Commit | Line | Data |
---|---|---|
0a8165d7 | 1 | /* |
127e670a JK |
2 | * fs/f2fs/checkpoint.c |
3 | * | |
4 | * Copyright (c) 2012 Samsung Electronics Co., Ltd. | |
5 | * http://www.samsung.com/ | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License version 2 as | |
9 | * published by the Free Software Foundation. | |
10 | */ | |
11 | #include <linux/fs.h> | |
12 | #include <linux/bio.h> | |
13 | #include <linux/mpage.h> | |
14 | #include <linux/writeback.h> | |
15 | #include <linux/blkdev.h> | |
16 | #include <linux/f2fs_fs.h> | |
17 | #include <linux/pagevec.h> | |
18 | #include <linux/swap.h> | |
19 | ||
20 | #include "f2fs.h" | |
21 | #include "node.h" | |
22 | #include "segment.h" | |
9e4ded3f | 23 | #include "trace.h" |
2af4bd6c | 24 | #include <trace/events/f2fs.h> |
127e670a | 25 | |
6451e041 | 26 | static struct kmem_cache *ino_entry_slab; |
06292073 | 27 | struct kmem_cache *inode_entry_slab; |
127e670a | 28 | |
0a8165d7 | 29 | /* |
127e670a JK |
30 | * We guarantee no failure on the returned page. |
31 | */ | |
32 | struct page *grab_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
33 | { | |
9df27d98 | 34 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
35 | struct page *page = NULL; |
36 | repeat: | |
bde44686 | 37 | page = grab_cache_page(mapping, index); |
127e670a JK |
38 | if (!page) { |
39 | cond_resched(); | |
40 | goto repeat; | |
41 | } | |
bde44686 | 42 | f2fs_wait_on_page_writeback(page, META); |
127e670a JK |
43 | SetPageUptodate(page); |
44 | return page; | |
45 | } | |
46 | ||
0a8165d7 | 47 | /* |
127e670a JK |
48 | * We guarantee no failure on the returned page. |
49 | */ | |
50 | struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
51 | { | |
9df27d98 | 52 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a | 53 | struct page *page; |
cf04e8eb | 54 | struct f2fs_io_info fio = { |
05ca3632 | 55 | .sbi = sbi, |
cf04e8eb JK |
56 | .type = META, |
57 | .rw = READ_SYNC | REQ_META | REQ_PRIO, | |
58 | .blk_addr = index, | |
59 | }; | |
127e670a JK |
60 | repeat: |
61 | page = grab_cache_page(mapping, index); | |
62 | if (!page) { | |
63 | cond_resched(); | |
64 | goto repeat; | |
65 | } | |
393ff91f JK |
66 | if (PageUptodate(page)) |
67 | goto out; | |
68 | ||
05ca3632 JK |
69 | fio.page = page; |
70 | ||
71 | if (f2fs_submit_page_bio(&fio)) | |
127e670a | 72 | goto repeat; |
127e670a | 73 | |
393ff91f | 74 | lock_page(page); |
6bacf52f | 75 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
76 | f2fs_put_page(page, 1); |
77 | goto repeat; | |
78 | } | |
393ff91f | 79 | out: |
127e670a JK |
80 | return page; |
81 | } | |
82 | ||
f0c9cada | 83 | bool is_valid_blkaddr(struct f2fs_sb_info *sbi, block_t blkaddr, int type) |
662befda CY |
84 | { |
85 | switch (type) { | |
86 | case META_NAT: | |
66b00c18 | 87 | break; |
662befda | 88 | case META_SIT: |
66b00c18 CY |
89 | if (unlikely(blkaddr >= SIT_BLK_CNT(sbi))) |
90 | return false; | |
91 | break; | |
81c1a0f1 | 92 | case META_SSA: |
66b00c18 CY |
93 | if (unlikely(blkaddr >= MAIN_BLKADDR(sbi) || |
94 | blkaddr < SM_I(sbi)->ssa_blkaddr)) | |
95 | return false; | |
96 | break; | |
662befda | 97 | case META_CP: |
66b00c18 CY |
98 | if (unlikely(blkaddr >= SIT_I(sbi)->sit_base_addr || |
99 | blkaddr < __start_cp_addr(sbi))) | |
100 | return false; | |
101 | break; | |
4c521f49 | 102 | case META_POR: |
66b00c18 CY |
103 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
104 | blkaddr < MAIN_BLKADDR(sbi))) | |
105 | return false; | |
106 | break; | |
662befda CY |
107 | default: |
108 | BUG(); | |
109 | } | |
66b00c18 CY |
110 | |
111 | return true; | |
662befda CY |
112 | } |
113 | ||
114 | /* | |
81c1a0f1 | 115 | * Readahead CP/NAT/SIT/SSA pages |
662befda | 116 | */ |
4c521f49 | 117 | int ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages, int type) |
662befda CY |
118 | { |
119 | block_t prev_blk_addr = 0; | |
120 | struct page *page; | |
4c521f49 | 121 | block_t blkno = start; |
662befda | 122 | struct f2fs_io_info fio = { |
05ca3632 | 123 | .sbi = sbi, |
662befda CY |
124 | .type = META, |
125 | .rw = READ_SYNC | REQ_META | REQ_PRIO | |
126 | }; | |
127 | ||
128 | for (; nrpages-- > 0; blkno++) { | |
662befda | 129 | |
66b00c18 CY |
130 | if (!is_valid_blkaddr(sbi, blkno, type)) |
131 | goto out; | |
132 | ||
662befda CY |
133 | switch (type) { |
134 | case META_NAT: | |
66b00c18 CY |
135 | if (unlikely(blkno >= |
136 | NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))) | |
662befda | 137 | blkno = 0; |
66b00c18 | 138 | /* get nat block addr */ |
cf04e8eb | 139 | fio.blk_addr = current_nat_addr(sbi, |
662befda CY |
140 | blkno * NAT_ENTRY_PER_BLOCK); |
141 | break; | |
142 | case META_SIT: | |
143 | /* get sit block addr */ | |
cf04e8eb | 144 | fio.blk_addr = current_sit_addr(sbi, |
662befda | 145 | blkno * SIT_ENTRY_PER_BLOCK); |
cf04e8eb | 146 | if (blkno != start && prev_blk_addr + 1 != fio.blk_addr) |
662befda | 147 | goto out; |
cf04e8eb | 148 | prev_blk_addr = fio.blk_addr; |
662befda | 149 | break; |
81c1a0f1 | 150 | case META_SSA: |
662befda | 151 | case META_CP: |
4c521f49 | 152 | case META_POR: |
cf04e8eb | 153 | fio.blk_addr = blkno; |
662befda CY |
154 | break; |
155 | default: | |
156 | BUG(); | |
157 | } | |
158 | ||
cf04e8eb | 159 | page = grab_cache_page(META_MAPPING(sbi), fio.blk_addr); |
662befda CY |
160 | if (!page) |
161 | continue; | |
162 | if (PageUptodate(page)) { | |
662befda CY |
163 | f2fs_put_page(page, 1); |
164 | continue; | |
165 | } | |
166 | ||
05ca3632 JK |
167 | fio.page = page; |
168 | f2fs_submit_page_mbio(&fio); | |
662befda CY |
169 | f2fs_put_page(page, 0); |
170 | } | |
171 | out: | |
172 | f2fs_submit_merged_bio(sbi, META, READ); | |
173 | return blkno - start; | |
174 | } | |
175 | ||
635aee1f CY |
176 | void ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index) |
177 | { | |
178 | struct page *page; | |
179 | bool readahead = false; | |
180 | ||
181 | page = find_get_page(META_MAPPING(sbi), index); | |
182 | if (!page || (page && !PageUptodate(page))) | |
183 | readahead = true; | |
184 | f2fs_put_page(page, 0); | |
185 | ||
186 | if (readahead) | |
187 | ra_meta_pages(sbi, index, MAX_BIO_BLOCKS(sbi), META_POR); | |
188 | } | |
189 | ||
127e670a JK |
190 | static int f2fs_write_meta_page(struct page *page, |
191 | struct writeback_control *wbc) | |
192 | { | |
4081363f | 193 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 194 | |
ecda0de3 CY |
195 | trace_f2fs_writepage(page, META); |
196 | ||
caf0047e | 197 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
cfb271d4 | 198 | goto redirty_out; |
857dc4e0 | 199 | if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0)) |
cfb271d4 | 200 | goto redirty_out; |
1e968fdf | 201 | if (unlikely(f2fs_cp_error(sbi))) |
cf779cab | 202 | goto redirty_out; |
127e670a | 203 | |
3cb5ad15 | 204 | f2fs_wait_on_page_writeback(page, META); |
577e3495 JK |
205 | write_meta_page(sbi, page); |
206 | dec_page_count(sbi, F2FS_DIRTY_META); | |
207 | unlock_page(page); | |
857dc4e0 JK |
208 | |
209 | if (wbc->for_reclaim) | |
210 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
577e3495 | 211 | return 0; |
cfb271d4 CY |
212 | |
213 | redirty_out: | |
76f60268 | 214 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 215 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
216 | } |
217 | ||
218 | static int f2fs_write_meta_pages(struct address_space *mapping, | |
219 | struct writeback_control *wbc) | |
220 | { | |
4081363f | 221 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 222 | long diff, written; |
127e670a | 223 | |
e5748434 CY |
224 | trace_f2fs_writepages(mapping->host, wbc, META); |
225 | ||
5459aa97 | 226 | /* collect a number of dirty meta pages and write together */ |
50c8cdb3 JK |
227 | if (wbc->for_kupdate || |
228 | get_pages(sbi, F2FS_DIRTY_META) < nr_pages_to_skip(sbi, META)) | |
d3baf95d | 229 | goto skip_write; |
127e670a JK |
230 | |
231 | /* if mounting is failed, skip writing node pages */ | |
232 | mutex_lock(&sbi->cp_mutex); | |
50c8cdb3 JK |
233 | diff = nr_pages_to_write(sbi, META, wbc); |
234 | written = sync_meta_pages(sbi, META, wbc->nr_to_write); | |
127e670a | 235 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 236 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 237 | return 0; |
d3baf95d JK |
238 | |
239 | skip_write: | |
240 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
241 | return 0; | |
127e670a JK |
242 | } |
243 | ||
244 | long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, | |
245 | long nr_to_write) | |
246 | { | |
9df27d98 | 247 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
248 | pgoff_t index = 0, end = LONG_MAX; |
249 | struct pagevec pvec; | |
250 | long nwritten = 0; | |
251 | struct writeback_control wbc = { | |
252 | .for_reclaim = 0, | |
253 | }; | |
254 | ||
255 | pagevec_init(&pvec, 0); | |
256 | ||
257 | while (index <= end) { | |
258 | int i, nr_pages; | |
259 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, | |
260 | PAGECACHE_TAG_DIRTY, | |
261 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
cfb271d4 | 262 | if (unlikely(nr_pages == 0)) |
127e670a JK |
263 | break; |
264 | ||
265 | for (i = 0; i < nr_pages; i++) { | |
266 | struct page *page = pvec.pages[i]; | |
203681f6 | 267 | |
127e670a | 268 | lock_page(page); |
203681f6 JK |
269 | |
270 | if (unlikely(page->mapping != mapping)) { | |
271 | continue_unlock: | |
272 | unlock_page(page); | |
273 | continue; | |
274 | } | |
275 | if (!PageDirty(page)) { | |
276 | /* someone wrote it for us */ | |
277 | goto continue_unlock; | |
278 | } | |
279 | ||
280 | if (!clear_page_dirty_for_io(page)) | |
281 | goto continue_unlock; | |
282 | ||
97dc3fd2 | 283 | if (mapping->a_ops->writepage(page, &wbc)) { |
577e3495 JK |
284 | unlock_page(page); |
285 | break; | |
286 | } | |
cfb271d4 CY |
287 | nwritten++; |
288 | if (unlikely(nwritten >= nr_to_write)) | |
127e670a JK |
289 | break; |
290 | } | |
291 | pagevec_release(&pvec); | |
292 | cond_resched(); | |
293 | } | |
294 | ||
295 | if (nwritten) | |
458e6197 | 296 | f2fs_submit_merged_bio(sbi, type, WRITE); |
127e670a JK |
297 | |
298 | return nwritten; | |
299 | } | |
300 | ||
301 | static int f2fs_set_meta_page_dirty(struct page *page) | |
302 | { | |
26c6b887 JK |
303 | trace_f2fs_set_page_dirty(page, META); |
304 | ||
127e670a JK |
305 | SetPageUptodate(page); |
306 | if (!PageDirty(page)) { | |
307 | __set_page_dirty_nobuffers(page); | |
4081363f | 308 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); |
1601839e | 309 | SetPagePrivate(page); |
9e4ded3f | 310 | f2fs_trace_pid(page); |
127e670a JK |
311 | return 1; |
312 | } | |
313 | return 0; | |
314 | } | |
315 | ||
316 | const struct address_space_operations f2fs_meta_aops = { | |
317 | .writepage = f2fs_write_meta_page, | |
318 | .writepages = f2fs_write_meta_pages, | |
319 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
487261f3 CY |
320 | .invalidatepage = f2fs_invalidate_page, |
321 | .releasepage = f2fs_release_page, | |
127e670a JK |
322 | }; |
323 | ||
6451e041 | 324 | static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 325 | { |
67298804 | 326 | struct inode_management *im = &sbi->im[type]; |
39efac41 JK |
327 | struct ino_entry *e; |
328 | retry: | |
769ec6e5 JK |
329 | if (radix_tree_preload(GFP_NOFS)) { |
330 | cond_resched(); | |
331 | goto retry; | |
332 | } | |
333 | ||
67298804 | 334 | spin_lock(&im->ino_lock); |
39efac41 | 335 | |
67298804 | 336 | e = radix_tree_lookup(&im->ino_root, ino); |
39efac41 JK |
337 | if (!e) { |
338 | e = kmem_cache_alloc(ino_entry_slab, GFP_ATOMIC); | |
339 | if (!e) { | |
67298804 | 340 | spin_unlock(&im->ino_lock); |
769ec6e5 | 341 | radix_tree_preload_end(); |
39efac41 | 342 | goto retry; |
953e6cc6 | 343 | } |
67298804 CY |
344 | if (radix_tree_insert(&im->ino_root, ino, e)) { |
345 | spin_unlock(&im->ino_lock); | |
39efac41 | 346 | kmem_cache_free(ino_entry_slab, e); |
769ec6e5 | 347 | radix_tree_preload_end(); |
39efac41 JK |
348 | goto retry; |
349 | } | |
350 | memset(e, 0, sizeof(struct ino_entry)); | |
351 | e->ino = ino; | |
953e6cc6 | 352 | |
67298804 | 353 | list_add_tail(&e->list, &im->ino_list); |
8c402946 | 354 | if (type != ORPHAN_INO) |
67298804 | 355 | im->ino_num++; |
39efac41 | 356 | } |
67298804 | 357 | spin_unlock(&im->ino_lock); |
769ec6e5 | 358 | radix_tree_preload_end(); |
953e6cc6 JK |
359 | } |
360 | ||
6451e041 | 361 | static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 362 | { |
67298804 | 363 | struct inode_management *im = &sbi->im[type]; |
6451e041 | 364 | struct ino_entry *e; |
953e6cc6 | 365 | |
67298804 CY |
366 | spin_lock(&im->ino_lock); |
367 | e = radix_tree_lookup(&im->ino_root, ino); | |
39efac41 JK |
368 | if (e) { |
369 | list_del(&e->list); | |
67298804 CY |
370 | radix_tree_delete(&im->ino_root, ino); |
371 | im->ino_num--; | |
372 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
373 | kmem_cache_free(ino_entry_slab, e); |
374 | return; | |
953e6cc6 | 375 | } |
67298804 | 376 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
377 | } |
378 | ||
fff04f90 JK |
379 | void add_dirty_inode(struct f2fs_sb_info *sbi, nid_t ino, int type) |
380 | { | |
381 | /* add new dirty ino entry into list */ | |
382 | __add_ino_entry(sbi, ino, type); | |
383 | } | |
384 | ||
385 | void remove_dirty_inode(struct f2fs_sb_info *sbi, nid_t ino, int type) | |
386 | { | |
387 | /* remove dirty ino entry from list */ | |
388 | __remove_ino_entry(sbi, ino, type); | |
389 | } | |
390 | ||
391 | /* mode should be APPEND_INO or UPDATE_INO */ | |
392 | bool exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode) | |
393 | { | |
67298804 | 394 | struct inode_management *im = &sbi->im[mode]; |
fff04f90 | 395 | struct ino_entry *e; |
67298804 CY |
396 | |
397 | spin_lock(&im->ino_lock); | |
398 | e = radix_tree_lookup(&im->ino_root, ino); | |
399 | spin_unlock(&im->ino_lock); | |
fff04f90 JK |
400 | return e ? true : false; |
401 | } | |
402 | ||
6f12ac25 | 403 | void release_dirty_inode(struct f2fs_sb_info *sbi) |
fff04f90 JK |
404 | { |
405 | struct ino_entry *e, *tmp; | |
406 | int i; | |
407 | ||
408 | for (i = APPEND_INO; i <= UPDATE_INO; i++) { | |
67298804 CY |
409 | struct inode_management *im = &sbi->im[i]; |
410 | ||
411 | spin_lock(&im->ino_lock); | |
412 | list_for_each_entry_safe(e, tmp, &im->ino_list, list) { | |
fff04f90 | 413 | list_del(&e->list); |
67298804 | 414 | radix_tree_delete(&im->ino_root, e->ino); |
fff04f90 | 415 | kmem_cache_free(ino_entry_slab, e); |
67298804 | 416 | im->ino_num--; |
fff04f90 | 417 | } |
67298804 | 418 | spin_unlock(&im->ino_lock); |
fff04f90 JK |
419 | } |
420 | } | |
421 | ||
cbd56e7d | 422 | int acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 423 | { |
67298804 | 424 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a JK |
425 | int err = 0; |
426 | ||
67298804 CY |
427 | spin_lock(&im->ino_lock); |
428 | if (unlikely(im->ino_num >= sbi->max_orphans)) | |
127e670a | 429 | err = -ENOSPC; |
cbd56e7d | 430 | else |
67298804 CY |
431 | im->ino_num++; |
432 | spin_unlock(&im->ino_lock); | |
0d47c1ad | 433 | |
127e670a JK |
434 | return err; |
435 | } | |
436 | ||
cbd56e7d JK |
437 | void release_orphan_inode(struct f2fs_sb_info *sbi) |
438 | { | |
67298804 CY |
439 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
440 | ||
441 | spin_lock(&im->ino_lock); | |
442 | f2fs_bug_on(sbi, im->ino_num == 0); | |
443 | im->ino_num--; | |
444 | spin_unlock(&im->ino_lock); | |
cbd56e7d JK |
445 | } |
446 | ||
127e670a JK |
447 | void add_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
448 | { | |
39efac41 | 449 | /* add new orphan ino entry into list */ |
6451e041 | 450 | __add_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
451 | } |
452 | ||
453 | void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
454 | { | |
953e6cc6 | 455 | /* remove orphan entry from orphan list */ |
6451e041 | 456 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
457 | } |
458 | ||
459 | static void recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
460 | { | |
461 | struct inode *inode = f2fs_iget(sbi->sb, ino); | |
9850cf4a | 462 | f2fs_bug_on(sbi, IS_ERR(inode)); |
127e670a JK |
463 | clear_nlink(inode); |
464 | ||
465 | /* truncate all the data during iput */ | |
466 | iput(inode); | |
467 | } | |
468 | ||
8f99a946 | 469 | void recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a | 470 | { |
3c642985 | 471 | block_t start_blk, orphan_blocks, i, j; |
127e670a | 472 | |
25ca923b | 473 | if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG)) |
8f99a946 | 474 | return; |
127e670a | 475 | |
caf0047e | 476 | set_sbi_flag(sbi, SBI_POR_DOING); |
1dbe4152 | 477 | |
55141486 | 478 | start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi); |
3c642985 | 479 | orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi); |
127e670a | 480 | |
3c642985 | 481 | ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP); |
662befda | 482 | |
3c642985 | 483 | for (i = 0; i < orphan_blocks; i++) { |
127e670a JK |
484 | struct page *page = get_meta_page(sbi, start_blk + i); |
485 | struct f2fs_orphan_block *orphan_blk; | |
486 | ||
487 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); | |
488 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
489 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
490 | recover_orphan_inode(sbi, ino); | |
491 | } | |
492 | f2fs_put_page(page, 1); | |
493 | } | |
494 | /* clear Orphan Flag */ | |
25ca923b | 495 | clear_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG); |
caf0047e | 496 | clear_sbi_flag(sbi, SBI_POR_DOING); |
8f99a946 | 497 | return; |
127e670a JK |
498 | } |
499 | ||
500 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
501 | { | |
502c6e0b | 502 | struct list_head *head; |
127e670a | 503 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 504 | unsigned int nentries = 0; |
4531929e | 505 | unsigned short index; |
8c402946 | 506 | unsigned short orphan_blocks; |
4531929e | 507 | struct page *page = NULL; |
6451e041 | 508 | struct ino_entry *orphan = NULL; |
67298804 | 509 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 510 | |
67298804 | 511 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 512 | |
4531929e | 513 | for (index = 0; index < orphan_blocks; index++) |
63f5384c | 514 | grab_meta_page(sbi, start_blk + index); |
127e670a | 515 | |
4531929e | 516 | index = 1; |
d6c67a4f JK |
517 | |
518 | /* | |
519 | * we don't need to do spin_lock(&im->ino_lock) here, since all the | |
520 | * orphan inode operations are covered under f2fs_lock_op(). | |
521 | * And, spin_lock should be avoided due to page operations below. | |
522 | */ | |
67298804 | 523 | head = &im->ino_list; |
127e670a JK |
524 | |
525 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
526 | list_for_each_entry(orphan, head, list) { |
527 | if (!page) { | |
63f5384c | 528 | page = find_get_page(META_MAPPING(sbi), start_blk++); |
9850cf4a | 529 | f2fs_bug_on(sbi, !page); |
502c6e0b GZ |
530 | orphan_blk = |
531 | (struct f2fs_orphan_block *)page_address(page); | |
532 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
63f5384c | 533 | f2fs_put_page(page, 0); |
502c6e0b | 534 | } |
127e670a | 535 | |
36795567 | 536 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 537 | |
36795567 | 538 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
539 | /* |
540 | * an orphan block is full of 1020 entries, | |
541 | * then we need to flush current orphan blocks | |
542 | * and bring another one in memory | |
543 | */ | |
544 | orphan_blk->blk_addr = cpu_to_le16(index); | |
545 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
546 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
547 | set_page_dirty(page); | |
548 | f2fs_put_page(page, 1); | |
549 | index++; | |
127e670a JK |
550 | nentries = 0; |
551 | page = NULL; | |
552 | } | |
502c6e0b | 553 | } |
127e670a | 554 | |
502c6e0b GZ |
555 | if (page) { |
556 | orphan_blk->blk_addr = cpu_to_le16(index); | |
557 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
558 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
559 | set_page_dirty(page); | |
560 | f2fs_put_page(page, 1); | |
127e670a | 561 | } |
127e670a JK |
562 | } |
563 | ||
564 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, | |
565 | block_t cp_addr, unsigned long long *version) | |
566 | { | |
567 | struct page *cp_page_1, *cp_page_2 = NULL; | |
568 | unsigned long blk_size = sbi->blocksize; | |
569 | struct f2fs_checkpoint *cp_block; | |
570 | unsigned long long cur_version = 0, pre_version = 0; | |
127e670a | 571 | size_t crc_offset; |
7e586fa0 | 572 | __u32 crc = 0; |
127e670a JK |
573 | |
574 | /* Read the 1st cp block in this CP pack */ | |
575 | cp_page_1 = get_meta_page(sbi, cp_addr); | |
576 | ||
577 | /* get the version number */ | |
578 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_1); | |
579 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
580 | if (crc_offset >= blk_size) | |
581 | goto invalid_cp1; | |
582 | ||
29e7043f | 583 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
584 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
585 | goto invalid_cp1; | |
586 | ||
d71b5564 | 587 | pre_version = cur_cp_version(cp_block); |
127e670a JK |
588 | |
589 | /* Read the 2nd cp block in this CP pack */ | |
25ca923b | 590 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
127e670a JK |
591 | cp_page_2 = get_meta_page(sbi, cp_addr); |
592 | ||
593 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_2); | |
594 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
595 | if (crc_offset >= blk_size) | |
596 | goto invalid_cp2; | |
597 | ||
29e7043f | 598 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
599 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
600 | goto invalid_cp2; | |
601 | ||
d71b5564 | 602 | cur_version = cur_cp_version(cp_block); |
127e670a JK |
603 | |
604 | if (cur_version == pre_version) { | |
605 | *version = cur_version; | |
606 | f2fs_put_page(cp_page_2, 1); | |
607 | return cp_page_1; | |
608 | } | |
609 | invalid_cp2: | |
610 | f2fs_put_page(cp_page_2, 1); | |
611 | invalid_cp1: | |
612 | f2fs_put_page(cp_page_1, 1); | |
613 | return NULL; | |
614 | } | |
615 | ||
616 | int get_valid_checkpoint(struct f2fs_sb_info *sbi) | |
617 | { | |
618 | struct f2fs_checkpoint *cp_block; | |
619 | struct f2fs_super_block *fsb = sbi->raw_super; | |
620 | struct page *cp1, *cp2, *cur_page; | |
621 | unsigned long blk_size = sbi->blocksize; | |
622 | unsigned long long cp1_version = 0, cp2_version = 0; | |
623 | unsigned long long cp_start_blk_no; | |
55141486 | 624 | unsigned int cp_blks = 1 + __cp_payload(sbi); |
1dbe4152 CL |
625 | block_t cp_blk_no; |
626 | int i; | |
127e670a | 627 | |
1dbe4152 | 628 | sbi->ckpt = kzalloc(cp_blks * blk_size, GFP_KERNEL); |
127e670a JK |
629 | if (!sbi->ckpt) |
630 | return -ENOMEM; | |
631 | /* | |
632 | * Finding out valid cp block involves read both | |
633 | * sets( cp pack1 and cp pack 2) | |
634 | */ | |
635 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
636 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
637 | ||
638 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
639 | cp_start_blk_no += ((unsigned long long)1) << |
640 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
641 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
642 | ||
643 | if (cp1 && cp2) { | |
644 | if (ver_after(cp2_version, cp1_version)) | |
645 | cur_page = cp2; | |
646 | else | |
647 | cur_page = cp1; | |
648 | } else if (cp1) { | |
649 | cur_page = cp1; | |
650 | } else if (cp2) { | |
651 | cur_page = cp2; | |
652 | } else { | |
653 | goto fail_no_cp; | |
654 | } | |
655 | ||
656 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
657 | memcpy(sbi->ckpt, cp_block, blk_size); | |
658 | ||
1dbe4152 CL |
659 | if (cp_blks <= 1) |
660 | goto done; | |
661 | ||
662 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
663 | if (cur_page == cp2) | |
664 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
665 | ||
666 | for (i = 1; i < cp_blks; i++) { | |
667 | void *sit_bitmap_ptr; | |
668 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
669 | ||
670 | cur_page = get_meta_page(sbi, cp_blk_no + i); | |
671 | sit_bitmap_ptr = page_address(cur_page); | |
672 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
673 | f2fs_put_page(cur_page, 1); | |
674 | } | |
675 | done: | |
127e670a JK |
676 | f2fs_put_page(cp1, 1); |
677 | f2fs_put_page(cp2, 1); | |
678 | return 0; | |
679 | ||
680 | fail_no_cp: | |
681 | kfree(sbi->ckpt); | |
682 | return -EINVAL; | |
683 | } | |
684 | ||
06292073 | 685 | static int __add_dirty_inode(struct inode *inode, struct inode_entry *new) |
127e670a | 686 | { |
4081363f | 687 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
127e670a | 688 | |
ed57c27f JK |
689 | if (is_inode_flag_set(F2FS_I(inode), FI_DIRTY_DIR)) |
690 | return -EEXIST; | |
2d7b822a | 691 | |
ed57c27f JK |
692 | set_inode_flag(F2FS_I(inode), FI_DIRTY_DIR); |
693 | F2FS_I(inode)->dirty_dir = new; | |
694 | list_add_tail(&new->list, &sbi->dir_inode_list); | |
dcdfff65 | 695 | stat_inc_dirty_dir(sbi); |
5deb8267 JK |
696 | return 0; |
697 | } | |
698 | ||
a7ffdbe2 | 699 | void update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 700 | { |
4081363f | 701 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
06292073 | 702 | struct inode_entry *new; |
cf0ee0f0 | 703 | int ret = 0; |
5deb8267 | 704 | |
a7ffdbe2 | 705 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode)) |
127e670a | 706 | return; |
7bd59381 | 707 | |
a7ffdbe2 JK |
708 | if (!S_ISDIR(inode->i_mode)) { |
709 | inode_inc_dirty_pages(inode); | |
710 | goto out; | |
711 | } | |
712 | ||
7bd59381 | 713 | new = f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS); |
127e670a JK |
714 | new->inode = inode; |
715 | INIT_LIST_HEAD(&new->list); | |
716 | ||
717 | spin_lock(&sbi->dir_inode_lock); | |
cf0ee0f0 | 718 | ret = __add_dirty_inode(inode, new); |
a7ffdbe2 | 719 | inode_inc_dirty_pages(inode); |
5deb8267 | 720 | spin_unlock(&sbi->dir_inode_lock); |
cf0ee0f0 CY |
721 | |
722 | if (ret) | |
723 | kmem_cache_free(inode_entry_slab, new); | |
a7ffdbe2 JK |
724 | out: |
725 | SetPagePrivate(page); | |
9e4ded3f | 726 | f2fs_trace_pid(page); |
5deb8267 JK |
727 | } |
728 | ||
729 | void add_dirty_dir_inode(struct inode *inode) | |
730 | { | |
4081363f | 731 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
06292073 | 732 | struct inode_entry *new = |
7bd59381 | 733 | f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS); |
cf0ee0f0 | 734 | int ret = 0; |
7bd59381 | 735 | |
5deb8267 JK |
736 | new->inode = inode; |
737 | INIT_LIST_HEAD(&new->list); | |
127e670a | 738 | |
5deb8267 | 739 | spin_lock(&sbi->dir_inode_lock); |
cf0ee0f0 | 740 | ret = __add_dirty_inode(inode, new); |
127e670a | 741 | spin_unlock(&sbi->dir_inode_lock); |
cf0ee0f0 CY |
742 | |
743 | if (ret) | |
744 | kmem_cache_free(inode_entry_slab, new); | |
127e670a JK |
745 | } |
746 | ||
747 | void remove_dirty_dir_inode(struct inode *inode) | |
748 | { | |
4081363f | 749 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
06292073 | 750 | struct inode_entry *entry; |
127e670a JK |
751 | |
752 | if (!S_ISDIR(inode->i_mode)) | |
753 | return; | |
754 | ||
755 | spin_lock(&sbi->dir_inode_lock); | |
a7ffdbe2 | 756 | if (get_dirty_pages(inode) || |
ed57c27f | 757 | !is_inode_flag_set(F2FS_I(inode), FI_DIRTY_DIR)) { |
3b10b1fd JK |
758 | spin_unlock(&sbi->dir_inode_lock); |
759 | return; | |
760 | } | |
127e670a | 761 | |
ed57c27f JK |
762 | entry = F2FS_I(inode)->dirty_dir; |
763 | list_del(&entry->list); | |
764 | F2FS_I(inode)->dirty_dir = NULL; | |
765 | clear_inode_flag(F2FS_I(inode), FI_DIRTY_DIR); | |
766 | stat_dec_dirty_dir(sbi); | |
127e670a | 767 | spin_unlock(&sbi->dir_inode_lock); |
ed57c27f | 768 | kmem_cache_free(inode_entry_slab, entry); |
74d0b917 JK |
769 | |
770 | /* Only from the recovery routine */ | |
afc3eda2 JK |
771 | if (is_inode_flag_set(F2FS_I(inode), FI_DELAY_IPUT)) { |
772 | clear_inode_flag(F2FS_I(inode), FI_DELAY_IPUT); | |
74d0b917 | 773 | iput(inode); |
afc3eda2 | 774 | } |
74d0b917 JK |
775 | } |
776 | ||
127e670a JK |
777 | void sync_dirty_dir_inodes(struct f2fs_sb_info *sbi) |
778 | { | |
ce3b7d80 | 779 | struct list_head *head; |
06292073 | 780 | struct inode_entry *entry; |
127e670a JK |
781 | struct inode *inode; |
782 | retry: | |
af41d3ee JK |
783 | if (unlikely(f2fs_cp_error(sbi))) |
784 | return; | |
785 | ||
127e670a | 786 | spin_lock(&sbi->dir_inode_lock); |
ce3b7d80 GZ |
787 | |
788 | head = &sbi->dir_inode_list; | |
127e670a JK |
789 | if (list_empty(head)) { |
790 | spin_unlock(&sbi->dir_inode_lock); | |
791 | return; | |
792 | } | |
06292073 | 793 | entry = list_entry(head->next, struct inode_entry, list); |
127e670a JK |
794 | inode = igrab(entry->inode); |
795 | spin_unlock(&sbi->dir_inode_lock); | |
796 | if (inode) { | |
87d6f890 | 797 | filemap_fdatawrite(inode->i_mapping); |
127e670a JK |
798 | iput(inode); |
799 | } else { | |
800 | /* | |
801 | * We should submit bio, since it exists several | |
802 | * wribacking dentry pages in the freeing inode. | |
803 | */ | |
458e6197 | 804 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
7ecebe5e | 805 | cond_resched(); |
127e670a JK |
806 | } |
807 | goto retry; | |
808 | } | |
809 | ||
0a8165d7 | 810 | /* |
127e670a JK |
811 | * Freeze all the FS-operations for checkpoint. |
812 | */ | |
cf779cab | 813 | static int block_operations(struct f2fs_sb_info *sbi) |
127e670a | 814 | { |
127e670a JK |
815 | struct writeback_control wbc = { |
816 | .sync_mode = WB_SYNC_ALL, | |
817 | .nr_to_write = LONG_MAX, | |
818 | .for_reclaim = 0, | |
819 | }; | |
c718379b | 820 | struct blk_plug plug; |
cf779cab | 821 | int err = 0; |
c718379b JK |
822 | |
823 | blk_start_plug(&plug); | |
824 | ||
39936837 | 825 | retry_flush_dents: |
e479556b | 826 | f2fs_lock_all(sbi); |
127e670a | 827 | /* write all the dirty dentry pages */ |
127e670a | 828 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 829 | f2fs_unlock_all(sbi); |
39936837 | 830 | sync_dirty_dir_inodes(sbi); |
cf779cab JK |
831 | if (unlikely(f2fs_cp_error(sbi))) { |
832 | err = -EIO; | |
833 | goto out; | |
834 | } | |
39936837 | 835 | goto retry_flush_dents; |
127e670a JK |
836 | } |
837 | ||
127e670a | 838 | /* |
e1c42045 | 839 | * POR: we should ensure that there are no dirty node pages |
127e670a JK |
840 | * until finishing nat/sit flush. |
841 | */ | |
39936837 | 842 | retry_flush_nodes: |
b3582c68 | 843 | down_write(&sbi->node_write); |
127e670a JK |
844 | |
845 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
b3582c68 | 846 | up_write(&sbi->node_write); |
39936837 | 847 | sync_node_pages(sbi, 0, &wbc); |
cf779cab JK |
848 | if (unlikely(f2fs_cp_error(sbi))) { |
849 | f2fs_unlock_all(sbi); | |
850 | err = -EIO; | |
851 | goto out; | |
852 | } | |
39936837 | 853 | goto retry_flush_nodes; |
127e670a | 854 | } |
cf779cab | 855 | out: |
c718379b | 856 | blk_finish_plug(&plug); |
cf779cab | 857 | return err; |
127e670a JK |
858 | } |
859 | ||
860 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
861 | { | |
b3582c68 | 862 | up_write(&sbi->node_write); |
e479556b | 863 | f2fs_unlock_all(sbi); |
127e670a JK |
864 | } |
865 | ||
fb51b5ef CL |
866 | static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi) |
867 | { | |
868 | DEFINE_WAIT(wait); | |
869 | ||
870 | for (;;) { | |
871 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
872 | ||
873 | if (!get_pages(sbi, F2FS_WRITEBACK)) | |
874 | break; | |
875 | ||
876 | io_schedule(); | |
877 | } | |
878 | finish_wait(&sbi->cp_wait, &wait); | |
879 | } | |
880 | ||
75ab4cb8 | 881 | static void do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
882 | { |
883 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
cf2271e7 | 884 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); |
77041823 | 885 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
67298804 | 886 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; |
77041823 | 887 | nid_t last_nid = nm_i->next_scan_nid; |
127e670a JK |
888 | block_t start_blk; |
889 | struct page *cp_page; | |
890 | unsigned int data_sum_blocks, orphan_blocks; | |
7e586fa0 | 891 | __u32 crc32 = 0; |
127e670a | 892 | void *kaddr; |
127e670a | 893 | int i; |
55141486 | 894 | int cp_payload_blks = __cp_payload(sbi); |
127e670a | 895 | |
1e87a78d JK |
896 | /* |
897 | * This avoids to conduct wrong roll-forward operations and uses | |
898 | * metapages, so should be called prior to sync_meta_pages below. | |
899 | */ | |
cf2271e7 | 900 | discard_next_dnode(sbi, NEXT_FREE_BLKADDR(sbi, curseg)); |
127e670a JK |
901 | |
902 | /* Flush all the NAT/SIT pages */ | |
cf779cab | 903 | while (get_pages(sbi, F2FS_DIRTY_META)) { |
127e670a | 904 | sync_meta_pages(sbi, META, LONG_MAX); |
cf779cab JK |
905 | if (unlikely(f2fs_cp_error(sbi))) |
906 | return; | |
907 | } | |
127e670a JK |
908 | |
909 | next_free_nid(sbi, &last_nid); | |
910 | ||
911 | /* | |
912 | * modify checkpoint | |
913 | * version number is already updated | |
914 | */ | |
915 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi)); | |
916 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
917 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); | |
b5b82205 | 918 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
919 | ckpt->cur_node_segno[i] = |
920 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
921 | ckpt->cur_node_blkoff[i] = | |
922 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
923 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
924 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
925 | } | |
b5b82205 | 926 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
927 | ckpt->cur_data_segno[i] = |
928 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
929 | ckpt->cur_data_blkoff[i] = | |
930 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
931 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
932 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
933 | } | |
934 | ||
935 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
936 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
937 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
938 | ||
939 | /* 2 cp + n data seg summary + orphan inode blocks */ | |
3fa06d7b | 940 | data_sum_blocks = npages_for_summary_flush(sbi, false); |
b5b82205 | 941 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
25ca923b | 942 | set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 943 | else |
25ca923b | 944 | clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 945 | |
67298804 | 946 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
947 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
948 | orphan_blocks); | |
127e670a | 949 | |
119ee914 | 950 | if (__remain_node_summaries(cpc->reason)) |
b5b82205 | 951 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
952 | cp_payload_blks + data_sum_blocks + |
953 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
119ee914 | 954 | else |
b5b82205 | 955 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS + |
1dbe4152 CL |
956 | cp_payload_blks + data_sum_blocks + |
957 | orphan_blocks); | |
119ee914 JK |
958 | |
959 | if (cpc->reason == CP_UMOUNT) | |
960 | set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
961 | else | |
962 | clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
963 | ||
964 | if (cpc->reason == CP_FASTBOOT) | |
965 | set_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
966 | else | |
967 | clear_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
127e670a | 968 | |
67298804 | 969 | if (orphan_num) |
25ca923b | 970 | set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 971 | else |
25ca923b | 972 | clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 973 | |
caf0047e | 974 | if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) |
2ae4c673 JK |
975 | set_ckpt_flags(ckpt, CP_FSCK_FLAG); |
976 | ||
127e670a JK |
977 | /* update SIT/NAT bitmap */ |
978 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
979 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
980 | ||
981 | crc32 = f2fs_crc32(ckpt, le32_to_cpu(ckpt->checksum_offset)); | |
7e586fa0 JK |
982 | *((__le32 *)((unsigned char *)ckpt + |
983 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
984 | = cpu_to_le32(crc32); |
985 | ||
986 | start_blk = __start_cp_addr(sbi); | |
987 | ||
988 | /* write out checkpoint buffer at block 0 */ | |
989 | cp_page = grab_meta_page(sbi, start_blk++); | |
990 | kaddr = page_address(cp_page); | |
f7ef9b83 | 991 | memcpy(kaddr, ckpt, F2FS_BLKSIZE); |
127e670a JK |
992 | set_page_dirty(cp_page); |
993 | f2fs_put_page(cp_page, 1); | |
994 | ||
1dbe4152 CL |
995 | for (i = 1; i < 1 + cp_payload_blks; i++) { |
996 | cp_page = grab_meta_page(sbi, start_blk++); | |
997 | kaddr = page_address(cp_page); | |
f7ef9b83 | 998 | memcpy(kaddr, (char *)ckpt + i * F2FS_BLKSIZE, F2FS_BLKSIZE); |
1dbe4152 CL |
999 | set_page_dirty(cp_page); |
1000 | f2fs_put_page(cp_page, 1); | |
1001 | } | |
1002 | ||
67298804 | 1003 | if (orphan_num) { |
127e670a JK |
1004 | write_orphan_inodes(sbi, start_blk); |
1005 | start_blk += orphan_blocks; | |
1006 | } | |
1007 | ||
1008 | write_data_summaries(sbi, start_blk); | |
1009 | start_blk += data_sum_blocks; | |
119ee914 | 1010 | if (__remain_node_summaries(cpc->reason)) { |
127e670a JK |
1011 | write_node_summaries(sbi, start_blk); |
1012 | start_blk += NR_CURSEG_NODE_TYPE; | |
1013 | } | |
1014 | ||
1015 | /* writeout checkpoint block */ | |
1016 | cp_page = grab_meta_page(sbi, start_blk); | |
1017 | kaddr = page_address(cp_page); | |
f7ef9b83 | 1018 | memcpy(kaddr, ckpt, F2FS_BLKSIZE); |
127e670a JK |
1019 | set_page_dirty(cp_page); |
1020 | f2fs_put_page(cp_page, 1); | |
1021 | ||
1022 | /* wait for previous submitted node/meta pages writeback */ | |
fb51b5ef | 1023 | wait_on_all_pages_writeback(sbi); |
127e670a | 1024 | |
cf779cab JK |
1025 | if (unlikely(f2fs_cp_error(sbi))) |
1026 | return; | |
1027 | ||
4ef51a8f | 1028 | filemap_fdatawait_range(NODE_MAPPING(sbi), 0, LONG_MAX); |
9df27d98 | 1029 | filemap_fdatawait_range(META_MAPPING(sbi), 0, LONG_MAX); |
127e670a JK |
1030 | |
1031 | /* update user_block_counts */ | |
1032 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
1033 | sbi->alloc_valid_block_count = 0; | |
1034 | ||
1035 | /* Here, we only have one bio having CP pack */ | |
577e3495 | 1036 | sync_meta_pages(sbi, META_FLUSH, LONG_MAX); |
127e670a | 1037 | |
6a8f8ca5 JK |
1038 | /* wait for previous submitted meta pages writeback */ |
1039 | wait_on_all_pages_writeback(sbi); | |
1040 | ||
cf779cab JK |
1041 | release_dirty_inode(sbi); |
1042 | ||
1043 | if (unlikely(f2fs_cp_error(sbi))) | |
1044 | return; | |
1045 | ||
1046 | clear_prefree_segments(sbi); | |
caf0047e | 1047 | clear_sbi_flag(sbi, SBI_IS_DIRTY); |
127e670a JK |
1048 | } |
1049 | ||
0a8165d7 | 1050 | /* |
e1c42045 | 1051 | * We guarantee that this checkpoint procedure will not fail. |
127e670a | 1052 | */ |
75ab4cb8 | 1053 | void write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1054 | { |
1055 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1056 | unsigned long long ckpt_ver; | |
1057 | ||
43727527 | 1058 | mutex_lock(&sbi->cp_mutex); |
8501017e | 1059 | |
caf0047e | 1060 | if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) && |
a66cdd98 JK |
1061 | (cpc->reason == CP_FASTBOOT || cpc->reason == CP_SYNC || |
1062 | (cpc->reason == CP_DISCARD && !sbi->discard_blks))) | |
8501017e | 1063 | goto out; |
cf779cab JK |
1064 | if (unlikely(f2fs_cp_error(sbi))) |
1065 | goto out; | |
11504a8e JK |
1066 | if (f2fs_readonly(sbi->sb)) |
1067 | goto out; | |
2bda542d WL |
1068 | |
1069 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); | |
1070 | ||
cf779cab JK |
1071 | if (block_operations(sbi)) |
1072 | goto out; | |
127e670a | 1073 | |
75ab4cb8 | 1074 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1075 | |
458e6197 JK |
1076 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
1077 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
1078 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
127e670a JK |
1079 | |
1080 | /* | |
1081 | * update checkpoint pack index | |
1082 | * Increase the version number so that | |
1083 | * SIT entries and seg summaries are written at correct place | |
1084 | */ | |
d71b5564 | 1085 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1086 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1087 | ||
1088 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
1089 | flush_nat_entries(sbi); | |
4b2fecc8 | 1090 | flush_sit_entries(sbi, cpc); |
127e670a | 1091 | |
127e670a | 1092 | /* unlock all the fs_lock[] in do_checkpoint() */ |
75ab4cb8 | 1093 | do_checkpoint(sbi, cpc); |
127e670a JK |
1094 | |
1095 | unblock_operations(sbi); | |
942e0be6 | 1096 | stat_inc_cp_count(sbi->stat_info); |
10027551 JK |
1097 | |
1098 | if (cpc->reason == CP_RECOVERY) | |
1099 | f2fs_msg(sbi->sb, KERN_NOTICE, | |
1100 | "checkpoint: version = %llx", ckpt_ver); | |
8501017e JK |
1101 | out: |
1102 | mutex_unlock(&sbi->cp_mutex); | |
75ab4cb8 | 1103 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
127e670a JK |
1104 | } |
1105 | ||
6451e041 | 1106 | void init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1107 | { |
6451e041 JK |
1108 | int i; |
1109 | ||
1110 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1111 | struct inode_management *im = &sbi->im[i]; |
1112 | ||
1113 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1114 | spin_lock_init(&im->ino_lock); | |
1115 | INIT_LIST_HEAD(&im->ino_list); | |
1116 | im->ino_num = 0; | |
6451e041 JK |
1117 | } |
1118 | ||
b5b82205 | 1119 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
14b42817 WL |
1120 | NR_CURSEG_TYPE - __cp_payload(sbi)) * |
1121 | F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
1122 | } |
1123 | ||
6e6093a8 | 1124 | int __init create_checkpoint_caches(void) |
127e670a | 1125 | { |
6451e041 JK |
1126 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1127 | sizeof(struct ino_entry)); | |
1128 | if (!ino_entry_slab) | |
127e670a | 1129 | return -ENOMEM; |
06292073 CY |
1130 | inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry", |
1131 | sizeof(struct inode_entry)); | |
6bacf52f | 1132 | if (!inode_entry_slab) { |
6451e041 | 1133 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1134 | return -ENOMEM; |
1135 | } | |
1136 | return 0; | |
1137 | } | |
1138 | ||
1139 | void destroy_checkpoint_caches(void) | |
1140 | { | |
6451e041 | 1141 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1142 | kmem_cache_destroy(inode_entry_slab); |
1143 | } |