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 | |
38f91ca8 JK |
29 | void f2fs_stop_checkpoint(struct f2fs_sb_info *sbi, bool end_io) |
30 | { | |
31 | set_ckpt_flags(sbi->ckpt, CP_ERROR_FLAG); | |
32 | sbi->sb->s_flags |= MS_RDONLY; | |
33 | if (!end_io) | |
34 | f2fs_flush_merged_bios(sbi); | |
35 | } | |
36 | ||
0a8165d7 | 37 | /* |
127e670a JK |
38 | * We guarantee no failure on the returned page. |
39 | */ | |
40 | struct page *grab_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
41 | { | |
9df27d98 | 42 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
43 | struct page *page = NULL; |
44 | repeat: | |
300e129c | 45 | page = f2fs_grab_cache_page(mapping, index, false); |
127e670a JK |
46 | if (!page) { |
47 | cond_resched(); | |
48 | goto repeat; | |
49 | } | |
fec1d657 | 50 | f2fs_wait_on_page_writeback(page, META, true); |
237c0790 JK |
51 | if (!PageUptodate(page)) |
52 | SetPageUptodate(page); | |
127e670a JK |
53 | return page; |
54 | } | |
55 | ||
0a8165d7 | 56 | /* |
127e670a JK |
57 | * We guarantee no failure on the returned page. |
58 | */ | |
2b947003 CY |
59 | static struct page *__get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index, |
60 | bool is_meta) | |
127e670a | 61 | { |
9df27d98 | 62 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a | 63 | struct page *page; |
cf04e8eb | 64 | struct f2fs_io_info fio = { |
05ca3632 | 65 | .sbi = sbi, |
cf04e8eb JK |
66 | .type = META, |
67 | .rw = READ_SYNC | REQ_META | REQ_PRIO, | |
7a9d7548 CY |
68 | .old_blkaddr = index, |
69 | .new_blkaddr = index, | |
4375a336 | 70 | .encrypted_page = NULL, |
cf04e8eb | 71 | }; |
2b947003 CY |
72 | |
73 | if (unlikely(!is_meta)) | |
74 | fio.rw &= ~REQ_META; | |
127e670a | 75 | repeat: |
300e129c | 76 | page = f2fs_grab_cache_page(mapping, index, false); |
127e670a JK |
77 | if (!page) { |
78 | cond_resched(); | |
79 | goto repeat; | |
80 | } | |
393ff91f JK |
81 | if (PageUptodate(page)) |
82 | goto out; | |
83 | ||
05ca3632 JK |
84 | fio.page = page; |
85 | ||
86531d6b JK |
86 | if (f2fs_submit_page_bio(&fio)) { |
87 | f2fs_put_page(page, 1); | |
127e670a | 88 | goto repeat; |
86531d6b | 89 | } |
127e670a | 90 | |
393ff91f | 91 | lock_page(page); |
6bacf52f | 92 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
93 | f2fs_put_page(page, 1); |
94 | goto repeat; | |
95 | } | |
f3f338ca CY |
96 | |
97 | /* | |
98 | * if there is any IO error when accessing device, make our filesystem | |
99 | * readonly and make sure do not write checkpoint with non-uptodate | |
100 | * meta page. | |
101 | */ | |
102 | if (unlikely(!PageUptodate(page))) | |
38f91ca8 | 103 | f2fs_stop_checkpoint(sbi, false); |
393ff91f | 104 | out: |
127e670a JK |
105 | return page; |
106 | } | |
107 | ||
2b947003 CY |
108 | struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) |
109 | { | |
110 | return __get_meta_page(sbi, index, true); | |
111 | } | |
112 | ||
113 | /* for POR only */ | |
114 | struct page *get_tmp_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
115 | { | |
116 | return __get_meta_page(sbi, index, false); | |
117 | } | |
118 | ||
f0c9cada | 119 | bool is_valid_blkaddr(struct f2fs_sb_info *sbi, block_t blkaddr, int type) |
662befda CY |
120 | { |
121 | switch (type) { | |
122 | case META_NAT: | |
66b00c18 | 123 | break; |
662befda | 124 | case META_SIT: |
66b00c18 CY |
125 | if (unlikely(blkaddr >= SIT_BLK_CNT(sbi))) |
126 | return false; | |
127 | break; | |
81c1a0f1 | 128 | case META_SSA: |
66b00c18 CY |
129 | if (unlikely(blkaddr >= MAIN_BLKADDR(sbi) || |
130 | blkaddr < SM_I(sbi)->ssa_blkaddr)) | |
131 | return false; | |
132 | break; | |
662befda | 133 | case META_CP: |
66b00c18 CY |
134 | if (unlikely(blkaddr >= SIT_I(sbi)->sit_base_addr || |
135 | blkaddr < __start_cp_addr(sbi))) | |
136 | return false; | |
137 | break; | |
4c521f49 | 138 | case META_POR: |
66b00c18 CY |
139 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
140 | blkaddr < MAIN_BLKADDR(sbi))) | |
141 | return false; | |
142 | break; | |
662befda CY |
143 | default: |
144 | BUG(); | |
145 | } | |
66b00c18 CY |
146 | |
147 | return true; | |
662befda CY |
148 | } |
149 | ||
150 | /* | |
81c1a0f1 | 151 | * Readahead CP/NAT/SIT/SSA pages |
662befda | 152 | */ |
26879fb1 CY |
153 | int ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages, |
154 | int type, bool sync) | |
662befda | 155 | { |
662befda | 156 | struct page *page; |
4c521f49 | 157 | block_t blkno = start; |
662befda | 158 | struct f2fs_io_info fio = { |
05ca3632 | 159 | .sbi = sbi, |
662befda | 160 | .type = META, |
26879fb1 | 161 | .rw = sync ? (READ_SYNC | REQ_META | REQ_PRIO) : READA, |
4375a336 | 162 | .encrypted_page = NULL, |
662befda | 163 | }; |
e9f5b8b8 | 164 | struct blk_plug plug; |
662befda | 165 | |
2b947003 CY |
166 | if (unlikely(type == META_POR)) |
167 | fio.rw &= ~REQ_META; | |
168 | ||
e9f5b8b8 | 169 | blk_start_plug(&plug); |
662befda | 170 | for (; nrpages-- > 0; blkno++) { |
662befda | 171 | |
66b00c18 CY |
172 | if (!is_valid_blkaddr(sbi, blkno, type)) |
173 | goto out; | |
174 | ||
662befda CY |
175 | switch (type) { |
176 | case META_NAT: | |
66b00c18 CY |
177 | if (unlikely(blkno >= |
178 | NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))) | |
662befda | 179 | blkno = 0; |
66b00c18 | 180 | /* get nat block addr */ |
7a9d7548 | 181 | fio.new_blkaddr = current_nat_addr(sbi, |
662befda CY |
182 | blkno * NAT_ENTRY_PER_BLOCK); |
183 | break; | |
184 | case META_SIT: | |
185 | /* get sit block addr */ | |
7a9d7548 | 186 | fio.new_blkaddr = current_sit_addr(sbi, |
662befda | 187 | blkno * SIT_ENTRY_PER_BLOCK); |
662befda | 188 | break; |
81c1a0f1 | 189 | case META_SSA: |
662befda | 190 | case META_CP: |
4c521f49 | 191 | case META_POR: |
7a9d7548 | 192 | fio.new_blkaddr = blkno; |
662befda CY |
193 | break; |
194 | default: | |
195 | BUG(); | |
196 | } | |
197 | ||
300e129c JK |
198 | page = f2fs_grab_cache_page(META_MAPPING(sbi), |
199 | fio.new_blkaddr, false); | |
662befda CY |
200 | if (!page) |
201 | continue; | |
202 | if (PageUptodate(page)) { | |
662befda CY |
203 | f2fs_put_page(page, 1); |
204 | continue; | |
205 | } | |
206 | ||
05ca3632 | 207 | fio.page = page; |
7a9d7548 | 208 | fio.old_blkaddr = fio.new_blkaddr; |
05ca3632 | 209 | f2fs_submit_page_mbio(&fio); |
662befda CY |
210 | f2fs_put_page(page, 0); |
211 | } | |
212 | out: | |
213 | f2fs_submit_merged_bio(sbi, META, READ); | |
e9f5b8b8 | 214 | blk_finish_plug(&plug); |
662befda CY |
215 | return blkno - start; |
216 | } | |
217 | ||
635aee1f CY |
218 | void ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index) |
219 | { | |
220 | struct page *page; | |
221 | bool readahead = false; | |
222 | ||
223 | page = find_get_page(META_MAPPING(sbi), index); | |
4da7bf5a | 224 | if (!page || !PageUptodate(page)) |
635aee1f CY |
225 | readahead = true; |
226 | f2fs_put_page(page, 0); | |
227 | ||
228 | if (readahead) | |
26879fb1 | 229 | ra_meta_pages(sbi, index, MAX_BIO_BLOCKS(sbi), META_POR, true); |
635aee1f CY |
230 | } |
231 | ||
127e670a JK |
232 | static int f2fs_write_meta_page(struct page *page, |
233 | struct writeback_control *wbc) | |
234 | { | |
4081363f | 235 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 236 | |
ecda0de3 CY |
237 | trace_f2fs_writepage(page, META); |
238 | ||
caf0047e | 239 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
cfb271d4 | 240 | goto redirty_out; |
857dc4e0 | 241 | if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0)) |
cfb271d4 | 242 | goto redirty_out; |
1e968fdf | 243 | if (unlikely(f2fs_cp_error(sbi))) |
cf779cab | 244 | goto redirty_out; |
127e670a | 245 | |
577e3495 JK |
246 | write_meta_page(sbi, page); |
247 | dec_page_count(sbi, F2FS_DIRTY_META); | |
0c3a5797 CY |
248 | |
249 | if (wbc->for_reclaim) | |
250 | f2fs_submit_merged_bio_cond(sbi, NULL, page, 0, META, WRITE); | |
251 | ||
577e3495 | 252 | unlock_page(page); |
857dc4e0 | 253 | |
0c3a5797 | 254 | if (unlikely(f2fs_cp_error(sbi))) |
857dc4e0 | 255 | f2fs_submit_merged_bio(sbi, META, WRITE); |
0c3a5797 | 256 | |
577e3495 | 257 | return 0; |
cfb271d4 CY |
258 | |
259 | redirty_out: | |
76f60268 | 260 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 261 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
262 | } |
263 | ||
264 | static int f2fs_write_meta_pages(struct address_space *mapping, | |
265 | struct writeback_control *wbc) | |
266 | { | |
4081363f | 267 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 268 | long diff, written; |
127e670a | 269 | |
5459aa97 | 270 | /* collect a number of dirty meta pages and write together */ |
50c8cdb3 JK |
271 | if (wbc->for_kupdate || |
272 | get_pages(sbi, F2FS_DIRTY_META) < nr_pages_to_skip(sbi, META)) | |
d3baf95d | 273 | goto skip_write; |
127e670a | 274 | |
d31c7c3f YH |
275 | trace_f2fs_writepages(mapping->host, wbc, META); |
276 | ||
127e670a JK |
277 | /* if mounting is failed, skip writing node pages */ |
278 | mutex_lock(&sbi->cp_mutex); | |
50c8cdb3 JK |
279 | diff = nr_pages_to_write(sbi, META, wbc); |
280 | written = sync_meta_pages(sbi, META, wbc->nr_to_write); | |
127e670a | 281 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 282 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 283 | return 0; |
d3baf95d JK |
284 | |
285 | skip_write: | |
286 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
d31c7c3f | 287 | trace_f2fs_writepages(mapping->host, wbc, META); |
d3baf95d | 288 | return 0; |
127e670a JK |
289 | } |
290 | ||
291 | long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, | |
292 | long nr_to_write) | |
293 | { | |
9df27d98 | 294 | struct address_space *mapping = META_MAPPING(sbi); |
80dd9c0e | 295 | pgoff_t index = 0, end = ULONG_MAX, prev = ULONG_MAX; |
127e670a JK |
296 | struct pagevec pvec; |
297 | long nwritten = 0; | |
298 | struct writeback_control wbc = { | |
299 | .for_reclaim = 0, | |
300 | }; | |
e9f5b8b8 | 301 | struct blk_plug plug; |
127e670a JK |
302 | |
303 | pagevec_init(&pvec, 0); | |
304 | ||
e9f5b8b8 CY |
305 | blk_start_plug(&plug); |
306 | ||
127e670a JK |
307 | while (index <= end) { |
308 | int i, nr_pages; | |
309 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, | |
310 | PAGECACHE_TAG_DIRTY, | |
311 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
cfb271d4 | 312 | if (unlikely(nr_pages == 0)) |
127e670a JK |
313 | break; |
314 | ||
315 | for (i = 0; i < nr_pages; i++) { | |
316 | struct page *page = pvec.pages[i]; | |
203681f6 | 317 | |
80dd9c0e | 318 | if (prev == ULONG_MAX) |
6066d8cd JK |
319 | prev = page->index - 1; |
320 | if (nr_to_write != LONG_MAX && page->index != prev + 1) { | |
321 | pagevec_release(&pvec); | |
322 | goto stop; | |
323 | } | |
324 | ||
127e670a | 325 | lock_page(page); |
203681f6 JK |
326 | |
327 | if (unlikely(page->mapping != mapping)) { | |
328 | continue_unlock: | |
329 | unlock_page(page); | |
330 | continue; | |
331 | } | |
332 | if (!PageDirty(page)) { | |
333 | /* someone wrote it for us */ | |
334 | goto continue_unlock; | |
335 | } | |
336 | ||
fa3d2bdf JK |
337 | f2fs_wait_on_page_writeback(page, META, true); |
338 | ||
339 | BUG_ON(PageWriteback(page)); | |
203681f6 JK |
340 | if (!clear_page_dirty_for_io(page)) |
341 | goto continue_unlock; | |
342 | ||
97dc3fd2 | 343 | if (mapping->a_ops->writepage(page, &wbc)) { |
577e3495 JK |
344 | unlock_page(page); |
345 | break; | |
346 | } | |
cfb271d4 | 347 | nwritten++; |
6066d8cd | 348 | prev = page->index; |
cfb271d4 | 349 | if (unlikely(nwritten >= nr_to_write)) |
127e670a JK |
350 | break; |
351 | } | |
352 | pagevec_release(&pvec); | |
353 | cond_resched(); | |
354 | } | |
6066d8cd | 355 | stop: |
127e670a | 356 | if (nwritten) |
458e6197 | 357 | f2fs_submit_merged_bio(sbi, type, WRITE); |
127e670a | 358 | |
e9f5b8b8 CY |
359 | blk_finish_plug(&plug); |
360 | ||
127e670a JK |
361 | return nwritten; |
362 | } | |
363 | ||
364 | static int f2fs_set_meta_page_dirty(struct page *page) | |
365 | { | |
26c6b887 JK |
366 | trace_f2fs_set_page_dirty(page, META); |
367 | ||
237c0790 JK |
368 | if (!PageUptodate(page)) |
369 | SetPageUptodate(page); | |
127e670a | 370 | if (!PageDirty(page)) { |
fe76b796 | 371 | f2fs_set_page_dirty_nobuffers(page); |
4081363f | 372 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); |
1601839e | 373 | SetPagePrivate(page); |
9e4ded3f | 374 | f2fs_trace_pid(page); |
127e670a JK |
375 | return 1; |
376 | } | |
377 | return 0; | |
378 | } | |
379 | ||
380 | const struct address_space_operations f2fs_meta_aops = { | |
381 | .writepage = f2fs_write_meta_page, | |
382 | .writepages = f2fs_write_meta_pages, | |
383 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
487261f3 CY |
384 | .invalidatepage = f2fs_invalidate_page, |
385 | .releasepage = f2fs_release_page, | |
127e670a JK |
386 | }; |
387 | ||
6451e041 | 388 | static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 389 | { |
67298804 | 390 | struct inode_management *im = &sbi->im[type]; |
80c54505 JK |
391 | struct ino_entry *e, *tmp; |
392 | ||
393 | tmp = f2fs_kmem_cache_alloc(ino_entry_slab, GFP_NOFS); | |
39efac41 | 394 | retry: |
80c54505 | 395 | radix_tree_preload(GFP_NOFS | __GFP_NOFAIL); |
769ec6e5 | 396 | |
67298804 | 397 | spin_lock(&im->ino_lock); |
67298804 | 398 | e = radix_tree_lookup(&im->ino_root, ino); |
39efac41 | 399 | if (!e) { |
80c54505 | 400 | e = tmp; |
67298804 CY |
401 | if (radix_tree_insert(&im->ino_root, ino, e)) { |
402 | spin_unlock(&im->ino_lock); | |
769ec6e5 | 403 | radix_tree_preload_end(); |
39efac41 JK |
404 | goto retry; |
405 | } | |
406 | memset(e, 0, sizeof(struct ino_entry)); | |
407 | e->ino = ino; | |
953e6cc6 | 408 | |
67298804 | 409 | list_add_tail(&e->list, &im->ino_list); |
8c402946 | 410 | if (type != ORPHAN_INO) |
67298804 | 411 | im->ino_num++; |
39efac41 | 412 | } |
67298804 | 413 | spin_unlock(&im->ino_lock); |
769ec6e5 | 414 | radix_tree_preload_end(); |
80c54505 JK |
415 | |
416 | if (e != tmp) | |
417 | kmem_cache_free(ino_entry_slab, tmp); | |
953e6cc6 JK |
418 | } |
419 | ||
6451e041 | 420 | static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 421 | { |
67298804 | 422 | struct inode_management *im = &sbi->im[type]; |
6451e041 | 423 | struct ino_entry *e; |
953e6cc6 | 424 | |
67298804 CY |
425 | spin_lock(&im->ino_lock); |
426 | e = radix_tree_lookup(&im->ino_root, ino); | |
39efac41 JK |
427 | if (e) { |
428 | list_del(&e->list); | |
67298804 CY |
429 | radix_tree_delete(&im->ino_root, ino); |
430 | im->ino_num--; | |
431 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
432 | kmem_cache_free(ino_entry_slab, e); |
433 | return; | |
953e6cc6 | 434 | } |
67298804 | 435 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
436 | } |
437 | ||
a49324f1 | 438 | void add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
439 | { |
440 | /* add new dirty ino entry into list */ | |
441 | __add_ino_entry(sbi, ino, type); | |
442 | } | |
443 | ||
a49324f1 | 444 | void remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
445 | { |
446 | /* remove dirty ino entry from list */ | |
447 | __remove_ino_entry(sbi, ino, type); | |
448 | } | |
449 | ||
450 | /* mode should be APPEND_INO or UPDATE_INO */ | |
451 | bool exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode) | |
452 | { | |
67298804 | 453 | struct inode_management *im = &sbi->im[mode]; |
fff04f90 | 454 | struct ino_entry *e; |
67298804 CY |
455 | |
456 | spin_lock(&im->ino_lock); | |
457 | e = radix_tree_lookup(&im->ino_root, ino); | |
458 | spin_unlock(&im->ino_lock); | |
fff04f90 JK |
459 | return e ? true : false; |
460 | } | |
461 | ||
74ef9241 | 462 | void release_ino_entry(struct f2fs_sb_info *sbi, bool all) |
fff04f90 JK |
463 | { |
464 | struct ino_entry *e, *tmp; | |
465 | int i; | |
466 | ||
74ef9241 | 467 | for (i = all ? ORPHAN_INO: APPEND_INO; i <= UPDATE_INO; i++) { |
67298804 CY |
468 | struct inode_management *im = &sbi->im[i]; |
469 | ||
470 | spin_lock(&im->ino_lock); | |
471 | list_for_each_entry_safe(e, tmp, &im->ino_list, list) { | |
fff04f90 | 472 | list_del(&e->list); |
67298804 | 473 | radix_tree_delete(&im->ino_root, e->ino); |
fff04f90 | 474 | kmem_cache_free(ino_entry_slab, e); |
67298804 | 475 | im->ino_num--; |
fff04f90 | 476 | } |
67298804 | 477 | spin_unlock(&im->ino_lock); |
fff04f90 JK |
478 | } |
479 | } | |
480 | ||
cbd56e7d | 481 | int acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 482 | { |
67298804 | 483 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a JK |
484 | int err = 0; |
485 | ||
67298804 | 486 | spin_lock(&im->ino_lock); |
cb78942b JK |
487 | |
488 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
489 | if (time_to_inject(FAULT_ORPHAN)) { | |
490 | spin_unlock(&im->ino_lock); | |
491 | return -ENOSPC; | |
492 | } | |
493 | #endif | |
67298804 | 494 | if (unlikely(im->ino_num >= sbi->max_orphans)) |
127e670a | 495 | err = -ENOSPC; |
cbd56e7d | 496 | else |
67298804 CY |
497 | im->ino_num++; |
498 | spin_unlock(&im->ino_lock); | |
0d47c1ad | 499 | |
127e670a JK |
500 | return err; |
501 | } | |
502 | ||
cbd56e7d JK |
503 | void release_orphan_inode(struct f2fs_sb_info *sbi) |
504 | { | |
67298804 CY |
505 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
506 | ||
507 | spin_lock(&im->ino_lock); | |
508 | f2fs_bug_on(sbi, im->ino_num == 0); | |
509 | im->ino_num--; | |
510 | spin_unlock(&im->ino_lock); | |
cbd56e7d JK |
511 | } |
512 | ||
67c3758d | 513 | void add_orphan_inode(struct inode *inode) |
127e670a | 514 | { |
39efac41 | 515 | /* add new orphan ino entry into list */ |
67c3758d JK |
516 | __add_ino_entry(F2FS_I_SB(inode), inode->i_ino, ORPHAN_INO); |
517 | update_inode_page(inode); | |
127e670a JK |
518 | } |
519 | ||
520 | void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
521 | { | |
953e6cc6 | 522 | /* remove orphan entry from orphan list */ |
6451e041 | 523 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
524 | } |
525 | ||
8c14bfad | 526 | static int recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 527 | { |
8c14bfad CY |
528 | struct inode *inode; |
529 | ||
530 | inode = f2fs_iget(sbi->sb, ino); | |
531 | if (IS_ERR(inode)) { | |
532 | /* | |
533 | * there should be a bug that we can't find the entry | |
534 | * to orphan inode. | |
535 | */ | |
536 | f2fs_bug_on(sbi, PTR_ERR(inode) == -ENOENT); | |
537 | return PTR_ERR(inode); | |
538 | } | |
539 | ||
127e670a JK |
540 | clear_nlink(inode); |
541 | ||
542 | /* truncate all the data during iput */ | |
543 | iput(inode); | |
8c14bfad | 544 | return 0; |
127e670a JK |
545 | } |
546 | ||
8c14bfad | 547 | int recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a | 548 | { |
3c642985 | 549 | block_t start_blk, orphan_blocks, i, j; |
8c14bfad | 550 | int err; |
127e670a | 551 | |
25ca923b | 552 | if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG)) |
8c14bfad | 553 | return 0; |
127e670a | 554 | |
55141486 | 555 | start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi); |
3c642985 | 556 | orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi); |
127e670a | 557 | |
26879fb1 | 558 | ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP, true); |
662befda | 559 | |
3c642985 | 560 | for (i = 0; i < orphan_blocks; i++) { |
127e670a JK |
561 | struct page *page = get_meta_page(sbi, start_blk + i); |
562 | struct f2fs_orphan_block *orphan_blk; | |
563 | ||
564 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); | |
565 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
566 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
8c14bfad CY |
567 | err = recover_orphan_inode(sbi, ino); |
568 | if (err) { | |
569 | f2fs_put_page(page, 1); | |
8c14bfad CY |
570 | return err; |
571 | } | |
127e670a JK |
572 | } |
573 | f2fs_put_page(page, 1); | |
574 | } | |
575 | /* clear Orphan Flag */ | |
25ca923b | 576 | clear_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG); |
8c14bfad | 577 | return 0; |
127e670a JK |
578 | } |
579 | ||
580 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
581 | { | |
502c6e0b | 582 | struct list_head *head; |
127e670a | 583 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 584 | unsigned int nentries = 0; |
bd936f84 | 585 | unsigned short index = 1; |
8c402946 | 586 | unsigned short orphan_blocks; |
4531929e | 587 | struct page *page = NULL; |
6451e041 | 588 | struct ino_entry *orphan = NULL; |
67298804 | 589 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 590 | |
67298804 | 591 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 592 | |
d6c67a4f JK |
593 | /* |
594 | * we don't need to do spin_lock(&im->ino_lock) here, since all the | |
595 | * orphan inode operations are covered under f2fs_lock_op(). | |
596 | * And, spin_lock should be avoided due to page operations below. | |
597 | */ | |
67298804 | 598 | head = &im->ino_list; |
127e670a JK |
599 | |
600 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
601 | list_for_each_entry(orphan, head, list) { |
602 | if (!page) { | |
bd936f84 | 603 | page = grab_meta_page(sbi, start_blk++); |
502c6e0b GZ |
604 | orphan_blk = |
605 | (struct f2fs_orphan_block *)page_address(page); | |
606 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
607 | } | |
127e670a | 608 | |
36795567 | 609 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 610 | |
36795567 | 611 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
612 | /* |
613 | * an orphan block is full of 1020 entries, | |
614 | * then we need to flush current orphan blocks | |
615 | * and bring another one in memory | |
616 | */ | |
617 | orphan_blk->blk_addr = cpu_to_le16(index); | |
618 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
619 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
620 | set_page_dirty(page); | |
621 | f2fs_put_page(page, 1); | |
622 | index++; | |
127e670a JK |
623 | nentries = 0; |
624 | page = NULL; | |
625 | } | |
502c6e0b | 626 | } |
127e670a | 627 | |
502c6e0b GZ |
628 | if (page) { |
629 | orphan_blk->blk_addr = cpu_to_le16(index); | |
630 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
631 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
632 | set_page_dirty(page); | |
633 | f2fs_put_page(page, 1); | |
127e670a | 634 | } |
127e670a JK |
635 | } |
636 | ||
637 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, | |
638 | block_t cp_addr, unsigned long long *version) | |
639 | { | |
640 | struct page *cp_page_1, *cp_page_2 = NULL; | |
641 | unsigned long blk_size = sbi->blocksize; | |
642 | struct f2fs_checkpoint *cp_block; | |
643 | unsigned long long cur_version = 0, pre_version = 0; | |
127e670a | 644 | size_t crc_offset; |
7e586fa0 | 645 | __u32 crc = 0; |
127e670a JK |
646 | |
647 | /* Read the 1st cp block in this CP pack */ | |
648 | cp_page_1 = get_meta_page(sbi, cp_addr); | |
649 | ||
650 | /* get the version number */ | |
651 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_1); | |
652 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
653 | if (crc_offset >= blk_size) | |
654 | goto invalid_cp1; | |
655 | ||
29e7043f | 656 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
43b6573b | 657 | if (!f2fs_crc_valid(sbi, crc, cp_block, crc_offset)) |
127e670a JK |
658 | goto invalid_cp1; |
659 | ||
d71b5564 | 660 | pre_version = cur_cp_version(cp_block); |
127e670a JK |
661 | |
662 | /* Read the 2nd cp block in this CP pack */ | |
25ca923b | 663 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
127e670a JK |
664 | cp_page_2 = get_meta_page(sbi, cp_addr); |
665 | ||
666 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_2); | |
667 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
668 | if (crc_offset >= blk_size) | |
669 | goto invalid_cp2; | |
670 | ||
29e7043f | 671 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
43b6573b | 672 | if (!f2fs_crc_valid(sbi, crc, cp_block, crc_offset)) |
127e670a JK |
673 | goto invalid_cp2; |
674 | ||
d71b5564 | 675 | cur_version = cur_cp_version(cp_block); |
127e670a JK |
676 | |
677 | if (cur_version == pre_version) { | |
678 | *version = cur_version; | |
679 | f2fs_put_page(cp_page_2, 1); | |
680 | return cp_page_1; | |
681 | } | |
682 | invalid_cp2: | |
683 | f2fs_put_page(cp_page_2, 1); | |
684 | invalid_cp1: | |
685 | f2fs_put_page(cp_page_1, 1); | |
686 | return NULL; | |
687 | } | |
688 | ||
689 | int get_valid_checkpoint(struct f2fs_sb_info *sbi) | |
690 | { | |
691 | struct f2fs_checkpoint *cp_block; | |
692 | struct f2fs_super_block *fsb = sbi->raw_super; | |
693 | struct page *cp1, *cp2, *cur_page; | |
694 | unsigned long blk_size = sbi->blocksize; | |
695 | unsigned long long cp1_version = 0, cp2_version = 0; | |
696 | unsigned long long cp_start_blk_no; | |
55141486 | 697 | unsigned int cp_blks = 1 + __cp_payload(sbi); |
1dbe4152 CL |
698 | block_t cp_blk_no; |
699 | int i; | |
127e670a | 700 | |
1dbe4152 | 701 | sbi->ckpt = kzalloc(cp_blks * blk_size, GFP_KERNEL); |
127e670a JK |
702 | if (!sbi->ckpt) |
703 | return -ENOMEM; | |
704 | /* | |
705 | * Finding out valid cp block involves read both | |
706 | * sets( cp pack1 and cp pack 2) | |
707 | */ | |
708 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
709 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
710 | ||
711 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
712 | cp_start_blk_no += ((unsigned long long)1) << |
713 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
714 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
715 | ||
716 | if (cp1 && cp2) { | |
717 | if (ver_after(cp2_version, cp1_version)) | |
718 | cur_page = cp2; | |
719 | else | |
720 | cur_page = cp1; | |
721 | } else if (cp1) { | |
722 | cur_page = cp1; | |
723 | } else if (cp2) { | |
724 | cur_page = cp2; | |
725 | } else { | |
726 | goto fail_no_cp; | |
727 | } | |
728 | ||
729 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
730 | memcpy(sbi->ckpt, cp_block, blk_size); | |
731 | ||
984ec63c SL |
732 | /* Sanity checking of checkpoint */ |
733 | if (sanity_check_ckpt(sbi)) | |
734 | goto fail_no_cp; | |
735 | ||
1dbe4152 CL |
736 | if (cp_blks <= 1) |
737 | goto done; | |
738 | ||
739 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
740 | if (cur_page == cp2) | |
741 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
742 | ||
743 | for (i = 1; i < cp_blks; i++) { | |
744 | void *sit_bitmap_ptr; | |
745 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
746 | ||
747 | cur_page = get_meta_page(sbi, cp_blk_no + i); | |
748 | sit_bitmap_ptr = page_address(cur_page); | |
749 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
750 | f2fs_put_page(cur_page, 1); | |
751 | } | |
752 | done: | |
127e670a JK |
753 | f2fs_put_page(cp1, 1); |
754 | f2fs_put_page(cp2, 1); | |
755 | return 0; | |
756 | ||
757 | fail_no_cp: | |
758 | kfree(sbi->ckpt); | |
759 | return -EINVAL; | |
760 | } | |
761 | ||
c227f912 | 762 | static void __add_dirty_inode(struct inode *inode, enum inode_type type) |
127e670a | 763 | { |
4081363f | 764 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 765 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
127e670a | 766 | |
91942321 | 767 | if (is_inode_flag_set(inode, flag)) |
2710fd7e | 768 | return; |
2d7b822a | 769 | |
91942321 JK |
770 | set_inode_flag(inode, flag); |
771 | list_add_tail(&F2FS_I(inode)->dirty_list, &sbi->inode_list[type]); | |
33fbd510 | 772 | stat_inc_dirty_inode(sbi, type); |
5deb8267 JK |
773 | } |
774 | ||
c227f912 | 775 | static void __remove_dirty_inode(struct inode *inode, enum inode_type type) |
6ad7609a | 776 | { |
c227f912 | 777 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
6ad7609a | 778 | |
91942321 | 779 | if (get_dirty_pages(inode) || !is_inode_flag_set(inode, flag)) |
6ad7609a CY |
780 | return; |
781 | ||
91942321 JK |
782 | list_del_init(&F2FS_I(inode)->dirty_list); |
783 | clear_inode_flag(inode, flag); | |
33fbd510 | 784 | stat_dec_dirty_inode(F2FS_I_SB(inode), type); |
6ad7609a CY |
785 | } |
786 | ||
a7ffdbe2 | 787 | void update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 788 | { |
4081363f | 789 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 790 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
5deb8267 | 791 | |
5ac9f36f CY |
792 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
793 | !S_ISLNK(inode->i_mode)) | |
127e670a | 794 | return; |
7bd59381 | 795 | |
1c4bf763 JK |
796 | spin_lock(&sbi->inode_lock[type]); |
797 | if (type != FILE_INODE || test_opt(sbi, DATA_FLUSH)) | |
10aa97c3 | 798 | __add_dirty_inode(inode, type); |
b951a4ec | 799 | inode_inc_dirty_pages(inode); |
1c4bf763 JK |
800 | spin_unlock(&sbi->inode_lock[type]); |
801 | ||
a7ffdbe2 | 802 | SetPagePrivate(page); |
9e4ded3f | 803 | f2fs_trace_pid(page); |
5deb8267 JK |
804 | } |
805 | ||
c227f912 | 806 | void remove_dirty_inode(struct inode *inode) |
127e670a | 807 | { |
4081363f | 808 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 809 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
127e670a | 810 | |
c227f912 CY |
811 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
812 | !S_ISLNK(inode->i_mode)) | |
127e670a JK |
813 | return; |
814 | ||
10aa97c3 JK |
815 | if (type == FILE_INODE && !test_opt(sbi, DATA_FLUSH)) |
816 | return; | |
817 | ||
c227f912 CY |
818 | spin_lock(&sbi->inode_lock[type]); |
819 | __remove_dirty_inode(inode, type); | |
820 | spin_unlock(&sbi->inode_lock[type]); | |
74d0b917 JK |
821 | } |
822 | ||
6d5a1495 | 823 | int sync_dirty_inodes(struct f2fs_sb_info *sbi, enum inode_type type) |
127e670a | 824 | { |
ce3b7d80 | 825 | struct list_head *head; |
127e670a | 826 | struct inode *inode; |
2710fd7e | 827 | struct f2fs_inode_info *fi; |
4cf18537 CY |
828 | bool is_dir = (type == DIR_INODE); |
829 | ||
830 | trace_f2fs_sync_dirty_inodes_enter(sbi->sb, is_dir, | |
831 | get_pages(sbi, is_dir ? | |
832 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
127e670a | 833 | retry: |
af41d3ee | 834 | if (unlikely(f2fs_cp_error(sbi))) |
6d5a1495 | 835 | return -EIO; |
af41d3ee | 836 | |
c227f912 | 837 | spin_lock(&sbi->inode_lock[type]); |
ce3b7d80 | 838 | |
c227f912 | 839 | head = &sbi->inode_list[type]; |
127e670a | 840 | if (list_empty(head)) { |
c227f912 | 841 | spin_unlock(&sbi->inode_lock[type]); |
4cf18537 CY |
842 | trace_f2fs_sync_dirty_inodes_exit(sbi->sb, is_dir, |
843 | get_pages(sbi, is_dir ? | |
844 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
6d5a1495 | 845 | return 0; |
127e670a | 846 | } |
2710fd7e CY |
847 | fi = list_entry(head->next, struct f2fs_inode_info, dirty_list); |
848 | inode = igrab(&fi->vfs_inode); | |
c227f912 | 849 | spin_unlock(&sbi->inode_lock[type]); |
127e670a | 850 | if (inode) { |
87d6f890 | 851 | filemap_fdatawrite(inode->i_mapping); |
127e670a JK |
852 | iput(inode); |
853 | } else { | |
854 | /* | |
855 | * We should submit bio, since it exists several | |
856 | * wribacking dentry pages in the freeing inode. | |
857 | */ | |
458e6197 | 858 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
7ecebe5e | 859 | cond_resched(); |
127e670a JK |
860 | } |
861 | goto retry; | |
862 | } | |
863 | ||
0f18b462 JK |
864 | int f2fs_sync_inode_meta(struct f2fs_sb_info *sbi) |
865 | { | |
866 | struct list_head *head = &sbi->inode_list[DIRTY_META]; | |
867 | struct inode *inode; | |
868 | struct f2fs_inode_info *fi; | |
869 | s64 total = get_pages(sbi, F2FS_DIRTY_IMETA); | |
870 | ||
871 | while (total--) { | |
872 | if (unlikely(f2fs_cp_error(sbi))) | |
873 | return -EIO; | |
874 | ||
875 | spin_lock(&sbi->inode_lock[DIRTY_META]); | |
876 | if (list_empty(head)) { | |
877 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
878 | return 0; | |
879 | } | |
880 | fi = list_entry(head->next, struct f2fs_inode_info, | |
881 | gdirty_list); | |
882 | inode = igrab(&fi->vfs_inode); | |
883 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
884 | if (inode) { | |
885 | update_inode_page(inode); | |
886 | iput(inode); | |
887 | } | |
888 | }; | |
889 | return 0; | |
890 | } | |
891 | ||
0a8165d7 | 892 | /* |
127e670a JK |
893 | * Freeze all the FS-operations for checkpoint. |
894 | */ | |
cf779cab | 895 | static int block_operations(struct f2fs_sb_info *sbi) |
127e670a | 896 | { |
127e670a JK |
897 | struct writeback_control wbc = { |
898 | .sync_mode = WB_SYNC_ALL, | |
899 | .nr_to_write = LONG_MAX, | |
900 | .for_reclaim = 0, | |
901 | }; | |
cf779cab | 902 | int err = 0; |
c718379b | 903 | |
39936837 | 904 | retry_flush_dents: |
e479556b | 905 | f2fs_lock_all(sbi); |
127e670a | 906 | /* write all the dirty dentry pages */ |
127e670a | 907 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 908 | f2fs_unlock_all(sbi); |
6d5a1495 CY |
909 | err = sync_dirty_inodes(sbi, DIR_INODE); |
910 | if (err) | |
cf779cab | 911 | goto out; |
39936837 | 912 | goto retry_flush_dents; |
127e670a JK |
913 | } |
914 | ||
0f18b462 JK |
915 | if (get_pages(sbi, F2FS_DIRTY_IMETA)) { |
916 | f2fs_unlock_all(sbi); | |
917 | err = f2fs_sync_inode_meta(sbi); | |
918 | if (err) | |
919 | goto out; | |
920 | goto retry_flush_dents; | |
921 | } | |
922 | ||
127e670a | 923 | /* |
e1c42045 | 924 | * POR: we should ensure that there are no dirty node pages |
127e670a JK |
925 | * until finishing nat/sit flush. |
926 | */ | |
39936837 | 927 | retry_flush_nodes: |
b3582c68 | 928 | down_write(&sbi->node_write); |
127e670a JK |
929 | |
930 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
b3582c68 | 931 | up_write(&sbi->node_write); |
52681375 | 932 | err = sync_node_pages(sbi, &wbc); |
6d5a1495 | 933 | if (err) { |
cf779cab | 934 | f2fs_unlock_all(sbi); |
cf779cab JK |
935 | goto out; |
936 | } | |
39936837 | 937 | goto retry_flush_nodes; |
127e670a | 938 | } |
cf779cab | 939 | out: |
cf779cab | 940 | return err; |
127e670a JK |
941 | } |
942 | ||
943 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
944 | { | |
b3582c68 | 945 | up_write(&sbi->node_write); |
ad4edb83 JK |
946 | |
947 | build_free_nids(sbi); | |
e479556b | 948 | f2fs_unlock_all(sbi); |
127e670a JK |
949 | } |
950 | ||
fb51b5ef CL |
951 | static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi) |
952 | { | |
953 | DEFINE_WAIT(wait); | |
954 | ||
955 | for (;;) { | |
956 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
957 | ||
f5730184 | 958 | if (!atomic_read(&sbi->nr_wb_bios)) |
fb51b5ef CL |
959 | break; |
960 | ||
0ff21646 | 961 | io_schedule_timeout(5*HZ); |
fb51b5ef CL |
962 | } |
963 | finish_wait(&sbi->cp_wait, &wait); | |
964 | } | |
965 | ||
c34f42e2 | 966 | static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
967 | { |
968 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
cf2271e7 | 969 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); |
77041823 | 970 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
67298804 | 971 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; |
77041823 | 972 | nid_t last_nid = nm_i->next_scan_nid; |
127e670a | 973 | block_t start_blk; |
127e670a | 974 | unsigned int data_sum_blocks, orphan_blocks; |
7e586fa0 | 975 | __u32 crc32 = 0; |
127e670a | 976 | int i; |
55141486 | 977 | int cp_payload_blks = __cp_payload(sbi); |
e90c2d28 CY |
978 | block_t discard_blk = NEXT_FREE_BLKADDR(sbi, curseg); |
979 | bool invalidate = false; | |
8f1dbbbb SL |
980 | struct super_block *sb = sbi->sb; |
981 | struct curseg_info *seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); | |
982 | u64 kbytes_written; | |
127e670a | 983 | |
1e87a78d JK |
984 | /* |
985 | * This avoids to conduct wrong roll-forward operations and uses | |
986 | * metapages, so should be called prior to sync_meta_pages below. | |
987 | */ | |
36abef4e | 988 | if (!test_opt(sbi, LFS) && discard_next_dnode(sbi, discard_blk)) |
e90c2d28 | 989 | invalidate = true; |
127e670a JK |
990 | |
991 | /* Flush all the NAT/SIT pages */ | |
cf779cab | 992 | while (get_pages(sbi, F2FS_DIRTY_META)) { |
127e670a | 993 | sync_meta_pages(sbi, META, LONG_MAX); |
cf779cab | 994 | if (unlikely(f2fs_cp_error(sbi))) |
c34f42e2 | 995 | return -EIO; |
cf779cab | 996 | } |
127e670a JK |
997 | |
998 | next_free_nid(sbi, &last_nid); | |
999 | ||
1000 | /* | |
1001 | * modify checkpoint | |
1002 | * version number is already updated | |
1003 | */ | |
1004 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi)); | |
1005 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
1006 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); | |
b5b82205 | 1007 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
1008 | ckpt->cur_node_segno[i] = |
1009 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
1010 | ckpt->cur_node_blkoff[i] = | |
1011 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
1012 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
1013 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
1014 | } | |
b5b82205 | 1015 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
1016 | ckpt->cur_data_segno[i] = |
1017 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
1018 | ckpt->cur_data_blkoff[i] = | |
1019 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
1020 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
1021 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
1022 | } | |
1023 | ||
1024 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
1025 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
1026 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
1027 | ||
1028 | /* 2 cp + n data seg summary + orphan inode blocks */ | |
3fa06d7b | 1029 | data_sum_blocks = npages_for_summary_flush(sbi, false); |
b5b82205 | 1030 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
25ca923b | 1031 | set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 1032 | else |
25ca923b | 1033 | clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 1034 | |
67298804 | 1035 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
1036 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
1037 | orphan_blocks); | |
127e670a | 1038 | |
119ee914 | 1039 | if (__remain_node_summaries(cpc->reason)) |
b5b82205 | 1040 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
1041 | cp_payload_blks + data_sum_blocks + |
1042 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
119ee914 | 1043 | else |
b5b82205 | 1044 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS + |
1dbe4152 CL |
1045 | cp_payload_blks + data_sum_blocks + |
1046 | orphan_blocks); | |
119ee914 JK |
1047 | |
1048 | if (cpc->reason == CP_UMOUNT) | |
1049 | set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
1050 | else | |
1051 | clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
1052 | ||
1053 | if (cpc->reason == CP_FASTBOOT) | |
1054 | set_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
1055 | else | |
1056 | clear_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
127e670a | 1057 | |
67298804 | 1058 | if (orphan_num) |
25ca923b | 1059 | set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1060 | else |
25ca923b | 1061 | clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1062 | |
caf0047e | 1063 | if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) |
2ae4c673 JK |
1064 | set_ckpt_flags(ckpt, CP_FSCK_FLAG); |
1065 | ||
127e670a JK |
1066 | /* update SIT/NAT bitmap */ |
1067 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
1068 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
1069 | ||
43b6573b | 1070 | crc32 = f2fs_crc32(sbi, ckpt, le32_to_cpu(ckpt->checksum_offset)); |
7e586fa0 JK |
1071 | *((__le32 *)((unsigned char *)ckpt + |
1072 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
1073 | = cpu_to_le32(crc32); |
1074 | ||
1075 | start_blk = __start_cp_addr(sbi); | |
1076 | ||
a7230d16 JK |
1077 | /* need to wait for end_io results */ |
1078 | wait_on_all_pages_writeback(sbi); | |
1079 | if (unlikely(f2fs_cp_error(sbi))) | |
c34f42e2 | 1080 | return -EIO; |
a7230d16 | 1081 | |
127e670a | 1082 | /* write out checkpoint buffer at block 0 */ |
381722d2 CY |
1083 | update_meta_page(sbi, ckpt, start_blk++); |
1084 | ||
1085 | for (i = 1; i < 1 + cp_payload_blks; i++) | |
1086 | update_meta_page(sbi, (char *)ckpt + i * F2FS_BLKSIZE, | |
1087 | start_blk++); | |
1dbe4152 | 1088 | |
67298804 | 1089 | if (orphan_num) { |
127e670a JK |
1090 | write_orphan_inodes(sbi, start_blk); |
1091 | start_blk += orphan_blocks; | |
1092 | } | |
1093 | ||
1094 | write_data_summaries(sbi, start_blk); | |
1095 | start_blk += data_sum_blocks; | |
8f1dbbbb SL |
1096 | |
1097 | /* Record write statistics in the hot node summary */ | |
1098 | kbytes_written = sbi->kbytes_written; | |
1099 | if (sb->s_bdev->bd_part) | |
1100 | kbytes_written += BD_PART_WRITTEN(sbi); | |
1101 | ||
b7ad7512 | 1102 | seg_i->journal->info.kbytes_written = cpu_to_le64(kbytes_written); |
8f1dbbbb | 1103 | |
119ee914 | 1104 | if (__remain_node_summaries(cpc->reason)) { |
127e670a JK |
1105 | write_node_summaries(sbi, start_blk); |
1106 | start_blk += NR_CURSEG_NODE_TYPE; | |
1107 | } | |
1108 | ||
1109 | /* writeout checkpoint block */ | |
381722d2 | 1110 | update_meta_page(sbi, ckpt, start_blk); |
127e670a JK |
1111 | |
1112 | /* wait for previous submitted node/meta pages writeback */ | |
fb51b5ef | 1113 | wait_on_all_pages_writeback(sbi); |
127e670a | 1114 | |
cf779cab | 1115 | if (unlikely(f2fs_cp_error(sbi))) |
c34f42e2 | 1116 | return -EIO; |
cf779cab | 1117 | |
80dd9c0e CY |
1118 | filemap_fdatawait_range(NODE_MAPPING(sbi), 0, LLONG_MAX); |
1119 | filemap_fdatawait_range(META_MAPPING(sbi), 0, LLONG_MAX); | |
127e670a JK |
1120 | |
1121 | /* update user_block_counts */ | |
1122 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
41382ec4 | 1123 | percpu_counter_set(&sbi->alloc_valid_block_count, 0); |
127e670a JK |
1124 | |
1125 | /* Here, we only have one bio having CP pack */ | |
577e3495 | 1126 | sync_meta_pages(sbi, META_FLUSH, LONG_MAX); |
127e670a | 1127 | |
6a8f8ca5 JK |
1128 | /* wait for previous submitted meta pages writeback */ |
1129 | wait_on_all_pages_writeback(sbi); | |
1130 | ||
e90c2d28 CY |
1131 | /* |
1132 | * invalidate meta page which is used temporarily for zeroing out | |
1133 | * block at the end of warm node chain. | |
1134 | */ | |
1135 | if (invalidate) | |
1136 | invalidate_mapping_pages(META_MAPPING(sbi), discard_blk, | |
1137 | discard_blk); | |
1138 | ||
74ef9241 | 1139 | release_ino_entry(sbi, false); |
cf779cab JK |
1140 | |
1141 | if (unlikely(f2fs_cp_error(sbi))) | |
c34f42e2 | 1142 | return -EIO; |
cf779cab | 1143 | |
836b5a63 | 1144 | clear_prefree_segments(sbi, cpc); |
caf0047e | 1145 | clear_sbi_flag(sbi, SBI_IS_DIRTY); |
c34f42e2 CY |
1146 | |
1147 | return 0; | |
127e670a JK |
1148 | } |
1149 | ||
0a8165d7 | 1150 | /* |
e1c42045 | 1151 | * We guarantee that this checkpoint procedure will not fail. |
127e670a | 1152 | */ |
c34f42e2 | 1153 | int write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1154 | { |
1155 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1156 | unsigned long long ckpt_ver; | |
c34f42e2 | 1157 | int err = 0; |
127e670a | 1158 | |
43727527 | 1159 | mutex_lock(&sbi->cp_mutex); |
8501017e | 1160 | |
caf0047e | 1161 | if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) && |
a66cdd98 JK |
1162 | (cpc->reason == CP_FASTBOOT || cpc->reason == CP_SYNC || |
1163 | (cpc->reason == CP_DISCARD && !sbi->discard_blks))) | |
8501017e | 1164 | goto out; |
c34f42e2 CY |
1165 | if (unlikely(f2fs_cp_error(sbi))) { |
1166 | err = -EIO; | |
cf779cab | 1167 | goto out; |
c34f42e2 CY |
1168 | } |
1169 | if (f2fs_readonly(sbi->sb)) { | |
1170 | err = -EROFS; | |
11504a8e | 1171 | goto out; |
c34f42e2 | 1172 | } |
2bda542d WL |
1173 | |
1174 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); | |
1175 | ||
c34f42e2 CY |
1176 | err = block_operations(sbi); |
1177 | if (err) | |
cf779cab | 1178 | goto out; |
127e670a | 1179 | |
75ab4cb8 | 1180 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1181 | |
406657dd | 1182 | f2fs_flush_merged_bios(sbi); |
127e670a JK |
1183 | |
1184 | /* | |
1185 | * update checkpoint pack index | |
1186 | * Increase the version number so that | |
1187 | * SIT entries and seg summaries are written at correct place | |
1188 | */ | |
d71b5564 | 1189 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1190 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1191 | ||
1192 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
1193 | flush_nat_entries(sbi); | |
4b2fecc8 | 1194 | flush_sit_entries(sbi, cpc); |
127e670a | 1195 | |
127e670a | 1196 | /* unlock all the fs_lock[] in do_checkpoint() */ |
c34f42e2 | 1197 | err = do_checkpoint(sbi, cpc); |
127e670a JK |
1198 | |
1199 | unblock_operations(sbi); | |
942e0be6 | 1200 | stat_inc_cp_count(sbi->stat_info); |
10027551 JK |
1201 | |
1202 | if (cpc->reason == CP_RECOVERY) | |
1203 | f2fs_msg(sbi->sb, KERN_NOTICE, | |
1204 | "checkpoint: version = %llx", ckpt_ver); | |
60b99b48 JK |
1205 | |
1206 | /* do checkpoint periodically */ | |
6beceb54 | 1207 | f2fs_update_time(sbi, CP_TIME); |
55d1cdb2 | 1208 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
8501017e JK |
1209 | out: |
1210 | mutex_unlock(&sbi->cp_mutex); | |
c34f42e2 | 1211 | return err; |
127e670a JK |
1212 | } |
1213 | ||
6451e041 | 1214 | void init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1215 | { |
6451e041 JK |
1216 | int i; |
1217 | ||
1218 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1219 | struct inode_management *im = &sbi->im[i]; |
1220 | ||
1221 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1222 | spin_lock_init(&im->ino_lock); | |
1223 | INIT_LIST_HEAD(&im->ino_list); | |
1224 | im->ino_num = 0; | |
6451e041 JK |
1225 | } |
1226 | ||
b5b82205 | 1227 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
14b42817 WL |
1228 | NR_CURSEG_TYPE - __cp_payload(sbi)) * |
1229 | F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
1230 | } |
1231 | ||
6e6093a8 | 1232 | int __init create_checkpoint_caches(void) |
127e670a | 1233 | { |
6451e041 JK |
1234 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1235 | sizeof(struct ino_entry)); | |
1236 | if (!ino_entry_slab) | |
127e670a | 1237 | return -ENOMEM; |
06292073 CY |
1238 | inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry", |
1239 | sizeof(struct inode_entry)); | |
6bacf52f | 1240 | if (!inode_entry_slab) { |
6451e041 | 1241 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1242 | return -ENOMEM; |
1243 | } | |
1244 | return 0; | |
1245 | } | |
1246 | ||
1247 | void destroy_checkpoint_caches(void) | |
1248 | { | |
6451e041 | 1249 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1250 | kmem_cache_destroy(inode_entry_slab); |
1251 | } |