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: | |
3cb5ad15 | 36 | page = grab_cache_page_write_begin(mapping, index, AOP_FLAG_NOFS); |
127e670a JK |
37 | if (!page) { |
38 | cond_resched(); | |
39 | goto repeat; | |
40 | } | |
41 | ||
127e670a JK |
42 | SetPageUptodate(page); |
43 | return page; | |
44 | } | |
45 | ||
0a8165d7 | 46 | /* |
127e670a JK |
47 | * We guarantee no failure on the returned page. |
48 | */ | |
49 | struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
50 | { | |
9df27d98 | 51 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
52 | struct page *page; |
53 | repeat: | |
54 | page = grab_cache_page(mapping, index); | |
55 | if (!page) { | |
56 | cond_resched(); | |
57 | goto repeat; | |
58 | } | |
393ff91f JK |
59 | if (PageUptodate(page)) |
60 | goto out; | |
61 | ||
93dfe2ac JK |
62 | if (f2fs_submit_page_bio(sbi, page, index, |
63 | READ_SYNC | REQ_META | REQ_PRIO)) | |
127e670a | 64 | goto repeat; |
127e670a | 65 | |
393ff91f | 66 | lock_page(page); |
6bacf52f | 67 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
68 | f2fs_put_page(page, 1); |
69 | goto repeat; | |
70 | } | |
393ff91f | 71 | out: |
127e670a JK |
72 | return page; |
73 | } | |
74 | ||
662befda CY |
75 | inline int get_max_meta_blks(struct f2fs_sb_info *sbi, int type) |
76 | { | |
77 | switch (type) { | |
78 | case META_NAT: | |
79 | return NM_I(sbi)->max_nid / NAT_ENTRY_PER_BLOCK; | |
80 | case META_SIT: | |
81 | return SIT_BLK_CNT(sbi); | |
81c1a0f1 | 82 | case META_SSA: |
662befda CY |
83 | case META_CP: |
84 | return 0; | |
85 | default: | |
86 | BUG(); | |
87 | } | |
88 | } | |
89 | ||
90 | /* | |
81c1a0f1 | 91 | * Readahead CP/NAT/SIT/SSA pages |
662befda CY |
92 | */ |
93 | int ra_meta_pages(struct f2fs_sb_info *sbi, int start, int nrpages, int type) | |
94 | { | |
95 | block_t prev_blk_addr = 0; | |
96 | struct page *page; | |
97 | int blkno = start; | |
98 | int max_blks = get_max_meta_blks(sbi, type); | |
99 | ||
100 | struct f2fs_io_info fio = { | |
101 | .type = META, | |
102 | .rw = READ_SYNC | REQ_META | REQ_PRIO | |
103 | }; | |
104 | ||
105 | for (; nrpages-- > 0; blkno++) { | |
106 | block_t blk_addr; | |
107 | ||
108 | switch (type) { | |
109 | case META_NAT: | |
110 | /* get nat block addr */ | |
111 | if (unlikely(blkno >= max_blks)) | |
112 | blkno = 0; | |
113 | blk_addr = current_nat_addr(sbi, | |
114 | blkno * NAT_ENTRY_PER_BLOCK); | |
115 | break; | |
116 | case META_SIT: | |
117 | /* get sit block addr */ | |
118 | if (unlikely(blkno >= max_blks)) | |
119 | goto out; | |
120 | blk_addr = current_sit_addr(sbi, | |
121 | blkno * SIT_ENTRY_PER_BLOCK); | |
122 | if (blkno != start && prev_blk_addr + 1 != blk_addr) | |
123 | goto out; | |
124 | prev_blk_addr = blk_addr; | |
125 | break; | |
81c1a0f1 | 126 | case META_SSA: |
662befda | 127 | case META_CP: |
81c1a0f1 | 128 | /* get ssa/cp block addr */ |
662befda CY |
129 | blk_addr = blkno; |
130 | break; | |
131 | default: | |
132 | BUG(); | |
133 | } | |
134 | ||
135 | page = grab_cache_page(META_MAPPING(sbi), blk_addr); | |
136 | if (!page) | |
137 | continue; | |
138 | if (PageUptodate(page)) { | |
662befda CY |
139 | f2fs_put_page(page, 1); |
140 | continue; | |
141 | } | |
142 | ||
143 | f2fs_submit_page_mbio(sbi, page, blk_addr, &fio); | |
662befda CY |
144 | f2fs_put_page(page, 0); |
145 | } | |
146 | out: | |
147 | f2fs_submit_merged_bio(sbi, META, READ); | |
148 | return blkno - start; | |
149 | } | |
150 | ||
127e670a JK |
151 | static int f2fs_write_meta_page(struct page *page, |
152 | struct writeback_control *wbc) | |
153 | { | |
154 | struct inode *inode = page->mapping->host; | |
155 | struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); | |
127e670a | 156 | |
203681f6 | 157 | if (unlikely(sbi->por_doing)) |
cfb271d4 | 158 | goto redirty_out; |
cfb271d4 CY |
159 | if (wbc->for_reclaim) |
160 | goto redirty_out; | |
127e670a | 161 | |
203681f6 JK |
162 | /* Should not write any meta pages, if any IO error was occurred */ |
163 | if (unlikely(is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ERROR_FLAG))) | |
164 | goto no_write; | |
127e670a | 165 | |
3cb5ad15 | 166 | f2fs_wait_on_page_writeback(page, META); |
577e3495 | 167 | write_meta_page(sbi, page); |
203681f6 | 168 | no_write: |
577e3495 JK |
169 | dec_page_count(sbi, F2FS_DIRTY_META); |
170 | unlock_page(page); | |
171 | return 0; | |
cfb271d4 CY |
172 | |
173 | redirty_out: | |
174 | dec_page_count(sbi, F2FS_DIRTY_META); | |
175 | wbc->pages_skipped++; | |
9cf3c389 | 176 | account_page_redirty(page); |
cfb271d4 CY |
177 | set_page_dirty(page); |
178 | return AOP_WRITEPAGE_ACTIVATE; | |
127e670a JK |
179 | } |
180 | ||
181 | static int f2fs_write_meta_pages(struct address_space *mapping, | |
182 | struct writeback_control *wbc) | |
183 | { | |
184 | struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb); | |
50c8cdb3 | 185 | long diff, written; |
127e670a | 186 | |
5459aa97 | 187 | /* collect a number of dirty meta pages and write together */ |
50c8cdb3 JK |
188 | if (wbc->for_kupdate || |
189 | get_pages(sbi, F2FS_DIRTY_META) < nr_pages_to_skip(sbi, META)) | |
d3baf95d | 190 | goto skip_write; |
127e670a JK |
191 | |
192 | /* if mounting is failed, skip writing node pages */ | |
193 | mutex_lock(&sbi->cp_mutex); | |
50c8cdb3 JK |
194 | diff = nr_pages_to_write(sbi, META, wbc); |
195 | written = sync_meta_pages(sbi, META, wbc->nr_to_write); | |
127e670a | 196 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 197 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 198 | return 0; |
d3baf95d JK |
199 | |
200 | skip_write: | |
201 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
202 | return 0; | |
127e670a JK |
203 | } |
204 | ||
205 | long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, | |
206 | long nr_to_write) | |
207 | { | |
9df27d98 | 208 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
209 | pgoff_t index = 0, end = LONG_MAX; |
210 | struct pagevec pvec; | |
211 | long nwritten = 0; | |
212 | struct writeback_control wbc = { | |
213 | .for_reclaim = 0, | |
214 | }; | |
215 | ||
216 | pagevec_init(&pvec, 0); | |
217 | ||
218 | while (index <= end) { | |
219 | int i, nr_pages; | |
220 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, | |
221 | PAGECACHE_TAG_DIRTY, | |
222 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
cfb271d4 | 223 | if (unlikely(nr_pages == 0)) |
127e670a JK |
224 | break; |
225 | ||
226 | for (i = 0; i < nr_pages; i++) { | |
227 | struct page *page = pvec.pages[i]; | |
203681f6 | 228 | |
127e670a | 229 | lock_page(page); |
203681f6 JK |
230 | |
231 | if (unlikely(page->mapping != mapping)) { | |
232 | continue_unlock: | |
233 | unlock_page(page); | |
234 | continue; | |
235 | } | |
236 | if (!PageDirty(page)) { | |
237 | /* someone wrote it for us */ | |
238 | goto continue_unlock; | |
239 | } | |
240 | ||
241 | if (!clear_page_dirty_for_io(page)) | |
242 | goto continue_unlock; | |
243 | ||
577e3495 JK |
244 | if (f2fs_write_meta_page(page, &wbc)) { |
245 | unlock_page(page); | |
246 | break; | |
247 | } | |
cfb271d4 CY |
248 | nwritten++; |
249 | if (unlikely(nwritten >= nr_to_write)) | |
127e670a JK |
250 | break; |
251 | } | |
252 | pagevec_release(&pvec); | |
253 | cond_resched(); | |
254 | } | |
255 | ||
256 | if (nwritten) | |
458e6197 | 257 | f2fs_submit_merged_bio(sbi, type, WRITE); |
127e670a JK |
258 | |
259 | return nwritten; | |
260 | } | |
261 | ||
262 | static int f2fs_set_meta_page_dirty(struct page *page) | |
263 | { | |
264 | struct address_space *mapping = page->mapping; | |
265 | struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb); | |
266 | ||
26c6b887 JK |
267 | trace_f2fs_set_page_dirty(page, META); |
268 | ||
127e670a JK |
269 | SetPageUptodate(page); |
270 | if (!PageDirty(page)) { | |
271 | __set_page_dirty_nobuffers(page); | |
272 | inc_page_count(sbi, F2FS_DIRTY_META); | |
127e670a JK |
273 | return 1; |
274 | } | |
275 | return 0; | |
276 | } | |
277 | ||
278 | const struct address_space_operations f2fs_meta_aops = { | |
279 | .writepage = f2fs_write_meta_page, | |
280 | .writepages = f2fs_write_meta_pages, | |
281 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
282 | }; | |
283 | ||
cbd56e7d | 284 | int acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 285 | { |
127e670a JK |
286 | int err = 0; |
287 | ||
17b692f6 | 288 | spin_lock(&sbi->orphan_inode_lock); |
0d47c1ad | 289 | if (unlikely(sbi->n_orphans >= sbi->max_orphans)) |
127e670a | 290 | err = -ENOSPC; |
cbd56e7d JK |
291 | else |
292 | sbi->n_orphans++; | |
17b692f6 | 293 | spin_unlock(&sbi->orphan_inode_lock); |
0d47c1ad | 294 | |
127e670a JK |
295 | return err; |
296 | } | |
297 | ||
cbd56e7d JK |
298 | void release_orphan_inode(struct f2fs_sb_info *sbi) |
299 | { | |
17b692f6 | 300 | spin_lock(&sbi->orphan_inode_lock); |
5d56b671 | 301 | f2fs_bug_on(sbi->n_orphans == 0); |
cbd56e7d | 302 | sbi->n_orphans--; |
17b692f6 | 303 | spin_unlock(&sbi->orphan_inode_lock); |
cbd56e7d JK |
304 | } |
305 | ||
127e670a JK |
306 | void add_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
307 | { | |
2d7b822a CY |
308 | struct list_head *head; |
309 | struct orphan_inode_entry *new, *orphan; | |
127e670a | 310 | |
c1ef3725 GZ |
311 | new = f2fs_kmem_cache_alloc(orphan_entry_slab, GFP_ATOMIC); |
312 | new->ino = ino; | |
313 | ||
17b692f6 | 314 | spin_lock(&sbi->orphan_inode_lock); |
127e670a | 315 | head = &sbi->orphan_inode_list; |
2d7b822a | 316 | list_for_each_entry(orphan, head, list) { |
c1ef3725 | 317 | if (orphan->ino == ino) { |
17b692f6 | 318 | spin_unlock(&sbi->orphan_inode_lock); |
c1ef3725 GZ |
319 | kmem_cache_free(orphan_entry_slab, new); |
320 | return; | |
321 | } | |
322 | ||
127e670a JK |
323 | if (orphan->ino > ino) |
324 | break; | |
127e670a | 325 | } |
7bd59381 | 326 | |
2d7b822a CY |
327 | /* add new orphan entry into list which is sorted by inode number */ |
328 | list_add_tail(&new->list, &orphan->list); | |
17b692f6 | 329 | spin_unlock(&sbi->orphan_inode_lock); |
127e670a JK |
330 | } |
331 | ||
332 | void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
333 | { | |
60ed9a0f | 334 | struct list_head *head; |
127e670a JK |
335 | struct orphan_inode_entry *orphan; |
336 | ||
17b692f6 | 337 | spin_lock(&sbi->orphan_inode_lock); |
127e670a | 338 | head = &sbi->orphan_inode_list; |
60ed9a0f | 339 | list_for_each_entry(orphan, head, list) { |
127e670a JK |
340 | if (orphan->ino == ino) { |
341 | list_del(&orphan->list); | |
5d56b671 | 342 | f2fs_bug_on(sbi->n_orphans == 0); |
127e670a | 343 | sbi->n_orphans--; |
cf0ee0f0 CY |
344 | spin_unlock(&sbi->orphan_inode_lock); |
345 | kmem_cache_free(orphan_entry_slab, orphan); | |
346 | return; | |
127e670a JK |
347 | } |
348 | } | |
17b692f6 | 349 | spin_unlock(&sbi->orphan_inode_lock); |
127e670a JK |
350 | } |
351 | ||
352 | static void recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
353 | { | |
354 | struct inode *inode = f2fs_iget(sbi->sb, ino); | |
5d56b671 | 355 | f2fs_bug_on(IS_ERR(inode)); |
127e670a JK |
356 | clear_nlink(inode); |
357 | ||
358 | /* truncate all the data during iput */ | |
359 | iput(inode); | |
360 | } | |
361 | ||
8f99a946 | 362 | void recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a JK |
363 | { |
364 | block_t start_blk, orphan_blkaddr, i, j; | |
365 | ||
25ca923b | 366 | if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG)) |
8f99a946 | 367 | return; |
127e670a | 368 | |
aabe5136 | 369 | sbi->por_doing = true; |
127e670a JK |
370 | start_blk = __start_cp_addr(sbi) + 1; |
371 | orphan_blkaddr = __start_sum_addr(sbi) - 1; | |
372 | ||
662befda CY |
373 | ra_meta_pages(sbi, start_blk, orphan_blkaddr, META_CP); |
374 | ||
127e670a JK |
375 | for (i = 0; i < orphan_blkaddr; i++) { |
376 | struct page *page = get_meta_page(sbi, start_blk + i); | |
377 | struct f2fs_orphan_block *orphan_blk; | |
378 | ||
379 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); | |
380 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
381 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
382 | recover_orphan_inode(sbi, ino); | |
383 | } | |
384 | f2fs_put_page(page, 1); | |
385 | } | |
386 | /* clear Orphan Flag */ | |
25ca923b | 387 | clear_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG); |
aabe5136 | 388 | sbi->por_doing = false; |
8f99a946 | 389 | return; |
127e670a JK |
390 | } |
391 | ||
392 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
393 | { | |
502c6e0b | 394 | struct list_head *head; |
127e670a | 395 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 396 | unsigned int nentries = 0; |
4531929e GZ |
397 | unsigned short index; |
398 | unsigned short orphan_blocks = (unsigned short)((sbi->n_orphans + | |
399 | (F2FS_ORPHANS_PER_BLOCK - 1)) / F2FS_ORPHANS_PER_BLOCK); | |
400 | struct page *page = NULL; | |
502c6e0b | 401 | struct orphan_inode_entry *orphan = NULL; |
127e670a | 402 | |
4531929e | 403 | for (index = 0; index < orphan_blocks; index++) |
63f5384c | 404 | grab_meta_page(sbi, start_blk + index); |
127e670a | 405 | |
4531929e | 406 | index = 1; |
17b692f6 | 407 | spin_lock(&sbi->orphan_inode_lock); |
127e670a JK |
408 | head = &sbi->orphan_inode_list; |
409 | ||
410 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
411 | list_for_each_entry(orphan, head, list) { |
412 | if (!page) { | |
63f5384c GZ |
413 | page = find_get_page(META_MAPPING(sbi), start_blk++); |
414 | f2fs_bug_on(!page); | |
502c6e0b GZ |
415 | orphan_blk = |
416 | (struct f2fs_orphan_block *)page_address(page); | |
417 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
63f5384c | 418 | f2fs_put_page(page, 0); |
502c6e0b | 419 | } |
127e670a | 420 | |
36795567 | 421 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 422 | |
36795567 | 423 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
424 | /* |
425 | * an orphan block is full of 1020 entries, | |
426 | * then we need to flush current orphan blocks | |
427 | * and bring another one in memory | |
428 | */ | |
429 | orphan_blk->blk_addr = cpu_to_le16(index); | |
430 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
431 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
432 | set_page_dirty(page); | |
433 | f2fs_put_page(page, 1); | |
434 | index++; | |
127e670a JK |
435 | nentries = 0; |
436 | page = NULL; | |
437 | } | |
502c6e0b | 438 | } |
127e670a | 439 | |
502c6e0b GZ |
440 | if (page) { |
441 | orphan_blk->blk_addr = cpu_to_le16(index); | |
442 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
443 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
444 | set_page_dirty(page); | |
445 | f2fs_put_page(page, 1); | |
127e670a | 446 | } |
502c6e0b | 447 | |
17b692f6 | 448 | spin_unlock(&sbi->orphan_inode_lock); |
127e670a JK |
449 | } |
450 | ||
451 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, | |
452 | block_t cp_addr, unsigned long long *version) | |
453 | { | |
454 | struct page *cp_page_1, *cp_page_2 = NULL; | |
455 | unsigned long blk_size = sbi->blocksize; | |
456 | struct f2fs_checkpoint *cp_block; | |
457 | unsigned long long cur_version = 0, pre_version = 0; | |
127e670a | 458 | size_t crc_offset; |
7e586fa0 | 459 | __u32 crc = 0; |
127e670a JK |
460 | |
461 | /* Read the 1st cp block in this CP pack */ | |
462 | cp_page_1 = get_meta_page(sbi, cp_addr); | |
463 | ||
464 | /* get the version number */ | |
465 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_1); | |
466 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
467 | if (crc_offset >= blk_size) | |
468 | goto invalid_cp1; | |
469 | ||
7e586fa0 | 470 | crc = le32_to_cpu(*((__u32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
471 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
472 | goto invalid_cp1; | |
473 | ||
d71b5564 | 474 | pre_version = cur_cp_version(cp_block); |
127e670a JK |
475 | |
476 | /* Read the 2nd cp block in this CP pack */ | |
25ca923b | 477 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
127e670a JK |
478 | cp_page_2 = get_meta_page(sbi, cp_addr); |
479 | ||
480 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_2); | |
481 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
482 | if (crc_offset >= blk_size) | |
483 | goto invalid_cp2; | |
484 | ||
7e586fa0 | 485 | crc = le32_to_cpu(*((__u32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
486 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
487 | goto invalid_cp2; | |
488 | ||
d71b5564 | 489 | cur_version = cur_cp_version(cp_block); |
127e670a JK |
490 | |
491 | if (cur_version == pre_version) { | |
492 | *version = cur_version; | |
493 | f2fs_put_page(cp_page_2, 1); | |
494 | return cp_page_1; | |
495 | } | |
496 | invalid_cp2: | |
497 | f2fs_put_page(cp_page_2, 1); | |
498 | invalid_cp1: | |
499 | f2fs_put_page(cp_page_1, 1); | |
500 | return NULL; | |
501 | } | |
502 | ||
503 | int get_valid_checkpoint(struct f2fs_sb_info *sbi) | |
504 | { | |
505 | struct f2fs_checkpoint *cp_block; | |
506 | struct f2fs_super_block *fsb = sbi->raw_super; | |
507 | struct page *cp1, *cp2, *cur_page; | |
508 | unsigned long blk_size = sbi->blocksize; | |
509 | unsigned long long cp1_version = 0, cp2_version = 0; | |
510 | unsigned long long cp_start_blk_no; | |
511 | ||
512 | sbi->ckpt = kzalloc(blk_size, GFP_KERNEL); | |
513 | if (!sbi->ckpt) | |
514 | return -ENOMEM; | |
515 | /* | |
516 | * Finding out valid cp block involves read both | |
517 | * sets( cp pack1 and cp pack 2) | |
518 | */ | |
519 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
520 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
521 | ||
522 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
523 | cp_start_blk_no += ((unsigned long long)1) << |
524 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
525 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
526 | ||
527 | if (cp1 && cp2) { | |
528 | if (ver_after(cp2_version, cp1_version)) | |
529 | cur_page = cp2; | |
530 | else | |
531 | cur_page = cp1; | |
532 | } else if (cp1) { | |
533 | cur_page = cp1; | |
534 | } else if (cp2) { | |
535 | cur_page = cp2; | |
536 | } else { | |
537 | goto fail_no_cp; | |
538 | } | |
539 | ||
540 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
541 | memcpy(sbi->ckpt, cp_block, blk_size); | |
542 | ||
543 | f2fs_put_page(cp1, 1); | |
544 | f2fs_put_page(cp2, 1); | |
545 | return 0; | |
546 | ||
547 | fail_no_cp: | |
548 | kfree(sbi->ckpt); | |
549 | return -EINVAL; | |
550 | } | |
551 | ||
5deb8267 | 552 | static int __add_dirty_inode(struct inode *inode, struct dir_inode_entry *new) |
127e670a JK |
553 | { |
554 | struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); | |
555 | struct list_head *head = &sbi->dir_inode_list; | |
2d7b822a | 556 | struct dir_inode_entry *entry; |
127e670a | 557 | |
2d7b822a | 558 | list_for_each_entry(entry, head, list) |
6bacf52f | 559 | if (unlikely(entry->inode == inode)) |
5deb8267 | 560 | return -EEXIST; |
2d7b822a | 561 | |
5deb8267 | 562 | list_add_tail(&new->list, head); |
dcdfff65 | 563 | stat_inc_dirty_dir(sbi); |
5deb8267 JK |
564 | return 0; |
565 | } | |
566 | ||
567 | void set_dirty_dir_page(struct inode *inode, struct page *page) | |
568 | { | |
569 | struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); | |
570 | struct dir_inode_entry *new; | |
cf0ee0f0 | 571 | int ret = 0; |
5deb8267 | 572 | |
127e670a JK |
573 | if (!S_ISDIR(inode->i_mode)) |
574 | return; | |
7bd59381 GZ |
575 | |
576 | new = f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS); | |
127e670a JK |
577 | new->inode = inode; |
578 | INIT_LIST_HEAD(&new->list); | |
579 | ||
580 | spin_lock(&sbi->dir_inode_lock); | |
cf0ee0f0 | 581 | ret = __add_dirty_inode(inode, new); |
127e670a JK |
582 | inode_inc_dirty_dents(inode); |
583 | SetPagePrivate(page); | |
5deb8267 | 584 | spin_unlock(&sbi->dir_inode_lock); |
cf0ee0f0 CY |
585 | |
586 | if (ret) | |
587 | kmem_cache_free(inode_entry_slab, new); | |
5deb8267 JK |
588 | } |
589 | ||
590 | void add_dirty_dir_inode(struct inode *inode) | |
591 | { | |
592 | struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); | |
7bd59381 GZ |
593 | struct dir_inode_entry *new = |
594 | f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS); | |
cf0ee0f0 | 595 | int ret = 0; |
7bd59381 | 596 | |
5deb8267 JK |
597 | new->inode = inode; |
598 | INIT_LIST_HEAD(&new->list); | |
127e670a | 599 | |
5deb8267 | 600 | spin_lock(&sbi->dir_inode_lock); |
cf0ee0f0 | 601 | ret = __add_dirty_inode(inode, new); |
127e670a | 602 | spin_unlock(&sbi->dir_inode_lock); |
cf0ee0f0 CY |
603 | |
604 | if (ret) | |
605 | kmem_cache_free(inode_entry_slab, new); | |
127e670a JK |
606 | } |
607 | ||
608 | void remove_dirty_dir_inode(struct inode *inode) | |
609 | { | |
610 | struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); | |
2d7b822a CY |
611 | struct list_head *head; |
612 | struct dir_inode_entry *entry; | |
127e670a JK |
613 | |
614 | if (!S_ISDIR(inode->i_mode)) | |
615 | return; | |
616 | ||
617 | spin_lock(&sbi->dir_inode_lock); | |
f8b2c1f9 | 618 | if (get_dirty_dents(inode)) { |
3b10b1fd JK |
619 | spin_unlock(&sbi->dir_inode_lock); |
620 | return; | |
621 | } | |
127e670a | 622 | |
ce3b7d80 | 623 | head = &sbi->dir_inode_list; |
2d7b822a | 624 | list_for_each_entry(entry, head, list) { |
127e670a JK |
625 | if (entry->inode == inode) { |
626 | list_del(&entry->list); | |
dcdfff65 | 627 | stat_dec_dirty_dir(sbi); |
cf0ee0f0 CY |
628 | spin_unlock(&sbi->dir_inode_lock); |
629 | kmem_cache_free(inode_entry_slab, entry); | |
630 | goto done; | |
127e670a JK |
631 | } |
632 | } | |
127e670a | 633 | spin_unlock(&sbi->dir_inode_lock); |
74d0b917 | 634 | |
cf0ee0f0 | 635 | done: |
74d0b917 | 636 | /* Only from the recovery routine */ |
afc3eda2 JK |
637 | if (is_inode_flag_set(F2FS_I(inode), FI_DELAY_IPUT)) { |
638 | clear_inode_flag(F2FS_I(inode), FI_DELAY_IPUT); | |
74d0b917 | 639 | iput(inode); |
afc3eda2 | 640 | } |
74d0b917 JK |
641 | } |
642 | ||
643 | struct inode *check_dirty_dir_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
644 | { | |
ce3b7d80 | 645 | |
2d7b822a | 646 | struct list_head *head; |
74d0b917 | 647 | struct inode *inode = NULL; |
2d7b822a | 648 | struct dir_inode_entry *entry; |
74d0b917 JK |
649 | |
650 | spin_lock(&sbi->dir_inode_lock); | |
ce3b7d80 GZ |
651 | |
652 | head = &sbi->dir_inode_list; | |
2d7b822a | 653 | list_for_each_entry(entry, head, list) { |
74d0b917 JK |
654 | if (entry->inode->i_ino == ino) { |
655 | inode = entry->inode; | |
656 | break; | |
657 | } | |
658 | } | |
659 | spin_unlock(&sbi->dir_inode_lock); | |
660 | return inode; | |
127e670a JK |
661 | } |
662 | ||
663 | void sync_dirty_dir_inodes(struct f2fs_sb_info *sbi) | |
664 | { | |
ce3b7d80 | 665 | struct list_head *head; |
127e670a JK |
666 | struct dir_inode_entry *entry; |
667 | struct inode *inode; | |
668 | retry: | |
669 | spin_lock(&sbi->dir_inode_lock); | |
ce3b7d80 GZ |
670 | |
671 | head = &sbi->dir_inode_list; | |
127e670a JK |
672 | if (list_empty(head)) { |
673 | spin_unlock(&sbi->dir_inode_lock); | |
674 | return; | |
675 | } | |
676 | entry = list_entry(head->next, struct dir_inode_entry, list); | |
677 | inode = igrab(entry->inode); | |
678 | spin_unlock(&sbi->dir_inode_lock); | |
679 | if (inode) { | |
87d6f890 | 680 | filemap_fdatawrite(inode->i_mapping); |
127e670a JK |
681 | iput(inode); |
682 | } else { | |
683 | /* | |
684 | * We should submit bio, since it exists several | |
685 | * wribacking dentry pages in the freeing inode. | |
686 | */ | |
458e6197 | 687 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
127e670a JK |
688 | } |
689 | goto retry; | |
690 | } | |
691 | ||
0a8165d7 | 692 | /* |
127e670a JK |
693 | * Freeze all the FS-operations for checkpoint. |
694 | */ | |
43727527 | 695 | static void block_operations(struct f2fs_sb_info *sbi) |
127e670a | 696 | { |
127e670a JK |
697 | struct writeback_control wbc = { |
698 | .sync_mode = WB_SYNC_ALL, | |
699 | .nr_to_write = LONG_MAX, | |
700 | .for_reclaim = 0, | |
701 | }; | |
c718379b JK |
702 | struct blk_plug plug; |
703 | ||
704 | blk_start_plug(&plug); | |
705 | ||
39936837 | 706 | retry_flush_dents: |
e479556b | 707 | f2fs_lock_all(sbi); |
127e670a | 708 | /* write all the dirty dentry pages */ |
127e670a | 709 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 710 | f2fs_unlock_all(sbi); |
39936837 JK |
711 | sync_dirty_dir_inodes(sbi); |
712 | goto retry_flush_dents; | |
127e670a JK |
713 | } |
714 | ||
127e670a JK |
715 | /* |
716 | * POR: we should ensure that there is no dirty node pages | |
717 | * until finishing nat/sit flush. | |
718 | */ | |
39936837 JK |
719 | retry_flush_nodes: |
720 | mutex_lock(&sbi->node_write); | |
127e670a JK |
721 | |
722 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
39936837 JK |
723 | mutex_unlock(&sbi->node_write); |
724 | sync_node_pages(sbi, 0, &wbc); | |
725 | goto retry_flush_nodes; | |
127e670a | 726 | } |
c718379b | 727 | blk_finish_plug(&plug); |
127e670a JK |
728 | } |
729 | ||
730 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
731 | { | |
39936837 | 732 | mutex_unlock(&sbi->node_write); |
e479556b | 733 | f2fs_unlock_all(sbi); |
127e670a JK |
734 | } |
735 | ||
fb51b5ef CL |
736 | static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi) |
737 | { | |
738 | DEFINE_WAIT(wait); | |
739 | ||
740 | for (;;) { | |
741 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
742 | ||
743 | if (!get_pages(sbi, F2FS_WRITEBACK)) | |
744 | break; | |
745 | ||
746 | io_schedule(); | |
747 | } | |
748 | finish_wait(&sbi->cp_wait, &wait); | |
749 | } | |
750 | ||
127e670a JK |
751 | static void do_checkpoint(struct f2fs_sb_info *sbi, bool is_umount) |
752 | { | |
753 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
754 | nid_t last_nid = 0; | |
755 | block_t start_blk; | |
756 | struct page *cp_page; | |
757 | unsigned int data_sum_blocks, orphan_blocks; | |
7e586fa0 | 758 | __u32 crc32 = 0; |
127e670a | 759 | void *kaddr; |
127e670a JK |
760 | int i; |
761 | ||
762 | /* Flush all the NAT/SIT pages */ | |
763 | while (get_pages(sbi, F2FS_DIRTY_META)) | |
764 | sync_meta_pages(sbi, META, LONG_MAX); | |
765 | ||
766 | next_free_nid(sbi, &last_nid); | |
767 | ||
768 | /* | |
769 | * modify checkpoint | |
770 | * version number is already updated | |
771 | */ | |
772 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi)); | |
773 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
774 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); | |
775 | for (i = 0; i < 3; i++) { | |
776 | ckpt->cur_node_segno[i] = | |
777 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
778 | ckpt->cur_node_blkoff[i] = | |
779 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
780 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
781 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
782 | } | |
783 | for (i = 0; i < 3; i++) { | |
784 | ckpt->cur_data_segno[i] = | |
785 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
786 | ckpt->cur_data_blkoff[i] = | |
787 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
788 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
789 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
790 | } | |
791 | ||
792 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
793 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
794 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
795 | ||
796 | /* 2 cp + n data seg summary + orphan inode blocks */ | |
797 | data_sum_blocks = npages_for_summary_flush(sbi); | |
798 | if (data_sum_blocks < 3) | |
25ca923b | 799 | set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 800 | else |
25ca923b | 801 | clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a JK |
802 | |
803 | orphan_blocks = (sbi->n_orphans + F2FS_ORPHANS_PER_BLOCK - 1) | |
804 | / F2FS_ORPHANS_PER_BLOCK; | |
25ca923b | 805 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + orphan_blocks); |
127e670a JK |
806 | |
807 | if (is_umount) { | |
25ca923b JK |
808 | set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); |
809 | ckpt->cp_pack_total_block_count = cpu_to_le32(2 + | |
810 | data_sum_blocks + orphan_blocks + NR_CURSEG_NODE_TYPE); | |
127e670a | 811 | } else { |
25ca923b JK |
812 | clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); |
813 | ckpt->cp_pack_total_block_count = cpu_to_le32(2 + | |
814 | data_sum_blocks + orphan_blocks); | |
127e670a JK |
815 | } |
816 | ||
817 | if (sbi->n_orphans) | |
25ca923b | 818 | set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 819 | else |
25ca923b | 820 | clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a JK |
821 | |
822 | /* update SIT/NAT bitmap */ | |
823 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
824 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
825 | ||
826 | crc32 = f2fs_crc32(ckpt, le32_to_cpu(ckpt->checksum_offset)); | |
7e586fa0 JK |
827 | *((__le32 *)((unsigned char *)ckpt + |
828 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
829 | = cpu_to_le32(crc32); |
830 | ||
831 | start_blk = __start_cp_addr(sbi); | |
832 | ||
833 | /* write out checkpoint buffer at block 0 */ | |
834 | cp_page = grab_meta_page(sbi, start_blk++); | |
835 | kaddr = page_address(cp_page); | |
836 | memcpy(kaddr, ckpt, (1 << sbi->log_blocksize)); | |
837 | set_page_dirty(cp_page); | |
838 | f2fs_put_page(cp_page, 1); | |
839 | ||
840 | if (sbi->n_orphans) { | |
841 | write_orphan_inodes(sbi, start_blk); | |
842 | start_blk += orphan_blocks; | |
843 | } | |
844 | ||
845 | write_data_summaries(sbi, start_blk); | |
846 | start_blk += data_sum_blocks; | |
847 | if (is_umount) { | |
848 | write_node_summaries(sbi, start_blk); | |
849 | start_blk += NR_CURSEG_NODE_TYPE; | |
850 | } | |
851 | ||
852 | /* writeout checkpoint block */ | |
853 | cp_page = grab_meta_page(sbi, start_blk); | |
854 | kaddr = page_address(cp_page); | |
855 | memcpy(kaddr, ckpt, (1 << sbi->log_blocksize)); | |
856 | set_page_dirty(cp_page); | |
857 | f2fs_put_page(cp_page, 1); | |
858 | ||
859 | /* wait for previous submitted node/meta pages writeback */ | |
fb51b5ef | 860 | wait_on_all_pages_writeback(sbi); |
127e670a | 861 | |
4ef51a8f | 862 | filemap_fdatawait_range(NODE_MAPPING(sbi), 0, LONG_MAX); |
9df27d98 | 863 | filemap_fdatawait_range(META_MAPPING(sbi), 0, LONG_MAX); |
127e670a JK |
864 | |
865 | /* update user_block_counts */ | |
866 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
867 | sbi->alloc_valid_block_count = 0; | |
868 | ||
869 | /* Here, we only have one bio having CP pack */ | |
577e3495 | 870 | sync_meta_pages(sbi, META_FLUSH, LONG_MAX); |
127e670a | 871 | |
6bacf52f | 872 | if (unlikely(!is_set_ckpt_flags(ckpt, CP_ERROR_FLAG))) { |
577e3495 JK |
873 | clear_prefree_segments(sbi); |
874 | F2FS_RESET_SB_DIRT(sbi); | |
875 | } | |
127e670a JK |
876 | } |
877 | ||
0a8165d7 | 878 | /* |
127e670a JK |
879 | * We guarantee that this checkpoint procedure should not fail. |
880 | */ | |
43727527 | 881 | void write_checkpoint(struct f2fs_sb_info *sbi, bool is_umount) |
127e670a JK |
882 | { |
883 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
884 | unsigned long long ckpt_ver; | |
885 | ||
2af4bd6c NJ |
886 | trace_f2fs_write_checkpoint(sbi->sb, is_umount, "start block_ops"); |
887 | ||
43727527 JK |
888 | mutex_lock(&sbi->cp_mutex); |
889 | block_operations(sbi); | |
127e670a | 890 | |
2af4bd6c NJ |
891 | trace_f2fs_write_checkpoint(sbi->sb, is_umount, "finish block_ops"); |
892 | ||
458e6197 JK |
893 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
894 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
895 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
127e670a JK |
896 | |
897 | /* | |
898 | * update checkpoint pack index | |
899 | * Increase the version number so that | |
900 | * SIT entries and seg summaries are written at correct place | |
901 | */ | |
d71b5564 | 902 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
903 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
904 | ||
905 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
906 | flush_nat_entries(sbi); | |
907 | flush_sit_entries(sbi); | |
908 | ||
127e670a JK |
909 | /* unlock all the fs_lock[] in do_checkpoint() */ |
910 | do_checkpoint(sbi, is_umount); | |
911 | ||
912 | unblock_operations(sbi); | |
913 | mutex_unlock(&sbi->cp_mutex); | |
2af4bd6c | 914 | |
942e0be6 | 915 | stat_inc_cp_count(sbi->stat_info); |
2af4bd6c | 916 | trace_f2fs_write_checkpoint(sbi->sb, is_umount, "finish checkpoint"); |
127e670a JK |
917 | } |
918 | ||
919 | void init_orphan_info(struct f2fs_sb_info *sbi) | |
920 | { | |
17b692f6 | 921 | spin_lock_init(&sbi->orphan_inode_lock); |
127e670a JK |
922 | INIT_LIST_HEAD(&sbi->orphan_inode_list); |
923 | sbi->n_orphans = 0; | |
0d47c1ad GZ |
924 | /* |
925 | * considering 512 blocks in a segment 8 blocks are needed for cp | |
926 | * and log segment summaries. Remaining blocks are used to keep | |
927 | * orphan entries with the limitation one reserved segment | |
928 | * for cp pack we can have max 1020*504 orphan entries | |
929 | */ | |
930 | sbi->max_orphans = (sbi->blocks_per_seg - 2 - NR_CURSEG_TYPE) | |
931 | * F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
932 | } |
933 | ||
6e6093a8 | 934 | int __init create_checkpoint_caches(void) |
127e670a JK |
935 | { |
936 | orphan_entry_slab = f2fs_kmem_cache_create("f2fs_orphan_entry", | |
e8512d2e | 937 | sizeof(struct orphan_inode_entry)); |
6bacf52f | 938 | if (!orphan_entry_slab) |
127e670a JK |
939 | return -ENOMEM; |
940 | inode_entry_slab = f2fs_kmem_cache_create("f2fs_dirty_dir_entry", | |
e8512d2e | 941 | sizeof(struct dir_inode_entry)); |
6bacf52f | 942 | if (!inode_entry_slab) { |
127e670a JK |
943 | kmem_cache_destroy(orphan_entry_slab); |
944 | return -ENOMEM; | |
945 | } | |
946 | return 0; | |
947 | } | |
948 | ||
949 | void destroy_checkpoint_caches(void) | |
950 | { | |
951 | kmem_cache_destroy(orphan_entry_slab); | |
952 | kmem_cache_destroy(inode_entry_slab); | |
953 | } |