikely(!PageUptodate(page))) {
1012
			f2fs_put_page(page, 1);
1013 1014
			err = -EIO;
			goto fail;
1015
		}
1016
		if (unlikely(page->mapping != mapping)) {
1017 1018
			f2fs_put_page(page, 1);
			goto repeat;
1019 1020
		}
	}
1021
out:
1022 1023 1024
	SetPageUptodate(page);
	clear_cold_data(page);
	return 0;
1025

1026 1027
put_fail:
	f2fs_put_dnode(&dn);
1028 1029
unlock_fail:
	f2fs_unlock_op(sbi);
1030
	f2fs_put_page(page, 1);
1031 1032 1033
fail:
	f2fs_write_failed(mapping, pos + len);
	return err;
1034 1035
}

1036 1037 1038 1039 1040 1041 1042
static int f2fs_write_end(struct file *file,
			struct address_space *mapping,
			loff_t pos, unsigned len, unsigned copied,
			struct page *page, void *fsdata)
{
	struct inode *inode = page->mapping->host;

1043 1044
	trace_f2fs_write_end(inode, pos, len, copied);

1045
	set_page_dirty(page);
1046 1047 1048 1049 1050 1051 1052

	if (pos + copied > i_size_read(inode)) {
		i_size_write(inode, pos + copied);
		mark_inode_dirty(inode);
		update_inode_page(inode);
	}

1053
	f2fs_put_page(page, 1);
1054 1055 1056
	return copied;
}

1057
static int check_direct_IO(struct inode *inode, int rw,
A
Al Viro 已提交
1058
		struct iov_iter *iter, loff_t offset)
1059 1060 1061 1062 1063 1064 1065 1066 1067
{
	unsigned blocksize_mask = inode->i_sb->s_blocksize - 1;

	if (rw == READ)
		return 0;

	if (offset & blocksize_mask)
		return -EINVAL;

A
Al Viro 已提交
1068 1069 1070
	if (iov_iter_alignment(iter) & blocksize_mask)
		return -EINVAL;

1071 1072 1073
	return 0;
}

1074
static ssize_t f2fs_direct_IO(int rw, struct kiocb *iocb,
A
Al Viro 已提交
1075
		struct iov_iter *iter, loff_t offset)
1076 1077
{
	struct file *file = iocb->ki_filp;
1078 1079 1080 1081
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
	size_t count = iov_iter_count(iter);
	int err;
1082

1083 1084 1085 1086 1087 1088
	/* we don't need to use inline_data strictly */
	if (f2fs_has_inline_data(inode)) {
		err = f2fs_convert_inline_inode(inode);
		if (err)
			return err;
	}
H
Huajun Li 已提交
1089

A
Al Viro 已提交
1090
	if (check_direct_IO(inode, rw, iter, offset))
1091 1092
		return 0;

1093 1094
	trace_f2fs_direct_IO_enter(inode, offset, count, rw);

1095 1096 1097
	err = blockdev_direct_IO(rw, iocb, inode, iter, offset, get_data_block);
	if (err < 0 && (rw & WRITE))
		f2fs_write_failed(mapping, offset + count);
1098 1099 1100

	trace_f2fs_direct_IO_exit(inode, offset, count, rw, err);

1101
	return err;
1102 1103
}

1104 1105
static void f2fs_invalidate_data_page(struct page *page, unsigned int offset,
				      unsigned int length)
1106 1107
{
	struct inode *inode = page->mapping->host;
1108 1109 1110 1111

	if (offset % PAGE_CACHE_SIZE || length != PAGE_CACHE_SIZE)
		return;

J
Jaegeuk Kim 已提交
1112 1113 1114
	if (f2fs_is_atomic_file(inode) || f2fs_is_volatile_file(inode))
		invalidate_inmem_page(inode, page);

1115
	if (PageDirty(page))
1116
		inode_dec_dirty_pages(inode);
1117 1118 1119 1120 1121 1122
	ClearPagePrivate(page);
}

static int f2fs_release_data_page(struct page *page, gfp_t wait)
{
	ClearPagePrivate(page);
1123
	return 1;
1124 1125 1126 1127 1128 1129 1130
}

static int f2fs_set_data_page_dirty(struct page *page)
{
	struct address_space *mapping = page->mapping;
	struct inode *inode = mapping->host;

1131 1132
	trace_f2fs_set_page_dirty(page, DATA);

1133
	SetPageUptodate(page);
1134 1135 1136 1137 1138 1139

	if (f2fs_is_atomic_file(inode) || f2fs_is_volatile_file(inode)) {
		register_inmem_page(inode, page);
		return 1;
	}

1140 1141
	mark_inode_dirty(inode);

1142 1143
	if (!PageDirty(page)) {
		__set_page_dirty_nobuffers(page);
1144
		update_dirty_page(inode, page);
1145 1146 1147 1148 1149
		return 1;
	}
	return 0;
}

J
Jaegeuk Kim 已提交
1150 1151
static sector_t f2fs_bmap(struct address_space *mapping, sector_t block)
{
1152 1153
	struct inode *inode = mapping->host;

1154 1155 1156 1157 1158 1159
	/* we don't need to use inline_data strictly */
	if (f2fs_has_inline_data(inode)) {
		int err = f2fs_convert_inline_inode(inode);
		if (err)
			return err;
	}
1160
	return generic_block_bmap(mapping, block, get_data_block);
J
Jaegeuk Kim 已提交
1161 1162
}

1163 1164 1165 1166 1167 1168
const struct address_space_operations f2fs_dblock_aops = {
	.readpage	= f2fs_read_data_page,
	.readpages	= f2fs_read_data_pages,
	.writepage	= f2fs_write_data_page,
	.writepages	= f2fs_write_data_pages,
	.write_begin	= f2fs_write_begin,
1169
	.write_end	= f2fs_write_end,
1170 1171 1172 1173
	.set_page_dirty	= f2fs_set_data_page_dirty,
	.invalidatepage	= f2fs_invalidate_data_page,
	.releasepage	= f2fs_release_data_page,
	.direct_IO	= f2fs_direct_IO,
J
Jaegeuk Kim 已提交
1174
	.bmap		= f2fs_bmap,
1175
};