Commit | Line | Data |
---|---|---|
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> |
f1e88660 | 22 | #include <linux/cleancache.h> |
eb47b800 JK |
23 | |
24 | #include "f2fs.h" | |
25 | #include "node.h" | |
26 | #include "segment.h" | |
db9f7c1a | 27 | #include "trace.h" |
848753aa | 28 | #include <trace/events/f2fs.h> |
eb47b800 | 29 | |
4246a0b6 | 30 | static void f2fs_read_end_io(struct bio *bio) |
93dfe2ac | 31 | { |
f568849e LT |
32 | struct bio_vec *bvec; |
33 | int i; | |
93dfe2ac | 34 | |
4375a336 | 35 | if (f2fs_bio_encrypted(bio)) { |
4246a0b6 | 36 | if (bio->bi_error) { |
4375a336 JK |
37 | f2fs_release_crypto_ctx(bio->bi_private); |
38 | } else { | |
39 | f2fs_end_io_crypto_work(bio->bi_private, bio); | |
40 | return; | |
41 | } | |
42 | } | |
43 | ||
12377024 CY |
44 | bio_for_each_segment_all(bvec, bio, i) { |
45 | struct page *page = bvec->bv_page; | |
f1e88660 | 46 | |
4246a0b6 | 47 | if (!bio->bi_error) { |
f1e88660 JK |
48 | SetPageUptodate(page); |
49 | } else { | |
50 | ClearPageUptodate(page); | |
51 | SetPageError(page); | |
52 | } | |
53 | unlock_page(page); | |
54 | } | |
f1e88660 JK |
55 | bio_put(bio); |
56 | } | |
57 | ||
4246a0b6 | 58 | static void f2fs_write_end_io(struct bio *bio) |
93dfe2ac | 59 | { |
1b1f559f | 60 | struct f2fs_sb_info *sbi = bio->bi_private; |
f568849e LT |
61 | struct bio_vec *bvec; |
62 | int i; | |
93dfe2ac | 63 | |
f568849e | 64 | bio_for_each_segment_all(bvec, bio, i) { |
93dfe2ac JK |
65 | struct page *page = bvec->bv_page; |
66 | ||
4375a336 JK |
67 | f2fs_restore_and_release_control_page(&page); |
68 | ||
4246a0b6 | 69 | if (unlikely(bio->bi_error)) { |
93dfe2ac | 70 | set_bit(AS_EIO, &page->mapping->flags); |
744602cf | 71 | f2fs_stop_checkpoint(sbi); |
93dfe2ac JK |
72 | } |
73 | end_page_writeback(page); | |
74 | dec_page_count(sbi, F2FS_WRITEBACK); | |
f568849e | 75 | } |
93dfe2ac | 76 | |
7c506896 | 77 | if (!get_pages(sbi, F2FS_WRITEBACK) && wq_has_sleeper(&sbi->cp_wait)) |
93dfe2ac JK |
78 | wake_up(&sbi->cp_wait); |
79 | ||
80 | bio_put(bio); | |
81 | } | |
82 | ||
940a6d34 GZ |
83 | /* |
84 | * Low-level block read/write IO operations. | |
85 | */ | |
86 | static struct bio *__bio_alloc(struct f2fs_sb_info *sbi, block_t blk_addr, | |
87 | int npages, bool is_read) | |
88 | { | |
89 | struct bio *bio; | |
90 | ||
740432f8 | 91 | bio = f2fs_bio_alloc(npages); |
940a6d34 GZ |
92 | |
93 | bio->bi_bdev = sbi->sb->s_bdev; | |
55cf9cb6 | 94 | bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr); |
940a6d34 | 95 | bio->bi_end_io = is_read ? f2fs_read_end_io : f2fs_write_end_io; |
12377024 | 96 | bio->bi_private = is_read ? NULL : sbi; |
940a6d34 GZ |
97 | |
98 | return bio; | |
99 | } | |
100 | ||
458e6197 | 101 | static void __submit_merged_bio(struct f2fs_bio_info *io) |
93dfe2ac | 102 | { |
458e6197 | 103 | struct f2fs_io_info *fio = &io->fio; |
93dfe2ac JK |
104 | |
105 | if (!io->bio) | |
106 | return; | |
107 | ||
6a8f8ca5 | 108 | if (is_read_io(fio->rw)) |
2ace38e0 | 109 | trace_f2fs_submit_read_bio(io->sbi->sb, fio, io->bio); |
6a8f8ca5 | 110 | else |
2ace38e0 | 111 | trace_f2fs_submit_write_bio(io->sbi->sb, fio, io->bio); |
940a6d34 | 112 | |
6a8f8ca5 | 113 | submit_bio(fio->rw, io->bio); |
93dfe2ac JK |
114 | io->bio = NULL; |
115 | } | |
116 | ||
0fd785eb CY |
117 | bool is_merged_page(struct f2fs_sb_info *sbi, struct page *page, |
118 | enum page_type type) | |
119 | { | |
120 | enum page_type btype = PAGE_TYPE_OF_BIO(type); | |
121 | struct f2fs_bio_info *io = &sbi->write_io[btype]; | |
122 | struct bio_vec *bvec; | |
123 | struct page *target; | |
124 | int i; | |
125 | ||
126 | down_read(&io->io_rwsem); | |
127 | if (!io->bio) { | |
128 | up_read(&io->io_rwsem); | |
129 | return false; | |
130 | } | |
131 | ||
132 | bio_for_each_segment_all(bvec, io->bio, i) { | |
133 | ||
134 | if (bvec->bv_page->mapping) { | |
135 | target = bvec->bv_page; | |
136 | } else { | |
137 | struct f2fs_crypto_ctx *ctx; | |
138 | ||
139 | /* encrypted page */ | |
140 | ctx = (struct f2fs_crypto_ctx *)page_private( | |
141 | bvec->bv_page); | |
142 | target = ctx->w.control_page; | |
143 | } | |
144 | ||
145 | if (page == target) { | |
146 | up_read(&io->io_rwsem); | |
147 | return true; | |
148 | } | |
149 | } | |
150 | ||
151 | up_read(&io->io_rwsem); | |
152 | return false; | |
153 | } | |
154 | ||
93dfe2ac | 155 | void f2fs_submit_merged_bio(struct f2fs_sb_info *sbi, |
458e6197 | 156 | enum page_type type, int rw) |
93dfe2ac JK |
157 | { |
158 | enum page_type btype = PAGE_TYPE_OF_BIO(type); | |
159 | struct f2fs_bio_info *io; | |
160 | ||
161 | io = is_read_io(rw) ? &sbi->read_io : &sbi->write_io[btype]; | |
162 | ||
df0f8dc0 | 163 | down_write(&io->io_rwsem); |
458e6197 JK |
164 | |
165 | /* change META to META_FLUSH in the checkpoint procedure */ | |
166 | if (type >= META_FLUSH) { | |
167 | io->fio.type = META_FLUSH; | |
0f7b2abd JK |
168 | if (test_opt(sbi, NOBARRIER)) |
169 | io->fio.rw = WRITE_FLUSH | REQ_META | REQ_PRIO; | |
170 | else | |
171 | io->fio.rw = WRITE_FLUSH_FUA | REQ_META | REQ_PRIO; | |
458e6197 JK |
172 | } |
173 | __submit_merged_bio(io); | |
df0f8dc0 | 174 | up_write(&io->io_rwsem); |
93dfe2ac JK |
175 | } |
176 | ||
177 | /* | |
178 | * Fill the locked page with data located in the block address. | |
179 | * Return unlocked page. | |
180 | */ | |
05ca3632 | 181 | int f2fs_submit_page_bio(struct f2fs_io_info *fio) |
93dfe2ac | 182 | { |
93dfe2ac | 183 | struct bio *bio; |
4375a336 | 184 | struct page *page = fio->encrypted_page ? fio->encrypted_page : fio->page; |
93dfe2ac | 185 | |
2ace38e0 | 186 | trace_f2fs_submit_page_bio(page, fio); |
05ca3632 | 187 | f2fs_trace_ios(fio, 0); |
93dfe2ac JK |
188 | |
189 | /* Allocate a new bio */ | |
05ca3632 | 190 | bio = __bio_alloc(fio->sbi, fio->blk_addr, 1, is_read_io(fio->rw)); |
93dfe2ac JK |
191 | |
192 | if (bio_add_page(bio, page, PAGE_CACHE_SIZE, 0) < PAGE_CACHE_SIZE) { | |
193 | bio_put(bio); | |
93dfe2ac JK |
194 | return -EFAULT; |
195 | } | |
196 | ||
cf04e8eb | 197 | submit_bio(fio->rw, bio); |
93dfe2ac JK |
198 | return 0; |
199 | } | |
200 | ||
05ca3632 | 201 | void f2fs_submit_page_mbio(struct f2fs_io_info *fio) |
93dfe2ac | 202 | { |
05ca3632 | 203 | struct f2fs_sb_info *sbi = fio->sbi; |
458e6197 | 204 | enum page_type btype = PAGE_TYPE_OF_BIO(fio->type); |
93dfe2ac | 205 | struct f2fs_bio_info *io; |
940a6d34 | 206 | bool is_read = is_read_io(fio->rw); |
4375a336 | 207 | struct page *bio_page; |
93dfe2ac | 208 | |
940a6d34 | 209 | io = is_read ? &sbi->read_io : &sbi->write_io[btype]; |
93dfe2ac | 210 | |
cf04e8eb | 211 | verify_block_addr(sbi, fio->blk_addr); |
93dfe2ac | 212 | |
df0f8dc0 | 213 | down_write(&io->io_rwsem); |
93dfe2ac | 214 | |
940a6d34 | 215 | if (!is_read) |
93dfe2ac JK |
216 | inc_page_count(sbi, F2FS_WRITEBACK); |
217 | ||
cf04e8eb | 218 | if (io->bio && (io->last_block_in_bio != fio->blk_addr - 1 || |
458e6197 JK |
219 | io->fio.rw != fio->rw)) |
220 | __submit_merged_bio(io); | |
93dfe2ac JK |
221 | alloc_new: |
222 | if (io->bio == NULL) { | |
90a893c7 | 223 | int bio_blocks = MAX_BIO_BLOCKS(sbi); |
940a6d34 | 224 | |
cf04e8eb | 225 | io->bio = __bio_alloc(sbi, fio->blk_addr, bio_blocks, is_read); |
458e6197 | 226 | io->fio = *fio; |
93dfe2ac JK |
227 | } |
228 | ||
4375a336 JK |
229 | bio_page = fio->encrypted_page ? fio->encrypted_page : fio->page; |
230 | ||
231 | if (bio_add_page(io->bio, bio_page, PAGE_CACHE_SIZE, 0) < | |
93dfe2ac | 232 | PAGE_CACHE_SIZE) { |
458e6197 | 233 | __submit_merged_bio(io); |
93dfe2ac JK |
234 | goto alloc_new; |
235 | } | |
236 | ||
cf04e8eb | 237 | io->last_block_in_bio = fio->blk_addr; |
05ca3632 | 238 | f2fs_trace_ios(fio, 0); |
93dfe2ac | 239 | |
df0f8dc0 | 240 | up_write(&io->io_rwsem); |
05ca3632 | 241 | trace_f2fs_submit_page_mbio(fio->page, fio); |
93dfe2ac JK |
242 | } |
243 | ||
0a8165d7 | 244 | /* |
eb47b800 JK |
245 | * Lock ordering for the change of data block address: |
246 | * ->data_page | |
247 | * ->node_page | |
248 | * update block addresses in the node page | |
249 | */ | |
216a620a | 250 | void set_data_blkaddr(struct dnode_of_data *dn) |
eb47b800 JK |
251 | { |
252 | struct f2fs_node *rn; | |
253 | __le32 *addr_array; | |
254 | struct page *node_page = dn->node_page; | |
255 | unsigned int ofs_in_node = dn->ofs_in_node; | |
256 | ||
fec1d657 | 257 | f2fs_wait_on_page_writeback(node_page, NODE, true); |
eb47b800 | 258 | |
45590710 | 259 | rn = F2FS_NODE(node_page); |
eb47b800 JK |
260 | |
261 | /* Get physical address of data block */ | |
262 | addr_array = blkaddr_in_node(rn); | |
e1509cf2 | 263 | addr_array[ofs_in_node] = cpu_to_le32(dn->data_blkaddr); |
12719ae1 JK |
264 | if (set_page_dirty(node_page)) |
265 | dn->node_changed = true; | |
eb47b800 JK |
266 | } |
267 | ||
268 | int reserve_new_block(struct dnode_of_data *dn) | |
269 | { | |
4081363f | 270 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
eb47b800 | 271 | |
6bacf52f | 272 | if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))) |
eb47b800 | 273 | return -EPERM; |
cfb271d4 | 274 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, 1))) |
eb47b800 JK |
275 | return -ENOSPC; |
276 | ||
c01e2853 NJ |
277 | trace_f2fs_reserve_new_block(dn->inode, dn->nid, dn->ofs_in_node); |
278 | ||
eb47b800 | 279 | dn->data_blkaddr = NEW_ADDR; |
216a620a | 280 | set_data_blkaddr(dn); |
a18ff063 | 281 | mark_inode_dirty(dn->inode); |
eb47b800 JK |
282 | sync_inode_page(dn); |
283 | return 0; | |
284 | } | |
285 | ||
b600965c HL |
286 | int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index) |
287 | { | |
288 | bool need_put = dn->inode_page ? false : true; | |
289 | int err; | |
290 | ||
291 | err = get_dnode_of_data(dn, index, ALLOC_NODE); | |
292 | if (err) | |
293 | return err; | |
a8865372 | 294 | |
b600965c HL |
295 | if (dn->data_blkaddr == NULL_ADDR) |
296 | err = reserve_new_block(dn); | |
a8865372 | 297 | if (err || need_put) |
b600965c HL |
298 | f2fs_put_dnode(dn); |
299 | return err; | |
300 | } | |
301 | ||
759af1c9 | 302 | int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index) |
eb47b800 | 303 | { |
028a41e8 | 304 | struct extent_info ei; |
759af1c9 | 305 | struct inode *inode = dn->inode; |
028a41e8 | 306 | |
759af1c9 FL |
307 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { |
308 | dn->data_blkaddr = ei.blk + index - ei.fofs; | |
309 | return 0; | |
429511cd | 310 | } |
028a41e8 | 311 | |
759af1c9 | 312 | return f2fs_reserve_block(dn, index); |
eb47b800 JK |
313 | } |
314 | ||
a56c7c6f JK |
315 | struct page *get_read_data_page(struct inode *inode, pgoff_t index, |
316 | int rw, bool for_write) | |
eb47b800 | 317 | { |
eb47b800 JK |
318 | struct address_space *mapping = inode->i_mapping; |
319 | struct dnode_of_data dn; | |
320 | struct page *page; | |
cb3bc9ee | 321 | struct extent_info ei; |
eb47b800 | 322 | int err; |
cf04e8eb | 323 | struct f2fs_io_info fio = { |
05ca3632 | 324 | .sbi = F2FS_I_SB(inode), |
cf04e8eb | 325 | .type = DATA, |
43f3eae1 | 326 | .rw = rw, |
4375a336 | 327 | .encrypted_page = NULL, |
cf04e8eb | 328 | }; |
eb47b800 | 329 | |
4375a336 JK |
330 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) |
331 | return read_mapping_page(mapping, index, NULL); | |
332 | ||
a56c7c6f | 333 | page = f2fs_grab_cache_page(mapping, index, for_write); |
650495de JK |
334 | if (!page) |
335 | return ERR_PTR(-ENOMEM); | |
336 | ||
cb3bc9ee CY |
337 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { |
338 | dn.data_blkaddr = ei.blk + index - ei.fofs; | |
339 | goto got_it; | |
340 | } | |
341 | ||
eb47b800 | 342 | set_new_dnode(&dn, inode, NULL, NULL, 0); |
266e97a8 | 343 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); |
86531d6b JK |
344 | if (err) |
345 | goto put_err; | |
eb47b800 JK |
346 | f2fs_put_dnode(&dn); |
347 | ||
6bacf52f | 348 | if (unlikely(dn.data_blkaddr == NULL_ADDR)) { |
86531d6b JK |
349 | err = -ENOENT; |
350 | goto put_err; | |
650495de | 351 | } |
cb3bc9ee | 352 | got_it: |
43f3eae1 JK |
353 | if (PageUptodate(page)) { |
354 | unlock_page(page); | |
eb47b800 | 355 | return page; |
43f3eae1 | 356 | } |
eb47b800 | 357 | |
d59ff4df JK |
358 | /* |
359 | * A new dentry page is allocated but not able to be written, since its | |
360 | * new inode page couldn't be allocated due to -ENOSPC. | |
361 | * In such the case, its blkaddr can be remained as NEW_ADDR. | |
362 | * see, f2fs_add_link -> get_new_data_page -> init_inode_metadata. | |
363 | */ | |
364 | if (dn.data_blkaddr == NEW_ADDR) { | |
365 | zero_user_segment(page, 0, PAGE_CACHE_SIZE); | |
366 | SetPageUptodate(page); | |
43f3eae1 | 367 | unlock_page(page); |
d59ff4df JK |
368 | return page; |
369 | } | |
eb47b800 | 370 | |
cf04e8eb | 371 | fio.blk_addr = dn.data_blkaddr; |
05ca3632 JK |
372 | fio.page = page; |
373 | err = f2fs_submit_page_bio(&fio); | |
393ff91f | 374 | if (err) |
86531d6b | 375 | goto put_err; |
43f3eae1 | 376 | return page; |
86531d6b JK |
377 | |
378 | put_err: | |
379 | f2fs_put_page(page, 1); | |
380 | return ERR_PTR(err); | |
43f3eae1 JK |
381 | } |
382 | ||
383 | struct page *find_data_page(struct inode *inode, pgoff_t index) | |
384 | { | |
385 | struct address_space *mapping = inode->i_mapping; | |
386 | struct page *page; | |
387 | ||
388 | page = find_get_page(mapping, index); | |
389 | if (page && PageUptodate(page)) | |
390 | return page; | |
391 | f2fs_put_page(page, 0); | |
392 | ||
a56c7c6f | 393 | page = get_read_data_page(inode, index, READ_SYNC, false); |
43f3eae1 JK |
394 | if (IS_ERR(page)) |
395 | return page; | |
396 | ||
397 | if (PageUptodate(page)) | |
398 | return page; | |
399 | ||
400 | wait_on_page_locked(page); | |
401 | if (unlikely(!PageUptodate(page))) { | |
402 | f2fs_put_page(page, 0); | |
403 | return ERR_PTR(-EIO); | |
404 | } | |
405 | return page; | |
406 | } | |
407 | ||
408 | /* | |
409 | * If it tries to access a hole, return an error. | |
410 | * Because, the callers, functions in dir.c and GC, should be able to know | |
411 | * whether this page exists or not. | |
412 | */ | |
a56c7c6f JK |
413 | struct page *get_lock_data_page(struct inode *inode, pgoff_t index, |
414 | bool for_write) | |
43f3eae1 JK |
415 | { |
416 | struct address_space *mapping = inode->i_mapping; | |
417 | struct page *page; | |
418 | repeat: | |
a56c7c6f | 419 | page = get_read_data_page(inode, index, READ_SYNC, for_write); |
43f3eae1 JK |
420 | if (IS_ERR(page)) |
421 | return page; | |
393ff91f | 422 | |
43f3eae1 | 423 | /* wait for read completion */ |
393ff91f | 424 | lock_page(page); |
6bacf52f | 425 | if (unlikely(!PageUptodate(page))) { |
393ff91f JK |
426 | f2fs_put_page(page, 1); |
427 | return ERR_PTR(-EIO); | |
eb47b800 | 428 | } |
6bacf52f | 429 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
430 | f2fs_put_page(page, 1); |
431 | goto repeat; | |
eb47b800 JK |
432 | } |
433 | return page; | |
434 | } | |
435 | ||
0a8165d7 | 436 | /* |
eb47b800 JK |
437 | * Caller ensures that this data page is never allocated. |
438 | * A new zero-filled data page is allocated in the page cache. | |
39936837 | 439 | * |
4f4124d0 CY |
440 | * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and |
441 | * f2fs_unlock_op(). | |
470f00e9 CY |
442 | * Note that, ipage is set only by make_empty_dir, and if any error occur, |
443 | * ipage should be released by this function. | |
eb47b800 | 444 | */ |
64aa7ed9 | 445 | struct page *get_new_data_page(struct inode *inode, |
a8865372 | 446 | struct page *ipage, pgoff_t index, bool new_i_size) |
eb47b800 | 447 | { |
eb47b800 JK |
448 | struct address_space *mapping = inode->i_mapping; |
449 | struct page *page; | |
450 | struct dnode_of_data dn; | |
451 | int err; | |
7612118a | 452 | |
a56c7c6f | 453 | page = f2fs_grab_cache_page(mapping, index, true); |
470f00e9 CY |
454 | if (!page) { |
455 | /* | |
456 | * before exiting, we should make sure ipage will be released | |
457 | * if any error occur. | |
458 | */ | |
459 | f2fs_put_page(ipage, 1); | |
01f28610 | 460 | return ERR_PTR(-ENOMEM); |
470f00e9 | 461 | } |
eb47b800 | 462 | |
a8865372 | 463 | set_new_dnode(&dn, inode, ipage, NULL, 0); |
b600965c | 464 | err = f2fs_reserve_block(&dn, index); |
01f28610 JK |
465 | if (err) { |
466 | f2fs_put_page(page, 1); | |
eb47b800 | 467 | return ERR_PTR(err); |
a8865372 | 468 | } |
01f28610 JK |
469 | if (!ipage) |
470 | f2fs_put_dnode(&dn); | |
eb47b800 JK |
471 | |
472 | if (PageUptodate(page)) | |
01f28610 | 473 | goto got_it; |
eb47b800 JK |
474 | |
475 | if (dn.data_blkaddr == NEW_ADDR) { | |
476 | zero_user_segment(page, 0, PAGE_CACHE_SIZE); | |
393ff91f | 477 | SetPageUptodate(page); |
eb47b800 | 478 | } else { |
4375a336 | 479 | f2fs_put_page(page, 1); |
a8865372 | 480 | |
7612118a JK |
481 | /* if ipage exists, blkaddr should be NEW_ADDR */ |
482 | f2fs_bug_on(F2FS_I_SB(inode), ipage); | |
483 | page = get_lock_data_page(inode, index, true); | |
4375a336 | 484 | if (IS_ERR(page)) |
7612118a | 485 | return page; |
eb47b800 | 486 | } |
01f28610 | 487 | got_it: |
9edcdabf CY |
488 | if (new_i_size && i_size_read(inode) < |
489 | ((loff_t)(index + 1) << PAGE_CACHE_SHIFT)) { | |
490 | i_size_write(inode, ((loff_t)(index + 1) << PAGE_CACHE_SHIFT)); | |
699489bb JK |
491 | /* Only the directory inode sets new_i_size */ |
492 | set_inode_flag(F2FS_I(inode), FI_UPDATE_DIR); | |
eb47b800 JK |
493 | } |
494 | return page; | |
495 | } | |
496 | ||
bfad7c2d JK |
497 | static int __allocate_data_block(struct dnode_of_data *dn) |
498 | { | |
4081363f | 499 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
bfad7c2d | 500 | struct f2fs_summary sum; |
bfad7c2d | 501 | struct node_info ni; |
38aa0889 | 502 | int seg = CURSEG_WARM_DATA; |
976e4c50 | 503 | pgoff_t fofs; |
bfad7c2d JK |
504 | |
505 | if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))) | |
506 | return -EPERM; | |
df6136ef CY |
507 | |
508 | dn->data_blkaddr = datablock_addr(dn->node_page, dn->ofs_in_node); | |
509 | if (dn->data_blkaddr == NEW_ADDR) | |
510 | goto alloc; | |
511 | ||
bfad7c2d JK |
512 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, 1))) |
513 | return -ENOSPC; | |
514 | ||
df6136ef | 515 | alloc: |
bfad7c2d JK |
516 | get_node_info(sbi, dn->nid, &ni); |
517 | set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version); | |
518 | ||
38aa0889 JK |
519 | if (dn->ofs_in_node == 0 && dn->inode_page == dn->node_page) |
520 | seg = CURSEG_DIRECT_IO; | |
521 | ||
df6136ef CY |
522 | allocate_data_block(sbi, NULL, dn->data_blkaddr, &dn->data_blkaddr, |
523 | &sum, seg); | |
216a620a | 524 | set_data_blkaddr(dn); |
bfad7c2d | 525 | |
976e4c50 | 526 | /* update i_size */ |
81ca7350 | 527 | fofs = start_bidx_of_node(ofs_of_node(dn->node_page), dn->inode) + |
976e4c50 | 528 | dn->ofs_in_node; |
9edcdabf CY |
529 | if (i_size_read(dn->inode) < ((loff_t)(fofs + 1) << PAGE_CACHE_SHIFT)) |
530 | i_size_write(dn->inode, | |
531 | ((loff_t)(fofs + 1) << PAGE_CACHE_SHIFT)); | |
bfad7c2d JK |
532 | return 0; |
533 | } | |
534 | ||
6d5a1495 | 535 | static int __allocate_data_blocks(struct inode *inode, loff_t offset, |
59b802e5 JK |
536 | size_t count) |
537 | { | |
5b8db7fa | 538 | struct f2fs_map_blocks map; |
59b802e5 | 539 | |
5b8db7fa CY |
540 | map.m_lblk = F2FS_BYTES_TO_BLK(offset); |
541 | map.m_len = F2FS_BYTES_TO_BLK(count); | |
2a340760 | 542 | |
5b8db7fa | 543 | return f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_DIO); |
59b802e5 JK |
544 | } |
545 | ||
0a8165d7 | 546 | /* |
003a3e1d JK |
547 | * f2fs_map_blocks() now supported readahead/bmap/rw direct_IO with |
548 | * f2fs_map_blocks structure. | |
4f4124d0 CY |
549 | * If original data blocks are allocated, then give them to blockdev. |
550 | * Otherwise, | |
551 | * a. preallocate requested block addresses | |
552 | * b. do not use extent cache for better performance | |
553 | * c. give the block addresses to blockdev | |
eb47b800 | 554 | */ |
d323d005 | 555 | int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map, |
e2b4e2bc | 556 | int create, int flag) |
eb47b800 | 557 | { |
003a3e1d | 558 | unsigned int maxblocks = map->m_len; |
eb47b800 | 559 | struct dnode_of_data dn; |
f9811703 | 560 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
bfad7c2d JK |
561 | int mode = create ? ALLOC_NODE : LOOKUP_NODE_RA; |
562 | pgoff_t pgofs, end_offset; | |
563 | int err = 0, ofs = 1; | |
a2e7d1bf | 564 | struct extent_info ei; |
bfad7c2d | 565 | bool allocated = false; |
7df3a431 | 566 | block_t blkaddr; |
eb47b800 | 567 | |
003a3e1d JK |
568 | map->m_len = 0; |
569 | map->m_flags = 0; | |
570 | ||
571 | /* it only supports block size == page size */ | |
572 | pgofs = (pgoff_t)map->m_lblk; | |
eb47b800 | 573 | |
7e4dde79 | 574 | if (f2fs_lookup_extent_cache(inode, pgofs, &ei)) { |
003a3e1d JK |
575 | map->m_pblk = ei.blk + pgofs - ei.fofs; |
576 | map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs); | |
577 | map->m_flags = F2FS_MAP_MAPPED; | |
bfad7c2d | 578 | goto out; |
a2e7d1bf | 579 | } |
bfad7c2d | 580 | |
4fe71e88 | 581 | next_dnode: |
59b802e5 | 582 | if (create) |
3104af35 | 583 | f2fs_lock_op(sbi); |
eb47b800 JK |
584 | |
585 | /* When reading holes, we need its node page */ | |
586 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
bfad7c2d | 587 | err = get_dnode_of_data(&dn, pgofs, mode); |
1ec79083 | 588 | if (err) { |
bfad7c2d JK |
589 | if (err == -ENOENT) |
590 | err = 0; | |
591 | goto unlock_out; | |
848753aa | 592 | } |
973163fc | 593 | |
81ca7350 | 594 | end_offset = ADDRS_PER_PAGE(dn.node_page, inode); |
4fe71e88 CY |
595 | |
596 | next_block: | |
597 | blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node); | |
598 | ||
599 | if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR) { | |
973163fc | 600 | if (create) { |
f9811703 CY |
601 | if (unlikely(f2fs_cp_error(sbi))) { |
602 | err = -EIO; | |
4fe71e88 | 603 | goto sync_out; |
f9811703 | 604 | } |
973163fc CY |
605 | err = __allocate_data_block(&dn); |
606 | if (err) | |
4fe71e88 | 607 | goto sync_out; |
973163fc CY |
608 | allocated = true; |
609 | map->m_flags = F2FS_MAP_NEW; | |
4fe71e88 | 610 | blkaddr = dn.data_blkaddr; |
973163fc CY |
611 | } else { |
612 | if (flag != F2FS_GET_BLOCK_FIEMAP || | |
4fe71e88 | 613 | blkaddr != NEW_ADDR) { |
973163fc CY |
614 | if (flag == F2FS_GET_BLOCK_BMAP) |
615 | err = -ENOENT; | |
4fe71e88 | 616 | goto sync_out; |
973163fc | 617 | } |
e2b4e2bc | 618 | } |
e2b4e2bc | 619 | } |
eb47b800 | 620 | |
4fe71e88 CY |
621 | if (map->m_len == 0) { |
622 | /* preallocated unwritten block should be mapped for fiemap. */ | |
623 | if (blkaddr == NEW_ADDR) | |
624 | map->m_flags |= F2FS_MAP_UNWRITTEN; | |
625 | map->m_flags |= F2FS_MAP_MAPPED; | |
626 | ||
627 | map->m_pblk = blkaddr; | |
628 | map->m_len = 1; | |
629 | } else if ((map->m_pblk != NEW_ADDR && | |
630 | blkaddr == (map->m_pblk + ofs)) || | |
631 | (map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR)) { | |
632 | ofs++; | |
633 | map->m_len++; | |
634 | } else { | |
635 | goto sync_out; | |
636 | } | |
bfad7c2d | 637 | |
bfad7c2d JK |
638 | dn.ofs_in_node++; |
639 | pgofs++; | |
640 | ||
4fe71e88 CY |
641 | if (map->m_len < maxblocks) { |
642 | if (dn.ofs_in_node < end_offset) | |
643 | goto next_block; | |
7df3a431 | 644 | |
bfad7c2d JK |
645 | if (allocated) |
646 | sync_inode_page(&dn); | |
bfad7c2d JK |
647 | f2fs_put_dnode(&dn); |
648 | ||
3104af35 CY |
649 | if (create) { |
650 | f2fs_unlock_op(sbi); | |
3c082b7b | 651 | f2fs_balance_fs(sbi, allocated); |
3104af35 | 652 | } |
3c082b7b | 653 | allocated = false; |
4fe71e88 | 654 | goto next_dnode; |
eb47b800 | 655 | } |
7df3a431 | 656 | |
bfad7c2d JK |
657 | sync_out: |
658 | if (allocated) | |
659 | sync_inode_page(&dn); | |
eb47b800 | 660 | f2fs_put_dnode(&dn); |
bfad7c2d | 661 | unlock_out: |
2a340760 | 662 | if (create) { |
3104af35 | 663 | f2fs_unlock_op(sbi); |
3c082b7b | 664 | f2fs_balance_fs(sbi, allocated); |
2a340760 | 665 | } |
bfad7c2d | 666 | out: |
003a3e1d | 667 | trace_f2fs_map_blocks(inode, map, err); |
bfad7c2d | 668 | return err; |
eb47b800 JK |
669 | } |
670 | ||
003a3e1d | 671 | static int __get_data_block(struct inode *inode, sector_t iblock, |
e2b4e2bc | 672 | struct buffer_head *bh, int create, int flag) |
003a3e1d JK |
673 | { |
674 | struct f2fs_map_blocks map; | |
675 | int ret; | |
676 | ||
677 | map.m_lblk = iblock; | |
678 | map.m_len = bh->b_size >> inode->i_blkbits; | |
679 | ||
e2b4e2bc | 680 | ret = f2fs_map_blocks(inode, &map, create, flag); |
003a3e1d JK |
681 | if (!ret) { |
682 | map_bh(bh, inode->i_sb, map.m_pblk); | |
683 | bh->b_state = (bh->b_state & ~F2FS_MAP_FLAGS) | map.m_flags; | |
684 | bh->b_size = map.m_len << inode->i_blkbits; | |
685 | } | |
686 | return ret; | |
687 | } | |
688 | ||
ccfb3000 | 689 | static int get_data_block(struct inode *inode, sector_t iblock, |
e2b4e2bc CY |
690 | struct buffer_head *bh_result, int create, int flag) |
691 | { | |
692 | return __get_data_block(inode, iblock, bh_result, create, flag); | |
693 | } | |
694 | ||
695 | static int get_data_block_dio(struct inode *inode, sector_t iblock, | |
ccfb3000 JK |
696 | struct buffer_head *bh_result, int create) |
697 | { | |
e2b4e2bc CY |
698 | return __get_data_block(inode, iblock, bh_result, create, |
699 | F2FS_GET_BLOCK_DIO); | |
ccfb3000 JK |
700 | } |
701 | ||
e2b4e2bc | 702 | static int get_data_block_bmap(struct inode *inode, sector_t iblock, |
ccfb3000 JK |
703 | struct buffer_head *bh_result, int create) |
704 | { | |
179448bf | 705 | /* Block number less than F2FS MAX BLOCKS */ |
e0afc4d6 | 706 | if (unlikely(iblock >= F2FS_I_SB(inode)->max_file_blocks)) |
179448bf YH |
707 | return -EFBIG; |
708 | ||
e2b4e2bc CY |
709 | return __get_data_block(inode, iblock, bh_result, create, |
710 | F2FS_GET_BLOCK_BMAP); | |
ccfb3000 JK |
711 | } |
712 | ||
7f63eb77 JK |
713 | static inline sector_t logical_to_blk(struct inode *inode, loff_t offset) |
714 | { | |
715 | return (offset >> inode->i_blkbits); | |
716 | } | |
717 | ||
718 | static inline loff_t blk_to_logical(struct inode *inode, sector_t blk) | |
719 | { | |
720 | return (blk << inode->i_blkbits); | |
721 | } | |
722 | ||
9ab70134 JK |
723 | int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, |
724 | u64 start, u64 len) | |
725 | { | |
7f63eb77 JK |
726 | struct buffer_head map_bh; |
727 | sector_t start_blk, last_blk; | |
de1475cc | 728 | loff_t isize; |
7f63eb77 JK |
729 | u64 logical = 0, phys = 0, size = 0; |
730 | u32 flags = 0; | |
7f63eb77 JK |
731 | int ret = 0; |
732 | ||
733 | ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC); | |
734 | if (ret) | |
735 | return ret; | |
736 | ||
67f8cf3c JK |
737 | if (f2fs_has_inline_data(inode)) { |
738 | ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len); | |
739 | if (ret != -EAGAIN) | |
740 | return ret; | |
741 | } | |
742 | ||
5955102c | 743 | inode_lock(inode); |
de1475cc FL |
744 | |
745 | isize = i_size_read(inode); | |
9a950d52 FL |
746 | if (start >= isize) |
747 | goto out; | |
7f63eb77 | 748 | |
9a950d52 FL |
749 | if (start + len > isize) |
750 | len = isize - start; | |
7f63eb77 JK |
751 | |
752 | if (logical_to_blk(inode, len) == 0) | |
753 | len = blk_to_logical(inode, 1); | |
754 | ||
755 | start_blk = logical_to_blk(inode, start); | |
756 | last_blk = logical_to_blk(inode, start + len - 1); | |
9a950d52 | 757 | |
7f63eb77 JK |
758 | next: |
759 | memset(&map_bh, 0, sizeof(struct buffer_head)); | |
760 | map_bh.b_size = len; | |
761 | ||
e2b4e2bc CY |
762 | ret = get_data_block(inode, start_blk, &map_bh, 0, |
763 | F2FS_GET_BLOCK_FIEMAP); | |
7f63eb77 JK |
764 | if (ret) |
765 | goto out; | |
766 | ||
767 | /* HOLE */ | |
768 | if (!buffer_mapped(&map_bh)) { | |
9a950d52 FL |
769 | /* Go through holes util pass the EOF */ |
770 | if (blk_to_logical(inode, start_blk++) < isize) | |
771 | goto prep_next; | |
772 | /* Found a hole beyond isize means no more extents. | |
773 | * Note that the premise is that filesystems don't | |
774 | * punch holes beyond isize and keep size unchanged. | |
775 | */ | |
776 | flags |= FIEMAP_EXTENT_LAST; | |
777 | } | |
7f63eb77 | 778 | |
da5af127 CY |
779 | if (size) { |
780 | if (f2fs_encrypted_inode(inode)) | |
781 | flags |= FIEMAP_EXTENT_DATA_ENCRYPTED; | |
782 | ||
9a950d52 FL |
783 | ret = fiemap_fill_next_extent(fieinfo, logical, |
784 | phys, size, flags); | |
da5af127 | 785 | } |
7f63eb77 | 786 | |
9a950d52 FL |
787 | if (start_blk > last_blk || ret) |
788 | goto out; | |
7f63eb77 | 789 | |
9a950d52 FL |
790 | logical = blk_to_logical(inode, start_blk); |
791 | phys = blk_to_logical(inode, map_bh.b_blocknr); | |
792 | size = map_bh.b_size; | |
793 | flags = 0; | |
794 | if (buffer_unwritten(&map_bh)) | |
795 | flags = FIEMAP_EXTENT_UNWRITTEN; | |
7f63eb77 | 796 | |
9a950d52 | 797 | start_blk += logical_to_blk(inode, size); |
7f63eb77 | 798 | |
9a950d52 | 799 | prep_next: |
7f63eb77 JK |
800 | cond_resched(); |
801 | if (fatal_signal_pending(current)) | |
802 | ret = -EINTR; | |
803 | else | |
804 | goto next; | |
805 | out: | |
806 | if (ret == 1) | |
807 | ret = 0; | |
808 | ||
5955102c | 809 | inode_unlock(inode); |
7f63eb77 | 810 | return ret; |
9ab70134 JK |
811 | } |
812 | ||
f1e88660 JK |
813 | /* |
814 | * This function was originally taken from fs/mpage.c, and customized for f2fs. | |
815 | * Major change was from block_size == page_size in f2fs by default. | |
816 | */ | |
817 | static int f2fs_mpage_readpages(struct address_space *mapping, | |
818 | struct list_head *pages, struct page *page, | |
819 | unsigned nr_pages) | |
820 | { | |
821 | struct bio *bio = NULL; | |
822 | unsigned page_idx; | |
823 | sector_t last_block_in_bio = 0; | |
824 | struct inode *inode = mapping->host; | |
825 | const unsigned blkbits = inode->i_blkbits; | |
826 | const unsigned blocksize = 1 << blkbits; | |
827 | sector_t block_in_file; | |
828 | sector_t last_block; | |
829 | sector_t last_block_in_file; | |
830 | sector_t block_nr; | |
831 | struct block_device *bdev = inode->i_sb->s_bdev; | |
832 | struct f2fs_map_blocks map; | |
833 | ||
834 | map.m_pblk = 0; | |
835 | map.m_lblk = 0; | |
836 | map.m_len = 0; | |
837 | map.m_flags = 0; | |
838 | ||
839 | for (page_idx = 0; nr_pages; page_idx++, nr_pages--) { | |
840 | ||
841 | prefetchw(&page->flags); | |
842 | if (pages) { | |
843 | page = list_entry(pages->prev, struct page, lru); | |
844 | list_del(&page->lru); | |
845 | if (add_to_page_cache_lru(page, mapping, | |
846 | page->index, GFP_KERNEL)) | |
847 | goto next_page; | |
848 | } | |
849 | ||
850 | block_in_file = (sector_t)page->index; | |
851 | last_block = block_in_file + nr_pages; | |
852 | last_block_in_file = (i_size_read(inode) + blocksize - 1) >> | |
853 | blkbits; | |
854 | if (last_block > last_block_in_file) | |
855 | last_block = last_block_in_file; | |
856 | ||
857 | /* | |
858 | * Map blocks using the previous result first. | |
859 | */ | |
860 | if ((map.m_flags & F2FS_MAP_MAPPED) && | |
861 | block_in_file > map.m_lblk && | |
862 | block_in_file < (map.m_lblk + map.m_len)) | |
863 | goto got_it; | |
864 | ||
865 | /* | |
866 | * Then do more f2fs_map_blocks() calls until we are | |
867 | * done with this page. | |
868 | */ | |
869 | map.m_flags = 0; | |
870 | ||
871 | if (block_in_file < last_block) { | |
872 | map.m_lblk = block_in_file; | |
873 | map.m_len = last_block - block_in_file; | |
874 | ||
46c9e141 CY |
875 | if (f2fs_map_blocks(inode, &map, 0, |
876 | F2FS_GET_BLOCK_READ)) | |
f1e88660 JK |
877 | goto set_error_page; |
878 | } | |
879 | got_it: | |
880 | if ((map.m_flags & F2FS_MAP_MAPPED)) { | |
881 | block_nr = map.m_pblk + block_in_file - map.m_lblk; | |
882 | SetPageMappedToDisk(page); | |
883 | ||
884 | if (!PageUptodate(page) && !cleancache_get_page(page)) { | |
885 | SetPageUptodate(page); | |
886 | goto confused; | |
887 | } | |
888 | } else { | |
889 | zero_user_segment(page, 0, PAGE_CACHE_SIZE); | |
890 | SetPageUptodate(page); | |
891 | unlock_page(page); | |
892 | goto next_page; | |
893 | } | |
894 | ||
895 | /* | |
896 | * This page will go to BIO. Do we need to send this | |
897 | * BIO off first? | |
898 | */ | |
899 | if (bio && (last_block_in_bio != block_nr - 1)) { | |
900 | submit_and_realloc: | |
901 | submit_bio(READ, bio); | |
902 | bio = NULL; | |
903 | } | |
904 | if (bio == NULL) { | |
4375a336 JK |
905 | struct f2fs_crypto_ctx *ctx = NULL; |
906 | ||
907 | if (f2fs_encrypted_inode(inode) && | |
908 | S_ISREG(inode->i_mode)) { | |
4375a336 JK |
909 | |
910 | ctx = f2fs_get_crypto_ctx(inode); | |
911 | if (IS_ERR(ctx)) | |
912 | goto set_error_page; | |
913 | ||
914 | /* wait the page to be moved by cleaning */ | |
08b39fbd CY |
915 | f2fs_wait_on_encrypted_page_writeback( |
916 | F2FS_I_SB(inode), block_nr); | |
4375a336 JK |
917 | } |
918 | ||
f1e88660 | 919 | bio = bio_alloc(GFP_KERNEL, |
b54ffb73 | 920 | min_t(int, nr_pages, BIO_MAX_PAGES)); |
4375a336 JK |
921 | if (!bio) { |
922 | if (ctx) | |
923 | f2fs_release_crypto_ctx(ctx); | |
f1e88660 | 924 | goto set_error_page; |
4375a336 | 925 | } |
f1e88660 JK |
926 | bio->bi_bdev = bdev; |
927 | bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(block_nr); | |
12377024 | 928 | bio->bi_end_io = f2fs_read_end_io; |
4375a336 | 929 | bio->bi_private = ctx; |
f1e88660 JK |
930 | } |
931 | ||
932 | if (bio_add_page(bio, page, blocksize, 0) < blocksize) | |
933 | goto submit_and_realloc; | |
934 | ||
935 | last_block_in_bio = block_nr; | |
936 | goto next_page; | |
937 | set_error_page: | |
938 | SetPageError(page); | |
939 | zero_user_segment(page, 0, PAGE_CACHE_SIZE); | |
940 | unlock_page(page); | |
941 | goto next_page; | |
942 | confused: | |
943 | if (bio) { | |
944 | submit_bio(READ, bio); | |
945 | bio = NULL; | |
946 | } | |
947 | unlock_page(page); | |
948 | next_page: | |
949 | if (pages) | |
950 | page_cache_release(page); | |
951 | } | |
952 | BUG_ON(pages && !list_empty(pages)); | |
953 | if (bio) | |
954 | submit_bio(READ, bio); | |
955 | return 0; | |
956 | } | |
957 | ||
eb47b800 JK |
958 | static int f2fs_read_data_page(struct file *file, struct page *page) |
959 | { | |
9ffe0fb5 | 960 | struct inode *inode = page->mapping->host; |
b3d208f9 | 961 | int ret = -EAGAIN; |
9ffe0fb5 | 962 | |
c20e89cd CY |
963 | trace_f2fs_readpage(page, DATA); |
964 | ||
e1c42045 | 965 | /* If the file has inline data, try to read it directly */ |
9ffe0fb5 HL |
966 | if (f2fs_has_inline_data(inode)) |
967 | ret = f2fs_read_inline_data(inode, page); | |
b3d208f9 | 968 | if (ret == -EAGAIN) |
f1e88660 | 969 | ret = f2fs_mpage_readpages(page->mapping, NULL, page, 1); |
9ffe0fb5 | 970 | return ret; |
eb47b800 JK |
971 | } |
972 | ||
973 | static int f2fs_read_data_pages(struct file *file, | |
974 | struct address_space *mapping, | |
975 | struct list_head *pages, unsigned nr_pages) | |
976 | { | |
9ffe0fb5 | 977 | struct inode *inode = file->f_mapping->host; |
b8c29400 CY |
978 | struct page *page = list_entry(pages->prev, struct page, lru); |
979 | ||
980 | trace_f2fs_readpages(inode, page, nr_pages); | |
9ffe0fb5 HL |
981 | |
982 | /* If the file has inline data, skip readpages */ | |
983 | if (f2fs_has_inline_data(inode)) | |
984 | return 0; | |
985 | ||
f1e88660 | 986 | return f2fs_mpage_readpages(mapping, pages, NULL, nr_pages); |
eb47b800 JK |
987 | } |
988 | ||
05ca3632 | 989 | int do_write_data_page(struct f2fs_io_info *fio) |
eb47b800 | 990 | { |
05ca3632 | 991 | struct page *page = fio->page; |
eb47b800 | 992 | struct inode *inode = page->mapping->host; |
eb47b800 JK |
993 | struct dnode_of_data dn; |
994 | int err = 0; | |
995 | ||
996 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
266e97a8 | 997 | err = get_dnode_of_data(&dn, page->index, LOOKUP_NODE); |
eb47b800 JK |
998 | if (err) |
999 | return err; | |
1000 | ||
cf04e8eb | 1001 | fio->blk_addr = dn.data_blkaddr; |
eb47b800 JK |
1002 | |
1003 | /* This page is already truncated */ | |
2bca1e23 JK |
1004 | if (fio->blk_addr == NULL_ADDR) { |
1005 | ClearPageUptodate(page); | |
eb47b800 | 1006 | goto out_writepage; |
2bca1e23 | 1007 | } |
eb47b800 | 1008 | |
4375a336 | 1009 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) { |
08b39fbd CY |
1010 | |
1011 | /* wait for GCed encrypted page writeback */ | |
1012 | f2fs_wait_on_encrypted_page_writeback(F2FS_I_SB(inode), | |
1013 | fio->blk_addr); | |
1014 | ||
4375a336 JK |
1015 | fio->encrypted_page = f2fs_encrypt(inode, fio->page); |
1016 | if (IS_ERR(fio->encrypted_page)) { | |
1017 | err = PTR_ERR(fio->encrypted_page); | |
1018 | goto out_writepage; | |
1019 | } | |
1020 | } | |
1021 | ||
eb47b800 JK |
1022 | set_page_writeback(page); |
1023 | ||
1024 | /* | |
1025 | * If current allocation needs SSR, | |
1026 | * it had better in-place writes for updated data. | |
1027 | */ | |
cf04e8eb | 1028 | if (unlikely(fio->blk_addr != NEW_ADDR && |
b25958b6 | 1029 | !is_cold_data(page) && |
2da3e027 | 1030 | !IS_ATOMIC_WRITTEN_PAGE(page) && |
b25958b6 | 1031 | need_inplace_update(inode))) { |
05ca3632 | 1032 | rewrite_data_page(fio); |
fff04f90 | 1033 | set_inode_flag(F2FS_I(inode), FI_UPDATE_WRITE); |
8ce67cb0 | 1034 | trace_f2fs_do_write_data_page(page, IPU); |
eb47b800 | 1035 | } else { |
05ca3632 | 1036 | write_data_page(&dn, fio); |
216a620a | 1037 | set_data_blkaddr(&dn); |
7e4dde79 | 1038 | f2fs_update_extent_cache(&dn); |
8ce67cb0 | 1039 | trace_f2fs_do_write_data_page(page, OPU); |
fff04f90 | 1040 | set_inode_flag(F2FS_I(inode), FI_APPEND_WRITE); |
3c6c2beb JK |
1041 | if (page->index == 0) |
1042 | set_inode_flag(F2FS_I(inode), FI_FIRST_BLOCK_WRITTEN); | |
eb47b800 JK |
1043 | } |
1044 | out_writepage: | |
1045 | f2fs_put_dnode(&dn); | |
1046 | return err; | |
1047 | } | |
1048 | ||
1049 | static int f2fs_write_data_page(struct page *page, | |
1050 | struct writeback_control *wbc) | |
1051 | { | |
1052 | struct inode *inode = page->mapping->host; | |
4081363f | 1053 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
eb47b800 JK |
1054 | loff_t i_size = i_size_read(inode); |
1055 | const pgoff_t end_index = ((unsigned long long) i_size) | |
1056 | >> PAGE_CACHE_SHIFT; | |
9ffe0fb5 | 1057 | unsigned offset = 0; |
39936837 | 1058 | bool need_balance_fs = false; |
eb47b800 | 1059 | int err = 0; |
458e6197 | 1060 | struct f2fs_io_info fio = { |
05ca3632 | 1061 | .sbi = sbi, |
458e6197 | 1062 | .type = DATA, |
6c311ec6 | 1063 | .rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE, |
05ca3632 | 1064 | .page = page, |
4375a336 | 1065 | .encrypted_page = NULL, |
458e6197 | 1066 | }; |
eb47b800 | 1067 | |
ecda0de3 CY |
1068 | trace_f2fs_writepage(page, DATA); |
1069 | ||
eb47b800 | 1070 | if (page->index < end_index) |
39936837 | 1071 | goto write; |
eb47b800 JK |
1072 | |
1073 | /* | |
1074 | * If the offset is out-of-range of file size, | |
1075 | * this page does not have to be written to disk. | |
1076 | */ | |
1077 | offset = i_size & (PAGE_CACHE_SIZE - 1); | |
76f60268 | 1078 | if ((page->index >= end_index + 1) || !offset) |
39936837 | 1079 | goto out; |
eb47b800 JK |
1080 | |
1081 | zero_user_segment(page, offset, PAGE_CACHE_SIZE); | |
39936837 | 1082 | write: |
caf0047e | 1083 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
eb47b800 | 1084 | goto redirty_out; |
1e84371f JK |
1085 | if (f2fs_is_drop_cache(inode)) |
1086 | goto out; | |
1087 | if (f2fs_is_volatile_file(inode) && !wbc->for_reclaim && | |
1088 | available_free_memory(sbi, BASE_CHECK)) | |
1089 | goto redirty_out; | |
eb47b800 | 1090 | |
39936837 | 1091 | /* Dentry blocks are controlled by checkpoint */ |
eb47b800 | 1092 | if (S_ISDIR(inode->i_mode)) { |
cf779cab JK |
1093 | if (unlikely(f2fs_cp_error(sbi))) |
1094 | goto redirty_out; | |
05ca3632 | 1095 | err = do_write_data_page(&fio); |
8618b881 JK |
1096 | goto done; |
1097 | } | |
9ffe0fb5 | 1098 | |
cf779cab JK |
1099 | /* we should bypass data pages to proceed the kworkder jobs */ |
1100 | if (unlikely(f2fs_cp_error(sbi))) { | |
1101 | SetPageError(page); | |
a7ffdbe2 | 1102 | goto out; |
cf779cab JK |
1103 | } |
1104 | ||
8618b881 | 1105 | if (!wbc->for_reclaim) |
39936837 | 1106 | need_balance_fs = true; |
8618b881 | 1107 | else if (has_not_enough_free_secs(sbi, 0)) |
39936837 | 1108 | goto redirty_out; |
eb47b800 | 1109 | |
b3d208f9 | 1110 | err = -EAGAIN; |
8618b881 | 1111 | f2fs_lock_op(sbi); |
b3d208f9 JK |
1112 | if (f2fs_has_inline_data(inode)) |
1113 | err = f2fs_write_inline_data(inode, page); | |
1114 | if (err == -EAGAIN) | |
05ca3632 | 1115 | err = do_write_data_page(&fio); |
8618b881 JK |
1116 | f2fs_unlock_op(sbi); |
1117 | done: | |
1118 | if (err && err != -ENOENT) | |
1119 | goto redirty_out; | |
eb47b800 | 1120 | |
eb47b800 | 1121 | clear_cold_data(page); |
39936837 | 1122 | out: |
a7ffdbe2 | 1123 | inode_dec_dirty_pages(inode); |
2bca1e23 JK |
1124 | if (err) |
1125 | ClearPageUptodate(page); | |
eb47b800 | 1126 | unlock_page(page); |
2c4db1a6 | 1127 | f2fs_balance_fs(sbi, need_balance_fs); |
8d4ea29b | 1128 | if (wbc->for_reclaim || unlikely(f2fs_cp_error(sbi))) { |
2aea39ec | 1129 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
c227f912 | 1130 | remove_dirty_inode(inode); |
eb7e813c | 1131 | } |
eb47b800 JK |
1132 | return 0; |
1133 | ||
eb47b800 | 1134 | redirty_out: |
76f60268 | 1135 | redirty_page_for_writepage(wbc, page); |
8618b881 | 1136 | return AOP_WRITEPAGE_ACTIVATE; |
eb47b800 JK |
1137 | } |
1138 | ||
fa9150a8 NJ |
1139 | static int __f2fs_writepage(struct page *page, struct writeback_control *wbc, |
1140 | void *data) | |
1141 | { | |
1142 | struct address_space *mapping = data; | |
1143 | int ret = mapping->a_ops->writepage(page, wbc); | |
1144 | mapping_set_error(mapping, ret); | |
1145 | return ret; | |
1146 | } | |
1147 | ||
8f46dcae CY |
1148 | /* |
1149 | * This function was copied from write_cche_pages from mm/page-writeback.c. | |
1150 | * The major change is making write step of cold data page separately from | |
1151 | * warm/hot data page. | |
1152 | */ | |
1153 | static int f2fs_write_cache_pages(struct address_space *mapping, | |
1154 | struct writeback_control *wbc, writepage_t writepage, | |
1155 | void *data) | |
1156 | { | |
1157 | int ret = 0; | |
1158 | int done = 0; | |
1159 | struct pagevec pvec; | |
1160 | int nr_pages; | |
1161 | pgoff_t uninitialized_var(writeback_index); | |
1162 | pgoff_t index; | |
1163 | pgoff_t end; /* Inclusive */ | |
1164 | pgoff_t done_index; | |
1165 | int cycled; | |
1166 | int range_whole = 0; | |
1167 | int tag; | |
1168 | int step = 0; | |
1169 | ||
1170 | pagevec_init(&pvec, 0); | |
1171 | next: | |
1172 | if (wbc->range_cyclic) { | |
1173 | writeback_index = mapping->writeback_index; /* prev offset */ | |
1174 | index = writeback_index; | |
1175 | if (index == 0) | |
1176 | cycled = 1; | |
1177 | else | |
1178 | cycled = 0; | |
1179 | end = -1; | |
1180 | } else { | |
1181 | index = wbc->range_start >> PAGE_CACHE_SHIFT; | |
1182 | end = wbc->range_end >> PAGE_CACHE_SHIFT; | |
1183 | if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX) | |
1184 | range_whole = 1; | |
1185 | cycled = 1; /* ignore range_cyclic tests */ | |
1186 | } | |
1187 | if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages) | |
1188 | tag = PAGECACHE_TAG_TOWRITE; | |
1189 | else | |
1190 | tag = PAGECACHE_TAG_DIRTY; | |
1191 | retry: | |
1192 | if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages) | |
1193 | tag_pages_for_writeback(mapping, index, end); | |
1194 | done_index = index; | |
1195 | while (!done && (index <= end)) { | |
1196 | int i; | |
1197 | ||
1198 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag, | |
1199 | min(end - index, (pgoff_t)PAGEVEC_SIZE - 1) + 1); | |
1200 | if (nr_pages == 0) | |
1201 | break; | |
1202 | ||
1203 | for (i = 0; i < nr_pages; i++) { | |
1204 | struct page *page = pvec.pages[i]; | |
1205 | ||
1206 | if (page->index > end) { | |
1207 | done = 1; | |
1208 | break; | |
1209 | } | |
1210 | ||
1211 | done_index = page->index; | |
1212 | ||
1213 | lock_page(page); | |
1214 | ||
1215 | if (unlikely(page->mapping != mapping)) { | |
1216 | continue_unlock: | |
1217 | unlock_page(page); | |
1218 | continue; | |
1219 | } | |
1220 | ||
1221 | if (!PageDirty(page)) { | |
1222 | /* someone wrote it for us */ | |
1223 | goto continue_unlock; | |
1224 | } | |
1225 | ||
737f1899 | 1226 | if (step == is_cold_data(page)) |
8f46dcae CY |
1227 | goto continue_unlock; |
1228 | ||
1229 | if (PageWriteback(page)) { | |
1230 | if (wbc->sync_mode != WB_SYNC_NONE) | |
fec1d657 JK |
1231 | f2fs_wait_on_page_writeback(page, |
1232 | DATA, true); | |
8f46dcae CY |
1233 | else |
1234 | goto continue_unlock; | |
1235 | } | |
1236 | ||
1237 | BUG_ON(PageWriteback(page)); | |
1238 | if (!clear_page_dirty_for_io(page)) | |
1239 | goto continue_unlock; | |
1240 | ||
1241 | ret = (*writepage)(page, wbc, data); | |
1242 | if (unlikely(ret)) { | |
1243 | if (ret == AOP_WRITEPAGE_ACTIVATE) { | |
1244 | unlock_page(page); | |
1245 | ret = 0; | |
1246 | } else { | |
1247 | done_index = page->index + 1; | |
1248 | done = 1; | |
1249 | break; | |
1250 | } | |
1251 | } | |
1252 | ||
1253 | if (--wbc->nr_to_write <= 0 && | |
1254 | wbc->sync_mode == WB_SYNC_NONE) { | |
1255 | done = 1; | |
1256 | break; | |
1257 | } | |
1258 | } | |
1259 | pagevec_release(&pvec); | |
1260 | cond_resched(); | |
1261 | } | |
1262 | ||
1263 | if (step < 1) { | |
1264 | step++; | |
1265 | goto next; | |
1266 | } | |
1267 | ||
1268 | if (!cycled && !done) { | |
1269 | cycled = 1; | |
1270 | index = 0; | |
1271 | end = writeback_index - 1; | |
1272 | goto retry; | |
1273 | } | |
1274 | if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) | |
1275 | mapping->writeback_index = done_index; | |
1276 | ||
1277 | return ret; | |
1278 | } | |
1279 | ||
25ca923b | 1280 | static int f2fs_write_data_pages(struct address_space *mapping, |
eb47b800 JK |
1281 | struct writeback_control *wbc) |
1282 | { | |
1283 | struct inode *inode = mapping->host; | |
4081363f | 1284 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
5463e7c1 | 1285 | bool locked = false; |
eb47b800 | 1286 | int ret; |
50c8cdb3 | 1287 | long diff; |
eb47b800 | 1288 | |
e5748434 CY |
1289 | trace_f2fs_writepages(mapping->host, wbc, DATA); |
1290 | ||
cfb185a1 | 1291 | /* deal with chardevs and other special file */ |
1292 | if (!mapping->a_ops->writepage) | |
1293 | return 0; | |
1294 | ||
6a290544 CY |
1295 | /* skip writing if there is no dirty page in this inode */ |
1296 | if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE) | |
1297 | return 0; | |
1298 | ||
a1257023 JK |
1299 | if (S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_NONE && |
1300 | get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) && | |
1301 | available_free_memory(sbi, DIRTY_DENTS)) | |
1302 | goto skip_write; | |
1303 | ||
d323d005 CY |
1304 | /* skip writing during file defragment */ |
1305 | if (is_inode_flag_set(F2FS_I(inode), FI_DO_DEFRAG)) | |
1306 | goto skip_write; | |
1307 | ||
d5669f7b JK |
1308 | /* during POR, we don't need to trigger writepage at all. */ |
1309 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) | |
1310 | goto skip_write; | |
1311 | ||
50c8cdb3 | 1312 | diff = nr_pages_to_write(sbi, DATA, wbc); |
eb47b800 | 1313 | |
25c13551 | 1314 | if (!S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_ALL) { |
5463e7c1 JK |
1315 | mutex_lock(&sbi->writepages); |
1316 | locked = true; | |
1317 | } | |
8f46dcae | 1318 | ret = f2fs_write_cache_pages(mapping, wbc, __f2fs_writepage, mapping); |
bb96a8d5 | 1319 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
5463e7c1 JK |
1320 | if (locked) |
1321 | mutex_unlock(&sbi->writepages); | |
458e6197 | 1322 | |
c227f912 | 1323 | remove_dirty_inode(inode); |
eb47b800 | 1324 | |
50c8cdb3 | 1325 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff); |
eb47b800 | 1326 | return ret; |
d3baf95d JK |
1327 | |
1328 | skip_write: | |
a7ffdbe2 | 1329 | wbc->pages_skipped += get_dirty_pages(inode); |
d3baf95d | 1330 | return 0; |
eb47b800 JK |
1331 | } |
1332 | ||
3aab8f82 CY |
1333 | static void f2fs_write_failed(struct address_space *mapping, loff_t to) |
1334 | { | |
1335 | struct inode *inode = mapping->host; | |
819d9153 | 1336 | loff_t i_size = i_size_read(inode); |
3aab8f82 | 1337 | |
819d9153 JK |
1338 | if (to > i_size) { |
1339 | truncate_pagecache(inode, i_size); | |
1340 | truncate_blocks(inode, i_size, true); | |
3aab8f82 CY |
1341 | } |
1342 | } | |
1343 | ||
2aadac08 JK |
1344 | static int prepare_write_begin(struct f2fs_sb_info *sbi, |
1345 | struct page *page, loff_t pos, unsigned len, | |
1346 | block_t *blk_addr, bool *node_changed) | |
1347 | { | |
1348 | struct inode *inode = page->mapping->host; | |
1349 | pgoff_t index = page->index; | |
1350 | struct dnode_of_data dn; | |
1351 | struct page *ipage; | |
b4d07a3e JK |
1352 | bool locked = false; |
1353 | struct extent_info ei; | |
2aadac08 JK |
1354 | int err = 0; |
1355 | ||
b4d07a3e JK |
1356 | if (f2fs_has_inline_data(inode) || |
1357 | (pos & PAGE_CACHE_MASK) >= i_size_read(inode)) { | |
1358 | f2fs_lock_op(sbi); | |
1359 | locked = true; | |
1360 | } | |
1361 | restart: | |
2aadac08 JK |
1362 | /* check inline_data */ |
1363 | ipage = get_node_page(sbi, inode->i_ino); | |
1364 | if (IS_ERR(ipage)) { | |
1365 | err = PTR_ERR(ipage); | |
1366 | goto unlock_out; | |
1367 | } | |
1368 | ||
1369 | set_new_dnode(&dn, inode, ipage, ipage, 0); | |
1370 | ||
1371 | if (f2fs_has_inline_data(inode)) { | |
1372 | if (pos + len <= MAX_INLINE_DATA) { | |
1373 | read_inline_data(page, ipage); | |
1374 | set_inode_flag(F2FS_I(inode), FI_DATA_EXIST); | |
2049d4fc | 1375 | set_inline_node(ipage); |
2aadac08 JK |
1376 | } else { |
1377 | err = f2fs_convert_inline_page(&dn, page); | |
1378 | if (err) | |
b4d07a3e JK |
1379 | goto out; |
1380 | if (dn.data_blkaddr == NULL_ADDR) | |
1381 | err = f2fs_get_block(&dn, index); | |
1382 | } | |
1383 | } else if (locked) { | |
1384 | err = f2fs_get_block(&dn, index); | |
1385 | } else { | |
1386 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { | |
1387 | dn.data_blkaddr = ei.blk + index - ei.fofs; | |
1388 | } else { | |
b4d07a3e JK |
1389 | /* hole case */ |
1390 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); | |
b483fadf | 1391 | if (err || (!err && dn.data_blkaddr == NULL_ADDR)) { |
b4d07a3e JK |
1392 | f2fs_put_dnode(&dn); |
1393 | f2fs_lock_op(sbi); | |
1394 | locked = true; | |
1395 | goto restart; | |
1396 | } | |
2aadac08 JK |
1397 | } |
1398 | } | |
b4d07a3e | 1399 | |
2aadac08 JK |
1400 | /* convert_inline_page can make node_changed */ |
1401 | *blk_addr = dn.data_blkaddr; | |
1402 | *node_changed = dn.node_changed; | |
b4d07a3e | 1403 | out: |
2aadac08 JK |
1404 | f2fs_put_dnode(&dn); |
1405 | unlock_out: | |
b4d07a3e JK |
1406 | if (locked) |
1407 | f2fs_unlock_op(sbi); | |
2aadac08 JK |
1408 | return err; |
1409 | } | |
1410 | ||
eb47b800 JK |
1411 | static int f2fs_write_begin(struct file *file, struct address_space *mapping, |
1412 | loff_t pos, unsigned len, unsigned flags, | |
1413 | struct page **pagep, void **fsdata) | |
1414 | { | |
1415 | struct inode *inode = mapping->host; | |
4081363f | 1416 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
86531d6b | 1417 | struct page *page = NULL; |
eb47b800 | 1418 | pgoff_t index = ((unsigned long long) pos) >> PAGE_CACHE_SHIFT; |
2aadac08 JK |
1419 | bool need_balance = false; |
1420 | block_t blkaddr = NULL_ADDR; | |
eb47b800 JK |
1421 | int err = 0; |
1422 | ||
62aed044 CY |
1423 | trace_f2fs_write_begin(inode, pos, len, flags); |
1424 | ||
5f727395 JK |
1425 | /* |
1426 | * We should check this at this moment to avoid deadlock on inode page | |
1427 | * and #0 page. The locking rule for inline_data conversion should be: | |
1428 | * lock_page(page #0) -> lock_page(inode_page) | |
1429 | */ | |
1430 | if (index != 0) { | |
1431 | err = f2fs_convert_inline_inode(inode); | |
1432 | if (err) | |
1433 | goto fail; | |
1434 | } | |
afcb7ca0 | 1435 | repeat: |
eb47b800 | 1436 | page = grab_cache_page_write_begin(mapping, index, flags); |
3aab8f82 CY |
1437 | if (!page) { |
1438 | err = -ENOMEM; | |
1439 | goto fail; | |
1440 | } | |
d5f66990 | 1441 | |
eb47b800 JK |
1442 | *pagep = page; |
1443 | ||
2aadac08 JK |
1444 | err = prepare_write_begin(sbi, page, pos, len, |
1445 | &blkaddr, &need_balance); | |
9ba69cf9 | 1446 | if (err) |
2aadac08 | 1447 | goto fail; |
9ba69cf9 | 1448 | |
2aadac08 | 1449 | if (need_balance && has_not_enough_free_secs(sbi, 0)) { |
2a340760 | 1450 | unlock_page(page); |
2c4db1a6 | 1451 | f2fs_balance_fs(sbi, true); |
2a340760 JK |
1452 | lock_page(page); |
1453 | if (page->mapping != mapping) { | |
1454 | /* The page got truncated from under us */ | |
1455 | f2fs_put_page(page, 1); | |
1456 | goto repeat; | |
1457 | } | |
1458 | } | |
1459 | ||
fec1d657 | 1460 | f2fs_wait_on_page_writeback(page, DATA, false); |
b3d208f9 | 1461 | |
08b39fbd CY |
1462 | /* wait for GCed encrypted page writeback */ |
1463 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) | |
2aadac08 | 1464 | f2fs_wait_on_encrypted_page_writeback(sbi, blkaddr); |
08b39fbd | 1465 | |
90d4388a CY |
1466 | if (len == PAGE_CACHE_SIZE) |
1467 | goto out_update; | |
1468 | if (PageUptodate(page)) | |
1469 | goto out_clear; | |
eb47b800 JK |
1470 | |
1471 | if ((pos & PAGE_CACHE_MASK) >= i_size_read(inode)) { | |
1472 | unsigned start = pos & (PAGE_CACHE_SIZE - 1); | |
1473 | unsigned end = start + len; | |
1474 | ||
1475 | /* Reading beyond i_size is simple: memset to zero */ | |
1476 | zero_user_segments(page, 0, start, end, PAGE_CACHE_SIZE); | |
90d4388a | 1477 | goto out_update; |
eb47b800 JK |
1478 | } |
1479 | ||
2aadac08 | 1480 | if (blkaddr == NEW_ADDR) { |
eb47b800 JK |
1481 | zero_user_segment(page, 0, PAGE_CACHE_SIZE); |
1482 | } else { | |
cf04e8eb | 1483 | struct f2fs_io_info fio = { |
05ca3632 | 1484 | .sbi = sbi, |
cf04e8eb JK |
1485 | .type = DATA, |
1486 | .rw = READ_SYNC, | |
2aadac08 | 1487 | .blk_addr = blkaddr, |
05ca3632 | 1488 | .page = page, |
4375a336 | 1489 | .encrypted_page = NULL, |
cf04e8eb | 1490 | }; |
05ca3632 | 1491 | err = f2fs_submit_page_bio(&fio); |
9234f319 JK |
1492 | if (err) |
1493 | goto fail; | |
d54c795b | 1494 | |
393ff91f | 1495 | lock_page(page); |
6bacf52f | 1496 | if (unlikely(!PageUptodate(page))) { |
3aab8f82 CY |
1497 | err = -EIO; |
1498 | goto fail; | |
eb47b800 | 1499 | } |
6bacf52f | 1500 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
1501 | f2fs_put_page(page, 1); |
1502 | goto repeat; | |
eb47b800 | 1503 | } |
4375a336 JK |
1504 | |
1505 | /* avoid symlink page */ | |
1506 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) { | |
1507 | err = f2fs_decrypt_one(inode, page); | |
86531d6b | 1508 | if (err) |
4375a336 | 1509 | goto fail; |
4375a336 | 1510 | } |
eb47b800 | 1511 | } |
90d4388a | 1512 | out_update: |
eb47b800 | 1513 | SetPageUptodate(page); |
90d4388a | 1514 | out_clear: |
eb47b800 JK |
1515 | clear_cold_data(page); |
1516 | return 0; | |
9ba69cf9 | 1517 | |
3aab8f82 | 1518 | fail: |
86531d6b | 1519 | f2fs_put_page(page, 1); |
3aab8f82 CY |
1520 | f2fs_write_failed(mapping, pos + len); |
1521 | return err; | |
eb47b800 JK |
1522 | } |
1523 | ||
a1dd3c13 JK |
1524 | static int f2fs_write_end(struct file *file, |
1525 | struct address_space *mapping, | |
1526 | loff_t pos, unsigned len, unsigned copied, | |
1527 | struct page *page, void *fsdata) | |
1528 | { | |
1529 | struct inode *inode = page->mapping->host; | |
1530 | ||
dfb2bf38 CY |
1531 | trace_f2fs_write_end(inode, pos, len, copied); |
1532 | ||
34ba94ba | 1533 | set_page_dirty(page); |
a1dd3c13 JK |
1534 | |
1535 | if (pos + copied > i_size_read(inode)) { | |
1536 | i_size_write(inode, pos + copied); | |
1537 | mark_inode_dirty(inode); | |
a1dd3c13 JK |
1538 | } |
1539 | ||
75c3c8bc | 1540 | f2fs_put_page(page, 1); |
d0239e1b | 1541 | f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); |
a1dd3c13 JK |
1542 | return copied; |
1543 | } | |
1544 | ||
6f673763 OS |
1545 | static int check_direct_IO(struct inode *inode, struct iov_iter *iter, |
1546 | loff_t offset) | |
944fcfc1 JK |
1547 | { |
1548 | unsigned blocksize_mask = inode->i_sb->s_blocksize - 1; | |
944fcfc1 | 1549 | |
944fcfc1 JK |
1550 | if (offset & blocksize_mask) |
1551 | return -EINVAL; | |
1552 | ||
5b46f25d AV |
1553 | if (iov_iter_alignment(iter) & blocksize_mask) |
1554 | return -EINVAL; | |
1555 | ||
944fcfc1 JK |
1556 | return 0; |
1557 | } | |
1558 | ||
22c6186e OS |
1559 | static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter, |
1560 | loff_t offset) | |
eb47b800 JK |
1561 | { |
1562 | struct file *file = iocb->ki_filp; | |
3aab8f82 CY |
1563 | struct address_space *mapping = file->f_mapping; |
1564 | struct inode *inode = mapping->host; | |
1565 | size_t count = iov_iter_count(iter); | |
1566 | int err; | |
944fcfc1 | 1567 | |
b3d208f9 | 1568 | /* we don't need to use inline_data strictly */ |
b9d777b8 JK |
1569 | err = f2fs_convert_inline_inode(inode); |
1570 | if (err) | |
1571 | return err; | |
9ffe0fb5 | 1572 | |
fcc85a4d JK |
1573 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) |
1574 | return 0; | |
1575 | ||
c15e8599 CY |
1576 | err = check_direct_IO(inode, iter, offset); |
1577 | if (err) | |
1578 | return err; | |
944fcfc1 | 1579 | |
6f673763 | 1580 | trace_f2fs_direct_IO_enter(inode, offset, count, iov_iter_rw(iter)); |
70407fad | 1581 | |
f9811703 | 1582 | if (iov_iter_rw(iter) == WRITE) { |
6d5a1495 CY |
1583 | err = __allocate_data_blocks(inode, offset, count); |
1584 | if (err) | |
f9811703 | 1585 | goto out; |
f9811703 | 1586 | } |
59b802e5 | 1587 | |
e2b4e2bc | 1588 | err = blockdev_direct_IO(iocb, inode, iter, offset, get_data_block_dio); |
f9811703 | 1589 | out: |
6f673763 | 1590 | if (err < 0 && iov_iter_rw(iter) == WRITE) |
3aab8f82 | 1591 | f2fs_write_failed(mapping, offset + count); |
70407fad | 1592 | |
6f673763 | 1593 | trace_f2fs_direct_IO_exit(inode, offset, count, iov_iter_rw(iter), err); |
70407fad | 1594 | |
3aab8f82 | 1595 | return err; |
eb47b800 JK |
1596 | } |
1597 | ||
487261f3 CY |
1598 | void f2fs_invalidate_page(struct page *page, unsigned int offset, |
1599 | unsigned int length) | |
eb47b800 JK |
1600 | { |
1601 | struct inode *inode = page->mapping->host; | |
487261f3 | 1602 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
a7ffdbe2 | 1603 | |
487261f3 CY |
1604 | if (inode->i_ino >= F2FS_ROOT_INO(sbi) && |
1605 | (offset % PAGE_CACHE_SIZE || length != PAGE_CACHE_SIZE)) | |
a7ffdbe2 JK |
1606 | return; |
1607 | ||
487261f3 CY |
1608 | if (PageDirty(page)) { |
1609 | if (inode->i_ino == F2FS_META_INO(sbi)) | |
1610 | dec_page_count(sbi, F2FS_DIRTY_META); | |
1611 | else if (inode->i_ino == F2FS_NODE_INO(sbi)) | |
1612 | dec_page_count(sbi, F2FS_DIRTY_NODES); | |
1613 | else | |
1614 | inode_dec_dirty_pages(inode); | |
1615 | } | |
decd36b6 CY |
1616 | |
1617 | /* This is atomic written page, keep Private */ | |
1618 | if (IS_ATOMIC_WRITTEN_PAGE(page)) | |
1619 | return; | |
1620 | ||
eb47b800 JK |
1621 | ClearPagePrivate(page); |
1622 | } | |
1623 | ||
487261f3 | 1624 | int f2fs_release_page(struct page *page, gfp_t wait) |
eb47b800 | 1625 | { |
f68daeeb JK |
1626 | /* If this is dirty page, keep PagePrivate */ |
1627 | if (PageDirty(page)) | |
1628 | return 0; | |
1629 | ||
decd36b6 CY |
1630 | /* This is atomic written page, keep Private */ |
1631 | if (IS_ATOMIC_WRITTEN_PAGE(page)) | |
1632 | return 0; | |
1633 | ||
eb47b800 | 1634 | ClearPagePrivate(page); |
c3850aa1 | 1635 | return 1; |
eb47b800 JK |
1636 | } |
1637 | ||
1638 | static int f2fs_set_data_page_dirty(struct page *page) | |
1639 | { | |
1640 | struct address_space *mapping = page->mapping; | |
1641 | struct inode *inode = mapping->host; | |
1642 | ||
26c6b887 JK |
1643 | trace_f2fs_set_page_dirty(page, DATA); |
1644 | ||
eb47b800 | 1645 | SetPageUptodate(page); |
34ba94ba | 1646 | |
1e84371f | 1647 | if (f2fs_is_atomic_file(inode)) { |
decd36b6 CY |
1648 | if (!IS_ATOMIC_WRITTEN_PAGE(page)) { |
1649 | register_inmem_page(inode, page); | |
1650 | return 1; | |
1651 | } | |
1652 | /* | |
1653 | * Previously, this page has been registered, we just | |
1654 | * return here. | |
1655 | */ | |
1656 | return 0; | |
34ba94ba JK |
1657 | } |
1658 | ||
eb47b800 JK |
1659 | if (!PageDirty(page)) { |
1660 | __set_page_dirty_nobuffers(page); | |
a7ffdbe2 | 1661 | update_dirty_page(inode, page); |
eb47b800 JK |
1662 | return 1; |
1663 | } | |
1664 | return 0; | |
1665 | } | |
1666 | ||
c01e54b7 JK |
1667 | static sector_t f2fs_bmap(struct address_space *mapping, sector_t block) |
1668 | { | |
454ae7e5 CY |
1669 | struct inode *inode = mapping->host; |
1670 | ||
1d373a0e JK |
1671 | if (f2fs_has_inline_data(inode)) |
1672 | return 0; | |
1673 | ||
1674 | /* make sure allocating whole blocks */ | |
1675 | if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) | |
1676 | filemap_write_and_wait(mapping); | |
1677 | ||
e2b4e2bc | 1678 | return generic_block_bmap(mapping, block, get_data_block_bmap); |
429511cd CY |
1679 | } |
1680 | ||
eb47b800 JK |
1681 | const struct address_space_operations f2fs_dblock_aops = { |
1682 | .readpage = f2fs_read_data_page, | |
1683 | .readpages = f2fs_read_data_pages, | |
1684 | .writepage = f2fs_write_data_page, | |
1685 | .writepages = f2fs_write_data_pages, | |
1686 | .write_begin = f2fs_write_begin, | |
a1dd3c13 | 1687 | .write_end = f2fs_write_end, |
eb47b800 | 1688 | .set_page_dirty = f2fs_set_data_page_dirty, |
487261f3 CY |
1689 | .invalidatepage = f2fs_invalidate_page, |
1690 | .releasepage = f2fs_release_page, | |
eb47b800 | 1691 | .direct_IO = f2fs_direct_IO, |
c01e54b7 | 1692 | .bmap = f2fs_bmap, |
eb47b800 | 1693 | }; |