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