]>
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); |
127e670a JK |
51 | SetPageUptodate(page); |
52 | return page; | |
53 | } | |
54 | ||
0a8165d7 | 55 | /* |
127e670a JK |
56 | * We guarantee no failure on the returned page. |
57 | */ | |
2b947003 CY |
58 | static struct page *__get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index, |
59 | bool is_meta) | |
127e670a | 60 | { |
9df27d98 | 61 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a | 62 | struct page *page; |
cf04e8eb | 63 | struct f2fs_io_info fio = { |
05ca3632 | 64 | .sbi = sbi, |
cf04e8eb JK |
65 | .type = META, |
66 | .rw = READ_SYNC | REQ_META | REQ_PRIO, | |
7a9d7548 CY |
67 | .old_blkaddr = index, |
68 | .new_blkaddr = index, | |
4375a336 | 69 | .encrypted_page = NULL, |
cf04e8eb | 70 | }; |
2b947003 CY |
71 | |
72 | if (unlikely(!is_meta)) | |
73 | fio.rw &= ~REQ_META; | |
127e670a | 74 | repeat: |
300e129c | 75 | page = f2fs_grab_cache_page(mapping, index, false); |
127e670a JK |
76 | if (!page) { |
77 | cond_resched(); | |
78 | goto repeat; | |
79 | } | |
393ff91f JK |
80 | if (PageUptodate(page)) |
81 | goto out; | |
82 | ||
05ca3632 JK |
83 | fio.page = page; |
84 | ||
86531d6b JK |
85 | if (f2fs_submit_page_bio(&fio)) { |
86 | f2fs_put_page(page, 1); | |
127e670a | 87 | goto repeat; |
86531d6b | 88 | } |
127e670a | 89 | |
393ff91f | 90 | lock_page(page); |
6bacf52f | 91 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
92 | f2fs_put_page(page, 1); |
93 | goto repeat; | |
94 | } | |
f3f338ca CY |
95 | |
96 | /* | |
97 | * if there is any IO error when accessing device, make our filesystem | |
98 | * readonly and make sure do not write checkpoint with non-uptodate | |
99 | * meta page. | |
100 | */ | |
101 | if (unlikely(!PageUptodate(page))) | |
38f91ca8 | 102 | f2fs_stop_checkpoint(sbi, false); |
393ff91f | 103 | out: |
127e670a JK |
104 | return page; |
105 | } | |
106 | ||
2b947003 CY |
107 | struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) |
108 | { | |
109 | return __get_meta_page(sbi, index, true); | |
110 | } | |
111 | ||
112 | /* for POR only */ | |
113 | struct page *get_tmp_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
114 | { | |
115 | return __get_meta_page(sbi, index, false); | |
116 | } | |
117 | ||
f0c9cada | 118 | bool is_valid_blkaddr(struct f2fs_sb_info *sbi, block_t blkaddr, int type) |
662befda CY |
119 | { |
120 | switch (type) { | |
121 | case META_NAT: | |
66b00c18 | 122 | break; |
662befda | 123 | case META_SIT: |
66b00c18 CY |
124 | if (unlikely(blkaddr >= SIT_BLK_CNT(sbi))) |
125 | return false; | |
126 | break; | |
81c1a0f1 | 127 | case META_SSA: |
66b00c18 CY |
128 | if (unlikely(blkaddr >= MAIN_BLKADDR(sbi) || |
129 | blkaddr < SM_I(sbi)->ssa_blkaddr)) | |
130 | return false; | |
131 | break; | |
662befda | 132 | case META_CP: |
66b00c18 CY |
133 | if (unlikely(blkaddr >= SIT_I(sbi)->sit_base_addr || |
134 | blkaddr < __start_cp_addr(sbi))) | |
135 | return false; | |
136 | break; | |
4c521f49 | 137 | case META_POR: |
66b00c18 CY |
138 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
139 | blkaddr < MAIN_BLKADDR(sbi))) | |
140 | return false; | |
141 | break; | |
662befda CY |
142 | default: |
143 | BUG(); | |
144 | } | |
66b00c18 CY |
145 | |
146 | return true; | |
662befda CY |
147 | } |
148 | ||
149 | /* | |
81c1a0f1 | 150 | * Readahead CP/NAT/SIT/SSA pages |
662befda | 151 | */ |
26879fb1 CY |
152 | int ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages, |
153 | int type, bool sync) | |
662befda | 154 | { |
662befda | 155 | struct page *page; |
4c521f49 | 156 | block_t blkno = start; |
662befda | 157 | struct f2fs_io_info fio = { |
05ca3632 | 158 | .sbi = sbi, |
662befda | 159 | .type = META, |
26879fb1 | 160 | .rw = sync ? (READ_SYNC | REQ_META | REQ_PRIO) : READA, |
4375a336 | 161 | .encrypted_page = NULL, |
662befda | 162 | }; |
e9f5b8b8 | 163 | struct blk_plug plug; |
662befda | 164 | |
2b947003 CY |
165 | if (unlikely(type == META_POR)) |
166 | fio.rw &= ~REQ_META; | |
167 | ||
e9f5b8b8 | 168 | blk_start_plug(&plug); |
662befda | 169 | for (; nrpages-- > 0; blkno++) { |
662befda | 170 | |
66b00c18 CY |
171 | if (!is_valid_blkaddr(sbi, blkno, type)) |
172 | goto out; | |
173 | ||
662befda CY |
174 | switch (type) { |
175 | case META_NAT: | |
66b00c18 CY |
176 | if (unlikely(blkno >= |
177 | NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))) | |
662befda | 178 | blkno = 0; |
66b00c18 | 179 | /* get nat block addr */ |
7a9d7548 | 180 | fio.new_blkaddr = current_nat_addr(sbi, |
662befda CY |
181 | blkno * NAT_ENTRY_PER_BLOCK); |
182 | break; | |
183 | case META_SIT: | |
184 | /* get sit block addr */ | |
7a9d7548 | 185 | fio.new_blkaddr = current_sit_addr(sbi, |
662befda | 186 | blkno * SIT_ENTRY_PER_BLOCK); |
662befda | 187 | break; |
81c1a0f1 | 188 | case META_SSA: |
662befda | 189 | case META_CP: |
4c521f49 | 190 | case META_POR: |
7a9d7548 | 191 | fio.new_blkaddr = blkno; |
662befda CY |
192 | break; |
193 | default: | |
194 | BUG(); | |
195 | } | |
196 | ||
300e129c JK |
197 | page = f2fs_grab_cache_page(META_MAPPING(sbi), |
198 | fio.new_blkaddr, false); | |
662befda CY |
199 | if (!page) |
200 | continue; | |
201 | if (PageUptodate(page)) { | |
662befda CY |
202 | f2fs_put_page(page, 1); |
203 | continue; | |
204 | } | |
205 | ||
05ca3632 | 206 | fio.page = page; |
7a9d7548 | 207 | fio.old_blkaddr = fio.new_blkaddr; |
05ca3632 | 208 | f2fs_submit_page_mbio(&fio); |
662befda CY |
209 | f2fs_put_page(page, 0); |
210 | } | |
211 | out: | |
212 | f2fs_submit_merged_bio(sbi, META, READ); | |
e9f5b8b8 | 213 | blk_finish_plug(&plug); |
662befda CY |
214 | return blkno - start; |
215 | } | |
216 | ||
635aee1f CY |
217 | void ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index) |
218 | { | |
219 | struct page *page; | |
220 | bool readahead = false; | |
221 | ||
222 | page = find_get_page(META_MAPPING(sbi), index); | |
4da7bf5a | 223 | if (!page || !PageUptodate(page)) |
635aee1f CY |
224 | readahead = true; |
225 | f2fs_put_page(page, 0); | |
226 | ||
227 | if (readahead) | |
26879fb1 | 228 | ra_meta_pages(sbi, index, MAX_BIO_BLOCKS(sbi), META_POR, true); |
635aee1f CY |
229 | } |
230 | ||
127e670a JK |
231 | static int f2fs_write_meta_page(struct page *page, |
232 | struct writeback_control *wbc) | |
233 | { | |
4081363f | 234 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 235 | |
ecda0de3 CY |
236 | trace_f2fs_writepage(page, META); |
237 | ||
caf0047e | 238 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
cfb271d4 | 239 | goto redirty_out; |
857dc4e0 | 240 | if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0)) |
cfb271d4 | 241 | goto redirty_out; |
1e968fdf | 242 | if (unlikely(f2fs_cp_error(sbi))) |
cf779cab | 243 | goto redirty_out; |
127e670a | 244 | |
577e3495 JK |
245 | write_meta_page(sbi, page); |
246 | dec_page_count(sbi, F2FS_DIRTY_META); | |
0c3a5797 CY |
247 | |
248 | if (wbc->for_reclaim) | |
249 | f2fs_submit_merged_bio_cond(sbi, NULL, page, 0, META, WRITE); | |
250 | ||
577e3495 | 251 | unlock_page(page); |
857dc4e0 | 252 | |
0c3a5797 | 253 | if (unlikely(f2fs_cp_error(sbi))) |
857dc4e0 | 254 | f2fs_submit_merged_bio(sbi, META, WRITE); |
0c3a5797 | 255 | |
577e3495 | 256 | return 0; |
cfb271d4 CY |
257 | |
258 | redirty_out: | |
76f60268 | 259 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 260 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
261 | } |
262 | ||
263 | static int f2fs_write_meta_pages(struct address_space *mapping, | |
264 | struct writeback_control *wbc) | |
265 | { | |
4081363f | 266 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 267 | long diff, written; |
127e670a | 268 | |
5459aa97 | 269 | /* collect a number of dirty meta pages and write together */ |
50c8cdb3 JK |
270 | if (wbc->for_kupdate || |
271 | get_pages(sbi, F2FS_DIRTY_META) < nr_pages_to_skip(sbi, META)) | |
d3baf95d | 272 | goto skip_write; |
127e670a | 273 | |
d31c7c3f YH |
274 | trace_f2fs_writepages(mapping->host, wbc, META); |
275 | ||
127e670a JK |
276 | /* if mounting is failed, skip writing node pages */ |
277 | mutex_lock(&sbi->cp_mutex); | |
50c8cdb3 JK |
278 | diff = nr_pages_to_write(sbi, META, wbc); |
279 | written = sync_meta_pages(sbi, META, wbc->nr_to_write); | |
127e670a | 280 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 281 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 282 | return 0; |
d3baf95d JK |
283 | |
284 | skip_write: | |
285 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
d31c7c3f | 286 | trace_f2fs_writepages(mapping->host, wbc, META); |
d3baf95d | 287 | return 0; |
127e670a JK |
288 | } |
289 | ||
290 | long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, | |
291 | long nr_to_write) | |
292 | { | |
9df27d98 | 293 | struct address_space *mapping = META_MAPPING(sbi); |
80dd9c0e | 294 | pgoff_t index = 0, end = ULONG_MAX, prev = ULONG_MAX; |
127e670a JK |
295 | struct pagevec pvec; |
296 | long nwritten = 0; | |
297 | struct writeback_control wbc = { | |
298 | .for_reclaim = 0, | |
299 | }; | |
e9f5b8b8 | 300 | struct blk_plug plug; |
127e670a JK |
301 | |
302 | pagevec_init(&pvec, 0); | |
303 | ||
e9f5b8b8 CY |
304 | blk_start_plug(&plug); |
305 | ||
127e670a JK |
306 | while (index <= end) { |
307 | int i, nr_pages; | |
308 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, | |
309 | PAGECACHE_TAG_DIRTY, | |
310 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
cfb271d4 | 311 | if (unlikely(nr_pages == 0)) |
127e670a JK |
312 | break; |
313 | ||
314 | for (i = 0; i < nr_pages; i++) { | |
315 | struct page *page = pvec.pages[i]; | |
203681f6 | 316 | |
80dd9c0e | 317 | if (prev == ULONG_MAX) |
6066d8cd JK |
318 | prev = page->index - 1; |
319 | if (nr_to_write != LONG_MAX && page->index != prev + 1) { | |
320 | pagevec_release(&pvec); | |
321 | goto stop; | |
322 | } | |
323 | ||
127e670a | 324 | lock_page(page); |
203681f6 JK |
325 | |
326 | if (unlikely(page->mapping != mapping)) { | |
327 | continue_unlock: | |
328 | unlock_page(page); | |
329 | continue; | |
330 | } | |
331 | if (!PageDirty(page)) { | |
332 | /* someone wrote it for us */ | |
333 | goto continue_unlock; | |
334 | } | |
335 | ||
fa3d2bdf JK |
336 | f2fs_wait_on_page_writeback(page, META, true); |
337 | ||
338 | BUG_ON(PageWriteback(page)); | |
203681f6 JK |
339 | if (!clear_page_dirty_for_io(page)) |
340 | goto continue_unlock; | |
341 | ||
97dc3fd2 | 342 | if (mapping->a_ops->writepage(page, &wbc)) { |
577e3495 JK |
343 | unlock_page(page); |
344 | break; | |
345 | } | |
cfb271d4 | 346 | nwritten++; |
6066d8cd | 347 | prev = page->index; |
cfb271d4 | 348 | if (unlikely(nwritten >= nr_to_write)) |
127e670a JK |
349 | break; |
350 | } | |
351 | pagevec_release(&pvec); | |
352 | cond_resched(); | |
353 | } | |
6066d8cd | 354 | stop: |
127e670a | 355 | if (nwritten) |
458e6197 | 356 | f2fs_submit_merged_bio(sbi, type, WRITE); |
127e670a | 357 | |
e9f5b8b8 CY |
358 | blk_finish_plug(&plug); |
359 | ||
127e670a JK |
360 | return nwritten; |
361 | } | |
362 | ||
363 | static int f2fs_set_meta_page_dirty(struct page *page) | |
364 | { | |
26c6b887 JK |
365 | trace_f2fs_set_page_dirty(page, META); |
366 | ||
127e670a JK |
367 | SetPageUptodate(page); |
368 | if (!PageDirty(page)) { | |
369 | __set_page_dirty_nobuffers(page); | |
4081363f | 370 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); |
1601839e | 371 | SetPagePrivate(page); |
9e4ded3f | 372 | f2fs_trace_pid(page); |
127e670a JK |
373 | return 1; |
374 | } | |
375 | return 0; | |
376 | } | |
377 | ||
378 | const struct address_space_operations f2fs_meta_aops = { | |
379 | .writepage = f2fs_write_meta_page, | |
380 | .writepages = f2fs_write_meta_pages, | |
381 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
487261f3 CY |
382 | .invalidatepage = f2fs_invalidate_page, |
383 | .releasepage = f2fs_release_page, | |
127e670a JK |
384 | }; |
385 | ||
6451e041 | 386 | static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 387 | { |
67298804 | 388 | struct inode_management *im = &sbi->im[type]; |
80c54505 JK |
389 | struct ino_entry *e, *tmp; |
390 | ||
391 | tmp = f2fs_kmem_cache_alloc(ino_entry_slab, GFP_NOFS); | |
39efac41 | 392 | retry: |
80c54505 | 393 | radix_tree_preload(GFP_NOFS | __GFP_NOFAIL); |
769ec6e5 | 394 | |
67298804 | 395 | spin_lock(&im->ino_lock); |
67298804 | 396 | e = radix_tree_lookup(&im->ino_root, ino); |
39efac41 | 397 | if (!e) { |
80c54505 | 398 | e = tmp; |
67298804 CY |
399 | if (radix_tree_insert(&im->ino_root, ino, e)) { |
400 | spin_unlock(&im->ino_lock); | |
769ec6e5 | 401 | radix_tree_preload_end(); |
39efac41 JK |
402 | goto retry; |
403 | } | |
404 | memset(e, 0, sizeof(struct ino_entry)); | |
405 | e->ino = ino; | |
953e6cc6 | 406 | |
67298804 | 407 | list_add_tail(&e->list, &im->ino_list); |
8c402946 | 408 | if (type != ORPHAN_INO) |
67298804 | 409 | im->ino_num++; |
39efac41 | 410 | } |
67298804 | 411 | spin_unlock(&im->ino_lock); |
769ec6e5 | 412 | radix_tree_preload_end(); |
80c54505 JK |
413 | |
414 | if (e != tmp) | |
415 | kmem_cache_free(ino_entry_slab, tmp); | |
953e6cc6 JK |
416 | } |
417 | ||
6451e041 | 418 | static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 419 | { |
67298804 | 420 | struct inode_management *im = &sbi->im[type]; |
6451e041 | 421 | struct ino_entry *e; |
953e6cc6 | 422 | |
67298804 CY |
423 | spin_lock(&im->ino_lock); |
424 | e = radix_tree_lookup(&im->ino_root, ino); | |
39efac41 JK |
425 | if (e) { |
426 | list_del(&e->list); | |
67298804 CY |
427 | radix_tree_delete(&im->ino_root, ino); |
428 | im->ino_num--; | |
429 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
430 | kmem_cache_free(ino_entry_slab, e); |
431 | return; | |
953e6cc6 | 432 | } |
67298804 | 433 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
434 | } |
435 | ||
a49324f1 | 436 | void add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
437 | { |
438 | /* add new dirty ino entry into list */ | |
439 | __add_ino_entry(sbi, ino, type); | |
440 | } | |
441 | ||
a49324f1 | 442 | void remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
443 | { |
444 | /* remove dirty ino entry from list */ | |
445 | __remove_ino_entry(sbi, ino, type); | |
446 | } | |
447 | ||
448 | /* mode should be APPEND_INO or UPDATE_INO */ | |
449 | bool exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode) | |
450 | { | |
67298804 | 451 | struct inode_management *im = &sbi->im[mode]; |
fff04f90 | 452 | struct ino_entry *e; |
67298804 CY |
453 | |
454 | spin_lock(&im->ino_lock); | |
455 | e = radix_tree_lookup(&im->ino_root, ino); | |
456 | spin_unlock(&im->ino_lock); | |
fff04f90 JK |
457 | return e ? true : false; |
458 | } | |
459 | ||
74ef9241 | 460 | void release_ino_entry(struct f2fs_sb_info *sbi, bool all) |
fff04f90 JK |
461 | { |
462 | struct ino_entry *e, *tmp; | |
463 | int i; | |
464 | ||
74ef9241 | 465 | for (i = all ? ORPHAN_INO: APPEND_INO; i <= UPDATE_INO; i++) { |
67298804 CY |
466 | struct inode_management *im = &sbi->im[i]; |
467 | ||
468 | spin_lock(&im->ino_lock); | |
469 | list_for_each_entry_safe(e, tmp, &im->ino_list, list) { | |
fff04f90 | 470 | list_del(&e->list); |
67298804 | 471 | radix_tree_delete(&im->ino_root, e->ino); |
fff04f90 | 472 | kmem_cache_free(ino_entry_slab, e); |
67298804 | 473 | im->ino_num--; |
fff04f90 | 474 | } |
67298804 | 475 | spin_unlock(&im->ino_lock); |
fff04f90 JK |
476 | } |
477 | } | |
478 | ||
cbd56e7d | 479 | int acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 480 | { |
67298804 | 481 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a JK |
482 | int err = 0; |
483 | ||
67298804 | 484 | spin_lock(&im->ino_lock); |
cb78942b JK |
485 | |
486 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
487 | if (time_to_inject(FAULT_ORPHAN)) { | |
488 | spin_unlock(&im->ino_lock); | |
489 | return -ENOSPC; | |
490 | } | |
491 | #endif | |
67298804 | 492 | if (unlikely(im->ino_num >= sbi->max_orphans)) |
127e670a | 493 | err = -ENOSPC; |
cbd56e7d | 494 | else |
67298804 CY |
495 | im->ino_num++; |
496 | spin_unlock(&im->ino_lock); | |
0d47c1ad | 497 | |
127e670a JK |
498 | return err; |
499 | } | |
500 | ||
cbd56e7d JK |
501 | void release_orphan_inode(struct f2fs_sb_info *sbi) |
502 | { | |
67298804 CY |
503 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
504 | ||
505 | spin_lock(&im->ino_lock); | |
506 | f2fs_bug_on(sbi, im->ino_num == 0); | |
507 | im->ino_num--; | |
508 | spin_unlock(&im->ino_lock); | |
cbd56e7d JK |
509 | } |
510 | ||
127e670a JK |
511 | void add_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
512 | { | |
39efac41 | 513 | /* add new orphan ino entry into list */ |
6451e041 | 514 | __add_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
515 | } |
516 | ||
517 | void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
518 | { | |
953e6cc6 | 519 | /* remove orphan entry from orphan list */ |
6451e041 | 520 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
521 | } |
522 | ||
8c14bfad | 523 | static int recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 524 | { |
8c14bfad CY |
525 | struct inode *inode; |
526 | ||
527 | inode = f2fs_iget(sbi->sb, ino); | |
528 | if (IS_ERR(inode)) { | |
529 | /* | |
530 | * there should be a bug that we can't find the entry | |
531 | * to orphan inode. | |
532 | */ | |
533 | f2fs_bug_on(sbi, PTR_ERR(inode) == -ENOENT); | |
534 | return PTR_ERR(inode); | |
535 | } | |
536 | ||
127e670a | 537 | clear_nlink(inode); |
a1961246 | 538 | mark_inode_dirty_sync(inode); |
127e670a JK |
539 | |
540 | /* truncate all the data during iput */ | |
541 | iput(inode); | |
8c14bfad | 542 | return 0; |
127e670a JK |
543 | } |
544 | ||
8c14bfad | 545 | int recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a | 546 | { |
3c642985 | 547 | block_t start_blk, orphan_blocks, i, j; |
8c14bfad | 548 | int err; |
127e670a | 549 | |
25ca923b | 550 | if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG)) |
8c14bfad | 551 | return 0; |
127e670a | 552 | |
55141486 | 553 | start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi); |
3c642985 | 554 | orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi); |
127e670a | 555 | |
26879fb1 | 556 | ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP, true); |
662befda | 557 | |
3c642985 | 558 | for (i = 0; i < orphan_blocks; i++) { |
127e670a JK |
559 | struct page *page = get_meta_page(sbi, start_blk + i); |
560 | struct f2fs_orphan_block *orphan_blk; | |
561 | ||
562 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); | |
563 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
564 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
8c14bfad CY |
565 | err = recover_orphan_inode(sbi, ino); |
566 | if (err) { | |
567 | f2fs_put_page(page, 1); | |
8c14bfad CY |
568 | return err; |
569 | } | |
127e670a JK |
570 | } |
571 | f2fs_put_page(page, 1); | |
572 | } | |
573 | /* clear Orphan Flag */ | |
25ca923b | 574 | clear_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG); |
8c14bfad | 575 | return 0; |
127e670a JK |
576 | } |
577 | ||
578 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
579 | { | |
502c6e0b | 580 | struct list_head *head; |
127e670a | 581 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 582 | unsigned int nentries = 0; |
bd936f84 | 583 | unsigned short index = 1; |
8c402946 | 584 | unsigned short orphan_blocks; |
4531929e | 585 | struct page *page = NULL; |
6451e041 | 586 | struct ino_entry *orphan = NULL; |
67298804 | 587 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 588 | |
67298804 | 589 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 590 | |
d6c67a4f JK |
591 | /* |
592 | * we don't need to do spin_lock(&im->ino_lock) here, since all the | |
593 | * orphan inode operations are covered under f2fs_lock_op(). | |
594 | * And, spin_lock should be avoided due to page operations below. | |
595 | */ | |
67298804 | 596 | head = &im->ino_list; |
127e670a JK |
597 | |
598 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
599 | list_for_each_entry(orphan, head, list) { |
600 | if (!page) { | |
bd936f84 | 601 | page = grab_meta_page(sbi, start_blk++); |
502c6e0b GZ |
602 | orphan_blk = |
603 | (struct f2fs_orphan_block *)page_address(page); | |
604 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
605 | } | |
127e670a | 606 | |
36795567 | 607 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 608 | |
36795567 | 609 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
610 | /* |
611 | * an orphan block is full of 1020 entries, | |
612 | * then we need to flush current orphan blocks | |
613 | * and bring another one in memory | |
614 | */ | |
615 | orphan_blk->blk_addr = cpu_to_le16(index); | |
616 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
617 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
618 | set_page_dirty(page); | |
619 | f2fs_put_page(page, 1); | |
620 | index++; | |
127e670a JK |
621 | nentries = 0; |
622 | page = NULL; | |
623 | } | |
502c6e0b | 624 | } |
127e670a | 625 | |
502c6e0b GZ |
626 | if (page) { |
627 | orphan_blk->blk_addr = cpu_to_le16(index); | |
628 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
629 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
630 | set_page_dirty(page); | |
631 | f2fs_put_page(page, 1); | |
127e670a | 632 | } |
127e670a JK |
633 | } |
634 | ||
635 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, | |
636 | block_t cp_addr, unsigned long long *version) | |
637 | { | |
638 | struct page *cp_page_1, *cp_page_2 = NULL; | |
639 | unsigned long blk_size = sbi->blocksize; | |
640 | struct f2fs_checkpoint *cp_block; | |
641 | unsigned long long cur_version = 0, pre_version = 0; | |
127e670a | 642 | size_t crc_offset; |
7e586fa0 | 643 | __u32 crc = 0; |
127e670a JK |
644 | |
645 | /* Read the 1st cp block in this CP pack */ | |
646 | cp_page_1 = get_meta_page(sbi, cp_addr); | |
647 | ||
648 | /* get the version number */ | |
649 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_1); | |
650 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
651 | if (crc_offset >= blk_size) | |
652 | goto invalid_cp1; | |
653 | ||
29e7043f | 654 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
43b6573b | 655 | if (!f2fs_crc_valid(sbi, crc, cp_block, crc_offset)) |
127e670a JK |
656 | goto invalid_cp1; |
657 | ||
d71b5564 | 658 | pre_version = cur_cp_version(cp_block); |
127e670a JK |
659 | |
660 | /* Read the 2nd cp block in this CP pack */ | |
25ca923b | 661 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
127e670a JK |
662 | cp_page_2 = get_meta_page(sbi, cp_addr); |
663 | ||
664 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_2); | |
665 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
666 | if (crc_offset >= blk_size) | |
667 | goto invalid_cp2; | |
668 | ||
29e7043f | 669 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
43b6573b | 670 | if (!f2fs_crc_valid(sbi, crc, cp_block, crc_offset)) |
127e670a JK |
671 | goto invalid_cp2; |
672 | ||
d71b5564 | 673 | cur_version = cur_cp_version(cp_block); |
127e670a JK |
674 | |
675 | if (cur_version == pre_version) { | |
676 | *version = cur_version; | |
677 | f2fs_put_page(cp_page_2, 1); | |
678 | return cp_page_1; | |
679 | } | |
680 | invalid_cp2: | |
681 | f2fs_put_page(cp_page_2, 1); | |
682 | invalid_cp1: | |
683 | f2fs_put_page(cp_page_1, 1); | |
684 | return NULL; | |
685 | } | |
686 | ||
687 | int get_valid_checkpoint(struct f2fs_sb_info *sbi) | |
688 | { | |
689 | struct f2fs_checkpoint *cp_block; | |
690 | struct f2fs_super_block *fsb = sbi->raw_super; | |
691 | struct page *cp1, *cp2, *cur_page; | |
692 | unsigned long blk_size = sbi->blocksize; | |
693 | unsigned long long cp1_version = 0, cp2_version = 0; | |
694 | unsigned long long cp_start_blk_no; | |
55141486 | 695 | unsigned int cp_blks = 1 + __cp_payload(sbi); |
1dbe4152 CL |
696 | block_t cp_blk_no; |
697 | int i; | |
127e670a | 698 | |
1dbe4152 | 699 | sbi->ckpt = kzalloc(cp_blks * blk_size, GFP_KERNEL); |
127e670a JK |
700 | if (!sbi->ckpt) |
701 | return -ENOMEM; | |
702 | /* | |
703 | * Finding out valid cp block involves read both | |
704 | * sets( cp pack1 and cp pack 2) | |
705 | */ | |
706 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
707 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
708 | ||
709 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
710 | cp_start_blk_no += ((unsigned long long)1) << |
711 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
712 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
713 | ||
714 | if (cp1 && cp2) { | |
715 | if (ver_after(cp2_version, cp1_version)) | |
716 | cur_page = cp2; | |
717 | else | |
718 | cur_page = cp1; | |
719 | } else if (cp1) { | |
720 | cur_page = cp1; | |
721 | } else if (cp2) { | |
722 | cur_page = cp2; | |
723 | } else { | |
724 | goto fail_no_cp; | |
725 | } | |
726 | ||
727 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
728 | memcpy(sbi->ckpt, cp_block, blk_size); | |
729 | ||
984ec63c SL |
730 | /* Sanity checking of checkpoint */ |
731 | if (sanity_check_ckpt(sbi)) | |
732 | goto fail_no_cp; | |
733 | ||
1dbe4152 CL |
734 | if (cp_blks <= 1) |
735 | goto done; | |
736 | ||
737 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
738 | if (cur_page == cp2) | |
739 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
740 | ||
741 | for (i = 1; i < cp_blks; i++) { | |
742 | void *sit_bitmap_ptr; | |
743 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
744 | ||
745 | cur_page = get_meta_page(sbi, cp_blk_no + i); | |
746 | sit_bitmap_ptr = page_address(cur_page); | |
747 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
748 | f2fs_put_page(cur_page, 1); | |
749 | } | |
750 | done: | |
127e670a JK |
751 | f2fs_put_page(cp1, 1); |
752 | f2fs_put_page(cp2, 1); | |
753 | return 0; | |
754 | ||
755 | fail_no_cp: | |
756 | kfree(sbi->ckpt); | |
757 | return -EINVAL; | |
758 | } | |
759 | ||
c227f912 | 760 | static void __add_dirty_inode(struct inode *inode, enum inode_type type) |
127e670a | 761 | { |
4081363f | 762 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 763 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
127e670a | 764 | |
91942321 | 765 | if (is_inode_flag_set(inode, flag)) |
2710fd7e | 766 | return; |
2d7b822a | 767 | |
91942321 JK |
768 | set_inode_flag(inode, flag); |
769 | list_add_tail(&F2FS_I(inode)->dirty_list, &sbi->inode_list[type]); | |
33fbd510 | 770 | stat_inc_dirty_inode(sbi, type); |
5deb8267 JK |
771 | } |
772 | ||
c227f912 | 773 | static void __remove_dirty_inode(struct inode *inode, enum inode_type type) |
6ad7609a | 774 | { |
c227f912 | 775 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
6ad7609a | 776 | |
91942321 | 777 | if (get_dirty_pages(inode) || !is_inode_flag_set(inode, flag)) |
6ad7609a CY |
778 | return; |
779 | ||
91942321 JK |
780 | list_del_init(&F2FS_I(inode)->dirty_list); |
781 | clear_inode_flag(inode, flag); | |
33fbd510 | 782 | stat_dec_dirty_inode(F2FS_I_SB(inode), type); |
6ad7609a CY |
783 | } |
784 | ||
a7ffdbe2 | 785 | void update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 786 | { |
4081363f | 787 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 788 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
5deb8267 | 789 | |
5ac9f36f CY |
790 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
791 | !S_ISLNK(inode->i_mode)) | |
127e670a | 792 | return; |
7bd59381 | 793 | |
1c4bf763 JK |
794 | spin_lock(&sbi->inode_lock[type]); |
795 | if (type != FILE_INODE || test_opt(sbi, DATA_FLUSH)) | |
10aa97c3 | 796 | __add_dirty_inode(inode, type); |
b951a4ec | 797 | inode_inc_dirty_pages(inode); |
1c4bf763 JK |
798 | spin_unlock(&sbi->inode_lock[type]); |
799 | ||
a7ffdbe2 | 800 | SetPagePrivate(page); |
9e4ded3f | 801 | f2fs_trace_pid(page); |
5deb8267 JK |
802 | } |
803 | ||
c227f912 | 804 | void remove_dirty_inode(struct inode *inode) |
127e670a | 805 | { |
4081363f | 806 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 807 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
127e670a | 808 | |
c227f912 CY |
809 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
810 | !S_ISLNK(inode->i_mode)) | |
127e670a JK |
811 | return; |
812 | ||
10aa97c3 JK |
813 | if (type == FILE_INODE && !test_opt(sbi, DATA_FLUSH)) |
814 | return; | |
815 | ||
c227f912 CY |
816 | spin_lock(&sbi->inode_lock[type]); |
817 | __remove_dirty_inode(inode, type); | |
818 | spin_unlock(&sbi->inode_lock[type]); | |
74d0b917 JK |
819 | } |
820 | ||
6d5a1495 | 821 | int sync_dirty_inodes(struct f2fs_sb_info *sbi, enum inode_type type) |
127e670a | 822 | { |
ce3b7d80 | 823 | struct list_head *head; |
127e670a | 824 | struct inode *inode; |
2710fd7e | 825 | struct f2fs_inode_info *fi; |
4cf18537 CY |
826 | bool is_dir = (type == DIR_INODE); |
827 | ||
828 | trace_f2fs_sync_dirty_inodes_enter(sbi->sb, is_dir, | |
829 | get_pages(sbi, is_dir ? | |
830 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
127e670a | 831 | retry: |
af41d3ee | 832 | if (unlikely(f2fs_cp_error(sbi))) |
6d5a1495 | 833 | return -EIO; |
af41d3ee | 834 | |
c227f912 | 835 | spin_lock(&sbi->inode_lock[type]); |
ce3b7d80 | 836 | |
c227f912 | 837 | head = &sbi->inode_list[type]; |
127e670a | 838 | if (list_empty(head)) { |
c227f912 | 839 | spin_unlock(&sbi->inode_lock[type]); |
4cf18537 CY |
840 | trace_f2fs_sync_dirty_inodes_exit(sbi->sb, is_dir, |
841 | get_pages(sbi, is_dir ? | |
842 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
6d5a1495 | 843 | return 0; |
127e670a | 844 | } |
2710fd7e CY |
845 | fi = list_entry(head->next, struct f2fs_inode_info, dirty_list); |
846 | inode = igrab(&fi->vfs_inode); | |
c227f912 | 847 | spin_unlock(&sbi->inode_lock[type]); |
127e670a | 848 | if (inode) { |
87d6f890 | 849 | filemap_fdatawrite(inode->i_mapping); |
127e670a JK |
850 | iput(inode); |
851 | } else { | |
852 | /* | |
853 | * We should submit bio, since it exists several | |
854 | * wribacking dentry pages in the freeing inode. | |
855 | */ | |
458e6197 | 856 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
7ecebe5e | 857 | cond_resched(); |
127e670a JK |
858 | } |
859 | goto retry; | |
860 | } | |
861 | ||
0f18b462 JK |
862 | int f2fs_sync_inode_meta(struct f2fs_sb_info *sbi) |
863 | { | |
864 | struct list_head *head = &sbi->inode_list[DIRTY_META]; | |
865 | struct inode *inode; | |
866 | struct f2fs_inode_info *fi; | |
867 | s64 total = get_pages(sbi, F2FS_DIRTY_IMETA); | |
868 | ||
869 | while (total--) { | |
870 | if (unlikely(f2fs_cp_error(sbi))) | |
871 | return -EIO; | |
872 | ||
873 | spin_lock(&sbi->inode_lock[DIRTY_META]); | |
874 | if (list_empty(head)) { | |
875 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
876 | return 0; | |
877 | } | |
878 | fi = list_entry(head->next, struct f2fs_inode_info, | |
879 | gdirty_list); | |
880 | inode = igrab(&fi->vfs_inode); | |
881 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
882 | if (inode) { | |
883 | update_inode_page(inode); | |
884 | iput(inode); | |
885 | } | |
886 | }; | |
887 | return 0; | |
888 | } | |
889 | ||
0a8165d7 | 890 | /* |
127e670a JK |
891 | * Freeze all the FS-operations for checkpoint. |
892 | */ | |
cf779cab | 893 | static int block_operations(struct f2fs_sb_info *sbi) |
127e670a | 894 | { |
127e670a JK |
895 | struct writeback_control wbc = { |
896 | .sync_mode = WB_SYNC_ALL, | |
897 | .nr_to_write = LONG_MAX, | |
898 | .for_reclaim = 0, | |
899 | }; | |
cf779cab | 900 | int err = 0; |
c718379b | 901 | |
39936837 | 902 | retry_flush_dents: |
e479556b | 903 | f2fs_lock_all(sbi); |
127e670a | 904 | /* write all the dirty dentry pages */ |
127e670a | 905 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 906 | f2fs_unlock_all(sbi); |
6d5a1495 CY |
907 | err = sync_dirty_inodes(sbi, DIR_INODE); |
908 | if (err) | |
cf779cab | 909 | goto out; |
39936837 | 910 | goto retry_flush_dents; |
127e670a JK |
911 | } |
912 | ||
0f18b462 JK |
913 | if (get_pages(sbi, F2FS_DIRTY_IMETA)) { |
914 | f2fs_unlock_all(sbi); | |
915 | err = f2fs_sync_inode_meta(sbi); | |
916 | if (err) | |
917 | goto out; | |
918 | goto retry_flush_dents; | |
919 | } | |
920 | ||
127e670a | 921 | /* |
e1c42045 | 922 | * POR: we should ensure that there are no dirty node pages |
127e670a JK |
923 | * until finishing nat/sit flush. |
924 | */ | |
39936837 | 925 | retry_flush_nodes: |
b3582c68 | 926 | down_write(&sbi->node_write); |
127e670a JK |
927 | |
928 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
b3582c68 | 929 | up_write(&sbi->node_write); |
52681375 | 930 | err = sync_node_pages(sbi, &wbc); |
6d5a1495 | 931 | if (err) { |
cf779cab | 932 | f2fs_unlock_all(sbi); |
cf779cab JK |
933 | goto out; |
934 | } | |
39936837 | 935 | goto retry_flush_nodes; |
127e670a | 936 | } |
cf779cab | 937 | out: |
cf779cab | 938 | return err; |
127e670a JK |
939 | } |
940 | ||
941 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
942 | { | |
b3582c68 | 943 | up_write(&sbi->node_write); |
e479556b | 944 | f2fs_unlock_all(sbi); |
127e670a JK |
945 | } |
946 | ||
fb51b5ef CL |
947 | static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi) |
948 | { | |
949 | DEFINE_WAIT(wait); | |
950 | ||
951 | for (;;) { | |
952 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
953 | ||
f5730184 | 954 | if (!atomic_read(&sbi->nr_wb_bios)) |
fb51b5ef CL |
955 | break; |
956 | ||
0ff21646 | 957 | io_schedule_timeout(5*HZ); |
fb51b5ef CL |
958 | } |
959 | finish_wait(&sbi->cp_wait, &wait); | |
960 | } | |
961 | ||
c34f42e2 | 962 | static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
963 | { |
964 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
cf2271e7 | 965 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); |
77041823 | 966 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
67298804 | 967 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; |
77041823 | 968 | nid_t last_nid = nm_i->next_scan_nid; |
127e670a | 969 | block_t start_blk; |
127e670a | 970 | unsigned int data_sum_blocks, orphan_blocks; |
7e586fa0 | 971 | __u32 crc32 = 0; |
127e670a | 972 | int i; |
55141486 | 973 | int cp_payload_blks = __cp_payload(sbi); |
e90c2d28 CY |
974 | block_t discard_blk = NEXT_FREE_BLKADDR(sbi, curseg); |
975 | bool invalidate = false; | |
8f1dbbbb SL |
976 | struct super_block *sb = sbi->sb; |
977 | struct curseg_info *seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); | |
978 | u64 kbytes_written; | |
127e670a | 979 | |
1e87a78d JK |
980 | /* |
981 | * This avoids to conduct wrong roll-forward operations and uses | |
982 | * metapages, so should be called prior to sync_meta_pages below. | |
983 | */ | |
e90c2d28 CY |
984 | if (discard_next_dnode(sbi, discard_blk)) |
985 | invalidate = true; | |
127e670a JK |
986 | |
987 | /* Flush all the NAT/SIT pages */ | |
cf779cab | 988 | while (get_pages(sbi, F2FS_DIRTY_META)) { |
127e670a | 989 | sync_meta_pages(sbi, META, LONG_MAX); |
cf779cab | 990 | if (unlikely(f2fs_cp_error(sbi))) |
c34f42e2 | 991 | return -EIO; |
cf779cab | 992 | } |
127e670a JK |
993 | |
994 | next_free_nid(sbi, &last_nid); | |
995 | ||
996 | /* | |
997 | * modify checkpoint | |
998 | * version number is already updated | |
999 | */ | |
1000 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi)); | |
1001 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
1002 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); | |
b5b82205 | 1003 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
1004 | ckpt->cur_node_segno[i] = |
1005 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
1006 | ckpt->cur_node_blkoff[i] = | |
1007 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
1008 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
1009 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
1010 | } | |
b5b82205 | 1011 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
1012 | ckpt->cur_data_segno[i] = |
1013 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
1014 | ckpt->cur_data_blkoff[i] = | |
1015 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
1016 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
1017 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
1018 | } | |
1019 | ||
1020 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
1021 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
1022 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
1023 | ||
1024 | /* 2 cp + n data seg summary + orphan inode blocks */ | |
3fa06d7b | 1025 | data_sum_blocks = npages_for_summary_flush(sbi, false); |
b5b82205 | 1026 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
25ca923b | 1027 | set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 1028 | else |
25ca923b | 1029 | clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 1030 | |
67298804 | 1031 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
1032 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
1033 | orphan_blocks); | |
127e670a | 1034 | |
119ee914 | 1035 | if (__remain_node_summaries(cpc->reason)) |
b5b82205 | 1036 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
1037 | cp_payload_blks + data_sum_blocks + |
1038 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
119ee914 | 1039 | else |
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); | |
119ee914 JK |
1043 | |
1044 | if (cpc->reason == CP_UMOUNT) | |
1045 | set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
1046 | else | |
1047 | clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
1048 | ||
1049 | if (cpc->reason == CP_FASTBOOT) | |
1050 | set_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
1051 | else | |
1052 | clear_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
127e670a | 1053 | |
67298804 | 1054 | if (orphan_num) |
25ca923b | 1055 | set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1056 | else |
25ca923b | 1057 | clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1058 | |
caf0047e | 1059 | if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) |
2ae4c673 JK |
1060 | set_ckpt_flags(ckpt, CP_FSCK_FLAG); |
1061 | ||
127e670a JK |
1062 | /* update SIT/NAT bitmap */ |
1063 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
1064 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
1065 | ||
43b6573b | 1066 | crc32 = f2fs_crc32(sbi, ckpt, le32_to_cpu(ckpt->checksum_offset)); |
7e586fa0 JK |
1067 | *((__le32 *)((unsigned char *)ckpt + |
1068 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
1069 | = cpu_to_le32(crc32); |
1070 | ||
1071 | start_blk = __start_cp_addr(sbi); | |
1072 | ||
a7230d16 JK |
1073 | /* need to wait for end_io results */ |
1074 | wait_on_all_pages_writeback(sbi); | |
1075 | if (unlikely(f2fs_cp_error(sbi))) | |
c34f42e2 | 1076 | return -EIO; |
a7230d16 | 1077 | |
127e670a | 1078 | /* write out checkpoint buffer at block 0 */ |
381722d2 CY |
1079 | update_meta_page(sbi, ckpt, start_blk++); |
1080 | ||
1081 | for (i = 1; i < 1 + cp_payload_blks; i++) | |
1082 | update_meta_page(sbi, (char *)ckpt + i * F2FS_BLKSIZE, | |
1083 | start_blk++); | |
1dbe4152 | 1084 | |
67298804 | 1085 | if (orphan_num) { |
127e670a JK |
1086 | write_orphan_inodes(sbi, start_blk); |
1087 | start_blk += orphan_blocks; | |
1088 | } | |
1089 | ||
1090 | write_data_summaries(sbi, start_blk); | |
1091 | start_blk += data_sum_blocks; | |
8f1dbbbb SL |
1092 | |
1093 | /* Record write statistics in the hot node summary */ | |
1094 | kbytes_written = sbi->kbytes_written; | |
1095 | if (sb->s_bdev->bd_part) | |
1096 | kbytes_written += BD_PART_WRITTEN(sbi); | |
1097 | ||
b7ad7512 | 1098 | seg_i->journal->info.kbytes_written = cpu_to_le64(kbytes_written); |
8f1dbbbb | 1099 | |
119ee914 | 1100 | if (__remain_node_summaries(cpc->reason)) { |
127e670a JK |
1101 | write_node_summaries(sbi, start_blk); |
1102 | start_blk += NR_CURSEG_NODE_TYPE; | |
1103 | } | |
1104 | ||
1105 | /* writeout checkpoint block */ | |
381722d2 | 1106 | update_meta_page(sbi, ckpt, start_blk); |
127e670a JK |
1107 | |
1108 | /* wait for previous submitted node/meta pages writeback */ | |
fb51b5ef | 1109 | wait_on_all_pages_writeback(sbi); |
127e670a | 1110 | |
cf779cab | 1111 | if (unlikely(f2fs_cp_error(sbi))) |
c34f42e2 | 1112 | return -EIO; |
cf779cab | 1113 | |
80dd9c0e CY |
1114 | filemap_fdatawait_range(NODE_MAPPING(sbi), 0, LLONG_MAX); |
1115 | filemap_fdatawait_range(META_MAPPING(sbi), 0, LLONG_MAX); | |
127e670a JK |
1116 | |
1117 | /* update user_block_counts */ | |
1118 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
41382ec4 | 1119 | percpu_counter_set(&sbi->alloc_valid_block_count, 0); |
127e670a JK |
1120 | |
1121 | /* Here, we only have one bio having CP pack */ | |
577e3495 | 1122 | sync_meta_pages(sbi, META_FLUSH, LONG_MAX); |
127e670a | 1123 | |
6a8f8ca5 JK |
1124 | /* wait for previous submitted meta pages writeback */ |
1125 | wait_on_all_pages_writeback(sbi); | |
1126 | ||
e90c2d28 CY |
1127 | /* |
1128 | * invalidate meta page which is used temporarily for zeroing out | |
1129 | * block at the end of warm node chain. | |
1130 | */ | |
1131 | if (invalidate) | |
1132 | invalidate_mapping_pages(META_MAPPING(sbi), discard_blk, | |
1133 | discard_blk); | |
1134 | ||
74ef9241 | 1135 | release_ino_entry(sbi, false); |
cf779cab JK |
1136 | |
1137 | if (unlikely(f2fs_cp_error(sbi))) | |
c34f42e2 | 1138 | return -EIO; |
cf779cab | 1139 | |
836b5a63 | 1140 | clear_prefree_segments(sbi, cpc); |
caf0047e | 1141 | clear_sbi_flag(sbi, SBI_IS_DIRTY); |
c34f42e2 CY |
1142 | |
1143 | return 0; | |
127e670a JK |
1144 | } |
1145 | ||
0a8165d7 | 1146 | /* |
e1c42045 | 1147 | * We guarantee that this checkpoint procedure will not fail. |
127e670a | 1148 | */ |
c34f42e2 | 1149 | int write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1150 | { |
1151 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1152 | unsigned long long ckpt_ver; | |
c34f42e2 | 1153 | int err = 0; |
127e670a | 1154 | |
43727527 | 1155 | mutex_lock(&sbi->cp_mutex); |
8501017e | 1156 | |
caf0047e | 1157 | if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) && |
a66cdd98 JK |
1158 | (cpc->reason == CP_FASTBOOT || cpc->reason == CP_SYNC || |
1159 | (cpc->reason == CP_DISCARD && !sbi->discard_blks))) | |
8501017e | 1160 | goto out; |
c34f42e2 CY |
1161 | if (unlikely(f2fs_cp_error(sbi))) { |
1162 | err = -EIO; | |
cf779cab | 1163 | goto out; |
c34f42e2 CY |
1164 | } |
1165 | if (f2fs_readonly(sbi->sb)) { | |
1166 | err = -EROFS; | |
11504a8e | 1167 | goto out; |
c34f42e2 | 1168 | } |
2bda542d WL |
1169 | |
1170 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); | |
1171 | ||
c34f42e2 CY |
1172 | err = block_operations(sbi); |
1173 | if (err) | |
cf779cab | 1174 | goto out; |
127e670a | 1175 | |
75ab4cb8 | 1176 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1177 | |
406657dd | 1178 | f2fs_flush_merged_bios(sbi); |
127e670a JK |
1179 | |
1180 | /* | |
1181 | * update checkpoint pack index | |
1182 | * Increase the version number so that | |
1183 | * SIT entries and seg summaries are written at correct place | |
1184 | */ | |
d71b5564 | 1185 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1186 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1187 | ||
1188 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
1189 | flush_nat_entries(sbi); | |
4b2fecc8 | 1190 | flush_sit_entries(sbi, cpc); |
127e670a | 1191 | |
127e670a | 1192 | /* unlock all the fs_lock[] in do_checkpoint() */ |
c34f42e2 | 1193 | err = do_checkpoint(sbi, cpc); |
127e670a JK |
1194 | |
1195 | unblock_operations(sbi); | |
942e0be6 | 1196 | stat_inc_cp_count(sbi->stat_info); |
10027551 JK |
1197 | |
1198 | if (cpc->reason == CP_RECOVERY) | |
1199 | f2fs_msg(sbi->sb, KERN_NOTICE, | |
1200 | "checkpoint: version = %llx", ckpt_ver); | |
60b99b48 JK |
1201 | |
1202 | /* do checkpoint periodically */ | |
6beceb54 | 1203 | f2fs_update_time(sbi, CP_TIME); |
55d1cdb2 | 1204 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
8501017e JK |
1205 | out: |
1206 | mutex_unlock(&sbi->cp_mutex); | |
c34f42e2 | 1207 | return err; |
127e670a JK |
1208 | } |
1209 | ||
6451e041 | 1210 | void init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1211 | { |
6451e041 JK |
1212 | int i; |
1213 | ||
1214 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1215 | struct inode_management *im = &sbi->im[i]; |
1216 | ||
1217 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1218 | spin_lock_init(&im->ino_lock); | |
1219 | INIT_LIST_HEAD(&im->ino_list); | |
1220 | im->ino_num = 0; | |
6451e041 JK |
1221 | } |
1222 | ||
b5b82205 | 1223 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
14b42817 WL |
1224 | NR_CURSEG_TYPE - __cp_payload(sbi)) * |
1225 | F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
1226 | } |
1227 | ||
6e6093a8 | 1228 | int __init create_checkpoint_caches(void) |
127e670a | 1229 | { |
6451e041 JK |
1230 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1231 | sizeof(struct ino_entry)); | |
1232 | if (!ino_entry_slab) | |
127e670a | 1233 | return -ENOMEM; |
06292073 CY |
1234 | inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry", |
1235 | sizeof(struct inode_entry)); | |
6bacf52f | 1236 | if (!inode_entry_slab) { |
6451e041 | 1237 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1238 | return -ENOMEM; |
1239 | } | |
1240 | return 0; | |
1241 | } | |
1242 | ||
1243 | void destroy_checkpoint_caches(void) | |
1244 | { | |
6451e041 | 1245 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1246 | kmem_cache_destroy(inode_entry_slab); |
1247 | } |