]>
Commit | Line | Data |
---|---|---|
7c1a000d | 1 | // SPDX-License-Identifier: GPL-2.0 |
0a8165d7 | 2 | /* |
d624c96f JK |
3 | * fs/f2fs/recovery.c |
4 | * | |
5 | * Copyright (c) 2012 Samsung Electronics Co., Ltd. | |
6 | * http://www.samsung.com/ | |
d624c96f JK |
7 | */ |
8 | #include <linux/fs.h> | |
9 | #include <linux/f2fs_fs.h> | |
10 | #include "f2fs.h" | |
11 | #include "node.h" | |
12 | #include "segment.h" | |
13 | ||
441ac5cb JK |
14 | /* |
15 | * Roll forward recovery scenarios. | |
16 | * | |
17 | * [Term] F: fsync_mark, D: dentry_mark | |
18 | * | |
19 | * 1. inode(x) | CP | inode(x) | dnode(F) | |
20 | * -> Update the latest inode(x). | |
21 | * | |
22 | * 2. inode(x) | CP | inode(F) | dnode(F) | |
23 | * -> No problem. | |
24 | * | |
25 | * 3. inode(x) | CP | dnode(F) | inode(x) | |
26 | * -> Recover to the latest dnode(F), and drop the last inode(x) | |
27 | * | |
28 | * 4. inode(x) | CP | dnode(F) | inode(F) | |
29 | * -> No problem. | |
30 | * | |
31 | * 5. CP | inode(x) | dnode(F) | |
32 | * -> The inode(DF) was missing. Should drop this dnode(F). | |
33 | * | |
34 | * 6. CP | inode(DF) | dnode(F) | |
35 | * -> No problem. | |
36 | * | |
37 | * 7. CP | dnode(F) | inode(DF) | |
38 | * -> If f2fs_iget fails, then goto next to find inode(DF). | |
39 | * | |
40 | * 8. CP | dnode(F) | inode(x) | |
41 | * -> If f2fs_iget fails, then goto next to find inode(DF). | |
42 | * But it will fail due to no inode(DF). | |
43 | */ | |
44 | ||
d624c96f JK |
45 | static struct kmem_cache *fsync_entry_slab; |
46 | ||
4d57b86d | 47 | bool f2fs_space_for_roll_forward(struct f2fs_sb_info *sbi) |
d624c96f | 48 | { |
41382ec4 JK |
49 | s64 nalloc = percpu_counter_sum_positive(&sbi->alloc_valid_block_count); |
50 | ||
51 | if (sbi->last_valid_block_count + nalloc > sbi->user_block_count) | |
d624c96f JK |
52 | return false; |
53 | return true; | |
54 | } | |
55 | ||
56 | static struct fsync_inode_entry *get_fsync_inode(struct list_head *head, | |
57 | nid_t ino) | |
58 | { | |
d624c96f JK |
59 | struct fsync_inode_entry *entry; |
60 | ||
2d7b822a | 61 | list_for_each_entry(entry, head, list) |
d624c96f JK |
62 | if (entry->inode->i_ino == ino) |
63 | return entry; | |
2d7b822a | 64 | |
d624c96f JK |
65 | return NULL; |
66 | } | |
67 | ||
f4702d61 | 68 | static struct fsync_inode_entry *add_fsync_inode(struct f2fs_sb_info *sbi, |
4b2414d0 | 69 | struct list_head *head, nid_t ino, bool quota_inode) |
3f8ab270 | 70 | { |
e8ea9b3d | 71 | struct inode *inode; |
3f8ab270 | 72 | struct fsync_inode_entry *entry; |
4b2414d0 | 73 | int err; |
3f8ab270 | 74 | |
e8ea9b3d | 75 | inode = f2fs_iget_retry(sbi->sb, ino); |
f4702d61 JK |
76 | if (IS_ERR(inode)) |
77 | return ERR_CAST(inode); | |
78 | ||
4b2414d0 CY |
79 | err = dquot_initialize(inode); |
80 | if (err) | |
81 | goto err_out; | |
82 | ||
83 | if (quota_inode) { | |
84 | err = dquot_alloc_inode(inode); | |
85 | if (err) | |
86 | goto err_out; | |
87 | } | |
88 | ||
e8ea9b3d | 89 | entry = f2fs_kmem_cache_alloc(fsync_entry_slab, GFP_F2FS_ZERO); |
3f8ab270 CY |
90 | entry->inode = inode; |
91 | list_add_tail(&entry->list, head); | |
92 | ||
93 | return entry; | |
4b2414d0 CY |
94 | err_out: |
95 | iput(inode); | |
96 | return ERR_PTR(err); | |
3f8ab270 CY |
97 | } |
98 | ||
26b5a079 | 99 | static void del_fsync_inode(struct fsync_inode_entry *entry, int drop) |
3f8ab270 | 100 | { |
26b5a079 SY |
101 | if (drop) { |
102 | /* inode should not be recovered, drop it */ | |
103 | f2fs_inode_synced(entry->inode); | |
104 | } | |
3f8ab270 CY |
105 | iput(entry->inode); |
106 | list_del(&entry->list); | |
107 | kmem_cache_free(fsync_entry_slab, entry); | |
108 | } | |
109 | ||
f61cce5b CY |
110 | static int recover_dentry(struct inode *inode, struct page *ipage, |
111 | struct list_head *dir_list) | |
d624c96f | 112 | { |
58bfaf44 | 113 | struct f2fs_inode *raw_inode = F2FS_INODE(ipage); |
74d0b917 | 114 | nid_t pino = le32_to_cpu(raw_inode->i_pino); |
6b8213d9 | 115 | struct f2fs_dir_entry *de; |
e7ba108a | 116 | struct fscrypt_name fname; |
d624c96f | 117 | struct page *page; |
6b8213d9 | 118 | struct inode *dir, *einode; |
f61cce5b | 119 | struct fsync_inode_entry *entry; |
d624c96f | 120 | int err = 0; |
e7ba108a | 121 | char *name; |
d624c96f | 122 | |
f61cce5b CY |
123 | entry = get_fsync_inode(dir_list, pino); |
124 | if (!entry) { | |
4b2414d0 CY |
125 | entry = add_fsync_inode(F2FS_I_SB(inode), dir_list, |
126 | pino, false); | |
f4702d61 JK |
127 | if (IS_ERR(entry)) { |
128 | dir = ERR_CAST(entry); | |
129 | err = PTR_ERR(entry); | |
f61cce5b CY |
130 | goto out; |
131 | } | |
ed57c27f JK |
132 | } |
133 | ||
f61cce5b CY |
134 | dir = entry->inode; |
135 | ||
e7ba108a SL |
136 | memset(&fname, 0, sizeof(struct fscrypt_name)); |
137 | fname.disk_name.len = le32_to_cpu(raw_inode->i_namelen); | |
138 | fname.disk_name.name = raw_inode->i_name; | |
d96b1431 | 139 | |
e7ba108a | 140 | if (unlikely(fname.disk_name.len > F2FS_NAME_LEN)) { |
d96b1431 CY |
141 | WARN_ON(1); |
142 | err = -ENAMETOOLONG; | |
f61cce5b | 143 | goto out; |
d96b1431 | 144 | } |
6b8213d9 | 145 | retry: |
e7ba108a | 146 | de = __f2fs_find_entry(dir, &fname, &page); |
418f6c27 | 147 | if (de && inode->i_ino == le32_to_cpu(de->ino)) |
bdbc90fa | 148 | goto out_put; |
418f6c27 | 149 | |
6b8213d9 | 150 | if (de) { |
e8ea9b3d | 151 | einode = f2fs_iget_retry(inode->i_sb, le32_to_cpu(de->ino)); |
6b8213d9 JK |
152 | if (IS_ERR(einode)) { |
153 | WARN_ON(1); | |
5c1f9927 CY |
154 | err = PTR_ERR(einode); |
155 | if (err == -ENOENT) | |
6b8213d9 | 156 | err = -EEXIST; |
bdbc90fa | 157 | goto out_put; |
2e5558f4 | 158 | } |
4b2414d0 CY |
159 | |
160 | err = dquot_initialize(einode); | |
161 | if (err) { | |
162 | iput(einode); | |
bdbc90fa | 163 | goto out_put; |
4b2414d0 CY |
164 | } |
165 | ||
4d57b86d | 166 | err = f2fs_acquire_orphan_inode(F2FS_I_SB(inode)); |
2e5558f4 RK |
167 | if (err) { |
168 | iput(einode); | |
bdbc90fa | 169 | goto out_put; |
6b8213d9 | 170 | } |
dbeacf02 | 171 | f2fs_delete_entry(de, page, dir, einode); |
6b8213d9 JK |
172 | iput(einode); |
173 | goto retry; | |
91246c21 CY |
174 | } else if (IS_ERR(page)) { |
175 | err = PTR_ERR(page); | |
176 | } else { | |
4d57b86d | 177 | err = f2fs_add_dentry(dir, &fname, inode, |
91246c21 | 178 | inode->i_ino, inode->i_mode); |
d624c96f | 179 | } |
e8ea9b3d JK |
180 | if (err == -ENOMEM) |
181 | goto retry; | |
2e5558f4 RK |
182 | goto out; |
183 | ||
bdbc90fa | 184 | out_put: |
2e5558f4 | 185 | f2fs_put_page(page, 0); |
d624c96f | 186 | out: |
e7ba108a SL |
187 | if (file_enc_name(inode)) |
188 | name = "<encrypted>"; | |
189 | else | |
190 | name = raw_inode->i_name; | |
6c311ec6 CF |
191 | f2fs_msg(inode->i_sb, KERN_NOTICE, |
192 | "%s: ino = %x, name = %s, dir = %lx, err = %d", | |
e7ba108a | 193 | __func__, ino_of_node(ipage), name, |
f28c06fa | 194 | IS_ERR(dir) ? 0 : dir->i_ino, err); |
d624c96f JK |
195 | return err; |
196 | } | |
197 | ||
af033b2a CY |
198 | static int recover_quota_data(struct inode *inode, struct page *page) |
199 | { | |
200 | struct f2fs_inode *raw = F2FS_INODE(page); | |
201 | struct iattr attr; | |
202 | uid_t i_uid = le32_to_cpu(raw->i_uid); | |
203 | gid_t i_gid = le32_to_cpu(raw->i_gid); | |
204 | int err; | |
205 | ||
206 | memset(&attr, 0, sizeof(attr)); | |
207 | ||
208 | attr.ia_uid = make_kuid(inode->i_sb->s_user_ns, i_uid); | |
209 | attr.ia_gid = make_kgid(inode->i_sb->s_user_ns, i_gid); | |
210 | ||
211 | if (!uid_eq(attr.ia_uid, inode->i_uid)) | |
212 | attr.ia_valid |= ATTR_UID; | |
213 | if (!gid_eq(attr.ia_gid, inode->i_gid)) | |
214 | attr.ia_valid |= ATTR_GID; | |
215 | ||
216 | if (!attr.ia_valid) | |
217 | return 0; | |
218 | ||
219 | err = dquot_transfer(inode, &attr); | |
220 | if (err) | |
221 | set_sbi_flag(F2FS_I_SB(inode), SBI_QUOTA_NEED_REPAIR); | |
222 | return err; | |
223 | } | |
224 | ||
37a086f0 JK |
225 | static void recover_inline_flags(struct inode *inode, struct f2fs_inode *ri) |
226 | { | |
227 | if (ri->i_inline & F2FS_PIN_FILE) | |
228 | set_inode_flag(inode, FI_PIN_FILE); | |
229 | else | |
230 | clear_inode_flag(inode, FI_PIN_FILE); | |
231 | if (ri->i_inline & F2FS_DATA_EXIST) | |
232 | set_inode_flag(inode, FI_DATA_EXIST); | |
233 | else | |
234 | clear_inode_flag(inode, FI_DATA_EXIST); | |
37a086f0 JK |
235 | } |
236 | ||
af033b2a | 237 | static int recover_inode(struct inode *inode, struct page *page) |
d624c96f | 238 | { |
441ac5cb | 239 | struct f2fs_inode *raw = F2FS_INODE(page); |
e7d55452 | 240 | char *name; |
af033b2a | 241 | int err; |
441ac5cb JK |
242 | |
243 | inode->i_mode = le16_to_cpu(raw->i_mode); | |
af033b2a CY |
244 | |
245 | err = recover_quota_data(inode, page); | |
246 | if (err) | |
247 | return err; | |
248 | ||
dc4cd125 CY |
249 | i_uid_write(inode, le32_to_cpu(raw->i_uid)); |
250 | i_gid_write(inode, le32_to_cpu(raw->i_gid)); | |
f4474aa6 CY |
251 | |
252 | if (raw->i_inline & F2FS_EXTRA_ATTR) { | |
7beb01f7 | 253 | if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)) && |
f4474aa6 CY |
254 | F2FS_FITS_IN_INODE(raw, le16_to_cpu(raw->i_extra_isize), |
255 | i_projid)) { | |
256 | projid_t i_projid; | |
78130819 | 257 | kprojid_t kprojid; |
f4474aa6 CY |
258 | |
259 | i_projid = (projid_t)le32_to_cpu(raw->i_projid); | |
78130819 CY |
260 | kprojid = make_kprojid(&init_user_ns, i_projid); |
261 | ||
262 | if (!projid_eq(kprojid, F2FS_I(inode)->i_projid)) { | |
263 | err = f2fs_transfer_project_quota(inode, | |
264 | kprojid); | |
265 | if (err) | |
266 | return err; | |
267 | F2FS_I(inode)->i_projid = kprojid; | |
268 | } | |
f4474aa6 CY |
269 | } |
270 | } | |
271 | ||
fc9581c8 | 272 | f2fs_i_size_write(inode, le64_to_cpu(raw->i_size)); |
9f0552e0 | 273 | inode->i_atime.tv_sec = le64_to_cpu(raw->i_atime); |
441ac5cb JK |
274 | inode->i_ctime.tv_sec = le64_to_cpu(raw->i_ctime); |
275 | inode->i_mtime.tv_sec = le64_to_cpu(raw->i_mtime); | |
9f0552e0 | 276 | inode->i_atime.tv_nsec = le32_to_cpu(raw->i_atime_nsec); |
441ac5cb JK |
277 | inode->i_ctime.tv_nsec = le32_to_cpu(raw->i_ctime_nsec); |
278 | inode->i_mtime.tv_nsec = le32_to_cpu(raw->i_mtime_nsec); | |
f356fe0c | 279 | |
26787236 | 280 | F2FS_I(inode)->i_advise = raw->i_advise; |
19c73a69 | 281 | F2FS_I(inode)->i_flags = le32_to_cpu(raw->i_flags); |
0c093b59 | 282 | f2fs_set_inode_flags(inode); |
7de36cf3 CY |
283 | F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN] = |
284 | le16_to_cpu(raw->i_gc_failures); | |
26787236 | 285 | |
37a086f0 JK |
286 | recover_inline_flags(inode, raw); |
287 | ||
4a1728ca CY |
288 | f2fs_mark_inode_dirty_sync(inode, true); |
289 | ||
e7d55452 JK |
290 | if (file_enc_name(inode)) |
291 | name = "<encrypted>"; | |
292 | else | |
293 | name = F2FS_INODE(page)->i_name; | |
294 | ||
37a086f0 JK |
295 | f2fs_msg(inode->i_sb, KERN_NOTICE, |
296 | "recover_inode: ino = %x, name = %s, inline = %x", | |
297 | ino_of_node(page), name, raw->i_inline); | |
af033b2a | 298 | return 0; |
d624c96f JK |
299 | } |
300 | ||
d40d30c5 JK |
301 | static int find_fsync_dnodes(struct f2fs_sb_info *sbi, struct list_head *head, |
302 | bool check_only) | |
d624c96f | 303 | { |
d624c96f | 304 | struct curseg_info *curseg; |
4c521f49 | 305 | struct page *page = NULL; |
d624c96f | 306 | block_t blkaddr; |
fb0e72c8 | 307 | unsigned int loop_cnt = 0; |
82902c06 CY |
308 | unsigned int free_blocks = MAIN_SEGS(sbi) * sbi->blocks_per_seg - |
309 | valid_user_blocks(sbi); | |
d624c96f JK |
310 | int err = 0; |
311 | ||
312 | /* get node pages in the current segment */ | |
313 | curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); | |
695fd1ed | 314 | blkaddr = NEXT_FREE_BLKADDR(sbi, curseg); |
d624c96f | 315 | |
d624c96f JK |
316 | while (1) { |
317 | struct fsync_inode_entry *entry; | |
318 | ||
e1da7872 | 319 | if (!f2fs_is_valid_blkaddr(sbi, blkaddr, META_POR)) |
4c521f49 | 320 | return 0; |
d624c96f | 321 | |
4d57b86d | 322 | page = f2fs_get_tmp_page(sbi, blkaddr); |
7735730d CY |
323 | if (IS_ERR(page)) { |
324 | err = PTR_ERR(page); | |
325 | break; | |
326 | } | |
393ff91f | 327 | |
a468f0ef | 328 | if (!is_recoverable_dnode(page)) |
f356fe0c | 329 | break; |
d624c96f JK |
330 | |
331 | if (!is_fsync_dnode(page)) | |
332 | goto next; | |
333 | ||
334 | entry = get_fsync_inode(head, ino_of_node(page)); | |
d47b8715 | 335 | if (!entry) { |
4b2414d0 CY |
336 | bool quota_inode = false; |
337 | ||
d40d30c5 JK |
338 | if (!check_only && |
339 | IS_INODE(page) && is_dent_dnode(page)) { | |
4d57b86d | 340 | err = f2fs_recover_inode_page(sbi, page); |
6ead1142 | 341 | if (err) |
f356fe0c | 342 | break; |
4b2414d0 | 343 | quota_inode = true; |
d624c96f JK |
344 | } |
345 | ||
441ac5cb JK |
346 | /* |
347 | * CP | dnode(F) | inode(DF) | |
348 | * For this case, we should not give up now. | |
349 | */ | |
4b2414d0 CY |
350 | entry = add_fsync_inode(sbi, head, ino_of_node(page), |
351 | quota_inode); | |
f4702d61 JK |
352 | if (IS_ERR(entry)) { |
353 | err = PTR_ERR(entry); | |
8fbc418f JK |
354 | if (err == -ENOENT) { |
355 | err = 0; | |
441ac5cb | 356 | goto next; |
8fbc418f | 357 | } |
f356fe0c | 358 | break; |
d624c96f | 359 | } |
d624c96f | 360 | } |
addbe45b JK |
361 | entry->blkaddr = blkaddr; |
362 | ||
608514de JK |
363 | if (IS_INODE(page) && is_dent_dnode(page)) |
364 | entry->last_dentry = blkaddr; | |
d624c96f | 365 | next: |
fb0e72c8 CY |
366 | /* sanity check in order to detect looped node chain */ |
367 | if (++loop_cnt >= free_blocks || | |
368 | blkaddr == next_blkaddr_of_node(page)) { | |
369 | f2fs_msg(sbi->sb, KERN_NOTICE, | |
370 | "%s: detect looped node chain, " | |
371 | "blkaddr:%u, next:%u", | |
372 | __func__, blkaddr, next_blkaddr_of_node(page)); | |
373 | err = -EINVAL; | |
374 | break; | |
375 | } | |
376 | ||
d624c96f JK |
377 | /* check next segment */ |
378 | blkaddr = next_blkaddr_of_node(page); | |
4c521f49 | 379 | f2fs_put_page(page, 1); |
635aee1f | 380 | |
4d57b86d | 381 | f2fs_ra_meta_pages_cond(sbi, blkaddr); |
d624c96f | 382 | } |
4c521f49 | 383 | f2fs_put_page(page, 1); |
d624c96f JK |
384 | return err; |
385 | } | |
386 | ||
26b5a079 | 387 | static void destroy_fsync_dnodes(struct list_head *head, int drop) |
d624c96f | 388 | { |
d8b79b2f DC |
389 | struct fsync_inode_entry *entry, *tmp; |
390 | ||
3f8ab270 | 391 | list_for_each_entry_safe(entry, tmp, head, list) |
26b5a079 | 392 | del_fsync_inode(entry, drop); |
d624c96f JK |
393 | } |
394 | ||
39cf72cf | 395 | static int check_index_in_prev_nodes(struct f2fs_sb_info *sbi, |
b292dcab | 396 | block_t blkaddr, struct dnode_of_data *dn) |
d624c96f JK |
397 | { |
398 | struct seg_entry *sentry; | |
399 | unsigned int segno = GET_SEGNO(sbi, blkaddr); | |
491c0854 | 400 | unsigned short blkoff = GET_BLKOFF_FROM_SEG0(sbi, blkaddr); |
f6517cfc | 401 | struct f2fs_summary_block *sum_node; |
d624c96f | 402 | struct f2fs_summary sum; |
f6517cfc | 403 | struct page *sum_page, *node_page; |
c9ef4810 | 404 | struct dnode_of_data tdn = *dn; |
b292dcab | 405 | nid_t ino, nid; |
d624c96f | 406 | struct inode *inode; |
de93653f | 407 | unsigned int offset; |
d624c96f JK |
408 | block_t bidx; |
409 | int i; | |
410 | ||
411 | sentry = get_seg_entry(sbi, segno); | |
412 | if (!f2fs_test_bit(blkoff, sentry->cur_valid_map)) | |
39cf72cf | 413 | return 0; |
d624c96f JK |
414 | |
415 | /* Get the previous summary */ | |
125c9fb1 | 416 | for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) { |
d624c96f JK |
417 | struct curseg_info *curseg = CURSEG_I(sbi, i); |
418 | if (curseg->segno == segno) { | |
419 | sum = curseg->sum_blk->entries[blkoff]; | |
f6517cfc | 420 | goto got_it; |
d624c96f JK |
421 | } |
422 | } | |
d624c96f | 423 | |
4d57b86d | 424 | sum_page = f2fs_get_sum_page(sbi, segno); |
edc55aaf JK |
425 | if (IS_ERR(sum_page)) |
426 | return PTR_ERR(sum_page); | |
f6517cfc JK |
427 | sum_node = (struct f2fs_summary_block *)page_address(sum_page); |
428 | sum = sum_node->entries[blkoff]; | |
429 | f2fs_put_page(sum_page, 1); | |
430 | got_it: | |
b292dcab JK |
431 | /* Use the locked dnode page and inode */ |
432 | nid = le32_to_cpu(sum.nid); | |
433 | if (dn->inode->i_ino == nid) { | |
b292dcab | 434 | tdn.nid = nid; |
c9ef4810 JK |
435 | if (!dn->inode_page_locked) |
436 | lock_page(dn->inode_page); | |
b292dcab | 437 | tdn.node_page = dn->inode_page; |
060dd67b | 438 | tdn.ofs_in_node = le16_to_cpu(sum.ofs_in_node); |
c9ef4810 | 439 | goto truncate_out; |
b292dcab | 440 | } else if (dn->nid == nid) { |
060dd67b | 441 | tdn.ofs_in_node = le16_to_cpu(sum.ofs_in_node); |
c9ef4810 | 442 | goto truncate_out; |
b292dcab JK |
443 | } |
444 | ||
d624c96f | 445 | /* Get the node page */ |
4d57b86d | 446 | node_page = f2fs_get_node_page(sbi, nid); |
39cf72cf JK |
447 | if (IS_ERR(node_page)) |
448 | return PTR_ERR(node_page); | |
de93653f JK |
449 | |
450 | offset = ofs_of_node(node_page); | |
d624c96f JK |
451 | ino = ino_of_node(node_page); |
452 | f2fs_put_page(node_page, 1); | |
453 | ||
60979115 | 454 | if (ino != dn->inode->i_ino) { |
4b2414d0 CY |
455 | int ret; |
456 | ||
60979115 | 457 | /* Deallocate previous index in the node page */ |
e8ea9b3d | 458 | inode = f2fs_iget_retry(sbi->sb, ino); |
60979115 JK |
459 | if (IS_ERR(inode)) |
460 | return PTR_ERR(inode); | |
4b2414d0 CY |
461 | |
462 | ret = dquot_initialize(inode); | |
463 | if (ret) { | |
464 | iput(inode); | |
465 | return ret; | |
466 | } | |
60979115 JK |
467 | } else { |
468 | inode = dn->inode; | |
469 | } | |
06025f4d | 470 | |
4d57b86d CY |
471 | bidx = f2fs_start_bidx_of_node(offset, inode) + |
472 | le16_to_cpu(sum.ofs_in_node); | |
de93653f | 473 | |
c9ef4810 JK |
474 | /* |
475 | * if inode page is locked, unlock temporarily, but its reference | |
476 | * count keeps alive. | |
477 | */ | |
478 | if (ino == dn->inode->i_ino && dn->inode_page_locked) | |
479 | unlock_page(dn->inode_page); | |
480 | ||
481 | set_new_dnode(&tdn, inode, NULL, NULL, 0); | |
4d57b86d | 482 | if (f2fs_get_dnode_of_data(&tdn, bidx, LOOKUP_NODE)) |
c9ef4810 JK |
483 | goto out; |
484 | ||
485 | if (tdn.data_blkaddr == blkaddr) | |
4d57b86d | 486 | f2fs_truncate_data_blocks_range(&tdn, 1); |
c9ef4810 JK |
487 | |
488 | f2fs_put_dnode(&tdn); | |
489 | out: | |
490 | if (ino != dn->inode->i_ino) | |
60979115 | 491 | iput(inode); |
c9ef4810 JK |
492 | else if (dn->inode_page_locked) |
493 | lock_page(dn->inode_page); | |
494 | return 0; | |
495 | ||
496 | truncate_out: | |
7a2af766 CY |
497 | if (datablock_addr(tdn.inode, tdn.node_page, |
498 | tdn.ofs_in_node) == blkaddr) | |
4d57b86d | 499 | f2fs_truncate_data_blocks_range(&tdn, 1); |
c9ef4810 JK |
500 | if (dn->inode->i_ino == nid && !dn->inode_page_locked) |
501 | unlock_page(dn->inode_page); | |
39cf72cf | 502 | return 0; |
d624c96f JK |
503 | } |
504 | ||
6ead1142 | 505 | static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode, |
e17d488b | 506 | struct page *page) |
d624c96f | 507 | { |
d624c96f | 508 | struct dnode_of_data dn; |
d624c96f | 509 | struct node_info ni; |
81ca7350 | 510 | unsigned int start, end; |
f356fe0c | 511 | int err = 0, recovered = 0; |
d624c96f | 512 | |
1c35a90e JK |
513 | /* step 1: recover xattr */ |
514 | if (IS_INODE(page)) { | |
4d57b86d | 515 | f2fs_recover_inline_xattr(inode, page); |
1c35a90e | 516 | } else if (f2fs_has_xattr_block(ofs_of_node(page))) { |
4d57b86d | 517 | err = f2fs_recover_xattr_data(inode, page); |
d260081c CY |
518 | if (!err) |
519 | recovered++; | |
1e1bb4ba | 520 | goto out; |
1c35a90e | 521 | } |
1e1bb4ba | 522 | |
1c35a90e | 523 | /* step 2: recover inline data */ |
4d57b86d | 524 | if (f2fs_recover_inline_data(inode, page)) |
abb2366c JK |
525 | goto out; |
526 | ||
1c35a90e | 527 | /* step 3: recover data indices */ |
4d57b86d | 528 | start = f2fs_start_bidx_of_node(ofs_of_node(page), inode); |
81ca7350 | 529 | end = start + ADDRS_PER_PAGE(page, inode); |
d624c96f JK |
530 | |
531 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
e8ea9b3d | 532 | retry_dn: |
4d57b86d | 533 | err = f2fs_get_dnode_of_data(&dn, start, ALLOC_NODE); |
e8ea9b3d JK |
534 | if (err) { |
535 | if (err == -ENOMEM) { | |
536 | congestion_wait(BLK_RW_ASYNC, HZ/50); | |
537 | goto retry_dn; | |
538 | } | |
1e1bb4ba | 539 | goto out; |
e8ea9b3d | 540 | } |
d624c96f | 541 | |
bae0ee7a | 542 | f2fs_wait_on_page_writeback(dn.node_page, NODE, true, true); |
d624c96f | 543 | |
7735730d CY |
544 | err = f2fs_get_node_info(sbi, dn.nid, &ni); |
545 | if (err) | |
546 | goto err; | |
547 | ||
9850cf4a JK |
548 | f2fs_bug_on(sbi, ni.ino != ino_of_node(page)); |
549 | f2fs_bug_on(sbi, ofs_of_node(dn.node_page) != ofs_of_node(page)); | |
d624c96f | 550 | |
12a8343e | 551 | for (; start < end; start++, dn.ofs_in_node++) { |
d624c96f JK |
552 | block_t src, dest; |
553 | ||
7a2af766 CY |
554 | src = datablock_addr(dn.inode, dn.node_page, dn.ofs_in_node); |
555 | dest = datablock_addr(dn.inode, page, dn.ofs_in_node); | |
d624c96f | 556 | |
12a8343e CY |
557 | /* skip recovering if dest is the same as src */ |
558 | if (src == dest) | |
559 | continue; | |
560 | ||
561 | /* dest is invalid, just invalidate src block */ | |
562 | if (dest == NULL_ADDR) { | |
4d57b86d | 563 | f2fs_truncate_data_blocks_range(&dn, 1); |
12a8343e CY |
564 | continue; |
565 | } | |
566 | ||
26787236 | 567 | if (!file_keep_isize(inode) && |
dba79f38 CY |
568 | (i_size_read(inode) <= ((loff_t)start << PAGE_SHIFT))) |
569 | f2fs_i_size_write(inode, | |
570 | (loff_t)(start + 1) << PAGE_SHIFT); | |
26de9b11 | 571 | |
12a8343e CY |
572 | /* |
573 | * dest is reserved block, invalidate src block | |
574 | * and then reserve one new block in dnode page. | |
575 | */ | |
576 | if (dest == NEW_ADDR) { | |
4d57b86d CY |
577 | f2fs_truncate_data_blocks_range(&dn, 1); |
578 | f2fs_reserve_new_block(&dn); | |
12a8343e CY |
579 | continue; |
580 | } | |
581 | ||
582 | /* dest is valid block, try to recover from src to dest */ | |
e1da7872 | 583 | if (f2fs_is_valid_blkaddr(sbi, dest, META_POR)) { |
e03b07d9 | 584 | |
d624c96f | 585 | if (src == NULL_ADDR) { |
4d57b86d | 586 | err = f2fs_reserve_new_block(&dn); |
7fa750a1 AB |
587 | while (err && |
588 | IS_ENABLED(CONFIG_F2FS_FAULT_INJECTION)) | |
4d57b86d | 589 | err = f2fs_reserve_new_block(&dn); |
d624c96f | 590 | /* We should not get -ENOSPC */ |
9850cf4a | 591 | f2fs_bug_on(sbi, err); |
6f3ec995 JK |
592 | if (err) |
593 | goto err; | |
d624c96f | 594 | } |
e8ea9b3d | 595 | retry_prev: |
d624c96f | 596 | /* Check the previous node page having this index */ |
39cf72cf | 597 | err = check_index_in_prev_nodes(sbi, dest, &dn); |
e8ea9b3d JK |
598 | if (err) { |
599 | if (err == -ENOMEM) { | |
600 | congestion_wait(BLK_RW_ASYNC, HZ/50); | |
601 | goto retry_prev; | |
602 | } | |
39cf72cf | 603 | goto err; |
e8ea9b3d | 604 | } |
d624c96f | 605 | |
d624c96f | 606 | /* write dummy data page */ |
528e3459 | 607 | f2fs_replace_block(sbi, &dn, src, dest, |
28bc106b | 608 | ni.version, false, false); |
f356fe0c | 609 | recovered++; |
d624c96f | 610 | } |
d624c96f JK |
611 | } |
612 | ||
d624c96f JK |
613 | copy_node_footer(dn.node_page, page); |
614 | fill_node_footer(dn.node_page, dn.nid, ni.ino, | |
615 | ofs_of_node(page), false); | |
616 | set_page_dirty(dn.node_page); | |
39cf72cf | 617 | err: |
d624c96f | 618 | f2fs_put_dnode(&dn); |
1e1bb4ba | 619 | out: |
6c311ec6 | 620 | f2fs_msg(sbi->sb, KERN_NOTICE, |
26787236 JK |
621 | "recover_data: ino = %lx (i_size: %s) recovered = %d, err = %d", |
622 | inode->i_ino, | |
623 | file_keep_isize(inode) ? "keep" : "recover", | |
624 | recovered, err); | |
39cf72cf | 625 | return err; |
d624c96f JK |
626 | } |
627 | ||
f61cce5b | 628 | static int recover_data(struct f2fs_sb_info *sbi, struct list_head *inode_list, |
26b5a079 | 629 | struct list_head *tmp_inode_list, struct list_head *dir_list) |
d624c96f | 630 | { |
d624c96f | 631 | struct curseg_info *curseg; |
4c521f49 | 632 | struct page *page = NULL; |
6ead1142 | 633 | int err = 0; |
d624c96f JK |
634 | block_t blkaddr; |
635 | ||
636 | /* get node pages in the current segment */ | |
b7973f23 | 637 | curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); |
d624c96f JK |
638 | blkaddr = NEXT_FREE_BLKADDR(sbi, curseg); |
639 | ||
d624c96f JK |
640 | while (1) { |
641 | struct fsync_inode_entry *entry; | |
642 | ||
e1da7872 | 643 | if (!f2fs_is_valid_blkaddr(sbi, blkaddr, META_POR)) |
4c521f49 | 644 | break; |
d624c96f | 645 | |
4d57b86d | 646 | f2fs_ra_meta_pages_cond(sbi, blkaddr); |
635aee1f | 647 | |
4d57b86d | 648 | page = f2fs_get_tmp_page(sbi, blkaddr); |
7735730d CY |
649 | if (IS_ERR(page)) { |
650 | err = PTR_ERR(page); | |
651 | break; | |
652 | } | |
393ff91f | 653 | |
a468f0ef | 654 | if (!is_recoverable_dnode(page)) { |
4c521f49 | 655 | f2fs_put_page(page, 1); |
45856aff | 656 | break; |
4c521f49 | 657 | } |
d624c96f | 658 | |
f61cce5b | 659 | entry = get_fsync_inode(inode_list, ino_of_node(page)); |
d624c96f JK |
660 | if (!entry) |
661 | goto next; | |
441ac5cb JK |
662 | /* |
663 | * inode(x) | CP | inode(x) | dnode(F) | |
664 | * In this case, we can lose the latest inode(x). | |
c52e1b10 | 665 | * So, call recover_inode for the inode update. |
441ac5cb | 666 | */ |
af033b2a CY |
667 | if (IS_INODE(page)) { |
668 | err = recover_inode(entry->inode, page); | |
669 | if (err) | |
670 | break; | |
671 | } | |
c52e1b10 | 672 | if (entry->last_dentry == blkaddr) { |
f61cce5b | 673 | err = recover_dentry(entry->inode, page, dir_list); |
c52e1b10 JK |
674 | if (err) { |
675 | f2fs_put_page(page, 1); | |
676 | break; | |
677 | } | |
678 | } | |
e17d488b | 679 | err = do_recover_data(sbi, entry->inode, page); |
4c521f49 JK |
680 | if (err) { |
681 | f2fs_put_page(page, 1); | |
45856aff | 682 | break; |
4c521f49 | 683 | } |
d624c96f | 684 | |
3f8ab270 | 685 | if (entry->blkaddr == blkaddr) |
26b5a079 | 686 | list_move_tail(&entry->list, tmp_inode_list); |
d624c96f JK |
687 | next: |
688 | /* check next segment */ | |
689 | blkaddr = next_blkaddr_of_node(page); | |
4c521f49 | 690 | f2fs_put_page(page, 1); |
d624c96f | 691 | } |
6ead1142 | 692 | if (!err) |
4d57b86d | 693 | f2fs_allocate_new_segments(sbi); |
6ead1142 | 694 | return err; |
d624c96f JK |
695 | } |
696 | ||
4d57b86d | 697 | int f2fs_recover_fsync_data(struct f2fs_sb_info *sbi, bool check_only) |
d624c96f | 698 | { |
26b5a079 | 699 | struct list_head inode_list, tmp_inode_list; |
f61cce5b | 700 | struct list_head dir_list; |
6ead1142 | 701 | int err; |
6781eabb | 702 | int ret = 0; |
4b2414d0 | 703 | unsigned long s_flags = sbi->sb->s_flags; |
aabe5136 | 704 | bool need_writecp = false; |
ea676733 JK |
705 | #ifdef CONFIG_QUOTA |
706 | int quota_enabled; | |
707 | #endif | |
d624c96f | 708 | |
1751e8a6 | 709 | if (s_flags & SB_RDONLY) { |
e6b0b159 YH |
710 | f2fs_msg(sbi->sb, KERN_INFO, |
711 | "recover fsync data on readonly fs"); | |
1751e8a6 | 712 | sbi->sb->s_flags &= ~SB_RDONLY; |
4b2414d0 CY |
713 | } |
714 | ||
715 | #ifdef CONFIG_QUOTA | |
716 | /* Needed for iput() to work correctly and not trash data */ | |
1751e8a6 | 717 | sbi->sb->s_flags |= SB_ACTIVE; |
4b2414d0 | 718 | /* Turn on quotas so that they are updated correctly */ |
1751e8a6 | 719 | quota_enabled = f2fs_enable_quota_files(sbi, s_flags & SB_RDONLY); |
4b2414d0 CY |
720 | #endif |
721 | ||
d624c96f | 722 | fsync_entry_slab = f2fs_kmem_cache_create("f2fs_fsync_inode_entry", |
e8512d2e | 723 | sizeof(struct fsync_inode_entry)); |
4b2414d0 CY |
724 | if (!fsync_entry_slab) { |
725 | err = -ENOMEM; | |
726 | goto out; | |
727 | } | |
d624c96f JK |
728 | |
729 | INIT_LIST_HEAD(&inode_list); | |
26b5a079 | 730 | INIT_LIST_HEAD(&tmp_inode_list); |
f61cce5b | 731 | INIT_LIST_HEAD(&dir_list); |
d624c96f | 732 | |
14f4e690 JK |
733 | /* prevent checkpoint */ |
734 | mutex_lock(&sbi->cp_mutex); | |
735 | ||
315df839 | 736 | /* step #1: find fsynced inode numbers */ |
d40d30c5 | 737 | err = find_fsync_dnodes(sbi, &inode_list, check_only); |
6781eabb | 738 | if (err || list_empty(&inode_list)) |
4b2414d0 | 739 | goto skip; |
d624c96f | 740 | |
6781eabb JK |
741 | if (check_only) { |
742 | ret = 1; | |
4b2414d0 | 743 | goto skip; |
6781eabb | 744 | } |
d624c96f | 745 | |
aabe5136 | 746 | need_writecp = true; |
691c6fd2 | 747 | |
d624c96f | 748 | /* step #2: recover data */ |
26b5a079 | 749 | err = recover_data(sbi, &inode_list, &tmp_inode_list, &dir_list); |
b307384e | 750 | if (!err) |
9850cf4a | 751 | f2fs_bug_on(sbi, !list_empty(&inode_list)); |
26b5a079 SY |
752 | else { |
753 | /* restore s_flags to let iput() trash data */ | |
754 | sbi->sb->s_flags = s_flags; | |
755 | } | |
4b2414d0 | 756 | skip: |
26b5a079 SY |
757 | destroy_fsync_dnodes(&inode_list, err); |
758 | destroy_fsync_dnodes(&tmp_inode_list, err); | |
cf2271e7 | 759 | |
4c521f49 JK |
760 | /* truncate meta pages to be used by the recovery */ |
761 | truncate_inode_pages_range(META_MAPPING(sbi), | |
09cbfeaf | 762 | (loff_t)MAIN_BLKADDR(sbi) << PAGE_SHIFT, -1); |
4c521f49 | 763 | |
cf2271e7 JK |
764 | if (err) { |
765 | truncate_inode_pages_final(NODE_MAPPING(sbi)); | |
766 | truncate_inode_pages_final(META_MAPPING(sbi)); | |
26b5a079 SY |
767 | } else { |
768 | clear_sbi_flag(sbi, SBI_POR_DOING); | |
cf2271e7 | 769 | } |
a468f0ef JK |
770 | mutex_unlock(&sbi->cp_mutex); |
771 | ||
9e1e6df4 | 772 | /* let's drop all the directory inodes for clean checkpoint */ |
26b5a079 | 773 | destroy_fsync_dnodes(&dir_list, err); |
9e1e6df4 | 774 | |
1378752b CY |
775 | if (need_writecp) { |
776 | set_sbi_flag(sbi, SBI_IS_RECOVERED); | |
777 | ||
778 | if (!err) { | |
779 | struct cp_control cpc = { | |
780 | .reason = CP_RECOVERY, | |
781 | }; | |
782 | err = f2fs_write_checkpoint(sbi, &cpc); | |
783 | } | |
cf2271e7 | 784 | } |
f61cce5b | 785 | |
f61cce5b | 786 | kmem_cache_destroy(fsync_entry_slab); |
4b2414d0 CY |
787 | out: |
788 | #ifdef CONFIG_QUOTA | |
789 | /* Turn quotas off */ | |
ea676733 JK |
790 | if (quota_enabled) |
791 | f2fs_quota_off_umount(sbi->sb); | |
4b2414d0 | 792 | #endif |
1751e8a6 | 793 | sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ |
4b2414d0 | 794 | |
6781eabb | 795 | return ret ? ret: err; |
d624c96f | 796 | } |