data.c 29.1 KB
Newer Older
J
Jaegeuk Kim 已提交
1
/*
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
 * 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>
#include <linux/writeback.h>
#include <linux/backing-dev.h>
#include <linux/blkdev.h>
#include <linux/bio.h>
19
#include <linux/prefetch.h>
20
#include <linux/uio.h>
21 22 23 24

#include "f2fs.h"
#include "node.h"
#include "segment.h"
J
Jaegeuk Kim 已提交
25
#include "trace.h"
26
#include <trace/events/f2fs.h>
27

28 29
static void f2fs_read_end_io(struct bio *bio, int err)
{
30 31
	struct bio_vec *bvec;
	int i;
32

33
	bio_for_each_segment_all(bvec, bio, i) {
34 35
		struct page *page = bvec->bv_page;

36 37 38
		if (!err) {
			SetPageUptodate(page);
		} else {
39 40 41 42
			ClearPageUptodate(page);
			SetPageError(page);
		}
		unlock_page(page);
43
	}
44 45 46 47 48
	bio_put(bio);
}

static void f2fs_write_end_io(struct bio *bio, int err)
{
49
	struct f2fs_sb_info *sbi = bio->bi_private;
50 51
	struct bio_vec *bvec;
	int i;
52

53
	bio_for_each_segment_all(bvec, bio, i) {
54 55
		struct page *page = bvec->bv_page;

56
		if (unlikely(err)) {
57
			set_page_dirty(page);
58
			set_bit(AS_EIO, &page->mapping->flags);
59
			f2fs_stop_checkpoint(sbi);
60 61 62
		}
		end_page_writeback(page);
		dec_page_count(sbi, F2FS_WRITEBACK);
63
	}
64 65 66 67 68 69 70 71

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

	bio_put(bio);
}

72 73 74 75 76 77 78 79 80 81 82 83
/*
 * 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
84
	bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr);
85
	bio->bi_end_io = is_read ? f2fs_read_end_io : f2fs_write_end_io;
86
	bio->bi_private = sbi;
87 88 89 90

	return bio;
}

J
Jaegeuk Kim 已提交
91
static void __submit_merged_bio(struct f2fs_bio_info *io)
92
{
J
Jaegeuk Kim 已提交
93
	struct f2fs_io_info *fio = &io->fio;
94 95 96 97

	if (!io->bio)
		return;

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

103
	submit_bio(fio->rw, io->bio);
104 105 106 107
	io->bio = NULL;
}

void f2fs_submit_merged_bio(struct f2fs_sb_info *sbi,
J
Jaegeuk Kim 已提交
108
				enum page_type type, int rw)
109 110 111 112 113 114
{
	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];

115
	down_write(&io->io_rwsem);
J
Jaegeuk Kim 已提交
116 117 118 119

	/* change META to META_FLUSH in the checkpoint procedure */
	if (type >= META_FLUSH) {
		io->fio.type = META_FLUSH;
J
Jaegeuk Kim 已提交
120 121 122 123
		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
124 125
	}
	__submit_merged_bio(io);
126
	up_write(&io->io_rwsem);
127 128 129 130 131 132 133
}

/*
 * 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,
134
					struct f2fs_io_info *fio)
135 136 137
{
	struct bio *bio;

138
	trace_f2fs_submit_page_bio(page, fio);
J
Jaegeuk Kim 已提交
139
	f2fs_trace_ios(page, fio, 0);
140 141

	/* Allocate a new bio */
142
	bio = __bio_alloc(sbi, fio->blk_addr, 1, is_read_io(fio->rw));
143 144 145 146 147 148 149

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

150
	submit_bio(fio->rw, bio);
151 152 153 154
	return 0;
}

void f2fs_submit_page_mbio(struct f2fs_sb_info *sbi, struct page *page,
155
					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
	verify_block_addr(sbi, fio->blk_addr);
164

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 != fio->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, fio->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
		goto alloc_new;
	}

187
	io->last_block_in_bio = fio->blk_addr;
J
Jaegeuk Kim 已提交
188
	f2fs_trace_ios(page, fio, 0);
189

190
	up_write(&io->io_rwsem);
191
	trace_f2fs_submit_page_mbio(page, fio);
192 193
}

J
Jaegeuk Kim 已提交
194
/*
195 196 197 198 199
 * Lock ordering for the change of data block address:
 * ->data_page
 *  ->node_page
 *    update block addresses in the node page
 */
J
Jaegeuk Kim 已提交
200
static void __set_data_blkaddr(struct dnode_of_data *dn)
201 202 203 204 205 206
{
	struct f2fs_node *rn;
	__le32 *addr_array;
	struct page *node_page = dn->node_page;
	unsigned int ofs_in_node = dn->ofs_in_node;

207
	f2fs_wait_on_page_writeback(node_page, NODE);
208

209
	rn = F2FS_NODE(node_page);
210 211 212

	/* Get physical address of data block */
	addr_array = blkaddr_in_node(rn);
J
Jaegeuk Kim 已提交
213
	addr_array[ofs_in_node] = cpu_to_le32(dn->data_blkaddr);
214 215 216 217 218
	set_page_dirty(node_page);
}

int reserve_new_block(struct dnode_of_data *dn)
{
219
	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
220

221
	if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC)))
222
		return -EPERM;
223
	if (unlikely(!inc_valid_block_count(sbi, dn->inode, 1)))
224 225
		return -ENOSPC;

226 227
	trace_f2fs_reserve_new_block(dn->inode, dn->nid, dn->ofs_in_node);

228
	dn->data_blkaddr = NEW_ADDR;
J
Jaegeuk Kim 已提交
229
	__set_data_blkaddr(dn);
230
	mark_inode_dirty(dn->inode);
231 232 233 234
	sync_inode_page(dn);
	return 0;
}

235 236 237 238 239 240 241 242
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;
243

244 245
	if (dn->data_blkaddr == NULL_ADDR)
		err = reserve_new_block(dn);
246
	if (err || need_put)
247 248 249 250
		f2fs_put_dnode(dn);
	return err;
}

251 252 253 254 255 256 257
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;

258 259 260
	if (is_inode_flag_set(fi, FI_NO_EXTENT))
		return 0;

261 262 263 264 265 266
	read_lock(&fi->ext.ext_lock);
	if (fi->ext.len == 0) {
		read_unlock(&fi->ext.ext_lock);
		return 0;
	}

267 268
	stat_inc_total_hit(inode->i_sb);

269 270 271 272 273 274 275 276
	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;

277
		set_buffer_new(bh_result);
278 279 280 281 282 283 284 285
		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;

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

J
Jaegeuk Kim 已提交
294
void update_extent_cache(struct dnode_of_data *dn)
295 296 297 298
{
	struct f2fs_inode_info *fi = F2FS_I(dn->inode);
	pgoff_t fofs, start_fofs, end_fofs;
	block_t start_blkaddr, end_blkaddr;
299
	int need_update = true;
300

J
Jaegeuk Kim 已提交
301
	f2fs_bug_on(F2FS_I_SB(dn->inode), dn->data_blkaddr == NEW_ADDR);
302 303

	/* Update the page address in the parent node */
J
Jaegeuk Kim 已提交
304
	__set_data_blkaddr(dn);
305

306 307 308
	if (is_inode_flag_set(fi, FI_NO_EXTENT))
		return;

309 310 311
	fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) +
							dn->ofs_in_node;

312 313 314 315 316 317 318 319 320 321 322 323 324
	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) {
J
Jaegeuk Kim 已提交
325
		if (dn->data_blkaddr != NULL_ADDR) {
326
			fi->ext.fofs = fofs;
J
Jaegeuk Kim 已提交
327
			fi->ext.blk_addr = dn->data_blkaddr;
328 329 330 331 332
			fi->ext.len = 1;
		}
		goto end_update;
	}

N
Namjae Jeon 已提交
333
	/* Front merge */
J
Jaegeuk Kim 已提交
334
	if (fofs == start_fofs - 1 && dn->data_blkaddr == start_blkaddr - 1) {
335 336 337 338 339 340 341
		fi->ext.fofs--;
		fi->ext.blk_addr--;
		fi->ext.len++;
		goto end_update;
	}

	/* Back merge */
J
Jaegeuk Kim 已提交
342
	if (fofs == end_fofs + 1 && dn->data_blkaddr == end_blkaddr + 1) {
343 344 345 346 347 348 349 350 351 352 353 354 355 356 357
		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;
		}
358 359
	} else {
		need_update = false;
360 361
	}

362 363 364 365 366 367
	/* 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;
	}
368 369
end_update:
	write_unlock(&fi->ext.ext_lock);
370 371 372
	if (need_update)
		sync_inode_page(dn);
	return;
373 374
}

375
struct page *find_data_page(struct inode *inode, pgoff_t index, bool sync)
376 377 378 379 380
{
	struct address_space *mapping = inode->i_mapping;
	struct dnode_of_data dn;
	struct page *page;
	int err;
381 382 383 384
	struct f2fs_io_info fio = {
		.type = DATA,
		.rw = sync ? READ_SYNC : READA,
	};
385 386 387 388 389 390 391

	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);
392
	err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
393 394 395 396 397 398 399 400
	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 */
401
	if (unlikely(dn.data_blkaddr == NEW_ADDR))
402 403
		return ERR_PTR(-EINVAL);

404
	page = grab_cache_page(mapping, index);
405 406 407
	if (!page)
		return ERR_PTR(-ENOMEM);

408 409 410 411 412
	if (PageUptodate(page)) {
		unlock_page(page);
		return page;
	}

413 414
	fio.blk_addr = dn.data_blkaddr;
	err = f2fs_submit_page_bio(F2FS_I_SB(inode), page, &fio);
415 416 417
	if (err)
		return ERR_PTR(err);

418 419
	if (sync) {
		wait_on_page_locked(page);
420
		if (unlikely(!PageUptodate(page))) {
421 422 423
			f2fs_put_page(page, 0);
			return ERR_PTR(-EIO);
		}
424 425 426 427
	}
	return page;
}

J
Jaegeuk Kim 已提交
428
/*
429 430 431 432 433 434 435 436 437 438
 * 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;
439 440 441 442
	struct f2fs_io_info fio = {
		.type = DATA,
		.rw = READ_SYNC,
	};
443
repeat:
444
	page = grab_cache_page(mapping, index);
445 446 447
	if (!page)
		return ERR_PTR(-ENOMEM);

448
	set_new_dnode(&dn, inode, NULL, NULL, 0);
449
	err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
450 451
	if (err) {
		f2fs_put_page(page, 1);
452
		return ERR_PTR(err);
453
	}
454 455
	f2fs_put_dnode(&dn);

456
	if (unlikely(dn.data_blkaddr == NULL_ADDR)) {
457
		f2fs_put_page(page, 1);
458
		return ERR_PTR(-ENOENT);
459
	}
460 461 462 463

	if (PageUptodate(page))
		return page;

J
Jaegeuk Kim 已提交
464 465 466 467 468 469 470 471 472 473 474
	/*
	 * 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;
	}
475

476 477
	fio.blk_addr = dn.data_blkaddr;
	err = f2fs_submit_page_bio(F2FS_I_SB(inode), page, &fio);
478
	if (err)
479
		return ERR_PTR(err);
480 481

	lock_page(page);
482
	if (unlikely(!PageUptodate(page))) {
483 484
		f2fs_put_page(page, 1);
		return ERR_PTR(-EIO);
485
	}
486
	if (unlikely(page->mapping != mapping)) {
487 488
		f2fs_put_page(page, 1);
		goto repeat;
489 490 491 492
	}
	return page;
}

J
Jaegeuk Kim 已提交
493
/*
494 495
 * Caller ensures that this data page is never allocated.
 * A new zero-filled data page is allocated in the page cache.
496
 *
C
Chao Yu 已提交
497 498
 * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and
 * f2fs_unlock_op().
499
 * Note that, ipage is set only by make_empty_dir.
500
 */
501
struct page *get_new_data_page(struct inode *inode,
502
		struct page *ipage, pgoff_t index, bool new_i_size)
503 504 505 506 507 508
{
	struct address_space *mapping = inode->i_mapping;
	struct page *page;
	struct dnode_of_data dn;
	int err;

509
	set_new_dnode(&dn, inode, ipage, NULL, 0);
510
	err = f2fs_reserve_block(&dn, index);
511 512
	if (err)
		return ERR_PTR(err);
513
repeat:
514
	page = grab_cache_page(mapping, index);
515 516 517 518
	if (!page) {
		err = -ENOMEM;
		goto put_err;
	}
519 520 521 522 523 524

	if (PageUptodate(page))
		return page;

	if (dn.data_blkaddr == NEW_ADDR) {
		zero_user_segment(page, 0, PAGE_CACHE_SIZE);
525
		SetPageUptodate(page);
526
	} else {
527 528 529 530 531 532
		struct f2fs_io_info fio = {
			.type = DATA,
			.rw = READ_SYNC,
			.blk_addr = dn.data_blkaddr,
		};
		err = f2fs_submit_page_bio(F2FS_I_SB(inode), page, &fio);
533
		if (err)
534 535
			goto put_err;

536
		lock_page(page);
537
		if (unlikely(!PageUptodate(page))) {
538
			f2fs_put_page(page, 1);
539 540
			err = -EIO;
			goto put_err;
541
		}
542
		if (unlikely(page->mapping != mapping)) {
543 544
			f2fs_put_page(page, 1);
			goto repeat;
545 546 547 548 549 550
		}
	}

	if (new_i_size &&
		i_size_read(inode) < ((index + 1) << PAGE_CACHE_SHIFT)) {
		i_size_write(inode, ((index + 1) << PAGE_CACHE_SHIFT));
551 552
		/* Only the directory inode sets new_i_size */
		set_inode_flag(F2FS_I(inode), FI_UPDATE_DIR);
553 554
	}
	return page;
555 556 557 558

put_err:
	f2fs_put_dnode(&dn);
	return ERR_PTR(err);
559 560
}

561 562
static int __allocate_data_block(struct dnode_of_data *dn)
{
563
	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
564
	struct f2fs_inode_info *fi = F2FS_I(dn->inode);
565 566
	struct f2fs_summary sum;
	struct node_info ni;
567
	int seg = CURSEG_WARM_DATA;
568
	pgoff_t fofs;
569 570 571 572 573 574 575 576 577

	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;

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

578 579 580 581
	if (dn->ofs_in_node == 0 && dn->inode_page == dn->node_page)
		seg = CURSEG_DIRECT_IO;

	allocate_data_block(sbi, NULL, NULL_ADDR, &dn->data_blkaddr, &sum, seg);
582 583

	/* direct IO doesn't use extent cache to maximize the performance */
J
Jaegeuk Kim 已提交
584
	__set_data_blkaddr(dn);
585

586 587 588 589 590 591
	/* 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));

592 593 594
	return 0;
}

595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644
static void __allocate_data_blocks(struct inode *inode, loff_t offset,
							size_t count)
{
	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
	struct dnode_of_data dn;
	u64 start = F2FS_BYTES_TO_BLK(offset);
	u64 len = F2FS_BYTES_TO_BLK(count);
	bool allocated;
	u64 end_offset;

	while (len) {
		f2fs_balance_fs(sbi);
		f2fs_lock_op(sbi);

		/* When reading holes, we need its node page */
		set_new_dnode(&dn, inode, NULL, NULL, 0);
		if (get_dnode_of_data(&dn, start, ALLOC_NODE))
			goto out;

		allocated = false;
		end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));

		while (dn.ofs_in_node < end_offset && len) {
			if (dn.data_blkaddr == NULL_ADDR) {
				if (__allocate_data_block(&dn))
					goto sync_out;
				allocated = true;
			}
			len--;
			start++;
			dn.ofs_in_node++;
		}

		if (allocated)
			sync_inode_page(&dn);

		f2fs_put_dnode(&dn);
		f2fs_unlock_op(sbi);
	}
	return;

sync_out:
	if (allocated)
		sync_inode_page(&dn);
	f2fs_put_dnode(&dn);
out:
	f2fs_unlock_op(sbi);
	return;
}

J
Jaegeuk Kim 已提交
645
/*
C
Chao Yu 已提交
646 647 648 649 650 651
 * 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
652
 */
653 654
static int __get_data_block(struct inode *inode, sector_t iblock,
			struct buffer_head *bh_result, int create, bool fiemap)
655 656 657 658
{
	unsigned int blkbits = inode->i_sb->s_blocksize_bits;
	unsigned maxblocks = bh_result->b_size >> blkbits;
	struct dnode_of_data dn;
659 660 661 662
	int mode = create ? ALLOC_NODE : LOOKUP_NODE_RA;
	pgoff_t pgofs, end_offset;
	int err = 0, ofs = 1;
	bool allocated = false;
663 664 665 666

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

667 668 669
	if (check_extent_cache(inode, pgofs, bh_result))
		goto out;

670
	if (create)
671
		f2fs_lock_op(F2FS_I_SB(inode));
672 673 674

	/* When reading holes, we need its node page */
	set_new_dnode(&dn, inode, NULL, NULL, 0);
675
	err = get_dnode_of_data(&dn, pgofs, mode);
676
	if (err) {
677 678 679
		if (err == -ENOENT)
			err = 0;
		goto unlock_out;
680
	}
681
	if (dn.data_blkaddr == NEW_ADDR && !fiemap)
682
		goto put_out;
683

684
	if (dn.data_blkaddr != NULL_ADDR) {
685
		set_buffer_new(bh_result);
686 687 688 689 690 691
		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;
692
		set_buffer_new(bh_result);
693 694 695 696 697
		map_bh(bh_result, inode->i_sb, dn.data_blkaddr);
	} else {
		goto put_out;
	}

698
	end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
699 700 701 702 703 704 705 706 707 708 709 710 711
	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);
712
		if (err) {
713 714 715 716
			if (err == -ENOENT)
				err = 0;
			goto unlock_out;
		}
717
		if (dn.data_blkaddr == NEW_ADDR && !fiemap)
718 719
			goto put_out;

720
		end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
721
	}
722

723 724 725 726 727 728 729 730 731
	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
732
		/* Give more consecutive addresses for the readahead */
733 734 735 736 737 738 739
		if (blkaddr == (bh_result->b_blocknr + ofs)) {
			ofs++;
			dn.ofs_in_node++;
			pgofs++;
			bh_result->b_size += (((size_t)1) << blkbits);
			goto get_next;
		}
740
	}
741 742 743 744
sync_out:
	if (allocated)
		sync_inode_page(&dn);
put_out:
745
	f2fs_put_dnode(&dn);
746 747
unlock_out:
	if (create)
748
		f2fs_unlock_op(F2FS_I_SB(inode));
749 750 751
out:
	trace_f2fs_get_data_block(inode, iblock, bh_result, err);
	return err;
752 753
}

754 755 756 757 758 759 760 761 762 763 764 765
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
766 767 768
int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
		u64 start, u64 len)
{
769 770
	return generic_block_fiemap(inode, fieinfo,
				start, len, get_data_block_fiemap);
J
Jaegeuk Kim 已提交
771 772
}

773 774
static int f2fs_read_data_page(struct file *file, struct page *page)
{
H
Huajun Li 已提交
775
	struct inode *inode = page->mapping->host;
776
	int ret = -EAGAIN;
H
Huajun Li 已提交
777

778 779
	trace_f2fs_readpage(page, DATA);

A
arter97 已提交
780
	/* If the file has inline data, try to read it directly */
H
Huajun Li 已提交
781 782
	if (f2fs_has_inline_data(inode))
		ret = f2fs_read_inline_data(inode, page);
783
	if (ret == -EAGAIN)
H
Huajun Li 已提交
784 785 786
		ret = mpage_readpage(page, get_data_block);

	return ret;
787 788 789 790 791 792
}

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
793 794 795 796 797 798
	struct inode *inode = file->f_mapping->host;

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

799
	return mpage_readpages(mapping, pages, nr_pages, get_data_block);
800 801
}

J
Jaegeuk Kim 已提交
802
int do_write_data_page(struct page *page, struct f2fs_io_info *fio)
803 804 805 806 807 808
{
	struct inode *inode = page->mapping->host;
	struct dnode_of_data dn;
	int err = 0;

	set_new_dnode(&dn, inode, NULL, NULL, 0);
809
	err = get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
810 811 812
	if (err)
		return err;

813
	fio->blk_addr = dn.data_blkaddr;
814 815

	/* This page is already truncated */
816
	if (fio->blk_addr == NULL_ADDR)
817 818 819 820 821 822 823 824
		goto out_writepage;

	set_page_writeback(page);

	/*
	 * If current allocation needs SSR,
	 * it had better in-place writes for updated data.
	 */
825
	if (unlikely(fio->blk_addr != NEW_ADDR &&
826 827
			!is_cold_data(page) &&
			need_inplace_update(inode))) {
828
		rewrite_data_page(page, fio);
829
		set_inode_flag(F2FS_I(inode), FI_UPDATE_WRITE);
830
	} else {
831
		write_data_page(page, &dn, fio);
J
Jaegeuk Kim 已提交
832
		update_extent_cache(&dn);
833
		set_inode_flag(F2FS_I(inode), FI_APPEND_WRITE);
834 835 836 837 838 839 840 841 842 843
	}
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;
844
	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
845 846 847
	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
848
	unsigned offset = 0;
849
	bool need_balance_fs = false;
850
	int err = 0;
J
Jaegeuk Kim 已提交
851 852
	struct f2fs_io_info fio = {
		.type = DATA,
C
Chris Fries 已提交
853
		.rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE,
J
Jaegeuk Kim 已提交
854
	};
855

856 857
	trace_f2fs_writepage(page, DATA);

858
	if (page->index < end_index)
859
		goto write;
860 861 862 863 864 865

	/*
	 * 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);
866
	if ((page->index >= end_index + 1) || !offset)
867
		goto out;
868 869

	zero_user_segment(page, offset, PAGE_CACHE_SIZE);
870
write:
871
	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
872
		goto redirty_out;
873 874 875 876 877
	if (f2fs_is_drop_cache(inode))
		goto out;
	if (f2fs_is_volatile_file(inode) && !wbc->for_reclaim &&
			available_free_memory(sbi, BASE_CHECK))
		goto redirty_out;
878

879
	/* Dentry blocks are controlled by checkpoint */
880
	if (S_ISDIR(inode->i_mode)) {
881 882
		if (unlikely(f2fs_cp_error(sbi)))
			goto redirty_out;
J
Jaegeuk Kim 已提交
883
		err = do_write_data_page(page, &fio);
884 885
		goto done;
	}
H
Huajun Li 已提交
886

887 888 889
	/* we should bypass data pages to proceed the kworkder jobs */
	if (unlikely(f2fs_cp_error(sbi))) {
		SetPageError(page);
890
		goto out;
891 892
	}

893
	if (!wbc->for_reclaim)
894
		need_balance_fs = true;
895
	else if (has_not_enough_free_secs(sbi, 0))
896
		goto redirty_out;
897

898
	err = -EAGAIN;
899
	f2fs_lock_op(sbi);
900 901 902
	if (f2fs_has_inline_data(inode))
		err = f2fs_write_inline_data(inode, page);
	if (err == -EAGAIN)
903 904 905 906 907
		err = do_write_data_page(page, &fio);
	f2fs_unlock_op(sbi);
done:
	if (err && err != -ENOENT)
		goto redirty_out;
908 909

	clear_cold_data(page);
910
out:
911
	inode_dec_dirty_pages(inode);
912
	unlock_page(page);
913
	if (need_balance_fs)
914
		f2fs_balance_fs(sbi);
915 916
	if (wbc->for_reclaim)
		f2fs_submit_merged_bio(sbi, DATA, WRITE);
917 918 919
	return 0;

redirty_out:
920
	redirty_page_for_writepage(wbc, page);
921
	return AOP_WRITEPAGE_ACTIVATE;
922 923
}

924 925 926 927 928 929 930 931 932
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;
}

933
static int f2fs_write_data_pages(struct address_space *mapping,
934 935 936
			    struct writeback_control *wbc)
{
	struct inode *inode = mapping->host;
937
	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
938
	bool locked = false;
939
	int ret;
940
	long diff;
941

942 943
	trace_f2fs_writepages(mapping->host, wbc, DATA);

P
P J P 已提交
944 945 946 947
	/* deal with chardevs and other special file */
	if (!mapping->a_ops->writepage)
		return 0;

948
	if (S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_NONE &&
949
			get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) &&
950
			available_free_memory(sbi, DIRTY_DENTS))
951
		goto skip_write;
952

953
	diff = nr_pages_to_write(sbi, DATA, wbc);
954

955
	if (!S_ISDIR(inode->i_mode)) {
956
		mutex_lock(&sbi->writepages);
957 958
		locked = true;
	}
959
	ret = write_cache_pages(mapping, wbc, __f2fs_writepage, mapping);
960
	if (locked)
961
		mutex_unlock(&sbi->writepages);
J
Jaegeuk Kim 已提交
962 963

	f2fs_submit_merged_bio(sbi, DATA, WRITE);
964 965 966

	remove_dirty_dir_inode(inode);

967
	wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff);
968
	return ret;
969 970

skip_write:
971
	wbc->pages_skipped += get_dirty_pages(inode);
972
	return 0;
973 974
}

975 976 977 978 979 980
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);
981
		truncate_blocks(inode, inode->i_size, true);
982 983 984
	}
}

985 986 987 988 989
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;
990
	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
991
	struct page *page, *ipage;
992 993 994 995
	pgoff_t index = ((unsigned long long) pos) >> PAGE_CACHE_SHIFT;
	struct dnode_of_data dn;
	int err = 0;

996 997
	trace_f2fs_write_begin(inode, pos, len, flags);

998
	f2fs_balance_fs(sbi);
999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009

	/*
	 * 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;
	}
1010
repeat:
1011
	page = grab_cache_page_write_begin(mapping, index, flags);
1012 1013 1014 1015
	if (!page) {
		err = -ENOMEM;
		goto fail;
	}
1016

1017 1018
	*pagep = page;

1019
	f2fs_lock_op(sbi);
1020 1021 1022

	/* check inline_data */
	ipage = get_node_page(sbi, inode->i_ino);
1023 1024
	if (IS_ERR(ipage)) {
		err = PTR_ERR(ipage);
1025
		goto unlock_fail;
1026
	}
1027

1028 1029
	set_new_dnode(&dn, inode, ipage, ipage, 0);

1030
	if (f2fs_has_inline_data(inode)) {
1031 1032 1033 1034 1035 1036
		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;
		}
1037 1038 1039
		err = f2fs_convert_inline_page(&dn, page);
		if (err)
			goto put_fail;
1040
	}
1041 1042
	err = f2fs_reserve_block(&dn, index);
	if (err)
1043
		goto put_fail;
1044
put_next:
1045 1046 1047
	f2fs_put_dnode(&dn);
	f2fs_unlock_op(sbi);

1048 1049 1050
	if ((len == PAGE_CACHE_SIZE) || PageUptodate(page))
		return 0;

1051 1052
	f2fs_wait_on_page_writeback(page, DATA);

1053 1054 1055 1056 1057 1058
	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);
1059
		goto out;
1060 1061
	}

1062
	if (dn.data_blkaddr == NEW_ADDR) {
1063 1064
		zero_user_segment(page, 0, PAGE_CACHE_SIZE);
	} else {
1065 1066 1067 1068 1069 1070
		struct f2fs_io_info fio = {
			.type = DATA,
			.rw = READ_SYNC,
			.blk_addr = dn.data_blkaddr,
		};
		err = f2fs_submit_page_bio(sbi, page, &fio);
1071 1072
		if (err)
			goto fail;
1073

1074
		lock_page(page);
1075
		if (unlikely(!PageUptodate(page))) {
1076
			f2fs_put_page(page, 1);
1077 1078
			err = -EIO;
			goto fail;
1079
		}
1080
		if (unlikely(page->mapping != mapping)) {
1081 1082
			f2fs_put_page(page, 1);
			goto repeat;
1083 1084
		}
	}
1085
out:
1086 1087 1088
	SetPageUptodate(page);
	clear_cold_data(page);
	return 0;
1089

1090 1091
put_fail:
	f2fs_put_dnode(&dn);
1092 1093
unlock_fail:
	f2fs_unlock_op(sbi);
1094
	f2fs_put_page(page, 1);
1095 1096 1097
fail:
	f2fs_write_failed(mapping, pos + len);
	return err;
1098 1099
}

1100 1101 1102 1103 1104 1105 1106
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;

1107 1108
	trace_f2fs_write_end(inode, pos, len, copied);

1109
	set_page_dirty(page);
1110 1111 1112 1113 1114 1115 1116

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

1117
	f2fs_put_page(page, 1);
1118 1119 1120
	return copied;
}

1121 1122
static int check_direct_IO(struct inode *inode, struct iov_iter *iter,
			   loff_t offset)
1123 1124 1125
{
	unsigned blocksize_mask = inode->i_sb->s_blocksize - 1;

1126
	if (iov_iter_rw(iter) == READ)
1127 1128 1129 1130 1131
		return 0;

	if (offset & blocksize_mask)
		return -EINVAL;

A
Al Viro 已提交
1132 1133 1134
	if (iov_iter_alignment(iter) & blocksize_mask)
		return -EINVAL;

1135 1136 1137
	return 0;
}

1138
static ssize_t f2fs_direct_IO(int rw, struct kiocb *iocb,
A
Al Viro 已提交
1139
		struct iov_iter *iter, loff_t offset)
1140 1141
{
	struct file *file = iocb->ki_filp;
1142 1143 1144 1145
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
	size_t count = iov_iter_count(iter);
	int err;
1146

1147 1148 1149 1150 1151 1152
	/* 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
1153

1154
	if (check_direct_IO(inode, iter, offset))
1155 1156
		return 0;

1157
	trace_f2fs_direct_IO_enter(inode, offset, count, iov_iter_rw(iter));
1158

1159
	if (iov_iter_rw(iter) == WRITE)
1160 1161
		__allocate_data_blocks(inode, offset, count);

1162
	err = blockdev_direct_IO(iocb, inode, iter, offset, get_data_block);
1163
	if (err < 0 && iov_iter_rw(iter) == WRITE)
1164
		f2fs_write_failed(mapping, offset + count);
1165

1166
	trace_f2fs_direct_IO_exit(inode, offset, count, iov_iter_rw(iter), err);
1167

1168
	return err;
1169 1170
}

1171 1172
void f2fs_invalidate_page(struct page *page, unsigned int offset,
							unsigned int length)
1173 1174
{
	struct inode *inode = page->mapping->host;
1175
	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1176

1177 1178
	if (inode->i_ino >= F2FS_ROOT_INO(sbi) &&
		(offset % PAGE_CACHE_SIZE || length != PAGE_CACHE_SIZE))
1179 1180
		return;

1181 1182 1183 1184 1185 1186 1187 1188
	if (PageDirty(page)) {
		if (inode->i_ino == F2FS_META_INO(sbi))
			dec_page_count(sbi, F2FS_DIRTY_META);
		else if (inode->i_ino == F2FS_NODE_INO(sbi))
			dec_page_count(sbi, F2FS_DIRTY_NODES);
		else
			inode_dec_dirty_pages(inode);
	}
1189 1190 1191
	ClearPagePrivate(page);
}

1192
int f2fs_release_page(struct page *page, gfp_t wait)
1193
{
1194 1195 1196 1197
	/* If this is dirty page, keep PagePrivate */
	if (PageDirty(page))
		return 0;

1198
	ClearPagePrivate(page);
1199
	return 1;
1200 1201 1202 1203 1204 1205 1206
}

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

1207 1208
	trace_f2fs_set_page_dirty(page, DATA);

1209
	SetPageUptodate(page);
1210

1211
	if (f2fs_is_atomic_file(inode)) {
1212 1213 1214 1215
		register_inmem_page(inode, page);
		return 1;
	}

1216 1217
	mark_inode_dirty(inode);

1218 1219
	if (!PageDirty(page)) {
		__set_page_dirty_nobuffers(page);
1220
		update_dirty_page(inode, page);
1221 1222 1223 1224 1225
		return 1;
	}
	return 0;
}

J
Jaegeuk Kim 已提交
1226 1227
static sector_t f2fs_bmap(struct address_space *mapping, sector_t block)
{
1228 1229
	struct inode *inode = mapping->host;

1230 1231 1232 1233 1234 1235
	/* 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;
	}
1236
	return generic_block_bmap(mapping, block, get_data_block);
J
Jaegeuk Kim 已提交
1237 1238
}

1239 1240 1241 1242 1243 1244
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,
1245
	.write_end	= f2fs_write_end,
1246
	.set_page_dirty	= f2fs_set_data_page_dirty,
1247 1248
	.invalidatepage	= f2fs_invalidate_page,
	.releasepage	= f2fs_release_page,
1249
	.direct_IO	= f2fs_direct_IO,
J
Jaegeuk Kim 已提交
1250
	.bmap		= f2fs_bmap,
1251
};