]>
Commit | Line | Data |
---|---|---|
0a8165d7 | 1 | /* |
d624c96f JK |
2 | * fs/f2fs/recovery.c |
3 | * | |
4 | * Copyright (c) 2012 Samsung Electronics Co., Ltd. | |
5 | * http://www.samsung.com/ | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License version 2 as | |
9 | * published by the Free Software Foundation. | |
10 | */ | |
11 | #include <linux/fs.h> | |
12 | #include <linux/f2fs_fs.h> | |
13 | #include "f2fs.h" | |
14 | #include "node.h" | |
15 | #include "segment.h" | |
16 | ||
17 | static struct kmem_cache *fsync_entry_slab; | |
18 | ||
19 | bool space_for_roll_forward(struct f2fs_sb_info *sbi) | |
20 | { | |
21 | if (sbi->last_valid_block_count + sbi->alloc_valid_block_count | |
22 | > sbi->user_block_count) | |
23 | return false; | |
24 | return true; | |
25 | } | |
26 | ||
27 | static struct fsync_inode_entry *get_fsync_inode(struct list_head *head, | |
28 | nid_t ino) | |
29 | { | |
30 | struct list_head *this; | |
31 | struct fsync_inode_entry *entry; | |
32 | ||
33 | list_for_each(this, head) { | |
34 | entry = list_entry(this, struct fsync_inode_entry, list); | |
35 | if (entry->inode->i_ino == ino) | |
36 | return entry; | |
37 | } | |
38 | return NULL; | |
39 | } | |
40 | ||
41 | static int recover_dentry(struct page *ipage, struct inode *inode) | |
42 | { | |
58bfaf44 | 43 | struct f2fs_inode *raw_inode = F2FS_INODE(ipage); |
74d0b917 | 44 | nid_t pino = le32_to_cpu(raw_inode->i_pino); |
6b8213d9 | 45 | struct f2fs_dir_entry *de; |
b7f7a5e0 | 46 | struct qstr name; |
d624c96f | 47 | struct page *page; |
6b8213d9 | 48 | struct inode *dir, *einode; |
d624c96f JK |
49 | int err = 0; |
50 | ||
74d0b917 JK |
51 | dir = check_dirty_dir_inode(F2FS_SB(inode->i_sb), pino); |
52 | if (!dir) { | |
53 | dir = f2fs_iget(inode->i_sb, pino); | |
54 | if (IS_ERR(dir)) { | |
55 | err = PTR_ERR(dir); | |
56 | goto out; | |
57 | } | |
58 | set_inode_flag(F2FS_I(dir), FI_DELAY_IPUT); | |
5deb8267 | 59 | add_dirty_dir_inode(dir); |
d624c96f JK |
60 | } |
61 | ||
b7f7a5e0 AV |
62 | name.len = le32_to_cpu(raw_inode->i_namelen); |
63 | name.name = raw_inode->i_name; | |
d96b1431 CY |
64 | |
65 | if (unlikely(name.len > F2FS_NAME_LEN)) { | |
66 | WARN_ON(1); | |
67 | err = -ENAMETOOLONG; | |
68 | goto out; | |
69 | } | |
6b8213d9 JK |
70 | retry: |
71 | de = f2fs_find_entry(dir, &name, &page); | |
2e5558f4 RK |
72 | if (de && inode->i_ino == le32_to_cpu(de->ino)) |
73 | goto out_unmap_put; | |
6b8213d9 JK |
74 | if (de) { |
75 | einode = f2fs_iget(inode->i_sb, le32_to_cpu(de->ino)); | |
76 | if (IS_ERR(einode)) { | |
77 | WARN_ON(1); | |
78 | if (PTR_ERR(einode) == -ENOENT) | |
79 | err = -EEXIST; | |
2e5558f4 RK |
80 | goto out_unmap_put; |
81 | } | |
82 | err = acquire_orphan_inode(F2FS_SB(inode->i_sb)); | |
83 | if (err) { | |
84 | iput(einode); | |
85 | goto out_unmap_put; | |
6b8213d9 JK |
86 | } |
87 | f2fs_delete_entry(de, page, einode); | |
88 | iput(einode); | |
89 | goto retry; | |
d624c96f | 90 | } |
6b8213d9 | 91 | err = __f2fs_add_link(dir, &name, inode); |
2e5558f4 RK |
92 | goto out; |
93 | ||
94 | out_unmap_put: | |
95 | kunmap(page); | |
96 | f2fs_put_page(page, 0); | |
d624c96f | 97 | out: |
f356fe0c JK |
98 | f2fs_msg(inode->i_sb, KERN_NOTICE, "recover_inode and its dentry: " |
99 | "ino = %x, name = %s, dir = %lx, err = %d", | |
f28c06fa DC |
100 | ino_of_node(ipage), raw_inode->i_name, |
101 | IS_ERR(dir) ? 0 : dir->i_ino, err); | |
d624c96f JK |
102 | return err; |
103 | } | |
104 | ||
105 | static int recover_inode(struct inode *inode, struct page *node_page) | |
106 | { | |
58bfaf44 | 107 | struct f2fs_inode *raw_inode = F2FS_INODE(node_page); |
d624c96f | 108 | |
f356fe0c JK |
109 | if (!IS_INODE(node_page)) |
110 | return 0; | |
111 | ||
25ca923b | 112 | inode->i_mode = le16_to_cpu(raw_inode->i_mode); |
d624c96f JK |
113 | i_size_write(inode, le64_to_cpu(raw_inode->i_size)); |
114 | inode->i_atime.tv_sec = le64_to_cpu(raw_inode->i_mtime); | |
115 | inode->i_ctime.tv_sec = le64_to_cpu(raw_inode->i_ctime); | |
116 | inode->i_mtime.tv_sec = le64_to_cpu(raw_inode->i_mtime); | |
117 | inode->i_atime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); | |
118 | inode->i_ctime.tv_nsec = le32_to_cpu(raw_inode->i_ctime_nsec); | |
119 | inode->i_mtime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); | |
120 | ||
f356fe0c JK |
121 | if (is_dent_dnode(node_page)) |
122 | return recover_dentry(node_page, inode); | |
123 | ||
124 | f2fs_msg(inode->i_sb, KERN_NOTICE, "recover_inode: ino = %x, name = %s", | |
125 | ino_of_node(node_page), raw_inode->i_name); | |
126 | return 0; | |
d624c96f JK |
127 | } |
128 | ||
129 | static int find_fsync_dnodes(struct f2fs_sb_info *sbi, struct list_head *head) | |
130 | { | |
d71b5564 | 131 | unsigned long long cp_ver = cur_cp_version(F2FS_CKPT(sbi)); |
d624c96f JK |
132 | struct curseg_info *curseg; |
133 | struct page *page; | |
134 | block_t blkaddr; | |
135 | int err = 0; | |
136 | ||
137 | /* get node pages in the current segment */ | |
138 | curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); | |
139 | blkaddr = START_BLOCK(sbi, curseg->segno) + curseg->next_blkoff; | |
140 | ||
141 | /* read node page */ | |
142 | page = alloc_page(GFP_F2FS_ZERO); | |
e27dae4d DC |
143 | if (!page) |
144 | return -ENOMEM; | |
d624c96f JK |
145 | lock_page(page); |
146 | ||
147 | while (1) { | |
148 | struct fsync_inode_entry *entry; | |
149 | ||
93dfe2ac | 150 | err = f2fs_submit_page_bio(sbi, page, blkaddr, READ_SYNC); |
393ff91f | 151 | if (err) |
b9987a27 | 152 | return err; |
d624c96f | 153 | |
393ff91f JK |
154 | lock_page(page); |
155 | ||
6ead1142 | 156 | if (cp_ver != cpver_of_node(page)) |
f356fe0c | 157 | break; |
d624c96f JK |
158 | |
159 | if (!is_fsync_dnode(page)) | |
160 | goto next; | |
161 | ||
162 | entry = get_fsync_inode(head, ino_of_node(page)); | |
163 | if (entry) { | |
d624c96f JK |
164 | if (IS_INODE(page) && is_dent_dnode(page)) |
165 | set_inode_flag(F2FS_I(entry->inode), | |
166 | FI_INC_LINK); | |
167 | } else { | |
168 | if (IS_INODE(page) && is_dent_dnode(page)) { | |
6ead1142 JK |
169 | err = recover_inode_page(sbi, page); |
170 | if (err) | |
f356fe0c | 171 | break; |
d624c96f JK |
172 | } |
173 | ||
174 | /* add this fsync inode to the list */ | |
175 | entry = kmem_cache_alloc(fsync_entry_slab, GFP_NOFS); | |
176 | if (!entry) { | |
177 | err = -ENOMEM; | |
f356fe0c | 178 | break; |
d624c96f JK |
179 | } |
180 | ||
d624c96f JK |
181 | entry->inode = f2fs_iget(sbi->sb, ino_of_node(page)); |
182 | if (IS_ERR(entry->inode)) { | |
183 | err = PTR_ERR(entry->inode); | |
fd8bb65f | 184 | kmem_cache_free(fsync_entry_slab, entry); |
f356fe0c | 185 | break; |
d624c96f | 186 | } |
fd8bb65f | 187 | list_add_tail(&entry->list, head); |
d624c96f | 188 | } |
addbe45b JK |
189 | entry->blkaddr = blkaddr; |
190 | ||
f356fe0c JK |
191 | err = recover_inode(entry->inode, page); |
192 | if (err && err != -ENOENT) | |
193 | break; | |
d624c96f JK |
194 | next: |
195 | /* check next segment */ | |
196 | blkaddr = next_blkaddr_of_node(page); | |
d624c96f | 197 | } |
b9987a27 | 198 | |
d624c96f JK |
199 | unlock_page(page); |
200 | __free_pages(page, 0); | |
b9987a27 | 201 | |
d624c96f JK |
202 | return err; |
203 | } | |
204 | ||
5ebefc5b | 205 | static void destroy_fsync_dnodes(struct list_head *head) |
d624c96f | 206 | { |
d8b79b2f DC |
207 | struct fsync_inode_entry *entry, *tmp; |
208 | ||
209 | list_for_each_entry_safe(entry, tmp, head, list) { | |
d624c96f JK |
210 | iput(entry->inode); |
211 | list_del(&entry->list); | |
212 | kmem_cache_free(fsync_entry_slab, entry); | |
213 | } | |
214 | } | |
215 | ||
39cf72cf | 216 | static int check_index_in_prev_nodes(struct f2fs_sb_info *sbi, |
b292dcab | 217 | block_t blkaddr, struct dnode_of_data *dn) |
d624c96f JK |
218 | { |
219 | struct seg_entry *sentry; | |
220 | unsigned int segno = GET_SEGNO(sbi, blkaddr); | |
221 | unsigned short blkoff = GET_SEGOFF_FROM_SEG0(sbi, blkaddr) & | |
222 | (sbi->blocks_per_seg - 1); | |
223 | struct f2fs_summary sum; | |
b292dcab | 224 | nid_t ino, nid; |
d624c96f JK |
225 | void *kaddr; |
226 | struct inode *inode; | |
227 | struct page *node_page; | |
de93653f | 228 | unsigned int offset; |
d624c96f JK |
229 | block_t bidx; |
230 | int i; | |
231 | ||
232 | sentry = get_seg_entry(sbi, segno); | |
233 | if (!f2fs_test_bit(blkoff, sentry->cur_valid_map)) | |
39cf72cf | 234 | return 0; |
d624c96f JK |
235 | |
236 | /* Get the previous summary */ | |
237 | for (i = CURSEG_WARM_DATA; i <= CURSEG_COLD_DATA; i++) { | |
238 | struct curseg_info *curseg = CURSEG_I(sbi, i); | |
239 | if (curseg->segno == segno) { | |
240 | sum = curseg->sum_blk->entries[blkoff]; | |
241 | break; | |
242 | } | |
243 | } | |
244 | if (i > CURSEG_COLD_DATA) { | |
245 | struct page *sum_page = get_sum_page(sbi, segno); | |
246 | struct f2fs_summary_block *sum_node; | |
247 | kaddr = page_address(sum_page); | |
248 | sum_node = (struct f2fs_summary_block *)kaddr; | |
249 | sum = sum_node->entries[blkoff]; | |
250 | f2fs_put_page(sum_page, 1); | |
251 | } | |
252 | ||
b292dcab JK |
253 | /* Use the locked dnode page and inode */ |
254 | nid = le32_to_cpu(sum.nid); | |
255 | if (dn->inode->i_ino == nid) { | |
256 | struct dnode_of_data tdn = *dn; | |
257 | tdn.nid = nid; | |
258 | tdn.node_page = dn->inode_page; | |
060dd67b | 259 | tdn.ofs_in_node = le16_to_cpu(sum.ofs_in_node); |
b292dcab | 260 | truncate_data_blocks_range(&tdn, 1); |
39cf72cf | 261 | return 0; |
b292dcab JK |
262 | } else if (dn->nid == nid) { |
263 | struct dnode_of_data tdn = *dn; | |
060dd67b | 264 | tdn.ofs_in_node = le16_to_cpu(sum.ofs_in_node); |
b292dcab | 265 | truncate_data_blocks_range(&tdn, 1); |
39cf72cf | 266 | return 0; |
b292dcab JK |
267 | } |
268 | ||
d624c96f | 269 | /* Get the node page */ |
b292dcab | 270 | node_page = get_node_page(sbi, nid); |
39cf72cf JK |
271 | if (IS_ERR(node_page)) |
272 | return PTR_ERR(node_page); | |
de93653f JK |
273 | |
274 | offset = ofs_of_node(node_page); | |
d624c96f JK |
275 | ino = ino_of_node(node_page); |
276 | f2fs_put_page(node_page, 1); | |
277 | ||
278 | /* Deallocate previous index in the node page */ | |
d4686d56 | 279 | inode = f2fs_iget(sbi->sb, ino); |
06025f4d | 280 | if (IS_ERR(inode)) |
39cf72cf | 281 | return PTR_ERR(inode); |
06025f4d | 282 | |
de93653f JK |
283 | bidx = start_bidx_of_node(offset, F2FS_I(inode)) + |
284 | le16_to_cpu(sum.ofs_in_node); | |
285 | ||
d624c96f JK |
286 | truncate_hole(inode, bidx, bidx + 1); |
287 | iput(inode); | |
39cf72cf | 288 | return 0; |
d624c96f JK |
289 | } |
290 | ||
6ead1142 | 291 | static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode, |
d624c96f JK |
292 | struct page *page, block_t blkaddr) |
293 | { | |
de93653f | 294 | struct f2fs_inode_info *fi = F2FS_I(inode); |
d624c96f JK |
295 | unsigned int start, end; |
296 | struct dnode_of_data dn; | |
297 | struct f2fs_summary sum; | |
298 | struct node_info ni; | |
f356fe0c | 299 | int err = 0, recovered = 0; |
d624c96f | 300 | |
de93653f | 301 | start = start_bidx_of_node(ofs_of_node(page), fi); |
d624c96f | 302 | if (IS_INODE(page)) |
de93653f | 303 | end = start + ADDRS_PER_INODE(fi); |
d624c96f JK |
304 | else |
305 | end = start + ADDRS_PER_BLOCK; | |
306 | ||
e479556b | 307 | f2fs_lock_op(sbi); |
d624c96f | 308 | set_new_dnode(&dn, inode, NULL, NULL, 0); |
39936837 | 309 | |
6ead1142 | 310 | err = get_dnode_of_data(&dn, start, ALLOC_NODE); |
39936837 | 311 | if (err) { |
e479556b | 312 | f2fs_unlock_op(sbi); |
6ead1142 | 313 | return err; |
39936837 | 314 | } |
d624c96f JK |
315 | |
316 | wait_on_page_writeback(dn.node_page); | |
317 | ||
318 | get_node_info(sbi, dn.nid, &ni); | |
5d56b671 JK |
319 | f2fs_bug_on(ni.ino != ino_of_node(page)); |
320 | f2fs_bug_on(ofs_of_node(dn.node_page) != ofs_of_node(page)); | |
d624c96f JK |
321 | |
322 | for (; start < end; start++) { | |
323 | block_t src, dest; | |
324 | ||
325 | src = datablock_addr(dn.node_page, dn.ofs_in_node); | |
326 | dest = datablock_addr(page, dn.ofs_in_node); | |
327 | ||
328 | if (src != dest && dest != NEW_ADDR && dest != NULL_ADDR) { | |
329 | if (src == NULL_ADDR) { | |
5d56b671 | 330 | err = reserve_new_block(&dn); |
d624c96f | 331 | /* We should not get -ENOSPC */ |
5d56b671 | 332 | f2fs_bug_on(err); |
d624c96f JK |
333 | } |
334 | ||
335 | /* Check the previous node page having this index */ | |
39cf72cf JK |
336 | err = check_index_in_prev_nodes(sbi, dest, &dn); |
337 | if (err) | |
338 | goto err; | |
d624c96f JK |
339 | |
340 | set_summary(&sum, dn.nid, dn.ofs_in_node, ni.version); | |
341 | ||
342 | /* write dummy data page */ | |
343 | recover_data_page(sbi, NULL, &sum, src, dest); | |
344 | update_extent_cache(dest, &dn); | |
f356fe0c | 345 | recovered++; |
d624c96f JK |
346 | } |
347 | dn.ofs_in_node++; | |
348 | } | |
349 | ||
350 | /* write node page in place */ | |
351 | set_summary(&sum, dn.nid, 0, 0); | |
352 | if (IS_INODE(dn.node_page)) | |
353 | sync_inode_page(&dn); | |
354 | ||
355 | copy_node_footer(dn.node_page, page); | |
356 | fill_node_footer(dn.node_page, dn.nid, ni.ino, | |
357 | ofs_of_node(page), false); | |
358 | set_page_dirty(dn.node_page); | |
359 | ||
360 | recover_node_page(sbi, dn.node_page, &sum, &ni, blkaddr); | |
39cf72cf | 361 | err: |
d624c96f | 362 | f2fs_put_dnode(&dn); |
e479556b | 363 | f2fs_unlock_op(sbi); |
f356fe0c JK |
364 | |
365 | f2fs_msg(sbi->sb, KERN_NOTICE, "recover_data: ino = %lx, " | |
39cf72cf JK |
366 | "recovered_data = %d blocks, err = %d", |
367 | inode->i_ino, recovered, err); | |
368 | return err; | |
d624c96f JK |
369 | } |
370 | ||
6ead1142 | 371 | static int recover_data(struct f2fs_sb_info *sbi, |
d624c96f JK |
372 | struct list_head *head, int type) |
373 | { | |
d71b5564 | 374 | unsigned long long cp_ver = cur_cp_version(F2FS_CKPT(sbi)); |
d624c96f JK |
375 | struct curseg_info *curseg; |
376 | struct page *page; | |
6ead1142 | 377 | int err = 0; |
d624c96f JK |
378 | block_t blkaddr; |
379 | ||
380 | /* get node pages in the current segment */ | |
381 | curseg = CURSEG_I(sbi, type); | |
382 | blkaddr = NEXT_FREE_BLKADDR(sbi, curseg); | |
383 | ||
384 | /* read node page */ | |
a0acdfe0 | 385 | page = alloc_page(GFP_F2FS_ZERO); |
e27dae4d | 386 | if (!page) |
6ead1142 JK |
387 | return -ENOMEM; |
388 | ||
d624c96f JK |
389 | lock_page(page); |
390 | ||
391 | while (1) { | |
392 | struct fsync_inode_entry *entry; | |
393 | ||
93dfe2ac | 394 | err = f2fs_submit_page_bio(sbi, page, blkaddr, READ_SYNC); |
6ead1142 | 395 | if (err) |
b9987a27 | 396 | return err; |
d624c96f | 397 | |
393ff91f JK |
398 | lock_page(page); |
399 | ||
d624c96f | 400 | if (cp_ver != cpver_of_node(page)) |
45856aff | 401 | break; |
d624c96f JK |
402 | |
403 | entry = get_fsync_inode(head, ino_of_node(page)); | |
404 | if (!entry) | |
405 | goto next; | |
406 | ||
6ead1142 JK |
407 | err = do_recover_data(sbi, entry->inode, page, blkaddr); |
408 | if (err) | |
45856aff | 409 | break; |
d624c96f JK |
410 | |
411 | if (entry->blkaddr == blkaddr) { | |
412 | iput(entry->inode); | |
413 | list_del(&entry->list); | |
414 | kmem_cache_free(fsync_entry_slab, entry); | |
415 | } | |
416 | next: | |
417 | /* check next segment */ | |
418 | blkaddr = next_blkaddr_of_node(page); | |
d624c96f | 419 | } |
b9987a27 | 420 | |
d624c96f JK |
421 | unlock_page(page); |
422 | __free_pages(page, 0); | |
423 | ||
6ead1142 JK |
424 | if (!err) |
425 | allocate_new_segments(sbi); | |
426 | return err; | |
d624c96f JK |
427 | } |
428 | ||
6ead1142 | 429 | int recover_fsync_data(struct f2fs_sb_info *sbi) |
d624c96f JK |
430 | { |
431 | struct list_head inode_list; | |
6ead1142 | 432 | int err; |
aabe5136 | 433 | bool need_writecp = false; |
d624c96f JK |
434 | |
435 | fsync_entry_slab = f2fs_kmem_cache_create("f2fs_fsync_inode_entry", | |
436 | sizeof(struct fsync_inode_entry), NULL); | |
6bacf52f | 437 | if (!fsync_entry_slab) |
6ead1142 | 438 | return -ENOMEM; |
d624c96f JK |
439 | |
440 | INIT_LIST_HEAD(&inode_list); | |
441 | ||
442 | /* step #1: find fsynced inode numbers */ | |
aabe5136 | 443 | sbi->por_doing = true; |
6ead1142 JK |
444 | err = find_fsync_dnodes(sbi, &inode_list); |
445 | if (err) | |
d624c96f JK |
446 | goto out; |
447 | ||
448 | if (list_empty(&inode_list)) | |
449 | goto out; | |
450 | ||
aabe5136 | 451 | need_writecp = true; |
691c6fd2 | 452 | |
d624c96f | 453 | /* step #2: recover data */ |
6ead1142 | 454 | err = recover_data(sbi, &inode_list, CURSEG_WARM_NODE); |
5d56b671 | 455 | f2fs_bug_on(!list_empty(&inode_list)); |
d624c96f | 456 | out: |
5ebefc5b | 457 | destroy_fsync_dnodes(&inode_list); |
d624c96f | 458 | kmem_cache_destroy(fsync_entry_slab); |
aabe5136 | 459 | sbi->por_doing = false; |
691c6fd2 | 460 | if (!err && need_writecp) |
2c2c149f | 461 | write_checkpoint(sbi, false); |
6ead1142 | 462 | return err; |
d624c96f | 463 | } |