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