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: |
6c311ec6 CF |
98 | f2fs_msg(inode->i_sb, KERN_NOTICE, |
99 | "%s: ino = %x, name = %s, dir = %lx, err = %d", | |
100 | __func__, ino_of_node(ipage), raw_inode->i_name, | |
f28c06fa | 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); | |
695fd1ed | 139 | blkaddr = NEXT_FREE_BLKADDR(sbi, curseg); |
d624c96f JK |
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); | |
491c0854 | 221 | unsigned short blkoff = GET_BLKOFF_FROM_SEG0(sbi, blkaddr); |
f6517cfc | 222 | struct f2fs_summary_block *sum_node; |
d624c96f | 223 | struct f2fs_summary sum; |
f6517cfc | 224 | struct page *sum_page, *node_page; |
b292dcab | 225 | nid_t ino, nid; |
d624c96f | 226 | struct inode *inode; |
de93653f | 227 | unsigned int offset; |
d624c96f JK |
228 | block_t bidx; |
229 | int i; | |
230 | ||
231 | sentry = get_seg_entry(sbi, segno); | |
232 | if (!f2fs_test_bit(blkoff, sentry->cur_valid_map)) | |
39cf72cf | 233 | return 0; |
d624c96f JK |
234 | |
235 | /* Get the previous summary */ | |
236 | for (i = CURSEG_WARM_DATA; i <= CURSEG_COLD_DATA; i++) { | |
237 | struct curseg_info *curseg = CURSEG_I(sbi, i); | |
238 | if (curseg->segno == segno) { | |
239 | sum = curseg->sum_blk->entries[blkoff]; | |
f6517cfc | 240 | goto got_it; |
d624c96f JK |
241 | } |
242 | } | |
d624c96f | 243 | |
f6517cfc JK |
244 | sum_page = get_sum_page(sbi, segno); |
245 | sum_node = (struct f2fs_summary_block *)page_address(sum_page); | |
246 | sum = sum_node->entries[blkoff]; | |
247 | f2fs_put_page(sum_page, 1); | |
248 | got_it: | |
b292dcab JK |
249 | /* Use the locked dnode page and inode */ |
250 | nid = le32_to_cpu(sum.nid); | |
251 | if (dn->inode->i_ino == nid) { | |
252 | struct dnode_of_data tdn = *dn; | |
253 | tdn.nid = nid; | |
254 | tdn.node_page = dn->inode_page; | |
060dd67b | 255 | tdn.ofs_in_node = le16_to_cpu(sum.ofs_in_node); |
b292dcab | 256 | truncate_data_blocks_range(&tdn, 1); |
39cf72cf | 257 | return 0; |
b292dcab JK |
258 | } else if (dn->nid == nid) { |
259 | struct dnode_of_data tdn = *dn; | |
060dd67b | 260 | tdn.ofs_in_node = le16_to_cpu(sum.ofs_in_node); |
b292dcab | 261 | truncate_data_blocks_range(&tdn, 1); |
39cf72cf | 262 | return 0; |
b292dcab JK |
263 | } |
264 | ||
d624c96f | 265 | /* Get the node page */ |
b292dcab | 266 | node_page = get_node_page(sbi, nid); |
39cf72cf JK |
267 | if (IS_ERR(node_page)) |
268 | return PTR_ERR(node_page); | |
de93653f JK |
269 | |
270 | offset = ofs_of_node(node_page); | |
d624c96f JK |
271 | ino = ino_of_node(node_page); |
272 | f2fs_put_page(node_page, 1); | |
273 | ||
274 | /* Deallocate previous index in the node page */ | |
d4686d56 | 275 | inode = f2fs_iget(sbi->sb, ino); |
06025f4d | 276 | if (IS_ERR(inode)) |
39cf72cf | 277 | return PTR_ERR(inode); |
06025f4d | 278 | |
de93653f JK |
279 | bidx = start_bidx_of_node(offset, F2FS_I(inode)) + |
280 | le16_to_cpu(sum.ofs_in_node); | |
281 | ||
d624c96f JK |
282 | truncate_hole(inode, bidx, bidx + 1); |
283 | iput(inode); | |
39cf72cf | 284 | return 0; |
d624c96f JK |
285 | } |
286 | ||
6ead1142 | 287 | static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode, |
d624c96f JK |
288 | struct page *page, block_t blkaddr) |
289 | { | |
de93653f | 290 | struct f2fs_inode_info *fi = F2FS_I(inode); |
d624c96f JK |
291 | unsigned int start, end; |
292 | struct dnode_of_data dn; | |
293 | struct f2fs_summary sum; | |
294 | struct node_info ni; | |
f356fe0c | 295 | int err = 0, recovered = 0; |
d624c96f | 296 | |
1e1bb4ba JK |
297 | if (recover_inline_data(inode, page)) |
298 | goto out; | |
299 | ||
abb2366c JK |
300 | if (recover_xattr_data(inode, page, blkaddr)) |
301 | goto out; | |
302 | ||
de93653f | 303 | start = start_bidx_of_node(ofs_of_node(page), fi); |
d624c96f | 304 | if (IS_INODE(page)) |
de93653f | 305 | end = start + ADDRS_PER_INODE(fi); |
d624c96f JK |
306 | else |
307 | end = start + ADDRS_PER_BLOCK; | |
308 | ||
e479556b | 309 | f2fs_lock_op(sbi); |
1e1bb4ba | 310 | |
d624c96f | 311 | set_new_dnode(&dn, inode, NULL, NULL, 0); |
39936837 | 312 | |
6ead1142 | 313 | err = get_dnode_of_data(&dn, start, ALLOC_NODE); |
39936837 | 314 | if (err) { |
e479556b | 315 | f2fs_unlock_op(sbi); |
1e1bb4ba | 316 | goto out; |
39936837 | 317 | } |
d624c96f JK |
318 | |
319 | wait_on_page_writeback(dn.node_page); | |
320 | ||
321 | get_node_info(sbi, dn.nid, &ni); | |
5d56b671 JK |
322 | f2fs_bug_on(ni.ino != ino_of_node(page)); |
323 | f2fs_bug_on(ofs_of_node(dn.node_page) != ofs_of_node(page)); | |
d624c96f JK |
324 | |
325 | for (; start < end; start++) { | |
326 | block_t src, dest; | |
327 | ||
328 | src = datablock_addr(dn.node_page, dn.ofs_in_node); | |
329 | dest = datablock_addr(page, dn.ofs_in_node); | |
330 | ||
331 | if (src != dest && dest != NEW_ADDR && dest != NULL_ADDR) { | |
332 | if (src == NULL_ADDR) { | |
5d56b671 | 333 | err = reserve_new_block(&dn); |
d624c96f | 334 | /* We should not get -ENOSPC */ |
5d56b671 | 335 | f2fs_bug_on(err); |
d624c96f JK |
336 | } |
337 | ||
338 | /* Check the previous node page having this index */ | |
39cf72cf JK |
339 | err = check_index_in_prev_nodes(sbi, dest, &dn); |
340 | if (err) | |
341 | goto err; | |
d624c96f JK |
342 | |
343 | set_summary(&sum, dn.nid, dn.ofs_in_node, ni.version); | |
344 | ||
345 | /* write dummy data page */ | |
346 | recover_data_page(sbi, NULL, &sum, src, dest); | |
347 | update_extent_cache(dest, &dn); | |
f356fe0c | 348 | recovered++; |
d624c96f JK |
349 | } |
350 | dn.ofs_in_node++; | |
351 | } | |
352 | ||
353 | /* write node page in place */ | |
354 | set_summary(&sum, dn.nid, 0, 0); | |
355 | if (IS_INODE(dn.node_page)) | |
356 | sync_inode_page(&dn); | |
357 | ||
358 | copy_node_footer(dn.node_page, page); | |
359 | fill_node_footer(dn.node_page, dn.nid, ni.ino, | |
360 | ofs_of_node(page), false); | |
361 | set_page_dirty(dn.node_page); | |
362 | ||
363 | recover_node_page(sbi, dn.node_page, &sum, &ni, blkaddr); | |
39cf72cf | 364 | err: |
d624c96f | 365 | f2fs_put_dnode(&dn); |
e479556b | 366 | f2fs_unlock_op(sbi); |
1e1bb4ba | 367 | out: |
6c311ec6 CF |
368 | f2fs_msg(sbi->sb, KERN_NOTICE, |
369 | "recover_data: ino = %lx, recovered = %d blocks, err = %d", | |
370 | inode->i_ino, recovered, err); | |
39cf72cf | 371 | return err; |
d624c96f JK |
372 | } |
373 | ||
6ead1142 | 374 | static int recover_data(struct f2fs_sb_info *sbi, |
d624c96f JK |
375 | struct list_head *head, int type) |
376 | { | |
d71b5564 | 377 | unsigned long long cp_ver = cur_cp_version(F2FS_CKPT(sbi)); |
d624c96f JK |
378 | struct curseg_info *curseg; |
379 | struct page *page; | |
6ead1142 | 380 | int err = 0; |
d624c96f JK |
381 | block_t blkaddr; |
382 | ||
383 | /* get node pages in the current segment */ | |
384 | curseg = CURSEG_I(sbi, type); | |
385 | blkaddr = NEXT_FREE_BLKADDR(sbi, curseg); | |
386 | ||
387 | /* read node page */ | |
a0acdfe0 | 388 | page = alloc_page(GFP_F2FS_ZERO); |
e27dae4d | 389 | if (!page) |
6ead1142 JK |
390 | return -ENOMEM; |
391 | ||
d624c96f JK |
392 | lock_page(page); |
393 | ||
394 | while (1) { | |
395 | struct fsync_inode_entry *entry; | |
396 | ||
93dfe2ac | 397 | err = f2fs_submit_page_bio(sbi, page, blkaddr, READ_SYNC); |
6ead1142 | 398 | if (err) |
b9987a27 | 399 | return err; |
d624c96f | 400 | |
393ff91f JK |
401 | lock_page(page); |
402 | ||
d624c96f | 403 | if (cp_ver != cpver_of_node(page)) |
45856aff | 404 | break; |
d624c96f JK |
405 | |
406 | entry = get_fsync_inode(head, ino_of_node(page)); | |
407 | if (!entry) | |
408 | goto next; | |
409 | ||
6ead1142 JK |
410 | err = do_recover_data(sbi, entry->inode, page, blkaddr); |
411 | if (err) | |
45856aff | 412 | break; |
d624c96f JK |
413 | |
414 | if (entry->blkaddr == blkaddr) { | |
415 | iput(entry->inode); | |
416 | list_del(&entry->list); | |
417 | kmem_cache_free(fsync_entry_slab, entry); | |
418 | } | |
419 | next: | |
420 | /* check next segment */ | |
421 | blkaddr = next_blkaddr_of_node(page); | |
d624c96f | 422 | } |
b9987a27 | 423 | |
d624c96f JK |
424 | unlock_page(page); |
425 | __free_pages(page, 0); | |
426 | ||
6ead1142 JK |
427 | if (!err) |
428 | allocate_new_segments(sbi); | |
429 | return err; | |
d624c96f JK |
430 | } |
431 | ||
6ead1142 | 432 | int recover_fsync_data(struct f2fs_sb_info *sbi) |
d624c96f JK |
433 | { |
434 | struct list_head inode_list; | |
6ead1142 | 435 | int err; |
aabe5136 | 436 | bool need_writecp = false; |
d624c96f JK |
437 | |
438 | fsync_entry_slab = f2fs_kmem_cache_create("f2fs_fsync_inode_entry", | |
439 | sizeof(struct fsync_inode_entry), NULL); | |
6bacf52f | 440 | if (!fsync_entry_slab) |
6ead1142 | 441 | return -ENOMEM; |
d624c96f JK |
442 | |
443 | INIT_LIST_HEAD(&inode_list); | |
444 | ||
445 | /* step #1: find fsynced inode numbers */ | |
aabe5136 | 446 | sbi->por_doing = true; |
6ead1142 JK |
447 | err = find_fsync_dnodes(sbi, &inode_list); |
448 | if (err) | |
d624c96f JK |
449 | goto out; |
450 | ||
451 | if (list_empty(&inode_list)) | |
452 | goto out; | |
453 | ||
aabe5136 | 454 | need_writecp = true; |
691c6fd2 | 455 | |
d624c96f | 456 | /* step #2: recover data */ |
6ead1142 | 457 | err = recover_data(sbi, &inode_list, CURSEG_WARM_NODE); |
5d56b671 | 458 | f2fs_bug_on(!list_empty(&inode_list)); |
d624c96f | 459 | out: |
5ebefc5b | 460 | destroy_fsync_dnodes(&inode_list); |
d624c96f | 461 | kmem_cache_destroy(fsync_entry_slab); |
aabe5136 | 462 | sbi->por_doing = false; |
691c6fd2 | 463 | if (!err && need_writecp) |
2c2c149f | 464 | write_checkpoint(sbi, false); |
6ead1142 | 465 | return err; |
d624c96f | 466 | } |