]>
Commit | Line | Data |
---|---|---|
83d290c5 | 1 | // SPDX-License-Identifier: GPL-2.0+ |
9eefe2a2 SR |
2 | /* |
3 | * This file is part of UBIFS. | |
4 | * | |
5 | * Copyright (C) 2006-2008 Nokia Corporation. | |
6 | * | |
9eefe2a2 SR |
7 | * Author: Adrian Hunter |
8 | */ | |
9 | ||
ff94bc40 | 10 | #include <linux/err.h> |
9eefe2a2 SR |
11 | #include "ubifs.h" |
12 | ||
13 | /* | |
14 | * An orphan is an inode number whose inode node has been committed to the index | |
15 | * with a link count of zero. That happens when an open file is deleted | |
16 | * (unlinked) and then a commit is run. In the normal course of events the inode | |
17 | * would be deleted when the file is closed. However in the case of an unclean | |
18 | * unmount, orphans need to be accounted for. After an unclean unmount, the | |
19 | * orphans' inodes must be deleted which means either scanning the entire index | |
20 | * looking for them, or keeping a list on flash somewhere. This unit implements | |
21 | * the latter approach. | |
22 | * | |
23 | * The orphan area is a fixed number of LEBs situated between the LPT area and | |
24 | * the main area. The number of orphan area LEBs is specified when the file | |
25 | * system is created. The minimum number is 1. The size of the orphan area | |
26 | * should be so that it can hold the maximum number of orphans that are expected | |
27 | * to ever exist at one time. | |
28 | * | |
29 | * The number of orphans that can fit in a LEB is: | |
30 | * | |
31 | * (c->leb_size - UBIFS_ORPH_NODE_SZ) / sizeof(__le64) | |
32 | * | |
33 | * For example: a 15872 byte LEB can fit 1980 orphans so 1 LEB may be enough. | |
34 | * | |
35 | * Orphans are accumulated in a rb-tree. When an inode's link count drops to | |
36 | * zero, the inode number is added to the rb-tree. It is removed from the tree | |
37 | * when the inode is deleted. Any new orphans that are in the orphan tree when | |
38 | * the commit is run, are written to the orphan area in 1 or more orphan nodes. | |
39 | * If the orphan area is full, it is consolidated to make space. There is | |
40 | * always enough space because validation prevents the user from creating more | |
41 | * than the maximum number of orphans allowed. | |
42 | */ | |
43 | ||
ff94bc40 HS |
44 | static int dbg_check_orphans(struct ubifs_info *c); |
45 | ||
46 | /** | |
47 | * ubifs_add_orphan - add an orphan. | |
48 | * @c: UBIFS file-system description object | |
49 | * @inum: orphan inode number | |
50 | * | |
51 | * Add an orphan. This function is called when an inodes link count drops to | |
52 | * zero. | |
53 | */ | |
54 | int ubifs_add_orphan(struct ubifs_info *c, ino_t inum) | |
55 | { | |
56 | struct ubifs_orphan *orphan, *o; | |
57 | struct rb_node **p, *parent = NULL; | |
58 | ||
59 | orphan = kzalloc(sizeof(struct ubifs_orphan), GFP_NOFS); | |
60 | if (!orphan) | |
61 | return -ENOMEM; | |
62 | orphan->inum = inum; | |
63 | orphan->new = 1; | |
64 | ||
65 | spin_lock(&c->orphan_lock); | |
66 | if (c->tot_orphans >= c->max_orphans) { | |
67 | spin_unlock(&c->orphan_lock); | |
68 | kfree(orphan); | |
69 | return -ENFILE; | |
70 | } | |
71 | p = &c->orph_tree.rb_node; | |
72 | while (*p) { | |
73 | parent = *p; | |
74 | o = rb_entry(parent, struct ubifs_orphan, rb); | |
75 | if (inum < o->inum) | |
76 | p = &(*p)->rb_left; | |
77 | else if (inum > o->inum) | |
78 | p = &(*p)->rb_right; | |
79 | else { | |
0195a7bb | 80 | ubifs_err(c, "orphaned twice"); |
ff94bc40 HS |
81 | spin_unlock(&c->orphan_lock); |
82 | kfree(orphan); | |
83 | return 0; | |
84 | } | |
85 | } | |
86 | c->tot_orphans += 1; | |
87 | c->new_orphans += 1; | |
88 | rb_link_node(&orphan->rb, parent, p); | |
89 | rb_insert_color(&orphan->rb, &c->orph_tree); | |
90 | list_add_tail(&orphan->list, &c->orph_list); | |
91 | list_add_tail(&orphan->new_list, &c->orph_new); | |
92 | spin_unlock(&c->orphan_lock); | |
93 | dbg_gen("ino %lu", (unsigned long)inum); | |
94 | return 0; | |
95 | } | |
96 | ||
97 | /** | |
98 | * ubifs_delete_orphan - delete an orphan. | |
99 | * @c: UBIFS file-system description object | |
100 | * @inum: orphan inode number | |
101 | * | |
102 | * Delete an orphan. This function is called when an inode is deleted. | |
103 | */ | |
104 | void ubifs_delete_orphan(struct ubifs_info *c, ino_t inum) | |
105 | { | |
106 | struct ubifs_orphan *o; | |
107 | struct rb_node *p; | |
108 | ||
109 | spin_lock(&c->orphan_lock); | |
110 | p = c->orph_tree.rb_node; | |
111 | while (p) { | |
112 | o = rb_entry(p, struct ubifs_orphan, rb); | |
113 | if (inum < o->inum) | |
114 | p = p->rb_left; | |
115 | else if (inum > o->inum) | |
116 | p = p->rb_right; | |
117 | else { | |
118 | if (o->del) { | |
119 | spin_unlock(&c->orphan_lock); | |
120 | dbg_gen("deleted twice ino %lu", | |
121 | (unsigned long)inum); | |
122 | return; | |
123 | } | |
124 | if (o->cmt) { | |
125 | o->del = 1; | |
126 | o->dnext = c->orph_dnext; | |
127 | c->orph_dnext = o; | |
128 | spin_unlock(&c->orphan_lock); | |
129 | dbg_gen("delete later ino %lu", | |
130 | (unsigned long)inum); | |
131 | return; | |
132 | } | |
133 | rb_erase(p, &c->orph_tree); | |
134 | list_del(&o->list); | |
135 | c->tot_orphans -= 1; | |
136 | if (o->new) { | |
137 | list_del(&o->new_list); | |
138 | c->new_orphans -= 1; | |
139 | } | |
140 | spin_unlock(&c->orphan_lock); | |
141 | kfree(o); | |
142 | dbg_gen("inum %lu", (unsigned long)inum); | |
143 | return; | |
144 | } | |
145 | } | |
146 | spin_unlock(&c->orphan_lock); | |
0195a7bb | 147 | ubifs_err(c, "missing orphan ino %lu", (unsigned long)inum); |
ff94bc40 HS |
148 | dump_stack(); |
149 | } | |
150 | ||
151 | /** | |
152 | * ubifs_orphan_start_commit - start commit of orphans. | |
153 | * @c: UBIFS file-system description object | |
154 | * | |
155 | * Start commit of orphans. | |
156 | */ | |
157 | int ubifs_orphan_start_commit(struct ubifs_info *c) | |
158 | { | |
159 | struct ubifs_orphan *orphan, **last; | |
160 | ||
161 | spin_lock(&c->orphan_lock); | |
162 | last = &c->orph_cnext; | |
163 | list_for_each_entry(orphan, &c->orph_new, new_list) { | |
164 | ubifs_assert(orphan->new); | |
165 | ubifs_assert(!orphan->cmt); | |
166 | orphan->new = 0; | |
167 | orphan->cmt = 1; | |
168 | *last = orphan; | |
169 | last = &orphan->cnext; | |
170 | } | |
171 | *last = NULL; | |
172 | c->cmt_orphans = c->new_orphans; | |
173 | c->new_orphans = 0; | |
174 | dbg_cmt("%d orphans to commit", c->cmt_orphans); | |
175 | INIT_LIST_HEAD(&c->orph_new); | |
176 | if (c->tot_orphans == 0) | |
177 | c->no_orphs = 1; | |
178 | else | |
179 | c->no_orphs = 0; | |
180 | spin_unlock(&c->orphan_lock); | |
181 | return 0; | |
182 | } | |
183 | ||
184 | /** | |
185 | * avail_orphs - calculate available space. | |
186 | * @c: UBIFS file-system description object | |
187 | * | |
188 | * This function returns the number of orphans that can be written in the | |
189 | * available space. | |
190 | */ | |
191 | static int avail_orphs(struct ubifs_info *c) | |
192 | { | |
193 | int avail_lebs, avail, gap; | |
194 | ||
195 | avail_lebs = c->orph_lebs - (c->ohead_lnum - c->orph_first) - 1; | |
196 | avail = avail_lebs * | |
197 | ((c->leb_size - UBIFS_ORPH_NODE_SZ) / sizeof(__le64)); | |
198 | gap = c->leb_size - c->ohead_offs; | |
199 | if (gap >= UBIFS_ORPH_NODE_SZ + sizeof(__le64)) | |
200 | avail += (gap - UBIFS_ORPH_NODE_SZ) / sizeof(__le64); | |
201 | return avail; | |
202 | } | |
203 | ||
9eefe2a2 SR |
204 | /** |
205 | * tot_avail_orphs - calculate total space. | |
206 | * @c: UBIFS file-system description object | |
207 | * | |
208 | * This function returns the number of orphans that can be written in half | |
209 | * the total space. That leaves half the space for adding new orphans. | |
210 | */ | |
211 | static int tot_avail_orphs(struct ubifs_info *c) | |
212 | { | |
213 | int avail_lebs, avail; | |
214 | ||
215 | avail_lebs = c->orph_lebs; | |
216 | avail = avail_lebs * | |
217 | ((c->leb_size - UBIFS_ORPH_NODE_SZ) / sizeof(__le64)); | |
218 | return avail / 2; | |
219 | } | |
220 | ||
ff94bc40 HS |
221 | /** |
222 | * do_write_orph_node - write a node to the orphan head. | |
223 | * @c: UBIFS file-system description object | |
224 | * @len: length of node | |
225 | * @atomic: write atomically | |
226 | * | |
227 | * This function writes a node to the orphan head from the orphan buffer. If | |
228 | * %atomic is not zero, then the write is done atomically. On success, %0 is | |
229 | * returned, otherwise a negative error code is returned. | |
230 | */ | |
231 | static int do_write_orph_node(struct ubifs_info *c, int len, int atomic) | |
232 | { | |
233 | int err = 0; | |
234 | ||
235 | if (atomic) { | |
236 | ubifs_assert(c->ohead_offs == 0); | |
237 | ubifs_prepare_node(c, c->orph_buf, len, 1); | |
238 | len = ALIGN(len, c->min_io_size); | |
239 | err = ubifs_leb_change(c, c->ohead_lnum, c->orph_buf, len); | |
240 | } else { | |
241 | if (c->ohead_offs == 0) { | |
242 | /* Ensure LEB has been unmapped */ | |
243 | err = ubifs_leb_unmap(c, c->ohead_lnum); | |
244 | if (err) | |
245 | return err; | |
246 | } | |
247 | err = ubifs_write_node(c, c->orph_buf, len, c->ohead_lnum, | |
248 | c->ohead_offs); | |
249 | } | |
250 | return err; | |
251 | } | |
252 | ||
253 | /** | |
254 | * write_orph_node - write an orphan node. | |
255 | * @c: UBIFS file-system description object | |
256 | * @atomic: write atomically | |
257 | * | |
258 | * This function builds an orphan node from the cnext list and writes it to the | |
259 | * orphan head. On success, %0 is returned, otherwise a negative error code | |
260 | * is returned. | |
261 | */ | |
262 | static int write_orph_node(struct ubifs_info *c, int atomic) | |
263 | { | |
264 | struct ubifs_orphan *orphan, *cnext; | |
265 | struct ubifs_orph_node *orph; | |
266 | int gap, err, len, cnt, i; | |
267 | ||
268 | ubifs_assert(c->cmt_orphans > 0); | |
269 | gap = c->leb_size - c->ohead_offs; | |
270 | if (gap < UBIFS_ORPH_NODE_SZ + sizeof(__le64)) { | |
271 | c->ohead_lnum += 1; | |
272 | c->ohead_offs = 0; | |
273 | gap = c->leb_size; | |
274 | if (c->ohead_lnum > c->orph_last) { | |
275 | /* | |
276 | * We limit the number of orphans so that this should | |
277 | * never happen. | |
278 | */ | |
0195a7bb | 279 | ubifs_err(c, "out of space in orphan area"); |
ff94bc40 HS |
280 | return -EINVAL; |
281 | } | |
282 | } | |
283 | cnt = (gap - UBIFS_ORPH_NODE_SZ) / sizeof(__le64); | |
284 | if (cnt > c->cmt_orphans) | |
285 | cnt = c->cmt_orphans; | |
286 | len = UBIFS_ORPH_NODE_SZ + cnt * sizeof(__le64); | |
287 | ubifs_assert(c->orph_buf); | |
288 | orph = c->orph_buf; | |
289 | orph->ch.node_type = UBIFS_ORPH_NODE; | |
290 | spin_lock(&c->orphan_lock); | |
291 | cnext = c->orph_cnext; | |
292 | for (i = 0; i < cnt; i++) { | |
293 | orphan = cnext; | |
294 | ubifs_assert(orphan->cmt); | |
295 | orph->inos[i] = cpu_to_le64(orphan->inum); | |
296 | orphan->cmt = 0; | |
297 | cnext = orphan->cnext; | |
298 | orphan->cnext = NULL; | |
299 | } | |
300 | c->orph_cnext = cnext; | |
301 | c->cmt_orphans -= cnt; | |
302 | spin_unlock(&c->orphan_lock); | |
303 | if (c->cmt_orphans) | |
304 | orph->cmt_no = cpu_to_le64(c->cmt_no); | |
305 | else | |
306 | /* Mark the last node of the commit */ | |
307 | orph->cmt_no = cpu_to_le64((c->cmt_no) | (1ULL << 63)); | |
308 | ubifs_assert(c->ohead_offs + len <= c->leb_size); | |
309 | ubifs_assert(c->ohead_lnum >= c->orph_first); | |
310 | ubifs_assert(c->ohead_lnum <= c->orph_last); | |
311 | err = do_write_orph_node(c, len, atomic); | |
312 | c->ohead_offs += ALIGN(len, c->min_io_size); | |
313 | c->ohead_offs = ALIGN(c->ohead_offs, 8); | |
314 | return err; | |
315 | } | |
316 | ||
317 | /** | |
318 | * write_orph_nodes - write orphan nodes until there are no more to commit. | |
319 | * @c: UBIFS file-system description object | |
320 | * @atomic: write atomically | |
321 | * | |
322 | * This function writes orphan nodes for all the orphans to commit. On success, | |
323 | * %0 is returned, otherwise a negative error code is returned. | |
324 | */ | |
325 | static int write_orph_nodes(struct ubifs_info *c, int atomic) | |
326 | { | |
327 | int err; | |
328 | ||
329 | while (c->cmt_orphans > 0) { | |
330 | err = write_orph_node(c, atomic); | |
331 | if (err) | |
332 | return err; | |
333 | } | |
334 | if (atomic) { | |
335 | int lnum; | |
336 | ||
337 | /* Unmap any unused LEBs after consolidation */ | |
ff94bc40 HS |
338 | for (lnum = c->ohead_lnum + 1; lnum <= c->orph_last; lnum++) { |
339 | err = ubifs_leb_unmap(c, lnum); | |
340 | if (err) | |
341 | return err; | |
342 | } | |
343 | } | |
344 | return 0; | |
345 | } | |
346 | ||
347 | /** | |
348 | * consolidate - consolidate the orphan area. | |
349 | * @c: UBIFS file-system description object | |
350 | * | |
351 | * This function enables consolidation by putting all the orphans into the list | |
352 | * to commit. The list is in the order that the orphans were added, and the | |
353 | * LEBs are written atomically in order, so at no time can orphans be lost by | |
354 | * an unclean unmount. | |
355 | * | |
356 | * This function returns %0 on success and a negative error code on failure. | |
357 | */ | |
358 | static int consolidate(struct ubifs_info *c) | |
359 | { | |
360 | int tot_avail = tot_avail_orphs(c), err = 0; | |
361 | ||
362 | spin_lock(&c->orphan_lock); | |
363 | dbg_cmt("there is space for %d orphans and there are %d", | |
364 | tot_avail, c->tot_orphans); | |
365 | if (c->tot_orphans - c->new_orphans <= tot_avail) { | |
366 | struct ubifs_orphan *orphan, **last; | |
367 | int cnt = 0; | |
368 | ||
369 | /* Change the cnext list to include all non-new orphans */ | |
370 | last = &c->orph_cnext; | |
371 | list_for_each_entry(orphan, &c->orph_list, list) { | |
372 | if (orphan->new) | |
373 | continue; | |
374 | orphan->cmt = 1; | |
375 | *last = orphan; | |
376 | last = &orphan->cnext; | |
377 | cnt += 1; | |
378 | } | |
379 | *last = NULL; | |
380 | ubifs_assert(cnt == c->tot_orphans - c->new_orphans); | |
381 | c->cmt_orphans = cnt; | |
382 | c->ohead_lnum = c->orph_first; | |
383 | c->ohead_offs = 0; | |
384 | } else { | |
385 | /* | |
386 | * We limit the number of orphans so that this should | |
387 | * never happen. | |
388 | */ | |
0195a7bb | 389 | ubifs_err(c, "out of space in orphan area"); |
ff94bc40 HS |
390 | err = -EINVAL; |
391 | } | |
392 | spin_unlock(&c->orphan_lock); | |
393 | return err; | |
394 | } | |
395 | ||
396 | /** | |
397 | * commit_orphans - commit orphans. | |
398 | * @c: UBIFS file-system description object | |
399 | * | |
400 | * This function commits orphans to flash. On success, %0 is returned, | |
401 | * otherwise a negative error code is returned. | |
402 | */ | |
403 | static int commit_orphans(struct ubifs_info *c) | |
404 | { | |
405 | int avail, atomic = 0, err; | |
406 | ||
407 | ubifs_assert(c->cmt_orphans > 0); | |
408 | avail = avail_orphs(c); | |
409 | if (avail < c->cmt_orphans) { | |
410 | /* Not enough space to write new orphans, so consolidate */ | |
411 | err = consolidate(c); | |
412 | if (err) | |
413 | return err; | |
414 | atomic = 1; | |
415 | } | |
416 | err = write_orph_nodes(c, atomic); | |
417 | return err; | |
418 | } | |
419 | ||
420 | /** | |
421 | * erase_deleted - erase the orphans marked for deletion. | |
422 | * @c: UBIFS file-system description object | |
423 | * | |
424 | * During commit, the orphans being committed cannot be deleted, so they are | |
425 | * marked for deletion and deleted by this function. Also, the recovery | |
426 | * adds killed orphans to the deletion list, and therefore they are deleted | |
427 | * here too. | |
428 | */ | |
429 | static void erase_deleted(struct ubifs_info *c) | |
430 | { | |
431 | struct ubifs_orphan *orphan, *dnext; | |
432 | ||
433 | spin_lock(&c->orphan_lock); | |
434 | dnext = c->orph_dnext; | |
435 | while (dnext) { | |
436 | orphan = dnext; | |
437 | dnext = orphan->dnext; | |
438 | ubifs_assert(!orphan->new); | |
439 | ubifs_assert(orphan->del); | |
440 | rb_erase(&orphan->rb, &c->orph_tree); | |
441 | list_del(&orphan->list); | |
442 | c->tot_orphans -= 1; | |
443 | dbg_gen("deleting orphan ino %lu", (unsigned long)orphan->inum); | |
444 | kfree(orphan); | |
445 | } | |
446 | c->orph_dnext = NULL; | |
447 | spin_unlock(&c->orphan_lock); | |
448 | } | |
449 | ||
450 | /** | |
451 | * ubifs_orphan_end_commit - end commit of orphans. | |
452 | * @c: UBIFS file-system description object | |
453 | * | |
454 | * End commit of orphans. | |
455 | */ | |
456 | int ubifs_orphan_end_commit(struct ubifs_info *c) | |
457 | { | |
458 | int err; | |
459 | ||
460 | if (c->cmt_orphans != 0) { | |
461 | err = commit_orphans(c); | |
462 | if (err) | |
463 | return err; | |
464 | } | |
465 | erase_deleted(c); | |
466 | err = dbg_check_orphans(c); | |
467 | return err; | |
468 | } | |
469 | ||
9eefe2a2 SR |
470 | /** |
471 | * ubifs_clear_orphans - erase all LEBs used for orphans. | |
472 | * @c: UBIFS file-system description object | |
473 | * | |
474 | * If recovery is not required, then the orphans from the previous session | |
475 | * are not needed. This function locates the LEBs used to record | |
476 | * orphans, and un-maps them. | |
477 | */ | |
478 | int ubifs_clear_orphans(struct ubifs_info *c) | |
479 | { | |
480 | int lnum, err; | |
481 | ||
482 | for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) { | |
483 | err = ubifs_leb_unmap(c, lnum); | |
484 | if (err) | |
485 | return err; | |
486 | } | |
487 | c->ohead_lnum = c->orph_first; | |
488 | c->ohead_offs = 0; | |
489 | return 0; | |
490 | } | |
491 | ||
492 | /** | |
493 | * insert_dead_orphan - insert an orphan. | |
494 | * @c: UBIFS file-system description object | |
495 | * @inum: orphan inode number | |
496 | * | |
497 | * This function is a helper to the 'do_kill_orphans()' function. The orphan | |
498 | * must be kept until the next commit, so it is added to the rb-tree and the | |
499 | * deletion list. | |
500 | */ | |
501 | static int insert_dead_orphan(struct ubifs_info *c, ino_t inum) | |
502 | { | |
503 | struct ubifs_orphan *orphan, *o; | |
504 | struct rb_node **p, *parent = NULL; | |
505 | ||
506 | orphan = kzalloc(sizeof(struct ubifs_orphan), GFP_KERNEL); | |
507 | if (!orphan) | |
508 | return -ENOMEM; | |
509 | orphan->inum = inum; | |
510 | ||
511 | p = &c->orph_tree.rb_node; | |
512 | while (*p) { | |
513 | parent = *p; | |
514 | o = rb_entry(parent, struct ubifs_orphan, rb); | |
515 | if (inum < o->inum) | |
516 | p = &(*p)->rb_left; | |
517 | else if (inum > o->inum) | |
518 | p = &(*p)->rb_right; | |
519 | else { | |
520 | /* Already added - no problem */ | |
521 | kfree(orphan); | |
522 | return 0; | |
523 | } | |
524 | } | |
525 | c->tot_orphans += 1; | |
526 | rb_link_node(&orphan->rb, parent, p); | |
527 | rb_insert_color(&orphan->rb, &c->orph_tree); | |
528 | list_add_tail(&orphan->list, &c->orph_list); | |
ff94bc40 | 529 | orphan->del = 1; |
9eefe2a2 SR |
530 | orphan->dnext = c->orph_dnext; |
531 | c->orph_dnext = orphan; | |
532 | dbg_mnt("ino %lu, new %d, tot %d", (unsigned long)inum, | |
533 | c->new_orphans, c->tot_orphans); | |
534 | return 0; | |
535 | } | |
536 | ||
537 | /** | |
538 | * do_kill_orphans - remove orphan inodes from the index. | |
539 | * @c: UBIFS file-system description object | |
540 | * @sleb: scanned LEB | |
541 | * @last_cmt_no: cmt_no of last orphan node read is passed and returned here | |
542 | * @outofdate: whether the LEB is out of date is returned here | |
543 | * @last_flagged: whether the end orphan node is encountered | |
544 | * | |
545 | * This function is a helper to the 'kill_orphans()' function. It goes through | |
546 | * every orphan node in a LEB and for every inode number recorded, removes | |
547 | * all keys for that inode from the TNC. | |
548 | */ | |
549 | static int do_kill_orphans(struct ubifs_info *c, struct ubifs_scan_leb *sleb, | |
550 | unsigned long long *last_cmt_no, int *outofdate, | |
551 | int *last_flagged) | |
552 | { | |
553 | struct ubifs_scan_node *snod; | |
554 | struct ubifs_orph_node *orph; | |
555 | unsigned long long cmt_no; | |
556 | ino_t inum; | |
557 | int i, n, err, first = 1; | |
558 | ||
559 | list_for_each_entry(snod, &sleb->nodes, list) { | |
560 | if (snod->type != UBIFS_ORPH_NODE) { | |
0195a7bb | 561 | ubifs_err(c, "invalid node type %d in orphan area at %d:%d", |
ff94bc40 HS |
562 | snod->type, sleb->lnum, snod->offs); |
563 | ubifs_dump_node(c, snod->node); | |
9eefe2a2 SR |
564 | return -EINVAL; |
565 | } | |
566 | ||
567 | orph = snod->node; | |
568 | ||
569 | /* Check commit number */ | |
570 | cmt_no = le64_to_cpu(orph->cmt_no) & LLONG_MAX; | |
571 | /* | |
572 | * The commit number on the master node may be less, because | |
573 | * of a failed commit. If there are several failed commits in a | |
574 | * row, the commit number written on orphan nodes will continue | |
575 | * to increase (because the commit number is adjusted here) even | |
576 | * though the commit number on the master node stays the same | |
577 | * because the master node has not been re-written. | |
578 | */ | |
579 | if (cmt_no > c->cmt_no) | |
580 | c->cmt_no = cmt_no; | |
581 | if (cmt_no < *last_cmt_no && *last_flagged) { | |
582 | /* | |
583 | * The last orphan node had a higher commit number and | |
584 | * was flagged as the last written for that commit | |
585 | * number. That makes this orphan node, out of date. | |
586 | */ | |
587 | if (!first) { | |
0195a7bb | 588 | ubifs_err(c, "out of order commit number %llu in orphan node at %d:%d", |
9eefe2a2 | 589 | cmt_no, sleb->lnum, snod->offs); |
ff94bc40 | 590 | ubifs_dump_node(c, snod->node); |
9eefe2a2 SR |
591 | return -EINVAL; |
592 | } | |
593 | dbg_rcvry("out of date LEB %d", sleb->lnum); | |
594 | *outofdate = 1; | |
595 | return 0; | |
596 | } | |
597 | ||
598 | if (first) | |
599 | first = 0; | |
600 | ||
601 | n = (le32_to_cpu(orph->ch.len) - UBIFS_ORPH_NODE_SZ) >> 3; | |
602 | for (i = 0; i < n; i++) { | |
603 | inum = le64_to_cpu(orph->inos[i]); | |
604 | dbg_rcvry("deleting orphaned inode %lu", | |
605 | (unsigned long)inum); | |
606 | err = ubifs_tnc_remove_ino(c, inum); | |
607 | if (err) | |
608 | return err; | |
609 | err = insert_dead_orphan(c, inum); | |
610 | if (err) | |
611 | return err; | |
612 | } | |
613 | ||
614 | *last_cmt_no = cmt_no; | |
615 | if (le64_to_cpu(orph->cmt_no) & (1ULL << 63)) { | |
616 | dbg_rcvry("last orph node for commit %llu at %d:%d", | |
617 | cmt_no, sleb->lnum, snod->offs); | |
618 | *last_flagged = 1; | |
619 | } else | |
620 | *last_flagged = 0; | |
621 | } | |
622 | ||
623 | return 0; | |
624 | } | |
625 | ||
626 | /** | |
627 | * kill_orphans - remove all orphan inodes from the index. | |
628 | * @c: UBIFS file-system description object | |
629 | * | |
630 | * If recovery is required, then orphan inodes recorded during the previous | |
631 | * session (which ended with an unclean unmount) must be deleted from the index. | |
632 | * This is done by updating the TNC, but since the index is not updated until | |
633 | * the next commit, the LEBs where the orphan information is recorded are not | |
634 | * erased until the next commit. | |
635 | */ | |
636 | static int kill_orphans(struct ubifs_info *c) | |
637 | { | |
638 | unsigned long long last_cmt_no = 0; | |
639 | int lnum, err = 0, outofdate = 0, last_flagged = 0; | |
640 | ||
641 | c->ohead_lnum = c->orph_first; | |
642 | c->ohead_offs = 0; | |
643 | /* Check no-orphans flag and skip this if no orphans */ | |
644 | if (c->no_orphs) { | |
645 | dbg_rcvry("no orphans"); | |
646 | return 0; | |
647 | } | |
648 | /* | |
649 | * Orph nodes always start at c->orph_first and are written to each | |
650 | * successive LEB in turn. Generally unused LEBs will have been unmapped | |
651 | * but may contain out of date orphan nodes if the unmap didn't go | |
652 | * through. In addition, the last orphan node written for each commit is | |
653 | * marked (top bit of orph->cmt_no is set to 1). It is possible that | |
654 | * there are orphan nodes from the next commit (i.e. the commit did not | |
655 | * complete successfully). In that case, no orphans will have been lost | |
656 | * due to the way that orphans are written, and any orphans added will | |
657 | * be valid orphans anyway and so can be deleted. | |
658 | */ | |
659 | for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) { | |
660 | struct ubifs_scan_leb *sleb; | |
661 | ||
662 | dbg_rcvry("LEB %d", lnum); | |
ff94bc40 | 663 | sleb = ubifs_scan(c, lnum, 0, c->sbuf, 1); |
9eefe2a2 | 664 | if (IS_ERR(sleb)) { |
ff94bc40 HS |
665 | if (PTR_ERR(sleb) == -EUCLEAN) |
666 | sleb = ubifs_recover_leb(c, lnum, 0, | |
667 | c->sbuf, -1); | |
9eefe2a2 SR |
668 | if (IS_ERR(sleb)) { |
669 | err = PTR_ERR(sleb); | |
670 | break; | |
671 | } | |
672 | } | |
673 | err = do_kill_orphans(c, sleb, &last_cmt_no, &outofdate, | |
674 | &last_flagged); | |
675 | if (err || outofdate) { | |
676 | ubifs_scan_destroy(sleb); | |
677 | break; | |
678 | } | |
679 | if (sleb->endpt) { | |
680 | c->ohead_lnum = lnum; | |
681 | c->ohead_offs = sleb->endpt; | |
682 | } | |
683 | ubifs_scan_destroy(sleb); | |
684 | } | |
685 | return err; | |
686 | } | |
687 | ||
688 | /** | |
689 | * ubifs_mount_orphans - delete orphan inodes and erase LEBs that recorded them. | |
690 | * @c: UBIFS file-system description object | |
691 | * @unclean: indicates recovery from unclean unmount | |
692 | * @read_only: indicates read only mount | |
693 | * | |
694 | * This function is called when mounting to erase orphans from the previous | |
695 | * session. If UBIFS was not unmounted cleanly, then the inodes recorded as | |
696 | * orphans are deleted. | |
697 | */ | |
698 | int ubifs_mount_orphans(struct ubifs_info *c, int unclean, int read_only) | |
699 | { | |
700 | int err = 0; | |
701 | ||
702 | c->max_orphans = tot_avail_orphs(c); | |
703 | ||
704 | if (!read_only) { | |
705 | c->orph_buf = vmalloc(c->leb_size); | |
706 | if (!c->orph_buf) | |
707 | return -ENOMEM; | |
708 | } | |
709 | ||
710 | if (unclean) | |
711 | err = kill_orphans(c); | |
712 | else if (!read_only) | |
713 | err = ubifs_clear_orphans(c); | |
714 | ||
715 | return err; | |
716 | } | |
ff94bc40 HS |
717 | |
718 | /* | |
719 | * Everything below is related to debugging. | |
720 | */ | |
721 | ||
722 | struct check_orphan { | |
723 | struct rb_node rb; | |
724 | ino_t inum; | |
725 | }; | |
726 | ||
727 | struct check_info { | |
728 | unsigned long last_ino; | |
729 | unsigned long tot_inos; | |
730 | unsigned long missing; | |
731 | unsigned long long leaf_cnt; | |
732 | struct ubifs_ino_node *node; | |
733 | struct rb_root root; | |
734 | }; | |
735 | ||
736 | static int dbg_find_orphan(struct ubifs_info *c, ino_t inum) | |
737 | { | |
738 | struct ubifs_orphan *o; | |
739 | struct rb_node *p; | |
740 | ||
741 | spin_lock(&c->orphan_lock); | |
742 | p = c->orph_tree.rb_node; | |
743 | while (p) { | |
744 | o = rb_entry(p, struct ubifs_orphan, rb); | |
745 | if (inum < o->inum) | |
746 | p = p->rb_left; | |
747 | else if (inum > o->inum) | |
748 | p = p->rb_right; | |
749 | else { | |
750 | spin_unlock(&c->orphan_lock); | |
751 | return 1; | |
752 | } | |
753 | } | |
754 | spin_unlock(&c->orphan_lock); | |
755 | return 0; | |
756 | } | |
757 | ||
758 | static int dbg_ins_check_orphan(struct rb_root *root, ino_t inum) | |
759 | { | |
760 | struct check_orphan *orphan, *o; | |
761 | struct rb_node **p, *parent = NULL; | |
762 | ||
763 | orphan = kzalloc(sizeof(struct check_orphan), GFP_NOFS); | |
764 | if (!orphan) | |
765 | return -ENOMEM; | |
766 | orphan->inum = inum; | |
767 | ||
768 | p = &root->rb_node; | |
769 | while (*p) { | |
770 | parent = *p; | |
771 | o = rb_entry(parent, struct check_orphan, rb); | |
772 | if (inum < o->inum) | |
773 | p = &(*p)->rb_left; | |
774 | else if (inum > o->inum) | |
775 | p = &(*p)->rb_right; | |
776 | else { | |
777 | kfree(orphan); | |
778 | return 0; | |
779 | } | |
780 | } | |
781 | rb_link_node(&orphan->rb, parent, p); | |
782 | rb_insert_color(&orphan->rb, root); | |
783 | return 0; | |
784 | } | |
785 | ||
786 | static int dbg_find_check_orphan(struct rb_root *root, ino_t inum) | |
787 | { | |
788 | struct check_orphan *o; | |
789 | struct rb_node *p; | |
790 | ||
791 | p = root->rb_node; | |
792 | while (p) { | |
793 | o = rb_entry(p, struct check_orphan, rb); | |
794 | if (inum < o->inum) | |
795 | p = p->rb_left; | |
796 | else if (inum > o->inum) | |
797 | p = p->rb_right; | |
798 | else | |
799 | return 1; | |
800 | } | |
801 | return 0; | |
802 | } | |
803 | ||
804 | static void dbg_free_check_tree(struct rb_root *root) | |
805 | { | |
806 | struct check_orphan *o, *n; | |
807 | ||
808 | rbtree_postorder_for_each_entry_safe(o, n, root, rb) | |
809 | kfree(o); | |
810 | } | |
811 | ||
812 | static int dbg_orphan_check(struct ubifs_info *c, struct ubifs_zbranch *zbr, | |
813 | void *priv) | |
814 | { | |
815 | struct check_info *ci = priv; | |
816 | ino_t inum; | |
817 | int err; | |
818 | ||
819 | inum = key_inum(c, &zbr->key); | |
820 | if (inum != ci->last_ino) { | |
821 | /* Lowest node type is the inode node, so it comes first */ | |
822 | if (key_type(c, &zbr->key) != UBIFS_INO_KEY) | |
0195a7bb | 823 | ubifs_err(c, "found orphan node ino %lu, type %d", |
ff94bc40 HS |
824 | (unsigned long)inum, key_type(c, &zbr->key)); |
825 | ci->last_ino = inum; | |
826 | ci->tot_inos += 1; | |
827 | err = ubifs_tnc_read_node(c, zbr, ci->node); | |
828 | if (err) { | |
0195a7bb | 829 | ubifs_err(c, "node read failed, error %d", err); |
ff94bc40 HS |
830 | return err; |
831 | } | |
832 | if (ci->node->nlink == 0) | |
833 | /* Must be recorded as an orphan */ | |
834 | if (!dbg_find_check_orphan(&ci->root, inum) && | |
835 | !dbg_find_orphan(c, inum)) { | |
0195a7bb | 836 | ubifs_err(c, "missing orphan, ino %lu", |
ff94bc40 HS |
837 | (unsigned long)inum); |
838 | ci->missing += 1; | |
839 | } | |
840 | } | |
841 | ci->leaf_cnt += 1; | |
842 | return 0; | |
843 | } | |
844 | ||
845 | static int dbg_read_orphans(struct check_info *ci, struct ubifs_scan_leb *sleb) | |
846 | { | |
847 | struct ubifs_scan_node *snod; | |
848 | struct ubifs_orph_node *orph; | |
849 | ino_t inum; | |
850 | int i, n, err; | |
851 | ||
852 | list_for_each_entry(snod, &sleb->nodes, list) { | |
853 | cond_resched(); | |
854 | if (snod->type != UBIFS_ORPH_NODE) | |
855 | continue; | |
856 | orph = snod->node; | |
857 | n = (le32_to_cpu(orph->ch.len) - UBIFS_ORPH_NODE_SZ) >> 3; | |
858 | for (i = 0; i < n; i++) { | |
859 | inum = le64_to_cpu(orph->inos[i]); | |
860 | err = dbg_ins_check_orphan(&ci->root, inum); | |
861 | if (err) | |
862 | return err; | |
863 | } | |
864 | } | |
865 | return 0; | |
866 | } | |
867 | ||
868 | static int dbg_scan_orphans(struct ubifs_info *c, struct check_info *ci) | |
869 | { | |
870 | int lnum, err = 0; | |
871 | void *buf; | |
872 | ||
873 | /* Check no-orphans flag and skip this if no orphans */ | |
874 | if (c->no_orphs) | |
875 | return 0; | |
876 | ||
877 | buf = __vmalloc(c->leb_size, GFP_NOFS, PAGE_KERNEL); | |
878 | if (!buf) { | |
0195a7bb | 879 | ubifs_err(c, "cannot allocate memory to check orphans"); |
ff94bc40 HS |
880 | return 0; |
881 | } | |
882 | ||
883 | for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) { | |
884 | struct ubifs_scan_leb *sleb; | |
885 | ||
886 | sleb = ubifs_scan(c, lnum, 0, buf, 0); | |
887 | if (IS_ERR(sleb)) { | |
888 | err = PTR_ERR(sleb); | |
889 | break; | |
890 | } | |
891 | ||
892 | err = dbg_read_orphans(ci, sleb); | |
893 | ubifs_scan_destroy(sleb); | |
894 | if (err) | |
895 | break; | |
896 | } | |
897 | ||
898 | vfree(buf); | |
899 | return err; | |
900 | } | |
901 | ||
902 | static int dbg_check_orphans(struct ubifs_info *c) | |
903 | { | |
904 | struct check_info ci; | |
905 | int err; | |
906 | ||
907 | if (!dbg_is_chk_orph(c)) | |
908 | return 0; | |
909 | ||
910 | ci.last_ino = 0; | |
911 | ci.tot_inos = 0; | |
912 | ci.missing = 0; | |
913 | ci.leaf_cnt = 0; | |
914 | ci.root = RB_ROOT; | |
915 | ci.node = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS); | |
916 | if (!ci.node) { | |
0195a7bb | 917 | ubifs_err(c, "out of memory"); |
ff94bc40 HS |
918 | return -ENOMEM; |
919 | } | |
920 | ||
921 | err = dbg_scan_orphans(c, &ci); | |
922 | if (err) | |
923 | goto out; | |
924 | ||
925 | err = dbg_walk_index(c, &dbg_orphan_check, NULL, &ci); | |
926 | if (err) { | |
0195a7bb | 927 | ubifs_err(c, "cannot scan TNC, error %d", err); |
ff94bc40 HS |
928 | goto out; |
929 | } | |
930 | ||
931 | if (ci.missing) { | |
0195a7bb | 932 | ubifs_err(c, "%lu missing orphan(s)", ci.missing); |
ff94bc40 HS |
933 | err = -EINVAL; |
934 | goto out; | |
935 | } | |
936 | ||
937 | dbg_cmt("last inode number is %lu", ci.last_ino); | |
938 | dbg_cmt("total number of inodes is %lu", ci.tot_inos); | |
939 | dbg_cmt("total number of leaf nodes is %llu", ci.leaf_cnt); | |
940 | ||
941 | out: | |
942 | dbg_free_check_tree(&ci.root); | |
943 | kfree(ci.node); | |
944 | return err; | |
945 | } |