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 | */ | |
2b947003 CY |
50 | static struct page *__get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index, |
51 | bool is_meta) | |
127e670a | 52 | { |
9df27d98 | 53 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a | 54 | struct page *page; |
cf04e8eb | 55 | struct f2fs_io_info fio = { |
05ca3632 | 56 | .sbi = sbi, |
cf04e8eb JK |
57 | .type = META, |
58 | .rw = READ_SYNC | REQ_META | REQ_PRIO, | |
59 | .blk_addr = index, | |
4375a336 | 60 | .encrypted_page = NULL, |
cf04e8eb | 61 | }; |
2b947003 CY |
62 | |
63 | if (unlikely(!is_meta)) | |
64 | fio.rw &= ~REQ_META; | |
127e670a JK |
65 | repeat: |
66 | page = grab_cache_page(mapping, index); | |
67 | if (!page) { | |
68 | cond_resched(); | |
69 | goto repeat; | |
70 | } | |
393ff91f JK |
71 | if (PageUptodate(page)) |
72 | goto out; | |
73 | ||
05ca3632 JK |
74 | fio.page = page; |
75 | ||
86531d6b JK |
76 | if (f2fs_submit_page_bio(&fio)) { |
77 | f2fs_put_page(page, 1); | |
127e670a | 78 | goto repeat; |
86531d6b | 79 | } |
127e670a | 80 | |
393ff91f | 81 | lock_page(page); |
6bacf52f | 82 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
83 | f2fs_put_page(page, 1); |
84 | goto repeat; | |
85 | } | |
f3f338ca CY |
86 | |
87 | /* | |
88 | * if there is any IO error when accessing device, make our filesystem | |
89 | * readonly and make sure do not write checkpoint with non-uptodate | |
90 | * meta page. | |
91 | */ | |
92 | if (unlikely(!PageUptodate(page))) | |
93 | f2fs_stop_checkpoint(sbi); | |
393ff91f | 94 | out: |
127e670a JK |
95 | return page; |
96 | } | |
97 | ||
2b947003 CY |
98 | struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) |
99 | { | |
100 | return __get_meta_page(sbi, index, true); | |
101 | } | |
102 | ||
103 | /* for POR only */ | |
104 | struct page *get_tmp_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
105 | { | |
106 | return __get_meta_page(sbi, index, false); | |
107 | } | |
108 | ||
f0c9cada | 109 | bool is_valid_blkaddr(struct f2fs_sb_info *sbi, block_t blkaddr, int type) |
662befda CY |
110 | { |
111 | switch (type) { | |
112 | case META_NAT: | |
66b00c18 | 113 | break; |
662befda | 114 | case META_SIT: |
66b00c18 CY |
115 | if (unlikely(blkaddr >= SIT_BLK_CNT(sbi))) |
116 | return false; | |
117 | break; | |
81c1a0f1 | 118 | case META_SSA: |
66b00c18 CY |
119 | if (unlikely(blkaddr >= MAIN_BLKADDR(sbi) || |
120 | blkaddr < SM_I(sbi)->ssa_blkaddr)) | |
121 | return false; | |
122 | break; | |
662befda | 123 | case META_CP: |
66b00c18 CY |
124 | if (unlikely(blkaddr >= SIT_I(sbi)->sit_base_addr || |
125 | blkaddr < __start_cp_addr(sbi))) | |
126 | return false; | |
127 | break; | |
4c521f49 | 128 | case META_POR: |
66b00c18 CY |
129 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
130 | blkaddr < MAIN_BLKADDR(sbi))) | |
131 | return false; | |
132 | break; | |
662befda CY |
133 | default: |
134 | BUG(); | |
135 | } | |
66b00c18 CY |
136 | |
137 | return true; | |
662befda CY |
138 | } |
139 | ||
140 | /* | |
81c1a0f1 | 141 | * Readahead CP/NAT/SIT/SSA pages |
662befda | 142 | */ |
26879fb1 CY |
143 | int ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages, |
144 | int type, bool sync) | |
662befda CY |
145 | { |
146 | block_t prev_blk_addr = 0; | |
147 | struct page *page; | |
4c521f49 | 148 | block_t blkno = start; |
662befda | 149 | struct f2fs_io_info fio = { |
05ca3632 | 150 | .sbi = sbi, |
662befda | 151 | .type = META, |
26879fb1 | 152 | .rw = sync ? (READ_SYNC | REQ_META | REQ_PRIO) : READA, |
4375a336 | 153 | .encrypted_page = NULL, |
662befda CY |
154 | }; |
155 | ||
2b947003 CY |
156 | if (unlikely(type == META_POR)) |
157 | fio.rw &= ~REQ_META; | |
158 | ||
662befda | 159 | for (; nrpages-- > 0; blkno++) { |
662befda | 160 | |
66b00c18 CY |
161 | if (!is_valid_blkaddr(sbi, blkno, type)) |
162 | goto out; | |
163 | ||
662befda CY |
164 | switch (type) { |
165 | case META_NAT: | |
66b00c18 CY |
166 | if (unlikely(blkno >= |
167 | NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))) | |
662befda | 168 | blkno = 0; |
66b00c18 | 169 | /* get nat block addr */ |
cf04e8eb | 170 | fio.blk_addr = current_nat_addr(sbi, |
662befda CY |
171 | blkno * NAT_ENTRY_PER_BLOCK); |
172 | break; | |
173 | case META_SIT: | |
174 | /* get sit block addr */ | |
cf04e8eb | 175 | fio.blk_addr = current_sit_addr(sbi, |
662befda | 176 | blkno * SIT_ENTRY_PER_BLOCK); |
cf04e8eb | 177 | if (blkno != start && prev_blk_addr + 1 != fio.blk_addr) |
662befda | 178 | goto out; |
cf04e8eb | 179 | prev_blk_addr = fio.blk_addr; |
662befda | 180 | break; |
81c1a0f1 | 181 | case META_SSA: |
662befda | 182 | case META_CP: |
4c521f49 | 183 | case META_POR: |
cf04e8eb | 184 | fio.blk_addr = blkno; |
662befda CY |
185 | break; |
186 | default: | |
187 | BUG(); | |
188 | } | |
189 | ||
cf04e8eb | 190 | page = grab_cache_page(META_MAPPING(sbi), fio.blk_addr); |
662befda CY |
191 | if (!page) |
192 | continue; | |
193 | if (PageUptodate(page)) { | |
662befda CY |
194 | f2fs_put_page(page, 1); |
195 | continue; | |
196 | } | |
197 | ||
05ca3632 JK |
198 | fio.page = page; |
199 | f2fs_submit_page_mbio(&fio); | |
662befda CY |
200 | f2fs_put_page(page, 0); |
201 | } | |
202 | out: | |
203 | f2fs_submit_merged_bio(sbi, META, READ); | |
204 | return blkno - start; | |
205 | } | |
206 | ||
635aee1f CY |
207 | void ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index) |
208 | { | |
209 | struct page *page; | |
210 | bool readahead = false; | |
211 | ||
212 | page = find_get_page(META_MAPPING(sbi), index); | |
213 | if (!page || (page && !PageUptodate(page))) | |
214 | readahead = true; | |
215 | f2fs_put_page(page, 0); | |
216 | ||
217 | if (readahead) | |
26879fb1 | 218 | ra_meta_pages(sbi, index, MAX_BIO_BLOCKS(sbi), META_POR, true); |
635aee1f CY |
219 | } |
220 | ||
127e670a JK |
221 | static int f2fs_write_meta_page(struct page *page, |
222 | struct writeback_control *wbc) | |
223 | { | |
4081363f | 224 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 225 | |
ecda0de3 CY |
226 | trace_f2fs_writepage(page, META); |
227 | ||
caf0047e | 228 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
cfb271d4 | 229 | goto redirty_out; |
857dc4e0 | 230 | if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0)) |
cfb271d4 | 231 | goto redirty_out; |
1e968fdf | 232 | if (unlikely(f2fs_cp_error(sbi))) |
cf779cab | 233 | goto redirty_out; |
127e670a | 234 | |
3cb5ad15 | 235 | f2fs_wait_on_page_writeback(page, META); |
577e3495 JK |
236 | write_meta_page(sbi, page); |
237 | dec_page_count(sbi, F2FS_DIRTY_META); | |
238 | unlock_page(page); | |
857dc4e0 | 239 | |
8d4ea29b | 240 | if (wbc->for_reclaim || unlikely(f2fs_cp_error(sbi))) |
857dc4e0 | 241 | f2fs_submit_merged_bio(sbi, META, WRITE); |
577e3495 | 242 | return 0; |
cfb271d4 CY |
243 | |
244 | redirty_out: | |
76f60268 | 245 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 246 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
247 | } |
248 | ||
249 | static int f2fs_write_meta_pages(struct address_space *mapping, | |
250 | struct writeback_control *wbc) | |
251 | { | |
4081363f | 252 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 253 | long diff, written; |
127e670a | 254 | |
e5748434 CY |
255 | trace_f2fs_writepages(mapping->host, wbc, META); |
256 | ||
5459aa97 | 257 | /* collect a number of dirty meta pages and write together */ |
50c8cdb3 JK |
258 | if (wbc->for_kupdate || |
259 | get_pages(sbi, F2FS_DIRTY_META) < nr_pages_to_skip(sbi, META)) | |
d3baf95d | 260 | goto skip_write; |
127e670a JK |
261 | |
262 | /* if mounting is failed, skip writing node pages */ | |
263 | mutex_lock(&sbi->cp_mutex); | |
50c8cdb3 JK |
264 | diff = nr_pages_to_write(sbi, META, wbc); |
265 | written = sync_meta_pages(sbi, META, wbc->nr_to_write); | |
127e670a | 266 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 267 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 268 | return 0; |
d3baf95d JK |
269 | |
270 | skip_write: | |
271 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
272 | return 0; | |
127e670a JK |
273 | } |
274 | ||
275 | long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, | |
276 | long nr_to_write) | |
277 | { | |
9df27d98 | 278 | struct address_space *mapping = META_MAPPING(sbi); |
6066d8cd | 279 | pgoff_t index = 0, end = LONG_MAX, prev = LONG_MAX; |
127e670a JK |
280 | struct pagevec pvec; |
281 | long nwritten = 0; | |
282 | struct writeback_control wbc = { | |
283 | .for_reclaim = 0, | |
284 | }; | |
285 | ||
286 | pagevec_init(&pvec, 0); | |
287 | ||
288 | while (index <= end) { | |
289 | int i, nr_pages; | |
290 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, | |
291 | PAGECACHE_TAG_DIRTY, | |
292 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
cfb271d4 | 293 | if (unlikely(nr_pages == 0)) |
127e670a JK |
294 | break; |
295 | ||
296 | for (i = 0; i < nr_pages; i++) { | |
297 | struct page *page = pvec.pages[i]; | |
203681f6 | 298 | |
6066d8cd JK |
299 | if (prev == LONG_MAX) |
300 | prev = page->index - 1; | |
301 | if (nr_to_write != LONG_MAX && page->index != prev + 1) { | |
302 | pagevec_release(&pvec); | |
303 | goto stop; | |
304 | } | |
305 | ||
127e670a | 306 | lock_page(page); |
203681f6 JK |
307 | |
308 | if (unlikely(page->mapping != mapping)) { | |
309 | continue_unlock: | |
310 | unlock_page(page); | |
311 | continue; | |
312 | } | |
313 | if (!PageDirty(page)) { | |
314 | /* someone wrote it for us */ | |
315 | goto continue_unlock; | |
316 | } | |
317 | ||
318 | if (!clear_page_dirty_for_io(page)) | |
319 | goto continue_unlock; | |
320 | ||
97dc3fd2 | 321 | if (mapping->a_ops->writepage(page, &wbc)) { |
577e3495 JK |
322 | unlock_page(page); |
323 | break; | |
324 | } | |
cfb271d4 | 325 | nwritten++; |
6066d8cd | 326 | prev = page->index; |
cfb271d4 | 327 | if (unlikely(nwritten >= nr_to_write)) |
127e670a JK |
328 | break; |
329 | } | |
330 | pagevec_release(&pvec); | |
331 | cond_resched(); | |
332 | } | |
6066d8cd | 333 | stop: |
127e670a | 334 | if (nwritten) |
458e6197 | 335 | f2fs_submit_merged_bio(sbi, type, WRITE); |
127e670a JK |
336 | |
337 | return nwritten; | |
338 | } | |
339 | ||
340 | static int f2fs_set_meta_page_dirty(struct page *page) | |
341 | { | |
26c6b887 JK |
342 | trace_f2fs_set_page_dirty(page, META); |
343 | ||
127e670a JK |
344 | SetPageUptodate(page); |
345 | if (!PageDirty(page)) { | |
346 | __set_page_dirty_nobuffers(page); | |
4081363f | 347 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); |
1601839e | 348 | SetPagePrivate(page); |
9e4ded3f | 349 | f2fs_trace_pid(page); |
127e670a JK |
350 | return 1; |
351 | } | |
352 | return 0; | |
353 | } | |
354 | ||
355 | const struct address_space_operations f2fs_meta_aops = { | |
356 | .writepage = f2fs_write_meta_page, | |
357 | .writepages = f2fs_write_meta_pages, | |
358 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
487261f3 CY |
359 | .invalidatepage = f2fs_invalidate_page, |
360 | .releasepage = f2fs_release_page, | |
127e670a JK |
361 | }; |
362 | ||
6451e041 | 363 | static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 364 | { |
67298804 | 365 | struct inode_management *im = &sbi->im[type]; |
80c54505 JK |
366 | struct ino_entry *e, *tmp; |
367 | ||
368 | tmp = f2fs_kmem_cache_alloc(ino_entry_slab, GFP_NOFS); | |
39efac41 | 369 | retry: |
80c54505 | 370 | radix_tree_preload(GFP_NOFS | __GFP_NOFAIL); |
769ec6e5 | 371 | |
67298804 | 372 | spin_lock(&im->ino_lock); |
67298804 | 373 | e = radix_tree_lookup(&im->ino_root, ino); |
39efac41 | 374 | if (!e) { |
80c54505 | 375 | e = tmp; |
67298804 CY |
376 | if (radix_tree_insert(&im->ino_root, ino, e)) { |
377 | spin_unlock(&im->ino_lock); | |
769ec6e5 | 378 | radix_tree_preload_end(); |
39efac41 JK |
379 | goto retry; |
380 | } | |
381 | memset(e, 0, sizeof(struct ino_entry)); | |
382 | e->ino = ino; | |
953e6cc6 | 383 | |
67298804 | 384 | list_add_tail(&e->list, &im->ino_list); |
8c402946 | 385 | if (type != ORPHAN_INO) |
67298804 | 386 | im->ino_num++; |
39efac41 | 387 | } |
67298804 | 388 | spin_unlock(&im->ino_lock); |
769ec6e5 | 389 | radix_tree_preload_end(); |
80c54505 JK |
390 | |
391 | if (e != tmp) | |
392 | kmem_cache_free(ino_entry_slab, tmp); | |
953e6cc6 JK |
393 | } |
394 | ||
6451e041 | 395 | static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 396 | { |
67298804 | 397 | struct inode_management *im = &sbi->im[type]; |
6451e041 | 398 | struct ino_entry *e; |
953e6cc6 | 399 | |
67298804 CY |
400 | spin_lock(&im->ino_lock); |
401 | e = radix_tree_lookup(&im->ino_root, ino); | |
39efac41 JK |
402 | if (e) { |
403 | list_del(&e->list); | |
67298804 CY |
404 | radix_tree_delete(&im->ino_root, ino); |
405 | im->ino_num--; | |
406 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
407 | kmem_cache_free(ino_entry_slab, e); |
408 | return; | |
953e6cc6 | 409 | } |
67298804 | 410 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
411 | } |
412 | ||
a49324f1 | 413 | void add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
414 | { |
415 | /* add new dirty ino entry into list */ | |
416 | __add_ino_entry(sbi, ino, type); | |
417 | } | |
418 | ||
a49324f1 | 419 | void remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
420 | { |
421 | /* remove dirty ino entry from list */ | |
422 | __remove_ino_entry(sbi, ino, type); | |
423 | } | |
424 | ||
425 | /* mode should be APPEND_INO or UPDATE_INO */ | |
426 | bool exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode) | |
427 | { | |
67298804 | 428 | struct inode_management *im = &sbi->im[mode]; |
fff04f90 | 429 | struct ino_entry *e; |
67298804 CY |
430 | |
431 | spin_lock(&im->ino_lock); | |
432 | e = radix_tree_lookup(&im->ino_root, ino); | |
433 | spin_unlock(&im->ino_lock); | |
fff04f90 JK |
434 | return e ? true : false; |
435 | } | |
436 | ||
a49324f1 | 437 | void release_ino_entry(struct f2fs_sb_info *sbi) |
fff04f90 JK |
438 | { |
439 | struct ino_entry *e, *tmp; | |
440 | int i; | |
441 | ||
442 | for (i = APPEND_INO; i <= UPDATE_INO; i++) { | |
67298804 CY |
443 | struct inode_management *im = &sbi->im[i]; |
444 | ||
445 | spin_lock(&im->ino_lock); | |
446 | list_for_each_entry_safe(e, tmp, &im->ino_list, list) { | |
fff04f90 | 447 | list_del(&e->list); |
67298804 | 448 | radix_tree_delete(&im->ino_root, e->ino); |
fff04f90 | 449 | kmem_cache_free(ino_entry_slab, e); |
67298804 | 450 | im->ino_num--; |
fff04f90 | 451 | } |
67298804 | 452 | spin_unlock(&im->ino_lock); |
fff04f90 JK |
453 | } |
454 | } | |
455 | ||
cbd56e7d | 456 | int acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 457 | { |
67298804 | 458 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a JK |
459 | int err = 0; |
460 | ||
67298804 CY |
461 | spin_lock(&im->ino_lock); |
462 | if (unlikely(im->ino_num >= sbi->max_orphans)) | |
127e670a | 463 | err = -ENOSPC; |
cbd56e7d | 464 | else |
67298804 CY |
465 | im->ino_num++; |
466 | spin_unlock(&im->ino_lock); | |
0d47c1ad | 467 | |
127e670a JK |
468 | return err; |
469 | } | |
470 | ||
cbd56e7d JK |
471 | void release_orphan_inode(struct f2fs_sb_info *sbi) |
472 | { | |
67298804 CY |
473 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
474 | ||
475 | spin_lock(&im->ino_lock); | |
476 | f2fs_bug_on(sbi, im->ino_num == 0); | |
477 | im->ino_num--; | |
478 | spin_unlock(&im->ino_lock); | |
cbd56e7d JK |
479 | } |
480 | ||
127e670a JK |
481 | void add_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
482 | { | |
39efac41 | 483 | /* add new orphan ino entry into list */ |
6451e041 | 484 | __add_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
485 | } |
486 | ||
487 | void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
488 | { | |
953e6cc6 | 489 | /* remove orphan entry from orphan list */ |
6451e041 | 490 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
491 | } |
492 | ||
8c14bfad | 493 | static int recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 494 | { |
8c14bfad CY |
495 | struct inode *inode; |
496 | ||
497 | inode = f2fs_iget(sbi->sb, ino); | |
498 | if (IS_ERR(inode)) { | |
499 | /* | |
500 | * there should be a bug that we can't find the entry | |
501 | * to orphan inode. | |
502 | */ | |
503 | f2fs_bug_on(sbi, PTR_ERR(inode) == -ENOENT); | |
504 | return PTR_ERR(inode); | |
505 | } | |
506 | ||
127e670a JK |
507 | clear_nlink(inode); |
508 | ||
509 | /* truncate all the data during iput */ | |
510 | iput(inode); | |
8c14bfad | 511 | return 0; |
127e670a JK |
512 | } |
513 | ||
8c14bfad | 514 | int recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a | 515 | { |
3c642985 | 516 | block_t start_blk, orphan_blocks, i, j; |
8c14bfad | 517 | int err; |
127e670a | 518 | |
25ca923b | 519 | if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG)) |
8c14bfad | 520 | return 0; |
127e670a | 521 | |
55141486 | 522 | start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi); |
3c642985 | 523 | orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi); |
127e670a | 524 | |
26879fb1 | 525 | ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP, true); |
662befda | 526 | |
3c642985 | 527 | for (i = 0; i < orphan_blocks; i++) { |
127e670a JK |
528 | struct page *page = get_meta_page(sbi, start_blk + i); |
529 | struct f2fs_orphan_block *orphan_blk; | |
530 | ||
531 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); | |
532 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
533 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
8c14bfad CY |
534 | err = recover_orphan_inode(sbi, ino); |
535 | if (err) { | |
536 | f2fs_put_page(page, 1); | |
8c14bfad CY |
537 | return err; |
538 | } | |
127e670a JK |
539 | } |
540 | f2fs_put_page(page, 1); | |
541 | } | |
542 | /* clear Orphan Flag */ | |
25ca923b | 543 | clear_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG); |
8c14bfad | 544 | return 0; |
127e670a JK |
545 | } |
546 | ||
547 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
548 | { | |
502c6e0b | 549 | struct list_head *head; |
127e670a | 550 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 551 | unsigned int nentries = 0; |
bd936f84 | 552 | unsigned short index = 1; |
8c402946 | 553 | unsigned short orphan_blocks; |
4531929e | 554 | struct page *page = NULL; |
6451e041 | 555 | struct ino_entry *orphan = NULL; |
67298804 | 556 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 557 | |
67298804 | 558 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 559 | |
d6c67a4f JK |
560 | /* |
561 | * we don't need to do spin_lock(&im->ino_lock) here, since all the | |
562 | * orphan inode operations are covered under f2fs_lock_op(). | |
563 | * And, spin_lock should be avoided due to page operations below. | |
564 | */ | |
67298804 | 565 | head = &im->ino_list; |
127e670a JK |
566 | |
567 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
568 | list_for_each_entry(orphan, head, list) { |
569 | if (!page) { | |
bd936f84 | 570 | page = grab_meta_page(sbi, start_blk++); |
502c6e0b GZ |
571 | orphan_blk = |
572 | (struct f2fs_orphan_block *)page_address(page); | |
573 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
574 | } | |
127e670a | 575 | |
36795567 | 576 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 577 | |
36795567 | 578 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
579 | /* |
580 | * an orphan block is full of 1020 entries, | |
581 | * then we need to flush current orphan blocks | |
582 | * and bring another one in memory | |
583 | */ | |
584 | orphan_blk->blk_addr = cpu_to_le16(index); | |
585 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
586 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
587 | set_page_dirty(page); | |
588 | f2fs_put_page(page, 1); | |
589 | index++; | |
127e670a JK |
590 | nentries = 0; |
591 | page = NULL; | |
592 | } | |
502c6e0b | 593 | } |
127e670a | 594 | |
502c6e0b GZ |
595 | if (page) { |
596 | orphan_blk->blk_addr = cpu_to_le16(index); | |
597 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
598 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
599 | set_page_dirty(page); | |
600 | f2fs_put_page(page, 1); | |
127e670a | 601 | } |
127e670a JK |
602 | } |
603 | ||
604 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, | |
605 | block_t cp_addr, unsigned long long *version) | |
606 | { | |
607 | struct page *cp_page_1, *cp_page_2 = NULL; | |
608 | unsigned long blk_size = sbi->blocksize; | |
609 | struct f2fs_checkpoint *cp_block; | |
610 | unsigned long long cur_version = 0, pre_version = 0; | |
127e670a | 611 | size_t crc_offset; |
7e586fa0 | 612 | __u32 crc = 0; |
127e670a JK |
613 | |
614 | /* Read the 1st cp block in this CP pack */ | |
615 | cp_page_1 = get_meta_page(sbi, cp_addr); | |
616 | ||
617 | /* get the version number */ | |
618 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_1); | |
619 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
620 | if (crc_offset >= blk_size) | |
621 | goto invalid_cp1; | |
622 | ||
29e7043f | 623 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
624 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
625 | goto invalid_cp1; | |
626 | ||
d71b5564 | 627 | pre_version = cur_cp_version(cp_block); |
127e670a JK |
628 | |
629 | /* Read the 2nd cp block in this CP pack */ | |
25ca923b | 630 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
127e670a JK |
631 | cp_page_2 = get_meta_page(sbi, cp_addr); |
632 | ||
633 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_2); | |
634 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
635 | if (crc_offset >= blk_size) | |
636 | goto invalid_cp2; | |
637 | ||
29e7043f | 638 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
639 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
640 | goto invalid_cp2; | |
641 | ||
d71b5564 | 642 | cur_version = cur_cp_version(cp_block); |
127e670a JK |
643 | |
644 | if (cur_version == pre_version) { | |
645 | *version = cur_version; | |
646 | f2fs_put_page(cp_page_2, 1); | |
647 | return cp_page_1; | |
648 | } | |
649 | invalid_cp2: | |
650 | f2fs_put_page(cp_page_2, 1); | |
651 | invalid_cp1: | |
652 | f2fs_put_page(cp_page_1, 1); | |
653 | return NULL; | |
654 | } | |
655 | ||
656 | int get_valid_checkpoint(struct f2fs_sb_info *sbi) | |
657 | { | |
658 | struct f2fs_checkpoint *cp_block; | |
659 | struct f2fs_super_block *fsb = sbi->raw_super; | |
660 | struct page *cp1, *cp2, *cur_page; | |
661 | unsigned long blk_size = sbi->blocksize; | |
662 | unsigned long long cp1_version = 0, cp2_version = 0; | |
663 | unsigned long long cp_start_blk_no; | |
55141486 | 664 | unsigned int cp_blks = 1 + __cp_payload(sbi); |
1dbe4152 CL |
665 | block_t cp_blk_no; |
666 | int i; | |
127e670a | 667 | |
1dbe4152 | 668 | sbi->ckpt = kzalloc(cp_blks * blk_size, GFP_KERNEL); |
127e670a JK |
669 | if (!sbi->ckpt) |
670 | return -ENOMEM; | |
671 | /* | |
672 | * Finding out valid cp block involves read both | |
673 | * sets( cp pack1 and cp pack 2) | |
674 | */ | |
675 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
676 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
677 | ||
678 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
679 | cp_start_blk_no += ((unsigned long long)1) << |
680 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
681 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
682 | ||
683 | if (cp1 && cp2) { | |
684 | if (ver_after(cp2_version, cp1_version)) | |
685 | cur_page = cp2; | |
686 | else | |
687 | cur_page = cp1; | |
688 | } else if (cp1) { | |
689 | cur_page = cp1; | |
690 | } else if (cp2) { | |
691 | cur_page = cp2; | |
692 | } else { | |
693 | goto fail_no_cp; | |
694 | } | |
695 | ||
696 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
697 | memcpy(sbi->ckpt, cp_block, blk_size); | |
698 | ||
1dbe4152 CL |
699 | if (cp_blks <= 1) |
700 | goto done; | |
701 | ||
702 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
703 | if (cur_page == cp2) | |
704 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
705 | ||
706 | for (i = 1; i < cp_blks; i++) { | |
707 | void *sit_bitmap_ptr; | |
708 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
709 | ||
710 | cur_page = get_meta_page(sbi, cp_blk_no + i); | |
711 | sit_bitmap_ptr = page_address(cur_page); | |
712 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
713 | f2fs_put_page(cur_page, 1); | |
714 | } | |
715 | done: | |
127e670a JK |
716 | f2fs_put_page(cp1, 1); |
717 | f2fs_put_page(cp2, 1); | |
718 | return 0; | |
719 | ||
720 | fail_no_cp: | |
721 | kfree(sbi->ckpt); | |
722 | return -EINVAL; | |
723 | } | |
724 | ||
c227f912 | 725 | static void __add_dirty_inode(struct inode *inode, enum inode_type type) |
127e670a | 726 | { |
4081363f | 727 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
2710fd7e | 728 | struct f2fs_inode_info *fi = F2FS_I(inode); |
c227f912 | 729 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
127e670a | 730 | |
c227f912 | 731 | if (is_inode_flag_set(fi, flag)) |
2710fd7e | 732 | return; |
2d7b822a | 733 | |
c227f912 CY |
734 | set_inode_flag(fi, flag); |
735 | list_add_tail(&fi->dirty_list, &sbi->inode_list[type]); | |
33fbd510 | 736 | stat_inc_dirty_inode(sbi, type); |
5deb8267 JK |
737 | } |
738 | ||
c227f912 | 739 | static void __remove_dirty_inode(struct inode *inode, enum inode_type type) |
6ad7609a | 740 | { |
6ad7609a | 741 | struct f2fs_inode_info *fi = F2FS_I(inode); |
c227f912 | 742 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
6ad7609a CY |
743 | |
744 | if (get_dirty_pages(inode) || | |
c227f912 | 745 | !is_inode_flag_set(F2FS_I(inode), flag)) |
6ad7609a CY |
746 | return; |
747 | ||
748 | list_del_init(&fi->dirty_list); | |
c227f912 | 749 | clear_inode_flag(fi, flag); |
33fbd510 | 750 | stat_dec_dirty_inode(F2FS_I_SB(inode), type); |
6ad7609a CY |
751 | } |
752 | ||
a7ffdbe2 | 753 | void update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 754 | { |
4081363f | 755 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 756 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
5deb8267 | 757 | |
5ac9f36f CY |
758 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
759 | !S_ISLNK(inode->i_mode)) | |
127e670a | 760 | return; |
7bd59381 | 761 | |
c227f912 CY |
762 | spin_lock(&sbi->inode_lock[type]); |
763 | __add_dirty_inode(inode, type); | |
a7ffdbe2 | 764 | inode_inc_dirty_pages(inode); |
c227f912 | 765 | spin_unlock(&sbi->inode_lock[type]); |
cf0ee0f0 | 766 | |
a7ffdbe2 | 767 | SetPagePrivate(page); |
9e4ded3f | 768 | f2fs_trace_pid(page); |
5deb8267 JK |
769 | } |
770 | ||
771 | void add_dirty_dir_inode(struct inode *inode) | |
772 | { | |
4081363f | 773 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
127e670a | 774 | |
c227f912 CY |
775 | spin_lock(&sbi->inode_lock[DIR_INODE]); |
776 | __add_dirty_inode(inode, DIR_INODE); | |
777 | spin_unlock(&sbi->inode_lock[DIR_INODE]); | |
127e670a JK |
778 | } |
779 | ||
c227f912 | 780 | void remove_dirty_inode(struct inode *inode) |
127e670a | 781 | { |
4081363f | 782 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
2710fd7e | 783 | struct f2fs_inode_info *fi = F2FS_I(inode); |
c227f912 | 784 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
127e670a | 785 | |
c227f912 CY |
786 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
787 | !S_ISLNK(inode->i_mode)) | |
127e670a JK |
788 | return; |
789 | ||
c227f912 CY |
790 | spin_lock(&sbi->inode_lock[type]); |
791 | __remove_dirty_inode(inode, type); | |
792 | spin_unlock(&sbi->inode_lock[type]); | |
74d0b917 JK |
793 | |
794 | /* Only from the recovery routine */ | |
2710fd7e CY |
795 | if (is_inode_flag_set(fi, FI_DELAY_IPUT)) { |
796 | clear_inode_flag(fi, FI_DELAY_IPUT); | |
74d0b917 | 797 | iput(inode); |
afc3eda2 | 798 | } |
74d0b917 JK |
799 | } |
800 | ||
6d5a1495 | 801 | int sync_dirty_inodes(struct f2fs_sb_info *sbi, enum inode_type type) |
127e670a | 802 | { |
ce3b7d80 | 803 | struct list_head *head; |
127e670a | 804 | struct inode *inode; |
2710fd7e | 805 | struct f2fs_inode_info *fi; |
4cf18537 CY |
806 | bool is_dir = (type == DIR_INODE); |
807 | ||
808 | trace_f2fs_sync_dirty_inodes_enter(sbi->sb, is_dir, | |
809 | get_pages(sbi, is_dir ? | |
810 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
127e670a | 811 | retry: |
af41d3ee | 812 | if (unlikely(f2fs_cp_error(sbi))) |
6d5a1495 | 813 | return -EIO; |
af41d3ee | 814 | |
c227f912 | 815 | spin_lock(&sbi->inode_lock[type]); |
ce3b7d80 | 816 | |
c227f912 | 817 | head = &sbi->inode_list[type]; |
127e670a | 818 | if (list_empty(head)) { |
c227f912 | 819 | spin_unlock(&sbi->inode_lock[type]); |
4cf18537 CY |
820 | trace_f2fs_sync_dirty_inodes_exit(sbi->sb, is_dir, |
821 | get_pages(sbi, is_dir ? | |
822 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
6d5a1495 | 823 | return 0; |
127e670a | 824 | } |
2710fd7e CY |
825 | fi = list_entry(head->next, struct f2fs_inode_info, dirty_list); |
826 | inode = igrab(&fi->vfs_inode); | |
c227f912 | 827 | spin_unlock(&sbi->inode_lock[type]); |
127e670a | 828 | if (inode) { |
87d6f890 | 829 | filemap_fdatawrite(inode->i_mapping); |
127e670a JK |
830 | iput(inode); |
831 | } else { | |
832 | /* | |
833 | * We should submit bio, since it exists several | |
834 | * wribacking dentry pages in the freeing inode. | |
835 | */ | |
458e6197 | 836 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
7ecebe5e | 837 | cond_resched(); |
127e670a JK |
838 | } |
839 | goto retry; | |
840 | } | |
841 | ||
0a8165d7 | 842 | /* |
127e670a JK |
843 | * Freeze all the FS-operations for checkpoint. |
844 | */ | |
cf779cab | 845 | static int block_operations(struct f2fs_sb_info *sbi) |
127e670a | 846 | { |
127e670a JK |
847 | struct writeback_control wbc = { |
848 | .sync_mode = WB_SYNC_ALL, | |
849 | .nr_to_write = LONG_MAX, | |
850 | .for_reclaim = 0, | |
851 | }; | |
c718379b | 852 | struct blk_plug plug; |
cf779cab | 853 | int err = 0; |
c718379b JK |
854 | |
855 | blk_start_plug(&plug); | |
856 | ||
39936837 | 857 | retry_flush_dents: |
e479556b | 858 | f2fs_lock_all(sbi); |
127e670a | 859 | /* write all the dirty dentry pages */ |
127e670a | 860 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 861 | f2fs_unlock_all(sbi); |
6d5a1495 CY |
862 | err = sync_dirty_inodes(sbi, DIR_INODE); |
863 | if (err) | |
cf779cab | 864 | goto out; |
39936837 | 865 | goto retry_flush_dents; |
127e670a JK |
866 | } |
867 | ||
127e670a | 868 | /* |
e1c42045 | 869 | * POR: we should ensure that there are no dirty node pages |
127e670a JK |
870 | * until finishing nat/sit flush. |
871 | */ | |
39936837 | 872 | retry_flush_nodes: |
b3582c68 | 873 | down_write(&sbi->node_write); |
127e670a JK |
874 | |
875 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
b3582c68 | 876 | up_write(&sbi->node_write); |
6d5a1495 CY |
877 | err = sync_node_pages(sbi, 0, &wbc); |
878 | if (err) { | |
cf779cab | 879 | f2fs_unlock_all(sbi); |
cf779cab JK |
880 | goto out; |
881 | } | |
39936837 | 882 | goto retry_flush_nodes; |
127e670a | 883 | } |
cf779cab | 884 | out: |
c718379b | 885 | blk_finish_plug(&plug); |
cf779cab | 886 | return err; |
127e670a JK |
887 | } |
888 | ||
889 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
890 | { | |
b3582c68 | 891 | up_write(&sbi->node_write); |
e479556b | 892 | f2fs_unlock_all(sbi); |
127e670a JK |
893 | } |
894 | ||
fb51b5ef CL |
895 | static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi) |
896 | { | |
897 | DEFINE_WAIT(wait); | |
898 | ||
899 | for (;;) { | |
900 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
901 | ||
902 | if (!get_pages(sbi, F2FS_WRITEBACK)) | |
903 | break; | |
904 | ||
905 | io_schedule(); | |
906 | } | |
907 | finish_wait(&sbi->cp_wait, &wait); | |
908 | } | |
909 | ||
c34f42e2 | 910 | static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
911 | { |
912 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
cf2271e7 | 913 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); |
77041823 | 914 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
67298804 | 915 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; |
77041823 | 916 | nid_t last_nid = nm_i->next_scan_nid; |
127e670a | 917 | block_t start_blk; |
127e670a | 918 | unsigned int data_sum_blocks, orphan_blocks; |
7e586fa0 | 919 | __u32 crc32 = 0; |
127e670a | 920 | int i; |
55141486 | 921 | int cp_payload_blks = __cp_payload(sbi); |
e90c2d28 CY |
922 | block_t discard_blk = NEXT_FREE_BLKADDR(sbi, curseg); |
923 | bool invalidate = false; | |
127e670a | 924 | |
1e87a78d JK |
925 | /* |
926 | * This avoids to conduct wrong roll-forward operations and uses | |
927 | * metapages, so should be called prior to sync_meta_pages below. | |
928 | */ | |
e90c2d28 CY |
929 | if (discard_next_dnode(sbi, discard_blk)) |
930 | invalidate = true; | |
127e670a JK |
931 | |
932 | /* Flush all the NAT/SIT pages */ | |
cf779cab | 933 | while (get_pages(sbi, F2FS_DIRTY_META)) { |
127e670a | 934 | sync_meta_pages(sbi, META, LONG_MAX); |
cf779cab | 935 | if (unlikely(f2fs_cp_error(sbi))) |
c34f42e2 | 936 | return -EIO; |
cf779cab | 937 | } |
127e670a JK |
938 | |
939 | next_free_nid(sbi, &last_nid); | |
940 | ||
941 | /* | |
942 | * modify checkpoint | |
943 | * version number is already updated | |
944 | */ | |
945 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi)); | |
946 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
947 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); | |
b5b82205 | 948 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
949 | ckpt->cur_node_segno[i] = |
950 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
951 | ckpt->cur_node_blkoff[i] = | |
952 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
953 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
954 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
955 | } | |
b5b82205 | 956 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
957 | ckpt->cur_data_segno[i] = |
958 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
959 | ckpt->cur_data_blkoff[i] = | |
960 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
961 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
962 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
963 | } | |
964 | ||
965 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
966 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
967 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
968 | ||
969 | /* 2 cp + n data seg summary + orphan inode blocks */ | |
3fa06d7b | 970 | data_sum_blocks = npages_for_summary_flush(sbi, false); |
b5b82205 | 971 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
25ca923b | 972 | set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 973 | else |
25ca923b | 974 | clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 975 | |
67298804 | 976 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
977 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
978 | orphan_blocks); | |
127e670a | 979 | |
119ee914 | 980 | if (__remain_node_summaries(cpc->reason)) |
b5b82205 | 981 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
982 | cp_payload_blks + data_sum_blocks + |
983 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
119ee914 | 984 | else |
b5b82205 | 985 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS + |
1dbe4152 CL |
986 | cp_payload_blks + data_sum_blocks + |
987 | orphan_blocks); | |
119ee914 JK |
988 | |
989 | if (cpc->reason == CP_UMOUNT) | |
990 | set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
991 | else | |
992 | clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
993 | ||
994 | if (cpc->reason == CP_FASTBOOT) | |
995 | set_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
996 | else | |
997 | clear_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
127e670a | 998 | |
67298804 | 999 | if (orphan_num) |
25ca923b | 1000 | set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1001 | else |
25ca923b | 1002 | clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1003 | |
caf0047e | 1004 | if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) |
2ae4c673 JK |
1005 | set_ckpt_flags(ckpt, CP_FSCK_FLAG); |
1006 | ||
127e670a JK |
1007 | /* update SIT/NAT bitmap */ |
1008 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
1009 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
1010 | ||
1011 | crc32 = f2fs_crc32(ckpt, le32_to_cpu(ckpt->checksum_offset)); | |
7e586fa0 JK |
1012 | *((__le32 *)((unsigned char *)ckpt + |
1013 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
1014 | = cpu_to_le32(crc32); |
1015 | ||
1016 | start_blk = __start_cp_addr(sbi); | |
1017 | ||
a7230d16 JK |
1018 | /* need to wait for end_io results */ |
1019 | wait_on_all_pages_writeback(sbi); | |
1020 | if (unlikely(f2fs_cp_error(sbi))) | |
c34f42e2 | 1021 | return -EIO; |
a7230d16 | 1022 | |
127e670a | 1023 | /* write out checkpoint buffer at block 0 */ |
381722d2 CY |
1024 | update_meta_page(sbi, ckpt, start_blk++); |
1025 | ||
1026 | for (i = 1; i < 1 + cp_payload_blks; i++) | |
1027 | update_meta_page(sbi, (char *)ckpt + i * F2FS_BLKSIZE, | |
1028 | start_blk++); | |
1dbe4152 | 1029 | |
67298804 | 1030 | if (orphan_num) { |
127e670a JK |
1031 | write_orphan_inodes(sbi, start_blk); |
1032 | start_blk += orphan_blocks; | |
1033 | } | |
1034 | ||
1035 | write_data_summaries(sbi, start_blk); | |
1036 | start_blk += data_sum_blocks; | |
119ee914 | 1037 | if (__remain_node_summaries(cpc->reason)) { |
127e670a JK |
1038 | write_node_summaries(sbi, start_blk); |
1039 | start_blk += NR_CURSEG_NODE_TYPE; | |
1040 | } | |
1041 | ||
1042 | /* writeout checkpoint block */ | |
381722d2 | 1043 | update_meta_page(sbi, ckpt, start_blk); |
127e670a JK |
1044 | |
1045 | /* wait for previous submitted node/meta pages writeback */ | |
fb51b5ef | 1046 | wait_on_all_pages_writeback(sbi); |
127e670a | 1047 | |
cf779cab | 1048 | if (unlikely(f2fs_cp_error(sbi))) |
c34f42e2 | 1049 | return -EIO; |
cf779cab | 1050 | |
4ef51a8f | 1051 | filemap_fdatawait_range(NODE_MAPPING(sbi), 0, LONG_MAX); |
9df27d98 | 1052 | filemap_fdatawait_range(META_MAPPING(sbi), 0, LONG_MAX); |
127e670a JK |
1053 | |
1054 | /* update user_block_counts */ | |
1055 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
1056 | sbi->alloc_valid_block_count = 0; | |
1057 | ||
1058 | /* Here, we only have one bio having CP pack */ | |
577e3495 | 1059 | sync_meta_pages(sbi, META_FLUSH, LONG_MAX); |
127e670a | 1060 | |
6a8f8ca5 JK |
1061 | /* wait for previous submitted meta pages writeback */ |
1062 | wait_on_all_pages_writeback(sbi); | |
1063 | ||
e90c2d28 CY |
1064 | /* |
1065 | * invalidate meta page which is used temporarily for zeroing out | |
1066 | * block at the end of warm node chain. | |
1067 | */ | |
1068 | if (invalidate) | |
1069 | invalidate_mapping_pages(META_MAPPING(sbi), discard_blk, | |
1070 | discard_blk); | |
1071 | ||
a49324f1 | 1072 | release_ino_entry(sbi); |
cf779cab JK |
1073 | |
1074 | if (unlikely(f2fs_cp_error(sbi))) | |
c34f42e2 | 1075 | return -EIO; |
cf779cab | 1076 | |
836b5a63 | 1077 | clear_prefree_segments(sbi, cpc); |
caf0047e | 1078 | clear_sbi_flag(sbi, SBI_IS_DIRTY); |
c34f42e2 CY |
1079 | |
1080 | return 0; | |
127e670a JK |
1081 | } |
1082 | ||
0a8165d7 | 1083 | /* |
e1c42045 | 1084 | * We guarantee that this checkpoint procedure will not fail. |
127e670a | 1085 | */ |
c34f42e2 | 1086 | int write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1087 | { |
1088 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1089 | unsigned long long ckpt_ver; | |
c34f42e2 | 1090 | int err = 0; |
127e670a | 1091 | |
43727527 | 1092 | mutex_lock(&sbi->cp_mutex); |
8501017e | 1093 | |
caf0047e | 1094 | if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) && |
a66cdd98 JK |
1095 | (cpc->reason == CP_FASTBOOT || cpc->reason == CP_SYNC || |
1096 | (cpc->reason == CP_DISCARD && !sbi->discard_blks))) | |
8501017e | 1097 | goto out; |
c34f42e2 CY |
1098 | if (unlikely(f2fs_cp_error(sbi))) { |
1099 | err = -EIO; | |
cf779cab | 1100 | goto out; |
c34f42e2 CY |
1101 | } |
1102 | if (f2fs_readonly(sbi->sb)) { | |
1103 | err = -EROFS; | |
11504a8e | 1104 | goto out; |
c34f42e2 | 1105 | } |
2bda542d WL |
1106 | |
1107 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); | |
1108 | ||
c34f42e2 CY |
1109 | err = block_operations(sbi); |
1110 | if (err) | |
cf779cab | 1111 | goto out; |
127e670a | 1112 | |
75ab4cb8 | 1113 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1114 | |
458e6197 JK |
1115 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
1116 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
1117 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
127e670a JK |
1118 | |
1119 | /* | |
1120 | * update checkpoint pack index | |
1121 | * Increase the version number so that | |
1122 | * SIT entries and seg summaries are written at correct place | |
1123 | */ | |
d71b5564 | 1124 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1125 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1126 | ||
1127 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
1128 | flush_nat_entries(sbi); | |
4b2fecc8 | 1129 | flush_sit_entries(sbi, cpc); |
127e670a | 1130 | |
127e670a | 1131 | /* unlock all the fs_lock[] in do_checkpoint() */ |
c34f42e2 | 1132 | err = do_checkpoint(sbi, cpc); |
127e670a JK |
1133 | |
1134 | unblock_operations(sbi); | |
942e0be6 | 1135 | stat_inc_cp_count(sbi->stat_info); |
10027551 JK |
1136 | |
1137 | if (cpc->reason == CP_RECOVERY) | |
1138 | f2fs_msg(sbi->sb, KERN_NOTICE, | |
1139 | "checkpoint: version = %llx", ckpt_ver); | |
60b99b48 JK |
1140 | |
1141 | /* do checkpoint periodically */ | |
6beceb54 | 1142 | f2fs_update_time(sbi, CP_TIME); |
55d1cdb2 | 1143 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
8501017e JK |
1144 | out: |
1145 | mutex_unlock(&sbi->cp_mutex); | |
c34f42e2 | 1146 | return err; |
127e670a JK |
1147 | } |
1148 | ||
6451e041 | 1149 | void init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1150 | { |
6451e041 JK |
1151 | int i; |
1152 | ||
1153 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1154 | struct inode_management *im = &sbi->im[i]; |
1155 | ||
1156 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1157 | spin_lock_init(&im->ino_lock); | |
1158 | INIT_LIST_HEAD(&im->ino_list); | |
1159 | im->ino_num = 0; | |
6451e041 JK |
1160 | } |
1161 | ||
b5b82205 | 1162 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
14b42817 WL |
1163 | NR_CURSEG_TYPE - __cp_payload(sbi)) * |
1164 | F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
1165 | } |
1166 | ||
6e6093a8 | 1167 | int __init create_checkpoint_caches(void) |
127e670a | 1168 | { |
6451e041 JK |
1169 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1170 | sizeof(struct ino_entry)); | |
1171 | if (!ino_entry_slab) | |
127e670a | 1172 | return -ENOMEM; |
06292073 CY |
1173 | inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry", |
1174 | sizeof(struct inode_entry)); | |
6bacf52f | 1175 | if (!inode_entry_slab) { |
6451e041 | 1176 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1177 | return -ENOMEM; |
1178 | } | |
1179 | return 0; | |
1180 | } | |
1181 | ||
1182 | void destroy_checkpoint_caches(void) | |
1183 | { | |
6451e041 | 1184 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1185 | kmem_cache_destroy(inode_entry_slab); |
1186 | } |