]>
Commit | Line | Data |
---|---|---|
801c135c AB |
1 | /* |
2 | * Copyright (c) International Business Machines Corp., 2006 | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or modify | |
5 | * it under the terms of the GNU General Public License as published by | |
6 | * the Free Software Foundation; either version 2 of the License, or | |
7 | * (at your option) any later version. | |
8 | * | |
9 | * This program is distributed in the hope that it will be useful, | |
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See | |
12 | * the GNU General Public License for more details. | |
13 | * | |
14 | * You should have received a copy of the GNU General Public License | |
15 | * along with this program; if not, write to the Free Software | |
16 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
17 | * | |
18 | * Authors: Artem Bityutskiy (Битюцкий Артём), Thomas Gleixner | |
19 | */ | |
20 | ||
21 | /* | |
85c6e6e2 | 22 | * UBI wear-leveling sub-system. |
801c135c | 23 | * |
85c6e6e2 AB |
24 | * This sub-system is responsible for wear-leveling. It works in terms of |
25 | * physical* eraseblocks and erase counters and knows nothing about logical | |
26 | * eraseblocks, volumes, etc. From this sub-system's perspective all physical | |
27 | * eraseblocks are of two types - used and free. Used physical eraseblocks are | |
28 | * those that were "get" by the 'ubi_wl_get_peb()' function, and free physical | |
29 | * eraseblocks are those that were put by the 'ubi_wl_put_peb()' function. | |
801c135c AB |
30 | * |
31 | * Physical eraseblocks returned by 'ubi_wl_get_peb()' have only erase counter | |
85c6e6e2 | 32 | * header. The rest of the physical eraseblock contains only %0xFF bytes. |
801c135c | 33 | * |
85c6e6e2 | 34 | * When physical eraseblocks are returned to the WL sub-system by means of the |
801c135c AB |
35 | * 'ubi_wl_put_peb()' function, they are scheduled for erasure. The erasure is |
36 | * done asynchronously in context of the per-UBI device background thread, | |
85c6e6e2 | 37 | * which is also managed by the WL sub-system. |
801c135c AB |
38 | * |
39 | * The wear-leveling is ensured by means of moving the contents of used | |
40 | * physical eraseblocks with low erase counter to free physical eraseblocks | |
41 | * with high erase counter. | |
42 | * | |
43 | * The 'ubi_wl_get_peb()' function accepts data type hints which help to pick | |
44 | * an "optimal" physical eraseblock. For example, when it is known that the | |
45 | * physical eraseblock will be "put" soon because it contains short-term data, | |
85c6e6e2 AB |
46 | * the WL sub-system may pick a free physical eraseblock with low erase |
47 | * counter, and so forth. | |
801c135c | 48 | * |
85c6e6e2 AB |
49 | * If the WL sub-system fails to erase a physical eraseblock, it marks it as |
50 | * bad. | |
801c135c | 51 | * |
85c6e6e2 AB |
52 | * This sub-system is also responsible for scrubbing. If a bit-flip is detected |
53 | * in a physical eraseblock, it has to be moved. Technically this is the same | |
54 | * as moving it for wear-leveling reasons. | |
801c135c | 55 | * |
85c6e6e2 AB |
56 | * As it was said, for the UBI sub-system all physical eraseblocks are either |
57 | * "free" or "used". Free eraseblock are kept in the @wl->free RB-tree, while | |
58 | * used eraseblocks are kept in a set of different RB-trees: @wl->used, | |
801c135c AB |
59 | * @wl->prot.pnum, @wl->prot.aec, and @wl->scrub. |
60 | * | |
61 | * Note, in this implementation, we keep a small in-RAM object for each physical | |
62 | * eraseblock. This is surely not a scalable solution. But it appears to be good | |
63 | * enough for moderately large flashes and it is simple. In future, one may | |
85c6e6e2 | 64 | * re-work this sub-system and make it more scalable. |
801c135c | 65 | * |
85c6e6e2 AB |
66 | * At the moment this sub-system does not utilize the sequence number, which |
67 | * was introduced relatively recently. But it would be wise to do this because | |
68 | * the sequence number of a logical eraseblock characterizes how old is it. For | |
801c135c AB |
69 | * example, when we move a PEB with low erase counter, and we need to pick the |
70 | * target PEB, we pick a PEB with the highest EC if our PEB is "old" and we | |
71 | * pick target PEB with an average EC if our PEB is not very "old". This is a | |
85c6e6e2 | 72 | * room for future re-works of the WL sub-system. |
801c135c | 73 | * |
85c6e6e2 AB |
74 | * Note: the stuff with protection trees looks too complex and is difficult to |
75 | * understand. Should be fixed. | |
801c135c AB |
76 | */ |
77 | ||
78 | #include <linux/slab.h> | |
79 | #include <linux/crc32.h> | |
80 | #include <linux/freezer.h> | |
81 | #include <linux/kthread.h> | |
82 | #include "ubi.h" | |
83 | ||
84 | /* Number of physical eraseblocks reserved for wear-leveling purposes */ | |
85 | #define WL_RESERVED_PEBS 1 | |
86 | ||
87 | /* | |
88 | * How many erase cycles are short term, unknown, and long term physical | |
89 | * eraseblocks protected. | |
90 | */ | |
91 | #define ST_PROTECTION 16 | |
92 | #define U_PROTECTION 10 | |
93 | #define LT_PROTECTION 4 | |
94 | ||
95 | /* | |
96 | * Maximum difference between two erase counters. If this threshold is | |
85c6e6e2 AB |
97 | * exceeded, the WL sub-system starts moving data from used physical |
98 | * eraseblocks with low erase counter to free physical eraseblocks with high | |
99 | * erase counter. | |
801c135c AB |
100 | */ |
101 | #define UBI_WL_THRESHOLD CONFIG_MTD_UBI_WL_THRESHOLD | |
102 | ||
103 | /* | |
85c6e6e2 | 104 | * When a physical eraseblock is moved, the WL sub-system has to pick the target |
801c135c AB |
105 | * physical eraseblock to move to. The simplest way would be just to pick the |
106 | * one with the highest erase counter. But in certain workloads this could lead | |
107 | * to an unlimited wear of one or few physical eraseblock. Indeed, imagine a | |
108 | * situation when the picked physical eraseblock is constantly erased after the | |
109 | * data is written to it. So, we have a constant which limits the highest erase | |
85c6e6e2 AB |
110 | * counter of the free physical eraseblock to pick. Namely, the WL sub-system |
111 | * does not pick eraseblocks with erase counter greater then the lowest erase | |
801c135c AB |
112 | * counter plus %WL_FREE_MAX_DIFF. |
113 | */ | |
114 | #define WL_FREE_MAX_DIFF (2*UBI_WL_THRESHOLD) | |
115 | ||
116 | /* | |
117 | * Maximum number of consecutive background thread failures which is enough to | |
118 | * switch to read-only mode. | |
119 | */ | |
120 | #define WL_MAX_FAILURES 32 | |
121 | ||
801c135c AB |
122 | /** |
123 | * struct ubi_wl_prot_entry - PEB protection entry. | |
124 | * @rb_pnum: link in the @wl->prot.pnum RB-tree | |
125 | * @rb_aec: link in the @wl->prot.aec RB-tree | |
126 | * @abs_ec: the absolute erase counter value when the protection ends | |
127 | * @e: the wear-leveling entry of the physical eraseblock under protection | |
128 | * | |
85c6e6e2 AB |
129 | * When the WL sub-system returns a physical eraseblock, the physical |
130 | * eraseblock is protected from being moved for some "time". For this reason, | |
131 | * the physical eraseblock is not directly moved from the @wl->free tree to the | |
132 | * @wl->used tree. There is one more tree in between where this physical | |
133 | * eraseblock is temporarily stored (@wl->prot). | |
801c135c AB |
134 | * |
135 | * All this protection stuff is needed because: | |
136 | * o we don't want to move physical eraseblocks just after we have given them | |
137 | * to the user; instead, we first want to let users fill them up with data; | |
138 | * | |
139 | * o there is a chance that the user will put the physical eraseblock very | |
140 | * soon, so it makes sense not to move it for some time, but wait; this is | |
141 | * especially important in case of "short term" physical eraseblocks. | |
142 | * | |
143 | * Physical eraseblocks stay protected only for limited time. But the "time" is | |
144 | * measured in erase cycles in this case. This is implemented with help of the | |
145 | * absolute erase counter (@wl->abs_ec). When it reaches certain value, the | |
146 | * physical eraseblocks are moved from the protection trees (@wl->prot.*) to | |
147 | * the @wl->used tree. | |
148 | * | |
149 | * Protected physical eraseblocks are searched by physical eraseblock number | |
150 | * (when they are put) and by the absolute erase counter (to check if it is | |
151 | * time to move them to the @wl->used tree). So there are actually 2 RB-trees | |
152 | * storing the protected physical eraseblocks: @wl->prot.pnum and | |
153 | * @wl->prot.aec. They are referred to as the "protection" trees. The | |
154 | * first one is indexed by the physical eraseblock number. The second one is | |
155 | * indexed by the absolute erase counter. Both trees store | |
156 | * &struct ubi_wl_prot_entry objects. | |
157 | * | |
158 | * Each physical eraseblock has 2 main states: free and used. The former state | |
159 | * corresponds to the @wl->free tree. The latter state is split up on several | |
160 | * sub-states: | |
161 | * o the WL movement is allowed (@wl->used tree); | |
162 | * o the WL movement is temporarily prohibited (@wl->prot.pnum and | |
163 | * @wl->prot.aec trees); | |
164 | * o scrubbing is needed (@wl->scrub tree). | |
165 | * | |
166 | * Depending on the sub-state, wear-leveling entries of the used physical | |
167 | * eraseblocks may be kept in one of those trees. | |
168 | */ | |
169 | struct ubi_wl_prot_entry { | |
170 | struct rb_node rb_pnum; | |
171 | struct rb_node rb_aec; | |
172 | unsigned long long abs_ec; | |
173 | struct ubi_wl_entry *e; | |
174 | }; | |
175 | ||
176 | /** | |
177 | * struct ubi_work - UBI work description data structure. | |
178 | * @list: a link in the list of pending works | |
179 | * @func: worker function | |
180 | * @priv: private data of the worker function | |
801c135c AB |
181 | * @e: physical eraseblock to erase |
182 | * @torture: if the physical eraseblock has to be tortured | |
183 | * | |
184 | * The @func pointer points to the worker function. If the @cancel argument is | |
185 | * not zero, the worker has to free the resources and exit immediately. The | |
186 | * worker has to return zero in case of success and a negative error code in | |
187 | * case of failure. | |
188 | */ | |
189 | struct ubi_work { | |
190 | struct list_head list; | |
191 | int (*func)(struct ubi_device *ubi, struct ubi_work *wrk, int cancel); | |
192 | /* The below fields are only relevant to erasure works */ | |
193 | struct ubi_wl_entry *e; | |
194 | int torture; | |
195 | }; | |
196 | ||
197 | #ifdef CONFIG_MTD_UBI_DEBUG_PARANOID | |
e88d6e10 | 198 | static int paranoid_check_ec(struct ubi_device *ubi, int pnum, int ec); |
801c135c AB |
199 | static int paranoid_check_in_wl_tree(struct ubi_wl_entry *e, |
200 | struct rb_root *root); | |
201 | #else | |
202 | #define paranoid_check_ec(ubi, pnum, ec) 0 | |
203 | #define paranoid_check_in_wl_tree(e, root) | |
204 | #endif | |
205 | ||
801c135c AB |
206 | /** |
207 | * wl_tree_add - add a wear-leveling entry to a WL RB-tree. | |
208 | * @e: the wear-leveling entry to add | |
209 | * @root: the root of the tree | |
210 | * | |
211 | * Note, we use (erase counter, physical eraseblock number) pairs as keys in | |
212 | * the @ubi->used and @ubi->free RB-trees. | |
213 | */ | |
214 | static void wl_tree_add(struct ubi_wl_entry *e, struct rb_root *root) | |
215 | { | |
216 | struct rb_node **p, *parent = NULL; | |
217 | ||
218 | p = &root->rb_node; | |
219 | while (*p) { | |
220 | struct ubi_wl_entry *e1; | |
221 | ||
222 | parent = *p; | |
223 | e1 = rb_entry(parent, struct ubi_wl_entry, rb); | |
224 | ||
225 | if (e->ec < e1->ec) | |
226 | p = &(*p)->rb_left; | |
227 | else if (e->ec > e1->ec) | |
228 | p = &(*p)->rb_right; | |
229 | else { | |
230 | ubi_assert(e->pnum != e1->pnum); | |
231 | if (e->pnum < e1->pnum) | |
232 | p = &(*p)->rb_left; | |
233 | else | |
234 | p = &(*p)->rb_right; | |
235 | } | |
236 | } | |
237 | ||
238 | rb_link_node(&e->rb, parent, p); | |
239 | rb_insert_color(&e->rb, root); | |
240 | } | |
241 | ||
801c135c AB |
242 | /** |
243 | * do_work - do one pending work. | |
244 | * @ubi: UBI device description object | |
245 | * | |
246 | * This function returns zero in case of success and a negative error code in | |
247 | * case of failure. | |
248 | */ | |
249 | static int do_work(struct ubi_device *ubi) | |
250 | { | |
251 | int err; | |
252 | struct ubi_work *wrk; | |
253 | ||
43f9b25a AB |
254 | cond_resched(); |
255 | ||
593dd33c AB |
256 | /* |
257 | * @ubi->work_sem is used to synchronize with the workers. Workers take | |
258 | * it in read mode, so many of them may be doing works at a time. But | |
259 | * the queue flush code has to be sure the whole queue of works is | |
260 | * done, and it takes the mutex in write mode. | |
261 | */ | |
262 | down_read(&ubi->work_sem); | |
801c135c | 263 | spin_lock(&ubi->wl_lock); |
801c135c AB |
264 | if (list_empty(&ubi->works)) { |
265 | spin_unlock(&ubi->wl_lock); | |
593dd33c | 266 | up_read(&ubi->work_sem); |
801c135c AB |
267 | return 0; |
268 | } | |
269 | ||
270 | wrk = list_entry(ubi->works.next, struct ubi_work, list); | |
271 | list_del(&wrk->list); | |
16f557ec AB |
272 | ubi->works_count -= 1; |
273 | ubi_assert(ubi->works_count >= 0); | |
801c135c AB |
274 | spin_unlock(&ubi->wl_lock); |
275 | ||
276 | /* | |
277 | * Call the worker function. Do not touch the work structure | |
278 | * after this call as it will have been freed or reused by that | |
279 | * time by the worker function. | |
280 | */ | |
281 | err = wrk->func(ubi, wrk, 0); | |
282 | if (err) | |
283 | ubi_err("work failed with error code %d", err); | |
593dd33c | 284 | up_read(&ubi->work_sem); |
16f557ec | 285 | |
801c135c AB |
286 | return err; |
287 | } | |
288 | ||
289 | /** | |
290 | * produce_free_peb - produce a free physical eraseblock. | |
291 | * @ubi: UBI device description object | |
292 | * | |
293 | * This function tries to make a free PEB by means of synchronous execution of | |
294 | * pending works. This may be needed if, for example the background thread is | |
295 | * disabled. Returns zero in case of success and a negative error code in case | |
296 | * of failure. | |
297 | */ | |
298 | static int produce_free_peb(struct ubi_device *ubi) | |
299 | { | |
300 | int err; | |
301 | ||
302 | spin_lock(&ubi->wl_lock); | |
5abde384 | 303 | while (!ubi->free.rb_node) { |
801c135c AB |
304 | spin_unlock(&ubi->wl_lock); |
305 | ||
306 | dbg_wl("do one work synchronously"); | |
307 | err = do_work(ubi); | |
308 | if (err) | |
309 | return err; | |
310 | ||
311 | spin_lock(&ubi->wl_lock); | |
312 | } | |
313 | spin_unlock(&ubi->wl_lock); | |
314 | ||
315 | return 0; | |
316 | } | |
317 | ||
318 | /** | |
319 | * in_wl_tree - check if wear-leveling entry is present in a WL RB-tree. | |
320 | * @e: the wear-leveling entry to check | |
321 | * @root: the root of the tree | |
322 | * | |
323 | * This function returns non-zero if @e is in the @root RB-tree and zero if it | |
324 | * is not. | |
325 | */ | |
326 | static int in_wl_tree(struct ubi_wl_entry *e, struct rb_root *root) | |
327 | { | |
328 | struct rb_node *p; | |
329 | ||
330 | p = root->rb_node; | |
331 | while (p) { | |
332 | struct ubi_wl_entry *e1; | |
333 | ||
334 | e1 = rb_entry(p, struct ubi_wl_entry, rb); | |
335 | ||
336 | if (e->pnum == e1->pnum) { | |
337 | ubi_assert(e == e1); | |
338 | return 1; | |
339 | } | |
340 | ||
341 | if (e->ec < e1->ec) | |
342 | p = p->rb_left; | |
343 | else if (e->ec > e1->ec) | |
344 | p = p->rb_right; | |
345 | else { | |
346 | ubi_assert(e->pnum != e1->pnum); | |
347 | if (e->pnum < e1->pnum) | |
348 | p = p->rb_left; | |
349 | else | |
350 | p = p->rb_right; | |
351 | } | |
352 | } | |
353 | ||
354 | return 0; | |
355 | } | |
356 | ||
357 | /** | |
358 | * prot_tree_add - add physical eraseblock to protection trees. | |
359 | * @ubi: UBI device description object | |
360 | * @e: the physical eraseblock to add | |
361 | * @pe: protection entry object to use | |
362 | * @abs_ec: absolute erase counter value when this physical eraseblock has | |
363 | * to be removed from the protection trees. | |
364 | * | |
365 | * @wl->lock has to be locked. | |
366 | */ | |
367 | static void prot_tree_add(struct ubi_device *ubi, struct ubi_wl_entry *e, | |
368 | struct ubi_wl_prot_entry *pe, int abs_ec) | |
369 | { | |
370 | struct rb_node **p, *parent = NULL; | |
371 | struct ubi_wl_prot_entry *pe1; | |
372 | ||
373 | pe->e = e; | |
374 | pe->abs_ec = ubi->abs_ec + abs_ec; | |
375 | ||
376 | p = &ubi->prot.pnum.rb_node; | |
377 | while (*p) { | |
378 | parent = *p; | |
379 | pe1 = rb_entry(parent, struct ubi_wl_prot_entry, rb_pnum); | |
380 | ||
381 | if (e->pnum < pe1->e->pnum) | |
382 | p = &(*p)->rb_left; | |
383 | else | |
384 | p = &(*p)->rb_right; | |
385 | } | |
386 | rb_link_node(&pe->rb_pnum, parent, p); | |
387 | rb_insert_color(&pe->rb_pnum, &ubi->prot.pnum); | |
388 | ||
389 | p = &ubi->prot.aec.rb_node; | |
390 | parent = NULL; | |
391 | while (*p) { | |
392 | parent = *p; | |
393 | pe1 = rb_entry(parent, struct ubi_wl_prot_entry, rb_aec); | |
394 | ||
395 | if (pe->abs_ec < pe1->abs_ec) | |
396 | p = &(*p)->rb_left; | |
397 | else | |
398 | p = &(*p)->rb_right; | |
399 | } | |
400 | rb_link_node(&pe->rb_aec, parent, p); | |
401 | rb_insert_color(&pe->rb_aec, &ubi->prot.aec); | |
402 | } | |
403 | ||
404 | /** | |
405 | * find_wl_entry - find wear-leveling entry closest to certain erase counter. | |
406 | * @root: the RB-tree where to look for | |
407 | * @max: highest possible erase counter | |
408 | * | |
409 | * This function looks for a wear leveling entry with erase counter closest to | |
410 | * @max and less then @max. | |
411 | */ | |
412 | static struct ubi_wl_entry *find_wl_entry(struct rb_root *root, int max) | |
413 | { | |
414 | struct rb_node *p; | |
415 | struct ubi_wl_entry *e; | |
416 | ||
417 | e = rb_entry(rb_first(root), struct ubi_wl_entry, rb); | |
418 | max += e->ec; | |
419 | ||
420 | p = root->rb_node; | |
421 | while (p) { | |
422 | struct ubi_wl_entry *e1; | |
423 | ||
424 | e1 = rb_entry(p, struct ubi_wl_entry, rb); | |
425 | if (e1->ec >= max) | |
426 | p = p->rb_left; | |
427 | else { | |
428 | p = p->rb_right; | |
429 | e = e1; | |
430 | } | |
431 | } | |
432 | ||
433 | return e; | |
434 | } | |
435 | ||
436 | /** | |
437 | * ubi_wl_get_peb - get a physical eraseblock. | |
438 | * @ubi: UBI device description object | |
439 | * @dtype: type of data which will be stored in this physical eraseblock | |
440 | * | |
441 | * This function returns a physical eraseblock in case of success and a | |
442 | * negative error code in case of failure. Might sleep. | |
443 | */ | |
444 | int ubi_wl_get_peb(struct ubi_device *ubi, int dtype) | |
445 | { | |
446 | int err, protect, medium_ec; | |
447 | struct ubi_wl_entry *e, *first, *last; | |
448 | struct ubi_wl_prot_entry *pe; | |
449 | ||
450 | ubi_assert(dtype == UBI_LONGTERM || dtype == UBI_SHORTTERM || | |
451 | dtype == UBI_UNKNOWN); | |
452 | ||
33818bbb | 453 | pe = kmalloc(sizeof(struct ubi_wl_prot_entry), GFP_NOFS); |
801c135c AB |
454 | if (!pe) |
455 | return -ENOMEM; | |
456 | ||
457 | retry: | |
458 | spin_lock(&ubi->wl_lock); | |
5abde384 | 459 | if (!ubi->free.rb_node) { |
801c135c AB |
460 | if (ubi->works_count == 0) { |
461 | ubi_assert(list_empty(&ubi->works)); | |
462 | ubi_err("no free eraseblocks"); | |
463 | spin_unlock(&ubi->wl_lock); | |
464 | kfree(pe); | |
465 | return -ENOSPC; | |
466 | } | |
467 | spin_unlock(&ubi->wl_lock); | |
468 | ||
469 | err = produce_free_peb(ubi); | |
470 | if (err < 0) { | |
471 | kfree(pe); | |
472 | return err; | |
473 | } | |
474 | goto retry; | |
475 | } | |
476 | ||
477 | switch (dtype) { | |
478 | case UBI_LONGTERM: | |
479 | /* | |
480 | * For long term data we pick a physical eraseblock | |
481 | * with high erase counter. But the highest erase | |
482 | * counter we can pick is bounded by the the lowest | |
483 | * erase counter plus %WL_FREE_MAX_DIFF. | |
484 | */ | |
485 | e = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); | |
486 | protect = LT_PROTECTION; | |
487 | break; | |
488 | case UBI_UNKNOWN: | |
489 | /* | |
490 | * For unknown data we pick a physical eraseblock with | |
491 | * medium erase counter. But we by no means can pick a | |
492 | * physical eraseblock with erase counter greater or | |
493 | * equivalent than the lowest erase counter plus | |
494 | * %WL_FREE_MAX_DIFF. | |
495 | */ | |
496 | first = rb_entry(rb_first(&ubi->free), | |
497 | struct ubi_wl_entry, rb); | |
498 | last = rb_entry(rb_last(&ubi->free), | |
499 | struct ubi_wl_entry, rb); | |
500 | ||
501 | if (last->ec - first->ec < WL_FREE_MAX_DIFF) | |
502 | e = rb_entry(ubi->free.rb_node, | |
503 | struct ubi_wl_entry, rb); | |
504 | else { | |
505 | medium_ec = (first->ec + WL_FREE_MAX_DIFF)/2; | |
506 | e = find_wl_entry(&ubi->free, medium_ec); | |
507 | } | |
508 | protect = U_PROTECTION; | |
509 | break; | |
510 | case UBI_SHORTTERM: | |
511 | /* | |
512 | * For short term data we pick a physical eraseblock | |
513 | * with the lowest erase counter as we expect it will | |
514 | * be erased soon. | |
515 | */ | |
516 | e = rb_entry(rb_first(&ubi->free), | |
517 | struct ubi_wl_entry, rb); | |
518 | protect = ST_PROTECTION; | |
519 | break; | |
520 | default: | |
521 | protect = 0; | |
522 | e = NULL; | |
523 | BUG(); | |
524 | } | |
525 | ||
526 | /* | |
527 | * Move the physical eraseblock to the protection trees where it will | |
528 | * be protected from being moved for some time. | |
529 | */ | |
5abde384 AB |
530 | paranoid_check_in_wl_tree(e, &ubi->free); |
531 | rb_erase(&e->rb, &ubi->free); | |
801c135c AB |
532 | prot_tree_add(ubi, e, pe, protect); |
533 | ||
534 | dbg_wl("PEB %d EC %d, protection %d", e->pnum, e->ec, protect); | |
535 | spin_unlock(&ubi->wl_lock); | |
536 | ||
537 | return e->pnum; | |
538 | } | |
539 | ||
540 | /** | |
541 | * prot_tree_del - remove a physical eraseblock from the protection trees | |
542 | * @ubi: UBI device description object | |
543 | * @pnum: the physical eraseblock to remove | |
43f9b25a AB |
544 | * |
545 | * This function returns PEB @pnum from the protection trees and returns zero | |
546 | * in case of success and %-ENODEV if the PEB was not found in the protection | |
547 | * trees. | |
801c135c | 548 | */ |
43f9b25a | 549 | static int prot_tree_del(struct ubi_device *ubi, int pnum) |
801c135c AB |
550 | { |
551 | struct rb_node *p; | |
552 | struct ubi_wl_prot_entry *pe = NULL; | |
553 | ||
554 | p = ubi->prot.pnum.rb_node; | |
555 | while (p) { | |
556 | ||
557 | pe = rb_entry(p, struct ubi_wl_prot_entry, rb_pnum); | |
558 | ||
559 | if (pnum == pe->e->pnum) | |
43f9b25a | 560 | goto found; |
801c135c AB |
561 | |
562 | if (pnum < pe->e->pnum) | |
563 | p = p->rb_left; | |
564 | else | |
565 | p = p->rb_right; | |
566 | } | |
567 | ||
43f9b25a AB |
568 | return -ENODEV; |
569 | ||
570 | found: | |
801c135c AB |
571 | ubi_assert(pe->e->pnum == pnum); |
572 | rb_erase(&pe->rb_aec, &ubi->prot.aec); | |
573 | rb_erase(&pe->rb_pnum, &ubi->prot.pnum); | |
574 | kfree(pe); | |
43f9b25a | 575 | return 0; |
801c135c AB |
576 | } |
577 | ||
578 | /** | |
579 | * sync_erase - synchronously erase a physical eraseblock. | |
580 | * @ubi: UBI device description object | |
581 | * @e: the the physical eraseblock to erase | |
582 | * @torture: if the physical eraseblock has to be tortured | |
583 | * | |
584 | * This function returns zero in case of success and a negative error code in | |
585 | * case of failure. | |
586 | */ | |
587 | static int sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e, int torture) | |
588 | { | |
589 | int err; | |
590 | struct ubi_ec_hdr *ec_hdr; | |
591 | unsigned long long ec = e->ec; | |
592 | ||
593 | dbg_wl("erase PEB %d, old EC %llu", e->pnum, ec); | |
594 | ||
595 | err = paranoid_check_ec(ubi, e->pnum, e->ec); | |
596 | if (err > 0) | |
597 | return -EINVAL; | |
598 | ||
33818bbb | 599 | ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_NOFS); |
801c135c AB |
600 | if (!ec_hdr) |
601 | return -ENOMEM; | |
602 | ||
603 | err = ubi_io_sync_erase(ubi, e->pnum, torture); | |
604 | if (err < 0) | |
605 | goto out_free; | |
606 | ||
607 | ec += err; | |
608 | if (ec > UBI_MAX_ERASECOUNTER) { | |
609 | /* | |
610 | * Erase counter overflow. Upgrade UBI and use 64-bit | |
611 | * erase counters internally. | |
612 | */ | |
613 | ubi_err("erase counter overflow at PEB %d, EC %llu", | |
614 | e->pnum, ec); | |
615 | err = -EINVAL; | |
616 | goto out_free; | |
617 | } | |
618 | ||
619 | dbg_wl("erased PEB %d, new EC %llu", e->pnum, ec); | |
620 | ||
3261ebd7 | 621 | ec_hdr->ec = cpu_to_be64(ec); |
801c135c AB |
622 | |
623 | err = ubi_io_write_ec_hdr(ubi, e->pnum, ec_hdr); | |
624 | if (err) | |
625 | goto out_free; | |
626 | ||
627 | e->ec = ec; | |
628 | spin_lock(&ubi->wl_lock); | |
629 | if (e->ec > ubi->max_ec) | |
630 | ubi->max_ec = e->ec; | |
631 | spin_unlock(&ubi->wl_lock); | |
632 | ||
633 | out_free: | |
634 | kfree(ec_hdr); | |
635 | return err; | |
636 | } | |
637 | ||
638 | /** | |
639 | * check_protection_over - check if it is time to stop protecting some | |
640 | * physical eraseblocks. | |
641 | * @ubi: UBI device description object | |
642 | * | |
643 | * This function is called after each erase operation, when the absolute erase | |
644 | * counter is incremented, to check if some physical eraseblock have not to be | |
645 | * protected any longer. These physical eraseblocks are moved from the | |
646 | * protection trees to the used tree. | |
647 | */ | |
648 | static void check_protection_over(struct ubi_device *ubi) | |
649 | { | |
650 | struct ubi_wl_prot_entry *pe; | |
651 | ||
652 | /* | |
653 | * There may be several protected physical eraseblock to remove, | |
654 | * process them all. | |
655 | */ | |
656 | while (1) { | |
657 | spin_lock(&ubi->wl_lock); | |
5abde384 | 658 | if (!ubi->prot.aec.rb_node) { |
801c135c AB |
659 | spin_unlock(&ubi->wl_lock); |
660 | break; | |
661 | } | |
662 | ||
663 | pe = rb_entry(rb_first(&ubi->prot.aec), | |
664 | struct ubi_wl_prot_entry, rb_aec); | |
665 | ||
666 | if (pe->abs_ec > ubi->abs_ec) { | |
667 | spin_unlock(&ubi->wl_lock); | |
668 | break; | |
669 | } | |
670 | ||
671 | dbg_wl("PEB %d protection over, abs_ec %llu, PEB abs_ec %llu", | |
672 | pe->e->pnum, ubi->abs_ec, pe->abs_ec); | |
673 | rb_erase(&pe->rb_aec, &ubi->prot.aec); | |
674 | rb_erase(&pe->rb_pnum, &ubi->prot.pnum); | |
5abde384 | 675 | wl_tree_add(pe->e, &ubi->used); |
801c135c AB |
676 | spin_unlock(&ubi->wl_lock); |
677 | ||
678 | kfree(pe); | |
679 | cond_resched(); | |
680 | } | |
681 | } | |
682 | ||
683 | /** | |
684 | * schedule_ubi_work - schedule a work. | |
685 | * @ubi: UBI device description object | |
686 | * @wrk: the work to schedule | |
687 | * | |
688 | * This function enqueues a work defined by @wrk to the tail of the pending | |
689 | * works list. | |
690 | */ | |
691 | static void schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk) | |
692 | { | |
693 | spin_lock(&ubi->wl_lock); | |
694 | list_add_tail(&wrk->list, &ubi->works); | |
695 | ubi_assert(ubi->works_count >= 0); | |
696 | ubi->works_count += 1; | |
697 | if (ubi->thread_enabled) | |
698 | wake_up_process(ubi->bgt_thread); | |
699 | spin_unlock(&ubi->wl_lock); | |
700 | } | |
701 | ||
702 | static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk, | |
703 | int cancel); | |
704 | ||
705 | /** | |
706 | * schedule_erase - schedule an erase work. | |
707 | * @ubi: UBI device description object | |
708 | * @e: the WL entry of the physical eraseblock to erase | |
709 | * @torture: if the physical eraseblock has to be tortured | |
710 | * | |
711 | * This function returns zero in case of success and a %-ENOMEM in case of | |
712 | * failure. | |
713 | */ | |
714 | static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e, | |
715 | int torture) | |
716 | { | |
717 | struct ubi_work *wl_wrk; | |
718 | ||
719 | dbg_wl("schedule erasure of PEB %d, EC %d, torture %d", | |
720 | e->pnum, e->ec, torture); | |
721 | ||
33818bbb | 722 | wl_wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS); |
801c135c AB |
723 | if (!wl_wrk) |
724 | return -ENOMEM; | |
725 | ||
726 | wl_wrk->func = &erase_worker; | |
727 | wl_wrk->e = e; | |
728 | wl_wrk->torture = torture; | |
729 | ||
730 | schedule_ubi_work(ubi, wl_wrk); | |
731 | return 0; | |
732 | } | |
733 | ||
734 | /** | |
735 | * wear_leveling_worker - wear-leveling worker function. | |
736 | * @ubi: UBI device description object | |
737 | * @wrk: the work object | |
738 | * @cancel: non-zero if the worker has to free memory and exit | |
739 | * | |
740 | * This function copies a more worn out physical eraseblock to a less worn out | |
741 | * one. Returns zero in case of success and a negative error code in case of | |
742 | * failure. | |
743 | */ | |
744 | static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk, | |
745 | int cancel) | |
746 | { | |
43f9b25a | 747 | int err, put = 0, scrubbing = 0, protect = 0; |
c18a8418 | 748 | struct ubi_wl_prot_entry *uninitialized_var(pe); |
801c135c AB |
749 | struct ubi_wl_entry *e1, *e2; |
750 | struct ubi_vid_hdr *vid_hdr; | |
751 | ||
752 | kfree(wrk); | |
753 | ||
754 | if (cancel) | |
755 | return 0; | |
756 | ||
33818bbb | 757 | vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS); |
801c135c AB |
758 | if (!vid_hdr) |
759 | return -ENOMEM; | |
760 | ||
43f9b25a | 761 | mutex_lock(&ubi->move_mutex); |
801c135c | 762 | spin_lock(&ubi->wl_lock); |
43f9b25a AB |
763 | ubi_assert(!ubi->move_from && !ubi->move_to); |
764 | ubi_assert(!ubi->move_to_put); | |
801c135c | 765 | |
43f9b25a | 766 | if (!ubi->free.rb_node || |
5abde384 | 767 | (!ubi->used.rb_node && !ubi->scrub.rb_node)) { |
801c135c | 768 | /* |
43f9b25a AB |
769 | * No free physical eraseblocks? Well, they must be waiting in |
770 | * the queue to be erased. Cancel movement - it will be | |
771 | * triggered again when a free physical eraseblock appears. | |
801c135c AB |
772 | * |
773 | * No used physical eraseblocks? They must be temporarily | |
774 | * protected from being moved. They will be moved to the | |
775 | * @ubi->used tree later and the wear-leveling will be | |
776 | * triggered again. | |
777 | */ | |
778 | dbg_wl("cancel WL, a list is empty: free %d, used %d", | |
5abde384 | 779 | !ubi->free.rb_node, !ubi->used.rb_node); |
43f9b25a | 780 | goto out_cancel; |
801c135c AB |
781 | } |
782 | ||
5abde384 | 783 | if (!ubi->scrub.rb_node) { |
801c135c AB |
784 | /* |
785 | * Now pick the least worn-out used physical eraseblock and a | |
786 | * highly worn-out free physical eraseblock. If the erase | |
787 | * counters differ much enough, start wear-leveling. | |
788 | */ | |
789 | e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, rb); | |
790 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); | |
791 | ||
792 | if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD)) { | |
793 | dbg_wl("no WL needed: min used EC %d, max free EC %d", | |
794 | e1->ec, e2->ec); | |
43f9b25a | 795 | goto out_cancel; |
801c135c | 796 | } |
5abde384 AB |
797 | paranoid_check_in_wl_tree(e1, &ubi->used); |
798 | rb_erase(&e1->rb, &ubi->used); | |
801c135c AB |
799 | dbg_wl("move PEB %d EC %d to PEB %d EC %d", |
800 | e1->pnum, e1->ec, e2->pnum, e2->ec); | |
801 | } else { | |
43f9b25a AB |
802 | /* Perform scrubbing */ |
803 | scrubbing = 1; | |
801c135c AB |
804 | e1 = rb_entry(rb_first(&ubi->scrub), struct ubi_wl_entry, rb); |
805 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); | |
5abde384 | 806 | paranoid_check_in_wl_tree(e1, &ubi->scrub); |
d2c46855 | 807 | rb_erase(&e1->rb, &ubi->scrub); |
801c135c AB |
808 | dbg_wl("scrub PEB %d to PEB %d", e1->pnum, e2->pnum); |
809 | } | |
810 | ||
5abde384 AB |
811 | paranoid_check_in_wl_tree(e2, &ubi->free); |
812 | rb_erase(&e2->rb, &ubi->free); | |
801c135c AB |
813 | ubi->move_from = e1; |
814 | ubi->move_to = e2; | |
815 | spin_unlock(&ubi->wl_lock); | |
816 | ||
817 | /* | |
818 | * Now we are going to copy physical eraseblock @e1->pnum to @e2->pnum. | |
819 | * We so far do not know which logical eraseblock our physical | |
820 | * eraseblock (@e1) belongs to. We have to read the volume identifier | |
821 | * header first. | |
43f9b25a AB |
822 | * |
823 | * Note, we are protected from this PEB being unmapped and erased. The | |
824 | * 'ubi_wl_put_peb()' would wait for moving to be finished if the PEB | |
825 | * which is being moved was unmapped. | |
801c135c AB |
826 | */ |
827 | ||
828 | err = ubi_io_read_vid_hdr(ubi, e1->pnum, vid_hdr, 0); | |
829 | if (err && err != UBI_IO_BITFLIPS) { | |
830 | if (err == UBI_IO_PEB_FREE) { | |
831 | /* | |
832 | * We are trying to move PEB without a VID header. UBI | |
833 | * always write VID headers shortly after the PEB was | |
834 | * given, so we have a situation when it did not have | |
835 | * chance to write it down because it was preempted. | |
836 | * Just re-schedule the work, so that next time it will | |
837 | * likely have the VID header in place. | |
838 | */ | |
839 | dbg_wl("PEB %d has no VID header", e1->pnum); | |
43f9b25a | 840 | goto out_not_moved; |
801c135c | 841 | } |
43f9b25a AB |
842 | |
843 | ubi_err("error %d while reading VID header from PEB %d", | |
844 | err, e1->pnum); | |
845 | if (err > 0) | |
846 | err = -EIO; | |
847 | goto out_error; | |
801c135c AB |
848 | } |
849 | ||
850 | err = ubi_eba_copy_leb(ubi, e1->pnum, e2->pnum, vid_hdr); | |
851 | if (err) { | |
43f9b25a AB |
852 | |
853 | if (err < 0) | |
854 | goto out_error; | |
855 | if (err == 1) | |
856 | goto out_not_moved; | |
857 | ||
858 | /* | |
859 | * For some reason the LEB was not moved - it might be because | |
860 | * the volume is being deleted. We should prevent this PEB from | |
861 | * being selected for wear-levelling movement for some "time", | |
862 | * so put it to the protection tree. | |
863 | */ | |
864 | ||
865 | dbg_wl("cancelled moving PEB %d", e1->pnum); | |
866 | pe = kmalloc(sizeof(struct ubi_wl_prot_entry), GFP_NOFS); | |
867 | if (!pe) { | |
868 | err = -ENOMEM; | |
869 | goto out_error; | |
870 | } | |
871 | ||
872 | protect = 1; | |
801c135c AB |
873 | } |
874 | ||
875 | ubi_free_vid_hdr(ubi, vid_hdr); | |
8c1e6ee1 AB |
876 | if (scrubbing && !protect) |
877 | ubi_msg("scrubbed PEB %d, data moved to PEB %d", | |
878 | e1->pnum, e2->pnum); | |
879 | ||
801c135c | 880 | spin_lock(&ubi->wl_lock); |
43f9b25a AB |
881 | if (protect) |
882 | prot_tree_add(ubi, e1, pe, protect); | |
801c135c | 883 | if (!ubi->move_to_put) |
5abde384 | 884 | wl_tree_add(e2, &ubi->used); |
801c135c AB |
885 | else |
886 | put = 1; | |
887 | ubi->move_from = ubi->move_to = NULL; | |
43f9b25a | 888 | ubi->move_to_put = ubi->wl_scheduled = 0; |
801c135c AB |
889 | spin_unlock(&ubi->wl_lock); |
890 | ||
891 | if (put) { | |
892 | /* | |
893 | * Well, the target PEB was put meanwhile, schedule it for | |
894 | * erasure. | |
895 | */ | |
896 | dbg_wl("PEB %d was put meanwhile, erase", e2->pnum); | |
897 | err = schedule_erase(ubi, e2, 0); | |
43f9b25a AB |
898 | if (err) |
899 | goto out_error; | |
801c135c AB |
900 | } |
901 | ||
43f9b25a AB |
902 | if (!protect) { |
903 | err = schedule_erase(ubi, e1, 0); | |
904 | if (err) | |
905 | goto out_error; | |
801c135c AB |
906 | } |
907 | ||
43f9b25a | 908 | |
801c135c | 909 | dbg_wl("done"); |
43f9b25a AB |
910 | mutex_unlock(&ubi->move_mutex); |
911 | return 0; | |
801c135c AB |
912 | |
913 | /* | |
43f9b25a AB |
914 | * For some reasons the LEB was not moved, might be an error, might be |
915 | * something else. @e1 was not changed, so return it back. @e2 might | |
916 | * be changed, schedule it for erasure. | |
801c135c | 917 | */ |
43f9b25a | 918 | out_not_moved: |
801c135c AB |
919 | ubi_free_vid_hdr(ubi, vid_hdr); |
920 | spin_lock(&ubi->wl_lock); | |
43f9b25a AB |
921 | if (scrubbing) |
922 | wl_tree_add(e1, &ubi->scrub); | |
801c135c | 923 | else |
5abde384 | 924 | wl_tree_add(e1, &ubi->used); |
801c135c | 925 | ubi->move_from = ubi->move_to = NULL; |
43f9b25a | 926 | ubi->move_to_put = ubi->wl_scheduled = 0; |
801c135c AB |
927 | spin_unlock(&ubi->wl_lock); |
928 | ||
801c135c | 929 | err = schedule_erase(ubi, e2, 0); |
43f9b25a AB |
930 | if (err) |
931 | goto out_error; | |
932 | ||
933 | mutex_unlock(&ubi->move_mutex); | |
934 | return 0; | |
935 | ||
936 | out_error: | |
937 | ubi_err("error %d while moving PEB %d to PEB %d", | |
938 | err, e1->pnum, e2->pnum); | |
801c135c | 939 | |
43f9b25a AB |
940 | ubi_free_vid_hdr(ubi, vid_hdr); |
941 | spin_lock(&ubi->wl_lock); | |
942 | ubi->move_from = ubi->move_to = NULL; | |
943 | ubi->move_to_put = ubi->wl_scheduled = 0; | |
944 | spin_unlock(&ubi->wl_lock); | |
945 | ||
946 | kmem_cache_free(ubi_wl_entry_slab, e1); | |
947 | kmem_cache_free(ubi_wl_entry_slab, e2); | |
948 | ubi_ro_mode(ubi); | |
949 | ||
950 | mutex_unlock(&ubi->move_mutex); | |
801c135c | 951 | return err; |
43f9b25a AB |
952 | |
953 | out_cancel: | |
954 | ubi->wl_scheduled = 0; | |
955 | spin_unlock(&ubi->wl_lock); | |
956 | mutex_unlock(&ubi->move_mutex); | |
957 | ubi_free_vid_hdr(ubi, vid_hdr); | |
958 | return 0; | |
801c135c AB |
959 | } |
960 | ||
961 | /** | |
962 | * ensure_wear_leveling - schedule wear-leveling if it is needed. | |
963 | * @ubi: UBI device description object | |
964 | * | |
965 | * This function checks if it is time to start wear-leveling and schedules it | |
966 | * if yes. This function returns zero in case of success and a negative error | |
967 | * code in case of failure. | |
968 | */ | |
969 | static int ensure_wear_leveling(struct ubi_device *ubi) | |
970 | { | |
971 | int err = 0; | |
972 | struct ubi_wl_entry *e1; | |
973 | struct ubi_wl_entry *e2; | |
974 | struct ubi_work *wrk; | |
975 | ||
976 | spin_lock(&ubi->wl_lock); | |
977 | if (ubi->wl_scheduled) | |
978 | /* Wear-leveling is already in the work queue */ | |
979 | goto out_unlock; | |
980 | ||
981 | /* | |
982 | * If the ubi->scrub tree is not empty, scrubbing is needed, and the | |
983 | * the WL worker has to be scheduled anyway. | |
984 | */ | |
5abde384 AB |
985 | if (!ubi->scrub.rb_node) { |
986 | if (!ubi->used.rb_node || !ubi->free.rb_node) | |
801c135c AB |
987 | /* No physical eraseblocks - no deal */ |
988 | goto out_unlock; | |
989 | ||
990 | /* | |
991 | * We schedule wear-leveling only if the difference between the | |
992 | * lowest erase counter of used physical eraseblocks and a high | |
993 | * erase counter of free physical eraseblocks is greater then | |
994 | * %UBI_WL_THRESHOLD. | |
995 | */ | |
996 | e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, rb); | |
997 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); | |
998 | ||
999 | if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD)) | |
1000 | goto out_unlock; | |
1001 | dbg_wl("schedule wear-leveling"); | |
1002 | } else | |
1003 | dbg_wl("schedule scrubbing"); | |
1004 | ||
1005 | ubi->wl_scheduled = 1; | |
1006 | spin_unlock(&ubi->wl_lock); | |
1007 | ||
33818bbb | 1008 | wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS); |
801c135c AB |
1009 | if (!wrk) { |
1010 | err = -ENOMEM; | |
1011 | goto out_cancel; | |
1012 | } | |
1013 | ||
1014 | wrk->func = &wear_leveling_worker; | |
1015 | schedule_ubi_work(ubi, wrk); | |
1016 | return err; | |
1017 | ||
1018 | out_cancel: | |
1019 | spin_lock(&ubi->wl_lock); | |
1020 | ubi->wl_scheduled = 0; | |
1021 | out_unlock: | |
1022 | spin_unlock(&ubi->wl_lock); | |
1023 | return err; | |
1024 | } | |
1025 | ||
1026 | /** | |
1027 | * erase_worker - physical eraseblock erase worker function. | |
1028 | * @ubi: UBI device description object | |
1029 | * @wl_wrk: the work object | |
1030 | * @cancel: non-zero if the worker has to free memory and exit | |
1031 | * | |
1032 | * This function erases a physical eraseblock and perform torture testing if | |
1033 | * needed. It also takes care about marking the physical eraseblock bad if | |
1034 | * needed. Returns zero in case of success and a negative error code in case of | |
1035 | * failure. | |
1036 | */ | |
1037 | static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk, | |
1038 | int cancel) | |
1039 | { | |
801c135c | 1040 | struct ubi_wl_entry *e = wl_wrk->e; |
784c1454 | 1041 | int pnum = e->pnum, err, need; |
801c135c AB |
1042 | |
1043 | if (cancel) { | |
1044 | dbg_wl("cancel erasure of PEB %d EC %d", pnum, e->ec); | |
1045 | kfree(wl_wrk); | |
06b68ba1 | 1046 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1047 | return 0; |
1048 | } | |
1049 | ||
1050 | dbg_wl("erase PEB %d EC %d", pnum, e->ec); | |
1051 | ||
1052 | err = sync_erase(ubi, e, wl_wrk->torture); | |
1053 | if (!err) { | |
1054 | /* Fine, we've erased it successfully */ | |
1055 | kfree(wl_wrk); | |
1056 | ||
1057 | spin_lock(&ubi->wl_lock); | |
1058 | ubi->abs_ec += 1; | |
5abde384 | 1059 | wl_tree_add(e, &ubi->free); |
801c135c AB |
1060 | spin_unlock(&ubi->wl_lock); |
1061 | ||
1062 | /* | |
1063 | * One more erase operation has happened, take care about protected | |
1064 | * physical eraseblocks. | |
1065 | */ | |
1066 | check_protection_over(ubi); | |
1067 | ||
1068 | /* And take care about wear-leveling */ | |
1069 | err = ensure_wear_leveling(ubi); | |
1070 | return err; | |
1071 | } | |
1072 | ||
8d2d4011 | 1073 | ubi_err("failed to erase PEB %d, error %d", pnum, err); |
801c135c | 1074 | kfree(wl_wrk); |
06b68ba1 | 1075 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c | 1076 | |
784c1454 AB |
1077 | if (err == -EINTR || err == -ENOMEM || err == -EAGAIN || |
1078 | err == -EBUSY) { | |
1079 | int err1; | |
1080 | ||
1081 | /* Re-schedule the LEB for erasure */ | |
1082 | err1 = schedule_erase(ubi, e, 0); | |
1083 | if (err1) { | |
1084 | err = err1; | |
1085 | goto out_ro; | |
1086 | } | |
1087 | return err; | |
1088 | } else if (err != -EIO) { | |
801c135c AB |
1089 | /* |
1090 | * If this is not %-EIO, we have no idea what to do. Scheduling | |
1091 | * this physical eraseblock for erasure again would cause | |
1092 | * errors again and again. Well, lets switch to RO mode. | |
1093 | */ | |
784c1454 | 1094 | goto out_ro; |
801c135c AB |
1095 | } |
1096 | ||
1097 | /* It is %-EIO, the PEB went bad */ | |
1098 | ||
1099 | if (!ubi->bad_allowed) { | |
1100 | ubi_err("bad physical eraseblock %d detected", pnum); | |
784c1454 AB |
1101 | goto out_ro; |
1102 | } | |
801c135c | 1103 | |
784c1454 AB |
1104 | spin_lock(&ubi->volumes_lock); |
1105 | need = ubi->beb_rsvd_level - ubi->beb_rsvd_pebs + 1; | |
1106 | if (need > 0) { | |
1107 | need = ubi->avail_pebs >= need ? need : ubi->avail_pebs; | |
1108 | ubi->avail_pebs -= need; | |
1109 | ubi->rsvd_pebs += need; | |
1110 | ubi->beb_rsvd_pebs += need; | |
1111 | if (need > 0) | |
1112 | ubi_msg("reserve more %d PEBs", need); | |
1113 | } | |
801c135c | 1114 | |
784c1454 | 1115 | if (ubi->beb_rsvd_pebs == 0) { |
801c135c | 1116 | spin_unlock(&ubi->volumes_lock); |
784c1454 AB |
1117 | ubi_err("no reserved physical eraseblocks"); |
1118 | goto out_ro; | |
1119 | } | |
801c135c | 1120 | |
784c1454 AB |
1121 | spin_unlock(&ubi->volumes_lock); |
1122 | ubi_msg("mark PEB %d as bad", pnum); | |
801c135c | 1123 | |
784c1454 AB |
1124 | err = ubi_io_mark_bad(ubi, pnum); |
1125 | if (err) | |
1126 | goto out_ro; | |
1127 | ||
1128 | spin_lock(&ubi->volumes_lock); | |
1129 | ubi->beb_rsvd_pebs -= 1; | |
1130 | ubi->bad_peb_count += 1; | |
1131 | ubi->good_peb_count -= 1; | |
1132 | ubi_calculate_reserved(ubi); | |
1133 | if (ubi->beb_rsvd_pebs == 0) | |
1134 | ubi_warn("last PEB from the reserved pool was used"); | |
1135 | spin_unlock(&ubi->volumes_lock); | |
1136 | ||
1137 | return err; | |
801c135c | 1138 | |
784c1454 AB |
1139 | out_ro: |
1140 | ubi_ro_mode(ubi); | |
801c135c AB |
1141 | return err; |
1142 | } | |
1143 | ||
1144 | /** | |
85c6e6e2 | 1145 | * ubi_wl_put_peb - return a PEB to the wear-leveling sub-system. |
801c135c AB |
1146 | * @ubi: UBI device description object |
1147 | * @pnum: physical eraseblock to return | |
1148 | * @torture: if this physical eraseblock has to be tortured | |
1149 | * | |
1150 | * This function is called to return physical eraseblock @pnum to the pool of | |
1151 | * free physical eraseblocks. The @torture flag has to be set if an I/O error | |
1152 | * occurred to this @pnum and it has to be tested. This function returns zero | |
43f9b25a | 1153 | * in case of success, and a negative error code in case of failure. |
801c135c AB |
1154 | */ |
1155 | int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture) | |
1156 | { | |
1157 | int err; | |
1158 | struct ubi_wl_entry *e; | |
1159 | ||
1160 | dbg_wl("PEB %d", pnum); | |
1161 | ubi_assert(pnum >= 0); | |
1162 | ubi_assert(pnum < ubi->peb_count); | |
1163 | ||
43f9b25a | 1164 | retry: |
801c135c | 1165 | spin_lock(&ubi->wl_lock); |
801c135c AB |
1166 | e = ubi->lookuptbl[pnum]; |
1167 | if (e == ubi->move_from) { | |
1168 | /* | |
1169 | * User is putting the physical eraseblock which was selected to | |
1170 | * be moved. It will be scheduled for erasure in the | |
1171 | * wear-leveling worker. | |
1172 | */ | |
43f9b25a | 1173 | dbg_wl("PEB %d is being moved, wait", pnum); |
801c135c | 1174 | spin_unlock(&ubi->wl_lock); |
43f9b25a AB |
1175 | |
1176 | /* Wait for the WL worker by taking the @ubi->move_mutex */ | |
1177 | mutex_lock(&ubi->move_mutex); | |
1178 | mutex_unlock(&ubi->move_mutex); | |
1179 | goto retry; | |
801c135c AB |
1180 | } else if (e == ubi->move_to) { |
1181 | /* | |
1182 | * User is putting the physical eraseblock which was selected | |
1183 | * as the target the data is moved to. It may happen if the EBA | |
85c6e6e2 AB |
1184 | * sub-system already re-mapped the LEB in 'ubi_eba_copy_leb()' |
1185 | * but the WL sub-system has not put the PEB to the "used" tree | |
1186 | * yet, but it is about to do this. So we just set a flag which | |
1187 | * will tell the WL worker that the PEB is not needed anymore | |
1188 | * and should be scheduled for erasure. | |
801c135c AB |
1189 | */ |
1190 | dbg_wl("PEB %d is the target of data moving", pnum); | |
1191 | ubi_assert(!ubi->move_to_put); | |
1192 | ubi->move_to_put = 1; | |
1193 | spin_unlock(&ubi->wl_lock); | |
1194 | return 0; | |
1195 | } else { | |
5abde384 AB |
1196 | if (in_wl_tree(e, &ubi->used)) { |
1197 | paranoid_check_in_wl_tree(e, &ubi->used); | |
1198 | rb_erase(&e->rb, &ubi->used); | |
1199 | } else if (in_wl_tree(e, &ubi->scrub)) { | |
1200 | paranoid_check_in_wl_tree(e, &ubi->scrub); | |
1201 | rb_erase(&e->rb, &ubi->scrub); | |
43f9b25a AB |
1202 | } else { |
1203 | err = prot_tree_del(ubi, e->pnum); | |
1204 | if (err) { | |
1205 | ubi_err("PEB %d not found", pnum); | |
1206 | ubi_ro_mode(ubi); | |
1207 | spin_unlock(&ubi->wl_lock); | |
1208 | return err; | |
1209 | } | |
1210 | } | |
801c135c AB |
1211 | } |
1212 | spin_unlock(&ubi->wl_lock); | |
1213 | ||
1214 | err = schedule_erase(ubi, e, torture); | |
1215 | if (err) { | |
1216 | spin_lock(&ubi->wl_lock); | |
5abde384 | 1217 | wl_tree_add(e, &ubi->used); |
801c135c AB |
1218 | spin_unlock(&ubi->wl_lock); |
1219 | } | |
1220 | ||
1221 | return err; | |
1222 | } | |
1223 | ||
1224 | /** | |
1225 | * ubi_wl_scrub_peb - schedule a physical eraseblock for scrubbing. | |
1226 | * @ubi: UBI device description object | |
1227 | * @pnum: the physical eraseblock to schedule | |
1228 | * | |
1229 | * If a bit-flip in a physical eraseblock is detected, this physical eraseblock | |
1230 | * needs scrubbing. This function schedules a physical eraseblock for | |
1231 | * scrubbing which is done in background. This function returns zero in case of | |
1232 | * success and a negative error code in case of failure. | |
1233 | */ | |
1234 | int ubi_wl_scrub_peb(struct ubi_device *ubi, int pnum) | |
1235 | { | |
1236 | struct ubi_wl_entry *e; | |
1237 | ||
8c1e6ee1 | 1238 | dbg_msg("schedule PEB %d for scrubbing", pnum); |
801c135c AB |
1239 | |
1240 | retry: | |
1241 | spin_lock(&ubi->wl_lock); | |
1242 | e = ubi->lookuptbl[pnum]; | |
1243 | if (e == ubi->move_from || in_wl_tree(e, &ubi->scrub)) { | |
1244 | spin_unlock(&ubi->wl_lock); | |
1245 | return 0; | |
1246 | } | |
1247 | ||
1248 | if (e == ubi->move_to) { | |
1249 | /* | |
1250 | * This physical eraseblock was used to move data to. The data | |
1251 | * was moved but the PEB was not yet inserted to the proper | |
1252 | * tree. We should just wait a little and let the WL worker | |
1253 | * proceed. | |
1254 | */ | |
1255 | spin_unlock(&ubi->wl_lock); | |
1256 | dbg_wl("the PEB %d is not in proper tree, retry", pnum); | |
1257 | yield(); | |
1258 | goto retry; | |
1259 | } | |
1260 | ||
5abde384 AB |
1261 | if (in_wl_tree(e, &ubi->used)) { |
1262 | paranoid_check_in_wl_tree(e, &ubi->used); | |
1263 | rb_erase(&e->rb, &ubi->used); | |
43f9b25a AB |
1264 | } else { |
1265 | int err; | |
1266 | ||
1267 | err = prot_tree_del(ubi, e->pnum); | |
1268 | if (err) { | |
1269 | ubi_err("PEB %d not found", pnum); | |
1270 | ubi_ro_mode(ubi); | |
1271 | spin_unlock(&ubi->wl_lock); | |
1272 | return err; | |
1273 | } | |
1274 | } | |
801c135c | 1275 | |
5abde384 | 1276 | wl_tree_add(e, &ubi->scrub); |
801c135c AB |
1277 | spin_unlock(&ubi->wl_lock); |
1278 | ||
1279 | /* | |
1280 | * Technically scrubbing is the same as wear-leveling, so it is done | |
1281 | * by the WL worker. | |
1282 | */ | |
1283 | return ensure_wear_leveling(ubi); | |
1284 | } | |
1285 | ||
1286 | /** | |
1287 | * ubi_wl_flush - flush all pending works. | |
1288 | * @ubi: UBI device description object | |
1289 | * | |
1290 | * This function returns zero in case of success and a negative error code in | |
1291 | * case of failure. | |
1292 | */ | |
1293 | int ubi_wl_flush(struct ubi_device *ubi) | |
1294 | { | |
593dd33c | 1295 | int err; |
801c135c AB |
1296 | |
1297 | /* | |
1298 | * Erase while the pending works queue is not empty, but not more then | |
1299 | * the number of currently pending works. | |
1300 | */ | |
593dd33c AB |
1301 | dbg_wl("flush (%d pending works)", ubi->works_count); |
1302 | while (ubi->works_count) { | |
1303 | err = do_work(ubi); | |
1304 | if (err) | |
1305 | return err; | |
1306 | } | |
1307 | ||
1308 | /* | |
1309 | * Make sure all the works which have been done in parallel are | |
1310 | * finished. | |
1311 | */ | |
1312 | down_write(&ubi->work_sem); | |
1313 | up_write(&ubi->work_sem); | |
1314 | ||
1315 | /* | |
1316 | * And in case last was the WL worker and it cancelled the LEB | |
1317 | * movement, flush again. | |
1318 | */ | |
1319 | while (ubi->works_count) { | |
1320 | dbg_wl("flush more (%d pending works)", ubi->works_count); | |
801c135c AB |
1321 | err = do_work(ubi); |
1322 | if (err) | |
1323 | return err; | |
1324 | } | |
1325 | ||
1326 | return 0; | |
1327 | } | |
1328 | ||
1329 | /** | |
1330 | * tree_destroy - destroy an RB-tree. | |
1331 | * @root: the root of the tree to destroy | |
1332 | */ | |
1333 | static void tree_destroy(struct rb_root *root) | |
1334 | { | |
1335 | struct rb_node *rb; | |
1336 | struct ubi_wl_entry *e; | |
1337 | ||
1338 | rb = root->rb_node; | |
1339 | while (rb) { | |
1340 | if (rb->rb_left) | |
1341 | rb = rb->rb_left; | |
1342 | else if (rb->rb_right) | |
1343 | rb = rb->rb_right; | |
1344 | else { | |
1345 | e = rb_entry(rb, struct ubi_wl_entry, rb); | |
1346 | ||
1347 | rb = rb_parent(rb); | |
1348 | if (rb) { | |
1349 | if (rb->rb_left == &e->rb) | |
1350 | rb->rb_left = NULL; | |
1351 | else | |
1352 | rb->rb_right = NULL; | |
1353 | } | |
1354 | ||
06b68ba1 | 1355 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1356 | } |
1357 | } | |
1358 | } | |
1359 | ||
1360 | /** | |
1361 | * ubi_thread - UBI background thread. | |
1362 | * @u: the UBI device description object pointer | |
1363 | */ | |
cdfa788a | 1364 | int ubi_thread(void *u) |
801c135c AB |
1365 | { |
1366 | int failures = 0; | |
1367 | struct ubi_device *ubi = u; | |
1368 | ||
1369 | ubi_msg("background thread \"%s\" started, PID %d", | |
ba25f9dc | 1370 | ubi->bgt_name, task_pid_nr(current)); |
801c135c | 1371 | |
83144186 | 1372 | set_freezable(); |
801c135c AB |
1373 | for (;;) { |
1374 | int err; | |
1375 | ||
1376 | if (kthread_should_stop()) | |
cadb40cc | 1377 | break; |
801c135c AB |
1378 | |
1379 | if (try_to_freeze()) | |
1380 | continue; | |
1381 | ||
1382 | spin_lock(&ubi->wl_lock); | |
1383 | if (list_empty(&ubi->works) || ubi->ro_mode || | |
1384 | !ubi->thread_enabled) { | |
1385 | set_current_state(TASK_INTERRUPTIBLE); | |
1386 | spin_unlock(&ubi->wl_lock); | |
1387 | schedule(); | |
1388 | continue; | |
1389 | } | |
1390 | spin_unlock(&ubi->wl_lock); | |
1391 | ||
1392 | err = do_work(ubi); | |
1393 | if (err) { | |
1394 | ubi_err("%s: work failed with error code %d", | |
1395 | ubi->bgt_name, err); | |
1396 | if (failures++ > WL_MAX_FAILURES) { | |
1397 | /* | |
1398 | * Too many failures, disable the thread and | |
1399 | * switch to read-only mode. | |
1400 | */ | |
1401 | ubi_msg("%s: %d consecutive failures", | |
1402 | ubi->bgt_name, WL_MAX_FAILURES); | |
1403 | ubi_ro_mode(ubi); | |
1404 | break; | |
1405 | } | |
1406 | } else | |
1407 | failures = 0; | |
1408 | ||
1409 | cond_resched(); | |
1410 | } | |
1411 | ||
801c135c AB |
1412 | dbg_wl("background thread \"%s\" is killed", ubi->bgt_name); |
1413 | return 0; | |
1414 | } | |
1415 | ||
1416 | /** | |
1417 | * cancel_pending - cancel all pending works. | |
1418 | * @ubi: UBI device description object | |
1419 | */ | |
1420 | static void cancel_pending(struct ubi_device *ubi) | |
1421 | { | |
1422 | while (!list_empty(&ubi->works)) { | |
1423 | struct ubi_work *wrk; | |
1424 | ||
1425 | wrk = list_entry(ubi->works.next, struct ubi_work, list); | |
1426 | list_del(&wrk->list); | |
1427 | wrk->func(ubi, wrk, 1); | |
1428 | ubi->works_count -= 1; | |
1429 | ubi_assert(ubi->works_count >= 0); | |
1430 | } | |
1431 | } | |
1432 | ||
1433 | /** | |
85c6e6e2 | 1434 | * ubi_wl_init_scan - initialize the WL sub-system using scanning information. |
801c135c AB |
1435 | * @ubi: UBI device description object |
1436 | * @si: scanning information | |
1437 | * | |
1438 | * This function returns zero in case of success, and a negative error code in | |
1439 | * case of failure. | |
1440 | */ | |
1441 | int ubi_wl_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si) | |
1442 | { | |
1443 | int err; | |
1444 | struct rb_node *rb1, *rb2; | |
1445 | struct ubi_scan_volume *sv; | |
1446 | struct ubi_scan_leb *seb, *tmp; | |
1447 | struct ubi_wl_entry *e; | |
1448 | ||
1449 | ||
1450 | ubi->used = ubi->free = ubi->scrub = RB_ROOT; | |
1451 | ubi->prot.pnum = ubi->prot.aec = RB_ROOT; | |
1452 | spin_lock_init(&ubi->wl_lock); | |
43f9b25a | 1453 | mutex_init(&ubi->move_mutex); |
593dd33c | 1454 | init_rwsem(&ubi->work_sem); |
801c135c AB |
1455 | ubi->max_ec = si->max_ec; |
1456 | INIT_LIST_HEAD(&ubi->works); | |
1457 | ||
1458 | sprintf(ubi->bgt_name, UBI_BGT_NAME_PATTERN, ubi->ubi_num); | |
1459 | ||
801c135c AB |
1460 | err = -ENOMEM; |
1461 | ubi->lookuptbl = kzalloc(ubi->peb_count * sizeof(void *), GFP_KERNEL); | |
1462 | if (!ubi->lookuptbl) | |
cdfa788a | 1463 | return err; |
801c135c AB |
1464 | |
1465 | list_for_each_entry_safe(seb, tmp, &si->erase, u.list) { | |
1466 | cond_resched(); | |
1467 | ||
06b68ba1 | 1468 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1469 | if (!e) |
1470 | goto out_free; | |
1471 | ||
1472 | e->pnum = seb->pnum; | |
1473 | e->ec = seb->ec; | |
1474 | ubi->lookuptbl[e->pnum] = e; | |
1475 | if (schedule_erase(ubi, e, 0)) { | |
06b68ba1 | 1476 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1477 | goto out_free; |
1478 | } | |
1479 | } | |
1480 | ||
1481 | list_for_each_entry(seb, &si->free, u.list) { | |
1482 | cond_resched(); | |
1483 | ||
06b68ba1 | 1484 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1485 | if (!e) |
1486 | goto out_free; | |
1487 | ||
1488 | e->pnum = seb->pnum; | |
1489 | e->ec = seb->ec; | |
1490 | ubi_assert(e->ec >= 0); | |
5abde384 | 1491 | wl_tree_add(e, &ubi->free); |
801c135c AB |
1492 | ubi->lookuptbl[e->pnum] = e; |
1493 | } | |
1494 | ||
1495 | list_for_each_entry(seb, &si->corr, u.list) { | |
1496 | cond_resched(); | |
1497 | ||
06b68ba1 | 1498 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1499 | if (!e) |
1500 | goto out_free; | |
1501 | ||
1502 | e->pnum = seb->pnum; | |
1503 | e->ec = seb->ec; | |
1504 | ubi->lookuptbl[e->pnum] = e; | |
1505 | if (schedule_erase(ubi, e, 0)) { | |
06b68ba1 | 1506 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1507 | goto out_free; |
1508 | } | |
1509 | } | |
1510 | ||
1511 | ubi_rb_for_each_entry(rb1, sv, &si->volumes, rb) { | |
1512 | ubi_rb_for_each_entry(rb2, seb, &sv->root, u.rb) { | |
1513 | cond_resched(); | |
1514 | ||
06b68ba1 | 1515 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1516 | if (!e) |
1517 | goto out_free; | |
1518 | ||
1519 | e->pnum = seb->pnum; | |
1520 | e->ec = seb->ec; | |
1521 | ubi->lookuptbl[e->pnum] = e; | |
1522 | if (!seb->scrub) { | |
1523 | dbg_wl("add PEB %d EC %d to the used tree", | |
1524 | e->pnum, e->ec); | |
5abde384 | 1525 | wl_tree_add(e, &ubi->used); |
801c135c AB |
1526 | } else { |
1527 | dbg_wl("add PEB %d EC %d to the scrub tree", | |
1528 | e->pnum, e->ec); | |
5abde384 | 1529 | wl_tree_add(e, &ubi->scrub); |
801c135c AB |
1530 | } |
1531 | } | |
1532 | } | |
1533 | ||
5abde384 | 1534 | if (ubi->avail_pebs < WL_RESERVED_PEBS) { |
801c135c AB |
1535 | ubi_err("no enough physical eraseblocks (%d, need %d)", |
1536 | ubi->avail_pebs, WL_RESERVED_PEBS); | |
1537 | goto out_free; | |
1538 | } | |
1539 | ubi->avail_pebs -= WL_RESERVED_PEBS; | |
1540 | ubi->rsvd_pebs += WL_RESERVED_PEBS; | |
1541 | ||
1542 | /* Schedule wear-leveling if needed */ | |
1543 | err = ensure_wear_leveling(ubi); | |
1544 | if (err) | |
1545 | goto out_free; | |
1546 | ||
1547 | return 0; | |
1548 | ||
1549 | out_free: | |
1550 | cancel_pending(ubi); | |
1551 | tree_destroy(&ubi->used); | |
1552 | tree_destroy(&ubi->free); | |
1553 | tree_destroy(&ubi->scrub); | |
1554 | kfree(ubi->lookuptbl); | |
801c135c AB |
1555 | return err; |
1556 | } | |
1557 | ||
1558 | /** | |
1559 | * protection_trees_destroy - destroy the protection RB-trees. | |
1560 | * @ubi: UBI device description object | |
1561 | */ | |
1562 | static void protection_trees_destroy(struct ubi_device *ubi) | |
1563 | { | |
1564 | struct rb_node *rb; | |
1565 | struct ubi_wl_prot_entry *pe; | |
1566 | ||
1567 | rb = ubi->prot.aec.rb_node; | |
1568 | while (rb) { | |
1569 | if (rb->rb_left) | |
1570 | rb = rb->rb_left; | |
1571 | else if (rb->rb_right) | |
1572 | rb = rb->rb_right; | |
1573 | else { | |
1574 | pe = rb_entry(rb, struct ubi_wl_prot_entry, rb_aec); | |
1575 | ||
1576 | rb = rb_parent(rb); | |
1577 | if (rb) { | |
1578 | if (rb->rb_left == &pe->rb_aec) | |
1579 | rb->rb_left = NULL; | |
1580 | else | |
1581 | rb->rb_right = NULL; | |
1582 | } | |
1583 | ||
06b68ba1 | 1584 | kmem_cache_free(ubi_wl_entry_slab, pe->e); |
801c135c AB |
1585 | kfree(pe); |
1586 | } | |
1587 | } | |
1588 | } | |
1589 | ||
1590 | /** | |
85c6e6e2 | 1591 | * ubi_wl_close - close the wear-leveling sub-system. |
801c135c AB |
1592 | * @ubi: UBI device description object |
1593 | */ | |
1594 | void ubi_wl_close(struct ubi_device *ubi) | |
1595 | { | |
85c6e6e2 | 1596 | dbg_wl("close the WL sub-system"); |
801c135c AB |
1597 | cancel_pending(ubi); |
1598 | protection_trees_destroy(ubi); | |
1599 | tree_destroy(&ubi->used); | |
1600 | tree_destroy(&ubi->free); | |
1601 | tree_destroy(&ubi->scrub); | |
1602 | kfree(ubi->lookuptbl); | |
801c135c AB |
1603 | } |
1604 | ||
1605 | #ifdef CONFIG_MTD_UBI_DEBUG_PARANOID | |
1606 | ||
1607 | /** | |
1608 | * paranoid_check_ec - make sure that the erase counter of a physical eraseblock | |
1609 | * is correct. | |
1610 | * @ubi: UBI device description object | |
1611 | * @pnum: the physical eraseblock number to check | |
1612 | * @ec: the erase counter to check | |
1613 | * | |
1614 | * This function returns zero if the erase counter of physical eraseblock @pnum | |
1615 | * is equivalent to @ec, %1 if not, and a negative error code if an error | |
1616 | * occurred. | |
1617 | */ | |
e88d6e10 | 1618 | static int paranoid_check_ec(struct ubi_device *ubi, int pnum, int ec) |
801c135c AB |
1619 | { |
1620 | int err; | |
1621 | long long read_ec; | |
1622 | struct ubi_ec_hdr *ec_hdr; | |
1623 | ||
33818bbb | 1624 | ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_NOFS); |
801c135c AB |
1625 | if (!ec_hdr) |
1626 | return -ENOMEM; | |
1627 | ||
1628 | err = ubi_io_read_ec_hdr(ubi, pnum, ec_hdr, 0); | |
1629 | if (err && err != UBI_IO_BITFLIPS) { | |
1630 | /* The header does not have to exist */ | |
1631 | err = 0; | |
1632 | goto out_free; | |
1633 | } | |
1634 | ||
3261ebd7 | 1635 | read_ec = be64_to_cpu(ec_hdr->ec); |
801c135c AB |
1636 | if (ec != read_ec) { |
1637 | ubi_err("paranoid check failed for PEB %d", pnum); | |
1638 | ubi_err("read EC is %lld, should be %d", read_ec, ec); | |
1639 | ubi_dbg_dump_stack(); | |
1640 | err = 1; | |
1641 | } else | |
1642 | err = 0; | |
1643 | ||
1644 | out_free: | |
1645 | kfree(ec_hdr); | |
1646 | return err; | |
1647 | } | |
1648 | ||
1649 | /** | |
1650 | * paranoid_check_in_wl_tree - make sure that a wear-leveling entry is present | |
1651 | * in a WL RB-tree. | |
1652 | * @e: the wear-leveling entry to check | |
1653 | * @root: the root of the tree | |
1654 | * | |
1655 | * This function returns zero if @e is in the @root RB-tree and %1 if it | |
1656 | * is not. | |
1657 | */ | |
1658 | static int paranoid_check_in_wl_tree(struct ubi_wl_entry *e, | |
1659 | struct rb_root *root) | |
1660 | { | |
1661 | if (in_wl_tree(e, root)) | |
1662 | return 0; | |
1663 | ||
1664 | ubi_err("paranoid check failed for PEB %d, EC %d, RB-tree %p ", | |
1665 | e->pnum, e->ec, root); | |
1666 | ubi_dbg_dump_stack(); | |
1667 | return 1; | |
1668 | } | |
1669 | ||
1670 | #endif /* CONFIG_MTD_UBI_DEBUG_PARANOID */ |