Commit | Line | Data |
---|---|---|
7c1a000d | 1 | // SPDX-License-Identifier: GPL-2.0 |
0a8165d7 | 2 | /* |
127e670a JK |
3 | * fs/f2fs/checkpoint.c |
4 | * | |
5 | * Copyright (c) 2012 Samsung Electronics Co., Ltd. | |
6 | * http://www.samsung.com/ | |
127e670a JK |
7 | */ |
8 | #include <linux/fs.h> | |
9 | #include <linux/bio.h> | |
10 | #include <linux/mpage.h> | |
11 | #include <linux/writeback.h> | |
12 | #include <linux/blkdev.h> | |
13 | #include <linux/f2fs_fs.h> | |
14 | #include <linux/pagevec.h> | |
15 | #include <linux/swap.h> | |
16 | ||
17 | #include "f2fs.h" | |
18 | #include "node.h" | |
19 | #include "segment.h" | |
9e4ded3f | 20 | #include "trace.h" |
2af4bd6c | 21 | #include <trace/events/f2fs.h> |
127e670a | 22 | |
6451e041 | 23 | static struct kmem_cache *ino_entry_slab; |
4d57b86d | 24 | struct kmem_cache *f2fs_inode_entry_slab; |
127e670a | 25 | |
38f91ca8 JK |
26 | void f2fs_stop_checkpoint(struct f2fs_sb_info *sbi, bool end_io) |
27 | { | |
d494500a | 28 | f2fs_build_fault_attr(sbi, 0, 0); |
aaec2b1d | 29 | set_ckpt_flags(sbi, CP_ERROR_FLAG); |
38f91ca8 | 30 | if (!end_io) |
b9109b0e | 31 | f2fs_flush_merged_writes(sbi); |
38f91ca8 JK |
32 | } |
33 | ||
0a8165d7 | 34 | /* |
127e670a JK |
35 | * We guarantee no failure on the returned page. |
36 | */ | |
4d57b86d | 37 | struct page *f2fs_grab_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) |
127e670a | 38 | { |
9df27d98 | 39 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a JK |
40 | struct page *page = NULL; |
41 | repeat: | |
300e129c | 42 | page = f2fs_grab_cache_page(mapping, index, false); |
127e670a JK |
43 | if (!page) { |
44 | cond_resched(); | |
45 | goto repeat; | |
46 | } | |
bae0ee7a | 47 | f2fs_wait_on_page_writeback(page, META, true, true); |
237c0790 JK |
48 | if (!PageUptodate(page)) |
49 | SetPageUptodate(page); | |
127e670a JK |
50 | return page; |
51 | } | |
52 | ||
0a8165d7 | 53 | /* |
127e670a JK |
54 | * We guarantee no failure on the returned page. |
55 | */ | |
2b947003 CY |
56 | static struct page *__get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index, |
57 | bool is_meta) | |
127e670a | 58 | { |
9df27d98 | 59 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a | 60 | struct page *page; |
cf04e8eb | 61 | struct f2fs_io_info fio = { |
05ca3632 | 62 | .sbi = sbi, |
cf04e8eb | 63 | .type = META, |
04d328de | 64 | .op = REQ_OP_READ, |
70fd7614 | 65 | .op_flags = REQ_META | REQ_PRIO, |
7a9d7548 CY |
66 | .old_blkaddr = index, |
67 | .new_blkaddr = index, | |
4375a336 | 68 | .encrypted_page = NULL, |
6dc3a126 | 69 | .is_por = !is_meta, |
cf04e8eb | 70 | }; |
7735730d | 71 | int err; |
2b947003 CY |
72 | |
73 | if (unlikely(!is_meta)) | |
04d328de | 74 | fio.op_flags &= ~REQ_META; |
127e670a | 75 | repeat: |
300e129c | 76 | page = f2fs_grab_cache_page(mapping, index, false); |
127e670a JK |
77 | if (!page) { |
78 | cond_resched(); | |
79 | goto repeat; | |
80 | } | |
393ff91f JK |
81 | if (PageUptodate(page)) |
82 | goto out; | |
83 | ||
05ca3632 JK |
84 | fio.page = page; |
85 | ||
7735730d CY |
86 | err = f2fs_submit_page_bio(&fio); |
87 | if (err) { | |
88 | f2fs_put_page(page, 1); | |
89 | return ERR_PTR(err); | |
86531d6b | 90 | } |
127e670a | 91 | |
393ff91f | 92 | lock_page(page); |
6bacf52f | 93 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
94 | f2fs_put_page(page, 1); |
95 | goto repeat; | |
96 | } | |
f3f338ca | 97 | |
81114baa | 98 | if (unlikely(!PageUptodate(page))) { |
7735730d CY |
99 | f2fs_put_page(page, 1); |
100 | return ERR_PTR(-EIO); | |
81114baa | 101 | } |
393ff91f | 102 | out: |
127e670a JK |
103 | return page; |
104 | } | |
105 | ||
4d57b86d | 106 | struct page *f2fs_get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) |
2b947003 CY |
107 | { |
108 | return __get_meta_page(sbi, index, true); | |
109 | } | |
110 | ||
7735730d CY |
111 | struct page *f2fs_get_meta_page_nofail(struct f2fs_sb_info *sbi, pgoff_t index) |
112 | { | |
113 | struct page *page; | |
114 | int count = 0; | |
115 | ||
116 | retry: | |
117 | page = __get_meta_page(sbi, index, true); | |
118 | if (IS_ERR(page)) { | |
119 | if (PTR_ERR(page) == -EIO && | |
120 | ++count <= DEFAULT_RETRY_IO_COUNT) | |
121 | goto retry; | |
7735730d | 122 | f2fs_stop_checkpoint(sbi, false); |
7735730d | 123 | } |
7735730d CY |
124 | return page; |
125 | } | |
126 | ||
2b947003 | 127 | /* for POR only */ |
4d57b86d | 128 | struct page *f2fs_get_tmp_page(struct f2fs_sb_info *sbi, pgoff_t index) |
2b947003 CY |
129 | { |
130 | return __get_meta_page(sbi, index, false); | |
131 | } | |
132 | ||
e1da7872 | 133 | bool f2fs_is_valid_blkaddr(struct f2fs_sb_info *sbi, |
4d57b86d | 134 | block_t blkaddr, int type) |
662befda CY |
135 | { |
136 | switch (type) { | |
137 | case META_NAT: | |
66b00c18 | 138 | break; |
662befda | 139 | case META_SIT: |
66b00c18 CY |
140 | if (unlikely(blkaddr >= SIT_BLK_CNT(sbi))) |
141 | return false; | |
142 | break; | |
81c1a0f1 | 143 | case META_SSA: |
66b00c18 CY |
144 | if (unlikely(blkaddr >= MAIN_BLKADDR(sbi) || |
145 | blkaddr < SM_I(sbi)->ssa_blkaddr)) | |
146 | return false; | |
147 | break; | |
662befda | 148 | case META_CP: |
66b00c18 CY |
149 | if (unlikely(blkaddr >= SIT_I(sbi)->sit_base_addr || |
150 | blkaddr < __start_cp_addr(sbi))) | |
151 | return false; | |
152 | break; | |
4c521f49 | 153 | case META_POR: |
e1da7872 | 154 | case DATA_GENERIC: |
66b00c18 | 155 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
c9b60788 CY |
156 | blkaddr < MAIN_BLKADDR(sbi))) { |
157 | if (type == DATA_GENERIC) { | |
158 | f2fs_msg(sbi->sb, KERN_WARNING, | |
159 | "access invalid blkaddr:%u", blkaddr); | |
160 | WARN_ON(1); | |
161 | } | |
66b00c18 | 162 | return false; |
c9b60788 | 163 | } |
66b00c18 | 164 | break; |
e1da7872 CY |
165 | case META_GENERIC: |
166 | if (unlikely(blkaddr < SEG0_BLKADDR(sbi) || | |
167 | blkaddr >= MAIN_BLKADDR(sbi))) | |
168 | return false; | |
169 | break; | |
662befda CY |
170 | default: |
171 | BUG(); | |
172 | } | |
66b00c18 CY |
173 | |
174 | return true; | |
662befda CY |
175 | } |
176 | ||
177 | /* | |
81c1a0f1 | 178 | * Readahead CP/NAT/SIT/SSA pages |
662befda | 179 | */ |
4d57b86d | 180 | int f2fs_ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages, |
26879fb1 | 181 | int type, bool sync) |
662befda | 182 | { |
662befda | 183 | struct page *page; |
4c521f49 | 184 | block_t blkno = start; |
662befda | 185 | struct f2fs_io_info fio = { |
05ca3632 | 186 | .sbi = sbi, |
662befda | 187 | .type = META, |
04d328de | 188 | .op = REQ_OP_READ, |
70fd7614 | 189 | .op_flags = sync ? (REQ_META | REQ_PRIO) : REQ_RAHEAD, |
4375a336 | 190 | .encrypted_page = NULL, |
fb830fc5 | 191 | .in_list = false, |
6dc3a126 | 192 | .is_por = (type == META_POR), |
662befda | 193 | }; |
e9f5b8b8 | 194 | struct blk_plug plug; |
662befda | 195 | |
2b947003 | 196 | if (unlikely(type == META_POR)) |
04d328de | 197 | fio.op_flags &= ~REQ_META; |
2b947003 | 198 | |
e9f5b8b8 | 199 | blk_start_plug(&plug); |
662befda | 200 | for (; nrpages-- > 0; blkno++) { |
662befda | 201 | |
e1da7872 | 202 | if (!f2fs_is_valid_blkaddr(sbi, blkno, type)) |
66b00c18 CY |
203 | goto out; |
204 | ||
662befda CY |
205 | switch (type) { |
206 | case META_NAT: | |
66b00c18 CY |
207 | if (unlikely(blkno >= |
208 | NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))) | |
662befda | 209 | blkno = 0; |
66b00c18 | 210 | /* get nat block addr */ |
7a9d7548 | 211 | fio.new_blkaddr = current_nat_addr(sbi, |
662befda CY |
212 | blkno * NAT_ENTRY_PER_BLOCK); |
213 | break; | |
214 | case META_SIT: | |
215 | /* get sit block addr */ | |
7a9d7548 | 216 | fio.new_blkaddr = current_sit_addr(sbi, |
662befda | 217 | blkno * SIT_ENTRY_PER_BLOCK); |
662befda | 218 | break; |
81c1a0f1 | 219 | case META_SSA: |
662befda | 220 | case META_CP: |
4c521f49 | 221 | case META_POR: |
7a9d7548 | 222 | fio.new_blkaddr = blkno; |
662befda CY |
223 | break; |
224 | default: | |
225 | BUG(); | |
226 | } | |
227 | ||
300e129c JK |
228 | page = f2fs_grab_cache_page(META_MAPPING(sbi), |
229 | fio.new_blkaddr, false); | |
662befda CY |
230 | if (!page) |
231 | continue; | |
232 | if (PageUptodate(page)) { | |
662befda CY |
233 | f2fs_put_page(page, 1); |
234 | continue; | |
235 | } | |
236 | ||
05ca3632 | 237 | fio.page = page; |
1919ffc0 | 238 | f2fs_submit_page_bio(&fio); |
662befda CY |
239 | f2fs_put_page(page, 0); |
240 | } | |
241 | out: | |
e9f5b8b8 | 242 | blk_finish_plug(&plug); |
662befda CY |
243 | return blkno - start; |
244 | } | |
245 | ||
4d57b86d | 246 | void f2fs_ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index) |
635aee1f CY |
247 | { |
248 | struct page *page; | |
249 | bool readahead = false; | |
250 | ||
251 | page = find_get_page(META_MAPPING(sbi), index); | |
4da7bf5a | 252 | if (!page || !PageUptodate(page)) |
635aee1f CY |
253 | readahead = true; |
254 | f2fs_put_page(page, 0); | |
255 | ||
256 | if (readahead) | |
4d57b86d | 257 | f2fs_ra_meta_pages(sbi, index, BIO_MAX_PAGES, META_POR, true); |
635aee1f CY |
258 | } |
259 | ||
b0af6d49 CY |
260 | static int __f2fs_write_meta_page(struct page *page, |
261 | struct writeback_control *wbc, | |
262 | enum iostat_type io_type) | |
127e670a | 263 | { |
4081363f | 264 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 265 | |
ecda0de3 CY |
266 | trace_f2fs_writepage(page, META); |
267 | ||
af697c0f JK |
268 | if (unlikely(f2fs_cp_error(sbi))) |
269 | goto redirty_out; | |
caf0047e | 270 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
cfb271d4 | 271 | goto redirty_out; |
857dc4e0 | 272 | if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0)) |
cfb271d4 | 273 | goto redirty_out; |
127e670a | 274 | |
4d57b86d | 275 | f2fs_do_write_meta_page(sbi, page, io_type); |
577e3495 | 276 | dec_page_count(sbi, F2FS_DIRTY_META); |
0c3a5797 CY |
277 | |
278 | if (wbc->for_reclaim) | |
bab475c5 | 279 | f2fs_submit_merged_write_cond(sbi, NULL, page, 0, META); |
0c3a5797 | 280 | |
577e3495 | 281 | unlock_page(page); |
857dc4e0 | 282 | |
0c3a5797 | 283 | if (unlikely(f2fs_cp_error(sbi))) |
b9109b0e | 284 | f2fs_submit_merged_write(sbi, META); |
0c3a5797 | 285 | |
577e3495 | 286 | return 0; |
cfb271d4 CY |
287 | |
288 | redirty_out: | |
76f60268 | 289 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 290 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
291 | } |
292 | ||
b0af6d49 CY |
293 | static int f2fs_write_meta_page(struct page *page, |
294 | struct writeback_control *wbc) | |
295 | { | |
296 | return __f2fs_write_meta_page(page, wbc, FS_META_IO); | |
297 | } | |
298 | ||
127e670a JK |
299 | static int f2fs_write_meta_pages(struct address_space *mapping, |
300 | struct writeback_control *wbc) | |
301 | { | |
4081363f | 302 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 303 | long diff, written; |
127e670a | 304 | |
0771fcc7 CY |
305 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
306 | goto skip_write; | |
307 | ||
5459aa97 | 308 | /* collect a number of dirty meta pages and write together */ |
812a9597 JK |
309 | if (wbc->sync_mode != WB_SYNC_ALL && |
310 | get_pages(sbi, F2FS_DIRTY_META) < | |
311 | nr_pages_to_skip(sbi, META)) | |
d3baf95d | 312 | goto skip_write; |
127e670a | 313 | |
a29d0e0b YH |
314 | /* if locked failed, cp will flush dirty pages instead */ |
315 | if (!mutex_trylock(&sbi->cp_mutex)) | |
316 | goto skip_write; | |
d31c7c3f | 317 | |
a29d0e0b | 318 | trace_f2fs_writepages(mapping->host, wbc, META); |
50c8cdb3 | 319 | diff = nr_pages_to_write(sbi, META, wbc); |
4d57b86d | 320 | written = f2fs_sync_meta_pages(sbi, META, wbc->nr_to_write, FS_META_IO); |
127e670a | 321 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 322 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 323 | return 0; |
d3baf95d JK |
324 | |
325 | skip_write: | |
326 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
d31c7c3f | 327 | trace_f2fs_writepages(mapping->host, wbc, META); |
d3baf95d | 328 | return 0; |
127e670a JK |
329 | } |
330 | ||
4d57b86d | 331 | long f2fs_sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, |
b0af6d49 | 332 | long nr_to_write, enum iostat_type io_type) |
127e670a | 333 | { |
9df27d98 | 334 | struct address_space *mapping = META_MAPPING(sbi); |
028a63a6 | 335 | pgoff_t index = 0, prev = ULONG_MAX; |
127e670a JK |
336 | struct pagevec pvec; |
337 | long nwritten = 0; | |
028a63a6 | 338 | int nr_pages; |
127e670a JK |
339 | struct writeback_control wbc = { |
340 | .for_reclaim = 0, | |
341 | }; | |
e9f5b8b8 | 342 | struct blk_plug plug; |
127e670a | 343 | |
86679820 | 344 | pagevec_init(&pvec); |
127e670a | 345 | |
e9f5b8b8 CY |
346 | blk_start_plug(&plug); |
347 | ||
028a63a6 | 348 | while ((nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, |
67fd707f | 349 | PAGECACHE_TAG_DIRTY))) { |
028a63a6 | 350 | int i; |
127e670a JK |
351 | |
352 | for (i = 0; i < nr_pages; i++) { | |
353 | struct page *page = pvec.pages[i]; | |
203681f6 | 354 | |
80dd9c0e | 355 | if (prev == ULONG_MAX) |
6066d8cd JK |
356 | prev = page->index - 1; |
357 | if (nr_to_write != LONG_MAX && page->index != prev + 1) { | |
358 | pagevec_release(&pvec); | |
359 | goto stop; | |
360 | } | |
361 | ||
127e670a | 362 | lock_page(page); |
203681f6 JK |
363 | |
364 | if (unlikely(page->mapping != mapping)) { | |
365 | continue_unlock: | |
366 | unlock_page(page); | |
367 | continue; | |
368 | } | |
369 | if (!PageDirty(page)) { | |
370 | /* someone wrote it for us */ | |
371 | goto continue_unlock; | |
372 | } | |
373 | ||
bae0ee7a | 374 | f2fs_wait_on_page_writeback(page, META, true, true); |
fa3d2bdf | 375 | |
203681f6 JK |
376 | if (!clear_page_dirty_for_io(page)) |
377 | goto continue_unlock; | |
378 | ||
b0af6d49 | 379 | if (__f2fs_write_meta_page(page, &wbc, io_type)) { |
577e3495 JK |
380 | unlock_page(page); |
381 | break; | |
382 | } | |
cfb271d4 | 383 | nwritten++; |
6066d8cd | 384 | prev = page->index; |
cfb271d4 | 385 | if (unlikely(nwritten >= nr_to_write)) |
127e670a JK |
386 | break; |
387 | } | |
388 | pagevec_release(&pvec); | |
389 | cond_resched(); | |
390 | } | |
6066d8cd | 391 | stop: |
127e670a | 392 | if (nwritten) |
b9109b0e | 393 | f2fs_submit_merged_write(sbi, type); |
127e670a | 394 | |
e9f5b8b8 CY |
395 | blk_finish_plug(&plug); |
396 | ||
127e670a JK |
397 | return nwritten; |
398 | } | |
399 | ||
400 | static int f2fs_set_meta_page_dirty(struct page *page) | |
401 | { | |
26c6b887 JK |
402 | trace_f2fs_set_page_dirty(page, META); |
403 | ||
237c0790 JK |
404 | if (!PageUptodate(page)) |
405 | SetPageUptodate(page); | |
127e670a | 406 | if (!PageDirty(page)) { |
b87078ad | 407 | __set_page_dirty_nobuffers(page); |
4081363f | 408 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); |
240a5915 | 409 | f2fs_set_page_private(page, 0); |
9e4ded3f | 410 | f2fs_trace_pid(page); |
127e670a JK |
411 | return 1; |
412 | } | |
413 | return 0; | |
414 | } | |
415 | ||
416 | const struct address_space_operations f2fs_meta_aops = { | |
417 | .writepage = f2fs_write_meta_page, | |
418 | .writepages = f2fs_write_meta_pages, | |
419 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
487261f3 CY |
420 | .invalidatepage = f2fs_invalidate_page, |
421 | .releasepage = f2fs_release_page, | |
5b7a487c WG |
422 | #ifdef CONFIG_MIGRATION |
423 | .migratepage = f2fs_migrate_page, | |
424 | #endif | |
127e670a JK |
425 | }; |
426 | ||
39d787be CY |
427 | static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, |
428 | unsigned int devidx, int type) | |
953e6cc6 | 429 | { |
67298804 | 430 | struct inode_management *im = &sbi->im[type]; |
80c54505 JK |
431 | struct ino_entry *e, *tmp; |
432 | ||
433 | tmp = f2fs_kmem_cache_alloc(ino_entry_slab, GFP_NOFS); | |
19526d74 | 434 | |
80c54505 | 435 | radix_tree_preload(GFP_NOFS | __GFP_NOFAIL); |
769ec6e5 | 436 | |
67298804 | 437 | spin_lock(&im->ino_lock); |
67298804 | 438 | e = radix_tree_lookup(&im->ino_root, ino); |
39efac41 | 439 | if (!e) { |
80c54505 | 440 | e = tmp; |
19526d74 CY |
441 | if (unlikely(radix_tree_insert(&im->ino_root, ino, e))) |
442 | f2fs_bug_on(sbi, 1); | |
443 | ||
39efac41 JK |
444 | memset(e, 0, sizeof(struct ino_entry)); |
445 | e->ino = ino; | |
953e6cc6 | 446 | |
67298804 | 447 | list_add_tail(&e->list, &im->ino_list); |
8c402946 | 448 | if (type != ORPHAN_INO) |
67298804 | 449 | im->ino_num++; |
39efac41 | 450 | } |
39d787be CY |
451 | |
452 | if (type == FLUSH_INO) | |
453 | f2fs_set_bit(devidx, (char *)&e->dirty_device); | |
454 | ||
67298804 | 455 | spin_unlock(&im->ino_lock); |
769ec6e5 | 456 | radix_tree_preload_end(); |
80c54505 JK |
457 | |
458 | if (e != tmp) | |
459 | kmem_cache_free(ino_entry_slab, tmp); | |
953e6cc6 JK |
460 | } |
461 | ||
6451e041 | 462 | static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 463 | { |
67298804 | 464 | struct inode_management *im = &sbi->im[type]; |
6451e041 | 465 | struct ino_entry *e; |
953e6cc6 | 466 | |
67298804 CY |
467 | spin_lock(&im->ino_lock); |
468 | e = radix_tree_lookup(&im->ino_root, ino); | |
39efac41 JK |
469 | if (e) { |
470 | list_del(&e->list); | |
67298804 CY |
471 | radix_tree_delete(&im->ino_root, ino); |
472 | im->ino_num--; | |
473 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
474 | kmem_cache_free(ino_entry_slab, e); |
475 | return; | |
953e6cc6 | 476 | } |
67298804 | 477 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
478 | } |
479 | ||
4d57b86d | 480 | void f2fs_add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
481 | { |
482 | /* add new dirty ino entry into list */ | |
39d787be | 483 | __add_ino_entry(sbi, ino, 0, type); |
fff04f90 JK |
484 | } |
485 | ||
4d57b86d | 486 | void f2fs_remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
487 | { |
488 | /* remove dirty ino entry from list */ | |
489 | __remove_ino_entry(sbi, ino, type); | |
490 | } | |
491 | ||
492 | /* mode should be APPEND_INO or UPDATE_INO */ | |
4d57b86d | 493 | bool f2fs_exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode) |
fff04f90 | 494 | { |
67298804 | 495 | struct inode_management *im = &sbi->im[mode]; |
fff04f90 | 496 | struct ino_entry *e; |
67298804 CY |
497 | |
498 | spin_lock(&im->ino_lock); | |
499 | e = radix_tree_lookup(&im->ino_root, ino); | |
500 | spin_unlock(&im->ino_lock); | |
fff04f90 JK |
501 | return e ? true : false; |
502 | } | |
503 | ||
4d57b86d | 504 | void f2fs_release_ino_entry(struct f2fs_sb_info *sbi, bool all) |
fff04f90 JK |
505 | { |
506 | struct ino_entry *e, *tmp; | |
507 | int i; | |
508 | ||
39d787be | 509 | for (i = all ? ORPHAN_INO : APPEND_INO; i < MAX_INO_ENTRY; i++) { |
67298804 CY |
510 | struct inode_management *im = &sbi->im[i]; |
511 | ||
512 | spin_lock(&im->ino_lock); | |
513 | list_for_each_entry_safe(e, tmp, &im->ino_list, list) { | |
fff04f90 | 514 | list_del(&e->list); |
67298804 | 515 | radix_tree_delete(&im->ino_root, e->ino); |
fff04f90 | 516 | kmem_cache_free(ino_entry_slab, e); |
67298804 | 517 | im->ino_num--; |
fff04f90 | 518 | } |
67298804 | 519 | spin_unlock(&im->ino_lock); |
fff04f90 JK |
520 | } |
521 | } | |
522 | ||
4d57b86d | 523 | void f2fs_set_dirty_device(struct f2fs_sb_info *sbi, nid_t ino, |
39d787be CY |
524 | unsigned int devidx, int type) |
525 | { | |
526 | __add_ino_entry(sbi, ino, devidx, type); | |
527 | } | |
528 | ||
4d57b86d | 529 | bool f2fs_is_dirty_device(struct f2fs_sb_info *sbi, nid_t ino, |
39d787be CY |
530 | unsigned int devidx, int type) |
531 | { | |
532 | struct inode_management *im = &sbi->im[type]; | |
533 | struct ino_entry *e; | |
534 | bool is_dirty = false; | |
535 | ||
536 | spin_lock(&im->ino_lock); | |
537 | e = radix_tree_lookup(&im->ino_root, ino); | |
538 | if (e && f2fs_test_bit(devidx, (char *)&e->dirty_device)) | |
539 | is_dirty = true; | |
540 | spin_unlock(&im->ino_lock); | |
541 | return is_dirty; | |
542 | } | |
543 | ||
4d57b86d | 544 | int f2fs_acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 545 | { |
67298804 | 546 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a JK |
547 | int err = 0; |
548 | ||
67298804 | 549 | spin_lock(&im->ino_lock); |
cb78942b | 550 | |
1ecc0c5c | 551 | if (time_to_inject(sbi, FAULT_ORPHAN)) { |
cb78942b | 552 | spin_unlock(&im->ino_lock); |
55523519 | 553 | f2fs_show_injection_info(FAULT_ORPHAN); |
cb78942b JK |
554 | return -ENOSPC; |
555 | } | |
7fa750a1 | 556 | |
67298804 | 557 | if (unlikely(im->ino_num >= sbi->max_orphans)) |
127e670a | 558 | err = -ENOSPC; |
cbd56e7d | 559 | else |
67298804 CY |
560 | im->ino_num++; |
561 | spin_unlock(&im->ino_lock); | |
0d47c1ad | 562 | |
127e670a JK |
563 | return err; |
564 | } | |
565 | ||
4d57b86d | 566 | void f2fs_release_orphan_inode(struct f2fs_sb_info *sbi) |
cbd56e7d | 567 | { |
67298804 CY |
568 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
569 | ||
570 | spin_lock(&im->ino_lock); | |
571 | f2fs_bug_on(sbi, im->ino_num == 0); | |
572 | im->ino_num--; | |
573 | spin_unlock(&im->ino_lock); | |
cbd56e7d JK |
574 | } |
575 | ||
4d57b86d | 576 | void f2fs_add_orphan_inode(struct inode *inode) |
127e670a | 577 | { |
39efac41 | 578 | /* add new orphan ino entry into list */ |
39d787be | 579 | __add_ino_entry(F2FS_I_SB(inode), inode->i_ino, 0, ORPHAN_INO); |
4d57b86d | 580 | f2fs_update_inode_page(inode); |
127e670a JK |
581 | } |
582 | ||
4d57b86d | 583 | void f2fs_remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 584 | { |
953e6cc6 | 585 | /* remove orphan entry from orphan list */ |
6451e041 | 586 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
587 | } |
588 | ||
8c14bfad | 589 | static int recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 590 | { |
8c14bfad | 591 | struct inode *inode; |
5905f9af | 592 | struct node_info ni; |
76a45e3c | 593 | int err; |
8c14bfad | 594 | |
5905f9af | 595 | inode = f2fs_iget_retry(sbi->sb, ino); |
8c14bfad CY |
596 | if (IS_ERR(inode)) { |
597 | /* | |
598 | * there should be a bug that we can't find the entry | |
599 | * to orphan inode. | |
600 | */ | |
601 | f2fs_bug_on(sbi, PTR_ERR(inode) == -ENOENT); | |
602 | return PTR_ERR(inode); | |
603 | } | |
604 | ||
0f9ec2a8 | 605 | err = dquot_initialize(inode); |
a515d12f SY |
606 | if (err) { |
607 | iput(inode); | |
0f9ec2a8 | 608 | goto err_out; |
a515d12f | 609 | } |
0f9ec2a8 | 610 | |
127e670a JK |
611 | clear_nlink(inode); |
612 | ||
613 | /* truncate all the data during iput */ | |
614 | iput(inode); | |
5905f9af | 615 | |
7735730d CY |
616 | err = f2fs_get_node_info(sbi, ino, &ni); |
617 | if (err) | |
618 | goto err_out; | |
5905f9af JK |
619 | |
620 | /* ENOMEM was fully retried in f2fs_evict_inode. */ | |
621 | if (ni.blk_addr != NULL_ADDR) { | |
0f9ec2a8 JK |
622 | err = -EIO; |
623 | goto err_out; | |
5905f9af | 624 | } |
8c14bfad | 625 | return 0; |
0f9ec2a8 JK |
626 | |
627 | err_out: | |
628 | set_sbi_flag(sbi, SBI_NEED_FSCK); | |
629 | f2fs_msg(sbi->sb, KERN_WARNING, | |
630 | "%s: orphan failed (ino=%x), run fsck to fix.", | |
631 | __func__, ino); | |
632 | return err; | |
127e670a JK |
633 | } |
634 | ||
4d57b86d | 635 | int f2fs_recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a | 636 | { |
3c642985 | 637 | block_t start_blk, orphan_blocks, i, j; |
4b2414d0 CY |
638 | unsigned int s_flags = sbi->sb->s_flags; |
639 | int err = 0; | |
ea676733 JK |
640 | #ifdef CONFIG_QUOTA |
641 | int quota_enabled; | |
642 | #endif | |
127e670a | 643 | |
aaec2b1d | 644 | if (!is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG)) |
8c14bfad | 645 | return 0; |
127e670a | 646 | |
1751e8a6 | 647 | if (s_flags & SB_RDONLY) { |
4b2414d0 | 648 | f2fs_msg(sbi->sb, KERN_INFO, "orphan cleanup on readonly fs"); |
1751e8a6 | 649 | sbi->sb->s_flags &= ~SB_RDONLY; |
4b2414d0 CY |
650 | } |
651 | ||
652 | #ifdef CONFIG_QUOTA | |
653 | /* Needed for iput() to work correctly and not trash data */ | |
1751e8a6 | 654 | sbi->sb->s_flags |= SB_ACTIVE; |
ea676733 | 655 | |
76cf05d7 SY |
656 | /* |
657 | * Turn on quotas which were not enabled for read-only mounts if | |
658 | * filesystem has quota feature, so that they are updated correctly. | |
659 | */ | |
1751e8a6 | 660 | quota_enabled = f2fs_enable_quota_files(sbi, s_flags & SB_RDONLY); |
4b2414d0 CY |
661 | #endif |
662 | ||
55141486 | 663 | start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi); |
3c642985 | 664 | orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi); |
127e670a | 665 | |
4d57b86d | 666 | f2fs_ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP, true); |
662befda | 667 | |
3c642985 | 668 | for (i = 0; i < orphan_blocks; i++) { |
7735730d | 669 | struct page *page; |
127e670a JK |
670 | struct f2fs_orphan_block *orphan_blk; |
671 | ||
7735730d CY |
672 | page = f2fs_get_meta_page(sbi, start_blk + i); |
673 | if (IS_ERR(page)) { | |
674 | err = PTR_ERR(page); | |
675 | goto out; | |
676 | } | |
677 | ||
127e670a JK |
678 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); |
679 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
680 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
8c14bfad CY |
681 | err = recover_orphan_inode(sbi, ino); |
682 | if (err) { | |
683 | f2fs_put_page(page, 1); | |
4b2414d0 | 684 | goto out; |
8c14bfad | 685 | } |
127e670a JK |
686 | } |
687 | f2fs_put_page(page, 1); | |
688 | } | |
689 | /* clear Orphan Flag */ | |
aaec2b1d | 690 | clear_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG); |
4b2414d0 | 691 | out: |
1378752b CY |
692 | set_sbi_flag(sbi, SBI_IS_RECOVERED); |
693 | ||
4b2414d0 CY |
694 | #ifdef CONFIG_QUOTA |
695 | /* Turn quotas off */ | |
ea676733 JK |
696 | if (quota_enabled) |
697 | f2fs_quota_off_umount(sbi->sb); | |
4b2414d0 | 698 | #endif |
1751e8a6 | 699 | sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ |
4b2414d0 CY |
700 | |
701 | return err; | |
127e670a JK |
702 | } |
703 | ||
704 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
705 | { | |
502c6e0b | 706 | struct list_head *head; |
127e670a | 707 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 708 | unsigned int nentries = 0; |
bd936f84 | 709 | unsigned short index = 1; |
8c402946 | 710 | unsigned short orphan_blocks; |
4531929e | 711 | struct page *page = NULL; |
6451e041 | 712 | struct ino_entry *orphan = NULL; |
67298804 | 713 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 714 | |
67298804 | 715 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 716 | |
d6c67a4f JK |
717 | /* |
718 | * we don't need to do spin_lock(&im->ino_lock) here, since all the | |
719 | * orphan inode operations are covered under f2fs_lock_op(). | |
720 | * And, spin_lock should be avoided due to page operations below. | |
721 | */ | |
67298804 | 722 | head = &im->ino_list; |
127e670a JK |
723 | |
724 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
725 | list_for_each_entry(orphan, head, list) { |
726 | if (!page) { | |
4d57b86d | 727 | page = f2fs_grab_meta_page(sbi, start_blk++); |
502c6e0b GZ |
728 | orphan_blk = |
729 | (struct f2fs_orphan_block *)page_address(page); | |
730 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
731 | } | |
127e670a | 732 | |
36795567 | 733 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 734 | |
36795567 | 735 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
736 | /* |
737 | * an orphan block is full of 1020 entries, | |
738 | * then we need to flush current orphan blocks | |
739 | * and bring another one in memory | |
740 | */ | |
741 | orphan_blk->blk_addr = cpu_to_le16(index); | |
742 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
743 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
744 | set_page_dirty(page); | |
745 | f2fs_put_page(page, 1); | |
746 | index++; | |
127e670a JK |
747 | nentries = 0; |
748 | page = NULL; | |
749 | } | |
502c6e0b | 750 | } |
127e670a | 751 | |
502c6e0b GZ |
752 | if (page) { |
753 | orphan_blk->blk_addr = cpu_to_le16(index); | |
754 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
755 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
756 | set_page_dirty(page); | |
757 | f2fs_put_page(page, 1); | |
127e670a | 758 | } |
127e670a JK |
759 | } |
760 | ||
d7eb8f1c CY |
761 | static __u32 f2fs_checkpoint_chksum(struct f2fs_sb_info *sbi, |
762 | struct f2fs_checkpoint *ckpt) | |
763 | { | |
764 | unsigned int chksum_ofs = le32_to_cpu(ckpt->checksum_offset); | |
765 | __u32 chksum; | |
766 | ||
767 | chksum = f2fs_crc32(sbi, ckpt, chksum_ofs); | |
768 | if (chksum_ofs < CP_CHKSUM_OFFSET) { | |
769 | chksum_ofs += sizeof(chksum); | |
770 | chksum = f2fs_chksum(sbi, chksum, (__u8 *)ckpt + chksum_ofs, | |
771 | F2FS_BLKSIZE - chksum_ofs); | |
772 | } | |
773 | return chksum; | |
774 | } | |
775 | ||
fc0065ad TY |
776 | static int get_checkpoint_version(struct f2fs_sb_info *sbi, block_t cp_addr, |
777 | struct f2fs_checkpoint **cp_block, struct page **cp_page, | |
778 | unsigned long long *version) | |
127e670a | 779 | { |
fc0065ad | 780 | size_t crc_offset = 0; |
d7eb8f1c | 781 | __u32 crc; |
127e670a | 782 | |
4d57b86d | 783 | *cp_page = f2fs_get_meta_page(sbi, cp_addr); |
7735730d CY |
784 | if (IS_ERR(*cp_page)) |
785 | return PTR_ERR(*cp_page); | |
786 | ||
fc0065ad | 787 | *cp_block = (struct f2fs_checkpoint *)page_address(*cp_page); |
127e670a | 788 | |
fc0065ad | 789 | crc_offset = le32_to_cpu((*cp_block)->checksum_offset); |
d7eb8f1c CY |
790 | if (crc_offset < CP_MIN_CHKSUM_OFFSET || |
791 | crc_offset > CP_CHKSUM_OFFSET) { | |
d3f07c04 | 792 | f2fs_put_page(*cp_page, 1); |
fc0065ad TY |
793 | f2fs_msg(sbi->sb, KERN_WARNING, |
794 | "invalid crc_offset: %zu", crc_offset); | |
795 | return -EINVAL; | |
796 | } | |
127e670a | 797 | |
b471eb99 CY |
798 | if (__is_set_ckpt_flags(*cp_block, CP_LARGE_NAT_BITMAP_FLAG)) { |
799 | if (crc_offset != CP_MIN_CHKSUM_OFFSET) { | |
800 | f2fs_put_page(*cp_page, 1); | |
801 | f2fs_msg(sbi->sb, KERN_WARNING, | |
802 | "layout of large_nat_bitmap is deprecated, " | |
803 | "run fsck to repair, chksum_offset: %zu", | |
804 | crc_offset); | |
805 | return -EINVAL; | |
806 | } | |
807 | } | |
808 | ||
d7eb8f1c CY |
809 | crc = f2fs_checkpoint_chksum(sbi, *cp_block); |
810 | if (crc != cur_cp_crc(*cp_block)) { | |
d3f07c04 | 811 | f2fs_put_page(*cp_page, 1); |
fc0065ad TY |
812 | f2fs_msg(sbi->sb, KERN_WARNING, "invalid crc value"); |
813 | return -EINVAL; | |
814 | } | |
127e670a | 815 | |
fc0065ad TY |
816 | *version = cur_cp_version(*cp_block); |
817 | return 0; | |
818 | } | |
127e670a | 819 | |
fc0065ad TY |
820 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, |
821 | block_t cp_addr, unsigned long long *version) | |
822 | { | |
823 | struct page *cp_page_1 = NULL, *cp_page_2 = NULL; | |
824 | struct f2fs_checkpoint *cp_block = NULL; | |
825 | unsigned long long cur_version = 0, pre_version = 0; | |
826 | int err; | |
127e670a | 827 | |
fc0065ad TY |
828 | err = get_checkpoint_version(sbi, cp_addr, &cp_block, |
829 | &cp_page_1, version); | |
830 | if (err) | |
d3f07c04 | 831 | return NULL; |
c9b60788 CY |
832 | |
833 | if (le32_to_cpu(cp_block->cp_pack_total_block_count) > | |
834 | sbi->blocks_per_seg) { | |
835 | f2fs_msg(sbi->sb, KERN_WARNING, | |
836 | "invalid cp_pack_total_block_count:%u", | |
837 | le32_to_cpu(cp_block->cp_pack_total_block_count)); | |
d3f07c04 | 838 | goto invalid_cp; |
c9b60788 | 839 | } |
fc0065ad | 840 | pre_version = *version; |
127e670a | 841 | |
fc0065ad TY |
842 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
843 | err = get_checkpoint_version(sbi, cp_addr, &cp_block, | |
844 | &cp_page_2, version); | |
845 | if (err) | |
d3f07c04 | 846 | goto invalid_cp; |
fc0065ad | 847 | cur_version = *version; |
127e670a JK |
848 | |
849 | if (cur_version == pre_version) { | |
850 | *version = cur_version; | |
851 | f2fs_put_page(cp_page_2, 1); | |
852 | return cp_page_1; | |
853 | } | |
127e670a | 854 | f2fs_put_page(cp_page_2, 1); |
d3f07c04 | 855 | invalid_cp: |
127e670a JK |
856 | f2fs_put_page(cp_page_1, 1); |
857 | return NULL; | |
858 | } | |
859 | ||
4d57b86d | 860 | int f2fs_get_valid_checkpoint(struct f2fs_sb_info *sbi) |
127e670a JK |
861 | { |
862 | struct f2fs_checkpoint *cp_block; | |
863 | struct f2fs_super_block *fsb = sbi->raw_super; | |
864 | struct page *cp1, *cp2, *cur_page; | |
865 | unsigned long blk_size = sbi->blocksize; | |
866 | unsigned long long cp1_version = 0, cp2_version = 0; | |
867 | unsigned long long cp_start_blk_no; | |
55141486 | 868 | unsigned int cp_blks = 1 + __cp_payload(sbi); |
1dbe4152 CL |
869 | block_t cp_blk_no; |
870 | int i; | |
127e670a | 871 | |
026f0507 KC |
872 | sbi->ckpt = f2fs_kzalloc(sbi, array_size(blk_size, cp_blks), |
873 | GFP_KERNEL); | |
127e670a JK |
874 | if (!sbi->ckpt) |
875 | return -ENOMEM; | |
876 | /* | |
877 | * Finding out valid cp block involves read both | |
878 | * sets( cp pack1 and cp pack 2) | |
879 | */ | |
880 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
881 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
882 | ||
883 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
884 | cp_start_blk_no += ((unsigned long long)1) << |
885 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
886 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
887 | ||
888 | if (cp1 && cp2) { | |
889 | if (ver_after(cp2_version, cp1_version)) | |
890 | cur_page = cp2; | |
891 | else | |
892 | cur_page = cp1; | |
893 | } else if (cp1) { | |
894 | cur_page = cp1; | |
895 | } else if (cp2) { | |
896 | cur_page = cp2; | |
897 | } else { | |
898 | goto fail_no_cp; | |
899 | } | |
900 | ||
901 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
902 | memcpy(sbi->ckpt, cp_block, blk_size); | |
903 | ||
8508e44a JK |
904 | if (cur_page == cp1) |
905 | sbi->cur_cp_pack = 1; | |
906 | else | |
907 | sbi->cur_cp_pack = 2; | |
984ec63c | 908 | |
e494c2f9 CY |
909 | /* Sanity checking of checkpoint */ |
910 | if (f2fs_sanity_check_ckpt(sbi)) | |
911 | goto free_fail_no_cp; | |
912 | ||
1dbe4152 CL |
913 | if (cp_blks <= 1) |
914 | goto done; | |
915 | ||
916 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
917 | if (cur_page == cp2) | |
918 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
919 | ||
920 | for (i = 1; i < cp_blks; i++) { | |
921 | void *sit_bitmap_ptr; | |
922 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
923 | ||
4d57b86d | 924 | cur_page = f2fs_get_meta_page(sbi, cp_blk_no + i); |
7735730d CY |
925 | if (IS_ERR(cur_page)) |
926 | goto free_fail_no_cp; | |
1dbe4152 CL |
927 | sit_bitmap_ptr = page_address(cur_page); |
928 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
929 | f2fs_put_page(cur_page, 1); | |
930 | } | |
931 | done: | |
127e670a JK |
932 | f2fs_put_page(cp1, 1); |
933 | f2fs_put_page(cp2, 1); | |
934 | return 0; | |
935 | ||
a2125ff7 JK |
936 | free_fail_no_cp: |
937 | f2fs_put_page(cp1, 1); | |
938 | f2fs_put_page(cp2, 1); | |
127e670a | 939 | fail_no_cp: |
5222595d | 940 | kvfree(sbi->ckpt); |
127e670a JK |
941 | return -EINVAL; |
942 | } | |
943 | ||
c227f912 | 944 | static void __add_dirty_inode(struct inode *inode, enum inode_type type) |
127e670a | 945 | { |
4081363f | 946 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 947 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
127e670a | 948 | |
91942321 | 949 | if (is_inode_flag_set(inode, flag)) |
2710fd7e | 950 | return; |
2d7b822a | 951 | |
91942321 | 952 | set_inode_flag(inode, flag); |
99f4b917 CY |
953 | if (!f2fs_is_volatile_file(inode)) |
954 | list_add_tail(&F2FS_I(inode)->dirty_list, | |
955 | &sbi->inode_list[type]); | |
33fbd510 | 956 | stat_inc_dirty_inode(sbi, type); |
5deb8267 JK |
957 | } |
958 | ||
c227f912 | 959 | static void __remove_dirty_inode(struct inode *inode, enum inode_type type) |
6ad7609a | 960 | { |
c227f912 | 961 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
6ad7609a | 962 | |
91942321 | 963 | if (get_dirty_pages(inode) || !is_inode_flag_set(inode, flag)) |
6ad7609a CY |
964 | return; |
965 | ||
91942321 JK |
966 | list_del_init(&F2FS_I(inode)->dirty_list); |
967 | clear_inode_flag(inode, flag); | |
33fbd510 | 968 | stat_dec_dirty_inode(F2FS_I_SB(inode), type); |
6ad7609a CY |
969 | } |
970 | ||
4d57b86d | 971 | void f2fs_update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 972 | { |
4081363f | 973 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 974 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
5deb8267 | 975 | |
5ac9f36f CY |
976 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
977 | !S_ISLNK(inode->i_mode)) | |
127e670a | 978 | return; |
7bd59381 | 979 | |
1c4bf763 JK |
980 | spin_lock(&sbi->inode_lock[type]); |
981 | if (type != FILE_INODE || test_opt(sbi, DATA_FLUSH)) | |
10aa97c3 | 982 | __add_dirty_inode(inode, type); |
b951a4ec | 983 | inode_inc_dirty_pages(inode); |
1c4bf763 JK |
984 | spin_unlock(&sbi->inode_lock[type]); |
985 | ||
240a5915 | 986 | f2fs_set_page_private(page, 0); |
9e4ded3f | 987 | f2fs_trace_pid(page); |
5deb8267 JK |
988 | } |
989 | ||
4d57b86d | 990 | void f2fs_remove_dirty_inode(struct inode *inode) |
127e670a | 991 | { |
4081363f | 992 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 993 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
127e670a | 994 | |
c227f912 CY |
995 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
996 | !S_ISLNK(inode->i_mode)) | |
127e670a JK |
997 | return; |
998 | ||
10aa97c3 JK |
999 | if (type == FILE_INODE && !test_opt(sbi, DATA_FLUSH)) |
1000 | return; | |
1001 | ||
c227f912 CY |
1002 | spin_lock(&sbi->inode_lock[type]); |
1003 | __remove_dirty_inode(inode, type); | |
1004 | spin_unlock(&sbi->inode_lock[type]); | |
74d0b917 JK |
1005 | } |
1006 | ||
4d57b86d | 1007 | int f2fs_sync_dirty_inodes(struct f2fs_sb_info *sbi, enum inode_type type) |
127e670a | 1008 | { |
ce3b7d80 | 1009 | struct list_head *head; |
127e670a | 1010 | struct inode *inode; |
2710fd7e | 1011 | struct f2fs_inode_info *fi; |
4cf18537 | 1012 | bool is_dir = (type == DIR_INODE); |
4db08d01 | 1013 | unsigned long ino = 0; |
4cf18537 CY |
1014 | |
1015 | trace_f2fs_sync_dirty_inodes_enter(sbi->sb, is_dir, | |
1016 | get_pages(sbi, is_dir ? | |
1017 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
127e670a | 1018 | retry: |
af41d3ee | 1019 | if (unlikely(f2fs_cp_error(sbi))) |
6d5a1495 | 1020 | return -EIO; |
af41d3ee | 1021 | |
c227f912 | 1022 | spin_lock(&sbi->inode_lock[type]); |
ce3b7d80 | 1023 | |
c227f912 | 1024 | head = &sbi->inode_list[type]; |
127e670a | 1025 | if (list_empty(head)) { |
c227f912 | 1026 | spin_unlock(&sbi->inode_lock[type]); |
4cf18537 CY |
1027 | trace_f2fs_sync_dirty_inodes_exit(sbi->sb, is_dir, |
1028 | get_pages(sbi, is_dir ? | |
1029 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
6d5a1495 | 1030 | return 0; |
127e670a | 1031 | } |
939afa94 | 1032 | fi = list_first_entry(head, struct f2fs_inode_info, dirty_list); |
2710fd7e | 1033 | inode = igrab(&fi->vfs_inode); |
c227f912 | 1034 | spin_unlock(&sbi->inode_lock[type]); |
127e670a | 1035 | if (inode) { |
4db08d01 JK |
1036 | unsigned long cur_ino = inode->i_ino; |
1037 | ||
186857c5 | 1038 | F2FS_I(inode)->cp_task = current; |
b0af6d49 | 1039 | |
87d6f890 | 1040 | filemap_fdatawrite(inode->i_mapping); |
b0af6d49 | 1041 | |
186857c5 | 1042 | F2FS_I(inode)->cp_task = NULL; |
b0af6d49 | 1043 | |
127e670a | 1044 | iput(inode); |
4db08d01 | 1045 | /* We need to give cpu to another writers. */ |
2a63531a | 1046 | if (ino == cur_ino) |
4db08d01 | 1047 | cond_resched(); |
2a63531a | 1048 | else |
4db08d01 | 1049 | ino = cur_ino; |
127e670a JK |
1050 | } else { |
1051 | /* | |
1052 | * We should submit bio, since it exists several | |
1053 | * wribacking dentry pages in the freeing inode. | |
1054 | */ | |
b9109b0e | 1055 | f2fs_submit_merged_write(sbi, DATA); |
7ecebe5e | 1056 | cond_resched(); |
127e670a JK |
1057 | } |
1058 | goto retry; | |
1059 | } | |
1060 | ||
0f18b462 JK |
1061 | int f2fs_sync_inode_meta(struct f2fs_sb_info *sbi) |
1062 | { | |
1063 | struct list_head *head = &sbi->inode_list[DIRTY_META]; | |
1064 | struct inode *inode; | |
1065 | struct f2fs_inode_info *fi; | |
1066 | s64 total = get_pages(sbi, F2FS_DIRTY_IMETA); | |
1067 | ||
1068 | while (total--) { | |
1069 | if (unlikely(f2fs_cp_error(sbi))) | |
1070 | return -EIO; | |
1071 | ||
1072 | spin_lock(&sbi->inode_lock[DIRTY_META]); | |
1073 | if (list_empty(head)) { | |
1074 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
1075 | return 0; | |
1076 | } | |
939afa94 | 1077 | fi = list_first_entry(head, struct f2fs_inode_info, |
0f18b462 JK |
1078 | gdirty_list); |
1079 | inode = igrab(&fi->vfs_inode); | |
1080 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
1081 | if (inode) { | |
18340edc JK |
1082 | sync_inode_metadata(inode, 0); |
1083 | ||
1084 | /* it's on eviction */ | |
1085 | if (is_inode_flag_set(inode, FI_DIRTY_INODE)) | |
4d57b86d | 1086 | f2fs_update_inode_page(inode); |
0f18b462 JK |
1087 | iput(inode); |
1088 | } | |
dee668c1 | 1089 | } |
0f18b462 JK |
1090 | return 0; |
1091 | } | |
1092 | ||
59c9081b YH |
1093 | static void __prepare_cp_block(struct f2fs_sb_info *sbi) |
1094 | { | |
1095 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1096 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1097 | nid_t last_nid = nm_i->next_scan_nid; | |
1098 | ||
1099 | next_free_nid(sbi, &last_nid); | |
1100 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
1101 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
1102 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
1103 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
1104 | } | |
1105 | ||
af033b2a CY |
1106 | static bool __need_flush_quota(struct f2fs_sb_info *sbi) |
1107 | { | |
1108 | if (!is_journalled_quota(sbi)) | |
1109 | return false; | |
1110 | if (is_sbi_flag_set(sbi, SBI_QUOTA_SKIP_FLUSH)) | |
1111 | return false; | |
1112 | if (is_sbi_flag_set(sbi, SBI_QUOTA_NEED_REPAIR)) | |
1113 | return false; | |
1114 | if (is_sbi_flag_set(sbi, SBI_QUOTA_NEED_FLUSH)) | |
1115 | return true; | |
1116 | if (get_pages(sbi, F2FS_DIRTY_QDATA)) | |
1117 | return true; | |
1118 | return false; | |
1119 | } | |
1120 | ||
0a8165d7 | 1121 | /* |
127e670a JK |
1122 | * Freeze all the FS-operations for checkpoint. |
1123 | */ | |
cf779cab | 1124 | static int block_operations(struct f2fs_sb_info *sbi) |
127e670a | 1125 | { |
127e670a JK |
1126 | struct writeback_control wbc = { |
1127 | .sync_mode = WB_SYNC_ALL, | |
1128 | .nr_to_write = LONG_MAX, | |
1129 | .for_reclaim = 0, | |
1130 | }; | |
c718379b | 1131 | struct blk_plug plug; |
af033b2a | 1132 | int err = 0, cnt = 0; |
c718379b JK |
1133 | |
1134 | blk_start_plug(&plug); | |
1135 | ||
af033b2a CY |
1136 | retry_flush_quotas: |
1137 | if (__need_flush_quota(sbi)) { | |
1138 | int locked; | |
1139 | ||
1140 | if (++cnt > DEFAULT_RETRY_QUOTA_FLUSH_COUNT) { | |
1141 | set_sbi_flag(sbi, SBI_QUOTA_SKIP_FLUSH); | |
1142 | f2fs_lock_all(sbi); | |
1143 | goto retry_flush_dents; | |
1144 | } | |
1145 | clear_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); | |
1146 | ||
1147 | /* only failed during mount/umount/freeze/quotactl */ | |
1148 | locked = down_read_trylock(&sbi->sb->s_umount); | |
1149 | f2fs_quota_sync(sbi->sb, -1); | |
1150 | if (locked) | |
1151 | up_read(&sbi->sb->s_umount); | |
1152 | } | |
1153 | ||
e479556b | 1154 | f2fs_lock_all(sbi); |
af033b2a CY |
1155 | if (__need_flush_quota(sbi)) { |
1156 | f2fs_unlock_all(sbi); | |
1157 | cond_resched(); | |
1158 | goto retry_flush_quotas; | |
1159 | } | |
1160 | ||
1161 | retry_flush_dents: | |
127e670a | 1162 | /* write all the dirty dentry pages */ |
127e670a | 1163 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 1164 | f2fs_unlock_all(sbi); |
4d57b86d | 1165 | err = f2fs_sync_dirty_inodes(sbi, DIR_INODE); |
6d5a1495 | 1166 | if (err) |
cf779cab | 1167 | goto out; |
30973883 | 1168 | cond_resched(); |
af033b2a | 1169 | goto retry_flush_quotas; |
127e670a JK |
1170 | } |
1171 | ||
59c9081b YH |
1172 | /* |
1173 | * POR: we should ensure that there are no dirty node pages | |
1174 | * until finishing nat/sit flush. inode->i_blocks can be updated. | |
1175 | */ | |
1176 | down_write(&sbi->node_change); | |
1177 | ||
af033b2a CY |
1178 | if (__need_flush_quota(sbi)) { |
1179 | up_write(&sbi->node_change); | |
1180 | f2fs_unlock_all(sbi); | |
1181 | goto retry_flush_quotas; | |
1182 | } | |
1183 | ||
0f18b462 | 1184 | if (get_pages(sbi, F2FS_DIRTY_IMETA)) { |
59c9081b | 1185 | up_write(&sbi->node_change); |
0f18b462 JK |
1186 | f2fs_unlock_all(sbi); |
1187 | err = f2fs_sync_inode_meta(sbi); | |
1188 | if (err) | |
1189 | goto out; | |
30973883 | 1190 | cond_resched(); |
af033b2a | 1191 | goto retry_flush_quotas; |
0f18b462 JK |
1192 | } |
1193 | ||
39936837 | 1194 | retry_flush_nodes: |
b3582c68 | 1195 | down_write(&sbi->node_write); |
127e670a JK |
1196 | |
1197 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
b3582c68 | 1198 | up_write(&sbi->node_write); |
c29fd0c0 | 1199 | atomic_inc(&sbi->wb_sync_req[NODE]); |
4d57b86d | 1200 | err = f2fs_sync_node_pages(sbi, &wbc, false, FS_CP_NODE_IO); |
c29fd0c0 | 1201 | atomic_dec(&sbi->wb_sync_req[NODE]); |
6d5a1495 | 1202 | if (err) { |
59c9081b | 1203 | up_write(&sbi->node_change); |
cf779cab | 1204 | f2fs_unlock_all(sbi); |
cf779cab JK |
1205 | goto out; |
1206 | } | |
30973883 | 1207 | cond_resched(); |
39936837 | 1208 | goto retry_flush_nodes; |
127e670a | 1209 | } |
59c9081b YH |
1210 | |
1211 | /* | |
1212 | * sbi->node_change is used only for AIO write_begin path which produces | |
1213 | * dirty node blocks and some checkpoint values by block allocation. | |
1214 | */ | |
1215 | __prepare_cp_block(sbi); | |
1216 | up_write(&sbi->node_change); | |
cf779cab | 1217 | out: |
c718379b | 1218 | blk_finish_plug(&plug); |
cf779cab | 1219 | return err; |
127e670a JK |
1220 | } |
1221 | ||
1222 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
1223 | { | |
b3582c68 | 1224 | up_write(&sbi->node_write); |
e479556b | 1225 | f2fs_unlock_all(sbi); |
127e670a JK |
1226 | } |
1227 | ||
50fa53ec | 1228 | void f2fs_wait_on_all_pages_writeback(struct f2fs_sb_info *sbi) |
fb51b5ef CL |
1229 | { |
1230 | DEFINE_WAIT(wait); | |
1231 | ||
1232 | for (;;) { | |
1233 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
1234 | ||
36951b38 | 1235 | if (!get_pages(sbi, F2FS_WB_CP_DATA)) |
fb51b5ef CL |
1236 | break; |
1237 | ||
af697c0f JK |
1238 | if (unlikely(f2fs_cp_error(sbi))) |
1239 | break; | |
1240 | ||
0ff21646 | 1241 | io_schedule_timeout(5*HZ); |
fb51b5ef CL |
1242 | } |
1243 | finish_wait(&sbi->cp_wait, &wait); | |
1244 | } | |
1245 | ||
e4c5d848 JK |
1246 | static void update_ckpt_flags(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
1247 | { | |
1248 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; | |
1249 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
d1aa2453 | 1250 | unsigned long flags; |
e4c5d848 | 1251 | |
d1aa2453 | 1252 | spin_lock_irqsave(&sbi->cp_lock, flags); |
e4c5d848 | 1253 | |
c473f1a9 | 1254 | if ((cpc->reason & CP_UMOUNT) && |
10047f53 | 1255 | le32_to_cpu(ckpt->cp_pack_total_block_count) > |
22ad0b6a JK |
1256 | sbi->blocks_per_seg - NM_I(sbi)->nat_bits_blocks) |
1257 | disable_nat_bits(sbi, false); | |
1258 | ||
1f43e2ad CY |
1259 | if (cpc->reason & CP_TRIMMED) |
1260 | __set_ckpt_flags(ckpt, CP_TRIMMED_FLAG); | |
cd36d7a1 CY |
1261 | else |
1262 | __clear_ckpt_flags(ckpt, CP_TRIMMED_FLAG); | |
1f43e2ad | 1263 | |
c473f1a9 | 1264 | if (cpc->reason & CP_UMOUNT) |
e4c5d848 JK |
1265 | __set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); |
1266 | else | |
1267 | __clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
1268 | ||
c473f1a9 | 1269 | if (cpc->reason & CP_FASTBOOT) |
e4c5d848 JK |
1270 | __set_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); |
1271 | else | |
1272 | __clear_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
1273 | ||
1274 | if (orphan_num) | |
1275 | __set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); | |
1276 | else | |
1277 | __clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); | |
1278 | ||
1279 | if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) | |
1280 | __set_ckpt_flags(ckpt, CP_FSCK_FLAG); | |
1281 | ||
4354994f DR |
1282 | if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) |
1283 | __set_ckpt_flags(ckpt, CP_DISABLED_FLAG); | |
1284 | else | |
1285 | __clear_ckpt_flags(ckpt, CP_DISABLED_FLAG); | |
1286 | ||
db610a64 JK |
1287 | if (is_sbi_flag_set(sbi, SBI_CP_DISABLED_QUICK)) |
1288 | __set_ckpt_flags(ckpt, CP_DISABLED_QUICK_FLAG); | |
1289 | else | |
1290 | __clear_ckpt_flags(ckpt, CP_DISABLED_QUICK_FLAG); | |
1291 | ||
af033b2a CY |
1292 | if (is_sbi_flag_set(sbi, SBI_QUOTA_SKIP_FLUSH)) |
1293 | __set_ckpt_flags(ckpt, CP_QUOTA_NEED_FSCK_FLAG); | |
fb40d618 JK |
1294 | /* |
1295 | * TODO: we count on fsck.f2fs to clear this flag until we figure out | |
1296 | * missing cases which clear it incorrectly. | |
1297 | */ | |
af033b2a CY |
1298 | |
1299 | if (is_sbi_flag_set(sbi, SBI_QUOTA_NEED_REPAIR)) | |
1300 | __set_ckpt_flags(ckpt, CP_QUOTA_NEED_FSCK_FLAG); | |
1301 | ||
e4c5d848 JK |
1302 | /* set this flag to activate crc|cp_ver for recovery */ |
1303 | __set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG); | |
f2367923 | 1304 | __clear_ckpt_flags(ckpt, CP_NOCRC_RECOVERY_FLAG); |
e4c5d848 | 1305 | |
d1aa2453 | 1306 | spin_unlock_irqrestore(&sbi->cp_lock, flags); |
e4c5d848 JK |
1307 | } |
1308 | ||
46706d59 GX |
1309 | static void commit_checkpoint(struct f2fs_sb_info *sbi, |
1310 | void *src, block_t blk_addr) | |
1311 | { | |
1312 | struct writeback_control wbc = { | |
1313 | .for_reclaim = 0, | |
1314 | }; | |
1315 | ||
1316 | /* | |
1317 | * pagevec_lookup_tag and lock_page again will take | |
4d57b86d CY |
1318 | * some extra time. Therefore, f2fs_update_meta_pages and |
1319 | * f2fs_sync_meta_pages are combined in this function. | |
46706d59 | 1320 | */ |
4d57b86d | 1321 | struct page *page = f2fs_grab_meta_page(sbi, blk_addr); |
46706d59 GX |
1322 | int err; |
1323 | ||
bae0ee7a | 1324 | f2fs_wait_on_page_writeback(page, META, true, true); |
8d64d365 CY |
1325 | |
1326 | memcpy(page_address(page), src, PAGE_SIZE); | |
1327 | ||
1328 | set_page_dirty(page); | |
46706d59 GX |
1329 | if (unlikely(!clear_page_dirty_for_io(page))) |
1330 | f2fs_bug_on(sbi, 1); | |
1331 | ||
1332 | /* writeout cp pack 2 page */ | |
1333 | err = __f2fs_write_meta_page(page, &wbc, FS_CP_META_IO); | |
af697c0f JK |
1334 | if (unlikely(err && f2fs_cp_error(sbi))) { |
1335 | f2fs_put_page(page, 1); | |
1336 | return; | |
1337 | } | |
46706d59 | 1338 | |
af697c0f | 1339 | f2fs_bug_on(sbi, err); |
46706d59 GX |
1340 | f2fs_put_page(page, 0); |
1341 | ||
1342 | /* submit checkpoint (with barrier if NOBARRIER is not set) */ | |
1343 | f2fs_submit_merged_write(sbi, META_FLUSH); | |
1344 | } | |
1345 | ||
c34f42e2 | 1346 | static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1347 | { |
1348 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
77041823 | 1349 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
d1aa2453 | 1350 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num, flags; |
127e670a | 1351 | block_t start_blk; |
127e670a | 1352 | unsigned int data_sum_blocks, orphan_blocks; |
7e586fa0 | 1353 | __u32 crc32 = 0; |
127e670a | 1354 | int i; |
55141486 | 1355 | int cp_payload_blks = __cp_payload(sbi); |
8f1dbbbb SL |
1356 | struct super_block *sb = sbi->sb; |
1357 | struct curseg_info *seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); | |
1358 | u64 kbytes_written; | |
1228b482 | 1359 | int err; |
127e670a JK |
1360 | |
1361 | /* Flush all the NAT/SIT pages */ | |
8ec18bff CY |
1362 | f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_CP_META_IO); |
1363 | f2fs_bug_on(sbi, get_pages(sbi, F2FS_DIRTY_META) && | |
1364 | !f2fs_cp_error(sbi)); | |
127e670a | 1365 | |
127e670a JK |
1366 | /* |
1367 | * modify checkpoint | |
1368 | * version number is already updated | |
1369 | */ | |
a1f72ac2 | 1370 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi, true)); |
127e670a | 1371 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); |
b5b82205 | 1372 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
1373 | ckpt->cur_node_segno[i] = |
1374 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
1375 | ckpt->cur_node_blkoff[i] = | |
1376 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
1377 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
1378 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
1379 | } | |
b5b82205 | 1380 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
1381 | ckpt->cur_data_segno[i] = |
1382 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
1383 | ckpt->cur_data_blkoff[i] = | |
1384 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
1385 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
1386 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
1387 | } | |
1388 | ||
127e670a | 1389 | /* 2 cp + n data seg summary + orphan inode blocks */ |
4d57b86d | 1390 | data_sum_blocks = f2fs_npages_for_summary_flush(sbi, false); |
d1aa2453 | 1391 | spin_lock_irqsave(&sbi->cp_lock, flags); |
b5b82205 | 1392 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
aaec2b1d | 1393 | __set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 1394 | else |
aaec2b1d | 1395 | __clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
d1aa2453 | 1396 | spin_unlock_irqrestore(&sbi->cp_lock, flags); |
127e670a | 1397 | |
67298804 | 1398 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
1399 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
1400 | orphan_blocks); | |
127e670a | 1401 | |
119ee914 | 1402 | if (__remain_node_summaries(cpc->reason)) |
b5b82205 | 1403 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
1404 | cp_payload_blks + data_sum_blocks + |
1405 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
119ee914 | 1406 | else |
b5b82205 | 1407 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS + |
1dbe4152 CL |
1408 | cp_payload_blks + data_sum_blocks + |
1409 | orphan_blocks); | |
119ee914 | 1410 | |
e4c5d848 JK |
1411 | /* update ckpt flag for checkpoint */ |
1412 | update_ckpt_flags(sbi, cpc); | |
a468f0ef | 1413 | |
127e670a JK |
1414 | /* update SIT/NAT bitmap */ |
1415 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
1416 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
1417 | ||
d7eb8f1c | 1418 | crc32 = f2fs_checkpoint_chksum(sbi, ckpt); |
7e586fa0 JK |
1419 | *((__le32 *)((unsigned char *)ckpt + |
1420 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
1421 | = cpu_to_le32(crc32); |
1422 | ||
8508e44a | 1423 | start_blk = __start_cp_next_addr(sbi); |
127e670a | 1424 | |
22ad0b6a JK |
1425 | /* write nat bits */ |
1426 | if (enabled_nat_bits(sbi, cpc)) { | |
1427 | __u64 cp_ver = cur_cp_version(ckpt); | |
22ad0b6a JK |
1428 | block_t blk; |
1429 | ||
1430 | cp_ver |= ((__u64)crc32 << 32); | |
1431 | *(__le64 *)nm_i->nat_bits = cpu_to_le64(cp_ver); | |
1432 | ||
1433 | blk = start_blk + sbi->blocks_per_seg - nm_i->nat_bits_blocks; | |
1434 | for (i = 0; i < nm_i->nat_bits_blocks; i++) | |
4d57b86d | 1435 | f2fs_update_meta_page(sbi, nm_i->nat_bits + |
22ad0b6a | 1436 | (i << F2FS_BLKSIZE_BITS), blk + i); |
22ad0b6a JK |
1437 | } |
1438 | ||
127e670a | 1439 | /* write out checkpoint buffer at block 0 */ |
4d57b86d | 1440 | f2fs_update_meta_page(sbi, ckpt, start_blk++); |
381722d2 CY |
1441 | |
1442 | for (i = 1; i < 1 + cp_payload_blks; i++) | |
4d57b86d | 1443 | f2fs_update_meta_page(sbi, (char *)ckpt + i * F2FS_BLKSIZE, |
381722d2 | 1444 | start_blk++); |
1dbe4152 | 1445 | |
67298804 | 1446 | if (orphan_num) { |
127e670a JK |
1447 | write_orphan_inodes(sbi, start_blk); |
1448 | start_blk += orphan_blocks; | |
1449 | } | |
1450 | ||
4d57b86d | 1451 | f2fs_write_data_summaries(sbi, start_blk); |
127e670a | 1452 | start_blk += data_sum_blocks; |
8f1dbbbb SL |
1453 | |
1454 | /* Record write statistics in the hot node summary */ | |
1455 | kbytes_written = sbi->kbytes_written; | |
1456 | if (sb->s_bdev->bd_part) | |
1457 | kbytes_written += BD_PART_WRITTEN(sbi); | |
1458 | ||
b7ad7512 | 1459 | seg_i->journal->info.kbytes_written = cpu_to_le64(kbytes_written); |
8f1dbbbb | 1460 | |
119ee914 | 1461 | if (__remain_node_summaries(cpc->reason)) { |
4d57b86d | 1462 | f2fs_write_node_summaries(sbi, start_blk); |
127e670a JK |
1463 | start_blk += NR_CURSEG_NODE_TYPE; |
1464 | } | |
1465 | ||
46706d59 GX |
1466 | /* update user_block_counts */ |
1467 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
1468 | percpu_counter_set(&sbi->alloc_valid_block_count, 0); | |
127e670a | 1469 | |
46706d59 | 1470 | /* Here, we have one bio having CP pack except cp pack 2 page */ |
4d57b86d | 1471 | f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_CP_META_IO); |
8ec18bff CY |
1472 | f2fs_bug_on(sbi, get_pages(sbi, F2FS_DIRTY_META) && |
1473 | !f2fs_cp_error(sbi)); | |
46706d59 GX |
1474 | |
1475 | /* wait for previous submitted meta pages writeback */ | |
50fa53ec | 1476 | f2fs_wait_on_all_pages_writeback(sbi); |
127e670a | 1477 | |
46706d59 GX |
1478 | /* flush all device cache */ |
1479 | err = f2fs_flush_device_cache(sbi); | |
1480 | if (err) | |
1481 | return err; | |
127e670a | 1482 | |
46706d59 GX |
1483 | /* barrier and flush checkpoint cp pack 2 page if it can */ |
1484 | commit_checkpoint(sbi, ckpt, start_blk); | |
50fa53ec | 1485 | f2fs_wait_on_all_pages_writeback(sbi); |
6a8f8ca5 | 1486 | |
18767e62 CY |
1487 | /* |
1488 | * invalidate intermediate page cache borrowed from meta inode | |
1489 | * which are used for migration of encrypted inode's blocks. | |
1490 | */ | |
7beb01f7 | 1491 | if (f2fs_sb_has_encrypt(sbi)) |
18767e62 CY |
1492 | invalidate_mapping_pages(META_MAPPING(sbi), |
1493 | MAIN_BLKADDR(sbi), MAX_BLKADDR(sbi) - 1); | |
1494 | ||
4d57b86d | 1495 | f2fs_release_ino_entry(sbi, false); |
cf779cab | 1496 | |
50fa53ec CY |
1497 | f2fs_reset_fsync_node_info(sbi); |
1498 | ||
caf0047e | 1499 | clear_sbi_flag(sbi, SBI_IS_DIRTY); |
bbf156f7 | 1500 | clear_sbi_flag(sbi, SBI_NEED_CP); |
af033b2a | 1501 | clear_sbi_flag(sbi, SBI_QUOTA_SKIP_FLUSH); |
4354994f | 1502 | sbi->unusable_block_count = 0; |
8508e44a | 1503 | __set_cp_next_pack(sbi); |
c34f42e2 | 1504 | |
c2a080ae CY |
1505 | /* |
1506 | * redirty superblock if metadata like node page or inode cache is | |
1507 | * updated during writing checkpoint. | |
1508 | */ | |
1509 | if (get_pages(sbi, F2FS_DIRTY_NODES) || | |
1510 | get_pages(sbi, F2FS_DIRTY_IMETA)) | |
1511 | set_sbi_flag(sbi, SBI_IS_DIRTY); | |
1512 | ||
1513 | f2fs_bug_on(sbi, get_pages(sbi, F2FS_DIRTY_DENTS)); | |
1514 | ||
af697c0f | 1515 | return unlikely(f2fs_cp_error(sbi)) ? -EIO : 0; |
127e670a JK |
1516 | } |
1517 | ||
0a8165d7 | 1518 | /* |
e1c42045 | 1519 | * We guarantee that this checkpoint procedure will not fail. |
127e670a | 1520 | */ |
4d57b86d | 1521 | int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1522 | { |
1523 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1524 | unsigned long long ckpt_ver; | |
c34f42e2 | 1525 | int err = 0; |
127e670a | 1526 | |
4354994f DR |
1527 | if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { |
1528 | if (cpc->reason != CP_PAUSE) | |
1529 | return 0; | |
1530 | f2fs_msg(sbi->sb, KERN_WARNING, | |
1531 | "Start checkpoint disabled!"); | |
1532 | } | |
43727527 | 1533 | mutex_lock(&sbi->cp_mutex); |
8501017e | 1534 | |
caf0047e | 1535 | if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) && |
c473f1a9 CY |
1536 | ((cpc->reason & CP_FASTBOOT) || (cpc->reason & CP_SYNC) || |
1537 | ((cpc->reason & CP_DISCARD) && !sbi->discard_blks))) | |
8501017e | 1538 | goto out; |
c34f42e2 CY |
1539 | if (unlikely(f2fs_cp_error(sbi))) { |
1540 | err = -EIO; | |
cf779cab | 1541 | goto out; |
c34f42e2 CY |
1542 | } |
1543 | if (f2fs_readonly(sbi->sb)) { | |
1544 | err = -EROFS; | |
11504a8e | 1545 | goto out; |
c34f42e2 | 1546 | } |
2bda542d WL |
1547 | |
1548 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); | |
1549 | ||
c34f42e2 CY |
1550 | err = block_operations(sbi); |
1551 | if (err) | |
cf779cab | 1552 | goto out; |
127e670a | 1553 | |
75ab4cb8 | 1554 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1555 | |
b9109b0e | 1556 | f2fs_flush_merged_writes(sbi); |
127e670a | 1557 | |
58cce381 | 1558 | /* this is the case of multiple fstrims without any changes */ |
c473f1a9 | 1559 | if (cpc->reason & CP_DISCARD) { |
4d57b86d | 1560 | if (!f2fs_exist_trim_candidates(sbi, cpc)) { |
25290fa5 JK |
1561 | unblock_operations(sbi); |
1562 | goto out; | |
1563 | } | |
1564 | ||
0333ad4e JK |
1565 | if (NM_I(sbi)->dirty_nat_cnt == 0 && |
1566 | SIT_I(sbi)->dirty_sentries == 0 && | |
1567 | prefree_segments(sbi) == 0) { | |
4d57b86d CY |
1568 | f2fs_flush_sit_entries(sbi, cpc); |
1569 | f2fs_clear_prefree_segments(sbi, cpc); | |
0333ad4e JK |
1570 | unblock_operations(sbi); |
1571 | goto out; | |
1572 | } | |
58cce381 YH |
1573 | } |
1574 | ||
127e670a JK |
1575 | /* |
1576 | * update checkpoint pack index | |
1577 | * Increase the version number so that | |
1578 | * SIT entries and seg summaries are written at correct place | |
1579 | */ | |
d71b5564 | 1580 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1581 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1582 | ||
1583 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
edc55aaf JK |
1584 | err = f2fs_flush_nat_entries(sbi, cpc); |
1585 | if (err) | |
1586 | goto stop; | |
1587 | ||
4d57b86d | 1588 | f2fs_flush_sit_entries(sbi, cpc); |
127e670a | 1589 | |
127e670a | 1590 | /* unlock all the fs_lock[] in do_checkpoint() */ |
c34f42e2 | 1591 | err = do_checkpoint(sbi, cpc); |
4e6a8d9b | 1592 | if (err) |
4d57b86d | 1593 | f2fs_release_discard_addrs(sbi); |
4e6a8d9b | 1594 | else |
4d57b86d | 1595 | f2fs_clear_prefree_segments(sbi, cpc); |
edc55aaf | 1596 | stop: |
127e670a | 1597 | unblock_operations(sbi); |
942e0be6 | 1598 | stat_inc_cp_count(sbi->stat_info); |
10027551 | 1599 | |
c473f1a9 | 1600 | if (cpc->reason & CP_RECOVERY) |
10027551 JK |
1601 | f2fs_msg(sbi->sb, KERN_NOTICE, |
1602 | "checkpoint: version = %llx", ckpt_ver); | |
60b99b48 JK |
1603 | |
1604 | /* do checkpoint periodically */ | |
6beceb54 | 1605 | f2fs_update_time(sbi, CP_TIME); |
55d1cdb2 | 1606 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
8501017e JK |
1607 | out: |
1608 | mutex_unlock(&sbi->cp_mutex); | |
c34f42e2 | 1609 | return err; |
127e670a JK |
1610 | } |
1611 | ||
4d57b86d | 1612 | void f2fs_init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1613 | { |
6451e041 JK |
1614 | int i; |
1615 | ||
1616 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1617 | struct inode_management *im = &sbi->im[i]; |
1618 | ||
1619 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1620 | spin_lock_init(&im->ino_lock); | |
1621 | INIT_LIST_HEAD(&im->ino_list); | |
1622 | im->ino_num = 0; | |
6451e041 JK |
1623 | } |
1624 | ||
b5b82205 | 1625 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
14b42817 WL |
1626 | NR_CURSEG_TYPE - __cp_payload(sbi)) * |
1627 | F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
1628 | } |
1629 | ||
4d57b86d | 1630 | int __init f2fs_create_checkpoint_caches(void) |
127e670a | 1631 | { |
6451e041 JK |
1632 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1633 | sizeof(struct ino_entry)); | |
1634 | if (!ino_entry_slab) | |
127e670a | 1635 | return -ENOMEM; |
4d57b86d | 1636 | f2fs_inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry", |
06292073 | 1637 | sizeof(struct inode_entry)); |
4d57b86d | 1638 | if (!f2fs_inode_entry_slab) { |
6451e041 | 1639 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1640 | return -ENOMEM; |
1641 | } | |
1642 | return 0; | |
1643 | } | |
1644 | ||
4d57b86d | 1645 | void f2fs_destroy_checkpoint_caches(void) |
127e670a | 1646 | { |
6451e041 | 1647 | kmem_cache_destroy(ino_entry_slab); |
4d57b86d | 1648 | kmem_cache_destroy(f2fs_inode_entry_slab); |
127e670a | 1649 | } |