]>
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 | 24 | * This sub-system is responsible for wear-leveling. It works in terms of |
7b6c32da | 25 | * physical eraseblocks and erase counters and knows nothing about logical |
85c6e6e2 AB |
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 | |
b86a2c56 AB |
58 | * used eraseblocks are kept in @wl->used, @wl->erroneous, or @wl->scrub |
59 | * RB-trees, as well as (temporarily) in the @wl->pq queue. | |
7b6c32da XX |
60 | * |
61 | * When the WL sub-system returns a physical eraseblock, the physical | |
62 | * eraseblock is protected from being moved for some "time". For this reason, | |
63 | * the physical eraseblock is not directly moved from the @wl->free tree to the | |
64 | * @wl->used tree. There is a protection queue in between where this | |
65 | * physical eraseblock is temporarily stored (@wl->pq). | |
66 | * | |
67 | * All this protection stuff is needed because: | |
68 | * o we don't want to move physical eraseblocks just after we have given them | |
69 | * to the user; instead, we first want to let users fill them up with data; | |
70 | * | |
71 | * o there is a chance that the user will put the physical eraseblock very | |
72 | * soon, so it makes sense not to move it for some time, but wait; this is | |
73 | * especially important in case of "short term" physical eraseblocks. | |
74 | * | |
75 | * Physical eraseblocks stay protected only for limited time. But the "time" is | |
76 | * measured in erase cycles in this case. This is implemented with help of the | |
77 | * protection queue. Eraseblocks are put to the tail of this queue when they | |
78 | * are returned by the 'ubi_wl_get_peb()', and eraseblocks are removed from the | |
79 | * head of the queue on each erase operation (for any eraseblock). So the | |
80 | * length of the queue defines how may (global) erase cycles PEBs are protected. | |
81 | * | |
82 | * To put it differently, each physical eraseblock has 2 main states: free and | |
83 | * used. The former state corresponds to the @wl->free tree. The latter state | |
84 | * is split up on several sub-states: | |
85 | * o the WL movement is allowed (@wl->used tree); | |
b86a2c56 AB |
86 | * o the WL movement is disallowed (@wl->erroneous) becouse the PEB is |
87 | * erroneous - e.g., there was a read error; | |
7b6c32da XX |
88 | * o the WL movement is temporarily prohibited (@wl->pq queue); |
89 | * o scrubbing is needed (@wl->scrub tree). | |
90 | * | |
91 | * Depending on the sub-state, wear-leveling entries of the used physical | |
92 | * eraseblocks may be kept in one of those structures. | |
801c135c AB |
93 | * |
94 | * Note, in this implementation, we keep a small in-RAM object for each physical | |
95 | * eraseblock. This is surely not a scalable solution. But it appears to be good | |
96 | * enough for moderately large flashes and it is simple. In future, one may | |
85c6e6e2 | 97 | * re-work this sub-system and make it more scalable. |
801c135c | 98 | * |
85c6e6e2 AB |
99 | * At the moment this sub-system does not utilize the sequence number, which |
100 | * was introduced relatively recently. But it would be wise to do this because | |
101 | * the sequence number of a logical eraseblock characterizes how old is it. For | |
801c135c AB |
102 | * example, when we move a PEB with low erase counter, and we need to pick the |
103 | * target PEB, we pick a PEB with the highest EC if our PEB is "old" and we | |
104 | * pick target PEB with an average EC if our PEB is not very "old". This is a | |
85c6e6e2 | 105 | * room for future re-works of the WL sub-system. |
801c135c AB |
106 | */ |
107 | ||
108 | #include <linux/slab.h> | |
109 | #include <linux/crc32.h> | |
110 | #include <linux/freezer.h> | |
111 | #include <linux/kthread.h> | |
112 | #include "ubi.h" | |
113 | ||
114 | /* Number of physical eraseblocks reserved for wear-leveling purposes */ | |
115 | #define WL_RESERVED_PEBS 1 | |
116 | ||
801c135c AB |
117 | /* |
118 | * Maximum difference between two erase counters. If this threshold is | |
85c6e6e2 AB |
119 | * exceeded, the WL sub-system starts moving data from used physical |
120 | * eraseblocks with low erase counter to free physical eraseblocks with high | |
121 | * erase counter. | |
801c135c AB |
122 | */ |
123 | #define UBI_WL_THRESHOLD CONFIG_MTD_UBI_WL_THRESHOLD | |
124 | ||
125 | /* | |
85c6e6e2 | 126 | * When a physical eraseblock is moved, the WL sub-system has to pick the target |
801c135c AB |
127 | * physical eraseblock to move to. The simplest way would be just to pick the |
128 | * one with the highest erase counter. But in certain workloads this could lead | |
129 | * to an unlimited wear of one or few physical eraseblock. Indeed, imagine a | |
130 | * situation when the picked physical eraseblock is constantly erased after the | |
131 | * data is written to it. So, we have a constant which limits the highest erase | |
85c6e6e2 | 132 | * counter of the free physical eraseblock to pick. Namely, the WL sub-system |
025dfdaf | 133 | * does not pick eraseblocks with erase counter greater than the lowest erase |
801c135c AB |
134 | * counter plus %WL_FREE_MAX_DIFF. |
135 | */ | |
136 | #define WL_FREE_MAX_DIFF (2*UBI_WL_THRESHOLD) | |
137 | ||
138 | /* | |
139 | * Maximum number of consecutive background thread failures which is enough to | |
140 | * switch to read-only mode. | |
141 | */ | |
142 | #define WL_MAX_FAILURES 32 | |
143 | ||
801c135c AB |
144 | /** |
145 | * struct ubi_work - UBI work description data structure. | |
146 | * @list: a link in the list of pending works | |
147 | * @func: worker function | |
801c135c AB |
148 | * @e: physical eraseblock to erase |
149 | * @torture: if the physical eraseblock has to be tortured | |
150 | * | |
151 | * The @func pointer points to the worker function. If the @cancel argument is | |
152 | * not zero, the worker has to free the resources and exit immediately. The | |
153 | * worker has to return zero in case of success and a negative error code in | |
154 | * case of failure. | |
155 | */ | |
156 | struct ubi_work { | |
157 | struct list_head list; | |
158 | int (*func)(struct ubi_device *ubi, struct ubi_work *wrk, int cancel); | |
159 | /* The below fields are only relevant to erasure works */ | |
160 | struct ubi_wl_entry *e; | |
161 | int torture; | |
162 | }; | |
163 | ||
164 | #ifdef CONFIG_MTD_UBI_DEBUG_PARANOID | |
e88d6e10 | 165 | static int paranoid_check_ec(struct ubi_device *ubi, int pnum, int ec); |
801c135c AB |
166 | static int paranoid_check_in_wl_tree(struct ubi_wl_entry *e, |
167 | struct rb_root *root); | |
7b6c32da | 168 | static int paranoid_check_in_pq(struct ubi_device *ubi, struct ubi_wl_entry *e); |
801c135c AB |
169 | #else |
170 | #define paranoid_check_ec(ubi, pnum, ec) 0 | |
171 | #define paranoid_check_in_wl_tree(e, root) | |
7b6c32da | 172 | #define paranoid_check_in_pq(ubi, e) 0 |
801c135c AB |
173 | #endif |
174 | ||
801c135c AB |
175 | /** |
176 | * wl_tree_add - add a wear-leveling entry to a WL RB-tree. | |
177 | * @e: the wear-leveling entry to add | |
178 | * @root: the root of the tree | |
179 | * | |
180 | * Note, we use (erase counter, physical eraseblock number) pairs as keys in | |
181 | * the @ubi->used and @ubi->free RB-trees. | |
182 | */ | |
183 | static void wl_tree_add(struct ubi_wl_entry *e, struct rb_root *root) | |
184 | { | |
185 | struct rb_node **p, *parent = NULL; | |
186 | ||
187 | p = &root->rb_node; | |
188 | while (*p) { | |
189 | struct ubi_wl_entry *e1; | |
190 | ||
191 | parent = *p; | |
23553b2c | 192 | e1 = rb_entry(parent, struct ubi_wl_entry, u.rb); |
801c135c AB |
193 | |
194 | if (e->ec < e1->ec) | |
195 | p = &(*p)->rb_left; | |
196 | else if (e->ec > e1->ec) | |
197 | p = &(*p)->rb_right; | |
198 | else { | |
199 | ubi_assert(e->pnum != e1->pnum); | |
200 | if (e->pnum < e1->pnum) | |
201 | p = &(*p)->rb_left; | |
202 | else | |
203 | p = &(*p)->rb_right; | |
204 | } | |
205 | } | |
206 | ||
23553b2c XX |
207 | rb_link_node(&e->u.rb, parent, p); |
208 | rb_insert_color(&e->u.rb, root); | |
801c135c AB |
209 | } |
210 | ||
801c135c AB |
211 | /** |
212 | * do_work - do one pending work. | |
213 | * @ubi: UBI device description object | |
214 | * | |
215 | * This function returns zero in case of success and a negative error code in | |
216 | * case of failure. | |
217 | */ | |
218 | static int do_work(struct ubi_device *ubi) | |
219 | { | |
220 | int err; | |
221 | struct ubi_work *wrk; | |
222 | ||
43f9b25a AB |
223 | cond_resched(); |
224 | ||
593dd33c AB |
225 | /* |
226 | * @ubi->work_sem is used to synchronize with the workers. Workers take | |
227 | * it in read mode, so many of them may be doing works at a time. But | |
228 | * the queue flush code has to be sure the whole queue of works is | |
229 | * done, and it takes the mutex in write mode. | |
230 | */ | |
231 | down_read(&ubi->work_sem); | |
801c135c | 232 | spin_lock(&ubi->wl_lock); |
801c135c AB |
233 | if (list_empty(&ubi->works)) { |
234 | spin_unlock(&ubi->wl_lock); | |
593dd33c | 235 | up_read(&ubi->work_sem); |
801c135c AB |
236 | return 0; |
237 | } | |
238 | ||
239 | wrk = list_entry(ubi->works.next, struct ubi_work, list); | |
240 | list_del(&wrk->list); | |
16f557ec AB |
241 | ubi->works_count -= 1; |
242 | ubi_assert(ubi->works_count >= 0); | |
801c135c AB |
243 | spin_unlock(&ubi->wl_lock); |
244 | ||
245 | /* | |
246 | * Call the worker function. Do not touch the work structure | |
247 | * after this call as it will have been freed or reused by that | |
248 | * time by the worker function. | |
249 | */ | |
250 | err = wrk->func(ubi, wrk, 0); | |
251 | if (err) | |
252 | ubi_err("work failed with error code %d", err); | |
593dd33c | 253 | up_read(&ubi->work_sem); |
16f557ec | 254 | |
801c135c AB |
255 | return err; |
256 | } | |
257 | ||
258 | /** | |
259 | * produce_free_peb - produce a free physical eraseblock. | |
260 | * @ubi: UBI device description object | |
261 | * | |
262 | * This function tries to make a free PEB by means of synchronous execution of | |
263 | * pending works. This may be needed if, for example the background thread is | |
264 | * disabled. Returns zero in case of success and a negative error code in case | |
265 | * of failure. | |
266 | */ | |
267 | static int produce_free_peb(struct ubi_device *ubi) | |
268 | { | |
269 | int err; | |
270 | ||
271 | spin_lock(&ubi->wl_lock); | |
5abde384 | 272 | while (!ubi->free.rb_node) { |
801c135c AB |
273 | spin_unlock(&ubi->wl_lock); |
274 | ||
275 | dbg_wl("do one work synchronously"); | |
276 | err = do_work(ubi); | |
277 | if (err) | |
278 | return err; | |
279 | ||
280 | spin_lock(&ubi->wl_lock); | |
281 | } | |
282 | spin_unlock(&ubi->wl_lock); | |
283 | ||
284 | return 0; | |
285 | } | |
286 | ||
287 | /** | |
288 | * in_wl_tree - check if wear-leveling entry is present in a WL RB-tree. | |
289 | * @e: the wear-leveling entry to check | |
290 | * @root: the root of the tree | |
291 | * | |
292 | * This function returns non-zero if @e is in the @root RB-tree and zero if it | |
293 | * is not. | |
294 | */ | |
295 | static int in_wl_tree(struct ubi_wl_entry *e, struct rb_root *root) | |
296 | { | |
297 | struct rb_node *p; | |
298 | ||
299 | p = root->rb_node; | |
300 | while (p) { | |
301 | struct ubi_wl_entry *e1; | |
302 | ||
23553b2c | 303 | e1 = rb_entry(p, struct ubi_wl_entry, u.rb); |
801c135c AB |
304 | |
305 | if (e->pnum == e1->pnum) { | |
306 | ubi_assert(e == e1); | |
307 | return 1; | |
308 | } | |
309 | ||
310 | if (e->ec < e1->ec) | |
311 | p = p->rb_left; | |
312 | else if (e->ec > e1->ec) | |
313 | p = p->rb_right; | |
314 | else { | |
315 | ubi_assert(e->pnum != e1->pnum); | |
316 | if (e->pnum < e1->pnum) | |
317 | p = p->rb_left; | |
318 | else | |
319 | p = p->rb_right; | |
320 | } | |
321 | } | |
322 | ||
323 | return 0; | |
324 | } | |
325 | ||
326 | /** | |
7b6c32da | 327 | * prot_queue_add - add physical eraseblock to the protection queue. |
801c135c AB |
328 | * @ubi: UBI device description object |
329 | * @e: the physical eraseblock to add | |
801c135c | 330 | * |
7b6c32da XX |
331 | * This function adds @e to the tail of the protection queue @ubi->pq, where |
332 | * @e will stay for %UBI_PROT_QUEUE_LEN erase operations and will be | |
333 | * temporarily protected from the wear-leveling worker. Note, @wl->lock has to | |
334 | * be locked. | |
801c135c | 335 | */ |
7b6c32da | 336 | static void prot_queue_add(struct ubi_device *ubi, struct ubi_wl_entry *e) |
801c135c | 337 | { |
7b6c32da | 338 | int pq_tail = ubi->pq_head - 1; |
801c135c | 339 | |
7b6c32da XX |
340 | if (pq_tail < 0) |
341 | pq_tail = UBI_PROT_QUEUE_LEN - 1; | |
342 | ubi_assert(pq_tail >= 0 && pq_tail < UBI_PROT_QUEUE_LEN); | |
343 | list_add_tail(&e->u.list, &ubi->pq[pq_tail]); | |
344 | dbg_wl("added PEB %d EC %d to the protection queue", e->pnum, e->ec); | |
801c135c AB |
345 | } |
346 | ||
347 | /** | |
348 | * find_wl_entry - find wear-leveling entry closest to certain erase counter. | |
349 | * @root: the RB-tree where to look for | |
350 | * @max: highest possible erase counter | |
351 | * | |
352 | * This function looks for a wear leveling entry with erase counter closest to | |
353 | * @max and less then @max. | |
354 | */ | |
355 | static struct ubi_wl_entry *find_wl_entry(struct rb_root *root, int max) | |
356 | { | |
357 | struct rb_node *p; | |
358 | struct ubi_wl_entry *e; | |
359 | ||
23553b2c | 360 | e = rb_entry(rb_first(root), struct ubi_wl_entry, u.rb); |
801c135c AB |
361 | max += e->ec; |
362 | ||
363 | p = root->rb_node; | |
364 | while (p) { | |
365 | struct ubi_wl_entry *e1; | |
366 | ||
23553b2c | 367 | e1 = rb_entry(p, struct ubi_wl_entry, u.rb); |
801c135c AB |
368 | if (e1->ec >= max) |
369 | p = p->rb_left; | |
370 | else { | |
371 | p = p->rb_right; | |
372 | e = e1; | |
373 | } | |
374 | } | |
375 | ||
376 | return e; | |
377 | } | |
378 | ||
379 | /** | |
380 | * ubi_wl_get_peb - get a physical eraseblock. | |
381 | * @ubi: UBI device description object | |
382 | * @dtype: type of data which will be stored in this physical eraseblock | |
383 | * | |
384 | * This function returns a physical eraseblock in case of success and a | |
385 | * negative error code in case of failure. Might sleep. | |
386 | */ | |
387 | int ubi_wl_get_peb(struct ubi_device *ubi, int dtype) | |
388 | { | |
7b6c32da | 389 | int err, medium_ec; |
801c135c | 390 | struct ubi_wl_entry *e, *first, *last; |
801c135c AB |
391 | |
392 | ubi_assert(dtype == UBI_LONGTERM || dtype == UBI_SHORTTERM || | |
393 | dtype == UBI_UNKNOWN); | |
394 | ||
801c135c AB |
395 | retry: |
396 | spin_lock(&ubi->wl_lock); | |
5abde384 | 397 | if (!ubi->free.rb_node) { |
801c135c AB |
398 | if (ubi->works_count == 0) { |
399 | ubi_assert(list_empty(&ubi->works)); | |
400 | ubi_err("no free eraseblocks"); | |
401 | spin_unlock(&ubi->wl_lock); | |
801c135c AB |
402 | return -ENOSPC; |
403 | } | |
404 | spin_unlock(&ubi->wl_lock); | |
405 | ||
406 | err = produce_free_peb(ubi); | |
7b6c32da | 407 | if (err < 0) |
801c135c | 408 | return err; |
801c135c AB |
409 | goto retry; |
410 | } | |
411 | ||
412 | switch (dtype) { | |
9c9ec147 AB |
413 | case UBI_LONGTERM: |
414 | /* | |
415 | * For long term data we pick a physical eraseblock with high | |
416 | * erase counter. But the highest erase counter we can pick is | |
417 | * bounded by the the lowest erase counter plus | |
418 | * %WL_FREE_MAX_DIFF. | |
419 | */ | |
420 | e = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); | |
9c9ec147 AB |
421 | break; |
422 | case UBI_UNKNOWN: | |
423 | /* | |
424 | * For unknown data we pick a physical eraseblock with medium | |
425 | * erase counter. But we by no means can pick a physical | |
426 | * eraseblock with erase counter greater or equivalent than the | |
427 | * lowest erase counter plus %WL_FREE_MAX_DIFF. | |
428 | */ | |
23553b2c XX |
429 | first = rb_entry(rb_first(&ubi->free), struct ubi_wl_entry, |
430 | u.rb); | |
431 | last = rb_entry(rb_last(&ubi->free), struct ubi_wl_entry, u.rb); | |
801c135c | 432 | |
9c9ec147 AB |
433 | if (last->ec - first->ec < WL_FREE_MAX_DIFF) |
434 | e = rb_entry(ubi->free.rb_node, | |
23553b2c | 435 | struct ubi_wl_entry, u.rb); |
9c9ec147 AB |
436 | else { |
437 | medium_ec = (first->ec + WL_FREE_MAX_DIFF)/2; | |
438 | e = find_wl_entry(&ubi->free, medium_ec); | |
439 | } | |
9c9ec147 AB |
440 | break; |
441 | case UBI_SHORTTERM: | |
442 | /* | |
443 | * For short term data we pick a physical eraseblock with the | |
444 | * lowest erase counter as we expect it will be erased soon. | |
445 | */ | |
23553b2c | 446 | e = rb_entry(rb_first(&ubi->free), struct ubi_wl_entry, u.rb); |
9c9ec147 AB |
447 | break; |
448 | default: | |
9c9ec147 | 449 | BUG(); |
801c135c AB |
450 | } |
451 | ||
7b6c32da XX |
452 | paranoid_check_in_wl_tree(e, &ubi->free); |
453 | ||
801c135c | 454 | /* |
7b6c32da | 455 | * Move the physical eraseblock to the protection queue where it will |
801c135c AB |
456 | * be protected from being moved for some time. |
457 | */ | |
23553b2c | 458 | rb_erase(&e->u.rb, &ubi->free); |
7b6c32da XX |
459 | dbg_wl("PEB %d EC %d", e->pnum, e->ec); |
460 | prot_queue_add(ubi, e); | |
801c135c | 461 | spin_unlock(&ubi->wl_lock); |
801c135c AB |
462 | return e->pnum; |
463 | } | |
464 | ||
465 | /** | |
7b6c32da | 466 | * prot_queue_del - remove a physical eraseblock from the protection queue. |
801c135c AB |
467 | * @ubi: UBI device description object |
468 | * @pnum: the physical eraseblock to remove | |
43f9b25a | 469 | * |
7b6c32da XX |
470 | * This function deletes PEB @pnum from the protection queue and returns zero |
471 | * in case of success and %-ENODEV if the PEB was not found. | |
801c135c | 472 | */ |
7b6c32da | 473 | static int prot_queue_del(struct ubi_device *ubi, int pnum) |
801c135c | 474 | { |
7b6c32da | 475 | struct ubi_wl_entry *e; |
801c135c | 476 | |
7b6c32da XX |
477 | e = ubi->lookuptbl[pnum]; |
478 | if (!e) | |
479 | return -ENODEV; | |
801c135c | 480 | |
7b6c32da XX |
481 | if (paranoid_check_in_pq(ubi, e)) |
482 | return -ENODEV; | |
43f9b25a | 483 | |
7b6c32da XX |
484 | list_del(&e->u.list); |
485 | dbg_wl("deleted PEB %d from the protection queue", e->pnum); | |
43f9b25a | 486 | return 0; |
801c135c AB |
487 | } |
488 | ||
489 | /** | |
490 | * sync_erase - synchronously erase a physical eraseblock. | |
491 | * @ubi: UBI device description object | |
492 | * @e: the the physical eraseblock to erase | |
493 | * @torture: if the physical eraseblock has to be tortured | |
494 | * | |
495 | * This function returns zero in case of success and a negative error code in | |
496 | * case of failure. | |
497 | */ | |
9c9ec147 AB |
498 | static int sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e, |
499 | int torture) | |
801c135c AB |
500 | { |
501 | int err; | |
502 | struct ubi_ec_hdr *ec_hdr; | |
503 | unsigned long long ec = e->ec; | |
504 | ||
505 | dbg_wl("erase PEB %d, old EC %llu", e->pnum, ec); | |
506 | ||
507 | err = paranoid_check_ec(ubi, e->pnum, e->ec); | |
508 | if (err > 0) | |
509 | return -EINVAL; | |
510 | ||
33818bbb | 511 | ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_NOFS); |
801c135c AB |
512 | if (!ec_hdr) |
513 | return -ENOMEM; | |
514 | ||
515 | err = ubi_io_sync_erase(ubi, e->pnum, torture); | |
516 | if (err < 0) | |
517 | goto out_free; | |
518 | ||
519 | ec += err; | |
520 | if (ec > UBI_MAX_ERASECOUNTER) { | |
521 | /* | |
522 | * Erase counter overflow. Upgrade UBI and use 64-bit | |
523 | * erase counters internally. | |
524 | */ | |
525 | ubi_err("erase counter overflow at PEB %d, EC %llu", | |
526 | e->pnum, ec); | |
527 | err = -EINVAL; | |
528 | goto out_free; | |
529 | } | |
530 | ||
531 | dbg_wl("erased PEB %d, new EC %llu", e->pnum, ec); | |
532 | ||
3261ebd7 | 533 | ec_hdr->ec = cpu_to_be64(ec); |
801c135c AB |
534 | |
535 | err = ubi_io_write_ec_hdr(ubi, e->pnum, ec_hdr); | |
536 | if (err) | |
537 | goto out_free; | |
538 | ||
539 | e->ec = ec; | |
540 | spin_lock(&ubi->wl_lock); | |
541 | if (e->ec > ubi->max_ec) | |
542 | ubi->max_ec = e->ec; | |
543 | spin_unlock(&ubi->wl_lock); | |
544 | ||
545 | out_free: | |
546 | kfree(ec_hdr); | |
547 | return err; | |
548 | } | |
549 | ||
550 | /** | |
7b6c32da | 551 | * serve_prot_queue - check if it is time to stop protecting PEBs. |
801c135c AB |
552 | * @ubi: UBI device description object |
553 | * | |
7b6c32da XX |
554 | * This function is called after each erase operation and removes PEBs from the |
555 | * tail of the protection queue. These PEBs have been protected for long enough | |
556 | * and should be moved to the used tree. | |
801c135c | 557 | */ |
7b6c32da | 558 | static void serve_prot_queue(struct ubi_device *ubi) |
801c135c | 559 | { |
7b6c32da XX |
560 | struct ubi_wl_entry *e, *tmp; |
561 | int count; | |
801c135c AB |
562 | |
563 | /* | |
564 | * There may be several protected physical eraseblock to remove, | |
565 | * process them all. | |
566 | */ | |
7b6c32da XX |
567 | repeat: |
568 | count = 0; | |
569 | spin_lock(&ubi->wl_lock); | |
570 | list_for_each_entry_safe(e, tmp, &ubi->pq[ubi->pq_head], u.list) { | |
571 | dbg_wl("PEB %d EC %d protection over, move to used tree", | |
572 | e->pnum, e->ec); | |
801c135c | 573 | |
7b6c32da XX |
574 | list_del(&e->u.list); |
575 | wl_tree_add(e, &ubi->used); | |
576 | if (count++ > 32) { | |
577 | /* | |
578 | * Let's be nice and avoid holding the spinlock for | |
579 | * too long. | |
580 | */ | |
801c135c | 581 | spin_unlock(&ubi->wl_lock); |
7b6c32da XX |
582 | cond_resched(); |
583 | goto repeat; | |
801c135c | 584 | } |
801c135c | 585 | } |
7b6c32da XX |
586 | |
587 | ubi->pq_head += 1; | |
588 | if (ubi->pq_head == UBI_PROT_QUEUE_LEN) | |
589 | ubi->pq_head = 0; | |
590 | ubi_assert(ubi->pq_head >= 0 && ubi->pq_head < UBI_PROT_QUEUE_LEN); | |
591 | spin_unlock(&ubi->wl_lock); | |
801c135c AB |
592 | } |
593 | ||
594 | /** | |
595 | * schedule_ubi_work - schedule a work. | |
596 | * @ubi: UBI device description object | |
597 | * @wrk: the work to schedule | |
598 | * | |
7b6c32da XX |
599 | * This function adds a work defined by @wrk to the tail of the pending works |
600 | * list. | |
801c135c AB |
601 | */ |
602 | static void schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk) | |
603 | { | |
604 | spin_lock(&ubi->wl_lock); | |
605 | list_add_tail(&wrk->list, &ubi->works); | |
606 | ubi_assert(ubi->works_count >= 0); | |
607 | ubi->works_count += 1; | |
608 | if (ubi->thread_enabled) | |
609 | wake_up_process(ubi->bgt_thread); | |
610 | spin_unlock(&ubi->wl_lock); | |
611 | } | |
612 | ||
613 | static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk, | |
614 | int cancel); | |
615 | ||
616 | /** | |
617 | * schedule_erase - schedule an erase work. | |
618 | * @ubi: UBI device description object | |
619 | * @e: the WL entry of the physical eraseblock to erase | |
620 | * @torture: if the physical eraseblock has to be tortured | |
621 | * | |
622 | * This function returns zero in case of success and a %-ENOMEM in case of | |
623 | * failure. | |
624 | */ | |
625 | static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e, | |
626 | int torture) | |
627 | { | |
628 | struct ubi_work *wl_wrk; | |
629 | ||
630 | dbg_wl("schedule erasure of PEB %d, EC %d, torture %d", | |
631 | e->pnum, e->ec, torture); | |
632 | ||
33818bbb | 633 | wl_wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS); |
801c135c AB |
634 | if (!wl_wrk) |
635 | return -ENOMEM; | |
636 | ||
637 | wl_wrk->func = &erase_worker; | |
638 | wl_wrk->e = e; | |
639 | wl_wrk->torture = torture; | |
640 | ||
641 | schedule_ubi_work(ubi, wl_wrk); | |
642 | return 0; | |
643 | } | |
644 | ||
645 | /** | |
646 | * wear_leveling_worker - wear-leveling worker function. | |
647 | * @ubi: UBI device description object | |
648 | * @wrk: the work object | |
649 | * @cancel: non-zero if the worker has to free memory and exit | |
650 | * | |
651 | * This function copies a more worn out physical eraseblock to a less worn out | |
652 | * one. Returns zero in case of success and a negative error code in case of | |
653 | * failure. | |
654 | */ | |
655 | static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk, | |
656 | int cancel) | |
657 | { | |
b86a2c56 | 658 | int err, scrubbing = 0, torture = 0, protect = 0, erroneous = 0; |
9c259a52 | 659 | int vol_id = -1, uninitialized_var(lnum); |
801c135c AB |
660 | struct ubi_wl_entry *e1, *e2; |
661 | struct ubi_vid_hdr *vid_hdr; | |
662 | ||
663 | kfree(wrk); | |
801c135c AB |
664 | if (cancel) |
665 | return 0; | |
666 | ||
33818bbb | 667 | vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS); |
801c135c AB |
668 | if (!vid_hdr) |
669 | return -ENOMEM; | |
670 | ||
43f9b25a | 671 | mutex_lock(&ubi->move_mutex); |
801c135c | 672 | spin_lock(&ubi->wl_lock); |
43f9b25a AB |
673 | ubi_assert(!ubi->move_from && !ubi->move_to); |
674 | ubi_assert(!ubi->move_to_put); | |
801c135c | 675 | |
43f9b25a | 676 | if (!ubi->free.rb_node || |
5abde384 | 677 | (!ubi->used.rb_node && !ubi->scrub.rb_node)) { |
801c135c | 678 | /* |
43f9b25a AB |
679 | * No free physical eraseblocks? Well, they must be waiting in |
680 | * the queue to be erased. Cancel movement - it will be | |
681 | * triggered again when a free physical eraseblock appears. | |
801c135c AB |
682 | * |
683 | * No used physical eraseblocks? They must be temporarily | |
684 | * protected from being moved. They will be moved to the | |
685 | * @ubi->used tree later and the wear-leveling will be | |
686 | * triggered again. | |
687 | */ | |
688 | dbg_wl("cancel WL, a list is empty: free %d, used %d", | |
5abde384 | 689 | !ubi->free.rb_node, !ubi->used.rb_node); |
43f9b25a | 690 | goto out_cancel; |
801c135c AB |
691 | } |
692 | ||
5abde384 | 693 | if (!ubi->scrub.rb_node) { |
801c135c AB |
694 | /* |
695 | * Now pick the least worn-out used physical eraseblock and a | |
696 | * highly worn-out free physical eraseblock. If the erase | |
697 | * counters differ much enough, start wear-leveling. | |
698 | */ | |
23553b2c | 699 | e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, u.rb); |
801c135c AB |
700 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); |
701 | ||
702 | if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD)) { | |
703 | dbg_wl("no WL needed: min used EC %d, max free EC %d", | |
704 | e1->ec, e2->ec); | |
43f9b25a | 705 | goto out_cancel; |
801c135c | 706 | } |
5abde384 | 707 | paranoid_check_in_wl_tree(e1, &ubi->used); |
23553b2c | 708 | rb_erase(&e1->u.rb, &ubi->used); |
801c135c AB |
709 | dbg_wl("move PEB %d EC %d to PEB %d EC %d", |
710 | e1->pnum, e1->ec, e2->pnum, e2->ec); | |
711 | } else { | |
43f9b25a AB |
712 | /* Perform scrubbing */ |
713 | scrubbing = 1; | |
23553b2c | 714 | e1 = rb_entry(rb_first(&ubi->scrub), struct ubi_wl_entry, u.rb); |
801c135c | 715 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); |
5abde384 | 716 | paranoid_check_in_wl_tree(e1, &ubi->scrub); |
23553b2c | 717 | rb_erase(&e1->u.rb, &ubi->scrub); |
801c135c AB |
718 | dbg_wl("scrub PEB %d to PEB %d", e1->pnum, e2->pnum); |
719 | } | |
720 | ||
5abde384 | 721 | paranoid_check_in_wl_tree(e2, &ubi->free); |
23553b2c | 722 | rb_erase(&e2->u.rb, &ubi->free); |
801c135c AB |
723 | ubi->move_from = e1; |
724 | ubi->move_to = e2; | |
725 | spin_unlock(&ubi->wl_lock); | |
726 | ||
727 | /* | |
728 | * Now we are going to copy physical eraseblock @e1->pnum to @e2->pnum. | |
729 | * We so far do not know which logical eraseblock our physical | |
730 | * eraseblock (@e1) belongs to. We have to read the volume identifier | |
731 | * header first. | |
43f9b25a AB |
732 | * |
733 | * Note, we are protected from this PEB being unmapped and erased. The | |
734 | * 'ubi_wl_put_peb()' would wait for moving to be finished if the PEB | |
735 | * which is being moved was unmapped. | |
801c135c AB |
736 | */ |
737 | ||
738 | err = ubi_io_read_vid_hdr(ubi, e1->pnum, vid_hdr, 0); | |
739 | if (err && err != UBI_IO_BITFLIPS) { | |
740 | if (err == UBI_IO_PEB_FREE) { | |
741 | /* | |
742 | * We are trying to move PEB without a VID header. UBI | |
743 | * always write VID headers shortly after the PEB was | |
87960c0b AB |
744 | * given, so we have a situation when it has not yet |
745 | * had a chance to write it, because it was preempted. | |
746 | * So add this PEB to the protection queue so far, | |
747 | * because presubably more data will be written there | |
748 | * (including the missin VID header), and then we'll | |
749 | * move it. | |
801c135c AB |
750 | */ |
751 | dbg_wl("PEB %d has no VID header", e1->pnum); | |
87960c0b | 752 | protect = 1; |
43f9b25a | 753 | goto out_not_moved; |
801c135c | 754 | } |
43f9b25a AB |
755 | |
756 | ubi_err("error %d while reading VID header from PEB %d", | |
757 | err, e1->pnum); | |
43f9b25a | 758 | goto out_error; |
801c135c AB |
759 | } |
760 | ||
9c259a52 AB |
761 | vol_id = be32_to_cpu(vid_hdr->vol_id); |
762 | lnum = be32_to_cpu(vid_hdr->lnum); | |
763 | ||
801c135c AB |
764 | err = ubi_eba_copy_leb(ubi, e1->pnum, e2->pnum, vid_hdr); |
765 | if (err) { | |
87960c0b AB |
766 | if (err == MOVE_CANCEL_RACE) { |
767 | /* | |
768 | * The LEB has not been moved because the volume is | |
769 | * being deleted or the PEB has been put meanwhile. We | |
770 | * should prevent this PEB from being selected for | |
771 | * wear-leveling movement again, so put it to the | |
772 | * protection queue. | |
773 | */ | |
774 | protect = 1; | |
775 | goto out_not_moved; | |
776 | } | |
777 | ||
b86a2c56 AB |
778 | if (err == MOVE_CANCEL_BITFLIPS || err == MOVE_TARGET_WR_ERR || |
779 | err == MOVE_TARGET_RD_ERR) { | |
9c259a52 AB |
780 | /* |
781 | * Target PEB had bit-flips or write error - torture it. | |
782 | */ | |
6fa6f5bb | 783 | torture = 1; |
43f9b25a | 784 | goto out_not_moved; |
6fa6f5bb | 785 | } |
87960c0b | 786 | |
b86a2c56 AB |
787 | if (err == MOVE_SOURCE_RD_ERR) { |
788 | /* | |
789 | * An error happened while reading the source PEB. Do | |
790 | * not switch to R/O mode in this case, and give the | |
791 | * upper layers a possibility to recover from this, | |
792 | * e.g. by unmapping corresponding LEB. Instead, just | |
793 | * put thie PEB to the @ubi->erroneus list to prevent | |
794 | * UBI from trying to move the over and over again. | |
795 | */ | |
796 | if (ubi->erroneous_peb_count > ubi->max_erroneous) { | |
797 | ubi_err("too many erroneous eraseblocks (%d)", | |
798 | ubi->erroneous_peb_count); | |
799 | goto out_error; | |
800 | } | |
801 | erroneous = 1; | |
802 | goto out_not_moved; | |
803 | } | |
804 | ||
90bf0265 AB |
805 | if (err < 0) |
806 | goto out_error; | |
43f9b25a | 807 | |
87960c0b | 808 | ubi_assert(0); |
801c135c AB |
809 | } |
810 | ||
6a8f483f | 811 | /* The PEB has been successfully moved */ |
6a8f483f | 812 | if (scrubbing) |
9c259a52 AB |
813 | ubi_msg("scrubbed PEB %d (LEB %d:%d), data moved to PEB %d", |
814 | e1->pnum, vol_id, lnum, e2->pnum); | |
815 | ubi_free_vid_hdr(ubi, vid_hdr); | |
8c1e6ee1 | 816 | |
801c135c | 817 | spin_lock(&ubi->wl_lock); |
3c98b0a0 | 818 | if (!ubi->move_to_put) { |
5abde384 | 819 | wl_tree_add(e2, &ubi->used); |
3c98b0a0 AB |
820 | e2 = NULL; |
821 | } | |
801c135c | 822 | ubi->move_from = ubi->move_to = NULL; |
43f9b25a | 823 | ubi->move_to_put = ubi->wl_scheduled = 0; |
801c135c AB |
824 | spin_unlock(&ubi->wl_lock); |
825 | ||
6a8f483f | 826 | err = schedule_erase(ubi, e1, 0); |
3c98b0a0 | 827 | if (err) { |
87960c0b AB |
828 | kmem_cache_free(ubi_wl_entry_slab, e1); |
829 | kmem_cache_free(ubi_wl_entry_slab, e2); | |
830 | goto out_ro; | |
3c98b0a0 | 831 | } |
6a8f483f | 832 | |
3c98b0a0 | 833 | if (e2) { |
801c135c AB |
834 | /* |
835 | * Well, the target PEB was put meanwhile, schedule it for | |
836 | * erasure. | |
837 | */ | |
9c259a52 AB |
838 | dbg_wl("PEB %d (LEB %d:%d) was put meanwhile, erase", |
839 | e2->pnum, vol_id, lnum); | |
801c135c | 840 | err = schedule_erase(ubi, e2, 0); |
87960c0b AB |
841 | if (err) { |
842 | kmem_cache_free(ubi_wl_entry_slab, e2); | |
843 | goto out_ro; | |
844 | } | |
801c135c AB |
845 | } |
846 | ||
801c135c | 847 | dbg_wl("done"); |
43f9b25a AB |
848 | mutex_unlock(&ubi->move_mutex); |
849 | return 0; | |
801c135c AB |
850 | |
851 | /* | |
43f9b25a AB |
852 | * For some reasons the LEB was not moved, might be an error, might be |
853 | * something else. @e1 was not changed, so return it back. @e2 might | |
6fa6f5bb | 854 | * have been changed, schedule it for erasure. |
801c135c | 855 | */ |
43f9b25a | 856 | out_not_moved: |
9c259a52 AB |
857 | if (vol_id != -1) |
858 | dbg_wl("cancel moving PEB %d (LEB %d:%d) to PEB %d (%d)", | |
859 | e1->pnum, vol_id, lnum, e2->pnum, err); | |
860 | else | |
861 | dbg_wl("cancel moving PEB %d to PEB %d (%d)", | |
862 | e1->pnum, e2->pnum, err); | |
801c135c | 863 | spin_lock(&ubi->wl_lock); |
87960c0b AB |
864 | if (protect) |
865 | prot_queue_add(ubi, e1); | |
b86a2c56 AB |
866 | else if (erroneous) { |
867 | wl_tree_add(e1, &ubi->erroneous); | |
868 | ubi->erroneous_peb_count += 1; | |
869 | } else if (scrubbing) | |
43f9b25a | 870 | wl_tree_add(e1, &ubi->scrub); |
801c135c | 871 | else |
5abde384 | 872 | wl_tree_add(e1, &ubi->used); |
6fa6f5bb | 873 | ubi_assert(!ubi->move_to_put); |
801c135c | 874 | ubi->move_from = ubi->move_to = NULL; |
6fa6f5bb | 875 | ubi->wl_scheduled = 0; |
801c135c AB |
876 | spin_unlock(&ubi->wl_lock); |
877 | ||
87960c0b | 878 | ubi_free_vid_hdr(ubi, vid_hdr); |
6fa6f5bb | 879 | err = schedule_erase(ubi, e2, torture); |
87960c0b AB |
880 | if (err) { |
881 | kmem_cache_free(ubi_wl_entry_slab, e2); | |
882 | goto out_ro; | |
883 | } | |
43f9b25a AB |
884 | mutex_unlock(&ubi->move_mutex); |
885 | return 0; | |
886 | ||
887 | out_error: | |
9c259a52 AB |
888 | if (vol_id != -1) |
889 | ubi_err("error %d while moving PEB %d to PEB %d", | |
890 | err, e1->pnum, e2->pnum); | |
891 | else | |
892 | ubi_err("error %d while moving PEB %d (LEB %d:%d) to PEB %d", | |
893 | err, e1->pnum, vol_id, lnum, e2->pnum); | |
43f9b25a AB |
894 | spin_lock(&ubi->wl_lock); |
895 | ubi->move_from = ubi->move_to = NULL; | |
896 | ubi->move_to_put = ubi->wl_scheduled = 0; | |
897 | spin_unlock(&ubi->wl_lock); | |
898 | ||
87960c0b AB |
899 | ubi_free_vid_hdr(ubi, vid_hdr); |
900 | kmem_cache_free(ubi_wl_entry_slab, e1); | |
901 | kmem_cache_free(ubi_wl_entry_slab, e2); | |
43f9b25a | 902 | |
87960c0b AB |
903 | out_ro: |
904 | ubi_ro_mode(ubi); | |
43f9b25a | 905 | mutex_unlock(&ubi->move_mutex); |
87960c0b AB |
906 | ubi_assert(err != 0); |
907 | return err < 0 ? err : -EIO; | |
43f9b25a AB |
908 | |
909 | out_cancel: | |
910 | ubi->wl_scheduled = 0; | |
911 | spin_unlock(&ubi->wl_lock); | |
912 | mutex_unlock(&ubi->move_mutex); | |
913 | ubi_free_vid_hdr(ubi, vid_hdr); | |
914 | return 0; | |
801c135c AB |
915 | } |
916 | ||
917 | /** | |
918 | * ensure_wear_leveling - schedule wear-leveling if it is needed. | |
919 | * @ubi: UBI device description object | |
920 | * | |
921 | * This function checks if it is time to start wear-leveling and schedules it | |
922 | * if yes. This function returns zero in case of success and a negative error | |
923 | * code in case of failure. | |
924 | */ | |
925 | static int ensure_wear_leveling(struct ubi_device *ubi) | |
926 | { | |
927 | int err = 0; | |
928 | struct ubi_wl_entry *e1; | |
929 | struct ubi_wl_entry *e2; | |
930 | struct ubi_work *wrk; | |
931 | ||
932 | spin_lock(&ubi->wl_lock); | |
933 | if (ubi->wl_scheduled) | |
934 | /* Wear-leveling is already in the work queue */ | |
935 | goto out_unlock; | |
936 | ||
937 | /* | |
938 | * If the ubi->scrub tree is not empty, scrubbing is needed, and the | |
939 | * the WL worker has to be scheduled anyway. | |
940 | */ | |
5abde384 AB |
941 | if (!ubi->scrub.rb_node) { |
942 | if (!ubi->used.rb_node || !ubi->free.rb_node) | |
801c135c AB |
943 | /* No physical eraseblocks - no deal */ |
944 | goto out_unlock; | |
945 | ||
946 | /* | |
947 | * We schedule wear-leveling only if the difference between the | |
948 | * lowest erase counter of used physical eraseblocks and a high | |
025dfdaf | 949 | * erase counter of free physical eraseblocks is greater than |
801c135c AB |
950 | * %UBI_WL_THRESHOLD. |
951 | */ | |
23553b2c | 952 | e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, u.rb); |
801c135c AB |
953 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); |
954 | ||
955 | if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD)) | |
956 | goto out_unlock; | |
957 | dbg_wl("schedule wear-leveling"); | |
958 | } else | |
959 | dbg_wl("schedule scrubbing"); | |
960 | ||
961 | ubi->wl_scheduled = 1; | |
962 | spin_unlock(&ubi->wl_lock); | |
963 | ||
33818bbb | 964 | wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS); |
801c135c AB |
965 | if (!wrk) { |
966 | err = -ENOMEM; | |
967 | goto out_cancel; | |
968 | } | |
969 | ||
970 | wrk->func = &wear_leveling_worker; | |
971 | schedule_ubi_work(ubi, wrk); | |
972 | return err; | |
973 | ||
974 | out_cancel: | |
975 | spin_lock(&ubi->wl_lock); | |
976 | ubi->wl_scheduled = 0; | |
977 | out_unlock: | |
978 | spin_unlock(&ubi->wl_lock); | |
979 | return err; | |
980 | } | |
981 | ||
982 | /** | |
983 | * erase_worker - physical eraseblock erase worker function. | |
984 | * @ubi: UBI device description object | |
985 | * @wl_wrk: the work object | |
986 | * @cancel: non-zero if the worker has to free memory and exit | |
987 | * | |
988 | * This function erases a physical eraseblock and perform torture testing if | |
989 | * needed. It also takes care about marking the physical eraseblock bad if | |
990 | * needed. Returns zero in case of success and a negative error code in case of | |
991 | * failure. | |
992 | */ | |
993 | static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk, | |
994 | int cancel) | |
995 | { | |
801c135c | 996 | struct ubi_wl_entry *e = wl_wrk->e; |
784c1454 | 997 | int pnum = e->pnum, err, need; |
801c135c AB |
998 | |
999 | if (cancel) { | |
1000 | dbg_wl("cancel erasure of PEB %d EC %d", pnum, e->ec); | |
1001 | kfree(wl_wrk); | |
06b68ba1 | 1002 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1003 | return 0; |
1004 | } | |
1005 | ||
1006 | dbg_wl("erase PEB %d EC %d", pnum, e->ec); | |
1007 | ||
1008 | err = sync_erase(ubi, e, wl_wrk->torture); | |
1009 | if (!err) { | |
1010 | /* Fine, we've erased it successfully */ | |
1011 | kfree(wl_wrk); | |
1012 | ||
1013 | spin_lock(&ubi->wl_lock); | |
5abde384 | 1014 | wl_tree_add(e, &ubi->free); |
801c135c AB |
1015 | spin_unlock(&ubi->wl_lock); |
1016 | ||
1017 | /* | |
9c9ec147 AB |
1018 | * One more erase operation has happened, take care about |
1019 | * protected physical eraseblocks. | |
801c135c | 1020 | */ |
7b6c32da | 1021 | serve_prot_queue(ubi); |
801c135c AB |
1022 | |
1023 | /* And take care about wear-leveling */ | |
1024 | err = ensure_wear_leveling(ubi); | |
1025 | return err; | |
1026 | } | |
1027 | ||
8d2d4011 | 1028 | ubi_err("failed to erase PEB %d, error %d", pnum, err); |
801c135c | 1029 | kfree(wl_wrk); |
06b68ba1 | 1030 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c | 1031 | |
784c1454 AB |
1032 | if (err == -EINTR || err == -ENOMEM || err == -EAGAIN || |
1033 | err == -EBUSY) { | |
1034 | int err1; | |
1035 | ||
1036 | /* Re-schedule the LEB for erasure */ | |
1037 | err1 = schedule_erase(ubi, e, 0); | |
1038 | if (err1) { | |
1039 | err = err1; | |
1040 | goto out_ro; | |
1041 | } | |
1042 | return err; | |
1043 | } else if (err != -EIO) { | |
801c135c AB |
1044 | /* |
1045 | * If this is not %-EIO, we have no idea what to do. Scheduling | |
1046 | * this physical eraseblock for erasure again would cause | |
1047 | * errors again and again. Well, lets switch to RO mode. | |
1048 | */ | |
784c1454 | 1049 | goto out_ro; |
801c135c AB |
1050 | } |
1051 | ||
1052 | /* It is %-EIO, the PEB went bad */ | |
1053 | ||
1054 | if (!ubi->bad_allowed) { | |
1055 | ubi_err("bad physical eraseblock %d detected", pnum); | |
784c1454 AB |
1056 | goto out_ro; |
1057 | } | |
801c135c | 1058 | |
784c1454 AB |
1059 | spin_lock(&ubi->volumes_lock); |
1060 | need = ubi->beb_rsvd_level - ubi->beb_rsvd_pebs + 1; | |
1061 | if (need > 0) { | |
1062 | need = ubi->avail_pebs >= need ? need : ubi->avail_pebs; | |
1063 | ubi->avail_pebs -= need; | |
1064 | ubi->rsvd_pebs += need; | |
1065 | ubi->beb_rsvd_pebs += need; | |
1066 | if (need > 0) | |
1067 | ubi_msg("reserve more %d PEBs", need); | |
1068 | } | |
801c135c | 1069 | |
784c1454 | 1070 | if (ubi->beb_rsvd_pebs == 0) { |
801c135c | 1071 | spin_unlock(&ubi->volumes_lock); |
784c1454 AB |
1072 | ubi_err("no reserved physical eraseblocks"); |
1073 | goto out_ro; | |
1074 | } | |
801c135c | 1075 | |
784c1454 AB |
1076 | spin_unlock(&ubi->volumes_lock); |
1077 | ubi_msg("mark PEB %d as bad", pnum); | |
801c135c | 1078 | |
784c1454 AB |
1079 | err = ubi_io_mark_bad(ubi, pnum); |
1080 | if (err) | |
1081 | goto out_ro; | |
1082 | ||
1083 | spin_lock(&ubi->volumes_lock); | |
1084 | ubi->beb_rsvd_pebs -= 1; | |
1085 | ubi->bad_peb_count += 1; | |
1086 | ubi->good_peb_count -= 1; | |
1087 | ubi_calculate_reserved(ubi); | |
1088 | if (ubi->beb_rsvd_pebs == 0) | |
1089 | ubi_warn("last PEB from the reserved pool was used"); | |
1090 | spin_unlock(&ubi->volumes_lock); | |
1091 | ||
1092 | return err; | |
801c135c | 1093 | |
784c1454 AB |
1094 | out_ro: |
1095 | ubi_ro_mode(ubi); | |
801c135c AB |
1096 | return err; |
1097 | } | |
1098 | ||
1099 | /** | |
85c6e6e2 | 1100 | * ubi_wl_put_peb - return a PEB to the wear-leveling sub-system. |
801c135c AB |
1101 | * @ubi: UBI device description object |
1102 | * @pnum: physical eraseblock to return | |
1103 | * @torture: if this physical eraseblock has to be tortured | |
1104 | * | |
1105 | * This function is called to return physical eraseblock @pnum to the pool of | |
1106 | * free physical eraseblocks. The @torture flag has to be set if an I/O error | |
1107 | * occurred to this @pnum and it has to be tested. This function returns zero | |
43f9b25a | 1108 | * in case of success, and a negative error code in case of failure. |
801c135c AB |
1109 | */ |
1110 | int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture) | |
1111 | { | |
1112 | int err; | |
1113 | struct ubi_wl_entry *e; | |
1114 | ||
1115 | dbg_wl("PEB %d", pnum); | |
1116 | ubi_assert(pnum >= 0); | |
1117 | ubi_assert(pnum < ubi->peb_count); | |
1118 | ||
43f9b25a | 1119 | retry: |
801c135c | 1120 | spin_lock(&ubi->wl_lock); |
801c135c AB |
1121 | e = ubi->lookuptbl[pnum]; |
1122 | if (e == ubi->move_from) { | |
1123 | /* | |
1124 | * User is putting the physical eraseblock which was selected to | |
1125 | * be moved. It will be scheduled for erasure in the | |
1126 | * wear-leveling worker. | |
1127 | */ | |
43f9b25a | 1128 | dbg_wl("PEB %d is being moved, wait", pnum); |
801c135c | 1129 | spin_unlock(&ubi->wl_lock); |
43f9b25a AB |
1130 | |
1131 | /* Wait for the WL worker by taking the @ubi->move_mutex */ | |
1132 | mutex_lock(&ubi->move_mutex); | |
1133 | mutex_unlock(&ubi->move_mutex); | |
1134 | goto retry; | |
801c135c AB |
1135 | } else if (e == ubi->move_to) { |
1136 | /* | |
1137 | * User is putting the physical eraseblock which was selected | |
1138 | * as the target the data is moved to. It may happen if the EBA | |
85c6e6e2 AB |
1139 | * sub-system already re-mapped the LEB in 'ubi_eba_copy_leb()' |
1140 | * but the WL sub-system has not put the PEB to the "used" tree | |
1141 | * yet, but it is about to do this. So we just set a flag which | |
1142 | * will tell the WL worker that the PEB is not needed anymore | |
1143 | * and should be scheduled for erasure. | |
801c135c AB |
1144 | */ |
1145 | dbg_wl("PEB %d is the target of data moving", pnum); | |
1146 | ubi_assert(!ubi->move_to_put); | |
1147 | ubi->move_to_put = 1; | |
1148 | spin_unlock(&ubi->wl_lock); | |
1149 | return 0; | |
1150 | } else { | |
5abde384 AB |
1151 | if (in_wl_tree(e, &ubi->used)) { |
1152 | paranoid_check_in_wl_tree(e, &ubi->used); | |
23553b2c | 1153 | rb_erase(&e->u.rb, &ubi->used); |
5abde384 AB |
1154 | } else if (in_wl_tree(e, &ubi->scrub)) { |
1155 | paranoid_check_in_wl_tree(e, &ubi->scrub); | |
23553b2c | 1156 | rb_erase(&e->u.rb, &ubi->scrub); |
b86a2c56 AB |
1157 | } else if (in_wl_tree(e, &ubi->erroneous)) { |
1158 | paranoid_check_in_wl_tree(e, &ubi->erroneous); | |
1159 | rb_erase(&e->u.rb, &ubi->erroneous); | |
1160 | ubi->erroneous_peb_count -= 1; | |
1161 | ubi_assert(ubi->erroneous_peb_count >= 0); | |
1162 | /* Erronious PEBs should be tortured */ | |
1163 | torture = 1; | |
43f9b25a | 1164 | } else { |
7b6c32da | 1165 | err = prot_queue_del(ubi, e->pnum); |
43f9b25a AB |
1166 | if (err) { |
1167 | ubi_err("PEB %d not found", pnum); | |
1168 | ubi_ro_mode(ubi); | |
1169 | spin_unlock(&ubi->wl_lock); | |
1170 | return err; | |
1171 | } | |
1172 | } | |
801c135c AB |
1173 | } |
1174 | spin_unlock(&ubi->wl_lock); | |
1175 | ||
1176 | err = schedule_erase(ubi, e, torture); | |
1177 | if (err) { | |
1178 | spin_lock(&ubi->wl_lock); | |
5abde384 | 1179 | wl_tree_add(e, &ubi->used); |
801c135c AB |
1180 | spin_unlock(&ubi->wl_lock); |
1181 | } | |
1182 | ||
1183 | return err; | |
1184 | } | |
1185 | ||
1186 | /** | |
1187 | * ubi_wl_scrub_peb - schedule a physical eraseblock for scrubbing. | |
1188 | * @ubi: UBI device description object | |
1189 | * @pnum: the physical eraseblock to schedule | |
1190 | * | |
1191 | * If a bit-flip in a physical eraseblock is detected, this physical eraseblock | |
1192 | * needs scrubbing. This function schedules a physical eraseblock for | |
1193 | * scrubbing which is done in background. This function returns zero in case of | |
1194 | * success and a negative error code in case of failure. | |
1195 | */ | |
1196 | int ubi_wl_scrub_peb(struct ubi_device *ubi, int pnum) | |
1197 | { | |
1198 | struct ubi_wl_entry *e; | |
1199 | ||
8c1e6ee1 | 1200 | dbg_msg("schedule PEB %d for scrubbing", pnum); |
801c135c AB |
1201 | |
1202 | retry: | |
1203 | spin_lock(&ubi->wl_lock); | |
1204 | e = ubi->lookuptbl[pnum]; | |
1205 | if (e == ubi->move_from || in_wl_tree(e, &ubi->scrub)) { | |
1206 | spin_unlock(&ubi->wl_lock); | |
1207 | return 0; | |
1208 | } | |
1209 | ||
1210 | if (e == ubi->move_to) { | |
1211 | /* | |
1212 | * This physical eraseblock was used to move data to. The data | |
1213 | * was moved but the PEB was not yet inserted to the proper | |
1214 | * tree. We should just wait a little and let the WL worker | |
1215 | * proceed. | |
1216 | */ | |
1217 | spin_unlock(&ubi->wl_lock); | |
1218 | dbg_wl("the PEB %d is not in proper tree, retry", pnum); | |
1219 | yield(); | |
1220 | goto retry; | |
1221 | } | |
1222 | ||
5abde384 AB |
1223 | if (in_wl_tree(e, &ubi->used)) { |
1224 | paranoid_check_in_wl_tree(e, &ubi->used); | |
23553b2c | 1225 | rb_erase(&e->u.rb, &ubi->used); |
43f9b25a AB |
1226 | } else { |
1227 | int err; | |
1228 | ||
7b6c32da | 1229 | err = prot_queue_del(ubi, e->pnum); |
43f9b25a AB |
1230 | if (err) { |
1231 | ubi_err("PEB %d not found", pnum); | |
1232 | ubi_ro_mode(ubi); | |
1233 | spin_unlock(&ubi->wl_lock); | |
1234 | return err; | |
1235 | } | |
1236 | } | |
801c135c | 1237 | |
5abde384 | 1238 | wl_tree_add(e, &ubi->scrub); |
801c135c AB |
1239 | spin_unlock(&ubi->wl_lock); |
1240 | ||
1241 | /* | |
1242 | * Technically scrubbing is the same as wear-leveling, so it is done | |
1243 | * by the WL worker. | |
1244 | */ | |
1245 | return ensure_wear_leveling(ubi); | |
1246 | } | |
1247 | ||
1248 | /** | |
1249 | * ubi_wl_flush - flush all pending works. | |
1250 | * @ubi: UBI device description object | |
1251 | * | |
1252 | * This function returns zero in case of success and a negative error code in | |
1253 | * case of failure. | |
1254 | */ | |
1255 | int ubi_wl_flush(struct ubi_device *ubi) | |
1256 | { | |
593dd33c | 1257 | int err; |
801c135c AB |
1258 | |
1259 | /* | |
7b6c32da | 1260 | * Erase while the pending works queue is not empty, but not more than |
801c135c AB |
1261 | * the number of currently pending works. |
1262 | */ | |
593dd33c AB |
1263 | dbg_wl("flush (%d pending works)", ubi->works_count); |
1264 | while (ubi->works_count) { | |
1265 | err = do_work(ubi); | |
1266 | if (err) | |
1267 | return err; | |
1268 | } | |
1269 | ||
1270 | /* | |
1271 | * Make sure all the works which have been done in parallel are | |
1272 | * finished. | |
1273 | */ | |
1274 | down_write(&ubi->work_sem); | |
1275 | up_write(&ubi->work_sem); | |
1276 | ||
1277 | /* | |
6fa6f5bb | 1278 | * And in case last was the WL worker and it canceled the LEB |
593dd33c AB |
1279 | * movement, flush again. |
1280 | */ | |
1281 | while (ubi->works_count) { | |
1282 | dbg_wl("flush more (%d pending works)", ubi->works_count); | |
801c135c AB |
1283 | err = do_work(ubi); |
1284 | if (err) | |
1285 | return err; | |
1286 | } | |
1287 | ||
1288 | return 0; | |
1289 | } | |
1290 | ||
1291 | /** | |
1292 | * tree_destroy - destroy an RB-tree. | |
1293 | * @root: the root of the tree to destroy | |
1294 | */ | |
1295 | static void tree_destroy(struct rb_root *root) | |
1296 | { | |
1297 | struct rb_node *rb; | |
1298 | struct ubi_wl_entry *e; | |
1299 | ||
1300 | rb = root->rb_node; | |
1301 | while (rb) { | |
1302 | if (rb->rb_left) | |
1303 | rb = rb->rb_left; | |
1304 | else if (rb->rb_right) | |
1305 | rb = rb->rb_right; | |
1306 | else { | |
23553b2c | 1307 | e = rb_entry(rb, struct ubi_wl_entry, u.rb); |
801c135c AB |
1308 | |
1309 | rb = rb_parent(rb); | |
1310 | if (rb) { | |
23553b2c | 1311 | if (rb->rb_left == &e->u.rb) |
801c135c AB |
1312 | rb->rb_left = NULL; |
1313 | else | |
1314 | rb->rb_right = NULL; | |
1315 | } | |
1316 | ||
06b68ba1 | 1317 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1318 | } |
1319 | } | |
1320 | } | |
1321 | ||
1322 | /** | |
1323 | * ubi_thread - UBI background thread. | |
1324 | * @u: the UBI device description object pointer | |
1325 | */ | |
cdfa788a | 1326 | int ubi_thread(void *u) |
801c135c AB |
1327 | { |
1328 | int failures = 0; | |
1329 | struct ubi_device *ubi = u; | |
1330 | ||
1331 | ubi_msg("background thread \"%s\" started, PID %d", | |
ba25f9dc | 1332 | ubi->bgt_name, task_pid_nr(current)); |
801c135c | 1333 | |
83144186 | 1334 | set_freezable(); |
801c135c AB |
1335 | for (;;) { |
1336 | int err; | |
1337 | ||
1338 | if (kthread_should_stop()) | |
cadb40cc | 1339 | break; |
801c135c AB |
1340 | |
1341 | if (try_to_freeze()) | |
1342 | continue; | |
1343 | ||
1344 | spin_lock(&ubi->wl_lock); | |
1345 | if (list_empty(&ubi->works) || ubi->ro_mode || | |
1346 | !ubi->thread_enabled) { | |
1347 | set_current_state(TASK_INTERRUPTIBLE); | |
1348 | spin_unlock(&ubi->wl_lock); | |
1349 | schedule(); | |
1350 | continue; | |
1351 | } | |
1352 | spin_unlock(&ubi->wl_lock); | |
1353 | ||
1354 | err = do_work(ubi); | |
1355 | if (err) { | |
1356 | ubi_err("%s: work failed with error code %d", | |
1357 | ubi->bgt_name, err); | |
1358 | if (failures++ > WL_MAX_FAILURES) { | |
1359 | /* | |
1360 | * Too many failures, disable the thread and | |
1361 | * switch to read-only mode. | |
1362 | */ | |
1363 | ubi_msg("%s: %d consecutive failures", | |
1364 | ubi->bgt_name, WL_MAX_FAILURES); | |
1365 | ubi_ro_mode(ubi); | |
2ad49887 VG |
1366 | ubi->thread_enabled = 0; |
1367 | continue; | |
801c135c AB |
1368 | } |
1369 | } else | |
1370 | failures = 0; | |
1371 | ||
1372 | cond_resched(); | |
1373 | } | |
1374 | ||
801c135c AB |
1375 | dbg_wl("background thread \"%s\" is killed", ubi->bgt_name); |
1376 | return 0; | |
1377 | } | |
1378 | ||
1379 | /** | |
1380 | * cancel_pending - cancel all pending works. | |
1381 | * @ubi: UBI device description object | |
1382 | */ | |
1383 | static void cancel_pending(struct ubi_device *ubi) | |
1384 | { | |
1385 | while (!list_empty(&ubi->works)) { | |
1386 | struct ubi_work *wrk; | |
1387 | ||
1388 | wrk = list_entry(ubi->works.next, struct ubi_work, list); | |
1389 | list_del(&wrk->list); | |
1390 | wrk->func(ubi, wrk, 1); | |
1391 | ubi->works_count -= 1; | |
1392 | ubi_assert(ubi->works_count >= 0); | |
1393 | } | |
1394 | } | |
1395 | ||
1396 | /** | |
85c6e6e2 | 1397 | * ubi_wl_init_scan - initialize the WL sub-system using scanning information. |
801c135c AB |
1398 | * @ubi: UBI device description object |
1399 | * @si: scanning information | |
1400 | * | |
1401 | * This function returns zero in case of success, and a negative error code in | |
1402 | * case of failure. | |
1403 | */ | |
1404 | int ubi_wl_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si) | |
1405 | { | |
7b6c32da | 1406 | int err, i; |
801c135c AB |
1407 | struct rb_node *rb1, *rb2; |
1408 | struct ubi_scan_volume *sv; | |
1409 | struct ubi_scan_leb *seb, *tmp; | |
1410 | struct ubi_wl_entry *e; | |
1411 | ||
b86a2c56 | 1412 | ubi->used = ubi->erroneous = ubi->free = ubi->scrub = RB_ROOT; |
801c135c | 1413 | spin_lock_init(&ubi->wl_lock); |
43f9b25a | 1414 | mutex_init(&ubi->move_mutex); |
593dd33c | 1415 | init_rwsem(&ubi->work_sem); |
801c135c AB |
1416 | ubi->max_ec = si->max_ec; |
1417 | INIT_LIST_HEAD(&ubi->works); | |
1418 | ||
1419 | sprintf(ubi->bgt_name, UBI_BGT_NAME_PATTERN, ubi->ubi_num); | |
1420 | ||
801c135c AB |
1421 | err = -ENOMEM; |
1422 | ubi->lookuptbl = kzalloc(ubi->peb_count * sizeof(void *), GFP_KERNEL); | |
1423 | if (!ubi->lookuptbl) | |
cdfa788a | 1424 | return err; |
801c135c | 1425 | |
7b6c32da XX |
1426 | for (i = 0; i < UBI_PROT_QUEUE_LEN; i++) |
1427 | INIT_LIST_HEAD(&ubi->pq[i]); | |
1428 | ubi->pq_head = 0; | |
1429 | ||
801c135c AB |
1430 | list_for_each_entry_safe(seb, tmp, &si->erase, u.list) { |
1431 | cond_resched(); | |
1432 | ||
06b68ba1 | 1433 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1434 | if (!e) |
1435 | goto out_free; | |
1436 | ||
1437 | e->pnum = seb->pnum; | |
1438 | e->ec = seb->ec; | |
1439 | ubi->lookuptbl[e->pnum] = e; | |
1440 | if (schedule_erase(ubi, e, 0)) { | |
06b68ba1 | 1441 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1442 | goto out_free; |
1443 | } | |
1444 | } | |
1445 | ||
1446 | list_for_each_entry(seb, &si->free, u.list) { | |
1447 | cond_resched(); | |
1448 | ||
06b68ba1 | 1449 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1450 | if (!e) |
1451 | goto out_free; | |
1452 | ||
1453 | e->pnum = seb->pnum; | |
1454 | e->ec = seb->ec; | |
1455 | ubi_assert(e->ec >= 0); | |
5abde384 | 1456 | wl_tree_add(e, &ubi->free); |
801c135c AB |
1457 | ubi->lookuptbl[e->pnum] = e; |
1458 | } | |
1459 | ||
1460 | list_for_each_entry(seb, &si->corr, u.list) { | |
1461 | cond_resched(); | |
1462 | ||
06b68ba1 | 1463 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1464 | if (!e) |
1465 | goto out_free; | |
1466 | ||
1467 | e->pnum = seb->pnum; | |
1468 | e->ec = seb->ec; | |
1469 | ubi->lookuptbl[e->pnum] = e; | |
1470 | if (schedule_erase(ubi, e, 0)) { | |
06b68ba1 | 1471 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1472 | goto out_free; |
1473 | } | |
1474 | } | |
1475 | ||
1476 | ubi_rb_for_each_entry(rb1, sv, &si->volumes, rb) { | |
1477 | ubi_rb_for_each_entry(rb2, seb, &sv->root, u.rb) { | |
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->lookuptbl[e->pnum] = e; | |
1487 | if (!seb->scrub) { | |
1488 | dbg_wl("add PEB %d EC %d to the used tree", | |
1489 | e->pnum, e->ec); | |
5abde384 | 1490 | wl_tree_add(e, &ubi->used); |
801c135c AB |
1491 | } else { |
1492 | dbg_wl("add PEB %d EC %d to the scrub tree", | |
1493 | e->pnum, e->ec); | |
5abde384 | 1494 | wl_tree_add(e, &ubi->scrub); |
801c135c AB |
1495 | } |
1496 | } | |
1497 | } | |
1498 | ||
5abde384 | 1499 | if (ubi->avail_pebs < WL_RESERVED_PEBS) { |
801c135c AB |
1500 | ubi_err("no enough physical eraseblocks (%d, need %d)", |
1501 | ubi->avail_pebs, WL_RESERVED_PEBS); | |
1502 | goto out_free; | |
1503 | } | |
1504 | ubi->avail_pebs -= WL_RESERVED_PEBS; | |
1505 | ubi->rsvd_pebs += WL_RESERVED_PEBS; | |
1506 | ||
1507 | /* Schedule wear-leveling if needed */ | |
1508 | err = ensure_wear_leveling(ubi); | |
1509 | if (err) | |
1510 | goto out_free; | |
1511 | ||
1512 | return 0; | |
1513 | ||
1514 | out_free: | |
1515 | cancel_pending(ubi); | |
1516 | tree_destroy(&ubi->used); | |
1517 | tree_destroy(&ubi->free); | |
1518 | tree_destroy(&ubi->scrub); | |
1519 | kfree(ubi->lookuptbl); | |
801c135c AB |
1520 | return err; |
1521 | } | |
1522 | ||
1523 | /** | |
7b6c32da | 1524 | * protection_queue_destroy - destroy the protection queue. |
801c135c AB |
1525 | * @ubi: UBI device description object |
1526 | */ | |
7b6c32da | 1527 | static void protection_queue_destroy(struct ubi_device *ubi) |
801c135c | 1528 | { |
7b6c32da XX |
1529 | int i; |
1530 | struct ubi_wl_entry *e, *tmp; | |
801c135c | 1531 | |
7b6c32da XX |
1532 | for (i = 0; i < UBI_PROT_QUEUE_LEN; ++i) { |
1533 | list_for_each_entry_safe(e, tmp, &ubi->pq[i], u.list) { | |
1534 | list_del(&e->u.list); | |
1535 | kmem_cache_free(ubi_wl_entry_slab, e); | |
801c135c AB |
1536 | } |
1537 | } | |
1538 | } | |
1539 | ||
1540 | /** | |
85c6e6e2 | 1541 | * ubi_wl_close - close the wear-leveling sub-system. |
801c135c AB |
1542 | * @ubi: UBI device description object |
1543 | */ | |
1544 | void ubi_wl_close(struct ubi_device *ubi) | |
1545 | { | |
85c6e6e2 | 1546 | dbg_wl("close the WL sub-system"); |
801c135c | 1547 | cancel_pending(ubi); |
7b6c32da | 1548 | protection_queue_destroy(ubi); |
801c135c | 1549 | tree_destroy(&ubi->used); |
b86a2c56 | 1550 | tree_destroy(&ubi->erroneous); |
801c135c AB |
1551 | tree_destroy(&ubi->free); |
1552 | tree_destroy(&ubi->scrub); | |
1553 | kfree(ubi->lookuptbl); | |
801c135c AB |
1554 | } |
1555 | ||
1556 | #ifdef CONFIG_MTD_UBI_DEBUG_PARANOID | |
1557 | ||
1558 | /** | |
ebaaf1af | 1559 | * paranoid_check_ec - make sure that the erase counter of a PEB is correct. |
801c135c AB |
1560 | * @ubi: UBI device description object |
1561 | * @pnum: the physical eraseblock number to check | |
1562 | * @ec: the erase counter to check | |
1563 | * | |
1564 | * This function returns zero if the erase counter of physical eraseblock @pnum | |
1565 | * is equivalent to @ec, %1 if not, and a negative error code if an error | |
1566 | * occurred. | |
1567 | */ | |
e88d6e10 | 1568 | static int paranoid_check_ec(struct ubi_device *ubi, int pnum, int ec) |
801c135c AB |
1569 | { |
1570 | int err; | |
1571 | long long read_ec; | |
1572 | struct ubi_ec_hdr *ec_hdr; | |
1573 | ||
33818bbb | 1574 | ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_NOFS); |
801c135c AB |
1575 | if (!ec_hdr) |
1576 | return -ENOMEM; | |
1577 | ||
1578 | err = ubi_io_read_ec_hdr(ubi, pnum, ec_hdr, 0); | |
1579 | if (err && err != UBI_IO_BITFLIPS) { | |
1580 | /* The header does not have to exist */ | |
1581 | err = 0; | |
1582 | goto out_free; | |
1583 | } | |
1584 | ||
3261ebd7 | 1585 | read_ec = be64_to_cpu(ec_hdr->ec); |
801c135c AB |
1586 | if (ec != read_ec) { |
1587 | ubi_err("paranoid check failed for PEB %d", pnum); | |
1588 | ubi_err("read EC is %lld, should be %d", read_ec, ec); | |
1589 | ubi_dbg_dump_stack(); | |
1590 | err = 1; | |
1591 | } else | |
1592 | err = 0; | |
1593 | ||
1594 | out_free: | |
1595 | kfree(ec_hdr); | |
1596 | return err; | |
1597 | } | |
1598 | ||
1599 | /** | |
ebaaf1af | 1600 | * paranoid_check_in_wl_tree - check that wear-leveling entry is in WL RB-tree. |
801c135c AB |
1601 | * @e: the wear-leveling entry to check |
1602 | * @root: the root of the tree | |
1603 | * | |
ebaaf1af AB |
1604 | * This function returns zero if @e is in the @root RB-tree and %1 if it is |
1605 | * not. | |
801c135c AB |
1606 | */ |
1607 | static int paranoid_check_in_wl_tree(struct ubi_wl_entry *e, | |
1608 | struct rb_root *root) | |
1609 | { | |
1610 | if (in_wl_tree(e, root)) | |
1611 | return 0; | |
1612 | ||
1613 | ubi_err("paranoid check failed for PEB %d, EC %d, RB-tree %p ", | |
1614 | e->pnum, e->ec, root); | |
1615 | ubi_dbg_dump_stack(); | |
1616 | return 1; | |
1617 | } | |
1618 | ||
7b6c32da XX |
1619 | /** |
1620 | * paranoid_check_in_pq - check if wear-leveling entry is in the protection | |
1621 | * queue. | |
1622 | * @ubi: UBI device description object | |
1623 | * @e: the wear-leveling entry to check | |
1624 | * | |
1625 | * This function returns zero if @e is in @ubi->pq and %1 if it is not. | |
1626 | */ | |
1627 | static int paranoid_check_in_pq(struct ubi_device *ubi, struct ubi_wl_entry *e) | |
1628 | { | |
1629 | struct ubi_wl_entry *p; | |
1630 | int i; | |
1631 | ||
1632 | for (i = 0; i < UBI_PROT_QUEUE_LEN; ++i) | |
1633 | list_for_each_entry(p, &ubi->pq[i], u.list) | |
1634 | if (p == e) | |
1635 | return 0; | |
1636 | ||
1637 | ubi_err("paranoid check failed for PEB %d, EC %d, Protect queue", | |
1638 | e->pnum, e->ec); | |
1639 | ubi_dbg_dump_stack(); | |
1640 | return 1; | |
1641 | } | |
801c135c | 1642 | #endif /* CONFIG_MTD_UBI_DEBUG_PARANOID */ |