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