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); |
38f91ca8 | 71 | f2fs_stop_checkpoint(sbi, true); |
93dfe2ac JK |
72 | } |
73 | end_page_writeback(page); | |
f568849e | 74 | } |
f5730184 JK |
75 | if (atomic_dec_and_test(&sbi->nr_wb_bios) && |
76 | wq_has_sleeper(&sbi->cp_wait)) | |
93dfe2ac JK |
77 | wake_up(&sbi->cp_wait); |
78 | ||
79 | bio_put(bio); | |
80 | } | |
81 | ||
940a6d34 GZ |
82 | /* |
83 | * Low-level block read/write IO operations. | |
84 | */ | |
85 | static struct bio *__bio_alloc(struct f2fs_sb_info *sbi, block_t blk_addr, | |
86 | int npages, bool is_read) | |
87 | { | |
88 | struct bio *bio; | |
89 | ||
740432f8 | 90 | bio = f2fs_bio_alloc(npages); |
940a6d34 GZ |
91 | |
92 | bio->bi_bdev = sbi->sb->s_bdev; | |
55cf9cb6 | 93 | bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr); |
940a6d34 | 94 | bio->bi_end_io = is_read ? f2fs_read_end_io : f2fs_write_end_io; |
12377024 | 95 | bio->bi_private = is_read ? NULL : sbi; |
940a6d34 GZ |
96 | |
97 | return bio; | |
98 | } | |
99 | ||
f5730184 JK |
100 | static inline void __submit_bio(struct f2fs_sb_info *sbi, int rw, |
101 | struct bio *bio) | |
102 | { | |
103 | if (!is_read_io(rw)) | |
104 | atomic_inc(&sbi->nr_wb_bios); | |
4e49ea4a MC |
105 | bio->bi_rw = rw; |
106 | submit_bio(bio); | |
f5730184 JK |
107 | } |
108 | ||
458e6197 | 109 | static void __submit_merged_bio(struct f2fs_bio_info *io) |
93dfe2ac | 110 | { |
458e6197 | 111 | struct f2fs_io_info *fio = &io->fio; |
93dfe2ac JK |
112 | |
113 | if (!io->bio) | |
114 | return; | |
115 | ||
6a8f8ca5 | 116 | if (is_read_io(fio->rw)) |
2ace38e0 | 117 | trace_f2fs_submit_read_bio(io->sbi->sb, fio, io->bio); |
6a8f8ca5 | 118 | else |
2ace38e0 | 119 | trace_f2fs_submit_write_bio(io->sbi->sb, fio, io->bio); |
940a6d34 | 120 | |
f5730184 | 121 | __submit_bio(io->sbi, fio->rw, io->bio); |
93dfe2ac JK |
122 | io->bio = NULL; |
123 | } | |
124 | ||
0c3a5797 CY |
125 | static bool __has_merged_page(struct f2fs_bio_info *io, struct inode *inode, |
126 | struct page *page, nid_t ino) | |
0fd785eb | 127 | { |
0fd785eb CY |
128 | struct bio_vec *bvec; |
129 | struct page *target; | |
130 | int i; | |
131 | ||
0c3a5797 | 132 | if (!io->bio) |
0fd785eb | 133 | return false; |
0c3a5797 CY |
134 | |
135 | if (!inode && !page && !ino) | |
136 | return true; | |
0fd785eb CY |
137 | |
138 | bio_for_each_segment_all(bvec, io->bio, i) { | |
139 | ||
0b81d077 | 140 | if (bvec->bv_page->mapping) |
0fd785eb | 141 | target = bvec->bv_page; |
0b81d077 JK |
142 | else |
143 | target = fscrypt_control_page(bvec->bv_page); | |
0fd785eb | 144 | |
0c3a5797 CY |
145 | if (inode && inode == target->mapping->host) |
146 | return true; | |
147 | if (page && page == target) | |
148 | return true; | |
149 | if (ino && ino == ino_of_node(target)) | |
0fd785eb | 150 | return true; |
0fd785eb CY |
151 | } |
152 | ||
0fd785eb CY |
153 | return false; |
154 | } | |
155 | ||
0c3a5797 CY |
156 | static bool has_merged_page(struct f2fs_sb_info *sbi, struct inode *inode, |
157 | struct page *page, nid_t ino, | |
158 | enum page_type type) | |
159 | { | |
160 | enum page_type btype = PAGE_TYPE_OF_BIO(type); | |
161 | struct f2fs_bio_info *io = &sbi->write_io[btype]; | |
162 | bool ret; | |
163 | ||
164 | down_read(&io->io_rwsem); | |
165 | ret = __has_merged_page(io, inode, page, ino); | |
166 | up_read(&io->io_rwsem); | |
167 | return ret; | |
168 | } | |
169 | ||
170 | static void __f2fs_submit_merged_bio(struct f2fs_sb_info *sbi, | |
171 | struct inode *inode, struct page *page, | |
172 | nid_t ino, enum page_type type, int rw) | |
93dfe2ac JK |
173 | { |
174 | enum page_type btype = PAGE_TYPE_OF_BIO(type); | |
175 | struct f2fs_bio_info *io; | |
176 | ||
177 | io = is_read_io(rw) ? &sbi->read_io : &sbi->write_io[btype]; | |
178 | ||
df0f8dc0 | 179 | down_write(&io->io_rwsem); |
458e6197 | 180 | |
0c3a5797 CY |
181 | if (!__has_merged_page(io, inode, page, ino)) |
182 | goto out; | |
183 | ||
458e6197 JK |
184 | /* change META to META_FLUSH in the checkpoint procedure */ |
185 | if (type >= META_FLUSH) { | |
186 | io->fio.type = META_FLUSH; | |
0f7b2abd JK |
187 | if (test_opt(sbi, NOBARRIER)) |
188 | io->fio.rw = WRITE_FLUSH | REQ_META | REQ_PRIO; | |
189 | else | |
190 | io->fio.rw = WRITE_FLUSH_FUA | REQ_META | REQ_PRIO; | |
458e6197 JK |
191 | } |
192 | __submit_merged_bio(io); | |
0c3a5797 | 193 | out: |
df0f8dc0 | 194 | up_write(&io->io_rwsem); |
93dfe2ac JK |
195 | } |
196 | ||
0c3a5797 CY |
197 | void f2fs_submit_merged_bio(struct f2fs_sb_info *sbi, enum page_type type, |
198 | int rw) | |
199 | { | |
200 | __f2fs_submit_merged_bio(sbi, NULL, NULL, 0, type, rw); | |
201 | } | |
202 | ||
203 | void f2fs_submit_merged_bio_cond(struct f2fs_sb_info *sbi, | |
204 | struct inode *inode, struct page *page, | |
205 | nid_t ino, enum page_type type, int rw) | |
206 | { | |
207 | if (has_merged_page(sbi, inode, page, ino, type)) | |
208 | __f2fs_submit_merged_bio(sbi, inode, page, ino, type, rw); | |
209 | } | |
210 | ||
406657dd CY |
211 | void f2fs_flush_merged_bios(struct f2fs_sb_info *sbi) |
212 | { | |
213 | f2fs_submit_merged_bio(sbi, DATA, WRITE); | |
214 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
215 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
216 | } | |
217 | ||
93dfe2ac JK |
218 | /* |
219 | * Fill the locked page with data located in the block address. | |
220 | * Return unlocked page. | |
221 | */ | |
05ca3632 | 222 | int f2fs_submit_page_bio(struct f2fs_io_info *fio) |
93dfe2ac | 223 | { |
93dfe2ac | 224 | struct bio *bio; |
0b81d077 JK |
225 | struct page *page = fio->encrypted_page ? |
226 | fio->encrypted_page : fio->page; | |
93dfe2ac | 227 | |
2ace38e0 | 228 | trace_f2fs_submit_page_bio(page, fio); |
05ca3632 | 229 | f2fs_trace_ios(fio, 0); |
93dfe2ac JK |
230 | |
231 | /* Allocate a new bio */ | |
7a9d7548 | 232 | bio = __bio_alloc(fio->sbi, fio->new_blkaddr, 1, is_read_io(fio->rw)); |
93dfe2ac | 233 | |
09cbfeaf | 234 | if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) { |
93dfe2ac | 235 | bio_put(bio); |
93dfe2ac JK |
236 | return -EFAULT; |
237 | } | |
238 | ||
f5730184 | 239 | __submit_bio(fio->sbi, fio->rw, bio); |
93dfe2ac JK |
240 | return 0; |
241 | } | |
242 | ||
05ca3632 | 243 | void f2fs_submit_page_mbio(struct f2fs_io_info *fio) |
93dfe2ac | 244 | { |
05ca3632 | 245 | struct f2fs_sb_info *sbi = fio->sbi; |
458e6197 | 246 | enum page_type btype = PAGE_TYPE_OF_BIO(fio->type); |
93dfe2ac | 247 | struct f2fs_bio_info *io; |
940a6d34 | 248 | bool is_read = is_read_io(fio->rw); |
4375a336 | 249 | struct page *bio_page; |
93dfe2ac | 250 | |
940a6d34 | 251 | io = is_read ? &sbi->read_io : &sbi->write_io[btype]; |
93dfe2ac | 252 | |
7a9d7548 CY |
253 | if (fio->old_blkaddr != NEW_ADDR) |
254 | verify_block_addr(sbi, fio->old_blkaddr); | |
255 | verify_block_addr(sbi, fio->new_blkaddr); | |
93dfe2ac | 256 | |
df0f8dc0 | 257 | down_write(&io->io_rwsem); |
93dfe2ac | 258 | |
7a9d7548 | 259 | if (io->bio && (io->last_block_in_bio != fio->new_blkaddr - 1 || |
458e6197 JK |
260 | io->fio.rw != fio->rw)) |
261 | __submit_merged_bio(io); | |
93dfe2ac JK |
262 | alloc_new: |
263 | if (io->bio == NULL) { | |
90a893c7 | 264 | int bio_blocks = MAX_BIO_BLOCKS(sbi); |
940a6d34 | 265 | |
7a9d7548 CY |
266 | io->bio = __bio_alloc(sbi, fio->new_blkaddr, |
267 | bio_blocks, is_read); | |
458e6197 | 268 | io->fio = *fio; |
93dfe2ac JK |
269 | } |
270 | ||
4375a336 JK |
271 | bio_page = fio->encrypted_page ? fio->encrypted_page : fio->page; |
272 | ||
09cbfeaf KS |
273 | if (bio_add_page(io->bio, bio_page, PAGE_SIZE, 0) < |
274 | PAGE_SIZE) { | |
458e6197 | 275 | __submit_merged_bio(io); |
93dfe2ac JK |
276 | goto alloc_new; |
277 | } | |
278 | ||
7a9d7548 | 279 | io->last_block_in_bio = fio->new_blkaddr; |
05ca3632 | 280 | f2fs_trace_ios(fio, 0); |
93dfe2ac | 281 | |
df0f8dc0 | 282 | up_write(&io->io_rwsem); |
05ca3632 | 283 | trace_f2fs_submit_page_mbio(fio->page, fio); |
93dfe2ac JK |
284 | } |
285 | ||
46008c6d CY |
286 | static void __set_data_blkaddr(struct dnode_of_data *dn) |
287 | { | |
288 | struct f2fs_node *rn = F2FS_NODE(dn->node_page); | |
289 | __le32 *addr_array; | |
290 | ||
291 | /* Get physical address of data block */ | |
292 | addr_array = blkaddr_in_node(rn); | |
293 | addr_array[dn->ofs_in_node] = cpu_to_le32(dn->data_blkaddr); | |
294 | } | |
295 | ||
0a8165d7 | 296 | /* |
eb47b800 JK |
297 | * Lock ordering for the change of data block address: |
298 | * ->data_page | |
299 | * ->node_page | |
300 | * update block addresses in the node page | |
301 | */ | |
216a620a | 302 | void set_data_blkaddr(struct dnode_of_data *dn) |
eb47b800 | 303 | { |
46008c6d CY |
304 | f2fs_wait_on_page_writeback(dn->node_page, NODE, true); |
305 | __set_data_blkaddr(dn); | |
306 | if (set_page_dirty(dn->node_page)) | |
12719ae1 | 307 | dn->node_changed = true; |
eb47b800 JK |
308 | } |
309 | ||
f28b3434 CY |
310 | void f2fs_update_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr) |
311 | { | |
312 | dn->data_blkaddr = blkaddr; | |
313 | set_data_blkaddr(dn); | |
314 | f2fs_update_extent_cache(dn); | |
315 | } | |
316 | ||
46008c6d CY |
317 | /* dn->ofs_in_node will be returned with up-to-date last block pointer */ |
318 | int reserve_new_blocks(struct dnode_of_data *dn, blkcnt_t count) | |
eb47b800 | 319 | { |
4081363f | 320 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
eb47b800 | 321 | |
46008c6d CY |
322 | if (!count) |
323 | return 0; | |
324 | ||
6bacf52f | 325 | if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))) |
eb47b800 | 326 | return -EPERM; |
46008c6d | 327 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, &count))) |
eb47b800 JK |
328 | return -ENOSPC; |
329 | ||
46008c6d CY |
330 | trace_f2fs_reserve_new_blocks(dn->inode, dn->nid, |
331 | dn->ofs_in_node, count); | |
332 | ||
333 | f2fs_wait_on_page_writeback(dn->node_page, NODE, true); | |
334 | ||
335 | for (; count > 0; dn->ofs_in_node++) { | |
336 | block_t blkaddr = | |
337 | datablock_addr(dn->node_page, dn->ofs_in_node); | |
338 | if (blkaddr == NULL_ADDR) { | |
339 | dn->data_blkaddr = NEW_ADDR; | |
340 | __set_data_blkaddr(dn); | |
341 | count--; | |
342 | } | |
343 | } | |
344 | ||
345 | if (set_page_dirty(dn->node_page)) | |
346 | dn->node_changed = true; | |
c01e2853 | 347 | |
a18ff063 | 348 | mark_inode_dirty(dn->inode); |
eb47b800 JK |
349 | sync_inode_page(dn); |
350 | return 0; | |
351 | } | |
352 | ||
46008c6d CY |
353 | /* Should keep dn->ofs_in_node unchanged */ |
354 | int reserve_new_block(struct dnode_of_data *dn) | |
355 | { | |
356 | unsigned int ofs_in_node = dn->ofs_in_node; | |
357 | int ret; | |
358 | ||
359 | ret = reserve_new_blocks(dn, 1); | |
360 | dn->ofs_in_node = ofs_in_node; | |
361 | return ret; | |
362 | } | |
363 | ||
b600965c HL |
364 | int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index) |
365 | { | |
366 | bool need_put = dn->inode_page ? false : true; | |
367 | int err; | |
368 | ||
369 | err = get_dnode_of_data(dn, index, ALLOC_NODE); | |
370 | if (err) | |
371 | return err; | |
a8865372 | 372 | |
b600965c HL |
373 | if (dn->data_blkaddr == NULL_ADDR) |
374 | err = reserve_new_block(dn); | |
a8865372 | 375 | if (err || need_put) |
b600965c HL |
376 | f2fs_put_dnode(dn); |
377 | return err; | |
378 | } | |
379 | ||
759af1c9 | 380 | int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index) |
eb47b800 | 381 | { |
028a41e8 | 382 | struct extent_info ei; |
759af1c9 | 383 | struct inode *inode = dn->inode; |
028a41e8 | 384 | |
759af1c9 FL |
385 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { |
386 | dn->data_blkaddr = ei.blk + index - ei.fofs; | |
387 | return 0; | |
429511cd | 388 | } |
028a41e8 | 389 | |
759af1c9 | 390 | return f2fs_reserve_block(dn, index); |
eb47b800 JK |
391 | } |
392 | ||
a56c7c6f JK |
393 | struct page *get_read_data_page(struct inode *inode, pgoff_t index, |
394 | int rw, bool for_write) | |
eb47b800 | 395 | { |
eb47b800 JK |
396 | struct address_space *mapping = inode->i_mapping; |
397 | struct dnode_of_data dn; | |
398 | struct page *page; | |
cb3bc9ee | 399 | struct extent_info ei; |
eb47b800 | 400 | int err; |
cf04e8eb | 401 | struct f2fs_io_info fio = { |
05ca3632 | 402 | .sbi = F2FS_I_SB(inode), |
cf04e8eb | 403 | .type = DATA, |
43f3eae1 | 404 | .rw = rw, |
4375a336 | 405 | .encrypted_page = NULL, |
cf04e8eb | 406 | }; |
eb47b800 | 407 | |
4375a336 JK |
408 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) |
409 | return read_mapping_page(mapping, index, NULL); | |
410 | ||
a56c7c6f | 411 | page = f2fs_grab_cache_page(mapping, index, for_write); |
650495de JK |
412 | if (!page) |
413 | return ERR_PTR(-ENOMEM); | |
414 | ||
cb3bc9ee CY |
415 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { |
416 | dn.data_blkaddr = ei.blk + index - ei.fofs; | |
417 | goto got_it; | |
418 | } | |
419 | ||
eb47b800 | 420 | set_new_dnode(&dn, inode, NULL, NULL, 0); |
266e97a8 | 421 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); |
86531d6b JK |
422 | if (err) |
423 | goto put_err; | |
eb47b800 JK |
424 | f2fs_put_dnode(&dn); |
425 | ||
6bacf52f | 426 | if (unlikely(dn.data_blkaddr == NULL_ADDR)) { |
86531d6b JK |
427 | err = -ENOENT; |
428 | goto put_err; | |
650495de | 429 | } |
cb3bc9ee | 430 | got_it: |
43f3eae1 JK |
431 | if (PageUptodate(page)) { |
432 | unlock_page(page); | |
eb47b800 | 433 | return page; |
43f3eae1 | 434 | } |
eb47b800 | 435 | |
d59ff4df JK |
436 | /* |
437 | * A new dentry page is allocated but not able to be written, since its | |
438 | * new inode page couldn't be allocated due to -ENOSPC. | |
439 | * In such the case, its blkaddr can be remained as NEW_ADDR. | |
440 | * see, f2fs_add_link -> get_new_data_page -> init_inode_metadata. | |
441 | */ | |
442 | if (dn.data_blkaddr == NEW_ADDR) { | |
09cbfeaf | 443 | zero_user_segment(page, 0, PAGE_SIZE); |
d59ff4df | 444 | SetPageUptodate(page); |
43f3eae1 | 445 | unlock_page(page); |
d59ff4df JK |
446 | return page; |
447 | } | |
eb47b800 | 448 | |
7a9d7548 | 449 | fio.new_blkaddr = fio.old_blkaddr = dn.data_blkaddr; |
05ca3632 JK |
450 | fio.page = page; |
451 | err = f2fs_submit_page_bio(&fio); | |
393ff91f | 452 | if (err) |
86531d6b | 453 | goto put_err; |
43f3eae1 | 454 | return page; |
86531d6b JK |
455 | |
456 | put_err: | |
457 | f2fs_put_page(page, 1); | |
458 | return ERR_PTR(err); | |
43f3eae1 JK |
459 | } |
460 | ||
461 | struct page *find_data_page(struct inode *inode, pgoff_t index) | |
462 | { | |
463 | struct address_space *mapping = inode->i_mapping; | |
464 | struct page *page; | |
465 | ||
466 | page = find_get_page(mapping, index); | |
467 | if (page && PageUptodate(page)) | |
468 | return page; | |
469 | f2fs_put_page(page, 0); | |
470 | ||
a56c7c6f | 471 | page = get_read_data_page(inode, index, READ_SYNC, false); |
43f3eae1 JK |
472 | if (IS_ERR(page)) |
473 | return page; | |
474 | ||
475 | if (PageUptodate(page)) | |
476 | return page; | |
477 | ||
478 | wait_on_page_locked(page); | |
479 | if (unlikely(!PageUptodate(page))) { | |
480 | f2fs_put_page(page, 0); | |
481 | return ERR_PTR(-EIO); | |
482 | } | |
483 | return page; | |
484 | } | |
485 | ||
486 | /* | |
487 | * If it tries to access a hole, return an error. | |
488 | * Because, the callers, functions in dir.c and GC, should be able to know | |
489 | * whether this page exists or not. | |
490 | */ | |
a56c7c6f JK |
491 | struct page *get_lock_data_page(struct inode *inode, pgoff_t index, |
492 | bool for_write) | |
43f3eae1 JK |
493 | { |
494 | struct address_space *mapping = inode->i_mapping; | |
495 | struct page *page; | |
496 | repeat: | |
a56c7c6f | 497 | page = get_read_data_page(inode, index, READ_SYNC, for_write); |
43f3eae1 JK |
498 | if (IS_ERR(page)) |
499 | return page; | |
393ff91f | 500 | |
43f3eae1 | 501 | /* wait for read completion */ |
393ff91f | 502 | lock_page(page); |
6bacf52f | 503 | if (unlikely(!PageUptodate(page))) { |
393ff91f JK |
504 | f2fs_put_page(page, 1); |
505 | return ERR_PTR(-EIO); | |
eb47b800 | 506 | } |
6bacf52f | 507 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
508 | f2fs_put_page(page, 1); |
509 | goto repeat; | |
eb47b800 JK |
510 | } |
511 | return page; | |
512 | } | |
513 | ||
0a8165d7 | 514 | /* |
eb47b800 JK |
515 | * Caller ensures that this data page is never allocated. |
516 | * A new zero-filled data page is allocated in the page cache. | |
39936837 | 517 | * |
4f4124d0 CY |
518 | * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and |
519 | * f2fs_unlock_op(). | |
470f00e9 CY |
520 | * Note that, ipage is set only by make_empty_dir, and if any error occur, |
521 | * ipage should be released by this function. | |
eb47b800 | 522 | */ |
64aa7ed9 | 523 | struct page *get_new_data_page(struct inode *inode, |
a8865372 | 524 | struct page *ipage, pgoff_t index, bool new_i_size) |
eb47b800 | 525 | { |
eb47b800 JK |
526 | struct address_space *mapping = inode->i_mapping; |
527 | struct page *page; | |
528 | struct dnode_of_data dn; | |
529 | int err; | |
7612118a | 530 | |
a56c7c6f | 531 | page = f2fs_grab_cache_page(mapping, index, true); |
470f00e9 CY |
532 | if (!page) { |
533 | /* | |
534 | * before exiting, we should make sure ipage will be released | |
535 | * if any error occur. | |
536 | */ | |
537 | f2fs_put_page(ipage, 1); | |
01f28610 | 538 | return ERR_PTR(-ENOMEM); |
470f00e9 | 539 | } |
eb47b800 | 540 | |
a8865372 | 541 | set_new_dnode(&dn, inode, ipage, NULL, 0); |
b600965c | 542 | err = f2fs_reserve_block(&dn, index); |
01f28610 JK |
543 | if (err) { |
544 | f2fs_put_page(page, 1); | |
eb47b800 | 545 | return ERR_PTR(err); |
a8865372 | 546 | } |
01f28610 JK |
547 | if (!ipage) |
548 | f2fs_put_dnode(&dn); | |
eb47b800 JK |
549 | |
550 | if (PageUptodate(page)) | |
01f28610 | 551 | goto got_it; |
eb47b800 JK |
552 | |
553 | if (dn.data_blkaddr == NEW_ADDR) { | |
09cbfeaf | 554 | zero_user_segment(page, 0, PAGE_SIZE); |
393ff91f | 555 | SetPageUptodate(page); |
eb47b800 | 556 | } else { |
4375a336 | 557 | f2fs_put_page(page, 1); |
a8865372 | 558 | |
7612118a JK |
559 | /* if ipage exists, blkaddr should be NEW_ADDR */ |
560 | f2fs_bug_on(F2FS_I_SB(inode), ipage); | |
561 | page = get_lock_data_page(inode, index, true); | |
4375a336 | 562 | if (IS_ERR(page)) |
7612118a | 563 | return page; |
eb47b800 | 564 | } |
01f28610 | 565 | got_it: |
9edcdabf | 566 | if (new_i_size && i_size_read(inode) < |
09cbfeaf KS |
567 | ((loff_t)(index + 1) << PAGE_SHIFT)) { |
568 | i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT)); | |
699489bb JK |
569 | /* Only the directory inode sets new_i_size */ |
570 | set_inode_flag(F2FS_I(inode), FI_UPDATE_DIR); | |
eb47b800 JK |
571 | } |
572 | return page; | |
573 | } | |
574 | ||
bfad7c2d JK |
575 | static int __allocate_data_block(struct dnode_of_data *dn) |
576 | { | |
4081363f | 577 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
bfad7c2d | 578 | struct f2fs_summary sum; |
bfad7c2d | 579 | struct node_info ni; |
38aa0889 | 580 | int seg = CURSEG_WARM_DATA; |
976e4c50 | 581 | pgoff_t fofs; |
46008c6d | 582 | blkcnt_t count = 1; |
bfad7c2d JK |
583 | |
584 | if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))) | |
585 | return -EPERM; | |
df6136ef CY |
586 | |
587 | dn->data_blkaddr = datablock_addr(dn->node_page, dn->ofs_in_node); | |
588 | if (dn->data_blkaddr == NEW_ADDR) | |
589 | goto alloc; | |
590 | ||
46008c6d | 591 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, &count))) |
bfad7c2d JK |
592 | return -ENOSPC; |
593 | ||
df6136ef | 594 | alloc: |
bfad7c2d JK |
595 | get_node_info(sbi, dn->nid, &ni); |
596 | set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version); | |
597 | ||
38aa0889 JK |
598 | if (dn->ofs_in_node == 0 && dn->inode_page == dn->node_page) |
599 | seg = CURSEG_DIRECT_IO; | |
600 | ||
df6136ef CY |
601 | allocate_data_block(sbi, NULL, dn->data_blkaddr, &dn->data_blkaddr, |
602 | &sum, seg); | |
216a620a | 603 | set_data_blkaddr(dn); |
bfad7c2d | 604 | |
976e4c50 | 605 | /* update i_size */ |
81ca7350 | 606 | fofs = start_bidx_of_node(ofs_of_node(dn->node_page), dn->inode) + |
976e4c50 | 607 | dn->ofs_in_node; |
09cbfeaf | 608 | if (i_size_read(dn->inode) < ((loff_t)(fofs + 1) << PAGE_SHIFT)) |
9edcdabf | 609 | i_size_write(dn->inode, |
09cbfeaf | 610 | ((loff_t)(fofs + 1) << PAGE_SHIFT)); |
bfad7c2d JK |
611 | return 0; |
612 | } | |
613 | ||
b439b103 | 614 | ssize_t f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *from) |
59b802e5 | 615 | { |
b439b103 | 616 | struct inode *inode = file_inode(iocb->ki_filp); |
5b8db7fa | 617 | struct f2fs_map_blocks map; |
b439b103 | 618 | ssize_t ret = 0; |
59b802e5 | 619 | |
0080c507 JK |
620 | map.m_lblk = F2FS_BLK_ALIGN(iocb->ki_pos); |
621 | map.m_len = F2FS_BYTES_TO_BLK(iov_iter_count(from)); | |
da85985c | 622 | map.m_next_pgofs = NULL; |
2a340760 | 623 | |
24b84912 JK |
624 | if (f2fs_encrypted_inode(inode)) |
625 | return 0; | |
626 | ||
627 | if (iocb->ki_flags & IOCB_DIRECT) { | |
628 | ret = f2fs_convert_inline_inode(inode); | |
629 | if (ret) | |
630 | return ret; | |
631 | return f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO); | |
632 | } | |
633 | if (iocb->ki_pos + iov_iter_count(from) > MAX_INLINE_DATA) { | |
b439b103 JK |
634 | ret = f2fs_convert_inline_inode(inode); |
635 | if (ret) | |
636 | return ret; | |
b439b103 | 637 | } |
24b84912 JK |
638 | if (!f2fs_has_inline_data(inode)) |
639 | return f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO); | |
b439b103 | 640 | return ret; |
59b802e5 JK |
641 | } |
642 | ||
0a8165d7 | 643 | /* |
003a3e1d JK |
644 | * f2fs_map_blocks() now supported readahead/bmap/rw direct_IO with |
645 | * f2fs_map_blocks structure. | |
4f4124d0 CY |
646 | * If original data blocks are allocated, then give them to blockdev. |
647 | * Otherwise, | |
648 | * a. preallocate requested block addresses | |
649 | * b. do not use extent cache for better performance | |
650 | * c. give the block addresses to blockdev | |
eb47b800 | 651 | */ |
d323d005 | 652 | int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map, |
e2b4e2bc | 653 | int create, int flag) |
eb47b800 | 654 | { |
003a3e1d | 655 | unsigned int maxblocks = map->m_len; |
eb47b800 | 656 | struct dnode_of_data dn; |
f9811703 | 657 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
bfad7c2d | 658 | int mode = create ? ALLOC_NODE : LOOKUP_NODE_RA; |
46008c6d | 659 | pgoff_t pgofs, end_offset, end; |
bfad7c2d | 660 | int err = 0, ofs = 1; |
46008c6d CY |
661 | unsigned int ofs_in_node, last_ofs_in_node; |
662 | blkcnt_t prealloc; | |
a2e7d1bf | 663 | struct extent_info ei; |
bfad7c2d | 664 | bool allocated = false; |
7df3a431 | 665 | block_t blkaddr; |
eb47b800 | 666 | |
003a3e1d JK |
667 | map->m_len = 0; |
668 | map->m_flags = 0; | |
669 | ||
670 | /* it only supports block size == page size */ | |
671 | pgofs = (pgoff_t)map->m_lblk; | |
46008c6d | 672 | end = pgofs + maxblocks; |
eb47b800 | 673 | |
24b84912 | 674 | if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) { |
003a3e1d JK |
675 | map->m_pblk = ei.blk + pgofs - ei.fofs; |
676 | map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs); | |
677 | map->m_flags = F2FS_MAP_MAPPED; | |
bfad7c2d | 678 | goto out; |
a2e7d1bf | 679 | } |
bfad7c2d | 680 | |
4fe71e88 | 681 | next_dnode: |
59b802e5 | 682 | if (create) |
3104af35 | 683 | f2fs_lock_op(sbi); |
eb47b800 JK |
684 | |
685 | /* When reading holes, we need its node page */ | |
686 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
bfad7c2d | 687 | err = get_dnode_of_data(&dn, pgofs, mode); |
1ec79083 | 688 | if (err) { |
43473f96 CY |
689 | if (flag == F2FS_GET_BLOCK_BMAP) |
690 | map->m_pblk = 0; | |
da85985c | 691 | if (err == -ENOENT) { |
bfad7c2d | 692 | err = 0; |
da85985c CY |
693 | if (map->m_next_pgofs) |
694 | *map->m_next_pgofs = | |
695 | get_next_page_offset(&dn, pgofs); | |
696 | } | |
bfad7c2d | 697 | goto unlock_out; |
848753aa | 698 | } |
973163fc | 699 | |
46008c6d CY |
700 | prealloc = 0; |
701 | ofs_in_node = dn.ofs_in_node; | |
81ca7350 | 702 | end_offset = ADDRS_PER_PAGE(dn.node_page, inode); |
4fe71e88 CY |
703 | |
704 | next_block: | |
705 | blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node); | |
706 | ||
707 | if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR) { | |
973163fc | 708 | if (create) { |
f9811703 CY |
709 | if (unlikely(f2fs_cp_error(sbi))) { |
710 | err = -EIO; | |
4fe71e88 | 711 | goto sync_out; |
f9811703 | 712 | } |
24b84912 | 713 | if (flag == F2FS_GET_BLOCK_PRE_AIO) { |
46008c6d CY |
714 | if (blkaddr == NULL_ADDR) { |
715 | prealloc++; | |
716 | last_ofs_in_node = dn.ofs_in_node; | |
717 | } | |
24b84912 JK |
718 | } else { |
719 | err = __allocate_data_block(&dn); | |
46008c6d | 720 | if (!err) { |
6bfc4919 JK |
721 | set_inode_flag(F2FS_I(inode), |
722 | FI_APPEND_WRITE); | |
46008c6d CY |
723 | allocated = true; |
724 | } | |
24b84912 | 725 | } |
973163fc | 726 | if (err) |
4fe71e88 | 727 | goto sync_out; |
973163fc | 728 | map->m_flags = F2FS_MAP_NEW; |
4fe71e88 | 729 | blkaddr = dn.data_blkaddr; |
973163fc | 730 | } else { |
43473f96 CY |
731 | if (flag == F2FS_GET_BLOCK_BMAP) { |
732 | map->m_pblk = 0; | |
733 | goto sync_out; | |
734 | } | |
da85985c CY |
735 | if (flag == F2FS_GET_BLOCK_FIEMAP && |
736 | blkaddr == NULL_ADDR) { | |
737 | if (map->m_next_pgofs) | |
738 | *map->m_next_pgofs = pgofs + 1; | |
739 | } | |
973163fc | 740 | if (flag != F2FS_GET_BLOCK_FIEMAP || |
43473f96 | 741 | blkaddr != NEW_ADDR) |
4fe71e88 | 742 | goto sync_out; |
e2b4e2bc | 743 | } |
e2b4e2bc | 744 | } |
eb47b800 | 745 | |
46008c6d CY |
746 | if (flag == F2FS_GET_BLOCK_PRE_AIO) |
747 | goto skip; | |
748 | ||
4fe71e88 CY |
749 | if (map->m_len == 0) { |
750 | /* preallocated unwritten block should be mapped for fiemap. */ | |
751 | if (blkaddr == NEW_ADDR) | |
752 | map->m_flags |= F2FS_MAP_UNWRITTEN; | |
753 | map->m_flags |= F2FS_MAP_MAPPED; | |
754 | ||
755 | map->m_pblk = blkaddr; | |
756 | map->m_len = 1; | |
757 | } else if ((map->m_pblk != NEW_ADDR && | |
758 | blkaddr == (map->m_pblk + ofs)) || | |
b439b103 | 759 | (map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR) || |
46008c6d | 760 | flag == F2FS_GET_BLOCK_PRE_DIO) { |
4fe71e88 CY |
761 | ofs++; |
762 | map->m_len++; | |
763 | } else { | |
764 | goto sync_out; | |
765 | } | |
bfad7c2d | 766 | |
46008c6d | 767 | skip: |
bfad7c2d JK |
768 | dn.ofs_in_node++; |
769 | pgofs++; | |
770 | ||
46008c6d CY |
771 | /* preallocate blocks in batch for one dnode page */ |
772 | if (flag == F2FS_GET_BLOCK_PRE_AIO && | |
773 | (pgofs == end || dn.ofs_in_node == end_offset)) { | |
7df3a431 | 774 | |
46008c6d CY |
775 | dn.ofs_in_node = ofs_in_node; |
776 | err = reserve_new_blocks(&dn, prealloc); | |
777 | if (err) | |
778 | goto sync_out; | |
bfad7c2d | 779 | |
46008c6d CY |
780 | map->m_len += dn.ofs_in_node - ofs_in_node; |
781 | if (prealloc && dn.ofs_in_node != last_ofs_in_node + 1) { | |
782 | err = -ENOSPC; | |
783 | goto sync_out; | |
3104af35 | 784 | } |
46008c6d CY |
785 | dn.ofs_in_node = end_offset; |
786 | } | |
787 | ||
788 | if (pgofs >= end) | |
789 | goto sync_out; | |
790 | else if (dn.ofs_in_node < end_offset) | |
791 | goto next_block; | |
792 | ||
793 | if (allocated) | |
794 | sync_inode_page(&dn); | |
795 | f2fs_put_dnode(&dn); | |
796 | ||
797 | if (create) { | |
798 | f2fs_unlock_op(sbi); | |
799 | f2fs_balance_fs(sbi, allocated); | |
eb47b800 | 800 | } |
46008c6d CY |
801 | allocated = false; |
802 | goto next_dnode; | |
7df3a431 | 803 | |
bfad7c2d JK |
804 | sync_out: |
805 | if (allocated) | |
806 | sync_inode_page(&dn); | |
eb47b800 | 807 | f2fs_put_dnode(&dn); |
bfad7c2d | 808 | unlock_out: |
2a340760 | 809 | if (create) { |
3104af35 | 810 | f2fs_unlock_op(sbi); |
3c082b7b | 811 | f2fs_balance_fs(sbi, allocated); |
2a340760 | 812 | } |
bfad7c2d | 813 | out: |
003a3e1d | 814 | trace_f2fs_map_blocks(inode, map, err); |
bfad7c2d | 815 | return err; |
eb47b800 JK |
816 | } |
817 | ||
003a3e1d | 818 | static int __get_data_block(struct inode *inode, sector_t iblock, |
da85985c CY |
819 | struct buffer_head *bh, int create, int flag, |
820 | pgoff_t *next_pgofs) | |
003a3e1d JK |
821 | { |
822 | struct f2fs_map_blocks map; | |
823 | int ret; | |
824 | ||
825 | map.m_lblk = iblock; | |
826 | map.m_len = bh->b_size >> inode->i_blkbits; | |
da85985c | 827 | map.m_next_pgofs = next_pgofs; |
003a3e1d | 828 | |
e2b4e2bc | 829 | ret = f2fs_map_blocks(inode, &map, create, flag); |
003a3e1d JK |
830 | if (!ret) { |
831 | map_bh(bh, inode->i_sb, map.m_pblk); | |
832 | bh->b_state = (bh->b_state & ~F2FS_MAP_FLAGS) | map.m_flags; | |
833 | bh->b_size = map.m_len << inode->i_blkbits; | |
834 | } | |
835 | return ret; | |
836 | } | |
837 | ||
ccfb3000 | 838 | static int get_data_block(struct inode *inode, sector_t iblock, |
da85985c CY |
839 | struct buffer_head *bh_result, int create, int flag, |
840 | pgoff_t *next_pgofs) | |
e2b4e2bc | 841 | { |
da85985c CY |
842 | return __get_data_block(inode, iblock, bh_result, create, |
843 | flag, next_pgofs); | |
e2b4e2bc CY |
844 | } |
845 | ||
846 | static int get_data_block_dio(struct inode *inode, sector_t iblock, | |
ccfb3000 JK |
847 | struct buffer_head *bh_result, int create) |
848 | { | |
e2b4e2bc | 849 | return __get_data_block(inode, iblock, bh_result, create, |
da85985c | 850 | F2FS_GET_BLOCK_DIO, NULL); |
ccfb3000 JK |
851 | } |
852 | ||
e2b4e2bc | 853 | static int get_data_block_bmap(struct inode *inode, sector_t iblock, |
ccfb3000 JK |
854 | struct buffer_head *bh_result, int create) |
855 | { | |
179448bf | 856 | /* Block number less than F2FS MAX BLOCKS */ |
e0afc4d6 | 857 | if (unlikely(iblock >= F2FS_I_SB(inode)->max_file_blocks)) |
179448bf YH |
858 | return -EFBIG; |
859 | ||
e2b4e2bc | 860 | return __get_data_block(inode, iblock, bh_result, create, |
da85985c | 861 | F2FS_GET_BLOCK_BMAP, NULL); |
ccfb3000 JK |
862 | } |
863 | ||
7f63eb77 JK |
864 | static inline sector_t logical_to_blk(struct inode *inode, loff_t offset) |
865 | { | |
866 | return (offset >> inode->i_blkbits); | |
867 | } | |
868 | ||
869 | static inline loff_t blk_to_logical(struct inode *inode, sector_t blk) | |
870 | { | |
871 | return (blk << inode->i_blkbits); | |
872 | } | |
873 | ||
9ab70134 JK |
874 | int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, |
875 | u64 start, u64 len) | |
876 | { | |
7f63eb77 JK |
877 | struct buffer_head map_bh; |
878 | sector_t start_blk, last_blk; | |
da85985c | 879 | pgoff_t next_pgofs; |
de1475cc | 880 | loff_t isize; |
7f63eb77 JK |
881 | u64 logical = 0, phys = 0, size = 0; |
882 | u32 flags = 0; | |
7f63eb77 JK |
883 | int ret = 0; |
884 | ||
885 | ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC); | |
886 | if (ret) | |
887 | return ret; | |
888 | ||
67f8cf3c JK |
889 | if (f2fs_has_inline_data(inode)) { |
890 | ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len); | |
891 | if (ret != -EAGAIN) | |
892 | return ret; | |
893 | } | |
894 | ||
5955102c | 895 | inode_lock(inode); |
de1475cc FL |
896 | |
897 | isize = i_size_read(inode); | |
9a950d52 FL |
898 | if (start >= isize) |
899 | goto out; | |
7f63eb77 | 900 | |
9a950d52 FL |
901 | if (start + len > isize) |
902 | len = isize - start; | |
7f63eb77 JK |
903 | |
904 | if (logical_to_blk(inode, len) == 0) | |
905 | len = blk_to_logical(inode, 1); | |
906 | ||
907 | start_blk = logical_to_blk(inode, start); | |
908 | last_blk = logical_to_blk(inode, start + len - 1); | |
9a950d52 | 909 | |
7f63eb77 JK |
910 | next: |
911 | memset(&map_bh, 0, sizeof(struct buffer_head)); | |
912 | map_bh.b_size = len; | |
913 | ||
e2b4e2bc | 914 | ret = get_data_block(inode, start_blk, &map_bh, 0, |
da85985c | 915 | F2FS_GET_BLOCK_FIEMAP, &next_pgofs); |
7f63eb77 JK |
916 | if (ret) |
917 | goto out; | |
918 | ||
919 | /* HOLE */ | |
920 | if (!buffer_mapped(&map_bh)) { | |
da85985c | 921 | start_blk = next_pgofs; |
9a950d52 | 922 | /* Go through holes util pass the EOF */ |
da85985c | 923 | if (blk_to_logical(inode, start_blk) < isize) |
9a950d52 FL |
924 | goto prep_next; |
925 | /* Found a hole beyond isize means no more extents. | |
926 | * Note that the premise is that filesystems don't | |
927 | * punch holes beyond isize and keep size unchanged. | |
928 | */ | |
929 | flags |= FIEMAP_EXTENT_LAST; | |
930 | } | |
7f63eb77 | 931 | |
da5af127 CY |
932 | if (size) { |
933 | if (f2fs_encrypted_inode(inode)) | |
934 | flags |= FIEMAP_EXTENT_DATA_ENCRYPTED; | |
935 | ||
9a950d52 FL |
936 | ret = fiemap_fill_next_extent(fieinfo, logical, |
937 | phys, size, flags); | |
da5af127 | 938 | } |
7f63eb77 | 939 | |
9a950d52 FL |
940 | if (start_blk > last_blk || ret) |
941 | goto out; | |
7f63eb77 | 942 | |
9a950d52 FL |
943 | logical = blk_to_logical(inode, start_blk); |
944 | phys = blk_to_logical(inode, map_bh.b_blocknr); | |
945 | size = map_bh.b_size; | |
946 | flags = 0; | |
947 | if (buffer_unwritten(&map_bh)) | |
948 | flags = FIEMAP_EXTENT_UNWRITTEN; | |
7f63eb77 | 949 | |
9a950d52 | 950 | start_blk += logical_to_blk(inode, size); |
7f63eb77 | 951 | |
9a950d52 | 952 | prep_next: |
7f63eb77 JK |
953 | cond_resched(); |
954 | if (fatal_signal_pending(current)) | |
955 | ret = -EINTR; | |
956 | else | |
957 | goto next; | |
958 | out: | |
959 | if (ret == 1) | |
960 | ret = 0; | |
961 | ||
5955102c | 962 | inode_unlock(inode); |
7f63eb77 | 963 | return ret; |
9ab70134 JK |
964 | } |
965 | ||
f1e88660 JK |
966 | /* |
967 | * This function was originally taken from fs/mpage.c, and customized for f2fs. | |
968 | * Major change was from block_size == page_size in f2fs by default. | |
969 | */ | |
970 | static int f2fs_mpage_readpages(struct address_space *mapping, | |
971 | struct list_head *pages, struct page *page, | |
972 | unsigned nr_pages) | |
973 | { | |
974 | struct bio *bio = NULL; | |
975 | unsigned page_idx; | |
976 | sector_t last_block_in_bio = 0; | |
977 | struct inode *inode = mapping->host; | |
978 | const unsigned blkbits = inode->i_blkbits; | |
979 | const unsigned blocksize = 1 << blkbits; | |
980 | sector_t block_in_file; | |
981 | sector_t last_block; | |
982 | sector_t last_block_in_file; | |
983 | sector_t block_nr; | |
984 | struct block_device *bdev = inode->i_sb->s_bdev; | |
985 | struct f2fs_map_blocks map; | |
986 | ||
987 | map.m_pblk = 0; | |
988 | map.m_lblk = 0; | |
989 | map.m_len = 0; | |
990 | map.m_flags = 0; | |
da85985c | 991 | map.m_next_pgofs = NULL; |
f1e88660 JK |
992 | |
993 | for (page_idx = 0; nr_pages; page_idx++, nr_pages--) { | |
994 | ||
995 | prefetchw(&page->flags); | |
996 | if (pages) { | |
997 | page = list_entry(pages->prev, struct page, lru); | |
998 | list_del(&page->lru); | |
999 | if (add_to_page_cache_lru(page, mapping, | |
1000 | page->index, GFP_KERNEL)) | |
1001 | goto next_page; | |
1002 | } | |
1003 | ||
1004 | block_in_file = (sector_t)page->index; | |
1005 | last_block = block_in_file + nr_pages; | |
1006 | last_block_in_file = (i_size_read(inode) + blocksize - 1) >> | |
1007 | blkbits; | |
1008 | if (last_block > last_block_in_file) | |
1009 | last_block = last_block_in_file; | |
1010 | ||
1011 | /* | |
1012 | * Map blocks using the previous result first. | |
1013 | */ | |
1014 | if ((map.m_flags & F2FS_MAP_MAPPED) && | |
1015 | block_in_file > map.m_lblk && | |
1016 | block_in_file < (map.m_lblk + map.m_len)) | |
1017 | goto got_it; | |
1018 | ||
1019 | /* | |
1020 | * Then do more f2fs_map_blocks() calls until we are | |
1021 | * done with this page. | |
1022 | */ | |
1023 | map.m_flags = 0; | |
1024 | ||
1025 | if (block_in_file < last_block) { | |
1026 | map.m_lblk = block_in_file; | |
1027 | map.m_len = last_block - block_in_file; | |
1028 | ||
46c9e141 | 1029 | if (f2fs_map_blocks(inode, &map, 0, |
da85985c | 1030 | F2FS_GET_BLOCK_READ)) |
f1e88660 JK |
1031 | goto set_error_page; |
1032 | } | |
1033 | got_it: | |
1034 | if ((map.m_flags & F2FS_MAP_MAPPED)) { | |
1035 | block_nr = map.m_pblk + block_in_file - map.m_lblk; | |
1036 | SetPageMappedToDisk(page); | |
1037 | ||
1038 | if (!PageUptodate(page) && !cleancache_get_page(page)) { | |
1039 | SetPageUptodate(page); | |
1040 | goto confused; | |
1041 | } | |
1042 | } else { | |
09cbfeaf | 1043 | zero_user_segment(page, 0, PAGE_SIZE); |
f1e88660 JK |
1044 | SetPageUptodate(page); |
1045 | unlock_page(page); | |
1046 | goto next_page; | |
1047 | } | |
1048 | ||
1049 | /* | |
1050 | * This page will go to BIO. Do we need to send this | |
1051 | * BIO off first? | |
1052 | */ | |
1053 | if (bio && (last_block_in_bio != block_nr - 1)) { | |
1054 | submit_and_realloc: | |
f5730184 | 1055 | __submit_bio(F2FS_I_SB(inode), READ, bio); |
f1e88660 JK |
1056 | bio = NULL; |
1057 | } | |
1058 | if (bio == NULL) { | |
0b81d077 | 1059 | struct fscrypt_ctx *ctx = NULL; |
4375a336 JK |
1060 | |
1061 | if (f2fs_encrypted_inode(inode) && | |
1062 | S_ISREG(inode->i_mode)) { | |
4375a336 | 1063 | |
b32e4482 | 1064 | ctx = fscrypt_get_ctx(inode, GFP_NOFS); |
4375a336 JK |
1065 | if (IS_ERR(ctx)) |
1066 | goto set_error_page; | |
1067 | ||
1068 | /* wait the page to be moved by cleaning */ | |
08b39fbd CY |
1069 | f2fs_wait_on_encrypted_page_writeback( |
1070 | F2FS_I_SB(inode), block_nr); | |
4375a336 JK |
1071 | } |
1072 | ||
f1e88660 | 1073 | bio = bio_alloc(GFP_KERNEL, |
b54ffb73 | 1074 | min_t(int, nr_pages, BIO_MAX_PAGES)); |
4375a336 JK |
1075 | if (!bio) { |
1076 | if (ctx) | |
0b81d077 | 1077 | fscrypt_release_ctx(ctx); |
f1e88660 | 1078 | goto set_error_page; |
4375a336 | 1079 | } |
f1e88660 JK |
1080 | bio->bi_bdev = bdev; |
1081 | bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(block_nr); | |
12377024 | 1082 | bio->bi_end_io = f2fs_read_end_io; |
4375a336 | 1083 | bio->bi_private = ctx; |
4e49ea4a | 1084 | bio->bi_rw = READ; |
f1e88660 JK |
1085 | } |
1086 | ||
1087 | if (bio_add_page(bio, page, blocksize, 0) < blocksize) | |
1088 | goto submit_and_realloc; | |
1089 | ||
1090 | last_block_in_bio = block_nr; | |
1091 | goto next_page; | |
1092 | set_error_page: | |
1093 | SetPageError(page); | |
09cbfeaf | 1094 | zero_user_segment(page, 0, PAGE_SIZE); |
f1e88660 JK |
1095 | unlock_page(page); |
1096 | goto next_page; | |
1097 | confused: | |
1098 | if (bio) { | |
f5730184 | 1099 | __submit_bio(F2FS_I_SB(inode), READ, bio); |
f1e88660 JK |
1100 | bio = NULL; |
1101 | } | |
1102 | unlock_page(page); | |
1103 | next_page: | |
1104 | if (pages) | |
09cbfeaf | 1105 | put_page(page); |
f1e88660 JK |
1106 | } |
1107 | BUG_ON(pages && !list_empty(pages)); | |
1108 | if (bio) | |
f5730184 | 1109 | __submit_bio(F2FS_I_SB(inode), READ, bio); |
f1e88660 JK |
1110 | return 0; |
1111 | } | |
1112 | ||
eb47b800 JK |
1113 | static int f2fs_read_data_page(struct file *file, struct page *page) |
1114 | { | |
9ffe0fb5 | 1115 | struct inode *inode = page->mapping->host; |
b3d208f9 | 1116 | int ret = -EAGAIN; |
9ffe0fb5 | 1117 | |
c20e89cd CY |
1118 | trace_f2fs_readpage(page, DATA); |
1119 | ||
e1c42045 | 1120 | /* If the file has inline data, try to read it directly */ |
9ffe0fb5 HL |
1121 | if (f2fs_has_inline_data(inode)) |
1122 | ret = f2fs_read_inline_data(inode, page); | |
b3d208f9 | 1123 | if (ret == -EAGAIN) |
f1e88660 | 1124 | ret = f2fs_mpage_readpages(page->mapping, NULL, page, 1); |
9ffe0fb5 | 1125 | return ret; |
eb47b800 JK |
1126 | } |
1127 | ||
1128 | static int f2fs_read_data_pages(struct file *file, | |
1129 | struct address_space *mapping, | |
1130 | struct list_head *pages, unsigned nr_pages) | |
1131 | { | |
9ffe0fb5 | 1132 | struct inode *inode = file->f_mapping->host; |
b8c29400 CY |
1133 | struct page *page = list_entry(pages->prev, struct page, lru); |
1134 | ||
1135 | trace_f2fs_readpages(inode, page, nr_pages); | |
9ffe0fb5 HL |
1136 | |
1137 | /* If the file has inline data, skip readpages */ | |
1138 | if (f2fs_has_inline_data(inode)) | |
1139 | return 0; | |
1140 | ||
f1e88660 | 1141 | return f2fs_mpage_readpages(mapping, pages, NULL, nr_pages); |
eb47b800 JK |
1142 | } |
1143 | ||
05ca3632 | 1144 | int do_write_data_page(struct f2fs_io_info *fio) |
eb47b800 | 1145 | { |
05ca3632 | 1146 | struct page *page = fio->page; |
eb47b800 | 1147 | struct inode *inode = page->mapping->host; |
eb47b800 JK |
1148 | struct dnode_of_data dn; |
1149 | int err = 0; | |
1150 | ||
1151 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
266e97a8 | 1152 | err = get_dnode_of_data(&dn, page->index, LOOKUP_NODE); |
eb47b800 JK |
1153 | if (err) |
1154 | return err; | |
1155 | ||
28bc106b | 1156 | fio->old_blkaddr = dn.data_blkaddr; |
eb47b800 JK |
1157 | |
1158 | /* This page is already truncated */ | |
7a9d7548 | 1159 | if (fio->old_blkaddr == NULL_ADDR) { |
2bca1e23 | 1160 | ClearPageUptodate(page); |
eb47b800 | 1161 | goto out_writepage; |
2bca1e23 | 1162 | } |
eb47b800 | 1163 | |
4375a336 | 1164 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) { |
b32e4482 | 1165 | gfp_t gfp_flags = GFP_NOFS; |
08b39fbd CY |
1166 | |
1167 | /* wait for GCed encrypted page writeback */ | |
1168 | f2fs_wait_on_encrypted_page_writeback(F2FS_I_SB(inode), | |
7a9d7548 | 1169 | fio->old_blkaddr); |
b32e4482 JK |
1170 | retry_encrypt: |
1171 | fio->encrypted_page = fscrypt_encrypt_page(inode, fio->page, | |
1172 | gfp_flags); | |
4375a336 JK |
1173 | if (IS_ERR(fio->encrypted_page)) { |
1174 | err = PTR_ERR(fio->encrypted_page); | |
b32e4482 JK |
1175 | if (err == -ENOMEM) { |
1176 | /* flush pending ios and wait for a while */ | |
1177 | f2fs_flush_merged_bios(F2FS_I_SB(inode)); | |
1178 | congestion_wait(BLK_RW_ASYNC, HZ/50); | |
1179 | gfp_flags |= __GFP_NOFAIL; | |
1180 | err = 0; | |
1181 | goto retry_encrypt; | |
1182 | } | |
4375a336 JK |
1183 | goto out_writepage; |
1184 | } | |
1185 | } | |
1186 | ||
eb47b800 JK |
1187 | set_page_writeback(page); |
1188 | ||
1189 | /* | |
1190 | * If current allocation needs SSR, | |
1191 | * it had better in-place writes for updated data. | |
1192 | */ | |
7a9d7548 | 1193 | if (unlikely(fio->old_blkaddr != NEW_ADDR && |
b25958b6 | 1194 | !is_cold_data(page) && |
2da3e027 | 1195 | !IS_ATOMIC_WRITTEN_PAGE(page) && |
b25958b6 | 1196 | need_inplace_update(inode))) { |
05ca3632 | 1197 | rewrite_data_page(fio); |
fff04f90 | 1198 | set_inode_flag(F2FS_I(inode), FI_UPDATE_WRITE); |
8ce67cb0 | 1199 | trace_f2fs_do_write_data_page(page, IPU); |
eb47b800 | 1200 | } else { |
05ca3632 | 1201 | write_data_page(&dn, fio); |
8ce67cb0 | 1202 | trace_f2fs_do_write_data_page(page, OPU); |
fff04f90 | 1203 | set_inode_flag(F2FS_I(inode), FI_APPEND_WRITE); |
3c6c2beb JK |
1204 | if (page->index == 0) |
1205 | set_inode_flag(F2FS_I(inode), FI_FIRST_BLOCK_WRITTEN); | |
eb47b800 JK |
1206 | } |
1207 | out_writepage: | |
1208 | f2fs_put_dnode(&dn); | |
1209 | return err; | |
1210 | } | |
1211 | ||
1212 | static int f2fs_write_data_page(struct page *page, | |
1213 | struct writeback_control *wbc) | |
1214 | { | |
1215 | struct inode *inode = page->mapping->host; | |
4081363f | 1216 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
eb47b800 JK |
1217 | loff_t i_size = i_size_read(inode); |
1218 | const pgoff_t end_index = ((unsigned long long) i_size) | |
09cbfeaf | 1219 | >> PAGE_SHIFT; |
9ffe0fb5 | 1220 | unsigned offset = 0; |
39936837 | 1221 | bool need_balance_fs = false; |
eb47b800 | 1222 | int err = 0; |
458e6197 | 1223 | struct f2fs_io_info fio = { |
05ca3632 | 1224 | .sbi = sbi, |
458e6197 | 1225 | .type = DATA, |
6c311ec6 | 1226 | .rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE, |
05ca3632 | 1227 | .page = page, |
4375a336 | 1228 | .encrypted_page = NULL, |
458e6197 | 1229 | }; |
eb47b800 | 1230 | |
ecda0de3 CY |
1231 | trace_f2fs_writepage(page, DATA); |
1232 | ||
eb47b800 | 1233 | if (page->index < end_index) |
39936837 | 1234 | goto write; |
eb47b800 JK |
1235 | |
1236 | /* | |
1237 | * If the offset is out-of-range of file size, | |
1238 | * this page does not have to be written to disk. | |
1239 | */ | |
09cbfeaf | 1240 | offset = i_size & (PAGE_SIZE - 1); |
76f60268 | 1241 | if ((page->index >= end_index + 1) || !offset) |
39936837 | 1242 | goto out; |
eb47b800 | 1243 | |
09cbfeaf | 1244 | zero_user_segment(page, offset, PAGE_SIZE); |
39936837 | 1245 | write: |
caf0047e | 1246 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
eb47b800 | 1247 | goto redirty_out; |
1e84371f JK |
1248 | if (f2fs_is_drop_cache(inode)) |
1249 | goto out; | |
e6e5f561 JK |
1250 | /* we should not write 0'th page having journal header */ |
1251 | if (f2fs_is_volatile_file(inode) && (!page->index || | |
1252 | (!wbc->for_reclaim && | |
1253 | available_free_memory(sbi, BASE_CHECK)))) | |
1e84371f | 1254 | goto redirty_out; |
eb47b800 | 1255 | |
39936837 | 1256 | /* Dentry blocks are controlled by checkpoint */ |
eb47b800 | 1257 | if (S_ISDIR(inode->i_mode)) { |
cf779cab JK |
1258 | if (unlikely(f2fs_cp_error(sbi))) |
1259 | goto redirty_out; | |
05ca3632 | 1260 | err = do_write_data_page(&fio); |
8618b881 JK |
1261 | goto done; |
1262 | } | |
9ffe0fb5 | 1263 | |
cf779cab JK |
1264 | /* we should bypass data pages to proceed the kworkder jobs */ |
1265 | if (unlikely(f2fs_cp_error(sbi))) { | |
1266 | SetPageError(page); | |
a7ffdbe2 | 1267 | goto out; |
cf779cab JK |
1268 | } |
1269 | ||
8618b881 | 1270 | if (!wbc->for_reclaim) |
39936837 | 1271 | need_balance_fs = true; |
8618b881 | 1272 | else if (has_not_enough_free_secs(sbi, 0)) |
39936837 | 1273 | goto redirty_out; |
eb47b800 | 1274 | |
b3d208f9 | 1275 | err = -EAGAIN; |
8618b881 | 1276 | f2fs_lock_op(sbi); |
b3d208f9 JK |
1277 | if (f2fs_has_inline_data(inode)) |
1278 | err = f2fs_write_inline_data(inode, page); | |
1279 | if (err == -EAGAIN) | |
05ca3632 | 1280 | err = do_write_data_page(&fio); |
8618b881 JK |
1281 | f2fs_unlock_op(sbi); |
1282 | done: | |
1283 | if (err && err != -ENOENT) | |
1284 | goto redirty_out; | |
eb47b800 | 1285 | |
eb47b800 | 1286 | clear_cold_data(page); |
39936837 | 1287 | out: |
a7ffdbe2 | 1288 | inode_dec_dirty_pages(inode); |
2bca1e23 JK |
1289 | if (err) |
1290 | ClearPageUptodate(page); | |
0c3a5797 CY |
1291 | |
1292 | if (wbc->for_reclaim) { | |
1293 | f2fs_submit_merged_bio_cond(sbi, NULL, page, 0, DATA, WRITE); | |
1294 | remove_dirty_inode(inode); | |
1295 | } | |
1296 | ||
eb47b800 | 1297 | unlock_page(page); |
2c4db1a6 | 1298 | f2fs_balance_fs(sbi, need_balance_fs); |
0c3a5797 CY |
1299 | |
1300 | if (unlikely(f2fs_cp_error(sbi))) | |
2aea39ec | 1301 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
0c3a5797 | 1302 | |
eb47b800 JK |
1303 | return 0; |
1304 | ||
eb47b800 | 1305 | redirty_out: |
76f60268 | 1306 | redirty_page_for_writepage(wbc, page); |
8618b881 | 1307 | return AOP_WRITEPAGE_ACTIVATE; |
eb47b800 JK |
1308 | } |
1309 | ||
fa9150a8 NJ |
1310 | static int __f2fs_writepage(struct page *page, struct writeback_control *wbc, |
1311 | void *data) | |
1312 | { | |
1313 | struct address_space *mapping = data; | |
1314 | int ret = mapping->a_ops->writepage(page, wbc); | |
1315 | mapping_set_error(mapping, ret); | |
1316 | return ret; | |
1317 | } | |
1318 | ||
8f46dcae CY |
1319 | /* |
1320 | * This function was copied from write_cche_pages from mm/page-writeback.c. | |
1321 | * The major change is making write step of cold data page separately from | |
1322 | * warm/hot data page. | |
1323 | */ | |
1324 | static int f2fs_write_cache_pages(struct address_space *mapping, | |
1325 | struct writeback_control *wbc, writepage_t writepage, | |
1326 | void *data) | |
1327 | { | |
1328 | int ret = 0; | |
1329 | int done = 0; | |
1330 | struct pagevec pvec; | |
1331 | int nr_pages; | |
1332 | pgoff_t uninitialized_var(writeback_index); | |
1333 | pgoff_t index; | |
1334 | pgoff_t end; /* Inclusive */ | |
1335 | pgoff_t done_index; | |
1336 | int cycled; | |
1337 | int range_whole = 0; | |
1338 | int tag; | |
1339 | int step = 0; | |
1340 | ||
1341 | pagevec_init(&pvec, 0); | |
1342 | next: | |
1343 | if (wbc->range_cyclic) { | |
1344 | writeback_index = mapping->writeback_index; /* prev offset */ | |
1345 | index = writeback_index; | |
1346 | if (index == 0) | |
1347 | cycled = 1; | |
1348 | else | |
1349 | cycled = 0; | |
1350 | end = -1; | |
1351 | } else { | |
09cbfeaf KS |
1352 | index = wbc->range_start >> PAGE_SHIFT; |
1353 | end = wbc->range_end >> PAGE_SHIFT; | |
8f46dcae CY |
1354 | if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX) |
1355 | range_whole = 1; | |
1356 | cycled = 1; /* ignore range_cyclic tests */ | |
1357 | } | |
1358 | if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages) | |
1359 | tag = PAGECACHE_TAG_TOWRITE; | |
1360 | else | |
1361 | tag = PAGECACHE_TAG_DIRTY; | |
1362 | retry: | |
1363 | if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages) | |
1364 | tag_pages_for_writeback(mapping, index, end); | |
1365 | done_index = index; | |
1366 | while (!done && (index <= end)) { | |
1367 | int i; | |
1368 | ||
1369 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag, | |
1370 | min(end - index, (pgoff_t)PAGEVEC_SIZE - 1) + 1); | |
1371 | if (nr_pages == 0) | |
1372 | break; | |
1373 | ||
1374 | for (i = 0; i < nr_pages; i++) { | |
1375 | struct page *page = pvec.pages[i]; | |
1376 | ||
1377 | if (page->index > end) { | |
1378 | done = 1; | |
1379 | break; | |
1380 | } | |
1381 | ||
1382 | done_index = page->index; | |
1383 | ||
1384 | lock_page(page); | |
1385 | ||
1386 | if (unlikely(page->mapping != mapping)) { | |
1387 | continue_unlock: | |
1388 | unlock_page(page); | |
1389 | continue; | |
1390 | } | |
1391 | ||
1392 | if (!PageDirty(page)) { | |
1393 | /* someone wrote it for us */ | |
1394 | goto continue_unlock; | |
1395 | } | |
1396 | ||
737f1899 | 1397 | if (step == is_cold_data(page)) |
8f46dcae CY |
1398 | goto continue_unlock; |
1399 | ||
1400 | if (PageWriteback(page)) { | |
1401 | if (wbc->sync_mode != WB_SYNC_NONE) | |
fec1d657 JK |
1402 | f2fs_wait_on_page_writeback(page, |
1403 | DATA, true); | |
8f46dcae CY |
1404 | else |
1405 | goto continue_unlock; | |
1406 | } | |
1407 | ||
1408 | BUG_ON(PageWriteback(page)); | |
1409 | if (!clear_page_dirty_for_io(page)) | |
1410 | goto continue_unlock; | |
1411 | ||
1412 | ret = (*writepage)(page, wbc, data); | |
1413 | if (unlikely(ret)) { | |
1414 | if (ret == AOP_WRITEPAGE_ACTIVATE) { | |
1415 | unlock_page(page); | |
1416 | ret = 0; | |
1417 | } else { | |
1418 | done_index = page->index + 1; | |
1419 | done = 1; | |
1420 | break; | |
1421 | } | |
1422 | } | |
1423 | ||
1424 | if (--wbc->nr_to_write <= 0 && | |
1425 | wbc->sync_mode == WB_SYNC_NONE) { | |
1426 | done = 1; | |
1427 | break; | |
1428 | } | |
1429 | } | |
1430 | pagevec_release(&pvec); | |
1431 | cond_resched(); | |
1432 | } | |
1433 | ||
1434 | if (step < 1) { | |
1435 | step++; | |
1436 | goto next; | |
1437 | } | |
1438 | ||
1439 | if (!cycled && !done) { | |
1440 | cycled = 1; | |
1441 | index = 0; | |
1442 | end = writeback_index - 1; | |
1443 | goto retry; | |
1444 | } | |
1445 | if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) | |
1446 | mapping->writeback_index = done_index; | |
1447 | ||
1448 | return ret; | |
1449 | } | |
1450 | ||
25ca923b | 1451 | static int f2fs_write_data_pages(struct address_space *mapping, |
eb47b800 JK |
1452 | struct writeback_control *wbc) |
1453 | { | |
1454 | struct inode *inode = mapping->host; | |
4081363f | 1455 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
5463e7c1 | 1456 | bool locked = false; |
eb47b800 | 1457 | int ret; |
50c8cdb3 | 1458 | long diff; |
eb47b800 | 1459 | |
cfb185a1 | 1460 | /* deal with chardevs and other special file */ |
1461 | if (!mapping->a_ops->writepage) | |
1462 | return 0; | |
1463 | ||
6a290544 CY |
1464 | /* skip writing if there is no dirty page in this inode */ |
1465 | if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE) | |
1466 | return 0; | |
1467 | ||
a1257023 JK |
1468 | if (S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_NONE && |
1469 | get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) && | |
1470 | available_free_memory(sbi, DIRTY_DENTS)) | |
1471 | goto skip_write; | |
1472 | ||
d323d005 CY |
1473 | /* skip writing during file defragment */ |
1474 | if (is_inode_flag_set(F2FS_I(inode), FI_DO_DEFRAG)) | |
1475 | goto skip_write; | |
1476 | ||
d5669f7b JK |
1477 | /* during POR, we don't need to trigger writepage at all. */ |
1478 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) | |
1479 | goto skip_write; | |
1480 | ||
d31c7c3f YH |
1481 | trace_f2fs_writepages(mapping->host, wbc, DATA); |
1482 | ||
50c8cdb3 | 1483 | diff = nr_pages_to_write(sbi, DATA, wbc); |
eb47b800 | 1484 | |
25c13551 | 1485 | if (!S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_ALL) { |
5463e7c1 JK |
1486 | mutex_lock(&sbi->writepages); |
1487 | locked = true; | |
1488 | } | |
8f46dcae | 1489 | ret = f2fs_write_cache_pages(mapping, wbc, __f2fs_writepage, mapping); |
0c3a5797 | 1490 | f2fs_submit_merged_bio_cond(sbi, inode, NULL, 0, DATA, WRITE); |
5463e7c1 JK |
1491 | if (locked) |
1492 | mutex_unlock(&sbi->writepages); | |
458e6197 | 1493 | |
c227f912 | 1494 | remove_dirty_inode(inode); |
eb47b800 | 1495 | |
50c8cdb3 | 1496 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff); |
eb47b800 | 1497 | return ret; |
d3baf95d JK |
1498 | |
1499 | skip_write: | |
a7ffdbe2 | 1500 | wbc->pages_skipped += get_dirty_pages(inode); |
d31c7c3f | 1501 | trace_f2fs_writepages(mapping->host, wbc, DATA); |
d3baf95d | 1502 | return 0; |
eb47b800 JK |
1503 | } |
1504 | ||
3aab8f82 CY |
1505 | static void f2fs_write_failed(struct address_space *mapping, loff_t to) |
1506 | { | |
1507 | struct inode *inode = mapping->host; | |
819d9153 | 1508 | loff_t i_size = i_size_read(inode); |
3aab8f82 | 1509 | |
819d9153 JK |
1510 | if (to > i_size) { |
1511 | truncate_pagecache(inode, i_size); | |
1512 | truncate_blocks(inode, i_size, true); | |
3aab8f82 CY |
1513 | } |
1514 | } | |
1515 | ||
2aadac08 JK |
1516 | static int prepare_write_begin(struct f2fs_sb_info *sbi, |
1517 | struct page *page, loff_t pos, unsigned len, | |
1518 | block_t *blk_addr, bool *node_changed) | |
1519 | { | |
1520 | struct inode *inode = page->mapping->host; | |
1521 | pgoff_t index = page->index; | |
1522 | struct dnode_of_data dn; | |
1523 | struct page *ipage; | |
b4d07a3e JK |
1524 | bool locked = false; |
1525 | struct extent_info ei; | |
2aadac08 JK |
1526 | int err = 0; |
1527 | ||
24b84912 JK |
1528 | /* |
1529 | * we already allocated all the blocks, so we don't need to get | |
1530 | * the block addresses when there is no need to fill the page. | |
1531 | */ | |
1532 | if (!f2fs_has_inline_data(inode) && !f2fs_encrypted_inode(inode) && | |
09cbfeaf | 1533 | len == PAGE_SIZE) |
24b84912 JK |
1534 | return 0; |
1535 | ||
b4d07a3e | 1536 | if (f2fs_has_inline_data(inode) || |
09cbfeaf | 1537 | (pos & PAGE_MASK) >= i_size_read(inode)) { |
b4d07a3e JK |
1538 | f2fs_lock_op(sbi); |
1539 | locked = true; | |
1540 | } | |
1541 | restart: | |
2aadac08 JK |
1542 | /* check inline_data */ |
1543 | ipage = get_node_page(sbi, inode->i_ino); | |
1544 | if (IS_ERR(ipage)) { | |
1545 | err = PTR_ERR(ipage); | |
1546 | goto unlock_out; | |
1547 | } | |
1548 | ||
1549 | set_new_dnode(&dn, inode, ipage, ipage, 0); | |
1550 | ||
1551 | if (f2fs_has_inline_data(inode)) { | |
1552 | if (pos + len <= MAX_INLINE_DATA) { | |
1553 | read_inline_data(page, ipage); | |
1554 | set_inode_flag(F2FS_I(inode), FI_DATA_EXIST); | |
ab47036d CY |
1555 | if (inode->i_nlink) |
1556 | set_inline_node(ipage); | |
2aadac08 JK |
1557 | } else { |
1558 | err = f2fs_convert_inline_page(&dn, page); | |
1559 | if (err) | |
b4d07a3e JK |
1560 | goto out; |
1561 | if (dn.data_blkaddr == NULL_ADDR) | |
1562 | err = f2fs_get_block(&dn, index); | |
1563 | } | |
1564 | } else if (locked) { | |
1565 | err = f2fs_get_block(&dn, index); | |
1566 | } else { | |
1567 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { | |
1568 | dn.data_blkaddr = ei.blk + index - ei.fofs; | |
1569 | } else { | |
b4d07a3e JK |
1570 | /* hole case */ |
1571 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); | |
4da7bf5a | 1572 | if (err || dn.data_blkaddr == NULL_ADDR) { |
b4d07a3e JK |
1573 | f2fs_put_dnode(&dn); |
1574 | f2fs_lock_op(sbi); | |
1575 | locked = true; | |
1576 | goto restart; | |
1577 | } | |
2aadac08 JK |
1578 | } |
1579 | } | |
b4d07a3e | 1580 | |
2aadac08 JK |
1581 | /* convert_inline_page can make node_changed */ |
1582 | *blk_addr = dn.data_blkaddr; | |
1583 | *node_changed = dn.node_changed; | |
b4d07a3e | 1584 | out: |
2aadac08 JK |
1585 | f2fs_put_dnode(&dn); |
1586 | unlock_out: | |
b4d07a3e JK |
1587 | if (locked) |
1588 | f2fs_unlock_op(sbi); | |
2aadac08 JK |
1589 | return err; |
1590 | } | |
1591 | ||
eb47b800 JK |
1592 | static int f2fs_write_begin(struct file *file, struct address_space *mapping, |
1593 | loff_t pos, unsigned len, unsigned flags, | |
1594 | struct page **pagep, void **fsdata) | |
1595 | { | |
1596 | struct inode *inode = mapping->host; | |
4081363f | 1597 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
86531d6b | 1598 | struct page *page = NULL; |
09cbfeaf | 1599 | pgoff_t index = ((unsigned long long) pos) >> PAGE_SHIFT; |
2aadac08 JK |
1600 | bool need_balance = false; |
1601 | block_t blkaddr = NULL_ADDR; | |
eb47b800 JK |
1602 | int err = 0; |
1603 | ||
62aed044 CY |
1604 | trace_f2fs_write_begin(inode, pos, len, flags); |
1605 | ||
5f727395 JK |
1606 | /* |
1607 | * We should check this at this moment to avoid deadlock on inode page | |
1608 | * and #0 page. The locking rule for inline_data conversion should be: | |
1609 | * lock_page(page #0) -> lock_page(inode_page) | |
1610 | */ | |
1611 | if (index != 0) { | |
1612 | err = f2fs_convert_inline_inode(inode); | |
1613 | if (err) | |
1614 | goto fail; | |
1615 | } | |
afcb7ca0 | 1616 | repeat: |
eb47b800 | 1617 | page = grab_cache_page_write_begin(mapping, index, flags); |
3aab8f82 CY |
1618 | if (!page) { |
1619 | err = -ENOMEM; | |
1620 | goto fail; | |
1621 | } | |
d5f66990 | 1622 | |
eb47b800 JK |
1623 | *pagep = page; |
1624 | ||
2aadac08 JK |
1625 | err = prepare_write_begin(sbi, page, pos, len, |
1626 | &blkaddr, &need_balance); | |
9ba69cf9 | 1627 | if (err) |
2aadac08 | 1628 | goto fail; |
9ba69cf9 | 1629 | |
2aadac08 | 1630 | if (need_balance && has_not_enough_free_secs(sbi, 0)) { |
2a340760 | 1631 | unlock_page(page); |
2c4db1a6 | 1632 | f2fs_balance_fs(sbi, true); |
2a340760 JK |
1633 | lock_page(page); |
1634 | if (page->mapping != mapping) { | |
1635 | /* The page got truncated from under us */ | |
1636 | f2fs_put_page(page, 1); | |
1637 | goto repeat; | |
1638 | } | |
1639 | } | |
1640 | ||
fec1d657 | 1641 | f2fs_wait_on_page_writeback(page, DATA, false); |
b3d208f9 | 1642 | |
08b39fbd CY |
1643 | /* wait for GCed encrypted page writeback */ |
1644 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) | |
2aadac08 | 1645 | f2fs_wait_on_encrypted_page_writeback(sbi, blkaddr); |
08b39fbd | 1646 | |
09cbfeaf | 1647 | if (len == PAGE_SIZE) |
90d4388a CY |
1648 | goto out_update; |
1649 | if (PageUptodate(page)) | |
1650 | goto out_clear; | |
eb47b800 | 1651 | |
09cbfeaf KS |
1652 | if ((pos & PAGE_MASK) >= i_size_read(inode)) { |
1653 | unsigned start = pos & (PAGE_SIZE - 1); | |
eb47b800 JK |
1654 | unsigned end = start + len; |
1655 | ||
1656 | /* Reading beyond i_size is simple: memset to zero */ | |
09cbfeaf | 1657 | zero_user_segments(page, 0, start, end, PAGE_SIZE); |
90d4388a | 1658 | goto out_update; |
eb47b800 JK |
1659 | } |
1660 | ||
2aadac08 | 1661 | if (blkaddr == NEW_ADDR) { |
09cbfeaf | 1662 | zero_user_segment(page, 0, PAGE_SIZE); |
eb47b800 | 1663 | } else { |
cf04e8eb | 1664 | struct f2fs_io_info fio = { |
05ca3632 | 1665 | .sbi = sbi, |
cf04e8eb JK |
1666 | .type = DATA, |
1667 | .rw = READ_SYNC, | |
7a9d7548 CY |
1668 | .old_blkaddr = blkaddr, |
1669 | .new_blkaddr = blkaddr, | |
05ca3632 | 1670 | .page = page, |
4375a336 | 1671 | .encrypted_page = NULL, |
cf04e8eb | 1672 | }; |
05ca3632 | 1673 | err = f2fs_submit_page_bio(&fio); |
9234f319 JK |
1674 | if (err) |
1675 | goto fail; | |
d54c795b | 1676 | |
393ff91f | 1677 | lock_page(page); |
6bacf52f | 1678 | if (unlikely(!PageUptodate(page))) { |
3aab8f82 CY |
1679 | err = -EIO; |
1680 | goto fail; | |
eb47b800 | 1681 | } |
6bacf52f | 1682 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
1683 | f2fs_put_page(page, 1); |
1684 | goto repeat; | |
eb47b800 | 1685 | } |
4375a336 JK |
1686 | |
1687 | /* avoid symlink page */ | |
1688 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) { | |
0b81d077 | 1689 | err = fscrypt_decrypt_page(page); |
86531d6b | 1690 | if (err) |
4375a336 | 1691 | goto fail; |
4375a336 | 1692 | } |
eb47b800 | 1693 | } |
90d4388a | 1694 | out_update: |
eb47b800 | 1695 | SetPageUptodate(page); |
90d4388a | 1696 | out_clear: |
eb47b800 JK |
1697 | clear_cold_data(page); |
1698 | return 0; | |
9ba69cf9 | 1699 | |
3aab8f82 | 1700 | fail: |
86531d6b | 1701 | f2fs_put_page(page, 1); |
3aab8f82 CY |
1702 | f2fs_write_failed(mapping, pos + len); |
1703 | return err; | |
eb47b800 JK |
1704 | } |
1705 | ||
a1dd3c13 JK |
1706 | static int f2fs_write_end(struct file *file, |
1707 | struct address_space *mapping, | |
1708 | loff_t pos, unsigned len, unsigned copied, | |
1709 | struct page *page, void *fsdata) | |
1710 | { | |
1711 | struct inode *inode = page->mapping->host; | |
1712 | ||
dfb2bf38 CY |
1713 | trace_f2fs_write_end(inode, pos, len, copied); |
1714 | ||
34ba94ba | 1715 | set_page_dirty(page); |
a1dd3c13 JK |
1716 | |
1717 | if (pos + copied > i_size_read(inode)) { | |
1718 | i_size_write(inode, pos + copied); | |
1719 | mark_inode_dirty(inode); | |
a1dd3c13 JK |
1720 | } |
1721 | ||
75c3c8bc | 1722 | f2fs_put_page(page, 1); |
d0239e1b | 1723 | f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); |
a1dd3c13 JK |
1724 | return copied; |
1725 | } | |
1726 | ||
6f673763 OS |
1727 | static int check_direct_IO(struct inode *inode, struct iov_iter *iter, |
1728 | loff_t offset) | |
944fcfc1 JK |
1729 | { |
1730 | unsigned blocksize_mask = inode->i_sb->s_blocksize - 1; | |
944fcfc1 | 1731 | |
944fcfc1 JK |
1732 | if (offset & blocksize_mask) |
1733 | return -EINVAL; | |
1734 | ||
5b46f25d AV |
1735 | if (iov_iter_alignment(iter) & blocksize_mask) |
1736 | return -EINVAL; | |
1737 | ||
944fcfc1 JK |
1738 | return 0; |
1739 | } | |
1740 | ||
c8b8e32d | 1741 | static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter) |
eb47b800 | 1742 | { |
b439b103 | 1743 | struct address_space *mapping = iocb->ki_filp->f_mapping; |
3aab8f82 CY |
1744 | struct inode *inode = mapping->host; |
1745 | size_t count = iov_iter_count(iter); | |
c8b8e32d | 1746 | loff_t offset = iocb->ki_pos; |
3aab8f82 | 1747 | int err; |
944fcfc1 | 1748 | |
b439b103 | 1749 | err = check_direct_IO(inode, iter, offset); |
b9d777b8 JK |
1750 | if (err) |
1751 | return err; | |
9ffe0fb5 | 1752 | |
fcc85a4d JK |
1753 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) |
1754 | return 0; | |
1755 | ||
6f673763 | 1756 | trace_f2fs_direct_IO_enter(inode, offset, count, iov_iter_rw(iter)); |
70407fad | 1757 | |
c8b8e32d | 1758 | err = blockdev_direct_IO(iocb, inode, iter, get_data_block_dio); |
6bfc4919 JK |
1759 | if (iov_iter_rw(iter) == WRITE) { |
1760 | if (err > 0) | |
1761 | set_inode_flag(F2FS_I(inode), FI_UPDATE_WRITE); | |
1762 | else if (err < 0) | |
1763 | f2fs_write_failed(mapping, offset + count); | |
1764 | } | |
70407fad | 1765 | |
6f673763 | 1766 | trace_f2fs_direct_IO_exit(inode, offset, count, iov_iter_rw(iter), err); |
70407fad | 1767 | |
3aab8f82 | 1768 | return err; |
eb47b800 JK |
1769 | } |
1770 | ||
487261f3 CY |
1771 | void f2fs_invalidate_page(struct page *page, unsigned int offset, |
1772 | unsigned int length) | |
eb47b800 JK |
1773 | { |
1774 | struct inode *inode = page->mapping->host; | |
487261f3 | 1775 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
a7ffdbe2 | 1776 | |
487261f3 | 1777 | if (inode->i_ino >= F2FS_ROOT_INO(sbi) && |
09cbfeaf | 1778 | (offset % PAGE_SIZE || length != PAGE_SIZE)) |
a7ffdbe2 JK |
1779 | return; |
1780 | ||
487261f3 CY |
1781 | if (PageDirty(page)) { |
1782 | if (inode->i_ino == F2FS_META_INO(sbi)) | |
1783 | dec_page_count(sbi, F2FS_DIRTY_META); | |
1784 | else if (inode->i_ino == F2FS_NODE_INO(sbi)) | |
1785 | dec_page_count(sbi, F2FS_DIRTY_NODES); | |
1786 | else | |
1787 | inode_dec_dirty_pages(inode); | |
1788 | } | |
decd36b6 CY |
1789 | |
1790 | /* This is atomic written page, keep Private */ | |
1791 | if (IS_ATOMIC_WRITTEN_PAGE(page)) | |
1792 | return; | |
1793 | ||
23dc974e | 1794 | set_page_private(page, 0); |
eb47b800 JK |
1795 | ClearPagePrivate(page); |
1796 | } | |
1797 | ||
487261f3 | 1798 | int f2fs_release_page(struct page *page, gfp_t wait) |
eb47b800 | 1799 | { |
f68daeeb JK |
1800 | /* If this is dirty page, keep PagePrivate */ |
1801 | if (PageDirty(page)) | |
1802 | return 0; | |
1803 | ||
decd36b6 CY |
1804 | /* This is atomic written page, keep Private */ |
1805 | if (IS_ATOMIC_WRITTEN_PAGE(page)) | |
1806 | return 0; | |
1807 | ||
23dc974e | 1808 | set_page_private(page, 0); |
eb47b800 | 1809 | ClearPagePrivate(page); |
c3850aa1 | 1810 | return 1; |
eb47b800 JK |
1811 | } |
1812 | ||
1813 | static int f2fs_set_data_page_dirty(struct page *page) | |
1814 | { | |
1815 | struct address_space *mapping = page->mapping; | |
1816 | struct inode *inode = mapping->host; | |
1817 | ||
26c6b887 JK |
1818 | trace_f2fs_set_page_dirty(page, DATA); |
1819 | ||
eb47b800 | 1820 | SetPageUptodate(page); |
34ba94ba | 1821 | |
1e84371f | 1822 | if (f2fs_is_atomic_file(inode)) { |
decd36b6 CY |
1823 | if (!IS_ATOMIC_WRITTEN_PAGE(page)) { |
1824 | register_inmem_page(inode, page); | |
1825 | return 1; | |
1826 | } | |
1827 | /* | |
1828 | * Previously, this page has been registered, we just | |
1829 | * return here. | |
1830 | */ | |
1831 | return 0; | |
34ba94ba JK |
1832 | } |
1833 | ||
eb47b800 JK |
1834 | if (!PageDirty(page)) { |
1835 | __set_page_dirty_nobuffers(page); | |
a7ffdbe2 | 1836 | update_dirty_page(inode, page); |
eb47b800 JK |
1837 | return 1; |
1838 | } | |
1839 | return 0; | |
1840 | } | |
1841 | ||
c01e54b7 JK |
1842 | static sector_t f2fs_bmap(struct address_space *mapping, sector_t block) |
1843 | { | |
454ae7e5 CY |
1844 | struct inode *inode = mapping->host; |
1845 | ||
1d373a0e JK |
1846 | if (f2fs_has_inline_data(inode)) |
1847 | return 0; | |
1848 | ||
1849 | /* make sure allocating whole blocks */ | |
1850 | if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) | |
1851 | filemap_write_and_wait(mapping); | |
1852 | ||
e2b4e2bc | 1853 | return generic_block_bmap(mapping, block, get_data_block_bmap); |
429511cd CY |
1854 | } |
1855 | ||
eb47b800 JK |
1856 | const struct address_space_operations f2fs_dblock_aops = { |
1857 | .readpage = f2fs_read_data_page, | |
1858 | .readpages = f2fs_read_data_pages, | |
1859 | .writepage = f2fs_write_data_page, | |
1860 | .writepages = f2fs_write_data_pages, | |
1861 | .write_begin = f2fs_write_begin, | |
a1dd3c13 | 1862 | .write_end = f2fs_write_end, |
eb47b800 | 1863 | .set_page_dirty = f2fs_set_data_page_dirty, |
487261f3 CY |
1864 | .invalidatepage = f2fs_invalidate_page, |
1865 | .releasepage = f2fs_release_page, | |
eb47b800 | 1866 | .direct_IO = f2fs_direct_IO, |
c01e54b7 | 1867 | .bmap = f2fs_bmap, |
eb47b800 | 1868 | }; |