]>
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 | { | |
43 | struct f2fs_node *raw_node = (struct f2fs_node *)kmap(ipage); | |
44 | struct f2fs_inode *raw_inode = &(raw_node->i); | |
b7f7a5e0 | 45 | struct qstr name; |
d624c96f JK |
46 | struct f2fs_dir_entry *de; |
47 | struct page *page; | |
48 | struct inode *dir; | |
49 | int err = 0; | |
50 | ||
51 | if (!is_dent_dnode(ipage)) | |
52 | goto out; | |
53 | ||
54 | dir = f2fs_iget(inode->i_sb, le32_to_cpu(raw_inode->i_pino)); | |
55 | if (IS_ERR(dir)) { | |
56 | err = -EINVAL; | |
57 | goto out; | |
58 | } | |
59 | ||
b7f7a5e0 AV |
60 | name.len = le32_to_cpu(raw_inode->i_namelen); |
61 | name.name = raw_inode->i_name; | |
d624c96f | 62 | |
b7f7a5e0 | 63 | de = f2fs_find_entry(dir, &name, &page); |
d624c96f JK |
64 | if (de) { |
65 | kunmap(page); | |
66 | f2fs_put_page(page, 0); | |
67 | } else { | |
90b2fc64 | 68 | err = __f2fs_add_link(dir, &name, inode); |
d624c96f JK |
69 | } |
70 | iput(dir); | |
71 | out: | |
72 | kunmap(ipage); | |
73 | return err; | |
74 | } | |
75 | ||
76 | static int recover_inode(struct inode *inode, struct page *node_page) | |
77 | { | |
78 | void *kaddr = page_address(node_page); | |
79 | struct f2fs_node *raw_node = (struct f2fs_node *)kaddr; | |
80 | struct f2fs_inode *raw_inode = &(raw_node->i); | |
81 | ||
25ca923b | 82 | inode->i_mode = le16_to_cpu(raw_inode->i_mode); |
d624c96f JK |
83 | i_size_write(inode, le64_to_cpu(raw_inode->i_size)); |
84 | inode->i_atime.tv_sec = le64_to_cpu(raw_inode->i_mtime); | |
85 | inode->i_ctime.tv_sec = le64_to_cpu(raw_inode->i_ctime); | |
86 | inode->i_mtime.tv_sec = le64_to_cpu(raw_inode->i_mtime); | |
87 | inode->i_atime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); | |
88 | inode->i_ctime.tv_nsec = le32_to_cpu(raw_inode->i_ctime_nsec); | |
89 | inode->i_mtime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); | |
90 | ||
91 | return recover_dentry(node_page, inode); | |
92 | } | |
93 | ||
94 | static int find_fsync_dnodes(struct f2fs_sb_info *sbi, struct list_head *head) | |
95 | { | |
96 | unsigned long long cp_ver = le64_to_cpu(sbi->ckpt->checkpoint_ver); | |
97 | struct curseg_info *curseg; | |
98 | struct page *page; | |
99 | block_t blkaddr; | |
100 | int err = 0; | |
101 | ||
102 | /* get node pages in the current segment */ | |
103 | curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); | |
104 | blkaddr = START_BLOCK(sbi, curseg->segno) + curseg->next_blkoff; | |
105 | ||
106 | /* read node page */ | |
107 | page = alloc_page(GFP_F2FS_ZERO); | |
108 | if (IS_ERR(page)) | |
109 | return PTR_ERR(page); | |
110 | lock_page(page); | |
111 | ||
112 | while (1) { | |
113 | struct fsync_inode_entry *entry; | |
114 | ||
393ff91f JK |
115 | err = f2fs_readpage(sbi, page, blkaddr, READ_SYNC); |
116 | if (err) | |
d624c96f JK |
117 | goto out; |
118 | ||
393ff91f JK |
119 | lock_page(page); |
120 | ||
6ead1142 | 121 | if (cp_ver != cpver_of_node(page)) |
393ff91f | 122 | goto unlock_out; |
d624c96f JK |
123 | |
124 | if (!is_fsync_dnode(page)) | |
125 | goto next; | |
126 | ||
127 | entry = get_fsync_inode(head, ino_of_node(page)); | |
128 | if (entry) { | |
129 | entry->blkaddr = blkaddr; | |
130 | if (IS_INODE(page) && is_dent_dnode(page)) | |
131 | set_inode_flag(F2FS_I(entry->inode), | |
132 | FI_INC_LINK); | |
133 | } else { | |
134 | if (IS_INODE(page) && is_dent_dnode(page)) { | |
6ead1142 JK |
135 | err = recover_inode_page(sbi, page); |
136 | if (err) | |
393ff91f | 137 | goto unlock_out; |
d624c96f JK |
138 | } |
139 | ||
140 | /* add this fsync inode to the list */ | |
141 | entry = kmem_cache_alloc(fsync_entry_slab, GFP_NOFS); | |
142 | if (!entry) { | |
143 | err = -ENOMEM; | |
393ff91f | 144 | goto unlock_out; |
d624c96f JK |
145 | } |
146 | ||
d624c96f JK |
147 | entry->inode = f2fs_iget(sbi->sb, ino_of_node(page)); |
148 | if (IS_ERR(entry->inode)) { | |
149 | err = PTR_ERR(entry->inode); | |
fd8bb65f | 150 | kmem_cache_free(fsync_entry_slab, entry); |
393ff91f | 151 | goto unlock_out; |
d624c96f | 152 | } |
fd8bb65f | 153 | |
fd8bb65f | 154 | list_add_tail(&entry->list, head); |
d624c96f JK |
155 | entry->blkaddr = blkaddr; |
156 | } | |
157 | if (IS_INODE(page)) { | |
158 | err = recover_inode(entry->inode, page); | |
159 | if (err) | |
393ff91f | 160 | goto unlock_out; |
d624c96f JK |
161 | } |
162 | next: | |
163 | /* check next segment */ | |
164 | blkaddr = next_blkaddr_of_node(page); | |
d624c96f | 165 | } |
393ff91f | 166 | unlock_out: |
d624c96f | 167 | unlock_page(page); |
393ff91f | 168 | out: |
d624c96f JK |
169 | __free_pages(page, 0); |
170 | return err; | |
171 | } | |
172 | ||
173 | static void destroy_fsync_dnodes(struct f2fs_sb_info *sbi, | |
174 | struct list_head *head) | |
175 | { | |
d8b79b2f DC |
176 | struct fsync_inode_entry *entry, *tmp; |
177 | ||
178 | list_for_each_entry_safe(entry, tmp, head, list) { | |
d624c96f JK |
179 | iput(entry->inode); |
180 | list_del(&entry->list); | |
181 | kmem_cache_free(fsync_entry_slab, entry); | |
182 | } | |
183 | } | |
184 | ||
185 | static void check_index_in_prev_nodes(struct f2fs_sb_info *sbi, | |
186 | block_t blkaddr) | |
187 | { | |
188 | struct seg_entry *sentry; | |
189 | unsigned int segno = GET_SEGNO(sbi, blkaddr); | |
190 | unsigned short blkoff = GET_SEGOFF_FROM_SEG0(sbi, blkaddr) & | |
191 | (sbi->blocks_per_seg - 1); | |
192 | struct f2fs_summary sum; | |
193 | nid_t ino; | |
194 | void *kaddr; | |
195 | struct inode *inode; | |
196 | struct page *node_page; | |
197 | block_t bidx; | |
198 | int i; | |
199 | ||
200 | sentry = get_seg_entry(sbi, segno); | |
201 | if (!f2fs_test_bit(blkoff, sentry->cur_valid_map)) | |
202 | return; | |
203 | ||
204 | /* Get the previous summary */ | |
205 | for (i = CURSEG_WARM_DATA; i <= CURSEG_COLD_DATA; i++) { | |
206 | struct curseg_info *curseg = CURSEG_I(sbi, i); | |
207 | if (curseg->segno == segno) { | |
208 | sum = curseg->sum_blk->entries[blkoff]; | |
209 | break; | |
210 | } | |
211 | } | |
212 | if (i > CURSEG_COLD_DATA) { | |
213 | struct page *sum_page = get_sum_page(sbi, segno); | |
214 | struct f2fs_summary_block *sum_node; | |
215 | kaddr = page_address(sum_page); | |
216 | sum_node = (struct f2fs_summary_block *)kaddr; | |
217 | sum = sum_node->entries[blkoff]; | |
218 | f2fs_put_page(sum_page, 1); | |
219 | } | |
220 | ||
221 | /* Get the node page */ | |
222 | node_page = get_node_page(sbi, le32_to_cpu(sum.nid)); | |
223 | bidx = start_bidx_of_node(ofs_of_node(node_page)) + | |
224 | le16_to_cpu(sum.ofs_in_node); | |
225 | ino = ino_of_node(node_page); | |
226 | f2fs_put_page(node_page, 1); | |
227 | ||
228 | /* Deallocate previous index in the node page */ | |
d4686d56 | 229 | inode = f2fs_iget(sbi->sb, ino); |
06025f4d NJ |
230 | if (IS_ERR(inode)) |
231 | return; | |
232 | ||
d624c96f JK |
233 | truncate_hole(inode, bidx, bidx + 1); |
234 | iput(inode); | |
235 | } | |
236 | ||
6ead1142 | 237 | static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode, |
d624c96f JK |
238 | struct page *page, block_t blkaddr) |
239 | { | |
240 | unsigned int start, end; | |
241 | struct dnode_of_data dn; | |
242 | struct f2fs_summary sum; | |
243 | struct node_info ni; | |
6ead1142 | 244 | int err = 0; |
39936837 | 245 | int ilock; |
d624c96f JK |
246 | |
247 | start = start_bidx_of_node(ofs_of_node(page)); | |
248 | if (IS_INODE(page)) | |
249 | end = start + ADDRS_PER_INODE; | |
250 | else | |
251 | end = start + ADDRS_PER_BLOCK; | |
252 | ||
39936837 | 253 | ilock = mutex_lock_op(sbi); |
d624c96f | 254 | set_new_dnode(&dn, inode, NULL, NULL, 0); |
39936837 | 255 | |
6ead1142 | 256 | err = get_dnode_of_data(&dn, start, ALLOC_NODE); |
39936837 JK |
257 | if (err) { |
258 | mutex_unlock_op(sbi, ilock); | |
6ead1142 | 259 | return err; |
39936837 | 260 | } |
d624c96f JK |
261 | |
262 | wait_on_page_writeback(dn.node_page); | |
263 | ||
264 | get_node_info(sbi, dn.nid, &ni); | |
265 | BUG_ON(ni.ino != ino_of_node(page)); | |
266 | BUG_ON(ofs_of_node(dn.node_page) != ofs_of_node(page)); | |
267 | ||
268 | for (; start < end; start++) { | |
269 | block_t src, dest; | |
270 | ||
271 | src = datablock_addr(dn.node_page, dn.ofs_in_node); | |
272 | dest = datablock_addr(page, dn.ofs_in_node); | |
273 | ||
274 | if (src != dest && dest != NEW_ADDR && dest != NULL_ADDR) { | |
275 | if (src == NULL_ADDR) { | |
276 | int err = reserve_new_block(&dn); | |
277 | /* We should not get -ENOSPC */ | |
278 | BUG_ON(err); | |
279 | } | |
280 | ||
281 | /* Check the previous node page having this index */ | |
282 | check_index_in_prev_nodes(sbi, dest); | |
283 | ||
284 | set_summary(&sum, dn.nid, dn.ofs_in_node, ni.version); | |
285 | ||
286 | /* write dummy data page */ | |
287 | recover_data_page(sbi, NULL, &sum, src, dest); | |
288 | update_extent_cache(dest, &dn); | |
289 | } | |
290 | dn.ofs_in_node++; | |
291 | } | |
292 | ||
293 | /* write node page in place */ | |
294 | set_summary(&sum, dn.nid, 0, 0); | |
295 | if (IS_INODE(dn.node_page)) | |
296 | sync_inode_page(&dn); | |
297 | ||
298 | copy_node_footer(dn.node_page, page); | |
299 | fill_node_footer(dn.node_page, dn.nid, ni.ino, | |
300 | ofs_of_node(page), false); | |
301 | set_page_dirty(dn.node_page); | |
302 | ||
303 | recover_node_page(sbi, dn.node_page, &sum, &ni, blkaddr); | |
304 | f2fs_put_dnode(&dn); | |
39936837 | 305 | mutex_unlock_op(sbi, ilock); |
6ead1142 | 306 | return 0; |
d624c96f JK |
307 | } |
308 | ||
6ead1142 | 309 | static int recover_data(struct f2fs_sb_info *sbi, |
d624c96f JK |
310 | struct list_head *head, int type) |
311 | { | |
312 | unsigned long long cp_ver = le64_to_cpu(sbi->ckpt->checkpoint_ver); | |
313 | struct curseg_info *curseg; | |
314 | struct page *page; | |
6ead1142 | 315 | int err = 0; |
d624c96f JK |
316 | block_t blkaddr; |
317 | ||
318 | /* get node pages in the current segment */ | |
319 | curseg = CURSEG_I(sbi, type); | |
320 | blkaddr = NEXT_FREE_BLKADDR(sbi, curseg); | |
321 | ||
322 | /* read node page */ | |
323 | page = alloc_page(GFP_NOFS | __GFP_ZERO); | |
324 | if (IS_ERR(page)) | |
6ead1142 JK |
325 | return -ENOMEM; |
326 | ||
d624c96f JK |
327 | lock_page(page); |
328 | ||
329 | while (1) { | |
330 | struct fsync_inode_entry *entry; | |
331 | ||
6ead1142 JK |
332 | err = f2fs_readpage(sbi, page, blkaddr, READ_SYNC); |
333 | if (err) | |
d624c96f JK |
334 | goto out; |
335 | ||
393ff91f JK |
336 | lock_page(page); |
337 | ||
d624c96f | 338 | if (cp_ver != cpver_of_node(page)) |
393ff91f | 339 | goto unlock_out; |
d624c96f JK |
340 | |
341 | entry = get_fsync_inode(head, ino_of_node(page)); | |
342 | if (!entry) | |
343 | goto next; | |
344 | ||
6ead1142 JK |
345 | err = do_recover_data(sbi, entry->inode, page, blkaddr); |
346 | if (err) | |
347 | goto out; | |
d624c96f JK |
348 | |
349 | if (entry->blkaddr == blkaddr) { | |
350 | iput(entry->inode); | |
351 | list_del(&entry->list); | |
352 | kmem_cache_free(fsync_entry_slab, entry); | |
353 | } | |
354 | next: | |
355 | /* check next segment */ | |
356 | blkaddr = next_blkaddr_of_node(page); | |
d624c96f | 357 | } |
393ff91f | 358 | unlock_out: |
d624c96f | 359 | unlock_page(page); |
393ff91f | 360 | out: |
d624c96f JK |
361 | __free_pages(page, 0); |
362 | ||
6ead1142 JK |
363 | if (!err) |
364 | allocate_new_segments(sbi); | |
365 | return err; | |
d624c96f JK |
366 | } |
367 | ||
6ead1142 | 368 | int recover_fsync_data(struct f2fs_sb_info *sbi) |
d624c96f JK |
369 | { |
370 | struct list_head inode_list; | |
6ead1142 | 371 | int err; |
d624c96f JK |
372 | |
373 | fsync_entry_slab = f2fs_kmem_cache_create("f2fs_fsync_inode_entry", | |
374 | sizeof(struct fsync_inode_entry), NULL); | |
375 | if (unlikely(!fsync_entry_slab)) | |
6ead1142 | 376 | return -ENOMEM; |
d624c96f JK |
377 | |
378 | INIT_LIST_HEAD(&inode_list); | |
379 | ||
380 | /* step #1: find fsynced inode numbers */ | |
6ead1142 JK |
381 | err = find_fsync_dnodes(sbi, &inode_list); |
382 | if (err) | |
d624c96f JK |
383 | goto out; |
384 | ||
385 | if (list_empty(&inode_list)) | |
386 | goto out; | |
387 | ||
388 | /* step #2: recover data */ | |
389 | sbi->por_doing = 1; | |
6ead1142 | 390 | err = recover_data(sbi, &inode_list, CURSEG_WARM_NODE); |
d624c96f JK |
391 | sbi->por_doing = 0; |
392 | BUG_ON(!list_empty(&inode_list)); | |
393 | out: | |
394 | destroy_fsync_dnodes(sbi, &inode_list); | |
395 | kmem_cache_destroy(fsync_entry_slab); | |
43727527 | 396 | write_checkpoint(sbi, false); |
6ead1142 | 397 | return err; |
d624c96f | 398 | } |