]>
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> | |
261eeb9c | 16 | #include <linux/kthread.h> |
127e670a JK |
17 | |
18 | #include "f2fs.h" | |
19 | #include "node.h" | |
20 | #include "segment.h" | |
2af4bd6c | 21 | #include <trace/events/f2fs.h> |
127e670a | 22 | |
261eeb9c DJ |
23 | #define DEFAULT_CHECKPOINT_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) |
24 | ||
6451e041 | 25 | static struct kmem_cache *ino_entry_slab; |
4d57b86d | 26 | struct kmem_cache *f2fs_inode_entry_slab; |
127e670a | 27 | |
38f91ca8 JK |
28 | void f2fs_stop_checkpoint(struct f2fs_sb_info *sbi, bool end_io) |
29 | { | |
d494500a | 30 | f2fs_build_fault_attr(sbi, 0, 0); |
aaec2b1d | 31 | set_ckpt_flags(sbi, CP_ERROR_FLAG); |
38f91ca8 | 32 | if (!end_io) |
b9109b0e | 33 | f2fs_flush_merged_writes(sbi); |
38f91ca8 JK |
34 | } |
35 | ||
0a8165d7 | 36 | /* |
127e670a JK |
37 | * We guarantee no failure on the returned page. |
38 | */ | |
4d57b86d | 39 | struct page *f2fs_grab_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) |
127e670a | 40 | { |
9df27d98 | 41 | struct address_space *mapping = META_MAPPING(sbi); |
beb78181 | 42 | struct page *page; |
127e670a | 43 | repeat: |
300e129c | 44 | page = f2fs_grab_cache_page(mapping, index, false); |
127e670a JK |
45 | if (!page) { |
46 | cond_resched(); | |
47 | goto repeat; | |
48 | } | |
bae0ee7a | 49 | f2fs_wait_on_page_writeback(page, META, true, true); |
237c0790 JK |
50 | if (!PageUptodate(page)) |
51 | SetPageUptodate(page); | |
127e670a JK |
52 | return page; |
53 | } | |
54 | ||
2b947003 CY |
55 | static struct page *__get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index, |
56 | bool is_meta) | |
127e670a | 57 | { |
9df27d98 | 58 | struct address_space *mapping = META_MAPPING(sbi); |
127e670a | 59 | struct page *page; |
cf04e8eb | 60 | struct f2fs_io_info fio = { |
05ca3632 | 61 | .sbi = sbi, |
cf04e8eb | 62 | .type = META, |
04d328de | 63 | .op = REQ_OP_READ, |
70fd7614 | 64 | .op_flags = REQ_META | REQ_PRIO, |
7a9d7548 CY |
65 | .old_blkaddr = index, |
66 | .new_blkaddr = index, | |
4375a336 | 67 | .encrypted_page = NULL, |
6dc3a126 | 68 | .is_por = !is_meta, |
cf04e8eb | 69 | }; |
7735730d | 70 | int err; |
2b947003 CY |
71 | |
72 | if (unlikely(!is_meta)) | |
04d328de | 73 | fio.op_flags &= ~REQ_META; |
127e670a | 74 | repeat: |
300e129c | 75 | page = f2fs_grab_cache_page(mapping, index, false); |
127e670a JK |
76 | if (!page) { |
77 | cond_resched(); | |
78 | goto repeat; | |
79 | } | |
393ff91f JK |
80 | if (PageUptodate(page)) |
81 | goto out; | |
82 | ||
05ca3632 JK |
83 | fio.page = page; |
84 | ||
7735730d CY |
85 | err = f2fs_submit_page_bio(&fio); |
86 | if (err) { | |
87 | f2fs_put_page(page, 1); | |
88 | return ERR_PTR(err); | |
86531d6b | 89 | } |
127e670a | 90 | |
8b83ac81 CY |
91 | f2fs_update_iostat(sbi, FS_META_READ_IO, F2FS_BLKSIZE); |
92 | ||
393ff91f | 93 | lock_page(page); |
6bacf52f | 94 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
95 | f2fs_put_page(page, 1); |
96 | goto repeat; | |
97 | } | |
f3f338ca | 98 | |
81114baa | 99 | if (unlikely(!PageUptodate(page))) { |
7735730d CY |
100 | f2fs_put_page(page, 1); |
101 | return ERR_PTR(-EIO); | |
81114baa | 102 | } |
393ff91f | 103 | out: |
127e670a JK |
104 | return page; |
105 | } | |
106 | ||
4d57b86d | 107 | struct page *f2fs_get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index) |
2b947003 CY |
108 | { |
109 | return __get_meta_page(sbi, index, true); | |
110 | } | |
111 | ||
86f33603 | 112 | struct page *f2fs_get_meta_page_retry(struct f2fs_sb_info *sbi, pgoff_t index) |
7735730d CY |
113 | { |
114 | struct page *page; | |
115 | int count = 0; | |
116 | ||
117 | retry: | |
118 | page = __get_meta_page(sbi, index, true); | |
119 | if (IS_ERR(page)) { | |
120 | if (PTR_ERR(page) == -EIO && | |
121 | ++count <= DEFAULT_RETRY_IO_COUNT) | |
122 | goto retry; | |
7735730d | 123 | f2fs_stop_checkpoint(sbi, false); |
7735730d | 124 | } |
7735730d CY |
125 | return page; |
126 | } | |
127 | ||
2b947003 | 128 | /* for POR only */ |
4d57b86d | 129 | struct page *f2fs_get_tmp_page(struct f2fs_sb_info *sbi, pgoff_t index) |
2b947003 CY |
130 | { |
131 | return __get_meta_page(sbi, index, false); | |
132 | } | |
133 | ||
93770ab7 CY |
134 | static bool __is_bitmap_valid(struct f2fs_sb_info *sbi, block_t blkaddr, |
135 | int type) | |
136 | { | |
137 | struct seg_entry *se; | |
138 | unsigned int segno, offset; | |
139 | bool exist; | |
140 | ||
141 | if (type != DATA_GENERIC_ENHANCE && type != DATA_GENERIC_ENHANCE_READ) | |
142 | return true; | |
143 | ||
144 | segno = GET_SEGNO(sbi, blkaddr); | |
145 | offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr); | |
146 | se = get_seg_entry(sbi, segno); | |
147 | ||
148 | exist = f2fs_test_bit(offset, se->cur_valid_map); | |
149 | if (!exist && type == DATA_GENERIC_ENHANCE) { | |
dcbb4c10 JP |
150 | f2fs_err(sbi, "Inconsistent error blkaddr:%u, sit bitmap:%d", |
151 | blkaddr, exist); | |
93770ab7 CY |
152 | set_sbi_flag(sbi, SBI_NEED_FSCK); |
153 | WARN_ON(1); | |
154 | } | |
155 | return exist; | |
156 | } | |
157 | ||
e1da7872 | 158 | bool f2fs_is_valid_blkaddr(struct f2fs_sb_info *sbi, |
4d57b86d | 159 | block_t blkaddr, int type) |
662befda CY |
160 | { |
161 | switch (type) { | |
162 | case META_NAT: | |
66b00c18 | 163 | break; |
662befda | 164 | case META_SIT: |
66b00c18 CY |
165 | if (unlikely(blkaddr >= SIT_BLK_CNT(sbi))) |
166 | return false; | |
167 | break; | |
81c1a0f1 | 168 | case META_SSA: |
66b00c18 CY |
169 | if (unlikely(blkaddr >= MAIN_BLKADDR(sbi) || |
170 | blkaddr < SM_I(sbi)->ssa_blkaddr)) | |
171 | return false; | |
172 | break; | |
662befda | 173 | case META_CP: |
66b00c18 CY |
174 | if (unlikely(blkaddr >= SIT_I(sbi)->sit_base_addr || |
175 | blkaddr < __start_cp_addr(sbi))) | |
176 | return false; | |
177 | break; | |
4c521f49 | 178 | case META_POR: |
93770ab7 CY |
179 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
180 | blkaddr < MAIN_BLKADDR(sbi))) | |
181 | return false; | |
182 | break; | |
e1da7872 | 183 | case DATA_GENERIC: |
93770ab7 CY |
184 | case DATA_GENERIC_ENHANCE: |
185 | case DATA_GENERIC_ENHANCE_READ: | |
66b00c18 | 186 | if (unlikely(blkaddr >= MAX_BLKADDR(sbi) || |
93770ab7 | 187 | blkaddr < MAIN_BLKADDR(sbi))) { |
dcbb4c10 JP |
188 | f2fs_warn(sbi, "access invalid blkaddr:%u", |
189 | blkaddr); | |
93770ab7 CY |
190 | set_sbi_flag(sbi, SBI_NEED_FSCK); |
191 | WARN_ON(1); | |
66b00c18 | 192 | return false; |
93770ab7 CY |
193 | } else { |
194 | return __is_bitmap_valid(sbi, blkaddr, type); | |
c9b60788 | 195 | } |
66b00c18 | 196 | break; |
e1da7872 CY |
197 | case META_GENERIC: |
198 | if (unlikely(blkaddr < SEG0_BLKADDR(sbi) || | |
199 | blkaddr >= MAIN_BLKADDR(sbi))) | |
200 | return false; | |
201 | break; | |
662befda CY |
202 | default: |
203 | BUG(); | |
204 | } | |
66b00c18 CY |
205 | |
206 | return true; | |
662befda CY |
207 | } |
208 | ||
209 | /* | |
7a88ddb5 | 210 | * Readahead CP/NAT/SIT/SSA/POR pages |
662befda | 211 | */ |
4d57b86d | 212 | int f2fs_ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages, |
26879fb1 | 213 | int type, bool sync) |
662befda | 214 | { |
662befda | 215 | struct page *page; |
4c521f49 | 216 | block_t blkno = start; |
662befda | 217 | struct f2fs_io_info fio = { |
05ca3632 | 218 | .sbi = sbi, |
662befda | 219 | .type = META, |
04d328de | 220 | .op = REQ_OP_READ, |
70fd7614 | 221 | .op_flags = sync ? (REQ_META | REQ_PRIO) : REQ_RAHEAD, |
4375a336 | 222 | .encrypted_page = NULL, |
fb830fc5 | 223 | .in_list = false, |
6dc3a126 | 224 | .is_por = (type == META_POR), |
662befda | 225 | }; |
e9f5b8b8 | 226 | struct blk_plug plug; |
ce4c638c | 227 | int err; |
662befda | 228 | |
2b947003 | 229 | if (unlikely(type == META_POR)) |
04d328de | 230 | fio.op_flags &= ~REQ_META; |
2b947003 | 231 | |
e9f5b8b8 | 232 | blk_start_plug(&plug); |
662befda | 233 | for (; nrpages-- > 0; blkno++) { |
662befda | 234 | |
e1da7872 | 235 | if (!f2fs_is_valid_blkaddr(sbi, blkno, type)) |
66b00c18 CY |
236 | goto out; |
237 | ||
662befda CY |
238 | switch (type) { |
239 | case META_NAT: | |
66b00c18 CY |
240 | if (unlikely(blkno >= |
241 | NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))) | |
662befda | 242 | blkno = 0; |
66b00c18 | 243 | /* get nat block addr */ |
7a9d7548 | 244 | fio.new_blkaddr = current_nat_addr(sbi, |
662befda CY |
245 | blkno * NAT_ENTRY_PER_BLOCK); |
246 | break; | |
247 | case META_SIT: | |
6a257471 CY |
248 | if (unlikely(blkno >= TOTAL_SEGS(sbi))) |
249 | goto out; | |
662befda | 250 | /* get sit block addr */ |
7a9d7548 | 251 | fio.new_blkaddr = current_sit_addr(sbi, |
662befda | 252 | blkno * SIT_ENTRY_PER_BLOCK); |
662befda | 253 | break; |
81c1a0f1 | 254 | case META_SSA: |
662befda | 255 | case META_CP: |
4c521f49 | 256 | case META_POR: |
7a9d7548 | 257 | fio.new_blkaddr = blkno; |
662befda CY |
258 | break; |
259 | default: | |
260 | BUG(); | |
261 | } | |
262 | ||
300e129c JK |
263 | page = f2fs_grab_cache_page(META_MAPPING(sbi), |
264 | fio.new_blkaddr, false); | |
662befda CY |
265 | if (!page) |
266 | continue; | |
267 | if (PageUptodate(page)) { | |
662befda CY |
268 | f2fs_put_page(page, 1); |
269 | continue; | |
270 | } | |
271 | ||
05ca3632 | 272 | fio.page = page; |
ce4c638c CY |
273 | err = f2fs_submit_page_bio(&fio); |
274 | f2fs_put_page(page, err ? 1 : 0); | |
8b83ac81 CY |
275 | |
276 | if (!err) | |
277 | f2fs_update_iostat(sbi, FS_META_READ_IO, F2FS_BLKSIZE); | |
662befda CY |
278 | } |
279 | out: | |
e9f5b8b8 | 280 | blk_finish_plug(&plug); |
662befda CY |
281 | return blkno - start; |
282 | } | |
283 | ||
4d57b86d | 284 | void f2fs_ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index) |
635aee1f CY |
285 | { |
286 | struct page *page; | |
287 | bool readahead = false; | |
288 | ||
289 | page = find_get_page(META_MAPPING(sbi), index); | |
4da7bf5a | 290 | if (!page || !PageUptodate(page)) |
635aee1f CY |
291 | readahead = true; |
292 | f2fs_put_page(page, 0); | |
293 | ||
294 | if (readahead) | |
a8affc03 | 295 | f2fs_ra_meta_pages(sbi, index, BIO_MAX_VECS, META_POR, true); |
635aee1f CY |
296 | } |
297 | ||
b0af6d49 CY |
298 | static int __f2fs_write_meta_page(struct page *page, |
299 | struct writeback_control *wbc, | |
300 | enum iostat_type io_type) | |
127e670a | 301 | { |
4081363f | 302 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
127e670a | 303 | |
ecda0de3 CY |
304 | trace_f2fs_writepage(page, META); |
305 | ||
af697c0f JK |
306 | if (unlikely(f2fs_cp_error(sbi))) |
307 | goto redirty_out; | |
caf0047e | 308 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
cfb271d4 | 309 | goto redirty_out; |
857dc4e0 | 310 | if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0)) |
cfb271d4 | 311 | goto redirty_out; |
127e670a | 312 | |
4d57b86d | 313 | f2fs_do_write_meta_page(sbi, page, io_type); |
577e3495 | 314 | dec_page_count(sbi, F2FS_DIRTY_META); |
0c3a5797 CY |
315 | |
316 | if (wbc->for_reclaim) | |
bab475c5 | 317 | f2fs_submit_merged_write_cond(sbi, NULL, page, 0, META); |
0c3a5797 | 318 | |
577e3495 | 319 | unlock_page(page); |
857dc4e0 | 320 | |
0c3a5797 | 321 | if (unlikely(f2fs_cp_error(sbi))) |
b9109b0e | 322 | f2fs_submit_merged_write(sbi, META); |
0c3a5797 | 323 | |
577e3495 | 324 | return 0; |
cfb271d4 CY |
325 | |
326 | redirty_out: | |
76f60268 | 327 | redirty_page_for_writepage(wbc, page); |
cfb271d4 | 328 | return AOP_WRITEPAGE_ACTIVATE; |
127e670a JK |
329 | } |
330 | ||
b0af6d49 CY |
331 | static int f2fs_write_meta_page(struct page *page, |
332 | struct writeback_control *wbc) | |
333 | { | |
334 | return __f2fs_write_meta_page(page, wbc, FS_META_IO); | |
335 | } | |
336 | ||
127e670a JK |
337 | static int f2fs_write_meta_pages(struct address_space *mapping, |
338 | struct writeback_control *wbc) | |
339 | { | |
4081363f | 340 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 341 | long diff, written; |
127e670a | 342 | |
0771fcc7 CY |
343 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
344 | goto skip_write; | |
345 | ||
5459aa97 | 346 | /* collect a number of dirty meta pages and write together */ |
812a9597 JK |
347 | if (wbc->sync_mode != WB_SYNC_ALL && |
348 | get_pages(sbi, F2FS_DIRTY_META) < | |
349 | nr_pages_to_skip(sbi, META)) | |
d3baf95d | 350 | goto skip_write; |
127e670a | 351 | |
a29d0e0b | 352 | /* if locked failed, cp will flush dirty pages instead */ |
8769918b | 353 | if (!down_write_trylock(&sbi->cp_global_sem)) |
a29d0e0b | 354 | goto skip_write; |
d31c7c3f | 355 | |
a29d0e0b | 356 | trace_f2fs_writepages(mapping->host, wbc, META); |
50c8cdb3 | 357 | diff = nr_pages_to_write(sbi, META, wbc); |
4d57b86d | 358 | written = f2fs_sync_meta_pages(sbi, META, wbc->nr_to_write, FS_META_IO); |
8769918b | 359 | up_write(&sbi->cp_global_sem); |
50c8cdb3 | 360 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff); |
127e670a | 361 | return 0; |
d3baf95d JK |
362 | |
363 | skip_write: | |
364 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META); | |
d31c7c3f | 365 | trace_f2fs_writepages(mapping->host, wbc, META); |
d3baf95d | 366 | return 0; |
127e670a JK |
367 | } |
368 | ||
4d57b86d | 369 | long f2fs_sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type, |
b0af6d49 | 370 | long nr_to_write, enum iostat_type io_type) |
127e670a | 371 | { |
9df27d98 | 372 | struct address_space *mapping = META_MAPPING(sbi); |
028a63a6 | 373 | pgoff_t index = 0, prev = ULONG_MAX; |
127e670a JK |
374 | struct pagevec pvec; |
375 | long nwritten = 0; | |
028a63a6 | 376 | int nr_pages; |
127e670a JK |
377 | struct writeback_control wbc = { |
378 | .for_reclaim = 0, | |
379 | }; | |
e9f5b8b8 | 380 | struct blk_plug plug; |
127e670a | 381 | |
86679820 | 382 | pagevec_init(&pvec); |
127e670a | 383 | |
e9f5b8b8 CY |
384 | blk_start_plug(&plug); |
385 | ||
028a63a6 | 386 | while ((nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, |
67fd707f | 387 | PAGECACHE_TAG_DIRTY))) { |
028a63a6 | 388 | int i; |
127e670a JK |
389 | |
390 | for (i = 0; i < nr_pages; i++) { | |
391 | struct page *page = pvec.pages[i]; | |
203681f6 | 392 | |
80dd9c0e | 393 | if (prev == ULONG_MAX) |
6066d8cd JK |
394 | prev = page->index - 1; |
395 | if (nr_to_write != LONG_MAX && page->index != prev + 1) { | |
396 | pagevec_release(&pvec); | |
397 | goto stop; | |
398 | } | |
399 | ||
127e670a | 400 | lock_page(page); |
203681f6 JK |
401 | |
402 | if (unlikely(page->mapping != mapping)) { | |
403 | continue_unlock: | |
404 | unlock_page(page); | |
405 | continue; | |
406 | } | |
407 | if (!PageDirty(page)) { | |
408 | /* someone wrote it for us */ | |
409 | goto continue_unlock; | |
410 | } | |
411 | ||
bae0ee7a | 412 | f2fs_wait_on_page_writeback(page, META, true, true); |
fa3d2bdf | 413 | |
203681f6 JK |
414 | if (!clear_page_dirty_for_io(page)) |
415 | goto continue_unlock; | |
416 | ||
b0af6d49 | 417 | if (__f2fs_write_meta_page(page, &wbc, io_type)) { |
577e3495 JK |
418 | unlock_page(page); |
419 | break; | |
420 | } | |
cfb271d4 | 421 | nwritten++; |
6066d8cd | 422 | prev = page->index; |
cfb271d4 | 423 | if (unlikely(nwritten >= nr_to_write)) |
127e670a JK |
424 | break; |
425 | } | |
426 | pagevec_release(&pvec); | |
427 | cond_resched(); | |
428 | } | |
6066d8cd | 429 | stop: |
127e670a | 430 | if (nwritten) |
b9109b0e | 431 | f2fs_submit_merged_write(sbi, type); |
127e670a | 432 | |
e9f5b8b8 CY |
433 | blk_finish_plug(&plug); |
434 | ||
127e670a JK |
435 | return nwritten; |
436 | } | |
437 | ||
438 | static int f2fs_set_meta_page_dirty(struct page *page) | |
439 | { | |
26c6b887 JK |
440 | trace_f2fs_set_page_dirty(page, META); |
441 | ||
237c0790 JK |
442 | if (!PageUptodate(page)) |
443 | SetPageUptodate(page); | |
127e670a | 444 | if (!PageDirty(page)) { |
b87078ad | 445 | __set_page_dirty_nobuffers(page); |
4081363f | 446 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); |
240a5915 | 447 | f2fs_set_page_private(page, 0); |
127e670a JK |
448 | return 1; |
449 | } | |
450 | return 0; | |
451 | } | |
452 | ||
453 | const struct address_space_operations f2fs_meta_aops = { | |
454 | .writepage = f2fs_write_meta_page, | |
455 | .writepages = f2fs_write_meta_pages, | |
456 | .set_page_dirty = f2fs_set_meta_page_dirty, | |
487261f3 CY |
457 | .invalidatepage = f2fs_invalidate_page, |
458 | .releasepage = f2fs_release_page, | |
5b7a487c WG |
459 | #ifdef CONFIG_MIGRATION |
460 | .migratepage = f2fs_migrate_page, | |
461 | #endif | |
127e670a JK |
462 | }; |
463 | ||
39d787be CY |
464 | static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, |
465 | unsigned int devidx, int type) | |
953e6cc6 | 466 | { |
67298804 | 467 | struct inode_management *im = &sbi->im[type]; |
80c54505 JK |
468 | struct ino_entry *e, *tmp; |
469 | ||
470 | tmp = f2fs_kmem_cache_alloc(ino_entry_slab, GFP_NOFS); | |
19526d74 | 471 | |
80c54505 | 472 | radix_tree_preload(GFP_NOFS | __GFP_NOFAIL); |
769ec6e5 | 473 | |
67298804 | 474 | spin_lock(&im->ino_lock); |
67298804 | 475 | e = radix_tree_lookup(&im->ino_root, ino); |
39efac41 | 476 | if (!e) { |
80c54505 | 477 | e = tmp; |
19526d74 CY |
478 | if (unlikely(radix_tree_insert(&im->ino_root, ino, e))) |
479 | f2fs_bug_on(sbi, 1); | |
480 | ||
39efac41 JK |
481 | memset(e, 0, sizeof(struct ino_entry)); |
482 | e->ino = ino; | |
953e6cc6 | 483 | |
67298804 | 484 | list_add_tail(&e->list, &im->ino_list); |
8c402946 | 485 | if (type != ORPHAN_INO) |
67298804 | 486 | im->ino_num++; |
39efac41 | 487 | } |
39d787be CY |
488 | |
489 | if (type == FLUSH_INO) | |
490 | f2fs_set_bit(devidx, (char *)&e->dirty_device); | |
491 | ||
67298804 | 492 | spin_unlock(&im->ino_lock); |
769ec6e5 | 493 | radix_tree_preload_end(); |
80c54505 JK |
494 | |
495 | if (e != tmp) | |
496 | kmem_cache_free(ino_entry_slab, tmp); | |
953e6cc6 JK |
497 | } |
498 | ||
6451e041 | 499 | static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
953e6cc6 | 500 | { |
67298804 | 501 | struct inode_management *im = &sbi->im[type]; |
6451e041 | 502 | struct ino_entry *e; |
953e6cc6 | 503 | |
67298804 CY |
504 | spin_lock(&im->ino_lock); |
505 | e = radix_tree_lookup(&im->ino_root, ino); | |
39efac41 JK |
506 | if (e) { |
507 | list_del(&e->list); | |
67298804 CY |
508 | radix_tree_delete(&im->ino_root, ino); |
509 | im->ino_num--; | |
510 | spin_unlock(&im->ino_lock); | |
39efac41 JK |
511 | kmem_cache_free(ino_entry_slab, e); |
512 | return; | |
953e6cc6 | 513 | } |
67298804 | 514 | spin_unlock(&im->ino_lock); |
953e6cc6 JK |
515 | } |
516 | ||
4d57b86d | 517 | void f2fs_add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
518 | { |
519 | /* add new dirty ino entry into list */ | |
39d787be | 520 | __add_ino_entry(sbi, ino, 0, type); |
fff04f90 JK |
521 | } |
522 | ||
4d57b86d | 523 | void f2fs_remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type) |
fff04f90 JK |
524 | { |
525 | /* remove dirty ino entry from list */ | |
526 | __remove_ino_entry(sbi, ino, type); | |
527 | } | |
528 | ||
1f07cc58 | 529 | /* mode should be APPEND_INO, UPDATE_INO or TRANS_DIR_INO */ |
4d57b86d | 530 | bool f2fs_exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode) |
fff04f90 | 531 | { |
67298804 | 532 | struct inode_management *im = &sbi->im[mode]; |
fff04f90 | 533 | struct ino_entry *e; |
67298804 CY |
534 | |
535 | spin_lock(&im->ino_lock); | |
536 | e = radix_tree_lookup(&im->ino_root, ino); | |
537 | spin_unlock(&im->ino_lock); | |
fff04f90 JK |
538 | return e ? true : false; |
539 | } | |
540 | ||
4d57b86d | 541 | void f2fs_release_ino_entry(struct f2fs_sb_info *sbi, bool all) |
fff04f90 JK |
542 | { |
543 | struct ino_entry *e, *tmp; | |
544 | int i; | |
545 | ||
39d787be | 546 | for (i = all ? ORPHAN_INO : APPEND_INO; i < MAX_INO_ENTRY; i++) { |
67298804 CY |
547 | struct inode_management *im = &sbi->im[i]; |
548 | ||
549 | spin_lock(&im->ino_lock); | |
550 | list_for_each_entry_safe(e, tmp, &im->ino_list, list) { | |
fff04f90 | 551 | list_del(&e->list); |
67298804 | 552 | radix_tree_delete(&im->ino_root, e->ino); |
fff04f90 | 553 | kmem_cache_free(ino_entry_slab, e); |
67298804 | 554 | im->ino_num--; |
fff04f90 | 555 | } |
67298804 | 556 | spin_unlock(&im->ino_lock); |
fff04f90 JK |
557 | } |
558 | } | |
559 | ||
4d57b86d | 560 | void f2fs_set_dirty_device(struct f2fs_sb_info *sbi, nid_t ino, |
39d787be CY |
561 | unsigned int devidx, int type) |
562 | { | |
563 | __add_ino_entry(sbi, ino, devidx, type); | |
564 | } | |
565 | ||
4d57b86d | 566 | bool f2fs_is_dirty_device(struct f2fs_sb_info *sbi, nid_t ino, |
39d787be CY |
567 | unsigned int devidx, int type) |
568 | { | |
569 | struct inode_management *im = &sbi->im[type]; | |
570 | struct ino_entry *e; | |
571 | bool is_dirty = false; | |
572 | ||
573 | spin_lock(&im->ino_lock); | |
574 | e = radix_tree_lookup(&im->ino_root, ino); | |
575 | if (e && f2fs_test_bit(devidx, (char *)&e->dirty_device)) | |
576 | is_dirty = true; | |
577 | spin_unlock(&im->ino_lock); | |
578 | return is_dirty; | |
579 | } | |
580 | ||
4d57b86d | 581 | int f2fs_acquire_orphan_inode(struct f2fs_sb_info *sbi) |
127e670a | 582 | { |
67298804 | 583 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a JK |
584 | int err = 0; |
585 | ||
67298804 | 586 | spin_lock(&im->ino_lock); |
cb78942b | 587 | |
1ecc0c5c | 588 | if (time_to_inject(sbi, FAULT_ORPHAN)) { |
cb78942b | 589 | spin_unlock(&im->ino_lock); |
c45d6002 | 590 | f2fs_show_injection_info(sbi, FAULT_ORPHAN); |
cb78942b JK |
591 | return -ENOSPC; |
592 | } | |
7fa750a1 | 593 | |
67298804 | 594 | if (unlikely(im->ino_num >= sbi->max_orphans)) |
127e670a | 595 | err = -ENOSPC; |
cbd56e7d | 596 | else |
67298804 CY |
597 | im->ino_num++; |
598 | spin_unlock(&im->ino_lock); | |
0d47c1ad | 599 | |
127e670a JK |
600 | return err; |
601 | } | |
602 | ||
4d57b86d | 603 | void f2fs_release_orphan_inode(struct f2fs_sb_info *sbi) |
cbd56e7d | 604 | { |
67298804 CY |
605 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
606 | ||
607 | spin_lock(&im->ino_lock); | |
608 | f2fs_bug_on(sbi, im->ino_num == 0); | |
609 | im->ino_num--; | |
610 | spin_unlock(&im->ino_lock); | |
cbd56e7d JK |
611 | } |
612 | ||
4d57b86d | 613 | void f2fs_add_orphan_inode(struct inode *inode) |
127e670a | 614 | { |
39efac41 | 615 | /* add new orphan ino entry into list */ |
39d787be | 616 | __add_ino_entry(F2FS_I_SB(inode), inode->i_ino, 0, ORPHAN_INO); |
4d57b86d | 617 | f2fs_update_inode_page(inode); |
127e670a JK |
618 | } |
619 | ||
4d57b86d | 620 | void f2fs_remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 621 | { |
953e6cc6 | 622 | /* remove orphan entry from orphan list */ |
6451e041 | 623 | __remove_ino_entry(sbi, ino, ORPHAN_INO); |
127e670a JK |
624 | } |
625 | ||
8c14bfad | 626 | static int recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) |
127e670a | 627 | { |
8c14bfad | 628 | struct inode *inode; |
5905f9af | 629 | struct node_info ni; |
76a45e3c | 630 | int err; |
8c14bfad | 631 | |
5905f9af | 632 | inode = f2fs_iget_retry(sbi->sb, ino); |
8c14bfad CY |
633 | if (IS_ERR(inode)) { |
634 | /* | |
635 | * there should be a bug that we can't find the entry | |
636 | * to orphan inode. | |
637 | */ | |
638 | f2fs_bug_on(sbi, PTR_ERR(inode) == -ENOENT); | |
639 | return PTR_ERR(inode); | |
640 | } | |
641 | ||
0f9ec2a8 | 642 | err = dquot_initialize(inode); |
a515d12f SY |
643 | if (err) { |
644 | iput(inode); | |
0f9ec2a8 | 645 | goto err_out; |
a515d12f | 646 | } |
0f9ec2a8 | 647 | |
127e670a JK |
648 | clear_nlink(inode); |
649 | ||
650 | /* truncate all the data during iput */ | |
651 | iput(inode); | |
5905f9af | 652 | |
7735730d CY |
653 | err = f2fs_get_node_info(sbi, ino, &ni); |
654 | if (err) | |
655 | goto err_out; | |
5905f9af JK |
656 | |
657 | /* ENOMEM was fully retried in f2fs_evict_inode. */ | |
658 | if (ni.blk_addr != NULL_ADDR) { | |
0f9ec2a8 JK |
659 | err = -EIO; |
660 | goto err_out; | |
5905f9af | 661 | } |
8c14bfad | 662 | return 0; |
0f9ec2a8 JK |
663 | |
664 | err_out: | |
665 | set_sbi_flag(sbi, SBI_NEED_FSCK); | |
dcbb4c10 JP |
666 | f2fs_warn(sbi, "%s: orphan failed (ino=%x), run fsck to fix.", |
667 | __func__, ino); | |
0f9ec2a8 | 668 | return err; |
127e670a JK |
669 | } |
670 | ||
4d57b86d | 671 | int f2fs_recover_orphan_inodes(struct f2fs_sb_info *sbi) |
127e670a | 672 | { |
3c642985 | 673 | block_t start_blk, orphan_blocks, i, j; |
4b2414d0 CY |
674 | unsigned int s_flags = sbi->sb->s_flags; |
675 | int err = 0; | |
ea676733 JK |
676 | #ifdef CONFIG_QUOTA |
677 | int quota_enabled; | |
678 | #endif | |
127e670a | 679 | |
aaec2b1d | 680 | if (!is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG)) |
8c14bfad | 681 | return 0; |
127e670a | 682 | |
b61af314 | 683 | if (bdev_read_only(sbi->sb->s_bdev)) { |
dcbb4c10 | 684 | f2fs_info(sbi, "write access unavailable, skipping orphan cleanup"); |
b61af314 CY |
685 | return 0; |
686 | } | |
687 | ||
1751e8a6 | 688 | if (s_flags & SB_RDONLY) { |
dcbb4c10 | 689 | f2fs_info(sbi, "orphan cleanup on readonly fs"); |
1751e8a6 | 690 | sbi->sb->s_flags &= ~SB_RDONLY; |
4b2414d0 CY |
691 | } |
692 | ||
693 | #ifdef CONFIG_QUOTA | |
694 | /* Needed for iput() to work correctly and not trash data */ | |
1751e8a6 | 695 | sbi->sb->s_flags |= SB_ACTIVE; |
ea676733 | 696 | |
76cf05d7 SY |
697 | /* |
698 | * Turn on quotas which were not enabled for read-only mounts if | |
699 | * filesystem has quota feature, so that they are updated correctly. | |
700 | */ | |
1751e8a6 | 701 | quota_enabled = f2fs_enable_quota_files(sbi, s_flags & SB_RDONLY); |
4b2414d0 CY |
702 | #endif |
703 | ||
55141486 | 704 | start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi); |
3c642985 | 705 | orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi); |
127e670a | 706 | |
4d57b86d | 707 | f2fs_ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP, true); |
662befda | 708 | |
3c642985 | 709 | for (i = 0; i < orphan_blocks; i++) { |
7735730d | 710 | struct page *page; |
127e670a JK |
711 | struct f2fs_orphan_block *orphan_blk; |
712 | ||
7735730d CY |
713 | page = f2fs_get_meta_page(sbi, start_blk + i); |
714 | if (IS_ERR(page)) { | |
715 | err = PTR_ERR(page); | |
716 | goto out; | |
717 | } | |
718 | ||
127e670a JK |
719 | orphan_blk = (struct f2fs_orphan_block *)page_address(page); |
720 | for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) { | |
721 | nid_t ino = le32_to_cpu(orphan_blk->ino[j]); | |
8c14bfad CY |
722 | err = recover_orphan_inode(sbi, ino); |
723 | if (err) { | |
724 | f2fs_put_page(page, 1); | |
4b2414d0 | 725 | goto out; |
8c14bfad | 726 | } |
127e670a JK |
727 | } |
728 | f2fs_put_page(page, 1); | |
729 | } | |
730 | /* clear Orphan Flag */ | |
aaec2b1d | 731 | clear_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG); |
4b2414d0 | 732 | out: |
1378752b CY |
733 | set_sbi_flag(sbi, SBI_IS_RECOVERED); |
734 | ||
4b2414d0 CY |
735 | #ifdef CONFIG_QUOTA |
736 | /* Turn quotas off */ | |
ea676733 JK |
737 | if (quota_enabled) |
738 | f2fs_quota_off_umount(sbi->sb); | |
4b2414d0 | 739 | #endif |
1751e8a6 | 740 | sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ |
4b2414d0 CY |
741 | |
742 | return err; | |
127e670a JK |
743 | } |
744 | ||
745 | static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk) | |
746 | { | |
502c6e0b | 747 | struct list_head *head; |
127e670a | 748 | struct f2fs_orphan_block *orphan_blk = NULL; |
127e670a | 749 | unsigned int nentries = 0; |
bd936f84 | 750 | unsigned short index = 1; |
8c402946 | 751 | unsigned short orphan_blocks; |
4531929e | 752 | struct page *page = NULL; |
6451e041 | 753 | struct ino_entry *orphan = NULL; |
67298804 | 754 | struct inode_management *im = &sbi->im[ORPHAN_INO]; |
127e670a | 755 | |
67298804 | 756 | orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num); |
8c402946 | 757 | |
d6c67a4f JK |
758 | /* |
759 | * we don't need to do spin_lock(&im->ino_lock) here, since all the | |
760 | * orphan inode operations are covered under f2fs_lock_op(). | |
761 | * And, spin_lock should be avoided due to page operations below. | |
762 | */ | |
67298804 | 763 | head = &im->ino_list; |
127e670a JK |
764 | |
765 | /* loop for each orphan inode entry and write them in Jornal block */ | |
502c6e0b GZ |
766 | list_for_each_entry(orphan, head, list) { |
767 | if (!page) { | |
4d57b86d | 768 | page = f2fs_grab_meta_page(sbi, start_blk++); |
502c6e0b GZ |
769 | orphan_blk = |
770 | (struct f2fs_orphan_block *)page_address(page); | |
771 | memset(orphan_blk, 0, sizeof(*orphan_blk)); | |
772 | } | |
127e670a | 773 | |
36795567 | 774 | orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino); |
127e670a | 775 | |
36795567 | 776 | if (nentries == F2FS_ORPHANS_PER_BLOCK) { |
127e670a JK |
777 | /* |
778 | * an orphan block is full of 1020 entries, | |
779 | * then we need to flush current orphan blocks | |
780 | * and bring another one in memory | |
781 | */ | |
782 | orphan_blk->blk_addr = cpu_to_le16(index); | |
783 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
784 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
785 | set_page_dirty(page); | |
786 | f2fs_put_page(page, 1); | |
787 | index++; | |
127e670a JK |
788 | nentries = 0; |
789 | page = NULL; | |
790 | } | |
502c6e0b | 791 | } |
127e670a | 792 | |
502c6e0b GZ |
793 | if (page) { |
794 | orphan_blk->blk_addr = cpu_to_le16(index); | |
795 | orphan_blk->blk_count = cpu_to_le16(orphan_blocks); | |
796 | orphan_blk->entry_count = cpu_to_le32(nentries); | |
797 | set_page_dirty(page); | |
798 | f2fs_put_page(page, 1); | |
127e670a | 799 | } |
127e670a JK |
800 | } |
801 | ||
d7eb8f1c CY |
802 | static __u32 f2fs_checkpoint_chksum(struct f2fs_sb_info *sbi, |
803 | struct f2fs_checkpoint *ckpt) | |
804 | { | |
805 | unsigned int chksum_ofs = le32_to_cpu(ckpt->checksum_offset); | |
806 | __u32 chksum; | |
807 | ||
808 | chksum = f2fs_crc32(sbi, ckpt, chksum_ofs); | |
809 | if (chksum_ofs < CP_CHKSUM_OFFSET) { | |
810 | chksum_ofs += sizeof(chksum); | |
811 | chksum = f2fs_chksum(sbi, chksum, (__u8 *)ckpt + chksum_ofs, | |
812 | F2FS_BLKSIZE - chksum_ofs); | |
813 | } | |
814 | return chksum; | |
815 | } | |
816 | ||
fc0065ad TY |
817 | static int get_checkpoint_version(struct f2fs_sb_info *sbi, block_t cp_addr, |
818 | struct f2fs_checkpoint **cp_block, struct page **cp_page, | |
819 | unsigned long long *version) | |
127e670a | 820 | { |
fc0065ad | 821 | size_t crc_offset = 0; |
d7eb8f1c | 822 | __u32 crc; |
127e670a | 823 | |
4d57b86d | 824 | *cp_page = f2fs_get_meta_page(sbi, cp_addr); |
7735730d CY |
825 | if (IS_ERR(*cp_page)) |
826 | return PTR_ERR(*cp_page); | |
827 | ||
fc0065ad | 828 | *cp_block = (struct f2fs_checkpoint *)page_address(*cp_page); |
127e670a | 829 | |
fc0065ad | 830 | crc_offset = le32_to_cpu((*cp_block)->checksum_offset); |
d7eb8f1c CY |
831 | if (crc_offset < CP_MIN_CHKSUM_OFFSET || |
832 | crc_offset > CP_CHKSUM_OFFSET) { | |
d3f07c04 | 833 | f2fs_put_page(*cp_page, 1); |
dcbb4c10 | 834 | f2fs_warn(sbi, "invalid crc_offset: %zu", crc_offset); |
fc0065ad TY |
835 | return -EINVAL; |
836 | } | |
127e670a | 837 | |
d7eb8f1c CY |
838 | crc = f2fs_checkpoint_chksum(sbi, *cp_block); |
839 | if (crc != cur_cp_crc(*cp_block)) { | |
d3f07c04 | 840 | f2fs_put_page(*cp_page, 1); |
dcbb4c10 | 841 | f2fs_warn(sbi, "invalid crc value"); |
fc0065ad TY |
842 | return -EINVAL; |
843 | } | |
127e670a | 844 | |
fc0065ad TY |
845 | *version = cur_cp_version(*cp_block); |
846 | return 0; | |
847 | } | |
127e670a | 848 | |
fc0065ad TY |
849 | static struct page *validate_checkpoint(struct f2fs_sb_info *sbi, |
850 | block_t cp_addr, unsigned long long *version) | |
851 | { | |
852 | struct page *cp_page_1 = NULL, *cp_page_2 = NULL; | |
853 | struct f2fs_checkpoint *cp_block = NULL; | |
854 | unsigned long long cur_version = 0, pre_version = 0; | |
855 | int err; | |
127e670a | 856 | |
fc0065ad TY |
857 | err = get_checkpoint_version(sbi, cp_addr, &cp_block, |
858 | &cp_page_1, version); | |
859 | if (err) | |
d3f07c04 | 860 | return NULL; |
c9b60788 CY |
861 | |
862 | if (le32_to_cpu(cp_block->cp_pack_total_block_count) > | |
863 | sbi->blocks_per_seg) { | |
dcbb4c10 JP |
864 | f2fs_warn(sbi, "invalid cp_pack_total_block_count:%u", |
865 | le32_to_cpu(cp_block->cp_pack_total_block_count)); | |
d3f07c04 | 866 | goto invalid_cp; |
c9b60788 | 867 | } |
fc0065ad | 868 | pre_version = *version; |
127e670a | 869 | |
fc0065ad TY |
870 | cp_addr += le32_to_cpu(cp_block->cp_pack_total_block_count) - 1; |
871 | err = get_checkpoint_version(sbi, cp_addr, &cp_block, | |
872 | &cp_page_2, version); | |
873 | if (err) | |
d3f07c04 | 874 | goto invalid_cp; |
fc0065ad | 875 | cur_version = *version; |
127e670a JK |
876 | |
877 | if (cur_version == pre_version) { | |
878 | *version = cur_version; | |
879 | f2fs_put_page(cp_page_2, 1); | |
880 | return cp_page_1; | |
881 | } | |
127e670a | 882 | f2fs_put_page(cp_page_2, 1); |
d3f07c04 | 883 | invalid_cp: |
127e670a JK |
884 | f2fs_put_page(cp_page_1, 1); |
885 | return NULL; | |
886 | } | |
887 | ||
4d57b86d | 888 | int f2fs_get_valid_checkpoint(struct f2fs_sb_info *sbi) |
127e670a JK |
889 | { |
890 | struct f2fs_checkpoint *cp_block; | |
891 | struct f2fs_super_block *fsb = sbi->raw_super; | |
892 | struct page *cp1, *cp2, *cur_page; | |
893 | unsigned long blk_size = sbi->blocksize; | |
894 | unsigned long long cp1_version = 0, cp2_version = 0; | |
895 | unsigned long long cp_start_blk_no; | |
55141486 | 896 | unsigned int cp_blks = 1 + __cp_payload(sbi); |
1dbe4152 CL |
897 | block_t cp_blk_no; |
898 | int i; | |
10f966bb | 899 | int err; |
127e670a | 900 | |
0b6d4ca0 EB |
901 | sbi->ckpt = f2fs_kvzalloc(sbi, array_size(blk_size, cp_blks), |
902 | GFP_KERNEL); | |
127e670a JK |
903 | if (!sbi->ckpt) |
904 | return -ENOMEM; | |
905 | /* | |
906 | * Finding out valid cp block involves read both | |
7a88ddb5 | 907 | * sets( cp pack 1 and cp pack 2) |
127e670a JK |
908 | */ |
909 | cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
910 | cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version); | |
911 | ||
912 | /* The second checkpoint pack should start at the next segment */ | |
f9a4e6df JK |
913 | cp_start_blk_no += ((unsigned long long)1) << |
914 | le32_to_cpu(fsb->log_blocks_per_seg); | |
127e670a JK |
915 | cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version); |
916 | ||
917 | if (cp1 && cp2) { | |
918 | if (ver_after(cp2_version, cp1_version)) | |
919 | cur_page = cp2; | |
920 | else | |
921 | cur_page = cp1; | |
922 | } else if (cp1) { | |
923 | cur_page = cp1; | |
924 | } else if (cp2) { | |
925 | cur_page = cp2; | |
926 | } else { | |
10f966bb | 927 | err = -EFSCORRUPTED; |
127e670a JK |
928 | goto fail_no_cp; |
929 | } | |
930 | ||
931 | cp_block = (struct f2fs_checkpoint *)page_address(cur_page); | |
932 | memcpy(sbi->ckpt, cp_block, blk_size); | |
933 | ||
8508e44a JK |
934 | if (cur_page == cp1) |
935 | sbi->cur_cp_pack = 1; | |
936 | else | |
937 | sbi->cur_cp_pack = 2; | |
984ec63c | 938 | |
e494c2f9 | 939 | /* Sanity checking of checkpoint */ |
10f966bb CY |
940 | if (f2fs_sanity_check_ckpt(sbi)) { |
941 | err = -EFSCORRUPTED; | |
e494c2f9 | 942 | goto free_fail_no_cp; |
10f966bb | 943 | } |
e494c2f9 | 944 | |
1dbe4152 CL |
945 | if (cp_blks <= 1) |
946 | goto done; | |
947 | ||
948 | cp_blk_no = le32_to_cpu(fsb->cp_blkaddr); | |
949 | if (cur_page == cp2) | |
950 | cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg); | |
951 | ||
952 | for (i = 1; i < cp_blks; i++) { | |
953 | void *sit_bitmap_ptr; | |
954 | unsigned char *ckpt = (unsigned char *)sbi->ckpt; | |
955 | ||
4d57b86d | 956 | cur_page = f2fs_get_meta_page(sbi, cp_blk_no + i); |
10f966bb CY |
957 | if (IS_ERR(cur_page)) { |
958 | err = PTR_ERR(cur_page); | |
7735730d | 959 | goto free_fail_no_cp; |
10f966bb | 960 | } |
1dbe4152 CL |
961 | sit_bitmap_ptr = page_address(cur_page); |
962 | memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size); | |
963 | f2fs_put_page(cur_page, 1); | |
964 | } | |
965 | done: | |
127e670a JK |
966 | f2fs_put_page(cp1, 1); |
967 | f2fs_put_page(cp2, 1); | |
968 | return 0; | |
969 | ||
a2125ff7 JK |
970 | free_fail_no_cp: |
971 | f2fs_put_page(cp1, 1); | |
972 | f2fs_put_page(cp2, 1); | |
127e670a | 973 | fail_no_cp: |
5222595d | 974 | kvfree(sbi->ckpt); |
10f966bb | 975 | return err; |
127e670a JK |
976 | } |
977 | ||
c227f912 | 978 | static void __add_dirty_inode(struct inode *inode, enum inode_type type) |
127e670a | 979 | { |
4081363f | 980 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 981 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
127e670a | 982 | |
91942321 | 983 | if (is_inode_flag_set(inode, flag)) |
2710fd7e | 984 | return; |
2d7b822a | 985 | |
91942321 | 986 | set_inode_flag(inode, flag); |
99f4b917 CY |
987 | if (!f2fs_is_volatile_file(inode)) |
988 | list_add_tail(&F2FS_I(inode)->dirty_list, | |
989 | &sbi->inode_list[type]); | |
33fbd510 | 990 | stat_inc_dirty_inode(sbi, type); |
5deb8267 JK |
991 | } |
992 | ||
c227f912 | 993 | static void __remove_dirty_inode(struct inode *inode, enum inode_type type) |
6ad7609a | 994 | { |
c227f912 | 995 | int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE; |
6ad7609a | 996 | |
91942321 | 997 | if (get_dirty_pages(inode) || !is_inode_flag_set(inode, flag)) |
6ad7609a CY |
998 | return; |
999 | ||
91942321 JK |
1000 | list_del_init(&F2FS_I(inode)->dirty_list); |
1001 | clear_inode_flag(inode, flag); | |
33fbd510 | 1002 | stat_dec_dirty_inode(F2FS_I_SB(inode), type); |
6ad7609a CY |
1003 | } |
1004 | ||
4d57b86d | 1005 | void f2fs_update_dirty_page(struct inode *inode, struct page *page) |
5deb8267 | 1006 | { |
4081363f | 1007 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
c227f912 | 1008 | enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE; |
5deb8267 | 1009 | |
5ac9f36f CY |
1010 | if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) && |
1011 | !S_ISLNK(inode->i_mode)) | |
127e670a | 1012 | return; |
7bd59381 | 1013 | |
1c4bf763 JK |
1014 | spin_lock(&sbi->inode_lock[type]); |
1015 | if (type != FILE_INODE || test_opt(sbi, DATA_FLUSH)) | |
10aa97c3 | 1016 | __add_dirty_inode(inode, type); |
b951a4ec | 1017 | inode_inc_dirty_pages(inode); |
1c4bf763 JK |
1018 | spin_unlock(&sbi->inode_lock[type]); |
1019 | ||
240a5915 | 1020 | f2fs_set_page_private(page, 0); |
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 | ||
3a0a9cbc CY |
1388 | static inline u64 get_sectors_written(struct block_device *bdev) |
1389 | { | |
ff49c86f | 1390 | return (u64)part_stat_read(bdev, sectors[STAT_WRITE]); |
3a0a9cbc CY |
1391 | } |
1392 | ||
1393 | u64 f2fs_get_sectors_written(struct f2fs_sb_info *sbi) | |
1394 | { | |
1395 | if (f2fs_is_multi_device(sbi)) { | |
1396 | u64 sectors = 0; | |
1397 | int i; | |
1398 | ||
1399 | for (i = 0; i < sbi->s_ndevs; i++) | |
1400 | sectors += get_sectors_written(FDEV(i).bdev); | |
1401 | ||
1402 | return sectors; | |
1403 | } | |
1404 | ||
1405 | return get_sectors_written(sbi->sb->s_bdev); | |
1406 | } | |
1407 | ||
c34f42e2 | 1408 | static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1409 | { |
1410 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
77041823 | 1411 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
d1aa2453 | 1412 | unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num, flags; |
127e670a | 1413 | block_t start_blk; |
127e670a | 1414 | unsigned int data_sum_blocks, orphan_blocks; |
7e586fa0 | 1415 | __u32 crc32 = 0; |
127e670a | 1416 | int i; |
55141486 | 1417 | int cp_payload_blks = __cp_payload(sbi); |
8f1dbbbb SL |
1418 | struct curseg_info *seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); |
1419 | u64 kbytes_written; | |
1228b482 | 1420 | int err; |
127e670a JK |
1421 | |
1422 | /* Flush all the NAT/SIT pages */ | |
8ec18bff | 1423 | f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_CP_META_IO); |
127e670a | 1424 | |
7a88ddb5 | 1425 | /* start to update checkpoint, cp ver is already updated previously */ |
a1f72ac2 | 1426 | ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi, true)); |
127e670a | 1427 | ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); |
b5b82205 | 1428 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { |
127e670a JK |
1429 | ckpt->cur_node_segno[i] = |
1430 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE)); | |
1431 | ckpt->cur_node_blkoff[i] = | |
1432 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE)); | |
1433 | ckpt->alloc_type[i + CURSEG_HOT_NODE] = | |
1434 | curseg_alloc_type(sbi, i + CURSEG_HOT_NODE); | |
1435 | } | |
b5b82205 | 1436 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { |
127e670a JK |
1437 | ckpt->cur_data_segno[i] = |
1438 | cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA)); | |
1439 | ckpt->cur_data_blkoff[i] = | |
1440 | cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA)); | |
1441 | ckpt->alloc_type[i + CURSEG_HOT_DATA] = | |
1442 | curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); | |
1443 | } | |
1444 | ||
a87aff1d | 1445 | /* 2 cp + n data seg summary + orphan inode blocks */ |
4d57b86d | 1446 | data_sum_blocks = f2fs_npages_for_summary_flush(sbi, false); |
d1aa2453 | 1447 | spin_lock_irqsave(&sbi->cp_lock, flags); |
b5b82205 | 1448 | if (data_sum_blocks < NR_CURSEG_DATA_TYPE) |
aaec2b1d | 1449 | __set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
127e670a | 1450 | else |
aaec2b1d | 1451 | __clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG); |
d1aa2453 | 1452 | spin_unlock_irqrestore(&sbi->cp_lock, flags); |
127e670a | 1453 | |
67298804 | 1454 | orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num); |
1dbe4152 CL |
1455 | ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks + |
1456 | orphan_blocks); | |
127e670a | 1457 | |
119ee914 | 1458 | if (__remain_node_summaries(cpc->reason)) |
b5b82205 | 1459 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS+ |
1dbe4152 CL |
1460 | cp_payload_blks + data_sum_blocks + |
1461 | orphan_blocks + NR_CURSEG_NODE_TYPE); | |
119ee914 | 1462 | else |
b5b82205 | 1463 | ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS + |
1dbe4152 CL |
1464 | cp_payload_blks + data_sum_blocks + |
1465 | orphan_blocks); | |
119ee914 | 1466 | |
e4c5d848 JK |
1467 | /* update ckpt flag for checkpoint */ |
1468 | update_ckpt_flags(sbi, cpc); | |
a468f0ef | 1469 | |
127e670a JK |
1470 | /* update SIT/NAT bitmap */ |
1471 | get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP)); | |
1472 | get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP)); | |
1473 | ||
d7eb8f1c | 1474 | crc32 = f2fs_checkpoint_chksum(sbi, ckpt); |
7e586fa0 JK |
1475 | *((__le32 *)((unsigned char *)ckpt + |
1476 | le32_to_cpu(ckpt->checksum_offset))) | |
127e670a JK |
1477 | = cpu_to_le32(crc32); |
1478 | ||
8508e44a | 1479 | start_blk = __start_cp_next_addr(sbi); |
127e670a | 1480 | |
22ad0b6a JK |
1481 | /* write nat bits */ |
1482 | if (enabled_nat_bits(sbi, cpc)) { | |
1483 | __u64 cp_ver = cur_cp_version(ckpt); | |
22ad0b6a JK |
1484 | block_t blk; |
1485 | ||
1486 | cp_ver |= ((__u64)crc32 << 32); | |
1487 | *(__le64 *)nm_i->nat_bits = cpu_to_le64(cp_ver); | |
1488 | ||
1489 | blk = start_blk + sbi->blocks_per_seg - nm_i->nat_bits_blocks; | |
1490 | for (i = 0; i < nm_i->nat_bits_blocks; i++) | |
4d57b86d | 1491 | f2fs_update_meta_page(sbi, nm_i->nat_bits + |
22ad0b6a | 1492 | (i << F2FS_BLKSIZE_BITS), blk + i); |
22ad0b6a JK |
1493 | } |
1494 | ||
127e670a | 1495 | /* write out checkpoint buffer at block 0 */ |
4d57b86d | 1496 | f2fs_update_meta_page(sbi, ckpt, start_blk++); |
381722d2 CY |
1497 | |
1498 | for (i = 1; i < 1 + cp_payload_blks; i++) | |
4d57b86d | 1499 | f2fs_update_meta_page(sbi, (char *)ckpt + i * F2FS_BLKSIZE, |
381722d2 | 1500 | start_blk++); |
1dbe4152 | 1501 | |
67298804 | 1502 | if (orphan_num) { |
127e670a JK |
1503 | write_orphan_inodes(sbi, start_blk); |
1504 | start_blk += orphan_blocks; | |
1505 | } | |
1506 | ||
4d57b86d | 1507 | f2fs_write_data_summaries(sbi, start_blk); |
127e670a | 1508 | start_blk += data_sum_blocks; |
8f1dbbbb SL |
1509 | |
1510 | /* Record write statistics in the hot node summary */ | |
1511 | kbytes_written = sbi->kbytes_written; | |
3a0a9cbc CY |
1512 | kbytes_written += (f2fs_get_sectors_written(sbi) - |
1513 | sbi->sectors_written_start) >> 1; | |
b7ad7512 | 1514 | seg_i->journal->info.kbytes_written = cpu_to_le64(kbytes_written); |
8f1dbbbb | 1515 | |
119ee914 | 1516 | if (__remain_node_summaries(cpc->reason)) { |
4d57b86d | 1517 | f2fs_write_node_summaries(sbi, start_blk); |
127e670a JK |
1518 | start_blk += NR_CURSEG_NODE_TYPE; |
1519 | } | |
1520 | ||
46706d59 GX |
1521 | /* update user_block_counts */ |
1522 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
1523 | percpu_counter_set(&sbi->alloc_valid_block_count, 0); | |
127e670a | 1524 | |
46706d59 | 1525 | /* Here, we have one bio having CP pack except cp pack 2 page */ |
4d57b86d | 1526 | f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_CP_META_IO); |
bf22c3cc ST |
1527 | /* Wait for all dirty meta pages to be submitted for IO */ |
1528 | f2fs_wait_on_all_pages(sbi, F2FS_DIRTY_META); | |
46706d59 GX |
1529 | |
1530 | /* wait for previous submitted meta pages writeback */ | |
bf22c3cc | 1531 | f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); |
127e670a | 1532 | |
46706d59 GX |
1533 | /* flush all device cache */ |
1534 | err = f2fs_flush_device_cache(sbi); | |
1535 | if (err) | |
1536 | return err; | |
127e670a | 1537 | |
46706d59 GX |
1538 | /* barrier and flush checkpoint cp pack 2 page if it can */ |
1539 | commit_checkpoint(sbi, ckpt, start_blk); | |
bf22c3cc | 1540 | f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); |
6a8f8ca5 | 1541 | |
18767e62 | 1542 | /* |
542989b6 | 1543 | * invalidate intermediate page cache borrowed from meta inode which are |
aff6fbbe | 1544 | * used for migration of encrypted, verity or compressed inode's blocks. |
18767e62 | 1545 | */ |
aff6fbbe CY |
1546 | if (f2fs_sb_has_encrypt(sbi) || f2fs_sb_has_verity(sbi) || |
1547 | f2fs_sb_has_compression(sbi)) | |
18767e62 CY |
1548 | invalidate_mapping_pages(META_MAPPING(sbi), |
1549 | MAIN_BLKADDR(sbi), MAX_BLKADDR(sbi) - 1); | |
1550 | ||
4d57b86d | 1551 | f2fs_release_ino_entry(sbi, false); |
cf779cab | 1552 | |
50fa53ec CY |
1553 | f2fs_reset_fsync_node_info(sbi); |
1554 | ||
caf0047e | 1555 | clear_sbi_flag(sbi, SBI_IS_DIRTY); |
bbf156f7 | 1556 | clear_sbi_flag(sbi, SBI_NEED_CP); |
af033b2a | 1557 | clear_sbi_flag(sbi, SBI_QUOTA_SKIP_FLUSH); |
c9c8ed50 CY |
1558 | |
1559 | spin_lock(&sbi->stat_lock); | |
4354994f | 1560 | sbi->unusable_block_count = 0; |
c9c8ed50 CY |
1561 | spin_unlock(&sbi->stat_lock); |
1562 | ||
8508e44a | 1563 | __set_cp_next_pack(sbi); |
c34f42e2 | 1564 | |
c2a080ae CY |
1565 | /* |
1566 | * redirty superblock if metadata like node page or inode cache is | |
1567 | * updated during writing checkpoint. | |
1568 | */ | |
1569 | if (get_pages(sbi, F2FS_DIRTY_NODES) || | |
1570 | get_pages(sbi, F2FS_DIRTY_IMETA)) | |
1571 | set_sbi_flag(sbi, SBI_IS_DIRTY); | |
1572 | ||
1573 | f2fs_bug_on(sbi, get_pages(sbi, F2FS_DIRTY_DENTS)); | |
1574 | ||
af697c0f | 1575 | return unlikely(f2fs_cp_error(sbi)) ? -EIO : 0; |
127e670a JK |
1576 | } |
1577 | ||
4d57b86d | 1578 | int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
127e670a JK |
1579 | { |
1580 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1581 | unsigned long long ckpt_ver; | |
c34f42e2 | 1582 | int err = 0; |
127e670a | 1583 | |
f5a131bb CY |
1584 | if (f2fs_readonly(sbi->sb) || f2fs_hw_is_readonly(sbi)) |
1585 | return -EROFS; | |
1586 | ||
4354994f DR |
1587 | if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { |
1588 | if (cpc->reason != CP_PAUSE) | |
1589 | return 0; | |
dcbb4c10 | 1590 | f2fs_warn(sbi, "Start checkpoint disabled!"); |
4354994f | 1591 | } |
b4b10061 | 1592 | if (cpc->reason != CP_RESIZE) |
8769918b | 1593 | down_write(&sbi->cp_global_sem); |
8501017e | 1594 | |
caf0047e | 1595 | if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) && |
c473f1a9 CY |
1596 | ((cpc->reason & CP_FASTBOOT) || (cpc->reason & CP_SYNC) || |
1597 | ((cpc->reason & CP_DISCARD) && !sbi->discard_blks))) | |
8501017e | 1598 | goto out; |
c34f42e2 CY |
1599 | if (unlikely(f2fs_cp_error(sbi))) { |
1600 | err = -EIO; | |
cf779cab | 1601 | goto out; |
c34f42e2 | 1602 | } |
2bda542d WL |
1603 | |
1604 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops"); | |
1605 | ||
c34f42e2 CY |
1606 | err = block_operations(sbi); |
1607 | if (err) | |
cf779cab | 1608 | goto out; |
127e670a | 1609 | |
75ab4cb8 | 1610 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops"); |
2af4bd6c | 1611 | |
b9109b0e | 1612 | f2fs_flush_merged_writes(sbi); |
127e670a | 1613 | |
58cce381 | 1614 | /* this is the case of multiple fstrims without any changes */ |
c473f1a9 | 1615 | if (cpc->reason & CP_DISCARD) { |
4d57b86d | 1616 | if (!f2fs_exist_trim_candidates(sbi, cpc)) { |
25290fa5 JK |
1617 | unblock_operations(sbi); |
1618 | goto out; | |
1619 | } | |
1620 | ||
a95ba66a | 1621 | if (NM_I(sbi)->nat_cnt[DIRTY_NAT] == 0 && |
0333ad4e JK |
1622 | SIT_I(sbi)->dirty_sentries == 0 && |
1623 | prefree_segments(sbi) == 0) { | |
4d57b86d CY |
1624 | f2fs_flush_sit_entries(sbi, cpc); |
1625 | f2fs_clear_prefree_segments(sbi, cpc); | |
0333ad4e JK |
1626 | unblock_operations(sbi); |
1627 | goto out; | |
1628 | } | |
58cce381 YH |
1629 | } |
1630 | ||
127e670a JK |
1631 | /* |
1632 | * update checkpoint pack index | |
1633 | * Increase the version number so that | |
1634 | * SIT entries and seg summaries are written at correct place | |
1635 | */ | |
d71b5564 | 1636 | ckpt_ver = cur_cp_version(ckpt); |
127e670a JK |
1637 | ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver); |
1638 | ||
1639 | /* write cached NAT/SIT entries to NAT/SIT area */ | |
edc55aaf JK |
1640 | err = f2fs_flush_nat_entries(sbi, cpc); |
1641 | if (err) | |
1642 | goto stop; | |
1643 | ||
4d57b86d | 1644 | f2fs_flush_sit_entries(sbi, cpc); |
127e670a | 1645 | |
d0b9e42a | 1646 | /* save inmem log status */ |
093749e2 | 1647 | f2fs_save_inmem_curseg(sbi); |
d0b9e42a | 1648 | |
c34f42e2 | 1649 | err = do_checkpoint(sbi, cpc); |
4e6a8d9b | 1650 | if (err) |
4d57b86d | 1651 | f2fs_release_discard_addrs(sbi); |
4e6a8d9b | 1652 | else |
4d57b86d | 1653 | f2fs_clear_prefree_segments(sbi, cpc); |
d0b9e42a | 1654 | |
093749e2 | 1655 | f2fs_restore_inmem_curseg(sbi); |
edc55aaf | 1656 | stop: |
127e670a | 1657 | unblock_operations(sbi); |
942e0be6 | 1658 | stat_inc_cp_count(sbi->stat_info); |
10027551 | 1659 | |
c473f1a9 | 1660 | if (cpc->reason & CP_RECOVERY) |
dcbb4c10 | 1661 | f2fs_notice(sbi, "checkpoint: version = %llx", ckpt_ver); |
60b99b48 | 1662 | |
7a88ddb5 | 1663 | /* update CP_TIME to trigger checkpoint periodically */ |
6beceb54 | 1664 | f2fs_update_time(sbi, CP_TIME); |
55d1cdb2 | 1665 | trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint"); |
8501017e | 1666 | out: |
b4b10061 | 1667 | if (cpc->reason != CP_RESIZE) |
8769918b | 1668 | up_write(&sbi->cp_global_sem); |
c34f42e2 | 1669 | return err; |
127e670a JK |
1670 | } |
1671 | ||
4d57b86d | 1672 | void f2fs_init_ino_entry_info(struct f2fs_sb_info *sbi) |
127e670a | 1673 | { |
6451e041 JK |
1674 | int i; |
1675 | ||
1676 | for (i = 0; i < MAX_INO_ENTRY; i++) { | |
67298804 CY |
1677 | struct inode_management *im = &sbi->im[i]; |
1678 | ||
1679 | INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC); | |
1680 | spin_lock_init(&im->ino_lock); | |
1681 | INIT_LIST_HEAD(&im->ino_list); | |
1682 | im->ino_num = 0; | |
6451e041 JK |
1683 | } |
1684 | ||
b5b82205 | 1685 | sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS - |
d0b9e42a | 1686 | NR_CURSEG_PERSIST_TYPE - __cp_payload(sbi)) * |
14b42817 | 1687 | F2FS_ORPHANS_PER_BLOCK; |
127e670a JK |
1688 | } |
1689 | ||
4d57b86d | 1690 | int __init f2fs_create_checkpoint_caches(void) |
127e670a | 1691 | { |
6451e041 JK |
1692 | ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry", |
1693 | sizeof(struct ino_entry)); | |
1694 | if (!ino_entry_slab) | |
127e670a | 1695 | return -ENOMEM; |
4d57b86d | 1696 | f2fs_inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry", |
06292073 | 1697 | sizeof(struct inode_entry)); |
4d57b86d | 1698 | if (!f2fs_inode_entry_slab) { |
6451e041 | 1699 | kmem_cache_destroy(ino_entry_slab); |
127e670a JK |
1700 | return -ENOMEM; |
1701 | } | |
1702 | return 0; | |
1703 | } | |
1704 | ||
4d57b86d | 1705 | void f2fs_destroy_checkpoint_caches(void) |
127e670a | 1706 | { |
6451e041 | 1707 | kmem_cache_destroy(ino_entry_slab); |
4d57b86d | 1708 | kmem_cache_destroy(f2fs_inode_entry_slab); |
127e670a | 1709 | } |
261eeb9c DJ |
1710 | |
1711 | static int __write_checkpoint_sync(struct f2fs_sb_info *sbi) | |
1712 | { | |
1713 | struct cp_control cpc = { .reason = CP_SYNC, }; | |
1714 | int err; | |
1715 | ||
1716 | down_write(&sbi->gc_lock); | |
1717 | err = f2fs_write_checkpoint(sbi, &cpc); | |
1718 | up_write(&sbi->gc_lock); | |
1719 | ||
1720 | return err; | |
1721 | } | |
1722 | ||
1723 | static void __checkpoint_and_complete_reqs(struct f2fs_sb_info *sbi) | |
1724 | { | |
1725 | struct ckpt_req_control *cprc = &sbi->cprc_info; | |
1726 | struct ckpt_req *req, *next; | |
1727 | struct llist_node *dispatch_list; | |
1728 | u64 sum_diff = 0, diff, count = 0; | |
1729 | int ret; | |
1730 | ||
1731 | dispatch_list = llist_del_all(&cprc->issue_list); | |
1732 | if (!dispatch_list) | |
1733 | return; | |
1734 | dispatch_list = llist_reverse_order(dispatch_list); | |
1735 | ||
1736 | ret = __write_checkpoint_sync(sbi); | |
1737 | atomic_inc(&cprc->issued_ckpt); | |
1738 | ||
1739 | llist_for_each_entry_safe(req, next, dispatch_list, llnode) { | |
1740 | diff = (u64)ktime_ms_delta(ktime_get(), req->queue_time); | |
1741 | req->ret = ret; | |
1742 | complete(&req->wait); | |
1743 | ||
1744 | sum_diff += diff; | |
1745 | count++; | |
1746 | } | |
1747 | atomic_sub(count, &cprc->queued_ckpt); | |
1748 | atomic_add(count, &cprc->total_ckpt); | |
1749 | ||
1750 | spin_lock(&cprc->stat_lock); | |
1751 | cprc->cur_time = (unsigned int)div64_u64(sum_diff, count); | |
1752 | if (cprc->peak_time < cprc->cur_time) | |
1753 | cprc->peak_time = cprc->cur_time; | |
1754 | spin_unlock(&cprc->stat_lock); | |
1755 | } | |
1756 | ||
1757 | static int issue_checkpoint_thread(void *data) | |
1758 | { | |
1759 | struct f2fs_sb_info *sbi = data; | |
1760 | struct ckpt_req_control *cprc = &sbi->cprc_info; | |
1761 | wait_queue_head_t *q = &cprc->ckpt_wait_queue; | |
1762 | repeat: | |
1763 | if (kthread_should_stop()) | |
1764 | return 0; | |
1765 | ||
261eeb9c DJ |
1766 | if (!llist_empty(&cprc->issue_list)) |
1767 | __checkpoint_and_complete_reqs(sbi); | |
1768 | ||
261eeb9c DJ |
1769 | wait_event_interruptible(*q, |
1770 | kthread_should_stop() || !llist_empty(&cprc->issue_list)); | |
1771 | goto repeat; | |
1772 | } | |
1773 | ||
1774 | static void flush_remained_ckpt_reqs(struct f2fs_sb_info *sbi, | |
1775 | struct ckpt_req *wait_req) | |
1776 | { | |
1777 | struct ckpt_req_control *cprc = &sbi->cprc_info; | |
1778 | ||
1779 | if (!llist_empty(&cprc->issue_list)) { | |
1780 | __checkpoint_and_complete_reqs(sbi); | |
1781 | } else { | |
1782 | /* already dispatched by issue_checkpoint_thread */ | |
1783 | if (wait_req) | |
1784 | wait_for_completion(&wait_req->wait); | |
1785 | } | |
1786 | } | |
1787 | ||
1788 | static void init_ckpt_req(struct ckpt_req *req) | |
1789 | { | |
1790 | memset(req, 0, sizeof(struct ckpt_req)); | |
1791 | ||
1792 | init_completion(&req->wait); | |
1793 | req->queue_time = ktime_get(); | |
1794 | } | |
1795 | ||
1796 | int f2fs_issue_checkpoint(struct f2fs_sb_info *sbi) | |
1797 | { | |
1798 | struct ckpt_req_control *cprc = &sbi->cprc_info; | |
1799 | struct ckpt_req req; | |
1800 | struct cp_control cpc; | |
1801 | ||
1802 | cpc.reason = __get_cp_reason(sbi); | |
1803 | if (!test_opt(sbi, MERGE_CHECKPOINT) || cpc.reason != CP_SYNC) { | |
1804 | int ret; | |
1805 | ||
1806 | down_write(&sbi->gc_lock); | |
1807 | ret = f2fs_write_checkpoint(sbi, &cpc); | |
1808 | up_write(&sbi->gc_lock); | |
1809 | ||
1810 | return ret; | |
1811 | } | |
1812 | ||
1813 | if (!cprc->f2fs_issue_ckpt) | |
1814 | return __write_checkpoint_sync(sbi); | |
1815 | ||
1816 | init_ckpt_req(&req); | |
1817 | ||
1818 | llist_add(&req.llnode, &cprc->issue_list); | |
1819 | atomic_inc(&cprc->queued_ckpt); | |
1820 | ||
1821 | /* update issue_list before we wake up issue_checkpoint thread */ | |
1822 | smp_mb(); | |
1823 | ||
1824 | if (waitqueue_active(&cprc->ckpt_wait_queue)) | |
1825 | wake_up(&cprc->ckpt_wait_queue); | |
1826 | ||
1827 | if (cprc->f2fs_issue_ckpt) | |
1828 | wait_for_completion(&req.wait); | |
1829 | else | |
1830 | flush_remained_ckpt_reqs(sbi, &req); | |
1831 | ||
1832 | return req.ret; | |
1833 | } | |
1834 | ||
1835 | int f2fs_start_ckpt_thread(struct f2fs_sb_info *sbi) | |
1836 | { | |
1837 | dev_t dev = sbi->sb->s_bdev->bd_dev; | |
1838 | struct ckpt_req_control *cprc = &sbi->cprc_info; | |
1839 | ||
1840 | if (cprc->f2fs_issue_ckpt) | |
1841 | return 0; | |
1842 | ||
1843 | cprc->f2fs_issue_ckpt = kthread_run(issue_checkpoint_thread, sbi, | |
1844 | "f2fs_ckpt-%u:%u", MAJOR(dev), MINOR(dev)); | |
1845 | if (IS_ERR(cprc->f2fs_issue_ckpt)) { | |
1846 | cprc->f2fs_issue_ckpt = NULL; | |
1847 | return -ENOMEM; | |
1848 | } | |
1849 | ||
e6592066 | 1850 | set_task_ioprio(cprc->f2fs_issue_ckpt, cprc->ckpt_thread_ioprio); |
261eeb9c DJ |
1851 | |
1852 | return 0; | |
1853 | } | |
1854 | ||
1855 | void f2fs_stop_ckpt_thread(struct f2fs_sb_info *sbi) | |
1856 | { | |
1857 | struct ckpt_req_control *cprc = &sbi->cprc_info; | |
1858 | ||
1859 | if (cprc->f2fs_issue_ckpt) { | |
1860 | struct task_struct *ckpt_task = cprc->f2fs_issue_ckpt; | |
1861 | ||
1862 | cprc->f2fs_issue_ckpt = NULL; | |
1863 | kthread_stop(ckpt_task); | |
1864 | ||
1865 | flush_remained_ckpt_reqs(sbi, NULL); | |
1866 | } | |
1867 | } | |
1868 | ||
1869 | void f2fs_init_ckpt_req_control(struct f2fs_sb_info *sbi) | |
1870 | { | |
1871 | struct ckpt_req_control *cprc = &sbi->cprc_info; | |
1872 | ||
1873 | atomic_set(&cprc->issued_ckpt, 0); | |
1874 | atomic_set(&cprc->total_ckpt, 0); | |
1875 | atomic_set(&cprc->queued_ckpt, 0); | |
e6592066 | 1876 | cprc->ckpt_thread_ioprio = DEFAULT_CHECKPOINT_IOPRIO; |
261eeb9c DJ |
1877 | init_waitqueue_head(&cprc->ckpt_wait_queue); |
1878 | init_llist_head(&cprc->issue_list); | |
1879 | spin_lock_init(&cprc->stat_lock); | |
1880 | } |