]> git.ipfire.org Git - people/ms/linux.git/blame - fs/f2fs/data.c
projects / people / ms / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: introduce DEFAULT_IO_TIMEOUT
[people/ms/linux.git] / fs / f2fs / data.c
CommitLineData
7c1a000d 1// SPDX-License-Identifier: GPL-2.0
0a8165d7 2/*
eb47b800
JK		 3 * fs/f2fs/data.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
eb47b800
JK		 7 */
8#include <linux/fs.h>
9#include <linux/f2fs_fs.h>
10#include <linux/buffer_head.h>
11#include <linux/mpage.h>
12#include <linux/writeback.h>
13#include <linux/backing-dev.h>
8f46dcae 14#include <linux/pagevec.h>
eb47b800
JK		 15#include <linux/blkdev.h>
16#include <linux/bio.h>
4969c06a 17#include <linux/swap.h>
690e4a3e 18#include <linux/prefetch.h>
e2e40f2c 19#include <linux/uio.h>
f1e88660 20#include <linux/cleancache.h>
174cd4b1 21#include <linux/sched/signal.h>
eb47b800
JK		 22
23#include "f2fs.h"
24#include "node.h"
25#include "segment.h"
db9f7c1a 26#include "trace.h"
848753aa 27#include <trace/events/f2fs.h>
eb47b800 28
6dbb1796
EB		 29#define NUM_PREALLOC_POST_READ_CTXS 128
30
31static struct kmem_cache *bio_post_read_ctx_cache;
0b20fcec 32static struct kmem_cache *bio_entry_slab;
6dbb1796 33static mempool_t *bio_post_read_ctx_pool;
f543805f
CY		 34static struct bio_set f2fs_bioset;
35
36#define F2FS_BIO_POOL_SIZE NR_CURSEG_TYPE
37
38int __init f2fs_init_bioset(void)
39{
40 if (bioset_init(&f2fs_bioset, F2FS_BIO_POOL_SIZE,
41 0, BIOSET_NEED_BVECS))
42 return -ENOMEM;
43 return 0;
44}
45
46void f2fs_destroy_bioset(void)
47{
48 bioset_exit(&f2fs_bioset);
49}
50
51static inline struct bio *__f2fs_bio_alloc(gfp_t gfp_mask,
52 unsigned int nr_iovecs)
53{
54 return bio_alloc_bioset(gfp_mask, nr_iovecs, &f2fs_bioset);
55}
56
57struct bio *f2fs_bio_alloc(struct f2fs_sb_info *sbi, int npages, bool no_fail)
58{
59 struct bio *bio;
60
61 if (no_fail) {
62 /* No failure on bio allocation */
63 bio = __f2fs_bio_alloc(GFP_NOIO, npages);
64 if (!bio)
65 bio = __f2fs_bio_alloc(GFP_NOIO | __GFP_NOFAIL, npages);
66 return bio;
67 }
68 if (time_to_inject(sbi, FAULT_ALLOC_BIO)) {
69 f2fs_show_injection_info(sbi, FAULT_ALLOC_BIO);
70 return NULL;
71 }
72
73 return __f2fs_bio_alloc(GFP_KERNEL, npages);
74}
6dbb1796 75
36951b38
CY		 76static bool __is_cp_guaranteed(struct page *page)
77{
78 struct address_space *mapping = page->mapping;
79 struct inode *inode;
80 struct f2fs_sb_info *sbi;
81
82 if (!mapping)
83 return false;
84
4c8ff709
CY		 85 if (f2fs_is_compressed_page(page))
86 return false;
87
36951b38
CY		 88 inode = mapping->host;
89 sbi = F2FS_I_SB(inode);
90
91 if (inode->i_ino == F2FS_META_INO(sbi) ||
92 inode->i_ino == F2FS_NODE_INO(sbi) ||
93 S_ISDIR(inode->i_mode) ||
e7a4feb0 94 (S_ISREG(inode->i_mode) &&
af033b2a 95 (f2fs_is_atomic_file(inode) || IS_NOQUOTA(inode))) ||
36951b38
CY		 96 is_cold_data(page))
97 return true;
98 return false;
99}
100
5f9abab4
JK		 101static enum count_type __read_io_type(struct page *page)
102{
4969c06a 103 struct address_space *mapping = page_file_mapping(page);
5f9abab4
JK		 104
105 if (mapping) {
106 struct inode *inode = mapping->host;
107 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
108
109 if (inode->i_ino == F2FS_META_INO(sbi))
110 return F2FS_RD_META;
111
112 if (inode->i_ino == F2FS_NODE_INO(sbi))
113 return F2FS_RD_NODE;
114 }
115 return F2FS_RD_DATA;
116}
117
6dbb1796
EB		 118/* postprocessing steps for read bios */
119enum bio_post_read_step {
6dbb1796 120 STEP_DECRYPT,
4c8ff709 121 STEP_DECOMPRESS,
95ae251f 122 STEP_VERITY,
6dbb1796
EB		 123};
124
125struct bio_post_read_ctx {
126 struct bio *bio;
4c8ff709 127 struct f2fs_sb_info *sbi;
6dbb1796 128 struct work_struct work;
6dbb1796
EB		 129 unsigned int enabled_steps;
130};
131
4c8ff709 132static void __read_end_io(struct bio *bio, bool compr, bool verity)
93dfe2ac 133{
6dbb1796
EB		 134 struct page *page;
135 struct bio_vec *bv;
6dc4f100 136 struct bvec_iter_all iter_all;
93dfe2ac 137
2b070cfe 138 bio_for_each_segment_all(bv, bio, iter_all) {
6dbb1796
EB		 139 page = bv->bv_page;
140
4c8ff709
CY		 141#ifdef CONFIG_F2FS_FS_COMPRESSION
142 if (compr && f2fs_is_compressed_page(page)) {
143 f2fs_decompress_pages(bio, page, verity);
144 continue;
145 }
146#endif
147
6dbb1796
EB		 148 /* PG_error was set if any post_read step failed */
149 if (bio->bi_status || PageError(page)) {
150 ClearPageUptodate(page);
fb7d70db
JK		 151 /* will re-read again later */
152 ClearPageError(page);
6dbb1796
EB		 153 } else {
154 SetPageUptodate(page);
155 }
5f9abab4 156 dec_page_count(F2FS_P_SB(page), __read_io_type(page));
6dbb1796
EB		 157 unlock_page(page);
158 }
4c8ff709
CY		 159}
160
161static void f2fs_release_read_bio(struct bio *bio);
162static void __f2fs_read_end_io(struct bio *bio, bool compr, bool verity)
163{
164 if (!compr)
165 __read_end_io(bio, false, verity);
166 f2fs_release_read_bio(bio);
167}
168
169static void f2fs_decompress_bio(struct bio *bio, bool verity)
170{
171 __read_end_io(bio, true, verity);
6dbb1796
EB		 172}
173
174static void bio_post_read_processing(struct bio_post_read_ctx *ctx);
175
4c8ff709
CY		 176static void f2fs_decrypt_work(struct bio_post_read_ctx *ctx)
177{
178 fscrypt_decrypt_bio(ctx->bio);
179}
180
181static void f2fs_decompress_work(struct bio_post_read_ctx *ctx)
182{
183 f2fs_decompress_bio(ctx->bio, ctx->enabled_steps & (1 << STEP_VERITY));
184}
185
186#ifdef CONFIG_F2FS_FS_COMPRESSION
187static void f2fs_verify_pages(struct page **rpages, unsigned int cluster_size)
188{
189 f2fs_decompress_end_io(rpages, cluster_size, false, true);
190}
191
192static void f2fs_verify_bio(struct bio *bio)
193{
194 struct page *page = bio_first_page_all(bio);
195 struct decompress_io_ctx *dic =
196 (struct decompress_io_ctx *)page_private(page);
197
198 f2fs_verify_pages(dic->rpages, dic->cluster_size);
199 f2fs_free_dic(dic);
200}
201#endif
202
203static void f2fs_verity_work(struct work_struct *work)
6dbb1796
EB		 204{
205 struct bio_post_read_ctx *ctx =
206 container_of(work, struct bio_post_read_ctx, work);
644c8c92
EB		 207 struct bio *bio = ctx->bio;
208#ifdef CONFIG_F2FS_FS_COMPRESSION
209 unsigned int enabled_steps = ctx->enabled_steps;
210#endif
211
212 /*
213 * fsverity_verify_bio() may call readpages() again, and while verity
214 * will be disabled for this, decryption may still be needed, resulting
215 * in another bio_post_read_ctx being allocated. So to prevent
216 * deadlocks we need to release the current ctx to the mempool first.
217 * This assumes that verity is the last post-read step.
218 */
219 mempool_free(ctx, bio_post_read_ctx_pool);
220 bio->bi_private = NULL;
6dbb1796 221
4c8ff709
CY		 222#ifdef CONFIG_F2FS_FS_COMPRESSION
223 /* previous step is decompression */
644c8c92
EB		 224 if (enabled_steps & (1 << STEP_DECOMPRESS)) {
225 f2fs_verify_bio(bio);
226 f2fs_release_read_bio(bio);
4c8ff709
CY		 227 return;
228 }
229#endif
6dbb1796 230
644c8c92
EB		 231 fsverity_verify_bio(bio);
232 __f2fs_read_end_io(bio, false, false);
6dbb1796
EB		 233}
234
4c8ff709 235static void f2fs_post_read_work(struct work_struct *work)
95ae251f
EB		 236{
237 struct bio_post_read_ctx *ctx =
238 container_of(work, struct bio_post_read_ctx, work);
239
4c8ff709
CY		 240 if (ctx->enabled_steps & (1 << STEP_DECRYPT))
241 f2fs_decrypt_work(ctx);
242
243 if (ctx->enabled_steps & (1 << STEP_DECOMPRESS))
244 f2fs_decompress_work(ctx);
245
246 if (ctx->enabled_steps & (1 << STEP_VERITY)) {
247 INIT_WORK(&ctx->work, f2fs_verity_work);
248 fsverity_enqueue_verify_work(&ctx->work);
249 return;
250 }
251
252 __f2fs_read_end_io(ctx->bio,
253 ctx->enabled_steps & (1 << STEP_DECOMPRESS), false);
254}
95ae251f 255
4c8ff709
CY		 256static void f2fs_enqueue_post_read_work(struct f2fs_sb_info *sbi,
257 struct work_struct *work)
258{
259 queue_work(sbi->post_read_wq, work);
95ae251f
EB		 260}
261
6dbb1796
EB		 262static void bio_post_read_processing(struct bio_post_read_ctx *ctx)
263{
95ae251f
EB		 264 /*
265 * We use different work queues for decryption and for verity because
266 * verity may require reading metadata pages that need decryption, and
267 * we shouldn't recurse to the same workqueue.
268 */
4c8ff709
CY		 269
270 if (ctx->enabled_steps & (1 << STEP_DECRYPT) ||
271 ctx->enabled_steps & (1 << STEP_DECOMPRESS)) {
272 INIT_WORK(&ctx->work, f2fs_post_read_work);
273 f2fs_enqueue_post_read_work(ctx->sbi, &ctx->work);
274 return;
275 }
276
277 if (ctx->enabled_steps & (1 << STEP_VERITY)) {
278 INIT_WORK(&ctx->work, f2fs_verity_work);
279 fsverity_enqueue_verify_work(&ctx->work);
280 return;
6dbb1796 281 }
4c8ff709
CY		 282
283 __f2fs_read_end_io(ctx->bio, false, false);
6dbb1796
EB		 284}
285
286static bool f2fs_bio_post_read_required(struct bio *bio)
287{
4c8ff709 288 return bio->bi_private;
6dbb1796
EB		 289}
290
291static void f2fs_read_end_io(struct bio *bio)
292{
c45d6002
CY		 293 struct f2fs_sb_info *sbi = F2FS_P_SB(bio_first_page_all(bio));
294
295 if (time_to_inject(sbi, FAULT_READ_IO)) {
296 f2fs_show_injection_info(sbi, FAULT_READ_IO);
4e4cbee9 297 bio->bi_status = BLK_STS_IOERR;
55523519 298 }
8b038c70 299
6dbb1796
EB		 300 if (f2fs_bio_post_read_required(bio)) {
301 struct bio_post_read_ctx *ctx = bio->bi_private;
f1e88660 302
6dbb1796
EB		 303 bio_post_read_processing(ctx);
304 return;
f1e88660 305 }
6dbb1796 306
4c8ff709 307 __f2fs_read_end_io(bio, false, false);
f1e88660
JK		 308}
309
4246a0b6 310static void f2fs_write_end_io(struct bio *bio)
93dfe2ac 311{
1b1f559f 312 struct f2fs_sb_info *sbi = bio->bi_private;
f568849e 313 struct bio_vec *bvec;
6dc4f100 314 struct bvec_iter_all iter_all;
93dfe2ac 315
6f5c2ed0 316 if (time_to_inject(sbi, FAULT_WRITE_IO)) {
c45d6002 317 f2fs_show_injection_info(sbi, FAULT_WRITE_IO);
6f5c2ed0
CY		 318 bio->bi_status = BLK_STS_IOERR;
319 }
320
2b070cfe 321 bio_for_each_segment_all(bvec, bio, iter_all) {
93dfe2ac 322 struct page *page = bvec->bv_page;
36951b38 323 enum count_type type = WB_DATA_TYPE(page);
93dfe2ac 324
0a595eba
JK		 325 if (IS_DUMMY_WRITTEN_PAGE(page)) {
326 set_page_private(page, (unsigned long)NULL);
327 ClearPagePrivate(page);
328 unlock_page(page);
329 mempool_free(page, sbi->write_io_dummy);
330
4e4cbee9 331 if (unlikely(bio->bi_status))
0a595eba
JK		 332 f2fs_stop_checkpoint(sbi, true);
333 continue;
334 }
335
d2d0727b 336 fscrypt_finalize_bounce_page(&page);
4375a336 337
4c8ff709
CY		 338#ifdef CONFIG_F2FS_FS_COMPRESSION
339 if (f2fs_is_compressed_page(page)) {
340 f2fs_compress_write_end_io(bio, page);
341 continue;
342 }
343#endif
344
4e4cbee9 345 if (unlikely(bio->bi_status)) {
5114a97a 346 mapping_set_error(page->mapping, -EIO);
b1ca321d
JK		 347 if (type == F2FS_WB_CP_DATA)
348 f2fs_stop_checkpoint(sbi, true);
93dfe2ac 349 }
7dff55d2
YH		 350
351 f2fs_bug_on(sbi, page->mapping == NODE_MAPPING(sbi) &&
352 page->index != nid_of_node(page));
353
36951b38 354 dec_page_count(sbi, type);
50fa53ec
CY		 355 if (f2fs_in_warm_node_list(sbi, page))
356 f2fs_del_fsync_node_entry(sbi, page);
36951b38 357 clear_cold_data(page);
93dfe2ac 358 end_page_writeback(page);
f568849e 359 }
36951b38 360 if (!get_pages(sbi, F2FS_WB_CP_DATA) &&
f5730184 361 wq_has_sleeper(&sbi->cp_wait))
93dfe2ac
JK		 362 wake_up(&sbi->cp_wait);
363
364 bio_put(bio);
365}
366
3c62be17
JK		 367struct block_device *f2fs_target_device(struct f2fs_sb_info *sbi,
368 block_t blk_addr, struct bio *bio)
369{
370 struct block_device *bdev = sbi->sb->s_bdev;
371 int i;
372
0916878d
DLM		 373 if (f2fs_is_multi_device(sbi)) {
374 for (i = 0; i < sbi->s_ndevs; i++) {
375 if (FDEV(i).start_blk <= blk_addr &&
376 FDEV(i).end_blk >= blk_addr) {
377 blk_addr -= FDEV(i).start_blk;
378 bdev = FDEV(i).bdev;
379 break;
380 }
3c62be17
JK		 381 }
382 }
383 if (bio) {
74d46992 384 bio_set_dev(bio, bdev);
3c62be17
JK		 385 bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr);
386 }
387 return bdev;
388}
389
390int f2fs_target_device_index(struct f2fs_sb_info *sbi, block_t blkaddr)
391{
392 int i;
393
0916878d
DLM		 394 if (!f2fs_is_multi_device(sbi))
395 return 0;
396
3c62be17
JK		 397 for (i = 0; i < sbi->s_ndevs; i++)
398 if (FDEV(i).start_blk <= blkaddr && FDEV(i).end_blk >= blkaddr)
399 return i;
400 return 0;
401}
402
7a88ddb5
CY		 403/*
404 * Return true, if pre_bio's bdev is same as its target device.
405 */
3c62be17
JK		 406static bool __same_bdev(struct f2fs_sb_info *sbi,
407 block_t blk_addr, struct bio *bio)
408{
74d46992
CH		 409 struct block_device *b = f2fs_target_device(sbi, blk_addr, NULL);
410 return bio->bi_disk == b->bd_disk && bio->bi_partno == b->bd_partno;
3c62be17
JK		 411}
412
b757f6ed 413static struct bio *__bio_alloc(struct f2fs_io_info *fio, int npages)
940a6d34 414{
b757f6ed 415 struct f2fs_sb_info *sbi = fio->sbi;
940a6d34
GZ		 416 struct bio *bio;
417
d62fe971 418 bio = f2fs_bio_alloc(sbi, npages, true);
940a6d34 419
b757f6ed
CY		 420 f2fs_target_device(sbi, fio->new_blkaddr, bio);
421 if (is_read_io(fio->op)) {
0cdd3195
HL		 422 bio->bi_end_io = f2fs_read_end_io;
423 bio->bi_private = NULL;
424 } else {
425 bio->bi_end_io = f2fs_write_end_io;
426 bio->bi_private = sbi;
b757f6ed
CY		 427 bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi,
428 fio->type, fio->temp);
0cdd3195 429 }
b757f6ed
CY		 430 if (fio->io_wbc)
431 wbc_init_bio(fio->io_wbc, bio);
940a6d34
GZ		 432
433 return bio;
434}
435
4fc29c1a
LT		 436static inline void __submit_bio(struct f2fs_sb_info *sbi,
437 struct bio *bio, enum page_type type)
f5730184 438{
4fc29c1a 439 if (!is_read_io(bio_op(bio))) {
0a595eba
JK		 440 unsigned int start;
441
0a595eba
JK		 442 if (type != DATA && type != NODE)
443 goto submit_io;
444
b0332a0f 445 if (f2fs_lfs_mode(sbi) && current->plug)
3bb09a0e
TY		 446 blk_finish_plug(current->plug);
447
8223ecc4
CY		 448 if (F2FS_IO_ALIGNED(sbi))
449 goto submit_io;
450
0a595eba
JK		 451 start = bio->bi_iter.bi_size >> F2FS_BLKSIZE_BITS;
452 start %= F2FS_IO_SIZE(sbi);
453
454 if (start == 0)
455 goto submit_io;
456
457 /* fill dummy pages */
458 for (; start < F2FS_IO_SIZE(sbi); start++) {
459 struct page *page =
460 mempool_alloc(sbi->write_io_dummy,
bc73a4b2 461 GFP_NOIO | __GFP_NOFAIL);
0a595eba
JK		 462 f2fs_bug_on(sbi, !page);
463
bc73a4b2 464 zero_user_segment(page, 0, PAGE_SIZE);
0a595eba
JK		 465 SetPagePrivate(page);
466 set_page_private(page, (unsigned long)DUMMY_WRITTEN_PAGE);
467 lock_page(page);
468 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE)
469 f2fs_bug_on(sbi, 1);
470 }
471 /*
472 * In the NODE case, we lose next block address chain. So, we
473 * need to do checkpoint in f2fs_sync_file.
474 */
475 if (type == NODE)
476 set_sbi_flag(sbi, SBI_NEED_CP);
19a5f5e2 477 }
0a595eba 478submit_io:
554b5125
JK		 479 if (is_read_io(bio_op(bio)))
480 trace_f2fs_submit_read_bio(sbi->sb, type, bio);
481 else
482 trace_f2fs_submit_write_bio(sbi->sb, type, bio);
4e49ea4a 483 submit_bio(bio);
f5730184
JK		 484}
485
4c8ff709
CY		 486void f2fs_submit_bio(struct f2fs_sb_info *sbi,
487 struct bio *bio, enum page_type type)
488{
489 __submit_bio(sbi, bio, type);
490}
491
458e6197 492static void __submit_merged_bio(struct f2fs_bio_info *io)
93dfe2ac 493{
458e6197 494 struct f2fs_io_info *fio = &io->fio;
93dfe2ac
JK		 495
496 if (!io->bio)
497 return;
498
554b5125
JK		 499 bio_set_op_attrs(io->bio, fio->op, fio->op_flags);
500
04d328de 501 if (is_read_io(fio->op))
554b5125 502 trace_f2fs_prepare_read_bio(io->sbi->sb, fio->type, io->bio);
6a8f8ca5 503 else
554b5125 504 trace_f2fs_prepare_write_bio(io->sbi->sb, fio->type, io->bio);
04d328de 505
4fc29c1a 506 __submit_bio(io->sbi, io->bio, fio->type);
93dfe2ac
JK		 507 io->bio = NULL;
508}
509
8648de2c 510static bool __has_merged_page(struct bio *bio, struct inode *inode,
bab475c5 511 struct page *page, nid_t ino)
0fd785eb 512{
0fd785eb 513 struct bio_vec *bvec;
6dc4f100 514 struct bvec_iter_all iter_all;
0fd785eb 515
8648de2c 516 if (!bio)
0fd785eb 517 return false;
0c3a5797 518
bab475c5 519 if (!inode && !page && !ino)
0c3a5797 520 return true;
0fd785eb 521
8648de2c 522 bio_for_each_segment_all(bvec, bio, iter_all) {
4c8ff709 523 struct page *target = bvec->bv_page;
0fd785eb 524
4c8ff709 525 if (fscrypt_is_bounce_page(target)) {
d2d0727b 526 target = fscrypt_pagecache_page(target);
4c8ff709
CY		 527 if (IS_ERR(target))
528 continue;
529 }
530 if (f2fs_is_compressed_page(target)) {
531 target = f2fs_compress_control_page(target);
532 if (IS_ERR(target))
533 continue;
534 }
0fd785eb 535
0c3a5797
CY		 536 if (inode && inode == target->mapping->host)
537 return true;
bab475c5
CY		 538 if (page && page == target)
539 return true;
0c3a5797 540 if (ino && ino == ino_of_node(target))
0fd785eb 541 return true;
0fd785eb
CY		 542 }
543
0fd785eb
CY		 544 return false;
545}
546
b9109b0e 547static void __f2fs_submit_merged_write(struct f2fs_sb_info *sbi,
a912b54d 548 enum page_type type, enum temp_type temp)
93dfe2ac
JK		 549{
550 enum page_type btype = PAGE_TYPE_OF_BIO(type);
a912b54d 551 struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
93dfe2ac 552
df0f8dc0 553 down_write(&io->io_rwsem);
458e6197
JK		 554
555 /* change META to META_FLUSH in the checkpoint procedure */
556 if (type >= META_FLUSH) {
557 io->fio.type = META_FLUSH;
04d328de 558 io->fio.op = REQ_OP_WRITE;
3adc5fcb 559 io->fio.op_flags = REQ_META | REQ_PRIO | REQ_SYNC;
70fd7614 560 if (!test_opt(sbi, NOBARRIER))
7f54f51f 561 io->fio.op_flags |= REQ_PREFLUSH | REQ_FUA;
458e6197
JK		 562 }
563 __submit_merged_bio(io);
df0f8dc0 564 up_write(&io->io_rwsem);
93dfe2ac
JK		 565}
566
a912b54d 567static void __submit_merged_write_cond(struct f2fs_sb_info *sbi,
bab475c5
CY		 568 struct inode *inode, struct page *page,
569 nid_t ino, enum page_type type, bool force)
0c3a5797 570{
a912b54d 571 enum temp_type temp;
1e771e83 572 bool ret = true;
a912b54d
JK		 573
574 for (temp = HOT; temp < NR_TEMP_TYPE; temp++) {
1e771e83
YS		 575 if (!force) {
576 enum page_type btype = PAGE_TYPE_OF_BIO(type);
577 struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
a912b54d 578
1e771e83 579 down_read(&io->io_rwsem);
8648de2c 580 ret = __has_merged_page(io->bio, inode, page, ino);
1e771e83
YS		 581 up_read(&io->io_rwsem);
582 }
583 if (ret)
584 __f2fs_submit_merged_write(sbi, type, temp);
a912b54d
JK		 585
586 /* TODO: use HOT temp only for meta pages now. */
587 if (type >= META)
588 break;
589 }
0c3a5797
CY		 590}
591
b9109b0e 592void f2fs_submit_merged_write(struct f2fs_sb_info *sbi, enum page_type type)
0c3a5797 593{
adcc00f7 594 __submit_merged_write_cond(sbi, NULL, NULL, 0, type, true);
0c3a5797
CY		 595}
596
b9109b0e 597void f2fs_submit_merged_write_cond(struct f2fs_sb_info *sbi,
bab475c5
CY		 598 struct inode *inode, struct page *page,
599 nid_t ino, enum page_type type)
0c3a5797 600{
bab475c5 601 __submit_merged_write_cond(sbi, inode, page, ino, type, false);
0c3a5797
CY		 602}
603
b9109b0e 604void f2fs_flush_merged_writes(struct f2fs_sb_info *sbi)
406657dd 605{
b9109b0e
JK		 606 f2fs_submit_merged_write(sbi, DATA);
607 f2fs_submit_merged_write(sbi, NODE);
608 f2fs_submit_merged_write(sbi, META);
406657dd
CY		 609}
610
93dfe2ac
JK		 611/*
612 * Fill the locked page with data located in the block address.
771a9a71 613 * A caller needs to unlock the page on failure.
93dfe2ac 614 */
05ca3632 615int f2fs_submit_page_bio(struct f2fs_io_info *fio)
93dfe2ac 616{
93dfe2ac 617 struct bio *bio;
0b81d077
JK		 618 struct page *page = fio->encrypted_page ?
619 fio->encrypted_page : fio->page;
93dfe2ac 620
c9b60788 621 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
93770ab7
CY		 622 fio->is_por ? META_POR : (__is_meta_io(fio) ?
623 META_GENERIC : DATA_GENERIC_ENHANCE)))
10f966bb 624 return -EFSCORRUPTED;
c9b60788 625
2ace38e0 626 trace_f2fs_submit_page_bio(page, fio);
05ca3632 627 f2fs_trace_ios(fio, 0);
93dfe2ac
JK		 628
629 /* Allocate a new bio */
b757f6ed 630 bio = __bio_alloc(fio, 1);
93dfe2ac 631
09cbfeaf 632 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
93dfe2ac 633 bio_put(bio);
93dfe2ac
JK		 634 return -EFAULT;
635 }
78efac53
CY		 636
637 if (fio->io_wbc && !is_read_io(fio->op))
34e51a5e 638 wbc_account_cgroup_owner(fio->io_wbc, page, PAGE_SIZE);
78efac53 639
04d328de 640 bio_set_op_attrs(bio, fio->op, fio->op_flags);
93dfe2ac 641
5f9abab4
JK		 642 inc_page_count(fio->sbi, is_read_io(fio->op) ?
643 __read_io_type(page): WB_DATA_TYPE(fio->page));
4c58ed07
CY		 644
645 __submit_bio(fio->sbi, bio, fio->type);
93dfe2ac
JK		 646 return 0;
647}
648
8896cbdf
CY		 649static bool page_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
650 block_t last_blkaddr, block_t cur_blkaddr)
651{
652 if (last_blkaddr + 1 != cur_blkaddr)
653 return false;
654 return __same_bdev(sbi, cur_blkaddr, bio);
655}
656
657static bool io_type_is_mergeable(struct f2fs_bio_info *io,
658 struct f2fs_io_info *fio)
659{
660 if (io->fio.op != fio->op)
661 return false;
662 return io->fio.op_flags == fio->op_flags;
663}
664
665static bool io_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
666 struct f2fs_bio_info *io,
667 struct f2fs_io_info *fio,
668 block_t last_blkaddr,
669 block_t cur_blkaddr)
670{
c72db71e
CY		 671 if (F2FS_IO_ALIGNED(sbi) && (fio->type == DATA || fio->type == NODE)) {
672 unsigned int filled_blocks =
673 F2FS_BYTES_TO_BLK(bio->bi_iter.bi_size);
674 unsigned int io_size = F2FS_IO_SIZE(sbi);
675 unsigned int left_vecs = bio->bi_max_vecs - bio->bi_vcnt;
676
677 /* IOs in bio is aligned and left space of vectors is not enough */
678 if (!(filled_blocks % io_size) && left_vecs < io_size)
679 return false;
680 }
8896cbdf
CY		 681 if (!page_is_mergeable(sbi, bio, last_blkaddr, cur_blkaddr))
682 return false;
683 return io_type_is_mergeable(io, fio);
684}
685
0b20fcec
CY		 686static void add_bio_entry(struct f2fs_sb_info *sbi, struct bio *bio,
687 struct page *page, enum temp_type temp)
688{
689 struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
690 struct bio_entry *be;
691
692 be = f2fs_kmem_cache_alloc(bio_entry_slab, GFP_NOFS);
693 be->bio = bio;
694 bio_get(bio);
695
696 if (bio_add_page(bio, page, PAGE_SIZE, 0) != PAGE_SIZE)
697 f2fs_bug_on(sbi, 1);
698
699 down_write(&io->bio_list_lock);
700 list_add_tail(&be->list, &io->bio_list);
701 up_write(&io->bio_list_lock);
702}
703
704static void del_bio_entry(struct bio_entry *be)
705{
706 list_del(&be->list);
707 kmem_cache_free(bio_entry_slab, be);
708}
709
710static int add_ipu_page(struct f2fs_sb_info *sbi, struct bio **bio,
711 struct page *page)
712{
713 enum temp_type temp;
714 bool found = false;
715 int ret = -EAGAIN;
716
717 for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
718 struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
719 struct list_head *head = &io->bio_list;
720 struct bio_entry *be;
721
722 down_write(&io->bio_list_lock);
723 list_for_each_entry(be, head, list) {
724 if (be->bio != *bio)
725 continue;
726
727 found = true;
728
4c8ff709
CY		 729 if (bio_add_page(*bio, page, PAGE_SIZE, 0) ==
730 PAGE_SIZE) {
0b20fcec
CY		 731 ret = 0;
732 break;
733 }
734
735 /* bio is full */
736 del_bio_entry(be);
737 __submit_bio(sbi, *bio, DATA);
738 break;
739 }
740 up_write(&io->bio_list_lock);
741 }
742
743 if (ret) {
744 bio_put(*bio);
745 *bio = NULL;
746 }
747
748 return ret;
749}
750
751void f2fs_submit_merged_ipu_write(struct f2fs_sb_info *sbi,
752 struct bio **bio, struct page *page)
753{
754 enum temp_type temp;
755 bool found = false;
756 struct bio *target = bio ? *bio : NULL;
757
758 for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
759 struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
760 struct list_head *head = &io->bio_list;
761 struct bio_entry *be;
762
763 if (list_empty(head))
764 continue;
765
766 down_read(&io->bio_list_lock);
767 list_for_each_entry(be, head, list) {
768 if (target)
769 found = (target == be->bio);
770 else
771 found = __has_merged_page(be->bio, NULL,
772 page, 0);
773 if (found)
774 break;
775 }
776 up_read(&io->bio_list_lock);
777
778 if (!found)
779 continue;
780
781 found = false;
782
783 down_write(&io->bio_list_lock);
784 list_for_each_entry(be, head, list) {
785 if (target)
786 found = (target == be->bio);
787 else
788 found = __has_merged_page(be->bio, NULL,
789 page, 0);
790 if (found) {
791 target = be->bio;
792 del_bio_entry(be);
793 break;
794 }
795 }
796 up_write(&io->bio_list_lock);
797 }
798
799 if (found)
800 __submit_bio(sbi, target, DATA);
801 if (bio && *bio) {
802 bio_put(*bio);
803 *bio = NULL;
804 }
805}
806
8648de2c
CY		 807int f2fs_merge_page_bio(struct f2fs_io_info *fio)
808{
809 struct bio *bio = *fio->bio;
810 struct page *page = fio->encrypted_page ?
811 fio->encrypted_page : fio->page;
812
813 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
814 __is_meta_io(fio) ? META_GENERIC : DATA_GENERIC))
10f966bb 815 return -EFSCORRUPTED;
8648de2c
CY		 816
817 trace_f2fs_submit_page_bio(page, fio);
818 f2fs_trace_ios(fio, 0);
819
8896cbdf 820 if (bio && !page_is_mergeable(fio->sbi, bio, *fio->last_block,
0b20fcec
CY		 821 fio->new_blkaddr))
822 f2fs_submit_merged_ipu_write(fio->sbi, &bio, NULL);
8648de2c
CY		 823alloc_new:
824 if (!bio) {
b757f6ed 825 bio = __bio_alloc(fio, BIO_MAX_PAGES);
8648de2c 826 bio_set_op_attrs(bio, fio->op, fio->op_flags);
8648de2c 827
0b20fcec
CY		 828 add_bio_entry(fio->sbi, bio, page, fio->temp);
829 } else {
830 if (add_ipu_page(fio->sbi, &bio, page))
831 goto alloc_new;
8648de2c
CY		 832 }
833
834 if (fio->io_wbc)
9637d517 835 wbc_account_cgroup_owner(fio->io_wbc, page, PAGE_SIZE);
8648de2c
CY		 836
837 inc_page_count(fio->sbi, WB_DATA_TYPE(page));
838
839 *fio->last_block = fio->new_blkaddr;
840 *fio->bio = bio;
841
842 return 0;
843}
844
fe16efe6 845void f2fs_submit_page_write(struct f2fs_io_info *fio)
93dfe2ac 846{
05ca3632 847 struct f2fs_sb_info *sbi = fio->sbi;
458e6197 848 enum page_type btype = PAGE_TYPE_OF_BIO(fio->type);
a912b54d 849 struct f2fs_bio_info *io = sbi->write_io[btype] + fio->temp;
4375a336 850 struct page *bio_page;
93dfe2ac 851
b9109b0e 852 f2fs_bug_on(sbi, is_read_io(fio->op));
93dfe2ac 853
fb830fc5
CY		 854 down_write(&io->io_rwsem);
855next:
856 if (fio->in_list) {
857 spin_lock(&io->io_lock);
858 if (list_empty(&io->io_list)) {
859 spin_unlock(&io->io_lock);
fe16efe6 860 goto out;
fb830fc5
CY		 861 }
862 fio = list_first_entry(&io->io_list,
863 struct f2fs_io_info, list);
864 list_del(&fio->list);
865 spin_unlock(&io->io_lock);
866 }
93dfe2ac 867
93770ab7 868 verify_fio_blkaddr(fio);
93dfe2ac 869
4c8ff709
CY		 870 if (fio->encrypted_page)
871 bio_page = fio->encrypted_page;
872 else if (fio->compressed_page)
873 bio_page = fio->compressed_page;
874 else
875 bio_page = fio->page;
36951b38 876
ebf7c522
TM		 877 /* set submitted = true as a return value */
878 fio->submitted = true;
d68f735b 879
b9109b0e 880 inc_page_count(sbi, WB_DATA_TYPE(bio_page));
93dfe2ac 881
8896cbdf
CY		 882 if (io->bio && !io_is_mergeable(sbi, io->bio, io, fio,
883 io->last_block_in_bio, fio->new_blkaddr))
458e6197 884 __submit_merged_bio(io);
93dfe2ac
JK		 885alloc_new:
886 if (io->bio == NULL) {
8223ecc4
CY		 887 if (F2FS_IO_ALIGNED(sbi) &&
888 (fio->type == DATA || fio->type == NODE) &&
0a595eba 889 fio->new_blkaddr & F2FS_IO_SIZE_MASK(sbi)) {
b9109b0e 890 dec_page_count(sbi, WB_DATA_TYPE(bio_page));
fe16efe6
CY		 891 fio->retry = true;
892 goto skip;
0a595eba 893 }
b757f6ed 894 io->bio = __bio_alloc(fio, BIO_MAX_PAGES);
458e6197 895 io->fio = *fio;
93dfe2ac
JK		 896 }
897
a912b54d 898 if (bio_add_page(io->bio, bio_page, PAGE_SIZE, 0) < PAGE_SIZE) {
458e6197 899 __submit_merged_bio(io);
93dfe2ac
JK		 900 goto alloc_new;
901 }
902
578c6478 903 if (fio->io_wbc)
34e51a5e 904 wbc_account_cgroup_owner(fio->io_wbc, bio_page, PAGE_SIZE);
578c6478 905
7a9d7548 906 io->last_block_in_bio = fio->new_blkaddr;
05ca3632 907 f2fs_trace_ios(fio, 0);
fb830fc5
CY		 908
909 trace_f2fs_submit_page_write(fio->page, fio);
fe16efe6 910skip:
fb830fc5
CY		 911 if (fio->in_list)
912 goto next;
fe16efe6 913out:
4354994f 914 if (is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN) ||
00e09c0b 915 !f2fs_is_checkpoint_ready(sbi))
5ce80586 916 __submit_merged_bio(io);
df0f8dc0 917 up_write(&io->io_rwsem);
93dfe2ac
JK		 918}
919
95ae251f
EB		 920static inline bool f2fs_need_verity(const struct inode *inode, pgoff_t idx)
921{
922 return fsverity_active(inode) &&
923 idx < DIV_ROUND_UP(inode->i_size, PAGE_SIZE);
924}
925
13ba41e3 926static struct bio *f2fs_grab_read_bio(struct inode *inode, block_t blkaddr,
95ae251f
EB		 927 unsigned nr_pages, unsigned op_flag,
928 pgoff_t first_idx)
13ba41e3
JK		 929{
930 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
13ba41e3 931 struct bio *bio;
6dbb1796
EB		 932 struct bio_post_read_ctx *ctx;
933 unsigned int post_read_steps = 0;
13ba41e3 934
d62fe971 935 bio = f2fs_bio_alloc(sbi, min_t(int, nr_pages, BIO_MAX_PAGES), false);
6dbb1796 936 if (!bio)
13ba41e3 937 return ERR_PTR(-ENOMEM);
13ba41e3
JK		 938 f2fs_target_device(sbi, blkaddr, bio);
939 bio->bi_end_io = f2fs_read_end_io;
e2e59414 940 bio_set_op_attrs(bio, REQ_OP_READ, op_flag);
13ba41e3 941
6dbb1796
EB		 942 if (f2fs_encrypted_file(inode))
943 post_read_steps |= 1 << STEP_DECRYPT;
4c8ff709
CY		 944 if (f2fs_compressed_file(inode))
945 post_read_steps |= 1 << STEP_DECOMPRESS;
95ae251f
EB		 946 if (f2fs_need_verity(inode, first_idx))
947 post_read_steps |= 1 << STEP_VERITY;
948
6dbb1796 949 if (post_read_steps) {
e8ce5749 950 /* Due to the mempool, this never fails. */
6dbb1796 951 ctx = mempool_alloc(bio_post_read_ctx_pool, GFP_NOFS);
6dbb1796 952 ctx->bio = bio;
4c8ff709 953 ctx->sbi = sbi;
6dbb1796
EB		 954 ctx->enabled_steps = post_read_steps;
955 bio->bi_private = ctx;
6dbb1796
EB		 956 }
957
13ba41e3
JK		 958 return bio;
959}
960
4c8ff709
CY		 961static void f2fs_release_read_bio(struct bio *bio)
962{
963 if (bio->bi_private)
964 mempool_free(bio->bi_private, bio_post_read_ctx_pool);
965 bio_put(bio);
966}
967
13ba41e3
JK		 968/* This can handle encryption stuffs */
969static int f2fs_submit_page_read(struct inode *inode, struct page *page,
970 block_t blkaddr)
971{
93770ab7
CY		 972 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
973 struct bio *bio;
13ba41e3 974
95ae251f 975 bio = f2fs_grab_read_bio(inode, blkaddr, 1, 0, page->index);
13ba41e3
JK		 976 if (IS_ERR(bio))
977 return PTR_ERR(bio);
978
0ded69f6
JK		 979 /* wait for GCed page writeback via META_MAPPING */
980 f2fs_wait_on_block_writeback(inode, blkaddr);
981
13ba41e3
JK		 982 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
983 bio_put(bio);
984 return -EFAULT;
985 }
fb7d70db 986 ClearPageError(page);
93770ab7
CY		 987 inc_page_count(sbi, F2FS_RD_DATA);
988 __submit_bio(sbi, bio, DATA);
13ba41e3
JK		 989 return 0;
990}
991
46008c6d
CY		 992static void __set_data_blkaddr(struct dnode_of_data *dn)
993{
994 struct f2fs_node *rn = F2FS_NODE(dn->node_page);
995 __le32 *addr_array;
7a2af766
CY		 996 int base = 0;
997
998 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
999 base = get_extra_isize(dn->inode);
46008c6d
CY		 1000
1001 /* Get physical address of data block */
1002 addr_array = blkaddr_in_node(rn);
7a2af766 1003 addr_array[base + dn->ofs_in_node] = cpu_to_le32(dn->data_blkaddr);
46008c6d
CY		 1004}
1005
0a8165d7 1006/*
eb47b800
JK		 1007 * Lock ordering for the change of data block address:
1008 * ->data_page
1009 * ->node_page
1010 * update block addresses in the node page
1011 */
4d57b86d 1012void f2fs_set_data_blkaddr(struct dnode_of_data *dn)
eb47b800 1013{
bae0ee7a 1014 f2fs_wait_on_page_writeback(dn->node_page, NODE, true, true);
46008c6d
CY		 1015 __set_data_blkaddr(dn);
1016 if (set_page_dirty(dn->node_page))
12719ae1 1017 dn->node_changed = true;
eb47b800
JK		 1018}
1019
f28b3434
CY		 1020void f2fs_update_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr)
1021{
1022 dn->data_blkaddr = blkaddr;
4d57b86d 1023 f2fs_set_data_blkaddr(dn);
f28b3434
CY		 1024 f2fs_update_extent_cache(dn);
1025}
1026
46008c6d 1027/* dn->ofs_in_node will be returned with up-to-date last block pointer */
4d57b86d 1028int f2fs_reserve_new_blocks(struct dnode_of_data *dn, blkcnt_t count)
eb47b800 1029{
4081363f 1030 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
0abd675e 1031 int err;
eb47b800 1032
46008c6d
CY		 1033 if (!count)
1034 return 0;
1035
91942321 1036 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
eb47b800 1037 return -EPERM;
0abd675e
CY		 1038 if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
1039 return err;
eb47b800 1040
46008c6d
CY		 1041 trace_f2fs_reserve_new_blocks(dn->inode, dn->nid,
1042 dn->ofs_in_node, count);
1043
bae0ee7a 1044 f2fs_wait_on_page_writeback(dn->node_page, NODE, true, true);
46008c6d
CY		 1045
1046 for (; count > 0; dn->ofs_in_node++) {
a2ced1ce 1047 block_t blkaddr = f2fs_data_blkaddr(dn);
46008c6d
CY		 1048 if (blkaddr == NULL_ADDR) {
1049 dn->data_blkaddr = NEW_ADDR;
1050 __set_data_blkaddr(dn);
1051 count--;
1052 }
1053 }
1054
1055 if (set_page_dirty(dn->node_page))
1056 dn->node_changed = true;
eb47b800
JK		 1057 return 0;
1058}
1059
46008c6d 1060/* Should keep dn->ofs_in_node unchanged */
4d57b86d 1061int f2fs_reserve_new_block(struct dnode_of_data *dn)
46008c6d
CY		 1062{
1063 unsigned int ofs_in_node = dn->ofs_in_node;
1064 int ret;
1065
4d57b86d 1066 ret = f2fs_reserve_new_blocks(dn, 1);
46008c6d
CY		 1067 dn->ofs_in_node = ofs_in_node;
1068 return ret;
1069}
1070
b600965c
HL		 1071int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
1072{
1073 bool need_put = dn->inode_page ? false : true;
1074 int err;
1075
4d57b86d 1076 err = f2fs_get_dnode_of_data(dn, index, ALLOC_NODE);
b600965c
HL		 1077 if (err)
1078 return err;
a8865372 1079
b600965c 1080 if (dn->data_blkaddr == NULL_ADDR)
4d57b86d 1081 err = f2fs_reserve_new_block(dn);
a8865372 1082 if (err || need_put)
b600965c
HL		 1083 f2fs_put_dnode(dn);
1084 return err;
1085}
1086
759af1c9 1087int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index)
eb47b800 1088{
e15882b6 1089 struct extent_info ei = {0,0,0};
759af1c9 1090 struct inode *inode = dn->inode;
028a41e8 1091
759af1c9
FL		 1092 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
1093 dn->data_blkaddr = ei.blk + index - ei.fofs;
1094 return 0;
429511cd 1095 }
028a41e8 1096
759af1c9 1097 return f2fs_reserve_block(dn, index);
eb47b800
JK		 1098}
1099
4d57b86d 1100struct page *f2fs_get_read_data_page(struct inode *inode, pgoff_t index,
04d328de 1101 int op_flags, bool for_write)
eb47b800 1102{
eb47b800
JK		 1103 struct address_space *mapping = inode->i_mapping;
1104 struct dnode_of_data dn;
1105 struct page *page;
e15882b6 1106 struct extent_info ei = {0,0,0};
eb47b800 1107 int err;
4375a336 1108
a56c7c6f 1109 page = f2fs_grab_cache_page(mapping, index, for_write);
650495de
JK		 1110 if (!page)
1111 return ERR_PTR(-ENOMEM);
1112
cb3bc9ee
CY		 1113 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
1114 dn.data_blkaddr = ei.blk + index - ei.fofs;
93770ab7
CY		 1115 if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), dn.data_blkaddr,
1116 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 1117 err = -EFSCORRUPTED;
93770ab7
CY		 1118 goto put_err;
1119 }
cb3bc9ee
CY		 1120 goto got_it;
1121 }
1122
eb47b800 1123 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1124 err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
86531d6b
JK		 1125 if (err)
1126 goto put_err;
eb47b800
JK		 1127 f2fs_put_dnode(&dn);
1128
6bacf52f 1129 if (unlikely(dn.data_blkaddr == NULL_ADDR)) {
86531d6b
JK		 1130 err = -ENOENT;
1131 goto put_err;
650495de 1132 }
93770ab7
CY		 1133 if (dn.data_blkaddr != NEW_ADDR &&
1134 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
1135 dn.data_blkaddr,
1136 DATA_GENERIC_ENHANCE)) {
10f966bb 1137 err = -EFSCORRUPTED;
93770ab7
CY		 1138 goto put_err;
1139 }
cb3bc9ee 1140got_it:
43f3eae1
JK		 1141 if (PageUptodate(page)) {
1142 unlock_page(page);
eb47b800 1143 return page;
43f3eae1 1144 }
eb47b800 1145
d59ff4df
JK		 1146 /*
1147 * A new dentry page is allocated but not able to be written, since its
1148 * new inode page couldn't be allocated due to -ENOSPC.
1149 * In such the case, its blkaddr can be remained as NEW_ADDR.
4d57b86d
CY		 1150 * see, f2fs_add_link -> f2fs_get_new_data_page ->
1151 * f2fs_init_inode_metadata.
d59ff4df
JK		 1152 */
1153 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 1154 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 1155 if (!PageUptodate(page))
1156 SetPageUptodate(page);
43f3eae1 1157 unlock_page(page);
d59ff4df
JK		 1158 return page;
1159 }
eb47b800 1160
13ba41e3 1161 err = f2fs_submit_page_read(inode, page, dn.data_blkaddr);
393ff91f 1162 if (err)
86531d6b 1163 goto put_err;
43f3eae1 1164 return page;
86531d6b
JK		 1165
1166put_err:
1167 f2fs_put_page(page, 1);
1168 return ERR_PTR(err);
43f3eae1
JK		 1169}
1170
4d57b86d 1171struct page *f2fs_find_data_page(struct inode *inode, pgoff_t index)
43f3eae1
JK		 1172{
1173 struct address_space *mapping = inode->i_mapping;
1174 struct page *page;
1175
1176 page = find_get_page(mapping, index);
1177 if (page && PageUptodate(page))
1178 return page;
1179 f2fs_put_page(page, 0);
1180
4d57b86d 1181 page = f2fs_get_read_data_page(inode, index, 0, false);
43f3eae1
JK		 1182 if (IS_ERR(page))
1183 return page;
1184
1185 if (PageUptodate(page))
1186 return page;
1187
1188 wait_on_page_locked(page);
1189 if (unlikely(!PageUptodate(page))) {
1190 f2fs_put_page(page, 0);
1191 return ERR_PTR(-EIO);
1192 }
1193 return page;
1194}
1195
1196/*
1197 * If it tries to access a hole, return an error.
1198 * Because, the callers, functions in dir.c and GC, should be able to know
1199 * whether this page exists or not.
1200 */
4d57b86d 1201struct page *f2fs_get_lock_data_page(struct inode *inode, pgoff_t index,
a56c7c6f 1202 bool for_write)
43f3eae1
JK		 1203{
1204 struct address_space *mapping = inode->i_mapping;
1205 struct page *page;
1206repeat:
4d57b86d 1207 page = f2fs_get_read_data_page(inode, index, 0, for_write);
43f3eae1
JK		 1208 if (IS_ERR(page))
1209 return page;
393ff91f 1210
43f3eae1 1211 /* wait for read completion */
393ff91f 1212 lock_page(page);
6bacf52f 1213 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 1214 f2fs_put_page(page, 1);
1215 goto repeat;
eb47b800 1216 }
1563ac75
CY		 1217 if (unlikely(!PageUptodate(page))) {
1218 f2fs_put_page(page, 1);
1219 return ERR_PTR(-EIO);
1220 }
eb47b800
JK		 1221 return page;
1222}
1223
0a8165d7 1224/*
eb47b800
JK		 1225 * Caller ensures that this data page is never allocated.
1226 * A new zero-filled data page is allocated in the page cache.
39936837 1227 *
4f4124d0
CY		 1228 * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and
1229 * f2fs_unlock_op().
470f00e9
CY		 1230 * Note that, ipage is set only by make_empty_dir, and if any error occur,
1231 * ipage should be released by this function.
eb47b800 1232 */
4d57b86d 1233struct page *f2fs_get_new_data_page(struct inode *inode,
a8865372 1234 struct page *ipage, pgoff_t index, bool new_i_size)
eb47b800 1235{
eb47b800
JK		 1236 struct address_space *mapping = inode->i_mapping;
1237 struct page *page;
1238 struct dnode_of_data dn;
1239 int err;
7612118a 1240
a56c7c6f 1241 page = f2fs_grab_cache_page(mapping, index, true);
470f00e9
CY		 1242 if (!page) {
1243 /*
1244 * before exiting, we should make sure ipage will be released
1245 * if any error occur.
1246 */
1247 f2fs_put_page(ipage, 1);
01f28610 1248 return ERR_PTR(-ENOMEM);
470f00e9 1249 }
eb47b800 1250
a8865372 1251 set_new_dnode(&dn, inode, ipage, NULL, 0);
b600965c 1252 err = f2fs_reserve_block(&dn, index);
01f28610
JK		 1253 if (err) {
1254 f2fs_put_page(page, 1);
eb47b800 1255 return ERR_PTR(err);
a8865372 1256 }
01f28610
JK		 1257 if (!ipage)
1258 f2fs_put_dnode(&dn);
eb47b800
JK		 1259
1260 if (PageUptodate(page))
01f28610 1261 goto got_it;
eb47b800
JK		 1262
1263 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 1264 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 1265 if (!PageUptodate(page))
1266 SetPageUptodate(page);
eb47b800 1267 } else {
4375a336 1268 f2fs_put_page(page, 1);
a8865372 1269
7612118a
JK		 1270 /* if ipage exists, blkaddr should be NEW_ADDR */
1271 f2fs_bug_on(F2FS_I_SB(inode), ipage);
4d57b86d 1272 page = f2fs_get_lock_data_page(inode, index, true);
4375a336 1273 if (IS_ERR(page))
7612118a 1274 return page;
eb47b800 1275 }
01f28610 1276got_it:
9edcdabf 1277 if (new_i_size && i_size_read(inode) <
ee6d182f 1278 ((loff_t)(index + 1) << PAGE_SHIFT))
fc9581c8 1279 f2fs_i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT));
eb47b800
JK		 1280 return page;
1281}
1282
d5097be5 1283static int __allocate_data_block(struct dnode_of_data *dn, int seg_type)
bfad7c2d 1284{
4081363f 1285 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
bfad7c2d 1286 struct f2fs_summary sum;
bfad7c2d 1287 struct node_info ni;
6aa58d8a 1288 block_t old_blkaddr;
46008c6d 1289 blkcnt_t count = 1;
0abd675e 1290 int err;
bfad7c2d 1291
91942321 1292 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
bfad7c2d 1293 return -EPERM;
df6136ef 1294
7735730d
CY		 1295 err = f2fs_get_node_info(sbi, dn->nid, &ni);
1296 if (err)
1297 return err;
1298
a2ced1ce 1299 dn->data_blkaddr = f2fs_data_blkaddr(dn);
f847c699 1300 if (dn->data_blkaddr != NULL_ADDR)
df6136ef
CY		 1301 goto alloc;
1302
0abd675e
CY		 1303 if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
1304 return err;
bfad7c2d 1305
df6136ef 1306alloc:
bfad7c2d 1307 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
6aa58d8a
CY		 1308 old_blkaddr = dn->data_blkaddr;
1309 f2fs_allocate_data_block(sbi, NULL, old_blkaddr, &dn->data_blkaddr,
d5097be5 1310 &sum, seg_type, NULL, false);
6aa58d8a
CY		 1311 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
1312 invalidate_mapping_pages(META_MAPPING(sbi),
1313 old_blkaddr, old_blkaddr);
86f35dc3 1314 f2fs_update_data_blkaddr(dn, dn->data_blkaddr);
bfad7c2d 1315
0a4daae5
JK		 1316 /*
1317 * i_size will be updated by direct_IO. Otherwise, we'll get stale
1318 * data from unwritten block via dio_read.
1319 */
bfad7c2d
JK		 1320 return 0;
1321}
1322
a7de6086 1323int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *from)
59b802e5 1324{
b439b103 1325 struct inode *inode = file_inode(iocb->ki_filp);
5b8db7fa 1326 struct f2fs_map_blocks map;
d6d478a1 1327 int flag;
a7de6086 1328 int err = 0;
d6d478a1 1329 bool direct_io = iocb->ki_flags & IOCB_DIRECT;
59b802e5 1330
0080c507 1331 map.m_lblk = F2FS_BLK_ALIGN(iocb->ki_pos);
dfd02e4d
CY		 1332 map.m_len = F2FS_BYTES_TO_BLK(iocb->ki_pos + iov_iter_count(from));
1333 if (map.m_len > map.m_lblk)
1334 map.m_len -= map.m_lblk;
1335 else
1336 map.m_len = 0;
1337
da85985c 1338 map.m_next_pgofs = NULL;
c4020b2d 1339 map.m_next_extent = NULL;
d5097be5 1340 map.m_seg_type = NO_CHECK_TYPE;
f9d6d059 1341 map.m_may_create = true;
2a340760 1342
d6d478a1 1343 if (direct_io) {
4d57b86d 1344 map.m_seg_type = f2fs_rw_hint_to_seg_type(iocb->ki_hint);
f847c699 1345 flag = f2fs_force_buffered_io(inode, iocb, from) ?
d6d478a1
CY		 1346 F2FS_GET_BLOCK_PRE_AIO :
1347 F2FS_GET_BLOCK_PRE_DIO;
1348 goto map_blocks;
d5097be5 1349 }
f2470371 1350 if (iocb->ki_pos + iov_iter_count(from) > MAX_INLINE_DATA(inode)) {
a7de6086
JK		 1351 err = f2fs_convert_inline_inode(inode);
1352 if (err)
1353 return err;
b439b103 1354 }
d6d478a1 1355 if (f2fs_has_inline_data(inode))
25006645 1356 return err;
d6d478a1
CY		 1357
1358 flag = F2FS_GET_BLOCK_PRE_AIO;
1359
1360map_blocks:
1361 err = f2fs_map_blocks(inode, &map, 1, flag);
1362 if (map.m_len > 0 && err == -ENOSPC) {
1363 if (!direct_io)
1364 set_inode_flag(inode, FI_NO_PREALLOC);
1365 err = 0;
25006645 1366 }
a7de6086 1367 return err;
59b802e5
JK		 1368}
1369
39a86958 1370void __do_map_lock(struct f2fs_sb_info *sbi, int flag, bool lock)
59c9081b
YH		 1371{
1372 if (flag == F2FS_GET_BLOCK_PRE_AIO) {
1373 if (lock)
1374 down_read(&sbi->node_change);
1375 else
1376 up_read(&sbi->node_change);
1377 } else {
1378 if (lock)
1379 f2fs_lock_op(sbi);
1380 else
1381 f2fs_unlock_op(sbi);
1382 }
1383}
1384
0a8165d7 1385/*
7a88ddb5
CY		 1386 * f2fs_map_blocks() tries to find or build mapping relationship which
1387 * maps continuous logical blocks to physical blocks, and return such
1388 * info via f2fs_map_blocks structure.
eb47b800 1389 */
d323d005 1390int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
e2b4e2bc 1391 int create, int flag)
eb47b800 1392{
003a3e1d 1393 unsigned int maxblocks = map->m_len;
eb47b800 1394 struct dnode_of_data dn;
f9811703 1395 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
f9d6d059 1396 int mode = map->m_may_create ? ALLOC_NODE : LOOKUP_NODE;
46008c6d 1397 pgoff_t pgofs, end_offset, end;
bfad7c2d 1398 int err = 0, ofs = 1;
46008c6d
CY		 1399 unsigned int ofs_in_node, last_ofs_in_node;
1400 blkcnt_t prealloc;
e15882b6 1401 struct extent_info ei = {0,0,0};
7df3a431 1402 block_t blkaddr;
c4020b2d 1403 unsigned int start_pgofs;
eb47b800 1404
dfd02e4d
CY		 1405 if (!maxblocks)
1406 return 0;
1407
003a3e1d
JK		 1408 map->m_len = 0;
1409 map->m_flags = 0;
1410
1411 /* it only supports block size == page size */
1412 pgofs = (pgoff_t)map->m_lblk;
46008c6d 1413 end = pgofs + maxblocks;
eb47b800 1414
24b84912 1415 if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) {
b0332a0f 1416 if (f2fs_lfs_mode(sbi) && flag == F2FS_GET_BLOCK_DIO &&
f4f0b677
JZ		 1417 map->m_may_create)
1418 goto next_dnode;
1419
003a3e1d
JK		 1420 map->m_pblk = ei.blk + pgofs - ei.fofs;
1421 map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs);
1422 map->m_flags = F2FS_MAP_MAPPED;
c4020b2d
CY		 1423 if (map->m_next_extent)
1424 *map->m_next_extent = pgofs + map->m_len;
1e78e8bd
ST		 1425
1426 /* for hardware encryption, but to avoid potential issue in future */
1427 if (flag == F2FS_GET_BLOCK_DIO)
1428 f2fs_wait_on_block_writeback_range(inode,
1429 map->m_pblk, map->m_len);
bfad7c2d 1430 goto out;
a2e7d1bf 1431 }
bfad7c2d 1432
4fe71e88 1433next_dnode:
f9d6d059 1434 if (map->m_may_create)
59c9081b 1435 __do_map_lock(sbi, flag, true);
eb47b800
JK		 1436
1437 /* When reading holes, we need its node page */
1438 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1439 err = f2fs_get_dnode_of_data(&dn, pgofs, mode);
1ec79083 1440 if (err) {
43473f96
CY		 1441 if (flag == F2FS_GET_BLOCK_BMAP)
1442 map->m_pblk = 0;
da85985c 1443 if (err == -ENOENT) {
bfad7c2d 1444 err = 0;
da85985c
CY		 1445 if (map->m_next_pgofs)
1446 *map->m_next_pgofs =
4d57b86d 1447 f2fs_get_next_page_offset(&dn, pgofs);
c4020b2d
CY		 1448 if (map->m_next_extent)
1449 *map->m_next_extent =
4d57b86d 1450 f2fs_get_next_page_offset(&dn, pgofs);
da85985c 1451 }
bfad7c2d 1452 goto unlock_out;
848753aa 1453 }
973163fc 1454
c4020b2d 1455 start_pgofs = pgofs;
46008c6d 1456 prealloc = 0;
230436b3 1457 last_ofs_in_node = ofs_in_node = dn.ofs_in_node;
81ca7350 1458 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
4fe71e88
CY		 1459
1460next_block:
a2ced1ce 1461 blkaddr = f2fs_data_blkaddr(&dn);
4fe71e88 1462
c9b60788 1463 if (__is_valid_data_blkaddr(blkaddr) &&
93770ab7 1464 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC_ENHANCE)) {
10f966bb 1465 err = -EFSCORRUPTED;
c9b60788
CY		 1466 goto sync_out;
1467 }
1468
93770ab7 1469 if (__is_valid_data_blkaddr(blkaddr)) {
f847c699 1470 /* use out-place-update for driect IO under LFS mode */
b0332a0f 1471 if (f2fs_lfs_mode(sbi) && flag == F2FS_GET_BLOCK_DIO &&
f9d6d059 1472 map->m_may_create) {
f847c699 1473 err = __allocate_data_block(&dn, map->m_seg_type);
05e36006
CY		 1474 if (err)
1475 goto sync_out;
1476 blkaddr = dn.data_blkaddr;
1477 set_inode_flag(inode, FI_APPEND_WRITE);
f847c699
CY		 1478 }
1479 } else {
973163fc 1480 if (create) {
f9811703
CY		 1481 if (unlikely(f2fs_cp_error(sbi))) {
1482 err = -EIO;
4fe71e88 1483 goto sync_out;
f9811703 1484 }
24b84912 1485 if (flag == F2FS_GET_BLOCK_PRE_AIO) {
46008c6d
CY		 1486 if (blkaddr == NULL_ADDR) {
1487 prealloc++;
1488 last_ofs_in_node = dn.ofs_in_node;
1489 }
24b84912 1490 } else {
0a4daae5
JK		 1491 WARN_ON(flag != F2FS_GET_BLOCK_PRE_DIO &&
1492 flag != F2FS_GET_BLOCK_DIO);
d5097be5
HL		 1493 err = __allocate_data_block(&dn,
1494 map->m_seg_type);
6f2d8ed6 1495 if (!err)
91942321 1496 set_inode_flag(inode, FI_APPEND_WRITE);
24b84912 1497 }
973163fc 1498 if (err)
4fe71e88 1499 goto sync_out;
3f2be043 1500 map->m_flags |= F2FS_MAP_NEW;
4fe71e88 1501 blkaddr = dn.data_blkaddr;
973163fc 1502 } else {
43473f96
CY		 1503 if (flag == F2FS_GET_BLOCK_BMAP) {
1504 map->m_pblk = 0;
1505 goto sync_out;
1506 }
c4020b2d
CY		 1507 if (flag == F2FS_GET_BLOCK_PRECACHE)
1508 goto sync_out;
da85985c
CY		 1509 if (flag == F2FS_GET_BLOCK_FIEMAP &&
1510 blkaddr == NULL_ADDR) {
1511 if (map->m_next_pgofs)
1512 *map->m_next_pgofs = pgofs + 1;
4c2ac6a8 1513 goto sync_out;
da85985c 1514 }
f3d98e74
CY		 1515 if (flag != F2FS_GET_BLOCK_FIEMAP) {
1516 /* for defragment case */
1517 if (map->m_next_pgofs)
1518 *map->m_next_pgofs = pgofs + 1;
4fe71e88 1519 goto sync_out;
f3d98e74 1520 }
e2b4e2bc 1521 }
e2b4e2bc 1522 }
eb47b800 1523
46008c6d
CY		 1524 if (flag == F2FS_GET_BLOCK_PRE_AIO)
1525 goto skip;
1526
4fe71e88
CY		 1527 if (map->m_len == 0) {
1528 /* preallocated unwritten block should be mapped for fiemap. */
1529 if (blkaddr == NEW_ADDR)
1530 map->m_flags |= F2FS_MAP_UNWRITTEN;
1531 map->m_flags |= F2FS_MAP_MAPPED;
1532
1533 map->m_pblk = blkaddr;
1534 map->m_len = 1;
1535 } else if ((map->m_pblk != NEW_ADDR &&
1536 blkaddr == (map->m_pblk + ofs)) ||
b439b103 1537 (map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR) ||
46008c6d 1538 flag == F2FS_GET_BLOCK_PRE_DIO) {
4fe71e88
CY		 1539 ofs++;
1540 map->m_len++;
1541 } else {
1542 goto sync_out;
1543 }
bfad7c2d 1544
46008c6d 1545skip:
bfad7c2d
JK		 1546 dn.ofs_in_node++;
1547 pgofs++;
1548
46008c6d
CY		 1549 /* preallocate blocks in batch for one dnode page */
1550 if (flag == F2FS_GET_BLOCK_PRE_AIO &&
1551 (pgofs == end || dn.ofs_in_node == end_offset)) {
7df3a431 1552
46008c6d 1553 dn.ofs_in_node = ofs_in_node;
4d57b86d 1554 err = f2fs_reserve_new_blocks(&dn, prealloc);
46008c6d
CY		 1555 if (err)
1556 goto sync_out;
bfad7c2d 1557
46008c6d
CY		 1558 map->m_len += dn.ofs_in_node - ofs_in_node;
1559 if (prealloc && dn.ofs_in_node != last_ofs_in_node + 1) {
1560 err = -ENOSPC;
1561 goto sync_out;
3104af35 1562 }
46008c6d
CY		 1563 dn.ofs_in_node = end_offset;
1564 }
1565
1566 if (pgofs >= end)
1567 goto sync_out;
1568 else if (dn.ofs_in_node < end_offset)
1569 goto next_block;
1570
c4020b2d
CY		 1571 if (flag == F2FS_GET_BLOCK_PRECACHE) {
1572 if (map->m_flags & F2FS_MAP_MAPPED) {
1573 unsigned int ofs = start_pgofs - map->m_lblk;
1574
1575 f2fs_update_extent_cache_range(&dn,
1576 start_pgofs, map->m_pblk + ofs,
1577 map->m_len - ofs);
1578 }
1579 }
1580
46008c6d
CY		 1581 f2fs_put_dnode(&dn);
1582
f9d6d059 1583 if (map->m_may_create) {
59c9081b 1584 __do_map_lock(sbi, flag, false);
6f2d8ed6 1585 f2fs_balance_fs(sbi, dn.node_changed);
eb47b800 1586 }
46008c6d 1587 goto next_dnode;
7df3a431 1588
bfad7c2d 1589sync_out:
1e78e8bd
ST		 1590
1591 /* for hardware encryption, but to avoid potential issue in future */
1592 if (flag == F2FS_GET_BLOCK_DIO && map->m_flags & F2FS_MAP_MAPPED)
1593 f2fs_wait_on_block_writeback_range(inode,
1594 map->m_pblk, map->m_len);
1595
c4020b2d
CY		 1596 if (flag == F2FS_GET_BLOCK_PRECACHE) {
1597 if (map->m_flags & F2FS_MAP_MAPPED) {
1598 unsigned int ofs = start_pgofs - map->m_lblk;
1599
1600 f2fs_update_extent_cache_range(&dn,
1601 start_pgofs, map->m_pblk + ofs,
1602 map->m_len - ofs);
1603 }
1604 if (map->m_next_extent)
1605 *map->m_next_extent = pgofs + 1;
1606 }
eb47b800 1607 f2fs_put_dnode(&dn);
bfad7c2d 1608unlock_out:
f9d6d059 1609 if (map->m_may_create) {
59c9081b 1610 __do_map_lock(sbi, flag, false);
6f2d8ed6 1611 f2fs_balance_fs(sbi, dn.node_changed);
2a340760 1612 }
bfad7c2d 1613out:
003a3e1d 1614 trace_f2fs_map_blocks(inode, map, err);
bfad7c2d 1615 return err;
eb47b800
JK		 1616}
1617
b91050a8
HL		 1618bool f2fs_overwrite_io(struct inode *inode, loff_t pos, size_t len)
1619{
1620 struct f2fs_map_blocks map;
1621 block_t last_lblk;
1622 int err;
1623
1624 if (pos + len > i_size_read(inode))
1625 return false;
1626
1627 map.m_lblk = F2FS_BYTES_TO_BLK(pos);
1628 map.m_next_pgofs = NULL;
1629 map.m_next_extent = NULL;
1630 map.m_seg_type = NO_CHECK_TYPE;
f4f0b677 1631 map.m_may_create = false;
b91050a8
HL		 1632 last_lblk = F2FS_BLK_ALIGN(pos + len);
1633
1634 while (map.m_lblk < last_lblk) {
1635 map.m_len = last_lblk - map.m_lblk;
1636 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
1637 if (err || map.m_len == 0)
1638 return false;
1639 map.m_lblk += map.m_len;
1640 }
1641 return true;
1642}
1643
003a3e1d 1644static int __get_data_block(struct inode *inode, sector_t iblock,
da85985c 1645 struct buffer_head *bh, int create, int flag,
f9d6d059 1646 pgoff_t *next_pgofs, int seg_type, bool may_write)
003a3e1d
JK		 1647{
1648 struct f2fs_map_blocks map;
a7de6086 1649 int err;
003a3e1d
JK		 1650
1651 map.m_lblk = iblock;
1652 map.m_len = bh->b_size >> inode->i_blkbits;
da85985c 1653 map.m_next_pgofs = next_pgofs;
c4020b2d 1654 map.m_next_extent = NULL;
d5097be5 1655 map.m_seg_type = seg_type;
f9d6d059 1656 map.m_may_create = may_write;
003a3e1d 1657
a7de6086
JK		 1658 err = f2fs_map_blocks(inode, &map, create, flag);
1659 if (!err) {
003a3e1d
JK		 1660 map_bh(bh, inode->i_sb, map.m_pblk);
1661 bh->b_state = (bh->b_state & ~F2FS_MAP_FLAGS) | map.m_flags;
b86e3307 1662 bh->b_size = (u64)map.m_len << inode->i_blkbits;
003a3e1d 1663 }
a7de6086 1664 return err;
003a3e1d
JK		 1665}
1666
ccfb3000 1667static int get_data_block(struct inode *inode, sector_t iblock,
da85985c
CY		 1668 struct buffer_head *bh_result, int create, int flag,
1669 pgoff_t *next_pgofs)
e2b4e2bc 1670{
da85985c 1671 return __get_data_block(inode, iblock, bh_result, create,
d5097be5 1672 flag, next_pgofs,
f9d6d059
CY		 1673 NO_CHECK_TYPE, create);
1674}
1675
1676static int get_data_block_dio_write(struct inode *inode, sector_t iblock,
1677 struct buffer_head *bh_result, int create)
1678{
1679 return __get_data_block(inode, iblock, bh_result, create,
1680 F2FS_GET_BLOCK_DIO, NULL,
1681 f2fs_rw_hint_to_seg_type(inode->i_write_hint),
75a037f3 1682 IS_SWAPFILE(inode) ? false : true);
e2b4e2bc
CY		 1683}
1684
1685static int get_data_block_dio(struct inode *inode, sector_t iblock,
ccfb3000
JK		 1686 struct buffer_head *bh_result, int create)
1687{
e2b4e2bc 1688 return __get_data_block(inode, iblock, bh_result, create,
f9d6d059
CY		 1689 F2FS_GET_BLOCK_DIO, NULL,
1690 f2fs_rw_hint_to_seg_type(inode->i_write_hint),
1691 false);
ccfb3000
JK		 1692}
1693
e2b4e2bc 1694static int get_data_block_bmap(struct inode *inode, sector_t iblock,
ccfb3000
JK		 1695 struct buffer_head *bh_result, int create)
1696{
179448bf 1697 /* Block number less than F2FS MAX BLOCKS */
e0afc4d6 1698 if (unlikely(iblock >= F2FS_I_SB(inode)->max_file_blocks))
179448bf
YH		 1699 return -EFBIG;
1700
e2b4e2bc 1701 return __get_data_block(inode, iblock, bh_result, create,
d5097be5 1702 F2FS_GET_BLOCK_BMAP, NULL,
f9d6d059 1703 NO_CHECK_TYPE, create);
ccfb3000
JK		 1704}
1705
7f63eb77
JK		 1706static inline sector_t logical_to_blk(struct inode *inode, loff_t offset)
1707{
1708 return (offset >> inode->i_blkbits);
1709}
1710
1711static inline loff_t blk_to_logical(struct inode *inode, sector_t blk)
1712{
1713 return (blk << inode->i_blkbits);
1714}
1715
442a9dbd
CY		 1716static int f2fs_xattr_fiemap(struct inode *inode,
1717 struct fiemap_extent_info *fieinfo)
1718{
1719 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1720 struct page *page;
1721 struct node_info ni;
1722 __u64 phys = 0, len;
1723 __u32 flags;
1724 nid_t xnid = F2FS_I(inode)->i_xattr_nid;
1725 int err = 0;
1726
1727 if (f2fs_has_inline_xattr(inode)) {
1728 int offset;
1729
1730 page = f2fs_grab_cache_page(NODE_MAPPING(sbi),
1731 inode->i_ino, false);
1732 if (!page)
1733 return -ENOMEM;
1734
7735730d
CY		 1735 err = f2fs_get_node_info(sbi, inode->i_ino, &ni);
1736 if (err) {
1737 f2fs_put_page(page, 1);
1738 return err;
1739 }
442a9dbd
CY		 1740
1741 phys = (__u64)blk_to_logical(inode, ni.blk_addr);
1742 offset = offsetof(struct f2fs_inode, i_addr) +
1743 sizeof(__le32) * (DEF_ADDRS_PER_INODE -
b323fd28 1744 get_inline_xattr_addrs(inode));
442a9dbd
CY		 1745
1746 phys += offset;
1747 len = inline_xattr_size(inode);
1748
1749 f2fs_put_page(page, 1);
1750
1751 flags = FIEMAP_EXTENT_DATA_INLINE | FIEMAP_EXTENT_NOT_ALIGNED;
1752
1753 if (!xnid)
1754 flags |= FIEMAP_EXTENT_LAST;
1755
1756 err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
1757 if (err || err == 1)
1758 return err;
1759 }
1760
1761 if (xnid) {
1762 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), xnid, false);
1763 if (!page)
1764 return -ENOMEM;
1765
7735730d
CY		 1766 err = f2fs_get_node_info(sbi, xnid, &ni);
1767 if (err) {
1768 f2fs_put_page(page, 1);
1769 return err;
1770 }
442a9dbd
CY		 1771
1772 phys = (__u64)blk_to_logical(inode, ni.blk_addr);
1773 len = inode->i_sb->s_blocksize;
1774
1775 f2fs_put_page(page, 1);
1776
1777 flags = FIEMAP_EXTENT_LAST;
1778 }
1779
1780 if (phys)
1781 err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
1782
1783 return (err < 0 ? err : 0);
1784}
1785
9ab70134
JK		 1786int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1787 u64 start, u64 len)
1788{
7f63eb77
JK		 1789 struct buffer_head map_bh;
1790 sector_t start_blk, last_blk;
da85985c 1791 pgoff_t next_pgofs;
7f63eb77
JK		 1792 u64 logical = 0, phys = 0, size = 0;
1793 u32 flags = 0;
7f63eb77
JK		 1794 int ret = 0;
1795
c4020b2d
CY		 1796 if (fieinfo->fi_flags & FIEMAP_FLAG_CACHE) {
1797 ret = f2fs_precache_extents(inode);
1798 if (ret)
1799 return ret;
1800 }
1801
442a9dbd 1802 ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC | FIEMAP_FLAG_XATTR);
7f63eb77
JK		 1803 if (ret)
1804 return ret;
1805
f1b43d4c
CY		 1806 inode_lock(inode);
1807
442a9dbd
CY		 1808 if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
1809 ret = f2fs_xattr_fiemap(inode, fieinfo);
1810 goto out;
1811 }
1812
7975f349 1813 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
67f8cf3c
JK		 1814 ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len);
1815 if (ret != -EAGAIN)
f1b43d4c 1816 goto out;
67f8cf3c
JK		 1817 }
1818
7f63eb77
JK		 1819 if (logical_to_blk(inode, len) == 0)
1820 len = blk_to_logical(inode, 1);
1821
1822 start_blk = logical_to_blk(inode, start);
1823 last_blk = logical_to_blk(inode, start + len - 1);
9a950d52 1824
7f63eb77
JK		 1825next:
1826 memset(&map_bh, 0, sizeof(struct buffer_head));
1827 map_bh.b_size = len;
1828
e2b4e2bc 1829 ret = get_data_block(inode, start_blk, &map_bh, 0,
da85985c 1830 F2FS_GET_BLOCK_FIEMAP, &next_pgofs);
7f63eb77
JK		 1831 if (ret)
1832 goto out;
1833
1834 /* HOLE */
1835 if (!buffer_mapped(&map_bh)) {
da85985c 1836 start_blk = next_pgofs;
58736fa6
CY		 1837
1838 if (blk_to_logical(inode, start_blk) < blk_to_logical(inode,
1839 F2FS_I_SB(inode)->max_file_blocks))
9a950d52 1840 goto prep_next;
58736fa6 1841
9a950d52
FL		 1842 flags |= FIEMAP_EXTENT_LAST;
1843 }
7f63eb77 1844
da5af127 1845 if (size) {
62230e0d 1846 if (IS_ENCRYPTED(inode))
da5af127
CY		 1847 flags |= FIEMAP_EXTENT_DATA_ENCRYPTED;
1848
9a950d52
FL		 1849 ret = fiemap_fill_next_extent(fieinfo, logical,
1850 phys, size, flags);
da5af127 1851 }
7f63eb77 1852
9a950d52
FL		 1853 if (start_blk > last_blk || ret)
1854 goto out;
7f63eb77 1855
9a950d52
FL		 1856 logical = blk_to_logical(inode, start_blk);
1857 phys = blk_to_logical(inode, map_bh.b_blocknr);
1858 size = map_bh.b_size;
1859 flags = 0;
1860 if (buffer_unwritten(&map_bh))
1861 flags = FIEMAP_EXTENT_UNWRITTEN;
7f63eb77 1862
9a950d52 1863 start_blk += logical_to_blk(inode, size);
7f63eb77 1864
9a950d52 1865prep_next:
7f63eb77
JK		 1866 cond_resched();
1867 if (fatal_signal_pending(current))
1868 ret = -EINTR;
1869 else
1870 goto next;
1871out:
1872 if (ret == 1)
1873 ret = 0;
1874
5955102c 1875 inode_unlock(inode);
7f63eb77 1876 return ret;
9ab70134
JK		 1877}
1878
95ae251f
EB		 1879static inline loff_t f2fs_readpage_limit(struct inode *inode)
1880{
1881 if (IS_ENABLED(CONFIG_FS_VERITY) &&
1882 (IS_VERITY(inode) || f2fs_verity_in_progress(inode)))
1883 return inode->i_sb->s_maxbytes;
1884
1885 return i_size_read(inode);
1886}
1887
2df0ab04
CY		 1888static int f2fs_read_single_page(struct inode *inode, struct page *page,
1889 unsigned nr_pages,
1890 struct f2fs_map_blocks *map,
1891 struct bio **bio_ret,
1892 sector_t *last_block_in_bio,
1893 bool is_readahead)
1894{
1895 struct bio *bio = *bio_ret;
1896 const unsigned blkbits = inode->i_blkbits;
1897 const unsigned blocksize = 1 << blkbits;
1898 sector_t block_in_file;
1899 sector_t last_block;
1900 sector_t last_block_in_file;
1901 sector_t block_nr;
1902 int ret = 0;
1903
4969c06a 1904 block_in_file = (sector_t)page_index(page);
2df0ab04 1905 last_block = block_in_file + nr_pages;
95ae251f 1906 last_block_in_file = (f2fs_readpage_limit(inode) + blocksize - 1) >>
2df0ab04
CY		 1907 blkbits;
1908 if (last_block > last_block_in_file)
1909 last_block = last_block_in_file;
1910
1911 /* just zeroing out page which is beyond EOF */
1912 if (block_in_file >= last_block)
1913 goto zero_out;
1914 /*
1915 * Map blocks using the previous result first.
1916 */
1917 if ((map->m_flags & F2FS_MAP_MAPPED) &&
1918 block_in_file > map->m_lblk &&
1919 block_in_file < (map->m_lblk + map->m_len))
1920 goto got_it;
1921
1922 /*
1923 * Then do more f2fs_map_blocks() calls until we are
1924 * done with this page.
1925 */
1926 map->m_lblk = block_in_file;
1927 map->m_len = last_block - block_in_file;
1928
1929 ret = f2fs_map_blocks(inode, map, 0, F2FS_GET_BLOCK_DEFAULT);
1930 if (ret)
1931 goto out;
1932got_it:
1933 if ((map->m_flags & F2FS_MAP_MAPPED)) {
1934 block_nr = map->m_pblk + block_in_file - map->m_lblk;
1935 SetPageMappedToDisk(page);
1936
4969c06a
JK		 1937 if (!PageUptodate(page) && (!PageSwapCache(page) &&
1938 !cleancache_get_page(page))) {
2df0ab04
CY		 1939 SetPageUptodate(page);
1940 goto confused;
1941 }
1942
1943 if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), block_nr,
93770ab7 1944 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 1945 ret = -EFSCORRUPTED;
2df0ab04
CY		 1946 goto out;
1947 }
1948 } else {
1949zero_out:
1950 zero_user_segment(page, 0, PAGE_SIZE);
95ae251f
EB		 1951 if (f2fs_need_verity(inode, page->index) &&
1952 !fsverity_verify_page(page)) {
1953 ret = -EIO;
1954 goto out;
1955 }
2df0ab04
CY		 1956 if (!PageUptodate(page))
1957 SetPageUptodate(page);
1958 unlock_page(page);
1959 goto out;
1960 }
1961
1962 /*
1963 * This page will go to BIO. Do we need to send this
1964 * BIO off first?
1965 */
8896cbdf
CY		 1966 if (bio && !page_is_mergeable(F2FS_I_SB(inode), bio,
1967 *last_block_in_bio, block_nr)) {
2df0ab04
CY		 1968submit_and_realloc:
1969 __submit_bio(F2FS_I_SB(inode), bio, DATA);
1970 bio = NULL;
1971 }
1972 if (bio == NULL) {
1973 bio = f2fs_grab_read_bio(inode, block_nr, nr_pages,
95ae251f 1974 is_readahead ? REQ_RAHEAD : 0, page->index);
2df0ab04
CY		 1975 if (IS_ERR(bio)) {
1976 ret = PTR_ERR(bio);
1977 bio = NULL;
1978 goto out;
1979 }
1980 }
1981
1982 /*
1983 * If the page is under writeback, we need to wait for
1984 * its completion to see the correct decrypted data.
1985 */
1986 f2fs_wait_on_block_writeback(inode, block_nr);
1987
1988 if (bio_add_page(bio, page, blocksize, 0) < blocksize)
1989 goto submit_and_realloc;
1990
1991 inc_page_count(F2FS_I_SB(inode), F2FS_RD_DATA);
1992 ClearPageError(page);
1993 *last_block_in_bio = block_nr;
1994 goto out;
1995confused:
1996 if (bio) {
1997 __submit_bio(F2FS_I_SB(inode), bio, DATA);
1998 bio = NULL;
1999 }
2000 unlock_page(page);
2001out:
2002 *bio_ret = bio;
2003 return ret;
2004}
2005
4c8ff709
CY		 2006#ifdef CONFIG_F2FS_FS_COMPRESSION
2007int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret,
2008 unsigned nr_pages, sector_t *last_block_in_bio,
2009 bool is_readahead)
2010{
2011 struct dnode_of_data dn;
2012 struct inode *inode = cc->inode;
2013 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2014 struct bio *bio = *bio_ret;
2015 unsigned int start_idx = cc->cluster_idx << cc->log_cluster_size;
2016 sector_t last_block_in_file;
2017 const unsigned blkbits = inode->i_blkbits;
2018 const unsigned blocksize = 1 << blkbits;
2019 struct decompress_io_ctx *dic = NULL;
2020 int i;
2021 int ret = 0;
2022
2023 f2fs_bug_on(sbi, f2fs_cluster_is_empty(cc));
2024
2025 last_block_in_file = (i_size_read(inode) + blocksize - 1) >> blkbits;
2026
2027 /* get rid of pages beyond EOF */
2028 for (i = 0; i < cc->cluster_size; i++) {
2029 struct page *page = cc->rpages[i];
2030
2031 if (!page)
2032 continue;
2033 if ((sector_t)page->index >= last_block_in_file) {
2034 zero_user_segment(page, 0, PAGE_SIZE);
2035 if (!PageUptodate(page))
2036 SetPageUptodate(page);
2037 } else if (!PageUptodate(page)) {
2038 continue;
2039 }
2040 unlock_page(page);
2041 cc->rpages[i] = NULL;
2042 cc->nr_rpages--;
2043 }
2044
2045 /* we are done since all pages are beyond EOF */
2046 if (f2fs_cluster_is_empty(cc))
2047 goto out;
2048
2049 set_new_dnode(&dn, inode, NULL, NULL, 0);
2050 ret = f2fs_get_dnode_of_data(&dn, start_idx, LOOKUP_NODE);
2051 if (ret)
2052 goto out;
2053
2054 /* cluster was overwritten as normal cluster */
2055 if (dn.data_blkaddr != COMPRESS_ADDR)
2056 goto out;
2057
2058 for (i = 1; i < cc->cluster_size; i++) {
2059 block_t blkaddr;
2060
a2ced1ce 2061 blkaddr = data_blkaddr(dn.inode, dn.node_page,
4c8ff709
CY		 2062 dn.ofs_in_node + i);
2063
2064 if (!__is_valid_data_blkaddr(blkaddr))
2065 break;
2066
2067 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC)) {
2068 ret = -EFAULT;
2069 goto out_put_dnode;
2070 }
2071 cc->nr_cpages++;
2072 }
2073
2074 /* nothing to decompress */
2075 if (cc->nr_cpages == 0) {
2076 ret = 0;
2077 goto out_put_dnode;
2078 }
2079
2080 dic = f2fs_alloc_dic(cc);
2081 if (IS_ERR(dic)) {
2082 ret = PTR_ERR(dic);
2083 goto out_put_dnode;
2084 }
2085
2086 for (i = 0; i < dic->nr_cpages; i++) {
2087 struct page *page = dic->cpages[i];
2088 block_t blkaddr;
2089
a2ced1ce 2090 blkaddr = data_blkaddr(dn.inode, dn.node_page,
4c8ff709
CY		 2091 dn.ofs_in_node + i + 1);
2092
2093 if (bio && !page_is_mergeable(sbi, bio,
2094 *last_block_in_bio, blkaddr)) {
2095submit_and_realloc:
2096 __submit_bio(sbi, bio, DATA);
2097 bio = NULL;
2098 }
2099
2100 if (!bio) {
2101 bio = f2fs_grab_read_bio(inode, blkaddr, nr_pages,
2102 is_readahead ? REQ_RAHEAD : 0,
2103 page->index);
2104 if (IS_ERR(bio)) {
2105 ret = PTR_ERR(bio);
2106 bio = NULL;
2107 dic->failed = true;
2108 if (refcount_sub_and_test(dic->nr_cpages - i,
2109 &dic->ref))
2110 f2fs_decompress_end_io(dic->rpages,
2111 cc->cluster_size, true,
2112 false);
2113 f2fs_free_dic(dic);
2114 f2fs_put_dnode(&dn);
2115 *bio_ret = bio;
2116 return ret;
2117 }
2118 }
2119
2120 f2fs_wait_on_block_writeback(inode, blkaddr);
2121
2122 if (bio_add_page(bio, page, blocksize, 0) < blocksize)
2123 goto submit_and_realloc;
2124
2125 inc_page_count(sbi, F2FS_RD_DATA);
2126 ClearPageError(page);
2127 *last_block_in_bio = blkaddr;
2128 }
2129
2130 f2fs_put_dnode(&dn);
2131
2132 *bio_ret = bio;
2133 return 0;
2134
2135out_put_dnode:
2136 f2fs_put_dnode(&dn);
2137out:
2138 f2fs_decompress_end_io(cc->rpages, cc->cluster_size, true, false);
2139 *bio_ret = bio;
2140 return ret;
2141}
2142#endif
2143
f1e88660
JK		 2144/*
2145 * This function was originally taken from fs/mpage.c, and customized for f2fs.
2146 * Major change was from block_size == page_size in f2fs by default.
e2e59414
JK		 2147 *
2148 * Note that the aops->readpages() function is ONLY used for read-ahead. If
2149 * this function ever deviates from doing just read-ahead, it should either
2150 * use ->readpage() or do the necessary surgery to decouple ->readpages()
2151 * from read-ahead.
f1e88660 2152 */
4c8ff709 2153int f2fs_mpage_readpages(struct address_space *mapping,
f1e88660 2154 struct list_head *pages, struct page *page,
e2e59414 2155 unsigned nr_pages, bool is_readahead)
f1e88660
JK		 2156{
2157 struct bio *bio = NULL;
f1e88660
JK		 2158 sector_t last_block_in_bio = 0;
2159 struct inode *inode = mapping->host;
f1e88660 2160 struct f2fs_map_blocks map;
4c8ff709
CY		 2161#ifdef CONFIG_F2FS_FS_COMPRESSION
2162 struct compress_ctx cc = {
2163 .inode = inode,
2164 .log_cluster_size = F2FS_I(inode)->i_log_cluster_size,
2165 .cluster_size = F2FS_I(inode)->i_cluster_size,
2166 .cluster_idx = NULL_CLUSTER,
2167 .rpages = NULL,
2168 .cpages = NULL,
2169 .nr_rpages = 0,
2170 .nr_cpages = 0,
2171 };
2172#endif
2173 unsigned max_nr_pages = nr_pages;
2df0ab04 2174 int ret = 0;
f1e88660
JK		 2175
2176 map.m_pblk = 0;
2177 map.m_lblk = 0;
2178 map.m_len = 0;
2179 map.m_flags = 0;
da85985c 2180 map.m_next_pgofs = NULL;
c4020b2d 2181 map.m_next_extent = NULL;
d5097be5 2182 map.m_seg_type = NO_CHECK_TYPE;
f9d6d059 2183 map.m_may_create = false;
f1e88660 2184
736c0a74 2185 for (; nr_pages; nr_pages--) {
f1e88660 2186 if (pages) {
939afa94 2187 page = list_last_entry(pages, struct page, lru);
a83d50bc
KM		 2188
2189 prefetchw(&page->flags);
f1e88660
JK		 2190 list_del(&page->lru);
2191 if (add_to_page_cache_lru(page, mapping,
4969c06a 2192 page_index(page),
8a5c743e 2193 readahead_gfp_mask(mapping)))
f1e88660
JK		 2194 goto next_page;
2195 }
2196
4c8ff709
CY		 2197#ifdef CONFIG_F2FS_FS_COMPRESSION
2198 if (f2fs_compressed_file(inode)) {
2199 /* there are remained comressed pages, submit them */
2200 if (!f2fs_cluster_can_merge_page(&cc, page->index)) {
2201 ret = f2fs_read_multi_pages(&cc, &bio,
2202 max_nr_pages,
2203 &last_block_in_bio,
2204 is_readahead);
2205 f2fs_destroy_compress_ctx(&cc);
2206 if (ret)
2207 goto set_error_page;
2208 }
2209 ret = f2fs_is_compressed_cluster(inode, page->index);
2210 if (ret < 0)
2211 goto set_error_page;
2212 else if (!ret)
2213 goto read_single_page;
2214
2215 ret = f2fs_init_compress_ctx(&cc);
2216 if (ret)
2217 goto set_error_page;
2218
2219 f2fs_compress_ctx_add_page(&cc, page);
2220
2221 goto next_page;
2222 }
2223read_single_page:
2224#endif
2225
2226 ret = f2fs_read_single_page(inode, page, max_nr_pages, &map,
2227 &bio, &last_block_in_bio, is_readahead);
2df0ab04 2228 if (ret) {
4c8ff709
CY		 2229#ifdef CONFIG_F2FS_FS_COMPRESSION
2230set_error_page:
2231#endif
2df0ab04 2232 SetPageError(page);
09cbfeaf 2233 zero_user_segment(page, 0, PAGE_SIZE);
f1e88660 2234 unlock_page(page);
f1e88660 2235 }
f1e88660
JK		 2236next_page:
2237 if (pages)
09cbfeaf 2238 put_page(page);
4c8ff709
CY		 2239
2240#ifdef CONFIG_F2FS_FS_COMPRESSION
2241 if (f2fs_compressed_file(inode)) {
2242 /* last page */
2243 if (nr_pages == 1 && !f2fs_cluster_is_empty(&cc)) {
2244 ret = f2fs_read_multi_pages(&cc, &bio,
2245 max_nr_pages,
2246 &last_block_in_bio,
2247 is_readahead);
2248 f2fs_destroy_compress_ctx(&cc);
2249 }
2250 }
2251#endif
f1e88660
JK		 2252 }
2253 BUG_ON(pages && !list_empty(pages));
2254 if (bio)
4fc29c1a 2255 __submit_bio(F2FS_I_SB(inode), bio, DATA);
2df0ab04 2256 return pages ? 0 : ret;
f1e88660
JK		 2257}
2258
eb47b800
JK		 2259static int f2fs_read_data_page(struct file *file, struct page *page)
2260{
4969c06a 2261 struct inode *inode = page_file_mapping(page)->host;
b3d208f9 2262 int ret = -EAGAIN;
9ffe0fb5 2263
c20e89cd
CY		 2264 trace_f2fs_readpage(page, DATA);
2265
4c8ff709
CY		 2266 if (!f2fs_is_compress_backend_ready(inode)) {
2267 unlock_page(page);
2268 return -EOPNOTSUPP;
2269 }
2270
e1c42045 2271 /* If the file has inline data, try to read it directly */
9ffe0fb5
HL		 2272 if (f2fs_has_inline_data(inode))
2273 ret = f2fs_read_inline_data(inode, page);
b3d208f9 2274 if (ret == -EAGAIN)
4969c06a
JK		 2275 ret = f2fs_mpage_readpages(page_file_mapping(page),
2276 NULL, page, 1, false);
9ffe0fb5 2277 return ret;
eb47b800
JK		 2278}
2279
2280static int f2fs_read_data_pages(struct file *file,
2281 struct address_space *mapping,
2282 struct list_head *pages, unsigned nr_pages)
2283{
71cb4aff 2284 struct inode *inode = mapping->host;
939afa94 2285 struct page *page = list_last_entry(pages, struct page, lru);
b8c29400
CY		 2286
2287 trace_f2fs_readpages(inode, page, nr_pages);
9ffe0fb5 2288
4c8ff709
CY		 2289 if (!f2fs_is_compress_backend_ready(inode))
2290 return 0;
2291
9ffe0fb5
HL		 2292 /* If the file has inline data, skip readpages */
2293 if (f2fs_has_inline_data(inode))
2294 return 0;
2295
e2e59414 2296 return f2fs_mpage_readpages(mapping, pages, NULL, nr_pages, true);
eb47b800
JK		 2297}
2298
4c8ff709 2299int f2fs_encrypt_one_page(struct f2fs_io_info *fio)
7eab0c0d
HP		 2300{
2301 struct inode *inode = fio->page->mapping->host;
4c8ff709 2302 struct page *mpage, *page;
7eab0c0d
HP		 2303 gfp_t gfp_flags = GFP_NOFS;
2304
1958593e 2305 if (!f2fs_encrypted_file(inode))
7eab0c0d
HP		 2306 return 0;
2307
4c8ff709
CY		 2308 page = fio->compressed_page ? fio->compressed_page : fio->page;
2309
6dbb1796 2310 /* wait for GCed page writeback via META_MAPPING */
0ded69f6 2311 f2fs_wait_on_block_writeback(inode, fio->old_blkaddr);
7eab0c0d
HP		 2312
2313retry_encrypt:
4c8ff709
CY		 2314 fio->encrypted_page = fscrypt_encrypt_pagecache_blocks(page,
2315 PAGE_SIZE, 0, gfp_flags);
6aa58d8a
CY		 2316 if (IS_ERR(fio->encrypted_page)) {
2317 /* flush pending IOs and wait for a while in the ENOMEM case */
2318 if (PTR_ERR(fio->encrypted_page) == -ENOMEM) {
2319 f2fs_flush_merged_writes(fio->sbi);
5df7731f 2320 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
6aa58d8a
CY		 2321 gfp_flags |= __GFP_NOFAIL;
2322 goto retry_encrypt;
2323 }
2324 return PTR_ERR(fio->encrypted_page);
2325 }
7eab0c0d 2326
6aa58d8a
CY		 2327 mpage = find_lock_page(META_MAPPING(fio->sbi), fio->old_blkaddr);
2328 if (mpage) {
2329 if (PageUptodate(mpage))
2330 memcpy(page_address(mpage),
2331 page_address(fio->encrypted_page), PAGE_SIZE);
2332 f2fs_put_page(mpage, 1);
7eab0c0d 2333 }
6aa58d8a 2334 return 0;
7eab0c0d
HP		 2335}
2336
bb9e3bb8
CY		 2337static inline bool check_inplace_update_policy(struct inode *inode,
2338 struct f2fs_io_info *fio)
7eab0c0d 2339{
bb9e3bb8
CY		 2340 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2341 unsigned int policy = SM_I(sbi)->ipu_policy;
7eab0c0d 2342
bb9e3bb8
CY		 2343 if (policy & (0x1 << F2FS_IPU_FORCE))
2344 return true;
4d57b86d 2345 if (policy & (0x1 << F2FS_IPU_SSR) && f2fs_need_SSR(sbi))
bb9e3bb8
CY		 2346 return true;
2347 if (policy & (0x1 << F2FS_IPU_UTIL) &&
2348 utilization(sbi) > SM_I(sbi)->min_ipu_util)
2349 return true;
4d57b86d 2350 if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && f2fs_need_SSR(sbi) &&
bb9e3bb8
CY		 2351 utilization(sbi) > SM_I(sbi)->min_ipu_util)
2352 return true;
2353
2354 /*
2355 * IPU for rewrite async pages
2356 */
2357 if (policy & (0x1 << F2FS_IPU_ASYNC) &&
2358 fio && fio->op == REQ_OP_WRITE &&
2359 !(fio->op_flags & REQ_SYNC) &&
62230e0d 2360 !IS_ENCRYPTED(inode))
bb9e3bb8
CY		 2361 return true;
2362
2363 /* this is only set during fdatasync */
2364 if (policy & (0x1 << F2FS_IPU_FSYNC) &&
2365 is_inode_flag_set(inode, FI_NEED_IPU))
2366 return true;
2367
4354994f
DR		 2368 if (unlikely(fio && is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2369 !f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2370 return true;
2371
bb9e3bb8
CY		 2372 return false;
2373}
2374
4d57b86d 2375bool f2fs_should_update_inplace(struct inode *inode, struct f2fs_io_info *fio)
bb9e3bb8 2376{
1ad71a27
JK		 2377 if (f2fs_is_pinned_file(inode))
2378 return true;
bb9e3bb8
CY		 2379
2380 /* if this is cold file, we should overwrite to avoid fragmentation */
2381 if (file_is_cold(inode))
2382 return true;
2383
2384 return check_inplace_update_policy(inode, fio);
2385}
2386
4d57b86d 2387bool f2fs_should_update_outplace(struct inode *inode, struct f2fs_io_info *fio)
bb9e3bb8
CY		 2388{
2389 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2390
b0332a0f 2391 if (f2fs_lfs_mode(sbi))
bb9e3bb8
CY		 2392 return true;
2393 if (S_ISDIR(inode->i_mode))
2394 return true;
af033b2a
CY		 2395 if (IS_NOQUOTA(inode))
2396 return true;
bb9e3bb8
CY		 2397 if (f2fs_is_atomic_file(inode))
2398 return true;
2399 if (fio) {
2400 if (is_cold_data(fio->page))
2401 return true;
2402 if (IS_ATOMIC_WRITTEN_PAGE(fio->page))
2403 return true;
4354994f
DR		 2404 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2405 f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2406 return true;
bb9e3bb8
CY		 2407 }
2408 return false;
2409}
2410
7eab0c0d
HP		 2411static inline bool need_inplace_update(struct f2fs_io_info *fio)
2412{
2413 struct inode *inode = fio->page->mapping->host;
2414
4d57b86d 2415 if (f2fs_should_update_outplace(inode, fio))
7eab0c0d
HP		 2416 return false;
2417
4d57b86d 2418 return f2fs_should_update_inplace(inode, fio);
7eab0c0d
HP		 2419}
2420
4d57b86d 2421int f2fs_do_write_data_page(struct f2fs_io_info *fio)
eb47b800 2422{
05ca3632 2423 struct page *page = fio->page;
eb47b800 2424 struct inode *inode = page->mapping->host;
eb47b800 2425 struct dnode_of_data dn;
e959c8f5 2426 struct extent_info ei = {0,0,0};
7735730d 2427 struct node_info ni;
e959c8f5 2428 bool ipu_force = false;
eb47b800
JK		 2429 int err = 0;
2430
2431 set_new_dnode(&dn, inode, NULL, NULL, 0);
e959c8f5
HP		 2432 if (need_inplace_update(fio) &&
2433 f2fs_lookup_extent_cache(inode, page->index, &ei)) {
2434 fio->old_blkaddr = ei.blk + page->index - ei.fofs;
a817737e 2435
c9b60788 2436 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
93770ab7 2437 DATA_GENERIC_ENHANCE))
10f966bb 2438 return -EFSCORRUPTED;
c9b60788
CY		 2439
2440 ipu_force = true;
2441 fio->need_lock = LOCK_DONE;
2442 goto got_it;
e959c8f5 2443 }
279d6df2 2444
d29460e5
JK		 2445 /* Deadlock due to between page->lock and f2fs_lock_op */
2446 if (fio->need_lock == LOCK_REQ && !f2fs_trylock_op(fio->sbi))
2447 return -EAGAIN;
279d6df2 2448
4d57b86d 2449 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
eb47b800 2450 if (err)
279d6df2 2451 goto out;
eb47b800 2452
28bc106b 2453 fio->old_blkaddr = dn.data_blkaddr;
eb47b800
JK		 2454
2455 /* This page is already truncated */
7a9d7548 2456 if (fio->old_blkaddr == NULL_ADDR) {
2bca1e23 2457 ClearPageUptodate(page);
2baf0781 2458 clear_cold_data(page);
eb47b800 2459 goto out_writepage;
2bca1e23 2460 }
e959c8f5 2461got_it:
c9b60788
CY		 2462 if (__is_valid_data_blkaddr(fio->old_blkaddr) &&
2463 !f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
93770ab7 2464 DATA_GENERIC_ENHANCE)) {
10f966bb 2465 err = -EFSCORRUPTED;
c9b60788
CY		 2466 goto out_writepage;
2467 }
eb47b800
JK		 2468 /*
2469 * If current allocation needs SSR,
2470 * it had better in-place writes for updated data.
2471 */
93770ab7
CY		 2472 if (ipu_force ||
2473 (__is_valid_data_blkaddr(fio->old_blkaddr) &&
7b525dd0 2474 need_inplace_update(fio))) {
4c8ff709 2475 err = f2fs_encrypt_one_page(fio);
cc15620b
JK		 2476 if (err)
2477 goto out_writepage;
2478
2479 set_page_writeback(page);
17c50035 2480 ClearPageError(page);
279d6df2 2481 f2fs_put_dnode(&dn);
cc15620b 2482 if (fio->need_lock == LOCK_REQ)
279d6df2 2483 f2fs_unlock_op(fio->sbi);
4d57b86d 2484 err = f2fs_inplace_write_data(fio);
6492a335
CY		 2485 if (err) {
2486 if (f2fs_encrypted_file(inode))
d2d0727b 2487 fscrypt_finalize_bounce_page(&fio->encrypted_page);
6492a335
CY		 2488 if (PageWriteback(page))
2489 end_page_writeback(page);
cd23ffa9
CY		 2490 } else {
2491 set_inode_flag(inode, FI_UPDATE_WRITE);
6492a335 2492 }
7eab0c0d 2493 trace_f2fs_do_write_data_page(fio->page, IPU);
279d6df2 2494 return err;
eb47b800 2495 }
279d6df2 2496
cc15620b
JK		 2497 if (fio->need_lock == LOCK_RETRY) {
2498 if (!f2fs_trylock_op(fio->sbi)) {
2499 err = -EAGAIN;
2500 goto out_writepage;
2501 }
2502 fio->need_lock = LOCK_REQ;
2503 }
2504
7735730d
CY		 2505 err = f2fs_get_node_info(fio->sbi, dn.nid, &ni);
2506 if (err)
2507 goto out_writepage;
2508
2509 fio->version = ni.version;
2510
4c8ff709 2511 err = f2fs_encrypt_one_page(fio);
cc15620b
JK		 2512 if (err)
2513 goto out_writepage;
2514
2515 set_page_writeback(page);
17c50035 2516 ClearPageError(page);
cc15620b 2517
4c8ff709
CY		 2518 if (fio->compr_blocks && fio->old_blkaddr == COMPRESS_ADDR)
2519 f2fs_i_compr_blocks_update(inode, fio->compr_blocks - 1, false);
2520
279d6df2 2521 /* LFS mode write path */
4d57b86d 2522 f2fs_outplace_write_data(&dn, fio);
279d6df2
HP		 2523 trace_f2fs_do_write_data_page(page, OPU);
2524 set_inode_flag(inode, FI_APPEND_WRITE);
2525 if (page->index == 0)
2526 set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
eb47b800
JK		 2527out_writepage:
2528 f2fs_put_dnode(&dn);
279d6df2 2529out:
cc15620b 2530 if (fio->need_lock == LOCK_REQ)
279d6df2 2531 f2fs_unlock_op(fio->sbi);
eb47b800
JK		 2532 return err;
2533}
2534
4c8ff709 2535int f2fs_write_single_data_page(struct page *page, int *submitted,
8648de2c
CY		 2536 struct bio **bio,
2537 sector_t *last_block,
b0af6d49 2538 struct writeback_control *wbc,
4c8ff709
CY		 2539 enum iostat_type io_type,
2540 int compr_blocks)
eb47b800
JK		 2541{
2542 struct inode *inode = page->mapping->host;
4081363f 2543 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
eb47b800 2544 loff_t i_size = i_size_read(inode);
4c8ff709 2545 const pgoff_t end_index = ((unsigned long long)i_size)
09cbfeaf 2546 >> PAGE_SHIFT;
1f0d5c91 2547 loff_t psize = (loff_t)(page->index + 1) << PAGE_SHIFT;
9ffe0fb5 2548 unsigned offset = 0;
39936837 2549 bool need_balance_fs = false;
eb47b800 2550 int err = 0;
458e6197 2551 struct f2fs_io_info fio = {
05ca3632 2552 .sbi = sbi,
39d787be 2553 .ino = inode->i_ino,
458e6197 2554 .type = DATA,
04d328de 2555 .op = REQ_OP_WRITE,
7637241e 2556 .op_flags = wbc_to_write_flags(wbc),
e959c8f5 2557 .old_blkaddr = NULL_ADDR,
05ca3632 2558 .page = page,
4375a336 2559 .encrypted_page = NULL,
d68f735b 2560 .submitted = false,
4c8ff709 2561 .compr_blocks = compr_blocks,
cc15620b 2562 .need_lock = LOCK_RETRY,
b0af6d49 2563 .io_type = io_type,
578c6478 2564 .io_wbc = wbc,
8648de2c
CY		 2565 .bio = bio,
2566 .last_block = last_block,
458e6197 2567 };
eb47b800 2568
ecda0de3
CY		 2569 trace_f2fs_writepage(page, DATA);
2570
db198ae0
CY		 2571 /* we should bypass data pages to proceed the kworkder jobs */
2572 if (unlikely(f2fs_cp_error(sbi))) {
2573 mapping_set_error(page->mapping, -EIO);
1174abfd
CY		 2574 /*
2575 * don't drop any dirty dentry pages for keeping lastest
2576 * directory structure.
2577 */
2578 if (S_ISDIR(inode->i_mode))
2579 goto redirty_out;
db198ae0
CY		 2580 goto out;
2581 }
2582
0771fcc7
CY		 2583 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
2584 goto redirty_out;
2585
4c8ff709
CY		 2586 if (page->index < end_index ||
2587 f2fs_verity_in_progress(inode) ||
2588 compr_blocks)
39936837 2589 goto write;
eb47b800
JK		 2590
2591 /*
2592 * If the offset is out-of-range of file size,
2593 * this page does not have to be written to disk.
2594 */
09cbfeaf 2595 offset = i_size & (PAGE_SIZE - 1);
76f60268 2596 if ((page->index >= end_index + 1) || !offset)
39936837 2597 goto out;
eb47b800 2598
09cbfeaf 2599 zero_user_segment(page, offset, PAGE_SIZE);
39936837 2600write:
1e84371f
JK		 2601 if (f2fs_is_drop_cache(inode))
2602 goto out;
e6e5f561
JK		 2603 /* we should not write 0'th page having journal header */
2604 if (f2fs_is_volatile_file(inode) && (!page->index ||
2605 (!wbc->for_reclaim &&
4d57b86d 2606 f2fs_available_free_memory(sbi, BASE_CHECK))))
1e84371f 2607 goto redirty_out;
eb47b800 2608
39936837 2609 /* Dentry blocks are controlled by checkpoint */
eb47b800 2610 if (S_ISDIR(inode->i_mode)) {
cc15620b 2611 fio.need_lock = LOCK_DONE;
4d57b86d 2612 err = f2fs_do_write_data_page(&fio);
8618b881
JK		 2613 goto done;
2614 }
9ffe0fb5 2615
8618b881 2616 if (!wbc->for_reclaim)
39936837 2617 need_balance_fs = true;
7f3037a5 2618 else if (has_not_enough_free_secs(sbi, 0, 0))
39936837 2619 goto redirty_out;
ef095d19
JK		 2620 else
2621 set_inode_flag(inode, FI_HOT_DATA);
eb47b800 2622
b3d208f9 2623 err = -EAGAIN;
dd7b2333 2624 if (f2fs_has_inline_data(inode)) {
b3d208f9 2625 err = f2fs_write_inline_data(inode, page);
dd7b2333
YH		 2626 if (!err)
2627 goto out;
2628 }
279d6df2 2629
cc15620b 2630 if (err == -EAGAIN) {
4d57b86d 2631 err = f2fs_do_write_data_page(&fio);
cc15620b
JK		 2632 if (err == -EAGAIN) {
2633 fio.need_lock = LOCK_REQ;
4d57b86d 2634 err = f2fs_do_write_data_page(&fio);
cc15620b
JK		 2635 }
2636 }
a0d00fad 2637
eb449797
CY		 2638 if (err) {
2639 file_set_keep_isize(inode);
2640 } else {
c10c9820 2641 spin_lock(&F2FS_I(inode)->i_size_lock);
eb449797
CY		 2642 if (F2FS_I(inode)->last_disk_size < psize)
2643 F2FS_I(inode)->last_disk_size = psize;
c10c9820 2644 spin_unlock(&F2FS_I(inode)->i_size_lock);
eb449797 2645 }
279d6df2 2646
8618b881
JK		 2647done:
2648 if (err && err != -ENOENT)
2649 goto redirty_out;
eb47b800 2650
39936837 2651out:
a7ffdbe2 2652 inode_dec_dirty_pages(inode);
2baf0781 2653 if (err) {
2bca1e23 2654 ClearPageUptodate(page);
2baf0781
CY		 2655 clear_cold_data(page);
2656 }
0c3a5797
CY		 2657
2658 if (wbc->for_reclaim) {
bab475c5 2659 f2fs_submit_merged_write_cond(sbi, NULL, page, 0, DATA);
ef095d19 2660 clear_inode_flag(inode, FI_HOT_DATA);
4d57b86d 2661 f2fs_remove_dirty_inode(inode);
d68f735b 2662 submitted = NULL;
0c3a5797 2663 }
eb47b800 2664 unlock_page(page);
186857c5 2665 if (!S_ISDIR(inode->i_mode) && !IS_NOQUOTA(inode) &&
0b20fcec 2666 !F2FS_I(inode)->cp_task)
a7881893 2667 f2fs_balance_fs(sbi, need_balance_fs);
0c3a5797 2668
d68f735b 2669 if (unlikely(f2fs_cp_error(sbi))) {
b9109b0e 2670 f2fs_submit_merged_write(sbi, DATA);
0b20fcec 2671 f2fs_submit_merged_ipu_write(sbi, bio, NULL);
d68f735b
JK		 2672 submitted = NULL;
2673 }
2674
2675 if (submitted)
4c8ff709 2676 *submitted = fio.submitted ? 1 : 0;
0c3a5797 2677
eb47b800
JK		 2678 return 0;
2679
eb47b800 2680redirty_out:
76f60268 2681 redirty_page_for_writepage(wbc, page);
5b19d284
JK		 2682 /*
2683 * pageout() in MM traslates EAGAIN, so calls handle_write_error()
2684 * -> mapping_set_error() -> set_bit(AS_EIO, ...).
2685 * file_write_and_wait_range() will see EIO error, which is critical
2686 * to return value of fsync() followed by atomic_write failure to user.
2687 */
2688 if (!err || wbc->for_reclaim)
0002b61b 2689 return AOP_WRITEPAGE_ACTIVATE;
b230e6ca
JK		 2690 unlock_page(page);
2691 return err;
fa9150a8
NJ		 2692}
2693
f566bae8
JK		 2694static int f2fs_write_data_page(struct page *page,
2695 struct writeback_control *wbc)
2696{
4c8ff709
CY		 2697#ifdef CONFIG_F2FS_FS_COMPRESSION
2698 struct inode *inode = page->mapping->host;
2699
2700 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2701 goto out;
2702
2703 if (f2fs_compressed_file(inode)) {
2704 if (f2fs_is_compressed_cluster(inode, page->index)) {
2705 redirty_page_for_writepage(wbc, page);
2706 return AOP_WRITEPAGE_ACTIVATE;
2707 }
2708 }
2709out:
2710#endif
2711
2712 return f2fs_write_single_data_page(page, NULL, NULL, NULL,
2713 wbc, FS_DATA_IO, 0);
f566bae8
JK		 2714}
2715
8f46dcae
CY		 2716/*
2717 * This function was copied from write_cche_pages from mm/page-writeback.c.
2718 * The major change is making write step of cold data page separately from
2719 * warm/hot data page.
2720 */
2721static int f2fs_write_cache_pages(struct address_space *mapping,
b0af6d49
CY		 2722 struct writeback_control *wbc,
2723 enum iostat_type io_type)
8f46dcae
CY		 2724{
2725 int ret = 0;
4c8ff709 2726 int done = 0, retry = 0;
8f46dcae 2727 struct pagevec pvec;
c29fd0c0 2728 struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
8648de2c
CY		 2729 struct bio *bio = NULL;
2730 sector_t last_block;
4c8ff709
CY		 2731#ifdef CONFIG_F2FS_FS_COMPRESSION
2732 struct inode *inode = mapping->host;
2733 struct compress_ctx cc = {
2734 .inode = inode,
2735 .log_cluster_size = F2FS_I(inode)->i_log_cluster_size,
2736 .cluster_size = F2FS_I(inode)->i_cluster_size,
2737 .cluster_idx = NULL_CLUSTER,
2738 .rpages = NULL,
2739 .nr_rpages = 0,
2740 .cpages = NULL,
2741 .rbuf = NULL,
2742 .cbuf = NULL,
2743 .rlen = PAGE_SIZE * F2FS_I(inode)->i_cluster_size,
2744 .private = NULL,
2745 };
2746#endif
8f46dcae
CY		 2747 int nr_pages;
2748 pgoff_t uninitialized_var(writeback_index);
2749 pgoff_t index;
2750 pgoff_t end; /* Inclusive */
2751 pgoff_t done_index;
2752 int cycled;
2753 int range_whole = 0;
10bbd235 2754 xa_mark_t tag;
bab475c5 2755 int nwritten = 0;
4c8ff709
CY		 2756 int submitted = 0;
2757 int i;
8f46dcae 2758
86679820 2759 pagevec_init(&pvec);
46ae957f 2760
ef095d19
JK		 2761 if (get_dirty_pages(mapping->host) <=
2762 SM_I(F2FS_M_SB(mapping))->min_hot_blocks)
2763 set_inode_flag(mapping->host, FI_HOT_DATA);
2764 else
2765 clear_inode_flag(mapping->host, FI_HOT_DATA);
2766
8f46dcae
CY		 2767 if (wbc->range_cyclic) {
2768 writeback_index = mapping->writeback_index; /* prev offset */
2769 index = writeback_index;
2770 if (index == 0)
2771 cycled = 1;
2772 else
2773 cycled = 0;
2774 end = -1;
2775 } else {
09cbfeaf
KS		 2776 index = wbc->range_start >> PAGE_SHIFT;
2777 end = wbc->range_end >> PAGE_SHIFT;
8f46dcae
CY		 2778 if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
2779 range_whole = 1;
2780 cycled = 1; /* ignore range_cyclic tests */
2781 }
2782 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2783 tag = PAGECACHE_TAG_TOWRITE;
2784 else
2785 tag = PAGECACHE_TAG_DIRTY;
2786retry:
4c8ff709 2787 retry = 0;
8f46dcae
CY		 2788 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2789 tag_pages_for_writeback(mapping, index, end);
2790 done_index = index;
4c8ff709 2791 while (!done && !retry && (index <= end)) {
69c4f35d 2792 nr_pages = pagevec_lookup_range_tag(&pvec, mapping, &index, end,
67fd707f 2793 tag);
8f46dcae
CY		 2794 if (nr_pages == 0)
2795 break;
2796
2797 for (i = 0; i < nr_pages; i++) {
2798 struct page *page = pvec.pages[i];
4c8ff709
CY		 2799 bool need_readd;
2800readd:
2801 need_readd = false;
2802#ifdef CONFIG_F2FS_FS_COMPRESSION
2803 if (f2fs_compressed_file(inode)) {
2804 ret = f2fs_init_compress_ctx(&cc);
2805 if (ret) {
2806 done = 1;
2807 break;
2808 }
2809
2810 if (!f2fs_cluster_can_merge_page(&cc,
2811 page->index)) {
2812 ret = f2fs_write_multi_pages(&cc,
2813 &submitted, wbc, io_type);
2814 if (!ret)
2815 need_readd = true;
2816 goto result;
2817 }
8f46dcae 2818
4c8ff709
CY		 2819 if (unlikely(f2fs_cp_error(sbi)))
2820 goto lock_page;
2821
2822 if (f2fs_cluster_is_empty(&cc)) {
2823 void *fsdata = NULL;
2824 struct page *pagep;
2825 int ret2;
2826
2827 ret2 = f2fs_prepare_compress_overwrite(
2828 inode, &pagep,
2829 page->index, &fsdata);
2830 if (ret2 < 0) {
2831 ret = ret2;
2832 done = 1;
2833 break;
2834 } else if (ret2 &&
2835 !f2fs_compress_write_end(inode,
2836 fsdata, page->index,
2837 1)) {
2838 retry = 1;
2839 break;
2840 }
2841 } else {
2842 goto lock_page;
2843 }
2844 }
2845#endif
f8de4331 2846 /* give a priority to WB_SYNC threads */
c29fd0c0 2847 if (atomic_read(&sbi->wb_sync_req[DATA]) &&
f8de4331
CY		 2848 wbc->sync_mode == WB_SYNC_NONE) {
2849 done = 1;
2850 break;
2851 }
4c8ff709
CY		 2852#ifdef CONFIG_F2FS_FS_COMPRESSION
2853lock_page:
2854#endif
8f46dcae 2855 done_index = page->index;
d29460e5 2856retry_write:
8f46dcae
CY		 2857 lock_page(page);
2858
2859 if (unlikely(page->mapping != mapping)) {
2860continue_unlock:
2861 unlock_page(page);
2862 continue;
2863 }
2864
2865 if (!PageDirty(page)) {
2866 /* someone wrote it for us */
2867 goto continue_unlock;
2868 }
2869
8f46dcae 2870 if (PageWriteback(page)) {
0b20fcec 2871 if (wbc->sync_mode != WB_SYNC_NONE)
fec1d657 2872 f2fs_wait_on_page_writeback(page,
bae0ee7a 2873 DATA, true, true);
0b20fcec 2874 else
8f46dcae
CY		 2875 goto continue_unlock;
2876 }
2877
8f46dcae
CY		 2878 if (!clear_page_dirty_for_io(page))
2879 goto continue_unlock;
2880
4c8ff709
CY		 2881#ifdef CONFIG_F2FS_FS_COMPRESSION
2882 if (f2fs_compressed_file(inode)) {
2883 get_page(page);
2884 f2fs_compress_ctx_add_page(&cc, page);
2885 continue;
2886 }
2887#endif
2888 ret = f2fs_write_single_data_page(page, &submitted,
2889 &bio, &last_block, wbc, io_type, 0);
2890 if (ret == AOP_WRITEPAGE_ACTIVATE)
2891 unlock_page(page);
2892#ifdef CONFIG_F2FS_FS_COMPRESSION
2893result:
2894#endif
2895 nwritten += submitted;
2896 wbc->nr_to_write -= submitted;
2897
8f46dcae 2898 if (unlikely(ret)) {
0002b61b
CY		 2899 /*
2900 * keep nr_to_write, since vfs uses this to
2901 * get # of written pages.
2902 */
2903 if (ret == AOP_WRITEPAGE_ACTIVATE) {
0002b61b 2904 ret = 0;
4c8ff709 2905 goto next;
d29460e5
JK		 2906 } else if (ret == -EAGAIN) {
2907 ret = 0;
2908 if (wbc->sync_mode == WB_SYNC_ALL) {
2909 cond_resched();
2910 congestion_wait(BLK_RW_ASYNC,
5df7731f 2911 DEFAULT_IO_TIMEOUT);
d29460e5
JK		 2912 goto retry_write;
2913 }
4c8ff709 2914 goto next;
0002b61b 2915 }
b230e6ca
JK		 2916 done_index = page->index + 1;
2917 done = 1;
2918 break;
8f46dcae
CY		 2919 }
2920
4c8ff709 2921 if (wbc->nr_to_write <= 0 &&
687de7f1 2922 wbc->sync_mode == WB_SYNC_NONE) {
8f46dcae
CY		 2923 done = 1;
2924 break;
2925 }
4c8ff709
CY		 2926next:
2927 if (need_readd)
2928 goto readd;
8f46dcae
CY		 2929 }
2930 pagevec_release(&pvec);
2931 cond_resched();
2932 }
4c8ff709
CY		 2933#ifdef CONFIG_F2FS_FS_COMPRESSION
2934 /* flush remained pages in compress cluster */
2935 if (f2fs_compressed_file(inode) && !f2fs_cluster_is_empty(&cc)) {
2936 ret = f2fs_write_multi_pages(&cc, &submitted, wbc, io_type);
2937 nwritten += submitted;
2938 wbc->nr_to_write -= submitted;
2939 if (ret) {
2940 done = 1;
2941 retry = 0;
2942 }
2943 }
2944#endif
2945 if ((!cycled && !done) || retry) {
8f46dcae
CY		 2946 cycled = 1;
2947 index = 0;
2948 end = writeback_index - 1;
2949 goto retry;
2950 }
2951 if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
2952 mapping->writeback_index = done_index;
2953
bab475c5 2954 if (nwritten)
b9109b0e 2955 f2fs_submit_merged_write_cond(F2FS_M_SB(mapping), mapping->host,
bab475c5 2956 NULL, 0, DATA);
8648de2c
CY		 2957 /* submit cached bio of IPU write */
2958 if (bio)
0b20fcec 2959 f2fs_submit_merged_ipu_write(sbi, &bio, NULL);
6ca56ca4 2960
8f46dcae
CY		 2961 return ret;
2962}
2963
853137ce
JK		 2964static inline bool __should_serialize_io(struct inode *inode,
2965 struct writeback_control *wbc)
2966{
2967 if (!S_ISREG(inode->i_mode))
2968 return false;
4c8ff709
CY		 2969 if (f2fs_compressed_file(inode))
2970 return true;
af033b2a
CY		 2971 if (IS_NOQUOTA(inode))
2972 return false;
040d2bb3
CY		 2973 /* to avoid deadlock in path of data flush */
2974 if (F2FS_I(inode)->cp_task)
2975 return false;
853137ce
JK		 2976 if (wbc->sync_mode != WB_SYNC_ALL)
2977 return true;
2978 if (get_dirty_pages(inode) >= SM_I(F2FS_I_SB(inode))->min_seq_blocks)
2979 return true;
2980 return false;
2981}
2982
fc99fe27 2983static int __f2fs_write_data_pages(struct address_space *mapping,
b0af6d49
CY		 2984 struct writeback_control *wbc,
2985 enum iostat_type io_type)
eb47b800
JK		 2986{
2987 struct inode *inode = mapping->host;
4081363f 2988 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
9dfa1baf 2989 struct blk_plug plug;
eb47b800 2990 int ret;
853137ce 2991 bool locked = false;
eb47b800 2992
cfb185a1 2993 /* deal with chardevs and other special file */
2994 if (!mapping->a_ops->writepage)
2995 return 0;
2996
6a290544
CY		 2997 /* skip writing if there is no dirty page in this inode */
2998 if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE)
2999 return 0;
3000
0771fcc7
CY		 3001 /* during POR, we don't need to trigger writepage at all. */
3002 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
3003 goto skip_write;
3004
af033b2a
CY		 3005 if ((S_ISDIR(inode->i_mode) || IS_NOQUOTA(inode)) &&
3006 wbc->sync_mode == WB_SYNC_NONE &&
a1257023 3007 get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) &&
4d57b86d 3008 f2fs_available_free_memory(sbi, DIRTY_DENTS))
a1257023
JK		 3009 goto skip_write;
3010
d323d005 3011 /* skip writing during file defragment */
91942321 3012 if (is_inode_flag_set(inode, FI_DO_DEFRAG))
d323d005
CY		 3013 goto skip_write;
3014
d31c7c3f
YH		 3015 trace_f2fs_writepages(mapping->host, wbc, DATA);
3016
687de7f1
JK		 3017 /* to avoid spliting IOs due to mixed WB_SYNC_ALL and WB_SYNC_NONE */
3018 if (wbc->sync_mode == WB_SYNC_ALL)
c29fd0c0
CY		 3019 atomic_inc(&sbi->wb_sync_req[DATA]);
3020 else if (atomic_read(&sbi->wb_sync_req[DATA]))
687de7f1
JK		 3021 goto skip_write;
3022
853137ce
JK		 3023 if (__should_serialize_io(inode, wbc)) {
3024 mutex_lock(&sbi->writepages);
3025 locked = true;
3026 }
3027
9dfa1baf 3028 blk_start_plug(&plug);
b0af6d49 3029 ret = f2fs_write_cache_pages(mapping, wbc, io_type);
9dfa1baf 3030 blk_finish_plug(&plug);
687de7f1 3031
853137ce
JK		 3032 if (locked)
3033 mutex_unlock(&sbi->writepages);
3034
687de7f1 3035 if (wbc->sync_mode == WB_SYNC_ALL)
c29fd0c0 3036 atomic_dec(&sbi->wb_sync_req[DATA]);
28ea6162
JK		 3037 /*
3038 * if some pages were truncated, we cannot guarantee its mapping->host
3039 * to detect pending bios.
3040 */
458e6197 3041
4d57b86d 3042 f2fs_remove_dirty_inode(inode);
eb47b800 3043 return ret;
d3baf95d
JK		 3044
3045skip_write:
a7ffdbe2 3046 wbc->pages_skipped += get_dirty_pages(inode);
d31c7c3f 3047 trace_f2fs_writepages(mapping->host, wbc, DATA);
d3baf95d 3048 return 0;
eb47b800
JK		 3049}
3050
b0af6d49
CY		 3051static int f2fs_write_data_pages(struct address_space *mapping,
3052 struct writeback_control *wbc)
3053{
3054 struct inode *inode = mapping->host;
3055
3056 return __f2fs_write_data_pages(mapping, wbc,
3057 F2FS_I(inode)->cp_task == current ?
3058 FS_CP_DATA_IO : FS_DATA_IO);
3059}
3060
3aab8f82
CY		 3061static void f2fs_write_failed(struct address_space *mapping, loff_t to)
3062{
3063 struct inode *inode = mapping->host;
819d9153 3064 loff_t i_size = i_size_read(inode);
3aab8f82 3065
3f188c23
JK		 3066 if (IS_NOQUOTA(inode))
3067 return;
3068
95ae251f
EB		 3069 /* In the fs-verity case, f2fs_end_enable_verity() does the truncate */
3070 if (to > i_size && !f2fs_verity_in_progress(inode)) {
a33c1502 3071 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
6f8d4455 3072 down_write(&F2FS_I(inode)->i_mmap_sem);
a33c1502 3073
819d9153 3074 truncate_pagecache(inode, i_size);
3f188c23 3075 f2fs_truncate_blocks(inode, i_size, true);
a33c1502 3076
5a3a2d83 3077 up_write(&F2FS_I(inode)->i_mmap_sem);
6f8d4455 3078 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3aab8f82
CY		 3079 }
3080}
3081
2aadac08
JK		 3082static int prepare_write_begin(struct f2fs_sb_info *sbi,
3083 struct page *page, loff_t pos, unsigned len,
3084 block_t *blk_addr, bool *node_changed)
3085{
3086 struct inode *inode = page->mapping->host;
3087 pgoff_t index = page->index;
3088 struct dnode_of_data dn;
3089 struct page *ipage;
b4d07a3e 3090 bool locked = false;
e15882b6 3091 struct extent_info ei = {0,0,0};
2aadac08 3092 int err = 0;
2866fb16 3093 int flag;
2aadac08 3094
24b84912
JK		 3095 /*
3096 * we already allocated all the blocks, so we don't need to get
3097 * the block addresses when there is no need to fill the page.
3098 */
dc91de78 3099 if (!f2fs_has_inline_data(inode) && len == PAGE_SIZE &&
95ae251f
EB		 3100 !is_inode_flag_set(inode, FI_NO_PREALLOC) &&
3101 !f2fs_verity_in_progress(inode))
24b84912
JK		 3102 return 0;
3103
2866fb16
SY		 3104 /* f2fs_lock_op avoids race between write CP and convert_inline_page */
3105 if (f2fs_has_inline_data(inode) && pos + len > MAX_INLINE_DATA(inode))
3106 flag = F2FS_GET_BLOCK_DEFAULT;
3107 else
3108 flag = F2FS_GET_BLOCK_PRE_AIO;
3109
b4d07a3e 3110 if (f2fs_has_inline_data(inode) ||
09cbfeaf 3111 (pos & PAGE_MASK) >= i_size_read(inode)) {
2866fb16 3112 __do_map_lock(sbi, flag, true);
b4d07a3e
JK		 3113 locked = true;
3114 }
4c8ff709 3115
b4d07a3e 3116restart:
2aadac08 3117 /* check inline_data */
4d57b86d 3118 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2aadac08
JK		 3119 if (IS_ERR(ipage)) {
3120 err = PTR_ERR(ipage);
3121 goto unlock_out;
3122 }
3123
3124 set_new_dnode(&dn, inode, ipage, ipage, 0);
3125
3126 if (f2fs_has_inline_data(inode)) {
f2470371 3127 if (pos + len <= MAX_INLINE_DATA(inode)) {
4d57b86d 3128 f2fs_do_read_inline_data(page, ipage);
91942321 3129 set_inode_flag(inode, FI_DATA_EXIST);
ab47036d
CY		 3130 if (inode->i_nlink)
3131 set_inline_node(ipage);
2aadac08
JK		 3132 } else {
3133 err = f2fs_convert_inline_page(&dn, page);
3134 if (err)
b4d07a3e
JK		 3135 goto out;
3136 if (dn.data_blkaddr == NULL_ADDR)
3137 err = f2fs_get_block(&dn, index);
3138 }
3139 } else if (locked) {
3140 err = f2fs_get_block(&dn, index);
3141 } else {
3142 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
3143 dn.data_blkaddr = ei.blk + index - ei.fofs;
3144 } else {
b4d07a3e 3145 /* hole case */
4d57b86d 3146 err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
4da7bf5a 3147 if (err || dn.data_blkaddr == NULL_ADDR) {
b4d07a3e 3148 f2fs_put_dnode(&dn);
59c9081b
YH		 3149 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO,
3150 true);
2866fb16 3151 WARN_ON(flag != F2FS_GET_BLOCK_PRE_AIO);
b4d07a3e
JK		 3152 locked = true;
3153 goto restart;
3154 }
2aadac08
JK		 3155 }
3156 }
b4d07a3e 3157
2aadac08
JK		 3158 /* convert_inline_page can make node_changed */
3159 *blk_addr = dn.data_blkaddr;
3160 *node_changed = dn.node_changed;
b4d07a3e 3161out:
2aadac08
JK		 3162 f2fs_put_dnode(&dn);
3163unlock_out:
b4d07a3e 3164 if (locked)
2866fb16 3165 __do_map_lock(sbi, flag, false);
2aadac08
JK		 3166 return err;
3167}
3168
eb47b800
JK		 3169static int f2fs_write_begin(struct file *file, struct address_space *mapping,
3170 loff_t pos, unsigned len, unsigned flags,
3171 struct page **pagep, void **fsdata)
3172{
3173 struct inode *inode = mapping->host;
4081363f 3174 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
86531d6b 3175 struct page *page = NULL;
09cbfeaf 3176 pgoff_t index = ((unsigned long long) pos) >> PAGE_SHIFT;
a2e2e76b 3177 bool need_balance = false, drop_atomic = false;
2aadac08 3178 block_t blkaddr = NULL_ADDR;
eb47b800
JK		 3179 int err = 0;
3180
62aed044
CY		 3181 trace_f2fs_write_begin(inode, pos, len, flags);
3182
00e09c0b
CY		 3183 if (!f2fs_is_checkpoint_ready(sbi)) {
3184 err = -ENOSPC;
4354994f 3185 goto fail;
00e09c0b 3186 }
4354994f 3187
455e3a58
JK		 3188 if ((f2fs_is_atomic_file(inode) &&
3189 !f2fs_available_free_memory(sbi, INMEM_PAGES)) ||
3190 is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
57864ae5 3191 err = -ENOMEM;
a2e2e76b 3192 drop_atomic = true;
57864ae5
JK		 3193 goto fail;
3194 }
3195
5f727395
JK		 3196 /*
3197 * We should check this at this moment to avoid deadlock on inode page
3198 * and #0 page. The locking rule for inline_data conversion should be:
3199 * lock_page(page #0) -> lock_page(inode_page)
3200 */
3201 if (index != 0) {
3202 err = f2fs_convert_inline_inode(inode);
3203 if (err)
3204 goto fail;
3205 }
4c8ff709
CY		 3206
3207#ifdef CONFIG_F2FS_FS_COMPRESSION
3208 if (f2fs_compressed_file(inode)) {
3209 int ret;
3210
3211 *fsdata = NULL;
3212
3213 ret = f2fs_prepare_compress_overwrite(inode, pagep,
3214 index, fsdata);
3215 if (ret < 0) {
3216 err = ret;
3217 goto fail;
3218 } else if (ret) {
3219 return 0;
3220 }
3221 }
3222#endif
3223
afcb7ca0 3224repeat:
86d54795
JK		 3225 /*
3226 * Do not use grab_cache_page_write_begin() to avoid deadlock due to
3227 * wait_for_stable_page. Will wait that below with our IO control.
3228 */
01eccef7 3229 page = f2fs_pagecache_get_page(mapping, index,
86d54795 3230 FGP_LOCK | FGP_WRITE | FGP_CREAT, GFP_NOFS);
3aab8f82
CY		 3231 if (!page) {
3232 err = -ENOMEM;
3233 goto fail;
3234 }
d5f66990 3235
4c8ff709
CY		 3236 /* TODO: cluster can be compressed due to race with .writepage */
3237
eb47b800
JK		 3238 *pagep = page;
3239
2aadac08
JK		 3240 err = prepare_write_begin(sbi, page, pos, len,
3241 &blkaddr, &need_balance);
9ba69cf9 3242 if (err)
2aadac08 3243 goto fail;
9ba69cf9 3244
af033b2a
CY		 3245 if (need_balance && !IS_NOQUOTA(inode) &&
3246 has_not_enough_free_secs(sbi, 0, 0)) {
2a340760 3247 unlock_page(page);
2c4db1a6 3248 f2fs_balance_fs(sbi, true);
2a340760
JK		 3249 lock_page(page);
3250 if (page->mapping != mapping) {
3251 /* The page got truncated from under us */
3252 f2fs_put_page(page, 1);
3253 goto repeat;
3254 }
3255 }
3256
bae0ee7a 3257 f2fs_wait_on_page_writeback(page, DATA, false, true);
b3d208f9 3258
649d7df2
JK		 3259 if (len == PAGE_SIZE || PageUptodate(page))
3260 return 0;
eb47b800 3261
95ae251f
EB		 3262 if (!(pos & (PAGE_SIZE - 1)) && (pos + len) >= i_size_read(inode) &&
3263 !f2fs_verity_in_progress(inode)) {
746e2403
YH		 3264 zero_user_segment(page, len, PAGE_SIZE);
3265 return 0;
3266 }
3267
2aadac08 3268 if (blkaddr == NEW_ADDR) {
09cbfeaf 3269 zero_user_segment(page, 0, PAGE_SIZE);
649d7df2 3270 SetPageUptodate(page);
eb47b800 3271 } else {
93770ab7
CY		 3272 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3273 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 3274 err = -EFSCORRUPTED;
93770ab7
CY		 3275 goto fail;
3276 }
13ba41e3
JK		 3277 err = f2fs_submit_page_read(inode, page, blkaddr);
3278 if (err)
78682f79 3279 goto fail;
d54c795b 3280
393ff91f 3281 lock_page(page);
6bacf52f 3282 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 3283 f2fs_put_page(page, 1);
3284 goto repeat;
eb47b800 3285 }
1563ac75
CY		 3286 if (unlikely(!PageUptodate(page))) {
3287 err = -EIO;
3288 goto fail;
4375a336 3289 }
eb47b800 3290 }
eb47b800 3291 return 0;
9ba69cf9 3292
3aab8f82 3293fail:
86531d6b 3294 f2fs_put_page(page, 1);
3aab8f82 3295 f2fs_write_failed(mapping, pos + len);
a2e2e76b 3296 if (drop_atomic)
4d57b86d 3297 f2fs_drop_inmem_pages_all(sbi, false);
3aab8f82 3298 return err;
eb47b800
JK		 3299}
3300
a1dd3c13
JK		 3301static int f2fs_write_end(struct file *file,
3302 struct address_space *mapping,
3303 loff_t pos, unsigned len, unsigned copied,
3304 struct page *page, void *fsdata)
3305{
3306 struct inode *inode = page->mapping->host;
3307
dfb2bf38
CY		 3308 trace_f2fs_write_end(inode, pos, len, copied);
3309
649d7df2
JK		 3310 /*
3311 * This should be come from len == PAGE_SIZE, and we expect copied
3312 * should be PAGE_SIZE. Otherwise, we treat it with zero copied and
3313 * let generic_perform_write() try to copy data again through copied=0.
3314 */
3315 if (!PageUptodate(page)) {
746e2403 3316 if (unlikely(copied != len))
649d7df2
JK		 3317 copied = 0;
3318 else
3319 SetPageUptodate(page);
3320 }
4c8ff709
CY		 3321
3322#ifdef CONFIG_F2FS_FS_COMPRESSION
3323 /* overwrite compressed file */
3324 if (f2fs_compressed_file(inode) && fsdata) {
3325 f2fs_compress_write_end(inode, fsdata, page->index, copied);
3326 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3327 return copied;
3328 }
3329#endif
3330
649d7df2
JK		 3331 if (!copied)
3332 goto unlock_out;
3333
34ba94ba 3334 set_page_dirty(page);
a1dd3c13 3335
95ae251f
EB		 3336 if (pos + copied > i_size_read(inode) &&
3337 !f2fs_verity_in_progress(inode))
fc9581c8 3338 f2fs_i_size_write(inode, pos + copied);
649d7df2 3339unlock_out:
3024c9a1 3340 f2fs_put_page(page, 1);
d0239e1b 3341 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
a1dd3c13
JK		 3342 return copied;
3343}
3344
6f673763
OS		 3345static int check_direct_IO(struct inode *inode, struct iov_iter *iter,
3346 loff_t offset)
944fcfc1 3347{
8a56dd96
JK		 3348 unsigned i_blkbits = READ_ONCE(inode->i_blkbits);
3349 unsigned blkbits = i_blkbits;
3350 unsigned blocksize_mask = (1 << blkbits) - 1;
3351 unsigned long align = offset | iov_iter_alignment(iter);
3352 struct block_device *bdev = inode->i_sb->s_bdev;
3353
3354 if (align & blocksize_mask) {
3355 if (bdev)
3356 blkbits = blksize_bits(bdev_logical_block_size(bdev));
3357 blocksize_mask = (1 << blkbits) - 1;
3358 if (align & blocksize_mask)
3359 return -EINVAL;
3360 return 1;
3361 }
944fcfc1
JK		 3362 return 0;
3363}
3364
02b16d0a
CY		 3365static void f2fs_dio_end_io(struct bio *bio)
3366{
3367 struct f2fs_private_dio *dio = bio->bi_private;
3368
3369 dec_page_count(F2FS_I_SB(dio->inode),
3370 dio->write ? F2FS_DIO_WRITE : F2FS_DIO_READ);
3371
3372 bio->bi_private = dio->orig_private;
3373 bio->bi_end_io = dio->orig_end_io;
3374
5222595d 3375 kvfree(dio);
02b16d0a
CY		 3376
3377 bio_endio(bio);
3378}
3379
3380static void f2fs_dio_submit_bio(struct bio *bio, struct inode *inode,
3381 loff_t file_offset)
3382{
3383 struct f2fs_private_dio *dio;
3384 bool write = (bio_op(bio) == REQ_OP_WRITE);
02b16d0a
CY		 3385
3386 dio = f2fs_kzalloc(F2FS_I_SB(inode),
3387 sizeof(struct f2fs_private_dio), GFP_NOFS);
8e114038 3388 if (!dio)
02b16d0a 3389 goto out;
02b16d0a
CY		 3390
3391 dio->inode = inode;
3392 dio->orig_end_io = bio->bi_end_io;
3393 dio->orig_private = bio->bi_private;
3394 dio->write = write;
3395
3396 bio->bi_end_io = f2fs_dio_end_io;
3397 bio->bi_private = dio;
3398
3399 inc_page_count(F2FS_I_SB(inode),
3400 write ? F2FS_DIO_WRITE : F2FS_DIO_READ);
3401
3402 submit_bio(bio);
3403 return;
3404out:
3405 bio->bi_status = BLK_STS_IOERR;
3406 bio_endio(bio);
3407}
3408
c8b8e32d 3409static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
eb47b800 3410{
b439b103 3411 struct address_space *mapping = iocb->ki_filp->f_mapping;
3aab8f82 3412 struct inode *inode = mapping->host;
0cdd3195 3413 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
f847c699 3414 struct f2fs_inode_info *fi = F2FS_I(inode);
3aab8f82 3415 size_t count = iov_iter_count(iter);
c8b8e32d 3416 loff_t offset = iocb->ki_pos;
82e0a5aa 3417 int rw = iov_iter_rw(iter);
3aab8f82 3418 int err;
0cdd3195 3419 enum rw_hint hint = iocb->ki_hint;
63189b78 3420 int whint_mode = F2FS_OPTION(sbi).whint_mode;
f847c699 3421 bool do_opu;
944fcfc1 3422
b439b103 3423 err = check_direct_IO(inode, iter, offset);
b9d777b8 3424 if (err)
8a56dd96 3425 return err < 0 ? err : 0;
9ffe0fb5 3426
f847c699 3427 if (f2fs_force_buffered_io(inode, iocb, iter))
36abef4e 3428 return 0;
fcc85a4d 3429
f847c699
CY		 3430 do_opu = allow_outplace_dio(inode, iocb, iter);
3431
5302fb00 3432 trace_f2fs_direct_IO_enter(inode, offset, count, rw);
70407fad 3433
0cdd3195
HL		 3434 if (rw == WRITE && whint_mode == WHINT_MODE_OFF)
3435 iocb->ki_hint = WRITE_LIFE_NOT_SET;
3436
f847c699
CY		 3437 if (iocb->ki_flags & IOCB_NOWAIT) {
3438 if (!down_read_trylock(&fi->i_gc_rwsem[rw])) {
3439 iocb->ki_hint = hint;
3440 err = -EAGAIN;
3441 goto out;
3442 }
3443 if (do_opu && !down_read_trylock(&fi->i_gc_rwsem[READ])) {
3444 up_read(&fi->i_gc_rwsem[rw]);
b91050a8
HL		 3445 iocb->ki_hint = hint;
3446 err = -EAGAIN;
3447 goto out;
3448 }
f847c699
CY		 3449 } else {
3450 down_read(&fi->i_gc_rwsem[rw]);
3451 if (do_opu)
3452 down_read(&fi->i_gc_rwsem[READ]);
b91050a8
HL		 3453 }
3454
02b16d0a 3455 err = __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev,
f9d6d059
CY		 3456 iter, rw == WRITE ? get_data_block_dio_write :
3457 get_data_block_dio, NULL, f2fs_dio_submit_bio,
02b16d0a 3458 DIO_LOCKING | DIO_SKIP_HOLES);
f847c699
CY		 3459
3460 if (do_opu)
3461 up_read(&fi->i_gc_rwsem[READ]);
3462
3463 up_read(&fi->i_gc_rwsem[rw]);
82e0a5aa
CY		 3464
3465 if (rw == WRITE) {
0cdd3195
HL		 3466 if (whint_mode == WHINT_MODE_OFF)
3467 iocb->ki_hint = hint;
b0af6d49
CY		 3468 if (err > 0) {
3469 f2fs_update_iostat(F2FS_I_SB(inode), APP_DIRECT_IO,
3470 err);
f847c699
CY		 3471 if (!do_opu)
3472 set_inode_flag(inode, FI_UPDATE_WRITE);
b0af6d49 3473 } else if (err < 0) {
6bfc4919 3474 f2fs_write_failed(mapping, offset + count);
b0af6d49 3475 }
6bfc4919 3476 }
70407fad 3477
b91050a8 3478out:
5302fb00 3479 trace_f2fs_direct_IO_exit(inode, offset, count, rw, err);
70407fad 3480
3aab8f82 3481 return err;
eb47b800
JK		 3482}
3483
487261f3
CY		 3484void f2fs_invalidate_page(struct page *page, unsigned int offset,
3485 unsigned int length)
eb47b800
JK		 3486{
3487 struct inode *inode = page->mapping->host;
487261f3 3488 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
a7ffdbe2 3489
487261f3 3490 if (inode->i_ino >= F2FS_ROOT_INO(sbi) &&
09cbfeaf 3491 (offset % PAGE_SIZE || length != PAGE_SIZE))
a7ffdbe2
JK		 3492 return;
3493
487261f3 3494 if (PageDirty(page)) {
933439c8 3495 if (inode->i_ino == F2FS_META_INO(sbi)) {
487261f3 3496 dec_page_count(sbi, F2FS_DIRTY_META);
933439c8 3497 } else if (inode->i_ino == F2FS_NODE_INO(sbi)) {
487261f3 3498 dec_page_count(sbi, F2FS_DIRTY_NODES);
933439c8 3499 } else {
487261f3 3500 inode_dec_dirty_pages(inode);
4d57b86d 3501 f2fs_remove_dirty_inode(inode);
933439c8 3502 }
487261f3 3503 }
decd36b6 3504
2baf0781
CY		 3505 clear_cold_data(page);
3506
decd36b6 3507 if (IS_ATOMIC_WRITTEN_PAGE(page))
4d57b86d 3508 return f2fs_drop_inmem_page(inode, page);
decd36b6 3509
240a5915 3510 f2fs_clear_page_private(page);
eb47b800
JK		 3511}
3512
487261f3 3513int f2fs_release_page(struct page *page, gfp_t wait)
eb47b800 3514{
f68daeeb
JK		 3515 /* If this is dirty page, keep PagePrivate */
3516 if (PageDirty(page))
3517 return 0;
3518
decd36b6
CY		 3519 /* This is atomic written page, keep Private */
3520 if (IS_ATOMIC_WRITTEN_PAGE(page))
3521 return 0;
3522
2baf0781 3523 clear_cold_data(page);
240a5915 3524 f2fs_clear_page_private(page);
c3850aa1 3525 return 1;
eb47b800
JK		 3526}
3527
3528static int f2fs_set_data_page_dirty(struct page *page)
3529{
4969c06a 3530 struct inode *inode = page_file_mapping(page)->host;
eb47b800 3531
26c6b887
JK		 3532 trace_f2fs_set_page_dirty(page, DATA);
3533
237c0790
JK		 3534 if (!PageUptodate(page))
3535 SetPageUptodate(page);
4969c06a
JK		 3536 if (PageSwapCache(page))
3537 return __set_page_dirty_nobuffers(page);
34ba94ba 3538
5fe45743 3539 if (f2fs_is_atomic_file(inode) && !f2fs_is_commit_atomic_write(inode)) {
decd36b6 3540 if (!IS_ATOMIC_WRITTEN_PAGE(page)) {
4d57b86d 3541 f2fs_register_inmem_page(inode, page);
decd36b6
CY		 3542 return 1;
3543 }
3544 /*
3545 * Previously, this page has been registered, we just
3546 * return here.
3547 */
3548 return 0;
34ba94ba
JK		 3549 }
3550
eb47b800 3551 if (!PageDirty(page)) {
b87078ad 3552 __set_page_dirty_nobuffers(page);
4d57b86d 3553 f2fs_update_dirty_page(inode, page);
eb47b800
JK		 3554 return 1;
3555 }
3556 return 0;
3557}
3558
c01e54b7
JK		 3559static sector_t f2fs_bmap(struct address_space *mapping, sector_t block)
3560{
454ae7e5
CY		 3561 struct inode *inode = mapping->host;
3562
1d373a0e
JK		 3563 if (f2fs_has_inline_data(inode))
3564 return 0;
3565
3566 /* make sure allocating whole blocks */
3567 if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
3568 filemap_write_and_wait(mapping);
3569
e2b4e2bc 3570 return generic_block_bmap(mapping, block, get_data_block_bmap);
429511cd
CY		 3571}
3572
5b7a487c
WG		 3573#ifdef CONFIG_MIGRATION
3574#include <linux/migrate.h>
3575
3576int f2fs_migrate_page(struct address_space *mapping,
3577 struct page *newpage, struct page *page, enum migrate_mode mode)
3578{
3579 int rc, extra_count;
3580 struct f2fs_inode_info *fi = F2FS_I(mapping->host);
3581 bool atomic_written = IS_ATOMIC_WRITTEN_PAGE(page);
3582
3583 BUG_ON(PageWriteback(page));
3584
3585 /* migrating an atomic written page is safe with the inmem_lock hold */
ff1048e7
JK		 3586 if (atomic_written) {
3587 if (mode != MIGRATE_SYNC)
3588 return -EBUSY;
3589 if (!mutex_trylock(&fi->inmem_lock))
3590 return -EAGAIN;
3591 }
5b7a487c 3592
240a5915
CY		 3593 /* one extra reference was held for atomic_write page */
3594 extra_count = atomic_written ? 1 : 0;
5b7a487c 3595 rc = migrate_page_move_mapping(mapping, newpage,
37109694 3596 page, extra_count);
5b7a487c
WG		 3597 if (rc != MIGRATEPAGE_SUCCESS) {
3598 if (atomic_written)
3599 mutex_unlock(&fi->inmem_lock);
3600 return rc;
3601 }
3602
3603 if (atomic_written) {
3604 struct inmem_pages *cur;
3605 list_for_each_entry(cur, &fi->inmem_pages, list)
3606 if (cur->page == page) {
3607 cur->page = newpage;
3608 break;
3609 }
3610 mutex_unlock(&fi->inmem_lock);
3611 put_page(page);
3612 get_page(newpage);
3613 }
3614
240a5915
CY		 3615 if (PagePrivate(page)) {
3616 f2fs_set_page_private(newpage, page_private(page));
3617 f2fs_clear_page_private(page);
3618 }
5b7a487c 3619
2916ecc0
JG		 3620 if (mode != MIGRATE_SYNC_NO_COPY)
3621 migrate_page_copy(newpage, page);
3622 else
3623 migrate_page_states(newpage, page);
5b7a487c
WG		 3624
3625 return MIGRATEPAGE_SUCCESS;
3626}
3627#endif
3628
4969c06a
JK		 3629#ifdef CONFIG_SWAP
3630/* Copied from generic_swapfile_activate() to check any holes */
3e5e479a
CY		 3631static int check_swap_activate(struct swap_info_struct *sis,
3632 struct file *swap_file, sector_t *span)
4969c06a
JK		 3633{
3634 struct address_space *mapping = swap_file->f_mapping;
3635 struct inode *inode = mapping->host;
3636 unsigned blocks_per_page;
3637 unsigned long page_no;
3638 unsigned blkbits;
3639 sector_t probe_block;
3640 sector_t last_block;
3641 sector_t lowest_block = -1;
3642 sector_t highest_block = 0;
3e5e479a
CY		 3643 int nr_extents = 0;
3644 int ret;
4969c06a
JK		 3645
3646 blkbits = inode->i_blkbits;
3647 blocks_per_page = PAGE_SIZE >> blkbits;
3648
3649 /*
3650 * Map all the blocks into the extent list. This code doesn't try
3651 * to be very smart.
3652 */
3653 probe_block = 0;
3654 page_no = 0;
3655 last_block = i_size_read(inode) >> blkbits;
3e5e479a
CY		 3656 while ((probe_block + blocks_per_page) <= last_block &&
3657 page_no < sis->max) {
4969c06a
JK		 3658 unsigned block_in_page;
3659 sector_t first_block;
30460e1e
CM		 3660 sector_t block = 0;
3661 int err = 0;
4969c06a
JK		 3662
3663 cond_resched();
3664
30460e1e
CM		 3665 block = probe_block;
3666 err = bmap(inode, &block);
3667 if (err || !block)
4969c06a 3668 goto bad_bmap;
30460e1e 3669 first_block = block;
4969c06a
JK		 3670
3671 /*
3672 * It must be PAGE_SIZE aligned on-disk
3673 */
3674 if (first_block & (blocks_per_page - 1)) {
3675 probe_block++;
3676 goto reprobe;
3677 }
3678
3679 for (block_in_page = 1; block_in_page < blocks_per_page;
3680 block_in_page++) {
4969c06a 3681
30460e1e
CM		 3682 block = probe_block + block_in_page;
3683 err = bmap(inode, &block);
3684
3685 if (err || !block)
4969c06a 3686 goto bad_bmap;
30460e1e 3687
4969c06a
JK		 3688 if (block != first_block + block_in_page) {
3689 /* Discontiguity */
3690 probe_block++;
3691 goto reprobe;
3692 }
3693 }
3694
3695 first_block >>= (PAGE_SHIFT - blkbits);
3696 if (page_no) { /* exclude the header page */
3697 if (first_block < lowest_block)
3698 lowest_block = first_block;
3699 if (first_block > highest_block)
3700 highest_block = first_block;
3701 }
3702
3e5e479a
CY		 3703 /*
3704 * We found a PAGE_SIZE-length, PAGE_SIZE-aligned run of blocks
3705 */
3706 ret = add_swap_extent(sis, page_no, 1, first_block);
3707 if (ret < 0)
3708 goto out;
3709 nr_extents += ret;
4969c06a
JK		 3710 page_no++;
3711 probe_block += blocks_per_page;
3712reprobe:
3713 continue;
3714 }
3e5e479a
CY		 3715 ret = nr_extents;
3716 *span = 1 + highest_block - lowest_block;
3717 if (page_no == 0)
3718 page_no = 1; /* force Empty message */
3719 sis->max = page_no;
3720 sis->pages = page_no - 1;
3721 sis->highest_bit = page_no - 1;
3722out:
3723 return ret;
4969c06a
JK		 3724bad_bmap:
3725 pr_err("swapon: swapfile has holes\n");
3726 return -EINVAL;
3727}
3728
3729static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
3730 sector_t *span)
3731{
3732 struct inode *inode = file_inode(file);
3733 int ret;
3734
3735 if (!S_ISREG(inode->i_mode))
3736 return -EINVAL;
3737
3738 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3739 return -EROFS;
3740
3741 ret = f2fs_convert_inline_inode(inode);
3742 if (ret)
3743 return ret;
3744
4c8ff709
CY		 3745 if (f2fs_disable_compressed_file(inode))
3746 return -EINVAL;
3747
3e5e479a
CY		 3748 ret = check_swap_activate(sis, file, span);
3749 if (ret < 0)
4969c06a
JK		 3750 return ret;
3751
3752 set_inode_flag(inode, FI_PIN_FILE);
3753 f2fs_precache_extents(inode);
3754 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3e5e479a 3755 return ret;
4969c06a
JK		 3756}
3757
3758static void f2fs_swap_deactivate(struct file *file)
3759{
3760 struct inode *inode = file_inode(file);
3761
3762 clear_inode_flag(inode, FI_PIN_FILE);
3763}
3764#else
3765static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
3766 sector_t *span)
3767{
3768 return -EOPNOTSUPP;
3769}
3770
3771static void f2fs_swap_deactivate(struct file *file)
3772{
3773}
3774#endif
3775
eb47b800
JK		 3776const struct address_space_operations f2fs_dblock_aops = {
3777 .readpage = f2fs_read_data_page,
3778 .readpages = f2fs_read_data_pages,
3779 .writepage = f2fs_write_data_page,
3780 .writepages = f2fs_write_data_pages,
3781 .write_begin = f2fs_write_begin,
a1dd3c13 3782 .write_end = f2fs_write_end,
eb47b800 3783 .set_page_dirty = f2fs_set_data_page_dirty,
487261f3
CY		 3784 .invalidatepage = f2fs_invalidate_page,
3785 .releasepage = f2fs_release_page,
eb47b800 3786 .direct_IO = f2fs_direct_IO,
c01e54b7 3787 .bmap = f2fs_bmap,
4969c06a
JK		 3788 .swap_activate = f2fs_swap_activate,
3789 .swap_deactivate = f2fs_swap_deactivate,
5b7a487c
WG		 3790#ifdef CONFIG_MIGRATION
3791 .migratepage = f2fs_migrate_page,
3792#endif
eb47b800 3793};
6dbb1796 3794
5ec2d99d 3795void f2fs_clear_page_cache_dirty_tag(struct page *page)
aec2f729
CY		 3796{
3797 struct address_space *mapping = page_mapping(page);
3798 unsigned long flags;
3799
3800 xa_lock_irqsave(&mapping->i_pages, flags);
5ec2d99d 3801 __xa_clear_mark(&mapping->i_pages, page_index(page),
aec2f729
CY		 3802 PAGECACHE_TAG_DIRTY);
3803 xa_unlock_irqrestore(&mapping->i_pages, flags);
3804}
3805
6dbb1796
EB		 3806int __init f2fs_init_post_read_processing(void)
3807{
95ae251f
EB		 3808 bio_post_read_ctx_cache =
3809 kmem_cache_create("f2fs_bio_post_read_ctx",
3810 sizeof(struct bio_post_read_ctx), 0, 0, NULL);
6dbb1796
EB		 3811 if (!bio_post_read_ctx_cache)
3812 goto fail;
3813 bio_post_read_ctx_pool =
3814 mempool_create_slab_pool(NUM_PREALLOC_POST_READ_CTXS,
3815 bio_post_read_ctx_cache);
3816 if (!bio_post_read_ctx_pool)
3817 goto fail_free_cache;
3818 return 0;
3819
3820fail_free_cache:
3821 kmem_cache_destroy(bio_post_read_ctx_cache);
3822fail:
3823 return -ENOMEM;
3824}
3825
0b20fcec 3826void f2fs_destroy_post_read_processing(void)
6dbb1796
EB		 3827{
3828 mempool_destroy(bio_post_read_ctx_pool);
3829 kmem_cache_destroy(bio_post_read_ctx_cache);
3830}
0b20fcec 3831
4c8ff709
CY		 3832int f2fs_init_post_read_wq(struct f2fs_sb_info *sbi)
3833{
3834 if (!f2fs_sb_has_encrypt(sbi) &&
3835 !f2fs_sb_has_verity(sbi) &&
3836 !f2fs_sb_has_compression(sbi))
3837 return 0;
3838
3839 sbi->post_read_wq = alloc_workqueue("f2fs_post_read_wq",
3840 WQ_UNBOUND | WQ_HIGHPRI,
3841 num_online_cpus());
3842 if (!sbi->post_read_wq)
3843 return -ENOMEM;
3844 return 0;
3845}
3846
3847void f2fs_destroy_post_read_wq(struct f2fs_sb_info *sbi)
3848{
3849 if (sbi->post_read_wq)
3850 destroy_workqueue(sbi->post_read_wq);
3851}
3852
0b20fcec
CY		 3853int __init f2fs_init_bio_entry_cache(void)
3854{
3855 bio_entry_slab = f2fs_kmem_cache_create("bio_entry_slab",
3856 sizeof(struct bio_entry));
3857 if (!bio_entry_slab)
3858 return -ENOMEM;
3859 return 0;
3860}
3861
f543805f 3862void f2fs_destroy_bio_entry_cache(void)
0b20fcec
CY		 3863{
3864 kmem_cache_destroy(bio_entry_slab);
3865}
Michael's Linux tree.