]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/f2fs/data.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: remove unsafe bitmap checking
[mirror_ubuntu-artful-kernel.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>
eb47b800
JK		 25
26#include "f2fs.h"
27#include "node.h"
28#include "segment.h"
db9f7c1a 29#include "trace.h"
848753aa 30#include <trace/events/f2fs.h>
eb47b800 31
36951b38
CY		 32static bool __is_cp_guaranteed(struct page *page)
33{
34 struct address_space *mapping = page->mapping;
35 struct inode *inode;
36 struct f2fs_sb_info *sbi;
37
38 if (!mapping)
39 return false;
40
41 inode = mapping->host;
42 sbi = F2FS_I_SB(inode);
43
44 if (inode->i_ino == F2FS_META_INO(sbi) ||
45 inode->i_ino == F2FS_NODE_INO(sbi) ||
46 S_ISDIR(inode->i_mode) ||
47 is_cold_data(page))
48 return true;
49 return false;
50}
51
4246a0b6 52static void f2fs_read_end_io(struct bio *bio)
93dfe2ac 53{
f568849e
LT		 54 struct bio_vec *bvec;
55 int i;
93dfe2ac 56
8b038c70 57#ifdef CONFIG_F2FS_FAULT_INJECTION
1ecc0c5c 58 if (time_to_inject(F2FS_P_SB(bio->bi_io_vec->bv_page), FAULT_IO))
8b038c70
CY		 59 bio->bi_error = -EIO;
60#endif
61
4375a336 62 if (f2fs_bio_encrypted(bio)) {
4246a0b6 63 if (bio->bi_error) {
0b81d077 64 fscrypt_release_ctx(bio->bi_private);
4375a336 65 } else {
0b81d077 66 fscrypt_decrypt_bio_pages(bio->bi_private, bio);
4375a336
JK		 67 return;
68 }
69 }
70
12377024
CY		 71 bio_for_each_segment_all(bvec, bio, i) {
72 struct page *page = bvec->bv_page;
f1e88660 73
4246a0b6 74 if (!bio->bi_error) {
237c0790
JK		 75 if (!PageUptodate(page))
76 SetPageUptodate(page);
f1e88660
JK		 77 } else {
78 ClearPageUptodate(page);
79 SetPageError(page);
80 }
81 unlock_page(page);
82 }
f1e88660
JK		 83 bio_put(bio);
84}
85
4246a0b6 86static void f2fs_write_end_io(struct bio *bio)
93dfe2ac 87{
1b1f559f 88 struct f2fs_sb_info *sbi = bio->bi_private;
f568849e
LT		 89 struct bio_vec *bvec;
90 int i;
93dfe2ac 91
f568849e 92 bio_for_each_segment_all(bvec, bio, i) {
93dfe2ac 93 struct page *page = bvec->bv_page;
36951b38 94 enum count_type type = WB_DATA_TYPE(page);
93dfe2ac 95
0a595eba
JK		 96 if (IS_DUMMY_WRITTEN_PAGE(page)) {
97 set_page_private(page, (unsigned long)NULL);
98 ClearPagePrivate(page);
99 unlock_page(page);
100 mempool_free(page, sbi->write_io_dummy);
101
102 if (unlikely(bio->bi_error))
103 f2fs_stop_checkpoint(sbi, true);
104 continue;
105 }
106
0b81d077 107 fscrypt_pullback_bio_page(&page, true);
4375a336 108
4246a0b6 109 if (unlikely(bio->bi_error)) {
5114a97a 110 mapping_set_error(page->mapping, -EIO);
38f91ca8 111 f2fs_stop_checkpoint(sbi, true);
93dfe2ac 112 }
36951b38
CY		 113 dec_page_count(sbi, type);
114 clear_cold_data(page);
93dfe2ac 115 end_page_writeback(page);
f568849e 116 }
36951b38 117 if (!get_pages(sbi, F2FS_WB_CP_DATA) &&
f5730184 118 wq_has_sleeper(&sbi->cp_wait))
93dfe2ac
JK		 119 wake_up(&sbi->cp_wait);
120
121 bio_put(bio);
122}
123
3c62be17
JK		 124/*
125 * Return true, if pre_bio's bdev is same as its target device.
126 */
127struct block_device *f2fs_target_device(struct f2fs_sb_info *sbi,
128 block_t blk_addr, struct bio *bio)
129{
130 struct block_device *bdev = sbi->sb->s_bdev;
131 int i;
132
133 for (i = 0; i < sbi->s_ndevs; i++) {
134 if (FDEV(i).start_blk <= blk_addr &&
135 FDEV(i).end_blk >= blk_addr) {
136 blk_addr -= FDEV(i).start_blk;
137 bdev = FDEV(i).bdev;
138 break;
139 }
140 }
141 if (bio) {
142 bio->bi_bdev = bdev;
143 bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr);
144 }
145 return bdev;
146}
147
148int f2fs_target_device_index(struct f2fs_sb_info *sbi, block_t blkaddr)
149{
150 int i;
151
152 for (i = 0; i < sbi->s_ndevs; i++)
153 if (FDEV(i).start_blk <= blkaddr && FDEV(i).end_blk >= blkaddr)
154 return i;
155 return 0;
156}
157
158static bool __same_bdev(struct f2fs_sb_info *sbi,
159 block_t blk_addr, struct bio *bio)
160{
161 return f2fs_target_device(sbi, blk_addr, NULL) == bio->bi_bdev;
162}
163
940a6d34
GZ		 164/*
165 * Low-level block read/write IO operations.
166 */
167static struct bio *__bio_alloc(struct f2fs_sb_info *sbi, block_t blk_addr,
168 int npages, bool is_read)
169{
170 struct bio *bio;
171
740432f8 172 bio = f2fs_bio_alloc(npages);
940a6d34 173
3c62be17 174 f2fs_target_device(sbi, blk_addr, bio);
940a6d34 175 bio->bi_end_io = is_read ? f2fs_read_end_io : f2fs_write_end_io;
12377024 176 bio->bi_private = is_read ? NULL : sbi;
940a6d34
GZ		 177
178 return bio;
179}
180
4fc29c1a
LT		 181static inline void __submit_bio(struct f2fs_sb_info *sbi,
182 struct bio *bio, enum page_type type)
f5730184 183{
4fc29c1a 184 if (!is_read_io(bio_op(bio))) {
0a595eba
JK		 185 unsigned int start;
186
0bfd7a09 187 if (f2fs_sb_mounted_blkzoned(sbi->sb) &&
52763a4b 188 current->plug && (type == DATA || type == NODE))
19a5f5e2 189 blk_finish_plug(current->plug);
0a595eba
JK		 190
191 if (type != DATA && type != NODE)
192 goto submit_io;
193
194 start = bio->bi_iter.bi_size >> F2FS_BLKSIZE_BITS;
195 start %= F2FS_IO_SIZE(sbi);
196
197 if (start == 0)
198 goto submit_io;
199
200 /* fill dummy pages */
201 for (; start < F2FS_IO_SIZE(sbi); start++) {
202 struct page *page =
203 mempool_alloc(sbi->write_io_dummy,
204 GFP_NOIO | __GFP_ZERO | __GFP_NOFAIL);
205 f2fs_bug_on(sbi, !page);
206
207 SetPagePrivate(page);
208 set_page_private(page, (unsigned long)DUMMY_WRITTEN_PAGE);
209 lock_page(page);
210 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE)
211 f2fs_bug_on(sbi, 1);
212 }
213 /*
214 * In the NODE case, we lose next block address chain. So, we
215 * need to do checkpoint in f2fs_sync_file.
216 */
217 if (type == NODE)
218 set_sbi_flag(sbi, SBI_NEED_CP);
19a5f5e2 219 }
0a595eba 220submit_io:
554b5125
JK		 221 if (is_read_io(bio_op(bio)))
222 trace_f2fs_submit_read_bio(sbi->sb, type, bio);
223 else
224 trace_f2fs_submit_write_bio(sbi->sb, type, bio);
4e49ea4a 225 submit_bio(bio);
f5730184
JK		 226}
227
458e6197 228static void __submit_merged_bio(struct f2fs_bio_info *io)
93dfe2ac 229{
458e6197 230 struct f2fs_io_info *fio = &io->fio;
93dfe2ac
JK		 231
232 if (!io->bio)
233 return;
234
554b5125
JK		 235 bio_set_op_attrs(io->bio, fio->op, fio->op_flags);
236
04d328de 237 if (is_read_io(fio->op))
554b5125 238 trace_f2fs_prepare_read_bio(io->sbi->sb, fio->type, io->bio);
6a8f8ca5 239 else
554b5125 240 trace_f2fs_prepare_write_bio(io->sbi->sb, fio->type, io->bio);
04d328de 241
4fc29c1a 242 __submit_bio(io->sbi, io->bio, fio->type);
93dfe2ac
JK		 243 io->bio = NULL;
244}
245
942fd319
JK		 246static bool __has_merged_page(struct f2fs_bio_info *io,
247 struct inode *inode, nid_t ino, pgoff_t idx)
0fd785eb 248{
0fd785eb
CY		 249 struct bio_vec *bvec;
250 struct page *target;
251 int i;
252
0c3a5797 253 if (!io->bio)
0fd785eb 254 return false;
0c3a5797 255
942fd319 256 if (!inode && !ino)
0c3a5797 257 return true;
0fd785eb
CY		 258
259 bio_for_each_segment_all(bvec, io->bio, i) {
260
0b81d077 261 if (bvec->bv_page->mapping)
0fd785eb 262 target = bvec->bv_page;
0b81d077
JK		 263 else
264 target = fscrypt_control_page(bvec->bv_page);
0fd785eb 265
942fd319
JK		 266 if (idx != target->index)
267 continue;
268
0c3a5797
CY		 269 if (inode && inode == target->mapping->host)
270 return true;
0c3a5797 271 if (ino && ino == ino_of_node(target))
0fd785eb 272 return true;
0fd785eb
CY		 273 }
274
0fd785eb
CY		 275 return false;
276}
277
0c3a5797 278static bool has_merged_page(struct f2fs_sb_info *sbi, struct inode *inode,
942fd319 279 nid_t ino, pgoff_t idx, enum page_type type)
0c3a5797
CY		 280{
281 enum page_type btype = PAGE_TYPE_OF_BIO(type);
282 struct f2fs_bio_info *io = &sbi->write_io[btype];
283 bool ret;
284
285 down_read(&io->io_rwsem);
942fd319 286 ret = __has_merged_page(io, inode, ino, idx);
0c3a5797
CY		 287 up_read(&io->io_rwsem);
288 return ret;
289}
290
291static void __f2fs_submit_merged_bio(struct f2fs_sb_info *sbi,
942fd319
JK		 292 struct inode *inode, nid_t ino, pgoff_t idx,
293 enum page_type type, int rw)
93dfe2ac
JK		 294{
295 enum page_type btype = PAGE_TYPE_OF_BIO(type);
296 struct f2fs_bio_info *io;
297
298 io = is_read_io(rw) ? &sbi->read_io : &sbi->write_io[btype];
299
df0f8dc0 300 down_write(&io->io_rwsem);
458e6197 301
942fd319 302 if (!__has_merged_page(io, inode, ino, idx))
0c3a5797
CY		 303 goto out;
304
458e6197
JK		 305 /* change META to META_FLUSH in the checkpoint procedure */
306 if (type >= META_FLUSH) {
307 io->fio.type = META_FLUSH;
04d328de 308 io->fio.op = REQ_OP_WRITE;
7f54f51f 309 io->fio.op_flags = REQ_META | REQ_PRIO;
70fd7614 310 if (!test_opt(sbi, NOBARRIER))
7f54f51f 311 io->fio.op_flags |= REQ_PREFLUSH | REQ_FUA;
458e6197
JK		 312 }
313 __submit_merged_bio(io);
0c3a5797 314out:
df0f8dc0 315 up_write(&io->io_rwsem);
93dfe2ac
JK		 316}
317
0c3a5797
CY		 318void f2fs_submit_merged_bio(struct f2fs_sb_info *sbi, enum page_type type,
319 int rw)
320{
942fd319 321 __f2fs_submit_merged_bio(sbi, NULL, 0, 0, type, rw);
0c3a5797
CY		 322}
323
324void f2fs_submit_merged_bio_cond(struct f2fs_sb_info *sbi,
942fd319
JK		 325 struct inode *inode, nid_t ino, pgoff_t idx,
326 enum page_type type, int rw)
0c3a5797 327{
942fd319
JK		 328 if (has_merged_page(sbi, inode, ino, idx, type))
329 __f2fs_submit_merged_bio(sbi, inode, ino, idx, type, rw);
0c3a5797
CY		 330}
331
406657dd
CY		 332void f2fs_flush_merged_bios(struct f2fs_sb_info *sbi)
333{
334 f2fs_submit_merged_bio(sbi, DATA, WRITE);
335 f2fs_submit_merged_bio(sbi, NODE, WRITE);
336 f2fs_submit_merged_bio(sbi, META, WRITE);
337}
338
93dfe2ac
JK		 339/*
340 * Fill the locked page with data located in the block address.
341 * Return unlocked page.
342 */
05ca3632 343int f2fs_submit_page_bio(struct f2fs_io_info *fio)
93dfe2ac 344{
93dfe2ac 345 struct bio *bio;
0b81d077
JK		 346 struct page *page = fio->encrypted_page ?
347 fio->encrypted_page : fio->page;
93dfe2ac 348
2ace38e0 349 trace_f2fs_submit_page_bio(page, fio);
05ca3632 350 f2fs_trace_ios(fio, 0);
93dfe2ac
JK		 351
352 /* Allocate a new bio */
04d328de 353 bio = __bio_alloc(fio->sbi, fio->new_blkaddr, 1, is_read_io(fio->op));
93dfe2ac 354
09cbfeaf 355 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
93dfe2ac 356 bio_put(bio);
93dfe2ac
JK		 357 return -EFAULT;
358 }
04d328de 359 bio_set_op_attrs(bio, fio->op, fio->op_flags);
93dfe2ac 360
4fc29c1a 361 __submit_bio(fio->sbi, bio, fio->type);
93dfe2ac
JK		 362 return 0;
363}
364
0a595eba 365int f2fs_submit_page_mbio(struct f2fs_io_info *fio)
93dfe2ac 366{
05ca3632 367 struct f2fs_sb_info *sbi = fio->sbi;
458e6197 368 enum page_type btype = PAGE_TYPE_OF_BIO(fio->type);
93dfe2ac 369 struct f2fs_bio_info *io;
04d328de 370 bool is_read = is_read_io(fio->op);
4375a336 371 struct page *bio_page;
0a595eba 372 int err = 0;
93dfe2ac 373
940a6d34 374 io = is_read ? &sbi->read_io : &sbi->write_io[btype];
93dfe2ac 375
7a9d7548
CY		 376 if (fio->old_blkaddr != NEW_ADDR)
377 verify_block_addr(sbi, fio->old_blkaddr);
378 verify_block_addr(sbi, fio->new_blkaddr);
93dfe2ac 379
36951b38
CY		 380 bio_page = fio->encrypted_page ? fio->encrypted_page : fio->page;
381
d68f735b
JK		 382 /* set submitted = 1 as a return value */
383 fio->submitted = 1;
384
36951b38
CY		 385 if (!is_read)
386 inc_page_count(sbi, WB_DATA_TYPE(bio_page));
387
df0f8dc0 388 down_write(&io->io_rwsem);
93dfe2ac 389
7a9d7548 390 if (io->bio && (io->last_block_in_bio != fio->new_blkaddr - 1 ||
3c62be17
JK		 391 (io->fio.op != fio->op || io->fio.op_flags != fio->op_flags) ||
392 !__same_bdev(sbi, fio->new_blkaddr, io->bio)))
458e6197 393 __submit_merged_bio(io);
93dfe2ac
JK		 394alloc_new:
395 if (io->bio == NULL) {
0a595eba
JK		 396 if ((fio->type == DATA || fio->type == NODE) &&
397 fio->new_blkaddr & F2FS_IO_SIZE_MASK(sbi)) {
398 err = -EAGAIN;
399 dec_page_count(sbi, WB_DATA_TYPE(bio_page));
400 goto out_fail;
401 }
7a9d7548 402 io->bio = __bio_alloc(sbi, fio->new_blkaddr,
664ba972 403 BIO_MAX_PAGES, is_read);
458e6197 404 io->fio = *fio;
93dfe2ac
JK		 405 }
406
09cbfeaf
KS		 407 if (bio_add_page(io->bio, bio_page, PAGE_SIZE, 0) <
408 PAGE_SIZE) {
458e6197 409 __submit_merged_bio(io);
93dfe2ac
JK		 410 goto alloc_new;
411 }
412
7a9d7548 413 io->last_block_in_bio = fio->new_blkaddr;
05ca3632 414 f2fs_trace_ios(fio, 0);
0a595eba 415out_fail:
df0f8dc0 416 up_write(&io->io_rwsem);
05ca3632 417 trace_f2fs_submit_page_mbio(fio->page, fio);
0a595eba 418 return err;
93dfe2ac
JK		 419}
420
46008c6d
CY		 421static void __set_data_blkaddr(struct dnode_of_data *dn)
422{
423 struct f2fs_node *rn = F2FS_NODE(dn->node_page);
424 __le32 *addr_array;
425
426 /* Get physical address of data block */
427 addr_array = blkaddr_in_node(rn);
428 addr_array[dn->ofs_in_node] = cpu_to_le32(dn->data_blkaddr);
429}
430
0a8165d7 431/*
eb47b800
JK		 432 * Lock ordering for the change of data block address:
433 * ->data_page
434 * ->node_page
435 * update block addresses in the node page
436 */
216a620a 437void set_data_blkaddr(struct dnode_of_data *dn)
eb47b800 438{
46008c6d
CY		 439 f2fs_wait_on_page_writeback(dn->node_page, NODE, true);
440 __set_data_blkaddr(dn);
441 if (set_page_dirty(dn->node_page))
12719ae1 442 dn->node_changed = true;
eb47b800
JK		 443}
444
f28b3434
CY		 445void f2fs_update_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr)
446{
447 dn->data_blkaddr = blkaddr;
448 set_data_blkaddr(dn);
449 f2fs_update_extent_cache(dn);
450}
451
46008c6d
CY		 452/* dn->ofs_in_node will be returned with up-to-date last block pointer */
453int reserve_new_blocks(struct dnode_of_data *dn, blkcnt_t count)
eb47b800 454{
4081363f 455 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
eb47b800 456
46008c6d
CY		 457 if (!count)
458 return 0;
459
91942321 460 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
eb47b800 461 return -EPERM;
46008c6d 462 if (unlikely(!inc_valid_block_count(sbi, dn->inode, &count)))
eb47b800
JK		 463 return -ENOSPC;
464
46008c6d
CY		 465 trace_f2fs_reserve_new_blocks(dn->inode, dn->nid,
466 dn->ofs_in_node, count);
467
468 f2fs_wait_on_page_writeback(dn->node_page, NODE, true);
469
470 for (; count > 0; dn->ofs_in_node++) {
471 block_t blkaddr =
472 datablock_addr(dn->node_page, dn->ofs_in_node);
473 if (blkaddr == NULL_ADDR) {
474 dn->data_blkaddr = NEW_ADDR;
475 __set_data_blkaddr(dn);
476 count--;
477 }
478 }
479
480 if (set_page_dirty(dn->node_page))
481 dn->node_changed = true;
eb47b800
JK		 482 return 0;
483}
484
46008c6d
CY		 485/* Should keep dn->ofs_in_node unchanged */
486int reserve_new_block(struct dnode_of_data *dn)
487{
488 unsigned int ofs_in_node = dn->ofs_in_node;
489 int ret;
490
491 ret = reserve_new_blocks(dn, 1);
492 dn->ofs_in_node = ofs_in_node;
493 return ret;
494}
495
b600965c
HL		 496int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
497{
498 bool need_put = dn->inode_page ? false : true;
499 int err;
500
501 err = get_dnode_of_data(dn, index, ALLOC_NODE);
502 if (err)
503 return err;
a8865372 504
b600965c
HL		 505 if (dn->data_blkaddr == NULL_ADDR)
506 err = reserve_new_block(dn);
a8865372 507 if (err || need_put)
b600965c
HL		 508 f2fs_put_dnode(dn);
509 return err;
510}
511
759af1c9 512int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index)
eb47b800 513{
e15882b6 514 struct extent_info ei = {0,0,0};
759af1c9 515 struct inode *inode = dn->inode;
028a41e8 516
759af1c9
FL		 517 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
518 dn->data_blkaddr = ei.blk + index - ei.fofs;
519 return 0;
429511cd 520 }
028a41e8 521
759af1c9 522 return f2fs_reserve_block(dn, index);
eb47b800
JK		 523}
524
a56c7c6f 525struct page *get_read_data_page(struct inode *inode, pgoff_t index,
04d328de 526 int op_flags, bool for_write)
eb47b800 527{
eb47b800
JK		 528 struct address_space *mapping = inode->i_mapping;
529 struct dnode_of_data dn;
530 struct page *page;
e15882b6 531 struct extent_info ei = {0,0,0};
eb47b800 532 int err;
cf04e8eb 533 struct f2fs_io_info fio = {
05ca3632 534 .sbi = F2FS_I_SB(inode),
cf04e8eb 535 .type = DATA,
04d328de
MC		 536 .op = REQ_OP_READ,
537 .op_flags = op_flags,
4375a336 538 .encrypted_page = NULL,
cf04e8eb 539 };
eb47b800 540
4375a336
JK		 541 if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
542 return read_mapping_page(mapping, index, NULL);
543
a56c7c6f 544 page = f2fs_grab_cache_page(mapping, index, for_write);
650495de
JK		 545 if (!page)
546 return ERR_PTR(-ENOMEM);
547
cb3bc9ee
CY		 548 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
549 dn.data_blkaddr = ei.blk + index - ei.fofs;
550 goto got_it;
551 }
552
eb47b800 553 set_new_dnode(&dn, inode, NULL, NULL, 0);
266e97a8 554 err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
86531d6b
JK		 555 if (err)
556 goto put_err;
eb47b800
JK		 557 f2fs_put_dnode(&dn);
558
6bacf52f 559 if (unlikely(dn.data_blkaddr == NULL_ADDR)) {
86531d6b
JK		 560 err = -ENOENT;
561 goto put_err;
650495de 562 }
cb3bc9ee 563got_it:
43f3eae1
JK		 564 if (PageUptodate(page)) {
565 unlock_page(page);
eb47b800 566 return page;
43f3eae1 567 }
eb47b800 568
d59ff4df
JK		 569 /*
570 * A new dentry page is allocated but not able to be written, since its
571 * new inode page couldn't be allocated due to -ENOSPC.
572 * In such the case, its blkaddr can be remained as NEW_ADDR.
573 * see, f2fs_add_link -> get_new_data_page -> init_inode_metadata.
574 */
575 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 576 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 577 if (!PageUptodate(page))
578 SetPageUptodate(page);
43f3eae1 579 unlock_page(page);
d59ff4df
JK		 580 return page;
581 }
eb47b800 582
7a9d7548 583 fio.new_blkaddr = fio.old_blkaddr = dn.data_blkaddr;
05ca3632
JK		 584 fio.page = page;
585 err = f2fs_submit_page_bio(&fio);
393ff91f 586 if (err)
86531d6b 587 goto put_err;
43f3eae1 588 return page;
86531d6b
JK		 589
590put_err:
591 f2fs_put_page(page, 1);
592 return ERR_PTR(err);
43f3eae1
JK		 593}
594
595struct page *find_data_page(struct inode *inode, pgoff_t index)
596{
597 struct address_space *mapping = inode->i_mapping;
598 struct page *page;
599
600 page = find_get_page(mapping, index);
601 if (page && PageUptodate(page))
602 return page;
603 f2fs_put_page(page, 0);
604
70fd7614 605 page = get_read_data_page(inode, index, 0, false);
43f3eae1
JK		 606 if (IS_ERR(page))
607 return page;
608
609 if (PageUptodate(page))
610 return page;
611
612 wait_on_page_locked(page);
613 if (unlikely(!PageUptodate(page))) {
614 f2fs_put_page(page, 0);
615 return ERR_PTR(-EIO);
616 }
617 return page;
618}
619
620/*
621 * If it tries to access a hole, return an error.
622 * Because, the callers, functions in dir.c and GC, should be able to know
623 * whether this page exists or not.
624 */
a56c7c6f
JK		 625struct page *get_lock_data_page(struct inode *inode, pgoff_t index,
626 bool for_write)
43f3eae1
JK		 627{
628 struct address_space *mapping = inode->i_mapping;
629 struct page *page;
630repeat:
70fd7614 631 page = get_read_data_page(inode, index, 0, for_write);
43f3eae1
JK		 632 if (IS_ERR(page))
633 return page;
393ff91f 634
43f3eae1 635 /* wait for read completion */
393ff91f 636 lock_page(page);
6bacf52f 637 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 638 f2fs_put_page(page, 1);
639 goto repeat;
eb47b800 640 }
1563ac75
CY		 641 if (unlikely(!PageUptodate(page))) {
642 f2fs_put_page(page, 1);
643 return ERR_PTR(-EIO);
644 }
eb47b800
JK		 645 return page;
646}
647
0a8165d7 648/*
eb47b800
JK		 649 * Caller ensures that this data page is never allocated.
650 * A new zero-filled data page is allocated in the page cache.
39936837 651 *
4f4124d0
CY		 652 * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and
653 * f2fs_unlock_op().
470f00e9
CY		 654 * Note that, ipage is set only by make_empty_dir, and if any error occur,
655 * ipage should be released by this function.
eb47b800 656 */
64aa7ed9 657struct page *get_new_data_page(struct inode *inode,
a8865372 658 struct page *ipage, pgoff_t index, bool new_i_size)
eb47b800 659{
eb47b800
JK		 660 struct address_space *mapping = inode->i_mapping;
661 struct page *page;
662 struct dnode_of_data dn;
663 int err;
7612118a 664
a56c7c6f 665 page = f2fs_grab_cache_page(mapping, index, true);
470f00e9
CY		 666 if (!page) {
667 /*
668 * before exiting, we should make sure ipage will be released
669 * if any error occur.
670 */
671 f2fs_put_page(ipage, 1);
01f28610 672 return ERR_PTR(-ENOMEM);
470f00e9 673 }
eb47b800 674
a8865372 675 set_new_dnode(&dn, inode, ipage, NULL, 0);
b600965c 676 err = f2fs_reserve_block(&dn, index);
01f28610
JK		 677 if (err) {
678 f2fs_put_page(page, 1);
eb47b800 679 return ERR_PTR(err);
a8865372 680 }
01f28610
JK		 681 if (!ipage)
682 f2fs_put_dnode(&dn);
eb47b800
JK		 683
684 if (PageUptodate(page))
01f28610 685 goto got_it;
eb47b800
JK		 686
687 if (dn.data_blkaddr == NEW_ADDR) {
09cbfeaf 688 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 689 if (!PageUptodate(page))
690 SetPageUptodate(page);
eb47b800 691 } else {
4375a336 692 f2fs_put_page(page, 1);
a8865372 693
7612118a
JK		 694 /* if ipage exists, blkaddr should be NEW_ADDR */
695 f2fs_bug_on(F2FS_I_SB(inode), ipage);
696 page = get_lock_data_page(inode, index, true);
4375a336 697 if (IS_ERR(page))
7612118a 698 return page;
eb47b800 699 }
01f28610 700got_it:
9edcdabf 701 if (new_i_size && i_size_read(inode) <
ee6d182f 702 ((loff_t)(index + 1) << PAGE_SHIFT))
fc9581c8 703 f2fs_i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT));
eb47b800
JK		 704 return page;
705}
706
bfad7c2d
JK		 707static int __allocate_data_block(struct dnode_of_data *dn)
708{
4081363f 709 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
bfad7c2d 710 struct f2fs_summary sum;
bfad7c2d 711 struct node_info ni;
976e4c50 712 pgoff_t fofs;
46008c6d 713 blkcnt_t count = 1;
bfad7c2d 714
91942321 715 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
bfad7c2d 716 return -EPERM;
df6136ef
CY		 717
718 dn->data_blkaddr = datablock_addr(dn->node_page, dn->ofs_in_node);
719 if (dn->data_blkaddr == NEW_ADDR)
720 goto alloc;
721
46008c6d 722 if (unlikely(!inc_valid_block_count(sbi, dn->inode, &count)))
bfad7c2d
JK		 723 return -ENOSPC;
724
df6136ef 725alloc:
bfad7c2d
JK		 726 get_node_info(sbi, dn->nid, &ni);
727 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
728
df6136ef 729 allocate_data_block(sbi, NULL, dn->data_blkaddr, &dn->data_blkaddr,
6ae1be13 730 &sum, CURSEG_WARM_DATA);
216a620a 731 set_data_blkaddr(dn);
bfad7c2d 732
976e4c50 733 /* update i_size */
81ca7350 734 fofs = start_bidx_of_node(ofs_of_node(dn->node_page), dn->inode) +
976e4c50 735 dn->ofs_in_node;
09cbfeaf 736 if (i_size_read(dn->inode) < ((loff_t)(fofs + 1) << PAGE_SHIFT))
fc9581c8 737 f2fs_i_size_write(dn->inode,
09cbfeaf 738 ((loff_t)(fofs + 1) << PAGE_SHIFT));
bfad7c2d
JK		 739 return 0;
740}
741
c040ff9d
JK		 742static inline bool __force_buffered_io(struct inode *inode, int rw)
743{
744 return ((f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) ||
745 (rw == WRITE && test_opt(F2FS_I_SB(inode), LFS)) ||
746 F2FS_I_SB(inode)->s_ndevs);
747}
748
a7de6086 749int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *from)
59b802e5 750{
b439b103 751 struct inode *inode = file_inode(iocb->ki_filp);
5b8db7fa 752 struct f2fs_map_blocks map;
a7de6086 753 int err = 0;
59b802e5 754
dc91de78
JK		 755 if (is_inode_flag_set(inode, FI_NO_PREALLOC))
756 return 0;
757
0080c507 758 map.m_lblk = F2FS_BLK_ALIGN(iocb->ki_pos);
dfd02e4d
CY		 759 map.m_len = F2FS_BYTES_TO_BLK(iocb->ki_pos + iov_iter_count(from));
760 if (map.m_len > map.m_lblk)
761 map.m_len -= map.m_lblk;
762 else
763 map.m_len = 0;
764
da85985c 765 map.m_next_pgofs = NULL;
2a340760 766
24b84912 767 if (iocb->ki_flags & IOCB_DIRECT) {
a7de6086
JK		 768 err = f2fs_convert_inline_inode(inode);
769 if (err)
770 return err;
c040ff9d
JK		 771 return f2fs_map_blocks(inode, &map, 1,
772 __force_buffered_io(inode, WRITE) ?
773 F2FS_GET_BLOCK_PRE_AIO :
774 F2FS_GET_BLOCK_PRE_DIO);
24b84912
JK		 775 }
776 if (iocb->ki_pos + iov_iter_count(from) > MAX_INLINE_DATA) {
a7de6086
JK		 777 err = f2fs_convert_inline_inode(inode);
778 if (err)
779 return err;
b439b103 780 }
24b84912
JK		 781 if (!f2fs_has_inline_data(inode))
782 return f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
a7de6086 783 return err;
59b802e5
JK		 784}
785
0a8165d7 786/*
003a3e1d
JK		 787 * f2fs_map_blocks() now supported readahead/bmap/rw direct_IO with
788 * f2fs_map_blocks structure.
4f4124d0
CY		 789 * If original data blocks are allocated, then give them to blockdev.
790 * Otherwise,
791 * a. preallocate requested block addresses
792 * b. do not use extent cache for better performance
793 * c. give the block addresses to blockdev
eb47b800 794 */
d323d005 795int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
e2b4e2bc 796 int create, int flag)
eb47b800 797{
003a3e1d 798 unsigned int maxblocks = map->m_len;
eb47b800 799 struct dnode_of_data dn;
f9811703 800 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
ac6f1999 801 int mode = create ? ALLOC_NODE : LOOKUP_NODE;
46008c6d 802 pgoff_t pgofs, end_offset, end;
bfad7c2d 803 int err = 0, ofs = 1;
46008c6d
CY		 804 unsigned int ofs_in_node, last_ofs_in_node;
805 blkcnt_t prealloc;
e15882b6 806 struct extent_info ei = {0,0,0};
7df3a431 807 block_t blkaddr;
eb47b800 808
dfd02e4d
CY		 809 if (!maxblocks)
810 return 0;
811
003a3e1d
JK		 812 map->m_len = 0;
813 map->m_flags = 0;
814
815 /* it only supports block size == page size */
816 pgofs = (pgoff_t)map->m_lblk;
46008c6d 817 end = pgofs + maxblocks;
eb47b800 818
24b84912 819 if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) {
003a3e1d
JK		 820 map->m_pblk = ei.blk + pgofs - ei.fofs;
821 map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs);
822 map->m_flags = F2FS_MAP_MAPPED;
bfad7c2d 823 goto out;
a2e7d1bf 824 }
bfad7c2d 825
4fe71e88 826next_dnode:
59b802e5 827 if (create)
3104af35 828 f2fs_lock_op(sbi);
eb47b800
JK		 829
830 /* When reading holes, we need its node page */
831 set_new_dnode(&dn, inode, NULL, NULL, 0);
bfad7c2d 832 err = get_dnode_of_data(&dn, pgofs, mode);
1ec79083 833 if (err) {
43473f96
CY		 834 if (flag == F2FS_GET_BLOCK_BMAP)
835 map->m_pblk = 0;
da85985c 836 if (err == -ENOENT) {
bfad7c2d 837 err = 0;
da85985c
CY		 838 if (map->m_next_pgofs)
839 *map->m_next_pgofs =
840 get_next_page_offset(&dn, pgofs);
841 }
bfad7c2d 842 goto unlock_out;
848753aa 843 }
973163fc 844
46008c6d 845 prealloc = 0;
230436b3 846 last_ofs_in_node = ofs_in_node = dn.ofs_in_node;
81ca7350 847 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
4fe71e88
CY		 848
849next_block:
850 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
851
852 if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR) {
973163fc 853 if (create) {
f9811703
CY		 854 if (unlikely(f2fs_cp_error(sbi))) {
855 err = -EIO;
4fe71e88 856 goto sync_out;
f9811703 857 }
24b84912 858 if (flag == F2FS_GET_BLOCK_PRE_AIO) {
46008c6d
CY		 859 if (blkaddr == NULL_ADDR) {
860 prealloc++;
861 last_ofs_in_node = dn.ofs_in_node;
862 }
24b84912
JK		 863 } else {
864 err = __allocate_data_block(&dn);
6f2d8ed6 865 if (!err)
91942321 866 set_inode_flag(inode, FI_APPEND_WRITE);
24b84912 867 }
973163fc 868 if (err)
4fe71e88 869 goto sync_out;
973163fc 870 map->m_flags = F2FS_MAP_NEW;
4fe71e88 871 blkaddr = dn.data_blkaddr;
973163fc 872 } else {
43473f96
CY		 873 if (flag == F2FS_GET_BLOCK_BMAP) {
874 map->m_pblk = 0;
875 goto sync_out;
876 }
da85985c
CY		 877 if (flag == F2FS_GET_BLOCK_FIEMAP &&
878 blkaddr == NULL_ADDR) {
879 if (map->m_next_pgofs)
880 *map->m_next_pgofs = pgofs + 1;
881 }
973163fc 882 if (flag != F2FS_GET_BLOCK_FIEMAP ||
43473f96 883 blkaddr != NEW_ADDR)
4fe71e88 884 goto sync_out;
e2b4e2bc 885 }
e2b4e2bc 886 }
eb47b800 887
46008c6d
CY		 888 if (flag == F2FS_GET_BLOCK_PRE_AIO)
889 goto skip;
890
4fe71e88
CY		 891 if (map->m_len == 0) {
892 /* preallocated unwritten block should be mapped for fiemap. */
893 if (blkaddr == NEW_ADDR)
894 map->m_flags |= F2FS_MAP_UNWRITTEN;
895 map->m_flags |= F2FS_MAP_MAPPED;
896
897 map->m_pblk = blkaddr;
898 map->m_len = 1;
899 } else if ((map->m_pblk != NEW_ADDR &&
900 blkaddr == (map->m_pblk + ofs)) ||
b439b103 901 (map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR) ||
46008c6d 902 flag == F2FS_GET_BLOCK_PRE_DIO) {
4fe71e88
CY		 903 ofs++;
904 map->m_len++;
905 } else {
906 goto sync_out;
907 }
bfad7c2d 908
46008c6d 909skip:
bfad7c2d
JK		 910 dn.ofs_in_node++;
911 pgofs++;
912
46008c6d
CY		 913 /* preallocate blocks in batch for one dnode page */
914 if (flag == F2FS_GET_BLOCK_PRE_AIO &&
915 (pgofs == end || dn.ofs_in_node == end_offset)) {
7df3a431 916
46008c6d
CY		 917 dn.ofs_in_node = ofs_in_node;
918 err = reserve_new_blocks(&dn, prealloc);
919 if (err)
920 goto sync_out;
bfad7c2d 921
46008c6d
CY		 922 map->m_len += dn.ofs_in_node - ofs_in_node;
923 if (prealloc && dn.ofs_in_node != last_ofs_in_node + 1) {
924 err = -ENOSPC;
925 goto sync_out;
3104af35 926 }
46008c6d
CY		 927 dn.ofs_in_node = end_offset;
928 }
929
930 if (pgofs >= end)
931 goto sync_out;
932 else if (dn.ofs_in_node < end_offset)
933 goto next_block;
934
46008c6d
CY		 935 f2fs_put_dnode(&dn);
936
937 if (create) {
938 f2fs_unlock_op(sbi);
6f2d8ed6 939 f2fs_balance_fs(sbi, dn.node_changed);
eb47b800 940 }
46008c6d 941 goto next_dnode;
7df3a431 942
bfad7c2d 943sync_out:
eb47b800 944 f2fs_put_dnode(&dn);
bfad7c2d 945unlock_out:
2a340760 946 if (create) {
3104af35 947 f2fs_unlock_op(sbi);
6f2d8ed6 948 f2fs_balance_fs(sbi, dn.node_changed);
2a340760 949 }
bfad7c2d 950out:
003a3e1d 951 trace_f2fs_map_blocks(inode, map, err);
bfad7c2d 952 return err;
eb47b800
JK		 953}
954
003a3e1d 955static int __get_data_block(struct inode *inode, sector_t iblock,
da85985c
CY		 956 struct buffer_head *bh, int create, int flag,
957 pgoff_t *next_pgofs)
003a3e1d
JK		 958{
959 struct f2fs_map_blocks map;
a7de6086 960 int err;
003a3e1d
JK		 961
962 map.m_lblk = iblock;
963 map.m_len = bh->b_size >> inode->i_blkbits;
da85985c 964 map.m_next_pgofs = next_pgofs;
003a3e1d 965
a7de6086
JK		 966 err = f2fs_map_blocks(inode, &map, create, flag);
967 if (!err) {
003a3e1d
JK		 968 map_bh(bh, inode->i_sb, map.m_pblk);
969 bh->b_state = (bh->b_state & ~F2FS_MAP_FLAGS) | map.m_flags;
b86e3307 970 bh->b_size = (u64)map.m_len << inode->i_blkbits;
003a3e1d 971 }
a7de6086 972 return err;
003a3e1d
JK		 973}
974
ccfb3000 975static int get_data_block(struct inode *inode, sector_t iblock,
da85985c
CY		 976 struct buffer_head *bh_result, int create, int flag,
977 pgoff_t *next_pgofs)
e2b4e2bc 978{
da85985c
CY		 979 return __get_data_block(inode, iblock, bh_result, create,
980 flag, next_pgofs);
e2b4e2bc
CY		 981}
982
983static int get_data_block_dio(struct inode *inode, sector_t iblock,
ccfb3000
JK		 984 struct buffer_head *bh_result, int create)
985{
e2b4e2bc 986 return __get_data_block(inode, iblock, bh_result, create,
da85985c 987 F2FS_GET_BLOCK_DIO, NULL);
ccfb3000
JK		 988}
989
e2b4e2bc 990static int get_data_block_bmap(struct inode *inode, sector_t iblock,
ccfb3000
JK		 991 struct buffer_head *bh_result, int create)
992{
179448bf 993 /* Block number less than F2FS MAX BLOCKS */
e0afc4d6 994 if (unlikely(iblock >= F2FS_I_SB(inode)->max_file_blocks))
179448bf
YH		 995 return -EFBIG;
996
e2b4e2bc 997 return __get_data_block(inode, iblock, bh_result, create,
da85985c 998 F2FS_GET_BLOCK_BMAP, NULL);
ccfb3000
JK		 999}
1000
7f63eb77
JK		 1001static inline sector_t logical_to_blk(struct inode *inode, loff_t offset)
1002{
1003 return (offset >> inode->i_blkbits);
1004}
1005
1006static inline loff_t blk_to_logical(struct inode *inode, sector_t blk)
1007{
1008 return (blk << inode->i_blkbits);
1009}
1010
9ab70134
JK		 1011int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1012 u64 start, u64 len)
1013{
7f63eb77
JK		 1014 struct buffer_head map_bh;
1015 sector_t start_blk, last_blk;
da85985c 1016 pgoff_t next_pgofs;
7f63eb77
JK		 1017 u64 logical = 0, phys = 0, size = 0;
1018 u32 flags = 0;
7f63eb77
JK		 1019 int ret = 0;
1020
1021 ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC);
1022 if (ret)
1023 return ret;
1024
67f8cf3c
JK		 1025 if (f2fs_has_inline_data(inode)) {
1026 ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len);
1027 if (ret != -EAGAIN)
1028 return ret;
1029 }
1030
5955102c 1031 inode_lock(inode);
de1475cc 1032
7f63eb77
JK		 1033 if (logical_to_blk(inode, len) == 0)
1034 len = blk_to_logical(inode, 1);
1035
1036 start_blk = logical_to_blk(inode, start);
1037 last_blk = logical_to_blk(inode, start + len - 1);
9a950d52 1038
7f63eb77
JK		 1039next:
1040 memset(&map_bh, 0, sizeof(struct buffer_head));
1041 map_bh.b_size = len;
1042
e2b4e2bc 1043 ret = get_data_block(inode, start_blk, &map_bh, 0,
da85985c 1044 F2FS_GET_BLOCK_FIEMAP, &next_pgofs);
7f63eb77
JK		 1045 if (ret)
1046 goto out;
1047
1048 /* HOLE */
1049 if (!buffer_mapped(&map_bh)) {
da85985c 1050 start_blk = next_pgofs;
58736fa6
CY		 1051
1052 if (blk_to_logical(inode, start_blk) < blk_to_logical(inode,
1053 F2FS_I_SB(inode)->max_file_blocks))
9a950d52 1054 goto prep_next;
58736fa6 1055
9a950d52
FL		 1056 flags |= FIEMAP_EXTENT_LAST;
1057 }
7f63eb77 1058
da5af127
CY		 1059 if (size) {
1060 if (f2fs_encrypted_inode(inode))
1061 flags |= FIEMAP_EXTENT_DATA_ENCRYPTED;
1062
9a950d52
FL		 1063 ret = fiemap_fill_next_extent(fieinfo, logical,
1064 phys, size, flags);
da5af127 1065 }
7f63eb77 1066
9a950d52
FL		 1067 if (start_blk > last_blk || ret)
1068 goto out;
7f63eb77 1069
9a950d52
FL		 1070 logical = blk_to_logical(inode, start_blk);
1071 phys = blk_to_logical(inode, map_bh.b_blocknr);
1072 size = map_bh.b_size;
1073 flags = 0;
1074 if (buffer_unwritten(&map_bh))
1075 flags = FIEMAP_EXTENT_UNWRITTEN;
7f63eb77 1076
9a950d52 1077 start_blk += logical_to_blk(inode, size);
7f63eb77 1078
9a950d52 1079prep_next:
7f63eb77
JK		 1080 cond_resched();
1081 if (fatal_signal_pending(current))
1082 ret = -EINTR;
1083 else
1084 goto next;
1085out:
1086 if (ret == 1)
1087 ret = 0;
1088
5955102c 1089 inode_unlock(inode);
7f63eb77 1090 return ret;
9ab70134
JK		 1091}
1092
6a7a3aed
WY		 1093static struct bio *f2fs_grab_bio(struct inode *inode, block_t blkaddr,
1094 unsigned nr_pages)
78682f79
CY		 1095{
1096 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1097 struct fscrypt_ctx *ctx = NULL;
78682f79
CY		 1098 struct bio *bio;
1099
1100 if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) {
1101 ctx = fscrypt_get_ctx(inode, GFP_NOFS);
1102 if (IS_ERR(ctx))
1103 return ERR_CAST(ctx);
1104
1105 /* wait the page to be moved by cleaning */
1106 f2fs_wait_on_encrypted_page_writeback(sbi, blkaddr);
1107 }
1108
1109 bio = bio_alloc(GFP_KERNEL, min_t(int, nr_pages, BIO_MAX_PAGES));
1110 if (!bio) {
1111 if (ctx)
1112 fscrypt_release_ctx(ctx);
1113 return ERR_PTR(-ENOMEM);
1114 }
3c62be17 1115 f2fs_target_device(sbi, blkaddr, bio);
78682f79
CY		 1116 bio->bi_end_io = f2fs_read_end_io;
1117 bio->bi_private = ctx;
1118
1119 return bio;
1120}
1121
f1e88660
JK		 1122/*
1123 * This function was originally taken from fs/mpage.c, and customized for f2fs.
1124 * Major change was from block_size == page_size in f2fs by default.
1125 */
1126static int f2fs_mpage_readpages(struct address_space *mapping,
1127 struct list_head *pages, struct page *page,
1128 unsigned nr_pages)
1129{
1130 struct bio *bio = NULL;
1131 unsigned page_idx;
1132 sector_t last_block_in_bio = 0;
1133 struct inode *inode = mapping->host;
1134 const unsigned blkbits = inode->i_blkbits;
1135 const unsigned blocksize = 1 << blkbits;
1136 sector_t block_in_file;
1137 sector_t last_block;
1138 sector_t last_block_in_file;
1139 sector_t block_nr;
f1e88660
JK		 1140 struct f2fs_map_blocks map;
1141
1142 map.m_pblk = 0;
1143 map.m_lblk = 0;
1144 map.m_len = 0;
1145 map.m_flags = 0;
da85985c 1146 map.m_next_pgofs = NULL;
f1e88660
JK		 1147
1148 for (page_idx = 0; nr_pages; page_idx++, nr_pages--) {
1149
1150 prefetchw(&page->flags);
1151 if (pages) {
939afa94 1152 page = list_last_entry(pages, struct page, lru);
f1e88660
JK		 1153 list_del(&page->lru);
1154 if (add_to_page_cache_lru(page, mapping,
8a5c743e
MH		 1155 page->index,
1156 readahead_gfp_mask(mapping)))
f1e88660
JK		 1157 goto next_page;
1158 }
1159
1160 block_in_file = (sector_t)page->index;
1161 last_block = block_in_file + nr_pages;
1162 last_block_in_file = (i_size_read(inode) + blocksize - 1) >>
1163 blkbits;
1164 if (last_block > last_block_in_file)
1165 last_block = last_block_in_file;
1166
1167 /*
1168 * Map blocks using the previous result first.
1169 */
1170 if ((map.m_flags & F2FS_MAP_MAPPED) &&
1171 block_in_file > map.m_lblk &&
1172 block_in_file < (map.m_lblk + map.m_len))
1173 goto got_it;
1174
1175 /*
1176 * Then do more f2fs_map_blocks() calls until we are
1177 * done with this page.
1178 */
1179 map.m_flags = 0;
1180
1181 if (block_in_file < last_block) {
1182 map.m_lblk = block_in_file;
1183 map.m_len = last_block - block_in_file;
1184
46c9e141 1185 if (f2fs_map_blocks(inode, &map, 0,
da85985c 1186 F2FS_GET_BLOCK_READ))
f1e88660
JK		 1187 goto set_error_page;
1188 }
1189got_it:
1190 if ((map.m_flags & F2FS_MAP_MAPPED)) {
1191 block_nr = map.m_pblk + block_in_file - map.m_lblk;
1192 SetPageMappedToDisk(page);
1193
1194 if (!PageUptodate(page) && !cleancache_get_page(page)) {
1195 SetPageUptodate(page);
1196 goto confused;
1197 }
1198 } else {
09cbfeaf 1199 zero_user_segment(page, 0, PAGE_SIZE);
237c0790
JK		 1200 if (!PageUptodate(page))
1201 SetPageUptodate(page);
f1e88660
JK		 1202 unlock_page(page);
1203 goto next_page;
1204 }
1205
1206 /*
1207 * This page will go to BIO. Do we need to send this
1208 * BIO off first?
1209 */
3c62be17
JK		 1210 if (bio && (last_block_in_bio != block_nr - 1 ||
1211 !__same_bdev(F2FS_I_SB(inode), block_nr, bio))) {
f1e88660 1212submit_and_realloc:
4fc29c1a 1213 __submit_bio(F2FS_I_SB(inode), bio, DATA);
f1e88660
JK		 1214 bio = NULL;
1215 }
1216 if (bio == NULL) {
78682f79 1217 bio = f2fs_grab_bio(inode, block_nr, nr_pages);
1d353eb7
JK		 1218 if (IS_ERR(bio)) {
1219 bio = NULL;
f1e88660 1220 goto set_error_page;
4375a336 1221 }
04d328de 1222 bio_set_op_attrs(bio, REQ_OP_READ, 0);
f1e88660
JK		 1223 }
1224
1225 if (bio_add_page(bio, page, blocksize, 0) < blocksize)
1226 goto submit_and_realloc;
1227
1228 last_block_in_bio = block_nr;
1229 goto next_page;
1230set_error_page:
1231 SetPageError(page);
09cbfeaf 1232 zero_user_segment(page, 0, PAGE_SIZE);
f1e88660
JK		 1233 unlock_page(page);
1234 goto next_page;
1235confused:
1236 if (bio) {
4fc29c1a 1237 __submit_bio(F2FS_I_SB(inode), bio, DATA);
f1e88660
JK		 1238 bio = NULL;
1239 }
1240 unlock_page(page);
1241next_page:
1242 if (pages)
09cbfeaf 1243 put_page(page);
f1e88660
JK		 1244 }
1245 BUG_ON(pages && !list_empty(pages));
1246 if (bio)
4fc29c1a 1247 __submit_bio(F2FS_I_SB(inode), bio, DATA);
f1e88660
JK		 1248 return 0;
1249}
1250
eb47b800
JK		 1251static int f2fs_read_data_page(struct file *file, struct page *page)
1252{
9ffe0fb5 1253 struct inode *inode = page->mapping->host;
b3d208f9 1254 int ret = -EAGAIN;
9ffe0fb5 1255
c20e89cd
CY		 1256 trace_f2fs_readpage(page, DATA);
1257
e1c42045 1258 /* If the file has inline data, try to read it directly */
9ffe0fb5
HL		 1259 if (f2fs_has_inline_data(inode))
1260 ret = f2fs_read_inline_data(inode, page);
b3d208f9 1261 if (ret == -EAGAIN)
f1e88660 1262 ret = f2fs_mpage_readpages(page->mapping, NULL, page, 1);
9ffe0fb5 1263 return ret;
eb47b800
JK		 1264}
1265
1266static int f2fs_read_data_pages(struct file *file,
1267 struct address_space *mapping,
1268 struct list_head *pages, unsigned nr_pages)
1269{
9ffe0fb5 1270 struct inode *inode = file->f_mapping->host;
939afa94 1271 struct page *page = list_last_entry(pages, struct page, lru);
b8c29400
CY		 1272
1273 trace_f2fs_readpages(inode, page, nr_pages);
9ffe0fb5
HL		 1274
1275 /* If the file has inline data, skip readpages */
1276 if (f2fs_has_inline_data(inode))
1277 return 0;
1278
f1e88660 1279 return f2fs_mpage_readpages(mapping, pages, NULL, nr_pages);
eb47b800
JK		 1280}
1281
05ca3632 1282int do_write_data_page(struct f2fs_io_info *fio)
eb47b800 1283{
05ca3632 1284 struct page *page = fio->page;
eb47b800 1285 struct inode *inode = page->mapping->host;
eb47b800
JK		 1286 struct dnode_of_data dn;
1287 int err = 0;
1288
1289 set_new_dnode(&dn, inode, NULL, NULL, 0);
266e97a8 1290 err = get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
eb47b800
JK		 1291 if (err)
1292 return err;
1293
28bc106b 1294 fio->old_blkaddr = dn.data_blkaddr;
eb47b800
JK		 1295
1296 /* This page is already truncated */
7a9d7548 1297 if (fio->old_blkaddr == NULL_ADDR) {
2bca1e23 1298 ClearPageUptodate(page);
eb47b800 1299 goto out_writepage;
2bca1e23 1300 }
eb47b800 1301
4375a336 1302 if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) {
b32e4482 1303 gfp_t gfp_flags = GFP_NOFS;
08b39fbd
CY		 1304
1305 /* wait for GCed encrypted page writeback */
1306 f2fs_wait_on_encrypted_page_writeback(F2FS_I_SB(inode),
7a9d7548 1307 fio->old_blkaddr);
b32e4482
JK		 1308retry_encrypt:
1309 fio->encrypted_page = fscrypt_encrypt_page(inode, fio->page,
9c4bb8a3
DG		 1310 PAGE_SIZE, 0,
1311 fio->page->index,
1312 gfp_flags);
4375a336
JK		 1313 if (IS_ERR(fio->encrypted_page)) {
1314 err = PTR_ERR(fio->encrypted_page);
b32e4482
JK		 1315 if (err == -ENOMEM) {
1316 /* flush pending ios and wait for a while */
1317 f2fs_flush_merged_bios(F2FS_I_SB(inode));
1318 congestion_wait(BLK_RW_ASYNC, HZ/50);
1319 gfp_flags |= __GFP_NOFAIL;
1320 err = 0;
1321 goto retry_encrypt;
1322 }
4375a336
JK		 1323 goto out_writepage;
1324 }
1325 }
1326
eb47b800
JK		 1327 set_page_writeback(page);
1328
1329 /*
1330 * If current allocation needs SSR,
1331 * it had better in-place writes for updated data.
1332 */
7a9d7548 1333 if (unlikely(fio->old_blkaddr != NEW_ADDR &&
b25958b6 1334 !is_cold_data(page) &&
2da3e027 1335 !IS_ATOMIC_WRITTEN_PAGE(page) &&
b25958b6 1336 need_inplace_update(inode))) {
05ca3632 1337 rewrite_data_page(fio);
91942321 1338 set_inode_flag(inode, FI_UPDATE_WRITE);
8ce67cb0 1339 trace_f2fs_do_write_data_page(page, IPU);
eb47b800 1340 } else {
05ca3632 1341 write_data_page(&dn, fio);
8ce67cb0 1342 trace_f2fs_do_write_data_page(page, OPU);
91942321 1343 set_inode_flag(inode, FI_APPEND_WRITE);
3c6c2beb 1344 if (page->index == 0)
91942321 1345 set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
eb47b800
JK		 1346 }
1347out_writepage:
1348 f2fs_put_dnode(&dn);
1349 return err;
1350}
1351
d68f735b
JK		 1352static int __write_data_page(struct page *page, bool *submitted,
1353 struct writeback_control *wbc)
eb47b800
JK		 1354{
1355 struct inode *inode = page->mapping->host;
4081363f 1356 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
eb47b800
JK		 1357 loff_t i_size = i_size_read(inode);
1358 const pgoff_t end_index = ((unsigned long long) i_size)
09cbfeaf 1359 >> PAGE_SHIFT;
26de9b11 1360 loff_t psize = (page->index + 1) << PAGE_SHIFT;
9ffe0fb5 1361 unsigned offset = 0;
39936837 1362 bool need_balance_fs = false;
eb47b800 1363 int err = 0;
458e6197 1364 struct f2fs_io_info fio = {
05ca3632 1365 .sbi = sbi,
458e6197 1366 .type = DATA,
04d328de 1367 .op = REQ_OP_WRITE,
7637241e 1368 .op_flags = wbc_to_write_flags(wbc),
05ca3632 1369 .page = page,
4375a336 1370 .encrypted_page = NULL,
d68f735b 1371 .submitted = false,
458e6197 1372 };
eb47b800 1373
ecda0de3
CY		 1374 trace_f2fs_writepage(page, DATA);
1375
eb47b800 1376 if (page->index < end_index)
39936837 1377 goto write;
eb47b800
JK		 1378
1379 /*
1380 * If the offset is out-of-range of file size,
1381 * this page does not have to be written to disk.
1382 */
09cbfeaf 1383 offset = i_size & (PAGE_SIZE - 1);
76f60268 1384 if ((page->index >= end_index + 1) || !offset)
39936837 1385 goto out;
eb47b800 1386
09cbfeaf 1387 zero_user_segment(page, offset, PAGE_SIZE);
39936837 1388write:
caf0047e 1389 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
eb47b800 1390 goto redirty_out;
1e84371f
JK		 1391 if (f2fs_is_drop_cache(inode))
1392 goto out;
e6e5f561
JK		 1393 /* we should not write 0'th page having journal header */
1394 if (f2fs_is_volatile_file(inode) && (!page->index ||
1395 (!wbc->for_reclaim &&
1396 available_free_memory(sbi, BASE_CHECK))))
1e84371f 1397 goto redirty_out;
eb47b800 1398
cf779cab
JK		 1399 /* we should bypass data pages to proceed the kworkder jobs */
1400 if (unlikely(f2fs_cp_error(sbi))) {
7f319975 1401 mapping_set_error(page->mapping, -EIO);
a7ffdbe2 1402 goto out;
cf779cab
JK		 1403 }
1404
39936837 1405 /* Dentry blocks are controlled by checkpoint */
eb47b800 1406 if (S_ISDIR(inode->i_mode)) {
05ca3632 1407 err = do_write_data_page(&fio);
8618b881
JK		 1408 goto done;
1409 }
9ffe0fb5 1410
8618b881 1411 if (!wbc->for_reclaim)
39936837 1412 need_balance_fs = true;
7f3037a5 1413 else if (has_not_enough_free_secs(sbi, 0, 0))
39936837 1414 goto redirty_out;
eb47b800 1415
b3d208f9 1416 err = -EAGAIN;
8618b881 1417 f2fs_lock_op(sbi);
b3d208f9
JK		 1418 if (f2fs_has_inline_data(inode))
1419 err = f2fs_write_inline_data(inode, page);
1420 if (err == -EAGAIN)
05ca3632 1421 err = do_write_data_page(&fio);
26de9b11
JK		 1422 if (F2FS_I(inode)->last_disk_size < psize)
1423 F2FS_I(inode)->last_disk_size = psize;
8618b881
JK		 1424 f2fs_unlock_op(sbi);
1425done:
1426 if (err && err != -ENOENT)
1427 goto redirty_out;
eb47b800 1428
39936837 1429out:
a7ffdbe2 1430 inode_dec_dirty_pages(inode);
2bca1e23
JK		 1431 if (err)
1432 ClearPageUptodate(page);
0c3a5797
CY		 1433
1434 if (wbc->for_reclaim) {
942fd319
JK		 1435 f2fs_submit_merged_bio_cond(sbi, inode, 0, page->index,
1436 DATA, WRITE);
0c3a5797 1437 remove_dirty_inode(inode);
d68f735b 1438 submitted = NULL;
0c3a5797
CY		 1439 }
1440
eb47b800 1441 unlock_page(page);
2c4db1a6 1442 f2fs_balance_fs(sbi, need_balance_fs);
0c3a5797 1443
d68f735b 1444 if (unlikely(f2fs_cp_error(sbi))) {
2aea39ec 1445 f2fs_submit_merged_bio(sbi, DATA, WRITE);
d68f735b
JK		 1446 submitted = NULL;
1447 }
1448
1449 if (submitted)
1450 *submitted = fio.submitted;
0c3a5797 1451
eb47b800
JK		 1452 return 0;
1453
eb47b800 1454redirty_out:
76f60268 1455 redirty_page_for_writepage(wbc, page);
0002b61b
CY		 1456 if (!err)
1457 return AOP_WRITEPAGE_ACTIVATE;
b230e6ca
JK		 1458 unlock_page(page);
1459 return err;
fa9150a8
NJ		 1460}
1461
f566bae8
JK		 1462static int f2fs_write_data_page(struct page *page,
1463 struct writeback_control *wbc)
1464{
d68f735b 1465 return __write_data_page(page, NULL, wbc);
f566bae8
JK		 1466}
1467
8f46dcae
CY		 1468/*
1469 * This function was copied from write_cche_pages from mm/page-writeback.c.
1470 * The major change is making write step of cold data page separately from
1471 * warm/hot data page.
1472 */
1473static int f2fs_write_cache_pages(struct address_space *mapping,
b230e6ca 1474 struct writeback_control *wbc)
8f46dcae
CY		 1475{
1476 int ret = 0;
1477 int done = 0;
1478 struct pagevec pvec;
1479 int nr_pages;
1480 pgoff_t uninitialized_var(writeback_index);
1481 pgoff_t index;
1482 pgoff_t end; /* Inclusive */
1483 pgoff_t done_index;
942fd319 1484 pgoff_t last_idx = ULONG_MAX;
8f46dcae
CY		 1485 int cycled;
1486 int range_whole = 0;
1487 int tag;
8f46dcae
CY		 1488
1489 pagevec_init(&pvec, 0);
46ae957f 1490
8f46dcae
CY		 1491 if (wbc->range_cyclic) {
1492 writeback_index = mapping->writeback_index; /* prev offset */
1493 index = writeback_index;
1494 if (index == 0)
1495 cycled = 1;
1496 else
1497 cycled = 0;
1498 end = -1;
1499 } else {
09cbfeaf
KS		 1500 index = wbc->range_start >> PAGE_SHIFT;
1501 end = wbc->range_end >> PAGE_SHIFT;
8f46dcae
CY		 1502 if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
1503 range_whole = 1;
1504 cycled = 1; /* ignore range_cyclic tests */
1505 }
1506 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
1507 tag = PAGECACHE_TAG_TOWRITE;
1508 else
1509 tag = PAGECACHE_TAG_DIRTY;
1510retry:
1511 if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
1512 tag_pages_for_writeback(mapping, index, end);
1513 done_index = index;
1514 while (!done && (index <= end)) {
1515 int i;
1516
1517 nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
1518 min(end - index, (pgoff_t)PAGEVEC_SIZE - 1) + 1);
1519 if (nr_pages == 0)
1520 break;
1521
1522 for (i = 0; i < nr_pages; i++) {
1523 struct page *page = pvec.pages[i];
d68f735b 1524 bool submitted = false;
8f46dcae
CY		 1525
1526 if (page->index > end) {
1527 done = 1;
1528 break;
1529 }
1530
1531 done_index = page->index;
1532
1533 lock_page(page);
1534
1535 if (unlikely(page->mapping != mapping)) {
1536continue_unlock:
1537 unlock_page(page);
1538 continue;
1539 }
1540
1541 if (!PageDirty(page)) {
1542 /* someone wrote it for us */
1543 goto continue_unlock;
1544 }
1545
8f46dcae
CY		 1546 if (PageWriteback(page)) {
1547 if (wbc->sync_mode != WB_SYNC_NONE)
fec1d657
JK		 1548 f2fs_wait_on_page_writeback(page,
1549 DATA, true);
8f46dcae
CY		 1550 else
1551 goto continue_unlock;
1552 }
1553
1554 BUG_ON(PageWriteback(page));
1555 if (!clear_page_dirty_for_io(page))
1556 goto continue_unlock;
1557
d68f735b 1558 ret = __write_data_page(page, &submitted, wbc);
8f46dcae 1559 if (unlikely(ret)) {
0002b61b
CY		 1560 /*
1561 * keep nr_to_write, since vfs uses this to
1562 * get # of written pages.
1563 */
1564 if (ret == AOP_WRITEPAGE_ACTIVATE) {
1565 unlock_page(page);
1566 ret = 0;
1567 continue;
1568 }
b230e6ca
JK		 1569 done_index = page->index + 1;
1570 done = 1;
1571 break;
d68f735b 1572 } else if (submitted) {
942fd319 1573 last_idx = page->index;
8f46dcae
CY		 1574 }
1575
1576 if (--wbc->nr_to_write <= 0 &&
1577 wbc->sync_mode == WB_SYNC_NONE) {
1578 done = 1;
1579 break;
1580 }
1581 }
1582 pagevec_release(&pvec);
1583 cond_resched();
1584 }
1585
8f46dcae
CY		 1586 if (!cycled && !done) {
1587 cycled = 1;
1588 index = 0;
1589 end = writeback_index - 1;
1590 goto retry;
1591 }
1592 if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
1593 mapping->writeback_index = done_index;
1594
942fd319 1595 if (last_idx != ULONG_MAX)
6ca56ca4 1596 f2fs_submit_merged_bio_cond(F2FS_M_SB(mapping), mapping->host,
942fd319 1597 0, last_idx, DATA, WRITE);
6ca56ca4 1598
8f46dcae
CY		 1599 return ret;
1600}
1601
25ca923b 1602static int f2fs_write_data_pages(struct address_space *mapping,
eb47b800
JK		 1603 struct writeback_control *wbc)
1604{
1605 struct inode *inode = mapping->host;
4081363f 1606 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
9dfa1baf 1607 struct blk_plug plug;
eb47b800 1608 int ret;
eb47b800 1609
cfb185a1 1610 /* deal with chardevs and other special file */
1611 if (!mapping->a_ops->writepage)
1612 return 0;
1613
6a290544
CY		 1614 /* skip writing if there is no dirty page in this inode */
1615 if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE)
1616 return 0;
1617
a1257023
JK		 1618 if (S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_NONE &&
1619 get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) &&
1620 available_free_memory(sbi, DIRTY_DENTS))
1621 goto skip_write;
1622
d323d005 1623 /* skip writing during file defragment */
91942321 1624 if (is_inode_flag_set(inode, FI_DO_DEFRAG))
d323d005
CY		 1625 goto skip_write;
1626
d5669f7b
JK		 1627 /* during POR, we don't need to trigger writepage at all. */
1628 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1629 goto skip_write;
1630
d31c7c3f
YH		 1631 trace_f2fs_writepages(mapping->host, wbc, DATA);
1632
9dfa1baf 1633 blk_start_plug(&plug);
b230e6ca 1634 ret = f2fs_write_cache_pages(mapping, wbc);
9dfa1baf 1635 blk_finish_plug(&plug);
28ea6162
JK		 1636 /*
1637 * if some pages were truncated, we cannot guarantee its mapping->host
1638 * to detect pending bios.
1639 */
458e6197 1640
c227f912 1641 remove_dirty_inode(inode);
eb47b800 1642 return ret;
d3baf95d
JK		 1643
1644skip_write:
a7ffdbe2 1645 wbc->pages_skipped += get_dirty_pages(inode);
d31c7c3f 1646 trace_f2fs_writepages(mapping->host, wbc, DATA);
d3baf95d 1647 return 0;
eb47b800
JK		 1648}
1649
3aab8f82
CY		 1650static void f2fs_write_failed(struct address_space *mapping, loff_t to)
1651{
1652 struct inode *inode = mapping->host;
819d9153 1653 loff_t i_size = i_size_read(inode);
3aab8f82 1654
819d9153
JK		 1655 if (to > i_size) {
1656 truncate_pagecache(inode, i_size);
1657 truncate_blocks(inode, i_size, true);
3aab8f82
CY		 1658 }
1659}
1660
2aadac08
JK		 1661static int prepare_write_begin(struct f2fs_sb_info *sbi,
1662 struct page *page, loff_t pos, unsigned len,
1663 block_t *blk_addr, bool *node_changed)
1664{
1665 struct inode *inode = page->mapping->host;
1666 pgoff_t index = page->index;
1667 struct dnode_of_data dn;
1668 struct page *ipage;
b4d07a3e 1669 bool locked = false;
e15882b6 1670 struct extent_info ei = {0,0,0};
2aadac08
JK		 1671 int err = 0;
1672
24b84912
JK		 1673 /*
1674 * we already allocated all the blocks, so we don't need to get
1675 * the block addresses when there is no need to fill the page.
1676 */
dc91de78
JK		 1677 if (!f2fs_has_inline_data(inode) && len == PAGE_SIZE &&
1678 !is_inode_flag_set(inode, FI_NO_PREALLOC))
24b84912
JK		 1679 return 0;
1680
b4d07a3e 1681 if (f2fs_has_inline_data(inode) ||
09cbfeaf 1682 (pos & PAGE_MASK) >= i_size_read(inode)) {
b4d07a3e
JK		 1683 f2fs_lock_op(sbi);
1684 locked = true;
1685 }
1686restart:
2aadac08
JK		 1687 /* check inline_data */
1688 ipage = get_node_page(sbi, inode->i_ino);
1689 if (IS_ERR(ipage)) {
1690 err = PTR_ERR(ipage);
1691 goto unlock_out;
1692 }
1693
1694 set_new_dnode(&dn, inode, ipage, ipage, 0);
1695
1696 if (f2fs_has_inline_data(inode)) {
1697 if (pos + len <= MAX_INLINE_DATA) {
1698 read_inline_data(page, ipage);
91942321 1699 set_inode_flag(inode, FI_DATA_EXIST);
ab47036d
CY		 1700 if (inode->i_nlink)
1701 set_inline_node(ipage);
2aadac08
JK		 1702 } else {
1703 err = f2fs_convert_inline_page(&dn, page);
1704 if (err)
b4d07a3e
JK		 1705 goto out;
1706 if (dn.data_blkaddr == NULL_ADDR)
1707 err = f2fs_get_block(&dn, index);
1708 }
1709 } else if (locked) {
1710 err = f2fs_get_block(&dn, index);
1711 } else {
1712 if (f2fs_lookup_extent_cache(inode, index, &ei)) {
1713 dn.data_blkaddr = ei.blk + index - ei.fofs;
1714 } else {
b4d07a3e
JK		 1715 /* hole case */
1716 err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
4da7bf5a 1717 if (err || dn.data_blkaddr == NULL_ADDR) {
b4d07a3e
JK		 1718 f2fs_put_dnode(&dn);
1719 f2fs_lock_op(sbi);
1720 locked = true;
1721 goto restart;
1722 }
2aadac08
JK		 1723 }
1724 }
b4d07a3e 1725
2aadac08
JK		 1726 /* convert_inline_page can make node_changed */
1727 *blk_addr = dn.data_blkaddr;
1728 *node_changed = dn.node_changed;
b4d07a3e 1729out:
2aadac08
JK		 1730 f2fs_put_dnode(&dn);
1731unlock_out:
b4d07a3e
JK		 1732 if (locked)
1733 f2fs_unlock_op(sbi);
2aadac08
JK		 1734 return err;
1735}
1736
eb47b800
JK		 1737static int f2fs_write_begin(struct file *file, struct address_space *mapping,
1738 loff_t pos, unsigned len, unsigned flags,
1739 struct page **pagep, void **fsdata)
1740{
1741 struct inode *inode = mapping->host;
4081363f 1742 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
86531d6b 1743 struct page *page = NULL;
09cbfeaf 1744 pgoff_t index = ((unsigned long long) pos) >> PAGE_SHIFT;
2aadac08
JK		 1745 bool need_balance = false;
1746 block_t blkaddr = NULL_ADDR;
eb47b800
JK		 1747 int err = 0;
1748
62aed044
CY		 1749 trace_f2fs_write_begin(inode, pos, len, flags);
1750
5f727395
JK		 1751 /*
1752 * We should check this at this moment to avoid deadlock on inode page
1753 * and #0 page. The locking rule for inline_data conversion should be:
1754 * lock_page(page #0) -> lock_page(inode_page)
1755 */
1756 if (index != 0) {
1757 err = f2fs_convert_inline_inode(inode);
1758 if (err)
1759 goto fail;
1760 }
afcb7ca0 1761repeat:
86d54795
JK		 1762 /*
1763 * Do not use grab_cache_page_write_begin() to avoid deadlock due to
1764 * wait_for_stable_page. Will wait that below with our IO control.
1765 */
1766 page = pagecache_get_page(mapping, index,
1767 FGP_LOCK | FGP_WRITE | FGP_CREAT, GFP_NOFS);
3aab8f82
CY		 1768 if (!page) {
1769 err = -ENOMEM;
1770 goto fail;
1771 }
d5f66990 1772
eb47b800
JK		 1773 *pagep = page;
1774
2aadac08
JK		 1775 err = prepare_write_begin(sbi, page, pos, len,
1776 &blkaddr, &need_balance);
9ba69cf9 1777 if (err)
2aadac08 1778 goto fail;
9ba69cf9 1779
7f3037a5 1780 if (need_balance && has_not_enough_free_secs(sbi, 0, 0)) {
2a340760 1781 unlock_page(page);
2c4db1a6 1782 f2fs_balance_fs(sbi, true);
2a340760
JK		 1783 lock_page(page);
1784 if (page->mapping != mapping) {
1785 /* The page got truncated from under us */
1786 f2fs_put_page(page, 1);
1787 goto repeat;
1788 }
1789 }
1790
fec1d657 1791 f2fs_wait_on_page_writeback(page, DATA, false);
b3d208f9 1792
08b39fbd
CY		 1793 /* wait for GCed encrypted page writeback */
1794 if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
2aadac08 1795 f2fs_wait_on_encrypted_page_writeback(sbi, blkaddr);
08b39fbd 1796
649d7df2
JK		 1797 if (len == PAGE_SIZE || PageUptodate(page))
1798 return 0;
eb47b800 1799
746e2403
YH		 1800 if (!(pos & (PAGE_SIZE - 1)) && (pos + len) >= i_size_read(inode)) {
1801 zero_user_segment(page, len, PAGE_SIZE);
1802 return 0;
1803 }
1804
2aadac08 1805 if (blkaddr == NEW_ADDR) {
09cbfeaf 1806 zero_user_segment(page, 0, PAGE_SIZE);
649d7df2 1807 SetPageUptodate(page);
eb47b800 1808 } else {
78682f79 1809 struct bio *bio;
d54c795b 1810
78682f79
CY		 1811 bio = f2fs_grab_bio(inode, blkaddr, 1);
1812 if (IS_ERR(bio)) {
1813 err = PTR_ERR(bio);
3aab8f82 1814 goto fail;
eb47b800 1815 }
70fd7614 1816 bio->bi_opf = REQ_OP_READ;
78682f79
CY		 1817 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
1818 bio_put(bio);
1819 err = -EFAULT;
1820 goto fail;
1821 }
1822
4fc29c1a 1823 __submit_bio(sbi, bio, DATA);
d54c795b 1824
393ff91f 1825 lock_page(page);
6bacf52f 1826 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 1827 f2fs_put_page(page, 1);
1828 goto repeat;
eb47b800 1829 }
1563ac75
CY		 1830 if (unlikely(!PageUptodate(page))) {
1831 err = -EIO;
1832 goto fail;
4375a336 1833 }
eb47b800 1834 }
eb47b800 1835 return 0;
9ba69cf9 1836
3aab8f82 1837fail:
86531d6b 1838 f2fs_put_page(page, 1);
3aab8f82
CY		 1839 f2fs_write_failed(mapping, pos + len);
1840 return err;
eb47b800
JK		 1841}
1842
a1dd3c13
JK		 1843static int f2fs_write_end(struct file *file,
1844 struct address_space *mapping,
1845 loff_t pos, unsigned len, unsigned copied,
1846 struct page *page, void *fsdata)
1847{
1848 struct inode *inode = page->mapping->host;
1849
dfb2bf38
CY		 1850 trace_f2fs_write_end(inode, pos, len, copied);
1851
649d7df2
JK		 1852 /*
1853 * This should be come from len == PAGE_SIZE, and we expect copied
1854 * should be PAGE_SIZE. Otherwise, we treat it with zero copied and
1855 * let generic_perform_write() try to copy data again through copied=0.
1856 */
1857 if (!PageUptodate(page)) {
746e2403 1858 if (unlikely(copied != len))
649d7df2
JK		 1859 copied = 0;
1860 else
1861 SetPageUptodate(page);
1862 }
1863 if (!copied)
1864 goto unlock_out;
1865
34ba94ba 1866 set_page_dirty(page);
a1dd3c13 1867
fc9581c8
JK		 1868 if (pos + copied > i_size_read(inode))
1869 f2fs_i_size_write(inode, pos + copied);
649d7df2 1870unlock_out:
3024c9a1 1871 f2fs_put_page(page, 1);
d0239e1b 1872 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
a1dd3c13
JK		 1873 return copied;
1874}
1875
6f673763
OS		 1876static int check_direct_IO(struct inode *inode, struct iov_iter *iter,
1877 loff_t offset)
944fcfc1
JK		 1878{
1879 unsigned blocksize_mask = inode->i_sb->s_blocksize - 1;
944fcfc1 1880
944fcfc1
JK		 1881 if (offset & blocksize_mask)
1882 return -EINVAL;
1883
5b46f25d
AV		 1884 if (iov_iter_alignment(iter) & blocksize_mask)
1885 return -EINVAL;
1886
944fcfc1
JK		 1887 return 0;
1888}
1889
c8b8e32d 1890static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
eb47b800 1891{
b439b103 1892 struct address_space *mapping = iocb->ki_filp->f_mapping;
3aab8f82
CY		 1893 struct inode *inode = mapping->host;
1894 size_t count = iov_iter_count(iter);
c8b8e32d 1895 loff_t offset = iocb->ki_pos;
82e0a5aa 1896 int rw = iov_iter_rw(iter);
3aab8f82 1897 int err;
944fcfc1 1898
b439b103 1899 err = check_direct_IO(inode, iter, offset);
b9d777b8
JK		 1900 if (err)
1901 return err;
9ffe0fb5 1902
c040ff9d 1903 if (__force_buffered_io(inode, rw))
36abef4e 1904 return 0;
fcc85a4d 1905
5302fb00 1906 trace_f2fs_direct_IO_enter(inode, offset, count, rw);
70407fad 1907
82e0a5aa 1908 down_read(&F2FS_I(inode)->dio_rwsem[rw]);
c8b8e32d 1909 err = blockdev_direct_IO(iocb, inode, iter, get_data_block_dio);
82e0a5aa
CY		 1910 up_read(&F2FS_I(inode)->dio_rwsem[rw]);
1911
1912 if (rw == WRITE) {
6bfc4919 1913 if (err > 0)
91942321 1914 set_inode_flag(inode, FI_UPDATE_WRITE);
6bfc4919
JK		 1915 else if (err < 0)
1916 f2fs_write_failed(mapping, offset + count);
1917 }
70407fad 1918
5302fb00 1919 trace_f2fs_direct_IO_exit(inode, offset, count, rw, err);
70407fad 1920
3aab8f82 1921 return err;
eb47b800
JK		 1922}
1923
487261f3
CY		 1924void f2fs_invalidate_page(struct page *page, unsigned int offset,
1925 unsigned int length)
eb47b800
JK		 1926{
1927 struct inode *inode = page->mapping->host;
487261f3 1928 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
a7ffdbe2 1929
487261f3 1930 if (inode->i_ino >= F2FS_ROOT_INO(sbi) &&
09cbfeaf 1931 (offset % PAGE_SIZE || length != PAGE_SIZE))
a7ffdbe2
JK		 1932 return;
1933
487261f3 1934 if (PageDirty(page)) {
933439c8 1935 if (inode->i_ino == F2FS_META_INO(sbi)) {
487261f3 1936 dec_page_count(sbi, F2FS_DIRTY_META);
933439c8 1937 } else if (inode->i_ino == F2FS_NODE_INO(sbi)) {
487261f3 1938 dec_page_count(sbi, F2FS_DIRTY_NODES);
933439c8 1939 } else {
487261f3 1940 inode_dec_dirty_pages(inode);
933439c8
CY		 1941 remove_dirty_inode(inode);
1942 }
487261f3 1943 }
decd36b6
CY		 1944
1945 /* This is atomic written page, keep Private */
1946 if (IS_ATOMIC_WRITTEN_PAGE(page))
1947 return;
1948
23dc974e 1949 set_page_private(page, 0);
eb47b800
JK		 1950 ClearPagePrivate(page);
1951}
1952
487261f3 1953int f2fs_release_page(struct page *page, gfp_t wait)
eb47b800 1954{
f68daeeb
JK		 1955 /* If this is dirty page, keep PagePrivate */
1956 if (PageDirty(page))
1957 return 0;
1958
decd36b6
CY		 1959 /* This is atomic written page, keep Private */
1960 if (IS_ATOMIC_WRITTEN_PAGE(page))
1961 return 0;
1962
23dc974e 1963 set_page_private(page, 0);
eb47b800 1964 ClearPagePrivate(page);
c3850aa1 1965 return 1;
eb47b800
JK		 1966}
1967
fe76b796
JK		 1968/*
1969 * This was copied from __set_page_dirty_buffers which gives higher performance
1970 * in very high speed storages. (e.g., pmem)
1971 */
1972void f2fs_set_page_dirty_nobuffers(struct page *page)
1973{
1974 struct address_space *mapping = page->mapping;
1975 unsigned long flags;
1976
1977 if (unlikely(!mapping))
1978 return;
1979
1980 spin_lock(&mapping->private_lock);
1981 lock_page_memcg(page);
1982 SetPageDirty(page);
1983 spin_unlock(&mapping->private_lock);
1984
1985 spin_lock_irqsave(&mapping->tree_lock, flags);
1986 WARN_ON_ONCE(!PageUptodate(page));
1987 account_page_dirtied(page, mapping);
1988 radix_tree_tag_set(&mapping->page_tree,
1989 page_index(page), PAGECACHE_TAG_DIRTY);
1990 spin_unlock_irqrestore(&mapping->tree_lock, flags);
1991 unlock_page_memcg(page);
1992
1993 __mark_inode_dirty(mapping->host, I_DIRTY_PAGES);
1994 return;
1995}
1996
eb47b800
JK		 1997static int f2fs_set_data_page_dirty(struct page *page)
1998{
1999 struct address_space *mapping = page->mapping;
2000 struct inode *inode = mapping->host;
2001
26c6b887
JK		 2002 trace_f2fs_set_page_dirty(page, DATA);
2003
237c0790
JK		 2004 if (!PageUptodate(page))
2005 SetPageUptodate(page);
34ba94ba 2006
5fe45743 2007 if (f2fs_is_atomic_file(inode) && !f2fs_is_commit_atomic_write(inode)) {
decd36b6
CY		 2008 if (!IS_ATOMIC_WRITTEN_PAGE(page)) {
2009 register_inmem_page(inode, page);
2010 return 1;
2011 }
2012 /*
2013 * Previously, this page has been registered, we just
2014 * return here.
2015 */
2016 return 0;
34ba94ba
JK		 2017 }
2018
eb47b800 2019 if (!PageDirty(page)) {
fe76b796 2020 f2fs_set_page_dirty_nobuffers(page);
a7ffdbe2 2021 update_dirty_page(inode, page);
eb47b800
JK		 2022 return 1;
2023 }
2024 return 0;
2025}
2026
c01e54b7
JK		 2027static sector_t f2fs_bmap(struct address_space *mapping, sector_t block)
2028{
454ae7e5
CY		 2029 struct inode *inode = mapping->host;
2030
1d373a0e
JK		 2031 if (f2fs_has_inline_data(inode))
2032 return 0;
2033
2034 /* make sure allocating whole blocks */
2035 if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
2036 filemap_write_and_wait(mapping);
2037
e2b4e2bc 2038 return generic_block_bmap(mapping, block, get_data_block_bmap);
429511cd
CY		 2039}
2040
5b7a487c
WG		 2041#ifdef CONFIG_MIGRATION
2042#include <linux/migrate.h>
2043
2044int f2fs_migrate_page(struct address_space *mapping,
2045 struct page *newpage, struct page *page, enum migrate_mode mode)
2046{
2047 int rc, extra_count;
2048 struct f2fs_inode_info *fi = F2FS_I(mapping->host);
2049 bool atomic_written = IS_ATOMIC_WRITTEN_PAGE(page);
2050
2051 BUG_ON(PageWriteback(page));
2052
2053 /* migrating an atomic written page is safe with the inmem_lock hold */
2054 if (atomic_written && !mutex_trylock(&fi->inmem_lock))
2055 return -EAGAIN;
2056
2057 /*
2058 * A reference is expected if PagePrivate set when move mapping,
2059 * however F2FS breaks this for maintaining dirty page counts when
2060 * truncating pages. So here adjusting the 'extra_count' make it work.
2061 */
2062 extra_count = (atomic_written ? 1 : 0) - page_has_private(page);
2063 rc = migrate_page_move_mapping(mapping, newpage,
2064 page, NULL, mode, extra_count);
2065 if (rc != MIGRATEPAGE_SUCCESS) {
2066 if (atomic_written)
2067 mutex_unlock(&fi->inmem_lock);
2068 return rc;
2069 }
2070
2071 if (atomic_written) {
2072 struct inmem_pages *cur;
2073 list_for_each_entry(cur, &fi->inmem_pages, list)
2074 if (cur->page == page) {
2075 cur->page = newpage;
2076 break;
2077 }
2078 mutex_unlock(&fi->inmem_lock);
2079 put_page(page);
2080 get_page(newpage);
2081 }
2082
2083 if (PagePrivate(page))
2084 SetPagePrivate(newpage);
2085 set_page_private(newpage, page_private(page));
2086
2087 migrate_page_copy(newpage, page);
2088
2089 return MIGRATEPAGE_SUCCESS;
2090}
2091#endif
2092
eb47b800
JK		 2093const struct address_space_operations f2fs_dblock_aops = {
2094 .readpage = f2fs_read_data_page,
2095 .readpages = f2fs_read_data_pages,
2096 .writepage = f2fs_write_data_page,
2097 .writepages = f2fs_write_data_pages,
2098 .write_begin = f2fs_write_begin,
a1dd3c13 2099 .write_end = f2fs_write_end,
eb47b800 2100 .set_page_dirty = f2fs_set_data_page_dirty,
487261f3
CY		 2101 .invalidatepage = f2fs_invalidate_page,
2102 .releasepage = f2fs_release_page,
eb47b800 2103 .direct_IO = f2fs_direct_IO,
c01e54b7 2104 .bmap = f2fs_bmap,
5b7a487c
WG		 2105#ifdef CONFIG_MIGRATION
2106 .migratepage = f2fs_migrate_page,
2107#endif
eb47b800 2108};
Ubuntu Artful kernel mirror