projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: Kconfig: clean up config options about compression
[linux-2.6-block.git] / fs / f2fs / data.c
CommitLineData
7c1a000d 1// SPDX-License-Identifier: GPL-2.0
0a8165d7 2/*
eb47b800
JK		 3 * fs/f2fs/data.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
eb47b800
JK		 7 */
8#include <linux/fs.h>
9#include <linux/f2fs_fs.h>
10#include <linux/buffer_head.h>
11#include <linux/mpage.h>
12#include <linux/writeback.h>
13#include <linux/backing-dev.h>
8f46dcae 14#include <linux/pagevec.h>
eb47b800
JK		 15#include <linux/blkdev.h>
16#include <linux/bio.h>
27aacd28 17#include <linux/blk-crypto.h>
4969c06a 18#include <linux/swap.h>
690e4a3e 19#include <linux/prefetch.h>
e2e40f2c 20#include <linux/uio.h>
f1e88660 21#include <linux/cleancache.h>
174cd4b1 22#include <linux/sched/signal.h>
10c5db28 23#include <linux/fiemap.h>
eb47b800
JK		 24
25#include "f2fs.h"
26#include "node.h"
27#include "segment.h"
848753aa 28#include <trace/events/f2fs.h>
eb47b800 29
6dbb1796
EB		 30#define NUM_PREALLOC_POST_READ_CTXS 128
31
32static struct kmem_cache *bio_post_read_ctx_cache;
0b20fcec 33static struct kmem_cache *bio_entry_slab;
6dbb1796 34static mempool_t *bio_post_read_ctx_pool;
f543805f
CY		 35static struct bio_set f2fs_bioset;
36
37#define F2FS_BIO_POOL_SIZE NR_CURSEG_TYPE
38
39int __init f2fs_init_bioset(void)
40{
41 if (bioset_init(&f2fs_bioset, F2FS_BIO_POOL_SIZE,
42 0, BIOSET_NEED_BVECS))
43 return -ENOMEM;
44 return 0;
45}
46
47void f2fs_destroy_bioset(void)
48{
49 bioset_exit(&f2fs_bioset);
50}
51
36951b38
CY		 52static bool __is_cp_guaranteed(struct page *page)
53{
54 struct address_space *mapping = page->mapping;
55 struct inode *inode;
56 struct f2fs_sb_info *sbi;
57
58 if (!mapping)
59 return false;
60
61 inode = mapping->host;
62 sbi = F2FS_I_SB(inode);
63
64 if (inode->i_ino == F2FS_META_INO(sbi) ||
a87aff1d 65 inode->i_ino == F2FS_NODE_INO(sbi) ||
b763f3be
CY		 66 S_ISDIR(inode->i_mode))
67 return true;
68
69 if (f2fs_is_compressed_page(page))
70 return false;
71 if ((S_ISREG(inode->i_mode) &&
af033b2a 72 (f2fs_is_atomic_file(inode) || IS_NOQUOTA(inode))) ||
b763f3be 73 page_private_gcing(page))
36951b38
CY		 74 return true;
75 return false;
76}
77
5f9abab4
JK		 78static enum count_type __read_io_type(struct page *page)
79{
4969c06a 80 struct address_space *mapping = page_file_mapping(page);
5f9abab4
JK		 81
82 if (mapping) {
83 struct inode *inode = mapping->host;
84 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
85
86 if (inode->i_ino == F2FS_META_INO(sbi))
87 return F2FS_RD_META;
88
89 if (inode->i_ino == F2FS_NODE_INO(sbi))
90 return F2FS_RD_NODE;
91 }
92 return F2FS_RD_DATA;
93}
94
6dbb1796
EB		 95/* postprocessing steps for read bios */
96enum bio_post_read_step {
7f59b277
EB		 97#ifdef CONFIG_FS_ENCRYPTION
98 STEP_DECRYPT = 1 << 0,
99#else
100 STEP_DECRYPT = 0, /* compile out the decryption-related code */
101#endif
102#ifdef CONFIG_F2FS_FS_COMPRESSION
103 STEP_DECOMPRESS = 1 << 1,
104#else
105 STEP_DECOMPRESS = 0, /* compile out the decompression-related code */
106#endif
107#ifdef CONFIG_FS_VERITY
108 STEP_VERITY = 1 << 2,
109#else
110 STEP_VERITY = 0, /* compile out the verity-related code */
111#endif
6dbb1796
EB		 112};
113
114struct bio_post_read_ctx {
115 struct bio *bio;
4c8ff709 116 struct f2fs_sb_info *sbi;
6dbb1796 117 struct work_struct work;
6dbb1796 118 unsigned int enabled_steps;
4931e0c9 119 block_t fs_blkaddr;
6dbb1796
EB		 120};
121
7f59b277 122static void f2fs_finish_read_bio(struct bio *bio)
93dfe2ac 123{
6dbb1796 124 struct bio_vec *bv;
6dc4f100 125 struct bvec_iter_all iter_all;
93dfe2ac 126
7f59b277
EB		 127 /*
128 * Update and unlock the bio's pagecache pages, and put the
129 * decompression context for any compressed pages.
130 */
2b070cfe 131 bio_for_each_segment_all(bv, bio, iter_all) {
7f59b277 132 struct page *page = bv->bv_page;
6dbb1796 133
7f59b277
EB		 134 if (f2fs_is_compressed_page(page)) {
135 if (bio->bi_status)
6ce19aff 136 f2fs_end_read_compressed_page(page, true, 0);
7f59b277 137 f2fs_put_page_dic(page);
4c8ff709
CY		 138 continue;
139 }
4c8ff709 140
7f59b277 141 /* PG_error was set if decryption or verity failed. */
6dbb1796
EB		 142 if (bio->bi_status || PageError(page)) {
143 ClearPageUptodate(page);
fb7d70db
JK		 144 /* will re-read again later */
145 ClearPageError(page);
6dbb1796
EB		 146 } else {
147 SetPageUptodate(page);
148 }
5f9abab4 149 dec_page_count(F2FS_P_SB(page), __read_io_type(page));
6dbb1796
EB		 150 unlock_page(page);
151 }
79bbefb1 152
7f59b277
EB		 153 if (bio->bi_private)
154 mempool_free(bio->bi_private, bio_post_read_ctx_pool);
155 bio_put(bio);
4c8ff709 156}
4c8ff709 157
7f59b277 158static void f2fs_verify_bio(struct work_struct *work)
6dbb1796
EB		 159{
160 struct bio_post_read_ctx *ctx =
161 container_of(work, struct bio_post_read_ctx, work);
644c8c92 162 struct bio *bio = ctx->bio;
7f59b277 163 bool may_have_compressed_pages = (ctx->enabled_steps & STEP_DECOMPRESS);
644c8c92
EB		 164
165 /*
166 * fsverity_verify_bio() may call readpages() again, and while verity
7f59b277
EB		 167 * will be disabled for this, decryption and/or decompression may still
168 * be needed, resulting in another bio_post_read_ctx being allocated.
169 * So to prevent deadlocks we need to release the current ctx to the
170 * mempool first. This assumes that verity is the last post-read step.
644c8c92
EB		 171 */
172 mempool_free(ctx, bio_post_read_ctx_pool);
173 bio->bi_private = NULL;
6dbb1796 174
7f59b277
EB		 175 /*
176 * Verify the bio's pages with fs-verity. Exclude compressed pages,
177 * as those were handled separately by f2fs_end_read_compressed_page().
178 */
179 if (may_have_compressed_pages) {
180 struct bio_vec *bv;
181 struct bvec_iter_all iter_all;
182
183 bio_for_each_segment_all(bv, bio, iter_all) {
184 struct page *page = bv->bv_page;
185
186 if (!f2fs_is_compressed_page(page) &&
187 !PageError(page) && !fsverity_verify_page(page))
188 SetPageError(page);
189 }
190 } else {
191 fsverity_verify_bio(bio);
4c8ff709 192 }
6dbb1796 193
7f59b277 194 f2fs_finish_read_bio(bio);
6dbb1796
EB		 195}
196
7f59b277
EB		 197/*
198 * If the bio's data needs to be verified with fs-verity, then enqueue the
199 * verity work for the bio. Otherwise finish the bio now.
200 *
201 * Note that to avoid deadlocks, the verity work can't be done on the
202 * decryption/decompression workqueue. This is because verifying the data pages
203 * can involve reading verity metadata pages from the file, and these verity
204 * metadata pages may be encrypted and/or compressed.
205 */
206static void f2fs_verify_and_finish_bio(struct bio *bio)
95ae251f 207{
7f59b277 208 struct bio_post_read_ctx *ctx = bio->bi_private;
4c8ff709 209
7f59b277
EB		 210 if (ctx && (ctx->enabled_steps & STEP_VERITY)) {
211 INIT_WORK(&ctx->work, f2fs_verify_bio);
4c8ff709 212 fsverity_enqueue_verify_work(&ctx->work);
7f59b277
EB		 213 } else {
214 f2fs_finish_read_bio(bio);
4c8ff709 215 }
4c8ff709 216}
95ae251f 217
7f59b277
EB		 218/*
219 * Handle STEP_DECOMPRESS by decompressing any compressed clusters whose last
220 * remaining page was read by @ctx->bio.
221 *
222 * Note that a bio may span clusters (even a mix of compressed and uncompressed
223 * clusters) or be for just part of a cluster. STEP_DECOMPRESS just indicates
224 * that the bio includes at least one compressed page. The actual decompression
225 * is done on a per-cluster basis, not a per-bio basis.
226 */
227static void f2fs_handle_step_decompress(struct bio_post_read_ctx *ctx)
6dbb1796 228{
7f59b277
EB		 229 struct bio_vec *bv;
230 struct bvec_iter_all iter_all;
231 bool all_compressed = true;
4931e0c9 232 block_t blkaddr = ctx->fs_blkaddr;
4c8ff709 233
7f59b277
EB		 234 bio_for_each_segment_all(bv, ctx->bio, iter_all) {
235 struct page *page = bv->bv_page;
4c8ff709 236
7f59b277
EB		 237 /* PG_error was set if decryption failed. */
238 if (f2fs_is_compressed_page(page))
6ce19aff
CY		 239 f2fs_end_read_compressed_page(page, PageError(page),
240 blkaddr);
7f59b277
EB		 241 else
242 all_compressed = false;
6ce19aff
CY		 243
244 blkaddr++;
6dbb1796 245 }
4c8ff709 246
7f59b277
EB		 247 /*
248 * Optimization: if all the bio's pages are compressed, then scheduling
249 * the per-bio verity work is unnecessary, as verity will be fully
250 * handled at the compression cluster level.
251 */
252 if (all_compressed)
253 ctx->enabled_steps &= ~STEP_VERITY;
6dbb1796
EB		 254}
255
7f59b277 256static void f2fs_post_read_work(struct work_struct *work)
6dbb1796 257{
7f59b277
EB		 258 struct bio_post_read_ctx *ctx =
259 container_of(work, struct bio_post_read_ctx, work);
260
261 if (ctx->enabled_steps & STEP_DECRYPT)
262 fscrypt_decrypt_bio(ctx->bio);
263
264 if (ctx->enabled_steps & STEP_DECOMPRESS)
265 f2fs_handle_step_decompress(ctx);
266
267 f2fs_verify_and_finish_bio(ctx->bio);
6dbb1796
EB		 268}
269
270static void f2fs_read_end_io(struct bio *bio)
271{
c45d6002 272 struct f2fs_sb_info *sbi = F2FS_P_SB(bio_first_page_all(bio));
7f59b277 273 struct bio_post_read_ctx *ctx = bio->bi_private;
c45d6002
CY		 274
275 if (time_to_inject(sbi, FAULT_READ_IO)) {
276 f2fs_show_injection_info(sbi, FAULT_READ_IO);
4e4cbee9 277 bio->bi_status = BLK_STS_IOERR;
55523519 278 }
8b038c70 279
7f59b277
EB		 280 if (bio->bi_status) {
281 f2fs_finish_read_bio(bio);
6dbb1796 282 return;
f1e88660 283 }
6dbb1796 284
7f59b277
EB		 285 if (ctx && (ctx->enabled_steps & (STEP_DECRYPT | STEP_DECOMPRESS))) {
286 INIT_WORK(&ctx->work, f2fs_post_read_work);
287 queue_work(ctx->sbi->post_read_wq, &ctx->work);
288 } else {
289 f2fs_verify_and_finish_bio(bio);
290 }
f1e88660
JK		 291}
292
4246a0b6 293static void f2fs_write_end_io(struct bio *bio)
93dfe2ac 294{
1b1f559f 295 struct f2fs_sb_info *sbi = bio->bi_private;
f568849e 296 struct bio_vec *bvec;
6dc4f100 297 struct bvec_iter_all iter_all;
93dfe2ac 298
6f5c2ed0 299 if (time_to_inject(sbi, FAULT_WRITE_IO)) {
c45d6002 300 f2fs_show_injection_info(sbi, FAULT_WRITE_IO);
6f5c2ed0
CY		 301 bio->bi_status = BLK_STS_IOERR;
302 }
303
2b070cfe 304 bio_for_each_segment_all(bvec, bio, iter_all) {
93dfe2ac 305 struct page *page = bvec->bv_page;
36951b38 306 enum count_type type = WB_DATA_TYPE(page);
93dfe2ac 307
b763f3be
CY		 308 if (page_private_dummy(page)) {
309 clear_page_private_dummy(page);
0a595eba
JK		 310 unlock_page(page);
311 mempool_free(page, sbi->write_io_dummy);
312
4e4cbee9 313 if (unlikely(bio->bi_status))
0a595eba
JK		 314 f2fs_stop_checkpoint(sbi, true);
315 continue;
316 }
317
d2d0727b 318 fscrypt_finalize_bounce_page(&page);
4375a336 319
4c8ff709
CY		 320#ifdef CONFIG_F2FS_FS_COMPRESSION
321 if (f2fs_is_compressed_page(page)) {
322 f2fs_compress_write_end_io(bio, page);
323 continue;
324 }
325#endif
326
4e4cbee9 327 if (unlikely(bio->bi_status)) {
5114a97a 328 mapping_set_error(page->mapping, -EIO);
b1ca321d
JK		 329 if (type == F2FS_WB_CP_DATA)
330 f2fs_stop_checkpoint(sbi, true);
93dfe2ac 331 }
7dff55d2
YH		 332
333 f2fs_bug_on(sbi, page->mapping == NODE_MAPPING(sbi) &&
334 page->index != nid_of_node(page));
335
36951b38 336 dec_page_count(sbi, type);
50fa53ec
CY		 337 if (f2fs_in_warm_node_list(sbi, page))
338 f2fs_del_fsync_node_entry(sbi, page);
b763f3be 339 clear_page_private_gcing(page);
93dfe2ac 340 end_page_writeback(page);
f568849e 341 }
36951b38 342 if (!get_pages(sbi, F2FS_WB_CP_DATA) &&
f5730184 343 wq_has_sleeper(&sbi->cp_wait))
93dfe2ac
JK		 344 wake_up(&sbi->cp_wait);
345
346 bio_put(bio);
347}
348
3c62be17
JK		 349struct block_device *f2fs_target_device(struct f2fs_sb_info *sbi,
350 block_t blk_addr, struct bio *bio)
351{
352 struct block_device *bdev = sbi->sb->s_bdev;
353 int i;
354
0916878d
DLM		 355 if (f2fs_is_multi_device(sbi)) {
356 for (i = 0; i < sbi->s_ndevs; i++) {
357 if (FDEV(i).start_blk <= blk_addr &&
358 FDEV(i).end_blk >= blk_addr) {
359 blk_addr -= FDEV(i).start_blk;
360 bdev = FDEV(i).bdev;
361 break;
362 }
3c62be17
JK		 363 }
364 }
365 if (bio) {
74d46992 366 bio_set_dev(bio, bdev);
3c62be17
JK		 367 bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr);
368 }
369 return bdev;
370}
371
372int f2fs_target_device_index(struct f2fs_sb_info *sbi, block_t blkaddr)
373{
374 int i;
375
0916878d
DLM		 376 if (!f2fs_is_multi_device(sbi))
377 return 0;
378
3c62be17
JK		 379 for (i = 0; i < sbi->s_ndevs; i++)
380 if (FDEV(i).start_blk <= blkaddr && FDEV(i).end_blk >= blkaddr)
381 return i;
382 return 0;
383}
384
b757f6ed 385static struct bio *__bio_alloc(struct f2fs_io_info *fio, int npages)
940a6d34 386{
b757f6ed 387 struct f2fs_sb_info *sbi = fio->sbi;
940a6d34
GZ		 388 struct bio *bio;
389
67883ade 390 bio = bio_alloc_bioset(GFP_NOIO, npages, &f2fs_bioset);
940a6d34 391
b757f6ed
CY		 392 f2fs_target_device(sbi, fio->new_blkaddr, bio);
393 if (is_read_io(fio->op)) {
0cdd3195
HL		 394 bio->bi_end_io = f2fs_read_end_io;
395 bio->bi_private = NULL;
396 } else {
397 bio->bi_end_io = f2fs_write_end_io;
398 bio->bi_private = sbi;
b757f6ed
CY		 399 bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi,
400 fio->type, fio->temp);
0cdd3195 401 }
b757f6ed
CY		 402 if (fio->io_wbc)
403 wbc_init_bio(fio->io_wbc, bio);
940a6d34
GZ		 404
405 return bio;
406}
407
27aacd28
ST		 408static void f2fs_set_bio_crypt_ctx(struct bio *bio, const struct inode *inode,
409 pgoff_t first_idx,
410 const struct f2fs_io_info *fio,
411 gfp_t gfp_mask)
412{
413 /*
414 * The f2fs garbage collector sets ->encrypted_page when it wants to
415 * read/write raw data without encryption.
416 */
417 if (!fio || !fio->encrypted_page)
418 fscrypt_set_bio_crypt_ctx(bio, inode, first_idx, gfp_mask);
419}
420
421static bool f2fs_crypt_mergeable_bio(struct bio *bio, const struct inode *inode,
422 pgoff_t next_idx,
423 const struct f2fs_io_info *fio)
424{
425 /*
426 * The f2fs garbage collector sets ->encrypted_page when it wants to
427 * read/write raw data without encryption.
428 */
429 if (fio && fio->encrypted_page)
430 return !bio_has_crypt_ctx(bio);
431
432 return fscrypt_mergeable_bio(bio, inode, next_idx);
433}
434
4fc29c1a
LT		 435static inline void __submit_bio(struct f2fs_sb_info *sbi,
436 struct bio *bio, enum page_type type)
f5730184 437{
4fc29c1a 438 if (!is_read_io(bio_op(bio))) {
0a595eba
JK		 439 unsigned int start;
440
0a595eba
JK		 441 if (type != DATA && type != NODE)
442 goto submit_io;
443
b0332a0f 444 if (f2fs_lfs_mode(sbi) && current->plug)
3bb09a0e
TY		 445 blk_finish_plug(current->plug);
446
39f71b7e 447 if (!F2FS_IO_ALIGNED(sbi))
8223ecc4
CY		 448 goto submit_io;
449
0a595eba
JK		 450 start = bio->bi_iter.bi_size >> F2FS_BLKSIZE_BITS;
451 start %= F2FS_IO_SIZE(sbi);
452
453 if (start == 0)
454 goto submit_io;
455
456 /* fill dummy pages */
457 for (; start < F2FS_IO_SIZE(sbi); start++) {
458 struct page *page =
459 mempool_alloc(sbi->write_io_dummy,
bc73a4b2 460 GFP_NOIO | __GFP_NOFAIL);
0a595eba
JK		 461 f2fs_bug_on(sbi, !page);
462
0a595eba 463 lock_page(page);
b763f3be
CY		 464
465 zero_user_segment(page, 0, PAGE_SIZE);
466 set_page_private_dummy(page);
467
0a595eba
JK		 468 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE)
469 f2fs_bug_on(sbi, 1);
470 }
471 /*
472 * In the NODE case, we lose next block address chain. So, we
473 * need to do checkpoint in f2fs_sync_file.
474 */
475 if (type == NODE)
476 set_sbi_flag(sbi, SBI_NEED_CP);
19a5f5e2 477 }
0a595eba 478submit_io:
554b5125
JK		 479 if (is_read_io(bio_op(bio)))
480 trace_f2fs_submit_read_bio(sbi->sb, type, bio);
481 else
482 trace_f2fs_submit_write_bio(sbi->sb, type, bio);
4e49ea4a 483 submit_bio(bio);
f5730184
JK		 484}
485
4c8ff709
CY		 486void f2fs_submit_bio(struct f2fs_sb_info *sbi,
487 struct bio *bio, enum page_type type)
488{
489 __submit_bio(sbi, bio, type);
490}
491
32b6aba8 492static void __attach_io_flag(struct f2fs_io_info *fio)
da9953b7
JK		 493{
494 struct f2fs_sb_info *sbi = fio->sbi;
495 unsigned int temp_mask = (1 << NR_TEMP_TYPE) - 1;
32b6aba8
JK		 496 unsigned int io_flag, fua_flag, meta_flag;
497
498 if (fio->type == DATA)
499 io_flag = sbi->data_io_flag;
500 else if (fio->type == NODE)
501 io_flag = sbi->node_io_flag;
502 else
503 return;
504
505 fua_flag = io_flag & temp_mask;
506 meta_flag = (io_flag >> NR_TEMP_TYPE) & temp_mask;
507
da9953b7 508 /*
32b6aba8 509 * data/node io flag bits per temp:
da9953b7
JK		 510 * REQ_META | REQ_FUA |
511 * 5 | 4 | 3 | 2 | 1 | 0 |
512 * Cold | Warm | Hot | Cold | Warm | Hot |
513 */
da9953b7
JK		 514 if ((1 << fio->temp) & meta_flag)
515 fio->op_flags |= REQ_META;
516 if ((1 << fio->temp) & fua_flag)
517 fio->op_flags |= REQ_FUA;
518}
519
458e6197 520static void __submit_merged_bio(struct f2fs_bio_info *io)
93dfe2ac 521{
458e6197 522 struct f2fs_io_info *fio = &io->fio;
93dfe2ac
JK		 523
524 if (!io->bio)
525 return;
526
32b6aba8 527 __attach_io_flag(fio);
554b5125
JK		 528 bio_set_op_attrs(io->bio, fio->op, fio->op_flags);
529
04d328de 530 if (is_read_io(fio->op))
554b5125 531 trace_f2fs_prepare_read_bio(io->sbi->sb, fio->type, io->bio);
6a8f8ca5 532 else
554b5125 533 trace_f2fs_prepare_write_bio(io->sbi->sb, fio->type, io->bio);
04d328de 534
4fc29c1a 535 __submit_bio(io->sbi, io->bio, fio->type);
93dfe2ac
JK		 536 io->bio = NULL;
537}
538
8648de2c 539static bool __has_merged_page(struct bio *bio, struct inode *inode,
bab475c5 540 struct page *page, nid_t ino)
0fd785eb 541{
0fd785eb 542 struct bio_vec *bvec;
6dc4f100 543 struct bvec_iter_all iter_all;
0fd785eb 544
8648de2c 545 if (!bio)
0fd785eb 546 return false;
0c3a5797 547
bab475c5 548 if (!inode && !page && !ino)
0c3a5797 549 return true;
0fd785eb 550
8648de2c 551 bio_for_each_segment_all(bvec, bio, iter_all) {
4c8ff709 552 struct page *target = bvec->bv_page;
0fd785eb 553
4c8ff709 554 if (fscrypt_is_bounce_page(target)) {
d2d0727b 555 target = fscrypt_pagecache_page(target);
4c8ff709
CY		 556 if (IS_ERR(target))
557 continue;
558 }
559 if (f2fs_is_compressed_page(target)) {
560 target = f2fs_compress_control_page(target);
561 if (IS_ERR(target))
562 continue;
563 }
0fd785eb 564
0c3a5797
CY		 565 if (inode && inode == target->mapping->host)
566 return true;
bab475c5
CY		 567 if (page && page == target)
568 return true;
0c3a5797 569 if (ino && ino == ino_of_node(target))
0fd785eb 570 return true;
0fd785eb
CY		 571 }
572
0fd785eb
CY		 573 return false;
574}
575
b9109b0e 576static void __f2fs_submit_merged_write(struct f2fs_sb_info *sbi,
a912b54d 577 enum page_type type, enum temp_type temp)
93dfe2ac
JK		 578{
579 enum page_type btype = PAGE_TYPE_OF_BIO(type);
a912b54d 580 struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
93dfe2ac 581
df0f8dc0 582 down_write(&io->io_rwsem);
458e6197
JK		 583
584 /* change META to META_FLUSH in the checkpoint procedure */
585 if (type >= META_FLUSH) {
586 io->fio.type = META_FLUSH;
04d328de 587 io->fio.op = REQ_OP_WRITE;
3adc5fcb 588 io->fio.op_flags = REQ_META | REQ_PRIO | REQ_SYNC;
70fd7614 589 if (!test_opt(sbi, NOBARRIER))
7f54f51f 590 io->fio.op_flags |= REQ_PREFLUSH | REQ_FUA;
458e6197
JK		 591 }
592 __submit_merged_bio(io);
df0f8dc0 593 up_write(&io->io_rwsem);
93dfe2ac
JK		 594}
595
a912b54d 596static void __submit_merged_write_cond(struct f2fs_sb_info *sbi,
bab475c5
CY		 597 struct inode *inode, struct page *page,
598 nid_t ino, enum page_type type, bool force)
0c3a5797 599{
a912b54d 600 enum temp_type temp;
1e771e83 601 bool ret = true;
a912b54d
JK		 602
603 for (temp = HOT; temp < NR_TEMP_TYPE; temp++) {
1e771e83
YS		 604 if (!force) {
605 enum page_type btype = PAGE_TYPE_OF_BIO(type);
606 struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
a912b54d 607
1e771e83 608 down_read(&io->io_rwsem);
8648de2c 609 ret = __has_merged_page(io->bio, inode, page, ino);
1e771e83
YS		 610 up_read(&io->io_rwsem);
611 }
612 if (ret)
613 __f2fs_submit_merged_write(sbi, type, temp);
a912b54d
JK		 614
615 /* TODO: use HOT temp only for meta pages now. */
616 if (type >= META)
617 break;
618 }
0c3a5797
CY		 619}
620
b9109b0e 621void f2fs_submit_merged_write(struct f2fs_sb_info *sbi, enum page_type type)
0c3a5797 622{
adcc00f7 623 __submit_merged_write_cond(sbi, NULL, NULL, 0, type, true);
0c3a5797
CY		 624}
625
b9109b0e 626void f2fs_submit_merged_write_cond(struct f2fs_sb_info *sbi,
bab475c5
CY		 627 struct inode *inode, struct page *page,
628 nid_t ino, enum page_type type)
0c3a5797 629{
bab475c5 630 __submit_merged_write_cond(sbi, inode, page, ino, type, false);
0c3a5797
CY		 631}
632
b9109b0e 633void f2fs_flush_merged_writes(struct f2fs_sb_info *sbi)
406657dd 634{
b9109b0e
JK		 635 f2fs_submit_merged_write(sbi, DATA);
636 f2fs_submit_merged_write(sbi, NODE);
637 f2fs_submit_merged_write(sbi, META);
406657dd
CY		 638}
639
93dfe2ac
JK		 640/*
641 * Fill the locked page with data located in the block address.
771a9a71 642 * A caller needs to unlock the page on failure.
93dfe2ac 643 */
05ca3632 644int f2fs_submit_page_bio(struct f2fs_io_info *fio)
93dfe2ac 645{
93dfe2ac 646 struct bio *bio;
0b81d077
JK		 647 struct page *page = fio->encrypted_page ?
648 fio->encrypted_page : fio->page;
93dfe2ac 649
c9b60788 650 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
93770ab7
CY		 651 fio->is_por ? META_POR : (__is_meta_io(fio) ?
652 META_GENERIC : DATA_GENERIC_ENHANCE)))
10f966bb 653 return -EFSCORRUPTED;
c9b60788 654
2ace38e0 655 trace_f2fs_submit_page_bio(page, fio);
93dfe2ac
JK		 656
657 /* Allocate a new bio */
b757f6ed 658 bio = __bio_alloc(fio, 1);
93dfe2ac 659
27aacd28
ST		 660 f2fs_set_bio_crypt_ctx(bio, fio->page->mapping->host,
661 fio->page->index, fio, GFP_NOIO);
662
09cbfeaf 663 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
93dfe2ac 664 bio_put(bio);
93dfe2ac
JK		 665 return -EFAULT;
666 }
78efac53
CY		 667
668 if (fio->io_wbc && !is_read_io(fio->op))
34e51a5e 669 wbc_account_cgroup_owner(fio->io_wbc, page, PAGE_SIZE);
78efac53 670
b7b911d5 671 __attach_io_flag(fio);
04d328de 672 bio_set_op_attrs(bio, fio->op, fio->op_flags);
93dfe2ac 673
5f9abab4
JK		 674 inc_page_count(fio->sbi, is_read_io(fio->op) ?
675 __read_io_type(page): WB_DATA_TYPE(fio->page));
4c58ed07
CY		 676
677 __submit_bio(fio->sbi, bio, fio->type);
93dfe2ac
JK		 678 return 0;
679}
680
8896cbdf
CY		 681static bool page_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
682 block_t last_blkaddr, block_t cur_blkaddr)
683{
10208567
JK		 684 if (unlikely(sbi->max_io_bytes &&
685 bio->bi_iter.bi_size >= sbi->max_io_bytes))
686 return false;
8896cbdf
CY		 687 if (last_blkaddr + 1 != cur_blkaddr)
688 return false;
309dca30 689 return bio->bi_bdev == f2fs_target_device(sbi, cur_blkaddr, NULL);
8896cbdf
CY		 690}
691
692static bool io_type_is_mergeable(struct f2fs_bio_info *io,
693 struct f2fs_io_info *fio)
694{
695 if (io->fio.op != fio->op)
696 return false;
697 return io->fio.op_flags == fio->op_flags;
698}
699
700static bool io_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
701 struct f2fs_bio_info *io,
702 struct f2fs_io_info *fio,
703 block_t last_blkaddr,
704 block_t cur_blkaddr)
705{
c72db71e
CY		 706 if (F2FS_IO_ALIGNED(sbi) && (fio->type == DATA || fio->type == NODE)) {
707 unsigned int filled_blocks =
708 F2FS_BYTES_TO_BLK(bio->bi_iter.bi_size);
709 unsigned int io_size = F2FS_IO_SIZE(sbi);
710 unsigned int left_vecs = bio->bi_max_vecs - bio->bi_vcnt;
711
712 /* IOs in bio is aligned and left space of vectors is not enough */
713 if (!(filled_blocks % io_size) && left_vecs < io_size)
714 return false;
715 }
8896cbdf
CY		 716 if (!page_is_mergeable(sbi, bio, last_blkaddr, cur_blkaddr))
717 return false;
718 return io_type_is_mergeable(io, fio);
719}
720
0b20fcec
CY		 721static void add_bio_entry(struct f2fs_sb_info *sbi, struct bio *bio,
722 struct page *page, enum temp_type temp)
723{
724 struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
725 struct bio_entry *be;
726
727 be = f2fs_kmem_cache_alloc(bio_entry_slab, GFP_NOFS);
728 be->bio = bio;
729 bio_get(bio);
730
731 if (bio_add_page(bio, page, PAGE_SIZE, 0) != PAGE_SIZE)
732 f2fs_bug_on(sbi, 1);
733
734 down_write(&io->bio_list_lock);
735 list_add_tail(&be->list, &io->bio_list);
736 up_write(&io->bio_list_lock);
737}
738
739static void del_bio_entry(struct bio_entry *be)
740{
741 list_del(&be->list);
742 kmem_cache_free(bio_entry_slab, be);
743}
744
27aacd28 745static int add_ipu_page(struct f2fs_io_info *fio, struct bio **bio,
0b20fcec
CY		 746 struct page *page)
747{
27aacd28 748 struct f2fs_sb_info *sbi = fio->sbi;
0b20fcec
CY		 749 enum temp_type temp;
750 bool found = false;
751 int ret = -EAGAIN;
752
753 for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
754 struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
755 struct list_head *head = &io->bio_list;
756 struct bio_entry *be;
757
758 down_write(&io->bio_list_lock);
759 list_for_each_entry(be, head, list) {
760 if (be->bio != *bio)
761 continue;
762
763 found = true;
764
27aacd28
ST		 765 f2fs_bug_on(sbi, !page_is_mergeable(sbi, *bio,
766 *fio->last_block,
767 fio->new_blkaddr));
768 if (f2fs_crypt_mergeable_bio(*bio,
769 fio->page->mapping->host,
770 fio->page->index, fio) &&
771 bio_add_page(*bio, page, PAGE_SIZE, 0) ==
772 PAGE_SIZE) {
0b20fcec
CY		 773 ret = 0;
774 break;
775 }
776
27aacd28 777 /* page can't be merged into bio; submit the bio */
0b20fcec
CY		 778 del_bio_entry(be);
779 __submit_bio(sbi, *bio, DATA);
780 break;
781 }
782 up_write(&io->bio_list_lock);
783 }
784
785 if (ret) {
786 bio_put(*bio);
787 *bio = NULL;
788 }
789
790 return ret;
791}
792
793void f2fs_submit_merged_ipu_write(struct f2fs_sb_info *sbi,
794 struct bio **bio, struct page *page)
795{
796 enum temp_type temp;
797 bool found = false;
798 struct bio *target = bio ? *bio : NULL;
799
800 for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
801 struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
802 struct list_head *head = &io->bio_list;
803 struct bio_entry *be;
804
805 if (list_empty(head))
806 continue;
807
808 down_read(&io->bio_list_lock);
809 list_for_each_entry(be, head, list) {
810 if (target)
811 found = (target == be->bio);
812 else
813 found = __has_merged_page(be->bio, NULL,
814 page, 0);
815 if (found)
816 break;
817 }
818 up_read(&io->bio_list_lock);
819
820 if (!found)
821 continue;
822
823 found = false;
824
825 down_write(&io->bio_list_lock);
826 list_for_each_entry(be, head, list) {
827 if (target)
828 found = (target == be->bio);
829 else
830 found = __has_merged_page(be->bio, NULL,
831 page, 0);
832 if (found) {
833 target = be->bio;
834 del_bio_entry(be);
835 break;
836 }
837 }
838 up_write(&io->bio_list_lock);
839 }
840
841 if (found)
842 __submit_bio(sbi, target, DATA);
843 if (bio && *bio) {
844 bio_put(*bio);
845 *bio = NULL;
846 }
847}
848
8648de2c
CY		 849int f2fs_merge_page_bio(struct f2fs_io_info *fio)
850{
851 struct bio *bio = *fio->bio;
852 struct page *page = fio->encrypted_page ?
853 fio->encrypted_page : fio->page;
854
855 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
856 __is_meta_io(fio) ? META_GENERIC : DATA_GENERIC))
10f966bb 857 return -EFSCORRUPTED;
8648de2c
CY		 858
859 trace_f2fs_submit_page_bio(page, fio);
8648de2c 860
8896cbdf 861 if (bio && !page_is_mergeable(fio->sbi, bio, *fio->last_block,
0b20fcec
CY		 862 fio->new_blkaddr))
863 f2fs_submit_merged_ipu_write(fio->sbi, &bio, NULL);
8648de2c
CY		 864alloc_new:
865 if (!bio) {
a8affc03 866 bio = __bio_alloc(fio, BIO_MAX_VECS);
b7b911d5 867 __attach_io_flag(fio);
27aacd28
ST		 868 f2fs_set_bio_crypt_ctx(bio, fio->page->mapping->host,
869 fio->page->index, fio, GFP_NOIO);
8648de2c 870 bio_set_op_attrs(bio, fio->op, fio->op_flags);
8648de2c 871
0b20fcec
CY		 872 add_bio_entry(fio->sbi, bio, page, fio->temp);
873 } else {
27aacd28 874 if (add_ipu_page(fio, &bio, page))
0b20fcec 875 goto alloc_new;
8648de2c
CY		 876 }
877
878 if (fio->io_wbc)
9637d517 879 wbc_account_cgroup_owner(fio->io_wbc, page, PAGE_SIZE);
8648de2c
CY		 880
881 inc_page_count(fio->sbi, WB_DATA_TYPE(page));
882
883 *fio->last_block = fio->new_blkaddr;
884 *fio->bio = bio;
885
886 return 0;
887}
888
fe16efe6 889void f2fs_submit_page_write(struct f2fs_io_info *fio)
93dfe2ac 890{
05ca3632 891 struct f2fs_sb_info *sbi = fio->sbi;
458e6197 892 enum page_type btype = PAGE_TYPE_OF_BIO(fio->type);
a912b54d 893 struct f2fs_bio_info *io = sbi->write_io[btype] + fio->temp;
4375a336 894 struct page *bio_page;
93dfe2ac 895
b9109b0e 896 f2fs_bug_on(sbi, is_read_io(fio->op));
93dfe2ac 897
fb830fc5
CY		 898 down_write(&io->io_rwsem);
899next:
900 if (fio->in_list) {
901 spin_lock(&io->io_lock);
902 if (list_empty(&io->io_list)) {
903 spin_unlock(&io->io_lock);
fe16efe6 904 goto out;
fb830fc5
CY		 905 }
906 fio = list_first_entry(&io->io_list,
907 struct f2fs_io_info, list);
908 list_del(&fio->list);
909 spin_unlock(&io->io_lock);
910 }
93dfe2ac 911
93770ab7 912 verify_fio_blkaddr(fio);
93dfe2ac 913
4c8ff709
CY		 914 if (fio->encrypted_page)
915 bio_page = fio->encrypted_page;
916 else if (fio->compressed_page)
917 bio_page = fio->compressed_page;
918 else
919 bio_page = fio->page;
36951b38 920
ebf7c522
TM		 921 /* set submitted = true as a return value */
922 fio->submitted = true;
d68f735b 923
b9109b0e 924 inc_page_count(sbi, WB_DATA_TYPE(bio_page));
93dfe2ac 925
27aacd28
ST		 926 if (io->bio &&
927 (!io_is_mergeable(sbi, io->bio, io, fio, io->last_block_in_bio,
928 fio->new_blkaddr) ||
929 !f2fs_crypt_mergeable_bio(io->bio, fio->page->mapping->host,
930 bio_page->index, fio)))
458e6197 931 __submit_merged_bio(io);
93dfe2ac
JK		 932alloc_new:
933 if (io->bio == NULL) {
8223ecc4
CY		 934 if (F2FS_IO_ALIGNED(sbi) &&
935 (fio->type == DATA || fio->type == NODE) &&
0a595eba 936 fio->new_blkaddr & F2FS_IO_SIZE_MASK(sbi)) {
b9109b0e 937 dec_page_count(sbi, WB_DATA_TYPE(bio_page));
fe16efe6
CY		 938 fio->retry = true;
939 goto skip;
0a595eba 940 }
a8affc03 941 io->bio = __bio_alloc(fio, BIO_MAX_VECS);
27aacd28
ST		 942 f2fs_set_bio_crypt_ctx(io->bio, fio->page->mapping->host,
943 bio_page->index, fio, GFP_NOIO);
458e6197 944 io->fio = *fio;
93dfe2ac
JK		 945 }
946
a912b54d 947 if (bio_add_page(io->bio, bio_page, PAGE_SIZE, 0) < PAGE_SIZE) {
458e6197 948 __submit_merged_bio(io);
93dfe2ac
JK		 949 goto alloc_new;
950 }
951
578c6478 952 if (fio->io_wbc)
34e51a5e 953 wbc_account_cgroup_owner(fio->io_wbc, bio_page, PAGE_SIZE);
578c6478 954
7a9d7548 955 io->last_block_in_bio = fio->new_blkaddr;
fb830fc5
CY		 956
957 trace_f2fs_submit_page_write(fio->page, fio);
fe16efe6 958skip:
fb830fc5
CY		 959 if (fio->in_list)
960 goto next;
fe16efe6 961out:
4354994f 962 if (is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN) ||
00e09c0b 963 !f2fs_is_checkpoint_ready(sbi))
5ce80586 964 __submit_merged_bio(io);
df0f8dc0 965 up_write(&io->io_rwsem);
93dfe2ac
JK		 966}
967
13ba41e3 968static struct bio *f2fs_grab_read_bio(struct inode *inode, block_t blkaddr,
95ae251f 969 unsigned nr_pages, unsigned op_flag,
7f59b277 970 pgoff_t first_idx, bool for_write)
13ba41e3
JK		 971{
972 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
13ba41e3 973 struct bio *bio;
6dbb1796
EB		 974 struct bio_post_read_ctx *ctx;
975 unsigned int post_read_steps = 0;
13ba41e3 976
67883ade 977 bio = bio_alloc_bioset(for_write ? GFP_NOIO : GFP_KERNEL,
5f7136db 978 bio_max_segs(nr_pages), &f2fs_bioset);
6dbb1796 979 if (!bio)
13ba41e3 980 return ERR_PTR(-ENOMEM);
27aacd28
ST		 981
982 f2fs_set_bio_crypt_ctx(bio, inode, first_idx, NULL, GFP_NOFS);
983
13ba41e3
JK		 984 f2fs_target_device(sbi, blkaddr, bio);
985 bio->bi_end_io = f2fs_read_end_io;
e2e59414 986 bio_set_op_attrs(bio, REQ_OP_READ, op_flag);
13ba41e3 987
27aacd28 988 if (fscrypt_inode_uses_fs_layer_crypto(inode))
7f59b277 989 post_read_steps |= STEP_DECRYPT;
95ae251f 990
7f59b277
EB		 991 if (f2fs_need_verity(inode, first_idx))
992 post_read_steps |= STEP_VERITY;
993
994 /*
995 * STEP_DECOMPRESS is handled specially, since a compressed file might
996 * contain both compressed and uncompressed clusters. We'll allocate a
997 * bio_post_read_ctx if the file is compressed, but the caller is
998 * responsible for enabling STEP_DECOMPRESS if it's actually needed.
999 */
1000
1001 if (post_read_steps || f2fs_compressed_file(inode)) {
e8ce5749 1002 /* Due to the mempool, this never fails. */
6dbb1796 1003 ctx = mempool_alloc(bio_post_read_ctx_pool, GFP_NOFS);
6dbb1796 1004 ctx->bio = bio;
4c8ff709 1005 ctx->sbi = sbi;
6dbb1796 1006 ctx->enabled_steps = post_read_steps;
4931e0c9 1007 ctx->fs_blkaddr = blkaddr;
6dbb1796 1008 bio->bi_private = ctx;
6dbb1796
EB		 1009 }
1010
13ba41e3
JK		 1011 return bio;
1012}
1013
1014/* This can handle encryption stuffs */
1015static int f2fs_submit_page_read(struct inode *inode, struct page *page,
b7973091 1016 block_t blkaddr, int op_flags, bool for_write)
13ba41e3 1017{
93770ab7
CY		 1018 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1019 struct bio *bio;
13ba41e3 1020
b7973091 1021 bio = f2fs_grab_read_bio(inode, blkaddr, 1, op_flags,
7f59b277 1022 page->index, for_write);
13ba41e3
JK		 1023 if (IS_ERR(bio))
1024 return PTR_ERR(bio);
1025
0ded69f6
JK		 1026 /* wait for GCed page writeback via META_MAPPING */
1027 f2fs_wait_on_block_writeback(inode, blkaddr);
1028
13ba41e3
JK		 1029 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
1030 bio_put(bio);
1031 return -EFAULT;
1032 }
fb7d70db 1033 ClearPageError(page);
93770ab7 1034 inc_page_count(sbi, F2FS_RD_DATA);
8b83ac81 1035 f2fs_update_iostat(sbi, FS_DATA_READ_IO, F2FS_BLKSIZE);
93770ab7 1036 __submit_bio(sbi, bio, DATA);
13ba41e3
JK		 1037 return 0;
1038}
1039
46008c6d
CY		 1040static void __set_data_blkaddr(struct dnode_of_data *dn)
1041{
1042 struct f2fs_node *rn = F2FS_NODE(dn->node_page);
1043 __le32 *addr_array;
7a2af766
CY		 1044 int base = 0;
1045
1046 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
1047 base = get_extra_isize(dn->inode);
46008c6d
CY		 1048
1049 /* Get physical address of data block */
1050 addr_array = blkaddr_in_node(rn);
7a2af766 1051 addr_array[base + dn->ofs_in_node] = cpu_to_le32(dn->data_blkaddr);
46008c6d
CY		 1052}
1053
0a8165d7 1054/*
eb47b800
JK		 1055 * Lock ordering for the change of data block address:
1056 * ->data_page
1057 * ->node_page
1058 * update block addresses in the node page
1059 */
4d57b86d 1060void f2fs_set_data_blkaddr(struct dnode_of_data *dn)
eb47b800 1061{
bae0ee7a 1062 f2fs_wait_on_page_writeback(dn->node_page, NODE, true, true);
46008c6d
CY		 1063 __set_data_blkaddr(dn);
1064 if (set_page_dirty(dn->node_page))
12719ae1 1065 dn->node_changed = true;
eb47b800
JK		 1066}
1067
f28b3434
CY		 1068void f2fs_update_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr)
1069{
1070 dn->data_blkaddr = blkaddr;
4d57b86d 1071 f2fs_set_data_blkaddr(dn);
f28b3434
CY		 1072 f2fs_update_extent_cache(dn);
1073}
1074
46008c6d 1075/* dn->ofs_in_node will be returned with up-to-date last block pointer */
4d57b86d 1076int f2fs_reserve_new_blocks(struct dnode_of_data *dn, blkcnt_t count)
eb47b800 1077{
4081363f 1078 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
0abd675e 1079 int err;
eb47b800 1080
46008c6d
CY		 1081 if (!count)
1082 return 0;
1083
91942321 1084 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
eb47b800 1085 return -EPERM;
0abd675e
CY		 1086 if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
1087 return err;
eb47b800 1088
46008c6d
CY		 1089 trace_f2fs_reserve_new_blocks(dn->inode, dn->nid,
1090 dn->ofs_in_node, count);
1091
bae0ee7a 1092 f2fs_wait_on_page_writeback(dn->node_page, NODE, true, true);
46008c6d
CY		 1093
1094 for (; count > 0; dn->ofs_in_node++) {
a2ced1ce 1095 block_t blkaddr = f2fs_data_blkaddr(dn);
5f029c04 1096
46008c6d
CY		 1097 if (blkaddr == NULL_ADDR) {
1098 dn->data_blkaddr = NEW_ADDR;
1099 __set_data_blkaddr(dn);
1100 count--;
1101 }
1102 }
1103
1104 if (set_page_dirty(dn->node_page))
1105 dn->node_changed = true;
eb47b800
JK		 1106 return 0;
1107}
1108
46008c6d 1109/* Should keep dn->ofs_in_node unchanged */
4d57b86d 1110int f2fs_reserve_new_block(struct dnode_of_data *dn)
46008c6d
CY		 1111{
1112 unsigned int ofs_in_node = dn->ofs_in_node;
1113 int ret;
1114
4d57b86d 1115 ret = f2fs_reserve_new_blocks(dn, 1);
46008c6d
CY		 1116 dn->ofs_in_node = ofs_in_node;
1117 return ret;
1118}
1119
b600965c
HL		 1120int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
1121{
1122 bool need_put = dn->inode_page ? false : true;
1123 int err;
1124
4d57b86d 1125 err = f2fs_get_dnode_of_data(dn, index, ALLOC_NODE);
b600965c
HL		 1126 if (err)
1127 return err;
a8865372 1128
b600965c 1129 if (dn->data_blkaddr == NULL_ADDR)
4d57b86d 1130 err = f2fs_reserve_new_block(dn);
a8865372 1131 if (err || need_put)
b600965c
HL		 1132 f2fs_put_dnode(dn);
1133 return err;
1134}
1135
759af1c9 1136int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index)
eb47b800 1137{
a87aff1d 1138 struct extent_info ei = {0, 0, 0};
759af1c9 1139 struct inode *inode = dn->inode;
028a41e8 1140
759af1c9
FL		 1141 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
1142 dn->data_blkaddr = ei.blk + index - ei.fofs;
1143 return 0;
429511cd 1144 }
028a41e8 1145
759af1c9 1146 return f2fs_reserve_block(dn, index);
eb47b800
JK		 1147}
1148
4d57b86d 1149struct page *f2fs_get_read_data_page(struct inode *inode, pgoff_t index,
04d328de 1150 int op_flags, bool for_write)
eb47b800 1151{
eb47b800
JK		 1152 struct address_space *mapping = inode->i_mapping;
1153 struct dnode_of_data dn;
1154 struct page *page;
e15882b6 1155 struct extent_info ei = {0,0,0};
eb47b800 1156 int err;
4375a336 1157
a56c7c6f 1158 page = f2fs_grab_cache_page(mapping, index, for_write);
650495de
JK		 1159 if (!page)
1160 return ERR_PTR(-ENOMEM);
1161
cb3bc9ee
CY		 1162 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
1163 dn.data_blkaddr = ei.blk + index - ei.fofs;
93770ab7
CY		 1164 if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), dn.data_blkaddr,
1165 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 1166 err = -EFSCORRUPTED;
93770ab7
CY		 1167 goto put_err;
1168 }
cb3bc9ee
CY		 1169 goto got_it;
1170 }
1171
eb47b800 1172 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1173 err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
86531d6b
JK		 1174 if (err)
1175 goto put_err;
eb47b800
JK		 1176 f2fs_put_dnode(&dn);
1177
6bacf52f 1178 if (unlikely(dn.data_blkaddr == NULL_ADDR)) {
86531d6b
JK		 1179 err = -ENOENT;
1180 goto put_err;
650495de 1181 }
93770ab7
CY		 1182 if (dn.data_blkaddr != NEW_ADDR &&
1183 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
1184 dn.data_blkaddr,
1185 DATA_GENERIC_ENHANCE)) {
10f966bb 1186 err = -EFSCORRUPTED;
93770ab7
CY		 1187 goto put_err;
1188 }
cb3bc9ee 1189got_it:
43f3eae1
JK		 1190 if (PageUptodate(page)) {
1191 unlock_page(page);
eb47b800 1192 return page;
43f3eae1 1193 }
eb47b800 1194
d59ff4df
JK		 1195 /*
1196 * A new dentry page is allocated but not able to be written, since its
1197 * new inode page couldn't be allocated due to -ENOSPC.
1198 * In such the case, its blkaddr can be remained as NEW_ADDR.
4d57b86d
CY		 1199 * see, f2fs_add_link -> f2fs_get_new_data_page ->
1200 * f2fs_init_inode_metadata.
d59ff4df
JK		 1201 */
1202 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 1203 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 1204 if (!PageUptodate(page))
1205 SetPageUptodate(page);
43f3eae1 1206 unlock_page(page);
d59ff4df
JK		 1207 return page;
1208 }
eb47b800 1209
b7973091
JY		 1210 err = f2fs_submit_page_read(inode, page, dn.data_blkaddr,
1211 op_flags, for_write);
393ff91f 1212 if (err)
86531d6b 1213 goto put_err;
43f3eae1 1214 return page;
86531d6b
JK		 1215
1216put_err:
1217 f2fs_put_page(page, 1);
1218 return ERR_PTR(err);
43f3eae1
JK		 1219}
1220
4d57b86d 1221struct page *f2fs_find_data_page(struct inode *inode, pgoff_t index)
43f3eae1
JK		 1222{
1223 struct address_space *mapping = inode->i_mapping;
1224 struct page *page;
1225
1226 page = find_get_page(mapping, index);
1227 if (page && PageUptodate(page))
1228 return page;
1229 f2fs_put_page(page, 0);
1230
4d57b86d 1231 page = f2fs_get_read_data_page(inode, index, 0, false);
43f3eae1
JK		 1232 if (IS_ERR(page))
1233 return page;
1234
1235 if (PageUptodate(page))
1236 return page;
1237
1238 wait_on_page_locked(page);
1239 if (unlikely(!PageUptodate(page))) {
1240 f2fs_put_page(page, 0);
1241 return ERR_PTR(-EIO);
1242 }
1243 return page;
1244}
1245
1246/*
1247 * If it tries to access a hole, return an error.
1248 * Because, the callers, functions in dir.c and GC, should be able to know
1249 * whether this page exists or not.
1250 */
4d57b86d 1251struct page *f2fs_get_lock_data_page(struct inode *inode, pgoff_t index,
a56c7c6f 1252 bool for_write)
43f3eae1
JK		 1253{
1254 struct address_space *mapping = inode->i_mapping;
1255 struct page *page;
1256repeat:
4d57b86d 1257 page = f2fs_get_read_data_page(inode, index, 0, for_write);
43f3eae1
JK		 1258 if (IS_ERR(page))
1259 return page;
393ff91f 1260
43f3eae1 1261 /* wait for read completion */
393ff91f 1262 lock_page(page);
6bacf52f 1263 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 1264 f2fs_put_page(page, 1);
1265 goto repeat;
eb47b800 1266 }
1563ac75
CY		 1267 if (unlikely(!PageUptodate(page))) {
1268 f2fs_put_page(page, 1);
1269 return ERR_PTR(-EIO);
1270 }
eb47b800
JK		 1271 return page;
1272}
1273
0a8165d7 1274/*
eb47b800
JK		 1275 * Caller ensures that this data page is never allocated.
1276 * A new zero-filled data page is allocated in the page cache.
39936837 1277 *
4f4124d0
CY		 1278 * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and
1279 * f2fs_unlock_op().
470f00e9
CY		 1280 * Note that, ipage is set only by make_empty_dir, and if any error occur,
1281 * ipage should be released by this function.
eb47b800 1282 */
4d57b86d 1283struct page *f2fs_get_new_data_page(struct inode *inode,
a8865372 1284 struct page *ipage, pgoff_t index, bool new_i_size)
eb47b800 1285{
eb47b800
JK		 1286 struct address_space *mapping = inode->i_mapping;
1287 struct page *page;
1288 struct dnode_of_data dn;
1289 int err;
7612118a 1290
a56c7c6f 1291 page = f2fs_grab_cache_page(mapping, index, true);
470f00e9
CY		 1292 if (!page) {
1293 /*
1294 * before exiting, we should make sure ipage will be released
1295 * if any error occur.
1296 */
1297 f2fs_put_page(ipage, 1);
01f28610 1298 return ERR_PTR(-ENOMEM);
470f00e9 1299 }
eb47b800 1300
a8865372 1301 set_new_dnode(&dn, inode, ipage, NULL, 0);
b600965c 1302 err = f2fs_reserve_block(&dn, index);
01f28610
JK		 1303 if (err) {
1304 f2fs_put_page(page, 1);
eb47b800 1305 return ERR_PTR(err);
a8865372 1306 }
01f28610
JK		 1307 if (!ipage)
1308 f2fs_put_dnode(&dn);
eb47b800
JK		 1309
1310 if (PageUptodate(page))
01f28610 1311 goto got_it;
eb47b800
JK		 1312
1313 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 1314 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 1315 if (!PageUptodate(page))
1316 SetPageUptodate(page);
eb47b800 1317 } else {
4375a336 1318 f2fs_put_page(page, 1);
a8865372 1319
7612118a
JK		 1320 /* if ipage exists, blkaddr should be NEW_ADDR */
1321 f2fs_bug_on(F2FS_I_SB(inode), ipage);
4d57b86d 1322 page = f2fs_get_lock_data_page(inode, index, true);
4375a336 1323 if (IS_ERR(page))
7612118a 1324 return page;
eb47b800 1325 }
01f28610 1326got_it:
9edcdabf 1327 if (new_i_size && i_size_read(inode) <
ee6d182f 1328 ((loff_t)(index + 1) << PAGE_SHIFT))
fc9581c8 1329 f2fs_i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT));
eb47b800
JK		 1330 return page;
1331}
1332
d5097be5 1333static int __allocate_data_block(struct dnode_of_data *dn, int seg_type)
bfad7c2d 1334{
4081363f 1335 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
bfad7c2d 1336 struct f2fs_summary sum;
bfad7c2d 1337 struct node_info ni;
6aa58d8a 1338 block_t old_blkaddr;
46008c6d 1339 blkcnt_t count = 1;
0abd675e 1340 int err;
bfad7c2d 1341
91942321 1342 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
bfad7c2d 1343 return -EPERM;
df6136ef 1344
7735730d
CY		 1345 err = f2fs_get_node_info(sbi, dn->nid, &ni);
1346 if (err)
1347 return err;
1348
a2ced1ce 1349 dn->data_blkaddr = f2fs_data_blkaddr(dn);
f847c699 1350 if (dn->data_blkaddr != NULL_ADDR)
df6136ef
CY		 1351 goto alloc;
1352
0abd675e
CY		 1353 if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
1354 return err;
bfad7c2d 1355
df6136ef 1356alloc:
bfad7c2d 1357 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
6aa58d8a
CY		 1358 old_blkaddr = dn->data_blkaddr;
1359 f2fs_allocate_data_block(sbi, NULL, old_blkaddr, &dn->data_blkaddr,
093749e2 1360 &sum, seg_type, NULL);
6ce19aff 1361 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO) {
6aa58d8a
CY		 1362 invalidate_mapping_pages(META_MAPPING(sbi),
1363 old_blkaddr, old_blkaddr);
6ce19aff
CY		 1364 f2fs_invalidate_compress_page(sbi, old_blkaddr);
1365 }
86f35dc3 1366 f2fs_update_data_blkaddr(dn, dn->data_blkaddr);
bfad7c2d 1367
0a4daae5
JK		 1368 /*
1369 * i_size will be updated by direct_IO. Otherwise, we'll get stale
1370 * data from unwritten block via dio_read.
1371 */
bfad7c2d
JK		 1372 return 0;
1373}
1374
a7de6086 1375int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *from)
59b802e5 1376{
b439b103 1377 struct inode *inode = file_inode(iocb->ki_filp);
5b8db7fa 1378 struct f2fs_map_blocks map;
d6d478a1 1379 int flag;
a7de6086 1380 int err = 0;
d6d478a1 1381 bool direct_io = iocb->ki_flags & IOCB_DIRECT;
59b802e5 1382
0080c507 1383 map.m_lblk = F2FS_BLK_ALIGN(iocb->ki_pos);
dfd02e4d
CY		 1384 map.m_len = F2FS_BYTES_TO_BLK(iocb->ki_pos + iov_iter_count(from));
1385 if (map.m_len > map.m_lblk)
1386 map.m_len -= map.m_lblk;
1387 else
1388 map.m_len = 0;
1389
da85985c 1390 map.m_next_pgofs = NULL;
c4020b2d 1391 map.m_next_extent = NULL;
d5097be5 1392 map.m_seg_type = NO_CHECK_TYPE;
f9d6d059 1393 map.m_may_create = true;
2a340760 1394
d6d478a1 1395 if (direct_io) {
4d57b86d 1396 map.m_seg_type = f2fs_rw_hint_to_seg_type(iocb->ki_hint);
f847c699 1397 flag = f2fs_force_buffered_io(inode, iocb, from) ?
d6d478a1
CY		 1398 F2FS_GET_BLOCK_PRE_AIO :
1399 F2FS_GET_BLOCK_PRE_DIO;
1400 goto map_blocks;
d5097be5 1401 }
f2470371 1402 if (iocb->ki_pos + iov_iter_count(from) > MAX_INLINE_DATA(inode)) {
a7de6086
JK		 1403 err = f2fs_convert_inline_inode(inode);
1404 if (err)
1405 return err;
b439b103 1406 }
d6d478a1 1407 if (f2fs_has_inline_data(inode))
25006645 1408 return err;
d6d478a1
CY		 1409
1410 flag = F2FS_GET_BLOCK_PRE_AIO;
1411
1412map_blocks:
1413 err = f2fs_map_blocks(inode, &map, 1, flag);
1414 if (map.m_len > 0 && err == -ENOSPC) {
1415 if (!direct_io)
1416 set_inode_flag(inode, FI_NO_PREALLOC);
1417 err = 0;
25006645 1418 }
a7de6086 1419 return err;
59b802e5
JK		 1420}
1421
0ef81833 1422void f2fs_do_map_lock(struct f2fs_sb_info *sbi, int flag, bool lock)
59c9081b
YH		 1423{
1424 if (flag == F2FS_GET_BLOCK_PRE_AIO) {
1425 if (lock)
1426 down_read(&sbi->node_change);
1427 else
1428 up_read(&sbi->node_change);
1429 } else {
1430 if (lock)
1431 f2fs_lock_op(sbi);
1432 else
1433 f2fs_unlock_op(sbi);
1434 }
1435}
1436
0a8165d7 1437/*
7a88ddb5
CY		 1438 * f2fs_map_blocks() tries to find or build mapping relationship which
1439 * maps continuous logical blocks to physical blocks, and return such
1440 * info via f2fs_map_blocks structure.
eb47b800 1441 */
d323d005 1442int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
e2b4e2bc 1443 int create, int flag)
eb47b800 1444{
003a3e1d 1445 unsigned int maxblocks = map->m_len;
eb47b800 1446 struct dnode_of_data dn;
f9811703 1447 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
f9d6d059 1448 int mode = map->m_may_create ? ALLOC_NODE : LOOKUP_NODE;
46008c6d 1449 pgoff_t pgofs, end_offset, end;
bfad7c2d 1450 int err = 0, ofs = 1;
46008c6d
CY		 1451 unsigned int ofs_in_node, last_ofs_in_node;
1452 blkcnt_t prealloc;
e15882b6 1453 struct extent_info ei = {0,0,0};
7df3a431 1454 block_t blkaddr;
c4020b2d 1455 unsigned int start_pgofs;
eb47b800 1456
dfd02e4d
CY		 1457 if (!maxblocks)
1458 return 0;
1459
003a3e1d
JK		 1460 map->m_len = 0;
1461 map->m_flags = 0;
1462
1463 /* it only supports block size == page size */
1464 pgofs = (pgoff_t)map->m_lblk;
46008c6d 1465 end = pgofs + maxblocks;
eb47b800 1466
24b84912 1467 if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) {
b0332a0f 1468 if (f2fs_lfs_mode(sbi) && flag == F2FS_GET_BLOCK_DIO &&
f4f0b677
JZ		 1469 map->m_may_create)
1470 goto next_dnode;
1471
003a3e1d
JK		 1472 map->m_pblk = ei.blk + pgofs - ei.fofs;
1473 map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs);
1474 map->m_flags = F2FS_MAP_MAPPED;
c4020b2d
CY		 1475 if (map->m_next_extent)
1476 *map->m_next_extent = pgofs + map->m_len;
1e78e8bd
ST		 1477
1478 /* for hardware encryption, but to avoid potential issue in future */
1479 if (flag == F2FS_GET_BLOCK_DIO)
1480 f2fs_wait_on_block_writeback_range(inode,
1481 map->m_pblk, map->m_len);
bfad7c2d 1482 goto out;
a2e7d1bf 1483 }
bfad7c2d 1484
4fe71e88 1485next_dnode:
f9d6d059 1486 if (map->m_may_create)
0ef81833 1487 f2fs_do_map_lock(sbi, flag, true);
eb47b800
JK		 1488
1489 /* When reading holes, we need its node page */
1490 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1491 err = f2fs_get_dnode_of_data(&dn, pgofs, mode);
1ec79083 1492 if (err) {
43473f96
CY		 1493 if (flag == F2FS_GET_BLOCK_BMAP)
1494 map->m_pblk = 0;
da85985c 1495 if (err == -ENOENT) {
bfad7c2d 1496 err = 0;
da85985c
CY		 1497 if (map->m_next_pgofs)
1498 *map->m_next_pgofs =
4d57b86d 1499 f2fs_get_next_page_offset(&dn, pgofs);
c4020b2d
CY		 1500 if (map->m_next_extent)
1501 *map->m_next_extent =
4d57b86d 1502 f2fs_get_next_page_offset(&dn, pgofs);
da85985c 1503 }
bfad7c2d 1504 goto unlock_out;
848753aa 1505 }
973163fc 1506
c4020b2d 1507 start_pgofs = pgofs;
46008c6d 1508 prealloc = 0;
230436b3 1509 last_ofs_in_node = ofs_in_node = dn.ofs_in_node;
81ca7350 1510 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
4fe71e88
CY		 1511
1512next_block:
a2ced1ce 1513 blkaddr = f2fs_data_blkaddr(&dn);
4fe71e88 1514
c9b60788 1515 if (__is_valid_data_blkaddr(blkaddr) &&
93770ab7 1516 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC_ENHANCE)) {
10f966bb 1517 err = -EFSCORRUPTED;
c9b60788
CY		 1518 goto sync_out;
1519 }
1520
93770ab7 1521 if (__is_valid_data_blkaddr(blkaddr)) {
f847c699 1522 /* use out-place-update for driect IO under LFS mode */
b0332a0f 1523 if (f2fs_lfs_mode(sbi) && flag == F2FS_GET_BLOCK_DIO &&
f9d6d059 1524 map->m_may_create) {
f847c699 1525 err = __allocate_data_block(&dn, map->m_seg_type);
05e36006
CY		 1526 if (err)
1527 goto sync_out;
1528 blkaddr = dn.data_blkaddr;
1529 set_inode_flag(inode, FI_APPEND_WRITE);
f847c699
CY		 1530 }
1531 } else {
973163fc 1532 if (create) {
f9811703
CY		 1533 if (unlikely(f2fs_cp_error(sbi))) {
1534 err = -EIO;
4fe71e88 1535 goto sync_out;
f9811703 1536 }
24b84912 1537 if (flag == F2FS_GET_BLOCK_PRE_AIO) {
46008c6d
CY		 1538 if (blkaddr == NULL_ADDR) {
1539 prealloc++;
1540 last_ofs_in_node = dn.ofs_in_node;
1541 }
24b84912 1542 } else {
0a4daae5
JK		 1543 WARN_ON(flag != F2FS_GET_BLOCK_PRE_DIO &&
1544 flag != F2FS_GET_BLOCK_DIO);
d5097be5
HL		 1545 err = __allocate_data_block(&dn,
1546 map->m_seg_type);
6f2d8ed6 1547 if (!err)
91942321 1548 set_inode_flag(inode, FI_APPEND_WRITE);
24b84912 1549 }
973163fc 1550 if (err)
4fe71e88 1551 goto sync_out;
3f2be043 1552 map->m_flags |= F2FS_MAP_NEW;
4fe71e88 1553 blkaddr = dn.data_blkaddr;
973163fc 1554 } else {
43473f96
CY		 1555 if (flag == F2FS_GET_BLOCK_BMAP) {
1556 map->m_pblk = 0;
1557 goto sync_out;
1558 }
c4020b2d
CY		 1559 if (flag == F2FS_GET_BLOCK_PRECACHE)
1560 goto sync_out;
da85985c
CY		 1561 if (flag == F2FS_GET_BLOCK_FIEMAP &&
1562 blkaddr == NULL_ADDR) {
1563 if (map->m_next_pgofs)
1564 *map->m_next_pgofs = pgofs + 1;
4c2ac6a8 1565 goto sync_out;
da85985c 1566 }
f3d98e74
CY		 1567 if (flag != F2FS_GET_BLOCK_FIEMAP) {
1568 /* for defragment case */
1569 if (map->m_next_pgofs)
1570 *map->m_next_pgofs = pgofs + 1;
4fe71e88 1571 goto sync_out;
f3d98e74 1572 }
e2b4e2bc 1573 }
e2b4e2bc 1574 }
eb47b800 1575
46008c6d
CY		 1576 if (flag == F2FS_GET_BLOCK_PRE_AIO)
1577 goto skip;
1578
4fe71e88
CY		 1579 if (map->m_len == 0) {
1580 /* preallocated unwritten block should be mapped for fiemap. */
1581 if (blkaddr == NEW_ADDR)
1582 map->m_flags |= F2FS_MAP_UNWRITTEN;
1583 map->m_flags |= F2FS_MAP_MAPPED;
1584
1585 map->m_pblk = blkaddr;
1586 map->m_len = 1;
1587 } else if ((map->m_pblk != NEW_ADDR &&
1588 blkaddr == (map->m_pblk + ofs)) ||
b439b103 1589 (map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR) ||
46008c6d 1590 flag == F2FS_GET_BLOCK_PRE_DIO) {
4fe71e88
CY		 1591 ofs++;
1592 map->m_len++;
1593 } else {
1594 goto sync_out;
1595 }
bfad7c2d 1596
46008c6d 1597skip:
bfad7c2d
JK		 1598 dn.ofs_in_node++;
1599 pgofs++;
1600
46008c6d
CY		 1601 /* preallocate blocks in batch for one dnode page */
1602 if (flag == F2FS_GET_BLOCK_PRE_AIO &&
1603 (pgofs == end || dn.ofs_in_node == end_offset)) {
7df3a431 1604
46008c6d 1605 dn.ofs_in_node = ofs_in_node;
4d57b86d 1606 err = f2fs_reserve_new_blocks(&dn, prealloc);
46008c6d
CY		 1607 if (err)
1608 goto sync_out;
bfad7c2d 1609
46008c6d
CY		 1610 map->m_len += dn.ofs_in_node - ofs_in_node;
1611 if (prealloc && dn.ofs_in_node != last_ofs_in_node + 1) {
1612 err = -ENOSPC;
1613 goto sync_out;
3104af35 1614 }
46008c6d
CY		 1615 dn.ofs_in_node = end_offset;
1616 }
1617
1618 if (pgofs >= end)
1619 goto sync_out;
1620 else if (dn.ofs_in_node < end_offset)
1621 goto next_block;
1622
c4020b2d
CY		 1623 if (flag == F2FS_GET_BLOCK_PRECACHE) {
1624 if (map->m_flags & F2FS_MAP_MAPPED) {
1625 unsigned int ofs = start_pgofs - map->m_lblk;
1626
1627 f2fs_update_extent_cache_range(&dn,
1628 start_pgofs, map->m_pblk + ofs,
1629 map->m_len - ofs);
1630 }
1631 }
1632
46008c6d
CY		 1633 f2fs_put_dnode(&dn);
1634
f9d6d059 1635 if (map->m_may_create) {
0ef81833 1636 f2fs_do_map_lock(sbi, flag, false);
6f2d8ed6 1637 f2fs_balance_fs(sbi, dn.node_changed);
eb47b800 1638 }
46008c6d 1639 goto next_dnode;
7df3a431 1640
bfad7c2d 1641sync_out:
1e78e8bd
ST		 1642
1643 /* for hardware encryption, but to avoid potential issue in future */
1644 if (flag == F2FS_GET_BLOCK_DIO && map->m_flags & F2FS_MAP_MAPPED)
1645 f2fs_wait_on_block_writeback_range(inode,
1646 map->m_pblk, map->m_len);
1647
c4020b2d
CY		 1648 if (flag == F2FS_GET_BLOCK_PRECACHE) {
1649 if (map->m_flags & F2FS_MAP_MAPPED) {
1650 unsigned int ofs = start_pgofs - map->m_lblk;
1651
1652 f2fs_update_extent_cache_range(&dn,
1653 start_pgofs, map->m_pblk + ofs,
1654 map->m_len - ofs);
1655 }
1656 if (map->m_next_extent)
1657 *map->m_next_extent = pgofs + 1;
1658 }
eb47b800 1659 f2fs_put_dnode(&dn);
bfad7c2d 1660unlock_out:
f9d6d059 1661 if (map->m_may_create) {
0ef81833 1662 f2fs_do_map_lock(sbi, flag, false);
6f2d8ed6 1663 f2fs_balance_fs(sbi, dn.node_changed);
2a340760 1664 }
bfad7c2d 1665out:
003a3e1d 1666 trace_f2fs_map_blocks(inode, map, err);
bfad7c2d 1667 return err;
eb47b800
JK		 1668}
1669
b91050a8
HL		 1670bool f2fs_overwrite_io(struct inode *inode, loff_t pos, size_t len)
1671{
1672 struct f2fs_map_blocks map;
1673 block_t last_lblk;
1674 int err;
1675
1676 if (pos + len > i_size_read(inode))
1677 return false;
1678
1679 map.m_lblk = F2FS_BYTES_TO_BLK(pos);
1680 map.m_next_pgofs = NULL;
1681 map.m_next_extent = NULL;
1682 map.m_seg_type = NO_CHECK_TYPE;
f4f0b677 1683 map.m_may_create = false;
b91050a8
HL		 1684 last_lblk = F2FS_BLK_ALIGN(pos + len);
1685
1686 while (map.m_lblk < last_lblk) {
1687 map.m_len = last_lblk - map.m_lblk;
1688 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
1689 if (err || map.m_len == 0)
1690 return false;
1691 map.m_lblk += map.m_len;
1692 }
1693 return true;
1694}
1695
43b9d4b4
JK		 1696static inline u64 bytes_to_blks(struct inode *inode, u64 bytes)
1697{
1698 return (bytes >> inode->i_blkbits);
1699}
1700
1701static inline u64 blks_to_bytes(struct inode *inode, u64 blks)
1702{
1703 return (blks << inode->i_blkbits);
1704}
1705
003a3e1d 1706static int __get_data_block(struct inode *inode, sector_t iblock,
da85985c 1707 struct buffer_head *bh, int create, int flag,
f9d6d059 1708 pgoff_t *next_pgofs, int seg_type, bool may_write)
003a3e1d
JK		 1709{
1710 struct f2fs_map_blocks map;
a7de6086 1711 int err;
003a3e1d
JK		 1712
1713 map.m_lblk = iblock;
43b9d4b4 1714 map.m_len = bytes_to_blks(inode, bh->b_size);
da85985c 1715 map.m_next_pgofs = next_pgofs;
c4020b2d 1716 map.m_next_extent = NULL;
d5097be5 1717 map.m_seg_type = seg_type;
f9d6d059 1718 map.m_may_create = may_write;
003a3e1d 1719
a7de6086
JK		 1720 err = f2fs_map_blocks(inode, &map, create, flag);
1721 if (!err) {
003a3e1d
JK		 1722 map_bh(bh, inode->i_sb, map.m_pblk);
1723 bh->b_state = (bh->b_state & ~F2FS_MAP_FLAGS) | map.m_flags;
43b9d4b4 1724 bh->b_size = blks_to_bytes(inode, map.m_len);
003a3e1d 1725 }
a7de6086 1726 return err;
003a3e1d
JK		 1727}
1728
f9d6d059
CY		 1729static int get_data_block_dio_write(struct inode *inode, sector_t iblock,
1730 struct buffer_head *bh_result, int create)
1731{
1732 return __get_data_block(inode, iblock, bh_result, create,
1733 F2FS_GET_BLOCK_DIO, NULL,
1734 f2fs_rw_hint_to_seg_type(inode->i_write_hint),
ebc29b62 1735 true);
e2b4e2bc
CY		 1736}
1737
1738static int get_data_block_dio(struct inode *inode, sector_t iblock,
ccfb3000
JK		 1739 struct buffer_head *bh_result, int create)
1740{
e2b4e2bc 1741 return __get_data_block(inode, iblock, bh_result, create,
f9d6d059
CY		 1742 F2FS_GET_BLOCK_DIO, NULL,
1743 f2fs_rw_hint_to_seg_type(inode->i_write_hint),
1744 false);
ccfb3000
JK		 1745}
1746
442a9dbd
CY		 1747static int f2fs_xattr_fiemap(struct inode *inode,
1748 struct fiemap_extent_info *fieinfo)
1749{
1750 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1751 struct page *page;
1752 struct node_info ni;
1753 __u64 phys = 0, len;
1754 __u32 flags;
1755 nid_t xnid = F2FS_I(inode)->i_xattr_nid;
1756 int err = 0;
1757
1758 if (f2fs_has_inline_xattr(inode)) {
1759 int offset;
1760
1761 page = f2fs_grab_cache_page(NODE_MAPPING(sbi),
1762 inode->i_ino, false);
1763 if (!page)
1764 return -ENOMEM;
1765
7735730d
CY		 1766 err = f2fs_get_node_info(sbi, inode->i_ino, &ni);
1767 if (err) {
1768 f2fs_put_page(page, 1);
1769 return err;
1770 }
442a9dbd 1771
6cbfcab5 1772 phys = blks_to_bytes(inode, ni.blk_addr);
442a9dbd
CY		 1773 offset = offsetof(struct f2fs_inode, i_addr) +
1774 sizeof(__le32) * (DEF_ADDRS_PER_INODE -
b323fd28 1775 get_inline_xattr_addrs(inode));
442a9dbd
CY		 1776
1777 phys += offset;
1778 len = inline_xattr_size(inode);
1779
1780 f2fs_put_page(page, 1);
1781
1782 flags = FIEMAP_EXTENT_DATA_INLINE | FIEMAP_EXTENT_NOT_ALIGNED;
1783
1784 if (!xnid)
1785 flags |= FIEMAP_EXTENT_LAST;
1786
1787 err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
dd5a09bd 1788 trace_f2fs_fiemap(inode, 0, phys, len, flags, err);
442a9dbd
CY		 1789 if (err || err == 1)
1790 return err;
1791 }
1792
1793 if (xnid) {
1794 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), xnid, false);
1795 if (!page)
1796 return -ENOMEM;
1797
7735730d
CY		 1798 err = f2fs_get_node_info(sbi, xnid, &ni);
1799 if (err) {
1800 f2fs_put_page(page, 1);
1801 return err;
1802 }
442a9dbd 1803
6cbfcab5 1804 phys = blks_to_bytes(inode, ni.blk_addr);
442a9dbd
CY		 1805 len = inode->i_sb->s_blocksize;
1806
1807 f2fs_put_page(page, 1);
1808
1809 flags = FIEMAP_EXTENT_LAST;
1810 }
1811
dd5a09bd 1812 if (phys) {
442a9dbd 1813 err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
dd5a09bd
CY		 1814 trace_f2fs_fiemap(inode, 0, phys, len, flags, err);
1815 }
442a9dbd
CY		 1816
1817 return (err < 0 ? err : 0);
1818}
1819
bf38fbad
CY		 1820static loff_t max_inode_blocks(struct inode *inode)
1821{
1822 loff_t result = ADDRS_PER_INODE(inode);
1823 loff_t leaf_count = ADDRS_PER_BLOCK(inode);
1824
1825 /* two direct node blocks */
1826 result += (leaf_count * 2);
1827
1828 /* two indirect node blocks */
1829 leaf_count *= NIDS_PER_BLOCK;
1830 result += (leaf_count * 2);
1831
1832 /* one double indirect node block */
1833 leaf_count *= NIDS_PER_BLOCK;
1834 result += leaf_count;
1835
1836 return result;
1837}
1838
9ab70134
JK		 1839int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1840 u64 start, u64 len)
1841{
b876f4c9 1842 struct f2fs_map_blocks map;
7f63eb77 1843 sector_t start_blk, last_blk;
da85985c 1844 pgoff_t next_pgofs;
7f63eb77
JK		 1845 u64 logical = 0, phys = 0, size = 0;
1846 u32 flags = 0;
7f63eb77 1847 int ret = 0;
093f0bac 1848 bool compr_cluster = false, compr_appended;
bf38fbad 1849 unsigned int cluster_size = F2FS_I(inode)->i_cluster_size;
093f0bac 1850 unsigned int count_in_cluster = 0;
0bb2045c 1851 loff_t maxbytes;
7f63eb77 1852
c4020b2d
CY		 1853 if (fieinfo->fi_flags & FIEMAP_FLAG_CACHE) {
1854 ret = f2fs_precache_extents(inode);
1855 if (ret)
1856 return ret;
1857 }
1858
45dd052e 1859 ret = fiemap_prep(inode, fieinfo, start, &len, FIEMAP_FLAG_XATTR);
7f63eb77
JK		 1860 if (ret)
1861 return ret;
1862
f1b43d4c
CY		 1863 inode_lock(inode);
1864
0bb2045c
CX		 1865 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
1866 if (start > maxbytes) {
1867 ret = -EFBIG;
1868 goto out;
1869 }
1870
1871 if (len > maxbytes || (maxbytes - len) < start)
1872 len = maxbytes - start;
1873
442a9dbd
CY		 1874 if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
1875 ret = f2fs_xattr_fiemap(inode, fieinfo);
1876 goto out;
1877 }
1878
7975f349 1879 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
67f8cf3c
JK		 1880 ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len);
1881 if (ret != -EAGAIN)
f1b43d4c 1882 goto out;
67f8cf3c
JK		 1883 }
1884
6cbfcab5
JK		 1885 if (bytes_to_blks(inode, len) == 0)
1886 len = blks_to_bytes(inode, 1);
7f63eb77 1887
6cbfcab5
JK		 1888 start_blk = bytes_to_blks(inode, start);
1889 last_blk = bytes_to_blks(inode, start + len - 1);
9a950d52 1890
7f63eb77 1891next:
b876f4c9
JK		 1892 memset(&map, 0, sizeof(map));
1893 map.m_lblk = start_blk;
1894 map.m_len = bytes_to_blks(inode, len);
1895 map.m_next_pgofs = &next_pgofs;
1896 map.m_seg_type = NO_CHECK_TYPE;
7f63eb77 1897
093f0bac
DJ		 1898 if (compr_cluster) {
1899 map.m_lblk += 1;
1900 map.m_len = cluster_size - count_in_cluster;
1901 }
bf38fbad 1902
b876f4c9 1903 ret = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_FIEMAP);
7f63eb77
JK		 1904 if (ret)
1905 goto out;
1906
1907 /* HOLE */
093f0bac 1908 if (!compr_cluster && !(map.m_flags & F2FS_MAP_FLAGS)) {
da85985c 1909 start_blk = next_pgofs;
58736fa6 1910
6cbfcab5 1911 if (blks_to_bytes(inode, start_blk) < blks_to_bytes(inode,
bf38fbad 1912 max_inode_blocks(inode)))
9a950d52 1913 goto prep_next;
58736fa6 1914
9a950d52
FL		 1915 flags |= FIEMAP_EXTENT_LAST;
1916 }
7f63eb77 1917
093f0bac
DJ		 1918 compr_appended = false;
1919 /* In a case of compressed cluster, append this to the last extent */
1920 if (compr_cluster && ((map.m_flags & F2FS_MAP_UNWRITTEN) ||
1921 !(map.m_flags & F2FS_MAP_FLAGS))) {
1922 compr_appended = true;
1923 goto skip_fill;
1924 }
1925
da5af127 1926 if (size) {
0953fe86 1927 flags |= FIEMAP_EXTENT_MERGED;
62230e0d 1928 if (IS_ENCRYPTED(inode))
da5af127
CY		 1929 flags |= FIEMAP_EXTENT_DATA_ENCRYPTED;
1930
9a950d52
FL		 1931 ret = fiemap_fill_next_extent(fieinfo, logical,
1932 phys, size, flags);
dd5a09bd 1933 trace_f2fs_fiemap(inode, logical, phys, size, flags, ret);
bf38fbad
CY		 1934 if (ret)
1935 goto out;
1936 size = 0;
da5af127 1937 }
7f63eb77 1938
bf38fbad 1939 if (start_blk > last_blk)
9a950d52 1940 goto out;
7f63eb77 1941
093f0bac 1942skip_fill:
b876f4c9 1943 if (map.m_pblk == COMPRESS_ADDR) {
bf38fbad 1944 compr_cluster = true;
093f0bac
DJ		 1945 count_in_cluster = 1;
1946 } else if (compr_appended) {
1947 unsigned int appended_blks = cluster_size -
1948 count_in_cluster + 1;
1949 size += blks_to_bytes(inode, appended_blks);
1950 start_blk += appended_blks;
1951 compr_cluster = false;
1952 } else {
1953 logical = blks_to_bytes(inode, start_blk);
1954 phys = __is_valid_data_blkaddr(map.m_pblk) ?
1955 blks_to_bytes(inode, map.m_pblk) : 0;
1956 size = blks_to_bytes(inode, map.m_len);
1957 flags = 0;
1958
1959 if (compr_cluster) {
1960 flags = FIEMAP_EXTENT_ENCODED;
1961 count_in_cluster += map.m_len;
1962 if (count_in_cluster == cluster_size) {
1963 compr_cluster = false;
1964 size += blks_to_bytes(inode, 1);
1965 }
1966 } else if (map.m_flags & F2FS_MAP_UNWRITTEN) {
1967 flags = FIEMAP_EXTENT_UNWRITTEN;
1968 }
7f63eb77 1969
093f0bac
DJ		 1970 start_blk += bytes_to_blks(inode, size);
1971 }
7f63eb77 1972
9a950d52 1973prep_next:
7f63eb77
JK		 1974 cond_resched();
1975 if (fatal_signal_pending(current))
1976 ret = -EINTR;
1977 else
1978 goto next;
1979out:
1980 if (ret == 1)
1981 ret = 0;
1982
5955102c 1983 inode_unlock(inode);
7f63eb77 1984 return ret;
9ab70134
JK		 1985}
1986
95ae251f
EB		 1987static inline loff_t f2fs_readpage_limit(struct inode *inode)
1988{
1989 if (IS_ENABLED(CONFIG_FS_VERITY) &&
1990 (IS_VERITY(inode) || f2fs_verity_in_progress(inode)))
1991 return inode->i_sb->s_maxbytes;
1992
1993 return i_size_read(inode);
1994}
1995
2df0ab04
CY		 1996static int f2fs_read_single_page(struct inode *inode, struct page *page,
1997 unsigned nr_pages,
1998 struct f2fs_map_blocks *map,
1999 struct bio **bio_ret,
2000 sector_t *last_block_in_bio,
2001 bool is_readahead)
2002{
2003 struct bio *bio = *bio_ret;
43b9d4b4 2004 const unsigned blocksize = blks_to_bytes(inode, 1);
2df0ab04
CY		 2005 sector_t block_in_file;
2006 sector_t last_block;
2007 sector_t last_block_in_file;
2008 sector_t block_nr;
2009 int ret = 0;
2010
4969c06a 2011 block_in_file = (sector_t)page_index(page);
2df0ab04 2012 last_block = block_in_file + nr_pages;
43b9d4b4
JK		 2013 last_block_in_file = bytes_to_blks(inode,
2014 f2fs_readpage_limit(inode) + blocksize - 1);
2df0ab04
CY		 2015 if (last_block > last_block_in_file)
2016 last_block = last_block_in_file;
2017
2018 /* just zeroing out page which is beyond EOF */
2019 if (block_in_file >= last_block)
2020 goto zero_out;
2021 /*
2022 * Map blocks using the previous result first.
2023 */
2024 if ((map->m_flags & F2FS_MAP_MAPPED) &&
2025 block_in_file > map->m_lblk &&
2026 block_in_file < (map->m_lblk + map->m_len))
2027 goto got_it;
2028
2029 /*
2030 * Then do more f2fs_map_blocks() calls until we are
2031 * done with this page.
2032 */
2033 map->m_lblk = block_in_file;
2034 map->m_len = last_block - block_in_file;
2035
2036 ret = f2fs_map_blocks(inode, map, 0, F2FS_GET_BLOCK_DEFAULT);
2037 if (ret)
2038 goto out;
2039got_it:
2040 if ((map->m_flags & F2FS_MAP_MAPPED)) {
2041 block_nr = map->m_pblk + block_in_file - map->m_lblk;
2042 SetPageMappedToDisk(page);
2043
4969c06a
JK		 2044 if (!PageUptodate(page) && (!PageSwapCache(page) &&
2045 !cleancache_get_page(page))) {
2df0ab04
CY		 2046 SetPageUptodate(page);
2047 goto confused;
2048 }
2049
2050 if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), block_nr,
93770ab7 2051 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 2052 ret = -EFSCORRUPTED;
2df0ab04
CY		 2053 goto out;
2054 }
2055 } else {
2056zero_out:
2057 zero_user_segment(page, 0, PAGE_SIZE);
95ae251f
EB		 2058 if (f2fs_need_verity(inode, page->index) &&
2059 !fsverity_verify_page(page)) {
2060 ret = -EIO;
2061 goto out;
2062 }
2df0ab04
CY		 2063 if (!PageUptodate(page))
2064 SetPageUptodate(page);
2065 unlock_page(page);
2066 goto out;
2067 }
2068
2069 /*
2070 * This page will go to BIO. Do we need to send this
2071 * BIO off first?
2072 */
27aacd28
ST		 2073 if (bio && (!page_is_mergeable(F2FS_I_SB(inode), bio,
2074 *last_block_in_bio, block_nr) ||
2075 !f2fs_crypt_mergeable_bio(bio, inode, page->index, NULL))) {
2df0ab04
CY		 2076submit_and_realloc:
2077 __submit_bio(F2FS_I_SB(inode), bio, DATA);
2078 bio = NULL;
2079 }
2080 if (bio == NULL) {
2081 bio = f2fs_grab_read_bio(inode, block_nr, nr_pages,
0683728a 2082 is_readahead ? REQ_RAHEAD : 0, page->index,
7f59b277 2083 false);
2df0ab04
CY		 2084 if (IS_ERR(bio)) {
2085 ret = PTR_ERR(bio);
2086 bio = NULL;
2087 goto out;
2088 }
2089 }
2090
2091 /*
2092 * If the page is under writeback, we need to wait for
2093 * its completion to see the correct decrypted data.
2094 */
2095 f2fs_wait_on_block_writeback(inode, block_nr);
2096
2097 if (bio_add_page(bio, page, blocksize, 0) < blocksize)
2098 goto submit_and_realloc;
2099
2100 inc_page_count(F2FS_I_SB(inode), F2FS_RD_DATA);
8b83ac81 2101 f2fs_update_iostat(F2FS_I_SB(inode), FS_DATA_READ_IO, F2FS_BLKSIZE);
2df0ab04
CY		 2102 ClearPageError(page);
2103 *last_block_in_bio = block_nr;
2104 goto out;
2105confused:
2106 if (bio) {
2107 __submit_bio(F2FS_I_SB(inode), bio, DATA);
2108 bio = NULL;
2109 }
2110 unlock_page(page);
2111out:
2112 *bio_ret = bio;
2113 return ret;
2114}
2115
4c8ff709
CY		 2116#ifdef CONFIG_F2FS_FS_COMPRESSION
2117int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret,
2118 unsigned nr_pages, sector_t *last_block_in_bio,
0683728a 2119 bool is_readahead, bool for_write)
4c8ff709
CY		 2120{
2121 struct dnode_of_data dn;
2122 struct inode *inode = cc->inode;
2123 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2124 struct bio *bio = *bio_ret;
2125 unsigned int start_idx = cc->cluster_idx << cc->log_cluster_size;
2126 sector_t last_block_in_file;
43b9d4b4 2127 const unsigned blocksize = blks_to_bytes(inode, 1);
4c8ff709
CY		 2128 struct decompress_io_ctx *dic = NULL;
2129 int i;
2130 int ret = 0;
2131
2132 f2fs_bug_on(sbi, f2fs_cluster_is_empty(cc));
2133
43b9d4b4
JK		 2134 last_block_in_file = bytes_to_blks(inode,
2135 f2fs_readpage_limit(inode) + blocksize - 1);
4c8ff709
CY		 2136
2137 /* get rid of pages beyond EOF */
2138 for (i = 0; i < cc->cluster_size; i++) {
2139 struct page *page = cc->rpages[i];
2140
2141 if (!page)
2142 continue;
2143 if ((sector_t)page->index >= last_block_in_file) {
2144 zero_user_segment(page, 0, PAGE_SIZE);
2145 if (!PageUptodate(page))
2146 SetPageUptodate(page);
2147 } else if (!PageUptodate(page)) {
2148 continue;
2149 }
2150 unlock_page(page);
2151 cc->rpages[i] = NULL;
2152 cc->nr_rpages--;
2153 }
2154
2155 /* we are done since all pages are beyond EOF */
2156 if (f2fs_cluster_is_empty(cc))
2157 goto out;
2158
2159 set_new_dnode(&dn, inode, NULL, NULL, 0);
2160 ret = f2fs_get_dnode_of_data(&dn, start_idx, LOOKUP_NODE);
2161 if (ret)
2162 goto out;
2163
a86d27dd 2164 f2fs_bug_on(sbi, dn.data_blkaddr != COMPRESS_ADDR);
4c8ff709
CY		 2165
2166 for (i = 1; i < cc->cluster_size; i++) {
2167 block_t blkaddr;
2168
a2ced1ce 2169 blkaddr = data_blkaddr(dn.inode, dn.node_page,
4c8ff709
CY		 2170 dn.ofs_in_node + i);
2171
2172 if (!__is_valid_data_blkaddr(blkaddr))
2173 break;
2174
2175 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC)) {
2176 ret = -EFAULT;
2177 goto out_put_dnode;
2178 }
2179 cc->nr_cpages++;
2180 }
2181
2182 /* nothing to decompress */
2183 if (cc->nr_cpages == 0) {
2184 ret = 0;
2185 goto out_put_dnode;
2186 }
2187
2188 dic = f2fs_alloc_dic(cc);
2189 if (IS_ERR(dic)) {
2190 ret = PTR_ERR(dic);
2191 goto out_put_dnode;
2192 }
2193
6ce19aff 2194 for (i = 0; i < cc->nr_cpages; i++) {
4c8ff709
CY		 2195 struct page *page = dic->cpages[i];
2196 block_t blkaddr;
7f59b277 2197 struct bio_post_read_ctx *ctx;
4c8ff709 2198
a2ced1ce 2199 blkaddr = data_blkaddr(dn.inode, dn.node_page,
4c8ff709
CY		 2200 dn.ofs_in_node + i + 1);
2201
6ce19aff
CY		 2202 f2fs_wait_on_block_writeback(inode, blkaddr);
2203
2204 if (f2fs_load_compressed_page(sbi, page, blkaddr)) {
2205 if (atomic_dec_and_test(&dic->remaining_pages))
2206 f2fs_decompress_cluster(dic);
2207 continue;
2208 }
2209
27aacd28
ST		 2210 if (bio && (!page_is_mergeable(sbi, bio,
2211 *last_block_in_bio, blkaddr) ||
2212 !f2fs_crypt_mergeable_bio(bio, inode, page->index, NULL))) {
4c8ff709
CY		 2213submit_and_realloc:
2214 __submit_bio(sbi, bio, DATA);
2215 bio = NULL;
2216 }
2217
2218 if (!bio) {
2219 bio = f2fs_grab_read_bio(inode, blkaddr, nr_pages,
2220 is_readahead ? REQ_RAHEAD : 0,
7f59b277 2221 page->index, for_write);
4c8ff709
CY		 2222 if (IS_ERR(bio)) {
2223 ret = PTR_ERR(bio);
7f59b277 2224 f2fs_decompress_end_io(dic, ret);
4c8ff709 2225 f2fs_put_dnode(&dn);
f3494345 2226 *bio_ret = NULL;
4c8ff709
CY		 2227 return ret;
2228 }
2229 }
2230
4c8ff709
CY		 2231 if (bio_add_page(bio, page, blocksize, 0) < blocksize)
2232 goto submit_and_realloc;
2233
03382f1a 2234 ctx = bio->bi_private;
7f59b277
EB		 2235 ctx->enabled_steps |= STEP_DECOMPRESS;
2236 refcount_inc(&dic->refcnt);
03382f1a 2237
4c8ff709 2238 inc_page_count(sbi, F2FS_RD_DATA);
8b83ac81 2239 f2fs_update_iostat(sbi, FS_DATA_READ_IO, F2FS_BLKSIZE);
9c122384 2240 f2fs_update_iostat(sbi, FS_CDATA_READ_IO, F2FS_BLKSIZE);
4c8ff709
CY		 2241 ClearPageError(page);
2242 *last_block_in_bio = blkaddr;
2243 }
2244
2245 f2fs_put_dnode(&dn);
2246
2247 *bio_ret = bio;
2248 return 0;
2249
2250out_put_dnode:
2251 f2fs_put_dnode(&dn);
2252out:
7f59b277
EB		 2253 for (i = 0; i < cc->cluster_size; i++) {
2254 if (cc->rpages[i]) {
2255 ClearPageUptodate(cc->rpages[i]);
2256 ClearPageError(cc->rpages[i]);
2257 unlock_page(cc->rpages[i]);
2258 }
2259 }
4c8ff709
CY		 2260 *bio_ret = bio;
2261 return ret;
2262}
2263#endif
2264
f1e88660
JK		 2265/*
2266 * This function was originally taken from fs/mpage.c, and customized for f2fs.
2267 * Major change was from block_size == page_size in f2fs by default.
2268 */
e20a7693 2269static int f2fs_mpage_readpages(struct inode *inode,
23323196 2270 struct readahead_control *rac, struct page *page)
f1e88660
JK		 2271{
2272 struct bio *bio = NULL;
f1e88660 2273 sector_t last_block_in_bio = 0;
f1e88660 2274 struct f2fs_map_blocks map;
4c8ff709
CY		 2275#ifdef CONFIG_F2FS_FS_COMPRESSION
2276 struct compress_ctx cc = {
2277 .inode = inode,
2278 .log_cluster_size = F2FS_I(inode)->i_log_cluster_size,
2279 .cluster_size = F2FS_I(inode)->i_cluster_size,
2280 .cluster_idx = NULL_CLUSTER,
2281 .rpages = NULL,
2282 .cpages = NULL,
2283 .nr_rpages = 0,
2284 .nr_cpages = 0,
2285 };
2286#endif
23323196 2287 unsigned nr_pages = rac ? readahead_count(rac) : 1;
4c8ff709 2288 unsigned max_nr_pages = nr_pages;
2df0ab04 2289 int ret = 0;
f1e88660
JK		 2290
2291 map.m_pblk = 0;
2292 map.m_lblk = 0;
2293 map.m_len = 0;
2294 map.m_flags = 0;
da85985c 2295 map.m_next_pgofs = NULL;
c4020b2d 2296 map.m_next_extent = NULL;
d5097be5 2297 map.m_seg_type = NO_CHECK_TYPE;
f9d6d059 2298 map.m_may_create = false;
f1e88660 2299
736c0a74 2300 for (; nr_pages; nr_pages--) {
23323196
MWO		 2301 if (rac) {
2302 page = readahead_page(rac);
a83d50bc 2303 prefetchw(&page->flags);
f1e88660
JK		 2304 }
2305
4c8ff709
CY		 2306#ifdef CONFIG_F2FS_FS_COMPRESSION
2307 if (f2fs_compressed_file(inode)) {
2308 /* there are remained comressed pages, submit them */
2309 if (!f2fs_cluster_can_merge_page(&cc, page->index)) {
2310 ret = f2fs_read_multi_pages(&cc, &bio,
2311 max_nr_pages,
2312 &last_block_in_bio,
23323196 2313 rac != NULL, false);
8bfbfb0d 2314 f2fs_destroy_compress_ctx(&cc, false);
4c8ff709
CY		 2315 if (ret)
2316 goto set_error_page;
2317 }
2318 ret = f2fs_is_compressed_cluster(inode, page->index);
2319 if (ret < 0)
2320 goto set_error_page;
2321 else if (!ret)
2322 goto read_single_page;
2323
2324 ret = f2fs_init_compress_ctx(&cc);
2325 if (ret)
2326 goto set_error_page;
2327
2328 f2fs_compress_ctx_add_page(&cc, page);
2329
2330 goto next_page;
2331 }
2332read_single_page:
2333#endif
2334
2335 ret = f2fs_read_single_page(inode, page, max_nr_pages, &map,
23323196 2336 &bio, &last_block_in_bio, rac);
2df0ab04 2337 if (ret) {
4c8ff709
CY		 2338#ifdef CONFIG_F2FS_FS_COMPRESSION
2339set_error_page:
2340#endif
2df0ab04 2341 SetPageError(page);
09cbfeaf 2342 zero_user_segment(page, 0, PAGE_SIZE);
f1e88660 2343 unlock_page(page);
f1e88660 2344 }
23323196 2345#ifdef CONFIG_F2FS_FS_COMPRESSION
f1e88660 2346next_page:
23323196
MWO		 2347#endif
2348 if (rac)
09cbfeaf 2349 put_page(page);
4c8ff709
CY		 2350
2351#ifdef CONFIG_F2FS_FS_COMPRESSION
2352 if (f2fs_compressed_file(inode)) {
2353 /* last page */
2354 if (nr_pages == 1 && !f2fs_cluster_is_empty(&cc)) {
2355 ret = f2fs_read_multi_pages(&cc, &bio,
2356 max_nr_pages,
2357 &last_block_in_bio,
23323196 2358 rac != NULL, false);
8bfbfb0d 2359 f2fs_destroy_compress_ctx(&cc, false);
4c8ff709
CY		 2360 }
2361 }
2362#endif
f1e88660 2363 }
f1e88660 2364 if (bio)
4fc29c1a 2365 __submit_bio(F2FS_I_SB(inode), bio, DATA);
23323196 2366 return ret;
f1e88660
JK		 2367}
2368
eb47b800
JK		 2369static int f2fs_read_data_page(struct file *file, struct page *page)
2370{
4969c06a 2371 struct inode *inode = page_file_mapping(page)->host;
b3d208f9 2372 int ret = -EAGAIN;
9ffe0fb5 2373
c20e89cd
CY		 2374 trace_f2fs_readpage(page, DATA);
2375
4c8ff709
CY		 2376 if (!f2fs_is_compress_backend_ready(inode)) {
2377 unlock_page(page);
2378 return -EOPNOTSUPP;
2379 }
2380
e1c42045 2381 /* If the file has inline data, try to read it directly */
9ffe0fb5
HL		 2382 if (f2fs_has_inline_data(inode))
2383 ret = f2fs_read_inline_data(inode, page);
b3d208f9 2384 if (ret == -EAGAIN)
e20a7693 2385 ret = f2fs_mpage_readpages(inode, NULL, page);
9ffe0fb5 2386 return ret;
eb47b800
JK		 2387}
2388
23323196 2389static void f2fs_readahead(struct readahead_control *rac)
eb47b800 2390{
23323196 2391 struct inode *inode = rac->mapping->host;
b8c29400 2392
23323196 2393 trace_f2fs_readpages(inode, readahead_index(rac), readahead_count(rac));
9ffe0fb5 2394
4c8ff709 2395 if (!f2fs_is_compress_backend_ready(inode))
23323196 2396 return;
4c8ff709 2397
9ffe0fb5
HL		 2398 /* If the file has inline data, skip readpages */
2399 if (f2fs_has_inline_data(inode))
23323196 2400 return;
9ffe0fb5 2401
e20a7693 2402 f2fs_mpage_readpages(inode, rac, NULL);
eb47b800
JK		 2403}
2404
4c8ff709 2405int f2fs_encrypt_one_page(struct f2fs_io_info *fio)
7eab0c0d
HP		 2406{
2407 struct inode *inode = fio->page->mapping->host;
4c8ff709 2408 struct page *mpage, *page;
7eab0c0d
HP		 2409 gfp_t gfp_flags = GFP_NOFS;
2410
1958593e 2411 if (!f2fs_encrypted_file(inode))
7eab0c0d
HP		 2412 return 0;
2413
4c8ff709
CY		 2414 page = fio->compressed_page ? fio->compressed_page : fio->page;
2415
6dbb1796 2416 /* wait for GCed page writeback via META_MAPPING */
0ded69f6 2417 f2fs_wait_on_block_writeback(inode, fio->old_blkaddr);
7eab0c0d 2418
27aacd28
ST		 2419 if (fscrypt_inode_uses_inline_crypto(inode))
2420 return 0;
2421
7eab0c0d 2422retry_encrypt:
4c8ff709
CY		 2423 fio->encrypted_page = fscrypt_encrypt_pagecache_blocks(page,
2424 PAGE_SIZE, 0, gfp_flags);
6aa58d8a
CY		 2425 if (IS_ERR(fio->encrypted_page)) {
2426 /* flush pending IOs and wait for a while in the ENOMEM case */
2427 if (PTR_ERR(fio->encrypted_page) == -ENOMEM) {
2428 f2fs_flush_merged_writes(fio->sbi);
5df7731f 2429 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
6aa58d8a
CY		 2430 gfp_flags |= __GFP_NOFAIL;
2431 goto retry_encrypt;
2432 }
2433 return PTR_ERR(fio->encrypted_page);
2434 }
7eab0c0d 2435
6aa58d8a
CY		 2436 mpage = find_lock_page(META_MAPPING(fio->sbi), fio->old_blkaddr);
2437 if (mpage) {
2438 if (PageUptodate(mpage))
2439 memcpy(page_address(mpage),
2440 page_address(fio->encrypted_page), PAGE_SIZE);
2441 f2fs_put_page(mpage, 1);
7eab0c0d 2442 }
6aa58d8a 2443 return 0;
7eab0c0d
HP		 2444}
2445
bb9e3bb8
CY		 2446static inline bool check_inplace_update_policy(struct inode *inode,
2447 struct f2fs_io_info *fio)
7eab0c0d 2448{
bb9e3bb8
CY		 2449 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2450 unsigned int policy = SM_I(sbi)->ipu_policy;
7eab0c0d 2451
bb9e3bb8
CY		 2452 if (policy & (0x1 << F2FS_IPU_FORCE))
2453 return true;
4d57b86d 2454 if (policy & (0x1 << F2FS_IPU_SSR) && f2fs_need_SSR(sbi))
bb9e3bb8
CY		 2455 return true;
2456 if (policy & (0x1 << F2FS_IPU_UTIL) &&
2457 utilization(sbi) > SM_I(sbi)->min_ipu_util)
2458 return true;
4d57b86d 2459 if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && f2fs_need_SSR(sbi) &&
bb9e3bb8
CY		 2460 utilization(sbi) > SM_I(sbi)->min_ipu_util)
2461 return true;
2462
2463 /*
2464 * IPU for rewrite async pages
2465 */
2466 if (policy & (0x1 << F2FS_IPU_ASYNC) &&
2467 fio && fio->op == REQ_OP_WRITE &&
2468 !(fio->op_flags & REQ_SYNC) &&
62230e0d 2469 !IS_ENCRYPTED(inode))
bb9e3bb8
CY		 2470 return true;
2471
2472 /* this is only set during fdatasync */
2473 if (policy & (0x1 << F2FS_IPU_FSYNC) &&
2474 is_inode_flag_set(inode, FI_NEED_IPU))
2475 return true;
2476
4354994f
DR		 2477 if (unlikely(fio && is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2478 !f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2479 return true;
2480
bb9e3bb8
CY		 2481 return false;
2482}
2483
4d57b86d 2484bool f2fs_should_update_inplace(struct inode *inode, struct f2fs_io_info *fio)
bb9e3bb8 2485{
859fca6b
CY		 2486 /* swap file is migrating in aligned write mode */
2487 if (is_inode_flag_set(inode, FI_ALIGNED_WRITE))
2488 return false;
2489
1ad71a27
JK		 2490 if (f2fs_is_pinned_file(inode))
2491 return true;
bb9e3bb8
CY		 2492
2493 /* if this is cold file, we should overwrite to avoid fragmentation */
2494 if (file_is_cold(inode))
2495 return true;
2496
2497 return check_inplace_update_policy(inode, fio);
2498}
2499
4d57b86d 2500bool f2fs_should_update_outplace(struct inode *inode, struct f2fs_io_info *fio)
bb9e3bb8
CY		 2501{
2502 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2503
b0332a0f 2504 if (f2fs_lfs_mode(sbi))
bb9e3bb8
CY		 2505 return true;
2506 if (S_ISDIR(inode->i_mode))
2507 return true;
af033b2a
CY		 2508 if (IS_NOQUOTA(inode))
2509 return true;
bb9e3bb8
CY		 2510 if (f2fs_is_atomic_file(inode))
2511 return true;
1ffc8f5f
JK		 2512 if (is_sbi_flag_set(sbi, SBI_NEED_FSCK))
2513 return true;
859fca6b
CY		 2514
2515 /* swap file is migrating in aligned write mode */
2516 if (is_inode_flag_set(inode, FI_ALIGNED_WRITE))
2517 return true;
2518
bb9e3bb8 2519 if (fio) {
b763f3be 2520 if (page_private_gcing(fio->page))
bb9e3bb8 2521 return true;
b763f3be 2522 if (page_private_dummy(fio->page))
bb9e3bb8 2523 return true;
4354994f
DR		 2524 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2525 f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2526 return true;
bb9e3bb8
CY		 2527 }
2528 return false;
2529}
2530
7eab0c0d
HP		 2531static inline bool need_inplace_update(struct f2fs_io_info *fio)
2532{
2533 struct inode *inode = fio->page->mapping->host;
2534
4d57b86d 2535 if (f2fs_should_update_outplace(inode, fio))
7eab0c0d
HP		 2536 return false;
2537
4d57b86d 2538 return f2fs_should_update_inplace(inode, fio);
7eab0c0d
HP		 2539}
2540
4d57b86d 2541int f2fs_do_write_data_page(struct f2fs_io_info *fio)
eb47b800 2542{
05ca3632 2543 struct page *page = fio->page;
eb47b800 2544 struct inode *inode = page->mapping->host;
eb47b800 2545 struct dnode_of_data dn;
e959c8f5 2546 struct extent_info ei = {0,0,0};
7735730d 2547 struct node_info ni;
e959c8f5 2548 bool ipu_force = false;
eb47b800
JK		 2549 int err = 0;
2550
2551 set_new_dnode(&dn, inode, NULL, NULL, 0);
e959c8f5
HP		 2552 if (need_inplace_update(fio) &&
2553 f2fs_lookup_extent_cache(inode, page->index, &ei)) {
2554 fio->old_blkaddr = ei.blk + page->index - ei.fofs;
a817737e 2555
c9b60788 2556 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
93770ab7 2557 DATA_GENERIC_ENHANCE))
10f966bb 2558 return -EFSCORRUPTED;
c9b60788
CY		 2559
2560 ipu_force = true;
2561 fio->need_lock = LOCK_DONE;
2562 goto got_it;
e959c8f5 2563 }
279d6df2 2564
d29460e5
JK		 2565 /* Deadlock due to between page->lock and f2fs_lock_op */
2566 if (fio->need_lock == LOCK_REQ && !f2fs_trylock_op(fio->sbi))
2567 return -EAGAIN;
279d6df2 2568
4d57b86d 2569 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
eb47b800 2570 if (err)
279d6df2 2571 goto out;
eb47b800 2572
28bc106b 2573 fio->old_blkaddr = dn.data_blkaddr;
eb47b800
JK		 2574
2575 /* This page is already truncated */
7a9d7548 2576 if (fio->old_blkaddr == NULL_ADDR) {
2bca1e23 2577 ClearPageUptodate(page);
b763f3be 2578 clear_page_private_gcing(page);
eb47b800 2579 goto out_writepage;
2bca1e23 2580 }
e959c8f5 2581got_it:
c9b60788
CY		 2582 if (__is_valid_data_blkaddr(fio->old_blkaddr) &&
2583 !f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
93770ab7 2584 DATA_GENERIC_ENHANCE)) {
10f966bb 2585 err = -EFSCORRUPTED;
c9b60788
CY		 2586 goto out_writepage;
2587 }
eb47b800
JK		 2588 /*
2589 * If current allocation needs SSR,
2590 * it had better in-place writes for updated data.
2591 */
93770ab7
CY		 2592 if (ipu_force ||
2593 (__is_valid_data_blkaddr(fio->old_blkaddr) &&
7b525dd0 2594 need_inplace_update(fio))) {
4c8ff709 2595 err = f2fs_encrypt_one_page(fio);
cc15620b
JK		 2596 if (err)
2597 goto out_writepage;
2598
2599 set_page_writeback(page);
17c50035 2600 ClearPageError(page);
279d6df2 2601 f2fs_put_dnode(&dn);
cc15620b 2602 if (fio->need_lock == LOCK_REQ)
279d6df2 2603 f2fs_unlock_op(fio->sbi);
4d57b86d 2604 err = f2fs_inplace_write_data(fio);
6492a335 2605 if (err) {
27aacd28 2606 if (fscrypt_inode_uses_fs_layer_crypto(inode))
d2d0727b 2607 fscrypt_finalize_bounce_page(&fio->encrypted_page);
6492a335
CY		 2608 if (PageWriteback(page))
2609 end_page_writeback(page);
cd23ffa9
CY		 2610 } else {
2611 set_inode_flag(inode, FI_UPDATE_WRITE);
6492a335 2612 }
7eab0c0d 2613 trace_f2fs_do_write_data_page(fio->page, IPU);
279d6df2 2614 return err;
eb47b800 2615 }
279d6df2 2616
cc15620b
JK		 2617 if (fio->need_lock == LOCK_RETRY) {
2618 if (!f2fs_trylock_op(fio->sbi)) {
2619 err = -EAGAIN;
2620 goto out_writepage;
2621 }
2622 fio->need_lock = LOCK_REQ;
2623 }
2624
7735730d
CY		 2625 err = f2fs_get_node_info(fio->sbi, dn.nid, &ni);
2626 if (err)
2627 goto out_writepage;
2628
2629 fio->version = ni.version;
2630
4c8ff709 2631 err = f2fs_encrypt_one_page(fio);
cc15620b
JK		 2632 if (err)
2633 goto out_writepage;
2634
2635 set_page_writeback(page);
17c50035 2636 ClearPageError(page);
cc15620b 2637
4c8ff709
CY		 2638 if (fio->compr_blocks && fio->old_blkaddr == COMPRESS_ADDR)
2639 f2fs_i_compr_blocks_update(inode, fio->compr_blocks - 1, false);
2640
279d6df2 2641 /* LFS mode write path */
4d57b86d 2642 f2fs_outplace_write_data(&dn, fio);
279d6df2
HP		 2643 trace_f2fs_do_write_data_page(page, OPU);
2644 set_inode_flag(inode, FI_APPEND_WRITE);
2645 if (page->index == 0)
2646 set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
eb47b800
JK		 2647out_writepage:
2648 f2fs_put_dnode(&dn);
279d6df2 2649out:
cc15620b 2650 if (fio->need_lock == LOCK_REQ)
279d6df2 2651 f2fs_unlock_op(fio->sbi);
eb47b800
JK		 2652 return err;
2653}
2654
4c8ff709 2655int f2fs_write_single_data_page(struct page *page, int *submitted,
8648de2c
CY		 2656 struct bio **bio,
2657 sector_t *last_block,
b0af6d49 2658 struct writeback_control *wbc,
4c8ff709 2659 enum iostat_type io_type,
3afae09f
CY		 2660 int compr_blocks,
2661 bool allow_balance)
eb47b800
JK		 2662{
2663 struct inode *inode = page->mapping->host;
4081363f 2664 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
eb47b800 2665 loff_t i_size = i_size_read(inode);
4c8ff709 2666 const pgoff_t end_index = ((unsigned long long)i_size)
09cbfeaf 2667 >> PAGE_SHIFT;
1f0d5c91 2668 loff_t psize = (loff_t)(page->index + 1) << PAGE_SHIFT;
9ffe0fb5 2669 unsigned offset = 0;
39936837 2670 bool need_balance_fs = false;
eb47b800 2671 int err = 0;
458e6197 2672 struct f2fs_io_info fio = {
05ca3632 2673 .sbi = sbi,
39d787be 2674 .ino = inode->i_ino,
458e6197 2675 .type = DATA,
04d328de 2676 .op = REQ_OP_WRITE,
7637241e 2677 .op_flags = wbc_to_write_flags(wbc),
e959c8f5 2678 .old_blkaddr = NULL_ADDR,
05ca3632 2679 .page = page,
4375a336 2680 .encrypted_page = NULL,
d68f735b 2681 .submitted = false,
4c8ff709 2682 .compr_blocks = compr_blocks,
cc15620b 2683 .need_lock = LOCK_RETRY,
b0af6d49 2684 .io_type = io_type,
578c6478 2685 .io_wbc = wbc,
8648de2c
CY		 2686 .bio = bio,
2687 .last_block = last_block,
458e6197 2688 };
eb47b800 2689
ecda0de3
CY		 2690 trace_f2fs_writepage(page, DATA);
2691
db198ae0
CY		 2692 /* we should bypass data pages to proceed the kworkder jobs */
2693 if (unlikely(f2fs_cp_error(sbi))) {
2694 mapping_set_error(page->mapping, -EIO);
1174abfd
CY		 2695 /*
2696 * don't drop any dirty dentry pages for keeping lastest
2697 * directory structure.
2698 */
2699 if (S_ISDIR(inode->i_mode))
2700 goto redirty_out;
db198ae0
CY		 2701 goto out;
2702 }
2703
0771fcc7
CY		 2704 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
2705 goto redirty_out;
2706
4c8ff709
CY		 2707 if (page->index < end_index ||
2708 f2fs_verity_in_progress(inode) ||
2709 compr_blocks)
39936837 2710 goto write;
eb47b800
JK		 2711
2712 /*
2713 * If the offset is out-of-range of file size,
2714 * this page does not have to be written to disk.
2715 */
09cbfeaf 2716 offset = i_size & (PAGE_SIZE - 1);
76f60268 2717 if ((page->index >= end_index + 1) || !offset)
39936837 2718 goto out;
eb47b800 2719
09cbfeaf 2720 zero_user_segment(page, offset, PAGE_SIZE);
39936837 2721write:
1e84371f
JK		 2722 if (f2fs_is_drop_cache(inode))
2723 goto out;
e6e5f561
JK		 2724 /* we should not write 0'th page having journal header */
2725 if (f2fs_is_volatile_file(inode) && (!page->index ||
2726 (!wbc->for_reclaim &&
4d57b86d 2727 f2fs_available_free_memory(sbi, BASE_CHECK))))
1e84371f 2728 goto redirty_out;
eb47b800 2729
435cbab9
JK		 2730 /* Dentry/quota blocks are controlled by checkpoint */
2731 if (S_ISDIR(inode->i_mode) || IS_NOQUOTA(inode)) {
79963d96
CY		 2732 /*
2733 * We need to wait for node_write to avoid block allocation during
2734 * checkpoint. This can only happen to quota writes which can cause
2735 * the below discard race condition.
2736 */
2737 if (IS_NOQUOTA(inode))
2738 down_read(&sbi->node_write);
2739
cc15620b 2740 fio.need_lock = LOCK_DONE;
4d57b86d 2741 err = f2fs_do_write_data_page(&fio);
79963d96
CY		 2742
2743 if (IS_NOQUOTA(inode))
2744 up_read(&sbi->node_write);
2745
8618b881
JK		 2746 goto done;
2747 }
9ffe0fb5 2748
8618b881 2749 if (!wbc->for_reclaim)
39936837 2750 need_balance_fs = true;
7f3037a5 2751 else if (has_not_enough_free_secs(sbi, 0, 0))
39936837 2752 goto redirty_out;
ef095d19
JK		 2753 else
2754 set_inode_flag(inode, FI_HOT_DATA);
eb47b800 2755
b3d208f9 2756 err = -EAGAIN;
dd7b2333 2757 if (f2fs_has_inline_data(inode)) {
b3d208f9 2758 err = f2fs_write_inline_data(inode, page);
dd7b2333
YH		 2759 if (!err)
2760 goto out;
2761 }
279d6df2 2762
cc15620b 2763 if (err == -EAGAIN) {
4d57b86d 2764 err = f2fs_do_write_data_page(&fio);
cc15620b
JK		 2765 if (err == -EAGAIN) {
2766 fio.need_lock = LOCK_REQ;
4d57b86d 2767 err = f2fs_do_write_data_page(&fio);
cc15620b
JK		 2768 }
2769 }
a0d00fad 2770
eb449797
CY		 2771 if (err) {
2772 file_set_keep_isize(inode);
2773 } else {
c10c9820 2774 spin_lock(&F2FS_I(inode)->i_size_lock);
eb449797
CY		 2775 if (F2FS_I(inode)->last_disk_size < psize)
2776 F2FS_I(inode)->last_disk_size = psize;
c10c9820 2777 spin_unlock(&F2FS_I(inode)->i_size_lock);
eb449797 2778 }
279d6df2 2779
8618b881
JK		 2780done:
2781 if (err && err != -ENOENT)
2782 goto redirty_out;
eb47b800 2783
39936837 2784out:
a7ffdbe2 2785 inode_dec_dirty_pages(inode);
2baf0781 2786 if (err) {
2bca1e23 2787 ClearPageUptodate(page);
b763f3be 2788 clear_page_private_gcing(page);
2baf0781 2789 }
0c3a5797
CY		 2790
2791 if (wbc->for_reclaim) {
bab475c5 2792 f2fs_submit_merged_write_cond(sbi, NULL, page, 0, DATA);
ef095d19 2793 clear_inode_flag(inode, FI_HOT_DATA);
4d57b86d 2794 f2fs_remove_dirty_inode(inode);
d68f735b 2795 submitted = NULL;
0c3a5797 2796 }
eb47b800 2797 unlock_page(page);
186857c5 2798 if (!S_ISDIR(inode->i_mode) && !IS_NOQUOTA(inode) &&
3afae09f 2799 !F2FS_I(inode)->cp_task && allow_balance)
a7881893 2800 f2fs_balance_fs(sbi, need_balance_fs);
0c3a5797 2801
d68f735b 2802 if (unlikely(f2fs_cp_error(sbi))) {
b9109b0e 2803 f2fs_submit_merged_write(sbi, DATA);
0b20fcec 2804 f2fs_submit_merged_ipu_write(sbi, bio, NULL);
d68f735b
JK		 2805 submitted = NULL;
2806 }
2807
2808 if (submitted)
4c8ff709 2809 *submitted = fio.submitted ? 1 : 0;
0c3a5797 2810
eb47b800
JK		 2811 return 0;
2812
eb47b800 2813redirty_out:
76f60268 2814 redirty_page_for_writepage(wbc, page);
5b19d284
JK		 2815 /*
2816 * pageout() in MM traslates EAGAIN, so calls handle_write_error()
2817 * -> mapping_set_error() -> set_bit(AS_EIO, ...).
2818 * file_write_and_wait_range() will see EIO error, which is critical
2819 * to return value of fsync() followed by atomic_write failure to user.
2820 */
2821 if (!err || wbc->for_reclaim)
0002b61b 2822 return AOP_WRITEPAGE_ACTIVATE;
b230e6ca
JK		 2823 unlock_page(page);
2824 return err;
fa9150a8
NJ		 2825}
2826
f566bae8
JK		 2827static int f2fs_write_data_page(struct page *page,
2828 struct writeback_control *wbc)
2829{
4c8ff709
CY		 2830#ifdef CONFIG_F2FS_FS_COMPRESSION
2831 struct inode *inode = page->mapping->host;
2832
2833 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2834 goto out;
2835
2836 if (f2fs_compressed_file(inode)) {
2837 if (f2fs_is_compressed_cluster(inode, page->index)) {
2838 redirty_page_for_writepage(wbc, page);
2839 return AOP_WRITEPAGE_ACTIVATE;
2840 }
2841 }
2842out:
2843#endif
2844
2845 return f2fs_write_single_data_page(page, NULL, NULL, NULL,
3afae09f 2846 wbc, FS_DATA_IO, 0, true);
f566bae8
JK		 2847}
2848
8f46dcae
CY		 2849/*
2850 * This function was copied from write_cche_pages from mm/page-writeback.c.
2851 * The major change is making write step of cold data page separately from
2852 * warm/hot data page.
2853 */
2854static int f2fs_write_cache_pages(struct address_space *mapping,
b0af6d49
CY		 2855 struct writeback_control *wbc,
2856 enum iostat_type io_type)
8f46dcae
CY		 2857{
2858 int ret = 0;
4c8ff709 2859 int done = 0, retry = 0;
8f46dcae 2860 struct pagevec pvec;
c29fd0c0 2861 struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
8648de2c
CY		 2862 struct bio *bio = NULL;
2863 sector_t last_block;
4c8ff709
CY		 2864#ifdef CONFIG_F2FS_FS_COMPRESSION
2865 struct inode *inode = mapping->host;
2866 struct compress_ctx cc = {
2867 .inode = inode,
2868 .log_cluster_size = F2FS_I(inode)->i_log_cluster_size,
2869 .cluster_size = F2FS_I(inode)->i_cluster_size,
2870 .cluster_idx = NULL_CLUSTER,
2871 .rpages = NULL,
2872 .nr_rpages = 0,
2873 .cpages = NULL,
2874 .rbuf = NULL,
2875 .cbuf = NULL,
2876 .rlen = PAGE_SIZE * F2FS_I(inode)->i_cluster_size,
2877 .private = NULL,
2878 };
2879#endif
8f46dcae 2880 int nr_pages;
8f46dcae
CY		 2881 pgoff_t index;
2882 pgoff_t end; /* Inclusive */
2883 pgoff_t done_index;
8f46dcae 2884 int range_whole = 0;
10bbd235 2885 xa_mark_t tag;
bab475c5 2886 int nwritten = 0;
4c8ff709
CY		 2887 int submitted = 0;
2888 int i;
8f46dcae 2889
86679820 2890 pagevec_init(&pvec);
46ae957f 2891
ef095d19
JK		 2892 if (get_dirty_pages(mapping->host) <=
2893 SM_I(F2FS_M_SB(mapping))->min_hot_blocks)
2894 set_inode_flag(mapping->host, FI_HOT_DATA);
2895 else
2896 clear_inode_flag(mapping->host, FI_HOT_DATA);
2897
8f46dcae 2898 if (wbc->range_cyclic) {
4df7a75f 2899 index = mapping->writeback_index; /* prev offset */
8f46dcae
CY		 2900 end = -1;
2901 } else {
09cbfeaf
KS		 2902 index = wbc->range_start >> PAGE_SHIFT;
2903 end = wbc->range_end >> PAGE_SHIFT;
8f46dcae
CY		 2904 if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
2905 range_whole = 1;
8f46dcae
CY		 2906 }
2907 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2908 tag = PAGECACHE_TAG_TOWRITE;
2909 else
2910 tag = PAGECACHE_TAG_DIRTY;
2911retry:
4c8ff709 2912 retry = 0;
8f46dcae
CY		 2913 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2914 tag_pages_for_writeback(mapping, index, end);
2915 done_index = index;
4c8ff709 2916 while (!done && !retry && (index <= end)) {
69c4f35d 2917 nr_pages = pagevec_lookup_range_tag(&pvec, mapping, &index, end,
67fd707f 2918 tag);
8f46dcae
CY		 2919 if (nr_pages == 0)
2920 break;
2921
2922 for (i = 0; i < nr_pages; i++) {
2923 struct page *page = pvec.pages[i];
4c8ff709
CY		 2924 bool need_readd;
2925readd:
2926 need_readd = false;
2927#ifdef CONFIG_F2FS_FS_COMPRESSION
2928 if (f2fs_compressed_file(inode)) {
2929 ret = f2fs_init_compress_ctx(&cc);
2930 if (ret) {
2931 done = 1;
2932 break;
2933 }
2934
2935 if (!f2fs_cluster_can_merge_page(&cc,
2936 page->index)) {
2937 ret = f2fs_write_multi_pages(&cc,
2938 &submitted, wbc, io_type);
2939 if (!ret)
2940 need_readd = true;
2941 goto result;
2942 }
8f46dcae 2943
4c8ff709
CY		 2944 if (unlikely(f2fs_cp_error(sbi)))
2945 goto lock_page;
2946
2947 if (f2fs_cluster_is_empty(&cc)) {
2948 void *fsdata = NULL;
2949 struct page *pagep;
2950 int ret2;
2951
2952 ret2 = f2fs_prepare_compress_overwrite(
2953 inode, &pagep,
2954 page->index, &fsdata);
2955 if (ret2 < 0) {
2956 ret = ret2;
2957 done = 1;
2958 break;
2959 } else if (ret2 &&
2960 !f2fs_compress_write_end(inode,
2961 fsdata, page->index,
2962 1)) {
2963 retry = 1;
2964 break;
2965 }
2966 } else {
2967 goto lock_page;
2968 }
2969 }
2970#endif
f8de4331 2971 /* give a priority to WB_SYNC threads */
c29fd0c0 2972 if (atomic_read(&sbi->wb_sync_req[DATA]) &&
f8de4331
CY		 2973 wbc->sync_mode == WB_SYNC_NONE) {
2974 done = 1;
2975 break;
2976 }
4c8ff709
CY		 2977#ifdef CONFIG_F2FS_FS_COMPRESSION
2978lock_page:
2979#endif
8f46dcae 2980 done_index = page->index;
d29460e5 2981retry_write:
8f46dcae
CY		 2982 lock_page(page);
2983
2984 if (unlikely(page->mapping != mapping)) {
2985continue_unlock:
2986 unlock_page(page);
2987 continue;
2988 }
2989
2990 if (!PageDirty(page)) {
2991 /* someone wrote it for us */
2992 goto continue_unlock;
2993 }
2994
8f46dcae 2995 if (PageWriteback(page)) {
0b20fcec 2996 if (wbc->sync_mode != WB_SYNC_NONE)
fec1d657 2997 f2fs_wait_on_page_writeback(page,
bae0ee7a 2998 DATA, true, true);
0b20fcec 2999 else
8f46dcae
CY		 3000 goto continue_unlock;
3001 }
3002
8f46dcae
CY		 3003 if (!clear_page_dirty_for_io(page))
3004 goto continue_unlock;
3005
4c8ff709
CY		 3006#ifdef CONFIG_F2FS_FS_COMPRESSION
3007 if (f2fs_compressed_file(inode)) {
3008 get_page(page);
3009 f2fs_compress_ctx_add_page(&cc, page);
3010 continue;
3011 }
3012#endif
3013 ret = f2fs_write_single_data_page(page, &submitted,
3afae09f
CY		 3014 &bio, &last_block, wbc, io_type,
3015 0, true);
4c8ff709
CY		 3016 if (ret == AOP_WRITEPAGE_ACTIVATE)
3017 unlock_page(page);
3018#ifdef CONFIG_F2FS_FS_COMPRESSION
3019result:
3020#endif
3021 nwritten += submitted;
3022 wbc->nr_to_write -= submitted;
3023
8f46dcae 3024 if (unlikely(ret)) {
0002b61b
CY		 3025 /*
3026 * keep nr_to_write, since vfs uses this to
3027 * get # of written pages.
3028 */
3029 if (ret == AOP_WRITEPAGE_ACTIVATE) {
0002b61b 3030 ret = 0;
4c8ff709 3031 goto next;
d29460e5
JK		 3032 } else if (ret == -EAGAIN) {
3033 ret = 0;
3034 if (wbc->sync_mode == WB_SYNC_ALL) {
3035 cond_resched();
3036 congestion_wait(BLK_RW_ASYNC,
5df7731f 3037 DEFAULT_IO_TIMEOUT);
d29460e5
JK		 3038 goto retry_write;
3039 }
4c8ff709 3040 goto next;
0002b61b 3041 }
b230e6ca
JK		 3042 done_index = page->index + 1;
3043 done = 1;
3044 break;
8f46dcae
CY		 3045 }
3046
4c8ff709 3047 if (wbc->nr_to_write <= 0 &&
687de7f1 3048 wbc->sync_mode == WB_SYNC_NONE) {
8f46dcae
CY		 3049 done = 1;
3050 break;
3051 }
4c8ff709
CY		 3052next:
3053 if (need_readd)
3054 goto readd;
8f46dcae
CY		 3055 }
3056 pagevec_release(&pvec);
3057 cond_resched();
3058 }
4c8ff709
CY		 3059#ifdef CONFIG_F2FS_FS_COMPRESSION
3060 /* flush remained pages in compress cluster */
3061 if (f2fs_compressed_file(inode) && !f2fs_cluster_is_empty(&cc)) {
3062 ret = f2fs_write_multi_pages(&cc, &submitted, wbc, io_type);
3063 nwritten += submitted;
3064 wbc->nr_to_write -= submitted;
3065 if (ret) {
3066 done = 1;
3067 retry = 0;
3068 }
3069 }
adfc6943 3070 if (f2fs_compressed_file(inode))
8bfbfb0d 3071 f2fs_destroy_compress_ctx(&cc, false);
4c8ff709 3072#endif
e78790f8 3073 if (retry) {
8f46dcae 3074 index = 0;
e78790f8 3075 end = -1;
8f46dcae
CY		 3076 goto retry;
3077 }
e78790f8
ST		 3078 if (wbc->range_cyclic && !done)
3079 done_index = 0;
8f46dcae
CY		 3080 if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
3081 mapping->writeback_index = done_index;
3082
bab475c5 3083 if (nwritten)
b9109b0e 3084 f2fs_submit_merged_write_cond(F2FS_M_SB(mapping), mapping->host,
bab475c5 3085 NULL, 0, DATA);
8648de2c
CY		 3086 /* submit cached bio of IPU write */
3087 if (bio)
0b20fcec 3088 f2fs_submit_merged_ipu_write(sbi, &bio, NULL);
6ca56ca4 3089
8f46dcae
CY		 3090 return ret;
3091}
3092
853137ce
JK		 3093static inline bool __should_serialize_io(struct inode *inode,
3094 struct writeback_control *wbc)
3095{
b13f67ff
CY		 3096 /* to avoid deadlock in path of data flush */
3097 if (F2FS_I(inode)->cp_task)
3098 return false;
3099
853137ce
JK		 3100 if (!S_ISREG(inode->i_mode))
3101 return false;
af033b2a
CY		 3102 if (IS_NOQUOTA(inode))
3103 return false;
b13f67ff 3104
602a16d5 3105 if (f2fs_need_compress_data(inode))
b13f67ff 3106 return true;
853137ce
JK		 3107 if (wbc->sync_mode != WB_SYNC_ALL)
3108 return true;
3109 if (get_dirty_pages(inode) >= SM_I(F2FS_I_SB(inode))->min_seq_blocks)
3110 return true;
3111 return false;
3112}
3113
fc99fe27 3114static int __f2fs_write_data_pages(struct address_space *mapping,
b0af6d49
CY		 3115 struct writeback_control *wbc,
3116 enum iostat_type io_type)
eb47b800
JK		 3117{
3118 struct inode *inode = mapping->host;
4081363f 3119 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
9dfa1baf 3120 struct blk_plug plug;
eb47b800 3121 int ret;
853137ce 3122 bool locked = false;
eb47b800 3123
cfb185a1 3124 /* deal with chardevs and other special file */
3125 if (!mapping->a_ops->writepage)
3126 return 0;
3127
6a290544
CY		 3128 /* skip writing if there is no dirty page in this inode */
3129 if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE)
3130 return 0;
3131
0771fcc7
CY		 3132 /* during POR, we don't need to trigger writepage at all. */
3133 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
3134 goto skip_write;
3135
af033b2a
CY		 3136 if ((S_ISDIR(inode->i_mode) || IS_NOQUOTA(inode)) &&
3137 wbc->sync_mode == WB_SYNC_NONE &&
a1257023 3138 get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) &&
4d57b86d 3139 f2fs_available_free_memory(sbi, DIRTY_DENTS))
a1257023
JK		 3140 goto skip_write;
3141
d323d005 3142 /* skip writing during file defragment */
91942321 3143 if (is_inode_flag_set(inode, FI_DO_DEFRAG))
d323d005
CY		 3144 goto skip_write;
3145
d31c7c3f
YH		 3146 trace_f2fs_writepages(mapping->host, wbc, DATA);
3147
687de7f1
JK		 3148 /* to avoid spliting IOs due to mixed WB_SYNC_ALL and WB_SYNC_NONE */
3149 if (wbc->sync_mode == WB_SYNC_ALL)
c29fd0c0
CY		 3150 atomic_inc(&sbi->wb_sync_req[DATA]);
3151 else if (atomic_read(&sbi->wb_sync_req[DATA]))
687de7f1
JK		 3152 goto skip_write;
3153
853137ce
JK		 3154 if (__should_serialize_io(inode, wbc)) {
3155 mutex_lock(&sbi->writepages);
3156 locked = true;
3157 }
3158
9dfa1baf 3159 blk_start_plug(&plug);
b0af6d49 3160 ret = f2fs_write_cache_pages(mapping, wbc, io_type);
9dfa1baf 3161 blk_finish_plug(&plug);
687de7f1 3162
853137ce
JK		 3163 if (locked)
3164 mutex_unlock(&sbi->writepages);
3165
687de7f1 3166 if (wbc->sync_mode == WB_SYNC_ALL)
c29fd0c0 3167 atomic_dec(&sbi->wb_sync_req[DATA]);
28ea6162
JK		 3168 /*
3169 * if some pages were truncated, we cannot guarantee its mapping->host
3170 * to detect pending bios.
3171 */
458e6197 3172
4d57b86d 3173 f2fs_remove_dirty_inode(inode);
eb47b800 3174 return ret;
d3baf95d
JK		 3175
3176skip_write:
a7ffdbe2 3177 wbc->pages_skipped += get_dirty_pages(inode);
d31c7c3f 3178 trace_f2fs_writepages(mapping->host, wbc, DATA);
d3baf95d 3179 return 0;
eb47b800
JK		 3180}
3181
b0af6d49
CY		 3182static int f2fs_write_data_pages(struct address_space *mapping,
3183 struct writeback_control *wbc)
3184{
3185 struct inode *inode = mapping->host;
3186
3187 return __f2fs_write_data_pages(mapping, wbc,
3188 F2FS_I(inode)->cp_task == current ?
3189 FS_CP_DATA_IO : FS_DATA_IO);
3190}
3191
3e679dc7 3192static void f2fs_write_failed(struct inode *inode, loff_t to)
3aab8f82 3193{
819d9153 3194 loff_t i_size = i_size_read(inode);
3aab8f82 3195
3f188c23
JK		 3196 if (IS_NOQUOTA(inode))
3197 return;
3198
95ae251f
EB		 3199 /* In the fs-verity case, f2fs_end_enable_verity() does the truncate */
3200 if (to > i_size && !f2fs_verity_in_progress(inode)) {
a33c1502 3201 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
6f8d4455 3202 down_write(&F2FS_I(inode)->i_mmap_sem);
a33c1502 3203
819d9153 3204 truncate_pagecache(inode, i_size);
3f188c23 3205 f2fs_truncate_blocks(inode, i_size, true);
a33c1502 3206
5a3a2d83 3207 up_write(&F2FS_I(inode)->i_mmap_sem);
6f8d4455 3208 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3aab8f82
CY		 3209 }
3210}
3211
2aadac08
JK		 3212static int prepare_write_begin(struct f2fs_sb_info *sbi,
3213 struct page *page, loff_t pos, unsigned len,
3214 block_t *blk_addr, bool *node_changed)
3215{
3216 struct inode *inode = page->mapping->host;
3217 pgoff_t index = page->index;
3218 struct dnode_of_data dn;
3219 struct page *ipage;
b4d07a3e 3220 bool locked = false;
e15882b6 3221 struct extent_info ei = {0,0,0};
2aadac08 3222 int err = 0;
2866fb16 3223 int flag;
2aadac08 3224
24b84912
JK		 3225 /*
3226 * we already allocated all the blocks, so we don't need to get
3227 * the block addresses when there is no need to fill the page.
3228 */
dc91de78 3229 if (!f2fs_has_inline_data(inode) && len == PAGE_SIZE &&
95ae251f
EB		 3230 !is_inode_flag_set(inode, FI_NO_PREALLOC) &&
3231 !f2fs_verity_in_progress(inode))
24b84912
JK		 3232 return 0;
3233
2866fb16
SY		 3234 /* f2fs_lock_op avoids race between write CP and convert_inline_page */
3235 if (f2fs_has_inline_data(inode) && pos + len > MAX_INLINE_DATA(inode))
3236 flag = F2FS_GET_BLOCK_DEFAULT;
3237 else
3238 flag = F2FS_GET_BLOCK_PRE_AIO;
3239
b4d07a3e 3240 if (f2fs_has_inline_data(inode) ||
09cbfeaf 3241 (pos & PAGE_MASK) >= i_size_read(inode)) {
0ef81833 3242 f2fs_do_map_lock(sbi, flag, true);
b4d07a3e
JK		 3243 locked = true;
3244 }
4c8ff709 3245
b4d07a3e 3246restart:
2aadac08 3247 /* check inline_data */
4d57b86d 3248 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2aadac08
JK		 3249 if (IS_ERR(ipage)) {
3250 err = PTR_ERR(ipage);
3251 goto unlock_out;
3252 }
3253
3254 set_new_dnode(&dn, inode, ipage, ipage, 0);
3255
3256 if (f2fs_has_inline_data(inode)) {
f2470371 3257 if (pos + len <= MAX_INLINE_DATA(inode)) {
4d57b86d 3258 f2fs_do_read_inline_data(page, ipage);
91942321 3259 set_inode_flag(inode, FI_DATA_EXIST);
ab47036d 3260 if (inode->i_nlink)
b763f3be 3261 set_page_private_inline(ipage);
2aadac08
JK		 3262 } else {
3263 err = f2fs_convert_inline_page(&dn, page);
3264 if (err)
b4d07a3e
JK		 3265 goto out;
3266 if (dn.data_blkaddr == NULL_ADDR)
3267 err = f2fs_get_block(&dn, index);
3268 }
3269 } else if (locked) {
3270 err = f2fs_get_block(&dn, index);
3271 } else {
3272 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
3273 dn.data_blkaddr = ei.blk + index - ei.fofs;
3274 } else {
b4d07a3e 3275 /* hole case */
4d57b86d 3276 err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
4da7bf5a 3277 if (err || dn.data_blkaddr == NULL_ADDR) {
b4d07a3e 3278 f2fs_put_dnode(&dn);
0ef81833 3279 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO,
59c9081b 3280 true);
2866fb16 3281 WARN_ON(flag != F2FS_GET_BLOCK_PRE_AIO);
b4d07a3e
JK		 3282 locked = true;
3283 goto restart;
3284 }
2aadac08
JK		 3285 }
3286 }
b4d07a3e 3287
2aadac08
JK		 3288 /* convert_inline_page can make node_changed */
3289 *blk_addr = dn.data_blkaddr;
3290 *node_changed = dn.node_changed;
b4d07a3e 3291out:
2aadac08
JK		 3292 f2fs_put_dnode(&dn);
3293unlock_out:
b4d07a3e 3294 if (locked)
0ef81833 3295 f2fs_do_map_lock(sbi, flag, false);
2aadac08
JK		 3296 return err;
3297}
3298
eb47b800
JK		 3299static int f2fs_write_begin(struct file *file, struct address_space *mapping,
3300 loff_t pos, unsigned len, unsigned flags,
3301 struct page **pagep, void **fsdata)
3302{
3303 struct inode *inode = mapping->host;
4081363f 3304 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
86531d6b 3305 struct page *page = NULL;
09cbfeaf 3306 pgoff_t index = ((unsigned long long) pos) >> PAGE_SHIFT;
a2e2e76b 3307 bool need_balance = false, drop_atomic = false;
2aadac08 3308 block_t blkaddr = NULL_ADDR;
eb47b800
JK		 3309 int err = 0;
3310
62aed044
CY		 3311 trace_f2fs_write_begin(inode, pos, len, flags);
3312
00e09c0b
CY		 3313 if (!f2fs_is_checkpoint_ready(sbi)) {
3314 err = -ENOSPC;
4354994f 3315 goto fail;
00e09c0b 3316 }
4354994f 3317
455e3a58
JK		 3318 if ((f2fs_is_atomic_file(inode) &&
3319 !f2fs_available_free_memory(sbi, INMEM_PAGES)) ||
3320 is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
57864ae5 3321 err = -ENOMEM;
a2e2e76b 3322 drop_atomic = true;
57864ae5
JK		 3323 goto fail;
3324 }
3325
5f727395
JK		 3326 /*
3327 * We should check this at this moment to avoid deadlock on inode page
3328 * and #0 page. The locking rule for inline_data conversion should be:
3329 * lock_page(page #0) -> lock_page(inode_page)
3330 */
3331 if (index != 0) {
3332 err = f2fs_convert_inline_inode(inode);
3333 if (err)
3334 goto fail;
3335 }
4c8ff709
CY		 3336
3337#ifdef CONFIG_F2FS_FS_COMPRESSION
3338 if (f2fs_compressed_file(inode)) {
3339 int ret;
3340
3341 *fsdata = NULL;
3342
7eab7a69
FC		 3343 if (len == PAGE_SIZE)
3344 goto repeat;
3345
4c8ff709
CY		 3346 ret = f2fs_prepare_compress_overwrite(inode, pagep,
3347 index, fsdata);
3348 if (ret < 0) {
3349 err = ret;
3350 goto fail;
3351 } else if (ret) {
3352 return 0;
3353 }
3354 }
3355#endif
3356
afcb7ca0 3357repeat:
86d54795
JK		 3358 /*
3359 * Do not use grab_cache_page_write_begin() to avoid deadlock due to
3360 * wait_for_stable_page. Will wait that below with our IO control.
3361 */
01eccef7 3362 page = f2fs_pagecache_get_page(mapping, index,
86d54795 3363 FGP_LOCK | FGP_WRITE | FGP_CREAT, GFP_NOFS);
3aab8f82
CY		 3364 if (!page) {
3365 err = -ENOMEM;
3366 goto fail;
3367 }
d5f66990 3368
4c8ff709
CY		 3369 /* TODO: cluster can be compressed due to race with .writepage */
3370
eb47b800
JK		 3371 *pagep = page;
3372
2aadac08
JK		 3373 err = prepare_write_begin(sbi, page, pos, len,
3374 &blkaddr, &need_balance);
9ba69cf9 3375 if (err)
2aadac08 3376 goto fail;
9ba69cf9 3377
af033b2a
CY		 3378 if (need_balance && !IS_NOQUOTA(inode) &&
3379 has_not_enough_free_secs(sbi, 0, 0)) {
2a340760 3380 unlock_page(page);
2c4db1a6 3381 f2fs_balance_fs(sbi, true);
2a340760
JK		 3382 lock_page(page);
3383 if (page->mapping != mapping) {
3384 /* The page got truncated from under us */
3385 f2fs_put_page(page, 1);
3386 goto repeat;
3387 }
3388 }
3389
bae0ee7a 3390 f2fs_wait_on_page_writeback(page, DATA, false, true);
b3d208f9 3391
649d7df2
JK		 3392 if (len == PAGE_SIZE || PageUptodate(page))
3393 return 0;
eb47b800 3394
95ae251f
EB		 3395 if (!(pos & (PAGE_SIZE - 1)) && (pos + len) >= i_size_read(inode) &&
3396 !f2fs_verity_in_progress(inode)) {
746e2403
YH		 3397 zero_user_segment(page, len, PAGE_SIZE);
3398 return 0;
3399 }
3400
2aadac08 3401 if (blkaddr == NEW_ADDR) {
09cbfeaf 3402 zero_user_segment(page, 0, PAGE_SIZE);
649d7df2 3403 SetPageUptodate(page);
eb47b800 3404 } else {
93770ab7
CY		 3405 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3406 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 3407 err = -EFSCORRUPTED;
93770ab7
CY		 3408 goto fail;
3409 }
b7973091 3410 err = f2fs_submit_page_read(inode, page, blkaddr, 0, true);
13ba41e3 3411 if (err)
78682f79 3412 goto fail;
d54c795b 3413
393ff91f 3414 lock_page(page);
6bacf52f 3415 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 3416 f2fs_put_page(page, 1);
3417 goto repeat;
eb47b800 3418 }
1563ac75
CY		 3419 if (unlikely(!PageUptodate(page))) {
3420 err = -EIO;
3421 goto fail;
4375a336 3422 }
eb47b800 3423 }
eb47b800 3424 return 0;
9ba69cf9 3425
3aab8f82 3426fail:
86531d6b 3427 f2fs_put_page(page, 1);
3e679dc7 3428 f2fs_write_failed(inode, pos + len);
a2e2e76b 3429 if (drop_atomic)
4d57b86d 3430 f2fs_drop_inmem_pages_all(sbi, false);
3aab8f82 3431 return err;
eb47b800
JK		 3432}
3433
a1dd3c13
JK		 3434static int f2fs_write_end(struct file *file,
3435 struct address_space *mapping,
3436 loff_t pos, unsigned len, unsigned copied,
3437 struct page *page, void *fsdata)
3438{
3439 struct inode *inode = page->mapping->host;
3440
dfb2bf38
CY		 3441 trace_f2fs_write_end(inode, pos, len, copied);
3442
649d7df2
JK		 3443 /*
3444 * This should be come from len == PAGE_SIZE, and we expect copied
3445 * should be PAGE_SIZE. Otherwise, we treat it with zero copied and
3446 * let generic_perform_write() try to copy data again through copied=0.
3447 */
3448 if (!PageUptodate(page)) {
746e2403 3449 if (unlikely(copied != len))
649d7df2
JK		 3450 copied = 0;
3451 else
3452 SetPageUptodate(page);
3453 }
4c8ff709
CY		 3454
3455#ifdef CONFIG_F2FS_FS_COMPRESSION
3456 /* overwrite compressed file */
3457 if (f2fs_compressed_file(inode) && fsdata) {
3458 f2fs_compress_write_end(inode, fsdata, page->index, copied);
3459 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
944dd22e
CY		 3460
3461 if (pos + copied > i_size_read(inode) &&
3462 !f2fs_verity_in_progress(inode))
3463 f2fs_i_size_write(inode, pos + copied);
4c8ff709
CY		 3464 return copied;
3465 }
3466#endif
3467
649d7df2
JK		 3468 if (!copied)
3469 goto unlock_out;
3470
34ba94ba 3471 set_page_dirty(page);
a1dd3c13 3472
95ae251f
EB		 3473 if (pos + copied > i_size_read(inode) &&
3474 !f2fs_verity_in_progress(inode))
fc9581c8 3475 f2fs_i_size_write(inode, pos + copied);
649d7df2 3476unlock_out:
3024c9a1 3477 f2fs_put_page(page, 1);
d0239e1b 3478 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
a1dd3c13
JK		 3479 return copied;
3480}
3481
6f673763
OS		 3482static int check_direct_IO(struct inode *inode, struct iov_iter *iter,
3483 loff_t offset)
944fcfc1 3484{
8a56dd96
JK		 3485 unsigned i_blkbits = READ_ONCE(inode->i_blkbits);
3486 unsigned blkbits = i_blkbits;
3487 unsigned blocksize_mask = (1 << blkbits) - 1;
3488 unsigned long align = offset | iov_iter_alignment(iter);
3489 struct block_device *bdev = inode->i_sb->s_bdev;
3490
20d0a107
GKB		 3491 if (iov_iter_rw(iter) == READ && offset >= i_size_read(inode))
3492 return 1;
3493
8a56dd96
JK		 3494 if (align & blocksize_mask) {
3495 if (bdev)
3496 blkbits = blksize_bits(bdev_logical_block_size(bdev));
3497 blocksize_mask = (1 << blkbits) - 1;
3498 if (align & blocksize_mask)
3499 return -EINVAL;
3500 return 1;
3501 }
944fcfc1
JK		 3502 return 0;
3503}
3504
02b16d0a
CY		 3505static void f2fs_dio_end_io(struct bio *bio)
3506{
3507 struct f2fs_private_dio *dio = bio->bi_private;
3508
3509 dec_page_count(F2FS_I_SB(dio->inode),
3510 dio->write ? F2FS_DIO_WRITE : F2FS_DIO_READ);
3511
3512 bio->bi_private = dio->orig_private;
3513 bio->bi_end_io = dio->orig_end_io;
3514
c8eb7024 3515 kfree(dio);
02b16d0a
CY		 3516
3517 bio_endio(bio);
3518}
3519
3520static void f2fs_dio_submit_bio(struct bio *bio, struct inode *inode,
3521 loff_t file_offset)
3522{
3523 struct f2fs_private_dio *dio;
3524 bool write = (bio_op(bio) == REQ_OP_WRITE);
02b16d0a
CY		 3525
3526 dio = f2fs_kzalloc(F2FS_I_SB(inode),
3527 sizeof(struct f2fs_private_dio), GFP_NOFS);
8e114038 3528 if (!dio)
02b16d0a 3529 goto out;
02b16d0a
CY		 3530
3531 dio->inode = inode;
3532 dio->orig_end_io = bio->bi_end_io;
3533 dio->orig_private = bio->bi_private;
3534 dio->write = write;
3535
3536 bio->bi_end_io = f2fs_dio_end_io;
3537 bio->bi_private = dio;
3538
3539 inc_page_count(F2FS_I_SB(inode),
3540 write ? F2FS_DIO_WRITE : F2FS_DIO_READ);
3541
3542 submit_bio(bio);
3543 return;
3544out:
3545 bio->bi_status = BLK_STS_IOERR;
3546 bio_endio(bio);
3547}
3548
c8b8e32d 3549static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
eb47b800 3550{
b439b103 3551 struct address_space *mapping = iocb->ki_filp->f_mapping;
3aab8f82 3552 struct inode *inode = mapping->host;
0cdd3195 3553 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
f847c699 3554 struct f2fs_inode_info *fi = F2FS_I(inode);
3aab8f82 3555 size_t count = iov_iter_count(iter);
c8b8e32d 3556 loff_t offset = iocb->ki_pos;
82e0a5aa 3557 int rw = iov_iter_rw(iter);
3aab8f82 3558 int err;
0cdd3195 3559 enum rw_hint hint = iocb->ki_hint;
63189b78 3560 int whint_mode = F2FS_OPTION(sbi).whint_mode;
f847c699 3561 bool do_opu;
944fcfc1 3562
b439b103 3563 err = check_direct_IO(inode, iter, offset);
b9d777b8 3564 if (err)
8a56dd96 3565 return err < 0 ? err : 0;
9ffe0fb5 3566
f847c699 3567 if (f2fs_force_buffered_io(inode, iocb, iter))
36abef4e 3568 return 0;
fcc85a4d 3569
6de8687c 3570 do_opu = rw == WRITE && f2fs_lfs_mode(sbi);
f847c699 3571
5302fb00 3572 trace_f2fs_direct_IO_enter(inode, offset, count, rw);
70407fad 3573
0cdd3195
HL		 3574 if (rw == WRITE && whint_mode == WHINT_MODE_OFF)
3575 iocb->ki_hint = WRITE_LIFE_NOT_SET;
3576
f847c699
CY		 3577 if (iocb->ki_flags & IOCB_NOWAIT) {
3578 if (!down_read_trylock(&fi->i_gc_rwsem[rw])) {
3579 iocb->ki_hint = hint;
3580 err = -EAGAIN;
3581 goto out;
3582 }
3583 if (do_opu && !down_read_trylock(&fi->i_gc_rwsem[READ])) {
3584 up_read(&fi->i_gc_rwsem[rw]);
b91050a8
HL		 3585 iocb->ki_hint = hint;
3586 err = -EAGAIN;
3587 goto out;
3588 }
f847c699
CY		 3589 } else {
3590 down_read(&fi->i_gc_rwsem[rw]);
3591 if (do_opu)
3592 down_read(&fi->i_gc_rwsem[READ]);
b91050a8
HL		 3593 }
3594
02b16d0a 3595 err = __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev,
f9d6d059
CY		 3596 iter, rw == WRITE ? get_data_block_dio_write :
3597 get_data_block_dio, NULL, f2fs_dio_submit_bio,
ad8d6a02
D		 3598 rw == WRITE ? DIO_LOCKING | DIO_SKIP_HOLES :
3599 DIO_SKIP_HOLES);
f847c699
CY		 3600
3601 if (do_opu)
3602 up_read(&fi->i_gc_rwsem[READ]);
3603
3604 up_read(&fi->i_gc_rwsem[rw]);
82e0a5aa
CY		 3605
3606 if (rw == WRITE) {
0cdd3195
HL		 3607 if (whint_mode == WHINT_MODE_OFF)
3608 iocb->ki_hint = hint;
b0af6d49
CY		 3609 if (err > 0) {
3610 f2fs_update_iostat(F2FS_I_SB(inode), APP_DIRECT_IO,
3611 err);
f847c699
CY		 3612 if (!do_opu)
3613 set_inode_flag(inode, FI_UPDATE_WRITE);
335cac8b
JQ		 3614 } else if (err == -EIOCBQUEUED) {
3615 f2fs_update_iostat(F2FS_I_SB(inode), APP_DIRECT_IO,
3616 count - iov_iter_count(iter));
b0af6d49 3617 } else if (err < 0) {
3e679dc7 3618 f2fs_write_failed(inode, offset + count);
b0af6d49 3619 }
8b83ac81
CY		 3620 } else {
3621 if (err > 0)
3622 f2fs_update_iostat(sbi, APP_DIRECT_READ_IO, err);
335cac8b
JQ		 3623 else if (err == -EIOCBQUEUED)
3624 f2fs_update_iostat(F2FS_I_SB(inode), APP_DIRECT_READ_IO,
3625 count - iov_iter_count(iter));
6bfc4919 3626 }
70407fad 3627
b91050a8 3628out:
5302fb00 3629 trace_f2fs_direct_IO_exit(inode, offset, count, rw, err);
70407fad 3630
3aab8f82 3631 return err;
eb47b800
JK		 3632}
3633
487261f3
CY		 3634void f2fs_invalidate_page(struct page *page, unsigned int offset,
3635 unsigned int length)
eb47b800
JK		 3636{
3637 struct inode *inode = page->mapping->host;
487261f3 3638 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
a7ffdbe2 3639
487261f3 3640 if (inode->i_ino >= F2FS_ROOT_INO(sbi) &&
09cbfeaf 3641 (offset % PAGE_SIZE || length != PAGE_SIZE))
a7ffdbe2
JK		 3642 return;
3643
487261f3 3644 if (PageDirty(page)) {
933439c8 3645 if (inode->i_ino == F2FS_META_INO(sbi)) {
487261f3 3646 dec_page_count(sbi, F2FS_DIRTY_META);
933439c8 3647 } else if (inode->i_ino == F2FS_NODE_INO(sbi)) {
487261f3 3648 dec_page_count(sbi, F2FS_DIRTY_NODES);
933439c8 3649 } else {
487261f3 3650 inode_dec_dirty_pages(inode);
4d57b86d 3651 f2fs_remove_dirty_inode(inode);
933439c8 3652 }
487261f3 3653 }
decd36b6 3654
b763f3be 3655 clear_page_private_gcing(page);
2baf0781 3656
6ce19aff
CY		 3657 if (test_opt(sbi, COMPRESS_CACHE)) {
3658 if (f2fs_compressed_file(inode))
3659 f2fs_invalidate_compress_pages(sbi, inode->i_ino);
3660 if (inode->i_ino == F2FS_COMPRESS_INO(sbi))
3661 clear_page_private_data(page);
3662 }
3663
b763f3be 3664 if (page_private_atomic(page))
4d57b86d 3665 return f2fs_drop_inmem_page(inode, page);
decd36b6 3666
b763f3be
CY		 3667 detach_page_private(page);
3668 set_page_private(page, 0);
eb47b800
JK		 3669}
3670
487261f3 3671int f2fs_release_page(struct page *page, gfp_t wait)
eb47b800 3672{
f68daeeb
JK		 3673 /* If this is dirty page, keep PagePrivate */
3674 if (PageDirty(page))
3675 return 0;
3676
decd36b6 3677 /* This is atomic written page, keep Private */
b763f3be 3678 if (page_private_atomic(page))
decd36b6
CY		 3679 return 0;
3680
6ce19aff
CY		 3681 if (test_opt(F2FS_P_SB(page), COMPRESS_CACHE)) {
3682 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
3683 struct inode *inode = page->mapping->host;
3684
3685 if (f2fs_compressed_file(inode))
3686 f2fs_invalidate_compress_pages(sbi, inode->i_ino);
3687 if (inode->i_ino == F2FS_COMPRESS_INO(sbi))
3688 clear_page_private_data(page);
3689 }
3690
b763f3be
CY		 3691 clear_page_private_gcing(page);
3692
3693 detach_page_private(page);
3694 set_page_private(page, 0);
c3850aa1 3695 return 1;
eb47b800
JK		 3696}
3697
3698static int f2fs_set_data_page_dirty(struct page *page)
3699{
4969c06a 3700 struct inode *inode = page_file_mapping(page)->host;
eb47b800 3701
26c6b887
JK		 3702 trace_f2fs_set_page_dirty(page, DATA);
3703
237c0790
JK		 3704 if (!PageUptodate(page))
3705 SetPageUptodate(page);
4969c06a
JK		 3706 if (PageSwapCache(page))
3707 return __set_page_dirty_nobuffers(page);
34ba94ba 3708
5fe45743 3709 if (f2fs_is_atomic_file(inode) && !f2fs_is_commit_atomic_write(inode)) {
b763f3be 3710 if (!page_private_atomic(page)) {
4d57b86d 3711 f2fs_register_inmem_page(inode, page);
decd36b6
CY		 3712 return 1;
3713 }
3714 /*
3715 * Previously, this page has been registered, we just
3716 * return here.
3717 */
3718 return 0;
34ba94ba
JK		 3719 }
3720
eb47b800 3721 if (!PageDirty(page)) {
b87078ad 3722 __set_page_dirty_nobuffers(page);
4d57b86d 3723 f2fs_update_dirty_page(inode, page);
eb47b800
JK		 3724 return 1;
3725 }
3726 return 0;
3727}
3728
c1c63387
CY		 3729
3730static sector_t f2fs_bmap_compress(struct inode *inode, sector_t block)
3731{
3732#ifdef CONFIG_F2FS_FS_COMPRESSION
3733 struct dnode_of_data dn;
3734 sector_t start_idx, blknr = 0;
3735 int ret;
3736
3737 start_idx = round_down(block, F2FS_I(inode)->i_cluster_size);
3738
3739 set_new_dnode(&dn, inode, NULL, NULL, 0);
3740 ret = f2fs_get_dnode_of_data(&dn, start_idx, LOOKUP_NODE);
3741 if (ret)
3742 return 0;
3743
3744 if (dn.data_blkaddr != COMPRESS_ADDR) {
3745 dn.ofs_in_node += block - start_idx;
3746 blknr = f2fs_data_blkaddr(&dn);
3747 if (!__is_valid_data_blkaddr(blknr))
3748 blknr = 0;
3749 }
3750
3751 f2fs_put_dnode(&dn);
c1c63387
CY		 3752 return blknr;
3753#else
250e84d7 3754 return 0;
c1c63387
CY		 3755#endif
3756}
3757
3758
c01e54b7
JK		 3759static sector_t f2fs_bmap(struct address_space *mapping, sector_t block)
3760{
454ae7e5 3761 struct inode *inode = mapping->host;
b79b0a31 3762 sector_t blknr = 0;
454ae7e5 3763
1d373a0e 3764 if (f2fs_has_inline_data(inode))
b79b0a31 3765 goto out;
1d373a0e
JK		 3766
3767 /* make sure allocating whole blocks */
3768 if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
3769 filemap_write_and_wait(mapping);
3770
4eda1682 3771 /* Block number less than F2FS MAX BLOCKS */
6d1451bf 3772 if (unlikely(block >= max_file_blocks(inode)))
4eda1682 3773 goto out;
c1c63387 3774
4eda1682
DJ		 3775 if (f2fs_compressed_file(inode)) {
3776 blknr = f2fs_bmap_compress(inode, block);
3777 } else {
b876f4c9
JK		 3778 struct f2fs_map_blocks map;
3779
3780 memset(&map, 0, sizeof(map));
3781 map.m_lblk = block;
3782 map.m_len = 1;
3783 map.m_next_pgofs = NULL;
3784 map.m_seg_type = NO_CHECK_TYPE;
3785
3786 if (!f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_BMAP))
3787 blknr = map.m_pblk;
4eda1682 3788 }
b79b0a31
CY		 3789out:
3790 trace_f2fs_bmap(inode, block, blknr);
3791 return blknr;
429511cd
CY		 3792}
3793
5b7a487c
WG		 3794#ifdef CONFIG_MIGRATION
3795#include <linux/migrate.h>
3796
3797int f2fs_migrate_page(struct address_space *mapping,
3798 struct page *newpage, struct page *page, enum migrate_mode mode)
3799{
3800 int rc, extra_count;
3801 struct f2fs_inode_info *fi = F2FS_I(mapping->host);
b763f3be 3802 bool atomic_written = page_private_atomic(page);
5b7a487c
WG		 3803
3804 BUG_ON(PageWriteback(page));
3805
3806 /* migrating an atomic written page is safe with the inmem_lock hold */
ff1048e7
JK		 3807 if (atomic_written) {
3808 if (mode != MIGRATE_SYNC)
3809 return -EBUSY;
3810 if (!mutex_trylock(&fi->inmem_lock))
3811 return -EAGAIN;
3812 }
5b7a487c 3813
240a5915
CY		 3814 /* one extra reference was held for atomic_write page */
3815 extra_count = atomic_written ? 1 : 0;
5b7a487c 3816 rc = migrate_page_move_mapping(mapping, newpage,
37109694 3817 page, extra_count);
5b7a487c
WG		 3818 if (rc != MIGRATEPAGE_SUCCESS) {
3819 if (atomic_written)
3820 mutex_unlock(&fi->inmem_lock);
3821 return rc;
3822 }
3823
3824 if (atomic_written) {
3825 struct inmem_pages *cur;
5f029c04 3826
5b7a487c
WG		 3827 list_for_each_entry(cur, &fi->inmem_pages, list)
3828 if (cur->page == page) {
3829 cur->page = newpage;
3830 break;
3831 }
3832 mutex_unlock(&fi->inmem_lock);
3833 put_page(page);
3834 get_page(newpage);
3835 }
3836
c9ebd3df
JK		 3837 /* guarantee to start from no stale private field */
3838 set_page_private(newpage, 0);
240a5915 3839 if (PagePrivate(page)) {
b763f3be
CY		 3840 set_page_private(newpage, page_private(page));
3841 SetPagePrivate(newpage);
3842 get_page(newpage);
3843
3844 set_page_private(page, 0);
3845 ClearPagePrivate(page);
3846 put_page(page);
240a5915 3847 }
5b7a487c 3848
2916ecc0
JG		 3849 if (mode != MIGRATE_SYNC_NO_COPY)
3850 migrate_page_copy(newpage, page);
3851 else
3852 migrate_page_states(newpage, page);
5b7a487c
WG		 3853
3854 return MIGRATEPAGE_SUCCESS;
3855}
3856#endif
3857
4969c06a 3858#ifdef CONFIG_SWAP
859fca6b
CY		 3859static int f2fs_migrate_blocks(struct inode *inode, block_t start_blk,
3860 unsigned int blkcnt)
3861{
3862 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3863 unsigned int blkofs;
3864 unsigned int blk_per_sec = BLKS_PER_SEC(sbi);
3865 unsigned int secidx = start_blk / blk_per_sec;
3866 unsigned int end_sec = secidx + blkcnt / blk_per_sec;
3867 int ret = 0;
3868
3869 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3870 down_write(&F2FS_I(inode)->i_mmap_sem);
3871
3872 set_inode_flag(inode, FI_ALIGNED_WRITE);
3873
3874 for (; secidx < end_sec; secidx++) {
3875 down_write(&sbi->pin_sem);
3876
3877 f2fs_lock_op(sbi);
3878 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
3879 f2fs_unlock_op(sbi);
3880
3881 set_inode_flag(inode, FI_DO_DEFRAG);
3882
3883 for (blkofs = 0; blkofs < blk_per_sec; blkofs++) {
3884 struct page *page;
3885 unsigned int blkidx = secidx * blk_per_sec + blkofs;
3886
3887 page = f2fs_get_lock_data_page(inode, blkidx, true);
3888 if (IS_ERR(page)) {
3889 up_write(&sbi->pin_sem);
3890 ret = PTR_ERR(page);
3891 goto done;
3892 }
3893
3894 set_page_dirty(page);
3895 f2fs_put_page(page, 1);
3896 }
3897
3898 clear_inode_flag(inode, FI_DO_DEFRAG);
3899
3900 ret = filemap_fdatawrite(inode->i_mapping);
3901
3902 up_write(&sbi->pin_sem);
3903
3904 if (ret)
3905 break;
3906 }
3907
3908done:
3909 clear_inode_flag(inode, FI_DO_DEFRAG);
3910 clear_inode_flag(inode, FI_ALIGNED_WRITE);
3911
3912 up_write(&F2FS_I(inode)->i_mmap_sem);
3913 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3914
3915 return ret;
3916}
3917
0b8fc006 3918static int check_swap_activate(struct swap_info_struct *sis,
af4b6b8e
CY		 3919 struct file *swap_file, sector_t *span)
3920{
3921 struct address_space *mapping = swap_file->f_mapping;
3922 struct inode *inode = mapping->host;
36e4d958 3923 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
af4b6b8e
CY		 3924 sector_t cur_lblock;
3925 sector_t last_lblock;
3926 sector_t pblock;
3927 sector_t lowest_pblock = -1;
3928 sector_t highest_pblock = 0;
3929 int nr_extents = 0;
3930 unsigned long nr_pblocks;
859fca6b
CY		 3931 unsigned int blks_per_sec = BLKS_PER_SEC(sbi);
3932 unsigned int sec_blks_mask = BLKS_PER_SEC(sbi) - 1;
ca298241 3933 unsigned int not_aligned = 0;
36e4d958 3934 int ret = 0;
af4b6b8e
CY		 3935
3936 /*
3937 * Map all the blocks into the extent list. This code doesn't try
3938 * to be very smart.
3939 */
3940 cur_lblock = 0;
6cbfcab5 3941 last_lblock = bytes_to_blks(inode, i_size_read(inode));
af4b6b8e 3942
1da66103 3943 while (cur_lblock < last_lblock && cur_lblock < sis->max) {
b876f4c9 3944 struct f2fs_map_blocks map;
859fca6b 3945retry:
af4b6b8e
CY		 3946 cond_resched();
3947
b876f4c9
JK		 3948 memset(&map, 0, sizeof(map));
3949 map.m_lblk = cur_lblock;
36e4d958 3950 map.m_len = last_lblock - cur_lblock;
3951 map.m_next_pgofs = NULL;
3952 map.m_next_extent = NULL;
b876f4c9 3953 map.m_seg_type = NO_CHECK_TYPE;
36e4d958 3954 map.m_may_create = false;
af4b6b8e 3955
b876f4c9 3956 ret = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_FIEMAP);
af4b6b8e 3957 if (ret)
36e4d958 3958 goto out;
af4b6b8e
CY		 3959
3960 /* hole */
36e4d958 3961 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
833dcd35 3962 f2fs_err(sbi, "Swapfile has holes");
f395183f 3963 ret = -EINVAL;
36e4d958 3964 goto out;
3965 }
af4b6b8e 3966
b876f4c9
JK		 3967 pblock = map.m_pblk;
3968 nr_pblocks = map.m_len;
af4b6b8e 3969
859fca6b
CY		 3970 if ((pblock - SM_I(sbi)->main_blkaddr) & sec_blks_mask ||
3971 nr_pblocks & sec_blks_mask) {
ca298241 3972 not_aligned++;
36e4d958 3973
859fca6b
CY		 3974 nr_pblocks = roundup(nr_pblocks, blks_per_sec);
3975 if (cur_lblock + nr_pblocks > sis->max)
3976 nr_pblocks -= blks_per_sec;
3977
3978 if (!nr_pblocks) {
3979 /* this extent is last one */
3980 nr_pblocks = map.m_len;
3981 f2fs_warn(sbi, "Swapfile: last extent is not aligned to section");
3982 goto next;
3983 }
3984
3985 ret = f2fs_migrate_blocks(inode, cur_lblock,
3986 nr_pblocks);
3987 if (ret)
3988 goto out;
3989 goto retry;
3990 }
3991next:
af4b6b8e
CY		 3992 if (cur_lblock + nr_pblocks >= sis->max)
3993 nr_pblocks = sis->max - cur_lblock;
3994
3995 if (cur_lblock) { /* exclude the header page */
3996 if (pblock < lowest_pblock)
3997 lowest_pblock = pblock;
3998 if (pblock + nr_pblocks - 1 > highest_pblock)
3999 highest_pblock = pblock + nr_pblocks - 1;
4000 }
4001
4002 /*
4003 * We found a PAGE_SIZE-length, PAGE_SIZE-aligned run of blocks
4004 */
4005 ret = add_swap_extent(sis, cur_lblock, nr_pblocks, pblock);
4006 if (ret < 0)
4007 goto out;
4008 nr_extents += ret;
4009 cur_lblock += nr_pblocks;
4010 }
4011 ret = nr_extents;
4012 *span = 1 + highest_pblock - lowest_pblock;
4013 if (cur_lblock == 0)
4014 cur_lblock = 1; /* force Empty message */
4015 sis->max = cur_lblock;
4016 sis->pages = cur_lblock - 1;
4017 sis->highest_bit = cur_lblock - 1;
4018out:
859fca6b
CY		 4019 if (not_aligned)
4020 f2fs_warn(sbi, "Swapfile (%u) is not align to section: 1) creat(), 2) ioctl(F2FS_IOC_SET_PIN_FILE), 3) fallocate(%u * N)",
4021 not_aligned, blks_per_sec * F2FS_BLKSIZE);
af4b6b8e 4022 return ret;
af4b6b8e
CY		 4023}
4024
4969c06a
JK		 4025static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
4026 sector_t *span)
4027{
4028 struct inode *inode = file_inode(file);
4029 int ret;
4030
4031 if (!S_ISREG(inode->i_mode))
4032 return -EINVAL;
4033
4034 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
4035 return -EROFS;
4036
d927ccfc
SK		 4037 if (f2fs_lfs_mode(F2FS_I_SB(inode))) {
4038 f2fs_err(F2FS_I_SB(inode),
4039 "Swapfile not supported in LFS mode");
4040 return -EINVAL;
4041 }
4042
4969c06a
JK		 4043 ret = f2fs_convert_inline_inode(inode);
4044 if (ret)
4045 return ret;
4046
78134d03 4047 if (!f2fs_disable_compressed_file(inode))
4c8ff709
CY		 4048 return -EINVAL;
4049
0b979f1b
CY		 4050 f2fs_precache_extents(inode);
4051
3e5e479a
CY		 4052 ret = check_swap_activate(sis, file, span);
4053 if (ret < 0)
4969c06a
JK		 4054 return ret;
4055
4056 set_inode_flag(inode, FI_PIN_FILE);
4969c06a 4057 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3e5e479a 4058 return ret;
4969c06a
JK		 4059}
4060
4061static void f2fs_swap_deactivate(struct file *file)
4062{
4063 struct inode *inode = file_inode(file);
4064
4065 clear_inode_flag(inode, FI_PIN_FILE);
4066}
4067#else
4068static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
4069 sector_t *span)
4070{
4071 return -EOPNOTSUPP;
4072}
4073
4074static void f2fs_swap_deactivate(struct file *file)
4075{
4076}
4077#endif
4078
eb47b800
JK		 4079const struct address_space_operations f2fs_dblock_aops = {
4080 .readpage = f2fs_read_data_page,
23323196 4081 .readahead = f2fs_readahead,
eb47b800
JK		 4082 .writepage = f2fs_write_data_page,
4083 .writepages = f2fs_write_data_pages,
4084 .write_begin = f2fs_write_begin,
a1dd3c13 4085 .write_end = f2fs_write_end,
eb47b800 4086 .set_page_dirty = f2fs_set_data_page_dirty,
487261f3
CY		 4087 .invalidatepage = f2fs_invalidate_page,
4088 .releasepage = f2fs_release_page,
eb47b800 4089 .direct_IO = f2fs_direct_IO,
c01e54b7 4090 .bmap = f2fs_bmap,
4969c06a
JK		 4091 .swap_activate = f2fs_swap_activate,
4092 .swap_deactivate = f2fs_swap_deactivate,
5b7a487c
WG		 4093#ifdef CONFIG_MIGRATION
4094 .migratepage = f2fs_migrate_page,
4095#endif
eb47b800 4096};
6dbb1796 4097
5ec2d99d 4098void f2fs_clear_page_cache_dirty_tag(struct page *page)
aec2f729
CY		 4099{
4100 struct address_space *mapping = page_mapping(page);
4101 unsigned long flags;
4102
4103 xa_lock_irqsave(&mapping->i_pages, flags);
5ec2d99d 4104 __xa_clear_mark(&mapping->i_pages, page_index(page),
aec2f729
CY		 4105 PAGECACHE_TAG_DIRTY);
4106 xa_unlock_irqrestore(&mapping->i_pages, flags);
4107}
4108
6dbb1796
EB		 4109int __init f2fs_init_post_read_processing(void)
4110{
95ae251f
EB		 4111 bio_post_read_ctx_cache =
4112 kmem_cache_create("f2fs_bio_post_read_ctx",
4113 sizeof(struct bio_post_read_ctx), 0, 0, NULL);
6dbb1796
EB		 4114 if (!bio_post_read_ctx_cache)
4115 goto fail;
4116 bio_post_read_ctx_pool =
4117 mempool_create_slab_pool(NUM_PREALLOC_POST_READ_CTXS,
4118 bio_post_read_ctx_cache);
4119 if (!bio_post_read_ctx_pool)
4120 goto fail_free_cache;
4121 return 0;
4122
4123fail_free_cache:
4124 kmem_cache_destroy(bio_post_read_ctx_cache);
4125fail:
4126 return -ENOMEM;
4127}
4128
0b20fcec 4129void f2fs_destroy_post_read_processing(void)
6dbb1796
EB		 4130{
4131 mempool_destroy(bio_post_read_ctx_pool);
4132 kmem_cache_destroy(bio_post_read_ctx_cache);
4133}
0b20fcec 4134
4c8ff709
CY		 4135int f2fs_init_post_read_wq(struct f2fs_sb_info *sbi)
4136{
4137 if (!f2fs_sb_has_encrypt(sbi) &&
4138 !f2fs_sb_has_verity(sbi) &&
4139 !f2fs_sb_has_compression(sbi))
4140 return 0;
4141
4142 sbi->post_read_wq = alloc_workqueue("f2fs_post_read_wq",
4143 WQ_UNBOUND | WQ_HIGHPRI,
4144 num_online_cpus());
4145 if (!sbi->post_read_wq)
4146 return -ENOMEM;
4147 return 0;
4148}
4149
4150void f2fs_destroy_post_read_wq(struct f2fs_sb_info *sbi)
4151{
4152 if (sbi->post_read_wq)
4153 destroy_workqueue(sbi->post_read_wq);
4154}
4155
0b20fcec
CY		 4156int __init f2fs_init_bio_entry_cache(void)
4157{
98510003 4158 bio_entry_slab = f2fs_kmem_cache_create("f2fs_bio_entry_slab",
0b20fcec
CY		 4159 sizeof(struct bio_entry));
4160 if (!bio_entry_slab)
4161 return -ENOMEM;
4162 return 0;
4163}
4164
f543805f 4165void f2fs_destroy_bio_entry_cache(void)
0b20fcec
CY		 4166{
4167 kmem_cache_destroy(bio_entry_slab);
4168}
Linux 6.x block layer and io_uring tree(s)