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 | 24 | |
6451e041 | 25 | static struct kmem_cache *ino_entry_slab; |
127e670a JK |
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: | |
bde44686 | 36 | page = grab_cache_page(mapping, index); |
127e670a JK |
37 | if (!page) { |
38 | cond_resched(); | |
39 | goto repeat; | |
40 | } | |
bde44686 | 41 | f2fs_wait_on_page_writeback(page, META); |
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 | ||
4c521f49 JK |
75 | struct page *get_meta_page_ra(struct f2fs_sb_info *sbi, pgoff_t index) |
76 | { | |
77 | bool readahead = false; | |
78 | struct page *page; | |
79 | ||
80 | page = find_get_page(META_MAPPING(sbi), index); | |
81 | if (!page || (page && !PageUptodate(page))) | |
82 | readahead = true; | |
83 | f2fs_put_page(page, 0); | |
84 | ||
85 | if (readahead) | |
90a893c7 | 86 | ra_meta_pages(sbi, index, MAX_BIO_BLOCKS(sbi), META_POR); |
4c521f49 JK |
87 | return get_meta_page(sbi, index); |
88 | } | |
89 | ||
90 | static inline block_t get_max_meta_blks(struct f2fs_sb_info *sbi, int type) | |
662befda CY |
91 | { |
92 | switch (type) { | |
93 | case META_NAT: | |
94 | return NM_I(sbi)->max_nid / NAT_ENTRY_PER_BLOCK; | |
95 | case META_SIT: | |
96 | return SIT_BLK_CNT(sbi); | |
81c1a0f1 | 97 | case META_SSA: |
662befda CY |
98 | case META_CP: |
99 | return 0; | |
4c521f49 | 100 | case META_POR: |
7cd8558b | 101 | return MAX_BLKADDR(sbi); |
662befda CY |
102 | default: |
103 | BUG(); | |
104 | } | |
105 | } | |
106 | ||
107 | /* | |
81c1a0f1 | 108 | * Readahead CP/NAT/SIT/SSA pages |
662befda | 109 | */ |
4c521f49 | 110 | int ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages, int type) |
662befda CY |
111 | { |
112 | block_t prev_blk_addr = 0; | |
113 | struct page *page; | |
4c521f49 JK |
114 | block_t blkno = start; |
115 | block_t max_blks = get_max_meta_blks(sbi, type); | |
662befda CY |
116 | |
117 | struct f2fs_io_info fio = { | |
118 | .type = META, | |
119 | .rw = READ_SYNC | REQ_META | REQ_PRIO | |
120 | }; | |
121 | ||
122 | for (; nrpages-- > 0; blkno++) { | |
123 | block_t blk_addr; | |
124 | ||
125 | switch (type) { | |
126 | case META_NAT: | |
127 | /* get nat block addr */ | |
128 | if (unlikely(blkno >= max_blks)) | |
129 | blkno = 0; | |
130 | blk_addr = current_nat_addr(sbi, | |
131 | blkno * NAT_ENTRY_PER_BLOCK); | |
132 | break; | |
133 | case META_SIT: | |
134 | /* get sit block addr */ | |
135 | if (unlikely(blkno >= max_blks)) | |
136 | goto out; | |
137 | blk_addr = current_sit_addr(sbi, | |
138 | blkno * SIT_ENTRY_PER_BLOCK); | |
139 | if (blkno != start && prev_blk_addr + 1 != blk_addr) | |
140 | goto out; | |
141 | prev_blk_addr = blk_addr; | |
142 | break; | |
81c1a0f1 | 143 | case META_SSA: |
662befda | 144 | case META_CP: |
4c521f49 JK |
145 | case META_POR: |
146 | if (unlikely(blkno >= max_blks)) | |
147 | goto out; | |
7cd8558b | 148 | if (unlikely(blkno < SEG0_BLKADDR(sbi))) |
4c521f49 | 149 | goto out; |
662befda CY |
150 | blk_addr = blkno; |
151 | break; | |
152 | default: | |
153 | BUG(); | |
154 | } | |
155 | ||
156 | page = grab_cache_page(META_MAPPING(sbi), blk_addr); | |
157 | if (!page) | |
158 | continue; | |
159 | if (PageUptodate(page)) { | |
662befda CY |
160 | f2fs_put_page(page, 1); |
161 | continue; | |
162 | } | |
163 | ||
164 | f2fs_submit_page_mbio(sbi, page, blk_addr, &fio); | |
662befda CY |
165 | f2fs_put_page(page, 0); |
166 | } | |
167 | out: | |
168 | f2fs_submit_merged_bio(sbi, META, READ); | |
169 | return blkno - start; | |
170 | } | |
171 | ||
127e670a JK |
172 | static int f2fs_write_meta_page(struct page *page, |
173 | struct writeback_control *wbc) | |
174 | { | |
4081363f | 175 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 176 | |
ecda0de3 CY |
177 | trace_f2fs_writepage(page, META); |
178 | ||
203681f6 | 179 | if (unlikely(sbi->por_doing)) |
cfb271d4 | 180 | goto redirty_out; |
cfb271d4 CY |
181 | if (wbc->for_reclaim) |
182 | goto redirty_out; | |
1e968fdf | 183 | if (unlikely(f2fs_cp_error(sbi))) |
cf779cab | 184 | goto redirty_out; |
127e670a | 185 | |
3cb5ad15 | 186 | f2fs_wait_on_page_writeback(page, META); |
577e3495 JK |
187 | write_meta_page(sbi, page); |
188 | dec_page_count(sbi, F2FS_DIRTY_META); | |
189 | unlock_page(page); | |
190 | return 0; | |
cfb271d4 CY |
191 | |
192 | redirty_out: | |
76f60268 | 193 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 194 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
195 | } |
196 | ||
197 | static int f2fs_write_meta_pages(struct address_space *mapping, | |
198 | struct writeback_control *wbc) | |
199 | { | |
4081363f | 200 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 201 | long diff, written; |
127e670a | 202 | |
e5748434 CY |
203 | trace_f2fs_writepages(mapping->host, wbc, META); |
204 | ||
5459aa97 | 205 | /* collect a number of dirty meta pages and write together */ |
50c8cdb3 JK |
206 | if (wbc->for_kupdate || |
207 | get_pages(sbi, F2FS_DIRTY_META) < nr_pages_to_skip(sbi, META)) | |
d3baf95d | 208 | goto skip_write; |
127e670a JK |
209 | |
210 | /* if mounting is failed, skip writing node pages */ | |
211 | mutex_lock(&sbi->cp_mutex); | |
50c8cdb3 JK |
212 | diff = nr_pages_to_write(sbi, META, wbc); |
213 | written = sync_meta_pages(sbi, META, wbc->nr_to_write); | |
127e670a | 214 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 215 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 216 | return 0; |
d3baf95d JK |
217 | |
218 | skip_write: | |
219 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
220 | return 0; | |
127e670a JK |
221 | } |
222 | ||
223 | long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, | |
224 | long nr_to_write) | |
225 | { | |
9df27d98 | 226 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
227 | pgoff_t index = 0, end = LONG_MAX; |
228 | struct pagevec pvec; | |
229 | long nwritten = 0; | |
230 | struct writeback_control wbc = { | |
231 | .for_reclaim = 0, | |
232 | }; | |
233 | ||
234 | pagevec_init(&pvec, 0); | |
235 | ||
236 | while (index <= end) { | |
237 | int i, nr_pages; | |
238 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, | |
239 | PAGECACHE_TAG_DIRTY, | |
240 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
cfb271d4 | 241 | if (unlikely(nr_pages == 0)) |
127e670a JK |
242 | break; |
243 | ||
244 | for (i = 0; i < nr_pages; i++) { | |
245 | struct page *page = pvec.pages[i]; | |
203681f6 | 246 | |
127e670a | 247 | lock_page(page); |
203681f6 JK |
248 | |
249 | if (unlikely(page->mapping != mapping)) { | |
250 | continue_unlock: | |
251 | unlock_page(page); | |
252 | continue; | |
253 | } | |
254 | if (!PageDirty(page)) { | |
255 | /* someone wrote it for us */ | |
256 | goto continue_unlock; | |
257 | } | |
258 | ||
259 | if (!clear_page_dirty_for_io(page)) | |
260 | goto continue_unlock; | |
261 | ||
577e3495 JK |
262 | if (f2fs_write_meta_page(page, &wbc)) { |
263 | unlock_page(page); | |
264 | break; | |
265 | } | |
cfb271d4 CY |
266 | nwritten++; |
267 | if (unlikely(nwritten >= nr_to_write)) | |
127e670a JK |
268 | break; |
269 | } | |
270 | pagevec_release(&pvec); | |
271 | cond_resched(); | |
272 | } | |
273 | ||
274 | if (nwritten) | |
458e6197 | 275 | f2fs_submit_merged_bio(sbi, type, WRITE); |
127e670a JK |
276 | |
277 | return nwritten; | |
278 | } | |
279 | ||
280 | static int f2fs_set_meta_page_dirty(struct page *page) | |
281 | { | |
26c6b887 JK |
282 | trace_f2fs_set_page_dirty(page, META); |
283 | ||
127e670a JK |
284 | SetPageUptodate(page); |
285 | if (!PageDirty(page)) { | |
286 | __set_page_dirty_nobuffers(page); | |
4081363f | 287 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); |
127e670a JK |
288 | return 1; |
289 | } | |
290 | return 0; | |
291 | } | |
292 | ||
293 | const struct address_space_operations f2fs_meta_aops = { | |
294 | .writepage = f2fs_write_meta_page, | |
295 | .writepages = f2fs_write_meta_pages, | |
296 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
297 | }; | |
298 | ||
6451e041 | 299 | static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 300 | { |
67298804 | 301 | struct inode_management *im = &sbi->im[type]; |
39efac41 JK |
302 | struct ino_entry *e; |
303 | retry: | |
67298804 | 304 | spin_lock(&im->ino_lock); |
39efac41 | 305 | |
67298804 | 306 | e = radix_tree_lookup(&im->ino_root, ino); |
39efac41 JK |
307 | if (!e) { |
308 | e = kmem_cache_alloc(ino_entry_slab, GFP_ATOMIC); | |
309 | if (!e) { | |
67298804 | 310 | spin_unlock(&im->ino_lock); |
39efac41 | 311 | goto retry; |
953e6cc6 | 312 | } |
67298804 CY |
313 | if (radix_tree_insert(&im->ino_root, ino, e)) { |
314 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
315 | kmem_cache_free(ino_entry_slab, e); |
316 | goto retry; | |
317 | } | |
318 | memset(e, 0, sizeof(struct ino_entry)); | |
319 | e->ino = ino; | |
953e6cc6 | 320 | |
67298804 | 321 | list_add_tail(&e->list, &im->ino_list); |
8c402946 | 322 | if (type != ORPHAN_INO) |
67298804 | 323 | im->ino_num++; |
39efac41 | 324 | } |
67298804 | 325 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
326 | } |
327 | ||
6451e041 | 328 | static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 329 | { |
67298804 | 330 | struct inode_management *im = &sbi->im[type]; |
6451e041 | 331 | struct ino_entry *e; |
953e6cc6 | 332 | |
67298804 CY |
333 | spin_lock(&im->ino_lock); |
334 | e = radix_tree_lookup(&im->ino_root, ino); | |
39efac41 JK |
335 | if (e) { |
336 | list_del(&e->list); | |
67298804 CY |
337 | radix_tree_delete(&im->ino_root, ino); |
338 | im->ino_num--; | |
339 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
340 | kmem_cache_free(ino_entry_slab, e); |
341 | return; | |
953e6cc6 | 342 | } |
67298804 | 343 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
344 | } |
345 | ||
fff04f90 JK |
346 | void add_dirty_inode(struct f2fs_sb_info *sbi, nid_t ino, int type) |
347 | { | |
348 | /* add new dirty ino entry into list */ | |
349 | __add_ino_entry(sbi, ino, type); | |
350 | } | |
351 | ||
352 | void remove_dirty_inode(struct f2fs_sb_info *sbi, nid_t ino, int type) | |
353 | { | |
354 | /* remove dirty ino entry from list */ | |
355 | __remove_ino_entry(sbi, ino, type); | |
356 | } | |
357 | ||
358 | /* mode should be APPEND_INO or UPDATE_INO */ | |
359 | bool exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode) | |
360 | { | |
67298804 | 361 | struct inode_management *im = &sbi->im[mode]; |
fff04f90 | 362 | struct ino_entry *e; |
67298804 CY |
363 | |
364 | spin_lock(&im->ino_lock); | |
365 | e = radix_tree_lookup(&im->ino_root, ino); | |
366 | spin_unlock(&im->ino_lock); | |
fff04f90 JK |
367 | return e ? true : false; |
368 | } | |
369 | ||
6f12ac25 | 370 | void release_dirty_inode(struct f2fs_sb_info *sbi) |
fff04f90 JK |
371 | { |
372 | struct ino_entry *e, *tmp; | |
373 | int i; | |
374 | ||
375 | for (i = APPEND_INO; i <= UPDATE_INO; i++) { | |
67298804 CY |
376 | struct inode_management *im = &sbi->im[i]; |
377 | ||
378 | spin_lock(&im->ino_lock); | |
379 | list_for_each_entry_safe(e, tmp, &im->ino_list, list) { | |
fff04f90 | 380 | list_del(&e->list); |
67298804 | 381 | radix_tree_delete(&im->ino_root, e->ino); |
fff04f90 | 382 | kmem_cache_free(ino_entry_slab, e); |
67298804 | 383 | im->ino_num--; |
fff04f90 | 384 | } |
67298804 | 385 | spin_unlock(&im->ino_lock); |
fff04f90 JK |
386 | } |
387 | } | |
388 | ||
cbd56e7d | 389 | int acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 390 | { |
67298804 | 391 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a JK |
392 | int err = 0; |
393 | ||
67298804 CY |
394 | spin_lock(&im->ino_lock); |
395 | if (unlikely(im->ino_num >= sbi->max_orphans)) | |
127e670a | 396 | err = -ENOSPC; |
cbd56e7d | 397 | else |
67298804 CY |
398 | im->ino_num++; |
399 | spin_unlock(&im->ino_lock); | |
0d47c1ad | 400 | |
127e670a JK |
401 | return err; |
402 | } | |
403 | ||
cbd56e7d JK |
404 | void release_orphan_inode(struct f2fs_sb_info *sbi) |
405 | { | |
67298804 CY |
406 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
407 | ||
408 | spin_lock(&im->ino_lock); | |
409 | f2fs_bug_on(sbi, im->ino_num == 0); | |
410 | im->ino_num--; | |
411 | spin_unlock(&im->ino_lock); | |
cbd56e7d JK |
412 | } |
413 | ||
127e670a JK |
414 | void add_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
415 | { | |
39efac41 | 416 | /* add new orphan ino entry into list */ |
6451e041 | 417 | __add_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
418 | } |
419 | ||
420 | void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
421 | { | |
953e6cc6 | 422 | /* remove orphan entry from orphan list */ |
6451e041 | 423 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
424 | } |
425 | ||
426 | static void recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
427 | { | |
428 | struct inode *inode = f2fs_iget(sbi->sb, ino); | |
9850cf4a | 429 | f2fs_bug_on(sbi, IS_ERR(inode)); |
127e670a JK |
430 | clear_nlink(inode); |
431 | ||
432 | /* truncate all the data during iput */ | |
433 | iput(inode); | |
434 | } | |
435 | ||
8f99a946 | 436 | void recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a JK |
437 | { |
438 | block_t start_blk, orphan_blkaddr, i, j; | |
439 | ||
25ca923b | 440 | if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG)) |
8f99a946 | 441 | return; |
127e670a | 442 | |
aabe5136 | 443 | sbi->por_doing = true; |
1dbe4152 CL |
444 | |
445 | start_blk = __start_cp_addr(sbi) + 1 + | |
446 | le32_to_cpu(F2FS_RAW_SUPER(sbi)->cp_payload); | |
127e670a JK |
447 | orphan_blkaddr = __start_sum_addr(sbi) - 1; |
448 | ||
662befda CY |
449 | ra_meta_pages(sbi, start_blk, orphan_blkaddr, META_CP); |
450 | ||
127e670a JK |
451 | for (i = 0; i < orphan_blkaddr; i++) { |
452 | struct page *page = get_meta_page(sbi, start_blk + i); | |
453 | struct f2fs_orphan_block *orphan_blk; | |
454 | ||
455 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); | |
456 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
457 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
458 | recover_orphan_inode(sbi, ino); | |
459 | } | |
460 | f2fs_put_page(page, 1); | |
461 | } | |
462 | /* clear Orphan Flag */ | |
25ca923b | 463 | clear_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG); |
aabe5136 | 464 | sbi->por_doing = false; |
8f99a946 | 465 | return; |
127e670a JK |
466 | } |
467 | ||
468 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
469 | { | |
502c6e0b | 470 | struct list_head *head; |
127e670a | 471 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 472 | unsigned int nentries = 0; |
4531929e | 473 | unsigned short index; |
8c402946 | 474 | unsigned short orphan_blocks; |
4531929e | 475 | struct page *page = NULL; |
6451e041 | 476 | struct ino_entry *orphan = NULL; |
67298804 | 477 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 478 | |
67298804 | 479 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 480 | |
4531929e | 481 | for (index = 0; index < orphan_blocks; index++) |
63f5384c | 482 | grab_meta_page(sbi, start_blk + index); |
127e670a | 483 | |
4531929e | 484 | index = 1; |
67298804 CY |
485 | spin_lock(&im->ino_lock); |
486 | head = &im->ino_list; | |
127e670a JK |
487 | |
488 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
489 | list_for_each_entry(orphan, head, list) { |
490 | if (!page) { | |
63f5384c | 491 | page = find_get_page(META_MAPPING(sbi), start_blk++); |
9850cf4a | 492 | f2fs_bug_on(sbi, !page); |
502c6e0b GZ |
493 | orphan_blk = |
494 | (struct f2fs_orphan_block *)page_address(page); | |
495 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
63f5384c | 496 | f2fs_put_page(page, 0); |
502c6e0b | 497 | } |
127e670a | 498 | |
36795567 | 499 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 500 | |
36795567 | 501 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
502 | /* |
503 | * an orphan block is full of 1020 entries, | |
504 | * then we need to flush current orphan blocks | |
505 | * and bring another one in memory | |
506 | */ | |
507 | orphan_blk->blk_addr = cpu_to_le16(index); | |
508 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
509 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
510 | set_page_dirty(page); | |
511 | f2fs_put_page(page, 1); | |
512 | index++; | |
127e670a JK |
513 | nentries = 0; |
514 | page = NULL; | |
515 | } | |
502c6e0b | 516 | } |
127e670a | 517 | |
502c6e0b GZ |
518 | if (page) { |
519 | orphan_blk->blk_addr = cpu_to_le16(index); | |
520 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
521 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
522 | set_page_dirty(page); | |
523 | f2fs_put_page(page, 1); | |
127e670a | 524 | } |
502c6e0b | 525 | |
67298804 | 526 | spin_unlock(&im->ino_lock); |
127e670a JK |
527 | } |
528 | ||
529 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, | |
530 | block_t cp_addr, unsigned long long *version) | |
531 | { | |
532 | struct page *cp_page_1, *cp_page_2 = NULL; | |
533 | unsigned long blk_size = sbi->blocksize; | |
534 | struct f2fs_checkpoint *cp_block; | |
535 | unsigned long long cur_version = 0, pre_version = 0; | |
127e670a | 536 | size_t crc_offset; |
7e586fa0 | 537 | __u32 crc = 0; |
127e670a JK |
538 | |
539 | /* Read the 1st cp block in this CP pack */ | |
540 | cp_page_1 = get_meta_page(sbi, cp_addr); | |
541 | ||
542 | /* get the version number */ | |
543 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_1); | |
544 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
545 | if (crc_offset >= blk_size) | |
546 | goto invalid_cp1; | |
547 | ||
7e586fa0 | 548 | crc = le32_to_cpu(*((__u32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
549 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
550 | goto invalid_cp1; | |
551 | ||
d71b5564 | 552 | pre_version = cur_cp_version(cp_block); |
127e670a JK |
553 | |
554 | /* Read the 2nd cp block in this CP pack */ | |
25ca923b | 555 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
127e670a JK |
556 | cp_page_2 = get_meta_page(sbi, cp_addr); |
557 | ||
558 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_2); | |
559 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
560 | if (crc_offset >= blk_size) | |
561 | goto invalid_cp2; | |
562 | ||
7e586fa0 | 563 | crc = le32_to_cpu(*((__u32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
564 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
565 | goto invalid_cp2; | |
566 | ||
d71b5564 | 567 | cur_version = cur_cp_version(cp_block); |
127e670a JK |
568 | |
569 | if (cur_version == pre_version) { | |
570 | *version = cur_version; | |
571 | f2fs_put_page(cp_page_2, 1); | |
572 | return cp_page_1; | |
573 | } | |
574 | invalid_cp2: | |
575 | f2fs_put_page(cp_page_2, 1); | |
576 | invalid_cp1: | |
577 | f2fs_put_page(cp_page_1, 1); | |
578 | return NULL; | |
579 | } | |
580 | ||
581 | int get_valid_checkpoint(struct f2fs_sb_info *sbi) | |
582 | { | |
583 | struct f2fs_checkpoint *cp_block; | |
584 | struct f2fs_super_block *fsb = sbi->raw_super; | |
585 | struct page *cp1, *cp2, *cur_page; | |
586 | unsigned long blk_size = sbi->blocksize; | |
587 | unsigned long long cp1_version = 0, cp2_version = 0; | |
588 | unsigned long long cp_start_blk_no; | |
1dbe4152 CL |
589 | unsigned int cp_blks = 1 + le32_to_cpu(F2FS_RAW_SUPER(sbi)->cp_payload); |
590 | block_t cp_blk_no; | |
591 | int i; | |
127e670a | 592 | |
1dbe4152 | 593 | sbi->ckpt = kzalloc(cp_blks * blk_size, GFP_KERNEL); |
127e670a JK |
594 | if (!sbi->ckpt) |
595 | return -ENOMEM; | |
596 | /* | |
597 | * Finding out valid cp block involves read both | |
598 | * sets( cp pack1 and cp pack 2) | |
599 | */ | |
600 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
601 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
602 | ||
603 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
604 | cp_start_blk_no += ((unsigned long long)1) << |
605 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
606 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
607 | ||
608 | if (cp1 && cp2) { | |
609 | if (ver_after(cp2_version, cp1_version)) | |
610 | cur_page = cp2; | |
611 | else | |
612 | cur_page = cp1; | |
613 | } else if (cp1) { | |
614 | cur_page = cp1; | |
615 | } else if (cp2) { | |
616 | cur_page = cp2; | |
617 | } else { | |
618 | goto fail_no_cp; | |
619 | } | |
620 | ||
621 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
622 | memcpy(sbi->ckpt, cp_block, blk_size); | |
623 | ||
1dbe4152 CL |
624 | if (cp_blks <= 1) |
625 | goto done; | |
626 | ||
627 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
628 | if (cur_page == cp2) | |
629 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
630 | ||
631 | for (i = 1; i < cp_blks; i++) { | |
632 | void *sit_bitmap_ptr; | |
633 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
634 | ||
635 | cur_page = get_meta_page(sbi, cp_blk_no + i); | |
636 | sit_bitmap_ptr = page_address(cur_page); | |
637 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
638 | f2fs_put_page(cur_page, 1); | |
639 | } | |
640 | done: | |
127e670a JK |
641 | f2fs_put_page(cp1, 1); |
642 | f2fs_put_page(cp2, 1); | |
643 | return 0; | |
644 | ||
645 | fail_no_cp: | |
646 | kfree(sbi->ckpt); | |
647 | return -EINVAL; | |
648 | } | |
649 | ||
5deb8267 | 650 | static int __add_dirty_inode(struct inode *inode, struct dir_inode_entry *new) |
127e670a | 651 | { |
4081363f | 652 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
127e670a | 653 | |
ed57c27f JK |
654 | if (is_inode_flag_set(F2FS_I(inode), FI_DIRTY_DIR)) |
655 | return -EEXIST; | |
2d7b822a | 656 | |
ed57c27f JK |
657 | set_inode_flag(F2FS_I(inode), FI_DIRTY_DIR); |
658 | F2FS_I(inode)->dirty_dir = new; | |
659 | list_add_tail(&new->list, &sbi->dir_inode_list); | |
dcdfff65 | 660 | stat_inc_dirty_dir(sbi); |
5deb8267 JK |
661 | return 0; |
662 | } | |
663 | ||
a7ffdbe2 | 664 | void update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 665 | { |
4081363f | 666 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
5deb8267 | 667 | struct dir_inode_entry *new; |
cf0ee0f0 | 668 | int ret = 0; |
5deb8267 | 669 | |
a7ffdbe2 | 670 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode)) |
127e670a | 671 | return; |
7bd59381 | 672 | |
a7ffdbe2 JK |
673 | if (!S_ISDIR(inode->i_mode)) { |
674 | inode_inc_dirty_pages(inode); | |
675 | goto out; | |
676 | } | |
677 | ||
7bd59381 | 678 | new = f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS); |
127e670a JK |
679 | new->inode = inode; |
680 | INIT_LIST_HEAD(&new->list); | |
681 | ||
682 | spin_lock(&sbi->dir_inode_lock); | |
cf0ee0f0 | 683 | ret = __add_dirty_inode(inode, new); |
a7ffdbe2 | 684 | inode_inc_dirty_pages(inode); |
5deb8267 | 685 | spin_unlock(&sbi->dir_inode_lock); |
cf0ee0f0 CY |
686 | |
687 | if (ret) | |
688 | kmem_cache_free(inode_entry_slab, new); | |
a7ffdbe2 JK |
689 | out: |
690 | SetPagePrivate(page); | |
5deb8267 JK |
691 | } |
692 | ||
693 | void add_dirty_dir_inode(struct inode *inode) | |
694 | { | |
4081363f | 695 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
7bd59381 GZ |
696 | struct dir_inode_entry *new = |
697 | f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS); | |
cf0ee0f0 | 698 | int ret = 0; |
7bd59381 | 699 | |
5deb8267 JK |
700 | new->inode = inode; |
701 | INIT_LIST_HEAD(&new->list); | |
127e670a | 702 | |
5deb8267 | 703 | spin_lock(&sbi->dir_inode_lock); |
cf0ee0f0 | 704 | ret = __add_dirty_inode(inode, new); |
127e670a | 705 | spin_unlock(&sbi->dir_inode_lock); |
cf0ee0f0 CY |
706 | |
707 | if (ret) | |
708 | kmem_cache_free(inode_entry_slab, new); | |
127e670a JK |
709 | } |
710 | ||
711 | void remove_dirty_dir_inode(struct inode *inode) | |
712 | { | |
4081363f | 713 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
2d7b822a | 714 | struct dir_inode_entry *entry; |
127e670a JK |
715 | |
716 | if (!S_ISDIR(inode->i_mode)) | |
717 | return; | |
718 | ||
719 | spin_lock(&sbi->dir_inode_lock); | |
a7ffdbe2 | 720 | if (get_dirty_pages(inode) || |
ed57c27f | 721 | !is_inode_flag_set(F2FS_I(inode), FI_DIRTY_DIR)) { |
3b10b1fd JK |
722 | spin_unlock(&sbi->dir_inode_lock); |
723 | return; | |
724 | } | |
127e670a | 725 | |
ed57c27f JK |
726 | entry = F2FS_I(inode)->dirty_dir; |
727 | list_del(&entry->list); | |
728 | F2FS_I(inode)->dirty_dir = NULL; | |
729 | clear_inode_flag(F2FS_I(inode), FI_DIRTY_DIR); | |
730 | stat_dec_dirty_dir(sbi); | |
127e670a | 731 | spin_unlock(&sbi->dir_inode_lock); |
ed57c27f | 732 | kmem_cache_free(inode_entry_slab, entry); |
74d0b917 JK |
733 | |
734 | /* Only from the recovery routine */ | |
afc3eda2 JK |
735 | if (is_inode_flag_set(F2FS_I(inode), FI_DELAY_IPUT)) { |
736 | clear_inode_flag(F2FS_I(inode), FI_DELAY_IPUT); | |
74d0b917 | 737 | iput(inode); |
afc3eda2 | 738 | } |
74d0b917 JK |
739 | } |
740 | ||
127e670a JK |
741 | void sync_dirty_dir_inodes(struct f2fs_sb_info *sbi) |
742 | { | |
ce3b7d80 | 743 | struct list_head *head; |
127e670a JK |
744 | struct dir_inode_entry *entry; |
745 | struct inode *inode; | |
746 | retry: | |
af41d3ee JK |
747 | if (unlikely(f2fs_cp_error(sbi))) |
748 | return; | |
749 | ||
127e670a | 750 | spin_lock(&sbi->dir_inode_lock); |
ce3b7d80 GZ |
751 | |
752 | head = &sbi->dir_inode_list; | |
127e670a JK |
753 | if (list_empty(head)) { |
754 | spin_unlock(&sbi->dir_inode_lock); | |
755 | return; | |
756 | } | |
757 | entry = list_entry(head->next, struct dir_inode_entry, list); | |
758 | inode = igrab(entry->inode); | |
759 | spin_unlock(&sbi->dir_inode_lock); | |
760 | if (inode) { | |
87d6f890 | 761 | filemap_fdatawrite(inode->i_mapping); |
127e670a JK |
762 | iput(inode); |
763 | } else { | |
764 | /* | |
765 | * We should submit bio, since it exists several | |
766 | * wribacking dentry pages in the freeing inode. | |
767 | */ | |
458e6197 | 768 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
127e670a JK |
769 | } |
770 | goto retry; | |
771 | } | |
772 | ||
0a8165d7 | 773 | /* |
127e670a JK |
774 | * Freeze all the FS-operations for checkpoint. |
775 | */ | |
cf779cab | 776 | static int block_operations(struct f2fs_sb_info *sbi) |
127e670a | 777 | { |
127e670a JK |
778 | struct writeback_control wbc = { |
779 | .sync_mode = WB_SYNC_ALL, | |
780 | .nr_to_write = LONG_MAX, | |
781 | .for_reclaim = 0, | |
782 | }; | |
c718379b | 783 | struct blk_plug plug; |
cf779cab | 784 | int err = 0; |
c718379b JK |
785 | |
786 | blk_start_plug(&plug); | |
787 | ||
39936837 | 788 | retry_flush_dents: |
e479556b | 789 | f2fs_lock_all(sbi); |
127e670a | 790 | /* write all the dirty dentry pages */ |
127e670a | 791 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 792 | f2fs_unlock_all(sbi); |
39936837 | 793 | sync_dirty_dir_inodes(sbi); |
cf779cab JK |
794 | if (unlikely(f2fs_cp_error(sbi))) { |
795 | err = -EIO; | |
796 | goto out; | |
797 | } | |
39936837 | 798 | goto retry_flush_dents; |
127e670a JK |
799 | } |
800 | ||
127e670a | 801 | /* |
e1c42045 | 802 | * POR: we should ensure that there are no dirty node pages |
127e670a JK |
803 | * until finishing nat/sit flush. |
804 | */ | |
39936837 | 805 | retry_flush_nodes: |
b3582c68 | 806 | down_write(&sbi->node_write); |
127e670a JK |
807 | |
808 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
b3582c68 | 809 | up_write(&sbi->node_write); |
39936837 | 810 | sync_node_pages(sbi, 0, &wbc); |
cf779cab JK |
811 | if (unlikely(f2fs_cp_error(sbi))) { |
812 | f2fs_unlock_all(sbi); | |
813 | err = -EIO; | |
814 | goto out; | |
815 | } | |
39936837 | 816 | goto retry_flush_nodes; |
127e670a | 817 | } |
cf779cab | 818 | out: |
c718379b | 819 | blk_finish_plug(&plug); |
cf779cab | 820 | return err; |
127e670a JK |
821 | } |
822 | ||
823 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
824 | { | |
b3582c68 | 825 | up_write(&sbi->node_write); |
e479556b | 826 | f2fs_unlock_all(sbi); |
127e670a JK |
827 | } |
828 | ||
fb51b5ef CL |
829 | static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi) |
830 | { | |
831 | DEFINE_WAIT(wait); | |
832 | ||
833 | for (;;) { | |
834 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
835 | ||
836 | if (!get_pages(sbi, F2FS_WRITEBACK)) | |
837 | break; | |
838 | ||
839 | io_schedule(); | |
840 | } | |
841 | finish_wait(&sbi->cp_wait, &wait); | |
842 | } | |
843 | ||
75ab4cb8 | 844 | static void do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
845 | { |
846 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
cf2271e7 | 847 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); |
77041823 | 848 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
67298804 | 849 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; |
77041823 | 850 | nid_t last_nid = nm_i->next_scan_nid; |
127e670a JK |
851 | block_t start_blk; |
852 | struct page *cp_page; | |
853 | unsigned int data_sum_blocks, orphan_blocks; | |
7e586fa0 | 854 | __u32 crc32 = 0; |
127e670a | 855 | void *kaddr; |
127e670a | 856 | int i; |
1dbe4152 | 857 | int cp_payload_blks = le32_to_cpu(F2FS_RAW_SUPER(sbi)->cp_payload); |
127e670a | 858 | |
1e87a78d JK |
859 | /* |
860 | * This avoids to conduct wrong roll-forward operations and uses | |
861 | * metapages, so should be called prior to sync_meta_pages below. | |
862 | */ | |
cf2271e7 | 863 | discard_next_dnode(sbi, NEXT_FREE_BLKADDR(sbi, curseg)); |
127e670a JK |
864 | |
865 | /* Flush all the NAT/SIT pages */ | |
cf779cab | 866 | while (get_pages(sbi, F2FS_DIRTY_META)) { |
127e670a | 867 | sync_meta_pages(sbi, META, LONG_MAX); |
cf779cab JK |
868 | if (unlikely(f2fs_cp_error(sbi))) |
869 | return; | |
870 | } | |
127e670a JK |
871 | |
872 | next_free_nid(sbi, &last_nid); | |
873 | ||
874 | /* | |
875 | * modify checkpoint | |
876 | * version number is already updated | |
877 | */ | |
878 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi)); | |
879 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
880 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); | |
b5b82205 | 881 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
882 | ckpt->cur_node_segno[i] = |
883 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
884 | ckpt->cur_node_blkoff[i] = | |
885 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
886 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
887 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
888 | } | |
b5b82205 | 889 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
890 | ckpt->cur_data_segno[i] = |
891 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
892 | ckpt->cur_data_blkoff[i] = | |
893 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
894 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
895 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
896 | } | |
897 | ||
898 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
899 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
900 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
901 | ||
902 | /* 2 cp + n data seg summary + orphan inode blocks */ | |
903 | data_sum_blocks = npages_for_summary_flush(sbi); | |
b5b82205 | 904 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
25ca923b | 905 | set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 906 | else |
25ca923b | 907 | clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 908 | |
67298804 | 909 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
910 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
911 | orphan_blocks); | |
127e670a | 912 | |
75ab4cb8 | 913 | if (cpc->reason == CP_UMOUNT) { |
25ca923b | 914 | set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); |
b5b82205 | 915 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
916 | cp_payload_blks + data_sum_blocks + |
917 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
127e670a | 918 | } else { |
25ca923b | 919 | clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); |
b5b82205 | 920 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS + |
1dbe4152 CL |
921 | cp_payload_blks + data_sum_blocks + |
922 | orphan_blocks); | |
127e670a JK |
923 | } |
924 | ||
67298804 | 925 | if (orphan_num) |
25ca923b | 926 | set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 927 | else |
25ca923b | 928 | clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 929 | |
2ae4c673 JK |
930 | if (sbi->need_fsck) |
931 | set_ckpt_flags(ckpt, CP_FSCK_FLAG); | |
932 | ||
127e670a JK |
933 | /* update SIT/NAT bitmap */ |
934 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
935 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
936 | ||
937 | crc32 = f2fs_crc32(ckpt, le32_to_cpu(ckpt->checksum_offset)); | |
7e586fa0 JK |
938 | *((__le32 *)((unsigned char *)ckpt + |
939 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
940 | = cpu_to_le32(crc32); |
941 | ||
942 | start_blk = __start_cp_addr(sbi); | |
943 | ||
944 | /* write out checkpoint buffer at block 0 */ | |
945 | cp_page = grab_meta_page(sbi, start_blk++); | |
946 | kaddr = page_address(cp_page); | |
947 | memcpy(kaddr, ckpt, (1 << sbi->log_blocksize)); | |
948 | set_page_dirty(cp_page); | |
949 | f2fs_put_page(cp_page, 1); | |
950 | ||
1dbe4152 CL |
951 | for (i = 1; i < 1 + cp_payload_blks; i++) { |
952 | cp_page = grab_meta_page(sbi, start_blk++); | |
953 | kaddr = page_address(cp_page); | |
954 | memcpy(kaddr, (char *)ckpt + i * F2FS_BLKSIZE, | |
955 | (1 << sbi->log_blocksize)); | |
956 | set_page_dirty(cp_page); | |
957 | f2fs_put_page(cp_page, 1); | |
958 | } | |
959 | ||
67298804 | 960 | if (orphan_num) { |
127e670a JK |
961 | write_orphan_inodes(sbi, start_blk); |
962 | start_blk += orphan_blocks; | |
963 | } | |
964 | ||
965 | write_data_summaries(sbi, start_blk); | |
966 | start_blk += data_sum_blocks; | |
75ab4cb8 | 967 | if (cpc->reason == CP_UMOUNT) { |
127e670a JK |
968 | write_node_summaries(sbi, start_blk); |
969 | start_blk += NR_CURSEG_NODE_TYPE; | |
970 | } | |
971 | ||
972 | /* writeout checkpoint block */ | |
973 | cp_page = grab_meta_page(sbi, start_blk); | |
974 | kaddr = page_address(cp_page); | |
975 | memcpy(kaddr, ckpt, (1 << sbi->log_blocksize)); | |
976 | set_page_dirty(cp_page); | |
977 | f2fs_put_page(cp_page, 1); | |
978 | ||
979 | /* wait for previous submitted node/meta pages writeback */ | |
fb51b5ef | 980 | wait_on_all_pages_writeback(sbi); |
127e670a | 981 | |
cf779cab JK |
982 | if (unlikely(f2fs_cp_error(sbi))) |
983 | return; | |
984 | ||
4ef51a8f | 985 | filemap_fdatawait_range(NODE_MAPPING(sbi), 0, LONG_MAX); |
9df27d98 | 986 | filemap_fdatawait_range(META_MAPPING(sbi), 0, LONG_MAX); |
127e670a JK |
987 | |
988 | /* update user_block_counts */ | |
989 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
990 | sbi->alloc_valid_block_count = 0; | |
991 | ||
992 | /* Here, we only have one bio having CP pack */ | |
577e3495 | 993 | sync_meta_pages(sbi, META_FLUSH, LONG_MAX); |
127e670a | 994 | |
6a8f8ca5 JK |
995 | /* wait for previous submitted meta pages writeback */ |
996 | wait_on_all_pages_writeback(sbi); | |
997 | ||
cf779cab JK |
998 | release_dirty_inode(sbi); |
999 | ||
1000 | if (unlikely(f2fs_cp_error(sbi))) | |
1001 | return; | |
1002 | ||
1003 | clear_prefree_segments(sbi); | |
1004 | F2FS_RESET_SB_DIRT(sbi); | |
127e670a JK |
1005 | } |
1006 | ||
0a8165d7 | 1007 | /* |
e1c42045 | 1008 | * We guarantee that this checkpoint procedure will not fail. |
127e670a | 1009 | */ |
75ab4cb8 | 1010 | void write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1011 | { |
1012 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1013 | unsigned long long ckpt_ver; | |
1014 | ||
75ab4cb8 | 1015 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); |
2af4bd6c | 1016 | |
43727527 | 1017 | mutex_lock(&sbi->cp_mutex); |
8501017e | 1018 | |
4b2fecc8 | 1019 | if (!sbi->s_dirty && cpc->reason != CP_DISCARD) |
8501017e | 1020 | goto out; |
cf779cab JK |
1021 | if (unlikely(f2fs_cp_error(sbi))) |
1022 | goto out; | |
1023 | if (block_operations(sbi)) | |
1024 | goto out; | |
127e670a | 1025 | |
75ab4cb8 | 1026 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1027 | |
458e6197 JK |
1028 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
1029 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
1030 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
127e670a JK |
1031 | |
1032 | /* | |
1033 | * update checkpoint pack index | |
1034 | * Increase the version number so that | |
1035 | * SIT entries and seg summaries are written at correct place | |
1036 | */ | |
d71b5564 | 1037 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1038 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1039 | ||
1040 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
1041 | flush_nat_entries(sbi); | |
4b2fecc8 | 1042 | flush_sit_entries(sbi, cpc); |
127e670a | 1043 | |
127e670a | 1044 | /* unlock all the fs_lock[] in do_checkpoint() */ |
75ab4cb8 | 1045 | do_checkpoint(sbi, cpc); |
127e670a JK |
1046 | |
1047 | unblock_operations(sbi); | |
942e0be6 | 1048 | stat_inc_cp_count(sbi->stat_info); |
8501017e JK |
1049 | out: |
1050 | mutex_unlock(&sbi->cp_mutex); | |
75ab4cb8 | 1051 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
127e670a JK |
1052 | } |
1053 | ||
6451e041 | 1054 | void init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1055 | { |
6451e041 JK |
1056 | int i; |
1057 | ||
1058 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1059 | struct inode_management *im = &sbi->im[i]; |
1060 | ||
1061 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1062 | spin_lock_init(&im->ino_lock); | |
1063 | INIT_LIST_HEAD(&im->ino_list); | |
1064 | im->ino_num = 0; | |
6451e041 JK |
1065 | } |
1066 | ||
0d47c1ad GZ |
1067 | /* |
1068 | * considering 512 blocks in a segment 8 blocks are needed for cp | |
1069 | * and log segment summaries. Remaining blocks are used to keep | |
1070 | * orphan entries with the limitation one reserved segment | |
1071 | * for cp pack we can have max 1020*504 orphan entries | |
1072 | */ | |
b5b82205 CY |
1073 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
1074 | NR_CURSEG_TYPE) * F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
1075 | } |
1076 | ||
6e6093a8 | 1077 | int __init create_checkpoint_caches(void) |
127e670a | 1078 | { |
6451e041 JK |
1079 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1080 | sizeof(struct ino_entry)); | |
1081 | if (!ino_entry_slab) | |
127e670a JK |
1082 | return -ENOMEM; |
1083 | inode_entry_slab = f2fs_kmem_cache_create("f2fs_dirty_dir_entry", | |
e8512d2e | 1084 | sizeof(struct dir_inode_entry)); |
6bacf52f | 1085 | if (!inode_entry_slab) { |
6451e041 | 1086 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1087 | return -ENOMEM; |
1088 | } | |
1089 | return 0; | |
1090 | } | |
1091 | ||
1092 | void destroy_checkpoint_caches(void) | |
1093 | { | |
6451e041 | 1094 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1095 | kmem_cache_destroy(inode_entry_slab); |
1096 | } |