]>
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 | |
6a8f483f | 362 | * @ec: for how many erase operations this PEB should be protected |
801c135c AB |
363 | * |
364 | * @wl->lock has to be locked. | |
365 | */ | |
366 | static void prot_tree_add(struct ubi_device *ubi, struct ubi_wl_entry *e, | |
6a8f483f | 367 | struct ubi_wl_prot_entry *pe, int ec) |
801c135c AB |
368 | { |
369 | struct rb_node **p, *parent = NULL; | |
370 | struct ubi_wl_prot_entry *pe1; | |
371 | ||
372 | pe->e = e; | |
6a8f483f | 373 | pe->abs_ec = ubi->abs_ec + ec; |
801c135c AB |
374 | |
375 | p = &ubi->prot.pnum.rb_node; | |
376 | while (*p) { | |
377 | parent = *p; | |
378 | pe1 = rb_entry(parent, struct ubi_wl_prot_entry, rb_pnum); | |
379 | ||
380 | if (e->pnum < pe1->e->pnum) | |
381 | p = &(*p)->rb_left; | |
382 | else | |
383 | p = &(*p)->rb_right; | |
384 | } | |
385 | rb_link_node(&pe->rb_pnum, parent, p); | |
386 | rb_insert_color(&pe->rb_pnum, &ubi->prot.pnum); | |
387 | ||
388 | p = &ubi->prot.aec.rb_node; | |
389 | parent = NULL; | |
390 | while (*p) { | |
391 | parent = *p; | |
392 | pe1 = rb_entry(parent, struct ubi_wl_prot_entry, rb_aec); | |
393 | ||
394 | if (pe->abs_ec < pe1->abs_ec) | |
395 | p = &(*p)->rb_left; | |
396 | else | |
397 | p = &(*p)->rb_right; | |
398 | } | |
399 | rb_link_node(&pe->rb_aec, parent, p); | |
400 | rb_insert_color(&pe->rb_aec, &ubi->prot.aec); | |
401 | } | |
402 | ||
403 | /** | |
404 | * find_wl_entry - find wear-leveling entry closest to certain erase counter. | |
405 | * @root: the RB-tree where to look for | |
406 | * @max: highest possible erase counter | |
407 | * | |
408 | * This function looks for a wear leveling entry with erase counter closest to | |
409 | * @max and less then @max. | |
410 | */ | |
411 | static struct ubi_wl_entry *find_wl_entry(struct rb_root *root, int max) | |
412 | { | |
413 | struct rb_node *p; | |
414 | struct ubi_wl_entry *e; | |
415 | ||
416 | e = rb_entry(rb_first(root), struct ubi_wl_entry, rb); | |
417 | max += e->ec; | |
418 | ||
419 | p = root->rb_node; | |
420 | while (p) { | |
421 | struct ubi_wl_entry *e1; | |
422 | ||
423 | e1 = rb_entry(p, struct ubi_wl_entry, rb); | |
424 | if (e1->ec >= max) | |
425 | p = p->rb_left; | |
426 | else { | |
427 | p = p->rb_right; | |
428 | e = e1; | |
429 | } | |
430 | } | |
431 | ||
432 | return e; | |
433 | } | |
434 | ||
435 | /** | |
436 | * ubi_wl_get_peb - get a physical eraseblock. | |
437 | * @ubi: UBI device description object | |
438 | * @dtype: type of data which will be stored in this physical eraseblock | |
439 | * | |
440 | * This function returns a physical eraseblock in case of success and a | |
441 | * negative error code in case of failure. Might sleep. | |
442 | */ | |
443 | int ubi_wl_get_peb(struct ubi_device *ubi, int dtype) | |
444 | { | |
445 | int err, protect, medium_ec; | |
446 | struct ubi_wl_entry *e, *first, *last; | |
447 | struct ubi_wl_prot_entry *pe; | |
448 | ||
449 | ubi_assert(dtype == UBI_LONGTERM || dtype == UBI_SHORTTERM || | |
450 | dtype == UBI_UNKNOWN); | |
451 | ||
33818bbb | 452 | pe = kmalloc(sizeof(struct ubi_wl_prot_entry), GFP_NOFS); |
801c135c AB |
453 | if (!pe) |
454 | return -ENOMEM; | |
455 | ||
456 | retry: | |
457 | spin_lock(&ubi->wl_lock); | |
5abde384 | 458 | if (!ubi->free.rb_node) { |
801c135c AB |
459 | if (ubi->works_count == 0) { |
460 | ubi_assert(list_empty(&ubi->works)); | |
461 | ubi_err("no free eraseblocks"); | |
462 | spin_unlock(&ubi->wl_lock); | |
463 | kfree(pe); | |
464 | return -ENOSPC; | |
465 | } | |
466 | spin_unlock(&ubi->wl_lock); | |
467 | ||
468 | err = produce_free_peb(ubi); | |
469 | if (err < 0) { | |
470 | kfree(pe); | |
471 | return err; | |
472 | } | |
473 | goto retry; | |
474 | } | |
475 | ||
476 | switch (dtype) { | |
9c9ec147 AB |
477 | case UBI_LONGTERM: |
478 | /* | |
479 | * For long term data we pick a physical eraseblock with high | |
480 | * erase counter. But the highest erase counter we can pick is | |
481 | * bounded by the the lowest erase counter plus | |
482 | * %WL_FREE_MAX_DIFF. | |
483 | */ | |
484 | e = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); | |
485 | protect = LT_PROTECTION; | |
486 | break; | |
487 | case UBI_UNKNOWN: | |
488 | /* | |
489 | * For unknown data we pick a physical eraseblock with medium | |
490 | * erase counter. But we by no means can pick a physical | |
491 | * eraseblock with erase counter greater or equivalent than the | |
492 | * lowest erase counter plus %WL_FREE_MAX_DIFF. | |
493 | */ | |
494 | first = rb_entry(rb_first(&ubi->free), struct ubi_wl_entry, rb); | |
495 | last = rb_entry(rb_last(&ubi->free), struct ubi_wl_entry, rb); | |
801c135c | 496 | |
9c9ec147 AB |
497 | if (last->ec - first->ec < WL_FREE_MAX_DIFF) |
498 | e = rb_entry(ubi->free.rb_node, | |
499 | struct ubi_wl_entry, rb); | |
500 | else { | |
501 | medium_ec = (first->ec + WL_FREE_MAX_DIFF)/2; | |
502 | e = find_wl_entry(&ubi->free, medium_ec); | |
503 | } | |
504 | protect = U_PROTECTION; | |
505 | break; | |
506 | case UBI_SHORTTERM: | |
507 | /* | |
508 | * For short term data we pick a physical eraseblock with the | |
509 | * lowest erase counter as we expect it will be erased soon. | |
510 | */ | |
511 | e = rb_entry(rb_first(&ubi->free), struct ubi_wl_entry, rb); | |
512 | protect = ST_PROTECTION; | |
513 | break; | |
514 | default: | |
515 | protect = 0; | |
516 | e = NULL; | |
517 | BUG(); | |
801c135c AB |
518 | } |
519 | ||
520 | /* | |
521 | * Move the physical eraseblock to the protection trees where it will | |
522 | * be protected from being moved for some time. | |
523 | */ | |
5abde384 AB |
524 | paranoid_check_in_wl_tree(e, &ubi->free); |
525 | rb_erase(&e->rb, &ubi->free); | |
801c135c AB |
526 | prot_tree_add(ubi, e, pe, protect); |
527 | ||
528 | dbg_wl("PEB %d EC %d, protection %d", e->pnum, e->ec, protect); | |
529 | spin_unlock(&ubi->wl_lock); | |
530 | ||
531 | return e->pnum; | |
532 | } | |
533 | ||
534 | /** | |
535 | * prot_tree_del - remove a physical eraseblock from the protection trees | |
536 | * @ubi: UBI device description object | |
537 | * @pnum: the physical eraseblock to remove | |
43f9b25a AB |
538 | * |
539 | * This function returns PEB @pnum from the protection trees and returns zero | |
540 | * in case of success and %-ENODEV if the PEB was not found in the protection | |
541 | * trees. | |
801c135c | 542 | */ |
43f9b25a | 543 | static int prot_tree_del(struct ubi_device *ubi, int pnum) |
801c135c AB |
544 | { |
545 | struct rb_node *p; | |
546 | struct ubi_wl_prot_entry *pe = NULL; | |
547 | ||
548 | p = ubi->prot.pnum.rb_node; | |
549 | while (p) { | |
550 | ||
551 | pe = rb_entry(p, struct ubi_wl_prot_entry, rb_pnum); | |
552 | ||
553 | if (pnum == pe->e->pnum) | |
43f9b25a | 554 | goto found; |
801c135c AB |
555 | |
556 | if (pnum < pe->e->pnum) | |
557 | p = p->rb_left; | |
558 | else | |
559 | p = p->rb_right; | |
560 | } | |
561 | ||
43f9b25a AB |
562 | return -ENODEV; |
563 | ||
564 | found: | |
801c135c AB |
565 | ubi_assert(pe->e->pnum == pnum); |
566 | rb_erase(&pe->rb_aec, &ubi->prot.aec); | |
567 | rb_erase(&pe->rb_pnum, &ubi->prot.pnum); | |
568 | kfree(pe); | |
43f9b25a | 569 | return 0; |
801c135c AB |
570 | } |
571 | ||
572 | /** | |
573 | * sync_erase - synchronously erase a physical eraseblock. | |
574 | * @ubi: UBI device description object | |
575 | * @e: the the physical eraseblock to erase | |
576 | * @torture: if the physical eraseblock has to be tortured | |
577 | * | |
578 | * This function returns zero in case of success and a negative error code in | |
579 | * case of failure. | |
580 | */ | |
9c9ec147 AB |
581 | static int sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e, |
582 | int torture) | |
801c135c AB |
583 | { |
584 | int err; | |
585 | struct ubi_ec_hdr *ec_hdr; | |
586 | unsigned long long ec = e->ec; | |
587 | ||
588 | dbg_wl("erase PEB %d, old EC %llu", e->pnum, ec); | |
589 | ||
590 | err = paranoid_check_ec(ubi, e->pnum, e->ec); | |
591 | if (err > 0) | |
592 | return -EINVAL; | |
593 | ||
33818bbb | 594 | ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_NOFS); |
801c135c AB |
595 | if (!ec_hdr) |
596 | return -ENOMEM; | |
597 | ||
598 | err = ubi_io_sync_erase(ubi, e->pnum, torture); | |
599 | if (err < 0) | |
600 | goto out_free; | |
601 | ||
602 | ec += err; | |
603 | if (ec > UBI_MAX_ERASECOUNTER) { | |
604 | /* | |
605 | * Erase counter overflow. Upgrade UBI and use 64-bit | |
606 | * erase counters internally. | |
607 | */ | |
608 | ubi_err("erase counter overflow at PEB %d, EC %llu", | |
609 | e->pnum, ec); | |
610 | err = -EINVAL; | |
611 | goto out_free; | |
612 | } | |
613 | ||
614 | dbg_wl("erased PEB %d, new EC %llu", e->pnum, ec); | |
615 | ||
3261ebd7 | 616 | ec_hdr->ec = cpu_to_be64(ec); |
801c135c AB |
617 | |
618 | err = ubi_io_write_ec_hdr(ubi, e->pnum, ec_hdr); | |
619 | if (err) | |
620 | goto out_free; | |
621 | ||
622 | e->ec = ec; | |
623 | spin_lock(&ubi->wl_lock); | |
624 | if (e->ec > ubi->max_ec) | |
625 | ubi->max_ec = e->ec; | |
626 | spin_unlock(&ubi->wl_lock); | |
627 | ||
628 | out_free: | |
629 | kfree(ec_hdr); | |
630 | return err; | |
631 | } | |
632 | ||
633 | /** | |
ebaaf1af | 634 | * check_protection_over - check if it is time to stop protecting some PEBs. |
801c135c AB |
635 | * @ubi: UBI device description object |
636 | * | |
637 | * This function is called after each erase operation, when the absolute erase | |
638 | * counter is incremented, to check if some physical eraseblock have not to be | |
639 | * protected any longer. These physical eraseblocks are moved from the | |
640 | * protection trees to the used tree. | |
641 | */ | |
642 | static void check_protection_over(struct ubi_device *ubi) | |
643 | { | |
644 | struct ubi_wl_prot_entry *pe; | |
645 | ||
646 | /* | |
647 | * There may be several protected physical eraseblock to remove, | |
648 | * process them all. | |
649 | */ | |
650 | while (1) { | |
651 | spin_lock(&ubi->wl_lock); | |
5abde384 | 652 | if (!ubi->prot.aec.rb_node) { |
801c135c AB |
653 | spin_unlock(&ubi->wl_lock); |
654 | break; | |
655 | } | |
656 | ||
657 | pe = rb_entry(rb_first(&ubi->prot.aec), | |
658 | struct ubi_wl_prot_entry, rb_aec); | |
659 | ||
660 | if (pe->abs_ec > ubi->abs_ec) { | |
661 | spin_unlock(&ubi->wl_lock); | |
662 | break; | |
663 | } | |
664 | ||
665 | dbg_wl("PEB %d protection over, abs_ec %llu, PEB abs_ec %llu", | |
666 | pe->e->pnum, ubi->abs_ec, pe->abs_ec); | |
667 | rb_erase(&pe->rb_aec, &ubi->prot.aec); | |
668 | rb_erase(&pe->rb_pnum, &ubi->prot.pnum); | |
5abde384 | 669 | wl_tree_add(pe->e, &ubi->used); |
801c135c AB |
670 | spin_unlock(&ubi->wl_lock); |
671 | ||
672 | kfree(pe); | |
673 | cond_resched(); | |
674 | } | |
675 | } | |
676 | ||
677 | /** | |
678 | * schedule_ubi_work - schedule a work. | |
679 | * @ubi: UBI device description object | |
680 | * @wrk: the work to schedule | |
681 | * | |
682 | * This function enqueues a work defined by @wrk to the tail of the pending | |
683 | * works list. | |
684 | */ | |
685 | static void schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk) | |
686 | { | |
687 | spin_lock(&ubi->wl_lock); | |
688 | list_add_tail(&wrk->list, &ubi->works); | |
689 | ubi_assert(ubi->works_count >= 0); | |
690 | ubi->works_count += 1; | |
691 | if (ubi->thread_enabled) | |
692 | wake_up_process(ubi->bgt_thread); | |
693 | spin_unlock(&ubi->wl_lock); | |
694 | } | |
695 | ||
696 | static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk, | |
697 | int cancel); | |
698 | ||
699 | /** | |
700 | * schedule_erase - schedule an erase work. | |
701 | * @ubi: UBI device description object | |
702 | * @e: the WL entry of the physical eraseblock to erase | |
703 | * @torture: if the physical eraseblock has to be tortured | |
704 | * | |
705 | * This function returns zero in case of success and a %-ENOMEM in case of | |
706 | * failure. | |
707 | */ | |
708 | static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e, | |
709 | int torture) | |
710 | { | |
711 | struct ubi_work *wl_wrk; | |
712 | ||
713 | dbg_wl("schedule erasure of PEB %d, EC %d, torture %d", | |
714 | e->pnum, e->ec, torture); | |
715 | ||
33818bbb | 716 | wl_wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS); |
801c135c AB |
717 | if (!wl_wrk) |
718 | return -ENOMEM; | |
719 | ||
720 | wl_wrk->func = &erase_worker; | |
721 | wl_wrk->e = e; | |
722 | wl_wrk->torture = torture; | |
723 | ||
724 | schedule_ubi_work(ubi, wl_wrk); | |
725 | return 0; | |
726 | } | |
727 | ||
728 | /** | |
729 | * wear_leveling_worker - wear-leveling worker function. | |
730 | * @ubi: UBI device description object | |
731 | * @wrk: the work object | |
732 | * @cancel: non-zero if the worker has to free memory and exit | |
733 | * | |
734 | * This function copies a more worn out physical eraseblock to a less worn out | |
735 | * one. Returns zero in case of success and a negative error code in case of | |
736 | * failure. | |
737 | */ | |
738 | static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk, | |
739 | int cancel) | |
740 | { | |
6a8f483f | 741 | int err, put = 0, scrubbing = 0; |
c18a8418 | 742 | struct ubi_wl_prot_entry *uninitialized_var(pe); |
801c135c AB |
743 | struct ubi_wl_entry *e1, *e2; |
744 | struct ubi_vid_hdr *vid_hdr; | |
745 | ||
746 | kfree(wrk); | |
747 | ||
748 | if (cancel) | |
749 | return 0; | |
750 | ||
33818bbb | 751 | vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS); |
801c135c AB |
752 | if (!vid_hdr) |
753 | return -ENOMEM; | |
754 | ||
43f9b25a | 755 | mutex_lock(&ubi->move_mutex); |
801c135c | 756 | spin_lock(&ubi->wl_lock); |
43f9b25a AB |
757 | ubi_assert(!ubi->move_from && !ubi->move_to); |
758 | ubi_assert(!ubi->move_to_put); | |
801c135c | 759 | |
43f9b25a | 760 | if (!ubi->free.rb_node || |
5abde384 | 761 | (!ubi->used.rb_node && !ubi->scrub.rb_node)) { |
801c135c | 762 | /* |
43f9b25a AB |
763 | * No free physical eraseblocks? Well, they must be waiting in |
764 | * the queue to be erased. Cancel movement - it will be | |
765 | * triggered again when a free physical eraseblock appears. | |
801c135c AB |
766 | * |
767 | * No used physical eraseblocks? They must be temporarily | |
768 | * protected from being moved. They will be moved to the | |
769 | * @ubi->used tree later and the wear-leveling will be | |
770 | * triggered again. | |
771 | */ | |
772 | dbg_wl("cancel WL, a list is empty: free %d, used %d", | |
5abde384 | 773 | !ubi->free.rb_node, !ubi->used.rb_node); |
43f9b25a | 774 | goto out_cancel; |
801c135c AB |
775 | } |
776 | ||
5abde384 | 777 | if (!ubi->scrub.rb_node) { |
801c135c AB |
778 | /* |
779 | * Now pick the least worn-out used physical eraseblock and a | |
780 | * highly worn-out free physical eraseblock. If the erase | |
781 | * counters differ much enough, start wear-leveling. | |
782 | */ | |
783 | e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, rb); | |
784 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); | |
785 | ||
786 | if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD)) { | |
787 | dbg_wl("no WL needed: min used EC %d, max free EC %d", | |
788 | e1->ec, e2->ec); | |
43f9b25a | 789 | goto out_cancel; |
801c135c | 790 | } |
5abde384 AB |
791 | paranoid_check_in_wl_tree(e1, &ubi->used); |
792 | rb_erase(&e1->rb, &ubi->used); | |
801c135c AB |
793 | dbg_wl("move PEB %d EC %d to PEB %d EC %d", |
794 | e1->pnum, e1->ec, e2->pnum, e2->ec); | |
795 | } else { | |
43f9b25a AB |
796 | /* Perform scrubbing */ |
797 | scrubbing = 1; | |
801c135c AB |
798 | e1 = rb_entry(rb_first(&ubi->scrub), struct ubi_wl_entry, rb); |
799 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); | |
5abde384 | 800 | paranoid_check_in_wl_tree(e1, &ubi->scrub); |
d2c46855 | 801 | rb_erase(&e1->rb, &ubi->scrub); |
801c135c AB |
802 | dbg_wl("scrub PEB %d to PEB %d", e1->pnum, e2->pnum); |
803 | } | |
804 | ||
5abde384 AB |
805 | paranoid_check_in_wl_tree(e2, &ubi->free); |
806 | rb_erase(&e2->rb, &ubi->free); | |
801c135c AB |
807 | ubi->move_from = e1; |
808 | ubi->move_to = e2; | |
809 | spin_unlock(&ubi->wl_lock); | |
810 | ||
811 | /* | |
812 | * Now we are going to copy physical eraseblock @e1->pnum to @e2->pnum. | |
813 | * We so far do not know which logical eraseblock our physical | |
814 | * eraseblock (@e1) belongs to. We have to read the volume identifier | |
815 | * header first. | |
43f9b25a AB |
816 | * |
817 | * Note, we are protected from this PEB being unmapped and erased. The | |
818 | * 'ubi_wl_put_peb()' would wait for moving to be finished if the PEB | |
819 | * which is being moved was unmapped. | |
801c135c AB |
820 | */ |
821 | ||
822 | err = ubi_io_read_vid_hdr(ubi, e1->pnum, vid_hdr, 0); | |
823 | if (err && err != UBI_IO_BITFLIPS) { | |
824 | if (err == UBI_IO_PEB_FREE) { | |
825 | /* | |
826 | * We are trying to move PEB without a VID header. UBI | |
827 | * always write VID headers shortly after the PEB was | |
828 | * given, so we have a situation when it did not have | |
829 | * chance to write it down because it was preempted. | |
830 | * Just re-schedule the work, so that next time it will | |
831 | * likely have the VID header in place. | |
832 | */ | |
833 | dbg_wl("PEB %d has no VID header", e1->pnum); | |
43f9b25a | 834 | goto out_not_moved; |
801c135c | 835 | } |
43f9b25a AB |
836 | |
837 | ubi_err("error %d while reading VID header from PEB %d", | |
838 | err, e1->pnum); | |
839 | if (err > 0) | |
840 | err = -EIO; | |
841 | goto out_error; | |
801c135c AB |
842 | } |
843 | ||
844 | err = ubi_eba_copy_leb(ubi, e1->pnum, e2->pnum, vid_hdr); | |
845 | if (err) { | |
43f9b25a AB |
846 | |
847 | if (err < 0) | |
848 | goto out_error; | |
849 | if (err == 1) | |
850 | goto out_not_moved; | |
851 | ||
852 | /* | |
853 | * For some reason the LEB was not moved - it might be because | |
854 | * the volume is being deleted. We should prevent this PEB from | |
855 | * being selected for wear-levelling movement for some "time", | |
856 | * so put it to the protection tree. | |
857 | */ | |
858 | ||
859 | dbg_wl("cancelled moving PEB %d", e1->pnum); | |
860 | pe = kmalloc(sizeof(struct ubi_wl_prot_entry), GFP_NOFS); | |
861 | if (!pe) { | |
862 | err = -ENOMEM; | |
863 | goto out_error; | |
864 | } | |
865 | ||
6a8f483f AB |
866 | ubi_free_vid_hdr(ubi, vid_hdr); |
867 | spin_lock(&ubi->wl_lock); | |
868 | prot_tree_add(ubi, e1, pe, U_PROTECTION); | |
869 | ubi_assert(!ubi->move_to_put); | |
870 | ubi->move_from = ubi->move_to = NULL; | |
871 | ubi->wl_scheduled = 0; | |
872 | spin_unlock(&ubi->wl_lock); | |
873 | ||
874 | err = schedule_erase(ubi, e2, 0); | |
875 | if (err) | |
876 | goto out_error; | |
877 | mutex_unlock(&ubi->move_mutex); | |
878 | return 0; | |
801c135c AB |
879 | } |
880 | ||
6a8f483f | 881 | /* The PEB has been successfully moved */ |
801c135c | 882 | ubi_free_vid_hdr(ubi, vid_hdr); |
6a8f483f | 883 | if (scrubbing) |
8c1e6ee1 AB |
884 | ubi_msg("scrubbed PEB %d, data moved to PEB %d", |
885 | e1->pnum, e2->pnum); | |
886 | ||
801c135c AB |
887 | spin_lock(&ubi->wl_lock); |
888 | if (!ubi->move_to_put) | |
5abde384 | 889 | wl_tree_add(e2, &ubi->used); |
801c135c AB |
890 | else |
891 | put = 1; | |
892 | ubi->move_from = ubi->move_to = NULL; | |
43f9b25a | 893 | ubi->move_to_put = ubi->wl_scheduled = 0; |
801c135c AB |
894 | spin_unlock(&ubi->wl_lock); |
895 | ||
6a8f483f AB |
896 | err = schedule_erase(ubi, e1, 0); |
897 | if (err) | |
898 | goto out_error; | |
899 | ||
801c135c AB |
900 | if (put) { |
901 | /* | |
902 | * Well, the target PEB was put meanwhile, schedule it for | |
903 | * erasure. | |
904 | */ | |
905 | dbg_wl("PEB %d was put meanwhile, erase", e2->pnum); | |
906 | err = schedule_erase(ubi, e2, 0); | |
43f9b25a AB |
907 | if (err) |
908 | goto out_error; | |
801c135c AB |
909 | } |
910 | ||
801c135c | 911 | dbg_wl("done"); |
43f9b25a AB |
912 | mutex_unlock(&ubi->move_mutex); |
913 | return 0; | |
801c135c AB |
914 | |
915 | /* | |
43f9b25a AB |
916 | * For some reasons the LEB was not moved, might be an error, might be |
917 | * something else. @e1 was not changed, so return it back. @e2 might | |
918 | * be changed, schedule it for erasure. | |
801c135c | 919 | */ |
43f9b25a | 920 | out_not_moved: |
801c135c AB |
921 | ubi_free_vid_hdr(ubi, vid_hdr); |
922 | spin_lock(&ubi->wl_lock); | |
43f9b25a AB |
923 | if (scrubbing) |
924 | wl_tree_add(e1, &ubi->scrub); | |
801c135c | 925 | else |
5abde384 | 926 | wl_tree_add(e1, &ubi->used); |
801c135c | 927 | ubi->move_from = ubi->move_to = NULL; |
43f9b25a | 928 | ubi->move_to_put = ubi->wl_scheduled = 0; |
801c135c AB |
929 | spin_unlock(&ubi->wl_lock); |
930 | ||
801c135c | 931 | err = schedule_erase(ubi, e2, 0); |
43f9b25a AB |
932 | if (err) |
933 | goto out_error; | |
934 | ||
935 | mutex_unlock(&ubi->move_mutex); | |
936 | return 0; | |
937 | ||
938 | out_error: | |
939 | ubi_err("error %d while moving PEB %d to PEB %d", | |
940 | err, e1->pnum, e2->pnum); | |
801c135c | 941 | |
43f9b25a AB |
942 | ubi_free_vid_hdr(ubi, vid_hdr); |
943 | spin_lock(&ubi->wl_lock); | |
944 | ubi->move_from = ubi->move_to = NULL; | |
945 | ubi->move_to_put = ubi->wl_scheduled = 0; | |
946 | spin_unlock(&ubi->wl_lock); | |
947 | ||
948 | kmem_cache_free(ubi_wl_entry_slab, e1); | |
949 | kmem_cache_free(ubi_wl_entry_slab, e2); | |
950 | ubi_ro_mode(ubi); | |
951 | ||
952 | mutex_unlock(&ubi->move_mutex); | |
801c135c | 953 | return err; |
43f9b25a AB |
954 | |
955 | out_cancel: | |
956 | ubi->wl_scheduled = 0; | |
957 | spin_unlock(&ubi->wl_lock); | |
958 | mutex_unlock(&ubi->move_mutex); | |
959 | ubi_free_vid_hdr(ubi, vid_hdr); | |
960 | return 0; | |
801c135c AB |
961 | } |
962 | ||
963 | /** | |
964 | * ensure_wear_leveling - schedule wear-leveling if it is needed. | |
965 | * @ubi: UBI device description object | |
966 | * | |
967 | * This function checks if it is time to start wear-leveling and schedules it | |
968 | * if yes. This function returns zero in case of success and a negative error | |
969 | * code in case of failure. | |
970 | */ | |
971 | static int ensure_wear_leveling(struct ubi_device *ubi) | |
972 | { | |
973 | int err = 0; | |
974 | struct ubi_wl_entry *e1; | |
975 | struct ubi_wl_entry *e2; | |
976 | struct ubi_work *wrk; | |
977 | ||
978 | spin_lock(&ubi->wl_lock); | |
979 | if (ubi->wl_scheduled) | |
980 | /* Wear-leveling is already in the work queue */ | |
981 | goto out_unlock; | |
982 | ||
983 | /* | |
984 | * If the ubi->scrub tree is not empty, scrubbing is needed, and the | |
985 | * the WL worker has to be scheduled anyway. | |
986 | */ | |
5abde384 AB |
987 | if (!ubi->scrub.rb_node) { |
988 | if (!ubi->used.rb_node || !ubi->free.rb_node) | |
801c135c AB |
989 | /* No physical eraseblocks - no deal */ |
990 | goto out_unlock; | |
991 | ||
992 | /* | |
993 | * We schedule wear-leveling only if the difference between the | |
994 | * lowest erase counter of used physical eraseblocks and a high | |
995 | * erase counter of free physical eraseblocks is greater then | |
996 | * %UBI_WL_THRESHOLD. | |
997 | */ | |
998 | e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, rb); | |
999 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); | |
1000 | ||
1001 | if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD)) | |
1002 | goto out_unlock; | |
1003 | dbg_wl("schedule wear-leveling"); | |
1004 | } else | |
1005 | dbg_wl("schedule scrubbing"); | |
1006 | ||
1007 | ubi->wl_scheduled = 1; | |
1008 | spin_unlock(&ubi->wl_lock); | |
1009 | ||
33818bbb | 1010 | wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS); |
801c135c AB |
1011 | if (!wrk) { |
1012 | err = -ENOMEM; | |
1013 | goto out_cancel; | |
1014 | } | |
1015 | ||
1016 | wrk->func = &wear_leveling_worker; | |
1017 | schedule_ubi_work(ubi, wrk); | |
1018 | return err; | |
1019 | ||
1020 | out_cancel: | |
1021 | spin_lock(&ubi->wl_lock); | |
1022 | ubi->wl_scheduled = 0; | |
1023 | out_unlock: | |
1024 | spin_unlock(&ubi->wl_lock); | |
1025 | return err; | |
1026 | } | |
1027 | ||
1028 | /** | |
1029 | * erase_worker - physical eraseblock erase worker function. | |
1030 | * @ubi: UBI device description object | |
1031 | * @wl_wrk: the work object | |
1032 | * @cancel: non-zero if the worker has to free memory and exit | |
1033 | * | |
1034 | * This function erases a physical eraseblock and perform torture testing if | |
1035 | * needed. It also takes care about marking the physical eraseblock bad if | |
1036 | * needed. Returns zero in case of success and a negative error code in case of | |
1037 | * failure. | |
1038 | */ | |
1039 | static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk, | |
1040 | int cancel) | |
1041 | { | |
801c135c | 1042 | struct ubi_wl_entry *e = wl_wrk->e; |
784c1454 | 1043 | int pnum = e->pnum, err, need; |
801c135c AB |
1044 | |
1045 | if (cancel) { | |
1046 | dbg_wl("cancel erasure of PEB %d EC %d", pnum, e->ec); | |
1047 | kfree(wl_wrk); | |
06b68ba1 | 1048 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1049 | return 0; |
1050 | } | |
1051 | ||
1052 | dbg_wl("erase PEB %d EC %d", pnum, e->ec); | |
1053 | ||
1054 | err = sync_erase(ubi, e, wl_wrk->torture); | |
1055 | if (!err) { | |
1056 | /* Fine, we've erased it successfully */ | |
1057 | kfree(wl_wrk); | |
1058 | ||
1059 | spin_lock(&ubi->wl_lock); | |
1060 | ubi->abs_ec += 1; | |
5abde384 | 1061 | wl_tree_add(e, &ubi->free); |
801c135c AB |
1062 | spin_unlock(&ubi->wl_lock); |
1063 | ||
1064 | /* | |
9c9ec147 AB |
1065 | * One more erase operation has happened, take care about |
1066 | * protected physical eraseblocks. | |
801c135c AB |
1067 | */ |
1068 | check_protection_over(ubi); | |
1069 | ||
1070 | /* And take care about wear-leveling */ | |
1071 | err = ensure_wear_leveling(ubi); | |
1072 | return err; | |
1073 | } | |
1074 | ||
8d2d4011 | 1075 | ubi_err("failed to erase PEB %d, error %d", pnum, err); |
801c135c | 1076 | kfree(wl_wrk); |
06b68ba1 | 1077 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c | 1078 | |
784c1454 AB |
1079 | if (err == -EINTR || err == -ENOMEM || err == -EAGAIN || |
1080 | err == -EBUSY) { | |
1081 | int err1; | |
1082 | ||
1083 | /* Re-schedule the LEB for erasure */ | |
1084 | err1 = schedule_erase(ubi, e, 0); | |
1085 | if (err1) { | |
1086 | err = err1; | |
1087 | goto out_ro; | |
1088 | } | |
1089 | return err; | |
1090 | } else if (err != -EIO) { | |
801c135c AB |
1091 | /* |
1092 | * If this is not %-EIO, we have no idea what to do. Scheduling | |
1093 | * this physical eraseblock for erasure again would cause | |
1094 | * errors again and again. Well, lets switch to RO mode. | |
1095 | */ | |
784c1454 | 1096 | goto out_ro; |
801c135c AB |
1097 | } |
1098 | ||
1099 | /* It is %-EIO, the PEB went bad */ | |
1100 | ||
1101 | if (!ubi->bad_allowed) { | |
1102 | ubi_err("bad physical eraseblock %d detected", pnum); | |
784c1454 AB |
1103 | goto out_ro; |
1104 | } | |
801c135c | 1105 | |
784c1454 AB |
1106 | spin_lock(&ubi->volumes_lock); |
1107 | need = ubi->beb_rsvd_level - ubi->beb_rsvd_pebs + 1; | |
1108 | if (need > 0) { | |
1109 | need = ubi->avail_pebs >= need ? need : ubi->avail_pebs; | |
1110 | ubi->avail_pebs -= need; | |
1111 | ubi->rsvd_pebs += need; | |
1112 | ubi->beb_rsvd_pebs += need; | |
1113 | if (need > 0) | |
1114 | ubi_msg("reserve more %d PEBs", need); | |
1115 | } | |
801c135c | 1116 | |
784c1454 | 1117 | if (ubi->beb_rsvd_pebs == 0) { |
801c135c | 1118 | spin_unlock(&ubi->volumes_lock); |
784c1454 AB |
1119 | ubi_err("no reserved physical eraseblocks"); |
1120 | goto out_ro; | |
1121 | } | |
801c135c | 1122 | |
784c1454 AB |
1123 | spin_unlock(&ubi->volumes_lock); |
1124 | ubi_msg("mark PEB %d as bad", pnum); | |
801c135c | 1125 | |
784c1454 AB |
1126 | err = ubi_io_mark_bad(ubi, pnum); |
1127 | if (err) | |
1128 | goto out_ro; | |
1129 | ||
1130 | spin_lock(&ubi->volumes_lock); | |
1131 | ubi->beb_rsvd_pebs -= 1; | |
1132 | ubi->bad_peb_count += 1; | |
1133 | ubi->good_peb_count -= 1; | |
1134 | ubi_calculate_reserved(ubi); | |
1135 | if (ubi->beb_rsvd_pebs == 0) | |
1136 | ubi_warn("last PEB from the reserved pool was used"); | |
1137 | spin_unlock(&ubi->volumes_lock); | |
1138 | ||
1139 | return err; | |
801c135c | 1140 | |
784c1454 AB |
1141 | out_ro: |
1142 | ubi_ro_mode(ubi); | |
801c135c AB |
1143 | return err; |
1144 | } | |
1145 | ||
1146 | /** | |
85c6e6e2 | 1147 | * ubi_wl_put_peb - return a PEB to the wear-leveling sub-system. |
801c135c AB |
1148 | * @ubi: UBI device description object |
1149 | * @pnum: physical eraseblock to return | |
1150 | * @torture: if this physical eraseblock has to be tortured | |
1151 | * | |
1152 | * This function is called to return physical eraseblock @pnum to the pool of | |
1153 | * free physical eraseblocks. The @torture flag has to be set if an I/O error | |
1154 | * occurred to this @pnum and it has to be tested. This function returns zero | |
43f9b25a | 1155 | * in case of success, and a negative error code in case of failure. |
801c135c AB |
1156 | */ |
1157 | int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture) | |
1158 | { | |
1159 | int err; | |
1160 | struct ubi_wl_entry *e; | |
1161 | ||
1162 | dbg_wl("PEB %d", pnum); | |
1163 | ubi_assert(pnum >= 0); | |
1164 | ubi_assert(pnum < ubi->peb_count); | |
1165 | ||
43f9b25a | 1166 | retry: |
801c135c | 1167 | spin_lock(&ubi->wl_lock); |
801c135c AB |
1168 | e = ubi->lookuptbl[pnum]; |
1169 | if (e == ubi->move_from) { | |
1170 | /* | |
1171 | * User is putting the physical eraseblock which was selected to | |
1172 | * be moved. It will be scheduled for erasure in the | |
1173 | * wear-leveling worker. | |
1174 | */ | |
43f9b25a | 1175 | dbg_wl("PEB %d is being moved, wait", pnum); |
801c135c | 1176 | spin_unlock(&ubi->wl_lock); |
43f9b25a AB |
1177 | |
1178 | /* Wait for the WL worker by taking the @ubi->move_mutex */ | |
1179 | mutex_lock(&ubi->move_mutex); | |
1180 | mutex_unlock(&ubi->move_mutex); | |
1181 | goto retry; | |
801c135c AB |
1182 | } else if (e == ubi->move_to) { |
1183 | /* | |
1184 | * User is putting the physical eraseblock which was selected | |
1185 | * as the target the data is moved to. It may happen if the EBA | |
85c6e6e2 AB |
1186 | * sub-system already re-mapped the LEB in 'ubi_eba_copy_leb()' |
1187 | * but the WL sub-system has not put the PEB to the "used" tree | |
1188 | * yet, but it is about to do this. So we just set a flag which | |
1189 | * will tell the WL worker that the PEB is not needed anymore | |
1190 | * and should be scheduled for erasure. | |
801c135c AB |
1191 | */ |
1192 | dbg_wl("PEB %d is the target of data moving", pnum); | |
1193 | ubi_assert(!ubi->move_to_put); | |
1194 | ubi->move_to_put = 1; | |
1195 | spin_unlock(&ubi->wl_lock); | |
1196 | return 0; | |
1197 | } else { | |
5abde384 AB |
1198 | if (in_wl_tree(e, &ubi->used)) { |
1199 | paranoid_check_in_wl_tree(e, &ubi->used); | |
1200 | rb_erase(&e->rb, &ubi->used); | |
1201 | } else if (in_wl_tree(e, &ubi->scrub)) { | |
1202 | paranoid_check_in_wl_tree(e, &ubi->scrub); | |
1203 | rb_erase(&e->rb, &ubi->scrub); | |
43f9b25a AB |
1204 | } else { |
1205 | err = prot_tree_del(ubi, e->pnum); | |
1206 | if (err) { | |
1207 | ubi_err("PEB %d not found", pnum); | |
1208 | ubi_ro_mode(ubi); | |
1209 | spin_unlock(&ubi->wl_lock); | |
1210 | return err; | |
1211 | } | |
1212 | } | |
801c135c AB |
1213 | } |
1214 | spin_unlock(&ubi->wl_lock); | |
1215 | ||
1216 | err = schedule_erase(ubi, e, torture); | |
1217 | if (err) { | |
1218 | spin_lock(&ubi->wl_lock); | |
5abde384 | 1219 | wl_tree_add(e, &ubi->used); |
801c135c AB |
1220 | spin_unlock(&ubi->wl_lock); |
1221 | } | |
1222 | ||
1223 | return err; | |
1224 | } | |
1225 | ||
1226 | /** | |
1227 | * ubi_wl_scrub_peb - schedule a physical eraseblock for scrubbing. | |
1228 | * @ubi: UBI device description object | |
1229 | * @pnum: the physical eraseblock to schedule | |
1230 | * | |
1231 | * If a bit-flip in a physical eraseblock is detected, this physical eraseblock | |
1232 | * needs scrubbing. This function schedules a physical eraseblock for | |
1233 | * scrubbing which is done in background. This function returns zero in case of | |
1234 | * success and a negative error code in case of failure. | |
1235 | */ | |
1236 | int ubi_wl_scrub_peb(struct ubi_device *ubi, int pnum) | |
1237 | { | |
1238 | struct ubi_wl_entry *e; | |
1239 | ||
8c1e6ee1 | 1240 | dbg_msg("schedule PEB %d for scrubbing", pnum); |
801c135c AB |
1241 | |
1242 | retry: | |
1243 | spin_lock(&ubi->wl_lock); | |
1244 | e = ubi->lookuptbl[pnum]; | |
1245 | if (e == ubi->move_from || in_wl_tree(e, &ubi->scrub)) { | |
1246 | spin_unlock(&ubi->wl_lock); | |
1247 | return 0; | |
1248 | } | |
1249 | ||
1250 | if (e == ubi->move_to) { | |
1251 | /* | |
1252 | * This physical eraseblock was used to move data to. The data | |
1253 | * was moved but the PEB was not yet inserted to the proper | |
1254 | * tree. We should just wait a little and let the WL worker | |
1255 | * proceed. | |
1256 | */ | |
1257 | spin_unlock(&ubi->wl_lock); | |
1258 | dbg_wl("the PEB %d is not in proper tree, retry", pnum); | |
1259 | yield(); | |
1260 | goto retry; | |
1261 | } | |
1262 | ||
5abde384 AB |
1263 | if (in_wl_tree(e, &ubi->used)) { |
1264 | paranoid_check_in_wl_tree(e, &ubi->used); | |
1265 | rb_erase(&e->rb, &ubi->used); | |
43f9b25a AB |
1266 | } else { |
1267 | int err; | |
1268 | ||
1269 | err = prot_tree_del(ubi, e->pnum); | |
1270 | if (err) { | |
1271 | ubi_err("PEB %d not found", pnum); | |
1272 | ubi_ro_mode(ubi); | |
1273 | spin_unlock(&ubi->wl_lock); | |
1274 | return err; | |
1275 | } | |
1276 | } | |
801c135c | 1277 | |
5abde384 | 1278 | wl_tree_add(e, &ubi->scrub); |
801c135c AB |
1279 | spin_unlock(&ubi->wl_lock); |
1280 | ||
1281 | /* | |
1282 | * Technically scrubbing is the same as wear-leveling, so it is done | |
1283 | * by the WL worker. | |
1284 | */ | |
1285 | return ensure_wear_leveling(ubi); | |
1286 | } | |
1287 | ||
1288 | /** | |
1289 | * ubi_wl_flush - flush all pending works. | |
1290 | * @ubi: UBI device description object | |
1291 | * | |
1292 | * This function returns zero in case of success and a negative error code in | |
1293 | * case of failure. | |
1294 | */ | |
1295 | int ubi_wl_flush(struct ubi_device *ubi) | |
1296 | { | |
593dd33c | 1297 | int err; |
801c135c AB |
1298 | |
1299 | /* | |
1300 | * Erase while the pending works queue is not empty, but not more then | |
1301 | * the number of currently pending works. | |
1302 | */ | |
593dd33c AB |
1303 | dbg_wl("flush (%d pending works)", ubi->works_count); |
1304 | while (ubi->works_count) { | |
1305 | err = do_work(ubi); | |
1306 | if (err) | |
1307 | return err; | |
1308 | } | |
1309 | ||
1310 | /* | |
1311 | * Make sure all the works which have been done in parallel are | |
1312 | * finished. | |
1313 | */ | |
1314 | down_write(&ubi->work_sem); | |
1315 | up_write(&ubi->work_sem); | |
1316 | ||
1317 | /* | |
1318 | * And in case last was the WL worker and it cancelled the LEB | |
1319 | * movement, flush again. | |
1320 | */ | |
1321 | while (ubi->works_count) { | |
1322 | dbg_wl("flush more (%d pending works)", ubi->works_count); | |
801c135c AB |
1323 | err = do_work(ubi); |
1324 | if (err) | |
1325 | return err; | |
1326 | } | |
1327 | ||
1328 | return 0; | |
1329 | } | |
1330 | ||
1331 | /** | |
1332 | * tree_destroy - destroy an RB-tree. | |
1333 | * @root: the root of the tree to destroy | |
1334 | */ | |
1335 | static void tree_destroy(struct rb_root *root) | |
1336 | { | |
1337 | struct rb_node *rb; | |
1338 | struct ubi_wl_entry *e; | |
1339 | ||
1340 | rb = root->rb_node; | |
1341 | while (rb) { | |
1342 | if (rb->rb_left) | |
1343 | rb = rb->rb_left; | |
1344 | else if (rb->rb_right) | |
1345 | rb = rb->rb_right; | |
1346 | else { | |
1347 | e = rb_entry(rb, struct ubi_wl_entry, rb); | |
1348 | ||
1349 | rb = rb_parent(rb); | |
1350 | if (rb) { | |
1351 | if (rb->rb_left == &e->rb) | |
1352 | rb->rb_left = NULL; | |
1353 | else | |
1354 | rb->rb_right = NULL; | |
1355 | } | |
1356 | ||
06b68ba1 | 1357 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1358 | } |
1359 | } | |
1360 | } | |
1361 | ||
1362 | /** | |
1363 | * ubi_thread - UBI background thread. | |
1364 | * @u: the UBI device description object pointer | |
1365 | */ | |
cdfa788a | 1366 | int ubi_thread(void *u) |
801c135c AB |
1367 | { |
1368 | int failures = 0; | |
1369 | struct ubi_device *ubi = u; | |
1370 | ||
1371 | ubi_msg("background thread \"%s\" started, PID %d", | |
ba25f9dc | 1372 | ubi->bgt_name, task_pid_nr(current)); |
801c135c | 1373 | |
83144186 | 1374 | set_freezable(); |
801c135c AB |
1375 | for (;;) { |
1376 | int err; | |
1377 | ||
1378 | if (kthread_should_stop()) | |
cadb40cc | 1379 | break; |
801c135c AB |
1380 | |
1381 | if (try_to_freeze()) | |
1382 | continue; | |
1383 | ||
1384 | spin_lock(&ubi->wl_lock); | |
1385 | if (list_empty(&ubi->works) || ubi->ro_mode || | |
1386 | !ubi->thread_enabled) { | |
1387 | set_current_state(TASK_INTERRUPTIBLE); | |
1388 | spin_unlock(&ubi->wl_lock); | |
1389 | schedule(); | |
1390 | continue; | |
1391 | } | |
1392 | spin_unlock(&ubi->wl_lock); | |
1393 | ||
1394 | err = do_work(ubi); | |
1395 | if (err) { | |
1396 | ubi_err("%s: work failed with error code %d", | |
1397 | ubi->bgt_name, err); | |
1398 | if (failures++ > WL_MAX_FAILURES) { | |
1399 | /* | |
1400 | * Too many failures, disable the thread and | |
1401 | * switch to read-only mode. | |
1402 | */ | |
1403 | ubi_msg("%s: %d consecutive failures", | |
1404 | ubi->bgt_name, WL_MAX_FAILURES); | |
1405 | ubi_ro_mode(ubi); | |
2ad49887 VG |
1406 | ubi->thread_enabled = 0; |
1407 | continue; | |
801c135c AB |
1408 | } |
1409 | } else | |
1410 | failures = 0; | |
1411 | ||
1412 | cond_resched(); | |
1413 | } | |
1414 | ||
801c135c AB |
1415 | dbg_wl("background thread \"%s\" is killed", ubi->bgt_name); |
1416 | return 0; | |
1417 | } | |
1418 | ||
1419 | /** | |
1420 | * cancel_pending - cancel all pending works. | |
1421 | * @ubi: UBI device description object | |
1422 | */ | |
1423 | static void cancel_pending(struct ubi_device *ubi) | |
1424 | { | |
1425 | while (!list_empty(&ubi->works)) { | |
1426 | struct ubi_work *wrk; | |
1427 | ||
1428 | wrk = list_entry(ubi->works.next, struct ubi_work, list); | |
1429 | list_del(&wrk->list); | |
1430 | wrk->func(ubi, wrk, 1); | |
1431 | ubi->works_count -= 1; | |
1432 | ubi_assert(ubi->works_count >= 0); | |
1433 | } | |
1434 | } | |
1435 | ||
1436 | /** | |
85c6e6e2 | 1437 | * ubi_wl_init_scan - initialize the WL sub-system using scanning information. |
801c135c AB |
1438 | * @ubi: UBI device description object |
1439 | * @si: scanning information | |
1440 | * | |
1441 | * This function returns zero in case of success, and a negative error code in | |
1442 | * case of failure. | |
1443 | */ | |
1444 | int ubi_wl_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si) | |
1445 | { | |
1446 | int err; | |
1447 | struct rb_node *rb1, *rb2; | |
1448 | struct ubi_scan_volume *sv; | |
1449 | struct ubi_scan_leb *seb, *tmp; | |
1450 | struct ubi_wl_entry *e; | |
1451 | ||
1452 | ||
1453 | ubi->used = ubi->free = ubi->scrub = RB_ROOT; | |
1454 | ubi->prot.pnum = ubi->prot.aec = RB_ROOT; | |
1455 | spin_lock_init(&ubi->wl_lock); | |
43f9b25a | 1456 | mutex_init(&ubi->move_mutex); |
593dd33c | 1457 | init_rwsem(&ubi->work_sem); |
801c135c AB |
1458 | ubi->max_ec = si->max_ec; |
1459 | INIT_LIST_HEAD(&ubi->works); | |
1460 | ||
1461 | sprintf(ubi->bgt_name, UBI_BGT_NAME_PATTERN, ubi->ubi_num); | |
1462 | ||
801c135c AB |
1463 | err = -ENOMEM; |
1464 | ubi->lookuptbl = kzalloc(ubi->peb_count * sizeof(void *), GFP_KERNEL); | |
1465 | if (!ubi->lookuptbl) | |
cdfa788a | 1466 | return err; |
801c135c AB |
1467 | |
1468 | list_for_each_entry_safe(seb, tmp, &si->erase, u.list) { | |
1469 | cond_resched(); | |
1470 | ||
06b68ba1 | 1471 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1472 | if (!e) |
1473 | goto out_free; | |
1474 | ||
1475 | e->pnum = seb->pnum; | |
1476 | e->ec = seb->ec; | |
1477 | ubi->lookuptbl[e->pnum] = e; | |
1478 | if (schedule_erase(ubi, e, 0)) { | |
06b68ba1 | 1479 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1480 | goto out_free; |
1481 | } | |
1482 | } | |
1483 | ||
1484 | list_for_each_entry(seb, &si->free, u.list) { | |
1485 | cond_resched(); | |
1486 | ||
06b68ba1 | 1487 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1488 | if (!e) |
1489 | goto out_free; | |
1490 | ||
1491 | e->pnum = seb->pnum; | |
1492 | e->ec = seb->ec; | |
1493 | ubi_assert(e->ec >= 0); | |
5abde384 | 1494 | wl_tree_add(e, &ubi->free); |
801c135c AB |
1495 | ubi->lookuptbl[e->pnum] = e; |
1496 | } | |
1497 | ||
1498 | list_for_each_entry(seb, &si->corr, u.list) { | |
1499 | cond_resched(); | |
1500 | ||
06b68ba1 | 1501 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1502 | if (!e) |
1503 | goto out_free; | |
1504 | ||
1505 | e->pnum = seb->pnum; | |
1506 | e->ec = seb->ec; | |
1507 | ubi->lookuptbl[e->pnum] = e; | |
1508 | if (schedule_erase(ubi, e, 0)) { | |
06b68ba1 | 1509 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1510 | goto out_free; |
1511 | } | |
1512 | } | |
1513 | ||
1514 | ubi_rb_for_each_entry(rb1, sv, &si->volumes, rb) { | |
1515 | ubi_rb_for_each_entry(rb2, seb, &sv->root, u.rb) { | |
1516 | cond_resched(); | |
1517 | ||
06b68ba1 | 1518 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1519 | if (!e) |
1520 | goto out_free; | |
1521 | ||
1522 | e->pnum = seb->pnum; | |
1523 | e->ec = seb->ec; | |
1524 | ubi->lookuptbl[e->pnum] = e; | |
1525 | if (!seb->scrub) { | |
1526 | dbg_wl("add PEB %d EC %d to the used tree", | |
1527 | e->pnum, e->ec); | |
5abde384 | 1528 | wl_tree_add(e, &ubi->used); |
801c135c AB |
1529 | } else { |
1530 | dbg_wl("add PEB %d EC %d to the scrub tree", | |
1531 | e->pnum, e->ec); | |
5abde384 | 1532 | wl_tree_add(e, &ubi->scrub); |
801c135c AB |
1533 | } |
1534 | } | |
1535 | } | |
1536 | ||
5abde384 | 1537 | if (ubi->avail_pebs < WL_RESERVED_PEBS) { |
801c135c AB |
1538 | ubi_err("no enough physical eraseblocks (%d, need %d)", |
1539 | ubi->avail_pebs, WL_RESERVED_PEBS); | |
1540 | goto out_free; | |
1541 | } | |
1542 | ubi->avail_pebs -= WL_RESERVED_PEBS; | |
1543 | ubi->rsvd_pebs += WL_RESERVED_PEBS; | |
1544 | ||
1545 | /* Schedule wear-leveling if needed */ | |
1546 | err = ensure_wear_leveling(ubi); | |
1547 | if (err) | |
1548 | goto out_free; | |
1549 | ||
1550 | return 0; | |
1551 | ||
1552 | out_free: | |
1553 | cancel_pending(ubi); | |
1554 | tree_destroy(&ubi->used); | |
1555 | tree_destroy(&ubi->free); | |
1556 | tree_destroy(&ubi->scrub); | |
1557 | kfree(ubi->lookuptbl); | |
801c135c AB |
1558 | return err; |
1559 | } | |
1560 | ||
1561 | /** | |
1562 | * protection_trees_destroy - destroy the protection RB-trees. | |
1563 | * @ubi: UBI device description object | |
1564 | */ | |
1565 | static void protection_trees_destroy(struct ubi_device *ubi) | |
1566 | { | |
1567 | struct rb_node *rb; | |
1568 | struct ubi_wl_prot_entry *pe; | |
1569 | ||
1570 | rb = ubi->prot.aec.rb_node; | |
1571 | while (rb) { | |
1572 | if (rb->rb_left) | |
1573 | rb = rb->rb_left; | |
1574 | else if (rb->rb_right) | |
1575 | rb = rb->rb_right; | |
1576 | else { | |
1577 | pe = rb_entry(rb, struct ubi_wl_prot_entry, rb_aec); | |
1578 | ||
1579 | rb = rb_parent(rb); | |
1580 | if (rb) { | |
1581 | if (rb->rb_left == &pe->rb_aec) | |
1582 | rb->rb_left = NULL; | |
1583 | else | |
1584 | rb->rb_right = NULL; | |
1585 | } | |
1586 | ||
06b68ba1 | 1587 | kmem_cache_free(ubi_wl_entry_slab, pe->e); |
801c135c AB |
1588 | kfree(pe); |
1589 | } | |
1590 | } | |
1591 | } | |
1592 | ||
1593 | /** | |
85c6e6e2 | 1594 | * ubi_wl_close - close the wear-leveling sub-system. |
801c135c AB |
1595 | * @ubi: UBI device description object |
1596 | */ | |
1597 | void ubi_wl_close(struct ubi_device *ubi) | |
1598 | { | |
85c6e6e2 | 1599 | dbg_wl("close the WL sub-system"); |
801c135c AB |
1600 | cancel_pending(ubi); |
1601 | protection_trees_destroy(ubi); | |
1602 | tree_destroy(&ubi->used); | |
1603 | tree_destroy(&ubi->free); | |
1604 | tree_destroy(&ubi->scrub); | |
1605 | kfree(ubi->lookuptbl); | |
801c135c AB |
1606 | } |
1607 | ||
1608 | #ifdef CONFIG_MTD_UBI_DEBUG_PARANOID | |
1609 | ||
1610 | /** | |
ebaaf1af | 1611 | * paranoid_check_ec - make sure that the erase counter of a PEB is correct. |
801c135c AB |
1612 | * @ubi: UBI device description object |
1613 | * @pnum: the physical eraseblock number to check | |
1614 | * @ec: the erase counter to check | |
1615 | * | |
1616 | * This function returns zero if the erase counter of physical eraseblock @pnum | |
1617 | * is equivalent to @ec, %1 if not, and a negative error code if an error | |
1618 | * occurred. | |
1619 | */ | |
e88d6e10 | 1620 | static int paranoid_check_ec(struct ubi_device *ubi, int pnum, int ec) |
801c135c AB |
1621 | { |
1622 | int err; | |
1623 | long long read_ec; | |
1624 | struct ubi_ec_hdr *ec_hdr; | |
1625 | ||
33818bbb | 1626 | ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_NOFS); |
801c135c AB |
1627 | if (!ec_hdr) |
1628 | return -ENOMEM; | |
1629 | ||
1630 | err = ubi_io_read_ec_hdr(ubi, pnum, ec_hdr, 0); | |
1631 | if (err && err != UBI_IO_BITFLIPS) { | |
1632 | /* The header does not have to exist */ | |
1633 | err = 0; | |
1634 | goto out_free; | |
1635 | } | |
1636 | ||
3261ebd7 | 1637 | read_ec = be64_to_cpu(ec_hdr->ec); |
801c135c AB |
1638 | if (ec != read_ec) { |
1639 | ubi_err("paranoid check failed for PEB %d", pnum); | |
1640 | ubi_err("read EC is %lld, should be %d", read_ec, ec); | |
1641 | ubi_dbg_dump_stack(); | |
1642 | err = 1; | |
1643 | } else | |
1644 | err = 0; | |
1645 | ||
1646 | out_free: | |
1647 | kfree(ec_hdr); | |
1648 | return err; | |
1649 | } | |
1650 | ||
1651 | /** | |
ebaaf1af | 1652 | * paranoid_check_in_wl_tree - check that wear-leveling entry is in WL RB-tree. |
801c135c AB |
1653 | * @e: the wear-leveling entry to check |
1654 | * @root: the root of the tree | |
1655 | * | |
ebaaf1af AB |
1656 | * This function returns zero if @e is in the @root RB-tree and %1 if it is |
1657 | * not. | |
801c135c AB |
1658 | */ |
1659 | static int paranoid_check_in_wl_tree(struct ubi_wl_entry *e, | |
1660 | struct rb_root *root) | |
1661 | { | |
1662 | if (in_wl_tree(e, root)) | |
1663 | return 0; | |
1664 | ||
1665 | ubi_err("paranoid check failed for PEB %d, EC %d, RB-tree %p ", | |
1666 | e->pnum, e->ec, root); | |
1667 | ubi_dbg_dump_stack(); | |
1668 | return 1; | |
1669 | } | |
1670 | ||
1671 | #endif /* CONFIG_MTD_UBI_DEBUG_PARANOID */ |