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" | |
2af4bd6c | 23 | #include <trace/events/f2fs.h> |
127e670a JK |
24 | |
25 | static struct kmem_cache *orphan_entry_slab; | |
26 | static struct kmem_cache *inode_entry_slab; | |
27 | ||
0a8165d7 | 28 | /* |
127e670a JK |
29 | * We guarantee no failure on the returned page. |
30 | */ | |
31 | struct page *grab_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
32 | { | |
9df27d98 | 33 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
34 | struct page *page = NULL; |
35 | repeat: | |
36 | page = grab_cache_page(mapping, index); | |
37 | if (!page) { | |
38 | cond_resched(); | |
39 | goto repeat; | |
40 | } | |
41 | ||
42 | /* We wait writeback only inside grab_meta_page() */ | |
43 | wait_on_page_writeback(page); | |
44 | SetPageUptodate(page); | |
45 | return page; | |
46 | } | |
47 | ||
0a8165d7 | 48 | /* |
127e670a JK |
49 | * We guarantee no failure on the returned page. |
50 | */ | |
51 | struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
52 | { | |
9df27d98 | 53 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
54 | struct page *page; |
55 | repeat: | |
56 | page = grab_cache_page(mapping, index); | |
57 | if (!page) { | |
58 | cond_resched(); | |
59 | goto repeat; | |
60 | } | |
393ff91f JK |
61 | if (PageUptodate(page)) |
62 | goto out; | |
63 | ||
93dfe2ac JK |
64 | if (f2fs_submit_page_bio(sbi, page, index, |
65 | READ_SYNC | REQ_META | REQ_PRIO)) | |
127e670a | 66 | goto repeat; |
127e670a | 67 | |
393ff91f | 68 | lock_page(page); |
6bacf52f | 69 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
70 | f2fs_put_page(page, 1); |
71 | goto repeat; | |
72 | } | |
393ff91f JK |
73 | out: |
74 | mark_page_accessed(page); | |
127e670a JK |
75 | return page; |
76 | } | |
77 | ||
662befda CY |
78 | inline int get_max_meta_blks(struct f2fs_sb_info *sbi, int type) |
79 | { | |
80 | switch (type) { | |
81 | case META_NAT: | |
82 | return NM_I(sbi)->max_nid / NAT_ENTRY_PER_BLOCK; | |
83 | case META_SIT: | |
84 | return SIT_BLK_CNT(sbi); | |
81c1a0f1 | 85 | case META_SSA: |
662befda CY |
86 | case META_CP: |
87 | return 0; | |
88 | default: | |
89 | BUG(); | |
90 | } | |
91 | } | |
92 | ||
93 | /* | |
81c1a0f1 | 94 | * Readahead CP/NAT/SIT/SSA pages |
662befda CY |
95 | */ |
96 | int ra_meta_pages(struct f2fs_sb_info *sbi, int start, int nrpages, int type) | |
97 | { | |
98 | block_t prev_blk_addr = 0; | |
99 | struct page *page; | |
100 | int blkno = start; | |
101 | int max_blks = get_max_meta_blks(sbi, type); | |
102 | ||
103 | struct f2fs_io_info fio = { | |
104 | .type = META, | |
105 | .rw = READ_SYNC | REQ_META | REQ_PRIO | |
106 | }; | |
107 | ||
108 | for (; nrpages-- > 0; blkno++) { | |
109 | block_t blk_addr; | |
110 | ||
111 | switch (type) { | |
112 | case META_NAT: | |
113 | /* get nat block addr */ | |
114 | if (unlikely(blkno >= max_blks)) | |
115 | blkno = 0; | |
116 | blk_addr = current_nat_addr(sbi, | |
117 | blkno * NAT_ENTRY_PER_BLOCK); | |
118 | break; | |
119 | case META_SIT: | |
120 | /* get sit block addr */ | |
121 | if (unlikely(blkno >= max_blks)) | |
122 | goto out; | |
123 | blk_addr = current_sit_addr(sbi, | |
124 | blkno * SIT_ENTRY_PER_BLOCK); | |
125 | if (blkno != start && prev_blk_addr + 1 != blk_addr) | |
126 | goto out; | |
127 | prev_blk_addr = blk_addr; | |
128 | break; | |
81c1a0f1 | 129 | case META_SSA: |
662befda | 130 | case META_CP: |
81c1a0f1 | 131 | /* get ssa/cp block addr */ |
662befda CY |
132 | blk_addr = blkno; |
133 | break; | |
134 | default: | |
135 | BUG(); | |
136 | } | |
137 | ||
138 | page = grab_cache_page(META_MAPPING(sbi), blk_addr); | |
139 | if (!page) | |
140 | continue; | |
141 | if (PageUptodate(page)) { | |
142 | mark_page_accessed(page); | |
143 | f2fs_put_page(page, 1); | |
144 | continue; | |
145 | } | |
146 | ||
147 | f2fs_submit_page_mbio(sbi, page, blk_addr, &fio); | |
148 | mark_page_accessed(page); | |
149 | f2fs_put_page(page, 0); | |
150 | } | |
151 | out: | |
152 | f2fs_submit_merged_bio(sbi, META, READ); | |
153 | return blkno - start; | |
154 | } | |
155 | ||
127e670a JK |
156 | static int f2fs_write_meta_page(struct page *page, |
157 | struct writeback_control *wbc) | |
158 | { | |
159 | struct inode *inode = page->mapping->host; | |
160 | struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); | |
127e670a | 161 | |
203681f6 | 162 | if (unlikely(sbi->por_doing)) |
cfb271d4 | 163 | goto redirty_out; |
cfb271d4 CY |
164 | if (wbc->for_reclaim) |
165 | goto redirty_out; | |
127e670a | 166 | |
203681f6 JK |
167 | /* Should not write any meta pages, if any IO error was occurred */ |
168 | if (unlikely(is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ERROR_FLAG))) | |
169 | goto no_write; | |
127e670a | 170 | |
203681f6 | 171 | wait_on_page_writeback(page); |
577e3495 | 172 | write_meta_page(sbi, page); |
203681f6 | 173 | no_write: |
577e3495 JK |
174 | dec_page_count(sbi, F2FS_DIRTY_META); |
175 | unlock_page(page); | |
176 | return 0; | |
cfb271d4 CY |
177 | |
178 | redirty_out: | |
179 | dec_page_count(sbi, F2FS_DIRTY_META); | |
180 | wbc->pages_skipped++; | |
9cf3c389 | 181 | account_page_redirty(page); |
cfb271d4 CY |
182 | set_page_dirty(page); |
183 | return AOP_WRITEPAGE_ACTIVATE; | |
127e670a JK |
184 | } |
185 | ||
186 | static int f2fs_write_meta_pages(struct address_space *mapping, | |
187 | struct writeback_control *wbc) | |
188 | { | |
189 | struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb); | |
50c8cdb3 | 190 | long diff, written; |
127e670a | 191 | |
5459aa97 | 192 | /* collect a number of dirty meta pages and write together */ |
50c8cdb3 JK |
193 | if (wbc->for_kupdate || |
194 | get_pages(sbi, F2FS_DIRTY_META) < nr_pages_to_skip(sbi, META)) | |
d3baf95d | 195 | goto skip_write; |
127e670a JK |
196 | |
197 | /* if mounting is failed, skip writing node pages */ | |
198 | mutex_lock(&sbi->cp_mutex); | |
50c8cdb3 JK |
199 | diff = nr_pages_to_write(sbi, META, wbc); |
200 | written = sync_meta_pages(sbi, META, wbc->nr_to_write); | |
127e670a | 201 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 202 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 203 | return 0; |
d3baf95d JK |
204 | |
205 | skip_write: | |
206 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
207 | return 0; | |
127e670a JK |
208 | } |
209 | ||
210 | long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, | |
211 | long nr_to_write) | |
212 | { | |
9df27d98 | 213 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
214 | pgoff_t index = 0, end = LONG_MAX; |
215 | struct pagevec pvec; | |
216 | long nwritten = 0; | |
217 | struct writeback_control wbc = { | |
218 | .for_reclaim = 0, | |
219 | }; | |
220 | ||
221 | pagevec_init(&pvec, 0); | |
222 | ||
223 | while (index <= end) { | |
224 | int i, nr_pages; | |
225 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, | |
226 | PAGECACHE_TAG_DIRTY, | |
227 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
cfb271d4 | 228 | if (unlikely(nr_pages == 0)) |
127e670a JK |
229 | break; |
230 | ||
231 | for (i = 0; i < nr_pages; i++) { | |
232 | struct page *page = pvec.pages[i]; | |
203681f6 | 233 | |
127e670a | 234 | lock_page(page); |
203681f6 JK |
235 | |
236 | if (unlikely(page->mapping != mapping)) { | |
237 | continue_unlock: | |
238 | unlock_page(page); | |
239 | continue; | |
240 | } | |
241 | if (!PageDirty(page)) { | |
242 | /* someone wrote it for us */ | |
243 | goto continue_unlock; | |
244 | } | |
245 | ||
246 | if (!clear_page_dirty_for_io(page)) | |
247 | goto continue_unlock; | |
248 | ||
577e3495 JK |
249 | if (f2fs_write_meta_page(page, &wbc)) { |
250 | unlock_page(page); | |
251 | break; | |
252 | } | |
cfb271d4 CY |
253 | nwritten++; |
254 | if (unlikely(nwritten >= nr_to_write)) | |
127e670a JK |
255 | break; |
256 | } | |
257 | pagevec_release(&pvec); | |
258 | cond_resched(); | |
259 | } | |
260 | ||
261 | if (nwritten) | |
458e6197 | 262 | f2fs_submit_merged_bio(sbi, type, WRITE); |
127e670a JK |
263 | |
264 | return nwritten; | |
265 | } | |
266 | ||
267 | static int f2fs_set_meta_page_dirty(struct page *page) | |
268 | { | |
269 | struct address_space *mapping = page->mapping; | |
270 | struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb); | |
271 | ||
26c6b887 JK |
272 | trace_f2fs_set_page_dirty(page, META); |
273 | ||
127e670a JK |
274 | SetPageUptodate(page); |
275 | if (!PageDirty(page)) { | |
276 | __set_page_dirty_nobuffers(page); | |
277 | inc_page_count(sbi, F2FS_DIRTY_META); | |
127e670a JK |
278 | return 1; |
279 | } | |
280 | return 0; | |
281 | } | |
282 | ||
283 | const struct address_space_operations f2fs_meta_aops = { | |
284 | .writepage = f2fs_write_meta_page, | |
285 | .writepages = f2fs_write_meta_pages, | |
286 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
287 | }; | |
288 | ||
cbd56e7d | 289 | int acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 290 | { |
127e670a JK |
291 | int err = 0; |
292 | ||
17b692f6 | 293 | spin_lock(&sbi->orphan_inode_lock); |
0d47c1ad | 294 | if (unlikely(sbi->n_orphans >= sbi->max_orphans)) |
127e670a | 295 | err = -ENOSPC; |
cbd56e7d JK |
296 | else |
297 | sbi->n_orphans++; | |
17b692f6 | 298 | spin_unlock(&sbi->orphan_inode_lock); |
0d47c1ad | 299 | |
127e670a JK |
300 | return err; |
301 | } | |
302 | ||
cbd56e7d JK |
303 | void release_orphan_inode(struct f2fs_sb_info *sbi) |
304 | { | |
17b692f6 | 305 | spin_lock(&sbi->orphan_inode_lock); |
5d56b671 | 306 | f2fs_bug_on(sbi->n_orphans == 0); |
cbd56e7d | 307 | sbi->n_orphans--; |
17b692f6 | 308 | spin_unlock(&sbi->orphan_inode_lock); |
cbd56e7d JK |
309 | } |
310 | ||
127e670a JK |
311 | void add_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
312 | { | |
313 | struct list_head *head, *this; | |
314 | struct orphan_inode_entry *new = NULL, *orphan = NULL; | |
315 | ||
c1ef3725 GZ |
316 | new = f2fs_kmem_cache_alloc(orphan_entry_slab, GFP_ATOMIC); |
317 | new->ino = ino; | |
318 | ||
17b692f6 | 319 | spin_lock(&sbi->orphan_inode_lock); |
127e670a JK |
320 | head = &sbi->orphan_inode_list; |
321 | list_for_each(this, head) { | |
322 | orphan = list_entry(this, struct orphan_inode_entry, list); | |
c1ef3725 | 323 | if (orphan->ino == ino) { |
17b692f6 | 324 | spin_unlock(&sbi->orphan_inode_lock); |
c1ef3725 GZ |
325 | kmem_cache_free(orphan_entry_slab, new); |
326 | return; | |
327 | } | |
328 | ||
127e670a JK |
329 | if (orphan->ino > ino) |
330 | break; | |
331 | orphan = NULL; | |
332 | } | |
7bd59381 | 333 | |
127e670a | 334 | /* add new_oentry into list which is sorted by inode number */ |
a2617dc6 | 335 | if (orphan) |
336 | list_add(&new->list, this->prev); | |
337 | else | |
127e670a | 338 | list_add_tail(&new->list, head); |
17b692f6 | 339 | spin_unlock(&sbi->orphan_inode_lock); |
127e670a JK |
340 | } |
341 | ||
342 | void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
343 | { | |
60ed9a0f | 344 | struct list_head *head; |
127e670a JK |
345 | struct orphan_inode_entry *orphan; |
346 | ||
17b692f6 | 347 | spin_lock(&sbi->orphan_inode_lock); |
127e670a | 348 | head = &sbi->orphan_inode_list; |
60ed9a0f | 349 | list_for_each_entry(orphan, head, list) { |
127e670a JK |
350 | if (orphan->ino == ino) { |
351 | list_del(&orphan->list); | |
352 | kmem_cache_free(orphan_entry_slab, orphan); | |
5d56b671 | 353 | f2fs_bug_on(sbi->n_orphans == 0); |
127e670a JK |
354 | sbi->n_orphans--; |
355 | break; | |
356 | } | |
357 | } | |
17b692f6 | 358 | spin_unlock(&sbi->orphan_inode_lock); |
127e670a JK |
359 | } |
360 | ||
361 | static void recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
362 | { | |
363 | struct inode *inode = f2fs_iget(sbi->sb, ino); | |
5d56b671 | 364 | f2fs_bug_on(IS_ERR(inode)); |
127e670a JK |
365 | clear_nlink(inode); |
366 | ||
367 | /* truncate all the data during iput */ | |
368 | iput(inode); | |
369 | } | |
370 | ||
8f99a946 | 371 | void recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a JK |
372 | { |
373 | block_t start_blk, orphan_blkaddr, i, j; | |
374 | ||
25ca923b | 375 | if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG)) |
8f99a946 | 376 | return; |
127e670a | 377 | |
aabe5136 | 378 | sbi->por_doing = true; |
127e670a JK |
379 | start_blk = __start_cp_addr(sbi) + 1; |
380 | orphan_blkaddr = __start_sum_addr(sbi) - 1; | |
381 | ||
662befda CY |
382 | ra_meta_pages(sbi, start_blk, orphan_blkaddr, META_CP); |
383 | ||
127e670a JK |
384 | for (i = 0; i < orphan_blkaddr; i++) { |
385 | struct page *page = get_meta_page(sbi, start_blk + i); | |
386 | struct f2fs_orphan_block *orphan_blk; | |
387 | ||
388 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); | |
389 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
390 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
391 | recover_orphan_inode(sbi, ino); | |
392 | } | |
393 | f2fs_put_page(page, 1); | |
394 | } | |
395 | /* clear Orphan Flag */ | |
25ca923b | 396 | clear_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG); |
aabe5136 | 397 | sbi->por_doing = false; |
8f99a946 | 398 | return; |
127e670a JK |
399 | } |
400 | ||
401 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
402 | { | |
502c6e0b | 403 | struct list_head *head; |
127e670a | 404 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 405 | unsigned int nentries = 0; |
4531929e GZ |
406 | unsigned short index; |
407 | unsigned short orphan_blocks = (unsigned short)((sbi->n_orphans + | |
408 | (F2FS_ORPHANS_PER_BLOCK - 1)) / F2FS_ORPHANS_PER_BLOCK); | |
409 | struct page *page = NULL; | |
502c6e0b | 410 | struct orphan_inode_entry *orphan = NULL; |
127e670a | 411 | |
4531929e | 412 | for (index = 0; index < orphan_blocks; index++) |
63f5384c | 413 | grab_meta_page(sbi, start_blk + index); |
127e670a | 414 | |
4531929e | 415 | index = 1; |
17b692f6 | 416 | spin_lock(&sbi->orphan_inode_lock); |
127e670a JK |
417 | head = &sbi->orphan_inode_list; |
418 | ||
419 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
420 | list_for_each_entry(orphan, head, list) { |
421 | if (!page) { | |
63f5384c GZ |
422 | page = find_get_page(META_MAPPING(sbi), start_blk++); |
423 | f2fs_bug_on(!page); | |
502c6e0b GZ |
424 | orphan_blk = |
425 | (struct f2fs_orphan_block *)page_address(page); | |
426 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
63f5384c | 427 | f2fs_put_page(page, 0); |
502c6e0b | 428 | } |
127e670a | 429 | |
36795567 | 430 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 431 | |
36795567 | 432 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
433 | /* |
434 | * an orphan block is full of 1020 entries, | |
435 | * then we need to flush current orphan blocks | |
436 | * and bring another one in memory | |
437 | */ | |
438 | orphan_blk->blk_addr = cpu_to_le16(index); | |
439 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
440 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
441 | set_page_dirty(page); | |
442 | f2fs_put_page(page, 1); | |
443 | index++; | |
127e670a JK |
444 | nentries = 0; |
445 | page = NULL; | |
446 | } | |
502c6e0b | 447 | } |
127e670a | 448 | |
502c6e0b GZ |
449 | if (page) { |
450 | orphan_blk->blk_addr = cpu_to_le16(index); | |
451 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
452 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
453 | set_page_dirty(page); | |
454 | f2fs_put_page(page, 1); | |
127e670a | 455 | } |
502c6e0b | 456 | |
17b692f6 | 457 | spin_unlock(&sbi->orphan_inode_lock); |
127e670a JK |
458 | } |
459 | ||
460 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, | |
461 | block_t cp_addr, unsigned long long *version) | |
462 | { | |
463 | struct page *cp_page_1, *cp_page_2 = NULL; | |
464 | unsigned long blk_size = sbi->blocksize; | |
465 | struct f2fs_checkpoint *cp_block; | |
466 | unsigned long long cur_version = 0, pre_version = 0; | |
127e670a | 467 | size_t crc_offset; |
7e586fa0 | 468 | __u32 crc = 0; |
127e670a JK |
469 | |
470 | /* Read the 1st cp block in this CP pack */ | |
471 | cp_page_1 = get_meta_page(sbi, cp_addr); | |
472 | ||
473 | /* get the version number */ | |
474 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_1); | |
475 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
476 | if (crc_offset >= blk_size) | |
477 | goto invalid_cp1; | |
478 | ||
7e586fa0 | 479 | crc = le32_to_cpu(*((__u32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
480 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
481 | goto invalid_cp1; | |
482 | ||
d71b5564 | 483 | pre_version = cur_cp_version(cp_block); |
127e670a JK |
484 | |
485 | /* Read the 2nd cp block in this CP pack */ | |
25ca923b | 486 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
127e670a JK |
487 | cp_page_2 = get_meta_page(sbi, cp_addr); |
488 | ||
489 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_2); | |
490 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
491 | if (crc_offset >= blk_size) | |
492 | goto invalid_cp2; | |
493 | ||
7e586fa0 | 494 | crc = le32_to_cpu(*((__u32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
495 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
496 | goto invalid_cp2; | |
497 | ||
d71b5564 | 498 | cur_version = cur_cp_version(cp_block); |
127e670a JK |
499 | |
500 | if (cur_version == pre_version) { | |
501 | *version = cur_version; | |
502 | f2fs_put_page(cp_page_2, 1); | |
503 | return cp_page_1; | |
504 | } | |
505 | invalid_cp2: | |
506 | f2fs_put_page(cp_page_2, 1); | |
507 | invalid_cp1: | |
508 | f2fs_put_page(cp_page_1, 1); | |
509 | return NULL; | |
510 | } | |
511 | ||
512 | int get_valid_checkpoint(struct f2fs_sb_info *sbi) | |
513 | { | |
514 | struct f2fs_checkpoint *cp_block; | |
515 | struct f2fs_super_block *fsb = sbi->raw_super; | |
516 | struct page *cp1, *cp2, *cur_page; | |
517 | unsigned long blk_size = sbi->blocksize; | |
518 | unsigned long long cp1_version = 0, cp2_version = 0; | |
519 | unsigned long long cp_start_blk_no; | |
520 | ||
521 | sbi->ckpt = kzalloc(blk_size, GFP_KERNEL); | |
522 | if (!sbi->ckpt) | |
523 | return -ENOMEM; | |
524 | /* | |
525 | * Finding out valid cp block involves read both | |
526 | * sets( cp pack1 and cp pack 2) | |
527 | */ | |
528 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
529 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
530 | ||
531 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
532 | cp_start_blk_no += ((unsigned long long)1) << |
533 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
534 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
535 | ||
536 | if (cp1 && cp2) { | |
537 | if (ver_after(cp2_version, cp1_version)) | |
538 | cur_page = cp2; | |
539 | else | |
540 | cur_page = cp1; | |
541 | } else if (cp1) { | |
542 | cur_page = cp1; | |
543 | } else if (cp2) { | |
544 | cur_page = cp2; | |
545 | } else { | |
546 | goto fail_no_cp; | |
547 | } | |
548 | ||
549 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
550 | memcpy(sbi->ckpt, cp_block, blk_size); | |
551 | ||
552 | f2fs_put_page(cp1, 1); | |
553 | f2fs_put_page(cp2, 1); | |
554 | return 0; | |
555 | ||
556 | fail_no_cp: | |
557 | kfree(sbi->ckpt); | |
558 | return -EINVAL; | |
559 | } | |
560 | ||
5deb8267 | 561 | static int __add_dirty_inode(struct inode *inode, struct dir_inode_entry *new) |
127e670a JK |
562 | { |
563 | struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); | |
564 | struct list_head *head = &sbi->dir_inode_list; | |
127e670a JK |
565 | struct list_head *this; |
566 | ||
5deb8267 JK |
567 | list_for_each(this, head) { |
568 | struct dir_inode_entry *entry; | |
569 | entry = list_entry(this, struct dir_inode_entry, list); | |
6bacf52f | 570 | if (unlikely(entry->inode == inode)) |
5deb8267 JK |
571 | return -EEXIST; |
572 | } | |
573 | list_add_tail(&new->list, head); | |
dcdfff65 | 574 | stat_inc_dirty_dir(sbi); |
5deb8267 JK |
575 | return 0; |
576 | } | |
577 | ||
578 | void set_dirty_dir_page(struct inode *inode, struct page *page) | |
579 | { | |
580 | struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); | |
581 | struct dir_inode_entry *new; | |
582 | ||
127e670a JK |
583 | if (!S_ISDIR(inode->i_mode)) |
584 | return; | |
7bd59381 GZ |
585 | |
586 | new = f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS); | |
127e670a JK |
587 | new->inode = inode; |
588 | INIT_LIST_HEAD(&new->list); | |
589 | ||
590 | spin_lock(&sbi->dir_inode_lock); | |
5deb8267 JK |
591 | if (__add_dirty_inode(inode, new)) |
592 | kmem_cache_free(inode_entry_slab, new); | |
127e670a | 593 | |
127e670a JK |
594 | inode_inc_dirty_dents(inode); |
595 | SetPagePrivate(page); | |
5deb8267 JK |
596 | spin_unlock(&sbi->dir_inode_lock); |
597 | } | |
598 | ||
599 | void add_dirty_dir_inode(struct inode *inode) | |
600 | { | |
601 | struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); | |
7bd59381 GZ |
602 | struct dir_inode_entry *new = |
603 | f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS); | |
604 | ||
5deb8267 JK |
605 | new->inode = inode; |
606 | INIT_LIST_HEAD(&new->list); | |
127e670a | 607 | |
5deb8267 JK |
608 | spin_lock(&sbi->dir_inode_lock); |
609 | if (__add_dirty_inode(inode, new)) | |
610 | kmem_cache_free(inode_entry_slab, new); | |
127e670a JK |
611 | spin_unlock(&sbi->dir_inode_lock); |
612 | } | |
613 | ||
614 | void remove_dirty_dir_inode(struct inode *inode) | |
615 | { | |
616 | struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); | |
ce3b7d80 GZ |
617 | |
618 | struct list_head *this, *head; | |
127e670a JK |
619 | |
620 | if (!S_ISDIR(inode->i_mode)) | |
621 | return; | |
622 | ||
623 | spin_lock(&sbi->dir_inode_lock); | |
f8b2c1f9 | 624 | if (get_dirty_dents(inode)) { |
3b10b1fd JK |
625 | spin_unlock(&sbi->dir_inode_lock); |
626 | return; | |
627 | } | |
127e670a | 628 | |
ce3b7d80 | 629 | head = &sbi->dir_inode_list; |
127e670a JK |
630 | list_for_each(this, head) { |
631 | struct dir_inode_entry *entry; | |
632 | entry = list_entry(this, struct dir_inode_entry, list); | |
633 | if (entry->inode == inode) { | |
634 | list_del(&entry->list); | |
635 | kmem_cache_free(inode_entry_slab, entry); | |
dcdfff65 | 636 | stat_dec_dirty_dir(sbi); |
127e670a JK |
637 | break; |
638 | } | |
639 | } | |
127e670a | 640 | spin_unlock(&sbi->dir_inode_lock); |
74d0b917 JK |
641 | |
642 | /* Only from the recovery routine */ | |
afc3eda2 JK |
643 | if (is_inode_flag_set(F2FS_I(inode), FI_DELAY_IPUT)) { |
644 | clear_inode_flag(F2FS_I(inode), FI_DELAY_IPUT); | |
74d0b917 | 645 | iput(inode); |
afc3eda2 | 646 | } |
74d0b917 JK |
647 | } |
648 | ||
649 | struct inode *check_dirty_dir_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
650 | { | |
ce3b7d80 GZ |
651 | |
652 | struct list_head *this, *head; | |
74d0b917 JK |
653 | struct inode *inode = NULL; |
654 | ||
655 | spin_lock(&sbi->dir_inode_lock); | |
ce3b7d80 GZ |
656 | |
657 | head = &sbi->dir_inode_list; | |
74d0b917 JK |
658 | list_for_each(this, head) { |
659 | struct dir_inode_entry *entry; | |
660 | entry = list_entry(this, struct dir_inode_entry, list); | |
661 | if (entry->inode->i_ino == ino) { | |
662 | inode = entry->inode; | |
663 | break; | |
664 | } | |
665 | } | |
666 | spin_unlock(&sbi->dir_inode_lock); | |
667 | return inode; | |
127e670a JK |
668 | } |
669 | ||
670 | void sync_dirty_dir_inodes(struct f2fs_sb_info *sbi) | |
671 | { | |
ce3b7d80 | 672 | struct list_head *head; |
127e670a JK |
673 | struct dir_inode_entry *entry; |
674 | struct inode *inode; | |
675 | retry: | |
676 | spin_lock(&sbi->dir_inode_lock); | |
ce3b7d80 GZ |
677 | |
678 | head = &sbi->dir_inode_list; | |
127e670a JK |
679 | if (list_empty(head)) { |
680 | spin_unlock(&sbi->dir_inode_lock); | |
681 | return; | |
682 | } | |
683 | entry = list_entry(head->next, struct dir_inode_entry, list); | |
684 | inode = igrab(entry->inode); | |
685 | spin_unlock(&sbi->dir_inode_lock); | |
686 | if (inode) { | |
87d6f890 | 687 | filemap_fdatawrite(inode->i_mapping); |
127e670a JK |
688 | iput(inode); |
689 | } else { | |
690 | /* | |
691 | * We should submit bio, since it exists several | |
692 | * wribacking dentry pages in the freeing inode. | |
693 | */ | |
458e6197 | 694 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
127e670a JK |
695 | } |
696 | goto retry; | |
697 | } | |
698 | ||
0a8165d7 | 699 | /* |
127e670a JK |
700 | * Freeze all the FS-operations for checkpoint. |
701 | */ | |
43727527 | 702 | static void block_operations(struct f2fs_sb_info *sbi) |
127e670a | 703 | { |
127e670a JK |
704 | struct writeback_control wbc = { |
705 | .sync_mode = WB_SYNC_ALL, | |
706 | .nr_to_write = LONG_MAX, | |
707 | .for_reclaim = 0, | |
708 | }; | |
c718379b JK |
709 | struct blk_plug plug; |
710 | ||
711 | blk_start_plug(&plug); | |
712 | ||
39936837 | 713 | retry_flush_dents: |
e479556b | 714 | f2fs_lock_all(sbi); |
127e670a | 715 | /* write all the dirty dentry pages */ |
127e670a | 716 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 717 | f2fs_unlock_all(sbi); |
39936837 JK |
718 | sync_dirty_dir_inodes(sbi); |
719 | goto retry_flush_dents; | |
127e670a JK |
720 | } |
721 | ||
127e670a JK |
722 | /* |
723 | * POR: we should ensure that there is no dirty node pages | |
724 | * until finishing nat/sit flush. | |
725 | */ | |
39936837 JK |
726 | retry_flush_nodes: |
727 | mutex_lock(&sbi->node_write); | |
127e670a JK |
728 | |
729 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
39936837 JK |
730 | mutex_unlock(&sbi->node_write); |
731 | sync_node_pages(sbi, 0, &wbc); | |
732 | goto retry_flush_nodes; | |
127e670a | 733 | } |
c718379b | 734 | blk_finish_plug(&plug); |
127e670a JK |
735 | } |
736 | ||
737 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
738 | { | |
39936837 | 739 | mutex_unlock(&sbi->node_write); |
e479556b | 740 | f2fs_unlock_all(sbi); |
127e670a JK |
741 | } |
742 | ||
fb51b5ef CL |
743 | static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi) |
744 | { | |
745 | DEFINE_WAIT(wait); | |
746 | ||
747 | for (;;) { | |
748 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
749 | ||
750 | if (!get_pages(sbi, F2FS_WRITEBACK)) | |
751 | break; | |
752 | ||
753 | io_schedule(); | |
754 | } | |
755 | finish_wait(&sbi->cp_wait, &wait); | |
756 | } | |
757 | ||
127e670a JK |
758 | static void do_checkpoint(struct f2fs_sb_info *sbi, bool is_umount) |
759 | { | |
760 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
761 | nid_t last_nid = 0; | |
762 | block_t start_blk; | |
763 | struct page *cp_page; | |
764 | unsigned int data_sum_blocks, orphan_blocks; | |
7e586fa0 | 765 | __u32 crc32 = 0; |
127e670a | 766 | void *kaddr; |
127e670a JK |
767 | int i; |
768 | ||
769 | /* Flush all the NAT/SIT pages */ | |
770 | while (get_pages(sbi, F2FS_DIRTY_META)) | |
771 | sync_meta_pages(sbi, META, LONG_MAX); | |
772 | ||
773 | next_free_nid(sbi, &last_nid); | |
774 | ||
775 | /* | |
776 | * modify checkpoint | |
777 | * version number is already updated | |
778 | */ | |
779 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi)); | |
780 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
781 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); | |
782 | for (i = 0; i < 3; i++) { | |
783 | ckpt->cur_node_segno[i] = | |
784 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
785 | ckpt->cur_node_blkoff[i] = | |
786 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
787 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
788 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
789 | } | |
790 | for (i = 0; i < 3; i++) { | |
791 | ckpt->cur_data_segno[i] = | |
792 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
793 | ckpt->cur_data_blkoff[i] = | |
794 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
795 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
796 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
797 | } | |
798 | ||
799 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
800 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
801 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
802 | ||
803 | /* 2 cp + n data seg summary + orphan inode blocks */ | |
804 | data_sum_blocks = npages_for_summary_flush(sbi); | |
805 | if (data_sum_blocks < 3) | |
25ca923b | 806 | set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 807 | else |
25ca923b | 808 | clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a JK |
809 | |
810 | orphan_blocks = (sbi->n_orphans + F2FS_ORPHANS_PER_BLOCK - 1) | |
811 | / F2FS_ORPHANS_PER_BLOCK; | |
25ca923b | 812 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + orphan_blocks); |
127e670a JK |
813 | |
814 | if (is_umount) { | |
25ca923b JK |
815 | set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); |
816 | ckpt->cp_pack_total_block_count = cpu_to_le32(2 + | |
817 | data_sum_blocks + orphan_blocks + NR_CURSEG_NODE_TYPE); | |
127e670a | 818 | } else { |
25ca923b JK |
819 | clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); |
820 | ckpt->cp_pack_total_block_count = cpu_to_le32(2 + | |
821 | data_sum_blocks + orphan_blocks); | |
127e670a JK |
822 | } |
823 | ||
824 | if (sbi->n_orphans) | |
25ca923b | 825 | set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 826 | else |
25ca923b | 827 | clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a JK |
828 | |
829 | /* update SIT/NAT bitmap */ | |
830 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
831 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
832 | ||
833 | crc32 = f2fs_crc32(ckpt, le32_to_cpu(ckpt->checksum_offset)); | |
7e586fa0 JK |
834 | *((__le32 *)((unsigned char *)ckpt + |
835 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
836 | = cpu_to_le32(crc32); |
837 | ||
838 | start_blk = __start_cp_addr(sbi); | |
839 | ||
840 | /* write out checkpoint buffer at block 0 */ | |
841 | cp_page = grab_meta_page(sbi, start_blk++); | |
842 | kaddr = page_address(cp_page); | |
843 | memcpy(kaddr, ckpt, (1 << sbi->log_blocksize)); | |
844 | set_page_dirty(cp_page); | |
845 | f2fs_put_page(cp_page, 1); | |
846 | ||
847 | if (sbi->n_orphans) { | |
848 | write_orphan_inodes(sbi, start_blk); | |
849 | start_blk += orphan_blocks; | |
850 | } | |
851 | ||
852 | write_data_summaries(sbi, start_blk); | |
853 | start_blk += data_sum_blocks; | |
854 | if (is_umount) { | |
855 | write_node_summaries(sbi, start_blk); | |
856 | start_blk += NR_CURSEG_NODE_TYPE; | |
857 | } | |
858 | ||
859 | /* writeout checkpoint block */ | |
860 | cp_page = grab_meta_page(sbi, start_blk); | |
861 | kaddr = page_address(cp_page); | |
862 | memcpy(kaddr, ckpt, (1 << sbi->log_blocksize)); | |
863 | set_page_dirty(cp_page); | |
864 | f2fs_put_page(cp_page, 1); | |
865 | ||
866 | /* wait for previous submitted node/meta pages writeback */ | |
fb51b5ef | 867 | wait_on_all_pages_writeback(sbi); |
127e670a | 868 | |
4ef51a8f | 869 | filemap_fdatawait_range(NODE_MAPPING(sbi), 0, LONG_MAX); |
9df27d98 | 870 | filemap_fdatawait_range(META_MAPPING(sbi), 0, LONG_MAX); |
127e670a JK |
871 | |
872 | /* update user_block_counts */ | |
873 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
874 | sbi->alloc_valid_block_count = 0; | |
875 | ||
876 | /* Here, we only have one bio having CP pack */ | |
577e3495 | 877 | sync_meta_pages(sbi, META_FLUSH, LONG_MAX); |
127e670a | 878 | |
6bacf52f | 879 | if (unlikely(!is_set_ckpt_flags(ckpt, CP_ERROR_FLAG))) { |
577e3495 JK |
880 | clear_prefree_segments(sbi); |
881 | F2FS_RESET_SB_DIRT(sbi); | |
882 | } | |
127e670a JK |
883 | } |
884 | ||
0a8165d7 | 885 | /* |
127e670a JK |
886 | * We guarantee that this checkpoint procedure should not fail. |
887 | */ | |
43727527 | 888 | void write_checkpoint(struct f2fs_sb_info *sbi, bool is_umount) |
127e670a JK |
889 | { |
890 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
891 | unsigned long long ckpt_ver; | |
892 | ||
2af4bd6c NJ |
893 | trace_f2fs_write_checkpoint(sbi->sb, is_umount, "start block_ops"); |
894 | ||
43727527 JK |
895 | mutex_lock(&sbi->cp_mutex); |
896 | block_operations(sbi); | |
127e670a | 897 | |
2af4bd6c NJ |
898 | trace_f2fs_write_checkpoint(sbi->sb, is_umount, "finish block_ops"); |
899 | ||
458e6197 JK |
900 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
901 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
902 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
127e670a JK |
903 | |
904 | /* | |
905 | * update checkpoint pack index | |
906 | * Increase the version number so that | |
907 | * SIT entries and seg summaries are written at correct place | |
908 | */ | |
d71b5564 | 909 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
910 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
911 | ||
912 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
913 | flush_nat_entries(sbi); | |
914 | flush_sit_entries(sbi); | |
915 | ||
127e670a JK |
916 | /* unlock all the fs_lock[] in do_checkpoint() */ |
917 | do_checkpoint(sbi, is_umount); | |
918 | ||
919 | unblock_operations(sbi); | |
920 | mutex_unlock(&sbi->cp_mutex); | |
2af4bd6c | 921 | |
942e0be6 | 922 | stat_inc_cp_count(sbi->stat_info); |
2af4bd6c | 923 | trace_f2fs_write_checkpoint(sbi->sb, is_umount, "finish checkpoint"); |
127e670a JK |
924 | } |
925 | ||
926 | void init_orphan_info(struct f2fs_sb_info *sbi) | |
927 | { | |
17b692f6 | 928 | spin_lock_init(&sbi->orphan_inode_lock); |
127e670a JK |
929 | INIT_LIST_HEAD(&sbi->orphan_inode_list); |
930 | sbi->n_orphans = 0; | |
0d47c1ad GZ |
931 | /* |
932 | * considering 512 blocks in a segment 8 blocks are needed for cp | |
933 | * and log segment summaries. Remaining blocks are used to keep | |
934 | * orphan entries with the limitation one reserved segment | |
935 | * for cp pack we can have max 1020*504 orphan entries | |
936 | */ | |
937 | sbi->max_orphans = (sbi->blocks_per_seg - 2 - NR_CURSEG_TYPE) | |
938 | * F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
939 | } |
940 | ||
6e6093a8 | 941 | int __init create_checkpoint_caches(void) |
127e670a JK |
942 | { |
943 | orphan_entry_slab = f2fs_kmem_cache_create("f2fs_orphan_entry", | |
e8512d2e | 944 | sizeof(struct orphan_inode_entry)); |
6bacf52f | 945 | if (!orphan_entry_slab) |
127e670a JK |
946 | return -ENOMEM; |
947 | inode_entry_slab = f2fs_kmem_cache_create("f2fs_dirty_dir_entry", | |
e8512d2e | 948 | sizeof(struct dir_inode_entry)); |
6bacf52f | 949 | if (!inode_entry_slab) { |
127e670a JK |
950 | kmem_cache_destroy(orphan_entry_slab); |
951 | return -ENOMEM; | |
952 | } | |
953 | return 0; | |
954 | } | |
955 | ||
956 | void destroy_checkpoint_caches(void) | |
957 | { | |
958 | kmem_cache_destroy(orphan_entry_slab); | |
959 | kmem_cache_destroy(inode_entry_slab); | |
960 | } |