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" | |
2af4bd6c | 23 | #include <trace/events/f2fs.h> |
127e670a | 24 | |
6451e041 | 25 | static struct kmem_cache *ino_entry_slab; |
127e670a JK |
26 | static struct kmem_cache *inode_entry_slab; |
27 | ||
0a8165d7 | 28 | /* |
127e670a JK |
29 | * We guarantee no failure on the returned page. |
30 | */ | |
31 | struct page *grab_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
32 | { | |
9df27d98 | 33 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
34 | struct page *page = NULL; |
35 | repeat: | |
bde44686 | 36 | page = grab_cache_page(mapping, index); |
127e670a JK |
37 | if (!page) { |
38 | cond_resched(); | |
39 | goto repeat; | |
40 | } | |
bde44686 | 41 | f2fs_wait_on_page_writeback(page, META); |
127e670a JK |
42 | SetPageUptodate(page); |
43 | return page; | |
44 | } | |
45 | ||
0a8165d7 | 46 | /* |
127e670a JK |
47 | * We guarantee no failure on the returned page. |
48 | */ | |
49 | struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
50 | { | |
9df27d98 | 51 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
52 | struct page *page; |
53 | repeat: | |
54 | page = grab_cache_page(mapping, index); | |
55 | if (!page) { | |
56 | cond_resched(); | |
57 | goto repeat; | |
58 | } | |
393ff91f JK |
59 | if (PageUptodate(page)) |
60 | goto out; | |
61 | ||
93dfe2ac JK |
62 | if (f2fs_submit_page_bio(sbi, page, index, |
63 | READ_SYNC | REQ_META | REQ_PRIO)) | |
127e670a | 64 | goto repeat; |
127e670a | 65 | |
393ff91f | 66 | lock_page(page); |
6bacf52f | 67 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
68 | f2fs_put_page(page, 1); |
69 | goto repeat; | |
70 | } | |
393ff91f | 71 | out: |
127e670a JK |
72 | return page; |
73 | } | |
74 | ||
4c521f49 JK |
75 | struct page *get_meta_page_ra(struct f2fs_sb_info *sbi, pgoff_t index) |
76 | { | |
77 | bool readahead = false; | |
78 | struct page *page; | |
79 | ||
80 | page = find_get_page(META_MAPPING(sbi), index); | |
81 | if (!page || (page && !PageUptodate(page))) | |
82 | readahead = true; | |
83 | f2fs_put_page(page, 0); | |
84 | ||
85 | if (readahead) | |
90a893c7 | 86 | ra_meta_pages(sbi, index, MAX_BIO_BLOCKS(sbi), META_POR); |
4c521f49 JK |
87 | return get_meta_page(sbi, index); |
88 | } | |
89 | ||
90 | static inline block_t get_max_meta_blks(struct f2fs_sb_info *sbi, int type) | |
662befda CY |
91 | { |
92 | switch (type) { | |
93 | case META_NAT: | |
94 | return NM_I(sbi)->max_nid / NAT_ENTRY_PER_BLOCK; | |
95 | case META_SIT: | |
96 | return SIT_BLK_CNT(sbi); | |
81c1a0f1 | 97 | case META_SSA: |
662befda CY |
98 | case META_CP: |
99 | return 0; | |
4c521f49 | 100 | case META_POR: |
7cd8558b | 101 | return MAX_BLKADDR(sbi); |
662befda CY |
102 | default: |
103 | BUG(); | |
104 | } | |
105 | } | |
106 | ||
107 | /* | |
81c1a0f1 | 108 | * Readahead CP/NAT/SIT/SSA pages |
662befda | 109 | */ |
4c521f49 | 110 | int ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages, int type) |
662befda CY |
111 | { |
112 | block_t prev_blk_addr = 0; | |
113 | struct page *page; | |
4c521f49 JK |
114 | block_t blkno = start; |
115 | block_t max_blks = get_max_meta_blks(sbi, type); | |
662befda CY |
116 | |
117 | struct f2fs_io_info fio = { | |
118 | .type = META, | |
119 | .rw = READ_SYNC | REQ_META | REQ_PRIO | |
120 | }; | |
121 | ||
122 | for (; nrpages-- > 0; blkno++) { | |
123 | block_t blk_addr; | |
124 | ||
125 | switch (type) { | |
126 | case META_NAT: | |
127 | /* get nat block addr */ | |
128 | if (unlikely(blkno >= max_blks)) | |
129 | blkno = 0; | |
130 | blk_addr = current_nat_addr(sbi, | |
131 | blkno * NAT_ENTRY_PER_BLOCK); | |
132 | break; | |
133 | case META_SIT: | |
134 | /* get sit block addr */ | |
135 | if (unlikely(blkno >= max_blks)) | |
136 | goto out; | |
137 | blk_addr = current_sit_addr(sbi, | |
138 | blkno * SIT_ENTRY_PER_BLOCK); | |
139 | if (blkno != start && prev_blk_addr + 1 != blk_addr) | |
140 | goto out; | |
141 | prev_blk_addr = blk_addr; | |
142 | break; | |
81c1a0f1 | 143 | case META_SSA: |
662befda | 144 | case META_CP: |
4c521f49 JK |
145 | case META_POR: |
146 | if (unlikely(blkno >= max_blks)) | |
147 | goto out; | |
7cd8558b | 148 | if (unlikely(blkno < SEG0_BLKADDR(sbi))) |
4c521f49 | 149 | goto out; |
662befda CY |
150 | blk_addr = blkno; |
151 | break; | |
152 | default: | |
153 | BUG(); | |
154 | } | |
155 | ||
156 | page = grab_cache_page(META_MAPPING(sbi), blk_addr); | |
157 | if (!page) | |
158 | continue; | |
159 | if (PageUptodate(page)) { | |
662befda CY |
160 | f2fs_put_page(page, 1); |
161 | continue; | |
162 | } | |
163 | ||
164 | f2fs_submit_page_mbio(sbi, page, blk_addr, &fio); | |
662befda CY |
165 | f2fs_put_page(page, 0); |
166 | } | |
167 | out: | |
168 | f2fs_submit_merged_bio(sbi, META, READ); | |
169 | return blkno - start; | |
170 | } | |
171 | ||
127e670a JK |
172 | static int f2fs_write_meta_page(struct page *page, |
173 | struct writeback_control *wbc) | |
174 | { | |
4081363f | 175 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 176 | |
ecda0de3 CY |
177 | trace_f2fs_writepage(page, META); |
178 | ||
203681f6 | 179 | if (unlikely(sbi->por_doing)) |
cfb271d4 | 180 | goto redirty_out; |
857dc4e0 | 181 | if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0)) |
cfb271d4 | 182 | goto redirty_out; |
1e968fdf | 183 | if (unlikely(f2fs_cp_error(sbi))) |
cf779cab | 184 | goto redirty_out; |
127e670a | 185 | |
3cb5ad15 | 186 | f2fs_wait_on_page_writeback(page, META); |
577e3495 JK |
187 | write_meta_page(sbi, page); |
188 | dec_page_count(sbi, F2FS_DIRTY_META); | |
189 | unlock_page(page); | |
857dc4e0 JK |
190 | |
191 | if (wbc->for_reclaim) | |
192 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
577e3495 | 193 | return 0; |
cfb271d4 CY |
194 | |
195 | redirty_out: | |
76f60268 | 196 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 197 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
198 | } |
199 | ||
200 | static int f2fs_write_meta_pages(struct address_space *mapping, | |
201 | struct writeback_control *wbc) | |
202 | { | |
4081363f | 203 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 204 | long diff, written; |
127e670a | 205 | |
e5748434 CY |
206 | trace_f2fs_writepages(mapping->host, wbc, META); |
207 | ||
5459aa97 | 208 | /* collect a number of dirty meta pages and write together */ |
50c8cdb3 JK |
209 | if (wbc->for_kupdate || |
210 | get_pages(sbi, F2FS_DIRTY_META) < nr_pages_to_skip(sbi, META)) | |
d3baf95d | 211 | goto skip_write; |
127e670a JK |
212 | |
213 | /* if mounting is failed, skip writing node pages */ | |
214 | mutex_lock(&sbi->cp_mutex); | |
50c8cdb3 JK |
215 | diff = nr_pages_to_write(sbi, META, wbc); |
216 | written = sync_meta_pages(sbi, META, wbc->nr_to_write); | |
127e670a | 217 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 218 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 219 | return 0; |
d3baf95d JK |
220 | |
221 | skip_write: | |
222 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
223 | return 0; | |
127e670a JK |
224 | } |
225 | ||
226 | long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, | |
227 | long nr_to_write) | |
228 | { | |
9df27d98 | 229 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
230 | pgoff_t index = 0, end = LONG_MAX; |
231 | struct pagevec pvec; | |
232 | long nwritten = 0; | |
233 | struct writeback_control wbc = { | |
234 | .for_reclaim = 0, | |
235 | }; | |
236 | ||
237 | pagevec_init(&pvec, 0); | |
238 | ||
239 | while (index <= end) { | |
240 | int i, nr_pages; | |
241 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, | |
242 | PAGECACHE_TAG_DIRTY, | |
243 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
cfb271d4 | 244 | if (unlikely(nr_pages == 0)) |
127e670a JK |
245 | break; |
246 | ||
247 | for (i = 0; i < nr_pages; i++) { | |
248 | struct page *page = pvec.pages[i]; | |
203681f6 | 249 | |
127e670a | 250 | lock_page(page); |
203681f6 JK |
251 | |
252 | if (unlikely(page->mapping != mapping)) { | |
253 | continue_unlock: | |
254 | unlock_page(page); | |
255 | continue; | |
256 | } | |
257 | if (!PageDirty(page)) { | |
258 | /* someone wrote it for us */ | |
259 | goto continue_unlock; | |
260 | } | |
261 | ||
262 | if (!clear_page_dirty_for_io(page)) | |
263 | goto continue_unlock; | |
264 | ||
577e3495 JK |
265 | if (f2fs_write_meta_page(page, &wbc)) { |
266 | unlock_page(page); | |
267 | break; | |
268 | } | |
cfb271d4 CY |
269 | nwritten++; |
270 | if (unlikely(nwritten >= nr_to_write)) | |
127e670a JK |
271 | break; |
272 | } | |
273 | pagevec_release(&pvec); | |
274 | cond_resched(); | |
275 | } | |
276 | ||
277 | if (nwritten) | |
458e6197 | 278 | f2fs_submit_merged_bio(sbi, type, WRITE); |
127e670a JK |
279 | |
280 | return nwritten; | |
281 | } | |
282 | ||
283 | static int f2fs_set_meta_page_dirty(struct page *page) | |
284 | { | |
26c6b887 JK |
285 | trace_f2fs_set_page_dirty(page, META); |
286 | ||
127e670a JK |
287 | SetPageUptodate(page); |
288 | if (!PageDirty(page)) { | |
289 | __set_page_dirty_nobuffers(page); | |
4081363f | 290 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); |
127e670a JK |
291 | return 1; |
292 | } | |
293 | return 0; | |
294 | } | |
295 | ||
296 | const struct address_space_operations f2fs_meta_aops = { | |
297 | .writepage = f2fs_write_meta_page, | |
298 | .writepages = f2fs_write_meta_pages, | |
299 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
300 | }; | |
301 | ||
6451e041 | 302 | static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 303 | { |
67298804 | 304 | struct inode_management *im = &sbi->im[type]; |
39efac41 JK |
305 | struct ino_entry *e; |
306 | retry: | |
67298804 | 307 | spin_lock(&im->ino_lock); |
39efac41 | 308 | |
67298804 | 309 | e = radix_tree_lookup(&im->ino_root, ino); |
39efac41 JK |
310 | if (!e) { |
311 | e = kmem_cache_alloc(ino_entry_slab, GFP_ATOMIC); | |
312 | if (!e) { | |
67298804 | 313 | spin_unlock(&im->ino_lock); |
39efac41 | 314 | goto retry; |
953e6cc6 | 315 | } |
67298804 CY |
316 | if (radix_tree_insert(&im->ino_root, ino, e)) { |
317 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
318 | kmem_cache_free(ino_entry_slab, e); |
319 | goto retry; | |
320 | } | |
321 | memset(e, 0, sizeof(struct ino_entry)); | |
322 | e->ino = ino; | |
953e6cc6 | 323 | |
67298804 | 324 | list_add_tail(&e->list, &im->ino_list); |
8c402946 | 325 | if (type != ORPHAN_INO) |
67298804 | 326 | im->ino_num++; |
39efac41 | 327 | } |
67298804 | 328 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
329 | } |
330 | ||
6451e041 | 331 | static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 332 | { |
67298804 | 333 | struct inode_management *im = &sbi->im[type]; |
6451e041 | 334 | struct ino_entry *e; |
953e6cc6 | 335 | |
67298804 CY |
336 | spin_lock(&im->ino_lock); |
337 | e = radix_tree_lookup(&im->ino_root, ino); | |
39efac41 JK |
338 | if (e) { |
339 | list_del(&e->list); | |
67298804 CY |
340 | radix_tree_delete(&im->ino_root, ino); |
341 | im->ino_num--; | |
342 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
343 | kmem_cache_free(ino_entry_slab, e); |
344 | return; | |
953e6cc6 | 345 | } |
67298804 | 346 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
347 | } |
348 | ||
fff04f90 JK |
349 | void add_dirty_inode(struct f2fs_sb_info *sbi, nid_t ino, int type) |
350 | { | |
351 | /* add new dirty ino entry into list */ | |
352 | __add_ino_entry(sbi, ino, type); | |
353 | } | |
354 | ||
355 | void remove_dirty_inode(struct f2fs_sb_info *sbi, nid_t ino, int type) | |
356 | { | |
357 | /* remove dirty ino entry from list */ | |
358 | __remove_ino_entry(sbi, ino, type); | |
359 | } | |
360 | ||
361 | /* mode should be APPEND_INO or UPDATE_INO */ | |
362 | bool exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode) | |
363 | { | |
67298804 | 364 | struct inode_management *im = &sbi->im[mode]; |
fff04f90 | 365 | struct ino_entry *e; |
67298804 CY |
366 | |
367 | spin_lock(&im->ino_lock); | |
368 | e = radix_tree_lookup(&im->ino_root, ino); | |
369 | spin_unlock(&im->ino_lock); | |
fff04f90 JK |
370 | return e ? true : false; |
371 | } | |
372 | ||
6f12ac25 | 373 | void release_dirty_inode(struct f2fs_sb_info *sbi) |
fff04f90 JK |
374 | { |
375 | struct ino_entry *e, *tmp; | |
376 | int i; | |
377 | ||
378 | for (i = APPEND_INO; i <= UPDATE_INO; i++) { | |
67298804 CY |
379 | struct inode_management *im = &sbi->im[i]; |
380 | ||
381 | spin_lock(&im->ino_lock); | |
382 | list_for_each_entry_safe(e, tmp, &im->ino_list, list) { | |
fff04f90 | 383 | list_del(&e->list); |
67298804 | 384 | radix_tree_delete(&im->ino_root, e->ino); |
fff04f90 | 385 | kmem_cache_free(ino_entry_slab, e); |
67298804 | 386 | im->ino_num--; |
fff04f90 | 387 | } |
67298804 | 388 | spin_unlock(&im->ino_lock); |
fff04f90 JK |
389 | } |
390 | } | |
391 | ||
cbd56e7d | 392 | int acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 393 | { |
67298804 | 394 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a JK |
395 | int err = 0; |
396 | ||
67298804 CY |
397 | spin_lock(&im->ino_lock); |
398 | if (unlikely(im->ino_num >= sbi->max_orphans)) | |
127e670a | 399 | err = -ENOSPC; |
cbd56e7d | 400 | else |
67298804 CY |
401 | im->ino_num++; |
402 | spin_unlock(&im->ino_lock); | |
0d47c1ad | 403 | |
127e670a JK |
404 | return err; |
405 | } | |
406 | ||
cbd56e7d JK |
407 | void release_orphan_inode(struct f2fs_sb_info *sbi) |
408 | { | |
67298804 CY |
409 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
410 | ||
411 | spin_lock(&im->ino_lock); | |
412 | f2fs_bug_on(sbi, im->ino_num == 0); | |
413 | im->ino_num--; | |
414 | spin_unlock(&im->ino_lock); | |
cbd56e7d JK |
415 | } |
416 | ||
127e670a JK |
417 | void add_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
418 | { | |
39efac41 | 419 | /* add new orphan ino entry into list */ |
6451e041 | 420 | __add_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
421 | } |
422 | ||
423 | void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
424 | { | |
953e6cc6 | 425 | /* remove orphan entry from orphan list */ |
6451e041 | 426 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
427 | } |
428 | ||
429 | static void recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
430 | { | |
431 | struct inode *inode = f2fs_iget(sbi->sb, ino); | |
9850cf4a | 432 | f2fs_bug_on(sbi, IS_ERR(inode)); |
127e670a JK |
433 | clear_nlink(inode); |
434 | ||
435 | /* truncate all the data during iput */ | |
436 | iput(inode); | |
437 | } | |
438 | ||
8f99a946 | 439 | void recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a JK |
440 | { |
441 | block_t start_blk, orphan_blkaddr, i, j; | |
442 | ||
25ca923b | 443 | if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG)) |
8f99a946 | 444 | return; |
127e670a | 445 | |
aabe5136 | 446 | sbi->por_doing = true; |
1dbe4152 CL |
447 | |
448 | start_blk = __start_cp_addr(sbi) + 1 + | |
449 | le32_to_cpu(F2FS_RAW_SUPER(sbi)->cp_payload); | |
127e670a JK |
450 | orphan_blkaddr = __start_sum_addr(sbi) - 1; |
451 | ||
662befda CY |
452 | ra_meta_pages(sbi, start_blk, orphan_blkaddr, META_CP); |
453 | ||
127e670a JK |
454 | for (i = 0; i < orphan_blkaddr; i++) { |
455 | struct page *page = get_meta_page(sbi, start_blk + i); | |
456 | struct f2fs_orphan_block *orphan_blk; | |
457 | ||
458 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); | |
459 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
460 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
461 | recover_orphan_inode(sbi, ino); | |
462 | } | |
463 | f2fs_put_page(page, 1); | |
464 | } | |
465 | /* clear Orphan Flag */ | |
25ca923b | 466 | clear_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG); |
aabe5136 | 467 | sbi->por_doing = false; |
8f99a946 | 468 | return; |
127e670a JK |
469 | } |
470 | ||
471 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
472 | { | |
502c6e0b | 473 | struct list_head *head; |
127e670a | 474 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 475 | unsigned int nentries = 0; |
4531929e | 476 | unsigned short index; |
8c402946 | 477 | unsigned short orphan_blocks; |
4531929e | 478 | struct page *page = NULL; |
6451e041 | 479 | struct ino_entry *orphan = NULL; |
67298804 | 480 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 481 | |
67298804 | 482 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 483 | |
4531929e | 484 | for (index = 0; index < orphan_blocks; index++) |
63f5384c | 485 | grab_meta_page(sbi, start_blk + index); |
127e670a | 486 | |
4531929e | 487 | index = 1; |
67298804 CY |
488 | spin_lock(&im->ino_lock); |
489 | head = &im->ino_list; | |
127e670a JK |
490 | |
491 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
492 | list_for_each_entry(orphan, head, list) { |
493 | if (!page) { | |
63f5384c | 494 | page = find_get_page(META_MAPPING(sbi), start_blk++); |
9850cf4a | 495 | f2fs_bug_on(sbi, !page); |
502c6e0b GZ |
496 | orphan_blk = |
497 | (struct f2fs_orphan_block *)page_address(page); | |
498 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
63f5384c | 499 | f2fs_put_page(page, 0); |
502c6e0b | 500 | } |
127e670a | 501 | |
36795567 | 502 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 503 | |
36795567 | 504 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
505 | /* |
506 | * an orphan block is full of 1020 entries, | |
507 | * then we need to flush current orphan blocks | |
508 | * and bring another one in memory | |
509 | */ | |
510 | orphan_blk->blk_addr = cpu_to_le16(index); | |
511 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
512 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
513 | set_page_dirty(page); | |
514 | f2fs_put_page(page, 1); | |
515 | index++; | |
127e670a JK |
516 | nentries = 0; |
517 | page = NULL; | |
518 | } | |
502c6e0b | 519 | } |
127e670a | 520 | |
502c6e0b GZ |
521 | if (page) { |
522 | orphan_blk->blk_addr = cpu_to_le16(index); | |
523 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
524 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
525 | set_page_dirty(page); | |
526 | f2fs_put_page(page, 1); | |
127e670a | 527 | } |
502c6e0b | 528 | |
67298804 | 529 | spin_unlock(&im->ino_lock); |
127e670a JK |
530 | } |
531 | ||
532 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, | |
533 | block_t cp_addr, unsigned long long *version) | |
534 | { | |
535 | struct page *cp_page_1, *cp_page_2 = NULL; | |
536 | unsigned long blk_size = sbi->blocksize; | |
537 | struct f2fs_checkpoint *cp_block; | |
538 | unsigned long long cur_version = 0, pre_version = 0; | |
127e670a | 539 | size_t crc_offset; |
7e586fa0 | 540 | __u32 crc = 0; |
127e670a JK |
541 | |
542 | /* Read the 1st cp block in this CP pack */ | |
543 | cp_page_1 = get_meta_page(sbi, cp_addr); | |
544 | ||
545 | /* get the version number */ | |
546 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_1); | |
547 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
548 | if (crc_offset >= blk_size) | |
549 | goto invalid_cp1; | |
550 | ||
7e586fa0 | 551 | crc = le32_to_cpu(*((__u32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
552 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
553 | goto invalid_cp1; | |
554 | ||
d71b5564 | 555 | pre_version = cur_cp_version(cp_block); |
127e670a JK |
556 | |
557 | /* Read the 2nd cp block in this CP pack */ | |
25ca923b | 558 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
127e670a JK |
559 | cp_page_2 = get_meta_page(sbi, cp_addr); |
560 | ||
561 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_2); | |
562 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
563 | if (crc_offset >= blk_size) | |
564 | goto invalid_cp2; | |
565 | ||
7e586fa0 | 566 | crc = le32_to_cpu(*((__u32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
567 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
568 | goto invalid_cp2; | |
569 | ||
d71b5564 | 570 | cur_version = cur_cp_version(cp_block); |
127e670a JK |
571 | |
572 | if (cur_version == pre_version) { | |
573 | *version = cur_version; | |
574 | f2fs_put_page(cp_page_2, 1); | |
575 | return cp_page_1; | |
576 | } | |
577 | invalid_cp2: | |
578 | f2fs_put_page(cp_page_2, 1); | |
579 | invalid_cp1: | |
580 | f2fs_put_page(cp_page_1, 1); | |
581 | return NULL; | |
582 | } | |
583 | ||
584 | int get_valid_checkpoint(struct f2fs_sb_info *sbi) | |
585 | { | |
586 | struct f2fs_checkpoint *cp_block; | |
587 | struct f2fs_super_block *fsb = sbi->raw_super; | |
588 | struct page *cp1, *cp2, *cur_page; | |
589 | unsigned long blk_size = sbi->blocksize; | |
590 | unsigned long long cp1_version = 0, cp2_version = 0; | |
591 | unsigned long long cp_start_blk_no; | |
1dbe4152 CL |
592 | unsigned int cp_blks = 1 + le32_to_cpu(F2FS_RAW_SUPER(sbi)->cp_payload); |
593 | block_t cp_blk_no; | |
594 | int i; | |
127e670a | 595 | |
1dbe4152 | 596 | sbi->ckpt = kzalloc(cp_blks * blk_size, GFP_KERNEL); |
127e670a JK |
597 | if (!sbi->ckpt) |
598 | return -ENOMEM; | |
599 | /* | |
600 | * Finding out valid cp block involves read both | |
601 | * sets( cp pack1 and cp pack 2) | |
602 | */ | |
603 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
604 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
605 | ||
606 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
607 | cp_start_blk_no += ((unsigned long long)1) << |
608 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
609 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
610 | ||
611 | if (cp1 && cp2) { | |
612 | if (ver_after(cp2_version, cp1_version)) | |
613 | cur_page = cp2; | |
614 | else | |
615 | cur_page = cp1; | |
616 | } else if (cp1) { | |
617 | cur_page = cp1; | |
618 | } else if (cp2) { | |
619 | cur_page = cp2; | |
620 | } else { | |
621 | goto fail_no_cp; | |
622 | } | |
623 | ||
624 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
625 | memcpy(sbi->ckpt, cp_block, blk_size); | |
626 | ||
1dbe4152 CL |
627 | if (cp_blks <= 1) |
628 | goto done; | |
629 | ||
630 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
631 | if (cur_page == cp2) | |
632 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
633 | ||
634 | for (i = 1; i < cp_blks; i++) { | |
635 | void *sit_bitmap_ptr; | |
636 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
637 | ||
638 | cur_page = get_meta_page(sbi, cp_blk_no + i); | |
639 | sit_bitmap_ptr = page_address(cur_page); | |
640 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
641 | f2fs_put_page(cur_page, 1); | |
642 | } | |
643 | done: | |
127e670a JK |
644 | f2fs_put_page(cp1, 1); |
645 | f2fs_put_page(cp2, 1); | |
646 | return 0; | |
647 | ||
648 | fail_no_cp: | |
649 | kfree(sbi->ckpt); | |
650 | return -EINVAL; | |
651 | } | |
652 | ||
5deb8267 | 653 | static int __add_dirty_inode(struct inode *inode, struct dir_inode_entry *new) |
127e670a | 654 | { |
4081363f | 655 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
127e670a | 656 | |
ed57c27f JK |
657 | if (is_inode_flag_set(F2FS_I(inode), FI_DIRTY_DIR)) |
658 | return -EEXIST; | |
2d7b822a | 659 | |
ed57c27f JK |
660 | set_inode_flag(F2FS_I(inode), FI_DIRTY_DIR); |
661 | F2FS_I(inode)->dirty_dir = new; | |
662 | list_add_tail(&new->list, &sbi->dir_inode_list); | |
dcdfff65 | 663 | stat_inc_dirty_dir(sbi); |
5deb8267 JK |
664 | return 0; |
665 | } | |
666 | ||
a7ffdbe2 | 667 | void update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 668 | { |
4081363f | 669 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
5deb8267 | 670 | struct dir_inode_entry *new; |
cf0ee0f0 | 671 | int ret = 0; |
5deb8267 | 672 | |
a7ffdbe2 | 673 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode)) |
127e670a | 674 | return; |
7bd59381 | 675 | |
a7ffdbe2 JK |
676 | if (!S_ISDIR(inode->i_mode)) { |
677 | inode_inc_dirty_pages(inode); | |
678 | goto out; | |
679 | } | |
680 | ||
7bd59381 | 681 | new = f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS); |
127e670a JK |
682 | new->inode = inode; |
683 | INIT_LIST_HEAD(&new->list); | |
684 | ||
685 | spin_lock(&sbi->dir_inode_lock); | |
cf0ee0f0 | 686 | ret = __add_dirty_inode(inode, new); |
a7ffdbe2 | 687 | inode_inc_dirty_pages(inode); |
5deb8267 | 688 | spin_unlock(&sbi->dir_inode_lock); |
cf0ee0f0 CY |
689 | |
690 | if (ret) | |
691 | kmem_cache_free(inode_entry_slab, new); | |
a7ffdbe2 JK |
692 | out: |
693 | SetPagePrivate(page); | |
5deb8267 JK |
694 | } |
695 | ||
696 | void add_dirty_dir_inode(struct inode *inode) | |
697 | { | |
4081363f | 698 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
7bd59381 GZ |
699 | struct dir_inode_entry *new = |
700 | f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS); | |
cf0ee0f0 | 701 | int ret = 0; |
7bd59381 | 702 | |
5deb8267 JK |
703 | new->inode = inode; |
704 | INIT_LIST_HEAD(&new->list); | |
127e670a | 705 | |
5deb8267 | 706 | spin_lock(&sbi->dir_inode_lock); |
cf0ee0f0 | 707 | ret = __add_dirty_inode(inode, new); |
127e670a | 708 | spin_unlock(&sbi->dir_inode_lock); |
cf0ee0f0 CY |
709 | |
710 | if (ret) | |
711 | kmem_cache_free(inode_entry_slab, new); | |
127e670a JK |
712 | } |
713 | ||
714 | void remove_dirty_dir_inode(struct inode *inode) | |
715 | { | |
4081363f | 716 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
2d7b822a | 717 | struct dir_inode_entry *entry; |
127e670a JK |
718 | |
719 | if (!S_ISDIR(inode->i_mode)) | |
720 | return; | |
721 | ||
722 | spin_lock(&sbi->dir_inode_lock); | |
a7ffdbe2 | 723 | if (get_dirty_pages(inode) || |
ed57c27f | 724 | !is_inode_flag_set(F2FS_I(inode), FI_DIRTY_DIR)) { |
3b10b1fd JK |
725 | spin_unlock(&sbi->dir_inode_lock); |
726 | return; | |
727 | } | |
127e670a | 728 | |
ed57c27f JK |
729 | entry = F2FS_I(inode)->dirty_dir; |
730 | list_del(&entry->list); | |
731 | F2FS_I(inode)->dirty_dir = NULL; | |
732 | clear_inode_flag(F2FS_I(inode), FI_DIRTY_DIR); | |
733 | stat_dec_dirty_dir(sbi); | |
127e670a | 734 | spin_unlock(&sbi->dir_inode_lock); |
ed57c27f | 735 | kmem_cache_free(inode_entry_slab, entry); |
74d0b917 JK |
736 | |
737 | /* Only from the recovery routine */ | |
afc3eda2 JK |
738 | if (is_inode_flag_set(F2FS_I(inode), FI_DELAY_IPUT)) { |
739 | clear_inode_flag(F2FS_I(inode), FI_DELAY_IPUT); | |
74d0b917 | 740 | iput(inode); |
afc3eda2 | 741 | } |
74d0b917 JK |
742 | } |
743 | ||
127e670a JK |
744 | void sync_dirty_dir_inodes(struct f2fs_sb_info *sbi) |
745 | { | |
ce3b7d80 | 746 | struct list_head *head; |
127e670a JK |
747 | struct dir_inode_entry *entry; |
748 | struct inode *inode; | |
749 | retry: | |
af41d3ee JK |
750 | if (unlikely(f2fs_cp_error(sbi))) |
751 | return; | |
752 | ||
127e670a | 753 | spin_lock(&sbi->dir_inode_lock); |
ce3b7d80 GZ |
754 | |
755 | head = &sbi->dir_inode_list; | |
127e670a JK |
756 | if (list_empty(head)) { |
757 | spin_unlock(&sbi->dir_inode_lock); | |
758 | return; | |
759 | } | |
760 | entry = list_entry(head->next, struct dir_inode_entry, list); | |
761 | inode = igrab(entry->inode); | |
762 | spin_unlock(&sbi->dir_inode_lock); | |
763 | if (inode) { | |
87d6f890 | 764 | filemap_fdatawrite(inode->i_mapping); |
127e670a JK |
765 | iput(inode); |
766 | } else { | |
767 | /* | |
768 | * We should submit bio, since it exists several | |
769 | * wribacking dentry pages in the freeing inode. | |
770 | */ | |
458e6197 | 771 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
127e670a JK |
772 | } |
773 | goto retry; | |
774 | } | |
775 | ||
0a8165d7 | 776 | /* |
127e670a JK |
777 | * Freeze all the FS-operations for checkpoint. |
778 | */ | |
cf779cab | 779 | static int block_operations(struct f2fs_sb_info *sbi) |
127e670a | 780 | { |
127e670a JK |
781 | struct writeback_control wbc = { |
782 | .sync_mode = WB_SYNC_ALL, | |
783 | .nr_to_write = LONG_MAX, | |
784 | .for_reclaim = 0, | |
785 | }; | |
c718379b | 786 | struct blk_plug plug; |
cf779cab | 787 | int err = 0; |
c718379b JK |
788 | |
789 | blk_start_plug(&plug); | |
790 | ||
39936837 | 791 | retry_flush_dents: |
e479556b | 792 | f2fs_lock_all(sbi); |
127e670a | 793 | /* write all the dirty dentry pages */ |
127e670a | 794 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 795 | f2fs_unlock_all(sbi); |
39936837 | 796 | sync_dirty_dir_inodes(sbi); |
cf779cab JK |
797 | if (unlikely(f2fs_cp_error(sbi))) { |
798 | err = -EIO; | |
799 | goto out; | |
800 | } | |
39936837 | 801 | goto retry_flush_dents; |
127e670a JK |
802 | } |
803 | ||
127e670a | 804 | /* |
e1c42045 | 805 | * POR: we should ensure that there are no dirty node pages |
127e670a JK |
806 | * until finishing nat/sit flush. |
807 | */ | |
39936837 | 808 | retry_flush_nodes: |
b3582c68 | 809 | down_write(&sbi->node_write); |
127e670a JK |
810 | |
811 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
b3582c68 | 812 | up_write(&sbi->node_write); |
39936837 | 813 | sync_node_pages(sbi, 0, &wbc); |
cf779cab JK |
814 | if (unlikely(f2fs_cp_error(sbi))) { |
815 | f2fs_unlock_all(sbi); | |
816 | err = -EIO; | |
817 | goto out; | |
818 | } | |
39936837 | 819 | goto retry_flush_nodes; |
127e670a | 820 | } |
cf779cab | 821 | out: |
c718379b | 822 | blk_finish_plug(&plug); |
cf779cab | 823 | return err; |
127e670a JK |
824 | } |
825 | ||
826 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
827 | { | |
b3582c68 | 828 | up_write(&sbi->node_write); |
e479556b | 829 | f2fs_unlock_all(sbi); |
127e670a JK |
830 | } |
831 | ||
fb51b5ef CL |
832 | static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi) |
833 | { | |
834 | DEFINE_WAIT(wait); | |
835 | ||
836 | for (;;) { | |
837 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
838 | ||
839 | if (!get_pages(sbi, F2FS_WRITEBACK)) | |
840 | break; | |
841 | ||
842 | io_schedule(); | |
843 | } | |
844 | finish_wait(&sbi->cp_wait, &wait); | |
845 | } | |
846 | ||
75ab4cb8 | 847 | static void do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
848 | { |
849 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
cf2271e7 | 850 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); |
77041823 | 851 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
67298804 | 852 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; |
77041823 | 853 | nid_t last_nid = nm_i->next_scan_nid; |
127e670a JK |
854 | block_t start_blk; |
855 | struct page *cp_page; | |
856 | unsigned int data_sum_blocks, orphan_blocks; | |
7e586fa0 | 857 | __u32 crc32 = 0; |
127e670a | 858 | void *kaddr; |
127e670a | 859 | int i; |
1dbe4152 | 860 | int cp_payload_blks = le32_to_cpu(F2FS_RAW_SUPER(sbi)->cp_payload); |
127e670a | 861 | |
1e87a78d JK |
862 | /* |
863 | * This avoids to conduct wrong roll-forward operations and uses | |
864 | * metapages, so should be called prior to sync_meta_pages below. | |
865 | */ | |
cf2271e7 | 866 | discard_next_dnode(sbi, NEXT_FREE_BLKADDR(sbi, curseg)); |
127e670a JK |
867 | |
868 | /* Flush all the NAT/SIT pages */ | |
cf779cab | 869 | while (get_pages(sbi, F2FS_DIRTY_META)) { |
127e670a | 870 | sync_meta_pages(sbi, META, LONG_MAX); |
cf779cab JK |
871 | if (unlikely(f2fs_cp_error(sbi))) |
872 | return; | |
873 | } | |
127e670a JK |
874 | |
875 | next_free_nid(sbi, &last_nid); | |
876 | ||
877 | /* | |
878 | * modify checkpoint | |
879 | * version number is already updated | |
880 | */ | |
881 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi)); | |
882 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
883 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); | |
b5b82205 | 884 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
885 | ckpt->cur_node_segno[i] = |
886 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
887 | ckpt->cur_node_blkoff[i] = | |
888 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
889 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
890 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
891 | } | |
b5b82205 | 892 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
893 | ckpt->cur_data_segno[i] = |
894 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
895 | ckpt->cur_data_blkoff[i] = | |
896 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
897 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
898 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
899 | } | |
900 | ||
901 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
902 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
903 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
904 | ||
905 | /* 2 cp + n data seg summary + orphan inode blocks */ | |
906 | data_sum_blocks = npages_for_summary_flush(sbi); | |
b5b82205 | 907 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
25ca923b | 908 | set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 909 | else |
25ca923b | 910 | clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 911 | |
67298804 | 912 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
913 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
914 | orphan_blocks); | |
127e670a | 915 | |
75ab4cb8 | 916 | if (cpc->reason == CP_UMOUNT) { |
25ca923b | 917 | set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); |
b5b82205 | 918 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
919 | cp_payload_blks + data_sum_blocks + |
920 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
127e670a | 921 | } else { |
25ca923b | 922 | clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); |
b5b82205 | 923 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS + |
1dbe4152 CL |
924 | cp_payload_blks + data_sum_blocks + |
925 | orphan_blocks); | |
127e670a JK |
926 | } |
927 | ||
67298804 | 928 | if (orphan_num) |
25ca923b | 929 | set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 930 | else |
25ca923b | 931 | clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 932 | |
2ae4c673 JK |
933 | if (sbi->need_fsck) |
934 | set_ckpt_flags(ckpt, CP_FSCK_FLAG); | |
935 | ||
127e670a JK |
936 | /* update SIT/NAT bitmap */ |
937 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
938 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
939 | ||
940 | crc32 = f2fs_crc32(ckpt, le32_to_cpu(ckpt->checksum_offset)); | |
7e586fa0 JK |
941 | *((__le32 *)((unsigned char *)ckpt + |
942 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
943 | = cpu_to_le32(crc32); |
944 | ||
945 | start_blk = __start_cp_addr(sbi); | |
946 | ||
947 | /* write out checkpoint buffer at block 0 */ | |
948 | cp_page = grab_meta_page(sbi, start_blk++); | |
949 | kaddr = page_address(cp_page); | |
950 | memcpy(kaddr, ckpt, (1 << sbi->log_blocksize)); | |
951 | set_page_dirty(cp_page); | |
952 | f2fs_put_page(cp_page, 1); | |
953 | ||
1dbe4152 CL |
954 | for (i = 1; i < 1 + cp_payload_blks; i++) { |
955 | cp_page = grab_meta_page(sbi, start_blk++); | |
956 | kaddr = page_address(cp_page); | |
957 | memcpy(kaddr, (char *)ckpt + i * F2FS_BLKSIZE, | |
958 | (1 << sbi->log_blocksize)); | |
959 | set_page_dirty(cp_page); | |
960 | f2fs_put_page(cp_page, 1); | |
961 | } | |
962 | ||
67298804 | 963 | if (orphan_num) { |
127e670a JK |
964 | write_orphan_inodes(sbi, start_blk); |
965 | start_blk += orphan_blocks; | |
966 | } | |
967 | ||
968 | write_data_summaries(sbi, start_blk); | |
969 | start_blk += data_sum_blocks; | |
75ab4cb8 | 970 | if (cpc->reason == CP_UMOUNT) { |
127e670a JK |
971 | write_node_summaries(sbi, start_blk); |
972 | start_blk += NR_CURSEG_NODE_TYPE; | |
973 | } | |
974 | ||
975 | /* writeout checkpoint block */ | |
976 | cp_page = grab_meta_page(sbi, start_blk); | |
977 | kaddr = page_address(cp_page); | |
978 | memcpy(kaddr, ckpt, (1 << sbi->log_blocksize)); | |
979 | set_page_dirty(cp_page); | |
980 | f2fs_put_page(cp_page, 1); | |
981 | ||
982 | /* wait for previous submitted node/meta pages writeback */ | |
fb51b5ef | 983 | wait_on_all_pages_writeback(sbi); |
127e670a | 984 | |
cf779cab JK |
985 | if (unlikely(f2fs_cp_error(sbi))) |
986 | return; | |
987 | ||
4ef51a8f | 988 | filemap_fdatawait_range(NODE_MAPPING(sbi), 0, LONG_MAX); |
9df27d98 | 989 | filemap_fdatawait_range(META_MAPPING(sbi), 0, LONG_MAX); |
127e670a JK |
990 | |
991 | /* update user_block_counts */ | |
992 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
993 | sbi->alloc_valid_block_count = 0; | |
994 | ||
995 | /* Here, we only have one bio having CP pack */ | |
577e3495 | 996 | sync_meta_pages(sbi, META_FLUSH, LONG_MAX); |
127e670a | 997 | |
6a8f8ca5 JK |
998 | /* wait for previous submitted meta pages writeback */ |
999 | wait_on_all_pages_writeback(sbi); | |
1000 | ||
cf779cab JK |
1001 | release_dirty_inode(sbi); |
1002 | ||
1003 | if (unlikely(f2fs_cp_error(sbi))) | |
1004 | return; | |
1005 | ||
1006 | clear_prefree_segments(sbi); | |
1007 | F2FS_RESET_SB_DIRT(sbi); | |
127e670a JK |
1008 | } |
1009 | ||
0a8165d7 | 1010 | /* |
e1c42045 | 1011 | * We guarantee that this checkpoint procedure will not fail. |
127e670a | 1012 | */ |
75ab4cb8 | 1013 | void write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1014 | { |
1015 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1016 | unsigned long long ckpt_ver; | |
1017 | ||
75ab4cb8 | 1018 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); |
2af4bd6c | 1019 | |
43727527 | 1020 | mutex_lock(&sbi->cp_mutex); |
8501017e | 1021 | |
4b2fecc8 | 1022 | if (!sbi->s_dirty && cpc->reason != CP_DISCARD) |
8501017e | 1023 | goto out; |
cf779cab JK |
1024 | if (unlikely(f2fs_cp_error(sbi))) |
1025 | goto out; | |
1026 | if (block_operations(sbi)) | |
1027 | goto out; | |
127e670a | 1028 | |
75ab4cb8 | 1029 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1030 | |
458e6197 JK |
1031 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
1032 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
1033 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
127e670a JK |
1034 | |
1035 | /* | |
1036 | * update checkpoint pack index | |
1037 | * Increase the version number so that | |
1038 | * SIT entries and seg summaries are written at correct place | |
1039 | */ | |
d71b5564 | 1040 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1041 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1042 | ||
1043 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
1044 | flush_nat_entries(sbi); | |
4b2fecc8 | 1045 | flush_sit_entries(sbi, cpc); |
127e670a | 1046 | |
127e670a | 1047 | /* unlock all the fs_lock[] in do_checkpoint() */ |
75ab4cb8 | 1048 | do_checkpoint(sbi, cpc); |
127e670a JK |
1049 | |
1050 | unblock_operations(sbi); | |
942e0be6 | 1051 | stat_inc_cp_count(sbi->stat_info); |
8501017e JK |
1052 | out: |
1053 | mutex_unlock(&sbi->cp_mutex); | |
75ab4cb8 | 1054 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
127e670a JK |
1055 | } |
1056 | ||
6451e041 | 1057 | void init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1058 | { |
6451e041 JK |
1059 | int i; |
1060 | ||
1061 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1062 | struct inode_management *im = &sbi->im[i]; |
1063 | ||
1064 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1065 | spin_lock_init(&im->ino_lock); | |
1066 | INIT_LIST_HEAD(&im->ino_list); | |
1067 | im->ino_num = 0; | |
6451e041 JK |
1068 | } |
1069 | ||
0d47c1ad GZ |
1070 | /* |
1071 | * considering 512 blocks in a segment 8 blocks are needed for cp | |
1072 | * and log segment summaries. Remaining blocks are used to keep | |
1073 | * orphan entries with the limitation one reserved segment | |
1074 | * for cp pack we can have max 1020*504 orphan entries | |
1075 | */ | |
b5b82205 CY |
1076 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
1077 | NR_CURSEG_TYPE) * F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
1078 | } |
1079 | ||
6e6093a8 | 1080 | int __init create_checkpoint_caches(void) |
127e670a | 1081 | { |
6451e041 JK |
1082 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1083 | sizeof(struct ino_entry)); | |
1084 | if (!ino_entry_slab) | |
127e670a JK |
1085 | return -ENOMEM; |
1086 | inode_entry_slab = f2fs_kmem_cache_create("f2fs_dirty_dir_entry", | |
e8512d2e | 1087 | sizeof(struct dir_inode_entry)); |
6bacf52f | 1088 | if (!inode_entry_slab) { |
6451e041 | 1089 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1090 | return -ENOMEM; |
1091 | } | |
1092 | return 0; | |
1093 | } | |
1094 | ||
1095 | void destroy_checkpoint_caches(void) | |
1096 | { | |
6451e041 | 1097 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1098 | kmem_cache_destroy(inode_entry_slab); |
1099 | } |