]> git.ipfire.org Git - people/ms/linux.git/blame - fs/f2fs/data.c
projects / people / ms / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: fix to caclulate required free section correctly
[people/ms/linux.git] / fs / f2fs / data.c
CommitLineData
0a8165d7 1/*
eb47b800
JK		 2 * fs/f2fs/data.c
3 *
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 */
11#include <linux/fs.h>
12#include <linux/f2fs_fs.h>
13#include <linux/buffer_head.h>
14#include <linux/mpage.h>
15#include <linux/writeback.h>
16#include <linux/backing-dev.h>
8f46dcae 17#include <linux/pagevec.h>
eb47b800
JK		 18#include <linux/blkdev.h>
19#include <linux/bio.h>
690e4a3e 20#include <linux/prefetch.h>
e2e40f2c 21#include <linux/uio.h>
fe76b796
JK		 22#include <linux/mm.h>
23#include <linux/memcontrol.h>
f1e88660 24#include <linux/cleancache.h>
174cd4b1 25#include <linux/sched/signal.h>
eb47b800
JK		 26
27#include "f2fs.h"
28#include "node.h"
29#include "segment.h"
db9f7c1a 30#include "trace.h"
848753aa 31#include <trace/events/f2fs.h>
eb47b800 32
36951b38
CY		 33static bool __is_cp_guaranteed(struct page *page)
34{
35 struct address_space *mapping = page->mapping;
36 struct inode *inode;
37 struct f2fs_sb_info *sbi;
38
39 if (!mapping)
40 return false;
41
42 inode = mapping->host;
43 sbi = F2FS_I_SB(inode);
44
45 if (inode->i_ino == F2FS_META_INO(sbi) ||
46 inode->i_ino == F2FS_NODE_INO(sbi) ||
47 S_ISDIR(inode->i_mode) ||
48 is_cold_data(page))
49 return true;
50 return false;
51}
52
4246a0b6 53static void f2fs_read_end_io(struct bio *bio)
93dfe2ac 54{
f568849e
LT		 55 struct bio_vec *bvec;
56 int i;
93dfe2ac 57
8b038c70 58#ifdef CONFIG_F2FS_FAULT_INJECTION
55523519
CY		 59 if (time_to_inject(F2FS_P_SB(bio->bi_io_vec->bv_page), FAULT_IO)) {
60 f2fs_show_injection_info(FAULT_IO);
4e4cbee9 61 bio->bi_status = BLK_STS_IOERR;
55523519 62 }
8b038c70
CY		 63#endif
64
4375a336 65 if (f2fs_bio_encrypted(bio)) {
4e4cbee9 66 if (bio->bi_status) {
0b81d077 67 fscrypt_release_ctx(bio->bi_private);
4375a336 68 } else {
0b81d077 69 fscrypt_decrypt_bio_pages(bio->bi_private, bio);
4375a336
JK		 70 return;
71 }
72 }
73
12377024
CY		 74 bio_for_each_segment_all(bvec, bio, i) {
75 struct page *page = bvec->bv_page;
f1e88660 76
4e4cbee9 77 if (!bio->bi_status) {
237c0790
JK		 78 if (!PageUptodate(page))
79 SetPageUptodate(page);
f1e88660
JK		 80 } else {
81 ClearPageUptodate(page);
82 SetPageError(page);
83 }
84 unlock_page(page);
85 }
f1e88660
JK		 86 bio_put(bio);
87}
88
4246a0b6 89static void f2fs_write_end_io(struct bio *bio)
93dfe2ac 90{
1b1f559f 91 struct f2fs_sb_info *sbi = bio->bi_private;
f568849e
LT		 92 struct bio_vec *bvec;
93 int i;
93dfe2ac 94
f568849e 95 bio_for_each_segment_all(bvec, bio, i) {
93dfe2ac 96 struct page *page = bvec->bv_page;
36951b38 97 enum count_type type = WB_DATA_TYPE(page);
93dfe2ac 98
0a595eba
JK		 99 if (IS_DUMMY_WRITTEN_PAGE(page)) {
100 set_page_private(page, (unsigned long)NULL);
101 ClearPagePrivate(page);
102 unlock_page(page);
103 mempool_free(page, sbi->write_io_dummy);
104
4e4cbee9 105 if (unlikely(bio->bi_status))
0a595eba
JK		 106 f2fs_stop_checkpoint(sbi, true);
107 continue;
108 }
109
0b81d077 110 fscrypt_pullback_bio_page(&page, true);
4375a336 111
4e4cbee9 112 if (unlikely(bio->bi_status)) {
5114a97a 113 mapping_set_error(page->mapping, -EIO);
b1ca321d
JK		 114 if (type == F2FS_WB_CP_DATA)
115 f2fs_stop_checkpoint(sbi, true);
93dfe2ac 116 }
36951b38
CY		 117 dec_page_count(sbi, type);
118 clear_cold_data(page);
93dfe2ac 119 end_page_writeback(page);
f568849e 120 }
36951b38 121 if (!get_pages(sbi, F2FS_WB_CP_DATA) &&
f5730184 122 wq_has_sleeper(&sbi->cp_wait))
93dfe2ac
JK		 123 wake_up(&sbi->cp_wait);
124
125 bio_put(bio);
126}
127
3c62be17
JK		 128/*
129 * Return true, if pre_bio's bdev is same as its target device.
130 */
131struct block_device *f2fs_target_device(struct f2fs_sb_info *sbi,
132 block_t blk_addr, struct bio *bio)
133{
134 struct block_device *bdev = sbi->sb->s_bdev;
135 int i;
136
137 for (i = 0; i < sbi->s_ndevs; i++) {
138 if (FDEV(i).start_blk <= blk_addr &&
139 FDEV(i).end_blk >= blk_addr) {
140 blk_addr -= FDEV(i).start_blk;
141 bdev = FDEV(i).bdev;
142 break;
143 }
144 }
145 if (bio) {
74d46992 146 bio_set_dev(bio, bdev);
3c62be17
JK		 147 bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr);
148 }
149 return bdev;
150}
151
152int f2fs_target_device_index(struct f2fs_sb_info *sbi, block_t blkaddr)
153{
154 int i;
155
156 for (i = 0; i < sbi->s_ndevs; i++)
157 if (FDEV(i).start_blk <= blkaddr && FDEV(i).end_blk >= blkaddr)
158 return i;
159 return 0;
160}
161
162static bool __same_bdev(struct f2fs_sb_info *sbi,
163 block_t blk_addr, struct bio *bio)
164{
74d46992
CH		 165 struct block_device *b = f2fs_target_device(sbi, blk_addr, NULL);
166 return bio->bi_disk == b->bd_disk && bio->bi_partno == b->bd_partno;
3c62be17
JK		 167}
168
940a6d34
GZ		 169/*
170 * Low-level block read/write IO operations.
171 */
172static struct bio *__bio_alloc(struct f2fs_sb_info *sbi, block_t blk_addr,
578c6478 173 struct writeback_control *wbc,
940a6d34
GZ		 174 int npages, bool is_read)
175{
176 struct bio *bio;
177
d62fe971 178 bio = f2fs_bio_alloc(sbi, npages, true);
940a6d34 179
3c62be17 180 f2fs_target_device(sbi, blk_addr, bio);
940a6d34 181 bio->bi_end_io = is_read ? f2fs_read_end_io : f2fs_write_end_io;
12377024 182 bio->bi_private = is_read ? NULL : sbi;
578c6478
YY		 183 if (wbc)
184 wbc_init_bio(wbc, bio);
940a6d34
GZ		 185
186 return bio;
187}
188
4fc29c1a
LT		 189static inline void __submit_bio(struct f2fs_sb_info *sbi,
190 struct bio *bio, enum page_type type)
f5730184 191{
4fc29c1a 192 if (!is_read_io(bio_op(bio))) {
0a595eba
JK		 193 unsigned int start;
194
0bfd7a09 195 if (f2fs_sb_mounted_blkzoned(sbi->sb) &&
52763a4b 196 current->plug && (type == DATA || type == NODE))
19a5f5e2 197 blk_finish_plug(current->plug);
0a595eba
JK		 198
199 if (type != DATA && type != NODE)
200 goto submit_io;
201
202 start = bio->bi_iter.bi_size >> F2FS_BLKSIZE_BITS;
203 start %= F2FS_IO_SIZE(sbi);
204
205 if (start == 0)
206 goto submit_io;
207
208 /* fill dummy pages */
209 for (; start < F2FS_IO_SIZE(sbi); start++) {
210 struct page *page =
211 mempool_alloc(sbi->write_io_dummy,
212 GFP_NOIO | __GFP_ZERO | __GFP_NOFAIL);
213 f2fs_bug_on(sbi, !page);
214
215 SetPagePrivate(page);
216 set_page_private(page, (unsigned long)DUMMY_WRITTEN_PAGE);
217 lock_page(page);
218 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE)
219 f2fs_bug_on(sbi, 1);
220 }
221 /*
222 * In the NODE case, we lose next block address chain. So, we
223 * need to do checkpoint in f2fs_sync_file.
224 */
225 if (type == NODE)
226 set_sbi_flag(sbi, SBI_NEED_CP);
19a5f5e2 227 }
0a595eba 228submit_io:
554b5125
JK		 229 if (is_read_io(bio_op(bio)))
230 trace_f2fs_submit_read_bio(sbi->sb, type, bio);
231 else
232 trace_f2fs_submit_write_bio(sbi->sb, type, bio);
4e49ea4a 233 submit_bio(bio);
f5730184
JK		 234}
235
458e6197 236static void __submit_merged_bio(struct f2fs_bio_info *io)
93dfe2ac 237{
458e6197 238 struct f2fs_io_info *fio = &io->fio;
93dfe2ac
JK		 239
240 if (!io->bio)
241 return;
242
554b5125
JK		 243 bio_set_op_attrs(io->bio, fio->op, fio->op_flags);
244
04d328de 245 if (is_read_io(fio->op))
554b5125 246 trace_f2fs_prepare_read_bio(io->sbi->sb, fio->type, io->bio);
6a8f8ca5 247 else
554b5125 248 trace_f2fs_prepare_write_bio(io->sbi->sb, fio->type, io->bio);
04d328de 249
4fc29c1a 250 __submit_bio(io->sbi, io->bio, fio->type);
93dfe2ac
JK		 251 io->bio = NULL;
252}
253
942fd319
JK		 254static bool __has_merged_page(struct f2fs_bio_info *io,
255 struct inode *inode, nid_t ino, pgoff_t idx)
0fd785eb 256{
0fd785eb
CY		 257 struct bio_vec *bvec;
258 struct page *target;
259 int i;
260
0c3a5797 261 if (!io->bio)
0fd785eb 262 return false;
0c3a5797 263
942fd319 264 if (!inode && !ino)
0c3a5797 265 return true;
0fd785eb
CY		 266
267 bio_for_each_segment_all(bvec, io->bio, i) {
268
0b81d077 269 if (bvec->bv_page->mapping)
0fd785eb 270 target = bvec->bv_page;
0b81d077
JK		 271 else
272 target = fscrypt_control_page(bvec->bv_page);
0fd785eb 273
942fd319
JK		 274 if (idx != target->index)
275 continue;
276
0c3a5797
CY		 277 if (inode && inode == target->mapping->host)
278 return true;
0c3a5797 279 if (ino && ino == ino_of_node(target))
0fd785eb 280 return true;
0fd785eb
CY		 281 }
282
0fd785eb
CY		 283 return false;
284}
285
0c3a5797 286static bool has_merged_page(struct f2fs_sb_info *sbi, struct inode *inode,
942fd319 287 nid_t ino, pgoff_t idx, enum page_type type)
0c3a5797
CY		 288{
289 enum page_type btype = PAGE_TYPE_OF_BIO(type);
a912b54d
JK		 290 enum temp_type temp;
291 struct f2fs_bio_info *io;
292 bool ret = false;
0c3a5797 293
a912b54d
JK		 294 for (temp = HOT; temp < NR_TEMP_TYPE; temp++) {
295 io = sbi->write_io[btype] + temp;
296
297 down_read(&io->io_rwsem);
298 ret = __has_merged_page(io, inode, ino, idx);
299 up_read(&io->io_rwsem);
0c3a5797 300
a912b54d
JK		 301 /* TODO: use HOT temp only for meta pages now. */
302 if (ret || btype == META)
303 break;
304 }
0c3a5797
CY		 305 return ret;
306}
307
b9109b0e 308static void __f2fs_submit_merged_write(struct f2fs_sb_info *sbi,
a912b54d 309 enum page_type type, enum temp_type temp)
93dfe2ac
JK		 310{
311 enum page_type btype = PAGE_TYPE_OF_BIO(type);
a912b54d 312 struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
93dfe2ac 313
df0f8dc0 314 down_write(&io->io_rwsem);
458e6197
JK		 315
316 /* change META to META_FLUSH in the checkpoint procedure */
317 if (type >= META_FLUSH) {
318 io->fio.type = META_FLUSH;
04d328de 319 io->fio.op = REQ_OP_WRITE;
3adc5fcb 320 io->fio.op_flags = REQ_META | REQ_PRIO | REQ_SYNC;
70fd7614 321 if (!test_opt(sbi, NOBARRIER))
7f54f51f 322 io->fio.op_flags |= REQ_PREFLUSH | REQ_FUA;
458e6197
JK		 323 }
324 __submit_merged_bio(io);
df0f8dc0 325 up_write(&io->io_rwsem);
93dfe2ac
JK		 326}
327
a912b54d
JK		 328static void __submit_merged_write_cond(struct f2fs_sb_info *sbi,
329 struct inode *inode, nid_t ino, pgoff_t idx,
330 enum page_type type, bool force)
0c3a5797 331{
a912b54d
JK		 332 enum temp_type temp;
333
334 if (!force && !has_merged_page(sbi, inode, ino, idx, type))
335 return;
336
337 for (temp = HOT; temp < NR_TEMP_TYPE; temp++) {
338
339 __f2fs_submit_merged_write(sbi, type, temp);
340
341 /* TODO: use HOT temp only for meta pages now. */
342 if (type >= META)
343 break;
344 }
0c3a5797
CY		 345}
346
b9109b0e 347void f2fs_submit_merged_write(struct f2fs_sb_info *sbi, enum page_type type)
0c3a5797 348{
a912b54d 349 __submit_merged_write_cond(sbi, NULL, 0, 0, type, true);
0c3a5797
CY		 350}
351
b9109b0e 352void f2fs_submit_merged_write_cond(struct f2fs_sb_info *sbi,
942fd319 353 struct inode *inode, nid_t ino, pgoff_t idx,
b9109b0e 354 enum page_type type)
0c3a5797 355{
a912b54d 356 __submit_merged_write_cond(sbi, inode, ino, idx, type, false);
0c3a5797
CY		 357}
358
b9109b0e 359void f2fs_flush_merged_writes(struct f2fs_sb_info *sbi)
406657dd 360{
b9109b0e
JK		 361 f2fs_submit_merged_write(sbi, DATA);
362 f2fs_submit_merged_write(sbi, NODE);
363 f2fs_submit_merged_write(sbi, META);
406657dd
CY		 364}
365
93dfe2ac
JK		 366/*
367 * Fill the locked page with data located in the block address.
771a9a71 368 * A caller needs to unlock the page on failure.
93dfe2ac 369 */
05ca3632 370int f2fs_submit_page_bio(struct f2fs_io_info *fio)
93dfe2ac 371{
93dfe2ac 372 struct bio *bio;
0b81d077
JK		 373 struct page *page = fio->encrypted_page ?
374 fio->encrypted_page : fio->page;
93dfe2ac 375
2ace38e0 376 trace_f2fs_submit_page_bio(page, fio);
05ca3632 377 f2fs_trace_ios(fio, 0);
93dfe2ac
JK		 378
379 /* Allocate a new bio */
578c6478
YY		 380 bio = __bio_alloc(fio->sbi, fio->new_blkaddr, fio->io_wbc,
381 1, is_read_io(fio->op));
93dfe2ac 382
09cbfeaf 383 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
93dfe2ac 384 bio_put(bio);
93dfe2ac
JK		 385 return -EFAULT;
386 }
04d328de 387 bio_set_op_attrs(bio, fio->op, fio->op_flags);
93dfe2ac 388
4fc29c1a 389 __submit_bio(fio->sbi, bio, fio->type);
d1b3e72d
JK		 390
391 if (!is_read_io(fio->op))
392 inc_page_count(fio->sbi, WB_DATA_TYPE(fio->page));
93dfe2ac
JK		 393 return 0;
394}
395
b9109b0e 396int f2fs_submit_page_write(struct f2fs_io_info *fio)
93dfe2ac 397{
05ca3632 398 struct f2fs_sb_info *sbi = fio->sbi;
458e6197 399 enum page_type btype = PAGE_TYPE_OF_BIO(fio->type);
a912b54d 400 struct f2fs_bio_info *io = sbi->write_io[btype] + fio->temp;
4375a336 401 struct page *bio_page;
0a595eba 402 int err = 0;
93dfe2ac 403
b9109b0e 404 f2fs_bug_on(sbi, is_read_io(fio->op));
93dfe2ac 405
fb830fc5
CY		 406 down_write(&io->io_rwsem);
407next:
408 if (fio->in_list) {
409 spin_lock(&io->io_lock);
410 if (list_empty(&io->io_list)) {
411 spin_unlock(&io->io_lock);
412 goto out_fail;
413 }
414 fio = list_first_entry(&io->io_list,
415 struct f2fs_io_info, list);
416 list_del(&fio->list);
417 spin_unlock(&io->io_lock);
418 }
93dfe2ac 419
7a9d7548
CY		 420 if (fio->old_blkaddr != NEW_ADDR)
421 verify_block_addr(sbi, fio->old_blkaddr);
422 verify_block_addr(sbi, fio->new_blkaddr);
93dfe2ac 423
36951b38
CY		 424 bio_page = fio->encrypted_page ? fio->encrypted_page : fio->page;
425
ebf7c522
TM		 426 /* set submitted = true as a return value */
427 fio->submitted = true;
d68f735b 428
b9109b0e 429 inc_page_count(sbi, WB_DATA_TYPE(bio_page));
93dfe2ac 430
7a9d7548 431 if (io->bio && (io->last_block_in_bio != fio->new_blkaddr - 1 ||
3c62be17
JK		 432 (io->fio.op != fio->op || io->fio.op_flags != fio->op_flags) ||
433 !__same_bdev(sbi, fio->new_blkaddr, io->bio)))
458e6197 434 __submit_merged_bio(io);
93dfe2ac
JK		 435alloc_new:
436 if (io->bio == NULL) {
0a595eba
JK		 437 if ((fio->type == DATA || fio->type == NODE) &&
438 fio->new_blkaddr & F2FS_IO_SIZE_MASK(sbi)) {
439 err = -EAGAIN;
b9109b0e 440 dec_page_count(sbi, WB_DATA_TYPE(bio_page));
0a595eba
JK		 441 goto out_fail;
442 }
578c6478 443 io->bio = __bio_alloc(sbi, fio->new_blkaddr, fio->io_wbc,
b9109b0e 444 BIO_MAX_PAGES, false);
458e6197 445 io->fio = *fio;
93dfe2ac
JK		 446 }
447
a912b54d 448 if (bio_add_page(io->bio, bio_page, PAGE_SIZE, 0) < PAGE_SIZE) {
458e6197 449 __submit_merged_bio(io);
93dfe2ac
JK		 450 goto alloc_new;
451 }
452
578c6478
YY		 453 if (fio->io_wbc)
454 wbc_account_io(fio->io_wbc, bio_page, PAGE_SIZE);
455
7a9d7548 456 io->last_block_in_bio = fio->new_blkaddr;
05ca3632 457 f2fs_trace_ios(fio, 0);
fb830fc5
CY		 458
459 trace_f2fs_submit_page_write(fio->page, fio);
460
461 if (fio->in_list)
462 goto next;
0a595eba 463out_fail:
df0f8dc0 464 up_write(&io->io_rwsem);
0a595eba 465 return err;
93dfe2ac
JK		 466}
467
13ba41e3
JK		 468static struct bio *f2fs_grab_read_bio(struct inode *inode, block_t blkaddr,
469 unsigned nr_pages)
470{
471 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
472 struct fscrypt_ctx *ctx = NULL;
473 struct bio *bio;
474
475 if (f2fs_encrypted_file(inode)) {
476 ctx = fscrypt_get_ctx(inode, GFP_NOFS);
477 if (IS_ERR(ctx))
478 return ERR_CAST(ctx);
479
480 /* wait the page to be moved by cleaning */
481 f2fs_wait_on_block_writeback(sbi, blkaddr);
482 }
483
d62fe971 484 bio = f2fs_bio_alloc(sbi, min_t(int, nr_pages, BIO_MAX_PAGES), false);
13ba41e3
JK		 485 if (!bio) {
486 if (ctx)
487 fscrypt_release_ctx(ctx);
488 return ERR_PTR(-ENOMEM);
489 }
490 f2fs_target_device(sbi, blkaddr, bio);
491 bio->bi_end_io = f2fs_read_end_io;
492 bio->bi_private = ctx;
493 bio_set_op_attrs(bio, REQ_OP_READ, 0);
494
495 return bio;
496}
497
498/* This can handle encryption stuffs */
499static int f2fs_submit_page_read(struct inode *inode, struct page *page,
500 block_t blkaddr)
501{
502 struct bio *bio = f2fs_grab_read_bio(inode, blkaddr, 1);
503
504 if (IS_ERR(bio))
505 return PTR_ERR(bio);
506
507 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
508 bio_put(bio);
509 return -EFAULT;
510 }
511 __submit_bio(F2FS_I_SB(inode), bio, DATA);
512 return 0;
513}
514
46008c6d
CY		 515static void __set_data_blkaddr(struct dnode_of_data *dn)
516{
517 struct f2fs_node *rn = F2FS_NODE(dn->node_page);
518 __le32 *addr_array;
7a2af766
CY		 519 int base = 0;
520
521 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
522 base = get_extra_isize(dn->inode);
46008c6d
CY		 523
524 /* Get physical address of data block */
525 addr_array = blkaddr_in_node(rn);
7a2af766 526 addr_array[base + dn->ofs_in_node] = cpu_to_le32(dn->data_blkaddr);
46008c6d
CY		 527}
528
0a8165d7 529/*
eb47b800
JK		 530 * Lock ordering for the change of data block address:
531 * ->data_page
532 * ->node_page
533 * update block addresses in the node page
534 */
216a620a 535void set_data_blkaddr(struct dnode_of_data *dn)
eb47b800 536{
46008c6d
CY		 537 f2fs_wait_on_page_writeback(dn->node_page, NODE, true);
538 __set_data_blkaddr(dn);
539 if (set_page_dirty(dn->node_page))
12719ae1 540 dn->node_changed = true;
eb47b800
JK		 541}
542
f28b3434
CY		 543void f2fs_update_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr)
544{
545 dn->data_blkaddr = blkaddr;
546 set_data_blkaddr(dn);
547 f2fs_update_extent_cache(dn);
548}
549
46008c6d
CY		 550/* dn->ofs_in_node will be returned with up-to-date last block pointer */
551int reserve_new_blocks(struct dnode_of_data *dn, blkcnt_t count)
eb47b800 552{
4081363f 553 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
0abd675e 554 int err;
eb47b800 555
46008c6d
CY		 556 if (!count)
557 return 0;
558
91942321 559 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
eb47b800 560 return -EPERM;
0abd675e
CY		 561 if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
562 return err;
eb47b800 563
46008c6d
CY		 564 trace_f2fs_reserve_new_blocks(dn->inode, dn->nid,
565 dn->ofs_in_node, count);
566
567 f2fs_wait_on_page_writeback(dn->node_page, NODE, true);
568
569 for (; count > 0; dn->ofs_in_node++) {
7a2af766
CY		 570 block_t blkaddr = datablock_addr(dn->inode,
571 dn->node_page, dn->ofs_in_node);
46008c6d
CY		 572 if (blkaddr == NULL_ADDR) {
573 dn->data_blkaddr = NEW_ADDR;
574 __set_data_blkaddr(dn);
575 count--;
576 }
577 }
578
579 if (set_page_dirty(dn->node_page))
580 dn->node_changed = true;
eb47b800
JK		 581 return 0;
582}
583
46008c6d
CY		 584/* Should keep dn->ofs_in_node unchanged */
585int reserve_new_block(struct dnode_of_data *dn)
586{
587 unsigned int ofs_in_node = dn->ofs_in_node;
588 int ret;
589
590 ret = reserve_new_blocks(dn, 1);
591 dn->ofs_in_node = ofs_in_node;
592 return ret;
593}
594
b600965c
HL		 595int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
596{
597 bool need_put = dn->inode_page ? false : true;
598 int err;
599
600 err = get_dnode_of_data(dn, index, ALLOC_NODE);
601 if (err)
602 return err;
a8865372 603
b600965c
HL		 604 if (dn->data_blkaddr == NULL_ADDR)
605 err = reserve_new_block(dn);
a8865372 606 if (err || need_put)
b600965c
HL		 607 f2fs_put_dnode(dn);
608 return err;
609}
610
759af1c9 611int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index)
eb47b800 612{
e15882b6 613 struct extent_info ei = {0,0,0};
759af1c9 614 struct inode *inode = dn->inode;
028a41e8 615
759af1c9
FL		 616 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
617 dn->data_blkaddr = ei.blk + index - ei.fofs;
618 return 0;
429511cd 619 }
028a41e8 620
759af1c9 621 return f2fs_reserve_block(dn, index);
eb47b800
JK		 622}
623
a56c7c6f 624struct page *get_read_data_page(struct inode *inode, pgoff_t index,
04d328de 625 int op_flags, bool for_write)
eb47b800 626{
eb47b800
JK		 627 struct address_space *mapping = inode->i_mapping;
628 struct dnode_of_data dn;
629 struct page *page;
e15882b6 630 struct extent_info ei = {0,0,0};
eb47b800 631 int err;
4375a336 632
a56c7c6f 633 page = f2fs_grab_cache_page(mapping, index, for_write);
650495de
JK		 634 if (!page)
635 return ERR_PTR(-ENOMEM);
636
cb3bc9ee
CY		 637 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
638 dn.data_blkaddr = ei.blk + index - ei.fofs;
639 goto got_it;
640 }
641
eb47b800 642 set_new_dnode(&dn, inode, NULL, NULL, 0);
266e97a8 643 err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
86531d6b
JK		 644 if (err)
645 goto put_err;
eb47b800
JK		 646 f2fs_put_dnode(&dn);
647
6bacf52f 648 if (unlikely(dn.data_blkaddr == NULL_ADDR)) {
86531d6b
JK		 649 err = -ENOENT;
650 goto put_err;
650495de 651 }
cb3bc9ee 652got_it:
43f3eae1
JK		 653 if (PageUptodate(page)) {
654 unlock_page(page);
eb47b800 655 return page;
43f3eae1 656 }
eb47b800 657
d59ff4df
JK		 658 /*
659 * A new dentry page is allocated but not able to be written, since its
660 * new inode page couldn't be allocated due to -ENOSPC.
661 * In such the case, its blkaddr can be remained as NEW_ADDR.
662 * see, f2fs_add_link -> get_new_data_page -> init_inode_metadata.
663 */
664 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 665 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 666 if (!PageUptodate(page))
667 SetPageUptodate(page);
43f3eae1 668 unlock_page(page);
d59ff4df
JK		 669 return page;
670 }
eb47b800 671
13ba41e3 672 err = f2fs_submit_page_read(inode, page, dn.data_blkaddr);
393ff91f 673 if (err)
86531d6b 674 goto put_err;
43f3eae1 675 return page;
86531d6b
JK		 676
677put_err:
678 f2fs_put_page(page, 1);
679 return ERR_PTR(err);
43f3eae1
JK		 680}
681
682struct page *find_data_page(struct inode *inode, pgoff_t index)
683{
684 struct address_space *mapping = inode->i_mapping;
685 struct page *page;
686
687 page = find_get_page(mapping, index);
688 if (page && PageUptodate(page))
689 return page;
690 f2fs_put_page(page, 0);
691
70fd7614 692 page = get_read_data_page(inode, index, 0, false);
43f3eae1
JK		 693 if (IS_ERR(page))
694 return page;
695
696 if (PageUptodate(page))
697 return page;
698
699 wait_on_page_locked(page);
700 if (unlikely(!PageUptodate(page))) {
701 f2fs_put_page(page, 0);
702 return ERR_PTR(-EIO);
703 }
704 return page;
705}
706
707/*
708 * If it tries to access a hole, return an error.
709 * Because, the callers, functions in dir.c and GC, should be able to know
710 * whether this page exists or not.
711 */
a56c7c6f
JK		 712struct page *get_lock_data_page(struct inode *inode, pgoff_t index,
713 bool for_write)
43f3eae1
JK		 714{
715 struct address_space *mapping = inode->i_mapping;
716 struct page *page;
717repeat:
70fd7614 718 page = get_read_data_page(inode, index, 0, for_write);
43f3eae1
JK		 719 if (IS_ERR(page))
720 return page;
393ff91f 721
43f3eae1 722 /* wait for read completion */
393ff91f 723 lock_page(page);
6bacf52f 724 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 725 f2fs_put_page(page, 1);
726 goto repeat;
eb47b800 727 }
1563ac75
CY		 728 if (unlikely(!PageUptodate(page))) {
729 f2fs_put_page(page, 1);
730 return ERR_PTR(-EIO);
731 }
eb47b800
JK		 732 return page;
733}
734
0a8165d7 735/*
eb47b800
JK		 736 * Caller ensures that this data page is never allocated.
737 * A new zero-filled data page is allocated in the page cache.
39936837 738 *
4f4124d0
CY		 739 * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and
740 * f2fs_unlock_op().
470f00e9
CY		 741 * Note that, ipage is set only by make_empty_dir, and if any error occur,
742 * ipage should be released by this function.
eb47b800 743 */
64aa7ed9 744struct page *get_new_data_page(struct inode *inode,
a8865372 745 struct page *ipage, pgoff_t index, bool new_i_size)
eb47b800 746{
eb47b800
JK		 747 struct address_space *mapping = inode->i_mapping;
748 struct page *page;
749 struct dnode_of_data dn;
750 int err;
7612118a 751
a56c7c6f 752 page = f2fs_grab_cache_page(mapping, index, true);
470f00e9
CY		 753 if (!page) {
754 /*
755 * before exiting, we should make sure ipage will be released
756 * if any error occur.
757 */
758 f2fs_put_page(ipage, 1);
01f28610 759 return ERR_PTR(-ENOMEM);
470f00e9 760 }
eb47b800 761
a8865372 762 set_new_dnode(&dn, inode, ipage, NULL, 0);
b600965c 763 err = f2fs_reserve_block(&dn, index);
01f28610
JK		 764 if (err) {
765 f2fs_put_page(page, 1);
eb47b800 766 return ERR_PTR(err);
a8865372 767 }
01f28610
JK		 768 if (!ipage)
769 f2fs_put_dnode(&dn);
eb47b800
JK		 770
771 if (PageUptodate(page))
01f28610 772 goto got_it;
eb47b800
JK		 773
774 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 775 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 776 if (!PageUptodate(page))
777 SetPageUptodate(page);
eb47b800 778 } else {
4375a336 779 f2fs_put_page(page, 1);
a8865372 780
7612118a
JK		 781 /* if ipage exists, blkaddr should be NEW_ADDR */
782 f2fs_bug_on(F2FS_I_SB(inode), ipage);
783 page = get_lock_data_page(inode, index, true);
4375a336 784 if (IS_ERR(page))
7612118a 785 return page;
eb47b800 786 }
01f28610 787got_it:
9edcdabf 788 if (new_i_size && i_size_read(inode) <
ee6d182f 789 ((loff_t)(index + 1) << PAGE_SHIFT))
fc9581c8 790 f2fs_i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT));
eb47b800
JK		 791 return page;
792}
793
d5097be5 794static int __allocate_data_block(struct dnode_of_data *dn, int seg_type)
bfad7c2d 795{
4081363f 796 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
bfad7c2d 797 struct f2fs_summary sum;
bfad7c2d 798 struct node_info ni;
976e4c50 799 pgoff_t fofs;
46008c6d 800 blkcnt_t count = 1;
0abd675e 801 int err;
bfad7c2d 802
91942321 803 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
bfad7c2d 804 return -EPERM;
df6136ef 805
7a2af766
CY		 806 dn->data_blkaddr = datablock_addr(dn->inode,
807 dn->node_page, dn->ofs_in_node);
df6136ef
CY		 808 if (dn->data_blkaddr == NEW_ADDR)
809 goto alloc;
810
0abd675e
CY		 811 if (unlikely((err = inc_valid_block_count(sbi, dn->inode, &count))))
812 return err;
bfad7c2d 813
df6136ef 814alloc:
bfad7c2d
JK		 815 get_node_info(sbi, dn->nid, &ni);
816 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
817
df6136ef 818 allocate_data_block(sbi, NULL, dn->data_blkaddr, &dn->data_blkaddr,
d5097be5 819 &sum, seg_type, NULL, false);
216a620a 820 set_data_blkaddr(dn);
bfad7c2d 821
976e4c50 822 /* update i_size */
81ca7350 823 fofs = start_bidx_of_node(ofs_of_node(dn->node_page), dn->inode) +
976e4c50 824 dn->ofs_in_node;
09cbfeaf 825 if (i_size_read(dn->inode) < ((loff_t)(fofs + 1) << PAGE_SHIFT))
fc9581c8 826 f2fs_i_size_write(dn->inode,
09cbfeaf 827 ((loff_t)(fofs + 1) << PAGE_SHIFT));
bfad7c2d
JK		 828 return 0;
829}
830
c040ff9d
JK		 831static inline bool __force_buffered_io(struct inode *inode, int rw)
832{
1958593e 833 return (f2fs_encrypted_file(inode) ||
c040ff9d
JK		 834 (rw == WRITE && test_opt(F2FS_I_SB(inode), LFS)) ||
835 F2FS_I_SB(inode)->s_ndevs);
836}
837
a7de6086 838int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *from)
59b802e5 839{
b439b103 840 struct inode *inode = file_inode(iocb->ki_filp);
5b8db7fa 841 struct f2fs_map_blocks map;
d6d478a1 842 int flag;
a7de6086 843 int err = 0;
d6d478a1 844 bool direct_io = iocb->ki_flags & IOCB_DIRECT;
59b802e5 845
71ad682c 846 /* convert inline data for Direct I/O*/
d6d478a1 847 if (direct_io) {
71ad682c
WG		 848 err = f2fs_convert_inline_inode(inode);
849 if (err)
850 return err;
851 }
852
dc91de78
JK		 853 if (is_inode_flag_set(inode, FI_NO_PREALLOC))
854 return 0;
855
0080c507 856 map.m_lblk = F2FS_BLK_ALIGN(iocb->ki_pos);
dfd02e4d
CY		 857 map.m_len = F2FS_BYTES_TO_BLK(iocb->ki_pos + iov_iter_count(from));
858 if (map.m_len > map.m_lblk)
859 map.m_len -= map.m_lblk;
860 else
861 map.m_len = 0;
862
da85985c 863 map.m_next_pgofs = NULL;
d5097be5 864 map.m_seg_type = NO_CHECK_TYPE;
2a340760 865
d6d478a1 866 if (direct_io) {
d5097be5 867 map.m_seg_type = rw_hint_to_seg_type(iocb->ki_hint);
d6d478a1
CY		 868 flag = __force_buffered_io(inode, WRITE) ?
869 F2FS_GET_BLOCK_PRE_AIO :
870 F2FS_GET_BLOCK_PRE_DIO;
871 goto map_blocks;
d5097be5 872 }
f2470371 873 if (iocb->ki_pos + iov_iter_count(from) > MAX_INLINE_DATA(inode)) {
a7de6086
JK		 874 err = f2fs_convert_inline_inode(inode);
875 if (err)
876 return err;
b439b103 877 }
d6d478a1 878 if (f2fs_has_inline_data(inode))
25006645 879 return err;
d6d478a1
CY		 880
881 flag = F2FS_GET_BLOCK_PRE_AIO;
882
883map_blocks:
884 err = f2fs_map_blocks(inode, &map, 1, flag);
885 if (map.m_len > 0 && err == -ENOSPC) {
886 if (!direct_io)
887 set_inode_flag(inode, FI_NO_PREALLOC);
888 err = 0;
25006645 889 }
a7de6086 890 return err;
59b802e5
JK		 891}
892
59c9081b
YH		 893static inline void __do_map_lock(struct f2fs_sb_info *sbi, int flag, bool lock)
894{
895 if (flag == F2FS_GET_BLOCK_PRE_AIO) {
896 if (lock)
897 down_read(&sbi->node_change);
898 else
899 up_read(&sbi->node_change);
900 } else {
901 if (lock)
902 f2fs_lock_op(sbi);
903 else
904 f2fs_unlock_op(sbi);
905 }
906}
907
0a8165d7 908/*
003a3e1d
JK		 909 * f2fs_map_blocks() now supported readahead/bmap/rw direct_IO with
910 * f2fs_map_blocks structure.
4f4124d0
CY		 911 * If original data blocks are allocated, then give them to blockdev.
912 * Otherwise,
913 * a. preallocate requested block addresses
914 * b. do not use extent cache for better performance
915 * c. give the block addresses to blockdev
eb47b800 916 */
d323d005 917int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
e2b4e2bc 918 int create, int flag)
eb47b800 919{
003a3e1d 920 unsigned int maxblocks = map->m_len;
eb47b800 921 struct dnode_of_data dn;
f9811703 922 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
ac6f1999 923 int mode = create ? ALLOC_NODE : LOOKUP_NODE;
46008c6d 924 pgoff_t pgofs, end_offset, end;
bfad7c2d 925 int err = 0, ofs = 1;
46008c6d
CY		 926 unsigned int ofs_in_node, last_ofs_in_node;
927 blkcnt_t prealloc;
e15882b6 928 struct extent_info ei = {0,0,0};
7df3a431 929 block_t blkaddr;
eb47b800 930
dfd02e4d
CY		 931 if (!maxblocks)
932 return 0;
933
003a3e1d
JK		 934 map->m_len = 0;
935 map->m_flags = 0;
936
937 /* it only supports block size == page size */
938 pgofs = (pgoff_t)map->m_lblk;
46008c6d 939 end = pgofs + maxblocks;
eb47b800 940
24b84912 941 if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) {
003a3e1d
JK		 942 map->m_pblk = ei.blk + pgofs - ei.fofs;
943 map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs);
944 map->m_flags = F2FS_MAP_MAPPED;
bfad7c2d 945 goto out;
a2e7d1bf 946 }
bfad7c2d 947
4fe71e88 948next_dnode:
59b802e5 949 if (create)
59c9081b 950 __do_map_lock(sbi, flag, true);
eb47b800
JK		 951
952 /* When reading holes, we need its node page */
953 set_new_dnode(&dn, inode, NULL, NULL, 0);
bfad7c2d 954 err = get_dnode_of_data(&dn, pgofs, mode);
1ec79083 955 if (err) {
43473f96
CY		 956 if (flag == F2FS_GET_BLOCK_BMAP)
957 map->m_pblk = 0;
da85985c 958 if (err == -ENOENT) {
bfad7c2d 959 err = 0;
da85985c
CY		 960 if (map->m_next_pgofs)
961 *map->m_next_pgofs =
962 get_next_page_offset(&dn, pgofs);
963 }
bfad7c2d 964 goto unlock_out;
848753aa 965 }
973163fc 966
46008c6d 967 prealloc = 0;
230436b3 968 last_ofs_in_node = ofs_in_node = dn.ofs_in_node;
81ca7350 969 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
4fe71e88
CY		 970
971next_block:
7a2af766 972 blkaddr = datablock_addr(dn.inode, dn.node_page, dn.ofs_in_node);
4fe71e88
CY		 973
974 if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR) {
973163fc 975 if (create) {
f9811703
CY		 976 if (unlikely(f2fs_cp_error(sbi))) {
977 err = -EIO;
4fe71e88 978 goto sync_out;
f9811703 979 }
24b84912 980 if (flag == F2FS_GET_BLOCK_PRE_AIO) {
46008c6d
CY		 981 if (blkaddr == NULL_ADDR) {
982 prealloc++;
983 last_ofs_in_node = dn.ofs_in_node;
984 }
24b84912 985 } else {
d5097be5
HL		 986 err = __allocate_data_block(&dn,
987 map->m_seg_type);
6f2d8ed6 988 if (!err)
91942321 989 set_inode_flag(inode, FI_APPEND_WRITE);
24b84912 990 }
973163fc 991 if (err)
4fe71e88 992 goto sync_out;
3f2be043 993 map->m_flags |= F2FS_MAP_NEW;
4fe71e88 994 blkaddr = dn.data_blkaddr;
973163fc 995 } else {
43473f96
CY		 996 if (flag == F2FS_GET_BLOCK_BMAP) {
997 map->m_pblk = 0;
998 goto sync_out;
999 }
da85985c
CY		 1000 if (flag == F2FS_GET_BLOCK_FIEMAP &&
1001 blkaddr == NULL_ADDR) {
1002 if (map->m_next_pgofs)
1003 *map->m_next_pgofs = pgofs + 1;
4c2ac6a8 1004 goto sync_out;
da85985c 1005 }
4c2ac6a8 1006 if (flag != F2FS_GET_BLOCK_FIEMAP)
4fe71e88 1007 goto sync_out;
e2b4e2bc 1008 }
e2b4e2bc 1009 }
eb47b800 1010
46008c6d
CY		 1011 if (flag == F2FS_GET_BLOCK_PRE_AIO)
1012 goto skip;
1013
4fe71e88
CY		 1014 if (map->m_len == 0) {
1015 /* preallocated unwritten block should be mapped for fiemap. */
1016 if (blkaddr == NEW_ADDR)
1017 map->m_flags |= F2FS_MAP_UNWRITTEN;
1018 map->m_flags |= F2FS_MAP_MAPPED;
1019
1020 map->m_pblk = blkaddr;
1021 map->m_len = 1;
1022 } else if ((map->m_pblk != NEW_ADDR &&
1023 blkaddr == (map->m_pblk + ofs)) ||
b439b103 1024 (map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR) ||
46008c6d 1025 flag == F2FS_GET_BLOCK_PRE_DIO) {
4fe71e88
CY		 1026 ofs++;
1027 map->m_len++;
1028 } else {
1029 goto sync_out;
1030 }
bfad7c2d 1031
46008c6d 1032skip:
bfad7c2d
JK		 1033 dn.ofs_in_node++;
1034 pgofs++;
1035
46008c6d
CY		 1036 /* preallocate blocks in batch for one dnode page */
1037 if (flag == F2FS_GET_BLOCK_PRE_AIO &&
1038 (pgofs == end || dn.ofs_in_node == end_offset)) {
7df3a431 1039
46008c6d
CY		 1040 dn.ofs_in_node = ofs_in_node;
1041 err = reserve_new_blocks(&dn, prealloc);
1042 if (err)
1043 goto sync_out;
bfad7c2d 1044
46008c6d
CY		 1045 map->m_len += dn.ofs_in_node - ofs_in_node;
1046 if (prealloc && dn.ofs_in_node != last_ofs_in_node + 1) {
1047 err = -ENOSPC;
1048 goto sync_out;
3104af35 1049 }
46008c6d
CY		 1050 dn.ofs_in_node = end_offset;
1051 }
1052
1053 if (pgofs >= end)
1054 goto sync_out;
1055 else if (dn.ofs_in_node < end_offset)
1056 goto next_block;
1057
46008c6d
CY		 1058 f2fs_put_dnode(&dn);
1059
1060 if (create) {
59c9081b 1061 __do_map_lock(sbi, flag, false);
6f2d8ed6 1062 f2fs_balance_fs(sbi, dn.node_changed);
eb47b800 1063 }
46008c6d 1064 goto next_dnode;
7df3a431 1065
bfad7c2d 1066sync_out:
eb47b800 1067 f2fs_put_dnode(&dn);
bfad7c2d 1068unlock_out:
2a340760 1069 if (create) {
59c9081b 1070 __do_map_lock(sbi, flag, false);
6f2d8ed6 1071 f2fs_balance_fs(sbi, dn.node_changed);
2a340760 1072 }
bfad7c2d 1073out:
003a3e1d 1074 trace_f2fs_map_blocks(inode, map, err);
bfad7c2d 1075 return err;
eb47b800
JK		 1076}
1077
003a3e1d 1078static int __get_data_block(struct inode *inode, sector_t iblock,
da85985c 1079 struct buffer_head *bh, int create, int flag,
d5097be5 1080 pgoff_t *next_pgofs, int seg_type)
003a3e1d
JK		 1081{
1082 struct f2fs_map_blocks map;
a7de6086 1083 int err;
003a3e1d
JK		 1084
1085 map.m_lblk = iblock;
1086 map.m_len = bh->b_size >> inode->i_blkbits;
da85985c 1087 map.m_next_pgofs = next_pgofs;
d5097be5 1088 map.m_seg_type = seg_type;
003a3e1d 1089
a7de6086
JK		 1090 err = f2fs_map_blocks(inode, &map, create, flag);
1091 if (!err) {
003a3e1d
JK		 1092 map_bh(bh, inode->i_sb, map.m_pblk);
1093 bh->b_state = (bh->b_state & ~F2FS_MAP_FLAGS) | map.m_flags;
b86e3307 1094 bh->b_size = (u64)map.m_len << inode->i_blkbits;
003a3e1d 1095 }
a7de6086 1096 return err;
003a3e1d
JK		 1097}
1098
ccfb3000 1099static int get_data_block(struct inode *inode, sector_t iblock,
da85985c
CY		 1100 struct buffer_head *bh_result, int create, int flag,
1101 pgoff_t *next_pgofs)
e2b4e2bc 1102{
da85985c 1103 return __get_data_block(inode, iblock, bh_result, create,
d5097be5
HL		 1104 flag, next_pgofs,
1105 NO_CHECK_TYPE);
e2b4e2bc
CY		 1106}
1107
1108static int get_data_block_dio(struct inode *inode, sector_t iblock,
ccfb3000
JK		 1109 struct buffer_head *bh_result, int create)
1110{
e2b4e2bc 1111 return __get_data_block(inode, iblock, bh_result, create,
d5097be5
HL		 1112 F2FS_GET_BLOCK_DEFAULT, NULL,
1113 rw_hint_to_seg_type(
1114 inode->i_write_hint));
ccfb3000
JK		 1115}
1116
e2b4e2bc 1117static int get_data_block_bmap(struct inode *inode, sector_t iblock,
ccfb3000
JK		 1118 struct buffer_head *bh_result, int create)
1119{
179448bf 1120 /* Block number less than F2FS MAX BLOCKS */
e0afc4d6 1121 if (unlikely(iblock >= F2FS_I_SB(inode)->max_file_blocks))
179448bf
YH		 1122 return -EFBIG;
1123
e2b4e2bc 1124 return __get_data_block(inode, iblock, bh_result, create,
d5097be5
HL		 1125 F2FS_GET_BLOCK_BMAP, NULL,
1126 NO_CHECK_TYPE);
ccfb3000
JK		 1127}
1128
7f63eb77
JK		 1129static inline sector_t logical_to_blk(struct inode *inode, loff_t offset)
1130{
1131 return (offset >> inode->i_blkbits);
1132}
1133
1134static inline loff_t blk_to_logical(struct inode *inode, sector_t blk)
1135{
1136 return (blk << inode->i_blkbits);
1137}
1138
9ab70134
JK		 1139int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1140 u64 start, u64 len)
1141{
7f63eb77
JK		 1142 struct buffer_head map_bh;
1143 sector_t start_blk, last_blk;
da85985c 1144 pgoff_t next_pgofs;
7f63eb77
JK		 1145 u64 logical = 0, phys = 0, size = 0;
1146 u32 flags = 0;
7f63eb77
JK		 1147 int ret = 0;
1148
1149 ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC);
1150 if (ret)
1151 return ret;
1152
67f8cf3c
JK		 1153 if (f2fs_has_inline_data(inode)) {
1154 ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len);
1155 if (ret != -EAGAIN)
1156 return ret;
1157 }
1158
5955102c 1159 inode_lock(inode);
de1475cc 1160
7f63eb77
JK		 1161 if (logical_to_blk(inode, len) == 0)
1162 len = blk_to_logical(inode, 1);
1163
1164 start_blk = logical_to_blk(inode, start);
1165 last_blk = logical_to_blk(inode, start + len - 1);
9a950d52 1166
7f63eb77
JK		 1167next:
1168 memset(&map_bh, 0, sizeof(struct buffer_head));
1169 map_bh.b_size = len;
1170
e2b4e2bc 1171 ret = get_data_block(inode, start_blk, &map_bh, 0,
da85985c 1172 F2FS_GET_BLOCK_FIEMAP, &next_pgofs);
7f63eb77
JK		 1173 if (ret)
1174 goto out;
1175
1176 /* HOLE */
1177 if (!buffer_mapped(&map_bh)) {
da85985c 1178 start_blk = next_pgofs;
58736fa6
CY		 1179
1180 if (blk_to_logical(inode, start_blk) < blk_to_logical(inode,
1181 F2FS_I_SB(inode)->max_file_blocks))
9a950d52 1182 goto prep_next;
58736fa6 1183
9a950d52
FL		 1184 flags |= FIEMAP_EXTENT_LAST;
1185 }
7f63eb77 1186
da5af127
CY		 1187 if (size) {
1188 if (f2fs_encrypted_inode(inode))
1189 flags |= FIEMAP_EXTENT_DATA_ENCRYPTED;
1190
9a950d52
FL		 1191 ret = fiemap_fill_next_extent(fieinfo, logical,
1192 phys, size, flags);
da5af127 1193 }
7f63eb77 1194
9a950d52
FL		 1195 if (start_blk > last_blk || ret)
1196 goto out;
7f63eb77 1197
9a950d52
FL		 1198 logical = blk_to_logical(inode, start_blk);
1199 phys = blk_to_logical(inode, map_bh.b_blocknr);
1200 size = map_bh.b_size;
1201 flags = 0;
1202 if (buffer_unwritten(&map_bh))
1203 flags = FIEMAP_EXTENT_UNWRITTEN;
7f63eb77 1204
9a950d52 1205 start_blk += logical_to_blk(inode, size);
7f63eb77 1206
9a950d52 1207prep_next:
7f63eb77
JK		 1208 cond_resched();
1209 if (fatal_signal_pending(current))
1210 ret = -EINTR;
1211 else
1212 goto next;
1213out:
1214 if (ret == 1)
1215 ret = 0;
1216
5955102c 1217 inode_unlock(inode);
7f63eb77 1218 return ret;
9ab70134
JK		 1219}
1220
f1e88660
JK		 1221/*
1222 * This function was originally taken from fs/mpage.c, and customized for f2fs.
1223 * Major change was from block_size == page_size in f2fs by default.
1224 */
1225static int f2fs_mpage_readpages(struct address_space *mapping,
1226 struct list_head *pages, struct page *page,
1227 unsigned nr_pages)
1228{
1229 struct bio *bio = NULL;
f1e88660
JK		 1230 sector_t last_block_in_bio = 0;
1231 struct inode *inode = mapping->host;
1232 const unsigned blkbits = inode->i_blkbits;
1233 const unsigned blocksize = 1 << blkbits;
1234 sector_t block_in_file;
1235 sector_t last_block;
1236 sector_t last_block_in_file;
1237 sector_t block_nr;
f1e88660
JK		 1238 struct f2fs_map_blocks map;
1239
1240 map.m_pblk = 0;
1241 map.m_lblk = 0;
1242 map.m_len = 0;
1243 map.m_flags = 0;
da85985c 1244 map.m_next_pgofs = NULL;
d5097be5 1245 map.m_seg_type = NO_CHECK_TYPE;
f1e88660 1246
736c0a74 1247 for (; nr_pages; nr_pages--) {
f1e88660 1248 if (pages) {
939afa94 1249 page = list_last_entry(pages, struct page, lru);
a83d50bc
KM		 1250
1251 prefetchw(&page->flags);
f1e88660
JK		 1252 list_del(&page->lru);
1253 if (add_to_page_cache_lru(page, mapping,
8a5c743e
MH		 1254 page->index,
1255 readahead_gfp_mask(mapping)))
f1e88660
JK		 1256 goto next_page;
1257 }
1258
1259 block_in_file = (sector_t)page->index;
1260 last_block = block_in_file + nr_pages;
1261 last_block_in_file = (i_size_read(inode) + blocksize - 1) >>
1262 blkbits;
1263 if (last_block > last_block_in_file)
1264 last_block = last_block_in_file;
1265
1266 /*
1267 * Map blocks using the previous result first.
1268 */
1269 if ((map.m_flags & F2FS_MAP_MAPPED) &&
1270 block_in_file > map.m_lblk &&
1271 block_in_file < (map.m_lblk + map.m_len))
1272 goto got_it;
1273
1274 /*
1275 * Then do more f2fs_map_blocks() calls until we are
1276 * done with this page.
1277 */
1278 map.m_flags = 0;
1279
1280 if (block_in_file < last_block) {
1281 map.m_lblk = block_in_file;
1282 map.m_len = last_block - block_in_file;
1283
46c9e141 1284 if (f2fs_map_blocks(inode, &map, 0,
f2220c7f 1285 F2FS_GET_BLOCK_DEFAULT))
f1e88660
JK		 1286 goto set_error_page;
1287 }
1288got_it:
1289 if ((map.m_flags & F2FS_MAP_MAPPED)) {
1290 block_nr = map.m_pblk + block_in_file - map.m_lblk;
1291 SetPageMappedToDisk(page);
1292
1293 if (!PageUptodate(page) && !cleancache_get_page(page)) {
1294 SetPageUptodate(page);
1295 goto confused;
1296 }
1297 } else {
09cbfeaf 1298 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 1299 if (!PageUptodate(page))
1300 SetPageUptodate(page);
f1e88660
JK		 1301 unlock_page(page);
1302 goto next_page;
1303 }
1304
1305 /*
1306 * This page will go to BIO. Do we need to send this
1307 * BIO off first?
1308 */
3c62be17
JK		 1309 if (bio && (last_block_in_bio != block_nr - 1 ||
1310 !__same_bdev(F2FS_I_SB(inode), block_nr, bio))) {
f1e88660 1311submit_and_realloc:
4fc29c1a 1312 __submit_bio(F2FS_I_SB(inode), bio, DATA);
f1e88660
JK		 1313 bio = NULL;
1314 }
1315 if (bio == NULL) {
13ba41e3 1316 bio = f2fs_grab_read_bio(inode, block_nr, nr_pages);
1d353eb7
JK		 1317 if (IS_ERR(bio)) {
1318 bio = NULL;
f1e88660 1319 goto set_error_page;
4375a336 1320 }
f1e88660
JK		 1321 }
1322
1323 if (bio_add_page(bio, page, blocksize, 0) < blocksize)
1324 goto submit_and_realloc;
1325
1326 last_block_in_bio = block_nr;
1327 goto next_page;
1328set_error_page:
1329 SetPageError(page);
09cbfeaf 1330 zero_user_segment(page, 0, PAGE_SIZE);
f1e88660
JK		 1331 unlock_page(page);
1332 goto next_page;
1333confused:
1334 if (bio) {
4fc29c1a 1335 __submit_bio(F2FS_I_SB(inode), bio, DATA);
f1e88660
JK		 1336 bio = NULL;
1337 }
1338 unlock_page(page);
1339next_page:
1340 if (pages)
09cbfeaf 1341 put_page(page);
f1e88660
JK		 1342 }
1343 BUG_ON(pages && !list_empty(pages));
1344 if (bio)
4fc29c1a 1345 __submit_bio(F2FS_I_SB(inode), bio, DATA);
f1e88660
JK		 1346 return 0;
1347}
1348
eb47b800
JK		 1349static int f2fs_read_data_page(struct file *file, struct page *page)
1350{
9ffe0fb5 1351 struct inode *inode = page->mapping->host;
b3d208f9 1352 int ret = -EAGAIN;
9ffe0fb5 1353
c20e89cd
CY		 1354 trace_f2fs_readpage(page, DATA);
1355
e1c42045 1356 /* If the file has inline data, try to read it directly */
9ffe0fb5
HL		 1357 if (f2fs_has_inline_data(inode))
1358 ret = f2fs_read_inline_data(inode, page);
b3d208f9 1359 if (ret == -EAGAIN)
f1e88660 1360 ret = f2fs_mpage_readpages(page->mapping, NULL, page, 1);
9ffe0fb5 1361 return ret;
eb47b800
JK		 1362}
1363
1364static int f2fs_read_data_pages(struct file *file,
1365 struct address_space *mapping,
1366 struct list_head *pages, unsigned nr_pages)
1367{
71cb4aff 1368 struct inode *inode = mapping->host;
939afa94 1369 struct page *page = list_last_entry(pages, struct page, lru);
b8c29400
CY		 1370
1371 trace_f2fs_readpages(inode, page, nr_pages);
9ffe0fb5
HL		 1372
1373 /* If the file has inline data, skip readpages */
1374 if (f2fs_has_inline_data(inode))
1375 return 0;
1376
f1e88660 1377 return f2fs_mpage_readpages(mapping, pages, NULL, nr_pages);
eb47b800
JK		 1378}
1379
7eab0c0d
HP		 1380static int encrypt_one_page(struct f2fs_io_info *fio)
1381{
1382 struct inode *inode = fio->page->mapping->host;
1383 gfp_t gfp_flags = GFP_NOFS;
1384
1958593e 1385 if (!f2fs_encrypted_file(inode))
7eab0c0d
HP		 1386 return 0;
1387
1388 /* wait for GCed encrypted page writeback */
d4c759ee 1389 f2fs_wait_on_block_writeback(fio->sbi, fio->old_blkaddr);
7eab0c0d
HP		 1390
1391retry_encrypt:
1392 fio->encrypted_page = fscrypt_encrypt_page(inode, fio->page,
1393 PAGE_SIZE, 0, fio->page->index, gfp_flags);
1394 if (!IS_ERR(fio->encrypted_page))
1395 return 0;
1396
1397 /* flush pending IOs and wait for a while in the ENOMEM case */
1398 if (PTR_ERR(fio->encrypted_page) == -ENOMEM) {
b9109b0e 1399 f2fs_flush_merged_writes(fio->sbi);
7eab0c0d
HP		 1400 congestion_wait(BLK_RW_ASYNC, HZ/50);
1401 gfp_flags |= __GFP_NOFAIL;
1402 goto retry_encrypt;
1403 }
1404 return PTR_ERR(fio->encrypted_page);
1405}
1406
1407static inline bool need_inplace_update(struct f2fs_io_info *fio)
1408{
1409 struct inode *inode = fio->page->mapping->host;
1410
7eab0c0d
HP		 1411 if (S_ISDIR(inode->i_mode) || f2fs_is_atomic_file(inode))
1412 return false;
1413 if (is_cold_data(fio->page))
1414 return false;
1415 if (IS_ATOMIC_WRITTEN_PAGE(fio->page))
1416 return false;
1417
1418 return need_inplace_update_policy(inode, fio);
1419}
1420
a817737e
JK		 1421static inline bool valid_ipu_blkaddr(struct f2fs_io_info *fio)
1422{
1423 if (fio->old_blkaddr == NEW_ADDR)
1424 return false;
1425 if (fio->old_blkaddr == NULL_ADDR)
1426 return false;
1427 return true;
1428}
1429
05ca3632 1430int do_write_data_page(struct f2fs_io_info *fio)
eb47b800 1431{
05ca3632 1432 struct page *page = fio->page;
eb47b800 1433 struct inode *inode = page->mapping->host;
eb47b800 1434 struct dnode_of_data dn;
e959c8f5
HP		 1435 struct extent_info ei = {0,0,0};
1436 bool ipu_force = false;
eb47b800
JK		 1437 int err = 0;
1438
1439 set_new_dnode(&dn, inode, NULL, NULL, 0);
e959c8f5
HP		 1440 if (need_inplace_update(fio) &&
1441 f2fs_lookup_extent_cache(inode, page->index, &ei)) {
1442 fio->old_blkaddr = ei.blk + page->index - ei.fofs;
a817737e
JK		 1443
1444 if (valid_ipu_blkaddr(fio)) {
e959c8f5 1445 ipu_force = true;
cc15620b 1446 fio->need_lock = LOCK_DONE;
e959c8f5
HP		 1447 goto got_it;
1448 }
1449 }
279d6df2 1450
d29460e5
JK		 1451 /* Deadlock due to between page->lock and f2fs_lock_op */
1452 if (fio->need_lock == LOCK_REQ && !f2fs_trylock_op(fio->sbi))
1453 return -EAGAIN;
279d6df2 1454
266e97a8 1455 err = get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
eb47b800 1456 if (err)
279d6df2 1457 goto out;
eb47b800 1458
28bc106b 1459 fio->old_blkaddr = dn.data_blkaddr;
eb47b800
JK		 1460
1461 /* This page is already truncated */
7a9d7548 1462 if (fio->old_blkaddr == NULL_ADDR) {
2bca1e23 1463 ClearPageUptodate(page);
eb47b800 1464 goto out_writepage;
2bca1e23 1465 }
e959c8f5 1466got_it:
eb47b800
JK		 1467 /*
1468 * If current allocation needs SSR,
1469 * it had better in-place writes for updated data.
1470 */
a817737e 1471 if (ipu_force || (valid_ipu_blkaddr(fio) && need_inplace_update(fio))) {
cc15620b
JK		 1472 err = encrypt_one_page(fio);
1473 if (err)
1474 goto out_writepage;
1475
1476 set_page_writeback(page);
279d6df2 1477 f2fs_put_dnode(&dn);
cc15620b 1478 if (fio->need_lock == LOCK_REQ)
279d6df2 1479 f2fs_unlock_op(fio->sbi);
d1b3e72d 1480 err = rewrite_data_page(fio);
7eab0c0d 1481 trace_f2fs_do_write_data_page(fio->page, IPU);
91942321 1482 set_inode_flag(inode, FI_UPDATE_WRITE);
279d6df2 1483 return err;
eb47b800 1484 }
279d6df2 1485
cc15620b
JK		 1486 if (fio->need_lock == LOCK_RETRY) {
1487 if (!f2fs_trylock_op(fio->sbi)) {
1488 err = -EAGAIN;
1489 goto out_writepage;
1490 }
1491 fio->need_lock = LOCK_REQ;
1492 }
1493
1494 err = encrypt_one_page(fio);
1495 if (err)
1496 goto out_writepage;
1497
1498 set_page_writeback(page);
1499
279d6df2
HP		 1500 /* LFS mode write path */
1501 write_data_page(&dn, fio);
1502 trace_f2fs_do_write_data_page(page, OPU);
1503 set_inode_flag(inode, FI_APPEND_WRITE);
1504 if (page->index == 0)
1505 set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
eb47b800
JK		 1506out_writepage:
1507 f2fs_put_dnode(&dn);
279d6df2 1508out:
cc15620b 1509 if (fio->need_lock == LOCK_REQ)
279d6df2 1510 f2fs_unlock_op(fio->sbi);
eb47b800
JK		 1511 return err;
1512}
1513
d68f735b 1514static int __write_data_page(struct page *page, bool *submitted,
b0af6d49
CY		 1515 struct writeback_control *wbc,
1516 enum iostat_type io_type)
eb47b800
JK		 1517{
1518 struct inode *inode = page->mapping->host;
4081363f 1519 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
eb47b800
JK		 1520 loff_t i_size = i_size_read(inode);
1521 const pgoff_t end_index = ((unsigned long long) i_size)
09cbfeaf 1522 >> PAGE_SHIFT;
26de9b11 1523 loff_t psize = (page->index + 1) << PAGE_SHIFT;
9ffe0fb5 1524 unsigned offset = 0;
39936837 1525 bool need_balance_fs = false;
eb47b800 1526 int err = 0;
458e6197 1527 struct f2fs_io_info fio = {
05ca3632 1528 .sbi = sbi,
39d787be 1529 .ino = inode->i_ino,
458e6197 1530 .type = DATA,
04d328de 1531 .op = REQ_OP_WRITE,
7637241e 1532 .op_flags = wbc_to_write_flags(wbc),
e959c8f5 1533 .old_blkaddr = NULL_ADDR,
05ca3632 1534 .page = page,
4375a336 1535 .encrypted_page = NULL,
d68f735b 1536 .submitted = false,
cc15620b 1537 .need_lock = LOCK_RETRY,
b0af6d49 1538 .io_type = io_type,
578c6478 1539 .io_wbc = wbc,
458e6197 1540 };
eb47b800 1541
ecda0de3
CY		 1542 trace_f2fs_writepage(page, DATA);
1543
0771fcc7
CY		 1544 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1545 goto redirty_out;
1546
eb47b800 1547 if (page->index < end_index)
39936837 1548 goto write;
eb47b800
JK		 1549
1550 /*
1551 * If the offset is out-of-range of file size,
1552 * this page does not have to be written to disk.
1553 */
09cbfeaf 1554 offset = i_size & (PAGE_SIZE - 1);
76f60268 1555 if ((page->index >= end_index + 1) || !offset)
39936837 1556 goto out;
eb47b800 1557
09cbfeaf 1558 zero_user_segment(page, offset, PAGE_SIZE);
39936837 1559write:
1e84371f
JK		 1560 if (f2fs_is_drop_cache(inode))
1561 goto out;
e6e5f561
JK		 1562 /* we should not write 0'th page having journal header */
1563 if (f2fs_is_volatile_file(inode) && (!page->index ||
1564 (!wbc->for_reclaim &&
1565 available_free_memory(sbi, BASE_CHECK))))
1e84371f 1566 goto redirty_out;
eb47b800 1567
cf779cab
JK		 1568 /* we should bypass data pages to proceed the kworkder jobs */
1569 if (unlikely(f2fs_cp_error(sbi))) {
7f319975 1570 mapping_set_error(page->mapping, -EIO);
a7ffdbe2 1571 goto out;
cf779cab
JK		 1572 }
1573
39936837 1574 /* Dentry blocks are controlled by checkpoint */
eb47b800 1575 if (S_ISDIR(inode->i_mode)) {
cc15620b 1576 fio.need_lock = LOCK_DONE;
05ca3632 1577 err = do_write_data_page(&fio);
8618b881
JK		 1578 goto done;
1579 }
9ffe0fb5 1580
8618b881 1581 if (!wbc->for_reclaim)
39936837 1582 need_balance_fs = true;
7f3037a5 1583 else if (has_not_enough_free_secs(sbi, 0, 0))
39936837 1584 goto redirty_out;
ef095d19
JK		 1585 else
1586 set_inode_flag(inode, FI_HOT_DATA);
eb47b800 1587
b3d208f9 1588 err = -EAGAIN;
dd7b2333 1589 if (f2fs_has_inline_data(inode)) {
b3d208f9 1590 err = f2fs_write_inline_data(inode, page);
dd7b2333
YH		 1591 if (!err)
1592 goto out;
1593 }
279d6df2 1594
cc15620b 1595 if (err == -EAGAIN) {
05ca3632 1596 err = do_write_data_page(&fio);
cc15620b
JK		 1597 if (err == -EAGAIN) {
1598 fio.need_lock = LOCK_REQ;
1599 err = do_write_data_page(&fio);
1600 }
1601 }
a0d00fad
CY		 1602
1603 down_write(&F2FS_I(inode)->i_sem);
26de9b11
JK		 1604 if (F2FS_I(inode)->last_disk_size < psize)
1605 F2FS_I(inode)->last_disk_size = psize;
a0d00fad 1606 up_write(&F2FS_I(inode)->i_sem);
279d6df2 1607
8618b881
JK		 1608done:
1609 if (err && err != -ENOENT)
1610 goto redirty_out;
eb47b800 1611
39936837 1612out:
a7ffdbe2 1613 inode_dec_dirty_pages(inode);
2bca1e23
JK		 1614 if (err)
1615 ClearPageUptodate(page);
0c3a5797
CY		 1616
1617 if (wbc->for_reclaim) {
b9109b0e 1618 f2fs_submit_merged_write_cond(sbi, inode, 0, page->index, DATA);
ef095d19 1619 clear_inode_flag(inode, FI_HOT_DATA);
0c3a5797 1620 remove_dirty_inode(inode);
d68f735b 1621 submitted = NULL;
0c3a5797
CY		 1622 }
1623
eb47b800 1624 unlock_page(page);
a7881893
JK		 1625 if (!S_ISDIR(inode->i_mode))
1626 f2fs_balance_fs(sbi, need_balance_fs);
0c3a5797 1627
d68f735b 1628 if (unlikely(f2fs_cp_error(sbi))) {
b9109b0e 1629 f2fs_submit_merged_write(sbi, DATA);
d68f735b
JK		 1630 submitted = NULL;
1631 }
1632
1633 if (submitted)
1634 *submitted = fio.submitted;
0c3a5797 1635
eb47b800
JK		 1636 return 0;
1637
eb47b800 1638redirty_out:
76f60268 1639 redirty_page_for_writepage(wbc, page);
0002b61b
CY		 1640 if (!err)
1641 return AOP_WRITEPAGE_ACTIVATE;
b230e6ca
JK		 1642 unlock_page(page);
1643 return err;
fa9150a8
NJ		 1644}
1645
f566bae8
JK		 1646static int f2fs_write_data_page(struct page *page,
1647 struct writeback_control *wbc)
1648{
b0af6d49 1649 return __write_data_page(page, NULL, wbc, FS_DATA_IO);
f566bae8
JK		 1650}
1651
8f46dcae
CY		 1652/*
1653 * This function was copied from write_cche_pages from mm/page-writeback.c.
1654 * The major change is making write step of cold data page separately from
1655 * warm/hot data page.
1656 */
1657static int f2fs_write_cache_pages(struct address_space *mapping,
b0af6d49
CY		 1658 struct writeback_control *wbc,
1659 enum iostat_type io_type)
8f46dcae
CY		 1660{
1661 int ret = 0;
1662 int done = 0;
1663 struct pagevec pvec;
1664 int nr_pages;
1665 pgoff_t uninitialized_var(writeback_index);
1666 pgoff_t index;
1667 pgoff_t end; /* Inclusive */
1668 pgoff_t done_index;
942fd319 1669 pgoff_t last_idx = ULONG_MAX;
8f46dcae
CY		 1670 int cycled;
1671 int range_whole = 0;
1672 int tag;
8f46dcae 1673
86679820 1674 pagevec_init(&pvec);
46ae957f 1675
ef095d19
JK		 1676 if (get_dirty_pages(mapping->host) <=
1677 SM_I(F2FS_M_SB(mapping))->min_hot_blocks)
1678 set_inode_flag(mapping->host, FI_HOT_DATA);
1679 else
1680 clear_inode_flag(mapping->host, FI_HOT_DATA);
1681
8f46dcae
CY		 1682 if (wbc->range_cyclic) {
1683 writeback_index = mapping->writeback_index; /* prev offset */
1684 index = writeback_index;
1685 if (index == 0)
1686 cycled = 1;
1687 else
1688 cycled = 0;
1689 end = -1;
1690 } else {
09cbfeaf
KS		 1691 index = wbc->range_start >> PAGE_SHIFT;
1692 end = wbc->range_end >> PAGE_SHIFT;
8f46dcae
CY		 1693 if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
1694 range_whole = 1;
1695 cycled = 1; /* ignore range_cyclic tests */
1696 }
1697 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
1698 tag = PAGECACHE_TAG_TOWRITE;
1699 else
1700 tag = PAGECACHE_TAG_DIRTY;
1701retry:
1702 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
1703 tag_pages_for_writeback(mapping, index, end);
1704 done_index = index;
1705 while (!done && (index <= end)) {
1706 int i;
1707
69c4f35d 1708 nr_pages = pagevec_lookup_range_tag(&pvec, mapping, &index, end,
67fd707f 1709 tag);
8f46dcae
CY		 1710 if (nr_pages == 0)
1711 break;
1712
1713 for (i = 0; i < nr_pages; i++) {
1714 struct page *page = pvec.pages[i];
d68f735b 1715 bool submitted = false;
8f46dcae 1716
8f46dcae 1717 done_index = page->index;
d29460e5 1718retry_write:
8f46dcae
CY		 1719 lock_page(page);
1720
1721 if (unlikely(page->mapping != mapping)) {
1722continue_unlock:
1723 unlock_page(page);
1724 continue;
1725 }
1726
1727 if (!PageDirty(page)) {
1728 /* someone wrote it for us */
1729 goto continue_unlock;
1730 }
1731
8f46dcae
CY		 1732 if (PageWriteback(page)) {
1733 if (wbc->sync_mode != WB_SYNC_NONE)
fec1d657
JK		 1734 f2fs_wait_on_page_writeback(page,
1735 DATA, true);
8f46dcae
CY		 1736 else
1737 goto continue_unlock;
1738 }
1739
1740 BUG_ON(PageWriteback(page));
1741 if (!clear_page_dirty_for_io(page))
1742 goto continue_unlock;
1743
b0af6d49 1744 ret = __write_data_page(page, &submitted, wbc, io_type);
8f46dcae 1745 if (unlikely(ret)) {
0002b61b
CY		 1746 /*
1747 * keep nr_to_write, since vfs uses this to
1748 * get # of written pages.
1749 */
1750 if (ret == AOP_WRITEPAGE_ACTIVATE) {
1751 unlock_page(page);
1752 ret = 0;
1753 continue;
d29460e5
JK		 1754 } else if (ret == -EAGAIN) {
1755 ret = 0;
1756 if (wbc->sync_mode == WB_SYNC_ALL) {
1757 cond_resched();
1758 congestion_wait(BLK_RW_ASYNC,
1759 HZ/50);
1760 goto retry_write;
1761 }
1762 continue;
0002b61b 1763 }
b230e6ca
JK		 1764 done_index = page->index + 1;
1765 done = 1;
1766 break;
d68f735b 1767 } else if (submitted) {
942fd319 1768 last_idx = page->index;
8f46dcae
CY		 1769 }
1770
687de7f1
JK		 1771 /* give a priority to WB_SYNC threads */
1772 if ((atomic_read(&F2FS_M_SB(mapping)->wb_sync_req) ||
1773 --wbc->nr_to_write <= 0) &&
1774 wbc->sync_mode == WB_SYNC_NONE) {
8f46dcae
CY		 1775 done = 1;
1776 break;
1777 }
1778 }
1779 pagevec_release(&pvec);
1780 cond_resched();
1781 }
1782
8f46dcae
CY		 1783 if (!cycled && !done) {
1784 cycled = 1;
1785 index = 0;
1786 end = writeback_index - 1;
1787 goto retry;
1788 }
1789 if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
1790 mapping->writeback_index = done_index;
1791
942fd319 1792 if (last_idx != ULONG_MAX)
b9109b0e
JK		 1793 f2fs_submit_merged_write_cond(F2FS_M_SB(mapping), mapping->host,
1794 0, last_idx, DATA);
6ca56ca4 1795
8f46dcae
CY		 1796 return ret;
1797}
1798
b0af6d49
CY		 1799int __f2fs_write_data_pages(struct address_space *mapping,
1800 struct writeback_control *wbc,
1801 enum iostat_type io_type)
eb47b800
JK		 1802{
1803 struct inode *inode = mapping->host;
4081363f 1804 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
9dfa1baf 1805 struct blk_plug plug;
eb47b800 1806 int ret;
eb47b800 1807
cfb185a1 1808 /* deal with chardevs and other special file */
1809 if (!mapping->a_ops->writepage)
1810 return 0;
1811
6a290544
CY		 1812 /* skip writing if there is no dirty page in this inode */
1813 if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE)
1814 return 0;
1815
0771fcc7
CY		 1816 /* during POR, we don't need to trigger writepage at all. */
1817 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1818 goto skip_write;
1819
a1257023
JK		 1820 if (S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_NONE &&
1821 get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) &&
1822 available_free_memory(sbi, DIRTY_DENTS))
1823 goto skip_write;
1824
d323d005 1825 /* skip writing during file defragment */
91942321 1826 if (is_inode_flag_set(inode, FI_DO_DEFRAG))
d323d005
CY		 1827 goto skip_write;
1828
d31c7c3f
YH		 1829 trace_f2fs_writepages(mapping->host, wbc, DATA);
1830
687de7f1
JK		 1831 /* to avoid spliting IOs due to mixed WB_SYNC_ALL and WB_SYNC_NONE */
1832 if (wbc->sync_mode == WB_SYNC_ALL)
1833 atomic_inc(&sbi->wb_sync_req);
1834 else if (atomic_read(&sbi->wb_sync_req))
1835 goto skip_write;
1836
9dfa1baf 1837 blk_start_plug(&plug);
b0af6d49 1838 ret = f2fs_write_cache_pages(mapping, wbc, io_type);
9dfa1baf 1839 blk_finish_plug(&plug);
687de7f1
JK		 1840
1841 if (wbc->sync_mode == WB_SYNC_ALL)
1842 atomic_dec(&sbi->wb_sync_req);
28ea6162
JK		 1843 /*
1844 * if some pages were truncated, we cannot guarantee its mapping->host
1845 * to detect pending bios.
1846 */
458e6197 1847
c227f912 1848 remove_dirty_inode(inode);
eb47b800 1849 return ret;
d3baf95d
JK		 1850
1851skip_write:
a7ffdbe2 1852 wbc->pages_skipped += get_dirty_pages(inode);
d31c7c3f 1853 trace_f2fs_writepages(mapping->host, wbc, DATA);
d3baf95d 1854 return 0;
eb47b800
JK		 1855}
1856
b0af6d49
CY		 1857static int f2fs_write_data_pages(struct address_space *mapping,
1858 struct writeback_control *wbc)
1859{
1860 struct inode *inode = mapping->host;
1861
1862 return __f2fs_write_data_pages(mapping, wbc,
1863 F2FS_I(inode)->cp_task == current ?
1864 FS_CP_DATA_IO : FS_DATA_IO);
1865}
1866
3aab8f82
CY		 1867static void f2fs_write_failed(struct address_space *mapping, loff_t to)
1868{
1869 struct inode *inode = mapping->host;
819d9153 1870 loff_t i_size = i_size_read(inode);
3aab8f82 1871
819d9153 1872 if (to > i_size) {
5a3a2d83 1873 down_write(&F2FS_I(inode)->i_mmap_sem);
819d9153
JK		 1874 truncate_pagecache(inode, i_size);
1875 truncate_blocks(inode, i_size, true);
5a3a2d83 1876 up_write(&F2FS_I(inode)->i_mmap_sem);
3aab8f82
CY		 1877 }
1878}
1879
2aadac08
JK		 1880static int prepare_write_begin(struct f2fs_sb_info *sbi,
1881 struct page *page, loff_t pos, unsigned len,
1882 block_t *blk_addr, bool *node_changed)
1883{
1884 struct inode *inode = page->mapping->host;
1885 pgoff_t index = page->index;
1886 struct dnode_of_data dn;
1887 struct page *ipage;
b4d07a3e 1888 bool locked = false;
e15882b6 1889 struct extent_info ei = {0,0,0};
2aadac08
JK		 1890 int err = 0;
1891
24b84912
JK		 1892 /*
1893 * we already allocated all the blocks, so we don't need to get
1894 * the block addresses when there is no need to fill the page.
1895 */
dc91de78
JK		 1896 if (!f2fs_has_inline_data(inode) && len == PAGE_SIZE &&
1897 !is_inode_flag_set(inode, FI_NO_PREALLOC))
24b84912
JK		 1898 return 0;
1899
b4d07a3e 1900 if (f2fs_has_inline_data(inode) ||
09cbfeaf 1901 (pos & PAGE_MASK) >= i_size_read(inode)) {
59c9081b 1902 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
b4d07a3e
JK		 1903 locked = true;
1904 }
1905restart:
2aadac08
JK		 1906 /* check inline_data */
1907 ipage = get_node_page(sbi, inode->i_ino);
1908 if (IS_ERR(ipage)) {
1909 err = PTR_ERR(ipage);
1910 goto unlock_out;
1911 }
1912
1913 set_new_dnode(&dn, inode, ipage, ipage, 0);
1914
1915 if (f2fs_has_inline_data(inode)) {
f2470371 1916 if (pos + len <= MAX_INLINE_DATA(inode)) {
2aadac08 1917 read_inline_data(page, ipage);
91942321 1918 set_inode_flag(inode, FI_DATA_EXIST);
ab47036d
CY		 1919 if (inode->i_nlink)
1920 set_inline_node(ipage);
2aadac08
JK		 1921 } else {
1922 err = f2fs_convert_inline_page(&dn, page);
1923 if (err)
b4d07a3e
JK		 1924 goto out;
1925 if (dn.data_blkaddr == NULL_ADDR)
1926 err = f2fs_get_block(&dn, index);
1927 }
1928 } else if (locked) {
1929 err = f2fs_get_block(&dn, index);
1930 } else {
1931 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
1932 dn.data_blkaddr = ei.blk + index - ei.fofs;
1933 } else {
b4d07a3e
JK		 1934 /* hole case */
1935 err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
4da7bf5a 1936 if (err || dn.data_blkaddr == NULL_ADDR) {
b4d07a3e 1937 f2fs_put_dnode(&dn);
59c9081b
YH		 1938 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO,
1939 true);
b4d07a3e
JK		 1940 locked = true;
1941 goto restart;
1942 }
2aadac08
JK		 1943 }
1944 }
b4d07a3e 1945
2aadac08
JK		 1946 /* convert_inline_page can make node_changed */
1947 *blk_addr = dn.data_blkaddr;
1948 *node_changed = dn.node_changed;
b4d07a3e 1949out:
2aadac08
JK		 1950 f2fs_put_dnode(&dn);
1951unlock_out:
b4d07a3e 1952 if (locked)
59c9081b 1953 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
2aadac08
JK		 1954 return err;
1955}
1956
eb47b800
JK		 1957static int f2fs_write_begin(struct file *file, struct address_space *mapping,
1958 loff_t pos, unsigned len, unsigned flags,
1959 struct page **pagep, void **fsdata)
1960{
1961 struct inode *inode = mapping->host;
4081363f 1962 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
86531d6b 1963 struct page *page = NULL;
09cbfeaf 1964 pgoff_t index = ((unsigned long long) pos) >> PAGE_SHIFT;
2aadac08
JK		 1965 bool need_balance = false;
1966 block_t blkaddr = NULL_ADDR;
eb47b800
JK		 1967 int err = 0;
1968
62aed044
CY		 1969 trace_f2fs_write_begin(inode, pos, len, flags);
1970
57864ae5
JK		 1971 if (f2fs_is_atomic_file(inode) &&
1972 !available_free_memory(sbi, INMEM_PAGES)) {
1973 err = -ENOMEM;
1974 goto fail;
1975 }
1976
5f727395
JK		 1977 /*
1978 * We should check this at this moment to avoid deadlock on inode page
1979 * and #0 page. The locking rule for inline_data conversion should be:
1980 * lock_page(page #0) -> lock_page(inode_page)
1981 */
1982 if (index != 0) {
1983 err = f2fs_convert_inline_inode(inode);
1984 if (err)
1985 goto fail;
1986 }
afcb7ca0 1987repeat:
86d54795
JK		 1988 /*
1989 * Do not use grab_cache_page_write_begin() to avoid deadlock due to
1990 * wait_for_stable_page. Will wait that below with our IO control.
1991 */
01eccef7 1992 page = f2fs_pagecache_get_page(mapping, index,
86d54795 1993 FGP_LOCK | FGP_WRITE | FGP_CREAT, GFP_NOFS);
3aab8f82
CY		 1994 if (!page) {
1995 err = -ENOMEM;
1996 goto fail;
1997 }
d5f66990 1998
eb47b800
JK		 1999 *pagep = page;
2000
2aadac08
JK		 2001 err = prepare_write_begin(sbi, page, pos, len,
2002 &blkaddr, &need_balance);
9ba69cf9 2003 if (err)
2aadac08 2004 goto fail;
9ba69cf9 2005
7f3037a5 2006 if (need_balance && has_not_enough_free_secs(sbi, 0, 0)) {
2a340760 2007 unlock_page(page);
2c4db1a6 2008 f2fs_balance_fs(sbi, true);
2a340760
JK		 2009 lock_page(page);
2010 if (page->mapping != mapping) {
2011 /* The page got truncated from under us */
2012 f2fs_put_page(page, 1);
2013 goto repeat;
2014 }
2015 }
2016
fec1d657 2017 f2fs_wait_on_page_writeback(page, DATA, false);
b3d208f9 2018
08b39fbd 2019 /* wait for GCed encrypted page writeback */
1958593e 2020 if (f2fs_encrypted_file(inode))
d4c759ee 2021 f2fs_wait_on_block_writeback(sbi, blkaddr);
08b39fbd 2022
649d7df2
JK		 2023 if (len == PAGE_SIZE || PageUptodate(page))
2024 return 0;
eb47b800 2025
746e2403
YH		 2026 if (!(pos & (PAGE_SIZE - 1)) && (pos + len) >= i_size_read(inode)) {
2027 zero_user_segment(page, len, PAGE_SIZE);
2028 return 0;
2029 }
2030
2aadac08 2031 if (blkaddr == NEW_ADDR) {
09cbfeaf 2032 zero_user_segment(page, 0, PAGE_SIZE);
649d7df2 2033 SetPageUptodate(page);
eb47b800 2034 } else {
13ba41e3
JK		 2035 err = f2fs_submit_page_read(inode, page, blkaddr);
2036 if (err)
78682f79 2037 goto fail;
d54c795b 2038
393ff91f 2039 lock_page(page);
6bacf52f 2040 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 2041 f2fs_put_page(page, 1);
2042 goto repeat;
eb47b800 2043 }
1563ac75
CY		 2044 if (unlikely(!PageUptodate(page))) {
2045 err = -EIO;
2046 goto fail;
4375a336 2047 }
eb47b800 2048 }
eb47b800 2049 return 0;
9ba69cf9 2050
3aab8f82 2051fail:
86531d6b 2052 f2fs_put_page(page, 1);
3aab8f82 2053 f2fs_write_failed(mapping, pos + len);
57864ae5
JK		 2054 if (f2fs_is_atomic_file(inode))
2055 drop_inmem_pages_all(sbi);
3aab8f82 2056 return err;
eb47b800
JK		 2057}
2058
a1dd3c13
JK		 2059static int f2fs_write_end(struct file *file,
2060 struct address_space *mapping,
2061 loff_t pos, unsigned len, unsigned copied,
2062 struct page *page, void *fsdata)
2063{
2064 struct inode *inode = page->mapping->host;
2065
dfb2bf38
CY		 2066 trace_f2fs_write_end(inode, pos, len, copied);
2067
649d7df2
JK		 2068 /*
2069 * This should be come from len == PAGE_SIZE, and we expect copied
2070 * should be PAGE_SIZE. Otherwise, we treat it with zero copied and
2071 * let generic_perform_write() try to copy data again through copied=0.
2072 */
2073 if (!PageUptodate(page)) {
746e2403 2074 if (unlikely(copied != len))
649d7df2
JK		 2075 copied = 0;
2076 else
2077 SetPageUptodate(page);
2078 }
2079 if (!copied)
2080 goto unlock_out;
2081
34ba94ba 2082 set_page_dirty(page);
a1dd3c13 2083
fc9581c8
JK		 2084 if (pos + copied > i_size_read(inode))
2085 f2fs_i_size_write(inode, pos + copied);
649d7df2 2086unlock_out:
3024c9a1 2087 f2fs_put_page(page, 1);
d0239e1b 2088 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
a1dd3c13
JK		 2089 return copied;
2090}
2091
6f673763
OS		 2092static int check_direct_IO(struct inode *inode, struct iov_iter *iter,
2093 loff_t offset)
944fcfc1
JK		 2094{
2095 unsigned blocksize_mask = inode->i_sb->s_blocksize - 1;
944fcfc1 2096
944fcfc1
JK		 2097 if (offset & blocksize_mask)
2098 return -EINVAL;
2099
5b46f25d
AV		 2100 if (iov_iter_alignment(iter) & blocksize_mask)
2101 return -EINVAL;
2102
944fcfc1
JK		 2103 return 0;
2104}
2105
c8b8e32d 2106static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
eb47b800 2107{
b439b103 2108 struct address_space *mapping = iocb->ki_filp->f_mapping;
3aab8f82
CY		 2109 struct inode *inode = mapping->host;
2110 size_t count = iov_iter_count(iter);
c8b8e32d 2111 loff_t offset = iocb->ki_pos;
82e0a5aa 2112 int rw = iov_iter_rw(iter);
3aab8f82 2113 int err;
944fcfc1 2114
b439b103 2115 err = check_direct_IO(inode, iter, offset);
b9d777b8
JK		 2116 if (err)
2117 return err;
9ffe0fb5 2118
c040ff9d 2119 if (__force_buffered_io(inode, rw))
36abef4e 2120 return 0;
fcc85a4d 2121
5302fb00 2122 trace_f2fs_direct_IO_enter(inode, offset, count, rw);
70407fad 2123
82e0a5aa 2124 down_read(&F2FS_I(inode)->dio_rwsem[rw]);
c8b8e32d 2125 err = blockdev_direct_IO(iocb, inode, iter, get_data_block_dio);
82e0a5aa
CY		 2126 up_read(&F2FS_I(inode)->dio_rwsem[rw]);
2127
2128 if (rw == WRITE) {
b0af6d49
CY		 2129 if (err > 0) {
2130 f2fs_update_iostat(F2FS_I_SB(inode), APP_DIRECT_IO,
2131 err);
91942321 2132 set_inode_flag(inode, FI_UPDATE_WRITE);
b0af6d49 2133 } else if (err < 0) {
6bfc4919 2134 f2fs_write_failed(mapping, offset + count);
b0af6d49 2135 }
6bfc4919 2136 }
70407fad 2137
5302fb00 2138 trace_f2fs_direct_IO_exit(inode, offset, count, rw, err);
70407fad 2139
3aab8f82 2140 return err;
eb47b800
JK		 2141}
2142
487261f3
CY		 2143void f2fs_invalidate_page(struct page *page, unsigned int offset,
2144 unsigned int length)
eb47b800
JK		 2145{
2146 struct inode *inode = page->mapping->host;
487261f3 2147 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
a7ffdbe2 2148
487261f3 2149 if (inode->i_ino >= F2FS_ROOT_INO(sbi) &&
09cbfeaf 2150 (offset % PAGE_SIZE || length != PAGE_SIZE))
a7ffdbe2
JK		 2151 return;
2152
487261f3 2153 if (PageDirty(page)) {
933439c8 2154 if (inode->i_ino == F2FS_META_INO(sbi)) {
487261f3 2155 dec_page_count(sbi, F2FS_DIRTY_META);
933439c8 2156 } else if (inode->i_ino == F2FS_NODE_INO(sbi)) {
487261f3 2157 dec_page_count(sbi, F2FS_DIRTY_NODES);
933439c8 2158 } else {
487261f3 2159 inode_dec_dirty_pages(inode);
933439c8
CY		 2160 remove_dirty_inode(inode);
2161 }
487261f3 2162 }
decd36b6
CY		 2163
2164 /* This is atomic written page, keep Private */
2165 if (IS_ATOMIC_WRITTEN_PAGE(page))
8c242db9 2166 return drop_inmem_page(inode, page);
decd36b6 2167
23dc974e 2168 set_page_private(page, 0);
eb47b800
JK		 2169 ClearPagePrivate(page);
2170}
2171
487261f3 2172int f2fs_release_page(struct page *page, gfp_t wait)
eb47b800 2173{
f68daeeb
JK		 2174 /* If this is dirty page, keep PagePrivate */
2175 if (PageDirty(page))
2176 return 0;
2177
decd36b6
CY		 2178 /* This is atomic written page, keep Private */
2179 if (IS_ATOMIC_WRITTEN_PAGE(page))
2180 return 0;
2181
23dc974e 2182 set_page_private(page, 0);
eb47b800 2183 ClearPagePrivate(page);
c3850aa1 2184 return 1;
eb47b800
JK		 2185}
2186
fe76b796
JK		 2187/*
2188 * This was copied from __set_page_dirty_buffers which gives higher performance
2189 * in very high speed storages. (e.g., pmem)
2190 */
2191void f2fs_set_page_dirty_nobuffers(struct page *page)
2192{
2193 struct address_space *mapping = page->mapping;
2194 unsigned long flags;
2195
2196 if (unlikely(!mapping))
2197 return;
2198
2199 spin_lock(&mapping->private_lock);
2200 lock_page_memcg(page);
2201 SetPageDirty(page);
2202 spin_unlock(&mapping->private_lock);
2203
2204 spin_lock_irqsave(&mapping->tree_lock, flags);
2205 WARN_ON_ONCE(!PageUptodate(page));
2206 account_page_dirtied(page, mapping);
2207 radix_tree_tag_set(&mapping->page_tree,
2208 page_index(page), PAGECACHE_TAG_DIRTY);
2209 spin_unlock_irqrestore(&mapping->tree_lock, flags);
2210 unlock_page_memcg(page);
2211
2212 __mark_inode_dirty(mapping->host, I_DIRTY_PAGES);
2213 return;
2214}
2215
eb47b800
JK		 2216static int f2fs_set_data_page_dirty(struct page *page)
2217{
2218 struct address_space *mapping = page->mapping;
2219 struct inode *inode = mapping->host;
2220
26c6b887
JK		 2221 trace_f2fs_set_page_dirty(page, DATA);
2222
237c0790
JK		 2223 if (!PageUptodate(page))
2224 SetPageUptodate(page);
34ba94ba 2225
5fe45743 2226 if (f2fs_is_atomic_file(inode) && !f2fs_is_commit_atomic_write(inode)) {
decd36b6
CY		 2227 if (!IS_ATOMIC_WRITTEN_PAGE(page)) {
2228 register_inmem_page(inode, page);
2229 return 1;
2230 }
2231 /*
2232 * Previously, this page has been registered, we just
2233 * return here.
2234 */
2235 return 0;
34ba94ba
JK		 2236 }
2237
eb47b800 2238 if (!PageDirty(page)) {
fe76b796 2239 f2fs_set_page_dirty_nobuffers(page);
a7ffdbe2 2240 update_dirty_page(inode, page);
eb47b800
JK		 2241 return 1;
2242 }
2243 return 0;
2244}
2245
c01e54b7
JK		 2246static sector_t f2fs_bmap(struct address_space *mapping, sector_t block)
2247{
454ae7e5
CY		 2248 struct inode *inode = mapping->host;
2249
1d373a0e
JK		 2250 if (f2fs_has_inline_data(inode))
2251 return 0;
2252
2253 /* make sure allocating whole blocks */
2254 if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
2255 filemap_write_and_wait(mapping);
2256
e2b4e2bc 2257 return generic_block_bmap(mapping, block, get_data_block_bmap);
429511cd
CY		 2258}
2259
5b7a487c
WG		 2260#ifdef CONFIG_MIGRATION
2261#include <linux/migrate.h>
2262
2263int f2fs_migrate_page(struct address_space *mapping,
2264 struct page *newpage, struct page *page, enum migrate_mode mode)
2265{
2266 int rc, extra_count;
2267 struct f2fs_inode_info *fi = F2FS_I(mapping->host);
2268 bool atomic_written = IS_ATOMIC_WRITTEN_PAGE(page);
2269
2270 BUG_ON(PageWriteback(page));
2271
2272 /* migrating an atomic written page is safe with the inmem_lock hold */
ff1048e7
JK		 2273 if (atomic_written) {
2274 if (mode != MIGRATE_SYNC)
2275 return -EBUSY;
2276 if (!mutex_trylock(&fi->inmem_lock))
2277 return -EAGAIN;
2278 }
5b7a487c
WG		 2279
2280 /*
2281 * A reference is expected if PagePrivate set when move mapping,
2282 * however F2FS breaks this for maintaining dirty page counts when
2283 * truncating pages. So here adjusting the 'extra_count' make it work.
2284 */
2285 extra_count = (atomic_written ? 1 : 0) - page_has_private(page);
2286 rc = migrate_page_move_mapping(mapping, newpage,
2287 page, NULL, mode, extra_count);
2288 if (rc != MIGRATEPAGE_SUCCESS) {
2289 if (atomic_written)
2290 mutex_unlock(&fi->inmem_lock);
2291 return rc;
2292 }
2293
2294 if (atomic_written) {
2295 struct inmem_pages *cur;
2296 list_for_each_entry(cur, &fi->inmem_pages, list)
2297 if (cur->page == page) {
2298 cur->page = newpage;
2299 break;
2300 }
2301 mutex_unlock(&fi->inmem_lock);
2302 put_page(page);
2303 get_page(newpage);
2304 }
2305
2306 if (PagePrivate(page))
2307 SetPagePrivate(newpage);
2308 set_page_private(newpage, page_private(page));
2309
2916ecc0
JG		 2310 if (mode != MIGRATE_SYNC_NO_COPY)
2311 migrate_page_copy(newpage, page);
2312 else
2313 migrate_page_states(newpage, page);
5b7a487c
WG		 2314
2315 return MIGRATEPAGE_SUCCESS;
2316}
2317#endif
2318
eb47b800
JK		 2319const struct address_space_operations f2fs_dblock_aops = {
2320 .readpage = f2fs_read_data_page,
2321 .readpages = f2fs_read_data_pages,
2322 .writepage = f2fs_write_data_page,
2323 .writepages = f2fs_write_data_pages,
2324 .write_begin = f2fs_write_begin,
a1dd3c13 2325 .write_end = f2fs_write_end,
eb47b800 2326 .set_page_dirty = f2fs_set_data_page_dirty,
487261f3
CY		 2327 .invalidatepage = f2fs_invalidate_page,
2328 .releasepage = f2fs_release_page,
eb47b800 2329 .direct_IO = f2fs_direct_IO,
c01e54b7 2330 .bmap = f2fs_bmap,
5b7a487c
WG		 2331#ifdef CONFIG_MIGRATION
2332 .migratepage = f2fs_migrate_page,
2333#endif
eb47b800 2334};
Michael's Linux tree.