Commit | Line | Data |
---|---|---|
801c135c | 1 | /* |
d99383b0 | 2 | * @ubi: UBI device description object |
801c135c AB |
3 | * Copyright (c) International Business Machines Corp., 2006 |
4 | * | |
5 | * This program is free software; you can redistribute it and/or modify | |
6 | * it under the terms of the GNU General Public License as published by | |
7 | * the Free Software Foundation; either version 2 of the License, or | |
8 | * (at your option) any later version. | |
9 | * | |
10 | * This program is distributed in the hope that it will be useful, | |
11 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
12 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See | |
13 | * the GNU General Public License for more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License | |
16 | * along with this program; if not, write to the Free Software | |
17 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
18 | * | |
19 | * Authors: Artem Bityutskiy (Битюцкий Артём), Thomas Gleixner | |
20 | */ | |
21 | ||
22 | /* | |
85c6e6e2 | 23 | * UBI wear-leveling sub-system. |
801c135c | 24 | * |
85c6e6e2 | 25 | * This sub-system is responsible for wear-leveling. It works in terms of |
7b6c32da | 26 | * physical eraseblocks and erase counters and knows nothing about logical |
85c6e6e2 AB |
27 | * eraseblocks, volumes, etc. From this sub-system's perspective all physical |
28 | * eraseblocks are of two types - used and free. Used physical eraseblocks are | |
29 | * those that were "get" by the 'ubi_wl_get_peb()' function, and free physical | |
30 | * eraseblocks are those that were put by the 'ubi_wl_put_peb()' function. | |
801c135c AB |
31 | * |
32 | * Physical eraseblocks returned by 'ubi_wl_get_peb()' have only erase counter | |
85c6e6e2 | 33 | * header. The rest of the physical eraseblock contains only %0xFF bytes. |
801c135c | 34 | * |
85c6e6e2 | 35 | * When physical eraseblocks are returned to the WL sub-system by means of the |
801c135c AB |
36 | * 'ubi_wl_put_peb()' function, they are scheduled for erasure. The erasure is |
37 | * done asynchronously in context of the per-UBI device background thread, | |
85c6e6e2 | 38 | * which is also managed by the WL sub-system. |
801c135c AB |
39 | * |
40 | * The wear-leveling is ensured by means of moving the contents of used | |
41 | * physical eraseblocks with low erase counter to free physical eraseblocks | |
42 | * with high erase counter. | |
43 | * | |
44 | * The 'ubi_wl_get_peb()' function accepts data type hints which help to pick | |
45 | * an "optimal" physical eraseblock. For example, when it is known that the | |
46 | * physical eraseblock will be "put" soon because it contains short-term data, | |
85c6e6e2 AB |
47 | * the WL sub-system may pick a free physical eraseblock with low erase |
48 | * counter, and so forth. | |
801c135c | 49 | * |
85c6e6e2 AB |
50 | * If the WL sub-system fails to erase a physical eraseblock, it marks it as |
51 | * bad. | |
801c135c | 52 | * |
85c6e6e2 AB |
53 | * This sub-system is also responsible for scrubbing. If a bit-flip is detected |
54 | * in a physical eraseblock, it has to be moved. Technically this is the same | |
55 | * as moving it for wear-leveling reasons. | |
801c135c | 56 | * |
85c6e6e2 AB |
57 | * As it was said, for the UBI sub-system all physical eraseblocks are either |
58 | * "free" or "used". Free eraseblock are kept in the @wl->free RB-tree, while | |
b86a2c56 AB |
59 | * used eraseblocks are kept in @wl->used, @wl->erroneous, or @wl->scrub |
60 | * RB-trees, as well as (temporarily) in the @wl->pq queue. | |
7b6c32da XX |
61 | * |
62 | * When the WL sub-system returns a physical eraseblock, the physical | |
63 | * eraseblock is protected from being moved for some "time". For this reason, | |
64 | * the physical eraseblock is not directly moved from the @wl->free tree to the | |
65 | * @wl->used tree. There is a protection queue in between where this | |
66 | * physical eraseblock is temporarily stored (@wl->pq). | |
67 | * | |
68 | * All this protection stuff is needed because: | |
69 | * o we don't want to move physical eraseblocks just after we have given them | |
70 | * to the user; instead, we first want to let users fill them up with data; | |
71 | * | |
72 | * o there is a chance that the user will put the physical eraseblock very | |
73 | * soon, so it makes sense not to move it for some time, but wait; this is | |
74 | * especially important in case of "short term" physical eraseblocks. | |
75 | * | |
76 | * Physical eraseblocks stay protected only for limited time. But the "time" is | |
77 | * measured in erase cycles in this case. This is implemented with help of the | |
78 | * protection queue. Eraseblocks are put to the tail of this queue when they | |
79 | * are returned by the 'ubi_wl_get_peb()', and eraseblocks are removed from the | |
80 | * head of the queue on each erase operation (for any eraseblock). So the | |
81 | * length of the queue defines how may (global) erase cycles PEBs are protected. | |
82 | * | |
83 | * To put it differently, each physical eraseblock has 2 main states: free and | |
84 | * used. The former state corresponds to the @wl->free tree. The latter state | |
85 | * is split up on several sub-states: | |
86 | * o the WL movement is allowed (@wl->used tree); | |
815bc5f8 | 87 | * o the WL movement is disallowed (@wl->erroneous) because the PEB is |
b86a2c56 | 88 | * erroneous - e.g., there was a read error; |
7b6c32da XX |
89 | * o the WL movement is temporarily prohibited (@wl->pq queue); |
90 | * o scrubbing is needed (@wl->scrub tree). | |
91 | * | |
92 | * Depending on the sub-state, wear-leveling entries of the used physical | |
93 | * eraseblocks may be kept in one of those structures. | |
801c135c AB |
94 | * |
95 | * Note, in this implementation, we keep a small in-RAM object for each physical | |
96 | * eraseblock. This is surely not a scalable solution. But it appears to be good | |
97 | * enough for moderately large flashes and it is simple. In future, one may | |
85c6e6e2 | 98 | * re-work this sub-system and make it more scalable. |
801c135c | 99 | * |
85c6e6e2 AB |
100 | * At the moment this sub-system does not utilize the sequence number, which |
101 | * was introduced relatively recently. But it would be wise to do this because | |
102 | * the sequence number of a logical eraseblock characterizes how old is it. For | |
801c135c AB |
103 | * example, when we move a PEB with low erase counter, and we need to pick the |
104 | * target PEB, we pick a PEB with the highest EC if our PEB is "old" and we | |
105 | * pick target PEB with an average EC if our PEB is not very "old". This is a | |
85c6e6e2 | 106 | * room for future re-works of the WL sub-system. |
801c135c AB |
107 | */ |
108 | ||
109 | #include <linux/slab.h> | |
110 | #include <linux/crc32.h> | |
111 | #include <linux/freezer.h> | |
112 | #include <linux/kthread.h> | |
113 | #include "ubi.h" | |
114 | ||
115 | /* Number of physical eraseblocks reserved for wear-leveling purposes */ | |
116 | #define WL_RESERVED_PEBS 1 | |
117 | ||
801c135c AB |
118 | /* |
119 | * Maximum difference between two erase counters. If this threshold is | |
85c6e6e2 AB |
120 | * exceeded, the WL sub-system starts moving data from used physical |
121 | * eraseblocks with low erase counter to free physical eraseblocks with high | |
122 | * erase counter. | |
801c135c AB |
123 | */ |
124 | #define UBI_WL_THRESHOLD CONFIG_MTD_UBI_WL_THRESHOLD | |
125 | ||
126 | /* | |
85c6e6e2 | 127 | * When a physical eraseblock is moved, the WL sub-system has to pick the target |
801c135c AB |
128 | * physical eraseblock to move to. The simplest way would be just to pick the |
129 | * one with the highest erase counter. But in certain workloads this could lead | |
130 | * to an unlimited wear of one or few physical eraseblock. Indeed, imagine a | |
131 | * situation when the picked physical eraseblock is constantly erased after the | |
132 | * data is written to it. So, we have a constant which limits the highest erase | |
85c6e6e2 | 133 | * counter of the free physical eraseblock to pick. Namely, the WL sub-system |
025dfdaf | 134 | * does not pick eraseblocks with erase counter greater than the lowest erase |
801c135c AB |
135 | * counter plus %WL_FREE_MAX_DIFF. |
136 | */ | |
137 | #define WL_FREE_MAX_DIFF (2*UBI_WL_THRESHOLD) | |
138 | ||
139 | /* | |
140 | * Maximum number of consecutive background thread failures which is enough to | |
141 | * switch to read-only mode. | |
142 | */ | |
143 | #define WL_MAX_FAILURES 32 | |
144 | ||
801c135c AB |
145 | /** |
146 | * struct ubi_work - UBI work description data structure. | |
147 | * @list: a link in the list of pending works | |
148 | * @func: worker function | |
801c135c AB |
149 | * @e: physical eraseblock to erase |
150 | * @torture: if the physical eraseblock has to be tortured | |
151 | * | |
152 | * The @func pointer points to the worker function. If the @cancel argument is | |
153 | * not zero, the worker has to free the resources and exit immediately. The | |
154 | * worker has to return zero in case of success and a negative error code in | |
155 | * case of failure. | |
156 | */ | |
157 | struct ubi_work { | |
158 | struct list_head list; | |
159 | int (*func)(struct ubi_device *ubi, struct ubi_work *wrk, int cancel); | |
160 | /* The below fields are only relevant to erasure works */ | |
161 | struct ubi_wl_entry *e; | |
162 | int torture; | |
163 | }; | |
164 | ||
92d124f5 | 165 | #ifdef CONFIG_MTD_UBI_DEBUG |
e88d6e10 | 166 | static int paranoid_check_ec(struct ubi_device *ubi, int pnum, int ec); |
d99383b0 AB |
167 | static int paranoid_check_in_wl_tree(const struct ubi_device *ubi, |
168 | struct ubi_wl_entry *e, | |
801c135c | 169 | struct rb_root *root); |
d99383b0 AB |
170 | static int paranoid_check_in_pq(const struct ubi_device *ubi, |
171 | struct ubi_wl_entry *e); | |
801c135c AB |
172 | #else |
173 | #define paranoid_check_ec(ubi, pnum, ec) 0 | |
d99383b0 | 174 | #define paranoid_check_in_wl_tree(ubi, e, root) |
7b6c32da | 175 | #define paranoid_check_in_pq(ubi, e) 0 |
801c135c AB |
176 | #endif |
177 | ||
801c135c AB |
178 | /** |
179 | * wl_tree_add - add a wear-leveling entry to a WL RB-tree. | |
180 | * @e: the wear-leveling entry to add | |
181 | * @root: the root of the tree | |
182 | * | |
183 | * Note, we use (erase counter, physical eraseblock number) pairs as keys in | |
184 | * the @ubi->used and @ubi->free RB-trees. | |
185 | */ | |
186 | static void wl_tree_add(struct ubi_wl_entry *e, struct rb_root *root) | |
187 | { | |
188 | struct rb_node **p, *parent = NULL; | |
189 | ||
190 | p = &root->rb_node; | |
191 | while (*p) { | |
192 | struct ubi_wl_entry *e1; | |
193 | ||
194 | parent = *p; | |
23553b2c | 195 | e1 = rb_entry(parent, struct ubi_wl_entry, u.rb); |
801c135c AB |
196 | |
197 | if (e->ec < e1->ec) | |
198 | p = &(*p)->rb_left; | |
199 | else if (e->ec > e1->ec) | |
200 | p = &(*p)->rb_right; | |
201 | else { | |
202 | ubi_assert(e->pnum != e1->pnum); | |
203 | if (e->pnum < e1->pnum) | |
204 | p = &(*p)->rb_left; | |
205 | else | |
206 | p = &(*p)->rb_right; | |
207 | } | |
208 | } | |
209 | ||
23553b2c XX |
210 | rb_link_node(&e->u.rb, parent, p); |
211 | rb_insert_color(&e->u.rb, root); | |
801c135c AB |
212 | } |
213 | ||
801c135c AB |
214 | /** |
215 | * do_work - do one pending work. | |
216 | * @ubi: UBI device description object | |
217 | * | |
218 | * This function returns zero in case of success and a negative error code in | |
219 | * case of failure. | |
220 | */ | |
221 | static int do_work(struct ubi_device *ubi) | |
222 | { | |
223 | int err; | |
224 | struct ubi_work *wrk; | |
225 | ||
43f9b25a AB |
226 | cond_resched(); |
227 | ||
593dd33c AB |
228 | /* |
229 | * @ubi->work_sem is used to synchronize with the workers. Workers take | |
230 | * it in read mode, so many of them may be doing works at a time. But | |
231 | * the queue flush code has to be sure the whole queue of works is | |
232 | * done, and it takes the mutex in write mode. | |
233 | */ | |
234 | down_read(&ubi->work_sem); | |
801c135c | 235 | spin_lock(&ubi->wl_lock); |
801c135c AB |
236 | if (list_empty(&ubi->works)) { |
237 | spin_unlock(&ubi->wl_lock); | |
593dd33c | 238 | up_read(&ubi->work_sem); |
801c135c AB |
239 | return 0; |
240 | } | |
241 | ||
242 | wrk = list_entry(ubi->works.next, struct ubi_work, list); | |
243 | list_del(&wrk->list); | |
16f557ec AB |
244 | ubi->works_count -= 1; |
245 | ubi_assert(ubi->works_count >= 0); | |
801c135c AB |
246 | spin_unlock(&ubi->wl_lock); |
247 | ||
248 | /* | |
249 | * Call the worker function. Do not touch the work structure | |
250 | * after this call as it will have been freed or reused by that | |
251 | * time by the worker function. | |
252 | */ | |
253 | err = wrk->func(ubi, wrk, 0); | |
254 | if (err) | |
255 | ubi_err("work failed with error code %d", err); | |
593dd33c | 256 | up_read(&ubi->work_sem); |
16f557ec | 257 | |
801c135c AB |
258 | return err; |
259 | } | |
260 | ||
261 | /** | |
262 | * produce_free_peb - produce a free physical eraseblock. | |
263 | * @ubi: UBI device description object | |
264 | * | |
265 | * This function tries to make a free PEB by means of synchronous execution of | |
266 | * pending works. This may be needed if, for example the background thread is | |
267 | * disabled. Returns zero in case of success and a negative error code in case | |
268 | * of failure. | |
269 | */ | |
270 | static int produce_free_peb(struct ubi_device *ubi) | |
271 | { | |
272 | int err; | |
273 | ||
274 | spin_lock(&ubi->wl_lock); | |
5abde384 | 275 | while (!ubi->free.rb_node) { |
801c135c AB |
276 | spin_unlock(&ubi->wl_lock); |
277 | ||
278 | dbg_wl("do one work synchronously"); | |
279 | err = do_work(ubi); | |
280 | if (err) | |
281 | return err; | |
282 | ||
283 | spin_lock(&ubi->wl_lock); | |
284 | } | |
285 | spin_unlock(&ubi->wl_lock); | |
286 | ||
287 | return 0; | |
288 | } | |
289 | ||
290 | /** | |
291 | * in_wl_tree - check if wear-leveling entry is present in a WL RB-tree. | |
292 | * @e: the wear-leveling entry to check | |
293 | * @root: the root of the tree | |
294 | * | |
295 | * This function returns non-zero if @e is in the @root RB-tree and zero if it | |
296 | * is not. | |
297 | */ | |
298 | static int in_wl_tree(struct ubi_wl_entry *e, struct rb_root *root) | |
299 | { | |
300 | struct rb_node *p; | |
301 | ||
302 | p = root->rb_node; | |
303 | while (p) { | |
304 | struct ubi_wl_entry *e1; | |
305 | ||
23553b2c | 306 | e1 = rb_entry(p, struct ubi_wl_entry, u.rb); |
801c135c AB |
307 | |
308 | if (e->pnum == e1->pnum) { | |
309 | ubi_assert(e == e1); | |
310 | return 1; | |
311 | } | |
312 | ||
313 | if (e->ec < e1->ec) | |
314 | p = p->rb_left; | |
315 | else if (e->ec > e1->ec) | |
316 | p = p->rb_right; | |
317 | else { | |
318 | ubi_assert(e->pnum != e1->pnum); | |
319 | if (e->pnum < e1->pnum) | |
320 | p = p->rb_left; | |
321 | else | |
322 | p = p->rb_right; | |
323 | } | |
324 | } | |
325 | ||
326 | return 0; | |
327 | } | |
328 | ||
329 | /** | |
7b6c32da | 330 | * prot_queue_add - add physical eraseblock to the protection queue. |
801c135c AB |
331 | * @ubi: UBI device description object |
332 | * @e: the physical eraseblock to add | |
801c135c | 333 | * |
7b6c32da XX |
334 | * This function adds @e to the tail of the protection queue @ubi->pq, where |
335 | * @e will stay for %UBI_PROT_QUEUE_LEN erase operations and will be | |
336 | * temporarily protected from the wear-leveling worker. Note, @wl->lock has to | |
337 | * be locked. | |
801c135c | 338 | */ |
7b6c32da | 339 | static void prot_queue_add(struct ubi_device *ubi, struct ubi_wl_entry *e) |
801c135c | 340 | { |
7b6c32da | 341 | int pq_tail = ubi->pq_head - 1; |
801c135c | 342 | |
7b6c32da XX |
343 | if (pq_tail < 0) |
344 | pq_tail = UBI_PROT_QUEUE_LEN - 1; | |
345 | ubi_assert(pq_tail >= 0 && pq_tail < UBI_PROT_QUEUE_LEN); | |
346 | list_add_tail(&e->u.list, &ubi->pq[pq_tail]); | |
347 | dbg_wl("added PEB %d EC %d to the protection queue", e->pnum, e->ec); | |
801c135c AB |
348 | } |
349 | ||
350 | /** | |
351 | * find_wl_entry - find wear-leveling entry closest to certain erase counter. | |
352 | * @root: the RB-tree where to look for | |
353 | * @max: highest possible erase counter | |
354 | * | |
355 | * This function looks for a wear leveling entry with erase counter closest to | |
3f502622 | 356 | * @max and less than @max. |
801c135c AB |
357 | */ |
358 | static struct ubi_wl_entry *find_wl_entry(struct rb_root *root, int max) | |
359 | { | |
360 | struct rb_node *p; | |
361 | struct ubi_wl_entry *e; | |
362 | ||
23553b2c | 363 | e = rb_entry(rb_first(root), struct ubi_wl_entry, u.rb); |
801c135c AB |
364 | max += e->ec; |
365 | ||
366 | p = root->rb_node; | |
367 | while (p) { | |
368 | struct ubi_wl_entry *e1; | |
369 | ||
23553b2c | 370 | e1 = rb_entry(p, struct ubi_wl_entry, u.rb); |
801c135c AB |
371 | if (e1->ec >= max) |
372 | p = p->rb_left; | |
373 | else { | |
374 | p = p->rb_right; | |
375 | e = e1; | |
376 | } | |
377 | } | |
378 | ||
379 | return e; | |
380 | } | |
381 | ||
382 | /** | |
383 | * ubi_wl_get_peb - get a physical eraseblock. | |
384 | * @ubi: UBI device description object | |
385 | * @dtype: type of data which will be stored in this physical eraseblock | |
386 | * | |
387 | * This function returns a physical eraseblock in case of success and a | |
388 | * negative error code in case of failure. Might sleep. | |
389 | */ | |
390 | int ubi_wl_get_peb(struct ubi_device *ubi, int dtype) | |
391 | { | |
7b6c32da | 392 | int err, medium_ec; |
801c135c | 393 | struct ubi_wl_entry *e, *first, *last; |
801c135c AB |
394 | |
395 | ubi_assert(dtype == UBI_LONGTERM || dtype == UBI_SHORTTERM || | |
396 | dtype == UBI_UNKNOWN); | |
397 | ||
801c135c AB |
398 | retry: |
399 | spin_lock(&ubi->wl_lock); | |
5abde384 | 400 | if (!ubi->free.rb_node) { |
801c135c AB |
401 | if (ubi->works_count == 0) { |
402 | ubi_assert(list_empty(&ubi->works)); | |
403 | ubi_err("no free eraseblocks"); | |
404 | spin_unlock(&ubi->wl_lock); | |
801c135c AB |
405 | return -ENOSPC; |
406 | } | |
407 | spin_unlock(&ubi->wl_lock); | |
408 | ||
409 | err = produce_free_peb(ubi); | |
7b6c32da | 410 | if (err < 0) |
801c135c | 411 | return err; |
801c135c AB |
412 | goto retry; |
413 | } | |
414 | ||
415 | switch (dtype) { | |
9c9ec147 AB |
416 | case UBI_LONGTERM: |
417 | /* | |
418 | * For long term data we pick a physical eraseblock with high | |
419 | * erase counter. But the highest erase counter we can pick is | |
420 | * bounded by the the lowest erase counter plus | |
421 | * %WL_FREE_MAX_DIFF. | |
422 | */ | |
423 | e = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); | |
9c9ec147 AB |
424 | break; |
425 | case UBI_UNKNOWN: | |
426 | /* | |
427 | * For unknown data we pick a physical eraseblock with medium | |
428 | * erase counter. But we by no means can pick a physical | |
429 | * eraseblock with erase counter greater or equivalent than the | |
430 | * lowest erase counter plus %WL_FREE_MAX_DIFF. | |
431 | */ | |
23553b2c XX |
432 | first = rb_entry(rb_first(&ubi->free), struct ubi_wl_entry, |
433 | u.rb); | |
434 | last = rb_entry(rb_last(&ubi->free), struct ubi_wl_entry, u.rb); | |
801c135c | 435 | |
9c9ec147 AB |
436 | if (last->ec - first->ec < WL_FREE_MAX_DIFF) |
437 | e = rb_entry(ubi->free.rb_node, | |
23553b2c | 438 | struct ubi_wl_entry, u.rb); |
9c9ec147 AB |
439 | else { |
440 | medium_ec = (first->ec + WL_FREE_MAX_DIFF)/2; | |
441 | e = find_wl_entry(&ubi->free, medium_ec); | |
442 | } | |
9c9ec147 AB |
443 | break; |
444 | case UBI_SHORTTERM: | |
445 | /* | |
446 | * For short term data we pick a physical eraseblock with the | |
447 | * lowest erase counter as we expect it will be erased soon. | |
448 | */ | |
23553b2c | 449 | e = rb_entry(rb_first(&ubi->free), struct ubi_wl_entry, u.rb); |
9c9ec147 AB |
450 | break; |
451 | default: | |
9c9ec147 | 452 | BUG(); |
801c135c AB |
453 | } |
454 | ||
d99383b0 | 455 | paranoid_check_in_wl_tree(ubi, e, &ubi->free); |
7b6c32da | 456 | |
801c135c | 457 | /* |
7b6c32da | 458 | * Move the physical eraseblock to the protection queue where it will |
801c135c AB |
459 | * be protected from being moved for some time. |
460 | */ | |
23553b2c | 461 | rb_erase(&e->u.rb, &ubi->free); |
7b6c32da XX |
462 | dbg_wl("PEB %d EC %d", e->pnum, e->ec); |
463 | prot_queue_add(ubi, e); | |
801c135c | 464 | spin_unlock(&ubi->wl_lock); |
40a71a87 AB |
465 | |
466 | err = ubi_dbg_check_all_ff(ubi, e->pnum, ubi->vid_hdr_aloffset, | |
467 | ubi->peb_size - ubi->vid_hdr_aloffset); | |
468 | if (err) { | |
1398788f | 469 | ubi_err("new PEB %d does not contain all 0xFF bytes", e->pnum); |
adbf05e3 | 470 | return err; |
40a71a87 AB |
471 | } |
472 | ||
801c135c AB |
473 | return e->pnum; |
474 | } | |
475 | ||
476 | /** | |
7b6c32da | 477 | * prot_queue_del - remove a physical eraseblock from the protection queue. |
801c135c AB |
478 | * @ubi: UBI device description object |
479 | * @pnum: the physical eraseblock to remove | |
43f9b25a | 480 | * |
7b6c32da XX |
481 | * This function deletes PEB @pnum from the protection queue and returns zero |
482 | * in case of success and %-ENODEV if the PEB was not found. | |
801c135c | 483 | */ |
7b6c32da | 484 | static int prot_queue_del(struct ubi_device *ubi, int pnum) |
801c135c | 485 | { |
7b6c32da | 486 | struct ubi_wl_entry *e; |
801c135c | 487 | |
7b6c32da XX |
488 | e = ubi->lookuptbl[pnum]; |
489 | if (!e) | |
490 | return -ENODEV; | |
801c135c | 491 | |
7b6c32da XX |
492 | if (paranoid_check_in_pq(ubi, e)) |
493 | return -ENODEV; | |
43f9b25a | 494 | |
7b6c32da XX |
495 | list_del(&e->u.list); |
496 | dbg_wl("deleted PEB %d from the protection queue", e->pnum); | |
43f9b25a | 497 | return 0; |
801c135c AB |
498 | } |
499 | ||
500 | /** | |
501 | * sync_erase - synchronously erase a physical eraseblock. | |
502 | * @ubi: UBI device description object | |
503 | * @e: the the physical eraseblock to erase | |
504 | * @torture: if the physical eraseblock has to be tortured | |
505 | * | |
506 | * This function returns zero in case of success and a negative error code in | |
507 | * case of failure. | |
508 | */ | |
9c9ec147 AB |
509 | static int sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e, |
510 | int torture) | |
801c135c AB |
511 | { |
512 | int err; | |
513 | struct ubi_ec_hdr *ec_hdr; | |
514 | unsigned long long ec = e->ec; | |
515 | ||
516 | dbg_wl("erase PEB %d, old EC %llu", e->pnum, ec); | |
517 | ||
518 | err = paranoid_check_ec(ubi, e->pnum, e->ec); | |
adbf05e3 | 519 | if (err) |
801c135c AB |
520 | return -EINVAL; |
521 | ||
33818bbb | 522 | ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_NOFS); |
801c135c AB |
523 | if (!ec_hdr) |
524 | return -ENOMEM; | |
525 | ||
526 | err = ubi_io_sync_erase(ubi, e->pnum, torture); | |
527 | if (err < 0) | |
528 | goto out_free; | |
529 | ||
530 | ec += err; | |
531 | if (ec > UBI_MAX_ERASECOUNTER) { | |
532 | /* | |
533 | * Erase counter overflow. Upgrade UBI and use 64-bit | |
534 | * erase counters internally. | |
535 | */ | |
536 | ubi_err("erase counter overflow at PEB %d, EC %llu", | |
537 | e->pnum, ec); | |
538 | err = -EINVAL; | |
539 | goto out_free; | |
540 | } | |
541 | ||
542 | dbg_wl("erased PEB %d, new EC %llu", e->pnum, ec); | |
543 | ||
3261ebd7 | 544 | ec_hdr->ec = cpu_to_be64(ec); |
801c135c AB |
545 | |
546 | err = ubi_io_write_ec_hdr(ubi, e->pnum, ec_hdr); | |
547 | if (err) | |
548 | goto out_free; | |
549 | ||
550 | e->ec = ec; | |
551 | spin_lock(&ubi->wl_lock); | |
552 | if (e->ec > ubi->max_ec) | |
553 | ubi->max_ec = e->ec; | |
554 | spin_unlock(&ubi->wl_lock); | |
555 | ||
556 | out_free: | |
557 | kfree(ec_hdr); | |
558 | return err; | |
559 | } | |
560 | ||
561 | /** | |
7b6c32da | 562 | * serve_prot_queue - check if it is time to stop protecting PEBs. |
801c135c AB |
563 | * @ubi: UBI device description object |
564 | * | |
7b6c32da XX |
565 | * This function is called after each erase operation and removes PEBs from the |
566 | * tail of the protection queue. These PEBs have been protected for long enough | |
567 | * and should be moved to the used tree. | |
801c135c | 568 | */ |
7b6c32da | 569 | static void serve_prot_queue(struct ubi_device *ubi) |
801c135c | 570 | { |
7b6c32da XX |
571 | struct ubi_wl_entry *e, *tmp; |
572 | int count; | |
801c135c AB |
573 | |
574 | /* | |
575 | * There may be several protected physical eraseblock to remove, | |
576 | * process them all. | |
577 | */ | |
7b6c32da XX |
578 | repeat: |
579 | count = 0; | |
580 | spin_lock(&ubi->wl_lock); | |
581 | list_for_each_entry_safe(e, tmp, &ubi->pq[ubi->pq_head], u.list) { | |
582 | dbg_wl("PEB %d EC %d protection over, move to used tree", | |
583 | e->pnum, e->ec); | |
801c135c | 584 | |
7b6c32da XX |
585 | list_del(&e->u.list); |
586 | wl_tree_add(e, &ubi->used); | |
587 | if (count++ > 32) { | |
588 | /* | |
589 | * Let's be nice and avoid holding the spinlock for | |
590 | * too long. | |
591 | */ | |
801c135c | 592 | spin_unlock(&ubi->wl_lock); |
7b6c32da XX |
593 | cond_resched(); |
594 | goto repeat; | |
801c135c | 595 | } |
801c135c | 596 | } |
7b6c32da XX |
597 | |
598 | ubi->pq_head += 1; | |
599 | if (ubi->pq_head == UBI_PROT_QUEUE_LEN) | |
600 | ubi->pq_head = 0; | |
601 | ubi_assert(ubi->pq_head >= 0 && ubi->pq_head < UBI_PROT_QUEUE_LEN); | |
602 | spin_unlock(&ubi->wl_lock); | |
801c135c AB |
603 | } |
604 | ||
605 | /** | |
606 | * schedule_ubi_work - schedule a work. | |
607 | * @ubi: UBI device description object | |
608 | * @wrk: the work to schedule | |
609 | * | |
7b6c32da XX |
610 | * This function adds a work defined by @wrk to the tail of the pending works |
611 | * list. | |
801c135c AB |
612 | */ |
613 | static void schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk) | |
614 | { | |
615 | spin_lock(&ubi->wl_lock); | |
616 | list_add_tail(&wrk->list, &ubi->works); | |
617 | ubi_assert(ubi->works_count >= 0); | |
618 | ubi->works_count += 1; | |
27a0f2a3 | 619 | if (ubi->thread_enabled && !ubi_dbg_is_bgt_disabled(ubi)) |
801c135c AB |
620 | wake_up_process(ubi->bgt_thread); |
621 | spin_unlock(&ubi->wl_lock); | |
622 | } | |
623 | ||
624 | static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk, | |
625 | int cancel); | |
626 | ||
627 | /** | |
628 | * schedule_erase - schedule an erase work. | |
629 | * @ubi: UBI device description object | |
630 | * @e: the WL entry of the physical eraseblock to erase | |
631 | * @torture: if the physical eraseblock has to be tortured | |
632 | * | |
633 | * This function returns zero in case of success and a %-ENOMEM in case of | |
634 | * failure. | |
635 | */ | |
636 | static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e, | |
637 | int torture) | |
638 | { | |
639 | struct ubi_work *wl_wrk; | |
640 | ||
641 | dbg_wl("schedule erasure of PEB %d, EC %d, torture %d", | |
642 | e->pnum, e->ec, torture); | |
643 | ||
33818bbb | 644 | wl_wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS); |
801c135c AB |
645 | if (!wl_wrk) |
646 | return -ENOMEM; | |
647 | ||
648 | wl_wrk->func = &erase_worker; | |
649 | wl_wrk->e = e; | |
650 | wl_wrk->torture = torture; | |
651 | ||
652 | schedule_ubi_work(ubi, wl_wrk); | |
653 | return 0; | |
654 | } | |
655 | ||
656 | /** | |
657 | * wear_leveling_worker - wear-leveling worker function. | |
658 | * @ubi: UBI device description object | |
659 | * @wrk: the work object | |
660 | * @cancel: non-zero if the worker has to free memory and exit | |
661 | * | |
662 | * This function copies a more worn out physical eraseblock to a less worn out | |
663 | * one. Returns zero in case of success and a negative error code in case of | |
664 | * failure. | |
665 | */ | |
666 | static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk, | |
667 | int cancel) | |
668 | { | |
b86a2c56 | 669 | int err, scrubbing = 0, torture = 0, protect = 0, erroneous = 0; |
9c259a52 | 670 | int vol_id = -1, uninitialized_var(lnum); |
801c135c AB |
671 | struct ubi_wl_entry *e1, *e2; |
672 | struct ubi_vid_hdr *vid_hdr; | |
673 | ||
674 | kfree(wrk); | |
801c135c AB |
675 | if (cancel) |
676 | return 0; | |
677 | ||
33818bbb | 678 | vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS); |
801c135c AB |
679 | if (!vid_hdr) |
680 | return -ENOMEM; | |
681 | ||
43f9b25a | 682 | mutex_lock(&ubi->move_mutex); |
801c135c | 683 | spin_lock(&ubi->wl_lock); |
43f9b25a AB |
684 | ubi_assert(!ubi->move_from && !ubi->move_to); |
685 | ubi_assert(!ubi->move_to_put); | |
801c135c | 686 | |
43f9b25a | 687 | if (!ubi->free.rb_node || |
5abde384 | 688 | (!ubi->used.rb_node && !ubi->scrub.rb_node)) { |
801c135c | 689 | /* |
43f9b25a AB |
690 | * No free physical eraseblocks? Well, they must be waiting in |
691 | * the queue to be erased. Cancel movement - it will be | |
692 | * triggered again when a free physical eraseblock appears. | |
801c135c AB |
693 | * |
694 | * No used physical eraseblocks? They must be temporarily | |
695 | * protected from being moved. They will be moved to the | |
696 | * @ubi->used tree later and the wear-leveling will be | |
697 | * triggered again. | |
698 | */ | |
699 | dbg_wl("cancel WL, a list is empty: free %d, used %d", | |
5abde384 | 700 | !ubi->free.rb_node, !ubi->used.rb_node); |
43f9b25a | 701 | goto out_cancel; |
801c135c AB |
702 | } |
703 | ||
5abde384 | 704 | if (!ubi->scrub.rb_node) { |
801c135c AB |
705 | /* |
706 | * Now pick the least worn-out used physical eraseblock and a | |
707 | * highly worn-out free physical eraseblock. If the erase | |
708 | * counters differ much enough, start wear-leveling. | |
709 | */ | |
23553b2c | 710 | e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, u.rb); |
801c135c AB |
711 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); |
712 | ||
713 | if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD)) { | |
714 | dbg_wl("no WL needed: min used EC %d, max free EC %d", | |
715 | e1->ec, e2->ec); | |
43f9b25a | 716 | goto out_cancel; |
801c135c | 717 | } |
d99383b0 | 718 | paranoid_check_in_wl_tree(ubi, e1, &ubi->used); |
23553b2c | 719 | rb_erase(&e1->u.rb, &ubi->used); |
801c135c AB |
720 | dbg_wl("move PEB %d EC %d to PEB %d EC %d", |
721 | e1->pnum, e1->ec, e2->pnum, e2->ec); | |
722 | } else { | |
43f9b25a AB |
723 | /* Perform scrubbing */ |
724 | scrubbing = 1; | |
23553b2c | 725 | e1 = rb_entry(rb_first(&ubi->scrub), struct ubi_wl_entry, u.rb); |
801c135c | 726 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); |
d99383b0 | 727 | paranoid_check_in_wl_tree(ubi, e1, &ubi->scrub); |
23553b2c | 728 | rb_erase(&e1->u.rb, &ubi->scrub); |
801c135c AB |
729 | dbg_wl("scrub PEB %d to PEB %d", e1->pnum, e2->pnum); |
730 | } | |
731 | ||
d99383b0 | 732 | paranoid_check_in_wl_tree(ubi, e2, &ubi->free); |
23553b2c | 733 | rb_erase(&e2->u.rb, &ubi->free); |
801c135c AB |
734 | ubi->move_from = e1; |
735 | ubi->move_to = e2; | |
736 | spin_unlock(&ubi->wl_lock); | |
737 | ||
738 | /* | |
739 | * Now we are going to copy physical eraseblock @e1->pnum to @e2->pnum. | |
740 | * We so far do not know which logical eraseblock our physical | |
741 | * eraseblock (@e1) belongs to. We have to read the volume identifier | |
742 | * header first. | |
43f9b25a AB |
743 | * |
744 | * Note, we are protected from this PEB being unmapped and erased. The | |
745 | * 'ubi_wl_put_peb()' would wait for moving to be finished if the PEB | |
746 | * which is being moved was unmapped. | |
801c135c AB |
747 | */ |
748 | ||
749 | err = ubi_io_read_vid_hdr(ubi, e1->pnum, vid_hdr, 0); | |
750 | if (err && err != UBI_IO_BITFLIPS) { | |
74d82d26 | 751 | if (err == UBI_IO_FF) { |
801c135c AB |
752 | /* |
753 | * We are trying to move PEB without a VID header. UBI | |
754 | * always write VID headers shortly after the PEB was | |
87960c0b AB |
755 | * given, so we have a situation when it has not yet |
756 | * had a chance to write it, because it was preempted. | |
757 | * So add this PEB to the protection queue so far, | |
815bc5f8 AB |
758 | * because presumably more data will be written there |
759 | * (including the missing VID header), and then we'll | |
87960c0b | 760 | * move it. |
801c135c AB |
761 | */ |
762 | dbg_wl("PEB %d has no VID header", e1->pnum); | |
87960c0b | 763 | protect = 1; |
43f9b25a | 764 | goto out_not_moved; |
92e1a7d9 AB |
765 | } else if (err == UBI_IO_FF_BITFLIPS) { |
766 | /* | |
767 | * The same situation as %UBI_IO_FF, but bit-flips were | |
768 | * detected. It is better to schedule this PEB for | |
769 | * scrubbing. | |
770 | */ | |
771 | dbg_wl("PEB %d has no VID header but has bit-flips", | |
772 | e1->pnum); | |
773 | scrubbing = 1; | |
774 | goto out_not_moved; | |
801c135c | 775 | } |
43f9b25a AB |
776 | |
777 | ubi_err("error %d while reading VID header from PEB %d", | |
778 | err, e1->pnum); | |
43f9b25a | 779 | goto out_error; |
801c135c AB |
780 | } |
781 | ||
9c259a52 AB |
782 | vol_id = be32_to_cpu(vid_hdr->vol_id); |
783 | lnum = be32_to_cpu(vid_hdr->lnum); | |
784 | ||
801c135c AB |
785 | err = ubi_eba_copy_leb(ubi, e1->pnum, e2->pnum, vid_hdr); |
786 | if (err) { | |
87960c0b AB |
787 | if (err == MOVE_CANCEL_RACE) { |
788 | /* | |
789 | * The LEB has not been moved because the volume is | |
790 | * being deleted or the PEB has been put meanwhile. We | |
791 | * should prevent this PEB from being selected for | |
792 | * wear-leveling movement again, so put it to the | |
793 | * protection queue. | |
794 | */ | |
795 | protect = 1; | |
796 | goto out_not_moved; | |
797 | } | |
798 | ||
b86a2c56 AB |
799 | if (err == MOVE_CANCEL_BITFLIPS || err == MOVE_TARGET_WR_ERR || |
800 | err == MOVE_TARGET_RD_ERR) { | |
9c259a52 AB |
801 | /* |
802 | * Target PEB had bit-flips or write error - torture it. | |
803 | */ | |
6fa6f5bb | 804 | torture = 1; |
43f9b25a | 805 | goto out_not_moved; |
6fa6f5bb | 806 | } |
87960c0b | 807 | |
b86a2c56 AB |
808 | if (err == MOVE_SOURCE_RD_ERR) { |
809 | /* | |
810 | * An error happened while reading the source PEB. Do | |
811 | * not switch to R/O mode in this case, and give the | |
812 | * upper layers a possibility to recover from this, | |
813 | * e.g. by unmapping corresponding LEB. Instead, just | |
815bc5f8 AB |
814 | * put this PEB to the @ubi->erroneous list to prevent |
815 | * UBI from trying to move it over and over again. | |
b86a2c56 AB |
816 | */ |
817 | if (ubi->erroneous_peb_count > ubi->max_erroneous) { | |
818 | ubi_err("too many erroneous eraseblocks (%d)", | |
819 | ubi->erroneous_peb_count); | |
820 | goto out_error; | |
821 | } | |
822 | erroneous = 1; | |
823 | goto out_not_moved; | |
824 | } | |
825 | ||
90bf0265 AB |
826 | if (err < 0) |
827 | goto out_error; | |
43f9b25a | 828 | |
87960c0b | 829 | ubi_assert(0); |
801c135c AB |
830 | } |
831 | ||
6a8f483f | 832 | /* The PEB has been successfully moved */ |
6a8f483f | 833 | if (scrubbing) |
9c259a52 AB |
834 | ubi_msg("scrubbed PEB %d (LEB %d:%d), data moved to PEB %d", |
835 | e1->pnum, vol_id, lnum, e2->pnum); | |
836 | ubi_free_vid_hdr(ubi, vid_hdr); | |
8c1e6ee1 | 837 | |
801c135c | 838 | spin_lock(&ubi->wl_lock); |
3c98b0a0 | 839 | if (!ubi->move_to_put) { |
5abde384 | 840 | wl_tree_add(e2, &ubi->used); |
3c98b0a0 AB |
841 | e2 = NULL; |
842 | } | |
801c135c | 843 | ubi->move_from = ubi->move_to = NULL; |
43f9b25a | 844 | ubi->move_to_put = ubi->wl_scheduled = 0; |
801c135c AB |
845 | spin_unlock(&ubi->wl_lock); |
846 | ||
6a8f483f | 847 | err = schedule_erase(ubi, e1, 0); |
3c98b0a0 | 848 | if (err) { |
87960c0b | 849 | kmem_cache_free(ubi_wl_entry_slab, e1); |
21d08bbc AB |
850 | if (e2) |
851 | kmem_cache_free(ubi_wl_entry_slab, e2); | |
87960c0b | 852 | goto out_ro; |
3c98b0a0 | 853 | } |
6a8f483f | 854 | |
3c98b0a0 | 855 | if (e2) { |
801c135c AB |
856 | /* |
857 | * Well, the target PEB was put meanwhile, schedule it for | |
858 | * erasure. | |
859 | */ | |
9c259a52 AB |
860 | dbg_wl("PEB %d (LEB %d:%d) was put meanwhile, erase", |
861 | e2->pnum, vol_id, lnum); | |
801c135c | 862 | err = schedule_erase(ubi, e2, 0); |
87960c0b AB |
863 | if (err) { |
864 | kmem_cache_free(ubi_wl_entry_slab, e2); | |
865 | goto out_ro; | |
866 | } | |
801c135c AB |
867 | } |
868 | ||
801c135c | 869 | dbg_wl("done"); |
43f9b25a AB |
870 | mutex_unlock(&ubi->move_mutex); |
871 | return 0; | |
801c135c AB |
872 | |
873 | /* | |
43f9b25a AB |
874 | * For some reasons the LEB was not moved, might be an error, might be |
875 | * something else. @e1 was not changed, so return it back. @e2 might | |
6fa6f5bb | 876 | * have been changed, schedule it for erasure. |
801c135c | 877 | */ |
43f9b25a | 878 | out_not_moved: |
9c259a52 AB |
879 | if (vol_id != -1) |
880 | dbg_wl("cancel moving PEB %d (LEB %d:%d) to PEB %d (%d)", | |
881 | e1->pnum, vol_id, lnum, e2->pnum, err); | |
882 | else | |
883 | dbg_wl("cancel moving PEB %d to PEB %d (%d)", | |
884 | e1->pnum, e2->pnum, err); | |
801c135c | 885 | spin_lock(&ubi->wl_lock); |
87960c0b AB |
886 | if (protect) |
887 | prot_queue_add(ubi, e1); | |
b86a2c56 AB |
888 | else if (erroneous) { |
889 | wl_tree_add(e1, &ubi->erroneous); | |
890 | ubi->erroneous_peb_count += 1; | |
891 | } else if (scrubbing) | |
43f9b25a | 892 | wl_tree_add(e1, &ubi->scrub); |
801c135c | 893 | else |
5abde384 | 894 | wl_tree_add(e1, &ubi->used); |
6fa6f5bb | 895 | ubi_assert(!ubi->move_to_put); |
801c135c | 896 | ubi->move_from = ubi->move_to = NULL; |
6fa6f5bb | 897 | ubi->wl_scheduled = 0; |
801c135c AB |
898 | spin_unlock(&ubi->wl_lock); |
899 | ||
87960c0b | 900 | ubi_free_vid_hdr(ubi, vid_hdr); |
6fa6f5bb | 901 | err = schedule_erase(ubi, e2, torture); |
87960c0b AB |
902 | if (err) { |
903 | kmem_cache_free(ubi_wl_entry_slab, e2); | |
904 | goto out_ro; | |
905 | } | |
43f9b25a AB |
906 | mutex_unlock(&ubi->move_mutex); |
907 | return 0; | |
908 | ||
909 | out_error: | |
9c259a52 AB |
910 | if (vol_id != -1) |
911 | ubi_err("error %d while moving PEB %d to PEB %d", | |
912 | err, e1->pnum, e2->pnum); | |
913 | else | |
914 | ubi_err("error %d while moving PEB %d (LEB %d:%d) to PEB %d", | |
915 | err, e1->pnum, vol_id, lnum, e2->pnum); | |
43f9b25a AB |
916 | spin_lock(&ubi->wl_lock); |
917 | ubi->move_from = ubi->move_to = NULL; | |
918 | ubi->move_to_put = ubi->wl_scheduled = 0; | |
919 | spin_unlock(&ubi->wl_lock); | |
920 | ||
87960c0b AB |
921 | ubi_free_vid_hdr(ubi, vid_hdr); |
922 | kmem_cache_free(ubi_wl_entry_slab, e1); | |
923 | kmem_cache_free(ubi_wl_entry_slab, e2); | |
43f9b25a | 924 | |
87960c0b AB |
925 | out_ro: |
926 | ubi_ro_mode(ubi); | |
43f9b25a | 927 | mutex_unlock(&ubi->move_mutex); |
87960c0b AB |
928 | ubi_assert(err != 0); |
929 | return err < 0 ? err : -EIO; | |
43f9b25a AB |
930 | |
931 | out_cancel: | |
932 | ubi->wl_scheduled = 0; | |
933 | spin_unlock(&ubi->wl_lock); | |
934 | mutex_unlock(&ubi->move_mutex); | |
935 | ubi_free_vid_hdr(ubi, vid_hdr); | |
936 | return 0; | |
801c135c AB |
937 | } |
938 | ||
939 | /** | |
940 | * ensure_wear_leveling - schedule wear-leveling if it is needed. | |
941 | * @ubi: UBI device description object | |
942 | * | |
943 | * This function checks if it is time to start wear-leveling and schedules it | |
944 | * if yes. This function returns zero in case of success and a negative error | |
945 | * code in case of failure. | |
946 | */ | |
947 | static int ensure_wear_leveling(struct ubi_device *ubi) | |
948 | { | |
949 | int err = 0; | |
950 | struct ubi_wl_entry *e1; | |
951 | struct ubi_wl_entry *e2; | |
952 | struct ubi_work *wrk; | |
953 | ||
954 | spin_lock(&ubi->wl_lock); | |
955 | if (ubi->wl_scheduled) | |
956 | /* Wear-leveling is already in the work queue */ | |
957 | goto out_unlock; | |
958 | ||
959 | /* | |
960 | * If the ubi->scrub tree is not empty, scrubbing is needed, and the | |
961 | * the WL worker has to be scheduled anyway. | |
962 | */ | |
5abde384 AB |
963 | if (!ubi->scrub.rb_node) { |
964 | if (!ubi->used.rb_node || !ubi->free.rb_node) | |
801c135c AB |
965 | /* No physical eraseblocks - no deal */ |
966 | goto out_unlock; | |
967 | ||
968 | /* | |
969 | * We schedule wear-leveling only if the difference between the | |
970 | * lowest erase counter of used physical eraseblocks and a high | |
025dfdaf | 971 | * erase counter of free physical eraseblocks is greater than |
801c135c AB |
972 | * %UBI_WL_THRESHOLD. |
973 | */ | |
23553b2c | 974 | e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, u.rb); |
801c135c AB |
975 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); |
976 | ||
977 | if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD)) | |
978 | goto out_unlock; | |
979 | dbg_wl("schedule wear-leveling"); | |
980 | } else | |
981 | dbg_wl("schedule scrubbing"); | |
982 | ||
983 | ubi->wl_scheduled = 1; | |
984 | spin_unlock(&ubi->wl_lock); | |
985 | ||
33818bbb | 986 | wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS); |
801c135c AB |
987 | if (!wrk) { |
988 | err = -ENOMEM; | |
989 | goto out_cancel; | |
990 | } | |
991 | ||
992 | wrk->func = &wear_leveling_worker; | |
993 | schedule_ubi_work(ubi, wrk); | |
994 | return err; | |
995 | ||
996 | out_cancel: | |
997 | spin_lock(&ubi->wl_lock); | |
998 | ubi->wl_scheduled = 0; | |
999 | out_unlock: | |
1000 | spin_unlock(&ubi->wl_lock); | |
1001 | return err; | |
1002 | } | |
1003 | ||
1004 | /** | |
1005 | * erase_worker - physical eraseblock erase worker function. | |
1006 | * @ubi: UBI device description object | |
1007 | * @wl_wrk: the work object | |
1008 | * @cancel: non-zero if the worker has to free memory and exit | |
1009 | * | |
1010 | * This function erases a physical eraseblock and perform torture testing if | |
1011 | * needed. It also takes care about marking the physical eraseblock bad if | |
1012 | * needed. Returns zero in case of success and a negative error code in case of | |
1013 | * failure. | |
1014 | */ | |
1015 | static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk, | |
1016 | int cancel) | |
1017 | { | |
801c135c | 1018 | struct ubi_wl_entry *e = wl_wrk->e; |
784c1454 | 1019 | int pnum = e->pnum, err, need; |
801c135c AB |
1020 | |
1021 | if (cancel) { | |
1022 | dbg_wl("cancel erasure of PEB %d EC %d", pnum, e->ec); | |
1023 | kfree(wl_wrk); | |
06b68ba1 | 1024 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1025 | return 0; |
1026 | } | |
1027 | ||
1028 | dbg_wl("erase PEB %d EC %d", pnum, e->ec); | |
1029 | ||
1030 | err = sync_erase(ubi, e, wl_wrk->torture); | |
1031 | if (!err) { | |
1032 | /* Fine, we've erased it successfully */ | |
1033 | kfree(wl_wrk); | |
1034 | ||
1035 | spin_lock(&ubi->wl_lock); | |
5abde384 | 1036 | wl_tree_add(e, &ubi->free); |
801c135c AB |
1037 | spin_unlock(&ubi->wl_lock); |
1038 | ||
1039 | /* | |
9c9ec147 AB |
1040 | * One more erase operation has happened, take care about |
1041 | * protected physical eraseblocks. | |
801c135c | 1042 | */ |
7b6c32da | 1043 | serve_prot_queue(ubi); |
801c135c AB |
1044 | |
1045 | /* And take care about wear-leveling */ | |
1046 | err = ensure_wear_leveling(ubi); | |
1047 | return err; | |
1048 | } | |
1049 | ||
8d2d4011 | 1050 | ubi_err("failed to erase PEB %d, error %d", pnum, err); |
801c135c | 1051 | kfree(wl_wrk); |
06b68ba1 | 1052 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c | 1053 | |
784c1454 AB |
1054 | if (err == -EINTR || err == -ENOMEM || err == -EAGAIN || |
1055 | err == -EBUSY) { | |
1056 | int err1; | |
1057 | ||
1058 | /* Re-schedule the LEB for erasure */ | |
1059 | err1 = schedule_erase(ubi, e, 0); | |
1060 | if (err1) { | |
1061 | err = err1; | |
1062 | goto out_ro; | |
1063 | } | |
1064 | return err; | |
1065 | } else if (err != -EIO) { | |
801c135c AB |
1066 | /* |
1067 | * If this is not %-EIO, we have no idea what to do. Scheduling | |
1068 | * this physical eraseblock for erasure again would cause | |
815bc5f8 | 1069 | * errors again and again. Well, lets switch to R/O mode. |
801c135c | 1070 | */ |
784c1454 | 1071 | goto out_ro; |
801c135c AB |
1072 | } |
1073 | ||
1074 | /* It is %-EIO, the PEB went bad */ | |
1075 | ||
1076 | if (!ubi->bad_allowed) { | |
1077 | ubi_err("bad physical eraseblock %d detected", pnum); | |
784c1454 AB |
1078 | goto out_ro; |
1079 | } | |
801c135c | 1080 | |
784c1454 AB |
1081 | spin_lock(&ubi->volumes_lock); |
1082 | need = ubi->beb_rsvd_level - ubi->beb_rsvd_pebs + 1; | |
1083 | if (need > 0) { | |
1084 | need = ubi->avail_pebs >= need ? need : ubi->avail_pebs; | |
1085 | ubi->avail_pebs -= need; | |
1086 | ubi->rsvd_pebs += need; | |
1087 | ubi->beb_rsvd_pebs += need; | |
1088 | if (need > 0) | |
1089 | ubi_msg("reserve more %d PEBs", need); | |
1090 | } | |
801c135c | 1091 | |
784c1454 | 1092 | if (ubi->beb_rsvd_pebs == 0) { |
801c135c | 1093 | spin_unlock(&ubi->volumes_lock); |
784c1454 AB |
1094 | ubi_err("no reserved physical eraseblocks"); |
1095 | goto out_ro; | |
1096 | } | |
784c1454 | 1097 | spin_unlock(&ubi->volumes_lock); |
801c135c | 1098 | |
52b605d1 | 1099 | ubi_msg("mark PEB %d as bad", pnum); |
784c1454 AB |
1100 | err = ubi_io_mark_bad(ubi, pnum); |
1101 | if (err) | |
1102 | goto out_ro; | |
1103 | ||
1104 | spin_lock(&ubi->volumes_lock); | |
1105 | ubi->beb_rsvd_pebs -= 1; | |
1106 | ubi->bad_peb_count += 1; | |
1107 | ubi->good_peb_count -= 1; | |
1108 | ubi_calculate_reserved(ubi); | |
52b605d1 AB |
1109 | if (ubi->beb_rsvd_pebs) |
1110 | ubi_msg("%d PEBs left in the reserve", ubi->beb_rsvd_pebs); | |
1111 | else | |
784c1454 AB |
1112 | ubi_warn("last PEB from the reserved pool was used"); |
1113 | spin_unlock(&ubi->volumes_lock); | |
1114 | ||
1115 | return err; | |
801c135c | 1116 | |
784c1454 AB |
1117 | out_ro: |
1118 | ubi_ro_mode(ubi); | |
801c135c AB |
1119 | return err; |
1120 | } | |
1121 | ||
1122 | /** | |
85c6e6e2 | 1123 | * ubi_wl_put_peb - return a PEB to the wear-leveling sub-system. |
801c135c AB |
1124 | * @ubi: UBI device description object |
1125 | * @pnum: physical eraseblock to return | |
1126 | * @torture: if this physical eraseblock has to be tortured | |
1127 | * | |
1128 | * This function is called to return physical eraseblock @pnum to the pool of | |
1129 | * free physical eraseblocks. The @torture flag has to be set if an I/O error | |
1130 | * occurred to this @pnum and it has to be tested. This function returns zero | |
43f9b25a | 1131 | * in case of success, and a negative error code in case of failure. |
801c135c AB |
1132 | */ |
1133 | int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture) | |
1134 | { | |
1135 | int err; | |
1136 | struct ubi_wl_entry *e; | |
1137 | ||
1138 | dbg_wl("PEB %d", pnum); | |
1139 | ubi_assert(pnum >= 0); | |
1140 | ubi_assert(pnum < ubi->peb_count); | |
1141 | ||
43f9b25a | 1142 | retry: |
801c135c | 1143 | spin_lock(&ubi->wl_lock); |
801c135c AB |
1144 | e = ubi->lookuptbl[pnum]; |
1145 | if (e == ubi->move_from) { | |
1146 | /* | |
1147 | * User is putting the physical eraseblock which was selected to | |
1148 | * be moved. It will be scheduled for erasure in the | |
1149 | * wear-leveling worker. | |
1150 | */ | |
43f9b25a | 1151 | dbg_wl("PEB %d is being moved, wait", pnum); |
801c135c | 1152 | spin_unlock(&ubi->wl_lock); |
43f9b25a AB |
1153 | |
1154 | /* Wait for the WL worker by taking the @ubi->move_mutex */ | |
1155 | mutex_lock(&ubi->move_mutex); | |
1156 | mutex_unlock(&ubi->move_mutex); | |
1157 | goto retry; | |
801c135c AB |
1158 | } else if (e == ubi->move_to) { |
1159 | /* | |
1160 | * User is putting the physical eraseblock which was selected | |
1161 | * as the target the data is moved to. It may happen if the EBA | |
85c6e6e2 AB |
1162 | * sub-system already re-mapped the LEB in 'ubi_eba_copy_leb()' |
1163 | * but the WL sub-system has not put the PEB to the "used" tree | |
1164 | * yet, but it is about to do this. So we just set a flag which | |
1165 | * will tell the WL worker that the PEB is not needed anymore | |
1166 | * and should be scheduled for erasure. | |
801c135c AB |
1167 | */ |
1168 | dbg_wl("PEB %d is the target of data moving", pnum); | |
1169 | ubi_assert(!ubi->move_to_put); | |
1170 | ubi->move_to_put = 1; | |
1171 | spin_unlock(&ubi->wl_lock); | |
1172 | return 0; | |
1173 | } else { | |
5abde384 | 1174 | if (in_wl_tree(e, &ubi->used)) { |
d99383b0 | 1175 | paranoid_check_in_wl_tree(ubi, e, &ubi->used); |
23553b2c | 1176 | rb_erase(&e->u.rb, &ubi->used); |
5abde384 | 1177 | } else if (in_wl_tree(e, &ubi->scrub)) { |
d99383b0 | 1178 | paranoid_check_in_wl_tree(ubi, e, &ubi->scrub); |
23553b2c | 1179 | rb_erase(&e->u.rb, &ubi->scrub); |
b86a2c56 | 1180 | } else if (in_wl_tree(e, &ubi->erroneous)) { |
d99383b0 | 1181 | paranoid_check_in_wl_tree(ubi, e, &ubi->erroneous); |
b86a2c56 AB |
1182 | rb_erase(&e->u.rb, &ubi->erroneous); |
1183 | ubi->erroneous_peb_count -= 1; | |
1184 | ubi_assert(ubi->erroneous_peb_count >= 0); | |
815bc5f8 | 1185 | /* Erroneous PEBs should be tortured */ |
b86a2c56 | 1186 | torture = 1; |
43f9b25a | 1187 | } else { |
7b6c32da | 1188 | err = prot_queue_del(ubi, e->pnum); |
43f9b25a AB |
1189 | if (err) { |
1190 | ubi_err("PEB %d not found", pnum); | |
1191 | ubi_ro_mode(ubi); | |
1192 | spin_unlock(&ubi->wl_lock); | |
1193 | return err; | |
1194 | } | |
1195 | } | |
801c135c AB |
1196 | } |
1197 | spin_unlock(&ubi->wl_lock); | |
1198 | ||
1199 | err = schedule_erase(ubi, e, torture); | |
1200 | if (err) { | |
1201 | spin_lock(&ubi->wl_lock); | |
5abde384 | 1202 | wl_tree_add(e, &ubi->used); |
801c135c AB |
1203 | spin_unlock(&ubi->wl_lock); |
1204 | } | |
1205 | ||
1206 | return err; | |
1207 | } | |
1208 | ||
1209 | /** | |
1210 | * ubi_wl_scrub_peb - schedule a physical eraseblock for scrubbing. | |
1211 | * @ubi: UBI device description object | |
1212 | * @pnum: the physical eraseblock to schedule | |
1213 | * | |
1214 | * If a bit-flip in a physical eraseblock is detected, this physical eraseblock | |
1215 | * needs scrubbing. This function schedules a physical eraseblock for | |
1216 | * scrubbing which is done in background. This function returns zero in case of | |
1217 | * success and a negative error code in case of failure. | |
1218 | */ | |
1219 | int ubi_wl_scrub_peb(struct ubi_device *ubi, int pnum) | |
1220 | { | |
1221 | struct ubi_wl_entry *e; | |
1222 | ||
8c1e6ee1 | 1223 | dbg_msg("schedule PEB %d for scrubbing", pnum); |
801c135c AB |
1224 | |
1225 | retry: | |
1226 | spin_lock(&ubi->wl_lock); | |
1227 | e = ubi->lookuptbl[pnum]; | |
d3f6e6c6 AB |
1228 | if (e == ubi->move_from || in_wl_tree(e, &ubi->scrub) || |
1229 | in_wl_tree(e, &ubi->erroneous)) { | |
801c135c AB |
1230 | spin_unlock(&ubi->wl_lock); |
1231 | return 0; | |
1232 | } | |
1233 | ||
1234 | if (e == ubi->move_to) { | |
1235 | /* | |
1236 | * This physical eraseblock was used to move data to. The data | |
1237 | * was moved but the PEB was not yet inserted to the proper | |
1238 | * tree. We should just wait a little and let the WL worker | |
1239 | * proceed. | |
1240 | */ | |
1241 | spin_unlock(&ubi->wl_lock); | |
1242 | dbg_wl("the PEB %d is not in proper tree, retry", pnum); | |
1243 | yield(); | |
1244 | goto retry; | |
1245 | } | |
1246 | ||
5abde384 | 1247 | if (in_wl_tree(e, &ubi->used)) { |
d99383b0 | 1248 | paranoid_check_in_wl_tree(ubi, e, &ubi->used); |
23553b2c | 1249 | rb_erase(&e->u.rb, &ubi->used); |
43f9b25a AB |
1250 | } else { |
1251 | int err; | |
1252 | ||
7b6c32da | 1253 | err = prot_queue_del(ubi, e->pnum); |
43f9b25a AB |
1254 | if (err) { |
1255 | ubi_err("PEB %d not found", pnum); | |
1256 | ubi_ro_mode(ubi); | |
1257 | spin_unlock(&ubi->wl_lock); | |
1258 | return err; | |
1259 | } | |
1260 | } | |
801c135c | 1261 | |
5abde384 | 1262 | wl_tree_add(e, &ubi->scrub); |
801c135c AB |
1263 | spin_unlock(&ubi->wl_lock); |
1264 | ||
1265 | /* | |
1266 | * Technically scrubbing is the same as wear-leveling, so it is done | |
1267 | * by the WL worker. | |
1268 | */ | |
1269 | return ensure_wear_leveling(ubi); | |
1270 | } | |
1271 | ||
1272 | /** | |
1273 | * ubi_wl_flush - flush all pending works. | |
1274 | * @ubi: UBI device description object | |
1275 | * | |
1276 | * This function returns zero in case of success and a negative error code in | |
1277 | * case of failure. | |
1278 | */ | |
1279 | int ubi_wl_flush(struct ubi_device *ubi) | |
1280 | { | |
593dd33c | 1281 | int err; |
801c135c AB |
1282 | |
1283 | /* | |
7b6c32da | 1284 | * Erase while the pending works queue is not empty, but not more than |
801c135c AB |
1285 | * the number of currently pending works. |
1286 | */ | |
593dd33c AB |
1287 | dbg_wl("flush (%d pending works)", ubi->works_count); |
1288 | while (ubi->works_count) { | |
1289 | err = do_work(ubi); | |
1290 | if (err) | |
1291 | return err; | |
1292 | } | |
1293 | ||
1294 | /* | |
1295 | * Make sure all the works which have been done in parallel are | |
1296 | * finished. | |
1297 | */ | |
1298 | down_write(&ubi->work_sem); | |
1299 | up_write(&ubi->work_sem); | |
1300 | ||
1301 | /* | |
6fa6f5bb | 1302 | * And in case last was the WL worker and it canceled the LEB |
593dd33c AB |
1303 | * movement, flush again. |
1304 | */ | |
1305 | while (ubi->works_count) { | |
1306 | dbg_wl("flush more (%d pending works)", ubi->works_count); | |
801c135c AB |
1307 | err = do_work(ubi); |
1308 | if (err) | |
1309 | return err; | |
1310 | } | |
1311 | ||
1312 | return 0; | |
1313 | } | |
1314 | ||
1315 | /** | |
1316 | * tree_destroy - destroy an RB-tree. | |
1317 | * @root: the root of the tree to destroy | |
1318 | */ | |
1319 | static void tree_destroy(struct rb_root *root) | |
1320 | { | |
1321 | struct rb_node *rb; | |
1322 | struct ubi_wl_entry *e; | |
1323 | ||
1324 | rb = root->rb_node; | |
1325 | while (rb) { | |
1326 | if (rb->rb_left) | |
1327 | rb = rb->rb_left; | |
1328 | else if (rb->rb_right) | |
1329 | rb = rb->rb_right; | |
1330 | else { | |
23553b2c | 1331 | e = rb_entry(rb, struct ubi_wl_entry, u.rb); |
801c135c AB |
1332 | |
1333 | rb = rb_parent(rb); | |
1334 | if (rb) { | |
23553b2c | 1335 | if (rb->rb_left == &e->u.rb) |
801c135c AB |
1336 | rb->rb_left = NULL; |
1337 | else | |
1338 | rb->rb_right = NULL; | |
1339 | } | |
1340 | ||
06b68ba1 | 1341 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1342 | } |
1343 | } | |
1344 | } | |
1345 | ||
1346 | /** | |
1347 | * ubi_thread - UBI background thread. | |
1348 | * @u: the UBI device description object pointer | |
1349 | */ | |
cdfa788a | 1350 | int ubi_thread(void *u) |
801c135c AB |
1351 | { |
1352 | int failures = 0; | |
1353 | struct ubi_device *ubi = u; | |
1354 | ||
1355 | ubi_msg("background thread \"%s\" started, PID %d", | |
ba25f9dc | 1356 | ubi->bgt_name, task_pid_nr(current)); |
801c135c | 1357 | |
83144186 | 1358 | set_freezable(); |
801c135c AB |
1359 | for (;;) { |
1360 | int err; | |
1361 | ||
1362 | if (kthread_should_stop()) | |
cadb40cc | 1363 | break; |
801c135c AB |
1364 | |
1365 | if (try_to_freeze()) | |
1366 | continue; | |
1367 | ||
1368 | spin_lock(&ubi->wl_lock); | |
1369 | if (list_empty(&ubi->works) || ubi->ro_mode || | |
27a0f2a3 | 1370 | !ubi->thread_enabled || ubi_dbg_is_bgt_disabled(ubi)) { |
801c135c AB |
1371 | set_current_state(TASK_INTERRUPTIBLE); |
1372 | spin_unlock(&ubi->wl_lock); | |
1373 | schedule(); | |
1374 | continue; | |
1375 | } | |
1376 | spin_unlock(&ubi->wl_lock); | |
1377 | ||
1378 | err = do_work(ubi); | |
1379 | if (err) { | |
1380 | ubi_err("%s: work failed with error code %d", | |
1381 | ubi->bgt_name, err); | |
1382 | if (failures++ > WL_MAX_FAILURES) { | |
1383 | /* | |
1384 | * Too many failures, disable the thread and | |
1385 | * switch to read-only mode. | |
1386 | */ | |
1387 | ubi_msg("%s: %d consecutive failures", | |
1388 | ubi->bgt_name, WL_MAX_FAILURES); | |
1389 | ubi_ro_mode(ubi); | |
2ad49887 VG |
1390 | ubi->thread_enabled = 0; |
1391 | continue; | |
801c135c AB |
1392 | } |
1393 | } else | |
1394 | failures = 0; | |
1395 | ||
1396 | cond_resched(); | |
1397 | } | |
1398 | ||
801c135c AB |
1399 | dbg_wl("background thread \"%s\" is killed", ubi->bgt_name); |
1400 | return 0; | |
1401 | } | |
1402 | ||
1403 | /** | |
1404 | * cancel_pending - cancel all pending works. | |
1405 | * @ubi: UBI device description object | |
1406 | */ | |
1407 | static void cancel_pending(struct ubi_device *ubi) | |
1408 | { | |
1409 | while (!list_empty(&ubi->works)) { | |
1410 | struct ubi_work *wrk; | |
1411 | ||
1412 | wrk = list_entry(ubi->works.next, struct ubi_work, list); | |
1413 | list_del(&wrk->list); | |
1414 | wrk->func(ubi, wrk, 1); | |
1415 | ubi->works_count -= 1; | |
1416 | ubi_assert(ubi->works_count >= 0); | |
1417 | } | |
1418 | } | |
1419 | ||
1420 | /** | |
85c6e6e2 | 1421 | * ubi_wl_init_scan - initialize the WL sub-system using scanning information. |
801c135c AB |
1422 | * @ubi: UBI device description object |
1423 | * @si: scanning information | |
1424 | * | |
1425 | * This function returns zero in case of success, and a negative error code in | |
1426 | * case of failure. | |
1427 | */ | |
1428 | int ubi_wl_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si) | |
1429 | { | |
7b6c32da | 1430 | int err, i; |
801c135c AB |
1431 | struct rb_node *rb1, *rb2; |
1432 | struct ubi_scan_volume *sv; | |
1433 | struct ubi_scan_leb *seb, *tmp; | |
1434 | struct ubi_wl_entry *e; | |
1435 | ||
b86a2c56 | 1436 | ubi->used = ubi->erroneous = ubi->free = ubi->scrub = RB_ROOT; |
801c135c | 1437 | spin_lock_init(&ubi->wl_lock); |
43f9b25a | 1438 | mutex_init(&ubi->move_mutex); |
593dd33c | 1439 | init_rwsem(&ubi->work_sem); |
801c135c AB |
1440 | ubi->max_ec = si->max_ec; |
1441 | INIT_LIST_HEAD(&ubi->works); | |
1442 | ||
1443 | sprintf(ubi->bgt_name, UBI_BGT_NAME_PATTERN, ubi->ubi_num); | |
1444 | ||
801c135c AB |
1445 | err = -ENOMEM; |
1446 | ubi->lookuptbl = kzalloc(ubi->peb_count * sizeof(void *), GFP_KERNEL); | |
1447 | if (!ubi->lookuptbl) | |
cdfa788a | 1448 | return err; |
801c135c | 1449 | |
7b6c32da XX |
1450 | for (i = 0; i < UBI_PROT_QUEUE_LEN; i++) |
1451 | INIT_LIST_HEAD(&ubi->pq[i]); | |
1452 | ubi->pq_head = 0; | |
1453 | ||
801c135c AB |
1454 | list_for_each_entry_safe(seb, tmp, &si->erase, u.list) { |
1455 | cond_resched(); | |
1456 | ||
06b68ba1 | 1457 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1458 | if (!e) |
1459 | goto out_free; | |
1460 | ||
1461 | e->pnum = seb->pnum; | |
1462 | e->ec = seb->ec; | |
1463 | ubi->lookuptbl[e->pnum] = e; | |
1464 | if (schedule_erase(ubi, e, 0)) { | |
06b68ba1 | 1465 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1466 | goto out_free; |
1467 | } | |
1468 | } | |
1469 | ||
1470 | list_for_each_entry(seb, &si->free, u.list) { | |
1471 | cond_resched(); | |
1472 | ||
06b68ba1 | 1473 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1474 | if (!e) |
1475 | goto out_free; | |
1476 | ||
1477 | e->pnum = seb->pnum; | |
1478 | e->ec = seb->ec; | |
1479 | ubi_assert(e->ec >= 0); | |
5abde384 | 1480 | wl_tree_add(e, &ubi->free); |
801c135c AB |
1481 | ubi->lookuptbl[e->pnum] = e; |
1482 | } | |
1483 | ||
801c135c AB |
1484 | ubi_rb_for_each_entry(rb1, sv, &si->volumes, rb) { |
1485 | ubi_rb_for_each_entry(rb2, seb, &sv->root, u.rb) { | |
1486 | cond_resched(); | |
1487 | ||
06b68ba1 | 1488 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1489 | if (!e) |
1490 | goto out_free; | |
1491 | ||
1492 | e->pnum = seb->pnum; | |
1493 | e->ec = seb->ec; | |
1494 | ubi->lookuptbl[e->pnum] = e; | |
1495 | if (!seb->scrub) { | |
1496 | dbg_wl("add PEB %d EC %d to the used tree", | |
1497 | e->pnum, e->ec); | |
5abde384 | 1498 | wl_tree_add(e, &ubi->used); |
801c135c AB |
1499 | } else { |
1500 | dbg_wl("add PEB %d EC %d to the scrub tree", | |
1501 | e->pnum, e->ec); | |
5abde384 | 1502 | wl_tree_add(e, &ubi->scrub); |
801c135c AB |
1503 | } |
1504 | } | |
1505 | } | |
1506 | ||
5abde384 | 1507 | if (ubi->avail_pebs < WL_RESERVED_PEBS) { |
801c135c AB |
1508 | ubi_err("no enough physical eraseblocks (%d, need %d)", |
1509 | ubi->avail_pebs, WL_RESERVED_PEBS); | |
5fc01ab6 AB |
1510 | if (ubi->corr_peb_count) |
1511 | ubi_err("%d PEBs are corrupted and not used", | |
1512 | ubi->corr_peb_count); | |
801c135c AB |
1513 | goto out_free; |
1514 | } | |
1515 | ubi->avail_pebs -= WL_RESERVED_PEBS; | |
1516 | ubi->rsvd_pebs += WL_RESERVED_PEBS; | |
1517 | ||
1518 | /* Schedule wear-leveling if needed */ | |
1519 | err = ensure_wear_leveling(ubi); | |
1520 | if (err) | |
1521 | goto out_free; | |
1522 | ||
1523 | return 0; | |
1524 | ||
1525 | out_free: | |
1526 | cancel_pending(ubi); | |
1527 | tree_destroy(&ubi->used); | |
1528 | tree_destroy(&ubi->free); | |
1529 | tree_destroy(&ubi->scrub); | |
1530 | kfree(ubi->lookuptbl); | |
801c135c AB |
1531 | return err; |
1532 | } | |
1533 | ||
1534 | /** | |
7b6c32da | 1535 | * protection_queue_destroy - destroy the protection queue. |
801c135c AB |
1536 | * @ubi: UBI device description object |
1537 | */ | |
7b6c32da | 1538 | static void protection_queue_destroy(struct ubi_device *ubi) |
801c135c | 1539 | { |
7b6c32da XX |
1540 | int i; |
1541 | struct ubi_wl_entry *e, *tmp; | |
801c135c | 1542 | |
7b6c32da XX |
1543 | for (i = 0; i < UBI_PROT_QUEUE_LEN; ++i) { |
1544 | list_for_each_entry_safe(e, tmp, &ubi->pq[i], u.list) { | |
1545 | list_del(&e->u.list); | |
1546 | kmem_cache_free(ubi_wl_entry_slab, e); | |
801c135c AB |
1547 | } |
1548 | } | |
1549 | } | |
1550 | ||
1551 | /** | |
85c6e6e2 | 1552 | * ubi_wl_close - close the wear-leveling sub-system. |
801c135c AB |
1553 | * @ubi: UBI device description object |
1554 | */ | |
1555 | void ubi_wl_close(struct ubi_device *ubi) | |
1556 | { | |
85c6e6e2 | 1557 | dbg_wl("close the WL sub-system"); |
801c135c | 1558 | cancel_pending(ubi); |
7b6c32da | 1559 | protection_queue_destroy(ubi); |
801c135c | 1560 | tree_destroy(&ubi->used); |
b86a2c56 | 1561 | tree_destroy(&ubi->erroneous); |
801c135c AB |
1562 | tree_destroy(&ubi->free); |
1563 | tree_destroy(&ubi->scrub); | |
1564 | kfree(ubi->lookuptbl); | |
801c135c AB |
1565 | } |
1566 | ||
92d124f5 | 1567 | #ifdef CONFIG_MTD_UBI_DEBUG |
801c135c AB |
1568 | |
1569 | /** | |
ebaaf1af | 1570 | * paranoid_check_ec - make sure that the erase counter of a PEB is correct. |
801c135c AB |
1571 | * @ubi: UBI device description object |
1572 | * @pnum: the physical eraseblock number to check | |
1573 | * @ec: the erase counter to check | |
1574 | * | |
1575 | * This function returns zero if the erase counter of physical eraseblock @pnum | |
feddbb34 AB |
1576 | * is equivalent to @ec, and a negative error code if not or if an error |
1577 | * occurred. | |
801c135c | 1578 | */ |
e88d6e10 | 1579 | static int paranoid_check_ec(struct ubi_device *ubi, int pnum, int ec) |
801c135c AB |
1580 | { |
1581 | int err; | |
1582 | long long read_ec; | |
1583 | struct ubi_ec_hdr *ec_hdr; | |
1584 | ||
2a734bb8 | 1585 | if (!ubi->dbg->chk_gen) |
92d124f5 AB |
1586 | return 0; |
1587 | ||
33818bbb | 1588 | ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_NOFS); |
801c135c AB |
1589 | if (!ec_hdr) |
1590 | return -ENOMEM; | |
1591 | ||
1592 | err = ubi_io_read_ec_hdr(ubi, pnum, ec_hdr, 0); | |
1593 | if (err && err != UBI_IO_BITFLIPS) { | |
1594 | /* The header does not have to exist */ | |
1595 | err = 0; | |
1596 | goto out_free; | |
1597 | } | |
1598 | ||
3261ebd7 | 1599 | read_ec = be64_to_cpu(ec_hdr->ec); |
801c135c AB |
1600 | if (ec != read_ec) { |
1601 | ubi_err("paranoid check failed for PEB %d", pnum); | |
1602 | ubi_err("read EC is %lld, should be %d", read_ec, ec); | |
1603 | ubi_dbg_dump_stack(); | |
1604 | err = 1; | |
1605 | } else | |
1606 | err = 0; | |
1607 | ||
1608 | out_free: | |
1609 | kfree(ec_hdr); | |
1610 | return err; | |
1611 | } | |
1612 | ||
1613 | /** | |
ebaaf1af | 1614 | * paranoid_check_in_wl_tree - check that wear-leveling entry is in WL RB-tree. |
d99383b0 | 1615 | * @ubi: UBI device description object |
801c135c AB |
1616 | * @e: the wear-leveling entry to check |
1617 | * @root: the root of the tree | |
1618 | * | |
adbf05e3 AB |
1619 | * This function returns zero if @e is in the @root RB-tree and %-EINVAL if it |
1620 | * is not. | |
801c135c | 1621 | */ |
d99383b0 AB |
1622 | static int paranoid_check_in_wl_tree(const struct ubi_device *ubi, |
1623 | struct ubi_wl_entry *e, | |
801c135c AB |
1624 | struct rb_root *root) |
1625 | { | |
2a734bb8 | 1626 | if (!ubi->dbg->chk_gen) |
92d124f5 AB |
1627 | return 0; |
1628 | ||
801c135c AB |
1629 | if (in_wl_tree(e, root)) |
1630 | return 0; | |
1631 | ||
1632 | ubi_err("paranoid check failed for PEB %d, EC %d, RB-tree %p ", | |
1633 | e->pnum, e->ec, root); | |
1634 | ubi_dbg_dump_stack(); | |
adbf05e3 | 1635 | return -EINVAL; |
801c135c AB |
1636 | } |
1637 | ||
7b6c32da XX |
1638 | /** |
1639 | * paranoid_check_in_pq - check if wear-leveling entry is in the protection | |
1640 | * queue. | |
1641 | * @ubi: UBI device description object | |
1642 | * @e: the wear-leveling entry to check | |
1643 | * | |
adbf05e3 | 1644 | * This function returns zero if @e is in @ubi->pq and %-EINVAL if it is not. |
7b6c32da | 1645 | */ |
d99383b0 AB |
1646 | static int paranoid_check_in_pq(const struct ubi_device *ubi, |
1647 | struct ubi_wl_entry *e) | |
7b6c32da XX |
1648 | { |
1649 | struct ubi_wl_entry *p; | |
1650 | int i; | |
1651 | ||
2a734bb8 | 1652 | if (!ubi->dbg->chk_gen) |
92d124f5 AB |
1653 | return 0; |
1654 | ||
7b6c32da XX |
1655 | for (i = 0; i < UBI_PROT_QUEUE_LEN; ++i) |
1656 | list_for_each_entry(p, &ubi->pq[i], u.list) | |
1657 | if (p == e) | |
1658 | return 0; | |
1659 | ||
1660 | ubi_err("paranoid check failed for PEB %d, EC %d, Protect queue", | |
1661 | e->pnum, e->ec); | |
1662 | ubi_dbg_dump_stack(); | |
adbf05e3 | 1663 | return -EINVAL; |
7b6c32da | 1664 | } |
92d124f5 AB |
1665 | |
1666 | #endif /* CONFIG_MTD_UBI_DEBUG */ |