]>
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 | |
36951b38 CY |
32 | static bool __is_cp_guaranteed(struct page *page) |
33 | { | |
34 | struct address_space *mapping = page->mapping; | |
35 | struct inode *inode; | |
36 | struct f2fs_sb_info *sbi; | |
37 | ||
38 | if (!mapping) | |
39 | return false; | |
40 | ||
41 | inode = mapping->host; | |
42 | sbi = F2FS_I_SB(inode); | |
43 | ||
44 | if (inode->i_ino == F2FS_META_INO(sbi) || | |
45 | inode->i_ino == F2FS_NODE_INO(sbi) || | |
46 | S_ISDIR(inode->i_mode) || | |
47 | is_cold_data(page)) | |
48 | return true; | |
49 | return false; | |
50 | } | |
51 | ||
4246a0b6 | 52 | static void f2fs_read_end_io(struct bio *bio) |
93dfe2ac | 53 | { |
f568849e LT |
54 | struct bio_vec *bvec; |
55 | int i; | |
93dfe2ac | 56 | |
8b038c70 | 57 | #ifdef CONFIG_F2FS_FAULT_INJECTION |
1ecc0c5c | 58 | if (time_to_inject(F2FS_P_SB(bio->bi_io_vec->bv_page), FAULT_IO)) |
8b038c70 CY |
59 | bio->bi_error = -EIO; |
60 | #endif | |
61 | ||
4375a336 | 62 | if (f2fs_bio_encrypted(bio)) { |
4246a0b6 | 63 | if (bio->bi_error) { |
0b81d077 | 64 | fscrypt_release_ctx(bio->bi_private); |
4375a336 | 65 | } else { |
0b81d077 | 66 | fscrypt_decrypt_bio_pages(bio->bi_private, bio); |
4375a336 JK |
67 | return; |
68 | } | |
69 | } | |
70 | ||
12377024 CY |
71 | bio_for_each_segment_all(bvec, bio, i) { |
72 | struct page *page = bvec->bv_page; | |
f1e88660 | 73 | |
4246a0b6 | 74 | if (!bio->bi_error) { |
237c0790 JK |
75 | if (!PageUptodate(page)) |
76 | SetPageUptodate(page); | |
f1e88660 JK |
77 | } else { |
78 | ClearPageUptodate(page); | |
79 | SetPageError(page); | |
80 | } | |
81 | unlock_page(page); | |
82 | } | |
f1e88660 JK |
83 | bio_put(bio); |
84 | } | |
85 | ||
4246a0b6 | 86 | static void f2fs_write_end_io(struct bio *bio) |
93dfe2ac | 87 | { |
1b1f559f | 88 | struct f2fs_sb_info *sbi = bio->bi_private; |
f568849e LT |
89 | struct bio_vec *bvec; |
90 | int i; | |
93dfe2ac | 91 | |
f568849e | 92 | bio_for_each_segment_all(bvec, bio, i) { |
93dfe2ac | 93 | struct page *page = bvec->bv_page; |
36951b38 | 94 | enum count_type type = WB_DATA_TYPE(page); |
93dfe2ac | 95 | |
0a595eba JK |
96 | if (IS_DUMMY_WRITTEN_PAGE(page)) { |
97 | set_page_private(page, (unsigned long)NULL); | |
98 | ClearPagePrivate(page); | |
99 | unlock_page(page); | |
100 | mempool_free(page, sbi->write_io_dummy); | |
101 | ||
102 | if (unlikely(bio->bi_error)) | |
103 | f2fs_stop_checkpoint(sbi, true); | |
104 | continue; | |
105 | } | |
106 | ||
0b81d077 | 107 | fscrypt_pullback_bio_page(&page, true); |
4375a336 | 108 | |
4246a0b6 | 109 | if (unlikely(bio->bi_error)) { |
5114a97a | 110 | mapping_set_error(page->mapping, -EIO); |
38f91ca8 | 111 | f2fs_stop_checkpoint(sbi, true); |
93dfe2ac | 112 | } |
36951b38 CY |
113 | dec_page_count(sbi, type); |
114 | clear_cold_data(page); | |
93dfe2ac | 115 | end_page_writeback(page); |
f568849e | 116 | } |
36951b38 | 117 | if (!get_pages(sbi, F2FS_WB_CP_DATA) && |
f5730184 | 118 | wq_has_sleeper(&sbi->cp_wait)) |
93dfe2ac JK |
119 | wake_up(&sbi->cp_wait); |
120 | ||
121 | bio_put(bio); | |
122 | } | |
123 | ||
3c62be17 JK |
124 | /* |
125 | * Return true, if pre_bio's bdev is same as its target device. | |
126 | */ | |
127 | struct block_device *f2fs_target_device(struct f2fs_sb_info *sbi, | |
128 | block_t blk_addr, struct bio *bio) | |
129 | { | |
130 | struct block_device *bdev = sbi->sb->s_bdev; | |
131 | int i; | |
132 | ||
133 | for (i = 0; i < sbi->s_ndevs; i++) { | |
134 | if (FDEV(i).start_blk <= blk_addr && | |
135 | FDEV(i).end_blk >= blk_addr) { | |
136 | blk_addr -= FDEV(i).start_blk; | |
137 | bdev = FDEV(i).bdev; | |
138 | break; | |
139 | } | |
140 | } | |
141 | if (bio) { | |
142 | bio->bi_bdev = bdev; | |
143 | bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr); | |
144 | } | |
145 | return bdev; | |
146 | } | |
147 | ||
148 | int f2fs_target_device_index(struct f2fs_sb_info *sbi, block_t blkaddr) | |
149 | { | |
150 | int i; | |
151 | ||
152 | for (i = 0; i < sbi->s_ndevs; i++) | |
153 | if (FDEV(i).start_blk <= blkaddr && FDEV(i).end_blk >= blkaddr) | |
154 | return i; | |
155 | return 0; | |
156 | } | |
157 | ||
158 | static bool __same_bdev(struct f2fs_sb_info *sbi, | |
159 | block_t blk_addr, struct bio *bio) | |
160 | { | |
161 | return f2fs_target_device(sbi, blk_addr, NULL) == bio->bi_bdev; | |
162 | } | |
163 | ||
940a6d34 GZ |
164 | /* |
165 | * Low-level block read/write IO operations. | |
166 | */ | |
167 | static struct bio *__bio_alloc(struct f2fs_sb_info *sbi, block_t blk_addr, | |
168 | int npages, bool is_read) | |
169 | { | |
170 | struct bio *bio; | |
171 | ||
740432f8 | 172 | bio = f2fs_bio_alloc(npages); |
940a6d34 | 173 | |
3c62be17 | 174 | f2fs_target_device(sbi, blk_addr, bio); |
940a6d34 | 175 | bio->bi_end_io = is_read ? f2fs_read_end_io : f2fs_write_end_io; |
12377024 | 176 | bio->bi_private = is_read ? NULL : sbi; |
940a6d34 GZ |
177 | |
178 | return bio; | |
179 | } | |
180 | ||
4fc29c1a LT |
181 | static inline void __submit_bio(struct f2fs_sb_info *sbi, |
182 | struct bio *bio, enum page_type type) | |
f5730184 | 183 | { |
4fc29c1a | 184 | if (!is_read_io(bio_op(bio))) { |
0a595eba JK |
185 | unsigned int start; |
186 | ||
0bfd7a09 | 187 | if (f2fs_sb_mounted_blkzoned(sbi->sb) && |
52763a4b | 188 | current->plug && (type == DATA || type == NODE)) |
19a5f5e2 | 189 | blk_finish_plug(current->plug); |
0a595eba JK |
190 | |
191 | if (type != DATA && type != NODE) | |
192 | goto submit_io; | |
193 | ||
194 | start = bio->bi_iter.bi_size >> F2FS_BLKSIZE_BITS; | |
195 | start %= F2FS_IO_SIZE(sbi); | |
196 | ||
197 | if (start == 0) | |
198 | goto submit_io; | |
199 | ||
200 | /* fill dummy pages */ | |
201 | for (; start < F2FS_IO_SIZE(sbi); start++) { | |
202 | struct page *page = | |
203 | mempool_alloc(sbi->write_io_dummy, | |
204 | GFP_NOIO | __GFP_ZERO | __GFP_NOFAIL); | |
205 | f2fs_bug_on(sbi, !page); | |
206 | ||
207 | SetPagePrivate(page); | |
208 | set_page_private(page, (unsigned long)DUMMY_WRITTEN_PAGE); | |
209 | lock_page(page); | |
210 | if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) | |
211 | f2fs_bug_on(sbi, 1); | |
212 | } | |
213 | /* | |
214 | * In the NODE case, we lose next block address chain. So, we | |
215 | * need to do checkpoint in f2fs_sync_file. | |
216 | */ | |
217 | if (type == NODE) | |
218 | set_sbi_flag(sbi, SBI_NEED_CP); | |
19a5f5e2 | 219 | } |
0a595eba | 220 | submit_io: |
554b5125 JK |
221 | if (is_read_io(bio_op(bio))) |
222 | trace_f2fs_submit_read_bio(sbi->sb, type, bio); | |
223 | else | |
224 | trace_f2fs_submit_write_bio(sbi->sb, type, bio); | |
4e49ea4a | 225 | submit_bio(bio); |
f5730184 JK |
226 | } |
227 | ||
458e6197 | 228 | static void __submit_merged_bio(struct f2fs_bio_info *io) |
93dfe2ac | 229 | { |
458e6197 | 230 | struct f2fs_io_info *fio = &io->fio; |
93dfe2ac JK |
231 | |
232 | if (!io->bio) | |
233 | return; | |
234 | ||
554b5125 JK |
235 | bio_set_op_attrs(io->bio, fio->op, fio->op_flags); |
236 | ||
04d328de | 237 | if (is_read_io(fio->op)) |
554b5125 | 238 | trace_f2fs_prepare_read_bio(io->sbi->sb, fio->type, io->bio); |
6a8f8ca5 | 239 | else |
554b5125 | 240 | trace_f2fs_prepare_write_bio(io->sbi->sb, fio->type, io->bio); |
04d328de | 241 | |
4fc29c1a | 242 | __submit_bio(io->sbi, io->bio, fio->type); |
93dfe2ac JK |
243 | io->bio = NULL; |
244 | } | |
245 | ||
0c3a5797 CY |
246 | static bool __has_merged_page(struct f2fs_bio_info *io, struct inode *inode, |
247 | struct page *page, nid_t ino) | |
0fd785eb | 248 | { |
0fd785eb CY |
249 | struct bio_vec *bvec; |
250 | struct page *target; | |
251 | int i; | |
252 | ||
0c3a5797 | 253 | if (!io->bio) |
0fd785eb | 254 | return false; |
0c3a5797 CY |
255 | |
256 | if (!inode && !page && !ino) | |
257 | return true; | |
0fd785eb CY |
258 | |
259 | bio_for_each_segment_all(bvec, io->bio, i) { | |
260 | ||
0b81d077 | 261 | if (bvec->bv_page->mapping) |
0fd785eb | 262 | target = bvec->bv_page; |
0b81d077 JK |
263 | else |
264 | target = fscrypt_control_page(bvec->bv_page); | |
0fd785eb | 265 | |
0c3a5797 CY |
266 | if (inode && inode == target->mapping->host) |
267 | return true; | |
268 | if (page && page == target) | |
269 | return true; | |
270 | if (ino && ino == ino_of_node(target)) | |
0fd785eb | 271 | return true; |
0fd785eb CY |
272 | } |
273 | ||
0fd785eb CY |
274 | return false; |
275 | } | |
276 | ||
0c3a5797 CY |
277 | static bool has_merged_page(struct f2fs_sb_info *sbi, struct inode *inode, |
278 | struct page *page, nid_t ino, | |
279 | enum page_type type) | |
280 | { | |
281 | enum page_type btype = PAGE_TYPE_OF_BIO(type); | |
282 | struct f2fs_bio_info *io = &sbi->write_io[btype]; | |
283 | bool ret; | |
284 | ||
285 | down_read(&io->io_rwsem); | |
286 | ret = __has_merged_page(io, inode, page, ino); | |
287 | up_read(&io->io_rwsem); | |
288 | return ret; | |
289 | } | |
290 | ||
291 | static void __f2fs_submit_merged_bio(struct f2fs_sb_info *sbi, | |
292 | struct inode *inode, struct page *page, | |
293 | nid_t ino, enum page_type type, int rw) | |
93dfe2ac JK |
294 | { |
295 | enum page_type btype = PAGE_TYPE_OF_BIO(type); | |
296 | struct f2fs_bio_info *io; | |
297 | ||
298 | io = is_read_io(rw) ? &sbi->read_io : &sbi->write_io[btype]; | |
299 | ||
df0f8dc0 | 300 | down_write(&io->io_rwsem); |
458e6197 | 301 | |
0c3a5797 CY |
302 | if (!__has_merged_page(io, inode, page, ino)) |
303 | goto out; | |
304 | ||
458e6197 JK |
305 | /* change META to META_FLUSH in the checkpoint procedure */ |
306 | if (type >= META_FLUSH) { | |
307 | io->fio.type = META_FLUSH; | |
04d328de | 308 | io->fio.op = REQ_OP_WRITE; |
70fd7614 CH |
309 | io->fio.op_flags = REQ_PREFLUSH | REQ_META | REQ_PRIO; |
310 | if (!test_opt(sbi, NOBARRIER)) | |
311 | io->fio.op_flags |= REQ_FUA; | |
458e6197 JK |
312 | } |
313 | __submit_merged_bio(io); | |
0c3a5797 | 314 | out: |
df0f8dc0 | 315 | up_write(&io->io_rwsem); |
93dfe2ac JK |
316 | } |
317 | ||
0c3a5797 CY |
318 | void f2fs_submit_merged_bio(struct f2fs_sb_info *sbi, enum page_type type, |
319 | int rw) | |
320 | { | |
321 | __f2fs_submit_merged_bio(sbi, NULL, NULL, 0, type, rw); | |
322 | } | |
323 | ||
324 | void f2fs_submit_merged_bio_cond(struct f2fs_sb_info *sbi, | |
325 | struct inode *inode, struct page *page, | |
326 | nid_t ino, enum page_type type, int rw) | |
327 | { | |
328 | if (has_merged_page(sbi, inode, page, ino, type)) | |
329 | __f2fs_submit_merged_bio(sbi, inode, page, ino, type, rw); | |
330 | } | |
331 | ||
406657dd CY |
332 | void f2fs_flush_merged_bios(struct f2fs_sb_info *sbi) |
333 | { | |
334 | f2fs_submit_merged_bio(sbi, DATA, WRITE); | |
335 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
336 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
337 | } | |
338 | ||
93dfe2ac JK |
339 | /* |
340 | * Fill the locked page with data located in the block address. | |
341 | * Return unlocked page. | |
342 | */ | |
05ca3632 | 343 | int f2fs_submit_page_bio(struct f2fs_io_info *fio) |
93dfe2ac | 344 | { |
93dfe2ac | 345 | struct bio *bio; |
0b81d077 JK |
346 | struct page *page = fio->encrypted_page ? |
347 | fio->encrypted_page : fio->page; | |
93dfe2ac | 348 | |
2ace38e0 | 349 | trace_f2fs_submit_page_bio(page, fio); |
05ca3632 | 350 | f2fs_trace_ios(fio, 0); |
93dfe2ac JK |
351 | |
352 | /* Allocate a new bio */ | |
04d328de | 353 | bio = __bio_alloc(fio->sbi, fio->new_blkaddr, 1, is_read_io(fio->op)); |
93dfe2ac | 354 | |
09cbfeaf | 355 | if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) { |
93dfe2ac | 356 | bio_put(bio); |
93dfe2ac JK |
357 | return -EFAULT; |
358 | } | |
04d328de | 359 | bio_set_op_attrs(bio, fio->op, fio->op_flags); |
93dfe2ac | 360 | |
4fc29c1a | 361 | __submit_bio(fio->sbi, bio, fio->type); |
93dfe2ac JK |
362 | return 0; |
363 | } | |
364 | ||
0a595eba | 365 | int f2fs_submit_page_mbio(struct f2fs_io_info *fio) |
93dfe2ac | 366 | { |
05ca3632 | 367 | struct f2fs_sb_info *sbi = fio->sbi; |
458e6197 | 368 | enum page_type btype = PAGE_TYPE_OF_BIO(fio->type); |
93dfe2ac | 369 | struct f2fs_bio_info *io; |
04d328de | 370 | bool is_read = is_read_io(fio->op); |
4375a336 | 371 | struct page *bio_page; |
0a595eba | 372 | int err = 0; |
93dfe2ac | 373 | |
940a6d34 | 374 | io = is_read ? &sbi->read_io : &sbi->write_io[btype]; |
93dfe2ac | 375 | |
7a9d7548 CY |
376 | if (fio->old_blkaddr != NEW_ADDR) |
377 | verify_block_addr(sbi, fio->old_blkaddr); | |
378 | verify_block_addr(sbi, fio->new_blkaddr); | |
93dfe2ac | 379 | |
36951b38 CY |
380 | bio_page = fio->encrypted_page ? fio->encrypted_page : fio->page; |
381 | ||
382 | if (!is_read) | |
383 | inc_page_count(sbi, WB_DATA_TYPE(bio_page)); | |
384 | ||
df0f8dc0 | 385 | down_write(&io->io_rwsem); |
93dfe2ac | 386 | |
7a9d7548 | 387 | if (io->bio && (io->last_block_in_bio != fio->new_blkaddr - 1 || |
3c62be17 JK |
388 | (io->fio.op != fio->op || io->fio.op_flags != fio->op_flags) || |
389 | !__same_bdev(sbi, fio->new_blkaddr, io->bio))) | |
458e6197 | 390 | __submit_merged_bio(io); |
93dfe2ac JK |
391 | alloc_new: |
392 | if (io->bio == NULL) { | |
0a595eba JK |
393 | if ((fio->type == DATA || fio->type == NODE) && |
394 | fio->new_blkaddr & F2FS_IO_SIZE_MASK(sbi)) { | |
395 | err = -EAGAIN; | |
396 | dec_page_count(sbi, WB_DATA_TYPE(bio_page)); | |
397 | goto out_fail; | |
398 | } | |
7a9d7548 | 399 | io->bio = __bio_alloc(sbi, fio->new_blkaddr, |
664ba972 | 400 | BIO_MAX_PAGES, is_read); |
458e6197 | 401 | io->fio = *fio; |
93dfe2ac JK |
402 | } |
403 | ||
09cbfeaf KS |
404 | if (bio_add_page(io->bio, bio_page, PAGE_SIZE, 0) < |
405 | PAGE_SIZE) { | |
458e6197 | 406 | __submit_merged_bio(io); |
93dfe2ac JK |
407 | goto alloc_new; |
408 | } | |
409 | ||
7a9d7548 | 410 | io->last_block_in_bio = fio->new_blkaddr; |
05ca3632 | 411 | f2fs_trace_ios(fio, 0); |
0a595eba | 412 | out_fail: |
df0f8dc0 | 413 | up_write(&io->io_rwsem); |
05ca3632 | 414 | trace_f2fs_submit_page_mbio(fio->page, fio); |
0a595eba | 415 | return err; |
93dfe2ac JK |
416 | } |
417 | ||
46008c6d CY |
418 | static void __set_data_blkaddr(struct dnode_of_data *dn) |
419 | { | |
420 | struct f2fs_node *rn = F2FS_NODE(dn->node_page); | |
421 | __le32 *addr_array; | |
422 | ||
423 | /* Get physical address of data block */ | |
424 | addr_array = blkaddr_in_node(rn); | |
425 | addr_array[dn->ofs_in_node] = cpu_to_le32(dn->data_blkaddr); | |
426 | } | |
427 | ||
0a8165d7 | 428 | /* |
eb47b800 JK |
429 | * Lock ordering for the change of data block address: |
430 | * ->data_page | |
431 | * ->node_page | |
432 | * update block addresses in the node page | |
433 | */ | |
216a620a | 434 | void set_data_blkaddr(struct dnode_of_data *dn) |
eb47b800 | 435 | { |
46008c6d CY |
436 | f2fs_wait_on_page_writeback(dn->node_page, NODE, true); |
437 | __set_data_blkaddr(dn); | |
438 | if (set_page_dirty(dn->node_page)) | |
12719ae1 | 439 | dn->node_changed = true; |
eb47b800 JK |
440 | } |
441 | ||
f28b3434 CY |
442 | void f2fs_update_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr) |
443 | { | |
444 | dn->data_blkaddr = blkaddr; | |
445 | set_data_blkaddr(dn); | |
446 | f2fs_update_extent_cache(dn); | |
447 | } | |
448 | ||
46008c6d CY |
449 | /* dn->ofs_in_node will be returned with up-to-date last block pointer */ |
450 | int reserve_new_blocks(struct dnode_of_data *dn, blkcnt_t count) | |
eb47b800 | 451 | { |
4081363f | 452 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
eb47b800 | 453 | |
46008c6d CY |
454 | if (!count) |
455 | return 0; | |
456 | ||
91942321 | 457 | if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC))) |
eb47b800 | 458 | return -EPERM; |
46008c6d | 459 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, &count))) |
eb47b800 JK |
460 | return -ENOSPC; |
461 | ||
46008c6d CY |
462 | trace_f2fs_reserve_new_blocks(dn->inode, dn->nid, |
463 | dn->ofs_in_node, count); | |
464 | ||
465 | f2fs_wait_on_page_writeback(dn->node_page, NODE, true); | |
466 | ||
467 | for (; count > 0; dn->ofs_in_node++) { | |
468 | block_t blkaddr = | |
469 | datablock_addr(dn->node_page, dn->ofs_in_node); | |
470 | if (blkaddr == NULL_ADDR) { | |
471 | dn->data_blkaddr = NEW_ADDR; | |
472 | __set_data_blkaddr(dn); | |
473 | count--; | |
474 | } | |
475 | } | |
476 | ||
477 | if (set_page_dirty(dn->node_page)) | |
478 | dn->node_changed = true; | |
eb47b800 JK |
479 | return 0; |
480 | } | |
481 | ||
46008c6d CY |
482 | /* Should keep dn->ofs_in_node unchanged */ |
483 | int reserve_new_block(struct dnode_of_data *dn) | |
484 | { | |
485 | unsigned int ofs_in_node = dn->ofs_in_node; | |
486 | int ret; | |
487 | ||
488 | ret = reserve_new_blocks(dn, 1); | |
489 | dn->ofs_in_node = ofs_in_node; | |
490 | return ret; | |
491 | } | |
492 | ||
b600965c HL |
493 | int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index) |
494 | { | |
495 | bool need_put = dn->inode_page ? false : true; | |
496 | int err; | |
497 | ||
498 | err = get_dnode_of_data(dn, index, ALLOC_NODE); | |
499 | if (err) | |
500 | return err; | |
a8865372 | 501 | |
b600965c HL |
502 | if (dn->data_blkaddr == NULL_ADDR) |
503 | err = reserve_new_block(dn); | |
a8865372 | 504 | if (err || need_put) |
b600965c HL |
505 | f2fs_put_dnode(dn); |
506 | return err; | |
507 | } | |
508 | ||
759af1c9 | 509 | int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index) |
eb47b800 | 510 | { |
028a41e8 | 511 | struct extent_info ei; |
759af1c9 | 512 | struct inode *inode = dn->inode; |
028a41e8 | 513 | |
759af1c9 FL |
514 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { |
515 | dn->data_blkaddr = ei.blk + index - ei.fofs; | |
516 | return 0; | |
429511cd | 517 | } |
028a41e8 | 518 | |
759af1c9 | 519 | return f2fs_reserve_block(dn, index); |
eb47b800 JK |
520 | } |
521 | ||
a56c7c6f | 522 | struct page *get_read_data_page(struct inode *inode, pgoff_t index, |
04d328de | 523 | int op_flags, bool for_write) |
eb47b800 | 524 | { |
eb47b800 JK |
525 | struct address_space *mapping = inode->i_mapping; |
526 | struct dnode_of_data dn; | |
527 | struct page *page; | |
cb3bc9ee | 528 | struct extent_info ei; |
eb47b800 | 529 | int err; |
cf04e8eb | 530 | struct f2fs_io_info fio = { |
05ca3632 | 531 | .sbi = F2FS_I_SB(inode), |
cf04e8eb | 532 | .type = DATA, |
04d328de MC |
533 | .op = REQ_OP_READ, |
534 | .op_flags = op_flags, | |
4375a336 | 535 | .encrypted_page = NULL, |
cf04e8eb | 536 | }; |
eb47b800 | 537 | |
4375a336 JK |
538 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) |
539 | return read_mapping_page(mapping, index, NULL); | |
540 | ||
a56c7c6f | 541 | page = f2fs_grab_cache_page(mapping, index, for_write); |
650495de JK |
542 | if (!page) |
543 | return ERR_PTR(-ENOMEM); | |
544 | ||
cb3bc9ee CY |
545 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { |
546 | dn.data_blkaddr = ei.blk + index - ei.fofs; | |
547 | goto got_it; | |
548 | } | |
549 | ||
eb47b800 | 550 | set_new_dnode(&dn, inode, NULL, NULL, 0); |
266e97a8 | 551 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); |
86531d6b JK |
552 | if (err) |
553 | goto put_err; | |
eb47b800 JK |
554 | f2fs_put_dnode(&dn); |
555 | ||
6bacf52f | 556 | if (unlikely(dn.data_blkaddr == NULL_ADDR)) { |
86531d6b JK |
557 | err = -ENOENT; |
558 | goto put_err; | |
650495de | 559 | } |
cb3bc9ee | 560 | got_it: |
43f3eae1 JK |
561 | if (PageUptodate(page)) { |
562 | unlock_page(page); | |
eb47b800 | 563 | return page; |
43f3eae1 | 564 | } |
eb47b800 | 565 | |
d59ff4df JK |
566 | /* |
567 | * A new dentry page is allocated but not able to be written, since its | |
568 | * new inode page couldn't be allocated due to -ENOSPC. | |
569 | * In such the case, its blkaddr can be remained as NEW_ADDR. | |
570 | * see, f2fs_add_link -> get_new_data_page -> init_inode_metadata. | |
571 | */ | |
572 | if (dn.data_blkaddr == NEW_ADDR) { | |
09cbfeaf | 573 | zero_user_segment(page, 0, PAGE_SIZE); |
237c0790 JK |
574 | if (!PageUptodate(page)) |
575 | SetPageUptodate(page); | |
43f3eae1 | 576 | unlock_page(page); |
d59ff4df JK |
577 | return page; |
578 | } | |
eb47b800 | 579 | |
7a9d7548 | 580 | fio.new_blkaddr = fio.old_blkaddr = dn.data_blkaddr; |
05ca3632 JK |
581 | fio.page = page; |
582 | err = f2fs_submit_page_bio(&fio); | |
393ff91f | 583 | if (err) |
86531d6b | 584 | goto put_err; |
43f3eae1 | 585 | return page; |
86531d6b JK |
586 | |
587 | put_err: | |
588 | f2fs_put_page(page, 1); | |
589 | return ERR_PTR(err); | |
43f3eae1 JK |
590 | } |
591 | ||
592 | struct page *find_data_page(struct inode *inode, pgoff_t index) | |
593 | { | |
594 | struct address_space *mapping = inode->i_mapping; | |
595 | struct page *page; | |
596 | ||
597 | page = find_get_page(mapping, index); | |
598 | if (page && PageUptodate(page)) | |
599 | return page; | |
600 | f2fs_put_page(page, 0); | |
601 | ||
70fd7614 | 602 | page = get_read_data_page(inode, index, 0, false); |
43f3eae1 JK |
603 | if (IS_ERR(page)) |
604 | return page; | |
605 | ||
606 | if (PageUptodate(page)) | |
607 | return page; | |
608 | ||
609 | wait_on_page_locked(page); | |
610 | if (unlikely(!PageUptodate(page))) { | |
611 | f2fs_put_page(page, 0); | |
612 | return ERR_PTR(-EIO); | |
613 | } | |
614 | return page; | |
615 | } | |
616 | ||
617 | /* | |
618 | * If it tries to access a hole, return an error. | |
619 | * Because, the callers, functions in dir.c and GC, should be able to know | |
620 | * whether this page exists or not. | |
621 | */ | |
a56c7c6f JK |
622 | struct page *get_lock_data_page(struct inode *inode, pgoff_t index, |
623 | bool for_write) | |
43f3eae1 JK |
624 | { |
625 | struct address_space *mapping = inode->i_mapping; | |
626 | struct page *page; | |
627 | repeat: | |
70fd7614 | 628 | page = get_read_data_page(inode, index, 0, for_write); |
43f3eae1 JK |
629 | if (IS_ERR(page)) |
630 | return page; | |
393ff91f | 631 | |
43f3eae1 | 632 | /* wait for read completion */ |
393ff91f | 633 | lock_page(page); |
6bacf52f | 634 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
635 | f2fs_put_page(page, 1); |
636 | goto repeat; | |
eb47b800 | 637 | } |
1563ac75 CY |
638 | if (unlikely(!PageUptodate(page))) { |
639 | f2fs_put_page(page, 1); | |
640 | return ERR_PTR(-EIO); | |
641 | } | |
eb47b800 JK |
642 | return page; |
643 | } | |
644 | ||
0a8165d7 | 645 | /* |
eb47b800 JK |
646 | * Caller ensures that this data page is never allocated. |
647 | * A new zero-filled data page is allocated in the page cache. | |
39936837 | 648 | * |
4f4124d0 CY |
649 | * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and |
650 | * f2fs_unlock_op(). | |
470f00e9 CY |
651 | * Note that, ipage is set only by make_empty_dir, and if any error occur, |
652 | * ipage should be released by this function. | |
eb47b800 | 653 | */ |
64aa7ed9 | 654 | struct page *get_new_data_page(struct inode *inode, |
a8865372 | 655 | struct page *ipage, pgoff_t index, bool new_i_size) |
eb47b800 | 656 | { |
eb47b800 JK |
657 | struct address_space *mapping = inode->i_mapping; |
658 | struct page *page; | |
659 | struct dnode_of_data dn; | |
660 | int err; | |
7612118a | 661 | |
a56c7c6f | 662 | page = f2fs_grab_cache_page(mapping, index, true); |
470f00e9 CY |
663 | if (!page) { |
664 | /* | |
665 | * before exiting, we should make sure ipage will be released | |
666 | * if any error occur. | |
667 | */ | |
668 | f2fs_put_page(ipage, 1); | |
01f28610 | 669 | return ERR_PTR(-ENOMEM); |
470f00e9 | 670 | } |
eb47b800 | 671 | |
a8865372 | 672 | set_new_dnode(&dn, inode, ipage, NULL, 0); |
b600965c | 673 | err = f2fs_reserve_block(&dn, index); |
01f28610 JK |
674 | if (err) { |
675 | f2fs_put_page(page, 1); | |
eb47b800 | 676 | return ERR_PTR(err); |
a8865372 | 677 | } |
01f28610 JK |
678 | if (!ipage) |
679 | f2fs_put_dnode(&dn); | |
eb47b800 JK |
680 | |
681 | if (PageUptodate(page)) | |
01f28610 | 682 | goto got_it; |
eb47b800 JK |
683 | |
684 | if (dn.data_blkaddr == NEW_ADDR) { | |
09cbfeaf | 685 | zero_user_segment(page, 0, PAGE_SIZE); |
237c0790 JK |
686 | if (!PageUptodate(page)) |
687 | SetPageUptodate(page); | |
eb47b800 | 688 | } else { |
4375a336 | 689 | f2fs_put_page(page, 1); |
a8865372 | 690 | |
7612118a JK |
691 | /* if ipage exists, blkaddr should be NEW_ADDR */ |
692 | f2fs_bug_on(F2FS_I_SB(inode), ipage); | |
693 | page = get_lock_data_page(inode, index, true); | |
4375a336 | 694 | if (IS_ERR(page)) |
7612118a | 695 | return page; |
eb47b800 | 696 | } |
01f28610 | 697 | got_it: |
9edcdabf | 698 | if (new_i_size && i_size_read(inode) < |
ee6d182f | 699 | ((loff_t)(index + 1) << PAGE_SHIFT)) |
fc9581c8 | 700 | f2fs_i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT)); |
eb47b800 JK |
701 | return page; |
702 | } | |
703 | ||
bfad7c2d JK |
704 | static int __allocate_data_block(struct dnode_of_data *dn) |
705 | { | |
4081363f | 706 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
bfad7c2d | 707 | struct f2fs_summary sum; |
bfad7c2d | 708 | struct node_info ni; |
976e4c50 | 709 | pgoff_t fofs; |
46008c6d | 710 | blkcnt_t count = 1; |
bfad7c2d | 711 | |
91942321 | 712 | if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC))) |
bfad7c2d | 713 | return -EPERM; |
df6136ef CY |
714 | |
715 | dn->data_blkaddr = datablock_addr(dn->node_page, dn->ofs_in_node); | |
716 | if (dn->data_blkaddr == NEW_ADDR) | |
717 | goto alloc; | |
718 | ||
46008c6d | 719 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, &count))) |
bfad7c2d JK |
720 | return -ENOSPC; |
721 | ||
df6136ef | 722 | alloc: |
bfad7c2d JK |
723 | get_node_info(sbi, dn->nid, &ni); |
724 | set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version); | |
725 | ||
df6136ef | 726 | allocate_data_block(sbi, NULL, dn->data_blkaddr, &dn->data_blkaddr, |
6ae1be13 | 727 | &sum, CURSEG_WARM_DATA); |
216a620a | 728 | set_data_blkaddr(dn); |
bfad7c2d | 729 | |
976e4c50 | 730 | /* update i_size */ |
81ca7350 | 731 | fofs = start_bidx_of_node(ofs_of_node(dn->node_page), dn->inode) + |
976e4c50 | 732 | dn->ofs_in_node; |
09cbfeaf | 733 | if (i_size_read(dn->inode) < ((loff_t)(fofs + 1) << PAGE_SHIFT)) |
fc9581c8 | 734 | f2fs_i_size_write(dn->inode, |
09cbfeaf | 735 | ((loff_t)(fofs + 1) << PAGE_SHIFT)); |
bfad7c2d JK |
736 | return 0; |
737 | } | |
738 | ||
c040ff9d JK |
739 | static inline bool __force_buffered_io(struct inode *inode, int rw) |
740 | { | |
741 | return ((f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) || | |
742 | (rw == WRITE && test_opt(F2FS_I_SB(inode), LFS)) || | |
743 | F2FS_I_SB(inode)->s_ndevs); | |
744 | } | |
745 | ||
a7de6086 | 746 | int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *from) |
59b802e5 | 747 | { |
b439b103 | 748 | struct inode *inode = file_inode(iocb->ki_filp); |
5b8db7fa | 749 | struct f2fs_map_blocks map; |
a7de6086 | 750 | int err = 0; |
59b802e5 | 751 | |
0080c507 | 752 | map.m_lblk = F2FS_BLK_ALIGN(iocb->ki_pos); |
dfd02e4d CY |
753 | map.m_len = F2FS_BYTES_TO_BLK(iocb->ki_pos + iov_iter_count(from)); |
754 | if (map.m_len > map.m_lblk) | |
755 | map.m_len -= map.m_lblk; | |
756 | else | |
757 | map.m_len = 0; | |
758 | ||
da85985c | 759 | map.m_next_pgofs = NULL; |
2a340760 | 760 | |
24b84912 | 761 | if (iocb->ki_flags & IOCB_DIRECT) { |
a7de6086 JK |
762 | err = f2fs_convert_inline_inode(inode); |
763 | if (err) | |
764 | return err; | |
c040ff9d JK |
765 | return f2fs_map_blocks(inode, &map, 1, |
766 | __force_buffered_io(inode, WRITE) ? | |
767 | F2FS_GET_BLOCK_PRE_AIO : | |
768 | F2FS_GET_BLOCK_PRE_DIO); | |
24b84912 JK |
769 | } |
770 | if (iocb->ki_pos + iov_iter_count(from) > MAX_INLINE_DATA) { | |
a7de6086 JK |
771 | err = f2fs_convert_inline_inode(inode); |
772 | if (err) | |
773 | return err; | |
b439b103 | 774 | } |
24b84912 JK |
775 | if (!f2fs_has_inline_data(inode)) |
776 | return f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO); | |
a7de6086 | 777 | return err; |
59b802e5 JK |
778 | } |
779 | ||
0a8165d7 | 780 | /* |
003a3e1d JK |
781 | * f2fs_map_blocks() now supported readahead/bmap/rw direct_IO with |
782 | * f2fs_map_blocks structure. | |
4f4124d0 CY |
783 | * If original data blocks are allocated, then give them to blockdev. |
784 | * Otherwise, | |
785 | * a. preallocate requested block addresses | |
786 | * b. do not use extent cache for better performance | |
787 | * c. give the block addresses to blockdev | |
eb47b800 | 788 | */ |
d323d005 | 789 | int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map, |
e2b4e2bc | 790 | int create, int flag) |
eb47b800 | 791 | { |
003a3e1d | 792 | unsigned int maxblocks = map->m_len; |
eb47b800 | 793 | struct dnode_of_data dn; |
f9811703 | 794 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
ac6f1999 | 795 | int mode = create ? ALLOC_NODE : LOOKUP_NODE; |
46008c6d | 796 | pgoff_t pgofs, end_offset, end; |
bfad7c2d | 797 | int err = 0, ofs = 1; |
46008c6d CY |
798 | unsigned int ofs_in_node, last_ofs_in_node; |
799 | blkcnt_t prealloc; | |
a2e7d1bf | 800 | struct extent_info ei; |
7df3a431 | 801 | block_t blkaddr; |
eb47b800 | 802 | |
dfd02e4d CY |
803 | if (!maxblocks) |
804 | return 0; | |
805 | ||
003a3e1d JK |
806 | map->m_len = 0; |
807 | map->m_flags = 0; | |
808 | ||
809 | /* it only supports block size == page size */ | |
810 | pgofs = (pgoff_t)map->m_lblk; | |
46008c6d | 811 | end = pgofs + maxblocks; |
eb47b800 | 812 | |
24b84912 | 813 | if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) { |
003a3e1d JK |
814 | map->m_pblk = ei.blk + pgofs - ei.fofs; |
815 | map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs); | |
816 | map->m_flags = F2FS_MAP_MAPPED; | |
bfad7c2d | 817 | goto out; |
a2e7d1bf | 818 | } |
bfad7c2d | 819 | |
4fe71e88 | 820 | next_dnode: |
59b802e5 | 821 | if (create) |
3104af35 | 822 | f2fs_lock_op(sbi); |
eb47b800 JK |
823 | |
824 | /* When reading holes, we need its node page */ | |
825 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
bfad7c2d | 826 | err = get_dnode_of_data(&dn, pgofs, mode); |
1ec79083 | 827 | if (err) { |
43473f96 CY |
828 | if (flag == F2FS_GET_BLOCK_BMAP) |
829 | map->m_pblk = 0; | |
da85985c | 830 | if (err == -ENOENT) { |
bfad7c2d | 831 | err = 0; |
da85985c CY |
832 | if (map->m_next_pgofs) |
833 | *map->m_next_pgofs = | |
834 | get_next_page_offset(&dn, pgofs); | |
835 | } | |
bfad7c2d | 836 | goto unlock_out; |
848753aa | 837 | } |
973163fc | 838 | |
46008c6d | 839 | prealloc = 0; |
230436b3 | 840 | last_ofs_in_node = ofs_in_node = dn.ofs_in_node; |
81ca7350 | 841 | end_offset = ADDRS_PER_PAGE(dn.node_page, inode); |
4fe71e88 CY |
842 | |
843 | next_block: | |
844 | blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node); | |
845 | ||
846 | if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR) { | |
973163fc | 847 | if (create) { |
f9811703 CY |
848 | if (unlikely(f2fs_cp_error(sbi))) { |
849 | err = -EIO; | |
4fe71e88 | 850 | goto sync_out; |
f9811703 | 851 | } |
24b84912 | 852 | if (flag == F2FS_GET_BLOCK_PRE_AIO) { |
46008c6d CY |
853 | if (blkaddr == NULL_ADDR) { |
854 | prealloc++; | |
855 | last_ofs_in_node = dn.ofs_in_node; | |
856 | } | |
24b84912 JK |
857 | } else { |
858 | err = __allocate_data_block(&dn); | |
6f2d8ed6 | 859 | if (!err) |
91942321 | 860 | set_inode_flag(inode, FI_APPEND_WRITE); |
24b84912 | 861 | } |
973163fc | 862 | if (err) |
4fe71e88 | 863 | goto sync_out; |
973163fc | 864 | map->m_flags = F2FS_MAP_NEW; |
4fe71e88 | 865 | blkaddr = dn.data_blkaddr; |
973163fc | 866 | } else { |
43473f96 CY |
867 | if (flag == F2FS_GET_BLOCK_BMAP) { |
868 | map->m_pblk = 0; | |
869 | goto sync_out; | |
870 | } | |
da85985c CY |
871 | if (flag == F2FS_GET_BLOCK_FIEMAP && |
872 | blkaddr == NULL_ADDR) { | |
873 | if (map->m_next_pgofs) | |
874 | *map->m_next_pgofs = pgofs + 1; | |
875 | } | |
973163fc | 876 | if (flag != F2FS_GET_BLOCK_FIEMAP || |
43473f96 | 877 | blkaddr != NEW_ADDR) |
4fe71e88 | 878 | goto sync_out; |
e2b4e2bc | 879 | } |
e2b4e2bc | 880 | } |
eb47b800 | 881 | |
46008c6d CY |
882 | if (flag == F2FS_GET_BLOCK_PRE_AIO) |
883 | goto skip; | |
884 | ||
4fe71e88 CY |
885 | if (map->m_len == 0) { |
886 | /* preallocated unwritten block should be mapped for fiemap. */ | |
887 | if (blkaddr == NEW_ADDR) | |
888 | map->m_flags |= F2FS_MAP_UNWRITTEN; | |
889 | map->m_flags |= F2FS_MAP_MAPPED; | |
890 | ||
891 | map->m_pblk = blkaddr; | |
892 | map->m_len = 1; | |
893 | } else if ((map->m_pblk != NEW_ADDR && | |
894 | blkaddr == (map->m_pblk + ofs)) || | |
b439b103 | 895 | (map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR) || |
46008c6d | 896 | flag == F2FS_GET_BLOCK_PRE_DIO) { |
4fe71e88 CY |
897 | ofs++; |
898 | map->m_len++; | |
899 | } else { | |
900 | goto sync_out; | |
901 | } | |
bfad7c2d | 902 | |
46008c6d | 903 | skip: |
bfad7c2d JK |
904 | dn.ofs_in_node++; |
905 | pgofs++; | |
906 | ||
46008c6d CY |
907 | /* preallocate blocks in batch for one dnode page */ |
908 | if (flag == F2FS_GET_BLOCK_PRE_AIO && | |
909 | (pgofs == end || dn.ofs_in_node == end_offset)) { | |
7df3a431 | 910 | |
46008c6d CY |
911 | dn.ofs_in_node = ofs_in_node; |
912 | err = reserve_new_blocks(&dn, prealloc); | |
913 | if (err) | |
914 | goto sync_out; | |
bfad7c2d | 915 | |
46008c6d CY |
916 | map->m_len += dn.ofs_in_node - ofs_in_node; |
917 | if (prealloc && dn.ofs_in_node != last_ofs_in_node + 1) { | |
918 | err = -ENOSPC; | |
919 | goto sync_out; | |
3104af35 | 920 | } |
46008c6d CY |
921 | dn.ofs_in_node = end_offset; |
922 | } | |
923 | ||
924 | if (pgofs >= end) | |
925 | goto sync_out; | |
926 | else if (dn.ofs_in_node < end_offset) | |
927 | goto next_block; | |
928 | ||
46008c6d CY |
929 | f2fs_put_dnode(&dn); |
930 | ||
931 | if (create) { | |
932 | f2fs_unlock_op(sbi); | |
6f2d8ed6 | 933 | f2fs_balance_fs(sbi, dn.node_changed); |
eb47b800 | 934 | } |
46008c6d | 935 | goto next_dnode; |
7df3a431 | 936 | |
bfad7c2d | 937 | sync_out: |
eb47b800 | 938 | f2fs_put_dnode(&dn); |
bfad7c2d | 939 | unlock_out: |
2a340760 | 940 | if (create) { |
3104af35 | 941 | f2fs_unlock_op(sbi); |
6f2d8ed6 | 942 | f2fs_balance_fs(sbi, dn.node_changed); |
2a340760 | 943 | } |
bfad7c2d | 944 | out: |
003a3e1d | 945 | trace_f2fs_map_blocks(inode, map, err); |
bfad7c2d | 946 | return err; |
eb47b800 JK |
947 | } |
948 | ||
003a3e1d | 949 | static int __get_data_block(struct inode *inode, sector_t iblock, |
da85985c CY |
950 | struct buffer_head *bh, int create, int flag, |
951 | pgoff_t *next_pgofs) | |
003a3e1d JK |
952 | { |
953 | struct f2fs_map_blocks map; | |
a7de6086 | 954 | int err; |
003a3e1d JK |
955 | |
956 | map.m_lblk = iblock; | |
957 | map.m_len = bh->b_size >> inode->i_blkbits; | |
da85985c | 958 | map.m_next_pgofs = next_pgofs; |
003a3e1d | 959 | |
a7de6086 JK |
960 | err = f2fs_map_blocks(inode, &map, create, flag); |
961 | if (!err) { | |
003a3e1d JK |
962 | map_bh(bh, inode->i_sb, map.m_pblk); |
963 | bh->b_state = (bh->b_state & ~F2FS_MAP_FLAGS) | map.m_flags; | |
964 | bh->b_size = map.m_len << inode->i_blkbits; | |
965 | } | |
a7de6086 | 966 | return err; |
003a3e1d JK |
967 | } |
968 | ||
ccfb3000 | 969 | static int get_data_block(struct inode *inode, sector_t iblock, |
da85985c CY |
970 | struct buffer_head *bh_result, int create, int flag, |
971 | pgoff_t *next_pgofs) | |
e2b4e2bc | 972 | { |
da85985c CY |
973 | return __get_data_block(inode, iblock, bh_result, create, |
974 | flag, next_pgofs); | |
e2b4e2bc CY |
975 | } |
976 | ||
977 | static int get_data_block_dio(struct inode *inode, sector_t iblock, | |
ccfb3000 JK |
978 | struct buffer_head *bh_result, int create) |
979 | { | |
e2b4e2bc | 980 | return __get_data_block(inode, iblock, bh_result, create, |
da85985c | 981 | F2FS_GET_BLOCK_DIO, NULL); |
ccfb3000 JK |
982 | } |
983 | ||
e2b4e2bc | 984 | static int get_data_block_bmap(struct inode *inode, sector_t iblock, |
ccfb3000 JK |
985 | struct buffer_head *bh_result, int create) |
986 | { | |
179448bf | 987 | /* Block number less than F2FS MAX BLOCKS */ |
e0afc4d6 | 988 | if (unlikely(iblock >= F2FS_I_SB(inode)->max_file_blocks)) |
179448bf YH |
989 | return -EFBIG; |
990 | ||
e2b4e2bc | 991 | return __get_data_block(inode, iblock, bh_result, create, |
da85985c | 992 | F2FS_GET_BLOCK_BMAP, NULL); |
ccfb3000 JK |
993 | } |
994 | ||
7f63eb77 JK |
995 | static inline sector_t logical_to_blk(struct inode *inode, loff_t offset) |
996 | { | |
997 | return (offset >> inode->i_blkbits); | |
998 | } | |
999 | ||
1000 | static inline loff_t blk_to_logical(struct inode *inode, sector_t blk) | |
1001 | { | |
1002 | return (blk << inode->i_blkbits); | |
1003 | } | |
1004 | ||
9ab70134 JK |
1005 | int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, |
1006 | u64 start, u64 len) | |
1007 | { | |
7f63eb77 JK |
1008 | struct buffer_head map_bh; |
1009 | sector_t start_blk, last_blk; | |
da85985c | 1010 | pgoff_t next_pgofs; |
7f63eb77 JK |
1011 | u64 logical = 0, phys = 0, size = 0; |
1012 | u32 flags = 0; | |
7f63eb77 JK |
1013 | int ret = 0; |
1014 | ||
1015 | ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC); | |
1016 | if (ret) | |
1017 | return ret; | |
1018 | ||
67f8cf3c JK |
1019 | if (f2fs_has_inline_data(inode)) { |
1020 | ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len); | |
1021 | if (ret != -EAGAIN) | |
1022 | return ret; | |
1023 | } | |
1024 | ||
5955102c | 1025 | inode_lock(inode); |
de1475cc | 1026 | |
7f63eb77 JK |
1027 | if (logical_to_blk(inode, len) == 0) |
1028 | len = blk_to_logical(inode, 1); | |
1029 | ||
1030 | start_blk = logical_to_blk(inode, start); | |
1031 | last_blk = logical_to_blk(inode, start + len - 1); | |
9a950d52 | 1032 | |
7f63eb77 JK |
1033 | next: |
1034 | memset(&map_bh, 0, sizeof(struct buffer_head)); | |
1035 | map_bh.b_size = len; | |
1036 | ||
e2b4e2bc | 1037 | ret = get_data_block(inode, start_blk, &map_bh, 0, |
da85985c | 1038 | F2FS_GET_BLOCK_FIEMAP, &next_pgofs); |
7f63eb77 JK |
1039 | if (ret) |
1040 | goto out; | |
1041 | ||
1042 | /* HOLE */ | |
1043 | if (!buffer_mapped(&map_bh)) { | |
da85985c | 1044 | start_blk = next_pgofs; |
58736fa6 CY |
1045 | |
1046 | if (blk_to_logical(inode, start_blk) < blk_to_logical(inode, | |
1047 | F2FS_I_SB(inode)->max_file_blocks)) | |
9a950d52 | 1048 | goto prep_next; |
58736fa6 | 1049 | |
9a950d52 FL |
1050 | flags |= FIEMAP_EXTENT_LAST; |
1051 | } | |
7f63eb77 | 1052 | |
da5af127 CY |
1053 | if (size) { |
1054 | if (f2fs_encrypted_inode(inode)) | |
1055 | flags |= FIEMAP_EXTENT_DATA_ENCRYPTED; | |
1056 | ||
9a950d52 FL |
1057 | ret = fiemap_fill_next_extent(fieinfo, logical, |
1058 | phys, size, flags); | |
da5af127 | 1059 | } |
7f63eb77 | 1060 | |
9a950d52 FL |
1061 | if (start_blk > last_blk || ret) |
1062 | goto out; | |
7f63eb77 | 1063 | |
9a950d52 FL |
1064 | logical = blk_to_logical(inode, start_blk); |
1065 | phys = blk_to_logical(inode, map_bh.b_blocknr); | |
1066 | size = map_bh.b_size; | |
1067 | flags = 0; | |
1068 | if (buffer_unwritten(&map_bh)) | |
1069 | flags = FIEMAP_EXTENT_UNWRITTEN; | |
7f63eb77 | 1070 | |
9a950d52 | 1071 | start_blk += logical_to_blk(inode, size); |
7f63eb77 | 1072 | |
9a950d52 | 1073 | prep_next: |
7f63eb77 JK |
1074 | cond_resched(); |
1075 | if (fatal_signal_pending(current)) | |
1076 | ret = -EINTR; | |
1077 | else | |
1078 | goto next; | |
1079 | out: | |
1080 | if (ret == 1) | |
1081 | ret = 0; | |
1082 | ||
5955102c | 1083 | inode_unlock(inode); |
7f63eb77 | 1084 | return ret; |
9ab70134 JK |
1085 | } |
1086 | ||
6a7a3aed WY |
1087 | static struct bio *f2fs_grab_bio(struct inode *inode, block_t blkaddr, |
1088 | unsigned nr_pages) | |
78682f79 CY |
1089 | { |
1090 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
1091 | struct fscrypt_ctx *ctx = NULL; | |
78682f79 CY |
1092 | struct bio *bio; |
1093 | ||
1094 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) { | |
1095 | ctx = fscrypt_get_ctx(inode, GFP_NOFS); | |
1096 | if (IS_ERR(ctx)) | |
1097 | return ERR_CAST(ctx); | |
1098 | ||
1099 | /* wait the page to be moved by cleaning */ | |
1100 | f2fs_wait_on_encrypted_page_writeback(sbi, blkaddr); | |
1101 | } | |
1102 | ||
1103 | bio = bio_alloc(GFP_KERNEL, min_t(int, nr_pages, BIO_MAX_PAGES)); | |
1104 | if (!bio) { | |
1105 | if (ctx) | |
1106 | fscrypt_release_ctx(ctx); | |
1107 | return ERR_PTR(-ENOMEM); | |
1108 | } | |
3c62be17 | 1109 | f2fs_target_device(sbi, blkaddr, bio); |
78682f79 CY |
1110 | bio->bi_end_io = f2fs_read_end_io; |
1111 | bio->bi_private = ctx; | |
1112 | ||
1113 | return bio; | |
1114 | } | |
1115 | ||
f1e88660 JK |
1116 | /* |
1117 | * This function was originally taken from fs/mpage.c, and customized for f2fs. | |
1118 | * Major change was from block_size == page_size in f2fs by default. | |
1119 | */ | |
1120 | static int f2fs_mpage_readpages(struct address_space *mapping, | |
1121 | struct list_head *pages, struct page *page, | |
1122 | unsigned nr_pages) | |
1123 | { | |
1124 | struct bio *bio = NULL; | |
1125 | unsigned page_idx; | |
1126 | sector_t last_block_in_bio = 0; | |
1127 | struct inode *inode = mapping->host; | |
1128 | const unsigned blkbits = inode->i_blkbits; | |
1129 | const unsigned blocksize = 1 << blkbits; | |
1130 | sector_t block_in_file; | |
1131 | sector_t last_block; | |
1132 | sector_t last_block_in_file; | |
1133 | sector_t block_nr; | |
f1e88660 JK |
1134 | struct f2fs_map_blocks map; |
1135 | ||
1136 | map.m_pblk = 0; | |
1137 | map.m_lblk = 0; | |
1138 | map.m_len = 0; | |
1139 | map.m_flags = 0; | |
da85985c | 1140 | map.m_next_pgofs = NULL; |
f1e88660 JK |
1141 | |
1142 | for (page_idx = 0; nr_pages; page_idx++, nr_pages--) { | |
1143 | ||
1144 | prefetchw(&page->flags); | |
1145 | if (pages) { | |
939afa94 | 1146 | page = list_last_entry(pages, struct page, lru); |
f1e88660 JK |
1147 | list_del(&page->lru); |
1148 | if (add_to_page_cache_lru(page, mapping, | |
8a5c743e MH |
1149 | page->index, |
1150 | readahead_gfp_mask(mapping))) | |
f1e88660 JK |
1151 | goto next_page; |
1152 | } | |
1153 | ||
1154 | block_in_file = (sector_t)page->index; | |
1155 | last_block = block_in_file + nr_pages; | |
1156 | last_block_in_file = (i_size_read(inode) + blocksize - 1) >> | |
1157 | blkbits; | |
1158 | if (last_block > last_block_in_file) | |
1159 | last_block = last_block_in_file; | |
1160 | ||
1161 | /* | |
1162 | * Map blocks using the previous result first. | |
1163 | */ | |
1164 | if ((map.m_flags & F2FS_MAP_MAPPED) && | |
1165 | block_in_file > map.m_lblk && | |
1166 | block_in_file < (map.m_lblk + map.m_len)) | |
1167 | goto got_it; | |
1168 | ||
1169 | /* | |
1170 | * Then do more f2fs_map_blocks() calls until we are | |
1171 | * done with this page. | |
1172 | */ | |
1173 | map.m_flags = 0; | |
1174 | ||
1175 | if (block_in_file < last_block) { | |
1176 | map.m_lblk = block_in_file; | |
1177 | map.m_len = last_block - block_in_file; | |
1178 | ||
46c9e141 | 1179 | if (f2fs_map_blocks(inode, &map, 0, |
da85985c | 1180 | F2FS_GET_BLOCK_READ)) |
f1e88660 JK |
1181 | goto set_error_page; |
1182 | } | |
1183 | got_it: | |
1184 | if ((map.m_flags & F2FS_MAP_MAPPED)) { | |
1185 | block_nr = map.m_pblk + block_in_file - map.m_lblk; | |
1186 | SetPageMappedToDisk(page); | |
1187 | ||
1188 | if (!PageUptodate(page) && !cleancache_get_page(page)) { | |
1189 | SetPageUptodate(page); | |
1190 | goto confused; | |
1191 | } | |
1192 | } else { | |
09cbfeaf | 1193 | zero_user_segment(page, 0, PAGE_SIZE); |
237c0790 JK |
1194 | if (!PageUptodate(page)) |
1195 | SetPageUptodate(page); | |
f1e88660 JK |
1196 | unlock_page(page); |
1197 | goto next_page; | |
1198 | } | |
1199 | ||
1200 | /* | |
1201 | * This page will go to BIO. Do we need to send this | |
1202 | * BIO off first? | |
1203 | */ | |
3c62be17 JK |
1204 | if (bio && (last_block_in_bio != block_nr - 1 || |
1205 | !__same_bdev(F2FS_I_SB(inode), block_nr, bio))) { | |
f1e88660 | 1206 | submit_and_realloc: |
4fc29c1a | 1207 | __submit_bio(F2FS_I_SB(inode), bio, DATA); |
f1e88660 JK |
1208 | bio = NULL; |
1209 | } | |
1210 | if (bio == NULL) { | |
78682f79 | 1211 | bio = f2fs_grab_bio(inode, block_nr, nr_pages); |
1d353eb7 JK |
1212 | if (IS_ERR(bio)) { |
1213 | bio = NULL; | |
f1e88660 | 1214 | goto set_error_page; |
4375a336 | 1215 | } |
04d328de | 1216 | bio_set_op_attrs(bio, REQ_OP_READ, 0); |
f1e88660 JK |
1217 | } |
1218 | ||
1219 | if (bio_add_page(bio, page, blocksize, 0) < blocksize) | |
1220 | goto submit_and_realloc; | |
1221 | ||
1222 | last_block_in_bio = block_nr; | |
1223 | goto next_page; | |
1224 | set_error_page: | |
1225 | SetPageError(page); | |
09cbfeaf | 1226 | zero_user_segment(page, 0, PAGE_SIZE); |
f1e88660 JK |
1227 | unlock_page(page); |
1228 | goto next_page; | |
1229 | confused: | |
1230 | if (bio) { | |
4fc29c1a | 1231 | __submit_bio(F2FS_I_SB(inode), bio, DATA); |
f1e88660 JK |
1232 | bio = NULL; |
1233 | } | |
1234 | unlock_page(page); | |
1235 | next_page: | |
1236 | if (pages) | |
09cbfeaf | 1237 | put_page(page); |
f1e88660 JK |
1238 | } |
1239 | BUG_ON(pages && !list_empty(pages)); | |
1240 | if (bio) | |
4fc29c1a | 1241 | __submit_bio(F2FS_I_SB(inode), bio, DATA); |
f1e88660 JK |
1242 | return 0; |
1243 | } | |
1244 | ||
eb47b800 JK |
1245 | static int f2fs_read_data_page(struct file *file, struct page *page) |
1246 | { | |
9ffe0fb5 | 1247 | struct inode *inode = page->mapping->host; |
b3d208f9 | 1248 | int ret = -EAGAIN; |
9ffe0fb5 | 1249 | |
c20e89cd CY |
1250 | trace_f2fs_readpage(page, DATA); |
1251 | ||
e1c42045 | 1252 | /* If the file has inline data, try to read it directly */ |
9ffe0fb5 HL |
1253 | if (f2fs_has_inline_data(inode)) |
1254 | ret = f2fs_read_inline_data(inode, page); | |
b3d208f9 | 1255 | if (ret == -EAGAIN) |
f1e88660 | 1256 | ret = f2fs_mpage_readpages(page->mapping, NULL, page, 1); |
9ffe0fb5 | 1257 | return ret; |
eb47b800 JK |
1258 | } |
1259 | ||
1260 | static int f2fs_read_data_pages(struct file *file, | |
1261 | struct address_space *mapping, | |
1262 | struct list_head *pages, unsigned nr_pages) | |
1263 | { | |
9ffe0fb5 | 1264 | struct inode *inode = file->f_mapping->host; |
939afa94 | 1265 | struct page *page = list_last_entry(pages, struct page, lru); |
b8c29400 CY |
1266 | |
1267 | trace_f2fs_readpages(inode, page, nr_pages); | |
9ffe0fb5 HL |
1268 | |
1269 | /* If the file has inline data, skip readpages */ | |
1270 | if (f2fs_has_inline_data(inode)) | |
1271 | return 0; | |
1272 | ||
f1e88660 | 1273 | return f2fs_mpage_readpages(mapping, pages, NULL, nr_pages); |
eb47b800 JK |
1274 | } |
1275 | ||
05ca3632 | 1276 | int do_write_data_page(struct f2fs_io_info *fio) |
eb47b800 | 1277 | { |
05ca3632 | 1278 | struct page *page = fio->page; |
eb47b800 | 1279 | struct inode *inode = page->mapping->host; |
eb47b800 JK |
1280 | struct dnode_of_data dn; |
1281 | int err = 0; | |
1282 | ||
1283 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
266e97a8 | 1284 | err = get_dnode_of_data(&dn, page->index, LOOKUP_NODE); |
eb47b800 JK |
1285 | if (err) |
1286 | return err; | |
1287 | ||
28bc106b | 1288 | fio->old_blkaddr = dn.data_blkaddr; |
eb47b800 JK |
1289 | |
1290 | /* This page is already truncated */ | |
7a9d7548 | 1291 | if (fio->old_blkaddr == NULL_ADDR) { |
2bca1e23 | 1292 | ClearPageUptodate(page); |
eb47b800 | 1293 | goto out_writepage; |
2bca1e23 | 1294 | } |
eb47b800 | 1295 | |
4375a336 | 1296 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) { |
b32e4482 | 1297 | gfp_t gfp_flags = GFP_NOFS; |
08b39fbd CY |
1298 | |
1299 | /* wait for GCed encrypted page writeback */ | |
1300 | f2fs_wait_on_encrypted_page_writeback(F2FS_I_SB(inode), | |
7a9d7548 | 1301 | fio->old_blkaddr); |
b32e4482 JK |
1302 | retry_encrypt: |
1303 | fio->encrypted_page = fscrypt_encrypt_page(inode, fio->page, | |
9c4bb8a3 DG |
1304 | PAGE_SIZE, 0, |
1305 | fio->page->index, | |
1306 | gfp_flags); | |
4375a336 JK |
1307 | if (IS_ERR(fio->encrypted_page)) { |
1308 | err = PTR_ERR(fio->encrypted_page); | |
b32e4482 JK |
1309 | if (err == -ENOMEM) { |
1310 | /* flush pending ios and wait for a while */ | |
1311 | f2fs_flush_merged_bios(F2FS_I_SB(inode)); | |
1312 | congestion_wait(BLK_RW_ASYNC, HZ/50); | |
1313 | gfp_flags |= __GFP_NOFAIL; | |
1314 | err = 0; | |
1315 | goto retry_encrypt; | |
1316 | } | |
4375a336 JK |
1317 | goto out_writepage; |
1318 | } | |
1319 | } | |
1320 | ||
eb47b800 JK |
1321 | set_page_writeback(page); |
1322 | ||
1323 | /* | |
1324 | * If current allocation needs SSR, | |
1325 | * it had better in-place writes for updated data. | |
1326 | */ | |
7a9d7548 | 1327 | if (unlikely(fio->old_blkaddr != NEW_ADDR && |
b25958b6 | 1328 | !is_cold_data(page) && |
2da3e027 | 1329 | !IS_ATOMIC_WRITTEN_PAGE(page) && |
b25958b6 | 1330 | need_inplace_update(inode))) { |
05ca3632 | 1331 | rewrite_data_page(fio); |
91942321 | 1332 | set_inode_flag(inode, FI_UPDATE_WRITE); |
8ce67cb0 | 1333 | trace_f2fs_do_write_data_page(page, IPU); |
eb47b800 | 1334 | } else { |
05ca3632 | 1335 | write_data_page(&dn, fio); |
8ce67cb0 | 1336 | trace_f2fs_do_write_data_page(page, OPU); |
91942321 | 1337 | set_inode_flag(inode, FI_APPEND_WRITE); |
3c6c2beb | 1338 | if (page->index == 0) |
91942321 | 1339 | set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN); |
eb47b800 JK |
1340 | } |
1341 | out_writepage: | |
1342 | f2fs_put_dnode(&dn); | |
1343 | return err; | |
1344 | } | |
1345 | ||
1346 | static int f2fs_write_data_page(struct page *page, | |
1347 | struct writeback_control *wbc) | |
1348 | { | |
1349 | struct inode *inode = page->mapping->host; | |
4081363f | 1350 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
eb47b800 JK |
1351 | loff_t i_size = i_size_read(inode); |
1352 | const pgoff_t end_index = ((unsigned long long) i_size) | |
09cbfeaf | 1353 | >> PAGE_SHIFT; |
26de9b11 | 1354 | loff_t psize = (page->index + 1) << PAGE_SHIFT; |
9ffe0fb5 | 1355 | unsigned offset = 0; |
39936837 | 1356 | bool need_balance_fs = false; |
eb47b800 | 1357 | int err = 0; |
458e6197 | 1358 | struct f2fs_io_info fio = { |
05ca3632 | 1359 | .sbi = sbi, |
458e6197 | 1360 | .type = DATA, |
04d328de | 1361 | .op = REQ_OP_WRITE, |
7637241e | 1362 | .op_flags = wbc_to_write_flags(wbc), |
05ca3632 | 1363 | .page = page, |
4375a336 | 1364 | .encrypted_page = NULL, |
458e6197 | 1365 | }; |
eb47b800 | 1366 | |
ecda0de3 CY |
1367 | trace_f2fs_writepage(page, DATA); |
1368 | ||
eb47b800 | 1369 | if (page->index < end_index) |
39936837 | 1370 | goto write; |
eb47b800 JK |
1371 | |
1372 | /* | |
1373 | * If the offset is out-of-range of file size, | |
1374 | * this page does not have to be written to disk. | |
1375 | */ | |
09cbfeaf | 1376 | offset = i_size & (PAGE_SIZE - 1); |
76f60268 | 1377 | if ((page->index >= end_index + 1) || !offset) |
39936837 | 1378 | goto out; |
eb47b800 | 1379 | |
09cbfeaf | 1380 | zero_user_segment(page, offset, PAGE_SIZE); |
39936837 | 1381 | write: |
caf0047e | 1382 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
eb47b800 | 1383 | goto redirty_out; |
1e84371f JK |
1384 | if (f2fs_is_drop_cache(inode)) |
1385 | goto out; | |
e6e5f561 JK |
1386 | /* we should not write 0'th page having journal header */ |
1387 | if (f2fs_is_volatile_file(inode) && (!page->index || | |
1388 | (!wbc->for_reclaim && | |
1389 | available_free_memory(sbi, BASE_CHECK)))) | |
1e84371f | 1390 | goto redirty_out; |
eb47b800 | 1391 | |
cf779cab JK |
1392 | /* we should bypass data pages to proceed the kworkder jobs */ |
1393 | if (unlikely(f2fs_cp_error(sbi))) { | |
7f319975 | 1394 | mapping_set_error(page->mapping, -EIO); |
a7ffdbe2 | 1395 | goto out; |
cf779cab JK |
1396 | } |
1397 | ||
39936837 | 1398 | /* Dentry blocks are controlled by checkpoint */ |
eb47b800 | 1399 | if (S_ISDIR(inode->i_mode)) { |
05ca3632 | 1400 | err = do_write_data_page(&fio); |
8618b881 JK |
1401 | goto done; |
1402 | } | |
9ffe0fb5 | 1403 | |
8618b881 | 1404 | if (!wbc->for_reclaim) |
39936837 | 1405 | need_balance_fs = true; |
7f3037a5 | 1406 | else if (has_not_enough_free_secs(sbi, 0, 0)) |
39936837 | 1407 | goto redirty_out; |
eb47b800 | 1408 | |
b3d208f9 | 1409 | err = -EAGAIN; |
8618b881 | 1410 | f2fs_lock_op(sbi); |
b3d208f9 JK |
1411 | if (f2fs_has_inline_data(inode)) |
1412 | err = f2fs_write_inline_data(inode, page); | |
1413 | if (err == -EAGAIN) | |
05ca3632 | 1414 | err = do_write_data_page(&fio); |
26de9b11 JK |
1415 | if (F2FS_I(inode)->last_disk_size < psize) |
1416 | F2FS_I(inode)->last_disk_size = psize; | |
8618b881 JK |
1417 | f2fs_unlock_op(sbi); |
1418 | done: | |
1419 | if (err && err != -ENOENT) | |
1420 | goto redirty_out; | |
eb47b800 | 1421 | |
39936837 | 1422 | out: |
a7ffdbe2 | 1423 | inode_dec_dirty_pages(inode); |
2bca1e23 JK |
1424 | if (err) |
1425 | ClearPageUptodate(page); | |
0c3a5797 CY |
1426 | |
1427 | if (wbc->for_reclaim) { | |
1428 | f2fs_submit_merged_bio_cond(sbi, NULL, page, 0, DATA, WRITE); | |
1429 | remove_dirty_inode(inode); | |
1430 | } | |
1431 | ||
eb47b800 | 1432 | unlock_page(page); |
2c4db1a6 | 1433 | f2fs_balance_fs(sbi, need_balance_fs); |
0c3a5797 CY |
1434 | |
1435 | if (unlikely(f2fs_cp_error(sbi))) | |
2aea39ec | 1436 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
0c3a5797 | 1437 | |
eb47b800 JK |
1438 | return 0; |
1439 | ||
eb47b800 | 1440 | redirty_out: |
76f60268 | 1441 | redirty_page_for_writepage(wbc, page); |
0002b61b CY |
1442 | if (!err) |
1443 | return AOP_WRITEPAGE_ACTIVATE; | |
b230e6ca JK |
1444 | unlock_page(page); |
1445 | return err; | |
fa9150a8 NJ |
1446 | } |
1447 | ||
8f46dcae CY |
1448 | /* |
1449 | * This function was copied from write_cche_pages from mm/page-writeback.c. | |
1450 | * The major change is making write step of cold data page separately from | |
1451 | * warm/hot data page. | |
1452 | */ | |
1453 | static int f2fs_write_cache_pages(struct address_space *mapping, | |
b230e6ca | 1454 | struct writeback_control *wbc) |
8f46dcae CY |
1455 | { |
1456 | int ret = 0; | |
1457 | int done = 0; | |
1458 | struct pagevec pvec; | |
1459 | int nr_pages; | |
1460 | pgoff_t uninitialized_var(writeback_index); | |
1461 | pgoff_t index; | |
1462 | pgoff_t end; /* Inclusive */ | |
1463 | pgoff_t done_index; | |
1464 | int cycled; | |
1465 | int range_whole = 0; | |
1466 | int tag; | |
6ca56ca4 | 1467 | int nwritten = 0; |
8f46dcae CY |
1468 | |
1469 | pagevec_init(&pvec, 0); | |
46ae957f | 1470 | |
8f46dcae CY |
1471 | if (wbc->range_cyclic) { |
1472 | writeback_index = mapping->writeback_index; /* prev offset */ | |
1473 | index = writeback_index; | |
1474 | if (index == 0) | |
1475 | cycled = 1; | |
1476 | else | |
1477 | cycled = 0; | |
1478 | end = -1; | |
1479 | } else { | |
09cbfeaf KS |
1480 | index = wbc->range_start >> PAGE_SHIFT; |
1481 | end = wbc->range_end >> PAGE_SHIFT; | |
8f46dcae CY |
1482 | if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX) |
1483 | range_whole = 1; | |
1484 | cycled = 1; /* ignore range_cyclic tests */ | |
1485 | } | |
1486 | if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages) | |
1487 | tag = PAGECACHE_TAG_TOWRITE; | |
1488 | else | |
1489 | tag = PAGECACHE_TAG_DIRTY; | |
1490 | retry: | |
1491 | if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages) | |
1492 | tag_pages_for_writeback(mapping, index, end); | |
1493 | done_index = index; | |
1494 | while (!done && (index <= end)) { | |
1495 | int i; | |
1496 | ||
1497 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag, | |
1498 | min(end - index, (pgoff_t)PAGEVEC_SIZE - 1) + 1); | |
1499 | if (nr_pages == 0) | |
1500 | break; | |
1501 | ||
1502 | for (i = 0; i < nr_pages; i++) { | |
1503 | struct page *page = pvec.pages[i]; | |
1504 | ||
1505 | if (page->index > end) { | |
1506 | done = 1; | |
1507 | break; | |
1508 | } | |
1509 | ||
1510 | done_index = page->index; | |
1511 | ||
1512 | lock_page(page); | |
1513 | ||
1514 | if (unlikely(page->mapping != mapping)) { | |
1515 | continue_unlock: | |
1516 | unlock_page(page); | |
1517 | continue; | |
1518 | } | |
1519 | ||
1520 | if (!PageDirty(page)) { | |
1521 | /* someone wrote it for us */ | |
1522 | goto continue_unlock; | |
1523 | } | |
1524 | ||
8f46dcae CY |
1525 | if (PageWriteback(page)) { |
1526 | if (wbc->sync_mode != WB_SYNC_NONE) | |
fec1d657 JK |
1527 | f2fs_wait_on_page_writeback(page, |
1528 | DATA, true); | |
8f46dcae CY |
1529 | else |
1530 | goto continue_unlock; | |
1531 | } | |
1532 | ||
1533 | BUG_ON(PageWriteback(page)); | |
1534 | if (!clear_page_dirty_for_io(page)) | |
1535 | goto continue_unlock; | |
1536 | ||
b230e6ca | 1537 | ret = mapping->a_ops->writepage(page, wbc); |
8f46dcae | 1538 | if (unlikely(ret)) { |
0002b61b CY |
1539 | /* |
1540 | * keep nr_to_write, since vfs uses this to | |
1541 | * get # of written pages. | |
1542 | */ | |
1543 | if (ret == AOP_WRITEPAGE_ACTIVATE) { | |
1544 | unlock_page(page); | |
1545 | ret = 0; | |
1546 | continue; | |
1547 | } | |
b230e6ca JK |
1548 | done_index = page->index + 1; |
1549 | done = 1; | |
1550 | break; | |
6ca56ca4 CY |
1551 | } else { |
1552 | nwritten++; | |
8f46dcae CY |
1553 | } |
1554 | ||
1555 | if (--wbc->nr_to_write <= 0 && | |
1556 | wbc->sync_mode == WB_SYNC_NONE) { | |
1557 | done = 1; | |
1558 | break; | |
1559 | } | |
1560 | } | |
1561 | pagevec_release(&pvec); | |
1562 | cond_resched(); | |
1563 | } | |
1564 | ||
8f46dcae CY |
1565 | if (!cycled && !done) { |
1566 | cycled = 1; | |
1567 | index = 0; | |
1568 | end = writeback_index - 1; | |
1569 | goto retry; | |
1570 | } | |
1571 | if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) | |
1572 | mapping->writeback_index = done_index; | |
1573 | ||
6ca56ca4 CY |
1574 | if (nwritten) |
1575 | f2fs_submit_merged_bio_cond(F2FS_M_SB(mapping), mapping->host, | |
1576 | NULL, 0, DATA, WRITE); | |
1577 | ||
8f46dcae CY |
1578 | return ret; |
1579 | } | |
1580 | ||
25ca923b | 1581 | static int f2fs_write_data_pages(struct address_space *mapping, |
eb47b800 JK |
1582 | struct writeback_control *wbc) |
1583 | { | |
1584 | struct inode *inode = mapping->host; | |
4081363f | 1585 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
9dfa1baf | 1586 | struct blk_plug plug; |
eb47b800 | 1587 | int ret; |
eb47b800 | 1588 | |
cfb185a1 | 1589 | /* deal with chardevs and other special file */ |
1590 | if (!mapping->a_ops->writepage) | |
1591 | return 0; | |
1592 | ||
6a290544 CY |
1593 | /* skip writing if there is no dirty page in this inode */ |
1594 | if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE) | |
1595 | return 0; | |
1596 | ||
a1257023 JK |
1597 | if (S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_NONE && |
1598 | get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) && | |
1599 | available_free_memory(sbi, DIRTY_DENTS)) | |
1600 | goto skip_write; | |
1601 | ||
d323d005 | 1602 | /* skip writing during file defragment */ |
91942321 | 1603 | if (is_inode_flag_set(inode, FI_DO_DEFRAG)) |
d323d005 CY |
1604 | goto skip_write; |
1605 | ||
d5669f7b JK |
1606 | /* during POR, we don't need to trigger writepage at all. */ |
1607 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) | |
1608 | goto skip_write; | |
1609 | ||
d31c7c3f YH |
1610 | trace_f2fs_writepages(mapping->host, wbc, DATA); |
1611 | ||
9dfa1baf | 1612 | blk_start_plug(&plug); |
b230e6ca | 1613 | ret = f2fs_write_cache_pages(mapping, wbc); |
9dfa1baf | 1614 | blk_finish_plug(&plug); |
28ea6162 JK |
1615 | /* |
1616 | * if some pages were truncated, we cannot guarantee its mapping->host | |
1617 | * to detect pending bios. | |
1618 | */ | |
458e6197 | 1619 | |
c227f912 | 1620 | remove_dirty_inode(inode); |
eb47b800 | 1621 | return ret; |
d3baf95d JK |
1622 | |
1623 | skip_write: | |
a7ffdbe2 | 1624 | wbc->pages_skipped += get_dirty_pages(inode); |
d31c7c3f | 1625 | trace_f2fs_writepages(mapping->host, wbc, DATA); |
d3baf95d | 1626 | return 0; |
eb47b800 JK |
1627 | } |
1628 | ||
3aab8f82 CY |
1629 | static void f2fs_write_failed(struct address_space *mapping, loff_t to) |
1630 | { | |
1631 | struct inode *inode = mapping->host; | |
819d9153 | 1632 | loff_t i_size = i_size_read(inode); |
3aab8f82 | 1633 | |
819d9153 JK |
1634 | if (to > i_size) { |
1635 | truncate_pagecache(inode, i_size); | |
1636 | truncate_blocks(inode, i_size, true); | |
3aab8f82 CY |
1637 | } |
1638 | } | |
1639 | ||
2aadac08 JK |
1640 | static int prepare_write_begin(struct f2fs_sb_info *sbi, |
1641 | struct page *page, loff_t pos, unsigned len, | |
1642 | block_t *blk_addr, bool *node_changed) | |
1643 | { | |
1644 | struct inode *inode = page->mapping->host; | |
1645 | pgoff_t index = page->index; | |
1646 | struct dnode_of_data dn; | |
1647 | struct page *ipage; | |
b4d07a3e JK |
1648 | bool locked = false; |
1649 | struct extent_info ei; | |
2aadac08 JK |
1650 | int err = 0; |
1651 | ||
24b84912 JK |
1652 | /* |
1653 | * we already allocated all the blocks, so we don't need to get | |
1654 | * the block addresses when there is no need to fill the page. | |
1655 | */ | |
5d4c0af4 | 1656 | if (!f2fs_has_inline_data(inode) && len == PAGE_SIZE) |
24b84912 JK |
1657 | return 0; |
1658 | ||
b4d07a3e | 1659 | if (f2fs_has_inline_data(inode) || |
09cbfeaf | 1660 | (pos & PAGE_MASK) >= i_size_read(inode)) { |
b4d07a3e JK |
1661 | f2fs_lock_op(sbi); |
1662 | locked = true; | |
1663 | } | |
1664 | restart: | |
2aadac08 JK |
1665 | /* check inline_data */ |
1666 | ipage = get_node_page(sbi, inode->i_ino); | |
1667 | if (IS_ERR(ipage)) { | |
1668 | err = PTR_ERR(ipage); | |
1669 | goto unlock_out; | |
1670 | } | |
1671 | ||
1672 | set_new_dnode(&dn, inode, ipage, ipage, 0); | |
1673 | ||
1674 | if (f2fs_has_inline_data(inode)) { | |
1675 | if (pos + len <= MAX_INLINE_DATA) { | |
1676 | read_inline_data(page, ipage); | |
91942321 | 1677 | set_inode_flag(inode, FI_DATA_EXIST); |
ab47036d CY |
1678 | if (inode->i_nlink) |
1679 | set_inline_node(ipage); | |
2aadac08 JK |
1680 | } else { |
1681 | err = f2fs_convert_inline_page(&dn, page); | |
1682 | if (err) | |
b4d07a3e JK |
1683 | goto out; |
1684 | if (dn.data_blkaddr == NULL_ADDR) | |
1685 | err = f2fs_get_block(&dn, index); | |
1686 | } | |
1687 | } else if (locked) { | |
1688 | err = f2fs_get_block(&dn, index); | |
1689 | } else { | |
1690 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { | |
1691 | dn.data_blkaddr = ei.blk + index - ei.fofs; | |
1692 | } else { | |
b4d07a3e JK |
1693 | /* hole case */ |
1694 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); | |
4da7bf5a | 1695 | if (err || dn.data_blkaddr == NULL_ADDR) { |
b4d07a3e JK |
1696 | f2fs_put_dnode(&dn); |
1697 | f2fs_lock_op(sbi); | |
1698 | locked = true; | |
1699 | goto restart; | |
1700 | } | |
2aadac08 JK |
1701 | } |
1702 | } | |
b4d07a3e | 1703 | |
2aadac08 JK |
1704 | /* convert_inline_page can make node_changed */ |
1705 | *blk_addr = dn.data_blkaddr; | |
1706 | *node_changed = dn.node_changed; | |
b4d07a3e | 1707 | out: |
2aadac08 JK |
1708 | f2fs_put_dnode(&dn); |
1709 | unlock_out: | |
b4d07a3e JK |
1710 | if (locked) |
1711 | f2fs_unlock_op(sbi); | |
2aadac08 JK |
1712 | return err; |
1713 | } | |
1714 | ||
eb47b800 JK |
1715 | static int f2fs_write_begin(struct file *file, struct address_space *mapping, |
1716 | loff_t pos, unsigned len, unsigned flags, | |
1717 | struct page **pagep, void **fsdata) | |
1718 | { | |
1719 | struct inode *inode = mapping->host; | |
4081363f | 1720 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
86531d6b | 1721 | struct page *page = NULL; |
09cbfeaf | 1722 | pgoff_t index = ((unsigned long long) pos) >> PAGE_SHIFT; |
2aadac08 JK |
1723 | bool need_balance = false; |
1724 | block_t blkaddr = NULL_ADDR; | |
eb47b800 JK |
1725 | int err = 0; |
1726 | ||
62aed044 CY |
1727 | trace_f2fs_write_begin(inode, pos, len, flags); |
1728 | ||
5f727395 JK |
1729 | /* |
1730 | * We should check this at this moment to avoid deadlock on inode page | |
1731 | * and #0 page. The locking rule for inline_data conversion should be: | |
1732 | * lock_page(page #0) -> lock_page(inode_page) | |
1733 | */ | |
1734 | if (index != 0) { | |
1735 | err = f2fs_convert_inline_inode(inode); | |
1736 | if (err) | |
1737 | goto fail; | |
1738 | } | |
afcb7ca0 | 1739 | repeat: |
eb47b800 | 1740 | page = grab_cache_page_write_begin(mapping, index, flags); |
3aab8f82 CY |
1741 | if (!page) { |
1742 | err = -ENOMEM; | |
1743 | goto fail; | |
1744 | } | |
d5f66990 | 1745 | |
eb47b800 JK |
1746 | *pagep = page; |
1747 | ||
2aadac08 JK |
1748 | err = prepare_write_begin(sbi, page, pos, len, |
1749 | &blkaddr, &need_balance); | |
9ba69cf9 | 1750 | if (err) |
2aadac08 | 1751 | goto fail; |
9ba69cf9 | 1752 | |
7f3037a5 | 1753 | if (need_balance && has_not_enough_free_secs(sbi, 0, 0)) { |
2a340760 | 1754 | unlock_page(page); |
2c4db1a6 | 1755 | f2fs_balance_fs(sbi, true); |
2a340760 JK |
1756 | lock_page(page); |
1757 | if (page->mapping != mapping) { | |
1758 | /* The page got truncated from under us */ | |
1759 | f2fs_put_page(page, 1); | |
1760 | goto repeat; | |
1761 | } | |
1762 | } | |
1763 | ||
fec1d657 | 1764 | f2fs_wait_on_page_writeback(page, DATA, false); |
b3d208f9 | 1765 | |
08b39fbd CY |
1766 | /* wait for GCed encrypted page writeback */ |
1767 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) | |
2aadac08 | 1768 | f2fs_wait_on_encrypted_page_writeback(sbi, blkaddr); |
08b39fbd | 1769 | |
649d7df2 JK |
1770 | if (len == PAGE_SIZE || PageUptodate(page)) |
1771 | return 0; | |
eb47b800 | 1772 | |
746e2403 YH |
1773 | if (!(pos & (PAGE_SIZE - 1)) && (pos + len) >= i_size_read(inode)) { |
1774 | zero_user_segment(page, len, PAGE_SIZE); | |
1775 | return 0; | |
1776 | } | |
1777 | ||
2aadac08 | 1778 | if (blkaddr == NEW_ADDR) { |
09cbfeaf | 1779 | zero_user_segment(page, 0, PAGE_SIZE); |
649d7df2 | 1780 | SetPageUptodate(page); |
eb47b800 | 1781 | } else { |
78682f79 | 1782 | struct bio *bio; |
d54c795b | 1783 | |
78682f79 CY |
1784 | bio = f2fs_grab_bio(inode, blkaddr, 1); |
1785 | if (IS_ERR(bio)) { | |
1786 | err = PTR_ERR(bio); | |
3aab8f82 | 1787 | goto fail; |
eb47b800 | 1788 | } |
70fd7614 | 1789 | bio->bi_opf = REQ_OP_READ; |
78682f79 CY |
1790 | if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) { |
1791 | bio_put(bio); | |
1792 | err = -EFAULT; | |
1793 | goto fail; | |
1794 | } | |
1795 | ||
4fc29c1a | 1796 | __submit_bio(sbi, bio, DATA); |
d54c795b | 1797 | |
393ff91f | 1798 | lock_page(page); |
6bacf52f | 1799 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
1800 | f2fs_put_page(page, 1); |
1801 | goto repeat; | |
eb47b800 | 1802 | } |
1563ac75 CY |
1803 | if (unlikely(!PageUptodate(page))) { |
1804 | err = -EIO; | |
1805 | goto fail; | |
4375a336 | 1806 | } |
eb47b800 | 1807 | } |
eb47b800 | 1808 | return 0; |
9ba69cf9 | 1809 | |
3aab8f82 | 1810 | fail: |
86531d6b | 1811 | f2fs_put_page(page, 1); |
3aab8f82 CY |
1812 | f2fs_write_failed(mapping, pos + len); |
1813 | return err; | |
eb47b800 JK |
1814 | } |
1815 | ||
a1dd3c13 JK |
1816 | static int f2fs_write_end(struct file *file, |
1817 | struct address_space *mapping, | |
1818 | loff_t pos, unsigned len, unsigned copied, | |
1819 | struct page *page, void *fsdata) | |
1820 | { | |
1821 | struct inode *inode = page->mapping->host; | |
1822 | ||
dfb2bf38 CY |
1823 | trace_f2fs_write_end(inode, pos, len, copied); |
1824 | ||
649d7df2 JK |
1825 | /* |
1826 | * This should be come from len == PAGE_SIZE, and we expect copied | |
1827 | * should be PAGE_SIZE. Otherwise, we treat it with zero copied and | |
1828 | * let generic_perform_write() try to copy data again through copied=0. | |
1829 | */ | |
1830 | if (!PageUptodate(page)) { | |
746e2403 | 1831 | if (unlikely(copied != len)) |
649d7df2 JK |
1832 | copied = 0; |
1833 | else | |
1834 | SetPageUptodate(page); | |
1835 | } | |
1836 | if (!copied) | |
1837 | goto unlock_out; | |
1838 | ||
34ba94ba | 1839 | set_page_dirty(page); |
a1dd3c13 | 1840 | |
fc9581c8 JK |
1841 | if (pos + copied > i_size_read(inode)) |
1842 | f2fs_i_size_write(inode, pos + copied); | |
649d7df2 | 1843 | unlock_out: |
3024c9a1 | 1844 | f2fs_put_page(page, 1); |
d0239e1b | 1845 | f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); |
a1dd3c13 JK |
1846 | return copied; |
1847 | } | |
1848 | ||
6f673763 OS |
1849 | static int check_direct_IO(struct inode *inode, struct iov_iter *iter, |
1850 | loff_t offset) | |
944fcfc1 JK |
1851 | { |
1852 | unsigned blocksize_mask = inode->i_sb->s_blocksize - 1; | |
944fcfc1 | 1853 | |
944fcfc1 JK |
1854 | if (offset & blocksize_mask) |
1855 | return -EINVAL; | |
1856 | ||
5b46f25d AV |
1857 | if (iov_iter_alignment(iter) & blocksize_mask) |
1858 | return -EINVAL; | |
1859 | ||
944fcfc1 JK |
1860 | return 0; |
1861 | } | |
1862 | ||
c8b8e32d | 1863 | static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter) |
eb47b800 | 1864 | { |
b439b103 | 1865 | struct address_space *mapping = iocb->ki_filp->f_mapping; |
3aab8f82 CY |
1866 | struct inode *inode = mapping->host; |
1867 | size_t count = iov_iter_count(iter); | |
c8b8e32d | 1868 | loff_t offset = iocb->ki_pos; |
82e0a5aa | 1869 | int rw = iov_iter_rw(iter); |
3aab8f82 | 1870 | int err; |
944fcfc1 | 1871 | |
b439b103 | 1872 | err = check_direct_IO(inode, iter, offset); |
b9d777b8 JK |
1873 | if (err) |
1874 | return err; | |
9ffe0fb5 | 1875 | |
c040ff9d | 1876 | if (__force_buffered_io(inode, rw)) |
36abef4e | 1877 | return 0; |
fcc85a4d | 1878 | |
5302fb00 | 1879 | trace_f2fs_direct_IO_enter(inode, offset, count, rw); |
70407fad | 1880 | |
82e0a5aa | 1881 | down_read(&F2FS_I(inode)->dio_rwsem[rw]); |
c8b8e32d | 1882 | err = blockdev_direct_IO(iocb, inode, iter, get_data_block_dio); |
82e0a5aa CY |
1883 | up_read(&F2FS_I(inode)->dio_rwsem[rw]); |
1884 | ||
1885 | if (rw == WRITE) { | |
6bfc4919 | 1886 | if (err > 0) |
91942321 | 1887 | set_inode_flag(inode, FI_UPDATE_WRITE); |
6bfc4919 JK |
1888 | else if (err < 0) |
1889 | f2fs_write_failed(mapping, offset + count); | |
1890 | } | |
70407fad | 1891 | |
5302fb00 | 1892 | trace_f2fs_direct_IO_exit(inode, offset, count, rw, err); |
70407fad | 1893 | |
3aab8f82 | 1894 | return err; |
eb47b800 JK |
1895 | } |
1896 | ||
487261f3 CY |
1897 | void f2fs_invalidate_page(struct page *page, unsigned int offset, |
1898 | unsigned int length) | |
eb47b800 JK |
1899 | { |
1900 | struct inode *inode = page->mapping->host; | |
487261f3 | 1901 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
a7ffdbe2 | 1902 | |
487261f3 | 1903 | if (inode->i_ino >= F2FS_ROOT_INO(sbi) && |
09cbfeaf | 1904 | (offset % PAGE_SIZE || length != PAGE_SIZE)) |
a7ffdbe2 JK |
1905 | return; |
1906 | ||
487261f3 | 1907 | if (PageDirty(page)) { |
933439c8 | 1908 | if (inode->i_ino == F2FS_META_INO(sbi)) { |
487261f3 | 1909 | dec_page_count(sbi, F2FS_DIRTY_META); |
933439c8 | 1910 | } else if (inode->i_ino == F2FS_NODE_INO(sbi)) { |
487261f3 | 1911 | dec_page_count(sbi, F2FS_DIRTY_NODES); |
933439c8 | 1912 | } else { |
487261f3 | 1913 | inode_dec_dirty_pages(inode); |
933439c8 CY |
1914 | remove_dirty_inode(inode); |
1915 | } | |
487261f3 | 1916 | } |
decd36b6 CY |
1917 | |
1918 | /* This is atomic written page, keep Private */ | |
1919 | if (IS_ATOMIC_WRITTEN_PAGE(page)) | |
1920 | return; | |
1921 | ||
23dc974e | 1922 | set_page_private(page, 0); |
eb47b800 JK |
1923 | ClearPagePrivate(page); |
1924 | } | |
1925 | ||
487261f3 | 1926 | int f2fs_release_page(struct page *page, gfp_t wait) |
eb47b800 | 1927 | { |
f68daeeb JK |
1928 | /* If this is dirty page, keep PagePrivate */ |
1929 | if (PageDirty(page)) | |
1930 | return 0; | |
1931 | ||
decd36b6 CY |
1932 | /* This is atomic written page, keep Private */ |
1933 | if (IS_ATOMIC_WRITTEN_PAGE(page)) | |
1934 | return 0; | |
1935 | ||
23dc974e | 1936 | set_page_private(page, 0); |
eb47b800 | 1937 | ClearPagePrivate(page); |
c3850aa1 | 1938 | return 1; |
eb47b800 JK |
1939 | } |
1940 | ||
fe76b796 JK |
1941 | /* |
1942 | * This was copied from __set_page_dirty_buffers which gives higher performance | |
1943 | * in very high speed storages. (e.g., pmem) | |
1944 | */ | |
1945 | void f2fs_set_page_dirty_nobuffers(struct page *page) | |
1946 | { | |
1947 | struct address_space *mapping = page->mapping; | |
1948 | unsigned long flags; | |
1949 | ||
1950 | if (unlikely(!mapping)) | |
1951 | return; | |
1952 | ||
1953 | spin_lock(&mapping->private_lock); | |
1954 | lock_page_memcg(page); | |
1955 | SetPageDirty(page); | |
1956 | spin_unlock(&mapping->private_lock); | |
1957 | ||
1958 | spin_lock_irqsave(&mapping->tree_lock, flags); | |
1959 | WARN_ON_ONCE(!PageUptodate(page)); | |
1960 | account_page_dirtied(page, mapping); | |
1961 | radix_tree_tag_set(&mapping->page_tree, | |
1962 | page_index(page), PAGECACHE_TAG_DIRTY); | |
1963 | spin_unlock_irqrestore(&mapping->tree_lock, flags); | |
1964 | unlock_page_memcg(page); | |
1965 | ||
1966 | __mark_inode_dirty(mapping->host, I_DIRTY_PAGES); | |
1967 | return; | |
1968 | } | |
1969 | ||
eb47b800 JK |
1970 | static int f2fs_set_data_page_dirty(struct page *page) |
1971 | { | |
1972 | struct address_space *mapping = page->mapping; | |
1973 | struct inode *inode = mapping->host; | |
1974 | ||
26c6b887 JK |
1975 | trace_f2fs_set_page_dirty(page, DATA); |
1976 | ||
237c0790 JK |
1977 | if (!PageUptodate(page)) |
1978 | SetPageUptodate(page); | |
34ba94ba | 1979 | |
1e84371f | 1980 | if (f2fs_is_atomic_file(inode)) { |
decd36b6 CY |
1981 | if (!IS_ATOMIC_WRITTEN_PAGE(page)) { |
1982 | register_inmem_page(inode, page); | |
1983 | return 1; | |
1984 | } | |
1985 | /* | |
1986 | * Previously, this page has been registered, we just | |
1987 | * return here. | |
1988 | */ | |
1989 | return 0; | |
34ba94ba JK |
1990 | } |
1991 | ||
eb47b800 | 1992 | if (!PageDirty(page)) { |
fe76b796 | 1993 | f2fs_set_page_dirty_nobuffers(page); |
a7ffdbe2 | 1994 | update_dirty_page(inode, page); |
eb47b800 JK |
1995 | return 1; |
1996 | } | |
1997 | return 0; | |
1998 | } | |
1999 | ||
c01e54b7 JK |
2000 | static sector_t f2fs_bmap(struct address_space *mapping, sector_t block) |
2001 | { | |
454ae7e5 CY |
2002 | struct inode *inode = mapping->host; |
2003 | ||
1d373a0e JK |
2004 | if (f2fs_has_inline_data(inode)) |
2005 | return 0; | |
2006 | ||
2007 | /* make sure allocating whole blocks */ | |
2008 | if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) | |
2009 | filemap_write_and_wait(mapping); | |
2010 | ||
e2b4e2bc | 2011 | return generic_block_bmap(mapping, block, get_data_block_bmap); |
429511cd CY |
2012 | } |
2013 | ||
5b7a487c WG |
2014 | #ifdef CONFIG_MIGRATION |
2015 | #include <linux/migrate.h> | |
2016 | ||
2017 | int f2fs_migrate_page(struct address_space *mapping, | |
2018 | struct page *newpage, struct page *page, enum migrate_mode mode) | |
2019 | { | |
2020 | int rc, extra_count; | |
2021 | struct f2fs_inode_info *fi = F2FS_I(mapping->host); | |
2022 | bool atomic_written = IS_ATOMIC_WRITTEN_PAGE(page); | |
2023 | ||
2024 | BUG_ON(PageWriteback(page)); | |
2025 | ||
2026 | /* migrating an atomic written page is safe with the inmem_lock hold */ | |
2027 | if (atomic_written && !mutex_trylock(&fi->inmem_lock)) | |
2028 | return -EAGAIN; | |
2029 | ||
2030 | /* | |
2031 | * A reference is expected if PagePrivate set when move mapping, | |
2032 | * however F2FS breaks this for maintaining dirty page counts when | |
2033 | * truncating pages. So here adjusting the 'extra_count' make it work. | |
2034 | */ | |
2035 | extra_count = (atomic_written ? 1 : 0) - page_has_private(page); | |
2036 | rc = migrate_page_move_mapping(mapping, newpage, | |
2037 | page, NULL, mode, extra_count); | |
2038 | if (rc != MIGRATEPAGE_SUCCESS) { | |
2039 | if (atomic_written) | |
2040 | mutex_unlock(&fi->inmem_lock); | |
2041 | return rc; | |
2042 | } | |
2043 | ||
2044 | if (atomic_written) { | |
2045 | struct inmem_pages *cur; | |
2046 | list_for_each_entry(cur, &fi->inmem_pages, list) | |
2047 | if (cur->page == page) { | |
2048 | cur->page = newpage; | |
2049 | break; | |
2050 | } | |
2051 | mutex_unlock(&fi->inmem_lock); | |
2052 | put_page(page); | |
2053 | get_page(newpage); | |
2054 | } | |
2055 | ||
2056 | if (PagePrivate(page)) | |
2057 | SetPagePrivate(newpage); | |
2058 | set_page_private(newpage, page_private(page)); | |
2059 | ||
2060 | migrate_page_copy(newpage, page); | |
2061 | ||
2062 | return MIGRATEPAGE_SUCCESS; | |
2063 | } | |
2064 | #endif | |
2065 | ||
eb47b800 JK |
2066 | const struct address_space_operations f2fs_dblock_aops = { |
2067 | .readpage = f2fs_read_data_page, | |
2068 | .readpages = f2fs_read_data_pages, | |
2069 | .writepage = f2fs_write_data_page, | |
2070 | .writepages = f2fs_write_data_pages, | |
2071 | .write_begin = f2fs_write_begin, | |
a1dd3c13 | 2072 | .write_end = f2fs_write_end, |
eb47b800 | 2073 | .set_page_dirty = f2fs_set_data_page_dirty, |
487261f3 CY |
2074 | .invalidatepage = f2fs_invalidate_page, |
2075 | .releasepage = f2fs_release_page, | |
eb47b800 | 2076 | .direct_IO = f2fs_direct_IO, |
c01e54b7 | 2077 | .bmap = f2fs_bmap, |
5b7a487c WG |
2078 | #ifdef CONFIG_MIGRATION |
2079 | .migratepage = f2fs_migrate_page, | |
2080 | #endif | |
eb47b800 | 2081 | }; |