]> git.ipfire.org Git - people/ms/linux.git/blame - fs/f2fs/data.c
projects / people / ms / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: reduce indentation in f2fs_file_write_iter()
[people/ms/linux.git] / fs / f2fs / data.c
CommitLineData
7c1a000d 1// SPDX-License-Identifier: GPL-2.0
0a8165d7 2/*
eb47b800
JK		 3 * fs/f2fs/data.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
eb47b800
JK		 7 */
8#include <linux/fs.h>
9#include <linux/f2fs_fs.h>
10#include <linux/buffer_head.h>
11#include <linux/mpage.h>
12#include <linux/writeback.h>
13#include <linux/backing-dev.h>
8f46dcae 14#include <linux/pagevec.h>
eb47b800
JK		 15#include <linux/blkdev.h>
16#include <linux/bio.h>
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"
52118743 28#include "iostat.h"
848753aa 29#include <trace/events/f2fs.h>
eb47b800 30
6dbb1796
EB		 31#define NUM_PREALLOC_POST_READ_CTXS 128
32
33static struct kmem_cache *bio_post_read_ctx_cache;
0b20fcec 34static struct kmem_cache *bio_entry_slab;
6dbb1796 35static mempool_t *bio_post_read_ctx_pool;
f543805f
CY		 36static struct bio_set f2fs_bioset;
37
38#define F2FS_BIO_POOL_SIZE NR_CURSEG_TYPE
39
40int __init f2fs_init_bioset(void)
41{
42 if (bioset_init(&f2fs_bioset, F2FS_BIO_POOL_SIZE,
43 0, BIOSET_NEED_BVECS))
44 return -ENOMEM;
45 return 0;
46}
47
48void f2fs_destroy_bioset(void)
49{
50 bioset_exit(&f2fs_bioset);
51}
52
36951b38
CY		 53static bool __is_cp_guaranteed(struct page *page)
54{
55 struct address_space *mapping = page->mapping;
56 struct inode *inode;
57 struct f2fs_sb_info *sbi;
58
59 if (!mapping)
60 return false;
61
62 inode = mapping->host;
63 sbi = F2FS_I_SB(inode);
64
65 if (inode->i_ino == F2FS_META_INO(sbi) ||
a87aff1d 66 inode->i_ino == F2FS_NODE_INO(sbi) ||
b763f3be
CY		 67 S_ISDIR(inode->i_mode))
68 return true;
69
70 if (f2fs_is_compressed_page(page))
71 return false;
72 if ((S_ISREG(inode->i_mode) &&
af033b2a 73 (f2fs_is_atomic_file(inode) || IS_NOQUOTA(inode))) ||
b763f3be 74 page_private_gcing(page))
36951b38
CY		 75 return true;
76 return false;
77}
78
5f9abab4
JK		 79static enum count_type __read_io_type(struct page *page)
80{
4969c06a 81 struct address_space *mapping = page_file_mapping(page);
5f9abab4
JK		 82
83 if (mapping) {
84 struct inode *inode = mapping->host;
85 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
86
87 if (inode->i_ino == F2FS_META_INO(sbi))
88 return F2FS_RD_META;
89
90 if (inode->i_ino == F2FS_NODE_INO(sbi))
91 return F2FS_RD_NODE;
92 }
93 return F2FS_RD_DATA;
94}
95
6dbb1796
EB		 96/* postprocessing steps for read bios */
97enum bio_post_read_step {
7f59b277
EB		 98#ifdef CONFIG_FS_ENCRYPTION
99 STEP_DECRYPT = 1 << 0,
100#else
101 STEP_DECRYPT = 0, /* compile out the decryption-related code */
102#endif
103#ifdef CONFIG_F2FS_FS_COMPRESSION
104 STEP_DECOMPRESS = 1 << 1,
105#else
106 STEP_DECOMPRESS = 0, /* compile out the decompression-related code */
107#endif
108#ifdef CONFIG_FS_VERITY
109 STEP_VERITY = 1 << 2,
110#else
111 STEP_VERITY = 0, /* compile out the verity-related code */
112#endif
6dbb1796
EB		 113};
114
115struct bio_post_read_ctx {
116 struct bio *bio;
4c8ff709 117 struct f2fs_sb_info *sbi;
6dbb1796 118 struct work_struct work;
6dbb1796 119 unsigned int enabled_steps;
4931e0c9 120 block_t fs_blkaddr;
6dbb1796
EB		 121};
122
7f59b277 123static void f2fs_finish_read_bio(struct bio *bio)
93dfe2ac 124{
6dbb1796 125 struct bio_vec *bv;
6dc4f100 126 struct bvec_iter_all iter_all;
93dfe2ac 127
7f59b277
EB		 128 /*
129 * Update and unlock the bio's pagecache pages, and put the
130 * decompression context for any compressed pages.
131 */
2b070cfe 132 bio_for_each_segment_all(bv, bio, iter_all) {
7f59b277 133 struct page *page = bv->bv_page;
6dbb1796 134
7f59b277
EB		 135 if (f2fs_is_compressed_page(page)) {
136 if (bio->bi_status)
6ce19aff 137 f2fs_end_read_compressed_page(page, true, 0);
7f59b277 138 f2fs_put_page_dic(page);
4c8ff709
CY		 139 continue;
140 }
4c8ff709 141
7f59b277 142 /* PG_error was set if decryption or verity failed. */
6dbb1796
EB		 143 if (bio->bi_status || PageError(page)) {
144 ClearPageUptodate(page);
fb7d70db
JK		 145 /* will re-read again later */
146 ClearPageError(page);
6dbb1796
EB		 147 } else {
148 SetPageUptodate(page);
149 }
5f9abab4 150 dec_page_count(F2FS_P_SB(page), __read_io_type(page));
6dbb1796
EB		 151 unlock_page(page);
152 }
79bbefb1 153
7f59b277
EB		 154 if (bio->bi_private)
155 mempool_free(bio->bi_private, bio_post_read_ctx_pool);
156 bio_put(bio);
4c8ff709 157}
4c8ff709 158
7f59b277 159static void f2fs_verify_bio(struct work_struct *work)
6dbb1796
EB		 160{
161 struct bio_post_read_ctx *ctx =
162 container_of(work, struct bio_post_read_ctx, work);
644c8c92 163 struct bio *bio = ctx->bio;
7f59b277 164 bool may_have_compressed_pages = (ctx->enabled_steps & STEP_DECOMPRESS);
644c8c92
EB		 165
166 /*
167 * fsverity_verify_bio() may call readpages() again, and while verity
7f59b277
EB		 168 * will be disabled for this, decryption and/or decompression may still
169 * be needed, resulting in another bio_post_read_ctx being allocated.
170 * So to prevent deadlocks we need to release the current ctx to the
171 * mempool first. This assumes that verity is the last post-read step.
644c8c92
EB		 172 */
173 mempool_free(ctx, bio_post_read_ctx_pool);
174 bio->bi_private = NULL;
6dbb1796 175
7f59b277
EB		 176 /*
177 * Verify the bio's pages with fs-verity. Exclude compressed pages,
178 * as those were handled separately by f2fs_end_read_compressed_page().
179 */
180 if (may_have_compressed_pages) {
181 struct bio_vec *bv;
182 struct bvec_iter_all iter_all;
183
184 bio_for_each_segment_all(bv, bio, iter_all) {
185 struct page *page = bv->bv_page;
186
187 if (!f2fs_is_compressed_page(page) &&
188 !PageError(page) && !fsverity_verify_page(page))
189 SetPageError(page);
190 }
191 } else {
192 fsverity_verify_bio(bio);
4c8ff709 193 }
6dbb1796 194
7f59b277 195 f2fs_finish_read_bio(bio);
6dbb1796
EB		 196}
197
7f59b277
EB		 198/*
199 * If the bio's data needs to be verified with fs-verity, then enqueue the
200 * verity work for the bio. Otherwise finish the bio now.
201 *
202 * Note that to avoid deadlocks, the verity work can't be done on the
203 * decryption/decompression workqueue. This is because verifying the data pages
204 * can involve reading verity metadata pages from the file, and these verity
205 * metadata pages may be encrypted and/or compressed.
206 */
207static void f2fs_verify_and_finish_bio(struct bio *bio)
95ae251f 208{
7f59b277 209 struct bio_post_read_ctx *ctx = bio->bi_private;
4c8ff709 210
7f59b277
EB		 211 if (ctx && (ctx->enabled_steps & STEP_VERITY)) {
212 INIT_WORK(&ctx->work, f2fs_verify_bio);
4c8ff709 213 fsverity_enqueue_verify_work(&ctx->work);
7f59b277
EB		 214 } else {
215 f2fs_finish_read_bio(bio);
4c8ff709 216 }
4c8ff709 217}
95ae251f 218
7f59b277
EB		 219/*
220 * Handle STEP_DECOMPRESS by decompressing any compressed clusters whose last
221 * remaining page was read by @ctx->bio.
222 *
223 * Note that a bio may span clusters (even a mix of compressed and uncompressed
224 * clusters) or be for just part of a cluster. STEP_DECOMPRESS just indicates
225 * that the bio includes at least one compressed page. The actual decompression
226 * is done on a per-cluster basis, not a per-bio basis.
227 */
228static void f2fs_handle_step_decompress(struct bio_post_read_ctx *ctx)
6dbb1796 229{
7f59b277
EB		 230 struct bio_vec *bv;
231 struct bvec_iter_all iter_all;
232 bool all_compressed = true;
4931e0c9 233 block_t blkaddr = ctx->fs_blkaddr;
4c8ff709 234
7f59b277
EB		 235 bio_for_each_segment_all(bv, ctx->bio, iter_all) {
236 struct page *page = bv->bv_page;
4c8ff709 237
7f59b277
EB		 238 /* PG_error was set if decryption failed. */
239 if (f2fs_is_compressed_page(page))
6ce19aff
CY		 240 f2fs_end_read_compressed_page(page, PageError(page),
241 blkaddr);
7f59b277
EB		 242 else
243 all_compressed = false;
6ce19aff
CY		 244
245 blkaddr++;
6dbb1796 246 }
4c8ff709 247
7f59b277
EB		 248 /*
249 * Optimization: if all the bio's pages are compressed, then scheduling
250 * the per-bio verity work is unnecessary, as verity will be fully
251 * handled at the compression cluster level.
252 */
253 if (all_compressed)
254 ctx->enabled_steps &= ~STEP_VERITY;
6dbb1796
EB		 255}
256
7f59b277 257static void f2fs_post_read_work(struct work_struct *work)
6dbb1796 258{
7f59b277
EB		 259 struct bio_post_read_ctx *ctx =
260 container_of(work, struct bio_post_read_ctx, work);
261
262 if (ctx->enabled_steps & STEP_DECRYPT)
263 fscrypt_decrypt_bio(ctx->bio);
264
265 if (ctx->enabled_steps & STEP_DECOMPRESS)
266 f2fs_handle_step_decompress(ctx);
267
268 f2fs_verify_and_finish_bio(ctx->bio);
6dbb1796
EB		 269}
270
271static void f2fs_read_end_io(struct bio *bio)
272{
c45d6002 273 struct f2fs_sb_info *sbi = F2FS_P_SB(bio_first_page_all(bio));
a4b68176
DJ		 274 struct bio_post_read_ctx *ctx;
275
276 iostat_update_and_unbind_ctx(bio, 0);
277 ctx = bio->bi_private;
c45d6002
CY		 278
279 if (time_to_inject(sbi, FAULT_READ_IO)) {
280 f2fs_show_injection_info(sbi, FAULT_READ_IO);
4e4cbee9 281 bio->bi_status = BLK_STS_IOERR;
55523519 282 }
8b038c70 283
7f59b277
EB		 284 if (bio->bi_status) {
285 f2fs_finish_read_bio(bio);
6dbb1796 286 return;
f1e88660 287 }
6dbb1796 288
7f59b277
EB		 289 if (ctx && (ctx->enabled_steps & (STEP_DECRYPT | STEP_DECOMPRESS))) {
290 INIT_WORK(&ctx->work, f2fs_post_read_work);
291 queue_work(ctx->sbi->post_read_wq, &ctx->work);
292 } else {
293 f2fs_verify_and_finish_bio(bio);
294 }
f1e88660
JK		 295}
296
4246a0b6 297static void f2fs_write_end_io(struct bio *bio)
93dfe2ac 298{
a4b68176 299 struct f2fs_sb_info *sbi;
f568849e 300 struct bio_vec *bvec;
6dc4f100 301 struct bvec_iter_all iter_all;
93dfe2ac 302
a4b68176
DJ		 303 iostat_update_and_unbind_ctx(bio, 1);
304 sbi = bio->bi_private;
305
6f5c2ed0 306 if (time_to_inject(sbi, FAULT_WRITE_IO)) {
c45d6002 307 f2fs_show_injection_info(sbi, FAULT_WRITE_IO);
6f5c2ed0
CY		 308 bio->bi_status = BLK_STS_IOERR;
309 }
310
2b070cfe 311 bio_for_each_segment_all(bvec, bio, iter_all) {
93dfe2ac 312 struct page *page = bvec->bv_page;
36951b38 313 enum count_type type = WB_DATA_TYPE(page);
93dfe2ac 314
b763f3be
CY		 315 if (page_private_dummy(page)) {
316 clear_page_private_dummy(page);
0a595eba
JK		 317 unlock_page(page);
318 mempool_free(page, sbi->write_io_dummy);
319
4e4cbee9 320 if (unlikely(bio->bi_status))
0a595eba
JK		 321 f2fs_stop_checkpoint(sbi, true);
322 continue;
323 }
324
d2d0727b 325 fscrypt_finalize_bounce_page(&page);
4375a336 326
4c8ff709
CY		 327#ifdef CONFIG_F2FS_FS_COMPRESSION
328 if (f2fs_is_compressed_page(page)) {
329 f2fs_compress_write_end_io(bio, page);
330 continue;
331 }
332#endif
333
4e4cbee9 334 if (unlikely(bio->bi_status)) {
5114a97a 335 mapping_set_error(page->mapping, -EIO);
b1ca321d
JK		 336 if (type == F2FS_WB_CP_DATA)
337 f2fs_stop_checkpoint(sbi, true);
93dfe2ac 338 }
7dff55d2
YH		 339
340 f2fs_bug_on(sbi, page->mapping == NODE_MAPPING(sbi) &&
341 page->index != nid_of_node(page));
342
36951b38 343 dec_page_count(sbi, type);
50fa53ec
CY		 344 if (f2fs_in_warm_node_list(sbi, page))
345 f2fs_del_fsync_node_entry(sbi, page);
b763f3be 346 clear_page_private_gcing(page);
93dfe2ac 347 end_page_writeback(page);
f568849e 348 }
36951b38 349 if (!get_pages(sbi, F2FS_WB_CP_DATA) &&
f5730184 350 wq_has_sleeper(&sbi->cp_wait))
93dfe2ac
JK		 351 wake_up(&sbi->cp_wait);
352
353 bio_put(bio);
354}
355
3c62be17
JK		 356struct block_device *f2fs_target_device(struct f2fs_sb_info *sbi,
357 block_t blk_addr, struct bio *bio)
358{
359 struct block_device *bdev = sbi->sb->s_bdev;
360 int i;
361
0916878d
DLM		 362 if (f2fs_is_multi_device(sbi)) {
363 for (i = 0; i < sbi->s_ndevs; i++) {
364 if (FDEV(i).start_blk <= blk_addr &&
365 FDEV(i).end_blk >= blk_addr) {
366 blk_addr -= FDEV(i).start_blk;
367 bdev = FDEV(i).bdev;
368 break;
369 }
3c62be17
JK		 370 }
371 }
372 if (bio) {
74d46992 373 bio_set_dev(bio, bdev);
3c62be17
JK		 374 bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr);
375 }
376 return bdev;
377}
378
379int f2fs_target_device_index(struct f2fs_sb_info *sbi, block_t blkaddr)
380{
381 int i;
382
0916878d
DLM		 383 if (!f2fs_is_multi_device(sbi))
384 return 0;
385
3c62be17
JK		 386 for (i = 0; i < sbi->s_ndevs; i++)
387 if (FDEV(i).start_blk <= blkaddr && FDEV(i).end_blk >= blkaddr)
388 return i;
389 return 0;
390}
391
b757f6ed 392static struct bio *__bio_alloc(struct f2fs_io_info *fio, int npages)
940a6d34 393{
b757f6ed 394 struct f2fs_sb_info *sbi = fio->sbi;
940a6d34
GZ		 395 struct bio *bio;
396
67883ade 397 bio = bio_alloc_bioset(GFP_NOIO, npages, &f2fs_bioset);
940a6d34 398
b757f6ed
CY		 399 f2fs_target_device(sbi, fio->new_blkaddr, bio);
400 if (is_read_io(fio->op)) {
0cdd3195
HL		 401 bio->bi_end_io = f2fs_read_end_io;
402 bio->bi_private = NULL;
403 } else {
404 bio->bi_end_io = f2fs_write_end_io;
405 bio->bi_private = sbi;
b757f6ed
CY		 406 bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi,
407 fio->type, fio->temp);
0cdd3195 408 }
a4b68176
DJ		 409 iostat_alloc_and_bind_ctx(sbi, bio, NULL);
410
b757f6ed
CY		 411 if (fio->io_wbc)
412 wbc_init_bio(fio->io_wbc, bio);
940a6d34
GZ		 413
414 return bio;
415}
416
27aacd28
ST		 417static void f2fs_set_bio_crypt_ctx(struct bio *bio, const struct inode *inode,
418 pgoff_t first_idx,
419 const struct f2fs_io_info *fio,
420 gfp_t gfp_mask)
421{
422 /*
423 * The f2fs garbage collector sets ->encrypted_page when it wants to
424 * read/write raw data without encryption.
425 */
426 if (!fio || !fio->encrypted_page)
427 fscrypt_set_bio_crypt_ctx(bio, inode, first_idx, gfp_mask);
428}
429
430static bool f2fs_crypt_mergeable_bio(struct bio *bio, const struct inode *inode,
431 pgoff_t next_idx,
432 const struct f2fs_io_info *fio)
433{
434 /*
435 * The f2fs garbage collector sets ->encrypted_page when it wants to
436 * read/write raw data without encryption.
437 */
438 if (fio && fio->encrypted_page)
439 return !bio_has_crypt_ctx(bio);
440
441 return fscrypt_mergeable_bio(bio, inode, next_idx);
442}
443
4fc29c1a
LT		 444static inline void __submit_bio(struct f2fs_sb_info *sbi,
445 struct bio *bio, enum page_type type)
f5730184 446{
4fc29c1a 447 if (!is_read_io(bio_op(bio))) {
0a595eba
JK		 448 unsigned int start;
449
0a595eba
JK		 450 if (type != DATA && type != NODE)
451 goto submit_io;
452
b0332a0f 453 if (f2fs_lfs_mode(sbi) && current->plug)
3bb09a0e
TY		 454 blk_finish_plug(current->plug);
455
39f71b7e 456 if (!F2FS_IO_ALIGNED(sbi))
8223ecc4
CY		 457 goto submit_io;
458
0a595eba
JK		 459 start = bio->bi_iter.bi_size >> F2FS_BLKSIZE_BITS;
460 start %= F2FS_IO_SIZE(sbi);
461
462 if (start == 0)
463 goto submit_io;
464
465 /* fill dummy pages */
466 for (; start < F2FS_IO_SIZE(sbi); start++) {
467 struct page *page =
468 mempool_alloc(sbi->write_io_dummy,
bc73a4b2 469 GFP_NOIO | __GFP_NOFAIL);
0a595eba
JK		 470 f2fs_bug_on(sbi, !page);
471
0a595eba 472 lock_page(page);
b763f3be
CY		 473
474 zero_user_segment(page, 0, PAGE_SIZE);
475 set_page_private_dummy(page);
476
0a595eba
JK		 477 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE)
478 f2fs_bug_on(sbi, 1);
479 }
480 /*
481 * In the NODE case, we lose next block address chain. So, we
482 * need to do checkpoint in f2fs_sync_file.
483 */
484 if (type == NODE)
485 set_sbi_flag(sbi, SBI_NEED_CP);
19a5f5e2 486 }
0a595eba 487submit_io:
554b5125
JK		 488 if (is_read_io(bio_op(bio)))
489 trace_f2fs_submit_read_bio(sbi->sb, type, bio);
490 else
491 trace_f2fs_submit_write_bio(sbi->sb, type, bio);
a4b68176
DJ		 492
493 iostat_update_submit_ctx(bio, type);
4e49ea4a 494 submit_bio(bio);
f5730184
JK		 495}
496
4c8ff709
CY		 497void f2fs_submit_bio(struct f2fs_sb_info *sbi,
498 struct bio *bio, enum page_type type)
499{
500 __submit_bio(sbi, bio, type);
501}
502
32b6aba8 503static void __attach_io_flag(struct f2fs_io_info *fio)
da9953b7
JK		 504{
505 struct f2fs_sb_info *sbi = fio->sbi;
506 unsigned int temp_mask = (1 << NR_TEMP_TYPE) - 1;
32b6aba8
JK		 507 unsigned int io_flag, fua_flag, meta_flag;
508
509 if (fio->type == DATA)
510 io_flag = sbi->data_io_flag;
511 else if (fio->type == NODE)
512 io_flag = sbi->node_io_flag;
513 else
514 return;
515
516 fua_flag = io_flag & temp_mask;
517 meta_flag = (io_flag >> NR_TEMP_TYPE) & temp_mask;
518
da9953b7 519 /*
32b6aba8 520 * data/node io flag bits per temp:
da9953b7
JK		 521 * REQ_META | REQ_FUA |
522 * 5 | 4 | 3 | 2 | 1 | 0 |
523 * Cold | Warm | Hot | Cold | Warm | Hot |
524 */
da9953b7
JK		 525 if ((1 << fio->temp) & meta_flag)
526 fio->op_flags |= REQ_META;
527 if ((1 << fio->temp) & fua_flag)
528 fio->op_flags |= REQ_FUA;
529}
530
458e6197 531static void __submit_merged_bio(struct f2fs_bio_info *io)
93dfe2ac 532{
458e6197 533 struct f2fs_io_info *fio = &io->fio;
93dfe2ac
JK		 534
535 if (!io->bio)
536 return;
537
32b6aba8 538 __attach_io_flag(fio);
554b5125
JK		 539 bio_set_op_attrs(io->bio, fio->op, fio->op_flags);
540
04d328de 541 if (is_read_io(fio->op))
554b5125 542 trace_f2fs_prepare_read_bio(io->sbi->sb, fio->type, io->bio);
6a8f8ca5 543 else
554b5125 544 trace_f2fs_prepare_write_bio(io->sbi->sb, fio->type, io->bio);
04d328de 545
4fc29c1a 546 __submit_bio(io->sbi, io->bio, fio->type);
93dfe2ac
JK		 547 io->bio = NULL;
548}
549
8648de2c 550static bool __has_merged_page(struct bio *bio, struct inode *inode,
bab475c5 551 struct page *page, nid_t ino)
0fd785eb 552{
0fd785eb 553 struct bio_vec *bvec;
6dc4f100 554 struct bvec_iter_all iter_all;
0fd785eb 555
8648de2c 556 if (!bio)
0fd785eb 557 return false;
0c3a5797 558
bab475c5 559 if (!inode && !page && !ino)
0c3a5797 560 return true;
0fd785eb 561
8648de2c 562 bio_for_each_segment_all(bvec, bio, iter_all) {
4c8ff709 563 struct page *target = bvec->bv_page;
0fd785eb 564
4c8ff709 565 if (fscrypt_is_bounce_page(target)) {
d2d0727b 566 target = fscrypt_pagecache_page(target);
4c8ff709
CY		 567 if (IS_ERR(target))
568 continue;
569 }
570 if (f2fs_is_compressed_page(target)) {
571 target = f2fs_compress_control_page(target);
572 if (IS_ERR(target))
573 continue;
574 }
0fd785eb 575
0c3a5797
CY		 576 if (inode && inode == target->mapping->host)
577 return true;
bab475c5
CY		 578 if (page && page == target)
579 return true;
0c3a5797 580 if (ino && ino == ino_of_node(target))
0fd785eb 581 return true;
0fd785eb
CY		 582 }
583
0fd785eb
CY		 584 return false;
585}
586
b9109b0e 587static void __f2fs_submit_merged_write(struct f2fs_sb_info *sbi,
a912b54d 588 enum page_type type, enum temp_type temp)
93dfe2ac
JK		 589{
590 enum page_type btype = PAGE_TYPE_OF_BIO(type);
a912b54d 591 struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
93dfe2ac 592
df0f8dc0 593 down_write(&io->io_rwsem);
458e6197
JK		 594
595 /* change META to META_FLUSH in the checkpoint procedure */
596 if (type >= META_FLUSH) {
597 io->fio.type = META_FLUSH;
04d328de 598 io->fio.op = REQ_OP_WRITE;
3adc5fcb 599 io->fio.op_flags = REQ_META | REQ_PRIO | REQ_SYNC;
70fd7614 600 if (!test_opt(sbi, NOBARRIER))
7f54f51f 601 io->fio.op_flags |= REQ_PREFLUSH | REQ_FUA;
458e6197
JK		 602 }
603 __submit_merged_bio(io);
df0f8dc0 604 up_write(&io->io_rwsem);
93dfe2ac
JK		 605}
606
a912b54d 607static void __submit_merged_write_cond(struct f2fs_sb_info *sbi,
bab475c5
CY		 608 struct inode *inode, struct page *page,
609 nid_t ino, enum page_type type, bool force)
0c3a5797 610{
a912b54d 611 enum temp_type temp;
1e771e83 612 bool ret = true;
a912b54d
JK		 613
614 for (temp = HOT; temp < NR_TEMP_TYPE; temp++) {
1e771e83
YS		 615 if (!force) {
616 enum page_type btype = PAGE_TYPE_OF_BIO(type);
617 struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
a912b54d 618
1e771e83 619 down_read(&io->io_rwsem);
8648de2c 620 ret = __has_merged_page(io->bio, inode, page, ino);
1e771e83
YS		 621 up_read(&io->io_rwsem);
622 }
623 if (ret)
624 __f2fs_submit_merged_write(sbi, type, temp);
a912b54d
JK		 625
626 /* TODO: use HOT temp only for meta pages now. */
627 if (type >= META)
628 break;
629 }
0c3a5797
CY		 630}
631
b9109b0e 632void f2fs_submit_merged_write(struct f2fs_sb_info *sbi, enum page_type type)
0c3a5797 633{
adcc00f7 634 __submit_merged_write_cond(sbi, NULL, NULL, 0, type, true);
0c3a5797
CY		 635}
636
b9109b0e 637void f2fs_submit_merged_write_cond(struct f2fs_sb_info *sbi,
bab475c5
CY		 638 struct inode *inode, struct page *page,
639 nid_t ino, enum page_type type)
0c3a5797 640{
bab475c5 641 __submit_merged_write_cond(sbi, inode, page, ino, type, false);
0c3a5797
CY		 642}
643
b9109b0e 644void f2fs_flush_merged_writes(struct f2fs_sb_info *sbi)
406657dd 645{
b9109b0e
JK		 646 f2fs_submit_merged_write(sbi, DATA);
647 f2fs_submit_merged_write(sbi, NODE);
648 f2fs_submit_merged_write(sbi, META);
406657dd
CY		 649}
650
93dfe2ac
JK		 651/*
652 * Fill the locked page with data located in the block address.
771a9a71 653 * A caller needs to unlock the page on failure.
93dfe2ac 654 */
05ca3632 655int f2fs_submit_page_bio(struct f2fs_io_info *fio)
93dfe2ac 656{
93dfe2ac 657 struct bio *bio;
0b81d077
JK		 658 struct page *page = fio->encrypted_page ?
659 fio->encrypted_page : fio->page;
93dfe2ac 660
c9b60788 661 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
93770ab7
CY		 662 fio->is_por ? META_POR : (__is_meta_io(fio) ?
663 META_GENERIC : DATA_GENERIC_ENHANCE)))
10f966bb 664 return -EFSCORRUPTED;
c9b60788 665
2ace38e0 666 trace_f2fs_submit_page_bio(page, fio);
93dfe2ac
JK		 667
668 /* Allocate a new bio */
b757f6ed 669 bio = __bio_alloc(fio, 1);
93dfe2ac 670
27aacd28
ST		 671 f2fs_set_bio_crypt_ctx(bio, fio->page->mapping->host,
672 fio->page->index, fio, GFP_NOIO);
673
09cbfeaf 674 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
93dfe2ac 675 bio_put(bio);
93dfe2ac
JK		 676 return -EFAULT;
677 }
78efac53
CY		 678
679 if (fio->io_wbc && !is_read_io(fio->op))
34e51a5e 680 wbc_account_cgroup_owner(fio->io_wbc, page, PAGE_SIZE);
78efac53 681
b7b911d5 682 __attach_io_flag(fio);
04d328de 683 bio_set_op_attrs(bio, fio->op, fio->op_flags);
93dfe2ac 684
5f9abab4
JK		 685 inc_page_count(fio->sbi, is_read_io(fio->op) ?
686 __read_io_type(page): WB_DATA_TYPE(fio->page));
4c58ed07
CY		 687
688 __submit_bio(fio->sbi, bio, fio->type);
93dfe2ac
JK		 689 return 0;
690}
691
8896cbdf
CY		 692static bool page_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
693 block_t last_blkaddr, block_t cur_blkaddr)
694{
10208567
JK		 695 if (unlikely(sbi->max_io_bytes &&
696 bio->bi_iter.bi_size >= sbi->max_io_bytes))
697 return false;
8896cbdf
CY		 698 if (last_blkaddr + 1 != cur_blkaddr)
699 return false;
309dca30 700 return bio->bi_bdev == f2fs_target_device(sbi, cur_blkaddr, NULL);
8896cbdf
CY		 701}
702
703static bool io_type_is_mergeable(struct f2fs_bio_info *io,
704 struct f2fs_io_info *fio)
705{
706 if (io->fio.op != fio->op)
707 return false;
708 return io->fio.op_flags == fio->op_flags;
709}
710
711static bool io_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
712 struct f2fs_bio_info *io,
713 struct f2fs_io_info *fio,
714 block_t last_blkaddr,
715 block_t cur_blkaddr)
716{
c72db71e
CY		 717 if (F2FS_IO_ALIGNED(sbi) && (fio->type == DATA || fio->type == NODE)) {
718 unsigned int filled_blocks =
719 F2FS_BYTES_TO_BLK(bio->bi_iter.bi_size);
720 unsigned int io_size = F2FS_IO_SIZE(sbi);
721 unsigned int left_vecs = bio->bi_max_vecs - bio->bi_vcnt;
722
723 /* IOs in bio is aligned and left space of vectors is not enough */
724 if (!(filled_blocks % io_size) && left_vecs < io_size)
725 return false;
726 }
8896cbdf
CY		 727 if (!page_is_mergeable(sbi, bio, last_blkaddr, cur_blkaddr))
728 return false;
729 return io_type_is_mergeable(io, fio);
730}
731
0b20fcec
CY		 732static void add_bio_entry(struct f2fs_sb_info *sbi, struct bio *bio,
733 struct page *page, enum temp_type temp)
734{
735 struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
736 struct bio_entry *be;
737
32410577 738 be = f2fs_kmem_cache_alloc(bio_entry_slab, GFP_NOFS, true, NULL);
0b20fcec
CY		 739 be->bio = bio;
740 bio_get(bio);
741
742 if (bio_add_page(bio, page, PAGE_SIZE, 0) != PAGE_SIZE)
743 f2fs_bug_on(sbi, 1);
744
745 down_write(&io->bio_list_lock);
746 list_add_tail(&be->list, &io->bio_list);
747 up_write(&io->bio_list_lock);
748}
749
750static void del_bio_entry(struct bio_entry *be)
751{
752 list_del(&be->list);
753 kmem_cache_free(bio_entry_slab, be);
754}
755
27aacd28 756static int add_ipu_page(struct f2fs_io_info *fio, struct bio **bio,
0b20fcec
CY		 757 struct page *page)
758{
27aacd28 759 struct f2fs_sb_info *sbi = fio->sbi;
0b20fcec
CY		 760 enum temp_type temp;
761 bool found = false;
762 int ret = -EAGAIN;
763
764 for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
765 struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
766 struct list_head *head = &io->bio_list;
767 struct bio_entry *be;
768
769 down_write(&io->bio_list_lock);
770 list_for_each_entry(be, head, list) {
771 if (be->bio != *bio)
772 continue;
773
774 found = true;
775
27aacd28
ST		 776 f2fs_bug_on(sbi, !page_is_mergeable(sbi, *bio,
777 *fio->last_block,
778 fio->new_blkaddr));
779 if (f2fs_crypt_mergeable_bio(*bio,
780 fio->page->mapping->host,
781 fio->page->index, fio) &&
782 bio_add_page(*bio, page, PAGE_SIZE, 0) ==
783 PAGE_SIZE) {
0b20fcec
CY		 784 ret = 0;
785 break;
786 }
787
27aacd28 788 /* page can't be merged into bio; submit the bio */
0b20fcec
CY		 789 del_bio_entry(be);
790 __submit_bio(sbi, *bio, DATA);
791 break;
792 }
793 up_write(&io->bio_list_lock);
794 }
795
796 if (ret) {
797 bio_put(*bio);
798 *bio = NULL;
799 }
800
801 return ret;
802}
803
804void f2fs_submit_merged_ipu_write(struct f2fs_sb_info *sbi,
805 struct bio **bio, struct page *page)
806{
807 enum temp_type temp;
808 bool found = false;
809 struct bio *target = bio ? *bio : NULL;
810
811 for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
812 struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
813 struct list_head *head = &io->bio_list;
814 struct bio_entry *be;
815
816 if (list_empty(head))
817 continue;
818
819 down_read(&io->bio_list_lock);
820 list_for_each_entry(be, head, list) {
821 if (target)
822 found = (target == be->bio);
823 else
824 found = __has_merged_page(be->bio, NULL,
825 page, 0);
826 if (found)
827 break;
828 }
829 up_read(&io->bio_list_lock);
830
831 if (!found)
832 continue;
833
834 found = false;
835
836 down_write(&io->bio_list_lock);
837 list_for_each_entry(be, head, list) {
838 if (target)
839 found = (target == be->bio);
840 else
841 found = __has_merged_page(be->bio, NULL,
842 page, 0);
843 if (found) {
844 target = be->bio;
845 del_bio_entry(be);
846 break;
847 }
848 }
849 up_write(&io->bio_list_lock);
850 }
851
852 if (found)
853 __submit_bio(sbi, target, DATA);
854 if (bio && *bio) {
855 bio_put(*bio);
856 *bio = NULL;
857 }
858}
859
8648de2c
CY		 860int f2fs_merge_page_bio(struct f2fs_io_info *fio)
861{
862 struct bio *bio = *fio->bio;
863 struct page *page = fio->encrypted_page ?
864 fio->encrypted_page : fio->page;
865
866 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
867 __is_meta_io(fio) ? META_GENERIC : DATA_GENERIC))
10f966bb 868 return -EFSCORRUPTED;
8648de2c
CY		 869
870 trace_f2fs_submit_page_bio(page, fio);
8648de2c 871
8896cbdf 872 if (bio && !page_is_mergeable(fio->sbi, bio, *fio->last_block,
0b20fcec
CY		 873 fio->new_blkaddr))
874 f2fs_submit_merged_ipu_write(fio->sbi, &bio, NULL);
8648de2c
CY		 875alloc_new:
876 if (!bio) {
a8affc03 877 bio = __bio_alloc(fio, BIO_MAX_VECS);
b7b911d5 878 __attach_io_flag(fio);
27aacd28
ST		 879 f2fs_set_bio_crypt_ctx(bio, fio->page->mapping->host,
880 fio->page->index, fio, GFP_NOIO);
8648de2c 881 bio_set_op_attrs(bio, fio->op, fio->op_flags);
8648de2c 882
0b20fcec
CY		 883 add_bio_entry(fio->sbi, bio, page, fio->temp);
884 } else {
27aacd28 885 if (add_ipu_page(fio, &bio, page))
0b20fcec 886 goto alloc_new;
8648de2c
CY		 887 }
888
889 if (fio->io_wbc)
9637d517 890 wbc_account_cgroup_owner(fio->io_wbc, page, PAGE_SIZE);
8648de2c
CY		 891
892 inc_page_count(fio->sbi, WB_DATA_TYPE(page));
893
894 *fio->last_block = fio->new_blkaddr;
895 *fio->bio = bio;
896
897 return 0;
898}
899
fe16efe6 900void f2fs_submit_page_write(struct f2fs_io_info *fio)
93dfe2ac 901{
05ca3632 902 struct f2fs_sb_info *sbi = fio->sbi;
458e6197 903 enum page_type btype = PAGE_TYPE_OF_BIO(fio->type);
a912b54d 904 struct f2fs_bio_info *io = sbi->write_io[btype] + fio->temp;
4375a336 905 struct page *bio_page;
93dfe2ac 906
b9109b0e 907 f2fs_bug_on(sbi, is_read_io(fio->op));
93dfe2ac 908
fb830fc5
CY		 909 down_write(&io->io_rwsem);
910next:
911 if (fio->in_list) {
912 spin_lock(&io->io_lock);
913 if (list_empty(&io->io_list)) {
914 spin_unlock(&io->io_lock);
fe16efe6 915 goto out;
fb830fc5
CY		 916 }
917 fio = list_first_entry(&io->io_list,
918 struct f2fs_io_info, list);
919 list_del(&fio->list);
920 spin_unlock(&io->io_lock);
921 }
93dfe2ac 922
93770ab7 923 verify_fio_blkaddr(fio);
93dfe2ac 924
4c8ff709
CY		 925 if (fio->encrypted_page)
926 bio_page = fio->encrypted_page;
927 else if (fio->compressed_page)
928 bio_page = fio->compressed_page;
929 else
930 bio_page = fio->page;
36951b38 931
ebf7c522
TM		 932 /* set submitted = true as a return value */
933 fio->submitted = true;
d68f735b 934
b9109b0e 935 inc_page_count(sbi, WB_DATA_TYPE(bio_page));
93dfe2ac 936
27aacd28
ST		 937 if (io->bio &&
938 (!io_is_mergeable(sbi, io->bio, io, fio, io->last_block_in_bio,
939 fio->new_blkaddr) ||
940 !f2fs_crypt_mergeable_bio(io->bio, fio->page->mapping->host,
941 bio_page->index, fio)))
458e6197 942 __submit_merged_bio(io);
93dfe2ac
JK		 943alloc_new:
944 if (io->bio == NULL) {
8223ecc4
CY		 945 if (F2FS_IO_ALIGNED(sbi) &&
946 (fio->type == DATA || fio->type == NODE) &&
0a595eba 947 fio->new_blkaddr & F2FS_IO_SIZE_MASK(sbi)) {
b9109b0e 948 dec_page_count(sbi, WB_DATA_TYPE(bio_page));
fe16efe6
CY		 949 fio->retry = true;
950 goto skip;
0a595eba 951 }
a8affc03 952 io->bio = __bio_alloc(fio, BIO_MAX_VECS);
27aacd28
ST		 953 f2fs_set_bio_crypt_ctx(io->bio, fio->page->mapping->host,
954 bio_page->index, fio, GFP_NOIO);
458e6197 955 io->fio = *fio;
93dfe2ac
JK		 956 }
957
a912b54d 958 if (bio_add_page(io->bio, bio_page, PAGE_SIZE, 0) < PAGE_SIZE) {
458e6197 959 __submit_merged_bio(io);
93dfe2ac
JK		 960 goto alloc_new;
961 }
962
578c6478 963 if (fio->io_wbc)
34e51a5e 964 wbc_account_cgroup_owner(fio->io_wbc, bio_page, PAGE_SIZE);
578c6478 965
7a9d7548 966 io->last_block_in_bio = fio->new_blkaddr;
fb830fc5
CY		 967
968 trace_f2fs_submit_page_write(fio->page, fio);
fe16efe6 969skip:
fb830fc5
CY		 970 if (fio->in_list)
971 goto next;
fe16efe6 972out:
4354994f 973 if (is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN) ||
00e09c0b 974 !f2fs_is_checkpoint_ready(sbi))
5ce80586 975 __submit_merged_bio(io);
df0f8dc0 976 up_write(&io->io_rwsem);
93dfe2ac
JK		 977}
978
13ba41e3 979static struct bio *f2fs_grab_read_bio(struct inode *inode, block_t blkaddr,
95ae251f 980 unsigned nr_pages, unsigned op_flag,
7f59b277 981 pgoff_t first_idx, bool for_write)
13ba41e3
JK		 982{
983 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
13ba41e3 984 struct bio *bio;
a4b68176 985 struct bio_post_read_ctx *ctx = NULL;
6dbb1796 986 unsigned int post_read_steps = 0;
13ba41e3 987
67883ade 988 bio = bio_alloc_bioset(for_write ? GFP_NOIO : GFP_KERNEL,
5f7136db 989 bio_max_segs(nr_pages), &f2fs_bioset);
6dbb1796 990 if (!bio)
13ba41e3 991 return ERR_PTR(-ENOMEM);
27aacd28
ST		 992
993 f2fs_set_bio_crypt_ctx(bio, inode, first_idx, NULL, GFP_NOFS);
994
13ba41e3
JK		 995 f2fs_target_device(sbi, blkaddr, bio);
996 bio->bi_end_io = f2fs_read_end_io;
e2e59414 997 bio_set_op_attrs(bio, REQ_OP_READ, op_flag);
13ba41e3 998
27aacd28 999 if (fscrypt_inode_uses_fs_layer_crypto(inode))
7f59b277 1000 post_read_steps |= STEP_DECRYPT;
95ae251f 1001
7f59b277
EB		 1002 if (f2fs_need_verity(inode, first_idx))
1003 post_read_steps |= STEP_VERITY;
1004
1005 /*
1006 * STEP_DECOMPRESS is handled specially, since a compressed file might
1007 * contain both compressed and uncompressed clusters. We'll allocate a
1008 * bio_post_read_ctx if the file is compressed, but the caller is
1009 * responsible for enabling STEP_DECOMPRESS if it's actually needed.
1010 */
1011
1012 if (post_read_steps || f2fs_compressed_file(inode)) {
e8ce5749 1013 /* Due to the mempool, this never fails. */
6dbb1796 1014 ctx = mempool_alloc(bio_post_read_ctx_pool, GFP_NOFS);
6dbb1796 1015 ctx->bio = bio;
4c8ff709 1016 ctx->sbi = sbi;
6dbb1796 1017 ctx->enabled_steps = post_read_steps;
4931e0c9 1018 ctx->fs_blkaddr = blkaddr;
6dbb1796 1019 bio->bi_private = ctx;
6dbb1796 1020 }
a4b68176 1021 iostat_alloc_and_bind_ctx(sbi, bio, ctx);
6dbb1796 1022
13ba41e3
JK		 1023 return bio;
1024}
1025
1026/* This can handle encryption stuffs */
1027static int f2fs_submit_page_read(struct inode *inode, struct page *page,
b7973091 1028 block_t blkaddr, int op_flags, bool for_write)
13ba41e3 1029{
93770ab7
CY		 1030 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1031 struct bio *bio;
13ba41e3 1032
b7973091 1033 bio = f2fs_grab_read_bio(inode, blkaddr, 1, op_flags,
7f59b277 1034 page->index, for_write);
13ba41e3
JK		 1035 if (IS_ERR(bio))
1036 return PTR_ERR(bio);
1037
0ded69f6
JK		 1038 /* wait for GCed page writeback via META_MAPPING */
1039 f2fs_wait_on_block_writeback(inode, blkaddr);
1040
13ba41e3
JK		 1041 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
1042 bio_put(bio);
1043 return -EFAULT;
1044 }
fb7d70db 1045 ClearPageError(page);
93770ab7 1046 inc_page_count(sbi, F2FS_RD_DATA);
8b83ac81 1047 f2fs_update_iostat(sbi, FS_DATA_READ_IO, F2FS_BLKSIZE);
93770ab7 1048 __submit_bio(sbi, bio, DATA);
13ba41e3
JK		 1049 return 0;
1050}
1051
46008c6d
CY		 1052static void __set_data_blkaddr(struct dnode_of_data *dn)
1053{
1054 struct f2fs_node *rn = F2FS_NODE(dn->node_page);
1055 __le32 *addr_array;
7a2af766
CY		 1056 int base = 0;
1057
1058 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
1059 base = get_extra_isize(dn->inode);
46008c6d
CY		 1060
1061 /* Get physical address of data block */
1062 addr_array = blkaddr_in_node(rn);
7a2af766 1063 addr_array[base + dn->ofs_in_node] = cpu_to_le32(dn->data_blkaddr);
46008c6d
CY		 1064}
1065
0a8165d7 1066/*
eb47b800
JK		 1067 * Lock ordering for the change of data block address:
1068 * ->data_page
1069 * ->node_page
1070 * update block addresses in the node page
1071 */
4d57b86d 1072void f2fs_set_data_blkaddr(struct dnode_of_data *dn)
eb47b800 1073{
bae0ee7a 1074 f2fs_wait_on_page_writeback(dn->node_page, NODE, true, true);
46008c6d
CY		 1075 __set_data_blkaddr(dn);
1076 if (set_page_dirty(dn->node_page))
12719ae1 1077 dn->node_changed = true;
eb47b800
JK		 1078}
1079
f28b3434
CY		 1080void f2fs_update_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr)
1081{
1082 dn->data_blkaddr = blkaddr;
4d57b86d 1083 f2fs_set_data_blkaddr(dn);
f28b3434
CY		 1084 f2fs_update_extent_cache(dn);
1085}
1086
46008c6d 1087/* dn->ofs_in_node will be returned with up-to-date last block pointer */
4d57b86d 1088int f2fs_reserve_new_blocks(struct dnode_of_data *dn, blkcnt_t count)
eb47b800 1089{
4081363f 1090 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
0abd675e 1091 int err;
eb47b800 1092
46008c6d
CY		 1093 if (!count)
1094 return 0;
1095
91942321 1096 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
eb47b800 1097 return -EPERM;
0abd675e
CY		 1098 if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
1099 return err;
eb47b800 1100
46008c6d
CY		 1101 trace_f2fs_reserve_new_blocks(dn->inode, dn->nid,
1102 dn->ofs_in_node, count);
1103
bae0ee7a 1104 f2fs_wait_on_page_writeback(dn->node_page, NODE, true, true);
46008c6d
CY		 1105
1106 for (; count > 0; dn->ofs_in_node++) {
a2ced1ce 1107 block_t blkaddr = f2fs_data_blkaddr(dn);
5f029c04 1108
46008c6d
CY		 1109 if (blkaddr == NULL_ADDR) {
1110 dn->data_blkaddr = NEW_ADDR;
1111 __set_data_blkaddr(dn);
1112 count--;
1113 }
1114 }
1115
1116 if (set_page_dirty(dn->node_page))
1117 dn->node_changed = true;
eb47b800
JK		 1118 return 0;
1119}
1120
46008c6d 1121/* Should keep dn->ofs_in_node unchanged */
4d57b86d 1122int f2fs_reserve_new_block(struct dnode_of_data *dn)
46008c6d
CY		 1123{
1124 unsigned int ofs_in_node = dn->ofs_in_node;
1125 int ret;
1126
4d57b86d 1127 ret = f2fs_reserve_new_blocks(dn, 1);
46008c6d
CY		 1128 dn->ofs_in_node = ofs_in_node;
1129 return ret;
1130}
1131
b600965c
HL		 1132int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
1133{
1134 bool need_put = dn->inode_page ? false : true;
1135 int err;
1136
4d57b86d 1137 err = f2fs_get_dnode_of_data(dn, index, ALLOC_NODE);
b600965c
HL		 1138 if (err)
1139 return err;
a8865372 1140
b600965c 1141 if (dn->data_blkaddr == NULL_ADDR)
4d57b86d 1142 err = f2fs_reserve_new_block(dn);
a8865372 1143 if (err || need_put)
b600965c
HL		 1144 f2fs_put_dnode(dn);
1145 return err;
1146}
1147
759af1c9 1148int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index)
eb47b800 1149{
94afd6d6 1150 struct extent_info ei = {0, };
759af1c9 1151 struct inode *inode = dn->inode;
028a41e8 1152
759af1c9
FL		 1153 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
1154 dn->data_blkaddr = ei.blk + index - ei.fofs;
1155 return 0;
429511cd 1156 }
028a41e8 1157
759af1c9 1158 return f2fs_reserve_block(dn, index);
eb47b800
JK		 1159}
1160
4d57b86d 1161struct page *f2fs_get_read_data_page(struct inode *inode, pgoff_t index,
04d328de 1162 int op_flags, bool for_write)
eb47b800 1163{
eb47b800
JK		 1164 struct address_space *mapping = inode->i_mapping;
1165 struct dnode_of_data dn;
1166 struct page *page;
94afd6d6 1167 struct extent_info ei = {0, };
eb47b800 1168 int err;
4375a336 1169
a56c7c6f 1170 page = f2fs_grab_cache_page(mapping, index, for_write);
650495de
JK		 1171 if (!page)
1172 return ERR_PTR(-ENOMEM);
1173
cb3bc9ee
CY		 1174 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
1175 dn.data_blkaddr = ei.blk + index - ei.fofs;
93770ab7
CY		 1176 if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), dn.data_blkaddr,
1177 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 1178 err = -EFSCORRUPTED;
93770ab7
CY		 1179 goto put_err;
1180 }
cb3bc9ee
CY		 1181 goto got_it;
1182 }
1183
eb47b800 1184 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1185 err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
86531d6b
JK		 1186 if (err)
1187 goto put_err;
eb47b800
JK		 1188 f2fs_put_dnode(&dn);
1189
6bacf52f 1190 if (unlikely(dn.data_blkaddr == NULL_ADDR)) {
86531d6b
JK		 1191 err = -ENOENT;
1192 goto put_err;
650495de 1193 }
93770ab7
CY		 1194 if (dn.data_blkaddr != NEW_ADDR &&
1195 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
1196 dn.data_blkaddr,
1197 DATA_GENERIC_ENHANCE)) {
10f966bb 1198 err = -EFSCORRUPTED;
93770ab7
CY		 1199 goto put_err;
1200 }
cb3bc9ee 1201got_it:
43f3eae1
JK		 1202 if (PageUptodate(page)) {
1203 unlock_page(page);
eb47b800 1204 return page;
43f3eae1 1205 }
eb47b800 1206
d59ff4df
JK		 1207 /*
1208 * A new dentry page is allocated but not able to be written, since its
1209 * new inode page couldn't be allocated due to -ENOSPC.
1210 * In such the case, its blkaddr can be remained as NEW_ADDR.
4d57b86d
CY		 1211 * see, f2fs_add_link -> f2fs_get_new_data_page ->
1212 * f2fs_init_inode_metadata.
d59ff4df
JK		 1213 */
1214 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 1215 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 1216 if (!PageUptodate(page))
1217 SetPageUptodate(page);
43f3eae1 1218 unlock_page(page);
d59ff4df
JK		 1219 return page;
1220 }
eb47b800 1221
b7973091
JY		 1222 err = f2fs_submit_page_read(inode, page, dn.data_blkaddr,
1223 op_flags, for_write);
393ff91f 1224 if (err)
86531d6b 1225 goto put_err;
43f3eae1 1226 return page;
86531d6b
JK		 1227
1228put_err:
1229 f2fs_put_page(page, 1);
1230 return ERR_PTR(err);
43f3eae1
JK		 1231}
1232
4d57b86d 1233struct page *f2fs_find_data_page(struct inode *inode, pgoff_t index)
43f3eae1
JK		 1234{
1235 struct address_space *mapping = inode->i_mapping;
1236 struct page *page;
1237
1238 page = find_get_page(mapping, index);
1239 if (page && PageUptodate(page))
1240 return page;
1241 f2fs_put_page(page, 0);
1242
4d57b86d 1243 page = f2fs_get_read_data_page(inode, index, 0, false);
43f3eae1
JK		 1244 if (IS_ERR(page))
1245 return page;
1246
1247 if (PageUptodate(page))
1248 return page;
1249
1250 wait_on_page_locked(page);
1251 if (unlikely(!PageUptodate(page))) {
1252 f2fs_put_page(page, 0);
1253 return ERR_PTR(-EIO);
1254 }
1255 return page;
1256}
1257
1258/*
1259 * If it tries to access a hole, return an error.
1260 * Because, the callers, functions in dir.c and GC, should be able to know
1261 * whether this page exists or not.
1262 */
4d57b86d 1263struct page *f2fs_get_lock_data_page(struct inode *inode, pgoff_t index,
a56c7c6f 1264 bool for_write)
43f3eae1
JK		 1265{
1266 struct address_space *mapping = inode->i_mapping;
1267 struct page *page;
1268repeat:
4d57b86d 1269 page = f2fs_get_read_data_page(inode, index, 0, for_write);
43f3eae1
JK		 1270 if (IS_ERR(page))
1271 return page;
393ff91f 1272
43f3eae1 1273 /* wait for read completion */
393ff91f 1274 lock_page(page);
6bacf52f 1275 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 1276 f2fs_put_page(page, 1);
1277 goto repeat;
eb47b800 1278 }
1563ac75
CY		 1279 if (unlikely(!PageUptodate(page))) {
1280 f2fs_put_page(page, 1);
1281 return ERR_PTR(-EIO);
1282 }
eb47b800
JK		 1283 return page;
1284}
1285
0a8165d7 1286/*
eb47b800
JK		 1287 * Caller ensures that this data page is never allocated.
1288 * A new zero-filled data page is allocated in the page cache.
39936837 1289 *
4f4124d0
CY		 1290 * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and
1291 * f2fs_unlock_op().
470f00e9
CY		 1292 * Note that, ipage is set only by make_empty_dir, and if any error occur,
1293 * ipage should be released by this function.
eb47b800 1294 */
4d57b86d 1295struct page *f2fs_get_new_data_page(struct inode *inode,
a8865372 1296 struct page *ipage, pgoff_t index, bool new_i_size)
eb47b800 1297{
eb47b800
JK		 1298 struct address_space *mapping = inode->i_mapping;
1299 struct page *page;
1300 struct dnode_of_data dn;
1301 int err;
7612118a 1302
a56c7c6f 1303 page = f2fs_grab_cache_page(mapping, index, true);
470f00e9
CY		 1304 if (!page) {
1305 /*
1306 * before exiting, we should make sure ipage will be released
1307 * if any error occur.
1308 */
1309 f2fs_put_page(ipage, 1);
01f28610 1310 return ERR_PTR(-ENOMEM);
470f00e9 1311 }
eb47b800 1312
a8865372 1313 set_new_dnode(&dn, inode, ipage, NULL, 0);
b600965c 1314 err = f2fs_reserve_block(&dn, index);
01f28610
JK		 1315 if (err) {
1316 f2fs_put_page(page, 1);
eb47b800 1317 return ERR_PTR(err);
a8865372 1318 }
01f28610
JK		 1319 if (!ipage)
1320 f2fs_put_dnode(&dn);
eb47b800
JK		 1321
1322 if (PageUptodate(page))
01f28610 1323 goto got_it;
eb47b800
JK		 1324
1325 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 1326 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 1327 if (!PageUptodate(page))
1328 SetPageUptodate(page);
eb47b800 1329 } else {
4375a336 1330 f2fs_put_page(page, 1);
a8865372 1331
7612118a
JK		 1332 /* if ipage exists, blkaddr should be NEW_ADDR */
1333 f2fs_bug_on(F2FS_I_SB(inode), ipage);
4d57b86d 1334 page = f2fs_get_lock_data_page(inode, index, true);
4375a336 1335 if (IS_ERR(page))
7612118a 1336 return page;
eb47b800 1337 }
01f28610 1338got_it:
9edcdabf 1339 if (new_i_size && i_size_read(inode) <
ee6d182f 1340 ((loff_t)(index + 1) << PAGE_SHIFT))
fc9581c8 1341 f2fs_i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT));
eb47b800
JK		 1342 return page;
1343}
1344
d5097be5 1345static int __allocate_data_block(struct dnode_of_data *dn, int seg_type)
bfad7c2d 1346{
4081363f 1347 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
bfad7c2d 1348 struct f2fs_summary sum;
bfad7c2d 1349 struct node_info ni;
6aa58d8a 1350 block_t old_blkaddr;
46008c6d 1351 blkcnt_t count = 1;
0abd675e 1352 int err;
bfad7c2d 1353
91942321 1354 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
bfad7c2d 1355 return -EPERM;
df6136ef 1356
7735730d
CY		 1357 err = f2fs_get_node_info(sbi, dn->nid, &ni);
1358 if (err)
1359 return err;
1360
a2ced1ce 1361 dn->data_blkaddr = f2fs_data_blkaddr(dn);
f847c699 1362 if (dn->data_blkaddr != NULL_ADDR)
df6136ef
CY		 1363 goto alloc;
1364
0abd675e
CY		 1365 if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
1366 return err;
bfad7c2d 1367
df6136ef 1368alloc:
bfad7c2d 1369 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
6aa58d8a
CY		 1370 old_blkaddr = dn->data_blkaddr;
1371 f2fs_allocate_data_block(sbi, NULL, old_blkaddr, &dn->data_blkaddr,
093749e2 1372 &sum, seg_type, NULL);
6ce19aff 1373 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO) {
6aa58d8a
CY		 1374 invalidate_mapping_pages(META_MAPPING(sbi),
1375 old_blkaddr, old_blkaddr);
6ce19aff
CY		 1376 f2fs_invalidate_compress_page(sbi, old_blkaddr);
1377 }
86f35dc3 1378 f2fs_update_data_blkaddr(dn, dn->data_blkaddr);
bfad7c2d 1379
0a4daae5
JK		 1380 /*
1381 * i_size will be updated by direct_IO. Otherwise, we'll get stale
1382 * data from unwritten block via dio_read.
1383 */
bfad7c2d
JK		 1384 return 0;
1385}
1386
0ef81833 1387void f2fs_do_map_lock(struct f2fs_sb_info *sbi, int flag, bool lock)
59c9081b
YH		 1388{
1389 if (flag == F2FS_GET_BLOCK_PRE_AIO) {
1390 if (lock)
1391 down_read(&sbi->node_change);
1392 else
1393 up_read(&sbi->node_change);
1394 } else {
1395 if (lock)
1396 f2fs_lock_op(sbi);
1397 else
1398 f2fs_unlock_op(sbi);
1399 }
1400}
1401
0a8165d7 1402/*
7a88ddb5
CY		 1403 * f2fs_map_blocks() tries to find or build mapping relationship which
1404 * maps continuous logical blocks to physical blocks, and return such
1405 * info via f2fs_map_blocks structure.
eb47b800 1406 */
d323d005 1407int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
e2b4e2bc 1408 int create, int flag)
eb47b800 1409{
003a3e1d 1410 unsigned int maxblocks = map->m_len;
eb47b800 1411 struct dnode_of_data dn;
f9811703 1412 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
f9d6d059 1413 int mode = map->m_may_create ? ALLOC_NODE : LOOKUP_NODE;
46008c6d 1414 pgoff_t pgofs, end_offset, end;
bfad7c2d 1415 int err = 0, ofs = 1;
46008c6d
CY		 1416 unsigned int ofs_in_node, last_ofs_in_node;
1417 blkcnt_t prealloc;
94afd6d6 1418 struct extent_info ei = {0, };
7df3a431 1419 block_t blkaddr;
c4020b2d 1420 unsigned int start_pgofs;
71f2c820 1421 int bidx = 0;
eb47b800 1422
dfd02e4d
CY		 1423 if (!maxblocks)
1424 return 0;
1425
71f2c820
CY		 1426 map->m_bdev = inode->i_sb->s_bdev;
1427 map->m_multidev_dio =
1428 f2fs_allow_multi_device_dio(F2FS_I_SB(inode), flag);
1429
003a3e1d
JK		 1430 map->m_len = 0;
1431 map->m_flags = 0;
1432
1433 /* it only supports block size == page size */
1434 pgofs = (pgoff_t)map->m_lblk;
46008c6d 1435 end = pgofs + maxblocks;
eb47b800 1436
24b84912 1437 if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) {
b0332a0f 1438 if (f2fs_lfs_mode(sbi) && flag == F2FS_GET_BLOCK_DIO &&
f4f0b677
JZ		 1439 map->m_may_create)
1440 goto next_dnode;
1441
003a3e1d
JK		 1442 map->m_pblk = ei.blk + pgofs - ei.fofs;
1443 map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs);
1444 map->m_flags = F2FS_MAP_MAPPED;
c4020b2d
CY		 1445 if (map->m_next_extent)
1446 *map->m_next_extent = pgofs + map->m_len;
1e78e8bd
ST		 1447
1448 /* for hardware encryption, but to avoid potential issue in future */
1449 if (flag == F2FS_GET_BLOCK_DIO)
1450 f2fs_wait_on_block_writeback_range(inode,
1451 map->m_pblk, map->m_len);
71f2c820
CY		 1452
1453 if (map->m_multidev_dio) {
1454 block_t blk_addr = map->m_pblk;
1455
1456 bidx = f2fs_target_device_index(sbi, map->m_pblk);
1457
1458 map->m_bdev = FDEV(bidx).bdev;
1459 map->m_pblk -= FDEV(bidx).start_blk;
1460 map->m_len = min(map->m_len,
1461 FDEV(bidx).end_blk + 1 - map->m_pblk);
1462
1463 if (map->m_may_create)
1464 f2fs_update_device_state(sbi, inode->i_ino,
1465 blk_addr, map->m_len);
1466 }
bfad7c2d 1467 goto out;
a2e7d1bf 1468 }
bfad7c2d 1469
4fe71e88 1470next_dnode:
f9d6d059 1471 if (map->m_may_create)
0ef81833 1472 f2fs_do_map_lock(sbi, flag, true);
eb47b800
JK		 1473
1474 /* When reading holes, we need its node page */
1475 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1476 err = f2fs_get_dnode_of_data(&dn, pgofs, mode);
1ec79083 1477 if (err) {
43473f96
CY		 1478 if (flag == F2FS_GET_BLOCK_BMAP)
1479 map->m_pblk = 0;
adf9ea89 1480
da85985c 1481 if (err == -ENOENT) {
adf9ea89
CY		 1482 /*
1483 * There is one exceptional case that read_node_page()
1484 * may return -ENOENT due to filesystem has been
1485 * shutdown or cp_error, so force to convert error
1486 * number to EIO for such case.
1487 */
1488 if (map->m_may_create &&
1489 (is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN) ||
1490 f2fs_cp_error(sbi))) {
1491 err = -EIO;
1492 goto unlock_out;
1493 }
1494
bfad7c2d 1495 err = 0;
da85985c
CY		 1496 if (map->m_next_pgofs)
1497 *map->m_next_pgofs =
4d57b86d 1498 f2fs_get_next_page_offset(&dn, pgofs);
c4020b2d
CY		 1499 if (map->m_next_extent)
1500 *map->m_next_extent =
4d57b86d 1501 f2fs_get_next_page_offset(&dn, pgofs);
da85985c 1502 }
bfad7c2d 1503 goto unlock_out;
848753aa 1504 }
973163fc 1505
c4020b2d 1506 start_pgofs = pgofs;
46008c6d 1507 prealloc = 0;
230436b3 1508 last_ofs_in_node = ofs_in_node = dn.ofs_in_node;
81ca7350 1509 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
4fe71e88
CY		 1510
1511next_block:
a2ced1ce 1512 blkaddr = f2fs_data_blkaddr(&dn);
4fe71e88 1513
c9b60788 1514 if (__is_valid_data_blkaddr(blkaddr) &&
93770ab7 1515 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC_ENHANCE)) {
10f966bb 1516 err = -EFSCORRUPTED;
c9b60788
CY		 1517 goto sync_out;
1518 }
1519
93770ab7 1520 if (__is_valid_data_blkaddr(blkaddr)) {
f847c699 1521 /* use out-place-update for driect IO under LFS mode */
b0332a0f 1522 if (f2fs_lfs_mode(sbi) && flag == F2FS_GET_BLOCK_DIO &&
f9d6d059 1523 map->m_may_create) {
f847c699 1524 err = __allocate_data_block(&dn, map->m_seg_type);
05e36006
CY		 1525 if (err)
1526 goto sync_out;
1527 blkaddr = dn.data_blkaddr;
1528 set_inode_flag(inode, FI_APPEND_WRITE);
f847c699
CY		 1529 }
1530 } else {
973163fc 1531 if (create) {
f9811703
CY		 1532 if (unlikely(f2fs_cp_error(sbi))) {
1533 err = -EIO;
4fe71e88 1534 goto sync_out;
f9811703 1535 }
24b84912 1536 if (flag == F2FS_GET_BLOCK_PRE_AIO) {
46008c6d
CY		 1537 if (blkaddr == NULL_ADDR) {
1538 prealloc++;
1539 last_ofs_in_node = dn.ofs_in_node;
1540 }
24b84912 1541 } else {
0a4daae5
JK		 1542 WARN_ON(flag != F2FS_GET_BLOCK_PRE_DIO &&
1543 flag != F2FS_GET_BLOCK_DIO);
d5097be5
HL		 1544 err = __allocate_data_block(&dn,
1545 map->m_seg_type);
6f2d8ed6 1546 if (!err)
91942321 1547 set_inode_flag(inode, FI_APPEND_WRITE);
24b84912 1548 }
973163fc 1549 if (err)
4fe71e88 1550 goto sync_out;
3f2be043 1551 map->m_flags |= F2FS_MAP_NEW;
4fe71e88 1552 blkaddr = dn.data_blkaddr;
973163fc 1553 } else {
bbe1da7e
CY		 1554 if (f2fs_compressed_file(inode) &&
1555 f2fs_sanity_check_cluster(&dn) &&
1556 (flag != F2FS_GET_BLOCK_FIEMAP ||
1557 IS_ENABLED(CONFIG_F2FS_CHECK_FS))) {
1558 err = -EFSCORRUPTED;
1559 goto sync_out;
1560 }
43473f96
CY		 1561 if (flag == F2FS_GET_BLOCK_BMAP) {
1562 map->m_pblk = 0;
1563 goto sync_out;
1564 }
c4020b2d
CY		 1565 if (flag == F2FS_GET_BLOCK_PRECACHE)
1566 goto sync_out;
da85985c
CY		 1567 if (flag == F2FS_GET_BLOCK_FIEMAP &&
1568 blkaddr == NULL_ADDR) {
1569 if (map->m_next_pgofs)
1570 *map->m_next_pgofs = pgofs + 1;
4c2ac6a8 1571 goto sync_out;
da85985c 1572 }
f3d98e74
CY		 1573 if (flag != F2FS_GET_BLOCK_FIEMAP) {
1574 /* for defragment case */
1575 if (map->m_next_pgofs)
1576 *map->m_next_pgofs = pgofs + 1;
4fe71e88 1577 goto sync_out;
f3d98e74 1578 }
e2b4e2bc 1579 }
e2b4e2bc 1580 }
eb47b800 1581
46008c6d
CY		 1582 if (flag == F2FS_GET_BLOCK_PRE_AIO)
1583 goto skip;
1584
71f2c820
CY		 1585 if (map->m_multidev_dio)
1586 bidx = f2fs_target_device_index(sbi, blkaddr);
1587
4fe71e88
CY		 1588 if (map->m_len == 0) {
1589 /* preallocated unwritten block should be mapped for fiemap. */
1590 if (blkaddr == NEW_ADDR)
1591 map->m_flags |= F2FS_MAP_UNWRITTEN;
1592 map->m_flags |= F2FS_MAP_MAPPED;
1593
1594 map->m_pblk = blkaddr;
1595 map->m_len = 1;
71f2c820
CY		 1596
1597 if (map->m_multidev_dio)
1598 map->m_bdev = FDEV(bidx).bdev;
4fe71e88
CY		 1599 } else if ((map->m_pblk != NEW_ADDR &&
1600 blkaddr == (map->m_pblk + ofs)) ||
b439b103 1601 (map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR) ||
46008c6d 1602 flag == F2FS_GET_BLOCK_PRE_DIO) {
71f2c820
CY		 1603 if (map->m_multidev_dio && map->m_bdev != FDEV(bidx).bdev)
1604 goto sync_out;
4fe71e88
CY		 1605 ofs++;
1606 map->m_len++;
1607 } else {
1608 goto sync_out;
1609 }
bfad7c2d 1610
46008c6d 1611skip:
bfad7c2d
JK		 1612 dn.ofs_in_node++;
1613 pgofs++;
1614
46008c6d
CY		 1615 /* preallocate blocks in batch for one dnode page */
1616 if (flag == F2FS_GET_BLOCK_PRE_AIO &&
1617 (pgofs == end || dn.ofs_in_node == end_offset)) {
7df3a431 1618
46008c6d 1619 dn.ofs_in_node = ofs_in_node;
4d57b86d 1620 err = f2fs_reserve_new_blocks(&dn, prealloc);
46008c6d
CY		 1621 if (err)
1622 goto sync_out;
bfad7c2d 1623
46008c6d
CY		 1624 map->m_len += dn.ofs_in_node - ofs_in_node;
1625 if (prealloc && dn.ofs_in_node != last_ofs_in_node + 1) {
1626 err = -ENOSPC;
1627 goto sync_out;
3104af35 1628 }
46008c6d
CY		 1629 dn.ofs_in_node = end_offset;
1630 }
1631
1632 if (pgofs >= end)
1633 goto sync_out;
1634 else if (dn.ofs_in_node < end_offset)
1635 goto next_block;
1636
c4020b2d
CY		 1637 if (flag == F2FS_GET_BLOCK_PRECACHE) {
1638 if (map->m_flags & F2FS_MAP_MAPPED) {
1639 unsigned int ofs = start_pgofs - map->m_lblk;
1640
1641 f2fs_update_extent_cache_range(&dn,
1642 start_pgofs, map->m_pblk + ofs,
1643 map->m_len - ofs);
1644 }
1645 }
1646
46008c6d
CY		 1647 f2fs_put_dnode(&dn);
1648
f9d6d059 1649 if (map->m_may_create) {
0ef81833 1650 f2fs_do_map_lock(sbi, flag, false);
6f2d8ed6 1651 f2fs_balance_fs(sbi, dn.node_changed);
eb47b800 1652 }
46008c6d 1653 goto next_dnode;
7df3a431 1654
bfad7c2d 1655sync_out:
1e78e8bd 1656
71f2c820
CY		 1657 if (flag == F2FS_GET_BLOCK_DIO && map->m_flags & F2FS_MAP_MAPPED) {
1658 /*
1659 * for hardware encryption, but to avoid potential issue
1660 * in future
1661 */
1e78e8bd
ST		 1662 f2fs_wait_on_block_writeback_range(inode,
1663 map->m_pblk, map->m_len);
e3b49ea3
HJK		 1664 invalidate_mapping_pages(META_MAPPING(sbi),
1665 map->m_pblk, map->m_pblk);
1e78e8bd 1666
71f2c820
CY		 1667 if (map->m_multidev_dio) {
1668 block_t blk_addr = map->m_pblk;
1669
1670 bidx = f2fs_target_device_index(sbi, map->m_pblk);
1671
1672 map->m_bdev = FDEV(bidx).bdev;
1673 map->m_pblk -= FDEV(bidx).start_blk;
1674
1675 if (map->m_may_create)
1676 f2fs_update_device_state(sbi, inode->i_ino,
1677 blk_addr, map->m_len);
1678
1679 f2fs_bug_on(sbi, blk_addr + map->m_len >
1680 FDEV(bidx).end_blk + 1);
1681 }
1682 }
1683
c4020b2d
CY		 1684 if (flag == F2FS_GET_BLOCK_PRECACHE) {
1685 if (map->m_flags & F2FS_MAP_MAPPED) {
1686 unsigned int ofs = start_pgofs - map->m_lblk;
1687
1688 f2fs_update_extent_cache_range(&dn,
1689 start_pgofs, map->m_pblk + ofs,
1690 map->m_len - ofs);
1691 }
1692 if (map->m_next_extent)
1693 *map->m_next_extent = pgofs + 1;
1694 }
eb47b800 1695 f2fs_put_dnode(&dn);
bfad7c2d 1696unlock_out:
f9d6d059 1697 if (map->m_may_create) {
0ef81833 1698 f2fs_do_map_lock(sbi, flag, false);
6f2d8ed6 1699 f2fs_balance_fs(sbi, dn.node_changed);
2a340760 1700 }
bfad7c2d 1701out:
71f2c820 1702 trace_f2fs_map_blocks(inode, map, create, flag, err);
bfad7c2d 1703 return err;
eb47b800
JK		 1704}
1705
b91050a8
HL		 1706bool f2fs_overwrite_io(struct inode *inode, loff_t pos, size_t len)
1707{
1708 struct f2fs_map_blocks map;
1709 block_t last_lblk;
1710 int err;
1711
1712 if (pos + len > i_size_read(inode))
1713 return false;
1714
1715 map.m_lblk = F2FS_BYTES_TO_BLK(pos);
1716 map.m_next_pgofs = NULL;
1717 map.m_next_extent = NULL;
1718 map.m_seg_type = NO_CHECK_TYPE;
f4f0b677 1719 map.m_may_create = false;
b91050a8
HL		 1720 last_lblk = F2FS_BLK_ALIGN(pos + len);
1721
1722 while (map.m_lblk < last_lblk) {
1723 map.m_len = last_lblk - map.m_lblk;
1724 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
1725 if (err || map.m_len == 0)
1726 return false;
1727 map.m_lblk += map.m_len;
1728 }
1729 return true;
1730}
1731
43b9d4b4
JK		 1732static inline u64 bytes_to_blks(struct inode *inode, u64 bytes)
1733{
1734 return (bytes >> inode->i_blkbits);
1735}
1736
1737static inline u64 blks_to_bytes(struct inode *inode, u64 blks)
1738{
1739 return (blks << inode->i_blkbits);
1740}
1741
003a3e1d 1742static int __get_data_block(struct inode *inode, sector_t iblock,
da85985c 1743 struct buffer_head *bh, int create, int flag,
f9d6d059 1744 pgoff_t *next_pgofs, int seg_type, bool may_write)
003a3e1d
JK		 1745{
1746 struct f2fs_map_blocks map;
a7de6086 1747 int err;
003a3e1d
JK		 1748
1749 map.m_lblk = iblock;
43b9d4b4 1750 map.m_len = bytes_to_blks(inode, bh->b_size);
da85985c 1751 map.m_next_pgofs = next_pgofs;
c4020b2d 1752 map.m_next_extent = NULL;
d5097be5 1753 map.m_seg_type = seg_type;
f9d6d059 1754 map.m_may_create = may_write;
003a3e1d 1755
a7de6086
JK		 1756 err = f2fs_map_blocks(inode, &map, create, flag);
1757 if (!err) {
003a3e1d
JK		 1758 map_bh(bh, inode->i_sb, map.m_pblk);
1759 bh->b_state = (bh->b_state & ~F2FS_MAP_FLAGS) | map.m_flags;
43b9d4b4 1760 bh->b_size = blks_to_bytes(inode, map.m_len);
71f2c820
CY		 1761
1762 if (map.m_multidev_dio)
1763 bh->b_bdev = map.m_bdev;
003a3e1d 1764 }
a7de6086 1765 return err;
003a3e1d
JK		 1766}
1767
f9d6d059
CY		 1768static int get_data_block_dio_write(struct inode *inode, sector_t iblock,
1769 struct buffer_head *bh_result, int create)
1770{
1771 return __get_data_block(inode, iblock, bh_result, create,
1772 F2FS_GET_BLOCK_DIO, NULL,
1773 f2fs_rw_hint_to_seg_type(inode->i_write_hint),
ebc29b62 1774 true);
e2b4e2bc
CY		 1775}
1776
1777static int get_data_block_dio(struct inode *inode, sector_t iblock,
ccfb3000
JK		 1778 struct buffer_head *bh_result, int create)
1779{
e2b4e2bc 1780 return __get_data_block(inode, iblock, bh_result, create,
f9d6d059
CY		 1781 F2FS_GET_BLOCK_DIO, NULL,
1782 f2fs_rw_hint_to_seg_type(inode->i_write_hint),
1783 false);
ccfb3000
JK		 1784}
1785
442a9dbd
CY		 1786static int f2fs_xattr_fiemap(struct inode *inode,
1787 struct fiemap_extent_info *fieinfo)
1788{
1789 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1790 struct page *page;
1791 struct node_info ni;
1792 __u64 phys = 0, len;
1793 __u32 flags;
1794 nid_t xnid = F2FS_I(inode)->i_xattr_nid;
1795 int err = 0;
1796
1797 if (f2fs_has_inline_xattr(inode)) {
1798 int offset;
1799
1800 page = f2fs_grab_cache_page(NODE_MAPPING(sbi),
1801 inode->i_ino, false);
1802 if (!page)
1803 return -ENOMEM;
1804
7735730d
CY		 1805 err = f2fs_get_node_info(sbi, inode->i_ino, &ni);
1806 if (err) {
1807 f2fs_put_page(page, 1);
1808 return err;
1809 }
442a9dbd 1810
6cbfcab5 1811 phys = blks_to_bytes(inode, ni.blk_addr);
442a9dbd
CY		 1812 offset = offsetof(struct f2fs_inode, i_addr) +
1813 sizeof(__le32) * (DEF_ADDRS_PER_INODE -
b323fd28 1814 get_inline_xattr_addrs(inode));
442a9dbd
CY		 1815
1816 phys += offset;
1817 len = inline_xattr_size(inode);
1818
1819 f2fs_put_page(page, 1);
1820
1821 flags = FIEMAP_EXTENT_DATA_INLINE | FIEMAP_EXTENT_NOT_ALIGNED;
1822
1823 if (!xnid)
1824 flags |= FIEMAP_EXTENT_LAST;
1825
1826 err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
dd5a09bd 1827 trace_f2fs_fiemap(inode, 0, phys, len, flags, err);
442a9dbd
CY		 1828 if (err || err == 1)
1829 return err;
1830 }
1831
1832 if (xnid) {
1833 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), xnid, false);
1834 if (!page)
1835 return -ENOMEM;
1836
7735730d
CY		 1837 err = f2fs_get_node_info(sbi, xnid, &ni);
1838 if (err) {
1839 f2fs_put_page(page, 1);
1840 return err;
1841 }
442a9dbd 1842
6cbfcab5 1843 phys = blks_to_bytes(inode, ni.blk_addr);
442a9dbd
CY		 1844 len = inode->i_sb->s_blocksize;
1845
1846 f2fs_put_page(page, 1);
1847
1848 flags = FIEMAP_EXTENT_LAST;
1849 }
1850
dd5a09bd 1851 if (phys) {
442a9dbd 1852 err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
dd5a09bd
CY		 1853 trace_f2fs_fiemap(inode, 0, phys, len, flags, err);
1854 }
442a9dbd
CY		 1855
1856 return (err < 0 ? err : 0);
1857}
1858
bf38fbad
CY		 1859static loff_t max_inode_blocks(struct inode *inode)
1860{
1861 loff_t result = ADDRS_PER_INODE(inode);
1862 loff_t leaf_count = ADDRS_PER_BLOCK(inode);
1863
1864 /* two direct node blocks */
1865 result += (leaf_count * 2);
1866
1867 /* two indirect node blocks */
1868 leaf_count *= NIDS_PER_BLOCK;
1869 result += (leaf_count * 2);
1870
1871 /* one double indirect node block */
1872 leaf_count *= NIDS_PER_BLOCK;
1873 result += leaf_count;
1874
1875 return result;
1876}
1877
9ab70134
JK		 1878int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1879 u64 start, u64 len)
1880{
b876f4c9 1881 struct f2fs_map_blocks map;
7f63eb77 1882 sector_t start_blk, last_blk;
da85985c 1883 pgoff_t next_pgofs;
7f63eb77
JK		 1884 u64 logical = 0, phys = 0, size = 0;
1885 u32 flags = 0;
7f63eb77 1886 int ret = 0;
093f0bac 1887 bool compr_cluster = false, compr_appended;
bf38fbad 1888 unsigned int cluster_size = F2FS_I(inode)->i_cluster_size;
093f0bac 1889 unsigned int count_in_cluster = 0;
0bb2045c 1890 loff_t maxbytes;
7f63eb77 1891
c4020b2d
CY		 1892 if (fieinfo->fi_flags & FIEMAP_FLAG_CACHE) {
1893 ret = f2fs_precache_extents(inode);
1894 if (ret)
1895 return ret;
1896 }
1897
45dd052e 1898 ret = fiemap_prep(inode, fieinfo, start, &len, FIEMAP_FLAG_XATTR);
7f63eb77
JK		 1899 if (ret)
1900 return ret;
1901
f1b43d4c
CY		 1902 inode_lock(inode);
1903
0bb2045c
CX		 1904 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
1905 if (start > maxbytes) {
1906 ret = -EFBIG;
1907 goto out;
1908 }
1909
1910 if (len > maxbytes || (maxbytes - len) < start)
1911 len = maxbytes - start;
1912
442a9dbd
CY		 1913 if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
1914 ret = f2fs_xattr_fiemap(inode, fieinfo);
1915 goto out;
1916 }
1917
7975f349 1918 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
67f8cf3c
JK		 1919 ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len);
1920 if (ret != -EAGAIN)
f1b43d4c 1921 goto out;
67f8cf3c
JK		 1922 }
1923
6cbfcab5
JK		 1924 if (bytes_to_blks(inode, len) == 0)
1925 len = blks_to_bytes(inode, 1);
7f63eb77 1926
6cbfcab5
JK		 1927 start_blk = bytes_to_blks(inode, start);
1928 last_blk = bytes_to_blks(inode, start + len - 1);
9a950d52 1929
7f63eb77 1930next:
b876f4c9
JK		 1931 memset(&map, 0, sizeof(map));
1932 map.m_lblk = start_blk;
1933 map.m_len = bytes_to_blks(inode, len);
1934 map.m_next_pgofs = &next_pgofs;
1935 map.m_seg_type = NO_CHECK_TYPE;
7f63eb77 1936
093f0bac
DJ		 1937 if (compr_cluster) {
1938 map.m_lblk += 1;
1939 map.m_len = cluster_size - count_in_cluster;
1940 }
bf38fbad 1941
b876f4c9 1942 ret = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_FIEMAP);
7f63eb77
JK		 1943 if (ret)
1944 goto out;
1945
1946 /* HOLE */
093f0bac 1947 if (!compr_cluster && !(map.m_flags & F2FS_MAP_FLAGS)) {
da85985c 1948 start_blk = next_pgofs;
58736fa6 1949
6cbfcab5 1950 if (blks_to_bytes(inode, start_blk) < blks_to_bytes(inode,
bf38fbad 1951 max_inode_blocks(inode)))
9a950d52 1952 goto prep_next;
58736fa6 1953
9a950d52
FL		 1954 flags |= FIEMAP_EXTENT_LAST;
1955 }
7f63eb77 1956
093f0bac
DJ		 1957 compr_appended = false;
1958 /* In a case of compressed cluster, append this to the last extent */
1959 if (compr_cluster && ((map.m_flags & F2FS_MAP_UNWRITTEN) ||
1960 !(map.m_flags & F2FS_MAP_FLAGS))) {
1961 compr_appended = true;
1962 goto skip_fill;
1963 }
1964
da5af127 1965 if (size) {
0953fe86 1966 flags |= FIEMAP_EXTENT_MERGED;
62230e0d 1967 if (IS_ENCRYPTED(inode))
da5af127
CY		 1968 flags |= FIEMAP_EXTENT_DATA_ENCRYPTED;
1969
9a950d52
FL		 1970 ret = fiemap_fill_next_extent(fieinfo, logical,
1971 phys, size, flags);
dd5a09bd 1972 trace_f2fs_fiemap(inode, logical, phys, size, flags, ret);
bf38fbad
CY		 1973 if (ret)
1974 goto out;
1975 size = 0;
da5af127 1976 }
7f63eb77 1977
bf38fbad 1978 if (start_blk > last_blk)
9a950d52 1979 goto out;
7f63eb77 1980
093f0bac 1981skip_fill:
b876f4c9 1982 if (map.m_pblk == COMPRESS_ADDR) {
bf38fbad 1983 compr_cluster = true;
093f0bac
DJ		 1984 count_in_cluster = 1;
1985 } else if (compr_appended) {
1986 unsigned int appended_blks = cluster_size -
1987 count_in_cluster + 1;
1988 size += blks_to_bytes(inode, appended_blks);
1989 start_blk += appended_blks;
1990 compr_cluster = false;
1991 } else {
1992 logical = blks_to_bytes(inode, start_blk);
1993 phys = __is_valid_data_blkaddr(map.m_pblk) ?
1994 blks_to_bytes(inode, map.m_pblk) : 0;
1995 size = blks_to_bytes(inode, map.m_len);
1996 flags = 0;
1997
1998 if (compr_cluster) {
1999 flags = FIEMAP_EXTENT_ENCODED;
2000 count_in_cluster += map.m_len;
2001 if (count_in_cluster == cluster_size) {
2002 compr_cluster = false;
2003 size += blks_to_bytes(inode, 1);
2004 }
2005 } else if (map.m_flags & F2FS_MAP_UNWRITTEN) {
2006 flags = FIEMAP_EXTENT_UNWRITTEN;
2007 }
7f63eb77 2008
093f0bac
DJ		 2009 start_blk += bytes_to_blks(inode, size);
2010 }
7f63eb77 2011
9a950d52 2012prep_next:
7f63eb77
JK		 2013 cond_resched();
2014 if (fatal_signal_pending(current))
2015 ret = -EINTR;
2016 else
2017 goto next;
2018out:
2019 if (ret == 1)
2020 ret = 0;
2021
5955102c 2022 inode_unlock(inode);
7f63eb77 2023 return ret;
9ab70134
JK		 2024}
2025
95ae251f
EB		 2026static inline loff_t f2fs_readpage_limit(struct inode *inode)
2027{
2028 if (IS_ENABLED(CONFIG_FS_VERITY) &&
2029 (IS_VERITY(inode) || f2fs_verity_in_progress(inode)))
2030 return inode->i_sb->s_maxbytes;
2031
2032 return i_size_read(inode);
2033}
2034
2df0ab04
CY		 2035static int f2fs_read_single_page(struct inode *inode, struct page *page,
2036 unsigned nr_pages,
2037 struct f2fs_map_blocks *map,
2038 struct bio **bio_ret,
2039 sector_t *last_block_in_bio,
2040 bool is_readahead)
2041{
2042 struct bio *bio = *bio_ret;
43b9d4b4 2043 const unsigned blocksize = blks_to_bytes(inode, 1);
2df0ab04
CY		 2044 sector_t block_in_file;
2045 sector_t last_block;
2046 sector_t last_block_in_file;
2047 sector_t block_nr;
2048 int ret = 0;
2049
4969c06a 2050 block_in_file = (sector_t)page_index(page);
2df0ab04 2051 last_block = block_in_file + nr_pages;
43b9d4b4
JK		 2052 last_block_in_file = bytes_to_blks(inode,
2053 f2fs_readpage_limit(inode) + blocksize - 1);
2df0ab04
CY		 2054 if (last_block > last_block_in_file)
2055 last_block = last_block_in_file;
2056
2057 /* just zeroing out page which is beyond EOF */
2058 if (block_in_file >= last_block)
2059 goto zero_out;
2060 /*
2061 * Map blocks using the previous result first.
2062 */
2063 if ((map->m_flags & F2FS_MAP_MAPPED) &&
2064 block_in_file > map->m_lblk &&
2065 block_in_file < (map->m_lblk + map->m_len))
2066 goto got_it;
2067
2068 /*
2069 * Then do more f2fs_map_blocks() calls until we are
2070 * done with this page.
2071 */
2072 map->m_lblk = block_in_file;
2073 map->m_len = last_block - block_in_file;
2074
2075 ret = f2fs_map_blocks(inode, map, 0, F2FS_GET_BLOCK_DEFAULT);
2076 if (ret)
2077 goto out;
2078got_it:
2079 if ((map->m_flags & F2FS_MAP_MAPPED)) {
2080 block_nr = map->m_pblk + block_in_file - map->m_lblk;
2081 SetPageMappedToDisk(page);
2082
4969c06a
JK		 2083 if (!PageUptodate(page) && (!PageSwapCache(page) &&
2084 !cleancache_get_page(page))) {
2df0ab04
CY		 2085 SetPageUptodate(page);
2086 goto confused;
2087 }
2088
2089 if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), block_nr,
93770ab7 2090 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 2091 ret = -EFSCORRUPTED;
2df0ab04
CY		 2092 goto out;
2093 }
2094 } else {
2095zero_out:
2096 zero_user_segment(page, 0, PAGE_SIZE);
95ae251f
EB		 2097 if (f2fs_need_verity(inode, page->index) &&
2098 !fsverity_verify_page(page)) {
2099 ret = -EIO;
2100 goto out;
2101 }
2df0ab04
CY		 2102 if (!PageUptodate(page))
2103 SetPageUptodate(page);
2104 unlock_page(page);
2105 goto out;
2106 }
2107
2108 /*
2109 * This page will go to BIO. Do we need to send this
2110 * BIO off first?
2111 */
27aacd28
ST		 2112 if (bio && (!page_is_mergeable(F2FS_I_SB(inode), bio,
2113 *last_block_in_bio, block_nr) ||
2114 !f2fs_crypt_mergeable_bio(bio, inode, page->index, NULL))) {
2df0ab04
CY		 2115submit_and_realloc:
2116 __submit_bio(F2FS_I_SB(inode), bio, DATA);
2117 bio = NULL;
2118 }
2119 if (bio == NULL) {
2120 bio = f2fs_grab_read_bio(inode, block_nr, nr_pages,
0683728a 2121 is_readahead ? REQ_RAHEAD : 0, page->index,
7f59b277 2122 false);
2df0ab04
CY		 2123 if (IS_ERR(bio)) {
2124 ret = PTR_ERR(bio);
2125 bio = NULL;
2126 goto out;
2127 }
2128 }
2129
2130 /*
2131 * If the page is under writeback, we need to wait for
2132 * its completion to see the correct decrypted data.
2133 */
2134 f2fs_wait_on_block_writeback(inode, block_nr);
2135
2136 if (bio_add_page(bio, page, blocksize, 0) < blocksize)
2137 goto submit_and_realloc;
2138
2139 inc_page_count(F2FS_I_SB(inode), F2FS_RD_DATA);
8b83ac81 2140 f2fs_update_iostat(F2FS_I_SB(inode), FS_DATA_READ_IO, F2FS_BLKSIZE);
2df0ab04
CY		 2141 ClearPageError(page);
2142 *last_block_in_bio = block_nr;
2143 goto out;
2144confused:
2145 if (bio) {
2146 __submit_bio(F2FS_I_SB(inode), bio, DATA);
2147 bio = NULL;
2148 }
2149 unlock_page(page);
2150out:
2151 *bio_ret = bio;
2152 return ret;
2153}
2154
4c8ff709
CY		 2155#ifdef CONFIG_F2FS_FS_COMPRESSION
2156int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret,
2157 unsigned nr_pages, sector_t *last_block_in_bio,
0683728a 2158 bool is_readahead, bool for_write)
4c8ff709
CY		 2159{
2160 struct dnode_of_data dn;
2161 struct inode *inode = cc->inode;
2162 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2163 struct bio *bio = *bio_ret;
2164 unsigned int start_idx = cc->cluster_idx << cc->log_cluster_size;
2165 sector_t last_block_in_file;
43b9d4b4 2166 const unsigned blocksize = blks_to_bytes(inode, 1);
4c8ff709 2167 struct decompress_io_ctx *dic = NULL;
94afd6d6
CY		 2168 struct extent_info ei = {0, };
2169 bool from_dnode = true;
4c8ff709
CY		 2170 int i;
2171 int ret = 0;
2172
2173 f2fs_bug_on(sbi, f2fs_cluster_is_empty(cc));
2174
43b9d4b4
JK		 2175 last_block_in_file = bytes_to_blks(inode,
2176 f2fs_readpage_limit(inode) + blocksize - 1);
4c8ff709
CY		 2177
2178 /* get rid of pages beyond EOF */
2179 for (i = 0; i < cc->cluster_size; i++) {
2180 struct page *page = cc->rpages[i];
2181
2182 if (!page)
2183 continue;
2184 if ((sector_t)page->index >= last_block_in_file) {
2185 zero_user_segment(page, 0, PAGE_SIZE);
2186 if (!PageUptodate(page))
2187 SetPageUptodate(page);
2188 } else if (!PageUptodate(page)) {
2189 continue;
2190 }
2191 unlock_page(page);
9605f75c
JK		 2192 if (for_write)
2193 put_page(page);
4c8ff709
CY		 2194 cc->rpages[i] = NULL;
2195 cc->nr_rpages--;
2196 }
2197
2198 /* we are done since all pages are beyond EOF */
2199 if (f2fs_cluster_is_empty(cc))
2200 goto out;
2201
94afd6d6
CY		 2202 if (f2fs_lookup_extent_cache(inode, start_idx, &ei))
2203 from_dnode = false;
2204
2205 if (!from_dnode)
2206 goto skip_reading_dnode;
2207
4c8ff709
CY		 2208 set_new_dnode(&dn, inode, NULL, NULL, 0);
2209 ret = f2fs_get_dnode_of_data(&dn, start_idx, LOOKUP_NODE);
2210 if (ret)
2211 goto out;
2212
a86d27dd 2213 f2fs_bug_on(sbi, dn.data_blkaddr != COMPRESS_ADDR);
4c8ff709 2214
94afd6d6 2215skip_reading_dnode:
4c8ff709
CY		 2216 for (i = 1; i < cc->cluster_size; i++) {
2217 block_t blkaddr;
2218
94afd6d6
CY		 2219 blkaddr = from_dnode ? data_blkaddr(dn.inode, dn.node_page,
2220 dn.ofs_in_node + i) :
2221 ei.blk + i - 1;
4c8ff709
CY		 2222
2223 if (!__is_valid_data_blkaddr(blkaddr))
2224 break;
2225
2226 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC)) {
2227 ret = -EFAULT;
2228 goto out_put_dnode;
2229 }
2230 cc->nr_cpages++;
94afd6d6
CY		 2231
2232 if (!from_dnode && i >= ei.c_len)
2233 break;
4c8ff709
CY		 2234 }
2235
2236 /* nothing to decompress */
2237 if (cc->nr_cpages == 0) {
2238 ret = 0;
2239 goto out_put_dnode;
2240 }
2241
2242 dic = f2fs_alloc_dic(cc);
2243 if (IS_ERR(dic)) {
2244 ret = PTR_ERR(dic);
2245 goto out_put_dnode;
2246 }
2247
6ce19aff 2248 for (i = 0; i < cc->nr_cpages; i++) {
4c8ff709
CY		 2249 struct page *page = dic->cpages[i];
2250 block_t blkaddr;
7f59b277 2251 struct bio_post_read_ctx *ctx;
4c8ff709 2252
94afd6d6
CY		 2253 blkaddr = from_dnode ? data_blkaddr(dn.inode, dn.node_page,
2254 dn.ofs_in_node + i + 1) :
2255 ei.blk + i;
4c8ff709 2256
6ce19aff
CY		 2257 f2fs_wait_on_block_writeback(inode, blkaddr);
2258
2259 if (f2fs_load_compressed_page(sbi, page, blkaddr)) {
2260 if (atomic_dec_and_test(&dic->remaining_pages))
2261 f2fs_decompress_cluster(dic);
2262 continue;
2263 }
2264
27aacd28
ST		 2265 if (bio && (!page_is_mergeable(sbi, bio,
2266 *last_block_in_bio, blkaddr) ||
2267 !f2fs_crypt_mergeable_bio(bio, inode, page->index, NULL))) {
4c8ff709
CY		 2268submit_and_realloc:
2269 __submit_bio(sbi, bio, DATA);
2270 bio = NULL;
2271 }
2272
2273 if (!bio) {
2274 bio = f2fs_grab_read_bio(inode, blkaddr, nr_pages,
2275 is_readahead ? REQ_RAHEAD : 0,
7f59b277 2276 page->index, for_write);
4c8ff709
CY		 2277 if (IS_ERR(bio)) {
2278 ret = PTR_ERR(bio);
7f59b277 2279 f2fs_decompress_end_io(dic, ret);
4c8ff709 2280 f2fs_put_dnode(&dn);
f3494345 2281 *bio_ret = NULL;
4c8ff709
CY		 2282 return ret;
2283 }
2284 }
2285
4c8ff709
CY		 2286 if (bio_add_page(bio, page, blocksize, 0) < blocksize)
2287 goto submit_and_realloc;
2288
a4b68176 2289 ctx = get_post_read_ctx(bio);
7f59b277
EB		 2290 ctx->enabled_steps |= STEP_DECOMPRESS;
2291 refcount_inc(&dic->refcnt);
03382f1a 2292
4c8ff709 2293 inc_page_count(sbi, F2FS_RD_DATA);
8b83ac81 2294 f2fs_update_iostat(sbi, FS_DATA_READ_IO, F2FS_BLKSIZE);
9c122384 2295 f2fs_update_iostat(sbi, FS_CDATA_READ_IO, F2FS_BLKSIZE);
4c8ff709
CY		 2296 ClearPageError(page);
2297 *last_block_in_bio = blkaddr;
2298 }
2299
94afd6d6
CY		 2300 if (from_dnode)
2301 f2fs_put_dnode(&dn);
4c8ff709
CY		 2302
2303 *bio_ret = bio;
2304 return 0;
2305
2306out_put_dnode:
94afd6d6
CY		 2307 if (from_dnode)
2308 f2fs_put_dnode(&dn);
4c8ff709 2309out:
7f59b277
EB		 2310 for (i = 0; i < cc->cluster_size; i++) {
2311 if (cc->rpages[i]) {
2312 ClearPageUptodate(cc->rpages[i]);
2313 ClearPageError(cc->rpages[i]);
2314 unlock_page(cc->rpages[i]);
2315 }
2316 }
4c8ff709
CY		 2317 *bio_ret = bio;
2318 return ret;
2319}
2320#endif
2321
f1e88660
JK		 2322/*
2323 * This function was originally taken from fs/mpage.c, and customized for f2fs.
2324 * Major change was from block_size == page_size in f2fs by default.
2325 */
e20a7693 2326static int f2fs_mpage_readpages(struct inode *inode,
23323196 2327 struct readahead_control *rac, struct page *page)
f1e88660
JK		 2328{
2329 struct bio *bio = NULL;
f1e88660 2330 sector_t last_block_in_bio = 0;
f1e88660 2331 struct f2fs_map_blocks map;
4c8ff709
CY		 2332#ifdef CONFIG_F2FS_FS_COMPRESSION
2333 struct compress_ctx cc = {
2334 .inode = inode,
2335 .log_cluster_size = F2FS_I(inode)->i_log_cluster_size,
2336 .cluster_size = F2FS_I(inode)->i_cluster_size,
2337 .cluster_idx = NULL_CLUSTER,
2338 .rpages = NULL,
2339 .cpages = NULL,
2340 .nr_rpages = 0,
2341 .nr_cpages = 0,
2342 };
a2649315 2343 pgoff_t nc_cluster_idx = NULL_CLUSTER;
4c8ff709 2344#endif
23323196 2345 unsigned nr_pages = rac ? readahead_count(rac) : 1;
4c8ff709 2346 unsigned max_nr_pages = nr_pages;
2df0ab04 2347 int ret = 0;
f1e88660
JK		 2348
2349 map.m_pblk = 0;
2350 map.m_lblk = 0;
2351 map.m_len = 0;
2352 map.m_flags = 0;
da85985c 2353 map.m_next_pgofs = NULL;
c4020b2d 2354 map.m_next_extent = NULL;
d5097be5 2355 map.m_seg_type = NO_CHECK_TYPE;
f9d6d059 2356 map.m_may_create = false;
f1e88660 2357
736c0a74 2358 for (; nr_pages; nr_pages--) {
23323196
MWO		 2359 if (rac) {
2360 page = readahead_page(rac);
a83d50bc 2361 prefetchw(&page->flags);
f1e88660
JK		 2362 }
2363
4c8ff709
CY		 2364#ifdef CONFIG_F2FS_FS_COMPRESSION
2365 if (f2fs_compressed_file(inode)) {
2366 /* there are remained comressed pages, submit them */
2367 if (!f2fs_cluster_can_merge_page(&cc, page->index)) {
2368 ret = f2fs_read_multi_pages(&cc, &bio,
2369 max_nr_pages,
2370 &last_block_in_bio,
23323196 2371 rac != NULL, false);
8bfbfb0d 2372 f2fs_destroy_compress_ctx(&cc, false);
4c8ff709
CY		 2373 if (ret)
2374 goto set_error_page;
2375 }
a2649315
FC		 2376 if (cc.cluster_idx == NULL_CLUSTER) {
2377 if (nc_cluster_idx ==
2378 page->index >> cc.log_cluster_size) {
2379 goto read_single_page;
2380 }
4c8ff709 2381
a2649315
FC		 2382 ret = f2fs_is_compressed_cluster(inode, page->index);
2383 if (ret < 0)
2384 goto set_error_page;
2385 else if (!ret) {
2386 nc_cluster_idx =
2387 page->index >> cc.log_cluster_size;
2388 goto read_single_page;
2389 }
4c8ff709 2390
a2649315
FC		 2391 nc_cluster_idx = NULL_CLUSTER;
2392 }
4c8ff709
CY		 2393 ret = f2fs_init_compress_ctx(&cc);
2394 if (ret)
2395 goto set_error_page;
2396
2397 f2fs_compress_ctx_add_page(&cc, page);
2398
2399 goto next_page;
2400 }
2401read_single_page:
2402#endif
2403
2404 ret = f2fs_read_single_page(inode, page, max_nr_pages, &map,
23323196 2405 &bio, &last_block_in_bio, rac);
2df0ab04 2406 if (ret) {
4c8ff709
CY		 2407#ifdef CONFIG_F2FS_FS_COMPRESSION
2408set_error_page:
2409#endif
2df0ab04 2410 SetPageError(page);
09cbfeaf 2411 zero_user_segment(page, 0, PAGE_SIZE);
f1e88660 2412 unlock_page(page);
f1e88660 2413 }
23323196 2414#ifdef CONFIG_F2FS_FS_COMPRESSION
f1e88660 2415next_page:
23323196
MWO		 2416#endif
2417 if (rac)
09cbfeaf 2418 put_page(page);
4c8ff709
CY		 2419
2420#ifdef CONFIG_F2FS_FS_COMPRESSION
2421 if (f2fs_compressed_file(inode)) {
2422 /* last page */
2423 if (nr_pages == 1 && !f2fs_cluster_is_empty(&cc)) {
2424 ret = f2fs_read_multi_pages(&cc, &bio,
2425 max_nr_pages,
2426 &last_block_in_bio,
23323196 2427 rac != NULL, false);
8bfbfb0d 2428 f2fs_destroy_compress_ctx(&cc, false);
4c8ff709
CY		 2429 }
2430 }
2431#endif
f1e88660 2432 }
f1e88660 2433 if (bio)
4fc29c1a 2434 __submit_bio(F2FS_I_SB(inode), bio, DATA);
23323196 2435 return ret;
f1e88660
JK		 2436}
2437
eb47b800
JK		 2438static int f2fs_read_data_page(struct file *file, struct page *page)
2439{
4969c06a 2440 struct inode *inode = page_file_mapping(page)->host;
b3d208f9 2441 int ret = -EAGAIN;
9ffe0fb5 2442
c20e89cd
CY		 2443 trace_f2fs_readpage(page, DATA);
2444
4c8ff709
CY		 2445 if (!f2fs_is_compress_backend_ready(inode)) {
2446 unlock_page(page);
2447 return -EOPNOTSUPP;
2448 }
2449
e1c42045 2450 /* If the file has inline data, try to read it directly */
9ffe0fb5
HL		 2451 if (f2fs_has_inline_data(inode))
2452 ret = f2fs_read_inline_data(inode, page);
b3d208f9 2453 if (ret == -EAGAIN)
e20a7693 2454 ret = f2fs_mpage_readpages(inode, NULL, page);
9ffe0fb5 2455 return ret;
eb47b800
JK		 2456}
2457
23323196 2458static void f2fs_readahead(struct readahead_control *rac)
eb47b800 2459{
23323196 2460 struct inode *inode = rac->mapping->host;
b8c29400 2461
23323196 2462 trace_f2fs_readpages(inode, readahead_index(rac), readahead_count(rac));
9ffe0fb5 2463
4c8ff709 2464 if (!f2fs_is_compress_backend_ready(inode))
23323196 2465 return;
4c8ff709 2466
9ffe0fb5
HL		 2467 /* If the file has inline data, skip readpages */
2468 if (f2fs_has_inline_data(inode))
23323196 2469 return;
9ffe0fb5 2470
e20a7693 2471 f2fs_mpage_readpages(inode, rac, NULL);
eb47b800
JK		 2472}
2473
4c8ff709 2474int f2fs_encrypt_one_page(struct f2fs_io_info *fio)
7eab0c0d
HP		 2475{
2476 struct inode *inode = fio->page->mapping->host;
4c8ff709 2477 struct page *mpage, *page;
7eab0c0d
HP		 2478 gfp_t gfp_flags = GFP_NOFS;
2479
1958593e 2480 if (!f2fs_encrypted_file(inode))
7eab0c0d
HP		 2481 return 0;
2482
4c8ff709
CY		 2483 page = fio->compressed_page ? fio->compressed_page : fio->page;
2484
6dbb1796 2485 /* wait for GCed page writeback via META_MAPPING */
0ded69f6 2486 f2fs_wait_on_block_writeback(inode, fio->old_blkaddr);
7eab0c0d 2487
27aacd28
ST		 2488 if (fscrypt_inode_uses_inline_crypto(inode))
2489 return 0;
2490
7eab0c0d 2491retry_encrypt:
4c8ff709
CY		 2492 fio->encrypted_page = fscrypt_encrypt_pagecache_blocks(page,
2493 PAGE_SIZE, 0, gfp_flags);
6aa58d8a
CY		 2494 if (IS_ERR(fio->encrypted_page)) {
2495 /* flush pending IOs and wait for a while in the ENOMEM case */
2496 if (PTR_ERR(fio->encrypted_page) == -ENOMEM) {
2497 f2fs_flush_merged_writes(fio->sbi);
5df7731f 2498 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
6aa58d8a
CY		 2499 gfp_flags |= __GFP_NOFAIL;
2500 goto retry_encrypt;
2501 }
2502 return PTR_ERR(fio->encrypted_page);
2503 }
7eab0c0d 2504
6aa58d8a
CY		 2505 mpage = find_lock_page(META_MAPPING(fio->sbi), fio->old_blkaddr);
2506 if (mpage) {
2507 if (PageUptodate(mpage))
2508 memcpy(page_address(mpage),
2509 page_address(fio->encrypted_page), PAGE_SIZE);
2510 f2fs_put_page(mpage, 1);
7eab0c0d 2511 }
6aa58d8a 2512 return 0;
7eab0c0d
HP		 2513}
2514
bb9e3bb8
CY		 2515static inline bool check_inplace_update_policy(struct inode *inode,
2516 struct f2fs_io_info *fio)
7eab0c0d 2517{
bb9e3bb8
CY		 2518 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2519 unsigned int policy = SM_I(sbi)->ipu_policy;
7eab0c0d 2520
bb9e3bb8
CY		 2521 if (policy & (0x1 << F2FS_IPU_FORCE))
2522 return true;
4d57b86d 2523 if (policy & (0x1 << F2FS_IPU_SSR) && f2fs_need_SSR(sbi))
bb9e3bb8
CY		 2524 return true;
2525 if (policy & (0x1 << F2FS_IPU_UTIL) &&
2526 utilization(sbi) > SM_I(sbi)->min_ipu_util)
2527 return true;
4d57b86d 2528 if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && f2fs_need_SSR(sbi) &&
bb9e3bb8
CY		 2529 utilization(sbi) > SM_I(sbi)->min_ipu_util)
2530 return true;
2531
2532 /*
2533 * IPU for rewrite async pages
2534 */
2535 if (policy & (0x1 << F2FS_IPU_ASYNC) &&
2536 fio && fio->op == REQ_OP_WRITE &&
2537 !(fio->op_flags & REQ_SYNC) &&
62230e0d 2538 !IS_ENCRYPTED(inode))
bb9e3bb8
CY		 2539 return true;
2540
2541 /* this is only set during fdatasync */
2542 if (policy & (0x1 << F2FS_IPU_FSYNC) &&
2543 is_inode_flag_set(inode, FI_NEED_IPU))
2544 return true;
2545
4354994f
DR		 2546 if (unlikely(fio && is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2547 !f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2548 return true;
2549
bb9e3bb8
CY		 2550 return false;
2551}
2552
4d57b86d 2553bool f2fs_should_update_inplace(struct inode *inode, struct f2fs_io_info *fio)
bb9e3bb8 2554{
859fca6b
CY		 2555 /* swap file is migrating in aligned write mode */
2556 if (is_inode_flag_set(inode, FI_ALIGNED_WRITE))
2557 return false;
2558
1ad71a27
JK		 2559 if (f2fs_is_pinned_file(inode))
2560 return true;
bb9e3bb8
CY		 2561
2562 /* if this is cold file, we should overwrite to avoid fragmentation */
2563 if (file_is_cold(inode))
2564 return true;
2565
2566 return check_inplace_update_policy(inode, fio);
2567}
2568
4d57b86d 2569bool f2fs_should_update_outplace(struct inode *inode, struct f2fs_io_info *fio)
bb9e3bb8
CY		 2570{
2571 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2572
b0332a0f 2573 if (f2fs_lfs_mode(sbi))
bb9e3bb8
CY		 2574 return true;
2575 if (S_ISDIR(inode->i_mode))
2576 return true;
af033b2a
CY		 2577 if (IS_NOQUOTA(inode))
2578 return true;
bb9e3bb8
CY		 2579 if (f2fs_is_atomic_file(inode))
2580 return true;
1ffc8f5f
JK		 2581 if (is_sbi_flag_set(sbi, SBI_NEED_FSCK))
2582 return true;
859fca6b
CY		 2583
2584 /* swap file is migrating in aligned write mode */
2585 if (is_inode_flag_set(inode, FI_ALIGNED_WRITE))
2586 return true;
2587
bb9e3bb8 2588 if (fio) {
b763f3be 2589 if (page_private_gcing(fio->page))
bb9e3bb8 2590 return true;
b763f3be 2591 if (page_private_dummy(fio->page))
bb9e3bb8 2592 return true;
4354994f
DR		 2593 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2594 f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
2595 return true;
bb9e3bb8
CY		 2596 }
2597 return false;
2598}
2599
7eab0c0d
HP		 2600static inline bool need_inplace_update(struct f2fs_io_info *fio)
2601{
2602 struct inode *inode = fio->page->mapping->host;
2603
4d57b86d 2604 if (f2fs_should_update_outplace(inode, fio))
7eab0c0d
HP		 2605 return false;
2606
4d57b86d 2607 return f2fs_should_update_inplace(inode, fio);
7eab0c0d
HP		 2608}
2609
4d57b86d 2610int f2fs_do_write_data_page(struct f2fs_io_info *fio)
eb47b800 2611{
05ca3632 2612 struct page *page = fio->page;
eb47b800 2613 struct inode *inode = page->mapping->host;
eb47b800 2614 struct dnode_of_data dn;
94afd6d6 2615 struct extent_info ei = {0, };
7735730d 2616 struct node_info ni;
e959c8f5 2617 bool ipu_force = false;
eb47b800
JK		 2618 int err = 0;
2619
2620 set_new_dnode(&dn, inode, NULL, NULL, 0);
e959c8f5
HP		 2621 if (need_inplace_update(fio) &&
2622 f2fs_lookup_extent_cache(inode, page->index, &ei)) {
2623 fio->old_blkaddr = ei.blk + page->index - ei.fofs;
a817737e 2624
c9b60788 2625 if (!f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
93770ab7 2626 DATA_GENERIC_ENHANCE))
10f966bb 2627 return -EFSCORRUPTED;
c9b60788
CY		 2628
2629 ipu_force = true;
2630 fio->need_lock = LOCK_DONE;
2631 goto got_it;
e959c8f5 2632 }
279d6df2 2633
d29460e5
JK		 2634 /* Deadlock due to between page->lock and f2fs_lock_op */
2635 if (fio->need_lock == LOCK_REQ && !f2fs_trylock_op(fio->sbi))
2636 return -EAGAIN;
279d6df2 2637
4d57b86d 2638 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
eb47b800 2639 if (err)
279d6df2 2640 goto out;
eb47b800 2641
28bc106b 2642 fio->old_blkaddr = dn.data_blkaddr;
eb47b800
JK		 2643
2644 /* This page is already truncated */
7a9d7548 2645 if (fio->old_blkaddr == NULL_ADDR) {
2bca1e23 2646 ClearPageUptodate(page);
b763f3be 2647 clear_page_private_gcing(page);
eb47b800 2648 goto out_writepage;
2bca1e23 2649 }
e959c8f5 2650got_it:
c9b60788
CY		 2651 if (__is_valid_data_blkaddr(fio->old_blkaddr) &&
2652 !f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
93770ab7 2653 DATA_GENERIC_ENHANCE)) {
10f966bb 2654 err = -EFSCORRUPTED;
c9b60788
CY		 2655 goto out_writepage;
2656 }
eb47b800
JK		 2657 /*
2658 * If current allocation needs SSR,
2659 * it had better in-place writes for updated data.
2660 */
93770ab7
CY		 2661 if (ipu_force ||
2662 (__is_valid_data_blkaddr(fio->old_blkaddr) &&
7b525dd0 2663 need_inplace_update(fio))) {
4c8ff709 2664 err = f2fs_encrypt_one_page(fio);
cc15620b
JK		 2665 if (err)
2666 goto out_writepage;
2667
2668 set_page_writeback(page);
17c50035 2669 ClearPageError(page);
279d6df2 2670 f2fs_put_dnode(&dn);
cc15620b 2671 if (fio->need_lock == LOCK_REQ)
279d6df2 2672 f2fs_unlock_op(fio->sbi);
4d57b86d 2673 err = f2fs_inplace_write_data(fio);
6492a335 2674 if (err) {
27aacd28 2675 if (fscrypt_inode_uses_fs_layer_crypto(inode))
d2d0727b 2676 fscrypt_finalize_bounce_page(&fio->encrypted_page);
6492a335
CY		 2677 if (PageWriteback(page))
2678 end_page_writeback(page);
cd23ffa9
CY		 2679 } else {
2680 set_inode_flag(inode, FI_UPDATE_WRITE);
6492a335 2681 }
7eab0c0d 2682 trace_f2fs_do_write_data_page(fio->page, IPU);
279d6df2 2683 return err;
eb47b800 2684 }
279d6df2 2685
cc15620b
JK		 2686 if (fio->need_lock == LOCK_RETRY) {
2687 if (!f2fs_trylock_op(fio->sbi)) {
2688 err = -EAGAIN;
2689 goto out_writepage;
2690 }
2691 fio->need_lock = LOCK_REQ;
2692 }
2693
7735730d
CY		 2694 err = f2fs_get_node_info(fio->sbi, dn.nid, &ni);
2695 if (err)
2696 goto out_writepage;
2697
2698 fio->version = ni.version;
2699
4c8ff709 2700 err = f2fs_encrypt_one_page(fio);
cc15620b
JK		 2701 if (err)
2702 goto out_writepage;
2703
2704 set_page_writeback(page);
17c50035 2705 ClearPageError(page);
cc15620b 2706
4c8ff709
CY		 2707 if (fio->compr_blocks && fio->old_blkaddr == COMPRESS_ADDR)
2708 f2fs_i_compr_blocks_update(inode, fio->compr_blocks - 1, false);
2709
279d6df2 2710 /* LFS mode write path */
4d57b86d 2711 f2fs_outplace_write_data(&dn, fio);
279d6df2
HP		 2712 trace_f2fs_do_write_data_page(page, OPU);
2713 set_inode_flag(inode, FI_APPEND_WRITE);
2714 if (page->index == 0)
2715 set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
eb47b800
JK		 2716out_writepage:
2717 f2fs_put_dnode(&dn);
279d6df2 2718out:
cc15620b 2719 if (fio->need_lock == LOCK_REQ)
279d6df2 2720 f2fs_unlock_op(fio->sbi);
eb47b800
JK		 2721 return err;
2722}
2723
4c8ff709 2724int f2fs_write_single_data_page(struct page *page, int *submitted,
8648de2c
CY		 2725 struct bio **bio,
2726 sector_t *last_block,
b0af6d49 2727 struct writeback_control *wbc,
4c8ff709 2728 enum iostat_type io_type,
3afae09f
CY		 2729 int compr_blocks,
2730 bool allow_balance)
eb47b800
JK		 2731{
2732 struct inode *inode = page->mapping->host;
4081363f 2733 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
eb47b800 2734 loff_t i_size = i_size_read(inode);
4c8ff709 2735 const pgoff_t end_index = ((unsigned long long)i_size)
09cbfeaf 2736 >> PAGE_SHIFT;
1f0d5c91 2737 loff_t psize = (loff_t)(page->index + 1) << PAGE_SHIFT;
9ffe0fb5 2738 unsigned offset = 0;
39936837 2739 bool need_balance_fs = false;
eb47b800 2740 int err = 0;
458e6197 2741 struct f2fs_io_info fio = {
05ca3632 2742 .sbi = sbi,
39d787be 2743 .ino = inode->i_ino,
458e6197 2744 .type = DATA,
04d328de 2745 .op = REQ_OP_WRITE,
7637241e 2746 .op_flags = wbc_to_write_flags(wbc),
e959c8f5 2747 .old_blkaddr = NULL_ADDR,
05ca3632 2748 .page = page,
4375a336 2749 .encrypted_page = NULL,
d68f735b 2750 .submitted = false,
4c8ff709 2751 .compr_blocks = compr_blocks,
cc15620b 2752 .need_lock = LOCK_RETRY,
b0af6d49 2753 .io_type = io_type,
578c6478 2754 .io_wbc = wbc,
8648de2c
CY		 2755 .bio = bio,
2756 .last_block = last_block,
458e6197 2757 };
eb47b800 2758
ecda0de3
CY		 2759 trace_f2fs_writepage(page, DATA);
2760
db198ae0
CY		 2761 /* we should bypass data pages to proceed the kworkder jobs */
2762 if (unlikely(f2fs_cp_error(sbi))) {
2763 mapping_set_error(page->mapping, -EIO);
1174abfd
CY		 2764 /*
2765 * don't drop any dirty dentry pages for keeping lastest
2766 * directory structure.
2767 */
2768 if (S_ISDIR(inode->i_mode))
2769 goto redirty_out;
db198ae0
CY		 2770 goto out;
2771 }
2772
0771fcc7
CY		 2773 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
2774 goto redirty_out;
2775
4c8ff709
CY		 2776 if (page->index < end_index ||
2777 f2fs_verity_in_progress(inode) ||
2778 compr_blocks)
39936837 2779 goto write;
eb47b800
JK		 2780
2781 /*
2782 * If the offset is out-of-range of file size,
2783 * this page does not have to be written to disk.
2784 */
09cbfeaf 2785 offset = i_size & (PAGE_SIZE - 1);
76f60268 2786 if ((page->index >= end_index + 1) || !offset)
39936837 2787 goto out;
eb47b800 2788
09cbfeaf 2789 zero_user_segment(page, offset, PAGE_SIZE);
39936837 2790write:
1e84371f
JK		 2791 if (f2fs_is_drop_cache(inode))
2792 goto out;
e6e5f561
JK		 2793 /* we should not write 0'th page having journal header */
2794 if (f2fs_is_volatile_file(inode) && (!page->index ||
2795 (!wbc->for_reclaim &&
4d57b86d 2796 f2fs_available_free_memory(sbi, BASE_CHECK))))
1e84371f 2797 goto redirty_out;
eb47b800 2798
435cbab9
JK		 2799 /* Dentry/quota blocks are controlled by checkpoint */
2800 if (S_ISDIR(inode->i_mode) || IS_NOQUOTA(inode)) {
79963d96
CY		 2801 /*
2802 * We need to wait for node_write to avoid block allocation during
2803 * checkpoint. This can only happen to quota writes which can cause
2804 * the below discard race condition.
2805 */
2806 if (IS_NOQUOTA(inode))
2807 down_read(&sbi->node_write);
2808
cc15620b 2809 fio.need_lock = LOCK_DONE;
4d57b86d 2810 err = f2fs_do_write_data_page(&fio);
79963d96
CY		 2811
2812 if (IS_NOQUOTA(inode))
2813 up_read(&sbi->node_write);
2814
8618b881
JK		 2815 goto done;
2816 }
9ffe0fb5 2817
8618b881 2818 if (!wbc->for_reclaim)
39936837 2819 need_balance_fs = true;
7f3037a5 2820 else if (has_not_enough_free_secs(sbi, 0, 0))
39936837 2821 goto redirty_out;
ef095d19
JK		 2822 else
2823 set_inode_flag(inode, FI_HOT_DATA);
eb47b800 2824
b3d208f9 2825 err = -EAGAIN;
dd7b2333 2826 if (f2fs_has_inline_data(inode)) {
b3d208f9 2827 err = f2fs_write_inline_data(inode, page);
dd7b2333
YH		 2828 if (!err)
2829 goto out;
2830 }
279d6df2 2831
cc15620b 2832 if (err == -EAGAIN) {
4d57b86d 2833 err = f2fs_do_write_data_page(&fio);
cc15620b
JK		 2834 if (err == -EAGAIN) {
2835 fio.need_lock = LOCK_REQ;
4d57b86d 2836 err = f2fs_do_write_data_page(&fio);
cc15620b
JK		 2837 }
2838 }
a0d00fad 2839
eb449797
CY		 2840 if (err) {
2841 file_set_keep_isize(inode);
2842 } else {
c10c9820 2843 spin_lock(&F2FS_I(inode)->i_size_lock);
eb449797
CY		 2844 if (F2FS_I(inode)->last_disk_size < psize)
2845 F2FS_I(inode)->last_disk_size = psize;
c10c9820 2846 spin_unlock(&F2FS_I(inode)->i_size_lock);
eb449797 2847 }
279d6df2 2848
8618b881
JK		 2849done:
2850 if (err && err != -ENOENT)
2851 goto redirty_out;
eb47b800 2852
39936837 2853out:
a7ffdbe2 2854 inode_dec_dirty_pages(inode);
2baf0781 2855 if (err) {
2bca1e23 2856 ClearPageUptodate(page);
b763f3be 2857 clear_page_private_gcing(page);
2baf0781 2858 }
0c3a5797
CY		 2859
2860 if (wbc->for_reclaim) {
bab475c5 2861 f2fs_submit_merged_write_cond(sbi, NULL, page, 0, DATA);
ef095d19 2862 clear_inode_flag(inode, FI_HOT_DATA);
4d57b86d 2863 f2fs_remove_dirty_inode(inode);
d68f735b 2864 submitted = NULL;
0c3a5797 2865 }
eb47b800 2866 unlock_page(page);
186857c5 2867 if (!S_ISDIR(inode->i_mode) && !IS_NOQUOTA(inode) &&
3afae09f 2868 !F2FS_I(inode)->cp_task && allow_balance)
a7881893 2869 f2fs_balance_fs(sbi, need_balance_fs);
0c3a5797 2870
d68f735b 2871 if (unlikely(f2fs_cp_error(sbi))) {
b9109b0e 2872 f2fs_submit_merged_write(sbi, DATA);
0b20fcec 2873 f2fs_submit_merged_ipu_write(sbi, bio, NULL);
d68f735b
JK		 2874 submitted = NULL;
2875 }
2876
2877 if (submitted)
4c8ff709 2878 *submitted = fio.submitted ? 1 : 0;
0c3a5797 2879
eb47b800
JK		 2880 return 0;
2881
eb47b800 2882redirty_out:
76f60268 2883 redirty_page_for_writepage(wbc, page);
5b19d284
JK		 2884 /*
2885 * pageout() in MM traslates EAGAIN, so calls handle_write_error()
2886 * -> mapping_set_error() -> set_bit(AS_EIO, ...).
2887 * file_write_and_wait_range() will see EIO error, which is critical
2888 * to return value of fsync() followed by atomic_write failure to user.
2889 */
2890 if (!err || wbc->for_reclaim)
0002b61b 2891 return AOP_WRITEPAGE_ACTIVATE;
b230e6ca
JK		 2892 unlock_page(page);
2893 return err;
fa9150a8
NJ		 2894}
2895
f566bae8
JK		 2896static int f2fs_write_data_page(struct page *page,
2897 struct writeback_control *wbc)
2898{
4c8ff709
CY		 2899#ifdef CONFIG_F2FS_FS_COMPRESSION
2900 struct inode *inode = page->mapping->host;
2901
2902 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2903 goto out;
2904
2905 if (f2fs_compressed_file(inode)) {
2906 if (f2fs_is_compressed_cluster(inode, page->index)) {
2907 redirty_page_for_writepage(wbc, page);
2908 return AOP_WRITEPAGE_ACTIVATE;
2909 }
2910 }
2911out:
2912#endif
2913
2914 return f2fs_write_single_data_page(page, NULL, NULL, NULL,
3afae09f 2915 wbc, FS_DATA_IO, 0, true);
f566bae8
JK		 2916}
2917
8f46dcae
CY		 2918/*
2919 * This function was copied from write_cche_pages from mm/page-writeback.c.
2920 * The major change is making write step of cold data page separately from
2921 * warm/hot data page.
2922 */
2923static int f2fs_write_cache_pages(struct address_space *mapping,
b0af6d49
CY		 2924 struct writeback_control *wbc,
2925 enum iostat_type io_type)
8f46dcae
CY		 2926{
2927 int ret = 0;
4c8ff709 2928 int done = 0, retry = 0;
8f46dcae 2929 struct pagevec pvec;
c29fd0c0 2930 struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
8648de2c
CY		 2931 struct bio *bio = NULL;
2932 sector_t last_block;
4c8ff709
CY		 2933#ifdef CONFIG_F2FS_FS_COMPRESSION
2934 struct inode *inode = mapping->host;
2935 struct compress_ctx cc = {
2936 .inode = inode,
2937 .log_cluster_size = F2FS_I(inode)->i_log_cluster_size,
2938 .cluster_size = F2FS_I(inode)->i_cluster_size,
2939 .cluster_idx = NULL_CLUSTER,
2940 .rpages = NULL,
2941 .nr_rpages = 0,
2942 .cpages = NULL,
3271d7eb 2943 .valid_nr_cpages = 0,
4c8ff709
CY		 2944 .rbuf = NULL,
2945 .cbuf = NULL,
2946 .rlen = PAGE_SIZE * F2FS_I(inode)->i_cluster_size,
2947 .private = NULL,
2948 };
2949#endif
8f46dcae 2950 int nr_pages;
8f46dcae
CY		 2951 pgoff_t index;
2952 pgoff_t end; /* Inclusive */
2953 pgoff_t done_index;
8f46dcae 2954 int range_whole = 0;
10bbd235 2955 xa_mark_t tag;
bab475c5 2956 int nwritten = 0;
4c8ff709
CY		 2957 int submitted = 0;
2958 int i;
8f46dcae 2959
86679820 2960 pagevec_init(&pvec);
46ae957f 2961
ef095d19
JK		 2962 if (get_dirty_pages(mapping->host) <=
2963 SM_I(F2FS_M_SB(mapping))->min_hot_blocks)
2964 set_inode_flag(mapping->host, FI_HOT_DATA);
2965 else
2966 clear_inode_flag(mapping->host, FI_HOT_DATA);
2967
8f46dcae 2968 if (wbc->range_cyclic) {
4df7a75f 2969 index = mapping->writeback_index; /* prev offset */
8f46dcae
CY		 2970 end = -1;
2971 } else {
09cbfeaf
KS		 2972 index = wbc->range_start >> PAGE_SHIFT;
2973 end = wbc->range_end >> PAGE_SHIFT;
8f46dcae
CY		 2974 if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
2975 range_whole = 1;
8f46dcae
CY		 2976 }
2977 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2978 tag = PAGECACHE_TAG_TOWRITE;
2979 else
2980 tag = PAGECACHE_TAG_DIRTY;
2981retry:
4c8ff709 2982 retry = 0;
8f46dcae
CY		 2983 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
2984 tag_pages_for_writeback(mapping, index, end);
2985 done_index = index;
4c8ff709 2986 while (!done && !retry && (index <= end)) {
69c4f35d 2987 nr_pages = pagevec_lookup_range_tag(&pvec, mapping, &index, end,
67fd707f 2988 tag);
8f46dcae
CY		 2989 if (nr_pages == 0)
2990 break;
2991
2992 for (i = 0; i < nr_pages; i++) {
2993 struct page *page = pvec.pages[i];
4c8ff709
CY		 2994 bool need_readd;
2995readd:
2996 need_readd = false;
2997#ifdef CONFIG_F2FS_FS_COMPRESSION
2998 if (f2fs_compressed_file(inode)) {
b368cc5e
FC		 2999 void *fsdata = NULL;
3000 struct page *pagep;
3001 int ret2;
3002
4c8ff709
CY		 3003 ret = f2fs_init_compress_ctx(&cc);
3004 if (ret) {
3005 done = 1;
3006 break;
3007 }
3008
3009 if (!f2fs_cluster_can_merge_page(&cc,
3010 page->index)) {
3011 ret = f2fs_write_multi_pages(&cc,
3012 &submitted, wbc, io_type);
3013 if (!ret)
3014 need_readd = true;
3015 goto result;
3016 }
8f46dcae 3017
4c8ff709
CY		 3018 if (unlikely(f2fs_cp_error(sbi)))
3019 goto lock_page;
3020
b368cc5e
FC		 3021 if (!f2fs_cluster_is_empty(&cc))
3022 goto lock_page;
4c8ff709 3023
b368cc5e 3024 ret2 = f2fs_prepare_compress_overwrite(
4c8ff709
CY		 3025 inode, &pagep,
3026 page->index, &fsdata);
b368cc5e
FC		 3027 if (ret2 < 0) {
3028 ret = ret2;
3029 done = 1;
3030 break;
3031 } else if (ret2 &&
3032 (!f2fs_compress_write_end(inode,
3033 fsdata, page->index, 1) ||
3034 !f2fs_all_cluster_page_loaded(&cc,
3035 &pvec, i, nr_pages))) {
3036 retry = 1;
3037 break;
4c8ff709
CY		 3038 }
3039 }
3040#endif
f8de4331 3041 /* give a priority to WB_SYNC threads */
c29fd0c0 3042 if (atomic_read(&sbi->wb_sync_req[DATA]) &&
f8de4331
CY		 3043 wbc->sync_mode == WB_SYNC_NONE) {
3044 done = 1;
3045 break;
3046 }
4c8ff709
CY		 3047#ifdef CONFIG_F2FS_FS_COMPRESSION
3048lock_page:
3049#endif
8f46dcae 3050 done_index = page->index;
d29460e5 3051retry_write:
8f46dcae
CY		 3052 lock_page(page);
3053
3054 if (unlikely(page->mapping != mapping)) {
3055continue_unlock:
3056 unlock_page(page);
3057 continue;
3058 }
3059
3060 if (!PageDirty(page)) {
3061 /* someone wrote it for us */
3062 goto continue_unlock;
3063 }
3064
8f46dcae 3065 if (PageWriteback(page)) {
0b20fcec 3066 if (wbc->sync_mode != WB_SYNC_NONE)
fec1d657 3067 f2fs_wait_on_page_writeback(page,
bae0ee7a 3068 DATA, true, true);
0b20fcec 3069 else
8f46dcae
CY		 3070 goto continue_unlock;
3071 }
3072
8f46dcae
CY		 3073 if (!clear_page_dirty_for_io(page))
3074 goto continue_unlock;
3075
4c8ff709
CY		 3076#ifdef CONFIG_F2FS_FS_COMPRESSION
3077 if (f2fs_compressed_file(inode)) {
3078 get_page(page);
3079 f2fs_compress_ctx_add_page(&cc, page);
3080 continue;
3081 }
3082#endif
3083 ret = f2fs_write_single_data_page(page, &submitted,
3afae09f
CY		 3084 &bio, &last_block, wbc, io_type,
3085 0, true);
4c8ff709
CY		 3086 if (ret == AOP_WRITEPAGE_ACTIVATE)
3087 unlock_page(page);
3088#ifdef CONFIG_F2FS_FS_COMPRESSION
3089result:
3090#endif
3091 nwritten += submitted;
3092 wbc->nr_to_write -= submitted;
3093
8f46dcae 3094 if (unlikely(ret)) {
0002b61b
CY		 3095 /*
3096 * keep nr_to_write, since vfs uses this to
3097 * get # of written pages.
3098 */
3099 if (ret == AOP_WRITEPAGE_ACTIVATE) {
0002b61b 3100 ret = 0;
4c8ff709 3101 goto next;
d29460e5
JK		 3102 } else if (ret == -EAGAIN) {
3103 ret = 0;
3104 if (wbc->sync_mode == WB_SYNC_ALL) {
3105 cond_resched();
3106 congestion_wait(BLK_RW_ASYNC,
5df7731f 3107 DEFAULT_IO_TIMEOUT);
d29460e5
JK		 3108 goto retry_write;
3109 }
4c8ff709 3110 goto next;
0002b61b 3111 }
b230e6ca
JK		 3112 done_index = page->index + 1;
3113 done = 1;
3114 break;
8f46dcae
CY		 3115 }
3116
4c8ff709 3117 if (wbc->nr_to_write <= 0 &&
687de7f1 3118 wbc->sync_mode == WB_SYNC_NONE) {
8f46dcae
CY		 3119 done = 1;
3120 break;
3121 }
4c8ff709
CY		 3122next:
3123 if (need_readd)
3124 goto readd;
8f46dcae
CY		 3125 }
3126 pagevec_release(&pvec);
3127 cond_resched();
3128 }
4c8ff709
CY		 3129#ifdef CONFIG_F2FS_FS_COMPRESSION
3130 /* flush remained pages in compress cluster */
3131 if (f2fs_compressed_file(inode) && !f2fs_cluster_is_empty(&cc)) {
3132 ret = f2fs_write_multi_pages(&cc, &submitted, wbc, io_type);
3133 nwritten += submitted;
3134 wbc->nr_to_write -= submitted;
3135 if (ret) {
3136 done = 1;
3137 retry = 0;
3138 }
3139 }
adfc6943 3140 if (f2fs_compressed_file(inode))
8bfbfb0d 3141 f2fs_destroy_compress_ctx(&cc, false);
4c8ff709 3142#endif
e78790f8 3143 if (retry) {
8f46dcae 3144 index = 0;
e78790f8 3145 end = -1;
8f46dcae
CY		 3146 goto retry;
3147 }
e78790f8
ST		 3148 if (wbc->range_cyclic && !done)
3149 done_index = 0;
8f46dcae
CY		 3150 if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
3151 mapping->writeback_index = done_index;
3152
bab475c5 3153 if (nwritten)
b9109b0e 3154 f2fs_submit_merged_write_cond(F2FS_M_SB(mapping), mapping->host,
bab475c5 3155 NULL, 0, DATA);
8648de2c
CY		 3156 /* submit cached bio of IPU write */
3157 if (bio)
0b20fcec 3158 f2fs_submit_merged_ipu_write(sbi, &bio, NULL);
6ca56ca4 3159
8f46dcae
CY		 3160 return ret;
3161}
3162
853137ce
JK		 3163static inline bool __should_serialize_io(struct inode *inode,
3164 struct writeback_control *wbc)
3165{
b13f67ff
CY		 3166 /* to avoid deadlock in path of data flush */
3167 if (F2FS_I(inode)->cp_task)
3168 return false;
3169
853137ce
JK		 3170 if (!S_ISREG(inode->i_mode))
3171 return false;
af033b2a
CY		 3172 if (IS_NOQUOTA(inode))
3173 return false;
b13f67ff 3174
602a16d5 3175 if (f2fs_need_compress_data(inode))
b13f67ff 3176 return true;
853137ce
JK		 3177 if (wbc->sync_mode != WB_SYNC_ALL)
3178 return true;
3179 if (get_dirty_pages(inode) >= SM_I(F2FS_I_SB(inode))->min_seq_blocks)
3180 return true;
3181 return false;
3182}
3183
fc99fe27 3184static int __f2fs_write_data_pages(struct address_space *mapping,
b0af6d49
CY		 3185 struct writeback_control *wbc,
3186 enum iostat_type io_type)
eb47b800
JK		 3187{
3188 struct inode *inode = mapping->host;
4081363f 3189 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
9dfa1baf 3190 struct blk_plug plug;
eb47b800 3191 int ret;
853137ce 3192 bool locked = false;
eb47b800 3193
cfb185a1 3194 /* deal with chardevs and other special file */
3195 if (!mapping->a_ops->writepage)
3196 return 0;
3197
6a290544
CY		 3198 /* skip writing if there is no dirty page in this inode */
3199 if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE)
3200 return 0;
3201
0771fcc7
CY		 3202 /* during POR, we don't need to trigger writepage at all. */
3203 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
3204 goto skip_write;
3205
af033b2a
CY		 3206 if ((S_ISDIR(inode->i_mode) || IS_NOQUOTA(inode)) &&
3207 wbc->sync_mode == WB_SYNC_NONE &&
a1257023 3208 get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) &&
4d57b86d 3209 f2fs_available_free_memory(sbi, DIRTY_DENTS))
a1257023
JK		 3210 goto skip_write;
3211
d323d005 3212 /* skip writing during file defragment */
91942321 3213 if (is_inode_flag_set(inode, FI_DO_DEFRAG))
d323d005
CY		 3214 goto skip_write;
3215
d31c7c3f
YH		 3216 trace_f2fs_writepages(mapping->host, wbc, DATA);
3217
687de7f1
JK		 3218 /* to avoid spliting IOs due to mixed WB_SYNC_ALL and WB_SYNC_NONE */
3219 if (wbc->sync_mode == WB_SYNC_ALL)
c29fd0c0
CY		 3220 atomic_inc(&sbi->wb_sync_req[DATA]);
3221 else if (atomic_read(&sbi->wb_sync_req[DATA]))
687de7f1
JK		 3222 goto skip_write;
3223
853137ce
JK		 3224 if (__should_serialize_io(inode, wbc)) {
3225 mutex_lock(&sbi->writepages);
3226 locked = true;
3227 }
3228
9dfa1baf 3229 blk_start_plug(&plug);
b0af6d49 3230 ret = f2fs_write_cache_pages(mapping, wbc, io_type);
9dfa1baf 3231 blk_finish_plug(&plug);
687de7f1 3232
853137ce
JK		 3233 if (locked)
3234 mutex_unlock(&sbi->writepages);
3235
687de7f1 3236 if (wbc->sync_mode == WB_SYNC_ALL)
c29fd0c0 3237 atomic_dec(&sbi->wb_sync_req[DATA]);
28ea6162
JK		 3238 /*
3239 * if some pages were truncated, we cannot guarantee its mapping->host
3240 * to detect pending bios.
3241 */
458e6197 3242
4d57b86d 3243 f2fs_remove_dirty_inode(inode);
eb47b800 3244 return ret;
d3baf95d
JK		 3245
3246skip_write:
a7ffdbe2 3247 wbc->pages_skipped += get_dirty_pages(inode);
d31c7c3f 3248 trace_f2fs_writepages(mapping->host, wbc, DATA);
d3baf95d 3249 return 0;
eb47b800
JK		 3250}
3251
b0af6d49
CY		 3252static int f2fs_write_data_pages(struct address_space *mapping,
3253 struct writeback_control *wbc)
3254{
3255 struct inode *inode = mapping->host;
3256
3257 return __f2fs_write_data_pages(mapping, wbc,
3258 F2FS_I(inode)->cp_task == current ?
3259 FS_CP_DATA_IO : FS_DATA_IO);
3260}
3261
3e679dc7 3262static void f2fs_write_failed(struct inode *inode, loff_t to)
3aab8f82 3263{
819d9153 3264 loff_t i_size = i_size_read(inode);
3aab8f82 3265
3f188c23
JK		 3266 if (IS_NOQUOTA(inode))
3267 return;
3268
95ae251f
EB		 3269 /* In the fs-verity case, f2fs_end_enable_verity() does the truncate */
3270 if (to > i_size && !f2fs_verity_in_progress(inode)) {
a33c1502 3271 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
6abaa83c 3272 filemap_invalidate_lock(inode->i_mapping);
a33c1502 3273
819d9153 3274 truncate_pagecache(inode, i_size);
3f188c23 3275 f2fs_truncate_blocks(inode, i_size, true);
a33c1502 3276
6abaa83c 3277 filemap_invalidate_unlock(inode->i_mapping);
6f8d4455 3278 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3aab8f82
CY		 3279 }
3280}
3281
2aadac08
JK		 3282static int prepare_write_begin(struct f2fs_sb_info *sbi,
3283 struct page *page, loff_t pos, unsigned len,
3284 block_t *blk_addr, bool *node_changed)
3285{
3286 struct inode *inode = page->mapping->host;
3287 pgoff_t index = page->index;
3288 struct dnode_of_data dn;
3289 struct page *ipage;
b4d07a3e 3290 bool locked = false;
94afd6d6 3291 struct extent_info ei = {0, };
2aadac08 3292 int err = 0;
2866fb16 3293 int flag;
2aadac08 3294
24b84912 3295 /*
3d697a4a
EB		 3296 * If a whole page is being written and we already preallocated all the
3297 * blocks, then there is no need to get a block address now.
24b84912 3298 */
3d697a4a 3299 if (len == PAGE_SIZE && is_inode_flag_set(inode, FI_PREALLOCATED_ALL))
24b84912
JK		 3300 return 0;
3301
2866fb16
SY		 3302 /* f2fs_lock_op avoids race between write CP and convert_inline_page */
3303 if (f2fs_has_inline_data(inode) && pos + len > MAX_INLINE_DATA(inode))
3304 flag = F2FS_GET_BLOCK_DEFAULT;
3305 else
3306 flag = F2FS_GET_BLOCK_PRE_AIO;
3307
b4d07a3e 3308 if (f2fs_has_inline_data(inode) ||
09cbfeaf 3309 (pos & PAGE_MASK) >= i_size_read(inode)) {
0ef81833 3310 f2fs_do_map_lock(sbi, flag, true);
b4d07a3e
JK		 3311 locked = true;
3312 }
4c8ff709 3313
b4d07a3e 3314restart:
2aadac08 3315 /* check inline_data */
4d57b86d 3316 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2aadac08
JK		 3317 if (IS_ERR(ipage)) {
3318 err = PTR_ERR(ipage);
3319 goto unlock_out;
3320 }
3321
3322 set_new_dnode(&dn, inode, ipage, ipage, 0);
3323
3324 if (f2fs_has_inline_data(inode)) {
f2470371 3325 if (pos + len <= MAX_INLINE_DATA(inode)) {
4d57b86d 3326 f2fs_do_read_inline_data(page, ipage);
91942321 3327 set_inode_flag(inode, FI_DATA_EXIST);
ab47036d 3328 if (inode->i_nlink)
b763f3be 3329 set_page_private_inline(ipage);
2aadac08
JK		 3330 } else {
3331 err = f2fs_convert_inline_page(&dn, page);
3332 if (err)
b4d07a3e
JK		 3333 goto out;
3334 if (dn.data_blkaddr == NULL_ADDR)
3335 err = f2fs_get_block(&dn, index);
3336 }
3337 } else if (locked) {
3338 err = f2fs_get_block(&dn, index);
3339 } else {
3340 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
3341 dn.data_blkaddr = ei.blk + index - ei.fofs;
3342 } else {
b4d07a3e 3343 /* hole case */
4d57b86d 3344 err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
4da7bf5a 3345 if (err || dn.data_blkaddr == NULL_ADDR) {
b4d07a3e 3346 f2fs_put_dnode(&dn);
0ef81833 3347 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO,
59c9081b 3348 true);
2866fb16 3349 WARN_ON(flag != F2FS_GET_BLOCK_PRE_AIO);
b4d07a3e
JK		 3350 locked = true;
3351 goto restart;
3352 }
2aadac08
JK		 3353 }
3354 }
b4d07a3e 3355
2aadac08
JK		 3356 /* convert_inline_page can make node_changed */
3357 *blk_addr = dn.data_blkaddr;
3358 *node_changed = dn.node_changed;
b4d07a3e 3359out:
2aadac08
JK		 3360 f2fs_put_dnode(&dn);
3361unlock_out:
b4d07a3e 3362 if (locked)
0ef81833 3363 f2fs_do_map_lock(sbi, flag, false);
2aadac08
JK		 3364 return err;
3365}
3366
eb47b800
JK		 3367static int f2fs_write_begin(struct file *file, struct address_space *mapping,
3368 loff_t pos, unsigned len, unsigned flags,
3369 struct page **pagep, void **fsdata)
3370{
3371 struct inode *inode = mapping->host;
4081363f 3372 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
86531d6b 3373 struct page *page = NULL;
09cbfeaf 3374 pgoff_t index = ((unsigned long long) pos) >> PAGE_SHIFT;
a2e2e76b 3375 bool need_balance = false, drop_atomic = false;
2aadac08 3376 block_t blkaddr = NULL_ADDR;
eb47b800
JK		 3377 int err = 0;
3378
62aed044
CY		 3379 trace_f2fs_write_begin(inode, pos, len, flags);
3380
00e09c0b
CY		 3381 if (!f2fs_is_checkpoint_ready(sbi)) {
3382 err = -ENOSPC;
4354994f 3383 goto fail;
00e09c0b 3384 }
4354994f 3385
455e3a58
JK		 3386 if ((f2fs_is_atomic_file(inode) &&
3387 !f2fs_available_free_memory(sbi, INMEM_PAGES)) ||
3388 is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
57864ae5 3389 err = -ENOMEM;
a2e2e76b 3390 drop_atomic = true;
57864ae5
JK		 3391 goto fail;
3392 }
3393
5f727395
JK		 3394 /*
3395 * We should check this at this moment to avoid deadlock on inode page
3396 * and #0 page. The locking rule for inline_data conversion should be:
3397 * lock_page(page #0) -> lock_page(inode_page)
3398 */
3399 if (index != 0) {
3400 err = f2fs_convert_inline_inode(inode);
3401 if (err)
3402 goto fail;
3403 }
4c8ff709
CY		 3404
3405#ifdef CONFIG_F2FS_FS_COMPRESSION
3406 if (f2fs_compressed_file(inode)) {
3407 int ret;
3408
3409 *fsdata = NULL;
3410
7eab7a69
FC		 3411 if (len == PAGE_SIZE)
3412 goto repeat;
3413
4c8ff709
CY		 3414 ret = f2fs_prepare_compress_overwrite(inode, pagep,
3415 index, fsdata);
3416 if (ret < 0) {
3417 err = ret;
3418 goto fail;
3419 } else if (ret) {
3420 return 0;
3421 }
3422 }
3423#endif
3424
afcb7ca0 3425repeat:
86d54795
JK		 3426 /*
3427 * Do not use grab_cache_page_write_begin() to avoid deadlock due to
3428 * wait_for_stable_page. Will wait that below with our IO control.
3429 */
01eccef7 3430 page = f2fs_pagecache_get_page(mapping, index,
86d54795 3431 FGP_LOCK | FGP_WRITE | FGP_CREAT, GFP_NOFS);
3aab8f82
CY		 3432 if (!page) {
3433 err = -ENOMEM;
3434 goto fail;
3435 }
d5f66990 3436
4c8ff709
CY		 3437 /* TODO: cluster can be compressed due to race with .writepage */
3438
eb47b800
JK		 3439 *pagep = page;
3440
2aadac08
JK		 3441 err = prepare_write_begin(sbi, page, pos, len,
3442 &blkaddr, &need_balance);
9ba69cf9 3443 if (err)
2aadac08 3444 goto fail;
9ba69cf9 3445
af033b2a
CY		 3446 if (need_balance && !IS_NOQUOTA(inode) &&
3447 has_not_enough_free_secs(sbi, 0, 0)) {
2a340760 3448 unlock_page(page);
2c4db1a6 3449 f2fs_balance_fs(sbi, true);
2a340760
JK		 3450 lock_page(page);
3451 if (page->mapping != mapping) {
3452 /* The page got truncated from under us */
3453 f2fs_put_page(page, 1);
3454 goto repeat;
3455 }
3456 }
3457
bae0ee7a 3458 f2fs_wait_on_page_writeback(page, DATA, false, true);
b3d208f9 3459
649d7df2
JK		 3460 if (len == PAGE_SIZE || PageUptodate(page))
3461 return 0;
eb47b800 3462
95ae251f
EB		 3463 if (!(pos & (PAGE_SIZE - 1)) && (pos + len) >= i_size_read(inode) &&
3464 !f2fs_verity_in_progress(inode)) {
746e2403
YH		 3465 zero_user_segment(page, len, PAGE_SIZE);
3466 return 0;
3467 }
3468
2aadac08 3469 if (blkaddr == NEW_ADDR) {
09cbfeaf 3470 zero_user_segment(page, 0, PAGE_SIZE);
649d7df2 3471 SetPageUptodate(page);
eb47b800 3472 } else {
93770ab7
CY		 3473 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3474 DATA_GENERIC_ENHANCE_READ)) {
10f966bb 3475 err = -EFSCORRUPTED;
93770ab7
CY		 3476 goto fail;
3477 }
b7973091 3478 err = f2fs_submit_page_read(inode, page, blkaddr, 0, true);
13ba41e3 3479 if (err)
78682f79 3480 goto fail;
d54c795b 3481
393ff91f 3482 lock_page(page);
6bacf52f 3483 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 3484 f2fs_put_page(page, 1);
3485 goto repeat;
eb47b800 3486 }
1563ac75
CY		 3487 if (unlikely(!PageUptodate(page))) {
3488 err = -EIO;
3489 goto fail;
4375a336 3490 }
eb47b800 3491 }
eb47b800 3492 return 0;
9ba69cf9 3493
3aab8f82 3494fail:
86531d6b 3495 f2fs_put_page(page, 1);
3e679dc7 3496 f2fs_write_failed(inode, pos + len);
a2e2e76b 3497 if (drop_atomic)
4d57b86d 3498 f2fs_drop_inmem_pages_all(sbi, false);
3aab8f82 3499 return err;
eb47b800
JK		 3500}
3501
a1dd3c13
JK		 3502static int f2fs_write_end(struct file *file,
3503 struct address_space *mapping,
3504 loff_t pos, unsigned len, unsigned copied,
3505 struct page *page, void *fsdata)
3506{
3507 struct inode *inode = page->mapping->host;
3508
dfb2bf38
CY		 3509 trace_f2fs_write_end(inode, pos, len, copied);
3510
649d7df2
JK		 3511 /*
3512 * This should be come from len == PAGE_SIZE, and we expect copied
3513 * should be PAGE_SIZE. Otherwise, we treat it with zero copied and
3514 * let generic_perform_write() try to copy data again through copied=0.
3515 */
3516 if (!PageUptodate(page)) {
746e2403 3517 if (unlikely(copied != len))
649d7df2
JK		 3518 copied = 0;
3519 else
3520 SetPageUptodate(page);
3521 }
4c8ff709
CY		 3522
3523#ifdef CONFIG_F2FS_FS_COMPRESSION
3524 /* overwrite compressed file */
3525 if (f2fs_compressed_file(inode) && fsdata) {
3526 f2fs_compress_write_end(inode, fsdata, page->index, copied);
3527 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
944dd22e
CY		 3528
3529 if (pos + copied > i_size_read(inode) &&
3530 !f2fs_verity_in_progress(inode))
3531 f2fs_i_size_write(inode, pos + copied);
4c8ff709
CY		 3532 return copied;
3533 }
3534#endif
3535
649d7df2
JK		 3536 if (!copied)
3537 goto unlock_out;
3538
34ba94ba 3539 set_page_dirty(page);
a1dd3c13 3540
95ae251f
EB		 3541 if (pos + copied > i_size_read(inode) &&
3542 !f2fs_verity_in_progress(inode))
fc9581c8 3543 f2fs_i_size_write(inode, pos + copied);
649d7df2 3544unlock_out:
3024c9a1 3545 f2fs_put_page(page, 1);
d0239e1b 3546 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
a1dd3c13
JK		 3547 return copied;
3548}
3549
6f673763
OS		 3550static int check_direct_IO(struct inode *inode, struct iov_iter *iter,
3551 loff_t offset)
944fcfc1 3552{
8a56dd96
JK		 3553 unsigned i_blkbits = READ_ONCE(inode->i_blkbits);
3554 unsigned blkbits = i_blkbits;
3555 unsigned blocksize_mask = (1 << blkbits) - 1;
3556 unsigned long align = offset | iov_iter_alignment(iter);
3557 struct block_device *bdev = inode->i_sb->s_bdev;
3558
20d0a107
GKB		 3559 if (iov_iter_rw(iter) == READ && offset >= i_size_read(inode))
3560 return 1;
3561
8a56dd96
JK		 3562 if (align & blocksize_mask) {
3563 if (bdev)
3564 blkbits = blksize_bits(bdev_logical_block_size(bdev));
3565 blocksize_mask = (1 << blkbits) - 1;
3566 if (align & blocksize_mask)
3567 return -EINVAL;
3568 return 1;
3569 }
944fcfc1
JK		 3570 return 0;
3571}
3572
02b16d0a
CY		 3573static void f2fs_dio_end_io(struct bio *bio)
3574{
3575 struct f2fs_private_dio *dio = bio->bi_private;
3576
3577 dec_page_count(F2FS_I_SB(dio->inode),
3578 dio->write ? F2FS_DIO_WRITE : F2FS_DIO_READ);
3579
3580 bio->bi_private = dio->orig_private;
3581 bio->bi_end_io = dio->orig_end_io;
3582
c8eb7024 3583 kfree(dio);
02b16d0a
CY		 3584
3585 bio_endio(bio);
3586}
3587
3588static void f2fs_dio_submit_bio(struct bio *bio, struct inode *inode,
3589 loff_t file_offset)
3590{
3591 struct f2fs_private_dio *dio;
3592 bool write = (bio_op(bio) == REQ_OP_WRITE);
02b16d0a
CY		 3593
3594 dio = f2fs_kzalloc(F2FS_I_SB(inode),
3595 sizeof(struct f2fs_private_dio), GFP_NOFS);
8e114038 3596 if (!dio)
02b16d0a 3597 goto out;
02b16d0a
CY		 3598
3599 dio->inode = inode;
3600 dio->orig_end_io = bio->bi_end_io;
3601 dio->orig_private = bio->bi_private;
3602 dio->write = write;
3603
3604 bio->bi_end_io = f2fs_dio_end_io;
3605 bio->bi_private = dio;
3606
3607 inc_page_count(F2FS_I_SB(inode),
3608 write ? F2FS_DIO_WRITE : F2FS_DIO_READ);
3609
3610 submit_bio(bio);
3611 return;
3612out:
3613 bio->bi_status = BLK_STS_IOERR;
3614 bio_endio(bio);
3615}
3616
c8b8e32d 3617static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
eb47b800 3618{
b439b103 3619 struct address_space *mapping = iocb->ki_filp->f_mapping;
3aab8f82 3620 struct inode *inode = mapping->host;
0cdd3195 3621 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
f847c699 3622 struct f2fs_inode_info *fi = F2FS_I(inode);
3aab8f82 3623 size_t count = iov_iter_count(iter);
c8b8e32d 3624 loff_t offset = iocb->ki_pos;
82e0a5aa 3625 int rw = iov_iter_rw(iter);
3aab8f82 3626 int err;
0cdd3195 3627 enum rw_hint hint = iocb->ki_hint;
63189b78 3628 int whint_mode = F2FS_OPTION(sbi).whint_mode;
f847c699 3629 bool do_opu;
944fcfc1 3630
b439b103 3631 err = check_direct_IO(inode, iter, offset);
b9d777b8 3632 if (err)
8a56dd96 3633 return err < 0 ? err : 0;
9ffe0fb5 3634
f847c699 3635 if (f2fs_force_buffered_io(inode, iocb, iter))
36abef4e 3636 return 0;
fcc85a4d 3637
6de8687c 3638 do_opu = rw == WRITE && f2fs_lfs_mode(sbi);
f847c699 3639
5302fb00 3640 trace_f2fs_direct_IO_enter(inode, offset, count, rw);
70407fad 3641
0cdd3195
HL		 3642 if (rw == WRITE && whint_mode == WHINT_MODE_OFF)
3643 iocb->ki_hint = WRITE_LIFE_NOT_SET;
3644
f847c699
CY		 3645 if (iocb->ki_flags & IOCB_NOWAIT) {
3646 if (!down_read_trylock(&fi->i_gc_rwsem[rw])) {
3647 iocb->ki_hint = hint;
3648 err = -EAGAIN;
3649 goto out;
3650 }
3651 if (do_opu && !down_read_trylock(&fi->i_gc_rwsem[READ])) {
3652 up_read(&fi->i_gc_rwsem[rw]);
b91050a8
HL		 3653 iocb->ki_hint = hint;
3654 err = -EAGAIN;
3655 goto out;
3656 }
f847c699
CY		 3657 } else {
3658 down_read(&fi->i_gc_rwsem[rw]);
3659 if (do_opu)
3660 down_read(&fi->i_gc_rwsem[READ]);
b91050a8
HL		 3661 }
3662
02b16d0a 3663 err = __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev,
f9d6d059
CY		 3664 iter, rw == WRITE ? get_data_block_dio_write :
3665 get_data_block_dio, NULL, f2fs_dio_submit_bio,
ad8d6a02
D		 3666 rw == WRITE ? DIO_LOCKING | DIO_SKIP_HOLES :
3667 DIO_SKIP_HOLES);
f847c699
CY		 3668
3669 if (do_opu)
3670 up_read(&fi->i_gc_rwsem[READ]);
3671
3672 up_read(&fi->i_gc_rwsem[rw]);
82e0a5aa
CY		 3673
3674 if (rw == WRITE) {
0cdd3195
HL		 3675 if (whint_mode == WHINT_MODE_OFF)
3676 iocb->ki_hint = hint;
b0af6d49
CY		 3677 if (err > 0) {
3678 f2fs_update_iostat(F2FS_I_SB(inode), APP_DIRECT_IO,
3679 err);
f847c699
CY		 3680 if (!do_opu)
3681 set_inode_flag(inode, FI_UPDATE_WRITE);
335cac8b
JQ		 3682 } else if (err == -EIOCBQUEUED) {
3683 f2fs_update_iostat(F2FS_I_SB(inode), APP_DIRECT_IO,
3684 count - iov_iter_count(iter));
b0af6d49 3685 } else if (err < 0) {
3e679dc7 3686 f2fs_write_failed(inode, offset + count);
b0af6d49 3687 }
8b83ac81
CY		 3688 } else {
3689 if (err > 0)
3690 f2fs_update_iostat(sbi, APP_DIRECT_READ_IO, err);
335cac8b
JQ		 3691 else if (err == -EIOCBQUEUED)
3692 f2fs_update_iostat(F2FS_I_SB(inode), APP_DIRECT_READ_IO,
3693 count - iov_iter_count(iter));
6bfc4919 3694 }
70407fad 3695
b91050a8 3696out:
5302fb00 3697 trace_f2fs_direct_IO_exit(inode, offset, count, rw, err);
70407fad 3698
3aab8f82 3699 return err;
eb47b800
JK		 3700}
3701
487261f3
CY		 3702void f2fs_invalidate_page(struct page *page, unsigned int offset,
3703 unsigned int length)
eb47b800
JK		 3704{
3705 struct inode *inode = page->mapping->host;
487261f3 3706 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
a7ffdbe2 3707
487261f3 3708 if (inode->i_ino >= F2FS_ROOT_INO(sbi) &&
09cbfeaf 3709 (offset % PAGE_SIZE || length != PAGE_SIZE))
a7ffdbe2
JK		 3710 return;
3711
487261f3 3712 if (PageDirty(page)) {
933439c8 3713 if (inode->i_ino == F2FS_META_INO(sbi)) {
487261f3 3714 dec_page_count(sbi, F2FS_DIRTY_META);
933439c8 3715 } else if (inode->i_ino == F2FS_NODE_INO(sbi)) {
487261f3 3716 dec_page_count(sbi, F2FS_DIRTY_NODES);
933439c8 3717 } else {
487261f3 3718 inode_dec_dirty_pages(inode);
4d57b86d 3719 f2fs_remove_dirty_inode(inode);
933439c8 3720 }
487261f3 3721 }
decd36b6 3722
b763f3be 3723 clear_page_private_gcing(page);
2baf0781 3724
6ce19aff
CY		 3725 if (test_opt(sbi, COMPRESS_CACHE)) {
3726 if (f2fs_compressed_file(inode))
3727 f2fs_invalidate_compress_pages(sbi, inode->i_ino);
3728 if (inode->i_ino == F2FS_COMPRESS_INO(sbi))
3729 clear_page_private_data(page);
3730 }
3731
b763f3be 3732 if (page_private_atomic(page))
4d57b86d 3733 return f2fs_drop_inmem_page(inode, page);
decd36b6 3734
b763f3be
CY		 3735 detach_page_private(page);
3736 set_page_private(page, 0);
eb47b800
JK		 3737}
3738
487261f3 3739int f2fs_release_page(struct page *page, gfp_t wait)
eb47b800 3740{
f68daeeb
JK		 3741 /* If this is dirty page, keep PagePrivate */
3742 if (PageDirty(page))
3743 return 0;
3744
decd36b6 3745 /* This is atomic written page, keep Private */
b763f3be 3746 if (page_private_atomic(page))
decd36b6
CY		 3747 return 0;
3748
6ce19aff
CY		 3749 if (test_opt(F2FS_P_SB(page), COMPRESS_CACHE)) {
3750 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
3751 struct inode *inode = page->mapping->host;
3752
3753 if (f2fs_compressed_file(inode))
3754 f2fs_invalidate_compress_pages(sbi, inode->i_ino);
3755 if (inode->i_ino == F2FS_COMPRESS_INO(sbi))
3756 clear_page_private_data(page);
3757 }
3758
b763f3be
CY		 3759 clear_page_private_gcing(page);
3760
3761 detach_page_private(page);
3762 set_page_private(page, 0);
c3850aa1 3763 return 1;
eb47b800
JK		 3764}
3765
3766static int f2fs_set_data_page_dirty(struct page *page)
3767{
4969c06a 3768 struct inode *inode = page_file_mapping(page)->host;
eb47b800 3769
26c6b887
JK		 3770 trace_f2fs_set_page_dirty(page, DATA);
3771
237c0790
JK		 3772 if (!PageUptodate(page))
3773 SetPageUptodate(page);
4969c06a
JK		 3774 if (PageSwapCache(page))
3775 return __set_page_dirty_nobuffers(page);
34ba94ba 3776
5fe45743 3777 if (f2fs_is_atomic_file(inode) && !f2fs_is_commit_atomic_write(inode)) {
b763f3be 3778 if (!page_private_atomic(page)) {
4d57b86d 3779 f2fs_register_inmem_page(inode, page);
decd36b6
CY		 3780 return 1;
3781 }
3782 /*
3783 * Previously, this page has been registered, we just
3784 * return here.
3785 */
3786 return 0;
34ba94ba
JK		 3787 }
3788
eb47b800 3789 if (!PageDirty(page)) {
b87078ad 3790 __set_page_dirty_nobuffers(page);
4d57b86d 3791 f2fs_update_dirty_page(inode, page);
eb47b800
JK		 3792 return 1;
3793 }
3794 return 0;
3795}
3796
c1c63387
CY		 3797
3798static sector_t f2fs_bmap_compress(struct inode *inode, sector_t block)
3799{
3800#ifdef CONFIG_F2FS_FS_COMPRESSION
3801 struct dnode_of_data dn;
3802 sector_t start_idx, blknr = 0;
3803 int ret;
3804
3805 start_idx = round_down(block, F2FS_I(inode)->i_cluster_size);
3806
3807 set_new_dnode(&dn, inode, NULL, NULL, 0);
3808 ret = f2fs_get_dnode_of_data(&dn, start_idx, LOOKUP_NODE);
3809 if (ret)
3810 return 0;
3811
3812 if (dn.data_blkaddr != COMPRESS_ADDR) {
3813 dn.ofs_in_node += block - start_idx;
3814 blknr = f2fs_data_blkaddr(&dn);
3815 if (!__is_valid_data_blkaddr(blknr))
3816 blknr = 0;
3817 }
3818
3819 f2fs_put_dnode(&dn);
c1c63387
CY		 3820 return blknr;
3821#else
250e84d7 3822 return 0;
c1c63387
CY		 3823#endif
3824}
3825
3826
c01e54b7
JK		 3827static sector_t f2fs_bmap(struct address_space *mapping, sector_t block)
3828{
454ae7e5 3829 struct inode *inode = mapping->host;
b79b0a31 3830 sector_t blknr = 0;
454ae7e5 3831
1d373a0e 3832 if (f2fs_has_inline_data(inode))
b79b0a31 3833 goto out;
1d373a0e
JK		 3834
3835 /* make sure allocating whole blocks */
3836 if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
3837 filemap_write_and_wait(mapping);
3838
4eda1682 3839 /* Block number less than F2FS MAX BLOCKS */
6d1451bf 3840 if (unlikely(block >= max_file_blocks(inode)))
4eda1682 3841 goto out;
c1c63387 3842
4eda1682
DJ		 3843 if (f2fs_compressed_file(inode)) {
3844 blknr = f2fs_bmap_compress(inode, block);
3845 } else {
b876f4c9
JK		 3846 struct f2fs_map_blocks map;
3847
3848 memset(&map, 0, sizeof(map));
3849 map.m_lblk = block;
3850 map.m_len = 1;
3851 map.m_next_pgofs = NULL;
3852 map.m_seg_type = NO_CHECK_TYPE;
3853
3854 if (!f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_BMAP))
3855 blknr = map.m_pblk;
4eda1682 3856 }
b79b0a31
CY		 3857out:
3858 trace_f2fs_bmap(inode, block, blknr);
3859 return blknr;
429511cd
CY		 3860}
3861
5b7a487c
WG		 3862#ifdef CONFIG_MIGRATION
3863#include <linux/migrate.h>
3864
3865int f2fs_migrate_page(struct address_space *mapping,
3866 struct page *newpage, struct page *page, enum migrate_mode mode)
3867{
3868 int rc, extra_count;
3869 struct f2fs_inode_info *fi = F2FS_I(mapping->host);
b763f3be 3870 bool atomic_written = page_private_atomic(page);
5b7a487c
WG		 3871
3872 BUG_ON(PageWriteback(page));
3873
3874 /* migrating an atomic written page is safe with the inmem_lock hold */
ff1048e7
JK		 3875 if (atomic_written) {
3876 if (mode != MIGRATE_SYNC)
3877 return -EBUSY;
3878 if (!mutex_trylock(&fi->inmem_lock))
3879 return -EAGAIN;
3880 }
5b7a487c 3881
240a5915
CY		 3882 /* one extra reference was held for atomic_write page */
3883 extra_count = atomic_written ? 1 : 0;
5b7a487c 3884 rc = migrate_page_move_mapping(mapping, newpage,
37109694 3885 page, extra_count);
5b7a487c
WG		 3886 if (rc != MIGRATEPAGE_SUCCESS) {
3887 if (atomic_written)
3888 mutex_unlock(&fi->inmem_lock);
3889 return rc;
3890 }
3891
3892 if (atomic_written) {
3893 struct inmem_pages *cur;
5f029c04 3894
5b7a487c
WG		 3895 list_for_each_entry(cur, &fi->inmem_pages, list)
3896 if (cur->page == page) {
3897 cur->page = newpage;
3898 break;
3899 }
3900 mutex_unlock(&fi->inmem_lock);
3901 put_page(page);
3902 get_page(newpage);
3903 }
3904
c9ebd3df
JK		 3905 /* guarantee to start from no stale private field */
3906 set_page_private(newpage, 0);
240a5915 3907 if (PagePrivate(page)) {
b763f3be
CY		 3908 set_page_private(newpage, page_private(page));
3909 SetPagePrivate(newpage);
3910 get_page(newpage);
3911
3912 set_page_private(page, 0);
3913 ClearPagePrivate(page);
3914 put_page(page);
240a5915 3915 }
5b7a487c 3916
2916ecc0
JG		 3917 if (mode != MIGRATE_SYNC_NO_COPY)
3918 migrate_page_copy(newpage, page);
3919 else
3920 migrate_page_states(newpage, page);
5b7a487c
WG		 3921
3922 return MIGRATEPAGE_SUCCESS;
3923}
3924#endif
3925
4969c06a 3926#ifdef CONFIG_SWAP
859fca6b
CY		 3927static int f2fs_migrate_blocks(struct inode *inode, block_t start_blk,
3928 unsigned int blkcnt)
3929{
3930 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3931 unsigned int blkofs;
3932 unsigned int blk_per_sec = BLKS_PER_SEC(sbi);
3933 unsigned int secidx = start_blk / blk_per_sec;
3934 unsigned int end_sec = secidx + blkcnt / blk_per_sec;
3935 int ret = 0;
3936
3937 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
edc6d01b 3938 filemap_invalidate_lock(inode->i_mapping);
859fca6b
CY		 3939
3940 set_inode_flag(inode, FI_ALIGNED_WRITE);
3941
3942 for (; secidx < end_sec; secidx++) {
3943 down_write(&sbi->pin_sem);
3944
3945 f2fs_lock_op(sbi);
3946 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
3947 f2fs_unlock_op(sbi);
3948
3949 set_inode_flag(inode, FI_DO_DEFRAG);
3950
3951 for (blkofs = 0; blkofs < blk_per_sec; blkofs++) {
3952 struct page *page;
3953 unsigned int blkidx = secidx * blk_per_sec + blkofs;
3954
3955 page = f2fs_get_lock_data_page(inode, blkidx, true);
3956 if (IS_ERR(page)) {
3957 up_write(&sbi->pin_sem);
3958 ret = PTR_ERR(page);
3959 goto done;
3960 }
3961
3962 set_page_dirty(page);
3963 f2fs_put_page(page, 1);
3964 }
3965
3966 clear_inode_flag(inode, FI_DO_DEFRAG);
3967
3968 ret = filemap_fdatawrite(inode->i_mapping);
3969
3970 up_write(&sbi->pin_sem);
3971
3972 if (ret)
3973 break;
3974 }
3975
3976done:
3977 clear_inode_flag(inode, FI_DO_DEFRAG);
3978 clear_inode_flag(inode, FI_ALIGNED_WRITE);
3979
edc6d01b 3980 filemap_invalidate_unlock(inode->i_mapping);
859fca6b
CY		 3981 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3982
3983 return ret;
3984}
3985
0b8fc006 3986static int check_swap_activate(struct swap_info_struct *sis,
af4b6b8e
CY		 3987 struct file *swap_file, sector_t *span)
3988{
3989 struct address_space *mapping = swap_file->f_mapping;
3990 struct inode *inode = mapping->host;
36e4d958 3991 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
af4b6b8e
CY		 3992 sector_t cur_lblock;
3993 sector_t last_lblock;
3994 sector_t pblock;
3995 sector_t lowest_pblock = -1;
3996 sector_t highest_pblock = 0;
3997 int nr_extents = 0;
3998 unsigned long nr_pblocks;
859fca6b
CY		 3999 unsigned int blks_per_sec = BLKS_PER_SEC(sbi);
4000 unsigned int sec_blks_mask = BLKS_PER_SEC(sbi) - 1;
ca298241 4001 unsigned int not_aligned = 0;
36e4d958 4002 int ret = 0;
af4b6b8e
CY		 4003
4004 /*
4005 * Map all the blocks into the extent list. This code doesn't try
4006 * to be very smart.
4007 */
4008 cur_lblock = 0;
6cbfcab5 4009 last_lblock = bytes_to_blks(inode, i_size_read(inode));
af4b6b8e 4010
1da66103 4011 while (cur_lblock < last_lblock && cur_lblock < sis->max) {
b876f4c9 4012 struct f2fs_map_blocks map;
859fca6b 4013retry:
af4b6b8e
CY		 4014 cond_resched();
4015
b876f4c9
JK		 4016 memset(&map, 0, sizeof(map));
4017 map.m_lblk = cur_lblock;
36e4d958 4018 map.m_len = last_lblock - cur_lblock;
4019 map.m_next_pgofs = NULL;
4020 map.m_next_extent = NULL;
b876f4c9 4021 map.m_seg_type = NO_CHECK_TYPE;
36e4d958 4022 map.m_may_create = false;
af4b6b8e 4023
b876f4c9 4024 ret = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_FIEMAP);
af4b6b8e 4025 if (ret)
36e4d958 4026 goto out;
af4b6b8e
CY		 4027
4028 /* hole */
36e4d958 4029 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
833dcd35 4030 f2fs_err(sbi, "Swapfile has holes");
f395183f 4031 ret = -EINVAL;
36e4d958 4032 goto out;
4033 }
af4b6b8e 4034
b876f4c9
JK		 4035 pblock = map.m_pblk;
4036 nr_pblocks = map.m_len;
af4b6b8e 4037
859fca6b
CY		 4038 if ((pblock - SM_I(sbi)->main_blkaddr) & sec_blks_mask ||
4039 nr_pblocks & sec_blks_mask) {
ca298241 4040 not_aligned++;
36e4d958 4041
859fca6b
CY		 4042 nr_pblocks = roundup(nr_pblocks, blks_per_sec);
4043 if (cur_lblock + nr_pblocks > sis->max)
4044 nr_pblocks -= blks_per_sec;
4045
4046 if (!nr_pblocks) {
4047 /* this extent is last one */
4048 nr_pblocks = map.m_len;
4049 f2fs_warn(sbi, "Swapfile: last extent is not aligned to section");
4050 goto next;
4051 }
4052
4053 ret = f2fs_migrate_blocks(inode, cur_lblock,
4054 nr_pblocks);
4055 if (ret)
4056 goto out;
4057 goto retry;
4058 }
4059next:
af4b6b8e
CY		 4060 if (cur_lblock + nr_pblocks >= sis->max)
4061 nr_pblocks = sis->max - cur_lblock;
4062
4063 if (cur_lblock) { /* exclude the header page */
4064 if (pblock < lowest_pblock)
4065 lowest_pblock = pblock;
4066 if (pblock + nr_pblocks - 1 > highest_pblock)
4067 highest_pblock = pblock + nr_pblocks - 1;
4068 }
4069
4070 /*
4071 * We found a PAGE_SIZE-length, PAGE_SIZE-aligned run of blocks
4072 */
4073 ret = add_swap_extent(sis, cur_lblock, nr_pblocks, pblock);
4074 if (ret < 0)
4075 goto out;
4076 nr_extents += ret;
4077 cur_lblock += nr_pblocks;
4078 }
4079 ret = nr_extents;
4080 *span = 1 + highest_pblock - lowest_pblock;
4081 if (cur_lblock == 0)
4082 cur_lblock = 1; /* force Empty message */
4083 sis->max = cur_lblock;
4084 sis->pages = cur_lblock - 1;
4085 sis->highest_bit = cur_lblock - 1;
4086out:
859fca6b
CY		 4087 if (not_aligned)
4088 f2fs_warn(sbi, "Swapfile (%u) is not align to section: 1) creat(), 2) ioctl(F2FS_IOC_SET_PIN_FILE), 3) fallocate(%u * N)",
4089 not_aligned, blks_per_sec * F2FS_BLKSIZE);
af4b6b8e 4090 return ret;
af4b6b8e
CY		 4091}
4092
4969c06a
JK		 4093static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
4094 sector_t *span)
4095{
4096 struct inode *inode = file_inode(file);
4097 int ret;
4098
4099 if (!S_ISREG(inode->i_mode))
4100 return -EINVAL;
4101
4102 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
4103 return -EROFS;
4104
d927ccfc
SK		 4105 if (f2fs_lfs_mode(F2FS_I_SB(inode))) {
4106 f2fs_err(F2FS_I_SB(inode),
4107 "Swapfile not supported in LFS mode");
4108 return -EINVAL;
4109 }
4110
4969c06a
JK		 4111 ret = f2fs_convert_inline_inode(inode);
4112 if (ret)
4113 return ret;
4114
78134d03 4115 if (!f2fs_disable_compressed_file(inode))
4c8ff709
CY		 4116 return -EINVAL;
4117
0b979f1b
CY		 4118 f2fs_precache_extents(inode);
4119
3e5e479a
CY		 4120 ret = check_swap_activate(sis, file, span);
4121 if (ret < 0)
4969c06a
JK		 4122 return ret;
4123
4124 set_inode_flag(inode, FI_PIN_FILE);
4969c06a 4125 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3e5e479a 4126 return ret;
4969c06a
JK		 4127}
4128
4129static void f2fs_swap_deactivate(struct file *file)
4130{
4131 struct inode *inode = file_inode(file);
4132
4133 clear_inode_flag(inode, FI_PIN_FILE);
4134}
4135#else
4136static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
4137 sector_t *span)
4138{
4139 return -EOPNOTSUPP;
4140}
4141
4142static void f2fs_swap_deactivate(struct file *file)
4143{
4144}
4145#endif
4146
eb47b800
JK		 4147const struct address_space_operations f2fs_dblock_aops = {
4148 .readpage = f2fs_read_data_page,
23323196 4149 .readahead = f2fs_readahead,
eb47b800
JK		 4150 .writepage = f2fs_write_data_page,
4151 .writepages = f2fs_write_data_pages,
4152 .write_begin = f2fs_write_begin,
a1dd3c13 4153 .write_end = f2fs_write_end,
eb47b800 4154 .set_page_dirty = f2fs_set_data_page_dirty,
487261f3
CY		 4155 .invalidatepage = f2fs_invalidate_page,
4156 .releasepage = f2fs_release_page,
eb47b800 4157 .direct_IO = f2fs_direct_IO,
c01e54b7 4158 .bmap = f2fs_bmap,
4969c06a
JK		 4159 .swap_activate = f2fs_swap_activate,
4160 .swap_deactivate = f2fs_swap_deactivate,
5b7a487c
WG		 4161#ifdef CONFIG_MIGRATION
4162 .migratepage = f2fs_migrate_page,
4163#endif
eb47b800 4164};
6dbb1796 4165
5ec2d99d 4166void f2fs_clear_page_cache_dirty_tag(struct page *page)
aec2f729
CY		 4167{
4168 struct address_space *mapping = page_mapping(page);
4169 unsigned long flags;
4170
4171 xa_lock_irqsave(&mapping->i_pages, flags);
5ec2d99d 4172 __xa_clear_mark(&mapping->i_pages, page_index(page),
aec2f729
CY		 4173 PAGECACHE_TAG_DIRTY);
4174 xa_unlock_irqrestore(&mapping->i_pages, flags);
4175}
4176
6dbb1796
EB		 4177int __init f2fs_init_post_read_processing(void)
4178{
95ae251f
EB		 4179 bio_post_read_ctx_cache =
4180 kmem_cache_create("f2fs_bio_post_read_ctx",
4181 sizeof(struct bio_post_read_ctx), 0, 0, NULL);
6dbb1796
EB		 4182 if (!bio_post_read_ctx_cache)
4183 goto fail;
4184 bio_post_read_ctx_pool =
4185 mempool_create_slab_pool(NUM_PREALLOC_POST_READ_CTXS,
4186 bio_post_read_ctx_cache);
4187 if (!bio_post_read_ctx_pool)
4188 goto fail_free_cache;
4189 return 0;
4190
4191fail_free_cache:
4192 kmem_cache_destroy(bio_post_read_ctx_cache);
4193fail:
4194 return -ENOMEM;
4195}
4196
0b20fcec 4197void f2fs_destroy_post_read_processing(void)
6dbb1796
EB		 4198{
4199 mempool_destroy(bio_post_read_ctx_pool);
4200 kmem_cache_destroy(bio_post_read_ctx_cache);
4201}
0b20fcec 4202
4c8ff709
CY		 4203int f2fs_init_post_read_wq(struct f2fs_sb_info *sbi)
4204{
4205 if (!f2fs_sb_has_encrypt(sbi) &&
4206 !f2fs_sb_has_verity(sbi) &&
4207 !f2fs_sb_has_compression(sbi))
4208 return 0;
4209
4210 sbi->post_read_wq = alloc_workqueue("f2fs_post_read_wq",
4211 WQ_UNBOUND | WQ_HIGHPRI,
4212 num_online_cpus());
4213 if (!sbi->post_read_wq)
4214 return -ENOMEM;
4215 return 0;
4216}
4217
4218void f2fs_destroy_post_read_wq(struct f2fs_sb_info *sbi)
4219{
4220 if (sbi->post_read_wq)
4221 destroy_workqueue(sbi->post_read_wq);
4222}
4223
0b20fcec
CY		 4224int __init f2fs_init_bio_entry_cache(void)
4225{
98510003 4226 bio_entry_slab = f2fs_kmem_cache_create("f2fs_bio_entry_slab",
0b20fcec
CY		 4227 sizeof(struct bio_entry));
4228 if (!bio_entry_slab)
4229 return -ENOMEM;
4230 return 0;
4231}
4232
f543805f 4233void f2fs_destroy_bio_entry_cache(void)
0b20fcec
CY		 4234{
4235 kmem_cache_destroy(bio_entry_slab);
4236}
Michael's Linux tree.