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