]>
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) { |
0b81d077 | 37 | fscrypt_release_ctx(bio->bi_private); |
4375a336 | 38 | } else { |
0b81d077 | 39 | fscrypt_decrypt_bio_pages(bio->bi_private, bio); |
4375a336 JK |
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 | ||
0b81d077 | 67 | fscrypt_pullback_bio_page(&page, true); |
4375a336 | 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 | ||
0c3a5797 CY |
117 | static bool __has_merged_page(struct f2fs_bio_info *io, struct inode *inode, |
118 | struct page *page, nid_t ino) | |
0fd785eb | 119 | { |
0fd785eb CY |
120 | struct bio_vec *bvec; |
121 | struct page *target; | |
122 | int i; | |
123 | ||
0c3a5797 | 124 | if (!io->bio) |
0fd785eb | 125 | return false; |
0c3a5797 CY |
126 | |
127 | if (!inode && !page && !ino) | |
128 | return true; | |
0fd785eb CY |
129 | |
130 | bio_for_each_segment_all(bvec, io->bio, i) { | |
131 | ||
0b81d077 | 132 | if (bvec->bv_page->mapping) |
0fd785eb | 133 | target = bvec->bv_page; |
0b81d077 JK |
134 | else |
135 | target = fscrypt_control_page(bvec->bv_page); | |
0fd785eb | 136 | |
0c3a5797 CY |
137 | if (inode && inode == target->mapping->host) |
138 | return true; | |
139 | if (page && page == target) | |
140 | return true; | |
141 | if (ino && ino == ino_of_node(target)) | |
0fd785eb | 142 | return true; |
0fd785eb CY |
143 | } |
144 | ||
0fd785eb CY |
145 | return false; |
146 | } | |
147 | ||
0c3a5797 CY |
148 | static bool has_merged_page(struct f2fs_sb_info *sbi, struct inode *inode, |
149 | struct page *page, nid_t ino, | |
150 | enum page_type type) | |
151 | { | |
152 | enum page_type btype = PAGE_TYPE_OF_BIO(type); | |
153 | struct f2fs_bio_info *io = &sbi->write_io[btype]; | |
154 | bool ret; | |
155 | ||
156 | down_read(&io->io_rwsem); | |
157 | ret = __has_merged_page(io, inode, page, ino); | |
158 | up_read(&io->io_rwsem); | |
159 | return ret; | |
160 | } | |
161 | ||
162 | static void __f2fs_submit_merged_bio(struct f2fs_sb_info *sbi, | |
163 | struct inode *inode, struct page *page, | |
164 | nid_t ino, enum page_type type, int rw) | |
93dfe2ac JK |
165 | { |
166 | enum page_type btype = PAGE_TYPE_OF_BIO(type); | |
167 | struct f2fs_bio_info *io; | |
168 | ||
169 | io = is_read_io(rw) ? &sbi->read_io : &sbi->write_io[btype]; | |
170 | ||
df0f8dc0 | 171 | down_write(&io->io_rwsem); |
458e6197 | 172 | |
0c3a5797 CY |
173 | if (!__has_merged_page(io, inode, page, ino)) |
174 | goto out; | |
175 | ||
458e6197 JK |
176 | /* change META to META_FLUSH in the checkpoint procedure */ |
177 | if (type >= META_FLUSH) { | |
178 | io->fio.type = META_FLUSH; | |
0f7b2abd JK |
179 | if (test_opt(sbi, NOBARRIER)) |
180 | io->fio.rw = WRITE_FLUSH | REQ_META | REQ_PRIO; | |
181 | else | |
182 | io->fio.rw = WRITE_FLUSH_FUA | REQ_META | REQ_PRIO; | |
458e6197 JK |
183 | } |
184 | __submit_merged_bio(io); | |
0c3a5797 | 185 | out: |
df0f8dc0 | 186 | up_write(&io->io_rwsem); |
93dfe2ac JK |
187 | } |
188 | ||
0c3a5797 CY |
189 | void f2fs_submit_merged_bio(struct f2fs_sb_info *sbi, enum page_type type, |
190 | int rw) | |
191 | { | |
192 | __f2fs_submit_merged_bio(sbi, NULL, NULL, 0, type, rw); | |
193 | } | |
194 | ||
195 | void f2fs_submit_merged_bio_cond(struct f2fs_sb_info *sbi, | |
196 | struct inode *inode, struct page *page, | |
197 | nid_t ino, enum page_type type, int rw) | |
198 | { | |
199 | if (has_merged_page(sbi, inode, page, ino, type)) | |
200 | __f2fs_submit_merged_bio(sbi, inode, page, ino, type, rw); | |
201 | } | |
202 | ||
406657dd CY |
203 | void f2fs_flush_merged_bios(struct f2fs_sb_info *sbi) |
204 | { | |
205 | f2fs_submit_merged_bio(sbi, DATA, WRITE); | |
206 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
207 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
208 | } | |
209 | ||
93dfe2ac JK |
210 | /* |
211 | * Fill the locked page with data located in the block address. | |
212 | * Return unlocked page. | |
213 | */ | |
05ca3632 | 214 | int f2fs_submit_page_bio(struct f2fs_io_info *fio) |
93dfe2ac | 215 | { |
93dfe2ac | 216 | struct bio *bio; |
0b81d077 JK |
217 | struct page *page = fio->encrypted_page ? |
218 | fio->encrypted_page : fio->page; | |
93dfe2ac | 219 | |
2ace38e0 | 220 | trace_f2fs_submit_page_bio(page, fio); |
05ca3632 | 221 | f2fs_trace_ios(fio, 0); |
93dfe2ac JK |
222 | |
223 | /* Allocate a new bio */ | |
7a9d7548 | 224 | bio = __bio_alloc(fio->sbi, fio->new_blkaddr, 1, is_read_io(fio->rw)); |
93dfe2ac | 225 | |
09cbfeaf | 226 | if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) { |
93dfe2ac | 227 | bio_put(bio); |
93dfe2ac JK |
228 | return -EFAULT; |
229 | } | |
230 | ||
cf04e8eb | 231 | submit_bio(fio->rw, bio); |
93dfe2ac JK |
232 | return 0; |
233 | } | |
234 | ||
05ca3632 | 235 | void f2fs_submit_page_mbio(struct f2fs_io_info *fio) |
93dfe2ac | 236 | { |
05ca3632 | 237 | struct f2fs_sb_info *sbi = fio->sbi; |
458e6197 | 238 | enum page_type btype = PAGE_TYPE_OF_BIO(fio->type); |
93dfe2ac | 239 | struct f2fs_bio_info *io; |
940a6d34 | 240 | bool is_read = is_read_io(fio->rw); |
4375a336 | 241 | struct page *bio_page; |
93dfe2ac | 242 | |
940a6d34 | 243 | io = is_read ? &sbi->read_io : &sbi->write_io[btype]; |
93dfe2ac | 244 | |
7a9d7548 CY |
245 | if (fio->old_blkaddr != NEW_ADDR) |
246 | verify_block_addr(sbi, fio->old_blkaddr); | |
247 | verify_block_addr(sbi, fio->new_blkaddr); | |
93dfe2ac | 248 | |
df0f8dc0 | 249 | down_write(&io->io_rwsem); |
93dfe2ac | 250 | |
940a6d34 | 251 | if (!is_read) |
93dfe2ac JK |
252 | inc_page_count(sbi, F2FS_WRITEBACK); |
253 | ||
7a9d7548 | 254 | if (io->bio && (io->last_block_in_bio != fio->new_blkaddr - 1 || |
458e6197 JK |
255 | io->fio.rw != fio->rw)) |
256 | __submit_merged_bio(io); | |
93dfe2ac JK |
257 | alloc_new: |
258 | if (io->bio == NULL) { | |
90a893c7 | 259 | int bio_blocks = MAX_BIO_BLOCKS(sbi); |
940a6d34 | 260 | |
7a9d7548 CY |
261 | io->bio = __bio_alloc(sbi, fio->new_blkaddr, |
262 | bio_blocks, is_read); | |
458e6197 | 263 | io->fio = *fio; |
93dfe2ac JK |
264 | } |
265 | ||
4375a336 JK |
266 | bio_page = fio->encrypted_page ? fio->encrypted_page : fio->page; |
267 | ||
09cbfeaf KS |
268 | if (bio_add_page(io->bio, bio_page, PAGE_SIZE, 0) < |
269 | PAGE_SIZE) { | |
458e6197 | 270 | __submit_merged_bio(io); |
93dfe2ac JK |
271 | goto alloc_new; |
272 | } | |
273 | ||
7a9d7548 | 274 | io->last_block_in_bio = fio->new_blkaddr; |
05ca3632 | 275 | f2fs_trace_ios(fio, 0); |
93dfe2ac | 276 | |
df0f8dc0 | 277 | up_write(&io->io_rwsem); |
05ca3632 | 278 | trace_f2fs_submit_page_mbio(fio->page, fio); |
93dfe2ac JK |
279 | } |
280 | ||
0a8165d7 | 281 | /* |
eb47b800 JK |
282 | * Lock ordering for the change of data block address: |
283 | * ->data_page | |
284 | * ->node_page | |
285 | * update block addresses in the node page | |
286 | */ | |
216a620a | 287 | void set_data_blkaddr(struct dnode_of_data *dn) |
eb47b800 JK |
288 | { |
289 | struct f2fs_node *rn; | |
290 | __le32 *addr_array; | |
291 | struct page *node_page = dn->node_page; | |
292 | unsigned int ofs_in_node = dn->ofs_in_node; | |
293 | ||
fec1d657 | 294 | f2fs_wait_on_page_writeback(node_page, NODE, true); |
eb47b800 | 295 | |
45590710 | 296 | rn = F2FS_NODE(node_page); |
eb47b800 JK |
297 | |
298 | /* Get physical address of data block */ | |
299 | addr_array = blkaddr_in_node(rn); | |
e1509cf2 | 300 | addr_array[ofs_in_node] = cpu_to_le32(dn->data_blkaddr); |
12719ae1 JK |
301 | if (set_page_dirty(node_page)) |
302 | dn->node_changed = true; | |
eb47b800 JK |
303 | } |
304 | ||
f28b3434 CY |
305 | void f2fs_update_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr) |
306 | { | |
307 | dn->data_blkaddr = blkaddr; | |
308 | set_data_blkaddr(dn); | |
309 | f2fs_update_extent_cache(dn); | |
310 | } | |
311 | ||
eb47b800 JK |
312 | int reserve_new_block(struct dnode_of_data *dn) |
313 | { | |
4081363f | 314 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
eb47b800 | 315 | |
6bacf52f | 316 | if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))) |
eb47b800 | 317 | return -EPERM; |
cfb271d4 | 318 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, 1))) |
eb47b800 JK |
319 | return -ENOSPC; |
320 | ||
c01e2853 NJ |
321 | trace_f2fs_reserve_new_block(dn->inode, dn->nid, dn->ofs_in_node); |
322 | ||
eb47b800 | 323 | dn->data_blkaddr = NEW_ADDR; |
216a620a | 324 | set_data_blkaddr(dn); |
a18ff063 | 325 | mark_inode_dirty(dn->inode); |
eb47b800 JK |
326 | sync_inode_page(dn); |
327 | return 0; | |
328 | } | |
329 | ||
b600965c HL |
330 | int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index) |
331 | { | |
332 | bool need_put = dn->inode_page ? false : true; | |
333 | int err; | |
334 | ||
335 | err = get_dnode_of_data(dn, index, ALLOC_NODE); | |
336 | if (err) | |
337 | return err; | |
a8865372 | 338 | |
b600965c HL |
339 | if (dn->data_blkaddr == NULL_ADDR) |
340 | err = reserve_new_block(dn); | |
a8865372 | 341 | if (err || need_put) |
b600965c HL |
342 | f2fs_put_dnode(dn); |
343 | return err; | |
344 | } | |
345 | ||
759af1c9 | 346 | int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index) |
eb47b800 | 347 | { |
028a41e8 | 348 | struct extent_info ei; |
759af1c9 | 349 | struct inode *inode = dn->inode; |
028a41e8 | 350 | |
759af1c9 FL |
351 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { |
352 | dn->data_blkaddr = ei.blk + index - ei.fofs; | |
353 | return 0; | |
429511cd | 354 | } |
028a41e8 | 355 | |
759af1c9 | 356 | return f2fs_reserve_block(dn, index); |
eb47b800 JK |
357 | } |
358 | ||
a56c7c6f JK |
359 | struct page *get_read_data_page(struct inode *inode, pgoff_t index, |
360 | int rw, bool for_write) | |
eb47b800 | 361 | { |
eb47b800 JK |
362 | struct address_space *mapping = inode->i_mapping; |
363 | struct dnode_of_data dn; | |
364 | struct page *page; | |
cb3bc9ee | 365 | struct extent_info ei; |
eb47b800 | 366 | int err; |
cf04e8eb | 367 | struct f2fs_io_info fio = { |
05ca3632 | 368 | .sbi = F2FS_I_SB(inode), |
cf04e8eb | 369 | .type = DATA, |
43f3eae1 | 370 | .rw = rw, |
4375a336 | 371 | .encrypted_page = NULL, |
cf04e8eb | 372 | }; |
eb47b800 | 373 | |
4375a336 JK |
374 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) |
375 | return read_mapping_page(mapping, index, NULL); | |
376 | ||
a56c7c6f | 377 | page = f2fs_grab_cache_page(mapping, index, for_write); |
650495de JK |
378 | if (!page) |
379 | return ERR_PTR(-ENOMEM); | |
380 | ||
cb3bc9ee CY |
381 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { |
382 | dn.data_blkaddr = ei.blk + index - ei.fofs; | |
383 | goto got_it; | |
384 | } | |
385 | ||
eb47b800 | 386 | set_new_dnode(&dn, inode, NULL, NULL, 0); |
266e97a8 | 387 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); |
86531d6b JK |
388 | if (err) |
389 | goto put_err; | |
eb47b800 JK |
390 | f2fs_put_dnode(&dn); |
391 | ||
6bacf52f | 392 | if (unlikely(dn.data_blkaddr == NULL_ADDR)) { |
86531d6b JK |
393 | err = -ENOENT; |
394 | goto put_err; | |
650495de | 395 | } |
cb3bc9ee | 396 | got_it: |
43f3eae1 JK |
397 | if (PageUptodate(page)) { |
398 | unlock_page(page); | |
eb47b800 | 399 | return page; |
43f3eae1 | 400 | } |
eb47b800 | 401 | |
d59ff4df JK |
402 | /* |
403 | * A new dentry page is allocated but not able to be written, since its | |
404 | * new inode page couldn't be allocated due to -ENOSPC. | |
405 | * In such the case, its blkaddr can be remained as NEW_ADDR. | |
406 | * see, f2fs_add_link -> get_new_data_page -> init_inode_metadata. | |
407 | */ | |
408 | if (dn.data_blkaddr == NEW_ADDR) { | |
09cbfeaf | 409 | zero_user_segment(page, 0, PAGE_SIZE); |
d59ff4df | 410 | SetPageUptodate(page); |
43f3eae1 | 411 | unlock_page(page); |
d59ff4df JK |
412 | return page; |
413 | } | |
eb47b800 | 414 | |
7a9d7548 | 415 | fio.new_blkaddr = fio.old_blkaddr = dn.data_blkaddr; |
05ca3632 JK |
416 | fio.page = page; |
417 | err = f2fs_submit_page_bio(&fio); | |
393ff91f | 418 | if (err) |
86531d6b | 419 | goto put_err; |
43f3eae1 | 420 | return page; |
86531d6b JK |
421 | |
422 | put_err: | |
423 | f2fs_put_page(page, 1); | |
424 | return ERR_PTR(err); | |
43f3eae1 JK |
425 | } |
426 | ||
427 | struct page *find_data_page(struct inode *inode, pgoff_t index) | |
428 | { | |
429 | struct address_space *mapping = inode->i_mapping; | |
430 | struct page *page; | |
431 | ||
432 | page = find_get_page(mapping, index); | |
433 | if (page && PageUptodate(page)) | |
434 | return page; | |
435 | f2fs_put_page(page, 0); | |
436 | ||
a56c7c6f | 437 | page = get_read_data_page(inode, index, READ_SYNC, false); |
43f3eae1 JK |
438 | if (IS_ERR(page)) |
439 | return page; | |
440 | ||
441 | if (PageUptodate(page)) | |
442 | return page; | |
443 | ||
444 | wait_on_page_locked(page); | |
445 | if (unlikely(!PageUptodate(page))) { | |
446 | f2fs_put_page(page, 0); | |
447 | return ERR_PTR(-EIO); | |
448 | } | |
449 | return page; | |
450 | } | |
451 | ||
452 | /* | |
453 | * If it tries to access a hole, return an error. | |
454 | * Because, the callers, functions in dir.c and GC, should be able to know | |
455 | * whether this page exists or not. | |
456 | */ | |
a56c7c6f JK |
457 | struct page *get_lock_data_page(struct inode *inode, pgoff_t index, |
458 | bool for_write) | |
43f3eae1 JK |
459 | { |
460 | struct address_space *mapping = inode->i_mapping; | |
461 | struct page *page; | |
462 | repeat: | |
a56c7c6f | 463 | page = get_read_data_page(inode, index, READ_SYNC, for_write); |
43f3eae1 JK |
464 | if (IS_ERR(page)) |
465 | return page; | |
393ff91f | 466 | |
43f3eae1 | 467 | /* wait for read completion */ |
393ff91f | 468 | lock_page(page); |
6bacf52f | 469 | if (unlikely(!PageUptodate(page))) { |
393ff91f JK |
470 | f2fs_put_page(page, 1); |
471 | return ERR_PTR(-EIO); | |
eb47b800 | 472 | } |
6bacf52f | 473 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
474 | f2fs_put_page(page, 1); |
475 | goto repeat; | |
eb47b800 JK |
476 | } |
477 | return page; | |
478 | } | |
479 | ||
0a8165d7 | 480 | /* |
eb47b800 JK |
481 | * Caller ensures that this data page is never allocated. |
482 | * A new zero-filled data page is allocated in the page cache. | |
39936837 | 483 | * |
4f4124d0 CY |
484 | * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and |
485 | * f2fs_unlock_op(). | |
470f00e9 CY |
486 | * Note that, ipage is set only by make_empty_dir, and if any error occur, |
487 | * ipage should be released by this function. | |
eb47b800 | 488 | */ |
64aa7ed9 | 489 | struct page *get_new_data_page(struct inode *inode, |
a8865372 | 490 | struct page *ipage, pgoff_t index, bool new_i_size) |
eb47b800 | 491 | { |
eb47b800 JK |
492 | struct address_space *mapping = inode->i_mapping; |
493 | struct page *page; | |
494 | struct dnode_of_data dn; | |
495 | int err; | |
7612118a | 496 | |
a56c7c6f | 497 | page = f2fs_grab_cache_page(mapping, index, true); |
470f00e9 CY |
498 | if (!page) { |
499 | /* | |
500 | * before exiting, we should make sure ipage will be released | |
501 | * if any error occur. | |
502 | */ | |
503 | f2fs_put_page(ipage, 1); | |
01f28610 | 504 | return ERR_PTR(-ENOMEM); |
470f00e9 | 505 | } |
eb47b800 | 506 | |
a8865372 | 507 | set_new_dnode(&dn, inode, ipage, NULL, 0); |
b600965c | 508 | err = f2fs_reserve_block(&dn, index); |
01f28610 JK |
509 | if (err) { |
510 | f2fs_put_page(page, 1); | |
eb47b800 | 511 | return ERR_PTR(err); |
a8865372 | 512 | } |
01f28610 JK |
513 | if (!ipage) |
514 | f2fs_put_dnode(&dn); | |
eb47b800 JK |
515 | |
516 | if (PageUptodate(page)) | |
01f28610 | 517 | goto got_it; |
eb47b800 JK |
518 | |
519 | if (dn.data_blkaddr == NEW_ADDR) { | |
09cbfeaf | 520 | zero_user_segment(page, 0, PAGE_SIZE); |
393ff91f | 521 | SetPageUptodate(page); |
eb47b800 | 522 | } else { |
4375a336 | 523 | f2fs_put_page(page, 1); |
a8865372 | 524 | |
7612118a JK |
525 | /* if ipage exists, blkaddr should be NEW_ADDR */ |
526 | f2fs_bug_on(F2FS_I_SB(inode), ipage); | |
527 | page = get_lock_data_page(inode, index, true); | |
4375a336 | 528 | if (IS_ERR(page)) |
7612118a | 529 | return page; |
eb47b800 | 530 | } |
01f28610 | 531 | got_it: |
9edcdabf | 532 | if (new_i_size && i_size_read(inode) < |
09cbfeaf KS |
533 | ((loff_t)(index + 1) << PAGE_SHIFT)) { |
534 | i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT)); | |
699489bb JK |
535 | /* Only the directory inode sets new_i_size */ |
536 | set_inode_flag(F2FS_I(inode), FI_UPDATE_DIR); | |
eb47b800 JK |
537 | } |
538 | return page; | |
539 | } | |
540 | ||
bfad7c2d JK |
541 | static int __allocate_data_block(struct dnode_of_data *dn) |
542 | { | |
4081363f | 543 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
bfad7c2d | 544 | struct f2fs_summary sum; |
bfad7c2d | 545 | struct node_info ni; |
38aa0889 | 546 | int seg = CURSEG_WARM_DATA; |
976e4c50 | 547 | pgoff_t fofs; |
bfad7c2d JK |
548 | |
549 | if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))) | |
550 | return -EPERM; | |
df6136ef CY |
551 | |
552 | dn->data_blkaddr = datablock_addr(dn->node_page, dn->ofs_in_node); | |
553 | if (dn->data_blkaddr == NEW_ADDR) | |
554 | goto alloc; | |
555 | ||
bfad7c2d JK |
556 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, 1))) |
557 | return -ENOSPC; | |
558 | ||
df6136ef | 559 | alloc: |
bfad7c2d JK |
560 | get_node_info(sbi, dn->nid, &ni); |
561 | set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version); | |
562 | ||
38aa0889 JK |
563 | if (dn->ofs_in_node == 0 && dn->inode_page == dn->node_page) |
564 | seg = CURSEG_DIRECT_IO; | |
565 | ||
df6136ef CY |
566 | allocate_data_block(sbi, NULL, dn->data_blkaddr, &dn->data_blkaddr, |
567 | &sum, seg); | |
216a620a | 568 | set_data_blkaddr(dn); |
bfad7c2d | 569 | |
976e4c50 | 570 | /* update i_size */ |
81ca7350 | 571 | fofs = start_bidx_of_node(ofs_of_node(dn->node_page), dn->inode) + |
976e4c50 | 572 | dn->ofs_in_node; |
09cbfeaf | 573 | if (i_size_read(dn->inode) < ((loff_t)(fofs + 1) << PAGE_SHIFT)) |
9edcdabf | 574 | i_size_write(dn->inode, |
09cbfeaf | 575 | ((loff_t)(fofs + 1) << PAGE_SHIFT)); |
bfad7c2d JK |
576 | return 0; |
577 | } | |
578 | ||
b439b103 | 579 | ssize_t f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *from) |
59b802e5 | 580 | { |
b439b103 | 581 | struct inode *inode = file_inode(iocb->ki_filp); |
5b8db7fa | 582 | struct f2fs_map_blocks map; |
b439b103 | 583 | ssize_t ret = 0; |
59b802e5 | 584 | |
b439b103 | 585 | map.m_lblk = F2FS_BYTES_TO_BLK(iocb->ki_pos); |
24b84912 | 586 | map.m_len = F2FS_BLK_ALIGN(iov_iter_count(from)); |
da85985c | 587 | map.m_next_pgofs = NULL; |
2a340760 | 588 | |
24b84912 JK |
589 | if (f2fs_encrypted_inode(inode)) |
590 | return 0; | |
591 | ||
592 | if (iocb->ki_flags & IOCB_DIRECT) { | |
593 | ret = f2fs_convert_inline_inode(inode); | |
594 | if (ret) | |
595 | return ret; | |
596 | return f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO); | |
597 | } | |
598 | if (iocb->ki_pos + iov_iter_count(from) > MAX_INLINE_DATA) { | |
b439b103 JK |
599 | ret = f2fs_convert_inline_inode(inode); |
600 | if (ret) | |
601 | return ret; | |
b439b103 | 602 | } |
24b84912 JK |
603 | if (!f2fs_has_inline_data(inode)) |
604 | return f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO); | |
b439b103 | 605 | return ret; |
59b802e5 JK |
606 | } |
607 | ||
0a8165d7 | 608 | /* |
003a3e1d JK |
609 | * f2fs_map_blocks() now supported readahead/bmap/rw direct_IO with |
610 | * f2fs_map_blocks structure. | |
4f4124d0 CY |
611 | * If original data blocks are allocated, then give them to blockdev. |
612 | * Otherwise, | |
613 | * a. preallocate requested block addresses | |
614 | * b. do not use extent cache for better performance | |
615 | * c. give the block addresses to blockdev | |
eb47b800 | 616 | */ |
d323d005 | 617 | int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map, |
e2b4e2bc | 618 | int create, int flag) |
eb47b800 | 619 | { |
003a3e1d | 620 | unsigned int maxblocks = map->m_len; |
eb47b800 | 621 | struct dnode_of_data dn; |
f9811703 | 622 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
bfad7c2d JK |
623 | int mode = create ? ALLOC_NODE : LOOKUP_NODE_RA; |
624 | pgoff_t pgofs, end_offset; | |
625 | int err = 0, ofs = 1; | |
a2e7d1bf | 626 | struct extent_info ei; |
bfad7c2d | 627 | bool allocated = false; |
7df3a431 | 628 | block_t blkaddr; |
eb47b800 | 629 | |
003a3e1d JK |
630 | map->m_len = 0; |
631 | map->m_flags = 0; | |
632 | ||
633 | /* it only supports block size == page size */ | |
634 | pgofs = (pgoff_t)map->m_lblk; | |
eb47b800 | 635 | |
24b84912 | 636 | if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) { |
003a3e1d JK |
637 | map->m_pblk = ei.blk + pgofs - ei.fofs; |
638 | map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs); | |
639 | map->m_flags = F2FS_MAP_MAPPED; | |
bfad7c2d | 640 | goto out; |
a2e7d1bf | 641 | } |
bfad7c2d | 642 | |
4fe71e88 | 643 | next_dnode: |
59b802e5 | 644 | if (create) |
3104af35 | 645 | f2fs_lock_op(sbi); |
eb47b800 JK |
646 | |
647 | /* When reading holes, we need its node page */ | |
648 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
bfad7c2d | 649 | err = get_dnode_of_data(&dn, pgofs, mode); |
1ec79083 | 650 | if (err) { |
da85985c | 651 | if (err == -ENOENT) { |
bfad7c2d | 652 | err = 0; |
da85985c CY |
653 | if (map->m_next_pgofs) |
654 | *map->m_next_pgofs = | |
655 | get_next_page_offset(&dn, pgofs); | |
656 | } | |
bfad7c2d | 657 | goto unlock_out; |
848753aa | 658 | } |
973163fc | 659 | |
81ca7350 | 660 | end_offset = ADDRS_PER_PAGE(dn.node_page, inode); |
4fe71e88 CY |
661 | |
662 | next_block: | |
663 | blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node); | |
664 | ||
665 | if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR) { | |
973163fc | 666 | if (create) { |
f9811703 CY |
667 | if (unlikely(f2fs_cp_error(sbi))) { |
668 | err = -EIO; | |
4fe71e88 | 669 | goto sync_out; |
f9811703 | 670 | } |
24b84912 JK |
671 | if (flag == F2FS_GET_BLOCK_PRE_AIO) { |
672 | if (blkaddr == NULL_ADDR) | |
673 | err = reserve_new_block(&dn); | |
674 | } else { | |
675 | err = __allocate_data_block(&dn); | |
676 | } | |
973163fc | 677 | if (err) |
4fe71e88 | 678 | goto sync_out; |
973163fc CY |
679 | allocated = true; |
680 | map->m_flags = F2FS_MAP_NEW; | |
4fe71e88 | 681 | blkaddr = dn.data_blkaddr; |
973163fc | 682 | } else { |
da85985c CY |
683 | if (flag == F2FS_GET_BLOCK_FIEMAP && |
684 | blkaddr == NULL_ADDR) { | |
685 | if (map->m_next_pgofs) | |
686 | *map->m_next_pgofs = pgofs + 1; | |
687 | } | |
973163fc | 688 | if (flag != F2FS_GET_BLOCK_FIEMAP || |
4fe71e88 | 689 | blkaddr != NEW_ADDR) { |
973163fc CY |
690 | if (flag == F2FS_GET_BLOCK_BMAP) |
691 | err = -ENOENT; | |
4fe71e88 | 692 | goto sync_out; |
973163fc | 693 | } |
e2b4e2bc | 694 | } |
e2b4e2bc | 695 | } |
eb47b800 | 696 | |
4fe71e88 CY |
697 | if (map->m_len == 0) { |
698 | /* preallocated unwritten block should be mapped for fiemap. */ | |
699 | if (blkaddr == NEW_ADDR) | |
700 | map->m_flags |= F2FS_MAP_UNWRITTEN; | |
701 | map->m_flags |= F2FS_MAP_MAPPED; | |
702 | ||
703 | map->m_pblk = blkaddr; | |
704 | map->m_len = 1; | |
705 | } else if ((map->m_pblk != NEW_ADDR && | |
706 | blkaddr == (map->m_pblk + ofs)) || | |
b439b103 | 707 | (map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR) || |
24b84912 JK |
708 | flag == F2FS_GET_BLOCK_PRE_DIO || |
709 | flag == F2FS_GET_BLOCK_PRE_AIO) { | |
4fe71e88 CY |
710 | ofs++; |
711 | map->m_len++; | |
712 | } else { | |
713 | goto sync_out; | |
714 | } | |
bfad7c2d | 715 | |
bfad7c2d JK |
716 | dn.ofs_in_node++; |
717 | pgofs++; | |
718 | ||
4fe71e88 CY |
719 | if (map->m_len < maxblocks) { |
720 | if (dn.ofs_in_node < end_offset) | |
721 | goto next_block; | |
7df3a431 | 722 | |
bfad7c2d JK |
723 | if (allocated) |
724 | sync_inode_page(&dn); | |
bfad7c2d JK |
725 | f2fs_put_dnode(&dn); |
726 | ||
3104af35 CY |
727 | if (create) { |
728 | f2fs_unlock_op(sbi); | |
3c082b7b | 729 | f2fs_balance_fs(sbi, allocated); |
3104af35 | 730 | } |
3c082b7b | 731 | allocated = false; |
4fe71e88 | 732 | goto next_dnode; |
eb47b800 | 733 | } |
7df3a431 | 734 | |
bfad7c2d JK |
735 | sync_out: |
736 | if (allocated) | |
737 | sync_inode_page(&dn); | |
eb47b800 | 738 | f2fs_put_dnode(&dn); |
bfad7c2d | 739 | unlock_out: |
2a340760 | 740 | if (create) { |
3104af35 | 741 | f2fs_unlock_op(sbi); |
3c082b7b | 742 | f2fs_balance_fs(sbi, allocated); |
2a340760 | 743 | } |
bfad7c2d | 744 | out: |
003a3e1d | 745 | trace_f2fs_map_blocks(inode, map, err); |
bfad7c2d | 746 | return err; |
eb47b800 JK |
747 | } |
748 | ||
003a3e1d | 749 | static int __get_data_block(struct inode *inode, sector_t iblock, |
da85985c CY |
750 | struct buffer_head *bh, int create, int flag, |
751 | pgoff_t *next_pgofs) | |
003a3e1d JK |
752 | { |
753 | struct f2fs_map_blocks map; | |
754 | int ret; | |
755 | ||
756 | map.m_lblk = iblock; | |
757 | map.m_len = bh->b_size >> inode->i_blkbits; | |
da85985c | 758 | map.m_next_pgofs = next_pgofs; |
003a3e1d | 759 | |
e2b4e2bc | 760 | ret = f2fs_map_blocks(inode, &map, create, flag); |
003a3e1d JK |
761 | if (!ret) { |
762 | map_bh(bh, inode->i_sb, map.m_pblk); | |
763 | bh->b_state = (bh->b_state & ~F2FS_MAP_FLAGS) | map.m_flags; | |
764 | bh->b_size = map.m_len << inode->i_blkbits; | |
765 | } | |
766 | return ret; | |
767 | } | |
768 | ||
ccfb3000 | 769 | static int get_data_block(struct inode *inode, sector_t iblock, |
da85985c CY |
770 | struct buffer_head *bh_result, int create, int flag, |
771 | pgoff_t *next_pgofs) | |
e2b4e2bc | 772 | { |
da85985c CY |
773 | return __get_data_block(inode, iblock, bh_result, create, |
774 | flag, next_pgofs); | |
e2b4e2bc CY |
775 | } |
776 | ||
777 | static int get_data_block_dio(struct inode *inode, sector_t iblock, | |
ccfb3000 JK |
778 | struct buffer_head *bh_result, int create) |
779 | { | |
e2b4e2bc | 780 | return __get_data_block(inode, iblock, bh_result, create, |
da85985c | 781 | F2FS_GET_BLOCK_DIO, NULL); |
ccfb3000 JK |
782 | } |
783 | ||
e2b4e2bc | 784 | static int get_data_block_bmap(struct inode *inode, sector_t iblock, |
ccfb3000 JK |
785 | struct buffer_head *bh_result, int create) |
786 | { | |
179448bf | 787 | /* Block number less than F2FS MAX BLOCKS */ |
e0afc4d6 | 788 | if (unlikely(iblock >= F2FS_I_SB(inode)->max_file_blocks)) |
179448bf YH |
789 | return -EFBIG; |
790 | ||
e2b4e2bc | 791 | return __get_data_block(inode, iblock, bh_result, create, |
da85985c | 792 | F2FS_GET_BLOCK_BMAP, NULL); |
ccfb3000 JK |
793 | } |
794 | ||
7f63eb77 JK |
795 | static inline sector_t logical_to_blk(struct inode *inode, loff_t offset) |
796 | { | |
797 | return (offset >> inode->i_blkbits); | |
798 | } | |
799 | ||
800 | static inline loff_t blk_to_logical(struct inode *inode, sector_t blk) | |
801 | { | |
802 | return (blk << inode->i_blkbits); | |
803 | } | |
804 | ||
9ab70134 JK |
805 | int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, |
806 | u64 start, u64 len) | |
807 | { | |
7f63eb77 JK |
808 | struct buffer_head map_bh; |
809 | sector_t start_blk, last_blk; | |
da85985c | 810 | pgoff_t next_pgofs; |
de1475cc | 811 | loff_t isize; |
7f63eb77 JK |
812 | u64 logical = 0, phys = 0, size = 0; |
813 | u32 flags = 0; | |
7f63eb77 JK |
814 | int ret = 0; |
815 | ||
816 | ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC); | |
817 | if (ret) | |
818 | return ret; | |
819 | ||
67f8cf3c JK |
820 | if (f2fs_has_inline_data(inode)) { |
821 | ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len); | |
822 | if (ret != -EAGAIN) | |
823 | return ret; | |
824 | } | |
825 | ||
5955102c | 826 | inode_lock(inode); |
de1475cc FL |
827 | |
828 | isize = i_size_read(inode); | |
9a950d52 FL |
829 | if (start >= isize) |
830 | goto out; | |
7f63eb77 | 831 | |
9a950d52 FL |
832 | if (start + len > isize) |
833 | len = isize - start; | |
7f63eb77 JK |
834 | |
835 | if (logical_to_blk(inode, len) == 0) | |
836 | len = blk_to_logical(inode, 1); | |
837 | ||
838 | start_blk = logical_to_blk(inode, start); | |
839 | last_blk = logical_to_blk(inode, start + len - 1); | |
9a950d52 | 840 | |
7f63eb77 JK |
841 | next: |
842 | memset(&map_bh, 0, sizeof(struct buffer_head)); | |
843 | map_bh.b_size = len; | |
844 | ||
e2b4e2bc | 845 | ret = get_data_block(inode, start_blk, &map_bh, 0, |
da85985c | 846 | F2FS_GET_BLOCK_FIEMAP, &next_pgofs); |
7f63eb77 JK |
847 | if (ret) |
848 | goto out; | |
849 | ||
850 | /* HOLE */ | |
851 | if (!buffer_mapped(&map_bh)) { | |
da85985c | 852 | start_blk = next_pgofs; |
9a950d52 | 853 | /* Go through holes util pass the EOF */ |
da85985c | 854 | if (blk_to_logical(inode, start_blk) < isize) |
9a950d52 FL |
855 | goto prep_next; |
856 | /* Found a hole beyond isize means no more extents. | |
857 | * Note that the premise is that filesystems don't | |
858 | * punch holes beyond isize and keep size unchanged. | |
859 | */ | |
860 | flags |= FIEMAP_EXTENT_LAST; | |
861 | } | |
7f63eb77 | 862 | |
da5af127 CY |
863 | if (size) { |
864 | if (f2fs_encrypted_inode(inode)) | |
865 | flags |= FIEMAP_EXTENT_DATA_ENCRYPTED; | |
866 | ||
9a950d52 FL |
867 | ret = fiemap_fill_next_extent(fieinfo, logical, |
868 | phys, size, flags); | |
da5af127 | 869 | } |
7f63eb77 | 870 | |
9a950d52 FL |
871 | if (start_blk > last_blk || ret) |
872 | goto out; | |
7f63eb77 | 873 | |
9a950d52 FL |
874 | logical = blk_to_logical(inode, start_blk); |
875 | phys = blk_to_logical(inode, map_bh.b_blocknr); | |
876 | size = map_bh.b_size; | |
877 | flags = 0; | |
878 | if (buffer_unwritten(&map_bh)) | |
879 | flags = FIEMAP_EXTENT_UNWRITTEN; | |
7f63eb77 | 880 | |
9a950d52 | 881 | start_blk += logical_to_blk(inode, size); |
7f63eb77 | 882 | |
9a950d52 | 883 | prep_next: |
7f63eb77 JK |
884 | cond_resched(); |
885 | if (fatal_signal_pending(current)) | |
886 | ret = -EINTR; | |
887 | else | |
888 | goto next; | |
889 | out: | |
890 | if (ret == 1) | |
891 | ret = 0; | |
892 | ||
5955102c | 893 | inode_unlock(inode); |
7f63eb77 | 894 | return ret; |
9ab70134 JK |
895 | } |
896 | ||
f1e88660 JK |
897 | /* |
898 | * This function was originally taken from fs/mpage.c, and customized for f2fs. | |
899 | * Major change was from block_size == page_size in f2fs by default. | |
900 | */ | |
901 | static int f2fs_mpage_readpages(struct address_space *mapping, | |
902 | struct list_head *pages, struct page *page, | |
903 | unsigned nr_pages) | |
904 | { | |
905 | struct bio *bio = NULL; | |
906 | unsigned page_idx; | |
907 | sector_t last_block_in_bio = 0; | |
908 | struct inode *inode = mapping->host; | |
909 | const unsigned blkbits = inode->i_blkbits; | |
910 | const unsigned blocksize = 1 << blkbits; | |
911 | sector_t block_in_file; | |
912 | sector_t last_block; | |
913 | sector_t last_block_in_file; | |
914 | sector_t block_nr; | |
915 | struct block_device *bdev = inode->i_sb->s_bdev; | |
916 | struct f2fs_map_blocks map; | |
917 | ||
918 | map.m_pblk = 0; | |
919 | map.m_lblk = 0; | |
920 | map.m_len = 0; | |
921 | map.m_flags = 0; | |
da85985c | 922 | map.m_next_pgofs = NULL; |
f1e88660 JK |
923 | |
924 | for (page_idx = 0; nr_pages; page_idx++, nr_pages--) { | |
925 | ||
926 | prefetchw(&page->flags); | |
927 | if (pages) { | |
928 | page = list_entry(pages->prev, struct page, lru); | |
929 | list_del(&page->lru); | |
930 | if (add_to_page_cache_lru(page, mapping, | |
931 | page->index, GFP_KERNEL)) | |
932 | goto next_page; | |
933 | } | |
934 | ||
935 | block_in_file = (sector_t)page->index; | |
936 | last_block = block_in_file + nr_pages; | |
937 | last_block_in_file = (i_size_read(inode) + blocksize - 1) >> | |
938 | blkbits; | |
939 | if (last_block > last_block_in_file) | |
940 | last_block = last_block_in_file; | |
941 | ||
942 | /* | |
943 | * Map blocks using the previous result first. | |
944 | */ | |
945 | if ((map.m_flags & F2FS_MAP_MAPPED) && | |
946 | block_in_file > map.m_lblk && | |
947 | block_in_file < (map.m_lblk + map.m_len)) | |
948 | goto got_it; | |
949 | ||
950 | /* | |
951 | * Then do more f2fs_map_blocks() calls until we are | |
952 | * done with this page. | |
953 | */ | |
954 | map.m_flags = 0; | |
955 | ||
956 | if (block_in_file < last_block) { | |
957 | map.m_lblk = block_in_file; | |
958 | map.m_len = last_block - block_in_file; | |
959 | ||
46c9e141 | 960 | if (f2fs_map_blocks(inode, &map, 0, |
da85985c | 961 | F2FS_GET_BLOCK_READ)) |
f1e88660 JK |
962 | goto set_error_page; |
963 | } | |
964 | got_it: | |
965 | if ((map.m_flags & F2FS_MAP_MAPPED)) { | |
966 | block_nr = map.m_pblk + block_in_file - map.m_lblk; | |
967 | SetPageMappedToDisk(page); | |
968 | ||
969 | if (!PageUptodate(page) && !cleancache_get_page(page)) { | |
970 | SetPageUptodate(page); | |
971 | goto confused; | |
972 | } | |
973 | } else { | |
09cbfeaf | 974 | zero_user_segment(page, 0, PAGE_SIZE); |
f1e88660 JK |
975 | SetPageUptodate(page); |
976 | unlock_page(page); | |
977 | goto next_page; | |
978 | } | |
979 | ||
980 | /* | |
981 | * This page will go to BIO. Do we need to send this | |
982 | * BIO off first? | |
983 | */ | |
984 | if (bio && (last_block_in_bio != block_nr - 1)) { | |
985 | submit_and_realloc: | |
986 | submit_bio(READ, bio); | |
987 | bio = NULL; | |
988 | } | |
989 | if (bio == NULL) { | |
0b81d077 | 990 | struct fscrypt_ctx *ctx = NULL; |
4375a336 JK |
991 | |
992 | if (f2fs_encrypted_inode(inode) && | |
993 | S_ISREG(inode->i_mode)) { | |
4375a336 | 994 | |
0b81d077 | 995 | ctx = fscrypt_get_ctx(inode); |
4375a336 JK |
996 | if (IS_ERR(ctx)) |
997 | goto set_error_page; | |
998 | ||
999 | /* wait the page to be moved by cleaning */ | |
08b39fbd CY |
1000 | f2fs_wait_on_encrypted_page_writeback( |
1001 | F2FS_I_SB(inode), block_nr); | |
4375a336 JK |
1002 | } |
1003 | ||
f1e88660 | 1004 | bio = bio_alloc(GFP_KERNEL, |
b54ffb73 | 1005 | min_t(int, nr_pages, BIO_MAX_PAGES)); |
4375a336 JK |
1006 | if (!bio) { |
1007 | if (ctx) | |
0b81d077 | 1008 | fscrypt_release_ctx(ctx); |
f1e88660 | 1009 | goto set_error_page; |
4375a336 | 1010 | } |
f1e88660 JK |
1011 | bio->bi_bdev = bdev; |
1012 | bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(block_nr); | |
12377024 | 1013 | bio->bi_end_io = f2fs_read_end_io; |
4375a336 | 1014 | bio->bi_private = ctx; |
f1e88660 JK |
1015 | } |
1016 | ||
1017 | if (bio_add_page(bio, page, blocksize, 0) < blocksize) | |
1018 | goto submit_and_realloc; | |
1019 | ||
1020 | last_block_in_bio = block_nr; | |
1021 | goto next_page; | |
1022 | set_error_page: | |
1023 | SetPageError(page); | |
09cbfeaf | 1024 | zero_user_segment(page, 0, PAGE_SIZE); |
f1e88660 JK |
1025 | unlock_page(page); |
1026 | goto next_page; | |
1027 | confused: | |
1028 | if (bio) { | |
1029 | submit_bio(READ, bio); | |
1030 | bio = NULL; | |
1031 | } | |
1032 | unlock_page(page); | |
1033 | next_page: | |
1034 | if (pages) | |
09cbfeaf | 1035 | put_page(page); |
f1e88660 JK |
1036 | } |
1037 | BUG_ON(pages && !list_empty(pages)); | |
1038 | if (bio) | |
1039 | submit_bio(READ, bio); | |
1040 | return 0; | |
1041 | } | |
1042 | ||
eb47b800 JK |
1043 | static int f2fs_read_data_page(struct file *file, struct page *page) |
1044 | { | |
9ffe0fb5 | 1045 | struct inode *inode = page->mapping->host; |
b3d208f9 | 1046 | int ret = -EAGAIN; |
9ffe0fb5 | 1047 | |
c20e89cd CY |
1048 | trace_f2fs_readpage(page, DATA); |
1049 | ||
e1c42045 | 1050 | /* If the file has inline data, try to read it directly */ |
9ffe0fb5 HL |
1051 | if (f2fs_has_inline_data(inode)) |
1052 | ret = f2fs_read_inline_data(inode, page); | |
b3d208f9 | 1053 | if (ret == -EAGAIN) |
f1e88660 | 1054 | ret = f2fs_mpage_readpages(page->mapping, NULL, page, 1); |
9ffe0fb5 | 1055 | return ret; |
eb47b800 JK |
1056 | } |
1057 | ||
1058 | static int f2fs_read_data_pages(struct file *file, | |
1059 | struct address_space *mapping, | |
1060 | struct list_head *pages, unsigned nr_pages) | |
1061 | { | |
9ffe0fb5 | 1062 | struct inode *inode = file->f_mapping->host; |
b8c29400 CY |
1063 | struct page *page = list_entry(pages->prev, struct page, lru); |
1064 | ||
1065 | trace_f2fs_readpages(inode, page, nr_pages); | |
9ffe0fb5 HL |
1066 | |
1067 | /* If the file has inline data, skip readpages */ | |
1068 | if (f2fs_has_inline_data(inode)) | |
1069 | return 0; | |
1070 | ||
f1e88660 | 1071 | return f2fs_mpage_readpages(mapping, pages, NULL, nr_pages); |
eb47b800 JK |
1072 | } |
1073 | ||
05ca3632 | 1074 | int do_write_data_page(struct f2fs_io_info *fio) |
eb47b800 | 1075 | { |
05ca3632 | 1076 | struct page *page = fio->page; |
eb47b800 | 1077 | struct inode *inode = page->mapping->host; |
eb47b800 JK |
1078 | struct dnode_of_data dn; |
1079 | int err = 0; | |
1080 | ||
1081 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
266e97a8 | 1082 | err = get_dnode_of_data(&dn, page->index, LOOKUP_NODE); |
eb47b800 JK |
1083 | if (err) |
1084 | return err; | |
1085 | ||
28bc106b | 1086 | fio->old_blkaddr = dn.data_blkaddr; |
eb47b800 JK |
1087 | |
1088 | /* This page is already truncated */ | |
7a9d7548 | 1089 | if (fio->old_blkaddr == NULL_ADDR) { |
2bca1e23 | 1090 | ClearPageUptodate(page); |
eb47b800 | 1091 | goto out_writepage; |
2bca1e23 | 1092 | } |
eb47b800 | 1093 | |
4375a336 | 1094 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) { |
08b39fbd CY |
1095 | |
1096 | /* wait for GCed encrypted page writeback */ | |
1097 | f2fs_wait_on_encrypted_page_writeback(F2FS_I_SB(inode), | |
7a9d7548 | 1098 | fio->old_blkaddr); |
08b39fbd | 1099 | |
0b81d077 | 1100 | fio->encrypted_page = fscrypt_encrypt_page(inode, fio->page); |
4375a336 JK |
1101 | if (IS_ERR(fio->encrypted_page)) { |
1102 | err = PTR_ERR(fio->encrypted_page); | |
1103 | goto out_writepage; | |
1104 | } | |
1105 | } | |
1106 | ||
eb47b800 JK |
1107 | set_page_writeback(page); |
1108 | ||
1109 | /* | |
1110 | * If current allocation needs SSR, | |
1111 | * it had better in-place writes for updated data. | |
1112 | */ | |
7a9d7548 | 1113 | if (unlikely(fio->old_blkaddr != NEW_ADDR && |
b25958b6 | 1114 | !is_cold_data(page) && |
2da3e027 | 1115 | !IS_ATOMIC_WRITTEN_PAGE(page) && |
b25958b6 | 1116 | need_inplace_update(inode))) { |
05ca3632 | 1117 | rewrite_data_page(fio); |
fff04f90 | 1118 | set_inode_flag(F2FS_I(inode), FI_UPDATE_WRITE); |
8ce67cb0 | 1119 | trace_f2fs_do_write_data_page(page, IPU); |
eb47b800 | 1120 | } else { |
05ca3632 | 1121 | write_data_page(&dn, fio); |
8ce67cb0 | 1122 | trace_f2fs_do_write_data_page(page, OPU); |
fff04f90 | 1123 | set_inode_flag(F2FS_I(inode), FI_APPEND_WRITE); |
3c6c2beb JK |
1124 | if (page->index == 0) |
1125 | set_inode_flag(F2FS_I(inode), FI_FIRST_BLOCK_WRITTEN); | |
eb47b800 JK |
1126 | } |
1127 | out_writepage: | |
1128 | f2fs_put_dnode(&dn); | |
1129 | return err; | |
1130 | } | |
1131 | ||
1132 | static int f2fs_write_data_page(struct page *page, | |
1133 | struct writeback_control *wbc) | |
1134 | { | |
1135 | struct inode *inode = page->mapping->host; | |
4081363f | 1136 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
eb47b800 JK |
1137 | loff_t i_size = i_size_read(inode); |
1138 | const pgoff_t end_index = ((unsigned long long) i_size) | |
09cbfeaf | 1139 | >> PAGE_SHIFT; |
9ffe0fb5 | 1140 | unsigned offset = 0; |
39936837 | 1141 | bool need_balance_fs = false; |
eb47b800 | 1142 | int err = 0; |
458e6197 | 1143 | struct f2fs_io_info fio = { |
05ca3632 | 1144 | .sbi = sbi, |
458e6197 | 1145 | .type = DATA, |
6c311ec6 | 1146 | .rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE, |
05ca3632 | 1147 | .page = page, |
4375a336 | 1148 | .encrypted_page = NULL, |
458e6197 | 1149 | }; |
eb47b800 | 1150 | |
ecda0de3 CY |
1151 | trace_f2fs_writepage(page, DATA); |
1152 | ||
eb47b800 | 1153 | if (page->index < end_index) |
39936837 | 1154 | goto write; |
eb47b800 JK |
1155 | |
1156 | /* | |
1157 | * If the offset is out-of-range of file size, | |
1158 | * this page does not have to be written to disk. | |
1159 | */ | |
09cbfeaf | 1160 | offset = i_size & (PAGE_SIZE - 1); |
76f60268 | 1161 | if ((page->index >= end_index + 1) || !offset) |
39936837 | 1162 | goto out; |
eb47b800 | 1163 | |
09cbfeaf | 1164 | zero_user_segment(page, offset, PAGE_SIZE); |
39936837 | 1165 | write: |
caf0047e | 1166 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
eb47b800 | 1167 | goto redirty_out; |
1e84371f JK |
1168 | if (f2fs_is_drop_cache(inode)) |
1169 | goto out; | |
1170 | if (f2fs_is_volatile_file(inode) && !wbc->for_reclaim && | |
1171 | available_free_memory(sbi, BASE_CHECK)) | |
1172 | goto redirty_out; | |
eb47b800 | 1173 | |
39936837 | 1174 | /* Dentry blocks are controlled by checkpoint */ |
eb47b800 | 1175 | if (S_ISDIR(inode->i_mode)) { |
cf779cab JK |
1176 | if (unlikely(f2fs_cp_error(sbi))) |
1177 | goto redirty_out; | |
05ca3632 | 1178 | err = do_write_data_page(&fio); |
8618b881 JK |
1179 | goto done; |
1180 | } | |
9ffe0fb5 | 1181 | |
cf779cab JK |
1182 | /* we should bypass data pages to proceed the kworkder jobs */ |
1183 | if (unlikely(f2fs_cp_error(sbi))) { | |
1184 | SetPageError(page); | |
a7ffdbe2 | 1185 | goto out; |
cf779cab JK |
1186 | } |
1187 | ||
8618b881 | 1188 | if (!wbc->for_reclaim) |
39936837 | 1189 | need_balance_fs = true; |
8618b881 | 1190 | else if (has_not_enough_free_secs(sbi, 0)) |
39936837 | 1191 | goto redirty_out; |
eb47b800 | 1192 | |
b3d208f9 | 1193 | err = -EAGAIN; |
8618b881 | 1194 | f2fs_lock_op(sbi); |
b3d208f9 JK |
1195 | if (f2fs_has_inline_data(inode)) |
1196 | err = f2fs_write_inline_data(inode, page); | |
1197 | if (err == -EAGAIN) | |
05ca3632 | 1198 | err = do_write_data_page(&fio); |
8618b881 JK |
1199 | f2fs_unlock_op(sbi); |
1200 | done: | |
1201 | if (err && err != -ENOENT) | |
1202 | goto redirty_out; | |
eb47b800 | 1203 | |
eb47b800 | 1204 | clear_cold_data(page); |
39936837 | 1205 | out: |
a7ffdbe2 | 1206 | inode_dec_dirty_pages(inode); |
2bca1e23 JK |
1207 | if (err) |
1208 | ClearPageUptodate(page); | |
0c3a5797 CY |
1209 | |
1210 | if (wbc->for_reclaim) { | |
1211 | f2fs_submit_merged_bio_cond(sbi, NULL, page, 0, DATA, WRITE); | |
1212 | remove_dirty_inode(inode); | |
1213 | } | |
1214 | ||
eb47b800 | 1215 | unlock_page(page); |
2c4db1a6 | 1216 | f2fs_balance_fs(sbi, need_balance_fs); |
0c3a5797 CY |
1217 | |
1218 | if (unlikely(f2fs_cp_error(sbi))) | |
2aea39ec | 1219 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
0c3a5797 | 1220 | |
eb47b800 JK |
1221 | return 0; |
1222 | ||
eb47b800 | 1223 | redirty_out: |
76f60268 | 1224 | redirty_page_for_writepage(wbc, page); |
8618b881 | 1225 | return AOP_WRITEPAGE_ACTIVATE; |
eb47b800 JK |
1226 | } |
1227 | ||
fa9150a8 NJ |
1228 | static int __f2fs_writepage(struct page *page, struct writeback_control *wbc, |
1229 | void *data) | |
1230 | { | |
1231 | struct address_space *mapping = data; | |
1232 | int ret = mapping->a_ops->writepage(page, wbc); | |
1233 | mapping_set_error(mapping, ret); | |
1234 | return ret; | |
1235 | } | |
1236 | ||
8f46dcae CY |
1237 | /* |
1238 | * This function was copied from write_cche_pages from mm/page-writeback.c. | |
1239 | * The major change is making write step of cold data page separately from | |
1240 | * warm/hot data page. | |
1241 | */ | |
1242 | static int f2fs_write_cache_pages(struct address_space *mapping, | |
1243 | struct writeback_control *wbc, writepage_t writepage, | |
1244 | void *data) | |
1245 | { | |
1246 | int ret = 0; | |
1247 | int done = 0; | |
1248 | struct pagevec pvec; | |
1249 | int nr_pages; | |
1250 | pgoff_t uninitialized_var(writeback_index); | |
1251 | pgoff_t index; | |
1252 | pgoff_t end; /* Inclusive */ | |
1253 | pgoff_t done_index; | |
1254 | int cycled; | |
1255 | int range_whole = 0; | |
1256 | int tag; | |
1257 | int step = 0; | |
1258 | ||
1259 | pagevec_init(&pvec, 0); | |
1260 | next: | |
1261 | if (wbc->range_cyclic) { | |
1262 | writeback_index = mapping->writeback_index; /* prev offset */ | |
1263 | index = writeback_index; | |
1264 | if (index == 0) | |
1265 | cycled = 1; | |
1266 | else | |
1267 | cycled = 0; | |
1268 | end = -1; | |
1269 | } else { | |
09cbfeaf KS |
1270 | index = wbc->range_start >> PAGE_SHIFT; |
1271 | end = wbc->range_end >> PAGE_SHIFT; | |
8f46dcae CY |
1272 | if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX) |
1273 | range_whole = 1; | |
1274 | cycled = 1; /* ignore range_cyclic tests */ | |
1275 | } | |
1276 | if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages) | |
1277 | tag = PAGECACHE_TAG_TOWRITE; | |
1278 | else | |
1279 | tag = PAGECACHE_TAG_DIRTY; | |
1280 | retry: | |
1281 | if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages) | |
1282 | tag_pages_for_writeback(mapping, index, end); | |
1283 | done_index = index; | |
1284 | while (!done && (index <= end)) { | |
1285 | int i; | |
1286 | ||
1287 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag, | |
1288 | min(end - index, (pgoff_t)PAGEVEC_SIZE - 1) + 1); | |
1289 | if (nr_pages == 0) | |
1290 | break; | |
1291 | ||
1292 | for (i = 0; i < nr_pages; i++) { | |
1293 | struct page *page = pvec.pages[i]; | |
1294 | ||
1295 | if (page->index > end) { | |
1296 | done = 1; | |
1297 | break; | |
1298 | } | |
1299 | ||
1300 | done_index = page->index; | |
1301 | ||
1302 | lock_page(page); | |
1303 | ||
1304 | if (unlikely(page->mapping != mapping)) { | |
1305 | continue_unlock: | |
1306 | unlock_page(page); | |
1307 | continue; | |
1308 | } | |
1309 | ||
1310 | if (!PageDirty(page)) { | |
1311 | /* someone wrote it for us */ | |
1312 | goto continue_unlock; | |
1313 | } | |
1314 | ||
737f1899 | 1315 | if (step == is_cold_data(page)) |
8f46dcae CY |
1316 | goto continue_unlock; |
1317 | ||
1318 | if (PageWriteback(page)) { | |
1319 | if (wbc->sync_mode != WB_SYNC_NONE) | |
fec1d657 JK |
1320 | f2fs_wait_on_page_writeback(page, |
1321 | DATA, true); | |
8f46dcae CY |
1322 | else |
1323 | goto continue_unlock; | |
1324 | } | |
1325 | ||
1326 | BUG_ON(PageWriteback(page)); | |
1327 | if (!clear_page_dirty_for_io(page)) | |
1328 | goto continue_unlock; | |
1329 | ||
1330 | ret = (*writepage)(page, wbc, data); | |
1331 | if (unlikely(ret)) { | |
1332 | if (ret == AOP_WRITEPAGE_ACTIVATE) { | |
1333 | unlock_page(page); | |
1334 | ret = 0; | |
1335 | } else { | |
1336 | done_index = page->index + 1; | |
1337 | done = 1; | |
1338 | break; | |
1339 | } | |
1340 | } | |
1341 | ||
1342 | if (--wbc->nr_to_write <= 0 && | |
1343 | wbc->sync_mode == WB_SYNC_NONE) { | |
1344 | done = 1; | |
1345 | break; | |
1346 | } | |
1347 | } | |
1348 | pagevec_release(&pvec); | |
1349 | cond_resched(); | |
1350 | } | |
1351 | ||
1352 | if (step < 1) { | |
1353 | step++; | |
1354 | goto next; | |
1355 | } | |
1356 | ||
1357 | if (!cycled && !done) { | |
1358 | cycled = 1; | |
1359 | index = 0; | |
1360 | end = writeback_index - 1; | |
1361 | goto retry; | |
1362 | } | |
1363 | if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) | |
1364 | mapping->writeback_index = done_index; | |
1365 | ||
1366 | return ret; | |
1367 | } | |
1368 | ||
25ca923b | 1369 | static int f2fs_write_data_pages(struct address_space *mapping, |
eb47b800 JK |
1370 | struct writeback_control *wbc) |
1371 | { | |
1372 | struct inode *inode = mapping->host; | |
4081363f | 1373 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
5463e7c1 | 1374 | bool locked = false; |
eb47b800 | 1375 | int ret; |
50c8cdb3 | 1376 | long diff; |
eb47b800 | 1377 | |
cfb185a1 | 1378 | /* deal with chardevs and other special file */ |
1379 | if (!mapping->a_ops->writepage) | |
1380 | return 0; | |
1381 | ||
6a290544 CY |
1382 | /* skip writing if there is no dirty page in this inode */ |
1383 | if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE) | |
1384 | return 0; | |
1385 | ||
a1257023 JK |
1386 | if (S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_NONE && |
1387 | get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) && | |
1388 | available_free_memory(sbi, DIRTY_DENTS)) | |
1389 | goto skip_write; | |
1390 | ||
d323d005 CY |
1391 | /* skip writing during file defragment */ |
1392 | if (is_inode_flag_set(F2FS_I(inode), FI_DO_DEFRAG)) | |
1393 | goto skip_write; | |
1394 | ||
d5669f7b JK |
1395 | /* during POR, we don't need to trigger writepage at all. */ |
1396 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) | |
1397 | goto skip_write; | |
1398 | ||
d31c7c3f YH |
1399 | trace_f2fs_writepages(mapping->host, wbc, DATA); |
1400 | ||
50c8cdb3 | 1401 | diff = nr_pages_to_write(sbi, DATA, wbc); |
eb47b800 | 1402 | |
25c13551 | 1403 | if (!S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_ALL) { |
5463e7c1 JK |
1404 | mutex_lock(&sbi->writepages); |
1405 | locked = true; | |
1406 | } | |
8f46dcae | 1407 | ret = f2fs_write_cache_pages(mapping, wbc, __f2fs_writepage, mapping); |
0c3a5797 | 1408 | f2fs_submit_merged_bio_cond(sbi, inode, NULL, 0, DATA, WRITE); |
5463e7c1 JK |
1409 | if (locked) |
1410 | mutex_unlock(&sbi->writepages); | |
458e6197 | 1411 | |
c227f912 | 1412 | remove_dirty_inode(inode); |
eb47b800 | 1413 | |
50c8cdb3 | 1414 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff); |
eb47b800 | 1415 | return ret; |
d3baf95d JK |
1416 | |
1417 | skip_write: | |
a7ffdbe2 | 1418 | wbc->pages_skipped += get_dirty_pages(inode); |
d31c7c3f | 1419 | trace_f2fs_writepages(mapping->host, wbc, DATA); |
d3baf95d | 1420 | return 0; |
eb47b800 JK |
1421 | } |
1422 | ||
3aab8f82 CY |
1423 | static void f2fs_write_failed(struct address_space *mapping, loff_t to) |
1424 | { | |
1425 | struct inode *inode = mapping->host; | |
819d9153 | 1426 | loff_t i_size = i_size_read(inode); |
3aab8f82 | 1427 | |
819d9153 JK |
1428 | if (to > i_size) { |
1429 | truncate_pagecache(inode, i_size); | |
1430 | truncate_blocks(inode, i_size, true); | |
3aab8f82 CY |
1431 | } |
1432 | } | |
1433 | ||
2aadac08 JK |
1434 | static int prepare_write_begin(struct f2fs_sb_info *sbi, |
1435 | struct page *page, loff_t pos, unsigned len, | |
1436 | block_t *blk_addr, bool *node_changed) | |
1437 | { | |
1438 | struct inode *inode = page->mapping->host; | |
1439 | pgoff_t index = page->index; | |
1440 | struct dnode_of_data dn; | |
1441 | struct page *ipage; | |
b4d07a3e JK |
1442 | bool locked = false; |
1443 | struct extent_info ei; | |
2aadac08 JK |
1444 | int err = 0; |
1445 | ||
24b84912 JK |
1446 | /* |
1447 | * we already allocated all the blocks, so we don't need to get | |
1448 | * the block addresses when there is no need to fill the page. | |
1449 | */ | |
1450 | if (!f2fs_has_inline_data(inode) && !f2fs_encrypted_inode(inode) && | |
09cbfeaf | 1451 | len == PAGE_SIZE) |
24b84912 JK |
1452 | return 0; |
1453 | ||
b4d07a3e | 1454 | if (f2fs_has_inline_data(inode) || |
09cbfeaf | 1455 | (pos & PAGE_MASK) >= i_size_read(inode)) { |
b4d07a3e JK |
1456 | f2fs_lock_op(sbi); |
1457 | locked = true; | |
1458 | } | |
1459 | restart: | |
2aadac08 JK |
1460 | /* check inline_data */ |
1461 | ipage = get_node_page(sbi, inode->i_ino); | |
1462 | if (IS_ERR(ipage)) { | |
1463 | err = PTR_ERR(ipage); | |
1464 | goto unlock_out; | |
1465 | } | |
1466 | ||
1467 | set_new_dnode(&dn, inode, ipage, ipage, 0); | |
1468 | ||
1469 | if (f2fs_has_inline_data(inode)) { | |
1470 | if (pos + len <= MAX_INLINE_DATA) { | |
1471 | read_inline_data(page, ipage); | |
1472 | set_inode_flag(F2FS_I(inode), FI_DATA_EXIST); | |
2049d4fc | 1473 | set_inline_node(ipage); |
2aadac08 JK |
1474 | } else { |
1475 | err = f2fs_convert_inline_page(&dn, page); | |
1476 | if (err) | |
b4d07a3e JK |
1477 | goto out; |
1478 | if (dn.data_blkaddr == NULL_ADDR) | |
1479 | err = f2fs_get_block(&dn, index); | |
1480 | } | |
1481 | } else if (locked) { | |
1482 | err = f2fs_get_block(&dn, index); | |
1483 | } else { | |
1484 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { | |
1485 | dn.data_blkaddr = ei.blk + index - ei.fofs; | |
1486 | } else { | |
b4d07a3e JK |
1487 | /* hole case */ |
1488 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); | |
b483fadf | 1489 | if (err || (!err && dn.data_blkaddr == NULL_ADDR)) { |
b4d07a3e JK |
1490 | f2fs_put_dnode(&dn); |
1491 | f2fs_lock_op(sbi); | |
1492 | locked = true; | |
1493 | goto restart; | |
1494 | } | |
2aadac08 JK |
1495 | } |
1496 | } | |
b4d07a3e | 1497 | |
2aadac08 JK |
1498 | /* convert_inline_page can make node_changed */ |
1499 | *blk_addr = dn.data_blkaddr; | |
1500 | *node_changed = dn.node_changed; | |
b4d07a3e | 1501 | out: |
2aadac08 JK |
1502 | f2fs_put_dnode(&dn); |
1503 | unlock_out: | |
b4d07a3e JK |
1504 | if (locked) |
1505 | f2fs_unlock_op(sbi); | |
2aadac08 JK |
1506 | return err; |
1507 | } | |
1508 | ||
eb47b800 JK |
1509 | static int f2fs_write_begin(struct file *file, struct address_space *mapping, |
1510 | loff_t pos, unsigned len, unsigned flags, | |
1511 | struct page **pagep, void **fsdata) | |
1512 | { | |
1513 | struct inode *inode = mapping->host; | |
4081363f | 1514 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
86531d6b | 1515 | struct page *page = NULL; |
09cbfeaf | 1516 | pgoff_t index = ((unsigned long long) pos) >> PAGE_SHIFT; |
2aadac08 JK |
1517 | bool need_balance = false; |
1518 | block_t blkaddr = NULL_ADDR; | |
eb47b800 JK |
1519 | int err = 0; |
1520 | ||
62aed044 CY |
1521 | trace_f2fs_write_begin(inode, pos, len, flags); |
1522 | ||
5f727395 JK |
1523 | /* |
1524 | * We should check this at this moment to avoid deadlock on inode page | |
1525 | * and #0 page. The locking rule for inline_data conversion should be: | |
1526 | * lock_page(page #0) -> lock_page(inode_page) | |
1527 | */ | |
1528 | if (index != 0) { | |
1529 | err = f2fs_convert_inline_inode(inode); | |
1530 | if (err) | |
1531 | goto fail; | |
1532 | } | |
afcb7ca0 | 1533 | repeat: |
eb47b800 | 1534 | page = grab_cache_page_write_begin(mapping, index, flags); |
3aab8f82 CY |
1535 | if (!page) { |
1536 | err = -ENOMEM; | |
1537 | goto fail; | |
1538 | } | |
d5f66990 | 1539 | |
eb47b800 JK |
1540 | *pagep = page; |
1541 | ||
2aadac08 JK |
1542 | err = prepare_write_begin(sbi, page, pos, len, |
1543 | &blkaddr, &need_balance); | |
9ba69cf9 | 1544 | if (err) |
2aadac08 | 1545 | goto fail; |
9ba69cf9 | 1546 | |
2aadac08 | 1547 | if (need_balance && has_not_enough_free_secs(sbi, 0)) { |
2a340760 | 1548 | unlock_page(page); |
2c4db1a6 | 1549 | f2fs_balance_fs(sbi, true); |
2a340760 JK |
1550 | lock_page(page); |
1551 | if (page->mapping != mapping) { | |
1552 | /* The page got truncated from under us */ | |
1553 | f2fs_put_page(page, 1); | |
1554 | goto repeat; | |
1555 | } | |
1556 | } | |
1557 | ||
fec1d657 | 1558 | f2fs_wait_on_page_writeback(page, DATA, false); |
b3d208f9 | 1559 | |
08b39fbd CY |
1560 | /* wait for GCed encrypted page writeback */ |
1561 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) | |
2aadac08 | 1562 | f2fs_wait_on_encrypted_page_writeback(sbi, blkaddr); |
08b39fbd | 1563 | |
09cbfeaf | 1564 | if (len == PAGE_SIZE) |
90d4388a CY |
1565 | goto out_update; |
1566 | if (PageUptodate(page)) | |
1567 | goto out_clear; | |
eb47b800 | 1568 | |
09cbfeaf KS |
1569 | if ((pos & PAGE_MASK) >= i_size_read(inode)) { |
1570 | unsigned start = pos & (PAGE_SIZE - 1); | |
eb47b800 JK |
1571 | unsigned end = start + len; |
1572 | ||
1573 | /* Reading beyond i_size is simple: memset to zero */ | |
09cbfeaf | 1574 | zero_user_segments(page, 0, start, end, PAGE_SIZE); |
90d4388a | 1575 | goto out_update; |
eb47b800 JK |
1576 | } |
1577 | ||
2aadac08 | 1578 | if (blkaddr == NEW_ADDR) { |
09cbfeaf | 1579 | zero_user_segment(page, 0, PAGE_SIZE); |
eb47b800 | 1580 | } else { |
cf04e8eb | 1581 | struct f2fs_io_info fio = { |
05ca3632 | 1582 | .sbi = sbi, |
cf04e8eb JK |
1583 | .type = DATA, |
1584 | .rw = READ_SYNC, | |
7a9d7548 CY |
1585 | .old_blkaddr = blkaddr, |
1586 | .new_blkaddr = blkaddr, | |
05ca3632 | 1587 | .page = page, |
4375a336 | 1588 | .encrypted_page = NULL, |
cf04e8eb | 1589 | }; |
05ca3632 | 1590 | err = f2fs_submit_page_bio(&fio); |
9234f319 JK |
1591 | if (err) |
1592 | goto fail; | |
d54c795b | 1593 | |
393ff91f | 1594 | lock_page(page); |
6bacf52f | 1595 | if (unlikely(!PageUptodate(page))) { |
3aab8f82 CY |
1596 | err = -EIO; |
1597 | goto fail; | |
eb47b800 | 1598 | } |
6bacf52f | 1599 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
1600 | f2fs_put_page(page, 1); |
1601 | goto repeat; | |
eb47b800 | 1602 | } |
4375a336 JK |
1603 | |
1604 | /* avoid symlink page */ | |
1605 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) { | |
0b81d077 | 1606 | err = fscrypt_decrypt_page(page); |
86531d6b | 1607 | if (err) |
4375a336 | 1608 | goto fail; |
4375a336 | 1609 | } |
eb47b800 | 1610 | } |
90d4388a | 1611 | out_update: |
eb47b800 | 1612 | SetPageUptodate(page); |
90d4388a | 1613 | out_clear: |
eb47b800 JK |
1614 | clear_cold_data(page); |
1615 | return 0; | |
9ba69cf9 | 1616 | |
3aab8f82 | 1617 | fail: |
86531d6b | 1618 | f2fs_put_page(page, 1); |
3aab8f82 CY |
1619 | f2fs_write_failed(mapping, pos + len); |
1620 | return err; | |
eb47b800 JK |
1621 | } |
1622 | ||
a1dd3c13 JK |
1623 | static int f2fs_write_end(struct file *file, |
1624 | struct address_space *mapping, | |
1625 | loff_t pos, unsigned len, unsigned copied, | |
1626 | struct page *page, void *fsdata) | |
1627 | { | |
1628 | struct inode *inode = page->mapping->host; | |
1629 | ||
dfb2bf38 CY |
1630 | trace_f2fs_write_end(inode, pos, len, copied); |
1631 | ||
34ba94ba | 1632 | set_page_dirty(page); |
a1dd3c13 JK |
1633 | |
1634 | if (pos + copied > i_size_read(inode)) { | |
1635 | i_size_write(inode, pos + copied); | |
1636 | mark_inode_dirty(inode); | |
a1dd3c13 JK |
1637 | } |
1638 | ||
75c3c8bc | 1639 | f2fs_put_page(page, 1); |
d0239e1b | 1640 | f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); |
a1dd3c13 JK |
1641 | return copied; |
1642 | } | |
1643 | ||
6f673763 OS |
1644 | static int check_direct_IO(struct inode *inode, struct iov_iter *iter, |
1645 | loff_t offset) | |
944fcfc1 JK |
1646 | { |
1647 | unsigned blocksize_mask = inode->i_sb->s_blocksize - 1; | |
944fcfc1 | 1648 | |
944fcfc1 JK |
1649 | if (offset & blocksize_mask) |
1650 | return -EINVAL; | |
1651 | ||
5b46f25d AV |
1652 | if (iov_iter_alignment(iter) & blocksize_mask) |
1653 | return -EINVAL; | |
1654 | ||
944fcfc1 JK |
1655 | return 0; |
1656 | } | |
1657 | ||
c8b8e32d | 1658 | static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter) |
eb47b800 | 1659 | { |
b439b103 | 1660 | struct address_space *mapping = iocb->ki_filp->f_mapping; |
3aab8f82 CY |
1661 | struct inode *inode = mapping->host; |
1662 | size_t count = iov_iter_count(iter); | |
c8b8e32d | 1663 | loff_t offset = iocb->ki_pos; |
3aab8f82 | 1664 | int err; |
944fcfc1 | 1665 | |
b439b103 | 1666 | err = check_direct_IO(inode, iter, offset); |
b9d777b8 JK |
1667 | if (err) |
1668 | return err; | |
9ffe0fb5 | 1669 | |
fcc85a4d JK |
1670 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) |
1671 | return 0; | |
1672 | ||
6f673763 | 1673 | trace_f2fs_direct_IO_enter(inode, offset, count, iov_iter_rw(iter)); |
70407fad | 1674 | |
c8b8e32d | 1675 | err = blockdev_direct_IO(iocb, inode, iter, get_data_block_dio); |
6f673763 | 1676 | if (err < 0 && iov_iter_rw(iter) == WRITE) |
3aab8f82 | 1677 | f2fs_write_failed(mapping, offset + count); |
70407fad | 1678 | |
6f673763 | 1679 | trace_f2fs_direct_IO_exit(inode, offset, count, iov_iter_rw(iter), err); |
70407fad | 1680 | |
3aab8f82 | 1681 | return err; |
eb47b800 JK |
1682 | } |
1683 | ||
487261f3 CY |
1684 | void f2fs_invalidate_page(struct page *page, unsigned int offset, |
1685 | unsigned int length) | |
eb47b800 JK |
1686 | { |
1687 | struct inode *inode = page->mapping->host; | |
487261f3 | 1688 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
a7ffdbe2 | 1689 | |
487261f3 | 1690 | if (inode->i_ino >= F2FS_ROOT_INO(sbi) && |
09cbfeaf | 1691 | (offset % PAGE_SIZE || length != PAGE_SIZE)) |
a7ffdbe2 JK |
1692 | return; |
1693 | ||
487261f3 CY |
1694 | if (PageDirty(page)) { |
1695 | if (inode->i_ino == F2FS_META_INO(sbi)) | |
1696 | dec_page_count(sbi, F2FS_DIRTY_META); | |
1697 | else if (inode->i_ino == F2FS_NODE_INO(sbi)) | |
1698 | dec_page_count(sbi, F2FS_DIRTY_NODES); | |
1699 | else | |
1700 | inode_dec_dirty_pages(inode); | |
1701 | } | |
decd36b6 CY |
1702 | |
1703 | /* This is atomic written page, keep Private */ | |
1704 | if (IS_ATOMIC_WRITTEN_PAGE(page)) | |
1705 | return; | |
1706 | ||
eb47b800 JK |
1707 | ClearPagePrivate(page); |
1708 | } | |
1709 | ||
487261f3 | 1710 | int f2fs_release_page(struct page *page, gfp_t wait) |
eb47b800 | 1711 | { |
f68daeeb JK |
1712 | /* If this is dirty page, keep PagePrivate */ |
1713 | if (PageDirty(page)) | |
1714 | return 0; | |
1715 | ||
decd36b6 CY |
1716 | /* This is atomic written page, keep Private */ |
1717 | if (IS_ATOMIC_WRITTEN_PAGE(page)) | |
1718 | return 0; | |
1719 | ||
eb47b800 | 1720 | ClearPagePrivate(page); |
c3850aa1 | 1721 | return 1; |
eb47b800 JK |
1722 | } |
1723 | ||
1724 | static int f2fs_set_data_page_dirty(struct page *page) | |
1725 | { | |
1726 | struct address_space *mapping = page->mapping; | |
1727 | struct inode *inode = mapping->host; | |
1728 | ||
26c6b887 JK |
1729 | trace_f2fs_set_page_dirty(page, DATA); |
1730 | ||
eb47b800 | 1731 | SetPageUptodate(page); |
34ba94ba | 1732 | |
1e84371f | 1733 | if (f2fs_is_atomic_file(inode)) { |
decd36b6 CY |
1734 | if (!IS_ATOMIC_WRITTEN_PAGE(page)) { |
1735 | register_inmem_page(inode, page); | |
1736 | return 1; | |
1737 | } | |
1738 | /* | |
1739 | * Previously, this page has been registered, we just | |
1740 | * return here. | |
1741 | */ | |
1742 | return 0; | |
34ba94ba JK |
1743 | } |
1744 | ||
eb47b800 JK |
1745 | if (!PageDirty(page)) { |
1746 | __set_page_dirty_nobuffers(page); | |
a7ffdbe2 | 1747 | update_dirty_page(inode, page); |
eb47b800 JK |
1748 | return 1; |
1749 | } | |
1750 | return 0; | |
1751 | } | |
1752 | ||
c01e54b7 JK |
1753 | static sector_t f2fs_bmap(struct address_space *mapping, sector_t block) |
1754 | { | |
454ae7e5 CY |
1755 | struct inode *inode = mapping->host; |
1756 | ||
1d373a0e JK |
1757 | if (f2fs_has_inline_data(inode)) |
1758 | return 0; | |
1759 | ||
1760 | /* make sure allocating whole blocks */ | |
1761 | if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) | |
1762 | filemap_write_and_wait(mapping); | |
1763 | ||
e2b4e2bc | 1764 | return generic_block_bmap(mapping, block, get_data_block_bmap); |
429511cd CY |
1765 | } |
1766 | ||
eb47b800 JK |
1767 | const struct address_space_operations f2fs_dblock_aops = { |
1768 | .readpage = f2fs_read_data_page, | |
1769 | .readpages = f2fs_read_data_pages, | |
1770 | .writepage = f2fs_write_data_page, | |
1771 | .writepages = f2fs_write_data_pages, | |
1772 | .write_begin = f2fs_write_begin, | |
a1dd3c13 | 1773 | .write_end = f2fs_write_end, |
eb47b800 | 1774 | .set_page_dirty = f2fs_set_data_page_dirty, |
487261f3 CY |
1775 | .invalidatepage = f2fs_invalidate_page, |
1776 | .releasepage = f2fs_release_page, | |
eb47b800 | 1777 | .direct_IO = f2fs_direct_IO, |
c01e54b7 | 1778 | .bmap = f2fs_bmap, |
eb47b800 | 1779 | }; |