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) { | |
9c9ec147 AB |
478 | case UBI_LONGTERM: |
479 | /* | |
480 | * For long term data we pick a physical eraseblock with high | |
481 | * erase counter. But the highest erase counter we can pick is | |
482 | * bounded by the the lowest erase counter plus | |
483 | * %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 medium | |
491 | * erase counter. But we by no means can pick a physical | |
492 | * eraseblock with erase counter greater or equivalent than the | |
493 | * lowest erase counter plus %WL_FREE_MAX_DIFF. | |
494 | */ | |
495 | first = rb_entry(rb_first(&ubi->free), struct ubi_wl_entry, rb); | |
496 | last = rb_entry(rb_last(&ubi->free), struct ubi_wl_entry, rb); | |
801c135c | 497 | |
9c9ec147 AB |
498 | if (last->ec - first->ec < WL_FREE_MAX_DIFF) |
499 | e = rb_entry(ubi->free.rb_node, | |
500 | struct ubi_wl_entry, rb); | |
501 | else { | |
502 | medium_ec = (first->ec + WL_FREE_MAX_DIFF)/2; | |
503 | e = find_wl_entry(&ubi->free, medium_ec); | |
504 | } | |
505 | protect = U_PROTECTION; | |
506 | break; | |
507 | case UBI_SHORTTERM: | |
508 | /* | |
509 | * For short term data we pick a physical eraseblock with the | |
510 | * lowest erase counter as we expect it will be erased soon. | |
511 | */ | |
512 | e = rb_entry(rb_first(&ubi->free), struct ubi_wl_entry, rb); | |
513 | protect = ST_PROTECTION; | |
514 | break; | |
515 | default: | |
516 | protect = 0; | |
517 | e = NULL; | |
518 | BUG(); | |
801c135c AB |
519 | } |
520 | ||
521 | /* | |
522 | * Move the physical eraseblock to the protection trees where it will | |
523 | * be protected from being moved for some time. | |
524 | */ | |
5abde384 AB |
525 | paranoid_check_in_wl_tree(e, &ubi->free); |
526 | rb_erase(&e->rb, &ubi->free); | |
801c135c AB |
527 | prot_tree_add(ubi, e, pe, protect); |
528 | ||
529 | dbg_wl("PEB %d EC %d, protection %d", e->pnum, e->ec, protect); | |
530 | spin_unlock(&ubi->wl_lock); | |
531 | ||
532 | return e->pnum; | |
533 | } | |
534 | ||
535 | /** | |
536 | * prot_tree_del - remove a physical eraseblock from the protection trees | |
537 | * @ubi: UBI device description object | |
538 | * @pnum: the physical eraseblock to remove | |
43f9b25a AB |
539 | * |
540 | * This function returns PEB @pnum from the protection trees and returns zero | |
541 | * in case of success and %-ENODEV if the PEB was not found in the protection | |
542 | * trees. | |
801c135c | 543 | */ |
43f9b25a | 544 | static int prot_tree_del(struct ubi_device *ubi, int pnum) |
801c135c AB |
545 | { |
546 | struct rb_node *p; | |
547 | struct ubi_wl_prot_entry *pe = NULL; | |
548 | ||
549 | p = ubi->prot.pnum.rb_node; | |
550 | while (p) { | |
551 | ||
552 | pe = rb_entry(p, struct ubi_wl_prot_entry, rb_pnum); | |
553 | ||
554 | if (pnum == pe->e->pnum) | |
43f9b25a | 555 | goto found; |
801c135c AB |
556 | |
557 | if (pnum < pe->e->pnum) | |
558 | p = p->rb_left; | |
559 | else | |
560 | p = p->rb_right; | |
561 | } | |
562 | ||
43f9b25a AB |
563 | return -ENODEV; |
564 | ||
565 | found: | |
801c135c AB |
566 | ubi_assert(pe->e->pnum == pnum); |
567 | rb_erase(&pe->rb_aec, &ubi->prot.aec); | |
568 | rb_erase(&pe->rb_pnum, &ubi->prot.pnum); | |
569 | kfree(pe); | |
43f9b25a | 570 | return 0; |
801c135c AB |
571 | } |
572 | ||
573 | /** | |
574 | * sync_erase - synchronously erase a physical eraseblock. | |
575 | * @ubi: UBI device description object | |
576 | * @e: the the physical eraseblock to erase | |
577 | * @torture: if the physical eraseblock has to be tortured | |
578 | * | |
579 | * This function returns zero in case of success and a negative error code in | |
580 | * case of failure. | |
581 | */ | |
9c9ec147 AB |
582 | static int sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e, |
583 | int torture) | |
801c135c AB |
584 | { |
585 | int err; | |
586 | struct ubi_ec_hdr *ec_hdr; | |
587 | unsigned long long ec = e->ec; | |
588 | ||
589 | dbg_wl("erase PEB %d, old EC %llu", e->pnum, ec); | |
590 | ||
591 | err = paranoid_check_ec(ubi, e->pnum, e->ec); | |
592 | if (err > 0) | |
593 | return -EINVAL; | |
594 | ||
33818bbb | 595 | ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_NOFS); |
801c135c AB |
596 | if (!ec_hdr) |
597 | return -ENOMEM; | |
598 | ||
599 | err = ubi_io_sync_erase(ubi, e->pnum, torture); | |
600 | if (err < 0) | |
601 | goto out_free; | |
602 | ||
603 | ec += err; | |
604 | if (ec > UBI_MAX_ERASECOUNTER) { | |
605 | /* | |
606 | * Erase counter overflow. Upgrade UBI and use 64-bit | |
607 | * erase counters internally. | |
608 | */ | |
609 | ubi_err("erase counter overflow at PEB %d, EC %llu", | |
610 | e->pnum, ec); | |
611 | err = -EINVAL; | |
612 | goto out_free; | |
613 | } | |
614 | ||
615 | dbg_wl("erased PEB %d, new EC %llu", e->pnum, ec); | |
616 | ||
3261ebd7 | 617 | ec_hdr->ec = cpu_to_be64(ec); |
801c135c AB |
618 | |
619 | err = ubi_io_write_ec_hdr(ubi, e->pnum, ec_hdr); | |
620 | if (err) | |
621 | goto out_free; | |
622 | ||
623 | e->ec = ec; | |
624 | spin_lock(&ubi->wl_lock); | |
625 | if (e->ec > ubi->max_ec) | |
626 | ubi->max_ec = e->ec; | |
627 | spin_unlock(&ubi->wl_lock); | |
628 | ||
629 | out_free: | |
630 | kfree(ec_hdr); | |
631 | return err; | |
632 | } | |
633 | ||
634 | /** | |
ebaaf1af | 635 | * check_protection_over - check if it is time to stop protecting some PEBs. |
801c135c AB |
636 | * @ubi: UBI device description object |
637 | * | |
638 | * This function is called after each erase operation, when the absolute erase | |
639 | * counter is incremented, to check if some physical eraseblock have not to be | |
640 | * protected any longer. These physical eraseblocks are moved from the | |
641 | * protection trees to the used tree. | |
642 | */ | |
643 | static void check_protection_over(struct ubi_device *ubi) | |
644 | { | |
645 | struct ubi_wl_prot_entry *pe; | |
646 | ||
647 | /* | |
648 | * There may be several protected physical eraseblock to remove, | |
649 | * process them all. | |
650 | */ | |
651 | while (1) { | |
652 | spin_lock(&ubi->wl_lock); | |
5abde384 | 653 | if (!ubi->prot.aec.rb_node) { |
801c135c AB |
654 | spin_unlock(&ubi->wl_lock); |
655 | break; | |
656 | } | |
657 | ||
658 | pe = rb_entry(rb_first(&ubi->prot.aec), | |
659 | struct ubi_wl_prot_entry, rb_aec); | |
660 | ||
661 | if (pe->abs_ec > ubi->abs_ec) { | |
662 | spin_unlock(&ubi->wl_lock); | |
663 | break; | |
664 | } | |
665 | ||
666 | dbg_wl("PEB %d protection over, abs_ec %llu, PEB abs_ec %llu", | |
667 | pe->e->pnum, ubi->abs_ec, pe->abs_ec); | |
668 | rb_erase(&pe->rb_aec, &ubi->prot.aec); | |
669 | rb_erase(&pe->rb_pnum, &ubi->prot.pnum); | |
5abde384 | 670 | wl_tree_add(pe->e, &ubi->used); |
801c135c AB |
671 | spin_unlock(&ubi->wl_lock); |
672 | ||
673 | kfree(pe); | |
674 | cond_resched(); | |
675 | } | |
676 | } | |
677 | ||
678 | /** | |
679 | * schedule_ubi_work - schedule a work. | |
680 | * @ubi: UBI device description object | |
681 | * @wrk: the work to schedule | |
682 | * | |
683 | * This function enqueues a work defined by @wrk to the tail of the pending | |
684 | * works list. | |
685 | */ | |
686 | static void schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk) | |
687 | { | |
688 | spin_lock(&ubi->wl_lock); | |
689 | list_add_tail(&wrk->list, &ubi->works); | |
690 | ubi_assert(ubi->works_count >= 0); | |
691 | ubi->works_count += 1; | |
692 | if (ubi->thread_enabled) | |
693 | wake_up_process(ubi->bgt_thread); | |
694 | spin_unlock(&ubi->wl_lock); | |
695 | } | |
696 | ||
697 | static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk, | |
698 | int cancel); | |
699 | ||
700 | /** | |
701 | * schedule_erase - schedule an erase work. | |
702 | * @ubi: UBI device description object | |
703 | * @e: the WL entry of the physical eraseblock to erase | |
704 | * @torture: if the physical eraseblock has to be tortured | |
705 | * | |
706 | * This function returns zero in case of success and a %-ENOMEM in case of | |
707 | * failure. | |
708 | */ | |
709 | static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e, | |
710 | int torture) | |
711 | { | |
712 | struct ubi_work *wl_wrk; | |
713 | ||
714 | dbg_wl("schedule erasure of PEB %d, EC %d, torture %d", | |
715 | e->pnum, e->ec, torture); | |
716 | ||
33818bbb | 717 | wl_wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS); |
801c135c AB |
718 | if (!wl_wrk) |
719 | return -ENOMEM; | |
720 | ||
721 | wl_wrk->func = &erase_worker; | |
722 | wl_wrk->e = e; | |
723 | wl_wrk->torture = torture; | |
724 | ||
725 | schedule_ubi_work(ubi, wl_wrk); | |
726 | return 0; | |
727 | } | |
728 | ||
729 | /** | |
730 | * wear_leveling_worker - wear-leveling worker function. | |
731 | * @ubi: UBI device description object | |
732 | * @wrk: the work object | |
733 | * @cancel: non-zero if the worker has to free memory and exit | |
734 | * | |
735 | * This function copies a more worn out physical eraseblock to a less worn out | |
736 | * one. Returns zero in case of success and a negative error code in case of | |
737 | * failure. | |
738 | */ | |
739 | static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk, | |
740 | int cancel) | |
741 | { | |
43f9b25a | 742 | int err, put = 0, scrubbing = 0, protect = 0; |
c18a8418 | 743 | struct ubi_wl_prot_entry *uninitialized_var(pe); |
801c135c AB |
744 | struct ubi_wl_entry *e1, *e2; |
745 | struct ubi_vid_hdr *vid_hdr; | |
746 | ||
747 | kfree(wrk); | |
748 | ||
749 | if (cancel) | |
750 | return 0; | |
751 | ||
33818bbb | 752 | vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS); |
801c135c AB |
753 | if (!vid_hdr) |
754 | return -ENOMEM; | |
755 | ||
43f9b25a | 756 | mutex_lock(&ubi->move_mutex); |
801c135c | 757 | spin_lock(&ubi->wl_lock); |
43f9b25a AB |
758 | ubi_assert(!ubi->move_from && !ubi->move_to); |
759 | ubi_assert(!ubi->move_to_put); | |
801c135c | 760 | |
43f9b25a | 761 | if (!ubi->free.rb_node || |
5abde384 | 762 | (!ubi->used.rb_node && !ubi->scrub.rb_node)) { |
801c135c | 763 | /* |
43f9b25a AB |
764 | * No free physical eraseblocks? Well, they must be waiting in |
765 | * the queue to be erased. Cancel movement - it will be | |
766 | * triggered again when a free physical eraseblock appears. | |
801c135c AB |
767 | * |
768 | * No used physical eraseblocks? They must be temporarily | |
769 | * protected from being moved. They will be moved to the | |
770 | * @ubi->used tree later and the wear-leveling will be | |
771 | * triggered again. | |
772 | */ | |
773 | dbg_wl("cancel WL, a list is empty: free %d, used %d", | |
5abde384 | 774 | !ubi->free.rb_node, !ubi->used.rb_node); |
43f9b25a | 775 | goto out_cancel; |
801c135c AB |
776 | } |
777 | ||
5abde384 | 778 | if (!ubi->scrub.rb_node) { |
801c135c AB |
779 | /* |
780 | * Now pick the least worn-out used physical eraseblock and a | |
781 | * highly worn-out free physical eraseblock. If the erase | |
782 | * counters differ much enough, start wear-leveling. | |
783 | */ | |
784 | e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, rb); | |
785 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); | |
786 | ||
787 | if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD)) { | |
788 | dbg_wl("no WL needed: min used EC %d, max free EC %d", | |
789 | e1->ec, e2->ec); | |
43f9b25a | 790 | goto out_cancel; |
801c135c | 791 | } |
5abde384 AB |
792 | paranoid_check_in_wl_tree(e1, &ubi->used); |
793 | rb_erase(&e1->rb, &ubi->used); | |
801c135c AB |
794 | dbg_wl("move PEB %d EC %d to PEB %d EC %d", |
795 | e1->pnum, e1->ec, e2->pnum, e2->ec); | |
796 | } else { | |
43f9b25a AB |
797 | /* Perform scrubbing */ |
798 | scrubbing = 1; | |
801c135c AB |
799 | e1 = rb_entry(rb_first(&ubi->scrub), struct ubi_wl_entry, rb); |
800 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); | |
5abde384 | 801 | paranoid_check_in_wl_tree(e1, &ubi->scrub); |
d2c46855 | 802 | rb_erase(&e1->rb, &ubi->scrub); |
801c135c AB |
803 | dbg_wl("scrub PEB %d to PEB %d", e1->pnum, e2->pnum); |
804 | } | |
805 | ||
5abde384 AB |
806 | paranoid_check_in_wl_tree(e2, &ubi->free); |
807 | rb_erase(&e2->rb, &ubi->free); | |
801c135c AB |
808 | ubi->move_from = e1; |
809 | ubi->move_to = e2; | |
810 | spin_unlock(&ubi->wl_lock); | |
811 | ||
812 | /* | |
813 | * Now we are going to copy physical eraseblock @e1->pnum to @e2->pnum. | |
814 | * We so far do not know which logical eraseblock our physical | |
815 | * eraseblock (@e1) belongs to. We have to read the volume identifier | |
816 | * header first. | |
43f9b25a AB |
817 | * |
818 | * Note, we are protected from this PEB being unmapped and erased. The | |
819 | * 'ubi_wl_put_peb()' would wait for moving to be finished if the PEB | |
820 | * which is being moved was unmapped. | |
801c135c AB |
821 | */ |
822 | ||
823 | err = ubi_io_read_vid_hdr(ubi, e1->pnum, vid_hdr, 0); | |
824 | if (err && err != UBI_IO_BITFLIPS) { | |
825 | if (err == UBI_IO_PEB_FREE) { | |
826 | /* | |
827 | * We are trying to move PEB without a VID header. UBI | |
828 | * always write VID headers shortly after the PEB was | |
829 | * given, so we have a situation when it did not have | |
830 | * chance to write it down because it was preempted. | |
831 | * Just re-schedule the work, so that next time it will | |
832 | * likely have the VID header in place. | |
833 | */ | |
834 | dbg_wl("PEB %d has no VID header", e1->pnum); | |
43f9b25a | 835 | goto out_not_moved; |
801c135c | 836 | } |
43f9b25a AB |
837 | |
838 | ubi_err("error %d while reading VID header from PEB %d", | |
839 | err, e1->pnum); | |
840 | if (err > 0) | |
841 | err = -EIO; | |
842 | goto out_error; | |
801c135c AB |
843 | } |
844 | ||
845 | err = ubi_eba_copy_leb(ubi, e1->pnum, e2->pnum, vid_hdr); | |
846 | if (err) { | |
43f9b25a AB |
847 | |
848 | if (err < 0) | |
849 | goto out_error; | |
850 | if (err == 1) | |
851 | goto out_not_moved; | |
852 | ||
853 | /* | |
854 | * For some reason the LEB was not moved - it might be because | |
855 | * the volume is being deleted. We should prevent this PEB from | |
856 | * being selected for wear-levelling movement for some "time", | |
857 | * so put it to the protection tree. | |
858 | */ | |
859 | ||
860 | dbg_wl("cancelled moving PEB %d", e1->pnum); | |
861 | pe = kmalloc(sizeof(struct ubi_wl_prot_entry), GFP_NOFS); | |
862 | if (!pe) { | |
863 | err = -ENOMEM; | |
864 | goto out_error; | |
865 | } | |
866 | ||
867 | protect = 1; | |
801c135c AB |
868 | } |
869 | ||
870 | ubi_free_vid_hdr(ubi, vid_hdr); | |
8c1e6ee1 AB |
871 | if (scrubbing && !protect) |
872 | ubi_msg("scrubbed PEB %d, data moved to PEB %d", | |
873 | e1->pnum, e2->pnum); | |
874 | ||
801c135c | 875 | spin_lock(&ubi->wl_lock); |
43f9b25a AB |
876 | if (protect) |
877 | prot_tree_add(ubi, e1, pe, protect); | |
801c135c | 878 | if (!ubi->move_to_put) |
5abde384 | 879 | wl_tree_add(e2, &ubi->used); |
801c135c AB |
880 | else |
881 | put = 1; | |
882 | ubi->move_from = ubi->move_to = NULL; | |
43f9b25a | 883 | ubi->move_to_put = ubi->wl_scheduled = 0; |
801c135c AB |
884 | spin_unlock(&ubi->wl_lock); |
885 | ||
886 | if (put) { | |
887 | /* | |
888 | * Well, the target PEB was put meanwhile, schedule it for | |
889 | * erasure. | |
890 | */ | |
891 | dbg_wl("PEB %d was put meanwhile, erase", e2->pnum); | |
892 | err = schedule_erase(ubi, e2, 0); | |
43f9b25a AB |
893 | if (err) |
894 | goto out_error; | |
801c135c AB |
895 | } |
896 | ||
43f9b25a AB |
897 | if (!protect) { |
898 | err = schedule_erase(ubi, e1, 0); | |
899 | if (err) | |
900 | goto out_error; | |
801c135c AB |
901 | } |
902 | ||
43f9b25a | 903 | |
801c135c | 904 | dbg_wl("done"); |
43f9b25a AB |
905 | mutex_unlock(&ubi->move_mutex); |
906 | return 0; | |
801c135c AB |
907 | |
908 | /* | |
43f9b25a AB |
909 | * For some reasons the LEB was not moved, might be an error, might be |
910 | * something else. @e1 was not changed, so return it back. @e2 might | |
911 | * be changed, schedule it for erasure. | |
801c135c | 912 | */ |
43f9b25a | 913 | out_not_moved: |
801c135c AB |
914 | ubi_free_vid_hdr(ubi, vid_hdr); |
915 | spin_lock(&ubi->wl_lock); | |
43f9b25a AB |
916 | if (scrubbing) |
917 | wl_tree_add(e1, &ubi->scrub); | |
801c135c | 918 | else |
5abde384 | 919 | wl_tree_add(e1, &ubi->used); |
801c135c | 920 | ubi->move_from = ubi->move_to = NULL; |
43f9b25a | 921 | ubi->move_to_put = ubi->wl_scheduled = 0; |
801c135c AB |
922 | spin_unlock(&ubi->wl_lock); |
923 | ||
801c135c | 924 | err = schedule_erase(ubi, e2, 0); |
43f9b25a AB |
925 | if (err) |
926 | goto out_error; | |
927 | ||
928 | mutex_unlock(&ubi->move_mutex); | |
929 | return 0; | |
930 | ||
931 | out_error: | |
932 | ubi_err("error %d while moving PEB %d to PEB %d", | |
933 | err, e1->pnum, e2->pnum); | |
801c135c | 934 | |
43f9b25a AB |
935 | ubi_free_vid_hdr(ubi, vid_hdr); |
936 | spin_lock(&ubi->wl_lock); | |
937 | ubi->move_from = ubi->move_to = NULL; | |
938 | ubi->move_to_put = ubi->wl_scheduled = 0; | |
939 | spin_unlock(&ubi->wl_lock); | |
940 | ||
941 | kmem_cache_free(ubi_wl_entry_slab, e1); | |
942 | kmem_cache_free(ubi_wl_entry_slab, e2); | |
943 | ubi_ro_mode(ubi); | |
944 | ||
945 | mutex_unlock(&ubi->move_mutex); | |
801c135c | 946 | return err; |
43f9b25a AB |
947 | |
948 | out_cancel: | |
949 | ubi->wl_scheduled = 0; | |
950 | spin_unlock(&ubi->wl_lock); | |
951 | mutex_unlock(&ubi->move_mutex); | |
952 | ubi_free_vid_hdr(ubi, vid_hdr); | |
953 | return 0; | |
801c135c AB |
954 | } |
955 | ||
956 | /** | |
957 | * ensure_wear_leveling - schedule wear-leveling if it is needed. | |
958 | * @ubi: UBI device description object | |
959 | * | |
960 | * This function checks if it is time to start wear-leveling and schedules it | |
961 | * if yes. This function returns zero in case of success and a negative error | |
962 | * code in case of failure. | |
963 | */ | |
964 | static int ensure_wear_leveling(struct ubi_device *ubi) | |
965 | { | |
966 | int err = 0; | |
967 | struct ubi_wl_entry *e1; | |
968 | struct ubi_wl_entry *e2; | |
969 | struct ubi_work *wrk; | |
970 | ||
971 | spin_lock(&ubi->wl_lock); | |
972 | if (ubi->wl_scheduled) | |
973 | /* Wear-leveling is already in the work queue */ | |
974 | goto out_unlock; | |
975 | ||
976 | /* | |
977 | * If the ubi->scrub tree is not empty, scrubbing is needed, and the | |
978 | * the WL worker has to be scheduled anyway. | |
979 | */ | |
5abde384 AB |
980 | if (!ubi->scrub.rb_node) { |
981 | if (!ubi->used.rb_node || !ubi->free.rb_node) | |
801c135c AB |
982 | /* No physical eraseblocks - no deal */ |
983 | goto out_unlock; | |
984 | ||
985 | /* | |
986 | * We schedule wear-leveling only if the difference between the | |
987 | * lowest erase counter of used physical eraseblocks and a high | |
988 | * erase counter of free physical eraseblocks is greater then | |
989 | * %UBI_WL_THRESHOLD. | |
990 | */ | |
991 | e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, rb); | |
992 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); | |
993 | ||
994 | if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD)) | |
995 | goto out_unlock; | |
996 | dbg_wl("schedule wear-leveling"); | |
997 | } else | |
998 | dbg_wl("schedule scrubbing"); | |
999 | ||
1000 | ubi->wl_scheduled = 1; | |
1001 | spin_unlock(&ubi->wl_lock); | |
1002 | ||
33818bbb | 1003 | wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS); |
801c135c AB |
1004 | if (!wrk) { |
1005 | err = -ENOMEM; | |
1006 | goto out_cancel; | |
1007 | } | |
1008 | ||
1009 | wrk->func = &wear_leveling_worker; | |
1010 | schedule_ubi_work(ubi, wrk); | |
1011 | return err; | |
1012 | ||
1013 | out_cancel: | |
1014 | spin_lock(&ubi->wl_lock); | |
1015 | ubi->wl_scheduled = 0; | |
1016 | out_unlock: | |
1017 | spin_unlock(&ubi->wl_lock); | |
1018 | return err; | |
1019 | } | |
1020 | ||
1021 | /** | |
1022 | * erase_worker - physical eraseblock erase worker function. | |
1023 | * @ubi: UBI device description object | |
1024 | * @wl_wrk: the work object | |
1025 | * @cancel: non-zero if the worker has to free memory and exit | |
1026 | * | |
1027 | * This function erases a physical eraseblock and perform torture testing if | |
1028 | * needed. It also takes care about marking the physical eraseblock bad if | |
1029 | * needed. Returns zero in case of success and a negative error code in case of | |
1030 | * failure. | |
1031 | */ | |
1032 | static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk, | |
1033 | int cancel) | |
1034 | { | |
801c135c | 1035 | struct ubi_wl_entry *e = wl_wrk->e; |
784c1454 | 1036 | int pnum = e->pnum, err, need; |
801c135c AB |
1037 | |
1038 | if (cancel) { | |
1039 | dbg_wl("cancel erasure of PEB %d EC %d", pnum, e->ec); | |
1040 | kfree(wl_wrk); | |
06b68ba1 | 1041 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1042 | return 0; |
1043 | } | |
1044 | ||
1045 | dbg_wl("erase PEB %d EC %d", pnum, e->ec); | |
1046 | ||
1047 | err = sync_erase(ubi, e, wl_wrk->torture); | |
1048 | if (!err) { | |
1049 | /* Fine, we've erased it successfully */ | |
1050 | kfree(wl_wrk); | |
1051 | ||
1052 | spin_lock(&ubi->wl_lock); | |
1053 | ubi->abs_ec += 1; | |
5abde384 | 1054 | wl_tree_add(e, &ubi->free); |
801c135c AB |
1055 | spin_unlock(&ubi->wl_lock); |
1056 | ||
1057 | /* | |
9c9ec147 AB |
1058 | * One more erase operation has happened, take care about |
1059 | * protected physical eraseblocks. | |
801c135c AB |
1060 | */ |
1061 | check_protection_over(ubi); | |
1062 | ||
1063 | /* And take care about wear-leveling */ | |
1064 | err = ensure_wear_leveling(ubi); | |
1065 | return err; | |
1066 | } | |
1067 | ||
8d2d4011 | 1068 | ubi_err("failed to erase PEB %d, error %d", pnum, err); |
801c135c | 1069 | kfree(wl_wrk); |
06b68ba1 | 1070 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c | 1071 | |
784c1454 AB |
1072 | if (err == -EINTR || err == -ENOMEM || err == -EAGAIN || |
1073 | err == -EBUSY) { | |
1074 | int err1; | |
1075 | ||
1076 | /* Re-schedule the LEB for erasure */ | |
1077 | err1 = schedule_erase(ubi, e, 0); | |
1078 | if (err1) { | |
1079 | err = err1; | |
1080 | goto out_ro; | |
1081 | } | |
1082 | return err; | |
1083 | } else if (err != -EIO) { | |
801c135c AB |
1084 | /* |
1085 | * If this is not %-EIO, we have no idea what to do. Scheduling | |
1086 | * this physical eraseblock for erasure again would cause | |
1087 | * errors again and again. Well, lets switch to RO mode. | |
1088 | */ | |
784c1454 | 1089 | goto out_ro; |
801c135c AB |
1090 | } |
1091 | ||
1092 | /* It is %-EIO, the PEB went bad */ | |
1093 | ||
1094 | if (!ubi->bad_allowed) { | |
1095 | ubi_err("bad physical eraseblock %d detected", pnum); | |
784c1454 AB |
1096 | goto out_ro; |
1097 | } | |
801c135c | 1098 | |
784c1454 AB |
1099 | spin_lock(&ubi->volumes_lock); |
1100 | need = ubi->beb_rsvd_level - ubi->beb_rsvd_pebs + 1; | |
1101 | if (need > 0) { | |
1102 | need = ubi->avail_pebs >= need ? need : ubi->avail_pebs; | |
1103 | ubi->avail_pebs -= need; | |
1104 | ubi->rsvd_pebs += need; | |
1105 | ubi->beb_rsvd_pebs += need; | |
1106 | if (need > 0) | |
1107 | ubi_msg("reserve more %d PEBs", need); | |
1108 | } | |
801c135c | 1109 | |
784c1454 | 1110 | if (ubi->beb_rsvd_pebs == 0) { |
801c135c | 1111 | spin_unlock(&ubi->volumes_lock); |
784c1454 AB |
1112 | ubi_err("no reserved physical eraseblocks"); |
1113 | goto out_ro; | |
1114 | } | |
801c135c | 1115 | |
784c1454 AB |
1116 | spin_unlock(&ubi->volumes_lock); |
1117 | ubi_msg("mark PEB %d as bad", pnum); | |
801c135c | 1118 | |
784c1454 AB |
1119 | err = ubi_io_mark_bad(ubi, pnum); |
1120 | if (err) | |
1121 | goto out_ro; | |
1122 | ||
1123 | spin_lock(&ubi->volumes_lock); | |
1124 | ubi->beb_rsvd_pebs -= 1; | |
1125 | ubi->bad_peb_count += 1; | |
1126 | ubi->good_peb_count -= 1; | |
1127 | ubi_calculate_reserved(ubi); | |
1128 | if (ubi->beb_rsvd_pebs == 0) | |
1129 | ubi_warn("last PEB from the reserved pool was used"); | |
1130 | spin_unlock(&ubi->volumes_lock); | |
1131 | ||
1132 | return err; | |
801c135c | 1133 | |
784c1454 AB |
1134 | out_ro: |
1135 | ubi_ro_mode(ubi); | |
801c135c AB |
1136 | return err; |
1137 | } | |
1138 | ||
1139 | /** | |
85c6e6e2 | 1140 | * ubi_wl_put_peb - return a PEB to the wear-leveling sub-system. |
801c135c AB |
1141 | * @ubi: UBI device description object |
1142 | * @pnum: physical eraseblock to return | |
1143 | * @torture: if this physical eraseblock has to be tortured | |
1144 | * | |
1145 | * This function is called to return physical eraseblock @pnum to the pool of | |
1146 | * free physical eraseblocks. The @torture flag has to be set if an I/O error | |
1147 | * occurred to this @pnum and it has to be tested. This function returns zero | |
43f9b25a | 1148 | * in case of success, and a negative error code in case of failure. |
801c135c AB |
1149 | */ |
1150 | int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture) | |
1151 | { | |
1152 | int err; | |
1153 | struct ubi_wl_entry *e; | |
1154 | ||
1155 | dbg_wl("PEB %d", pnum); | |
1156 | ubi_assert(pnum >= 0); | |
1157 | ubi_assert(pnum < ubi->peb_count); | |
1158 | ||
43f9b25a | 1159 | retry: |
801c135c | 1160 | spin_lock(&ubi->wl_lock); |
801c135c AB |
1161 | e = ubi->lookuptbl[pnum]; |
1162 | if (e == ubi->move_from) { | |
1163 | /* | |
1164 | * User is putting the physical eraseblock which was selected to | |
1165 | * be moved. It will be scheduled for erasure in the | |
1166 | * wear-leveling worker. | |
1167 | */ | |
43f9b25a | 1168 | dbg_wl("PEB %d is being moved, wait", pnum); |
801c135c | 1169 | spin_unlock(&ubi->wl_lock); |
43f9b25a AB |
1170 | |
1171 | /* Wait for the WL worker by taking the @ubi->move_mutex */ | |
1172 | mutex_lock(&ubi->move_mutex); | |
1173 | mutex_unlock(&ubi->move_mutex); | |
1174 | goto retry; | |
801c135c AB |
1175 | } else if (e == ubi->move_to) { |
1176 | /* | |
1177 | * User is putting the physical eraseblock which was selected | |
1178 | * as the target the data is moved to. It may happen if the EBA | |
85c6e6e2 AB |
1179 | * sub-system already re-mapped the LEB in 'ubi_eba_copy_leb()' |
1180 | * but the WL sub-system has not put the PEB to the "used" tree | |
1181 | * yet, but it is about to do this. So we just set a flag which | |
1182 | * will tell the WL worker that the PEB is not needed anymore | |
1183 | * and should be scheduled for erasure. | |
801c135c AB |
1184 | */ |
1185 | dbg_wl("PEB %d is the target of data moving", pnum); | |
1186 | ubi_assert(!ubi->move_to_put); | |
1187 | ubi->move_to_put = 1; | |
1188 | spin_unlock(&ubi->wl_lock); | |
1189 | return 0; | |
1190 | } else { | |
5abde384 AB |
1191 | if (in_wl_tree(e, &ubi->used)) { |
1192 | paranoid_check_in_wl_tree(e, &ubi->used); | |
1193 | rb_erase(&e->rb, &ubi->used); | |
1194 | } else if (in_wl_tree(e, &ubi->scrub)) { | |
1195 | paranoid_check_in_wl_tree(e, &ubi->scrub); | |
1196 | rb_erase(&e->rb, &ubi->scrub); | |
43f9b25a AB |
1197 | } else { |
1198 | err = prot_tree_del(ubi, e->pnum); | |
1199 | if (err) { | |
1200 | ubi_err("PEB %d not found", pnum); | |
1201 | ubi_ro_mode(ubi); | |
1202 | spin_unlock(&ubi->wl_lock); | |
1203 | return err; | |
1204 | } | |
1205 | } | |
801c135c AB |
1206 | } |
1207 | spin_unlock(&ubi->wl_lock); | |
1208 | ||
1209 | err = schedule_erase(ubi, e, torture); | |
1210 | if (err) { | |
1211 | spin_lock(&ubi->wl_lock); | |
5abde384 | 1212 | wl_tree_add(e, &ubi->used); |
801c135c AB |
1213 | spin_unlock(&ubi->wl_lock); |
1214 | } | |
1215 | ||
1216 | return err; | |
1217 | } | |
1218 | ||
1219 | /** | |
1220 | * ubi_wl_scrub_peb - schedule a physical eraseblock for scrubbing. | |
1221 | * @ubi: UBI device description object | |
1222 | * @pnum: the physical eraseblock to schedule | |
1223 | * | |
1224 | * If a bit-flip in a physical eraseblock is detected, this physical eraseblock | |
1225 | * needs scrubbing. This function schedules a physical eraseblock for | |
1226 | * scrubbing which is done in background. This function returns zero in case of | |
1227 | * success and a negative error code in case of failure. | |
1228 | */ | |
1229 | int ubi_wl_scrub_peb(struct ubi_device *ubi, int pnum) | |
1230 | { | |
1231 | struct ubi_wl_entry *e; | |
1232 | ||
8c1e6ee1 | 1233 | dbg_msg("schedule PEB %d for scrubbing", pnum); |
801c135c AB |
1234 | |
1235 | retry: | |
1236 | spin_lock(&ubi->wl_lock); | |
1237 | e = ubi->lookuptbl[pnum]; | |
1238 | if (e == ubi->move_from || in_wl_tree(e, &ubi->scrub)) { | |
1239 | spin_unlock(&ubi->wl_lock); | |
1240 | return 0; | |
1241 | } | |
1242 | ||
1243 | if (e == ubi->move_to) { | |
1244 | /* | |
1245 | * This physical eraseblock was used to move data to. The data | |
1246 | * was moved but the PEB was not yet inserted to the proper | |
1247 | * tree. We should just wait a little and let the WL worker | |
1248 | * proceed. | |
1249 | */ | |
1250 | spin_unlock(&ubi->wl_lock); | |
1251 | dbg_wl("the PEB %d is not in proper tree, retry", pnum); | |
1252 | yield(); | |
1253 | goto retry; | |
1254 | } | |
1255 | ||
5abde384 AB |
1256 | if (in_wl_tree(e, &ubi->used)) { |
1257 | paranoid_check_in_wl_tree(e, &ubi->used); | |
1258 | rb_erase(&e->rb, &ubi->used); | |
43f9b25a AB |
1259 | } else { |
1260 | int err; | |
1261 | ||
1262 | err = prot_tree_del(ubi, e->pnum); | |
1263 | if (err) { | |
1264 | ubi_err("PEB %d not found", pnum); | |
1265 | ubi_ro_mode(ubi); | |
1266 | spin_unlock(&ubi->wl_lock); | |
1267 | return err; | |
1268 | } | |
1269 | } | |
801c135c | 1270 | |
5abde384 | 1271 | wl_tree_add(e, &ubi->scrub); |
801c135c AB |
1272 | spin_unlock(&ubi->wl_lock); |
1273 | ||
1274 | /* | |
1275 | * Technically scrubbing is the same as wear-leveling, so it is done | |
1276 | * by the WL worker. | |
1277 | */ | |
1278 | return ensure_wear_leveling(ubi); | |
1279 | } | |
1280 | ||
1281 | /** | |
1282 | * ubi_wl_flush - flush all pending works. | |
1283 | * @ubi: UBI device description object | |
1284 | * | |
1285 | * This function returns zero in case of success and a negative error code in | |
1286 | * case of failure. | |
1287 | */ | |
1288 | int ubi_wl_flush(struct ubi_device *ubi) | |
1289 | { | |
593dd33c | 1290 | int err; |
801c135c AB |
1291 | |
1292 | /* | |
1293 | * Erase while the pending works queue is not empty, but not more then | |
1294 | * the number of currently pending works. | |
1295 | */ | |
593dd33c AB |
1296 | dbg_wl("flush (%d pending works)", ubi->works_count); |
1297 | while (ubi->works_count) { | |
1298 | err = do_work(ubi); | |
1299 | if (err) | |
1300 | return err; | |
1301 | } | |
1302 | ||
1303 | /* | |
1304 | * Make sure all the works which have been done in parallel are | |
1305 | * finished. | |
1306 | */ | |
1307 | down_write(&ubi->work_sem); | |
1308 | up_write(&ubi->work_sem); | |
1309 | ||
1310 | /* | |
1311 | * And in case last was the WL worker and it cancelled the LEB | |
1312 | * movement, flush again. | |
1313 | */ | |
1314 | while (ubi->works_count) { | |
1315 | dbg_wl("flush more (%d pending works)", ubi->works_count); | |
801c135c AB |
1316 | err = do_work(ubi); |
1317 | if (err) | |
1318 | return err; | |
1319 | } | |
1320 | ||
1321 | return 0; | |
1322 | } | |
1323 | ||
1324 | /** | |
1325 | * tree_destroy - destroy an RB-tree. | |
1326 | * @root: the root of the tree to destroy | |
1327 | */ | |
1328 | static void tree_destroy(struct rb_root *root) | |
1329 | { | |
1330 | struct rb_node *rb; | |
1331 | struct ubi_wl_entry *e; | |
1332 | ||
1333 | rb = root->rb_node; | |
1334 | while (rb) { | |
1335 | if (rb->rb_left) | |
1336 | rb = rb->rb_left; | |
1337 | else if (rb->rb_right) | |
1338 | rb = rb->rb_right; | |
1339 | else { | |
1340 | e = rb_entry(rb, struct ubi_wl_entry, rb); | |
1341 | ||
1342 | rb = rb_parent(rb); | |
1343 | if (rb) { | |
1344 | if (rb->rb_left == &e->rb) | |
1345 | rb->rb_left = NULL; | |
1346 | else | |
1347 | rb->rb_right = NULL; | |
1348 | } | |
1349 | ||
06b68ba1 | 1350 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1351 | } |
1352 | } | |
1353 | } | |
1354 | ||
1355 | /** | |
1356 | * ubi_thread - UBI background thread. | |
1357 | * @u: the UBI device description object pointer | |
1358 | */ | |
cdfa788a | 1359 | int ubi_thread(void *u) |
801c135c AB |
1360 | { |
1361 | int failures = 0; | |
1362 | struct ubi_device *ubi = u; | |
1363 | ||
1364 | ubi_msg("background thread \"%s\" started, PID %d", | |
ba25f9dc | 1365 | ubi->bgt_name, task_pid_nr(current)); |
801c135c | 1366 | |
83144186 | 1367 | set_freezable(); |
801c135c AB |
1368 | for (;;) { |
1369 | int err; | |
1370 | ||
1371 | if (kthread_should_stop()) | |
cadb40cc | 1372 | break; |
801c135c AB |
1373 | |
1374 | if (try_to_freeze()) | |
1375 | continue; | |
1376 | ||
1377 | spin_lock(&ubi->wl_lock); | |
1378 | if (list_empty(&ubi->works) || ubi->ro_mode || | |
1379 | !ubi->thread_enabled) { | |
1380 | set_current_state(TASK_INTERRUPTIBLE); | |
1381 | spin_unlock(&ubi->wl_lock); | |
1382 | schedule(); | |
1383 | continue; | |
1384 | } | |
1385 | spin_unlock(&ubi->wl_lock); | |
1386 | ||
1387 | err = do_work(ubi); | |
1388 | if (err) { | |
1389 | ubi_err("%s: work failed with error code %d", | |
1390 | ubi->bgt_name, err); | |
1391 | if (failures++ > WL_MAX_FAILURES) { | |
1392 | /* | |
1393 | * Too many failures, disable the thread and | |
1394 | * switch to read-only mode. | |
1395 | */ | |
1396 | ubi_msg("%s: %d consecutive failures", | |
1397 | ubi->bgt_name, WL_MAX_FAILURES); | |
1398 | ubi_ro_mode(ubi); | |
2ad49887 VG |
1399 | ubi->thread_enabled = 0; |
1400 | continue; | |
801c135c AB |
1401 | } |
1402 | } else | |
1403 | failures = 0; | |
1404 | ||
1405 | cond_resched(); | |
1406 | } | |
1407 | ||
801c135c AB |
1408 | dbg_wl("background thread \"%s\" is killed", ubi->bgt_name); |
1409 | return 0; | |
1410 | } | |
1411 | ||
1412 | /** | |
1413 | * cancel_pending - cancel all pending works. | |
1414 | * @ubi: UBI device description object | |
1415 | */ | |
1416 | static void cancel_pending(struct ubi_device *ubi) | |
1417 | { | |
1418 | while (!list_empty(&ubi->works)) { | |
1419 | struct ubi_work *wrk; | |
1420 | ||
1421 | wrk = list_entry(ubi->works.next, struct ubi_work, list); | |
1422 | list_del(&wrk->list); | |
1423 | wrk->func(ubi, wrk, 1); | |
1424 | ubi->works_count -= 1; | |
1425 | ubi_assert(ubi->works_count >= 0); | |
1426 | } | |
1427 | } | |
1428 | ||
1429 | /** | |
85c6e6e2 | 1430 | * ubi_wl_init_scan - initialize the WL sub-system using scanning information. |
801c135c AB |
1431 | * @ubi: UBI device description object |
1432 | * @si: scanning information | |
1433 | * | |
1434 | * This function returns zero in case of success, and a negative error code in | |
1435 | * case of failure. | |
1436 | */ | |
1437 | int ubi_wl_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si) | |
1438 | { | |
1439 | int err; | |
1440 | struct rb_node *rb1, *rb2; | |
1441 | struct ubi_scan_volume *sv; | |
1442 | struct ubi_scan_leb *seb, *tmp; | |
1443 | struct ubi_wl_entry *e; | |
1444 | ||
1445 | ||
1446 | ubi->used = ubi->free = ubi->scrub = RB_ROOT; | |
1447 | ubi->prot.pnum = ubi->prot.aec = RB_ROOT; | |
1448 | spin_lock_init(&ubi->wl_lock); | |
43f9b25a | 1449 | mutex_init(&ubi->move_mutex); |
593dd33c | 1450 | init_rwsem(&ubi->work_sem); |
801c135c AB |
1451 | ubi->max_ec = si->max_ec; |
1452 | INIT_LIST_HEAD(&ubi->works); | |
1453 | ||
1454 | sprintf(ubi->bgt_name, UBI_BGT_NAME_PATTERN, ubi->ubi_num); | |
1455 | ||
801c135c AB |
1456 | err = -ENOMEM; |
1457 | ubi->lookuptbl = kzalloc(ubi->peb_count * sizeof(void *), GFP_KERNEL); | |
1458 | if (!ubi->lookuptbl) | |
cdfa788a | 1459 | return err; |
801c135c AB |
1460 | |
1461 | list_for_each_entry_safe(seb, tmp, &si->erase, u.list) { | |
1462 | cond_resched(); | |
1463 | ||
06b68ba1 | 1464 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1465 | if (!e) |
1466 | goto out_free; | |
1467 | ||
1468 | e->pnum = seb->pnum; | |
1469 | e->ec = seb->ec; | |
1470 | ubi->lookuptbl[e->pnum] = e; | |
1471 | if (schedule_erase(ubi, e, 0)) { | |
06b68ba1 | 1472 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1473 | goto out_free; |
1474 | } | |
1475 | } | |
1476 | ||
1477 | list_for_each_entry(seb, &si->free, u.list) { | |
1478 | cond_resched(); | |
1479 | ||
06b68ba1 | 1480 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1481 | if (!e) |
1482 | goto out_free; | |
1483 | ||
1484 | e->pnum = seb->pnum; | |
1485 | e->ec = seb->ec; | |
1486 | ubi_assert(e->ec >= 0); | |
5abde384 | 1487 | wl_tree_add(e, &ubi->free); |
801c135c AB |
1488 | ubi->lookuptbl[e->pnum] = e; |
1489 | } | |
1490 | ||
1491 | list_for_each_entry(seb, &si->corr, u.list) { | |
1492 | cond_resched(); | |
1493 | ||
06b68ba1 | 1494 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1495 | if (!e) |
1496 | goto out_free; | |
1497 | ||
1498 | e->pnum = seb->pnum; | |
1499 | e->ec = seb->ec; | |
1500 | ubi->lookuptbl[e->pnum] = e; | |
1501 | if (schedule_erase(ubi, e, 0)) { | |
06b68ba1 | 1502 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1503 | goto out_free; |
1504 | } | |
1505 | } | |
1506 | ||
1507 | ubi_rb_for_each_entry(rb1, sv, &si->volumes, rb) { | |
1508 | ubi_rb_for_each_entry(rb2, seb, &sv->root, u.rb) { | |
1509 | cond_resched(); | |
1510 | ||
06b68ba1 | 1511 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1512 | if (!e) |
1513 | goto out_free; | |
1514 | ||
1515 | e->pnum = seb->pnum; | |
1516 | e->ec = seb->ec; | |
1517 | ubi->lookuptbl[e->pnum] = e; | |
1518 | if (!seb->scrub) { | |
1519 | dbg_wl("add PEB %d EC %d to the used tree", | |
1520 | e->pnum, e->ec); | |
5abde384 | 1521 | wl_tree_add(e, &ubi->used); |
801c135c AB |
1522 | } else { |
1523 | dbg_wl("add PEB %d EC %d to the scrub tree", | |
1524 | e->pnum, e->ec); | |
5abde384 | 1525 | wl_tree_add(e, &ubi->scrub); |
801c135c AB |
1526 | } |
1527 | } | |
1528 | } | |
1529 | ||
5abde384 | 1530 | if (ubi->avail_pebs < WL_RESERVED_PEBS) { |
801c135c AB |
1531 | ubi_err("no enough physical eraseblocks (%d, need %d)", |
1532 | ubi->avail_pebs, WL_RESERVED_PEBS); | |
1533 | goto out_free; | |
1534 | } | |
1535 | ubi->avail_pebs -= WL_RESERVED_PEBS; | |
1536 | ubi->rsvd_pebs += WL_RESERVED_PEBS; | |
1537 | ||
1538 | /* Schedule wear-leveling if needed */ | |
1539 | err = ensure_wear_leveling(ubi); | |
1540 | if (err) | |
1541 | goto out_free; | |
1542 | ||
1543 | return 0; | |
1544 | ||
1545 | out_free: | |
1546 | cancel_pending(ubi); | |
1547 | tree_destroy(&ubi->used); | |
1548 | tree_destroy(&ubi->free); | |
1549 | tree_destroy(&ubi->scrub); | |
1550 | kfree(ubi->lookuptbl); | |
801c135c AB |
1551 | return err; |
1552 | } | |
1553 | ||
1554 | /** | |
1555 | * protection_trees_destroy - destroy the protection RB-trees. | |
1556 | * @ubi: UBI device description object | |
1557 | */ | |
1558 | static void protection_trees_destroy(struct ubi_device *ubi) | |
1559 | { | |
1560 | struct rb_node *rb; | |
1561 | struct ubi_wl_prot_entry *pe; | |
1562 | ||
1563 | rb = ubi->prot.aec.rb_node; | |
1564 | while (rb) { | |
1565 | if (rb->rb_left) | |
1566 | rb = rb->rb_left; | |
1567 | else if (rb->rb_right) | |
1568 | rb = rb->rb_right; | |
1569 | else { | |
1570 | pe = rb_entry(rb, struct ubi_wl_prot_entry, rb_aec); | |
1571 | ||
1572 | rb = rb_parent(rb); | |
1573 | if (rb) { | |
1574 | if (rb->rb_left == &pe->rb_aec) | |
1575 | rb->rb_left = NULL; | |
1576 | else | |
1577 | rb->rb_right = NULL; | |
1578 | } | |
1579 | ||
06b68ba1 | 1580 | kmem_cache_free(ubi_wl_entry_slab, pe->e); |
801c135c AB |
1581 | kfree(pe); |
1582 | } | |
1583 | } | |
1584 | } | |
1585 | ||
1586 | /** | |
85c6e6e2 | 1587 | * ubi_wl_close - close the wear-leveling sub-system. |
801c135c AB |
1588 | * @ubi: UBI device description object |
1589 | */ | |
1590 | void ubi_wl_close(struct ubi_device *ubi) | |
1591 | { | |
85c6e6e2 | 1592 | dbg_wl("close the WL sub-system"); |
801c135c AB |
1593 | cancel_pending(ubi); |
1594 | protection_trees_destroy(ubi); | |
1595 | tree_destroy(&ubi->used); | |
1596 | tree_destroy(&ubi->free); | |
1597 | tree_destroy(&ubi->scrub); | |
1598 | kfree(ubi->lookuptbl); | |
801c135c AB |
1599 | } |
1600 | ||
1601 | #ifdef CONFIG_MTD_UBI_DEBUG_PARANOID | |
1602 | ||
1603 | /** | |
ebaaf1af | 1604 | * paranoid_check_ec - make sure that the erase counter of a PEB is correct. |
801c135c AB |
1605 | * @ubi: UBI device description object |
1606 | * @pnum: the physical eraseblock number to check | |
1607 | * @ec: the erase counter to check | |
1608 | * | |
1609 | * This function returns zero if the erase counter of physical eraseblock @pnum | |
1610 | * is equivalent to @ec, %1 if not, and a negative error code if an error | |
1611 | * occurred. | |
1612 | */ | |
e88d6e10 | 1613 | static int paranoid_check_ec(struct ubi_device *ubi, int pnum, int ec) |
801c135c AB |
1614 | { |
1615 | int err; | |
1616 | long long read_ec; | |
1617 | struct ubi_ec_hdr *ec_hdr; | |
1618 | ||
33818bbb | 1619 | ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_NOFS); |
801c135c AB |
1620 | if (!ec_hdr) |
1621 | return -ENOMEM; | |
1622 | ||
1623 | err = ubi_io_read_ec_hdr(ubi, pnum, ec_hdr, 0); | |
1624 | if (err && err != UBI_IO_BITFLIPS) { | |
1625 | /* The header does not have to exist */ | |
1626 | err = 0; | |
1627 | goto out_free; | |
1628 | } | |
1629 | ||
3261ebd7 | 1630 | read_ec = be64_to_cpu(ec_hdr->ec); |
801c135c AB |
1631 | if (ec != read_ec) { |
1632 | ubi_err("paranoid check failed for PEB %d", pnum); | |
1633 | ubi_err("read EC is %lld, should be %d", read_ec, ec); | |
1634 | ubi_dbg_dump_stack(); | |
1635 | err = 1; | |
1636 | } else | |
1637 | err = 0; | |
1638 | ||
1639 | out_free: | |
1640 | kfree(ec_hdr); | |
1641 | return err; | |
1642 | } | |
1643 | ||
1644 | /** | |
ebaaf1af | 1645 | * paranoid_check_in_wl_tree - check that wear-leveling entry is in WL RB-tree. |
801c135c AB |
1646 | * @e: the wear-leveling entry to check |
1647 | * @root: the root of the tree | |
1648 | * | |
ebaaf1af AB |
1649 | * This function returns zero if @e is in the @root RB-tree and %1 if it is |
1650 | * not. | |
801c135c AB |
1651 | */ |
1652 | static int paranoid_check_in_wl_tree(struct ubi_wl_entry *e, | |
1653 | struct rb_root *root) | |
1654 | { | |
1655 | if (in_wl_tree(e, root)) | |
1656 | return 0; | |
1657 | ||
1658 | ubi_err("paranoid check failed for PEB %d, EC %d, RB-tree %p ", | |
1659 | e->pnum, e->ec, root); | |
1660 | ubi_dbg_dump_stack(); | |
1661 | return 1; | |
1662 | } | |
1663 | ||
1664 | #endif /* CONFIG_MTD_UBI_DEBUG_PARANOID */ |