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 | } | |
fec1d657 | 42 | f2fs_wait_on_page_writeback(page, META, true); |
127e670a JK |
43 | SetPageUptodate(page); |
44 | return page; | |
45 | } | |
46 | ||
0a8165d7 | 47 | /* |
127e670a JK |
48 | * We guarantee no failure on the returned page. |
49 | */ | |
2b947003 CY |
50 | static struct page *__get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index, |
51 | bool is_meta) | |
127e670a | 52 | { |
9df27d98 | 53 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a | 54 | struct page *page; |
cf04e8eb | 55 | struct f2fs_io_info fio = { |
05ca3632 | 56 | .sbi = sbi, |
cf04e8eb JK |
57 | .type = META, |
58 | .rw = READ_SYNC | REQ_META | REQ_PRIO, | |
59 | .blk_addr = index, | |
4375a336 | 60 | .encrypted_page = NULL, |
cf04e8eb | 61 | }; |
2b947003 CY |
62 | |
63 | if (unlikely(!is_meta)) | |
64 | fio.rw &= ~REQ_META; | |
127e670a JK |
65 | repeat: |
66 | page = grab_cache_page(mapping, index); | |
67 | if (!page) { | |
68 | cond_resched(); | |
69 | goto repeat; | |
70 | } | |
393ff91f JK |
71 | if (PageUptodate(page)) |
72 | goto out; | |
73 | ||
05ca3632 JK |
74 | fio.page = page; |
75 | ||
86531d6b JK |
76 | if (f2fs_submit_page_bio(&fio)) { |
77 | f2fs_put_page(page, 1); | |
127e670a | 78 | goto repeat; |
86531d6b | 79 | } |
127e670a | 80 | |
393ff91f | 81 | lock_page(page); |
6bacf52f | 82 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
83 | f2fs_put_page(page, 1); |
84 | goto repeat; | |
85 | } | |
f3f338ca CY |
86 | |
87 | /* | |
88 | * if there is any IO error when accessing device, make our filesystem | |
89 | * readonly and make sure do not write checkpoint with non-uptodate | |
90 | * meta page. | |
91 | */ | |
92 | if (unlikely(!PageUptodate(page))) | |
93 | f2fs_stop_checkpoint(sbi); | |
393ff91f | 94 | out: |
127e670a JK |
95 | return page; |
96 | } | |
97 | ||
2b947003 CY |
98 | struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) |
99 | { | |
100 | return __get_meta_page(sbi, index, true); | |
101 | } | |
102 | ||
103 | /* for POR only */ | |
104 | struct page *get_tmp_page(struct f2fs_sb_info *sbi, pgoff_t index) | |
105 | { | |
106 | return __get_meta_page(sbi, index, false); | |
107 | } | |
108 | ||
f0c9cada | 109 | bool is_valid_blkaddr(struct f2fs_sb_info *sbi, block_t blkaddr, int type) |
662befda CY |
110 | { |
111 | switch (type) { | |
112 | case META_NAT: | |
66b00c18 | 113 | break; |
662befda | 114 | case META_SIT: |
66b00c18 CY |
115 | if (unlikely(blkaddr >= SIT_BLK_CNT(sbi))) |
116 | return false; | |
117 | break; | |
81c1a0f1 | 118 | case META_SSA: |
66b00c18 CY |
119 | if (unlikely(blkaddr >= MAIN_BLKADDR(sbi) || |
120 | blkaddr < SM_I(sbi)->ssa_blkaddr)) | |
121 | return false; | |
122 | break; | |
662befda | 123 | case META_CP: |
66b00c18 CY |
124 | if (unlikely(blkaddr >= SIT_I(sbi)->sit_base_addr || |
125 | blkaddr < __start_cp_addr(sbi))) | |
126 | return false; | |
127 | break; | |
4c521f49 | 128 | case META_POR: |
66b00c18 CY |
129 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
130 | blkaddr < MAIN_BLKADDR(sbi))) | |
131 | return false; | |
132 | break; | |
662befda CY |
133 | default: |
134 | BUG(); | |
135 | } | |
66b00c18 CY |
136 | |
137 | return true; | |
662befda CY |
138 | } |
139 | ||
140 | /* | |
81c1a0f1 | 141 | * Readahead CP/NAT/SIT/SSA pages |
662befda | 142 | */ |
26879fb1 CY |
143 | int ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages, |
144 | int type, bool sync) | |
662befda CY |
145 | { |
146 | block_t prev_blk_addr = 0; | |
147 | struct page *page; | |
4c521f49 | 148 | block_t blkno = start; |
662befda | 149 | struct f2fs_io_info fio = { |
05ca3632 | 150 | .sbi = sbi, |
662befda | 151 | .type = META, |
26879fb1 | 152 | .rw = sync ? (READ_SYNC | REQ_META | REQ_PRIO) : READA, |
4375a336 | 153 | .encrypted_page = NULL, |
662befda CY |
154 | }; |
155 | ||
2b947003 CY |
156 | if (unlikely(type == META_POR)) |
157 | fio.rw &= ~REQ_META; | |
158 | ||
662befda | 159 | for (; nrpages-- > 0; blkno++) { |
662befda | 160 | |
66b00c18 CY |
161 | if (!is_valid_blkaddr(sbi, blkno, type)) |
162 | goto out; | |
163 | ||
662befda CY |
164 | switch (type) { |
165 | case META_NAT: | |
66b00c18 CY |
166 | if (unlikely(blkno >= |
167 | NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))) | |
662befda | 168 | blkno = 0; |
66b00c18 | 169 | /* get nat block addr */ |
cf04e8eb | 170 | fio.blk_addr = current_nat_addr(sbi, |
662befda CY |
171 | blkno * NAT_ENTRY_PER_BLOCK); |
172 | break; | |
173 | case META_SIT: | |
174 | /* get sit block addr */ | |
cf04e8eb | 175 | fio.blk_addr = current_sit_addr(sbi, |
662befda | 176 | blkno * SIT_ENTRY_PER_BLOCK); |
cf04e8eb | 177 | if (blkno != start && prev_blk_addr + 1 != fio.blk_addr) |
662befda | 178 | goto out; |
cf04e8eb | 179 | prev_blk_addr = fio.blk_addr; |
662befda | 180 | break; |
81c1a0f1 | 181 | case META_SSA: |
662befda | 182 | case META_CP: |
4c521f49 | 183 | case META_POR: |
cf04e8eb | 184 | fio.blk_addr = blkno; |
662befda CY |
185 | break; |
186 | default: | |
187 | BUG(); | |
188 | } | |
189 | ||
cf04e8eb | 190 | page = grab_cache_page(META_MAPPING(sbi), fio.blk_addr); |
662befda CY |
191 | if (!page) |
192 | continue; | |
193 | if (PageUptodate(page)) { | |
662befda CY |
194 | f2fs_put_page(page, 1); |
195 | continue; | |
196 | } | |
197 | ||
05ca3632 JK |
198 | fio.page = page; |
199 | f2fs_submit_page_mbio(&fio); | |
662befda CY |
200 | f2fs_put_page(page, 0); |
201 | } | |
202 | out: | |
203 | f2fs_submit_merged_bio(sbi, META, READ); | |
204 | return blkno - start; | |
205 | } | |
206 | ||
635aee1f CY |
207 | void ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index) |
208 | { | |
209 | struct page *page; | |
210 | bool readahead = false; | |
211 | ||
212 | page = find_get_page(META_MAPPING(sbi), index); | |
213 | if (!page || (page && !PageUptodate(page))) | |
214 | readahead = true; | |
215 | f2fs_put_page(page, 0); | |
216 | ||
217 | if (readahead) | |
26879fb1 | 218 | ra_meta_pages(sbi, index, MAX_BIO_BLOCKS(sbi), META_POR, true); |
635aee1f CY |
219 | } |
220 | ||
127e670a JK |
221 | static int f2fs_write_meta_page(struct page *page, |
222 | struct writeback_control *wbc) | |
223 | { | |
4081363f | 224 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 225 | |
ecda0de3 CY |
226 | trace_f2fs_writepage(page, META); |
227 | ||
caf0047e | 228 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
cfb271d4 | 229 | goto redirty_out; |
857dc4e0 | 230 | if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0)) |
cfb271d4 | 231 | goto redirty_out; |
1e968fdf | 232 | if (unlikely(f2fs_cp_error(sbi))) |
cf779cab | 233 | goto redirty_out; |
127e670a | 234 | |
577e3495 JK |
235 | write_meta_page(sbi, page); |
236 | dec_page_count(sbi, F2FS_DIRTY_META); | |
0c3a5797 CY |
237 | |
238 | if (wbc->for_reclaim) | |
239 | f2fs_submit_merged_bio_cond(sbi, NULL, page, 0, META, WRITE); | |
240 | ||
577e3495 | 241 | unlock_page(page); |
857dc4e0 | 242 | |
0c3a5797 | 243 | if (unlikely(f2fs_cp_error(sbi))) |
857dc4e0 | 244 | f2fs_submit_merged_bio(sbi, META, WRITE); |
0c3a5797 | 245 | |
577e3495 | 246 | return 0; |
cfb271d4 CY |
247 | |
248 | redirty_out: | |
76f60268 | 249 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 250 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
251 | } |
252 | ||
253 | static int f2fs_write_meta_pages(struct address_space *mapping, | |
254 | struct writeback_control *wbc) | |
255 | { | |
4081363f | 256 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 257 | long diff, written; |
127e670a | 258 | |
e5748434 CY |
259 | trace_f2fs_writepages(mapping->host, wbc, META); |
260 | ||
5459aa97 | 261 | /* collect a number of dirty meta pages and write together */ |
50c8cdb3 JK |
262 | if (wbc->for_kupdate || |
263 | get_pages(sbi, F2FS_DIRTY_META) < nr_pages_to_skip(sbi, META)) | |
d3baf95d | 264 | goto skip_write; |
127e670a JK |
265 | |
266 | /* if mounting is failed, skip writing node pages */ | |
267 | mutex_lock(&sbi->cp_mutex); | |
50c8cdb3 JK |
268 | diff = nr_pages_to_write(sbi, META, wbc); |
269 | written = sync_meta_pages(sbi, META, wbc->nr_to_write); | |
127e670a | 270 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 271 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 272 | return 0; |
d3baf95d JK |
273 | |
274 | skip_write: | |
275 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
276 | return 0; | |
127e670a JK |
277 | } |
278 | ||
279 | long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, | |
280 | long nr_to_write) | |
281 | { | |
9df27d98 | 282 | struct address_space *mapping = META_MAPPING(sbi); |
6066d8cd | 283 | pgoff_t index = 0, end = LONG_MAX, prev = LONG_MAX; |
127e670a JK |
284 | struct pagevec pvec; |
285 | long nwritten = 0; | |
286 | struct writeback_control wbc = { | |
287 | .for_reclaim = 0, | |
288 | }; | |
289 | ||
290 | pagevec_init(&pvec, 0); | |
291 | ||
292 | while (index <= end) { | |
293 | int i, nr_pages; | |
294 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, | |
295 | PAGECACHE_TAG_DIRTY, | |
296 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
cfb271d4 | 297 | if (unlikely(nr_pages == 0)) |
127e670a JK |
298 | break; |
299 | ||
300 | for (i = 0; i < nr_pages; i++) { | |
301 | struct page *page = pvec.pages[i]; | |
203681f6 | 302 | |
6066d8cd JK |
303 | if (prev == LONG_MAX) |
304 | prev = page->index - 1; | |
305 | if (nr_to_write != LONG_MAX && page->index != prev + 1) { | |
306 | pagevec_release(&pvec); | |
307 | goto stop; | |
308 | } | |
309 | ||
127e670a | 310 | lock_page(page); |
203681f6 JK |
311 | |
312 | if (unlikely(page->mapping != mapping)) { | |
313 | continue_unlock: | |
314 | unlock_page(page); | |
315 | continue; | |
316 | } | |
317 | if (!PageDirty(page)) { | |
318 | /* someone wrote it for us */ | |
319 | goto continue_unlock; | |
320 | } | |
321 | ||
fa3d2bdf JK |
322 | f2fs_wait_on_page_writeback(page, META, true); |
323 | ||
324 | BUG_ON(PageWriteback(page)); | |
203681f6 JK |
325 | if (!clear_page_dirty_for_io(page)) |
326 | goto continue_unlock; | |
327 | ||
97dc3fd2 | 328 | if (mapping->a_ops->writepage(page, &wbc)) { |
577e3495 JK |
329 | unlock_page(page); |
330 | break; | |
331 | } | |
cfb271d4 | 332 | nwritten++; |
6066d8cd | 333 | prev = page->index; |
cfb271d4 | 334 | if (unlikely(nwritten >= nr_to_write)) |
127e670a JK |
335 | break; |
336 | } | |
337 | pagevec_release(&pvec); | |
338 | cond_resched(); | |
339 | } | |
6066d8cd | 340 | stop: |
127e670a | 341 | if (nwritten) |
458e6197 | 342 | f2fs_submit_merged_bio(sbi, type, WRITE); |
127e670a JK |
343 | |
344 | return nwritten; | |
345 | } | |
346 | ||
347 | static int f2fs_set_meta_page_dirty(struct page *page) | |
348 | { | |
26c6b887 JK |
349 | trace_f2fs_set_page_dirty(page, META); |
350 | ||
127e670a JK |
351 | SetPageUptodate(page); |
352 | if (!PageDirty(page)) { | |
353 | __set_page_dirty_nobuffers(page); | |
4081363f | 354 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); |
1601839e | 355 | SetPagePrivate(page); |
9e4ded3f | 356 | f2fs_trace_pid(page); |
127e670a JK |
357 | return 1; |
358 | } | |
359 | return 0; | |
360 | } | |
361 | ||
362 | const struct address_space_operations f2fs_meta_aops = { | |
363 | .writepage = f2fs_write_meta_page, | |
364 | .writepages = f2fs_write_meta_pages, | |
365 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
487261f3 CY |
366 | .invalidatepage = f2fs_invalidate_page, |
367 | .releasepage = f2fs_release_page, | |
127e670a JK |
368 | }; |
369 | ||
6451e041 | 370 | static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 371 | { |
67298804 | 372 | struct inode_management *im = &sbi->im[type]; |
80c54505 JK |
373 | struct ino_entry *e, *tmp; |
374 | ||
375 | tmp = f2fs_kmem_cache_alloc(ino_entry_slab, GFP_NOFS); | |
39efac41 | 376 | retry: |
80c54505 | 377 | radix_tree_preload(GFP_NOFS | __GFP_NOFAIL); |
769ec6e5 | 378 | |
67298804 | 379 | spin_lock(&im->ino_lock); |
67298804 | 380 | e = radix_tree_lookup(&im->ino_root, ino); |
39efac41 | 381 | if (!e) { |
80c54505 | 382 | e = tmp; |
67298804 CY |
383 | if (radix_tree_insert(&im->ino_root, ino, e)) { |
384 | spin_unlock(&im->ino_lock); | |
769ec6e5 | 385 | radix_tree_preload_end(); |
39efac41 JK |
386 | goto retry; |
387 | } | |
388 | memset(e, 0, sizeof(struct ino_entry)); | |
389 | e->ino = ino; | |
953e6cc6 | 390 | |
67298804 | 391 | list_add_tail(&e->list, &im->ino_list); |
8c402946 | 392 | if (type != ORPHAN_INO) |
67298804 | 393 | im->ino_num++; |
39efac41 | 394 | } |
67298804 | 395 | spin_unlock(&im->ino_lock); |
769ec6e5 | 396 | radix_tree_preload_end(); |
80c54505 JK |
397 | |
398 | if (e != tmp) | |
399 | kmem_cache_free(ino_entry_slab, tmp); | |
953e6cc6 JK |
400 | } |
401 | ||
6451e041 | 402 | static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 403 | { |
67298804 | 404 | struct inode_management *im = &sbi->im[type]; |
6451e041 | 405 | struct ino_entry *e; |
953e6cc6 | 406 | |
67298804 CY |
407 | spin_lock(&im->ino_lock); |
408 | e = radix_tree_lookup(&im->ino_root, ino); | |
39efac41 JK |
409 | if (e) { |
410 | list_del(&e->list); | |
67298804 CY |
411 | radix_tree_delete(&im->ino_root, ino); |
412 | im->ino_num--; | |
413 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
414 | kmem_cache_free(ino_entry_slab, e); |
415 | return; | |
953e6cc6 | 416 | } |
67298804 | 417 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
418 | } |
419 | ||
a49324f1 | 420 | void add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
421 | { |
422 | /* add new dirty ino entry into list */ | |
423 | __add_ino_entry(sbi, ino, type); | |
424 | } | |
425 | ||
a49324f1 | 426 | void remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
427 | { |
428 | /* remove dirty ino entry from list */ | |
429 | __remove_ino_entry(sbi, ino, type); | |
430 | } | |
431 | ||
432 | /* mode should be APPEND_INO or UPDATE_INO */ | |
433 | bool exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode) | |
434 | { | |
67298804 | 435 | struct inode_management *im = &sbi->im[mode]; |
fff04f90 | 436 | struct ino_entry *e; |
67298804 CY |
437 | |
438 | spin_lock(&im->ino_lock); | |
439 | e = radix_tree_lookup(&im->ino_root, ino); | |
440 | spin_unlock(&im->ino_lock); | |
fff04f90 JK |
441 | return e ? true : false; |
442 | } | |
443 | ||
a49324f1 | 444 | void release_ino_entry(struct f2fs_sb_info *sbi) |
fff04f90 JK |
445 | { |
446 | struct ino_entry *e, *tmp; | |
447 | int i; | |
448 | ||
449 | for (i = APPEND_INO; i <= UPDATE_INO; i++) { | |
67298804 CY |
450 | struct inode_management *im = &sbi->im[i]; |
451 | ||
452 | spin_lock(&im->ino_lock); | |
453 | list_for_each_entry_safe(e, tmp, &im->ino_list, list) { | |
fff04f90 | 454 | list_del(&e->list); |
67298804 | 455 | radix_tree_delete(&im->ino_root, e->ino); |
fff04f90 | 456 | kmem_cache_free(ino_entry_slab, e); |
67298804 | 457 | im->ino_num--; |
fff04f90 | 458 | } |
67298804 | 459 | spin_unlock(&im->ino_lock); |
fff04f90 JK |
460 | } |
461 | } | |
462 | ||
cbd56e7d | 463 | int acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 464 | { |
67298804 | 465 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a JK |
466 | int err = 0; |
467 | ||
67298804 CY |
468 | spin_lock(&im->ino_lock); |
469 | if (unlikely(im->ino_num >= sbi->max_orphans)) | |
127e670a | 470 | err = -ENOSPC; |
cbd56e7d | 471 | else |
67298804 CY |
472 | im->ino_num++; |
473 | spin_unlock(&im->ino_lock); | |
0d47c1ad | 474 | |
127e670a JK |
475 | return err; |
476 | } | |
477 | ||
cbd56e7d JK |
478 | void release_orphan_inode(struct f2fs_sb_info *sbi) |
479 | { | |
67298804 CY |
480 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
481 | ||
482 | spin_lock(&im->ino_lock); | |
483 | f2fs_bug_on(sbi, im->ino_num == 0); | |
484 | im->ino_num--; | |
485 | spin_unlock(&im->ino_lock); | |
cbd56e7d JK |
486 | } |
487 | ||
127e670a JK |
488 | void add_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
489 | { | |
39efac41 | 490 | /* add new orphan ino entry into list */ |
6451e041 | 491 | __add_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
492 | } |
493 | ||
494 | void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) | |
495 | { | |
953e6cc6 | 496 | /* remove orphan entry from orphan list */ |
6451e041 | 497 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
498 | } |
499 | ||
8c14bfad | 500 | static int recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 501 | { |
8c14bfad CY |
502 | struct inode *inode; |
503 | ||
504 | inode = f2fs_iget(sbi->sb, ino); | |
505 | if (IS_ERR(inode)) { | |
506 | /* | |
507 | * there should be a bug that we can't find the entry | |
508 | * to orphan inode. | |
509 | */ | |
510 | f2fs_bug_on(sbi, PTR_ERR(inode) == -ENOENT); | |
511 | return PTR_ERR(inode); | |
512 | } | |
513 | ||
127e670a JK |
514 | clear_nlink(inode); |
515 | ||
516 | /* truncate all the data during iput */ | |
517 | iput(inode); | |
8c14bfad | 518 | return 0; |
127e670a JK |
519 | } |
520 | ||
8c14bfad | 521 | int recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a | 522 | { |
3c642985 | 523 | block_t start_blk, orphan_blocks, i, j; |
8c14bfad | 524 | int err; |
127e670a | 525 | |
25ca923b | 526 | if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG)) |
8c14bfad | 527 | return 0; |
127e670a | 528 | |
55141486 | 529 | start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi); |
3c642985 | 530 | orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi); |
127e670a | 531 | |
26879fb1 | 532 | ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP, true); |
662befda | 533 | |
3c642985 | 534 | for (i = 0; i < orphan_blocks; i++) { |
127e670a JK |
535 | struct page *page = get_meta_page(sbi, start_blk + i); |
536 | struct f2fs_orphan_block *orphan_blk; | |
537 | ||
538 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); | |
539 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
540 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
8c14bfad CY |
541 | err = recover_orphan_inode(sbi, ino); |
542 | if (err) { | |
543 | f2fs_put_page(page, 1); | |
8c14bfad CY |
544 | return err; |
545 | } | |
127e670a JK |
546 | } |
547 | f2fs_put_page(page, 1); | |
548 | } | |
549 | /* clear Orphan Flag */ | |
25ca923b | 550 | clear_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG); |
8c14bfad | 551 | return 0; |
127e670a JK |
552 | } |
553 | ||
554 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
555 | { | |
502c6e0b | 556 | struct list_head *head; |
127e670a | 557 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 558 | unsigned int nentries = 0; |
bd936f84 | 559 | unsigned short index = 1; |
8c402946 | 560 | unsigned short orphan_blocks; |
4531929e | 561 | struct page *page = NULL; |
6451e041 | 562 | struct ino_entry *orphan = NULL; |
67298804 | 563 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 564 | |
67298804 | 565 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 566 | |
d6c67a4f JK |
567 | /* |
568 | * we don't need to do spin_lock(&im->ino_lock) here, since all the | |
569 | * orphan inode operations are covered under f2fs_lock_op(). | |
570 | * And, spin_lock should be avoided due to page operations below. | |
571 | */ | |
67298804 | 572 | head = &im->ino_list; |
127e670a JK |
573 | |
574 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
575 | list_for_each_entry(orphan, head, list) { |
576 | if (!page) { | |
bd936f84 | 577 | page = grab_meta_page(sbi, start_blk++); |
502c6e0b GZ |
578 | orphan_blk = |
579 | (struct f2fs_orphan_block *)page_address(page); | |
580 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
581 | } | |
127e670a | 582 | |
36795567 | 583 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 584 | |
36795567 | 585 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
586 | /* |
587 | * an orphan block is full of 1020 entries, | |
588 | * then we need to flush current orphan blocks | |
589 | * and bring another one in memory | |
590 | */ | |
591 | orphan_blk->blk_addr = cpu_to_le16(index); | |
592 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
593 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
594 | set_page_dirty(page); | |
595 | f2fs_put_page(page, 1); | |
596 | index++; | |
127e670a JK |
597 | nentries = 0; |
598 | page = NULL; | |
599 | } | |
502c6e0b | 600 | } |
127e670a | 601 | |
502c6e0b GZ |
602 | if (page) { |
603 | orphan_blk->blk_addr = cpu_to_le16(index); | |
604 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
605 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
606 | set_page_dirty(page); | |
607 | f2fs_put_page(page, 1); | |
127e670a | 608 | } |
127e670a JK |
609 | } |
610 | ||
611 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, | |
612 | block_t cp_addr, unsigned long long *version) | |
613 | { | |
614 | struct page *cp_page_1, *cp_page_2 = NULL; | |
615 | unsigned long blk_size = sbi->blocksize; | |
616 | struct f2fs_checkpoint *cp_block; | |
617 | unsigned long long cur_version = 0, pre_version = 0; | |
127e670a | 618 | size_t crc_offset; |
7e586fa0 | 619 | __u32 crc = 0; |
127e670a JK |
620 | |
621 | /* Read the 1st cp block in this CP pack */ | |
622 | cp_page_1 = get_meta_page(sbi, cp_addr); | |
623 | ||
624 | /* get the version number */ | |
625 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_1); | |
626 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
627 | if (crc_offset >= blk_size) | |
628 | goto invalid_cp1; | |
629 | ||
29e7043f | 630 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
631 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
632 | goto invalid_cp1; | |
633 | ||
d71b5564 | 634 | pre_version = cur_cp_version(cp_block); |
127e670a JK |
635 | |
636 | /* Read the 2nd cp block in this CP pack */ | |
25ca923b | 637 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
127e670a JK |
638 | cp_page_2 = get_meta_page(sbi, cp_addr); |
639 | ||
640 | cp_block = (struct f2fs_checkpoint *)page_address(cp_page_2); | |
641 | crc_offset = le32_to_cpu(cp_block->checksum_offset); | |
642 | if (crc_offset >= blk_size) | |
643 | goto invalid_cp2; | |
644 | ||
29e7043f | 645 | crc = le32_to_cpu(*((__le32 *)((unsigned char *)cp_block + crc_offset))); |
127e670a JK |
646 | if (!f2fs_crc_valid(crc, cp_block, crc_offset)) |
647 | goto invalid_cp2; | |
648 | ||
d71b5564 | 649 | cur_version = cur_cp_version(cp_block); |
127e670a JK |
650 | |
651 | if (cur_version == pre_version) { | |
652 | *version = cur_version; | |
653 | f2fs_put_page(cp_page_2, 1); | |
654 | return cp_page_1; | |
655 | } | |
656 | invalid_cp2: | |
657 | f2fs_put_page(cp_page_2, 1); | |
658 | invalid_cp1: | |
659 | f2fs_put_page(cp_page_1, 1); | |
660 | return NULL; | |
661 | } | |
662 | ||
663 | int get_valid_checkpoint(struct f2fs_sb_info *sbi) | |
664 | { | |
665 | struct f2fs_checkpoint *cp_block; | |
666 | struct f2fs_super_block *fsb = sbi->raw_super; | |
667 | struct page *cp1, *cp2, *cur_page; | |
668 | unsigned long blk_size = sbi->blocksize; | |
669 | unsigned long long cp1_version = 0, cp2_version = 0; | |
670 | unsigned long long cp_start_blk_no; | |
55141486 | 671 | unsigned int cp_blks = 1 + __cp_payload(sbi); |
1dbe4152 CL |
672 | block_t cp_blk_no; |
673 | int i; | |
127e670a | 674 | |
1dbe4152 | 675 | sbi->ckpt = kzalloc(cp_blks * blk_size, GFP_KERNEL); |
127e670a JK |
676 | if (!sbi->ckpt) |
677 | return -ENOMEM; | |
678 | /* | |
679 | * Finding out valid cp block involves read both | |
680 | * sets( cp pack1 and cp pack 2) | |
681 | */ | |
682 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
683 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
684 | ||
685 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
686 | cp_start_blk_no += ((unsigned long long)1) << |
687 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
688 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
689 | ||
690 | if (cp1 && cp2) { | |
691 | if (ver_after(cp2_version, cp1_version)) | |
692 | cur_page = cp2; | |
693 | else | |
694 | cur_page = cp1; | |
695 | } else if (cp1) { | |
696 | cur_page = cp1; | |
697 | } else if (cp2) { | |
698 | cur_page = cp2; | |
699 | } else { | |
700 | goto fail_no_cp; | |
701 | } | |
702 | ||
703 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
704 | memcpy(sbi->ckpt, cp_block, blk_size); | |
705 | ||
1dbe4152 CL |
706 | if (cp_blks <= 1) |
707 | goto done; | |
708 | ||
709 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
710 | if (cur_page == cp2) | |
711 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
712 | ||
713 | for (i = 1; i < cp_blks; i++) { | |
714 | void *sit_bitmap_ptr; | |
715 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
716 | ||
717 | cur_page = get_meta_page(sbi, cp_blk_no + i); | |
718 | sit_bitmap_ptr = page_address(cur_page); | |
719 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
720 | f2fs_put_page(cur_page, 1); | |
721 | } | |
722 | done: | |
127e670a JK |
723 | f2fs_put_page(cp1, 1); |
724 | f2fs_put_page(cp2, 1); | |
725 | return 0; | |
726 | ||
727 | fail_no_cp: | |
728 | kfree(sbi->ckpt); | |
729 | return -EINVAL; | |
730 | } | |
731 | ||
c227f912 | 732 | static void __add_dirty_inode(struct inode *inode, enum inode_type type) |
127e670a | 733 | { |
4081363f | 734 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
2710fd7e | 735 | struct f2fs_inode_info *fi = F2FS_I(inode); |
c227f912 | 736 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
127e670a | 737 | |
c227f912 | 738 | if (is_inode_flag_set(fi, flag)) |
2710fd7e | 739 | return; |
2d7b822a | 740 | |
c227f912 CY |
741 | set_inode_flag(fi, flag); |
742 | list_add_tail(&fi->dirty_list, &sbi->inode_list[type]); | |
33fbd510 | 743 | stat_inc_dirty_inode(sbi, type); |
5deb8267 JK |
744 | } |
745 | ||
c227f912 | 746 | static void __remove_dirty_inode(struct inode *inode, enum inode_type type) |
6ad7609a | 747 | { |
6ad7609a | 748 | struct f2fs_inode_info *fi = F2FS_I(inode); |
c227f912 | 749 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
6ad7609a CY |
750 | |
751 | if (get_dirty_pages(inode) || | |
c227f912 | 752 | !is_inode_flag_set(F2FS_I(inode), flag)) |
6ad7609a CY |
753 | return; |
754 | ||
755 | list_del_init(&fi->dirty_list); | |
c227f912 | 756 | clear_inode_flag(fi, flag); |
33fbd510 | 757 | stat_dec_dirty_inode(F2FS_I_SB(inode), type); |
6ad7609a CY |
758 | } |
759 | ||
a7ffdbe2 | 760 | void update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 761 | { |
4081363f | 762 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 763 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
5deb8267 | 764 | |
5ac9f36f CY |
765 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
766 | !S_ISLNK(inode->i_mode)) | |
127e670a | 767 | return; |
7bd59381 | 768 | |
c227f912 CY |
769 | spin_lock(&sbi->inode_lock[type]); |
770 | __add_dirty_inode(inode, type); | |
a7ffdbe2 | 771 | inode_inc_dirty_pages(inode); |
c227f912 | 772 | spin_unlock(&sbi->inode_lock[type]); |
cf0ee0f0 | 773 | |
a7ffdbe2 | 774 | SetPagePrivate(page); |
9e4ded3f | 775 | f2fs_trace_pid(page); |
5deb8267 JK |
776 | } |
777 | ||
778 | void add_dirty_dir_inode(struct inode *inode) | |
779 | { | |
4081363f | 780 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
127e670a | 781 | |
c227f912 CY |
782 | spin_lock(&sbi->inode_lock[DIR_INODE]); |
783 | __add_dirty_inode(inode, DIR_INODE); | |
784 | spin_unlock(&sbi->inode_lock[DIR_INODE]); | |
127e670a JK |
785 | } |
786 | ||
c227f912 | 787 | void remove_dirty_inode(struct inode *inode) |
127e670a | 788 | { |
4081363f | 789 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
2710fd7e | 790 | struct f2fs_inode_info *fi = F2FS_I(inode); |
c227f912 | 791 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
127e670a | 792 | |
c227f912 CY |
793 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
794 | !S_ISLNK(inode->i_mode)) | |
127e670a JK |
795 | return; |
796 | ||
c227f912 CY |
797 | spin_lock(&sbi->inode_lock[type]); |
798 | __remove_dirty_inode(inode, type); | |
799 | spin_unlock(&sbi->inode_lock[type]); | |
74d0b917 JK |
800 | |
801 | /* Only from the recovery routine */ | |
2710fd7e CY |
802 | if (is_inode_flag_set(fi, FI_DELAY_IPUT)) { |
803 | clear_inode_flag(fi, FI_DELAY_IPUT); | |
74d0b917 | 804 | iput(inode); |
afc3eda2 | 805 | } |
74d0b917 JK |
806 | } |
807 | ||
6d5a1495 | 808 | int sync_dirty_inodes(struct f2fs_sb_info *sbi, enum inode_type type) |
127e670a | 809 | { |
ce3b7d80 | 810 | struct list_head *head; |
127e670a | 811 | struct inode *inode; |
2710fd7e | 812 | struct f2fs_inode_info *fi; |
4cf18537 CY |
813 | bool is_dir = (type == DIR_INODE); |
814 | ||
815 | trace_f2fs_sync_dirty_inodes_enter(sbi->sb, is_dir, | |
816 | get_pages(sbi, is_dir ? | |
817 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
127e670a | 818 | retry: |
af41d3ee | 819 | if (unlikely(f2fs_cp_error(sbi))) |
6d5a1495 | 820 | return -EIO; |
af41d3ee | 821 | |
c227f912 | 822 | spin_lock(&sbi->inode_lock[type]); |
ce3b7d80 | 823 | |
c227f912 | 824 | head = &sbi->inode_list[type]; |
127e670a | 825 | if (list_empty(head)) { |
c227f912 | 826 | spin_unlock(&sbi->inode_lock[type]); |
4cf18537 CY |
827 | trace_f2fs_sync_dirty_inodes_exit(sbi->sb, is_dir, |
828 | get_pages(sbi, is_dir ? | |
829 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
6d5a1495 | 830 | return 0; |
127e670a | 831 | } |
2710fd7e CY |
832 | fi = list_entry(head->next, struct f2fs_inode_info, dirty_list); |
833 | inode = igrab(&fi->vfs_inode); | |
c227f912 | 834 | spin_unlock(&sbi->inode_lock[type]); |
127e670a | 835 | if (inode) { |
87d6f890 | 836 | filemap_fdatawrite(inode->i_mapping); |
127e670a JK |
837 | iput(inode); |
838 | } else { | |
839 | /* | |
840 | * We should submit bio, since it exists several | |
841 | * wribacking dentry pages in the freeing inode. | |
842 | */ | |
458e6197 | 843 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
7ecebe5e | 844 | cond_resched(); |
127e670a JK |
845 | } |
846 | goto retry; | |
847 | } | |
848 | ||
0a8165d7 | 849 | /* |
127e670a JK |
850 | * Freeze all the FS-operations for checkpoint. |
851 | */ | |
cf779cab | 852 | static int block_operations(struct f2fs_sb_info *sbi) |
127e670a | 853 | { |
127e670a JK |
854 | struct writeback_control wbc = { |
855 | .sync_mode = WB_SYNC_ALL, | |
856 | .nr_to_write = LONG_MAX, | |
857 | .for_reclaim = 0, | |
858 | }; | |
c718379b | 859 | struct blk_plug plug; |
cf779cab | 860 | int err = 0; |
c718379b JK |
861 | |
862 | blk_start_plug(&plug); | |
863 | ||
39936837 | 864 | retry_flush_dents: |
e479556b | 865 | f2fs_lock_all(sbi); |
127e670a | 866 | /* write all the dirty dentry pages */ |
127e670a | 867 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 868 | f2fs_unlock_all(sbi); |
6d5a1495 CY |
869 | err = sync_dirty_inodes(sbi, DIR_INODE); |
870 | if (err) | |
cf779cab | 871 | goto out; |
39936837 | 872 | goto retry_flush_dents; |
127e670a JK |
873 | } |
874 | ||
127e670a | 875 | /* |
e1c42045 | 876 | * POR: we should ensure that there are no dirty node pages |
127e670a JK |
877 | * until finishing nat/sit flush. |
878 | */ | |
39936837 | 879 | retry_flush_nodes: |
b3582c68 | 880 | down_write(&sbi->node_write); |
127e670a JK |
881 | |
882 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
b3582c68 | 883 | up_write(&sbi->node_write); |
6d5a1495 CY |
884 | err = sync_node_pages(sbi, 0, &wbc); |
885 | if (err) { | |
cf779cab | 886 | f2fs_unlock_all(sbi); |
cf779cab JK |
887 | goto out; |
888 | } | |
39936837 | 889 | goto retry_flush_nodes; |
127e670a | 890 | } |
cf779cab | 891 | out: |
c718379b | 892 | blk_finish_plug(&plug); |
cf779cab | 893 | return err; |
127e670a JK |
894 | } |
895 | ||
896 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
897 | { | |
b3582c68 | 898 | up_write(&sbi->node_write); |
e479556b | 899 | f2fs_unlock_all(sbi); |
127e670a JK |
900 | } |
901 | ||
fb51b5ef CL |
902 | static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi) |
903 | { | |
904 | DEFINE_WAIT(wait); | |
905 | ||
906 | for (;;) { | |
907 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
908 | ||
909 | if (!get_pages(sbi, F2FS_WRITEBACK)) | |
910 | break; | |
911 | ||
912 | io_schedule(); | |
913 | } | |
914 | finish_wait(&sbi->cp_wait, &wait); | |
915 | } | |
916 | ||
c34f42e2 | 917 | static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
918 | { |
919 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
cf2271e7 | 920 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); |
77041823 | 921 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
67298804 | 922 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; |
77041823 | 923 | nid_t last_nid = nm_i->next_scan_nid; |
127e670a | 924 | block_t start_blk; |
127e670a | 925 | unsigned int data_sum_blocks, orphan_blocks; |
7e586fa0 | 926 | __u32 crc32 = 0; |
127e670a | 927 | int i; |
55141486 | 928 | int cp_payload_blks = __cp_payload(sbi); |
e90c2d28 CY |
929 | block_t discard_blk = NEXT_FREE_BLKADDR(sbi, curseg); |
930 | bool invalidate = false; | |
8f1dbbbb SL |
931 | struct super_block *sb = sbi->sb; |
932 | struct curseg_info *seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); | |
933 | u64 kbytes_written; | |
127e670a | 934 | |
1e87a78d JK |
935 | /* |
936 | * This avoids to conduct wrong roll-forward operations and uses | |
937 | * metapages, so should be called prior to sync_meta_pages below. | |
938 | */ | |
e90c2d28 CY |
939 | if (discard_next_dnode(sbi, discard_blk)) |
940 | invalidate = true; | |
127e670a JK |
941 | |
942 | /* Flush all the NAT/SIT pages */ | |
cf779cab | 943 | while (get_pages(sbi, F2FS_DIRTY_META)) { |
127e670a | 944 | sync_meta_pages(sbi, META, LONG_MAX); |
cf779cab | 945 | if (unlikely(f2fs_cp_error(sbi))) |
c34f42e2 | 946 | return -EIO; |
cf779cab | 947 | } |
127e670a JK |
948 | |
949 | next_free_nid(sbi, &last_nid); | |
950 | ||
951 | /* | |
952 | * modify checkpoint | |
953 | * version number is already updated | |
954 | */ | |
955 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi)); | |
956 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
957 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); | |
b5b82205 | 958 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
959 | ckpt->cur_node_segno[i] = |
960 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
961 | ckpt->cur_node_blkoff[i] = | |
962 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
963 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
964 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
965 | } | |
b5b82205 | 966 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
967 | ckpt->cur_data_segno[i] = |
968 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
969 | ckpt->cur_data_blkoff[i] = | |
970 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
971 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
972 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
973 | } | |
974 | ||
975 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
976 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
977 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
978 | ||
979 | /* 2 cp + n data seg summary + orphan inode blocks */ | |
3fa06d7b | 980 | data_sum_blocks = npages_for_summary_flush(sbi, false); |
b5b82205 | 981 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
25ca923b | 982 | set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 983 | else |
25ca923b | 984 | clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 985 | |
67298804 | 986 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
987 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
988 | orphan_blocks); | |
127e670a | 989 | |
119ee914 | 990 | if (__remain_node_summaries(cpc->reason)) |
b5b82205 | 991 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
992 | cp_payload_blks + data_sum_blocks + |
993 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
119ee914 | 994 | else |
b5b82205 | 995 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS + |
1dbe4152 CL |
996 | cp_payload_blks + data_sum_blocks + |
997 | orphan_blocks); | |
119ee914 JK |
998 | |
999 | if (cpc->reason == CP_UMOUNT) | |
1000 | set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
1001 | else | |
1002 | clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
1003 | ||
1004 | if (cpc->reason == CP_FASTBOOT) | |
1005 | set_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
1006 | else | |
1007 | clear_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
127e670a | 1008 | |
67298804 | 1009 | if (orphan_num) |
25ca923b | 1010 | set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1011 | else |
25ca923b | 1012 | clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); |
127e670a | 1013 | |
caf0047e | 1014 | if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) |
2ae4c673 JK |
1015 | set_ckpt_flags(ckpt, CP_FSCK_FLAG); |
1016 | ||
127e670a JK |
1017 | /* update SIT/NAT bitmap */ |
1018 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
1019 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
1020 | ||
1021 | crc32 = f2fs_crc32(ckpt, le32_to_cpu(ckpt->checksum_offset)); | |
7e586fa0 JK |
1022 | *((__le32 *)((unsigned char *)ckpt + |
1023 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
1024 | = cpu_to_le32(crc32); |
1025 | ||
1026 | start_blk = __start_cp_addr(sbi); | |
1027 | ||
a7230d16 JK |
1028 | /* need to wait for end_io results */ |
1029 | wait_on_all_pages_writeback(sbi); | |
1030 | if (unlikely(f2fs_cp_error(sbi))) | |
c34f42e2 | 1031 | return -EIO; |
a7230d16 | 1032 | |
127e670a | 1033 | /* write out checkpoint buffer at block 0 */ |
381722d2 CY |
1034 | update_meta_page(sbi, ckpt, start_blk++); |
1035 | ||
1036 | for (i = 1; i < 1 + cp_payload_blks; i++) | |
1037 | update_meta_page(sbi, (char *)ckpt + i * F2FS_BLKSIZE, | |
1038 | start_blk++); | |
1dbe4152 | 1039 | |
67298804 | 1040 | if (orphan_num) { |
127e670a JK |
1041 | write_orphan_inodes(sbi, start_blk); |
1042 | start_blk += orphan_blocks; | |
1043 | } | |
1044 | ||
1045 | write_data_summaries(sbi, start_blk); | |
1046 | start_blk += data_sum_blocks; | |
8f1dbbbb SL |
1047 | |
1048 | /* Record write statistics in the hot node summary */ | |
1049 | kbytes_written = sbi->kbytes_written; | |
1050 | if (sb->s_bdev->bd_part) | |
1051 | kbytes_written += BD_PART_WRITTEN(sbi); | |
1052 | ||
1053 | seg_i->sum_blk->info.kbytes_written = cpu_to_le64(kbytes_written); | |
1054 | ||
119ee914 | 1055 | if (__remain_node_summaries(cpc->reason)) { |
127e670a JK |
1056 | write_node_summaries(sbi, start_blk); |
1057 | start_blk += NR_CURSEG_NODE_TYPE; | |
1058 | } | |
1059 | ||
1060 | /* writeout checkpoint block */ | |
381722d2 | 1061 | update_meta_page(sbi, ckpt, start_blk); |
127e670a JK |
1062 | |
1063 | /* wait for previous submitted node/meta pages writeback */ | |
fb51b5ef | 1064 | wait_on_all_pages_writeback(sbi); |
127e670a | 1065 | |
cf779cab | 1066 | if (unlikely(f2fs_cp_error(sbi))) |
c34f42e2 | 1067 | return -EIO; |
cf779cab | 1068 | |
4ef51a8f | 1069 | filemap_fdatawait_range(NODE_MAPPING(sbi), 0, LONG_MAX); |
9df27d98 | 1070 | filemap_fdatawait_range(META_MAPPING(sbi), 0, LONG_MAX); |
127e670a JK |
1071 | |
1072 | /* update user_block_counts */ | |
1073 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
1074 | sbi->alloc_valid_block_count = 0; | |
1075 | ||
1076 | /* Here, we only have one bio having CP pack */ | |
577e3495 | 1077 | sync_meta_pages(sbi, META_FLUSH, LONG_MAX); |
127e670a | 1078 | |
6a8f8ca5 JK |
1079 | /* wait for previous submitted meta pages writeback */ |
1080 | wait_on_all_pages_writeback(sbi); | |
1081 | ||
e90c2d28 CY |
1082 | /* |
1083 | * invalidate meta page which is used temporarily for zeroing out | |
1084 | * block at the end of warm node chain. | |
1085 | */ | |
1086 | if (invalidate) | |
1087 | invalidate_mapping_pages(META_MAPPING(sbi), discard_blk, | |
1088 | discard_blk); | |
1089 | ||
a49324f1 | 1090 | release_ino_entry(sbi); |
cf779cab JK |
1091 | |
1092 | if (unlikely(f2fs_cp_error(sbi))) | |
c34f42e2 | 1093 | return -EIO; |
cf779cab | 1094 | |
836b5a63 | 1095 | clear_prefree_segments(sbi, cpc); |
caf0047e | 1096 | clear_sbi_flag(sbi, SBI_IS_DIRTY); |
c34f42e2 CY |
1097 | |
1098 | return 0; | |
127e670a JK |
1099 | } |
1100 | ||
0a8165d7 | 1101 | /* |
e1c42045 | 1102 | * We guarantee that this checkpoint procedure will not fail. |
127e670a | 1103 | */ |
c34f42e2 | 1104 | int write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1105 | { |
1106 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1107 | unsigned long long ckpt_ver; | |
c34f42e2 | 1108 | int err = 0; |
127e670a | 1109 | |
43727527 | 1110 | mutex_lock(&sbi->cp_mutex); |
8501017e | 1111 | |
caf0047e | 1112 | if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) && |
a66cdd98 JK |
1113 | (cpc->reason == CP_FASTBOOT || cpc->reason == CP_SYNC || |
1114 | (cpc->reason == CP_DISCARD && !sbi->discard_blks))) | |
8501017e | 1115 | goto out; |
c34f42e2 CY |
1116 | if (unlikely(f2fs_cp_error(sbi))) { |
1117 | err = -EIO; | |
cf779cab | 1118 | goto out; |
c34f42e2 CY |
1119 | } |
1120 | if (f2fs_readonly(sbi->sb)) { | |
1121 | err = -EROFS; | |
11504a8e | 1122 | goto out; |
c34f42e2 | 1123 | } |
2bda542d WL |
1124 | |
1125 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); | |
1126 | ||
c34f42e2 CY |
1127 | err = block_operations(sbi); |
1128 | if (err) | |
cf779cab | 1129 | goto out; |
127e670a | 1130 | |
75ab4cb8 | 1131 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1132 | |
458e6197 JK |
1133 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
1134 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
1135 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
127e670a JK |
1136 | |
1137 | /* | |
1138 | * update checkpoint pack index | |
1139 | * Increase the version number so that | |
1140 | * SIT entries and seg summaries are written at correct place | |
1141 | */ | |
d71b5564 | 1142 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1143 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1144 | ||
1145 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
1146 | flush_nat_entries(sbi); | |
4b2fecc8 | 1147 | flush_sit_entries(sbi, cpc); |
127e670a | 1148 | |
127e670a | 1149 | /* unlock all the fs_lock[] in do_checkpoint() */ |
c34f42e2 | 1150 | err = do_checkpoint(sbi, cpc); |
127e670a JK |
1151 | |
1152 | unblock_operations(sbi); | |
942e0be6 | 1153 | stat_inc_cp_count(sbi->stat_info); |
10027551 JK |
1154 | |
1155 | if (cpc->reason == CP_RECOVERY) | |
1156 | f2fs_msg(sbi->sb, KERN_NOTICE, | |
1157 | "checkpoint: version = %llx", ckpt_ver); | |
60b99b48 JK |
1158 | |
1159 | /* do checkpoint periodically */ | |
6beceb54 | 1160 | f2fs_update_time(sbi, CP_TIME); |
55d1cdb2 | 1161 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
8501017e JK |
1162 | out: |
1163 | mutex_unlock(&sbi->cp_mutex); | |
c34f42e2 | 1164 | return err; |
127e670a JK |
1165 | } |
1166 | ||
6451e041 | 1167 | void init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1168 | { |
6451e041 JK |
1169 | int i; |
1170 | ||
1171 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1172 | struct inode_management *im = &sbi->im[i]; |
1173 | ||
1174 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1175 | spin_lock_init(&im->ino_lock); | |
1176 | INIT_LIST_HEAD(&im->ino_list); | |
1177 | im->ino_num = 0; | |
6451e041 JK |
1178 | } |
1179 | ||
b5b82205 | 1180 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
14b42817 WL |
1181 | NR_CURSEG_TYPE - __cp_payload(sbi)) * |
1182 | F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
1183 | } |
1184 | ||
6e6093a8 | 1185 | int __init create_checkpoint_caches(void) |
127e670a | 1186 | { |
6451e041 JK |
1187 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1188 | sizeof(struct ino_entry)); | |
1189 | if (!ino_entry_slab) | |
127e670a | 1190 | return -ENOMEM; |
06292073 CY |
1191 | inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry", |
1192 | sizeof(struct inode_entry)); | |
6bacf52f | 1193 | if (!inode_entry_slab) { |
6451e041 | 1194 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1195 | return -ENOMEM; |
1196 | } | |
1197 | return 0; | |
1198 | } | |
1199 | ||
1200 | void destroy_checkpoint_caches(void) | |
1201 | { | |
6451e041 | 1202 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1203 | kmem_cache_destroy(inode_entry_slab); |
1204 | } |