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 | ||
74d0b917 JK |
49 | dir = check_dirty_dir_inode(F2FS_SB(inode->i_sb), pino); |
50 | if (!dir) { | |
51 | dir = f2fs_iget(inode->i_sb, pino); | |
52 | if (IS_ERR(dir)) { | |
53 | err = PTR_ERR(dir); | |
54 | goto out; | |
55 | } | |
56 | set_inode_flag(F2FS_I(dir), FI_DELAY_IPUT); | |
5deb8267 | 57 | add_dirty_dir_inode(dir); |
d624c96f JK |
58 | } |
59 | ||
b7f7a5e0 AV |
60 | name.len = le32_to_cpu(raw_inode->i_namelen); |
61 | name.name = raw_inode->i_name; | |
d96b1431 CY |
62 | |
63 | if (unlikely(name.len > F2FS_NAME_LEN)) { | |
64 | WARN_ON(1); | |
65 | err = -ENAMETOOLONG; | |
66 | goto out; | |
67 | } | |
6b8213d9 JK |
68 | retry: |
69 | de = f2fs_find_entry(dir, &name, &page); | |
2e5558f4 RK |
70 | if (de && inode->i_ino == le32_to_cpu(de->ino)) |
71 | goto out_unmap_put; | |
6b8213d9 JK |
72 | if (de) { |
73 | einode = f2fs_iget(inode->i_sb, le32_to_cpu(de->ino)); | |
74 | if (IS_ERR(einode)) { | |
75 | WARN_ON(1); | |
76 | if (PTR_ERR(einode) == -ENOENT) | |
77 | err = -EEXIST; | |
2e5558f4 RK |
78 | goto out_unmap_put; |
79 | } | |
80 | err = acquire_orphan_inode(F2FS_SB(inode->i_sb)); | |
81 | if (err) { | |
82 | iput(einode); | |
83 | goto out_unmap_put; | |
6b8213d9 JK |
84 | } |
85 | f2fs_delete_entry(de, page, einode); | |
86 | iput(einode); | |
87 | goto retry; | |
d624c96f | 88 | } |
6b8213d9 | 89 | err = __f2fs_add_link(dir, &name, inode); |
2e5558f4 RK |
90 | goto out; |
91 | ||
92 | out_unmap_put: | |
93 | kunmap(page); | |
94 | f2fs_put_page(page, 0); | |
d624c96f | 95 | out: |
6c311ec6 CF |
96 | f2fs_msg(inode->i_sb, KERN_NOTICE, |
97 | "%s: ino = %x, name = %s, dir = %lx, err = %d", | |
98 | __func__, ino_of_node(ipage), raw_inode->i_name, | |
f28c06fa | 99 | IS_ERR(dir) ? 0 : dir->i_ino, err); |
d624c96f JK |
100 | return err; |
101 | } | |
102 | ||
103 | static int recover_inode(struct inode *inode, struct page *node_page) | |
104 | { | |
58bfaf44 | 105 | struct f2fs_inode *raw_inode = F2FS_INODE(node_page); |
d624c96f | 106 | |
f356fe0c JK |
107 | if (!IS_INODE(node_page)) |
108 | return 0; | |
109 | ||
25ca923b | 110 | inode->i_mode = le16_to_cpu(raw_inode->i_mode); |
d624c96f JK |
111 | i_size_write(inode, le64_to_cpu(raw_inode->i_size)); |
112 | inode->i_atime.tv_sec = le64_to_cpu(raw_inode->i_mtime); | |
113 | inode->i_ctime.tv_sec = le64_to_cpu(raw_inode->i_ctime); | |
114 | inode->i_mtime.tv_sec = le64_to_cpu(raw_inode->i_mtime); | |
115 | inode->i_atime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); | |
116 | inode->i_ctime.tv_nsec = le32_to_cpu(raw_inode->i_ctime_nsec); | |
117 | inode->i_mtime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); | |
118 | ||
f356fe0c JK |
119 | if (is_dent_dnode(node_page)) |
120 | return recover_dentry(node_page, inode); | |
121 | ||
122 | f2fs_msg(inode->i_sb, KERN_NOTICE, "recover_inode: ino = %x, name = %s", | |
123 | ino_of_node(node_page), raw_inode->i_name); | |
124 | return 0; | |
d624c96f JK |
125 | } |
126 | ||
127 | static int find_fsync_dnodes(struct f2fs_sb_info *sbi, struct list_head *head) | |
128 | { | |
d71b5564 | 129 | unsigned long long cp_ver = cur_cp_version(F2FS_CKPT(sbi)); |
d624c96f JK |
130 | struct curseg_info *curseg; |
131 | struct page *page; | |
132 | block_t blkaddr; | |
133 | int err = 0; | |
134 | ||
135 | /* get node pages in the current segment */ | |
136 | curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); | |
695fd1ed | 137 | blkaddr = NEXT_FREE_BLKADDR(sbi, curseg); |
d624c96f JK |
138 | |
139 | /* read node page */ | |
140 | page = alloc_page(GFP_F2FS_ZERO); | |
e27dae4d DC |
141 | if (!page) |
142 | return -ENOMEM; | |
d624c96f JK |
143 | lock_page(page); |
144 | ||
145 | while (1) { | |
146 | struct fsync_inode_entry *entry; | |
147 | ||
93dfe2ac | 148 | err = f2fs_submit_page_bio(sbi, page, blkaddr, READ_SYNC); |
393ff91f | 149 | if (err) |
b9987a27 | 150 | return err; |
d624c96f | 151 | |
393ff91f JK |
152 | lock_page(page); |
153 | ||
6ead1142 | 154 | if (cp_ver != cpver_of_node(page)) |
f356fe0c | 155 | break; |
d624c96f JK |
156 | |
157 | if (!is_fsync_dnode(page)) | |
158 | goto next; | |
159 | ||
160 | entry = get_fsync_inode(head, ino_of_node(page)); | |
161 | if (entry) { | |
d624c96f JK |
162 | if (IS_INODE(page) && is_dent_dnode(page)) |
163 | set_inode_flag(F2FS_I(entry->inode), | |
164 | FI_INC_LINK); | |
165 | } else { | |
166 | if (IS_INODE(page) && is_dent_dnode(page)) { | |
6ead1142 JK |
167 | err = recover_inode_page(sbi, page); |
168 | if (err) | |
f356fe0c | 169 | break; |
d624c96f JK |
170 | } |
171 | ||
172 | /* add this fsync inode to the list */ | |
173 | entry = kmem_cache_alloc(fsync_entry_slab, GFP_NOFS); | |
174 | if (!entry) { | |
175 | err = -ENOMEM; | |
f356fe0c | 176 | break; |
d624c96f JK |
177 | } |
178 | ||
d624c96f JK |
179 | entry->inode = f2fs_iget(sbi->sb, ino_of_node(page)); |
180 | if (IS_ERR(entry->inode)) { | |
181 | err = PTR_ERR(entry->inode); | |
fd8bb65f | 182 | kmem_cache_free(fsync_entry_slab, entry); |
f356fe0c | 183 | break; |
d624c96f | 184 | } |
fd8bb65f | 185 | list_add_tail(&entry->list, head); |
d624c96f | 186 | } |
addbe45b JK |
187 | entry->blkaddr = blkaddr; |
188 | ||
f356fe0c JK |
189 | err = recover_inode(entry->inode, page); |
190 | if (err && err != -ENOENT) | |
191 | break; | |
d624c96f JK |
192 | next: |
193 | /* check next segment */ | |
194 | blkaddr = next_blkaddr_of_node(page); | |
d624c96f | 195 | } |
b9987a27 | 196 | |
d624c96f JK |
197 | unlock_page(page); |
198 | __free_pages(page, 0); | |
b9987a27 | 199 | |
d624c96f JK |
200 | return err; |
201 | } | |
202 | ||
5ebefc5b | 203 | static void destroy_fsync_dnodes(struct list_head *head) |
d624c96f | 204 | { |
d8b79b2f DC |
205 | struct fsync_inode_entry *entry, *tmp; |
206 | ||
207 | list_for_each_entry_safe(entry, tmp, head, list) { | |
d624c96f JK |
208 | iput(entry->inode); |
209 | list_del(&entry->list); | |
210 | kmem_cache_free(fsync_entry_slab, entry); | |
211 | } | |
212 | } | |
213 | ||
39cf72cf | 214 | static int check_index_in_prev_nodes(struct f2fs_sb_info *sbi, |
b292dcab | 215 | block_t blkaddr, struct dnode_of_data *dn) |
d624c96f JK |
216 | { |
217 | struct seg_entry *sentry; | |
218 | unsigned int segno = GET_SEGNO(sbi, blkaddr); | |
491c0854 | 219 | unsigned short blkoff = GET_BLKOFF_FROM_SEG0(sbi, blkaddr); |
f6517cfc | 220 | struct f2fs_summary_block *sum_node; |
d624c96f | 221 | struct f2fs_summary sum; |
f6517cfc | 222 | struct page *sum_page, *node_page; |
b292dcab | 223 | nid_t ino, nid; |
d624c96f | 224 | struct inode *inode; |
de93653f | 225 | unsigned int offset; |
d624c96f JK |
226 | block_t bidx; |
227 | int i; | |
228 | ||
229 | sentry = get_seg_entry(sbi, segno); | |
230 | if (!f2fs_test_bit(blkoff, sentry->cur_valid_map)) | |
39cf72cf | 231 | return 0; |
d624c96f JK |
232 | |
233 | /* Get the previous summary */ | |
234 | for (i = CURSEG_WARM_DATA; i <= CURSEG_COLD_DATA; i++) { | |
235 | struct curseg_info *curseg = CURSEG_I(sbi, i); | |
236 | if (curseg->segno == segno) { | |
237 | sum = curseg->sum_blk->entries[blkoff]; | |
f6517cfc | 238 | goto got_it; |
d624c96f JK |
239 | } |
240 | } | |
d624c96f | 241 | |
f6517cfc JK |
242 | sum_page = get_sum_page(sbi, segno); |
243 | sum_node = (struct f2fs_summary_block *)page_address(sum_page); | |
244 | sum = sum_node->entries[blkoff]; | |
245 | f2fs_put_page(sum_page, 1); | |
246 | got_it: | |
b292dcab JK |
247 | /* Use the locked dnode page and inode */ |
248 | nid = le32_to_cpu(sum.nid); | |
249 | if (dn->inode->i_ino == nid) { | |
250 | struct dnode_of_data tdn = *dn; | |
251 | tdn.nid = nid; | |
252 | tdn.node_page = dn->inode_page; | |
060dd67b | 253 | tdn.ofs_in_node = le16_to_cpu(sum.ofs_in_node); |
b292dcab | 254 | truncate_data_blocks_range(&tdn, 1); |
39cf72cf | 255 | return 0; |
b292dcab JK |
256 | } else if (dn->nid == nid) { |
257 | struct dnode_of_data tdn = *dn; | |
060dd67b | 258 | tdn.ofs_in_node = le16_to_cpu(sum.ofs_in_node); |
b292dcab | 259 | truncate_data_blocks_range(&tdn, 1); |
39cf72cf | 260 | return 0; |
b292dcab JK |
261 | } |
262 | ||
d624c96f | 263 | /* Get the node page */ |
b292dcab | 264 | node_page = get_node_page(sbi, nid); |
39cf72cf JK |
265 | if (IS_ERR(node_page)) |
266 | return PTR_ERR(node_page); | |
de93653f JK |
267 | |
268 | offset = ofs_of_node(node_page); | |
d624c96f JK |
269 | ino = ino_of_node(node_page); |
270 | f2fs_put_page(node_page, 1); | |
271 | ||
272 | /* Deallocate previous index in the node page */ | |
d4686d56 | 273 | inode = f2fs_iget(sbi->sb, ino); |
06025f4d | 274 | if (IS_ERR(inode)) |
39cf72cf | 275 | return PTR_ERR(inode); |
06025f4d | 276 | |
de93653f JK |
277 | bidx = start_bidx_of_node(offset, F2FS_I(inode)) + |
278 | le16_to_cpu(sum.ofs_in_node); | |
279 | ||
d624c96f JK |
280 | truncate_hole(inode, bidx, bidx + 1); |
281 | iput(inode); | |
39cf72cf | 282 | return 0; |
d624c96f JK |
283 | } |
284 | ||
6ead1142 | 285 | static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode, |
d624c96f JK |
286 | struct page *page, block_t blkaddr) |
287 | { | |
de93653f | 288 | struct f2fs_inode_info *fi = F2FS_I(inode); |
d624c96f JK |
289 | unsigned int start, end; |
290 | struct dnode_of_data dn; | |
291 | struct f2fs_summary sum; | |
292 | struct node_info ni; | |
f356fe0c | 293 | int err = 0, recovered = 0; |
d624c96f | 294 | |
1e1bb4ba JK |
295 | if (recover_inline_data(inode, page)) |
296 | goto out; | |
297 | ||
abb2366c JK |
298 | if (recover_xattr_data(inode, page, blkaddr)) |
299 | goto out; | |
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); |
1e1bb4ba | 308 | |
d624c96f | 309 | set_new_dnode(&dn, inode, NULL, NULL, 0); |
39936837 | 310 | |
6ead1142 | 311 | err = get_dnode_of_data(&dn, start, ALLOC_NODE); |
39936837 | 312 | if (err) { |
e479556b | 313 | f2fs_unlock_op(sbi); |
1e1bb4ba | 314 | goto out; |
39936837 | 315 | } |
d624c96f | 316 | |
3cb5ad15 | 317 | f2fs_wait_on_page_writeback(dn.node_page, NODE); |
d624c96f JK |
318 | |
319 | get_node_info(sbi, dn.nid, &ni); | |
5d56b671 JK |
320 | f2fs_bug_on(ni.ino != ino_of_node(page)); |
321 | f2fs_bug_on(ofs_of_node(dn.node_page) != ofs_of_node(page)); | |
d624c96f JK |
322 | |
323 | for (; start < end; start++) { | |
324 | block_t src, dest; | |
325 | ||
326 | src = datablock_addr(dn.node_page, dn.ofs_in_node); | |
327 | dest = datablock_addr(page, dn.ofs_in_node); | |
328 | ||
329 | if (src != dest && dest != NEW_ADDR && dest != NULL_ADDR) { | |
330 | if (src == NULL_ADDR) { | |
5d56b671 | 331 | err = reserve_new_block(&dn); |
d624c96f | 332 | /* We should not get -ENOSPC */ |
5d56b671 | 333 | f2fs_bug_on(err); |
d624c96f JK |
334 | } |
335 | ||
336 | /* Check the previous node page having this index */ | |
39cf72cf JK |
337 | err = check_index_in_prev_nodes(sbi, dest, &dn); |
338 | if (err) | |
339 | goto err; | |
d624c96f JK |
340 | |
341 | set_summary(&sum, dn.nid, dn.ofs_in_node, ni.version); | |
342 | ||
343 | /* write dummy data page */ | |
344 | recover_data_page(sbi, NULL, &sum, src, dest); | |
345 | update_extent_cache(dest, &dn); | |
f356fe0c | 346 | recovered++; |
d624c96f JK |
347 | } |
348 | dn.ofs_in_node++; | |
349 | } | |
350 | ||
351 | /* write node page in place */ | |
352 | set_summary(&sum, dn.nid, 0, 0); | |
353 | if (IS_INODE(dn.node_page)) | |
354 | sync_inode_page(&dn); | |
355 | ||
356 | copy_node_footer(dn.node_page, page); | |
357 | fill_node_footer(dn.node_page, dn.nid, ni.ino, | |
358 | ofs_of_node(page), false); | |
359 | set_page_dirty(dn.node_page); | |
360 | ||
361 | recover_node_page(sbi, dn.node_page, &sum, &ni, blkaddr); | |
39cf72cf | 362 | err: |
d624c96f | 363 | f2fs_put_dnode(&dn); |
e479556b | 364 | f2fs_unlock_op(sbi); |
1e1bb4ba | 365 | out: |
6c311ec6 CF |
366 | f2fs_msg(sbi->sb, KERN_NOTICE, |
367 | "recover_data: ino = %lx, recovered = %d blocks, err = %d", | |
368 | inode->i_ino, recovered, err); | |
39cf72cf | 369 | return err; |
d624c96f JK |
370 | } |
371 | ||
6ead1142 | 372 | static int recover_data(struct f2fs_sb_info *sbi, |
d624c96f JK |
373 | struct list_head *head, int type) |
374 | { | |
d71b5564 | 375 | unsigned long long cp_ver = cur_cp_version(F2FS_CKPT(sbi)); |
d624c96f JK |
376 | struct curseg_info *curseg; |
377 | struct page *page; | |
6ead1142 | 378 | int err = 0; |
d624c96f JK |
379 | block_t blkaddr; |
380 | ||
381 | /* get node pages in the current segment */ | |
382 | curseg = CURSEG_I(sbi, type); | |
383 | blkaddr = NEXT_FREE_BLKADDR(sbi, curseg); | |
384 | ||
385 | /* read node page */ | |
a0acdfe0 | 386 | page = alloc_page(GFP_F2FS_ZERO); |
e27dae4d | 387 | if (!page) |
6ead1142 JK |
388 | return -ENOMEM; |
389 | ||
d624c96f JK |
390 | lock_page(page); |
391 | ||
392 | while (1) { | |
393 | struct fsync_inode_entry *entry; | |
394 | ||
93dfe2ac | 395 | err = f2fs_submit_page_bio(sbi, page, blkaddr, READ_SYNC); |
6ead1142 | 396 | if (err) |
b9987a27 | 397 | return err; |
d624c96f | 398 | |
393ff91f JK |
399 | lock_page(page); |
400 | ||
d624c96f | 401 | if (cp_ver != cpver_of_node(page)) |
45856aff | 402 | break; |
d624c96f JK |
403 | |
404 | entry = get_fsync_inode(head, ino_of_node(page)); | |
405 | if (!entry) | |
406 | goto next; | |
407 | ||
6ead1142 JK |
408 | err = do_recover_data(sbi, entry->inode, page, blkaddr); |
409 | if (err) | |
45856aff | 410 | break; |
d624c96f JK |
411 | |
412 | if (entry->blkaddr == blkaddr) { | |
413 | iput(entry->inode); | |
414 | list_del(&entry->list); | |
415 | kmem_cache_free(fsync_entry_slab, entry); | |
416 | } | |
417 | next: | |
418 | /* check next segment */ | |
419 | blkaddr = next_blkaddr_of_node(page); | |
d624c96f | 420 | } |
b9987a27 | 421 | |
d624c96f JK |
422 | unlock_page(page); |
423 | __free_pages(page, 0); | |
424 | ||
6ead1142 JK |
425 | if (!err) |
426 | allocate_new_segments(sbi); | |
427 | return err; | |
d624c96f JK |
428 | } |
429 | ||
6ead1142 | 430 | int recover_fsync_data(struct f2fs_sb_info *sbi) |
d624c96f JK |
431 | { |
432 | struct list_head inode_list; | |
6ead1142 | 433 | int err; |
aabe5136 | 434 | bool need_writecp = false; |
d624c96f JK |
435 | |
436 | fsync_entry_slab = f2fs_kmem_cache_create("f2fs_fsync_inode_entry", | |
e8512d2e | 437 | sizeof(struct fsync_inode_entry)); |
6bacf52f | 438 | if (!fsync_entry_slab) |
6ead1142 | 439 | return -ENOMEM; |
d624c96f JK |
440 | |
441 | INIT_LIST_HEAD(&inode_list); | |
442 | ||
443 | /* step #1: find fsynced inode numbers */ | |
aabe5136 | 444 | sbi->por_doing = true; |
6ead1142 JK |
445 | err = find_fsync_dnodes(sbi, &inode_list); |
446 | if (err) | |
d624c96f JK |
447 | goto out; |
448 | ||
449 | if (list_empty(&inode_list)) | |
450 | goto out; | |
451 | ||
aabe5136 | 452 | need_writecp = true; |
691c6fd2 | 453 | |
d624c96f | 454 | /* step #2: recover data */ |
6ead1142 | 455 | err = recover_data(sbi, &inode_list, CURSEG_WARM_NODE); |
5d56b671 | 456 | f2fs_bug_on(!list_empty(&inode_list)); |
d624c96f | 457 | out: |
5ebefc5b | 458 | destroy_fsync_dnodes(&inode_list); |
d624c96f | 459 | kmem_cache_destroy(fsync_entry_slab); |
aabe5136 | 460 | sbi->por_doing = false; |
691c6fd2 | 461 | if (!err && need_writecp) |
2c2c149f | 462 | write_checkpoint(sbi, false); |
6ead1142 | 463 | return err; |
d624c96f | 464 | } |