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 | |
8b83ac81 CY |
89 | f2fs_update_iostat(sbi, FS_META_READ_IO, F2FS_BLKSIZE); |
90 | ||
393ff91f | 91 | lock_page(page); |
6bacf52f | 92 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
93 | f2fs_put_page(page, 1); |
94 | goto repeat; | |
95 | } | |
f3f338ca | 96 | |
81114baa | 97 | if (unlikely(!PageUptodate(page))) { |
7735730d CY |
98 | f2fs_put_page(page, 1); |
99 | return ERR_PTR(-EIO); | |
81114baa | 100 | } |
393ff91f | 101 | out: |
127e670a JK |
102 | return page; |
103 | } | |
104 | ||
4d57b86d | 105 | struct page *f2fs_get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) |
2b947003 CY |
106 | { |
107 | return __get_meta_page(sbi, index, true); | |
108 | } | |
109 | ||
86f33603 | 110 | struct page *f2fs_get_meta_page_retry(struct f2fs_sb_info *sbi, pgoff_t index) |
7735730d CY |
111 | { |
112 | struct page *page; | |
113 | int count = 0; | |
114 | ||
115 | retry: | |
116 | page = __get_meta_page(sbi, index, true); | |
117 | if (IS_ERR(page)) { | |
118 | if (PTR_ERR(page) == -EIO && | |
119 | ++count <= DEFAULT_RETRY_IO_COUNT) | |
120 | goto retry; | |
7735730d | 121 | f2fs_stop_checkpoint(sbi, false); |
7735730d | 122 | } |
7735730d CY |
123 | return page; |
124 | } | |
125 | ||
2b947003 | 126 | /* for POR only */ |
4d57b86d | 127 | struct page *f2fs_get_tmp_page(struct f2fs_sb_info *sbi, pgoff_t index) |
2b947003 CY |
128 | { |
129 | return __get_meta_page(sbi, index, false); | |
130 | } | |
131 | ||
93770ab7 CY |
132 | static bool __is_bitmap_valid(struct f2fs_sb_info *sbi, block_t blkaddr, |
133 | int type) | |
134 | { | |
135 | struct seg_entry *se; | |
136 | unsigned int segno, offset; | |
137 | bool exist; | |
138 | ||
139 | if (type != DATA_GENERIC_ENHANCE && type != DATA_GENERIC_ENHANCE_READ) | |
140 | return true; | |
141 | ||
142 | segno = GET_SEGNO(sbi, blkaddr); | |
143 | offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr); | |
144 | se = get_seg_entry(sbi, segno); | |
145 | ||
146 | exist = f2fs_test_bit(offset, se->cur_valid_map); | |
147 | if (!exist && type == DATA_GENERIC_ENHANCE) { | |
dcbb4c10 JP |
148 | f2fs_err(sbi, "Inconsistent error blkaddr:%u, sit bitmap:%d", |
149 | blkaddr, exist); | |
93770ab7 CY |
150 | set_sbi_flag(sbi, SBI_NEED_FSCK); |
151 | WARN_ON(1); | |
152 | } | |
153 | return exist; | |
154 | } | |
155 | ||
e1da7872 | 156 | bool f2fs_is_valid_blkaddr(struct f2fs_sb_info *sbi, |
4d57b86d | 157 | block_t blkaddr, int type) |
662befda CY |
158 | { |
159 | switch (type) { | |
160 | case META_NAT: | |
66b00c18 | 161 | break; |
662befda | 162 | case META_SIT: |
66b00c18 CY |
163 | if (unlikely(blkaddr >= SIT_BLK_CNT(sbi))) |
164 | return false; | |
165 | break; | |
81c1a0f1 | 166 | case META_SSA: |
66b00c18 CY |
167 | if (unlikely(blkaddr >= MAIN_BLKADDR(sbi) || |
168 | blkaddr < SM_I(sbi)->ssa_blkaddr)) | |
169 | return false; | |
170 | break; | |
662befda | 171 | case META_CP: |
66b00c18 CY |
172 | if (unlikely(blkaddr >= SIT_I(sbi)->sit_base_addr || |
173 | blkaddr < __start_cp_addr(sbi))) | |
174 | return false; | |
175 | break; | |
4c521f49 | 176 | case META_POR: |
93770ab7 CY |
177 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
178 | blkaddr < MAIN_BLKADDR(sbi))) | |
179 | return false; | |
180 | break; | |
e1da7872 | 181 | case DATA_GENERIC: |
93770ab7 CY |
182 | case DATA_GENERIC_ENHANCE: |
183 | case DATA_GENERIC_ENHANCE_READ: | |
66b00c18 | 184 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
93770ab7 | 185 | blkaddr < MAIN_BLKADDR(sbi))) { |
dcbb4c10 JP |
186 | f2fs_warn(sbi, "access invalid blkaddr:%u", |
187 | blkaddr); | |
93770ab7 CY |
188 | set_sbi_flag(sbi, SBI_NEED_FSCK); |
189 | WARN_ON(1); | |
66b00c18 | 190 | return false; |
93770ab7 CY |
191 | } else { |
192 | return __is_bitmap_valid(sbi, blkaddr, type); | |
c9b60788 | 193 | } |
66b00c18 | 194 | break; |
e1da7872 CY |
195 | case META_GENERIC: |
196 | if (unlikely(blkaddr < SEG0_BLKADDR(sbi) || | |
197 | blkaddr >= MAIN_BLKADDR(sbi))) | |
198 | return false; | |
199 | break; | |
662befda CY |
200 | default: |
201 | BUG(); | |
202 | } | |
66b00c18 CY |
203 | |
204 | return true; | |
662befda CY |
205 | } |
206 | ||
207 | /* | |
7a88ddb5 | 208 | * Readahead CP/NAT/SIT/SSA/POR pages |
662befda | 209 | */ |
4d57b86d | 210 | int f2fs_ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages, |
26879fb1 | 211 | int type, bool sync) |
662befda | 212 | { |
662befda | 213 | struct page *page; |
4c521f49 | 214 | block_t blkno = start; |
662befda | 215 | struct f2fs_io_info fio = { |
05ca3632 | 216 | .sbi = sbi, |
662befda | 217 | .type = META, |
04d328de | 218 | .op = REQ_OP_READ, |
70fd7614 | 219 | .op_flags = sync ? (REQ_META | REQ_PRIO) : REQ_RAHEAD, |
4375a336 | 220 | .encrypted_page = NULL, |
fb830fc5 | 221 | .in_list = false, |
6dc3a126 | 222 | .is_por = (type == META_POR), |
662befda | 223 | }; |
e9f5b8b8 | 224 | struct blk_plug plug; |
ce4c638c | 225 | int err; |
662befda | 226 | |
2b947003 | 227 | if (unlikely(type == META_POR)) |
04d328de | 228 | fio.op_flags &= ~REQ_META; |
2b947003 | 229 | |
e9f5b8b8 | 230 | blk_start_plug(&plug); |
662befda | 231 | for (; nrpages-- > 0; blkno++) { |
662befda | 232 | |
e1da7872 | 233 | if (!f2fs_is_valid_blkaddr(sbi, blkno, type)) |
66b00c18 CY |
234 | goto out; |
235 | ||
662befda CY |
236 | switch (type) { |
237 | case META_NAT: | |
66b00c18 CY |
238 | if (unlikely(blkno >= |
239 | NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))) | |
662befda | 240 | blkno = 0; |
66b00c18 | 241 | /* get nat block addr */ |
7a9d7548 | 242 | fio.new_blkaddr = current_nat_addr(sbi, |
662befda CY |
243 | blkno * NAT_ENTRY_PER_BLOCK); |
244 | break; | |
245 | case META_SIT: | |
6a257471 CY |
246 | if (unlikely(blkno >= TOTAL_SEGS(sbi))) |
247 | goto out; | |
662befda | 248 | /* get sit block addr */ |
7a9d7548 | 249 | fio.new_blkaddr = current_sit_addr(sbi, |
662befda | 250 | blkno * SIT_ENTRY_PER_BLOCK); |
662befda | 251 | break; |
81c1a0f1 | 252 | case META_SSA: |
662befda | 253 | case META_CP: |
4c521f49 | 254 | case META_POR: |
7a9d7548 | 255 | fio.new_blkaddr = blkno; |
662befda CY |
256 | break; |
257 | default: | |
258 | BUG(); | |
259 | } | |
260 | ||
300e129c JK |
261 | page = f2fs_grab_cache_page(META_MAPPING(sbi), |
262 | fio.new_blkaddr, false); | |
662befda CY |
263 | if (!page) |
264 | continue; | |
265 | if (PageUptodate(page)) { | |
662befda CY |
266 | f2fs_put_page(page, 1); |
267 | continue; | |
268 | } | |
269 | ||
05ca3632 | 270 | fio.page = page; |
ce4c638c CY |
271 | err = f2fs_submit_page_bio(&fio); |
272 | f2fs_put_page(page, err ? 1 : 0); | |
8b83ac81 CY |
273 | |
274 | if (!err) | |
275 | f2fs_update_iostat(sbi, FS_META_READ_IO, F2FS_BLKSIZE); | |
662befda CY |
276 | } |
277 | out: | |
e9f5b8b8 | 278 | blk_finish_plug(&plug); |
662befda CY |
279 | return blkno - start; |
280 | } | |
281 | ||
4d57b86d | 282 | void f2fs_ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index) |
635aee1f CY |
283 | { |
284 | struct page *page; | |
285 | bool readahead = false; | |
286 | ||
287 | page = find_get_page(META_MAPPING(sbi), index); | |
4da7bf5a | 288 | if (!page || !PageUptodate(page)) |
635aee1f CY |
289 | readahead = true; |
290 | f2fs_put_page(page, 0); | |
291 | ||
292 | if (readahead) | |
4d57b86d | 293 | f2fs_ra_meta_pages(sbi, index, BIO_MAX_PAGES, META_POR, true); |
635aee1f CY |
294 | } |
295 | ||
b0af6d49 CY |
296 | static int __f2fs_write_meta_page(struct page *page, |
297 | struct writeback_control *wbc, | |
298 | enum iostat_type io_type) | |
127e670a | 299 | { |
4081363f | 300 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 301 | |
ecda0de3 CY |
302 | trace_f2fs_writepage(page, META); |
303 | ||
af697c0f JK |
304 | if (unlikely(f2fs_cp_error(sbi))) |
305 | goto redirty_out; | |
caf0047e | 306 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
cfb271d4 | 307 | goto redirty_out; |
857dc4e0 | 308 | if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0)) |
cfb271d4 | 309 | goto redirty_out; |
127e670a | 310 | |
4d57b86d | 311 | f2fs_do_write_meta_page(sbi, page, io_type); |
577e3495 | 312 | dec_page_count(sbi, F2FS_DIRTY_META); |
0c3a5797 CY |
313 | |
314 | if (wbc->for_reclaim) | |
bab475c5 | 315 | f2fs_submit_merged_write_cond(sbi, NULL, page, 0, META); |
0c3a5797 | 316 | |
577e3495 | 317 | unlock_page(page); |
857dc4e0 | 318 | |
0c3a5797 | 319 | if (unlikely(f2fs_cp_error(sbi))) |
b9109b0e | 320 | f2fs_submit_merged_write(sbi, META); |
0c3a5797 | 321 | |
577e3495 | 322 | return 0; |
cfb271d4 CY |
323 | |
324 | redirty_out: | |
76f60268 | 325 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 326 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
327 | } |
328 | ||
b0af6d49 CY |
329 | static int f2fs_write_meta_page(struct page *page, |
330 | struct writeback_control *wbc) | |
331 | { | |
332 | return __f2fs_write_meta_page(page, wbc, FS_META_IO); | |
333 | } | |
334 | ||
127e670a JK |
335 | static int f2fs_write_meta_pages(struct address_space *mapping, |
336 | struct writeback_control *wbc) | |
337 | { | |
4081363f | 338 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 339 | long diff, written; |
127e670a | 340 | |
0771fcc7 CY |
341 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
342 | goto skip_write; | |
343 | ||
5459aa97 | 344 | /* collect a number of dirty meta pages and write together */ |
812a9597 JK |
345 | if (wbc->sync_mode != WB_SYNC_ALL && |
346 | get_pages(sbi, F2FS_DIRTY_META) < | |
347 | nr_pages_to_skip(sbi, META)) | |
d3baf95d | 348 | goto skip_write; |
127e670a | 349 | |
a29d0e0b YH |
350 | /* if locked failed, cp will flush dirty pages instead */ |
351 | if (!mutex_trylock(&sbi->cp_mutex)) | |
352 | goto skip_write; | |
d31c7c3f | 353 | |
a29d0e0b | 354 | trace_f2fs_writepages(mapping->host, wbc, META); |
50c8cdb3 | 355 | diff = nr_pages_to_write(sbi, META, wbc); |
4d57b86d | 356 | written = f2fs_sync_meta_pages(sbi, META, wbc->nr_to_write, FS_META_IO); |
127e670a | 357 | mutex_unlock(&sbi->cp_mutex); |
50c8cdb3 | 358 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 359 | return 0; |
d3baf95d JK |
360 | |
361 | skip_write: | |
362 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
d31c7c3f | 363 | trace_f2fs_writepages(mapping->host, wbc, META); |
d3baf95d | 364 | return 0; |
127e670a JK |
365 | } |
366 | ||
4d57b86d | 367 | long f2fs_sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, |
b0af6d49 | 368 | long nr_to_write, enum iostat_type io_type) |
127e670a | 369 | { |
9df27d98 | 370 | struct address_space *mapping = META_MAPPING(sbi); |
028a63a6 | 371 | pgoff_t index = 0, prev = ULONG_MAX; |
127e670a JK |
372 | struct pagevec pvec; |
373 | long nwritten = 0; | |
028a63a6 | 374 | int nr_pages; |
127e670a JK |
375 | struct writeback_control wbc = { |
376 | .for_reclaim = 0, | |
377 | }; | |
e9f5b8b8 | 378 | struct blk_plug plug; |
127e670a | 379 | |
86679820 | 380 | pagevec_init(&pvec); |
127e670a | 381 | |
e9f5b8b8 CY |
382 | blk_start_plug(&plug); |
383 | ||
028a63a6 | 384 | while ((nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, |
67fd707f | 385 | PAGECACHE_TAG_DIRTY))) { |
028a63a6 | 386 | int i; |
127e670a JK |
387 | |
388 | for (i = 0; i < nr_pages; i++) { | |
389 | struct page *page = pvec.pages[i]; | |
203681f6 | 390 | |
80dd9c0e | 391 | if (prev == ULONG_MAX) |
6066d8cd JK |
392 | prev = page->index - 1; |
393 | if (nr_to_write != LONG_MAX && page->index != prev + 1) { | |
394 | pagevec_release(&pvec); | |
395 | goto stop; | |
396 | } | |
397 | ||
127e670a | 398 | lock_page(page); |
203681f6 JK |
399 | |
400 | if (unlikely(page->mapping != mapping)) { | |
401 | continue_unlock: | |
402 | unlock_page(page); | |
403 | continue; | |
404 | } | |
405 | if (!PageDirty(page)) { | |
406 | /* someone wrote it for us */ | |
407 | goto continue_unlock; | |
408 | } | |
409 | ||
bae0ee7a | 410 | f2fs_wait_on_page_writeback(page, META, true, true); |
fa3d2bdf | 411 | |
203681f6 JK |
412 | if (!clear_page_dirty_for_io(page)) |
413 | goto continue_unlock; | |
414 | ||
b0af6d49 | 415 | if (__f2fs_write_meta_page(page, &wbc, io_type)) { |
577e3495 JK |
416 | unlock_page(page); |
417 | break; | |
418 | } | |
cfb271d4 | 419 | nwritten++; |
6066d8cd | 420 | prev = page->index; |
cfb271d4 | 421 | if (unlikely(nwritten >= nr_to_write)) |
127e670a JK |
422 | break; |
423 | } | |
424 | pagevec_release(&pvec); | |
425 | cond_resched(); | |
426 | } | |
6066d8cd | 427 | stop: |
127e670a | 428 | if (nwritten) |
b9109b0e | 429 | f2fs_submit_merged_write(sbi, type); |
127e670a | 430 | |
e9f5b8b8 CY |
431 | blk_finish_plug(&plug); |
432 | ||
127e670a JK |
433 | return nwritten; |
434 | } | |
435 | ||
436 | static int f2fs_set_meta_page_dirty(struct page *page) | |
437 | { | |
26c6b887 JK |
438 | trace_f2fs_set_page_dirty(page, META); |
439 | ||
237c0790 JK |
440 | if (!PageUptodate(page)) |
441 | SetPageUptodate(page); | |
127e670a | 442 | if (!PageDirty(page)) { |
b87078ad | 443 | __set_page_dirty_nobuffers(page); |
4081363f | 444 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); |
240a5915 | 445 | f2fs_set_page_private(page, 0); |
9e4ded3f | 446 | f2fs_trace_pid(page); |
127e670a JK |
447 | return 1; |
448 | } | |
449 | return 0; | |
450 | } | |
451 | ||
452 | const struct address_space_operations f2fs_meta_aops = { | |
453 | .writepage = f2fs_write_meta_page, | |
454 | .writepages = f2fs_write_meta_pages, | |
455 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
487261f3 CY |
456 | .invalidatepage = f2fs_invalidate_page, |
457 | .releasepage = f2fs_release_page, | |
5b7a487c WG |
458 | #ifdef CONFIG_MIGRATION |
459 | .migratepage = f2fs_migrate_page, | |
460 | #endif | |
127e670a JK |
461 | }; |
462 | ||
39d787be CY |
463 | static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, |
464 | unsigned int devidx, int type) | |
953e6cc6 | 465 | { |
67298804 | 466 | struct inode_management *im = &sbi->im[type]; |
80c54505 JK |
467 | struct ino_entry *e, *tmp; |
468 | ||
469 | tmp = f2fs_kmem_cache_alloc(ino_entry_slab, GFP_NOFS); | |
19526d74 | 470 | |
80c54505 | 471 | radix_tree_preload(GFP_NOFS | __GFP_NOFAIL); |
769ec6e5 | 472 | |
67298804 | 473 | spin_lock(&im->ino_lock); |
67298804 | 474 | e = radix_tree_lookup(&im->ino_root, ino); |
39efac41 | 475 | if (!e) { |
80c54505 | 476 | e = tmp; |
19526d74 CY |
477 | if (unlikely(radix_tree_insert(&im->ino_root, ino, e))) |
478 | f2fs_bug_on(sbi, 1); | |
479 | ||
39efac41 JK |
480 | memset(e, 0, sizeof(struct ino_entry)); |
481 | e->ino = ino; | |
953e6cc6 | 482 | |
67298804 | 483 | list_add_tail(&e->list, &im->ino_list); |
8c402946 | 484 | if (type != ORPHAN_INO) |
67298804 | 485 | im->ino_num++; |
39efac41 | 486 | } |
39d787be CY |
487 | |
488 | if (type == FLUSH_INO) | |
489 | f2fs_set_bit(devidx, (char *)&e->dirty_device); | |
490 | ||
67298804 | 491 | spin_unlock(&im->ino_lock); |
769ec6e5 | 492 | radix_tree_preload_end(); |
80c54505 JK |
493 | |
494 | if (e != tmp) | |
495 | kmem_cache_free(ino_entry_slab, tmp); | |
953e6cc6 JK |
496 | } |
497 | ||
6451e041 | 498 | static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 499 | { |
67298804 | 500 | struct inode_management *im = &sbi->im[type]; |
6451e041 | 501 | struct ino_entry *e; |
953e6cc6 | 502 | |
67298804 CY |
503 | spin_lock(&im->ino_lock); |
504 | e = radix_tree_lookup(&im->ino_root, ino); | |
39efac41 JK |
505 | if (e) { |
506 | list_del(&e->list); | |
67298804 CY |
507 | radix_tree_delete(&im->ino_root, ino); |
508 | im->ino_num--; | |
509 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
510 | kmem_cache_free(ino_entry_slab, e); |
511 | return; | |
953e6cc6 | 512 | } |
67298804 | 513 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
514 | } |
515 | ||
4d57b86d | 516 | void f2fs_add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
517 | { |
518 | /* add new dirty ino entry into list */ | |
39d787be | 519 | __add_ino_entry(sbi, ino, 0, type); |
fff04f90 JK |
520 | } |
521 | ||
4d57b86d | 522 | void f2fs_remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
523 | { |
524 | /* remove dirty ino entry from list */ | |
525 | __remove_ino_entry(sbi, ino, type); | |
526 | } | |
527 | ||
1f07cc58 | 528 | /* mode should be APPEND_INO, UPDATE_INO or TRANS_DIR_INO */ |
4d57b86d | 529 | bool f2fs_exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode) |
fff04f90 | 530 | { |
67298804 | 531 | struct inode_management *im = &sbi->im[mode]; |
fff04f90 | 532 | struct ino_entry *e; |
67298804 CY |
533 | |
534 | spin_lock(&im->ino_lock); | |
535 | e = radix_tree_lookup(&im->ino_root, ino); | |
536 | spin_unlock(&im->ino_lock); | |
fff04f90 JK |
537 | return e ? true : false; |
538 | } | |
539 | ||
4d57b86d | 540 | void f2fs_release_ino_entry(struct f2fs_sb_info *sbi, bool all) |
fff04f90 JK |
541 | { |
542 | struct ino_entry *e, *tmp; | |
543 | int i; | |
544 | ||
39d787be | 545 | for (i = all ? ORPHAN_INO : APPEND_INO; i < MAX_INO_ENTRY; i++) { |
67298804 CY |
546 | struct inode_management *im = &sbi->im[i]; |
547 | ||
548 | spin_lock(&im->ino_lock); | |
549 | list_for_each_entry_safe(e, tmp, &im->ino_list, list) { | |
fff04f90 | 550 | list_del(&e->list); |
67298804 | 551 | radix_tree_delete(&im->ino_root, e->ino); |
fff04f90 | 552 | kmem_cache_free(ino_entry_slab, e); |
67298804 | 553 | im->ino_num--; |
fff04f90 | 554 | } |
67298804 | 555 | spin_unlock(&im->ino_lock); |
fff04f90 JK |
556 | } |
557 | } | |
558 | ||
4d57b86d | 559 | void f2fs_set_dirty_device(struct f2fs_sb_info *sbi, nid_t ino, |
39d787be CY |
560 | unsigned int devidx, int type) |
561 | { | |
562 | __add_ino_entry(sbi, ino, devidx, type); | |
563 | } | |
564 | ||
4d57b86d | 565 | bool f2fs_is_dirty_device(struct f2fs_sb_info *sbi, nid_t ino, |
39d787be CY |
566 | unsigned int devidx, int type) |
567 | { | |
568 | struct inode_management *im = &sbi->im[type]; | |
569 | struct ino_entry *e; | |
570 | bool is_dirty = false; | |
571 | ||
572 | spin_lock(&im->ino_lock); | |
573 | e = radix_tree_lookup(&im->ino_root, ino); | |
574 | if (e && f2fs_test_bit(devidx, (char *)&e->dirty_device)) | |
575 | is_dirty = true; | |
576 | spin_unlock(&im->ino_lock); | |
577 | return is_dirty; | |
578 | } | |
579 | ||
4d57b86d | 580 | int f2fs_acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 581 | { |
67298804 | 582 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a JK |
583 | int err = 0; |
584 | ||
67298804 | 585 | spin_lock(&im->ino_lock); |
cb78942b | 586 | |
1ecc0c5c | 587 | if (time_to_inject(sbi, FAULT_ORPHAN)) { |
cb78942b | 588 | spin_unlock(&im->ino_lock); |
c45d6002 | 589 | f2fs_show_injection_info(sbi, FAULT_ORPHAN); |
cb78942b JK |
590 | return -ENOSPC; |
591 | } | |
7fa750a1 | 592 | |
67298804 | 593 | if (unlikely(im->ino_num >= sbi->max_orphans)) |
127e670a | 594 | err = -ENOSPC; |
cbd56e7d | 595 | else |
67298804 CY |
596 | im->ino_num++; |
597 | spin_unlock(&im->ino_lock); | |
0d47c1ad | 598 | |
127e670a JK |
599 | return err; |
600 | } | |
601 | ||
4d57b86d | 602 | void f2fs_release_orphan_inode(struct f2fs_sb_info *sbi) |
cbd56e7d | 603 | { |
67298804 CY |
604 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
605 | ||
606 | spin_lock(&im->ino_lock); | |
607 | f2fs_bug_on(sbi, im->ino_num == 0); | |
608 | im->ino_num--; | |
609 | spin_unlock(&im->ino_lock); | |
cbd56e7d JK |
610 | } |
611 | ||
4d57b86d | 612 | void f2fs_add_orphan_inode(struct inode *inode) |
127e670a | 613 | { |
39efac41 | 614 | /* add new orphan ino entry into list */ |
39d787be | 615 | __add_ino_entry(F2FS_I_SB(inode), inode->i_ino, 0, ORPHAN_INO); |
4d57b86d | 616 | f2fs_update_inode_page(inode); |
127e670a JK |
617 | } |
618 | ||
4d57b86d | 619 | void f2fs_remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 620 | { |
953e6cc6 | 621 | /* remove orphan entry from orphan list */ |
6451e041 | 622 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
623 | } |
624 | ||
8c14bfad | 625 | static int recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 626 | { |
8c14bfad | 627 | struct inode *inode; |
5905f9af | 628 | struct node_info ni; |
76a45e3c | 629 | int err; |
8c14bfad | 630 | |
5905f9af | 631 | inode = f2fs_iget_retry(sbi->sb, ino); |
8c14bfad CY |
632 | if (IS_ERR(inode)) { |
633 | /* | |
634 | * there should be a bug that we can't find the entry | |
635 | * to orphan inode. | |
636 | */ | |
637 | f2fs_bug_on(sbi, PTR_ERR(inode) == -ENOENT); | |
638 | return PTR_ERR(inode); | |
639 | } | |
640 | ||
0f9ec2a8 | 641 | err = dquot_initialize(inode); |
a515d12f SY |
642 | if (err) { |
643 | iput(inode); | |
0f9ec2a8 | 644 | goto err_out; |
a515d12f | 645 | } |
0f9ec2a8 | 646 | |
127e670a JK |
647 | clear_nlink(inode); |
648 | ||
649 | /* truncate all the data during iput */ | |
650 | iput(inode); | |
5905f9af | 651 | |
7735730d CY |
652 | err = f2fs_get_node_info(sbi, ino, &ni); |
653 | if (err) | |
654 | goto err_out; | |
5905f9af JK |
655 | |
656 | /* ENOMEM was fully retried in f2fs_evict_inode. */ | |
657 | if (ni.blk_addr != NULL_ADDR) { | |
0f9ec2a8 JK |
658 | err = -EIO; |
659 | goto err_out; | |
5905f9af | 660 | } |
8c14bfad | 661 | return 0; |
0f9ec2a8 JK |
662 | |
663 | err_out: | |
664 | set_sbi_flag(sbi, SBI_NEED_FSCK); | |
dcbb4c10 JP |
665 | f2fs_warn(sbi, "%s: orphan failed (ino=%x), run fsck to fix.", |
666 | __func__, ino); | |
0f9ec2a8 | 667 | return err; |
127e670a JK |
668 | } |
669 | ||
4d57b86d | 670 | int f2fs_recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a | 671 | { |
3c642985 | 672 | block_t start_blk, orphan_blocks, i, j; |
4b2414d0 CY |
673 | unsigned int s_flags = sbi->sb->s_flags; |
674 | int err = 0; | |
ea676733 JK |
675 | #ifdef CONFIG_QUOTA |
676 | int quota_enabled; | |
677 | #endif | |
127e670a | 678 | |
aaec2b1d | 679 | if (!is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG)) |
8c14bfad | 680 | return 0; |
127e670a | 681 | |
b61af314 | 682 | if (bdev_read_only(sbi->sb->s_bdev)) { |
dcbb4c10 | 683 | f2fs_info(sbi, "write access unavailable, skipping orphan cleanup"); |
b61af314 CY |
684 | return 0; |
685 | } | |
686 | ||
1751e8a6 | 687 | if (s_flags & SB_RDONLY) { |
dcbb4c10 | 688 | f2fs_info(sbi, "orphan cleanup on readonly fs"); |
1751e8a6 | 689 | sbi->sb->s_flags &= ~SB_RDONLY; |
4b2414d0 CY |
690 | } |
691 | ||
692 | #ifdef CONFIG_QUOTA | |
693 | /* Needed for iput() to work correctly and not trash data */ | |
1751e8a6 | 694 | sbi->sb->s_flags |= SB_ACTIVE; |
ea676733 | 695 | |
76cf05d7 SY |
696 | /* |
697 | * Turn on quotas which were not enabled for read-only mounts if | |
698 | * filesystem has quota feature, so that they are updated correctly. | |
699 | */ | |
1751e8a6 | 700 | quota_enabled = f2fs_enable_quota_files(sbi, s_flags & SB_RDONLY); |
4b2414d0 CY |
701 | #endif |
702 | ||
55141486 | 703 | start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi); |
3c642985 | 704 | orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi); |
127e670a | 705 | |
4d57b86d | 706 | f2fs_ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP, true); |
662befda | 707 | |
3c642985 | 708 | for (i = 0; i < orphan_blocks; i++) { |
7735730d | 709 | struct page *page; |
127e670a JK |
710 | struct f2fs_orphan_block *orphan_blk; |
711 | ||
7735730d CY |
712 | page = f2fs_get_meta_page(sbi, start_blk + i); |
713 | if (IS_ERR(page)) { | |
714 | err = PTR_ERR(page); | |
715 | goto out; | |
716 | } | |
717 | ||
127e670a JK |
718 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); |
719 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
720 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
8c14bfad CY |
721 | err = recover_orphan_inode(sbi, ino); |
722 | if (err) { | |
723 | f2fs_put_page(page, 1); | |
4b2414d0 | 724 | goto out; |
8c14bfad | 725 | } |
127e670a JK |
726 | } |
727 | f2fs_put_page(page, 1); | |
728 | } | |
729 | /* clear Orphan Flag */ | |
aaec2b1d | 730 | clear_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG); |
4b2414d0 | 731 | out: |
1378752b CY |
732 | set_sbi_flag(sbi, SBI_IS_RECOVERED); |
733 | ||
4b2414d0 CY |
734 | #ifdef CONFIG_QUOTA |
735 | /* Turn quotas off */ | |
ea676733 JK |
736 | if (quota_enabled) |
737 | f2fs_quota_off_umount(sbi->sb); | |
4b2414d0 | 738 | #endif |
1751e8a6 | 739 | sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ |
4b2414d0 CY |
740 | |
741 | return err; | |
127e670a JK |
742 | } |
743 | ||
744 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
745 | { | |
502c6e0b | 746 | struct list_head *head; |
127e670a | 747 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 748 | unsigned int nentries = 0; |
bd936f84 | 749 | unsigned short index = 1; |
8c402946 | 750 | unsigned short orphan_blocks; |
4531929e | 751 | struct page *page = NULL; |
6451e041 | 752 | struct ino_entry *orphan = NULL; |
67298804 | 753 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 754 | |
67298804 | 755 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 756 | |
d6c67a4f JK |
757 | /* |
758 | * we don't need to do spin_lock(&im->ino_lock) here, since all the | |
759 | * orphan inode operations are covered under f2fs_lock_op(). | |
760 | * And, spin_lock should be avoided due to page operations below. | |
761 | */ | |
67298804 | 762 | head = &im->ino_list; |
127e670a JK |
763 | |
764 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
765 | list_for_each_entry(orphan, head, list) { |
766 | if (!page) { | |
4d57b86d | 767 | page = f2fs_grab_meta_page(sbi, start_blk++); |
502c6e0b GZ |
768 | orphan_blk = |
769 | (struct f2fs_orphan_block *)page_address(page); | |
770 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
771 | } | |
127e670a | 772 | |
36795567 | 773 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 774 | |
36795567 | 775 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
776 | /* |
777 | * an orphan block is full of 1020 entries, | |
778 | * then we need to flush current orphan blocks | |
779 | * and bring another one in memory | |
780 | */ | |
781 | orphan_blk->blk_addr = cpu_to_le16(index); | |
782 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
783 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
784 | set_page_dirty(page); | |
785 | f2fs_put_page(page, 1); | |
786 | index++; | |
127e670a JK |
787 | nentries = 0; |
788 | page = NULL; | |
789 | } | |
502c6e0b | 790 | } |
127e670a | 791 | |
502c6e0b GZ |
792 | if (page) { |
793 | orphan_blk->blk_addr = cpu_to_le16(index); | |
794 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
795 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
796 | set_page_dirty(page); | |
797 | f2fs_put_page(page, 1); | |
127e670a | 798 | } |
127e670a JK |
799 | } |
800 | ||
d7eb8f1c CY |
801 | static __u32 f2fs_checkpoint_chksum(struct f2fs_sb_info *sbi, |
802 | struct f2fs_checkpoint *ckpt) | |
803 | { | |
804 | unsigned int chksum_ofs = le32_to_cpu(ckpt->checksum_offset); | |
805 | __u32 chksum; | |
806 | ||
807 | chksum = f2fs_crc32(sbi, ckpt, chksum_ofs); | |
808 | if (chksum_ofs < CP_CHKSUM_OFFSET) { | |
809 | chksum_ofs += sizeof(chksum); | |
810 | chksum = f2fs_chksum(sbi, chksum, (__u8 *)ckpt + chksum_ofs, | |
811 | F2FS_BLKSIZE - chksum_ofs); | |
812 | } | |
813 | return chksum; | |
814 | } | |
815 | ||
fc0065ad TY |
816 | static int get_checkpoint_version(struct f2fs_sb_info *sbi, block_t cp_addr, |
817 | struct f2fs_checkpoint **cp_block, struct page **cp_page, | |
818 | unsigned long long *version) | |
127e670a | 819 | { |
fc0065ad | 820 | size_t crc_offset = 0; |
d7eb8f1c | 821 | __u32 crc; |
127e670a | 822 | |
4d57b86d | 823 | *cp_page = f2fs_get_meta_page(sbi, cp_addr); |
7735730d CY |
824 | if (IS_ERR(*cp_page)) |
825 | return PTR_ERR(*cp_page); | |
826 | ||
fc0065ad | 827 | *cp_block = (struct f2fs_checkpoint *)page_address(*cp_page); |
127e670a | 828 | |
fc0065ad | 829 | crc_offset = le32_to_cpu((*cp_block)->checksum_offset); |
d7eb8f1c CY |
830 | if (crc_offset < CP_MIN_CHKSUM_OFFSET || |
831 | crc_offset > CP_CHKSUM_OFFSET) { | |
d3f07c04 | 832 | f2fs_put_page(*cp_page, 1); |
dcbb4c10 | 833 | f2fs_warn(sbi, "invalid crc_offset: %zu", crc_offset); |
fc0065ad TY |
834 | return -EINVAL; |
835 | } | |
127e670a | 836 | |
d7eb8f1c CY |
837 | crc = f2fs_checkpoint_chksum(sbi, *cp_block); |
838 | if (crc != cur_cp_crc(*cp_block)) { | |
d3f07c04 | 839 | f2fs_put_page(*cp_page, 1); |
dcbb4c10 | 840 | f2fs_warn(sbi, "invalid crc value"); |
fc0065ad TY |
841 | return -EINVAL; |
842 | } | |
127e670a | 843 | |
fc0065ad TY |
844 | *version = cur_cp_version(*cp_block); |
845 | return 0; | |
846 | } | |
127e670a | 847 | |
fc0065ad TY |
848 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, |
849 | block_t cp_addr, unsigned long long *version) | |
850 | { | |
851 | struct page *cp_page_1 = NULL, *cp_page_2 = NULL; | |
852 | struct f2fs_checkpoint *cp_block = NULL; | |
853 | unsigned long long cur_version = 0, pre_version = 0; | |
854 | int err; | |
127e670a | 855 | |
fc0065ad TY |
856 | err = get_checkpoint_version(sbi, cp_addr, &cp_block, |
857 | &cp_page_1, version); | |
858 | if (err) | |
d3f07c04 | 859 | return NULL; |
c9b60788 CY |
860 | |
861 | if (le32_to_cpu(cp_block->cp_pack_total_block_count) > | |
862 | sbi->blocks_per_seg) { | |
dcbb4c10 JP |
863 | f2fs_warn(sbi, "invalid cp_pack_total_block_count:%u", |
864 | le32_to_cpu(cp_block->cp_pack_total_block_count)); | |
d3f07c04 | 865 | goto invalid_cp; |
c9b60788 | 866 | } |
fc0065ad | 867 | pre_version = *version; |
127e670a | 868 | |
fc0065ad TY |
869 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
870 | err = get_checkpoint_version(sbi, cp_addr, &cp_block, | |
871 | &cp_page_2, version); | |
872 | if (err) | |
d3f07c04 | 873 | goto invalid_cp; |
fc0065ad | 874 | cur_version = *version; |
127e670a JK |
875 | |
876 | if (cur_version == pre_version) { | |
877 | *version = cur_version; | |
878 | f2fs_put_page(cp_page_2, 1); | |
879 | return cp_page_1; | |
880 | } | |
127e670a | 881 | f2fs_put_page(cp_page_2, 1); |
d3f07c04 | 882 | invalid_cp: |
127e670a JK |
883 | f2fs_put_page(cp_page_1, 1); |
884 | return NULL; | |
885 | } | |
886 | ||
4d57b86d | 887 | int f2fs_get_valid_checkpoint(struct f2fs_sb_info *sbi) |
127e670a JK |
888 | { |
889 | struct f2fs_checkpoint *cp_block; | |
890 | struct f2fs_super_block *fsb = sbi->raw_super; | |
891 | struct page *cp1, *cp2, *cur_page; | |
892 | unsigned long blk_size = sbi->blocksize; | |
893 | unsigned long long cp1_version = 0, cp2_version = 0; | |
894 | unsigned long long cp_start_blk_no; | |
55141486 | 895 | unsigned int cp_blks = 1 + __cp_payload(sbi); |
1dbe4152 CL |
896 | block_t cp_blk_no; |
897 | int i; | |
10f966bb | 898 | int err; |
127e670a | 899 | |
0b6d4ca0 EB |
900 | sbi->ckpt = f2fs_kvzalloc(sbi, array_size(blk_size, cp_blks), |
901 | GFP_KERNEL); | |
127e670a JK |
902 | if (!sbi->ckpt) |
903 | return -ENOMEM; | |
904 | /* | |
905 | * Finding out valid cp block involves read both | |
7a88ddb5 | 906 | * sets( cp pack 1 and cp pack 2) |
127e670a JK |
907 | */ |
908 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
909 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
910 | ||
911 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
912 | cp_start_blk_no += ((unsigned long long)1) << |
913 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
914 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
915 | ||
916 | if (cp1 && cp2) { | |
917 | if (ver_after(cp2_version, cp1_version)) | |
918 | cur_page = cp2; | |
919 | else | |
920 | cur_page = cp1; | |
921 | } else if (cp1) { | |
922 | cur_page = cp1; | |
923 | } else if (cp2) { | |
924 | cur_page = cp2; | |
925 | } else { | |
10f966bb | 926 | err = -EFSCORRUPTED; |
127e670a JK |
927 | goto fail_no_cp; |
928 | } | |
929 | ||
930 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
931 | memcpy(sbi->ckpt, cp_block, blk_size); | |
932 | ||
8508e44a JK |
933 | if (cur_page == cp1) |
934 | sbi->cur_cp_pack = 1; | |
935 | else | |
936 | sbi->cur_cp_pack = 2; | |
984ec63c | 937 | |
e494c2f9 | 938 | /* Sanity checking of checkpoint */ |
10f966bb CY |
939 | if (f2fs_sanity_check_ckpt(sbi)) { |
940 | err = -EFSCORRUPTED; | |
e494c2f9 | 941 | goto free_fail_no_cp; |
10f966bb | 942 | } |
e494c2f9 | 943 | |
1dbe4152 CL |
944 | if (cp_blks <= 1) |
945 | goto done; | |
946 | ||
947 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
948 | if (cur_page == cp2) | |
949 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
950 | ||
951 | for (i = 1; i < cp_blks; i++) { | |
952 | void *sit_bitmap_ptr; | |
953 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
954 | ||
4d57b86d | 955 | cur_page = f2fs_get_meta_page(sbi, cp_blk_no + i); |
10f966bb CY |
956 | if (IS_ERR(cur_page)) { |
957 | err = PTR_ERR(cur_page); | |
7735730d | 958 | goto free_fail_no_cp; |
10f966bb | 959 | } |
1dbe4152 CL |
960 | sit_bitmap_ptr = page_address(cur_page); |
961 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
962 | f2fs_put_page(cur_page, 1); | |
963 | } | |
964 | done: | |
127e670a JK |
965 | f2fs_put_page(cp1, 1); |
966 | f2fs_put_page(cp2, 1); | |
967 | return 0; | |
968 | ||
a2125ff7 JK |
969 | free_fail_no_cp: |
970 | f2fs_put_page(cp1, 1); | |
971 | f2fs_put_page(cp2, 1); | |
127e670a | 972 | fail_no_cp: |
5222595d | 973 | kvfree(sbi->ckpt); |
10f966bb | 974 | return err; |
127e670a JK |
975 | } |
976 | ||
c227f912 | 977 | static void __add_dirty_inode(struct inode *inode, enum inode_type type) |
127e670a | 978 | { |
4081363f | 979 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 980 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
127e670a | 981 | |
91942321 | 982 | if (is_inode_flag_set(inode, flag)) |
2710fd7e | 983 | return; |
2d7b822a | 984 | |
91942321 | 985 | set_inode_flag(inode, flag); |
99f4b917 CY |
986 | if (!f2fs_is_volatile_file(inode)) |
987 | list_add_tail(&F2FS_I(inode)->dirty_list, | |
988 | &sbi->inode_list[type]); | |
33fbd510 | 989 | stat_inc_dirty_inode(sbi, type); |
5deb8267 JK |
990 | } |
991 | ||
c227f912 | 992 | static void __remove_dirty_inode(struct inode *inode, enum inode_type type) |
6ad7609a | 993 | { |
c227f912 | 994 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
6ad7609a | 995 | |
91942321 | 996 | if (get_dirty_pages(inode) || !is_inode_flag_set(inode, flag)) |
6ad7609a CY |
997 | return; |
998 | ||
91942321 JK |
999 | list_del_init(&F2FS_I(inode)->dirty_list); |
1000 | clear_inode_flag(inode, flag); | |
33fbd510 | 1001 | stat_dec_dirty_inode(F2FS_I_SB(inode), type); |
6ad7609a CY |
1002 | } |
1003 | ||
4d57b86d | 1004 | void f2fs_update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 1005 | { |
4081363f | 1006 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 1007 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
5deb8267 | 1008 | |
5ac9f36f CY |
1009 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
1010 | !S_ISLNK(inode->i_mode)) | |
127e670a | 1011 | return; |
7bd59381 | 1012 | |
1c4bf763 JK |
1013 | spin_lock(&sbi->inode_lock[type]); |
1014 | if (type != FILE_INODE || test_opt(sbi, DATA_FLUSH)) | |
10aa97c3 | 1015 | __add_dirty_inode(inode, type); |
b951a4ec | 1016 | inode_inc_dirty_pages(inode); |
1c4bf763 JK |
1017 | spin_unlock(&sbi->inode_lock[type]); |
1018 | ||
240a5915 | 1019 | f2fs_set_page_private(page, 0); |
9e4ded3f | 1020 | f2fs_trace_pid(page); |
5deb8267 JK |
1021 | } |
1022 | ||
4d57b86d | 1023 | void f2fs_remove_dirty_inode(struct inode *inode) |
127e670a | 1024 | { |
4081363f | 1025 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 1026 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
127e670a | 1027 | |
c227f912 CY |
1028 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
1029 | !S_ISLNK(inode->i_mode)) | |
127e670a JK |
1030 | return; |
1031 | ||
10aa97c3 JK |
1032 | if (type == FILE_INODE && !test_opt(sbi, DATA_FLUSH)) |
1033 | return; | |
1034 | ||
c227f912 CY |
1035 | spin_lock(&sbi->inode_lock[type]); |
1036 | __remove_dirty_inode(inode, type); | |
1037 | spin_unlock(&sbi->inode_lock[type]); | |
74d0b917 JK |
1038 | } |
1039 | ||
4d57b86d | 1040 | int f2fs_sync_dirty_inodes(struct f2fs_sb_info *sbi, enum inode_type type) |
127e670a | 1041 | { |
ce3b7d80 | 1042 | struct list_head *head; |
127e670a | 1043 | struct inode *inode; |
2710fd7e | 1044 | struct f2fs_inode_info *fi; |
4cf18537 | 1045 | bool is_dir = (type == DIR_INODE); |
4db08d01 | 1046 | unsigned long ino = 0; |
4cf18537 CY |
1047 | |
1048 | trace_f2fs_sync_dirty_inodes_enter(sbi->sb, is_dir, | |
1049 | get_pages(sbi, is_dir ? | |
1050 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
127e670a | 1051 | retry: |
9b664822 ZQ |
1052 | if (unlikely(f2fs_cp_error(sbi))) { |
1053 | trace_f2fs_sync_dirty_inodes_exit(sbi->sb, is_dir, | |
1054 | get_pages(sbi, is_dir ? | |
1055 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
6d5a1495 | 1056 | return -EIO; |
9b664822 | 1057 | } |
af41d3ee | 1058 | |
c227f912 | 1059 | spin_lock(&sbi->inode_lock[type]); |
ce3b7d80 | 1060 | |
c227f912 | 1061 | head = &sbi->inode_list[type]; |
127e670a | 1062 | if (list_empty(head)) { |
c227f912 | 1063 | spin_unlock(&sbi->inode_lock[type]); |
4cf18537 CY |
1064 | trace_f2fs_sync_dirty_inodes_exit(sbi->sb, is_dir, |
1065 | get_pages(sbi, is_dir ? | |
1066 | F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA)); | |
6d5a1495 | 1067 | return 0; |
127e670a | 1068 | } |
939afa94 | 1069 | fi = list_first_entry(head, struct f2fs_inode_info, dirty_list); |
2710fd7e | 1070 | inode = igrab(&fi->vfs_inode); |
c227f912 | 1071 | spin_unlock(&sbi->inode_lock[type]); |
127e670a | 1072 | if (inode) { |
4db08d01 JK |
1073 | unsigned long cur_ino = inode->i_ino; |
1074 | ||
186857c5 | 1075 | F2FS_I(inode)->cp_task = current; |
b0af6d49 | 1076 | |
87d6f890 | 1077 | filemap_fdatawrite(inode->i_mapping); |
b0af6d49 | 1078 | |
186857c5 | 1079 | F2FS_I(inode)->cp_task = NULL; |
b0af6d49 | 1080 | |
127e670a | 1081 | iput(inode); |
4db08d01 | 1082 | /* We need to give cpu to another writers. */ |
2a63531a | 1083 | if (ino == cur_ino) |
4db08d01 | 1084 | cond_resched(); |
2a63531a | 1085 | else |
4db08d01 | 1086 | ino = cur_ino; |
127e670a JK |
1087 | } else { |
1088 | /* | |
1089 | * We should submit bio, since it exists several | |
1090 | * wribacking dentry pages in the freeing inode. | |
1091 | */ | |
b9109b0e | 1092 | f2fs_submit_merged_write(sbi, DATA); |
7ecebe5e | 1093 | cond_resched(); |
127e670a JK |
1094 | } |
1095 | goto retry; | |
1096 | } | |
1097 | ||
0f18b462 JK |
1098 | int f2fs_sync_inode_meta(struct f2fs_sb_info *sbi) |
1099 | { | |
1100 | struct list_head *head = &sbi->inode_list[DIRTY_META]; | |
1101 | struct inode *inode; | |
1102 | struct f2fs_inode_info *fi; | |
1103 | s64 total = get_pages(sbi, F2FS_DIRTY_IMETA); | |
1104 | ||
1105 | while (total--) { | |
1106 | if (unlikely(f2fs_cp_error(sbi))) | |
1107 | return -EIO; | |
1108 | ||
1109 | spin_lock(&sbi->inode_lock[DIRTY_META]); | |
1110 | if (list_empty(head)) { | |
1111 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
1112 | return 0; | |
1113 | } | |
939afa94 | 1114 | fi = list_first_entry(head, struct f2fs_inode_info, |
0f18b462 JK |
1115 | gdirty_list); |
1116 | inode = igrab(&fi->vfs_inode); | |
1117 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
1118 | if (inode) { | |
18340edc JK |
1119 | sync_inode_metadata(inode, 0); |
1120 | ||
1121 | /* it's on eviction */ | |
1122 | if (is_inode_flag_set(inode, FI_DIRTY_INODE)) | |
4d57b86d | 1123 | f2fs_update_inode_page(inode); |
0f18b462 JK |
1124 | iput(inode); |
1125 | } | |
dee668c1 | 1126 | } |
0f18b462 JK |
1127 | return 0; |
1128 | } | |
1129 | ||
59c9081b YH |
1130 | static void __prepare_cp_block(struct f2fs_sb_info *sbi) |
1131 | { | |
1132 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1133 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1134 | nid_t last_nid = nm_i->next_scan_nid; | |
1135 | ||
1136 | next_free_nid(sbi, &last_nid); | |
1137 | ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); | |
1138 | ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); | |
1139 | ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); | |
1140 | ckpt->next_free_nid = cpu_to_le32(last_nid); | |
1141 | } | |
1142 | ||
af033b2a CY |
1143 | static bool __need_flush_quota(struct f2fs_sb_info *sbi) |
1144 | { | |
db6ec53b JK |
1145 | bool ret = false; |
1146 | ||
af033b2a CY |
1147 | if (!is_journalled_quota(sbi)) |
1148 | return false; | |
db6ec53b JK |
1149 | |
1150 | down_write(&sbi->quota_sem); | |
1151 | if (is_sbi_flag_set(sbi, SBI_QUOTA_SKIP_FLUSH)) { | |
1152 | ret = false; | |
1153 | } else if (is_sbi_flag_set(sbi, SBI_QUOTA_NEED_REPAIR)) { | |
1154 | ret = false; | |
1155 | } else if (is_sbi_flag_set(sbi, SBI_QUOTA_NEED_FLUSH)) { | |
1156 | clear_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); | |
1157 | ret = true; | |
1158 | } else if (get_pages(sbi, F2FS_DIRTY_QDATA)) { | |
1159 | ret = true; | |
1160 | } | |
1161 | up_write(&sbi->quota_sem); | |
1162 | return ret; | |
af033b2a CY |
1163 | } |
1164 | ||
0a8165d7 | 1165 | /* |
127e670a JK |
1166 | * Freeze all the FS-operations for checkpoint. |
1167 | */ | |
cf779cab | 1168 | static int block_operations(struct f2fs_sb_info *sbi) |
127e670a | 1169 | { |
127e670a JK |
1170 | struct writeback_control wbc = { |
1171 | .sync_mode = WB_SYNC_ALL, | |
1172 | .nr_to_write = LONG_MAX, | |
1173 | .for_reclaim = 0, | |
1174 | }; | |
af033b2a | 1175 | int err = 0, cnt = 0; |
c718379b | 1176 | |
34c061ad SL |
1177 | /* |
1178 | * Let's flush inline_data in dirty node pages. | |
1179 | */ | |
1180 | f2fs_flush_inline_data(sbi); | |
1181 | ||
af033b2a | 1182 | retry_flush_quotas: |
db6ec53b | 1183 | f2fs_lock_all(sbi); |
af033b2a CY |
1184 | if (__need_flush_quota(sbi)) { |
1185 | int locked; | |
1186 | ||
1187 | if (++cnt > DEFAULT_RETRY_QUOTA_FLUSH_COUNT) { | |
1188 | set_sbi_flag(sbi, SBI_QUOTA_SKIP_FLUSH); | |
db6ec53b | 1189 | set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); |
af033b2a CY |
1190 | goto retry_flush_dents; |
1191 | } | |
db6ec53b | 1192 | f2fs_unlock_all(sbi); |
af033b2a CY |
1193 | |
1194 | /* only failed during mount/umount/freeze/quotactl */ | |
1195 | locked = down_read_trylock(&sbi->sb->s_umount); | |
1196 | f2fs_quota_sync(sbi->sb, -1); | |
1197 | if (locked) | |
1198 | up_read(&sbi->sb->s_umount); | |
af033b2a CY |
1199 | cond_resched(); |
1200 | goto retry_flush_quotas; | |
1201 | } | |
1202 | ||
1203 | retry_flush_dents: | |
127e670a | 1204 | /* write all the dirty dentry pages */ |
127e670a | 1205 | if (get_pages(sbi, F2FS_DIRTY_DENTS)) { |
e479556b | 1206 | f2fs_unlock_all(sbi); |
4d57b86d | 1207 | err = f2fs_sync_dirty_inodes(sbi, DIR_INODE); |
6d5a1495 | 1208 | if (err) |
1f5f11a3 | 1209 | return err; |
30973883 | 1210 | cond_resched(); |
af033b2a | 1211 | goto retry_flush_quotas; |
127e670a JK |
1212 | } |
1213 | ||
59c9081b YH |
1214 | /* |
1215 | * POR: we should ensure that there are no dirty node pages | |
1216 | * until finishing nat/sit flush. inode->i_blocks can be updated. | |
1217 | */ | |
1218 | down_write(&sbi->node_change); | |
1219 | ||
0f18b462 | 1220 | if (get_pages(sbi, F2FS_DIRTY_IMETA)) { |
59c9081b | 1221 | up_write(&sbi->node_change); |
0f18b462 JK |
1222 | f2fs_unlock_all(sbi); |
1223 | err = f2fs_sync_inode_meta(sbi); | |
1224 | if (err) | |
1f5f11a3 | 1225 | return err; |
30973883 | 1226 | cond_resched(); |
af033b2a | 1227 | goto retry_flush_quotas; |
0f18b462 JK |
1228 | } |
1229 | ||
39936837 | 1230 | retry_flush_nodes: |
b3582c68 | 1231 | down_write(&sbi->node_write); |
127e670a JK |
1232 | |
1233 | if (get_pages(sbi, F2FS_DIRTY_NODES)) { | |
b3582c68 | 1234 | up_write(&sbi->node_write); |
c29fd0c0 | 1235 | atomic_inc(&sbi->wb_sync_req[NODE]); |
4d57b86d | 1236 | err = f2fs_sync_node_pages(sbi, &wbc, false, FS_CP_NODE_IO); |
c29fd0c0 | 1237 | atomic_dec(&sbi->wb_sync_req[NODE]); |
6d5a1495 | 1238 | if (err) { |
59c9081b | 1239 | up_write(&sbi->node_change); |
cf779cab | 1240 | f2fs_unlock_all(sbi); |
1f5f11a3 | 1241 | return err; |
cf779cab | 1242 | } |
30973883 | 1243 | cond_resched(); |
39936837 | 1244 | goto retry_flush_nodes; |
127e670a | 1245 | } |
59c9081b YH |
1246 | |
1247 | /* | |
1248 | * sbi->node_change is used only for AIO write_begin path which produces | |
1249 | * dirty node blocks and some checkpoint values by block allocation. | |
1250 | */ | |
1251 | __prepare_cp_block(sbi); | |
1252 | up_write(&sbi->node_change); | |
cf779cab | 1253 | return err; |
127e670a JK |
1254 | } |
1255 | ||
1256 | static void unblock_operations(struct f2fs_sb_info *sbi) | |
1257 | { | |
b3582c68 | 1258 | up_write(&sbi->node_write); |
e479556b | 1259 | f2fs_unlock_all(sbi); |
127e670a JK |
1260 | } |
1261 | ||
bf22c3cc | 1262 | void f2fs_wait_on_all_pages(struct f2fs_sb_info *sbi, int type) |
fb51b5ef CL |
1263 | { |
1264 | DEFINE_WAIT(wait); | |
1265 | ||
1266 | for (;;) { | |
bf22c3cc | 1267 | if (!get_pages(sbi, type)) |
fb51b5ef CL |
1268 | break; |
1269 | ||
af697c0f JK |
1270 | if (unlikely(f2fs_cp_error(sbi))) |
1271 | break; | |
1272 | ||
9c30df7c JK |
1273 | if (type == F2FS_DIRTY_META) |
1274 | f2fs_sync_meta_pages(sbi, META, LONG_MAX, | |
1275 | FS_CP_META_IO); | |
1fd28018 JK |
1276 | else if (type == F2FS_WB_CP_DATA) |
1277 | f2fs_submit_merged_write(sbi, DATA); | |
828add77 JK |
1278 | |
1279 | prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE); | |
5df7731f | 1280 | io_schedule_timeout(DEFAULT_IO_TIMEOUT); |
fb51b5ef CL |
1281 | } |
1282 | finish_wait(&sbi->cp_wait, &wait); | |
1283 | } | |
1284 | ||
e4c5d848 JK |
1285 | static void update_ckpt_flags(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
1286 | { | |
1287 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; | |
1288 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
d1aa2453 | 1289 | unsigned long flags; |
e4c5d848 | 1290 | |
d1aa2453 | 1291 | spin_lock_irqsave(&sbi->cp_lock, flags); |
e4c5d848 | 1292 | |
c473f1a9 | 1293 | if ((cpc->reason & CP_UMOUNT) && |
10047f53 | 1294 | le32_to_cpu(ckpt->cp_pack_total_block_count) > |
22ad0b6a JK |
1295 | sbi->blocks_per_seg - NM_I(sbi)->nat_bits_blocks) |
1296 | disable_nat_bits(sbi, false); | |
1297 | ||
1f43e2ad CY |
1298 | if (cpc->reason & CP_TRIMMED) |
1299 | __set_ckpt_flags(ckpt, CP_TRIMMED_FLAG); | |
cd36d7a1 CY |
1300 | else |
1301 | __clear_ckpt_flags(ckpt, CP_TRIMMED_FLAG); | |
1f43e2ad | 1302 | |
c473f1a9 | 1303 | if (cpc->reason & CP_UMOUNT) |
e4c5d848 JK |
1304 | __set_ckpt_flags(ckpt, CP_UMOUNT_FLAG); |
1305 | else | |
1306 | __clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG); | |
1307 | ||
c473f1a9 | 1308 | if (cpc->reason & CP_FASTBOOT) |
e4c5d848 JK |
1309 | __set_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); |
1310 | else | |
1311 | __clear_ckpt_flags(ckpt, CP_FASTBOOT_FLAG); | |
1312 | ||
1313 | if (orphan_num) | |
1314 | __set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); | |
1315 | else | |
1316 | __clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG); | |
1317 | ||
c84ef3c5 | 1318 | if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) |
e4c5d848 JK |
1319 | __set_ckpt_flags(ckpt, CP_FSCK_FLAG); |
1320 | ||
c84ef3c5 ST |
1321 | if (is_sbi_flag_set(sbi, SBI_IS_RESIZEFS)) |
1322 | __set_ckpt_flags(ckpt, CP_RESIZEFS_FLAG); | |
1323 | else | |
1324 | __clear_ckpt_flags(ckpt, CP_RESIZEFS_FLAG); | |
1325 | ||
4354994f DR |
1326 | if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) |
1327 | __set_ckpt_flags(ckpt, CP_DISABLED_FLAG); | |
1328 | else | |
1329 | __clear_ckpt_flags(ckpt, CP_DISABLED_FLAG); | |
1330 | ||
db610a64 JK |
1331 | if (is_sbi_flag_set(sbi, SBI_CP_DISABLED_QUICK)) |
1332 | __set_ckpt_flags(ckpt, CP_DISABLED_QUICK_FLAG); | |
1333 | else | |
1334 | __clear_ckpt_flags(ckpt, CP_DISABLED_QUICK_FLAG); | |
1335 | ||
af033b2a CY |
1336 | if (is_sbi_flag_set(sbi, SBI_QUOTA_SKIP_FLUSH)) |
1337 | __set_ckpt_flags(ckpt, CP_QUOTA_NEED_FSCK_FLAG); | |
bc88ac96 JK |
1338 | else |
1339 | __clear_ckpt_flags(ckpt, CP_QUOTA_NEED_FSCK_FLAG); | |
af033b2a CY |
1340 | |
1341 | if (is_sbi_flag_set(sbi, SBI_QUOTA_NEED_REPAIR)) | |
1342 | __set_ckpt_flags(ckpt, CP_QUOTA_NEED_FSCK_FLAG); | |
1343 | ||
e4c5d848 JK |
1344 | /* set this flag to activate crc|cp_ver for recovery */ |
1345 | __set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG); | |
f2367923 | 1346 | __clear_ckpt_flags(ckpt, CP_NOCRC_RECOVERY_FLAG); |
e4c5d848 | 1347 | |
d1aa2453 | 1348 | spin_unlock_irqrestore(&sbi->cp_lock, flags); |
e4c5d848 JK |
1349 | } |
1350 | ||
46706d59 GX |
1351 | static void commit_checkpoint(struct f2fs_sb_info *sbi, |
1352 | void *src, block_t blk_addr) | |
1353 | { | |
1354 | struct writeback_control wbc = { | |
1355 | .for_reclaim = 0, | |
1356 | }; | |
1357 | ||
1358 | /* | |
1359 | * pagevec_lookup_tag and lock_page again will take | |
4d57b86d CY |
1360 | * some extra time. Therefore, f2fs_update_meta_pages and |
1361 | * f2fs_sync_meta_pages are combined in this function. | |
46706d59 | 1362 | */ |
4d57b86d | 1363 | struct page *page = f2fs_grab_meta_page(sbi, blk_addr); |
46706d59 GX |
1364 | int err; |
1365 | ||
bae0ee7a | 1366 | f2fs_wait_on_page_writeback(page, META, true, true); |
8d64d365 CY |
1367 | |
1368 | memcpy(page_address(page), src, PAGE_SIZE); | |
1369 | ||
1370 | set_page_dirty(page); | |
46706d59 GX |
1371 | if (unlikely(!clear_page_dirty_for_io(page))) |
1372 | f2fs_bug_on(sbi, 1); | |
1373 | ||
1374 | /* writeout cp pack 2 page */ | |
1375 | err = __f2fs_write_meta_page(page, &wbc, FS_CP_META_IO); | |
af697c0f JK |
1376 | if (unlikely(err && f2fs_cp_error(sbi))) { |
1377 | f2fs_put_page(page, 1); | |
1378 | return; | |
1379 | } | |
46706d59 | 1380 | |
af697c0f | 1381 | f2fs_bug_on(sbi, err); |
46706d59 GX |
1382 | f2fs_put_page(page, 0); |
1383 | ||
1384 | /* submit checkpoint (with barrier if NOBARRIER is not set) */ | |
1385 | f2fs_submit_merged_write(sbi, META_FLUSH); | |
1386 | } | |
1387 | ||
c34f42e2 | 1388 | static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1389 | { |
1390 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
77041823 | 1391 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
d1aa2453 | 1392 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num, flags; |
127e670a | 1393 | block_t start_blk; |
127e670a | 1394 | unsigned int data_sum_blocks, orphan_blocks; |
7e586fa0 | 1395 | __u32 crc32 = 0; |
127e670a | 1396 | int i; |
55141486 | 1397 | int cp_payload_blks = __cp_payload(sbi); |
8f1dbbbb SL |
1398 | struct curseg_info *seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); |
1399 | u64 kbytes_written; | |
1228b482 | 1400 | int err; |
127e670a JK |
1401 | |
1402 | /* Flush all the NAT/SIT pages */ | |
8ec18bff | 1403 | f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_CP_META_IO); |
127e670a | 1404 | |
7a88ddb5 | 1405 | /* start to update checkpoint, cp ver is already updated previously */ |
a1f72ac2 | 1406 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi, true)); |
127e670a | 1407 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); |
b5b82205 | 1408 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
1409 | ckpt->cur_node_segno[i] = |
1410 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
1411 | ckpt->cur_node_blkoff[i] = | |
1412 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
1413 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
1414 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
1415 | } | |
b5b82205 | 1416 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
1417 | ckpt->cur_data_segno[i] = |
1418 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
1419 | ckpt->cur_data_blkoff[i] = | |
1420 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
1421 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
1422 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
1423 | } | |
1424 | ||
a87aff1d | 1425 | /* 2 cp + n data seg summary + orphan inode blocks */ |
4d57b86d | 1426 | data_sum_blocks = f2fs_npages_for_summary_flush(sbi, false); |
d1aa2453 | 1427 | spin_lock_irqsave(&sbi->cp_lock, flags); |
b5b82205 | 1428 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
aaec2b1d | 1429 | __set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 1430 | else |
aaec2b1d | 1431 | __clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
d1aa2453 | 1432 | spin_unlock_irqrestore(&sbi->cp_lock, flags); |
127e670a | 1433 | |
67298804 | 1434 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
1435 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
1436 | orphan_blocks); | |
127e670a | 1437 | |
119ee914 | 1438 | if (__remain_node_summaries(cpc->reason)) |
b5b82205 | 1439 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
1440 | cp_payload_blks + data_sum_blocks + |
1441 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
119ee914 | 1442 | else |
b5b82205 | 1443 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS + |
1dbe4152 CL |
1444 | cp_payload_blks + data_sum_blocks + |
1445 | orphan_blocks); | |
119ee914 | 1446 | |
e4c5d848 JK |
1447 | /* update ckpt flag for checkpoint */ |
1448 | update_ckpt_flags(sbi, cpc); | |
a468f0ef | 1449 | |
127e670a JK |
1450 | /* update SIT/NAT bitmap */ |
1451 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
1452 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
1453 | ||
d7eb8f1c | 1454 | crc32 = f2fs_checkpoint_chksum(sbi, ckpt); |
7e586fa0 JK |
1455 | *((__le32 *)((unsigned char *)ckpt + |
1456 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
1457 | = cpu_to_le32(crc32); |
1458 | ||
8508e44a | 1459 | start_blk = __start_cp_next_addr(sbi); |
127e670a | 1460 | |
22ad0b6a JK |
1461 | /* write nat bits */ |
1462 | if (enabled_nat_bits(sbi, cpc)) { | |
1463 | __u64 cp_ver = cur_cp_version(ckpt); | |
22ad0b6a JK |
1464 | block_t blk; |
1465 | ||
1466 | cp_ver |= ((__u64)crc32 << 32); | |
1467 | *(__le64 *)nm_i->nat_bits = cpu_to_le64(cp_ver); | |
1468 | ||
1469 | blk = start_blk + sbi->blocks_per_seg - nm_i->nat_bits_blocks; | |
1470 | for (i = 0; i < nm_i->nat_bits_blocks; i++) | |
4d57b86d | 1471 | f2fs_update_meta_page(sbi, nm_i->nat_bits + |
22ad0b6a | 1472 | (i << F2FS_BLKSIZE_BITS), blk + i); |
22ad0b6a JK |
1473 | } |
1474 | ||
127e670a | 1475 | /* write out checkpoint buffer at block 0 */ |
4d57b86d | 1476 | f2fs_update_meta_page(sbi, ckpt, start_blk++); |
381722d2 CY |
1477 | |
1478 | for (i = 1; i < 1 + cp_payload_blks; i++) | |
4d57b86d | 1479 | f2fs_update_meta_page(sbi, (char *)ckpt + i * F2FS_BLKSIZE, |
381722d2 | 1480 | start_blk++); |
1dbe4152 | 1481 | |
67298804 | 1482 | if (orphan_num) { |
127e670a JK |
1483 | write_orphan_inodes(sbi, start_blk); |
1484 | start_blk += orphan_blocks; | |
1485 | } | |
1486 | ||
4d57b86d | 1487 | f2fs_write_data_summaries(sbi, start_blk); |
127e670a | 1488 | start_blk += data_sum_blocks; |
8f1dbbbb SL |
1489 | |
1490 | /* Record write statistics in the hot node summary */ | |
9499ffc7 | 1491 | kbytes_written = sbi->kbytes_written + BD_PART_WRITTEN(sbi); |
8f1dbbbb | 1492 | |
b7ad7512 | 1493 | seg_i->journal->info.kbytes_written = cpu_to_le64(kbytes_written); |
8f1dbbbb | 1494 | |
119ee914 | 1495 | if (__remain_node_summaries(cpc->reason)) { |
4d57b86d | 1496 | f2fs_write_node_summaries(sbi, start_blk); |
127e670a JK |
1497 | start_blk += NR_CURSEG_NODE_TYPE; |
1498 | } | |
1499 | ||
46706d59 GX |
1500 | /* update user_block_counts */ |
1501 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
1502 | percpu_counter_set(&sbi->alloc_valid_block_count, 0); | |
127e670a | 1503 | |
46706d59 | 1504 | /* Here, we have one bio having CP pack except cp pack 2 page */ |
4d57b86d | 1505 | f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_CP_META_IO); |
bf22c3cc ST |
1506 | /* Wait for all dirty meta pages to be submitted for IO */ |
1507 | f2fs_wait_on_all_pages(sbi, F2FS_DIRTY_META); | |
46706d59 GX |
1508 | |
1509 | /* wait for previous submitted meta pages writeback */ | |
bf22c3cc | 1510 | f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); |
127e670a | 1511 | |
46706d59 GX |
1512 | /* flush all device cache */ |
1513 | err = f2fs_flush_device_cache(sbi); | |
1514 | if (err) | |
1515 | return err; | |
127e670a | 1516 | |
46706d59 GX |
1517 | /* barrier and flush checkpoint cp pack 2 page if it can */ |
1518 | commit_checkpoint(sbi, ckpt, start_blk); | |
bf22c3cc | 1519 | f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); |
6a8f8ca5 | 1520 | |
18767e62 | 1521 | /* |
542989b6 | 1522 | * invalidate intermediate page cache borrowed from meta inode which are |
aff6fbbe | 1523 | * used for migration of encrypted, verity or compressed inode's blocks. |
18767e62 | 1524 | */ |
aff6fbbe CY |
1525 | if (f2fs_sb_has_encrypt(sbi) || f2fs_sb_has_verity(sbi) || |
1526 | f2fs_sb_has_compression(sbi)) | |
18767e62 CY |
1527 | invalidate_mapping_pages(META_MAPPING(sbi), |
1528 | MAIN_BLKADDR(sbi), MAX_BLKADDR(sbi) - 1); | |
1529 | ||
4d57b86d | 1530 | f2fs_release_ino_entry(sbi, false); |
cf779cab | 1531 | |
50fa53ec CY |
1532 | f2fs_reset_fsync_node_info(sbi); |
1533 | ||
caf0047e | 1534 | clear_sbi_flag(sbi, SBI_IS_DIRTY); |
bbf156f7 | 1535 | clear_sbi_flag(sbi, SBI_NEED_CP); |
af033b2a | 1536 | clear_sbi_flag(sbi, SBI_QUOTA_SKIP_FLUSH); |
c9c8ed50 CY |
1537 | |
1538 | spin_lock(&sbi->stat_lock); | |
4354994f | 1539 | sbi->unusable_block_count = 0; |
c9c8ed50 CY |
1540 | spin_unlock(&sbi->stat_lock); |
1541 | ||
8508e44a | 1542 | __set_cp_next_pack(sbi); |
c34f42e2 | 1543 | |
c2a080ae CY |
1544 | /* |
1545 | * redirty superblock if metadata like node page or inode cache is | |
1546 | * updated during writing checkpoint. | |
1547 | */ | |
1548 | if (get_pages(sbi, F2FS_DIRTY_NODES) || | |
1549 | get_pages(sbi, F2FS_DIRTY_IMETA)) | |
1550 | set_sbi_flag(sbi, SBI_IS_DIRTY); | |
1551 | ||
1552 | f2fs_bug_on(sbi, get_pages(sbi, F2FS_DIRTY_DENTS)); | |
1553 | ||
af697c0f | 1554 | return unlikely(f2fs_cp_error(sbi)) ? -EIO : 0; |
127e670a JK |
1555 | } |
1556 | ||
4d57b86d | 1557 | int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1558 | { |
1559 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1560 | unsigned long long ckpt_ver; | |
c34f42e2 | 1561 | int err = 0; |
127e670a | 1562 | |
f5a131bb CY |
1563 | if (f2fs_readonly(sbi->sb) || f2fs_hw_is_readonly(sbi)) |
1564 | return -EROFS; | |
1565 | ||
4354994f DR |
1566 | if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { |
1567 | if (cpc->reason != CP_PAUSE) | |
1568 | return 0; | |
dcbb4c10 | 1569 | f2fs_warn(sbi, "Start checkpoint disabled!"); |
4354994f | 1570 | } |
b4b10061 JK |
1571 | if (cpc->reason != CP_RESIZE) |
1572 | mutex_lock(&sbi->cp_mutex); | |
8501017e | 1573 | |
caf0047e | 1574 | if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) && |
c473f1a9 CY |
1575 | ((cpc->reason & CP_FASTBOOT) || (cpc->reason & CP_SYNC) || |
1576 | ((cpc->reason & CP_DISCARD) && !sbi->discard_blks))) | |
8501017e | 1577 | goto out; |
c34f42e2 CY |
1578 | if (unlikely(f2fs_cp_error(sbi))) { |
1579 | err = -EIO; | |
cf779cab | 1580 | goto out; |
c34f42e2 | 1581 | } |
2bda542d WL |
1582 | |
1583 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); | |
1584 | ||
c34f42e2 CY |
1585 | err = block_operations(sbi); |
1586 | if (err) | |
cf779cab | 1587 | goto out; |
127e670a | 1588 | |
75ab4cb8 | 1589 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1590 | |
b9109b0e | 1591 | f2fs_flush_merged_writes(sbi); |
127e670a | 1592 | |
58cce381 | 1593 | /* this is the case of multiple fstrims without any changes */ |
c473f1a9 | 1594 | if (cpc->reason & CP_DISCARD) { |
4d57b86d | 1595 | if (!f2fs_exist_trim_candidates(sbi, cpc)) { |
25290fa5 JK |
1596 | unblock_operations(sbi); |
1597 | goto out; | |
1598 | } | |
1599 | ||
0333ad4e JK |
1600 | if (NM_I(sbi)->dirty_nat_cnt == 0 && |
1601 | SIT_I(sbi)->dirty_sentries == 0 && | |
1602 | prefree_segments(sbi) == 0) { | |
4d57b86d CY |
1603 | f2fs_flush_sit_entries(sbi, cpc); |
1604 | f2fs_clear_prefree_segments(sbi, cpc); | |
0333ad4e JK |
1605 | unblock_operations(sbi); |
1606 | goto out; | |
1607 | } | |
58cce381 YH |
1608 | } |
1609 | ||
127e670a JK |
1610 | /* |
1611 | * update checkpoint pack index | |
1612 | * Increase the version number so that | |
1613 | * SIT entries and seg summaries are written at correct place | |
1614 | */ | |
d71b5564 | 1615 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1616 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1617 | ||
1618 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
edc55aaf JK |
1619 | err = f2fs_flush_nat_entries(sbi, cpc); |
1620 | if (err) | |
1621 | goto stop; | |
1622 | ||
4d57b86d | 1623 | f2fs_flush_sit_entries(sbi, cpc); |
127e670a | 1624 | |
d0b9e42a | 1625 | /* save inmem log status */ |
093749e2 | 1626 | f2fs_save_inmem_curseg(sbi); |
d0b9e42a | 1627 | |
c34f42e2 | 1628 | err = do_checkpoint(sbi, cpc); |
4e6a8d9b | 1629 | if (err) |
4d57b86d | 1630 | f2fs_release_discard_addrs(sbi); |
4e6a8d9b | 1631 | else |
4d57b86d | 1632 | f2fs_clear_prefree_segments(sbi, cpc); |
d0b9e42a | 1633 | |
093749e2 | 1634 | f2fs_restore_inmem_curseg(sbi); |
edc55aaf | 1635 | stop: |
127e670a | 1636 | unblock_operations(sbi); |
942e0be6 | 1637 | stat_inc_cp_count(sbi->stat_info); |
10027551 | 1638 | |
c473f1a9 | 1639 | if (cpc->reason & CP_RECOVERY) |
dcbb4c10 | 1640 | f2fs_notice(sbi, "checkpoint: version = %llx", ckpt_ver); |
60b99b48 | 1641 | |
7a88ddb5 | 1642 | /* update CP_TIME to trigger checkpoint periodically */ |
6beceb54 | 1643 | f2fs_update_time(sbi, CP_TIME); |
55d1cdb2 | 1644 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
8501017e | 1645 | out: |
b4b10061 JK |
1646 | if (cpc->reason != CP_RESIZE) |
1647 | mutex_unlock(&sbi->cp_mutex); | |
c34f42e2 | 1648 | return err; |
127e670a JK |
1649 | } |
1650 | ||
4d57b86d | 1651 | void f2fs_init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1652 | { |
6451e041 JK |
1653 | int i; |
1654 | ||
1655 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1656 | struct inode_management *im = &sbi->im[i]; |
1657 | ||
1658 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1659 | spin_lock_init(&im->ino_lock); | |
1660 | INIT_LIST_HEAD(&im->ino_list); | |
1661 | im->ino_num = 0; | |
6451e041 JK |
1662 | } |
1663 | ||
b5b82205 | 1664 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
d0b9e42a | 1665 | NR_CURSEG_PERSIST_TYPE - __cp_payload(sbi)) * |
14b42817 | 1666 | F2FS_ORPHANS_PER_BLOCK; |
127e670a JK |
1667 | } |
1668 | ||
4d57b86d | 1669 | int __init f2fs_create_checkpoint_caches(void) |
127e670a | 1670 | { |
6451e041 JK |
1671 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1672 | sizeof(struct ino_entry)); | |
1673 | if (!ino_entry_slab) | |
127e670a | 1674 | return -ENOMEM; |
4d57b86d | 1675 | f2fs_inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry", |
06292073 | 1676 | sizeof(struct inode_entry)); |
4d57b86d | 1677 | if (!f2fs_inode_entry_slab) { |
6451e041 | 1678 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1679 | return -ENOMEM; |
1680 | } | |
1681 | return 0; | |
1682 | } | |
1683 | ||
4d57b86d | 1684 | void f2fs_destroy_checkpoint_caches(void) |
127e670a | 1685 | { |
6451e041 | 1686 | kmem_cache_destroy(ino_entry_slab); |
4d57b86d | 1687 | kmem_cache_destroy(f2fs_inode_entry_slab); |
127e670a | 1688 | } |