projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: cleanup duplicate stats for atomic files
[linux-block.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
EB		 33static mempool_t *bio_post_read_ctx_pool;
34
36951b38
CY		 35static bool __is_cp_guaranteed(struct page *page)
36{
37 struct address_space *mapping = page->mapping;
38 struct inode *inode;
39 struct f2fs_sb_info *sbi;
40
41 if (!mapping)
42 return false;
43
44 inode = mapping->host;
45 sbi = F2FS_I_SB(inode);
46
47 if (inode->i_ino == F2FS_META_INO(sbi) ||
48 inode->i_ino == F2FS_NODE_INO(sbi) ||
49 S_ISDIR(inode->i_mode) ||
e7a4feb0 50 (S_ISREG(inode->i_mode) &&
af033b2a 51 (f2fs_is_atomic_file(inode) || IS_NOQUOTA(inode))) ||
36951b38
CY		 52 is_cold_data(page))
53 return true;
54 return false;
55}
56
5f9abab4
JK		 57static enum count_type __read_io_type(struct page *page)
58{
4969c06a 59 struct address_space *mapping = page_file_mapping(page);
5f9abab4
JK		 60
61 if (mapping) {
62 struct inode *inode = mapping->host;
63 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
64
65 if (inode->i_ino == F2FS_META_INO(sbi))
66 return F2FS_RD_META;
67
68 if (inode->i_ino == F2FS_NODE_INO(sbi))
69 return F2FS_RD_NODE;
70 }
71 return F2FS_RD_DATA;
72}
73
6dbb1796
EB		 74/* postprocessing steps for read bios */
75enum bio_post_read_step {
76 STEP_INITIAL = 0,
77 STEP_DECRYPT,
95ae251f 78 STEP_VERITY,
6dbb1796
EB		 79};
80
81struct bio_post_read_ctx {
82 struct bio *bio;
83 struct work_struct work;
84 unsigned int cur_step;
85 unsigned int enabled_steps;
86};
87
88static void __read_end_io(struct bio *bio)
93dfe2ac 89{
6dbb1796
EB		 90 struct page *page;
91 struct bio_vec *bv;
6dc4f100 92 struct bvec_iter_all iter_all;
93dfe2ac 93
2b070cfe 94 bio_for_each_segment_all(bv, bio, iter_all) {
6dbb1796
EB		 95 page = bv->bv_page;
96
97 /* PG_error was set if any post_read step failed */
98 if (bio->bi_status || PageError(page)) {
99 ClearPageUptodate(page);
fb7d70db
JK		 100 /* will re-read again later */
101 ClearPageError(page);
6dbb1796
EB		 102 } else {
103 SetPageUptodate(page);
104 }
5f9abab4 105 dec_page_count(F2FS_P_SB(page), __read_io_type(page));
6dbb1796
EB		 106 unlock_page(page);
107 }
108 if (bio->bi_private)
109 mempool_free(bio->bi_private, bio_post_read_ctx_pool);
110 bio_put(bio);
111}
112
113static void bio_post_read_processing(struct bio_post_read_ctx *ctx);
114
115static void decrypt_work(struct work_struct *work)
116{
117 struct bio_post_read_ctx *ctx =
118 container_of(work, struct bio_post_read_ctx, work);
119
120 fscrypt_decrypt_bio(ctx->bio);
121
122 bio_post_read_processing(ctx);
123}
124
95ae251f
EB		 125static void verity_work(struct work_struct *work)
126{
127 struct bio_post_read_ctx *ctx =
128 container_of(work, struct bio_post_read_ctx, work);
129
130 fsverity_verify_bio(ctx->bio);
131
132 bio_post_read_processing(ctx);
133}
134
6dbb1796
EB		 135static void bio_post_read_processing(struct bio_post_read_ctx *ctx)
136{
95ae251f
EB		 137 /*
138 * We use different work queues for decryption and for verity because
139 * verity may require reading metadata pages that need decryption, and
140 * we shouldn't recurse to the same workqueue.
141 */
6dbb1796
EB		 142 switch (++ctx->cur_step) {
143 case STEP_DECRYPT:
144 if (ctx->enabled_steps & (1 << STEP_DECRYPT)) {
145 INIT_WORK(&ctx->work, decrypt_work);
146 fscrypt_enqueue_decrypt_work(&ctx->work);
147 return;
148 }
149 ctx->cur_step++;
150 /* fall-through */
95ae251f
EB		 151 case STEP_VERITY:
152 if (ctx->enabled_steps & (1 << STEP_VERITY)) {
153 INIT_WORK(&ctx->work, verity_work);
154 fsverity_enqueue_verify_work(&ctx->work);
155 return;
156 }
157 ctx->cur_step++;
158 /* fall-through */
6dbb1796
EB		 159 default:
160 __read_end_io(ctx->bio);
161 }
162}
163
164static bool f2fs_bio_post_read_required(struct bio *bio)
165{
166 return bio->bi_private && !bio->bi_status;
167}
168
169static void f2fs_read_end_io(struct bio *bio)
170{
c45d6002
CY		 171 struct f2fs_sb_info *sbi = F2FS_P_SB(bio_first_page_all(bio));
172
173 if (time_to_inject(sbi, FAULT_READ_IO)) {
174 f2fs_show_injection_info(sbi, FAULT_READ_IO);
4e4cbee9 175 bio->bi_status = BLK_STS_IOERR;
55523519 176 }
8b038c70 177
6dbb1796
EB		 178 if (f2fs_bio_post_read_required(bio)) {
179 struct bio_post_read_ctx *ctx = bio->bi_private;
f1e88660 180
6dbb1796
EB		 181 ctx->cur_step = STEP_INITIAL;
182 bio_post_read_processing(ctx);
183 return;
f1e88660 184 }
6dbb1796
EB		 185
186 __read_end_io(bio);
f1e88660
JK		 187}
188
4246a0b6 189static void f2fs_write_end_io(struct bio *bio)
93dfe2ac 190{
1b1f559f 191 struct f2fs_sb_info *sbi = bio->bi_private;
f568849e 192 struct bio_vec *bvec;
6dc4f100 193 struct bvec_iter_all iter_all;
93dfe2ac 194
6f5c2ed0 195 if (time_to_inject(sbi, FAULT_WRITE_IO)) {
c45d6002 196 f2fs_show_injection_info(sbi, FAULT_WRITE_IO);
6f5c2ed0
CY		 197 bio->bi_status = BLK_STS_IOERR;
198 }
199
2b070cfe 200 bio_for_each_segment_all(bvec, bio, iter_all) {
93dfe2ac 201 struct page *page = bvec->bv_page;
36951b38 202 enum count_type type = WB_DATA_TYPE(page);
93dfe2ac 203
0a595eba
JK		 204 if (IS_DUMMY_WRITTEN_PAGE(page)) {
205 set_page_private(page, (unsigned long)NULL);
206 ClearPagePrivate(page);
207 unlock_page(page);
208 mempool_free(page, sbi->write_io_dummy);
209
4e4cbee9 210 if (unlikely(bio->bi_status))
0a595eba
JK		 211 f2fs_stop_checkpoint(sbi, true);
212 continue;
213 }
214
d2d0727b 215 fscrypt_finalize_bounce_page(&page);
4375a336 216
4e4cbee9 217 if (unlikely(bio->bi_status)) {
5114a97a 218 mapping_set_error(page->mapping, -EIO);
b1ca321d
JK		 219 if (type == F2FS_WB_CP_DATA)
220 f2fs_stop_checkpoint(sbi, true);
93dfe2ac 221 }
7dff55d2
YH		 222
223 f2fs_bug_on(sbi, page->mapping == NODE_MAPPING(sbi) &&
224 page->index != nid_of_node(page));
225
36951b38 226 dec_page_count(sbi, type);
50fa53ec
CY		 227 if (f2fs_in_warm_node_list(sbi, page))
228 f2fs_del_fsync_node_entry(sbi, page);
36951b38 229 clear_cold_data(page);
93dfe2ac 230 end_page_writeback(page);
f568849e 231 }
36951b38 232 if (!get_pages(sbi, F2FS_WB_CP_DATA) &&
f5730184 233 wq_has_sleeper(&sbi->cp_wait))
93dfe2ac
JK		 234 wake_up(&sbi->cp_wait);
235
236 bio_put(bio);
237}
238
3c62be17
JK		 239/*
240 * Return true, if pre_bio's bdev is same as its target device.
241 */
242struct block_device *f2fs_target_device(struct f2fs_sb_info *sbi,
243 block_t blk_addr, struct bio *bio)
244{
245 struct block_device *bdev = sbi->sb->s_bdev;
246 int i;
247
0916878d
DLM		 248 if (f2fs_is_multi_device(sbi)) {
249 for (i = 0; i < sbi->s_ndevs; i++) {
250 if (FDEV(i).start_blk <= blk_addr &&
251 FDEV(i).end_blk >= blk_addr) {
252 blk_addr -= FDEV(i).start_blk;
253 bdev = FDEV(i).bdev;
254 break;
255 }
3c62be17
JK		 256 }
257 }
258 if (bio) {
74d46992 259 bio_set_dev(bio, bdev);
3c62be17
JK		 260 bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr);
261 }
262 return bdev;
263}
264
265int f2fs_target_device_index(struct f2fs_sb_info *sbi, block_t blkaddr)
266{
267 int i;
268
0916878d
DLM		 269 if (!f2fs_is_multi_device(sbi))
270 return 0;
271
3c62be17
JK		 272 for (i = 0; i < sbi->s_ndevs; i++)
273 if (FDEV(i).start_blk <= blkaddr && FDEV(i).end_blk >= blkaddr)
274 return i;
275 return 0;
276}
277
278static bool __same_bdev(struct f2fs_sb_info *sbi,
279 block_t blk_addr, struct bio *bio)
280{
74d46992
CH		 281 struct block_device *b = f2fs_target_device(sbi, blk_addr, NULL);
282 return bio->bi_disk == b->bd_disk && bio->bi_partno == b->bd_partno;
3c62be17
JK		 283}
284
940a6d34
GZ		 285/*
286 * Low-level block read/write IO operations.
287 */
b757f6ed 288static struct bio *__bio_alloc(struct f2fs_io_info *fio, int npages)
940a6d34 289{
b757f6ed 290 struct f2fs_sb_info *sbi = fio->sbi;
940a6d34
GZ		 291 struct bio *bio;
292
d62fe971 293 bio = f2fs_bio_alloc(sbi, npages, true);
940a6d34 294
b757f6ed
CY		 295 f2fs_target_device(sbi, fio->new_blkaddr, bio);
296 if (is_read_io(fio->op)) {
0cdd3195
HL		 297 bio->bi_end_io = f2fs_read_end_io;
298 bio->bi_private = NULL;
299 } else {
300 bio->bi_end_io = f2fs_write_end_io;
301 bio->bi_private = sbi;
b757f6ed
CY		 302 bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi,
303 fio->type, fio->temp);
0cdd3195 304 }
b757f6ed
CY		 305 if (fio->io_wbc)
306 wbc_init_bio(fio->io_wbc, bio);
940a6d34
GZ		 307
308 return bio;
309}
310
4fc29c1a
LT		 311static inline void __submit_bio(struct f2fs_sb_info *sbi,
312 struct bio *bio, enum page_type type)
f5730184 313{
4fc29c1a 314 if (!is_read_io(bio_op(bio))) {
0a595eba
JK		 315 unsigned int start;
316
0a595eba
JK		 317 if (type != DATA && type != NODE)
318 goto submit_io;
319
66415cee 320 if (test_opt(sbi, LFS) && current->plug)
3bb09a0e
TY		 321 blk_finish_plug(current->plug);
322
8223ecc4
CY		 323 if (F2FS_IO_ALIGNED(sbi))
324 goto submit_io;
325
0a595eba
JK		 326 start = bio->bi_iter.bi_size >> F2FS_BLKSIZE_BITS;
327 start %= F2FS_IO_SIZE(sbi);
328
329 if (start == 0)
330 goto submit_io;
331
332 /* fill dummy pages */
333 for (; start < F2FS_IO_SIZE(sbi); start++) {
334 struct page *page =
335 mempool_alloc(sbi->write_io_dummy,
bc73a4b2 336 GFP_NOIO | __GFP_NOFAIL);
0a595eba
JK		 337 f2fs_bug_on(sbi, !page);
338
bc73a4b2 339 zero_user_segment(page, 0, PAGE_SIZE);
0a595eba
JK		 340 SetPagePrivate(page);
341 set_page_private(page, (unsigned long)DUMMY_WRITTEN_PAGE);
342 lock_page(page);
343 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE)
344 f2fs_bug_on(sbi, 1);
345 }
346 /*
347 * In the NODE case, we lose next block address chain. So, we
348 * need to do checkpoint in f2fs_sync_file.
349 */
350 if (type == NODE)
351 set_sbi_flag(sbi, SBI_NEED_CP);
19a5f5e2 352 }
0a595eba 353submit_io:
554b5125
JK		 354 if (is_read_io(bio_op(bio)))
355 trace_f2fs_submit_read_bio(sbi->sb, type, bio);
356 else
357 trace_f2fs_submit_write_bio(sbi->sb, type, bio);
4e49ea4a 358 submit_bio(bio);
f5730184
JK		 359}
360
458e6197 361static void __submit_merged_bio(struct f2fs_bio_info *io)
93dfe2ac 362{
458e6197 363 struct f2fs_io_info *fio = &io->fio;
93dfe2ac
JK		 364
365 if (!io->bio)
366 return;
367
554b5125
JK		 368 bio_set_op_attrs(io->bio, fio->op, fio->op_flags);
369
04d328de 370 if (is_read_io(fio->op))
554b5125 371 trace_f2fs_prepare_read_bio(io->sbi->sb, fio->type, io->bio);
6a8f8ca5 372 else
554b5125 373 trace_f2fs_prepare_write_bio(io->sbi->sb, fio->type, io->bio);
04d328de 374
4fc29c1a 375 __submit_bio(io->sbi, io->bio, fio->type);
93dfe2ac
JK		 376 io->bio = NULL;
377}
378
8648de2c 379static bool __has_merged_page(struct bio *bio, struct inode *inode,
bab475c5 380 struct page *page, nid_t ino)
0fd785eb 381{
0fd785eb
CY		 382 struct bio_vec *bvec;
383 struct page *target;
6dc4f100 384 struct bvec_iter_all iter_all;
0fd785eb 385
8648de2c 386 if (!bio)
0fd785eb 387 return false;
0c3a5797 388
bab475c5 389 if (!inode && !page && !ino)
0c3a5797 390 return true;
0fd785eb 391
8648de2c 392 bio_for_each_segment_all(bvec, bio, iter_all) {
0fd785eb 393
d2d0727b
EB		 394 target = bvec->bv_page;
395 if (fscrypt_is_bounce_page(target))
396 target = fscrypt_pagecache_page(target);
0fd785eb 397
0c3a5797
CY		 398 if (inode && inode == target->mapping->host)
399 return true;
bab475c5
CY		 400 if (page && page == target)
401 return true;
0c3a5797 402 if (ino && ino == ino_of_node(target))
0fd785eb 403 return true;
0fd785eb
CY		 404 }
405
0fd785eb
CY		 406 return false;
407}
408
b9109b0e 409static void __f2fs_submit_merged_write(struct f2fs_sb_info *sbi,
a912b54d 410 enum page_type type, enum temp_type temp)
93dfe2ac
JK		 411{
412 enum page_type btype = PAGE_TYPE_OF_BIO(type);
a912b54d 413 struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
93dfe2ac 414
df0f8dc0 415 down_write(&io->io_rwsem);
458e6197
JK		 416
417 /* change META to META_FLUSH in the checkpoint procedure */
418 if (type >= META_FLUSH) {
419 io->fio.type = META_FLUSH;
04d328de 420 io->fio.op = REQ_OP_WRITE;
3adc5fcb 421 io->fio.op_flags = REQ_META | REQ_PRIO | REQ_SYNC;
70fd7614 422 if (!test_opt(sbi, NOBARRIER))
7f54f51f 423 io->fio.op_flags |= REQ_PREFLUSH | REQ_FUA;
458e6197
JK		 424 }
425 __submit_merged_bio(io);
df0f8dc0 426 up_write(&io->io_rwsem);
93dfe2ac
JK		 427}
428
a912b54d 429static void __submit_merged_write_cond(struct f2fs_sb_info *sbi,
bab475c5
CY		 430 struct inode *inode, struct page *page,
431 nid_t ino, enum page_type type, bool force)
0c3a5797 432{
a912b54d 433 enum temp_type temp;
1e771e83 434 bool ret = true;
a912b54d
JK		 435
436 for (temp = HOT; temp < NR_TEMP_TYPE; temp++) {
1e771e83
YS		 437 if (!force) {
438 enum page_type btype = PAGE_TYPE_OF_BIO(type);
439 struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
a912b54d 440
1e771e83 441 down_read(&io->io_rwsem);
8648de2c 442 ret = __has_merged_page(io->bio, inode, page, ino);
1e771e83
YS		 443 up_read(&io->io_rwsem);
444 }
445 if (ret)
446 __f2fs_submit_merged_write(sbi, type, temp);
a912b54d
JK		 447
448 /* TODO: use HOT temp only for meta pages now. */
449 if (type >= META)
450 break;
451 }
0c3a5797
CY		 452}
453
b9109b0e 454void f2fs_submit_merged_write(struct f2fs_sb_info *sbi, enum page_type type)
0c3a5797 455{
adcc00f7 456 __submit_merged_write_cond(sbi, NULL, NULL, 0, type, true);
0c3a5797
CY		 457}
458
b9109b0e 459void f2fs_submit_merged_write_cond(struct f2fs_sb_info *sbi,
bab475c5
CY		 460 struct inode *inode, struct page *page,
461 nid_t ino, enum page_type type)
0c3a5797 462{
bab475c5 463 __submit_merged_write_cond(sbi, inode, page, ino, type, false);
0c3a5797
CY		 464}
465
b9109b0e 466void f2fs_flush_merged_writes(struct f2fs_sb_info *sbi)
406657dd 467{
b9109b0e
JK		 468 f2fs_submit_merged_write(sbi, DATA);
469 f2fs_submit_merged_write(sbi, NODE);
470 f2fs_submit_merged_write(sbi, META);
406657dd
CY		 471}
472
93dfe2ac
JK		 473/*
474 * Fill the locked page with data located in the block address.
771a9a71 475 * A caller needs to unlock the page on failure.
93dfe2ac 476 */
05ca3632 477int f2fs_submit_page_bio(struct f2fs_io_info *fio)
93dfe2ac 478{
93dfe2ac 479 struct bio *bio;
0b81d077
JK		 480 struct page *page = fio->encrypted_page ?
481 fio->encrypted_page : fio->page;
93dfe2ac 482
c9b60788 483 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
93770ab7
CY		 484 fio->is_por ? META_POR : (__is_meta_io(fio) ?
485 META_GENERIC : DATA_GENERIC_ENHANCE)))
10f966bb 486 return -EFSCORRUPTED;
c9b60788 487
2ace38e0 488 trace_f2fs_submit_page_bio(page, fio);
05ca3632 489 f2fs_trace_ios(fio, 0);
93dfe2ac
JK		 490
491 /* Allocate a new bio */
b757f6ed 492 bio = __bio_alloc(fio, 1);
93dfe2ac 493
09cbfeaf 494 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
93dfe2ac 495 bio_put(bio);
93dfe2ac
JK		 496 return -EFAULT;
497 }
78efac53
CY		 498
499 if (fio->io_wbc && !is_read_io(fio->op))
34e51a5e 500 wbc_account_cgroup_owner(fio->io_wbc, page, PAGE_SIZE);
78efac53 501
04d328de 502 bio_set_op_attrs(bio, fio->op, fio->op_flags);
93dfe2ac 503
5f9abab4
JK		 504 inc_page_count(fio->sbi, is_read_io(fio->op) ?
505 __read_io_type(page): WB_DATA_TYPE(fio->page));
4c58ed07
CY		 506
507 __submit_bio(fio->sbi, bio, fio->type);
93dfe2ac
JK		 508 return 0;
509}
510
8896cbdf
CY		 511static bool page_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
512 block_t last_blkaddr, block_t cur_blkaddr)
513{
514 if (last_blkaddr + 1 != cur_blkaddr)
515 return false;
516 return __same_bdev(sbi, cur_blkaddr, bio);
517}
518
519static bool io_type_is_mergeable(struct f2fs_bio_info *io,
520 struct f2fs_io_info *fio)
521{
522 if (io->fio.op != fio->op)
523 return false;
524 return io->fio.op_flags == fio->op_flags;
525}
526
527static bool io_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
528 struct f2fs_bio_info *io,
529 struct f2fs_io_info *fio,
530 block_t last_blkaddr,
531 block_t cur_blkaddr)
532{
c72db71e
CY		 533 if (F2FS_IO_ALIGNED(sbi) && (fio->type == DATA || fio->type == NODE)) {
534 unsigned int filled_blocks =
535 F2FS_BYTES_TO_BLK(bio->bi_iter.bi_size);
536 unsigned int io_size = F2FS_IO_SIZE(sbi);
537 unsigned int left_vecs = bio->bi_max_vecs - bio->bi_vcnt;
538
539 /* IOs in bio is aligned and left space of vectors is not enough */
540 if (!(filled_blocks % io_size) && left_vecs < io_size)
541 return false;
542 }
8896cbdf
CY		 543 if (!page_is_mergeable(sbi, bio, last_blkaddr, cur_blkaddr))
544 return false;
545 return io_type_is_mergeable(io, fio);
546}
547
0b20fcec
CY		 548static void add_bio_entry(struct f2fs_sb_info *sbi, struct bio *bio,
549 struct page *page, enum temp_type temp)
550{
551 struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
552 struct bio_entry *be;
553
554 be = f2fs_kmem_cache_alloc(bio_entry_slab, GFP_NOFS);
555 be->bio = bio;
556 bio_get(bio);
557
558 if (bio_add_page(bio, page, PAGE_SIZE, 0) != PAGE_SIZE)
559 f2fs_bug_on(sbi, 1);
560
561 down_write(&io->bio_list_lock);
562 list_add_tail(&be->list, &io->bio_list);
563 up_write(&io->bio_list_lock);
564}
565
566static void del_bio_entry(struct bio_entry *be)
567{
568 list_del(&be->list);
569 kmem_cache_free(bio_entry_slab, be);
570}
571
572static int add_ipu_page(struct f2fs_sb_info *sbi, struct bio **bio,
573 struct page *page)
574{
575 enum temp_type temp;
576 bool found = false;
577 int ret = -EAGAIN;
578
579 for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
580 struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
581 struct list_head *head = &io->bio_list;
582 struct bio_entry *be;
583
584 down_write(&io->bio_list_lock);
585 list_for_each_entry(be, head, list) {
586 if (be->bio != *bio)
587 continue;
588
589 found = true;
590
591 if (bio_add_page(*bio, page, PAGE_SIZE, 0) == PAGE_SIZE) {
592 ret = 0;
593 break;
594 }
595
596 /* bio is full */
597 del_bio_entry(be);
598 __submit_bio(sbi, *bio, DATA);
599 break;
600 }
601 up_write(&io->bio_list_lock);
602 }
603
604 if (ret) {
605 bio_put(*bio);
606 *bio = NULL;
607 }
608
609 return ret;
610}
611
612void f2fs_submit_merged_ipu_write(struct f2fs_sb_info *sbi,
613 struct bio **bio, struct page *page)
614{
615 enum temp_type temp;
616 bool found = false;
617 struct bio *target = bio ? *bio : NULL;
618
619 for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
620 struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
621 struct list_head *head = &io->bio_list;
622 struct bio_entry *be;
623
624 if (list_empty(head))
625 continue;
626
627 down_read(&io->bio_list_lock);
628 list_for_each_entry(be, head, list) {
629 if (target)
630 found = (target == be->bio);
631 else
632 found = __has_merged_page(be->bio, NULL,
633 page, 0);
634 if (found)
635 break;
636 }
637 up_read(&io->bio_list_lock);
638
639 if (!found)
640 continue;
641
642 found = false;
643
644 down_write(&io->bio_list_lock);
645 list_for_each_entry(be, head, list) {
646 if (target)
647 found = (target == be->bio);
648 else
649 found = __has_merged_page(be->bio, NULL,
650 page, 0);
651 if (found) {
652 target = be->bio;
653 del_bio_entry(be);
654 break;
655 }
656 }
657 up_write(&io->bio_list_lock);
658 }
659
660 if (found)
661 __submit_bio(sbi, target, DATA);
662 if (bio && *bio) {
663 bio_put(*bio);
664 *bio = NULL;
665 }
666}
667
8648de2c
CY		 668int f2fs_merge_page_bio(struct f2fs_io_info *fio)
669{
670 struct bio *bio = *fio->bio;
671 struct page *page = fio->encrypted_page ?
672 fio->encrypted_page : fio->page;
673
674 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
675 __is_meta_io(fio) ? META_GENERIC : DATA_GENERIC))
10f966bb 676 return -EFSCORRUPTED;
8648de2c
CY		 677
678 trace_f2fs_submit_page_bio(page, fio);
679 f2fs_trace_ios(fio, 0);
680
8896cbdf 681 if (bio && !page_is_mergeable(fio->sbi, bio, *fio->last_block,
0b20fcec
CY		 682 fio->new_blkaddr))
683 f2fs_submit_merged_ipu_write(fio->sbi, &bio, NULL);
8648de2c
CY		 684alloc_new:
685 if (!bio) {
b757f6ed 686 bio = __bio_alloc(fio, BIO_MAX_PAGES);
8648de2c 687 bio_set_op_attrs(bio, fio->op, fio->op_flags);
8648de2c 688
0b20fcec
CY		 689 add_bio_entry(fio->sbi, bio, page, fio->temp);
690 } else {
691 if (add_ipu_page(fio->sbi, &bio, page))
692 goto alloc_new;
8648de2c
CY		 693 }
694
695 if (fio->io_wbc)
9637d517 696 wbc_account_cgroup_owner(fio->io_wbc, page, PAGE_SIZE);
8648de2c
CY		 697
698 inc_page_count(fio->sbi, WB_DATA_TYPE(page));
699
700 *fio->last_block = fio->new_blkaddr;
701 *fio->bio = bio;
702
703 return 0;
704}
705
fe16efe6 706void f2fs_submit_page_write(struct f2fs_io_info *fio)
93dfe2ac 707{
05ca3632 708 struct f2fs_sb_info *sbi = fio->sbi;
458e6197 709 enum page_type btype = PAGE_TYPE_OF_BIO(fio->type);
a912b54d 710 struct f2fs_bio_info *io = sbi->write_io[btype] + fio->temp;
4375a336 711 struct page *bio_page;
93dfe2ac 712
b9109b0e 713 f2fs_bug_on(sbi, is_read_io(fio->op));
93dfe2ac 714
fb830fc5
CY		 715 down_write(&io->io_rwsem);
716next:
717 if (fio->in_list) {
718 spin_lock(&io->io_lock);
719 if (list_empty(&io->io_list)) {
720 spin_unlock(&io->io_lock);
fe16efe6 721 goto out;
fb830fc5
CY		 722 }
723 fio = list_first_entry(&io->io_list,
724 struct f2fs_io_info, list);
725 list_del(&fio->list);
726 spin_unlock(&io->io_lock);
727 }
93dfe2ac 728
93770ab7 729 verify_fio_blkaddr(fio);
93dfe2ac 730
36951b38
CY		 731 bio_page = fio->encrypted_page ? fio->encrypted_page : fio->page;
732
ebf7c522
TM		 733 /* set submitted = true as a return value */
734 fio->submitted = true;
d68f735b 735
b9109b0e 736 inc_page_count(sbi, WB_DATA_TYPE(bio_page));
93dfe2ac 737
8896cbdf
CY		 738 if (io->bio && !io_is_mergeable(sbi, io->bio, io, fio,
739 io->last_block_in_bio, fio->new_blkaddr))
458e6197 740 __submit_merged_bio(io);
93dfe2ac
JK		 741alloc_new:
742 if (io->bio == NULL) {
8223ecc4
CY		 743 if (F2FS_IO_ALIGNED(sbi) &&
744 (fio->type == DATA || fio->type == NODE) &&
0a595eba 745 fio->new_blkaddr & F2FS_IO_SIZE_MASK(sbi)) {
b9109b0e 746 dec_page_count(sbi, WB_DATA_TYPE(bio_page));
fe16efe6
CY		 747 fio->retry = true;
748 goto skip;
0a595eba 749 }
b757f6ed 750 io->bio = __bio_alloc(fio, BIO_MAX_PAGES);
458e6197 751 io->fio = *fio;
93dfe2ac
JK		 752 }
753
a912b54d 754 if (bio_add_page(io->bio, bio_page, PAGE_SIZE, 0) < PAGE_SIZE) {
458e6197 755 __submit_merged_bio(io);
93dfe2ac
JK		 756 goto alloc_new;
757 }
758
578c6478 759 if (fio->io_wbc)
34e51a5e 760 wbc_account_cgroup_owner(fio->io_wbc, bio_page, PAGE_SIZE);
578c6478 761
7a9d7548 762 io->last_block_in_bio = fio->new_blkaddr;
05ca3632 763 f2fs_trace_ios(fio, 0);
fb830fc5
CY		 764
765 trace_f2fs_submit_page_write(fio->page, fio);
fe16efe6 766skip:
fb830fc5
CY		 767 if (fio->in_list)
768 goto next;
fe16efe6 769out:
4354994f 770 if (is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN) ||
00e09c0b 771 !f2fs_is_checkpoint_ready(sbi))
5ce80586 772 __submit_merged_bio(io);
df0f8dc0 773 up_write(&io->io_rwsem);
93dfe2ac
JK		 774}
775
95ae251f
EB		 776static inline bool f2fs_need_verity(const struct inode *inode, pgoff_t idx)
777{
778 return fsverity_active(inode) &&
779 idx < DIV_ROUND_UP(inode->i_size, PAGE_SIZE);
780}
781
13ba41e3 782static struct bio *f2fs_grab_read_bio(struct inode *inode, block_t blkaddr,
95ae251f
EB		 783 unsigned nr_pages, unsigned op_flag,
784 pgoff_t first_idx)
13ba41e3
JK		 785{
786 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
13ba41e3 787 struct bio *bio;
6dbb1796
EB		 788 struct bio_post_read_ctx *ctx;
789 unsigned int post_read_steps = 0;
13ba41e3 790
d62fe971 791 bio = f2fs_bio_alloc(sbi, min_t(int, nr_pages, BIO_MAX_PAGES), false);
6dbb1796 792 if (!bio)
13ba41e3 793 return ERR_PTR(-ENOMEM);
13ba41e3
JK		 794 f2fs_target_device(sbi, blkaddr, bio);
795 bio->bi_end_io = f2fs_read_end_io;
e2e59414 796 bio_set_op_attrs(bio, REQ_OP_READ, op_flag);
13ba41e3 797
6dbb1796
EB		 798 if (f2fs_encrypted_file(inode))
799 post_read_steps |= 1 << STEP_DECRYPT;
95ae251f
EB		 800
801 if (f2fs_need_verity(inode, first_idx))
802 post_read_steps |= 1 << STEP_VERITY;
803
6dbb1796
EB		 804 if (post_read_steps) {
805 ctx = mempool_alloc(bio_post_read_ctx_pool, GFP_NOFS);
806 if (!ctx) {
807 bio_put(bio);
808 return ERR_PTR(-ENOMEM);
809 }
810 ctx->bio = bio;
811 ctx->enabled_steps = post_read_steps;
812 bio->bi_private = ctx;
6dbb1796
EB		 813 }
814
13ba41e3
JK		 815 return bio;
816}
817
818/* This can handle encryption stuffs */
819static int f2fs_submit_page_read(struct inode *inode, struct page *page,
820 block_t blkaddr)
821{
93770ab7
CY		 822 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
823 struct bio *bio;
13ba41e3 824
95ae251f 825 bio = f2fs_grab_read_bio(inode, blkaddr, 1, 0, page->index);
13ba41e3
JK		 826 if (IS_ERR(bio))
827 return PTR_ERR(bio);
828
0ded69f6
JK		 829 /* wait for GCed page writeback via META_MAPPING */
830 f2fs_wait_on_block_writeback(inode, blkaddr);
831
13ba41e3
JK		 832 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
833 bio_put(bio);
834 return -EFAULT;
835 }
fb7d70db 836 ClearPageError(page);
93770ab7
CY		 837 inc_page_count(sbi, F2FS_RD_DATA);
838 __submit_bio(sbi, bio, DATA);
13ba41e3
JK		 839 return 0;
840}
841
46008c6d
CY		 842static void __set_data_blkaddr(struct dnode_of_data *dn)
843{
844 struct f2fs_node *rn = F2FS_NODE(dn->node_page);
845 __le32 *addr_array;
7a2af766
CY		 846 int base = 0;
847
848 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
849 base = get_extra_isize(dn->inode);
46008c6d
CY		 850
851 /* Get physical address of data block */
852 addr_array = blkaddr_in_node(rn);
7a2af766 853 addr_array[base + dn->ofs_in_node] = cpu_to_le32(dn->data_blkaddr);
46008c6d
CY		 854}
855
0a8165d7 856/*
eb47b800
JK		 857 * Lock ordering for the change of data block address:
858 * ->data_page
859 * ->node_page
860 * update block addresses in the node page
861 */
4d57b86d 862void f2fs_set_data_blkaddr(struct dnode_of_data *dn)
eb47b800 863{
bae0ee7a 864 f2fs_wait_on_page_writeback(dn->node_page, NODE, true, true);
46008c6d
CY		 865 __set_data_blkaddr(dn);
866 if (set_page_dirty(dn->node_page))
12719ae1 867 dn->node_changed = true;
eb47b800
JK		 868}
869
f28b3434
CY		 870void f2fs_update_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr)
871{
872 dn->data_blkaddr = blkaddr;
4d57b86d 873 f2fs_set_data_blkaddr(dn);
f28b3434
CY		 874 f2fs_update_extent_cache(dn);
875}
876
46008c6d 877/* dn->ofs_in_node will be returned with up-to-date last block pointer */
4d57b86d 878int f2fs_reserve_new_blocks(struct dnode_of_data *dn, blkcnt_t count)
eb47b800 879{
4081363f 880 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
0abd675e 881 int err;
eb47b800 882
46008c6d
CY		 883 if (!count)
884 return 0;
885
91942321 886 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
eb47b800 887 return -EPERM;
0abd675e
CY		 888 if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
889 return err;
eb47b800 890
46008c6d
CY		 891 trace_f2fs_reserve_new_blocks(dn->inode, dn->nid,
892 dn->ofs_in_node, count);
893
bae0ee7a 894 f2fs_wait_on_page_writeback(dn->node_page, NODE, true, true);
46008c6d
CY		 895
896 for (; count > 0; dn->ofs_in_node++) {
7a2af766
CY		 897 block_t blkaddr = datablock_addr(dn->inode,
898 dn->node_page, dn->ofs_in_node);
46008c6d
CY		 899 if (blkaddr == NULL_ADDR) {
900 dn->data_blkaddr = NEW_ADDR;
901 __set_data_blkaddr(dn);
902 count--;
903 }
904 }
905
906 if (set_page_dirty(dn->node_page))
907 dn->node_changed = true;
eb47b800
JK		 908 return 0;
909}
910
46008c6d 911/* Should keep dn->ofs_in_node unchanged */
4d57b86d 912int f2fs_reserve_new_block(struct dnode_of_data *dn)
46008c6d
CY		 913{
914 unsigned int ofs_in_node = dn->ofs_in_node;
915 int ret;
916
4d57b86d 917 ret = f2fs_reserve_new_blocks(dn, 1);
46008c6d
CY		 918 dn->ofs_in_node = ofs_in_node;
919 return ret;
920}
921
b600965c
HL		 922int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
923{
924 bool need_put = dn->inode_page ? false : true;
925 int err;
926
4d57b86d 927 err = f2fs_get_dnode_of_data(dn, index, ALLOC_NODE);
b600965c
HL		 928 if (err)
929 return err;
a8865372 930
b600965c 931 if (dn->data_blkaddr == NULL_ADDR)
4d57b86d 932 err = f2fs_reserve_new_block(dn);
a8865372 933 if (err || need_put)
b600965c
HL		 934 f2fs_put_dnode(dn);
935 return err;
936}
937
759af1c9 938int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index)
eb47b800 939{
e15882b6 940 struct extent_info ei = {0,0,0};
759af1c9 941 struct inode *inode = dn->inode;
028a41e8 942
759af1c9
FL		 943 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
944 dn->data_blkaddr = ei.blk + index - ei.fofs;
945 return 0;
429511cd 946 }
028a41e8 947
759af1c9 948 return f2fs_reserve_block(dn, index);
eb47b800
JK		 949}
950
4d57b86d 951struct page *f2fs_get_read_data_page(struct inode *inode, pgoff_t index,
04d328de 952 int op_flags, bool for_write)
eb47b800 953{
eb47b800
JK		 954 struct address_space *mapping = inode->i_mapping;
955 struct dnode_of_data dn;
956 struct page *page;
e15882b6 957 struct extent_info ei = {0,0,0};
eb47b800 958 int err;
4375a336 959
a56c7c6f 960 page = f2fs_grab_cache_page(mapping, index, for_write);
650495de
JK		 961 if (!page)
962 return ERR_PTR(-ENOMEM);
963
cb3bc9ee
CY		 964 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
965 dn.data_blkaddr = ei.blk + index - ei.fofs;
93770ab7
CY		 966 if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), dn.data_blkaddr,
967 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 968 err = -EFSCORRUPTED;
93770ab7
CY		 969 goto put_err;
970 }
cb3bc9ee
CY		 971 goto got_it;
972 }
973
eb47b800 974 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 975 err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
86531d6b
JK		 976 if (err)
977 goto put_err;
eb47b800
JK		 978 f2fs_put_dnode(&dn);
979
6bacf52f 980 if (unlikely(dn.data_blkaddr == NULL_ADDR)) {
86531d6b
JK		 981 err = -ENOENT;
982 goto put_err;
650495de 983 }
93770ab7
CY		 984 if (dn.data_blkaddr != NEW_ADDR &&
985 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
986 dn.data_blkaddr,
987 DATA_GENERIC_ENHANCE)) {
10f966bb 988 err = -EFSCORRUPTED;
93770ab7
CY		 989 goto put_err;
990 }
cb3bc9ee 991got_it:
43f3eae1
JK		 992 if (PageUptodate(page)) {
993 unlock_page(page);
eb47b800 994 return page;
43f3eae1 995 }
eb47b800 996
d59ff4df
JK		 997 /*
998 * A new dentry page is allocated but not able to be written, since its
999 * new inode page couldn't be allocated due to -ENOSPC.
1000 * In such the case, its blkaddr can be remained as NEW_ADDR.
4d57b86d
CY		 1001 * see, f2fs_add_link -> f2fs_get_new_data_page ->
1002 * f2fs_init_inode_metadata.
d59ff4df
JK		 1003 */
1004 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 1005 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 1006 if (!PageUptodate(page))
1007 SetPageUptodate(page);
43f3eae1 1008 unlock_page(page);
d59ff4df
JK		 1009 return page;
1010 }
eb47b800 1011
13ba41e3 1012 err = f2fs_submit_page_read(inode, page, dn.data_blkaddr);
393ff91f 1013 if (err)
86531d6b 1014 goto put_err;
43f3eae1 1015 return page;
86531d6b
JK		 1016
1017put_err:
1018 f2fs_put_page(page, 1);
1019 return ERR_PTR(err);
43f3eae1
JK		 1020}
1021
4d57b86d 1022struct page *f2fs_find_data_page(struct inode *inode, pgoff_t index)
43f3eae1
JK		 1023{
1024 struct address_space *mapping = inode->i_mapping;
1025 struct page *page;
1026
1027 page = find_get_page(mapping, index);
1028 if (page && PageUptodate(page))
1029 return page;
1030 f2fs_put_page(page, 0);
1031
4d57b86d 1032 page = f2fs_get_read_data_page(inode, index, 0, false);
43f3eae1
JK		 1033 if (IS_ERR(page))
1034 return page;
1035
1036 if (PageUptodate(page))
1037 return page;
1038
1039 wait_on_page_locked(page);
1040 if (unlikely(!PageUptodate(page))) {
1041 f2fs_put_page(page, 0);
1042 return ERR_PTR(-EIO);
1043 }
1044 return page;
1045}
1046
1047/*
1048 * If it tries to access a hole, return an error.
1049 * Because, the callers, functions in dir.c and GC, should be able to know
1050 * whether this page exists or not.
1051 */
4d57b86d 1052struct page *f2fs_get_lock_data_page(struct inode *inode, pgoff_t index,
a56c7c6f 1053 bool for_write)
43f3eae1
JK		 1054{
1055 struct address_space *mapping = inode->i_mapping;
1056 struct page *page;
1057repeat:
4d57b86d 1058 page = f2fs_get_read_data_page(inode, index, 0, for_write);
43f3eae1
JK		 1059 if (IS_ERR(page))
1060 return page;
393ff91f 1061
43f3eae1 1062 /* wait for read completion */
393ff91f 1063 lock_page(page);
6bacf52f 1064 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 1065 f2fs_put_page(page, 1);
1066 goto repeat;
eb47b800 1067 }
1563ac75
CY		 1068 if (unlikely(!PageUptodate(page))) {
1069 f2fs_put_page(page, 1);
1070 return ERR_PTR(-EIO);
1071 }
eb47b800
JK		 1072 return page;
1073}
1074
0a8165d7 1075/*
eb47b800
JK		 1076 * Caller ensures that this data page is never allocated.
1077 * A new zero-filled data page is allocated in the page cache.
39936837 1078 *
4f4124d0
CY		 1079 * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and
1080 * f2fs_unlock_op().
470f00e9
CY		 1081 * Note that, ipage is set only by make_empty_dir, and if any error occur,
1082 * ipage should be released by this function.
eb47b800 1083 */
4d57b86d 1084struct page *f2fs_get_new_data_page(struct inode *inode,
a8865372 1085 struct page *ipage, pgoff_t index, bool new_i_size)
eb47b800 1086{
eb47b800
JK		 1087 struct address_space *mapping = inode->i_mapping;
1088 struct page *page;
1089 struct dnode_of_data dn;
1090 int err;
7612118a 1091
a56c7c6f 1092 page = f2fs_grab_cache_page(mapping, index, true);
470f00e9
CY		 1093 if (!page) {
1094 /*
1095 * before exiting, we should make sure ipage will be released
1096 * if any error occur.
1097 */
1098 f2fs_put_page(ipage, 1);
01f28610 1099 return ERR_PTR(-ENOMEM);
470f00e9 1100 }
eb47b800 1101
a8865372 1102 set_new_dnode(&dn, inode, ipage, NULL, 0);
b600965c 1103 err = f2fs_reserve_block(&dn, index);
01f28610
JK		 1104 if (err) {
1105 f2fs_put_page(page, 1);
eb47b800 1106 return ERR_PTR(err);
a8865372 1107 }
01f28610
JK		 1108 if (!ipage)
1109 f2fs_put_dnode(&dn);
eb47b800
JK		 1110
1111 if (PageUptodate(page))
01f28610 1112 goto got_it;
eb47b800
JK		 1113
1114 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 1115 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 1116 if (!PageUptodate(page))
1117 SetPageUptodate(page);
eb47b800 1118 } else {
4375a336 1119 f2fs_put_page(page, 1);
a8865372 1120
7612118a
JK		 1121 /* if ipage exists, blkaddr should be NEW_ADDR */
1122 f2fs_bug_on(F2FS_I_SB(inode), ipage);
4d57b86d 1123 page = f2fs_get_lock_data_page(inode, index, true);
4375a336 1124 if (IS_ERR(page))
7612118a 1125 return page;
eb47b800 1126 }
01f28610 1127got_it:
9edcdabf 1128 if (new_i_size && i_size_read(inode) <
ee6d182f 1129 ((loff_t)(index + 1) << PAGE_SHIFT))
fc9581c8 1130 f2fs_i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT));
eb47b800
JK		 1131 return page;
1132}
1133
d5097be5 1134static int __allocate_data_block(struct dnode_of_data *dn, int seg_type)
bfad7c2d 1135{
4081363f 1136 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
bfad7c2d 1137 struct f2fs_summary sum;
bfad7c2d 1138 struct node_info ni;
6aa58d8a 1139 block_t old_blkaddr;
46008c6d 1140 blkcnt_t count = 1;
0abd675e 1141 int err;
bfad7c2d 1142
91942321 1143 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
bfad7c2d 1144 return -EPERM;
df6136ef 1145
7735730d
CY		 1146 err = f2fs_get_node_info(sbi, dn->nid, &ni);
1147 if (err)
1148 return err;
1149
7a2af766
CY		 1150 dn->data_blkaddr = datablock_addr(dn->inode,
1151 dn->node_page, dn->ofs_in_node);
f847c699 1152 if (dn->data_blkaddr != NULL_ADDR)
df6136ef
CY		 1153 goto alloc;
1154
0abd675e
CY		 1155 if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
1156 return err;
bfad7c2d 1157
df6136ef 1158alloc:
bfad7c2d 1159 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
6aa58d8a
CY		 1160 old_blkaddr = dn->data_blkaddr;
1161 f2fs_allocate_data_block(sbi, NULL, old_blkaddr, &dn->data_blkaddr,
d5097be5 1162 &sum, seg_type, NULL, false);
6aa58d8a
CY		 1163 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
1164 invalidate_mapping_pages(META_MAPPING(sbi),
1165 old_blkaddr, old_blkaddr);
86f35dc3 1166 f2fs_update_data_blkaddr(dn, dn->data_blkaddr);
bfad7c2d 1167
0a4daae5
JK		 1168 /*
1169 * i_size will be updated by direct_IO. Otherwise, we'll get stale
1170 * data from unwritten block via dio_read.
1171 */
bfad7c2d
JK		 1172 return 0;
1173}
1174
a7de6086 1175int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *from)
59b802e5 1176{
b439b103 1177 struct inode *inode = file_inode(iocb->ki_filp);
5b8db7fa 1178 struct f2fs_map_blocks map;
d6d478a1 1179 int flag;
a7de6086 1180 int err = 0;
d6d478a1 1181 bool direct_io = iocb->ki_flags & IOCB_DIRECT;
59b802e5 1182
0080c507 1183 map.m_lblk = F2FS_BLK_ALIGN(iocb->ki_pos);
dfd02e4d
CY		 1184 map.m_len = F2FS_BYTES_TO_BLK(iocb->ki_pos + iov_iter_count(from));
1185 if (map.m_len > map.m_lblk)
1186 map.m_len -= map.m_lblk;
1187 else
1188 map.m_len = 0;
1189
da85985c 1190 map.m_next_pgofs = NULL;
c4020b2d 1191 map.m_next_extent = NULL;
d5097be5 1192 map.m_seg_type = NO_CHECK_TYPE;
f9d6d059 1193 map.m_may_create = true;
2a340760 1194
d6d478a1 1195 if (direct_io) {
4d57b86d 1196 map.m_seg_type = f2fs_rw_hint_to_seg_type(iocb->ki_hint);
f847c699 1197 flag = f2fs_force_buffered_io(inode, iocb, from) ?
d6d478a1
CY		 1198 F2FS_GET_BLOCK_PRE_AIO :
1199 F2FS_GET_BLOCK_PRE_DIO;
1200 goto map_blocks;
d5097be5 1201 }
f2470371 1202 if (iocb->ki_pos + iov_iter_count(from) > MAX_INLINE_DATA(inode)) {
a7de6086
JK		 1203 err = f2fs_convert_inline_inode(inode);
1204 if (err)
1205 return err;
b439b103 1206 }
d6d478a1 1207 if (f2fs_has_inline_data(inode))
25006645 1208 return err;
d6d478a1
CY		 1209
1210 flag = F2FS_GET_BLOCK_PRE_AIO;
1211
1212map_blocks:
1213 err = f2fs_map_blocks(inode, &map, 1, flag);
1214 if (map.m_len > 0 && err == -ENOSPC) {
1215 if (!direct_io)
1216 set_inode_flag(inode, FI_NO_PREALLOC);
1217 err = 0;
25006645 1218 }
a7de6086 1219 return err;
59b802e5
JK		 1220}
1221
39a86958 1222void __do_map_lock(struct f2fs_sb_info *sbi, int flag, bool lock)
59c9081b
YH		 1223{
1224 if (flag == F2FS_GET_BLOCK_PRE_AIO) {
1225 if (lock)
1226 down_read(&sbi->node_change);
1227 else
1228 up_read(&sbi->node_change);
1229 } else {
1230 if (lock)
1231 f2fs_lock_op(sbi);
1232 else
1233 f2fs_unlock_op(sbi);
1234 }
1235}
1236
0a8165d7 1237/*
003a3e1d
JK		 1238 * f2fs_map_blocks() now supported readahead/bmap/rw direct_IO with
1239 * f2fs_map_blocks structure.
4f4124d0
CY		 1240 * If original data blocks are allocated, then give them to blockdev.
1241 * Otherwise,
1242 * a. preallocate requested block addresses
1243 * b. do not use extent cache for better performance
1244 * c. give the block addresses to blockdev
eb47b800 1245 */
d323d005 1246int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
e2b4e2bc 1247 int create, int flag)
eb47b800 1248{
003a3e1d 1249 unsigned int maxblocks = map->m_len;
eb47b800 1250 struct dnode_of_data dn;
f9811703 1251 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
f9d6d059 1252 int mode = map->m_may_create ? ALLOC_NODE : LOOKUP_NODE;
46008c6d 1253 pgoff_t pgofs, end_offset, end;
bfad7c2d 1254 int err = 0, ofs = 1;
46008c6d
CY		 1255 unsigned int ofs_in_node, last_ofs_in_node;
1256 blkcnt_t prealloc;
e15882b6 1257 struct extent_info ei = {0,0,0};
7df3a431 1258 block_t blkaddr;
c4020b2d 1259 unsigned int start_pgofs;
eb47b800 1260
dfd02e4d
CY		 1261 if (!maxblocks)
1262 return 0;
1263
003a3e1d
JK		 1264 map->m_len = 0;
1265 map->m_flags = 0;
1266
1267 /* it only supports block size == page size */
1268 pgofs = (pgoff_t)map->m_lblk;
46008c6d 1269 end = pgofs + maxblocks;
eb47b800 1270
24b84912 1271 if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) {
f4f0b677
JZ		 1272 if (test_opt(sbi, LFS) && flag == F2FS_GET_BLOCK_DIO &&
1273 map->m_may_create)
1274 goto next_dnode;
1275
003a3e1d
JK		 1276 map->m_pblk = ei.blk + pgofs - ei.fofs;
1277 map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs);
1278 map->m_flags = F2FS_MAP_MAPPED;
c4020b2d
CY		 1279 if (map->m_next_extent)
1280 *map->m_next_extent = pgofs + map->m_len;
1e78e8bd
ST		 1281
1282 /* for hardware encryption, but to avoid potential issue in future */
1283 if (flag == F2FS_GET_BLOCK_DIO)
1284 f2fs_wait_on_block_writeback_range(inode,
1285 map->m_pblk, map->m_len);
bfad7c2d 1286 goto out;
a2e7d1bf 1287 }
bfad7c2d 1288
4fe71e88 1289next_dnode:
f9d6d059 1290 if (map->m_may_create)
59c9081b 1291 __do_map_lock(sbi, flag, true);
eb47b800
JK		 1292
1293 /* When reading holes, we need its node page */
1294 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1295 err = f2fs_get_dnode_of_data(&dn, pgofs, mode);
1ec79083 1296 if (err) {
43473f96
CY		 1297 if (flag == F2FS_GET_BLOCK_BMAP)
1298 map->m_pblk = 0;
da85985c 1299 if (err == -ENOENT) {
bfad7c2d 1300 err = 0;
da85985c
CY		 1301 if (map->m_next_pgofs)
1302 *map->m_next_pgofs =
4d57b86d 1303 f2fs_get_next_page_offset(&dn, pgofs);
c4020b2d
CY		 1304 if (map->m_next_extent)
1305 *map->m_next_extent =
4d57b86d 1306 f2fs_get_next_page_offset(&dn, pgofs);
da85985c 1307 }
bfad7c2d 1308 goto unlock_out;
848753aa 1309 }
973163fc 1310
c4020b2d 1311 start_pgofs = pgofs;
46008c6d 1312 prealloc = 0;
230436b3 1313 last_ofs_in_node = ofs_in_node = dn.ofs_in_node;
81ca7350 1314 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
4fe71e88
CY		 1315
1316next_block:
7a2af766 1317 blkaddr = datablock_addr(dn.inode, dn.node_page, dn.ofs_in_node);
4fe71e88 1318
c9b60788 1319 if (__is_valid_data_blkaddr(blkaddr) &&
93770ab7 1320 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC_ENHANCE)) {
10f966bb 1321 err = -EFSCORRUPTED;
c9b60788
CY		 1322 goto sync_out;
1323 }
1324
93770ab7 1325 if (__is_valid_data_blkaddr(blkaddr)) {
f847c699 1326 /* use out-place-update for driect IO under LFS mode */
f9d6d059
CY		 1327 if (test_opt(sbi, LFS) && flag == F2FS_GET_BLOCK_DIO &&
1328 map->m_may_create) {
f847c699 1329 err = __allocate_data_block(&dn, map->m_seg_type);
05e36006
CY		 1330 if (err)
1331 goto sync_out;
1332 blkaddr = dn.data_blkaddr;
1333 set_inode_flag(inode, FI_APPEND_WRITE);
f847c699
CY		 1334 }
1335 } else {
973163fc 1336 if (create) {
f9811703
CY		 1337 if (unlikely(f2fs_cp_error(sbi))) {
1338 err = -EIO;
4fe71e88 1339 goto sync_out;
f9811703 1340 }
24b84912 1341 if (flag == F2FS_GET_BLOCK_PRE_AIO) {
46008c6d
CY		 1342 if (blkaddr == NULL_ADDR) {
1343 prealloc++;
1344 last_ofs_in_node = dn.ofs_in_node;
1345 }
24b84912 1346 } else {
0a4daae5
JK		 1347 WARN_ON(flag != F2FS_GET_BLOCK_PRE_DIO &&
1348 flag != F2FS_GET_BLOCK_DIO);
d5097be5
HL		 1349 err = __allocate_data_block(&dn,
1350 map->m_seg_type);
6f2d8ed6 1351 if (!err)
91942321 1352 set_inode_flag(inode, FI_APPEND_WRITE);
24b84912 1353 }
973163fc 1354 if (err)
4fe71e88 1355 goto sync_out;
3f2be043 1356 map->m_flags |= F2FS_MAP_NEW;
4fe71e88 1357 blkaddr = dn.data_blkaddr;
973163fc 1358 } else {
43473f96
CY		 1359 if (flag == F2FS_GET_BLOCK_BMAP) {
1360 map->m_pblk = 0;
1361 goto sync_out;
1362 }
c4020b2d
CY		 1363 if (flag == F2FS_GET_BLOCK_PRECACHE)
1364 goto sync_out;
da85985c
CY		 1365 if (flag == F2FS_GET_BLOCK_FIEMAP &&
1366 blkaddr == NULL_ADDR) {
1367 if (map->m_next_pgofs)
1368 *map->m_next_pgofs = pgofs + 1;
4c2ac6a8 1369 goto sync_out;
da85985c 1370 }
f3d98e74
CY		 1371 if (flag != F2FS_GET_BLOCK_FIEMAP) {
1372 /* for defragment case */
1373 if (map->m_next_pgofs)
1374 *map->m_next_pgofs = pgofs + 1;
4fe71e88 1375 goto sync_out;
f3d98e74 1376 }
e2b4e2bc 1377 }
e2b4e2bc 1378 }
eb47b800 1379
46008c6d
CY		 1380 if (flag == F2FS_GET_BLOCK_PRE_AIO)
1381 goto skip;
1382
4fe71e88
CY		 1383 if (map->m_len == 0) {
1384 /* preallocated unwritten block should be mapped for fiemap. */
1385 if (blkaddr == NEW_ADDR)
1386 map->m_flags |= F2FS_MAP_UNWRITTEN;
1387 map->m_flags |= F2FS_MAP_MAPPED;
1388
1389 map->m_pblk = blkaddr;
1390 map->m_len = 1;
1391 } else if ((map->m_pblk != NEW_ADDR &&
1392 blkaddr == (map->m_pblk + ofs)) ||
b439b103 1393 (map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR) ||
46008c6d 1394 flag == F2FS_GET_BLOCK_PRE_DIO) {
4fe71e88
CY		 1395 ofs++;
1396 map->m_len++;
1397 } else {
1398 goto sync_out;
1399 }
bfad7c2d 1400
46008c6d 1401skip:
bfad7c2d
JK		 1402 dn.ofs_in_node++;
1403 pgofs++;
1404
46008c6d
CY		 1405 /* preallocate blocks in batch for one dnode page */
1406 if (flag == F2FS_GET_BLOCK_PRE_AIO &&
1407 (pgofs == end || dn.ofs_in_node == end_offset)) {
7df3a431 1408
46008c6d 1409 dn.ofs_in_node = ofs_in_node;
4d57b86d 1410 err = f2fs_reserve_new_blocks(&dn, prealloc);
46008c6d
CY		 1411 if (err)
1412 goto sync_out;
bfad7c2d 1413
46008c6d
CY		 1414 map->m_len += dn.ofs_in_node - ofs_in_node;
1415 if (prealloc && dn.ofs_in_node != last_ofs_in_node + 1) {
1416 err = -ENOSPC;
1417 goto sync_out;
3104af35 1418 }
46008c6d
CY		 1419 dn.ofs_in_node = end_offset;
1420 }
1421
1422 if (pgofs >= end)
1423 goto sync_out;
1424 else if (dn.ofs_in_node < end_offset)
1425 goto next_block;
1426
c4020b2d
CY		 1427 if (flag == F2FS_GET_BLOCK_PRECACHE) {
1428 if (map->m_flags & F2FS_MAP_MAPPED) {
1429 unsigned int ofs = start_pgofs - map->m_lblk;
1430
1431 f2fs_update_extent_cache_range(&dn,
1432 start_pgofs, map->m_pblk + ofs,
1433 map->m_len - ofs);
1434 }
1435 }
1436
46008c6d
CY		 1437 f2fs_put_dnode(&dn);
1438
f9d6d059 1439 if (map->m_may_create) {
59c9081b 1440 __do_map_lock(sbi, flag, false);
6f2d8ed6 1441 f2fs_balance_fs(sbi, dn.node_changed);
eb47b800 1442 }
46008c6d 1443 goto next_dnode;
7df3a431 1444
bfad7c2d 1445sync_out:
1e78e8bd
ST		 1446
1447 /* for hardware encryption, but to avoid potential issue in future */
1448 if (flag == F2FS_GET_BLOCK_DIO && map->m_flags & F2FS_MAP_MAPPED)
1449 f2fs_wait_on_block_writeback_range(inode,
1450 map->m_pblk, map->m_len);
1451
c4020b2d
CY		 1452 if (flag == F2FS_GET_BLOCK_PRECACHE) {
1453 if (map->m_flags & F2FS_MAP_MAPPED) {
1454 unsigned int ofs = start_pgofs - map->m_lblk;
1455
1456 f2fs_update_extent_cache_range(&dn,
1457 start_pgofs, map->m_pblk + ofs,
1458 map->m_len - ofs);
1459 }
1460 if (map->m_next_extent)
1461 *map->m_next_extent = pgofs + 1;
1462 }
eb47b800 1463 f2fs_put_dnode(&dn);
bfad7c2d 1464unlock_out:
f9d6d059 1465 if (map->m_may_create) {
59c9081b 1466 __do_map_lock(sbi, flag, false);
6f2d8ed6 1467 f2fs_balance_fs(sbi, dn.node_changed);
2a340760 1468 }
bfad7c2d 1469out:
003a3e1d 1470 trace_f2fs_map_blocks(inode, map, err);
bfad7c2d 1471 return err;
eb47b800
JK		 1472}
1473
b91050a8
HL		 1474bool f2fs_overwrite_io(struct inode *inode, loff_t pos, size_t len)
1475{
1476 struct f2fs_map_blocks map;
1477 block_t last_lblk;
1478 int err;
1479
1480 if (pos + len > i_size_read(inode))
1481 return false;
1482
1483 map.m_lblk = F2FS_BYTES_TO_BLK(pos);
1484 map.m_next_pgofs = NULL;
1485 map.m_next_extent = NULL;
1486 map.m_seg_type = NO_CHECK_TYPE;
f4f0b677 1487 map.m_may_create = false;
b91050a8
HL		 1488 last_lblk = F2FS_BLK_ALIGN(pos + len);
1489
1490 while (map.m_lblk < last_lblk) {
1491 map.m_len = last_lblk - map.m_lblk;
1492 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
1493 if (err || map.m_len == 0)
1494 return false;
1495 map.m_lblk += map.m_len;
1496 }
1497 return true;
1498}
1499
003a3e1d 1500static int __get_data_block(struct inode *inode, sector_t iblock,
da85985c 1501 struct buffer_head *bh, int create, int flag,
f9d6d059 1502 pgoff_t *next_pgofs, int seg_type, bool may_write)
003a3e1d
JK		 1503{
1504 struct f2fs_map_blocks map;
a7de6086 1505 int err;
003a3e1d
JK		 1506
1507 map.m_lblk = iblock;
1508 map.m_len = bh->b_size >> inode->i_blkbits;
da85985c 1509 map.m_next_pgofs = next_pgofs;
c4020b2d 1510 map.m_next_extent = NULL;
d5097be5 1511 map.m_seg_type = seg_type;
f9d6d059 1512 map.m_may_create = may_write;
003a3e1d 1513
a7de6086
JK		 1514 err = f2fs_map_blocks(inode, &map, create, flag);
1515 if (!err) {
003a3e1d
JK		 1516 map_bh(bh, inode->i_sb, map.m_pblk);
1517 bh->b_state = (bh->b_state & ~F2FS_MAP_FLAGS) | map.m_flags;
b86e3307 1518 bh->b_size = (u64)map.m_len << inode->i_blkbits;
003a3e1d 1519 }
a7de6086 1520 return err;
003a3e1d
JK		 1521}
1522
ccfb3000 1523static int get_data_block(struct inode *inode, sector_t iblock,
da85985c
CY		 1524 struct buffer_head *bh_result, int create, int flag,
1525 pgoff_t *next_pgofs)
e2b4e2bc 1526{
da85985c 1527 return __get_data_block(inode, iblock, bh_result, create,
d5097be5 1528 flag, next_pgofs,
f9d6d059
CY		 1529 NO_CHECK_TYPE, create);
1530}
1531
1532static int get_data_block_dio_write(struct inode *inode, sector_t iblock,
1533 struct buffer_head *bh_result, int create)
1534{
1535 return __get_data_block(inode, iblock, bh_result, create,
1536 F2FS_GET_BLOCK_DIO, NULL,
1537 f2fs_rw_hint_to_seg_type(inode->i_write_hint),
75a037f3 1538 IS_SWAPFILE(inode) ? false : true);
e2b4e2bc
CY		 1539}
1540
1541static int get_data_block_dio(struct inode *inode, sector_t iblock,
ccfb3000
JK		 1542 struct buffer_head *bh_result, int create)
1543{
e2b4e2bc 1544 return __get_data_block(inode, iblock, bh_result, create,
f9d6d059
CY		 1545 F2FS_GET_BLOCK_DIO, NULL,
1546 f2fs_rw_hint_to_seg_type(inode->i_write_hint),
1547 false);
ccfb3000
JK		 1548}
1549
e2b4e2bc 1550static int get_data_block_bmap(struct inode *inode, sector_t iblock,
ccfb3000
JK		 1551 struct buffer_head *bh_result, int create)
1552{
179448bf 1553 /* Block number less than F2FS MAX BLOCKS */
e0afc4d6 1554 if (unlikely(iblock >= F2FS_I_SB(inode)->max_file_blocks))
179448bf
YH		 1555 return -EFBIG;
1556
e2b4e2bc 1557 return __get_data_block(inode, iblock, bh_result, create,
d5097be5 1558 F2FS_GET_BLOCK_BMAP, NULL,
f9d6d059 1559 NO_CHECK_TYPE, create);
ccfb3000
JK		 1560}
1561
7f63eb77
JK		 1562static inline sector_t logical_to_blk(struct inode *inode, loff_t offset)
1563{
1564 return (offset >> inode->i_blkbits);
1565}
1566
1567static inline loff_t blk_to_logical(struct inode *inode, sector_t blk)
1568{
1569 return (blk << inode->i_blkbits);
1570}
1571
442a9dbd
CY		 1572static int f2fs_xattr_fiemap(struct inode *inode,
1573 struct fiemap_extent_info *fieinfo)
1574{
1575 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1576 struct page *page;
1577 struct node_info ni;
1578 __u64 phys = 0, len;
1579 __u32 flags;
1580 nid_t xnid = F2FS_I(inode)->i_xattr_nid;
1581 int err = 0;
1582
1583 if (f2fs_has_inline_xattr(inode)) {
1584 int offset;
1585
1586 page = f2fs_grab_cache_page(NODE_MAPPING(sbi),
1587 inode->i_ino, false);
1588 if (!page)
1589 return -ENOMEM;
1590
7735730d
CY		 1591 err = f2fs_get_node_info(sbi, inode->i_ino, &ni);
1592 if (err) {
1593 f2fs_put_page(page, 1);
1594 return err;
1595 }
442a9dbd
CY		 1596
1597 phys = (__u64)blk_to_logical(inode, ni.blk_addr);
1598 offset = offsetof(struct f2fs_inode, i_addr) +
1599 sizeof(__le32) * (DEF_ADDRS_PER_INODE -
b323fd28 1600 get_inline_xattr_addrs(inode));
442a9dbd
CY		 1601
1602 phys += offset;
1603 len = inline_xattr_size(inode);
1604
1605 f2fs_put_page(page, 1);
1606
1607 flags = FIEMAP_EXTENT_DATA_INLINE | FIEMAP_EXTENT_NOT_ALIGNED;
1608
1609 if (!xnid)
1610 flags |= FIEMAP_EXTENT_LAST;
1611
1612 err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
1613 if (err || err == 1)
1614 return err;
1615 }
1616
1617 if (xnid) {
1618 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), xnid, false);
1619 if (!page)
1620 return -ENOMEM;
1621
7735730d
CY		 1622 err = f2fs_get_node_info(sbi, xnid, &ni);
1623 if (err) {
1624 f2fs_put_page(page, 1);
1625 return err;
1626 }
442a9dbd
CY		 1627
1628 phys = (__u64)blk_to_logical(inode, ni.blk_addr);
1629 len = inode->i_sb->s_blocksize;
1630
1631 f2fs_put_page(page, 1);
1632
1633 flags = FIEMAP_EXTENT_LAST;
1634 }
1635
1636 if (phys)
1637 err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
1638
1639 return (err < 0 ? err : 0);
1640}
1641
9ab70134
JK		 1642int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1643 u64 start, u64 len)
1644{
7f63eb77
JK		 1645 struct buffer_head map_bh;
1646 sector_t start_blk, last_blk;
da85985c 1647 pgoff_t next_pgofs;
7f63eb77
JK		 1648 u64 logical = 0, phys = 0, size = 0;
1649 u32 flags = 0;
7f63eb77
JK		 1650 int ret = 0;
1651
c4020b2d
CY		 1652 if (fieinfo->fi_flags & FIEMAP_FLAG_CACHE) {
1653 ret = f2fs_precache_extents(inode);
1654 if (ret)
1655 return ret;
1656 }
1657
442a9dbd 1658 ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC | FIEMAP_FLAG_XATTR);
7f63eb77
JK		 1659 if (ret)
1660 return ret;
1661
f1b43d4c
CY		 1662 inode_lock(inode);
1663
442a9dbd
CY		 1664 if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
1665 ret = f2fs_xattr_fiemap(inode, fieinfo);
1666 goto out;
1667 }
1668
7975f349 1669 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
67f8cf3c
JK		 1670 ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len);
1671 if (ret != -EAGAIN)
f1b43d4c 1672 goto out;
67f8cf3c
JK		 1673 }
1674
7f63eb77
JK		 1675 if (logical_to_blk(inode, len) == 0)
1676 len = blk_to_logical(inode, 1);
1677
1678 start_blk = logical_to_blk(inode, start);
1679 last_blk = logical_to_blk(inode, start + len - 1);
9a950d52 1680
7f63eb77
JK		 1681next:
1682 memset(&map_bh, 0, sizeof(struct buffer_head));
1683 map_bh.b_size = len;
1684
e2b4e2bc 1685 ret = get_data_block(inode, start_blk, &map_bh, 0,
da85985c 1686 F2FS_GET_BLOCK_FIEMAP, &next_pgofs);
7f63eb77
JK		 1687 if (ret)
1688 goto out;
1689
1690 /* HOLE */
1691 if (!buffer_mapped(&map_bh)) {
da85985c 1692 start_blk = next_pgofs;
58736fa6
CY		 1693
1694 if (blk_to_logical(inode, start_blk) < blk_to_logical(inode,
1695 F2FS_I_SB(inode)->max_file_blocks))
9a950d52 1696 goto prep_next;
58736fa6 1697
9a950d52
FL		 1698 flags |= FIEMAP_EXTENT_LAST;
1699 }
7f63eb77 1700
da5af127 1701 if (size) {
62230e0d 1702 if (IS_ENCRYPTED(inode))
da5af127
CY		 1703 flags |= FIEMAP_EXTENT_DATA_ENCRYPTED;
1704
9a950d52
FL		 1705 ret = fiemap_fill_next_extent(fieinfo, logical,
1706 phys, size, flags);
da5af127 1707 }
7f63eb77 1708
9a950d52
FL		 1709 if (start_blk > last_blk || ret)
1710 goto out;
7f63eb77 1711
9a950d52
FL		 1712 logical = blk_to_logical(inode, start_blk);
1713 phys = blk_to_logical(inode, map_bh.b_blocknr);
1714 size = map_bh.b_size;
1715 flags = 0;
1716 if (buffer_unwritten(&map_bh))
1717 flags = FIEMAP_EXTENT_UNWRITTEN;
7f63eb77 1718
9a950d52 1719 start_blk += logical_to_blk(inode, size);
7f63eb77 1720
9a950d52 1721prep_next:
7f63eb77
JK		 1722 cond_resched();
1723 if (fatal_signal_pending(current))
1724 ret = -EINTR;
1725 else
1726 goto next;
1727out:
1728 if (ret == 1)
1729 ret = 0;
1730
5955102c 1731 inode_unlock(inode);
7f63eb77 1732 return ret;
9ab70134
JK		 1733}
1734
95ae251f
EB		 1735static inline loff_t f2fs_readpage_limit(struct inode *inode)
1736{
1737 if (IS_ENABLED(CONFIG_FS_VERITY) &&
1738 (IS_VERITY(inode) || f2fs_verity_in_progress(inode)))
1739 return inode->i_sb->s_maxbytes;
1740
1741 return i_size_read(inode);
1742}
1743
2df0ab04
CY		 1744static int f2fs_read_single_page(struct inode *inode, struct page *page,
1745 unsigned nr_pages,
1746 struct f2fs_map_blocks *map,
1747 struct bio **bio_ret,
1748 sector_t *last_block_in_bio,
1749 bool is_readahead)
1750{
1751 struct bio *bio = *bio_ret;
1752 const unsigned blkbits = inode->i_blkbits;
1753 const unsigned blocksize = 1 << blkbits;
1754 sector_t block_in_file;
1755 sector_t last_block;
1756 sector_t last_block_in_file;
1757 sector_t block_nr;
1758 int ret = 0;
1759
4969c06a 1760 block_in_file = (sector_t)page_index(page);
2df0ab04 1761 last_block = block_in_file + nr_pages;
95ae251f 1762 last_block_in_file = (f2fs_readpage_limit(inode) + blocksize - 1) >>
2df0ab04
CY		 1763 blkbits;
1764 if (last_block > last_block_in_file)
1765 last_block = last_block_in_file;
1766
1767 /* just zeroing out page which is beyond EOF */
1768 if (block_in_file >= last_block)
1769 goto zero_out;
1770 /*
1771 * Map blocks using the previous result first.
1772 */
1773 if ((map->m_flags & F2FS_MAP_MAPPED) &&
1774 block_in_file > map->m_lblk &&
1775 block_in_file < (map->m_lblk + map->m_len))
1776 goto got_it;
1777
1778 /*
1779 * Then do more f2fs_map_blocks() calls until we are
1780 * done with this page.
1781 */
1782 map->m_lblk = block_in_file;
1783 map->m_len = last_block - block_in_file;
1784
1785 ret = f2fs_map_blocks(inode, map, 0, F2FS_GET_BLOCK_DEFAULT);
1786 if (ret)
1787 goto out;
1788got_it:
1789 if ((map->m_flags & F2FS_MAP_MAPPED)) {
1790 block_nr = map->m_pblk + block_in_file - map->m_lblk;
1791 SetPageMappedToDisk(page);
1792
4969c06a
JK		 1793 if (!PageUptodate(page) && (!PageSwapCache(page) &&
1794 !cleancache_get_page(page))) {
2df0ab04
CY		 1795 SetPageUptodate(page);
1796 goto confused;
1797 }
1798
1799 if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), block_nr,
93770ab7 1800 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 1801 ret = -EFSCORRUPTED;
2df0ab04
CY		 1802 goto out;
1803 }
1804 } else {
1805zero_out:
1806 zero_user_segment(page, 0, PAGE_SIZE);
95ae251f
EB		 1807 if (f2fs_need_verity(inode, page->index) &&
1808 !fsverity_verify_page(page)) {
1809 ret = -EIO;
1810 goto out;
1811 }
2df0ab04
CY		 1812 if (!PageUptodate(page))
1813 SetPageUptodate(page);
1814 unlock_page(page);
1815 goto out;
1816 }
1817
1818 /*
1819 * This page will go to BIO. Do we need to send this
1820 * BIO off first?
1821 */
8896cbdf
CY		 1822 if (bio && !page_is_mergeable(F2FS_I_SB(inode), bio,
1823 *last_block_in_bio, block_nr)) {
2df0ab04
CY		 1824submit_and_realloc:
1825 __submit_bio(F2FS_I_SB(inode), bio, DATA);
1826 bio = NULL;
1827 }
1828 if (bio == NULL) {
1829 bio = f2fs_grab_read_bio(inode, block_nr, nr_pages,
95ae251f 1830 is_readahead ? REQ_RAHEAD : 0, page->index);
2df0ab04
CY		 1831 if (IS_ERR(bio)) {
1832 ret = PTR_ERR(bio);
1833 bio = NULL;
1834 goto out;
1835 }
1836 }
1837
1838 /*
1839 * If the page is under writeback, we need to wait for
1840 * its completion to see the correct decrypted data.
1841 */
1842 f2fs_wait_on_block_writeback(inode, block_nr);
1843
1844 if (bio_add_page(bio, page, blocksize, 0) < blocksize)
1845 goto submit_and_realloc;
1846
1847 inc_page_count(F2FS_I_SB(inode), F2FS_RD_DATA);
1848 ClearPageError(page);
1849 *last_block_in_bio = block_nr;
1850 goto out;
1851confused:
1852 if (bio) {
1853 __submit_bio(F2FS_I_SB(inode), bio, DATA);
1854 bio = NULL;
1855 }
1856 unlock_page(page);
1857out:
1858 *bio_ret = bio;
1859 return ret;
1860}
1861
f1e88660
JK		 1862/*
1863 * This function was originally taken from fs/mpage.c, and customized for f2fs.
1864 * Major change was from block_size == page_size in f2fs by default.
e2e59414
JK		 1865 *
1866 * Note that the aops->readpages() function is ONLY used for read-ahead. If
1867 * this function ever deviates from doing just read-ahead, it should either
1868 * use ->readpage() or do the necessary surgery to decouple ->readpages()
1869 * from read-ahead.
f1e88660
JK		 1870 */
1871static int f2fs_mpage_readpages(struct address_space *mapping,
1872 struct list_head *pages, struct page *page,
e2e59414 1873 unsigned nr_pages, bool is_readahead)
f1e88660
JK		 1874{
1875 struct bio *bio = NULL;
f1e88660
JK		 1876 sector_t last_block_in_bio = 0;
1877 struct inode *inode = mapping->host;
f1e88660 1878 struct f2fs_map_blocks map;
2df0ab04 1879 int ret = 0;
f1e88660
JK		 1880
1881 map.m_pblk = 0;
1882 map.m_lblk = 0;
1883 map.m_len = 0;
1884 map.m_flags = 0;
da85985c 1885 map.m_next_pgofs = NULL;
c4020b2d 1886 map.m_next_extent = NULL;
d5097be5 1887 map.m_seg_type = NO_CHECK_TYPE;
f9d6d059 1888 map.m_may_create = false;
f1e88660 1889
736c0a74 1890 for (; nr_pages; nr_pages--) {
f1e88660 1891 if (pages) {
939afa94 1892 page = list_last_entry(pages, struct page, lru);
a83d50bc
KM		 1893
1894 prefetchw(&page->flags);
f1e88660
JK		 1895 list_del(&page->lru);
1896 if (add_to_page_cache_lru(page, mapping,
4969c06a 1897 page_index(page),
8a5c743e 1898 readahead_gfp_mask(mapping)))
f1e88660
JK		 1899 goto next_page;
1900 }
1901
2df0ab04
CY		 1902 ret = f2fs_read_single_page(inode, page, nr_pages, &map, &bio,
1903 &last_block_in_bio, is_readahead);
1904 if (ret) {
1905 SetPageError(page);
09cbfeaf 1906 zero_user_segment(page, 0, PAGE_SIZE);
f1e88660 1907 unlock_page(page);
f1e88660 1908 }
f1e88660
JK		 1909next_page:
1910 if (pages)
09cbfeaf 1911 put_page(page);
f1e88660
JK		 1912 }
1913 BUG_ON(pages && !list_empty(pages));
1914 if (bio)
4fc29c1a 1915 __submit_bio(F2FS_I_SB(inode), bio, DATA);
2df0ab04 1916 return pages ? 0 : ret;
f1e88660
JK		 1917}
1918
eb47b800
JK		 1919static int f2fs_read_data_page(struct file *file, struct page *page)
1920{
4969c06a 1921 struct inode *inode = page_file_mapping(page)->host;
b3d208f9 1922 int ret = -EAGAIN;
9ffe0fb5 1923
c20e89cd
CY		 1924 trace_f2fs_readpage(page, DATA);
1925
e1c42045 1926 /* If the file has inline data, try to read it directly */
9ffe0fb5
HL		 1927 if (f2fs_has_inline_data(inode))
1928 ret = f2fs_read_inline_data(inode, page);
b3d208f9 1929 if (ret == -EAGAIN)
4969c06a
JK		 1930 ret = f2fs_mpage_readpages(page_file_mapping(page),
1931 NULL, page, 1, false);
9ffe0fb5 1932 return ret;
eb47b800
JK		 1933}
1934
1935static int f2fs_read_data_pages(struct file *file,
1936 struct address_space *mapping,
1937 struct list_head *pages, unsigned nr_pages)
1938{
71cb4aff 1939 struct inode *inode = mapping->host;
939afa94 1940 struct page *page = list_last_entry(pages, struct page, lru);
b8c29400
CY		 1941
1942 trace_f2fs_readpages(inode, page, nr_pages);
9ffe0fb5
HL		 1943
1944 /* If the file has inline data, skip readpages */
1945 if (f2fs_has_inline_data(inode))
1946 return 0;
1947
e2e59414 1948 return f2fs_mpage_readpages(mapping, pages, NULL, nr_pages, true);
eb47b800
JK		 1949}
1950
7eab0c0d
HP		 1951static int encrypt_one_page(struct f2fs_io_info *fio)
1952{
1953 struct inode *inode = fio->page->mapping->host;
6aa58d8a 1954 struct page *mpage;
7eab0c0d
HP		 1955 gfp_t gfp_flags = GFP_NOFS;
1956
1958593e 1957 if (!f2fs_encrypted_file(inode))
7eab0c0d
HP		 1958 return 0;
1959
6dbb1796 1960 /* wait for GCed page writeback via META_MAPPING */
0ded69f6 1961 f2fs_wait_on_block_writeback(inode, fio->old_blkaddr);
7eab0c0d
HP		 1962
1963retry_encrypt:
53bc1d85
EB		 1964 fio->encrypted_page = fscrypt_encrypt_pagecache_blocks(fio->page,
1965 PAGE_SIZE, 0,
1966 gfp_flags);
6aa58d8a
CY		 1967 if (IS_ERR(fio->encrypted_page)) {
1968 /* flush pending IOs and wait for a while in the ENOMEM case */
1969 if (PTR_ERR(fio->encrypted_page) == -ENOMEM) {
1970 f2fs_flush_merged_writes(fio->sbi);
1971 congestion_wait(BLK_RW_ASYNC, HZ/50);
1972 gfp_flags |= __GFP_NOFAIL;
1973 goto retry_encrypt;
1974 }
1975 return PTR_ERR(fio->encrypted_page);
1976 }
7eab0c0d 1977
6aa58d8a
CY		 1978 mpage = find_lock_page(META_MAPPING(fio->sbi), fio->old_blkaddr);
1979 if (mpage) {
1980 if (PageUptodate(mpage))
1981 memcpy(page_address(mpage),
1982 page_address(fio->encrypted_page), PAGE_SIZE);
1983 f2fs_put_page(mpage, 1);
7eab0c0d 1984 }
6aa58d8a 1985 return 0;
7eab0c0d
HP		 1986}
1987
bb9e3bb8
CY		 1988static inline bool check_inplace_update_policy(struct inode *inode,
1989 struct f2fs_io_info *fio)
7eab0c0d 1990{
bb9e3bb8
CY		 1991 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1992 unsigned int policy = SM_I(sbi)->ipu_policy;
7eab0c0d 1993
bb9e3bb8
CY		 1994 if (policy & (0x1 << F2FS_IPU_FORCE))
1995 return true;
4d57b86d 1996 if (policy & (0x1 << F2FS_IPU_SSR) && f2fs_need_SSR(sbi))
bb9e3bb8
CY		 1997 return true;
1998 if (policy & (0x1 << F2FS_IPU_UTIL) &&
1999 utilization(sbi) > SM_I(sbi)->min_ipu_util)
2000 return true;
4d57b86d 2001 if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && f2fs_need_SSR(sbi) &&
bb9e3bb8
CY		 2002 utilization(sbi) > SM_I(sbi)->min_ipu_util)
2003 return true;
2004
2005 /*
2006 * IPU for rewrite async pages
2007 */
2008 if (policy & (0x1 << F2FS_IPU_ASYNC) &&
2009 fio && fio->op == REQ_OP_WRITE &&
2010 !(fio->op_flags & REQ_SYNC) &&
62230e0d 2011 !IS_ENCRYPTED(inode))
bb9e3bb8
CY		 2012 return true;
2013
2014 /* this is only set during fdatasync */
2015 if (policy & (0x1 << F2FS_IPU_FSYNC) &&
2016 is_inode_flag_set(inode, FI_NEED_IPU))
2017 return true;
2018
4354994f
DR		 2019 if (unlikely(fio && is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2020 !f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2021 return true;
2022
bb9e3bb8
CY		 2023 return false;
2024}
2025
4d57b86d 2026bool f2fs_should_update_inplace(struct inode *inode, struct f2fs_io_info *fio)
bb9e3bb8 2027{
1ad71a27
JK		 2028 if (f2fs_is_pinned_file(inode))
2029 return true;
bb9e3bb8
CY		 2030
2031 /* if this is cold file, we should overwrite to avoid fragmentation */
2032 if (file_is_cold(inode))
2033 return true;
2034
2035 return check_inplace_update_policy(inode, fio);
2036}
2037
4d57b86d 2038bool f2fs_should_update_outplace(struct inode *inode, struct f2fs_io_info *fio)
bb9e3bb8
CY		 2039{
2040 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2041
2042 if (test_opt(sbi, LFS))
2043 return true;
2044 if (S_ISDIR(inode->i_mode))
2045 return true;
af033b2a
CY		 2046 if (IS_NOQUOTA(inode))
2047 return true;
bb9e3bb8
CY		 2048 if (f2fs_is_atomic_file(inode))
2049 return true;
2050 if (fio) {
2051 if (is_cold_data(fio->page))
2052 return true;
2053 if (IS_ATOMIC_WRITTEN_PAGE(fio->page))
2054 return true;
4354994f
DR		 2055 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2056 f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2057 return true;
bb9e3bb8
CY		 2058 }
2059 return false;
2060}
2061
7eab0c0d
HP		 2062static inline bool need_inplace_update(struct f2fs_io_info *fio)
2063{
2064 struct inode *inode = fio->page->mapping->host;
2065
4d57b86d 2066 if (f2fs_should_update_outplace(inode, fio))
7eab0c0d
HP		 2067 return false;
2068
4d57b86d 2069 return f2fs_should_update_inplace(inode, fio);
7eab0c0d
HP		 2070}
2071
4d57b86d 2072int f2fs_do_write_data_page(struct f2fs_io_info *fio)
eb47b800 2073{
05ca3632 2074 struct page *page = fio->page;
eb47b800 2075 struct inode *inode = page->mapping->host;
eb47b800 2076 struct dnode_of_data dn;
e959c8f5 2077 struct extent_info ei = {0,0,0};
7735730d 2078 struct node_info ni;
e959c8f5 2079 bool ipu_force = false;
eb47b800
JK		 2080 int err = 0;
2081
2082 set_new_dnode(&dn, inode, NULL, NULL, 0);
e959c8f5
HP		 2083 if (need_inplace_update(fio) &&
2084 f2fs_lookup_extent_cache(inode, page->index, &ei)) {
2085 fio->old_blkaddr = ei.blk + page->index - ei.fofs;
a817737e 2086
c9b60788 2087 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
93770ab7 2088 DATA_GENERIC_ENHANCE))
10f966bb 2089 return -EFSCORRUPTED;
c9b60788
CY		 2090
2091 ipu_force = true;
2092 fio->need_lock = LOCK_DONE;
2093 goto got_it;
e959c8f5 2094 }
279d6df2 2095
d29460e5
JK		 2096 /* Deadlock due to between page->lock and f2fs_lock_op */
2097 if (fio->need_lock == LOCK_REQ && !f2fs_trylock_op(fio->sbi))
2098 return -EAGAIN;
279d6df2 2099
4d57b86d 2100 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
eb47b800 2101 if (err)
279d6df2 2102 goto out;
eb47b800 2103
28bc106b 2104 fio->old_blkaddr = dn.data_blkaddr;
eb47b800
JK		 2105
2106 /* This page is already truncated */
7a9d7548 2107 if (fio->old_blkaddr == NULL_ADDR) {
2bca1e23 2108 ClearPageUptodate(page);
2baf0781 2109 clear_cold_data(page);
eb47b800 2110 goto out_writepage;
2bca1e23 2111 }
e959c8f5 2112got_it:
c9b60788
CY		 2113 if (__is_valid_data_blkaddr(fio->old_blkaddr) &&
2114 !f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
93770ab7 2115 DATA_GENERIC_ENHANCE)) {
10f966bb 2116 err = -EFSCORRUPTED;
c9b60788
CY		 2117 goto out_writepage;
2118 }
eb47b800
JK		 2119 /*
2120 * If current allocation needs SSR,
2121 * it had better in-place writes for updated data.
2122 */
93770ab7
CY		 2123 if (ipu_force ||
2124 (__is_valid_data_blkaddr(fio->old_blkaddr) &&
7b525dd0 2125 need_inplace_update(fio))) {
cc15620b
JK		 2126 err = encrypt_one_page(fio);
2127 if (err)
2128 goto out_writepage;
2129
2130 set_page_writeback(page);
17c50035 2131 ClearPageError(page);
279d6df2 2132 f2fs_put_dnode(&dn);
cc15620b 2133 if (fio->need_lock == LOCK_REQ)
279d6df2 2134 f2fs_unlock_op(fio->sbi);
4d57b86d 2135 err = f2fs_inplace_write_data(fio);
6492a335
CY		 2136 if (err) {
2137 if (f2fs_encrypted_file(inode))
d2d0727b 2138 fscrypt_finalize_bounce_page(&fio->encrypted_page);
6492a335
CY		 2139 if (PageWriteback(page))
2140 end_page_writeback(page);
cd23ffa9
CY		 2141 } else {
2142 set_inode_flag(inode, FI_UPDATE_WRITE);
6492a335 2143 }
7eab0c0d 2144 trace_f2fs_do_write_data_page(fio->page, IPU);
279d6df2 2145 return err;
eb47b800 2146 }
279d6df2 2147
cc15620b
JK		 2148 if (fio->need_lock == LOCK_RETRY) {
2149 if (!f2fs_trylock_op(fio->sbi)) {
2150 err = -EAGAIN;
2151 goto out_writepage;
2152 }
2153 fio->need_lock = LOCK_REQ;
2154 }
2155
7735730d
CY		 2156 err = f2fs_get_node_info(fio->sbi, dn.nid, &ni);
2157 if (err)
2158 goto out_writepage;
2159
2160 fio->version = ni.version;
2161
cc15620b
JK		 2162 err = encrypt_one_page(fio);
2163 if (err)
2164 goto out_writepage;
2165
2166 set_page_writeback(page);
17c50035 2167 ClearPageError(page);
cc15620b 2168
279d6df2 2169 /* LFS mode write path */
4d57b86d 2170 f2fs_outplace_write_data(&dn, fio);
279d6df2
HP		 2171 trace_f2fs_do_write_data_page(page, OPU);
2172 set_inode_flag(inode, FI_APPEND_WRITE);
2173 if (page->index == 0)
2174 set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
eb47b800
JK		 2175out_writepage:
2176 f2fs_put_dnode(&dn);
279d6df2 2177out:
cc15620b 2178 if (fio->need_lock == LOCK_REQ)
279d6df2 2179 f2fs_unlock_op(fio->sbi);
eb47b800
JK		 2180 return err;
2181}
2182
d68f735b 2183static int __write_data_page(struct page *page, bool *submitted,
8648de2c
CY		 2184 struct bio **bio,
2185 sector_t *last_block,
b0af6d49
CY		 2186 struct writeback_control *wbc,
2187 enum iostat_type io_type)
eb47b800
JK		 2188{
2189 struct inode *inode = page->mapping->host;
4081363f 2190 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
eb47b800
JK		 2191 loff_t i_size = i_size_read(inode);
2192 const pgoff_t end_index = ((unsigned long long) i_size)
09cbfeaf 2193 >> PAGE_SHIFT;
1f0d5c91 2194 loff_t psize = (loff_t)(page->index + 1) << PAGE_SHIFT;
9ffe0fb5 2195 unsigned offset = 0;
39936837 2196 bool need_balance_fs = false;
eb47b800 2197 int err = 0;
458e6197 2198 struct f2fs_io_info fio = {
05ca3632 2199 .sbi = sbi,
39d787be 2200 .ino = inode->i_ino,
458e6197 2201 .type = DATA,
04d328de 2202 .op = REQ_OP_WRITE,
7637241e 2203 .op_flags = wbc_to_write_flags(wbc),
e959c8f5 2204 .old_blkaddr = NULL_ADDR,
05ca3632 2205 .page = page,
4375a336 2206 .encrypted_page = NULL,
d68f735b 2207 .submitted = false,
cc15620b 2208 .need_lock = LOCK_RETRY,
b0af6d49 2209 .io_type = io_type,
578c6478 2210 .io_wbc = wbc,
8648de2c
CY		 2211 .bio = bio,
2212 .last_block = last_block,
458e6197 2213 };
eb47b800 2214
ecda0de3
CY		 2215 trace_f2fs_writepage(page, DATA);
2216
db198ae0
CY		 2217 /* we should bypass data pages to proceed the kworkder jobs */
2218 if (unlikely(f2fs_cp_error(sbi))) {
2219 mapping_set_error(page->mapping, -EIO);
1174abfd
CY		 2220 /*
2221 * don't drop any dirty dentry pages for keeping lastest
2222 * directory structure.
2223 */
2224 if (S_ISDIR(inode->i_mode))
2225 goto redirty_out;
db198ae0
CY		 2226 goto out;
2227 }
2228
0771fcc7
CY		 2229 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
2230 goto redirty_out;
2231
95ae251f 2232 if (page->index < end_index || f2fs_verity_in_progress(inode))
39936837 2233 goto write;
eb47b800
JK		 2234
2235 /*
2236 * If the offset is out-of-range of file size,
2237 * this page does not have to be written to disk.
2238 */
09cbfeaf 2239 offset = i_size & (PAGE_SIZE - 1);
76f60268 2240 if ((page->index >= end_index + 1) || !offset)
39936837 2241 goto out;
eb47b800 2242
09cbfeaf 2243 zero_user_segment(page, offset, PAGE_SIZE);
39936837 2244write:
1e84371f
JK		 2245 if (f2fs_is_drop_cache(inode))
2246 goto out;
e6e5f561
JK		 2247 /* we should not write 0'th page having journal header */
2248 if (f2fs_is_volatile_file(inode) && (!page->index ||
2249 (!wbc->for_reclaim &&
4d57b86d 2250 f2fs_available_free_memory(sbi, BASE_CHECK))))
1e84371f 2251 goto redirty_out;
eb47b800 2252
39936837 2253 /* Dentry blocks are controlled by checkpoint */
eb47b800 2254 if (S_ISDIR(inode->i_mode)) {
cc15620b 2255 fio.need_lock = LOCK_DONE;
4d57b86d 2256 err = f2fs_do_write_data_page(&fio);
8618b881
JK		 2257 goto done;
2258 }
9ffe0fb5 2259
8618b881 2260 if (!wbc->for_reclaim)
39936837 2261 need_balance_fs = true;
7f3037a5 2262 else if (has_not_enough_free_secs(sbi, 0, 0))
39936837 2263 goto redirty_out;
ef095d19
JK		 2264 else
2265 set_inode_flag(inode, FI_HOT_DATA);
eb47b800 2266
b3d208f9 2267 err = -EAGAIN;
dd7b2333 2268 if (f2fs_has_inline_data(inode)) {
b3d208f9 2269 err = f2fs_write_inline_data(inode, page);
dd7b2333
YH		 2270 if (!err)
2271 goto out;
2272 }
279d6df2 2273
cc15620b 2274 if (err == -EAGAIN) {
4d57b86d 2275 err = f2fs_do_write_data_page(&fio);
cc15620b
JK		 2276 if (err == -EAGAIN) {
2277 fio.need_lock = LOCK_REQ;
4d57b86d 2278 err = f2fs_do_write_data_page(&fio);
cc15620b
JK		 2279 }
2280 }
a0d00fad 2281
eb449797
CY		 2282 if (err) {
2283 file_set_keep_isize(inode);
2284 } else {
2285 down_write(&F2FS_I(inode)->i_sem);
2286 if (F2FS_I(inode)->last_disk_size < psize)
2287 F2FS_I(inode)->last_disk_size = psize;
2288 up_write(&F2FS_I(inode)->i_sem);
2289 }
279d6df2 2290
8618b881
JK		 2291done:
2292 if (err && err != -ENOENT)
2293 goto redirty_out;
eb47b800 2294
39936837 2295out:
a7ffdbe2 2296 inode_dec_dirty_pages(inode);
2baf0781 2297 if (err) {
2bca1e23 2298 ClearPageUptodate(page);
2baf0781
CY		 2299 clear_cold_data(page);
2300 }
0c3a5797
CY		 2301
2302 if (wbc->for_reclaim) {
bab475c5 2303 f2fs_submit_merged_write_cond(sbi, NULL, page, 0, DATA);
ef095d19 2304 clear_inode_flag(inode, FI_HOT_DATA);
4d57b86d 2305 f2fs_remove_dirty_inode(inode);
d68f735b 2306 submitted = NULL;
0c3a5797
CY		 2307 }
2308
eb47b800 2309 unlock_page(page);
186857c5 2310 if (!S_ISDIR(inode->i_mode) && !IS_NOQUOTA(inode) &&
0b20fcec 2311 !F2FS_I(inode)->cp_task)
a7881893 2312 f2fs_balance_fs(sbi, need_balance_fs);
0c3a5797 2313
d68f735b 2314 if (unlikely(f2fs_cp_error(sbi))) {
b9109b0e 2315 f2fs_submit_merged_write(sbi, DATA);
0b20fcec 2316 f2fs_submit_merged_ipu_write(sbi, bio, NULL);
d68f735b
JK		 2317 submitted = NULL;
2318 }
2319
2320 if (submitted)
2321 *submitted = fio.submitted;
0c3a5797 2322
eb47b800
JK		 2323 return 0;
2324
eb47b800 2325redirty_out:
76f60268 2326 redirty_page_for_writepage(wbc, page);
5b19d284
JK		 2327 /*
2328 * pageout() in MM traslates EAGAIN, so calls handle_write_error()
2329 * -> mapping_set_error() -> set_bit(AS_EIO, ...).
2330 * file_write_and_wait_range() will see EIO error, which is critical
2331 * to return value of fsync() followed by atomic_write failure to user.
2332 */
2333 if (!err || wbc->for_reclaim)
0002b61b 2334 return AOP_WRITEPAGE_ACTIVATE;
b230e6ca
JK		 2335 unlock_page(page);
2336 return err;
fa9150a8
NJ		 2337}
2338
f566bae8
JK		 2339static int f2fs_write_data_page(struct page *page,
2340 struct writeback_control *wbc)
2341{
8648de2c 2342 return __write_data_page(page, NULL, NULL, NULL, wbc, FS_DATA_IO);
f566bae8
JK		 2343}
2344
8f46dcae
CY		 2345/*
2346 * This function was copied from write_cche_pages from mm/page-writeback.c.
2347 * The major change is making write step of cold data page separately from
2348 * warm/hot data page.
2349 */
2350static int f2fs_write_cache_pages(struct address_space *mapping,
b0af6d49
CY		 2351 struct writeback_control *wbc,
2352 enum iostat_type io_type)
8f46dcae
CY		 2353{
2354 int ret = 0;
2355 int done = 0;
2356 struct pagevec pvec;
c29fd0c0 2357 struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
8648de2c
CY		 2358 struct bio *bio = NULL;
2359 sector_t last_block;
8f46dcae
CY		 2360 int nr_pages;
2361 pgoff_t uninitialized_var(writeback_index);
2362 pgoff_t index;
2363 pgoff_t end; /* Inclusive */
2364 pgoff_t done_index;
2365 int cycled;
2366 int range_whole = 0;
10bbd235 2367 xa_mark_t tag;
bab475c5 2368 int nwritten = 0;
8f46dcae 2369
86679820 2370 pagevec_init(&pvec);
46ae957f 2371
ef095d19
JK		 2372 if (get_dirty_pages(mapping->host) <=
2373 SM_I(F2FS_M_SB(mapping))->min_hot_blocks)
2374 set_inode_flag(mapping->host, FI_HOT_DATA);
2375 else
2376 clear_inode_flag(mapping->host, FI_HOT_DATA);
2377
8f46dcae
CY		 2378 if (wbc->range_cyclic) {
2379 writeback_index = mapping->writeback_index; /* prev offset */
2380 index = writeback_index;
2381 if (index == 0)
2382 cycled = 1;
2383 else
2384 cycled = 0;
2385 end = -1;
2386 } else {
09cbfeaf
KS		 2387 index = wbc->range_start >> PAGE_SHIFT;
2388 end = wbc->range_end >> PAGE_SHIFT;
8f46dcae
CY		 2389 if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
2390 range_whole = 1;
2391 cycled = 1; /* ignore range_cyclic tests */
2392 }
2393 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2394 tag = PAGECACHE_TAG_TOWRITE;
2395 else
2396 tag = PAGECACHE_TAG_DIRTY;
2397retry:
2398 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2399 tag_pages_for_writeback(mapping, index, end);
2400 done_index = index;
2401 while (!done && (index <= end)) {
2402 int i;
2403
69c4f35d 2404 nr_pages = pagevec_lookup_range_tag(&pvec, mapping, &index, end,
67fd707f 2405 tag);
8f46dcae
CY		 2406 if (nr_pages == 0)
2407 break;
2408
2409 for (i = 0; i < nr_pages; i++) {
2410 struct page *page = pvec.pages[i];
d68f735b 2411 bool submitted = false;
8f46dcae 2412
f8de4331 2413 /* give a priority to WB_SYNC threads */
c29fd0c0 2414 if (atomic_read(&sbi->wb_sync_req[DATA]) &&
f8de4331
CY		 2415 wbc->sync_mode == WB_SYNC_NONE) {
2416 done = 1;
2417 break;
2418 }
2419
8f46dcae 2420 done_index = page->index;
d29460e5 2421retry_write:
8f46dcae
CY		 2422 lock_page(page);
2423
2424 if (unlikely(page->mapping != mapping)) {
2425continue_unlock:
2426 unlock_page(page);
2427 continue;
2428 }
2429
2430 if (!PageDirty(page)) {
2431 /* someone wrote it for us */
2432 goto continue_unlock;
2433 }
2434
8f46dcae 2435 if (PageWriteback(page)) {
0b20fcec 2436 if (wbc->sync_mode != WB_SYNC_NONE)
fec1d657 2437 f2fs_wait_on_page_writeback(page,
bae0ee7a 2438 DATA, true, true);
0b20fcec 2439 else
8f46dcae
CY		 2440 goto continue_unlock;
2441 }
2442
8f46dcae
CY		 2443 if (!clear_page_dirty_for_io(page))
2444 goto continue_unlock;
2445
8648de2c
CY		 2446 ret = __write_data_page(page, &submitted, &bio,
2447 &last_block, wbc, io_type);
8f46dcae 2448 if (unlikely(ret)) {
0002b61b
CY		 2449 /*
2450 * keep nr_to_write, since vfs uses this to
2451 * get # of written pages.
2452 */
2453 if (ret == AOP_WRITEPAGE_ACTIVATE) {
2454 unlock_page(page);
2455 ret = 0;
2456 continue;
d29460e5
JK		 2457 } else if (ret == -EAGAIN) {
2458 ret = 0;
2459 if (wbc->sync_mode == WB_SYNC_ALL) {
2460 cond_resched();
2461 congestion_wait(BLK_RW_ASYNC,
2462 HZ/50);
2463 goto retry_write;
2464 }
2465 continue;
0002b61b 2466 }
b230e6ca
JK		 2467 done_index = page->index + 1;
2468 done = 1;
2469 break;
d68f735b 2470 } else if (submitted) {
bab475c5 2471 nwritten++;
8f46dcae
CY		 2472 }
2473
f8de4331 2474 if (--wbc->nr_to_write <= 0 &&
687de7f1 2475 wbc->sync_mode == WB_SYNC_NONE) {
8f46dcae
CY		 2476 done = 1;
2477 break;
2478 }
2479 }
2480 pagevec_release(&pvec);
2481 cond_resched();
2482 }
2483
8f46dcae
CY		 2484 if (!cycled && !done) {
2485 cycled = 1;
2486 index = 0;
2487 end = writeback_index - 1;
2488 goto retry;
2489 }
2490 if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
2491 mapping->writeback_index = done_index;
2492
bab475c5 2493 if (nwritten)
b9109b0e 2494 f2fs_submit_merged_write_cond(F2FS_M_SB(mapping), mapping->host,
bab475c5 2495 NULL, 0, DATA);
8648de2c
CY		 2496 /* submit cached bio of IPU write */
2497 if (bio)
0b20fcec 2498 f2fs_submit_merged_ipu_write(sbi, &bio, NULL);
6ca56ca4 2499
8f46dcae
CY		 2500 return ret;
2501}
2502
853137ce
JK		 2503static inline bool __should_serialize_io(struct inode *inode,
2504 struct writeback_control *wbc)
2505{
2506 if (!S_ISREG(inode->i_mode))
2507 return false;
af033b2a
CY		 2508 if (IS_NOQUOTA(inode))
2509 return false;
040d2bb3
CY		 2510 /* to avoid deadlock in path of data flush */
2511 if (F2FS_I(inode)->cp_task)
2512 return false;
853137ce
JK		 2513 if (wbc->sync_mode != WB_SYNC_ALL)
2514 return true;
2515 if (get_dirty_pages(inode) >= SM_I(F2FS_I_SB(inode))->min_seq_blocks)
2516 return true;
2517 return false;
2518}
2519
fc99fe27 2520static int __f2fs_write_data_pages(struct address_space *mapping,
b0af6d49
CY		 2521 struct writeback_control *wbc,
2522 enum iostat_type io_type)
eb47b800
JK		 2523{
2524 struct inode *inode = mapping->host;
4081363f 2525 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
9dfa1baf 2526 struct blk_plug plug;
eb47b800 2527 int ret;
853137ce 2528 bool locked = false;
eb47b800 2529
cfb185a1 2530 /* deal with chardevs and other special file */
2531 if (!mapping->a_ops->writepage)
2532 return 0;
2533
6a290544
CY		 2534 /* skip writing if there is no dirty page in this inode */
2535 if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE)
2536 return 0;
2537
0771fcc7
CY		 2538 /* during POR, we don't need to trigger writepage at all. */
2539 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
2540 goto skip_write;
2541
af033b2a
CY		 2542 if ((S_ISDIR(inode->i_mode) || IS_NOQUOTA(inode)) &&
2543 wbc->sync_mode == WB_SYNC_NONE &&
a1257023 2544 get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) &&
4d57b86d 2545 f2fs_available_free_memory(sbi, DIRTY_DENTS))
a1257023
JK		 2546 goto skip_write;
2547
d323d005 2548 /* skip writing during file defragment */
91942321 2549 if (is_inode_flag_set(inode, FI_DO_DEFRAG))
d323d005
CY		 2550 goto skip_write;
2551
d31c7c3f
YH		 2552 trace_f2fs_writepages(mapping->host, wbc, DATA);
2553
687de7f1
JK		 2554 /* to avoid spliting IOs due to mixed WB_SYNC_ALL and WB_SYNC_NONE */
2555 if (wbc->sync_mode == WB_SYNC_ALL)
c29fd0c0
CY		 2556 atomic_inc(&sbi->wb_sync_req[DATA]);
2557 else if (atomic_read(&sbi->wb_sync_req[DATA]))
687de7f1
JK		 2558 goto skip_write;
2559
853137ce
JK		 2560 if (__should_serialize_io(inode, wbc)) {
2561 mutex_lock(&sbi->writepages);
2562 locked = true;
2563 }
2564
9dfa1baf 2565 blk_start_plug(&plug);
b0af6d49 2566 ret = f2fs_write_cache_pages(mapping, wbc, io_type);
9dfa1baf 2567 blk_finish_plug(&plug);
687de7f1 2568
853137ce
JK		 2569 if (locked)
2570 mutex_unlock(&sbi->writepages);
2571
687de7f1 2572 if (wbc->sync_mode == WB_SYNC_ALL)
c29fd0c0 2573 atomic_dec(&sbi->wb_sync_req[DATA]);
28ea6162
JK		 2574 /*
2575 * if some pages were truncated, we cannot guarantee its mapping->host
2576 * to detect pending bios.
2577 */
458e6197 2578
4d57b86d 2579 f2fs_remove_dirty_inode(inode);
eb47b800 2580 return ret;
d3baf95d
JK		 2581
2582skip_write:
a7ffdbe2 2583 wbc->pages_skipped += get_dirty_pages(inode);
d31c7c3f 2584 trace_f2fs_writepages(mapping->host, wbc, DATA);
d3baf95d 2585 return 0;
eb47b800
JK		 2586}
2587
b0af6d49
CY		 2588static int f2fs_write_data_pages(struct address_space *mapping,
2589 struct writeback_control *wbc)
2590{
2591 struct inode *inode = mapping->host;
2592
2593 return __f2fs_write_data_pages(mapping, wbc,
2594 F2FS_I(inode)->cp_task == current ?
2595 FS_CP_DATA_IO : FS_DATA_IO);
2596}
2597
3aab8f82
CY		 2598static void f2fs_write_failed(struct address_space *mapping, loff_t to)
2599{
2600 struct inode *inode = mapping->host;
819d9153 2601 loff_t i_size = i_size_read(inode);
3aab8f82 2602
3f188c23
JK		 2603 if (IS_NOQUOTA(inode))
2604 return;
2605
95ae251f
EB		 2606 /* In the fs-verity case, f2fs_end_enable_verity() does the truncate */
2607 if (to > i_size && !f2fs_verity_in_progress(inode)) {
a33c1502 2608 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
6f8d4455 2609 down_write(&F2FS_I(inode)->i_mmap_sem);
a33c1502 2610
819d9153 2611 truncate_pagecache(inode, i_size);
3f188c23 2612 f2fs_truncate_blocks(inode, i_size, true);
a33c1502 2613
5a3a2d83 2614 up_write(&F2FS_I(inode)->i_mmap_sem);
6f8d4455 2615 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3aab8f82
CY		 2616 }
2617}
2618
2aadac08
JK		 2619static int prepare_write_begin(struct f2fs_sb_info *sbi,
2620 struct page *page, loff_t pos, unsigned len,
2621 block_t *blk_addr, bool *node_changed)
2622{
2623 struct inode *inode = page->mapping->host;
2624 pgoff_t index = page->index;
2625 struct dnode_of_data dn;
2626 struct page *ipage;
b4d07a3e 2627 bool locked = false;
e15882b6 2628 struct extent_info ei = {0,0,0};
2aadac08 2629 int err = 0;
2866fb16 2630 int flag;
2aadac08 2631
24b84912
JK		 2632 /*
2633 * we already allocated all the blocks, so we don't need to get
2634 * the block addresses when there is no need to fill the page.
2635 */
dc91de78 2636 if (!f2fs_has_inline_data(inode) && len == PAGE_SIZE &&
95ae251f
EB		 2637 !is_inode_flag_set(inode, FI_NO_PREALLOC) &&
2638 !f2fs_verity_in_progress(inode))
24b84912
JK		 2639 return 0;
2640
2866fb16
SY		 2641 /* f2fs_lock_op avoids race between write CP and convert_inline_page */
2642 if (f2fs_has_inline_data(inode) && pos + len > MAX_INLINE_DATA(inode))
2643 flag = F2FS_GET_BLOCK_DEFAULT;
2644 else
2645 flag = F2FS_GET_BLOCK_PRE_AIO;
2646
b4d07a3e 2647 if (f2fs_has_inline_data(inode) ||
09cbfeaf 2648 (pos & PAGE_MASK) >= i_size_read(inode)) {
2866fb16 2649 __do_map_lock(sbi, flag, true);
b4d07a3e
JK		 2650 locked = true;
2651 }
2652restart:
2aadac08 2653 /* check inline_data */
4d57b86d 2654 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2aadac08
JK		 2655 if (IS_ERR(ipage)) {
2656 err = PTR_ERR(ipage);
2657 goto unlock_out;
2658 }
2659
2660 set_new_dnode(&dn, inode, ipage, ipage, 0);
2661
2662 if (f2fs_has_inline_data(inode)) {
f2470371 2663 if (pos + len <= MAX_INLINE_DATA(inode)) {
4d57b86d 2664 f2fs_do_read_inline_data(page, ipage);
91942321 2665 set_inode_flag(inode, FI_DATA_EXIST);
ab47036d
CY		 2666 if (inode->i_nlink)
2667 set_inline_node(ipage);
2aadac08
JK		 2668 } else {
2669 err = f2fs_convert_inline_page(&dn, page);
2670 if (err)
b4d07a3e
JK		 2671 goto out;
2672 if (dn.data_blkaddr == NULL_ADDR)
2673 err = f2fs_get_block(&dn, index);
2674 }
2675 } else if (locked) {
2676 err = f2fs_get_block(&dn, index);
2677 } else {
2678 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
2679 dn.data_blkaddr = ei.blk + index - ei.fofs;
2680 } else {
b4d07a3e 2681 /* hole case */
4d57b86d 2682 err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
4da7bf5a 2683 if (err || dn.data_blkaddr == NULL_ADDR) {
b4d07a3e 2684 f2fs_put_dnode(&dn);
59c9081b
YH		 2685 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO,
2686 true);
2866fb16 2687 WARN_ON(flag != F2FS_GET_BLOCK_PRE_AIO);
b4d07a3e
JK		 2688 locked = true;
2689 goto restart;
2690 }
2aadac08
JK		 2691 }
2692 }
b4d07a3e 2693
2aadac08
JK		 2694 /* convert_inline_page can make node_changed */
2695 *blk_addr = dn.data_blkaddr;
2696 *node_changed = dn.node_changed;
b4d07a3e 2697out:
2aadac08
JK		 2698 f2fs_put_dnode(&dn);
2699unlock_out:
b4d07a3e 2700 if (locked)
2866fb16 2701 __do_map_lock(sbi, flag, false);
2aadac08
JK		 2702 return err;
2703}
2704
eb47b800
JK		 2705static int f2fs_write_begin(struct file *file, struct address_space *mapping,
2706 loff_t pos, unsigned len, unsigned flags,
2707 struct page **pagep, void **fsdata)
2708{
2709 struct inode *inode = mapping->host;
4081363f 2710 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
86531d6b 2711 struct page *page = NULL;
09cbfeaf 2712 pgoff_t index = ((unsigned long long) pos) >> PAGE_SHIFT;
a2e2e76b 2713 bool need_balance = false, drop_atomic = false;
2aadac08 2714 block_t blkaddr = NULL_ADDR;
eb47b800
JK		 2715 int err = 0;
2716
62aed044
CY		 2717 trace_f2fs_write_begin(inode, pos, len, flags);
2718
00e09c0b
CY		 2719 if (!f2fs_is_checkpoint_ready(sbi)) {
2720 err = -ENOSPC;
4354994f 2721 goto fail;
00e09c0b 2722 }
4354994f 2723
455e3a58
JK		 2724 if ((f2fs_is_atomic_file(inode) &&
2725 !f2fs_available_free_memory(sbi, INMEM_PAGES)) ||
2726 is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
57864ae5 2727 err = -ENOMEM;
a2e2e76b 2728 drop_atomic = true;
57864ae5
JK		 2729 goto fail;
2730 }
2731
5f727395
JK		 2732 /*
2733 * We should check this at this moment to avoid deadlock on inode page
2734 * and #0 page. The locking rule for inline_data conversion should be:
2735 * lock_page(page #0) -> lock_page(inode_page)
2736 */
2737 if (index != 0) {
2738 err = f2fs_convert_inline_inode(inode);
2739 if (err)
2740 goto fail;
2741 }
afcb7ca0 2742repeat:
86d54795
JK		 2743 /*
2744 * Do not use grab_cache_page_write_begin() to avoid deadlock due to
2745 * wait_for_stable_page. Will wait that below with our IO control.
2746 */
01eccef7 2747 page = f2fs_pagecache_get_page(mapping, index,
86d54795 2748 FGP_LOCK | FGP_WRITE | FGP_CREAT, GFP_NOFS);
3aab8f82
CY		 2749 if (!page) {
2750 err = -ENOMEM;
2751 goto fail;
2752 }
d5f66990 2753
eb47b800
JK		 2754 *pagep = page;
2755
2aadac08
JK		 2756 err = prepare_write_begin(sbi, page, pos, len,
2757 &blkaddr, &need_balance);
9ba69cf9 2758 if (err)
2aadac08 2759 goto fail;
9ba69cf9 2760
af033b2a
CY		 2761 if (need_balance && !IS_NOQUOTA(inode) &&
2762 has_not_enough_free_secs(sbi, 0, 0)) {
2a340760 2763 unlock_page(page);
2c4db1a6 2764 f2fs_balance_fs(sbi, true);
2a340760
JK		 2765 lock_page(page);
2766 if (page->mapping != mapping) {
2767 /* The page got truncated from under us */
2768 f2fs_put_page(page, 1);
2769 goto repeat;
2770 }
2771 }
2772
bae0ee7a 2773 f2fs_wait_on_page_writeback(page, DATA, false, true);
b3d208f9 2774
649d7df2
JK		 2775 if (len == PAGE_SIZE || PageUptodate(page))
2776 return 0;
eb47b800 2777
95ae251f
EB		 2778 if (!(pos & (PAGE_SIZE - 1)) && (pos + len) >= i_size_read(inode) &&
2779 !f2fs_verity_in_progress(inode)) {
746e2403
YH		 2780 zero_user_segment(page, len, PAGE_SIZE);
2781 return 0;
2782 }
2783
2aadac08 2784 if (blkaddr == NEW_ADDR) {
09cbfeaf 2785 zero_user_segment(page, 0, PAGE_SIZE);
649d7df2 2786 SetPageUptodate(page);
eb47b800 2787 } else {
93770ab7
CY		 2788 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
2789 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 2790 err = -EFSCORRUPTED;
93770ab7
CY		 2791 goto fail;
2792 }
13ba41e3
JK		 2793 err = f2fs_submit_page_read(inode, page, blkaddr);
2794 if (err)
78682f79 2795 goto fail;
d54c795b 2796
393ff91f 2797 lock_page(page);
6bacf52f 2798 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 2799 f2fs_put_page(page, 1);
2800 goto repeat;
eb47b800 2801 }
1563ac75
CY		 2802 if (unlikely(!PageUptodate(page))) {
2803 err = -EIO;
2804 goto fail;
4375a336 2805 }
eb47b800 2806 }
eb47b800 2807 return 0;
9ba69cf9 2808
3aab8f82 2809fail:
86531d6b 2810 f2fs_put_page(page, 1);
3aab8f82 2811 f2fs_write_failed(mapping, pos + len);
a2e2e76b 2812 if (drop_atomic)
4d57b86d 2813 f2fs_drop_inmem_pages_all(sbi, false);
3aab8f82 2814 return err;
eb47b800
JK		 2815}
2816
a1dd3c13
JK		 2817static int f2fs_write_end(struct file *file,
2818 struct address_space *mapping,
2819 loff_t pos, unsigned len, unsigned copied,
2820 struct page *page, void *fsdata)
2821{
2822 struct inode *inode = page->mapping->host;
2823
dfb2bf38
CY		 2824 trace_f2fs_write_end(inode, pos, len, copied);
2825
649d7df2
JK		 2826 /*
2827 * This should be come from len == PAGE_SIZE, and we expect copied
2828 * should be PAGE_SIZE. Otherwise, we treat it with zero copied and
2829 * let generic_perform_write() try to copy data again through copied=0.
2830 */
2831 if (!PageUptodate(page)) {
746e2403 2832 if (unlikely(copied != len))
649d7df2
JK		 2833 copied = 0;
2834 else
2835 SetPageUptodate(page);
2836 }
2837 if (!copied)
2838 goto unlock_out;
2839
34ba94ba 2840 set_page_dirty(page);
a1dd3c13 2841
95ae251f
EB		 2842 if (pos + copied > i_size_read(inode) &&
2843 !f2fs_verity_in_progress(inode))
fc9581c8 2844 f2fs_i_size_write(inode, pos + copied);
649d7df2 2845unlock_out:
3024c9a1 2846 f2fs_put_page(page, 1);
d0239e1b 2847 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
a1dd3c13
JK		 2848 return copied;
2849}
2850
6f673763
OS		 2851static int check_direct_IO(struct inode *inode, struct iov_iter *iter,
2852 loff_t offset)
944fcfc1 2853{
8a56dd96
JK		 2854 unsigned i_blkbits = READ_ONCE(inode->i_blkbits);
2855 unsigned blkbits = i_blkbits;
2856 unsigned blocksize_mask = (1 << blkbits) - 1;
2857 unsigned long align = offset | iov_iter_alignment(iter);
2858 struct block_device *bdev = inode->i_sb->s_bdev;
2859
2860 if (align & blocksize_mask) {
2861 if (bdev)
2862 blkbits = blksize_bits(bdev_logical_block_size(bdev));
2863 blocksize_mask = (1 << blkbits) - 1;
2864 if (align & blocksize_mask)
2865 return -EINVAL;
2866 return 1;
2867 }
944fcfc1
JK		 2868 return 0;
2869}
2870
02b16d0a
CY		 2871static void f2fs_dio_end_io(struct bio *bio)
2872{
2873 struct f2fs_private_dio *dio = bio->bi_private;
2874
2875 dec_page_count(F2FS_I_SB(dio->inode),
2876 dio->write ? F2FS_DIO_WRITE : F2FS_DIO_READ);
2877
2878 bio->bi_private = dio->orig_private;
2879 bio->bi_end_io = dio->orig_end_io;
2880
5222595d 2881 kvfree(dio);
02b16d0a
CY		 2882
2883 bio_endio(bio);
2884}
2885
2886static void f2fs_dio_submit_bio(struct bio *bio, struct inode *inode,
2887 loff_t file_offset)
2888{
2889 struct f2fs_private_dio *dio;
2890 bool write = (bio_op(bio) == REQ_OP_WRITE);
02b16d0a
CY		 2891
2892 dio = f2fs_kzalloc(F2FS_I_SB(inode),
2893 sizeof(struct f2fs_private_dio), GFP_NOFS);
8e114038 2894 if (!dio)
02b16d0a 2895 goto out;
02b16d0a
CY		 2896
2897 dio->inode = inode;
2898 dio->orig_end_io = bio->bi_end_io;
2899 dio->orig_private = bio->bi_private;
2900 dio->write = write;
2901
2902 bio->bi_end_io = f2fs_dio_end_io;
2903 bio->bi_private = dio;
2904
2905 inc_page_count(F2FS_I_SB(inode),
2906 write ? F2FS_DIO_WRITE : F2FS_DIO_READ);
2907
2908 submit_bio(bio);
2909 return;
2910out:
2911 bio->bi_status = BLK_STS_IOERR;
2912 bio_endio(bio);
2913}
2914
c8b8e32d 2915static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
eb47b800 2916{
b439b103 2917 struct address_space *mapping = iocb->ki_filp->f_mapping;
3aab8f82 2918 struct inode *inode = mapping->host;
0cdd3195 2919 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
f847c699 2920 struct f2fs_inode_info *fi = F2FS_I(inode);
3aab8f82 2921 size_t count = iov_iter_count(iter);
c8b8e32d 2922 loff_t offset = iocb->ki_pos;
82e0a5aa 2923 int rw = iov_iter_rw(iter);
3aab8f82 2924 int err;
0cdd3195 2925 enum rw_hint hint = iocb->ki_hint;
63189b78 2926 int whint_mode = F2FS_OPTION(sbi).whint_mode;
f847c699 2927 bool do_opu;
944fcfc1 2928
b439b103 2929 err = check_direct_IO(inode, iter, offset);
b9d777b8 2930 if (err)
8a56dd96 2931 return err < 0 ? err : 0;
9ffe0fb5 2932
f847c699 2933 if (f2fs_force_buffered_io(inode, iocb, iter))
36abef4e 2934 return 0;
fcc85a4d 2935
f847c699
CY		 2936 do_opu = allow_outplace_dio(inode, iocb, iter);
2937
5302fb00 2938 trace_f2fs_direct_IO_enter(inode, offset, count, rw);
70407fad 2939
0cdd3195
HL		 2940 if (rw == WRITE && whint_mode == WHINT_MODE_OFF)
2941 iocb->ki_hint = WRITE_LIFE_NOT_SET;
2942
f847c699
CY		 2943 if (iocb->ki_flags & IOCB_NOWAIT) {
2944 if (!down_read_trylock(&fi->i_gc_rwsem[rw])) {
2945 iocb->ki_hint = hint;
2946 err = -EAGAIN;
2947 goto out;
2948 }
2949 if (do_opu && !down_read_trylock(&fi->i_gc_rwsem[READ])) {
2950 up_read(&fi->i_gc_rwsem[rw]);
b91050a8
HL		 2951 iocb->ki_hint = hint;
2952 err = -EAGAIN;
2953 goto out;
2954 }
f847c699
CY		 2955 } else {
2956 down_read(&fi->i_gc_rwsem[rw]);
2957 if (do_opu)
2958 down_read(&fi->i_gc_rwsem[READ]);
b91050a8
HL		 2959 }
2960
02b16d0a 2961 err = __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev,
f9d6d059
CY		 2962 iter, rw == WRITE ? get_data_block_dio_write :
2963 get_data_block_dio, NULL, f2fs_dio_submit_bio,
02b16d0a 2964 DIO_LOCKING | DIO_SKIP_HOLES);
f847c699
CY		 2965
2966 if (do_opu)
2967 up_read(&fi->i_gc_rwsem[READ]);
2968
2969 up_read(&fi->i_gc_rwsem[rw]);
82e0a5aa
CY		 2970
2971 if (rw == WRITE) {
0cdd3195
HL		 2972 if (whint_mode == WHINT_MODE_OFF)
2973 iocb->ki_hint = hint;
b0af6d49
CY		 2974 if (err > 0) {
2975 f2fs_update_iostat(F2FS_I_SB(inode), APP_DIRECT_IO,
2976 err);
f847c699
CY		 2977 if (!do_opu)
2978 set_inode_flag(inode, FI_UPDATE_WRITE);
b0af6d49 2979 } else if (err < 0) {
6bfc4919 2980 f2fs_write_failed(mapping, offset + count);
b0af6d49 2981 }
6bfc4919 2982 }
70407fad 2983
b91050a8 2984out:
5302fb00 2985 trace_f2fs_direct_IO_exit(inode, offset, count, rw, err);
70407fad 2986
3aab8f82 2987 return err;
eb47b800
JK		 2988}
2989
487261f3
CY		 2990void f2fs_invalidate_page(struct page *page, unsigned int offset,
2991 unsigned int length)
eb47b800
JK		 2992{
2993 struct inode *inode = page->mapping->host;
487261f3 2994 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
a7ffdbe2 2995
487261f3 2996 if (inode->i_ino >= F2FS_ROOT_INO(sbi) &&
09cbfeaf 2997 (offset % PAGE_SIZE || length != PAGE_SIZE))
a7ffdbe2
JK		 2998 return;
2999
487261f3 3000 if (PageDirty(page)) {
933439c8 3001 if (inode->i_ino == F2FS_META_INO(sbi)) {
487261f3 3002 dec_page_count(sbi, F2FS_DIRTY_META);
933439c8 3003 } else if (inode->i_ino == F2FS_NODE_INO(sbi)) {
487261f3 3004 dec_page_count(sbi, F2FS_DIRTY_NODES);
933439c8 3005 } else {
487261f3 3006 inode_dec_dirty_pages(inode);
4d57b86d 3007 f2fs_remove_dirty_inode(inode);
933439c8 3008 }
487261f3 3009 }
decd36b6 3010
2baf0781
CY		 3011 clear_cold_data(page);
3012
decd36b6 3013 if (IS_ATOMIC_WRITTEN_PAGE(page))
4d57b86d 3014 return f2fs_drop_inmem_page(inode, page);
decd36b6 3015
240a5915 3016 f2fs_clear_page_private(page);
eb47b800
JK		 3017}
3018
487261f3 3019int f2fs_release_page(struct page *page, gfp_t wait)
eb47b800 3020{
f68daeeb
JK		 3021 /* If this is dirty page, keep PagePrivate */
3022 if (PageDirty(page))
3023 return 0;
3024
decd36b6
CY		 3025 /* This is atomic written page, keep Private */
3026 if (IS_ATOMIC_WRITTEN_PAGE(page))
3027 return 0;
3028
2baf0781 3029 clear_cold_data(page);
240a5915 3030 f2fs_clear_page_private(page);
c3850aa1 3031 return 1;
eb47b800
JK		 3032}
3033
3034static int f2fs_set_data_page_dirty(struct page *page)
3035{
4969c06a 3036 struct inode *inode = page_file_mapping(page)->host;
eb47b800 3037
26c6b887
JK		 3038 trace_f2fs_set_page_dirty(page, DATA);
3039
237c0790
JK		 3040 if (!PageUptodate(page))
3041 SetPageUptodate(page);
4969c06a
JK		 3042 if (PageSwapCache(page))
3043 return __set_page_dirty_nobuffers(page);
34ba94ba 3044
5fe45743 3045 if (f2fs_is_atomic_file(inode) && !f2fs_is_commit_atomic_write(inode)) {
decd36b6 3046 if (!IS_ATOMIC_WRITTEN_PAGE(page)) {
4d57b86d 3047 f2fs_register_inmem_page(inode, page);
decd36b6
CY		 3048 return 1;
3049 }
3050 /*
3051 * Previously, this page has been registered, we just
3052 * return here.
3053 */
3054 return 0;
34ba94ba
JK		 3055 }
3056
eb47b800 3057 if (!PageDirty(page)) {
b87078ad 3058 __set_page_dirty_nobuffers(page);
4d57b86d 3059 f2fs_update_dirty_page(inode, page);
eb47b800
JK		 3060 return 1;
3061 }
3062 return 0;
3063}
3064
c01e54b7
JK		 3065static sector_t f2fs_bmap(struct address_space *mapping, sector_t block)
3066{
454ae7e5
CY		 3067 struct inode *inode = mapping->host;
3068
1d373a0e
JK		 3069 if (f2fs_has_inline_data(inode))
3070 return 0;
3071
3072 /* make sure allocating whole blocks */
3073 if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
3074 filemap_write_and_wait(mapping);
3075
e2b4e2bc 3076 return generic_block_bmap(mapping, block, get_data_block_bmap);
429511cd
CY		 3077}
3078
5b7a487c
WG		 3079#ifdef CONFIG_MIGRATION
3080#include <linux/migrate.h>
3081
3082int f2fs_migrate_page(struct address_space *mapping,
3083 struct page *newpage, struct page *page, enum migrate_mode mode)
3084{
3085 int rc, extra_count;
3086 struct f2fs_inode_info *fi = F2FS_I(mapping->host);
3087 bool atomic_written = IS_ATOMIC_WRITTEN_PAGE(page);
3088
3089 BUG_ON(PageWriteback(page));
3090
3091 /* migrating an atomic written page is safe with the inmem_lock hold */
ff1048e7
JK		 3092 if (atomic_written) {
3093 if (mode != MIGRATE_SYNC)
3094 return -EBUSY;
3095 if (!mutex_trylock(&fi->inmem_lock))
3096 return -EAGAIN;
3097 }
5b7a487c 3098
240a5915
CY		 3099 /* one extra reference was held for atomic_write page */
3100 extra_count = atomic_written ? 1 : 0;
5b7a487c 3101 rc = migrate_page_move_mapping(mapping, newpage,
37109694 3102 page, extra_count);
5b7a487c
WG		 3103 if (rc != MIGRATEPAGE_SUCCESS) {
3104 if (atomic_written)
3105 mutex_unlock(&fi->inmem_lock);
3106 return rc;
3107 }
3108
3109 if (atomic_written) {
3110 struct inmem_pages *cur;
3111 list_for_each_entry(cur, &fi->inmem_pages, list)
3112 if (cur->page == page) {
3113 cur->page = newpage;
3114 break;
3115 }
3116 mutex_unlock(&fi->inmem_lock);
3117 put_page(page);
3118 get_page(newpage);
3119 }
3120
240a5915
CY		 3121 if (PagePrivate(page)) {
3122 f2fs_set_page_private(newpage, page_private(page));
3123 f2fs_clear_page_private(page);
3124 }
5b7a487c 3125
2916ecc0
JG		 3126 if (mode != MIGRATE_SYNC_NO_COPY)
3127 migrate_page_copy(newpage, page);
3128 else
3129 migrate_page_states(newpage, page);
5b7a487c
WG		 3130
3131 return MIGRATEPAGE_SUCCESS;
3132}
3133#endif
3134
4969c06a
JK		 3135#ifdef CONFIG_SWAP
3136/* Copied from generic_swapfile_activate() to check any holes */
3137static int check_swap_activate(struct file *swap_file, unsigned int max)
3138{
3139 struct address_space *mapping = swap_file->f_mapping;
3140 struct inode *inode = mapping->host;
3141 unsigned blocks_per_page;
3142 unsigned long page_no;
3143 unsigned blkbits;
3144 sector_t probe_block;
3145 sector_t last_block;
3146 sector_t lowest_block = -1;
3147 sector_t highest_block = 0;
3148
3149 blkbits = inode->i_blkbits;
3150 blocks_per_page = PAGE_SIZE >> blkbits;
3151
3152 /*
3153 * Map all the blocks into the extent list. This code doesn't try
3154 * to be very smart.
3155 */
3156 probe_block = 0;
3157 page_no = 0;
3158 last_block = i_size_read(inode) >> blkbits;
3159 while ((probe_block + blocks_per_page) <= last_block && page_no < max) {
3160 unsigned block_in_page;
3161 sector_t first_block;
3162
3163 cond_resched();
3164
3165 first_block = bmap(inode, probe_block);
3166 if (first_block == 0)
3167 goto bad_bmap;
3168
3169 /*
3170 * It must be PAGE_SIZE aligned on-disk
3171 */
3172 if (first_block & (blocks_per_page - 1)) {
3173 probe_block++;
3174 goto reprobe;
3175 }
3176
3177 for (block_in_page = 1; block_in_page < blocks_per_page;
3178 block_in_page++) {
3179 sector_t block;
3180
3181 block = bmap(inode, probe_block + block_in_page);
3182 if (block == 0)
3183 goto bad_bmap;
3184 if (block != first_block + block_in_page) {
3185 /* Discontiguity */
3186 probe_block++;
3187 goto reprobe;
3188 }
3189 }
3190
3191 first_block >>= (PAGE_SHIFT - blkbits);
3192 if (page_no) { /* exclude the header page */
3193 if (first_block < lowest_block)
3194 lowest_block = first_block;
3195 if (first_block > highest_block)
3196 highest_block = first_block;
3197 }
3198
3199 page_no++;
3200 probe_block += blocks_per_page;
3201reprobe:
3202 continue;
3203 }
3204 return 0;
3205
3206bad_bmap:
3207 pr_err("swapon: swapfile has holes\n");
3208 return -EINVAL;
3209}
3210
3211static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
3212 sector_t *span)
3213{
3214 struct inode *inode = file_inode(file);
3215 int ret;
3216
3217 if (!S_ISREG(inode->i_mode))
3218 return -EINVAL;
3219
3220 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3221 return -EROFS;
3222
3223 ret = f2fs_convert_inline_inode(inode);
3224 if (ret)
3225 return ret;
3226
3227 ret = check_swap_activate(file, sis->max);
3228 if (ret)
3229 return ret;
3230
3231 set_inode_flag(inode, FI_PIN_FILE);
3232 f2fs_precache_extents(inode);
3233 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3234 return 0;
3235}
3236
3237static void f2fs_swap_deactivate(struct file *file)
3238{
3239 struct inode *inode = file_inode(file);
3240
3241 clear_inode_flag(inode, FI_PIN_FILE);
3242}
3243#else
3244static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
3245 sector_t *span)
3246{
3247 return -EOPNOTSUPP;
3248}
3249
3250static void f2fs_swap_deactivate(struct file *file)
3251{
3252}
3253#endif
3254
eb47b800
JK		 3255const struct address_space_operations f2fs_dblock_aops = {
3256 .readpage = f2fs_read_data_page,
3257 .readpages = f2fs_read_data_pages,
3258 .writepage = f2fs_write_data_page,
3259 .writepages = f2fs_write_data_pages,
3260 .write_begin = f2fs_write_begin,
a1dd3c13 3261 .write_end = f2fs_write_end,
eb47b800 3262 .set_page_dirty = f2fs_set_data_page_dirty,
487261f3
CY		 3263 .invalidatepage = f2fs_invalidate_page,
3264 .releasepage = f2fs_release_page,
eb47b800 3265 .direct_IO = f2fs_direct_IO,
c01e54b7 3266 .bmap = f2fs_bmap,
4969c06a
JK		 3267 .swap_activate = f2fs_swap_activate,
3268 .swap_deactivate = f2fs_swap_deactivate,
5b7a487c
WG		 3269#ifdef CONFIG_MIGRATION
3270 .migratepage = f2fs_migrate_page,
3271#endif
eb47b800 3272};
6dbb1796 3273
5ec2d99d 3274void f2fs_clear_page_cache_dirty_tag(struct page *page)
aec2f729
CY		 3275{
3276 struct address_space *mapping = page_mapping(page);
3277 unsigned long flags;
3278
3279 xa_lock_irqsave(&mapping->i_pages, flags);
5ec2d99d 3280 __xa_clear_mark(&mapping->i_pages, page_index(page),
aec2f729
CY		 3281 PAGECACHE_TAG_DIRTY);
3282 xa_unlock_irqrestore(&mapping->i_pages, flags);
3283}
3284
6dbb1796
EB		 3285int __init f2fs_init_post_read_processing(void)
3286{
95ae251f
EB		 3287 bio_post_read_ctx_cache =
3288 kmem_cache_create("f2fs_bio_post_read_ctx",
3289 sizeof(struct bio_post_read_ctx), 0, 0, NULL);
6dbb1796
EB		 3290 if (!bio_post_read_ctx_cache)
3291 goto fail;
3292 bio_post_read_ctx_pool =
3293 mempool_create_slab_pool(NUM_PREALLOC_POST_READ_CTXS,
3294 bio_post_read_ctx_cache);
3295 if (!bio_post_read_ctx_pool)
3296 goto fail_free_cache;
3297 return 0;
3298
3299fail_free_cache:
3300 kmem_cache_destroy(bio_post_read_ctx_cache);
3301fail:
3302 return -ENOMEM;
3303}
3304
0b20fcec 3305void f2fs_destroy_post_read_processing(void)
6dbb1796
EB		 3306{
3307 mempool_destroy(bio_post_read_ctx_pool);
3308 kmem_cache_destroy(bio_post_read_ctx_cache);
3309}
0b20fcec
CY		 3310
3311int __init f2fs_init_bio_entry_cache(void)
3312{
3313 bio_entry_slab = f2fs_kmem_cache_create("bio_entry_slab",
3314 sizeof(struct bio_entry));
3315 if (!bio_entry_slab)
3316 return -ENOMEM;
3317 return 0;
3318}
3319
3320void __exit f2fs_destroy_bio_entry_cache(void)
3321{
3322 kmem_cache_destroy(bio_entry_slab);
3323}
Linux 6.x block layer and io_uring tree(s)