]> 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: show mounted time
[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
66415cee 445 if (test_opt(sbi, LFS) && 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++) {
7a2af766
CY		 1047 block_t blkaddr = datablock_addr(dn->inode,
1048 dn->node_page, dn->ofs_in_node);
46008c6d
CY		 1049 if (blkaddr == NULL_ADDR) {
1050 dn->data_blkaddr = NEW_ADDR;
1051 __set_data_blkaddr(dn);
1052 count--;
1053 }
1054 }
1055
1056 if (set_page_dirty(dn->node_page))
1057 dn->node_changed = true;
eb47b800
JK		 1058 return 0;
1059}
1060
46008c6d 1061/* Should keep dn->ofs_in_node unchanged */
4d57b86d 1062int f2fs_reserve_new_block(struct dnode_of_data *dn)
46008c6d
CY		 1063{
1064 unsigned int ofs_in_node = dn->ofs_in_node;
1065 int ret;
1066
4d57b86d 1067 ret = f2fs_reserve_new_blocks(dn, 1);
46008c6d
CY		 1068 dn->ofs_in_node = ofs_in_node;
1069 return ret;
1070}
1071
b600965c
HL		 1072int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
1073{
1074 bool need_put = dn->inode_page ? false : true;
1075 int err;
1076
4d57b86d 1077 err = f2fs_get_dnode_of_data(dn, index, ALLOC_NODE);
b600965c
HL		 1078 if (err)
1079 return err;
a8865372 1080
b600965c 1081 if (dn->data_blkaddr == NULL_ADDR)
4d57b86d 1082 err = f2fs_reserve_new_block(dn);
a8865372 1083 if (err || need_put)
b600965c
HL		 1084 f2fs_put_dnode(dn);
1085 return err;
1086}
1087
759af1c9 1088int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index)
eb47b800 1089{
e15882b6 1090 struct extent_info ei = {0,0,0};
759af1c9 1091 struct inode *inode = dn->inode;
028a41e8 1092
759af1c9
FL		 1093 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
1094 dn->data_blkaddr = ei.blk + index - ei.fofs;
1095 return 0;
429511cd 1096 }
028a41e8 1097
759af1c9 1098 return f2fs_reserve_block(dn, index);
eb47b800
JK		 1099}
1100
4d57b86d 1101struct page *f2fs_get_read_data_page(struct inode *inode, pgoff_t index,
04d328de 1102 int op_flags, bool for_write)
eb47b800 1103{
eb47b800
JK		 1104 struct address_space *mapping = inode->i_mapping;
1105 struct dnode_of_data dn;
1106 struct page *page;
e15882b6 1107 struct extent_info ei = {0,0,0};
eb47b800 1108 int err;
4375a336 1109
a56c7c6f 1110 page = f2fs_grab_cache_page(mapping, index, for_write);
650495de
JK		 1111 if (!page)
1112 return ERR_PTR(-ENOMEM);
1113
cb3bc9ee
CY		 1114 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
1115 dn.data_blkaddr = ei.blk + index - ei.fofs;
93770ab7
CY		 1116 if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), dn.data_blkaddr,
1117 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 1118 err = -EFSCORRUPTED;
93770ab7
CY		 1119 goto put_err;
1120 }
cb3bc9ee
CY		 1121 goto got_it;
1122 }
1123
eb47b800 1124 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1125 err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
86531d6b
JK		 1126 if (err)
1127 goto put_err;
eb47b800
JK		 1128 f2fs_put_dnode(&dn);
1129
6bacf52f 1130 if (unlikely(dn.data_blkaddr == NULL_ADDR)) {
86531d6b
JK		 1131 err = -ENOENT;
1132 goto put_err;
650495de 1133 }
93770ab7
CY		 1134 if (dn.data_blkaddr != NEW_ADDR &&
1135 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
1136 dn.data_blkaddr,
1137 DATA_GENERIC_ENHANCE)) {
10f966bb 1138 err = -EFSCORRUPTED;
93770ab7
CY		 1139 goto put_err;
1140 }
cb3bc9ee 1141got_it:
43f3eae1
JK		 1142 if (PageUptodate(page)) {
1143 unlock_page(page);
eb47b800 1144 return page;
43f3eae1 1145 }
eb47b800 1146
d59ff4df
JK		 1147 /*
1148 * A new dentry page is allocated but not able to be written, since its
1149 * new inode page couldn't be allocated due to -ENOSPC.
1150 * In such the case, its blkaddr can be remained as NEW_ADDR.
4d57b86d
CY		 1151 * see, f2fs_add_link -> f2fs_get_new_data_page ->
1152 * f2fs_init_inode_metadata.
d59ff4df
JK		 1153 */
1154 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 1155 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 1156 if (!PageUptodate(page))
1157 SetPageUptodate(page);
43f3eae1 1158 unlock_page(page);
d59ff4df
JK		 1159 return page;
1160 }
eb47b800 1161
13ba41e3 1162 err = f2fs_submit_page_read(inode, page, dn.data_blkaddr);
393ff91f 1163 if (err)
86531d6b 1164 goto put_err;
43f3eae1 1165 return page;
86531d6b
JK		 1166
1167put_err:
1168 f2fs_put_page(page, 1);
1169 return ERR_PTR(err);
43f3eae1
JK		 1170}
1171
4d57b86d 1172struct page *f2fs_find_data_page(struct inode *inode, pgoff_t index)
43f3eae1
JK		 1173{
1174 struct address_space *mapping = inode->i_mapping;
1175 struct page *page;
1176
1177 page = find_get_page(mapping, index);
1178 if (page && PageUptodate(page))
1179 return page;
1180 f2fs_put_page(page, 0);
1181
4d57b86d 1182 page = f2fs_get_read_data_page(inode, index, 0, false);
43f3eae1
JK		 1183 if (IS_ERR(page))
1184 return page;
1185
1186 if (PageUptodate(page))
1187 return page;
1188
1189 wait_on_page_locked(page);
1190 if (unlikely(!PageUptodate(page))) {
1191 f2fs_put_page(page, 0);
1192 return ERR_PTR(-EIO);
1193 }
1194 return page;
1195}
1196
1197/*
1198 * If it tries to access a hole, return an error.
1199 * Because, the callers, functions in dir.c and GC, should be able to know
1200 * whether this page exists or not.
1201 */
4d57b86d 1202struct page *f2fs_get_lock_data_page(struct inode *inode, pgoff_t index,
a56c7c6f 1203 bool for_write)
43f3eae1
JK		 1204{
1205 struct address_space *mapping = inode->i_mapping;
1206 struct page *page;
1207repeat:
4d57b86d 1208 page = f2fs_get_read_data_page(inode, index, 0, for_write);
43f3eae1
JK		 1209 if (IS_ERR(page))
1210 return page;
393ff91f 1211
43f3eae1 1212 /* wait for read completion */
393ff91f 1213 lock_page(page);
6bacf52f 1214 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 1215 f2fs_put_page(page, 1);
1216 goto repeat;
eb47b800 1217 }
1563ac75
CY		 1218 if (unlikely(!PageUptodate(page))) {
1219 f2fs_put_page(page, 1);
1220 return ERR_PTR(-EIO);
1221 }
eb47b800
JK		 1222 return page;
1223}
1224
0a8165d7 1225/*
eb47b800
JK		 1226 * Caller ensures that this data page is never allocated.
1227 * A new zero-filled data page is allocated in the page cache.
39936837 1228 *
4f4124d0
CY		 1229 * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and
1230 * f2fs_unlock_op().
470f00e9
CY		 1231 * Note that, ipage is set only by make_empty_dir, and if any error occur,
1232 * ipage should be released by this function.
eb47b800 1233 */
4d57b86d 1234struct page *f2fs_get_new_data_page(struct inode *inode,
a8865372 1235 struct page *ipage, pgoff_t index, bool new_i_size)
eb47b800 1236{
eb47b800
JK		 1237 struct address_space *mapping = inode->i_mapping;
1238 struct page *page;
1239 struct dnode_of_data dn;
1240 int err;
7612118a 1241
a56c7c6f 1242 page = f2fs_grab_cache_page(mapping, index, true);
470f00e9
CY		 1243 if (!page) {
1244 /*
1245 * before exiting, we should make sure ipage will be released
1246 * if any error occur.
1247 */
1248 f2fs_put_page(ipage, 1);
01f28610 1249 return ERR_PTR(-ENOMEM);
470f00e9 1250 }
eb47b800 1251
a8865372 1252 set_new_dnode(&dn, inode, ipage, NULL, 0);
b600965c 1253 err = f2fs_reserve_block(&dn, index);
01f28610
JK		 1254 if (err) {
1255 f2fs_put_page(page, 1);
eb47b800 1256 return ERR_PTR(err);
a8865372 1257 }
01f28610
JK		 1258 if (!ipage)
1259 f2fs_put_dnode(&dn);
eb47b800
JK		 1260
1261 if (PageUptodate(page))
01f28610 1262 goto got_it;
eb47b800
JK		 1263
1264 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 1265 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 1266 if (!PageUptodate(page))
1267 SetPageUptodate(page);
eb47b800 1268 } else {
4375a336 1269 f2fs_put_page(page, 1);
a8865372 1270
7612118a
JK		 1271 /* if ipage exists, blkaddr should be NEW_ADDR */
1272 f2fs_bug_on(F2FS_I_SB(inode), ipage);
4d57b86d 1273 page = f2fs_get_lock_data_page(inode, index, true);
4375a336 1274 if (IS_ERR(page))
7612118a 1275 return page;
eb47b800 1276 }
01f28610 1277got_it:
9edcdabf 1278 if (new_i_size && i_size_read(inode) <
ee6d182f 1279 ((loff_t)(index + 1) << PAGE_SHIFT))
fc9581c8 1280 f2fs_i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT));
eb47b800
JK		 1281 return page;
1282}
1283
d5097be5 1284static int __allocate_data_block(struct dnode_of_data *dn, int seg_type)
bfad7c2d 1285{
4081363f 1286 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
bfad7c2d 1287 struct f2fs_summary sum;
bfad7c2d 1288 struct node_info ni;
6aa58d8a 1289 block_t old_blkaddr;
46008c6d 1290 blkcnt_t count = 1;
0abd675e 1291 int err;
bfad7c2d 1292
91942321 1293 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
bfad7c2d 1294 return -EPERM;
df6136ef 1295
7735730d
CY		 1296 err = f2fs_get_node_info(sbi, dn->nid, &ni);
1297 if (err)
1298 return err;
1299
7a2af766
CY		 1300 dn->data_blkaddr = datablock_addr(dn->inode,
1301 dn->node_page, dn->ofs_in_node);
f847c699 1302 if (dn->data_blkaddr != NULL_ADDR)
df6136ef
CY		 1303 goto alloc;
1304
0abd675e
CY		 1305 if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
1306 return err;
bfad7c2d 1307
df6136ef 1308alloc:
bfad7c2d 1309 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
6aa58d8a
CY		 1310 old_blkaddr = dn->data_blkaddr;
1311 f2fs_allocate_data_block(sbi, NULL, old_blkaddr, &dn->data_blkaddr,
d5097be5 1312 &sum, seg_type, NULL, false);
6aa58d8a
CY		 1313 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
1314 invalidate_mapping_pages(META_MAPPING(sbi),
1315 old_blkaddr, old_blkaddr);
86f35dc3 1316 f2fs_update_data_blkaddr(dn, dn->data_blkaddr);
bfad7c2d 1317
0a4daae5
JK		 1318 /*
1319 * i_size will be updated by direct_IO. Otherwise, we'll get stale
1320 * data from unwritten block via dio_read.
1321 */
bfad7c2d
JK		 1322 return 0;
1323}
1324
a7de6086 1325int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *from)
59b802e5 1326{
b439b103 1327 struct inode *inode = file_inode(iocb->ki_filp);
5b8db7fa 1328 struct f2fs_map_blocks map;
d6d478a1 1329 int flag;
a7de6086 1330 int err = 0;
d6d478a1 1331 bool direct_io = iocb->ki_flags & IOCB_DIRECT;
59b802e5 1332
0080c507 1333 map.m_lblk = F2FS_BLK_ALIGN(iocb->ki_pos);
dfd02e4d
CY		 1334 map.m_len = F2FS_BYTES_TO_BLK(iocb->ki_pos + iov_iter_count(from));
1335 if (map.m_len > map.m_lblk)
1336 map.m_len -= map.m_lblk;
1337 else
1338 map.m_len = 0;
1339
da85985c 1340 map.m_next_pgofs = NULL;
c4020b2d 1341 map.m_next_extent = NULL;
d5097be5 1342 map.m_seg_type = NO_CHECK_TYPE;
f9d6d059 1343 map.m_may_create = true;
2a340760 1344
d6d478a1 1345 if (direct_io) {
4d57b86d 1346 map.m_seg_type = f2fs_rw_hint_to_seg_type(iocb->ki_hint);
f847c699 1347 flag = f2fs_force_buffered_io(inode, iocb, from) ?
d6d478a1
CY		 1348 F2FS_GET_BLOCK_PRE_AIO :
1349 F2FS_GET_BLOCK_PRE_DIO;
1350 goto map_blocks;
d5097be5 1351 }
f2470371 1352 if (iocb->ki_pos + iov_iter_count(from) > MAX_INLINE_DATA(inode)) {
a7de6086
JK		 1353 err = f2fs_convert_inline_inode(inode);
1354 if (err)
1355 return err;
b439b103 1356 }
d6d478a1 1357 if (f2fs_has_inline_data(inode))
25006645 1358 return err;
d6d478a1
CY		 1359
1360 flag = F2FS_GET_BLOCK_PRE_AIO;
1361
1362map_blocks:
1363 err = f2fs_map_blocks(inode, &map, 1, flag);
1364 if (map.m_len > 0 && err == -ENOSPC) {
1365 if (!direct_io)
1366 set_inode_flag(inode, FI_NO_PREALLOC);
1367 err = 0;
25006645 1368 }
a7de6086 1369 return err;
59b802e5
JK		 1370}
1371
39a86958 1372void __do_map_lock(struct f2fs_sb_info *sbi, int flag, bool lock)
59c9081b
YH		 1373{
1374 if (flag == F2FS_GET_BLOCK_PRE_AIO) {
1375 if (lock)
1376 down_read(&sbi->node_change);
1377 else
1378 up_read(&sbi->node_change);
1379 } else {
1380 if (lock)
1381 f2fs_lock_op(sbi);
1382 else
1383 f2fs_unlock_op(sbi);
1384 }
1385}
1386
0a8165d7 1387/*
7a88ddb5
CY		 1388 * f2fs_map_blocks() tries to find or build mapping relationship which
1389 * maps continuous logical blocks to physical blocks, and return such
1390 * info via f2fs_map_blocks structure.
eb47b800 1391 */
d323d005 1392int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
e2b4e2bc 1393 int create, int flag)
eb47b800 1394{
003a3e1d 1395 unsigned int maxblocks = map->m_len;
eb47b800 1396 struct dnode_of_data dn;
f9811703 1397 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
f9d6d059 1398 int mode = map->m_may_create ? ALLOC_NODE : LOOKUP_NODE;
46008c6d 1399 pgoff_t pgofs, end_offset, end;
bfad7c2d 1400 int err = 0, ofs = 1;
46008c6d
CY		 1401 unsigned int ofs_in_node, last_ofs_in_node;
1402 blkcnt_t prealloc;
e15882b6 1403 struct extent_info ei = {0,0,0};
7df3a431 1404 block_t blkaddr;
c4020b2d 1405 unsigned int start_pgofs;
eb47b800 1406
dfd02e4d
CY		 1407 if (!maxblocks)
1408 return 0;
1409
003a3e1d
JK		 1410 map->m_len = 0;
1411 map->m_flags = 0;
1412
1413 /* it only supports block size == page size */
1414 pgofs = (pgoff_t)map->m_lblk;
46008c6d 1415 end = pgofs + maxblocks;
eb47b800 1416
24b84912 1417 if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) {
f4f0b677
JZ		 1418 if (test_opt(sbi, LFS) && flag == F2FS_GET_BLOCK_DIO &&
1419 map->m_may_create)
1420 goto next_dnode;
1421
003a3e1d
JK		 1422 map->m_pblk = ei.blk + pgofs - ei.fofs;
1423 map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs);
1424 map->m_flags = F2FS_MAP_MAPPED;
c4020b2d
CY		 1425 if (map->m_next_extent)
1426 *map->m_next_extent = pgofs + map->m_len;
1e78e8bd
ST		 1427
1428 /* for hardware encryption, but to avoid potential issue in future */
1429 if (flag == F2FS_GET_BLOCK_DIO)
1430 f2fs_wait_on_block_writeback_range(inode,
1431 map->m_pblk, map->m_len);
bfad7c2d 1432 goto out;
a2e7d1bf 1433 }
bfad7c2d 1434
4fe71e88 1435next_dnode:
f9d6d059 1436 if (map->m_may_create)
59c9081b 1437 __do_map_lock(sbi, flag, true);
eb47b800
JK		 1438
1439 /* When reading holes, we need its node page */
1440 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1441 err = f2fs_get_dnode_of_data(&dn, pgofs, mode);
1ec79083 1442 if (err) {
43473f96
CY		 1443 if (flag == F2FS_GET_BLOCK_BMAP)
1444 map->m_pblk = 0;
da85985c 1445 if (err == -ENOENT) {
bfad7c2d 1446 err = 0;
da85985c
CY		 1447 if (map->m_next_pgofs)
1448 *map->m_next_pgofs =
4d57b86d 1449 f2fs_get_next_page_offset(&dn, pgofs);
c4020b2d
CY		 1450 if (map->m_next_extent)
1451 *map->m_next_extent =
4d57b86d 1452 f2fs_get_next_page_offset(&dn, pgofs);
da85985c 1453 }
bfad7c2d 1454 goto unlock_out;
848753aa 1455 }
973163fc 1456
c4020b2d 1457 start_pgofs = pgofs;
46008c6d 1458 prealloc = 0;
230436b3 1459 last_ofs_in_node = ofs_in_node = dn.ofs_in_node;
81ca7350 1460 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
4fe71e88
CY		 1461
1462next_block:
7a2af766 1463 blkaddr = datablock_addr(dn.inode, dn.node_page, dn.ofs_in_node);
4fe71e88 1464
c9b60788 1465 if (__is_valid_data_blkaddr(blkaddr) &&
93770ab7 1466 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC_ENHANCE)) {
10f966bb 1467 err = -EFSCORRUPTED;
c9b60788
CY		 1468 goto sync_out;
1469 }
1470
93770ab7 1471 if (__is_valid_data_blkaddr(blkaddr)) {
f847c699 1472 /* use out-place-update for driect IO under LFS mode */
f9d6d059
CY		 1473 if (test_opt(sbi, LFS) && flag == F2FS_GET_BLOCK_DIO &&
1474 map->m_may_create) {
f847c699 1475 err = __allocate_data_block(&dn, map->m_seg_type);
05e36006
CY		 1476 if (err)
1477 goto sync_out;
1478 blkaddr = dn.data_blkaddr;
1479 set_inode_flag(inode, FI_APPEND_WRITE);
f847c699
CY		 1480 }
1481 } else {
973163fc 1482 if (create) {
f9811703
CY		 1483 if (unlikely(f2fs_cp_error(sbi))) {
1484 err = -EIO;
4fe71e88 1485 goto sync_out;
f9811703 1486 }
24b84912 1487 if (flag == F2FS_GET_BLOCK_PRE_AIO) {
46008c6d
CY		 1488 if (blkaddr == NULL_ADDR) {
1489 prealloc++;
1490 last_ofs_in_node = dn.ofs_in_node;
1491 }
24b84912 1492 } else {
0a4daae5
JK		 1493 WARN_ON(flag != F2FS_GET_BLOCK_PRE_DIO &&
1494 flag != F2FS_GET_BLOCK_DIO);
d5097be5
HL		 1495 err = __allocate_data_block(&dn,
1496 map->m_seg_type);
6f2d8ed6 1497 if (!err)
91942321 1498 set_inode_flag(inode, FI_APPEND_WRITE);
24b84912 1499 }
973163fc 1500 if (err)
4fe71e88 1501 goto sync_out;
3f2be043 1502 map->m_flags |= F2FS_MAP_NEW;
4fe71e88 1503 blkaddr = dn.data_blkaddr;
973163fc 1504 } else {
43473f96
CY		 1505 if (flag == F2FS_GET_BLOCK_BMAP) {
1506 map->m_pblk = 0;
1507 goto sync_out;
1508 }
c4020b2d
CY		 1509 if (flag == F2FS_GET_BLOCK_PRECACHE)
1510 goto sync_out;
da85985c
CY		 1511 if (flag == F2FS_GET_BLOCK_FIEMAP &&
1512 blkaddr == NULL_ADDR) {
1513 if (map->m_next_pgofs)
1514 *map->m_next_pgofs = pgofs + 1;
4c2ac6a8 1515 goto sync_out;
da85985c 1516 }
f3d98e74
CY		 1517 if (flag != F2FS_GET_BLOCK_FIEMAP) {
1518 /* for defragment case */
1519 if (map->m_next_pgofs)
1520 *map->m_next_pgofs = pgofs + 1;
4fe71e88 1521 goto sync_out;
f3d98e74 1522 }
e2b4e2bc 1523 }
e2b4e2bc 1524 }
eb47b800 1525
46008c6d
CY		 1526 if (flag == F2FS_GET_BLOCK_PRE_AIO)
1527 goto skip;
1528
4fe71e88
CY		 1529 if (map->m_len == 0) {
1530 /* preallocated unwritten block should be mapped for fiemap. */
1531 if (blkaddr == NEW_ADDR)
1532 map->m_flags |= F2FS_MAP_UNWRITTEN;
1533 map->m_flags |= F2FS_MAP_MAPPED;
1534
1535 map->m_pblk = blkaddr;
1536 map->m_len = 1;
1537 } else if ((map->m_pblk != NEW_ADDR &&
1538 blkaddr == (map->m_pblk + ofs)) ||
b439b103 1539 (map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR) ||
46008c6d 1540 flag == F2FS_GET_BLOCK_PRE_DIO) {
4fe71e88
CY		 1541 ofs++;
1542 map->m_len++;
1543 } else {
1544 goto sync_out;
1545 }
bfad7c2d 1546
46008c6d 1547skip:
bfad7c2d
JK		 1548 dn.ofs_in_node++;
1549 pgofs++;
1550
46008c6d
CY		 1551 /* preallocate blocks in batch for one dnode page */
1552 if (flag == F2FS_GET_BLOCK_PRE_AIO &&
1553 (pgofs == end || dn.ofs_in_node == end_offset)) {
7df3a431 1554
46008c6d 1555 dn.ofs_in_node = ofs_in_node;
4d57b86d 1556 err = f2fs_reserve_new_blocks(&dn, prealloc);
46008c6d
CY		 1557 if (err)
1558 goto sync_out;
bfad7c2d 1559
46008c6d
CY		 1560 map->m_len += dn.ofs_in_node - ofs_in_node;
1561 if (prealloc && dn.ofs_in_node != last_ofs_in_node + 1) {
1562 err = -ENOSPC;
1563 goto sync_out;
3104af35 1564 }
46008c6d
CY		 1565 dn.ofs_in_node = end_offset;
1566 }
1567
1568 if (pgofs >= end)
1569 goto sync_out;
1570 else if (dn.ofs_in_node < end_offset)
1571 goto next_block;
1572
c4020b2d
CY		 1573 if (flag == F2FS_GET_BLOCK_PRECACHE) {
1574 if (map->m_flags & F2FS_MAP_MAPPED) {
1575 unsigned int ofs = start_pgofs - map->m_lblk;
1576
1577 f2fs_update_extent_cache_range(&dn,
1578 start_pgofs, map->m_pblk + ofs,
1579 map->m_len - ofs);
1580 }
1581 }
1582
46008c6d
CY		 1583 f2fs_put_dnode(&dn);
1584
f9d6d059 1585 if (map->m_may_create) {
59c9081b 1586 __do_map_lock(sbi, flag, false);
6f2d8ed6 1587 f2fs_balance_fs(sbi, dn.node_changed);
eb47b800 1588 }
46008c6d 1589 goto next_dnode;
7df3a431 1590
bfad7c2d 1591sync_out:
1e78e8bd
ST		 1592
1593 /* for hardware encryption, but to avoid potential issue in future */
1594 if (flag == F2FS_GET_BLOCK_DIO && map->m_flags & F2FS_MAP_MAPPED)
1595 f2fs_wait_on_block_writeback_range(inode,
1596 map->m_pblk, map->m_len);
1597
c4020b2d
CY		 1598 if (flag == F2FS_GET_BLOCK_PRECACHE) {
1599 if (map->m_flags & F2FS_MAP_MAPPED) {
1600 unsigned int ofs = start_pgofs - map->m_lblk;
1601
1602 f2fs_update_extent_cache_range(&dn,
1603 start_pgofs, map->m_pblk + ofs,
1604 map->m_len - ofs);
1605 }
1606 if (map->m_next_extent)
1607 *map->m_next_extent = pgofs + 1;
1608 }
eb47b800 1609 f2fs_put_dnode(&dn);
bfad7c2d 1610unlock_out:
f9d6d059 1611 if (map->m_may_create) {
59c9081b 1612 __do_map_lock(sbi, flag, false);
6f2d8ed6 1613 f2fs_balance_fs(sbi, dn.node_changed);
2a340760 1614 }
bfad7c2d 1615out:
003a3e1d 1616 trace_f2fs_map_blocks(inode, map, err);
bfad7c2d 1617 return err;
eb47b800
JK		 1618}
1619
b91050a8
HL		 1620bool f2fs_overwrite_io(struct inode *inode, loff_t pos, size_t len)
1621{
1622 struct f2fs_map_blocks map;
1623 block_t last_lblk;
1624 int err;
1625
1626 if (pos + len > i_size_read(inode))
1627 return false;
1628
1629 map.m_lblk = F2FS_BYTES_TO_BLK(pos);
1630 map.m_next_pgofs = NULL;
1631 map.m_next_extent = NULL;
1632 map.m_seg_type = NO_CHECK_TYPE;
f4f0b677 1633 map.m_may_create = false;
b91050a8
HL		 1634 last_lblk = F2FS_BLK_ALIGN(pos + len);
1635
1636 while (map.m_lblk < last_lblk) {
1637 map.m_len = last_lblk - map.m_lblk;
1638 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
1639 if (err || map.m_len == 0)
1640 return false;
1641 map.m_lblk += map.m_len;
1642 }
1643 return true;
1644}
1645
003a3e1d 1646static int __get_data_block(struct inode *inode, sector_t iblock,
da85985c 1647 struct buffer_head *bh, int create, int flag,
f9d6d059 1648 pgoff_t *next_pgofs, int seg_type, bool may_write)
003a3e1d
JK		 1649{
1650 struct f2fs_map_blocks map;
a7de6086 1651 int err;
003a3e1d
JK		 1652
1653 map.m_lblk = iblock;
1654 map.m_len = bh->b_size >> inode->i_blkbits;
da85985c 1655 map.m_next_pgofs = next_pgofs;
c4020b2d 1656 map.m_next_extent = NULL;
d5097be5 1657 map.m_seg_type = seg_type;
f9d6d059 1658 map.m_may_create = may_write;
003a3e1d 1659
a7de6086
JK		 1660 err = f2fs_map_blocks(inode, &map, create, flag);
1661 if (!err) {
003a3e1d
JK		 1662 map_bh(bh, inode->i_sb, map.m_pblk);
1663 bh->b_state = (bh->b_state & ~F2FS_MAP_FLAGS) | map.m_flags;
b86e3307 1664 bh->b_size = (u64)map.m_len << inode->i_blkbits;
003a3e1d 1665 }
a7de6086 1666 return err;
003a3e1d
JK		 1667}
1668
ccfb3000 1669static int get_data_block(struct inode *inode, sector_t iblock,
da85985c
CY		 1670 struct buffer_head *bh_result, int create, int flag,
1671 pgoff_t *next_pgofs)
e2b4e2bc 1672{
da85985c 1673 return __get_data_block(inode, iblock, bh_result, create,
d5097be5 1674 flag, next_pgofs,
f9d6d059
CY		 1675 NO_CHECK_TYPE, create);
1676}
1677
1678static int get_data_block_dio_write(struct inode *inode, sector_t iblock,
1679 struct buffer_head *bh_result, int create)
1680{
1681 return __get_data_block(inode, iblock, bh_result, create,
1682 F2FS_GET_BLOCK_DIO, NULL,
1683 f2fs_rw_hint_to_seg_type(inode->i_write_hint),
75a037f3 1684 IS_SWAPFILE(inode) ? false : true);
e2b4e2bc
CY		 1685}
1686
1687static int get_data_block_dio(struct inode *inode, sector_t iblock,
ccfb3000
JK		 1688 struct buffer_head *bh_result, int create)
1689{
e2b4e2bc 1690 return __get_data_block(inode, iblock, bh_result, create,
f9d6d059
CY		 1691 F2FS_GET_BLOCK_DIO, NULL,
1692 f2fs_rw_hint_to_seg_type(inode->i_write_hint),
1693 false);
ccfb3000
JK		 1694}
1695
e2b4e2bc 1696static int get_data_block_bmap(struct inode *inode, sector_t iblock,
ccfb3000
JK		 1697 struct buffer_head *bh_result, int create)
1698{
179448bf 1699 /* Block number less than F2FS MAX BLOCKS */
e0afc4d6 1700 if (unlikely(iblock >= F2FS_I_SB(inode)->max_file_blocks))
179448bf
YH		 1701 return -EFBIG;
1702
e2b4e2bc 1703 return __get_data_block(inode, iblock, bh_result, create,
d5097be5 1704 F2FS_GET_BLOCK_BMAP, NULL,
f9d6d059 1705 NO_CHECK_TYPE, create);
ccfb3000
JK		 1706}
1707
7f63eb77
JK		 1708static inline sector_t logical_to_blk(struct inode *inode, loff_t offset)
1709{
1710 return (offset >> inode->i_blkbits);
1711}
1712
1713static inline loff_t blk_to_logical(struct inode *inode, sector_t blk)
1714{
1715 return (blk << inode->i_blkbits);
1716}
1717
442a9dbd
CY		 1718static int f2fs_xattr_fiemap(struct inode *inode,
1719 struct fiemap_extent_info *fieinfo)
1720{
1721 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1722 struct page *page;
1723 struct node_info ni;
1724 __u64 phys = 0, len;
1725 __u32 flags;
1726 nid_t xnid = F2FS_I(inode)->i_xattr_nid;
1727 int err = 0;
1728
1729 if (f2fs_has_inline_xattr(inode)) {
1730 int offset;
1731
1732 page = f2fs_grab_cache_page(NODE_MAPPING(sbi),
1733 inode->i_ino, false);
1734 if (!page)
1735 return -ENOMEM;
1736
7735730d
CY		 1737 err = f2fs_get_node_info(sbi, inode->i_ino, &ni);
1738 if (err) {
1739 f2fs_put_page(page, 1);
1740 return err;
1741 }
442a9dbd
CY		 1742
1743 phys = (__u64)blk_to_logical(inode, ni.blk_addr);
1744 offset = offsetof(struct f2fs_inode, i_addr) +
1745 sizeof(__le32) * (DEF_ADDRS_PER_INODE -
b323fd28 1746 get_inline_xattr_addrs(inode));
442a9dbd
CY		 1747
1748 phys += offset;
1749 len = inline_xattr_size(inode);
1750
1751 f2fs_put_page(page, 1);
1752
1753 flags = FIEMAP_EXTENT_DATA_INLINE | FIEMAP_EXTENT_NOT_ALIGNED;
1754
1755 if (!xnid)
1756 flags |= FIEMAP_EXTENT_LAST;
1757
1758 err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
1759 if (err || err == 1)
1760 return err;
1761 }
1762
1763 if (xnid) {
1764 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), xnid, false);
1765 if (!page)
1766 return -ENOMEM;
1767
7735730d
CY		 1768 err = f2fs_get_node_info(sbi, xnid, &ni);
1769 if (err) {
1770 f2fs_put_page(page, 1);
1771 return err;
1772 }
442a9dbd
CY		 1773
1774 phys = (__u64)blk_to_logical(inode, ni.blk_addr);
1775 len = inode->i_sb->s_blocksize;
1776
1777 f2fs_put_page(page, 1);
1778
1779 flags = FIEMAP_EXTENT_LAST;
1780 }
1781
1782 if (phys)
1783 err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
1784
1785 return (err < 0 ? err : 0);
1786}
1787
9ab70134
JK		 1788int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1789 u64 start, u64 len)
1790{
7f63eb77
JK		 1791 struct buffer_head map_bh;
1792 sector_t start_blk, last_blk;
da85985c 1793 pgoff_t next_pgofs;
7f63eb77
JK		 1794 u64 logical = 0, phys = 0, size = 0;
1795 u32 flags = 0;
7f63eb77
JK		 1796 int ret = 0;
1797
c4020b2d
CY		 1798 if (fieinfo->fi_flags & FIEMAP_FLAG_CACHE) {
1799 ret = f2fs_precache_extents(inode);
1800 if (ret)
1801 return ret;
1802 }
1803
442a9dbd 1804 ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC | FIEMAP_FLAG_XATTR);
7f63eb77
JK		 1805 if (ret)
1806 return ret;
1807
f1b43d4c
CY		 1808 inode_lock(inode);
1809
442a9dbd
CY		 1810 if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
1811 ret = f2fs_xattr_fiemap(inode, fieinfo);
1812 goto out;
1813 }
1814
7975f349 1815 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
67f8cf3c
JK		 1816 ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len);
1817 if (ret != -EAGAIN)
f1b43d4c 1818 goto out;
67f8cf3c
JK		 1819 }
1820
7f63eb77
JK		 1821 if (logical_to_blk(inode, len) == 0)
1822 len = blk_to_logical(inode, 1);
1823
1824 start_blk = logical_to_blk(inode, start);
1825 last_blk = logical_to_blk(inode, start + len - 1);
9a950d52 1826
7f63eb77
JK		 1827next:
1828 memset(&map_bh, 0, sizeof(struct buffer_head));
1829 map_bh.b_size = len;
1830
e2b4e2bc 1831 ret = get_data_block(inode, start_blk, &map_bh, 0,
da85985c 1832 F2FS_GET_BLOCK_FIEMAP, &next_pgofs);
7f63eb77
JK		 1833 if (ret)
1834 goto out;
1835
1836 /* HOLE */
1837 if (!buffer_mapped(&map_bh)) {
da85985c 1838 start_blk = next_pgofs;
58736fa6
CY		 1839
1840 if (blk_to_logical(inode, start_blk) < blk_to_logical(inode,
1841 F2FS_I_SB(inode)->max_file_blocks))
9a950d52 1842 goto prep_next;
58736fa6 1843
9a950d52
FL		 1844 flags |= FIEMAP_EXTENT_LAST;
1845 }
7f63eb77 1846
da5af127 1847 if (size) {
62230e0d 1848 if (IS_ENCRYPTED(inode))
da5af127
CY		 1849 flags |= FIEMAP_EXTENT_DATA_ENCRYPTED;
1850
9a950d52
FL		 1851 ret = fiemap_fill_next_extent(fieinfo, logical,
1852 phys, size, flags);
da5af127 1853 }
7f63eb77 1854
9a950d52
FL		 1855 if (start_blk > last_blk || ret)
1856 goto out;
7f63eb77 1857
9a950d52
FL		 1858 logical = blk_to_logical(inode, start_blk);
1859 phys = blk_to_logical(inode, map_bh.b_blocknr);
1860 size = map_bh.b_size;
1861 flags = 0;
1862 if (buffer_unwritten(&map_bh))
1863 flags = FIEMAP_EXTENT_UNWRITTEN;
7f63eb77 1864
9a950d52 1865 start_blk += logical_to_blk(inode, size);
7f63eb77 1866
9a950d52 1867prep_next:
7f63eb77
JK		 1868 cond_resched();
1869 if (fatal_signal_pending(current))
1870 ret = -EINTR;
1871 else
1872 goto next;
1873out:
1874 if (ret == 1)
1875 ret = 0;
1876
5955102c 1877 inode_unlock(inode);
7f63eb77 1878 return ret;
9ab70134
JK		 1879}
1880
95ae251f
EB		 1881static inline loff_t f2fs_readpage_limit(struct inode *inode)
1882{
1883 if (IS_ENABLED(CONFIG_FS_VERITY) &&
1884 (IS_VERITY(inode) || f2fs_verity_in_progress(inode)))
1885 return inode->i_sb->s_maxbytes;
1886
1887 return i_size_read(inode);
1888}
1889
2df0ab04
CY		 1890static int f2fs_read_single_page(struct inode *inode, struct page *page,
1891 unsigned nr_pages,
1892 struct f2fs_map_blocks *map,
1893 struct bio **bio_ret,
1894 sector_t *last_block_in_bio,
1895 bool is_readahead)
1896{
1897 struct bio *bio = *bio_ret;
1898 const unsigned blkbits = inode->i_blkbits;
1899 const unsigned blocksize = 1 << blkbits;
1900 sector_t block_in_file;
1901 sector_t last_block;
1902 sector_t last_block_in_file;
1903 sector_t block_nr;
1904 int ret = 0;
1905
4969c06a 1906 block_in_file = (sector_t)page_index(page);
2df0ab04 1907 last_block = block_in_file + nr_pages;
95ae251f 1908 last_block_in_file = (f2fs_readpage_limit(inode) + blocksize - 1) >>
2df0ab04
CY		 1909 blkbits;
1910 if (last_block > last_block_in_file)
1911 last_block = last_block_in_file;
1912
1913 /* just zeroing out page which is beyond EOF */
1914 if (block_in_file >= last_block)
1915 goto zero_out;
1916 /*
1917 * Map blocks using the previous result first.
1918 */
1919 if ((map->m_flags & F2FS_MAP_MAPPED) &&
1920 block_in_file > map->m_lblk &&
1921 block_in_file < (map->m_lblk + map->m_len))
1922 goto got_it;
1923
1924 /*
1925 * Then do more f2fs_map_blocks() calls until we are
1926 * done with this page.
1927 */
1928 map->m_lblk = block_in_file;
1929 map->m_len = last_block - block_in_file;
1930
1931 ret = f2fs_map_blocks(inode, map, 0, F2FS_GET_BLOCK_DEFAULT);
1932 if (ret)
1933 goto out;
1934got_it:
1935 if ((map->m_flags & F2FS_MAP_MAPPED)) {
1936 block_nr = map->m_pblk + block_in_file - map->m_lblk;
1937 SetPageMappedToDisk(page);
1938
4969c06a
JK		 1939 if (!PageUptodate(page) && (!PageSwapCache(page) &&
1940 !cleancache_get_page(page))) {
2df0ab04
CY		 1941 SetPageUptodate(page);
1942 goto confused;
1943 }
1944
1945 if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), block_nr,
93770ab7 1946 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 1947 ret = -EFSCORRUPTED;
2df0ab04
CY		 1948 goto out;
1949 }
1950 } else {
1951zero_out:
1952 zero_user_segment(page, 0, PAGE_SIZE);
95ae251f
EB		 1953 if (f2fs_need_verity(inode, page->index) &&
1954 !fsverity_verify_page(page)) {
1955 ret = -EIO;
1956 goto out;
1957 }
2df0ab04
CY		 1958 if (!PageUptodate(page))
1959 SetPageUptodate(page);
1960 unlock_page(page);
1961 goto out;
1962 }
1963
1964 /*
1965 * This page will go to BIO. Do we need to send this
1966 * BIO off first?
1967 */
8896cbdf
CY		 1968 if (bio && !page_is_mergeable(F2FS_I_SB(inode), bio,
1969 *last_block_in_bio, block_nr)) {
2df0ab04
CY		 1970submit_and_realloc:
1971 __submit_bio(F2FS_I_SB(inode), bio, DATA);
1972 bio = NULL;
1973 }
1974 if (bio == NULL) {
1975 bio = f2fs_grab_read_bio(inode, block_nr, nr_pages,
95ae251f 1976 is_readahead ? REQ_RAHEAD : 0, page->index);
2df0ab04
CY		 1977 if (IS_ERR(bio)) {
1978 ret = PTR_ERR(bio);
1979 bio = NULL;
1980 goto out;
1981 }
1982 }
1983
1984 /*
1985 * If the page is under writeback, we need to wait for
1986 * its completion to see the correct decrypted data.
1987 */
1988 f2fs_wait_on_block_writeback(inode, block_nr);
1989
1990 if (bio_add_page(bio, page, blocksize, 0) < blocksize)
1991 goto submit_and_realloc;
1992
1993 inc_page_count(F2FS_I_SB(inode), F2FS_RD_DATA);
1994 ClearPageError(page);
1995 *last_block_in_bio = block_nr;
1996 goto out;
1997confused:
1998 if (bio) {
1999 __submit_bio(F2FS_I_SB(inode), bio, DATA);
2000 bio = NULL;
2001 }
2002 unlock_page(page);
2003out:
2004 *bio_ret = bio;
2005 return ret;
2006}
2007
4c8ff709
CY		 2008#ifdef CONFIG_F2FS_FS_COMPRESSION
2009int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret,
2010 unsigned nr_pages, sector_t *last_block_in_bio,
2011 bool is_readahead)
2012{
2013 struct dnode_of_data dn;
2014 struct inode *inode = cc->inode;
2015 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2016 struct bio *bio = *bio_ret;
2017 unsigned int start_idx = cc->cluster_idx << cc->log_cluster_size;
2018 sector_t last_block_in_file;
2019 const unsigned blkbits = inode->i_blkbits;
2020 const unsigned blocksize = 1 << blkbits;
2021 struct decompress_io_ctx *dic = NULL;
2022 int i;
2023 int ret = 0;
2024
2025 f2fs_bug_on(sbi, f2fs_cluster_is_empty(cc));
2026
2027 last_block_in_file = (i_size_read(inode) + blocksize - 1) >> blkbits;
2028
2029 /* get rid of pages beyond EOF */
2030 for (i = 0; i < cc->cluster_size; i++) {
2031 struct page *page = cc->rpages[i];
2032
2033 if (!page)
2034 continue;
2035 if ((sector_t)page->index >= last_block_in_file) {
2036 zero_user_segment(page, 0, PAGE_SIZE);
2037 if (!PageUptodate(page))
2038 SetPageUptodate(page);
2039 } else if (!PageUptodate(page)) {
2040 continue;
2041 }
2042 unlock_page(page);
2043 cc->rpages[i] = NULL;
2044 cc->nr_rpages--;
2045 }
2046
2047 /* we are done since all pages are beyond EOF */
2048 if (f2fs_cluster_is_empty(cc))
2049 goto out;
2050
2051 set_new_dnode(&dn, inode, NULL, NULL, 0);
2052 ret = f2fs_get_dnode_of_data(&dn, start_idx, LOOKUP_NODE);
2053 if (ret)
2054 goto out;
2055
2056 /* cluster was overwritten as normal cluster */
2057 if (dn.data_blkaddr != COMPRESS_ADDR)
2058 goto out;
2059
2060 for (i = 1; i < cc->cluster_size; i++) {
2061 block_t blkaddr;
2062
2063 blkaddr = datablock_addr(dn.inode, dn.node_page,
2064 dn.ofs_in_node + i);
2065
2066 if (!__is_valid_data_blkaddr(blkaddr))
2067 break;
2068
2069 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC)) {
2070 ret = -EFAULT;
2071 goto out_put_dnode;
2072 }
2073 cc->nr_cpages++;
2074 }
2075
2076 /* nothing to decompress */
2077 if (cc->nr_cpages == 0) {
2078 ret = 0;
2079 goto out_put_dnode;
2080 }
2081
2082 dic = f2fs_alloc_dic(cc);
2083 if (IS_ERR(dic)) {
2084 ret = PTR_ERR(dic);
2085 goto out_put_dnode;
2086 }
2087
2088 for (i = 0; i < dic->nr_cpages; i++) {
2089 struct page *page = dic->cpages[i];
2090 block_t blkaddr;
2091
2092 blkaddr = datablock_addr(dn.inode, dn.node_page,
2093 dn.ofs_in_node + i + 1);
2094
2095 if (bio && !page_is_mergeable(sbi, bio,
2096 *last_block_in_bio, blkaddr)) {
2097submit_and_realloc:
2098 __submit_bio(sbi, bio, DATA);
2099 bio = NULL;
2100 }
2101
2102 if (!bio) {
2103 bio = f2fs_grab_read_bio(inode, blkaddr, nr_pages,
2104 is_readahead ? REQ_RAHEAD : 0,
2105 page->index);
2106 if (IS_ERR(bio)) {
2107 ret = PTR_ERR(bio);
2108 bio = NULL;
2109 dic->failed = true;
2110 if (refcount_sub_and_test(dic->nr_cpages - i,
2111 &dic->ref))
2112 f2fs_decompress_end_io(dic->rpages,
2113 cc->cluster_size, true,
2114 false);
2115 f2fs_free_dic(dic);
2116 f2fs_put_dnode(&dn);
2117 *bio_ret = bio;
2118 return ret;
2119 }
2120 }
2121
2122 f2fs_wait_on_block_writeback(inode, blkaddr);
2123
2124 if (bio_add_page(bio, page, blocksize, 0) < blocksize)
2125 goto submit_and_realloc;
2126
2127 inc_page_count(sbi, F2FS_RD_DATA);
2128 ClearPageError(page);
2129 *last_block_in_bio = blkaddr;
2130 }
2131
2132 f2fs_put_dnode(&dn);
2133
2134 *bio_ret = bio;
2135 return 0;
2136
2137out_put_dnode:
2138 f2fs_put_dnode(&dn);
2139out:
2140 f2fs_decompress_end_io(cc->rpages, cc->cluster_size, true, false);
2141 *bio_ret = bio;
2142 return ret;
2143}
2144#endif
2145
f1e88660
JK		 2146/*
2147 * This function was originally taken from fs/mpage.c, and customized for f2fs.
2148 * Major change was from block_size == page_size in f2fs by default.
e2e59414
JK		 2149 *
2150 * Note that the aops->readpages() function is ONLY used for read-ahead. If
2151 * this function ever deviates from doing just read-ahead, it should either
2152 * use ->readpage() or do the necessary surgery to decouple ->readpages()
2153 * from read-ahead.
f1e88660 2154 */
4c8ff709 2155int f2fs_mpage_readpages(struct address_space *mapping,
f1e88660 2156 struct list_head *pages, struct page *page,
e2e59414 2157 unsigned nr_pages, bool is_readahead)
f1e88660
JK		 2158{
2159 struct bio *bio = NULL;
f1e88660
JK		 2160 sector_t last_block_in_bio = 0;
2161 struct inode *inode = mapping->host;
f1e88660 2162 struct f2fs_map_blocks map;
4c8ff709
CY		 2163#ifdef CONFIG_F2FS_FS_COMPRESSION
2164 struct compress_ctx cc = {
2165 .inode = inode,
2166 .log_cluster_size = F2FS_I(inode)->i_log_cluster_size,
2167 .cluster_size = F2FS_I(inode)->i_cluster_size,
2168 .cluster_idx = NULL_CLUSTER,
2169 .rpages = NULL,
2170 .cpages = NULL,
2171 .nr_rpages = 0,
2172 .nr_cpages = 0,
2173 };
2174#endif
2175 unsigned max_nr_pages = nr_pages;
2df0ab04 2176 int ret = 0;
f1e88660
JK		 2177
2178 map.m_pblk = 0;
2179 map.m_lblk = 0;
2180 map.m_len = 0;
2181 map.m_flags = 0;
da85985c 2182 map.m_next_pgofs = NULL;
c4020b2d 2183 map.m_next_extent = NULL;
d5097be5 2184 map.m_seg_type = NO_CHECK_TYPE;
f9d6d059 2185 map.m_may_create = false;
f1e88660 2186
736c0a74 2187 for (; nr_pages; nr_pages--) {
f1e88660 2188 if (pages) {
939afa94 2189 page = list_last_entry(pages, struct page, lru);
a83d50bc
KM		 2190
2191 prefetchw(&page->flags);
f1e88660
JK		 2192 list_del(&page->lru);
2193 if (add_to_page_cache_lru(page, mapping,
4969c06a 2194 page_index(page),
8a5c743e 2195 readahead_gfp_mask(mapping)))
f1e88660
JK		 2196 goto next_page;
2197 }
2198
4c8ff709
CY		 2199#ifdef CONFIG_F2FS_FS_COMPRESSION
2200 if (f2fs_compressed_file(inode)) {
2201 /* there are remained comressed pages, submit them */
2202 if (!f2fs_cluster_can_merge_page(&cc, page->index)) {
2203 ret = f2fs_read_multi_pages(&cc, &bio,
2204 max_nr_pages,
2205 &last_block_in_bio,
2206 is_readahead);
2207 f2fs_destroy_compress_ctx(&cc);
2208 if (ret)
2209 goto set_error_page;
2210 }
2211 ret = f2fs_is_compressed_cluster(inode, page->index);
2212 if (ret < 0)
2213 goto set_error_page;
2214 else if (!ret)
2215 goto read_single_page;
2216
2217 ret = f2fs_init_compress_ctx(&cc);
2218 if (ret)
2219 goto set_error_page;
2220
2221 f2fs_compress_ctx_add_page(&cc, page);
2222
2223 goto next_page;
2224 }
2225read_single_page:
2226#endif
2227
2228 ret = f2fs_read_single_page(inode, page, max_nr_pages, &map,
2229 &bio, &last_block_in_bio, is_readahead);
2df0ab04 2230 if (ret) {
4c8ff709
CY		 2231#ifdef CONFIG_F2FS_FS_COMPRESSION
2232set_error_page:
2233#endif
2df0ab04 2234 SetPageError(page);
09cbfeaf 2235 zero_user_segment(page, 0, PAGE_SIZE);
f1e88660 2236 unlock_page(page);
f1e88660 2237 }
f1e88660
JK		 2238next_page:
2239 if (pages)
09cbfeaf 2240 put_page(page);
4c8ff709
CY		 2241
2242#ifdef CONFIG_F2FS_FS_COMPRESSION
2243 if (f2fs_compressed_file(inode)) {
2244 /* last page */
2245 if (nr_pages == 1 && !f2fs_cluster_is_empty(&cc)) {
2246 ret = f2fs_read_multi_pages(&cc, &bio,
2247 max_nr_pages,
2248 &last_block_in_bio,
2249 is_readahead);
2250 f2fs_destroy_compress_ctx(&cc);
2251 }
2252 }
2253#endif
f1e88660
JK		 2254 }
2255 BUG_ON(pages && !list_empty(pages));
2256 if (bio)
4fc29c1a 2257 __submit_bio(F2FS_I_SB(inode), bio, DATA);
2df0ab04 2258 return pages ? 0 : ret;
f1e88660
JK		 2259}
2260
eb47b800
JK		 2261static int f2fs_read_data_page(struct file *file, struct page *page)
2262{
4969c06a 2263 struct inode *inode = page_file_mapping(page)->host;
b3d208f9 2264 int ret = -EAGAIN;
9ffe0fb5 2265
c20e89cd
CY		 2266 trace_f2fs_readpage(page, DATA);
2267
4c8ff709
CY		 2268 if (!f2fs_is_compress_backend_ready(inode)) {
2269 unlock_page(page);
2270 return -EOPNOTSUPP;
2271 }
2272
e1c42045 2273 /* If the file has inline data, try to read it directly */
9ffe0fb5
HL		 2274 if (f2fs_has_inline_data(inode))
2275 ret = f2fs_read_inline_data(inode, page);
b3d208f9 2276 if (ret == -EAGAIN)
4969c06a
JK		 2277 ret = f2fs_mpage_readpages(page_file_mapping(page),
2278 NULL, page, 1, false);
9ffe0fb5 2279 return ret;
eb47b800
JK		 2280}
2281
2282static int f2fs_read_data_pages(struct file *file,
2283 struct address_space *mapping,
2284 struct list_head *pages, unsigned nr_pages)
2285{
71cb4aff 2286 struct inode *inode = mapping->host;
939afa94 2287 struct page *page = list_last_entry(pages, struct page, lru);
b8c29400
CY		 2288
2289 trace_f2fs_readpages(inode, page, nr_pages);
9ffe0fb5 2290
4c8ff709
CY		 2291 if (!f2fs_is_compress_backend_ready(inode))
2292 return 0;
2293
9ffe0fb5
HL		 2294 /* If the file has inline data, skip readpages */
2295 if (f2fs_has_inline_data(inode))
2296 return 0;
2297
e2e59414 2298 return f2fs_mpage_readpages(mapping, pages, NULL, nr_pages, true);
eb47b800
JK		 2299}
2300
4c8ff709 2301int f2fs_encrypt_one_page(struct f2fs_io_info *fio)
7eab0c0d
HP		 2302{
2303 struct inode *inode = fio->page->mapping->host;
4c8ff709 2304 struct page *mpage, *page;
7eab0c0d
HP		 2305 gfp_t gfp_flags = GFP_NOFS;
2306
1958593e 2307 if (!f2fs_encrypted_file(inode))
7eab0c0d
HP		 2308 return 0;
2309
4c8ff709
CY		 2310 page = fio->compressed_page ? fio->compressed_page : fio->page;
2311
6dbb1796 2312 /* wait for GCed page writeback via META_MAPPING */
0ded69f6 2313 f2fs_wait_on_block_writeback(inode, fio->old_blkaddr);
7eab0c0d
HP		 2314
2315retry_encrypt:
4c8ff709
CY		 2316 fio->encrypted_page = fscrypt_encrypt_pagecache_blocks(page,
2317 PAGE_SIZE, 0, gfp_flags);
6aa58d8a
CY		 2318 if (IS_ERR(fio->encrypted_page)) {
2319 /* flush pending IOs and wait for a while in the ENOMEM case */
2320 if (PTR_ERR(fio->encrypted_page) == -ENOMEM) {
2321 f2fs_flush_merged_writes(fio->sbi);
2322 congestion_wait(BLK_RW_ASYNC, HZ/50);
2323 gfp_flags |= __GFP_NOFAIL;
2324 goto retry_encrypt;
2325 }
2326 return PTR_ERR(fio->encrypted_page);
2327 }
7eab0c0d 2328
6aa58d8a
CY		 2329 mpage = find_lock_page(META_MAPPING(fio->sbi), fio->old_blkaddr);
2330 if (mpage) {
2331 if (PageUptodate(mpage))
2332 memcpy(page_address(mpage),
2333 page_address(fio->encrypted_page), PAGE_SIZE);
2334 f2fs_put_page(mpage, 1);
7eab0c0d 2335 }
6aa58d8a 2336 return 0;
7eab0c0d
HP		 2337}
2338
bb9e3bb8
CY		 2339static inline bool check_inplace_update_policy(struct inode *inode,
2340 struct f2fs_io_info *fio)
7eab0c0d 2341{
bb9e3bb8
CY		 2342 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2343 unsigned int policy = SM_I(sbi)->ipu_policy;
7eab0c0d 2344
bb9e3bb8
CY		 2345 if (policy & (0x1 << F2FS_IPU_FORCE))
2346 return true;
4d57b86d 2347 if (policy & (0x1 << F2FS_IPU_SSR) && f2fs_need_SSR(sbi))
bb9e3bb8
CY		 2348 return true;
2349 if (policy & (0x1 << F2FS_IPU_UTIL) &&
2350 utilization(sbi) > SM_I(sbi)->min_ipu_util)
2351 return true;
4d57b86d 2352 if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && f2fs_need_SSR(sbi) &&
bb9e3bb8
CY		 2353 utilization(sbi) > SM_I(sbi)->min_ipu_util)
2354 return true;
2355
2356 /*
2357 * IPU for rewrite async pages
2358 */
2359 if (policy & (0x1 << F2FS_IPU_ASYNC) &&
2360 fio && fio->op == REQ_OP_WRITE &&
2361 !(fio->op_flags & REQ_SYNC) &&
62230e0d 2362 !IS_ENCRYPTED(inode))
bb9e3bb8
CY		 2363 return true;
2364
2365 /* this is only set during fdatasync */
2366 if (policy & (0x1 << F2FS_IPU_FSYNC) &&
2367 is_inode_flag_set(inode, FI_NEED_IPU))
2368 return true;
2369
4354994f
DR		 2370 if (unlikely(fio && is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2371 !f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2372 return true;
2373
bb9e3bb8
CY		 2374 return false;
2375}
2376
4d57b86d 2377bool f2fs_should_update_inplace(struct inode *inode, struct f2fs_io_info *fio)
bb9e3bb8 2378{
1ad71a27
JK		 2379 if (f2fs_is_pinned_file(inode))
2380 return true;
bb9e3bb8
CY		 2381
2382 /* if this is cold file, we should overwrite to avoid fragmentation */
2383 if (file_is_cold(inode))
2384 return true;
2385
2386 return check_inplace_update_policy(inode, fio);
2387}
2388
4d57b86d 2389bool f2fs_should_update_outplace(struct inode *inode, struct f2fs_io_info *fio)
bb9e3bb8
CY		 2390{
2391 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2392
2393 if (test_opt(sbi, LFS))
2394 return true;
2395 if (S_ISDIR(inode->i_mode))
2396 return true;
af033b2a
CY		 2397 if (IS_NOQUOTA(inode))
2398 return true;
bb9e3bb8
CY		 2399 if (f2fs_is_atomic_file(inode))
2400 return true;
2401 if (fio) {
2402 if (is_cold_data(fio->page))
2403 return true;
2404 if (IS_ATOMIC_WRITTEN_PAGE(fio->page))
2405 return true;
4354994f
DR		 2406 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2407 f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2408 return true;
bb9e3bb8
CY		 2409 }
2410 return false;
2411}
2412
7eab0c0d
HP		 2413static inline bool need_inplace_update(struct f2fs_io_info *fio)
2414{
2415 struct inode *inode = fio->page->mapping->host;
2416
4d57b86d 2417 if (f2fs_should_update_outplace(inode, fio))
7eab0c0d
HP		 2418 return false;
2419
4d57b86d 2420 return f2fs_should_update_inplace(inode, fio);
7eab0c0d
HP		 2421}
2422
4d57b86d 2423int f2fs_do_write_data_page(struct f2fs_io_info *fio)
eb47b800 2424{
05ca3632 2425 struct page *page = fio->page;
eb47b800 2426 struct inode *inode = page->mapping->host;
eb47b800 2427 struct dnode_of_data dn;
e959c8f5 2428 struct extent_info ei = {0,0,0};
7735730d 2429 struct node_info ni;
e959c8f5 2430 bool ipu_force = false;
eb47b800
JK		 2431 int err = 0;
2432
2433 set_new_dnode(&dn, inode, NULL, NULL, 0);
e959c8f5
HP		 2434 if (need_inplace_update(fio) &&
2435 f2fs_lookup_extent_cache(inode, page->index, &ei)) {
2436 fio->old_blkaddr = ei.blk + page->index - ei.fofs;
a817737e 2437
c9b60788 2438 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
93770ab7 2439 DATA_GENERIC_ENHANCE))
10f966bb 2440 return -EFSCORRUPTED;
c9b60788
CY		 2441
2442 ipu_force = true;
2443 fio->need_lock = LOCK_DONE;
2444 goto got_it;
e959c8f5 2445 }
279d6df2 2446
d29460e5
JK		 2447 /* Deadlock due to between page->lock and f2fs_lock_op */
2448 if (fio->need_lock == LOCK_REQ && !f2fs_trylock_op(fio->sbi))
2449 return -EAGAIN;
279d6df2 2450
4d57b86d 2451 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
eb47b800 2452 if (err)
279d6df2 2453 goto out;
eb47b800 2454
28bc106b 2455 fio->old_blkaddr = dn.data_blkaddr;
eb47b800
JK		 2456
2457 /* This page is already truncated */
7a9d7548 2458 if (fio->old_blkaddr == NULL_ADDR) {
2bca1e23 2459 ClearPageUptodate(page);
2baf0781 2460 clear_cold_data(page);
eb47b800 2461 goto out_writepage;
2bca1e23 2462 }
e959c8f5 2463got_it:
c9b60788
CY		 2464 if (__is_valid_data_blkaddr(fio->old_blkaddr) &&
2465 !f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
93770ab7 2466 DATA_GENERIC_ENHANCE)) {
10f966bb 2467 err = -EFSCORRUPTED;
c9b60788
CY		 2468 goto out_writepage;
2469 }
eb47b800
JK		 2470 /*
2471 * If current allocation needs SSR,
2472 * it had better in-place writes for updated data.
2473 */
93770ab7
CY		 2474 if (ipu_force ||
2475 (__is_valid_data_blkaddr(fio->old_blkaddr) &&
7b525dd0 2476 need_inplace_update(fio))) {
4c8ff709 2477 err = f2fs_encrypt_one_page(fio);
cc15620b
JK		 2478 if (err)
2479 goto out_writepage;
2480
2481 set_page_writeback(page);
17c50035 2482 ClearPageError(page);
279d6df2 2483 f2fs_put_dnode(&dn);
cc15620b 2484 if (fio->need_lock == LOCK_REQ)
279d6df2 2485 f2fs_unlock_op(fio->sbi);
4d57b86d 2486 err = f2fs_inplace_write_data(fio);
6492a335
CY		 2487 if (err) {
2488 if (f2fs_encrypted_file(inode))
d2d0727b 2489 fscrypt_finalize_bounce_page(&fio->encrypted_page);
6492a335
CY		 2490 if (PageWriteback(page))
2491 end_page_writeback(page);
cd23ffa9
CY		 2492 } else {
2493 set_inode_flag(inode, FI_UPDATE_WRITE);
6492a335 2494 }
7eab0c0d 2495 trace_f2fs_do_write_data_page(fio->page, IPU);
279d6df2 2496 return err;
eb47b800 2497 }
279d6df2 2498
cc15620b
JK		 2499 if (fio->need_lock == LOCK_RETRY) {
2500 if (!f2fs_trylock_op(fio->sbi)) {
2501 err = -EAGAIN;
2502 goto out_writepage;
2503 }
2504 fio->need_lock = LOCK_REQ;
2505 }
2506
7735730d
CY		 2507 err = f2fs_get_node_info(fio->sbi, dn.nid, &ni);
2508 if (err)
2509 goto out_writepage;
2510
2511 fio->version = ni.version;
2512
4c8ff709 2513 err = f2fs_encrypt_one_page(fio);
cc15620b
JK		 2514 if (err)
2515 goto out_writepage;
2516
2517 set_page_writeback(page);
17c50035 2518 ClearPageError(page);
cc15620b 2519
4c8ff709
CY		 2520 if (fio->compr_blocks && fio->old_blkaddr == COMPRESS_ADDR)
2521 f2fs_i_compr_blocks_update(inode, fio->compr_blocks - 1, false);
2522
279d6df2 2523 /* LFS mode write path */
4d57b86d 2524 f2fs_outplace_write_data(&dn, fio);
279d6df2
HP		 2525 trace_f2fs_do_write_data_page(page, OPU);
2526 set_inode_flag(inode, FI_APPEND_WRITE);
2527 if (page->index == 0)
2528 set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
eb47b800
JK		 2529out_writepage:
2530 f2fs_put_dnode(&dn);
279d6df2 2531out:
cc15620b 2532 if (fio->need_lock == LOCK_REQ)
279d6df2 2533 f2fs_unlock_op(fio->sbi);
eb47b800
JK		 2534 return err;
2535}
2536
4c8ff709 2537int f2fs_write_single_data_page(struct page *page, int *submitted,
8648de2c
CY		 2538 struct bio **bio,
2539 sector_t *last_block,
b0af6d49 2540 struct writeback_control *wbc,
4c8ff709
CY		 2541 enum iostat_type io_type,
2542 int compr_blocks)
eb47b800
JK		 2543{
2544 struct inode *inode = page->mapping->host;
4081363f 2545 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
eb47b800 2546 loff_t i_size = i_size_read(inode);
4c8ff709 2547 const pgoff_t end_index = ((unsigned long long)i_size)
09cbfeaf 2548 >> PAGE_SHIFT;
1f0d5c91 2549 loff_t psize = (loff_t)(page->index + 1) << PAGE_SHIFT;
9ffe0fb5 2550 unsigned offset = 0;
39936837 2551 bool need_balance_fs = false;
eb47b800 2552 int err = 0;
458e6197 2553 struct f2fs_io_info fio = {
05ca3632 2554 .sbi = sbi,
39d787be 2555 .ino = inode->i_ino,
458e6197 2556 .type = DATA,
04d328de 2557 .op = REQ_OP_WRITE,
7637241e 2558 .op_flags = wbc_to_write_flags(wbc),
e959c8f5 2559 .old_blkaddr = NULL_ADDR,
05ca3632 2560 .page = page,
4375a336 2561 .encrypted_page = NULL,
d68f735b 2562 .submitted = false,
4c8ff709 2563 .compr_blocks = compr_blocks,
cc15620b 2564 .need_lock = LOCK_RETRY,
b0af6d49 2565 .io_type = io_type,
578c6478 2566 .io_wbc = wbc,
8648de2c
CY		 2567 .bio = bio,
2568 .last_block = last_block,
458e6197 2569 };
eb47b800 2570
ecda0de3
CY		 2571 trace_f2fs_writepage(page, DATA);
2572
db198ae0
CY		 2573 /* we should bypass data pages to proceed the kworkder jobs */
2574 if (unlikely(f2fs_cp_error(sbi))) {
2575 mapping_set_error(page->mapping, -EIO);
1174abfd
CY		 2576 /*
2577 * don't drop any dirty dentry pages for keeping lastest
2578 * directory structure.
2579 */
2580 if (S_ISDIR(inode->i_mode))
2581 goto redirty_out;
db198ae0
CY		 2582 goto out;
2583 }
2584
0771fcc7
CY		 2585 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
2586 goto redirty_out;
2587
4c8ff709
CY		 2588 if (page->index < end_index ||
2589 f2fs_verity_in_progress(inode) ||
2590 compr_blocks)
39936837 2591 goto write;
eb47b800
JK		 2592
2593 /*
2594 * If the offset is out-of-range of file size,
2595 * this page does not have to be written to disk.
2596 */
09cbfeaf 2597 offset = i_size & (PAGE_SIZE - 1);
76f60268 2598 if ((page->index >= end_index + 1) || !offset)
39936837 2599 goto out;
eb47b800 2600
09cbfeaf 2601 zero_user_segment(page, offset, PAGE_SIZE);
39936837 2602write:
1e84371f
JK		 2603 if (f2fs_is_drop_cache(inode))
2604 goto out;
e6e5f561
JK		 2605 /* we should not write 0'th page having journal header */
2606 if (f2fs_is_volatile_file(inode) && (!page->index ||
2607 (!wbc->for_reclaim &&
4d57b86d 2608 f2fs_available_free_memory(sbi, BASE_CHECK))))
1e84371f 2609 goto redirty_out;
eb47b800 2610
39936837 2611 /* Dentry blocks are controlled by checkpoint */
eb47b800 2612 if (S_ISDIR(inode->i_mode)) {
cc15620b 2613 fio.need_lock = LOCK_DONE;
4d57b86d 2614 err = f2fs_do_write_data_page(&fio);
8618b881
JK		 2615 goto done;
2616 }
9ffe0fb5 2617
8618b881 2618 if (!wbc->for_reclaim)
39936837 2619 need_balance_fs = true;
7f3037a5 2620 else if (has_not_enough_free_secs(sbi, 0, 0))
39936837 2621 goto redirty_out;
ef095d19
JK		 2622 else
2623 set_inode_flag(inode, FI_HOT_DATA);
eb47b800 2624
b3d208f9 2625 err = -EAGAIN;
dd7b2333 2626 if (f2fs_has_inline_data(inode)) {
b3d208f9 2627 err = f2fs_write_inline_data(inode, page);
dd7b2333
YH		 2628 if (!err)
2629 goto out;
2630 }
279d6df2 2631
cc15620b 2632 if (err == -EAGAIN) {
4d57b86d 2633 err = f2fs_do_write_data_page(&fio);
cc15620b
JK		 2634 if (err == -EAGAIN) {
2635 fio.need_lock = LOCK_REQ;
4d57b86d 2636 err = f2fs_do_write_data_page(&fio);
cc15620b
JK		 2637 }
2638 }
a0d00fad 2639
eb449797
CY		 2640 if (err) {
2641 file_set_keep_isize(inode);
2642 } else {
c10c9820 2643 spin_lock(&F2FS_I(inode)->i_size_lock);
eb449797
CY		 2644 if (F2FS_I(inode)->last_disk_size < psize)
2645 F2FS_I(inode)->last_disk_size = psize;
c10c9820 2646 spin_unlock(&F2FS_I(inode)->i_size_lock);
eb449797 2647 }
279d6df2 2648
8618b881
JK		 2649done:
2650 if (err && err != -ENOENT)
2651 goto redirty_out;
eb47b800 2652
39936837 2653out:
a7ffdbe2 2654 inode_dec_dirty_pages(inode);
2baf0781 2655 if (err) {
2bca1e23 2656 ClearPageUptodate(page);
2baf0781
CY		 2657 clear_cold_data(page);
2658 }
0c3a5797
CY		 2659
2660 if (wbc->for_reclaim) {
bab475c5 2661 f2fs_submit_merged_write_cond(sbi, NULL, page, 0, DATA);
ef095d19 2662 clear_inode_flag(inode, FI_HOT_DATA);
4d57b86d 2663 f2fs_remove_dirty_inode(inode);
d68f735b 2664 submitted = NULL;
0c3a5797 2665 }
eb47b800 2666 unlock_page(page);
186857c5 2667 if (!S_ISDIR(inode->i_mode) && !IS_NOQUOTA(inode) &&
0b20fcec 2668 !F2FS_I(inode)->cp_task)
a7881893 2669 f2fs_balance_fs(sbi, need_balance_fs);
0c3a5797 2670
d68f735b 2671 if (unlikely(f2fs_cp_error(sbi))) {
b9109b0e 2672 f2fs_submit_merged_write(sbi, DATA);
0b20fcec 2673 f2fs_submit_merged_ipu_write(sbi, bio, NULL);
d68f735b
JK		 2674 submitted = NULL;
2675 }
2676
2677 if (submitted)
4c8ff709 2678 *submitted = fio.submitted ? 1 : 0;
0c3a5797 2679
eb47b800
JK		 2680 return 0;
2681
eb47b800 2682redirty_out:
76f60268 2683 redirty_page_for_writepage(wbc, page);
5b19d284
JK		 2684 /*
2685 * pageout() in MM traslates EAGAIN, so calls handle_write_error()
2686 * -> mapping_set_error() -> set_bit(AS_EIO, ...).
2687 * file_write_and_wait_range() will see EIO error, which is critical
2688 * to return value of fsync() followed by atomic_write failure to user.
2689 */
2690 if (!err || wbc->for_reclaim)
0002b61b 2691 return AOP_WRITEPAGE_ACTIVATE;
b230e6ca
JK		 2692 unlock_page(page);
2693 return err;
fa9150a8
NJ		 2694}
2695
f566bae8
JK		 2696static int f2fs_write_data_page(struct page *page,
2697 struct writeback_control *wbc)
2698{
4c8ff709
CY		 2699#ifdef CONFIG_F2FS_FS_COMPRESSION
2700 struct inode *inode = page->mapping->host;
2701
2702 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2703 goto out;
2704
2705 if (f2fs_compressed_file(inode)) {
2706 if (f2fs_is_compressed_cluster(inode, page->index)) {
2707 redirty_page_for_writepage(wbc, page);
2708 return AOP_WRITEPAGE_ACTIVATE;
2709 }
2710 }
2711out:
2712#endif
2713
2714 return f2fs_write_single_data_page(page, NULL, NULL, NULL,
2715 wbc, FS_DATA_IO, 0);
f566bae8
JK		 2716}
2717
8f46dcae
CY		 2718/*
2719 * This function was copied from write_cche_pages from mm/page-writeback.c.
2720 * The major change is making write step of cold data page separately from
2721 * warm/hot data page.
2722 */
2723static int f2fs_write_cache_pages(struct address_space *mapping,
b0af6d49
CY		 2724 struct writeback_control *wbc,
2725 enum iostat_type io_type)
8f46dcae
CY		 2726{
2727 int ret = 0;
4c8ff709 2728 int done = 0, retry = 0;
8f46dcae 2729 struct pagevec pvec;
c29fd0c0 2730 struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
8648de2c
CY		 2731 struct bio *bio = NULL;
2732 sector_t last_block;
4c8ff709
CY		 2733#ifdef CONFIG_F2FS_FS_COMPRESSION
2734 struct inode *inode = mapping->host;
2735 struct compress_ctx cc = {
2736 .inode = inode,
2737 .log_cluster_size = F2FS_I(inode)->i_log_cluster_size,
2738 .cluster_size = F2FS_I(inode)->i_cluster_size,
2739 .cluster_idx = NULL_CLUSTER,
2740 .rpages = NULL,
2741 .nr_rpages = 0,
2742 .cpages = NULL,
2743 .rbuf = NULL,
2744 .cbuf = NULL,
2745 .rlen = PAGE_SIZE * F2FS_I(inode)->i_cluster_size,
2746 .private = NULL,
2747 };
2748#endif
8f46dcae
CY		 2749 int nr_pages;
2750 pgoff_t uninitialized_var(writeback_index);
2751 pgoff_t index;
2752 pgoff_t end; /* Inclusive */
2753 pgoff_t done_index;
2754 int cycled;
2755 int range_whole = 0;
10bbd235 2756 xa_mark_t tag;
bab475c5 2757 int nwritten = 0;
4c8ff709
CY		 2758 int submitted = 0;
2759 int i;
8f46dcae 2760
86679820 2761 pagevec_init(&pvec);
46ae957f 2762
ef095d19
JK		 2763 if (get_dirty_pages(mapping->host) <=
2764 SM_I(F2FS_M_SB(mapping))->min_hot_blocks)
2765 set_inode_flag(mapping->host, FI_HOT_DATA);
2766 else
2767 clear_inode_flag(mapping->host, FI_HOT_DATA);
2768
8f46dcae
CY		 2769 if (wbc->range_cyclic) {
2770 writeback_index = mapping->writeback_index; /* prev offset */
2771 index = writeback_index;
2772 if (index == 0)
2773 cycled = 1;
2774 else
2775 cycled = 0;
2776 end = -1;
2777 } else {
09cbfeaf
KS		 2778 index = wbc->range_start >> PAGE_SHIFT;
2779 end = wbc->range_end >> PAGE_SHIFT;
8f46dcae
CY		 2780 if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
2781 range_whole = 1;
2782 cycled = 1; /* ignore range_cyclic tests */
2783 }
2784 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2785 tag = PAGECACHE_TAG_TOWRITE;
2786 else
2787 tag = PAGECACHE_TAG_DIRTY;
2788retry:
4c8ff709 2789 retry = 0;
8f46dcae
CY		 2790 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2791 tag_pages_for_writeback(mapping, index, end);
2792 done_index = index;
4c8ff709 2793 while (!done && !retry && (index <= end)) {
69c4f35d 2794 nr_pages = pagevec_lookup_range_tag(&pvec, mapping, &index, end,
67fd707f 2795 tag);
8f46dcae
CY		 2796 if (nr_pages == 0)
2797 break;
2798
2799 for (i = 0; i < nr_pages; i++) {
2800 struct page *page = pvec.pages[i];
4c8ff709
CY		 2801 bool need_readd;
2802readd:
2803 need_readd = false;
2804#ifdef CONFIG_F2FS_FS_COMPRESSION
2805 if (f2fs_compressed_file(inode)) {
2806 ret = f2fs_init_compress_ctx(&cc);
2807 if (ret) {
2808 done = 1;
2809 break;
2810 }
2811
2812 if (!f2fs_cluster_can_merge_page(&cc,
2813 page->index)) {
2814 ret = f2fs_write_multi_pages(&cc,
2815 &submitted, wbc, io_type);
2816 if (!ret)
2817 need_readd = true;
2818 goto result;
2819 }
8f46dcae 2820
4c8ff709
CY		 2821 if (unlikely(f2fs_cp_error(sbi)))
2822 goto lock_page;
2823
2824 if (f2fs_cluster_is_empty(&cc)) {
2825 void *fsdata = NULL;
2826 struct page *pagep;
2827 int ret2;
2828
2829 ret2 = f2fs_prepare_compress_overwrite(
2830 inode, &pagep,
2831 page->index, &fsdata);
2832 if (ret2 < 0) {
2833 ret = ret2;
2834 done = 1;
2835 break;
2836 } else if (ret2 &&
2837 !f2fs_compress_write_end(inode,
2838 fsdata, page->index,
2839 1)) {
2840 retry = 1;
2841 break;
2842 }
2843 } else {
2844 goto lock_page;
2845 }
2846 }
2847#endif
f8de4331 2848 /* give a priority to WB_SYNC threads */
c29fd0c0 2849 if (atomic_read(&sbi->wb_sync_req[DATA]) &&
f8de4331
CY		 2850 wbc->sync_mode == WB_SYNC_NONE) {
2851 done = 1;
2852 break;
2853 }
4c8ff709
CY		 2854#ifdef CONFIG_F2FS_FS_COMPRESSION
2855lock_page:
2856#endif
8f46dcae 2857 done_index = page->index;
d29460e5 2858retry_write:
8f46dcae
CY		 2859 lock_page(page);
2860
2861 if (unlikely(page->mapping != mapping)) {
2862continue_unlock:
2863 unlock_page(page);
2864 continue;
2865 }
2866
2867 if (!PageDirty(page)) {
2868 /* someone wrote it for us */
2869 goto continue_unlock;
2870 }
2871
8f46dcae 2872 if (PageWriteback(page)) {
0b20fcec 2873 if (wbc->sync_mode != WB_SYNC_NONE)
fec1d657 2874 f2fs_wait_on_page_writeback(page,
bae0ee7a 2875 DATA, true, true);
0b20fcec 2876 else
8f46dcae
CY		 2877 goto continue_unlock;
2878 }
2879
8f46dcae
CY		 2880 if (!clear_page_dirty_for_io(page))
2881 goto continue_unlock;
2882
4c8ff709
CY		 2883#ifdef CONFIG_F2FS_FS_COMPRESSION
2884 if (f2fs_compressed_file(inode)) {
2885 get_page(page);
2886 f2fs_compress_ctx_add_page(&cc, page);
2887 continue;
2888 }
2889#endif
2890 ret = f2fs_write_single_data_page(page, &submitted,
2891 &bio, &last_block, wbc, io_type, 0);
2892 if (ret == AOP_WRITEPAGE_ACTIVATE)
2893 unlock_page(page);
2894#ifdef CONFIG_F2FS_FS_COMPRESSION
2895result:
2896#endif
2897 nwritten += submitted;
2898 wbc->nr_to_write -= submitted;
2899
8f46dcae 2900 if (unlikely(ret)) {
0002b61b
CY		 2901 /*
2902 * keep nr_to_write, since vfs uses this to
2903 * get # of written pages.
2904 */
2905 if (ret == AOP_WRITEPAGE_ACTIVATE) {
0002b61b 2906 ret = 0;
4c8ff709 2907 goto next;
d29460e5
JK		 2908 } else if (ret == -EAGAIN) {
2909 ret = 0;
2910 if (wbc->sync_mode == WB_SYNC_ALL) {
2911 cond_resched();
2912 congestion_wait(BLK_RW_ASYNC,
4c8ff709 2913 HZ/50);
d29460e5
JK		 2914 goto retry_write;
2915 }
4c8ff709 2916 goto next;
0002b61b 2917 }
b230e6ca
JK		 2918 done_index = page->index + 1;
2919 done = 1;
2920 break;
8f46dcae
CY		 2921 }
2922
4c8ff709 2923 if (wbc->nr_to_write <= 0 &&
687de7f1 2924 wbc->sync_mode == WB_SYNC_NONE) {
8f46dcae
CY		 2925 done = 1;
2926 break;
2927 }
4c8ff709
CY		 2928next:
2929 if (need_readd)
2930 goto readd;
8f46dcae
CY		 2931 }
2932 pagevec_release(&pvec);
2933 cond_resched();
2934 }
4c8ff709
CY		 2935#ifdef CONFIG_F2FS_FS_COMPRESSION
2936 /* flush remained pages in compress cluster */
2937 if (f2fs_compressed_file(inode) && !f2fs_cluster_is_empty(&cc)) {
2938 ret = f2fs_write_multi_pages(&cc, &submitted, wbc, io_type);
2939 nwritten += submitted;
2940 wbc->nr_to_write -= submitted;
2941 if (ret) {
2942 done = 1;
2943 retry = 0;
2944 }
2945 }
2946#endif
2947 if ((!cycled && !done) || retry) {
8f46dcae
CY		 2948 cycled = 1;
2949 index = 0;
2950 end = writeback_index - 1;
2951 goto retry;
2952 }
2953 if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
2954 mapping->writeback_index = done_index;
2955
bab475c5 2956 if (nwritten)
b9109b0e 2957 f2fs_submit_merged_write_cond(F2FS_M_SB(mapping), mapping->host,
bab475c5 2958 NULL, 0, DATA);
8648de2c
CY		 2959 /* submit cached bio of IPU write */
2960 if (bio)
0b20fcec 2961 f2fs_submit_merged_ipu_write(sbi, &bio, NULL);
6ca56ca4 2962
8f46dcae
CY		 2963 return ret;
2964}
2965
853137ce
JK		 2966static inline bool __should_serialize_io(struct inode *inode,
2967 struct writeback_control *wbc)
2968{
2969 if (!S_ISREG(inode->i_mode))
2970 return false;
4c8ff709
CY		 2971 if (f2fs_compressed_file(inode))
2972 return true;
af033b2a
CY		 2973 if (IS_NOQUOTA(inode))
2974 return false;
040d2bb3
CY		 2975 /* to avoid deadlock in path of data flush */
2976 if (F2FS_I(inode)->cp_task)
2977 return false;
853137ce
JK		 2978 if (wbc->sync_mode != WB_SYNC_ALL)
2979 return true;
2980 if (get_dirty_pages(inode) >= SM_I(F2FS_I_SB(inode))->min_seq_blocks)
2981 return true;
2982 return false;
2983}
2984
fc99fe27 2985static int __f2fs_write_data_pages(struct address_space *mapping,
b0af6d49
CY		 2986 struct writeback_control *wbc,
2987 enum iostat_type io_type)
eb47b800
JK		 2988{
2989 struct inode *inode = mapping->host;
4081363f 2990 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
9dfa1baf 2991 struct blk_plug plug;
eb47b800 2992 int ret;
853137ce 2993 bool locked = false;
eb47b800 2994
cfb185a1 2995 /* deal with chardevs and other special file */
2996 if (!mapping->a_ops->writepage)
2997 return 0;
2998
6a290544
CY		 2999 /* skip writing if there is no dirty page in this inode */
3000 if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE)
3001 return 0;
3002
0771fcc7
CY		 3003 /* during POR, we don't need to trigger writepage at all. */
3004 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
3005 goto skip_write;
3006
af033b2a
CY		 3007 if ((S_ISDIR(inode->i_mode) || IS_NOQUOTA(inode)) &&
3008 wbc->sync_mode == WB_SYNC_NONE &&
a1257023 3009 get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) &&
4d57b86d 3010 f2fs_available_free_memory(sbi, DIRTY_DENTS))
a1257023
JK		 3011 goto skip_write;
3012
d323d005 3013 /* skip writing during file defragment */
91942321 3014 if (is_inode_flag_set(inode, FI_DO_DEFRAG))
d323d005
CY		 3015 goto skip_write;
3016
d31c7c3f
YH		 3017 trace_f2fs_writepages(mapping->host, wbc, DATA);
3018
687de7f1
JK		 3019 /* to avoid spliting IOs due to mixed WB_SYNC_ALL and WB_SYNC_NONE */
3020 if (wbc->sync_mode == WB_SYNC_ALL)
c29fd0c0
CY		 3021 atomic_inc(&sbi->wb_sync_req[DATA]);
3022 else if (atomic_read(&sbi->wb_sync_req[DATA]))
687de7f1
JK		 3023 goto skip_write;
3024
853137ce
JK		 3025 if (__should_serialize_io(inode, wbc)) {
3026 mutex_lock(&sbi->writepages);
3027 locked = true;
3028 }
3029
9dfa1baf 3030 blk_start_plug(&plug);
b0af6d49 3031 ret = f2fs_write_cache_pages(mapping, wbc, io_type);
9dfa1baf 3032 blk_finish_plug(&plug);
687de7f1 3033
853137ce
JK		 3034 if (locked)
3035 mutex_unlock(&sbi->writepages);
3036
687de7f1 3037 if (wbc->sync_mode == WB_SYNC_ALL)
c29fd0c0 3038 atomic_dec(&sbi->wb_sync_req[DATA]);
28ea6162
JK		 3039 /*
3040 * if some pages were truncated, we cannot guarantee its mapping->host
3041 * to detect pending bios.
3042 */
458e6197 3043
4d57b86d 3044 f2fs_remove_dirty_inode(inode);
eb47b800 3045 return ret;
d3baf95d
JK		 3046
3047skip_write:
a7ffdbe2 3048 wbc->pages_skipped += get_dirty_pages(inode);
d31c7c3f 3049 trace_f2fs_writepages(mapping->host, wbc, DATA);
d3baf95d 3050 return 0;
eb47b800
JK		 3051}
3052
b0af6d49
CY		 3053static int f2fs_write_data_pages(struct address_space *mapping,
3054 struct writeback_control *wbc)
3055{
3056 struct inode *inode = mapping->host;
3057
3058 return __f2fs_write_data_pages(mapping, wbc,
3059 F2FS_I(inode)->cp_task == current ?
3060 FS_CP_DATA_IO : FS_DATA_IO);
3061}
3062
3aab8f82
CY		 3063static void f2fs_write_failed(struct address_space *mapping, loff_t to)
3064{
3065 struct inode *inode = mapping->host;
819d9153 3066 loff_t i_size = i_size_read(inode);
3aab8f82 3067
3f188c23
JK		 3068 if (IS_NOQUOTA(inode))
3069 return;
3070
95ae251f
EB		 3071 /* In the fs-verity case, f2fs_end_enable_verity() does the truncate */
3072 if (to > i_size && !f2fs_verity_in_progress(inode)) {
a33c1502 3073 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
6f8d4455 3074 down_write(&F2FS_I(inode)->i_mmap_sem);
a33c1502 3075
819d9153 3076 truncate_pagecache(inode, i_size);
3f188c23 3077 f2fs_truncate_blocks(inode, i_size, true);
a33c1502 3078
5a3a2d83 3079 up_write(&F2FS_I(inode)->i_mmap_sem);
6f8d4455 3080 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3aab8f82
CY		 3081 }
3082}
3083
2aadac08
JK		 3084static int prepare_write_begin(struct f2fs_sb_info *sbi,
3085 struct page *page, loff_t pos, unsigned len,
3086 block_t *blk_addr, bool *node_changed)
3087{
3088 struct inode *inode = page->mapping->host;
3089 pgoff_t index = page->index;
3090 struct dnode_of_data dn;
3091 struct page *ipage;
b4d07a3e 3092 bool locked = false;
e15882b6 3093 struct extent_info ei = {0,0,0};
2aadac08 3094 int err = 0;
2866fb16 3095 int flag;
2aadac08 3096
24b84912
JK		 3097 /*
3098 * we already allocated all the blocks, so we don't need to get
3099 * the block addresses when there is no need to fill the page.
3100 */
dc91de78 3101 if (!f2fs_has_inline_data(inode) && len == PAGE_SIZE &&
95ae251f
EB		 3102 !is_inode_flag_set(inode, FI_NO_PREALLOC) &&
3103 !f2fs_verity_in_progress(inode))
24b84912
JK		 3104 return 0;
3105
2866fb16
SY		 3106 /* f2fs_lock_op avoids race between write CP and convert_inline_page */
3107 if (f2fs_has_inline_data(inode) && pos + len > MAX_INLINE_DATA(inode))
3108 flag = F2FS_GET_BLOCK_DEFAULT;
3109 else
3110 flag = F2FS_GET_BLOCK_PRE_AIO;
3111
b4d07a3e 3112 if (f2fs_has_inline_data(inode) ||
09cbfeaf 3113 (pos & PAGE_MASK) >= i_size_read(inode)) {
2866fb16 3114 __do_map_lock(sbi, flag, true);
b4d07a3e
JK		 3115 locked = true;
3116 }
4c8ff709 3117
b4d07a3e 3118restart:
2aadac08 3119 /* check inline_data */
4d57b86d 3120 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2aadac08
JK		 3121 if (IS_ERR(ipage)) {
3122 err = PTR_ERR(ipage);
3123 goto unlock_out;
3124 }
3125
3126 set_new_dnode(&dn, inode, ipage, ipage, 0);
3127
3128 if (f2fs_has_inline_data(inode)) {
f2470371 3129 if (pos + len <= MAX_INLINE_DATA(inode)) {
4d57b86d 3130 f2fs_do_read_inline_data(page, ipage);
91942321 3131 set_inode_flag(inode, FI_DATA_EXIST);
ab47036d
CY		 3132 if (inode->i_nlink)
3133 set_inline_node(ipage);
2aadac08
JK		 3134 } else {
3135 err = f2fs_convert_inline_page(&dn, page);
3136 if (err)
b4d07a3e
JK		 3137 goto out;
3138 if (dn.data_blkaddr == NULL_ADDR)
3139 err = f2fs_get_block(&dn, index);
3140 }
3141 } else if (locked) {
3142 err = f2fs_get_block(&dn, index);
3143 } else {
3144 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
3145 dn.data_blkaddr = ei.blk + index - ei.fofs;
3146 } else {
b4d07a3e 3147 /* hole case */
4d57b86d 3148 err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
4da7bf5a 3149 if (err || dn.data_blkaddr == NULL_ADDR) {
b4d07a3e 3150 f2fs_put_dnode(&dn);
59c9081b
YH		 3151 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO,
3152 true);
2866fb16 3153 WARN_ON(flag != F2FS_GET_BLOCK_PRE_AIO);
b4d07a3e
JK		 3154 locked = true;
3155 goto restart;
3156 }
2aadac08
JK		 3157 }
3158 }
b4d07a3e 3159
2aadac08
JK		 3160 /* convert_inline_page can make node_changed */
3161 *blk_addr = dn.data_blkaddr;
3162 *node_changed = dn.node_changed;
b4d07a3e 3163out:
2aadac08
JK		 3164 f2fs_put_dnode(&dn);
3165unlock_out:
b4d07a3e 3166 if (locked)
2866fb16 3167 __do_map_lock(sbi, flag, false);
2aadac08
JK		 3168 return err;
3169}
3170
eb47b800
JK		 3171static int f2fs_write_begin(struct file *file, struct address_space *mapping,
3172 loff_t pos, unsigned len, unsigned flags,
3173 struct page **pagep, void **fsdata)
3174{
3175 struct inode *inode = mapping->host;
4081363f 3176 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
86531d6b 3177 struct page *page = NULL;
09cbfeaf 3178 pgoff_t index = ((unsigned long long) pos) >> PAGE_SHIFT;
a2e2e76b 3179 bool need_balance = false, drop_atomic = false;
2aadac08 3180 block_t blkaddr = NULL_ADDR;
eb47b800
JK		 3181 int err = 0;
3182
62aed044
CY		 3183 trace_f2fs_write_begin(inode, pos, len, flags);
3184
00e09c0b
CY		 3185 if (!f2fs_is_checkpoint_ready(sbi)) {
3186 err = -ENOSPC;
4354994f 3187 goto fail;
00e09c0b 3188 }
4354994f 3189
455e3a58
JK		 3190 if ((f2fs_is_atomic_file(inode) &&
3191 !f2fs_available_free_memory(sbi, INMEM_PAGES)) ||
3192 is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
57864ae5 3193 err = -ENOMEM;
a2e2e76b 3194 drop_atomic = true;
57864ae5
JK		 3195 goto fail;
3196 }
3197
5f727395
JK		 3198 /*
3199 * We should check this at this moment to avoid deadlock on inode page
3200 * and #0 page. The locking rule for inline_data conversion should be:
3201 * lock_page(page #0) -> lock_page(inode_page)
3202 */
3203 if (index != 0) {
3204 err = f2fs_convert_inline_inode(inode);
3205 if (err)
3206 goto fail;
3207 }
4c8ff709
CY		 3208
3209#ifdef CONFIG_F2FS_FS_COMPRESSION
3210 if (f2fs_compressed_file(inode)) {
3211 int ret;
3212
3213 *fsdata = NULL;
3214
3215 ret = f2fs_prepare_compress_overwrite(inode, pagep,
3216 index, fsdata);
3217 if (ret < 0) {
3218 err = ret;
3219 goto fail;
3220 } else if (ret) {
3221 return 0;
3222 }
3223 }
3224#endif
3225
afcb7ca0 3226repeat:
86d54795
JK		 3227 /*
3228 * Do not use grab_cache_page_write_begin() to avoid deadlock due to
3229 * wait_for_stable_page. Will wait that below with our IO control.
3230 */
01eccef7 3231 page = f2fs_pagecache_get_page(mapping, index,
86d54795 3232 FGP_LOCK | FGP_WRITE | FGP_CREAT, GFP_NOFS);
3aab8f82
CY		 3233 if (!page) {
3234 err = -ENOMEM;
3235 goto fail;
3236 }
d5f66990 3237
4c8ff709
CY		 3238 /* TODO: cluster can be compressed due to race with .writepage */
3239
eb47b800
JK		 3240 *pagep = page;
3241
2aadac08
JK		 3242 err = prepare_write_begin(sbi, page, pos, len,
3243 &blkaddr, &need_balance);
9ba69cf9 3244 if (err)
2aadac08 3245 goto fail;
9ba69cf9 3246
af033b2a
CY		 3247 if (need_balance && !IS_NOQUOTA(inode) &&
3248 has_not_enough_free_secs(sbi, 0, 0)) {
2a340760 3249 unlock_page(page);
2c4db1a6 3250 f2fs_balance_fs(sbi, true);
2a340760
JK		 3251 lock_page(page);
3252 if (page->mapping != mapping) {
3253 /* The page got truncated from under us */
3254 f2fs_put_page(page, 1);
3255 goto repeat;
3256 }
3257 }
3258
bae0ee7a 3259 f2fs_wait_on_page_writeback(page, DATA, false, true);
b3d208f9 3260
649d7df2
JK		 3261 if (len == PAGE_SIZE || PageUptodate(page))
3262 return 0;
eb47b800 3263
95ae251f
EB		 3264 if (!(pos & (PAGE_SIZE - 1)) && (pos + len) >= i_size_read(inode) &&
3265 !f2fs_verity_in_progress(inode)) {
746e2403
YH		 3266 zero_user_segment(page, len, PAGE_SIZE);
3267 return 0;
3268 }
3269
2aadac08 3270 if (blkaddr == NEW_ADDR) {
09cbfeaf 3271 zero_user_segment(page, 0, PAGE_SIZE);
649d7df2 3272 SetPageUptodate(page);
eb47b800 3273 } else {
93770ab7
CY		 3274 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3275 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 3276 err = -EFSCORRUPTED;
93770ab7
CY		 3277 goto fail;
3278 }
13ba41e3
JK		 3279 err = f2fs_submit_page_read(inode, page, blkaddr);
3280 if (err)
78682f79 3281 goto fail;
d54c795b 3282
393ff91f 3283 lock_page(page);
6bacf52f 3284 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 3285 f2fs_put_page(page, 1);
3286 goto repeat;
eb47b800 3287 }
1563ac75
CY		 3288 if (unlikely(!PageUptodate(page))) {
3289 err = -EIO;
3290 goto fail;
4375a336 3291 }
eb47b800 3292 }
eb47b800 3293 return 0;
9ba69cf9 3294
3aab8f82 3295fail:
86531d6b 3296 f2fs_put_page(page, 1);
3aab8f82 3297 f2fs_write_failed(mapping, pos + len);
a2e2e76b 3298 if (drop_atomic)
4d57b86d 3299 f2fs_drop_inmem_pages_all(sbi, false);
3aab8f82 3300 return err;
eb47b800
JK		 3301}
3302
a1dd3c13
JK		 3303static int f2fs_write_end(struct file *file,
3304 struct address_space *mapping,
3305 loff_t pos, unsigned len, unsigned copied,
3306 struct page *page, void *fsdata)
3307{
3308 struct inode *inode = page->mapping->host;
3309
dfb2bf38
CY		 3310 trace_f2fs_write_end(inode, pos, len, copied);
3311
649d7df2
JK		 3312 /*
3313 * This should be come from len == PAGE_SIZE, and we expect copied
3314 * should be PAGE_SIZE. Otherwise, we treat it with zero copied and
3315 * let generic_perform_write() try to copy data again through copied=0.
3316 */
3317 if (!PageUptodate(page)) {
746e2403 3318 if (unlikely(copied != len))
649d7df2
JK		 3319 copied = 0;
3320 else
3321 SetPageUptodate(page);
3322 }
4c8ff709
CY		 3323
3324#ifdef CONFIG_F2FS_FS_COMPRESSION
3325 /* overwrite compressed file */
3326 if (f2fs_compressed_file(inode) && fsdata) {
3327 f2fs_compress_write_end(inode, fsdata, page->index, copied);
3328 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3329 return copied;
3330 }
3331#endif
3332
649d7df2
JK		 3333 if (!copied)
3334 goto unlock_out;
3335
34ba94ba 3336 set_page_dirty(page);
a1dd3c13 3337
95ae251f
EB		 3338 if (pos + copied > i_size_read(inode) &&
3339 !f2fs_verity_in_progress(inode))
fc9581c8 3340 f2fs_i_size_write(inode, pos + copied);
649d7df2 3341unlock_out:
3024c9a1 3342 f2fs_put_page(page, 1);
d0239e1b 3343 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
a1dd3c13
JK		 3344 return copied;
3345}
3346
6f673763
OS		 3347static int check_direct_IO(struct inode *inode, struct iov_iter *iter,
3348 loff_t offset)
944fcfc1 3349{
8a56dd96
JK		 3350 unsigned i_blkbits = READ_ONCE(inode->i_blkbits);
3351 unsigned blkbits = i_blkbits;
3352 unsigned blocksize_mask = (1 << blkbits) - 1;
3353 unsigned long align = offset | iov_iter_alignment(iter);
3354 struct block_device *bdev = inode->i_sb->s_bdev;
3355
3356 if (align & blocksize_mask) {
3357 if (bdev)
3358 blkbits = blksize_bits(bdev_logical_block_size(bdev));
3359 blocksize_mask = (1 << blkbits) - 1;
3360 if (align & blocksize_mask)
3361 return -EINVAL;
3362 return 1;
3363 }
944fcfc1
JK		 3364 return 0;
3365}
3366
02b16d0a
CY		 3367static void f2fs_dio_end_io(struct bio *bio)
3368{
3369 struct f2fs_private_dio *dio = bio->bi_private;
3370
3371 dec_page_count(F2FS_I_SB(dio->inode),
3372 dio->write ? F2FS_DIO_WRITE : F2FS_DIO_READ);
3373
3374 bio->bi_private = dio->orig_private;
3375 bio->bi_end_io = dio->orig_end_io;
3376
5222595d 3377 kvfree(dio);
02b16d0a
CY		 3378
3379 bio_endio(bio);
3380}
3381
3382static void f2fs_dio_submit_bio(struct bio *bio, struct inode *inode,
3383 loff_t file_offset)
3384{
3385 struct f2fs_private_dio *dio;
3386 bool write = (bio_op(bio) == REQ_OP_WRITE);
02b16d0a
CY		 3387
3388 dio = f2fs_kzalloc(F2FS_I_SB(inode),
3389 sizeof(struct f2fs_private_dio), GFP_NOFS);
8e114038 3390 if (!dio)
02b16d0a 3391 goto out;
02b16d0a
CY		 3392
3393 dio->inode = inode;
3394 dio->orig_end_io = bio->bi_end_io;
3395 dio->orig_private = bio->bi_private;
3396 dio->write = write;
3397
3398 bio->bi_end_io = f2fs_dio_end_io;
3399 bio->bi_private = dio;
3400
3401 inc_page_count(F2FS_I_SB(inode),
3402 write ? F2FS_DIO_WRITE : F2FS_DIO_READ);
3403
3404 submit_bio(bio);
3405 return;
3406out:
3407 bio->bi_status = BLK_STS_IOERR;
3408 bio_endio(bio);
3409}
3410
c8b8e32d 3411static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
eb47b800 3412{
b439b103 3413 struct address_space *mapping = iocb->ki_filp->f_mapping;
3aab8f82 3414 struct inode *inode = mapping->host;
0cdd3195 3415 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
f847c699 3416 struct f2fs_inode_info *fi = F2FS_I(inode);
3aab8f82 3417 size_t count = iov_iter_count(iter);
c8b8e32d 3418 loff_t offset = iocb->ki_pos;
82e0a5aa 3419 int rw = iov_iter_rw(iter);
3aab8f82 3420 int err;
0cdd3195 3421 enum rw_hint hint = iocb->ki_hint;
63189b78 3422 int whint_mode = F2FS_OPTION(sbi).whint_mode;
f847c699 3423 bool do_opu;
944fcfc1 3424
b439b103 3425 err = check_direct_IO(inode, iter, offset);
b9d777b8 3426 if (err)
8a56dd96 3427 return err < 0 ? err : 0;
9ffe0fb5 3428
f847c699 3429 if (f2fs_force_buffered_io(inode, iocb, iter))
36abef4e 3430 return 0;
fcc85a4d 3431
f847c699
CY		 3432 do_opu = allow_outplace_dio(inode, iocb, iter);
3433
5302fb00 3434 trace_f2fs_direct_IO_enter(inode, offset, count, rw);
70407fad 3435
0cdd3195
HL		 3436 if (rw == WRITE && whint_mode == WHINT_MODE_OFF)
3437 iocb->ki_hint = WRITE_LIFE_NOT_SET;
3438
f847c699
CY		 3439 if (iocb->ki_flags & IOCB_NOWAIT) {
3440 if (!down_read_trylock(&fi->i_gc_rwsem[rw])) {
3441 iocb->ki_hint = hint;
3442 err = -EAGAIN;
3443 goto out;
3444 }
3445 if (do_opu && !down_read_trylock(&fi->i_gc_rwsem[READ])) {
3446 up_read(&fi->i_gc_rwsem[rw]);
b91050a8
HL		 3447 iocb->ki_hint = hint;
3448 err = -EAGAIN;
3449 goto out;
3450 }
f847c699
CY		 3451 } else {
3452 down_read(&fi->i_gc_rwsem[rw]);
3453 if (do_opu)
3454 down_read(&fi->i_gc_rwsem[READ]);
b91050a8
HL		 3455 }
3456
02b16d0a 3457 err = __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev,
f9d6d059
CY		 3458 iter, rw == WRITE ? get_data_block_dio_write :
3459 get_data_block_dio, NULL, f2fs_dio_submit_bio,
02b16d0a 3460 DIO_LOCKING | DIO_SKIP_HOLES);
f847c699
CY		 3461
3462 if (do_opu)
3463 up_read(&fi->i_gc_rwsem[READ]);
3464
3465 up_read(&fi->i_gc_rwsem[rw]);
82e0a5aa
CY		 3466
3467 if (rw == WRITE) {
0cdd3195
HL		 3468 if (whint_mode == WHINT_MODE_OFF)
3469 iocb->ki_hint = hint;
b0af6d49
CY		 3470 if (err > 0) {
3471 f2fs_update_iostat(F2FS_I_SB(inode), APP_DIRECT_IO,
3472 err);
f847c699
CY		 3473 if (!do_opu)
3474 set_inode_flag(inode, FI_UPDATE_WRITE);
b0af6d49 3475 } else if (err < 0) {
6bfc4919 3476 f2fs_write_failed(mapping, offset + count);
b0af6d49 3477 }
6bfc4919 3478 }
70407fad 3479
b91050a8 3480out:
5302fb00 3481 trace_f2fs_direct_IO_exit(inode, offset, count, rw, err);
70407fad 3482
3aab8f82 3483 return err;
eb47b800
JK		 3484}
3485
487261f3
CY		 3486void f2fs_invalidate_page(struct page *page, unsigned int offset,
3487 unsigned int length)
eb47b800
JK		 3488{
3489 struct inode *inode = page->mapping->host;
487261f3 3490 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
a7ffdbe2 3491
487261f3 3492 if (inode->i_ino >= F2FS_ROOT_INO(sbi) &&
09cbfeaf 3493 (offset % PAGE_SIZE || length != PAGE_SIZE))
a7ffdbe2
JK		 3494 return;
3495
487261f3 3496 if (PageDirty(page)) {
933439c8 3497 if (inode->i_ino == F2FS_META_INO(sbi)) {
487261f3 3498 dec_page_count(sbi, F2FS_DIRTY_META);
933439c8 3499 } else if (inode->i_ino == F2FS_NODE_INO(sbi)) {
487261f3 3500 dec_page_count(sbi, F2FS_DIRTY_NODES);
933439c8 3501 } else {
487261f3 3502 inode_dec_dirty_pages(inode);
4d57b86d 3503 f2fs_remove_dirty_inode(inode);
933439c8 3504 }
487261f3 3505 }
decd36b6 3506
2baf0781
CY		 3507 clear_cold_data(page);
3508
decd36b6 3509 if (IS_ATOMIC_WRITTEN_PAGE(page))
4d57b86d 3510 return f2fs_drop_inmem_page(inode, page);
decd36b6 3511
240a5915 3512 f2fs_clear_page_private(page);
eb47b800
JK		 3513}
3514
487261f3 3515int f2fs_release_page(struct page *page, gfp_t wait)
eb47b800 3516{
f68daeeb
JK		 3517 /* If this is dirty page, keep PagePrivate */
3518 if (PageDirty(page))
3519 return 0;
3520
decd36b6
CY		 3521 /* This is atomic written page, keep Private */
3522 if (IS_ATOMIC_WRITTEN_PAGE(page))
3523 return 0;
3524
2baf0781 3525 clear_cold_data(page);
240a5915 3526 f2fs_clear_page_private(page);
c3850aa1 3527 return 1;
eb47b800
JK		 3528}
3529
3530static int f2fs_set_data_page_dirty(struct page *page)
3531{
4969c06a 3532 struct inode *inode = page_file_mapping(page)->host;
eb47b800 3533
26c6b887
JK		 3534 trace_f2fs_set_page_dirty(page, DATA);
3535
237c0790
JK		 3536 if (!PageUptodate(page))
3537 SetPageUptodate(page);
4969c06a
JK		 3538 if (PageSwapCache(page))
3539 return __set_page_dirty_nobuffers(page);
34ba94ba 3540
5fe45743 3541 if (f2fs_is_atomic_file(inode) && !f2fs_is_commit_atomic_write(inode)) {
decd36b6 3542 if (!IS_ATOMIC_WRITTEN_PAGE(page)) {
4d57b86d 3543 f2fs_register_inmem_page(inode, page);
decd36b6
CY		 3544 return 1;
3545 }
3546 /*
3547 * Previously, this page has been registered, we just
3548 * return here.
3549 */
3550 return 0;
34ba94ba
JK		 3551 }
3552
eb47b800 3553 if (!PageDirty(page)) {
b87078ad 3554 __set_page_dirty_nobuffers(page);
4d57b86d 3555 f2fs_update_dirty_page(inode, page);
eb47b800
JK		 3556 return 1;
3557 }
3558 return 0;
3559}
3560
c01e54b7
JK		 3561static sector_t f2fs_bmap(struct address_space *mapping, sector_t block)
3562{
454ae7e5
CY		 3563 struct inode *inode = mapping->host;
3564
1d373a0e
JK		 3565 if (f2fs_has_inline_data(inode))
3566 return 0;
3567
3568 /* make sure allocating whole blocks */
3569 if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
3570 filemap_write_and_wait(mapping);
3571
e2b4e2bc 3572 return generic_block_bmap(mapping, block, get_data_block_bmap);
429511cd
CY		 3573}
3574
5b7a487c
WG		 3575#ifdef CONFIG_MIGRATION
3576#include <linux/migrate.h>
3577
3578int f2fs_migrate_page(struct address_space *mapping,
3579 struct page *newpage, struct page *page, enum migrate_mode mode)
3580{
3581 int rc, extra_count;
3582 struct f2fs_inode_info *fi = F2FS_I(mapping->host);
3583 bool atomic_written = IS_ATOMIC_WRITTEN_PAGE(page);
3584
3585 BUG_ON(PageWriteback(page));
3586
3587 /* migrating an atomic written page is safe with the inmem_lock hold */
ff1048e7
JK		 3588 if (atomic_written) {
3589 if (mode != MIGRATE_SYNC)
3590 return -EBUSY;
3591 if (!mutex_trylock(&fi->inmem_lock))
3592 return -EAGAIN;
3593 }
5b7a487c 3594
240a5915
CY		 3595 /* one extra reference was held for atomic_write page */
3596 extra_count = atomic_written ? 1 : 0;
5b7a487c 3597 rc = migrate_page_move_mapping(mapping, newpage,
37109694 3598 page, extra_count);
5b7a487c
WG		 3599 if (rc != MIGRATEPAGE_SUCCESS) {
3600 if (atomic_written)
3601 mutex_unlock(&fi->inmem_lock);
3602 return rc;
3603 }
3604
3605 if (atomic_written) {
3606 struct inmem_pages *cur;
3607 list_for_each_entry(cur, &fi->inmem_pages, list)
3608 if (cur->page == page) {
3609 cur->page = newpage;
3610 break;
3611 }
3612 mutex_unlock(&fi->inmem_lock);
3613 put_page(page);
3614 get_page(newpage);
3615 }
3616
240a5915
CY		 3617 if (PagePrivate(page)) {
3618 f2fs_set_page_private(newpage, page_private(page));
3619 f2fs_clear_page_private(page);
3620 }
5b7a487c 3621
2916ecc0
JG		 3622 if (mode != MIGRATE_SYNC_NO_COPY)
3623 migrate_page_copy(newpage, page);
3624 else
3625 migrate_page_states(newpage, page);
5b7a487c
WG		 3626
3627 return MIGRATEPAGE_SUCCESS;
3628}
3629#endif
3630
4969c06a
JK		 3631#ifdef CONFIG_SWAP
3632/* Copied from generic_swapfile_activate() to check any holes */
3e5e479a
CY		 3633static int check_swap_activate(struct swap_info_struct *sis,
3634 struct file *swap_file, sector_t *span)
4969c06a
JK		 3635{
3636 struct address_space *mapping = swap_file->f_mapping;
3637 struct inode *inode = mapping->host;
3638 unsigned blocks_per_page;
3639 unsigned long page_no;
3640 unsigned blkbits;
3641 sector_t probe_block;
3642 sector_t last_block;
3643 sector_t lowest_block = -1;
3644 sector_t highest_block = 0;
3e5e479a
CY		 3645 int nr_extents = 0;
3646 int ret;
4969c06a
JK		 3647
3648 blkbits = inode->i_blkbits;
3649 blocks_per_page = PAGE_SIZE >> blkbits;
3650
3651 /*
3652 * Map all the blocks into the extent list. This code doesn't try
3653 * to be very smart.
3654 */
3655 probe_block = 0;
3656 page_no = 0;
3657 last_block = i_size_read(inode) >> blkbits;
3e5e479a
CY		 3658 while ((probe_block + blocks_per_page) <= last_block &&
3659 page_no < sis->max) {
4969c06a
JK		 3660 unsigned block_in_page;
3661 sector_t first_block;
30460e1e
CM		 3662 sector_t block = 0;
3663 int err = 0;
4969c06a
JK		 3664
3665 cond_resched();
3666
30460e1e
CM		 3667 block = probe_block;
3668 err = bmap(inode, &block);
3669 if (err || !block)
4969c06a 3670 goto bad_bmap;
30460e1e 3671 first_block = block;
4969c06a
JK		 3672
3673 /*
3674 * It must be PAGE_SIZE aligned on-disk
3675 */
3676 if (first_block & (blocks_per_page - 1)) {
3677 probe_block++;
3678 goto reprobe;
3679 }
3680
3681 for (block_in_page = 1; block_in_page < blocks_per_page;
3682 block_in_page++) {
4969c06a 3683
30460e1e
CM		 3684 block = probe_block + block_in_page;
3685 err = bmap(inode, &block);
3686
3687 if (err || !block)
4969c06a 3688 goto bad_bmap;
30460e1e 3689
4969c06a
JK		 3690 if (block != first_block + block_in_page) {
3691 /* Discontiguity */
3692 probe_block++;
3693 goto reprobe;
3694 }
3695 }
3696
3697 first_block >>= (PAGE_SHIFT - blkbits);
3698 if (page_no) { /* exclude the header page */
3699 if (first_block < lowest_block)
3700 lowest_block = first_block;
3701 if (first_block > highest_block)
3702 highest_block = first_block;
3703 }
3704
3e5e479a
CY		 3705 /*
3706 * We found a PAGE_SIZE-length, PAGE_SIZE-aligned run of blocks
3707 */
3708 ret = add_swap_extent(sis, page_no, 1, first_block);
3709 if (ret < 0)
3710 goto out;
3711 nr_extents += ret;
4969c06a
JK		 3712 page_no++;
3713 probe_block += blocks_per_page;
3714reprobe:
3715 continue;
3716 }
3e5e479a
CY		 3717 ret = nr_extents;
3718 *span = 1 + highest_block - lowest_block;
3719 if (page_no == 0)
3720 page_no = 1; /* force Empty message */
3721 sis->max = page_no;
3722 sis->pages = page_no - 1;
3723 sis->highest_bit = page_no - 1;
3724out:
3725 return ret;
4969c06a
JK		 3726bad_bmap:
3727 pr_err("swapon: swapfile has holes\n");
3728 return -EINVAL;
3729}
3730
3731static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
3732 sector_t *span)
3733{
3734 struct inode *inode = file_inode(file);
3735 int ret;
3736
3737 if (!S_ISREG(inode->i_mode))
3738 return -EINVAL;
3739
3740 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3741 return -EROFS;
3742
3743 ret = f2fs_convert_inline_inode(inode);
3744 if (ret)
3745 return ret;
3746
4c8ff709
CY		 3747 if (f2fs_disable_compressed_file(inode))
3748 return -EINVAL;
3749
3e5e479a
CY		 3750 ret = check_swap_activate(sis, file, span);
3751 if (ret < 0)
4969c06a
JK		 3752 return ret;
3753
3754 set_inode_flag(inode, FI_PIN_FILE);
3755 f2fs_precache_extents(inode);
3756 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3e5e479a 3757 return ret;
4969c06a
JK		 3758}
3759
3760static void f2fs_swap_deactivate(struct file *file)
3761{
3762 struct inode *inode = file_inode(file);
3763
3764 clear_inode_flag(inode, FI_PIN_FILE);
3765}
3766#else
3767static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
3768 sector_t *span)
3769{
3770 return -EOPNOTSUPP;
3771}
3772
3773static void f2fs_swap_deactivate(struct file *file)
3774{
3775}
3776#endif
3777
eb47b800
JK		 3778const struct address_space_operations f2fs_dblock_aops = {
3779 .readpage = f2fs_read_data_page,
3780 .readpages = f2fs_read_data_pages,
3781 .writepage = f2fs_write_data_page,
3782 .writepages = f2fs_write_data_pages,
3783 .write_begin = f2fs_write_begin,
a1dd3c13 3784 .write_end = f2fs_write_end,
eb47b800 3785 .set_page_dirty = f2fs_set_data_page_dirty,
487261f3
CY		 3786 .invalidatepage = f2fs_invalidate_page,
3787 .releasepage = f2fs_release_page,
eb47b800 3788 .direct_IO = f2fs_direct_IO,
c01e54b7 3789 .bmap = f2fs_bmap,
4969c06a
JK		 3790 .swap_activate = f2fs_swap_activate,
3791 .swap_deactivate = f2fs_swap_deactivate,
5b7a487c
WG		 3792#ifdef CONFIG_MIGRATION
3793 .migratepage = f2fs_migrate_page,
3794#endif
eb47b800 3795};
6dbb1796 3796
5ec2d99d 3797void f2fs_clear_page_cache_dirty_tag(struct page *page)
aec2f729
CY		 3798{
3799 struct address_space *mapping = page_mapping(page);
3800 unsigned long flags;
3801
3802 xa_lock_irqsave(&mapping->i_pages, flags);
5ec2d99d 3803 __xa_clear_mark(&mapping->i_pages, page_index(page),
aec2f729
CY		 3804 PAGECACHE_TAG_DIRTY);
3805 xa_unlock_irqrestore(&mapping->i_pages, flags);
3806}
3807
6dbb1796
EB		 3808int __init f2fs_init_post_read_processing(void)
3809{
95ae251f
EB		 3810 bio_post_read_ctx_cache =
3811 kmem_cache_create("f2fs_bio_post_read_ctx",
3812 sizeof(struct bio_post_read_ctx), 0, 0, NULL);
6dbb1796
EB		 3813 if (!bio_post_read_ctx_cache)
3814 goto fail;
3815 bio_post_read_ctx_pool =
3816 mempool_create_slab_pool(NUM_PREALLOC_POST_READ_CTXS,
3817 bio_post_read_ctx_cache);
3818 if (!bio_post_read_ctx_pool)
3819 goto fail_free_cache;
3820 return 0;
3821
3822fail_free_cache:
3823 kmem_cache_destroy(bio_post_read_ctx_cache);
3824fail:
3825 return -ENOMEM;
3826}
3827
0b20fcec 3828void f2fs_destroy_post_read_processing(void)
6dbb1796
EB		 3829{
3830 mempool_destroy(bio_post_read_ctx_pool);
3831 kmem_cache_destroy(bio_post_read_ctx_cache);
3832}
0b20fcec 3833
4c8ff709
CY		 3834int f2fs_init_post_read_wq(struct f2fs_sb_info *sbi)
3835{
3836 if (!f2fs_sb_has_encrypt(sbi) &&
3837 !f2fs_sb_has_verity(sbi) &&
3838 !f2fs_sb_has_compression(sbi))
3839 return 0;
3840
3841 sbi->post_read_wq = alloc_workqueue("f2fs_post_read_wq",
3842 WQ_UNBOUND | WQ_HIGHPRI,
3843 num_online_cpus());
3844 if (!sbi->post_read_wq)
3845 return -ENOMEM;
3846 return 0;
3847}
3848
3849void f2fs_destroy_post_read_wq(struct f2fs_sb_info *sbi)
3850{
3851 if (sbi->post_read_wq)
3852 destroy_workqueue(sbi->post_read_wq);
3853}
3854
0b20fcec
CY		 3855int __init f2fs_init_bio_entry_cache(void)
3856{
3857 bio_entry_slab = f2fs_kmem_cache_create("bio_entry_slab",
3858 sizeof(struct bio_entry));
3859 if (!bio_entry_slab)
3860 return -ENOMEM;
3861 return 0;
3862}
3863
f543805f 3864void f2fs_destroy_bio_entry_cache(void)
0b20fcec
CY		 3865{
3866 kmem_cache_destroy(bio_entry_slab);
3867}
Michael's Linux tree.