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); | |
695facc0 JK |
65 | if (de && inode->i_ino == le32_to_cpu(de->ino)) { |
66 | clear_inode_flag(F2FS_I(inode), FI_INC_LINK); | |
2e5558f4 | 67 | goto out_unmap_put; |
695facc0 | 68 | } |
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); | |
5c1f9927 CY |
73 | err = PTR_ERR(einode); |
74 | if (err == -ENOENT) | |
6b8213d9 | 75 | err = -EEXIST; |
2e5558f4 RK |
76 | goto out_unmap_put; |
77 | } | |
78 | err = acquire_orphan_inode(F2FS_SB(inode->i_sb)); | |
79 | if (err) { | |
80 | iput(einode); | |
81 | goto out_unmap_put; | |
6b8213d9 JK |
82 | } |
83 | f2fs_delete_entry(de, page, einode); | |
84 | iput(einode); | |
85 | goto retry; | |
d624c96f | 86 | } |
6b8213d9 | 87 | err = __f2fs_add_link(dir, &name, inode); |
86928f98 JK |
88 | if (err) |
89 | goto out_err; | |
90 | ||
91 | if (is_inode_flag_set(F2FS_I(dir), FI_DELAY_IPUT)) { | |
92 | iput(dir); | |
93 | } else { | |
94 | add_dirty_dir_inode(dir); | |
95 | set_inode_flag(F2FS_I(dir), FI_DELAY_IPUT); | |
96 | } | |
97 | ||
2e5558f4 RK |
98 | goto out; |
99 | ||
100 | out_unmap_put: | |
101 | kunmap(page); | |
102 | f2fs_put_page(page, 0); | |
86928f98 JK |
103 | out_err: |
104 | iput(dir); | |
d624c96f | 105 | out: |
6c311ec6 CF |
106 | f2fs_msg(inode->i_sb, KERN_NOTICE, |
107 | "%s: ino = %x, name = %s, dir = %lx, err = %d", | |
108 | __func__, ino_of_node(ipage), raw_inode->i_name, | |
f28c06fa | 109 | IS_ERR(dir) ? 0 : dir->i_ino, err); |
d624c96f JK |
110 | return err; |
111 | } | |
112 | ||
113 | static int recover_inode(struct inode *inode, struct page *node_page) | |
114 | { | |
58bfaf44 | 115 | struct f2fs_inode *raw_inode = F2FS_INODE(node_page); |
d624c96f | 116 | |
f356fe0c JK |
117 | if (!IS_INODE(node_page)) |
118 | return 0; | |
119 | ||
25ca923b | 120 | inode->i_mode = le16_to_cpu(raw_inode->i_mode); |
d624c96f JK |
121 | i_size_write(inode, le64_to_cpu(raw_inode->i_size)); |
122 | inode->i_atime.tv_sec = le64_to_cpu(raw_inode->i_mtime); | |
123 | inode->i_ctime.tv_sec = le64_to_cpu(raw_inode->i_ctime); | |
124 | inode->i_mtime.tv_sec = le64_to_cpu(raw_inode->i_mtime); | |
125 | inode->i_atime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); | |
126 | inode->i_ctime.tv_nsec = le32_to_cpu(raw_inode->i_ctime_nsec); | |
127 | inode->i_mtime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); | |
128 | ||
f356fe0c JK |
129 | if (is_dent_dnode(node_page)) |
130 | return recover_dentry(node_page, inode); | |
131 | ||
132 | f2fs_msg(inode->i_sb, KERN_NOTICE, "recover_inode: ino = %x, name = %s", | |
133 | ino_of_node(node_page), raw_inode->i_name); | |
134 | return 0; | |
d624c96f JK |
135 | } |
136 | ||
137 | static int find_fsync_dnodes(struct f2fs_sb_info *sbi, struct list_head *head) | |
138 | { | |
d71b5564 | 139 | unsigned long long cp_ver = cur_cp_version(F2FS_CKPT(sbi)); |
d624c96f JK |
140 | struct curseg_info *curseg; |
141 | struct page *page; | |
142 | block_t blkaddr; | |
143 | int err = 0; | |
144 | ||
145 | /* get node pages in the current segment */ | |
146 | curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); | |
695fd1ed | 147 | blkaddr = NEXT_FREE_BLKADDR(sbi, curseg); |
d624c96f JK |
148 | |
149 | /* read node page */ | |
150 | page = alloc_page(GFP_F2FS_ZERO); | |
e27dae4d DC |
151 | if (!page) |
152 | return -ENOMEM; | |
d624c96f JK |
153 | lock_page(page); |
154 | ||
155 | while (1) { | |
156 | struct fsync_inode_entry *entry; | |
157 | ||
93dfe2ac | 158 | err = f2fs_submit_page_bio(sbi, page, blkaddr, READ_SYNC); |
393ff91f | 159 | if (err) |
b9987a27 | 160 | return err; |
d624c96f | 161 | |
393ff91f JK |
162 | lock_page(page); |
163 | ||
6ead1142 | 164 | if (cp_ver != cpver_of_node(page)) |
f356fe0c | 165 | break; |
d624c96f JK |
166 | |
167 | if (!is_fsync_dnode(page)) | |
168 | goto next; | |
169 | ||
170 | entry = get_fsync_inode(head, ino_of_node(page)); | |
171 | if (entry) { | |
d624c96f JK |
172 | if (IS_INODE(page) && is_dent_dnode(page)) |
173 | set_inode_flag(F2FS_I(entry->inode), | |
174 | FI_INC_LINK); | |
175 | } else { | |
176 | if (IS_INODE(page) && is_dent_dnode(page)) { | |
6ead1142 JK |
177 | err = recover_inode_page(sbi, page); |
178 | if (err) | |
f356fe0c | 179 | break; |
d624c96f JK |
180 | } |
181 | ||
182 | /* add this fsync inode to the list */ | |
183 | entry = kmem_cache_alloc(fsync_entry_slab, GFP_NOFS); | |
184 | if (!entry) { | |
185 | err = -ENOMEM; | |
f356fe0c | 186 | break; |
d624c96f JK |
187 | } |
188 | ||
d624c96f JK |
189 | entry->inode = f2fs_iget(sbi->sb, ino_of_node(page)); |
190 | if (IS_ERR(entry->inode)) { | |
191 | err = PTR_ERR(entry->inode); | |
fd8bb65f | 192 | kmem_cache_free(fsync_entry_slab, entry); |
f356fe0c | 193 | break; |
d624c96f | 194 | } |
fd8bb65f | 195 | list_add_tail(&entry->list, head); |
d624c96f | 196 | } |
addbe45b JK |
197 | entry->blkaddr = blkaddr; |
198 | ||
f356fe0c JK |
199 | err = recover_inode(entry->inode, page); |
200 | if (err && err != -ENOENT) | |
201 | break; | |
d624c96f JK |
202 | next: |
203 | /* check next segment */ | |
204 | blkaddr = next_blkaddr_of_node(page); | |
d624c96f | 205 | } |
b9987a27 | 206 | |
d624c96f JK |
207 | unlock_page(page); |
208 | __free_pages(page, 0); | |
b9987a27 | 209 | |
d624c96f JK |
210 | return err; |
211 | } | |
212 | ||
5ebefc5b | 213 | static void destroy_fsync_dnodes(struct list_head *head) |
d624c96f | 214 | { |
d8b79b2f DC |
215 | struct fsync_inode_entry *entry, *tmp; |
216 | ||
217 | list_for_each_entry_safe(entry, tmp, head, list) { | |
d624c96f JK |
218 | iput(entry->inode); |
219 | list_del(&entry->list); | |
220 | kmem_cache_free(fsync_entry_slab, entry); | |
221 | } | |
222 | } | |
223 | ||
39cf72cf | 224 | static int check_index_in_prev_nodes(struct f2fs_sb_info *sbi, |
b292dcab | 225 | block_t blkaddr, struct dnode_of_data *dn) |
d624c96f JK |
226 | { |
227 | struct seg_entry *sentry; | |
228 | unsigned int segno = GET_SEGNO(sbi, blkaddr); | |
491c0854 | 229 | unsigned short blkoff = GET_BLKOFF_FROM_SEG0(sbi, blkaddr); |
f6517cfc | 230 | struct f2fs_summary_block *sum_node; |
d624c96f | 231 | struct f2fs_summary sum; |
f6517cfc | 232 | struct page *sum_page, *node_page; |
b292dcab | 233 | nid_t ino, nid; |
d624c96f | 234 | struct inode *inode; |
de93653f | 235 | unsigned int offset; |
d624c96f JK |
236 | block_t bidx; |
237 | int i; | |
238 | ||
239 | sentry = get_seg_entry(sbi, segno); | |
240 | if (!f2fs_test_bit(blkoff, sentry->cur_valid_map)) | |
39cf72cf | 241 | return 0; |
d624c96f JK |
242 | |
243 | /* Get the previous summary */ | |
244 | for (i = CURSEG_WARM_DATA; i <= CURSEG_COLD_DATA; i++) { | |
245 | struct curseg_info *curseg = CURSEG_I(sbi, i); | |
246 | if (curseg->segno == segno) { | |
247 | sum = curseg->sum_blk->entries[blkoff]; | |
f6517cfc | 248 | goto got_it; |
d624c96f JK |
249 | } |
250 | } | |
d624c96f | 251 | |
f6517cfc JK |
252 | sum_page = get_sum_page(sbi, segno); |
253 | sum_node = (struct f2fs_summary_block *)page_address(sum_page); | |
254 | sum = sum_node->entries[blkoff]; | |
255 | f2fs_put_page(sum_page, 1); | |
256 | got_it: | |
b292dcab JK |
257 | /* Use the locked dnode page and inode */ |
258 | nid = le32_to_cpu(sum.nid); | |
259 | if (dn->inode->i_ino == nid) { | |
260 | struct dnode_of_data tdn = *dn; | |
261 | tdn.nid = nid; | |
262 | tdn.node_page = dn->inode_page; | |
060dd67b | 263 | tdn.ofs_in_node = le16_to_cpu(sum.ofs_in_node); |
b292dcab | 264 | truncate_data_blocks_range(&tdn, 1); |
39cf72cf | 265 | return 0; |
b292dcab JK |
266 | } else if (dn->nid == nid) { |
267 | struct dnode_of_data tdn = *dn; | |
060dd67b | 268 | tdn.ofs_in_node = le16_to_cpu(sum.ofs_in_node); |
b292dcab | 269 | truncate_data_blocks_range(&tdn, 1); |
39cf72cf | 270 | return 0; |
b292dcab JK |
271 | } |
272 | ||
d624c96f | 273 | /* Get the node page */ |
b292dcab | 274 | node_page = get_node_page(sbi, nid); |
39cf72cf JK |
275 | if (IS_ERR(node_page)) |
276 | return PTR_ERR(node_page); | |
de93653f JK |
277 | |
278 | offset = ofs_of_node(node_page); | |
d624c96f JK |
279 | ino = ino_of_node(node_page); |
280 | f2fs_put_page(node_page, 1); | |
281 | ||
282 | /* Deallocate previous index in the node page */ | |
d4686d56 | 283 | inode = f2fs_iget(sbi->sb, ino); |
06025f4d | 284 | if (IS_ERR(inode)) |
39cf72cf | 285 | return PTR_ERR(inode); |
06025f4d | 286 | |
de93653f JK |
287 | bidx = start_bidx_of_node(offset, F2FS_I(inode)) + |
288 | le16_to_cpu(sum.ofs_in_node); | |
289 | ||
d624c96f JK |
290 | truncate_hole(inode, bidx, bidx + 1); |
291 | iput(inode); | |
39cf72cf | 292 | return 0; |
d624c96f JK |
293 | } |
294 | ||
6ead1142 | 295 | static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode, |
d624c96f JK |
296 | struct page *page, block_t blkaddr) |
297 | { | |
de93653f | 298 | struct f2fs_inode_info *fi = F2FS_I(inode); |
d624c96f JK |
299 | unsigned int start, end; |
300 | struct dnode_of_data dn; | |
301 | struct f2fs_summary sum; | |
302 | struct node_info ni; | |
f356fe0c | 303 | int err = 0, recovered = 0; |
d624c96f | 304 | |
70cfed88 CY |
305 | recover_inline_xattr(inode, page); |
306 | ||
1e1bb4ba JK |
307 | if (recover_inline_data(inode, page)) |
308 | goto out; | |
309 | ||
abb2366c JK |
310 | if (recover_xattr_data(inode, page, blkaddr)) |
311 | goto out; | |
312 | ||
de93653f | 313 | start = start_bidx_of_node(ofs_of_node(page), fi); |
6403eb1f | 314 | end = start + ADDRS_PER_PAGE(page, fi); |
d624c96f | 315 | |
e479556b | 316 | f2fs_lock_op(sbi); |
1e1bb4ba | 317 | |
d624c96f | 318 | set_new_dnode(&dn, inode, NULL, NULL, 0); |
39936837 | 319 | |
6ead1142 | 320 | err = get_dnode_of_data(&dn, start, ALLOC_NODE); |
39936837 | 321 | if (err) { |
e479556b | 322 | f2fs_unlock_op(sbi); |
1e1bb4ba | 323 | goto out; |
39936837 | 324 | } |
d624c96f | 325 | |
3cb5ad15 | 326 | f2fs_wait_on_page_writeback(dn.node_page, NODE); |
d624c96f JK |
327 | |
328 | get_node_info(sbi, dn.nid, &ni); | |
5d56b671 JK |
329 | f2fs_bug_on(ni.ino != ino_of_node(page)); |
330 | f2fs_bug_on(ofs_of_node(dn.node_page) != ofs_of_node(page)); | |
d624c96f JK |
331 | |
332 | for (; start < end; start++) { | |
333 | block_t src, dest; | |
334 | ||
335 | src = datablock_addr(dn.node_page, dn.ofs_in_node); | |
336 | dest = datablock_addr(page, dn.ofs_in_node); | |
337 | ||
338 | if (src != dest && dest != NEW_ADDR && dest != NULL_ADDR) { | |
339 | if (src == NULL_ADDR) { | |
5d56b671 | 340 | err = reserve_new_block(&dn); |
d624c96f | 341 | /* We should not get -ENOSPC */ |
5d56b671 | 342 | f2fs_bug_on(err); |
d624c96f JK |
343 | } |
344 | ||
345 | /* Check the previous node page having this index */ | |
39cf72cf JK |
346 | err = check_index_in_prev_nodes(sbi, dest, &dn); |
347 | if (err) | |
348 | goto err; | |
d624c96f JK |
349 | |
350 | set_summary(&sum, dn.nid, dn.ofs_in_node, ni.version); | |
351 | ||
352 | /* write dummy data page */ | |
353 | recover_data_page(sbi, NULL, &sum, src, dest); | |
354 | update_extent_cache(dest, &dn); | |
f356fe0c | 355 | recovered++; |
d624c96f JK |
356 | } |
357 | dn.ofs_in_node++; | |
358 | } | |
359 | ||
360 | /* write node page in place */ | |
361 | set_summary(&sum, dn.nid, 0, 0); | |
362 | if (IS_INODE(dn.node_page)) | |
363 | sync_inode_page(&dn); | |
364 | ||
365 | copy_node_footer(dn.node_page, page); | |
366 | fill_node_footer(dn.node_page, dn.nid, ni.ino, | |
367 | ofs_of_node(page), false); | |
368 | set_page_dirty(dn.node_page); | |
369 | ||
370 | recover_node_page(sbi, dn.node_page, &sum, &ni, blkaddr); | |
39cf72cf | 371 | err: |
d624c96f | 372 | f2fs_put_dnode(&dn); |
e479556b | 373 | f2fs_unlock_op(sbi); |
1e1bb4ba | 374 | out: |
6c311ec6 CF |
375 | f2fs_msg(sbi->sb, KERN_NOTICE, |
376 | "recover_data: ino = %lx, recovered = %d blocks, err = %d", | |
377 | inode->i_ino, recovered, err); | |
39cf72cf | 378 | return err; |
d624c96f JK |
379 | } |
380 | ||
6ead1142 | 381 | static int recover_data(struct f2fs_sb_info *sbi, |
d624c96f JK |
382 | struct list_head *head, int type) |
383 | { | |
d71b5564 | 384 | unsigned long long cp_ver = cur_cp_version(F2FS_CKPT(sbi)); |
d624c96f JK |
385 | struct curseg_info *curseg; |
386 | struct page *page; | |
6ead1142 | 387 | int err = 0; |
d624c96f JK |
388 | block_t blkaddr; |
389 | ||
390 | /* get node pages in the current segment */ | |
391 | curseg = CURSEG_I(sbi, type); | |
392 | blkaddr = NEXT_FREE_BLKADDR(sbi, curseg); | |
393 | ||
394 | /* read node page */ | |
a0acdfe0 | 395 | page = alloc_page(GFP_F2FS_ZERO); |
e27dae4d | 396 | if (!page) |
6ead1142 JK |
397 | return -ENOMEM; |
398 | ||
d624c96f JK |
399 | lock_page(page); |
400 | ||
401 | while (1) { | |
402 | struct fsync_inode_entry *entry; | |
403 | ||
93dfe2ac | 404 | err = f2fs_submit_page_bio(sbi, page, blkaddr, READ_SYNC); |
6ead1142 | 405 | if (err) |
b9987a27 | 406 | return err; |
d624c96f | 407 | |
393ff91f JK |
408 | lock_page(page); |
409 | ||
d624c96f | 410 | if (cp_ver != cpver_of_node(page)) |
45856aff | 411 | break; |
d624c96f JK |
412 | |
413 | entry = get_fsync_inode(head, ino_of_node(page)); | |
414 | if (!entry) | |
415 | goto next; | |
416 | ||
6ead1142 JK |
417 | err = do_recover_data(sbi, entry->inode, page, blkaddr); |
418 | if (err) | |
45856aff | 419 | break; |
d624c96f JK |
420 | |
421 | if (entry->blkaddr == blkaddr) { | |
422 | iput(entry->inode); | |
423 | list_del(&entry->list); | |
424 | kmem_cache_free(fsync_entry_slab, entry); | |
425 | } | |
426 | next: | |
427 | /* check next segment */ | |
428 | blkaddr = next_blkaddr_of_node(page); | |
d624c96f | 429 | } |
b9987a27 | 430 | |
d624c96f JK |
431 | unlock_page(page); |
432 | __free_pages(page, 0); | |
433 | ||
6ead1142 JK |
434 | if (!err) |
435 | allocate_new_segments(sbi); | |
436 | return err; | |
d624c96f JK |
437 | } |
438 | ||
6ead1142 | 439 | int recover_fsync_data(struct f2fs_sb_info *sbi) |
d624c96f | 440 | { |
cf2271e7 | 441 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_WARM_NODE); |
d624c96f | 442 | struct list_head inode_list; |
cf2271e7 | 443 | block_t blkaddr; |
6ead1142 | 444 | int err; |
aabe5136 | 445 | bool need_writecp = false; |
d624c96f JK |
446 | |
447 | fsync_entry_slab = f2fs_kmem_cache_create("f2fs_fsync_inode_entry", | |
e8512d2e | 448 | sizeof(struct fsync_inode_entry)); |
6bacf52f | 449 | if (!fsync_entry_slab) |
6ead1142 | 450 | return -ENOMEM; |
d624c96f JK |
451 | |
452 | INIT_LIST_HEAD(&inode_list); | |
453 | ||
454 | /* step #1: find fsynced inode numbers */ | |
aabe5136 | 455 | sbi->por_doing = true; |
cf2271e7 JK |
456 | |
457 | blkaddr = NEXT_FREE_BLKADDR(sbi, curseg); | |
458 | ||
6ead1142 JK |
459 | err = find_fsync_dnodes(sbi, &inode_list); |
460 | if (err) | |
d624c96f JK |
461 | goto out; |
462 | ||
463 | if (list_empty(&inode_list)) | |
464 | goto out; | |
465 | ||
aabe5136 | 466 | need_writecp = true; |
691c6fd2 | 467 | |
d624c96f | 468 | /* step #2: recover data */ |
6ead1142 | 469 | err = recover_data(sbi, &inode_list, CURSEG_WARM_NODE); |
5d56b671 | 470 | f2fs_bug_on(!list_empty(&inode_list)); |
d624c96f | 471 | out: |
5ebefc5b | 472 | destroy_fsync_dnodes(&inode_list); |
d624c96f | 473 | kmem_cache_destroy(fsync_entry_slab); |
cf2271e7 JK |
474 | |
475 | if (err) { | |
476 | truncate_inode_pages_final(NODE_MAPPING(sbi)); | |
477 | truncate_inode_pages_final(META_MAPPING(sbi)); | |
478 | } | |
479 | ||
aabe5136 | 480 | sbi->por_doing = false; |
cf2271e7 JK |
481 | if (err) { |
482 | discard_next_dnode(sbi, blkaddr); | |
483 | ||
484 | /* Flush all the NAT/SIT pages */ | |
485 | while (get_pages(sbi, F2FS_DIRTY_META)) | |
486 | sync_meta_pages(sbi, META, LONG_MAX); | |
487 | } else if (need_writecp) { | |
2c2c149f | 488 | write_checkpoint(sbi, false); |
cf2271e7 | 489 | } |
6ead1142 | 490 | return err; |
d624c96f | 491 | } |