]> git.ipfire.org Git - people/ms/linux.git/blame - fs/f2fs/data.c
projects / people / ms / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: fix to update dir's i_pino during cross_rename
[people/ms/linux.git] / fs / f2fs / data.c
CommitLineData
7c1a000d 1// SPDX-License-Identifier: GPL-2.0
0a8165d7 2/*
eb47b800
JK		 3 * fs/f2fs/data.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
eb47b800
JK		 7 */
8#include <linux/fs.h>
9#include <linux/f2fs_fs.h>
10#include <linux/buffer_head.h>
11#include <linux/mpage.h>
12#include <linux/writeback.h>
13#include <linux/backing-dev.h>
8f46dcae 14#include <linux/pagevec.h>
eb47b800
JK		 15#include <linux/blkdev.h>
16#include <linux/bio.h>
4969c06a 17#include <linux/swap.h>
690e4a3e 18#include <linux/prefetch.h>
e2e40f2c 19#include <linux/uio.h>
f1e88660 20#include <linux/cleancache.h>
174cd4b1 21#include <linux/sched/signal.h>
eb47b800
JK		 22
23#include "f2fs.h"
24#include "node.h"
25#include "segment.h"
db9f7c1a 26#include "trace.h"
848753aa 27#include <trace/events/f2fs.h>
eb47b800 28
6dbb1796
EB		 29#define NUM_PREALLOC_POST_READ_CTXS 128
30
31static struct kmem_cache *bio_post_read_ctx_cache;
0b20fcec 32static struct kmem_cache *bio_entry_slab;
6dbb1796
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{
6f5c2ed0
CY		 171 if (time_to_inject(F2FS_P_SB(bio_first_page_all(bio)),
172 FAULT_READ_IO)) {
173 f2fs_show_injection_info(FAULT_READ_IO);
4e4cbee9 174 bio->bi_status = BLK_STS_IOERR;
55523519 175 }
8b038c70 176
6dbb1796
EB		 177 if (f2fs_bio_post_read_required(bio)) {
178 struct bio_post_read_ctx *ctx = bio->bi_private;
f1e88660 179
6dbb1796
EB		 180 ctx->cur_step = STEP_INITIAL;
181 bio_post_read_processing(ctx);
182 return;
f1e88660 183 }
6dbb1796
EB		 184
185 __read_end_io(bio);
f1e88660
JK		 186}
187
4246a0b6 188static void f2fs_write_end_io(struct bio *bio)
93dfe2ac 189{
1b1f559f 190 struct f2fs_sb_info *sbi = bio->bi_private;
f568849e 191 struct bio_vec *bvec;
6dc4f100 192 struct bvec_iter_all iter_all;
93dfe2ac 193
6f5c2ed0
CY		 194 if (time_to_inject(sbi, FAULT_WRITE_IO)) {
195 f2fs_show_injection_info(FAULT_WRITE_IO);
196 bio->bi_status = BLK_STS_IOERR;
197 }
198
2b070cfe 199 bio_for_each_segment_all(bvec, bio, iter_all) {
93dfe2ac 200 struct page *page = bvec->bv_page;
36951b38 201 enum count_type type = WB_DATA_TYPE(page);
93dfe2ac 202
0a595eba
JK		 203 if (IS_DUMMY_WRITTEN_PAGE(page)) {
204 set_page_private(page, (unsigned long)NULL);
205 ClearPagePrivate(page);
206 unlock_page(page);
207 mempool_free(page, sbi->write_io_dummy);
208
4e4cbee9 209 if (unlikely(bio->bi_status))
0a595eba
JK		 210 f2fs_stop_checkpoint(sbi, true);
211 continue;
212 }
213
d2d0727b 214 fscrypt_finalize_bounce_page(&page);
4375a336 215
4e4cbee9 216 if (unlikely(bio->bi_status)) {
5114a97a 217 mapping_set_error(page->mapping, -EIO);
b1ca321d
JK		 218 if (type == F2FS_WB_CP_DATA)
219 f2fs_stop_checkpoint(sbi, true);
93dfe2ac 220 }
7dff55d2
YH		 221
222 f2fs_bug_on(sbi, page->mapping == NODE_MAPPING(sbi) &&
223 page->index != nid_of_node(page));
224
36951b38 225 dec_page_count(sbi, type);
50fa53ec
CY		 226 if (f2fs_in_warm_node_list(sbi, page))
227 f2fs_del_fsync_node_entry(sbi, page);
36951b38 228 clear_cold_data(page);
93dfe2ac 229 end_page_writeback(page);
f568849e 230 }
36951b38 231 if (!get_pages(sbi, F2FS_WB_CP_DATA) &&
f5730184 232 wq_has_sleeper(&sbi->cp_wait))
93dfe2ac
JK		 233 wake_up(&sbi->cp_wait);
234
235 bio_put(bio);
236}
237
3c62be17
JK		 238/*
239 * Return true, if pre_bio's bdev is same as its target device.
240 */
241struct block_device *f2fs_target_device(struct f2fs_sb_info *sbi,
242 block_t blk_addr, struct bio *bio)
243{
244 struct block_device *bdev = sbi->sb->s_bdev;
245 int i;
246
0916878d
DLM		 247 if (f2fs_is_multi_device(sbi)) {
248 for (i = 0; i < sbi->s_ndevs; i++) {
249 if (FDEV(i).start_blk <= blk_addr &&
250 FDEV(i).end_blk >= blk_addr) {
251 blk_addr -= FDEV(i).start_blk;
252 bdev = FDEV(i).bdev;
253 break;
254 }
3c62be17
JK		 255 }
256 }
257 if (bio) {
74d46992 258 bio_set_dev(bio, bdev);
3c62be17
JK		 259 bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr);
260 }
261 return bdev;
262}
263
264int f2fs_target_device_index(struct f2fs_sb_info *sbi, block_t blkaddr)
265{
266 int i;
267
0916878d
DLM		 268 if (!f2fs_is_multi_device(sbi))
269 return 0;
270
3c62be17
JK		 271 for (i = 0; i < sbi->s_ndevs; i++)
272 if (FDEV(i).start_blk <= blkaddr && FDEV(i).end_blk >= blkaddr)
273 return i;
274 return 0;
275}
276
277static bool __same_bdev(struct f2fs_sb_info *sbi,
278 block_t blk_addr, struct bio *bio)
279{
74d46992
CH		 280 struct block_device *b = f2fs_target_device(sbi, blk_addr, NULL);
281 return bio->bi_disk == b->bd_disk && bio->bi_partno == b->bd_partno;
3c62be17
JK		 282}
283
940a6d34
GZ		 284/*
285 * Low-level block read/write IO operations.
286 */
b757f6ed 287static struct bio *__bio_alloc(struct f2fs_io_info *fio, int npages)
940a6d34 288{
b757f6ed 289 struct f2fs_sb_info *sbi = fio->sbi;
940a6d34
GZ		 290 struct bio *bio;
291
d62fe971 292 bio = f2fs_bio_alloc(sbi, npages, true);
940a6d34 293
b757f6ed
CY		 294 f2fs_target_device(sbi, fio->new_blkaddr, bio);
295 if (is_read_io(fio->op)) {
0cdd3195
HL		 296 bio->bi_end_io = f2fs_read_end_io;
297 bio->bi_private = NULL;
298 } else {
299 bio->bi_end_io = f2fs_write_end_io;
300 bio->bi_private = sbi;
b757f6ed
CY		 301 bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi,
302 fio->type, fio->temp);
0cdd3195 303 }
b757f6ed
CY		 304 if (fio->io_wbc)
305 wbc_init_bio(fio->io_wbc, bio);
940a6d34
GZ		 306
307 return bio;
308}
309
4fc29c1a
LT		 310static inline void __submit_bio(struct f2fs_sb_info *sbi,
311 struct bio *bio, enum page_type type)
f5730184 312{
4fc29c1a 313 if (!is_read_io(bio_op(bio))) {
0a595eba
JK		 314 unsigned int start;
315
0a595eba
JK		 316 if (type != DATA && type != NODE)
317 goto submit_io;
318
66415cee 319 if (test_opt(sbi, LFS) && current->plug)
3bb09a0e
TY		 320 blk_finish_plug(current->plug);
321
8223ecc4
CY		 322 if (F2FS_IO_ALIGNED(sbi))
323 goto submit_io;
324
0a595eba
JK		 325 start = bio->bi_iter.bi_size >> F2FS_BLKSIZE_BITS;
326 start %= F2FS_IO_SIZE(sbi);
327
328 if (start == 0)
329 goto submit_io;
330
331 /* fill dummy pages */
332 for (; start < F2FS_IO_SIZE(sbi); start++) {
333 struct page *page =
334 mempool_alloc(sbi->write_io_dummy,
bc73a4b2 335 GFP_NOIO | __GFP_NOFAIL);
0a595eba
JK		 336 f2fs_bug_on(sbi, !page);
337
bc73a4b2 338 zero_user_segment(page, 0, PAGE_SIZE);
0a595eba
JK		 339 SetPagePrivate(page);
340 set_page_private(page, (unsigned long)DUMMY_WRITTEN_PAGE);
341 lock_page(page);
342 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE)
343 f2fs_bug_on(sbi, 1);
344 }
345 /*
346 * In the NODE case, we lose next block address chain. So, we
347 * need to do checkpoint in f2fs_sync_file.
348 */
349 if (type == NODE)
350 set_sbi_flag(sbi, SBI_NEED_CP);
19a5f5e2 351 }
0a595eba 352submit_io:
554b5125
JK		 353 if (is_read_io(bio_op(bio)))
354 trace_f2fs_submit_read_bio(sbi->sb, type, bio);
355 else
356 trace_f2fs_submit_write_bio(sbi->sb, type, bio);
4e49ea4a 357 submit_bio(bio);
f5730184
JK		 358}
359
458e6197 360static void __submit_merged_bio(struct f2fs_bio_info *io)
93dfe2ac 361{
458e6197 362 struct f2fs_io_info *fio = &io->fio;
93dfe2ac
JK		 363
364 if (!io->bio)
365 return;
366
554b5125
JK		 367 bio_set_op_attrs(io->bio, fio->op, fio->op_flags);
368
04d328de 369 if (is_read_io(fio->op))
554b5125 370 trace_f2fs_prepare_read_bio(io->sbi->sb, fio->type, io->bio);
6a8f8ca5 371 else
554b5125 372 trace_f2fs_prepare_write_bio(io->sbi->sb, fio->type, io->bio);
04d328de 373
4fc29c1a 374 __submit_bio(io->sbi, io->bio, fio->type);
93dfe2ac
JK		 375 io->bio = NULL;
376}
377
8648de2c 378static bool __has_merged_page(struct bio *bio, struct inode *inode,
bab475c5 379 struct page *page, nid_t ino)
0fd785eb 380{
0fd785eb
CY		 381 struct bio_vec *bvec;
382 struct page *target;
6dc4f100 383 struct bvec_iter_all iter_all;
0fd785eb 384
8648de2c 385 if (!bio)
0fd785eb 386 return false;
0c3a5797 387
bab475c5 388 if (!inode && !page && !ino)
0c3a5797 389 return true;
0fd785eb 390
8648de2c 391 bio_for_each_segment_all(bvec, bio, iter_all) {
0fd785eb 392
d2d0727b
EB		 393 target = bvec->bv_page;
394 if (fscrypt_is_bounce_page(target))
395 target = fscrypt_pagecache_page(target);
0fd785eb 396
0c3a5797
CY		 397 if (inode && inode == target->mapping->host)
398 return true;
bab475c5
CY		 399 if (page && page == target)
400 return true;
0c3a5797 401 if (ino && ino == ino_of_node(target))
0fd785eb 402 return true;
0fd785eb
CY		 403 }
404
0fd785eb
CY		 405 return false;
406}
407
b9109b0e 408static void __f2fs_submit_merged_write(struct f2fs_sb_info *sbi,
a912b54d 409 enum page_type type, enum temp_type temp)
93dfe2ac
JK		 410{
411 enum page_type btype = PAGE_TYPE_OF_BIO(type);
a912b54d 412 struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
93dfe2ac 413
df0f8dc0 414 down_write(&io->io_rwsem);
458e6197
JK		 415
416 /* change META to META_FLUSH in the checkpoint procedure */
417 if (type >= META_FLUSH) {
418 io->fio.type = META_FLUSH;
04d328de 419 io->fio.op = REQ_OP_WRITE;
3adc5fcb 420 io->fio.op_flags = REQ_META | REQ_PRIO | REQ_SYNC;
70fd7614 421 if (!test_opt(sbi, NOBARRIER))
7f54f51f 422 io->fio.op_flags |= REQ_PREFLUSH | REQ_FUA;
458e6197
JK		 423 }
424 __submit_merged_bio(io);
df0f8dc0 425 up_write(&io->io_rwsem);
93dfe2ac
JK		 426}
427
a912b54d 428static void __submit_merged_write_cond(struct f2fs_sb_info *sbi,
bab475c5
CY		 429 struct inode *inode, struct page *page,
430 nid_t ino, enum page_type type, bool force)
0c3a5797 431{
a912b54d 432 enum temp_type temp;
1e771e83 433 bool ret = true;
a912b54d
JK		 434
435 for (temp = HOT; temp < NR_TEMP_TYPE; temp++) {
1e771e83
YS		 436 if (!force) {
437 enum page_type btype = PAGE_TYPE_OF_BIO(type);
438 struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
a912b54d 439
1e771e83 440 down_read(&io->io_rwsem);
8648de2c 441 ret = __has_merged_page(io->bio, inode, page, ino);
1e771e83
YS		 442 up_read(&io->io_rwsem);
443 }
444 if (ret)
445 __f2fs_submit_merged_write(sbi, type, temp);
a912b54d
JK		 446
447 /* TODO: use HOT temp only for meta pages now. */
448 if (type >= META)
449 break;
450 }
0c3a5797
CY		 451}
452
b9109b0e 453void f2fs_submit_merged_write(struct f2fs_sb_info *sbi, enum page_type type)
0c3a5797 454{
adcc00f7 455 __submit_merged_write_cond(sbi, NULL, NULL, 0, type, true);
0c3a5797
CY		 456}
457
b9109b0e 458void f2fs_submit_merged_write_cond(struct f2fs_sb_info *sbi,
bab475c5
CY		 459 struct inode *inode, struct page *page,
460 nid_t ino, enum page_type type)
0c3a5797 461{
bab475c5 462 __submit_merged_write_cond(sbi, inode, page, ino, type, false);
0c3a5797
CY		 463}
464
b9109b0e 465void f2fs_flush_merged_writes(struct f2fs_sb_info *sbi)
406657dd 466{
b9109b0e
JK		 467 f2fs_submit_merged_write(sbi, DATA);
468 f2fs_submit_merged_write(sbi, NODE);
469 f2fs_submit_merged_write(sbi, META);
406657dd
CY		 470}
471
93dfe2ac
JK		 472/*
473 * Fill the locked page with data located in the block address.
771a9a71 474 * A caller needs to unlock the page on failure.
93dfe2ac 475 */
05ca3632 476int f2fs_submit_page_bio(struct f2fs_io_info *fio)
93dfe2ac 477{
93dfe2ac 478 struct bio *bio;
0b81d077
JK		 479 struct page *page = fio->encrypted_page ?
480 fio->encrypted_page : fio->page;
93dfe2ac 481
c9b60788 482 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
93770ab7
CY		 483 fio->is_por ? META_POR : (__is_meta_io(fio) ?
484 META_GENERIC : DATA_GENERIC_ENHANCE)))
10f966bb 485 return -EFSCORRUPTED;
c9b60788 486
2ace38e0 487 trace_f2fs_submit_page_bio(page, fio);
05ca3632 488 f2fs_trace_ios(fio, 0);
93dfe2ac
JK		 489
490 /* Allocate a new bio */
b757f6ed 491 bio = __bio_alloc(fio, 1);
93dfe2ac 492
09cbfeaf 493 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
93dfe2ac 494 bio_put(bio);
93dfe2ac
JK		 495 return -EFAULT;
496 }
78efac53
CY		 497
498 if (fio->io_wbc && !is_read_io(fio->op))
34e51a5e 499 wbc_account_cgroup_owner(fio->io_wbc, page, PAGE_SIZE);
78efac53 500
04d328de 501 bio_set_op_attrs(bio, fio->op, fio->op_flags);
93dfe2ac 502
5f9abab4
JK		 503 inc_page_count(fio->sbi, is_read_io(fio->op) ?
504 __read_io_type(page): WB_DATA_TYPE(fio->page));
4c58ed07
CY		 505
506 __submit_bio(fio->sbi, bio, fio->type);
93dfe2ac
JK		 507 return 0;
508}
509
8896cbdf
CY		 510static bool page_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
511 block_t last_blkaddr, block_t cur_blkaddr)
512{
513 if (last_blkaddr + 1 != cur_blkaddr)
514 return false;
515 return __same_bdev(sbi, cur_blkaddr, bio);
516}
517
518static bool io_type_is_mergeable(struct f2fs_bio_info *io,
519 struct f2fs_io_info *fio)
520{
521 if (io->fio.op != fio->op)
522 return false;
523 return io->fio.op_flags == fio->op_flags;
524}
525
526static bool io_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
527 struct f2fs_bio_info *io,
528 struct f2fs_io_info *fio,
529 block_t last_blkaddr,
530 block_t cur_blkaddr)
531{
c72db71e
CY		 532 if (F2FS_IO_ALIGNED(sbi) && (fio->type == DATA || fio->type == NODE)) {
533 unsigned int filled_blocks =
534 F2FS_BYTES_TO_BLK(bio->bi_iter.bi_size);
535 unsigned int io_size = F2FS_IO_SIZE(sbi);
536 unsigned int left_vecs = bio->bi_max_vecs - bio->bi_vcnt;
537
538 /* IOs in bio is aligned and left space of vectors is not enough */
539 if (!(filled_blocks % io_size) && left_vecs < io_size)
540 return false;
541 }
8896cbdf
CY		 542 if (!page_is_mergeable(sbi, bio, last_blkaddr, cur_blkaddr))
543 return false;
544 return io_type_is_mergeable(io, fio);
545}
546
0b20fcec
CY		 547static void add_bio_entry(struct f2fs_sb_info *sbi, struct bio *bio,
548 struct page *page, enum temp_type temp)
549{
550 struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
551 struct bio_entry *be;
552
553 be = f2fs_kmem_cache_alloc(bio_entry_slab, GFP_NOFS);
554 be->bio = bio;
555 bio_get(bio);
556
557 if (bio_add_page(bio, page, PAGE_SIZE, 0) != PAGE_SIZE)
558 f2fs_bug_on(sbi, 1);
559
560 down_write(&io->bio_list_lock);
561 list_add_tail(&be->list, &io->bio_list);
562 up_write(&io->bio_list_lock);
563}
564
565static void del_bio_entry(struct bio_entry *be)
566{
567 list_del(&be->list);
568 kmem_cache_free(bio_entry_slab, be);
569}
570
571static int add_ipu_page(struct f2fs_sb_info *sbi, struct bio **bio,
572 struct page *page)
573{
574 enum temp_type temp;
575 bool found = false;
576 int ret = -EAGAIN;
577
578 for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
579 struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
580 struct list_head *head = &io->bio_list;
581 struct bio_entry *be;
582
583 down_write(&io->bio_list_lock);
584 list_for_each_entry(be, head, list) {
585 if (be->bio != *bio)
586 continue;
587
588 found = true;
589
590 if (bio_add_page(*bio, page, PAGE_SIZE, 0) == PAGE_SIZE) {
591 ret = 0;
592 break;
593 }
594
595 /* bio is full */
596 del_bio_entry(be);
597 __submit_bio(sbi, *bio, DATA);
598 break;
599 }
600 up_write(&io->bio_list_lock);
601 }
602
603 if (ret) {
604 bio_put(*bio);
605 *bio = NULL;
606 }
607
608 return ret;
609}
610
611void f2fs_submit_merged_ipu_write(struct f2fs_sb_info *sbi,
612 struct bio **bio, struct page *page)
613{
614 enum temp_type temp;
615 bool found = false;
616 struct bio *target = bio ? *bio : NULL;
617
618 for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
619 struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
620 struct list_head *head = &io->bio_list;
621 struct bio_entry *be;
622
623 if (list_empty(head))
624 continue;
625
626 down_read(&io->bio_list_lock);
627 list_for_each_entry(be, head, list) {
628 if (target)
629 found = (target == be->bio);
630 else
631 found = __has_merged_page(be->bio, NULL,
632 page, 0);
633 if (found)
634 break;
635 }
636 up_read(&io->bio_list_lock);
637
638 if (!found)
639 continue;
640
641 found = false;
642
643 down_write(&io->bio_list_lock);
644 list_for_each_entry(be, head, list) {
645 if (target)
646 found = (target == be->bio);
647 else
648 found = __has_merged_page(be->bio, NULL,
649 page, 0);
650 if (found) {
651 target = be->bio;
652 del_bio_entry(be);
653 break;
654 }
655 }
656 up_write(&io->bio_list_lock);
657 }
658
659 if (found)
660 __submit_bio(sbi, target, DATA);
661 if (bio && *bio) {
662 bio_put(*bio);
663 *bio = NULL;
664 }
665}
666
8648de2c
CY		 667int f2fs_merge_page_bio(struct f2fs_io_info *fio)
668{
669 struct bio *bio = *fio->bio;
670 struct page *page = fio->encrypted_page ?
671 fio->encrypted_page : fio->page;
672
673 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
674 __is_meta_io(fio) ? META_GENERIC : DATA_GENERIC))
10f966bb 675 return -EFSCORRUPTED;
8648de2c
CY		 676
677 trace_f2fs_submit_page_bio(page, fio);
678 f2fs_trace_ios(fio, 0);
679
8896cbdf 680 if (bio && !page_is_mergeable(fio->sbi, bio, *fio->last_block,
0b20fcec
CY		 681 fio->new_blkaddr))
682 f2fs_submit_merged_ipu_write(fio->sbi, &bio, NULL);
8648de2c
CY		 683alloc_new:
684 if (!bio) {
b757f6ed 685 bio = __bio_alloc(fio, BIO_MAX_PAGES);
8648de2c 686 bio_set_op_attrs(bio, fio->op, fio->op_flags);
8648de2c 687
0b20fcec
CY		 688 add_bio_entry(fio->sbi, bio, page, fio->temp);
689 } else {
690 if (add_ipu_page(fio->sbi, &bio, page))
691 goto alloc_new;
8648de2c
CY		 692 }
693
694 if (fio->io_wbc)
9637d517 695 wbc_account_cgroup_owner(fio->io_wbc, page, PAGE_SIZE);
8648de2c
CY		 696
697 inc_page_count(fio->sbi, WB_DATA_TYPE(page));
698
699 *fio->last_block = fio->new_blkaddr;
700 *fio->bio = bio;
701
702 return 0;
703}
704
fe16efe6 705void f2fs_submit_page_write(struct f2fs_io_info *fio)
93dfe2ac 706{
05ca3632 707 struct f2fs_sb_info *sbi = fio->sbi;
458e6197 708 enum page_type btype = PAGE_TYPE_OF_BIO(fio->type);
a912b54d 709 struct f2fs_bio_info *io = sbi->write_io[btype] + fio->temp;
4375a336 710 struct page *bio_page;
93dfe2ac 711
b9109b0e 712 f2fs_bug_on(sbi, is_read_io(fio->op));
93dfe2ac 713
fb830fc5
CY		 714 down_write(&io->io_rwsem);
715next:
716 if (fio->in_list) {
717 spin_lock(&io->io_lock);
718 if (list_empty(&io->io_list)) {
719 spin_unlock(&io->io_lock);
fe16efe6 720 goto out;
fb830fc5
CY		 721 }
722 fio = list_first_entry(&io->io_list,
723 struct f2fs_io_info, list);
724 list_del(&fio->list);
725 spin_unlock(&io->io_lock);
726 }
93dfe2ac 727
93770ab7 728 verify_fio_blkaddr(fio);
93dfe2ac 729
36951b38
CY		 730 bio_page = fio->encrypted_page ? fio->encrypted_page : fio->page;
731
ebf7c522
TM		 732 /* set submitted = true as a return value */
733 fio->submitted = true;
d68f735b 734
b9109b0e 735 inc_page_count(sbi, WB_DATA_TYPE(bio_page));
93dfe2ac 736
8896cbdf
CY		 737 if (io->bio && !io_is_mergeable(sbi, io->bio, io, fio,
738 io->last_block_in_bio, fio->new_blkaddr))
458e6197 739 __submit_merged_bio(io);
93dfe2ac
JK		 740alloc_new:
741 if (io->bio == NULL) {
8223ecc4
CY		 742 if (F2FS_IO_ALIGNED(sbi) &&
743 (fio->type == DATA || fio->type == NODE) &&
0a595eba 744 fio->new_blkaddr & F2FS_IO_SIZE_MASK(sbi)) {
b9109b0e 745 dec_page_count(sbi, WB_DATA_TYPE(bio_page));
fe16efe6
CY		 746 fio->retry = true;
747 goto skip;
0a595eba 748 }
b757f6ed 749 io->bio = __bio_alloc(fio, BIO_MAX_PAGES);
458e6197 750 io->fio = *fio;
93dfe2ac
JK		 751 }
752
a912b54d 753 if (bio_add_page(io->bio, bio_page, PAGE_SIZE, 0) < PAGE_SIZE) {
458e6197 754 __submit_merged_bio(io);
93dfe2ac
JK		 755 goto alloc_new;
756 }
757
578c6478 758 if (fio->io_wbc)
34e51a5e 759 wbc_account_cgroup_owner(fio->io_wbc, bio_page, PAGE_SIZE);
578c6478 760
7a9d7548 761 io->last_block_in_bio = fio->new_blkaddr;
05ca3632 762 f2fs_trace_ios(fio, 0);
fb830fc5
CY		 763
764 trace_f2fs_submit_page_write(fio->page, fio);
fe16efe6 765skip:
fb830fc5
CY		 766 if (fio->in_list)
767 goto next;
fe16efe6 768out:
4354994f 769 if (is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN) ||
00e09c0b 770 !f2fs_is_checkpoint_ready(sbi))
5ce80586 771 __submit_merged_bio(io);
df0f8dc0 772 up_write(&io->io_rwsem);
93dfe2ac
JK		 773}
774
95ae251f
EB		 775static inline bool f2fs_need_verity(const struct inode *inode, pgoff_t idx)
776{
777 return fsverity_active(inode) &&
778 idx < DIV_ROUND_UP(inode->i_size, PAGE_SIZE);
779}
780
13ba41e3 781static struct bio *f2fs_grab_read_bio(struct inode *inode, block_t blkaddr,
95ae251f
EB		 782 unsigned nr_pages, unsigned op_flag,
783 pgoff_t first_idx)
13ba41e3
JK		 784{
785 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
13ba41e3 786 struct bio *bio;
6dbb1796
EB		 787 struct bio_post_read_ctx *ctx;
788 unsigned int post_read_steps = 0;
13ba41e3 789
d62fe971 790 bio = f2fs_bio_alloc(sbi, min_t(int, nr_pages, BIO_MAX_PAGES), false);
6dbb1796 791 if (!bio)
13ba41e3 792 return ERR_PTR(-ENOMEM);
13ba41e3
JK		 793 f2fs_target_device(sbi, blkaddr, bio);
794 bio->bi_end_io = f2fs_read_end_io;
e2e59414 795 bio_set_op_attrs(bio, REQ_OP_READ, op_flag);
13ba41e3 796
6dbb1796
EB		 797 if (f2fs_encrypted_file(inode))
798 post_read_steps |= 1 << STEP_DECRYPT;
95ae251f
EB		 799
800 if (f2fs_need_verity(inode, first_idx))
801 post_read_steps |= 1 << STEP_VERITY;
802
6dbb1796
EB		 803 if (post_read_steps) {
804 ctx = mempool_alloc(bio_post_read_ctx_pool, GFP_NOFS);
805 if (!ctx) {
806 bio_put(bio);
807 return ERR_PTR(-ENOMEM);
808 }
809 ctx->bio = bio;
810 ctx->enabled_steps = post_read_steps;
811 bio->bi_private = ctx;
6dbb1796
EB		 812 }
813
13ba41e3
JK		 814 return bio;
815}
816
817/* This can handle encryption stuffs */
818static int f2fs_submit_page_read(struct inode *inode, struct page *page,
819 block_t blkaddr)
820{
93770ab7
CY		 821 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
822 struct bio *bio;
13ba41e3 823
95ae251f 824 bio = f2fs_grab_read_bio(inode, blkaddr, 1, 0, page->index);
13ba41e3
JK		 825 if (IS_ERR(bio))
826 return PTR_ERR(bio);
827
0ded69f6
JK		 828 /* wait for GCed page writeback via META_MAPPING */
829 f2fs_wait_on_block_writeback(inode, blkaddr);
830
13ba41e3
JK		 831 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
832 bio_put(bio);
833 return -EFAULT;
834 }
fb7d70db 835 ClearPageError(page);
93770ab7
CY		 836 inc_page_count(sbi, F2FS_RD_DATA);
837 __submit_bio(sbi, bio, DATA);
13ba41e3
JK		 838 return 0;
839}
840
46008c6d
CY		 841static void __set_data_blkaddr(struct dnode_of_data *dn)
842{
843 struct f2fs_node *rn = F2FS_NODE(dn->node_page);
844 __le32 *addr_array;
7a2af766
CY		 845 int base = 0;
846
847 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
848 base = get_extra_isize(dn->inode);
46008c6d
CY		 849
850 /* Get physical address of data block */
851 addr_array = blkaddr_in_node(rn);
7a2af766 852 addr_array[base + dn->ofs_in_node] = cpu_to_le32(dn->data_blkaddr);
46008c6d
CY		 853}
854
0a8165d7 855/*
eb47b800
JK		 856 * Lock ordering for the change of data block address:
857 * ->data_page
858 * ->node_page
859 * update block addresses in the node page
860 */
4d57b86d 861void f2fs_set_data_blkaddr(struct dnode_of_data *dn)
eb47b800 862{
bae0ee7a 863 f2fs_wait_on_page_writeback(dn->node_page, NODE, true, true);
46008c6d
CY		 864 __set_data_blkaddr(dn);
865 if (set_page_dirty(dn->node_page))
12719ae1 866 dn->node_changed = true;
eb47b800
JK		 867}
868
f28b3434
CY		 869void f2fs_update_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr)
870{
871 dn->data_blkaddr = blkaddr;
4d57b86d 872 f2fs_set_data_blkaddr(dn);
f28b3434
CY		 873 f2fs_update_extent_cache(dn);
874}
875
46008c6d 876/* dn->ofs_in_node will be returned with up-to-date last block pointer */
4d57b86d 877int f2fs_reserve_new_blocks(struct dnode_of_data *dn, blkcnt_t count)
eb47b800 878{
4081363f 879 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
0abd675e 880 int err;
eb47b800 881
46008c6d
CY		 882 if (!count)
883 return 0;
884
91942321 885 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
eb47b800 886 return -EPERM;
0abd675e
CY		 887 if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
888 return err;
eb47b800 889
46008c6d
CY		 890 trace_f2fs_reserve_new_blocks(dn->inode, dn->nid,
891 dn->ofs_in_node, count);
892
bae0ee7a 893 f2fs_wait_on_page_writeback(dn->node_page, NODE, true, true);
46008c6d
CY		 894
895 for (; count > 0; dn->ofs_in_node++) {
7a2af766
CY		 896 block_t blkaddr = datablock_addr(dn->inode,
897 dn->node_page, dn->ofs_in_node);
46008c6d
CY		 898 if (blkaddr == NULL_ADDR) {
899 dn->data_blkaddr = NEW_ADDR;
900 __set_data_blkaddr(dn);
901 count--;
902 }
903 }
904
905 if (set_page_dirty(dn->node_page))
906 dn->node_changed = true;
eb47b800
JK		 907 return 0;
908}
909
46008c6d 910/* Should keep dn->ofs_in_node unchanged */
4d57b86d 911int f2fs_reserve_new_block(struct dnode_of_data *dn)
46008c6d
CY		 912{
913 unsigned int ofs_in_node = dn->ofs_in_node;
914 int ret;
915
4d57b86d 916 ret = f2fs_reserve_new_blocks(dn, 1);
46008c6d
CY		 917 dn->ofs_in_node = ofs_in_node;
918 return ret;
919}
920
b600965c
HL		 921int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
922{
923 bool need_put = dn->inode_page ? false : true;
924 int err;
925
4d57b86d 926 err = f2fs_get_dnode_of_data(dn, index, ALLOC_NODE);
b600965c
HL		 927 if (err)
928 return err;
a8865372 929
b600965c 930 if (dn->data_blkaddr == NULL_ADDR)
4d57b86d 931 err = f2fs_reserve_new_block(dn);
a8865372 932 if (err || need_put)
b600965c
HL		 933 f2fs_put_dnode(dn);
934 return err;
935}
936
759af1c9 937int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index)
eb47b800 938{
e15882b6 939 struct extent_info ei = {0,0,0};
759af1c9 940 struct inode *inode = dn->inode;
028a41e8 941
759af1c9
FL		 942 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
943 dn->data_blkaddr = ei.blk + index - ei.fofs;
944 return 0;
429511cd 945 }
028a41e8 946
759af1c9 947 return f2fs_reserve_block(dn, index);
eb47b800
JK		 948}
949
4d57b86d 950struct page *f2fs_get_read_data_page(struct inode *inode, pgoff_t index,
04d328de 951 int op_flags, bool for_write)
eb47b800 952{
eb47b800
JK		 953 struct address_space *mapping = inode->i_mapping;
954 struct dnode_of_data dn;
955 struct page *page;
e15882b6 956 struct extent_info ei = {0,0,0};
eb47b800 957 int err;
4375a336 958
a56c7c6f 959 page = f2fs_grab_cache_page(mapping, index, for_write);
650495de
JK		 960 if (!page)
961 return ERR_PTR(-ENOMEM);
962
cb3bc9ee
CY		 963 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
964 dn.data_blkaddr = ei.blk + index - ei.fofs;
93770ab7
CY		 965 if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), dn.data_blkaddr,
966 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 967 err = -EFSCORRUPTED;
93770ab7
CY		 968 goto put_err;
969 }
cb3bc9ee
CY		 970 goto got_it;
971 }
972
eb47b800 973 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 974 err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
86531d6b
JK		 975 if (err)
976 goto put_err;
eb47b800
JK		 977 f2fs_put_dnode(&dn);
978
6bacf52f 979 if (unlikely(dn.data_blkaddr == NULL_ADDR)) {
86531d6b
JK		 980 err = -ENOENT;
981 goto put_err;
650495de 982 }
93770ab7
CY		 983 if (dn.data_blkaddr != NEW_ADDR &&
984 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
985 dn.data_blkaddr,
986 DATA_GENERIC_ENHANCE)) {
10f966bb 987 err = -EFSCORRUPTED;
93770ab7
CY		 988 goto put_err;
989 }
cb3bc9ee 990got_it:
43f3eae1
JK		 991 if (PageUptodate(page)) {
992 unlock_page(page);
eb47b800 993 return page;
43f3eae1 994 }
eb47b800 995
d59ff4df
JK		 996 /*
997 * A new dentry page is allocated but not able to be written, since its
998 * new inode page couldn't be allocated due to -ENOSPC.
999 * In such the case, its blkaddr can be remained as NEW_ADDR.
4d57b86d
CY		 1000 * see, f2fs_add_link -> f2fs_get_new_data_page ->
1001 * f2fs_init_inode_metadata.
d59ff4df
JK		 1002 */
1003 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 1004 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 1005 if (!PageUptodate(page))
1006 SetPageUptodate(page);
43f3eae1 1007 unlock_page(page);
d59ff4df
JK		 1008 return page;
1009 }
eb47b800 1010
13ba41e3 1011 err = f2fs_submit_page_read(inode, page, dn.data_blkaddr);
393ff91f 1012 if (err)
86531d6b 1013 goto put_err;
43f3eae1 1014 return page;
86531d6b
JK		 1015
1016put_err:
1017 f2fs_put_page(page, 1);
1018 return ERR_PTR(err);
43f3eae1
JK		 1019}
1020
4d57b86d 1021struct page *f2fs_find_data_page(struct inode *inode, pgoff_t index)
43f3eae1
JK		 1022{
1023 struct address_space *mapping = inode->i_mapping;
1024 struct page *page;
1025
1026 page = find_get_page(mapping, index);
1027 if (page && PageUptodate(page))
1028 return page;
1029 f2fs_put_page(page, 0);
1030
4d57b86d 1031 page = f2fs_get_read_data_page(inode, index, 0, false);
43f3eae1
JK		 1032 if (IS_ERR(page))
1033 return page;
1034
1035 if (PageUptodate(page))
1036 return page;
1037
1038 wait_on_page_locked(page);
1039 if (unlikely(!PageUptodate(page))) {
1040 f2fs_put_page(page, 0);
1041 return ERR_PTR(-EIO);
1042 }
1043 return page;
1044}
1045
1046/*
1047 * If it tries to access a hole, return an error.
1048 * Because, the callers, functions in dir.c and GC, should be able to know
1049 * whether this page exists or not.
1050 */
4d57b86d 1051struct page *f2fs_get_lock_data_page(struct inode *inode, pgoff_t index,
a56c7c6f 1052 bool for_write)
43f3eae1
JK		 1053{
1054 struct address_space *mapping = inode->i_mapping;
1055 struct page *page;
1056repeat:
4d57b86d 1057 page = f2fs_get_read_data_page(inode, index, 0, for_write);
43f3eae1
JK		 1058 if (IS_ERR(page))
1059 return page;
393ff91f 1060
43f3eae1 1061 /* wait for read completion */
393ff91f 1062 lock_page(page);
6bacf52f 1063 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 1064 f2fs_put_page(page, 1);
1065 goto repeat;
eb47b800 1066 }
1563ac75
CY		 1067 if (unlikely(!PageUptodate(page))) {
1068 f2fs_put_page(page, 1);
1069 return ERR_PTR(-EIO);
1070 }
eb47b800
JK		 1071 return page;
1072}
1073
0a8165d7 1074/*
eb47b800
JK		 1075 * Caller ensures that this data page is never allocated.
1076 * A new zero-filled data page is allocated in the page cache.
39936837 1077 *
4f4124d0
CY		 1078 * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and
1079 * f2fs_unlock_op().
470f00e9
CY		 1080 * Note that, ipage is set only by make_empty_dir, and if any error occur,
1081 * ipage should be released by this function.
eb47b800 1082 */
4d57b86d 1083struct page *f2fs_get_new_data_page(struct inode *inode,
a8865372 1084 struct page *ipage, pgoff_t index, bool new_i_size)
eb47b800 1085{
eb47b800
JK		 1086 struct address_space *mapping = inode->i_mapping;
1087 struct page *page;
1088 struct dnode_of_data dn;
1089 int err;
7612118a 1090
a56c7c6f 1091 page = f2fs_grab_cache_page(mapping, index, true);
470f00e9
CY		 1092 if (!page) {
1093 /*
1094 * before exiting, we should make sure ipage will be released
1095 * if any error occur.
1096 */
1097 f2fs_put_page(ipage, 1);
01f28610 1098 return ERR_PTR(-ENOMEM);
470f00e9 1099 }
eb47b800 1100
a8865372 1101 set_new_dnode(&dn, inode, ipage, NULL, 0);
b600965c 1102 err = f2fs_reserve_block(&dn, index);
01f28610
JK		 1103 if (err) {
1104 f2fs_put_page(page, 1);
eb47b800 1105 return ERR_PTR(err);
a8865372 1106 }
01f28610
JK		 1107 if (!ipage)
1108 f2fs_put_dnode(&dn);
eb47b800
JK		 1109
1110 if (PageUptodate(page))
01f28610 1111 goto got_it;
eb47b800
JK		 1112
1113 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 1114 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 1115 if (!PageUptodate(page))
1116 SetPageUptodate(page);
eb47b800 1117 } else {
4375a336 1118 f2fs_put_page(page, 1);
a8865372 1119
7612118a
JK		 1120 /* if ipage exists, blkaddr should be NEW_ADDR */
1121 f2fs_bug_on(F2FS_I_SB(inode), ipage);
4d57b86d 1122 page = f2fs_get_lock_data_page(inode, index, true);
4375a336 1123 if (IS_ERR(page))
7612118a 1124 return page;
eb47b800 1125 }
01f28610 1126got_it:
9edcdabf 1127 if (new_i_size && i_size_read(inode) <
ee6d182f 1128 ((loff_t)(index + 1) << PAGE_SHIFT))
fc9581c8 1129 f2fs_i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT));
eb47b800
JK		 1130 return page;
1131}
1132
d5097be5 1133static int __allocate_data_block(struct dnode_of_data *dn, int seg_type)
bfad7c2d 1134{
4081363f 1135 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
bfad7c2d 1136 struct f2fs_summary sum;
bfad7c2d 1137 struct node_info ni;
6aa58d8a 1138 block_t old_blkaddr;
46008c6d 1139 blkcnt_t count = 1;
0abd675e 1140 int err;
bfad7c2d 1141
91942321 1142 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
bfad7c2d 1143 return -EPERM;
df6136ef 1144
7735730d
CY		 1145 err = f2fs_get_node_info(sbi, dn->nid, &ni);
1146 if (err)
1147 return err;
1148
7a2af766
CY		 1149 dn->data_blkaddr = datablock_addr(dn->inode,
1150 dn->node_page, dn->ofs_in_node);
f847c699 1151 if (dn->data_blkaddr != NULL_ADDR)
df6136ef
CY		 1152 goto alloc;
1153
0abd675e
CY		 1154 if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
1155 return err;
bfad7c2d 1156
df6136ef 1157alloc:
bfad7c2d 1158 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
6aa58d8a
CY		 1159 old_blkaddr = dn->data_blkaddr;
1160 f2fs_allocate_data_block(sbi, NULL, old_blkaddr, &dn->data_blkaddr,
d5097be5 1161 &sum, seg_type, NULL, false);
6aa58d8a
CY		 1162 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
1163 invalidate_mapping_pages(META_MAPPING(sbi),
1164 old_blkaddr, old_blkaddr);
86f35dc3 1165 f2fs_update_data_blkaddr(dn, dn->data_blkaddr);
bfad7c2d 1166
0a4daae5
JK		 1167 /*
1168 * i_size will be updated by direct_IO. Otherwise, we'll get stale
1169 * data from unwritten block via dio_read.
1170 */
bfad7c2d
JK		 1171 return 0;
1172}
1173
a7de6086 1174int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *from)
59b802e5 1175{
b439b103 1176 struct inode *inode = file_inode(iocb->ki_filp);
5b8db7fa 1177 struct f2fs_map_blocks map;
d6d478a1 1178 int flag;
a7de6086 1179 int err = 0;
d6d478a1 1180 bool direct_io = iocb->ki_flags & IOCB_DIRECT;
59b802e5 1181
71ad682c 1182 /* convert inline data for Direct I/O*/
d6d478a1 1183 if (direct_io) {
71ad682c
WG		 1184 err = f2fs_convert_inline_inode(inode);
1185 if (err)
1186 return err;
1187 }
1188
f9d6d059
CY		 1189 if (direct_io && allow_outplace_dio(inode, iocb, from))
1190 return 0;
1191
dc91de78
JK		 1192 if (is_inode_flag_set(inode, FI_NO_PREALLOC))
1193 return 0;
1194
0080c507 1195 map.m_lblk = F2FS_BLK_ALIGN(iocb->ki_pos);
dfd02e4d
CY		 1196 map.m_len = F2FS_BYTES_TO_BLK(iocb->ki_pos + iov_iter_count(from));
1197 if (map.m_len > map.m_lblk)
1198 map.m_len -= map.m_lblk;
1199 else
1200 map.m_len = 0;
1201
da85985c 1202 map.m_next_pgofs = NULL;
c4020b2d 1203 map.m_next_extent = NULL;
d5097be5 1204 map.m_seg_type = NO_CHECK_TYPE;
f9d6d059 1205 map.m_may_create = true;
2a340760 1206
d6d478a1 1207 if (direct_io) {
4d57b86d 1208 map.m_seg_type = f2fs_rw_hint_to_seg_type(iocb->ki_hint);
f847c699 1209 flag = f2fs_force_buffered_io(inode, iocb, from) ?
d6d478a1
CY		 1210 F2FS_GET_BLOCK_PRE_AIO :
1211 F2FS_GET_BLOCK_PRE_DIO;
1212 goto map_blocks;
d5097be5 1213 }
f2470371 1214 if (iocb->ki_pos + iov_iter_count(from) > MAX_INLINE_DATA(inode)) {
a7de6086
JK		 1215 err = f2fs_convert_inline_inode(inode);
1216 if (err)
1217 return err;
b439b103 1218 }
d6d478a1 1219 if (f2fs_has_inline_data(inode))
25006645 1220 return err;
d6d478a1
CY		 1221
1222 flag = F2FS_GET_BLOCK_PRE_AIO;
1223
1224map_blocks:
1225 err = f2fs_map_blocks(inode, &map, 1, flag);
1226 if (map.m_len > 0 && err == -ENOSPC) {
1227 if (!direct_io)
1228 set_inode_flag(inode, FI_NO_PREALLOC);
1229 err = 0;
25006645 1230 }
a7de6086 1231 return err;
59b802e5
JK		 1232}
1233
39a86958 1234void __do_map_lock(struct f2fs_sb_info *sbi, int flag, bool lock)
59c9081b
YH		 1235{
1236 if (flag == F2FS_GET_BLOCK_PRE_AIO) {
1237 if (lock)
1238 down_read(&sbi->node_change);
1239 else
1240 up_read(&sbi->node_change);
1241 } else {
1242 if (lock)
1243 f2fs_lock_op(sbi);
1244 else
1245 f2fs_unlock_op(sbi);
1246 }
1247}
1248
0a8165d7 1249/*
003a3e1d
JK		 1250 * f2fs_map_blocks() now supported readahead/bmap/rw direct_IO with
1251 * f2fs_map_blocks structure.
4f4124d0
CY		 1252 * If original data blocks are allocated, then give them to blockdev.
1253 * Otherwise,
1254 * a. preallocate requested block addresses
1255 * b. do not use extent cache for better performance
1256 * c. give the block addresses to blockdev
eb47b800 1257 */
d323d005 1258int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
e2b4e2bc 1259 int create, int flag)
eb47b800 1260{
003a3e1d 1261 unsigned int maxblocks = map->m_len;
eb47b800 1262 struct dnode_of_data dn;
f9811703 1263 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
f9d6d059 1264 int mode = map->m_may_create ? ALLOC_NODE : LOOKUP_NODE;
46008c6d 1265 pgoff_t pgofs, end_offset, end;
bfad7c2d 1266 int err = 0, ofs = 1;
46008c6d
CY		 1267 unsigned int ofs_in_node, last_ofs_in_node;
1268 blkcnt_t prealloc;
e15882b6 1269 struct extent_info ei = {0,0,0};
7df3a431 1270 block_t blkaddr;
c4020b2d 1271 unsigned int start_pgofs;
eb47b800 1272
dfd02e4d
CY		 1273 if (!maxblocks)
1274 return 0;
1275
003a3e1d
JK		 1276 map->m_len = 0;
1277 map->m_flags = 0;
1278
1279 /* it only supports block size == page size */
1280 pgofs = (pgoff_t)map->m_lblk;
46008c6d 1281 end = pgofs + maxblocks;
eb47b800 1282
24b84912 1283 if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) {
f4f0b677
JZ		 1284 if (test_opt(sbi, LFS) && flag == F2FS_GET_BLOCK_DIO &&
1285 map->m_may_create)
1286 goto next_dnode;
1287
003a3e1d
JK		 1288 map->m_pblk = ei.blk + pgofs - ei.fofs;
1289 map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs);
1290 map->m_flags = F2FS_MAP_MAPPED;
c4020b2d
CY		 1291 if (map->m_next_extent)
1292 *map->m_next_extent = pgofs + map->m_len;
1e78e8bd
ST		 1293
1294 /* for hardware encryption, but to avoid potential issue in future */
1295 if (flag == F2FS_GET_BLOCK_DIO)
1296 f2fs_wait_on_block_writeback_range(inode,
1297 map->m_pblk, map->m_len);
bfad7c2d 1298 goto out;
a2e7d1bf 1299 }
bfad7c2d 1300
4fe71e88 1301next_dnode:
f9d6d059 1302 if (map->m_may_create)
59c9081b 1303 __do_map_lock(sbi, flag, true);
eb47b800
JK		 1304
1305 /* When reading holes, we need its node page */
1306 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1307 err = f2fs_get_dnode_of_data(&dn, pgofs, mode);
1ec79083 1308 if (err) {
43473f96
CY		 1309 if (flag == F2FS_GET_BLOCK_BMAP)
1310 map->m_pblk = 0;
da85985c 1311 if (err == -ENOENT) {
bfad7c2d 1312 err = 0;
da85985c
CY		 1313 if (map->m_next_pgofs)
1314 *map->m_next_pgofs =
4d57b86d 1315 f2fs_get_next_page_offset(&dn, pgofs);
c4020b2d
CY		 1316 if (map->m_next_extent)
1317 *map->m_next_extent =
4d57b86d 1318 f2fs_get_next_page_offset(&dn, pgofs);
da85985c 1319 }
bfad7c2d 1320 goto unlock_out;
848753aa 1321 }
973163fc 1322
c4020b2d 1323 start_pgofs = pgofs;
46008c6d 1324 prealloc = 0;
230436b3 1325 last_ofs_in_node = ofs_in_node = dn.ofs_in_node;
81ca7350 1326 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
4fe71e88
CY		 1327
1328next_block:
7a2af766 1329 blkaddr = datablock_addr(dn.inode, dn.node_page, dn.ofs_in_node);
4fe71e88 1330
c9b60788 1331 if (__is_valid_data_blkaddr(blkaddr) &&
93770ab7 1332 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC_ENHANCE)) {
10f966bb 1333 err = -EFSCORRUPTED;
c9b60788
CY		 1334 goto sync_out;
1335 }
1336
93770ab7 1337 if (__is_valid_data_blkaddr(blkaddr)) {
f847c699 1338 /* use out-place-update for driect IO under LFS mode */
f9d6d059
CY		 1339 if (test_opt(sbi, LFS) && flag == F2FS_GET_BLOCK_DIO &&
1340 map->m_may_create) {
f847c699 1341 err = __allocate_data_block(&dn, map->m_seg_type);
05e36006
CY		 1342 if (err)
1343 goto sync_out;
1344 blkaddr = dn.data_blkaddr;
1345 set_inode_flag(inode, FI_APPEND_WRITE);
f847c699
CY		 1346 }
1347 } else {
973163fc 1348 if (create) {
f9811703
CY		 1349 if (unlikely(f2fs_cp_error(sbi))) {
1350 err = -EIO;
4fe71e88 1351 goto sync_out;
f9811703 1352 }
24b84912 1353 if (flag == F2FS_GET_BLOCK_PRE_AIO) {
46008c6d
CY		 1354 if (blkaddr == NULL_ADDR) {
1355 prealloc++;
1356 last_ofs_in_node = dn.ofs_in_node;
1357 }
24b84912 1358 } else {
0a4daae5
JK		 1359 WARN_ON(flag != F2FS_GET_BLOCK_PRE_DIO &&
1360 flag != F2FS_GET_BLOCK_DIO);
d5097be5
HL		 1361 err = __allocate_data_block(&dn,
1362 map->m_seg_type);
6f2d8ed6 1363 if (!err)
91942321 1364 set_inode_flag(inode, FI_APPEND_WRITE);
24b84912 1365 }
973163fc 1366 if (err)
4fe71e88 1367 goto sync_out;
3f2be043 1368 map->m_flags |= F2FS_MAP_NEW;
4fe71e88 1369 blkaddr = dn.data_blkaddr;
973163fc 1370 } else {
43473f96
CY		 1371 if (flag == F2FS_GET_BLOCK_BMAP) {
1372 map->m_pblk = 0;
1373 goto sync_out;
1374 }
c4020b2d
CY		 1375 if (flag == F2FS_GET_BLOCK_PRECACHE)
1376 goto sync_out;
da85985c
CY		 1377 if (flag == F2FS_GET_BLOCK_FIEMAP &&
1378 blkaddr == NULL_ADDR) {
1379 if (map->m_next_pgofs)
1380 *map->m_next_pgofs = pgofs + 1;
4c2ac6a8 1381 goto sync_out;
da85985c 1382 }
f3d98e74
CY		 1383 if (flag != F2FS_GET_BLOCK_FIEMAP) {
1384 /* for defragment case */
1385 if (map->m_next_pgofs)
1386 *map->m_next_pgofs = pgofs + 1;
4fe71e88 1387 goto sync_out;
f3d98e74 1388 }
e2b4e2bc 1389 }
e2b4e2bc 1390 }
eb47b800 1391
46008c6d
CY		 1392 if (flag == F2FS_GET_BLOCK_PRE_AIO)
1393 goto skip;
1394
4fe71e88
CY		 1395 if (map->m_len == 0) {
1396 /* preallocated unwritten block should be mapped for fiemap. */
1397 if (blkaddr == NEW_ADDR)
1398 map->m_flags |= F2FS_MAP_UNWRITTEN;
1399 map->m_flags |= F2FS_MAP_MAPPED;
1400
1401 map->m_pblk = blkaddr;
1402 map->m_len = 1;
1403 } else if ((map->m_pblk != NEW_ADDR &&
1404 blkaddr == (map->m_pblk + ofs)) ||
b439b103 1405 (map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR) ||
46008c6d 1406 flag == F2FS_GET_BLOCK_PRE_DIO) {
4fe71e88
CY		 1407 ofs++;
1408 map->m_len++;
1409 } else {
1410 goto sync_out;
1411 }
bfad7c2d 1412
46008c6d 1413skip:
bfad7c2d
JK		 1414 dn.ofs_in_node++;
1415 pgofs++;
1416
46008c6d
CY		 1417 /* preallocate blocks in batch for one dnode page */
1418 if (flag == F2FS_GET_BLOCK_PRE_AIO &&
1419 (pgofs == end || dn.ofs_in_node == end_offset)) {
7df3a431 1420
46008c6d 1421 dn.ofs_in_node = ofs_in_node;
4d57b86d 1422 err = f2fs_reserve_new_blocks(&dn, prealloc);
46008c6d
CY		 1423 if (err)
1424 goto sync_out;
bfad7c2d 1425
46008c6d
CY		 1426 map->m_len += dn.ofs_in_node - ofs_in_node;
1427 if (prealloc && dn.ofs_in_node != last_ofs_in_node + 1) {
1428 err = -ENOSPC;
1429 goto sync_out;
3104af35 1430 }
46008c6d
CY		 1431 dn.ofs_in_node = end_offset;
1432 }
1433
1434 if (pgofs >= end)
1435 goto sync_out;
1436 else if (dn.ofs_in_node < end_offset)
1437 goto next_block;
1438
c4020b2d
CY		 1439 if (flag == F2FS_GET_BLOCK_PRECACHE) {
1440 if (map->m_flags & F2FS_MAP_MAPPED) {
1441 unsigned int ofs = start_pgofs - map->m_lblk;
1442
1443 f2fs_update_extent_cache_range(&dn,
1444 start_pgofs, map->m_pblk + ofs,
1445 map->m_len - ofs);
1446 }
1447 }
1448
46008c6d
CY		 1449 f2fs_put_dnode(&dn);
1450
f9d6d059 1451 if (map->m_may_create) {
59c9081b 1452 __do_map_lock(sbi, flag, false);
6f2d8ed6 1453 f2fs_balance_fs(sbi, dn.node_changed);
eb47b800 1454 }
46008c6d 1455 goto next_dnode;
7df3a431 1456
bfad7c2d 1457sync_out:
1e78e8bd
ST		 1458
1459 /* for hardware encryption, but to avoid potential issue in future */
1460 if (flag == F2FS_GET_BLOCK_DIO && map->m_flags & F2FS_MAP_MAPPED)
1461 f2fs_wait_on_block_writeback_range(inode,
1462 map->m_pblk, map->m_len);
1463
c4020b2d
CY		 1464 if (flag == F2FS_GET_BLOCK_PRECACHE) {
1465 if (map->m_flags & F2FS_MAP_MAPPED) {
1466 unsigned int ofs = start_pgofs - map->m_lblk;
1467
1468 f2fs_update_extent_cache_range(&dn,
1469 start_pgofs, map->m_pblk + ofs,
1470 map->m_len - ofs);
1471 }
1472 if (map->m_next_extent)
1473 *map->m_next_extent = pgofs + 1;
1474 }
eb47b800 1475 f2fs_put_dnode(&dn);
bfad7c2d 1476unlock_out:
f9d6d059 1477 if (map->m_may_create) {
59c9081b 1478 __do_map_lock(sbi, flag, false);
6f2d8ed6 1479 f2fs_balance_fs(sbi, dn.node_changed);
2a340760 1480 }
bfad7c2d 1481out:
003a3e1d 1482 trace_f2fs_map_blocks(inode, map, err);
bfad7c2d 1483 return err;
eb47b800
JK		 1484}
1485
b91050a8
HL		 1486bool f2fs_overwrite_io(struct inode *inode, loff_t pos, size_t len)
1487{
1488 struct f2fs_map_blocks map;
1489 block_t last_lblk;
1490 int err;
1491
1492 if (pos + len > i_size_read(inode))
1493 return false;
1494
1495 map.m_lblk = F2FS_BYTES_TO_BLK(pos);
1496 map.m_next_pgofs = NULL;
1497 map.m_next_extent = NULL;
1498 map.m_seg_type = NO_CHECK_TYPE;
f4f0b677 1499 map.m_may_create = false;
b91050a8
HL		 1500 last_lblk = F2FS_BLK_ALIGN(pos + len);
1501
1502 while (map.m_lblk < last_lblk) {
1503 map.m_len = last_lblk - map.m_lblk;
1504 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
1505 if (err || map.m_len == 0)
1506 return false;
1507 map.m_lblk += map.m_len;
1508 }
1509 return true;
1510}
1511
003a3e1d 1512static int __get_data_block(struct inode *inode, sector_t iblock,
da85985c 1513 struct buffer_head *bh, int create, int flag,
f9d6d059 1514 pgoff_t *next_pgofs, int seg_type, bool may_write)
003a3e1d
JK		 1515{
1516 struct f2fs_map_blocks map;
a7de6086 1517 int err;
003a3e1d
JK		 1518
1519 map.m_lblk = iblock;
1520 map.m_len = bh->b_size >> inode->i_blkbits;
da85985c 1521 map.m_next_pgofs = next_pgofs;
c4020b2d 1522 map.m_next_extent = NULL;
d5097be5 1523 map.m_seg_type = seg_type;
f9d6d059 1524 map.m_may_create = may_write;
003a3e1d 1525
a7de6086
JK		 1526 err = f2fs_map_blocks(inode, &map, create, flag);
1527 if (!err) {
003a3e1d
JK		 1528 map_bh(bh, inode->i_sb, map.m_pblk);
1529 bh->b_state = (bh->b_state & ~F2FS_MAP_FLAGS) | map.m_flags;
b86e3307 1530 bh->b_size = (u64)map.m_len << inode->i_blkbits;
003a3e1d 1531 }
a7de6086 1532 return err;
003a3e1d
JK		 1533}
1534
ccfb3000 1535static int get_data_block(struct inode *inode, sector_t iblock,
da85985c
CY		 1536 struct buffer_head *bh_result, int create, int flag,
1537 pgoff_t *next_pgofs)
e2b4e2bc 1538{
da85985c 1539 return __get_data_block(inode, iblock, bh_result, create,
d5097be5 1540 flag, next_pgofs,
f9d6d059
CY		 1541 NO_CHECK_TYPE, create);
1542}
1543
1544static int get_data_block_dio_write(struct inode *inode, sector_t iblock,
1545 struct buffer_head *bh_result, int create)
1546{
1547 return __get_data_block(inode, iblock, bh_result, create,
1548 F2FS_GET_BLOCK_DIO, NULL,
1549 f2fs_rw_hint_to_seg_type(inode->i_write_hint),
75a037f3 1550 IS_SWAPFILE(inode) ? false : true);
e2b4e2bc
CY		 1551}
1552
1553static int get_data_block_dio(struct inode *inode, sector_t iblock,
ccfb3000
JK		 1554 struct buffer_head *bh_result, int create)
1555{
e2b4e2bc 1556 return __get_data_block(inode, iblock, bh_result, create,
f9d6d059
CY		 1557 F2FS_GET_BLOCK_DIO, NULL,
1558 f2fs_rw_hint_to_seg_type(inode->i_write_hint),
1559 false);
ccfb3000
JK		 1560}
1561
e2b4e2bc 1562static int get_data_block_bmap(struct inode *inode, sector_t iblock,
ccfb3000
JK		 1563 struct buffer_head *bh_result, int create)
1564{
179448bf 1565 /* Block number less than F2FS MAX BLOCKS */
e0afc4d6 1566 if (unlikely(iblock >= F2FS_I_SB(inode)->max_file_blocks))
179448bf
YH		 1567 return -EFBIG;
1568
e2b4e2bc 1569 return __get_data_block(inode, iblock, bh_result, create,
d5097be5 1570 F2FS_GET_BLOCK_BMAP, NULL,
f9d6d059 1571 NO_CHECK_TYPE, create);
ccfb3000
JK		 1572}
1573
7f63eb77
JK		 1574static inline sector_t logical_to_blk(struct inode *inode, loff_t offset)
1575{
1576 return (offset >> inode->i_blkbits);
1577}
1578
1579static inline loff_t blk_to_logical(struct inode *inode, sector_t blk)
1580{
1581 return (blk << inode->i_blkbits);
1582}
1583
442a9dbd
CY		 1584static int f2fs_xattr_fiemap(struct inode *inode,
1585 struct fiemap_extent_info *fieinfo)
1586{
1587 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1588 struct page *page;
1589 struct node_info ni;
1590 __u64 phys = 0, len;
1591 __u32 flags;
1592 nid_t xnid = F2FS_I(inode)->i_xattr_nid;
1593 int err = 0;
1594
1595 if (f2fs_has_inline_xattr(inode)) {
1596 int offset;
1597
1598 page = f2fs_grab_cache_page(NODE_MAPPING(sbi),
1599 inode->i_ino, false);
1600 if (!page)
1601 return -ENOMEM;
1602
7735730d
CY		 1603 err = f2fs_get_node_info(sbi, inode->i_ino, &ni);
1604 if (err) {
1605 f2fs_put_page(page, 1);
1606 return err;
1607 }
442a9dbd
CY		 1608
1609 phys = (__u64)blk_to_logical(inode, ni.blk_addr);
1610 offset = offsetof(struct f2fs_inode, i_addr) +
1611 sizeof(__le32) * (DEF_ADDRS_PER_INODE -
b323fd28 1612 get_inline_xattr_addrs(inode));
442a9dbd
CY		 1613
1614 phys += offset;
1615 len = inline_xattr_size(inode);
1616
1617 f2fs_put_page(page, 1);
1618
1619 flags = FIEMAP_EXTENT_DATA_INLINE | FIEMAP_EXTENT_NOT_ALIGNED;
1620
1621 if (!xnid)
1622 flags |= FIEMAP_EXTENT_LAST;
1623
1624 err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
1625 if (err || err == 1)
1626 return err;
1627 }
1628
1629 if (xnid) {
1630 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), xnid, false);
1631 if (!page)
1632 return -ENOMEM;
1633
7735730d
CY		 1634 err = f2fs_get_node_info(sbi, xnid, &ni);
1635 if (err) {
1636 f2fs_put_page(page, 1);
1637 return err;
1638 }
442a9dbd
CY		 1639
1640 phys = (__u64)blk_to_logical(inode, ni.blk_addr);
1641 len = inode->i_sb->s_blocksize;
1642
1643 f2fs_put_page(page, 1);
1644
1645 flags = FIEMAP_EXTENT_LAST;
1646 }
1647
1648 if (phys)
1649 err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
1650
1651 return (err < 0 ? err : 0);
1652}
1653
9ab70134
JK		 1654int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1655 u64 start, u64 len)
1656{
7f63eb77
JK		 1657 struct buffer_head map_bh;
1658 sector_t start_blk, last_blk;
da85985c 1659 pgoff_t next_pgofs;
7f63eb77
JK		 1660 u64 logical = 0, phys = 0, size = 0;
1661 u32 flags = 0;
7f63eb77
JK		 1662 int ret = 0;
1663
c4020b2d
CY		 1664 if (fieinfo->fi_flags & FIEMAP_FLAG_CACHE) {
1665 ret = f2fs_precache_extents(inode);
1666 if (ret)
1667 return ret;
1668 }
1669
442a9dbd 1670 ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC | FIEMAP_FLAG_XATTR);
7f63eb77
JK		 1671 if (ret)
1672 return ret;
1673
f1b43d4c
CY		 1674 inode_lock(inode);
1675
442a9dbd
CY		 1676 if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
1677 ret = f2fs_xattr_fiemap(inode, fieinfo);
1678 goto out;
1679 }
1680
7975f349 1681 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
67f8cf3c
JK		 1682 ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len);
1683 if (ret != -EAGAIN)
f1b43d4c 1684 goto out;
67f8cf3c
JK		 1685 }
1686
7f63eb77
JK		 1687 if (logical_to_blk(inode, len) == 0)
1688 len = blk_to_logical(inode, 1);
1689
1690 start_blk = logical_to_blk(inode, start);
1691 last_blk = logical_to_blk(inode, start + len - 1);
9a950d52 1692
7f63eb77
JK		 1693next:
1694 memset(&map_bh, 0, sizeof(struct buffer_head));
1695 map_bh.b_size = len;
1696
e2b4e2bc 1697 ret = get_data_block(inode, start_blk, &map_bh, 0,
da85985c 1698 F2FS_GET_BLOCK_FIEMAP, &next_pgofs);
7f63eb77
JK		 1699 if (ret)
1700 goto out;
1701
1702 /* HOLE */
1703 if (!buffer_mapped(&map_bh)) {
da85985c 1704 start_blk = next_pgofs;
58736fa6
CY		 1705
1706 if (blk_to_logical(inode, start_blk) < blk_to_logical(inode,
1707 F2FS_I_SB(inode)->max_file_blocks))
9a950d52 1708 goto prep_next;
58736fa6 1709
9a950d52
FL		 1710 flags |= FIEMAP_EXTENT_LAST;
1711 }
7f63eb77 1712
da5af127 1713 if (size) {
62230e0d 1714 if (IS_ENCRYPTED(inode))
da5af127
CY		 1715 flags |= FIEMAP_EXTENT_DATA_ENCRYPTED;
1716
9a950d52
FL		 1717 ret = fiemap_fill_next_extent(fieinfo, logical,
1718 phys, size, flags);
da5af127 1719 }
7f63eb77 1720
9a950d52
FL		 1721 if (start_blk > last_blk || ret)
1722 goto out;
7f63eb77 1723
9a950d52
FL		 1724 logical = blk_to_logical(inode, start_blk);
1725 phys = blk_to_logical(inode, map_bh.b_blocknr);
1726 size = map_bh.b_size;
1727 flags = 0;
1728 if (buffer_unwritten(&map_bh))
1729 flags = FIEMAP_EXTENT_UNWRITTEN;
7f63eb77 1730
9a950d52 1731 start_blk += logical_to_blk(inode, size);
7f63eb77 1732
9a950d52 1733prep_next:
7f63eb77
JK		 1734 cond_resched();
1735 if (fatal_signal_pending(current))
1736 ret = -EINTR;
1737 else
1738 goto next;
1739out:
1740 if (ret == 1)
1741 ret = 0;
1742
5955102c 1743 inode_unlock(inode);
7f63eb77 1744 return ret;
9ab70134
JK		 1745}
1746
95ae251f
EB		 1747static inline loff_t f2fs_readpage_limit(struct inode *inode)
1748{
1749 if (IS_ENABLED(CONFIG_FS_VERITY) &&
1750 (IS_VERITY(inode) || f2fs_verity_in_progress(inode)))
1751 return inode->i_sb->s_maxbytes;
1752
1753 return i_size_read(inode);
1754}
1755
2df0ab04
CY		 1756static int f2fs_read_single_page(struct inode *inode, struct page *page,
1757 unsigned nr_pages,
1758 struct f2fs_map_blocks *map,
1759 struct bio **bio_ret,
1760 sector_t *last_block_in_bio,
1761 bool is_readahead)
1762{
1763 struct bio *bio = *bio_ret;
1764 const unsigned blkbits = inode->i_blkbits;
1765 const unsigned blocksize = 1 << blkbits;
1766 sector_t block_in_file;
1767 sector_t last_block;
1768 sector_t last_block_in_file;
1769 sector_t block_nr;
1770 int ret = 0;
1771
4969c06a 1772 block_in_file = (sector_t)page_index(page);
2df0ab04 1773 last_block = block_in_file + nr_pages;
95ae251f 1774 last_block_in_file = (f2fs_readpage_limit(inode) + blocksize - 1) >>
2df0ab04
CY		 1775 blkbits;
1776 if (last_block > last_block_in_file)
1777 last_block = last_block_in_file;
1778
1779 /* just zeroing out page which is beyond EOF */
1780 if (block_in_file >= last_block)
1781 goto zero_out;
1782 /*
1783 * Map blocks using the previous result first.
1784 */
1785 if ((map->m_flags & F2FS_MAP_MAPPED) &&
1786 block_in_file > map->m_lblk &&
1787 block_in_file < (map->m_lblk + map->m_len))
1788 goto got_it;
1789
1790 /*
1791 * Then do more f2fs_map_blocks() calls until we are
1792 * done with this page.
1793 */
1794 map->m_lblk = block_in_file;
1795 map->m_len = last_block - block_in_file;
1796
1797 ret = f2fs_map_blocks(inode, map, 0, F2FS_GET_BLOCK_DEFAULT);
1798 if (ret)
1799 goto out;
1800got_it:
1801 if ((map->m_flags & F2FS_MAP_MAPPED)) {
1802 block_nr = map->m_pblk + block_in_file - map->m_lblk;
1803 SetPageMappedToDisk(page);
1804
4969c06a
JK		 1805 if (!PageUptodate(page) && (!PageSwapCache(page) &&
1806 !cleancache_get_page(page))) {
2df0ab04
CY		 1807 SetPageUptodate(page);
1808 goto confused;
1809 }
1810
1811 if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), block_nr,
93770ab7 1812 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 1813 ret = -EFSCORRUPTED;
2df0ab04
CY		 1814 goto out;
1815 }
1816 } else {
1817zero_out:
1818 zero_user_segment(page, 0, PAGE_SIZE);
95ae251f
EB		 1819 if (f2fs_need_verity(inode, page->index) &&
1820 !fsverity_verify_page(page)) {
1821 ret = -EIO;
1822 goto out;
1823 }
2df0ab04
CY		 1824 if (!PageUptodate(page))
1825 SetPageUptodate(page);
1826 unlock_page(page);
1827 goto out;
1828 }
1829
1830 /*
1831 * This page will go to BIO. Do we need to send this
1832 * BIO off first?
1833 */
8896cbdf
CY		 1834 if (bio && !page_is_mergeable(F2FS_I_SB(inode), bio,
1835 *last_block_in_bio, block_nr)) {
2df0ab04
CY		 1836submit_and_realloc:
1837 __submit_bio(F2FS_I_SB(inode), bio, DATA);
1838 bio = NULL;
1839 }
1840 if (bio == NULL) {
1841 bio = f2fs_grab_read_bio(inode, block_nr, nr_pages,
95ae251f 1842 is_readahead ? REQ_RAHEAD : 0, page->index);
2df0ab04
CY		 1843 if (IS_ERR(bio)) {
1844 ret = PTR_ERR(bio);
1845 bio = NULL;
1846 goto out;
1847 }
1848 }
1849
1850 /*
1851 * If the page is under writeback, we need to wait for
1852 * its completion to see the correct decrypted data.
1853 */
1854 f2fs_wait_on_block_writeback(inode, block_nr);
1855
1856 if (bio_add_page(bio, page, blocksize, 0) < blocksize)
1857 goto submit_and_realloc;
1858
1859 inc_page_count(F2FS_I_SB(inode), F2FS_RD_DATA);
1860 ClearPageError(page);
1861 *last_block_in_bio = block_nr;
1862 goto out;
1863confused:
1864 if (bio) {
1865 __submit_bio(F2FS_I_SB(inode), bio, DATA);
1866 bio = NULL;
1867 }
1868 unlock_page(page);
1869out:
1870 *bio_ret = bio;
1871 return ret;
1872}
1873
f1e88660
JK		 1874/*
1875 * This function was originally taken from fs/mpage.c, and customized for f2fs.
1876 * Major change was from block_size == page_size in f2fs by default.
e2e59414
JK		 1877 *
1878 * Note that the aops->readpages() function is ONLY used for read-ahead. If
1879 * this function ever deviates from doing just read-ahead, it should either
1880 * use ->readpage() or do the necessary surgery to decouple ->readpages()
1881 * from read-ahead.
f1e88660
JK		 1882 */
1883static int f2fs_mpage_readpages(struct address_space *mapping,
1884 struct list_head *pages, struct page *page,
e2e59414 1885 unsigned nr_pages, bool is_readahead)
f1e88660
JK		 1886{
1887 struct bio *bio = NULL;
f1e88660
JK		 1888 sector_t last_block_in_bio = 0;
1889 struct inode *inode = mapping->host;
f1e88660 1890 struct f2fs_map_blocks map;
2df0ab04 1891 int ret = 0;
f1e88660
JK		 1892
1893 map.m_pblk = 0;
1894 map.m_lblk = 0;
1895 map.m_len = 0;
1896 map.m_flags = 0;
da85985c 1897 map.m_next_pgofs = NULL;
c4020b2d 1898 map.m_next_extent = NULL;
d5097be5 1899 map.m_seg_type = NO_CHECK_TYPE;
f9d6d059 1900 map.m_may_create = false;
f1e88660 1901
736c0a74 1902 for (; nr_pages; nr_pages--) {
f1e88660 1903 if (pages) {
939afa94 1904 page = list_last_entry(pages, struct page, lru);
a83d50bc
KM		 1905
1906 prefetchw(&page->flags);
f1e88660
JK		 1907 list_del(&page->lru);
1908 if (add_to_page_cache_lru(page, mapping,
4969c06a 1909 page_index(page),
8a5c743e 1910 readahead_gfp_mask(mapping)))
f1e88660
JK		 1911 goto next_page;
1912 }
1913
2df0ab04
CY		 1914 ret = f2fs_read_single_page(inode, page, nr_pages, &map, &bio,
1915 &last_block_in_bio, is_readahead);
1916 if (ret) {
1917 SetPageError(page);
09cbfeaf 1918 zero_user_segment(page, 0, PAGE_SIZE);
f1e88660 1919 unlock_page(page);
f1e88660 1920 }
f1e88660
JK		 1921next_page:
1922 if (pages)
09cbfeaf 1923 put_page(page);
f1e88660
JK		 1924 }
1925 BUG_ON(pages && !list_empty(pages));
1926 if (bio)
4fc29c1a 1927 __submit_bio(F2FS_I_SB(inode), bio, DATA);
2df0ab04 1928 return pages ? 0 : ret;
f1e88660
JK		 1929}
1930
eb47b800
JK		 1931static int f2fs_read_data_page(struct file *file, struct page *page)
1932{
4969c06a 1933 struct inode *inode = page_file_mapping(page)->host;
b3d208f9 1934 int ret = -EAGAIN;
9ffe0fb5 1935
c20e89cd
CY		 1936 trace_f2fs_readpage(page, DATA);
1937
e1c42045 1938 /* If the file has inline data, try to read it directly */
9ffe0fb5
HL		 1939 if (f2fs_has_inline_data(inode))
1940 ret = f2fs_read_inline_data(inode, page);
b3d208f9 1941 if (ret == -EAGAIN)
4969c06a
JK		 1942 ret = f2fs_mpage_readpages(page_file_mapping(page),
1943 NULL, page, 1, false);
9ffe0fb5 1944 return ret;
eb47b800
JK		 1945}
1946
1947static int f2fs_read_data_pages(struct file *file,
1948 struct address_space *mapping,
1949 struct list_head *pages, unsigned nr_pages)
1950{
71cb4aff 1951 struct inode *inode = mapping->host;
939afa94 1952 struct page *page = list_last_entry(pages, struct page, lru);
b8c29400
CY		 1953
1954 trace_f2fs_readpages(inode, page, nr_pages);
9ffe0fb5
HL		 1955
1956 /* If the file has inline data, skip readpages */
1957 if (f2fs_has_inline_data(inode))
1958 return 0;
1959
e2e59414 1960 return f2fs_mpage_readpages(mapping, pages, NULL, nr_pages, true);
eb47b800
JK		 1961}
1962
7eab0c0d
HP		 1963static int encrypt_one_page(struct f2fs_io_info *fio)
1964{
1965 struct inode *inode = fio->page->mapping->host;
6aa58d8a 1966 struct page *mpage;
7eab0c0d
HP		 1967 gfp_t gfp_flags = GFP_NOFS;
1968
1958593e 1969 if (!f2fs_encrypted_file(inode))
7eab0c0d
HP		 1970 return 0;
1971
6dbb1796 1972 /* wait for GCed page writeback via META_MAPPING */
0ded69f6 1973 f2fs_wait_on_block_writeback(inode, fio->old_blkaddr);
7eab0c0d
HP		 1974
1975retry_encrypt:
53bc1d85
EB		 1976 fio->encrypted_page = fscrypt_encrypt_pagecache_blocks(fio->page,
1977 PAGE_SIZE, 0,
1978 gfp_flags);
6aa58d8a
CY		 1979 if (IS_ERR(fio->encrypted_page)) {
1980 /* flush pending IOs and wait for a while in the ENOMEM case */
1981 if (PTR_ERR(fio->encrypted_page) == -ENOMEM) {
1982 f2fs_flush_merged_writes(fio->sbi);
1983 congestion_wait(BLK_RW_ASYNC, HZ/50);
1984 gfp_flags |= __GFP_NOFAIL;
1985 goto retry_encrypt;
1986 }
1987 return PTR_ERR(fio->encrypted_page);
1988 }
7eab0c0d 1989
6aa58d8a
CY		 1990 mpage = find_lock_page(META_MAPPING(fio->sbi), fio->old_blkaddr);
1991 if (mpage) {
1992 if (PageUptodate(mpage))
1993 memcpy(page_address(mpage),
1994 page_address(fio->encrypted_page), PAGE_SIZE);
1995 f2fs_put_page(mpage, 1);
7eab0c0d 1996 }
6aa58d8a 1997 return 0;
7eab0c0d
HP		 1998}
1999
bb9e3bb8
CY		 2000static inline bool check_inplace_update_policy(struct inode *inode,
2001 struct f2fs_io_info *fio)
7eab0c0d 2002{
bb9e3bb8
CY		 2003 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2004 unsigned int policy = SM_I(sbi)->ipu_policy;
7eab0c0d 2005
bb9e3bb8
CY		 2006 if (policy & (0x1 << F2FS_IPU_FORCE))
2007 return true;
4d57b86d 2008 if (policy & (0x1 << F2FS_IPU_SSR) && f2fs_need_SSR(sbi))
bb9e3bb8
CY		 2009 return true;
2010 if (policy & (0x1 << F2FS_IPU_UTIL) &&
2011 utilization(sbi) > SM_I(sbi)->min_ipu_util)
2012 return true;
4d57b86d 2013 if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && f2fs_need_SSR(sbi) &&
bb9e3bb8
CY		 2014 utilization(sbi) > SM_I(sbi)->min_ipu_util)
2015 return true;
2016
2017 /*
2018 * IPU for rewrite async pages
2019 */
2020 if (policy & (0x1 << F2FS_IPU_ASYNC) &&
2021 fio && fio->op == REQ_OP_WRITE &&
2022 !(fio->op_flags & REQ_SYNC) &&
62230e0d 2023 !IS_ENCRYPTED(inode))
bb9e3bb8
CY		 2024 return true;
2025
2026 /* this is only set during fdatasync */
2027 if (policy & (0x1 << F2FS_IPU_FSYNC) &&
2028 is_inode_flag_set(inode, FI_NEED_IPU))
2029 return true;
2030
4354994f
DR		 2031 if (unlikely(fio && is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2032 !f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2033 return true;
2034
bb9e3bb8
CY		 2035 return false;
2036}
2037
4d57b86d 2038bool f2fs_should_update_inplace(struct inode *inode, struct f2fs_io_info *fio)
bb9e3bb8 2039{
1ad71a27
JK		 2040 if (f2fs_is_pinned_file(inode))
2041 return true;
bb9e3bb8
CY		 2042
2043 /* if this is cold file, we should overwrite to avoid fragmentation */
2044 if (file_is_cold(inode))
2045 return true;
2046
2047 return check_inplace_update_policy(inode, fio);
2048}
2049
4d57b86d 2050bool f2fs_should_update_outplace(struct inode *inode, struct f2fs_io_info *fio)
bb9e3bb8
CY		 2051{
2052 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2053
2054 if (test_opt(sbi, LFS))
2055 return true;
2056 if (S_ISDIR(inode->i_mode))
2057 return true;
af033b2a
CY		 2058 if (IS_NOQUOTA(inode))
2059 return true;
bb9e3bb8
CY		 2060 if (f2fs_is_atomic_file(inode))
2061 return true;
2062 if (fio) {
2063 if (is_cold_data(fio->page))
2064 return true;
2065 if (IS_ATOMIC_WRITTEN_PAGE(fio->page))
2066 return true;
4354994f
DR		 2067 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2068 f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2069 return true;
bb9e3bb8
CY		 2070 }
2071 return false;
2072}
2073
7eab0c0d
HP		 2074static inline bool need_inplace_update(struct f2fs_io_info *fio)
2075{
2076 struct inode *inode = fio->page->mapping->host;
2077
4d57b86d 2078 if (f2fs_should_update_outplace(inode, fio))
7eab0c0d
HP		 2079 return false;
2080
4d57b86d 2081 return f2fs_should_update_inplace(inode, fio);
7eab0c0d
HP		 2082}
2083
4d57b86d 2084int f2fs_do_write_data_page(struct f2fs_io_info *fio)
eb47b800 2085{
05ca3632 2086 struct page *page = fio->page;
eb47b800 2087 struct inode *inode = page->mapping->host;
eb47b800 2088 struct dnode_of_data dn;
e959c8f5 2089 struct extent_info ei = {0,0,0};
7735730d 2090 struct node_info ni;
e959c8f5 2091 bool ipu_force = false;
eb47b800
JK		 2092 int err = 0;
2093
2094 set_new_dnode(&dn, inode, NULL, NULL, 0);
e959c8f5
HP		 2095 if (need_inplace_update(fio) &&
2096 f2fs_lookup_extent_cache(inode, page->index, &ei)) {
2097 fio->old_blkaddr = ei.blk + page->index - ei.fofs;
a817737e 2098
c9b60788 2099 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
93770ab7 2100 DATA_GENERIC_ENHANCE))
10f966bb 2101 return -EFSCORRUPTED;
c9b60788
CY		 2102
2103 ipu_force = true;
2104 fio->need_lock = LOCK_DONE;
2105 goto got_it;
e959c8f5 2106 }
279d6df2 2107
d29460e5
JK		 2108 /* Deadlock due to between page->lock and f2fs_lock_op */
2109 if (fio->need_lock == LOCK_REQ && !f2fs_trylock_op(fio->sbi))
2110 return -EAGAIN;
279d6df2 2111
4d57b86d 2112 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
eb47b800 2113 if (err)
279d6df2 2114 goto out;
eb47b800 2115
28bc106b 2116 fio->old_blkaddr = dn.data_blkaddr;
eb47b800
JK		 2117
2118 /* This page is already truncated */
7a9d7548 2119 if (fio->old_blkaddr == NULL_ADDR) {
2bca1e23 2120 ClearPageUptodate(page);
2baf0781 2121 clear_cold_data(page);
eb47b800 2122 goto out_writepage;
2bca1e23 2123 }
e959c8f5 2124got_it:
c9b60788
CY		 2125 if (__is_valid_data_blkaddr(fio->old_blkaddr) &&
2126 !f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
93770ab7 2127 DATA_GENERIC_ENHANCE)) {
10f966bb 2128 err = -EFSCORRUPTED;
c9b60788
CY		 2129 goto out_writepage;
2130 }
eb47b800
JK		 2131 /*
2132 * If current allocation needs SSR,
2133 * it had better in-place writes for updated data.
2134 */
93770ab7
CY		 2135 if (ipu_force ||
2136 (__is_valid_data_blkaddr(fio->old_blkaddr) &&
7b525dd0 2137 need_inplace_update(fio))) {
cc15620b
JK		 2138 err = encrypt_one_page(fio);
2139 if (err)
2140 goto out_writepage;
2141
2142 set_page_writeback(page);
17c50035 2143 ClearPageError(page);
279d6df2 2144 f2fs_put_dnode(&dn);
cc15620b 2145 if (fio->need_lock == LOCK_REQ)
279d6df2 2146 f2fs_unlock_op(fio->sbi);
4d57b86d 2147 err = f2fs_inplace_write_data(fio);
6492a335
CY		 2148 if (err) {
2149 if (f2fs_encrypted_file(inode))
d2d0727b 2150 fscrypt_finalize_bounce_page(&fio->encrypted_page);
6492a335
CY		 2151 if (PageWriteback(page))
2152 end_page_writeback(page);
cd23ffa9
CY		 2153 } else {
2154 set_inode_flag(inode, FI_UPDATE_WRITE);
6492a335 2155 }
7eab0c0d 2156 trace_f2fs_do_write_data_page(fio->page, IPU);
279d6df2 2157 return err;
eb47b800 2158 }
279d6df2 2159
cc15620b
JK		 2160 if (fio->need_lock == LOCK_RETRY) {
2161 if (!f2fs_trylock_op(fio->sbi)) {
2162 err = -EAGAIN;
2163 goto out_writepage;
2164 }
2165 fio->need_lock = LOCK_REQ;
2166 }
2167
7735730d
CY		 2168 err = f2fs_get_node_info(fio->sbi, dn.nid, &ni);
2169 if (err)
2170 goto out_writepage;
2171
2172 fio->version = ni.version;
2173
cc15620b
JK		 2174 err = encrypt_one_page(fio);
2175 if (err)
2176 goto out_writepage;
2177
2178 set_page_writeback(page);
17c50035 2179 ClearPageError(page);
cc15620b 2180
279d6df2 2181 /* LFS mode write path */
4d57b86d 2182 f2fs_outplace_write_data(&dn, fio);
279d6df2
HP		 2183 trace_f2fs_do_write_data_page(page, OPU);
2184 set_inode_flag(inode, FI_APPEND_WRITE);
2185 if (page->index == 0)
2186 set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
eb47b800
JK		 2187out_writepage:
2188 f2fs_put_dnode(&dn);
279d6df2 2189out:
cc15620b 2190 if (fio->need_lock == LOCK_REQ)
279d6df2 2191 f2fs_unlock_op(fio->sbi);
eb47b800
JK		 2192 return err;
2193}
2194
d68f735b 2195static int __write_data_page(struct page *page, bool *submitted,
8648de2c
CY		 2196 struct bio **bio,
2197 sector_t *last_block,
b0af6d49
CY		 2198 struct writeback_control *wbc,
2199 enum iostat_type io_type)
eb47b800
JK		 2200{
2201 struct inode *inode = page->mapping->host;
4081363f 2202 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
eb47b800
JK		 2203 loff_t i_size = i_size_read(inode);
2204 const pgoff_t end_index = ((unsigned long long) i_size)
09cbfeaf 2205 >> PAGE_SHIFT;
26de9b11 2206 loff_t psize = (page->index + 1) << PAGE_SHIFT;
9ffe0fb5 2207 unsigned offset = 0;
39936837 2208 bool need_balance_fs = false;
eb47b800 2209 int err = 0;
458e6197 2210 struct f2fs_io_info fio = {
05ca3632 2211 .sbi = sbi,
39d787be 2212 .ino = inode->i_ino,
458e6197 2213 .type = DATA,
04d328de 2214 .op = REQ_OP_WRITE,
7637241e 2215 .op_flags = wbc_to_write_flags(wbc),
e959c8f5 2216 .old_blkaddr = NULL_ADDR,
05ca3632 2217 .page = page,
4375a336 2218 .encrypted_page = NULL,
d68f735b 2219 .submitted = false,
cc15620b 2220 .need_lock = LOCK_RETRY,
b0af6d49 2221 .io_type = io_type,
578c6478 2222 .io_wbc = wbc,
8648de2c
CY		 2223 .bio = bio,
2224 .last_block = last_block,
458e6197 2225 };
eb47b800 2226
ecda0de3
CY		 2227 trace_f2fs_writepage(page, DATA);
2228
db198ae0
CY		 2229 /* we should bypass data pages to proceed the kworkder jobs */
2230 if (unlikely(f2fs_cp_error(sbi))) {
2231 mapping_set_error(page->mapping, -EIO);
1174abfd
CY		 2232 /*
2233 * don't drop any dirty dentry pages for keeping lastest
2234 * directory structure.
2235 */
2236 if (S_ISDIR(inode->i_mode))
2237 goto redirty_out;
db198ae0
CY		 2238 goto out;
2239 }
2240
0771fcc7
CY		 2241 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
2242 goto redirty_out;
2243
95ae251f 2244 if (page->index < end_index || f2fs_verity_in_progress(inode))
39936837 2245 goto write;
eb47b800
JK		 2246
2247 /*
2248 * If the offset is out-of-range of file size,
2249 * this page does not have to be written to disk.
2250 */
09cbfeaf 2251 offset = i_size & (PAGE_SIZE - 1);
76f60268 2252 if ((page->index >= end_index + 1) || !offset)
39936837 2253 goto out;
eb47b800 2254
09cbfeaf 2255 zero_user_segment(page, offset, PAGE_SIZE);
39936837 2256write:
1e84371f
JK		 2257 if (f2fs_is_drop_cache(inode))
2258 goto out;
e6e5f561
JK		 2259 /* we should not write 0'th page having journal header */
2260 if (f2fs_is_volatile_file(inode) && (!page->index ||
2261 (!wbc->for_reclaim &&
4d57b86d 2262 f2fs_available_free_memory(sbi, BASE_CHECK))))
1e84371f 2263 goto redirty_out;
eb47b800 2264
39936837 2265 /* Dentry blocks are controlled by checkpoint */
eb47b800 2266 if (S_ISDIR(inode->i_mode)) {
cc15620b 2267 fio.need_lock = LOCK_DONE;
4d57b86d 2268 err = f2fs_do_write_data_page(&fio);
8618b881
JK		 2269 goto done;
2270 }
9ffe0fb5 2271
8618b881 2272 if (!wbc->for_reclaim)
39936837 2273 need_balance_fs = true;
7f3037a5 2274 else if (has_not_enough_free_secs(sbi, 0, 0))
39936837 2275 goto redirty_out;
ef095d19
JK		 2276 else
2277 set_inode_flag(inode, FI_HOT_DATA);
eb47b800 2278
b3d208f9 2279 err = -EAGAIN;
dd7b2333 2280 if (f2fs_has_inline_data(inode)) {
b3d208f9 2281 err = f2fs_write_inline_data(inode, page);
dd7b2333
YH		 2282 if (!err)
2283 goto out;
2284 }
279d6df2 2285
cc15620b 2286 if (err == -EAGAIN) {
4d57b86d 2287 err = f2fs_do_write_data_page(&fio);
cc15620b
JK		 2288 if (err == -EAGAIN) {
2289 fio.need_lock = LOCK_REQ;
4d57b86d 2290 err = f2fs_do_write_data_page(&fio);
cc15620b
JK		 2291 }
2292 }
a0d00fad 2293
eb449797
CY		 2294 if (err) {
2295 file_set_keep_isize(inode);
2296 } else {
2297 down_write(&F2FS_I(inode)->i_sem);
2298 if (F2FS_I(inode)->last_disk_size < psize)
2299 F2FS_I(inode)->last_disk_size = psize;
2300 up_write(&F2FS_I(inode)->i_sem);
2301 }
279d6df2 2302
8618b881
JK		 2303done:
2304 if (err && err != -ENOENT)
2305 goto redirty_out;
eb47b800 2306
39936837 2307out:
a7ffdbe2 2308 inode_dec_dirty_pages(inode);
2baf0781 2309 if (err) {
2bca1e23 2310 ClearPageUptodate(page);
2baf0781
CY		 2311 clear_cold_data(page);
2312 }
0c3a5797
CY		 2313
2314 if (wbc->for_reclaim) {
bab475c5 2315 f2fs_submit_merged_write_cond(sbi, NULL, page, 0, DATA);
ef095d19 2316 clear_inode_flag(inode, FI_HOT_DATA);
4d57b86d 2317 f2fs_remove_dirty_inode(inode);
d68f735b 2318 submitted = NULL;
0c3a5797
CY		 2319 }
2320
eb47b800 2321 unlock_page(page);
186857c5 2322 if (!S_ISDIR(inode->i_mode) && !IS_NOQUOTA(inode) &&
0b20fcec 2323 !F2FS_I(inode)->cp_task)
a7881893 2324 f2fs_balance_fs(sbi, need_balance_fs);
0c3a5797 2325
d68f735b 2326 if (unlikely(f2fs_cp_error(sbi))) {
b9109b0e 2327 f2fs_submit_merged_write(sbi, DATA);
0b20fcec 2328 f2fs_submit_merged_ipu_write(sbi, bio, NULL);
d68f735b
JK		 2329 submitted = NULL;
2330 }
2331
2332 if (submitted)
2333 *submitted = fio.submitted;
0c3a5797 2334
eb47b800
JK		 2335 return 0;
2336
eb47b800 2337redirty_out:
76f60268 2338 redirty_page_for_writepage(wbc, page);
5b19d284
JK		 2339 /*
2340 * pageout() in MM traslates EAGAIN, so calls handle_write_error()
2341 * -> mapping_set_error() -> set_bit(AS_EIO, ...).
2342 * file_write_and_wait_range() will see EIO error, which is critical
2343 * to return value of fsync() followed by atomic_write failure to user.
2344 */
2345 if (!err || wbc->for_reclaim)
0002b61b 2346 return AOP_WRITEPAGE_ACTIVATE;
b230e6ca
JK		 2347 unlock_page(page);
2348 return err;
fa9150a8
NJ		 2349}
2350
f566bae8
JK		 2351static int f2fs_write_data_page(struct page *page,
2352 struct writeback_control *wbc)
2353{
8648de2c 2354 return __write_data_page(page, NULL, NULL, NULL, wbc, FS_DATA_IO);
f566bae8
JK		 2355}
2356
8f46dcae
CY		 2357/*
2358 * This function was copied from write_cche_pages from mm/page-writeback.c.
2359 * The major change is making write step of cold data page separately from
2360 * warm/hot data page.
2361 */
2362static int f2fs_write_cache_pages(struct address_space *mapping,
b0af6d49
CY		 2363 struct writeback_control *wbc,
2364 enum iostat_type io_type)
8f46dcae
CY		 2365{
2366 int ret = 0;
2367 int done = 0;
2368 struct pagevec pvec;
c29fd0c0 2369 struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
8648de2c
CY		 2370 struct bio *bio = NULL;
2371 sector_t last_block;
8f46dcae
CY		 2372 int nr_pages;
2373 pgoff_t uninitialized_var(writeback_index);
2374 pgoff_t index;
2375 pgoff_t end; /* Inclusive */
2376 pgoff_t done_index;
2377 int cycled;
2378 int range_whole = 0;
10bbd235 2379 xa_mark_t tag;
bab475c5 2380 int nwritten = 0;
8f46dcae 2381
86679820 2382 pagevec_init(&pvec);
46ae957f 2383
ef095d19
JK		 2384 if (get_dirty_pages(mapping->host) <=
2385 SM_I(F2FS_M_SB(mapping))->min_hot_blocks)
2386 set_inode_flag(mapping->host, FI_HOT_DATA);
2387 else
2388 clear_inode_flag(mapping->host, FI_HOT_DATA);
2389
8f46dcae
CY		 2390 if (wbc->range_cyclic) {
2391 writeback_index = mapping->writeback_index; /* prev offset */
2392 index = writeback_index;
2393 if (index == 0)
2394 cycled = 1;
2395 else
2396 cycled = 0;
2397 end = -1;
2398 } else {
09cbfeaf
KS		 2399 index = wbc->range_start >> PAGE_SHIFT;
2400 end = wbc->range_end >> PAGE_SHIFT;
8f46dcae
CY		 2401 if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
2402 range_whole = 1;
2403 cycled = 1; /* ignore range_cyclic tests */
2404 }
2405 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2406 tag = PAGECACHE_TAG_TOWRITE;
2407 else
2408 tag = PAGECACHE_TAG_DIRTY;
2409retry:
2410 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2411 tag_pages_for_writeback(mapping, index, end);
2412 done_index = index;
2413 while (!done && (index <= end)) {
2414 int i;
2415
69c4f35d 2416 nr_pages = pagevec_lookup_range_tag(&pvec, mapping, &index, end,
67fd707f 2417 tag);
8f46dcae
CY		 2418 if (nr_pages == 0)
2419 break;
2420
2421 for (i = 0; i < nr_pages; i++) {
2422 struct page *page = pvec.pages[i];
d68f735b 2423 bool submitted = false;
8f46dcae 2424
f8de4331 2425 /* give a priority to WB_SYNC threads */
c29fd0c0 2426 if (atomic_read(&sbi->wb_sync_req[DATA]) &&
f8de4331
CY		 2427 wbc->sync_mode == WB_SYNC_NONE) {
2428 done = 1;
2429 break;
2430 }
2431
8f46dcae 2432 done_index = page->index;
d29460e5 2433retry_write:
8f46dcae
CY		 2434 lock_page(page);
2435
2436 if (unlikely(page->mapping != mapping)) {
2437continue_unlock:
2438 unlock_page(page);
2439 continue;
2440 }
2441
2442 if (!PageDirty(page)) {
2443 /* someone wrote it for us */
2444 goto continue_unlock;
2445 }
2446
8f46dcae 2447 if (PageWriteback(page)) {
0b20fcec 2448 if (wbc->sync_mode != WB_SYNC_NONE)
fec1d657 2449 f2fs_wait_on_page_writeback(page,
bae0ee7a 2450 DATA, true, true);
0b20fcec 2451 else
8f46dcae
CY		 2452 goto continue_unlock;
2453 }
2454
8f46dcae
CY		 2455 if (!clear_page_dirty_for_io(page))
2456 goto continue_unlock;
2457
8648de2c
CY		 2458 ret = __write_data_page(page, &submitted, &bio,
2459 &last_block, wbc, io_type);
8f46dcae 2460 if (unlikely(ret)) {
0002b61b
CY		 2461 /*
2462 * keep nr_to_write, since vfs uses this to
2463 * get # of written pages.
2464 */
2465 if (ret == AOP_WRITEPAGE_ACTIVATE) {
2466 unlock_page(page);
2467 ret = 0;
2468 continue;
d29460e5
JK		 2469 } else if (ret == -EAGAIN) {
2470 ret = 0;
2471 if (wbc->sync_mode == WB_SYNC_ALL) {
2472 cond_resched();
2473 congestion_wait(BLK_RW_ASYNC,
2474 HZ/50);
2475 goto retry_write;
2476 }
2477 continue;
0002b61b 2478 }
b230e6ca
JK		 2479 done_index = page->index + 1;
2480 done = 1;
2481 break;
d68f735b 2482 } else if (submitted) {
bab475c5 2483 nwritten++;
8f46dcae
CY		 2484 }
2485
f8de4331 2486 if (--wbc->nr_to_write <= 0 &&
687de7f1 2487 wbc->sync_mode == WB_SYNC_NONE) {
8f46dcae
CY		 2488 done = 1;
2489 break;
2490 }
2491 }
2492 pagevec_release(&pvec);
2493 cond_resched();
2494 }
2495
8f46dcae
CY		 2496 if (!cycled && !done) {
2497 cycled = 1;
2498 index = 0;
2499 end = writeback_index - 1;
2500 goto retry;
2501 }
2502 if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
2503 mapping->writeback_index = done_index;
2504
bab475c5 2505 if (nwritten)
b9109b0e 2506 f2fs_submit_merged_write_cond(F2FS_M_SB(mapping), mapping->host,
bab475c5 2507 NULL, 0, DATA);
8648de2c
CY		 2508 /* submit cached bio of IPU write */
2509 if (bio)
0b20fcec 2510 f2fs_submit_merged_ipu_write(sbi, &bio, NULL);
6ca56ca4 2511
8f46dcae
CY		 2512 return ret;
2513}
2514
853137ce
JK		 2515static inline bool __should_serialize_io(struct inode *inode,
2516 struct writeback_control *wbc)
2517{
2518 if (!S_ISREG(inode->i_mode))
2519 return false;
af033b2a
CY		 2520 if (IS_NOQUOTA(inode))
2521 return false;
040d2bb3
CY		 2522 /* to avoid deadlock in path of data flush */
2523 if (F2FS_I(inode)->cp_task)
2524 return false;
853137ce
JK		 2525 if (wbc->sync_mode != WB_SYNC_ALL)
2526 return true;
2527 if (get_dirty_pages(inode) >= SM_I(F2FS_I_SB(inode))->min_seq_blocks)
2528 return true;
2529 return false;
2530}
2531
fc99fe27 2532static int __f2fs_write_data_pages(struct address_space *mapping,
b0af6d49
CY		 2533 struct writeback_control *wbc,
2534 enum iostat_type io_type)
eb47b800
JK		 2535{
2536 struct inode *inode = mapping->host;
4081363f 2537 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
9dfa1baf 2538 struct blk_plug plug;
eb47b800 2539 int ret;
853137ce 2540 bool locked = false;
eb47b800 2541
cfb185a1 2542 /* deal with chardevs and other special file */
2543 if (!mapping->a_ops->writepage)
2544 return 0;
2545
6a290544
CY		 2546 /* skip writing if there is no dirty page in this inode */
2547 if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE)
2548 return 0;
2549
0771fcc7
CY		 2550 /* during POR, we don't need to trigger writepage at all. */
2551 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
2552 goto skip_write;
2553
af033b2a
CY		 2554 if ((S_ISDIR(inode->i_mode) || IS_NOQUOTA(inode)) &&
2555 wbc->sync_mode == WB_SYNC_NONE &&
a1257023 2556 get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) &&
4d57b86d 2557 f2fs_available_free_memory(sbi, DIRTY_DENTS))
a1257023
JK		 2558 goto skip_write;
2559
d323d005 2560 /* skip writing during file defragment */
91942321 2561 if (is_inode_flag_set(inode, FI_DO_DEFRAG))
d323d005
CY		 2562 goto skip_write;
2563
d31c7c3f
YH		 2564 trace_f2fs_writepages(mapping->host, wbc, DATA);
2565
687de7f1
JK		 2566 /* to avoid spliting IOs due to mixed WB_SYNC_ALL and WB_SYNC_NONE */
2567 if (wbc->sync_mode == WB_SYNC_ALL)
c29fd0c0
CY		 2568 atomic_inc(&sbi->wb_sync_req[DATA]);
2569 else if (atomic_read(&sbi->wb_sync_req[DATA]))
687de7f1
JK		 2570 goto skip_write;
2571
853137ce
JK		 2572 if (__should_serialize_io(inode, wbc)) {
2573 mutex_lock(&sbi->writepages);
2574 locked = true;
2575 }
2576
9dfa1baf 2577 blk_start_plug(&plug);
b0af6d49 2578 ret = f2fs_write_cache_pages(mapping, wbc, io_type);
9dfa1baf 2579 blk_finish_plug(&plug);
687de7f1 2580
853137ce
JK		 2581 if (locked)
2582 mutex_unlock(&sbi->writepages);
2583
687de7f1 2584 if (wbc->sync_mode == WB_SYNC_ALL)
c29fd0c0 2585 atomic_dec(&sbi->wb_sync_req[DATA]);
28ea6162
JK		 2586 /*
2587 * if some pages were truncated, we cannot guarantee its mapping->host
2588 * to detect pending bios.
2589 */
458e6197 2590
4d57b86d 2591 f2fs_remove_dirty_inode(inode);
eb47b800 2592 return ret;
d3baf95d
JK		 2593
2594skip_write:
a7ffdbe2 2595 wbc->pages_skipped += get_dirty_pages(inode);
d31c7c3f 2596 trace_f2fs_writepages(mapping->host, wbc, DATA);
d3baf95d 2597 return 0;
eb47b800
JK		 2598}
2599
b0af6d49
CY		 2600static int f2fs_write_data_pages(struct address_space *mapping,
2601 struct writeback_control *wbc)
2602{
2603 struct inode *inode = mapping->host;
2604
2605 return __f2fs_write_data_pages(mapping, wbc,
2606 F2FS_I(inode)->cp_task == current ?
2607 FS_CP_DATA_IO : FS_DATA_IO);
2608}
2609
3aab8f82
CY		 2610static void f2fs_write_failed(struct address_space *mapping, loff_t to)
2611{
2612 struct inode *inode = mapping->host;
819d9153 2613 loff_t i_size = i_size_read(inode);
3aab8f82 2614
95ae251f
EB		 2615 /* In the fs-verity case, f2fs_end_enable_verity() does the truncate */
2616 if (to > i_size && !f2fs_verity_in_progress(inode)) {
a33c1502 2617 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
6f8d4455 2618 down_write(&F2FS_I(inode)->i_mmap_sem);
a33c1502 2619
819d9153 2620 truncate_pagecache(inode, i_size);
c42d28ce
CY		 2621 if (!IS_NOQUOTA(inode))
2622 f2fs_truncate_blocks(inode, i_size, true);
a33c1502 2623
5a3a2d83 2624 up_write(&F2FS_I(inode)->i_mmap_sem);
6f8d4455 2625 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3aab8f82
CY		 2626 }
2627}
2628
2aadac08
JK		 2629static int prepare_write_begin(struct f2fs_sb_info *sbi,
2630 struct page *page, loff_t pos, unsigned len,
2631 block_t *blk_addr, bool *node_changed)
2632{
2633 struct inode *inode = page->mapping->host;
2634 pgoff_t index = page->index;
2635 struct dnode_of_data dn;
2636 struct page *ipage;
b4d07a3e 2637 bool locked = false;
e15882b6 2638 struct extent_info ei = {0,0,0};
2aadac08 2639 int err = 0;
2866fb16 2640 int flag;
2aadac08 2641
24b84912
JK		 2642 /*
2643 * we already allocated all the blocks, so we don't need to get
2644 * the block addresses when there is no need to fill the page.
2645 */
dc91de78 2646 if (!f2fs_has_inline_data(inode) && len == PAGE_SIZE &&
95ae251f
EB		 2647 !is_inode_flag_set(inode, FI_NO_PREALLOC) &&
2648 !f2fs_verity_in_progress(inode))
24b84912
JK		 2649 return 0;
2650
2866fb16
SY		 2651 /* f2fs_lock_op avoids race between write CP and convert_inline_page */
2652 if (f2fs_has_inline_data(inode) && pos + len > MAX_INLINE_DATA(inode))
2653 flag = F2FS_GET_BLOCK_DEFAULT;
2654 else
2655 flag = F2FS_GET_BLOCK_PRE_AIO;
2656
b4d07a3e 2657 if (f2fs_has_inline_data(inode) ||
09cbfeaf 2658 (pos & PAGE_MASK) >= i_size_read(inode)) {
2866fb16 2659 __do_map_lock(sbi, flag, true);
b4d07a3e
JK		 2660 locked = true;
2661 }
2662restart:
2aadac08 2663 /* check inline_data */
4d57b86d 2664 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2aadac08
JK		 2665 if (IS_ERR(ipage)) {
2666 err = PTR_ERR(ipage);
2667 goto unlock_out;
2668 }
2669
2670 set_new_dnode(&dn, inode, ipage, ipage, 0);
2671
2672 if (f2fs_has_inline_data(inode)) {
f2470371 2673 if (pos + len <= MAX_INLINE_DATA(inode)) {
4d57b86d 2674 f2fs_do_read_inline_data(page, ipage);
91942321 2675 set_inode_flag(inode, FI_DATA_EXIST);
ab47036d
CY		 2676 if (inode->i_nlink)
2677 set_inline_node(ipage);
2aadac08
JK		 2678 } else {
2679 err = f2fs_convert_inline_page(&dn, page);
2680 if (err)
b4d07a3e
JK		 2681 goto out;
2682 if (dn.data_blkaddr == NULL_ADDR)
2683 err = f2fs_get_block(&dn, index);
2684 }
2685 } else if (locked) {
2686 err = f2fs_get_block(&dn, index);
2687 } else {
2688 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
2689 dn.data_blkaddr = ei.blk + index - ei.fofs;
2690 } else {
b4d07a3e 2691 /* hole case */
4d57b86d 2692 err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
4da7bf5a 2693 if (err || dn.data_blkaddr == NULL_ADDR) {
b4d07a3e 2694 f2fs_put_dnode(&dn);
59c9081b
YH		 2695 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO,
2696 true);
2866fb16 2697 WARN_ON(flag != F2FS_GET_BLOCK_PRE_AIO);
b4d07a3e
JK		 2698 locked = true;
2699 goto restart;
2700 }
2aadac08
JK		 2701 }
2702 }
b4d07a3e 2703
2aadac08
JK		 2704 /* convert_inline_page can make node_changed */
2705 *blk_addr = dn.data_blkaddr;
2706 *node_changed = dn.node_changed;
b4d07a3e 2707out:
2aadac08
JK		 2708 f2fs_put_dnode(&dn);
2709unlock_out:
b4d07a3e 2710 if (locked)
2866fb16 2711 __do_map_lock(sbi, flag, false);
2aadac08
JK		 2712 return err;
2713}
2714
eb47b800
JK		 2715static int f2fs_write_begin(struct file *file, struct address_space *mapping,
2716 loff_t pos, unsigned len, unsigned flags,
2717 struct page **pagep, void **fsdata)
2718{
2719 struct inode *inode = mapping->host;
4081363f 2720 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
86531d6b 2721 struct page *page = NULL;
09cbfeaf 2722 pgoff_t index = ((unsigned long long) pos) >> PAGE_SHIFT;
a2e2e76b 2723 bool need_balance = false, drop_atomic = false;
2aadac08 2724 block_t blkaddr = NULL_ADDR;
eb47b800
JK		 2725 int err = 0;
2726
62aed044
CY		 2727 trace_f2fs_write_begin(inode, pos, len, flags);
2728
00e09c0b
CY		 2729 if (!f2fs_is_checkpoint_ready(sbi)) {
2730 err = -ENOSPC;
4354994f 2731 goto fail;
00e09c0b 2732 }
4354994f 2733
455e3a58
JK		 2734 if ((f2fs_is_atomic_file(inode) &&
2735 !f2fs_available_free_memory(sbi, INMEM_PAGES)) ||
2736 is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
57864ae5 2737 err = -ENOMEM;
a2e2e76b 2738 drop_atomic = true;
57864ae5
JK		 2739 goto fail;
2740 }
2741
5f727395
JK		 2742 /*
2743 * We should check this at this moment to avoid deadlock on inode page
2744 * and #0 page. The locking rule for inline_data conversion should be:
2745 * lock_page(page #0) -> lock_page(inode_page)
2746 */
2747 if (index != 0) {
2748 err = f2fs_convert_inline_inode(inode);
2749 if (err)
2750 goto fail;
2751 }
afcb7ca0 2752repeat:
86d54795
JK		 2753 /*
2754 * Do not use grab_cache_page_write_begin() to avoid deadlock due to
2755 * wait_for_stable_page. Will wait that below with our IO control.
2756 */
01eccef7 2757 page = f2fs_pagecache_get_page(mapping, index,
86d54795 2758 FGP_LOCK | FGP_WRITE | FGP_CREAT, GFP_NOFS);
3aab8f82
CY		 2759 if (!page) {
2760 err = -ENOMEM;
2761 goto fail;
2762 }
d5f66990 2763
eb47b800
JK		 2764 *pagep = page;
2765
2aadac08
JK		 2766 err = prepare_write_begin(sbi, page, pos, len,
2767 &blkaddr, &need_balance);
9ba69cf9 2768 if (err)
2aadac08 2769 goto fail;
9ba69cf9 2770
af033b2a
CY		 2771 if (need_balance && !IS_NOQUOTA(inode) &&
2772 has_not_enough_free_secs(sbi, 0, 0)) {
2a340760 2773 unlock_page(page);
2c4db1a6 2774 f2fs_balance_fs(sbi, true);
2a340760
JK		 2775 lock_page(page);
2776 if (page->mapping != mapping) {
2777 /* The page got truncated from under us */
2778 f2fs_put_page(page, 1);
2779 goto repeat;
2780 }
2781 }
2782
bae0ee7a 2783 f2fs_wait_on_page_writeback(page, DATA, false, true);
b3d208f9 2784
649d7df2
JK		 2785 if (len == PAGE_SIZE || PageUptodate(page))
2786 return 0;
eb47b800 2787
95ae251f
EB		 2788 if (!(pos & (PAGE_SIZE - 1)) && (pos + len) >= i_size_read(inode) &&
2789 !f2fs_verity_in_progress(inode)) {
746e2403
YH		 2790 zero_user_segment(page, len, PAGE_SIZE);
2791 return 0;
2792 }
2793
2aadac08 2794 if (blkaddr == NEW_ADDR) {
09cbfeaf 2795 zero_user_segment(page, 0, PAGE_SIZE);
649d7df2 2796 SetPageUptodate(page);
eb47b800 2797 } else {
93770ab7
CY		 2798 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
2799 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 2800 err = -EFSCORRUPTED;
93770ab7
CY		 2801 goto fail;
2802 }
13ba41e3
JK		 2803 err = f2fs_submit_page_read(inode, page, blkaddr);
2804 if (err)
78682f79 2805 goto fail;
d54c795b 2806
393ff91f 2807 lock_page(page);
6bacf52f 2808 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 2809 f2fs_put_page(page, 1);
2810 goto repeat;
eb47b800 2811 }
1563ac75
CY		 2812 if (unlikely(!PageUptodate(page))) {
2813 err = -EIO;
2814 goto fail;
4375a336 2815 }
eb47b800 2816 }
eb47b800 2817 return 0;
9ba69cf9 2818
3aab8f82 2819fail:
86531d6b 2820 f2fs_put_page(page, 1);
3aab8f82 2821 f2fs_write_failed(mapping, pos + len);
a2e2e76b 2822 if (drop_atomic)
4d57b86d 2823 f2fs_drop_inmem_pages_all(sbi, false);
3aab8f82 2824 return err;
eb47b800
JK		 2825}
2826
a1dd3c13
JK		 2827static int f2fs_write_end(struct file *file,
2828 struct address_space *mapping,
2829 loff_t pos, unsigned len, unsigned copied,
2830 struct page *page, void *fsdata)
2831{
2832 struct inode *inode = page->mapping->host;
2833
dfb2bf38
CY		 2834 trace_f2fs_write_end(inode, pos, len, copied);
2835
649d7df2
JK		 2836 /*
2837 * This should be come from len == PAGE_SIZE, and we expect copied
2838 * should be PAGE_SIZE. Otherwise, we treat it with zero copied and
2839 * let generic_perform_write() try to copy data again through copied=0.
2840 */
2841 if (!PageUptodate(page)) {
746e2403 2842 if (unlikely(copied != len))
649d7df2
JK		 2843 copied = 0;
2844 else
2845 SetPageUptodate(page);
2846 }
2847 if (!copied)
2848 goto unlock_out;
2849
34ba94ba 2850 set_page_dirty(page);
a1dd3c13 2851
95ae251f
EB		 2852 if (pos + copied > i_size_read(inode) &&
2853 !f2fs_verity_in_progress(inode))
fc9581c8 2854 f2fs_i_size_write(inode, pos + copied);
649d7df2 2855unlock_out:
3024c9a1 2856 f2fs_put_page(page, 1);
d0239e1b 2857 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
a1dd3c13
JK		 2858 return copied;
2859}
2860
6f673763
OS		 2861static int check_direct_IO(struct inode *inode, struct iov_iter *iter,
2862 loff_t offset)
944fcfc1 2863{
8a56dd96
JK		 2864 unsigned i_blkbits = READ_ONCE(inode->i_blkbits);
2865 unsigned blkbits = i_blkbits;
2866 unsigned blocksize_mask = (1 << blkbits) - 1;
2867 unsigned long align = offset | iov_iter_alignment(iter);
2868 struct block_device *bdev = inode->i_sb->s_bdev;
2869
2870 if (align & blocksize_mask) {
2871 if (bdev)
2872 blkbits = blksize_bits(bdev_logical_block_size(bdev));
2873 blocksize_mask = (1 << blkbits) - 1;
2874 if (align & blocksize_mask)
2875 return -EINVAL;
2876 return 1;
2877 }
944fcfc1
JK		 2878 return 0;
2879}
2880
02b16d0a
CY		 2881static void f2fs_dio_end_io(struct bio *bio)
2882{
2883 struct f2fs_private_dio *dio = bio->bi_private;
2884
2885 dec_page_count(F2FS_I_SB(dio->inode),
2886 dio->write ? F2FS_DIO_WRITE : F2FS_DIO_READ);
2887
2888 bio->bi_private = dio->orig_private;
2889 bio->bi_end_io = dio->orig_end_io;
2890
5222595d 2891 kvfree(dio);
02b16d0a
CY		 2892
2893 bio_endio(bio);
2894}
2895
2896static void f2fs_dio_submit_bio(struct bio *bio, struct inode *inode,
2897 loff_t file_offset)
2898{
2899 struct f2fs_private_dio *dio;
2900 bool write = (bio_op(bio) == REQ_OP_WRITE);
02b16d0a
CY		 2901
2902 dio = f2fs_kzalloc(F2FS_I_SB(inode),
2903 sizeof(struct f2fs_private_dio), GFP_NOFS);
8e114038 2904 if (!dio)
02b16d0a 2905 goto out;
02b16d0a
CY		 2906
2907 dio->inode = inode;
2908 dio->orig_end_io = bio->bi_end_io;
2909 dio->orig_private = bio->bi_private;
2910 dio->write = write;
2911
2912 bio->bi_end_io = f2fs_dio_end_io;
2913 bio->bi_private = dio;
2914
2915 inc_page_count(F2FS_I_SB(inode),
2916 write ? F2FS_DIO_WRITE : F2FS_DIO_READ);
2917
2918 submit_bio(bio);
2919 return;
2920out:
2921 bio->bi_status = BLK_STS_IOERR;
2922 bio_endio(bio);
2923}
2924
c8b8e32d 2925static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
eb47b800 2926{
b439b103 2927 struct address_space *mapping = iocb->ki_filp->f_mapping;
3aab8f82 2928 struct inode *inode = mapping->host;
0cdd3195 2929 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
f847c699 2930 struct f2fs_inode_info *fi = F2FS_I(inode);
3aab8f82 2931 size_t count = iov_iter_count(iter);
c8b8e32d 2932 loff_t offset = iocb->ki_pos;
82e0a5aa 2933 int rw = iov_iter_rw(iter);
3aab8f82 2934 int err;
0cdd3195 2935 enum rw_hint hint = iocb->ki_hint;
63189b78 2936 int whint_mode = F2FS_OPTION(sbi).whint_mode;
f847c699 2937 bool do_opu;
944fcfc1 2938
b439b103 2939 err = check_direct_IO(inode, iter, offset);
b9d777b8 2940 if (err)
8a56dd96 2941 return err < 0 ? err : 0;
9ffe0fb5 2942
f847c699 2943 if (f2fs_force_buffered_io(inode, iocb, iter))
36abef4e 2944 return 0;
fcc85a4d 2945
f847c699
CY		 2946 do_opu = allow_outplace_dio(inode, iocb, iter);
2947
5302fb00 2948 trace_f2fs_direct_IO_enter(inode, offset, count, rw);
70407fad 2949
0cdd3195
HL		 2950 if (rw == WRITE && whint_mode == WHINT_MODE_OFF)
2951 iocb->ki_hint = WRITE_LIFE_NOT_SET;
2952
f847c699
CY		 2953 if (iocb->ki_flags & IOCB_NOWAIT) {
2954 if (!down_read_trylock(&fi->i_gc_rwsem[rw])) {
2955 iocb->ki_hint = hint;
2956 err = -EAGAIN;
2957 goto out;
2958 }
2959 if (do_opu && !down_read_trylock(&fi->i_gc_rwsem[READ])) {
2960 up_read(&fi->i_gc_rwsem[rw]);
b91050a8
HL		 2961 iocb->ki_hint = hint;
2962 err = -EAGAIN;
2963 goto out;
2964 }
f847c699
CY		 2965 } else {
2966 down_read(&fi->i_gc_rwsem[rw]);
2967 if (do_opu)
2968 down_read(&fi->i_gc_rwsem[READ]);
b91050a8
HL		 2969 }
2970
02b16d0a 2971 err = __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev,
f9d6d059
CY		 2972 iter, rw == WRITE ? get_data_block_dio_write :
2973 get_data_block_dio, NULL, f2fs_dio_submit_bio,
02b16d0a 2974 DIO_LOCKING | DIO_SKIP_HOLES);
f847c699
CY		 2975
2976 if (do_opu)
2977 up_read(&fi->i_gc_rwsem[READ]);
2978
2979 up_read(&fi->i_gc_rwsem[rw]);
82e0a5aa
CY		 2980
2981 if (rw == WRITE) {
0cdd3195
HL		 2982 if (whint_mode == WHINT_MODE_OFF)
2983 iocb->ki_hint = hint;
b0af6d49
CY		 2984 if (err > 0) {
2985 f2fs_update_iostat(F2FS_I_SB(inode), APP_DIRECT_IO,
2986 err);
f847c699
CY		 2987 if (!do_opu)
2988 set_inode_flag(inode, FI_UPDATE_WRITE);
b0af6d49 2989 } else if (err < 0) {
6bfc4919 2990 f2fs_write_failed(mapping, offset + count);
b0af6d49 2991 }
6bfc4919 2992 }
70407fad 2993
b91050a8 2994out:
5302fb00 2995 trace_f2fs_direct_IO_exit(inode, offset, count, rw, err);
70407fad 2996
3aab8f82 2997 return err;
eb47b800
JK		 2998}
2999
487261f3
CY		 3000void f2fs_invalidate_page(struct page *page, unsigned int offset,
3001 unsigned int length)
eb47b800
JK		 3002{
3003 struct inode *inode = page->mapping->host;
487261f3 3004 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
a7ffdbe2 3005
487261f3 3006 if (inode->i_ino >= F2FS_ROOT_INO(sbi) &&
09cbfeaf 3007 (offset % PAGE_SIZE || length != PAGE_SIZE))
a7ffdbe2
JK		 3008 return;
3009
487261f3 3010 if (PageDirty(page)) {
933439c8 3011 if (inode->i_ino == F2FS_META_INO(sbi)) {
487261f3 3012 dec_page_count(sbi, F2FS_DIRTY_META);
933439c8 3013 } else if (inode->i_ino == F2FS_NODE_INO(sbi)) {
487261f3 3014 dec_page_count(sbi, F2FS_DIRTY_NODES);
933439c8 3015 } else {
487261f3 3016 inode_dec_dirty_pages(inode);
4d57b86d 3017 f2fs_remove_dirty_inode(inode);
933439c8 3018 }
487261f3 3019 }
decd36b6 3020
2baf0781
CY		 3021 clear_cold_data(page);
3022
decd36b6 3023 if (IS_ATOMIC_WRITTEN_PAGE(page))
4d57b86d 3024 return f2fs_drop_inmem_page(inode, page);
decd36b6 3025
240a5915 3026 f2fs_clear_page_private(page);
eb47b800
JK		 3027}
3028
487261f3 3029int f2fs_release_page(struct page *page, gfp_t wait)
eb47b800 3030{
f68daeeb
JK		 3031 /* If this is dirty page, keep PagePrivate */
3032 if (PageDirty(page))
3033 return 0;
3034
decd36b6
CY		 3035 /* This is atomic written page, keep Private */
3036 if (IS_ATOMIC_WRITTEN_PAGE(page))
3037 return 0;
3038
2baf0781 3039 clear_cold_data(page);
240a5915 3040 f2fs_clear_page_private(page);
c3850aa1 3041 return 1;
eb47b800
JK		 3042}
3043
3044static int f2fs_set_data_page_dirty(struct page *page)
3045{
4969c06a 3046 struct inode *inode = page_file_mapping(page)->host;
eb47b800 3047
26c6b887
JK		 3048 trace_f2fs_set_page_dirty(page, DATA);
3049
237c0790
JK		 3050 if (!PageUptodate(page))
3051 SetPageUptodate(page);
4969c06a
JK		 3052 if (PageSwapCache(page))
3053 return __set_page_dirty_nobuffers(page);
34ba94ba 3054
5fe45743 3055 if (f2fs_is_atomic_file(inode) && !f2fs_is_commit_atomic_write(inode)) {
decd36b6 3056 if (!IS_ATOMIC_WRITTEN_PAGE(page)) {
4d57b86d 3057 f2fs_register_inmem_page(inode, page);
decd36b6
CY		 3058 return 1;
3059 }
3060 /*
3061 * Previously, this page has been registered, we just
3062 * return here.
3063 */
3064 return 0;
34ba94ba
JK		 3065 }
3066
eb47b800 3067 if (!PageDirty(page)) {
b87078ad 3068 __set_page_dirty_nobuffers(page);
4d57b86d 3069 f2fs_update_dirty_page(inode, page);
eb47b800
JK		 3070 return 1;
3071 }
3072 return 0;
3073}
3074
c01e54b7
JK		 3075static sector_t f2fs_bmap(struct address_space *mapping, sector_t block)
3076{
454ae7e5
CY		 3077 struct inode *inode = mapping->host;
3078
1d373a0e
JK		 3079 if (f2fs_has_inline_data(inode))
3080 return 0;
3081
3082 /* make sure allocating whole blocks */
3083 if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
3084 filemap_write_and_wait(mapping);
3085
e2b4e2bc 3086 return generic_block_bmap(mapping, block, get_data_block_bmap);
429511cd
CY		 3087}
3088
5b7a487c
WG		 3089#ifdef CONFIG_MIGRATION
3090#include <linux/migrate.h>
3091
3092int f2fs_migrate_page(struct address_space *mapping,
3093 struct page *newpage, struct page *page, enum migrate_mode mode)
3094{
3095 int rc, extra_count;
3096 struct f2fs_inode_info *fi = F2FS_I(mapping->host);
3097 bool atomic_written = IS_ATOMIC_WRITTEN_PAGE(page);
3098
3099 BUG_ON(PageWriteback(page));
3100
3101 /* migrating an atomic written page is safe with the inmem_lock hold */
ff1048e7
JK		 3102 if (atomic_written) {
3103 if (mode != MIGRATE_SYNC)
3104 return -EBUSY;
3105 if (!mutex_trylock(&fi->inmem_lock))
3106 return -EAGAIN;
3107 }
5b7a487c 3108
240a5915
CY		 3109 /* one extra reference was held for atomic_write page */
3110 extra_count = atomic_written ? 1 : 0;
5b7a487c 3111 rc = migrate_page_move_mapping(mapping, newpage,
37109694 3112 page, extra_count);
5b7a487c
WG		 3113 if (rc != MIGRATEPAGE_SUCCESS) {
3114 if (atomic_written)
3115 mutex_unlock(&fi->inmem_lock);
3116 return rc;
3117 }
3118
3119 if (atomic_written) {
3120 struct inmem_pages *cur;
3121 list_for_each_entry(cur, &fi->inmem_pages, list)
3122 if (cur->page == page) {
3123 cur->page = newpage;
3124 break;
3125 }
3126 mutex_unlock(&fi->inmem_lock);
3127 put_page(page);
3128 get_page(newpage);
3129 }
3130
240a5915
CY		 3131 if (PagePrivate(page)) {
3132 f2fs_set_page_private(newpage, page_private(page));
3133 f2fs_clear_page_private(page);
3134 }
5b7a487c 3135
2916ecc0
JG		 3136 if (mode != MIGRATE_SYNC_NO_COPY)
3137 migrate_page_copy(newpage, page);
3138 else
3139 migrate_page_states(newpage, page);
5b7a487c
WG		 3140
3141 return MIGRATEPAGE_SUCCESS;
3142}
3143#endif
3144
4969c06a
JK		 3145#ifdef CONFIG_SWAP
3146/* Copied from generic_swapfile_activate() to check any holes */
3147static int check_swap_activate(struct file *swap_file, unsigned int max)
3148{
3149 struct address_space *mapping = swap_file->f_mapping;
3150 struct inode *inode = mapping->host;
3151 unsigned blocks_per_page;
3152 unsigned long page_no;
3153 unsigned blkbits;
3154 sector_t probe_block;
3155 sector_t last_block;
3156 sector_t lowest_block = -1;
3157 sector_t highest_block = 0;
3158
3159 blkbits = inode->i_blkbits;
3160 blocks_per_page = PAGE_SIZE >> blkbits;
3161
3162 /*
3163 * Map all the blocks into the extent list. This code doesn't try
3164 * to be very smart.
3165 */
3166 probe_block = 0;
3167 page_no = 0;
3168 last_block = i_size_read(inode) >> blkbits;
3169 while ((probe_block + blocks_per_page) <= last_block && page_no < max) {
3170 unsigned block_in_page;
3171 sector_t first_block;
3172
3173 cond_resched();
3174
3175 first_block = bmap(inode, probe_block);
3176 if (first_block == 0)
3177 goto bad_bmap;
3178
3179 /*
3180 * It must be PAGE_SIZE aligned on-disk
3181 */
3182 if (first_block & (blocks_per_page - 1)) {
3183 probe_block++;
3184 goto reprobe;
3185 }
3186
3187 for (block_in_page = 1; block_in_page < blocks_per_page;
3188 block_in_page++) {
3189 sector_t block;
3190
3191 block = bmap(inode, probe_block + block_in_page);
3192 if (block == 0)
3193 goto bad_bmap;
3194 if (block != first_block + block_in_page) {
3195 /* Discontiguity */
3196 probe_block++;
3197 goto reprobe;
3198 }
3199 }
3200
3201 first_block >>= (PAGE_SHIFT - blkbits);
3202 if (page_no) { /* exclude the header page */
3203 if (first_block < lowest_block)
3204 lowest_block = first_block;
3205 if (first_block > highest_block)
3206 highest_block = first_block;
3207 }
3208
3209 page_no++;
3210 probe_block += blocks_per_page;
3211reprobe:
3212 continue;
3213 }
3214 return 0;
3215
3216bad_bmap:
3217 pr_err("swapon: swapfile has holes\n");
3218 return -EINVAL;
3219}
3220
3221static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
3222 sector_t *span)
3223{
3224 struct inode *inode = file_inode(file);
3225 int ret;
3226
3227 if (!S_ISREG(inode->i_mode))
3228 return -EINVAL;
3229
3230 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3231 return -EROFS;
3232
3233 ret = f2fs_convert_inline_inode(inode);
3234 if (ret)
3235 return ret;
3236
3237 ret = check_swap_activate(file, sis->max);
3238 if (ret)
3239 return ret;
3240
3241 set_inode_flag(inode, FI_PIN_FILE);
3242 f2fs_precache_extents(inode);
3243 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3244 return 0;
3245}
3246
3247static void f2fs_swap_deactivate(struct file *file)
3248{
3249 struct inode *inode = file_inode(file);
3250
3251 clear_inode_flag(inode, FI_PIN_FILE);
3252}
3253#else
3254static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
3255 sector_t *span)
3256{
3257 return -EOPNOTSUPP;
3258}
3259
3260static void f2fs_swap_deactivate(struct file *file)
3261{
3262}
3263#endif
3264
eb47b800
JK		 3265const struct address_space_operations f2fs_dblock_aops = {
3266 .readpage = f2fs_read_data_page,
3267 .readpages = f2fs_read_data_pages,
3268 .writepage = f2fs_write_data_page,
3269 .writepages = f2fs_write_data_pages,
3270 .write_begin = f2fs_write_begin,
a1dd3c13 3271 .write_end = f2fs_write_end,
eb47b800 3272 .set_page_dirty = f2fs_set_data_page_dirty,
487261f3
CY		 3273 .invalidatepage = f2fs_invalidate_page,
3274 .releasepage = f2fs_release_page,
eb47b800 3275 .direct_IO = f2fs_direct_IO,
c01e54b7 3276 .bmap = f2fs_bmap,
4969c06a
JK		 3277 .swap_activate = f2fs_swap_activate,
3278 .swap_deactivate = f2fs_swap_deactivate,
5b7a487c
WG		 3279#ifdef CONFIG_MIGRATION
3280 .migratepage = f2fs_migrate_page,
3281#endif
eb47b800 3282};
6dbb1796 3283
5ec2d99d 3284void f2fs_clear_page_cache_dirty_tag(struct page *page)
aec2f729
CY		 3285{
3286 struct address_space *mapping = page_mapping(page);
3287 unsigned long flags;
3288
3289 xa_lock_irqsave(&mapping->i_pages, flags);
5ec2d99d 3290 __xa_clear_mark(&mapping->i_pages, page_index(page),
aec2f729
CY		 3291 PAGECACHE_TAG_DIRTY);
3292 xa_unlock_irqrestore(&mapping->i_pages, flags);
3293}
3294
6dbb1796
EB		 3295int __init f2fs_init_post_read_processing(void)
3296{
95ae251f
EB		 3297 bio_post_read_ctx_cache =
3298 kmem_cache_create("f2fs_bio_post_read_ctx",
3299 sizeof(struct bio_post_read_ctx), 0, 0, NULL);
6dbb1796
EB		 3300 if (!bio_post_read_ctx_cache)
3301 goto fail;
3302 bio_post_read_ctx_pool =
3303 mempool_create_slab_pool(NUM_PREALLOC_POST_READ_CTXS,
3304 bio_post_read_ctx_cache);
3305 if (!bio_post_read_ctx_pool)
3306 goto fail_free_cache;
3307 return 0;
3308
3309fail_free_cache:
3310 kmem_cache_destroy(bio_post_read_ctx_cache);
3311fail:
3312 return -ENOMEM;
3313}
3314
0b20fcec 3315void f2fs_destroy_post_read_processing(void)
6dbb1796
EB		 3316{
3317 mempool_destroy(bio_post_read_ctx_pool);
3318 kmem_cache_destroy(bio_post_read_ctx_cache);
3319}
0b20fcec
CY		 3320
3321int __init f2fs_init_bio_entry_cache(void)
3322{
3323 bio_entry_slab = f2fs_kmem_cache_create("bio_entry_slab",
3324 sizeof(struct bio_entry));
3325 if (!bio_entry_slab)
3326 return -ENOMEM;
3327 return 0;
3328}
3329
3330void __exit f2fs_destroy_bio_entry_cache(void)
3331{
3332 kmem_cache_destroy(bio_entry_slab);
3333}
Michael's Linux tree.