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 | ||
2b947003 CY |
53 | static struct page *__get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index, |
54 | bool is_meta) | |
127e670a | 55 | { |
9df27d98 | 56 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a | 57 | struct page *page; |
cf04e8eb | 58 | struct f2fs_io_info fio = { |
05ca3632 | 59 | .sbi = sbi, |
cf04e8eb | 60 | .type = META, |
04d328de | 61 | .op = REQ_OP_READ, |
70fd7614 | 62 | .op_flags = REQ_META | REQ_PRIO, |
7a9d7548 CY |
63 | .old_blkaddr = index, |
64 | .new_blkaddr = index, | |
4375a336 | 65 | .encrypted_page = NULL, |
6dc3a126 | 66 | .is_por = !is_meta, |
cf04e8eb | 67 | }; |
7735730d | 68 | int err; |
2b947003 CY |
69 | |
70 | if (unlikely(!is_meta)) | |
04d328de | 71 | fio.op_flags &= ~REQ_META; |
127e670a | 72 | repeat: |
300e129c | 73 | page = f2fs_grab_cache_page(mapping, index, false); |
127e670a JK |
74 | if (!page) { |
75 | cond_resched(); | |
76 | goto repeat; | |
77 | } | |
393ff91f JK |
78 | if (PageUptodate(page)) |
79 | goto out; | |
80 | ||
05ca3632 JK |
81 | fio.page = page; |
82 | ||
7735730d CY |
83 | err = f2fs_submit_page_bio(&fio); |
84 | if (err) { | |
85 | f2fs_put_page(page, 1); | |
86 | return ERR_PTR(err); | |
86531d6b | 87 | } |
127e670a | 88 | |
393ff91f | 89 | lock_page(page); |
6bacf52f | 90 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
91 | f2fs_put_page(page, 1); |
92 | goto repeat; | |
93 | } | |
f3f338ca | 94 | |
81114baa | 95 | if (unlikely(!PageUptodate(page))) { |
7735730d CY |
96 | f2fs_put_page(page, 1); |
97 | return ERR_PTR(-EIO); | |
81114baa | 98 | } |
393ff91f | 99 | out: |
127e670a JK |
100 | return page; |
101 | } | |
102 | ||
4d57b86d | 103 | struct page *f2fs_get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) |
2b947003 CY |
104 | { |
105 | return __get_meta_page(sbi, index, true); | |
106 | } | |
107 | ||
7735730d CY |
108 | struct page *f2fs_get_meta_page_nofail(struct f2fs_sb_info *sbi, pgoff_t index) |
109 | { | |
110 | struct page *page; | |
111 | int count = 0; | |
112 | ||
113 | retry: | |
114 | page = __get_meta_page(sbi, index, true); | |
115 | if (IS_ERR(page)) { | |
116 | if (PTR_ERR(page) == -EIO && | |
117 | ++count <= DEFAULT_RETRY_IO_COUNT) | |
118 | goto retry; | |
7735730d | 119 | f2fs_stop_checkpoint(sbi, false); |
7735730d | 120 | } |
7735730d CY |
121 | return page; |
122 | } | |
123 | ||
2b947003 | 124 | /* for POR only */ |
4d57b86d | 125 | struct page *f2fs_get_tmp_page(struct f2fs_sb_info *sbi, pgoff_t index) |
2b947003 CY |
126 | { |
127 | return __get_meta_page(sbi, index, false); | |
128 | } | |
129 | ||
93770ab7 CY |
130 | static bool __is_bitmap_valid(struct f2fs_sb_info *sbi, block_t blkaddr, |
131 | int type) | |
132 | { | |
133 | struct seg_entry *se; | |
134 | unsigned int segno, offset; | |
135 | bool exist; | |
136 | ||
137 | if (type != DATA_GENERIC_ENHANCE && type != DATA_GENERIC_ENHANCE_READ) | |
138 | return true; | |
139 | ||
140 | segno = GET_SEGNO(sbi, blkaddr); | |
141 | offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr); | |
142 | se = get_seg_entry(sbi, segno); | |
143 | ||
144 | exist = f2fs_test_bit(offset, se->cur_valid_map); | |
145 | if (!exist && type == DATA_GENERIC_ENHANCE) { | |
dcbb4c10 JP |
146 | f2fs_err(sbi, "Inconsistent error blkaddr:%u, sit bitmap:%d", |
147 | blkaddr, exist); | |
93770ab7 CY |
148 | set_sbi_flag(sbi, SBI_NEED_FSCK); |
149 | WARN_ON(1); | |
150 | } | |
151 | return exist; | |
152 | } | |
153 | ||
e1da7872 | 154 | bool f2fs_is_valid_blkaddr(struct f2fs_sb_info *sbi, |
4d57b86d | 155 | block_t blkaddr, int type) |
662befda CY |
156 | { |
157 | switch (type) { | |
158 | case META_NAT: | |
66b00c18 | 159 | break; |
662befda | 160 | case META_SIT: |
66b00c18 CY |
161 | if (unlikely(blkaddr >= SIT_BLK_CNT(sbi))) |
162 | return false; | |
163 | break; | |
81c1a0f1 | 164 | case META_SSA: |
66b00c18 CY |
165 | if (unlikely(blkaddr >= MAIN_BLKADDR(sbi) || |
166 | blkaddr < SM_I(sbi)->ssa_blkaddr)) | |
167 | return false; | |
168 | break; | |
662befda | 169 | case META_CP: |
66b00c18 CY |
170 | if (unlikely(blkaddr >= SIT_I(sbi)->sit_base_addr || |
171 | blkaddr < __start_cp_addr(sbi))) | |
172 | return false; | |
173 | break; | |
4c521f49 | 174 | case META_POR: |
93770ab7 CY |
175 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
176 | blkaddr < MAIN_BLKADDR(sbi))) | |
177 | return false; | |
178 | break; | |
e1da7872 | 179 | case DATA_GENERIC: |
93770ab7 CY |
180 | case DATA_GENERIC_ENHANCE: |
181 | case DATA_GENERIC_ENHANCE_READ: | |
66b00c18 | 182 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
93770ab7 | 183 | blkaddr < MAIN_BLKADDR(sbi))) { |
dcbb4c10 JP |
184 | f2fs_warn(sbi, "access invalid blkaddr:%u", |
185 | blkaddr); | |
93770ab7 CY |
186 | set_sbi_flag(sbi, SBI_NEED_FSCK); |
187 | WARN_ON(1); | |
66b00c18 | 188 | return false; |
93770ab7 CY |
189 | } else { |
190 | return __is_bitmap_valid(sbi, blkaddr, type); | |
c9b60788 | 191 | } |
66b00c18 | 192 | break; |
e1da7872 CY |
193 | case META_GENERIC: |
194 | if (unlikely(blkaddr < SEG0_BLKADDR(sbi) || | |
195 | blkaddr >= MAIN_BLKADDR(sbi))) | |
196 | return false; | |
197 | break; | |
662befda CY |
198 | default: |
199 | BUG(); | |
200 | } | |
66b00c18 CY |
201 | |
202 | return true; | |
662befda CY |
203 | } |
204 | ||
205 | /* | |
7a88ddb5 | 206 | * Readahead CP/NAT/SIT/SSA/POR pages |
662befda | 207 | */ |
4d57b86d | 208 | int f2fs_ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages, |
26879fb1 | 209 | int type, bool sync) |
662befda | 210 | { |
662befda | 211 | struct page *page; |
4c521f49 | 212 | block_t blkno = start; |
662befda | 213 | struct f2fs_io_info fio = { |
05ca3632 | 214 | .sbi = sbi, |
662befda | 215 | .type = META, |
04d328de | 216 | .op = REQ_OP_READ, |
70fd7614 | 217 | .op_flags = sync ? (REQ_META | REQ_PRIO) : REQ_RAHEAD, |
4375a336 | 218 | .encrypted_page = NULL, |
fb830fc5 | 219 | .in_list = false, |
6dc3a126 | 220 | .is_por = (type == META_POR), |
662befda | 221 | }; |
e9f5b8b8 | 222 | struct blk_plug plug; |
662befda | 223 | |
2b947003 | 224 | if (unlikely(type == META_POR)) |
04d328de | 225 | fio.op_flags &= ~REQ_META; |
2b947003 | 226 | |
e9f5b8b8 | 227 | blk_start_plug(&plug); |
662befda | 228 | for (; nrpages-- > 0; blkno++) { |
662befda | 229 | |
e1da7872 | 230 | if (!f2fs_is_valid_blkaddr(sbi, blkno, type)) |
66b00c18 CY |
231 | goto out; |
232 | ||
662befda CY |
233 | switch (type) { |
234 | case META_NAT: | |
66b00c18 CY |
235 | if (unlikely(blkno >= |
236 | NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))) | |
662befda | 237 | blkno = 0; |
66b00c18 | 238 | /* get nat block addr */ |
7a9d7548 | 239 | fio.new_blkaddr = current_nat_addr(sbi, |
662befda CY |
240 | blkno * NAT_ENTRY_PER_BLOCK); |
241 | break; | |
242 | case META_SIT: | |
243 | /* get sit block addr */ | |
7a9d7548 | 244 | fio.new_blkaddr = current_sit_addr(sbi, |
662befda | 245 | blkno * SIT_ENTRY_PER_BLOCK); |
662befda | 246 | break; |
81c1a0f1 | 247 | case META_SSA: |
662befda | 248 | case META_CP: |
4c521f49 | 249 | case META_POR: |
7a9d7548 | 250 | fio.new_blkaddr = blkno; |
662befda CY |
251 | break; |
252 | default: | |
253 | BUG(); | |
254 | } | |
255 | ||
300e129c JK |
256 | page = f2fs_grab_cache_page(META_MAPPING(sbi), |
257 | fio.new_blkaddr, false); | |
662befda CY |
258 | if (!page) |
259 | continue; | |
260 | if (PageUptodate(page)) { | |
662befda CY |
261 | f2fs_put_page(page, 1); |
262 | continue; | |
263 | } | |
264 | ||
05ca3632 | 265 | fio.page = page; |
1919ffc0 | 266 | f2fs_submit_page_bio(&fio); |
662befda CY |
267 | f2fs_put_page(page, 0); |
268 | } | |
269 | out: | |
e9f5b8b8 | 270 | blk_finish_plug(&plug); |
662befda CY |
271 | return blkno - start; |
272 | } | |
273 | ||
4d57b86d | 274 | void f2fs_ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index) |
635aee1f CY |
275 | { |
276 | struct page *page; | |
277 | bool readahead = false; | |
278 | ||
279 | page = find_get_page(META_MAPPING(sbi), index); | |
4da7bf5a | 280 | if (!page || !PageUptodate(page)) |
635aee1f CY |
281 | readahead = true; |
282 | f2fs_put_page(page, 0); | |
283 | ||
284 | if (readahead) | |
4d57b86d | 285 | f2fs_ra_meta_pages(sbi, index, BIO_MAX_PAGES, META_POR, true); |
635aee1f CY |
286 | } |
287 | ||
b0af6d49 CY |
288 | static int __f2fs_write_meta_page(struct page *page, |
289 | struct writeback_control *wbc, | |
290 | enum iostat_type io_type) | |
127e670a | 291 | { |
4081363f | 292 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 293 | |
ecda0de3 CY |
294 | trace_f2fs_writepage(page, META); |
295 | ||
af697c0f JK |
296 | if (unlikely(f2fs_cp_error(sbi))) |
297 | goto redirty_out; | |
caf0047e | 298 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
cfb271d4 | 299 | goto redirty_out; |
857dc4e0 | 300 | if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0)) |
cfb271d4 | 301 | goto redirty_out; |
127e670a | 302 | |
4d57b86d | 303 | f2fs_do_write_meta_page(sbi, page, io_type); |
577e3495 | 304 | dec_page_count(sbi, F2FS_DIRTY_META); |
0c3a5797 CY |
305 | |
306 | if (wbc->for_reclaim) | |
bab475c5 | 307 | f2fs_submit_merged_write_cond(sbi, NULL, page, 0, META); |
0c3a5797 | 308 | |
577e3495 | 309 | unlock_page(page); |
857dc4e0 | 310 | |
0c3a5797 | 311 | if (unlikely(f2fs_cp_error(sbi))) |
b9109b0e | 312 | f2fs_submit_merged_write(sbi, META); |
0c3a5797 | 313 | |
577e3495 | 314 | return 0; |
cfb271d4 CY |
315 | |
316 | redirty_out: | |
76f60268 | 317 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 318 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
319 | } |
320 | ||
b0af6d49 CY |
321 | static int f2fs_write_meta_page(struct page *page, |
322 | struct writeback_control *wbc) | |
323 | { | |
324 | return __f2fs_write_meta_page(page, wbc, FS_META_IO); | |
325 | } | |
326 | ||
127e670a JK |
327 | static int f2fs_write_meta_pages(struct address_space *mapping, |
328 | struct writeback_control *wbc) | |
329 | { | |
4081363f | 330 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 331 | long diff, written; |
127e670a | 332 | |
0771fcc7 CY |
333 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
334 | goto skip_write; | |
335 | ||
5459aa97 | 336 | /* collect a number of dirty meta pages and write together */ |
812a9597 JK |
337 | if (wbc->sync_mode != WB_SYNC_ALL && |
338 | get_pages(sbi, F2FS_DIRTY_META) < | |
339 | nr_pages_to_skip(sbi, META)) | |
d3baf95d | 340 | goto skip_write; |
127e670a | 341 | |
a29d0e0b YH |
342 | /* if locked failed, cp will flush dirty pages instead */ |
343 | if (!mutex_trylock(&sbi->cp_mutex)) | |
344 | goto skip_write; | |
d31c7c3f | 345 | |
a29d0e0b | 346 | trace_f2fs_writepages(mapping->host, wbc, META); |
50c8cdb3 | 347 | diff = nr_pages_to_write(sbi, META, wbc); |
4d57b86d | 348 | written = f2fs_sync_meta_pages(sbi, META, wbc->nr_to_write, FS_META_IO); |
127e670a | 349 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 350 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 351 | return 0; |
d3baf95d JK |
352 | |
353 | skip_write: | |
354 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
d31c7c3f | 355 | trace_f2fs_writepages(mapping->host, wbc, META); |
d3baf95d | 356 | return 0; |
127e670a JK |
357 | } |
358 | ||
4d57b86d | 359 | long f2fs_sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, |
b0af6d49 | 360 | long nr_to_write, enum iostat_type io_type) |
127e670a | 361 | { |
9df27d98 | 362 | struct address_space *mapping = META_MAPPING(sbi); |
028a63a6 | 363 | pgoff_t index = 0, prev = ULONG_MAX; |
127e670a JK |
364 | struct pagevec pvec; |
365 | long nwritten = 0; | |
028a63a6 | 366 | int nr_pages; |
127e670a JK |
367 | struct writeback_control wbc = { |
368 | .for_reclaim = 0, | |
369 | }; | |
e9f5b8b8 | 370 | struct blk_plug plug; |
127e670a | 371 | |
86679820 | 372 | pagevec_init(&pvec); |
127e670a | 373 | |
e9f5b8b8 CY |
374 | blk_start_plug(&plug); |
375 | ||
028a63a6 | 376 | while ((nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, |
67fd707f | 377 | PAGECACHE_TAG_DIRTY))) { |
028a63a6 | 378 | int i; |
127e670a JK |
379 | |
380 | for (i = 0; i < nr_pages; i++) { | |
381 | struct page *page = pvec.pages[i]; | |
203681f6 | 382 | |
80dd9c0e | 383 | if (prev == ULONG_MAX) |
6066d8cd JK |
384 | prev = page->index - 1; |
385 | if (nr_to_write != LONG_MAX && page->index != prev + 1) { | |
386 | pagevec_release(&pvec); | |
387 | goto stop; | |
388 | } | |
389 | ||
127e670a | 390 | lock_page(page); |
203681f6 JK |
391 | |
392 | if (unlikely(page->mapping != mapping)) { | |
393 | continue_unlock: | |
394 | unlock_page(page); | |
395 | continue; | |
396 | } | |
397 | if (!PageDirty(page)) { | |
398 | /* someone wrote it for us */ | |
399 | goto continue_unlock; | |
400 | } | |
401 | ||
bae0ee7a | 402 | f2fs_wait_on_page_writeback(page, META, true, true); |
fa3d2bdf | 403 | |
203681f6 JK |
404 | if (!clear_page_dirty_for_io(page)) |
405 | goto continue_unlock; | |
406 | ||
b0af6d49 | 407 | if (__f2fs_write_meta_page(page, &wbc, io_type)) { |
577e3495 JK |
408 | unlock_page(page); |
409 | break; | |
410 | } | |
cfb271d4 | 411 | nwritten++; |
6066d8cd | 412 | prev = page->index; |
cfb271d4 | 413 | if (unlikely(nwritten >= nr_to_write)) |
127e670a JK |
414 | break; |
415 | } | |
416 | pagevec_release(&pvec); | |
417 | cond_resched(); | |
418 | } | |
6066d8cd | 419 | stop: |
127e670a | 420 | if (nwritten) |
b9109b0e | 421 | f2fs_submit_merged_write(sbi, type); |
127e670a | 422 | |
e9f5b8b8 CY |
423 | blk_finish_plug(&plug); |
424 | ||
127e670a JK |
425 | return nwritten; |
426 | } | |
427 | ||
428 | static int f2fs_set_meta_page_dirty(struct page *page) | |
429 | { | |
26c6b887 JK |
430 | trace_f2fs_set_page_dirty(page, META); |
431 | ||
237c0790 JK |
432 | if (!PageUptodate(page)) |
433 | SetPageUptodate(page); | |
127e670a | 434 | if (!PageDirty(page)) { |
b87078ad | 435 | __set_page_dirty_nobuffers(page); |
4081363f | 436 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); |
240a5915 | 437 | f2fs_set_page_private(page, 0); |
9e4ded3f | 438 | f2fs_trace_pid(page); |
127e670a JK |
439 | return 1; |
440 | } | |
441 | return 0; | |
442 | } | |
443 | ||
444 | const struct address_space_operations f2fs_meta_aops = { | |
445 | .writepage = f2fs_write_meta_page, | |
446 | .writepages = f2fs_write_meta_pages, | |
447 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
487261f3 CY |
448 | .invalidatepage = f2fs_invalidate_page, |
449 | .releasepage = f2fs_release_page, | |
5b7a487c WG |
450 | #ifdef CONFIG_MIGRATION |
451 | .migratepage = f2fs_migrate_page, | |
452 | #endif | |
127e670a JK |
453 | }; |
454 | ||
39d787be CY |
455 | static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, |
456 | unsigned int devidx, int type) | |
953e6cc6 | 457 | { |
67298804 | 458 | struct inode_management *im = &sbi->im[type]; |
80c54505 JK |
459 | struct ino_entry *e, *tmp; |
460 | ||
461 | tmp = f2fs_kmem_cache_alloc(ino_entry_slab, GFP_NOFS); | |
19526d74 | 462 | |
80c54505 | 463 | radix_tree_preload(GFP_NOFS | __GFP_NOFAIL); |
769ec6e5 | 464 | |
67298804 | 465 | spin_lock(&im->ino_lock); |
67298804 | 466 | e = radix_tree_lookup(&im->ino_root, ino); |
39efac41 | 467 | if (!e) { |
80c54505 | 468 | e = tmp; |
19526d74 CY |
469 | if (unlikely(radix_tree_insert(&im->ino_root, ino, e))) |
470 | f2fs_bug_on(sbi, 1); | |
471 | ||
39efac41 JK |
472 | memset(e, 0, sizeof(struct ino_entry)); |
473 | e->ino = ino; | |
953e6cc6 | 474 | |
67298804 | 475 | list_add_tail(&e->list, &im->ino_list); |
8c402946 | 476 | if (type != ORPHAN_INO) |
67298804 | 477 | im->ino_num++; |
39efac41 | 478 | } |
39d787be CY |
479 | |
480 | if (type == FLUSH_INO) | |
481 | f2fs_set_bit(devidx, (char *)&e->dirty_device); | |
482 | ||
67298804 | 483 | spin_unlock(&im->ino_lock); |
769ec6e5 | 484 | radix_tree_preload_end(); |
80c54505 JK |
485 | |
486 | if (e != tmp) | |
487 | kmem_cache_free(ino_entry_slab, tmp); | |
953e6cc6 JK |
488 | } |
489 | ||
6451e041 | 490 | static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 491 | { |
67298804 | 492 | struct inode_management *im = &sbi->im[type]; |
6451e041 | 493 | struct ino_entry *e; |
953e6cc6 | 494 | |
67298804 CY |
495 | spin_lock(&im->ino_lock); |
496 | e = radix_tree_lookup(&im->ino_root, ino); | |
39efac41 JK |
497 | if (e) { |
498 | list_del(&e->list); | |
67298804 CY |
499 | radix_tree_delete(&im->ino_root, ino); |
500 | im->ino_num--; | |
501 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
502 | kmem_cache_free(ino_entry_slab, e); |
503 | return; | |
953e6cc6 | 504 | } |
67298804 | 505 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
506 | } |
507 | ||
4d57b86d | 508 | void f2fs_add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
509 | { |
510 | /* add new dirty ino entry into list */ | |
39d787be | 511 | __add_ino_entry(sbi, ino, 0, type); |
fff04f90 JK |
512 | } |
513 | ||
4d57b86d | 514 | void f2fs_remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
515 | { |
516 | /* remove dirty ino entry from list */ | |
517 | __remove_ino_entry(sbi, ino, type); | |
518 | } | |
519 | ||
520 | /* mode should be APPEND_INO or UPDATE_INO */ | |
4d57b86d | 521 | bool f2fs_exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode) |
fff04f90 | 522 | { |
67298804 | 523 | struct inode_management *im = &sbi->im[mode]; |
fff04f90 | 524 | struct ino_entry *e; |
67298804 CY |
525 | |
526 | spin_lock(&im->ino_lock); | |
527 | e = radix_tree_lookup(&im->ino_root, ino); | |
528 | spin_unlock(&im->ino_lock); | |
fff04f90 JK |
529 | return e ? true : false; |
530 | } | |
531 | ||
4d57b86d | 532 | void f2fs_release_ino_entry(struct f2fs_sb_info *sbi, bool all) |
fff04f90 JK |
533 | { |
534 | struct ino_entry *e, *tmp; | |
535 | int i; | |
536 | ||
39d787be | 537 | for (i = all ? ORPHAN_INO : APPEND_INO; i < MAX_INO_ENTRY; i++) { |
67298804 CY |
538 | struct inode_management *im = &sbi->im[i]; |
539 | ||
540 | spin_lock(&im->ino_lock); | |
541 | list_for_each_entry_safe(e, tmp, &im->ino_list, list) { | |
fff04f90 | 542 | list_del(&e->list); |
67298804 | 543 | radix_tree_delete(&im->ino_root, e->ino); |
fff04f90 | 544 | kmem_cache_free(ino_entry_slab, e); |
67298804 | 545 | im->ino_num--; |
fff04f90 | 546 | } |
67298804 | 547 | spin_unlock(&im->ino_lock); |
fff04f90 JK |
548 | } |
549 | } | |
550 | ||
4d57b86d | 551 | void f2fs_set_dirty_device(struct f2fs_sb_info *sbi, nid_t ino, |
39d787be CY |
552 | unsigned int devidx, int type) |
553 | { | |
554 | __add_ino_entry(sbi, ino, devidx, type); | |
555 | } | |
556 | ||
4d57b86d | 557 | bool f2fs_is_dirty_device(struct f2fs_sb_info *sbi, nid_t ino, |
39d787be CY |
558 | unsigned int devidx, int type) |
559 | { | |
560 | struct inode_management *im = &sbi->im[type]; | |
561 | struct ino_entry *e; | |
562 | bool is_dirty = false; | |
563 | ||
564 | spin_lock(&im->ino_lock); | |
565 | e = radix_tree_lookup(&im->ino_root, ino); | |
566 | if (e && f2fs_test_bit(devidx, (char *)&e->dirty_device)) | |
567 | is_dirty = true; | |
568 | spin_unlock(&im->ino_lock); | |
569 | return is_dirty; | |
570 | } | |
571 | ||
4d57b86d | 572 | int f2fs_acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 573 | { |
67298804 | 574 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a JK |
575 | int err = 0; |
576 | ||
67298804 | 577 | spin_lock(&im->ino_lock); |
cb78942b | 578 | |
1ecc0c5c | 579 | if (time_to_inject(sbi, FAULT_ORPHAN)) { |
cb78942b | 580 | spin_unlock(&im->ino_lock); |
c45d6002 | 581 | f2fs_show_injection_info(sbi, FAULT_ORPHAN); |
cb78942b JK |
582 | return -ENOSPC; |
583 | } | |
7fa750a1 | 584 | |
67298804 | 585 | if (unlikely(im->ino_num >= sbi->max_orphans)) |
127e670a | 586 | err = -ENOSPC; |
cbd56e7d | 587 | else |
67298804 CY |
588 | im->ino_num++; |
589 | spin_unlock(&im->ino_lock); | |
0d47c1ad | 590 | |
127e670a JK |
591 | return err; |
592 | } | |
593 | ||
4d57b86d | 594 | void f2fs_release_orphan_inode(struct f2fs_sb_info *sbi) |
cbd56e7d | 595 | { |
67298804 CY |
596 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
597 | ||
598 | spin_lock(&im->ino_lock); | |
599 | f2fs_bug_on(sbi, im->ino_num == 0); | |
600 | im->ino_num--; | |
601 | spin_unlock(&im->ino_lock); | |
cbd56e7d JK |
602 | } |
603 | ||
4d57b86d | 604 | void f2fs_add_orphan_inode(struct inode *inode) |
127e670a | 605 | { |
39efac41 | 606 | /* add new orphan ino entry into list */ |
39d787be | 607 | __add_ino_entry(F2FS_I_SB(inode), inode->i_ino, 0, ORPHAN_INO); |
4d57b86d | 608 | f2fs_update_inode_page(inode); |
127e670a JK |
609 | } |
610 | ||
4d57b86d | 611 | void f2fs_remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 612 | { |
953e6cc6 | 613 | /* remove orphan entry from orphan list */ |
6451e041 | 614 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
615 | } |
616 | ||
8c14bfad | 617 | static int recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 618 | { |
8c14bfad | 619 | struct inode *inode; |
5905f9af | 620 | struct node_info ni; |
76a45e3c | 621 | int err; |
8c14bfad | 622 | |
5905f9af | 623 | inode = f2fs_iget_retry(sbi->sb, ino); |
8c14bfad CY |
624 | if (IS_ERR(inode)) { |
625 | /* | |
626 | * there should be a bug that we can't find the entry | |
627 | * to orphan inode. | |
628 | */ | |
629 | f2fs_bug_on(sbi, PTR_ERR(inode) == -ENOENT); | |
630 | return PTR_ERR(inode); | |
631 | } | |
632 | ||
0f9ec2a8 | 633 | err = dquot_initialize(inode); |
a515d12f SY |
634 | if (err) { |
635 | iput(inode); | |
0f9ec2a8 | 636 | goto err_out; |
a515d12f | 637 | } |
0f9ec2a8 | 638 | |
127e670a JK |
639 | clear_nlink(inode); |
640 | ||
641 | /* truncate all the data during iput */ | |
642 | iput(inode); | |
5905f9af | 643 | |
7735730d CY |
644 | err = f2fs_get_node_info(sbi, ino, &ni); |
645 | if (err) | |
646 | goto err_out; | |
5905f9af JK |
647 | |
648 | /* ENOMEM was fully retried in f2fs_evict_inode. */ | |
649 | if (ni.blk_addr != NULL_ADDR) { | |
0f9ec2a8 JK |
650 | err = -EIO; |
651 | goto err_out; | |
5905f9af | 652 | } |
8c14bfad | 653 | return 0; |
0f9ec2a8 JK |
654 | |
655 | err_out: | |
656 | set_sbi_flag(sbi, SBI_NEED_FSCK); | |
dcbb4c10 JP |
657 | f2fs_warn(sbi, "%s: orphan failed (ino=%x), run fsck to fix.", |
658 | __func__, ino); | |
0f9ec2a8 | 659 | return err; |
127e670a JK |
660 | } |
661 | ||
4d57b86d | 662 | int f2fs_recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a | 663 | { |
3c642985 | 664 | block_t start_blk, orphan_blocks, i, j; |
4b2414d0 CY |
665 | unsigned int s_flags = sbi->sb->s_flags; |
666 | int err = 0; | |
ea676733 JK |
667 | #ifdef CONFIG_QUOTA |
668 | int quota_enabled; | |
669 | #endif | |
127e670a | 670 | |
aaec2b1d | 671 | if (!is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG)) |
8c14bfad | 672 | return 0; |
127e670a | 673 | |
b61af314 | 674 | if (bdev_read_only(sbi->sb->s_bdev)) { |
dcbb4c10 | 675 | f2fs_info(sbi, "write access unavailable, skipping orphan cleanup"); |
b61af314 CY |
676 | return 0; |
677 | } | |
678 | ||
1751e8a6 | 679 | if (s_flags & SB_RDONLY) { |
dcbb4c10 | 680 | f2fs_info(sbi, "orphan cleanup on readonly fs"); |
1751e8a6 | 681 | sbi->sb->s_flags &= ~SB_RDONLY; |
4b2414d0 CY |
682 | } |
683 | ||
684 | #ifdef CONFIG_QUOTA | |
685 | /* Needed for iput() to work correctly and not trash data */ | |
1751e8a6 | 686 | sbi->sb->s_flags |= SB_ACTIVE; |
ea676733 | 687 | |
76cf05d7 SY |
688 | /* |
689 | * Turn on quotas which were not enabled for read-only mounts if | |
690 | * filesystem has quota feature, so that they are updated correctly. | |
691 | */ | |
1751e8a6 | 692 | quota_enabled = f2fs_enable_quota_files(sbi, s_flags & SB_RDONLY); |
4b2414d0 CY |
693 | #endif |
694 | ||
55141486 | 695 | start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi); |
3c642985 | 696 | orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi); |
127e670a | 697 | |
4d57b86d | 698 | f2fs_ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP, true); |
662befda | 699 | |
3c642985 | 700 | for (i = 0; i < orphan_blocks; i++) { |
7735730d | 701 | struct page *page; |
127e670a JK |
702 | struct f2fs_orphan_block *orphan_blk; |
703 | ||
7735730d CY |
704 | page = f2fs_get_meta_page(sbi, start_blk + i); |
705 | if (IS_ERR(page)) { | |
706 | err = PTR_ERR(page); | |
707 | goto out; | |
708 | } | |
709 | ||
127e670a JK |
710 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); |
711 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
712 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
8c14bfad CY |
713 | err = recover_orphan_inode(sbi, ino); |
714 | if (err) { | |
715 | f2fs_put_page(page, 1); | |
4b2414d0 | 716 | goto out; |
8c14bfad | 717 | } |
127e670a JK |
718 | } |
719 | f2fs_put_page(page, 1); | |
720 | } | |
721 | /* clear Orphan Flag */ | |
aaec2b1d | 722 | clear_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG); |
4b2414d0 | 723 | out: |
1378752b CY |
724 | set_sbi_flag(sbi, SBI_IS_RECOVERED); |
725 | ||
4b2414d0 CY |
726 | #ifdef CONFIG_QUOTA |
727 | /* Turn quotas off */ | |
ea676733 JK |
728 | if (quota_enabled) |
729 | f2fs_quota_off_umount(sbi->sb); | |
4b2414d0 | 730 | #endif |
1751e8a6 | 731 | sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ |
4b2414d0 CY |
732 | |
733 | return err; | |
127e670a JK |
734 | } |
735 | ||
736 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
737 | { | |
502c6e0b | 738 | struct list_head *head; |
127e670a | 739 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 740 | unsigned int nentries = 0; |
bd936f84 | 741 | unsigned short index = 1; |
8c402946 | 742 | unsigned short orphan_blocks; |
4531929e | 743 | struct page *page = NULL; |
6451e041 | 744 | struct ino_entry *orphan = NULL; |
67298804 | 745 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 746 | |
67298804 | 747 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 748 | |
d6c67a4f JK |
749 | /* |
750 | * we don't need to do spin_lock(&im->ino_lock) here, since all the | |
751 | * orphan inode operations are covered under f2fs_lock_op(). | |
752 | * And, spin_lock should be avoided due to page operations below. | |
753 | */ | |
67298804 | 754 | head = &im->ino_list; |
127e670a JK |
755 | |
756 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
757 | list_for_each_entry(orphan, head, list) { |
758 | if (!page) { | |
4d57b86d | 759 | page = f2fs_grab_meta_page(sbi, start_blk++); |
502c6e0b GZ |
760 | orphan_blk = |
761 | (struct f2fs_orphan_block *)page_address(page); | |
762 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
763 | } | |
127e670a | 764 | |
36795567 | 765 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 766 | |
36795567 | 767 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
768 | /* |
769 | * an orphan block is full of 1020 entries, | |
770 | * then we need to flush current orphan blocks | |
771 | * and bring another one in memory | |
772 | */ | |
773 | orphan_blk->blk_addr = cpu_to_le16(index); | |
774 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
775 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
776 | set_page_dirty(page); | |
777 | f2fs_put_page(page, 1); | |
778 | index++; | |
127e670a JK |
779 | nentries = 0; |
780 | page = NULL; | |
781 | } | |
502c6e0b | 782 | } |
127e670a | 783 | |
502c6e0b GZ |
784 | if (page) { |
785 | orphan_blk->blk_addr = cpu_to_le16(index); | |
786 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
787 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
788 | set_page_dirty(page); | |
789 | f2fs_put_page(page, 1); | |
127e670a | 790 | } |
127e670a JK |
791 | } |
792 | ||
d7eb8f1c CY |
793 | static __u32 f2fs_checkpoint_chksum(struct f2fs_sb_info *sbi, |
794 | struct f2fs_checkpoint *ckpt) | |
795 | { | |
796 | unsigned int chksum_ofs = le32_to_cpu(ckpt->checksum_offset); | |
797 | __u32 chksum; | |
798 | ||
799 | chksum = f2fs_crc32(sbi, ckpt, chksum_ofs); | |
800 | if (chksum_ofs < CP_CHKSUM_OFFSET) { | |
801 | chksum_ofs += sizeof(chksum); | |
802 | chksum = f2fs_chksum(sbi, chksum, (__u8 *)ckpt + chksum_ofs, | |
803 | F2FS_BLKSIZE - chksum_ofs); | |
804 | } | |
805 | return chksum; | |
806 | } | |
807 | ||
fc0065ad TY |
808 | static int get_checkpoint_version(struct f2fs_sb_info *sbi, block_t cp_addr, |
809 | struct f2fs_checkpoint **cp_block, struct page **cp_page, | |
810 | unsigned long long *version) | |
127e670a | 811 | { |
fc0065ad | 812 | size_t crc_offset = 0; |
d7eb8f1c | 813 | __u32 crc; |
127e670a | 814 | |
4d57b86d | 815 | *cp_page = f2fs_get_meta_page(sbi, cp_addr); |
7735730d CY |
816 | if (IS_ERR(*cp_page)) |
817 | return PTR_ERR(*cp_page); | |
818 | ||
fc0065ad | 819 | *cp_block = (struct f2fs_checkpoint *)page_address(*cp_page); |
127e670a | 820 | |
fc0065ad | 821 | crc_offset = le32_to_cpu((*cp_block)->checksum_offset); |
d7eb8f1c CY |
822 | if (crc_offset < CP_MIN_CHKSUM_OFFSET || |
823 | crc_offset > CP_CHKSUM_OFFSET) { | |
d3f07c04 | 824 | f2fs_put_page(*cp_page, 1); |
dcbb4c10 | 825 | f2fs_warn(sbi, "invalid crc_offset: %zu", crc_offset); |
fc0065ad TY |
826 | return -EINVAL; |
827 | } | |
127e670a | 828 | |
d7eb8f1c CY |
829 | crc = f2fs_checkpoint_chksum(sbi, *cp_block); |
830 | if (crc != cur_cp_crc(*cp_block)) { | |
d3f07c04 | 831 | f2fs_put_page(*cp_page, 1); |
dcbb4c10 | 832 | f2fs_warn(sbi, "invalid crc value"); |
fc0065ad TY |
833 | return -EINVAL; |
834 | } | |
127e670a | 835 | |
fc0065ad TY |
836 | *version = cur_cp_version(*cp_block); |
837 | return 0; | |
838 | } | |
127e670a | 839 | |
fc0065ad TY |
840 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, |
841 | block_t cp_addr, unsigned long long *version) | |
842 | { | |
843 | struct page *cp_page_1 = NULL, *cp_page_2 = NULL; | |
844 | struct f2fs_checkpoint *cp_block = NULL; | |
845 | unsigned long long cur_version = 0, pre_version = 0; | |
846 | int err; | |
127e670a | 847 | |
fc0065ad TY |
848 | err = get_checkpoint_version(sbi, cp_addr, &cp_block, |
849 | &cp_page_1, version); | |
850 | if (err) | |
d3f07c04 | 851 | return NULL; |
c9b60788 CY |
852 | |
853 | if (le32_to_cpu(cp_block->cp_pack_total_block_count) > | |
854 | sbi->blocks_per_seg) { | |
dcbb4c10 JP |
855 | f2fs_warn(sbi, "invalid cp_pack_total_block_count:%u", |
856 | le32_to_cpu(cp_block->cp_pack_total_block_count)); | |
d3f07c04 | 857 | goto invalid_cp; |
c9b60788 | 858 | } |
fc0065ad | 859 | pre_version = *version; |
127e670a | 860 | |
fc0065ad TY |
861 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
862 | err = get_checkpoint_version(sbi, cp_addr, &cp_block, | |
863 | &cp_page_2, version); | |
864 | if (err) | |
d3f07c04 | 865 | goto invalid_cp; |
fc0065ad | 866 | cur_version = *version; |
127e670a JK |
867 | |
868 | if (cur_version == pre_version) { | |
869 | *version = cur_version; | |
870 | f2fs_put_page(cp_page_2, 1); | |
871 | return cp_page_1; | |
872 | } | |
127e670a | 873 | f2fs_put_page(cp_page_2, 1); |
d3f07c04 | 874 | invalid_cp: |
127e670a JK |
875 | f2fs_put_page(cp_page_1, 1); |
876 | return NULL; | |
877 | } | |
878 | ||
4d57b86d | 879 | int f2fs_get_valid_checkpoint(struct f2fs_sb_info *sbi) |
127e670a JK |
880 | { |
881 | struct f2fs_checkpoint *cp_block; | |
882 | struct f2fs_super_block *fsb = sbi->raw_super; | |
883 | struct page *cp1, *cp2, *cur_page; | |
884 | unsigned long blk_size = sbi->blocksize; | |
885 | unsigned long long cp1_version = 0, cp2_version = 0; | |
886 | unsigned long long cp_start_blk_no; | |
55141486 | 887 | unsigned int cp_blks = 1 + __cp_payload(sbi); |
1dbe4152 CL |
888 | block_t cp_blk_no; |
889 | int i; | |
10f966bb | 890 | int err; |
127e670a | 891 | |
026f0507 KC |
892 | sbi->ckpt = f2fs_kzalloc(sbi, array_size(blk_size, cp_blks), |
893 | GFP_KERNEL); | |
127e670a JK |
894 | if (!sbi->ckpt) |
895 | return -ENOMEM; | |
896 | /* | |
897 | * Finding out valid cp block involves read both | |
7a88ddb5 | 898 | * sets( cp pack 1 and cp pack 2) |
127e670a JK |
899 | */ |
900 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
901 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
902 | ||
903 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
904 | cp_start_blk_no += ((unsigned long long)1) << |
905 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
906 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
907 | ||
908 | if (cp1 && cp2) { | |
909 | if (ver_after(cp2_version, cp1_version)) | |
910 | cur_page = cp2; | |
911 | else | |
912 | cur_page = cp1; | |
913 | } else if (cp1) { | |
914 | cur_page = cp1; | |
915 | } else if (cp2) { | |
916 | cur_page = cp2; | |
917 | } else { | |
10f966bb | 918 | err = -EFSCORRUPTED; |
127e670a JK |
919 | goto fail_no_cp; |
920 | } | |
921 | ||
922 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
923 | memcpy(sbi->ckpt, cp_block, blk_size); | |
924 | ||
8508e44a JK |
925 | if (cur_page == cp1) |
926 | sbi->cur_cp_pack = 1; | |
927 | else | |
928 | sbi->cur_cp_pack = 2; | |
984ec63c | 929 | |
e494c2f9 | 930 | /* Sanity checking of checkpoint */ |
10f966bb CY |
931 | if (f2fs_sanity_check_ckpt(sbi)) { |
932 | err = -EFSCORRUPTED; | |
e494c2f9 | 933 | goto free_fail_no_cp; |
10f966bb | 934 | } |
e494c2f9 | 935 | |
1dbe4152 CL |
936 | if (cp_blks <= 1) |
937 | goto done; | |
938 | ||
939 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
940 | if (cur_page == cp2) | |
941 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
942 | ||
943 | for (i = 1; i < cp_blks; i++) { | |
944 | void *sit_bitmap_ptr; | |
945 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
946 | ||
4d57b86d | 947 | cur_page = f2fs_get_meta_page(sbi, cp_blk_no + i); |
10f966bb CY |
948 | if (IS_ERR(cur_page)) { |
949 | err = PTR_ERR(cur_page); | |
7735730d | 950 | goto free_fail_no_cp; |
10f966bb | 951 | } |
1dbe4152 CL |
952 | sit_bitmap_ptr = page_address(cur_page); |
953 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
954 | f2fs_put_page(cur_page, 1); | |
955 | } | |
956 | done: | |
127e670a JK |
957 | f2fs_put_page(cp1, 1); |
958 | f2fs_put_page(cp2, 1); | |
959 | return 0; | |
960 | ||
a2125ff7 JK |
961 | free_fail_no_cp: |
962 | f2fs_put_page(cp1, 1); | |
963 | f2fs_put_page(cp2, 1); | |
127e670a | 964 | fail_no_cp: |
5222595d | 965 | kvfree(sbi->ckpt); |
10f966bb | 966 | return err; |
127e670a JK |
967 | } |
968 | ||
c227f912 | 969 | static void __add_dirty_inode(struct inode *inode, enum inode_type type) |
127e670a | 970 | { |
4081363f | 971 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 972 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
127e670a | 973 | |
91942321 | 974 | if (is_inode_flag_set(inode, flag)) |
2710fd7e | 975 | return; |
2d7b822a | 976 | |
91942321 | 977 | set_inode_flag(inode, flag); |
99f4b917 CY |
978 | if (!f2fs_is_volatile_file(inode)) |
979 | list_add_tail(&F2FS_I(inode)->dirty_list, | |
980 | &sbi->inode_list[type]); | |
33fbd510 | 981 | stat_inc_dirty_inode(sbi, type); |
5deb8267 JK |
982 | } |
983 | ||
c227f912 | 984 | static void __remove_dirty_inode(struct inode *inode, enum inode_type type) |
6ad7609a | 985 | { |
c227f912 | 986 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
6ad7609a | 987 | |
91942321 | 988 | if (get_dirty_pages(inode) || !is_inode_flag_set(inode, flag)) |
6ad7609a CY |
989 | return; |
990 | ||
91942321 JK |
991 | list_del_init(&F2FS_I(inode)->dirty_list); |
992 | clear_inode_flag(inode, flag); | |
33fbd510 | 993 | stat_dec_dirty_inode(F2FS_I_SB(inode), type); |
6ad7609a CY |
994 | } |
995 | ||
4d57b86d | 996 | void f2fs_update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 997 | { |
4081363f | 998 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 999 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
5deb8267 | 1000 | |
5ac9f36f CY |
1001 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
1002 | !S_ISLNK(inode->i_mode)) | |
127e670a | 1003 | return; |
7bd59381 | 1004 | |
1c4bf763 JK |
1005 | spin_lock(&sbi->inode_lock[type]); |
1006 | if (type != FILE_INODE || test_opt(sbi, DATA_FLUSH)) | |
10aa97c3 | 1007 | __add_dirty_inode(inode, type); |
b951a4ec | 1008 | inode_inc_dirty_pages(inode); |
1c4bf763 JK |
1009 | spin_unlock(&sbi->inode_lock[type]); |
1010 | ||
240a5915 | 1011 | f2fs_set_page_private(page, 0); |
9e4ded3f | 1012 | f2fs_trace_pid(page); |
5deb8267 JK |
1013 | } |
1014 | ||
4d57b86d | 1015 | void f2fs_remove_dirty_inode(struct inode *inode) |
127e670a | 1016 | { |
4081363f | 1017 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 1018 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
127e670a | 1019 | |
c227f912 CY |
1020 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
1021 | !S_ISLNK(inode->i_mode)) | |
127e670a JK |
1022 | return; |
1023 | ||
10aa97c3 JK |
1024 | if (type == FILE_INODE && !test_opt(sbi, DATA_FLUSH)) |
1025 | return; | |
1026 | ||
c227f912 CY |
1027 | spin_lock(&sbi->inode_lock[type]); |
1028 | __remove_dirty_inode(inode, type); | |
1029 | spin_unlock(&sbi->inode_lock[type]); | |
74d0b917 JK |
1030 | } |
1031 | ||
4d57b86d | 1032 | int f2fs_sync_dirty_inodes(struct f2fs_sb_info *sbi, enum inode_type type) |
127e670a | 1033 | { |
ce3b7d80 | 1034 | struct list_head *head; |
127e670a | 1035 | struct inode *inode; |
2710fd7e | 1036 | struct f2fs_inode_info *fi; |
4cf18537 | 1037 | bool is_dir = (type == DIR_INODE); |
4db08d01 | 1038 | unsigned long ino = 0; |
4cf18537 CY |
1039 | |
1040 | trace_f2fs_sync_dirty_inodes_enter(sbi->sb, is_dir, | |
1041 | get_pages(sbi, is_dir ? | |
1042 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
127e670a | 1043 | retry: |
af41d3ee | 1044 | if (unlikely(f2fs_cp_error(sbi))) |
6d5a1495 | 1045 | return -EIO; |
af41d3ee | 1046 | |
c227f912 | 1047 | spin_lock(&sbi->inode_lock[type]); |
ce3b7d80 | 1048 | |
c227f912 | 1049 | head = &sbi->inode_list[type]; |
127e670a | 1050 | if (list_empty(head)) { |
c227f912 | 1051 | spin_unlock(&sbi->inode_lock[type]); |
4cf18537 CY |
1052 | trace_f2fs_sync_dirty_inodes_exit(sbi->sb, is_dir, |
1053 | get_pages(sbi, is_dir ? | |
1054 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
6d5a1495 | 1055 | return 0; |
127e670a | 1056 | } |
939afa94 | 1057 | fi = list_first_entry(head, struct f2fs_inode_info, dirty_list); |
2710fd7e | 1058 | inode = igrab(&fi->vfs_inode); |
c227f912 | 1059 | spin_unlock(&sbi->inode_lock[type]); |
127e670a | 1060 | if (inode) { |
4db08d01 JK |
1061 | unsigned long cur_ino = inode->i_ino; |
1062 | ||
186857c5 | 1063 | F2FS_I(inode)->cp_task = current; |
b0af6d49 | 1064 | |
87d6f890 | 1065 | filemap_fdatawrite(inode->i_mapping); |
b0af6d49 | 1066 | |
186857c5 | 1067 | F2FS_I(inode)->cp_task = NULL; |
b0af6d49 | 1068 | |
127e670a | 1069 | iput(inode); |
4db08d01 | 1070 | /* We need to give cpu to another writers. */ |
2a63531a | 1071 | if (ino == cur_ino) |
4db08d01 | 1072 | cond_resched(); |
2a63531a | 1073 | else |
4db08d01 | 1074 | ino = cur_ino; |
127e670a JK |
1075 | } else { |
1076 | /* | |
1077 | * We should submit bio, since it exists several | |
1078 | * wribacking dentry pages in the freeing inode. | |
1079 | */ | |
b9109b0e | 1080 | f2fs_submit_merged_write(sbi, DATA); |
7ecebe5e | 1081 | cond_resched(); |
127e670a JK |
1082 | } |
1083 | goto retry; | |
1084 | } | |
1085 | ||
0f18b462 JK |
1086 | int f2fs_sync_inode_meta(struct f2fs_sb_info *sbi) |
1087 | { | |
1088 | struct list_head *head = &sbi->inode_list[DIRTY_META]; | |
1089 | struct inode *inode; | |
1090 | struct f2fs_inode_info *fi; | |
1091 | s64 total = get_pages(sbi, F2FS_DIRTY_IMETA); | |
1092 | ||
1093 | while (total--) { | |
1094 | if (unlikely(f2fs_cp_error(sbi))) | |
1095 | return -EIO; | |
1096 | ||
1097 | spin_lock(&sbi->inode_lock[DIRTY_META]); | |
1098 | if (list_empty(head)) { | |
1099 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
1100 | return 0; | |
1101 | } | |
939afa94 | 1102 | fi = list_first_entry(head, struct f2fs_inode_info, |
0f18b462 JK |
1103 | gdirty_list); |
1104 | inode = igrab(&fi->vfs_inode); | |
1105 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
1106 | if (inode) { | |
18340edc JK |
1107 | sync_inode_metadata(inode, 0); |
1108 | ||
1109 | /* it's on eviction */ | |
1110 | if (is_inode_flag_set(inode, FI_DIRTY_INODE)) | |
4d57b86d | 1111 | f2fs_update_inode_page(inode); |
0f18b462 JK |
1112 | iput(inode); |
1113 | } | |
dee668c1 | 1114 | } |
0f18b462 JK |
1115 | return 0; |
1116 | } | |
1117 | ||
59c9081b YH |
1118 | static void __prepare_cp_block(struct f2fs_sb_info *sbi) |
1119 | { | |
1120 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1121 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1122 | nid_t last_nid = nm_i->next_scan_nid; | |
1123 | ||
1124 | next_free_nid(sbi, &last_nid); | |
1125 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
1126 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
1127 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
1128 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
1129 | } | |
1130 | ||
af033b2a CY |
1131 | static bool __need_flush_quota(struct f2fs_sb_info *sbi) |
1132 | { | |
db6ec53b JK |
1133 | bool ret = false; |
1134 | ||
af033b2a CY |
1135 | if (!is_journalled_quota(sbi)) |
1136 | return false; | |
db6ec53b JK |
1137 | |
1138 | down_write(&sbi->quota_sem); | |
1139 | if (is_sbi_flag_set(sbi, SBI_QUOTA_SKIP_FLUSH)) { | |
1140 | ret = false; | |
1141 | } else if (is_sbi_flag_set(sbi, SBI_QUOTA_NEED_REPAIR)) { | |
1142 | ret = false; | |
1143 | } else if (is_sbi_flag_set(sbi, SBI_QUOTA_NEED_FLUSH)) { | |
1144 | clear_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); | |
1145 | ret = true; | |
1146 | } else if (get_pages(sbi, F2FS_DIRTY_QDATA)) { | |
1147 | ret = true; | |
1148 | } | |
1149 | up_write(&sbi->quota_sem); | |
1150 | return ret; | |
af033b2a CY |
1151 | } |
1152 | ||
0a8165d7 | 1153 | /* |
127e670a JK |
1154 | * Freeze all the FS-operations for checkpoint. |
1155 | */ | |
cf779cab | 1156 | static int block_operations(struct f2fs_sb_info *sbi) |
127e670a | 1157 | { |
127e670a JK |
1158 | struct writeback_control wbc = { |
1159 | .sync_mode = WB_SYNC_ALL, | |
1160 | .nr_to_write = LONG_MAX, | |
1161 | .for_reclaim = 0, | |
1162 | }; | |
c718379b | 1163 | struct blk_plug plug; |
af033b2a | 1164 | int err = 0, cnt = 0; |
c718379b JK |
1165 | |
1166 | blk_start_plug(&plug); | |
1167 | ||
af033b2a | 1168 | retry_flush_quotas: |
db6ec53b | 1169 | f2fs_lock_all(sbi); |
af033b2a CY |
1170 | if (__need_flush_quota(sbi)) { |
1171 | int locked; | |
1172 | ||
1173 | if (++cnt > DEFAULT_RETRY_QUOTA_FLUSH_COUNT) { | |
1174 | set_sbi_flag(sbi, SBI_QUOTA_SKIP_FLUSH); | |
db6ec53b | 1175 | set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); |
af033b2a CY |
1176 | goto retry_flush_dents; |
1177 | } | |
db6ec53b | 1178 | f2fs_unlock_all(sbi); |
af033b2a CY |
1179 | |
1180 | /* only failed during mount/umount/freeze/quotactl */ | |
1181 | locked = down_read_trylock(&sbi->sb->s_umount); | |
1182 | f2fs_quota_sync(sbi->sb, -1); | |
1183 | if (locked) | |
1184 | up_read(&sbi->sb->s_umount); | |
af033b2a CY |
1185 | cond_resched(); |
1186 | goto retry_flush_quotas; | |
1187 | } | |
1188 | ||
1189 | retry_flush_dents: | |
127e670a | 1190 | /* write all the dirty dentry pages */ |
127e670a | 1191 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 1192 | f2fs_unlock_all(sbi); |
4d57b86d | 1193 | err = f2fs_sync_dirty_inodes(sbi, DIR_INODE); |
6d5a1495 | 1194 | if (err) |
cf779cab | 1195 | goto out; |
30973883 | 1196 | cond_resched(); |
af033b2a | 1197 | goto retry_flush_quotas; |
127e670a JK |
1198 | } |
1199 | ||
59c9081b YH |
1200 | /* |
1201 | * POR: we should ensure that there are no dirty node pages | |
1202 | * until finishing nat/sit flush. inode->i_blocks can be updated. | |
1203 | */ | |
1204 | down_write(&sbi->node_change); | |
1205 | ||
0f18b462 | 1206 | if (get_pages(sbi, F2FS_DIRTY_IMETA)) { |
59c9081b | 1207 | up_write(&sbi->node_change); |
0f18b462 JK |
1208 | f2fs_unlock_all(sbi); |
1209 | err = f2fs_sync_inode_meta(sbi); | |
1210 | if (err) | |
1211 | goto out; | |
30973883 | 1212 | cond_resched(); |
af033b2a | 1213 | goto retry_flush_quotas; |
0f18b462 JK |
1214 | } |
1215 | ||
39936837 | 1216 | retry_flush_nodes: |
b3582c68 | 1217 | down_write(&sbi->node_write); |
127e670a JK |
1218 | |
1219 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
b3582c68 | 1220 | up_write(&sbi->node_write); |
c29fd0c0 | 1221 | atomic_inc(&sbi->wb_sync_req[NODE]); |
4d57b86d | 1222 | err = f2fs_sync_node_pages(sbi, &wbc, false, FS_CP_NODE_IO); |
c29fd0c0 | 1223 | atomic_dec(&sbi->wb_sync_req[NODE]); |
6d5a1495 | 1224 | if (err) { |
59c9081b | 1225 | up_write(&sbi->node_change); |
cf779cab | 1226 | f2fs_unlock_all(sbi); |
cf779cab JK |
1227 | goto out; |
1228 | } | |
30973883 | 1229 | cond_resched(); |
39936837 | 1230 | goto retry_flush_nodes; |
127e670a | 1231 | } |
59c9081b YH |
1232 | |
1233 | /* | |
1234 | * sbi->node_change is used only for AIO write_begin path which produces | |
1235 | * dirty node blocks and some checkpoint values by block allocation. | |
1236 | */ | |
1237 | __prepare_cp_block(sbi); | |
1238 | up_write(&sbi->node_change); | |
cf779cab | 1239 | out: |
c718379b | 1240 | blk_finish_plug(&plug); |
cf779cab | 1241 | return err; |
127e670a JK |
1242 | } |
1243 | ||
1244 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
1245 | { | |
b3582c68 | 1246 | up_write(&sbi->node_write); |
e479556b | 1247 | f2fs_unlock_all(sbi); |
127e670a JK |
1248 | } |
1249 | ||
bf22c3cc | 1250 | void f2fs_wait_on_all_pages(struct f2fs_sb_info *sbi, int type) |
fb51b5ef CL |
1251 | { |
1252 | DEFINE_WAIT(wait); | |
1253 | ||
1254 | for (;;) { | |
1255 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
1256 | ||
bf22c3cc | 1257 | if (!get_pages(sbi, type)) |
fb51b5ef CL |
1258 | break; |
1259 | ||
af697c0f JK |
1260 | if (unlikely(f2fs_cp_error(sbi))) |
1261 | break; | |
1262 | ||
5df7731f | 1263 | io_schedule_timeout(DEFAULT_IO_TIMEOUT); |
fb51b5ef CL |
1264 | } |
1265 | finish_wait(&sbi->cp_wait, &wait); | |
1266 | } | |
1267 | ||
e4c5d848 JK |
1268 | static void update_ckpt_flags(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
1269 | { | |
1270 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; | |
1271 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
d1aa2453 | 1272 | unsigned long flags; |
e4c5d848 | 1273 | |
d1aa2453 | 1274 | spin_lock_irqsave(&sbi->cp_lock, flags); |
e4c5d848 | 1275 | |
c473f1a9 | 1276 | if ((cpc->reason & CP_UMOUNT) && |
10047f53 | 1277 | le32_to_cpu(ckpt->cp_pack_total_block_count) > |
22ad0b6a JK |
1278 | sbi->blocks_per_seg - NM_I(sbi)->nat_bits_blocks) |
1279 | disable_nat_bits(sbi, false); | |
1280 | ||
1f43e2ad CY |
1281 | if (cpc->reason & CP_TRIMMED) |
1282 | __set_ckpt_flags(ckpt, CP_TRIMMED_FLAG); | |
cd36d7a1 CY |
1283 | else |
1284 | __clear_ckpt_flags(ckpt, CP_TRIMMED_FLAG); | |
1f43e2ad | 1285 | |
c473f1a9 | 1286 | if (cpc->reason & CP_UMOUNT) |
e4c5d848 JK |
1287 | __set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); |
1288 | else | |
1289 | __clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
1290 | ||
c473f1a9 | 1291 | if (cpc->reason & CP_FASTBOOT) |
e4c5d848 JK |
1292 | __set_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); |
1293 | else | |
1294 | __clear_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
1295 | ||
1296 | if (orphan_num) | |
1297 | __set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); | |
1298 | else | |
1299 | __clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); | |
1300 | ||
c84ef3c5 | 1301 | if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) |
e4c5d848 JK |
1302 | __set_ckpt_flags(ckpt, CP_FSCK_FLAG); |
1303 | ||
c84ef3c5 ST |
1304 | if (is_sbi_flag_set(sbi, SBI_IS_RESIZEFS)) |
1305 | __set_ckpt_flags(ckpt, CP_RESIZEFS_FLAG); | |
1306 | else | |
1307 | __clear_ckpt_flags(ckpt, CP_RESIZEFS_FLAG); | |
1308 | ||
4354994f DR |
1309 | if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) |
1310 | __set_ckpt_flags(ckpt, CP_DISABLED_FLAG); | |
1311 | else | |
1312 | __clear_ckpt_flags(ckpt, CP_DISABLED_FLAG); | |
1313 | ||
db610a64 JK |
1314 | if (is_sbi_flag_set(sbi, SBI_CP_DISABLED_QUICK)) |
1315 | __set_ckpt_flags(ckpt, CP_DISABLED_QUICK_FLAG); | |
1316 | else | |
1317 | __clear_ckpt_flags(ckpt, CP_DISABLED_QUICK_FLAG); | |
1318 | ||
af033b2a CY |
1319 | if (is_sbi_flag_set(sbi, SBI_QUOTA_SKIP_FLUSH)) |
1320 | __set_ckpt_flags(ckpt, CP_QUOTA_NEED_FSCK_FLAG); | |
bc88ac96 JK |
1321 | else |
1322 | __clear_ckpt_flags(ckpt, CP_QUOTA_NEED_FSCK_FLAG); | |
af033b2a CY |
1323 | |
1324 | if (is_sbi_flag_set(sbi, SBI_QUOTA_NEED_REPAIR)) | |
1325 | __set_ckpt_flags(ckpt, CP_QUOTA_NEED_FSCK_FLAG); | |
1326 | ||
e4c5d848 JK |
1327 | /* set this flag to activate crc|cp_ver for recovery */ |
1328 | __set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG); | |
f2367923 | 1329 | __clear_ckpt_flags(ckpt, CP_NOCRC_RECOVERY_FLAG); |
e4c5d848 | 1330 | |
d1aa2453 | 1331 | spin_unlock_irqrestore(&sbi->cp_lock, flags); |
e4c5d848 JK |
1332 | } |
1333 | ||
46706d59 GX |
1334 | static void commit_checkpoint(struct f2fs_sb_info *sbi, |
1335 | void *src, block_t blk_addr) | |
1336 | { | |
1337 | struct writeback_control wbc = { | |
1338 | .for_reclaim = 0, | |
1339 | }; | |
1340 | ||
1341 | /* | |
1342 | * pagevec_lookup_tag and lock_page again will take | |
4d57b86d CY |
1343 | * some extra time. Therefore, f2fs_update_meta_pages and |
1344 | * f2fs_sync_meta_pages are combined in this function. | |
46706d59 | 1345 | */ |
4d57b86d | 1346 | struct page *page = f2fs_grab_meta_page(sbi, blk_addr); |
46706d59 GX |
1347 | int err; |
1348 | ||
bae0ee7a | 1349 | f2fs_wait_on_page_writeback(page, META, true, true); |
8d64d365 CY |
1350 | |
1351 | memcpy(page_address(page), src, PAGE_SIZE); | |
1352 | ||
1353 | set_page_dirty(page); | |
46706d59 GX |
1354 | if (unlikely(!clear_page_dirty_for_io(page))) |
1355 | f2fs_bug_on(sbi, 1); | |
1356 | ||
1357 | /* writeout cp pack 2 page */ | |
1358 | err = __f2fs_write_meta_page(page, &wbc, FS_CP_META_IO); | |
af697c0f JK |
1359 | if (unlikely(err && f2fs_cp_error(sbi))) { |
1360 | f2fs_put_page(page, 1); | |
1361 | return; | |
1362 | } | |
46706d59 | 1363 | |
af697c0f | 1364 | f2fs_bug_on(sbi, err); |
46706d59 GX |
1365 | f2fs_put_page(page, 0); |
1366 | ||
1367 | /* submit checkpoint (with barrier if NOBARRIER is not set) */ | |
1368 | f2fs_submit_merged_write(sbi, META_FLUSH); | |
1369 | } | |
1370 | ||
c34f42e2 | 1371 | static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1372 | { |
1373 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
77041823 | 1374 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
d1aa2453 | 1375 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num, flags; |
127e670a | 1376 | block_t start_blk; |
127e670a | 1377 | unsigned int data_sum_blocks, orphan_blocks; |
7e586fa0 | 1378 | __u32 crc32 = 0; |
127e670a | 1379 | int i; |
55141486 | 1380 | int cp_payload_blks = __cp_payload(sbi); |
8f1dbbbb SL |
1381 | struct super_block *sb = sbi->sb; |
1382 | struct curseg_info *seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); | |
1383 | u64 kbytes_written; | |
1228b482 | 1384 | int err; |
127e670a JK |
1385 | |
1386 | /* Flush all the NAT/SIT pages */ | |
8ec18bff | 1387 | f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_CP_META_IO); |
127e670a | 1388 | |
7a88ddb5 | 1389 | /* start to update checkpoint, cp ver is already updated previously */ |
a1f72ac2 | 1390 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi, true)); |
127e670a | 1391 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); |
b5b82205 | 1392 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
1393 | ckpt->cur_node_segno[i] = |
1394 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
1395 | ckpt->cur_node_blkoff[i] = | |
1396 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
1397 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
1398 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
1399 | } | |
b5b82205 | 1400 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
1401 | ckpt->cur_data_segno[i] = |
1402 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
1403 | ckpt->cur_data_blkoff[i] = | |
1404 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
1405 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
1406 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
1407 | } | |
1408 | ||
127e670a | 1409 | /* 2 cp + n data seg summary + orphan inode blocks */ |
4d57b86d | 1410 | data_sum_blocks = f2fs_npages_for_summary_flush(sbi, false); |
d1aa2453 | 1411 | spin_lock_irqsave(&sbi->cp_lock, flags); |
b5b82205 | 1412 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
aaec2b1d | 1413 | __set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 1414 | else |
aaec2b1d | 1415 | __clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
d1aa2453 | 1416 | spin_unlock_irqrestore(&sbi->cp_lock, flags); |
127e670a | 1417 | |
67298804 | 1418 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
1419 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
1420 | orphan_blocks); | |
127e670a | 1421 | |
119ee914 | 1422 | if (__remain_node_summaries(cpc->reason)) |
b5b82205 | 1423 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
1424 | cp_payload_blks + data_sum_blocks + |
1425 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
119ee914 | 1426 | else |
b5b82205 | 1427 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS + |
1dbe4152 CL |
1428 | cp_payload_blks + data_sum_blocks + |
1429 | orphan_blocks); | |
119ee914 | 1430 | |
e4c5d848 JK |
1431 | /* update ckpt flag for checkpoint */ |
1432 | update_ckpt_flags(sbi, cpc); | |
a468f0ef | 1433 | |
127e670a JK |
1434 | /* update SIT/NAT bitmap */ |
1435 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
1436 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
1437 | ||
d7eb8f1c | 1438 | crc32 = f2fs_checkpoint_chksum(sbi, ckpt); |
7e586fa0 JK |
1439 | *((__le32 *)((unsigned char *)ckpt + |
1440 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
1441 | = cpu_to_le32(crc32); |
1442 | ||
8508e44a | 1443 | start_blk = __start_cp_next_addr(sbi); |
127e670a | 1444 | |
22ad0b6a JK |
1445 | /* write nat bits */ |
1446 | if (enabled_nat_bits(sbi, cpc)) { | |
1447 | __u64 cp_ver = cur_cp_version(ckpt); | |
22ad0b6a JK |
1448 | block_t blk; |
1449 | ||
1450 | cp_ver |= ((__u64)crc32 << 32); | |
1451 | *(__le64 *)nm_i->nat_bits = cpu_to_le64(cp_ver); | |
1452 | ||
1453 | blk = start_blk + sbi->blocks_per_seg - nm_i->nat_bits_blocks; | |
1454 | for (i = 0; i < nm_i->nat_bits_blocks; i++) | |
4d57b86d | 1455 | f2fs_update_meta_page(sbi, nm_i->nat_bits + |
22ad0b6a | 1456 | (i << F2FS_BLKSIZE_BITS), blk + i); |
22ad0b6a JK |
1457 | } |
1458 | ||
127e670a | 1459 | /* write out checkpoint buffer at block 0 */ |
4d57b86d | 1460 | f2fs_update_meta_page(sbi, ckpt, start_blk++); |
381722d2 CY |
1461 | |
1462 | for (i = 1; i < 1 + cp_payload_blks; i++) | |
4d57b86d | 1463 | f2fs_update_meta_page(sbi, (char *)ckpt + i * F2FS_BLKSIZE, |
381722d2 | 1464 | start_blk++); |
1dbe4152 | 1465 | |
67298804 | 1466 | if (orphan_num) { |
127e670a JK |
1467 | write_orphan_inodes(sbi, start_blk); |
1468 | start_blk += orphan_blocks; | |
1469 | } | |
1470 | ||
4d57b86d | 1471 | f2fs_write_data_summaries(sbi, start_blk); |
127e670a | 1472 | start_blk += data_sum_blocks; |
8f1dbbbb SL |
1473 | |
1474 | /* Record write statistics in the hot node summary */ | |
1475 | kbytes_written = sbi->kbytes_written; | |
1476 | if (sb->s_bdev->bd_part) | |
1477 | kbytes_written += BD_PART_WRITTEN(sbi); | |
1478 | ||
b7ad7512 | 1479 | seg_i->journal->info.kbytes_written = cpu_to_le64(kbytes_written); |
8f1dbbbb | 1480 | |
119ee914 | 1481 | if (__remain_node_summaries(cpc->reason)) { |
4d57b86d | 1482 | f2fs_write_node_summaries(sbi, start_blk); |
127e670a JK |
1483 | start_blk += NR_CURSEG_NODE_TYPE; |
1484 | } | |
1485 | ||
46706d59 GX |
1486 | /* update user_block_counts */ |
1487 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
1488 | percpu_counter_set(&sbi->alloc_valid_block_count, 0); | |
127e670a | 1489 | |
46706d59 | 1490 | /* Here, we have one bio having CP pack except cp pack 2 page */ |
4d57b86d | 1491 | f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_CP_META_IO); |
bf22c3cc ST |
1492 | /* Wait for all dirty meta pages to be submitted for IO */ |
1493 | f2fs_wait_on_all_pages(sbi, F2FS_DIRTY_META); | |
46706d59 GX |
1494 | |
1495 | /* wait for previous submitted meta pages writeback */ | |
bf22c3cc | 1496 | f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); |
127e670a | 1497 | |
46706d59 GX |
1498 | /* flush all device cache */ |
1499 | err = f2fs_flush_device_cache(sbi); | |
1500 | if (err) | |
1501 | return err; | |
127e670a | 1502 | |
46706d59 GX |
1503 | /* barrier and flush checkpoint cp pack 2 page if it can */ |
1504 | commit_checkpoint(sbi, ckpt, start_blk); | |
bf22c3cc | 1505 | f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); |
6a8f8ca5 | 1506 | |
18767e62 | 1507 | /* |
542989b6 EB |
1508 | * invalidate intermediate page cache borrowed from meta inode which are |
1509 | * used for migration of encrypted or verity inode's blocks. | |
18767e62 | 1510 | */ |
542989b6 | 1511 | if (f2fs_sb_has_encrypt(sbi) || f2fs_sb_has_verity(sbi)) |
18767e62 CY |
1512 | invalidate_mapping_pages(META_MAPPING(sbi), |
1513 | MAIN_BLKADDR(sbi), MAX_BLKADDR(sbi) - 1); | |
1514 | ||
4d57b86d | 1515 | f2fs_release_ino_entry(sbi, false); |
cf779cab | 1516 | |
50fa53ec CY |
1517 | f2fs_reset_fsync_node_info(sbi); |
1518 | ||
caf0047e | 1519 | clear_sbi_flag(sbi, SBI_IS_DIRTY); |
bbf156f7 | 1520 | clear_sbi_flag(sbi, SBI_NEED_CP); |
af033b2a | 1521 | clear_sbi_flag(sbi, SBI_QUOTA_SKIP_FLUSH); |
c9c8ed50 CY |
1522 | |
1523 | spin_lock(&sbi->stat_lock); | |
4354994f | 1524 | sbi->unusable_block_count = 0; |
c9c8ed50 CY |
1525 | spin_unlock(&sbi->stat_lock); |
1526 | ||
8508e44a | 1527 | __set_cp_next_pack(sbi); |
c34f42e2 | 1528 | |
c2a080ae CY |
1529 | /* |
1530 | * redirty superblock if metadata like node page or inode cache is | |
1531 | * updated during writing checkpoint. | |
1532 | */ | |
1533 | if (get_pages(sbi, F2FS_DIRTY_NODES) || | |
1534 | get_pages(sbi, F2FS_DIRTY_IMETA)) | |
1535 | set_sbi_flag(sbi, SBI_IS_DIRTY); | |
1536 | ||
1537 | f2fs_bug_on(sbi, get_pages(sbi, F2FS_DIRTY_DENTS)); | |
1538 | ||
af697c0f | 1539 | return unlikely(f2fs_cp_error(sbi)) ? -EIO : 0; |
127e670a JK |
1540 | } |
1541 | ||
4d57b86d | 1542 | int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1543 | { |
1544 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1545 | unsigned long long ckpt_ver; | |
c34f42e2 | 1546 | int err = 0; |
127e670a | 1547 | |
f5a131bb CY |
1548 | if (f2fs_readonly(sbi->sb) || f2fs_hw_is_readonly(sbi)) |
1549 | return -EROFS; | |
1550 | ||
4354994f DR |
1551 | if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { |
1552 | if (cpc->reason != CP_PAUSE) | |
1553 | return 0; | |
dcbb4c10 | 1554 | f2fs_warn(sbi, "Start checkpoint disabled!"); |
4354994f | 1555 | } |
43727527 | 1556 | mutex_lock(&sbi->cp_mutex); |
8501017e | 1557 | |
caf0047e | 1558 | if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) && |
c473f1a9 CY |
1559 | ((cpc->reason & CP_FASTBOOT) || (cpc->reason & CP_SYNC) || |
1560 | ((cpc->reason & CP_DISCARD) && !sbi->discard_blks))) | |
8501017e | 1561 | goto out; |
c34f42e2 CY |
1562 | if (unlikely(f2fs_cp_error(sbi))) { |
1563 | err = -EIO; | |
cf779cab | 1564 | goto out; |
c34f42e2 | 1565 | } |
2bda542d WL |
1566 | |
1567 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); | |
1568 | ||
c34f42e2 CY |
1569 | err = block_operations(sbi); |
1570 | if (err) | |
cf779cab | 1571 | goto out; |
127e670a | 1572 | |
75ab4cb8 | 1573 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1574 | |
b9109b0e | 1575 | f2fs_flush_merged_writes(sbi); |
127e670a | 1576 | |
58cce381 | 1577 | /* this is the case of multiple fstrims without any changes */ |
c473f1a9 | 1578 | if (cpc->reason & CP_DISCARD) { |
4d57b86d | 1579 | if (!f2fs_exist_trim_candidates(sbi, cpc)) { |
25290fa5 JK |
1580 | unblock_operations(sbi); |
1581 | goto out; | |
1582 | } | |
1583 | ||
0333ad4e JK |
1584 | if (NM_I(sbi)->dirty_nat_cnt == 0 && |
1585 | SIT_I(sbi)->dirty_sentries == 0 && | |
1586 | prefree_segments(sbi) == 0) { | |
4d57b86d CY |
1587 | f2fs_flush_sit_entries(sbi, cpc); |
1588 | f2fs_clear_prefree_segments(sbi, cpc); | |
0333ad4e JK |
1589 | unblock_operations(sbi); |
1590 | goto out; | |
1591 | } | |
58cce381 YH |
1592 | } |
1593 | ||
127e670a JK |
1594 | /* |
1595 | * update checkpoint pack index | |
1596 | * Increase the version number so that | |
1597 | * SIT entries and seg summaries are written at correct place | |
1598 | */ | |
d71b5564 | 1599 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1600 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1601 | ||
1602 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
edc55aaf JK |
1603 | err = f2fs_flush_nat_entries(sbi, cpc); |
1604 | if (err) | |
1605 | goto stop; | |
1606 | ||
4d57b86d | 1607 | f2fs_flush_sit_entries(sbi, cpc); |
127e670a | 1608 | |
c34f42e2 | 1609 | err = do_checkpoint(sbi, cpc); |
4e6a8d9b | 1610 | if (err) |
4d57b86d | 1611 | f2fs_release_discard_addrs(sbi); |
4e6a8d9b | 1612 | else |
4d57b86d | 1613 | f2fs_clear_prefree_segments(sbi, cpc); |
edc55aaf | 1614 | stop: |
127e670a | 1615 | unblock_operations(sbi); |
942e0be6 | 1616 | stat_inc_cp_count(sbi->stat_info); |
10027551 | 1617 | |
c473f1a9 | 1618 | if (cpc->reason & CP_RECOVERY) |
dcbb4c10 | 1619 | f2fs_notice(sbi, "checkpoint: version = %llx", ckpt_ver); |
60b99b48 | 1620 | |
7a88ddb5 | 1621 | /* update CP_TIME to trigger checkpoint periodically */ |
6beceb54 | 1622 | f2fs_update_time(sbi, CP_TIME); |
55d1cdb2 | 1623 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
8501017e JK |
1624 | out: |
1625 | mutex_unlock(&sbi->cp_mutex); | |
c34f42e2 | 1626 | return err; |
127e670a JK |
1627 | } |
1628 | ||
4d57b86d | 1629 | void f2fs_init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1630 | { |
6451e041 JK |
1631 | int i; |
1632 | ||
1633 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1634 | struct inode_management *im = &sbi->im[i]; |
1635 | ||
1636 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1637 | spin_lock_init(&im->ino_lock); | |
1638 | INIT_LIST_HEAD(&im->ino_list); | |
1639 | im->ino_num = 0; | |
6451e041 JK |
1640 | } |
1641 | ||
b5b82205 | 1642 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
14b42817 WL |
1643 | NR_CURSEG_TYPE - __cp_payload(sbi)) * |
1644 | F2FS_ORPHANS_PER_BLOCK; | |
127e670a JK |
1645 | } |
1646 | ||
4d57b86d | 1647 | int __init f2fs_create_checkpoint_caches(void) |
127e670a | 1648 | { |
6451e041 JK |
1649 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1650 | sizeof(struct ino_entry)); | |
1651 | if (!ino_entry_slab) | |
127e670a | 1652 | return -ENOMEM; |
4d57b86d | 1653 | f2fs_inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry", |
06292073 | 1654 | sizeof(struct inode_entry)); |
4d57b86d | 1655 | if (!f2fs_inode_entry_slab) { |
6451e041 | 1656 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1657 | return -ENOMEM; |
1658 | } | |
1659 | return 0; | |
1660 | } | |
1661 | ||
4d57b86d | 1662 | void f2fs_destroy_checkpoint_caches(void) |
127e670a | 1663 | { |
6451e041 | 1664 | kmem_cache_destroy(ino_entry_slab); |
4d57b86d | 1665 | kmem_cache_destroy(f2fs_inode_entry_slab); |
127e670a | 1666 | } |