xref: /openbmc/linux/fs/f2fs/data.c (revision 359745d7)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * fs/f2fs/data.c
4  *
5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6  *             http://www.samsung.com/
7  */
8 #include <linux/fs.h>
9 #include <linux/f2fs_fs.h>
10 #include <linux/buffer_head.h>
11 #include <linux/sched/mm.h>
12 #include <linux/mpage.h>
13 #include <linux/writeback.h>
14 #include <linux/pagevec.h>
15 #include <linux/blkdev.h>
16 #include <linux/bio.h>
17 #include <linux/blk-crypto.h>
18 #include <linux/swap.h>
19 #include <linux/prefetch.h>
20 #include <linux/uio.h>
21 #include <linux/cleancache.h>
22 #include <linux/sched/signal.h>
23 #include <linux/fiemap.h>
24 #include <linux/iomap.h>
25 
26 #include "f2fs.h"
27 #include "node.h"
28 #include "segment.h"
29 #include "iostat.h"
30 #include <trace/events/f2fs.h>
31 
32 #define NUM_PREALLOC_POST_READ_CTXS	128
33 
34 static struct kmem_cache *bio_post_read_ctx_cache;
35 static struct kmem_cache *bio_entry_slab;
36 static mempool_t *bio_post_read_ctx_pool;
37 static struct bio_set f2fs_bioset;
38 
39 #define	F2FS_BIO_POOL_SIZE	NR_CURSEG_TYPE
40 
41 int __init f2fs_init_bioset(void)
42 {
43 	if (bioset_init(&f2fs_bioset, F2FS_BIO_POOL_SIZE,
44 					0, BIOSET_NEED_BVECS))
45 		return -ENOMEM;
46 	return 0;
47 }
48 
49 void f2fs_destroy_bioset(void)
50 {
51 	bioset_exit(&f2fs_bioset);
52 }
53 
54 static bool __is_cp_guaranteed(struct page *page)
55 {
56 	struct address_space *mapping = page->mapping;
57 	struct inode *inode;
58 	struct f2fs_sb_info *sbi;
59 
60 	if (!mapping)
61 		return false;
62 
63 	inode = mapping->host;
64 	sbi = F2FS_I_SB(inode);
65 
66 	if (inode->i_ino == F2FS_META_INO(sbi) ||
67 			inode->i_ino == F2FS_NODE_INO(sbi) ||
68 			S_ISDIR(inode->i_mode))
69 		return true;
70 
71 	if (f2fs_is_compressed_page(page))
72 		return false;
73 	if ((S_ISREG(inode->i_mode) &&
74 			(f2fs_is_atomic_file(inode) || IS_NOQUOTA(inode))) ||
75 			page_private_gcing(page))
76 		return true;
77 	return false;
78 }
79 
80 static enum count_type __read_io_type(struct page *page)
81 {
82 	struct address_space *mapping = page_file_mapping(page);
83 
84 	if (mapping) {
85 		struct inode *inode = mapping->host;
86 		struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
87 
88 		if (inode->i_ino == F2FS_META_INO(sbi))
89 			return F2FS_RD_META;
90 
91 		if (inode->i_ino == F2FS_NODE_INO(sbi))
92 			return F2FS_RD_NODE;
93 	}
94 	return F2FS_RD_DATA;
95 }
96 
97 /* postprocessing steps for read bios */
98 enum bio_post_read_step {
99 #ifdef CONFIG_FS_ENCRYPTION
100 	STEP_DECRYPT	= 1 << 0,
101 #else
102 	STEP_DECRYPT	= 0,	/* compile out the decryption-related code */
103 #endif
104 #ifdef CONFIG_F2FS_FS_COMPRESSION
105 	STEP_DECOMPRESS	= 1 << 1,
106 #else
107 	STEP_DECOMPRESS	= 0,	/* compile out the decompression-related code */
108 #endif
109 #ifdef CONFIG_FS_VERITY
110 	STEP_VERITY	= 1 << 2,
111 #else
112 	STEP_VERITY	= 0,	/* compile out the verity-related code */
113 #endif
114 };
115 
116 struct bio_post_read_ctx {
117 	struct bio *bio;
118 	struct f2fs_sb_info *sbi;
119 	struct work_struct work;
120 	unsigned int enabled_steps;
121 	block_t fs_blkaddr;
122 };
123 
124 static void f2fs_finish_read_bio(struct bio *bio)
125 {
126 	struct bio_vec *bv;
127 	struct bvec_iter_all iter_all;
128 
129 	/*
130 	 * Update and unlock the bio's pagecache pages, and put the
131 	 * decompression context for any compressed pages.
132 	 */
133 	bio_for_each_segment_all(bv, bio, iter_all) {
134 		struct page *page = bv->bv_page;
135 
136 		if (f2fs_is_compressed_page(page)) {
137 			if (bio->bi_status)
138 				f2fs_end_read_compressed_page(page, true, 0);
139 			f2fs_put_page_dic(page);
140 			continue;
141 		}
142 
143 		/* PG_error was set if decryption or verity failed. */
144 		if (bio->bi_status || PageError(page)) {
145 			ClearPageUptodate(page);
146 			/* will re-read again later */
147 			ClearPageError(page);
148 		} else {
149 			SetPageUptodate(page);
150 		}
151 		dec_page_count(F2FS_P_SB(page), __read_io_type(page));
152 		unlock_page(page);
153 	}
154 
155 	if (bio->bi_private)
156 		mempool_free(bio->bi_private, bio_post_read_ctx_pool);
157 	bio_put(bio);
158 }
159 
160 static void f2fs_verify_bio(struct work_struct *work)
161 {
162 	struct bio_post_read_ctx *ctx =
163 		container_of(work, struct bio_post_read_ctx, work);
164 	struct bio *bio = ctx->bio;
165 	bool may_have_compressed_pages = (ctx->enabled_steps & STEP_DECOMPRESS);
166 
167 	/*
168 	 * fsverity_verify_bio() may call readpages() again, and while verity
169 	 * will be disabled for this, decryption and/or decompression may still
170 	 * be needed, resulting in another bio_post_read_ctx being allocated.
171 	 * So to prevent deadlocks we need to release the current ctx to the
172 	 * mempool first.  This assumes that verity is the last post-read step.
173 	 */
174 	mempool_free(ctx, bio_post_read_ctx_pool);
175 	bio->bi_private = NULL;
176 
177 	/*
178 	 * Verify the bio's pages with fs-verity.  Exclude compressed pages,
179 	 * as those were handled separately by f2fs_end_read_compressed_page().
180 	 */
181 	if (may_have_compressed_pages) {
182 		struct bio_vec *bv;
183 		struct bvec_iter_all iter_all;
184 
185 		bio_for_each_segment_all(bv, bio, iter_all) {
186 			struct page *page = bv->bv_page;
187 
188 			if (!f2fs_is_compressed_page(page) &&
189 			    !PageError(page) && !fsverity_verify_page(page))
190 				SetPageError(page);
191 		}
192 	} else {
193 		fsverity_verify_bio(bio);
194 	}
195 
196 	f2fs_finish_read_bio(bio);
197 }
198 
199 /*
200  * If the bio's data needs to be verified with fs-verity, then enqueue the
201  * verity work for the bio.  Otherwise finish the bio now.
202  *
203  * Note that to avoid deadlocks, the verity work can't be done on the
204  * decryption/decompression workqueue.  This is because verifying the data pages
205  * can involve reading verity metadata pages from the file, and these verity
206  * metadata pages may be encrypted and/or compressed.
207  */
208 static void f2fs_verify_and_finish_bio(struct bio *bio)
209 {
210 	struct bio_post_read_ctx *ctx = bio->bi_private;
211 
212 	if (ctx && (ctx->enabled_steps & STEP_VERITY)) {
213 		INIT_WORK(&ctx->work, f2fs_verify_bio);
214 		fsverity_enqueue_verify_work(&ctx->work);
215 	} else {
216 		f2fs_finish_read_bio(bio);
217 	}
218 }
219 
220 /*
221  * Handle STEP_DECOMPRESS by decompressing any compressed clusters whose last
222  * remaining page was read by @ctx->bio.
223  *
224  * Note that a bio may span clusters (even a mix of compressed and uncompressed
225  * clusters) or be for just part of a cluster.  STEP_DECOMPRESS just indicates
226  * that the bio includes at least one compressed page.  The actual decompression
227  * is done on a per-cluster basis, not a per-bio basis.
228  */
229 static void f2fs_handle_step_decompress(struct bio_post_read_ctx *ctx)
230 {
231 	struct bio_vec *bv;
232 	struct bvec_iter_all iter_all;
233 	bool all_compressed = true;
234 	block_t blkaddr = ctx->fs_blkaddr;
235 
236 	bio_for_each_segment_all(bv, ctx->bio, iter_all) {
237 		struct page *page = bv->bv_page;
238 
239 		/* PG_error was set if decryption failed. */
240 		if (f2fs_is_compressed_page(page))
241 			f2fs_end_read_compressed_page(page, PageError(page),
242 						blkaddr);
243 		else
244 			all_compressed = false;
245 
246 		blkaddr++;
247 	}
248 
249 	/*
250 	 * Optimization: if all the bio's pages are compressed, then scheduling
251 	 * the per-bio verity work is unnecessary, as verity will be fully
252 	 * handled at the compression cluster level.
253 	 */
254 	if (all_compressed)
255 		ctx->enabled_steps &= ~STEP_VERITY;
256 }
257 
258 static void f2fs_post_read_work(struct work_struct *work)
259 {
260 	struct bio_post_read_ctx *ctx =
261 		container_of(work, struct bio_post_read_ctx, work);
262 
263 	if (ctx->enabled_steps & STEP_DECRYPT)
264 		fscrypt_decrypt_bio(ctx->bio);
265 
266 	if (ctx->enabled_steps & STEP_DECOMPRESS)
267 		f2fs_handle_step_decompress(ctx);
268 
269 	f2fs_verify_and_finish_bio(ctx->bio);
270 }
271 
272 static void f2fs_read_end_io(struct bio *bio)
273 {
274 	struct f2fs_sb_info *sbi = F2FS_P_SB(bio_first_page_all(bio));
275 	struct bio_post_read_ctx *ctx;
276 
277 	iostat_update_and_unbind_ctx(bio, 0);
278 	ctx = bio->bi_private;
279 
280 	if (time_to_inject(sbi, FAULT_READ_IO)) {
281 		f2fs_show_injection_info(sbi, FAULT_READ_IO);
282 		bio->bi_status = BLK_STS_IOERR;
283 	}
284 
285 	if (bio->bi_status) {
286 		f2fs_finish_read_bio(bio);
287 		return;
288 	}
289 
290 	if (ctx && (ctx->enabled_steps & (STEP_DECRYPT | STEP_DECOMPRESS))) {
291 		INIT_WORK(&ctx->work, f2fs_post_read_work);
292 		queue_work(ctx->sbi->post_read_wq, &ctx->work);
293 	} else {
294 		f2fs_verify_and_finish_bio(bio);
295 	}
296 }
297 
298 static void f2fs_write_end_io(struct bio *bio)
299 {
300 	struct f2fs_sb_info *sbi;
301 	struct bio_vec *bvec;
302 	struct bvec_iter_all iter_all;
303 
304 	iostat_update_and_unbind_ctx(bio, 1);
305 	sbi = bio->bi_private;
306 
307 	if (time_to_inject(sbi, FAULT_WRITE_IO)) {
308 		f2fs_show_injection_info(sbi, FAULT_WRITE_IO);
309 		bio->bi_status = BLK_STS_IOERR;
310 	}
311 
312 	bio_for_each_segment_all(bvec, bio, iter_all) {
313 		struct page *page = bvec->bv_page;
314 		enum count_type type = WB_DATA_TYPE(page);
315 
316 		if (page_private_dummy(page)) {
317 			clear_page_private_dummy(page);
318 			unlock_page(page);
319 			mempool_free(page, sbi->write_io_dummy);
320 
321 			if (unlikely(bio->bi_status))
322 				f2fs_stop_checkpoint(sbi, true);
323 			continue;
324 		}
325 
326 		fscrypt_finalize_bounce_page(&page);
327 
328 #ifdef CONFIG_F2FS_FS_COMPRESSION
329 		if (f2fs_is_compressed_page(page)) {
330 			f2fs_compress_write_end_io(bio, page);
331 			continue;
332 		}
333 #endif
334 
335 		if (unlikely(bio->bi_status)) {
336 			mapping_set_error(page->mapping, -EIO);
337 			if (type == F2FS_WB_CP_DATA)
338 				f2fs_stop_checkpoint(sbi, true);
339 		}
340 
341 		f2fs_bug_on(sbi, page->mapping == NODE_MAPPING(sbi) &&
342 					page->index != nid_of_node(page));
343 
344 		dec_page_count(sbi, type);
345 		if (f2fs_in_warm_node_list(sbi, page))
346 			f2fs_del_fsync_node_entry(sbi, page);
347 		clear_page_private_gcing(page);
348 		end_page_writeback(page);
349 	}
350 	if (!get_pages(sbi, F2FS_WB_CP_DATA) &&
351 				wq_has_sleeper(&sbi->cp_wait))
352 		wake_up(&sbi->cp_wait);
353 
354 	bio_put(bio);
355 }
356 
357 struct block_device *f2fs_target_device(struct f2fs_sb_info *sbi,
358 				block_t blk_addr, struct bio *bio)
359 {
360 	struct block_device *bdev = sbi->sb->s_bdev;
361 	int i;
362 
363 	if (f2fs_is_multi_device(sbi)) {
364 		for (i = 0; i < sbi->s_ndevs; i++) {
365 			if (FDEV(i).start_blk <= blk_addr &&
366 			    FDEV(i).end_blk >= blk_addr) {
367 				blk_addr -= FDEV(i).start_blk;
368 				bdev = FDEV(i).bdev;
369 				break;
370 			}
371 		}
372 	}
373 	if (bio) {
374 		bio_set_dev(bio, bdev);
375 		bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr);
376 	}
377 	return bdev;
378 }
379 
380 int f2fs_target_device_index(struct f2fs_sb_info *sbi, block_t blkaddr)
381 {
382 	int i;
383 
384 	if (!f2fs_is_multi_device(sbi))
385 		return 0;
386 
387 	for (i = 0; i < sbi->s_ndevs; i++)
388 		if (FDEV(i).start_blk <= blkaddr && FDEV(i).end_blk >= blkaddr)
389 			return i;
390 	return 0;
391 }
392 
393 static struct bio *__bio_alloc(struct f2fs_io_info *fio, int npages)
394 {
395 	struct f2fs_sb_info *sbi = fio->sbi;
396 	struct bio *bio;
397 
398 	bio = bio_alloc_bioset(GFP_NOIO, npages, &f2fs_bioset);
399 
400 	f2fs_target_device(sbi, fio->new_blkaddr, bio);
401 	if (is_read_io(fio->op)) {
402 		bio->bi_end_io = f2fs_read_end_io;
403 		bio->bi_private = NULL;
404 	} else {
405 		bio->bi_end_io = f2fs_write_end_io;
406 		bio->bi_private = sbi;
407 		bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi,
408 						fio->type, fio->temp);
409 	}
410 	iostat_alloc_and_bind_ctx(sbi, bio, NULL);
411 
412 	if (fio->io_wbc)
413 		wbc_init_bio(fio->io_wbc, bio);
414 
415 	return bio;
416 }
417 
418 static void f2fs_set_bio_crypt_ctx(struct bio *bio, const struct inode *inode,
419 				  pgoff_t first_idx,
420 				  const struct f2fs_io_info *fio,
421 				  gfp_t gfp_mask)
422 {
423 	/*
424 	 * The f2fs garbage collector sets ->encrypted_page when it wants to
425 	 * read/write raw data without encryption.
426 	 */
427 	if (!fio || !fio->encrypted_page)
428 		fscrypt_set_bio_crypt_ctx(bio, inode, first_idx, gfp_mask);
429 }
430 
431 static bool f2fs_crypt_mergeable_bio(struct bio *bio, const struct inode *inode,
432 				     pgoff_t next_idx,
433 				     const struct f2fs_io_info *fio)
434 {
435 	/*
436 	 * The f2fs garbage collector sets ->encrypted_page when it wants to
437 	 * read/write raw data without encryption.
438 	 */
439 	if (fio && fio->encrypted_page)
440 		return !bio_has_crypt_ctx(bio);
441 
442 	return fscrypt_mergeable_bio(bio, inode, next_idx);
443 }
444 
445 static inline void __submit_bio(struct f2fs_sb_info *sbi,
446 				struct bio *bio, enum page_type type)
447 {
448 	if (!is_read_io(bio_op(bio))) {
449 		unsigned int start;
450 
451 		if (type != DATA && type != NODE)
452 			goto submit_io;
453 
454 		if (f2fs_lfs_mode(sbi) && current->plug)
455 			blk_finish_plug(current->plug);
456 
457 		if (!F2FS_IO_ALIGNED(sbi))
458 			goto submit_io;
459 
460 		start = bio->bi_iter.bi_size >> F2FS_BLKSIZE_BITS;
461 		start %= F2FS_IO_SIZE(sbi);
462 
463 		if (start == 0)
464 			goto submit_io;
465 
466 		/* fill dummy pages */
467 		for (; start < F2FS_IO_SIZE(sbi); start++) {
468 			struct page *page =
469 				mempool_alloc(sbi->write_io_dummy,
470 					      GFP_NOIO | __GFP_NOFAIL);
471 			f2fs_bug_on(sbi, !page);
472 
473 			lock_page(page);
474 
475 			zero_user_segment(page, 0, PAGE_SIZE);
476 			set_page_private_dummy(page);
477 
478 			if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE)
479 				f2fs_bug_on(sbi, 1);
480 		}
481 		/*
482 		 * In the NODE case, we lose next block address chain. So, we
483 		 * need to do checkpoint in f2fs_sync_file.
484 		 */
485 		if (type == NODE)
486 			set_sbi_flag(sbi, SBI_NEED_CP);
487 	}
488 submit_io:
489 	if (is_read_io(bio_op(bio)))
490 		trace_f2fs_submit_read_bio(sbi->sb, type, bio);
491 	else
492 		trace_f2fs_submit_write_bio(sbi->sb, type, bio);
493 
494 	iostat_update_submit_ctx(bio, type);
495 	submit_bio(bio);
496 }
497 
498 void f2fs_submit_bio(struct f2fs_sb_info *sbi,
499 				struct bio *bio, enum page_type type)
500 {
501 	__submit_bio(sbi, bio, type);
502 }
503 
504 static void __attach_io_flag(struct f2fs_io_info *fio)
505 {
506 	struct f2fs_sb_info *sbi = fio->sbi;
507 	unsigned int temp_mask = (1 << NR_TEMP_TYPE) - 1;
508 	unsigned int io_flag, fua_flag, meta_flag;
509 
510 	if (fio->type == DATA)
511 		io_flag = sbi->data_io_flag;
512 	else if (fio->type == NODE)
513 		io_flag = sbi->node_io_flag;
514 	else
515 		return;
516 
517 	fua_flag = io_flag & temp_mask;
518 	meta_flag = (io_flag >> NR_TEMP_TYPE) & temp_mask;
519 
520 	/*
521 	 * data/node io flag bits per temp:
522 	 *      REQ_META     |      REQ_FUA      |
523 	 *    5 |    4 |   3 |    2 |    1 |   0 |
524 	 * Cold | Warm | Hot | Cold | Warm | Hot |
525 	 */
526 	if ((1 << fio->temp) & meta_flag)
527 		fio->op_flags |= REQ_META;
528 	if ((1 << fio->temp) & fua_flag)
529 		fio->op_flags |= REQ_FUA;
530 }
531 
532 static void __submit_merged_bio(struct f2fs_bio_info *io)
533 {
534 	struct f2fs_io_info *fio = &io->fio;
535 
536 	if (!io->bio)
537 		return;
538 
539 	__attach_io_flag(fio);
540 	bio_set_op_attrs(io->bio, fio->op, fio->op_flags);
541 
542 	if (is_read_io(fio->op))
543 		trace_f2fs_prepare_read_bio(io->sbi->sb, fio->type, io->bio);
544 	else
545 		trace_f2fs_prepare_write_bio(io->sbi->sb, fio->type, io->bio);
546 
547 	__submit_bio(io->sbi, io->bio, fio->type);
548 	io->bio = NULL;
549 }
550 
551 static bool __has_merged_page(struct bio *bio, struct inode *inode,
552 						struct page *page, nid_t ino)
553 {
554 	struct bio_vec *bvec;
555 	struct bvec_iter_all iter_all;
556 
557 	if (!bio)
558 		return false;
559 
560 	if (!inode && !page && !ino)
561 		return true;
562 
563 	bio_for_each_segment_all(bvec, bio, iter_all) {
564 		struct page *target = bvec->bv_page;
565 
566 		if (fscrypt_is_bounce_page(target)) {
567 			target = fscrypt_pagecache_page(target);
568 			if (IS_ERR(target))
569 				continue;
570 		}
571 		if (f2fs_is_compressed_page(target)) {
572 			target = f2fs_compress_control_page(target);
573 			if (IS_ERR(target))
574 				continue;
575 		}
576 
577 		if (inode && inode == target->mapping->host)
578 			return true;
579 		if (page && page == target)
580 			return true;
581 		if (ino && ino == ino_of_node(target))
582 			return true;
583 	}
584 
585 	return false;
586 }
587 
588 static void __f2fs_submit_merged_write(struct f2fs_sb_info *sbi,
589 				enum page_type type, enum temp_type temp)
590 {
591 	enum page_type btype = PAGE_TYPE_OF_BIO(type);
592 	struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
593 
594 	down_write(&io->io_rwsem);
595 
596 	/* change META to META_FLUSH in the checkpoint procedure */
597 	if (type >= META_FLUSH) {
598 		io->fio.type = META_FLUSH;
599 		io->fio.op = REQ_OP_WRITE;
600 		io->fio.op_flags = REQ_META | REQ_PRIO | REQ_SYNC;
601 		if (!test_opt(sbi, NOBARRIER))
602 			io->fio.op_flags |= REQ_PREFLUSH | REQ_FUA;
603 	}
604 	__submit_merged_bio(io);
605 	up_write(&io->io_rwsem);
606 }
607 
608 static void __submit_merged_write_cond(struct f2fs_sb_info *sbi,
609 				struct inode *inode, struct page *page,
610 				nid_t ino, enum page_type type, bool force)
611 {
612 	enum temp_type temp;
613 	bool ret = true;
614 
615 	for (temp = HOT; temp < NR_TEMP_TYPE; temp++) {
616 		if (!force)	{
617 			enum page_type btype = PAGE_TYPE_OF_BIO(type);
618 			struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
619 
620 			down_read(&io->io_rwsem);
621 			ret = __has_merged_page(io->bio, inode, page, ino);
622 			up_read(&io->io_rwsem);
623 		}
624 		if (ret)
625 			__f2fs_submit_merged_write(sbi, type, temp);
626 
627 		/* TODO: use HOT temp only for meta pages now. */
628 		if (type >= META)
629 			break;
630 	}
631 }
632 
633 void f2fs_submit_merged_write(struct f2fs_sb_info *sbi, enum page_type type)
634 {
635 	__submit_merged_write_cond(sbi, NULL, NULL, 0, type, true);
636 }
637 
638 void f2fs_submit_merged_write_cond(struct f2fs_sb_info *sbi,
639 				struct inode *inode, struct page *page,
640 				nid_t ino, enum page_type type)
641 {
642 	__submit_merged_write_cond(sbi, inode, page, ino, type, false);
643 }
644 
645 void f2fs_flush_merged_writes(struct f2fs_sb_info *sbi)
646 {
647 	f2fs_submit_merged_write(sbi, DATA);
648 	f2fs_submit_merged_write(sbi, NODE);
649 	f2fs_submit_merged_write(sbi, META);
650 }
651 
652 /*
653  * Fill the locked page with data located in the block address.
654  * A caller needs to unlock the page on failure.
655  */
656 int f2fs_submit_page_bio(struct f2fs_io_info *fio)
657 {
658 	struct bio *bio;
659 	struct page *page = fio->encrypted_page ?
660 			fio->encrypted_page : fio->page;
661 
662 	if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
663 			fio->is_por ? META_POR : (__is_meta_io(fio) ?
664 			META_GENERIC : DATA_GENERIC_ENHANCE)))
665 		return -EFSCORRUPTED;
666 
667 	trace_f2fs_submit_page_bio(page, fio);
668 
669 	/* Allocate a new bio */
670 	bio = __bio_alloc(fio, 1);
671 
672 	f2fs_set_bio_crypt_ctx(bio, fio->page->mapping->host,
673 			       fio->page->index, fio, GFP_NOIO);
674 
675 	if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
676 		bio_put(bio);
677 		return -EFAULT;
678 	}
679 
680 	if (fio->io_wbc && !is_read_io(fio->op))
681 		wbc_account_cgroup_owner(fio->io_wbc, page, PAGE_SIZE);
682 
683 	__attach_io_flag(fio);
684 	bio_set_op_attrs(bio, fio->op, fio->op_flags);
685 
686 	inc_page_count(fio->sbi, is_read_io(fio->op) ?
687 			__read_io_type(page): WB_DATA_TYPE(fio->page));
688 
689 	__submit_bio(fio->sbi, bio, fio->type);
690 	return 0;
691 }
692 
693 static bool page_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
694 				block_t last_blkaddr, block_t cur_blkaddr)
695 {
696 	if (unlikely(sbi->max_io_bytes &&
697 			bio->bi_iter.bi_size >= sbi->max_io_bytes))
698 		return false;
699 	if (last_blkaddr + 1 != cur_blkaddr)
700 		return false;
701 	return bio->bi_bdev == f2fs_target_device(sbi, cur_blkaddr, NULL);
702 }
703 
704 static bool io_type_is_mergeable(struct f2fs_bio_info *io,
705 						struct f2fs_io_info *fio)
706 {
707 	if (io->fio.op != fio->op)
708 		return false;
709 	return io->fio.op_flags == fio->op_flags;
710 }
711 
712 static bool io_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
713 					struct f2fs_bio_info *io,
714 					struct f2fs_io_info *fio,
715 					block_t last_blkaddr,
716 					block_t cur_blkaddr)
717 {
718 	if (F2FS_IO_ALIGNED(sbi) && (fio->type == DATA || fio->type == NODE)) {
719 		unsigned int filled_blocks =
720 				F2FS_BYTES_TO_BLK(bio->bi_iter.bi_size);
721 		unsigned int io_size = F2FS_IO_SIZE(sbi);
722 		unsigned int left_vecs = bio->bi_max_vecs - bio->bi_vcnt;
723 
724 		/* IOs in bio is aligned and left space of vectors is not enough */
725 		if (!(filled_blocks % io_size) && left_vecs < io_size)
726 			return false;
727 	}
728 	if (!page_is_mergeable(sbi, bio, last_blkaddr, cur_blkaddr))
729 		return false;
730 	return io_type_is_mergeable(io, fio);
731 }
732 
733 static void add_bio_entry(struct f2fs_sb_info *sbi, struct bio *bio,
734 				struct page *page, enum temp_type temp)
735 {
736 	struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
737 	struct bio_entry *be;
738 
739 	be = f2fs_kmem_cache_alloc(bio_entry_slab, GFP_NOFS, true, NULL);
740 	be->bio = bio;
741 	bio_get(bio);
742 
743 	if (bio_add_page(bio, page, PAGE_SIZE, 0) != PAGE_SIZE)
744 		f2fs_bug_on(sbi, 1);
745 
746 	down_write(&io->bio_list_lock);
747 	list_add_tail(&be->list, &io->bio_list);
748 	up_write(&io->bio_list_lock);
749 }
750 
751 static void del_bio_entry(struct bio_entry *be)
752 {
753 	list_del(&be->list);
754 	kmem_cache_free(bio_entry_slab, be);
755 }
756 
757 static int add_ipu_page(struct f2fs_io_info *fio, struct bio **bio,
758 							struct page *page)
759 {
760 	struct f2fs_sb_info *sbi = fio->sbi;
761 	enum temp_type temp;
762 	bool found = false;
763 	int ret = -EAGAIN;
764 
765 	for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
766 		struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
767 		struct list_head *head = &io->bio_list;
768 		struct bio_entry *be;
769 
770 		down_write(&io->bio_list_lock);
771 		list_for_each_entry(be, head, list) {
772 			if (be->bio != *bio)
773 				continue;
774 
775 			found = true;
776 
777 			f2fs_bug_on(sbi, !page_is_mergeable(sbi, *bio,
778 							    *fio->last_block,
779 							    fio->new_blkaddr));
780 			if (f2fs_crypt_mergeable_bio(*bio,
781 					fio->page->mapping->host,
782 					fio->page->index, fio) &&
783 			    bio_add_page(*bio, page, PAGE_SIZE, 0) ==
784 					PAGE_SIZE) {
785 				ret = 0;
786 				break;
787 			}
788 
789 			/* page can't be merged into bio; submit the bio */
790 			del_bio_entry(be);
791 			__submit_bio(sbi, *bio, DATA);
792 			break;
793 		}
794 		up_write(&io->bio_list_lock);
795 	}
796 
797 	if (ret) {
798 		bio_put(*bio);
799 		*bio = NULL;
800 	}
801 
802 	return ret;
803 }
804 
805 void f2fs_submit_merged_ipu_write(struct f2fs_sb_info *sbi,
806 					struct bio **bio, struct page *page)
807 {
808 	enum temp_type temp;
809 	bool found = false;
810 	struct bio *target = bio ? *bio : NULL;
811 
812 	for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
813 		struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
814 		struct list_head *head = &io->bio_list;
815 		struct bio_entry *be;
816 
817 		if (list_empty(head))
818 			continue;
819 
820 		down_read(&io->bio_list_lock);
821 		list_for_each_entry(be, head, list) {
822 			if (target)
823 				found = (target == be->bio);
824 			else
825 				found = __has_merged_page(be->bio, NULL,
826 								page, 0);
827 			if (found)
828 				break;
829 		}
830 		up_read(&io->bio_list_lock);
831 
832 		if (!found)
833 			continue;
834 
835 		found = false;
836 
837 		down_write(&io->bio_list_lock);
838 		list_for_each_entry(be, head, list) {
839 			if (target)
840 				found = (target == be->bio);
841 			else
842 				found = __has_merged_page(be->bio, NULL,
843 								page, 0);
844 			if (found) {
845 				target = be->bio;
846 				del_bio_entry(be);
847 				break;
848 			}
849 		}
850 		up_write(&io->bio_list_lock);
851 	}
852 
853 	if (found)
854 		__submit_bio(sbi, target, DATA);
855 	if (bio && *bio) {
856 		bio_put(*bio);
857 		*bio = NULL;
858 	}
859 }
860 
861 int f2fs_merge_page_bio(struct f2fs_io_info *fio)
862 {
863 	struct bio *bio = *fio->bio;
864 	struct page *page = fio->encrypted_page ?
865 			fio->encrypted_page : fio->page;
866 
867 	if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
868 			__is_meta_io(fio) ? META_GENERIC : DATA_GENERIC))
869 		return -EFSCORRUPTED;
870 
871 	trace_f2fs_submit_page_bio(page, fio);
872 
873 	if (bio && !page_is_mergeable(fio->sbi, bio, *fio->last_block,
874 						fio->new_blkaddr))
875 		f2fs_submit_merged_ipu_write(fio->sbi, &bio, NULL);
876 alloc_new:
877 	if (!bio) {
878 		bio = __bio_alloc(fio, BIO_MAX_VECS);
879 		__attach_io_flag(fio);
880 		f2fs_set_bio_crypt_ctx(bio, fio->page->mapping->host,
881 				       fio->page->index, fio, GFP_NOIO);
882 		bio_set_op_attrs(bio, fio->op, fio->op_flags);
883 
884 		add_bio_entry(fio->sbi, bio, page, fio->temp);
885 	} else {
886 		if (add_ipu_page(fio, &bio, page))
887 			goto alloc_new;
888 	}
889 
890 	if (fio->io_wbc)
891 		wbc_account_cgroup_owner(fio->io_wbc, page, PAGE_SIZE);
892 
893 	inc_page_count(fio->sbi, WB_DATA_TYPE(page));
894 
895 	*fio->last_block = fio->new_blkaddr;
896 	*fio->bio = bio;
897 
898 	return 0;
899 }
900 
901 void f2fs_submit_page_write(struct f2fs_io_info *fio)
902 {
903 	struct f2fs_sb_info *sbi = fio->sbi;
904 	enum page_type btype = PAGE_TYPE_OF_BIO(fio->type);
905 	struct f2fs_bio_info *io = sbi->write_io[btype] + fio->temp;
906 	struct page *bio_page;
907 
908 	f2fs_bug_on(sbi, is_read_io(fio->op));
909 
910 	down_write(&io->io_rwsem);
911 next:
912 	if (fio->in_list) {
913 		spin_lock(&io->io_lock);
914 		if (list_empty(&io->io_list)) {
915 			spin_unlock(&io->io_lock);
916 			goto out;
917 		}
918 		fio = list_first_entry(&io->io_list,
919 						struct f2fs_io_info, list);
920 		list_del(&fio->list);
921 		spin_unlock(&io->io_lock);
922 	}
923 
924 	verify_fio_blkaddr(fio);
925 
926 	if (fio->encrypted_page)
927 		bio_page = fio->encrypted_page;
928 	else if (fio->compressed_page)
929 		bio_page = fio->compressed_page;
930 	else
931 		bio_page = fio->page;
932 
933 	/* set submitted = true as a return value */
934 	fio->submitted = true;
935 
936 	inc_page_count(sbi, WB_DATA_TYPE(bio_page));
937 
938 	if (io->bio &&
939 	    (!io_is_mergeable(sbi, io->bio, io, fio, io->last_block_in_bio,
940 			      fio->new_blkaddr) ||
941 	     !f2fs_crypt_mergeable_bio(io->bio, fio->page->mapping->host,
942 				       bio_page->index, fio)))
943 		__submit_merged_bio(io);
944 alloc_new:
945 	if (io->bio == NULL) {
946 		if (F2FS_IO_ALIGNED(sbi) &&
947 				(fio->type == DATA || fio->type == NODE) &&
948 				fio->new_blkaddr & F2FS_IO_SIZE_MASK(sbi)) {
949 			dec_page_count(sbi, WB_DATA_TYPE(bio_page));
950 			fio->retry = true;
951 			goto skip;
952 		}
953 		io->bio = __bio_alloc(fio, BIO_MAX_VECS);
954 		f2fs_set_bio_crypt_ctx(io->bio, fio->page->mapping->host,
955 				       bio_page->index, fio, GFP_NOIO);
956 		io->fio = *fio;
957 	}
958 
959 	if (bio_add_page(io->bio, bio_page, PAGE_SIZE, 0) < PAGE_SIZE) {
960 		__submit_merged_bio(io);
961 		goto alloc_new;
962 	}
963 
964 	if (fio->io_wbc)
965 		wbc_account_cgroup_owner(fio->io_wbc, bio_page, PAGE_SIZE);
966 
967 	io->last_block_in_bio = fio->new_blkaddr;
968 
969 	trace_f2fs_submit_page_write(fio->page, fio);
970 skip:
971 	if (fio->in_list)
972 		goto next;
973 out:
974 	if (is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN) ||
975 				!f2fs_is_checkpoint_ready(sbi))
976 		__submit_merged_bio(io);
977 	up_write(&io->io_rwsem);
978 }
979 
980 static struct bio *f2fs_grab_read_bio(struct inode *inode, block_t blkaddr,
981 				      unsigned nr_pages, unsigned op_flag,
982 				      pgoff_t first_idx, bool for_write)
983 {
984 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
985 	struct bio *bio;
986 	struct bio_post_read_ctx *ctx = NULL;
987 	unsigned int post_read_steps = 0;
988 
989 	bio = bio_alloc_bioset(for_write ? GFP_NOIO : GFP_KERNEL,
990 			       bio_max_segs(nr_pages), &f2fs_bioset);
991 	if (!bio)
992 		return ERR_PTR(-ENOMEM);
993 
994 	f2fs_set_bio_crypt_ctx(bio, inode, first_idx, NULL, GFP_NOFS);
995 
996 	f2fs_target_device(sbi, blkaddr, bio);
997 	bio->bi_end_io = f2fs_read_end_io;
998 	bio_set_op_attrs(bio, REQ_OP_READ, op_flag);
999 
1000 	if (fscrypt_inode_uses_fs_layer_crypto(inode))
1001 		post_read_steps |= STEP_DECRYPT;
1002 
1003 	if (f2fs_need_verity(inode, first_idx))
1004 		post_read_steps |= STEP_VERITY;
1005 
1006 	/*
1007 	 * STEP_DECOMPRESS is handled specially, since a compressed file might
1008 	 * contain both compressed and uncompressed clusters.  We'll allocate a
1009 	 * bio_post_read_ctx if the file is compressed, but the caller is
1010 	 * responsible for enabling STEP_DECOMPRESS if it's actually needed.
1011 	 */
1012 
1013 	if (post_read_steps || f2fs_compressed_file(inode)) {
1014 		/* Due to the mempool, this never fails. */
1015 		ctx = mempool_alloc(bio_post_read_ctx_pool, GFP_NOFS);
1016 		ctx->bio = bio;
1017 		ctx->sbi = sbi;
1018 		ctx->enabled_steps = post_read_steps;
1019 		ctx->fs_blkaddr = blkaddr;
1020 		bio->bi_private = ctx;
1021 	}
1022 	iostat_alloc_and_bind_ctx(sbi, bio, ctx);
1023 
1024 	return bio;
1025 }
1026 
1027 /* This can handle encryption stuffs */
1028 static int f2fs_submit_page_read(struct inode *inode, struct page *page,
1029 				 block_t blkaddr, int op_flags, bool for_write)
1030 {
1031 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1032 	struct bio *bio;
1033 
1034 	bio = f2fs_grab_read_bio(inode, blkaddr, 1, op_flags,
1035 					page->index, for_write);
1036 	if (IS_ERR(bio))
1037 		return PTR_ERR(bio);
1038 
1039 	/* wait for GCed page writeback via META_MAPPING */
1040 	f2fs_wait_on_block_writeback(inode, blkaddr);
1041 
1042 	if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
1043 		bio_put(bio);
1044 		return -EFAULT;
1045 	}
1046 	ClearPageError(page);
1047 	inc_page_count(sbi, F2FS_RD_DATA);
1048 	f2fs_update_iostat(sbi, FS_DATA_READ_IO, F2FS_BLKSIZE);
1049 	__submit_bio(sbi, bio, DATA);
1050 	return 0;
1051 }
1052 
1053 static void __set_data_blkaddr(struct dnode_of_data *dn)
1054 {
1055 	struct f2fs_node *rn = F2FS_NODE(dn->node_page);
1056 	__le32 *addr_array;
1057 	int base = 0;
1058 
1059 	if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
1060 		base = get_extra_isize(dn->inode);
1061 
1062 	/* Get physical address of data block */
1063 	addr_array = blkaddr_in_node(rn);
1064 	addr_array[base + dn->ofs_in_node] = cpu_to_le32(dn->data_blkaddr);
1065 }
1066 
1067 /*
1068  * Lock ordering for the change of data block address:
1069  * ->data_page
1070  *  ->node_page
1071  *    update block addresses in the node page
1072  */
1073 void f2fs_set_data_blkaddr(struct dnode_of_data *dn)
1074 {
1075 	f2fs_wait_on_page_writeback(dn->node_page, NODE, true, true);
1076 	__set_data_blkaddr(dn);
1077 	if (set_page_dirty(dn->node_page))
1078 		dn->node_changed = true;
1079 }
1080 
1081 void f2fs_update_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr)
1082 {
1083 	dn->data_blkaddr = blkaddr;
1084 	f2fs_set_data_blkaddr(dn);
1085 	f2fs_update_extent_cache(dn);
1086 }
1087 
1088 /* dn->ofs_in_node will be returned with up-to-date last block pointer */
1089 int f2fs_reserve_new_blocks(struct dnode_of_data *dn, blkcnt_t count)
1090 {
1091 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1092 	int err;
1093 
1094 	if (!count)
1095 		return 0;
1096 
1097 	if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
1098 		return -EPERM;
1099 	if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
1100 		return err;
1101 
1102 	trace_f2fs_reserve_new_blocks(dn->inode, dn->nid,
1103 						dn->ofs_in_node, count);
1104 
1105 	f2fs_wait_on_page_writeback(dn->node_page, NODE, true, true);
1106 
1107 	for (; count > 0; dn->ofs_in_node++) {
1108 		block_t blkaddr = f2fs_data_blkaddr(dn);
1109 
1110 		if (blkaddr == NULL_ADDR) {
1111 			dn->data_blkaddr = NEW_ADDR;
1112 			__set_data_blkaddr(dn);
1113 			count--;
1114 		}
1115 	}
1116 
1117 	if (set_page_dirty(dn->node_page))
1118 		dn->node_changed = true;
1119 	return 0;
1120 }
1121 
1122 /* Should keep dn->ofs_in_node unchanged */
1123 int f2fs_reserve_new_block(struct dnode_of_data *dn)
1124 {
1125 	unsigned int ofs_in_node = dn->ofs_in_node;
1126 	int ret;
1127 
1128 	ret = f2fs_reserve_new_blocks(dn, 1);
1129 	dn->ofs_in_node = ofs_in_node;
1130 	return ret;
1131 }
1132 
1133 int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
1134 {
1135 	bool need_put = dn->inode_page ? false : true;
1136 	int err;
1137 
1138 	err = f2fs_get_dnode_of_data(dn, index, ALLOC_NODE);
1139 	if (err)
1140 		return err;
1141 
1142 	if (dn->data_blkaddr == NULL_ADDR)
1143 		err = f2fs_reserve_new_block(dn);
1144 	if (err || need_put)
1145 		f2fs_put_dnode(dn);
1146 	return err;
1147 }
1148 
1149 int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index)
1150 {
1151 	struct extent_info ei = {0, };
1152 	struct inode *inode = dn->inode;
1153 
1154 	if (f2fs_lookup_extent_cache(inode, index, &ei)) {
1155 		dn->data_blkaddr = ei.blk + index - ei.fofs;
1156 		return 0;
1157 	}
1158 
1159 	return f2fs_reserve_block(dn, index);
1160 }
1161 
1162 struct page *f2fs_get_read_data_page(struct inode *inode, pgoff_t index,
1163 						int op_flags, bool for_write)
1164 {
1165 	struct address_space *mapping = inode->i_mapping;
1166 	struct dnode_of_data dn;
1167 	struct page *page;
1168 	struct extent_info ei = {0, };
1169 	int err;
1170 
1171 	page = f2fs_grab_cache_page(mapping, index, for_write);
1172 	if (!page)
1173 		return ERR_PTR(-ENOMEM);
1174 
1175 	if (f2fs_lookup_extent_cache(inode, index, &ei)) {
1176 		dn.data_blkaddr = ei.blk + index - ei.fofs;
1177 		if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), dn.data_blkaddr,
1178 						DATA_GENERIC_ENHANCE_READ)) {
1179 			err = -EFSCORRUPTED;
1180 			goto put_err;
1181 		}
1182 		goto got_it;
1183 	}
1184 
1185 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1186 	err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
1187 	if (err)
1188 		goto put_err;
1189 	f2fs_put_dnode(&dn);
1190 
1191 	if (unlikely(dn.data_blkaddr == NULL_ADDR)) {
1192 		err = -ENOENT;
1193 		goto put_err;
1194 	}
1195 	if (dn.data_blkaddr != NEW_ADDR &&
1196 			!f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
1197 						dn.data_blkaddr,
1198 						DATA_GENERIC_ENHANCE)) {
1199 		err = -EFSCORRUPTED;
1200 		goto put_err;
1201 	}
1202 got_it:
1203 	if (PageUptodate(page)) {
1204 		unlock_page(page);
1205 		return page;
1206 	}
1207 
1208 	/*
1209 	 * A new dentry page is allocated but not able to be written, since its
1210 	 * new inode page couldn't be allocated due to -ENOSPC.
1211 	 * In such the case, its blkaddr can be remained as NEW_ADDR.
1212 	 * see, f2fs_add_link -> f2fs_get_new_data_page ->
1213 	 * f2fs_init_inode_metadata.
1214 	 */
1215 	if (dn.data_blkaddr == NEW_ADDR) {
1216 		zero_user_segment(page, 0, PAGE_SIZE);
1217 		if (!PageUptodate(page))
1218 			SetPageUptodate(page);
1219 		unlock_page(page);
1220 		return page;
1221 	}
1222 
1223 	err = f2fs_submit_page_read(inode, page, dn.data_blkaddr,
1224 						op_flags, for_write);
1225 	if (err)
1226 		goto put_err;
1227 	return page;
1228 
1229 put_err:
1230 	f2fs_put_page(page, 1);
1231 	return ERR_PTR(err);
1232 }
1233 
1234 struct page *f2fs_find_data_page(struct inode *inode, pgoff_t index)
1235 {
1236 	struct address_space *mapping = inode->i_mapping;
1237 	struct page *page;
1238 
1239 	page = find_get_page(mapping, index);
1240 	if (page && PageUptodate(page))
1241 		return page;
1242 	f2fs_put_page(page, 0);
1243 
1244 	page = f2fs_get_read_data_page(inode, index, 0, false);
1245 	if (IS_ERR(page))
1246 		return page;
1247 
1248 	if (PageUptodate(page))
1249 		return page;
1250 
1251 	wait_on_page_locked(page);
1252 	if (unlikely(!PageUptodate(page))) {
1253 		f2fs_put_page(page, 0);
1254 		return ERR_PTR(-EIO);
1255 	}
1256 	return page;
1257 }
1258 
1259 /*
1260  * If it tries to access a hole, return an error.
1261  * Because, the callers, functions in dir.c and GC, should be able to know
1262  * whether this page exists or not.
1263  */
1264 struct page *f2fs_get_lock_data_page(struct inode *inode, pgoff_t index,
1265 							bool for_write)
1266 {
1267 	struct address_space *mapping = inode->i_mapping;
1268 	struct page *page;
1269 repeat:
1270 	page = f2fs_get_read_data_page(inode, index, 0, for_write);
1271 	if (IS_ERR(page))
1272 		return page;
1273 
1274 	/* wait for read completion */
1275 	lock_page(page);
1276 	if (unlikely(page->mapping != mapping)) {
1277 		f2fs_put_page(page, 1);
1278 		goto repeat;
1279 	}
1280 	if (unlikely(!PageUptodate(page))) {
1281 		f2fs_put_page(page, 1);
1282 		return ERR_PTR(-EIO);
1283 	}
1284 	return page;
1285 }
1286 
1287 /*
1288  * Caller ensures that this data page is never allocated.
1289  * A new zero-filled data page is allocated in the page cache.
1290  *
1291  * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and
1292  * f2fs_unlock_op().
1293  * Note that, ipage is set only by make_empty_dir, and if any error occur,
1294  * ipage should be released by this function.
1295  */
1296 struct page *f2fs_get_new_data_page(struct inode *inode,
1297 		struct page *ipage, pgoff_t index, bool new_i_size)
1298 {
1299 	struct address_space *mapping = inode->i_mapping;
1300 	struct page *page;
1301 	struct dnode_of_data dn;
1302 	int err;
1303 
1304 	page = f2fs_grab_cache_page(mapping, index, true);
1305 	if (!page) {
1306 		/*
1307 		 * before exiting, we should make sure ipage will be released
1308 		 * if any error occur.
1309 		 */
1310 		f2fs_put_page(ipage, 1);
1311 		return ERR_PTR(-ENOMEM);
1312 	}
1313 
1314 	set_new_dnode(&dn, inode, ipage, NULL, 0);
1315 	err = f2fs_reserve_block(&dn, index);
1316 	if (err) {
1317 		f2fs_put_page(page, 1);
1318 		return ERR_PTR(err);
1319 	}
1320 	if (!ipage)
1321 		f2fs_put_dnode(&dn);
1322 
1323 	if (PageUptodate(page))
1324 		goto got_it;
1325 
1326 	if (dn.data_blkaddr == NEW_ADDR) {
1327 		zero_user_segment(page, 0, PAGE_SIZE);
1328 		if (!PageUptodate(page))
1329 			SetPageUptodate(page);
1330 	} else {
1331 		f2fs_put_page(page, 1);
1332 
1333 		/* if ipage exists, blkaddr should be NEW_ADDR */
1334 		f2fs_bug_on(F2FS_I_SB(inode), ipage);
1335 		page = f2fs_get_lock_data_page(inode, index, true);
1336 		if (IS_ERR(page))
1337 			return page;
1338 	}
1339 got_it:
1340 	if (new_i_size && i_size_read(inode) <
1341 				((loff_t)(index + 1) << PAGE_SHIFT))
1342 		f2fs_i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT));
1343 	return page;
1344 }
1345 
1346 static int __allocate_data_block(struct dnode_of_data *dn, int seg_type)
1347 {
1348 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1349 	struct f2fs_summary sum;
1350 	struct node_info ni;
1351 	block_t old_blkaddr;
1352 	blkcnt_t count = 1;
1353 	int err;
1354 
1355 	if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
1356 		return -EPERM;
1357 
1358 	err = f2fs_get_node_info(sbi, dn->nid, &ni, false);
1359 	if (err)
1360 		return err;
1361 
1362 	dn->data_blkaddr = f2fs_data_blkaddr(dn);
1363 	if (dn->data_blkaddr != NULL_ADDR)
1364 		goto alloc;
1365 
1366 	if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
1367 		return err;
1368 
1369 alloc:
1370 	set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
1371 	old_blkaddr = dn->data_blkaddr;
1372 	f2fs_allocate_data_block(sbi, NULL, old_blkaddr, &dn->data_blkaddr,
1373 				&sum, seg_type, NULL);
1374 	if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO) {
1375 		invalidate_mapping_pages(META_MAPPING(sbi),
1376 					old_blkaddr, old_blkaddr);
1377 		f2fs_invalidate_compress_page(sbi, old_blkaddr);
1378 	}
1379 	f2fs_update_data_blkaddr(dn, dn->data_blkaddr);
1380 	return 0;
1381 }
1382 
1383 void f2fs_do_map_lock(struct f2fs_sb_info *sbi, int flag, bool lock)
1384 {
1385 	if (flag == F2FS_GET_BLOCK_PRE_AIO) {
1386 		if (lock)
1387 			down_read(&sbi->node_change);
1388 		else
1389 			up_read(&sbi->node_change);
1390 	} else {
1391 		if (lock)
1392 			f2fs_lock_op(sbi);
1393 		else
1394 			f2fs_unlock_op(sbi);
1395 	}
1396 }
1397 
1398 /*
1399  * f2fs_map_blocks() tries to find or build mapping relationship which
1400  * maps continuous logical blocks to physical blocks, and return such
1401  * info via f2fs_map_blocks structure.
1402  */
1403 int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
1404 						int create, int flag)
1405 {
1406 	unsigned int maxblocks = map->m_len;
1407 	struct dnode_of_data dn;
1408 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1409 	int mode = map->m_may_create ? ALLOC_NODE : LOOKUP_NODE;
1410 	pgoff_t pgofs, end_offset, end;
1411 	int err = 0, ofs = 1;
1412 	unsigned int ofs_in_node, last_ofs_in_node;
1413 	blkcnt_t prealloc;
1414 	struct extent_info ei = {0, };
1415 	block_t blkaddr;
1416 	unsigned int start_pgofs;
1417 	int bidx = 0;
1418 
1419 	if (!maxblocks)
1420 		return 0;
1421 
1422 	map->m_bdev = inode->i_sb->s_bdev;
1423 	map->m_multidev_dio =
1424 		f2fs_allow_multi_device_dio(F2FS_I_SB(inode), flag);
1425 
1426 	map->m_len = 0;
1427 	map->m_flags = 0;
1428 
1429 	/* it only supports block size == page size */
1430 	pgofs =	(pgoff_t)map->m_lblk;
1431 	end = pgofs + maxblocks;
1432 
1433 	if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) {
1434 		if (f2fs_lfs_mode(sbi) && flag == F2FS_GET_BLOCK_DIO &&
1435 							map->m_may_create)
1436 			goto next_dnode;
1437 
1438 		map->m_pblk = ei.blk + pgofs - ei.fofs;
1439 		map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs);
1440 		map->m_flags = F2FS_MAP_MAPPED;
1441 		if (map->m_next_extent)
1442 			*map->m_next_extent = pgofs + map->m_len;
1443 
1444 		/* for hardware encryption, but to avoid potential issue in future */
1445 		if (flag == F2FS_GET_BLOCK_DIO)
1446 			f2fs_wait_on_block_writeback_range(inode,
1447 						map->m_pblk, map->m_len);
1448 
1449 		if (map->m_multidev_dio) {
1450 			block_t blk_addr = map->m_pblk;
1451 
1452 			bidx = f2fs_target_device_index(sbi, map->m_pblk);
1453 
1454 			map->m_bdev = FDEV(bidx).bdev;
1455 			map->m_pblk -= FDEV(bidx).start_blk;
1456 			map->m_len = min(map->m_len,
1457 				FDEV(bidx).end_blk + 1 - map->m_pblk);
1458 
1459 			if (map->m_may_create)
1460 				f2fs_update_device_state(sbi, inode->i_ino,
1461 							blk_addr, map->m_len);
1462 		}
1463 		goto out;
1464 	}
1465 
1466 next_dnode:
1467 	if (map->m_may_create)
1468 		f2fs_do_map_lock(sbi, flag, true);
1469 
1470 	/* When reading holes, we need its node page */
1471 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1472 	err = f2fs_get_dnode_of_data(&dn, pgofs, mode);
1473 	if (err) {
1474 		if (flag == F2FS_GET_BLOCK_BMAP)
1475 			map->m_pblk = 0;
1476 
1477 		if (err == -ENOENT) {
1478 			/*
1479 			 * There is one exceptional case that read_node_page()
1480 			 * may return -ENOENT due to filesystem has been
1481 			 * shutdown or cp_error, so force to convert error
1482 			 * number to EIO for such case.
1483 			 */
1484 			if (map->m_may_create &&
1485 				(is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN) ||
1486 				f2fs_cp_error(sbi))) {
1487 				err = -EIO;
1488 				goto unlock_out;
1489 			}
1490 
1491 			err = 0;
1492 			if (map->m_next_pgofs)
1493 				*map->m_next_pgofs =
1494 					f2fs_get_next_page_offset(&dn, pgofs);
1495 			if (map->m_next_extent)
1496 				*map->m_next_extent =
1497 					f2fs_get_next_page_offset(&dn, pgofs);
1498 		}
1499 		goto unlock_out;
1500 	}
1501 
1502 	start_pgofs = pgofs;
1503 	prealloc = 0;
1504 	last_ofs_in_node = ofs_in_node = dn.ofs_in_node;
1505 	end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1506 
1507 next_block:
1508 	blkaddr = f2fs_data_blkaddr(&dn);
1509 
1510 	if (__is_valid_data_blkaddr(blkaddr) &&
1511 		!f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC_ENHANCE)) {
1512 		err = -EFSCORRUPTED;
1513 		goto sync_out;
1514 	}
1515 
1516 	if (__is_valid_data_blkaddr(blkaddr)) {
1517 		/* use out-place-update for driect IO under LFS mode */
1518 		if (f2fs_lfs_mode(sbi) && flag == F2FS_GET_BLOCK_DIO &&
1519 							map->m_may_create) {
1520 			err = __allocate_data_block(&dn, map->m_seg_type);
1521 			if (err)
1522 				goto sync_out;
1523 			blkaddr = dn.data_blkaddr;
1524 			set_inode_flag(inode, FI_APPEND_WRITE);
1525 		}
1526 	} else {
1527 		if (create) {
1528 			if (unlikely(f2fs_cp_error(sbi))) {
1529 				err = -EIO;
1530 				goto sync_out;
1531 			}
1532 			if (flag == F2FS_GET_BLOCK_PRE_AIO) {
1533 				if (blkaddr == NULL_ADDR) {
1534 					prealloc++;
1535 					last_ofs_in_node = dn.ofs_in_node;
1536 				}
1537 			} else {
1538 				WARN_ON(flag != F2FS_GET_BLOCK_PRE_DIO &&
1539 					flag != F2FS_GET_BLOCK_DIO);
1540 				err = __allocate_data_block(&dn,
1541 							map->m_seg_type);
1542 				if (!err) {
1543 					if (flag == F2FS_GET_BLOCK_PRE_DIO)
1544 						file_need_truncate(inode);
1545 					set_inode_flag(inode, FI_APPEND_WRITE);
1546 				}
1547 			}
1548 			if (err)
1549 				goto sync_out;
1550 			map->m_flags |= F2FS_MAP_NEW;
1551 			blkaddr = dn.data_blkaddr;
1552 		} else {
1553 			if (f2fs_compressed_file(inode) &&
1554 					f2fs_sanity_check_cluster(&dn) &&
1555 					(flag != F2FS_GET_BLOCK_FIEMAP ||
1556 					IS_ENABLED(CONFIG_F2FS_CHECK_FS))) {
1557 				err = -EFSCORRUPTED;
1558 				goto sync_out;
1559 			}
1560 			if (flag == F2FS_GET_BLOCK_BMAP) {
1561 				map->m_pblk = 0;
1562 				goto sync_out;
1563 			}
1564 			if (flag == F2FS_GET_BLOCK_PRECACHE)
1565 				goto sync_out;
1566 			if (flag == F2FS_GET_BLOCK_FIEMAP &&
1567 						blkaddr == NULL_ADDR) {
1568 				if (map->m_next_pgofs)
1569 					*map->m_next_pgofs = pgofs + 1;
1570 				goto sync_out;
1571 			}
1572 			if (flag != F2FS_GET_BLOCK_FIEMAP) {
1573 				/* for defragment case */
1574 				if (map->m_next_pgofs)
1575 					*map->m_next_pgofs = pgofs + 1;
1576 				goto sync_out;
1577 			}
1578 		}
1579 	}
1580 
1581 	if (flag == F2FS_GET_BLOCK_PRE_AIO)
1582 		goto skip;
1583 
1584 	if (map->m_multidev_dio)
1585 		bidx = f2fs_target_device_index(sbi, blkaddr);
1586 
1587 	if (map->m_len == 0) {
1588 		/* preallocated unwritten block should be mapped for fiemap. */
1589 		if (blkaddr == NEW_ADDR)
1590 			map->m_flags |= F2FS_MAP_UNWRITTEN;
1591 		map->m_flags |= F2FS_MAP_MAPPED;
1592 
1593 		map->m_pblk = blkaddr;
1594 		map->m_len = 1;
1595 
1596 		if (map->m_multidev_dio)
1597 			map->m_bdev = FDEV(bidx).bdev;
1598 	} else if ((map->m_pblk != NEW_ADDR &&
1599 			blkaddr == (map->m_pblk + ofs)) ||
1600 			(map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR) ||
1601 			flag == F2FS_GET_BLOCK_PRE_DIO) {
1602 		if (map->m_multidev_dio && map->m_bdev != FDEV(bidx).bdev)
1603 			goto sync_out;
1604 		ofs++;
1605 		map->m_len++;
1606 	} else {
1607 		goto sync_out;
1608 	}
1609 
1610 skip:
1611 	dn.ofs_in_node++;
1612 	pgofs++;
1613 
1614 	/* preallocate blocks in batch for one dnode page */
1615 	if (flag == F2FS_GET_BLOCK_PRE_AIO &&
1616 			(pgofs == end || dn.ofs_in_node == end_offset)) {
1617 
1618 		dn.ofs_in_node = ofs_in_node;
1619 		err = f2fs_reserve_new_blocks(&dn, prealloc);
1620 		if (err)
1621 			goto sync_out;
1622 
1623 		map->m_len += dn.ofs_in_node - ofs_in_node;
1624 		if (prealloc && dn.ofs_in_node != last_ofs_in_node + 1) {
1625 			err = -ENOSPC;
1626 			goto sync_out;
1627 		}
1628 		dn.ofs_in_node = end_offset;
1629 	}
1630 
1631 	if (pgofs >= end)
1632 		goto sync_out;
1633 	else if (dn.ofs_in_node < end_offset)
1634 		goto next_block;
1635 
1636 	if (flag == F2FS_GET_BLOCK_PRECACHE) {
1637 		if (map->m_flags & F2FS_MAP_MAPPED) {
1638 			unsigned int ofs = start_pgofs - map->m_lblk;
1639 
1640 			f2fs_update_extent_cache_range(&dn,
1641 				start_pgofs, map->m_pblk + ofs,
1642 				map->m_len - ofs);
1643 		}
1644 	}
1645 
1646 	f2fs_put_dnode(&dn);
1647 
1648 	if (map->m_may_create) {
1649 		f2fs_do_map_lock(sbi, flag, false);
1650 		f2fs_balance_fs(sbi, dn.node_changed);
1651 	}
1652 	goto next_dnode;
1653 
1654 sync_out:
1655 
1656 	if (flag == F2FS_GET_BLOCK_DIO && map->m_flags & F2FS_MAP_MAPPED) {
1657 		/*
1658 		 * for hardware encryption, but to avoid potential issue
1659 		 * in future
1660 		 */
1661 		f2fs_wait_on_block_writeback_range(inode,
1662 						map->m_pblk, map->m_len);
1663 		invalidate_mapping_pages(META_MAPPING(sbi),
1664 						map->m_pblk, map->m_pblk);
1665 
1666 		if (map->m_multidev_dio) {
1667 			block_t blk_addr = map->m_pblk;
1668 
1669 			bidx = f2fs_target_device_index(sbi, map->m_pblk);
1670 
1671 			map->m_bdev = FDEV(bidx).bdev;
1672 			map->m_pblk -= FDEV(bidx).start_blk;
1673 
1674 			if (map->m_may_create)
1675 				f2fs_update_device_state(sbi, inode->i_ino,
1676 							blk_addr, map->m_len);
1677 
1678 			f2fs_bug_on(sbi, blk_addr + map->m_len >
1679 						FDEV(bidx).end_blk + 1);
1680 		}
1681 	}
1682 
1683 	if (flag == F2FS_GET_BLOCK_PRECACHE) {
1684 		if (map->m_flags & F2FS_MAP_MAPPED) {
1685 			unsigned int ofs = start_pgofs - map->m_lblk;
1686 
1687 			f2fs_update_extent_cache_range(&dn,
1688 				start_pgofs, map->m_pblk + ofs,
1689 				map->m_len - ofs);
1690 		}
1691 		if (map->m_next_extent)
1692 			*map->m_next_extent = pgofs + 1;
1693 	}
1694 	f2fs_put_dnode(&dn);
1695 unlock_out:
1696 	if (map->m_may_create) {
1697 		f2fs_do_map_lock(sbi, flag, false);
1698 		f2fs_balance_fs(sbi, dn.node_changed);
1699 	}
1700 out:
1701 	trace_f2fs_map_blocks(inode, map, create, flag, err);
1702 	return err;
1703 }
1704 
1705 bool f2fs_overwrite_io(struct inode *inode, loff_t pos, size_t len)
1706 {
1707 	struct f2fs_map_blocks map;
1708 	block_t last_lblk;
1709 	int err;
1710 
1711 	if (pos + len > i_size_read(inode))
1712 		return false;
1713 
1714 	map.m_lblk = F2FS_BYTES_TO_BLK(pos);
1715 	map.m_next_pgofs = NULL;
1716 	map.m_next_extent = NULL;
1717 	map.m_seg_type = NO_CHECK_TYPE;
1718 	map.m_may_create = false;
1719 	last_lblk = F2FS_BLK_ALIGN(pos + len);
1720 
1721 	while (map.m_lblk < last_lblk) {
1722 		map.m_len = last_lblk - map.m_lblk;
1723 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
1724 		if (err || map.m_len == 0)
1725 			return false;
1726 		map.m_lblk += map.m_len;
1727 	}
1728 	return true;
1729 }
1730 
1731 static inline u64 bytes_to_blks(struct inode *inode, u64 bytes)
1732 {
1733 	return (bytes >> inode->i_blkbits);
1734 }
1735 
1736 static inline u64 blks_to_bytes(struct inode *inode, u64 blks)
1737 {
1738 	return (blks << inode->i_blkbits);
1739 }
1740 
1741 static int f2fs_xattr_fiemap(struct inode *inode,
1742 				struct fiemap_extent_info *fieinfo)
1743 {
1744 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1745 	struct page *page;
1746 	struct node_info ni;
1747 	__u64 phys = 0, len;
1748 	__u32 flags;
1749 	nid_t xnid = F2FS_I(inode)->i_xattr_nid;
1750 	int err = 0;
1751 
1752 	if (f2fs_has_inline_xattr(inode)) {
1753 		int offset;
1754 
1755 		page = f2fs_grab_cache_page(NODE_MAPPING(sbi),
1756 						inode->i_ino, false);
1757 		if (!page)
1758 			return -ENOMEM;
1759 
1760 		err = f2fs_get_node_info(sbi, inode->i_ino, &ni, false);
1761 		if (err) {
1762 			f2fs_put_page(page, 1);
1763 			return err;
1764 		}
1765 
1766 		phys = blks_to_bytes(inode, ni.blk_addr);
1767 		offset = offsetof(struct f2fs_inode, i_addr) +
1768 					sizeof(__le32) * (DEF_ADDRS_PER_INODE -
1769 					get_inline_xattr_addrs(inode));
1770 
1771 		phys += offset;
1772 		len = inline_xattr_size(inode);
1773 
1774 		f2fs_put_page(page, 1);
1775 
1776 		flags = FIEMAP_EXTENT_DATA_INLINE | FIEMAP_EXTENT_NOT_ALIGNED;
1777 
1778 		if (!xnid)
1779 			flags |= FIEMAP_EXTENT_LAST;
1780 
1781 		err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
1782 		trace_f2fs_fiemap(inode, 0, phys, len, flags, err);
1783 		if (err || err == 1)
1784 			return err;
1785 	}
1786 
1787 	if (xnid) {
1788 		page = f2fs_grab_cache_page(NODE_MAPPING(sbi), xnid, false);
1789 		if (!page)
1790 			return -ENOMEM;
1791 
1792 		err = f2fs_get_node_info(sbi, xnid, &ni, false);
1793 		if (err) {
1794 			f2fs_put_page(page, 1);
1795 			return err;
1796 		}
1797 
1798 		phys = blks_to_bytes(inode, ni.blk_addr);
1799 		len = inode->i_sb->s_blocksize;
1800 
1801 		f2fs_put_page(page, 1);
1802 
1803 		flags = FIEMAP_EXTENT_LAST;
1804 	}
1805 
1806 	if (phys) {
1807 		err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
1808 		trace_f2fs_fiemap(inode, 0, phys, len, flags, err);
1809 	}
1810 
1811 	return (err < 0 ? err : 0);
1812 }
1813 
1814 static loff_t max_inode_blocks(struct inode *inode)
1815 {
1816 	loff_t result = ADDRS_PER_INODE(inode);
1817 	loff_t leaf_count = ADDRS_PER_BLOCK(inode);
1818 
1819 	/* two direct node blocks */
1820 	result += (leaf_count * 2);
1821 
1822 	/* two indirect node blocks */
1823 	leaf_count *= NIDS_PER_BLOCK;
1824 	result += (leaf_count * 2);
1825 
1826 	/* one double indirect node block */
1827 	leaf_count *= NIDS_PER_BLOCK;
1828 	result += leaf_count;
1829 
1830 	return result;
1831 }
1832 
1833 int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1834 		u64 start, u64 len)
1835 {
1836 	struct f2fs_map_blocks map;
1837 	sector_t start_blk, last_blk;
1838 	pgoff_t next_pgofs;
1839 	u64 logical = 0, phys = 0, size = 0;
1840 	u32 flags = 0;
1841 	int ret = 0;
1842 	bool compr_cluster = false, compr_appended;
1843 	unsigned int cluster_size = F2FS_I(inode)->i_cluster_size;
1844 	unsigned int count_in_cluster = 0;
1845 	loff_t maxbytes;
1846 
1847 	if (fieinfo->fi_flags & FIEMAP_FLAG_CACHE) {
1848 		ret = f2fs_precache_extents(inode);
1849 		if (ret)
1850 			return ret;
1851 	}
1852 
1853 	ret = fiemap_prep(inode, fieinfo, start, &len, FIEMAP_FLAG_XATTR);
1854 	if (ret)
1855 		return ret;
1856 
1857 	inode_lock(inode);
1858 
1859 	maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
1860 	if (start > maxbytes) {
1861 		ret = -EFBIG;
1862 		goto out;
1863 	}
1864 
1865 	if (len > maxbytes || (maxbytes - len) < start)
1866 		len = maxbytes - start;
1867 
1868 	if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
1869 		ret = f2fs_xattr_fiemap(inode, fieinfo);
1870 		goto out;
1871 	}
1872 
1873 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
1874 		ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len);
1875 		if (ret != -EAGAIN)
1876 			goto out;
1877 	}
1878 
1879 	if (bytes_to_blks(inode, len) == 0)
1880 		len = blks_to_bytes(inode, 1);
1881 
1882 	start_blk = bytes_to_blks(inode, start);
1883 	last_blk = bytes_to_blks(inode, start + len - 1);
1884 
1885 next:
1886 	memset(&map, 0, sizeof(map));
1887 	map.m_lblk = start_blk;
1888 	map.m_len = bytes_to_blks(inode, len);
1889 	map.m_next_pgofs = &next_pgofs;
1890 	map.m_seg_type = NO_CHECK_TYPE;
1891 
1892 	if (compr_cluster) {
1893 		map.m_lblk += 1;
1894 		map.m_len = cluster_size - count_in_cluster;
1895 	}
1896 
1897 	ret = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_FIEMAP);
1898 	if (ret)
1899 		goto out;
1900 
1901 	/* HOLE */
1902 	if (!compr_cluster && !(map.m_flags & F2FS_MAP_FLAGS)) {
1903 		start_blk = next_pgofs;
1904 
1905 		if (blks_to_bytes(inode, start_blk) < blks_to_bytes(inode,
1906 						max_inode_blocks(inode)))
1907 			goto prep_next;
1908 
1909 		flags |= FIEMAP_EXTENT_LAST;
1910 	}
1911 
1912 	compr_appended = false;
1913 	/* In a case of compressed cluster, append this to the last extent */
1914 	if (compr_cluster && ((map.m_flags & F2FS_MAP_UNWRITTEN) ||
1915 			!(map.m_flags & F2FS_MAP_FLAGS))) {
1916 		compr_appended = true;
1917 		goto skip_fill;
1918 	}
1919 
1920 	if (size) {
1921 		flags |= FIEMAP_EXTENT_MERGED;
1922 		if (IS_ENCRYPTED(inode))
1923 			flags |= FIEMAP_EXTENT_DATA_ENCRYPTED;
1924 
1925 		ret = fiemap_fill_next_extent(fieinfo, logical,
1926 				phys, size, flags);
1927 		trace_f2fs_fiemap(inode, logical, phys, size, flags, ret);
1928 		if (ret)
1929 			goto out;
1930 		size = 0;
1931 	}
1932 
1933 	if (start_blk > last_blk)
1934 		goto out;
1935 
1936 skip_fill:
1937 	if (map.m_pblk == COMPRESS_ADDR) {
1938 		compr_cluster = true;
1939 		count_in_cluster = 1;
1940 	} else if (compr_appended) {
1941 		unsigned int appended_blks = cluster_size -
1942 						count_in_cluster + 1;
1943 		size += blks_to_bytes(inode, appended_blks);
1944 		start_blk += appended_blks;
1945 		compr_cluster = false;
1946 	} else {
1947 		logical = blks_to_bytes(inode, start_blk);
1948 		phys = __is_valid_data_blkaddr(map.m_pblk) ?
1949 			blks_to_bytes(inode, map.m_pblk) : 0;
1950 		size = blks_to_bytes(inode, map.m_len);
1951 		flags = 0;
1952 
1953 		if (compr_cluster) {
1954 			flags = FIEMAP_EXTENT_ENCODED;
1955 			count_in_cluster += map.m_len;
1956 			if (count_in_cluster == cluster_size) {
1957 				compr_cluster = false;
1958 				size += blks_to_bytes(inode, 1);
1959 			}
1960 		} else if (map.m_flags & F2FS_MAP_UNWRITTEN) {
1961 			flags = FIEMAP_EXTENT_UNWRITTEN;
1962 		}
1963 
1964 		start_blk += bytes_to_blks(inode, size);
1965 	}
1966 
1967 prep_next:
1968 	cond_resched();
1969 	if (fatal_signal_pending(current))
1970 		ret = -EINTR;
1971 	else
1972 		goto next;
1973 out:
1974 	if (ret == 1)
1975 		ret = 0;
1976 
1977 	inode_unlock(inode);
1978 	return ret;
1979 }
1980 
1981 static inline loff_t f2fs_readpage_limit(struct inode *inode)
1982 {
1983 	if (IS_ENABLED(CONFIG_FS_VERITY) &&
1984 	    (IS_VERITY(inode) || f2fs_verity_in_progress(inode)))
1985 		return inode->i_sb->s_maxbytes;
1986 
1987 	return i_size_read(inode);
1988 }
1989 
1990 static int f2fs_read_single_page(struct inode *inode, struct page *page,
1991 					unsigned nr_pages,
1992 					struct f2fs_map_blocks *map,
1993 					struct bio **bio_ret,
1994 					sector_t *last_block_in_bio,
1995 					bool is_readahead)
1996 {
1997 	struct bio *bio = *bio_ret;
1998 	const unsigned blocksize = blks_to_bytes(inode, 1);
1999 	sector_t block_in_file;
2000 	sector_t last_block;
2001 	sector_t last_block_in_file;
2002 	sector_t block_nr;
2003 	int ret = 0;
2004 
2005 	block_in_file = (sector_t)page_index(page);
2006 	last_block = block_in_file + nr_pages;
2007 	last_block_in_file = bytes_to_blks(inode,
2008 			f2fs_readpage_limit(inode) + blocksize - 1);
2009 	if (last_block > last_block_in_file)
2010 		last_block = last_block_in_file;
2011 
2012 	/* just zeroing out page which is beyond EOF */
2013 	if (block_in_file >= last_block)
2014 		goto zero_out;
2015 	/*
2016 	 * Map blocks using the previous result first.
2017 	 */
2018 	if ((map->m_flags & F2FS_MAP_MAPPED) &&
2019 			block_in_file > map->m_lblk &&
2020 			block_in_file < (map->m_lblk + map->m_len))
2021 		goto got_it;
2022 
2023 	/*
2024 	 * Then do more f2fs_map_blocks() calls until we are
2025 	 * done with this page.
2026 	 */
2027 	map->m_lblk = block_in_file;
2028 	map->m_len = last_block - block_in_file;
2029 
2030 	ret = f2fs_map_blocks(inode, map, 0, F2FS_GET_BLOCK_DEFAULT);
2031 	if (ret)
2032 		goto out;
2033 got_it:
2034 	if ((map->m_flags & F2FS_MAP_MAPPED)) {
2035 		block_nr = map->m_pblk + block_in_file - map->m_lblk;
2036 		SetPageMappedToDisk(page);
2037 
2038 		if (!PageUptodate(page) && (!PageSwapCache(page) &&
2039 					!cleancache_get_page(page))) {
2040 			SetPageUptodate(page);
2041 			goto confused;
2042 		}
2043 
2044 		if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), block_nr,
2045 						DATA_GENERIC_ENHANCE_READ)) {
2046 			ret = -EFSCORRUPTED;
2047 			goto out;
2048 		}
2049 	} else {
2050 zero_out:
2051 		zero_user_segment(page, 0, PAGE_SIZE);
2052 		if (f2fs_need_verity(inode, page->index) &&
2053 		    !fsverity_verify_page(page)) {
2054 			ret = -EIO;
2055 			goto out;
2056 		}
2057 		if (!PageUptodate(page))
2058 			SetPageUptodate(page);
2059 		unlock_page(page);
2060 		goto out;
2061 	}
2062 
2063 	/*
2064 	 * This page will go to BIO.  Do we need to send this
2065 	 * BIO off first?
2066 	 */
2067 	if (bio && (!page_is_mergeable(F2FS_I_SB(inode), bio,
2068 				       *last_block_in_bio, block_nr) ||
2069 		    !f2fs_crypt_mergeable_bio(bio, inode, page->index, NULL))) {
2070 submit_and_realloc:
2071 		__submit_bio(F2FS_I_SB(inode), bio, DATA);
2072 		bio = NULL;
2073 	}
2074 	if (bio == NULL) {
2075 		bio = f2fs_grab_read_bio(inode, block_nr, nr_pages,
2076 				is_readahead ? REQ_RAHEAD : 0, page->index,
2077 				false);
2078 		if (IS_ERR(bio)) {
2079 			ret = PTR_ERR(bio);
2080 			bio = NULL;
2081 			goto out;
2082 		}
2083 	}
2084 
2085 	/*
2086 	 * If the page is under writeback, we need to wait for
2087 	 * its completion to see the correct decrypted data.
2088 	 */
2089 	f2fs_wait_on_block_writeback(inode, block_nr);
2090 
2091 	if (bio_add_page(bio, page, blocksize, 0) < blocksize)
2092 		goto submit_and_realloc;
2093 
2094 	inc_page_count(F2FS_I_SB(inode), F2FS_RD_DATA);
2095 	f2fs_update_iostat(F2FS_I_SB(inode), FS_DATA_READ_IO, F2FS_BLKSIZE);
2096 	ClearPageError(page);
2097 	*last_block_in_bio = block_nr;
2098 	goto out;
2099 confused:
2100 	if (bio) {
2101 		__submit_bio(F2FS_I_SB(inode), bio, DATA);
2102 		bio = NULL;
2103 	}
2104 	unlock_page(page);
2105 out:
2106 	*bio_ret = bio;
2107 	return ret;
2108 }
2109 
2110 #ifdef CONFIG_F2FS_FS_COMPRESSION
2111 int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret,
2112 				unsigned nr_pages, sector_t *last_block_in_bio,
2113 				bool is_readahead, bool for_write)
2114 {
2115 	struct dnode_of_data dn;
2116 	struct inode *inode = cc->inode;
2117 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2118 	struct bio *bio = *bio_ret;
2119 	unsigned int start_idx = cc->cluster_idx << cc->log_cluster_size;
2120 	sector_t last_block_in_file;
2121 	const unsigned blocksize = blks_to_bytes(inode, 1);
2122 	struct decompress_io_ctx *dic = NULL;
2123 	struct extent_info ei = {0, };
2124 	bool from_dnode = true;
2125 	int i;
2126 	int ret = 0;
2127 
2128 	f2fs_bug_on(sbi, f2fs_cluster_is_empty(cc));
2129 
2130 	last_block_in_file = bytes_to_blks(inode,
2131 			f2fs_readpage_limit(inode) + blocksize - 1);
2132 
2133 	/* get rid of pages beyond EOF */
2134 	for (i = 0; i < cc->cluster_size; i++) {
2135 		struct page *page = cc->rpages[i];
2136 
2137 		if (!page)
2138 			continue;
2139 		if ((sector_t)page->index >= last_block_in_file) {
2140 			zero_user_segment(page, 0, PAGE_SIZE);
2141 			if (!PageUptodate(page))
2142 				SetPageUptodate(page);
2143 		} else if (!PageUptodate(page)) {
2144 			continue;
2145 		}
2146 		unlock_page(page);
2147 		if (for_write)
2148 			put_page(page);
2149 		cc->rpages[i] = NULL;
2150 		cc->nr_rpages--;
2151 	}
2152 
2153 	/* we are done since all pages are beyond EOF */
2154 	if (f2fs_cluster_is_empty(cc))
2155 		goto out;
2156 
2157 	if (f2fs_lookup_extent_cache(inode, start_idx, &ei))
2158 		from_dnode = false;
2159 
2160 	if (!from_dnode)
2161 		goto skip_reading_dnode;
2162 
2163 	set_new_dnode(&dn, inode, NULL, NULL, 0);
2164 	ret = f2fs_get_dnode_of_data(&dn, start_idx, LOOKUP_NODE);
2165 	if (ret)
2166 		goto out;
2167 
2168 	f2fs_bug_on(sbi, dn.data_blkaddr != COMPRESS_ADDR);
2169 
2170 skip_reading_dnode:
2171 	for (i = 1; i < cc->cluster_size; i++) {
2172 		block_t blkaddr;
2173 
2174 		blkaddr = from_dnode ? data_blkaddr(dn.inode, dn.node_page,
2175 					dn.ofs_in_node + i) :
2176 					ei.blk + i - 1;
2177 
2178 		if (!__is_valid_data_blkaddr(blkaddr))
2179 			break;
2180 
2181 		if (!f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC)) {
2182 			ret = -EFAULT;
2183 			goto out_put_dnode;
2184 		}
2185 		cc->nr_cpages++;
2186 
2187 		if (!from_dnode && i >= ei.c_len)
2188 			break;
2189 	}
2190 
2191 	/* nothing to decompress */
2192 	if (cc->nr_cpages == 0) {
2193 		ret = 0;
2194 		goto out_put_dnode;
2195 	}
2196 
2197 	dic = f2fs_alloc_dic(cc);
2198 	if (IS_ERR(dic)) {
2199 		ret = PTR_ERR(dic);
2200 		goto out_put_dnode;
2201 	}
2202 
2203 	for (i = 0; i < cc->nr_cpages; i++) {
2204 		struct page *page = dic->cpages[i];
2205 		block_t blkaddr;
2206 		struct bio_post_read_ctx *ctx;
2207 
2208 		blkaddr = from_dnode ? data_blkaddr(dn.inode, dn.node_page,
2209 					dn.ofs_in_node + i + 1) :
2210 					ei.blk + i;
2211 
2212 		f2fs_wait_on_block_writeback(inode, blkaddr);
2213 
2214 		if (f2fs_load_compressed_page(sbi, page, blkaddr)) {
2215 			if (atomic_dec_and_test(&dic->remaining_pages))
2216 				f2fs_decompress_cluster(dic);
2217 			continue;
2218 		}
2219 
2220 		if (bio && (!page_is_mergeable(sbi, bio,
2221 					*last_block_in_bio, blkaddr) ||
2222 		    !f2fs_crypt_mergeable_bio(bio, inode, page->index, NULL))) {
2223 submit_and_realloc:
2224 			__submit_bio(sbi, bio, DATA);
2225 			bio = NULL;
2226 		}
2227 
2228 		if (!bio) {
2229 			bio = f2fs_grab_read_bio(inode, blkaddr, nr_pages,
2230 					is_readahead ? REQ_RAHEAD : 0,
2231 					page->index, for_write);
2232 			if (IS_ERR(bio)) {
2233 				ret = PTR_ERR(bio);
2234 				f2fs_decompress_end_io(dic, ret);
2235 				f2fs_put_dnode(&dn);
2236 				*bio_ret = NULL;
2237 				return ret;
2238 			}
2239 		}
2240 
2241 		if (bio_add_page(bio, page, blocksize, 0) < blocksize)
2242 			goto submit_and_realloc;
2243 
2244 		ctx = get_post_read_ctx(bio);
2245 		ctx->enabled_steps |= STEP_DECOMPRESS;
2246 		refcount_inc(&dic->refcnt);
2247 
2248 		inc_page_count(sbi, F2FS_RD_DATA);
2249 		f2fs_update_iostat(sbi, FS_DATA_READ_IO, F2FS_BLKSIZE);
2250 		f2fs_update_iostat(sbi, FS_CDATA_READ_IO, F2FS_BLKSIZE);
2251 		ClearPageError(page);
2252 		*last_block_in_bio = blkaddr;
2253 	}
2254 
2255 	if (from_dnode)
2256 		f2fs_put_dnode(&dn);
2257 
2258 	*bio_ret = bio;
2259 	return 0;
2260 
2261 out_put_dnode:
2262 	if (from_dnode)
2263 		f2fs_put_dnode(&dn);
2264 out:
2265 	for (i = 0; i < cc->cluster_size; i++) {
2266 		if (cc->rpages[i]) {
2267 			ClearPageUptodate(cc->rpages[i]);
2268 			ClearPageError(cc->rpages[i]);
2269 			unlock_page(cc->rpages[i]);
2270 		}
2271 	}
2272 	*bio_ret = bio;
2273 	return ret;
2274 }
2275 #endif
2276 
2277 /*
2278  * This function was originally taken from fs/mpage.c, and customized for f2fs.
2279  * Major change was from block_size == page_size in f2fs by default.
2280  */
2281 static int f2fs_mpage_readpages(struct inode *inode,
2282 		struct readahead_control *rac, struct page *page)
2283 {
2284 	struct bio *bio = NULL;
2285 	sector_t last_block_in_bio = 0;
2286 	struct f2fs_map_blocks map;
2287 #ifdef CONFIG_F2FS_FS_COMPRESSION
2288 	struct compress_ctx cc = {
2289 		.inode = inode,
2290 		.log_cluster_size = F2FS_I(inode)->i_log_cluster_size,
2291 		.cluster_size = F2FS_I(inode)->i_cluster_size,
2292 		.cluster_idx = NULL_CLUSTER,
2293 		.rpages = NULL,
2294 		.cpages = NULL,
2295 		.nr_rpages = 0,
2296 		.nr_cpages = 0,
2297 	};
2298 	pgoff_t nc_cluster_idx = NULL_CLUSTER;
2299 #endif
2300 	unsigned nr_pages = rac ? readahead_count(rac) : 1;
2301 	unsigned max_nr_pages = nr_pages;
2302 	int ret = 0;
2303 
2304 	map.m_pblk = 0;
2305 	map.m_lblk = 0;
2306 	map.m_len = 0;
2307 	map.m_flags = 0;
2308 	map.m_next_pgofs = NULL;
2309 	map.m_next_extent = NULL;
2310 	map.m_seg_type = NO_CHECK_TYPE;
2311 	map.m_may_create = false;
2312 
2313 	for (; nr_pages; nr_pages--) {
2314 		if (rac) {
2315 			page = readahead_page(rac);
2316 			prefetchw(&page->flags);
2317 		}
2318 
2319 #ifdef CONFIG_F2FS_FS_COMPRESSION
2320 		if (f2fs_compressed_file(inode)) {
2321 			/* there are remained comressed pages, submit them */
2322 			if (!f2fs_cluster_can_merge_page(&cc, page->index)) {
2323 				ret = f2fs_read_multi_pages(&cc, &bio,
2324 							max_nr_pages,
2325 							&last_block_in_bio,
2326 							rac != NULL, false);
2327 				f2fs_destroy_compress_ctx(&cc, false);
2328 				if (ret)
2329 					goto set_error_page;
2330 			}
2331 			if (cc.cluster_idx == NULL_CLUSTER) {
2332 				if (nc_cluster_idx ==
2333 					page->index >> cc.log_cluster_size) {
2334 					goto read_single_page;
2335 				}
2336 
2337 				ret = f2fs_is_compressed_cluster(inode, page->index);
2338 				if (ret < 0)
2339 					goto set_error_page;
2340 				else if (!ret) {
2341 					nc_cluster_idx =
2342 						page->index >> cc.log_cluster_size;
2343 					goto read_single_page;
2344 				}
2345 
2346 				nc_cluster_idx = NULL_CLUSTER;
2347 			}
2348 			ret = f2fs_init_compress_ctx(&cc);
2349 			if (ret)
2350 				goto set_error_page;
2351 
2352 			f2fs_compress_ctx_add_page(&cc, page);
2353 
2354 			goto next_page;
2355 		}
2356 read_single_page:
2357 #endif
2358 
2359 		ret = f2fs_read_single_page(inode, page, max_nr_pages, &map,
2360 					&bio, &last_block_in_bio, rac);
2361 		if (ret) {
2362 #ifdef CONFIG_F2FS_FS_COMPRESSION
2363 set_error_page:
2364 #endif
2365 			SetPageError(page);
2366 			zero_user_segment(page, 0, PAGE_SIZE);
2367 			unlock_page(page);
2368 		}
2369 #ifdef CONFIG_F2FS_FS_COMPRESSION
2370 next_page:
2371 #endif
2372 		if (rac)
2373 			put_page(page);
2374 
2375 #ifdef CONFIG_F2FS_FS_COMPRESSION
2376 		if (f2fs_compressed_file(inode)) {
2377 			/* last page */
2378 			if (nr_pages == 1 && !f2fs_cluster_is_empty(&cc)) {
2379 				ret = f2fs_read_multi_pages(&cc, &bio,
2380 							max_nr_pages,
2381 							&last_block_in_bio,
2382 							rac != NULL, false);
2383 				f2fs_destroy_compress_ctx(&cc, false);
2384 			}
2385 		}
2386 #endif
2387 	}
2388 	if (bio)
2389 		__submit_bio(F2FS_I_SB(inode), bio, DATA);
2390 	return ret;
2391 }
2392 
2393 static int f2fs_read_data_page(struct file *file, struct page *page)
2394 {
2395 	struct inode *inode = page_file_mapping(page)->host;
2396 	int ret = -EAGAIN;
2397 
2398 	trace_f2fs_readpage(page, DATA);
2399 
2400 	if (!f2fs_is_compress_backend_ready(inode)) {
2401 		unlock_page(page);
2402 		return -EOPNOTSUPP;
2403 	}
2404 
2405 	/* If the file has inline data, try to read it directly */
2406 	if (f2fs_has_inline_data(inode))
2407 		ret = f2fs_read_inline_data(inode, page);
2408 	if (ret == -EAGAIN)
2409 		ret = f2fs_mpage_readpages(inode, NULL, page);
2410 	return ret;
2411 }
2412 
2413 static void f2fs_readahead(struct readahead_control *rac)
2414 {
2415 	struct inode *inode = rac->mapping->host;
2416 
2417 	trace_f2fs_readpages(inode, readahead_index(rac), readahead_count(rac));
2418 
2419 	if (!f2fs_is_compress_backend_ready(inode))
2420 		return;
2421 
2422 	/* If the file has inline data, skip readpages */
2423 	if (f2fs_has_inline_data(inode))
2424 		return;
2425 
2426 	f2fs_mpage_readpages(inode, rac, NULL);
2427 }
2428 
2429 int f2fs_encrypt_one_page(struct f2fs_io_info *fio)
2430 {
2431 	struct inode *inode = fio->page->mapping->host;
2432 	struct page *mpage, *page;
2433 	gfp_t gfp_flags = GFP_NOFS;
2434 
2435 	if (!f2fs_encrypted_file(inode))
2436 		return 0;
2437 
2438 	page = fio->compressed_page ? fio->compressed_page : fio->page;
2439 
2440 	/* wait for GCed page writeback via META_MAPPING */
2441 	f2fs_wait_on_block_writeback(inode, fio->old_blkaddr);
2442 
2443 	if (fscrypt_inode_uses_inline_crypto(inode))
2444 		return 0;
2445 
2446 retry_encrypt:
2447 	fio->encrypted_page = fscrypt_encrypt_pagecache_blocks(page,
2448 					PAGE_SIZE, 0, gfp_flags);
2449 	if (IS_ERR(fio->encrypted_page)) {
2450 		/* flush pending IOs and wait for a while in the ENOMEM case */
2451 		if (PTR_ERR(fio->encrypted_page) == -ENOMEM) {
2452 			f2fs_flush_merged_writes(fio->sbi);
2453 			memalloc_retry_wait(GFP_NOFS);
2454 			gfp_flags |= __GFP_NOFAIL;
2455 			goto retry_encrypt;
2456 		}
2457 		return PTR_ERR(fio->encrypted_page);
2458 	}
2459 
2460 	mpage = find_lock_page(META_MAPPING(fio->sbi), fio->old_blkaddr);
2461 	if (mpage) {
2462 		if (PageUptodate(mpage))
2463 			memcpy(page_address(mpage),
2464 				page_address(fio->encrypted_page), PAGE_SIZE);
2465 		f2fs_put_page(mpage, 1);
2466 	}
2467 	return 0;
2468 }
2469 
2470 static inline bool check_inplace_update_policy(struct inode *inode,
2471 				struct f2fs_io_info *fio)
2472 {
2473 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2474 	unsigned int policy = SM_I(sbi)->ipu_policy;
2475 
2476 	if (policy & (0x1 << F2FS_IPU_FORCE))
2477 		return true;
2478 	if (policy & (0x1 << F2FS_IPU_SSR) && f2fs_need_SSR(sbi))
2479 		return true;
2480 	if (policy & (0x1 << F2FS_IPU_UTIL) &&
2481 			utilization(sbi) > SM_I(sbi)->min_ipu_util)
2482 		return true;
2483 	if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && f2fs_need_SSR(sbi) &&
2484 			utilization(sbi) > SM_I(sbi)->min_ipu_util)
2485 		return true;
2486 
2487 	/*
2488 	 * IPU for rewrite async pages
2489 	 */
2490 	if (policy & (0x1 << F2FS_IPU_ASYNC) &&
2491 			fio && fio->op == REQ_OP_WRITE &&
2492 			!(fio->op_flags & REQ_SYNC) &&
2493 			!IS_ENCRYPTED(inode))
2494 		return true;
2495 
2496 	/* this is only set during fdatasync */
2497 	if (policy & (0x1 << F2FS_IPU_FSYNC) &&
2498 			is_inode_flag_set(inode, FI_NEED_IPU))
2499 		return true;
2500 
2501 	if (unlikely(fio && is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2502 			!f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2503 		return true;
2504 
2505 	return false;
2506 }
2507 
2508 bool f2fs_should_update_inplace(struct inode *inode, struct f2fs_io_info *fio)
2509 {
2510 	/* swap file is migrating in aligned write mode */
2511 	if (is_inode_flag_set(inode, FI_ALIGNED_WRITE))
2512 		return false;
2513 
2514 	if (f2fs_is_pinned_file(inode))
2515 		return true;
2516 
2517 	/* if this is cold file, we should overwrite to avoid fragmentation */
2518 	if (file_is_cold(inode))
2519 		return true;
2520 
2521 	return check_inplace_update_policy(inode, fio);
2522 }
2523 
2524 bool f2fs_should_update_outplace(struct inode *inode, struct f2fs_io_info *fio)
2525 {
2526 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2527 
2528 	/* The below cases were checked when setting it. */
2529 	if (f2fs_is_pinned_file(inode))
2530 		return false;
2531 	if (fio && is_sbi_flag_set(sbi, SBI_NEED_FSCK))
2532 		return true;
2533 	if (f2fs_lfs_mode(sbi))
2534 		return true;
2535 	if (S_ISDIR(inode->i_mode))
2536 		return true;
2537 	if (IS_NOQUOTA(inode))
2538 		return true;
2539 	if (f2fs_is_atomic_file(inode))
2540 		return true;
2541 
2542 	/* swap file is migrating in aligned write mode */
2543 	if (is_inode_flag_set(inode, FI_ALIGNED_WRITE))
2544 		return true;
2545 
2546 	if (fio) {
2547 		if (page_private_gcing(fio->page))
2548 			return true;
2549 		if (page_private_dummy(fio->page))
2550 			return true;
2551 		if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2552 			f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2553 			return true;
2554 	}
2555 	return false;
2556 }
2557 
2558 static inline bool need_inplace_update(struct f2fs_io_info *fio)
2559 {
2560 	struct inode *inode = fio->page->mapping->host;
2561 
2562 	if (f2fs_should_update_outplace(inode, fio))
2563 		return false;
2564 
2565 	return f2fs_should_update_inplace(inode, fio);
2566 }
2567 
2568 int f2fs_do_write_data_page(struct f2fs_io_info *fio)
2569 {
2570 	struct page *page = fio->page;
2571 	struct inode *inode = page->mapping->host;
2572 	struct dnode_of_data dn;
2573 	struct extent_info ei = {0, };
2574 	struct node_info ni;
2575 	bool ipu_force = false;
2576 	int err = 0;
2577 
2578 	set_new_dnode(&dn, inode, NULL, NULL, 0);
2579 	if (need_inplace_update(fio) &&
2580 			f2fs_lookup_extent_cache(inode, page->index, &ei)) {
2581 		fio->old_blkaddr = ei.blk + page->index - ei.fofs;
2582 
2583 		if (!f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
2584 						DATA_GENERIC_ENHANCE))
2585 			return -EFSCORRUPTED;
2586 
2587 		ipu_force = true;
2588 		fio->need_lock = LOCK_DONE;
2589 		goto got_it;
2590 	}
2591 
2592 	/* Deadlock due to between page->lock and f2fs_lock_op */
2593 	if (fio->need_lock == LOCK_REQ && !f2fs_trylock_op(fio->sbi))
2594 		return -EAGAIN;
2595 
2596 	err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
2597 	if (err)
2598 		goto out;
2599 
2600 	fio->old_blkaddr = dn.data_blkaddr;
2601 
2602 	/* This page is already truncated */
2603 	if (fio->old_blkaddr == NULL_ADDR) {
2604 		ClearPageUptodate(page);
2605 		clear_page_private_gcing(page);
2606 		goto out_writepage;
2607 	}
2608 got_it:
2609 	if (__is_valid_data_blkaddr(fio->old_blkaddr) &&
2610 		!f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
2611 						DATA_GENERIC_ENHANCE)) {
2612 		err = -EFSCORRUPTED;
2613 		goto out_writepage;
2614 	}
2615 	/*
2616 	 * If current allocation needs SSR,
2617 	 * it had better in-place writes for updated data.
2618 	 */
2619 	if (ipu_force ||
2620 		(__is_valid_data_blkaddr(fio->old_blkaddr) &&
2621 					need_inplace_update(fio))) {
2622 		err = f2fs_encrypt_one_page(fio);
2623 		if (err)
2624 			goto out_writepage;
2625 
2626 		set_page_writeback(page);
2627 		ClearPageError(page);
2628 		f2fs_put_dnode(&dn);
2629 		if (fio->need_lock == LOCK_REQ)
2630 			f2fs_unlock_op(fio->sbi);
2631 		err = f2fs_inplace_write_data(fio);
2632 		if (err) {
2633 			if (fscrypt_inode_uses_fs_layer_crypto(inode))
2634 				fscrypt_finalize_bounce_page(&fio->encrypted_page);
2635 			if (PageWriteback(page))
2636 				end_page_writeback(page);
2637 		} else {
2638 			set_inode_flag(inode, FI_UPDATE_WRITE);
2639 		}
2640 		trace_f2fs_do_write_data_page(fio->page, IPU);
2641 		return err;
2642 	}
2643 
2644 	if (fio->need_lock == LOCK_RETRY) {
2645 		if (!f2fs_trylock_op(fio->sbi)) {
2646 			err = -EAGAIN;
2647 			goto out_writepage;
2648 		}
2649 		fio->need_lock = LOCK_REQ;
2650 	}
2651 
2652 	err = f2fs_get_node_info(fio->sbi, dn.nid, &ni, false);
2653 	if (err)
2654 		goto out_writepage;
2655 
2656 	fio->version = ni.version;
2657 
2658 	err = f2fs_encrypt_one_page(fio);
2659 	if (err)
2660 		goto out_writepage;
2661 
2662 	set_page_writeback(page);
2663 	ClearPageError(page);
2664 
2665 	if (fio->compr_blocks && fio->old_blkaddr == COMPRESS_ADDR)
2666 		f2fs_i_compr_blocks_update(inode, fio->compr_blocks - 1, false);
2667 
2668 	/* LFS mode write path */
2669 	f2fs_outplace_write_data(&dn, fio);
2670 	trace_f2fs_do_write_data_page(page, OPU);
2671 	set_inode_flag(inode, FI_APPEND_WRITE);
2672 	if (page->index == 0)
2673 		set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
2674 out_writepage:
2675 	f2fs_put_dnode(&dn);
2676 out:
2677 	if (fio->need_lock == LOCK_REQ)
2678 		f2fs_unlock_op(fio->sbi);
2679 	return err;
2680 }
2681 
2682 int f2fs_write_single_data_page(struct page *page, int *submitted,
2683 				struct bio **bio,
2684 				sector_t *last_block,
2685 				struct writeback_control *wbc,
2686 				enum iostat_type io_type,
2687 				int compr_blocks,
2688 				bool allow_balance)
2689 {
2690 	struct inode *inode = page->mapping->host;
2691 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2692 	loff_t i_size = i_size_read(inode);
2693 	const pgoff_t end_index = ((unsigned long long)i_size)
2694 							>> PAGE_SHIFT;
2695 	loff_t psize = (loff_t)(page->index + 1) << PAGE_SHIFT;
2696 	unsigned offset = 0;
2697 	bool need_balance_fs = false;
2698 	int err = 0;
2699 	struct f2fs_io_info fio = {
2700 		.sbi = sbi,
2701 		.ino = inode->i_ino,
2702 		.type = DATA,
2703 		.op = REQ_OP_WRITE,
2704 		.op_flags = wbc_to_write_flags(wbc),
2705 		.old_blkaddr = NULL_ADDR,
2706 		.page = page,
2707 		.encrypted_page = NULL,
2708 		.submitted = false,
2709 		.compr_blocks = compr_blocks,
2710 		.need_lock = LOCK_RETRY,
2711 		.io_type = io_type,
2712 		.io_wbc = wbc,
2713 		.bio = bio,
2714 		.last_block = last_block,
2715 	};
2716 
2717 	trace_f2fs_writepage(page, DATA);
2718 
2719 	/* we should bypass data pages to proceed the kworkder jobs */
2720 	if (unlikely(f2fs_cp_error(sbi))) {
2721 		mapping_set_error(page->mapping, -EIO);
2722 		/*
2723 		 * don't drop any dirty dentry pages for keeping lastest
2724 		 * directory structure.
2725 		 */
2726 		if (S_ISDIR(inode->i_mode))
2727 			goto redirty_out;
2728 		goto out;
2729 	}
2730 
2731 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
2732 		goto redirty_out;
2733 
2734 	if (page->index < end_index ||
2735 			f2fs_verity_in_progress(inode) ||
2736 			compr_blocks)
2737 		goto write;
2738 
2739 	/*
2740 	 * If the offset is out-of-range of file size,
2741 	 * this page does not have to be written to disk.
2742 	 */
2743 	offset = i_size & (PAGE_SIZE - 1);
2744 	if ((page->index >= end_index + 1) || !offset)
2745 		goto out;
2746 
2747 	zero_user_segment(page, offset, PAGE_SIZE);
2748 write:
2749 	if (f2fs_is_drop_cache(inode))
2750 		goto out;
2751 	/* we should not write 0'th page having journal header */
2752 	if (f2fs_is_volatile_file(inode) && (!page->index ||
2753 			(!wbc->for_reclaim &&
2754 			f2fs_available_free_memory(sbi, BASE_CHECK))))
2755 		goto redirty_out;
2756 
2757 	/* Dentry/quota blocks are controlled by checkpoint */
2758 	if (S_ISDIR(inode->i_mode) || IS_NOQUOTA(inode)) {
2759 		/*
2760 		 * We need to wait for node_write to avoid block allocation during
2761 		 * checkpoint. This can only happen to quota writes which can cause
2762 		 * the below discard race condition.
2763 		 */
2764 		if (IS_NOQUOTA(inode))
2765 			down_read(&sbi->node_write);
2766 
2767 		fio.need_lock = LOCK_DONE;
2768 		err = f2fs_do_write_data_page(&fio);
2769 
2770 		if (IS_NOQUOTA(inode))
2771 			up_read(&sbi->node_write);
2772 
2773 		goto done;
2774 	}
2775 
2776 	if (!wbc->for_reclaim)
2777 		need_balance_fs = true;
2778 	else if (has_not_enough_free_secs(sbi, 0, 0))
2779 		goto redirty_out;
2780 	else
2781 		set_inode_flag(inode, FI_HOT_DATA);
2782 
2783 	err = -EAGAIN;
2784 	if (f2fs_has_inline_data(inode)) {
2785 		err = f2fs_write_inline_data(inode, page);
2786 		if (!err)
2787 			goto out;
2788 	}
2789 
2790 	if (err == -EAGAIN) {
2791 		err = f2fs_do_write_data_page(&fio);
2792 		if (err == -EAGAIN) {
2793 			fio.need_lock = LOCK_REQ;
2794 			err = f2fs_do_write_data_page(&fio);
2795 		}
2796 	}
2797 
2798 	if (err) {
2799 		file_set_keep_isize(inode);
2800 	} else {
2801 		spin_lock(&F2FS_I(inode)->i_size_lock);
2802 		if (F2FS_I(inode)->last_disk_size < psize)
2803 			F2FS_I(inode)->last_disk_size = psize;
2804 		spin_unlock(&F2FS_I(inode)->i_size_lock);
2805 	}
2806 
2807 done:
2808 	if (err && err != -ENOENT)
2809 		goto redirty_out;
2810 
2811 out:
2812 	inode_dec_dirty_pages(inode);
2813 	if (err) {
2814 		ClearPageUptodate(page);
2815 		clear_page_private_gcing(page);
2816 	}
2817 
2818 	if (wbc->for_reclaim) {
2819 		f2fs_submit_merged_write_cond(sbi, NULL, page, 0, DATA);
2820 		clear_inode_flag(inode, FI_HOT_DATA);
2821 		f2fs_remove_dirty_inode(inode);
2822 		submitted = NULL;
2823 	}
2824 	unlock_page(page);
2825 	if (!S_ISDIR(inode->i_mode) && !IS_NOQUOTA(inode) &&
2826 			!F2FS_I(inode)->cp_task && allow_balance)
2827 		f2fs_balance_fs(sbi, need_balance_fs);
2828 
2829 	if (unlikely(f2fs_cp_error(sbi))) {
2830 		f2fs_submit_merged_write(sbi, DATA);
2831 		f2fs_submit_merged_ipu_write(sbi, bio, NULL);
2832 		submitted = NULL;
2833 	}
2834 
2835 	if (submitted)
2836 		*submitted = fio.submitted ? 1 : 0;
2837 
2838 	return 0;
2839 
2840 redirty_out:
2841 	redirty_page_for_writepage(wbc, page);
2842 	/*
2843 	 * pageout() in MM traslates EAGAIN, so calls handle_write_error()
2844 	 * -> mapping_set_error() -> set_bit(AS_EIO, ...).
2845 	 * file_write_and_wait_range() will see EIO error, which is critical
2846 	 * to return value of fsync() followed by atomic_write failure to user.
2847 	 */
2848 	if (!err || wbc->for_reclaim)
2849 		return AOP_WRITEPAGE_ACTIVATE;
2850 	unlock_page(page);
2851 	return err;
2852 }
2853 
2854 static int f2fs_write_data_page(struct page *page,
2855 					struct writeback_control *wbc)
2856 {
2857 #ifdef CONFIG_F2FS_FS_COMPRESSION
2858 	struct inode *inode = page->mapping->host;
2859 
2860 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2861 		goto out;
2862 
2863 	if (f2fs_compressed_file(inode)) {
2864 		if (f2fs_is_compressed_cluster(inode, page->index)) {
2865 			redirty_page_for_writepage(wbc, page);
2866 			return AOP_WRITEPAGE_ACTIVATE;
2867 		}
2868 	}
2869 out:
2870 #endif
2871 
2872 	return f2fs_write_single_data_page(page, NULL, NULL, NULL,
2873 						wbc, FS_DATA_IO, 0, true);
2874 }
2875 
2876 /*
2877  * This function was copied from write_cche_pages from mm/page-writeback.c.
2878  * The major change is making write step of cold data page separately from
2879  * warm/hot data page.
2880  */
2881 static int f2fs_write_cache_pages(struct address_space *mapping,
2882 					struct writeback_control *wbc,
2883 					enum iostat_type io_type)
2884 {
2885 	int ret = 0;
2886 	int done = 0, retry = 0;
2887 	struct pagevec pvec;
2888 	struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
2889 	struct bio *bio = NULL;
2890 	sector_t last_block;
2891 #ifdef CONFIG_F2FS_FS_COMPRESSION
2892 	struct inode *inode = mapping->host;
2893 	struct compress_ctx cc = {
2894 		.inode = inode,
2895 		.log_cluster_size = F2FS_I(inode)->i_log_cluster_size,
2896 		.cluster_size = F2FS_I(inode)->i_cluster_size,
2897 		.cluster_idx = NULL_CLUSTER,
2898 		.rpages = NULL,
2899 		.nr_rpages = 0,
2900 		.cpages = NULL,
2901 		.valid_nr_cpages = 0,
2902 		.rbuf = NULL,
2903 		.cbuf = NULL,
2904 		.rlen = PAGE_SIZE * F2FS_I(inode)->i_cluster_size,
2905 		.private = NULL,
2906 	};
2907 #endif
2908 	int nr_pages;
2909 	pgoff_t index;
2910 	pgoff_t end;		/* Inclusive */
2911 	pgoff_t done_index;
2912 	int range_whole = 0;
2913 	xa_mark_t tag;
2914 	int nwritten = 0;
2915 	int submitted = 0;
2916 	int i;
2917 
2918 	pagevec_init(&pvec);
2919 
2920 	if (get_dirty_pages(mapping->host) <=
2921 				SM_I(F2FS_M_SB(mapping))->min_hot_blocks)
2922 		set_inode_flag(mapping->host, FI_HOT_DATA);
2923 	else
2924 		clear_inode_flag(mapping->host, FI_HOT_DATA);
2925 
2926 	if (wbc->range_cyclic) {
2927 		index = mapping->writeback_index; /* prev offset */
2928 		end = -1;
2929 	} else {
2930 		index = wbc->range_start >> PAGE_SHIFT;
2931 		end = wbc->range_end >> PAGE_SHIFT;
2932 		if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
2933 			range_whole = 1;
2934 	}
2935 	if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2936 		tag = PAGECACHE_TAG_TOWRITE;
2937 	else
2938 		tag = PAGECACHE_TAG_DIRTY;
2939 retry:
2940 	retry = 0;
2941 	if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2942 		tag_pages_for_writeback(mapping, index, end);
2943 	done_index = index;
2944 	while (!done && !retry && (index <= end)) {
2945 		nr_pages = pagevec_lookup_range_tag(&pvec, mapping, &index, end,
2946 				tag);
2947 		if (nr_pages == 0)
2948 			break;
2949 
2950 		for (i = 0; i < nr_pages; i++) {
2951 			struct page *page = pvec.pages[i];
2952 			bool need_readd;
2953 readd:
2954 			need_readd = false;
2955 #ifdef CONFIG_F2FS_FS_COMPRESSION
2956 			if (f2fs_compressed_file(inode)) {
2957 				void *fsdata = NULL;
2958 				struct page *pagep;
2959 				int ret2;
2960 
2961 				ret = f2fs_init_compress_ctx(&cc);
2962 				if (ret) {
2963 					done = 1;
2964 					break;
2965 				}
2966 
2967 				if (!f2fs_cluster_can_merge_page(&cc,
2968 								page->index)) {
2969 					ret = f2fs_write_multi_pages(&cc,
2970 						&submitted, wbc, io_type);
2971 					if (!ret)
2972 						need_readd = true;
2973 					goto result;
2974 				}
2975 
2976 				if (unlikely(f2fs_cp_error(sbi)))
2977 					goto lock_page;
2978 
2979 				if (!f2fs_cluster_is_empty(&cc))
2980 					goto lock_page;
2981 
2982 				ret2 = f2fs_prepare_compress_overwrite(
2983 							inode, &pagep,
2984 							page->index, &fsdata);
2985 				if (ret2 < 0) {
2986 					ret = ret2;
2987 					done = 1;
2988 					break;
2989 				} else if (ret2 &&
2990 					(!f2fs_compress_write_end(inode,
2991 						fsdata, page->index, 1) ||
2992 					 !f2fs_all_cluster_page_loaded(&cc,
2993 						&pvec, i, nr_pages))) {
2994 					retry = 1;
2995 					break;
2996 				}
2997 			}
2998 #endif
2999 			/* give a priority to WB_SYNC threads */
3000 			if (atomic_read(&sbi->wb_sync_req[DATA]) &&
3001 					wbc->sync_mode == WB_SYNC_NONE) {
3002 				done = 1;
3003 				break;
3004 			}
3005 #ifdef CONFIG_F2FS_FS_COMPRESSION
3006 lock_page:
3007 #endif
3008 			done_index = page->index;
3009 retry_write:
3010 			lock_page(page);
3011 
3012 			if (unlikely(page->mapping != mapping)) {
3013 continue_unlock:
3014 				unlock_page(page);
3015 				continue;
3016 			}
3017 
3018 			if (!PageDirty(page)) {
3019 				/* someone wrote it for us */
3020 				goto continue_unlock;
3021 			}
3022 
3023 			if (PageWriteback(page)) {
3024 				if (wbc->sync_mode != WB_SYNC_NONE)
3025 					f2fs_wait_on_page_writeback(page,
3026 							DATA, true, true);
3027 				else
3028 					goto continue_unlock;
3029 			}
3030 
3031 			if (!clear_page_dirty_for_io(page))
3032 				goto continue_unlock;
3033 
3034 #ifdef CONFIG_F2FS_FS_COMPRESSION
3035 			if (f2fs_compressed_file(inode)) {
3036 				get_page(page);
3037 				f2fs_compress_ctx_add_page(&cc, page);
3038 				continue;
3039 			}
3040 #endif
3041 			ret = f2fs_write_single_data_page(page, &submitted,
3042 					&bio, &last_block, wbc, io_type,
3043 					0, true);
3044 			if (ret == AOP_WRITEPAGE_ACTIVATE)
3045 				unlock_page(page);
3046 #ifdef CONFIG_F2FS_FS_COMPRESSION
3047 result:
3048 #endif
3049 			nwritten += submitted;
3050 			wbc->nr_to_write -= submitted;
3051 
3052 			if (unlikely(ret)) {
3053 				/*
3054 				 * keep nr_to_write, since vfs uses this to
3055 				 * get # of written pages.
3056 				 */
3057 				if (ret == AOP_WRITEPAGE_ACTIVATE) {
3058 					ret = 0;
3059 					goto next;
3060 				} else if (ret == -EAGAIN) {
3061 					ret = 0;
3062 					if (wbc->sync_mode == WB_SYNC_ALL) {
3063 						cond_resched();
3064 						congestion_wait(BLK_RW_ASYNC,
3065 							DEFAULT_IO_TIMEOUT);
3066 						goto retry_write;
3067 					}
3068 					goto next;
3069 				}
3070 				done_index = page->index + 1;
3071 				done = 1;
3072 				break;
3073 			}
3074 
3075 			if (wbc->nr_to_write <= 0 &&
3076 					wbc->sync_mode == WB_SYNC_NONE) {
3077 				done = 1;
3078 				break;
3079 			}
3080 next:
3081 			if (need_readd)
3082 				goto readd;
3083 		}
3084 		pagevec_release(&pvec);
3085 		cond_resched();
3086 	}
3087 #ifdef CONFIG_F2FS_FS_COMPRESSION
3088 	/* flush remained pages in compress cluster */
3089 	if (f2fs_compressed_file(inode) && !f2fs_cluster_is_empty(&cc)) {
3090 		ret = f2fs_write_multi_pages(&cc, &submitted, wbc, io_type);
3091 		nwritten += submitted;
3092 		wbc->nr_to_write -= submitted;
3093 		if (ret) {
3094 			done = 1;
3095 			retry = 0;
3096 		}
3097 	}
3098 	if (f2fs_compressed_file(inode))
3099 		f2fs_destroy_compress_ctx(&cc, false);
3100 #endif
3101 	if (retry) {
3102 		index = 0;
3103 		end = -1;
3104 		goto retry;
3105 	}
3106 	if (wbc->range_cyclic && !done)
3107 		done_index = 0;
3108 	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
3109 		mapping->writeback_index = done_index;
3110 
3111 	if (nwritten)
3112 		f2fs_submit_merged_write_cond(F2FS_M_SB(mapping), mapping->host,
3113 								NULL, 0, DATA);
3114 	/* submit cached bio of IPU write */
3115 	if (bio)
3116 		f2fs_submit_merged_ipu_write(sbi, &bio, NULL);
3117 
3118 	return ret;
3119 }
3120 
3121 static inline bool __should_serialize_io(struct inode *inode,
3122 					struct writeback_control *wbc)
3123 {
3124 	/* to avoid deadlock in path of data flush */
3125 	if (F2FS_I(inode)->cp_task)
3126 		return false;
3127 
3128 	if (!S_ISREG(inode->i_mode))
3129 		return false;
3130 	if (IS_NOQUOTA(inode))
3131 		return false;
3132 
3133 	if (f2fs_need_compress_data(inode))
3134 		return true;
3135 	if (wbc->sync_mode != WB_SYNC_ALL)
3136 		return true;
3137 	if (get_dirty_pages(inode) >= SM_I(F2FS_I_SB(inode))->min_seq_blocks)
3138 		return true;
3139 	return false;
3140 }
3141 
3142 static int __f2fs_write_data_pages(struct address_space *mapping,
3143 						struct writeback_control *wbc,
3144 						enum iostat_type io_type)
3145 {
3146 	struct inode *inode = mapping->host;
3147 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3148 	struct blk_plug plug;
3149 	int ret;
3150 	bool locked = false;
3151 
3152 	/* deal with chardevs and other special file */
3153 	if (!mapping->a_ops->writepage)
3154 		return 0;
3155 
3156 	/* skip writing if there is no dirty page in this inode */
3157 	if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE)
3158 		return 0;
3159 
3160 	/* during POR, we don't need to trigger writepage at all. */
3161 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
3162 		goto skip_write;
3163 
3164 	if ((S_ISDIR(inode->i_mode) || IS_NOQUOTA(inode)) &&
3165 			wbc->sync_mode == WB_SYNC_NONE &&
3166 			get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) &&
3167 			f2fs_available_free_memory(sbi, DIRTY_DENTS))
3168 		goto skip_write;
3169 
3170 	/* skip writing during file defragment */
3171 	if (is_inode_flag_set(inode, FI_DO_DEFRAG))
3172 		goto skip_write;
3173 
3174 	trace_f2fs_writepages(mapping->host, wbc, DATA);
3175 
3176 	/* to avoid spliting IOs due to mixed WB_SYNC_ALL and WB_SYNC_NONE */
3177 	if (wbc->sync_mode == WB_SYNC_ALL)
3178 		atomic_inc(&sbi->wb_sync_req[DATA]);
3179 	else if (atomic_read(&sbi->wb_sync_req[DATA]))
3180 		goto skip_write;
3181 
3182 	if (__should_serialize_io(inode, wbc)) {
3183 		mutex_lock(&sbi->writepages);
3184 		locked = true;
3185 	}
3186 
3187 	blk_start_plug(&plug);
3188 	ret = f2fs_write_cache_pages(mapping, wbc, io_type);
3189 	blk_finish_plug(&plug);
3190 
3191 	if (locked)
3192 		mutex_unlock(&sbi->writepages);
3193 
3194 	if (wbc->sync_mode == WB_SYNC_ALL)
3195 		atomic_dec(&sbi->wb_sync_req[DATA]);
3196 	/*
3197 	 * if some pages were truncated, we cannot guarantee its mapping->host
3198 	 * to detect pending bios.
3199 	 */
3200 
3201 	f2fs_remove_dirty_inode(inode);
3202 	return ret;
3203 
3204 skip_write:
3205 	wbc->pages_skipped += get_dirty_pages(inode);
3206 	trace_f2fs_writepages(mapping->host, wbc, DATA);
3207 	return 0;
3208 }
3209 
3210 static int f2fs_write_data_pages(struct address_space *mapping,
3211 			    struct writeback_control *wbc)
3212 {
3213 	struct inode *inode = mapping->host;
3214 
3215 	return __f2fs_write_data_pages(mapping, wbc,
3216 			F2FS_I(inode)->cp_task == current ?
3217 			FS_CP_DATA_IO : FS_DATA_IO);
3218 }
3219 
3220 void f2fs_write_failed(struct inode *inode, loff_t to)
3221 {
3222 	loff_t i_size = i_size_read(inode);
3223 
3224 	if (IS_NOQUOTA(inode))
3225 		return;
3226 
3227 	/* In the fs-verity case, f2fs_end_enable_verity() does the truncate */
3228 	if (to > i_size && !f2fs_verity_in_progress(inode)) {
3229 		down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3230 		filemap_invalidate_lock(inode->i_mapping);
3231 
3232 		truncate_pagecache(inode, i_size);
3233 		f2fs_truncate_blocks(inode, i_size, true);
3234 
3235 		filemap_invalidate_unlock(inode->i_mapping);
3236 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3237 	}
3238 }
3239 
3240 static int prepare_write_begin(struct f2fs_sb_info *sbi,
3241 			struct page *page, loff_t pos, unsigned len,
3242 			block_t *blk_addr, bool *node_changed)
3243 {
3244 	struct inode *inode = page->mapping->host;
3245 	pgoff_t index = page->index;
3246 	struct dnode_of_data dn;
3247 	struct page *ipage;
3248 	bool locked = false;
3249 	struct extent_info ei = {0, };
3250 	int err = 0;
3251 	int flag;
3252 
3253 	/*
3254 	 * If a whole page is being written and we already preallocated all the
3255 	 * blocks, then there is no need to get a block address now.
3256 	 */
3257 	if (len == PAGE_SIZE && is_inode_flag_set(inode, FI_PREALLOCATED_ALL))
3258 		return 0;
3259 
3260 	/* f2fs_lock_op avoids race between write CP and convert_inline_page */
3261 	if (f2fs_has_inline_data(inode) && pos + len > MAX_INLINE_DATA(inode))
3262 		flag = F2FS_GET_BLOCK_DEFAULT;
3263 	else
3264 		flag = F2FS_GET_BLOCK_PRE_AIO;
3265 
3266 	if (f2fs_has_inline_data(inode) ||
3267 			(pos & PAGE_MASK) >= i_size_read(inode)) {
3268 		f2fs_do_map_lock(sbi, flag, true);
3269 		locked = true;
3270 	}
3271 
3272 restart:
3273 	/* check inline_data */
3274 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
3275 	if (IS_ERR(ipage)) {
3276 		err = PTR_ERR(ipage);
3277 		goto unlock_out;
3278 	}
3279 
3280 	set_new_dnode(&dn, inode, ipage, ipage, 0);
3281 
3282 	if (f2fs_has_inline_data(inode)) {
3283 		if (pos + len <= MAX_INLINE_DATA(inode)) {
3284 			f2fs_do_read_inline_data(page, ipage);
3285 			set_inode_flag(inode, FI_DATA_EXIST);
3286 			if (inode->i_nlink)
3287 				set_page_private_inline(ipage);
3288 		} else {
3289 			err = f2fs_convert_inline_page(&dn, page);
3290 			if (err)
3291 				goto out;
3292 			if (dn.data_blkaddr == NULL_ADDR)
3293 				err = f2fs_get_block(&dn, index);
3294 		}
3295 	} else if (locked) {
3296 		err = f2fs_get_block(&dn, index);
3297 	} else {
3298 		if (f2fs_lookup_extent_cache(inode, index, &ei)) {
3299 			dn.data_blkaddr = ei.blk + index - ei.fofs;
3300 		} else {
3301 			/* hole case */
3302 			err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3303 			if (err || dn.data_blkaddr == NULL_ADDR) {
3304 				f2fs_put_dnode(&dn);
3305 				f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO,
3306 								true);
3307 				WARN_ON(flag != F2FS_GET_BLOCK_PRE_AIO);
3308 				locked = true;
3309 				goto restart;
3310 			}
3311 		}
3312 	}
3313 
3314 	/* convert_inline_page can make node_changed */
3315 	*blk_addr = dn.data_blkaddr;
3316 	*node_changed = dn.node_changed;
3317 out:
3318 	f2fs_put_dnode(&dn);
3319 unlock_out:
3320 	if (locked)
3321 		f2fs_do_map_lock(sbi, flag, false);
3322 	return err;
3323 }
3324 
3325 static int f2fs_write_begin(struct file *file, struct address_space *mapping,
3326 		loff_t pos, unsigned len, unsigned flags,
3327 		struct page **pagep, void **fsdata)
3328 {
3329 	struct inode *inode = mapping->host;
3330 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3331 	struct page *page = NULL;
3332 	pgoff_t index = ((unsigned long long) pos) >> PAGE_SHIFT;
3333 	bool need_balance = false, drop_atomic = false;
3334 	block_t blkaddr = NULL_ADDR;
3335 	int err = 0;
3336 
3337 	trace_f2fs_write_begin(inode, pos, len, flags);
3338 
3339 	if (!f2fs_is_checkpoint_ready(sbi)) {
3340 		err = -ENOSPC;
3341 		goto fail;
3342 	}
3343 
3344 	if ((f2fs_is_atomic_file(inode) &&
3345 			!f2fs_available_free_memory(sbi, INMEM_PAGES)) ||
3346 			is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
3347 		err = -ENOMEM;
3348 		drop_atomic = true;
3349 		goto fail;
3350 	}
3351 
3352 	/*
3353 	 * We should check this at this moment to avoid deadlock on inode page
3354 	 * and #0 page. The locking rule for inline_data conversion should be:
3355 	 * lock_page(page #0) -> lock_page(inode_page)
3356 	 */
3357 	if (index != 0) {
3358 		err = f2fs_convert_inline_inode(inode);
3359 		if (err)
3360 			goto fail;
3361 	}
3362 
3363 #ifdef CONFIG_F2FS_FS_COMPRESSION
3364 	if (f2fs_compressed_file(inode)) {
3365 		int ret;
3366 
3367 		*fsdata = NULL;
3368 
3369 		if (len == PAGE_SIZE)
3370 			goto repeat;
3371 
3372 		ret = f2fs_prepare_compress_overwrite(inode, pagep,
3373 							index, fsdata);
3374 		if (ret < 0) {
3375 			err = ret;
3376 			goto fail;
3377 		} else if (ret) {
3378 			return 0;
3379 		}
3380 	}
3381 #endif
3382 
3383 repeat:
3384 	/*
3385 	 * Do not use grab_cache_page_write_begin() to avoid deadlock due to
3386 	 * wait_for_stable_page. Will wait that below with our IO control.
3387 	 */
3388 	page = f2fs_pagecache_get_page(mapping, index,
3389 				FGP_LOCK | FGP_WRITE | FGP_CREAT, GFP_NOFS);
3390 	if (!page) {
3391 		err = -ENOMEM;
3392 		goto fail;
3393 	}
3394 
3395 	/* TODO: cluster can be compressed due to race with .writepage */
3396 
3397 	*pagep = page;
3398 
3399 	err = prepare_write_begin(sbi, page, pos, len,
3400 					&blkaddr, &need_balance);
3401 	if (err)
3402 		goto fail;
3403 
3404 	if (need_balance && !IS_NOQUOTA(inode) &&
3405 			has_not_enough_free_secs(sbi, 0, 0)) {
3406 		unlock_page(page);
3407 		f2fs_balance_fs(sbi, true);
3408 		lock_page(page);
3409 		if (page->mapping != mapping) {
3410 			/* The page got truncated from under us */
3411 			f2fs_put_page(page, 1);
3412 			goto repeat;
3413 		}
3414 	}
3415 
3416 	f2fs_wait_on_page_writeback(page, DATA, false, true);
3417 
3418 	if (len == PAGE_SIZE || PageUptodate(page))
3419 		return 0;
3420 
3421 	if (!(pos & (PAGE_SIZE - 1)) && (pos + len) >= i_size_read(inode) &&
3422 	    !f2fs_verity_in_progress(inode)) {
3423 		zero_user_segment(page, len, PAGE_SIZE);
3424 		return 0;
3425 	}
3426 
3427 	if (blkaddr == NEW_ADDR) {
3428 		zero_user_segment(page, 0, PAGE_SIZE);
3429 		SetPageUptodate(page);
3430 	} else {
3431 		if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3432 				DATA_GENERIC_ENHANCE_READ)) {
3433 			err = -EFSCORRUPTED;
3434 			goto fail;
3435 		}
3436 		err = f2fs_submit_page_read(inode, page, blkaddr, 0, true);
3437 		if (err)
3438 			goto fail;
3439 
3440 		lock_page(page);
3441 		if (unlikely(page->mapping != mapping)) {
3442 			f2fs_put_page(page, 1);
3443 			goto repeat;
3444 		}
3445 		if (unlikely(!PageUptodate(page))) {
3446 			err = -EIO;
3447 			goto fail;
3448 		}
3449 	}
3450 	return 0;
3451 
3452 fail:
3453 	f2fs_put_page(page, 1);
3454 	f2fs_write_failed(inode, pos + len);
3455 	if (drop_atomic)
3456 		f2fs_drop_inmem_pages_all(sbi, false);
3457 	return err;
3458 }
3459 
3460 static int f2fs_write_end(struct file *file,
3461 			struct address_space *mapping,
3462 			loff_t pos, unsigned len, unsigned copied,
3463 			struct page *page, void *fsdata)
3464 {
3465 	struct inode *inode = page->mapping->host;
3466 
3467 	trace_f2fs_write_end(inode, pos, len, copied);
3468 
3469 	/*
3470 	 * This should be come from len == PAGE_SIZE, and we expect copied
3471 	 * should be PAGE_SIZE. Otherwise, we treat it with zero copied and
3472 	 * let generic_perform_write() try to copy data again through copied=0.
3473 	 */
3474 	if (!PageUptodate(page)) {
3475 		if (unlikely(copied != len))
3476 			copied = 0;
3477 		else
3478 			SetPageUptodate(page);
3479 	}
3480 
3481 #ifdef CONFIG_F2FS_FS_COMPRESSION
3482 	/* overwrite compressed file */
3483 	if (f2fs_compressed_file(inode) && fsdata) {
3484 		f2fs_compress_write_end(inode, fsdata, page->index, copied);
3485 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3486 
3487 		if (pos + copied > i_size_read(inode) &&
3488 				!f2fs_verity_in_progress(inode))
3489 			f2fs_i_size_write(inode, pos + copied);
3490 		return copied;
3491 	}
3492 #endif
3493 
3494 	if (!copied)
3495 		goto unlock_out;
3496 
3497 	set_page_dirty(page);
3498 
3499 	if (pos + copied > i_size_read(inode) &&
3500 	    !f2fs_verity_in_progress(inode))
3501 		f2fs_i_size_write(inode, pos + copied);
3502 unlock_out:
3503 	f2fs_put_page(page, 1);
3504 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3505 	return copied;
3506 }
3507 
3508 void f2fs_invalidate_page(struct page *page, unsigned int offset,
3509 							unsigned int length)
3510 {
3511 	struct inode *inode = page->mapping->host;
3512 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3513 
3514 	if (inode->i_ino >= F2FS_ROOT_INO(sbi) &&
3515 		(offset % PAGE_SIZE || length != PAGE_SIZE))
3516 		return;
3517 
3518 	if (PageDirty(page)) {
3519 		if (inode->i_ino == F2FS_META_INO(sbi)) {
3520 			dec_page_count(sbi, F2FS_DIRTY_META);
3521 		} else if (inode->i_ino == F2FS_NODE_INO(sbi)) {
3522 			dec_page_count(sbi, F2FS_DIRTY_NODES);
3523 		} else {
3524 			inode_dec_dirty_pages(inode);
3525 			f2fs_remove_dirty_inode(inode);
3526 		}
3527 	}
3528 
3529 	clear_page_private_gcing(page);
3530 
3531 	if (test_opt(sbi, COMPRESS_CACHE) &&
3532 			inode->i_ino == F2FS_COMPRESS_INO(sbi))
3533 		clear_page_private_data(page);
3534 
3535 	if (page_private_atomic(page))
3536 		return f2fs_drop_inmem_page(inode, page);
3537 
3538 	detach_page_private(page);
3539 	set_page_private(page, 0);
3540 }
3541 
3542 int f2fs_release_page(struct page *page, gfp_t wait)
3543 {
3544 	/* If this is dirty page, keep PagePrivate */
3545 	if (PageDirty(page))
3546 		return 0;
3547 
3548 	/* This is atomic written page, keep Private */
3549 	if (page_private_atomic(page))
3550 		return 0;
3551 
3552 	if (test_opt(F2FS_P_SB(page), COMPRESS_CACHE)) {
3553 		struct inode *inode = page->mapping->host;
3554 
3555 		if (inode->i_ino == F2FS_COMPRESS_INO(F2FS_I_SB(inode)))
3556 			clear_page_private_data(page);
3557 	}
3558 
3559 	clear_page_private_gcing(page);
3560 
3561 	detach_page_private(page);
3562 	set_page_private(page, 0);
3563 	return 1;
3564 }
3565 
3566 static int f2fs_set_data_page_dirty(struct page *page)
3567 {
3568 	struct inode *inode = page_file_mapping(page)->host;
3569 
3570 	trace_f2fs_set_page_dirty(page, DATA);
3571 
3572 	if (!PageUptodate(page))
3573 		SetPageUptodate(page);
3574 	if (PageSwapCache(page))
3575 		return __set_page_dirty_nobuffers(page);
3576 
3577 	if (f2fs_is_atomic_file(inode) && !f2fs_is_commit_atomic_write(inode)) {
3578 		if (!page_private_atomic(page)) {
3579 			f2fs_register_inmem_page(inode, page);
3580 			return 1;
3581 		}
3582 		/*
3583 		 * Previously, this page has been registered, we just
3584 		 * return here.
3585 		 */
3586 		return 0;
3587 	}
3588 
3589 	if (!PageDirty(page)) {
3590 		__set_page_dirty_nobuffers(page);
3591 		f2fs_update_dirty_page(inode, page);
3592 		return 1;
3593 	}
3594 	return 0;
3595 }
3596 
3597 
3598 static sector_t f2fs_bmap_compress(struct inode *inode, sector_t block)
3599 {
3600 #ifdef CONFIG_F2FS_FS_COMPRESSION
3601 	struct dnode_of_data dn;
3602 	sector_t start_idx, blknr = 0;
3603 	int ret;
3604 
3605 	start_idx = round_down(block, F2FS_I(inode)->i_cluster_size);
3606 
3607 	set_new_dnode(&dn, inode, NULL, NULL, 0);
3608 	ret = f2fs_get_dnode_of_data(&dn, start_idx, LOOKUP_NODE);
3609 	if (ret)
3610 		return 0;
3611 
3612 	if (dn.data_blkaddr != COMPRESS_ADDR) {
3613 		dn.ofs_in_node += block - start_idx;
3614 		blknr = f2fs_data_blkaddr(&dn);
3615 		if (!__is_valid_data_blkaddr(blknr))
3616 			blknr = 0;
3617 	}
3618 
3619 	f2fs_put_dnode(&dn);
3620 	return blknr;
3621 #else
3622 	return 0;
3623 #endif
3624 }
3625 
3626 
3627 static sector_t f2fs_bmap(struct address_space *mapping, sector_t block)
3628 {
3629 	struct inode *inode = mapping->host;
3630 	sector_t blknr = 0;
3631 
3632 	if (f2fs_has_inline_data(inode))
3633 		goto out;
3634 
3635 	/* make sure allocating whole blocks */
3636 	if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
3637 		filemap_write_and_wait(mapping);
3638 
3639 	/* Block number less than F2FS MAX BLOCKS */
3640 	if (unlikely(block >= max_file_blocks(inode)))
3641 		goto out;
3642 
3643 	if (f2fs_compressed_file(inode)) {
3644 		blknr = f2fs_bmap_compress(inode, block);
3645 	} else {
3646 		struct f2fs_map_blocks map;
3647 
3648 		memset(&map, 0, sizeof(map));
3649 		map.m_lblk = block;
3650 		map.m_len = 1;
3651 		map.m_next_pgofs = NULL;
3652 		map.m_seg_type = NO_CHECK_TYPE;
3653 
3654 		if (!f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_BMAP))
3655 			blknr = map.m_pblk;
3656 	}
3657 out:
3658 	trace_f2fs_bmap(inode, block, blknr);
3659 	return blknr;
3660 }
3661 
3662 #ifdef CONFIG_MIGRATION
3663 #include <linux/migrate.h>
3664 
3665 int f2fs_migrate_page(struct address_space *mapping,
3666 		struct page *newpage, struct page *page, enum migrate_mode mode)
3667 {
3668 	int rc, extra_count;
3669 	struct f2fs_inode_info *fi = F2FS_I(mapping->host);
3670 	bool atomic_written = page_private_atomic(page);
3671 
3672 	BUG_ON(PageWriteback(page));
3673 
3674 	/* migrating an atomic written page is safe with the inmem_lock hold */
3675 	if (atomic_written) {
3676 		if (mode != MIGRATE_SYNC)
3677 			return -EBUSY;
3678 		if (!mutex_trylock(&fi->inmem_lock))
3679 			return -EAGAIN;
3680 	}
3681 
3682 	/* one extra reference was held for atomic_write page */
3683 	extra_count = atomic_written ? 1 : 0;
3684 	rc = migrate_page_move_mapping(mapping, newpage,
3685 				page, extra_count);
3686 	if (rc != MIGRATEPAGE_SUCCESS) {
3687 		if (atomic_written)
3688 			mutex_unlock(&fi->inmem_lock);
3689 		return rc;
3690 	}
3691 
3692 	if (atomic_written) {
3693 		struct inmem_pages *cur;
3694 
3695 		list_for_each_entry(cur, &fi->inmem_pages, list)
3696 			if (cur->page == page) {
3697 				cur->page = newpage;
3698 				break;
3699 			}
3700 		mutex_unlock(&fi->inmem_lock);
3701 		put_page(page);
3702 		get_page(newpage);
3703 	}
3704 
3705 	/* guarantee to start from no stale private field */
3706 	set_page_private(newpage, 0);
3707 	if (PagePrivate(page)) {
3708 		set_page_private(newpage, page_private(page));
3709 		SetPagePrivate(newpage);
3710 		get_page(newpage);
3711 
3712 		set_page_private(page, 0);
3713 		ClearPagePrivate(page);
3714 		put_page(page);
3715 	}
3716 
3717 	if (mode != MIGRATE_SYNC_NO_COPY)
3718 		migrate_page_copy(newpage, page);
3719 	else
3720 		migrate_page_states(newpage, page);
3721 
3722 	return MIGRATEPAGE_SUCCESS;
3723 }
3724 #endif
3725 
3726 #ifdef CONFIG_SWAP
3727 static int f2fs_migrate_blocks(struct inode *inode, block_t start_blk,
3728 							unsigned int blkcnt)
3729 {
3730 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3731 	unsigned int blkofs;
3732 	unsigned int blk_per_sec = BLKS_PER_SEC(sbi);
3733 	unsigned int secidx = start_blk / blk_per_sec;
3734 	unsigned int end_sec = secidx + blkcnt / blk_per_sec;
3735 	int ret = 0;
3736 
3737 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3738 	filemap_invalidate_lock(inode->i_mapping);
3739 
3740 	set_inode_flag(inode, FI_ALIGNED_WRITE);
3741 
3742 	for (; secidx < end_sec; secidx++) {
3743 		down_write(&sbi->pin_sem);
3744 
3745 		f2fs_lock_op(sbi);
3746 		f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
3747 		f2fs_unlock_op(sbi);
3748 
3749 		set_inode_flag(inode, FI_DO_DEFRAG);
3750 
3751 		for (blkofs = 0; blkofs < blk_per_sec; blkofs++) {
3752 			struct page *page;
3753 			unsigned int blkidx = secidx * blk_per_sec + blkofs;
3754 
3755 			page = f2fs_get_lock_data_page(inode, blkidx, true);
3756 			if (IS_ERR(page)) {
3757 				up_write(&sbi->pin_sem);
3758 				ret = PTR_ERR(page);
3759 				goto done;
3760 			}
3761 
3762 			set_page_dirty(page);
3763 			f2fs_put_page(page, 1);
3764 		}
3765 
3766 		clear_inode_flag(inode, FI_DO_DEFRAG);
3767 
3768 		ret = filemap_fdatawrite(inode->i_mapping);
3769 
3770 		up_write(&sbi->pin_sem);
3771 
3772 		if (ret)
3773 			break;
3774 	}
3775 
3776 done:
3777 	clear_inode_flag(inode, FI_DO_DEFRAG);
3778 	clear_inode_flag(inode, FI_ALIGNED_WRITE);
3779 
3780 	filemap_invalidate_unlock(inode->i_mapping);
3781 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3782 
3783 	return ret;
3784 }
3785 
3786 static int check_swap_activate(struct swap_info_struct *sis,
3787 				struct file *swap_file, sector_t *span)
3788 {
3789 	struct address_space *mapping = swap_file->f_mapping;
3790 	struct inode *inode = mapping->host;
3791 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3792 	sector_t cur_lblock;
3793 	sector_t last_lblock;
3794 	sector_t pblock;
3795 	sector_t lowest_pblock = -1;
3796 	sector_t highest_pblock = 0;
3797 	int nr_extents = 0;
3798 	unsigned long nr_pblocks;
3799 	unsigned int blks_per_sec = BLKS_PER_SEC(sbi);
3800 	unsigned int sec_blks_mask = BLKS_PER_SEC(sbi) - 1;
3801 	unsigned int not_aligned = 0;
3802 	int ret = 0;
3803 
3804 	/*
3805 	 * Map all the blocks into the extent list.  This code doesn't try
3806 	 * to be very smart.
3807 	 */
3808 	cur_lblock = 0;
3809 	last_lblock = bytes_to_blks(inode, i_size_read(inode));
3810 
3811 	while (cur_lblock < last_lblock && cur_lblock < sis->max) {
3812 		struct f2fs_map_blocks map;
3813 retry:
3814 		cond_resched();
3815 
3816 		memset(&map, 0, sizeof(map));
3817 		map.m_lblk = cur_lblock;
3818 		map.m_len = last_lblock - cur_lblock;
3819 		map.m_next_pgofs = NULL;
3820 		map.m_next_extent = NULL;
3821 		map.m_seg_type = NO_CHECK_TYPE;
3822 		map.m_may_create = false;
3823 
3824 		ret = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_FIEMAP);
3825 		if (ret)
3826 			goto out;
3827 
3828 		/* hole */
3829 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
3830 			f2fs_err(sbi, "Swapfile has holes");
3831 			ret = -EINVAL;
3832 			goto out;
3833 		}
3834 
3835 		pblock = map.m_pblk;
3836 		nr_pblocks = map.m_len;
3837 
3838 		if ((pblock - SM_I(sbi)->main_blkaddr) & sec_blks_mask ||
3839 				nr_pblocks & sec_blks_mask) {
3840 			not_aligned++;
3841 
3842 			nr_pblocks = roundup(nr_pblocks, blks_per_sec);
3843 			if (cur_lblock + nr_pblocks > sis->max)
3844 				nr_pblocks -= blks_per_sec;
3845 
3846 			if (!nr_pblocks) {
3847 				/* this extent is last one */
3848 				nr_pblocks = map.m_len;
3849 				f2fs_warn(sbi, "Swapfile: last extent is not aligned to section");
3850 				goto next;
3851 			}
3852 
3853 			ret = f2fs_migrate_blocks(inode, cur_lblock,
3854 							nr_pblocks);
3855 			if (ret)
3856 				goto out;
3857 			goto retry;
3858 		}
3859 next:
3860 		if (cur_lblock + nr_pblocks >= sis->max)
3861 			nr_pblocks = sis->max - cur_lblock;
3862 
3863 		if (cur_lblock) {	/* exclude the header page */
3864 			if (pblock < lowest_pblock)
3865 				lowest_pblock = pblock;
3866 			if (pblock + nr_pblocks - 1 > highest_pblock)
3867 				highest_pblock = pblock + nr_pblocks - 1;
3868 		}
3869 
3870 		/*
3871 		 * We found a PAGE_SIZE-length, PAGE_SIZE-aligned run of blocks
3872 		 */
3873 		ret = add_swap_extent(sis, cur_lblock, nr_pblocks, pblock);
3874 		if (ret < 0)
3875 			goto out;
3876 		nr_extents += ret;
3877 		cur_lblock += nr_pblocks;
3878 	}
3879 	ret = nr_extents;
3880 	*span = 1 + highest_pblock - lowest_pblock;
3881 	if (cur_lblock == 0)
3882 		cur_lblock = 1;	/* force Empty message */
3883 	sis->max = cur_lblock;
3884 	sis->pages = cur_lblock - 1;
3885 	sis->highest_bit = cur_lblock - 1;
3886 out:
3887 	if (not_aligned)
3888 		f2fs_warn(sbi, "Swapfile (%u) is not align to section: 1) creat(), 2) ioctl(F2FS_IOC_SET_PIN_FILE), 3) fallocate(%u * N)",
3889 			  not_aligned, blks_per_sec * F2FS_BLKSIZE);
3890 	return ret;
3891 }
3892 
3893 static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
3894 				sector_t *span)
3895 {
3896 	struct inode *inode = file_inode(file);
3897 	int ret;
3898 
3899 	if (!S_ISREG(inode->i_mode))
3900 		return -EINVAL;
3901 
3902 	if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3903 		return -EROFS;
3904 
3905 	if (f2fs_lfs_mode(F2FS_I_SB(inode))) {
3906 		f2fs_err(F2FS_I_SB(inode),
3907 			"Swapfile not supported in LFS mode");
3908 		return -EINVAL;
3909 	}
3910 
3911 	ret = f2fs_convert_inline_inode(inode);
3912 	if (ret)
3913 		return ret;
3914 
3915 	if (!f2fs_disable_compressed_file(inode))
3916 		return -EINVAL;
3917 
3918 	f2fs_precache_extents(inode);
3919 
3920 	ret = check_swap_activate(sis, file, span);
3921 	if (ret < 0)
3922 		return ret;
3923 
3924 	set_inode_flag(inode, FI_PIN_FILE);
3925 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3926 	return ret;
3927 }
3928 
3929 static void f2fs_swap_deactivate(struct file *file)
3930 {
3931 	struct inode *inode = file_inode(file);
3932 
3933 	clear_inode_flag(inode, FI_PIN_FILE);
3934 }
3935 #else
3936 static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
3937 				sector_t *span)
3938 {
3939 	return -EOPNOTSUPP;
3940 }
3941 
3942 static void f2fs_swap_deactivate(struct file *file)
3943 {
3944 }
3945 #endif
3946 
3947 const struct address_space_operations f2fs_dblock_aops = {
3948 	.readpage	= f2fs_read_data_page,
3949 	.readahead	= f2fs_readahead,
3950 	.writepage	= f2fs_write_data_page,
3951 	.writepages	= f2fs_write_data_pages,
3952 	.write_begin	= f2fs_write_begin,
3953 	.write_end	= f2fs_write_end,
3954 	.set_page_dirty	= f2fs_set_data_page_dirty,
3955 	.invalidatepage	= f2fs_invalidate_page,
3956 	.releasepage	= f2fs_release_page,
3957 	.direct_IO	= noop_direct_IO,
3958 	.bmap		= f2fs_bmap,
3959 	.swap_activate  = f2fs_swap_activate,
3960 	.swap_deactivate = f2fs_swap_deactivate,
3961 #ifdef CONFIG_MIGRATION
3962 	.migratepage    = f2fs_migrate_page,
3963 #endif
3964 };
3965 
3966 void f2fs_clear_page_cache_dirty_tag(struct page *page)
3967 {
3968 	struct address_space *mapping = page_mapping(page);
3969 	unsigned long flags;
3970 
3971 	xa_lock_irqsave(&mapping->i_pages, flags);
3972 	__xa_clear_mark(&mapping->i_pages, page_index(page),
3973 						PAGECACHE_TAG_DIRTY);
3974 	xa_unlock_irqrestore(&mapping->i_pages, flags);
3975 }
3976 
3977 int __init f2fs_init_post_read_processing(void)
3978 {
3979 	bio_post_read_ctx_cache =
3980 		kmem_cache_create("f2fs_bio_post_read_ctx",
3981 				  sizeof(struct bio_post_read_ctx), 0, 0, NULL);
3982 	if (!bio_post_read_ctx_cache)
3983 		goto fail;
3984 	bio_post_read_ctx_pool =
3985 		mempool_create_slab_pool(NUM_PREALLOC_POST_READ_CTXS,
3986 					 bio_post_read_ctx_cache);
3987 	if (!bio_post_read_ctx_pool)
3988 		goto fail_free_cache;
3989 	return 0;
3990 
3991 fail_free_cache:
3992 	kmem_cache_destroy(bio_post_read_ctx_cache);
3993 fail:
3994 	return -ENOMEM;
3995 }
3996 
3997 void f2fs_destroy_post_read_processing(void)
3998 {
3999 	mempool_destroy(bio_post_read_ctx_pool);
4000 	kmem_cache_destroy(bio_post_read_ctx_cache);
4001 }
4002 
4003 int f2fs_init_post_read_wq(struct f2fs_sb_info *sbi)
4004 {
4005 	if (!f2fs_sb_has_encrypt(sbi) &&
4006 		!f2fs_sb_has_verity(sbi) &&
4007 		!f2fs_sb_has_compression(sbi))
4008 		return 0;
4009 
4010 	sbi->post_read_wq = alloc_workqueue("f2fs_post_read_wq",
4011 						 WQ_UNBOUND | WQ_HIGHPRI,
4012 						 num_online_cpus());
4013 	if (!sbi->post_read_wq)
4014 		return -ENOMEM;
4015 	return 0;
4016 }
4017 
4018 void f2fs_destroy_post_read_wq(struct f2fs_sb_info *sbi)
4019 {
4020 	if (sbi->post_read_wq)
4021 		destroy_workqueue(sbi->post_read_wq);
4022 }
4023 
4024 int __init f2fs_init_bio_entry_cache(void)
4025 {
4026 	bio_entry_slab = f2fs_kmem_cache_create("f2fs_bio_entry_slab",
4027 			sizeof(struct bio_entry));
4028 	if (!bio_entry_slab)
4029 		return -ENOMEM;
4030 	return 0;
4031 }
4032 
4033 void f2fs_destroy_bio_entry_cache(void)
4034 {
4035 	kmem_cache_destroy(bio_entry_slab);
4036 }
4037 
4038 static int f2fs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
4039 			    unsigned int flags, struct iomap *iomap,
4040 			    struct iomap *srcmap)
4041 {
4042 	struct f2fs_map_blocks map = {};
4043 	pgoff_t next_pgofs = 0;
4044 	int err;
4045 
4046 	map.m_lblk = bytes_to_blks(inode, offset);
4047 	map.m_len = bytes_to_blks(inode, offset + length - 1) - map.m_lblk + 1;
4048 	map.m_next_pgofs = &next_pgofs;
4049 	map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4050 	if (flags & IOMAP_WRITE)
4051 		map.m_may_create = true;
4052 
4053 	err = f2fs_map_blocks(inode, &map, flags & IOMAP_WRITE,
4054 			      F2FS_GET_BLOCK_DIO);
4055 	if (err)
4056 		return err;
4057 
4058 	iomap->offset = blks_to_bytes(inode, map.m_lblk);
4059 
4060 	if (map.m_flags & (F2FS_MAP_MAPPED | F2FS_MAP_UNWRITTEN)) {
4061 		iomap->length = blks_to_bytes(inode, map.m_len);
4062 		if (map.m_flags & F2FS_MAP_MAPPED) {
4063 			iomap->type = IOMAP_MAPPED;
4064 			iomap->flags |= IOMAP_F_MERGED;
4065 		} else {
4066 			iomap->type = IOMAP_UNWRITTEN;
4067 		}
4068 		if (WARN_ON_ONCE(!__is_valid_data_blkaddr(map.m_pblk)))
4069 			return -EINVAL;
4070 
4071 		iomap->bdev = map.m_bdev;
4072 		iomap->addr = blks_to_bytes(inode, map.m_pblk);
4073 	} else {
4074 		iomap->length = blks_to_bytes(inode, next_pgofs) -
4075 				iomap->offset;
4076 		iomap->type = IOMAP_HOLE;
4077 		iomap->addr = IOMAP_NULL_ADDR;
4078 	}
4079 
4080 	if (map.m_flags & F2FS_MAP_NEW)
4081 		iomap->flags |= IOMAP_F_NEW;
4082 	if ((inode->i_state & I_DIRTY_DATASYNC) ||
4083 	    offset + length > i_size_read(inode))
4084 		iomap->flags |= IOMAP_F_DIRTY;
4085 
4086 	return 0;
4087 }
4088 
4089 const struct iomap_ops f2fs_iomap_ops = {
4090 	.iomap_begin	= f2fs_iomap_begin,
4091 };
4092