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> | |
a27bb332 | 15 | #include <linux/aio.h> |
eb47b800 JK |
16 | #include <linux/writeback.h> |
17 | #include <linux/backing-dev.h> | |
18 | #include <linux/blkdev.h> | |
19 | #include <linux/bio.h> | |
690e4a3e | 20 | #include <linux/prefetch.h> |
eb47b800 JK |
21 | |
22 | #include "f2fs.h" | |
23 | #include "node.h" | |
24 | #include "segment.h" | |
848753aa | 25 | #include <trace/events/f2fs.h> |
eb47b800 | 26 | |
93dfe2ac JK |
27 | static void f2fs_read_end_io(struct bio *bio, int err) |
28 | { | |
f568849e LT |
29 | struct bio_vec *bvec; |
30 | int i; | |
93dfe2ac | 31 | |
f568849e | 32 | bio_for_each_segment_all(bvec, bio, i) { |
93dfe2ac JK |
33 | struct page *page = bvec->bv_page; |
34 | ||
f568849e LT |
35 | if (!err) { |
36 | SetPageUptodate(page); | |
37 | } else { | |
93dfe2ac JK |
38 | ClearPageUptodate(page); |
39 | SetPageError(page); | |
40 | } | |
41 | unlock_page(page); | |
f568849e | 42 | } |
93dfe2ac JK |
43 | bio_put(bio); |
44 | } | |
45 | ||
46 | static void f2fs_write_end_io(struct bio *bio, int err) | |
47 | { | |
1b1f559f | 48 | struct f2fs_sb_info *sbi = bio->bi_private; |
f568849e LT |
49 | struct bio_vec *bvec; |
50 | int i; | |
93dfe2ac | 51 | |
f568849e | 52 | bio_for_each_segment_all(bvec, bio, i) { |
93dfe2ac JK |
53 | struct page *page = bvec->bv_page; |
54 | ||
f568849e | 55 | if (unlikely(err)) { |
cf779cab | 56 | set_page_dirty(page); |
93dfe2ac | 57 | set_bit(AS_EIO, &page->mapping->flags); |
744602cf | 58 | f2fs_stop_checkpoint(sbi); |
93dfe2ac JK |
59 | } |
60 | end_page_writeback(page); | |
61 | dec_page_count(sbi, F2FS_WRITEBACK); | |
f568849e | 62 | } |
93dfe2ac | 63 | |
1b1f559f JK |
64 | if (sbi->wait_io) { |
65 | complete(sbi->wait_io); | |
66 | sbi->wait_io = NULL; | |
67 | } | |
93dfe2ac JK |
68 | |
69 | if (!get_pages(sbi, F2FS_WRITEBACK) && | |
70 | !list_empty(&sbi->cp_wait.task_list)) | |
71 | wake_up(&sbi->cp_wait); | |
72 | ||
73 | bio_put(bio); | |
74 | } | |
75 | ||
940a6d34 GZ |
76 | /* |
77 | * Low-level block read/write IO operations. | |
78 | */ | |
79 | static struct bio *__bio_alloc(struct f2fs_sb_info *sbi, block_t blk_addr, | |
80 | int npages, bool is_read) | |
81 | { | |
82 | struct bio *bio; | |
83 | ||
84 | /* No failure on bio allocation */ | |
85 | bio = bio_alloc(GFP_NOIO, npages); | |
86 | ||
87 | bio->bi_bdev = sbi->sb->s_bdev; | |
55cf9cb6 | 88 | bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr); |
940a6d34 | 89 | bio->bi_end_io = is_read ? f2fs_read_end_io : f2fs_write_end_io; |
1b1f559f | 90 | bio->bi_private = sbi; |
940a6d34 GZ |
91 | |
92 | return bio; | |
93 | } | |
94 | ||
458e6197 | 95 | static void __submit_merged_bio(struct f2fs_bio_info *io) |
93dfe2ac | 96 | { |
458e6197 JK |
97 | struct f2fs_io_info *fio = &io->fio; |
98 | int rw; | |
93dfe2ac JK |
99 | |
100 | if (!io->bio) | |
101 | return; | |
102 | ||
7e8f2308 | 103 | rw = fio->rw; |
93dfe2ac JK |
104 | |
105 | if (is_read_io(rw)) { | |
940a6d34 GZ |
106 | trace_f2fs_submit_read_bio(io->sbi->sb, rw, |
107 | fio->type, io->bio); | |
76130cca | 108 | submit_bio(rw, io->bio); |
93dfe2ac | 109 | } else { |
940a6d34 GZ |
110 | trace_f2fs_submit_write_bio(io->sbi->sb, rw, |
111 | fio->type, io->bio); | |
112 | /* | |
113 | * META_FLUSH is only from the checkpoint procedure, and we | |
114 | * should wait this metadata bio for FS consistency. | |
115 | */ | |
116 | if (fio->type == META_FLUSH) { | |
117 | DECLARE_COMPLETION_ONSTACK(wait); | |
1b1f559f | 118 | io->sbi->wait_io = &wait; |
940a6d34 GZ |
119 | submit_bio(rw, io->bio); |
120 | wait_for_completion(&wait); | |
121 | } else { | |
122 | submit_bio(rw, io->bio); | |
123 | } | |
93dfe2ac | 124 | } |
940a6d34 | 125 | |
93dfe2ac JK |
126 | io->bio = NULL; |
127 | } | |
128 | ||
129 | void f2fs_submit_merged_bio(struct f2fs_sb_info *sbi, | |
458e6197 | 130 | enum page_type type, int rw) |
93dfe2ac JK |
131 | { |
132 | enum page_type btype = PAGE_TYPE_OF_BIO(type); | |
133 | struct f2fs_bio_info *io; | |
134 | ||
135 | io = is_read_io(rw) ? &sbi->read_io : &sbi->write_io[btype]; | |
136 | ||
df0f8dc0 | 137 | down_write(&io->io_rwsem); |
458e6197 JK |
138 | |
139 | /* change META to META_FLUSH in the checkpoint procedure */ | |
140 | if (type >= META_FLUSH) { | |
141 | io->fio.type = META_FLUSH; | |
0f7b2abd JK |
142 | if (test_opt(sbi, NOBARRIER)) |
143 | io->fio.rw = WRITE_FLUSH | REQ_META | REQ_PRIO; | |
144 | else | |
145 | io->fio.rw = WRITE_FLUSH_FUA | REQ_META | REQ_PRIO; | |
458e6197 JK |
146 | } |
147 | __submit_merged_bio(io); | |
df0f8dc0 | 148 | up_write(&io->io_rwsem); |
93dfe2ac JK |
149 | } |
150 | ||
151 | /* | |
152 | * Fill the locked page with data located in the block address. | |
153 | * Return unlocked page. | |
154 | */ | |
155 | int f2fs_submit_page_bio(struct f2fs_sb_info *sbi, struct page *page, | |
156 | block_t blk_addr, int rw) | |
157 | { | |
93dfe2ac JK |
158 | struct bio *bio; |
159 | ||
160 | trace_f2fs_submit_page_bio(page, blk_addr, rw); | |
161 | ||
162 | /* Allocate a new bio */ | |
940a6d34 | 163 | bio = __bio_alloc(sbi, blk_addr, 1, is_read_io(rw)); |
93dfe2ac JK |
164 | |
165 | if (bio_add_page(bio, page, PAGE_CACHE_SIZE, 0) < PAGE_CACHE_SIZE) { | |
166 | bio_put(bio); | |
167 | f2fs_put_page(page, 1); | |
168 | return -EFAULT; | |
169 | } | |
170 | ||
171 | submit_bio(rw, bio); | |
172 | return 0; | |
173 | } | |
174 | ||
175 | void f2fs_submit_page_mbio(struct f2fs_sb_info *sbi, struct page *page, | |
458e6197 | 176 | block_t blk_addr, struct f2fs_io_info *fio) |
93dfe2ac | 177 | { |
458e6197 | 178 | enum page_type btype = PAGE_TYPE_OF_BIO(fio->type); |
93dfe2ac | 179 | struct f2fs_bio_info *io; |
940a6d34 | 180 | bool is_read = is_read_io(fio->rw); |
93dfe2ac | 181 | |
940a6d34 | 182 | io = is_read ? &sbi->read_io : &sbi->write_io[btype]; |
93dfe2ac JK |
183 | |
184 | verify_block_addr(sbi, blk_addr); | |
185 | ||
df0f8dc0 | 186 | down_write(&io->io_rwsem); |
93dfe2ac | 187 | |
940a6d34 | 188 | if (!is_read) |
93dfe2ac JK |
189 | inc_page_count(sbi, F2FS_WRITEBACK); |
190 | ||
63a0b7cb | 191 | if (io->bio && (io->last_block_in_bio != blk_addr - 1 || |
458e6197 JK |
192 | io->fio.rw != fio->rw)) |
193 | __submit_merged_bio(io); | |
93dfe2ac JK |
194 | alloc_new: |
195 | if (io->bio == NULL) { | |
90a893c7 | 196 | int bio_blocks = MAX_BIO_BLOCKS(sbi); |
940a6d34 GZ |
197 | |
198 | io->bio = __bio_alloc(sbi, blk_addr, bio_blocks, is_read); | |
458e6197 | 199 | io->fio = *fio; |
93dfe2ac JK |
200 | } |
201 | ||
202 | if (bio_add_page(io->bio, page, PAGE_CACHE_SIZE, 0) < | |
203 | PAGE_CACHE_SIZE) { | |
458e6197 | 204 | __submit_merged_bio(io); |
93dfe2ac JK |
205 | goto alloc_new; |
206 | } | |
207 | ||
208 | io->last_block_in_bio = blk_addr; | |
209 | ||
df0f8dc0 | 210 | up_write(&io->io_rwsem); |
458e6197 | 211 | trace_f2fs_submit_page_mbio(page, fio->rw, fio->type, blk_addr); |
93dfe2ac JK |
212 | } |
213 | ||
0a8165d7 | 214 | /* |
eb47b800 JK |
215 | * Lock ordering for the change of data block address: |
216 | * ->data_page | |
217 | * ->node_page | |
218 | * update block addresses in the node page | |
219 | */ | |
220 | static void __set_data_blkaddr(struct dnode_of_data *dn, block_t new_addr) | |
221 | { | |
222 | struct f2fs_node *rn; | |
223 | __le32 *addr_array; | |
224 | struct page *node_page = dn->node_page; | |
225 | unsigned int ofs_in_node = dn->ofs_in_node; | |
226 | ||
5514f0aa | 227 | f2fs_wait_on_page_writeback(node_page, NODE); |
eb47b800 | 228 | |
45590710 | 229 | rn = F2FS_NODE(node_page); |
eb47b800 JK |
230 | |
231 | /* Get physical address of data block */ | |
232 | addr_array = blkaddr_in_node(rn); | |
233 | addr_array[ofs_in_node] = cpu_to_le32(new_addr); | |
234 | set_page_dirty(node_page); | |
235 | } | |
236 | ||
237 | int reserve_new_block(struct dnode_of_data *dn) | |
238 | { | |
4081363f | 239 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
eb47b800 | 240 | |
6bacf52f | 241 | if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))) |
eb47b800 | 242 | return -EPERM; |
cfb271d4 | 243 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, 1))) |
eb47b800 JK |
244 | return -ENOSPC; |
245 | ||
c01e2853 NJ |
246 | trace_f2fs_reserve_new_block(dn->inode, dn->nid, dn->ofs_in_node); |
247 | ||
eb47b800 JK |
248 | __set_data_blkaddr(dn, NEW_ADDR); |
249 | dn->data_blkaddr = NEW_ADDR; | |
a18ff063 | 250 | mark_inode_dirty(dn->inode); |
eb47b800 JK |
251 | sync_inode_page(dn); |
252 | return 0; | |
253 | } | |
254 | ||
b600965c HL |
255 | int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index) |
256 | { | |
257 | bool need_put = dn->inode_page ? false : true; | |
258 | int err; | |
259 | ||
260 | err = get_dnode_of_data(dn, index, ALLOC_NODE); | |
261 | if (err) | |
262 | return err; | |
a8865372 | 263 | |
b600965c HL |
264 | if (dn->data_blkaddr == NULL_ADDR) |
265 | err = reserve_new_block(dn); | |
a8865372 | 266 | if (err || need_put) |
b600965c HL |
267 | f2fs_put_dnode(dn); |
268 | return err; | |
269 | } | |
270 | ||
eb47b800 JK |
271 | static int check_extent_cache(struct inode *inode, pgoff_t pgofs, |
272 | struct buffer_head *bh_result) | |
273 | { | |
274 | struct f2fs_inode_info *fi = F2FS_I(inode); | |
eb47b800 JK |
275 | pgoff_t start_fofs, end_fofs; |
276 | block_t start_blkaddr; | |
277 | ||
c11abd1a JK |
278 | if (is_inode_flag_set(fi, FI_NO_EXTENT)) |
279 | return 0; | |
280 | ||
eb47b800 JK |
281 | read_lock(&fi->ext.ext_lock); |
282 | if (fi->ext.len == 0) { | |
283 | read_unlock(&fi->ext.ext_lock); | |
284 | return 0; | |
285 | } | |
286 | ||
dcdfff65 JK |
287 | stat_inc_total_hit(inode->i_sb); |
288 | ||
eb47b800 JK |
289 | start_fofs = fi->ext.fofs; |
290 | end_fofs = fi->ext.fofs + fi->ext.len - 1; | |
291 | start_blkaddr = fi->ext.blk_addr; | |
292 | ||
293 | if (pgofs >= start_fofs && pgofs <= end_fofs) { | |
294 | unsigned int blkbits = inode->i_sb->s_blocksize_bits; | |
295 | size_t count; | |
296 | ||
297 | clear_buffer_new(bh_result); | |
298 | map_bh(bh_result, inode->i_sb, | |
299 | start_blkaddr + pgofs - start_fofs); | |
300 | count = end_fofs - pgofs + 1; | |
301 | if (count < (UINT_MAX >> blkbits)) | |
302 | bh_result->b_size = (count << blkbits); | |
303 | else | |
304 | bh_result->b_size = UINT_MAX; | |
305 | ||
dcdfff65 | 306 | stat_inc_read_hit(inode->i_sb); |
eb47b800 JK |
307 | read_unlock(&fi->ext.ext_lock); |
308 | return 1; | |
309 | } | |
310 | read_unlock(&fi->ext.ext_lock); | |
311 | return 0; | |
312 | } | |
313 | ||
314 | void update_extent_cache(block_t blk_addr, struct dnode_of_data *dn) | |
315 | { | |
316 | struct f2fs_inode_info *fi = F2FS_I(dn->inode); | |
317 | pgoff_t fofs, start_fofs, end_fofs; | |
318 | block_t start_blkaddr, end_blkaddr; | |
c11abd1a | 319 | int need_update = true; |
eb47b800 | 320 | |
9850cf4a | 321 | f2fs_bug_on(F2FS_I_SB(dn->inode), blk_addr == NEW_ADDR); |
de93653f JK |
322 | fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) + |
323 | dn->ofs_in_node; | |
eb47b800 JK |
324 | |
325 | /* Update the page address in the parent node */ | |
326 | __set_data_blkaddr(dn, blk_addr); | |
327 | ||
c11abd1a JK |
328 | if (is_inode_flag_set(fi, FI_NO_EXTENT)) |
329 | return; | |
330 | ||
eb47b800 JK |
331 | write_lock(&fi->ext.ext_lock); |
332 | ||
333 | start_fofs = fi->ext.fofs; | |
334 | end_fofs = fi->ext.fofs + fi->ext.len - 1; | |
335 | start_blkaddr = fi->ext.blk_addr; | |
336 | end_blkaddr = fi->ext.blk_addr + fi->ext.len - 1; | |
337 | ||
338 | /* Drop and initialize the matched extent */ | |
339 | if (fi->ext.len == 1 && fofs == start_fofs) | |
340 | fi->ext.len = 0; | |
341 | ||
342 | /* Initial extent */ | |
343 | if (fi->ext.len == 0) { | |
344 | if (blk_addr != NULL_ADDR) { | |
345 | fi->ext.fofs = fofs; | |
346 | fi->ext.blk_addr = blk_addr; | |
347 | fi->ext.len = 1; | |
348 | } | |
349 | goto end_update; | |
350 | } | |
351 | ||
6224da87 | 352 | /* Front merge */ |
eb47b800 JK |
353 | if (fofs == start_fofs - 1 && blk_addr == start_blkaddr - 1) { |
354 | fi->ext.fofs--; | |
355 | fi->ext.blk_addr--; | |
356 | fi->ext.len++; | |
357 | goto end_update; | |
358 | } | |
359 | ||
360 | /* Back merge */ | |
361 | if (fofs == end_fofs + 1 && blk_addr == end_blkaddr + 1) { | |
362 | fi->ext.len++; | |
363 | goto end_update; | |
364 | } | |
365 | ||
366 | /* Split the existing extent */ | |
367 | if (fi->ext.len > 1 && | |
368 | fofs >= start_fofs && fofs <= end_fofs) { | |
369 | if ((end_fofs - fofs) < (fi->ext.len >> 1)) { | |
370 | fi->ext.len = fofs - start_fofs; | |
371 | } else { | |
372 | fi->ext.fofs = fofs + 1; | |
373 | fi->ext.blk_addr = start_blkaddr + | |
374 | fofs - start_fofs + 1; | |
375 | fi->ext.len -= fofs - start_fofs + 1; | |
376 | } | |
c11abd1a JK |
377 | } else { |
378 | need_update = false; | |
eb47b800 | 379 | } |
eb47b800 | 380 | |
c11abd1a JK |
381 | /* Finally, if the extent is very fragmented, let's drop the cache. */ |
382 | if (fi->ext.len < F2FS_MIN_EXTENT_LEN) { | |
383 | fi->ext.len = 0; | |
384 | set_inode_flag(fi, FI_NO_EXTENT); | |
385 | need_update = true; | |
386 | } | |
eb47b800 JK |
387 | end_update: |
388 | write_unlock(&fi->ext.ext_lock); | |
c11abd1a JK |
389 | if (need_update) |
390 | sync_inode_page(dn); | |
391 | return; | |
eb47b800 JK |
392 | } |
393 | ||
c718379b | 394 | struct page *find_data_page(struct inode *inode, pgoff_t index, bool sync) |
eb47b800 | 395 | { |
eb47b800 JK |
396 | struct address_space *mapping = inode->i_mapping; |
397 | struct dnode_of_data dn; | |
398 | struct page *page; | |
399 | int err; | |
400 | ||
401 | page = find_get_page(mapping, index); | |
402 | if (page && PageUptodate(page)) | |
403 | return page; | |
404 | f2fs_put_page(page, 0); | |
405 | ||
406 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
266e97a8 | 407 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); |
eb47b800 JK |
408 | if (err) |
409 | return ERR_PTR(err); | |
410 | f2fs_put_dnode(&dn); | |
411 | ||
412 | if (dn.data_blkaddr == NULL_ADDR) | |
413 | return ERR_PTR(-ENOENT); | |
414 | ||
415 | /* By fallocate(), there is no cached page, but with NEW_ADDR */ | |
6bacf52f | 416 | if (unlikely(dn.data_blkaddr == NEW_ADDR)) |
eb47b800 JK |
417 | return ERR_PTR(-EINVAL); |
418 | ||
9ac1349a | 419 | page = grab_cache_page(mapping, index); |
eb47b800 JK |
420 | if (!page) |
421 | return ERR_PTR(-ENOMEM); | |
422 | ||
393ff91f JK |
423 | if (PageUptodate(page)) { |
424 | unlock_page(page); | |
425 | return page; | |
426 | } | |
427 | ||
4081363f | 428 | err = f2fs_submit_page_bio(F2FS_I_SB(inode), page, dn.data_blkaddr, |
c718379b | 429 | sync ? READ_SYNC : READA); |
1069bbf7 CY |
430 | if (err) |
431 | return ERR_PTR(err); | |
432 | ||
c718379b JK |
433 | if (sync) { |
434 | wait_on_page_locked(page); | |
6bacf52f | 435 | if (unlikely(!PageUptodate(page))) { |
c718379b JK |
436 | f2fs_put_page(page, 0); |
437 | return ERR_PTR(-EIO); | |
438 | } | |
eb47b800 | 439 | } |
eb47b800 JK |
440 | return page; |
441 | } | |
442 | ||
0a8165d7 | 443 | /* |
eb47b800 JK |
444 | * If it tries to access a hole, return an error. |
445 | * Because, the callers, functions in dir.c and GC, should be able to know | |
446 | * whether this page exists or not. | |
447 | */ | |
448 | struct page *get_lock_data_page(struct inode *inode, pgoff_t index) | |
449 | { | |
eb47b800 JK |
450 | struct address_space *mapping = inode->i_mapping; |
451 | struct dnode_of_data dn; | |
452 | struct page *page; | |
453 | int err; | |
454 | ||
650495de | 455 | repeat: |
9ac1349a | 456 | page = grab_cache_page(mapping, index); |
650495de JK |
457 | if (!page) |
458 | return ERR_PTR(-ENOMEM); | |
459 | ||
eb47b800 | 460 | set_new_dnode(&dn, inode, NULL, NULL, 0); |
266e97a8 | 461 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); |
650495de JK |
462 | if (err) { |
463 | f2fs_put_page(page, 1); | |
eb47b800 | 464 | return ERR_PTR(err); |
650495de | 465 | } |
eb47b800 JK |
466 | f2fs_put_dnode(&dn); |
467 | ||
6bacf52f | 468 | if (unlikely(dn.data_blkaddr == NULL_ADDR)) { |
650495de | 469 | f2fs_put_page(page, 1); |
eb47b800 | 470 | return ERR_PTR(-ENOENT); |
650495de | 471 | } |
eb47b800 JK |
472 | |
473 | if (PageUptodate(page)) | |
474 | return page; | |
475 | ||
d59ff4df JK |
476 | /* |
477 | * A new dentry page is allocated but not able to be written, since its | |
478 | * new inode page couldn't be allocated due to -ENOSPC. | |
479 | * In such the case, its blkaddr can be remained as NEW_ADDR. | |
480 | * see, f2fs_add_link -> get_new_data_page -> init_inode_metadata. | |
481 | */ | |
482 | if (dn.data_blkaddr == NEW_ADDR) { | |
483 | zero_user_segment(page, 0, PAGE_CACHE_SIZE); | |
484 | SetPageUptodate(page); | |
485 | return page; | |
486 | } | |
eb47b800 | 487 | |
4081363f JK |
488 | err = f2fs_submit_page_bio(F2FS_I_SB(inode), page, |
489 | dn.data_blkaddr, READ_SYNC); | |
393ff91f | 490 | if (err) |
eb47b800 | 491 | return ERR_PTR(err); |
393ff91f JK |
492 | |
493 | lock_page(page); | |
6bacf52f | 494 | if (unlikely(!PageUptodate(page))) { |
393ff91f JK |
495 | f2fs_put_page(page, 1); |
496 | return ERR_PTR(-EIO); | |
eb47b800 | 497 | } |
6bacf52f | 498 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
499 | f2fs_put_page(page, 1); |
500 | goto repeat; | |
eb47b800 JK |
501 | } |
502 | return page; | |
503 | } | |
504 | ||
0a8165d7 | 505 | /* |
eb47b800 JK |
506 | * Caller ensures that this data page is never allocated. |
507 | * A new zero-filled data page is allocated in the page cache. | |
39936837 | 508 | * |
4f4124d0 CY |
509 | * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and |
510 | * f2fs_unlock_op(). | |
a8865372 | 511 | * Note that, ipage is set only by make_empty_dir. |
eb47b800 | 512 | */ |
64aa7ed9 | 513 | struct page *get_new_data_page(struct inode *inode, |
a8865372 | 514 | struct page *ipage, pgoff_t index, bool new_i_size) |
eb47b800 | 515 | { |
eb47b800 JK |
516 | struct address_space *mapping = inode->i_mapping; |
517 | struct page *page; | |
518 | struct dnode_of_data dn; | |
519 | int err; | |
520 | ||
a8865372 | 521 | set_new_dnode(&dn, inode, ipage, NULL, 0); |
b600965c | 522 | err = f2fs_reserve_block(&dn, index); |
eb47b800 JK |
523 | if (err) |
524 | return ERR_PTR(err); | |
afcb7ca0 | 525 | repeat: |
eb47b800 | 526 | page = grab_cache_page(mapping, index); |
a8865372 JK |
527 | if (!page) { |
528 | err = -ENOMEM; | |
529 | goto put_err; | |
530 | } | |
eb47b800 JK |
531 | |
532 | if (PageUptodate(page)) | |
533 | return page; | |
534 | ||
535 | if (dn.data_blkaddr == NEW_ADDR) { | |
536 | zero_user_segment(page, 0, PAGE_CACHE_SIZE); | |
393ff91f | 537 | SetPageUptodate(page); |
eb47b800 | 538 | } else { |
4081363f JK |
539 | err = f2fs_submit_page_bio(F2FS_I_SB(inode), page, |
540 | dn.data_blkaddr, READ_SYNC); | |
393ff91f | 541 | if (err) |
a8865372 JK |
542 | goto put_err; |
543 | ||
393ff91f | 544 | lock_page(page); |
6bacf52f | 545 | if (unlikely(!PageUptodate(page))) { |
393ff91f | 546 | f2fs_put_page(page, 1); |
a8865372 JK |
547 | err = -EIO; |
548 | goto put_err; | |
eb47b800 | 549 | } |
6bacf52f | 550 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
551 | f2fs_put_page(page, 1); |
552 | goto repeat; | |
eb47b800 JK |
553 | } |
554 | } | |
eb47b800 JK |
555 | |
556 | if (new_i_size && | |
557 | i_size_read(inode) < ((index + 1) << PAGE_CACHE_SHIFT)) { | |
558 | i_size_write(inode, ((index + 1) << PAGE_CACHE_SHIFT)); | |
699489bb JK |
559 | /* Only the directory inode sets new_i_size */ |
560 | set_inode_flag(F2FS_I(inode), FI_UPDATE_DIR); | |
eb47b800 JK |
561 | } |
562 | return page; | |
a8865372 JK |
563 | |
564 | put_err: | |
565 | f2fs_put_dnode(&dn); | |
566 | return ERR_PTR(err); | |
eb47b800 JK |
567 | } |
568 | ||
bfad7c2d JK |
569 | static int __allocate_data_block(struct dnode_of_data *dn) |
570 | { | |
4081363f | 571 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
976e4c50 | 572 | struct f2fs_inode_info *fi = F2FS_I(dn->inode); |
bfad7c2d JK |
573 | struct f2fs_summary sum; |
574 | block_t new_blkaddr; | |
575 | struct node_info ni; | |
976e4c50 | 576 | pgoff_t fofs; |
bfad7c2d JK |
577 | int type; |
578 | ||
579 | if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))) | |
580 | return -EPERM; | |
581 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, 1))) | |
582 | return -ENOSPC; | |
583 | ||
584 | __set_data_blkaddr(dn, NEW_ADDR); | |
585 | dn->data_blkaddr = NEW_ADDR; | |
586 | ||
587 | get_node_info(sbi, dn->nid, &ni); | |
588 | set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version); | |
589 | ||
590 | type = CURSEG_WARM_DATA; | |
591 | ||
592 | allocate_data_block(sbi, NULL, NULL_ADDR, &new_blkaddr, &sum, type); | |
593 | ||
594 | /* direct IO doesn't use extent cache to maximize the performance */ | |
595 | set_inode_flag(F2FS_I(dn->inode), FI_NO_EXTENT); | |
596 | update_extent_cache(new_blkaddr, dn); | |
597 | clear_inode_flag(F2FS_I(dn->inode), FI_NO_EXTENT); | |
598 | ||
976e4c50 JK |
599 | /* update i_size */ |
600 | fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) + | |
601 | dn->ofs_in_node; | |
602 | if (i_size_read(dn->inode) < ((fofs + 1) << PAGE_CACHE_SHIFT)) | |
603 | i_size_write(dn->inode, ((fofs + 1) << PAGE_CACHE_SHIFT)); | |
604 | ||
bfad7c2d JK |
605 | dn->data_blkaddr = new_blkaddr; |
606 | return 0; | |
607 | } | |
608 | ||
0a8165d7 | 609 | /* |
4f4124d0 CY |
610 | * get_data_block() now supported readahead/bmap/rw direct_IO with mapped bh. |
611 | * If original data blocks are allocated, then give them to blockdev. | |
612 | * Otherwise, | |
613 | * a. preallocate requested block addresses | |
614 | * b. do not use extent cache for better performance | |
615 | * c. give the block addresses to blockdev | |
eb47b800 | 616 | */ |
ccfb3000 JK |
617 | static int __get_data_block(struct inode *inode, sector_t iblock, |
618 | struct buffer_head *bh_result, int create, bool fiemap) | |
eb47b800 JK |
619 | { |
620 | unsigned int blkbits = inode->i_sb->s_blocksize_bits; | |
621 | unsigned maxblocks = bh_result->b_size >> blkbits; | |
622 | struct dnode_of_data dn; | |
bfad7c2d JK |
623 | int mode = create ? ALLOC_NODE : LOOKUP_NODE_RA; |
624 | pgoff_t pgofs, end_offset; | |
625 | int err = 0, ofs = 1; | |
626 | bool allocated = false; | |
eb47b800 JK |
627 | |
628 | /* Get the page offset from the block offset(iblock) */ | |
629 | pgofs = (pgoff_t)(iblock >> (PAGE_CACHE_SHIFT - blkbits)); | |
630 | ||
bfad7c2d JK |
631 | if (check_extent_cache(inode, pgofs, bh_result)) |
632 | goto out; | |
633 | ||
79e35dc3 | 634 | if (create) { |
4081363f JK |
635 | f2fs_balance_fs(F2FS_I_SB(inode)); |
636 | f2fs_lock_op(F2FS_I_SB(inode)); | |
79e35dc3 | 637 | } |
eb47b800 JK |
638 | |
639 | /* When reading holes, we need its node page */ | |
640 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
bfad7c2d | 641 | err = get_dnode_of_data(&dn, pgofs, mode); |
1ec79083 | 642 | if (err) { |
bfad7c2d JK |
643 | if (err == -ENOENT) |
644 | err = 0; | |
645 | goto unlock_out; | |
848753aa | 646 | } |
ccfb3000 | 647 | if (dn.data_blkaddr == NEW_ADDR && !fiemap) |
1ec79083 | 648 | goto put_out; |
eb47b800 | 649 | |
bfad7c2d JK |
650 | if (dn.data_blkaddr != NULL_ADDR) { |
651 | map_bh(bh_result, inode->i_sb, dn.data_blkaddr); | |
652 | } else if (create) { | |
653 | err = __allocate_data_block(&dn); | |
654 | if (err) | |
655 | goto put_out; | |
656 | allocated = true; | |
657 | map_bh(bh_result, inode->i_sb, dn.data_blkaddr); | |
658 | } else { | |
659 | goto put_out; | |
660 | } | |
661 | ||
6403eb1f | 662 | end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode)); |
bfad7c2d JK |
663 | bh_result->b_size = (((size_t)1) << blkbits); |
664 | dn.ofs_in_node++; | |
665 | pgofs++; | |
666 | ||
667 | get_next: | |
668 | if (dn.ofs_in_node >= end_offset) { | |
669 | if (allocated) | |
670 | sync_inode_page(&dn); | |
671 | allocated = false; | |
672 | f2fs_put_dnode(&dn); | |
673 | ||
674 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
675 | err = get_dnode_of_data(&dn, pgofs, mode); | |
1ec79083 | 676 | if (err) { |
bfad7c2d JK |
677 | if (err == -ENOENT) |
678 | err = 0; | |
679 | goto unlock_out; | |
680 | } | |
ccfb3000 | 681 | if (dn.data_blkaddr == NEW_ADDR && !fiemap) |
1ec79083 JK |
682 | goto put_out; |
683 | ||
6403eb1f | 684 | end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode)); |
bfad7c2d | 685 | } |
eb47b800 | 686 | |
bfad7c2d JK |
687 | if (maxblocks > (bh_result->b_size >> blkbits)) { |
688 | block_t blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node); | |
689 | if (blkaddr == NULL_ADDR && create) { | |
690 | err = __allocate_data_block(&dn); | |
691 | if (err) | |
692 | goto sync_out; | |
693 | allocated = true; | |
694 | blkaddr = dn.data_blkaddr; | |
695 | } | |
e1c42045 | 696 | /* Give more consecutive addresses for the readahead */ |
bfad7c2d JK |
697 | if (blkaddr == (bh_result->b_blocknr + ofs)) { |
698 | ofs++; | |
699 | dn.ofs_in_node++; | |
700 | pgofs++; | |
701 | bh_result->b_size += (((size_t)1) << blkbits); | |
702 | goto get_next; | |
703 | } | |
eb47b800 | 704 | } |
bfad7c2d JK |
705 | sync_out: |
706 | if (allocated) | |
707 | sync_inode_page(&dn); | |
708 | put_out: | |
eb47b800 | 709 | f2fs_put_dnode(&dn); |
bfad7c2d JK |
710 | unlock_out: |
711 | if (create) | |
4081363f | 712 | f2fs_unlock_op(F2FS_I_SB(inode)); |
bfad7c2d JK |
713 | out: |
714 | trace_f2fs_get_data_block(inode, iblock, bh_result, err); | |
715 | return err; | |
eb47b800 JK |
716 | } |
717 | ||
ccfb3000 JK |
718 | static int get_data_block(struct inode *inode, sector_t iblock, |
719 | struct buffer_head *bh_result, int create) | |
720 | { | |
721 | return __get_data_block(inode, iblock, bh_result, create, false); | |
722 | } | |
723 | ||
724 | static int get_data_block_fiemap(struct inode *inode, sector_t iblock, | |
725 | struct buffer_head *bh_result, int create) | |
726 | { | |
727 | return __get_data_block(inode, iblock, bh_result, create, true); | |
728 | } | |
729 | ||
9ab70134 JK |
730 | int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, |
731 | u64 start, u64 len) | |
732 | { | |
ccfb3000 JK |
733 | return generic_block_fiemap(inode, fieinfo, |
734 | start, len, get_data_block_fiemap); | |
9ab70134 JK |
735 | } |
736 | ||
eb47b800 JK |
737 | static int f2fs_read_data_page(struct file *file, struct page *page) |
738 | { | |
9ffe0fb5 HL |
739 | struct inode *inode = page->mapping->host; |
740 | int ret; | |
741 | ||
c20e89cd CY |
742 | trace_f2fs_readpage(page, DATA); |
743 | ||
e1c42045 | 744 | /* If the file has inline data, try to read it directly */ |
9ffe0fb5 HL |
745 | if (f2fs_has_inline_data(inode)) |
746 | ret = f2fs_read_inline_data(inode, page); | |
747 | else | |
748 | ret = mpage_readpage(page, get_data_block); | |
749 | ||
750 | return ret; | |
eb47b800 JK |
751 | } |
752 | ||
753 | static int f2fs_read_data_pages(struct file *file, | |
754 | struct address_space *mapping, | |
755 | struct list_head *pages, unsigned nr_pages) | |
756 | { | |
9ffe0fb5 HL |
757 | struct inode *inode = file->f_mapping->host; |
758 | ||
759 | /* If the file has inline data, skip readpages */ | |
760 | if (f2fs_has_inline_data(inode)) | |
761 | return 0; | |
762 | ||
bfad7c2d | 763 | return mpage_readpages(mapping, pages, nr_pages, get_data_block); |
eb47b800 JK |
764 | } |
765 | ||
458e6197 | 766 | int do_write_data_page(struct page *page, struct f2fs_io_info *fio) |
eb47b800 JK |
767 | { |
768 | struct inode *inode = page->mapping->host; | |
458e6197 | 769 | block_t old_blkaddr, new_blkaddr; |
eb47b800 JK |
770 | struct dnode_of_data dn; |
771 | int err = 0; | |
772 | ||
773 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
266e97a8 | 774 | err = get_dnode_of_data(&dn, page->index, LOOKUP_NODE); |
eb47b800 JK |
775 | if (err) |
776 | return err; | |
777 | ||
458e6197 | 778 | old_blkaddr = dn.data_blkaddr; |
eb47b800 JK |
779 | |
780 | /* This page is already truncated */ | |
458e6197 | 781 | if (old_blkaddr == NULL_ADDR) |
eb47b800 JK |
782 | goto out_writepage; |
783 | ||
784 | set_page_writeback(page); | |
785 | ||
786 | /* | |
787 | * If current allocation needs SSR, | |
788 | * it had better in-place writes for updated data. | |
789 | */ | |
458e6197 | 790 | if (unlikely(old_blkaddr != NEW_ADDR && |
b25958b6 HL |
791 | !is_cold_data(page) && |
792 | need_inplace_update(inode))) { | |
458e6197 | 793 | rewrite_data_page(page, old_blkaddr, fio); |
fff04f90 | 794 | set_inode_flag(F2FS_I(inode), FI_UPDATE_WRITE); |
eb47b800 | 795 | } else { |
458e6197 JK |
796 | write_data_page(page, &dn, &new_blkaddr, fio); |
797 | update_extent_cache(new_blkaddr, &dn); | |
fff04f90 | 798 | set_inode_flag(F2FS_I(inode), FI_APPEND_WRITE); |
eb47b800 JK |
799 | } |
800 | out_writepage: | |
801 | f2fs_put_dnode(&dn); | |
802 | return err; | |
803 | } | |
804 | ||
805 | static int f2fs_write_data_page(struct page *page, | |
806 | struct writeback_control *wbc) | |
807 | { | |
808 | struct inode *inode = page->mapping->host; | |
4081363f | 809 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
eb47b800 JK |
810 | loff_t i_size = i_size_read(inode); |
811 | const pgoff_t end_index = ((unsigned long long) i_size) | |
812 | >> PAGE_CACHE_SHIFT; | |
9ffe0fb5 | 813 | unsigned offset = 0; |
39936837 | 814 | bool need_balance_fs = false; |
eb47b800 | 815 | int err = 0; |
458e6197 JK |
816 | struct f2fs_io_info fio = { |
817 | .type = DATA, | |
6c311ec6 | 818 | .rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE, |
458e6197 | 819 | }; |
eb47b800 | 820 | |
ecda0de3 CY |
821 | trace_f2fs_writepage(page, DATA); |
822 | ||
eb47b800 | 823 | if (page->index < end_index) |
39936837 | 824 | goto write; |
eb47b800 JK |
825 | |
826 | /* | |
827 | * If the offset is out-of-range of file size, | |
828 | * this page does not have to be written to disk. | |
829 | */ | |
830 | offset = i_size & (PAGE_CACHE_SIZE - 1); | |
76f60268 | 831 | if ((page->index >= end_index + 1) || !offset) |
39936837 | 832 | goto out; |
eb47b800 JK |
833 | |
834 | zero_user_segment(page, offset, PAGE_CACHE_SIZE); | |
39936837 | 835 | write: |
8618b881 | 836 | if (unlikely(sbi->por_doing)) |
eb47b800 JK |
837 | goto redirty_out; |
838 | ||
39936837 | 839 | /* Dentry blocks are controlled by checkpoint */ |
eb47b800 | 840 | if (S_ISDIR(inode->i_mode)) { |
cf779cab JK |
841 | if (unlikely(f2fs_cp_error(sbi))) |
842 | goto redirty_out; | |
458e6197 | 843 | err = do_write_data_page(page, &fio); |
8618b881 JK |
844 | goto done; |
845 | } | |
9ffe0fb5 | 846 | |
cf779cab JK |
847 | /* we should bypass data pages to proceed the kworkder jobs */ |
848 | if (unlikely(f2fs_cp_error(sbi))) { | |
849 | SetPageError(page); | |
850 | unlock_page(page); | |
a7ffdbe2 | 851 | goto out; |
cf779cab JK |
852 | } |
853 | ||
8618b881 | 854 | if (!wbc->for_reclaim) |
39936837 | 855 | need_balance_fs = true; |
8618b881 | 856 | else if (has_not_enough_free_secs(sbi, 0)) |
39936837 | 857 | goto redirty_out; |
eb47b800 | 858 | |
8618b881 JK |
859 | f2fs_lock_op(sbi); |
860 | if (f2fs_has_inline_data(inode) || f2fs_may_inline(inode)) | |
861 | err = f2fs_write_inline_data(inode, page, offset); | |
862 | else | |
863 | err = do_write_data_page(page, &fio); | |
864 | f2fs_unlock_op(sbi); | |
865 | done: | |
866 | if (err && err != -ENOENT) | |
867 | goto redirty_out; | |
eb47b800 | 868 | |
eb47b800 | 869 | clear_cold_data(page); |
39936837 | 870 | out: |
a7ffdbe2 | 871 | inode_dec_dirty_pages(inode); |
eb47b800 | 872 | unlock_page(page); |
39936837 | 873 | if (need_balance_fs) |
eb47b800 | 874 | f2fs_balance_fs(sbi); |
2aea39ec JK |
875 | if (wbc->for_reclaim) |
876 | f2fs_submit_merged_bio(sbi, DATA, WRITE); | |
eb47b800 JK |
877 | return 0; |
878 | ||
eb47b800 | 879 | redirty_out: |
76f60268 | 880 | redirty_page_for_writepage(wbc, page); |
8618b881 | 881 | return AOP_WRITEPAGE_ACTIVATE; |
eb47b800 JK |
882 | } |
883 | ||
fa9150a8 NJ |
884 | static int __f2fs_writepage(struct page *page, struct writeback_control *wbc, |
885 | void *data) | |
886 | { | |
887 | struct address_space *mapping = data; | |
888 | int ret = mapping->a_ops->writepage(page, wbc); | |
889 | mapping_set_error(mapping, ret); | |
890 | return ret; | |
891 | } | |
892 | ||
25ca923b | 893 | static int f2fs_write_data_pages(struct address_space *mapping, |
eb47b800 JK |
894 | struct writeback_control *wbc) |
895 | { | |
896 | struct inode *inode = mapping->host; | |
4081363f | 897 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
531ad7d5 | 898 | bool locked = false; |
eb47b800 | 899 | int ret; |
50c8cdb3 | 900 | long diff; |
eb47b800 | 901 | |
e5748434 CY |
902 | trace_f2fs_writepages(mapping->host, wbc, DATA); |
903 | ||
cfb185a1 | 904 | /* deal with chardevs and other special file */ |
905 | if (!mapping->a_ops->writepage) | |
906 | return 0; | |
907 | ||
87d6f890 | 908 | if (S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_NONE && |
a7ffdbe2 | 909 | get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) && |
6fb03f3a | 910 | available_free_memory(sbi, DIRTY_DENTS)) |
d3baf95d | 911 | goto skip_write; |
87d6f890 | 912 | |
50c8cdb3 | 913 | diff = nr_pages_to_write(sbi, DATA, wbc); |
eb47b800 | 914 | |
531ad7d5 | 915 | if (!S_ISDIR(inode->i_mode)) { |
eb47b800 | 916 | mutex_lock(&sbi->writepages); |
531ad7d5 JK |
917 | locked = true; |
918 | } | |
fa9150a8 | 919 | ret = write_cache_pages(mapping, wbc, __f2fs_writepage, mapping); |
531ad7d5 | 920 | if (locked) |
eb47b800 | 921 | mutex_unlock(&sbi->writepages); |
458e6197 JK |
922 | |
923 | f2fs_submit_merged_bio(sbi, DATA, WRITE); | |
eb47b800 JK |
924 | |
925 | remove_dirty_dir_inode(inode); | |
926 | ||
50c8cdb3 | 927 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff); |
eb47b800 | 928 | return ret; |
d3baf95d JK |
929 | |
930 | skip_write: | |
a7ffdbe2 | 931 | wbc->pages_skipped += get_dirty_pages(inode); |
d3baf95d | 932 | return 0; |
eb47b800 JK |
933 | } |
934 | ||
3aab8f82 CY |
935 | static void f2fs_write_failed(struct address_space *mapping, loff_t to) |
936 | { | |
937 | struct inode *inode = mapping->host; | |
938 | ||
939 | if (to > inode->i_size) { | |
940 | truncate_pagecache(inode, inode->i_size); | |
764aa3e9 | 941 | truncate_blocks(inode, inode->i_size, true); |
3aab8f82 CY |
942 | } |
943 | } | |
944 | ||
eb47b800 JK |
945 | static int f2fs_write_begin(struct file *file, struct address_space *mapping, |
946 | loff_t pos, unsigned len, unsigned flags, | |
947 | struct page **pagep, void **fsdata) | |
948 | { | |
949 | struct inode *inode = mapping->host; | |
4081363f | 950 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
9ba69cf9 | 951 | struct page *page, *ipage; |
eb47b800 JK |
952 | pgoff_t index = ((unsigned long long) pos) >> PAGE_CACHE_SHIFT; |
953 | struct dnode_of_data dn; | |
954 | int err = 0; | |
955 | ||
62aed044 CY |
956 | trace_f2fs_write_begin(inode, pos, len, flags); |
957 | ||
eb47b800 | 958 | f2fs_balance_fs(sbi); |
afcb7ca0 | 959 | repeat: |
b067ba1f | 960 | err = f2fs_convert_inline_data(inode, pos + len, NULL); |
9e09fc85 | 961 | if (err) |
3aab8f82 | 962 | goto fail; |
9e09fc85 | 963 | |
eb47b800 | 964 | page = grab_cache_page_write_begin(mapping, index, flags); |
3aab8f82 CY |
965 | if (!page) { |
966 | err = -ENOMEM; | |
967 | goto fail; | |
968 | } | |
d5f66990 JK |
969 | |
970 | /* to avoid latency during memory pressure */ | |
971 | unlock_page(page); | |
972 | ||
eb47b800 JK |
973 | *pagep = page; |
974 | ||
9e09fc85 JK |
975 | if (f2fs_has_inline_data(inode) && (pos + len) <= MAX_INLINE_DATA) |
976 | goto inline_data; | |
9ffe0fb5 | 977 | |
e479556b | 978 | f2fs_lock_op(sbi); |
9ba69cf9 JK |
979 | |
980 | /* check inline_data */ | |
981 | ipage = get_node_page(sbi, inode->i_ino); | |
982 | if (IS_ERR(ipage)) | |
983 | goto unlock_fail; | |
984 | ||
985 | if (f2fs_has_inline_data(inode)) { | |
986 | f2fs_put_page(ipage, 1); | |
987 | f2fs_unlock_op(sbi); | |
d5f66990 | 988 | f2fs_put_page(page, 0); |
9ba69cf9 | 989 | goto repeat; |
b600965c | 990 | } |
9ba69cf9 JK |
991 | |
992 | set_new_dnode(&dn, inode, ipage, NULL, 0); | |
993 | err = f2fs_reserve_block(&dn, index); | |
994 | if (err) | |
995 | goto unlock_fail; | |
996 | f2fs_put_dnode(&dn); | |
997 | f2fs_unlock_op(sbi); | |
998 | ||
9ffe0fb5 | 999 | inline_data: |
d5f66990 JK |
1000 | lock_page(page); |
1001 | if (unlikely(page->mapping != mapping)) { | |
1002 | f2fs_put_page(page, 1); | |
1003 | goto repeat; | |
1004 | } | |
1005 | ||
1006 | f2fs_wait_on_page_writeback(page, DATA); | |
1007 | ||
eb47b800 JK |
1008 | if ((len == PAGE_CACHE_SIZE) || PageUptodate(page)) |
1009 | return 0; | |
1010 | ||
1011 | if ((pos & PAGE_CACHE_MASK) >= i_size_read(inode)) { | |
1012 | unsigned start = pos & (PAGE_CACHE_SIZE - 1); | |
1013 | unsigned end = start + len; | |
1014 | ||
1015 | /* Reading beyond i_size is simple: memset to zero */ | |
1016 | zero_user_segments(page, 0, start, end, PAGE_CACHE_SIZE); | |
393ff91f | 1017 | goto out; |
eb47b800 JK |
1018 | } |
1019 | ||
1020 | if (dn.data_blkaddr == NEW_ADDR) { | |
1021 | zero_user_segment(page, 0, PAGE_CACHE_SIZE); | |
1022 | } else { | |
d54c795b | 1023 | if (f2fs_has_inline_data(inode)) { |
9ffe0fb5 | 1024 | err = f2fs_read_inline_data(inode, page); |
d54c795b CY |
1025 | if (err) { |
1026 | page_cache_release(page); | |
3aab8f82 | 1027 | goto fail; |
d54c795b CY |
1028 | } |
1029 | } else { | |
9ffe0fb5 | 1030 | err = f2fs_submit_page_bio(sbi, page, dn.data_blkaddr, |
93dfe2ac | 1031 | READ_SYNC); |
d54c795b | 1032 | if (err) |
3aab8f82 | 1033 | goto fail; |
d54c795b CY |
1034 | } |
1035 | ||
393ff91f | 1036 | lock_page(page); |
6bacf52f | 1037 | if (unlikely(!PageUptodate(page))) { |
393ff91f | 1038 | f2fs_put_page(page, 1); |
3aab8f82 CY |
1039 | err = -EIO; |
1040 | goto fail; | |
eb47b800 | 1041 | } |
6bacf52f | 1042 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
1043 | f2fs_put_page(page, 1); |
1044 | goto repeat; | |
eb47b800 JK |
1045 | } |
1046 | } | |
393ff91f | 1047 | out: |
eb47b800 JK |
1048 | SetPageUptodate(page); |
1049 | clear_cold_data(page); | |
1050 | return 0; | |
9ba69cf9 JK |
1051 | |
1052 | unlock_fail: | |
1053 | f2fs_unlock_op(sbi); | |
1054 | f2fs_put_page(page, 0); | |
3aab8f82 CY |
1055 | fail: |
1056 | f2fs_write_failed(mapping, pos + len); | |
1057 | return err; | |
eb47b800 JK |
1058 | } |
1059 | ||
a1dd3c13 JK |
1060 | static int f2fs_write_end(struct file *file, |
1061 | struct address_space *mapping, | |
1062 | loff_t pos, unsigned len, unsigned copied, | |
1063 | struct page *page, void *fsdata) | |
1064 | { | |
1065 | struct inode *inode = page->mapping->host; | |
1066 | ||
dfb2bf38 CY |
1067 | trace_f2fs_write_end(inode, pos, len, copied); |
1068 | ||
34ba94ba | 1069 | set_page_dirty(page); |
a1dd3c13 JK |
1070 | |
1071 | if (pos + copied > i_size_read(inode)) { | |
1072 | i_size_write(inode, pos + copied); | |
1073 | mark_inode_dirty(inode); | |
1074 | update_inode_page(inode); | |
1075 | } | |
1076 | ||
75c3c8bc | 1077 | f2fs_put_page(page, 1); |
a1dd3c13 JK |
1078 | return copied; |
1079 | } | |
1080 | ||
944fcfc1 | 1081 | static int check_direct_IO(struct inode *inode, int rw, |
5b46f25d | 1082 | struct iov_iter *iter, loff_t offset) |
944fcfc1 JK |
1083 | { |
1084 | unsigned blocksize_mask = inode->i_sb->s_blocksize - 1; | |
944fcfc1 JK |
1085 | |
1086 | if (rw == READ) | |
1087 | return 0; | |
1088 | ||
1089 | if (offset & blocksize_mask) | |
1090 | return -EINVAL; | |
1091 | ||
5b46f25d AV |
1092 | if (iov_iter_alignment(iter) & blocksize_mask) |
1093 | return -EINVAL; | |
1094 | ||
944fcfc1 JK |
1095 | return 0; |
1096 | } | |
1097 | ||
eb47b800 | 1098 | static ssize_t f2fs_direct_IO(int rw, struct kiocb *iocb, |
d8d3d94b | 1099 | struct iov_iter *iter, loff_t offset) |
eb47b800 JK |
1100 | { |
1101 | struct file *file = iocb->ki_filp; | |
3aab8f82 CY |
1102 | struct address_space *mapping = file->f_mapping; |
1103 | struct inode *inode = mapping->host; | |
1104 | size_t count = iov_iter_count(iter); | |
1105 | int err; | |
944fcfc1 | 1106 | |
9ffe0fb5 HL |
1107 | /* Let buffer I/O handle the inline data case. */ |
1108 | if (f2fs_has_inline_data(inode)) | |
1109 | return 0; | |
1110 | ||
5b46f25d | 1111 | if (check_direct_IO(inode, rw, iter, offset)) |
944fcfc1 JK |
1112 | return 0; |
1113 | ||
70407fad CY |
1114 | trace_f2fs_direct_IO_enter(inode, offset, count, rw); |
1115 | ||
3aab8f82 CY |
1116 | err = blockdev_direct_IO(rw, iocb, inode, iter, offset, get_data_block); |
1117 | if (err < 0 && (rw & WRITE)) | |
1118 | f2fs_write_failed(mapping, offset + count); | |
70407fad CY |
1119 | |
1120 | trace_f2fs_direct_IO_exit(inode, offset, count, rw, err); | |
1121 | ||
3aab8f82 | 1122 | return err; |
eb47b800 JK |
1123 | } |
1124 | ||
d47992f8 LC |
1125 | static void f2fs_invalidate_data_page(struct page *page, unsigned int offset, |
1126 | unsigned int length) | |
eb47b800 JK |
1127 | { |
1128 | struct inode *inode = page->mapping->host; | |
a7ffdbe2 JK |
1129 | |
1130 | if (offset % PAGE_CACHE_SIZE || length != PAGE_CACHE_SIZE) | |
1131 | return; | |
1132 | ||
cbcb2872 JK |
1133 | if (f2fs_is_atomic_file(inode) || f2fs_is_volatile_file(inode)) |
1134 | invalidate_inmem_page(inode, page); | |
1135 | ||
1fe54f9d | 1136 | if (PageDirty(page)) |
a7ffdbe2 | 1137 | inode_dec_dirty_pages(inode); |
eb47b800 JK |
1138 | ClearPagePrivate(page); |
1139 | } | |
1140 | ||
1141 | static int f2fs_release_data_page(struct page *page, gfp_t wait) | |
1142 | { | |
1143 | ClearPagePrivate(page); | |
c3850aa1 | 1144 | return 1; |
eb47b800 JK |
1145 | } |
1146 | ||
1147 | static int f2fs_set_data_page_dirty(struct page *page) | |
1148 | { | |
1149 | struct address_space *mapping = page->mapping; | |
1150 | struct inode *inode = mapping->host; | |
1151 | ||
26c6b887 JK |
1152 | trace_f2fs_set_page_dirty(page, DATA); |
1153 | ||
eb47b800 | 1154 | SetPageUptodate(page); |
34ba94ba JK |
1155 | |
1156 | if (f2fs_is_atomic_file(inode) || f2fs_is_volatile_file(inode)) { | |
1157 | register_inmem_page(inode, page); | |
1158 | return 1; | |
1159 | } | |
1160 | ||
a18ff063 JK |
1161 | mark_inode_dirty(inode); |
1162 | ||
eb47b800 JK |
1163 | if (!PageDirty(page)) { |
1164 | __set_page_dirty_nobuffers(page); | |
a7ffdbe2 | 1165 | update_dirty_page(inode, page); |
eb47b800 JK |
1166 | return 1; |
1167 | } | |
1168 | return 0; | |
1169 | } | |
1170 | ||
c01e54b7 JK |
1171 | static sector_t f2fs_bmap(struct address_space *mapping, sector_t block) |
1172 | { | |
454ae7e5 CY |
1173 | struct inode *inode = mapping->host; |
1174 | ||
1175 | if (f2fs_has_inline_data(inode)) | |
1176 | return 0; | |
1177 | ||
bfad7c2d | 1178 | return generic_block_bmap(mapping, block, get_data_block); |
c01e54b7 JK |
1179 | } |
1180 | ||
eb47b800 JK |
1181 | const struct address_space_operations f2fs_dblock_aops = { |
1182 | .readpage = f2fs_read_data_page, | |
1183 | .readpages = f2fs_read_data_pages, | |
1184 | .writepage = f2fs_write_data_page, | |
1185 | .writepages = f2fs_write_data_pages, | |
1186 | .write_begin = f2fs_write_begin, | |
a1dd3c13 | 1187 | .write_end = f2fs_write_end, |
eb47b800 JK |
1188 | .set_page_dirty = f2fs_set_data_page_dirty, |
1189 | .invalidatepage = f2fs_invalidate_data_page, | |
1190 | .releasepage = f2fs_release_data_page, | |
1191 | .direct_IO = f2fs_direct_IO, | |
c01e54b7 | 1192 | .bmap = f2fs_bmap, |
eb47b800 | 1193 | }; |