xref: /openbmc/linux/fs/f2fs/data.c (revision fb4a5dfc)
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/sched/signal.h>
22 #include <linux/fiemap.h>
23 #include <linux/iomap.h>
24 
25 #include "f2fs.h"
26 #include "node.h"
27 #include "segment.h"
28 #include "iostat.h"
29 #include <trace/events/f2fs.h>
30 
31 #define NUM_PREALLOC_POST_READ_CTXS	128
32 
33 static struct kmem_cache *bio_post_read_ctx_cache;
34 static struct kmem_cache *bio_entry_slab;
35 static mempool_t *bio_post_read_ctx_pool;
36 static struct bio_set f2fs_bioset;
37 
38 #define	F2FS_BIO_POOL_SIZE	NR_CURSEG_TYPE
39 
40 int __init f2fs_init_bioset(void)
41 {
42 	if (bioset_init(&f2fs_bioset, F2FS_BIO_POOL_SIZE,
43 					0, BIOSET_NEED_BVECS))
44 		return -ENOMEM;
45 	return 0;
46 }
47 
48 void f2fs_destroy_bioset(void)
49 {
50 	bioset_exit(&f2fs_bioset);
51 }
52 
53 static bool __is_cp_guaranteed(struct page *page)
54 {
55 	struct address_space *mapping = page->mapping;
56 	struct inode *inode;
57 	struct f2fs_sb_info *sbi;
58 
59 	if (!mapping)
60 		return false;
61 
62 	inode = mapping->host;
63 	sbi = F2FS_I_SB(inode);
64 
65 	if (inode->i_ino == F2FS_META_INO(sbi) ||
66 			inode->i_ino == F2FS_NODE_INO(sbi) ||
67 			S_ISDIR(inode->i_mode))
68 		return true;
69 
70 	if (f2fs_is_compressed_page(page))
71 		return false;
72 	if ((S_ISREG(inode->i_mode) && IS_NOQUOTA(inode)) ||
73 			page_private_gcing(page))
74 		return true;
75 	return false;
76 }
77 
78 static enum count_type __read_io_type(struct page *page)
79 {
80 	struct address_space *mapping = page_file_mapping(page);
81 
82 	if (mapping) {
83 		struct inode *inode = mapping->host;
84 		struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
85 
86 		if (inode->i_ino == F2FS_META_INO(sbi))
87 			return F2FS_RD_META;
88 
89 		if (inode->i_ino == F2FS_NODE_INO(sbi))
90 			return F2FS_RD_NODE;
91 	}
92 	return F2FS_RD_DATA;
93 }
94 
95 /* postprocessing steps for read bios */
96 enum bio_post_read_step {
97 #ifdef CONFIG_FS_ENCRYPTION
98 	STEP_DECRYPT	= 1 << 0,
99 #else
100 	STEP_DECRYPT	= 0,	/* compile out the decryption-related code */
101 #endif
102 #ifdef CONFIG_F2FS_FS_COMPRESSION
103 	STEP_DECOMPRESS	= 1 << 1,
104 #else
105 	STEP_DECOMPRESS	= 0,	/* compile out the decompression-related code */
106 #endif
107 #ifdef CONFIG_FS_VERITY
108 	STEP_VERITY	= 1 << 2,
109 #else
110 	STEP_VERITY	= 0,	/* compile out the verity-related code */
111 #endif
112 };
113 
114 struct bio_post_read_ctx {
115 	struct bio *bio;
116 	struct f2fs_sb_info *sbi;
117 	struct work_struct work;
118 	unsigned int enabled_steps;
119 	block_t fs_blkaddr;
120 };
121 
122 static void f2fs_finish_read_bio(struct bio *bio, bool in_task)
123 {
124 	struct bio_vec *bv;
125 	struct bvec_iter_all iter_all;
126 
127 	/*
128 	 * Update and unlock the bio's pagecache pages, and put the
129 	 * decompression context for any compressed pages.
130 	 */
131 	bio_for_each_segment_all(bv, bio, iter_all) {
132 		struct page *page = bv->bv_page;
133 
134 		if (f2fs_is_compressed_page(page)) {
135 			if (bio->bi_status)
136 				f2fs_end_read_compressed_page(page, true, 0,
137 							in_task);
138 			f2fs_put_page_dic(page, in_task);
139 			continue;
140 		}
141 
142 		/* PG_error was set if verity failed. */
143 		if (bio->bi_status || PageError(page)) {
144 			ClearPageUptodate(page);
145 			/* will re-read again later */
146 			ClearPageError(page);
147 		} else {
148 			SetPageUptodate(page);
149 		}
150 		dec_page_count(F2FS_P_SB(page), __read_io_type(page));
151 		unlock_page(page);
152 	}
153 
154 	if (bio->bi_private)
155 		mempool_free(bio->bi_private, bio_post_read_ctx_pool);
156 	bio_put(bio);
157 }
158 
159 static void f2fs_verify_bio(struct work_struct *work)
160 {
161 	struct bio_post_read_ctx *ctx =
162 		container_of(work, struct bio_post_read_ctx, work);
163 	struct bio *bio = ctx->bio;
164 	bool may_have_compressed_pages = (ctx->enabled_steps & STEP_DECOMPRESS);
165 
166 	/*
167 	 * fsverity_verify_bio() may call readahead() again, and while verity
168 	 * will be disabled for this, decryption and/or decompression may still
169 	 * be needed, resulting in another bio_post_read_ctx being allocated.
170 	 * So to prevent deadlocks we need to release the current ctx to the
171 	 * mempool first.  This assumes that verity is the last post-read step.
172 	 */
173 	mempool_free(ctx, bio_post_read_ctx_pool);
174 	bio->bi_private = NULL;
175 
176 	/*
177 	 * Verify the bio's pages with fs-verity.  Exclude compressed pages,
178 	 * as those were handled separately by f2fs_end_read_compressed_page().
179 	 */
180 	if (may_have_compressed_pages) {
181 		struct bio_vec *bv;
182 		struct bvec_iter_all iter_all;
183 
184 		bio_for_each_segment_all(bv, bio, iter_all) {
185 			struct page *page = bv->bv_page;
186 
187 			if (!f2fs_is_compressed_page(page) &&
188 			    !fsverity_verify_page(page))
189 				SetPageError(page);
190 		}
191 	} else {
192 		fsverity_verify_bio(bio);
193 	}
194 
195 	f2fs_finish_read_bio(bio, true);
196 }
197 
198 /*
199  * If the bio's data needs to be verified with fs-verity, then enqueue the
200  * verity work for the bio.  Otherwise finish the bio now.
201  *
202  * Note that to avoid deadlocks, the verity work can't be done on the
203  * decryption/decompression workqueue.  This is because verifying the data pages
204  * can involve reading verity metadata pages from the file, and these verity
205  * metadata pages may be encrypted and/or compressed.
206  */
207 static void f2fs_verify_and_finish_bio(struct bio *bio, bool in_task)
208 {
209 	struct bio_post_read_ctx *ctx = bio->bi_private;
210 
211 	if (ctx && (ctx->enabled_steps & STEP_VERITY)) {
212 		INIT_WORK(&ctx->work, f2fs_verify_bio);
213 		fsverity_enqueue_verify_work(&ctx->work);
214 	} else {
215 		f2fs_finish_read_bio(bio, in_task);
216 	}
217 }
218 
219 /*
220  * Handle STEP_DECOMPRESS by decompressing any compressed clusters whose last
221  * remaining page was read by @ctx->bio.
222  *
223  * Note that a bio may span clusters (even a mix of compressed and uncompressed
224  * clusters) or be for just part of a cluster.  STEP_DECOMPRESS just indicates
225  * that the bio includes at least one compressed page.  The actual decompression
226  * is done on a per-cluster basis, not a per-bio basis.
227  */
228 static void f2fs_handle_step_decompress(struct bio_post_read_ctx *ctx,
229 		bool in_task)
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 		if (f2fs_is_compressed_page(page))
240 			f2fs_end_read_compressed_page(page, false, blkaddr,
241 						      in_task);
242 		else
243 			all_compressed = false;
244 
245 		blkaddr++;
246 	}
247 
248 	/*
249 	 * Optimization: if all the bio's pages are compressed, then scheduling
250 	 * the per-bio verity work is unnecessary, as verity will be fully
251 	 * handled at the compression cluster level.
252 	 */
253 	if (all_compressed)
254 		ctx->enabled_steps &= ~STEP_VERITY;
255 }
256 
257 static void f2fs_post_read_work(struct work_struct *work)
258 {
259 	struct bio_post_read_ctx *ctx =
260 		container_of(work, struct bio_post_read_ctx, work);
261 	struct bio *bio = ctx->bio;
262 
263 	if ((ctx->enabled_steps & STEP_DECRYPT) && !fscrypt_decrypt_bio(bio)) {
264 		f2fs_finish_read_bio(bio, true);
265 		return;
266 	}
267 
268 	if (ctx->enabled_steps & STEP_DECOMPRESS)
269 		f2fs_handle_step_decompress(ctx, true);
270 
271 	f2fs_verify_and_finish_bio(bio, true);
272 }
273 
274 static void f2fs_read_end_io(struct bio *bio)
275 {
276 	struct f2fs_sb_info *sbi = F2FS_P_SB(bio_first_page_all(bio));
277 	struct bio_post_read_ctx *ctx;
278 	bool intask = in_task();
279 
280 	iostat_update_and_unbind_ctx(bio, 0);
281 	ctx = bio->bi_private;
282 
283 	if (time_to_inject(sbi, FAULT_READ_IO)) {
284 		f2fs_show_injection_info(sbi, FAULT_READ_IO);
285 		bio->bi_status = BLK_STS_IOERR;
286 	}
287 
288 	if (bio->bi_status) {
289 		f2fs_finish_read_bio(bio, intask);
290 		return;
291 	}
292 
293 	if (ctx) {
294 		unsigned int enabled_steps = ctx->enabled_steps &
295 					(STEP_DECRYPT | STEP_DECOMPRESS);
296 
297 		/*
298 		 * If we have only decompression step between decompression and
299 		 * decrypt, we don't need post processing for this.
300 		 */
301 		if (enabled_steps == STEP_DECOMPRESS &&
302 				!f2fs_low_mem_mode(sbi)) {
303 			f2fs_handle_step_decompress(ctx, intask);
304 		} else if (enabled_steps) {
305 			INIT_WORK(&ctx->work, f2fs_post_read_work);
306 			queue_work(ctx->sbi->post_read_wq, &ctx->work);
307 			return;
308 		}
309 	}
310 
311 	f2fs_verify_and_finish_bio(bio, intask);
312 }
313 
314 static void f2fs_write_end_io(struct bio *bio)
315 {
316 	struct f2fs_sb_info *sbi;
317 	struct bio_vec *bvec;
318 	struct bvec_iter_all iter_all;
319 
320 	iostat_update_and_unbind_ctx(bio, 1);
321 	sbi = bio->bi_private;
322 
323 	if (time_to_inject(sbi, FAULT_WRITE_IO)) {
324 		f2fs_show_injection_info(sbi, FAULT_WRITE_IO);
325 		bio->bi_status = BLK_STS_IOERR;
326 	}
327 
328 	bio_for_each_segment_all(bvec, bio, iter_all) {
329 		struct page *page = bvec->bv_page;
330 		enum count_type type = WB_DATA_TYPE(page);
331 
332 		if (page_private_dummy(page)) {
333 			clear_page_private_dummy(page);
334 			unlock_page(page);
335 			mempool_free(page, sbi->write_io_dummy);
336 
337 			if (unlikely(bio->bi_status))
338 				f2fs_stop_checkpoint(sbi, true);
339 			continue;
340 		}
341 
342 		fscrypt_finalize_bounce_page(&page);
343 
344 #ifdef CONFIG_F2FS_FS_COMPRESSION
345 		if (f2fs_is_compressed_page(page)) {
346 			f2fs_compress_write_end_io(bio, page);
347 			continue;
348 		}
349 #endif
350 
351 		if (unlikely(bio->bi_status)) {
352 			mapping_set_error(page->mapping, -EIO);
353 			if (type == F2FS_WB_CP_DATA)
354 				f2fs_stop_checkpoint(sbi, true);
355 		}
356 
357 		f2fs_bug_on(sbi, page->mapping == NODE_MAPPING(sbi) &&
358 					page->index != nid_of_node(page));
359 
360 		dec_page_count(sbi, type);
361 		if (f2fs_in_warm_node_list(sbi, page))
362 			f2fs_del_fsync_node_entry(sbi, page);
363 		clear_page_private_gcing(page);
364 		end_page_writeback(page);
365 	}
366 	if (!get_pages(sbi, F2FS_WB_CP_DATA) &&
367 				wq_has_sleeper(&sbi->cp_wait))
368 		wake_up(&sbi->cp_wait);
369 
370 	bio_put(bio);
371 }
372 
373 struct block_device *f2fs_target_device(struct f2fs_sb_info *sbi,
374 		block_t blk_addr, sector_t *sector)
375 {
376 	struct block_device *bdev = sbi->sb->s_bdev;
377 	int i;
378 
379 	if (f2fs_is_multi_device(sbi)) {
380 		for (i = 0; i < sbi->s_ndevs; i++) {
381 			if (FDEV(i).start_blk <= blk_addr &&
382 			    FDEV(i).end_blk >= blk_addr) {
383 				blk_addr -= FDEV(i).start_blk;
384 				bdev = FDEV(i).bdev;
385 				break;
386 			}
387 		}
388 	}
389 
390 	if (sector)
391 		*sector = SECTOR_FROM_BLOCK(blk_addr);
392 	return bdev;
393 }
394 
395 int f2fs_target_device_index(struct f2fs_sb_info *sbi, block_t blkaddr)
396 {
397 	int i;
398 
399 	if (!f2fs_is_multi_device(sbi))
400 		return 0;
401 
402 	for (i = 0; i < sbi->s_ndevs; i++)
403 		if (FDEV(i).start_blk <= blkaddr && FDEV(i).end_blk >= blkaddr)
404 			return i;
405 	return 0;
406 }
407 
408 static blk_opf_t f2fs_io_flags(struct f2fs_io_info *fio)
409 {
410 	unsigned int temp_mask = (1 << NR_TEMP_TYPE) - 1;
411 	unsigned int fua_flag, meta_flag, io_flag;
412 	blk_opf_t op_flags = 0;
413 
414 	if (fio->op != REQ_OP_WRITE)
415 		return 0;
416 	if (fio->type == DATA)
417 		io_flag = fio->sbi->data_io_flag;
418 	else if (fio->type == NODE)
419 		io_flag = fio->sbi->node_io_flag;
420 	else
421 		return 0;
422 
423 	fua_flag = io_flag & temp_mask;
424 	meta_flag = (io_flag >> NR_TEMP_TYPE) & temp_mask;
425 
426 	/*
427 	 * data/node io flag bits per temp:
428 	 *      REQ_META     |      REQ_FUA      |
429 	 *    5 |    4 |   3 |    2 |    1 |   0 |
430 	 * Cold | Warm | Hot | Cold | Warm | Hot |
431 	 */
432 	if ((1 << fio->temp) & meta_flag)
433 		op_flags |= REQ_META;
434 	if ((1 << fio->temp) & fua_flag)
435 		op_flags |= REQ_FUA;
436 	return op_flags;
437 }
438 
439 static struct bio *__bio_alloc(struct f2fs_io_info *fio, int npages)
440 {
441 	struct f2fs_sb_info *sbi = fio->sbi;
442 	struct block_device *bdev;
443 	sector_t sector;
444 	struct bio *bio;
445 
446 	bdev = f2fs_target_device(sbi, fio->new_blkaddr, &sector);
447 	bio = bio_alloc_bioset(bdev, npages,
448 				fio->op | fio->op_flags | f2fs_io_flags(fio),
449 				GFP_NOIO, &f2fs_bioset);
450 	bio->bi_iter.bi_sector = sector;
451 	if (is_read_io(fio->op)) {
452 		bio->bi_end_io = f2fs_read_end_io;
453 		bio->bi_private = NULL;
454 	} else {
455 		bio->bi_end_io = f2fs_write_end_io;
456 		bio->bi_private = sbi;
457 	}
458 	iostat_alloc_and_bind_ctx(sbi, bio, NULL);
459 
460 	if (fio->io_wbc)
461 		wbc_init_bio(fio->io_wbc, bio);
462 
463 	return bio;
464 }
465 
466 static void f2fs_set_bio_crypt_ctx(struct bio *bio, const struct inode *inode,
467 				  pgoff_t first_idx,
468 				  const struct f2fs_io_info *fio,
469 				  gfp_t gfp_mask)
470 {
471 	/*
472 	 * The f2fs garbage collector sets ->encrypted_page when it wants to
473 	 * read/write raw data without encryption.
474 	 */
475 	if (!fio || !fio->encrypted_page)
476 		fscrypt_set_bio_crypt_ctx(bio, inode, first_idx, gfp_mask);
477 }
478 
479 static bool f2fs_crypt_mergeable_bio(struct bio *bio, const struct inode *inode,
480 				     pgoff_t next_idx,
481 				     const struct f2fs_io_info *fio)
482 {
483 	/*
484 	 * The f2fs garbage collector sets ->encrypted_page when it wants to
485 	 * read/write raw data without encryption.
486 	 */
487 	if (fio && fio->encrypted_page)
488 		return !bio_has_crypt_ctx(bio);
489 
490 	return fscrypt_mergeable_bio(bio, inode, next_idx);
491 }
492 
493 static inline void __submit_bio(struct f2fs_sb_info *sbi,
494 				struct bio *bio, enum page_type type)
495 {
496 	if (!is_read_io(bio_op(bio))) {
497 		unsigned int start;
498 
499 		if (type != DATA && type != NODE)
500 			goto submit_io;
501 
502 		if (f2fs_lfs_mode(sbi) && current->plug)
503 			blk_finish_plug(current->plug);
504 
505 		if (!F2FS_IO_ALIGNED(sbi))
506 			goto submit_io;
507 
508 		start = bio->bi_iter.bi_size >> F2FS_BLKSIZE_BITS;
509 		start %= F2FS_IO_SIZE(sbi);
510 
511 		if (start == 0)
512 			goto submit_io;
513 
514 		/* fill dummy pages */
515 		for (; start < F2FS_IO_SIZE(sbi); start++) {
516 			struct page *page =
517 				mempool_alloc(sbi->write_io_dummy,
518 					      GFP_NOIO | __GFP_NOFAIL);
519 			f2fs_bug_on(sbi, !page);
520 
521 			lock_page(page);
522 
523 			zero_user_segment(page, 0, PAGE_SIZE);
524 			set_page_private_dummy(page);
525 
526 			if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE)
527 				f2fs_bug_on(sbi, 1);
528 		}
529 		/*
530 		 * In the NODE case, we lose next block address chain. So, we
531 		 * need to do checkpoint in f2fs_sync_file.
532 		 */
533 		if (type == NODE)
534 			set_sbi_flag(sbi, SBI_NEED_CP);
535 	}
536 submit_io:
537 	if (is_read_io(bio_op(bio)))
538 		trace_f2fs_submit_read_bio(sbi->sb, type, bio);
539 	else
540 		trace_f2fs_submit_write_bio(sbi->sb, type, bio);
541 
542 	iostat_update_submit_ctx(bio, type);
543 	submit_bio(bio);
544 }
545 
546 void f2fs_submit_bio(struct f2fs_sb_info *sbi,
547 				struct bio *bio, enum page_type type)
548 {
549 	__submit_bio(sbi, bio, type);
550 }
551 
552 static void __submit_merged_bio(struct f2fs_bio_info *io)
553 {
554 	struct f2fs_io_info *fio = &io->fio;
555 
556 	if (!io->bio)
557 		return;
558 
559 	if (is_read_io(fio->op))
560 		trace_f2fs_prepare_read_bio(io->sbi->sb, fio->type, io->bio);
561 	else
562 		trace_f2fs_prepare_write_bio(io->sbi->sb, fio->type, io->bio);
563 
564 	__submit_bio(io->sbi, io->bio, fio->type);
565 	io->bio = NULL;
566 }
567 
568 static bool __has_merged_page(struct bio *bio, struct inode *inode,
569 						struct page *page, nid_t ino)
570 {
571 	struct bio_vec *bvec;
572 	struct bvec_iter_all iter_all;
573 
574 	if (!bio)
575 		return false;
576 
577 	if (!inode && !page && !ino)
578 		return true;
579 
580 	bio_for_each_segment_all(bvec, bio, iter_all) {
581 		struct page *target = bvec->bv_page;
582 
583 		if (fscrypt_is_bounce_page(target)) {
584 			target = fscrypt_pagecache_page(target);
585 			if (IS_ERR(target))
586 				continue;
587 		}
588 		if (f2fs_is_compressed_page(target)) {
589 			target = f2fs_compress_control_page(target);
590 			if (IS_ERR(target))
591 				continue;
592 		}
593 
594 		if (inode && inode == target->mapping->host)
595 			return true;
596 		if (page && page == target)
597 			return true;
598 		if (ino && ino == ino_of_node(target))
599 			return true;
600 	}
601 
602 	return false;
603 }
604 
605 int f2fs_init_write_merge_io(struct f2fs_sb_info *sbi)
606 {
607 	int i;
608 
609 	for (i = 0; i < NR_PAGE_TYPE; i++) {
610 		int n = (i == META) ? 1 : NR_TEMP_TYPE;
611 		int j;
612 
613 		sbi->write_io[i] = f2fs_kmalloc(sbi,
614 				array_size(n, sizeof(struct f2fs_bio_info)),
615 				GFP_KERNEL);
616 		if (!sbi->write_io[i])
617 			return -ENOMEM;
618 
619 		for (j = HOT; j < n; j++) {
620 			init_f2fs_rwsem(&sbi->write_io[i][j].io_rwsem);
621 			sbi->write_io[i][j].sbi = sbi;
622 			sbi->write_io[i][j].bio = NULL;
623 			spin_lock_init(&sbi->write_io[i][j].io_lock);
624 			INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
625 			INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
626 			init_f2fs_rwsem(&sbi->write_io[i][j].bio_list_lock);
627 		}
628 	}
629 
630 	return 0;
631 }
632 
633 static void __f2fs_submit_merged_write(struct f2fs_sb_info *sbi,
634 				enum page_type type, enum temp_type temp)
635 {
636 	enum page_type btype = PAGE_TYPE_OF_BIO(type);
637 	struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
638 
639 	f2fs_down_write(&io->io_rwsem);
640 
641 	/* change META to META_FLUSH in the checkpoint procedure */
642 	if (type >= META_FLUSH) {
643 		io->fio.type = META_FLUSH;
644 		io->bio->bi_opf |= REQ_META | REQ_PRIO | REQ_SYNC;
645 		if (!test_opt(sbi, NOBARRIER))
646 			io->bio->bi_opf |= REQ_PREFLUSH | REQ_FUA;
647 	}
648 	__submit_merged_bio(io);
649 	f2fs_up_write(&io->io_rwsem);
650 }
651 
652 static void __submit_merged_write_cond(struct f2fs_sb_info *sbi,
653 				struct inode *inode, struct page *page,
654 				nid_t ino, enum page_type type, bool force)
655 {
656 	enum temp_type temp;
657 	bool ret = true;
658 
659 	for (temp = HOT; temp < NR_TEMP_TYPE; temp++) {
660 		if (!force)	{
661 			enum page_type btype = PAGE_TYPE_OF_BIO(type);
662 			struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
663 
664 			f2fs_down_read(&io->io_rwsem);
665 			ret = __has_merged_page(io->bio, inode, page, ino);
666 			f2fs_up_read(&io->io_rwsem);
667 		}
668 		if (ret)
669 			__f2fs_submit_merged_write(sbi, type, temp);
670 
671 		/* TODO: use HOT temp only for meta pages now. */
672 		if (type >= META)
673 			break;
674 	}
675 }
676 
677 void f2fs_submit_merged_write(struct f2fs_sb_info *sbi, enum page_type type)
678 {
679 	__submit_merged_write_cond(sbi, NULL, NULL, 0, type, true);
680 }
681 
682 void f2fs_submit_merged_write_cond(struct f2fs_sb_info *sbi,
683 				struct inode *inode, struct page *page,
684 				nid_t ino, enum page_type type)
685 {
686 	__submit_merged_write_cond(sbi, inode, page, ino, type, false);
687 }
688 
689 void f2fs_flush_merged_writes(struct f2fs_sb_info *sbi)
690 {
691 	f2fs_submit_merged_write(sbi, DATA);
692 	f2fs_submit_merged_write(sbi, NODE);
693 	f2fs_submit_merged_write(sbi, META);
694 }
695 
696 /*
697  * Fill the locked page with data located in the block address.
698  * A caller needs to unlock the page on failure.
699  */
700 int f2fs_submit_page_bio(struct f2fs_io_info *fio)
701 {
702 	struct bio *bio;
703 	struct page *page = fio->encrypted_page ?
704 			fio->encrypted_page : fio->page;
705 
706 	if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
707 			fio->is_por ? META_POR : (__is_meta_io(fio) ?
708 			META_GENERIC : DATA_GENERIC_ENHANCE)))
709 		return -EFSCORRUPTED;
710 
711 	trace_f2fs_submit_page_bio(page, fio);
712 
713 	/* Allocate a new bio */
714 	bio = __bio_alloc(fio, 1);
715 
716 	f2fs_set_bio_crypt_ctx(bio, fio->page->mapping->host,
717 			       fio->page->index, fio, GFP_NOIO);
718 
719 	if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
720 		bio_put(bio);
721 		return -EFAULT;
722 	}
723 
724 	if (fio->io_wbc && !is_read_io(fio->op))
725 		wbc_account_cgroup_owner(fio->io_wbc, page, PAGE_SIZE);
726 
727 	inc_page_count(fio->sbi, is_read_io(fio->op) ?
728 			__read_io_type(page): WB_DATA_TYPE(fio->page));
729 
730 	__submit_bio(fio->sbi, bio, fio->type);
731 	return 0;
732 }
733 
734 static bool page_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
735 				block_t last_blkaddr, block_t cur_blkaddr)
736 {
737 	if (unlikely(sbi->max_io_bytes &&
738 			bio->bi_iter.bi_size >= sbi->max_io_bytes))
739 		return false;
740 	if (last_blkaddr + 1 != cur_blkaddr)
741 		return false;
742 	return bio->bi_bdev == f2fs_target_device(sbi, cur_blkaddr, NULL);
743 }
744 
745 static bool io_type_is_mergeable(struct f2fs_bio_info *io,
746 						struct f2fs_io_info *fio)
747 {
748 	if (io->fio.op != fio->op)
749 		return false;
750 	return io->fio.op_flags == fio->op_flags;
751 }
752 
753 static bool io_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
754 					struct f2fs_bio_info *io,
755 					struct f2fs_io_info *fio,
756 					block_t last_blkaddr,
757 					block_t cur_blkaddr)
758 {
759 	if (F2FS_IO_ALIGNED(sbi) && (fio->type == DATA || fio->type == NODE)) {
760 		unsigned int filled_blocks =
761 				F2FS_BYTES_TO_BLK(bio->bi_iter.bi_size);
762 		unsigned int io_size = F2FS_IO_SIZE(sbi);
763 		unsigned int left_vecs = bio->bi_max_vecs - bio->bi_vcnt;
764 
765 		/* IOs in bio is aligned and left space of vectors is not enough */
766 		if (!(filled_blocks % io_size) && left_vecs < io_size)
767 			return false;
768 	}
769 	if (!page_is_mergeable(sbi, bio, last_blkaddr, cur_blkaddr))
770 		return false;
771 	return io_type_is_mergeable(io, fio);
772 }
773 
774 static void add_bio_entry(struct f2fs_sb_info *sbi, struct bio *bio,
775 				struct page *page, enum temp_type temp)
776 {
777 	struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
778 	struct bio_entry *be;
779 
780 	be = f2fs_kmem_cache_alloc(bio_entry_slab, GFP_NOFS, true, NULL);
781 	be->bio = bio;
782 	bio_get(bio);
783 
784 	if (bio_add_page(bio, page, PAGE_SIZE, 0) != PAGE_SIZE)
785 		f2fs_bug_on(sbi, 1);
786 
787 	f2fs_down_write(&io->bio_list_lock);
788 	list_add_tail(&be->list, &io->bio_list);
789 	f2fs_up_write(&io->bio_list_lock);
790 }
791 
792 static void del_bio_entry(struct bio_entry *be)
793 {
794 	list_del(&be->list);
795 	kmem_cache_free(bio_entry_slab, be);
796 }
797 
798 static int add_ipu_page(struct f2fs_io_info *fio, struct bio **bio,
799 							struct page *page)
800 {
801 	struct f2fs_sb_info *sbi = fio->sbi;
802 	enum temp_type temp;
803 	bool found = false;
804 	int ret = -EAGAIN;
805 
806 	for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
807 		struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
808 		struct list_head *head = &io->bio_list;
809 		struct bio_entry *be;
810 
811 		f2fs_down_write(&io->bio_list_lock);
812 		list_for_each_entry(be, head, list) {
813 			if (be->bio != *bio)
814 				continue;
815 
816 			found = true;
817 
818 			f2fs_bug_on(sbi, !page_is_mergeable(sbi, *bio,
819 							    *fio->last_block,
820 							    fio->new_blkaddr));
821 			if (f2fs_crypt_mergeable_bio(*bio,
822 					fio->page->mapping->host,
823 					fio->page->index, fio) &&
824 			    bio_add_page(*bio, page, PAGE_SIZE, 0) ==
825 					PAGE_SIZE) {
826 				ret = 0;
827 				break;
828 			}
829 
830 			/* page can't be merged into bio; submit the bio */
831 			del_bio_entry(be);
832 			__submit_bio(sbi, *bio, DATA);
833 			break;
834 		}
835 		f2fs_up_write(&io->bio_list_lock);
836 	}
837 
838 	if (ret) {
839 		bio_put(*bio);
840 		*bio = NULL;
841 	}
842 
843 	return ret;
844 }
845 
846 void f2fs_submit_merged_ipu_write(struct f2fs_sb_info *sbi,
847 					struct bio **bio, struct page *page)
848 {
849 	enum temp_type temp;
850 	bool found = false;
851 	struct bio *target = bio ? *bio : NULL;
852 
853 	for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
854 		struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
855 		struct list_head *head = &io->bio_list;
856 		struct bio_entry *be;
857 
858 		if (list_empty(head))
859 			continue;
860 
861 		f2fs_down_read(&io->bio_list_lock);
862 		list_for_each_entry(be, head, list) {
863 			if (target)
864 				found = (target == be->bio);
865 			else
866 				found = __has_merged_page(be->bio, NULL,
867 								page, 0);
868 			if (found)
869 				break;
870 		}
871 		f2fs_up_read(&io->bio_list_lock);
872 
873 		if (!found)
874 			continue;
875 
876 		found = false;
877 
878 		f2fs_down_write(&io->bio_list_lock);
879 		list_for_each_entry(be, head, list) {
880 			if (target)
881 				found = (target == be->bio);
882 			else
883 				found = __has_merged_page(be->bio, NULL,
884 								page, 0);
885 			if (found) {
886 				target = be->bio;
887 				del_bio_entry(be);
888 				break;
889 			}
890 		}
891 		f2fs_up_write(&io->bio_list_lock);
892 	}
893 
894 	if (found)
895 		__submit_bio(sbi, target, DATA);
896 	if (bio && *bio) {
897 		bio_put(*bio);
898 		*bio = NULL;
899 	}
900 }
901 
902 int f2fs_merge_page_bio(struct f2fs_io_info *fio)
903 {
904 	struct bio *bio = *fio->bio;
905 	struct page *page = fio->encrypted_page ?
906 			fio->encrypted_page : fio->page;
907 
908 	if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
909 			__is_meta_io(fio) ? META_GENERIC : DATA_GENERIC))
910 		return -EFSCORRUPTED;
911 
912 	trace_f2fs_submit_page_bio(page, fio);
913 
914 	if (bio && !page_is_mergeable(fio->sbi, bio, *fio->last_block,
915 						fio->new_blkaddr))
916 		f2fs_submit_merged_ipu_write(fio->sbi, &bio, NULL);
917 alloc_new:
918 	if (!bio) {
919 		bio = __bio_alloc(fio, BIO_MAX_VECS);
920 		f2fs_set_bio_crypt_ctx(bio, fio->page->mapping->host,
921 				       fio->page->index, fio, GFP_NOIO);
922 
923 		add_bio_entry(fio->sbi, bio, page, fio->temp);
924 	} else {
925 		if (add_ipu_page(fio, &bio, page))
926 			goto alloc_new;
927 	}
928 
929 	if (fio->io_wbc)
930 		wbc_account_cgroup_owner(fio->io_wbc, page, PAGE_SIZE);
931 
932 	inc_page_count(fio->sbi, WB_DATA_TYPE(page));
933 
934 	*fio->last_block = fio->new_blkaddr;
935 	*fio->bio = bio;
936 
937 	return 0;
938 }
939 
940 void f2fs_submit_page_write(struct f2fs_io_info *fio)
941 {
942 	struct f2fs_sb_info *sbi = fio->sbi;
943 	enum page_type btype = PAGE_TYPE_OF_BIO(fio->type);
944 	struct f2fs_bio_info *io = sbi->write_io[btype] + fio->temp;
945 	struct page *bio_page;
946 
947 	f2fs_bug_on(sbi, is_read_io(fio->op));
948 
949 	f2fs_down_write(&io->io_rwsem);
950 next:
951 	if (fio->in_list) {
952 		spin_lock(&io->io_lock);
953 		if (list_empty(&io->io_list)) {
954 			spin_unlock(&io->io_lock);
955 			goto out;
956 		}
957 		fio = list_first_entry(&io->io_list,
958 						struct f2fs_io_info, list);
959 		list_del(&fio->list);
960 		spin_unlock(&io->io_lock);
961 	}
962 
963 	verify_fio_blkaddr(fio);
964 
965 	if (fio->encrypted_page)
966 		bio_page = fio->encrypted_page;
967 	else if (fio->compressed_page)
968 		bio_page = fio->compressed_page;
969 	else
970 		bio_page = fio->page;
971 
972 	/* set submitted = true as a return value */
973 	fio->submitted = true;
974 
975 	inc_page_count(sbi, WB_DATA_TYPE(bio_page));
976 
977 	if (io->bio &&
978 	    (!io_is_mergeable(sbi, io->bio, io, fio, io->last_block_in_bio,
979 			      fio->new_blkaddr) ||
980 	     !f2fs_crypt_mergeable_bio(io->bio, fio->page->mapping->host,
981 				       bio_page->index, fio)))
982 		__submit_merged_bio(io);
983 alloc_new:
984 	if (io->bio == NULL) {
985 		if (F2FS_IO_ALIGNED(sbi) &&
986 				(fio->type == DATA || fio->type == NODE) &&
987 				fio->new_blkaddr & F2FS_IO_SIZE_MASK(sbi)) {
988 			dec_page_count(sbi, WB_DATA_TYPE(bio_page));
989 			fio->retry = true;
990 			goto skip;
991 		}
992 		io->bio = __bio_alloc(fio, BIO_MAX_VECS);
993 		f2fs_set_bio_crypt_ctx(io->bio, fio->page->mapping->host,
994 				       bio_page->index, fio, GFP_NOIO);
995 		io->fio = *fio;
996 	}
997 
998 	if (bio_add_page(io->bio, bio_page, PAGE_SIZE, 0) < PAGE_SIZE) {
999 		__submit_merged_bio(io);
1000 		goto alloc_new;
1001 	}
1002 
1003 	if (fio->io_wbc)
1004 		wbc_account_cgroup_owner(fio->io_wbc, bio_page, PAGE_SIZE);
1005 
1006 	io->last_block_in_bio = fio->new_blkaddr;
1007 
1008 	trace_f2fs_submit_page_write(fio->page, fio);
1009 skip:
1010 	if (fio->in_list)
1011 		goto next;
1012 out:
1013 	if (is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN) ||
1014 				!f2fs_is_checkpoint_ready(sbi))
1015 		__submit_merged_bio(io);
1016 	f2fs_up_write(&io->io_rwsem);
1017 }
1018 
1019 static struct bio *f2fs_grab_read_bio(struct inode *inode, block_t blkaddr,
1020 				      unsigned nr_pages, blk_opf_t op_flag,
1021 				      pgoff_t first_idx, bool for_write)
1022 {
1023 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1024 	struct bio *bio;
1025 	struct bio_post_read_ctx *ctx = NULL;
1026 	unsigned int post_read_steps = 0;
1027 	sector_t sector;
1028 	struct block_device *bdev = f2fs_target_device(sbi, blkaddr, &sector);
1029 
1030 	bio = bio_alloc_bioset(bdev, bio_max_segs(nr_pages),
1031 			       REQ_OP_READ | op_flag,
1032 			       for_write ? GFP_NOIO : GFP_KERNEL, &f2fs_bioset);
1033 	if (!bio)
1034 		return ERR_PTR(-ENOMEM);
1035 	bio->bi_iter.bi_sector = sector;
1036 	f2fs_set_bio_crypt_ctx(bio, inode, first_idx, NULL, GFP_NOFS);
1037 	bio->bi_end_io = f2fs_read_end_io;
1038 
1039 	if (fscrypt_inode_uses_fs_layer_crypto(inode))
1040 		post_read_steps |= STEP_DECRYPT;
1041 
1042 	if (f2fs_need_verity(inode, first_idx))
1043 		post_read_steps |= STEP_VERITY;
1044 
1045 	/*
1046 	 * STEP_DECOMPRESS is handled specially, since a compressed file might
1047 	 * contain both compressed and uncompressed clusters.  We'll allocate a
1048 	 * bio_post_read_ctx if the file is compressed, but the caller is
1049 	 * responsible for enabling STEP_DECOMPRESS if it's actually needed.
1050 	 */
1051 
1052 	if (post_read_steps || f2fs_compressed_file(inode)) {
1053 		/* Due to the mempool, this never fails. */
1054 		ctx = mempool_alloc(bio_post_read_ctx_pool, GFP_NOFS);
1055 		ctx->bio = bio;
1056 		ctx->sbi = sbi;
1057 		ctx->enabled_steps = post_read_steps;
1058 		ctx->fs_blkaddr = blkaddr;
1059 		bio->bi_private = ctx;
1060 	}
1061 	iostat_alloc_and_bind_ctx(sbi, bio, ctx);
1062 
1063 	return bio;
1064 }
1065 
1066 /* This can handle encryption stuffs */
1067 static int f2fs_submit_page_read(struct inode *inode, struct page *page,
1068 				 block_t blkaddr, blk_opf_t op_flags,
1069 				 bool for_write)
1070 {
1071 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1072 	struct bio *bio;
1073 
1074 	bio = f2fs_grab_read_bio(inode, blkaddr, 1, op_flags,
1075 					page->index, for_write);
1076 	if (IS_ERR(bio))
1077 		return PTR_ERR(bio);
1078 
1079 	/* wait for GCed page writeback via META_MAPPING */
1080 	f2fs_wait_on_block_writeback(inode, blkaddr);
1081 
1082 	if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
1083 		bio_put(bio);
1084 		return -EFAULT;
1085 	}
1086 	ClearPageError(page);
1087 	inc_page_count(sbi, F2FS_RD_DATA);
1088 	f2fs_update_iostat(sbi, FS_DATA_READ_IO, F2FS_BLKSIZE);
1089 	__submit_bio(sbi, bio, DATA);
1090 	return 0;
1091 }
1092 
1093 static void __set_data_blkaddr(struct dnode_of_data *dn)
1094 {
1095 	struct f2fs_node *rn = F2FS_NODE(dn->node_page);
1096 	__le32 *addr_array;
1097 	int base = 0;
1098 
1099 	if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
1100 		base = get_extra_isize(dn->inode);
1101 
1102 	/* Get physical address of data block */
1103 	addr_array = blkaddr_in_node(rn);
1104 	addr_array[base + dn->ofs_in_node] = cpu_to_le32(dn->data_blkaddr);
1105 }
1106 
1107 /*
1108  * Lock ordering for the change of data block address:
1109  * ->data_page
1110  *  ->node_page
1111  *    update block addresses in the node page
1112  */
1113 void f2fs_set_data_blkaddr(struct dnode_of_data *dn)
1114 {
1115 	f2fs_wait_on_page_writeback(dn->node_page, NODE, true, true);
1116 	__set_data_blkaddr(dn);
1117 	if (set_page_dirty(dn->node_page))
1118 		dn->node_changed = true;
1119 }
1120 
1121 void f2fs_update_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr)
1122 {
1123 	dn->data_blkaddr = blkaddr;
1124 	f2fs_set_data_blkaddr(dn);
1125 	f2fs_update_extent_cache(dn);
1126 }
1127 
1128 /* dn->ofs_in_node will be returned with up-to-date last block pointer */
1129 int f2fs_reserve_new_blocks(struct dnode_of_data *dn, blkcnt_t count)
1130 {
1131 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1132 	int err;
1133 
1134 	if (!count)
1135 		return 0;
1136 
1137 	if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
1138 		return -EPERM;
1139 	if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
1140 		return err;
1141 
1142 	trace_f2fs_reserve_new_blocks(dn->inode, dn->nid,
1143 						dn->ofs_in_node, count);
1144 
1145 	f2fs_wait_on_page_writeback(dn->node_page, NODE, true, true);
1146 
1147 	for (; count > 0; dn->ofs_in_node++) {
1148 		block_t blkaddr = f2fs_data_blkaddr(dn);
1149 
1150 		if (blkaddr == NULL_ADDR) {
1151 			dn->data_blkaddr = NEW_ADDR;
1152 			__set_data_blkaddr(dn);
1153 			count--;
1154 		}
1155 	}
1156 
1157 	if (set_page_dirty(dn->node_page))
1158 		dn->node_changed = true;
1159 	return 0;
1160 }
1161 
1162 /* Should keep dn->ofs_in_node unchanged */
1163 int f2fs_reserve_new_block(struct dnode_of_data *dn)
1164 {
1165 	unsigned int ofs_in_node = dn->ofs_in_node;
1166 	int ret;
1167 
1168 	ret = f2fs_reserve_new_blocks(dn, 1);
1169 	dn->ofs_in_node = ofs_in_node;
1170 	return ret;
1171 }
1172 
1173 int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
1174 {
1175 	bool need_put = dn->inode_page ? false : true;
1176 	int err;
1177 
1178 	err = f2fs_get_dnode_of_data(dn, index, ALLOC_NODE);
1179 	if (err)
1180 		return err;
1181 
1182 	if (dn->data_blkaddr == NULL_ADDR)
1183 		err = f2fs_reserve_new_block(dn);
1184 	if (err || need_put)
1185 		f2fs_put_dnode(dn);
1186 	return err;
1187 }
1188 
1189 int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index)
1190 {
1191 	struct extent_info ei = {0, };
1192 	struct inode *inode = dn->inode;
1193 
1194 	if (f2fs_lookup_extent_cache(inode, index, &ei)) {
1195 		dn->data_blkaddr = ei.blk + index - ei.fofs;
1196 		return 0;
1197 	}
1198 
1199 	return f2fs_reserve_block(dn, index);
1200 }
1201 
1202 struct page *f2fs_get_read_data_page(struct inode *inode, pgoff_t index,
1203 				     blk_opf_t op_flags, bool for_write)
1204 {
1205 	struct address_space *mapping = inode->i_mapping;
1206 	struct dnode_of_data dn;
1207 	struct page *page;
1208 	struct extent_info ei = {0, };
1209 	int err;
1210 
1211 	page = f2fs_grab_cache_page(mapping, index, for_write);
1212 	if (!page)
1213 		return ERR_PTR(-ENOMEM);
1214 
1215 	if (f2fs_lookup_extent_cache(inode, index, &ei)) {
1216 		dn.data_blkaddr = ei.blk + index - ei.fofs;
1217 		if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), dn.data_blkaddr,
1218 						DATA_GENERIC_ENHANCE_READ)) {
1219 			err = -EFSCORRUPTED;
1220 			goto put_err;
1221 		}
1222 		goto got_it;
1223 	}
1224 
1225 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1226 	err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
1227 	if (err)
1228 		goto put_err;
1229 	f2fs_put_dnode(&dn);
1230 
1231 	if (unlikely(dn.data_blkaddr == NULL_ADDR)) {
1232 		err = -ENOENT;
1233 		goto put_err;
1234 	}
1235 	if (dn.data_blkaddr != NEW_ADDR &&
1236 			!f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
1237 						dn.data_blkaddr,
1238 						DATA_GENERIC_ENHANCE)) {
1239 		err = -EFSCORRUPTED;
1240 		goto put_err;
1241 	}
1242 got_it:
1243 	if (PageUptodate(page)) {
1244 		unlock_page(page);
1245 		return page;
1246 	}
1247 
1248 	/*
1249 	 * A new dentry page is allocated but not able to be written, since its
1250 	 * new inode page couldn't be allocated due to -ENOSPC.
1251 	 * In such the case, its blkaddr can be remained as NEW_ADDR.
1252 	 * see, f2fs_add_link -> f2fs_get_new_data_page ->
1253 	 * f2fs_init_inode_metadata.
1254 	 */
1255 	if (dn.data_blkaddr == NEW_ADDR) {
1256 		zero_user_segment(page, 0, PAGE_SIZE);
1257 		if (!PageUptodate(page))
1258 			SetPageUptodate(page);
1259 		unlock_page(page);
1260 		return page;
1261 	}
1262 
1263 	err = f2fs_submit_page_read(inode, page, dn.data_blkaddr,
1264 						op_flags, for_write);
1265 	if (err)
1266 		goto put_err;
1267 	return page;
1268 
1269 put_err:
1270 	f2fs_put_page(page, 1);
1271 	return ERR_PTR(err);
1272 }
1273 
1274 struct page *f2fs_find_data_page(struct inode *inode, pgoff_t index)
1275 {
1276 	struct address_space *mapping = inode->i_mapping;
1277 	struct page *page;
1278 
1279 	page = find_get_page(mapping, index);
1280 	if (page && PageUptodate(page))
1281 		return page;
1282 	f2fs_put_page(page, 0);
1283 
1284 	page = f2fs_get_read_data_page(inode, index, 0, false);
1285 	if (IS_ERR(page))
1286 		return page;
1287 
1288 	if (PageUptodate(page))
1289 		return page;
1290 
1291 	wait_on_page_locked(page);
1292 	if (unlikely(!PageUptodate(page))) {
1293 		f2fs_put_page(page, 0);
1294 		return ERR_PTR(-EIO);
1295 	}
1296 	return page;
1297 }
1298 
1299 /*
1300  * If it tries to access a hole, return an error.
1301  * Because, the callers, functions in dir.c and GC, should be able to know
1302  * whether this page exists or not.
1303  */
1304 struct page *f2fs_get_lock_data_page(struct inode *inode, pgoff_t index,
1305 							bool for_write)
1306 {
1307 	struct address_space *mapping = inode->i_mapping;
1308 	struct page *page;
1309 repeat:
1310 	page = f2fs_get_read_data_page(inode, index, 0, for_write);
1311 	if (IS_ERR(page))
1312 		return page;
1313 
1314 	/* wait for read completion */
1315 	lock_page(page);
1316 	if (unlikely(page->mapping != mapping)) {
1317 		f2fs_put_page(page, 1);
1318 		goto repeat;
1319 	}
1320 	if (unlikely(!PageUptodate(page))) {
1321 		f2fs_put_page(page, 1);
1322 		return ERR_PTR(-EIO);
1323 	}
1324 	return page;
1325 }
1326 
1327 /*
1328  * Caller ensures that this data page is never allocated.
1329  * A new zero-filled data page is allocated in the page cache.
1330  *
1331  * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and
1332  * f2fs_unlock_op().
1333  * Note that, ipage is set only by make_empty_dir, and if any error occur,
1334  * ipage should be released by this function.
1335  */
1336 struct page *f2fs_get_new_data_page(struct inode *inode,
1337 		struct page *ipage, pgoff_t index, bool new_i_size)
1338 {
1339 	struct address_space *mapping = inode->i_mapping;
1340 	struct page *page;
1341 	struct dnode_of_data dn;
1342 	int err;
1343 
1344 	page = f2fs_grab_cache_page(mapping, index, true);
1345 	if (!page) {
1346 		/*
1347 		 * before exiting, we should make sure ipage will be released
1348 		 * if any error occur.
1349 		 */
1350 		f2fs_put_page(ipage, 1);
1351 		return ERR_PTR(-ENOMEM);
1352 	}
1353 
1354 	set_new_dnode(&dn, inode, ipage, NULL, 0);
1355 	err = f2fs_reserve_block(&dn, index);
1356 	if (err) {
1357 		f2fs_put_page(page, 1);
1358 		return ERR_PTR(err);
1359 	}
1360 	if (!ipage)
1361 		f2fs_put_dnode(&dn);
1362 
1363 	if (PageUptodate(page))
1364 		goto got_it;
1365 
1366 	if (dn.data_blkaddr == NEW_ADDR) {
1367 		zero_user_segment(page, 0, PAGE_SIZE);
1368 		if (!PageUptodate(page))
1369 			SetPageUptodate(page);
1370 	} else {
1371 		f2fs_put_page(page, 1);
1372 
1373 		/* if ipage exists, blkaddr should be NEW_ADDR */
1374 		f2fs_bug_on(F2FS_I_SB(inode), ipage);
1375 		page = f2fs_get_lock_data_page(inode, index, true);
1376 		if (IS_ERR(page))
1377 			return page;
1378 	}
1379 got_it:
1380 	if (new_i_size && i_size_read(inode) <
1381 				((loff_t)(index + 1) << PAGE_SHIFT))
1382 		f2fs_i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT));
1383 	return page;
1384 }
1385 
1386 static int __allocate_data_block(struct dnode_of_data *dn, int seg_type)
1387 {
1388 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1389 	struct f2fs_summary sum;
1390 	struct node_info ni;
1391 	block_t old_blkaddr;
1392 	blkcnt_t count = 1;
1393 	int err;
1394 
1395 	if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
1396 		return -EPERM;
1397 
1398 	err = f2fs_get_node_info(sbi, dn->nid, &ni, false);
1399 	if (err)
1400 		return err;
1401 
1402 	dn->data_blkaddr = f2fs_data_blkaddr(dn);
1403 	if (dn->data_blkaddr != NULL_ADDR)
1404 		goto alloc;
1405 
1406 	if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
1407 		return err;
1408 
1409 alloc:
1410 	set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
1411 	old_blkaddr = dn->data_blkaddr;
1412 	f2fs_allocate_data_block(sbi, NULL, old_blkaddr, &dn->data_blkaddr,
1413 				&sum, seg_type, NULL);
1414 	if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO) {
1415 		invalidate_mapping_pages(META_MAPPING(sbi),
1416 					old_blkaddr, old_blkaddr);
1417 		f2fs_invalidate_compress_page(sbi, old_blkaddr);
1418 	}
1419 	f2fs_update_data_blkaddr(dn, dn->data_blkaddr);
1420 	return 0;
1421 }
1422 
1423 void f2fs_do_map_lock(struct f2fs_sb_info *sbi, int flag, bool lock)
1424 {
1425 	if (flag == F2FS_GET_BLOCK_PRE_AIO) {
1426 		if (lock)
1427 			f2fs_down_read(&sbi->node_change);
1428 		else
1429 			f2fs_up_read(&sbi->node_change);
1430 	} else {
1431 		if (lock)
1432 			f2fs_lock_op(sbi);
1433 		else
1434 			f2fs_unlock_op(sbi);
1435 	}
1436 }
1437 
1438 /*
1439  * f2fs_map_blocks() tries to find or build mapping relationship which
1440  * maps continuous logical blocks to physical blocks, and return such
1441  * info via f2fs_map_blocks structure.
1442  */
1443 int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
1444 						int create, int flag)
1445 {
1446 	unsigned int maxblocks = map->m_len;
1447 	struct dnode_of_data dn;
1448 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1449 	int mode = map->m_may_create ? ALLOC_NODE : LOOKUP_NODE;
1450 	pgoff_t pgofs, end_offset, end;
1451 	int err = 0, ofs = 1;
1452 	unsigned int ofs_in_node, last_ofs_in_node;
1453 	blkcnt_t prealloc;
1454 	struct extent_info ei = {0, };
1455 	block_t blkaddr;
1456 	unsigned int start_pgofs;
1457 	int bidx = 0;
1458 
1459 	if (!maxblocks)
1460 		return 0;
1461 
1462 	map->m_bdev = inode->i_sb->s_bdev;
1463 	map->m_multidev_dio =
1464 		f2fs_allow_multi_device_dio(F2FS_I_SB(inode), flag);
1465 
1466 	map->m_len = 0;
1467 	map->m_flags = 0;
1468 
1469 	/* it only supports block size == page size */
1470 	pgofs =	(pgoff_t)map->m_lblk;
1471 	end = pgofs + maxblocks;
1472 
1473 	if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) {
1474 		if (f2fs_lfs_mode(sbi) && flag == F2FS_GET_BLOCK_DIO &&
1475 							map->m_may_create)
1476 			goto next_dnode;
1477 
1478 		map->m_pblk = ei.blk + pgofs - ei.fofs;
1479 		map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs);
1480 		map->m_flags = F2FS_MAP_MAPPED;
1481 		if (map->m_next_extent)
1482 			*map->m_next_extent = pgofs + map->m_len;
1483 
1484 		/* for hardware encryption, but to avoid potential issue in future */
1485 		if (flag == F2FS_GET_BLOCK_DIO)
1486 			f2fs_wait_on_block_writeback_range(inode,
1487 						map->m_pblk, map->m_len);
1488 
1489 		if (map->m_multidev_dio) {
1490 			block_t blk_addr = map->m_pblk;
1491 
1492 			bidx = f2fs_target_device_index(sbi, map->m_pblk);
1493 
1494 			map->m_bdev = FDEV(bidx).bdev;
1495 			map->m_pblk -= FDEV(bidx).start_blk;
1496 			map->m_len = min(map->m_len,
1497 				FDEV(bidx).end_blk + 1 - map->m_pblk);
1498 
1499 			if (map->m_may_create)
1500 				f2fs_update_device_state(sbi, inode->i_ino,
1501 							blk_addr, map->m_len);
1502 		}
1503 		goto out;
1504 	}
1505 
1506 next_dnode:
1507 	if (map->m_may_create)
1508 		f2fs_do_map_lock(sbi, flag, true);
1509 
1510 	/* When reading holes, we need its node page */
1511 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1512 	err = f2fs_get_dnode_of_data(&dn, pgofs, mode);
1513 	if (err) {
1514 		if (flag == F2FS_GET_BLOCK_BMAP)
1515 			map->m_pblk = 0;
1516 
1517 		if (err == -ENOENT) {
1518 			/*
1519 			 * There is one exceptional case that read_node_page()
1520 			 * may return -ENOENT due to filesystem has been
1521 			 * shutdown or cp_error, so force to convert error
1522 			 * number to EIO for such case.
1523 			 */
1524 			if (map->m_may_create &&
1525 				(is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN) ||
1526 				f2fs_cp_error(sbi))) {
1527 				err = -EIO;
1528 				goto unlock_out;
1529 			}
1530 
1531 			err = 0;
1532 			if (map->m_next_pgofs)
1533 				*map->m_next_pgofs =
1534 					f2fs_get_next_page_offset(&dn, pgofs);
1535 			if (map->m_next_extent)
1536 				*map->m_next_extent =
1537 					f2fs_get_next_page_offset(&dn, pgofs);
1538 		}
1539 		goto unlock_out;
1540 	}
1541 
1542 	start_pgofs = pgofs;
1543 	prealloc = 0;
1544 	last_ofs_in_node = ofs_in_node = dn.ofs_in_node;
1545 	end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1546 
1547 next_block:
1548 	blkaddr = f2fs_data_blkaddr(&dn);
1549 
1550 	if (__is_valid_data_blkaddr(blkaddr) &&
1551 		!f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC_ENHANCE)) {
1552 		err = -EFSCORRUPTED;
1553 		goto sync_out;
1554 	}
1555 
1556 	if (__is_valid_data_blkaddr(blkaddr)) {
1557 		/* use out-place-update for driect IO under LFS mode */
1558 		if (f2fs_lfs_mode(sbi) && flag == F2FS_GET_BLOCK_DIO &&
1559 							map->m_may_create) {
1560 			err = __allocate_data_block(&dn, map->m_seg_type);
1561 			if (err)
1562 				goto sync_out;
1563 			blkaddr = dn.data_blkaddr;
1564 			set_inode_flag(inode, FI_APPEND_WRITE);
1565 		}
1566 	} else {
1567 		if (create) {
1568 			if (unlikely(f2fs_cp_error(sbi))) {
1569 				err = -EIO;
1570 				goto sync_out;
1571 			}
1572 			if (flag == F2FS_GET_BLOCK_PRE_AIO) {
1573 				if (blkaddr == NULL_ADDR) {
1574 					prealloc++;
1575 					last_ofs_in_node = dn.ofs_in_node;
1576 				}
1577 			} else {
1578 				WARN_ON(flag != F2FS_GET_BLOCK_PRE_DIO &&
1579 					flag != F2FS_GET_BLOCK_DIO);
1580 				err = __allocate_data_block(&dn,
1581 							map->m_seg_type);
1582 				if (!err) {
1583 					if (flag == F2FS_GET_BLOCK_PRE_DIO)
1584 						file_need_truncate(inode);
1585 					set_inode_flag(inode, FI_APPEND_WRITE);
1586 				}
1587 			}
1588 			if (err)
1589 				goto sync_out;
1590 			map->m_flags |= F2FS_MAP_NEW;
1591 			blkaddr = dn.data_blkaddr;
1592 		} else {
1593 			if (f2fs_compressed_file(inode) &&
1594 					f2fs_sanity_check_cluster(&dn) &&
1595 					(flag != F2FS_GET_BLOCK_FIEMAP ||
1596 					IS_ENABLED(CONFIG_F2FS_CHECK_FS))) {
1597 				err = -EFSCORRUPTED;
1598 				goto sync_out;
1599 			}
1600 			if (flag == F2FS_GET_BLOCK_BMAP) {
1601 				map->m_pblk = 0;
1602 				goto sync_out;
1603 			}
1604 			if (flag == F2FS_GET_BLOCK_PRECACHE)
1605 				goto sync_out;
1606 			if (flag == F2FS_GET_BLOCK_FIEMAP &&
1607 						blkaddr == NULL_ADDR) {
1608 				if (map->m_next_pgofs)
1609 					*map->m_next_pgofs = pgofs + 1;
1610 				goto sync_out;
1611 			}
1612 			if (flag != F2FS_GET_BLOCK_FIEMAP) {
1613 				/* for defragment case */
1614 				if (map->m_next_pgofs)
1615 					*map->m_next_pgofs = pgofs + 1;
1616 				goto sync_out;
1617 			}
1618 		}
1619 	}
1620 
1621 	if (flag == F2FS_GET_BLOCK_PRE_AIO)
1622 		goto skip;
1623 
1624 	if (map->m_multidev_dio)
1625 		bidx = f2fs_target_device_index(sbi, blkaddr);
1626 
1627 	if (map->m_len == 0) {
1628 		/* preallocated unwritten block should be mapped for fiemap. */
1629 		if (blkaddr == NEW_ADDR)
1630 			map->m_flags |= F2FS_MAP_UNWRITTEN;
1631 		map->m_flags |= F2FS_MAP_MAPPED;
1632 
1633 		map->m_pblk = blkaddr;
1634 		map->m_len = 1;
1635 
1636 		if (map->m_multidev_dio)
1637 			map->m_bdev = FDEV(bidx).bdev;
1638 	} else if ((map->m_pblk != NEW_ADDR &&
1639 			blkaddr == (map->m_pblk + ofs)) ||
1640 			(map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR) ||
1641 			flag == F2FS_GET_BLOCK_PRE_DIO) {
1642 		if (map->m_multidev_dio && map->m_bdev != FDEV(bidx).bdev)
1643 			goto sync_out;
1644 		ofs++;
1645 		map->m_len++;
1646 	} else {
1647 		goto sync_out;
1648 	}
1649 
1650 skip:
1651 	dn.ofs_in_node++;
1652 	pgofs++;
1653 
1654 	/* preallocate blocks in batch for one dnode page */
1655 	if (flag == F2FS_GET_BLOCK_PRE_AIO &&
1656 			(pgofs == end || dn.ofs_in_node == end_offset)) {
1657 
1658 		dn.ofs_in_node = ofs_in_node;
1659 		err = f2fs_reserve_new_blocks(&dn, prealloc);
1660 		if (err)
1661 			goto sync_out;
1662 
1663 		map->m_len += dn.ofs_in_node - ofs_in_node;
1664 		if (prealloc && dn.ofs_in_node != last_ofs_in_node + 1) {
1665 			err = -ENOSPC;
1666 			goto sync_out;
1667 		}
1668 		dn.ofs_in_node = end_offset;
1669 	}
1670 
1671 	if (pgofs >= end)
1672 		goto sync_out;
1673 	else if (dn.ofs_in_node < end_offset)
1674 		goto next_block;
1675 
1676 	if (flag == F2FS_GET_BLOCK_PRECACHE) {
1677 		if (map->m_flags & F2FS_MAP_MAPPED) {
1678 			unsigned int ofs = start_pgofs - map->m_lblk;
1679 
1680 			f2fs_update_extent_cache_range(&dn,
1681 				start_pgofs, map->m_pblk + ofs,
1682 				map->m_len - ofs);
1683 		}
1684 	}
1685 
1686 	f2fs_put_dnode(&dn);
1687 
1688 	if (map->m_may_create) {
1689 		f2fs_do_map_lock(sbi, flag, false);
1690 		f2fs_balance_fs(sbi, dn.node_changed);
1691 	}
1692 	goto next_dnode;
1693 
1694 sync_out:
1695 
1696 	if (flag == F2FS_GET_BLOCK_DIO && map->m_flags & F2FS_MAP_MAPPED) {
1697 		/*
1698 		 * for hardware encryption, but to avoid potential issue
1699 		 * in future
1700 		 */
1701 		f2fs_wait_on_block_writeback_range(inode,
1702 						map->m_pblk, map->m_len);
1703 
1704 		if (map->m_multidev_dio) {
1705 			block_t blk_addr = map->m_pblk;
1706 
1707 			bidx = f2fs_target_device_index(sbi, map->m_pblk);
1708 
1709 			map->m_bdev = FDEV(bidx).bdev;
1710 			map->m_pblk -= FDEV(bidx).start_blk;
1711 
1712 			if (map->m_may_create)
1713 				f2fs_update_device_state(sbi, inode->i_ino,
1714 							blk_addr, map->m_len);
1715 
1716 			f2fs_bug_on(sbi, blk_addr + map->m_len >
1717 						FDEV(bidx).end_blk + 1);
1718 		}
1719 	}
1720 
1721 	if (flag == F2FS_GET_BLOCK_PRECACHE) {
1722 		if (map->m_flags & F2FS_MAP_MAPPED) {
1723 			unsigned int ofs = start_pgofs - map->m_lblk;
1724 
1725 			f2fs_update_extent_cache_range(&dn,
1726 				start_pgofs, map->m_pblk + ofs,
1727 				map->m_len - ofs);
1728 		}
1729 		if (map->m_next_extent)
1730 			*map->m_next_extent = pgofs + 1;
1731 	}
1732 	f2fs_put_dnode(&dn);
1733 unlock_out:
1734 	if (map->m_may_create) {
1735 		f2fs_do_map_lock(sbi, flag, false);
1736 		f2fs_balance_fs(sbi, dn.node_changed);
1737 	}
1738 out:
1739 	trace_f2fs_map_blocks(inode, map, create, flag, err);
1740 	return err;
1741 }
1742 
1743 bool f2fs_overwrite_io(struct inode *inode, loff_t pos, size_t len)
1744 {
1745 	struct f2fs_map_blocks map;
1746 	block_t last_lblk;
1747 	int err;
1748 
1749 	if (pos + len > i_size_read(inode))
1750 		return false;
1751 
1752 	map.m_lblk = F2FS_BYTES_TO_BLK(pos);
1753 	map.m_next_pgofs = NULL;
1754 	map.m_next_extent = NULL;
1755 	map.m_seg_type = NO_CHECK_TYPE;
1756 	map.m_may_create = false;
1757 	last_lblk = F2FS_BLK_ALIGN(pos + len);
1758 
1759 	while (map.m_lblk < last_lblk) {
1760 		map.m_len = last_lblk - map.m_lblk;
1761 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
1762 		if (err || map.m_len == 0)
1763 			return false;
1764 		map.m_lblk += map.m_len;
1765 	}
1766 	return true;
1767 }
1768 
1769 static inline u64 bytes_to_blks(struct inode *inode, u64 bytes)
1770 {
1771 	return (bytes >> inode->i_blkbits);
1772 }
1773 
1774 static inline u64 blks_to_bytes(struct inode *inode, u64 blks)
1775 {
1776 	return (blks << inode->i_blkbits);
1777 }
1778 
1779 static int f2fs_xattr_fiemap(struct inode *inode,
1780 				struct fiemap_extent_info *fieinfo)
1781 {
1782 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1783 	struct page *page;
1784 	struct node_info ni;
1785 	__u64 phys = 0, len;
1786 	__u32 flags;
1787 	nid_t xnid = F2FS_I(inode)->i_xattr_nid;
1788 	int err = 0;
1789 
1790 	if (f2fs_has_inline_xattr(inode)) {
1791 		int offset;
1792 
1793 		page = f2fs_grab_cache_page(NODE_MAPPING(sbi),
1794 						inode->i_ino, false);
1795 		if (!page)
1796 			return -ENOMEM;
1797 
1798 		err = f2fs_get_node_info(sbi, inode->i_ino, &ni, false);
1799 		if (err) {
1800 			f2fs_put_page(page, 1);
1801 			return err;
1802 		}
1803 
1804 		phys = blks_to_bytes(inode, ni.blk_addr);
1805 		offset = offsetof(struct f2fs_inode, i_addr) +
1806 					sizeof(__le32) * (DEF_ADDRS_PER_INODE -
1807 					get_inline_xattr_addrs(inode));
1808 
1809 		phys += offset;
1810 		len = inline_xattr_size(inode);
1811 
1812 		f2fs_put_page(page, 1);
1813 
1814 		flags = FIEMAP_EXTENT_DATA_INLINE | FIEMAP_EXTENT_NOT_ALIGNED;
1815 
1816 		if (!xnid)
1817 			flags |= FIEMAP_EXTENT_LAST;
1818 
1819 		err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
1820 		trace_f2fs_fiemap(inode, 0, phys, len, flags, err);
1821 		if (err || err == 1)
1822 			return err;
1823 	}
1824 
1825 	if (xnid) {
1826 		page = f2fs_grab_cache_page(NODE_MAPPING(sbi), xnid, false);
1827 		if (!page)
1828 			return -ENOMEM;
1829 
1830 		err = f2fs_get_node_info(sbi, xnid, &ni, false);
1831 		if (err) {
1832 			f2fs_put_page(page, 1);
1833 			return err;
1834 		}
1835 
1836 		phys = blks_to_bytes(inode, ni.blk_addr);
1837 		len = inode->i_sb->s_blocksize;
1838 
1839 		f2fs_put_page(page, 1);
1840 
1841 		flags = FIEMAP_EXTENT_LAST;
1842 	}
1843 
1844 	if (phys) {
1845 		err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
1846 		trace_f2fs_fiemap(inode, 0, phys, len, flags, err);
1847 	}
1848 
1849 	return (err < 0 ? err : 0);
1850 }
1851 
1852 static loff_t max_inode_blocks(struct inode *inode)
1853 {
1854 	loff_t result = ADDRS_PER_INODE(inode);
1855 	loff_t leaf_count = ADDRS_PER_BLOCK(inode);
1856 
1857 	/* two direct node blocks */
1858 	result += (leaf_count * 2);
1859 
1860 	/* two indirect node blocks */
1861 	leaf_count *= NIDS_PER_BLOCK;
1862 	result += (leaf_count * 2);
1863 
1864 	/* one double indirect node block */
1865 	leaf_count *= NIDS_PER_BLOCK;
1866 	result += leaf_count;
1867 
1868 	return result;
1869 }
1870 
1871 int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1872 		u64 start, u64 len)
1873 {
1874 	struct f2fs_map_blocks map;
1875 	sector_t start_blk, last_blk;
1876 	pgoff_t next_pgofs;
1877 	u64 logical = 0, phys = 0, size = 0;
1878 	u32 flags = 0;
1879 	int ret = 0;
1880 	bool compr_cluster = false, compr_appended;
1881 	unsigned int cluster_size = F2FS_I(inode)->i_cluster_size;
1882 	unsigned int count_in_cluster = 0;
1883 	loff_t maxbytes;
1884 
1885 	if (fieinfo->fi_flags & FIEMAP_FLAG_CACHE) {
1886 		ret = f2fs_precache_extents(inode);
1887 		if (ret)
1888 			return ret;
1889 	}
1890 
1891 	ret = fiemap_prep(inode, fieinfo, start, &len, FIEMAP_FLAG_XATTR);
1892 	if (ret)
1893 		return ret;
1894 
1895 	inode_lock(inode);
1896 
1897 	maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
1898 	if (start > maxbytes) {
1899 		ret = -EFBIG;
1900 		goto out;
1901 	}
1902 
1903 	if (len > maxbytes || (maxbytes - len) < start)
1904 		len = maxbytes - start;
1905 
1906 	if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
1907 		ret = f2fs_xattr_fiemap(inode, fieinfo);
1908 		goto out;
1909 	}
1910 
1911 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
1912 		ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len);
1913 		if (ret != -EAGAIN)
1914 			goto out;
1915 	}
1916 
1917 	if (bytes_to_blks(inode, len) == 0)
1918 		len = blks_to_bytes(inode, 1);
1919 
1920 	start_blk = bytes_to_blks(inode, start);
1921 	last_blk = bytes_to_blks(inode, start + len - 1);
1922 
1923 next:
1924 	memset(&map, 0, sizeof(map));
1925 	map.m_lblk = start_blk;
1926 	map.m_len = bytes_to_blks(inode, len);
1927 	map.m_next_pgofs = &next_pgofs;
1928 	map.m_seg_type = NO_CHECK_TYPE;
1929 
1930 	if (compr_cluster) {
1931 		map.m_lblk += 1;
1932 		map.m_len = cluster_size - count_in_cluster;
1933 	}
1934 
1935 	ret = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_FIEMAP);
1936 	if (ret)
1937 		goto out;
1938 
1939 	/* HOLE */
1940 	if (!compr_cluster && !(map.m_flags & F2FS_MAP_FLAGS)) {
1941 		start_blk = next_pgofs;
1942 
1943 		if (blks_to_bytes(inode, start_blk) < blks_to_bytes(inode,
1944 						max_inode_blocks(inode)))
1945 			goto prep_next;
1946 
1947 		flags |= FIEMAP_EXTENT_LAST;
1948 	}
1949 
1950 	compr_appended = false;
1951 	/* In a case of compressed cluster, append this to the last extent */
1952 	if (compr_cluster && ((map.m_flags & F2FS_MAP_UNWRITTEN) ||
1953 			!(map.m_flags & F2FS_MAP_FLAGS))) {
1954 		compr_appended = true;
1955 		goto skip_fill;
1956 	}
1957 
1958 	if (size) {
1959 		flags |= FIEMAP_EXTENT_MERGED;
1960 		if (IS_ENCRYPTED(inode))
1961 			flags |= FIEMAP_EXTENT_DATA_ENCRYPTED;
1962 
1963 		ret = fiemap_fill_next_extent(fieinfo, logical,
1964 				phys, size, flags);
1965 		trace_f2fs_fiemap(inode, logical, phys, size, flags, ret);
1966 		if (ret)
1967 			goto out;
1968 		size = 0;
1969 	}
1970 
1971 	if (start_blk > last_blk)
1972 		goto out;
1973 
1974 skip_fill:
1975 	if (map.m_pblk == COMPRESS_ADDR) {
1976 		compr_cluster = true;
1977 		count_in_cluster = 1;
1978 	} else if (compr_appended) {
1979 		unsigned int appended_blks = cluster_size -
1980 						count_in_cluster + 1;
1981 		size += blks_to_bytes(inode, appended_blks);
1982 		start_blk += appended_blks;
1983 		compr_cluster = false;
1984 	} else {
1985 		logical = blks_to_bytes(inode, start_blk);
1986 		phys = __is_valid_data_blkaddr(map.m_pblk) ?
1987 			blks_to_bytes(inode, map.m_pblk) : 0;
1988 		size = blks_to_bytes(inode, map.m_len);
1989 		flags = 0;
1990 
1991 		if (compr_cluster) {
1992 			flags = FIEMAP_EXTENT_ENCODED;
1993 			count_in_cluster += map.m_len;
1994 			if (count_in_cluster == cluster_size) {
1995 				compr_cluster = false;
1996 				size += blks_to_bytes(inode, 1);
1997 			}
1998 		} else if (map.m_flags & F2FS_MAP_UNWRITTEN) {
1999 			flags = FIEMAP_EXTENT_UNWRITTEN;
2000 		}
2001 
2002 		start_blk += bytes_to_blks(inode, size);
2003 	}
2004 
2005 prep_next:
2006 	cond_resched();
2007 	if (fatal_signal_pending(current))
2008 		ret = -EINTR;
2009 	else
2010 		goto next;
2011 out:
2012 	if (ret == 1)
2013 		ret = 0;
2014 
2015 	inode_unlock(inode);
2016 	return ret;
2017 }
2018 
2019 static inline loff_t f2fs_readpage_limit(struct inode *inode)
2020 {
2021 	if (IS_ENABLED(CONFIG_FS_VERITY) &&
2022 	    (IS_VERITY(inode) || f2fs_verity_in_progress(inode)))
2023 		return inode->i_sb->s_maxbytes;
2024 
2025 	return i_size_read(inode);
2026 }
2027 
2028 static int f2fs_read_single_page(struct inode *inode, struct page *page,
2029 					unsigned nr_pages,
2030 					struct f2fs_map_blocks *map,
2031 					struct bio **bio_ret,
2032 					sector_t *last_block_in_bio,
2033 					bool is_readahead)
2034 {
2035 	struct bio *bio = *bio_ret;
2036 	const unsigned blocksize = blks_to_bytes(inode, 1);
2037 	sector_t block_in_file;
2038 	sector_t last_block;
2039 	sector_t last_block_in_file;
2040 	sector_t block_nr;
2041 	int ret = 0;
2042 
2043 	block_in_file = (sector_t)page_index(page);
2044 	last_block = block_in_file + nr_pages;
2045 	last_block_in_file = bytes_to_blks(inode,
2046 			f2fs_readpage_limit(inode) + blocksize - 1);
2047 	if (last_block > last_block_in_file)
2048 		last_block = last_block_in_file;
2049 
2050 	/* just zeroing out page which is beyond EOF */
2051 	if (block_in_file >= last_block)
2052 		goto zero_out;
2053 	/*
2054 	 * Map blocks using the previous result first.
2055 	 */
2056 	if ((map->m_flags & F2FS_MAP_MAPPED) &&
2057 			block_in_file > map->m_lblk &&
2058 			block_in_file < (map->m_lblk + map->m_len))
2059 		goto got_it;
2060 
2061 	/*
2062 	 * Then do more f2fs_map_blocks() calls until we are
2063 	 * done with this page.
2064 	 */
2065 	map->m_lblk = block_in_file;
2066 	map->m_len = last_block - block_in_file;
2067 
2068 	ret = f2fs_map_blocks(inode, map, 0, F2FS_GET_BLOCK_DEFAULT);
2069 	if (ret)
2070 		goto out;
2071 got_it:
2072 	if ((map->m_flags & F2FS_MAP_MAPPED)) {
2073 		block_nr = map->m_pblk + block_in_file - map->m_lblk;
2074 		SetPageMappedToDisk(page);
2075 
2076 		if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), block_nr,
2077 						DATA_GENERIC_ENHANCE_READ)) {
2078 			ret = -EFSCORRUPTED;
2079 			goto out;
2080 		}
2081 	} else {
2082 zero_out:
2083 		zero_user_segment(page, 0, PAGE_SIZE);
2084 		if (f2fs_need_verity(inode, page->index) &&
2085 		    !fsverity_verify_page(page)) {
2086 			ret = -EIO;
2087 			goto out;
2088 		}
2089 		if (!PageUptodate(page))
2090 			SetPageUptodate(page);
2091 		unlock_page(page);
2092 		goto out;
2093 	}
2094 
2095 	/*
2096 	 * This page will go to BIO.  Do we need to send this
2097 	 * BIO off first?
2098 	 */
2099 	if (bio && (!page_is_mergeable(F2FS_I_SB(inode), bio,
2100 				       *last_block_in_bio, block_nr) ||
2101 		    !f2fs_crypt_mergeable_bio(bio, inode, page->index, NULL))) {
2102 submit_and_realloc:
2103 		__submit_bio(F2FS_I_SB(inode), bio, DATA);
2104 		bio = NULL;
2105 	}
2106 	if (bio == NULL) {
2107 		bio = f2fs_grab_read_bio(inode, block_nr, nr_pages,
2108 				is_readahead ? REQ_RAHEAD : 0, page->index,
2109 				false);
2110 		if (IS_ERR(bio)) {
2111 			ret = PTR_ERR(bio);
2112 			bio = NULL;
2113 			goto out;
2114 		}
2115 	}
2116 
2117 	/*
2118 	 * If the page is under writeback, we need to wait for
2119 	 * its completion to see the correct decrypted data.
2120 	 */
2121 	f2fs_wait_on_block_writeback(inode, block_nr);
2122 
2123 	if (bio_add_page(bio, page, blocksize, 0) < blocksize)
2124 		goto submit_and_realloc;
2125 
2126 	inc_page_count(F2FS_I_SB(inode), F2FS_RD_DATA);
2127 	f2fs_update_iostat(F2FS_I_SB(inode), FS_DATA_READ_IO, F2FS_BLKSIZE);
2128 	ClearPageError(page);
2129 	*last_block_in_bio = block_nr;
2130 	goto out;
2131 out:
2132 	*bio_ret = bio;
2133 	return ret;
2134 }
2135 
2136 #ifdef CONFIG_F2FS_FS_COMPRESSION
2137 int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret,
2138 				unsigned nr_pages, sector_t *last_block_in_bio,
2139 				bool is_readahead, bool for_write)
2140 {
2141 	struct dnode_of_data dn;
2142 	struct inode *inode = cc->inode;
2143 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2144 	struct bio *bio = *bio_ret;
2145 	unsigned int start_idx = cc->cluster_idx << cc->log_cluster_size;
2146 	sector_t last_block_in_file;
2147 	const unsigned blocksize = blks_to_bytes(inode, 1);
2148 	struct decompress_io_ctx *dic = NULL;
2149 	struct extent_info ei = {0, };
2150 	bool from_dnode = true;
2151 	int i;
2152 	int ret = 0;
2153 
2154 	f2fs_bug_on(sbi, f2fs_cluster_is_empty(cc));
2155 
2156 	last_block_in_file = bytes_to_blks(inode,
2157 			f2fs_readpage_limit(inode) + blocksize - 1);
2158 
2159 	/* get rid of pages beyond EOF */
2160 	for (i = 0; i < cc->cluster_size; i++) {
2161 		struct page *page = cc->rpages[i];
2162 
2163 		if (!page)
2164 			continue;
2165 		if ((sector_t)page->index >= last_block_in_file) {
2166 			zero_user_segment(page, 0, PAGE_SIZE);
2167 			if (!PageUptodate(page))
2168 				SetPageUptodate(page);
2169 		} else if (!PageUptodate(page)) {
2170 			continue;
2171 		}
2172 		unlock_page(page);
2173 		if (for_write)
2174 			put_page(page);
2175 		cc->rpages[i] = NULL;
2176 		cc->nr_rpages--;
2177 	}
2178 
2179 	/* we are done since all pages are beyond EOF */
2180 	if (f2fs_cluster_is_empty(cc))
2181 		goto out;
2182 
2183 	if (f2fs_lookup_extent_cache(inode, start_idx, &ei))
2184 		from_dnode = false;
2185 
2186 	if (!from_dnode)
2187 		goto skip_reading_dnode;
2188 
2189 	set_new_dnode(&dn, inode, NULL, NULL, 0);
2190 	ret = f2fs_get_dnode_of_data(&dn, start_idx, LOOKUP_NODE);
2191 	if (ret)
2192 		goto out;
2193 
2194 	f2fs_bug_on(sbi, dn.data_blkaddr != COMPRESS_ADDR);
2195 
2196 skip_reading_dnode:
2197 	for (i = 1; i < cc->cluster_size; i++) {
2198 		block_t blkaddr;
2199 
2200 		blkaddr = from_dnode ? data_blkaddr(dn.inode, dn.node_page,
2201 					dn.ofs_in_node + i) :
2202 					ei.blk + i - 1;
2203 
2204 		if (!__is_valid_data_blkaddr(blkaddr))
2205 			break;
2206 
2207 		if (!f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC)) {
2208 			ret = -EFAULT;
2209 			goto out_put_dnode;
2210 		}
2211 		cc->nr_cpages++;
2212 
2213 		if (!from_dnode && i >= ei.c_len)
2214 			break;
2215 	}
2216 
2217 	/* nothing to decompress */
2218 	if (cc->nr_cpages == 0) {
2219 		ret = 0;
2220 		goto out_put_dnode;
2221 	}
2222 
2223 	dic = f2fs_alloc_dic(cc);
2224 	if (IS_ERR(dic)) {
2225 		ret = PTR_ERR(dic);
2226 		goto out_put_dnode;
2227 	}
2228 
2229 	for (i = 0; i < cc->nr_cpages; i++) {
2230 		struct page *page = dic->cpages[i];
2231 		block_t blkaddr;
2232 		struct bio_post_read_ctx *ctx;
2233 
2234 		blkaddr = from_dnode ? data_blkaddr(dn.inode, dn.node_page,
2235 					dn.ofs_in_node + i + 1) :
2236 					ei.blk + i;
2237 
2238 		f2fs_wait_on_block_writeback(inode, blkaddr);
2239 
2240 		if (f2fs_load_compressed_page(sbi, page, blkaddr)) {
2241 			if (atomic_dec_and_test(&dic->remaining_pages))
2242 				f2fs_decompress_cluster(dic, true);
2243 			continue;
2244 		}
2245 
2246 		if (bio && (!page_is_mergeable(sbi, bio,
2247 					*last_block_in_bio, blkaddr) ||
2248 		    !f2fs_crypt_mergeable_bio(bio, inode, page->index, NULL))) {
2249 submit_and_realloc:
2250 			__submit_bio(sbi, bio, DATA);
2251 			bio = NULL;
2252 		}
2253 
2254 		if (!bio) {
2255 			bio = f2fs_grab_read_bio(inode, blkaddr, nr_pages,
2256 					is_readahead ? REQ_RAHEAD : 0,
2257 					page->index, for_write);
2258 			if (IS_ERR(bio)) {
2259 				ret = PTR_ERR(bio);
2260 				f2fs_decompress_end_io(dic, ret, true);
2261 				f2fs_put_dnode(&dn);
2262 				*bio_ret = NULL;
2263 				return ret;
2264 			}
2265 		}
2266 
2267 		if (bio_add_page(bio, page, blocksize, 0) < blocksize)
2268 			goto submit_and_realloc;
2269 
2270 		ctx = get_post_read_ctx(bio);
2271 		ctx->enabled_steps |= STEP_DECOMPRESS;
2272 		refcount_inc(&dic->refcnt);
2273 
2274 		inc_page_count(sbi, F2FS_RD_DATA);
2275 		f2fs_update_iostat(sbi, FS_DATA_READ_IO, F2FS_BLKSIZE);
2276 		f2fs_update_iostat(sbi, FS_CDATA_READ_IO, F2FS_BLKSIZE);
2277 		ClearPageError(page);
2278 		*last_block_in_bio = blkaddr;
2279 	}
2280 
2281 	if (from_dnode)
2282 		f2fs_put_dnode(&dn);
2283 
2284 	*bio_ret = bio;
2285 	return 0;
2286 
2287 out_put_dnode:
2288 	if (from_dnode)
2289 		f2fs_put_dnode(&dn);
2290 out:
2291 	for (i = 0; i < cc->cluster_size; i++) {
2292 		if (cc->rpages[i]) {
2293 			ClearPageUptodate(cc->rpages[i]);
2294 			ClearPageError(cc->rpages[i]);
2295 			unlock_page(cc->rpages[i]);
2296 		}
2297 	}
2298 	*bio_ret = bio;
2299 	return ret;
2300 }
2301 #endif
2302 
2303 /*
2304  * This function was originally taken from fs/mpage.c, and customized for f2fs.
2305  * Major change was from block_size == page_size in f2fs by default.
2306  */
2307 static int f2fs_mpage_readpages(struct inode *inode,
2308 		struct readahead_control *rac, struct page *page)
2309 {
2310 	struct bio *bio = NULL;
2311 	sector_t last_block_in_bio = 0;
2312 	struct f2fs_map_blocks map;
2313 #ifdef CONFIG_F2FS_FS_COMPRESSION
2314 	struct compress_ctx cc = {
2315 		.inode = inode,
2316 		.log_cluster_size = F2FS_I(inode)->i_log_cluster_size,
2317 		.cluster_size = F2FS_I(inode)->i_cluster_size,
2318 		.cluster_idx = NULL_CLUSTER,
2319 		.rpages = NULL,
2320 		.cpages = NULL,
2321 		.nr_rpages = 0,
2322 		.nr_cpages = 0,
2323 	};
2324 	pgoff_t nc_cluster_idx = NULL_CLUSTER;
2325 #endif
2326 	unsigned nr_pages = rac ? readahead_count(rac) : 1;
2327 	unsigned max_nr_pages = nr_pages;
2328 	int ret = 0;
2329 
2330 	map.m_pblk = 0;
2331 	map.m_lblk = 0;
2332 	map.m_len = 0;
2333 	map.m_flags = 0;
2334 	map.m_next_pgofs = NULL;
2335 	map.m_next_extent = NULL;
2336 	map.m_seg_type = NO_CHECK_TYPE;
2337 	map.m_may_create = false;
2338 
2339 	for (; nr_pages; nr_pages--) {
2340 		if (rac) {
2341 			page = readahead_page(rac);
2342 			prefetchw(&page->flags);
2343 		}
2344 
2345 #ifdef CONFIG_F2FS_FS_COMPRESSION
2346 		if (f2fs_compressed_file(inode)) {
2347 			/* there are remained comressed pages, submit them */
2348 			if (!f2fs_cluster_can_merge_page(&cc, page->index)) {
2349 				ret = f2fs_read_multi_pages(&cc, &bio,
2350 							max_nr_pages,
2351 							&last_block_in_bio,
2352 							rac != NULL, false);
2353 				f2fs_destroy_compress_ctx(&cc, false);
2354 				if (ret)
2355 					goto set_error_page;
2356 			}
2357 			if (cc.cluster_idx == NULL_CLUSTER) {
2358 				if (nc_cluster_idx ==
2359 					page->index >> cc.log_cluster_size) {
2360 					goto read_single_page;
2361 				}
2362 
2363 				ret = f2fs_is_compressed_cluster(inode, page->index);
2364 				if (ret < 0)
2365 					goto set_error_page;
2366 				else if (!ret) {
2367 					nc_cluster_idx =
2368 						page->index >> cc.log_cluster_size;
2369 					goto read_single_page;
2370 				}
2371 
2372 				nc_cluster_idx = NULL_CLUSTER;
2373 			}
2374 			ret = f2fs_init_compress_ctx(&cc);
2375 			if (ret)
2376 				goto set_error_page;
2377 
2378 			f2fs_compress_ctx_add_page(&cc, page);
2379 
2380 			goto next_page;
2381 		}
2382 read_single_page:
2383 #endif
2384 
2385 		ret = f2fs_read_single_page(inode, page, max_nr_pages, &map,
2386 					&bio, &last_block_in_bio, rac);
2387 		if (ret) {
2388 #ifdef CONFIG_F2FS_FS_COMPRESSION
2389 set_error_page:
2390 #endif
2391 			SetPageError(page);
2392 			zero_user_segment(page, 0, PAGE_SIZE);
2393 			unlock_page(page);
2394 		}
2395 #ifdef CONFIG_F2FS_FS_COMPRESSION
2396 next_page:
2397 #endif
2398 		if (rac)
2399 			put_page(page);
2400 
2401 #ifdef CONFIG_F2FS_FS_COMPRESSION
2402 		if (f2fs_compressed_file(inode)) {
2403 			/* last page */
2404 			if (nr_pages == 1 && !f2fs_cluster_is_empty(&cc)) {
2405 				ret = f2fs_read_multi_pages(&cc, &bio,
2406 							max_nr_pages,
2407 							&last_block_in_bio,
2408 							rac != NULL, false);
2409 				f2fs_destroy_compress_ctx(&cc, false);
2410 			}
2411 		}
2412 #endif
2413 	}
2414 	if (bio)
2415 		__submit_bio(F2FS_I_SB(inode), bio, DATA);
2416 	return ret;
2417 }
2418 
2419 static int f2fs_read_data_folio(struct file *file, struct folio *folio)
2420 {
2421 	struct page *page = &folio->page;
2422 	struct inode *inode = page_file_mapping(page)->host;
2423 	int ret = -EAGAIN;
2424 
2425 	trace_f2fs_readpage(page, DATA);
2426 
2427 	if (!f2fs_is_compress_backend_ready(inode)) {
2428 		unlock_page(page);
2429 		return -EOPNOTSUPP;
2430 	}
2431 
2432 	/* If the file has inline data, try to read it directly */
2433 	if (f2fs_has_inline_data(inode))
2434 		ret = f2fs_read_inline_data(inode, page);
2435 	if (ret == -EAGAIN)
2436 		ret = f2fs_mpage_readpages(inode, NULL, page);
2437 	return ret;
2438 }
2439 
2440 static void f2fs_readahead(struct readahead_control *rac)
2441 {
2442 	struct inode *inode = rac->mapping->host;
2443 
2444 	trace_f2fs_readpages(inode, readahead_index(rac), readahead_count(rac));
2445 
2446 	if (!f2fs_is_compress_backend_ready(inode))
2447 		return;
2448 
2449 	/* If the file has inline data, skip readahead */
2450 	if (f2fs_has_inline_data(inode))
2451 		return;
2452 
2453 	f2fs_mpage_readpages(inode, rac, NULL);
2454 }
2455 
2456 int f2fs_encrypt_one_page(struct f2fs_io_info *fio)
2457 {
2458 	struct inode *inode = fio->page->mapping->host;
2459 	struct page *mpage, *page;
2460 	gfp_t gfp_flags = GFP_NOFS;
2461 
2462 	if (!f2fs_encrypted_file(inode))
2463 		return 0;
2464 
2465 	page = fio->compressed_page ? fio->compressed_page : fio->page;
2466 
2467 	/* wait for GCed page writeback via META_MAPPING */
2468 	f2fs_wait_on_block_writeback(inode, fio->old_blkaddr);
2469 
2470 	if (fscrypt_inode_uses_inline_crypto(inode))
2471 		return 0;
2472 
2473 retry_encrypt:
2474 	fio->encrypted_page = fscrypt_encrypt_pagecache_blocks(page,
2475 					PAGE_SIZE, 0, gfp_flags);
2476 	if (IS_ERR(fio->encrypted_page)) {
2477 		/* flush pending IOs and wait for a while in the ENOMEM case */
2478 		if (PTR_ERR(fio->encrypted_page) == -ENOMEM) {
2479 			f2fs_flush_merged_writes(fio->sbi);
2480 			memalloc_retry_wait(GFP_NOFS);
2481 			gfp_flags |= __GFP_NOFAIL;
2482 			goto retry_encrypt;
2483 		}
2484 		return PTR_ERR(fio->encrypted_page);
2485 	}
2486 
2487 	mpage = find_lock_page(META_MAPPING(fio->sbi), fio->old_blkaddr);
2488 	if (mpage) {
2489 		if (PageUptodate(mpage))
2490 			memcpy(page_address(mpage),
2491 				page_address(fio->encrypted_page), PAGE_SIZE);
2492 		f2fs_put_page(mpage, 1);
2493 	}
2494 	return 0;
2495 }
2496 
2497 static inline bool check_inplace_update_policy(struct inode *inode,
2498 				struct f2fs_io_info *fio)
2499 {
2500 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2501 	unsigned int policy = SM_I(sbi)->ipu_policy;
2502 
2503 	if (policy & (0x1 << F2FS_IPU_HONOR_OPU_WRITE) &&
2504 			is_inode_flag_set(inode, FI_OPU_WRITE))
2505 		return false;
2506 	if (policy & (0x1 << F2FS_IPU_FORCE))
2507 		return true;
2508 	if (policy & (0x1 << F2FS_IPU_SSR) && f2fs_need_SSR(sbi))
2509 		return true;
2510 	if (policy & (0x1 << F2FS_IPU_UTIL) &&
2511 			utilization(sbi) > SM_I(sbi)->min_ipu_util)
2512 		return true;
2513 	if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && f2fs_need_SSR(sbi) &&
2514 			utilization(sbi) > SM_I(sbi)->min_ipu_util)
2515 		return true;
2516 
2517 	/*
2518 	 * IPU for rewrite async pages
2519 	 */
2520 	if (policy & (0x1 << F2FS_IPU_ASYNC) &&
2521 			fio && fio->op == REQ_OP_WRITE &&
2522 			!(fio->op_flags & REQ_SYNC) &&
2523 			!IS_ENCRYPTED(inode))
2524 		return true;
2525 
2526 	/* this is only set during fdatasync */
2527 	if (policy & (0x1 << F2FS_IPU_FSYNC) &&
2528 			is_inode_flag_set(inode, FI_NEED_IPU))
2529 		return true;
2530 
2531 	if (unlikely(fio && is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2532 			!f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2533 		return true;
2534 
2535 	return false;
2536 }
2537 
2538 bool f2fs_should_update_inplace(struct inode *inode, struct f2fs_io_info *fio)
2539 {
2540 	/* swap file is migrating in aligned write mode */
2541 	if (is_inode_flag_set(inode, FI_ALIGNED_WRITE))
2542 		return false;
2543 
2544 	if (f2fs_is_pinned_file(inode))
2545 		return true;
2546 
2547 	/* if this is cold file, we should overwrite to avoid fragmentation */
2548 	if (file_is_cold(inode))
2549 		return true;
2550 
2551 	return check_inplace_update_policy(inode, fio);
2552 }
2553 
2554 bool f2fs_should_update_outplace(struct inode *inode, struct f2fs_io_info *fio)
2555 {
2556 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2557 
2558 	/* The below cases were checked when setting it. */
2559 	if (f2fs_is_pinned_file(inode))
2560 		return false;
2561 	if (fio && is_sbi_flag_set(sbi, SBI_NEED_FSCK))
2562 		return true;
2563 	if (f2fs_lfs_mode(sbi))
2564 		return true;
2565 	if (S_ISDIR(inode->i_mode))
2566 		return true;
2567 	if (IS_NOQUOTA(inode))
2568 		return true;
2569 	if (f2fs_is_atomic_file(inode))
2570 		return true;
2571 
2572 	/* swap file is migrating in aligned write mode */
2573 	if (is_inode_flag_set(inode, FI_ALIGNED_WRITE))
2574 		return true;
2575 
2576 	if (is_inode_flag_set(inode, FI_OPU_WRITE))
2577 		return true;
2578 
2579 	if (fio) {
2580 		if (page_private_gcing(fio->page))
2581 			return true;
2582 		if (page_private_dummy(fio->page))
2583 			return true;
2584 		if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2585 			f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2586 			return true;
2587 	}
2588 	return false;
2589 }
2590 
2591 static inline bool need_inplace_update(struct f2fs_io_info *fio)
2592 {
2593 	struct inode *inode = fio->page->mapping->host;
2594 
2595 	if (f2fs_should_update_outplace(inode, fio))
2596 		return false;
2597 
2598 	return f2fs_should_update_inplace(inode, fio);
2599 }
2600 
2601 int f2fs_do_write_data_page(struct f2fs_io_info *fio)
2602 {
2603 	struct page *page = fio->page;
2604 	struct inode *inode = page->mapping->host;
2605 	struct dnode_of_data dn;
2606 	struct extent_info ei = {0, };
2607 	struct node_info ni;
2608 	bool ipu_force = false;
2609 	int err = 0;
2610 
2611 	/* Use COW inode to make dnode_of_data for atomic write */
2612 	if (f2fs_is_atomic_file(inode))
2613 		set_new_dnode(&dn, F2FS_I(inode)->cow_inode, NULL, NULL, 0);
2614 	else
2615 		set_new_dnode(&dn, inode, NULL, NULL, 0);
2616 
2617 	if (need_inplace_update(fio) &&
2618 			f2fs_lookup_extent_cache(inode, page->index, &ei)) {
2619 		fio->old_blkaddr = ei.blk + page->index - ei.fofs;
2620 
2621 		if (!f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
2622 						DATA_GENERIC_ENHANCE))
2623 			return -EFSCORRUPTED;
2624 
2625 		ipu_force = true;
2626 		fio->need_lock = LOCK_DONE;
2627 		goto got_it;
2628 	}
2629 
2630 	/* Deadlock due to between page->lock and f2fs_lock_op */
2631 	if (fio->need_lock == LOCK_REQ && !f2fs_trylock_op(fio->sbi))
2632 		return -EAGAIN;
2633 
2634 	err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
2635 	if (err)
2636 		goto out;
2637 
2638 	fio->old_blkaddr = dn.data_blkaddr;
2639 
2640 	/* This page is already truncated */
2641 	if (fio->old_blkaddr == NULL_ADDR) {
2642 		ClearPageUptodate(page);
2643 		clear_page_private_gcing(page);
2644 		goto out_writepage;
2645 	}
2646 got_it:
2647 	if (__is_valid_data_blkaddr(fio->old_blkaddr) &&
2648 		!f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
2649 						DATA_GENERIC_ENHANCE)) {
2650 		err = -EFSCORRUPTED;
2651 		goto out_writepage;
2652 	}
2653 
2654 	/*
2655 	 * If current allocation needs SSR,
2656 	 * it had better in-place writes for updated data.
2657 	 */
2658 	if (ipu_force ||
2659 		(__is_valid_data_blkaddr(fio->old_blkaddr) &&
2660 					need_inplace_update(fio))) {
2661 		err = f2fs_encrypt_one_page(fio);
2662 		if (err)
2663 			goto out_writepage;
2664 
2665 		set_page_writeback(page);
2666 		ClearPageError(page);
2667 		f2fs_put_dnode(&dn);
2668 		if (fio->need_lock == LOCK_REQ)
2669 			f2fs_unlock_op(fio->sbi);
2670 		err = f2fs_inplace_write_data(fio);
2671 		if (err) {
2672 			if (fscrypt_inode_uses_fs_layer_crypto(inode))
2673 				fscrypt_finalize_bounce_page(&fio->encrypted_page);
2674 			if (PageWriteback(page))
2675 				end_page_writeback(page);
2676 		} else {
2677 			set_inode_flag(inode, FI_UPDATE_WRITE);
2678 		}
2679 		trace_f2fs_do_write_data_page(fio->page, IPU);
2680 		return err;
2681 	}
2682 
2683 	if (fio->need_lock == LOCK_RETRY) {
2684 		if (!f2fs_trylock_op(fio->sbi)) {
2685 			err = -EAGAIN;
2686 			goto out_writepage;
2687 		}
2688 		fio->need_lock = LOCK_REQ;
2689 	}
2690 
2691 	err = f2fs_get_node_info(fio->sbi, dn.nid, &ni, false);
2692 	if (err)
2693 		goto out_writepage;
2694 
2695 	fio->version = ni.version;
2696 
2697 	err = f2fs_encrypt_one_page(fio);
2698 	if (err)
2699 		goto out_writepage;
2700 
2701 	set_page_writeback(page);
2702 	ClearPageError(page);
2703 
2704 	if (fio->compr_blocks && fio->old_blkaddr == COMPRESS_ADDR)
2705 		f2fs_i_compr_blocks_update(inode, fio->compr_blocks - 1, false);
2706 
2707 	/* LFS mode write path */
2708 	f2fs_outplace_write_data(&dn, fio);
2709 	trace_f2fs_do_write_data_page(page, OPU);
2710 	set_inode_flag(inode, FI_APPEND_WRITE);
2711 	if (page->index == 0)
2712 		set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
2713 out_writepage:
2714 	f2fs_put_dnode(&dn);
2715 out:
2716 	if (fio->need_lock == LOCK_REQ)
2717 		f2fs_unlock_op(fio->sbi);
2718 	return err;
2719 }
2720 
2721 int f2fs_write_single_data_page(struct page *page, int *submitted,
2722 				struct bio **bio,
2723 				sector_t *last_block,
2724 				struct writeback_control *wbc,
2725 				enum iostat_type io_type,
2726 				int compr_blocks,
2727 				bool allow_balance)
2728 {
2729 	struct inode *inode = page->mapping->host;
2730 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2731 	loff_t i_size = i_size_read(inode);
2732 	const pgoff_t end_index = ((unsigned long long)i_size)
2733 							>> PAGE_SHIFT;
2734 	loff_t psize = (loff_t)(page->index + 1) << PAGE_SHIFT;
2735 	unsigned offset = 0;
2736 	bool need_balance_fs = false;
2737 	int err = 0;
2738 	struct f2fs_io_info fio = {
2739 		.sbi = sbi,
2740 		.ino = inode->i_ino,
2741 		.type = DATA,
2742 		.op = REQ_OP_WRITE,
2743 		.op_flags = wbc_to_write_flags(wbc),
2744 		.old_blkaddr = NULL_ADDR,
2745 		.page = page,
2746 		.encrypted_page = NULL,
2747 		.submitted = false,
2748 		.compr_blocks = compr_blocks,
2749 		.need_lock = LOCK_RETRY,
2750 		.post_read = f2fs_post_read_required(inode),
2751 		.io_type = io_type,
2752 		.io_wbc = wbc,
2753 		.bio = bio,
2754 		.last_block = last_block,
2755 	};
2756 
2757 	trace_f2fs_writepage(page, DATA);
2758 
2759 	/* we should bypass data pages to proceed the kworkder jobs */
2760 	if (unlikely(f2fs_cp_error(sbi))) {
2761 		mapping_set_error(page->mapping, -EIO);
2762 		/*
2763 		 * don't drop any dirty dentry pages for keeping lastest
2764 		 * directory structure.
2765 		 */
2766 		if (S_ISDIR(inode->i_mode))
2767 			goto redirty_out;
2768 		goto out;
2769 	}
2770 
2771 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
2772 		goto redirty_out;
2773 
2774 	if (page->index < end_index ||
2775 			f2fs_verity_in_progress(inode) ||
2776 			compr_blocks)
2777 		goto write;
2778 
2779 	/*
2780 	 * If the offset is out-of-range of file size,
2781 	 * this page does not have to be written to disk.
2782 	 */
2783 	offset = i_size & (PAGE_SIZE - 1);
2784 	if ((page->index >= end_index + 1) || !offset)
2785 		goto out;
2786 
2787 	zero_user_segment(page, offset, PAGE_SIZE);
2788 write:
2789 	if (f2fs_is_drop_cache(inode))
2790 		goto out;
2791 
2792 	/* Dentry/quota blocks are controlled by checkpoint */
2793 	if (S_ISDIR(inode->i_mode) || IS_NOQUOTA(inode)) {
2794 		/*
2795 		 * We need to wait for node_write to avoid block allocation during
2796 		 * checkpoint. This can only happen to quota writes which can cause
2797 		 * the below discard race condition.
2798 		 */
2799 		if (IS_NOQUOTA(inode))
2800 			f2fs_down_read(&sbi->node_write);
2801 
2802 		fio.need_lock = LOCK_DONE;
2803 		err = f2fs_do_write_data_page(&fio);
2804 
2805 		if (IS_NOQUOTA(inode))
2806 			f2fs_up_read(&sbi->node_write);
2807 
2808 		goto done;
2809 	}
2810 
2811 	if (!wbc->for_reclaim)
2812 		need_balance_fs = true;
2813 	else if (has_not_enough_free_secs(sbi, 0, 0))
2814 		goto redirty_out;
2815 	else
2816 		set_inode_flag(inode, FI_HOT_DATA);
2817 
2818 	err = -EAGAIN;
2819 	if (f2fs_has_inline_data(inode)) {
2820 		err = f2fs_write_inline_data(inode, page);
2821 		if (!err)
2822 			goto out;
2823 	}
2824 
2825 	if (err == -EAGAIN) {
2826 		err = f2fs_do_write_data_page(&fio);
2827 		if (err == -EAGAIN) {
2828 			fio.need_lock = LOCK_REQ;
2829 			err = f2fs_do_write_data_page(&fio);
2830 		}
2831 	}
2832 
2833 	if (err) {
2834 		file_set_keep_isize(inode);
2835 	} else {
2836 		spin_lock(&F2FS_I(inode)->i_size_lock);
2837 		if (F2FS_I(inode)->last_disk_size < psize)
2838 			F2FS_I(inode)->last_disk_size = psize;
2839 		spin_unlock(&F2FS_I(inode)->i_size_lock);
2840 	}
2841 
2842 done:
2843 	if (err && err != -ENOENT)
2844 		goto redirty_out;
2845 
2846 out:
2847 	inode_dec_dirty_pages(inode);
2848 	if (err) {
2849 		ClearPageUptodate(page);
2850 		clear_page_private_gcing(page);
2851 	}
2852 
2853 	if (wbc->for_reclaim) {
2854 		f2fs_submit_merged_write_cond(sbi, NULL, page, 0, DATA);
2855 		clear_inode_flag(inode, FI_HOT_DATA);
2856 		f2fs_remove_dirty_inode(inode);
2857 		submitted = NULL;
2858 	}
2859 	unlock_page(page);
2860 	if (!S_ISDIR(inode->i_mode) && !IS_NOQUOTA(inode) &&
2861 			!F2FS_I(inode)->cp_task && allow_balance)
2862 		f2fs_balance_fs(sbi, need_balance_fs);
2863 
2864 	if (unlikely(f2fs_cp_error(sbi))) {
2865 		f2fs_submit_merged_write(sbi, DATA);
2866 		f2fs_submit_merged_ipu_write(sbi, bio, NULL);
2867 		submitted = NULL;
2868 	}
2869 
2870 	if (submitted)
2871 		*submitted = fio.submitted ? 1 : 0;
2872 
2873 	return 0;
2874 
2875 redirty_out:
2876 	redirty_page_for_writepage(wbc, page);
2877 	/*
2878 	 * pageout() in MM traslates EAGAIN, so calls handle_write_error()
2879 	 * -> mapping_set_error() -> set_bit(AS_EIO, ...).
2880 	 * file_write_and_wait_range() will see EIO error, which is critical
2881 	 * to return value of fsync() followed by atomic_write failure to user.
2882 	 */
2883 	if (!err || wbc->for_reclaim)
2884 		return AOP_WRITEPAGE_ACTIVATE;
2885 	unlock_page(page);
2886 	return err;
2887 }
2888 
2889 static int f2fs_write_data_page(struct page *page,
2890 					struct writeback_control *wbc)
2891 {
2892 #ifdef CONFIG_F2FS_FS_COMPRESSION
2893 	struct inode *inode = page->mapping->host;
2894 
2895 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2896 		goto out;
2897 
2898 	if (f2fs_compressed_file(inode)) {
2899 		if (f2fs_is_compressed_cluster(inode, page->index)) {
2900 			redirty_page_for_writepage(wbc, page);
2901 			return AOP_WRITEPAGE_ACTIVATE;
2902 		}
2903 	}
2904 out:
2905 #endif
2906 
2907 	return f2fs_write_single_data_page(page, NULL, NULL, NULL,
2908 						wbc, FS_DATA_IO, 0, true);
2909 }
2910 
2911 /*
2912  * This function was copied from write_cche_pages from mm/page-writeback.c.
2913  * The major change is making write step of cold data page separately from
2914  * warm/hot data page.
2915  */
2916 static int f2fs_write_cache_pages(struct address_space *mapping,
2917 					struct writeback_control *wbc,
2918 					enum iostat_type io_type)
2919 {
2920 	int ret = 0;
2921 	int done = 0, retry = 0;
2922 	struct page *pages[F2FS_ONSTACK_PAGES];
2923 	struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
2924 	struct bio *bio = NULL;
2925 	sector_t last_block;
2926 #ifdef CONFIG_F2FS_FS_COMPRESSION
2927 	struct inode *inode = mapping->host;
2928 	struct compress_ctx cc = {
2929 		.inode = inode,
2930 		.log_cluster_size = F2FS_I(inode)->i_log_cluster_size,
2931 		.cluster_size = F2FS_I(inode)->i_cluster_size,
2932 		.cluster_idx = NULL_CLUSTER,
2933 		.rpages = NULL,
2934 		.nr_rpages = 0,
2935 		.cpages = NULL,
2936 		.valid_nr_cpages = 0,
2937 		.rbuf = NULL,
2938 		.cbuf = NULL,
2939 		.rlen = PAGE_SIZE * F2FS_I(inode)->i_cluster_size,
2940 		.private = NULL,
2941 	};
2942 #endif
2943 	int nr_pages;
2944 	pgoff_t index;
2945 	pgoff_t end;		/* Inclusive */
2946 	pgoff_t done_index;
2947 	int range_whole = 0;
2948 	xa_mark_t tag;
2949 	int nwritten = 0;
2950 	int submitted = 0;
2951 	int i;
2952 
2953 	if (get_dirty_pages(mapping->host) <=
2954 				SM_I(F2FS_M_SB(mapping))->min_hot_blocks)
2955 		set_inode_flag(mapping->host, FI_HOT_DATA);
2956 	else
2957 		clear_inode_flag(mapping->host, FI_HOT_DATA);
2958 
2959 	if (wbc->range_cyclic) {
2960 		index = mapping->writeback_index; /* prev offset */
2961 		end = -1;
2962 	} else {
2963 		index = wbc->range_start >> PAGE_SHIFT;
2964 		end = wbc->range_end >> PAGE_SHIFT;
2965 		if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
2966 			range_whole = 1;
2967 	}
2968 	if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2969 		tag = PAGECACHE_TAG_TOWRITE;
2970 	else
2971 		tag = PAGECACHE_TAG_DIRTY;
2972 retry:
2973 	retry = 0;
2974 	if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2975 		tag_pages_for_writeback(mapping, index, end);
2976 	done_index = index;
2977 	while (!done && !retry && (index <= end)) {
2978 		nr_pages = find_get_pages_range_tag(mapping, &index, end,
2979 				tag, F2FS_ONSTACK_PAGES, pages);
2980 		if (nr_pages == 0)
2981 			break;
2982 
2983 		for (i = 0; i < nr_pages; i++) {
2984 			struct page *page = pages[i];
2985 			bool need_readd;
2986 readd:
2987 			need_readd = false;
2988 #ifdef CONFIG_F2FS_FS_COMPRESSION
2989 			if (f2fs_compressed_file(inode)) {
2990 				void *fsdata = NULL;
2991 				struct page *pagep;
2992 				int ret2;
2993 
2994 				ret = f2fs_init_compress_ctx(&cc);
2995 				if (ret) {
2996 					done = 1;
2997 					break;
2998 				}
2999 
3000 				if (!f2fs_cluster_can_merge_page(&cc,
3001 								page->index)) {
3002 					ret = f2fs_write_multi_pages(&cc,
3003 						&submitted, wbc, io_type);
3004 					if (!ret)
3005 						need_readd = true;
3006 					goto result;
3007 				}
3008 
3009 				if (unlikely(f2fs_cp_error(sbi)))
3010 					goto lock_page;
3011 
3012 				if (!f2fs_cluster_is_empty(&cc))
3013 					goto lock_page;
3014 
3015 				if (f2fs_all_cluster_page_ready(&cc,
3016 					pages, i, nr_pages, true))
3017 					goto lock_page;
3018 
3019 				ret2 = f2fs_prepare_compress_overwrite(
3020 							inode, &pagep,
3021 							page->index, &fsdata);
3022 				if (ret2 < 0) {
3023 					ret = ret2;
3024 					done = 1;
3025 					break;
3026 				} else if (ret2 &&
3027 					(!f2fs_compress_write_end(inode,
3028 						fsdata, page->index, 1) ||
3029 					 !f2fs_all_cluster_page_ready(&cc,
3030 						pages, i, nr_pages, false))) {
3031 					retry = 1;
3032 					break;
3033 				}
3034 			}
3035 #endif
3036 			/* give a priority to WB_SYNC threads */
3037 			if (atomic_read(&sbi->wb_sync_req[DATA]) &&
3038 					wbc->sync_mode == WB_SYNC_NONE) {
3039 				done = 1;
3040 				break;
3041 			}
3042 #ifdef CONFIG_F2FS_FS_COMPRESSION
3043 lock_page:
3044 #endif
3045 			done_index = page->index;
3046 retry_write:
3047 			lock_page(page);
3048 
3049 			if (unlikely(page->mapping != mapping)) {
3050 continue_unlock:
3051 				unlock_page(page);
3052 				continue;
3053 			}
3054 
3055 			if (!PageDirty(page)) {
3056 				/* someone wrote it for us */
3057 				goto continue_unlock;
3058 			}
3059 
3060 			if (PageWriteback(page)) {
3061 				if (wbc->sync_mode != WB_SYNC_NONE)
3062 					f2fs_wait_on_page_writeback(page,
3063 							DATA, true, true);
3064 				else
3065 					goto continue_unlock;
3066 			}
3067 
3068 			if (!clear_page_dirty_for_io(page))
3069 				goto continue_unlock;
3070 
3071 #ifdef CONFIG_F2FS_FS_COMPRESSION
3072 			if (f2fs_compressed_file(inode)) {
3073 				get_page(page);
3074 				f2fs_compress_ctx_add_page(&cc, page);
3075 				continue;
3076 			}
3077 #endif
3078 			ret = f2fs_write_single_data_page(page, &submitted,
3079 					&bio, &last_block, wbc, io_type,
3080 					0, true);
3081 			if (ret == AOP_WRITEPAGE_ACTIVATE)
3082 				unlock_page(page);
3083 #ifdef CONFIG_F2FS_FS_COMPRESSION
3084 result:
3085 #endif
3086 			nwritten += submitted;
3087 			wbc->nr_to_write -= submitted;
3088 
3089 			if (unlikely(ret)) {
3090 				/*
3091 				 * keep nr_to_write, since vfs uses this to
3092 				 * get # of written pages.
3093 				 */
3094 				if (ret == AOP_WRITEPAGE_ACTIVATE) {
3095 					ret = 0;
3096 					goto next;
3097 				} else if (ret == -EAGAIN) {
3098 					ret = 0;
3099 					if (wbc->sync_mode == WB_SYNC_ALL) {
3100 						f2fs_io_schedule_timeout(
3101 							DEFAULT_IO_TIMEOUT);
3102 						goto retry_write;
3103 					}
3104 					goto next;
3105 				}
3106 				done_index = page->index + 1;
3107 				done = 1;
3108 				break;
3109 			}
3110 
3111 			if (wbc->nr_to_write <= 0 &&
3112 					wbc->sync_mode == WB_SYNC_NONE) {
3113 				done = 1;
3114 				break;
3115 			}
3116 next:
3117 			if (need_readd)
3118 				goto readd;
3119 		}
3120 		release_pages(pages, nr_pages);
3121 		cond_resched();
3122 	}
3123 #ifdef CONFIG_F2FS_FS_COMPRESSION
3124 	/* flush remained pages in compress cluster */
3125 	if (f2fs_compressed_file(inode) && !f2fs_cluster_is_empty(&cc)) {
3126 		ret = f2fs_write_multi_pages(&cc, &submitted, wbc, io_type);
3127 		nwritten += submitted;
3128 		wbc->nr_to_write -= submitted;
3129 		if (ret) {
3130 			done = 1;
3131 			retry = 0;
3132 		}
3133 	}
3134 	if (f2fs_compressed_file(inode))
3135 		f2fs_destroy_compress_ctx(&cc, false);
3136 #endif
3137 	if (retry) {
3138 		index = 0;
3139 		end = -1;
3140 		goto retry;
3141 	}
3142 	if (wbc->range_cyclic && !done)
3143 		done_index = 0;
3144 	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
3145 		mapping->writeback_index = done_index;
3146 
3147 	if (nwritten)
3148 		f2fs_submit_merged_write_cond(F2FS_M_SB(mapping), mapping->host,
3149 								NULL, 0, DATA);
3150 	/* submit cached bio of IPU write */
3151 	if (bio)
3152 		f2fs_submit_merged_ipu_write(sbi, &bio, NULL);
3153 
3154 	return ret;
3155 }
3156 
3157 static inline bool __should_serialize_io(struct inode *inode,
3158 					struct writeback_control *wbc)
3159 {
3160 	/* to avoid deadlock in path of data flush */
3161 	if (F2FS_I(inode)->cp_task)
3162 		return false;
3163 
3164 	if (!S_ISREG(inode->i_mode))
3165 		return false;
3166 	if (IS_NOQUOTA(inode))
3167 		return false;
3168 
3169 	if (f2fs_need_compress_data(inode))
3170 		return true;
3171 	if (wbc->sync_mode != WB_SYNC_ALL)
3172 		return true;
3173 	if (get_dirty_pages(inode) >= SM_I(F2FS_I_SB(inode))->min_seq_blocks)
3174 		return true;
3175 	return false;
3176 }
3177 
3178 static int __f2fs_write_data_pages(struct address_space *mapping,
3179 						struct writeback_control *wbc,
3180 						enum iostat_type io_type)
3181 {
3182 	struct inode *inode = mapping->host;
3183 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3184 	struct blk_plug plug;
3185 	int ret;
3186 	bool locked = false;
3187 
3188 	/* deal with chardevs and other special file */
3189 	if (!mapping->a_ops->writepage)
3190 		return 0;
3191 
3192 	/* skip writing if there is no dirty page in this inode */
3193 	if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE)
3194 		return 0;
3195 
3196 	/* during POR, we don't need to trigger writepage at all. */
3197 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
3198 		goto skip_write;
3199 
3200 	if ((S_ISDIR(inode->i_mode) || IS_NOQUOTA(inode)) &&
3201 			wbc->sync_mode == WB_SYNC_NONE &&
3202 			get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) &&
3203 			f2fs_available_free_memory(sbi, DIRTY_DENTS))
3204 		goto skip_write;
3205 
3206 	/* skip writing in file defragment preparing stage */
3207 	if (is_inode_flag_set(inode, FI_SKIP_WRITES))
3208 		goto skip_write;
3209 
3210 	trace_f2fs_writepages(mapping->host, wbc, DATA);
3211 
3212 	/* to avoid spliting IOs due to mixed WB_SYNC_ALL and WB_SYNC_NONE */
3213 	if (wbc->sync_mode == WB_SYNC_ALL)
3214 		atomic_inc(&sbi->wb_sync_req[DATA]);
3215 	else if (atomic_read(&sbi->wb_sync_req[DATA])) {
3216 		/* to avoid potential deadlock */
3217 		if (current->plug)
3218 			blk_finish_plug(current->plug);
3219 		goto skip_write;
3220 	}
3221 
3222 	if (__should_serialize_io(inode, wbc)) {
3223 		mutex_lock(&sbi->writepages);
3224 		locked = true;
3225 	}
3226 
3227 	blk_start_plug(&plug);
3228 	ret = f2fs_write_cache_pages(mapping, wbc, io_type);
3229 	blk_finish_plug(&plug);
3230 
3231 	if (locked)
3232 		mutex_unlock(&sbi->writepages);
3233 
3234 	if (wbc->sync_mode == WB_SYNC_ALL)
3235 		atomic_dec(&sbi->wb_sync_req[DATA]);
3236 	/*
3237 	 * if some pages were truncated, we cannot guarantee its mapping->host
3238 	 * to detect pending bios.
3239 	 */
3240 
3241 	f2fs_remove_dirty_inode(inode);
3242 	return ret;
3243 
3244 skip_write:
3245 	wbc->pages_skipped += get_dirty_pages(inode);
3246 	trace_f2fs_writepages(mapping->host, wbc, DATA);
3247 	return 0;
3248 }
3249 
3250 static int f2fs_write_data_pages(struct address_space *mapping,
3251 			    struct writeback_control *wbc)
3252 {
3253 	struct inode *inode = mapping->host;
3254 
3255 	return __f2fs_write_data_pages(mapping, wbc,
3256 			F2FS_I(inode)->cp_task == current ?
3257 			FS_CP_DATA_IO : FS_DATA_IO);
3258 }
3259 
3260 void f2fs_write_failed(struct inode *inode, loff_t to)
3261 {
3262 	loff_t i_size = i_size_read(inode);
3263 
3264 	if (IS_NOQUOTA(inode))
3265 		return;
3266 
3267 	/* In the fs-verity case, f2fs_end_enable_verity() does the truncate */
3268 	if (to > i_size && !f2fs_verity_in_progress(inode)) {
3269 		f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3270 		filemap_invalidate_lock(inode->i_mapping);
3271 
3272 		truncate_pagecache(inode, i_size);
3273 		f2fs_truncate_blocks(inode, i_size, true);
3274 
3275 		filemap_invalidate_unlock(inode->i_mapping);
3276 		f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3277 	}
3278 }
3279 
3280 static int prepare_write_begin(struct f2fs_sb_info *sbi,
3281 			struct page *page, loff_t pos, unsigned len,
3282 			block_t *blk_addr, bool *node_changed)
3283 {
3284 	struct inode *inode = page->mapping->host;
3285 	pgoff_t index = page->index;
3286 	struct dnode_of_data dn;
3287 	struct page *ipage;
3288 	bool locked = false;
3289 	struct extent_info ei = {0, };
3290 	int err = 0;
3291 	int flag;
3292 
3293 	/*
3294 	 * If a whole page is being written and we already preallocated all the
3295 	 * blocks, then there is no need to get a block address now.
3296 	 */
3297 	if (len == PAGE_SIZE && is_inode_flag_set(inode, FI_PREALLOCATED_ALL))
3298 		return 0;
3299 
3300 	/* f2fs_lock_op avoids race between write CP and convert_inline_page */
3301 	if (f2fs_has_inline_data(inode) && pos + len > MAX_INLINE_DATA(inode))
3302 		flag = F2FS_GET_BLOCK_DEFAULT;
3303 	else
3304 		flag = F2FS_GET_BLOCK_PRE_AIO;
3305 
3306 	if (f2fs_has_inline_data(inode) ||
3307 			(pos & PAGE_MASK) >= i_size_read(inode)) {
3308 		f2fs_do_map_lock(sbi, flag, true);
3309 		locked = true;
3310 	}
3311 
3312 restart:
3313 	/* check inline_data */
3314 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
3315 	if (IS_ERR(ipage)) {
3316 		err = PTR_ERR(ipage);
3317 		goto unlock_out;
3318 	}
3319 
3320 	set_new_dnode(&dn, inode, ipage, ipage, 0);
3321 
3322 	if (f2fs_has_inline_data(inode)) {
3323 		if (pos + len <= MAX_INLINE_DATA(inode)) {
3324 			f2fs_do_read_inline_data(page, ipage);
3325 			set_inode_flag(inode, FI_DATA_EXIST);
3326 			if (inode->i_nlink)
3327 				set_page_private_inline(ipage);
3328 		} else {
3329 			err = f2fs_convert_inline_page(&dn, page);
3330 			if (err)
3331 				goto out;
3332 			if (dn.data_blkaddr == NULL_ADDR)
3333 				err = f2fs_get_block(&dn, index);
3334 		}
3335 	} else if (locked) {
3336 		err = f2fs_get_block(&dn, index);
3337 	} else {
3338 		if (f2fs_lookup_extent_cache(inode, index, &ei)) {
3339 			dn.data_blkaddr = ei.blk + index - ei.fofs;
3340 		} else {
3341 			/* hole case */
3342 			err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3343 			if (err || dn.data_blkaddr == NULL_ADDR) {
3344 				f2fs_put_dnode(&dn);
3345 				f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO,
3346 								true);
3347 				WARN_ON(flag != F2FS_GET_BLOCK_PRE_AIO);
3348 				locked = true;
3349 				goto restart;
3350 			}
3351 		}
3352 	}
3353 
3354 	/* convert_inline_page can make node_changed */
3355 	*blk_addr = dn.data_blkaddr;
3356 	*node_changed = dn.node_changed;
3357 out:
3358 	f2fs_put_dnode(&dn);
3359 unlock_out:
3360 	if (locked)
3361 		f2fs_do_map_lock(sbi, flag, false);
3362 	return err;
3363 }
3364 
3365 static int __find_data_block(struct inode *inode, pgoff_t index,
3366 				block_t *blk_addr)
3367 {
3368 	struct dnode_of_data dn;
3369 	struct page *ipage;
3370 	struct extent_info ei = {0, };
3371 	int err = 0;
3372 
3373 	ipage = f2fs_get_node_page(F2FS_I_SB(inode), inode->i_ino);
3374 	if (IS_ERR(ipage))
3375 		return PTR_ERR(ipage);
3376 
3377 	set_new_dnode(&dn, inode, ipage, ipage, 0);
3378 
3379 	if (f2fs_lookup_extent_cache(inode, index, &ei)) {
3380 		dn.data_blkaddr = ei.blk + index - ei.fofs;
3381 	} else {
3382 		/* hole case */
3383 		err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3384 		if (err) {
3385 			dn.data_blkaddr = NULL_ADDR;
3386 			err = 0;
3387 		}
3388 	}
3389 	*blk_addr = dn.data_blkaddr;
3390 	f2fs_put_dnode(&dn);
3391 	return err;
3392 }
3393 
3394 static int __reserve_data_block(struct inode *inode, pgoff_t index,
3395 				block_t *blk_addr, bool *node_changed)
3396 {
3397 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3398 	struct dnode_of_data dn;
3399 	struct page *ipage;
3400 	int err = 0;
3401 
3402 	f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
3403 
3404 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
3405 	if (IS_ERR(ipage)) {
3406 		err = PTR_ERR(ipage);
3407 		goto unlock_out;
3408 	}
3409 	set_new_dnode(&dn, inode, ipage, ipage, 0);
3410 
3411 	err = f2fs_get_block(&dn, index);
3412 
3413 	*blk_addr = dn.data_blkaddr;
3414 	*node_changed = dn.node_changed;
3415 	f2fs_put_dnode(&dn);
3416 
3417 unlock_out:
3418 	f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
3419 	return err;
3420 }
3421 
3422 static int prepare_atomic_write_begin(struct f2fs_sb_info *sbi,
3423 			struct page *page, loff_t pos, unsigned int len,
3424 			block_t *blk_addr, bool *node_changed)
3425 {
3426 	struct inode *inode = page->mapping->host;
3427 	struct inode *cow_inode = F2FS_I(inode)->cow_inode;
3428 	pgoff_t index = page->index;
3429 	int err = 0;
3430 	block_t ori_blk_addr = NULL_ADDR;
3431 
3432 	/* If pos is beyond the end of file, reserve a new block in COW inode */
3433 	if ((pos & PAGE_MASK) >= i_size_read(inode))
3434 		goto reserve_block;
3435 
3436 	/* Look for the block in COW inode first */
3437 	err = __find_data_block(cow_inode, index, blk_addr);
3438 	if (err)
3439 		return err;
3440 	else if (*blk_addr != NULL_ADDR)
3441 		return 0;
3442 
3443 	/* Look for the block in the original inode */
3444 	err = __find_data_block(inode, index, &ori_blk_addr);
3445 	if (err)
3446 		return err;
3447 
3448 reserve_block:
3449 	/* Finally, we should reserve a new block in COW inode for the update */
3450 	err = __reserve_data_block(cow_inode, index, blk_addr, node_changed);
3451 	if (err)
3452 		return err;
3453 	inc_atomic_write_cnt(inode);
3454 
3455 	if (ori_blk_addr != NULL_ADDR)
3456 		*blk_addr = ori_blk_addr;
3457 	return 0;
3458 }
3459 
3460 static int f2fs_write_begin(struct file *file, struct address_space *mapping,
3461 		loff_t pos, unsigned len, struct page **pagep, void **fsdata)
3462 {
3463 	struct inode *inode = mapping->host;
3464 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3465 	struct page *page = NULL;
3466 	pgoff_t index = ((unsigned long long) pos) >> PAGE_SHIFT;
3467 	bool need_balance = false;
3468 	block_t blkaddr = NULL_ADDR;
3469 	int err = 0;
3470 
3471 	trace_f2fs_write_begin(inode, pos, len);
3472 
3473 	if (!f2fs_is_checkpoint_ready(sbi)) {
3474 		err = -ENOSPC;
3475 		goto fail;
3476 	}
3477 
3478 	/*
3479 	 * We should check this at this moment to avoid deadlock on inode page
3480 	 * and #0 page. The locking rule for inline_data conversion should be:
3481 	 * lock_page(page #0) -> lock_page(inode_page)
3482 	 */
3483 	if (index != 0) {
3484 		err = f2fs_convert_inline_inode(inode);
3485 		if (err)
3486 			goto fail;
3487 	}
3488 
3489 #ifdef CONFIG_F2FS_FS_COMPRESSION
3490 	if (f2fs_compressed_file(inode)) {
3491 		int ret;
3492 
3493 		*fsdata = NULL;
3494 
3495 		if (len == PAGE_SIZE && !(f2fs_is_atomic_file(inode)))
3496 			goto repeat;
3497 
3498 		ret = f2fs_prepare_compress_overwrite(inode, pagep,
3499 							index, fsdata);
3500 		if (ret < 0) {
3501 			err = ret;
3502 			goto fail;
3503 		} else if (ret) {
3504 			return 0;
3505 		}
3506 	}
3507 #endif
3508 
3509 repeat:
3510 	/*
3511 	 * Do not use grab_cache_page_write_begin() to avoid deadlock due to
3512 	 * wait_for_stable_page. Will wait that below with our IO control.
3513 	 */
3514 	page = f2fs_pagecache_get_page(mapping, index,
3515 				FGP_LOCK | FGP_WRITE | FGP_CREAT, GFP_NOFS);
3516 	if (!page) {
3517 		err = -ENOMEM;
3518 		goto fail;
3519 	}
3520 
3521 	/* TODO: cluster can be compressed due to race with .writepage */
3522 
3523 	*pagep = page;
3524 
3525 	if (f2fs_is_atomic_file(inode))
3526 		err = prepare_atomic_write_begin(sbi, page, pos, len,
3527 					&blkaddr, &need_balance);
3528 	else
3529 		err = prepare_write_begin(sbi, page, pos, len,
3530 					&blkaddr, &need_balance);
3531 	if (err)
3532 		goto fail;
3533 
3534 	if (need_balance && !IS_NOQUOTA(inode) &&
3535 			has_not_enough_free_secs(sbi, 0, 0)) {
3536 		unlock_page(page);
3537 		f2fs_balance_fs(sbi, true);
3538 		lock_page(page);
3539 		if (page->mapping != mapping) {
3540 			/* The page got truncated from under us */
3541 			f2fs_put_page(page, 1);
3542 			goto repeat;
3543 		}
3544 	}
3545 
3546 	f2fs_wait_on_page_writeback(page, DATA, false, true);
3547 
3548 	if (len == PAGE_SIZE || PageUptodate(page))
3549 		return 0;
3550 
3551 	if (!(pos & (PAGE_SIZE - 1)) && (pos + len) >= i_size_read(inode) &&
3552 	    !f2fs_verity_in_progress(inode)) {
3553 		zero_user_segment(page, len, PAGE_SIZE);
3554 		return 0;
3555 	}
3556 
3557 	if (blkaddr == NEW_ADDR) {
3558 		zero_user_segment(page, 0, PAGE_SIZE);
3559 		SetPageUptodate(page);
3560 	} else {
3561 		if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3562 				DATA_GENERIC_ENHANCE_READ)) {
3563 			err = -EFSCORRUPTED;
3564 			goto fail;
3565 		}
3566 		err = f2fs_submit_page_read(inode, page, blkaddr, 0, true);
3567 		if (err)
3568 			goto fail;
3569 
3570 		lock_page(page);
3571 		if (unlikely(page->mapping != mapping)) {
3572 			f2fs_put_page(page, 1);
3573 			goto repeat;
3574 		}
3575 		if (unlikely(!PageUptodate(page))) {
3576 			err = -EIO;
3577 			goto fail;
3578 		}
3579 	}
3580 	return 0;
3581 
3582 fail:
3583 	f2fs_put_page(page, 1);
3584 	f2fs_write_failed(inode, pos + len);
3585 	return err;
3586 }
3587 
3588 static int f2fs_write_end(struct file *file,
3589 			struct address_space *mapping,
3590 			loff_t pos, unsigned len, unsigned copied,
3591 			struct page *page, void *fsdata)
3592 {
3593 	struct inode *inode = page->mapping->host;
3594 
3595 	trace_f2fs_write_end(inode, pos, len, copied);
3596 
3597 	/*
3598 	 * This should be come from len == PAGE_SIZE, and we expect copied
3599 	 * should be PAGE_SIZE. Otherwise, we treat it with zero copied and
3600 	 * let generic_perform_write() try to copy data again through copied=0.
3601 	 */
3602 	if (!PageUptodate(page)) {
3603 		if (unlikely(copied != len))
3604 			copied = 0;
3605 		else
3606 			SetPageUptodate(page);
3607 	}
3608 
3609 #ifdef CONFIG_F2FS_FS_COMPRESSION
3610 	/* overwrite compressed file */
3611 	if (f2fs_compressed_file(inode) && fsdata) {
3612 		f2fs_compress_write_end(inode, fsdata, page->index, copied);
3613 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3614 
3615 		if (pos + copied > i_size_read(inode) &&
3616 				!f2fs_verity_in_progress(inode))
3617 			f2fs_i_size_write(inode, pos + copied);
3618 		return copied;
3619 	}
3620 #endif
3621 
3622 	if (!copied)
3623 		goto unlock_out;
3624 
3625 	set_page_dirty(page);
3626 
3627 	if (pos + copied > i_size_read(inode) &&
3628 	    !f2fs_verity_in_progress(inode)) {
3629 		f2fs_i_size_write(inode, pos + copied);
3630 		if (f2fs_is_atomic_file(inode))
3631 			f2fs_i_size_write(F2FS_I(inode)->cow_inode,
3632 					pos + copied);
3633 	}
3634 unlock_out:
3635 	f2fs_put_page(page, 1);
3636 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3637 	return copied;
3638 }
3639 
3640 void f2fs_invalidate_folio(struct folio *folio, size_t offset, size_t length)
3641 {
3642 	struct inode *inode = folio->mapping->host;
3643 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3644 
3645 	if (inode->i_ino >= F2FS_ROOT_INO(sbi) &&
3646 				(offset || length != folio_size(folio)))
3647 		return;
3648 
3649 	if (folio_test_dirty(folio)) {
3650 		if (inode->i_ino == F2FS_META_INO(sbi)) {
3651 			dec_page_count(sbi, F2FS_DIRTY_META);
3652 		} else if (inode->i_ino == F2FS_NODE_INO(sbi)) {
3653 			dec_page_count(sbi, F2FS_DIRTY_NODES);
3654 		} else {
3655 			inode_dec_dirty_pages(inode);
3656 			f2fs_remove_dirty_inode(inode);
3657 		}
3658 	}
3659 
3660 	clear_page_private_gcing(&folio->page);
3661 
3662 	if (test_opt(sbi, COMPRESS_CACHE) &&
3663 			inode->i_ino == F2FS_COMPRESS_INO(sbi))
3664 		clear_page_private_data(&folio->page);
3665 
3666 	folio_detach_private(folio);
3667 }
3668 
3669 bool f2fs_release_folio(struct folio *folio, gfp_t wait)
3670 {
3671 	struct f2fs_sb_info *sbi;
3672 
3673 	/* If this is dirty folio, keep private data */
3674 	if (folio_test_dirty(folio))
3675 		return false;
3676 
3677 	sbi = F2FS_M_SB(folio->mapping);
3678 	if (test_opt(sbi, COMPRESS_CACHE)) {
3679 		struct inode *inode = folio->mapping->host;
3680 
3681 		if (inode->i_ino == F2FS_COMPRESS_INO(sbi))
3682 			clear_page_private_data(&folio->page);
3683 	}
3684 
3685 	clear_page_private_gcing(&folio->page);
3686 
3687 	folio_detach_private(folio);
3688 	return true;
3689 }
3690 
3691 static bool f2fs_dirty_data_folio(struct address_space *mapping,
3692 		struct folio *folio)
3693 {
3694 	struct inode *inode = mapping->host;
3695 
3696 	trace_f2fs_set_page_dirty(&folio->page, DATA);
3697 
3698 	if (!folio_test_uptodate(folio))
3699 		folio_mark_uptodate(folio);
3700 	BUG_ON(folio_test_swapcache(folio));
3701 
3702 	if (!folio_test_dirty(folio)) {
3703 		filemap_dirty_folio(mapping, folio);
3704 		f2fs_update_dirty_folio(inode, folio);
3705 		return true;
3706 	}
3707 	return false;
3708 }
3709 
3710 
3711 static sector_t f2fs_bmap_compress(struct inode *inode, sector_t block)
3712 {
3713 #ifdef CONFIG_F2FS_FS_COMPRESSION
3714 	struct dnode_of_data dn;
3715 	sector_t start_idx, blknr = 0;
3716 	int ret;
3717 
3718 	start_idx = round_down(block, F2FS_I(inode)->i_cluster_size);
3719 
3720 	set_new_dnode(&dn, inode, NULL, NULL, 0);
3721 	ret = f2fs_get_dnode_of_data(&dn, start_idx, LOOKUP_NODE);
3722 	if (ret)
3723 		return 0;
3724 
3725 	if (dn.data_blkaddr != COMPRESS_ADDR) {
3726 		dn.ofs_in_node += block - start_idx;
3727 		blknr = f2fs_data_blkaddr(&dn);
3728 		if (!__is_valid_data_blkaddr(blknr))
3729 			blknr = 0;
3730 	}
3731 
3732 	f2fs_put_dnode(&dn);
3733 	return blknr;
3734 #else
3735 	return 0;
3736 #endif
3737 }
3738 
3739 
3740 static sector_t f2fs_bmap(struct address_space *mapping, sector_t block)
3741 {
3742 	struct inode *inode = mapping->host;
3743 	sector_t blknr = 0;
3744 
3745 	if (f2fs_has_inline_data(inode))
3746 		goto out;
3747 
3748 	/* make sure allocating whole blocks */
3749 	if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
3750 		filemap_write_and_wait(mapping);
3751 
3752 	/* Block number less than F2FS MAX BLOCKS */
3753 	if (unlikely(block >= max_file_blocks(inode)))
3754 		goto out;
3755 
3756 	if (f2fs_compressed_file(inode)) {
3757 		blknr = f2fs_bmap_compress(inode, block);
3758 	} else {
3759 		struct f2fs_map_blocks map;
3760 
3761 		memset(&map, 0, sizeof(map));
3762 		map.m_lblk = block;
3763 		map.m_len = 1;
3764 		map.m_next_pgofs = NULL;
3765 		map.m_seg_type = NO_CHECK_TYPE;
3766 
3767 		if (!f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_BMAP))
3768 			blknr = map.m_pblk;
3769 	}
3770 out:
3771 	trace_f2fs_bmap(inode, block, blknr);
3772 	return blknr;
3773 }
3774 
3775 #ifdef CONFIG_SWAP
3776 static int f2fs_migrate_blocks(struct inode *inode, block_t start_blk,
3777 							unsigned int blkcnt)
3778 {
3779 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3780 	unsigned int blkofs;
3781 	unsigned int blk_per_sec = BLKS_PER_SEC(sbi);
3782 	unsigned int secidx = start_blk / blk_per_sec;
3783 	unsigned int end_sec = secidx + blkcnt / blk_per_sec;
3784 	int ret = 0;
3785 
3786 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3787 	filemap_invalidate_lock(inode->i_mapping);
3788 
3789 	set_inode_flag(inode, FI_ALIGNED_WRITE);
3790 	set_inode_flag(inode, FI_OPU_WRITE);
3791 
3792 	for (; secidx < end_sec; secidx++) {
3793 		f2fs_down_write(&sbi->pin_sem);
3794 
3795 		f2fs_lock_op(sbi);
3796 		f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
3797 		f2fs_unlock_op(sbi);
3798 
3799 		set_inode_flag(inode, FI_SKIP_WRITES);
3800 
3801 		for (blkofs = 0; blkofs < blk_per_sec; blkofs++) {
3802 			struct page *page;
3803 			unsigned int blkidx = secidx * blk_per_sec + blkofs;
3804 
3805 			page = f2fs_get_lock_data_page(inode, blkidx, true);
3806 			if (IS_ERR(page)) {
3807 				f2fs_up_write(&sbi->pin_sem);
3808 				ret = PTR_ERR(page);
3809 				goto done;
3810 			}
3811 
3812 			set_page_dirty(page);
3813 			f2fs_put_page(page, 1);
3814 		}
3815 
3816 		clear_inode_flag(inode, FI_SKIP_WRITES);
3817 
3818 		ret = filemap_fdatawrite(inode->i_mapping);
3819 
3820 		f2fs_up_write(&sbi->pin_sem);
3821 
3822 		if (ret)
3823 			break;
3824 	}
3825 
3826 done:
3827 	clear_inode_flag(inode, FI_SKIP_WRITES);
3828 	clear_inode_flag(inode, FI_OPU_WRITE);
3829 	clear_inode_flag(inode, FI_ALIGNED_WRITE);
3830 
3831 	filemap_invalidate_unlock(inode->i_mapping);
3832 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3833 
3834 	return ret;
3835 }
3836 
3837 static int check_swap_activate(struct swap_info_struct *sis,
3838 				struct file *swap_file, sector_t *span)
3839 {
3840 	struct address_space *mapping = swap_file->f_mapping;
3841 	struct inode *inode = mapping->host;
3842 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3843 	sector_t cur_lblock;
3844 	sector_t last_lblock;
3845 	sector_t pblock;
3846 	sector_t lowest_pblock = -1;
3847 	sector_t highest_pblock = 0;
3848 	int nr_extents = 0;
3849 	unsigned long nr_pblocks;
3850 	unsigned int blks_per_sec = BLKS_PER_SEC(sbi);
3851 	unsigned int sec_blks_mask = BLKS_PER_SEC(sbi) - 1;
3852 	unsigned int not_aligned = 0;
3853 	int ret = 0;
3854 
3855 	/*
3856 	 * Map all the blocks into the extent list.  This code doesn't try
3857 	 * to be very smart.
3858 	 */
3859 	cur_lblock = 0;
3860 	last_lblock = bytes_to_blks(inode, i_size_read(inode));
3861 
3862 	while (cur_lblock < last_lblock && cur_lblock < sis->max) {
3863 		struct f2fs_map_blocks map;
3864 retry:
3865 		cond_resched();
3866 
3867 		memset(&map, 0, sizeof(map));
3868 		map.m_lblk = cur_lblock;
3869 		map.m_len = last_lblock - cur_lblock;
3870 		map.m_next_pgofs = NULL;
3871 		map.m_next_extent = NULL;
3872 		map.m_seg_type = NO_CHECK_TYPE;
3873 		map.m_may_create = false;
3874 
3875 		ret = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_FIEMAP);
3876 		if (ret)
3877 			goto out;
3878 
3879 		/* hole */
3880 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
3881 			f2fs_err(sbi, "Swapfile has holes");
3882 			ret = -EINVAL;
3883 			goto out;
3884 		}
3885 
3886 		pblock = map.m_pblk;
3887 		nr_pblocks = map.m_len;
3888 
3889 		if ((pblock - SM_I(sbi)->main_blkaddr) & sec_blks_mask ||
3890 				nr_pblocks & sec_blks_mask) {
3891 			not_aligned++;
3892 
3893 			nr_pblocks = roundup(nr_pblocks, blks_per_sec);
3894 			if (cur_lblock + nr_pblocks > sis->max)
3895 				nr_pblocks -= blks_per_sec;
3896 
3897 			if (!nr_pblocks) {
3898 				/* this extent is last one */
3899 				nr_pblocks = map.m_len;
3900 				f2fs_warn(sbi, "Swapfile: last extent is not aligned to section");
3901 				goto next;
3902 			}
3903 
3904 			ret = f2fs_migrate_blocks(inode, cur_lblock,
3905 							nr_pblocks);
3906 			if (ret)
3907 				goto out;
3908 			goto retry;
3909 		}
3910 next:
3911 		if (cur_lblock + nr_pblocks >= sis->max)
3912 			nr_pblocks = sis->max - cur_lblock;
3913 
3914 		if (cur_lblock) {	/* exclude the header page */
3915 			if (pblock < lowest_pblock)
3916 				lowest_pblock = pblock;
3917 			if (pblock + nr_pblocks - 1 > highest_pblock)
3918 				highest_pblock = pblock + nr_pblocks - 1;
3919 		}
3920 
3921 		/*
3922 		 * We found a PAGE_SIZE-length, PAGE_SIZE-aligned run of blocks
3923 		 */
3924 		ret = add_swap_extent(sis, cur_lblock, nr_pblocks, pblock);
3925 		if (ret < 0)
3926 			goto out;
3927 		nr_extents += ret;
3928 		cur_lblock += nr_pblocks;
3929 	}
3930 	ret = nr_extents;
3931 	*span = 1 + highest_pblock - lowest_pblock;
3932 	if (cur_lblock == 0)
3933 		cur_lblock = 1;	/* force Empty message */
3934 	sis->max = cur_lblock;
3935 	sis->pages = cur_lblock - 1;
3936 	sis->highest_bit = cur_lblock - 1;
3937 out:
3938 	if (not_aligned)
3939 		f2fs_warn(sbi, "Swapfile (%u) is not align to section: 1) creat(), 2) ioctl(F2FS_IOC_SET_PIN_FILE), 3) fallocate(%u * N)",
3940 			  not_aligned, blks_per_sec * F2FS_BLKSIZE);
3941 	return ret;
3942 }
3943 
3944 static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
3945 				sector_t *span)
3946 {
3947 	struct inode *inode = file_inode(file);
3948 	int ret;
3949 
3950 	if (!S_ISREG(inode->i_mode))
3951 		return -EINVAL;
3952 
3953 	if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3954 		return -EROFS;
3955 
3956 	if (f2fs_lfs_mode(F2FS_I_SB(inode))) {
3957 		f2fs_err(F2FS_I_SB(inode),
3958 			"Swapfile not supported in LFS mode");
3959 		return -EINVAL;
3960 	}
3961 
3962 	ret = f2fs_convert_inline_inode(inode);
3963 	if (ret)
3964 		return ret;
3965 
3966 	if (!f2fs_disable_compressed_file(inode))
3967 		return -EINVAL;
3968 
3969 	f2fs_precache_extents(inode);
3970 
3971 	ret = check_swap_activate(sis, file, span);
3972 	if (ret < 0)
3973 		return ret;
3974 
3975 	set_inode_flag(inode, FI_PIN_FILE);
3976 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3977 	return ret;
3978 }
3979 
3980 static void f2fs_swap_deactivate(struct file *file)
3981 {
3982 	struct inode *inode = file_inode(file);
3983 
3984 	clear_inode_flag(inode, FI_PIN_FILE);
3985 }
3986 #else
3987 static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
3988 				sector_t *span)
3989 {
3990 	return -EOPNOTSUPP;
3991 }
3992 
3993 static void f2fs_swap_deactivate(struct file *file)
3994 {
3995 }
3996 #endif
3997 
3998 const struct address_space_operations f2fs_dblock_aops = {
3999 	.read_folio	= f2fs_read_data_folio,
4000 	.readahead	= f2fs_readahead,
4001 	.writepage	= f2fs_write_data_page,
4002 	.writepages	= f2fs_write_data_pages,
4003 	.write_begin	= f2fs_write_begin,
4004 	.write_end	= f2fs_write_end,
4005 	.dirty_folio	= f2fs_dirty_data_folio,
4006 	.migrate_folio	= filemap_migrate_folio,
4007 	.invalidate_folio = f2fs_invalidate_folio,
4008 	.release_folio	= f2fs_release_folio,
4009 	.direct_IO	= noop_direct_IO,
4010 	.bmap		= f2fs_bmap,
4011 	.swap_activate  = f2fs_swap_activate,
4012 	.swap_deactivate = f2fs_swap_deactivate,
4013 };
4014 
4015 void f2fs_clear_page_cache_dirty_tag(struct page *page)
4016 {
4017 	struct address_space *mapping = page_mapping(page);
4018 	unsigned long flags;
4019 
4020 	xa_lock_irqsave(&mapping->i_pages, flags);
4021 	__xa_clear_mark(&mapping->i_pages, page_index(page),
4022 						PAGECACHE_TAG_DIRTY);
4023 	xa_unlock_irqrestore(&mapping->i_pages, flags);
4024 }
4025 
4026 int __init f2fs_init_post_read_processing(void)
4027 {
4028 	bio_post_read_ctx_cache =
4029 		kmem_cache_create("f2fs_bio_post_read_ctx",
4030 				  sizeof(struct bio_post_read_ctx), 0, 0, NULL);
4031 	if (!bio_post_read_ctx_cache)
4032 		goto fail;
4033 	bio_post_read_ctx_pool =
4034 		mempool_create_slab_pool(NUM_PREALLOC_POST_READ_CTXS,
4035 					 bio_post_read_ctx_cache);
4036 	if (!bio_post_read_ctx_pool)
4037 		goto fail_free_cache;
4038 	return 0;
4039 
4040 fail_free_cache:
4041 	kmem_cache_destroy(bio_post_read_ctx_cache);
4042 fail:
4043 	return -ENOMEM;
4044 }
4045 
4046 void f2fs_destroy_post_read_processing(void)
4047 {
4048 	mempool_destroy(bio_post_read_ctx_pool);
4049 	kmem_cache_destroy(bio_post_read_ctx_cache);
4050 }
4051 
4052 int f2fs_init_post_read_wq(struct f2fs_sb_info *sbi)
4053 {
4054 	if (!f2fs_sb_has_encrypt(sbi) &&
4055 		!f2fs_sb_has_verity(sbi) &&
4056 		!f2fs_sb_has_compression(sbi))
4057 		return 0;
4058 
4059 	sbi->post_read_wq = alloc_workqueue("f2fs_post_read_wq",
4060 						 WQ_UNBOUND | WQ_HIGHPRI,
4061 						 num_online_cpus());
4062 	if (!sbi->post_read_wq)
4063 		return -ENOMEM;
4064 	return 0;
4065 }
4066 
4067 void f2fs_destroy_post_read_wq(struct f2fs_sb_info *sbi)
4068 {
4069 	if (sbi->post_read_wq)
4070 		destroy_workqueue(sbi->post_read_wq);
4071 }
4072 
4073 int __init f2fs_init_bio_entry_cache(void)
4074 {
4075 	bio_entry_slab = f2fs_kmem_cache_create("f2fs_bio_entry_slab",
4076 			sizeof(struct bio_entry));
4077 	if (!bio_entry_slab)
4078 		return -ENOMEM;
4079 	return 0;
4080 }
4081 
4082 void f2fs_destroy_bio_entry_cache(void)
4083 {
4084 	kmem_cache_destroy(bio_entry_slab);
4085 }
4086 
4087 static int f2fs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
4088 			    unsigned int flags, struct iomap *iomap,
4089 			    struct iomap *srcmap)
4090 {
4091 	struct f2fs_map_blocks map = {};
4092 	pgoff_t next_pgofs = 0;
4093 	int err;
4094 
4095 	map.m_lblk = bytes_to_blks(inode, offset);
4096 	map.m_len = bytes_to_blks(inode, offset + length - 1) - map.m_lblk + 1;
4097 	map.m_next_pgofs = &next_pgofs;
4098 	map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4099 	if (flags & IOMAP_WRITE)
4100 		map.m_may_create = true;
4101 
4102 	err = f2fs_map_blocks(inode, &map, flags & IOMAP_WRITE,
4103 			      F2FS_GET_BLOCK_DIO);
4104 	if (err)
4105 		return err;
4106 
4107 	iomap->offset = blks_to_bytes(inode, map.m_lblk);
4108 
4109 	/*
4110 	 * When inline encryption is enabled, sometimes I/O to an encrypted file
4111 	 * has to be broken up to guarantee DUN contiguity.  Handle this by
4112 	 * limiting the length of the mapping returned.
4113 	 */
4114 	map.m_len = fscrypt_limit_io_blocks(inode, map.m_lblk, map.m_len);
4115 
4116 	if (map.m_flags & (F2FS_MAP_MAPPED | F2FS_MAP_UNWRITTEN)) {
4117 		iomap->length = blks_to_bytes(inode, map.m_len);
4118 		if (map.m_flags & F2FS_MAP_MAPPED) {
4119 			iomap->type = IOMAP_MAPPED;
4120 			iomap->flags |= IOMAP_F_MERGED;
4121 		} else {
4122 			iomap->type = IOMAP_UNWRITTEN;
4123 		}
4124 		if (WARN_ON_ONCE(!__is_valid_data_blkaddr(map.m_pblk)))
4125 			return -EINVAL;
4126 
4127 		iomap->bdev = map.m_bdev;
4128 		iomap->addr = blks_to_bytes(inode, map.m_pblk);
4129 	} else {
4130 		iomap->length = blks_to_bytes(inode, next_pgofs) -
4131 				iomap->offset;
4132 		iomap->type = IOMAP_HOLE;
4133 		iomap->addr = IOMAP_NULL_ADDR;
4134 	}
4135 
4136 	if (map.m_flags & F2FS_MAP_NEW)
4137 		iomap->flags |= IOMAP_F_NEW;
4138 	if ((inode->i_state & I_DIRTY_DATASYNC) ||
4139 	    offset + length > i_size_read(inode))
4140 		iomap->flags |= IOMAP_F_DIRTY;
4141 
4142 	return 0;
4143 }
4144 
4145 const struct iomap_ops f2fs_iomap_ops = {
4146 	.iomap_begin	= f2fs_iomap_begin,
4147 };
4148