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 | |
0a8165d7 | 29 | /* |
127e670a JK |
30 | * We guarantee no failure on the returned page. |
31 | */ | |
32 | struct page *grab_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
33 | { | |
9df27d98 | 34 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
35 | struct page *page = NULL; |
36 | repeat: | |
300e129c | 37 | page = f2fs_grab_cache_page(mapping, index, false); |
127e670a JK |
38 | if (!page) { |
39 | cond_resched(); | |
40 | goto repeat; | |
41 | } | |
fec1d657 | 42 | f2fs_wait_on_page_writeback(page, META, true); |
127e670a JK |
43 | SetPageUptodate(page); |
44 | return page; | |
45 | } | |
46 | ||
0a8165d7 | 47 | /* |
127e670a JK |
48 | * We guarantee no failure on the returned page. |
49 | */ | |
2b947003 CY |
50 | static struct page *__get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index, |
51 | bool is_meta) | |
127e670a | 52 | { |
9df27d98 | 53 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a | 54 | struct page *page; |
cf04e8eb | 55 | struct f2fs_io_info fio = { |
05ca3632 | 56 | .sbi = sbi, |
cf04e8eb JK |
57 | .type = META, |
58 | .rw = READ_SYNC | REQ_META | REQ_PRIO, | |
7a9d7548 CY |
59 | .old_blkaddr = index, |
60 | .new_blkaddr = index, | |
4375a336 | 61 | .encrypted_page = NULL, |
cf04e8eb | 62 | }; |
2b947003 CY |
63 | |
64 | if (unlikely(!is_meta)) | |
65 | fio.rw &= ~REQ_META; | |
127e670a | 66 | repeat: |
300e129c | 67 | page = f2fs_grab_cache_page(mapping, index, false); |
127e670a JK |
68 | if (!page) { |
69 | cond_resched(); | |
70 | goto repeat; | |
71 | } | |
393ff91f JK |
72 | if (PageUptodate(page)) |
73 | goto out; | |
74 | ||
05ca3632 JK |
75 | fio.page = page; |
76 | ||
86531d6b JK |
77 | if (f2fs_submit_page_bio(&fio)) { |
78 | f2fs_put_page(page, 1); | |
127e670a | 79 | goto repeat; |
86531d6b | 80 | } |
127e670a | 81 | |
393ff91f | 82 | lock_page(page); |
6bacf52f | 83 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
84 | f2fs_put_page(page, 1); |
85 | goto repeat; | |
86 | } | |
f3f338ca CY |
87 | |
88 | /* | |
89 | * if there is any IO error when accessing device, make our filesystem | |
90 | * readonly and make sure do not write checkpoint with non-uptodate | |
91 | * meta page. | |
92 | */ | |
93 | if (unlikely(!PageUptodate(page))) | |
94 | f2fs_stop_checkpoint(sbi); | |
393ff91f | 95 | out: |
127e670a JK |
96 | return page; |
97 | } | |
98 | ||
2b947003 CY |
99 | struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) |
100 | { | |
101 | return __get_meta_page(sbi, index, true); | |
102 | } | |
103 | ||
104 | /* for POR only */ | |
105 | struct page *get_tmp_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
106 | { | |
107 | return __get_meta_page(sbi, index, false); | |
108 | } | |
109 | ||
f0c9cada | 110 | bool is_valid_blkaddr(struct f2fs_sb_info *sbi, block_t blkaddr, int type) |
662befda CY |
111 | { |
112 | switch (type) { | |
113 | case META_NAT: | |
66b00c18 | 114 | break; |
662befda | 115 | case META_SIT: |
66b00c18 CY |
116 | if (unlikely(blkaddr >= SIT_BLK_CNT(sbi))) |
117 | return false; | |
118 | break; | |
81c1a0f1 | 119 | case META_SSA: |
66b00c18 CY |
120 | if (unlikely(blkaddr >= MAIN_BLKADDR(sbi) || |
121 | blkaddr < SM_I(sbi)->ssa_blkaddr)) | |
122 | return false; | |
123 | break; | |
662befda | 124 | case META_CP: |
66b00c18 CY |
125 | if (unlikely(blkaddr >= SIT_I(sbi)->sit_base_addr || |
126 | blkaddr < __start_cp_addr(sbi))) | |
127 | return false; | |
128 | break; | |
4c521f49 | 129 | case META_POR: |
66b00c18 CY |
130 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
131 | blkaddr < MAIN_BLKADDR(sbi))) | |
132 | return false; | |
133 | break; | |
662befda CY |
134 | default: |
135 | BUG(); | |
136 | } | |
66b00c18 CY |
137 | |
138 | return true; | |
662befda CY |
139 | } |
140 | ||
141 | /* | |
81c1a0f1 | 142 | * Readahead CP/NAT/SIT/SSA pages |
662befda | 143 | */ |
26879fb1 CY |
144 | int ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages, |
145 | int type, bool sync) | |
662befda | 146 | { |
662befda | 147 | struct page *page; |
4c521f49 | 148 | block_t blkno = start; |
662befda | 149 | struct f2fs_io_info fio = { |
05ca3632 | 150 | .sbi = sbi, |
662befda | 151 | .type = META, |
26879fb1 | 152 | .rw = sync ? (READ_SYNC | REQ_META | REQ_PRIO) : READA, |
4375a336 | 153 | .encrypted_page = NULL, |
662befda | 154 | }; |
e9f5b8b8 | 155 | struct blk_plug plug; |
662befda | 156 | |
2b947003 CY |
157 | if (unlikely(type == META_POR)) |
158 | fio.rw &= ~REQ_META; | |
159 | ||
e9f5b8b8 | 160 | blk_start_plug(&plug); |
662befda | 161 | for (; nrpages-- > 0; blkno++) { |
662befda | 162 | |
66b00c18 CY |
163 | if (!is_valid_blkaddr(sbi, blkno, type)) |
164 | goto out; | |
165 | ||
662befda CY |
166 | switch (type) { |
167 | case META_NAT: | |
66b00c18 CY |
168 | if (unlikely(blkno >= |
169 | NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))) | |
662befda | 170 | blkno = 0; |
66b00c18 | 171 | /* get nat block addr */ |
7a9d7548 | 172 | fio.new_blkaddr = current_nat_addr(sbi, |
662befda CY |
173 | blkno * NAT_ENTRY_PER_BLOCK); |
174 | break; | |
175 | case META_SIT: | |
176 | /* get sit block addr */ | |
7a9d7548 | 177 | fio.new_blkaddr = current_sit_addr(sbi, |
662befda | 178 | blkno * SIT_ENTRY_PER_BLOCK); |
662befda | 179 | break; |
81c1a0f1 | 180 | case META_SSA: |
662befda | 181 | case META_CP: |
4c521f49 | 182 | case META_POR: |
7a9d7548 | 183 | fio.new_blkaddr = blkno; |
662befda CY |
184 | break; |
185 | default: | |
186 | BUG(); | |
187 | } | |
188 | ||
300e129c JK |
189 | page = f2fs_grab_cache_page(META_MAPPING(sbi), |
190 | fio.new_blkaddr, false); | |
662befda CY |
191 | if (!page) |
192 | continue; | |
193 | if (PageUptodate(page)) { | |
662befda CY |
194 | f2fs_put_page(page, 1); |
195 | continue; | |
196 | } | |
197 | ||
05ca3632 | 198 | fio.page = page; |
7a9d7548 | 199 | fio.old_blkaddr = fio.new_blkaddr; |
05ca3632 | 200 | f2fs_submit_page_mbio(&fio); |
662befda CY |
201 | f2fs_put_page(page, 0); |
202 | } | |
203 | out: | |
204 | f2fs_submit_merged_bio(sbi, META, READ); | |
e9f5b8b8 | 205 | blk_finish_plug(&plug); |
662befda CY |
206 | return blkno - start; |
207 | } | |
208 | ||
635aee1f CY |
209 | void ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index) |
210 | { | |
211 | struct page *page; | |
212 | bool readahead = false; | |
213 | ||
214 | page = find_get_page(META_MAPPING(sbi), index); | |
4da7bf5a | 215 | if (!page || !PageUptodate(page)) |
635aee1f CY |
216 | readahead = true; |
217 | f2fs_put_page(page, 0); | |
218 | ||
219 | if (readahead) | |
26879fb1 | 220 | ra_meta_pages(sbi, index, MAX_BIO_BLOCKS(sbi), META_POR, true); |
635aee1f CY |
221 | } |
222 | ||
127e670a JK |
223 | static int f2fs_write_meta_page(struct page *page, |
224 | struct writeback_control *wbc) | |
225 | { | |
4081363f | 226 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 227 | |
ecda0de3 CY |
228 | trace_f2fs_writepage(page, META); |
229 | ||
caf0047e | 230 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
cfb271d4 | 231 | goto redirty_out; |
857dc4e0 | 232 | if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0)) |
cfb271d4 | 233 | goto redirty_out; |
1e968fdf | 234 | if (unlikely(f2fs_cp_error(sbi))) |
cf779cab | 235 | goto redirty_out; |
127e670a | 236 | |
577e3495 JK |
237 | write_meta_page(sbi, page); |
238 | dec_page_count(sbi, F2FS_DIRTY_META); | |
0c3a5797 CY |
239 | |
240 | if (wbc->for_reclaim) | |
241 | f2fs_submit_merged_bio_cond(sbi, NULL, page, 0, META, WRITE); | |
242 | ||
577e3495 | 243 | unlock_page(page); |
857dc4e0 | 244 | |
0c3a5797 | 245 | if (unlikely(f2fs_cp_error(sbi))) |
857dc4e0 | 246 | f2fs_submit_merged_bio(sbi, META, WRITE); |
0c3a5797 | 247 | |
577e3495 | 248 | return 0; |
cfb271d4 CY |
249 | |
250 | redirty_out: | |
76f60268 | 251 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 252 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
253 | } |
254 | ||
255 | static int f2fs_write_meta_pages(struct address_space *mapping, | |
256 | struct writeback_control *wbc) | |
257 | { | |
4081363f | 258 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 259 | long diff, written; |
127e670a | 260 | |
5459aa97 | 261 | /* collect a number of dirty meta pages and write together */ |
50c8cdb3 JK |
262 | if (wbc->for_kupdate || |
263 | get_pages(sbi, F2FS_DIRTY_META) < nr_pages_to_skip(sbi, META)) | |
d3baf95d | 264 | goto skip_write; |
127e670a | 265 | |
d31c7c3f YH |
266 | trace_f2fs_writepages(mapping->host, wbc, META); |
267 | ||
127e670a JK |
268 | /* if mounting is failed, skip writing node pages */ |
269 | mutex_lock(&sbi->cp_mutex); | |
50c8cdb3 JK |
270 | diff = nr_pages_to_write(sbi, META, wbc); |
271 | written = sync_meta_pages(sbi, META, wbc->nr_to_write); | |
127e670a | 272 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 273 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 274 | return 0; |
d3baf95d JK |
275 | |
276 | skip_write: | |
277 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
d31c7c3f | 278 | trace_f2fs_writepages(mapping->host, wbc, META); |
d3baf95d | 279 | return 0; |
127e670a JK |
280 | } |
281 | ||
282 | long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, | |
283 | long nr_to_write) | |
284 | { | |
9df27d98 | 285 | struct address_space *mapping = META_MAPPING(sbi); |
80dd9c0e | 286 | pgoff_t index = 0, end = ULONG_MAX, prev = ULONG_MAX; |
127e670a JK |
287 | struct pagevec pvec; |
288 | long nwritten = 0; | |
289 | struct writeback_control wbc = { | |
290 | .for_reclaim = 0, | |
291 | }; | |
e9f5b8b8 | 292 | struct blk_plug plug; |
127e670a JK |
293 | |
294 | pagevec_init(&pvec, 0); | |
295 | ||
e9f5b8b8 CY |
296 | blk_start_plug(&plug); |
297 | ||
127e670a JK |
298 | while (index <= end) { |
299 | int i, nr_pages; | |
300 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, | |
301 | PAGECACHE_TAG_DIRTY, | |
302 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
cfb271d4 | 303 | if (unlikely(nr_pages == 0)) |
127e670a JK |
304 | break; |
305 | ||
306 | for (i = 0; i < nr_pages; i++) { | |
307 | struct page *page = pvec.pages[i]; | |
203681f6 | 308 | |
80dd9c0e | 309 | if (prev == ULONG_MAX) |
6066d8cd JK |
310 | prev = page->index - 1; |
311 | if (nr_to_write != LONG_MAX && page->index != prev + 1) { | |
312 | pagevec_release(&pvec); | |
313 | goto stop; | |
314 | } | |
315 | ||
127e670a | 316 | lock_page(page); |
203681f6 JK |
317 | |
318 | if (unlikely(page->mapping != mapping)) { | |
319 | continue_unlock: | |
320 | unlock_page(page); | |
321 | continue; | |
322 | } | |
323 | if (!PageDirty(page)) { | |
324 | /* someone wrote it for us */ | |
325 | goto continue_unlock; | |
326 | } | |
327 | ||
fa3d2bdf JK |
328 | f2fs_wait_on_page_writeback(page, META, true); |
329 | ||
330 | BUG_ON(PageWriteback(page)); | |
203681f6 JK |
331 | if (!clear_page_dirty_for_io(page)) |
332 | goto continue_unlock; | |
333 | ||
97dc3fd2 | 334 | if (mapping->a_ops->writepage(page, &wbc)) { |
577e3495 JK |
335 | unlock_page(page); |
336 | break; | |
337 | } | |
cfb271d4 | 338 | nwritten++; |
6066d8cd | 339 | prev = page->index; |
cfb271d4 | 340 | if (unlikely(nwritten >= nr_to_write)) |
127e670a JK |
341 | break; |
342 | } | |
343 | pagevec_release(&pvec); | |
344 | cond_resched(); | |
345 | } | |
6066d8cd | 346 | stop: |
127e670a | 347 | if (nwritten) |
458e6197 | 348 | f2fs_submit_merged_bio(sbi, type, WRITE); |
127e670a | 349 | |
e9f5b8b8 CY |
350 | blk_finish_plug(&plug); |
351 | ||
127e670a JK |
352 | return nwritten; |
353 | } | |
354 | ||
355 | static int f2fs_set_meta_page_dirty(struct page *page) | |
356 | { | |
26c6b887 JK |
357 | trace_f2fs_set_page_dirty(page, META); |
358 | ||
127e670a JK |
359 | SetPageUptodate(page); |
360 | if (!PageDirty(page)) { | |
361 | __set_page_dirty_nobuffers(page); | |
4081363f | 362 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); |
1601839e | 363 | SetPagePrivate(page); |
9e4ded3f | 364 | f2fs_trace_pid(page); |
127e670a JK |
365 | return 1; |
366 | } | |
367 | return 0; | |
368 | } | |
369 | ||
370 | const struct address_space_operations f2fs_meta_aops = { | |
371 | .writepage = f2fs_write_meta_page, | |
372 | .writepages = f2fs_write_meta_pages, | |
373 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
487261f3 CY |
374 | .invalidatepage = f2fs_invalidate_page, |
375 | .releasepage = f2fs_release_page, | |
127e670a JK |
376 | }; |
377 | ||
6451e041 | 378 | static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 379 | { |
67298804 | 380 | struct inode_management *im = &sbi->im[type]; |
80c54505 JK |
381 | struct ino_entry *e, *tmp; |
382 | ||
383 | tmp = f2fs_kmem_cache_alloc(ino_entry_slab, GFP_NOFS); | |
39efac41 | 384 | retry: |
80c54505 | 385 | radix_tree_preload(GFP_NOFS | __GFP_NOFAIL); |
769ec6e5 | 386 | |
67298804 | 387 | spin_lock(&im->ino_lock); |
67298804 | 388 | e = radix_tree_lookup(&im->ino_root, ino); |
39efac41 | 389 | if (!e) { |
80c54505 | 390 | e = tmp; |
67298804 CY |
391 | if (radix_tree_insert(&im->ino_root, ino, e)) { |
392 | spin_unlock(&im->ino_lock); | |
769ec6e5 | 393 | radix_tree_preload_end(); |
39efac41 JK |
394 | goto retry; |
395 | } | |
396 | memset(e, 0, sizeof(struct ino_entry)); | |
397 | e->ino = ino; | |
953e6cc6 | 398 | |
67298804 | 399 | list_add_tail(&e->list, &im->ino_list); |
8c402946 | 400 | if (type != ORPHAN_INO) |
67298804 | 401 | im->ino_num++; |
39efac41 | 402 | } |
67298804 | 403 | spin_unlock(&im->ino_lock); |
769ec6e5 | 404 | radix_tree_preload_end(); |
80c54505 JK |
405 | |
406 | if (e != tmp) | |
407 | kmem_cache_free(ino_entry_slab, tmp); | |
953e6cc6 JK |
408 | } |
409 | ||
6451e041 | 410 | static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 411 | { |
67298804 | 412 | struct inode_management *im = &sbi->im[type]; |
6451e041 | 413 | struct ino_entry *e; |
953e6cc6 | 414 | |
67298804 CY |
415 | spin_lock(&im->ino_lock); |
416 | e = radix_tree_lookup(&im->ino_root, ino); | |
39efac41 JK |
417 | if (e) { |
418 | list_del(&e->list); | |
67298804 CY |
419 | radix_tree_delete(&im->ino_root, ino); |
420 | im->ino_num--; | |
421 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
422 | kmem_cache_free(ino_entry_slab, e); |
423 | return; | |
953e6cc6 | 424 | } |
67298804 | 425 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
426 | } |
427 | ||
a49324f1 | 428 | void add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
429 | { |
430 | /* add new dirty ino entry into list */ | |
431 | __add_ino_entry(sbi, ino, type); | |
432 | } | |
433 | ||
a49324f1 | 434 | void remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
435 | { |
436 | /* remove dirty ino entry from list */ | |
437 | __remove_ino_entry(sbi, ino, type); | |
438 | } | |
439 | ||
440 | /* mode should be APPEND_INO or UPDATE_INO */ | |
441 | bool exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode) | |
442 | { | |
67298804 | 443 | struct inode_management *im = &sbi->im[mode]; |
fff04f90 | 444 | struct ino_entry *e; |
67298804 CY |
445 | |
446 | spin_lock(&im->ino_lock); | |
447 | e = radix_tree_lookup(&im->ino_root, ino); | |
448 | spin_unlock(&im->ino_lock); | |
fff04f90 JK |
449 | return e ? true : false; |
450 | } | |
451 | ||
a49324f1 | 452 | void release_ino_entry(struct f2fs_sb_info *sbi) |
fff04f90 JK |
453 | { |
454 | struct ino_entry *e, *tmp; | |
455 | int i; | |
456 | ||
457 | for (i = APPEND_INO; i <= UPDATE_INO; i++) { | |
67298804 CY |
458 | struct inode_management *im = &sbi->im[i]; |
459 | ||
460 | spin_lock(&im->ino_lock); | |
461 | list_for_each_entry_safe(e, tmp, &im->ino_list, list) { | |
fff04f90 | 462 | list_del(&e->list); |
67298804 | 463 | radix_tree_delete(&im->ino_root, e->ino); |
fff04f90 | 464 | kmem_cache_free(ino_entry_slab, e); |
67298804 | 465 | im->ino_num--; |
fff04f90 | 466 | } |
67298804 | 467 | spin_unlock(&im->ino_lock); |
fff04f90 JK |
468 | } |
469 | } | |
470 | ||
cbd56e7d | 471 | int acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 472 | { |
67298804 | 473 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a JK |
474 | int err = 0; |
475 | ||
67298804 CY |
476 | spin_lock(&im->ino_lock); |
477 | if (unlikely(im->ino_num >= sbi->max_orphans)) | |
127e670a | 478 | err = -ENOSPC; |
cbd56e7d | 479 | else |
67298804 CY |
480 | im->ino_num++; |
481 | spin_unlock(&im->ino_lock); | |
0d47c1ad | 482 | |
127e670a JK |
483 | return err; |
484 | } | |
485 | ||
cbd56e7d JK |
486 | void release_orphan_inode(struct f2fs_sb_info *sbi) |
487 | { | |
67298804 CY |
488 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
489 | ||
490 | spin_lock(&im->ino_lock); | |
491 | f2fs_bug_on(sbi, im->ino_num == 0); | |
492 | im->ino_num--; | |
493 | spin_unlock(&im->ino_lock); | |
cbd56e7d JK |
494 | } |
495 | ||
127e670a JK |
496 | void add_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
497 | { | |
39efac41 | 498 | /* add new orphan ino entry into list */ |
6451e041 | 499 | __add_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
500 | } |
501 | ||
502 | void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
503 | { | |
953e6cc6 | 504 | /* remove orphan entry from orphan list */ |
6451e041 | 505 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
506 | } |
507 | ||
8c14bfad | 508 | static int recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 509 | { |
8c14bfad CY |
510 | struct inode *inode; |
511 | ||
512 | inode = f2fs_iget(sbi->sb, ino); | |
513 | if (IS_ERR(inode)) { | |
514 | /* | |
515 | * there should be a bug that we can't find the entry | |
516 | * to orphan inode. | |
517 | */ | |
518 | f2fs_bug_on(sbi, PTR_ERR(inode) == -ENOENT); | |
519 | return PTR_ERR(inode); | |
520 | } | |
521 | ||
127e670a JK |
522 | clear_nlink(inode); |
523 | ||
524 | /* truncate all the data during iput */ | |
525 | iput(inode); | |
8c14bfad | 526 | return 0; |
127e670a JK |
527 | } |
528 | ||
8c14bfad | 529 | int recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a | 530 | { |
3c642985 | 531 | block_t start_blk, orphan_blocks, i, j; |
8c14bfad | 532 | int err; |
127e670a | 533 | |
25ca923b | 534 | if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG)) |
8c14bfad | 535 | return 0; |
127e670a | 536 | |
55141486 | 537 | start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi); |
3c642985 | 538 | orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi); |
127e670a | 539 | |
26879fb1 | 540 | ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP, true); |
662befda | 541 | |
3c642985 | 542 | for (i = 0; i < orphan_blocks; i++) { |
127e670a JK |
543 | struct page *page = get_meta_page(sbi, start_blk + i); |
544 | struct f2fs_orphan_block *orphan_blk; | |
545 | ||
546 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); | |
547 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
548 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
8c14bfad CY |
549 | err = recover_orphan_inode(sbi, ino); |
550 | if (err) { | |
551 | f2fs_put_page(page, 1); | |
8c14bfad CY |
552 | return err; |
553 | } | |
127e670a JK |
554 | } |
555 | f2fs_put_page(page, 1); | |
556 | } | |
557 | /* clear Orphan Flag */ | |
25ca923b | 558 | clear_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG); |
8c14bfad | 559 | return 0; |
127e670a JK |
560 | } |
561 | ||
562 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
563 | { | |
502c6e0b | 564 | struct list_head *head; |
127e670a | 565 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 566 | unsigned int nentries = 0; |
bd936f84 | 567 | unsigned short index = 1; |
8c402946 | 568 | unsigned short orphan_blocks; |
4531929e | 569 | struct page *page = NULL; |
6451e041 | 570 | struct ino_entry *orphan = NULL; |
67298804 | 571 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 572 | |
67298804 | 573 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 574 | |
d6c67a4f JK |
575 | /* |
576 | * we don't need to do spin_lock(&im->ino_lock) here, since all the | |
577 | * orphan inode operations are covered under f2fs_lock_op(). | |
578 | * And, spin_lock should be avoided due to page operations below. | |
579 | */ | |
67298804 | 580 | head = &im->ino_list; |
127e670a JK |
581 | |
582 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
583 | list_for_each_entry(orphan, head, list) { |
584 | if (!page) { | |
bd936f84 | 585 | page = grab_meta_page(sbi, start_blk++); |
502c6e0b GZ |
586 | orphan_blk = |
587 | (struct f2fs_orphan_block *)page_address(page); | |
588 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
589 | } | |
127e670a | 590 | |
36795567 | 591 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 592 | |
36795567 | 593 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
594 | /* |
595 | * an orphan block is full of 1020 entries, | |
596 | * then we need to flush current orphan blocks | |
597 | * and bring another one in memory | |
598 | */ | |
599 | orphan_blk->blk_addr = cpu_to_le16(index); | |
600 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
601 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
602 | set_page_dirty(page); | |
603 | f2fs_put_page(page, 1); | |
604 | index++; | |
127e670a JK |
605 | nentries = 0; |
606 | page = NULL; | |
607 | } | |
502c6e0b | 608 | } |
127e670a | 609 | |
502c6e0b GZ |
610 | if (page) { |
611 | orphan_blk->blk_addr = cpu_to_le16(index); | |
612 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
613 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
614 | set_page_dirty(page); | |
615 | f2fs_put_page(page, 1); | |
127e670a | 616 | } |
127e670a JK |
617 | } |
618 | ||
619 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, | |
620 | block_t cp_addr, unsigned long long *version) | |
621 | { | |
622 | struct page *cp_page_1, *cp_page_2 = NULL; | |
623 | unsigned long blk_size = sbi->blocksize; | |
624 | struct f2fs_checkpoint *cp_block; | |
625 | unsigned long long cur_version = 0, pre_version = 0; | |
127e670a | 626 | size_t crc_offset; |
7e586fa0 | 627 | __u32 crc = 0; |
127e670a JK |
628 | |
629 | /* Read the 1st cp block in this CP pack */ | |
630 | cp_page_1 = get_meta_page(sbi, cp_addr); | |
631 | ||
632 | /* get the version number */ | |
633 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_1); | |
634 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
635 | if (crc_offset >= blk_size) | |
636 | goto invalid_cp1; | |
637 | ||
29e7043f | 638 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
43b6573b | 639 | if (!f2fs_crc_valid(sbi, crc, cp_block, crc_offset)) |
127e670a JK |
640 | goto invalid_cp1; |
641 | ||
d71b5564 | 642 | pre_version = cur_cp_version(cp_block); |
127e670a JK |
643 | |
644 | /* Read the 2nd cp block in this CP pack */ | |
25ca923b | 645 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
127e670a JK |
646 | cp_page_2 = get_meta_page(sbi, cp_addr); |
647 | ||
648 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_2); | |
649 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
650 | if (crc_offset >= blk_size) | |
651 | goto invalid_cp2; | |
652 | ||
29e7043f | 653 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
43b6573b | 654 | if (!f2fs_crc_valid(sbi, crc, cp_block, crc_offset)) |
127e670a JK |
655 | goto invalid_cp2; |
656 | ||
d71b5564 | 657 | cur_version = cur_cp_version(cp_block); |
127e670a JK |
658 | |
659 | if (cur_version == pre_version) { | |
660 | *version = cur_version; | |
661 | f2fs_put_page(cp_page_2, 1); | |
662 | return cp_page_1; | |
663 | } | |
664 | invalid_cp2: | |
665 | f2fs_put_page(cp_page_2, 1); | |
666 | invalid_cp1: | |
667 | f2fs_put_page(cp_page_1, 1); | |
668 | return NULL; | |
669 | } | |
670 | ||
671 | int get_valid_checkpoint(struct f2fs_sb_info *sbi) | |
672 | { | |
673 | struct f2fs_checkpoint *cp_block; | |
674 | struct f2fs_super_block *fsb = sbi->raw_super; | |
675 | struct page *cp1, *cp2, *cur_page; | |
676 | unsigned long blk_size = sbi->blocksize; | |
677 | unsigned long long cp1_version = 0, cp2_version = 0; | |
678 | unsigned long long cp_start_blk_no; | |
55141486 | 679 | unsigned int cp_blks = 1 + __cp_payload(sbi); |
1dbe4152 CL |
680 | block_t cp_blk_no; |
681 | int i; | |
127e670a | 682 | |
1dbe4152 | 683 | sbi->ckpt = kzalloc(cp_blks * blk_size, GFP_KERNEL); |
127e670a JK |
684 | if (!sbi->ckpt) |
685 | return -ENOMEM; | |
686 | /* | |
687 | * Finding out valid cp block involves read both | |
688 | * sets( cp pack1 and cp pack 2) | |
689 | */ | |
690 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
691 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
692 | ||
693 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
694 | cp_start_blk_no += ((unsigned long long)1) << |
695 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
696 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
697 | ||
698 | if (cp1 && cp2) { | |
699 | if (ver_after(cp2_version, cp1_version)) | |
700 | cur_page = cp2; | |
701 | else | |
702 | cur_page = cp1; | |
703 | } else if (cp1) { | |
704 | cur_page = cp1; | |
705 | } else if (cp2) { | |
706 | cur_page = cp2; | |
707 | } else { | |
708 | goto fail_no_cp; | |
709 | } | |
710 | ||
711 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
712 | memcpy(sbi->ckpt, cp_block, blk_size); | |
713 | ||
984ec63c SL |
714 | /* Sanity checking of checkpoint */ |
715 | if (sanity_check_ckpt(sbi)) | |
716 | goto fail_no_cp; | |
717 | ||
1dbe4152 CL |
718 | if (cp_blks <= 1) |
719 | goto done; | |
720 | ||
721 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
722 | if (cur_page == cp2) | |
723 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
724 | ||
725 | for (i = 1; i < cp_blks; i++) { | |
726 | void *sit_bitmap_ptr; | |
727 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
728 | ||
729 | cur_page = get_meta_page(sbi, cp_blk_no + i); | |
730 | sit_bitmap_ptr = page_address(cur_page); | |
731 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
732 | f2fs_put_page(cur_page, 1); | |
733 | } | |
734 | done: | |
127e670a JK |
735 | f2fs_put_page(cp1, 1); |
736 | f2fs_put_page(cp2, 1); | |
737 | return 0; | |
738 | ||
739 | fail_no_cp: | |
740 | kfree(sbi->ckpt); | |
741 | return -EINVAL; | |
742 | } | |
743 | ||
c227f912 | 744 | static void __add_dirty_inode(struct inode *inode, enum inode_type type) |
127e670a | 745 | { |
4081363f | 746 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
2710fd7e | 747 | struct f2fs_inode_info *fi = F2FS_I(inode); |
c227f912 | 748 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
127e670a | 749 | |
c227f912 | 750 | if (is_inode_flag_set(fi, flag)) |
2710fd7e | 751 | return; |
2d7b822a | 752 | |
c227f912 CY |
753 | set_inode_flag(fi, flag); |
754 | list_add_tail(&fi->dirty_list, &sbi->inode_list[type]); | |
33fbd510 | 755 | stat_inc_dirty_inode(sbi, type); |
5deb8267 JK |
756 | } |
757 | ||
c227f912 | 758 | static void __remove_dirty_inode(struct inode *inode, enum inode_type type) |
6ad7609a | 759 | { |
6ad7609a | 760 | struct f2fs_inode_info *fi = F2FS_I(inode); |
c227f912 | 761 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
6ad7609a CY |
762 | |
763 | if (get_dirty_pages(inode) || | |
c227f912 | 764 | !is_inode_flag_set(F2FS_I(inode), flag)) |
6ad7609a CY |
765 | return; |
766 | ||
767 | list_del_init(&fi->dirty_list); | |
c227f912 | 768 | clear_inode_flag(fi, flag); |
33fbd510 | 769 | stat_dec_dirty_inode(F2FS_I_SB(inode), type); |
6ad7609a CY |
770 | } |
771 | ||
a7ffdbe2 | 772 | void update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 773 | { |
4081363f | 774 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 775 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
5deb8267 | 776 | |
5ac9f36f CY |
777 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
778 | !S_ISLNK(inode->i_mode)) | |
127e670a | 779 | return; |
7bd59381 | 780 | |
c227f912 CY |
781 | spin_lock(&sbi->inode_lock[type]); |
782 | __add_dirty_inode(inode, type); | |
a7ffdbe2 | 783 | inode_inc_dirty_pages(inode); |
c227f912 | 784 | spin_unlock(&sbi->inode_lock[type]); |
cf0ee0f0 | 785 | |
a7ffdbe2 | 786 | SetPagePrivate(page); |
9e4ded3f | 787 | f2fs_trace_pid(page); |
5deb8267 JK |
788 | } |
789 | ||
790 | void add_dirty_dir_inode(struct inode *inode) | |
791 | { | |
4081363f | 792 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
127e670a | 793 | |
c227f912 CY |
794 | spin_lock(&sbi->inode_lock[DIR_INODE]); |
795 | __add_dirty_inode(inode, DIR_INODE); | |
796 | spin_unlock(&sbi->inode_lock[DIR_INODE]); | |
127e670a JK |
797 | } |
798 | ||
c227f912 | 799 | void remove_dirty_inode(struct inode *inode) |
127e670a | 800 | { |
4081363f | 801 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
2710fd7e | 802 | struct f2fs_inode_info *fi = F2FS_I(inode); |
c227f912 | 803 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
127e670a | 804 | |
c227f912 CY |
805 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
806 | !S_ISLNK(inode->i_mode)) | |
127e670a JK |
807 | return; |
808 | ||
c227f912 CY |
809 | spin_lock(&sbi->inode_lock[type]); |
810 | __remove_dirty_inode(inode, type); | |
811 | spin_unlock(&sbi->inode_lock[type]); | |
74d0b917 JK |
812 | |
813 | /* Only from the recovery routine */ | |
2710fd7e CY |
814 | if (is_inode_flag_set(fi, FI_DELAY_IPUT)) { |
815 | clear_inode_flag(fi, FI_DELAY_IPUT); | |
74d0b917 | 816 | iput(inode); |
afc3eda2 | 817 | } |
74d0b917 JK |
818 | } |
819 | ||
6d5a1495 | 820 | int sync_dirty_inodes(struct f2fs_sb_info *sbi, enum inode_type type) |
127e670a | 821 | { |
ce3b7d80 | 822 | struct list_head *head; |
127e670a | 823 | struct inode *inode; |
2710fd7e | 824 | struct f2fs_inode_info *fi; |
4cf18537 CY |
825 | bool is_dir = (type == DIR_INODE); |
826 | ||
827 | trace_f2fs_sync_dirty_inodes_enter(sbi->sb, is_dir, | |
828 | get_pages(sbi, is_dir ? | |
829 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
127e670a | 830 | retry: |
af41d3ee | 831 | if (unlikely(f2fs_cp_error(sbi))) |
6d5a1495 | 832 | return -EIO; |
af41d3ee | 833 | |
c227f912 | 834 | spin_lock(&sbi->inode_lock[type]); |
ce3b7d80 | 835 | |
c227f912 | 836 | head = &sbi->inode_list[type]; |
127e670a | 837 | if (list_empty(head)) { |
c227f912 | 838 | spin_unlock(&sbi->inode_lock[type]); |
4cf18537 CY |
839 | trace_f2fs_sync_dirty_inodes_exit(sbi->sb, is_dir, |
840 | get_pages(sbi, is_dir ? | |
841 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
6d5a1495 | 842 | return 0; |
127e670a | 843 | } |
2710fd7e CY |
844 | fi = list_entry(head->next, struct f2fs_inode_info, dirty_list); |
845 | inode = igrab(&fi->vfs_inode); | |
c227f912 | 846 | spin_unlock(&sbi->inode_lock[type]); |
127e670a | 847 | if (inode) { |
87d6f890 | 848 | filemap_fdatawrite(inode->i_mapping); |
127e670a JK |
849 | iput(inode); |
850 | } else { | |
851 | /* | |
852 | * We should submit bio, since it exists several | |
853 | * wribacking dentry pages in the freeing inode. | |
854 | */ | |
458e6197 | 855 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
7ecebe5e | 856 | cond_resched(); |
127e670a JK |
857 | } |
858 | goto retry; | |
859 | } | |
860 | ||
0a8165d7 | 861 | /* |
127e670a JK |
862 | * Freeze all the FS-operations for checkpoint. |
863 | */ | |
cf779cab | 864 | static int block_operations(struct f2fs_sb_info *sbi) |
127e670a | 865 | { |
127e670a JK |
866 | struct writeback_control wbc = { |
867 | .sync_mode = WB_SYNC_ALL, | |
868 | .nr_to_write = LONG_MAX, | |
869 | .for_reclaim = 0, | |
870 | }; | |
c718379b | 871 | struct blk_plug plug; |
cf779cab | 872 | int err = 0; |
c718379b JK |
873 | |
874 | blk_start_plug(&plug); | |
875 | ||
39936837 | 876 | retry_flush_dents: |
e479556b | 877 | f2fs_lock_all(sbi); |
127e670a | 878 | /* write all the dirty dentry pages */ |
127e670a | 879 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 880 | f2fs_unlock_all(sbi); |
6d5a1495 CY |
881 | err = sync_dirty_inodes(sbi, DIR_INODE); |
882 | if (err) | |
cf779cab | 883 | goto out; |
39936837 | 884 | goto retry_flush_dents; |
127e670a JK |
885 | } |
886 | ||
127e670a | 887 | /* |
e1c42045 | 888 | * POR: we should ensure that there are no dirty node pages |
127e670a JK |
889 | * until finishing nat/sit flush. |
890 | */ | |
39936837 | 891 | retry_flush_nodes: |
b3582c68 | 892 | down_write(&sbi->node_write); |
127e670a JK |
893 | |
894 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
b3582c68 | 895 | up_write(&sbi->node_write); |
52681375 | 896 | err = sync_node_pages(sbi, &wbc); |
6d5a1495 | 897 | if (err) { |
cf779cab | 898 | f2fs_unlock_all(sbi); |
cf779cab JK |
899 | goto out; |
900 | } | |
39936837 | 901 | goto retry_flush_nodes; |
127e670a | 902 | } |
cf779cab | 903 | out: |
c718379b | 904 | blk_finish_plug(&plug); |
cf779cab | 905 | return err; |
127e670a JK |
906 | } |
907 | ||
908 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
909 | { | |
b3582c68 | 910 | up_write(&sbi->node_write); |
e479556b | 911 | f2fs_unlock_all(sbi); |
127e670a JK |
912 | } |
913 | ||
fb51b5ef CL |
914 | static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi) |
915 | { | |
916 | DEFINE_WAIT(wait); | |
917 | ||
918 | for (;;) { | |
919 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
920 | ||
921 | if (!get_pages(sbi, F2FS_WRITEBACK)) | |
922 | break; | |
923 | ||
0ff21646 | 924 | io_schedule_timeout(5*HZ); |
fb51b5ef CL |
925 | } |
926 | finish_wait(&sbi->cp_wait, &wait); | |
927 | } | |
928 | ||
c34f42e2 | 929 | static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
930 | { |
931 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
cf2271e7 | 932 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); |
77041823 | 933 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
67298804 | 934 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; |
77041823 | 935 | nid_t last_nid = nm_i->next_scan_nid; |
127e670a | 936 | block_t start_blk; |
127e670a | 937 | unsigned int data_sum_blocks, orphan_blocks; |
7e586fa0 | 938 | __u32 crc32 = 0; |
127e670a | 939 | int i; |
55141486 | 940 | int cp_payload_blks = __cp_payload(sbi); |
e90c2d28 CY |
941 | block_t discard_blk = NEXT_FREE_BLKADDR(sbi, curseg); |
942 | bool invalidate = false; | |
8f1dbbbb SL |
943 | struct super_block *sb = sbi->sb; |
944 | struct curseg_info *seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); | |
945 | u64 kbytes_written; | |
127e670a | 946 | |
1e87a78d JK |
947 | /* |
948 | * This avoids to conduct wrong roll-forward operations and uses | |
949 | * metapages, so should be called prior to sync_meta_pages below. | |
950 | */ | |
e90c2d28 CY |
951 | if (discard_next_dnode(sbi, discard_blk)) |
952 | invalidate = true; | |
127e670a JK |
953 | |
954 | /* Flush all the NAT/SIT pages */ | |
cf779cab | 955 | while (get_pages(sbi, F2FS_DIRTY_META)) { |
127e670a | 956 | sync_meta_pages(sbi, META, LONG_MAX); |
cf779cab | 957 | if (unlikely(f2fs_cp_error(sbi))) |
c34f42e2 | 958 | return -EIO; |
cf779cab | 959 | } |
127e670a JK |
960 | |
961 | next_free_nid(sbi, &last_nid); | |
962 | ||
963 | /* | |
964 | * modify checkpoint | |
965 | * version number is already updated | |
966 | */ | |
967 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi)); | |
968 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
969 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); | |
b5b82205 | 970 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
971 | ckpt->cur_node_segno[i] = |
972 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
973 | ckpt->cur_node_blkoff[i] = | |
974 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
975 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
976 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
977 | } | |
b5b82205 | 978 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
979 | ckpt->cur_data_segno[i] = |
980 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
981 | ckpt->cur_data_blkoff[i] = | |
982 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
983 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
984 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
985 | } | |
986 | ||
987 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
988 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
989 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
990 | ||
991 | /* 2 cp + n data seg summary + orphan inode blocks */ | |
3fa06d7b | 992 | data_sum_blocks = npages_for_summary_flush(sbi, false); |
b5b82205 | 993 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
25ca923b | 994 | set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 995 | else |
25ca923b | 996 | clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 997 | |
67298804 | 998 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
999 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
1000 | orphan_blocks); | |
127e670a | 1001 | |
119ee914 | 1002 | if (__remain_node_summaries(cpc->reason)) |
b5b82205 | 1003 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
1004 | cp_payload_blks + data_sum_blocks + |
1005 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
119ee914 | 1006 | else |
b5b82205 | 1007 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS + |
1dbe4152 CL |
1008 | cp_payload_blks + data_sum_blocks + |
1009 | orphan_blocks); | |
119ee914 JK |
1010 | |
1011 | if (cpc->reason == CP_UMOUNT) | |
1012 | set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
1013 | else | |
1014 | clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
1015 | ||
1016 | if (cpc->reason == CP_FASTBOOT) | |
1017 | set_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
1018 | else | |
1019 | clear_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
127e670a | 1020 | |
67298804 | 1021 | if (orphan_num) |
25ca923b | 1022 | set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1023 | else |
25ca923b | 1024 | clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1025 | |
caf0047e | 1026 | if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) |
2ae4c673 JK |
1027 | set_ckpt_flags(ckpt, CP_FSCK_FLAG); |
1028 | ||
127e670a JK |
1029 | /* update SIT/NAT bitmap */ |
1030 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
1031 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
1032 | ||
43b6573b | 1033 | crc32 = f2fs_crc32(sbi, ckpt, le32_to_cpu(ckpt->checksum_offset)); |
7e586fa0 JK |
1034 | *((__le32 *)((unsigned char *)ckpt + |
1035 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
1036 | = cpu_to_le32(crc32); |
1037 | ||
1038 | start_blk = __start_cp_addr(sbi); | |
1039 | ||
a7230d16 JK |
1040 | /* need to wait for end_io results */ |
1041 | wait_on_all_pages_writeback(sbi); | |
1042 | if (unlikely(f2fs_cp_error(sbi))) | |
c34f42e2 | 1043 | return -EIO; |
a7230d16 | 1044 | |
127e670a | 1045 | /* write out checkpoint buffer at block 0 */ |
381722d2 CY |
1046 | update_meta_page(sbi, ckpt, start_blk++); |
1047 | ||
1048 | for (i = 1; i < 1 + cp_payload_blks; i++) | |
1049 | update_meta_page(sbi, (char *)ckpt + i * F2FS_BLKSIZE, | |
1050 | start_blk++); | |
1dbe4152 | 1051 | |
67298804 | 1052 | if (orphan_num) { |
127e670a JK |
1053 | write_orphan_inodes(sbi, start_blk); |
1054 | start_blk += orphan_blocks; | |
1055 | } | |
1056 | ||
1057 | write_data_summaries(sbi, start_blk); | |
1058 | start_blk += data_sum_blocks; | |
8f1dbbbb SL |
1059 | |
1060 | /* Record write statistics in the hot node summary */ | |
1061 | kbytes_written = sbi->kbytes_written; | |
1062 | if (sb->s_bdev->bd_part) | |
1063 | kbytes_written += BD_PART_WRITTEN(sbi); | |
1064 | ||
b7ad7512 | 1065 | seg_i->journal->info.kbytes_written = cpu_to_le64(kbytes_written); |
8f1dbbbb | 1066 | |
119ee914 | 1067 | if (__remain_node_summaries(cpc->reason)) { |
127e670a JK |
1068 | write_node_summaries(sbi, start_blk); |
1069 | start_blk += NR_CURSEG_NODE_TYPE; | |
1070 | } | |
1071 | ||
1072 | /* writeout checkpoint block */ | |
381722d2 | 1073 | update_meta_page(sbi, ckpt, start_blk); |
127e670a JK |
1074 | |
1075 | /* wait for previous submitted node/meta pages writeback */ | |
fb51b5ef | 1076 | wait_on_all_pages_writeback(sbi); |
127e670a | 1077 | |
cf779cab | 1078 | if (unlikely(f2fs_cp_error(sbi))) |
c34f42e2 | 1079 | return -EIO; |
cf779cab | 1080 | |
80dd9c0e CY |
1081 | filemap_fdatawait_range(NODE_MAPPING(sbi), 0, LLONG_MAX); |
1082 | filemap_fdatawait_range(META_MAPPING(sbi), 0, LLONG_MAX); | |
127e670a JK |
1083 | |
1084 | /* update user_block_counts */ | |
1085 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
1086 | sbi->alloc_valid_block_count = 0; | |
1087 | ||
1088 | /* Here, we only have one bio having CP pack */ | |
577e3495 | 1089 | sync_meta_pages(sbi, META_FLUSH, LONG_MAX); |
127e670a | 1090 | |
6a8f8ca5 JK |
1091 | /* wait for previous submitted meta pages writeback */ |
1092 | wait_on_all_pages_writeback(sbi); | |
1093 | ||
e90c2d28 CY |
1094 | /* |
1095 | * invalidate meta page which is used temporarily for zeroing out | |
1096 | * block at the end of warm node chain. | |
1097 | */ | |
1098 | if (invalidate) | |
1099 | invalidate_mapping_pages(META_MAPPING(sbi), discard_blk, | |
1100 | discard_blk); | |
1101 | ||
a49324f1 | 1102 | release_ino_entry(sbi); |
cf779cab JK |
1103 | |
1104 | if (unlikely(f2fs_cp_error(sbi))) | |
c34f42e2 | 1105 | return -EIO; |
cf779cab | 1106 | |
836b5a63 | 1107 | clear_prefree_segments(sbi, cpc); |
caf0047e | 1108 | clear_sbi_flag(sbi, SBI_IS_DIRTY); |
c34f42e2 CY |
1109 | |
1110 | return 0; | |
127e670a JK |
1111 | } |
1112 | ||
0a8165d7 | 1113 | /* |
e1c42045 | 1114 | * We guarantee that this checkpoint procedure will not fail. |
127e670a | 1115 | */ |
c34f42e2 | 1116 | int write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1117 | { |
1118 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1119 | unsigned long long ckpt_ver; | |
c34f42e2 | 1120 | int err = 0; |
127e670a | 1121 | |
43727527 | 1122 | mutex_lock(&sbi->cp_mutex); |
8501017e | 1123 | |
caf0047e | 1124 | if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) && |
a66cdd98 JK |
1125 | (cpc->reason == CP_FASTBOOT || cpc->reason == CP_SYNC || |
1126 | (cpc->reason == CP_DISCARD && !sbi->discard_blks))) | |
8501017e | 1127 | goto out; |
c34f42e2 CY |
1128 | if (unlikely(f2fs_cp_error(sbi))) { |
1129 | err = -EIO; | |
cf779cab | 1130 | goto out; |
c34f42e2 CY |
1131 | } |
1132 | if (f2fs_readonly(sbi->sb)) { | |
1133 | err = -EROFS; | |
11504a8e | 1134 | goto out; |
c34f42e2 | 1135 | } |
2bda542d WL |
1136 | |
1137 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); | |
1138 | ||
c34f42e2 CY |
1139 | err = block_operations(sbi); |
1140 | if (err) | |
cf779cab | 1141 | goto out; |
127e670a | 1142 | |
75ab4cb8 | 1143 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1144 | |
406657dd | 1145 | f2fs_flush_merged_bios(sbi); |
127e670a JK |
1146 | |
1147 | /* | |
1148 | * update checkpoint pack index | |
1149 | * Increase the version number so that | |
1150 | * SIT entries and seg summaries are written at correct place | |
1151 | */ | |
d71b5564 | 1152 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1153 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1154 | ||
1155 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
1156 | flush_nat_entries(sbi); | |
4b2fecc8 | 1157 | flush_sit_entries(sbi, cpc); |
127e670a | 1158 | |
127e670a | 1159 | /* unlock all the fs_lock[] in do_checkpoint() */ |
c34f42e2 | 1160 | err = do_checkpoint(sbi, cpc); |
127e670a JK |
1161 | |
1162 | unblock_operations(sbi); | |
942e0be6 | 1163 | stat_inc_cp_count(sbi->stat_info); |
10027551 JK |
1164 | |
1165 | if (cpc->reason == CP_RECOVERY) | |
1166 | f2fs_msg(sbi->sb, KERN_NOTICE, | |
1167 | "checkpoint: version = %llx", ckpt_ver); | |
60b99b48 JK |
1168 | |
1169 | /* do checkpoint periodically */ | |
6beceb54 | 1170 | f2fs_update_time(sbi, CP_TIME); |
55d1cdb2 | 1171 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
8501017e JK |
1172 | out: |
1173 | mutex_unlock(&sbi->cp_mutex); | |
c34f42e2 | 1174 | return err; |
127e670a JK |
1175 | } |
1176 | ||
6451e041 | 1177 | void init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1178 | { |
6451e041 JK |
1179 | int i; |
1180 | ||
1181 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1182 | struct inode_management *im = &sbi->im[i]; |
1183 | ||
1184 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1185 | spin_lock_init(&im->ino_lock); | |
1186 | INIT_LIST_HEAD(&im->ino_list); | |
1187 | im->ino_num = 0; | |
6451e041 JK |
1188 | } |
1189 | ||
b5b82205 | 1190 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
14b42817 WL |
1191 | NR_CURSEG_TYPE - __cp_payload(sbi)) * |
1192 | F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
1193 | } |
1194 | ||
6e6093a8 | 1195 | int __init create_checkpoint_caches(void) |
127e670a | 1196 | { |
6451e041 JK |
1197 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1198 | sizeof(struct ino_entry)); | |
1199 | if (!ino_entry_slab) | |
127e670a | 1200 | return -ENOMEM; |
06292073 CY |
1201 | inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry", |
1202 | sizeof(struct inode_entry)); | |
6bacf52f | 1203 | if (!inode_entry_slab) { |
6451e041 | 1204 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1205 | return -ENOMEM; |
1206 | } | |
1207 | return 0; | |
1208 | } | |
1209 | ||
1210 | void destroy_checkpoint_caches(void) | |
1211 | { | |
6451e041 | 1212 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1213 | kmem_cache_destroy(inode_entry_slab); |
1214 | } |