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 | */ | |
50 | struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
51 | { | |
9df27d98 | 52 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a | 53 | struct page *page; |
cf04e8eb | 54 | struct f2fs_io_info fio = { |
05ca3632 | 55 | .sbi = sbi, |
cf04e8eb JK |
56 | .type = META, |
57 | .rw = READ_SYNC | REQ_META | REQ_PRIO, | |
58 | .blk_addr = index, | |
4375a336 | 59 | .encrypted_page = NULL, |
cf04e8eb | 60 | }; |
127e670a JK |
61 | repeat: |
62 | page = grab_cache_page(mapping, index); | |
63 | if (!page) { | |
64 | cond_resched(); | |
65 | goto repeat; | |
66 | } | |
393ff91f JK |
67 | if (PageUptodate(page)) |
68 | goto out; | |
69 | ||
05ca3632 JK |
70 | fio.page = page; |
71 | ||
86531d6b JK |
72 | if (f2fs_submit_page_bio(&fio)) { |
73 | f2fs_put_page(page, 1); | |
127e670a | 74 | goto repeat; |
86531d6b | 75 | } |
127e670a | 76 | |
393ff91f | 77 | lock_page(page); |
6bacf52f | 78 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
79 | f2fs_put_page(page, 1); |
80 | goto repeat; | |
81 | } | |
393ff91f | 82 | out: |
127e670a JK |
83 | return page; |
84 | } | |
85 | ||
f0c9cada | 86 | bool is_valid_blkaddr(struct f2fs_sb_info *sbi, block_t blkaddr, int type) |
662befda CY |
87 | { |
88 | switch (type) { | |
89 | case META_NAT: | |
66b00c18 | 90 | break; |
662befda | 91 | case META_SIT: |
66b00c18 CY |
92 | if (unlikely(blkaddr >= SIT_BLK_CNT(sbi))) |
93 | return false; | |
94 | break; | |
81c1a0f1 | 95 | case META_SSA: |
66b00c18 CY |
96 | if (unlikely(blkaddr >= MAIN_BLKADDR(sbi) || |
97 | blkaddr < SM_I(sbi)->ssa_blkaddr)) | |
98 | return false; | |
99 | break; | |
662befda | 100 | case META_CP: |
66b00c18 CY |
101 | if (unlikely(blkaddr >= SIT_I(sbi)->sit_base_addr || |
102 | blkaddr < __start_cp_addr(sbi))) | |
103 | return false; | |
104 | break; | |
4c521f49 | 105 | case META_POR: |
66b00c18 CY |
106 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
107 | blkaddr < MAIN_BLKADDR(sbi))) | |
108 | return false; | |
109 | break; | |
662befda CY |
110 | default: |
111 | BUG(); | |
112 | } | |
66b00c18 CY |
113 | |
114 | return true; | |
662befda CY |
115 | } |
116 | ||
117 | /* | |
81c1a0f1 | 118 | * Readahead CP/NAT/SIT/SSA pages |
662befda | 119 | */ |
4c521f49 | 120 | int ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages, int type) |
662befda CY |
121 | { |
122 | block_t prev_blk_addr = 0; | |
123 | struct page *page; | |
4c521f49 | 124 | block_t blkno = start; |
662befda | 125 | struct f2fs_io_info fio = { |
05ca3632 | 126 | .sbi = sbi, |
662befda | 127 | .type = META, |
4375a336 JK |
128 | .rw = READ_SYNC | REQ_META | REQ_PRIO, |
129 | .encrypted_page = NULL, | |
662befda CY |
130 | }; |
131 | ||
132 | for (; nrpages-- > 0; blkno++) { | |
662befda | 133 | |
66b00c18 CY |
134 | if (!is_valid_blkaddr(sbi, blkno, type)) |
135 | goto out; | |
136 | ||
662befda CY |
137 | switch (type) { |
138 | case META_NAT: | |
66b00c18 CY |
139 | if (unlikely(blkno >= |
140 | NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))) | |
662befda | 141 | blkno = 0; |
66b00c18 | 142 | /* get nat block addr */ |
cf04e8eb | 143 | fio.blk_addr = current_nat_addr(sbi, |
662befda CY |
144 | blkno * NAT_ENTRY_PER_BLOCK); |
145 | break; | |
146 | case META_SIT: | |
147 | /* get sit block addr */ | |
cf04e8eb | 148 | fio.blk_addr = current_sit_addr(sbi, |
662befda | 149 | blkno * SIT_ENTRY_PER_BLOCK); |
cf04e8eb | 150 | if (blkno != start && prev_blk_addr + 1 != fio.blk_addr) |
662befda | 151 | goto out; |
cf04e8eb | 152 | prev_blk_addr = fio.blk_addr; |
662befda | 153 | break; |
81c1a0f1 | 154 | case META_SSA: |
662befda | 155 | case META_CP: |
4c521f49 | 156 | case META_POR: |
cf04e8eb | 157 | fio.blk_addr = blkno; |
662befda CY |
158 | break; |
159 | default: | |
160 | BUG(); | |
161 | } | |
162 | ||
cf04e8eb | 163 | page = grab_cache_page(META_MAPPING(sbi), fio.blk_addr); |
662befda CY |
164 | if (!page) |
165 | continue; | |
166 | if (PageUptodate(page)) { | |
662befda CY |
167 | f2fs_put_page(page, 1); |
168 | continue; | |
169 | } | |
170 | ||
05ca3632 JK |
171 | fio.page = page; |
172 | f2fs_submit_page_mbio(&fio); | |
662befda CY |
173 | f2fs_put_page(page, 0); |
174 | } | |
175 | out: | |
176 | f2fs_submit_merged_bio(sbi, META, READ); | |
177 | return blkno - start; | |
178 | } | |
179 | ||
635aee1f CY |
180 | void ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index) |
181 | { | |
182 | struct page *page; | |
183 | bool readahead = false; | |
184 | ||
185 | page = find_get_page(META_MAPPING(sbi), index); | |
186 | if (!page || (page && !PageUptodate(page))) | |
187 | readahead = true; | |
188 | f2fs_put_page(page, 0); | |
189 | ||
190 | if (readahead) | |
191 | ra_meta_pages(sbi, index, MAX_BIO_BLOCKS(sbi), META_POR); | |
192 | } | |
193 | ||
127e670a JK |
194 | static int f2fs_write_meta_page(struct page *page, |
195 | struct writeback_control *wbc) | |
196 | { | |
4081363f | 197 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 198 | |
ecda0de3 CY |
199 | trace_f2fs_writepage(page, META); |
200 | ||
caf0047e | 201 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
cfb271d4 | 202 | goto redirty_out; |
857dc4e0 | 203 | if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0)) |
cfb271d4 | 204 | goto redirty_out; |
1e968fdf | 205 | if (unlikely(f2fs_cp_error(sbi))) |
cf779cab | 206 | goto redirty_out; |
127e670a | 207 | |
3cb5ad15 | 208 | f2fs_wait_on_page_writeback(page, META); |
577e3495 JK |
209 | write_meta_page(sbi, page); |
210 | dec_page_count(sbi, F2FS_DIRTY_META); | |
211 | unlock_page(page); | |
857dc4e0 JK |
212 | |
213 | if (wbc->for_reclaim) | |
214 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
577e3495 | 215 | return 0; |
cfb271d4 CY |
216 | |
217 | redirty_out: | |
76f60268 | 218 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 219 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
220 | } |
221 | ||
222 | static int f2fs_write_meta_pages(struct address_space *mapping, | |
223 | struct writeback_control *wbc) | |
224 | { | |
4081363f | 225 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 226 | long diff, written; |
127e670a | 227 | |
e5748434 CY |
228 | trace_f2fs_writepages(mapping->host, wbc, META); |
229 | ||
5459aa97 | 230 | /* collect a number of dirty meta pages and write together */ |
50c8cdb3 JK |
231 | if (wbc->for_kupdate || |
232 | get_pages(sbi, F2FS_DIRTY_META) < nr_pages_to_skip(sbi, META)) | |
d3baf95d | 233 | goto skip_write; |
127e670a JK |
234 | |
235 | /* if mounting is failed, skip writing node pages */ | |
236 | mutex_lock(&sbi->cp_mutex); | |
50c8cdb3 JK |
237 | diff = nr_pages_to_write(sbi, META, wbc); |
238 | written = sync_meta_pages(sbi, META, wbc->nr_to_write); | |
127e670a | 239 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 240 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 241 | return 0; |
d3baf95d JK |
242 | |
243 | skip_write: | |
244 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
245 | return 0; | |
127e670a JK |
246 | } |
247 | ||
248 | long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, | |
249 | long nr_to_write) | |
250 | { | |
9df27d98 | 251 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
252 | pgoff_t index = 0, end = LONG_MAX; |
253 | struct pagevec pvec; | |
254 | long nwritten = 0; | |
255 | struct writeback_control wbc = { | |
256 | .for_reclaim = 0, | |
257 | }; | |
258 | ||
259 | pagevec_init(&pvec, 0); | |
260 | ||
261 | while (index <= end) { | |
262 | int i, nr_pages; | |
263 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, | |
264 | PAGECACHE_TAG_DIRTY, | |
265 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
cfb271d4 | 266 | if (unlikely(nr_pages == 0)) |
127e670a JK |
267 | break; |
268 | ||
269 | for (i = 0; i < nr_pages; i++) { | |
270 | struct page *page = pvec.pages[i]; | |
203681f6 | 271 | |
127e670a | 272 | lock_page(page); |
203681f6 JK |
273 | |
274 | if (unlikely(page->mapping != mapping)) { | |
275 | continue_unlock: | |
276 | unlock_page(page); | |
277 | continue; | |
278 | } | |
279 | if (!PageDirty(page)) { | |
280 | /* someone wrote it for us */ | |
281 | goto continue_unlock; | |
282 | } | |
283 | ||
284 | if (!clear_page_dirty_for_io(page)) | |
285 | goto continue_unlock; | |
286 | ||
97dc3fd2 | 287 | if (mapping->a_ops->writepage(page, &wbc)) { |
577e3495 JK |
288 | unlock_page(page); |
289 | break; | |
290 | } | |
cfb271d4 CY |
291 | nwritten++; |
292 | if (unlikely(nwritten >= nr_to_write)) | |
127e670a JK |
293 | break; |
294 | } | |
295 | pagevec_release(&pvec); | |
296 | cond_resched(); | |
297 | } | |
298 | ||
299 | if (nwritten) | |
458e6197 | 300 | f2fs_submit_merged_bio(sbi, type, WRITE); |
127e670a JK |
301 | |
302 | return nwritten; | |
303 | } | |
304 | ||
305 | static int f2fs_set_meta_page_dirty(struct page *page) | |
306 | { | |
26c6b887 JK |
307 | trace_f2fs_set_page_dirty(page, META); |
308 | ||
127e670a JK |
309 | SetPageUptodate(page); |
310 | if (!PageDirty(page)) { | |
311 | __set_page_dirty_nobuffers(page); | |
4081363f | 312 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); |
1601839e | 313 | SetPagePrivate(page); |
9e4ded3f | 314 | f2fs_trace_pid(page); |
127e670a JK |
315 | return 1; |
316 | } | |
317 | return 0; | |
318 | } | |
319 | ||
320 | const struct address_space_operations f2fs_meta_aops = { | |
321 | .writepage = f2fs_write_meta_page, | |
322 | .writepages = f2fs_write_meta_pages, | |
323 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
487261f3 CY |
324 | .invalidatepage = f2fs_invalidate_page, |
325 | .releasepage = f2fs_release_page, | |
127e670a JK |
326 | }; |
327 | ||
6451e041 | 328 | static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 329 | { |
67298804 | 330 | struct inode_management *im = &sbi->im[type]; |
39efac41 JK |
331 | struct ino_entry *e; |
332 | retry: | |
769ec6e5 JK |
333 | if (radix_tree_preload(GFP_NOFS)) { |
334 | cond_resched(); | |
335 | goto retry; | |
336 | } | |
337 | ||
67298804 | 338 | spin_lock(&im->ino_lock); |
39efac41 | 339 | |
67298804 | 340 | e = radix_tree_lookup(&im->ino_root, ino); |
39efac41 JK |
341 | if (!e) { |
342 | e = kmem_cache_alloc(ino_entry_slab, GFP_ATOMIC); | |
343 | if (!e) { | |
67298804 | 344 | spin_unlock(&im->ino_lock); |
769ec6e5 | 345 | radix_tree_preload_end(); |
39efac41 | 346 | goto retry; |
953e6cc6 | 347 | } |
67298804 CY |
348 | if (radix_tree_insert(&im->ino_root, ino, e)) { |
349 | spin_unlock(&im->ino_lock); | |
39efac41 | 350 | kmem_cache_free(ino_entry_slab, e); |
769ec6e5 | 351 | radix_tree_preload_end(); |
39efac41 JK |
352 | goto retry; |
353 | } | |
354 | memset(e, 0, sizeof(struct ino_entry)); | |
355 | e->ino = ino; | |
953e6cc6 | 356 | |
67298804 | 357 | list_add_tail(&e->list, &im->ino_list); |
8c402946 | 358 | if (type != ORPHAN_INO) |
67298804 | 359 | im->ino_num++; |
39efac41 | 360 | } |
67298804 | 361 | spin_unlock(&im->ino_lock); |
769ec6e5 | 362 | radix_tree_preload_end(); |
953e6cc6 JK |
363 | } |
364 | ||
6451e041 | 365 | static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 366 | { |
67298804 | 367 | struct inode_management *im = &sbi->im[type]; |
6451e041 | 368 | struct ino_entry *e; |
953e6cc6 | 369 | |
67298804 CY |
370 | spin_lock(&im->ino_lock); |
371 | e = radix_tree_lookup(&im->ino_root, ino); | |
39efac41 JK |
372 | if (e) { |
373 | list_del(&e->list); | |
67298804 CY |
374 | radix_tree_delete(&im->ino_root, ino); |
375 | im->ino_num--; | |
376 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
377 | kmem_cache_free(ino_entry_slab, e); |
378 | return; | |
953e6cc6 | 379 | } |
67298804 | 380 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
381 | } |
382 | ||
fff04f90 JK |
383 | void add_dirty_inode(struct f2fs_sb_info *sbi, nid_t ino, int type) |
384 | { | |
385 | /* add new dirty ino entry into list */ | |
386 | __add_ino_entry(sbi, ino, type); | |
387 | } | |
388 | ||
389 | void remove_dirty_inode(struct f2fs_sb_info *sbi, nid_t ino, int type) | |
390 | { | |
391 | /* remove dirty ino entry from list */ | |
392 | __remove_ino_entry(sbi, ino, type); | |
393 | } | |
394 | ||
395 | /* mode should be APPEND_INO or UPDATE_INO */ | |
396 | bool exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode) | |
397 | { | |
67298804 | 398 | struct inode_management *im = &sbi->im[mode]; |
fff04f90 | 399 | struct ino_entry *e; |
67298804 CY |
400 | |
401 | spin_lock(&im->ino_lock); | |
402 | e = radix_tree_lookup(&im->ino_root, ino); | |
403 | spin_unlock(&im->ino_lock); | |
fff04f90 JK |
404 | return e ? true : false; |
405 | } | |
406 | ||
6f12ac25 | 407 | void release_dirty_inode(struct f2fs_sb_info *sbi) |
fff04f90 JK |
408 | { |
409 | struct ino_entry *e, *tmp; | |
410 | int i; | |
411 | ||
412 | for (i = APPEND_INO; i <= UPDATE_INO; i++) { | |
67298804 CY |
413 | struct inode_management *im = &sbi->im[i]; |
414 | ||
415 | spin_lock(&im->ino_lock); | |
416 | list_for_each_entry_safe(e, tmp, &im->ino_list, list) { | |
fff04f90 | 417 | list_del(&e->list); |
67298804 | 418 | radix_tree_delete(&im->ino_root, e->ino); |
fff04f90 | 419 | kmem_cache_free(ino_entry_slab, e); |
67298804 | 420 | im->ino_num--; |
fff04f90 | 421 | } |
67298804 | 422 | spin_unlock(&im->ino_lock); |
fff04f90 JK |
423 | } |
424 | } | |
425 | ||
cbd56e7d | 426 | int acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 427 | { |
67298804 | 428 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a JK |
429 | int err = 0; |
430 | ||
67298804 CY |
431 | spin_lock(&im->ino_lock); |
432 | if (unlikely(im->ino_num >= sbi->max_orphans)) | |
127e670a | 433 | err = -ENOSPC; |
cbd56e7d | 434 | else |
67298804 CY |
435 | im->ino_num++; |
436 | spin_unlock(&im->ino_lock); | |
0d47c1ad | 437 | |
127e670a JK |
438 | return err; |
439 | } | |
440 | ||
cbd56e7d JK |
441 | void release_orphan_inode(struct f2fs_sb_info *sbi) |
442 | { | |
67298804 CY |
443 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
444 | ||
445 | spin_lock(&im->ino_lock); | |
446 | f2fs_bug_on(sbi, im->ino_num == 0); | |
447 | im->ino_num--; | |
448 | spin_unlock(&im->ino_lock); | |
cbd56e7d JK |
449 | } |
450 | ||
127e670a JK |
451 | void add_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
452 | { | |
39efac41 | 453 | /* add new orphan ino entry into list */ |
6451e041 | 454 | __add_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
455 | } |
456 | ||
457 | void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
458 | { | |
953e6cc6 | 459 | /* remove orphan entry from orphan list */ |
6451e041 | 460 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
461 | } |
462 | ||
463 | static void recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
464 | { | |
465 | struct inode *inode = f2fs_iget(sbi->sb, ino); | |
9850cf4a | 466 | f2fs_bug_on(sbi, IS_ERR(inode)); |
127e670a JK |
467 | clear_nlink(inode); |
468 | ||
469 | /* truncate all the data during iput */ | |
470 | iput(inode); | |
471 | } | |
472 | ||
8f99a946 | 473 | void recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a | 474 | { |
3c642985 | 475 | block_t start_blk, orphan_blocks, i, j; |
127e670a | 476 | |
25ca923b | 477 | if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG)) |
8f99a946 | 478 | return; |
127e670a | 479 | |
caf0047e | 480 | set_sbi_flag(sbi, SBI_POR_DOING); |
1dbe4152 | 481 | |
55141486 | 482 | start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi); |
3c642985 | 483 | orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi); |
127e670a | 484 | |
3c642985 | 485 | ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP); |
662befda | 486 | |
3c642985 | 487 | for (i = 0; i < orphan_blocks; i++) { |
127e670a JK |
488 | struct page *page = get_meta_page(sbi, start_blk + i); |
489 | struct f2fs_orphan_block *orphan_blk; | |
490 | ||
491 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); | |
492 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
493 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
494 | recover_orphan_inode(sbi, ino); | |
495 | } | |
496 | f2fs_put_page(page, 1); | |
497 | } | |
498 | /* clear Orphan Flag */ | |
25ca923b | 499 | clear_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG); |
caf0047e | 500 | clear_sbi_flag(sbi, SBI_POR_DOING); |
8f99a946 | 501 | return; |
127e670a JK |
502 | } |
503 | ||
504 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
505 | { | |
502c6e0b | 506 | struct list_head *head; |
127e670a | 507 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 508 | unsigned int nentries = 0; |
bd936f84 | 509 | unsigned short index = 1; |
8c402946 | 510 | unsigned short orphan_blocks; |
4531929e | 511 | struct page *page = NULL; |
6451e041 | 512 | struct ino_entry *orphan = NULL; |
67298804 | 513 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 514 | |
67298804 | 515 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 516 | |
d6c67a4f JK |
517 | /* |
518 | * we don't need to do spin_lock(&im->ino_lock) here, since all the | |
519 | * orphan inode operations are covered under f2fs_lock_op(). | |
520 | * And, spin_lock should be avoided due to page operations below. | |
521 | */ | |
67298804 | 522 | head = &im->ino_list; |
127e670a JK |
523 | |
524 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
525 | list_for_each_entry(orphan, head, list) { |
526 | if (!page) { | |
bd936f84 | 527 | page = grab_meta_page(sbi, start_blk++); |
502c6e0b GZ |
528 | orphan_blk = |
529 | (struct f2fs_orphan_block *)page_address(page); | |
530 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
531 | } | |
127e670a | 532 | |
36795567 | 533 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 534 | |
36795567 | 535 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
536 | /* |
537 | * an orphan block is full of 1020 entries, | |
538 | * then we need to flush current orphan blocks | |
539 | * and bring another one in memory | |
540 | */ | |
541 | orphan_blk->blk_addr = cpu_to_le16(index); | |
542 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
543 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
544 | set_page_dirty(page); | |
545 | f2fs_put_page(page, 1); | |
546 | index++; | |
127e670a JK |
547 | nentries = 0; |
548 | page = NULL; | |
549 | } | |
502c6e0b | 550 | } |
127e670a | 551 | |
502c6e0b GZ |
552 | if (page) { |
553 | orphan_blk->blk_addr = cpu_to_le16(index); | |
554 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
555 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
556 | set_page_dirty(page); | |
557 | f2fs_put_page(page, 1); | |
127e670a | 558 | } |
127e670a JK |
559 | } |
560 | ||
561 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, | |
562 | block_t cp_addr, unsigned long long *version) | |
563 | { | |
564 | struct page *cp_page_1, *cp_page_2 = NULL; | |
565 | unsigned long blk_size = sbi->blocksize; | |
566 | struct f2fs_checkpoint *cp_block; | |
567 | unsigned long long cur_version = 0, pre_version = 0; | |
127e670a | 568 | size_t crc_offset; |
7e586fa0 | 569 | __u32 crc = 0; |
127e670a JK |
570 | |
571 | /* Read the 1st cp block in this CP pack */ | |
572 | cp_page_1 = get_meta_page(sbi, cp_addr); | |
573 | ||
574 | /* get the version number */ | |
575 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_1); | |
576 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
577 | if (crc_offset >= blk_size) | |
578 | goto invalid_cp1; | |
579 | ||
29e7043f | 580 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
581 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
582 | goto invalid_cp1; | |
583 | ||
d71b5564 | 584 | pre_version = cur_cp_version(cp_block); |
127e670a JK |
585 | |
586 | /* Read the 2nd cp block in this CP pack */ | |
25ca923b | 587 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
127e670a JK |
588 | cp_page_2 = get_meta_page(sbi, cp_addr); |
589 | ||
590 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_2); | |
591 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
592 | if (crc_offset >= blk_size) | |
593 | goto invalid_cp2; | |
594 | ||
29e7043f | 595 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
596 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
597 | goto invalid_cp2; | |
598 | ||
d71b5564 | 599 | cur_version = cur_cp_version(cp_block); |
127e670a JK |
600 | |
601 | if (cur_version == pre_version) { | |
602 | *version = cur_version; | |
603 | f2fs_put_page(cp_page_2, 1); | |
604 | return cp_page_1; | |
605 | } | |
606 | invalid_cp2: | |
607 | f2fs_put_page(cp_page_2, 1); | |
608 | invalid_cp1: | |
609 | f2fs_put_page(cp_page_1, 1); | |
610 | return NULL; | |
611 | } | |
612 | ||
613 | int get_valid_checkpoint(struct f2fs_sb_info *sbi) | |
614 | { | |
615 | struct f2fs_checkpoint *cp_block; | |
616 | struct f2fs_super_block *fsb = sbi->raw_super; | |
617 | struct page *cp1, *cp2, *cur_page; | |
618 | unsigned long blk_size = sbi->blocksize; | |
619 | unsigned long long cp1_version = 0, cp2_version = 0; | |
620 | unsigned long long cp_start_blk_no; | |
55141486 | 621 | unsigned int cp_blks = 1 + __cp_payload(sbi); |
1dbe4152 CL |
622 | block_t cp_blk_no; |
623 | int i; | |
127e670a | 624 | |
1dbe4152 | 625 | sbi->ckpt = kzalloc(cp_blks * blk_size, GFP_KERNEL); |
127e670a JK |
626 | if (!sbi->ckpt) |
627 | return -ENOMEM; | |
628 | /* | |
629 | * Finding out valid cp block involves read both | |
630 | * sets( cp pack1 and cp pack 2) | |
631 | */ | |
632 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
633 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
634 | ||
635 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
636 | cp_start_blk_no += ((unsigned long long)1) << |
637 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
638 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
639 | ||
640 | if (cp1 && cp2) { | |
641 | if (ver_after(cp2_version, cp1_version)) | |
642 | cur_page = cp2; | |
643 | else | |
644 | cur_page = cp1; | |
645 | } else if (cp1) { | |
646 | cur_page = cp1; | |
647 | } else if (cp2) { | |
648 | cur_page = cp2; | |
649 | } else { | |
650 | goto fail_no_cp; | |
651 | } | |
652 | ||
653 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
654 | memcpy(sbi->ckpt, cp_block, blk_size); | |
655 | ||
1dbe4152 CL |
656 | if (cp_blks <= 1) |
657 | goto done; | |
658 | ||
659 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
660 | if (cur_page == cp2) | |
661 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
662 | ||
663 | for (i = 1; i < cp_blks; i++) { | |
664 | void *sit_bitmap_ptr; | |
665 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
666 | ||
667 | cur_page = get_meta_page(sbi, cp_blk_no + i); | |
668 | sit_bitmap_ptr = page_address(cur_page); | |
669 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
670 | f2fs_put_page(cur_page, 1); | |
671 | } | |
672 | done: | |
127e670a JK |
673 | f2fs_put_page(cp1, 1); |
674 | f2fs_put_page(cp2, 1); | |
675 | return 0; | |
676 | ||
677 | fail_no_cp: | |
678 | kfree(sbi->ckpt); | |
679 | return -EINVAL; | |
680 | } | |
681 | ||
06292073 | 682 | static int __add_dirty_inode(struct inode *inode, struct inode_entry *new) |
127e670a | 683 | { |
4081363f | 684 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
127e670a | 685 | |
ed57c27f JK |
686 | if (is_inode_flag_set(F2FS_I(inode), FI_DIRTY_DIR)) |
687 | return -EEXIST; | |
2d7b822a | 688 | |
ed57c27f JK |
689 | set_inode_flag(F2FS_I(inode), FI_DIRTY_DIR); |
690 | F2FS_I(inode)->dirty_dir = new; | |
691 | list_add_tail(&new->list, &sbi->dir_inode_list); | |
dcdfff65 | 692 | stat_inc_dirty_dir(sbi); |
5deb8267 JK |
693 | return 0; |
694 | } | |
695 | ||
a7ffdbe2 | 696 | void update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 697 | { |
4081363f | 698 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
06292073 | 699 | struct inode_entry *new; |
cf0ee0f0 | 700 | int ret = 0; |
5deb8267 | 701 | |
5ac9f36f CY |
702 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
703 | !S_ISLNK(inode->i_mode)) | |
127e670a | 704 | return; |
7bd59381 | 705 | |
a7ffdbe2 JK |
706 | if (!S_ISDIR(inode->i_mode)) { |
707 | inode_inc_dirty_pages(inode); | |
708 | goto out; | |
709 | } | |
710 | ||
7bd59381 | 711 | new = f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS); |
127e670a JK |
712 | new->inode = inode; |
713 | INIT_LIST_HEAD(&new->list); | |
714 | ||
715 | spin_lock(&sbi->dir_inode_lock); | |
cf0ee0f0 | 716 | ret = __add_dirty_inode(inode, new); |
a7ffdbe2 | 717 | inode_inc_dirty_pages(inode); |
5deb8267 | 718 | spin_unlock(&sbi->dir_inode_lock); |
cf0ee0f0 CY |
719 | |
720 | if (ret) | |
721 | kmem_cache_free(inode_entry_slab, new); | |
a7ffdbe2 JK |
722 | out: |
723 | SetPagePrivate(page); | |
9e4ded3f | 724 | f2fs_trace_pid(page); |
5deb8267 JK |
725 | } |
726 | ||
727 | void add_dirty_dir_inode(struct inode *inode) | |
728 | { | |
4081363f | 729 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
06292073 | 730 | struct inode_entry *new = |
7bd59381 | 731 | f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS); |
cf0ee0f0 | 732 | int ret = 0; |
7bd59381 | 733 | |
5deb8267 JK |
734 | new->inode = inode; |
735 | INIT_LIST_HEAD(&new->list); | |
127e670a | 736 | |
5deb8267 | 737 | spin_lock(&sbi->dir_inode_lock); |
cf0ee0f0 | 738 | ret = __add_dirty_inode(inode, new); |
127e670a | 739 | spin_unlock(&sbi->dir_inode_lock); |
cf0ee0f0 CY |
740 | |
741 | if (ret) | |
742 | kmem_cache_free(inode_entry_slab, new); | |
127e670a JK |
743 | } |
744 | ||
745 | void remove_dirty_dir_inode(struct inode *inode) | |
746 | { | |
4081363f | 747 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
06292073 | 748 | struct inode_entry *entry; |
127e670a JK |
749 | |
750 | if (!S_ISDIR(inode->i_mode)) | |
751 | return; | |
752 | ||
753 | spin_lock(&sbi->dir_inode_lock); | |
a7ffdbe2 | 754 | if (get_dirty_pages(inode) || |
ed57c27f | 755 | !is_inode_flag_set(F2FS_I(inode), FI_DIRTY_DIR)) { |
3b10b1fd JK |
756 | spin_unlock(&sbi->dir_inode_lock); |
757 | return; | |
758 | } | |
127e670a | 759 | |
ed57c27f JK |
760 | entry = F2FS_I(inode)->dirty_dir; |
761 | list_del(&entry->list); | |
762 | F2FS_I(inode)->dirty_dir = NULL; | |
763 | clear_inode_flag(F2FS_I(inode), FI_DIRTY_DIR); | |
764 | stat_dec_dirty_dir(sbi); | |
127e670a | 765 | spin_unlock(&sbi->dir_inode_lock); |
ed57c27f | 766 | kmem_cache_free(inode_entry_slab, entry); |
74d0b917 JK |
767 | |
768 | /* Only from the recovery routine */ | |
afc3eda2 JK |
769 | if (is_inode_flag_set(F2FS_I(inode), FI_DELAY_IPUT)) { |
770 | clear_inode_flag(F2FS_I(inode), FI_DELAY_IPUT); | |
74d0b917 | 771 | iput(inode); |
afc3eda2 | 772 | } |
74d0b917 JK |
773 | } |
774 | ||
127e670a JK |
775 | void sync_dirty_dir_inodes(struct f2fs_sb_info *sbi) |
776 | { | |
ce3b7d80 | 777 | struct list_head *head; |
06292073 | 778 | struct inode_entry *entry; |
127e670a JK |
779 | struct inode *inode; |
780 | retry: | |
af41d3ee JK |
781 | if (unlikely(f2fs_cp_error(sbi))) |
782 | return; | |
783 | ||
127e670a | 784 | spin_lock(&sbi->dir_inode_lock); |
ce3b7d80 GZ |
785 | |
786 | head = &sbi->dir_inode_list; | |
127e670a JK |
787 | if (list_empty(head)) { |
788 | spin_unlock(&sbi->dir_inode_lock); | |
789 | return; | |
790 | } | |
06292073 | 791 | entry = list_entry(head->next, struct inode_entry, list); |
127e670a JK |
792 | inode = igrab(entry->inode); |
793 | spin_unlock(&sbi->dir_inode_lock); | |
794 | if (inode) { | |
87d6f890 | 795 | filemap_fdatawrite(inode->i_mapping); |
127e670a JK |
796 | iput(inode); |
797 | } else { | |
798 | /* | |
799 | * We should submit bio, since it exists several | |
800 | * wribacking dentry pages in the freeing inode. | |
801 | */ | |
458e6197 | 802 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
7ecebe5e | 803 | cond_resched(); |
127e670a JK |
804 | } |
805 | goto retry; | |
806 | } | |
807 | ||
0a8165d7 | 808 | /* |
127e670a JK |
809 | * Freeze all the FS-operations for checkpoint. |
810 | */ | |
cf779cab | 811 | static int block_operations(struct f2fs_sb_info *sbi) |
127e670a | 812 | { |
127e670a JK |
813 | struct writeback_control wbc = { |
814 | .sync_mode = WB_SYNC_ALL, | |
815 | .nr_to_write = LONG_MAX, | |
816 | .for_reclaim = 0, | |
817 | }; | |
c718379b | 818 | struct blk_plug plug; |
cf779cab | 819 | int err = 0; |
c718379b JK |
820 | |
821 | blk_start_plug(&plug); | |
822 | ||
39936837 | 823 | retry_flush_dents: |
e479556b | 824 | f2fs_lock_all(sbi); |
127e670a | 825 | /* write all the dirty dentry pages */ |
127e670a | 826 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 827 | f2fs_unlock_all(sbi); |
39936837 | 828 | sync_dirty_dir_inodes(sbi); |
cf779cab JK |
829 | if (unlikely(f2fs_cp_error(sbi))) { |
830 | err = -EIO; | |
831 | goto out; | |
832 | } | |
39936837 | 833 | goto retry_flush_dents; |
127e670a JK |
834 | } |
835 | ||
127e670a | 836 | /* |
e1c42045 | 837 | * POR: we should ensure that there are no dirty node pages |
127e670a JK |
838 | * until finishing nat/sit flush. |
839 | */ | |
39936837 | 840 | retry_flush_nodes: |
b3582c68 | 841 | down_write(&sbi->node_write); |
127e670a JK |
842 | |
843 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
b3582c68 | 844 | up_write(&sbi->node_write); |
39936837 | 845 | sync_node_pages(sbi, 0, &wbc); |
cf779cab JK |
846 | if (unlikely(f2fs_cp_error(sbi))) { |
847 | f2fs_unlock_all(sbi); | |
848 | err = -EIO; | |
849 | goto out; | |
850 | } | |
39936837 | 851 | goto retry_flush_nodes; |
127e670a | 852 | } |
cf779cab | 853 | out: |
c718379b | 854 | blk_finish_plug(&plug); |
cf779cab | 855 | return err; |
127e670a JK |
856 | } |
857 | ||
858 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
859 | { | |
b3582c68 | 860 | up_write(&sbi->node_write); |
e479556b | 861 | f2fs_unlock_all(sbi); |
127e670a JK |
862 | } |
863 | ||
fb51b5ef CL |
864 | static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi) |
865 | { | |
866 | DEFINE_WAIT(wait); | |
867 | ||
868 | for (;;) { | |
869 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
870 | ||
871 | if (!get_pages(sbi, F2FS_WRITEBACK)) | |
872 | break; | |
873 | ||
874 | io_schedule(); | |
875 | } | |
876 | finish_wait(&sbi->cp_wait, &wait); | |
877 | } | |
878 | ||
75ab4cb8 | 879 | static void do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
880 | { |
881 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
cf2271e7 | 882 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); |
77041823 | 883 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
67298804 | 884 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; |
77041823 | 885 | nid_t last_nid = nm_i->next_scan_nid; |
127e670a | 886 | block_t start_blk; |
127e670a | 887 | unsigned int data_sum_blocks, orphan_blocks; |
7e586fa0 | 888 | __u32 crc32 = 0; |
127e670a | 889 | int i; |
55141486 | 890 | int cp_payload_blks = __cp_payload(sbi); |
127e670a | 891 | |
1e87a78d JK |
892 | /* |
893 | * This avoids to conduct wrong roll-forward operations and uses | |
894 | * metapages, so should be called prior to sync_meta_pages below. | |
895 | */ | |
cf2271e7 | 896 | discard_next_dnode(sbi, NEXT_FREE_BLKADDR(sbi, curseg)); |
127e670a JK |
897 | |
898 | /* Flush all the NAT/SIT pages */ | |
cf779cab | 899 | while (get_pages(sbi, F2FS_DIRTY_META)) { |
127e670a | 900 | sync_meta_pages(sbi, META, LONG_MAX); |
cf779cab JK |
901 | if (unlikely(f2fs_cp_error(sbi))) |
902 | return; | |
903 | } | |
127e670a JK |
904 | |
905 | next_free_nid(sbi, &last_nid); | |
906 | ||
907 | /* | |
908 | * modify checkpoint | |
909 | * version number is already updated | |
910 | */ | |
911 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi)); | |
912 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
913 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); | |
b5b82205 | 914 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
915 | ckpt->cur_node_segno[i] = |
916 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
917 | ckpt->cur_node_blkoff[i] = | |
918 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
919 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
920 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
921 | } | |
b5b82205 | 922 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
923 | ckpt->cur_data_segno[i] = |
924 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
925 | ckpt->cur_data_blkoff[i] = | |
926 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
927 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
928 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
929 | } | |
930 | ||
931 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
932 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
933 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
934 | ||
935 | /* 2 cp + n data seg summary + orphan inode blocks */ | |
3fa06d7b | 936 | data_sum_blocks = npages_for_summary_flush(sbi, false); |
b5b82205 | 937 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
25ca923b | 938 | set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 939 | else |
25ca923b | 940 | clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 941 | |
67298804 | 942 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
943 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
944 | orphan_blocks); | |
127e670a | 945 | |
119ee914 | 946 | if (__remain_node_summaries(cpc->reason)) |
b5b82205 | 947 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
948 | cp_payload_blks + data_sum_blocks + |
949 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
119ee914 | 950 | else |
b5b82205 | 951 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS + |
1dbe4152 CL |
952 | cp_payload_blks + data_sum_blocks + |
953 | orphan_blocks); | |
119ee914 JK |
954 | |
955 | if (cpc->reason == CP_UMOUNT) | |
956 | set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
957 | else | |
958 | clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
959 | ||
960 | if (cpc->reason == CP_FASTBOOT) | |
961 | set_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
962 | else | |
963 | clear_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
127e670a | 964 | |
67298804 | 965 | if (orphan_num) |
25ca923b | 966 | set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 967 | else |
25ca923b | 968 | clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 969 | |
caf0047e | 970 | if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) |
2ae4c673 JK |
971 | set_ckpt_flags(ckpt, CP_FSCK_FLAG); |
972 | ||
127e670a JK |
973 | /* update SIT/NAT bitmap */ |
974 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
975 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
976 | ||
977 | crc32 = f2fs_crc32(ckpt, le32_to_cpu(ckpt->checksum_offset)); | |
7e586fa0 JK |
978 | *((__le32 *)((unsigned char *)ckpt + |
979 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
980 | = cpu_to_le32(crc32); |
981 | ||
982 | start_blk = __start_cp_addr(sbi); | |
983 | ||
984 | /* write out checkpoint buffer at block 0 */ | |
381722d2 CY |
985 | update_meta_page(sbi, ckpt, start_blk++); |
986 | ||
987 | for (i = 1; i < 1 + cp_payload_blks; i++) | |
988 | update_meta_page(sbi, (char *)ckpt + i * F2FS_BLKSIZE, | |
989 | start_blk++); | |
1dbe4152 | 990 | |
67298804 | 991 | if (orphan_num) { |
127e670a JK |
992 | write_orphan_inodes(sbi, start_blk); |
993 | start_blk += orphan_blocks; | |
994 | } | |
995 | ||
996 | write_data_summaries(sbi, start_blk); | |
997 | start_blk += data_sum_blocks; | |
119ee914 | 998 | if (__remain_node_summaries(cpc->reason)) { |
127e670a JK |
999 | write_node_summaries(sbi, start_blk); |
1000 | start_blk += NR_CURSEG_NODE_TYPE; | |
1001 | } | |
1002 | ||
1003 | /* writeout checkpoint block */ | |
381722d2 | 1004 | update_meta_page(sbi, ckpt, start_blk); |
127e670a JK |
1005 | |
1006 | /* wait for previous submitted node/meta pages writeback */ | |
fb51b5ef | 1007 | wait_on_all_pages_writeback(sbi); |
127e670a | 1008 | |
cf779cab JK |
1009 | if (unlikely(f2fs_cp_error(sbi))) |
1010 | return; | |
1011 | ||
4ef51a8f | 1012 | filemap_fdatawait_range(NODE_MAPPING(sbi), 0, LONG_MAX); |
9df27d98 | 1013 | filemap_fdatawait_range(META_MAPPING(sbi), 0, LONG_MAX); |
127e670a JK |
1014 | |
1015 | /* update user_block_counts */ | |
1016 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
1017 | sbi->alloc_valid_block_count = 0; | |
1018 | ||
1019 | /* Here, we only have one bio having CP pack */ | |
577e3495 | 1020 | sync_meta_pages(sbi, META_FLUSH, LONG_MAX); |
127e670a | 1021 | |
6a8f8ca5 JK |
1022 | /* wait for previous submitted meta pages writeback */ |
1023 | wait_on_all_pages_writeback(sbi); | |
1024 | ||
cf779cab JK |
1025 | release_dirty_inode(sbi); |
1026 | ||
1027 | if (unlikely(f2fs_cp_error(sbi))) | |
1028 | return; | |
1029 | ||
836b5a63 | 1030 | clear_prefree_segments(sbi, cpc); |
caf0047e | 1031 | clear_sbi_flag(sbi, SBI_IS_DIRTY); |
127e670a JK |
1032 | } |
1033 | ||
0a8165d7 | 1034 | /* |
e1c42045 | 1035 | * We guarantee that this checkpoint procedure will not fail. |
127e670a | 1036 | */ |
75ab4cb8 | 1037 | void write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1038 | { |
1039 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1040 | unsigned long long ckpt_ver; | |
1041 | ||
43727527 | 1042 | mutex_lock(&sbi->cp_mutex); |
8501017e | 1043 | |
caf0047e | 1044 | if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) && |
a66cdd98 JK |
1045 | (cpc->reason == CP_FASTBOOT || cpc->reason == CP_SYNC || |
1046 | (cpc->reason == CP_DISCARD && !sbi->discard_blks))) | |
8501017e | 1047 | goto out; |
cf779cab JK |
1048 | if (unlikely(f2fs_cp_error(sbi))) |
1049 | goto out; | |
11504a8e JK |
1050 | if (f2fs_readonly(sbi->sb)) |
1051 | goto out; | |
2bda542d WL |
1052 | |
1053 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); | |
1054 | ||
cf779cab JK |
1055 | if (block_operations(sbi)) |
1056 | goto out; | |
127e670a | 1057 | |
75ab4cb8 | 1058 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1059 | |
458e6197 JK |
1060 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
1061 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
1062 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
127e670a JK |
1063 | |
1064 | /* | |
1065 | * update checkpoint pack index | |
1066 | * Increase the version number so that | |
1067 | * SIT entries and seg summaries are written at correct place | |
1068 | */ | |
d71b5564 | 1069 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1070 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1071 | ||
1072 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
1073 | flush_nat_entries(sbi); | |
4b2fecc8 | 1074 | flush_sit_entries(sbi, cpc); |
127e670a | 1075 | |
127e670a | 1076 | /* unlock all the fs_lock[] in do_checkpoint() */ |
75ab4cb8 | 1077 | do_checkpoint(sbi, cpc); |
127e670a JK |
1078 | |
1079 | unblock_operations(sbi); | |
942e0be6 | 1080 | stat_inc_cp_count(sbi->stat_info); |
10027551 JK |
1081 | |
1082 | if (cpc->reason == CP_RECOVERY) | |
1083 | f2fs_msg(sbi->sb, KERN_NOTICE, | |
1084 | "checkpoint: version = %llx", ckpt_ver); | |
8501017e JK |
1085 | out: |
1086 | mutex_unlock(&sbi->cp_mutex); | |
75ab4cb8 | 1087 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
127e670a JK |
1088 | } |
1089 | ||
6451e041 | 1090 | void init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1091 | { |
6451e041 JK |
1092 | int i; |
1093 | ||
1094 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1095 | struct inode_management *im = &sbi->im[i]; |
1096 | ||
1097 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1098 | spin_lock_init(&im->ino_lock); | |
1099 | INIT_LIST_HEAD(&im->ino_list); | |
1100 | im->ino_num = 0; | |
6451e041 JK |
1101 | } |
1102 | ||
b5b82205 | 1103 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
14b42817 WL |
1104 | NR_CURSEG_TYPE - __cp_payload(sbi)) * |
1105 | F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
1106 | } |
1107 | ||
6e6093a8 | 1108 | int __init create_checkpoint_caches(void) |
127e670a | 1109 | { |
6451e041 JK |
1110 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1111 | sizeof(struct ino_entry)); | |
1112 | if (!ino_entry_slab) | |
127e670a | 1113 | return -ENOMEM; |
06292073 CY |
1114 | inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry", |
1115 | sizeof(struct inode_entry)); | |
6bacf52f | 1116 | if (!inode_entry_slab) { |
6451e041 | 1117 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1118 | return -ENOMEM; |
1119 | } | |
1120 | return 0; | |
1121 | } | |
1122 | ||
1123 | void destroy_checkpoint_caches(void) | |
1124 | { | |
6451e041 | 1125 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1126 | kmem_cache_destroy(inode_entry_slab); |
1127 | } |