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 | { | |
d624c96f JK |
30 | struct fsync_inode_entry *entry; |
31 | ||
2d7b822a | 32 | list_for_each_entry(entry, head, list) |
d624c96f JK |
33 | if (entry->inode->i_ino == ino) |
34 | return entry; | |
2d7b822a | 35 | |
d624c96f JK |
36 | return NULL; |
37 | } | |
38 | ||
39 | static int recover_dentry(struct page *ipage, struct inode *inode) | |
40 | { | |
58bfaf44 | 41 | struct f2fs_inode *raw_inode = F2FS_INODE(ipage); |
74d0b917 | 42 | nid_t pino = le32_to_cpu(raw_inode->i_pino); |
6b8213d9 | 43 | struct f2fs_dir_entry *de; |
b7f7a5e0 | 44 | struct qstr name; |
d624c96f | 45 | struct page *page; |
6b8213d9 | 46 | struct inode *dir, *einode; |
d624c96f JK |
47 | int err = 0; |
48 | ||
ed57c27f JK |
49 | dir = f2fs_iget(inode->i_sb, pino); |
50 | if (IS_ERR(dir)) { | |
51 | err = PTR_ERR(dir); | |
52 | goto out; | |
53 | } | |
54 | ||
b7f7a5e0 AV |
55 | name.len = le32_to_cpu(raw_inode->i_namelen); |
56 | name.name = raw_inode->i_name; | |
d96b1431 CY |
57 | |
58 | if (unlikely(name.len > F2FS_NAME_LEN)) { | |
59 | WARN_ON(1); | |
60 | err = -ENAMETOOLONG; | |
86928f98 | 61 | goto out_err; |
d96b1431 | 62 | } |
6b8213d9 JK |
63 | retry: |
64 | de = f2fs_find_entry(dir, &name, &page); | |
2e5558f4 RK |
65 | if (de && inode->i_ino == le32_to_cpu(de->ino)) |
66 | goto out_unmap_put; | |
6b8213d9 JK |
67 | if (de) { |
68 | einode = f2fs_iget(inode->i_sb, le32_to_cpu(de->ino)); | |
69 | if (IS_ERR(einode)) { | |
70 | WARN_ON(1); | |
5c1f9927 CY |
71 | err = PTR_ERR(einode); |
72 | if (err == -ENOENT) | |
6b8213d9 | 73 | err = -EEXIST; |
2e5558f4 RK |
74 | goto out_unmap_put; |
75 | } | |
76 | err = acquire_orphan_inode(F2FS_SB(inode->i_sb)); | |
77 | if (err) { | |
78 | iput(einode); | |
79 | goto out_unmap_put; | |
6b8213d9 JK |
80 | } |
81 | f2fs_delete_entry(de, page, einode); | |
82 | iput(einode); | |
83 | goto retry; | |
d624c96f | 84 | } |
6b8213d9 | 85 | err = __f2fs_add_link(dir, &name, inode); |
86928f98 JK |
86 | if (err) |
87 | goto out_err; | |
88 | ||
89 | if (is_inode_flag_set(F2FS_I(dir), FI_DELAY_IPUT)) { | |
90 | iput(dir); | |
91 | } else { | |
92 | add_dirty_dir_inode(dir); | |
93 | set_inode_flag(F2FS_I(dir), FI_DELAY_IPUT); | |
94 | } | |
95 | ||
2e5558f4 RK |
96 | goto out; |
97 | ||
98 | out_unmap_put: | |
99 | kunmap(page); | |
100 | f2fs_put_page(page, 0); | |
86928f98 JK |
101 | out_err: |
102 | iput(dir); | |
d624c96f | 103 | out: |
6c311ec6 CF |
104 | f2fs_msg(inode->i_sb, KERN_NOTICE, |
105 | "%s: ino = %x, name = %s, dir = %lx, err = %d", | |
106 | __func__, ino_of_node(ipage), raw_inode->i_name, | |
f28c06fa | 107 | IS_ERR(dir) ? 0 : dir->i_ino, err); |
d624c96f JK |
108 | return err; |
109 | } | |
110 | ||
111 | static int recover_inode(struct inode *inode, struct page *node_page) | |
112 | { | |
58bfaf44 | 113 | struct f2fs_inode *raw_inode = F2FS_INODE(node_page); |
d624c96f | 114 | |
f356fe0c JK |
115 | if (!IS_INODE(node_page)) |
116 | return 0; | |
117 | ||
25ca923b | 118 | inode->i_mode = le16_to_cpu(raw_inode->i_mode); |
d624c96f JK |
119 | i_size_write(inode, le64_to_cpu(raw_inode->i_size)); |
120 | inode->i_atime.tv_sec = le64_to_cpu(raw_inode->i_mtime); | |
121 | inode->i_ctime.tv_sec = le64_to_cpu(raw_inode->i_ctime); | |
122 | inode->i_mtime.tv_sec = le64_to_cpu(raw_inode->i_mtime); | |
123 | inode->i_atime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); | |
124 | inode->i_ctime.tv_nsec = le32_to_cpu(raw_inode->i_ctime_nsec); | |
125 | inode->i_mtime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); | |
126 | ||
f356fe0c JK |
127 | if (is_dent_dnode(node_page)) |
128 | return recover_dentry(node_page, inode); | |
129 | ||
130 | f2fs_msg(inode->i_sb, KERN_NOTICE, "recover_inode: ino = %x, name = %s", | |
131 | ino_of_node(node_page), raw_inode->i_name); | |
132 | return 0; | |
d624c96f JK |
133 | } |
134 | ||
135 | static int find_fsync_dnodes(struct f2fs_sb_info *sbi, struct list_head *head) | |
136 | { | |
d71b5564 | 137 | unsigned long long cp_ver = cur_cp_version(F2FS_CKPT(sbi)); |
d624c96f JK |
138 | struct curseg_info *curseg; |
139 | struct page *page; | |
140 | block_t blkaddr; | |
141 | int err = 0; | |
142 | ||
143 | /* get node pages in the current segment */ | |
144 | curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); | |
695fd1ed | 145 | blkaddr = NEXT_FREE_BLKADDR(sbi, curseg); |
d624c96f JK |
146 | |
147 | /* read node page */ | |
148 | page = alloc_page(GFP_F2FS_ZERO); | |
e27dae4d DC |
149 | if (!page) |
150 | return -ENOMEM; | |
d624c96f JK |
151 | lock_page(page); |
152 | ||
153 | while (1) { | |
154 | struct fsync_inode_entry *entry; | |
155 | ||
93dfe2ac | 156 | err = f2fs_submit_page_bio(sbi, page, blkaddr, READ_SYNC); |
393ff91f | 157 | if (err) |
b9987a27 | 158 | return err; |
d624c96f | 159 | |
393ff91f JK |
160 | lock_page(page); |
161 | ||
6ead1142 | 162 | if (cp_ver != cpver_of_node(page)) |
f356fe0c | 163 | break; |
d624c96f JK |
164 | |
165 | if (!is_fsync_dnode(page)) | |
166 | goto next; | |
167 | ||
168 | entry = get_fsync_inode(head, ino_of_node(page)); | |
169 | if (entry) { | |
d624c96f JK |
170 | if (IS_INODE(page) && is_dent_dnode(page)) |
171 | set_inode_flag(F2FS_I(entry->inode), | |
172 | FI_INC_LINK); | |
173 | } else { | |
174 | if (IS_INODE(page) && is_dent_dnode(page)) { | |
6ead1142 JK |
175 | err = recover_inode_page(sbi, page); |
176 | if (err) | |
f356fe0c | 177 | break; |
d624c96f JK |
178 | } |
179 | ||
180 | /* add this fsync inode to the list */ | |
181 | entry = kmem_cache_alloc(fsync_entry_slab, GFP_NOFS); | |
182 | if (!entry) { | |
183 | err = -ENOMEM; | |
f356fe0c | 184 | break; |
d624c96f JK |
185 | } |
186 | ||
d624c96f JK |
187 | entry->inode = f2fs_iget(sbi->sb, ino_of_node(page)); |
188 | if (IS_ERR(entry->inode)) { | |
189 | err = PTR_ERR(entry->inode); | |
fd8bb65f | 190 | kmem_cache_free(fsync_entry_slab, entry); |
f356fe0c | 191 | break; |
d624c96f | 192 | } |
fd8bb65f | 193 | list_add_tail(&entry->list, head); |
d624c96f | 194 | } |
addbe45b JK |
195 | entry->blkaddr = blkaddr; |
196 | ||
f356fe0c JK |
197 | err = recover_inode(entry->inode, page); |
198 | if (err && err != -ENOENT) | |
199 | break; | |
d624c96f JK |
200 | next: |
201 | /* check next segment */ | |
202 | blkaddr = next_blkaddr_of_node(page); | |
d624c96f | 203 | } |
b9987a27 | 204 | |
d624c96f JK |
205 | unlock_page(page); |
206 | __free_pages(page, 0); | |
b9987a27 | 207 | |
d624c96f JK |
208 | return err; |
209 | } | |
210 | ||
5ebefc5b | 211 | static void destroy_fsync_dnodes(struct list_head *head) |
d624c96f | 212 | { |
d8b79b2f DC |
213 | struct fsync_inode_entry *entry, *tmp; |
214 | ||
215 | list_for_each_entry_safe(entry, tmp, head, list) { | |
d624c96f JK |
216 | iput(entry->inode); |
217 | list_del(&entry->list); | |
218 | kmem_cache_free(fsync_entry_slab, entry); | |
219 | } | |
220 | } | |
221 | ||
39cf72cf | 222 | static int check_index_in_prev_nodes(struct f2fs_sb_info *sbi, |
b292dcab | 223 | block_t blkaddr, struct dnode_of_data *dn) |
d624c96f JK |
224 | { |
225 | struct seg_entry *sentry; | |
226 | unsigned int segno = GET_SEGNO(sbi, blkaddr); | |
491c0854 | 227 | unsigned short blkoff = GET_BLKOFF_FROM_SEG0(sbi, blkaddr); |
f6517cfc | 228 | struct f2fs_summary_block *sum_node; |
d624c96f | 229 | struct f2fs_summary sum; |
f6517cfc | 230 | struct page *sum_page, *node_page; |
b292dcab | 231 | nid_t ino, nid; |
d624c96f | 232 | struct inode *inode; |
de93653f | 233 | unsigned int offset; |
d624c96f JK |
234 | block_t bidx; |
235 | int i; | |
236 | ||
237 | sentry = get_seg_entry(sbi, segno); | |
238 | if (!f2fs_test_bit(blkoff, sentry->cur_valid_map)) | |
39cf72cf | 239 | return 0; |
d624c96f JK |
240 | |
241 | /* Get the previous summary */ | |
242 | for (i = CURSEG_WARM_DATA; i <= CURSEG_COLD_DATA; i++) { | |
243 | struct curseg_info *curseg = CURSEG_I(sbi, i); | |
244 | if (curseg->segno == segno) { | |
245 | sum = curseg->sum_blk->entries[blkoff]; | |
f6517cfc | 246 | goto got_it; |
d624c96f JK |
247 | } |
248 | } | |
d624c96f | 249 | |
f6517cfc JK |
250 | sum_page = get_sum_page(sbi, segno); |
251 | sum_node = (struct f2fs_summary_block *)page_address(sum_page); | |
252 | sum = sum_node->entries[blkoff]; | |
253 | f2fs_put_page(sum_page, 1); | |
254 | got_it: | |
b292dcab JK |
255 | /* Use the locked dnode page and inode */ |
256 | nid = le32_to_cpu(sum.nid); | |
257 | if (dn->inode->i_ino == nid) { | |
258 | struct dnode_of_data tdn = *dn; | |
259 | tdn.nid = nid; | |
260 | tdn.node_page = dn->inode_page; | |
060dd67b | 261 | tdn.ofs_in_node = le16_to_cpu(sum.ofs_in_node); |
b292dcab | 262 | truncate_data_blocks_range(&tdn, 1); |
39cf72cf | 263 | return 0; |
b292dcab JK |
264 | } else if (dn->nid == nid) { |
265 | struct dnode_of_data tdn = *dn; | |
060dd67b | 266 | tdn.ofs_in_node = le16_to_cpu(sum.ofs_in_node); |
b292dcab | 267 | truncate_data_blocks_range(&tdn, 1); |
39cf72cf | 268 | return 0; |
b292dcab JK |
269 | } |
270 | ||
d624c96f | 271 | /* Get the node page */ |
b292dcab | 272 | node_page = get_node_page(sbi, nid); |
39cf72cf JK |
273 | if (IS_ERR(node_page)) |
274 | return PTR_ERR(node_page); | |
de93653f JK |
275 | |
276 | offset = ofs_of_node(node_page); | |
d624c96f JK |
277 | ino = ino_of_node(node_page); |
278 | f2fs_put_page(node_page, 1); | |
279 | ||
280 | /* Deallocate previous index in the node page */ | |
d4686d56 | 281 | inode = f2fs_iget(sbi->sb, ino); |
06025f4d | 282 | if (IS_ERR(inode)) |
39cf72cf | 283 | return PTR_ERR(inode); |
06025f4d | 284 | |
de93653f JK |
285 | bidx = start_bidx_of_node(offset, F2FS_I(inode)) + |
286 | le16_to_cpu(sum.ofs_in_node); | |
287 | ||
d624c96f JK |
288 | truncate_hole(inode, bidx, bidx + 1); |
289 | iput(inode); | |
39cf72cf | 290 | return 0; |
d624c96f JK |
291 | } |
292 | ||
6ead1142 | 293 | static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode, |
d624c96f JK |
294 | struct page *page, block_t blkaddr) |
295 | { | |
de93653f | 296 | struct f2fs_inode_info *fi = F2FS_I(inode); |
d624c96f JK |
297 | unsigned int start, end; |
298 | struct dnode_of_data dn; | |
299 | struct f2fs_summary sum; | |
300 | struct node_info ni; | |
f356fe0c | 301 | int err = 0, recovered = 0; |
d624c96f | 302 | |
1e1bb4ba JK |
303 | if (recover_inline_data(inode, page)) |
304 | goto out; | |
305 | ||
abb2366c JK |
306 | if (recover_xattr_data(inode, page, blkaddr)) |
307 | goto out; | |
308 | ||
de93653f | 309 | start = start_bidx_of_node(ofs_of_node(page), fi); |
6403eb1f | 310 | end = start + ADDRS_PER_PAGE(page, fi); |
d624c96f | 311 | |
e479556b | 312 | f2fs_lock_op(sbi); |
1e1bb4ba | 313 | |
d624c96f | 314 | set_new_dnode(&dn, inode, NULL, NULL, 0); |
39936837 | 315 | |
6ead1142 | 316 | err = get_dnode_of_data(&dn, start, ALLOC_NODE); |
39936837 | 317 | if (err) { |
e479556b | 318 | f2fs_unlock_op(sbi); |
1e1bb4ba | 319 | goto out; |
39936837 | 320 | } |
d624c96f | 321 | |
3cb5ad15 | 322 | f2fs_wait_on_page_writeback(dn.node_page, NODE); |
d624c96f JK |
323 | |
324 | get_node_info(sbi, dn.nid, &ni); | |
5d56b671 JK |
325 | f2fs_bug_on(ni.ino != ino_of_node(page)); |
326 | f2fs_bug_on(ofs_of_node(dn.node_page) != ofs_of_node(page)); | |
d624c96f JK |
327 | |
328 | for (; start < end; start++) { | |
329 | block_t src, dest; | |
330 | ||
331 | src = datablock_addr(dn.node_page, dn.ofs_in_node); | |
332 | dest = datablock_addr(page, dn.ofs_in_node); | |
333 | ||
334 | if (src != dest && dest != NEW_ADDR && dest != NULL_ADDR) { | |
335 | if (src == NULL_ADDR) { | |
5d56b671 | 336 | err = reserve_new_block(&dn); |
d624c96f | 337 | /* We should not get -ENOSPC */ |
5d56b671 | 338 | f2fs_bug_on(err); |
d624c96f JK |
339 | } |
340 | ||
341 | /* Check the previous node page having this index */ | |
39cf72cf JK |
342 | err = check_index_in_prev_nodes(sbi, dest, &dn); |
343 | if (err) | |
344 | goto err; | |
d624c96f JK |
345 | |
346 | set_summary(&sum, dn.nid, dn.ofs_in_node, ni.version); | |
347 | ||
348 | /* write dummy data page */ | |
349 | recover_data_page(sbi, NULL, &sum, src, dest); | |
350 | update_extent_cache(dest, &dn); | |
f356fe0c | 351 | recovered++; |
d624c96f JK |
352 | } |
353 | dn.ofs_in_node++; | |
354 | } | |
355 | ||
356 | /* write node page in place */ | |
357 | set_summary(&sum, dn.nid, 0, 0); | |
358 | if (IS_INODE(dn.node_page)) | |
359 | sync_inode_page(&dn); | |
360 | ||
361 | copy_node_footer(dn.node_page, page); | |
362 | fill_node_footer(dn.node_page, dn.nid, ni.ino, | |
363 | ofs_of_node(page), false); | |
364 | set_page_dirty(dn.node_page); | |
365 | ||
366 | recover_node_page(sbi, dn.node_page, &sum, &ni, blkaddr); | |
39cf72cf | 367 | err: |
d624c96f | 368 | f2fs_put_dnode(&dn); |
e479556b | 369 | f2fs_unlock_op(sbi); |
1e1bb4ba | 370 | out: |
6c311ec6 CF |
371 | f2fs_msg(sbi->sb, KERN_NOTICE, |
372 | "recover_data: ino = %lx, recovered = %d blocks, err = %d", | |
373 | inode->i_ino, recovered, err); | |
39cf72cf | 374 | return err; |
d624c96f JK |
375 | } |
376 | ||
6ead1142 | 377 | static int recover_data(struct f2fs_sb_info *sbi, |
d624c96f JK |
378 | struct list_head *head, int type) |
379 | { | |
d71b5564 | 380 | unsigned long long cp_ver = cur_cp_version(F2FS_CKPT(sbi)); |
d624c96f JK |
381 | struct curseg_info *curseg; |
382 | struct page *page; | |
6ead1142 | 383 | int err = 0; |
d624c96f JK |
384 | block_t blkaddr; |
385 | ||
386 | /* get node pages in the current segment */ | |
387 | curseg = CURSEG_I(sbi, type); | |
388 | blkaddr = NEXT_FREE_BLKADDR(sbi, curseg); | |
389 | ||
390 | /* read node page */ | |
a0acdfe0 | 391 | page = alloc_page(GFP_F2FS_ZERO); |
e27dae4d | 392 | if (!page) |
6ead1142 JK |
393 | return -ENOMEM; |
394 | ||
d624c96f JK |
395 | lock_page(page); |
396 | ||
397 | while (1) { | |
398 | struct fsync_inode_entry *entry; | |
399 | ||
93dfe2ac | 400 | err = f2fs_submit_page_bio(sbi, page, blkaddr, READ_SYNC); |
6ead1142 | 401 | if (err) |
b9987a27 | 402 | return err; |
d624c96f | 403 | |
393ff91f JK |
404 | lock_page(page); |
405 | ||
d624c96f | 406 | if (cp_ver != cpver_of_node(page)) |
45856aff | 407 | break; |
d624c96f JK |
408 | |
409 | entry = get_fsync_inode(head, ino_of_node(page)); | |
410 | if (!entry) | |
411 | goto next; | |
412 | ||
6ead1142 JK |
413 | err = do_recover_data(sbi, entry->inode, page, blkaddr); |
414 | if (err) | |
45856aff | 415 | break; |
d624c96f JK |
416 | |
417 | if (entry->blkaddr == blkaddr) { | |
418 | iput(entry->inode); | |
419 | list_del(&entry->list); | |
420 | kmem_cache_free(fsync_entry_slab, entry); | |
421 | } | |
422 | next: | |
423 | /* check next segment */ | |
424 | blkaddr = next_blkaddr_of_node(page); | |
d624c96f | 425 | } |
b9987a27 | 426 | |
d624c96f JK |
427 | unlock_page(page); |
428 | __free_pages(page, 0); | |
429 | ||
6ead1142 JK |
430 | if (!err) |
431 | allocate_new_segments(sbi); | |
432 | return err; | |
d624c96f JK |
433 | } |
434 | ||
6ead1142 | 435 | int recover_fsync_data(struct f2fs_sb_info *sbi) |
d624c96f JK |
436 | { |
437 | struct list_head inode_list; | |
6ead1142 | 438 | int err; |
aabe5136 | 439 | bool need_writecp = false; |
d624c96f JK |
440 | |
441 | fsync_entry_slab = f2fs_kmem_cache_create("f2fs_fsync_inode_entry", | |
e8512d2e | 442 | sizeof(struct fsync_inode_entry)); |
6bacf52f | 443 | if (!fsync_entry_slab) |
6ead1142 | 444 | return -ENOMEM; |
d624c96f JK |
445 | |
446 | INIT_LIST_HEAD(&inode_list); | |
447 | ||
448 | /* step #1: find fsynced inode numbers */ | |
aabe5136 | 449 | sbi->por_doing = true; |
6ead1142 JK |
450 | err = find_fsync_dnodes(sbi, &inode_list); |
451 | if (err) | |
d624c96f JK |
452 | goto out; |
453 | ||
454 | if (list_empty(&inode_list)) | |
455 | goto out; | |
456 | ||
aabe5136 | 457 | need_writecp = true; |
691c6fd2 | 458 | |
d624c96f | 459 | /* step #2: recover data */ |
6ead1142 | 460 | err = recover_data(sbi, &inode_list, CURSEG_WARM_NODE); |
5d56b671 | 461 | f2fs_bug_on(!list_empty(&inode_list)); |
d624c96f | 462 | out: |
5ebefc5b | 463 | destroy_fsync_dnodes(&inode_list); |
d624c96f | 464 | kmem_cache_destroy(fsync_entry_slab); |
aabe5136 | 465 | sbi->por_doing = false; |
691c6fd2 | 466 | if (!err && need_writecp) |
2c2c149f | 467 | write_checkpoint(sbi, false); |
6ead1142 | 468 | return err; |
d624c96f | 469 | } |