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 | |
93dfe2ac JK |
64 | if (!get_pages(sbi, F2FS_WRITEBACK) && |
65 | !list_empty(&sbi->cp_wait.task_list)) | |
66 | wake_up(&sbi->cp_wait); | |
67 | ||
68 | bio_put(bio); | |
69 | } | |
70 | ||
940a6d34 GZ |
71 | /* |
72 | * Low-level block read/write IO operations. | |
73 | */ | |
74 | static struct bio *__bio_alloc(struct f2fs_sb_info *sbi, block_t blk_addr, | |
75 | int npages, bool is_read) | |
76 | { | |
77 | struct bio *bio; | |
78 | ||
79 | /* No failure on bio allocation */ | |
80 | bio = bio_alloc(GFP_NOIO, npages); | |
81 | ||
82 | bio->bi_bdev = sbi->sb->s_bdev; | |
55cf9cb6 | 83 | bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr); |
940a6d34 | 84 | bio->bi_end_io = is_read ? f2fs_read_end_io : f2fs_write_end_io; |
1b1f559f | 85 | bio->bi_private = sbi; |
940a6d34 GZ |
86 | |
87 | return bio; | |
88 | } | |
89 | ||
458e6197 | 90 | static void __submit_merged_bio(struct f2fs_bio_info *io) |
93dfe2ac | 91 | { |
458e6197 | 92 | struct f2fs_io_info *fio = &io->fio; |
93dfe2ac JK |
93 | |
94 | if (!io->bio) | |
95 | return; | |
96 | ||
6a8f8ca5 JK |
97 | if (is_read_io(fio->rw)) |
98 | trace_f2fs_submit_read_bio(io->sbi->sb, fio->rw, | |
99 | fio->type, io->bio); | |
100 | else | |
101 | trace_f2fs_submit_write_bio(io->sbi->sb, fio->rw, | |
102 | fio->type, io->bio); | |
940a6d34 | 103 | |
6a8f8ca5 | 104 | submit_bio(fio->rw, io->bio); |
93dfe2ac JK |
105 | io->bio = NULL; |
106 | } | |
107 | ||
108 | void f2fs_submit_merged_bio(struct f2fs_sb_info *sbi, | |
458e6197 | 109 | enum page_type type, int rw) |
93dfe2ac JK |
110 | { |
111 | enum page_type btype = PAGE_TYPE_OF_BIO(type); | |
112 | struct f2fs_bio_info *io; | |
113 | ||
114 | io = is_read_io(rw) ? &sbi->read_io : &sbi->write_io[btype]; | |
115 | ||
df0f8dc0 | 116 | down_write(&io->io_rwsem); |
458e6197 JK |
117 | |
118 | /* change META to META_FLUSH in the checkpoint procedure */ | |
119 | if (type >= META_FLUSH) { | |
120 | io->fio.type = META_FLUSH; | |
0f7b2abd JK |
121 | if (test_opt(sbi, NOBARRIER)) |
122 | io->fio.rw = WRITE_FLUSH | REQ_META | REQ_PRIO; | |
123 | else | |
124 | io->fio.rw = WRITE_FLUSH_FUA | REQ_META | REQ_PRIO; | |
458e6197 JK |
125 | } |
126 | __submit_merged_bio(io); | |
df0f8dc0 | 127 | up_write(&io->io_rwsem); |
93dfe2ac JK |
128 | } |
129 | ||
130 | /* | |
131 | * Fill the locked page with data located in the block address. | |
132 | * Return unlocked page. | |
133 | */ | |
134 | int f2fs_submit_page_bio(struct f2fs_sb_info *sbi, struct page *page, | |
135 | block_t blk_addr, int rw) | |
136 | { | |
93dfe2ac JK |
137 | struct bio *bio; |
138 | ||
139 | trace_f2fs_submit_page_bio(page, blk_addr, rw); | |
140 | ||
141 | /* Allocate a new bio */ | |
940a6d34 | 142 | bio = __bio_alloc(sbi, blk_addr, 1, is_read_io(rw)); |
93dfe2ac JK |
143 | |
144 | if (bio_add_page(bio, page, PAGE_CACHE_SIZE, 0) < PAGE_CACHE_SIZE) { | |
145 | bio_put(bio); | |
146 | f2fs_put_page(page, 1); | |
147 | return -EFAULT; | |
148 | } | |
149 | ||
150 | submit_bio(rw, bio); | |
151 | return 0; | |
152 | } | |
153 | ||
154 | void f2fs_submit_page_mbio(struct f2fs_sb_info *sbi, struct page *page, | |
458e6197 | 155 | block_t blk_addr, struct f2fs_io_info *fio) |
93dfe2ac | 156 | { |
458e6197 | 157 | enum page_type btype = PAGE_TYPE_OF_BIO(fio->type); |
93dfe2ac | 158 | struct f2fs_bio_info *io; |
940a6d34 | 159 | bool is_read = is_read_io(fio->rw); |
93dfe2ac | 160 | |
940a6d34 | 161 | io = is_read ? &sbi->read_io : &sbi->write_io[btype]; |
93dfe2ac JK |
162 | |
163 | verify_block_addr(sbi, blk_addr); | |
164 | ||
df0f8dc0 | 165 | down_write(&io->io_rwsem); |
93dfe2ac | 166 | |
940a6d34 | 167 | if (!is_read) |
93dfe2ac JK |
168 | inc_page_count(sbi, F2FS_WRITEBACK); |
169 | ||
63a0b7cb | 170 | if (io->bio && (io->last_block_in_bio != blk_addr - 1 || |
458e6197 JK |
171 | io->fio.rw != fio->rw)) |
172 | __submit_merged_bio(io); | |
93dfe2ac JK |
173 | alloc_new: |
174 | if (io->bio == NULL) { | |
90a893c7 | 175 | int bio_blocks = MAX_BIO_BLOCKS(sbi); |
940a6d34 GZ |
176 | |
177 | io->bio = __bio_alloc(sbi, blk_addr, bio_blocks, is_read); | |
458e6197 | 178 | io->fio = *fio; |
93dfe2ac JK |
179 | } |
180 | ||
181 | if (bio_add_page(io->bio, page, PAGE_CACHE_SIZE, 0) < | |
182 | PAGE_CACHE_SIZE) { | |
458e6197 | 183 | __submit_merged_bio(io); |
93dfe2ac JK |
184 | goto alloc_new; |
185 | } | |
186 | ||
187 | io->last_block_in_bio = blk_addr; | |
188 | ||
df0f8dc0 | 189 | up_write(&io->io_rwsem); |
458e6197 | 190 | trace_f2fs_submit_page_mbio(page, fio->rw, fio->type, blk_addr); |
93dfe2ac JK |
191 | } |
192 | ||
0a8165d7 | 193 | /* |
eb47b800 JK |
194 | * Lock ordering for the change of data block address: |
195 | * ->data_page | |
196 | * ->node_page | |
197 | * update block addresses in the node page | |
198 | */ | |
199 | static void __set_data_blkaddr(struct dnode_of_data *dn, block_t new_addr) | |
200 | { | |
201 | struct f2fs_node *rn; | |
202 | __le32 *addr_array; | |
203 | struct page *node_page = dn->node_page; | |
204 | unsigned int ofs_in_node = dn->ofs_in_node; | |
205 | ||
5514f0aa | 206 | f2fs_wait_on_page_writeback(node_page, NODE); |
eb47b800 | 207 | |
45590710 | 208 | rn = F2FS_NODE(node_page); |
eb47b800 JK |
209 | |
210 | /* Get physical address of data block */ | |
211 | addr_array = blkaddr_in_node(rn); | |
212 | addr_array[ofs_in_node] = cpu_to_le32(new_addr); | |
213 | set_page_dirty(node_page); | |
214 | } | |
215 | ||
216 | int reserve_new_block(struct dnode_of_data *dn) | |
217 | { | |
4081363f | 218 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
eb47b800 | 219 | |
6bacf52f | 220 | if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))) |
eb47b800 | 221 | return -EPERM; |
cfb271d4 | 222 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, 1))) |
eb47b800 JK |
223 | return -ENOSPC; |
224 | ||
c01e2853 NJ |
225 | trace_f2fs_reserve_new_block(dn->inode, dn->nid, dn->ofs_in_node); |
226 | ||
eb47b800 JK |
227 | __set_data_blkaddr(dn, NEW_ADDR); |
228 | dn->data_blkaddr = NEW_ADDR; | |
a18ff063 | 229 | mark_inode_dirty(dn->inode); |
eb47b800 JK |
230 | sync_inode_page(dn); |
231 | return 0; | |
232 | } | |
233 | ||
b600965c HL |
234 | int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index) |
235 | { | |
236 | bool need_put = dn->inode_page ? false : true; | |
237 | int err; | |
238 | ||
239 | err = get_dnode_of_data(dn, index, ALLOC_NODE); | |
240 | if (err) | |
241 | return err; | |
a8865372 | 242 | |
b600965c HL |
243 | if (dn->data_blkaddr == NULL_ADDR) |
244 | err = reserve_new_block(dn); | |
a8865372 | 245 | if (err || need_put) |
b600965c HL |
246 | f2fs_put_dnode(dn); |
247 | return err; | |
248 | } | |
249 | ||
eb47b800 JK |
250 | static int check_extent_cache(struct inode *inode, pgoff_t pgofs, |
251 | struct buffer_head *bh_result) | |
252 | { | |
253 | struct f2fs_inode_info *fi = F2FS_I(inode); | |
eb47b800 JK |
254 | pgoff_t start_fofs, end_fofs; |
255 | block_t start_blkaddr; | |
256 | ||
c11abd1a JK |
257 | if (is_inode_flag_set(fi, FI_NO_EXTENT)) |
258 | return 0; | |
259 | ||
eb47b800 JK |
260 | read_lock(&fi->ext.ext_lock); |
261 | if (fi->ext.len == 0) { | |
262 | read_unlock(&fi->ext.ext_lock); | |
263 | return 0; | |
264 | } | |
265 | ||
dcdfff65 JK |
266 | stat_inc_total_hit(inode->i_sb); |
267 | ||
eb47b800 JK |
268 | start_fofs = fi->ext.fofs; |
269 | end_fofs = fi->ext.fofs + fi->ext.len - 1; | |
270 | start_blkaddr = fi->ext.blk_addr; | |
271 | ||
272 | if (pgofs >= start_fofs && pgofs <= end_fofs) { | |
273 | unsigned int blkbits = inode->i_sb->s_blocksize_bits; | |
274 | size_t count; | |
275 | ||
276 | clear_buffer_new(bh_result); | |
277 | map_bh(bh_result, inode->i_sb, | |
278 | start_blkaddr + pgofs - start_fofs); | |
279 | count = end_fofs - pgofs + 1; | |
280 | if (count < (UINT_MAX >> blkbits)) | |
281 | bh_result->b_size = (count << blkbits); | |
282 | else | |
283 | bh_result->b_size = UINT_MAX; | |
284 | ||
dcdfff65 | 285 | stat_inc_read_hit(inode->i_sb); |
eb47b800 JK |
286 | read_unlock(&fi->ext.ext_lock); |
287 | return 1; | |
288 | } | |
289 | read_unlock(&fi->ext.ext_lock); | |
290 | return 0; | |
291 | } | |
292 | ||
293 | void update_extent_cache(block_t blk_addr, struct dnode_of_data *dn) | |
294 | { | |
295 | struct f2fs_inode_info *fi = F2FS_I(dn->inode); | |
296 | pgoff_t fofs, start_fofs, end_fofs; | |
297 | block_t start_blkaddr, end_blkaddr; | |
c11abd1a | 298 | int need_update = true; |
eb47b800 | 299 | |
9850cf4a | 300 | f2fs_bug_on(F2FS_I_SB(dn->inode), blk_addr == NEW_ADDR); |
de93653f JK |
301 | fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) + |
302 | dn->ofs_in_node; | |
eb47b800 JK |
303 | |
304 | /* Update the page address in the parent node */ | |
305 | __set_data_blkaddr(dn, blk_addr); | |
306 | ||
c11abd1a JK |
307 | if (is_inode_flag_set(fi, FI_NO_EXTENT)) |
308 | return; | |
309 | ||
eb47b800 JK |
310 | write_lock(&fi->ext.ext_lock); |
311 | ||
312 | start_fofs = fi->ext.fofs; | |
313 | end_fofs = fi->ext.fofs + fi->ext.len - 1; | |
314 | start_blkaddr = fi->ext.blk_addr; | |
315 | end_blkaddr = fi->ext.blk_addr + fi->ext.len - 1; | |
316 | ||
317 | /* Drop and initialize the matched extent */ | |
318 | if (fi->ext.len == 1 && fofs == start_fofs) | |
319 | fi->ext.len = 0; | |
320 | ||
321 | /* Initial extent */ | |
322 | if (fi->ext.len == 0) { | |
323 | if (blk_addr != NULL_ADDR) { | |
324 | fi->ext.fofs = fofs; | |
325 | fi->ext.blk_addr = blk_addr; | |
326 | fi->ext.len = 1; | |
327 | } | |
328 | goto end_update; | |
329 | } | |
330 | ||
6224da87 | 331 | /* Front merge */ |
eb47b800 JK |
332 | if (fofs == start_fofs - 1 && blk_addr == start_blkaddr - 1) { |
333 | fi->ext.fofs--; | |
334 | fi->ext.blk_addr--; | |
335 | fi->ext.len++; | |
336 | goto end_update; | |
337 | } | |
338 | ||
339 | /* Back merge */ | |
340 | if (fofs == end_fofs + 1 && blk_addr == end_blkaddr + 1) { | |
341 | fi->ext.len++; | |
342 | goto end_update; | |
343 | } | |
344 | ||
345 | /* Split the existing extent */ | |
346 | if (fi->ext.len > 1 && | |
347 | fofs >= start_fofs && fofs <= end_fofs) { | |
348 | if ((end_fofs - fofs) < (fi->ext.len >> 1)) { | |
349 | fi->ext.len = fofs - start_fofs; | |
350 | } else { | |
351 | fi->ext.fofs = fofs + 1; | |
352 | fi->ext.blk_addr = start_blkaddr + | |
353 | fofs - start_fofs + 1; | |
354 | fi->ext.len -= fofs - start_fofs + 1; | |
355 | } | |
c11abd1a JK |
356 | } else { |
357 | need_update = false; | |
eb47b800 | 358 | } |
eb47b800 | 359 | |
c11abd1a JK |
360 | /* Finally, if the extent is very fragmented, let's drop the cache. */ |
361 | if (fi->ext.len < F2FS_MIN_EXTENT_LEN) { | |
362 | fi->ext.len = 0; | |
363 | set_inode_flag(fi, FI_NO_EXTENT); | |
364 | need_update = true; | |
365 | } | |
eb47b800 JK |
366 | end_update: |
367 | write_unlock(&fi->ext.ext_lock); | |
c11abd1a JK |
368 | if (need_update) |
369 | sync_inode_page(dn); | |
370 | return; | |
eb47b800 JK |
371 | } |
372 | ||
c718379b | 373 | struct page *find_data_page(struct inode *inode, pgoff_t index, bool sync) |
eb47b800 | 374 | { |
eb47b800 JK |
375 | struct address_space *mapping = inode->i_mapping; |
376 | struct dnode_of_data dn; | |
377 | struct page *page; | |
378 | int err; | |
379 | ||
380 | page = find_get_page(mapping, index); | |
381 | if (page && PageUptodate(page)) | |
382 | return page; | |
383 | f2fs_put_page(page, 0); | |
384 | ||
385 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
266e97a8 | 386 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); |
eb47b800 JK |
387 | if (err) |
388 | return ERR_PTR(err); | |
389 | f2fs_put_dnode(&dn); | |
390 | ||
391 | if (dn.data_blkaddr == NULL_ADDR) | |
392 | return ERR_PTR(-ENOENT); | |
393 | ||
394 | /* By fallocate(), there is no cached page, but with NEW_ADDR */ | |
6bacf52f | 395 | if (unlikely(dn.data_blkaddr == NEW_ADDR)) |
eb47b800 JK |
396 | return ERR_PTR(-EINVAL); |
397 | ||
9ac1349a | 398 | page = grab_cache_page(mapping, index); |
eb47b800 JK |
399 | if (!page) |
400 | return ERR_PTR(-ENOMEM); | |
401 | ||
393ff91f JK |
402 | if (PageUptodate(page)) { |
403 | unlock_page(page); | |
404 | return page; | |
405 | } | |
406 | ||
4081363f | 407 | err = f2fs_submit_page_bio(F2FS_I_SB(inode), page, dn.data_blkaddr, |
c718379b | 408 | sync ? READ_SYNC : READA); |
1069bbf7 CY |
409 | if (err) |
410 | return ERR_PTR(err); | |
411 | ||
c718379b JK |
412 | if (sync) { |
413 | wait_on_page_locked(page); | |
6bacf52f | 414 | if (unlikely(!PageUptodate(page))) { |
c718379b JK |
415 | f2fs_put_page(page, 0); |
416 | return ERR_PTR(-EIO); | |
417 | } | |
eb47b800 | 418 | } |
eb47b800 JK |
419 | return page; |
420 | } | |
421 | ||
0a8165d7 | 422 | /* |
eb47b800 JK |
423 | * If it tries to access a hole, return an error. |
424 | * Because, the callers, functions in dir.c and GC, should be able to know | |
425 | * whether this page exists or not. | |
426 | */ | |
427 | struct page *get_lock_data_page(struct inode *inode, pgoff_t index) | |
428 | { | |
eb47b800 JK |
429 | struct address_space *mapping = inode->i_mapping; |
430 | struct dnode_of_data dn; | |
431 | struct page *page; | |
432 | int err; | |
433 | ||
650495de | 434 | repeat: |
9ac1349a | 435 | page = grab_cache_page(mapping, index); |
650495de JK |
436 | if (!page) |
437 | return ERR_PTR(-ENOMEM); | |
438 | ||
eb47b800 | 439 | set_new_dnode(&dn, inode, NULL, NULL, 0); |
266e97a8 | 440 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); |
650495de JK |
441 | if (err) { |
442 | f2fs_put_page(page, 1); | |
eb47b800 | 443 | return ERR_PTR(err); |
650495de | 444 | } |
eb47b800 JK |
445 | f2fs_put_dnode(&dn); |
446 | ||
6bacf52f | 447 | if (unlikely(dn.data_blkaddr == NULL_ADDR)) { |
650495de | 448 | f2fs_put_page(page, 1); |
eb47b800 | 449 | return ERR_PTR(-ENOENT); |
650495de | 450 | } |
eb47b800 JK |
451 | |
452 | if (PageUptodate(page)) | |
453 | return page; | |
454 | ||
d59ff4df JK |
455 | /* |
456 | * A new dentry page is allocated but not able to be written, since its | |
457 | * new inode page couldn't be allocated due to -ENOSPC. | |
458 | * In such the case, its blkaddr can be remained as NEW_ADDR. | |
459 | * see, f2fs_add_link -> get_new_data_page -> init_inode_metadata. | |
460 | */ | |
461 | if (dn.data_blkaddr == NEW_ADDR) { | |
462 | zero_user_segment(page, 0, PAGE_CACHE_SIZE); | |
463 | SetPageUptodate(page); | |
464 | return page; | |
465 | } | |
eb47b800 | 466 | |
4081363f JK |
467 | err = f2fs_submit_page_bio(F2FS_I_SB(inode), page, |
468 | dn.data_blkaddr, READ_SYNC); | |
393ff91f | 469 | if (err) |
eb47b800 | 470 | return ERR_PTR(err); |
393ff91f JK |
471 | |
472 | lock_page(page); | |
6bacf52f | 473 | if (unlikely(!PageUptodate(page))) { |
393ff91f JK |
474 | f2fs_put_page(page, 1); |
475 | return ERR_PTR(-EIO); | |
eb47b800 | 476 | } |
6bacf52f | 477 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
478 | f2fs_put_page(page, 1); |
479 | goto repeat; | |
eb47b800 JK |
480 | } |
481 | return page; | |
482 | } | |
483 | ||
0a8165d7 | 484 | /* |
eb47b800 JK |
485 | * Caller ensures that this data page is never allocated. |
486 | * A new zero-filled data page is allocated in the page cache. | |
39936837 | 487 | * |
4f4124d0 CY |
488 | * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and |
489 | * f2fs_unlock_op(). | |
a8865372 | 490 | * Note that, ipage is set only by make_empty_dir. |
eb47b800 | 491 | */ |
64aa7ed9 | 492 | struct page *get_new_data_page(struct inode *inode, |
a8865372 | 493 | struct page *ipage, pgoff_t index, bool new_i_size) |
eb47b800 | 494 | { |
eb47b800 JK |
495 | struct address_space *mapping = inode->i_mapping; |
496 | struct page *page; | |
497 | struct dnode_of_data dn; | |
498 | int err; | |
499 | ||
a8865372 | 500 | set_new_dnode(&dn, inode, ipage, NULL, 0); |
b600965c | 501 | err = f2fs_reserve_block(&dn, index); |
eb47b800 JK |
502 | if (err) |
503 | return ERR_PTR(err); | |
afcb7ca0 | 504 | repeat: |
eb47b800 | 505 | page = grab_cache_page(mapping, index); |
a8865372 JK |
506 | if (!page) { |
507 | err = -ENOMEM; | |
508 | goto put_err; | |
509 | } | |
eb47b800 JK |
510 | |
511 | if (PageUptodate(page)) | |
512 | return page; | |
513 | ||
514 | if (dn.data_blkaddr == NEW_ADDR) { | |
515 | zero_user_segment(page, 0, PAGE_CACHE_SIZE); | |
393ff91f | 516 | SetPageUptodate(page); |
eb47b800 | 517 | } else { |
4081363f JK |
518 | err = f2fs_submit_page_bio(F2FS_I_SB(inode), page, |
519 | dn.data_blkaddr, READ_SYNC); | |
393ff91f | 520 | if (err) |
a8865372 JK |
521 | goto put_err; |
522 | ||
393ff91f | 523 | lock_page(page); |
6bacf52f | 524 | if (unlikely(!PageUptodate(page))) { |
393ff91f | 525 | f2fs_put_page(page, 1); |
a8865372 JK |
526 | err = -EIO; |
527 | goto put_err; | |
eb47b800 | 528 | } |
6bacf52f | 529 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
530 | f2fs_put_page(page, 1); |
531 | goto repeat; | |
eb47b800 JK |
532 | } |
533 | } | |
eb47b800 JK |
534 | |
535 | if (new_i_size && | |
536 | i_size_read(inode) < ((index + 1) << PAGE_CACHE_SHIFT)) { | |
537 | i_size_write(inode, ((index + 1) << PAGE_CACHE_SHIFT)); | |
699489bb JK |
538 | /* Only the directory inode sets new_i_size */ |
539 | set_inode_flag(F2FS_I(inode), FI_UPDATE_DIR); | |
eb47b800 JK |
540 | } |
541 | return page; | |
a8865372 JK |
542 | |
543 | put_err: | |
544 | f2fs_put_dnode(&dn); | |
545 | return ERR_PTR(err); | |
eb47b800 JK |
546 | } |
547 | ||
bfad7c2d JK |
548 | static int __allocate_data_block(struct dnode_of_data *dn) |
549 | { | |
4081363f | 550 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
976e4c50 | 551 | struct f2fs_inode_info *fi = F2FS_I(dn->inode); |
bfad7c2d JK |
552 | struct f2fs_summary sum; |
553 | block_t new_blkaddr; | |
554 | struct node_info ni; | |
976e4c50 | 555 | pgoff_t fofs; |
bfad7c2d JK |
556 | int type; |
557 | ||
558 | if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))) | |
559 | return -EPERM; | |
560 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, 1))) | |
561 | return -ENOSPC; | |
562 | ||
563 | __set_data_blkaddr(dn, NEW_ADDR); | |
564 | dn->data_blkaddr = NEW_ADDR; | |
565 | ||
566 | get_node_info(sbi, dn->nid, &ni); | |
567 | set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version); | |
568 | ||
569 | type = CURSEG_WARM_DATA; | |
570 | ||
571 | allocate_data_block(sbi, NULL, NULL_ADDR, &new_blkaddr, &sum, type); | |
572 | ||
573 | /* direct IO doesn't use extent cache to maximize the performance */ | |
574 | set_inode_flag(F2FS_I(dn->inode), FI_NO_EXTENT); | |
575 | update_extent_cache(new_blkaddr, dn); | |
576 | clear_inode_flag(F2FS_I(dn->inode), FI_NO_EXTENT); | |
577 | ||
976e4c50 JK |
578 | /* update i_size */ |
579 | fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) + | |
580 | dn->ofs_in_node; | |
581 | if (i_size_read(dn->inode) < ((fofs + 1) << PAGE_CACHE_SHIFT)) | |
582 | i_size_write(dn->inode, ((fofs + 1) << PAGE_CACHE_SHIFT)); | |
583 | ||
bfad7c2d JK |
584 | dn->data_blkaddr = new_blkaddr; |
585 | return 0; | |
586 | } | |
587 | ||
0a8165d7 | 588 | /* |
4f4124d0 CY |
589 | * get_data_block() now supported readahead/bmap/rw direct_IO with mapped bh. |
590 | * If original data blocks are allocated, then give them to blockdev. | |
591 | * Otherwise, | |
592 | * a. preallocate requested block addresses | |
593 | * b. do not use extent cache for better performance | |
594 | * c. give the block addresses to blockdev | |
eb47b800 | 595 | */ |
ccfb3000 JK |
596 | static int __get_data_block(struct inode *inode, sector_t iblock, |
597 | struct buffer_head *bh_result, int create, bool fiemap) | |
eb47b800 JK |
598 | { |
599 | unsigned int blkbits = inode->i_sb->s_blocksize_bits; | |
600 | unsigned maxblocks = bh_result->b_size >> blkbits; | |
601 | struct dnode_of_data dn; | |
bfad7c2d JK |
602 | int mode = create ? ALLOC_NODE : LOOKUP_NODE_RA; |
603 | pgoff_t pgofs, end_offset; | |
604 | int err = 0, ofs = 1; | |
605 | bool allocated = false; | |
eb47b800 JK |
606 | |
607 | /* Get the page offset from the block offset(iblock) */ | |
608 | pgofs = (pgoff_t)(iblock >> (PAGE_CACHE_SHIFT - blkbits)); | |
609 | ||
bfad7c2d JK |
610 | if (check_extent_cache(inode, pgofs, bh_result)) |
611 | goto out; | |
612 | ||
79e35dc3 | 613 | if (create) { |
4081363f JK |
614 | f2fs_balance_fs(F2FS_I_SB(inode)); |
615 | f2fs_lock_op(F2FS_I_SB(inode)); | |
79e35dc3 | 616 | } |
eb47b800 JK |
617 | |
618 | /* When reading holes, we need its node page */ | |
619 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
bfad7c2d | 620 | err = get_dnode_of_data(&dn, pgofs, mode); |
1ec79083 | 621 | if (err) { |
bfad7c2d JK |
622 | if (err == -ENOENT) |
623 | err = 0; | |
624 | goto unlock_out; | |
848753aa | 625 | } |
ccfb3000 | 626 | if (dn.data_blkaddr == NEW_ADDR && !fiemap) |
1ec79083 | 627 | goto put_out; |
eb47b800 | 628 | |
bfad7c2d JK |
629 | if (dn.data_blkaddr != NULL_ADDR) { |
630 | map_bh(bh_result, inode->i_sb, dn.data_blkaddr); | |
631 | } else if (create) { | |
632 | err = __allocate_data_block(&dn); | |
633 | if (err) | |
634 | goto put_out; | |
635 | allocated = true; | |
636 | map_bh(bh_result, inode->i_sb, dn.data_blkaddr); | |
637 | } else { | |
638 | goto put_out; | |
639 | } | |
640 | ||
6403eb1f | 641 | end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode)); |
bfad7c2d JK |
642 | bh_result->b_size = (((size_t)1) << blkbits); |
643 | dn.ofs_in_node++; | |
644 | pgofs++; | |
645 | ||
646 | get_next: | |
647 | if (dn.ofs_in_node >= end_offset) { | |
648 | if (allocated) | |
649 | sync_inode_page(&dn); | |
650 | allocated = false; | |
651 | f2fs_put_dnode(&dn); | |
652 | ||
653 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
654 | err = get_dnode_of_data(&dn, pgofs, mode); | |
1ec79083 | 655 | if (err) { |
bfad7c2d JK |
656 | if (err == -ENOENT) |
657 | err = 0; | |
658 | goto unlock_out; | |
659 | } | |
ccfb3000 | 660 | if (dn.data_blkaddr == NEW_ADDR && !fiemap) |
1ec79083 JK |
661 | goto put_out; |
662 | ||
6403eb1f | 663 | end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode)); |
bfad7c2d | 664 | } |
eb47b800 | 665 | |
bfad7c2d JK |
666 | if (maxblocks > (bh_result->b_size >> blkbits)) { |
667 | block_t blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node); | |
668 | if (blkaddr == NULL_ADDR && create) { | |
669 | err = __allocate_data_block(&dn); | |
670 | if (err) | |
671 | goto sync_out; | |
672 | allocated = true; | |
673 | blkaddr = dn.data_blkaddr; | |
674 | } | |
e1c42045 | 675 | /* Give more consecutive addresses for the readahead */ |
bfad7c2d JK |
676 | if (blkaddr == (bh_result->b_blocknr + ofs)) { |
677 | ofs++; | |
678 | dn.ofs_in_node++; | |
679 | pgofs++; | |
680 | bh_result->b_size += (((size_t)1) << blkbits); | |
681 | goto get_next; | |
682 | } | |
eb47b800 | 683 | } |
bfad7c2d JK |
684 | sync_out: |
685 | if (allocated) | |
686 | sync_inode_page(&dn); | |
687 | put_out: | |
eb47b800 | 688 | f2fs_put_dnode(&dn); |
bfad7c2d JK |
689 | unlock_out: |
690 | if (create) | |
4081363f | 691 | f2fs_unlock_op(F2FS_I_SB(inode)); |
bfad7c2d JK |
692 | out: |
693 | trace_f2fs_get_data_block(inode, iblock, bh_result, err); | |
694 | return err; | |
eb47b800 JK |
695 | } |
696 | ||
ccfb3000 JK |
697 | static int get_data_block(struct inode *inode, sector_t iblock, |
698 | struct buffer_head *bh_result, int create) | |
699 | { | |
700 | return __get_data_block(inode, iblock, bh_result, create, false); | |
701 | } | |
702 | ||
703 | static int get_data_block_fiemap(struct inode *inode, sector_t iblock, | |
704 | struct buffer_head *bh_result, int create) | |
705 | { | |
706 | return __get_data_block(inode, iblock, bh_result, create, true); | |
707 | } | |
708 | ||
9ab70134 JK |
709 | int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, |
710 | u64 start, u64 len) | |
711 | { | |
ccfb3000 JK |
712 | return generic_block_fiemap(inode, fieinfo, |
713 | start, len, get_data_block_fiemap); | |
9ab70134 JK |
714 | } |
715 | ||
eb47b800 JK |
716 | static int f2fs_read_data_page(struct file *file, struct page *page) |
717 | { | |
9ffe0fb5 | 718 | struct inode *inode = page->mapping->host; |
b3d208f9 | 719 | int ret = -EAGAIN; |
9ffe0fb5 | 720 | |
c20e89cd CY |
721 | trace_f2fs_readpage(page, DATA); |
722 | ||
e1c42045 | 723 | /* If the file has inline data, try to read it directly */ |
9ffe0fb5 HL |
724 | if (f2fs_has_inline_data(inode)) |
725 | ret = f2fs_read_inline_data(inode, page); | |
b3d208f9 | 726 | if (ret == -EAGAIN) |
9ffe0fb5 HL |
727 | ret = mpage_readpage(page, get_data_block); |
728 | ||
729 | return ret; | |
eb47b800 JK |
730 | } |
731 | ||
732 | static int f2fs_read_data_pages(struct file *file, | |
733 | struct address_space *mapping, | |
734 | struct list_head *pages, unsigned nr_pages) | |
735 | { | |
9ffe0fb5 HL |
736 | struct inode *inode = file->f_mapping->host; |
737 | ||
738 | /* If the file has inline data, skip readpages */ | |
739 | if (f2fs_has_inline_data(inode)) | |
740 | return 0; | |
741 | ||
bfad7c2d | 742 | return mpage_readpages(mapping, pages, nr_pages, get_data_block); |
eb47b800 JK |
743 | } |
744 | ||
458e6197 | 745 | int do_write_data_page(struct page *page, struct f2fs_io_info *fio) |
eb47b800 JK |
746 | { |
747 | struct inode *inode = page->mapping->host; | |
458e6197 | 748 | block_t old_blkaddr, new_blkaddr; |
eb47b800 JK |
749 | struct dnode_of_data dn; |
750 | int err = 0; | |
751 | ||
752 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
266e97a8 | 753 | err = get_dnode_of_data(&dn, page->index, LOOKUP_NODE); |
eb47b800 JK |
754 | if (err) |
755 | return err; | |
756 | ||
458e6197 | 757 | old_blkaddr = dn.data_blkaddr; |
eb47b800 JK |
758 | |
759 | /* This page is already truncated */ | |
458e6197 | 760 | if (old_blkaddr == NULL_ADDR) |
eb47b800 JK |
761 | goto out_writepage; |
762 | ||
763 | set_page_writeback(page); | |
764 | ||
765 | /* | |
766 | * If current allocation needs SSR, | |
767 | * it had better in-place writes for updated data. | |
768 | */ | |
458e6197 | 769 | if (unlikely(old_blkaddr != NEW_ADDR && |
b25958b6 HL |
770 | !is_cold_data(page) && |
771 | need_inplace_update(inode))) { | |
458e6197 | 772 | rewrite_data_page(page, old_blkaddr, fio); |
fff04f90 | 773 | set_inode_flag(F2FS_I(inode), FI_UPDATE_WRITE); |
eb47b800 | 774 | } else { |
458e6197 JK |
775 | write_data_page(page, &dn, &new_blkaddr, fio); |
776 | update_extent_cache(new_blkaddr, &dn); | |
fff04f90 | 777 | set_inode_flag(F2FS_I(inode), FI_APPEND_WRITE); |
eb47b800 JK |
778 | } |
779 | out_writepage: | |
780 | f2fs_put_dnode(&dn); | |
781 | return err; | |
782 | } | |
783 | ||
784 | static int f2fs_write_data_page(struct page *page, | |
785 | struct writeback_control *wbc) | |
786 | { | |
787 | struct inode *inode = page->mapping->host; | |
4081363f | 788 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
eb47b800 JK |
789 | loff_t i_size = i_size_read(inode); |
790 | const pgoff_t end_index = ((unsigned long long) i_size) | |
791 | >> PAGE_CACHE_SHIFT; | |
9ffe0fb5 | 792 | unsigned offset = 0; |
39936837 | 793 | bool need_balance_fs = false; |
eb47b800 | 794 | int err = 0; |
458e6197 JK |
795 | struct f2fs_io_info fio = { |
796 | .type = DATA, | |
6c311ec6 | 797 | .rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE, |
458e6197 | 798 | }; |
eb47b800 | 799 | |
ecda0de3 CY |
800 | trace_f2fs_writepage(page, DATA); |
801 | ||
eb47b800 | 802 | if (page->index < end_index) |
39936837 | 803 | goto write; |
eb47b800 JK |
804 | |
805 | /* | |
806 | * If the offset is out-of-range of file size, | |
807 | * this page does not have to be written to disk. | |
808 | */ | |
809 | offset = i_size & (PAGE_CACHE_SIZE - 1); | |
76f60268 | 810 | if ((page->index >= end_index + 1) || !offset) |
39936837 | 811 | goto out; |
eb47b800 JK |
812 | |
813 | zero_user_segment(page, offset, PAGE_CACHE_SIZE); | |
39936837 | 814 | write: |
8618b881 | 815 | if (unlikely(sbi->por_doing)) |
eb47b800 JK |
816 | goto redirty_out; |
817 | ||
39936837 | 818 | /* Dentry blocks are controlled by checkpoint */ |
eb47b800 | 819 | if (S_ISDIR(inode->i_mode)) { |
cf779cab JK |
820 | if (unlikely(f2fs_cp_error(sbi))) |
821 | goto redirty_out; | |
458e6197 | 822 | err = do_write_data_page(page, &fio); |
8618b881 JK |
823 | goto done; |
824 | } | |
9ffe0fb5 | 825 | |
cf779cab JK |
826 | /* we should bypass data pages to proceed the kworkder jobs */ |
827 | if (unlikely(f2fs_cp_error(sbi))) { | |
828 | SetPageError(page); | |
829 | unlock_page(page); | |
a7ffdbe2 | 830 | goto out; |
cf779cab JK |
831 | } |
832 | ||
8618b881 | 833 | if (!wbc->for_reclaim) |
39936837 | 834 | need_balance_fs = true; |
8618b881 | 835 | else if (has_not_enough_free_secs(sbi, 0)) |
39936837 | 836 | goto redirty_out; |
eb47b800 | 837 | |
b3d208f9 | 838 | err = -EAGAIN; |
8618b881 | 839 | f2fs_lock_op(sbi); |
b3d208f9 JK |
840 | if (f2fs_has_inline_data(inode)) |
841 | err = f2fs_write_inline_data(inode, page); | |
842 | if (err == -EAGAIN) | |
8618b881 JK |
843 | err = do_write_data_page(page, &fio); |
844 | f2fs_unlock_op(sbi); | |
845 | done: | |
846 | if (err && err != -ENOENT) | |
847 | goto redirty_out; | |
eb47b800 | 848 | |
eb47b800 | 849 | clear_cold_data(page); |
39936837 | 850 | out: |
a7ffdbe2 | 851 | inode_dec_dirty_pages(inode); |
eb47b800 | 852 | unlock_page(page); |
39936837 | 853 | if (need_balance_fs) |
eb47b800 | 854 | f2fs_balance_fs(sbi); |
2aea39ec JK |
855 | if (wbc->for_reclaim) |
856 | f2fs_submit_merged_bio(sbi, DATA, WRITE); | |
eb47b800 JK |
857 | return 0; |
858 | ||
eb47b800 | 859 | redirty_out: |
76f60268 | 860 | redirty_page_for_writepage(wbc, page); |
8618b881 | 861 | return AOP_WRITEPAGE_ACTIVATE; |
eb47b800 JK |
862 | } |
863 | ||
fa9150a8 NJ |
864 | static int __f2fs_writepage(struct page *page, struct writeback_control *wbc, |
865 | void *data) | |
866 | { | |
867 | struct address_space *mapping = data; | |
868 | int ret = mapping->a_ops->writepage(page, wbc); | |
869 | mapping_set_error(mapping, ret); | |
870 | return ret; | |
871 | } | |
872 | ||
25ca923b | 873 | static int f2fs_write_data_pages(struct address_space *mapping, |
eb47b800 JK |
874 | struct writeback_control *wbc) |
875 | { | |
876 | struct inode *inode = mapping->host; | |
4081363f | 877 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
531ad7d5 | 878 | bool locked = false; |
eb47b800 | 879 | int ret; |
50c8cdb3 | 880 | long diff; |
eb47b800 | 881 | |
e5748434 CY |
882 | trace_f2fs_writepages(mapping->host, wbc, DATA); |
883 | ||
cfb185a1 | 884 | /* deal with chardevs and other special file */ |
885 | if (!mapping->a_ops->writepage) | |
886 | return 0; | |
887 | ||
87d6f890 | 888 | if (S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_NONE && |
a7ffdbe2 | 889 | get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) && |
6fb03f3a | 890 | available_free_memory(sbi, DIRTY_DENTS)) |
d3baf95d | 891 | goto skip_write; |
87d6f890 | 892 | |
50c8cdb3 | 893 | diff = nr_pages_to_write(sbi, DATA, wbc); |
eb47b800 | 894 | |
531ad7d5 | 895 | if (!S_ISDIR(inode->i_mode)) { |
eb47b800 | 896 | mutex_lock(&sbi->writepages); |
531ad7d5 JK |
897 | locked = true; |
898 | } | |
fa9150a8 | 899 | ret = write_cache_pages(mapping, wbc, __f2fs_writepage, mapping); |
531ad7d5 | 900 | if (locked) |
eb47b800 | 901 | mutex_unlock(&sbi->writepages); |
458e6197 JK |
902 | |
903 | f2fs_submit_merged_bio(sbi, DATA, WRITE); | |
eb47b800 JK |
904 | |
905 | remove_dirty_dir_inode(inode); | |
906 | ||
50c8cdb3 | 907 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff); |
eb47b800 | 908 | return ret; |
d3baf95d JK |
909 | |
910 | skip_write: | |
a7ffdbe2 | 911 | wbc->pages_skipped += get_dirty_pages(inode); |
d3baf95d | 912 | return 0; |
eb47b800 JK |
913 | } |
914 | ||
3aab8f82 CY |
915 | static void f2fs_write_failed(struct address_space *mapping, loff_t to) |
916 | { | |
917 | struct inode *inode = mapping->host; | |
918 | ||
919 | if (to > inode->i_size) { | |
920 | truncate_pagecache(inode, inode->i_size); | |
764aa3e9 | 921 | truncate_blocks(inode, inode->i_size, true); |
3aab8f82 CY |
922 | } |
923 | } | |
924 | ||
eb47b800 JK |
925 | static int f2fs_write_begin(struct file *file, struct address_space *mapping, |
926 | loff_t pos, unsigned len, unsigned flags, | |
927 | struct page **pagep, void **fsdata) | |
928 | { | |
929 | struct inode *inode = mapping->host; | |
4081363f | 930 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
9ba69cf9 | 931 | struct page *page, *ipage; |
eb47b800 JK |
932 | pgoff_t index = ((unsigned long long) pos) >> PAGE_CACHE_SHIFT; |
933 | struct dnode_of_data dn; | |
934 | int err = 0; | |
935 | ||
62aed044 CY |
936 | trace_f2fs_write_begin(inode, pos, len, flags); |
937 | ||
eb47b800 | 938 | f2fs_balance_fs(sbi); |
afcb7ca0 | 939 | repeat: |
eb47b800 | 940 | page = grab_cache_page_write_begin(mapping, index, flags); |
3aab8f82 CY |
941 | if (!page) { |
942 | err = -ENOMEM; | |
943 | goto fail; | |
944 | } | |
d5f66990 | 945 | |
eb47b800 JK |
946 | *pagep = page; |
947 | ||
e479556b | 948 | f2fs_lock_op(sbi); |
9ba69cf9 JK |
949 | |
950 | /* check inline_data */ | |
951 | ipage = get_node_page(sbi, inode->i_ino); | |
952 | if (IS_ERR(ipage)) | |
953 | goto unlock_fail; | |
954 | ||
b3d208f9 JK |
955 | set_new_dnode(&dn, inode, ipage, ipage, 0); |
956 | ||
9ba69cf9 | 957 | if (f2fs_has_inline_data(inode)) { |
b3d208f9 JK |
958 | if (pos + len <= MAX_INLINE_DATA) { |
959 | read_inline_data(page, ipage); | |
960 | set_inode_flag(F2FS_I(inode), FI_DATA_EXIST); | |
961 | sync_inode_page(&dn); | |
962 | goto put_next; | |
963 | } else if (page->index == 0) { | |
964 | err = f2fs_convert_inline_page(&dn, page); | |
965 | if (err) | |
8cdcb713 | 966 | goto put_fail; |
b3d208f9 JK |
967 | } else { |
968 | struct page *p = grab_cache_page(inode->i_mapping, 0); | |
969 | if (!p) { | |
970 | err = -ENOMEM; | |
8cdcb713 | 971 | goto put_fail; |
b3d208f9 JK |
972 | } |
973 | err = f2fs_convert_inline_page(&dn, p); | |
974 | f2fs_put_page(p, 1); | |
975 | if (err) | |
8cdcb713 | 976 | goto put_fail; |
b3d208f9 | 977 | } |
b600965c | 978 | } |
9ba69cf9 JK |
979 | err = f2fs_reserve_block(&dn, index); |
980 | if (err) | |
8cdcb713 | 981 | goto put_fail; |
b3d208f9 | 982 | put_next: |
9ba69cf9 JK |
983 | f2fs_put_dnode(&dn); |
984 | f2fs_unlock_op(sbi); | |
985 | ||
eb47b800 JK |
986 | if ((len == PAGE_CACHE_SIZE) || PageUptodate(page)) |
987 | return 0; | |
988 | ||
b3d208f9 JK |
989 | f2fs_wait_on_page_writeback(page, DATA); |
990 | ||
eb47b800 JK |
991 | if ((pos & PAGE_CACHE_MASK) >= i_size_read(inode)) { |
992 | unsigned start = pos & (PAGE_CACHE_SIZE - 1); | |
993 | unsigned end = start + len; | |
994 | ||
995 | /* Reading beyond i_size is simple: memset to zero */ | |
996 | zero_user_segments(page, 0, start, end, PAGE_CACHE_SIZE); | |
393ff91f | 997 | goto out; |
eb47b800 JK |
998 | } |
999 | ||
b3d208f9 | 1000 | if (dn.data_blkaddr == NEW_ADDR) { |
eb47b800 JK |
1001 | zero_user_segment(page, 0, PAGE_CACHE_SIZE); |
1002 | } else { | |
9234f319 JK |
1003 | err = f2fs_submit_page_bio(sbi, page, dn.data_blkaddr, |
1004 | READ_SYNC); | |
1005 | if (err) | |
1006 | goto fail; | |
d54c795b | 1007 | |
393ff91f | 1008 | lock_page(page); |
6bacf52f | 1009 | if (unlikely(!PageUptodate(page))) { |
393ff91f | 1010 | f2fs_put_page(page, 1); |
3aab8f82 CY |
1011 | err = -EIO; |
1012 | goto fail; | |
eb47b800 | 1013 | } |
6bacf52f | 1014 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
1015 | f2fs_put_page(page, 1); |
1016 | goto repeat; | |
eb47b800 JK |
1017 | } |
1018 | } | |
393ff91f | 1019 | out: |
eb47b800 JK |
1020 | SetPageUptodate(page); |
1021 | clear_cold_data(page); | |
1022 | return 0; | |
9ba69cf9 | 1023 | |
8cdcb713 JK |
1024 | put_fail: |
1025 | f2fs_put_dnode(&dn); | |
9ba69cf9 JK |
1026 | unlock_fail: |
1027 | f2fs_unlock_op(sbi); | |
b3d208f9 | 1028 | f2fs_put_page(page, 1); |
3aab8f82 CY |
1029 | fail: |
1030 | f2fs_write_failed(mapping, pos + len); | |
1031 | return err; | |
eb47b800 JK |
1032 | } |
1033 | ||
a1dd3c13 JK |
1034 | static int f2fs_write_end(struct file *file, |
1035 | struct address_space *mapping, | |
1036 | loff_t pos, unsigned len, unsigned copied, | |
1037 | struct page *page, void *fsdata) | |
1038 | { | |
1039 | struct inode *inode = page->mapping->host; | |
1040 | ||
dfb2bf38 CY |
1041 | trace_f2fs_write_end(inode, pos, len, copied); |
1042 | ||
34ba94ba | 1043 | set_page_dirty(page); |
a1dd3c13 JK |
1044 | |
1045 | if (pos + copied > i_size_read(inode)) { | |
1046 | i_size_write(inode, pos + copied); | |
1047 | mark_inode_dirty(inode); | |
1048 | update_inode_page(inode); | |
1049 | } | |
1050 | ||
75c3c8bc | 1051 | f2fs_put_page(page, 1); |
a1dd3c13 JK |
1052 | return copied; |
1053 | } | |
1054 | ||
944fcfc1 | 1055 | static int check_direct_IO(struct inode *inode, int rw, |
5b46f25d | 1056 | struct iov_iter *iter, loff_t offset) |
944fcfc1 JK |
1057 | { |
1058 | unsigned blocksize_mask = inode->i_sb->s_blocksize - 1; | |
944fcfc1 JK |
1059 | |
1060 | if (rw == READ) | |
1061 | return 0; | |
1062 | ||
1063 | if (offset & blocksize_mask) | |
1064 | return -EINVAL; | |
1065 | ||
5b46f25d AV |
1066 | if (iov_iter_alignment(iter) & blocksize_mask) |
1067 | return -EINVAL; | |
1068 | ||
944fcfc1 JK |
1069 | return 0; |
1070 | } | |
1071 | ||
eb47b800 | 1072 | static ssize_t f2fs_direct_IO(int rw, struct kiocb *iocb, |
d8d3d94b | 1073 | struct iov_iter *iter, loff_t offset) |
eb47b800 JK |
1074 | { |
1075 | struct file *file = iocb->ki_filp; | |
3aab8f82 CY |
1076 | struct address_space *mapping = file->f_mapping; |
1077 | struct inode *inode = mapping->host; | |
1078 | size_t count = iov_iter_count(iter); | |
1079 | int err; | |
944fcfc1 | 1080 | |
b3d208f9 JK |
1081 | /* we don't need to use inline_data strictly */ |
1082 | if (f2fs_has_inline_data(inode)) { | |
1083 | err = f2fs_convert_inline_inode(inode); | |
1084 | if (err) | |
1085 | return err; | |
1086 | } | |
9ffe0fb5 | 1087 | |
5b46f25d | 1088 | if (check_direct_IO(inode, rw, iter, offset)) |
944fcfc1 JK |
1089 | return 0; |
1090 | ||
70407fad CY |
1091 | trace_f2fs_direct_IO_enter(inode, offset, count, rw); |
1092 | ||
3aab8f82 CY |
1093 | err = blockdev_direct_IO(rw, iocb, inode, iter, offset, get_data_block); |
1094 | if (err < 0 && (rw & WRITE)) | |
1095 | f2fs_write_failed(mapping, offset + count); | |
70407fad CY |
1096 | |
1097 | trace_f2fs_direct_IO_exit(inode, offset, count, rw, err); | |
1098 | ||
3aab8f82 | 1099 | return err; |
eb47b800 JK |
1100 | } |
1101 | ||
d47992f8 LC |
1102 | static void f2fs_invalidate_data_page(struct page *page, unsigned int offset, |
1103 | unsigned int length) | |
eb47b800 JK |
1104 | { |
1105 | struct inode *inode = page->mapping->host; | |
a7ffdbe2 JK |
1106 | |
1107 | if (offset % PAGE_CACHE_SIZE || length != PAGE_CACHE_SIZE) | |
1108 | return; | |
1109 | ||
cbcb2872 JK |
1110 | if (f2fs_is_atomic_file(inode) || f2fs_is_volatile_file(inode)) |
1111 | invalidate_inmem_page(inode, page); | |
1112 | ||
1fe54f9d | 1113 | if (PageDirty(page)) |
a7ffdbe2 | 1114 | inode_dec_dirty_pages(inode); |
eb47b800 JK |
1115 | ClearPagePrivate(page); |
1116 | } | |
1117 | ||
1118 | static int f2fs_release_data_page(struct page *page, gfp_t wait) | |
1119 | { | |
1120 | ClearPagePrivate(page); | |
c3850aa1 | 1121 | return 1; |
eb47b800 JK |
1122 | } |
1123 | ||
1124 | static int f2fs_set_data_page_dirty(struct page *page) | |
1125 | { | |
1126 | struct address_space *mapping = page->mapping; | |
1127 | struct inode *inode = mapping->host; | |
1128 | ||
26c6b887 JK |
1129 | trace_f2fs_set_page_dirty(page, DATA); |
1130 | ||
eb47b800 | 1131 | SetPageUptodate(page); |
34ba94ba JK |
1132 | |
1133 | if (f2fs_is_atomic_file(inode) || f2fs_is_volatile_file(inode)) { | |
1134 | register_inmem_page(inode, page); | |
1135 | return 1; | |
1136 | } | |
1137 | ||
a18ff063 JK |
1138 | mark_inode_dirty(inode); |
1139 | ||
eb47b800 JK |
1140 | if (!PageDirty(page)) { |
1141 | __set_page_dirty_nobuffers(page); | |
a7ffdbe2 | 1142 | update_dirty_page(inode, page); |
eb47b800 JK |
1143 | return 1; |
1144 | } | |
1145 | return 0; | |
1146 | } | |
1147 | ||
c01e54b7 JK |
1148 | static sector_t f2fs_bmap(struct address_space *mapping, sector_t block) |
1149 | { | |
454ae7e5 CY |
1150 | struct inode *inode = mapping->host; |
1151 | ||
b3d208f9 JK |
1152 | /* we don't need to use inline_data strictly */ |
1153 | if (f2fs_has_inline_data(inode)) { | |
1154 | int err = f2fs_convert_inline_inode(inode); | |
1155 | if (err) | |
1156 | return err; | |
1157 | } | |
bfad7c2d | 1158 | return generic_block_bmap(mapping, block, get_data_block); |
c01e54b7 JK |
1159 | } |
1160 | ||
eb47b800 JK |
1161 | const struct address_space_operations f2fs_dblock_aops = { |
1162 | .readpage = f2fs_read_data_page, | |
1163 | .readpages = f2fs_read_data_pages, | |
1164 | .writepage = f2fs_write_data_page, | |
1165 | .writepages = f2fs_write_data_pages, | |
1166 | .write_begin = f2fs_write_begin, | |
a1dd3c13 | 1167 | .write_end = f2fs_write_end, |
eb47b800 JK |
1168 | .set_page_dirty = f2fs_set_data_page_dirty, |
1169 | .invalidatepage = f2fs_invalidate_data_page, | |
1170 | .releasepage = f2fs_release_data_page, | |
1171 | .direct_IO = f2fs_direct_IO, | |
c01e54b7 | 1172 | .bmap = f2fs_bmap, |
eb47b800 | 1173 | }; |