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 | 65 | .type = META, |
04d328de MC |
66 | .op = REQ_OP_READ, |
67 | .op_flags = 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)) | |
04d328de | 74 | fio.op_flags &= ~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, |
04d328de | 161 | .op = REQ_OP_READ, |
70246286 | 162 | .op_flags = sync ? (READ_SYNC | REQ_META | REQ_PRIO) : REQ_RAHEAD, |
4375a336 | 163 | .encrypted_page = NULL, |
662befda | 164 | }; |
e9f5b8b8 | 165 | struct blk_plug plug; |
662befda | 166 | |
2b947003 | 167 | if (unlikely(type == META_POR)) |
04d328de | 168 | fio.op_flags &= ~REQ_META; |
2b947003 | 169 | |
e9f5b8b8 | 170 | blk_start_plug(&plug); |
662befda | 171 | for (; nrpages-- > 0; blkno++) { |
662befda | 172 | |
66b00c18 CY |
173 | if (!is_valid_blkaddr(sbi, blkno, type)) |
174 | goto out; | |
175 | ||
662befda CY |
176 | switch (type) { |
177 | case META_NAT: | |
66b00c18 CY |
178 | if (unlikely(blkno >= |
179 | NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))) | |
662befda | 180 | blkno = 0; |
66b00c18 | 181 | /* get nat block addr */ |
7a9d7548 | 182 | fio.new_blkaddr = current_nat_addr(sbi, |
662befda CY |
183 | blkno * NAT_ENTRY_PER_BLOCK); |
184 | break; | |
185 | case META_SIT: | |
186 | /* get sit block addr */ | |
7a9d7548 | 187 | fio.new_blkaddr = current_sit_addr(sbi, |
662befda | 188 | blkno * SIT_ENTRY_PER_BLOCK); |
662befda | 189 | break; |
81c1a0f1 | 190 | case META_SSA: |
662befda | 191 | case META_CP: |
4c521f49 | 192 | case META_POR: |
7a9d7548 | 193 | fio.new_blkaddr = blkno; |
662befda CY |
194 | break; |
195 | default: | |
196 | BUG(); | |
197 | } | |
198 | ||
300e129c JK |
199 | page = f2fs_grab_cache_page(META_MAPPING(sbi), |
200 | fio.new_blkaddr, false); | |
662befda CY |
201 | if (!page) |
202 | continue; | |
203 | if (PageUptodate(page)) { | |
662befda CY |
204 | f2fs_put_page(page, 1); |
205 | continue; | |
206 | } | |
207 | ||
05ca3632 | 208 | fio.page = page; |
7a9d7548 | 209 | fio.old_blkaddr = fio.new_blkaddr; |
05ca3632 | 210 | f2fs_submit_page_mbio(&fio); |
662befda CY |
211 | f2fs_put_page(page, 0); |
212 | } | |
213 | out: | |
214 | f2fs_submit_merged_bio(sbi, META, READ); | |
e9f5b8b8 | 215 | blk_finish_plug(&plug); |
662befda CY |
216 | return blkno - start; |
217 | } | |
218 | ||
635aee1f CY |
219 | void ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index) |
220 | { | |
221 | struct page *page; | |
222 | bool readahead = false; | |
223 | ||
224 | page = find_get_page(META_MAPPING(sbi), index); | |
4da7bf5a | 225 | if (!page || !PageUptodate(page)) |
635aee1f CY |
226 | readahead = true; |
227 | f2fs_put_page(page, 0); | |
228 | ||
229 | if (readahead) | |
26879fb1 | 230 | ra_meta_pages(sbi, index, MAX_BIO_BLOCKS(sbi), META_POR, true); |
635aee1f CY |
231 | } |
232 | ||
127e670a JK |
233 | static int f2fs_write_meta_page(struct page *page, |
234 | struct writeback_control *wbc) | |
235 | { | |
4081363f | 236 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 237 | |
ecda0de3 CY |
238 | trace_f2fs_writepage(page, META); |
239 | ||
caf0047e | 240 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
cfb271d4 | 241 | goto redirty_out; |
857dc4e0 | 242 | if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0)) |
cfb271d4 | 243 | goto redirty_out; |
1e968fdf | 244 | if (unlikely(f2fs_cp_error(sbi))) |
cf779cab | 245 | goto redirty_out; |
127e670a | 246 | |
577e3495 JK |
247 | write_meta_page(sbi, page); |
248 | dec_page_count(sbi, F2FS_DIRTY_META); | |
0c3a5797 CY |
249 | |
250 | if (wbc->for_reclaim) | |
251 | f2fs_submit_merged_bio_cond(sbi, NULL, page, 0, META, WRITE); | |
252 | ||
577e3495 | 253 | unlock_page(page); |
857dc4e0 | 254 | |
0c3a5797 | 255 | if (unlikely(f2fs_cp_error(sbi))) |
857dc4e0 | 256 | f2fs_submit_merged_bio(sbi, META, WRITE); |
0c3a5797 | 257 | |
577e3495 | 258 | return 0; |
cfb271d4 CY |
259 | |
260 | redirty_out: | |
76f60268 | 261 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 262 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
263 | } |
264 | ||
265 | static int f2fs_write_meta_pages(struct address_space *mapping, | |
266 | struct writeback_control *wbc) | |
267 | { | |
4081363f | 268 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 269 | long diff, written; |
127e670a | 270 | |
5459aa97 | 271 | /* collect a number of dirty meta pages and write together */ |
50c8cdb3 JK |
272 | if (wbc->for_kupdate || |
273 | get_pages(sbi, F2FS_DIRTY_META) < nr_pages_to_skip(sbi, META)) | |
d3baf95d | 274 | goto skip_write; |
127e670a | 275 | |
d31c7c3f YH |
276 | trace_f2fs_writepages(mapping->host, wbc, META); |
277 | ||
127e670a JK |
278 | /* if mounting is failed, skip writing node pages */ |
279 | mutex_lock(&sbi->cp_mutex); | |
50c8cdb3 JK |
280 | diff = nr_pages_to_write(sbi, META, wbc); |
281 | written = sync_meta_pages(sbi, META, wbc->nr_to_write); | |
127e670a | 282 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 283 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 284 | return 0; |
d3baf95d JK |
285 | |
286 | skip_write: | |
287 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
d31c7c3f | 288 | trace_f2fs_writepages(mapping->host, wbc, META); |
d3baf95d | 289 | return 0; |
127e670a JK |
290 | } |
291 | ||
292 | long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, | |
293 | long nr_to_write) | |
294 | { | |
9df27d98 | 295 | struct address_space *mapping = META_MAPPING(sbi); |
80dd9c0e | 296 | pgoff_t index = 0, end = ULONG_MAX, prev = ULONG_MAX; |
127e670a JK |
297 | struct pagevec pvec; |
298 | long nwritten = 0; | |
299 | struct writeback_control wbc = { | |
300 | .for_reclaim = 0, | |
301 | }; | |
e9f5b8b8 | 302 | struct blk_plug plug; |
127e670a JK |
303 | |
304 | pagevec_init(&pvec, 0); | |
305 | ||
e9f5b8b8 CY |
306 | blk_start_plug(&plug); |
307 | ||
127e670a JK |
308 | while (index <= end) { |
309 | int i, nr_pages; | |
310 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, | |
311 | PAGECACHE_TAG_DIRTY, | |
312 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
cfb271d4 | 313 | if (unlikely(nr_pages == 0)) |
127e670a JK |
314 | break; |
315 | ||
316 | for (i = 0; i < nr_pages; i++) { | |
317 | struct page *page = pvec.pages[i]; | |
203681f6 | 318 | |
80dd9c0e | 319 | if (prev == ULONG_MAX) |
6066d8cd JK |
320 | prev = page->index - 1; |
321 | if (nr_to_write != LONG_MAX && page->index != prev + 1) { | |
322 | pagevec_release(&pvec); | |
323 | goto stop; | |
324 | } | |
325 | ||
127e670a | 326 | lock_page(page); |
203681f6 JK |
327 | |
328 | if (unlikely(page->mapping != mapping)) { | |
329 | continue_unlock: | |
330 | unlock_page(page); | |
331 | continue; | |
332 | } | |
333 | if (!PageDirty(page)) { | |
334 | /* someone wrote it for us */ | |
335 | goto continue_unlock; | |
336 | } | |
337 | ||
fa3d2bdf JK |
338 | f2fs_wait_on_page_writeback(page, META, true); |
339 | ||
340 | BUG_ON(PageWriteback(page)); | |
203681f6 JK |
341 | if (!clear_page_dirty_for_io(page)) |
342 | goto continue_unlock; | |
343 | ||
97dc3fd2 | 344 | if (mapping->a_ops->writepage(page, &wbc)) { |
577e3495 JK |
345 | unlock_page(page); |
346 | break; | |
347 | } | |
cfb271d4 | 348 | nwritten++; |
6066d8cd | 349 | prev = page->index; |
cfb271d4 | 350 | if (unlikely(nwritten >= nr_to_write)) |
127e670a JK |
351 | break; |
352 | } | |
353 | pagevec_release(&pvec); | |
354 | cond_resched(); | |
355 | } | |
6066d8cd | 356 | stop: |
127e670a | 357 | if (nwritten) |
458e6197 | 358 | f2fs_submit_merged_bio(sbi, type, WRITE); |
127e670a | 359 | |
e9f5b8b8 CY |
360 | blk_finish_plug(&plug); |
361 | ||
127e670a JK |
362 | return nwritten; |
363 | } | |
364 | ||
365 | static int f2fs_set_meta_page_dirty(struct page *page) | |
366 | { | |
26c6b887 JK |
367 | trace_f2fs_set_page_dirty(page, META); |
368 | ||
127e670a JK |
369 | SetPageUptodate(page); |
370 | if (!PageDirty(page)) { | |
371 | __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 | ||
127e670a JK |
513 | void add_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
514 | { | |
39efac41 | 515 | /* add new orphan ino entry into list */ |
6451e041 | 516 | __add_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
517 | } |
518 | ||
519 | void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
520 | { | |
953e6cc6 | 521 | /* remove orphan entry from orphan list */ |
6451e041 | 522 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
523 | } |
524 | ||
8c14bfad | 525 | static int recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 526 | { |
8c14bfad CY |
527 | struct inode *inode; |
528 | ||
529 | inode = f2fs_iget(sbi->sb, ino); | |
530 | if (IS_ERR(inode)) { | |
531 | /* | |
532 | * there should be a bug that we can't find the entry | |
533 | * to orphan inode. | |
534 | */ | |
535 | f2fs_bug_on(sbi, PTR_ERR(inode) == -ENOENT); | |
536 | return PTR_ERR(inode); | |
537 | } | |
538 | ||
127e670a JK |
539 | clear_nlink(inode); |
540 | ||
541 | /* truncate all the data during iput */ | |
542 | iput(inode); | |
8c14bfad | 543 | return 0; |
127e670a JK |
544 | } |
545 | ||
8c14bfad | 546 | int recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a | 547 | { |
3c642985 | 548 | block_t start_blk, orphan_blocks, i, j; |
8c14bfad | 549 | int err; |
127e670a | 550 | |
25ca923b | 551 | if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG)) |
8c14bfad | 552 | return 0; |
127e670a | 553 | |
55141486 | 554 | start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi); |
3c642985 | 555 | orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi); |
127e670a | 556 | |
26879fb1 | 557 | ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP, true); |
662befda | 558 | |
3c642985 | 559 | for (i = 0; i < orphan_blocks; i++) { |
127e670a JK |
560 | struct page *page = get_meta_page(sbi, start_blk + i); |
561 | struct f2fs_orphan_block *orphan_blk; | |
562 | ||
563 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); | |
564 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
565 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
8c14bfad CY |
566 | err = recover_orphan_inode(sbi, ino); |
567 | if (err) { | |
568 | f2fs_put_page(page, 1); | |
8c14bfad CY |
569 | return err; |
570 | } | |
127e670a JK |
571 | } |
572 | f2fs_put_page(page, 1); | |
573 | } | |
574 | /* clear Orphan Flag */ | |
25ca923b | 575 | clear_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG); |
8c14bfad | 576 | return 0; |
127e670a JK |
577 | } |
578 | ||
579 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
580 | { | |
502c6e0b | 581 | struct list_head *head; |
127e670a | 582 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 583 | unsigned int nentries = 0; |
bd936f84 | 584 | unsigned short index = 1; |
8c402946 | 585 | unsigned short orphan_blocks; |
4531929e | 586 | struct page *page = NULL; |
6451e041 | 587 | struct ino_entry *orphan = NULL; |
67298804 | 588 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 589 | |
67298804 | 590 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 591 | |
d6c67a4f JK |
592 | /* |
593 | * we don't need to do spin_lock(&im->ino_lock) here, since all the | |
594 | * orphan inode operations are covered under f2fs_lock_op(). | |
595 | * And, spin_lock should be avoided due to page operations below. | |
596 | */ | |
67298804 | 597 | head = &im->ino_list; |
127e670a JK |
598 | |
599 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
600 | list_for_each_entry(orphan, head, list) { |
601 | if (!page) { | |
bd936f84 | 602 | page = grab_meta_page(sbi, start_blk++); |
502c6e0b GZ |
603 | orphan_blk = |
604 | (struct f2fs_orphan_block *)page_address(page); | |
605 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
606 | } | |
127e670a | 607 | |
36795567 | 608 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 609 | |
36795567 | 610 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
611 | /* |
612 | * an orphan block is full of 1020 entries, | |
613 | * then we need to flush current orphan blocks | |
614 | * and bring another one in memory | |
615 | */ | |
616 | orphan_blk->blk_addr = cpu_to_le16(index); | |
617 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
618 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
619 | set_page_dirty(page); | |
620 | f2fs_put_page(page, 1); | |
621 | index++; | |
127e670a JK |
622 | nentries = 0; |
623 | page = NULL; | |
624 | } | |
502c6e0b | 625 | } |
127e670a | 626 | |
502c6e0b GZ |
627 | if (page) { |
628 | orphan_blk->blk_addr = cpu_to_le16(index); | |
629 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
630 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
631 | set_page_dirty(page); | |
632 | f2fs_put_page(page, 1); | |
127e670a | 633 | } |
127e670a JK |
634 | } |
635 | ||
636 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, | |
637 | block_t cp_addr, unsigned long long *version) | |
638 | { | |
639 | struct page *cp_page_1, *cp_page_2 = NULL; | |
640 | unsigned long blk_size = sbi->blocksize; | |
641 | struct f2fs_checkpoint *cp_block; | |
642 | unsigned long long cur_version = 0, pre_version = 0; | |
127e670a | 643 | size_t crc_offset; |
7e586fa0 | 644 | __u32 crc = 0; |
127e670a JK |
645 | |
646 | /* Read the 1st cp block in this CP pack */ | |
647 | cp_page_1 = get_meta_page(sbi, cp_addr); | |
648 | ||
649 | /* get the version number */ | |
650 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_1); | |
651 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
652 | if (crc_offset >= blk_size) | |
653 | goto invalid_cp1; | |
654 | ||
29e7043f | 655 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
43b6573b | 656 | if (!f2fs_crc_valid(sbi, crc, cp_block, crc_offset)) |
127e670a JK |
657 | goto invalid_cp1; |
658 | ||
d71b5564 | 659 | pre_version = cur_cp_version(cp_block); |
127e670a JK |
660 | |
661 | /* Read the 2nd cp block in this CP pack */ | |
25ca923b | 662 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
127e670a JK |
663 | cp_page_2 = get_meta_page(sbi, cp_addr); |
664 | ||
665 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_2); | |
666 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
667 | if (crc_offset >= blk_size) | |
668 | goto invalid_cp2; | |
669 | ||
29e7043f | 670 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
43b6573b | 671 | if (!f2fs_crc_valid(sbi, crc, cp_block, crc_offset)) |
127e670a JK |
672 | goto invalid_cp2; |
673 | ||
d71b5564 | 674 | cur_version = cur_cp_version(cp_block); |
127e670a JK |
675 | |
676 | if (cur_version == pre_version) { | |
677 | *version = cur_version; | |
678 | f2fs_put_page(cp_page_2, 1); | |
679 | return cp_page_1; | |
680 | } | |
681 | invalid_cp2: | |
682 | f2fs_put_page(cp_page_2, 1); | |
683 | invalid_cp1: | |
684 | f2fs_put_page(cp_page_1, 1); | |
685 | return NULL; | |
686 | } | |
687 | ||
688 | int get_valid_checkpoint(struct f2fs_sb_info *sbi) | |
689 | { | |
690 | struct f2fs_checkpoint *cp_block; | |
691 | struct f2fs_super_block *fsb = sbi->raw_super; | |
692 | struct page *cp1, *cp2, *cur_page; | |
693 | unsigned long blk_size = sbi->blocksize; | |
694 | unsigned long long cp1_version = 0, cp2_version = 0; | |
695 | unsigned long long cp_start_blk_no; | |
55141486 | 696 | unsigned int cp_blks = 1 + __cp_payload(sbi); |
1dbe4152 CL |
697 | block_t cp_blk_no; |
698 | int i; | |
127e670a | 699 | |
1dbe4152 | 700 | sbi->ckpt = kzalloc(cp_blks * blk_size, GFP_KERNEL); |
127e670a JK |
701 | if (!sbi->ckpt) |
702 | return -ENOMEM; | |
703 | /* | |
704 | * Finding out valid cp block involves read both | |
705 | * sets( cp pack1 and cp pack 2) | |
706 | */ | |
707 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
708 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
709 | ||
710 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
711 | cp_start_blk_no += ((unsigned long long)1) << |
712 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
713 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
714 | ||
715 | if (cp1 && cp2) { | |
716 | if (ver_after(cp2_version, cp1_version)) | |
717 | cur_page = cp2; | |
718 | else | |
719 | cur_page = cp1; | |
720 | } else if (cp1) { | |
721 | cur_page = cp1; | |
722 | } else if (cp2) { | |
723 | cur_page = cp2; | |
724 | } else { | |
725 | goto fail_no_cp; | |
726 | } | |
727 | ||
728 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
729 | memcpy(sbi->ckpt, cp_block, blk_size); | |
730 | ||
984ec63c SL |
731 | /* Sanity checking of checkpoint */ |
732 | if (sanity_check_ckpt(sbi)) | |
733 | goto fail_no_cp; | |
734 | ||
1dbe4152 CL |
735 | if (cp_blks <= 1) |
736 | goto done; | |
737 | ||
738 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
739 | if (cur_page == cp2) | |
740 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
741 | ||
742 | for (i = 1; i < cp_blks; i++) { | |
743 | void *sit_bitmap_ptr; | |
744 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
745 | ||
746 | cur_page = get_meta_page(sbi, cp_blk_no + i); | |
747 | sit_bitmap_ptr = page_address(cur_page); | |
748 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
749 | f2fs_put_page(cur_page, 1); | |
750 | } | |
751 | done: | |
127e670a JK |
752 | f2fs_put_page(cp1, 1); |
753 | f2fs_put_page(cp2, 1); | |
754 | return 0; | |
755 | ||
756 | fail_no_cp: | |
757 | kfree(sbi->ckpt); | |
758 | return -EINVAL; | |
759 | } | |
760 | ||
c227f912 | 761 | static void __add_dirty_inode(struct inode *inode, enum inode_type type) |
127e670a | 762 | { |
4081363f | 763 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
2710fd7e | 764 | struct f2fs_inode_info *fi = F2FS_I(inode); |
c227f912 | 765 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
127e670a | 766 | |
c227f912 | 767 | if (is_inode_flag_set(fi, flag)) |
2710fd7e | 768 | return; |
2d7b822a | 769 | |
c227f912 CY |
770 | set_inode_flag(fi, flag); |
771 | list_add_tail(&fi->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 | { |
6ad7609a | 777 | struct f2fs_inode_info *fi = F2FS_I(inode); |
c227f912 | 778 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
6ad7609a CY |
779 | |
780 | if (get_dirty_pages(inode) || | |
c227f912 | 781 | !is_inode_flag_set(F2FS_I(inode), flag)) |
6ad7609a CY |
782 | return; |
783 | ||
784 | list_del_init(&fi->dirty_list); | |
c227f912 | 785 | clear_inode_flag(fi, flag); |
33fbd510 | 786 | stat_dec_dirty_inode(F2FS_I_SB(inode), type); |
6ad7609a CY |
787 | } |
788 | ||
a7ffdbe2 | 789 | void update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 790 | { |
4081363f | 791 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 792 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
5deb8267 | 793 | |
5ac9f36f CY |
794 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
795 | !S_ISLNK(inode->i_mode)) | |
127e670a | 796 | return; |
7bd59381 | 797 | |
10aa97c3 JK |
798 | if (type != FILE_INODE || test_opt(sbi, DATA_FLUSH)) { |
799 | spin_lock(&sbi->inode_lock[type]); | |
800 | __add_dirty_inode(inode, type); | |
801 | spin_unlock(&sbi->inode_lock[type]); | |
802 | } | |
cf0ee0f0 | 803 | |
b951a4ec | 804 | inode_inc_dirty_pages(inode); |
a7ffdbe2 | 805 | SetPagePrivate(page); |
9e4ded3f | 806 | f2fs_trace_pid(page); |
5deb8267 JK |
807 | } |
808 | ||
c227f912 | 809 | void remove_dirty_inode(struct inode *inode) |
127e670a | 810 | { |
4081363f | 811 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 812 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
127e670a | 813 | |
c227f912 CY |
814 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
815 | !S_ISLNK(inode->i_mode)) | |
127e670a JK |
816 | return; |
817 | ||
10aa97c3 JK |
818 | if (type == FILE_INODE && !test_opt(sbi, DATA_FLUSH)) |
819 | return; | |
820 | ||
c227f912 CY |
821 | spin_lock(&sbi->inode_lock[type]); |
822 | __remove_dirty_inode(inode, type); | |
823 | spin_unlock(&sbi->inode_lock[type]); | |
74d0b917 JK |
824 | } |
825 | ||
6d5a1495 | 826 | int sync_dirty_inodes(struct f2fs_sb_info *sbi, enum inode_type type) |
127e670a | 827 | { |
ce3b7d80 | 828 | struct list_head *head; |
127e670a | 829 | struct inode *inode; |
2710fd7e | 830 | struct f2fs_inode_info *fi; |
4cf18537 CY |
831 | bool is_dir = (type == DIR_INODE); |
832 | ||
833 | trace_f2fs_sync_dirty_inodes_enter(sbi->sb, is_dir, | |
834 | get_pages(sbi, is_dir ? | |
835 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
127e670a | 836 | retry: |
af41d3ee | 837 | if (unlikely(f2fs_cp_error(sbi))) |
6d5a1495 | 838 | return -EIO; |
af41d3ee | 839 | |
c227f912 | 840 | spin_lock(&sbi->inode_lock[type]); |
ce3b7d80 | 841 | |
c227f912 | 842 | head = &sbi->inode_list[type]; |
127e670a | 843 | if (list_empty(head)) { |
c227f912 | 844 | spin_unlock(&sbi->inode_lock[type]); |
4cf18537 CY |
845 | trace_f2fs_sync_dirty_inodes_exit(sbi->sb, is_dir, |
846 | get_pages(sbi, is_dir ? | |
847 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
6d5a1495 | 848 | return 0; |
127e670a | 849 | } |
2710fd7e CY |
850 | fi = list_entry(head->next, struct f2fs_inode_info, dirty_list); |
851 | inode = igrab(&fi->vfs_inode); | |
c227f912 | 852 | spin_unlock(&sbi->inode_lock[type]); |
127e670a | 853 | if (inode) { |
87d6f890 | 854 | filemap_fdatawrite(inode->i_mapping); |
127e670a JK |
855 | iput(inode); |
856 | } else { | |
857 | /* | |
858 | * We should submit bio, since it exists several | |
859 | * wribacking dentry pages in the freeing inode. | |
860 | */ | |
458e6197 | 861 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
7ecebe5e | 862 | cond_resched(); |
127e670a JK |
863 | } |
864 | goto retry; | |
865 | } | |
866 | ||
0a8165d7 | 867 | /* |
127e670a JK |
868 | * Freeze all the FS-operations for checkpoint. |
869 | */ | |
cf779cab | 870 | static int block_operations(struct f2fs_sb_info *sbi) |
127e670a | 871 | { |
127e670a JK |
872 | struct writeback_control wbc = { |
873 | .sync_mode = WB_SYNC_ALL, | |
874 | .nr_to_write = LONG_MAX, | |
875 | .for_reclaim = 0, | |
876 | }; | |
c718379b | 877 | struct blk_plug plug; |
cf779cab | 878 | int err = 0; |
c718379b JK |
879 | |
880 | blk_start_plug(&plug); | |
881 | ||
39936837 | 882 | retry_flush_dents: |
e479556b | 883 | f2fs_lock_all(sbi); |
127e670a | 884 | /* write all the dirty dentry pages */ |
127e670a | 885 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 886 | f2fs_unlock_all(sbi); |
6d5a1495 CY |
887 | err = sync_dirty_inodes(sbi, DIR_INODE); |
888 | if (err) | |
cf779cab | 889 | goto out; |
39936837 | 890 | goto retry_flush_dents; |
127e670a JK |
891 | } |
892 | ||
127e670a | 893 | /* |
e1c42045 | 894 | * POR: we should ensure that there are no dirty node pages |
127e670a JK |
895 | * until finishing nat/sit flush. |
896 | */ | |
39936837 | 897 | retry_flush_nodes: |
b3582c68 | 898 | down_write(&sbi->node_write); |
127e670a JK |
899 | |
900 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
b3582c68 | 901 | up_write(&sbi->node_write); |
52681375 | 902 | err = sync_node_pages(sbi, &wbc); |
6d5a1495 | 903 | if (err) { |
cf779cab | 904 | f2fs_unlock_all(sbi); |
cf779cab JK |
905 | goto out; |
906 | } | |
39936837 | 907 | goto retry_flush_nodes; |
127e670a | 908 | } |
cf779cab | 909 | out: |
c718379b | 910 | blk_finish_plug(&plug); |
cf779cab | 911 | return err; |
127e670a JK |
912 | } |
913 | ||
914 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
915 | { | |
b3582c68 | 916 | up_write(&sbi->node_write); |
e479556b | 917 | f2fs_unlock_all(sbi); |
127e670a JK |
918 | } |
919 | ||
fb51b5ef CL |
920 | static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi) |
921 | { | |
922 | DEFINE_WAIT(wait); | |
923 | ||
924 | for (;;) { | |
925 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
926 | ||
f5730184 | 927 | if (!atomic_read(&sbi->nr_wb_bios)) |
fb51b5ef CL |
928 | break; |
929 | ||
0ff21646 | 930 | io_schedule_timeout(5*HZ); |
fb51b5ef CL |
931 | } |
932 | finish_wait(&sbi->cp_wait, &wait); | |
933 | } | |
934 | ||
c34f42e2 | 935 | static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
936 | { |
937 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
cf2271e7 | 938 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); |
77041823 | 939 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
67298804 | 940 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; |
77041823 | 941 | nid_t last_nid = nm_i->next_scan_nid; |
127e670a | 942 | block_t start_blk; |
127e670a | 943 | unsigned int data_sum_blocks, orphan_blocks; |
7e586fa0 | 944 | __u32 crc32 = 0; |
127e670a | 945 | int i; |
55141486 | 946 | int cp_payload_blks = __cp_payload(sbi); |
e90c2d28 CY |
947 | block_t discard_blk = NEXT_FREE_BLKADDR(sbi, curseg); |
948 | bool invalidate = false; | |
8f1dbbbb SL |
949 | struct super_block *sb = sbi->sb; |
950 | struct curseg_info *seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); | |
951 | u64 kbytes_written; | |
127e670a | 952 | |
1e87a78d JK |
953 | /* |
954 | * This avoids to conduct wrong roll-forward operations and uses | |
955 | * metapages, so should be called prior to sync_meta_pages below. | |
956 | */ | |
e90c2d28 CY |
957 | if (discard_next_dnode(sbi, discard_blk)) |
958 | invalidate = true; | |
127e670a JK |
959 | |
960 | /* Flush all the NAT/SIT pages */ | |
cf779cab | 961 | while (get_pages(sbi, F2FS_DIRTY_META)) { |
127e670a | 962 | sync_meta_pages(sbi, META, LONG_MAX); |
cf779cab | 963 | if (unlikely(f2fs_cp_error(sbi))) |
c34f42e2 | 964 | return -EIO; |
cf779cab | 965 | } |
127e670a JK |
966 | |
967 | next_free_nid(sbi, &last_nid); | |
968 | ||
969 | /* | |
970 | * modify checkpoint | |
971 | * version number is already updated | |
972 | */ | |
973 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi)); | |
974 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
975 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); | |
b5b82205 | 976 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
977 | ckpt->cur_node_segno[i] = |
978 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
979 | ckpt->cur_node_blkoff[i] = | |
980 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
981 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
982 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
983 | } | |
b5b82205 | 984 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
985 | ckpt->cur_data_segno[i] = |
986 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
987 | ckpt->cur_data_blkoff[i] = | |
988 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
989 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
990 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
991 | } | |
992 | ||
993 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
994 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
995 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
996 | ||
997 | /* 2 cp + n data seg summary + orphan inode blocks */ | |
3fa06d7b | 998 | data_sum_blocks = npages_for_summary_flush(sbi, false); |
b5b82205 | 999 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
25ca923b | 1000 | set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 1001 | else |
25ca923b | 1002 | clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 1003 | |
67298804 | 1004 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
1005 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
1006 | orphan_blocks); | |
127e670a | 1007 | |
119ee914 | 1008 | if (__remain_node_summaries(cpc->reason)) |
b5b82205 | 1009 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
1010 | cp_payload_blks + data_sum_blocks + |
1011 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
119ee914 | 1012 | else |
b5b82205 | 1013 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS + |
1dbe4152 CL |
1014 | cp_payload_blks + data_sum_blocks + |
1015 | orphan_blocks); | |
119ee914 JK |
1016 | |
1017 | if (cpc->reason == CP_UMOUNT) | |
1018 | set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
1019 | else | |
1020 | clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
1021 | ||
1022 | if (cpc->reason == CP_FASTBOOT) | |
1023 | set_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
1024 | else | |
1025 | clear_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
127e670a | 1026 | |
67298804 | 1027 | if (orphan_num) |
25ca923b | 1028 | set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1029 | else |
25ca923b | 1030 | clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1031 | |
caf0047e | 1032 | if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) |
2ae4c673 JK |
1033 | set_ckpt_flags(ckpt, CP_FSCK_FLAG); |
1034 | ||
127e670a JK |
1035 | /* update SIT/NAT bitmap */ |
1036 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
1037 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
1038 | ||
43b6573b | 1039 | crc32 = f2fs_crc32(sbi, ckpt, le32_to_cpu(ckpt->checksum_offset)); |
7e586fa0 JK |
1040 | *((__le32 *)((unsigned char *)ckpt + |
1041 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
1042 | = cpu_to_le32(crc32); |
1043 | ||
1044 | start_blk = __start_cp_addr(sbi); | |
1045 | ||
a7230d16 JK |
1046 | /* need to wait for end_io results */ |
1047 | wait_on_all_pages_writeback(sbi); | |
1048 | if (unlikely(f2fs_cp_error(sbi))) | |
c34f42e2 | 1049 | return -EIO; |
a7230d16 | 1050 | |
127e670a | 1051 | /* write out checkpoint buffer at block 0 */ |
381722d2 CY |
1052 | update_meta_page(sbi, ckpt, start_blk++); |
1053 | ||
1054 | for (i = 1; i < 1 + cp_payload_blks; i++) | |
1055 | update_meta_page(sbi, (char *)ckpt + i * F2FS_BLKSIZE, | |
1056 | start_blk++); | |
1dbe4152 | 1057 | |
67298804 | 1058 | if (orphan_num) { |
127e670a JK |
1059 | write_orphan_inodes(sbi, start_blk); |
1060 | start_blk += orphan_blocks; | |
1061 | } | |
1062 | ||
1063 | write_data_summaries(sbi, start_blk); | |
1064 | start_blk += data_sum_blocks; | |
8f1dbbbb SL |
1065 | |
1066 | /* Record write statistics in the hot node summary */ | |
1067 | kbytes_written = sbi->kbytes_written; | |
1068 | if (sb->s_bdev->bd_part) | |
1069 | kbytes_written += BD_PART_WRITTEN(sbi); | |
1070 | ||
b7ad7512 | 1071 | seg_i->journal->info.kbytes_written = cpu_to_le64(kbytes_written); |
8f1dbbbb | 1072 | |
119ee914 | 1073 | if (__remain_node_summaries(cpc->reason)) { |
127e670a JK |
1074 | write_node_summaries(sbi, start_blk); |
1075 | start_blk += NR_CURSEG_NODE_TYPE; | |
1076 | } | |
1077 | ||
1078 | /* writeout checkpoint block */ | |
381722d2 | 1079 | update_meta_page(sbi, ckpt, start_blk); |
127e670a JK |
1080 | |
1081 | /* wait for previous submitted node/meta pages writeback */ | |
fb51b5ef | 1082 | wait_on_all_pages_writeback(sbi); |
127e670a | 1083 | |
cf779cab | 1084 | if (unlikely(f2fs_cp_error(sbi))) |
c34f42e2 | 1085 | return -EIO; |
cf779cab | 1086 | |
80dd9c0e CY |
1087 | filemap_fdatawait_range(NODE_MAPPING(sbi), 0, LLONG_MAX); |
1088 | filemap_fdatawait_range(META_MAPPING(sbi), 0, LLONG_MAX); | |
127e670a JK |
1089 | |
1090 | /* update user_block_counts */ | |
1091 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
41382ec4 | 1092 | percpu_counter_set(&sbi->alloc_valid_block_count, 0); |
127e670a JK |
1093 | |
1094 | /* Here, we only have one bio having CP pack */ | |
577e3495 | 1095 | sync_meta_pages(sbi, META_FLUSH, LONG_MAX); |
127e670a | 1096 | |
6a8f8ca5 JK |
1097 | /* wait for previous submitted meta pages writeback */ |
1098 | wait_on_all_pages_writeback(sbi); | |
1099 | ||
e90c2d28 CY |
1100 | /* |
1101 | * invalidate meta page which is used temporarily for zeroing out | |
1102 | * block at the end of warm node chain. | |
1103 | */ | |
1104 | if (invalidate) | |
1105 | invalidate_mapping_pages(META_MAPPING(sbi), discard_blk, | |
1106 | discard_blk); | |
1107 | ||
74ef9241 | 1108 | release_ino_entry(sbi, false); |
cf779cab JK |
1109 | |
1110 | if (unlikely(f2fs_cp_error(sbi))) | |
c34f42e2 | 1111 | return -EIO; |
cf779cab | 1112 | |
836b5a63 | 1113 | clear_prefree_segments(sbi, cpc); |
caf0047e | 1114 | clear_sbi_flag(sbi, SBI_IS_DIRTY); |
c34f42e2 CY |
1115 | |
1116 | return 0; | |
127e670a JK |
1117 | } |
1118 | ||
0a8165d7 | 1119 | /* |
e1c42045 | 1120 | * We guarantee that this checkpoint procedure will not fail. |
127e670a | 1121 | */ |
c34f42e2 | 1122 | int write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1123 | { |
1124 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1125 | unsigned long long ckpt_ver; | |
c34f42e2 | 1126 | int err = 0; |
127e670a | 1127 | |
43727527 | 1128 | mutex_lock(&sbi->cp_mutex); |
8501017e | 1129 | |
caf0047e | 1130 | if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) && |
a66cdd98 JK |
1131 | (cpc->reason == CP_FASTBOOT || cpc->reason == CP_SYNC || |
1132 | (cpc->reason == CP_DISCARD && !sbi->discard_blks))) | |
8501017e | 1133 | goto out; |
c34f42e2 CY |
1134 | if (unlikely(f2fs_cp_error(sbi))) { |
1135 | err = -EIO; | |
cf779cab | 1136 | goto out; |
c34f42e2 CY |
1137 | } |
1138 | if (f2fs_readonly(sbi->sb)) { | |
1139 | err = -EROFS; | |
11504a8e | 1140 | goto out; |
c34f42e2 | 1141 | } |
2bda542d WL |
1142 | |
1143 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); | |
1144 | ||
c34f42e2 CY |
1145 | err = block_operations(sbi); |
1146 | if (err) | |
cf779cab | 1147 | goto out; |
127e670a | 1148 | |
75ab4cb8 | 1149 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1150 | |
406657dd | 1151 | f2fs_flush_merged_bios(sbi); |
127e670a JK |
1152 | |
1153 | /* | |
1154 | * update checkpoint pack index | |
1155 | * Increase the version number so that | |
1156 | * SIT entries and seg summaries are written at correct place | |
1157 | */ | |
d71b5564 | 1158 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1159 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1160 | ||
1161 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
1162 | flush_nat_entries(sbi); | |
4b2fecc8 | 1163 | flush_sit_entries(sbi, cpc); |
127e670a | 1164 | |
127e670a | 1165 | /* unlock all the fs_lock[] in do_checkpoint() */ |
c34f42e2 | 1166 | err = do_checkpoint(sbi, cpc); |
127e670a JK |
1167 | |
1168 | unblock_operations(sbi); | |
942e0be6 | 1169 | stat_inc_cp_count(sbi->stat_info); |
10027551 JK |
1170 | |
1171 | if (cpc->reason == CP_RECOVERY) | |
1172 | f2fs_msg(sbi->sb, KERN_NOTICE, | |
1173 | "checkpoint: version = %llx", ckpt_ver); | |
60b99b48 JK |
1174 | |
1175 | /* do checkpoint periodically */ | |
6beceb54 | 1176 | f2fs_update_time(sbi, CP_TIME); |
55d1cdb2 | 1177 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
8501017e JK |
1178 | out: |
1179 | mutex_unlock(&sbi->cp_mutex); | |
c34f42e2 | 1180 | return err; |
127e670a JK |
1181 | } |
1182 | ||
6451e041 | 1183 | void init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1184 | { |
6451e041 JK |
1185 | int i; |
1186 | ||
1187 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1188 | struct inode_management *im = &sbi->im[i]; |
1189 | ||
1190 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1191 | spin_lock_init(&im->ino_lock); | |
1192 | INIT_LIST_HEAD(&im->ino_list); | |
1193 | im->ino_num = 0; | |
6451e041 JK |
1194 | } |
1195 | ||
b5b82205 | 1196 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
14b42817 WL |
1197 | NR_CURSEG_TYPE - __cp_payload(sbi)) * |
1198 | F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
1199 | } |
1200 | ||
6e6093a8 | 1201 | int __init create_checkpoint_caches(void) |
127e670a | 1202 | { |
6451e041 JK |
1203 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1204 | sizeof(struct ino_entry)); | |
1205 | if (!ino_entry_slab) | |
127e670a | 1206 | return -ENOMEM; |
06292073 CY |
1207 | inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry", |
1208 | sizeof(struct inode_entry)); | |
6bacf52f | 1209 | if (!inode_entry_slab) { |
6451e041 | 1210 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1211 | return -ENOMEM; |
1212 | } | |
1213 | return 0; | |
1214 | } | |
1215 | ||
1216 | void destroy_checkpoint_caches(void) | |
1217 | { | |
6451e041 | 1218 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1219 | kmem_cache_destroy(inode_entry_slab); |
1220 | } |