]>
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 | { | |
45590710 | 43 | struct f2fs_node *raw_node = F2FS_NODE(ipage); |
d624c96f | 44 | struct f2fs_inode *raw_inode = &(raw_node->i); |
74d0b917 | 45 | nid_t pino = le32_to_cpu(raw_inode->i_pino); |
6b8213d9 | 46 | struct f2fs_dir_entry *de; |
b7f7a5e0 | 47 | struct qstr name; |
d624c96f | 48 | struct page *page; |
6b8213d9 | 49 | struct inode *dir, *einode; |
d624c96f JK |
50 | int err = 0; |
51 | ||
74d0b917 JK |
52 | dir = check_dirty_dir_inode(F2FS_SB(inode->i_sb), pino); |
53 | if (!dir) { | |
54 | dir = f2fs_iget(inode->i_sb, pino); | |
55 | if (IS_ERR(dir)) { | |
56 | err = PTR_ERR(dir); | |
57 | goto out; | |
58 | } | |
59 | set_inode_flag(F2FS_I(dir), FI_DELAY_IPUT); | |
5deb8267 | 60 | add_dirty_dir_inode(dir); |
d624c96f JK |
61 | } |
62 | ||
b7f7a5e0 AV |
63 | name.len = le32_to_cpu(raw_inode->i_namelen); |
64 | name.name = raw_inode->i_name; | |
6b8213d9 JK |
65 | retry: |
66 | de = f2fs_find_entry(dir, &name, &page); | |
2e5558f4 RK |
67 | if (de && inode->i_ino == le32_to_cpu(de->ino)) |
68 | goto out_unmap_put; | |
6b8213d9 JK |
69 | if (de) { |
70 | einode = f2fs_iget(inode->i_sb, le32_to_cpu(de->ino)); | |
71 | if (IS_ERR(einode)) { | |
72 | WARN_ON(1); | |
73 | if (PTR_ERR(einode) == -ENOENT) | |
74 | err = -EEXIST; | |
2e5558f4 RK |
75 | goto out_unmap_put; |
76 | } | |
77 | err = acquire_orphan_inode(F2FS_SB(inode->i_sb)); | |
78 | if (err) { | |
79 | iput(einode); | |
80 | goto out_unmap_put; | |
6b8213d9 JK |
81 | } |
82 | f2fs_delete_entry(de, page, einode); | |
83 | iput(einode); | |
84 | goto retry; | |
d624c96f | 85 | } |
6b8213d9 | 86 | err = __f2fs_add_link(dir, &name, inode); |
2e5558f4 RK |
87 | goto out; |
88 | ||
89 | out_unmap_put: | |
90 | kunmap(page); | |
91 | f2fs_put_page(page, 0); | |
d624c96f | 92 | out: |
f356fe0c JK |
93 | f2fs_msg(inode->i_sb, KERN_NOTICE, "recover_inode and its dentry: " |
94 | "ino = %x, name = %s, dir = %lx, err = %d", | |
f28c06fa DC |
95 | ino_of_node(ipage), raw_inode->i_name, |
96 | IS_ERR(dir) ? 0 : dir->i_ino, err); | |
d624c96f JK |
97 | return err; |
98 | } | |
99 | ||
100 | static int recover_inode(struct inode *inode, struct page *node_page) | |
101 | { | |
45590710 | 102 | struct f2fs_node *raw_node = F2FS_NODE(node_page); |
d624c96f JK |
103 | struct f2fs_inode *raw_inode = &(raw_node->i); |
104 | ||
f356fe0c JK |
105 | if (!IS_INODE(node_page)) |
106 | return 0; | |
107 | ||
25ca923b | 108 | inode->i_mode = le16_to_cpu(raw_inode->i_mode); |
d624c96f JK |
109 | i_size_write(inode, le64_to_cpu(raw_inode->i_size)); |
110 | inode->i_atime.tv_sec = le64_to_cpu(raw_inode->i_mtime); | |
111 | inode->i_ctime.tv_sec = le64_to_cpu(raw_inode->i_ctime); | |
112 | inode->i_mtime.tv_sec = le64_to_cpu(raw_inode->i_mtime); | |
113 | inode->i_atime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); | |
114 | inode->i_ctime.tv_nsec = le32_to_cpu(raw_inode->i_ctime_nsec); | |
115 | inode->i_mtime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); | |
116 | ||
f356fe0c JK |
117 | if (is_dent_dnode(node_page)) |
118 | return recover_dentry(node_page, inode); | |
119 | ||
120 | f2fs_msg(inode->i_sb, KERN_NOTICE, "recover_inode: ino = %x, name = %s", | |
121 | ino_of_node(node_page), raw_inode->i_name); | |
122 | return 0; | |
d624c96f JK |
123 | } |
124 | ||
125 | static int find_fsync_dnodes(struct f2fs_sb_info *sbi, struct list_head *head) | |
126 | { | |
d71b5564 | 127 | unsigned long long cp_ver = cur_cp_version(F2FS_CKPT(sbi)); |
d624c96f JK |
128 | struct curseg_info *curseg; |
129 | struct page *page; | |
130 | block_t blkaddr; | |
131 | int err = 0; | |
132 | ||
133 | /* get node pages in the current segment */ | |
134 | curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); | |
135 | blkaddr = START_BLOCK(sbi, curseg->segno) + curseg->next_blkoff; | |
136 | ||
137 | /* read node page */ | |
138 | page = alloc_page(GFP_F2FS_ZERO); | |
e27dae4d DC |
139 | if (!page) |
140 | return -ENOMEM; | |
d624c96f JK |
141 | lock_page(page); |
142 | ||
143 | while (1) { | |
144 | struct fsync_inode_entry *entry; | |
145 | ||
393ff91f JK |
146 | err = f2fs_readpage(sbi, page, blkaddr, READ_SYNC); |
147 | if (err) | |
d624c96f JK |
148 | goto out; |
149 | ||
393ff91f JK |
150 | lock_page(page); |
151 | ||
6ead1142 | 152 | if (cp_ver != cpver_of_node(page)) |
f356fe0c | 153 | break; |
d624c96f JK |
154 | |
155 | if (!is_fsync_dnode(page)) | |
156 | goto next; | |
157 | ||
158 | entry = get_fsync_inode(head, ino_of_node(page)); | |
159 | if (entry) { | |
d624c96f JK |
160 | if (IS_INODE(page) && is_dent_dnode(page)) |
161 | set_inode_flag(F2FS_I(entry->inode), | |
162 | FI_INC_LINK); | |
163 | } else { | |
164 | if (IS_INODE(page) && is_dent_dnode(page)) { | |
6ead1142 JK |
165 | err = recover_inode_page(sbi, page); |
166 | if (err) | |
f356fe0c | 167 | break; |
d624c96f JK |
168 | } |
169 | ||
170 | /* add this fsync inode to the list */ | |
171 | entry = kmem_cache_alloc(fsync_entry_slab, GFP_NOFS); | |
172 | if (!entry) { | |
173 | err = -ENOMEM; | |
f356fe0c | 174 | break; |
d624c96f JK |
175 | } |
176 | ||
d624c96f JK |
177 | entry->inode = f2fs_iget(sbi->sb, ino_of_node(page)); |
178 | if (IS_ERR(entry->inode)) { | |
179 | err = PTR_ERR(entry->inode); | |
fd8bb65f | 180 | kmem_cache_free(fsync_entry_slab, entry); |
f356fe0c | 181 | break; |
d624c96f | 182 | } |
fd8bb65f | 183 | list_add_tail(&entry->list, head); |
d624c96f | 184 | } |
addbe45b JK |
185 | entry->blkaddr = blkaddr; |
186 | ||
f356fe0c JK |
187 | err = recover_inode(entry->inode, page); |
188 | if (err && err != -ENOENT) | |
189 | break; | |
d624c96f JK |
190 | next: |
191 | /* check next segment */ | |
192 | blkaddr = next_blkaddr_of_node(page); | |
d624c96f | 193 | } |
d624c96f | 194 | unlock_page(page); |
393ff91f | 195 | out: |
d624c96f JK |
196 | __free_pages(page, 0); |
197 | return err; | |
198 | } | |
199 | ||
5ebefc5b | 200 | static void destroy_fsync_dnodes(struct list_head *head) |
d624c96f | 201 | { |
d8b79b2f DC |
202 | struct fsync_inode_entry *entry, *tmp; |
203 | ||
204 | list_for_each_entry_safe(entry, tmp, head, list) { | |
d624c96f JK |
205 | iput(entry->inode); |
206 | list_del(&entry->list); | |
207 | kmem_cache_free(fsync_entry_slab, entry); | |
208 | } | |
209 | } | |
210 | ||
39cf72cf | 211 | static int check_index_in_prev_nodes(struct f2fs_sb_info *sbi, |
b292dcab | 212 | block_t blkaddr, struct dnode_of_data *dn) |
d624c96f JK |
213 | { |
214 | struct seg_entry *sentry; | |
215 | unsigned int segno = GET_SEGNO(sbi, blkaddr); | |
216 | unsigned short blkoff = GET_SEGOFF_FROM_SEG0(sbi, blkaddr) & | |
217 | (sbi->blocks_per_seg - 1); | |
218 | struct f2fs_summary sum; | |
b292dcab | 219 | nid_t ino, nid; |
d624c96f JK |
220 | void *kaddr; |
221 | struct inode *inode; | |
222 | struct page *node_page; | |
de93653f | 223 | unsigned int offset; |
d624c96f JK |
224 | block_t bidx; |
225 | int i; | |
226 | ||
227 | sentry = get_seg_entry(sbi, segno); | |
228 | if (!f2fs_test_bit(blkoff, sentry->cur_valid_map)) | |
39cf72cf | 229 | return 0; |
d624c96f JK |
230 | |
231 | /* Get the previous summary */ | |
232 | for (i = CURSEG_WARM_DATA; i <= CURSEG_COLD_DATA; i++) { | |
233 | struct curseg_info *curseg = CURSEG_I(sbi, i); | |
234 | if (curseg->segno == segno) { | |
235 | sum = curseg->sum_blk->entries[blkoff]; | |
236 | break; | |
237 | } | |
238 | } | |
239 | if (i > CURSEG_COLD_DATA) { | |
240 | struct page *sum_page = get_sum_page(sbi, segno); | |
241 | struct f2fs_summary_block *sum_node; | |
242 | kaddr = page_address(sum_page); | |
243 | sum_node = (struct f2fs_summary_block *)kaddr; | |
244 | sum = sum_node->entries[blkoff]; | |
245 | f2fs_put_page(sum_page, 1); | |
246 | } | |
247 | ||
b292dcab JK |
248 | /* Use the locked dnode page and inode */ |
249 | nid = le32_to_cpu(sum.nid); | |
250 | if (dn->inode->i_ino == nid) { | |
251 | struct dnode_of_data tdn = *dn; | |
252 | tdn.nid = nid; | |
253 | tdn.node_page = dn->inode_page; | |
060dd67b | 254 | tdn.ofs_in_node = le16_to_cpu(sum.ofs_in_node); |
b292dcab | 255 | truncate_data_blocks_range(&tdn, 1); |
39cf72cf | 256 | return 0; |
b292dcab JK |
257 | } else if (dn->nid == nid) { |
258 | struct dnode_of_data tdn = *dn; | |
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 | } |
263 | ||
d624c96f | 264 | /* Get the node page */ |
b292dcab | 265 | node_page = get_node_page(sbi, nid); |
39cf72cf JK |
266 | if (IS_ERR(node_page)) |
267 | return PTR_ERR(node_page); | |
de93653f JK |
268 | |
269 | offset = ofs_of_node(node_page); | |
d624c96f JK |
270 | ino = ino_of_node(node_page); |
271 | f2fs_put_page(node_page, 1); | |
272 | ||
273 | /* Deallocate previous index in the node page */ | |
d4686d56 | 274 | inode = f2fs_iget(sbi->sb, ino); |
06025f4d | 275 | if (IS_ERR(inode)) |
39cf72cf | 276 | return PTR_ERR(inode); |
06025f4d | 277 | |
de93653f JK |
278 | bidx = start_bidx_of_node(offset, F2FS_I(inode)) + |
279 | le16_to_cpu(sum.ofs_in_node); | |
280 | ||
d624c96f JK |
281 | truncate_hole(inode, bidx, bidx + 1); |
282 | iput(inode); | |
39cf72cf | 283 | return 0; |
d624c96f JK |
284 | } |
285 | ||
6ead1142 | 286 | static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode, |
d624c96f JK |
287 | struct page *page, block_t blkaddr) |
288 | { | |
de93653f | 289 | struct f2fs_inode_info *fi = F2FS_I(inode); |
d624c96f JK |
290 | unsigned int start, end; |
291 | struct dnode_of_data dn; | |
292 | struct f2fs_summary sum; | |
293 | struct node_info ni; | |
f356fe0c | 294 | int err = 0, recovered = 0; |
d624c96f | 295 | |
de93653f | 296 | start = start_bidx_of_node(ofs_of_node(page), fi); |
d624c96f | 297 | if (IS_INODE(page)) |
de93653f | 298 | end = start + ADDRS_PER_INODE(fi); |
d624c96f JK |
299 | else |
300 | end = start + ADDRS_PER_BLOCK; | |
301 | ||
e479556b | 302 | f2fs_lock_op(sbi); |
d624c96f | 303 | set_new_dnode(&dn, inode, NULL, NULL, 0); |
39936837 | 304 | |
6ead1142 | 305 | err = get_dnode_of_data(&dn, start, ALLOC_NODE); |
39936837 | 306 | if (err) { |
e479556b | 307 | f2fs_unlock_op(sbi); |
6ead1142 | 308 | return err; |
39936837 | 309 | } |
d624c96f JK |
310 | |
311 | wait_on_page_writeback(dn.node_page); | |
312 | ||
313 | get_node_info(sbi, dn.nid, &ni); | |
314 | BUG_ON(ni.ino != ino_of_node(page)); | |
315 | BUG_ON(ofs_of_node(dn.node_page) != ofs_of_node(page)); | |
316 | ||
317 | for (; start < end; start++) { | |
318 | block_t src, dest; | |
319 | ||
320 | src = datablock_addr(dn.node_page, dn.ofs_in_node); | |
321 | dest = datablock_addr(page, dn.ofs_in_node); | |
322 | ||
323 | if (src != dest && dest != NEW_ADDR && dest != NULL_ADDR) { | |
324 | if (src == NULL_ADDR) { | |
325 | int err = reserve_new_block(&dn); | |
326 | /* We should not get -ENOSPC */ | |
327 | BUG_ON(err); | |
328 | } | |
329 | ||
330 | /* Check the previous node page having this index */ | |
39cf72cf JK |
331 | err = check_index_in_prev_nodes(sbi, dest, &dn); |
332 | if (err) | |
333 | goto err; | |
d624c96f JK |
334 | |
335 | set_summary(&sum, dn.nid, dn.ofs_in_node, ni.version); | |
336 | ||
337 | /* write dummy data page */ | |
338 | recover_data_page(sbi, NULL, &sum, src, dest); | |
339 | update_extent_cache(dest, &dn); | |
f356fe0c | 340 | recovered++; |
d624c96f JK |
341 | } |
342 | dn.ofs_in_node++; | |
343 | } | |
344 | ||
345 | /* write node page in place */ | |
346 | set_summary(&sum, dn.nid, 0, 0); | |
347 | if (IS_INODE(dn.node_page)) | |
348 | sync_inode_page(&dn); | |
349 | ||
350 | copy_node_footer(dn.node_page, page); | |
351 | fill_node_footer(dn.node_page, dn.nid, ni.ino, | |
352 | ofs_of_node(page), false); | |
353 | set_page_dirty(dn.node_page); | |
354 | ||
355 | recover_node_page(sbi, dn.node_page, &sum, &ni, blkaddr); | |
39cf72cf | 356 | err: |
d624c96f | 357 | f2fs_put_dnode(&dn); |
e479556b | 358 | f2fs_unlock_op(sbi); |
f356fe0c JK |
359 | |
360 | f2fs_msg(sbi->sb, KERN_NOTICE, "recover_data: ino = %lx, " | |
39cf72cf JK |
361 | "recovered_data = %d blocks, err = %d", |
362 | inode->i_ino, recovered, err); | |
363 | return err; | |
d624c96f JK |
364 | } |
365 | ||
6ead1142 | 366 | static int recover_data(struct f2fs_sb_info *sbi, |
d624c96f JK |
367 | struct list_head *head, int type) |
368 | { | |
d71b5564 | 369 | unsigned long long cp_ver = cur_cp_version(F2FS_CKPT(sbi)); |
d624c96f JK |
370 | struct curseg_info *curseg; |
371 | struct page *page; | |
6ead1142 | 372 | int err = 0; |
d624c96f JK |
373 | block_t blkaddr; |
374 | ||
375 | /* get node pages in the current segment */ | |
376 | curseg = CURSEG_I(sbi, type); | |
377 | blkaddr = NEXT_FREE_BLKADDR(sbi, curseg); | |
378 | ||
379 | /* read node page */ | |
380 | page = alloc_page(GFP_NOFS | __GFP_ZERO); | |
e27dae4d | 381 | if (!page) |
6ead1142 JK |
382 | return -ENOMEM; |
383 | ||
d624c96f JK |
384 | lock_page(page); |
385 | ||
386 | while (1) { | |
387 | struct fsync_inode_entry *entry; | |
388 | ||
6ead1142 JK |
389 | err = f2fs_readpage(sbi, page, blkaddr, READ_SYNC); |
390 | if (err) | |
d624c96f JK |
391 | goto out; |
392 | ||
393ff91f JK |
393 | lock_page(page); |
394 | ||
d624c96f | 395 | if (cp_ver != cpver_of_node(page)) |
45856aff | 396 | break; |
d624c96f JK |
397 | |
398 | entry = get_fsync_inode(head, ino_of_node(page)); | |
399 | if (!entry) | |
400 | goto next; | |
401 | ||
6ead1142 JK |
402 | err = do_recover_data(sbi, entry->inode, page, blkaddr); |
403 | if (err) | |
45856aff | 404 | break; |
d624c96f JK |
405 | |
406 | if (entry->blkaddr == blkaddr) { | |
407 | iput(entry->inode); | |
408 | list_del(&entry->list); | |
409 | kmem_cache_free(fsync_entry_slab, entry); | |
410 | } | |
411 | next: | |
412 | /* check next segment */ | |
413 | blkaddr = next_blkaddr_of_node(page); | |
d624c96f | 414 | } |
d624c96f | 415 | unlock_page(page); |
393ff91f | 416 | out: |
d624c96f JK |
417 | __free_pages(page, 0); |
418 | ||
6ead1142 JK |
419 | if (!err) |
420 | allocate_new_segments(sbi); | |
421 | return err; | |
d624c96f JK |
422 | } |
423 | ||
6ead1142 | 424 | int recover_fsync_data(struct f2fs_sb_info *sbi) |
d624c96f JK |
425 | { |
426 | struct list_head inode_list; | |
6ead1142 | 427 | int err; |
691c6fd2 | 428 | int need_writecp = 0; |
d624c96f JK |
429 | |
430 | fsync_entry_slab = f2fs_kmem_cache_create("f2fs_fsync_inode_entry", | |
431 | sizeof(struct fsync_inode_entry), NULL); | |
432 | if (unlikely(!fsync_entry_slab)) | |
6ead1142 | 433 | return -ENOMEM; |
d624c96f JK |
434 | |
435 | INIT_LIST_HEAD(&inode_list); | |
436 | ||
437 | /* step #1: find fsynced inode numbers */ | |
8c26d7d5 | 438 | sbi->por_doing = 1; |
6ead1142 JK |
439 | err = find_fsync_dnodes(sbi, &inode_list); |
440 | if (err) | |
d624c96f JK |
441 | goto out; |
442 | ||
443 | if (list_empty(&inode_list)) | |
444 | goto out; | |
445 | ||
691c6fd2 CY |
446 | need_writecp = 1; |
447 | ||
d624c96f | 448 | /* step #2: recover data */ |
6ead1142 | 449 | err = recover_data(sbi, &inode_list, CURSEG_WARM_NODE); |
d624c96f JK |
450 | BUG_ON(!list_empty(&inode_list)); |
451 | out: | |
5ebefc5b | 452 | destroy_fsync_dnodes(&inode_list); |
d624c96f | 453 | kmem_cache_destroy(fsync_entry_slab); |
8c26d7d5 | 454 | sbi->por_doing = 0; |
691c6fd2 | 455 | if (!err && need_writecp) |
2c2c149f | 456 | write_checkpoint(sbi, false); |
6ead1142 | 457 | return err; |
d624c96f | 458 | } |