Commit | Line | Data |
---|---|---|
0a8165d7 | 1 | /* |
127e670a JK |
2 | * fs/f2fs/checkpoint.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/bio.h> | |
13 | #include <linux/mpage.h> | |
14 | #include <linux/writeback.h> | |
15 | #include <linux/blkdev.h> | |
16 | #include <linux/f2fs_fs.h> | |
17 | #include <linux/pagevec.h> | |
18 | #include <linux/swap.h> | |
19 | ||
20 | #include "f2fs.h" | |
21 | #include "node.h" | |
22 | #include "segment.h" | |
9e4ded3f | 23 | #include "trace.h" |
2af4bd6c | 24 | #include <trace/events/f2fs.h> |
127e670a | 25 | |
6451e041 | 26 | static struct kmem_cache *ino_entry_slab; |
06292073 | 27 | struct kmem_cache *inode_entry_slab; |
127e670a | 28 | |
38f91ca8 JK |
29 | void f2fs_stop_checkpoint(struct f2fs_sb_info *sbi, bool end_io) |
30 | { | |
aaec2b1d | 31 | set_ckpt_flags(sbi, CP_ERROR_FLAG); |
38f91ca8 JK |
32 | sbi->sb->s_flags |= MS_RDONLY; |
33 | if (!end_io) | |
34 | f2fs_flush_merged_bios(sbi); | |
35 | } | |
36 | ||
0a8165d7 | 37 | /* |
127e670a JK |
38 | * We guarantee no failure on the returned page. |
39 | */ | |
40 | struct page *grab_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
41 | { | |
9df27d98 | 42 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
43 | struct page *page = NULL; |
44 | repeat: | |
300e129c | 45 | page = f2fs_grab_cache_page(mapping, index, false); |
127e670a JK |
46 | if (!page) { |
47 | cond_resched(); | |
48 | goto repeat; | |
49 | } | |
fec1d657 | 50 | f2fs_wait_on_page_writeback(page, META, true); |
237c0790 JK |
51 | if (!PageUptodate(page)) |
52 | SetPageUptodate(page); | |
127e670a JK |
53 | return page; |
54 | } | |
55 | ||
0a8165d7 | 56 | /* |
127e670a JK |
57 | * We guarantee no failure on the returned page. |
58 | */ | |
2b947003 CY |
59 | static struct page *__get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index, |
60 | bool is_meta) | |
127e670a | 61 | { |
9df27d98 | 62 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a | 63 | struct page *page; |
cf04e8eb | 64 | struct f2fs_io_info fio = { |
05ca3632 | 65 | .sbi = sbi, |
cf04e8eb | 66 | .type = META, |
04d328de MC |
67 | .op = REQ_OP_READ, |
68 | .op_flags = READ_SYNC | REQ_META | REQ_PRIO, | |
7a9d7548 CY |
69 | .old_blkaddr = index, |
70 | .new_blkaddr = index, | |
4375a336 | 71 | .encrypted_page = NULL, |
cf04e8eb | 72 | }; |
2b947003 CY |
73 | |
74 | if (unlikely(!is_meta)) | |
04d328de | 75 | fio.op_flags &= ~REQ_META; |
127e670a | 76 | repeat: |
300e129c | 77 | page = f2fs_grab_cache_page(mapping, index, false); |
127e670a JK |
78 | if (!page) { |
79 | cond_resched(); | |
80 | goto repeat; | |
81 | } | |
393ff91f JK |
82 | if (PageUptodate(page)) |
83 | goto out; | |
84 | ||
05ca3632 JK |
85 | fio.page = page; |
86 | ||
86531d6b JK |
87 | if (f2fs_submit_page_bio(&fio)) { |
88 | f2fs_put_page(page, 1); | |
127e670a | 89 | goto repeat; |
86531d6b | 90 | } |
127e670a | 91 | |
393ff91f | 92 | lock_page(page); |
6bacf52f | 93 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
94 | f2fs_put_page(page, 1); |
95 | goto repeat; | |
96 | } | |
f3f338ca CY |
97 | |
98 | /* | |
99 | * if there is any IO error when accessing device, make our filesystem | |
100 | * readonly and make sure do not write checkpoint with non-uptodate | |
101 | * meta page. | |
102 | */ | |
103 | if (unlikely(!PageUptodate(page))) | |
38f91ca8 | 104 | f2fs_stop_checkpoint(sbi, false); |
393ff91f | 105 | out: |
127e670a JK |
106 | return page; |
107 | } | |
108 | ||
2b947003 CY |
109 | struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) |
110 | { | |
111 | return __get_meta_page(sbi, index, true); | |
112 | } | |
113 | ||
114 | /* for POR only */ | |
115 | struct page *get_tmp_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
116 | { | |
117 | return __get_meta_page(sbi, index, false); | |
118 | } | |
119 | ||
f0c9cada | 120 | bool is_valid_blkaddr(struct f2fs_sb_info *sbi, block_t blkaddr, int type) |
662befda CY |
121 | { |
122 | switch (type) { | |
123 | case META_NAT: | |
66b00c18 | 124 | break; |
662befda | 125 | case META_SIT: |
66b00c18 CY |
126 | if (unlikely(blkaddr >= SIT_BLK_CNT(sbi))) |
127 | return false; | |
128 | break; | |
81c1a0f1 | 129 | case META_SSA: |
66b00c18 CY |
130 | if (unlikely(blkaddr >= MAIN_BLKADDR(sbi) || |
131 | blkaddr < SM_I(sbi)->ssa_blkaddr)) | |
132 | return false; | |
133 | break; | |
662befda | 134 | case META_CP: |
66b00c18 CY |
135 | if (unlikely(blkaddr >= SIT_I(sbi)->sit_base_addr || |
136 | blkaddr < __start_cp_addr(sbi))) | |
137 | return false; | |
138 | break; | |
4c521f49 | 139 | case META_POR: |
66b00c18 CY |
140 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
141 | blkaddr < MAIN_BLKADDR(sbi))) | |
142 | return false; | |
143 | break; | |
662befda CY |
144 | default: |
145 | BUG(); | |
146 | } | |
66b00c18 CY |
147 | |
148 | return true; | |
662befda CY |
149 | } |
150 | ||
151 | /* | |
81c1a0f1 | 152 | * Readahead CP/NAT/SIT/SSA pages |
662befda | 153 | */ |
26879fb1 CY |
154 | int ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages, |
155 | int type, bool sync) | |
662befda | 156 | { |
662befda | 157 | struct page *page; |
4c521f49 | 158 | block_t blkno = start; |
662befda | 159 | struct f2fs_io_info fio = { |
05ca3632 | 160 | .sbi = sbi, |
662befda | 161 | .type = META, |
04d328de | 162 | .op = REQ_OP_READ, |
70246286 | 163 | .op_flags = sync ? (READ_SYNC | REQ_META | REQ_PRIO) : REQ_RAHEAD, |
4375a336 | 164 | .encrypted_page = NULL, |
662befda | 165 | }; |
e9f5b8b8 | 166 | struct blk_plug plug; |
662befda | 167 | |
2b947003 | 168 | if (unlikely(type == META_POR)) |
04d328de | 169 | fio.op_flags &= ~REQ_META; |
2b947003 | 170 | |
e9f5b8b8 | 171 | blk_start_plug(&plug); |
662befda | 172 | for (; nrpages-- > 0; blkno++) { |
662befda | 173 | |
66b00c18 CY |
174 | if (!is_valid_blkaddr(sbi, blkno, type)) |
175 | goto out; | |
176 | ||
662befda CY |
177 | switch (type) { |
178 | case META_NAT: | |
66b00c18 CY |
179 | if (unlikely(blkno >= |
180 | NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))) | |
662befda | 181 | blkno = 0; |
66b00c18 | 182 | /* get nat block addr */ |
7a9d7548 | 183 | fio.new_blkaddr = current_nat_addr(sbi, |
662befda CY |
184 | blkno * NAT_ENTRY_PER_BLOCK); |
185 | break; | |
186 | case META_SIT: | |
187 | /* get sit block addr */ | |
7a9d7548 | 188 | fio.new_blkaddr = current_sit_addr(sbi, |
662befda | 189 | blkno * SIT_ENTRY_PER_BLOCK); |
662befda | 190 | break; |
81c1a0f1 | 191 | case META_SSA: |
662befda | 192 | case META_CP: |
4c521f49 | 193 | case META_POR: |
7a9d7548 | 194 | fio.new_blkaddr = blkno; |
662befda CY |
195 | break; |
196 | default: | |
197 | BUG(); | |
198 | } | |
199 | ||
300e129c JK |
200 | page = f2fs_grab_cache_page(META_MAPPING(sbi), |
201 | fio.new_blkaddr, false); | |
662befda CY |
202 | if (!page) |
203 | continue; | |
204 | if (PageUptodate(page)) { | |
662befda CY |
205 | f2fs_put_page(page, 1); |
206 | continue; | |
207 | } | |
208 | ||
05ca3632 | 209 | fio.page = page; |
7a9d7548 | 210 | fio.old_blkaddr = fio.new_blkaddr; |
05ca3632 | 211 | f2fs_submit_page_mbio(&fio); |
662befda CY |
212 | f2fs_put_page(page, 0); |
213 | } | |
214 | out: | |
215 | f2fs_submit_merged_bio(sbi, META, READ); | |
e9f5b8b8 | 216 | blk_finish_plug(&plug); |
662befda CY |
217 | return blkno - start; |
218 | } | |
219 | ||
635aee1f CY |
220 | void ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index) |
221 | { | |
222 | struct page *page; | |
223 | bool readahead = false; | |
224 | ||
225 | page = find_get_page(META_MAPPING(sbi), index); | |
4da7bf5a | 226 | if (!page || !PageUptodate(page)) |
635aee1f CY |
227 | readahead = true; |
228 | f2fs_put_page(page, 0); | |
229 | ||
230 | if (readahead) | |
26879fb1 | 231 | ra_meta_pages(sbi, index, MAX_BIO_BLOCKS(sbi), META_POR, true); |
635aee1f CY |
232 | } |
233 | ||
127e670a JK |
234 | static int f2fs_write_meta_page(struct page *page, |
235 | struct writeback_control *wbc) | |
236 | { | |
4081363f | 237 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 238 | |
ecda0de3 CY |
239 | trace_f2fs_writepage(page, META); |
240 | ||
caf0047e | 241 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
cfb271d4 | 242 | goto redirty_out; |
857dc4e0 | 243 | if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0)) |
cfb271d4 | 244 | goto redirty_out; |
1e968fdf | 245 | if (unlikely(f2fs_cp_error(sbi))) |
cf779cab | 246 | goto redirty_out; |
127e670a | 247 | |
577e3495 JK |
248 | write_meta_page(sbi, page); |
249 | dec_page_count(sbi, F2FS_DIRTY_META); | |
0c3a5797 CY |
250 | |
251 | if (wbc->for_reclaim) | |
252 | f2fs_submit_merged_bio_cond(sbi, NULL, page, 0, META, WRITE); | |
253 | ||
577e3495 | 254 | unlock_page(page); |
857dc4e0 | 255 | |
0c3a5797 | 256 | if (unlikely(f2fs_cp_error(sbi))) |
857dc4e0 | 257 | f2fs_submit_merged_bio(sbi, META, WRITE); |
0c3a5797 | 258 | |
577e3495 | 259 | return 0; |
cfb271d4 CY |
260 | |
261 | redirty_out: | |
76f60268 | 262 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 263 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
264 | } |
265 | ||
266 | static int f2fs_write_meta_pages(struct address_space *mapping, | |
267 | struct writeback_control *wbc) | |
268 | { | |
4081363f | 269 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 270 | long diff, written; |
127e670a | 271 | |
5459aa97 | 272 | /* collect a number of dirty meta pages and write together */ |
50c8cdb3 JK |
273 | if (wbc->for_kupdate || |
274 | get_pages(sbi, F2FS_DIRTY_META) < nr_pages_to_skip(sbi, META)) | |
d3baf95d | 275 | goto skip_write; |
127e670a | 276 | |
d31c7c3f YH |
277 | trace_f2fs_writepages(mapping->host, wbc, META); |
278 | ||
127e670a JK |
279 | /* if mounting is failed, skip writing node pages */ |
280 | mutex_lock(&sbi->cp_mutex); | |
50c8cdb3 JK |
281 | diff = nr_pages_to_write(sbi, META, wbc); |
282 | written = sync_meta_pages(sbi, META, wbc->nr_to_write); | |
127e670a | 283 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 284 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 285 | return 0; |
d3baf95d JK |
286 | |
287 | skip_write: | |
288 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
d31c7c3f | 289 | trace_f2fs_writepages(mapping->host, wbc, META); |
d3baf95d | 290 | return 0; |
127e670a JK |
291 | } |
292 | ||
293 | long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, | |
294 | long nr_to_write) | |
295 | { | |
9df27d98 | 296 | struct address_space *mapping = META_MAPPING(sbi); |
80dd9c0e | 297 | pgoff_t index = 0, end = ULONG_MAX, prev = ULONG_MAX; |
127e670a JK |
298 | struct pagevec pvec; |
299 | long nwritten = 0; | |
300 | struct writeback_control wbc = { | |
301 | .for_reclaim = 0, | |
302 | }; | |
e9f5b8b8 | 303 | struct blk_plug plug; |
127e670a JK |
304 | |
305 | pagevec_init(&pvec, 0); | |
306 | ||
e9f5b8b8 CY |
307 | blk_start_plug(&plug); |
308 | ||
127e670a JK |
309 | while (index <= end) { |
310 | int i, nr_pages; | |
311 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, | |
312 | PAGECACHE_TAG_DIRTY, | |
313 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
cfb271d4 | 314 | if (unlikely(nr_pages == 0)) |
127e670a JK |
315 | break; |
316 | ||
317 | for (i = 0; i < nr_pages; i++) { | |
318 | struct page *page = pvec.pages[i]; | |
203681f6 | 319 | |
80dd9c0e | 320 | if (prev == ULONG_MAX) |
6066d8cd JK |
321 | prev = page->index - 1; |
322 | if (nr_to_write != LONG_MAX && page->index != prev + 1) { | |
323 | pagevec_release(&pvec); | |
324 | goto stop; | |
325 | } | |
326 | ||
127e670a | 327 | lock_page(page); |
203681f6 JK |
328 | |
329 | if (unlikely(page->mapping != mapping)) { | |
330 | continue_unlock: | |
331 | unlock_page(page); | |
332 | continue; | |
333 | } | |
334 | if (!PageDirty(page)) { | |
335 | /* someone wrote it for us */ | |
336 | goto continue_unlock; | |
337 | } | |
338 | ||
fa3d2bdf JK |
339 | f2fs_wait_on_page_writeback(page, META, true); |
340 | ||
341 | BUG_ON(PageWriteback(page)); | |
203681f6 JK |
342 | if (!clear_page_dirty_for_io(page)) |
343 | goto continue_unlock; | |
344 | ||
97dc3fd2 | 345 | if (mapping->a_ops->writepage(page, &wbc)) { |
577e3495 JK |
346 | unlock_page(page); |
347 | break; | |
348 | } | |
cfb271d4 | 349 | nwritten++; |
6066d8cd | 350 | prev = page->index; |
cfb271d4 | 351 | if (unlikely(nwritten >= nr_to_write)) |
127e670a JK |
352 | break; |
353 | } | |
354 | pagevec_release(&pvec); | |
355 | cond_resched(); | |
356 | } | |
6066d8cd | 357 | stop: |
127e670a | 358 | if (nwritten) |
458e6197 | 359 | f2fs_submit_merged_bio(sbi, type, WRITE); |
127e670a | 360 | |
e9f5b8b8 CY |
361 | blk_finish_plug(&plug); |
362 | ||
127e670a JK |
363 | return nwritten; |
364 | } | |
365 | ||
366 | static int f2fs_set_meta_page_dirty(struct page *page) | |
367 | { | |
26c6b887 JK |
368 | trace_f2fs_set_page_dirty(page, META); |
369 | ||
237c0790 JK |
370 | if (!PageUptodate(page)) |
371 | SetPageUptodate(page); | |
127e670a | 372 | if (!PageDirty(page)) { |
fe76b796 | 373 | f2fs_set_page_dirty_nobuffers(page); |
4081363f | 374 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); |
1601839e | 375 | SetPagePrivate(page); |
9e4ded3f | 376 | f2fs_trace_pid(page); |
127e670a JK |
377 | return 1; |
378 | } | |
379 | return 0; | |
380 | } | |
381 | ||
382 | const struct address_space_operations f2fs_meta_aops = { | |
383 | .writepage = f2fs_write_meta_page, | |
384 | .writepages = f2fs_write_meta_pages, | |
385 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
487261f3 CY |
386 | .invalidatepage = f2fs_invalidate_page, |
387 | .releasepage = f2fs_release_page, | |
5b7a487c WG |
388 | #ifdef CONFIG_MIGRATION |
389 | .migratepage = f2fs_migrate_page, | |
390 | #endif | |
127e670a JK |
391 | }; |
392 | ||
6451e041 | 393 | static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 394 | { |
67298804 | 395 | struct inode_management *im = &sbi->im[type]; |
80c54505 JK |
396 | struct ino_entry *e, *tmp; |
397 | ||
398 | tmp = f2fs_kmem_cache_alloc(ino_entry_slab, GFP_NOFS); | |
39efac41 | 399 | retry: |
80c54505 | 400 | radix_tree_preload(GFP_NOFS | __GFP_NOFAIL); |
769ec6e5 | 401 | |
67298804 | 402 | spin_lock(&im->ino_lock); |
67298804 | 403 | e = radix_tree_lookup(&im->ino_root, ino); |
39efac41 | 404 | if (!e) { |
80c54505 | 405 | e = tmp; |
67298804 CY |
406 | if (radix_tree_insert(&im->ino_root, ino, e)) { |
407 | spin_unlock(&im->ino_lock); | |
769ec6e5 | 408 | radix_tree_preload_end(); |
39efac41 JK |
409 | goto retry; |
410 | } | |
411 | memset(e, 0, sizeof(struct ino_entry)); | |
412 | e->ino = ino; | |
953e6cc6 | 413 | |
67298804 | 414 | list_add_tail(&e->list, &im->ino_list); |
8c402946 | 415 | if (type != ORPHAN_INO) |
67298804 | 416 | im->ino_num++; |
39efac41 | 417 | } |
67298804 | 418 | spin_unlock(&im->ino_lock); |
769ec6e5 | 419 | radix_tree_preload_end(); |
80c54505 JK |
420 | |
421 | if (e != tmp) | |
422 | kmem_cache_free(ino_entry_slab, tmp); | |
953e6cc6 JK |
423 | } |
424 | ||
6451e041 | 425 | static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 426 | { |
67298804 | 427 | struct inode_management *im = &sbi->im[type]; |
6451e041 | 428 | struct ino_entry *e; |
953e6cc6 | 429 | |
67298804 CY |
430 | spin_lock(&im->ino_lock); |
431 | e = radix_tree_lookup(&im->ino_root, ino); | |
39efac41 JK |
432 | if (e) { |
433 | list_del(&e->list); | |
67298804 CY |
434 | radix_tree_delete(&im->ino_root, ino); |
435 | im->ino_num--; | |
436 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
437 | kmem_cache_free(ino_entry_slab, e); |
438 | return; | |
953e6cc6 | 439 | } |
67298804 | 440 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
441 | } |
442 | ||
a49324f1 | 443 | void add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
444 | { |
445 | /* add new dirty ino entry into list */ | |
446 | __add_ino_entry(sbi, ino, type); | |
447 | } | |
448 | ||
a49324f1 | 449 | void remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
450 | { |
451 | /* remove dirty ino entry from list */ | |
452 | __remove_ino_entry(sbi, ino, type); | |
453 | } | |
454 | ||
455 | /* mode should be APPEND_INO or UPDATE_INO */ | |
456 | bool exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode) | |
457 | { | |
67298804 | 458 | struct inode_management *im = &sbi->im[mode]; |
fff04f90 | 459 | struct ino_entry *e; |
67298804 CY |
460 | |
461 | spin_lock(&im->ino_lock); | |
462 | e = radix_tree_lookup(&im->ino_root, ino); | |
463 | spin_unlock(&im->ino_lock); | |
fff04f90 JK |
464 | return e ? true : false; |
465 | } | |
466 | ||
74ef9241 | 467 | void release_ino_entry(struct f2fs_sb_info *sbi, bool all) |
fff04f90 JK |
468 | { |
469 | struct ino_entry *e, *tmp; | |
470 | int i; | |
471 | ||
74ef9241 | 472 | for (i = all ? ORPHAN_INO: APPEND_INO; i <= UPDATE_INO; i++) { |
67298804 CY |
473 | struct inode_management *im = &sbi->im[i]; |
474 | ||
475 | spin_lock(&im->ino_lock); | |
476 | list_for_each_entry_safe(e, tmp, &im->ino_list, list) { | |
fff04f90 | 477 | list_del(&e->list); |
67298804 | 478 | radix_tree_delete(&im->ino_root, e->ino); |
fff04f90 | 479 | kmem_cache_free(ino_entry_slab, e); |
67298804 | 480 | im->ino_num--; |
fff04f90 | 481 | } |
67298804 | 482 | spin_unlock(&im->ino_lock); |
fff04f90 JK |
483 | } |
484 | } | |
485 | ||
cbd56e7d | 486 | int acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 487 | { |
67298804 | 488 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a JK |
489 | int err = 0; |
490 | ||
67298804 | 491 | spin_lock(&im->ino_lock); |
cb78942b JK |
492 | |
493 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
1ecc0c5c | 494 | if (time_to_inject(sbi, FAULT_ORPHAN)) { |
cb78942b JK |
495 | spin_unlock(&im->ino_lock); |
496 | return -ENOSPC; | |
497 | } | |
498 | #endif | |
67298804 | 499 | if (unlikely(im->ino_num >= sbi->max_orphans)) |
127e670a | 500 | err = -ENOSPC; |
cbd56e7d | 501 | else |
67298804 CY |
502 | im->ino_num++; |
503 | spin_unlock(&im->ino_lock); | |
0d47c1ad | 504 | |
127e670a JK |
505 | return err; |
506 | } | |
507 | ||
cbd56e7d JK |
508 | void release_orphan_inode(struct f2fs_sb_info *sbi) |
509 | { | |
67298804 CY |
510 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
511 | ||
512 | spin_lock(&im->ino_lock); | |
513 | f2fs_bug_on(sbi, im->ino_num == 0); | |
514 | im->ino_num--; | |
515 | spin_unlock(&im->ino_lock); | |
cbd56e7d JK |
516 | } |
517 | ||
67c3758d | 518 | void add_orphan_inode(struct inode *inode) |
127e670a | 519 | { |
39efac41 | 520 | /* add new orphan ino entry into list */ |
67c3758d JK |
521 | __add_ino_entry(F2FS_I_SB(inode), inode->i_ino, ORPHAN_INO); |
522 | update_inode_page(inode); | |
127e670a JK |
523 | } |
524 | ||
525 | void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
526 | { | |
953e6cc6 | 527 | /* remove orphan entry from orphan list */ |
6451e041 | 528 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
529 | } |
530 | ||
8c14bfad | 531 | static int recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 532 | { |
8c14bfad | 533 | struct inode *inode; |
5905f9af | 534 | struct node_info ni; |
d41065e2 JK |
535 | int err = acquire_orphan_inode(sbi); |
536 | ||
537 | if (err) { | |
538 | set_sbi_flag(sbi, SBI_NEED_FSCK); | |
539 | f2fs_msg(sbi->sb, KERN_WARNING, | |
540 | "%s: orphan failed (ino=%x), run fsck to fix.", | |
541 | __func__, ino); | |
542 | return err; | |
543 | } | |
544 | ||
545 | __add_ino_entry(sbi, ino, ORPHAN_INO); | |
8c14bfad | 546 | |
5905f9af | 547 | inode = f2fs_iget_retry(sbi->sb, ino); |
8c14bfad CY |
548 | if (IS_ERR(inode)) { |
549 | /* | |
550 | * there should be a bug that we can't find the entry | |
551 | * to orphan inode. | |
552 | */ | |
553 | f2fs_bug_on(sbi, PTR_ERR(inode) == -ENOENT); | |
554 | return PTR_ERR(inode); | |
555 | } | |
556 | ||
127e670a JK |
557 | clear_nlink(inode); |
558 | ||
559 | /* truncate all the data during iput */ | |
560 | iput(inode); | |
5905f9af JK |
561 | |
562 | get_node_info(sbi, ino, &ni); | |
563 | ||
564 | /* ENOMEM was fully retried in f2fs_evict_inode. */ | |
565 | if (ni.blk_addr != NULL_ADDR) { | |
d41065e2 JK |
566 | set_sbi_flag(sbi, SBI_NEED_FSCK); |
567 | f2fs_msg(sbi->sb, KERN_WARNING, | |
568 | "%s: orphan failed (ino=%x), run fsck to fix.", | |
569 | __func__, ino); | |
570 | return -EIO; | |
5905f9af | 571 | } |
d41065e2 | 572 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
8c14bfad | 573 | return 0; |
127e670a JK |
574 | } |
575 | ||
8c14bfad | 576 | int recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a | 577 | { |
3c642985 | 578 | block_t start_blk, orphan_blocks, i, j; |
8c14bfad | 579 | int err; |
127e670a | 580 | |
aaec2b1d | 581 | if (!is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG)) |
8c14bfad | 582 | return 0; |
127e670a | 583 | |
55141486 | 584 | start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi); |
3c642985 | 585 | orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi); |
127e670a | 586 | |
26879fb1 | 587 | ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP, true); |
662befda | 588 | |
3c642985 | 589 | for (i = 0; i < orphan_blocks; i++) { |
127e670a JK |
590 | struct page *page = get_meta_page(sbi, start_blk + i); |
591 | struct f2fs_orphan_block *orphan_blk; | |
592 | ||
593 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); | |
594 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
595 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
8c14bfad CY |
596 | err = recover_orphan_inode(sbi, ino); |
597 | if (err) { | |
598 | f2fs_put_page(page, 1); | |
8c14bfad CY |
599 | return err; |
600 | } | |
127e670a JK |
601 | } |
602 | f2fs_put_page(page, 1); | |
603 | } | |
604 | /* clear Orphan Flag */ | |
aaec2b1d | 605 | clear_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG); |
8c14bfad | 606 | return 0; |
127e670a JK |
607 | } |
608 | ||
609 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
610 | { | |
502c6e0b | 611 | struct list_head *head; |
127e670a | 612 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 613 | unsigned int nentries = 0; |
bd936f84 | 614 | unsigned short index = 1; |
8c402946 | 615 | unsigned short orphan_blocks; |
4531929e | 616 | struct page *page = NULL; |
6451e041 | 617 | struct ino_entry *orphan = NULL; |
67298804 | 618 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 619 | |
67298804 | 620 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 621 | |
d6c67a4f JK |
622 | /* |
623 | * we don't need to do spin_lock(&im->ino_lock) here, since all the | |
624 | * orphan inode operations are covered under f2fs_lock_op(). | |
625 | * And, spin_lock should be avoided due to page operations below. | |
626 | */ | |
67298804 | 627 | head = &im->ino_list; |
127e670a JK |
628 | |
629 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
630 | list_for_each_entry(orphan, head, list) { |
631 | if (!page) { | |
bd936f84 | 632 | page = grab_meta_page(sbi, start_blk++); |
502c6e0b GZ |
633 | orphan_blk = |
634 | (struct f2fs_orphan_block *)page_address(page); | |
635 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
636 | } | |
127e670a | 637 | |
36795567 | 638 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 639 | |
36795567 | 640 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
641 | /* |
642 | * an orphan block is full of 1020 entries, | |
643 | * then we need to flush current orphan blocks | |
644 | * and bring another one in memory | |
645 | */ | |
646 | orphan_blk->blk_addr = cpu_to_le16(index); | |
647 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
648 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
649 | set_page_dirty(page); | |
650 | f2fs_put_page(page, 1); | |
651 | index++; | |
127e670a JK |
652 | nentries = 0; |
653 | page = NULL; | |
654 | } | |
502c6e0b | 655 | } |
127e670a | 656 | |
502c6e0b GZ |
657 | if (page) { |
658 | orphan_blk->blk_addr = cpu_to_le16(index); | |
659 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
660 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
661 | set_page_dirty(page); | |
662 | f2fs_put_page(page, 1); | |
127e670a | 663 | } |
127e670a JK |
664 | } |
665 | ||
fc0065ad TY |
666 | static int get_checkpoint_version(struct f2fs_sb_info *sbi, block_t cp_addr, |
667 | struct f2fs_checkpoint **cp_block, struct page **cp_page, | |
668 | unsigned long long *version) | |
127e670a | 669 | { |
127e670a | 670 | unsigned long blk_size = sbi->blocksize; |
fc0065ad | 671 | size_t crc_offset = 0; |
7e586fa0 | 672 | __u32 crc = 0; |
127e670a | 673 | |
fc0065ad TY |
674 | *cp_page = get_meta_page(sbi, cp_addr); |
675 | *cp_block = (struct f2fs_checkpoint *)page_address(*cp_page); | |
127e670a | 676 | |
fc0065ad TY |
677 | crc_offset = le32_to_cpu((*cp_block)->checksum_offset); |
678 | if (crc_offset >= blk_size) { | |
679 | f2fs_msg(sbi->sb, KERN_WARNING, | |
680 | "invalid crc_offset: %zu", crc_offset); | |
681 | return -EINVAL; | |
682 | } | |
127e670a | 683 | |
fc0065ad TY |
684 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)*cp_block |
685 | + crc_offset))); | |
686 | if (!f2fs_crc_valid(sbi, crc, *cp_block, crc_offset)) { | |
687 | f2fs_msg(sbi->sb, KERN_WARNING, "invalid crc value"); | |
688 | return -EINVAL; | |
689 | } | |
127e670a | 690 | |
fc0065ad TY |
691 | *version = cur_cp_version(*cp_block); |
692 | return 0; | |
693 | } | |
127e670a | 694 | |
fc0065ad TY |
695 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, |
696 | block_t cp_addr, unsigned long long *version) | |
697 | { | |
698 | struct page *cp_page_1 = NULL, *cp_page_2 = NULL; | |
699 | struct f2fs_checkpoint *cp_block = NULL; | |
700 | unsigned long long cur_version = 0, pre_version = 0; | |
701 | int err; | |
127e670a | 702 | |
fc0065ad TY |
703 | err = get_checkpoint_version(sbi, cp_addr, &cp_block, |
704 | &cp_page_1, version); | |
705 | if (err) | |
706 | goto invalid_cp1; | |
707 | pre_version = *version; | |
127e670a | 708 | |
fc0065ad TY |
709 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
710 | err = get_checkpoint_version(sbi, cp_addr, &cp_block, | |
711 | &cp_page_2, version); | |
712 | if (err) | |
127e670a | 713 | goto invalid_cp2; |
fc0065ad | 714 | cur_version = *version; |
127e670a JK |
715 | |
716 | if (cur_version == pre_version) { | |
717 | *version = cur_version; | |
718 | f2fs_put_page(cp_page_2, 1); | |
719 | return cp_page_1; | |
720 | } | |
721 | invalid_cp2: | |
722 | f2fs_put_page(cp_page_2, 1); | |
723 | invalid_cp1: | |
724 | f2fs_put_page(cp_page_1, 1); | |
725 | return NULL; | |
726 | } | |
727 | ||
728 | int get_valid_checkpoint(struct f2fs_sb_info *sbi) | |
729 | { | |
730 | struct f2fs_checkpoint *cp_block; | |
731 | struct f2fs_super_block *fsb = sbi->raw_super; | |
732 | struct page *cp1, *cp2, *cur_page; | |
733 | unsigned long blk_size = sbi->blocksize; | |
734 | unsigned long long cp1_version = 0, cp2_version = 0; | |
735 | unsigned long long cp_start_blk_no; | |
55141486 | 736 | unsigned int cp_blks = 1 + __cp_payload(sbi); |
1dbe4152 CL |
737 | block_t cp_blk_no; |
738 | int i; | |
127e670a | 739 | |
1dbe4152 | 740 | sbi->ckpt = kzalloc(cp_blks * blk_size, GFP_KERNEL); |
127e670a JK |
741 | if (!sbi->ckpt) |
742 | return -ENOMEM; | |
743 | /* | |
744 | * Finding out valid cp block involves read both | |
745 | * sets( cp pack1 and cp pack 2) | |
746 | */ | |
747 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
748 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
749 | ||
750 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
751 | cp_start_blk_no += ((unsigned long long)1) << |
752 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
753 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
754 | ||
755 | if (cp1 && cp2) { | |
756 | if (ver_after(cp2_version, cp1_version)) | |
757 | cur_page = cp2; | |
758 | else | |
759 | cur_page = cp1; | |
760 | } else if (cp1) { | |
761 | cur_page = cp1; | |
762 | } else if (cp2) { | |
763 | cur_page = cp2; | |
764 | } else { | |
765 | goto fail_no_cp; | |
766 | } | |
767 | ||
768 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
769 | memcpy(sbi->ckpt, cp_block, blk_size); | |
770 | ||
984ec63c SL |
771 | /* Sanity checking of checkpoint */ |
772 | if (sanity_check_ckpt(sbi)) | |
773 | goto fail_no_cp; | |
774 | ||
1dbe4152 CL |
775 | if (cp_blks <= 1) |
776 | goto done; | |
777 | ||
778 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
779 | if (cur_page == cp2) | |
780 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
781 | ||
782 | for (i = 1; i < cp_blks; i++) { | |
783 | void *sit_bitmap_ptr; | |
784 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
785 | ||
786 | cur_page = get_meta_page(sbi, cp_blk_no + i); | |
787 | sit_bitmap_ptr = page_address(cur_page); | |
788 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
789 | f2fs_put_page(cur_page, 1); | |
790 | } | |
791 | done: | |
127e670a JK |
792 | f2fs_put_page(cp1, 1); |
793 | f2fs_put_page(cp2, 1); | |
794 | return 0; | |
795 | ||
796 | fail_no_cp: | |
797 | kfree(sbi->ckpt); | |
798 | return -EINVAL; | |
799 | } | |
800 | ||
c227f912 | 801 | static void __add_dirty_inode(struct inode *inode, enum inode_type type) |
127e670a | 802 | { |
4081363f | 803 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 804 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
127e670a | 805 | |
91942321 | 806 | if (is_inode_flag_set(inode, flag)) |
2710fd7e | 807 | return; |
2d7b822a | 808 | |
91942321 JK |
809 | set_inode_flag(inode, flag); |
810 | list_add_tail(&F2FS_I(inode)->dirty_list, &sbi->inode_list[type]); | |
33fbd510 | 811 | stat_inc_dirty_inode(sbi, type); |
5deb8267 JK |
812 | } |
813 | ||
c227f912 | 814 | static void __remove_dirty_inode(struct inode *inode, enum inode_type type) |
6ad7609a | 815 | { |
c227f912 | 816 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
6ad7609a | 817 | |
91942321 | 818 | if (get_dirty_pages(inode) || !is_inode_flag_set(inode, flag)) |
6ad7609a CY |
819 | return; |
820 | ||
91942321 JK |
821 | list_del_init(&F2FS_I(inode)->dirty_list); |
822 | clear_inode_flag(inode, flag); | |
33fbd510 | 823 | stat_dec_dirty_inode(F2FS_I_SB(inode), type); |
6ad7609a CY |
824 | } |
825 | ||
a7ffdbe2 | 826 | void update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 827 | { |
4081363f | 828 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 829 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
5deb8267 | 830 | |
5ac9f36f CY |
831 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
832 | !S_ISLNK(inode->i_mode)) | |
127e670a | 833 | return; |
7bd59381 | 834 | |
1c4bf763 JK |
835 | spin_lock(&sbi->inode_lock[type]); |
836 | if (type != FILE_INODE || test_opt(sbi, DATA_FLUSH)) | |
10aa97c3 | 837 | __add_dirty_inode(inode, type); |
b951a4ec | 838 | inode_inc_dirty_pages(inode); |
1c4bf763 JK |
839 | spin_unlock(&sbi->inode_lock[type]); |
840 | ||
a7ffdbe2 | 841 | SetPagePrivate(page); |
9e4ded3f | 842 | f2fs_trace_pid(page); |
5deb8267 JK |
843 | } |
844 | ||
c227f912 | 845 | void remove_dirty_inode(struct inode *inode) |
127e670a | 846 | { |
4081363f | 847 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 848 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
127e670a | 849 | |
c227f912 CY |
850 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
851 | !S_ISLNK(inode->i_mode)) | |
127e670a JK |
852 | return; |
853 | ||
10aa97c3 JK |
854 | if (type == FILE_INODE && !test_opt(sbi, DATA_FLUSH)) |
855 | return; | |
856 | ||
c227f912 CY |
857 | spin_lock(&sbi->inode_lock[type]); |
858 | __remove_dirty_inode(inode, type); | |
859 | spin_unlock(&sbi->inode_lock[type]); | |
74d0b917 JK |
860 | } |
861 | ||
6d5a1495 | 862 | int sync_dirty_inodes(struct f2fs_sb_info *sbi, enum inode_type type) |
127e670a | 863 | { |
ce3b7d80 | 864 | struct list_head *head; |
127e670a | 865 | struct inode *inode; |
2710fd7e | 866 | struct f2fs_inode_info *fi; |
4cf18537 CY |
867 | bool is_dir = (type == DIR_INODE); |
868 | ||
869 | trace_f2fs_sync_dirty_inodes_enter(sbi->sb, is_dir, | |
870 | get_pages(sbi, is_dir ? | |
871 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
127e670a | 872 | retry: |
af41d3ee | 873 | if (unlikely(f2fs_cp_error(sbi))) |
6d5a1495 | 874 | return -EIO; |
af41d3ee | 875 | |
c227f912 | 876 | spin_lock(&sbi->inode_lock[type]); |
ce3b7d80 | 877 | |
c227f912 | 878 | head = &sbi->inode_list[type]; |
127e670a | 879 | if (list_empty(head)) { |
c227f912 | 880 | spin_unlock(&sbi->inode_lock[type]); |
4cf18537 CY |
881 | trace_f2fs_sync_dirty_inodes_exit(sbi->sb, is_dir, |
882 | get_pages(sbi, is_dir ? | |
883 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
6d5a1495 | 884 | return 0; |
127e670a | 885 | } |
2710fd7e CY |
886 | fi = list_entry(head->next, struct f2fs_inode_info, dirty_list); |
887 | inode = igrab(&fi->vfs_inode); | |
c227f912 | 888 | spin_unlock(&sbi->inode_lock[type]); |
127e670a | 889 | if (inode) { |
87d6f890 | 890 | filemap_fdatawrite(inode->i_mapping); |
127e670a JK |
891 | iput(inode); |
892 | } else { | |
893 | /* | |
894 | * We should submit bio, since it exists several | |
895 | * wribacking dentry pages in the freeing inode. | |
896 | */ | |
458e6197 | 897 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
7ecebe5e | 898 | cond_resched(); |
127e670a JK |
899 | } |
900 | goto retry; | |
901 | } | |
902 | ||
0f18b462 JK |
903 | int f2fs_sync_inode_meta(struct f2fs_sb_info *sbi) |
904 | { | |
905 | struct list_head *head = &sbi->inode_list[DIRTY_META]; | |
906 | struct inode *inode; | |
907 | struct f2fs_inode_info *fi; | |
908 | s64 total = get_pages(sbi, F2FS_DIRTY_IMETA); | |
909 | ||
910 | while (total--) { | |
911 | if (unlikely(f2fs_cp_error(sbi))) | |
912 | return -EIO; | |
913 | ||
914 | spin_lock(&sbi->inode_lock[DIRTY_META]); | |
915 | if (list_empty(head)) { | |
916 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
917 | return 0; | |
918 | } | |
919 | fi = list_entry(head->next, struct f2fs_inode_info, | |
920 | gdirty_list); | |
921 | inode = igrab(&fi->vfs_inode); | |
922 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
923 | if (inode) { | |
924 | update_inode_page(inode); | |
925 | iput(inode); | |
926 | } | |
927 | }; | |
928 | return 0; | |
929 | } | |
930 | ||
0a8165d7 | 931 | /* |
127e670a JK |
932 | * Freeze all the FS-operations for checkpoint. |
933 | */ | |
cf779cab | 934 | static int block_operations(struct f2fs_sb_info *sbi) |
127e670a | 935 | { |
127e670a JK |
936 | struct writeback_control wbc = { |
937 | .sync_mode = WB_SYNC_ALL, | |
938 | .nr_to_write = LONG_MAX, | |
939 | .for_reclaim = 0, | |
940 | }; | |
c718379b | 941 | struct blk_plug plug; |
cf779cab | 942 | int err = 0; |
c718379b JK |
943 | |
944 | blk_start_plug(&plug); | |
945 | ||
39936837 | 946 | retry_flush_dents: |
e479556b | 947 | f2fs_lock_all(sbi); |
127e670a | 948 | /* write all the dirty dentry pages */ |
127e670a | 949 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 950 | f2fs_unlock_all(sbi); |
6d5a1495 CY |
951 | err = sync_dirty_inodes(sbi, DIR_INODE); |
952 | if (err) | |
cf779cab | 953 | goto out; |
39936837 | 954 | goto retry_flush_dents; |
127e670a JK |
955 | } |
956 | ||
0f18b462 JK |
957 | if (get_pages(sbi, F2FS_DIRTY_IMETA)) { |
958 | f2fs_unlock_all(sbi); | |
959 | err = f2fs_sync_inode_meta(sbi); | |
960 | if (err) | |
961 | goto out; | |
962 | goto retry_flush_dents; | |
963 | } | |
964 | ||
127e670a | 965 | /* |
e1c42045 | 966 | * POR: we should ensure that there are no dirty node pages |
127e670a JK |
967 | * until finishing nat/sit flush. |
968 | */ | |
39936837 | 969 | retry_flush_nodes: |
b3582c68 | 970 | down_write(&sbi->node_write); |
127e670a JK |
971 | |
972 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
b3582c68 | 973 | up_write(&sbi->node_write); |
52681375 | 974 | err = sync_node_pages(sbi, &wbc); |
6d5a1495 | 975 | if (err) { |
cf779cab | 976 | f2fs_unlock_all(sbi); |
cf779cab JK |
977 | goto out; |
978 | } | |
39936837 | 979 | goto retry_flush_nodes; |
127e670a | 980 | } |
cf779cab | 981 | out: |
c718379b | 982 | blk_finish_plug(&plug); |
cf779cab | 983 | return err; |
127e670a JK |
984 | } |
985 | ||
986 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
987 | { | |
b3582c68 | 988 | up_write(&sbi->node_write); |
ad4edb83 JK |
989 | |
990 | build_free_nids(sbi); | |
e479556b | 991 | f2fs_unlock_all(sbi); |
127e670a JK |
992 | } |
993 | ||
fb51b5ef CL |
994 | static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi) |
995 | { | |
996 | DEFINE_WAIT(wait); | |
997 | ||
998 | for (;;) { | |
999 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
1000 | ||
f5730184 | 1001 | if (!atomic_read(&sbi->nr_wb_bios)) |
fb51b5ef CL |
1002 | break; |
1003 | ||
0ff21646 | 1004 | io_schedule_timeout(5*HZ); |
fb51b5ef CL |
1005 | } |
1006 | finish_wait(&sbi->cp_wait, &wait); | |
1007 | } | |
1008 | ||
c34f42e2 | 1009 | static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1010 | { |
1011 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
77041823 | 1012 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
67298804 | 1013 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; |
77041823 | 1014 | nid_t last_nid = nm_i->next_scan_nid; |
127e670a | 1015 | block_t start_blk; |
127e670a | 1016 | unsigned int data_sum_blocks, orphan_blocks; |
7e586fa0 | 1017 | __u32 crc32 = 0; |
127e670a | 1018 | int i; |
55141486 | 1019 | int cp_payload_blks = __cp_payload(sbi); |
8f1dbbbb SL |
1020 | struct super_block *sb = sbi->sb; |
1021 | struct curseg_info *seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); | |
1022 | u64 kbytes_written; | |
127e670a JK |
1023 | |
1024 | /* Flush all the NAT/SIT pages */ | |
cf779cab | 1025 | while (get_pages(sbi, F2FS_DIRTY_META)) { |
127e670a | 1026 | sync_meta_pages(sbi, META, LONG_MAX); |
cf779cab | 1027 | if (unlikely(f2fs_cp_error(sbi))) |
c34f42e2 | 1028 | return -EIO; |
cf779cab | 1029 | } |
127e670a JK |
1030 | |
1031 | next_free_nid(sbi, &last_nid); | |
1032 | ||
1033 | /* | |
1034 | * modify checkpoint | |
1035 | * version number is already updated | |
1036 | */ | |
1037 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi)); | |
1038 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
1039 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); | |
b5b82205 | 1040 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
1041 | ckpt->cur_node_segno[i] = |
1042 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
1043 | ckpt->cur_node_blkoff[i] = | |
1044 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
1045 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
1046 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
1047 | } | |
b5b82205 | 1048 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
1049 | ckpt->cur_data_segno[i] = |
1050 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
1051 | ckpt->cur_data_blkoff[i] = | |
1052 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
1053 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
1054 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
1055 | } | |
1056 | ||
1057 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
1058 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
1059 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
1060 | ||
1061 | /* 2 cp + n data seg summary + orphan inode blocks */ | |
3fa06d7b | 1062 | data_sum_blocks = npages_for_summary_flush(sbi, false); |
aaec2b1d | 1063 | spin_lock(&sbi->cp_lock); |
b5b82205 | 1064 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
aaec2b1d | 1065 | __set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 1066 | else |
aaec2b1d CY |
1067 | __clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
1068 | spin_unlock(&sbi->cp_lock); | |
127e670a | 1069 | |
67298804 | 1070 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
1071 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
1072 | orphan_blocks); | |
127e670a | 1073 | |
119ee914 | 1074 | if (__remain_node_summaries(cpc->reason)) |
b5b82205 | 1075 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
1076 | cp_payload_blks + data_sum_blocks + |
1077 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
119ee914 | 1078 | else |
b5b82205 | 1079 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS + |
1dbe4152 CL |
1080 | cp_payload_blks + data_sum_blocks + |
1081 | orphan_blocks); | |
119ee914 | 1082 | |
aaec2b1d | 1083 | spin_lock(&sbi->cp_lock); |
119ee914 | 1084 | if (cpc->reason == CP_UMOUNT) |
aaec2b1d | 1085 | __set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); |
119ee914 | 1086 | else |
aaec2b1d | 1087 | __clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); |
119ee914 JK |
1088 | |
1089 | if (cpc->reason == CP_FASTBOOT) | |
aaec2b1d | 1090 | __set_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); |
119ee914 | 1091 | else |
aaec2b1d | 1092 | __clear_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); |
127e670a | 1093 | |
67298804 | 1094 | if (orphan_num) |
aaec2b1d | 1095 | __set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1096 | else |
aaec2b1d | 1097 | __clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1098 | |
caf0047e | 1099 | if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) |
aaec2b1d | 1100 | __set_ckpt_flags(ckpt, CP_FSCK_FLAG); |
2ae4c673 | 1101 | |
a468f0ef | 1102 | /* set this flag to activate crc|cp_ver for recovery */ |
aaec2b1d CY |
1103 | __set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG); |
1104 | ||
1105 | spin_unlock(&sbi->cp_lock); | |
a468f0ef | 1106 | |
127e670a JK |
1107 | /* update SIT/NAT bitmap */ |
1108 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
1109 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
1110 | ||
43b6573b | 1111 | crc32 = f2fs_crc32(sbi, ckpt, le32_to_cpu(ckpt->checksum_offset)); |
7e586fa0 JK |
1112 | *((__le32 *)((unsigned char *)ckpt + |
1113 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
1114 | = cpu_to_le32(crc32); |
1115 | ||
1116 | start_blk = __start_cp_addr(sbi); | |
1117 | ||
a7230d16 JK |
1118 | /* need to wait for end_io results */ |
1119 | wait_on_all_pages_writeback(sbi); | |
1120 | if (unlikely(f2fs_cp_error(sbi))) | |
c34f42e2 | 1121 | return -EIO; |
a7230d16 | 1122 | |
127e670a | 1123 | /* write out checkpoint buffer at block 0 */ |
381722d2 CY |
1124 | update_meta_page(sbi, ckpt, start_blk++); |
1125 | ||
1126 | for (i = 1; i < 1 + cp_payload_blks; i++) | |
1127 | update_meta_page(sbi, (char *)ckpt + i * F2FS_BLKSIZE, | |
1128 | start_blk++); | |
1dbe4152 | 1129 | |
67298804 | 1130 | if (orphan_num) { |
127e670a JK |
1131 | write_orphan_inodes(sbi, start_blk); |
1132 | start_blk += orphan_blocks; | |
1133 | } | |
1134 | ||
1135 | write_data_summaries(sbi, start_blk); | |
1136 | start_blk += data_sum_blocks; | |
8f1dbbbb SL |
1137 | |
1138 | /* Record write statistics in the hot node summary */ | |
1139 | kbytes_written = sbi->kbytes_written; | |
1140 | if (sb->s_bdev->bd_part) | |
1141 | kbytes_written += BD_PART_WRITTEN(sbi); | |
1142 | ||
b7ad7512 | 1143 | seg_i->journal->info.kbytes_written = cpu_to_le64(kbytes_written); |
8f1dbbbb | 1144 | |
119ee914 | 1145 | if (__remain_node_summaries(cpc->reason)) { |
127e670a JK |
1146 | write_node_summaries(sbi, start_blk); |
1147 | start_blk += NR_CURSEG_NODE_TYPE; | |
1148 | } | |
1149 | ||
1150 | /* writeout checkpoint block */ | |
381722d2 | 1151 | update_meta_page(sbi, ckpt, start_blk); |
127e670a JK |
1152 | |
1153 | /* wait for previous submitted node/meta pages writeback */ | |
fb51b5ef | 1154 | wait_on_all_pages_writeback(sbi); |
127e670a | 1155 | |
cf779cab | 1156 | if (unlikely(f2fs_cp_error(sbi))) |
c34f42e2 | 1157 | return -EIO; |
cf779cab | 1158 | |
80dd9c0e CY |
1159 | filemap_fdatawait_range(NODE_MAPPING(sbi), 0, LLONG_MAX); |
1160 | filemap_fdatawait_range(META_MAPPING(sbi), 0, LLONG_MAX); | |
127e670a JK |
1161 | |
1162 | /* update user_block_counts */ | |
1163 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
41382ec4 | 1164 | percpu_counter_set(&sbi->alloc_valid_block_count, 0); |
127e670a JK |
1165 | |
1166 | /* Here, we only have one bio having CP pack */ | |
577e3495 | 1167 | sync_meta_pages(sbi, META_FLUSH, LONG_MAX); |
127e670a | 1168 | |
6a8f8ca5 JK |
1169 | /* wait for previous submitted meta pages writeback */ |
1170 | wait_on_all_pages_writeback(sbi); | |
1171 | ||
74ef9241 | 1172 | release_ino_entry(sbi, false); |
cf779cab JK |
1173 | |
1174 | if (unlikely(f2fs_cp_error(sbi))) | |
c34f42e2 | 1175 | return -EIO; |
cf779cab | 1176 | |
836b5a63 | 1177 | clear_prefree_segments(sbi, cpc); |
caf0047e | 1178 | clear_sbi_flag(sbi, SBI_IS_DIRTY); |
bbf156f7 | 1179 | clear_sbi_flag(sbi, SBI_NEED_CP); |
c34f42e2 | 1180 | |
c2a080ae CY |
1181 | /* |
1182 | * redirty superblock if metadata like node page or inode cache is | |
1183 | * updated during writing checkpoint. | |
1184 | */ | |
1185 | if (get_pages(sbi, F2FS_DIRTY_NODES) || | |
1186 | get_pages(sbi, F2FS_DIRTY_IMETA)) | |
1187 | set_sbi_flag(sbi, SBI_IS_DIRTY); | |
1188 | ||
1189 | f2fs_bug_on(sbi, get_pages(sbi, F2FS_DIRTY_DENTS)); | |
1190 | ||
c34f42e2 | 1191 | return 0; |
127e670a JK |
1192 | } |
1193 | ||
0a8165d7 | 1194 | /* |
e1c42045 | 1195 | * We guarantee that this checkpoint procedure will not fail. |
127e670a | 1196 | */ |
c34f42e2 | 1197 | int write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1198 | { |
1199 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1200 | unsigned long long ckpt_ver; | |
c34f42e2 | 1201 | int err = 0; |
127e670a | 1202 | |
43727527 | 1203 | mutex_lock(&sbi->cp_mutex); |
8501017e | 1204 | |
caf0047e | 1205 | if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) && |
a66cdd98 JK |
1206 | (cpc->reason == CP_FASTBOOT || cpc->reason == CP_SYNC || |
1207 | (cpc->reason == CP_DISCARD && !sbi->discard_blks))) | |
8501017e | 1208 | goto out; |
c34f42e2 CY |
1209 | if (unlikely(f2fs_cp_error(sbi))) { |
1210 | err = -EIO; | |
cf779cab | 1211 | goto out; |
c34f42e2 CY |
1212 | } |
1213 | if (f2fs_readonly(sbi->sb)) { | |
1214 | err = -EROFS; | |
11504a8e | 1215 | goto out; |
c34f42e2 | 1216 | } |
2bda542d WL |
1217 | |
1218 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); | |
1219 | ||
c34f42e2 CY |
1220 | err = block_operations(sbi); |
1221 | if (err) | |
cf779cab | 1222 | goto out; |
127e670a | 1223 | |
75ab4cb8 | 1224 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1225 | |
406657dd | 1226 | f2fs_flush_merged_bios(sbi); |
127e670a | 1227 | |
58cce381 YH |
1228 | /* this is the case of multiple fstrims without any changes */ |
1229 | if (cpc->reason == CP_DISCARD && !is_sbi_flag_set(sbi, SBI_IS_DIRTY)) { | |
1230 | f2fs_bug_on(sbi, NM_I(sbi)->dirty_nat_cnt); | |
1231 | f2fs_bug_on(sbi, SIT_I(sbi)->dirty_sentries); | |
1232 | f2fs_bug_on(sbi, prefree_segments(sbi)); | |
1233 | flush_sit_entries(sbi, cpc); | |
1234 | clear_prefree_segments(sbi, cpc); | |
275b66b0 | 1235 | f2fs_wait_all_discard_bio(sbi); |
58cce381 YH |
1236 | unblock_operations(sbi); |
1237 | goto out; | |
1238 | } | |
1239 | ||
127e670a JK |
1240 | /* |
1241 | * update checkpoint pack index | |
1242 | * Increase the version number so that | |
1243 | * SIT entries and seg summaries are written at correct place | |
1244 | */ | |
d71b5564 | 1245 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1246 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1247 | ||
1248 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
1249 | flush_nat_entries(sbi); | |
4b2fecc8 | 1250 | flush_sit_entries(sbi, cpc); |
127e670a | 1251 | |
127e670a | 1252 | /* unlock all the fs_lock[] in do_checkpoint() */ |
c34f42e2 | 1253 | err = do_checkpoint(sbi, cpc); |
127e670a | 1254 | |
275b66b0 CY |
1255 | f2fs_wait_all_discard_bio(sbi); |
1256 | ||
127e670a | 1257 | unblock_operations(sbi); |
942e0be6 | 1258 | stat_inc_cp_count(sbi->stat_info); |
10027551 JK |
1259 | |
1260 | if (cpc->reason == CP_RECOVERY) | |
1261 | f2fs_msg(sbi->sb, KERN_NOTICE, | |
1262 | "checkpoint: version = %llx", ckpt_ver); | |
60b99b48 JK |
1263 | |
1264 | /* do checkpoint periodically */ | |
6beceb54 | 1265 | f2fs_update_time(sbi, CP_TIME); |
55d1cdb2 | 1266 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
8501017e JK |
1267 | out: |
1268 | mutex_unlock(&sbi->cp_mutex); | |
c34f42e2 | 1269 | return err; |
127e670a JK |
1270 | } |
1271 | ||
6451e041 | 1272 | void init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1273 | { |
6451e041 JK |
1274 | int i; |
1275 | ||
1276 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1277 | struct inode_management *im = &sbi->im[i]; |
1278 | ||
1279 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1280 | spin_lock_init(&im->ino_lock); | |
1281 | INIT_LIST_HEAD(&im->ino_list); | |
1282 | im->ino_num = 0; | |
6451e041 JK |
1283 | } |
1284 | ||
b5b82205 | 1285 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
14b42817 WL |
1286 | NR_CURSEG_TYPE - __cp_payload(sbi)) * |
1287 | F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
1288 | } |
1289 | ||
6e6093a8 | 1290 | int __init create_checkpoint_caches(void) |
127e670a | 1291 | { |
6451e041 JK |
1292 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1293 | sizeof(struct ino_entry)); | |
1294 | if (!ino_entry_slab) | |
127e670a | 1295 | return -ENOMEM; |
06292073 CY |
1296 | inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry", |
1297 | sizeof(struct inode_entry)); | |
6bacf52f | 1298 | if (!inode_entry_slab) { |
6451e041 | 1299 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1300 | return -ENOMEM; |
1301 | } | |
1302 | return 0; | |
1303 | } | |
1304 | ||
1305 | void destroy_checkpoint_caches(void) | |
1306 | { | |
6451e041 | 1307 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1308 | kmem_cache_destroy(inode_entry_slab); |
1309 | } |