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: | |
bde44686 | 37 | page = grab_cache_page(mapping, index); |
127e670a JK |
38 | if (!page) { |
39 | cond_resched(); | |
40 | goto repeat; | |
41 | } | |
bde44686 | 42 | f2fs_wait_on_page_writeback(page, META); |
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, | |
59 | .blk_addr = index, | |
4375a336 | 60 | .encrypted_page = NULL, |
cf04e8eb | 61 | }; |
2b947003 CY |
62 | |
63 | if (unlikely(!is_meta)) | |
64 | fio.rw &= ~REQ_META; | |
127e670a JK |
65 | repeat: |
66 | page = grab_cache_page(mapping, index); | |
67 | if (!page) { | |
68 | cond_resched(); | |
69 | goto repeat; | |
70 | } | |
393ff91f JK |
71 | if (PageUptodate(page)) |
72 | goto out; | |
73 | ||
05ca3632 JK |
74 | fio.page = page; |
75 | ||
86531d6b JK |
76 | if (f2fs_submit_page_bio(&fio)) { |
77 | f2fs_put_page(page, 1); | |
127e670a | 78 | goto repeat; |
86531d6b | 79 | } |
127e670a | 80 | |
393ff91f | 81 | lock_page(page); |
6bacf52f | 82 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
83 | f2fs_put_page(page, 1); |
84 | goto repeat; | |
85 | } | |
f3f338ca CY |
86 | |
87 | /* | |
88 | * if there is any IO error when accessing device, make our filesystem | |
89 | * readonly and make sure do not write checkpoint with non-uptodate | |
90 | * meta page. | |
91 | */ | |
92 | if (unlikely(!PageUptodate(page))) | |
93 | f2fs_stop_checkpoint(sbi); | |
393ff91f | 94 | out: |
127e670a JK |
95 | return page; |
96 | } | |
97 | ||
2b947003 CY |
98 | struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) |
99 | { | |
100 | return __get_meta_page(sbi, index, true); | |
101 | } | |
102 | ||
103 | /* for POR only */ | |
104 | struct page *get_tmp_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
105 | { | |
106 | return __get_meta_page(sbi, index, false); | |
107 | } | |
108 | ||
f0c9cada | 109 | bool is_valid_blkaddr(struct f2fs_sb_info *sbi, block_t blkaddr, int type) |
662befda CY |
110 | { |
111 | switch (type) { | |
112 | case META_NAT: | |
66b00c18 | 113 | break; |
662befda | 114 | case META_SIT: |
66b00c18 CY |
115 | if (unlikely(blkaddr >= SIT_BLK_CNT(sbi))) |
116 | return false; | |
117 | break; | |
81c1a0f1 | 118 | case META_SSA: |
66b00c18 CY |
119 | if (unlikely(blkaddr >= MAIN_BLKADDR(sbi) || |
120 | blkaddr < SM_I(sbi)->ssa_blkaddr)) | |
121 | return false; | |
122 | break; | |
662befda | 123 | case META_CP: |
66b00c18 CY |
124 | if (unlikely(blkaddr >= SIT_I(sbi)->sit_base_addr || |
125 | blkaddr < __start_cp_addr(sbi))) | |
126 | return false; | |
127 | break; | |
4c521f49 | 128 | case META_POR: |
66b00c18 CY |
129 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
130 | blkaddr < MAIN_BLKADDR(sbi))) | |
131 | return false; | |
132 | break; | |
662befda CY |
133 | default: |
134 | BUG(); | |
135 | } | |
66b00c18 CY |
136 | |
137 | return true; | |
662befda CY |
138 | } |
139 | ||
140 | /* | |
81c1a0f1 | 141 | * Readahead CP/NAT/SIT/SSA pages |
662befda | 142 | */ |
4c521f49 | 143 | int ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages, int type) |
662befda CY |
144 | { |
145 | block_t prev_blk_addr = 0; | |
146 | struct page *page; | |
4c521f49 | 147 | block_t blkno = start; |
662befda | 148 | struct f2fs_io_info fio = { |
05ca3632 | 149 | .sbi = sbi, |
662befda | 150 | .type = META, |
4375a336 JK |
151 | .rw = READ_SYNC | REQ_META | REQ_PRIO, |
152 | .encrypted_page = NULL, | |
662befda CY |
153 | }; |
154 | ||
2b947003 CY |
155 | if (unlikely(type == META_POR)) |
156 | fio.rw &= ~REQ_META; | |
157 | ||
662befda | 158 | for (; nrpages-- > 0; blkno++) { |
662befda | 159 | |
66b00c18 CY |
160 | if (!is_valid_blkaddr(sbi, blkno, type)) |
161 | goto out; | |
162 | ||
662befda CY |
163 | switch (type) { |
164 | case META_NAT: | |
66b00c18 CY |
165 | if (unlikely(blkno >= |
166 | NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))) | |
662befda | 167 | blkno = 0; |
66b00c18 | 168 | /* get nat block addr */ |
cf04e8eb | 169 | fio.blk_addr = current_nat_addr(sbi, |
662befda CY |
170 | blkno * NAT_ENTRY_PER_BLOCK); |
171 | break; | |
172 | case META_SIT: | |
173 | /* get sit block addr */ | |
cf04e8eb | 174 | fio.blk_addr = current_sit_addr(sbi, |
662befda | 175 | blkno * SIT_ENTRY_PER_BLOCK); |
cf04e8eb | 176 | if (blkno != start && prev_blk_addr + 1 != fio.blk_addr) |
662befda | 177 | goto out; |
cf04e8eb | 178 | prev_blk_addr = fio.blk_addr; |
662befda | 179 | break; |
81c1a0f1 | 180 | case META_SSA: |
662befda | 181 | case META_CP: |
4c521f49 | 182 | case META_POR: |
cf04e8eb | 183 | fio.blk_addr = blkno; |
662befda CY |
184 | break; |
185 | default: | |
186 | BUG(); | |
187 | } | |
188 | ||
cf04e8eb | 189 | page = grab_cache_page(META_MAPPING(sbi), fio.blk_addr); |
662befda CY |
190 | if (!page) |
191 | continue; | |
192 | if (PageUptodate(page)) { | |
662befda CY |
193 | f2fs_put_page(page, 1); |
194 | continue; | |
195 | } | |
196 | ||
05ca3632 JK |
197 | fio.page = page; |
198 | f2fs_submit_page_mbio(&fio); | |
662befda CY |
199 | f2fs_put_page(page, 0); |
200 | } | |
201 | out: | |
202 | f2fs_submit_merged_bio(sbi, META, READ); | |
203 | return blkno - start; | |
204 | } | |
205 | ||
635aee1f CY |
206 | void ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index) |
207 | { | |
208 | struct page *page; | |
209 | bool readahead = false; | |
210 | ||
211 | page = find_get_page(META_MAPPING(sbi), index); | |
212 | if (!page || (page && !PageUptodate(page))) | |
213 | readahead = true; | |
214 | f2fs_put_page(page, 0); | |
215 | ||
216 | if (readahead) | |
217 | ra_meta_pages(sbi, index, MAX_BIO_BLOCKS(sbi), META_POR); | |
218 | } | |
219 | ||
127e670a JK |
220 | static int f2fs_write_meta_page(struct page *page, |
221 | struct writeback_control *wbc) | |
222 | { | |
4081363f | 223 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 224 | |
ecda0de3 CY |
225 | trace_f2fs_writepage(page, META); |
226 | ||
caf0047e | 227 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
cfb271d4 | 228 | goto redirty_out; |
857dc4e0 | 229 | if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0)) |
cfb271d4 | 230 | goto redirty_out; |
1e968fdf | 231 | if (unlikely(f2fs_cp_error(sbi))) |
cf779cab | 232 | goto redirty_out; |
127e670a | 233 | |
3cb5ad15 | 234 | f2fs_wait_on_page_writeback(page, META); |
577e3495 JK |
235 | write_meta_page(sbi, page); |
236 | dec_page_count(sbi, F2FS_DIRTY_META); | |
237 | unlock_page(page); | |
857dc4e0 JK |
238 | |
239 | if (wbc->for_reclaim) | |
240 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
577e3495 | 241 | return 0; |
cfb271d4 CY |
242 | |
243 | redirty_out: | |
76f60268 | 244 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 245 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
246 | } |
247 | ||
248 | static int f2fs_write_meta_pages(struct address_space *mapping, | |
249 | struct writeback_control *wbc) | |
250 | { | |
4081363f | 251 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 252 | long diff, written; |
127e670a | 253 | |
e5748434 CY |
254 | trace_f2fs_writepages(mapping->host, wbc, META); |
255 | ||
5459aa97 | 256 | /* collect a number of dirty meta pages and write together */ |
50c8cdb3 JK |
257 | if (wbc->for_kupdate || |
258 | get_pages(sbi, F2FS_DIRTY_META) < nr_pages_to_skip(sbi, META)) | |
d3baf95d | 259 | goto skip_write; |
127e670a JK |
260 | |
261 | /* if mounting is failed, skip writing node pages */ | |
262 | mutex_lock(&sbi->cp_mutex); | |
50c8cdb3 JK |
263 | diff = nr_pages_to_write(sbi, META, wbc); |
264 | written = sync_meta_pages(sbi, META, wbc->nr_to_write); | |
127e670a | 265 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 266 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 267 | return 0; |
d3baf95d JK |
268 | |
269 | skip_write: | |
270 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
271 | return 0; | |
127e670a JK |
272 | } |
273 | ||
274 | long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, | |
275 | long nr_to_write) | |
276 | { | |
9df27d98 | 277 | struct address_space *mapping = META_MAPPING(sbi); |
6066d8cd | 278 | pgoff_t index = 0, end = LONG_MAX, prev = LONG_MAX; |
127e670a JK |
279 | struct pagevec pvec; |
280 | long nwritten = 0; | |
281 | struct writeback_control wbc = { | |
282 | .for_reclaim = 0, | |
283 | }; | |
284 | ||
285 | pagevec_init(&pvec, 0); | |
286 | ||
287 | while (index <= end) { | |
288 | int i, nr_pages; | |
289 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, | |
290 | PAGECACHE_TAG_DIRTY, | |
291 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
cfb271d4 | 292 | if (unlikely(nr_pages == 0)) |
127e670a JK |
293 | break; |
294 | ||
295 | for (i = 0; i < nr_pages; i++) { | |
296 | struct page *page = pvec.pages[i]; | |
203681f6 | 297 | |
6066d8cd JK |
298 | if (prev == LONG_MAX) |
299 | prev = page->index - 1; | |
300 | if (nr_to_write != LONG_MAX && page->index != prev + 1) { | |
301 | pagevec_release(&pvec); | |
302 | goto stop; | |
303 | } | |
304 | ||
127e670a | 305 | lock_page(page); |
203681f6 JK |
306 | |
307 | if (unlikely(page->mapping != mapping)) { | |
308 | continue_unlock: | |
309 | unlock_page(page); | |
310 | continue; | |
311 | } | |
312 | if (!PageDirty(page)) { | |
313 | /* someone wrote it for us */ | |
314 | goto continue_unlock; | |
315 | } | |
316 | ||
317 | if (!clear_page_dirty_for_io(page)) | |
318 | goto continue_unlock; | |
319 | ||
97dc3fd2 | 320 | if (mapping->a_ops->writepage(page, &wbc)) { |
577e3495 JK |
321 | unlock_page(page); |
322 | break; | |
323 | } | |
cfb271d4 | 324 | nwritten++; |
6066d8cd | 325 | prev = page->index; |
cfb271d4 | 326 | if (unlikely(nwritten >= nr_to_write)) |
127e670a JK |
327 | break; |
328 | } | |
329 | pagevec_release(&pvec); | |
330 | cond_resched(); | |
331 | } | |
6066d8cd | 332 | stop: |
127e670a | 333 | if (nwritten) |
458e6197 | 334 | f2fs_submit_merged_bio(sbi, type, WRITE); |
127e670a JK |
335 | |
336 | return nwritten; | |
337 | } | |
338 | ||
339 | static int f2fs_set_meta_page_dirty(struct page *page) | |
340 | { | |
26c6b887 JK |
341 | trace_f2fs_set_page_dirty(page, META); |
342 | ||
127e670a JK |
343 | SetPageUptodate(page); |
344 | if (!PageDirty(page)) { | |
345 | __set_page_dirty_nobuffers(page); | |
4081363f | 346 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); |
1601839e | 347 | SetPagePrivate(page); |
9e4ded3f | 348 | f2fs_trace_pid(page); |
127e670a JK |
349 | return 1; |
350 | } | |
351 | return 0; | |
352 | } | |
353 | ||
354 | const struct address_space_operations f2fs_meta_aops = { | |
355 | .writepage = f2fs_write_meta_page, | |
356 | .writepages = f2fs_write_meta_pages, | |
357 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
487261f3 CY |
358 | .invalidatepage = f2fs_invalidate_page, |
359 | .releasepage = f2fs_release_page, | |
127e670a JK |
360 | }; |
361 | ||
6451e041 | 362 | static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 363 | { |
67298804 | 364 | struct inode_management *im = &sbi->im[type]; |
80c54505 JK |
365 | struct ino_entry *e, *tmp; |
366 | ||
367 | tmp = f2fs_kmem_cache_alloc(ino_entry_slab, GFP_NOFS); | |
39efac41 | 368 | retry: |
80c54505 | 369 | radix_tree_preload(GFP_NOFS | __GFP_NOFAIL); |
769ec6e5 | 370 | |
67298804 | 371 | spin_lock(&im->ino_lock); |
67298804 | 372 | e = radix_tree_lookup(&im->ino_root, ino); |
39efac41 | 373 | if (!e) { |
80c54505 | 374 | e = tmp; |
67298804 CY |
375 | if (radix_tree_insert(&im->ino_root, ino, e)) { |
376 | spin_unlock(&im->ino_lock); | |
769ec6e5 | 377 | radix_tree_preload_end(); |
39efac41 JK |
378 | goto retry; |
379 | } | |
380 | memset(e, 0, sizeof(struct ino_entry)); | |
381 | e->ino = ino; | |
953e6cc6 | 382 | |
67298804 | 383 | list_add_tail(&e->list, &im->ino_list); |
8c402946 | 384 | if (type != ORPHAN_INO) |
67298804 | 385 | im->ino_num++; |
39efac41 | 386 | } |
67298804 | 387 | spin_unlock(&im->ino_lock); |
769ec6e5 | 388 | radix_tree_preload_end(); |
80c54505 JK |
389 | |
390 | if (e != tmp) | |
391 | kmem_cache_free(ino_entry_slab, tmp); | |
953e6cc6 JK |
392 | } |
393 | ||
6451e041 | 394 | static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 395 | { |
67298804 | 396 | struct inode_management *im = &sbi->im[type]; |
6451e041 | 397 | struct ino_entry *e; |
953e6cc6 | 398 | |
67298804 CY |
399 | spin_lock(&im->ino_lock); |
400 | e = radix_tree_lookup(&im->ino_root, ino); | |
39efac41 JK |
401 | if (e) { |
402 | list_del(&e->list); | |
67298804 CY |
403 | radix_tree_delete(&im->ino_root, ino); |
404 | im->ino_num--; | |
405 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
406 | kmem_cache_free(ino_entry_slab, e); |
407 | return; | |
953e6cc6 | 408 | } |
67298804 | 409 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
410 | } |
411 | ||
fff04f90 JK |
412 | void add_dirty_inode(struct f2fs_sb_info *sbi, nid_t ino, int type) |
413 | { | |
414 | /* add new dirty ino entry into list */ | |
415 | __add_ino_entry(sbi, ino, type); | |
416 | } | |
417 | ||
418 | void remove_dirty_inode(struct f2fs_sb_info *sbi, nid_t ino, int type) | |
419 | { | |
420 | /* remove dirty ino entry from list */ | |
421 | __remove_ino_entry(sbi, ino, type); | |
422 | } | |
423 | ||
424 | /* mode should be APPEND_INO or UPDATE_INO */ | |
425 | bool exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode) | |
426 | { | |
67298804 | 427 | struct inode_management *im = &sbi->im[mode]; |
fff04f90 | 428 | struct ino_entry *e; |
67298804 CY |
429 | |
430 | spin_lock(&im->ino_lock); | |
431 | e = radix_tree_lookup(&im->ino_root, ino); | |
432 | spin_unlock(&im->ino_lock); | |
fff04f90 JK |
433 | return e ? true : false; |
434 | } | |
435 | ||
6f12ac25 | 436 | void release_dirty_inode(struct f2fs_sb_info *sbi) |
fff04f90 JK |
437 | { |
438 | struct ino_entry *e, *tmp; | |
439 | int i; | |
440 | ||
441 | for (i = APPEND_INO; i <= UPDATE_INO; i++) { | |
67298804 CY |
442 | struct inode_management *im = &sbi->im[i]; |
443 | ||
444 | spin_lock(&im->ino_lock); | |
445 | list_for_each_entry_safe(e, tmp, &im->ino_list, list) { | |
fff04f90 | 446 | list_del(&e->list); |
67298804 | 447 | radix_tree_delete(&im->ino_root, e->ino); |
fff04f90 | 448 | kmem_cache_free(ino_entry_slab, e); |
67298804 | 449 | im->ino_num--; |
fff04f90 | 450 | } |
67298804 | 451 | spin_unlock(&im->ino_lock); |
fff04f90 JK |
452 | } |
453 | } | |
454 | ||
cbd56e7d | 455 | int acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 456 | { |
67298804 | 457 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a JK |
458 | int err = 0; |
459 | ||
67298804 CY |
460 | spin_lock(&im->ino_lock); |
461 | if (unlikely(im->ino_num >= sbi->max_orphans)) | |
127e670a | 462 | err = -ENOSPC; |
cbd56e7d | 463 | else |
67298804 CY |
464 | im->ino_num++; |
465 | spin_unlock(&im->ino_lock); | |
0d47c1ad | 466 | |
127e670a JK |
467 | return err; |
468 | } | |
469 | ||
cbd56e7d JK |
470 | void release_orphan_inode(struct f2fs_sb_info *sbi) |
471 | { | |
67298804 CY |
472 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
473 | ||
474 | spin_lock(&im->ino_lock); | |
475 | f2fs_bug_on(sbi, im->ino_num == 0); | |
476 | im->ino_num--; | |
477 | spin_unlock(&im->ino_lock); | |
cbd56e7d JK |
478 | } |
479 | ||
127e670a JK |
480 | void add_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
481 | { | |
39efac41 | 482 | /* add new orphan ino entry into list */ |
6451e041 | 483 | __add_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
484 | } |
485 | ||
486 | void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
487 | { | |
953e6cc6 | 488 | /* remove orphan entry from orphan list */ |
6451e041 | 489 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
490 | } |
491 | ||
8c14bfad | 492 | static int recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 493 | { |
8c14bfad CY |
494 | struct inode *inode; |
495 | ||
496 | inode = f2fs_iget(sbi->sb, ino); | |
497 | if (IS_ERR(inode)) { | |
498 | /* | |
499 | * there should be a bug that we can't find the entry | |
500 | * to orphan inode. | |
501 | */ | |
502 | f2fs_bug_on(sbi, PTR_ERR(inode) == -ENOENT); | |
503 | return PTR_ERR(inode); | |
504 | } | |
505 | ||
127e670a JK |
506 | clear_nlink(inode); |
507 | ||
508 | /* truncate all the data during iput */ | |
509 | iput(inode); | |
8c14bfad | 510 | return 0; |
127e670a JK |
511 | } |
512 | ||
8c14bfad | 513 | int recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a | 514 | { |
3c642985 | 515 | block_t start_blk, orphan_blocks, i, j; |
8c14bfad | 516 | int err; |
127e670a | 517 | |
25ca923b | 518 | if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG)) |
8c14bfad | 519 | return 0; |
127e670a | 520 | |
55141486 | 521 | start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi); |
3c642985 | 522 | orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi); |
127e670a | 523 | |
3c642985 | 524 | ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP); |
662befda | 525 | |
3c642985 | 526 | for (i = 0; i < orphan_blocks; i++) { |
127e670a JK |
527 | struct page *page = get_meta_page(sbi, start_blk + i); |
528 | struct f2fs_orphan_block *orphan_blk; | |
529 | ||
530 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); | |
531 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
532 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
8c14bfad CY |
533 | err = recover_orphan_inode(sbi, ino); |
534 | if (err) { | |
535 | f2fs_put_page(page, 1); | |
8c14bfad CY |
536 | return err; |
537 | } | |
127e670a JK |
538 | } |
539 | f2fs_put_page(page, 1); | |
540 | } | |
541 | /* clear Orphan Flag */ | |
25ca923b | 542 | clear_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG); |
8c14bfad | 543 | return 0; |
127e670a JK |
544 | } |
545 | ||
546 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
547 | { | |
502c6e0b | 548 | struct list_head *head; |
127e670a | 549 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 550 | unsigned int nentries = 0; |
bd936f84 | 551 | unsigned short index = 1; |
8c402946 | 552 | unsigned short orphan_blocks; |
4531929e | 553 | struct page *page = NULL; |
6451e041 | 554 | struct ino_entry *orphan = NULL; |
67298804 | 555 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 556 | |
67298804 | 557 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 558 | |
d6c67a4f JK |
559 | /* |
560 | * we don't need to do spin_lock(&im->ino_lock) here, since all the | |
561 | * orphan inode operations are covered under f2fs_lock_op(). | |
562 | * And, spin_lock should be avoided due to page operations below. | |
563 | */ | |
67298804 | 564 | head = &im->ino_list; |
127e670a JK |
565 | |
566 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
567 | list_for_each_entry(orphan, head, list) { |
568 | if (!page) { | |
bd936f84 | 569 | page = grab_meta_page(sbi, start_blk++); |
502c6e0b GZ |
570 | orphan_blk = |
571 | (struct f2fs_orphan_block *)page_address(page); | |
572 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
573 | } | |
127e670a | 574 | |
36795567 | 575 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 576 | |
36795567 | 577 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
578 | /* |
579 | * an orphan block is full of 1020 entries, | |
580 | * then we need to flush current orphan blocks | |
581 | * and bring another one in memory | |
582 | */ | |
583 | orphan_blk->blk_addr = cpu_to_le16(index); | |
584 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
585 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
586 | set_page_dirty(page); | |
587 | f2fs_put_page(page, 1); | |
588 | index++; | |
127e670a JK |
589 | nentries = 0; |
590 | page = NULL; | |
591 | } | |
502c6e0b | 592 | } |
127e670a | 593 | |
502c6e0b GZ |
594 | if (page) { |
595 | orphan_blk->blk_addr = cpu_to_le16(index); | |
596 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
597 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
598 | set_page_dirty(page); | |
599 | f2fs_put_page(page, 1); | |
127e670a | 600 | } |
127e670a JK |
601 | } |
602 | ||
603 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, | |
604 | block_t cp_addr, unsigned long long *version) | |
605 | { | |
606 | struct page *cp_page_1, *cp_page_2 = NULL; | |
607 | unsigned long blk_size = sbi->blocksize; | |
608 | struct f2fs_checkpoint *cp_block; | |
609 | unsigned long long cur_version = 0, pre_version = 0; | |
127e670a | 610 | size_t crc_offset; |
7e586fa0 | 611 | __u32 crc = 0; |
127e670a JK |
612 | |
613 | /* Read the 1st cp block in this CP pack */ | |
614 | cp_page_1 = get_meta_page(sbi, cp_addr); | |
615 | ||
616 | /* get the version number */ | |
617 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_1); | |
618 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
619 | if (crc_offset >= blk_size) | |
620 | goto invalid_cp1; | |
621 | ||
29e7043f | 622 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
623 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
624 | goto invalid_cp1; | |
625 | ||
d71b5564 | 626 | pre_version = cur_cp_version(cp_block); |
127e670a JK |
627 | |
628 | /* Read the 2nd cp block in this CP pack */ | |
25ca923b | 629 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
127e670a JK |
630 | cp_page_2 = get_meta_page(sbi, cp_addr); |
631 | ||
632 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_2); | |
633 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
634 | if (crc_offset >= blk_size) | |
635 | goto invalid_cp2; | |
636 | ||
29e7043f | 637 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
638 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
639 | goto invalid_cp2; | |
640 | ||
d71b5564 | 641 | cur_version = cur_cp_version(cp_block); |
127e670a JK |
642 | |
643 | if (cur_version == pre_version) { | |
644 | *version = cur_version; | |
645 | f2fs_put_page(cp_page_2, 1); | |
646 | return cp_page_1; | |
647 | } | |
648 | invalid_cp2: | |
649 | f2fs_put_page(cp_page_2, 1); | |
650 | invalid_cp1: | |
651 | f2fs_put_page(cp_page_1, 1); | |
652 | return NULL; | |
653 | } | |
654 | ||
655 | int get_valid_checkpoint(struct f2fs_sb_info *sbi) | |
656 | { | |
657 | struct f2fs_checkpoint *cp_block; | |
658 | struct f2fs_super_block *fsb = sbi->raw_super; | |
659 | struct page *cp1, *cp2, *cur_page; | |
660 | unsigned long blk_size = sbi->blocksize; | |
661 | unsigned long long cp1_version = 0, cp2_version = 0; | |
662 | unsigned long long cp_start_blk_no; | |
55141486 | 663 | unsigned int cp_blks = 1 + __cp_payload(sbi); |
1dbe4152 CL |
664 | block_t cp_blk_no; |
665 | int i; | |
127e670a | 666 | |
1dbe4152 | 667 | sbi->ckpt = kzalloc(cp_blks * blk_size, GFP_KERNEL); |
127e670a JK |
668 | if (!sbi->ckpt) |
669 | return -ENOMEM; | |
670 | /* | |
671 | * Finding out valid cp block involves read both | |
672 | * sets( cp pack1 and cp pack 2) | |
673 | */ | |
674 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
675 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
676 | ||
677 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
678 | cp_start_blk_no += ((unsigned long long)1) << |
679 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
680 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
681 | ||
682 | if (cp1 && cp2) { | |
683 | if (ver_after(cp2_version, cp1_version)) | |
684 | cur_page = cp2; | |
685 | else | |
686 | cur_page = cp1; | |
687 | } else if (cp1) { | |
688 | cur_page = cp1; | |
689 | } else if (cp2) { | |
690 | cur_page = cp2; | |
691 | } else { | |
692 | goto fail_no_cp; | |
693 | } | |
694 | ||
695 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
696 | memcpy(sbi->ckpt, cp_block, blk_size); | |
697 | ||
1dbe4152 CL |
698 | if (cp_blks <= 1) |
699 | goto done; | |
700 | ||
701 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
702 | if (cur_page == cp2) | |
703 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
704 | ||
705 | for (i = 1; i < cp_blks; i++) { | |
706 | void *sit_bitmap_ptr; | |
707 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
708 | ||
709 | cur_page = get_meta_page(sbi, cp_blk_no + i); | |
710 | sit_bitmap_ptr = page_address(cur_page); | |
711 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
712 | f2fs_put_page(cur_page, 1); | |
713 | } | |
714 | done: | |
127e670a JK |
715 | f2fs_put_page(cp1, 1); |
716 | f2fs_put_page(cp2, 1); | |
717 | return 0; | |
718 | ||
719 | fail_no_cp: | |
720 | kfree(sbi->ckpt); | |
721 | return -EINVAL; | |
722 | } | |
723 | ||
06292073 | 724 | static int __add_dirty_inode(struct inode *inode, struct inode_entry *new) |
127e670a | 725 | { |
4081363f | 726 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
127e670a | 727 | |
ed57c27f JK |
728 | if (is_inode_flag_set(F2FS_I(inode), FI_DIRTY_DIR)) |
729 | return -EEXIST; | |
2d7b822a | 730 | |
ed57c27f JK |
731 | set_inode_flag(F2FS_I(inode), FI_DIRTY_DIR); |
732 | F2FS_I(inode)->dirty_dir = new; | |
733 | list_add_tail(&new->list, &sbi->dir_inode_list); | |
dcdfff65 | 734 | stat_inc_dirty_dir(sbi); |
5deb8267 JK |
735 | return 0; |
736 | } | |
737 | ||
a7ffdbe2 | 738 | void update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 739 | { |
4081363f | 740 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
06292073 | 741 | struct inode_entry *new; |
cf0ee0f0 | 742 | int ret = 0; |
5deb8267 | 743 | |
5ac9f36f CY |
744 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
745 | !S_ISLNK(inode->i_mode)) | |
127e670a | 746 | return; |
7bd59381 | 747 | |
a7ffdbe2 JK |
748 | if (!S_ISDIR(inode->i_mode)) { |
749 | inode_inc_dirty_pages(inode); | |
750 | goto out; | |
751 | } | |
752 | ||
7bd59381 | 753 | new = f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS); |
127e670a JK |
754 | new->inode = inode; |
755 | INIT_LIST_HEAD(&new->list); | |
756 | ||
757 | spin_lock(&sbi->dir_inode_lock); | |
cf0ee0f0 | 758 | ret = __add_dirty_inode(inode, new); |
a7ffdbe2 | 759 | inode_inc_dirty_pages(inode); |
5deb8267 | 760 | spin_unlock(&sbi->dir_inode_lock); |
cf0ee0f0 CY |
761 | |
762 | if (ret) | |
763 | kmem_cache_free(inode_entry_slab, new); | |
a7ffdbe2 JK |
764 | out: |
765 | SetPagePrivate(page); | |
9e4ded3f | 766 | f2fs_trace_pid(page); |
5deb8267 JK |
767 | } |
768 | ||
769 | void add_dirty_dir_inode(struct inode *inode) | |
770 | { | |
4081363f | 771 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
06292073 | 772 | struct inode_entry *new = |
7bd59381 | 773 | f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS); |
cf0ee0f0 | 774 | int ret = 0; |
7bd59381 | 775 | |
5deb8267 JK |
776 | new->inode = inode; |
777 | INIT_LIST_HEAD(&new->list); | |
127e670a | 778 | |
5deb8267 | 779 | spin_lock(&sbi->dir_inode_lock); |
cf0ee0f0 | 780 | ret = __add_dirty_inode(inode, new); |
127e670a | 781 | spin_unlock(&sbi->dir_inode_lock); |
cf0ee0f0 CY |
782 | |
783 | if (ret) | |
784 | kmem_cache_free(inode_entry_slab, new); | |
127e670a JK |
785 | } |
786 | ||
787 | void remove_dirty_dir_inode(struct inode *inode) | |
788 | { | |
4081363f | 789 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
06292073 | 790 | struct inode_entry *entry; |
127e670a JK |
791 | |
792 | if (!S_ISDIR(inode->i_mode)) | |
793 | return; | |
794 | ||
795 | spin_lock(&sbi->dir_inode_lock); | |
a7ffdbe2 | 796 | if (get_dirty_pages(inode) || |
ed57c27f | 797 | !is_inode_flag_set(F2FS_I(inode), FI_DIRTY_DIR)) { |
3b10b1fd JK |
798 | spin_unlock(&sbi->dir_inode_lock); |
799 | return; | |
800 | } | |
127e670a | 801 | |
ed57c27f JK |
802 | entry = F2FS_I(inode)->dirty_dir; |
803 | list_del(&entry->list); | |
804 | F2FS_I(inode)->dirty_dir = NULL; | |
805 | clear_inode_flag(F2FS_I(inode), FI_DIRTY_DIR); | |
806 | stat_dec_dirty_dir(sbi); | |
127e670a | 807 | spin_unlock(&sbi->dir_inode_lock); |
ed57c27f | 808 | kmem_cache_free(inode_entry_slab, entry); |
74d0b917 JK |
809 | |
810 | /* Only from the recovery routine */ | |
afc3eda2 JK |
811 | if (is_inode_flag_set(F2FS_I(inode), FI_DELAY_IPUT)) { |
812 | clear_inode_flag(F2FS_I(inode), FI_DELAY_IPUT); | |
74d0b917 | 813 | iput(inode); |
afc3eda2 | 814 | } |
74d0b917 JK |
815 | } |
816 | ||
127e670a JK |
817 | void sync_dirty_dir_inodes(struct f2fs_sb_info *sbi) |
818 | { | |
ce3b7d80 | 819 | struct list_head *head; |
06292073 | 820 | struct inode_entry *entry; |
127e670a JK |
821 | struct inode *inode; |
822 | retry: | |
af41d3ee JK |
823 | if (unlikely(f2fs_cp_error(sbi))) |
824 | return; | |
825 | ||
127e670a | 826 | spin_lock(&sbi->dir_inode_lock); |
ce3b7d80 GZ |
827 | |
828 | head = &sbi->dir_inode_list; | |
127e670a JK |
829 | if (list_empty(head)) { |
830 | spin_unlock(&sbi->dir_inode_lock); | |
831 | return; | |
832 | } | |
06292073 | 833 | entry = list_entry(head->next, struct inode_entry, list); |
127e670a JK |
834 | inode = igrab(entry->inode); |
835 | spin_unlock(&sbi->dir_inode_lock); | |
836 | if (inode) { | |
87d6f890 | 837 | filemap_fdatawrite(inode->i_mapping); |
127e670a JK |
838 | iput(inode); |
839 | } else { | |
840 | /* | |
841 | * We should submit bio, since it exists several | |
842 | * wribacking dentry pages in the freeing inode. | |
843 | */ | |
458e6197 | 844 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
7ecebe5e | 845 | cond_resched(); |
127e670a JK |
846 | } |
847 | goto retry; | |
848 | } | |
849 | ||
0a8165d7 | 850 | /* |
127e670a JK |
851 | * Freeze all the FS-operations for checkpoint. |
852 | */ | |
cf779cab | 853 | static int block_operations(struct f2fs_sb_info *sbi) |
127e670a | 854 | { |
127e670a JK |
855 | struct writeback_control wbc = { |
856 | .sync_mode = WB_SYNC_ALL, | |
857 | .nr_to_write = LONG_MAX, | |
858 | .for_reclaim = 0, | |
859 | }; | |
c718379b | 860 | struct blk_plug plug; |
cf779cab | 861 | int err = 0; |
c718379b JK |
862 | |
863 | blk_start_plug(&plug); | |
864 | ||
39936837 | 865 | retry_flush_dents: |
e479556b | 866 | f2fs_lock_all(sbi); |
127e670a | 867 | /* write all the dirty dentry pages */ |
127e670a | 868 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 869 | f2fs_unlock_all(sbi); |
39936837 | 870 | sync_dirty_dir_inodes(sbi); |
cf779cab JK |
871 | if (unlikely(f2fs_cp_error(sbi))) { |
872 | err = -EIO; | |
873 | goto out; | |
874 | } | |
39936837 | 875 | goto retry_flush_dents; |
127e670a JK |
876 | } |
877 | ||
127e670a | 878 | /* |
e1c42045 | 879 | * POR: we should ensure that there are no dirty node pages |
127e670a JK |
880 | * until finishing nat/sit flush. |
881 | */ | |
39936837 | 882 | retry_flush_nodes: |
b3582c68 | 883 | down_write(&sbi->node_write); |
127e670a JK |
884 | |
885 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
b3582c68 | 886 | up_write(&sbi->node_write); |
39936837 | 887 | sync_node_pages(sbi, 0, &wbc); |
cf779cab JK |
888 | if (unlikely(f2fs_cp_error(sbi))) { |
889 | f2fs_unlock_all(sbi); | |
890 | err = -EIO; | |
891 | goto out; | |
892 | } | |
39936837 | 893 | goto retry_flush_nodes; |
127e670a | 894 | } |
cf779cab | 895 | out: |
c718379b | 896 | blk_finish_plug(&plug); |
cf779cab | 897 | return err; |
127e670a JK |
898 | } |
899 | ||
900 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
901 | { | |
b3582c68 | 902 | up_write(&sbi->node_write); |
e479556b | 903 | f2fs_unlock_all(sbi); |
127e670a JK |
904 | } |
905 | ||
fb51b5ef CL |
906 | static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi) |
907 | { | |
908 | DEFINE_WAIT(wait); | |
909 | ||
910 | for (;;) { | |
911 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
912 | ||
913 | if (!get_pages(sbi, F2FS_WRITEBACK)) | |
914 | break; | |
915 | ||
916 | io_schedule(); | |
917 | } | |
918 | finish_wait(&sbi->cp_wait, &wait); | |
919 | } | |
920 | ||
75ab4cb8 | 921 | static void do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
922 | { |
923 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
cf2271e7 | 924 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); |
77041823 | 925 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
67298804 | 926 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; |
77041823 | 927 | nid_t last_nid = nm_i->next_scan_nid; |
127e670a | 928 | block_t start_blk; |
127e670a | 929 | unsigned int data_sum_blocks, orphan_blocks; |
7e586fa0 | 930 | __u32 crc32 = 0; |
127e670a | 931 | int i; |
55141486 | 932 | int cp_payload_blks = __cp_payload(sbi); |
e90c2d28 CY |
933 | block_t discard_blk = NEXT_FREE_BLKADDR(sbi, curseg); |
934 | bool invalidate = false; | |
127e670a | 935 | |
1e87a78d JK |
936 | /* |
937 | * This avoids to conduct wrong roll-forward operations and uses | |
938 | * metapages, so should be called prior to sync_meta_pages below. | |
939 | */ | |
e90c2d28 CY |
940 | if (discard_next_dnode(sbi, discard_blk)) |
941 | invalidate = true; | |
127e670a JK |
942 | |
943 | /* Flush all the NAT/SIT pages */ | |
cf779cab | 944 | while (get_pages(sbi, F2FS_DIRTY_META)) { |
127e670a | 945 | sync_meta_pages(sbi, META, LONG_MAX); |
cf779cab JK |
946 | if (unlikely(f2fs_cp_error(sbi))) |
947 | return; | |
948 | } | |
127e670a JK |
949 | |
950 | next_free_nid(sbi, &last_nid); | |
951 | ||
952 | /* | |
953 | * modify checkpoint | |
954 | * version number is already updated | |
955 | */ | |
956 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi)); | |
957 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
958 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); | |
b5b82205 | 959 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
960 | ckpt->cur_node_segno[i] = |
961 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
962 | ckpt->cur_node_blkoff[i] = | |
963 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
964 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
965 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
966 | } | |
b5b82205 | 967 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
968 | ckpt->cur_data_segno[i] = |
969 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
970 | ckpt->cur_data_blkoff[i] = | |
971 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
972 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
973 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
974 | } | |
975 | ||
976 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
977 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
978 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
979 | ||
980 | /* 2 cp + n data seg summary + orphan inode blocks */ | |
3fa06d7b | 981 | data_sum_blocks = npages_for_summary_flush(sbi, false); |
b5b82205 | 982 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
25ca923b | 983 | set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 984 | else |
25ca923b | 985 | clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 986 | |
67298804 | 987 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
988 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
989 | orphan_blocks); | |
127e670a | 990 | |
119ee914 | 991 | if (__remain_node_summaries(cpc->reason)) |
b5b82205 | 992 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
993 | cp_payload_blks + data_sum_blocks + |
994 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
119ee914 | 995 | else |
b5b82205 | 996 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS + |
1dbe4152 CL |
997 | cp_payload_blks + data_sum_blocks + |
998 | orphan_blocks); | |
119ee914 JK |
999 | |
1000 | if (cpc->reason == CP_UMOUNT) | |
1001 | set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
1002 | else | |
1003 | clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
1004 | ||
1005 | if (cpc->reason == CP_FASTBOOT) | |
1006 | set_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
1007 | else | |
1008 | clear_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
127e670a | 1009 | |
67298804 | 1010 | if (orphan_num) |
25ca923b | 1011 | set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1012 | else |
25ca923b | 1013 | clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1014 | |
caf0047e | 1015 | if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) |
2ae4c673 JK |
1016 | set_ckpt_flags(ckpt, CP_FSCK_FLAG); |
1017 | ||
127e670a JK |
1018 | /* update SIT/NAT bitmap */ |
1019 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
1020 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
1021 | ||
1022 | crc32 = f2fs_crc32(ckpt, le32_to_cpu(ckpt->checksum_offset)); | |
7e586fa0 JK |
1023 | *((__le32 *)((unsigned char *)ckpt + |
1024 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
1025 | = cpu_to_le32(crc32); |
1026 | ||
1027 | start_blk = __start_cp_addr(sbi); | |
1028 | ||
a7230d16 JK |
1029 | /* need to wait for end_io results */ |
1030 | wait_on_all_pages_writeback(sbi); | |
1031 | if (unlikely(f2fs_cp_error(sbi))) | |
1032 | return; | |
1033 | ||
127e670a | 1034 | /* write out checkpoint buffer at block 0 */ |
381722d2 CY |
1035 | update_meta_page(sbi, ckpt, start_blk++); |
1036 | ||
1037 | for (i = 1; i < 1 + cp_payload_blks; i++) | |
1038 | update_meta_page(sbi, (char *)ckpt + i * F2FS_BLKSIZE, | |
1039 | start_blk++); | |
1dbe4152 | 1040 | |
67298804 | 1041 | if (orphan_num) { |
127e670a JK |
1042 | write_orphan_inodes(sbi, start_blk); |
1043 | start_blk += orphan_blocks; | |
1044 | } | |
1045 | ||
1046 | write_data_summaries(sbi, start_blk); | |
1047 | start_blk += data_sum_blocks; | |
119ee914 | 1048 | if (__remain_node_summaries(cpc->reason)) { |
127e670a JK |
1049 | write_node_summaries(sbi, start_blk); |
1050 | start_blk += NR_CURSEG_NODE_TYPE; | |
1051 | } | |
1052 | ||
1053 | /* writeout checkpoint block */ | |
381722d2 | 1054 | update_meta_page(sbi, ckpt, start_blk); |
127e670a JK |
1055 | |
1056 | /* wait for previous submitted node/meta pages writeback */ | |
fb51b5ef | 1057 | wait_on_all_pages_writeback(sbi); |
127e670a | 1058 | |
cf779cab JK |
1059 | if (unlikely(f2fs_cp_error(sbi))) |
1060 | return; | |
1061 | ||
4ef51a8f | 1062 | filemap_fdatawait_range(NODE_MAPPING(sbi), 0, LONG_MAX); |
9df27d98 | 1063 | filemap_fdatawait_range(META_MAPPING(sbi), 0, LONG_MAX); |
127e670a JK |
1064 | |
1065 | /* update user_block_counts */ | |
1066 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
1067 | sbi->alloc_valid_block_count = 0; | |
1068 | ||
1069 | /* Here, we only have one bio having CP pack */ | |
577e3495 | 1070 | sync_meta_pages(sbi, META_FLUSH, LONG_MAX); |
127e670a | 1071 | |
6a8f8ca5 JK |
1072 | /* wait for previous submitted meta pages writeback */ |
1073 | wait_on_all_pages_writeback(sbi); | |
1074 | ||
e90c2d28 CY |
1075 | /* |
1076 | * invalidate meta page which is used temporarily for zeroing out | |
1077 | * block at the end of warm node chain. | |
1078 | */ | |
1079 | if (invalidate) | |
1080 | invalidate_mapping_pages(META_MAPPING(sbi), discard_blk, | |
1081 | discard_blk); | |
1082 | ||
cf779cab JK |
1083 | release_dirty_inode(sbi); |
1084 | ||
1085 | if (unlikely(f2fs_cp_error(sbi))) | |
1086 | return; | |
1087 | ||
836b5a63 | 1088 | clear_prefree_segments(sbi, cpc); |
caf0047e | 1089 | clear_sbi_flag(sbi, SBI_IS_DIRTY); |
127e670a JK |
1090 | } |
1091 | ||
0a8165d7 | 1092 | /* |
e1c42045 | 1093 | * We guarantee that this checkpoint procedure will not fail. |
127e670a | 1094 | */ |
75ab4cb8 | 1095 | void write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1096 | { |
1097 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1098 | unsigned long long ckpt_ver; | |
1099 | ||
43727527 | 1100 | mutex_lock(&sbi->cp_mutex); |
8501017e | 1101 | |
caf0047e | 1102 | if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) && |
a66cdd98 JK |
1103 | (cpc->reason == CP_FASTBOOT || cpc->reason == CP_SYNC || |
1104 | (cpc->reason == CP_DISCARD && !sbi->discard_blks))) | |
8501017e | 1105 | goto out; |
cf779cab JK |
1106 | if (unlikely(f2fs_cp_error(sbi))) |
1107 | goto out; | |
11504a8e JK |
1108 | if (f2fs_readonly(sbi->sb)) |
1109 | goto out; | |
2bda542d WL |
1110 | |
1111 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); | |
1112 | ||
cf779cab JK |
1113 | if (block_operations(sbi)) |
1114 | goto out; | |
127e670a | 1115 | |
75ab4cb8 | 1116 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1117 | |
458e6197 JK |
1118 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
1119 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
1120 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
127e670a JK |
1121 | |
1122 | /* | |
1123 | * update checkpoint pack index | |
1124 | * Increase the version number so that | |
1125 | * SIT entries and seg summaries are written at correct place | |
1126 | */ | |
d71b5564 | 1127 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1128 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1129 | ||
1130 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
1131 | flush_nat_entries(sbi); | |
4b2fecc8 | 1132 | flush_sit_entries(sbi, cpc); |
127e670a | 1133 | |
127e670a | 1134 | /* unlock all the fs_lock[] in do_checkpoint() */ |
75ab4cb8 | 1135 | do_checkpoint(sbi, cpc); |
127e670a JK |
1136 | |
1137 | unblock_operations(sbi); | |
942e0be6 | 1138 | stat_inc_cp_count(sbi->stat_info); |
10027551 JK |
1139 | |
1140 | if (cpc->reason == CP_RECOVERY) | |
1141 | f2fs_msg(sbi->sb, KERN_NOTICE, | |
1142 | "checkpoint: version = %llx", ckpt_ver); | |
60b99b48 JK |
1143 | |
1144 | /* do checkpoint periodically */ | |
1145 | sbi->cp_expires = round_jiffies_up(jiffies + HZ * sbi->cp_interval); | |
8501017e JK |
1146 | out: |
1147 | mutex_unlock(&sbi->cp_mutex); | |
75ab4cb8 | 1148 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
127e670a JK |
1149 | } |
1150 | ||
6451e041 | 1151 | void init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1152 | { |
6451e041 JK |
1153 | int i; |
1154 | ||
1155 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1156 | struct inode_management *im = &sbi->im[i]; |
1157 | ||
1158 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1159 | spin_lock_init(&im->ino_lock); | |
1160 | INIT_LIST_HEAD(&im->ino_list); | |
1161 | im->ino_num = 0; | |
6451e041 JK |
1162 | } |
1163 | ||
b5b82205 | 1164 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
14b42817 WL |
1165 | NR_CURSEG_TYPE - __cp_payload(sbi)) * |
1166 | F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
1167 | } |
1168 | ||
6e6093a8 | 1169 | int __init create_checkpoint_caches(void) |
127e670a | 1170 | { |
6451e041 JK |
1171 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1172 | sizeof(struct ino_entry)); | |
1173 | if (!ino_entry_slab) | |
127e670a | 1174 | return -ENOMEM; |
06292073 CY |
1175 | inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry", |
1176 | sizeof(struct inode_entry)); | |
6bacf52f | 1177 | if (!inode_entry_slab) { |
6451e041 | 1178 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1179 | return -ENOMEM; |
1180 | } | |
1181 | return 0; | |
1182 | } | |
1183 | ||
1184 | void destroy_checkpoint_caches(void) | |
1185 | { | |
6451e041 | 1186 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1187 | kmem_cache_destroy(inode_entry_slab); |
1188 | } |