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> |
fe76b796 JK |
22 | #include <linux/mm.h> |
23 | #include <linux/memcontrol.h> | |
f1e88660 | 24 | #include <linux/cleancache.h> |
eb47b800 JK |
25 | |
26 | #include "f2fs.h" | |
27 | #include "node.h" | |
28 | #include "segment.h" | |
db9f7c1a | 29 | #include "trace.h" |
848753aa | 30 | #include <trace/events/f2fs.h> |
eb47b800 | 31 | |
4246a0b6 | 32 | static void f2fs_read_end_io(struct bio *bio) |
93dfe2ac | 33 | { |
f568849e LT |
34 | struct bio_vec *bvec; |
35 | int i; | |
93dfe2ac | 36 | |
4375a336 | 37 | if (f2fs_bio_encrypted(bio)) { |
4246a0b6 | 38 | if (bio->bi_error) { |
0b81d077 | 39 | fscrypt_release_ctx(bio->bi_private); |
4375a336 | 40 | } else { |
0b81d077 | 41 | fscrypt_decrypt_bio_pages(bio->bi_private, bio); |
4375a336 JK |
42 | return; |
43 | } | |
44 | } | |
45 | ||
12377024 CY |
46 | bio_for_each_segment_all(bvec, bio, i) { |
47 | struct page *page = bvec->bv_page; | |
f1e88660 | 48 | |
4246a0b6 | 49 | if (!bio->bi_error) { |
237c0790 JK |
50 | if (!PageUptodate(page)) |
51 | SetPageUptodate(page); | |
f1e88660 JK |
52 | } else { |
53 | ClearPageUptodate(page); | |
54 | SetPageError(page); | |
55 | } | |
56 | unlock_page(page); | |
57 | } | |
f1e88660 JK |
58 | bio_put(bio); |
59 | } | |
60 | ||
4246a0b6 | 61 | static void f2fs_write_end_io(struct bio *bio) |
93dfe2ac | 62 | { |
1b1f559f | 63 | struct f2fs_sb_info *sbi = bio->bi_private; |
f568849e LT |
64 | struct bio_vec *bvec; |
65 | int i; | |
93dfe2ac | 66 | |
f568849e | 67 | bio_for_each_segment_all(bvec, bio, i) { |
93dfe2ac JK |
68 | struct page *page = bvec->bv_page; |
69 | ||
0b81d077 | 70 | fscrypt_pullback_bio_page(&page, true); |
4375a336 | 71 | |
4246a0b6 | 72 | if (unlikely(bio->bi_error)) { |
93dfe2ac | 73 | set_bit(AS_EIO, &page->mapping->flags); |
38f91ca8 | 74 | f2fs_stop_checkpoint(sbi, true); |
93dfe2ac JK |
75 | } |
76 | end_page_writeback(page); | |
f568849e | 77 | } |
f5730184 JK |
78 | if (atomic_dec_and_test(&sbi->nr_wb_bios) && |
79 | wq_has_sleeper(&sbi->cp_wait)) | |
93dfe2ac JK |
80 | wake_up(&sbi->cp_wait); |
81 | ||
82 | bio_put(bio); | |
83 | } | |
84 | ||
940a6d34 GZ |
85 | /* |
86 | * Low-level block read/write IO operations. | |
87 | */ | |
88 | static struct bio *__bio_alloc(struct f2fs_sb_info *sbi, block_t blk_addr, | |
89 | int npages, bool is_read) | |
90 | { | |
91 | struct bio *bio; | |
92 | ||
740432f8 | 93 | bio = f2fs_bio_alloc(npages); |
940a6d34 GZ |
94 | |
95 | bio->bi_bdev = sbi->sb->s_bdev; | |
55cf9cb6 | 96 | bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr); |
940a6d34 | 97 | bio->bi_end_io = is_read ? f2fs_read_end_io : f2fs_write_end_io; |
12377024 | 98 | bio->bi_private = is_read ? NULL : sbi; |
940a6d34 GZ |
99 | |
100 | return bio; | |
101 | } | |
102 | ||
f5730184 | 103 | static inline void __submit_bio(struct f2fs_sb_info *sbi, int rw, |
19a5f5e2 | 104 | struct bio *bio, enum page_type type) |
f5730184 | 105 | { |
19a5f5e2 | 106 | if (!is_read_io(rw)) { |
f5730184 | 107 | atomic_inc(&sbi->nr_wb_bios); |
52763a4b JK |
108 | if (f2fs_sb_mounted_hmsmr(sbi->sb) && |
109 | current->plug && (type == DATA || type == NODE)) | |
19a5f5e2 JK |
110 | blk_finish_plug(current->plug); |
111 | } | |
f5730184 JK |
112 | submit_bio(rw, bio); |
113 | } | |
114 | ||
458e6197 | 115 | static void __submit_merged_bio(struct f2fs_bio_info *io) |
93dfe2ac | 116 | { |
458e6197 | 117 | struct f2fs_io_info *fio = &io->fio; |
93dfe2ac JK |
118 | |
119 | if (!io->bio) | |
120 | return; | |
121 | ||
6a8f8ca5 | 122 | if (is_read_io(fio->rw)) |
2ace38e0 | 123 | trace_f2fs_submit_read_bio(io->sbi->sb, fio, io->bio); |
6a8f8ca5 | 124 | else |
2ace38e0 | 125 | trace_f2fs_submit_write_bio(io->sbi->sb, fio, io->bio); |
940a6d34 | 126 | |
19a5f5e2 | 127 | __submit_bio(io->sbi, fio->rw, io->bio, fio->type); |
93dfe2ac JK |
128 | io->bio = NULL; |
129 | } | |
130 | ||
0c3a5797 CY |
131 | static bool __has_merged_page(struct f2fs_bio_info *io, struct inode *inode, |
132 | struct page *page, nid_t ino) | |
0fd785eb | 133 | { |
0fd785eb CY |
134 | struct bio_vec *bvec; |
135 | struct page *target; | |
136 | int i; | |
137 | ||
0c3a5797 | 138 | if (!io->bio) |
0fd785eb | 139 | return false; |
0c3a5797 CY |
140 | |
141 | if (!inode && !page && !ino) | |
142 | return true; | |
0fd785eb CY |
143 | |
144 | bio_for_each_segment_all(bvec, io->bio, i) { | |
145 | ||
0b81d077 | 146 | if (bvec->bv_page->mapping) |
0fd785eb | 147 | target = bvec->bv_page; |
0b81d077 JK |
148 | else |
149 | target = fscrypt_control_page(bvec->bv_page); | |
0fd785eb | 150 | |
0c3a5797 CY |
151 | if (inode && inode == target->mapping->host) |
152 | return true; | |
153 | if (page && page == target) | |
154 | return true; | |
155 | if (ino && ino == ino_of_node(target)) | |
0fd785eb | 156 | return true; |
0fd785eb CY |
157 | } |
158 | ||
0fd785eb CY |
159 | return false; |
160 | } | |
161 | ||
0c3a5797 CY |
162 | static bool has_merged_page(struct f2fs_sb_info *sbi, struct inode *inode, |
163 | struct page *page, nid_t ino, | |
164 | enum page_type type) | |
165 | { | |
166 | enum page_type btype = PAGE_TYPE_OF_BIO(type); | |
167 | struct f2fs_bio_info *io = &sbi->write_io[btype]; | |
168 | bool ret; | |
169 | ||
170 | down_read(&io->io_rwsem); | |
171 | ret = __has_merged_page(io, inode, page, ino); | |
172 | up_read(&io->io_rwsem); | |
173 | return ret; | |
174 | } | |
175 | ||
176 | static void __f2fs_submit_merged_bio(struct f2fs_sb_info *sbi, | |
177 | struct inode *inode, struct page *page, | |
178 | nid_t ino, enum page_type type, int rw) | |
93dfe2ac JK |
179 | { |
180 | enum page_type btype = PAGE_TYPE_OF_BIO(type); | |
181 | struct f2fs_bio_info *io; | |
182 | ||
183 | io = is_read_io(rw) ? &sbi->read_io : &sbi->write_io[btype]; | |
184 | ||
df0f8dc0 | 185 | down_write(&io->io_rwsem); |
458e6197 | 186 | |
0c3a5797 CY |
187 | if (!__has_merged_page(io, inode, page, ino)) |
188 | goto out; | |
189 | ||
458e6197 JK |
190 | /* change META to META_FLUSH in the checkpoint procedure */ |
191 | if (type >= META_FLUSH) { | |
192 | io->fio.type = META_FLUSH; | |
0f7b2abd JK |
193 | if (test_opt(sbi, NOBARRIER)) |
194 | io->fio.rw = WRITE_FLUSH | REQ_META | REQ_PRIO; | |
195 | else | |
196 | io->fio.rw = WRITE_FLUSH_FUA | REQ_META | REQ_PRIO; | |
458e6197 JK |
197 | } |
198 | __submit_merged_bio(io); | |
0c3a5797 | 199 | out: |
df0f8dc0 | 200 | up_write(&io->io_rwsem); |
93dfe2ac JK |
201 | } |
202 | ||
0c3a5797 CY |
203 | void f2fs_submit_merged_bio(struct f2fs_sb_info *sbi, enum page_type type, |
204 | int rw) | |
205 | { | |
206 | __f2fs_submit_merged_bio(sbi, NULL, NULL, 0, type, rw); | |
207 | } | |
208 | ||
209 | void f2fs_submit_merged_bio_cond(struct f2fs_sb_info *sbi, | |
210 | struct inode *inode, struct page *page, | |
211 | nid_t ino, enum page_type type, int rw) | |
212 | { | |
213 | if (has_merged_page(sbi, inode, page, ino, type)) | |
214 | __f2fs_submit_merged_bio(sbi, inode, page, ino, type, rw); | |
215 | } | |
216 | ||
406657dd CY |
217 | void f2fs_flush_merged_bios(struct f2fs_sb_info *sbi) |
218 | { | |
219 | f2fs_submit_merged_bio(sbi, DATA, WRITE); | |
220 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
221 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
222 | } | |
223 | ||
93dfe2ac JK |
224 | /* |
225 | * Fill the locked page with data located in the block address. | |
226 | * Return unlocked page. | |
227 | */ | |
05ca3632 | 228 | int f2fs_submit_page_bio(struct f2fs_io_info *fio) |
93dfe2ac | 229 | { |
93dfe2ac | 230 | struct bio *bio; |
0b81d077 JK |
231 | struct page *page = fio->encrypted_page ? |
232 | fio->encrypted_page : fio->page; | |
93dfe2ac | 233 | |
2ace38e0 | 234 | trace_f2fs_submit_page_bio(page, fio); |
05ca3632 | 235 | f2fs_trace_ios(fio, 0); |
93dfe2ac JK |
236 | |
237 | /* Allocate a new bio */ | |
7a9d7548 | 238 | bio = __bio_alloc(fio->sbi, fio->new_blkaddr, 1, is_read_io(fio->rw)); |
93dfe2ac | 239 | |
09cbfeaf | 240 | if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) { |
93dfe2ac | 241 | bio_put(bio); |
93dfe2ac JK |
242 | return -EFAULT; |
243 | } | |
244 | ||
19a5f5e2 | 245 | __submit_bio(fio->sbi, fio->rw, bio, fio->type); |
93dfe2ac JK |
246 | return 0; |
247 | } | |
248 | ||
05ca3632 | 249 | void f2fs_submit_page_mbio(struct f2fs_io_info *fio) |
93dfe2ac | 250 | { |
05ca3632 | 251 | struct f2fs_sb_info *sbi = fio->sbi; |
458e6197 | 252 | enum page_type btype = PAGE_TYPE_OF_BIO(fio->type); |
93dfe2ac | 253 | struct f2fs_bio_info *io; |
940a6d34 | 254 | bool is_read = is_read_io(fio->rw); |
4375a336 | 255 | struct page *bio_page; |
93dfe2ac | 256 | |
940a6d34 | 257 | io = is_read ? &sbi->read_io : &sbi->write_io[btype]; |
93dfe2ac | 258 | |
7a9d7548 CY |
259 | if (fio->old_blkaddr != NEW_ADDR) |
260 | verify_block_addr(sbi, fio->old_blkaddr); | |
261 | verify_block_addr(sbi, fio->new_blkaddr); | |
93dfe2ac | 262 | |
df0f8dc0 | 263 | down_write(&io->io_rwsem); |
93dfe2ac | 264 | |
7a9d7548 | 265 | if (io->bio && (io->last_block_in_bio != fio->new_blkaddr - 1 || |
458e6197 JK |
266 | io->fio.rw != fio->rw)) |
267 | __submit_merged_bio(io); | |
93dfe2ac JK |
268 | alloc_new: |
269 | if (io->bio == NULL) { | |
90a893c7 | 270 | int bio_blocks = MAX_BIO_BLOCKS(sbi); |
940a6d34 | 271 | |
7a9d7548 CY |
272 | io->bio = __bio_alloc(sbi, fio->new_blkaddr, |
273 | bio_blocks, is_read); | |
458e6197 | 274 | io->fio = *fio; |
93dfe2ac JK |
275 | } |
276 | ||
4375a336 JK |
277 | bio_page = fio->encrypted_page ? fio->encrypted_page : fio->page; |
278 | ||
09cbfeaf KS |
279 | if (bio_add_page(io->bio, bio_page, PAGE_SIZE, 0) < |
280 | PAGE_SIZE) { | |
458e6197 | 281 | __submit_merged_bio(io); |
93dfe2ac JK |
282 | goto alloc_new; |
283 | } | |
284 | ||
7a9d7548 | 285 | io->last_block_in_bio = fio->new_blkaddr; |
05ca3632 | 286 | f2fs_trace_ios(fio, 0); |
93dfe2ac | 287 | |
df0f8dc0 | 288 | up_write(&io->io_rwsem); |
05ca3632 | 289 | trace_f2fs_submit_page_mbio(fio->page, fio); |
93dfe2ac JK |
290 | } |
291 | ||
46008c6d CY |
292 | static void __set_data_blkaddr(struct dnode_of_data *dn) |
293 | { | |
294 | struct f2fs_node *rn = F2FS_NODE(dn->node_page); | |
295 | __le32 *addr_array; | |
296 | ||
297 | /* Get physical address of data block */ | |
298 | addr_array = blkaddr_in_node(rn); | |
299 | addr_array[dn->ofs_in_node] = cpu_to_le32(dn->data_blkaddr); | |
300 | } | |
301 | ||
0a8165d7 | 302 | /* |
eb47b800 JK |
303 | * Lock ordering for the change of data block address: |
304 | * ->data_page | |
305 | * ->node_page | |
306 | * update block addresses in the node page | |
307 | */ | |
216a620a | 308 | void set_data_blkaddr(struct dnode_of_data *dn) |
eb47b800 | 309 | { |
46008c6d CY |
310 | f2fs_wait_on_page_writeback(dn->node_page, NODE, true); |
311 | __set_data_blkaddr(dn); | |
312 | if (set_page_dirty(dn->node_page)) | |
12719ae1 | 313 | dn->node_changed = true; |
eb47b800 JK |
314 | } |
315 | ||
f28b3434 CY |
316 | void f2fs_update_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr) |
317 | { | |
318 | dn->data_blkaddr = blkaddr; | |
319 | set_data_blkaddr(dn); | |
320 | f2fs_update_extent_cache(dn); | |
321 | } | |
322 | ||
46008c6d CY |
323 | /* dn->ofs_in_node will be returned with up-to-date last block pointer */ |
324 | int reserve_new_blocks(struct dnode_of_data *dn, blkcnt_t count) | |
eb47b800 | 325 | { |
4081363f | 326 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
eb47b800 | 327 | |
46008c6d CY |
328 | if (!count) |
329 | return 0; | |
330 | ||
91942321 | 331 | if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC))) |
eb47b800 | 332 | return -EPERM; |
46008c6d | 333 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, &count))) |
eb47b800 JK |
334 | return -ENOSPC; |
335 | ||
46008c6d CY |
336 | trace_f2fs_reserve_new_blocks(dn->inode, dn->nid, |
337 | dn->ofs_in_node, count); | |
338 | ||
339 | f2fs_wait_on_page_writeback(dn->node_page, NODE, true); | |
340 | ||
341 | for (; count > 0; dn->ofs_in_node++) { | |
342 | block_t blkaddr = | |
343 | datablock_addr(dn->node_page, dn->ofs_in_node); | |
344 | if (blkaddr == NULL_ADDR) { | |
345 | dn->data_blkaddr = NEW_ADDR; | |
346 | __set_data_blkaddr(dn); | |
347 | count--; | |
348 | } | |
349 | } | |
350 | ||
351 | if (set_page_dirty(dn->node_page)) | |
352 | dn->node_changed = true; | |
eb47b800 JK |
353 | return 0; |
354 | } | |
355 | ||
46008c6d CY |
356 | /* Should keep dn->ofs_in_node unchanged */ |
357 | int reserve_new_block(struct dnode_of_data *dn) | |
358 | { | |
359 | unsigned int ofs_in_node = dn->ofs_in_node; | |
360 | int ret; | |
361 | ||
362 | ret = reserve_new_blocks(dn, 1); | |
363 | dn->ofs_in_node = ofs_in_node; | |
364 | return ret; | |
365 | } | |
366 | ||
b600965c HL |
367 | int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index) |
368 | { | |
369 | bool need_put = dn->inode_page ? false : true; | |
370 | int err; | |
371 | ||
372 | err = get_dnode_of_data(dn, index, ALLOC_NODE); | |
373 | if (err) | |
374 | return err; | |
a8865372 | 375 | |
b600965c HL |
376 | if (dn->data_blkaddr == NULL_ADDR) |
377 | err = reserve_new_block(dn); | |
a8865372 | 378 | if (err || need_put) |
b600965c HL |
379 | f2fs_put_dnode(dn); |
380 | return err; | |
381 | } | |
382 | ||
759af1c9 | 383 | int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index) |
eb47b800 | 384 | { |
028a41e8 | 385 | struct extent_info ei; |
759af1c9 | 386 | struct inode *inode = dn->inode; |
028a41e8 | 387 | |
759af1c9 FL |
388 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { |
389 | dn->data_blkaddr = ei.blk + index - ei.fofs; | |
390 | return 0; | |
429511cd | 391 | } |
028a41e8 | 392 | |
759af1c9 | 393 | return f2fs_reserve_block(dn, index); |
eb47b800 JK |
394 | } |
395 | ||
a56c7c6f JK |
396 | struct page *get_read_data_page(struct inode *inode, pgoff_t index, |
397 | int rw, bool for_write) | |
eb47b800 | 398 | { |
eb47b800 JK |
399 | struct address_space *mapping = inode->i_mapping; |
400 | struct dnode_of_data dn; | |
401 | struct page *page; | |
cb3bc9ee | 402 | struct extent_info ei; |
eb47b800 | 403 | int err; |
cf04e8eb | 404 | struct f2fs_io_info fio = { |
05ca3632 | 405 | .sbi = F2FS_I_SB(inode), |
cf04e8eb | 406 | .type = DATA, |
43f3eae1 | 407 | .rw = rw, |
4375a336 | 408 | .encrypted_page = NULL, |
cf04e8eb | 409 | }; |
eb47b800 | 410 | |
4375a336 JK |
411 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) |
412 | return read_mapping_page(mapping, index, NULL); | |
413 | ||
a56c7c6f | 414 | page = f2fs_grab_cache_page(mapping, index, for_write); |
650495de JK |
415 | if (!page) |
416 | return ERR_PTR(-ENOMEM); | |
417 | ||
cb3bc9ee CY |
418 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { |
419 | dn.data_blkaddr = ei.blk + index - ei.fofs; | |
420 | goto got_it; | |
421 | } | |
422 | ||
eb47b800 | 423 | set_new_dnode(&dn, inode, NULL, NULL, 0); |
266e97a8 | 424 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); |
86531d6b JK |
425 | if (err) |
426 | goto put_err; | |
eb47b800 JK |
427 | f2fs_put_dnode(&dn); |
428 | ||
6bacf52f | 429 | if (unlikely(dn.data_blkaddr == NULL_ADDR)) { |
86531d6b JK |
430 | err = -ENOENT; |
431 | goto put_err; | |
650495de | 432 | } |
cb3bc9ee | 433 | got_it: |
43f3eae1 JK |
434 | if (PageUptodate(page)) { |
435 | unlock_page(page); | |
eb47b800 | 436 | return page; |
43f3eae1 | 437 | } |
eb47b800 | 438 | |
d59ff4df JK |
439 | /* |
440 | * A new dentry page is allocated but not able to be written, since its | |
441 | * new inode page couldn't be allocated due to -ENOSPC. | |
442 | * In such the case, its blkaddr can be remained as NEW_ADDR. | |
443 | * see, f2fs_add_link -> get_new_data_page -> init_inode_metadata. | |
444 | */ | |
445 | if (dn.data_blkaddr == NEW_ADDR) { | |
09cbfeaf | 446 | zero_user_segment(page, 0, PAGE_SIZE); |
237c0790 JK |
447 | if (!PageUptodate(page)) |
448 | SetPageUptodate(page); | |
43f3eae1 | 449 | unlock_page(page); |
d59ff4df JK |
450 | return page; |
451 | } | |
eb47b800 | 452 | |
7a9d7548 | 453 | fio.new_blkaddr = fio.old_blkaddr = dn.data_blkaddr; |
05ca3632 JK |
454 | fio.page = page; |
455 | err = f2fs_submit_page_bio(&fio); | |
393ff91f | 456 | if (err) |
86531d6b | 457 | goto put_err; |
43f3eae1 | 458 | return page; |
86531d6b JK |
459 | |
460 | put_err: | |
461 | f2fs_put_page(page, 1); | |
462 | return ERR_PTR(err); | |
43f3eae1 JK |
463 | } |
464 | ||
465 | struct page *find_data_page(struct inode *inode, pgoff_t index) | |
466 | { | |
467 | struct address_space *mapping = inode->i_mapping; | |
468 | struct page *page; | |
469 | ||
470 | page = find_get_page(mapping, index); | |
471 | if (page && PageUptodate(page)) | |
472 | return page; | |
473 | f2fs_put_page(page, 0); | |
474 | ||
a56c7c6f | 475 | page = get_read_data_page(inode, index, READ_SYNC, false); |
43f3eae1 JK |
476 | if (IS_ERR(page)) |
477 | return page; | |
478 | ||
479 | if (PageUptodate(page)) | |
480 | return page; | |
481 | ||
482 | wait_on_page_locked(page); | |
483 | if (unlikely(!PageUptodate(page))) { | |
484 | f2fs_put_page(page, 0); | |
485 | return ERR_PTR(-EIO); | |
486 | } | |
487 | return page; | |
488 | } | |
489 | ||
490 | /* | |
491 | * If it tries to access a hole, return an error. | |
492 | * Because, the callers, functions in dir.c and GC, should be able to know | |
493 | * whether this page exists or not. | |
494 | */ | |
a56c7c6f JK |
495 | struct page *get_lock_data_page(struct inode *inode, pgoff_t index, |
496 | bool for_write) | |
43f3eae1 JK |
497 | { |
498 | struct address_space *mapping = inode->i_mapping; | |
499 | struct page *page; | |
500 | repeat: | |
a56c7c6f | 501 | page = get_read_data_page(inode, index, READ_SYNC, for_write); |
43f3eae1 JK |
502 | if (IS_ERR(page)) |
503 | return page; | |
393ff91f | 504 | |
43f3eae1 | 505 | /* wait for read completion */ |
393ff91f | 506 | lock_page(page); |
6bacf52f | 507 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
508 | f2fs_put_page(page, 1); |
509 | goto repeat; | |
eb47b800 | 510 | } |
1563ac75 CY |
511 | if (unlikely(!PageUptodate(page))) { |
512 | f2fs_put_page(page, 1); | |
513 | return ERR_PTR(-EIO); | |
514 | } | |
eb47b800 JK |
515 | return page; |
516 | } | |
517 | ||
0a8165d7 | 518 | /* |
eb47b800 JK |
519 | * Caller ensures that this data page is never allocated. |
520 | * A new zero-filled data page is allocated in the page cache. | |
39936837 | 521 | * |
4f4124d0 CY |
522 | * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and |
523 | * f2fs_unlock_op(). | |
470f00e9 CY |
524 | * Note that, ipage is set only by make_empty_dir, and if any error occur, |
525 | * ipage should be released by this function. | |
eb47b800 | 526 | */ |
64aa7ed9 | 527 | struct page *get_new_data_page(struct inode *inode, |
a8865372 | 528 | struct page *ipage, pgoff_t index, bool new_i_size) |
eb47b800 | 529 | { |
eb47b800 JK |
530 | struct address_space *mapping = inode->i_mapping; |
531 | struct page *page; | |
532 | struct dnode_of_data dn; | |
533 | int err; | |
7612118a | 534 | |
a56c7c6f | 535 | page = f2fs_grab_cache_page(mapping, index, true); |
470f00e9 CY |
536 | if (!page) { |
537 | /* | |
538 | * before exiting, we should make sure ipage will be released | |
539 | * if any error occur. | |
540 | */ | |
541 | f2fs_put_page(ipage, 1); | |
01f28610 | 542 | return ERR_PTR(-ENOMEM); |
470f00e9 | 543 | } |
eb47b800 | 544 | |
a8865372 | 545 | set_new_dnode(&dn, inode, ipage, NULL, 0); |
b600965c | 546 | err = f2fs_reserve_block(&dn, index); |
01f28610 JK |
547 | if (err) { |
548 | f2fs_put_page(page, 1); | |
eb47b800 | 549 | return ERR_PTR(err); |
a8865372 | 550 | } |
01f28610 JK |
551 | if (!ipage) |
552 | f2fs_put_dnode(&dn); | |
eb47b800 JK |
553 | |
554 | if (PageUptodate(page)) | |
01f28610 | 555 | goto got_it; |
eb47b800 JK |
556 | |
557 | if (dn.data_blkaddr == NEW_ADDR) { | |
09cbfeaf | 558 | zero_user_segment(page, 0, PAGE_SIZE); |
237c0790 JK |
559 | if (!PageUptodate(page)) |
560 | SetPageUptodate(page); | |
eb47b800 | 561 | } else { |
4375a336 | 562 | f2fs_put_page(page, 1); |
a8865372 | 563 | |
7612118a JK |
564 | /* if ipage exists, blkaddr should be NEW_ADDR */ |
565 | f2fs_bug_on(F2FS_I_SB(inode), ipage); | |
566 | page = get_lock_data_page(inode, index, true); | |
4375a336 | 567 | if (IS_ERR(page)) |
7612118a | 568 | return page; |
eb47b800 | 569 | } |
01f28610 | 570 | got_it: |
9edcdabf | 571 | if (new_i_size && i_size_read(inode) < |
ee6d182f | 572 | ((loff_t)(index + 1) << PAGE_SHIFT)) |
fc9581c8 | 573 | f2fs_i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT)); |
eb47b800 JK |
574 | return page; |
575 | } | |
576 | ||
bfad7c2d JK |
577 | static int __allocate_data_block(struct dnode_of_data *dn) |
578 | { | |
4081363f | 579 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
bfad7c2d | 580 | struct f2fs_summary sum; |
bfad7c2d | 581 | struct node_info ni; |
38aa0889 | 582 | int seg = CURSEG_WARM_DATA; |
976e4c50 | 583 | pgoff_t fofs; |
46008c6d | 584 | blkcnt_t count = 1; |
bfad7c2d | 585 | |
91942321 | 586 | if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC))) |
bfad7c2d | 587 | return -EPERM; |
df6136ef CY |
588 | |
589 | dn->data_blkaddr = datablock_addr(dn->node_page, dn->ofs_in_node); | |
590 | if (dn->data_blkaddr == NEW_ADDR) | |
591 | goto alloc; | |
592 | ||
46008c6d | 593 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, &count))) |
bfad7c2d JK |
594 | return -ENOSPC; |
595 | ||
df6136ef | 596 | alloc: |
bfad7c2d JK |
597 | get_node_info(sbi, dn->nid, &ni); |
598 | set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version); | |
599 | ||
38aa0889 JK |
600 | if (dn->ofs_in_node == 0 && dn->inode_page == dn->node_page) |
601 | seg = CURSEG_DIRECT_IO; | |
602 | ||
df6136ef CY |
603 | allocate_data_block(sbi, NULL, dn->data_blkaddr, &dn->data_blkaddr, |
604 | &sum, seg); | |
216a620a | 605 | set_data_blkaddr(dn); |
bfad7c2d | 606 | |
976e4c50 | 607 | /* update i_size */ |
81ca7350 | 608 | fofs = start_bidx_of_node(ofs_of_node(dn->node_page), dn->inode) + |
976e4c50 | 609 | dn->ofs_in_node; |
09cbfeaf | 610 | if (i_size_read(dn->inode) < ((loff_t)(fofs + 1) << PAGE_SHIFT)) |
fc9581c8 | 611 | f2fs_i_size_write(dn->inode, |
09cbfeaf | 612 | ((loff_t)(fofs + 1) << PAGE_SHIFT)); |
bfad7c2d JK |
613 | return 0; |
614 | } | |
615 | ||
b439b103 | 616 | ssize_t f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *from) |
59b802e5 | 617 | { |
b439b103 | 618 | struct inode *inode = file_inode(iocb->ki_filp); |
5b8db7fa | 619 | struct f2fs_map_blocks map; |
b439b103 | 620 | ssize_t ret = 0; |
59b802e5 | 621 | |
0080c507 JK |
622 | map.m_lblk = F2FS_BLK_ALIGN(iocb->ki_pos); |
623 | map.m_len = F2FS_BYTES_TO_BLK(iov_iter_count(from)); | |
da85985c | 624 | map.m_next_pgofs = NULL; |
2a340760 | 625 | |
24b84912 JK |
626 | if (f2fs_encrypted_inode(inode)) |
627 | return 0; | |
628 | ||
629 | if (iocb->ki_flags & IOCB_DIRECT) { | |
630 | ret = f2fs_convert_inline_inode(inode); | |
631 | if (ret) | |
632 | return ret; | |
633 | return f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO); | |
634 | } | |
635 | if (iocb->ki_pos + iov_iter_count(from) > MAX_INLINE_DATA) { | |
b439b103 JK |
636 | ret = f2fs_convert_inline_inode(inode); |
637 | if (ret) | |
638 | return ret; | |
b439b103 | 639 | } |
24b84912 JK |
640 | if (!f2fs_has_inline_data(inode)) |
641 | return f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO); | |
b439b103 | 642 | return ret; |
59b802e5 JK |
643 | } |
644 | ||
0a8165d7 | 645 | /* |
003a3e1d JK |
646 | * f2fs_map_blocks() now supported readahead/bmap/rw direct_IO with |
647 | * f2fs_map_blocks structure. | |
4f4124d0 CY |
648 | * If original data blocks are allocated, then give them to blockdev. |
649 | * Otherwise, | |
650 | * a. preallocate requested block addresses | |
651 | * b. do not use extent cache for better performance | |
652 | * c. give the block addresses to blockdev | |
eb47b800 | 653 | */ |
d323d005 | 654 | int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map, |
e2b4e2bc | 655 | int create, int flag) |
eb47b800 | 656 | { |
003a3e1d | 657 | unsigned int maxblocks = map->m_len; |
eb47b800 | 658 | struct dnode_of_data dn; |
f9811703 | 659 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
ac6f1999 | 660 | int mode = create ? ALLOC_NODE : LOOKUP_NODE; |
46008c6d | 661 | pgoff_t pgofs, end_offset, end; |
bfad7c2d | 662 | int err = 0, ofs = 1; |
46008c6d CY |
663 | unsigned int ofs_in_node, last_ofs_in_node; |
664 | blkcnt_t prealloc; | |
a2e7d1bf | 665 | struct extent_info ei; |
bfad7c2d | 666 | bool allocated = false; |
7df3a431 | 667 | block_t blkaddr; |
eb47b800 | 668 | |
003a3e1d JK |
669 | map->m_len = 0; |
670 | map->m_flags = 0; | |
671 | ||
672 | /* it only supports block size == page size */ | |
673 | pgofs = (pgoff_t)map->m_lblk; | |
46008c6d | 674 | end = pgofs + maxblocks; |
eb47b800 | 675 | |
24b84912 | 676 | if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) { |
003a3e1d JK |
677 | map->m_pblk = ei.blk + pgofs - ei.fofs; |
678 | map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs); | |
679 | map->m_flags = F2FS_MAP_MAPPED; | |
bfad7c2d | 680 | goto out; |
a2e7d1bf | 681 | } |
bfad7c2d | 682 | |
4fe71e88 | 683 | next_dnode: |
59b802e5 | 684 | if (create) |
3104af35 | 685 | f2fs_lock_op(sbi); |
eb47b800 JK |
686 | |
687 | /* When reading holes, we need its node page */ | |
688 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
bfad7c2d | 689 | err = get_dnode_of_data(&dn, pgofs, mode); |
1ec79083 | 690 | if (err) { |
43473f96 CY |
691 | if (flag == F2FS_GET_BLOCK_BMAP) |
692 | map->m_pblk = 0; | |
da85985c | 693 | if (err == -ENOENT) { |
bfad7c2d | 694 | err = 0; |
da85985c CY |
695 | if (map->m_next_pgofs) |
696 | *map->m_next_pgofs = | |
697 | get_next_page_offset(&dn, pgofs); | |
698 | } | |
bfad7c2d | 699 | goto unlock_out; |
848753aa | 700 | } |
973163fc | 701 | |
46008c6d CY |
702 | prealloc = 0; |
703 | ofs_in_node = dn.ofs_in_node; | |
81ca7350 | 704 | end_offset = ADDRS_PER_PAGE(dn.node_page, inode); |
4fe71e88 CY |
705 | |
706 | next_block: | |
707 | blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node); | |
708 | ||
709 | if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR) { | |
973163fc | 710 | if (create) { |
f9811703 CY |
711 | if (unlikely(f2fs_cp_error(sbi))) { |
712 | err = -EIO; | |
4fe71e88 | 713 | goto sync_out; |
f9811703 | 714 | } |
24b84912 | 715 | if (flag == F2FS_GET_BLOCK_PRE_AIO) { |
46008c6d CY |
716 | if (blkaddr == NULL_ADDR) { |
717 | prealloc++; | |
718 | last_ofs_in_node = dn.ofs_in_node; | |
719 | } | |
24b84912 JK |
720 | } else { |
721 | err = __allocate_data_block(&dn); | |
46008c6d | 722 | if (!err) { |
91942321 | 723 | set_inode_flag(inode, FI_APPEND_WRITE); |
46008c6d CY |
724 | allocated = true; |
725 | } | |
24b84912 | 726 | } |
973163fc | 727 | if (err) |
4fe71e88 | 728 | goto sync_out; |
973163fc | 729 | map->m_flags = F2FS_MAP_NEW; |
4fe71e88 | 730 | blkaddr = dn.data_blkaddr; |
973163fc | 731 | } else { |
43473f96 CY |
732 | if (flag == F2FS_GET_BLOCK_BMAP) { |
733 | map->m_pblk = 0; | |
734 | goto sync_out; | |
735 | } | |
da85985c CY |
736 | if (flag == F2FS_GET_BLOCK_FIEMAP && |
737 | blkaddr == NULL_ADDR) { | |
738 | if (map->m_next_pgofs) | |
739 | *map->m_next_pgofs = pgofs + 1; | |
740 | } | |
973163fc | 741 | if (flag != F2FS_GET_BLOCK_FIEMAP || |
43473f96 | 742 | blkaddr != NEW_ADDR) |
4fe71e88 | 743 | goto sync_out; |
e2b4e2bc | 744 | } |
e2b4e2bc | 745 | } |
eb47b800 | 746 | |
46008c6d CY |
747 | if (flag == F2FS_GET_BLOCK_PRE_AIO) |
748 | goto skip; | |
749 | ||
4fe71e88 CY |
750 | if (map->m_len == 0) { |
751 | /* preallocated unwritten block should be mapped for fiemap. */ | |
752 | if (blkaddr == NEW_ADDR) | |
753 | map->m_flags |= F2FS_MAP_UNWRITTEN; | |
754 | map->m_flags |= F2FS_MAP_MAPPED; | |
755 | ||
756 | map->m_pblk = blkaddr; | |
757 | map->m_len = 1; | |
758 | } else if ((map->m_pblk != NEW_ADDR && | |
759 | blkaddr == (map->m_pblk + ofs)) || | |
b439b103 | 760 | (map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR) || |
46008c6d | 761 | flag == F2FS_GET_BLOCK_PRE_DIO) { |
4fe71e88 CY |
762 | ofs++; |
763 | map->m_len++; | |
764 | } else { | |
765 | goto sync_out; | |
766 | } | |
bfad7c2d | 767 | |
46008c6d | 768 | skip: |
bfad7c2d JK |
769 | dn.ofs_in_node++; |
770 | pgofs++; | |
771 | ||
46008c6d CY |
772 | /* preallocate blocks in batch for one dnode page */ |
773 | if (flag == F2FS_GET_BLOCK_PRE_AIO && | |
774 | (pgofs == end || dn.ofs_in_node == end_offset)) { | |
7df3a431 | 775 | |
46008c6d CY |
776 | dn.ofs_in_node = ofs_in_node; |
777 | err = reserve_new_blocks(&dn, prealloc); | |
778 | if (err) | |
779 | goto sync_out; | |
bfad7c2d | 780 | |
46008c6d CY |
781 | map->m_len += dn.ofs_in_node - ofs_in_node; |
782 | if (prealloc && dn.ofs_in_node != last_ofs_in_node + 1) { | |
783 | err = -ENOSPC; | |
784 | goto sync_out; | |
3104af35 | 785 | } |
46008c6d CY |
786 | dn.ofs_in_node = end_offset; |
787 | } | |
788 | ||
789 | if (pgofs >= end) | |
790 | goto sync_out; | |
791 | else if (dn.ofs_in_node < end_offset) | |
792 | goto next_block; | |
793 | ||
46008c6d CY |
794 | f2fs_put_dnode(&dn); |
795 | ||
796 | if (create) { | |
797 | f2fs_unlock_op(sbi); | |
798 | f2fs_balance_fs(sbi, allocated); | |
eb47b800 | 799 | } |
46008c6d CY |
800 | allocated = false; |
801 | goto next_dnode; | |
7df3a431 | 802 | |
bfad7c2d | 803 | sync_out: |
eb47b800 | 804 | f2fs_put_dnode(&dn); |
bfad7c2d | 805 | unlock_out: |
2a340760 | 806 | if (create) { |
3104af35 | 807 | f2fs_unlock_op(sbi); |
3c082b7b | 808 | f2fs_balance_fs(sbi, allocated); |
2a340760 | 809 | } |
bfad7c2d | 810 | out: |
003a3e1d | 811 | trace_f2fs_map_blocks(inode, map, err); |
bfad7c2d | 812 | return err; |
eb47b800 JK |
813 | } |
814 | ||
003a3e1d | 815 | static int __get_data_block(struct inode *inode, sector_t iblock, |
da85985c CY |
816 | struct buffer_head *bh, int create, int flag, |
817 | pgoff_t *next_pgofs) | |
003a3e1d JK |
818 | { |
819 | struct f2fs_map_blocks map; | |
820 | int ret; | |
821 | ||
822 | map.m_lblk = iblock; | |
823 | map.m_len = bh->b_size >> inode->i_blkbits; | |
da85985c | 824 | map.m_next_pgofs = next_pgofs; |
003a3e1d | 825 | |
e2b4e2bc | 826 | ret = f2fs_map_blocks(inode, &map, create, flag); |
003a3e1d JK |
827 | if (!ret) { |
828 | map_bh(bh, inode->i_sb, map.m_pblk); | |
829 | bh->b_state = (bh->b_state & ~F2FS_MAP_FLAGS) | map.m_flags; | |
830 | bh->b_size = map.m_len << inode->i_blkbits; | |
831 | } | |
832 | return ret; | |
833 | } | |
834 | ||
ccfb3000 | 835 | static int get_data_block(struct inode *inode, sector_t iblock, |
da85985c CY |
836 | struct buffer_head *bh_result, int create, int flag, |
837 | pgoff_t *next_pgofs) | |
e2b4e2bc | 838 | { |
da85985c CY |
839 | return __get_data_block(inode, iblock, bh_result, create, |
840 | flag, next_pgofs); | |
e2b4e2bc CY |
841 | } |
842 | ||
843 | static int get_data_block_dio(struct inode *inode, sector_t iblock, | |
ccfb3000 JK |
844 | struct buffer_head *bh_result, int create) |
845 | { | |
e2b4e2bc | 846 | return __get_data_block(inode, iblock, bh_result, create, |
da85985c | 847 | F2FS_GET_BLOCK_DIO, NULL); |
ccfb3000 JK |
848 | } |
849 | ||
e2b4e2bc | 850 | static int get_data_block_bmap(struct inode *inode, sector_t iblock, |
ccfb3000 JK |
851 | struct buffer_head *bh_result, int create) |
852 | { | |
179448bf | 853 | /* Block number less than F2FS MAX BLOCKS */ |
e0afc4d6 | 854 | if (unlikely(iblock >= F2FS_I_SB(inode)->max_file_blocks)) |
179448bf YH |
855 | return -EFBIG; |
856 | ||
e2b4e2bc | 857 | return __get_data_block(inode, iblock, bh_result, create, |
da85985c | 858 | F2FS_GET_BLOCK_BMAP, NULL); |
ccfb3000 JK |
859 | } |
860 | ||
7f63eb77 JK |
861 | static inline sector_t logical_to_blk(struct inode *inode, loff_t offset) |
862 | { | |
863 | return (offset >> inode->i_blkbits); | |
864 | } | |
865 | ||
866 | static inline loff_t blk_to_logical(struct inode *inode, sector_t blk) | |
867 | { | |
868 | return (blk << inode->i_blkbits); | |
869 | } | |
870 | ||
9ab70134 JK |
871 | int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, |
872 | u64 start, u64 len) | |
873 | { | |
7f63eb77 JK |
874 | struct buffer_head map_bh; |
875 | sector_t start_blk, last_blk; | |
da85985c | 876 | pgoff_t next_pgofs; |
de1475cc | 877 | loff_t isize; |
7f63eb77 JK |
878 | u64 logical = 0, phys = 0, size = 0; |
879 | u32 flags = 0; | |
7f63eb77 JK |
880 | int ret = 0; |
881 | ||
882 | ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC); | |
883 | if (ret) | |
884 | return ret; | |
885 | ||
67f8cf3c JK |
886 | if (f2fs_has_inline_data(inode)) { |
887 | ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len); | |
888 | if (ret != -EAGAIN) | |
889 | return ret; | |
890 | } | |
891 | ||
5955102c | 892 | inode_lock(inode); |
de1475cc FL |
893 | |
894 | isize = i_size_read(inode); | |
9a950d52 FL |
895 | if (start >= isize) |
896 | goto out; | |
7f63eb77 | 897 | |
9a950d52 FL |
898 | if (start + len > isize) |
899 | len = isize - start; | |
7f63eb77 JK |
900 | |
901 | if (logical_to_blk(inode, len) == 0) | |
902 | len = blk_to_logical(inode, 1); | |
903 | ||
904 | start_blk = logical_to_blk(inode, start); | |
905 | last_blk = logical_to_blk(inode, start + len - 1); | |
9a950d52 | 906 | |
7f63eb77 JK |
907 | next: |
908 | memset(&map_bh, 0, sizeof(struct buffer_head)); | |
909 | map_bh.b_size = len; | |
910 | ||
e2b4e2bc | 911 | ret = get_data_block(inode, start_blk, &map_bh, 0, |
da85985c | 912 | F2FS_GET_BLOCK_FIEMAP, &next_pgofs); |
7f63eb77 JK |
913 | if (ret) |
914 | goto out; | |
915 | ||
916 | /* HOLE */ | |
917 | if (!buffer_mapped(&map_bh)) { | |
da85985c | 918 | start_blk = next_pgofs; |
9a950d52 | 919 | /* Go through holes util pass the EOF */ |
da85985c | 920 | if (blk_to_logical(inode, start_blk) < isize) |
9a950d52 FL |
921 | goto prep_next; |
922 | /* Found a hole beyond isize means no more extents. | |
923 | * Note that the premise is that filesystems don't | |
924 | * punch holes beyond isize and keep size unchanged. | |
925 | */ | |
926 | flags |= FIEMAP_EXTENT_LAST; | |
927 | } | |
7f63eb77 | 928 | |
da5af127 CY |
929 | if (size) { |
930 | if (f2fs_encrypted_inode(inode)) | |
931 | flags |= FIEMAP_EXTENT_DATA_ENCRYPTED; | |
932 | ||
9a950d52 FL |
933 | ret = fiemap_fill_next_extent(fieinfo, logical, |
934 | phys, size, flags); | |
da5af127 | 935 | } |
7f63eb77 | 936 | |
9a950d52 FL |
937 | if (start_blk > last_blk || ret) |
938 | goto out; | |
7f63eb77 | 939 | |
9a950d52 FL |
940 | logical = blk_to_logical(inode, start_blk); |
941 | phys = blk_to_logical(inode, map_bh.b_blocknr); | |
942 | size = map_bh.b_size; | |
943 | flags = 0; | |
944 | if (buffer_unwritten(&map_bh)) | |
945 | flags = FIEMAP_EXTENT_UNWRITTEN; | |
7f63eb77 | 946 | |
9a950d52 | 947 | start_blk += logical_to_blk(inode, size); |
7f63eb77 | 948 | |
9a950d52 | 949 | prep_next: |
7f63eb77 JK |
950 | cond_resched(); |
951 | if (fatal_signal_pending(current)) | |
952 | ret = -EINTR; | |
953 | else | |
954 | goto next; | |
955 | out: | |
956 | if (ret == 1) | |
957 | ret = 0; | |
958 | ||
5955102c | 959 | inode_unlock(inode); |
7f63eb77 | 960 | return ret; |
9ab70134 JK |
961 | } |
962 | ||
78682f79 CY |
963 | struct bio *f2fs_grab_bio(struct inode *inode, block_t blkaddr, |
964 | unsigned nr_pages) | |
965 | { | |
966 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
967 | struct fscrypt_ctx *ctx = NULL; | |
968 | struct block_device *bdev = sbi->sb->s_bdev; | |
969 | struct bio *bio; | |
970 | ||
971 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) { | |
972 | ctx = fscrypt_get_ctx(inode, GFP_NOFS); | |
973 | if (IS_ERR(ctx)) | |
974 | return ERR_CAST(ctx); | |
975 | ||
976 | /* wait the page to be moved by cleaning */ | |
977 | f2fs_wait_on_encrypted_page_writeback(sbi, blkaddr); | |
978 | } | |
979 | ||
980 | bio = bio_alloc(GFP_KERNEL, min_t(int, nr_pages, BIO_MAX_PAGES)); | |
981 | if (!bio) { | |
982 | if (ctx) | |
983 | fscrypt_release_ctx(ctx); | |
984 | return ERR_PTR(-ENOMEM); | |
985 | } | |
986 | bio->bi_bdev = bdev; | |
987 | bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blkaddr); | |
988 | bio->bi_end_io = f2fs_read_end_io; | |
989 | bio->bi_private = ctx; | |
990 | ||
991 | return bio; | |
992 | } | |
993 | ||
f1e88660 JK |
994 | /* |
995 | * This function was originally taken from fs/mpage.c, and customized for f2fs. | |
996 | * Major change was from block_size == page_size in f2fs by default. | |
997 | */ | |
998 | static int f2fs_mpage_readpages(struct address_space *mapping, | |
999 | struct list_head *pages, struct page *page, | |
1000 | unsigned nr_pages) | |
1001 | { | |
1002 | struct bio *bio = NULL; | |
1003 | unsigned page_idx; | |
1004 | sector_t last_block_in_bio = 0; | |
1005 | struct inode *inode = mapping->host; | |
1006 | const unsigned blkbits = inode->i_blkbits; | |
1007 | const unsigned blocksize = 1 << blkbits; | |
1008 | sector_t block_in_file; | |
1009 | sector_t last_block; | |
1010 | sector_t last_block_in_file; | |
1011 | sector_t block_nr; | |
f1e88660 JK |
1012 | struct f2fs_map_blocks map; |
1013 | ||
1014 | map.m_pblk = 0; | |
1015 | map.m_lblk = 0; | |
1016 | map.m_len = 0; | |
1017 | map.m_flags = 0; | |
da85985c | 1018 | map.m_next_pgofs = NULL; |
f1e88660 JK |
1019 | |
1020 | for (page_idx = 0; nr_pages; page_idx++, nr_pages--) { | |
1021 | ||
1022 | prefetchw(&page->flags); | |
1023 | if (pages) { | |
1024 | page = list_entry(pages->prev, struct page, lru); | |
1025 | list_del(&page->lru); | |
1026 | if (add_to_page_cache_lru(page, mapping, | |
1027 | page->index, GFP_KERNEL)) | |
1028 | goto next_page; | |
1029 | } | |
1030 | ||
1031 | block_in_file = (sector_t)page->index; | |
1032 | last_block = block_in_file + nr_pages; | |
1033 | last_block_in_file = (i_size_read(inode) + blocksize - 1) >> | |
1034 | blkbits; | |
1035 | if (last_block > last_block_in_file) | |
1036 | last_block = last_block_in_file; | |
1037 | ||
1038 | /* | |
1039 | * Map blocks using the previous result first. | |
1040 | */ | |
1041 | if ((map.m_flags & F2FS_MAP_MAPPED) && | |
1042 | block_in_file > map.m_lblk && | |
1043 | block_in_file < (map.m_lblk + map.m_len)) | |
1044 | goto got_it; | |
1045 | ||
1046 | /* | |
1047 | * Then do more f2fs_map_blocks() calls until we are | |
1048 | * done with this page. | |
1049 | */ | |
1050 | map.m_flags = 0; | |
1051 | ||
1052 | if (block_in_file < last_block) { | |
1053 | map.m_lblk = block_in_file; | |
1054 | map.m_len = last_block - block_in_file; | |
1055 | ||
46c9e141 | 1056 | if (f2fs_map_blocks(inode, &map, 0, |
da85985c | 1057 | F2FS_GET_BLOCK_READ)) |
f1e88660 JK |
1058 | goto set_error_page; |
1059 | } | |
1060 | got_it: | |
1061 | if ((map.m_flags & F2FS_MAP_MAPPED)) { | |
1062 | block_nr = map.m_pblk + block_in_file - map.m_lblk; | |
1063 | SetPageMappedToDisk(page); | |
1064 | ||
1065 | if (!PageUptodate(page) && !cleancache_get_page(page)) { | |
1066 | SetPageUptodate(page); | |
1067 | goto confused; | |
1068 | } | |
1069 | } else { | |
09cbfeaf | 1070 | zero_user_segment(page, 0, PAGE_SIZE); |
237c0790 JK |
1071 | if (!PageUptodate(page)) |
1072 | SetPageUptodate(page); | |
f1e88660 JK |
1073 | unlock_page(page); |
1074 | goto next_page; | |
1075 | } | |
1076 | ||
1077 | /* | |
1078 | * This page will go to BIO. Do we need to send this | |
1079 | * BIO off first? | |
1080 | */ | |
1081 | if (bio && (last_block_in_bio != block_nr - 1)) { | |
1082 | submit_and_realloc: | |
19a5f5e2 | 1083 | __submit_bio(F2FS_I_SB(inode), READ, bio, DATA); |
f1e88660 JK |
1084 | bio = NULL; |
1085 | } | |
1086 | if (bio == NULL) { | |
78682f79 CY |
1087 | bio = f2fs_grab_bio(inode, block_nr, nr_pages); |
1088 | if (IS_ERR(bio)) | |
f1e88660 | 1089 | goto set_error_page; |
f1e88660 JK |
1090 | } |
1091 | ||
1092 | if (bio_add_page(bio, page, blocksize, 0) < blocksize) | |
1093 | goto submit_and_realloc; | |
1094 | ||
1095 | last_block_in_bio = block_nr; | |
1096 | goto next_page; | |
1097 | set_error_page: | |
1098 | SetPageError(page); | |
09cbfeaf | 1099 | zero_user_segment(page, 0, PAGE_SIZE); |
f1e88660 JK |
1100 | unlock_page(page); |
1101 | goto next_page; | |
1102 | confused: | |
1103 | if (bio) { | |
19a5f5e2 | 1104 | __submit_bio(F2FS_I_SB(inode), READ, bio, DATA); |
f1e88660 JK |
1105 | bio = NULL; |
1106 | } | |
1107 | unlock_page(page); | |
1108 | next_page: | |
1109 | if (pages) | |
09cbfeaf | 1110 | put_page(page); |
f1e88660 JK |
1111 | } |
1112 | BUG_ON(pages && !list_empty(pages)); | |
1113 | if (bio) | |
19a5f5e2 | 1114 | __submit_bio(F2FS_I_SB(inode), READ, bio, DATA); |
f1e88660 JK |
1115 | return 0; |
1116 | } | |
1117 | ||
eb47b800 JK |
1118 | static int f2fs_read_data_page(struct file *file, struct page *page) |
1119 | { | |
9ffe0fb5 | 1120 | struct inode *inode = page->mapping->host; |
b3d208f9 | 1121 | int ret = -EAGAIN; |
9ffe0fb5 | 1122 | |
c20e89cd CY |
1123 | trace_f2fs_readpage(page, DATA); |
1124 | ||
e1c42045 | 1125 | /* If the file has inline data, try to read it directly */ |
9ffe0fb5 HL |
1126 | if (f2fs_has_inline_data(inode)) |
1127 | ret = f2fs_read_inline_data(inode, page); | |
b3d208f9 | 1128 | if (ret == -EAGAIN) |
f1e88660 | 1129 | ret = f2fs_mpage_readpages(page->mapping, NULL, page, 1); |
9ffe0fb5 | 1130 | return ret; |
eb47b800 JK |
1131 | } |
1132 | ||
1133 | static int f2fs_read_data_pages(struct file *file, | |
1134 | struct address_space *mapping, | |
1135 | struct list_head *pages, unsigned nr_pages) | |
1136 | { | |
9ffe0fb5 | 1137 | struct inode *inode = file->f_mapping->host; |
b8c29400 CY |
1138 | struct page *page = list_entry(pages->prev, struct page, lru); |
1139 | ||
1140 | trace_f2fs_readpages(inode, page, nr_pages); | |
9ffe0fb5 HL |
1141 | |
1142 | /* If the file has inline data, skip readpages */ | |
1143 | if (f2fs_has_inline_data(inode)) | |
1144 | return 0; | |
1145 | ||
f1e88660 | 1146 | return f2fs_mpage_readpages(mapping, pages, NULL, nr_pages); |
eb47b800 JK |
1147 | } |
1148 | ||
05ca3632 | 1149 | int do_write_data_page(struct f2fs_io_info *fio) |
eb47b800 | 1150 | { |
05ca3632 | 1151 | struct page *page = fio->page; |
eb47b800 | 1152 | struct inode *inode = page->mapping->host; |
eb47b800 JK |
1153 | struct dnode_of_data dn; |
1154 | int err = 0; | |
1155 | ||
1156 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
266e97a8 | 1157 | err = get_dnode_of_data(&dn, page->index, LOOKUP_NODE); |
eb47b800 JK |
1158 | if (err) |
1159 | return err; | |
1160 | ||
28bc106b | 1161 | fio->old_blkaddr = dn.data_blkaddr; |
eb47b800 JK |
1162 | |
1163 | /* This page is already truncated */ | |
7a9d7548 | 1164 | if (fio->old_blkaddr == NULL_ADDR) { |
2bca1e23 | 1165 | ClearPageUptodate(page); |
eb47b800 | 1166 | goto out_writepage; |
2bca1e23 | 1167 | } |
eb47b800 | 1168 | |
4375a336 | 1169 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) { |
b32e4482 | 1170 | gfp_t gfp_flags = GFP_NOFS; |
08b39fbd CY |
1171 | |
1172 | /* wait for GCed encrypted page writeback */ | |
1173 | f2fs_wait_on_encrypted_page_writeback(F2FS_I_SB(inode), | |
7a9d7548 | 1174 | fio->old_blkaddr); |
b32e4482 JK |
1175 | retry_encrypt: |
1176 | fio->encrypted_page = fscrypt_encrypt_page(inode, fio->page, | |
1177 | gfp_flags); | |
4375a336 JK |
1178 | if (IS_ERR(fio->encrypted_page)) { |
1179 | err = PTR_ERR(fio->encrypted_page); | |
b32e4482 JK |
1180 | if (err == -ENOMEM) { |
1181 | /* flush pending ios and wait for a while */ | |
1182 | f2fs_flush_merged_bios(F2FS_I_SB(inode)); | |
1183 | congestion_wait(BLK_RW_ASYNC, HZ/50); | |
1184 | gfp_flags |= __GFP_NOFAIL; | |
1185 | err = 0; | |
1186 | goto retry_encrypt; | |
1187 | } | |
4375a336 JK |
1188 | goto out_writepage; |
1189 | } | |
1190 | } | |
1191 | ||
eb47b800 JK |
1192 | set_page_writeback(page); |
1193 | ||
1194 | /* | |
1195 | * If current allocation needs SSR, | |
1196 | * it had better in-place writes for updated data. | |
1197 | */ | |
7a9d7548 | 1198 | if (unlikely(fio->old_blkaddr != NEW_ADDR && |
b25958b6 | 1199 | !is_cold_data(page) && |
2da3e027 | 1200 | !IS_ATOMIC_WRITTEN_PAGE(page) && |
b25958b6 | 1201 | need_inplace_update(inode))) { |
05ca3632 | 1202 | rewrite_data_page(fio); |
91942321 | 1203 | set_inode_flag(inode, FI_UPDATE_WRITE); |
8ce67cb0 | 1204 | trace_f2fs_do_write_data_page(page, IPU); |
eb47b800 | 1205 | } else { |
05ca3632 | 1206 | write_data_page(&dn, fio); |
8ce67cb0 | 1207 | trace_f2fs_do_write_data_page(page, OPU); |
91942321 | 1208 | set_inode_flag(inode, FI_APPEND_WRITE); |
3c6c2beb | 1209 | if (page->index == 0) |
91942321 | 1210 | set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN); |
eb47b800 JK |
1211 | } |
1212 | out_writepage: | |
1213 | f2fs_put_dnode(&dn); | |
1214 | return err; | |
1215 | } | |
1216 | ||
1217 | static int f2fs_write_data_page(struct page *page, | |
1218 | struct writeback_control *wbc) | |
1219 | { | |
1220 | struct inode *inode = page->mapping->host; | |
4081363f | 1221 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
eb47b800 JK |
1222 | loff_t i_size = i_size_read(inode); |
1223 | const pgoff_t end_index = ((unsigned long long) i_size) | |
09cbfeaf | 1224 | >> PAGE_SHIFT; |
26de9b11 | 1225 | loff_t psize = (page->index + 1) << PAGE_SHIFT; |
9ffe0fb5 | 1226 | unsigned offset = 0; |
39936837 | 1227 | bool need_balance_fs = false; |
eb47b800 | 1228 | int err = 0; |
458e6197 | 1229 | struct f2fs_io_info fio = { |
05ca3632 | 1230 | .sbi = sbi, |
458e6197 | 1231 | .type = DATA, |
6c311ec6 | 1232 | .rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE, |
05ca3632 | 1233 | .page = page, |
4375a336 | 1234 | .encrypted_page = NULL, |
458e6197 | 1235 | }; |
eb47b800 | 1236 | |
ecda0de3 CY |
1237 | trace_f2fs_writepage(page, DATA); |
1238 | ||
eb47b800 | 1239 | if (page->index < end_index) |
39936837 | 1240 | goto write; |
eb47b800 JK |
1241 | |
1242 | /* | |
1243 | * If the offset is out-of-range of file size, | |
1244 | * this page does not have to be written to disk. | |
1245 | */ | |
09cbfeaf | 1246 | offset = i_size & (PAGE_SIZE - 1); |
76f60268 | 1247 | if ((page->index >= end_index + 1) || !offset) |
39936837 | 1248 | goto out; |
eb47b800 | 1249 | |
09cbfeaf | 1250 | zero_user_segment(page, offset, PAGE_SIZE); |
39936837 | 1251 | write: |
caf0047e | 1252 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
eb47b800 | 1253 | goto redirty_out; |
1e84371f JK |
1254 | if (f2fs_is_drop_cache(inode)) |
1255 | goto out; | |
e6e5f561 JK |
1256 | /* we should not write 0'th page having journal header */ |
1257 | if (f2fs_is_volatile_file(inode) && (!page->index || | |
1258 | (!wbc->for_reclaim && | |
1259 | available_free_memory(sbi, BASE_CHECK)))) | |
1e84371f | 1260 | goto redirty_out; |
eb47b800 | 1261 | |
cf779cab JK |
1262 | /* we should bypass data pages to proceed the kworkder jobs */ |
1263 | if (unlikely(f2fs_cp_error(sbi))) { | |
7f319975 | 1264 | mapping_set_error(page->mapping, -EIO); |
a7ffdbe2 | 1265 | goto out; |
cf779cab JK |
1266 | } |
1267 | ||
b230e6ca JK |
1268 | /* Dentry blocks are controlled by checkpoint */ |
1269 | if (S_ISDIR(inode->i_mode)) { | |
1270 | err = do_write_data_page(&fio); | |
1271 | goto done; | |
1272 | } | |
1273 | ||
8618b881 | 1274 | if (!wbc->for_reclaim) |
39936837 | 1275 | need_balance_fs = true; |
8618b881 | 1276 | else if (has_not_enough_free_secs(sbi, 0)) |
39936837 | 1277 | goto redirty_out; |
eb47b800 | 1278 | |
b3d208f9 | 1279 | err = -EAGAIN; |
8618b881 | 1280 | f2fs_lock_op(sbi); |
b3d208f9 JK |
1281 | if (f2fs_has_inline_data(inode)) |
1282 | err = f2fs_write_inline_data(inode, page); | |
1283 | if (err == -EAGAIN) | |
05ca3632 | 1284 | err = do_write_data_page(&fio); |
26de9b11 JK |
1285 | if (F2FS_I(inode)->last_disk_size < psize) |
1286 | F2FS_I(inode)->last_disk_size = psize; | |
8618b881 JK |
1287 | f2fs_unlock_op(sbi); |
1288 | done: | |
1289 | if (err && err != -ENOENT) | |
1290 | goto redirty_out; | |
eb47b800 | 1291 | |
eb47b800 | 1292 | clear_cold_data(page); |
39936837 | 1293 | out: |
a7ffdbe2 | 1294 | inode_dec_dirty_pages(inode); |
2bca1e23 JK |
1295 | if (err) |
1296 | ClearPageUptodate(page); | |
0c3a5797 CY |
1297 | |
1298 | if (wbc->for_reclaim) { | |
1299 | f2fs_submit_merged_bio_cond(sbi, NULL, page, 0, DATA, WRITE); | |
1300 | remove_dirty_inode(inode); | |
1301 | } | |
1302 | ||
eb47b800 | 1303 | unlock_page(page); |
2c4db1a6 | 1304 | f2fs_balance_fs(sbi, need_balance_fs); |
0c3a5797 CY |
1305 | |
1306 | if (unlikely(f2fs_cp_error(sbi))) | |
2aea39ec | 1307 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
0c3a5797 | 1308 | |
eb47b800 JK |
1309 | return 0; |
1310 | ||
eb47b800 | 1311 | redirty_out: |
76f60268 | 1312 | redirty_page_for_writepage(wbc, page); |
b230e6ca JK |
1313 | unlock_page(page); |
1314 | return err; | |
fa9150a8 NJ |
1315 | } |
1316 | ||
8f46dcae CY |
1317 | /* |
1318 | * This function was copied from write_cche_pages from mm/page-writeback.c. | |
1319 | * The major change is making write step of cold data page separately from | |
1320 | * warm/hot data page. | |
1321 | */ | |
1322 | static int f2fs_write_cache_pages(struct address_space *mapping, | |
b230e6ca | 1323 | struct writeback_control *wbc) |
8f46dcae CY |
1324 | { |
1325 | int ret = 0; | |
1326 | int done = 0; | |
1327 | struct pagevec pvec; | |
1328 | int nr_pages; | |
1329 | pgoff_t uninitialized_var(writeback_index); | |
1330 | pgoff_t index; | |
1331 | pgoff_t end; /* Inclusive */ | |
1332 | pgoff_t done_index; | |
1333 | int cycled; | |
1334 | int range_whole = 0; | |
1335 | int tag; | |
8f46dcae CY |
1336 | |
1337 | pagevec_init(&pvec, 0); | |
46ae957f | 1338 | |
8f46dcae CY |
1339 | if (wbc->range_cyclic) { |
1340 | writeback_index = mapping->writeback_index; /* prev offset */ | |
1341 | index = writeback_index; | |
1342 | if (index == 0) | |
1343 | cycled = 1; | |
1344 | else | |
1345 | cycled = 0; | |
1346 | end = -1; | |
1347 | } else { | |
09cbfeaf KS |
1348 | index = wbc->range_start >> PAGE_SHIFT; |
1349 | end = wbc->range_end >> PAGE_SHIFT; | |
8f46dcae CY |
1350 | if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX) |
1351 | range_whole = 1; | |
1352 | cycled = 1; /* ignore range_cyclic tests */ | |
1353 | } | |
1354 | if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages) | |
1355 | tag = PAGECACHE_TAG_TOWRITE; | |
1356 | else | |
1357 | tag = PAGECACHE_TAG_DIRTY; | |
1358 | retry: | |
1359 | if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages) | |
1360 | tag_pages_for_writeback(mapping, index, end); | |
1361 | done_index = index; | |
1362 | while (!done && (index <= end)) { | |
1363 | int i; | |
1364 | ||
1365 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag, | |
1366 | min(end - index, (pgoff_t)PAGEVEC_SIZE - 1) + 1); | |
1367 | if (nr_pages == 0) | |
1368 | break; | |
1369 | ||
1370 | for (i = 0; i < nr_pages; i++) { | |
1371 | struct page *page = pvec.pages[i]; | |
1372 | ||
1373 | if (page->index > end) { | |
1374 | done = 1; | |
1375 | break; | |
1376 | } | |
1377 | ||
1378 | done_index = page->index; | |
1379 | ||
1380 | lock_page(page); | |
1381 | ||
1382 | if (unlikely(page->mapping != mapping)) { | |
1383 | continue_unlock: | |
1384 | unlock_page(page); | |
1385 | continue; | |
1386 | } | |
1387 | ||
1388 | if (!PageDirty(page)) { | |
1389 | /* someone wrote it for us */ | |
1390 | goto continue_unlock; | |
1391 | } | |
1392 | ||
8f46dcae CY |
1393 | if (PageWriteback(page)) { |
1394 | if (wbc->sync_mode != WB_SYNC_NONE) | |
fec1d657 JK |
1395 | f2fs_wait_on_page_writeback(page, |
1396 | DATA, true); | |
8f46dcae CY |
1397 | else |
1398 | goto continue_unlock; | |
1399 | } | |
1400 | ||
1401 | BUG_ON(PageWriteback(page)); | |
1402 | if (!clear_page_dirty_for_io(page)) | |
1403 | goto continue_unlock; | |
1404 | ||
b230e6ca | 1405 | ret = mapping->a_ops->writepage(page, wbc); |
8f46dcae | 1406 | if (unlikely(ret)) { |
b230e6ca JK |
1407 | done_index = page->index + 1; |
1408 | done = 1; | |
1409 | break; | |
8f46dcae CY |
1410 | } |
1411 | ||
1412 | if (--wbc->nr_to_write <= 0 && | |
1413 | wbc->sync_mode == WB_SYNC_NONE) { | |
1414 | done = 1; | |
1415 | break; | |
1416 | } | |
1417 | } | |
1418 | pagevec_release(&pvec); | |
1419 | cond_resched(); | |
1420 | } | |
1421 | ||
8f46dcae CY |
1422 | if (!cycled && !done) { |
1423 | cycled = 1; | |
1424 | index = 0; | |
1425 | end = writeback_index - 1; | |
1426 | goto retry; | |
1427 | } | |
1428 | if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) | |
1429 | mapping->writeback_index = done_index; | |
1430 | ||
1431 | return ret; | |
1432 | } | |
1433 | ||
25ca923b | 1434 | static int f2fs_write_data_pages(struct address_space *mapping, |
eb47b800 JK |
1435 | struct writeback_control *wbc) |
1436 | { | |
1437 | struct inode *inode = mapping->host; | |
4081363f | 1438 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
eb47b800 | 1439 | int ret; |
eb47b800 | 1440 | |
cfb185a1 | 1441 | /* deal with chardevs and other special file */ |
1442 | if (!mapping->a_ops->writepage) | |
1443 | return 0; | |
1444 | ||
6a290544 CY |
1445 | /* skip writing if there is no dirty page in this inode */ |
1446 | if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE) | |
1447 | return 0; | |
1448 | ||
a1257023 JK |
1449 | if (S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_NONE && |
1450 | get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) && | |
1451 | available_free_memory(sbi, DIRTY_DENTS)) | |
1452 | goto skip_write; | |
1453 | ||
d323d005 | 1454 | /* skip writing during file defragment */ |
91942321 | 1455 | if (is_inode_flag_set(inode, FI_DO_DEFRAG)) |
d323d005 CY |
1456 | goto skip_write; |
1457 | ||
d5669f7b JK |
1458 | /* during POR, we don't need to trigger writepage at all. */ |
1459 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) | |
1460 | goto skip_write; | |
1461 | ||
d31c7c3f YH |
1462 | trace_f2fs_writepages(mapping->host, wbc, DATA); |
1463 | ||
b230e6ca | 1464 | ret = f2fs_write_cache_pages(mapping, wbc); |
28ea6162 JK |
1465 | /* |
1466 | * if some pages were truncated, we cannot guarantee its mapping->host | |
1467 | * to detect pending bios. | |
1468 | */ | |
1469 | f2fs_submit_merged_bio(sbi, DATA, WRITE); | |
458e6197 | 1470 | |
c227f912 | 1471 | remove_dirty_inode(inode); |
eb47b800 | 1472 | return ret; |
d3baf95d JK |
1473 | |
1474 | skip_write: | |
a7ffdbe2 | 1475 | wbc->pages_skipped += get_dirty_pages(inode); |
d31c7c3f | 1476 | trace_f2fs_writepages(mapping->host, wbc, DATA); |
d3baf95d | 1477 | return 0; |
eb47b800 JK |
1478 | } |
1479 | ||
3aab8f82 CY |
1480 | static void f2fs_write_failed(struct address_space *mapping, loff_t to) |
1481 | { | |
1482 | struct inode *inode = mapping->host; | |
819d9153 | 1483 | loff_t i_size = i_size_read(inode); |
3aab8f82 | 1484 | |
819d9153 JK |
1485 | if (to > i_size) { |
1486 | truncate_pagecache(inode, i_size); | |
1487 | truncate_blocks(inode, i_size, true); | |
3aab8f82 CY |
1488 | } |
1489 | } | |
1490 | ||
2aadac08 JK |
1491 | static int prepare_write_begin(struct f2fs_sb_info *sbi, |
1492 | struct page *page, loff_t pos, unsigned len, | |
1493 | block_t *blk_addr, bool *node_changed) | |
1494 | { | |
1495 | struct inode *inode = page->mapping->host; | |
1496 | pgoff_t index = page->index; | |
1497 | struct dnode_of_data dn; | |
1498 | struct page *ipage; | |
b4d07a3e JK |
1499 | bool locked = false; |
1500 | struct extent_info ei; | |
2aadac08 JK |
1501 | int err = 0; |
1502 | ||
24b84912 JK |
1503 | /* |
1504 | * we already allocated all the blocks, so we don't need to get | |
1505 | * the block addresses when there is no need to fill the page. | |
1506 | */ | |
1507 | if (!f2fs_has_inline_data(inode) && !f2fs_encrypted_inode(inode) && | |
09cbfeaf | 1508 | len == PAGE_SIZE) |
24b84912 JK |
1509 | return 0; |
1510 | ||
b4d07a3e | 1511 | if (f2fs_has_inline_data(inode) || |
09cbfeaf | 1512 | (pos & PAGE_MASK) >= i_size_read(inode)) { |
b4d07a3e JK |
1513 | f2fs_lock_op(sbi); |
1514 | locked = true; | |
1515 | } | |
1516 | restart: | |
2aadac08 JK |
1517 | /* check inline_data */ |
1518 | ipage = get_node_page(sbi, inode->i_ino); | |
1519 | if (IS_ERR(ipage)) { | |
1520 | err = PTR_ERR(ipage); | |
1521 | goto unlock_out; | |
1522 | } | |
1523 | ||
1524 | set_new_dnode(&dn, inode, ipage, ipage, 0); | |
1525 | ||
1526 | if (f2fs_has_inline_data(inode)) { | |
1527 | if (pos + len <= MAX_INLINE_DATA) { | |
1528 | read_inline_data(page, ipage); | |
91942321 | 1529 | set_inode_flag(inode, FI_DATA_EXIST); |
ab47036d CY |
1530 | if (inode->i_nlink) |
1531 | set_inline_node(ipage); | |
2aadac08 JK |
1532 | } else { |
1533 | err = f2fs_convert_inline_page(&dn, page); | |
1534 | if (err) | |
b4d07a3e JK |
1535 | goto out; |
1536 | if (dn.data_blkaddr == NULL_ADDR) | |
1537 | err = f2fs_get_block(&dn, index); | |
1538 | } | |
1539 | } else if (locked) { | |
1540 | err = f2fs_get_block(&dn, index); | |
1541 | } else { | |
1542 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { | |
1543 | dn.data_blkaddr = ei.blk + index - ei.fofs; | |
1544 | } else { | |
b4d07a3e JK |
1545 | /* hole case */ |
1546 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); | |
4da7bf5a | 1547 | if (err || dn.data_blkaddr == NULL_ADDR) { |
b4d07a3e JK |
1548 | f2fs_put_dnode(&dn); |
1549 | f2fs_lock_op(sbi); | |
1550 | locked = true; | |
1551 | goto restart; | |
1552 | } | |
2aadac08 JK |
1553 | } |
1554 | } | |
b4d07a3e | 1555 | |
2aadac08 JK |
1556 | /* convert_inline_page can make node_changed */ |
1557 | *blk_addr = dn.data_blkaddr; | |
1558 | *node_changed = dn.node_changed; | |
b4d07a3e | 1559 | out: |
2aadac08 JK |
1560 | f2fs_put_dnode(&dn); |
1561 | unlock_out: | |
b4d07a3e JK |
1562 | if (locked) |
1563 | f2fs_unlock_op(sbi); | |
2aadac08 JK |
1564 | return err; |
1565 | } | |
1566 | ||
eb47b800 JK |
1567 | static int f2fs_write_begin(struct file *file, struct address_space *mapping, |
1568 | loff_t pos, unsigned len, unsigned flags, | |
1569 | struct page **pagep, void **fsdata) | |
1570 | { | |
1571 | struct inode *inode = mapping->host; | |
4081363f | 1572 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
86531d6b | 1573 | struct page *page = NULL; |
09cbfeaf | 1574 | pgoff_t index = ((unsigned long long) pos) >> PAGE_SHIFT; |
2aadac08 JK |
1575 | bool need_balance = false; |
1576 | block_t blkaddr = NULL_ADDR; | |
eb47b800 JK |
1577 | int err = 0; |
1578 | ||
62aed044 CY |
1579 | trace_f2fs_write_begin(inode, pos, len, flags); |
1580 | ||
5f727395 JK |
1581 | /* |
1582 | * We should check this at this moment to avoid deadlock on inode page | |
1583 | * and #0 page. The locking rule for inline_data conversion should be: | |
1584 | * lock_page(page #0) -> lock_page(inode_page) | |
1585 | */ | |
1586 | if (index != 0) { | |
1587 | err = f2fs_convert_inline_inode(inode); | |
1588 | if (err) | |
1589 | goto fail; | |
1590 | } | |
afcb7ca0 | 1591 | repeat: |
eb47b800 | 1592 | page = grab_cache_page_write_begin(mapping, index, flags); |
3aab8f82 CY |
1593 | if (!page) { |
1594 | err = -ENOMEM; | |
1595 | goto fail; | |
1596 | } | |
d5f66990 | 1597 | |
eb47b800 JK |
1598 | *pagep = page; |
1599 | ||
2aadac08 JK |
1600 | err = prepare_write_begin(sbi, page, pos, len, |
1601 | &blkaddr, &need_balance); | |
9ba69cf9 | 1602 | if (err) |
2aadac08 | 1603 | goto fail; |
9ba69cf9 | 1604 | |
2aadac08 | 1605 | if (need_balance && has_not_enough_free_secs(sbi, 0)) { |
2a340760 | 1606 | unlock_page(page); |
2c4db1a6 | 1607 | f2fs_balance_fs(sbi, true); |
2a340760 JK |
1608 | lock_page(page); |
1609 | if (page->mapping != mapping) { | |
1610 | /* The page got truncated from under us */ | |
1611 | f2fs_put_page(page, 1); | |
1612 | goto repeat; | |
1613 | } | |
1614 | } | |
1615 | ||
fec1d657 | 1616 | f2fs_wait_on_page_writeback(page, DATA, false); |
b3d208f9 | 1617 | |
08b39fbd CY |
1618 | /* wait for GCed encrypted page writeback */ |
1619 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) | |
2aadac08 | 1620 | f2fs_wait_on_encrypted_page_writeback(sbi, blkaddr); |
08b39fbd | 1621 | |
09cbfeaf | 1622 | if (len == PAGE_SIZE) |
90d4388a CY |
1623 | goto out_update; |
1624 | if (PageUptodate(page)) | |
1625 | goto out_clear; | |
eb47b800 | 1626 | |
09cbfeaf KS |
1627 | if ((pos & PAGE_MASK) >= i_size_read(inode)) { |
1628 | unsigned start = pos & (PAGE_SIZE - 1); | |
eb47b800 JK |
1629 | unsigned end = start + len; |
1630 | ||
1631 | /* Reading beyond i_size is simple: memset to zero */ | |
09cbfeaf | 1632 | zero_user_segments(page, 0, start, end, PAGE_SIZE); |
90d4388a | 1633 | goto out_update; |
eb47b800 JK |
1634 | } |
1635 | ||
2aadac08 | 1636 | if (blkaddr == NEW_ADDR) { |
09cbfeaf | 1637 | zero_user_segment(page, 0, PAGE_SIZE); |
eb47b800 | 1638 | } else { |
78682f79 CY |
1639 | struct bio *bio; |
1640 | ||
1641 | bio = f2fs_grab_bio(inode, blkaddr, 1); | |
1642 | if (IS_ERR(bio)) { | |
1643 | err = PTR_ERR(bio); | |
9234f319 | 1644 | goto fail; |
78682f79 CY |
1645 | } |
1646 | ||
1647 | if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) { | |
1648 | bio_put(bio); | |
1649 | err = -EFAULT; | |
1650 | goto fail; | |
1651 | } | |
1652 | ||
1653 | __submit_bio(sbi, READ_SYNC, bio, DATA); | |
d54c795b | 1654 | |
393ff91f | 1655 | lock_page(page); |
6bacf52f | 1656 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
1657 | f2fs_put_page(page, 1); |
1658 | goto repeat; | |
eb47b800 | 1659 | } |
1563ac75 CY |
1660 | if (unlikely(!PageUptodate(page))) { |
1661 | err = -EIO; | |
1662 | goto fail; | |
1663 | } | |
eb47b800 | 1664 | } |
90d4388a | 1665 | out_update: |
237c0790 JK |
1666 | if (!PageUptodate(page)) |
1667 | SetPageUptodate(page); | |
90d4388a | 1668 | out_clear: |
eb47b800 JK |
1669 | clear_cold_data(page); |
1670 | return 0; | |
9ba69cf9 | 1671 | |
3aab8f82 | 1672 | fail: |
86531d6b | 1673 | f2fs_put_page(page, 1); |
3aab8f82 CY |
1674 | f2fs_write_failed(mapping, pos + len); |
1675 | return err; | |
eb47b800 JK |
1676 | } |
1677 | ||
a1dd3c13 JK |
1678 | static int f2fs_write_end(struct file *file, |
1679 | struct address_space *mapping, | |
1680 | loff_t pos, unsigned len, unsigned copied, | |
1681 | struct page *page, void *fsdata) | |
1682 | { | |
1683 | struct inode *inode = page->mapping->host; | |
1684 | ||
dfb2bf38 CY |
1685 | trace_f2fs_write_end(inode, pos, len, copied); |
1686 | ||
34ba94ba | 1687 | set_page_dirty(page); |
a1dd3c13 | 1688 | |
fc9581c8 JK |
1689 | if (pos + copied > i_size_read(inode)) |
1690 | f2fs_i_size_write(inode, pos + copied); | |
a1dd3c13 | 1691 | |
75c3c8bc | 1692 | f2fs_put_page(page, 1); |
d0239e1b | 1693 | f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); |
a1dd3c13 JK |
1694 | return copied; |
1695 | } | |
1696 | ||
6f673763 OS |
1697 | static int check_direct_IO(struct inode *inode, struct iov_iter *iter, |
1698 | loff_t offset) | |
944fcfc1 JK |
1699 | { |
1700 | unsigned blocksize_mask = inode->i_sb->s_blocksize - 1; | |
944fcfc1 | 1701 | |
944fcfc1 JK |
1702 | if (offset & blocksize_mask) |
1703 | return -EINVAL; | |
1704 | ||
5b46f25d AV |
1705 | if (iov_iter_alignment(iter) & blocksize_mask) |
1706 | return -EINVAL; | |
1707 | ||
944fcfc1 JK |
1708 | return 0; |
1709 | } | |
1710 | ||
c8b8e32d | 1711 | static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter) |
eb47b800 | 1712 | { |
b439b103 | 1713 | struct address_space *mapping = iocb->ki_filp->f_mapping; |
3aab8f82 CY |
1714 | struct inode *inode = mapping->host; |
1715 | size_t count = iov_iter_count(iter); | |
c8b8e32d | 1716 | loff_t offset = iocb->ki_pos; |
3aab8f82 | 1717 | int err; |
944fcfc1 | 1718 | |
b439b103 | 1719 | err = check_direct_IO(inode, iter, offset); |
b9d777b8 JK |
1720 | if (err) |
1721 | return err; | |
9ffe0fb5 | 1722 | |
fcc85a4d JK |
1723 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) |
1724 | return 0; | |
36abef4e JK |
1725 | if (test_opt(F2FS_I_SB(inode), LFS)) |
1726 | return 0; | |
fcc85a4d | 1727 | |
6f673763 | 1728 | trace_f2fs_direct_IO_enter(inode, offset, count, iov_iter_rw(iter)); |
70407fad | 1729 | |
c8b8e32d | 1730 | err = blockdev_direct_IO(iocb, inode, iter, get_data_block_dio); |
6bfc4919 JK |
1731 | if (iov_iter_rw(iter) == WRITE) { |
1732 | if (err > 0) | |
91942321 | 1733 | set_inode_flag(inode, FI_UPDATE_WRITE); |
6bfc4919 JK |
1734 | else if (err < 0) |
1735 | f2fs_write_failed(mapping, offset + count); | |
1736 | } | |
70407fad | 1737 | |
6f673763 | 1738 | trace_f2fs_direct_IO_exit(inode, offset, count, iov_iter_rw(iter), err); |
70407fad | 1739 | |
3aab8f82 | 1740 | return err; |
eb47b800 JK |
1741 | } |
1742 | ||
487261f3 CY |
1743 | void f2fs_invalidate_page(struct page *page, unsigned int offset, |
1744 | unsigned int length) | |
eb47b800 JK |
1745 | { |
1746 | struct inode *inode = page->mapping->host; | |
487261f3 | 1747 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
a7ffdbe2 | 1748 | |
487261f3 | 1749 | if (inode->i_ino >= F2FS_ROOT_INO(sbi) && |
09cbfeaf | 1750 | (offset % PAGE_SIZE || length != PAGE_SIZE)) |
a7ffdbe2 JK |
1751 | return; |
1752 | ||
487261f3 CY |
1753 | if (PageDirty(page)) { |
1754 | if (inode->i_ino == F2FS_META_INO(sbi)) | |
1755 | dec_page_count(sbi, F2FS_DIRTY_META); | |
1756 | else if (inode->i_ino == F2FS_NODE_INO(sbi)) | |
1757 | dec_page_count(sbi, F2FS_DIRTY_NODES); | |
1758 | else | |
1759 | inode_dec_dirty_pages(inode); | |
1760 | } | |
decd36b6 CY |
1761 | |
1762 | /* This is atomic written page, keep Private */ | |
1763 | if (IS_ATOMIC_WRITTEN_PAGE(page)) | |
1764 | return; | |
1765 | ||
23dc974e | 1766 | set_page_private(page, 0); |
eb47b800 JK |
1767 | ClearPagePrivate(page); |
1768 | } | |
1769 | ||
487261f3 | 1770 | int f2fs_release_page(struct page *page, gfp_t wait) |
eb47b800 | 1771 | { |
f68daeeb JK |
1772 | /* If this is dirty page, keep PagePrivate */ |
1773 | if (PageDirty(page)) | |
1774 | return 0; | |
1775 | ||
decd36b6 CY |
1776 | /* This is atomic written page, keep Private */ |
1777 | if (IS_ATOMIC_WRITTEN_PAGE(page)) | |
1778 | return 0; | |
1779 | ||
23dc974e | 1780 | set_page_private(page, 0); |
eb47b800 | 1781 | ClearPagePrivate(page); |
c3850aa1 | 1782 | return 1; |
eb47b800 JK |
1783 | } |
1784 | ||
fe76b796 JK |
1785 | /* |
1786 | * This was copied from __set_page_dirty_buffers which gives higher performance | |
1787 | * in very high speed storages. (e.g., pmem) | |
1788 | */ | |
1789 | void f2fs_set_page_dirty_nobuffers(struct page *page) | |
1790 | { | |
1791 | struct address_space *mapping = page->mapping; | |
1792 | unsigned long flags; | |
1793 | ||
1794 | if (unlikely(!mapping)) | |
1795 | return; | |
1796 | ||
1797 | spin_lock(&mapping->private_lock); | |
1798 | lock_page_memcg(page); | |
1799 | SetPageDirty(page); | |
1800 | spin_unlock(&mapping->private_lock); | |
1801 | ||
1802 | spin_lock_irqsave(&mapping->tree_lock, flags); | |
1803 | WARN_ON_ONCE(!PageUptodate(page)); | |
1804 | account_page_dirtied(page, mapping); | |
1805 | radix_tree_tag_set(&mapping->page_tree, | |
1806 | page_index(page), PAGECACHE_TAG_DIRTY); | |
1807 | spin_unlock_irqrestore(&mapping->tree_lock, flags); | |
1808 | unlock_page_memcg(page); | |
1809 | ||
1810 | __mark_inode_dirty(mapping->host, I_DIRTY_PAGES); | |
1811 | return; | |
1812 | } | |
1813 | ||
eb47b800 JK |
1814 | static int f2fs_set_data_page_dirty(struct page *page) |
1815 | { | |
1816 | struct address_space *mapping = page->mapping; | |
1817 | struct inode *inode = mapping->host; | |
1818 | ||
26c6b887 JK |
1819 | trace_f2fs_set_page_dirty(page, DATA); |
1820 | ||
237c0790 JK |
1821 | if (!PageUptodate(page)) |
1822 | SetPageUptodate(page); | |
34ba94ba | 1823 | |
1e84371f | 1824 | if (f2fs_is_atomic_file(inode)) { |
decd36b6 CY |
1825 | if (!IS_ATOMIC_WRITTEN_PAGE(page)) { |
1826 | register_inmem_page(inode, page); | |
1827 | return 1; | |
1828 | } | |
1829 | /* | |
1830 | * Previously, this page has been registered, we just | |
1831 | * return here. | |
1832 | */ | |
1833 | return 0; | |
34ba94ba JK |
1834 | } |
1835 | ||
eb47b800 | 1836 | if (!PageDirty(page)) { |
fe76b796 | 1837 | f2fs_set_page_dirty_nobuffers(page); |
a7ffdbe2 | 1838 | update_dirty_page(inode, page); |
eb47b800 JK |
1839 | return 1; |
1840 | } | |
1841 | return 0; | |
1842 | } | |
1843 | ||
c01e54b7 JK |
1844 | static sector_t f2fs_bmap(struct address_space *mapping, sector_t block) |
1845 | { | |
454ae7e5 CY |
1846 | struct inode *inode = mapping->host; |
1847 | ||
1d373a0e JK |
1848 | if (f2fs_has_inline_data(inode)) |
1849 | return 0; | |
1850 | ||
1851 | /* make sure allocating whole blocks */ | |
1852 | if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) | |
1853 | filemap_write_and_wait(mapping); | |
1854 | ||
e2b4e2bc | 1855 | return generic_block_bmap(mapping, block, get_data_block_bmap); |
429511cd CY |
1856 | } |
1857 | ||
eb47b800 JK |
1858 | const struct address_space_operations f2fs_dblock_aops = { |
1859 | .readpage = f2fs_read_data_page, | |
1860 | .readpages = f2fs_read_data_pages, | |
1861 | .writepage = f2fs_write_data_page, | |
1862 | .writepages = f2fs_write_data_pages, | |
1863 | .write_begin = f2fs_write_begin, | |
a1dd3c13 | 1864 | .write_end = f2fs_write_end, |
eb47b800 | 1865 | .set_page_dirty = f2fs_set_data_page_dirty, |
487261f3 CY |
1866 | .invalidatepage = f2fs_invalidate_page, |
1867 | .releasepage = f2fs_release_page, | |
eb47b800 | 1868 | .direct_IO = f2fs_direct_IO, |
c01e54b7 | 1869 | .bmap = f2fs_bmap, |
eb47b800 | 1870 | }; |