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 | * | |
85c6e6e2 AB |
44 | * If the WL sub-system fails to erase a physical eraseblock, it marks it as |
45 | * bad. | |
801c135c | 46 | * |
85c6e6e2 AB |
47 | * This sub-system is also responsible for scrubbing. If a bit-flip is detected |
48 | * in a physical eraseblock, it has to be moved. Technically this is the same | |
49 | * as moving it for wear-leveling reasons. | |
801c135c | 50 | * |
85c6e6e2 AB |
51 | * As it was said, for the UBI sub-system all physical eraseblocks are either |
52 | * "free" or "used". Free eraseblock are kept in the @wl->free RB-tree, while | |
b86a2c56 AB |
53 | * used eraseblocks are kept in @wl->used, @wl->erroneous, or @wl->scrub |
54 | * RB-trees, as well as (temporarily) in the @wl->pq queue. | |
7b6c32da XX |
55 | * |
56 | * When the WL sub-system returns a physical eraseblock, the physical | |
57 | * eraseblock is protected from being moved for some "time". For this reason, | |
58 | * the physical eraseblock is not directly moved from the @wl->free tree to the | |
59 | * @wl->used tree. There is a protection queue in between where this | |
60 | * physical eraseblock is temporarily stored (@wl->pq). | |
61 | * | |
62 | * All this protection stuff is needed because: | |
63 | * o we don't want to move physical eraseblocks just after we have given them | |
64 | * to the user; instead, we first want to let users fill them up with data; | |
65 | * | |
66 | * o there is a chance that the user will put the physical eraseblock very | |
44156267 | 67 | * soon, so it makes sense not to move it for some time, but wait. |
7b6c32da XX |
68 | * |
69 | * Physical eraseblocks stay protected only for limited time. But the "time" is | |
70 | * measured in erase cycles in this case. This is implemented with help of the | |
71 | * protection queue. Eraseblocks are put to the tail of this queue when they | |
72 | * are returned by the 'ubi_wl_get_peb()', and eraseblocks are removed from the | |
73 | * head of the queue on each erase operation (for any eraseblock). So the | |
74 | * length of the queue defines how may (global) erase cycles PEBs are protected. | |
75 | * | |
76 | * To put it differently, each physical eraseblock has 2 main states: free and | |
77 | * used. The former state corresponds to the @wl->free tree. The latter state | |
78 | * is split up on several sub-states: | |
79 | * o the WL movement is allowed (@wl->used tree); | |
815bc5f8 | 80 | * o the WL movement is disallowed (@wl->erroneous) because the PEB is |
b86a2c56 | 81 | * erroneous - e.g., there was a read error; |
7b6c32da XX |
82 | * o the WL movement is temporarily prohibited (@wl->pq queue); |
83 | * o scrubbing is needed (@wl->scrub tree). | |
84 | * | |
85 | * Depending on the sub-state, wear-leveling entries of the used physical | |
86 | * eraseblocks may be kept in one of those structures. | |
801c135c AB |
87 | * |
88 | * Note, in this implementation, we keep a small in-RAM object for each physical | |
89 | * eraseblock. This is surely not a scalable solution. But it appears to be good | |
90 | * enough for moderately large flashes and it is simple. In future, one may | |
85c6e6e2 | 91 | * re-work this sub-system and make it more scalable. |
801c135c | 92 | * |
85c6e6e2 AB |
93 | * At the moment this sub-system does not utilize the sequence number, which |
94 | * was introduced relatively recently. But it would be wise to do this because | |
95 | * the sequence number of a logical eraseblock characterizes how old is it. For | |
801c135c AB |
96 | * example, when we move a PEB with low erase counter, and we need to pick the |
97 | * target PEB, we pick a PEB with the highest EC if our PEB is "old" and we | |
98 | * pick target PEB with an average EC if our PEB is not very "old". This is a | |
85c6e6e2 | 99 | * room for future re-works of the WL sub-system. |
801c135c AB |
100 | */ |
101 | ||
102 | #include <linux/slab.h> | |
103 | #include <linux/crc32.h> | |
104 | #include <linux/freezer.h> | |
105 | #include <linux/kthread.h> | |
106 | #include "ubi.h" | |
107 | ||
108 | /* Number of physical eraseblocks reserved for wear-leveling purposes */ | |
109 | #define WL_RESERVED_PEBS 1 | |
110 | ||
801c135c AB |
111 | /* |
112 | * Maximum difference between two erase counters. If this threshold is | |
85c6e6e2 AB |
113 | * exceeded, the WL sub-system starts moving data from used physical |
114 | * eraseblocks with low erase counter to free physical eraseblocks with high | |
115 | * erase counter. | |
801c135c AB |
116 | */ |
117 | #define UBI_WL_THRESHOLD CONFIG_MTD_UBI_WL_THRESHOLD | |
118 | ||
119 | /* | |
85c6e6e2 | 120 | * When a physical eraseblock is moved, the WL sub-system has to pick the target |
801c135c AB |
121 | * physical eraseblock to move to. The simplest way would be just to pick the |
122 | * one with the highest erase counter. But in certain workloads this could lead | |
123 | * to an unlimited wear of one or few physical eraseblock. Indeed, imagine a | |
124 | * situation when the picked physical eraseblock is constantly erased after the | |
125 | * data is written to it. So, we have a constant which limits the highest erase | |
85c6e6e2 | 126 | * counter of the free physical eraseblock to pick. Namely, the WL sub-system |
025dfdaf | 127 | * does not pick eraseblocks with erase counter greater than the lowest erase |
801c135c AB |
128 | * counter plus %WL_FREE_MAX_DIFF. |
129 | */ | |
130 | #define WL_FREE_MAX_DIFF (2*UBI_WL_THRESHOLD) | |
131 | ||
132 | /* | |
133 | * Maximum number of consecutive background thread failures which is enough to | |
134 | * switch to read-only mode. | |
135 | */ | |
136 | #define WL_MAX_FAILURES 32 | |
137 | ||
801c135c AB |
138 | /** |
139 | * struct ubi_work - UBI work description data structure. | |
140 | * @list: a link in the list of pending works | |
141 | * @func: worker function | |
801c135c AB |
142 | * @e: physical eraseblock to erase |
143 | * @torture: if the physical eraseblock has to be tortured | |
144 | * | |
145 | * The @func pointer points to the worker function. If the @cancel argument is | |
146 | * not zero, the worker has to free the resources and exit immediately. The | |
147 | * worker has to return zero in case of success and a negative error code in | |
148 | * case of failure. | |
149 | */ | |
150 | struct ubi_work { | |
151 | struct list_head list; | |
152 | int (*func)(struct ubi_device *ubi, struct ubi_work *wrk, int cancel); | |
153 | /* The below fields are only relevant to erasure works */ | |
154 | struct ubi_wl_entry *e; | |
155 | int torture; | |
156 | }; | |
157 | ||
7bf523ae AB |
158 | static int self_check_ec(struct ubi_device *ubi, int pnum, int ec); |
159 | static int self_check_in_wl_tree(const struct ubi_device *ubi, | |
160 | struct ubi_wl_entry *e, struct rb_root *root); | |
161 | static int self_check_in_pq(const struct ubi_device *ubi, | |
162 | struct ubi_wl_entry *e); | |
801c135c | 163 | |
801c135c AB |
164 | /** |
165 | * wl_tree_add - add a wear-leveling entry to a WL RB-tree. | |
166 | * @e: the wear-leveling entry to add | |
167 | * @root: the root of the tree | |
168 | * | |
169 | * Note, we use (erase counter, physical eraseblock number) pairs as keys in | |
170 | * the @ubi->used and @ubi->free RB-trees. | |
171 | */ | |
172 | static void wl_tree_add(struct ubi_wl_entry *e, struct rb_root *root) | |
173 | { | |
174 | struct rb_node **p, *parent = NULL; | |
175 | ||
176 | p = &root->rb_node; | |
177 | while (*p) { | |
178 | struct ubi_wl_entry *e1; | |
179 | ||
180 | parent = *p; | |
23553b2c | 181 | e1 = rb_entry(parent, struct ubi_wl_entry, u.rb); |
801c135c AB |
182 | |
183 | if (e->ec < e1->ec) | |
184 | p = &(*p)->rb_left; | |
185 | else if (e->ec > e1->ec) | |
186 | p = &(*p)->rb_right; | |
187 | else { | |
188 | ubi_assert(e->pnum != e1->pnum); | |
189 | if (e->pnum < e1->pnum) | |
190 | p = &(*p)->rb_left; | |
191 | else | |
192 | p = &(*p)->rb_right; | |
193 | } | |
194 | } | |
195 | ||
23553b2c XX |
196 | rb_link_node(&e->u.rb, parent, p); |
197 | rb_insert_color(&e->u.rb, root); | |
801c135c AB |
198 | } |
199 | ||
801c135c AB |
200 | /** |
201 | * do_work - do one pending work. | |
202 | * @ubi: UBI device description object | |
203 | * | |
204 | * This function returns zero in case of success and a negative error code in | |
205 | * case of failure. | |
206 | */ | |
207 | static int do_work(struct ubi_device *ubi) | |
208 | { | |
209 | int err; | |
210 | struct ubi_work *wrk; | |
211 | ||
43f9b25a AB |
212 | cond_resched(); |
213 | ||
593dd33c AB |
214 | /* |
215 | * @ubi->work_sem is used to synchronize with the workers. Workers take | |
216 | * it in read mode, so many of them may be doing works at a time. But | |
217 | * the queue flush code has to be sure the whole queue of works is | |
218 | * done, and it takes the mutex in write mode. | |
219 | */ | |
220 | down_read(&ubi->work_sem); | |
801c135c | 221 | spin_lock(&ubi->wl_lock); |
801c135c AB |
222 | if (list_empty(&ubi->works)) { |
223 | spin_unlock(&ubi->wl_lock); | |
593dd33c | 224 | up_read(&ubi->work_sem); |
801c135c AB |
225 | return 0; |
226 | } | |
227 | ||
228 | wrk = list_entry(ubi->works.next, struct ubi_work, list); | |
229 | list_del(&wrk->list); | |
16f557ec AB |
230 | ubi->works_count -= 1; |
231 | ubi_assert(ubi->works_count >= 0); | |
801c135c AB |
232 | spin_unlock(&ubi->wl_lock); |
233 | ||
234 | /* | |
235 | * Call the worker function. Do not touch the work structure | |
236 | * after this call as it will have been freed or reused by that | |
237 | * time by the worker function. | |
238 | */ | |
239 | err = wrk->func(ubi, wrk, 0); | |
240 | if (err) | |
241 | ubi_err("work failed with error code %d", err); | |
593dd33c | 242 | up_read(&ubi->work_sem); |
16f557ec | 243 | |
801c135c AB |
244 | return err; |
245 | } | |
246 | ||
247 | /** | |
248 | * produce_free_peb - produce a free physical eraseblock. | |
249 | * @ubi: UBI device description object | |
250 | * | |
251 | * This function tries to make a free PEB by means of synchronous execution of | |
252 | * pending works. This may be needed if, for example the background thread is | |
253 | * disabled. Returns zero in case of success and a negative error code in case | |
254 | * of failure. | |
255 | */ | |
256 | static int produce_free_peb(struct ubi_device *ubi) | |
257 | { | |
258 | int err; | |
259 | ||
260 | spin_lock(&ubi->wl_lock); | |
5abde384 | 261 | while (!ubi->free.rb_node) { |
801c135c AB |
262 | spin_unlock(&ubi->wl_lock); |
263 | ||
264 | dbg_wl("do one work synchronously"); | |
265 | err = do_work(ubi); | |
266 | if (err) | |
267 | return err; | |
268 | ||
269 | spin_lock(&ubi->wl_lock); | |
270 | } | |
271 | spin_unlock(&ubi->wl_lock); | |
272 | ||
273 | return 0; | |
274 | } | |
275 | ||
276 | /** | |
277 | * in_wl_tree - check if wear-leveling entry is present in a WL RB-tree. | |
278 | * @e: the wear-leveling entry to check | |
279 | * @root: the root of the tree | |
280 | * | |
281 | * This function returns non-zero if @e is in the @root RB-tree and zero if it | |
282 | * is not. | |
283 | */ | |
284 | static int in_wl_tree(struct ubi_wl_entry *e, struct rb_root *root) | |
285 | { | |
286 | struct rb_node *p; | |
287 | ||
288 | p = root->rb_node; | |
289 | while (p) { | |
290 | struct ubi_wl_entry *e1; | |
291 | ||
23553b2c | 292 | e1 = rb_entry(p, struct ubi_wl_entry, u.rb); |
801c135c AB |
293 | |
294 | if (e->pnum == e1->pnum) { | |
295 | ubi_assert(e == e1); | |
296 | return 1; | |
297 | } | |
298 | ||
299 | if (e->ec < e1->ec) | |
300 | p = p->rb_left; | |
301 | else if (e->ec > e1->ec) | |
302 | p = p->rb_right; | |
303 | else { | |
304 | ubi_assert(e->pnum != e1->pnum); | |
305 | if (e->pnum < e1->pnum) | |
306 | p = p->rb_left; | |
307 | else | |
308 | p = p->rb_right; | |
309 | } | |
310 | } | |
311 | ||
312 | return 0; | |
313 | } | |
314 | ||
315 | /** | |
7b6c32da | 316 | * prot_queue_add - add physical eraseblock to the protection queue. |
801c135c AB |
317 | * @ubi: UBI device description object |
318 | * @e: the physical eraseblock to add | |
801c135c | 319 | * |
7b6c32da XX |
320 | * This function adds @e to the tail of the protection queue @ubi->pq, where |
321 | * @e will stay for %UBI_PROT_QUEUE_LEN erase operations and will be | |
322 | * temporarily protected from the wear-leveling worker. Note, @wl->lock has to | |
323 | * be locked. | |
801c135c | 324 | */ |
7b6c32da | 325 | static void prot_queue_add(struct ubi_device *ubi, struct ubi_wl_entry *e) |
801c135c | 326 | { |
7b6c32da | 327 | int pq_tail = ubi->pq_head - 1; |
801c135c | 328 | |
7b6c32da XX |
329 | if (pq_tail < 0) |
330 | pq_tail = UBI_PROT_QUEUE_LEN - 1; | |
331 | ubi_assert(pq_tail >= 0 && pq_tail < UBI_PROT_QUEUE_LEN); | |
332 | list_add_tail(&e->u.list, &ubi->pq[pq_tail]); | |
333 | dbg_wl("added PEB %d EC %d to the protection queue", e->pnum, e->ec); | |
801c135c AB |
334 | } |
335 | ||
336 | /** | |
337 | * find_wl_entry - find wear-leveling entry closest to certain erase counter. | |
338 | * @root: the RB-tree where to look for | |
add8287e | 339 | * @diff: maximum possible difference from the smallest erase counter |
801c135c AB |
340 | * |
341 | * This function looks for a wear leveling entry with erase counter closest to | |
add8287e | 342 | * min + @diff, where min is the smallest erase counter. |
801c135c | 343 | */ |
add8287e | 344 | static struct ubi_wl_entry *find_wl_entry(struct rb_root *root, int diff) |
801c135c AB |
345 | { |
346 | struct rb_node *p; | |
347 | struct ubi_wl_entry *e; | |
add8287e | 348 | int max; |
801c135c | 349 | |
23553b2c | 350 | e = rb_entry(rb_first(root), struct ubi_wl_entry, u.rb); |
add8287e | 351 | max = e->ec + diff; |
801c135c AB |
352 | |
353 | p = root->rb_node; | |
354 | while (p) { | |
355 | struct ubi_wl_entry *e1; | |
356 | ||
23553b2c | 357 | e1 = rb_entry(p, struct ubi_wl_entry, u.rb); |
801c135c AB |
358 | if (e1->ec >= max) |
359 | p = p->rb_left; | |
360 | else { | |
361 | p = p->rb_right; | |
362 | e = e1; | |
363 | } | |
364 | } | |
365 | ||
366 | return e; | |
367 | } | |
368 | ||
369 | /** | |
370 | * ubi_wl_get_peb - get a physical eraseblock. | |
371 | * @ubi: UBI device description object | |
801c135c AB |
372 | * |
373 | * This function returns a physical eraseblock in case of success and a | |
374 | * negative error code in case of failure. Might sleep. | |
375 | */ | |
b36a261e | 376 | int ubi_wl_get_peb(struct ubi_device *ubi) |
801c135c | 377 | { |
7eb3aa65 | 378 | int err; |
801c135c | 379 | struct ubi_wl_entry *e, *first, *last; |
801c135c | 380 | |
801c135c AB |
381 | retry: |
382 | spin_lock(&ubi->wl_lock); | |
5abde384 | 383 | if (!ubi->free.rb_node) { |
801c135c AB |
384 | if (ubi->works_count == 0) { |
385 | ubi_assert(list_empty(&ubi->works)); | |
386 | ubi_err("no free eraseblocks"); | |
387 | spin_unlock(&ubi->wl_lock); | |
801c135c AB |
388 | return -ENOSPC; |
389 | } | |
390 | spin_unlock(&ubi->wl_lock); | |
391 | ||
392 | err = produce_free_peb(ubi); | |
7b6c32da | 393 | if (err < 0) |
801c135c | 394 | return err; |
801c135c AB |
395 | goto retry; |
396 | } | |
397 | ||
b36a261e RW |
398 | first = rb_entry(rb_first(&ubi->free), struct ubi_wl_entry, u.rb); |
399 | last = rb_entry(rb_last(&ubi->free), struct ubi_wl_entry, u.rb); | |
400 | ||
401 | if (last->ec - first->ec < WL_FREE_MAX_DIFF) | |
402 | e = rb_entry(ubi->free.rb_node, struct ubi_wl_entry, u.rb); | |
403 | else | |
404 | e = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF/2); | |
801c135c | 405 | |
7bf523ae | 406 | self_check_in_wl_tree(ubi, e, &ubi->free); |
7b6c32da | 407 | |
801c135c | 408 | /* |
7b6c32da | 409 | * Move the physical eraseblock to the protection queue where it will |
801c135c AB |
410 | * be protected from being moved for some time. |
411 | */ | |
23553b2c | 412 | rb_erase(&e->u.rb, &ubi->free); |
7b6c32da XX |
413 | dbg_wl("PEB %d EC %d", e->pnum, e->ec); |
414 | prot_queue_add(ubi, e); | |
801c135c | 415 | spin_unlock(&ubi->wl_lock); |
40a71a87 | 416 | |
97d6104b AB |
417 | err = ubi_self_check_all_ff(ubi, e->pnum, ubi->vid_hdr_aloffset, |
418 | ubi->peb_size - ubi->vid_hdr_aloffset); | |
40a71a87 | 419 | if (err) { |
1398788f | 420 | ubi_err("new PEB %d does not contain all 0xFF bytes", e->pnum); |
adbf05e3 | 421 | return err; |
40a71a87 AB |
422 | } |
423 | ||
801c135c AB |
424 | return e->pnum; |
425 | } | |
426 | ||
427 | /** | |
7b6c32da | 428 | * prot_queue_del - remove a physical eraseblock from the protection queue. |
801c135c AB |
429 | * @ubi: UBI device description object |
430 | * @pnum: the physical eraseblock to remove | |
43f9b25a | 431 | * |
7b6c32da XX |
432 | * This function deletes PEB @pnum from the protection queue and returns zero |
433 | * in case of success and %-ENODEV if the PEB was not found. | |
801c135c | 434 | */ |
7b6c32da | 435 | static int prot_queue_del(struct ubi_device *ubi, int pnum) |
801c135c | 436 | { |
7b6c32da | 437 | struct ubi_wl_entry *e; |
801c135c | 438 | |
7b6c32da XX |
439 | e = ubi->lookuptbl[pnum]; |
440 | if (!e) | |
441 | return -ENODEV; | |
801c135c | 442 | |
7bf523ae | 443 | if (self_check_in_pq(ubi, e)) |
7b6c32da | 444 | return -ENODEV; |
43f9b25a | 445 | |
7b6c32da XX |
446 | list_del(&e->u.list); |
447 | dbg_wl("deleted PEB %d from the protection queue", e->pnum); | |
43f9b25a | 448 | return 0; |
801c135c AB |
449 | } |
450 | ||
451 | /** | |
452 | * sync_erase - synchronously erase a physical eraseblock. | |
453 | * @ubi: UBI device description object | |
454 | * @e: the the physical eraseblock to erase | |
455 | * @torture: if the physical eraseblock has to be tortured | |
456 | * | |
457 | * This function returns zero in case of success and a negative error code in | |
458 | * case of failure. | |
459 | */ | |
9c9ec147 AB |
460 | static int sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e, |
461 | int torture) | |
801c135c AB |
462 | { |
463 | int err; | |
464 | struct ubi_ec_hdr *ec_hdr; | |
465 | unsigned long long ec = e->ec; | |
466 | ||
467 | dbg_wl("erase PEB %d, old EC %llu", e->pnum, ec); | |
468 | ||
7bf523ae | 469 | err = self_check_ec(ubi, e->pnum, e->ec); |
adbf05e3 | 470 | if (err) |
801c135c AB |
471 | return -EINVAL; |
472 | ||
33818bbb | 473 | ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_NOFS); |
801c135c AB |
474 | if (!ec_hdr) |
475 | return -ENOMEM; | |
476 | ||
477 | err = ubi_io_sync_erase(ubi, e->pnum, torture); | |
478 | if (err < 0) | |
479 | goto out_free; | |
480 | ||
481 | ec += err; | |
482 | if (ec > UBI_MAX_ERASECOUNTER) { | |
483 | /* | |
484 | * Erase counter overflow. Upgrade UBI and use 64-bit | |
485 | * erase counters internally. | |
486 | */ | |
487 | ubi_err("erase counter overflow at PEB %d, EC %llu", | |
488 | e->pnum, ec); | |
489 | err = -EINVAL; | |
490 | goto out_free; | |
491 | } | |
492 | ||
493 | dbg_wl("erased PEB %d, new EC %llu", e->pnum, ec); | |
494 | ||
3261ebd7 | 495 | ec_hdr->ec = cpu_to_be64(ec); |
801c135c AB |
496 | |
497 | err = ubi_io_write_ec_hdr(ubi, e->pnum, ec_hdr); | |
498 | if (err) | |
499 | goto out_free; | |
500 | ||
501 | e->ec = ec; | |
502 | spin_lock(&ubi->wl_lock); | |
503 | if (e->ec > ubi->max_ec) | |
504 | ubi->max_ec = e->ec; | |
505 | spin_unlock(&ubi->wl_lock); | |
506 | ||
507 | out_free: | |
508 | kfree(ec_hdr); | |
509 | return err; | |
510 | } | |
511 | ||
512 | /** | |
7b6c32da | 513 | * serve_prot_queue - check if it is time to stop protecting PEBs. |
801c135c AB |
514 | * @ubi: UBI device description object |
515 | * | |
7b6c32da XX |
516 | * This function is called after each erase operation and removes PEBs from the |
517 | * tail of the protection queue. These PEBs have been protected for long enough | |
518 | * and should be moved to the used tree. | |
801c135c | 519 | */ |
7b6c32da | 520 | static void serve_prot_queue(struct ubi_device *ubi) |
801c135c | 521 | { |
7b6c32da XX |
522 | struct ubi_wl_entry *e, *tmp; |
523 | int count; | |
801c135c AB |
524 | |
525 | /* | |
526 | * There may be several protected physical eraseblock to remove, | |
527 | * process them all. | |
528 | */ | |
7b6c32da XX |
529 | repeat: |
530 | count = 0; | |
531 | spin_lock(&ubi->wl_lock); | |
532 | list_for_each_entry_safe(e, tmp, &ubi->pq[ubi->pq_head], u.list) { | |
533 | dbg_wl("PEB %d EC %d protection over, move to used tree", | |
534 | e->pnum, e->ec); | |
801c135c | 535 | |
7b6c32da XX |
536 | list_del(&e->u.list); |
537 | wl_tree_add(e, &ubi->used); | |
538 | if (count++ > 32) { | |
539 | /* | |
540 | * Let's be nice and avoid holding the spinlock for | |
541 | * too long. | |
542 | */ | |
801c135c | 543 | spin_unlock(&ubi->wl_lock); |
7b6c32da XX |
544 | cond_resched(); |
545 | goto repeat; | |
801c135c | 546 | } |
801c135c | 547 | } |
7b6c32da XX |
548 | |
549 | ubi->pq_head += 1; | |
550 | if (ubi->pq_head == UBI_PROT_QUEUE_LEN) | |
551 | ubi->pq_head = 0; | |
552 | ubi_assert(ubi->pq_head >= 0 && ubi->pq_head < UBI_PROT_QUEUE_LEN); | |
553 | spin_unlock(&ubi->wl_lock); | |
801c135c AB |
554 | } |
555 | ||
556 | /** | |
557 | * schedule_ubi_work - schedule a work. | |
558 | * @ubi: UBI device description object | |
559 | * @wrk: the work to schedule | |
560 | * | |
7b6c32da XX |
561 | * This function adds a work defined by @wrk to the tail of the pending works |
562 | * list. | |
801c135c AB |
563 | */ |
564 | static void schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk) | |
565 | { | |
566 | spin_lock(&ubi->wl_lock); | |
567 | list_add_tail(&wrk->list, &ubi->works); | |
568 | ubi_assert(ubi->works_count >= 0); | |
569 | ubi->works_count += 1; | |
27a0f2a3 | 570 | if (ubi->thread_enabled && !ubi_dbg_is_bgt_disabled(ubi)) |
801c135c AB |
571 | wake_up_process(ubi->bgt_thread); |
572 | spin_unlock(&ubi->wl_lock); | |
573 | } | |
574 | ||
575 | static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk, | |
576 | int cancel); | |
577 | ||
578 | /** | |
579 | * schedule_erase - schedule an erase work. | |
580 | * @ubi: UBI device description object | |
581 | * @e: the WL entry of the physical eraseblock to erase | |
582 | * @torture: if the physical eraseblock has to be tortured | |
583 | * | |
584 | * This function returns zero in case of success and a %-ENOMEM in case of | |
585 | * failure. | |
586 | */ | |
587 | static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e, | |
588 | int torture) | |
589 | { | |
590 | struct ubi_work *wl_wrk; | |
591 | ||
592 | dbg_wl("schedule erasure of PEB %d, EC %d, torture %d", | |
593 | e->pnum, e->ec, torture); | |
594 | ||
33818bbb | 595 | wl_wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS); |
801c135c AB |
596 | if (!wl_wrk) |
597 | return -ENOMEM; | |
598 | ||
599 | wl_wrk->func = &erase_worker; | |
600 | wl_wrk->e = e; | |
601 | wl_wrk->torture = torture; | |
602 | ||
603 | schedule_ubi_work(ubi, wl_wrk); | |
604 | return 0; | |
605 | } | |
606 | ||
607 | /** | |
608 | * wear_leveling_worker - wear-leveling worker function. | |
609 | * @ubi: UBI device description object | |
610 | * @wrk: the work object | |
611 | * @cancel: non-zero if the worker has to free memory and exit | |
612 | * | |
613 | * This function copies a more worn out physical eraseblock to a less worn out | |
614 | * one. Returns zero in case of success and a negative error code in case of | |
615 | * failure. | |
616 | */ | |
617 | static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk, | |
618 | int cancel) | |
619 | { | |
b86a2c56 | 620 | int err, scrubbing = 0, torture = 0, protect = 0, erroneous = 0; |
9c259a52 | 621 | int vol_id = -1, uninitialized_var(lnum); |
801c135c AB |
622 | struct ubi_wl_entry *e1, *e2; |
623 | struct ubi_vid_hdr *vid_hdr; | |
624 | ||
625 | kfree(wrk); | |
801c135c AB |
626 | if (cancel) |
627 | return 0; | |
628 | ||
33818bbb | 629 | vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS); |
801c135c AB |
630 | if (!vid_hdr) |
631 | return -ENOMEM; | |
632 | ||
43f9b25a | 633 | mutex_lock(&ubi->move_mutex); |
801c135c | 634 | spin_lock(&ubi->wl_lock); |
43f9b25a AB |
635 | ubi_assert(!ubi->move_from && !ubi->move_to); |
636 | ubi_assert(!ubi->move_to_put); | |
801c135c | 637 | |
43f9b25a | 638 | if (!ubi->free.rb_node || |
5abde384 | 639 | (!ubi->used.rb_node && !ubi->scrub.rb_node)) { |
801c135c | 640 | /* |
43f9b25a AB |
641 | * No free physical eraseblocks? Well, they must be waiting in |
642 | * the queue to be erased. Cancel movement - it will be | |
643 | * triggered again when a free physical eraseblock appears. | |
801c135c AB |
644 | * |
645 | * No used physical eraseblocks? They must be temporarily | |
646 | * protected from being moved. They will be moved to the | |
647 | * @ubi->used tree later and the wear-leveling will be | |
648 | * triggered again. | |
649 | */ | |
650 | dbg_wl("cancel WL, a list is empty: free %d, used %d", | |
5abde384 | 651 | !ubi->free.rb_node, !ubi->used.rb_node); |
43f9b25a | 652 | goto out_cancel; |
801c135c AB |
653 | } |
654 | ||
5abde384 | 655 | if (!ubi->scrub.rb_node) { |
801c135c AB |
656 | /* |
657 | * Now pick the least worn-out used physical eraseblock and a | |
658 | * highly worn-out free physical eraseblock. If the erase | |
659 | * counters differ much enough, start wear-leveling. | |
660 | */ | |
23553b2c | 661 | e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, u.rb); |
801c135c AB |
662 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); |
663 | ||
664 | if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD)) { | |
665 | dbg_wl("no WL needed: min used EC %d, max free EC %d", | |
666 | e1->ec, e2->ec); | |
43f9b25a | 667 | goto out_cancel; |
801c135c | 668 | } |
7bf523ae | 669 | self_check_in_wl_tree(ubi, e1, &ubi->used); |
23553b2c | 670 | rb_erase(&e1->u.rb, &ubi->used); |
801c135c AB |
671 | dbg_wl("move PEB %d EC %d to PEB %d EC %d", |
672 | e1->pnum, e1->ec, e2->pnum, e2->ec); | |
673 | } else { | |
43f9b25a AB |
674 | /* Perform scrubbing */ |
675 | scrubbing = 1; | |
23553b2c | 676 | e1 = rb_entry(rb_first(&ubi->scrub), struct ubi_wl_entry, u.rb); |
801c135c | 677 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); |
7bf523ae | 678 | self_check_in_wl_tree(ubi, e1, &ubi->scrub); |
23553b2c | 679 | rb_erase(&e1->u.rb, &ubi->scrub); |
801c135c AB |
680 | dbg_wl("scrub PEB %d to PEB %d", e1->pnum, e2->pnum); |
681 | } | |
682 | ||
7bf523ae | 683 | self_check_in_wl_tree(ubi, e2, &ubi->free); |
23553b2c | 684 | rb_erase(&e2->u.rb, &ubi->free); |
801c135c AB |
685 | ubi->move_from = e1; |
686 | ubi->move_to = e2; | |
687 | spin_unlock(&ubi->wl_lock); | |
688 | ||
689 | /* | |
690 | * Now we are going to copy physical eraseblock @e1->pnum to @e2->pnum. | |
691 | * We so far do not know which logical eraseblock our physical | |
692 | * eraseblock (@e1) belongs to. We have to read the volume identifier | |
693 | * header first. | |
43f9b25a AB |
694 | * |
695 | * Note, we are protected from this PEB being unmapped and erased. The | |
696 | * 'ubi_wl_put_peb()' would wait for moving to be finished if the PEB | |
697 | * which is being moved was unmapped. | |
801c135c AB |
698 | */ |
699 | ||
700 | err = ubi_io_read_vid_hdr(ubi, e1->pnum, vid_hdr, 0); | |
701 | if (err && err != UBI_IO_BITFLIPS) { | |
74d82d26 | 702 | if (err == UBI_IO_FF) { |
801c135c AB |
703 | /* |
704 | * We are trying to move PEB without a VID header. UBI | |
705 | * always write VID headers shortly after the PEB was | |
87960c0b AB |
706 | * given, so we have a situation when it has not yet |
707 | * had a chance to write it, because it was preempted. | |
708 | * So add this PEB to the protection queue so far, | |
815bc5f8 AB |
709 | * because presumably more data will be written there |
710 | * (including the missing VID header), and then we'll | |
87960c0b | 711 | * move it. |
801c135c AB |
712 | */ |
713 | dbg_wl("PEB %d has no VID header", e1->pnum); | |
87960c0b | 714 | protect = 1; |
43f9b25a | 715 | goto out_not_moved; |
92e1a7d9 AB |
716 | } else if (err == UBI_IO_FF_BITFLIPS) { |
717 | /* | |
718 | * The same situation as %UBI_IO_FF, but bit-flips were | |
719 | * detected. It is better to schedule this PEB for | |
720 | * scrubbing. | |
721 | */ | |
722 | dbg_wl("PEB %d has no VID header but has bit-flips", | |
723 | e1->pnum); | |
724 | scrubbing = 1; | |
725 | goto out_not_moved; | |
801c135c | 726 | } |
43f9b25a AB |
727 | |
728 | ubi_err("error %d while reading VID header from PEB %d", | |
729 | err, e1->pnum); | |
43f9b25a | 730 | goto out_error; |
801c135c AB |
731 | } |
732 | ||
9c259a52 AB |
733 | vol_id = be32_to_cpu(vid_hdr->vol_id); |
734 | lnum = be32_to_cpu(vid_hdr->lnum); | |
735 | ||
801c135c AB |
736 | err = ubi_eba_copy_leb(ubi, e1->pnum, e2->pnum, vid_hdr); |
737 | if (err) { | |
87960c0b AB |
738 | if (err == MOVE_CANCEL_RACE) { |
739 | /* | |
740 | * The LEB has not been moved because the volume is | |
741 | * being deleted or the PEB has been put meanwhile. We | |
742 | * should prevent this PEB from being selected for | |
743 | * wear-leveling movement again, so put it to the | |
744 | * protection queue. | |
745 | */ | |
746 | protect = 1; | |
747 | goto out_not_moved; | |
748 | } | |
e801e128 BP |
749 | if (err == MOVE_RETRY) { |
750 | scrubbing = 1; | |
751 | goto out_not_moved; | |
752 | } | |
cc831464 | 753 | if (err == MOVE_TARGET_BITFLIPS || err == MOVE_TARGET_WR_ERR || |
b86a2c56 | 754 | err == MOVE_TARGET_RD_ERR) { |
9c259a52 AB |
755 | /* |
756 | * Target PEB had bit-flips or write error - torture it. | |
757 | */ | |
6fa6f5bb | 758 | torture = 1; |
43f9b25a | 759 | goto out_not_moved; |
6fa6f5bb | 760 | } |
87960c0b | 761 | |
b86a2c56 AB |
762 | if (err == MOVE_SOURCE_RD_ERR) { |
763 | /* | |
764 | * An error happened while reading the source PEB. Do | |
765 | * not switch to R/O mode in this case, and give the | |
766 | * upper layers a possibility to recover from this, | |
767 | * e.g. by unmapping corresponding LEB. Instead, just | |
815bc5f8 AB |
768 | * put this PEB to the @ubi->erroneous list to prevent |
769 | * UBI from trying to move it over and over again. | |
b86a2c56 AB |
770 | */ |
771 | if (ubi->erroneous_peb_count > ubi->max_erroneous) { | |
772 | ubi_err("too many erroneous eraseblocks (%d)", | |
773 | ubi->erroneous_peb_count); | |
774 | goto out_error; | |
775 | } | |
776 | erroneous = 1; | |
777 | goto out_not_moved; | |
778 | } | |
779 | ||
90bf0265 AB |
780 | if (err < 0) |
781 | goto out_error; | |
43f9b25a | 782 | |
87960c0b | 783 | ubi_assert(0); |
801c135c AB |
784 | } |
785 | ||
6a8f483f | 786 | /* The PEB has been successfully moved */ |
6a8f483f | 787 | if (scrubbing) |
9c259a52 AB |
788 | ubi_msg("scrubbed PEB %d (LEB %d:%d), data moved to PEB %d", |
789 | e1->pnum, vol_id, lnum, e2->pnum); | |
790 | ubi_free_vid_hdr(ubi, vid_hdr); | |
8c1e6ee1 | 791 | |
801c135c | 792 | spin_lock(&ubi->wl_lock); |
3c98b0a0 | 793 | if (!ubi->move_to_put) { |
5abde384 | 794 | wl_tree_add(e2, &ubi->used); |
3c98b0a0 AB |
795 | e2 = NULL; |
796 | } | |
801c135c | 797 | ubi->move_from = ubi->move_to = NULL; |
43f9b25a | 798 | ubi->move_to_put = ubi->wl_scheduled = 0; |
801c135c AB |
799 | spin_unlock(&ubi->wl_lock); |
800 | ||
6a8f483f | 801 | err = schedule_erase(ubi, e1, 0); |
3c98b0a0 | 802 | if (err) { |
87960c0b | 803 | kmem_cache_free(ubi_wl_entry_slab, e1); |
21d08bbc AB |
804 | if (e2) |
805 | kmem_cache_free(ubi_wl_entry_slab, e2); | |
87960c0b | 806 | goto out_ro; |
3c98b0a0 | 807 | } |
6a8f483f | 808 | |
3c98b0a0 | 809 | if (e2) { |
801c135c AB |
810 | /* |
811 | * Well, the target PEB was put meanwhile, schedule it for | |
812 | * erasure. | |
813 | */ | |
9c259a52 AB |
814 | dbg_wl("PEB %d (LEB %d:%d) was put meanwhile, erase", |
815 | e2->pnum, vol_id, lnum); | |
801c135c | 816 | err = schedule_erase(ubi, e2, 0); |
87960c0b AB |
817 | if (err) { |
818 | kmem_cache_free(ubi_wl_entry_slab, e2); | |
819 | goto out_ro; | |
820 | } | |
801c135c AB |
821 | } |
822 | ||
801c135c | 823 | dbg_wl("done"); |
43f9b25a AB |
824 | mutex_unlock(&ubi->move_mutex); |
825 | return 0; | |
801c135c AB |
826 | |
827 | /* | |
43f9b25a AB |
828 | * For some reasons the LEB was not moved, might be an error, might be |
829 | * something else. @e1 was not changed, so return it back. @e2 might | |
6fa6f5bb | 830 | * have been changed, schedule it for erasure. |
801c135c | 831 | */ |
43f9b25a | 832 | out_not_moved: |
9c259a52 AB |
833 | if (vol_id != -1) |
834 | dbg_wl("cancel moving PEB %d (LEB %d:%d) to PEB %d (%d)", | |
835 | e1->pnum, vol_id, lnum, e2->pnum, err); | |
836 | else | |
837 | dbg_wl("cancel moving PEB %d to PEB %d (%d)", | |
838 | e1->pnum, e2->pnum, err); | |
801c135c | 839 | spin_lock(&ubi->wl_lock); |
87960c0b AB |
840 | if (protect) |
841 | prot_queue_add(ubi, e1); | |
b86a2c56 AB |
842 | else if (erroneous) { |
843 | wl_tree_add(e1, &ubi->erroneous); | |
844 | ubi->erroneous_peb_count += 1; | |
845 | } else if (scrubbing) | |
43f9b25a | 846 | wl_tree_add(e1, &ubi->scrub); |
801c135c | 847 | else |
5abde384 | 848 | wl_tree_add(e1, &ubi->used); |
6fa6f5bb | 849 | ubi_assert(!ubi->move_to_put); |
801c135c | 850 | ubi->move_from = ubi->move_to = NULL; |
6fa6f5bb | 851 | ubi->wl_scheduled = 0; |
801c135c AB |
852 | spin_unlock(&ubi->wl_lock); |
853 | ||
87960c0b | 854 | ubi_free_vid_hdr(ubi, vid_hdr); |
6fa6f5bb | 855 | err = schedule_erase(ubi, e2, torture); |
87960c0b AB |
856 | if (err) { |
857 | kmem_cache_free(ubi_wl_entry_slab, e2); | |
858 | goto out_ro; | |
859 | } | |
43f9b25a AB |
860 | mutex_unlock(&ubi->move_mutex); |
861 | return 0; | |
862 | ||
863 | out_error: | |
9c259a52 AB |
864 | if (vol_id != -1) |
865 | ubi_err("error %d while moving PEB %d to PEB %d", | |
866 | err, e1->pnum, e2->pnum); | |
867 | else | |
868 | ubi_err("error %d while moving PEB %d (LEB %d:%d) to PEB %d", | |
869 | err, e1->pnum, vol_id, lnum, e2->pnum); | |
43f9b25a AB |
870 | spin_lock(&ubi->wl_lock); |
871 | ubi->move_from = ubi->move_to = NULL; | |
872 | ubi->move_to_put = ubi->wl_scheduled = 0; | |
873 | spin_unlock(&ubi->wl_lock); | |
874 | ||
87960c0b AB |
875 | ubi_free_vid_hdr(ubi, vid_hdr); |
876 | kmem_cache_free(ubi_wl_entry_slab, e1); | |
877 | kmem_cache_free(ubi_wl_entry_slab, e2); | |
43f9b25a | 878 | |
87960c0b AB |
879 | out_ro: |
880 | ubi_ro_mode(ubi); | |
43f9b25a | 881 | mutex_unlock(&ubi->move_mutex); |
87960c0b AB |
882 | ubi_assert(err != 0); |
883 | return err < 0 ? err : -EIO; | |
43f9b25a AB |
884 | |
885 | out_cancel: | |
886 | ubi->wl_scheduled = 0; | |
887 | spin_unlock(&ubi->wl_lock); | |
888 | mutex_unlock(&ubi->move_mutex); | |
889 | ubi_free_vid_hdr(ubi, vid_hdr); | |
890 | return 0; | |
801c135c AB |
891 | } |
892 | ||
893 | /** | |
894 | * ensure_wear_leveling - schedule wear-leveling if it is needed. | |
895 | * @ubi: UBI device description object | |
896 | * | |
897 | * This function checks if it is time to start wear-leveling and schedules it | |
898 | * if yes. This function returns zero in case of success and a negative error | |
899 | * code in case of failure. | |
900 | */ | |
901 | static int ensure_wear_leveling(struct ubi_device *ubi) | |
902 | { | |
903 | int err = 0; | |
904 | struct ubi_wl_entry *e1; | |
905 | struct ubi_wl_entry *e2; | |
906 | struct ubi_work *wrk; | |
907 | ||
908 | spin_lock(&ubi->wl_lock); | |
909 | if (ubi->wl_scheduled) | |
910 | /* Wear-leveling is already in the work queue */ | |
911 | goto out_unlock; | |
912 | ||
913 | /* | |
914 | * If the ubi->scrub tree is not empty, scrubbing is needed, and the | |
915 | * the WL worker has to be scheduled anyway. | |
916 | */ | |
5abde384 AB |
917 | if (!ubi->scrub.rb_node) { |
918 | if (!ubi->used.rb_node || !ubi->free.rb_node) | |
801c135c AB |
919 | /* No physical eraseblocks - no deal */ |
920 | goto out_unlock; | |
921 | ||
922 | /* | |
923 | * We schedule wear-leveling only if the difference between the | |
924 | * lowest erase counter of used physical eraseblocks and a high | |
025dfdaf | 925 | * erase counter of free physical eraseblocks is greater than |
801c135c AB |
926 | * %UBI_WL_THRESHOLD. |
927 | */ | |
23553b2c | 928 | e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, u.rb); |
801c135c AB |
929 | e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF); |
930 | ||
931 | if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD)) | |
932 | goto out_unlock; | |
933 | dbg_wl("schedule wear-leveling"); | |
934 | } else | |
935 | dbg_wl("schedule scrubbing"); | |
936 | ||
937 | ubi->wl_scheduled = 1; | |
938 | spin_unlock(&ubi->wl_lock); | |
939 | ||
33818bbb | 940 | wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS); |
801c135c AB |
941 | if (!wrk) { |
942 | err = -ENOMEM; | |
943 | goto out_cancel; | |
944 | } | |
945 | ||
946 | wrk->func = &wear_leveling_worker; | |
947 | schedule_ubi_work(ubi, wrk); | |
948 | return err; | |
949 | ||
950 | out_cancel: | |
951 | spin_lock(&ubi->wl_lock); | |
952 | ubi->wl_scheduled = 0; | |
953 | out_unlock: | |
954 | spin_unlock(&ubi->wl_lock); | |
955 | return err; | |
956 | } | |
957 | ||
958 | /** | |
959 | * erase_worker - physical eraseblock erase worker function. | |
960 | * @ubi: UBI device description object | |
961 | * @wl_wrk: the work object | |
962 | * @cancel: non-zero if the worker has to free memory and exit | |
963 | * | |
964 | * This function erases a physical eraseblock and perform torture testing if | |
965 | * needed. It also takes care about marking the physical eraseblock bad if | |
966 | * needed. Returns zero in case of success and a negative error code in case of | |
967 | * failure. | |
968 | */ | |
969 | static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk, | |
970 | int cancel) | |
971 | { | |
801c135c | 972 | struct ubi_wl_entry *e = wl_wrk->e; |
784c1454 | 973 | int pnum = e->pnum, err, need; |
801c135c AB |
974 | |
975 | if (cancel) { | |
976 | dbg_wl("cancel erasure of PEB %d EC %d", pnum, e->ec); | |
977 | kfree(wl_wrk); | |
06b68ba1 | 978 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
979 | return 0; |
980 | } | |
981 | ||
982 | dbg_wl("erase PEB %d EC %d", pnum, e->ec); | |
983 | ||
984 | err = sync_erase(ubi, e, wl_wrk->torture); | |
985 | if (!err) { | |
986 | /* Fine, we've erased it successfully */ | |
987 | kfree(wl_wrk); | |
988 | ||
989 | spin_lock(&ubi->wl_lock); | |
5abde384 | 990 | wl_tree_add(e, &ubi->free); |
801c135c AB |
991 | spin_unlock(&ubi->wl_lock); |
992 | ||
993 | /* | |
9c9ec147 AB |
994 | * One more erase operation has happened, take care about |
995 | * protected physical eraseblocks. | |
801c135c | 996 | */ |
7b6c32da | 997 | serve_prot_queue(ubi); |
801c135c AB |
998 | |
999 | /* And take care about wear-leveling */ | |
1000 | err = ensure_wear_leveling(ubi); | |
1001 | return err; | |
1002 | } | |
1003 | ||
8d2d4011 | 1004 | ubi_err("failed to erase PEB %d, error %d", pnum, err); |
801c135c | 1005 | kfree(wl_wrk); |
801c135c | 1006 | |
784c1454 AB |
1007 | if (err == -EINTR || err == -ENOMEM || err == -EAGAIN || |
1008 | err == -EBUSY) { | |
1009 | int err1; | |
1010 | ||
1011 | /* Re-schedule the LEB for erasure */ | |
1012 | err1 = schedule_erase(ubi, e, 0); | |
1013 | if (err1) { | |
1014 | err = err1; | |
1015 | goto out_ro; | |
1016 | } | |
1017 | return err; | |
e57e0d8e AB |
1018 | } |
1019 | ||
1020 | kmem_cache_free(ubi_wl_entry_slab, e); | |
1021 | if (err != -EIO) | |
801c135c AB |
1022 | /* |
1023 | * If this is not %-EIO, we have no idea what to do. Scheduling | |
1024 | * this physical eraseblock for erasure again would cause | |
815bc5f8 | 1025 | * errors again and again. Well, lets switch to R/O mode. |
801c135c | 1026 | */ |
784c1454 | 1027 | goto out_ro; |
801c135c AB |
1028 | |
1029 | /* It is %-EIO, the PEB went bad */ | |
1030 | ||
1031 | if (!ubi->bad_allowed) { | |
1032 | ubi_err("bad physical eraseblock %d detected", pnum); | |
784c1454 AB |
1033 | goto out_ro; |
1034 | } | |
801c135c | 1035 | |
784c1454 AB |
1036 | spin_lock(&ubi->volumes_lock); |
1037 | need = ubi->beb_rsvd_level - ubi->beb_rsvd_pebs + 1; | |
1038 | if (need > 0) { | |
1039 | need = ubi->avail_pebs >= need ? need : ubi->avail_pebs; | |
1040 | ubi->avail_pebs -= need; | |
1041 | ubi->rsvd_pebs += need; | |
1042 | ubi->beb_rsvd_pebs += need; | |
1043 | if (need > 0) | |
1044 | ubi_msg("reserve more %d PEBs", need); | |
1045 | } | |
801c135c | 1046 | |
784c1454 | 1047 | if (ubi->beb_rsvd_pebs == 0) { |
801c135c | 1048 | spin_unlock(&ubi->volumes_lock); |
784c1454 AB |
1049 | ubi_err("no reserved physical eraseblocks"); |
1050 | goto out_ro; | |
1051 | } | |
784c1454 | 1052 | spin_unlock(&ubi->volumes_lock); |
801c135c | 1053 | |
52b605d1 | 1054 | ubi_msg("mark PEB %d as bad", pnum); |
784c1454 AB |
1055 | err = ubi_io_mark_bad(ubi, pnum); |
1056 | if (err) | |
1057 | goto out_ro; | |
1058 | ||
1059 | spin_lock(&ubi->volumes_lock); | |
1060 | ubi->beb_rsvd_pebs -= 1; | |
1061 | ubi->bad_peb_count += 1; | |
1062 | ubi->good_peb_count -= 1; | |
1063 | ubi_calculate_reserved(ubi); | |
52b605d1 AB |
1064 | if (ubi->beb_rsvd_pebs) |
1065 | ubi_msg("%d PEBs left in the reserve", ubi->beb_rsvd_pebs); | |
1066 | else | |
784c1454 AB |
1067 | ubi_warn("last PEB from the reserved pool was used"); |
1068 | spin_unlock(&ubi->volumes_lock); | |
1069 | ||
1070 | return err; | |
801c135c | 1071 | |
784c1454 AB |
1072 | out_ro: |
1073 | ubi_ro_mode(ubi); | |
801c135c AB |
1074 | return err; |
1075 | } | |
1076 | ||
1077 | /** | |
85c6e6e2 | 1078 | * ubi_wl_put_peb - return a PEB to the wear-leveling sub-system. |
801c135c AB |
1079 | * @ubi: UBI device description object |
1080 | * @pnum: physical eraseblock to return | |
1081 | * @torture: if this physical eraseblock has to be tortured | |
1082 | * | |
1083 | * This function is called to return physical eraseblock @pnum to the pool of | |
1084 | * free physical eraseblocks. The @torture flag has to be set if an I/O error | |
1085 | * occurred to this @pnum and it has to be tested. This function returns zero | |
43f9b25a | 1086 | * in case of success, and a negative error code in case of failure. |
801c135c AB |
1087 | */ |
1088 | int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture) | |
1089 | { | |
1090 | int err; | |
1091 | struct ubi_wl_entry *e; | |
1092 | ||
1093 | dbg_wl("PEB %d", pnum); | |
1094 | ubi_assert(pnum >= 0); | |
1095 | ubi_assert(pnum < ubi->peb_count); | |
1096 | ||
43f9b25a | 1097 | retry: |
801c135c | 1098 | spin_lock(&ubi->wl_lock); |
801c135c AB |
1099 | e = ubi->lookuptbl[pnum]; |
1100 | if (e == ubi->move_from) { | |
1101 | /* | |
1102 | * User is putting the physical eraseblock which was selected to | |
1103 | * be moved. It will be scheduled for erasure in the | |
1104 | * wear-leveling worker. | |
1105 | */ | |
43f9b25a | 1106 | dbg_wl("PEB %d is being moved, wait", pnum); |
801c135c | 1107 | spin_unlock(&ubi->wl_lock); |
43f9b25a AB |
1108 | |
1109 | /* Wait for the WL worker by taking the @ubi->move_mutex */ | |
1110 | mutex_lock(&ubi->move_mutex); | |
1111 | mutex_unlock(&ubi->move_mutex); | |
1112 | goto retry; | |
801c135c AB |
1113 | } else if (e == ubi->move_to) { |
1114 | /* | |
1115 | * User is putting the physical eraseblock which was selected | |
1116 | * as the target the data is moved to. It may happen if the EBA | |
85c6e6e2 AB |
1117 | * sub-system already re-mapped the LEB in 'ubi_eba_copy_leb()' |
1118 | * but the WL sub-system has not put the PEB to the "used" tree | |
1119 | * yet, but it is about to do this. So we just set a flag which | |
1120 | * will tell the WL worker that the PEB is not needed anymore | |
1121 | * and should be scheduled for erasure. | |
801c135c AB |
1122 | */ |
1123 | dbg_wl("PEB %d is the target of data moving", pnum); | |
1124 | ubi_assert(!ubi->move_to_put); | |
1125 | ubi->move_to_put = 1; | |
1126 | spin_unlock(&ubi->wl_lock); | |
1127 | return 0; | |
1128 | } else { | |
5abde384 | 1129 | if (in_wl_tree(e, &ubi->used)) { |
7bf523ae | 1130 | self_check_in_wl_tree(ubi, e, &ubi->used); |
23553b2c | 1131 | rb_erase(&e->u.rb, &ubi->used); |
5abde384 | 1132 | } else if (in_wl_tree(e, &ubi->scrub)) { |
7bf523ae | 1133 | self_check_in_wl_tree(ubi, e, &ubi->scrub); |
23553b2c | 1134 | rb_erase(&e->u.rb, &ubi->scrub); |
b86a2c56 | 1135 | } else if (in_wl_tree(e, &ubi->erroneous)) { |
7bf523ae | 1136 | self_check_in_wl_tree(ubi, e, &ubi->erroneous); |
b86a2c56 AB |
1137 | rb_erase(&e->u.rb, &ubi->erroneous); |
1138 | ubi->erroneous_peb_count -= 1; | |
1139 | ubi_assert(ubi->erroneous_peb_count >= 0); | |
815bc5f8 | 1140 | /* Erroneous PEBs should be tortured */ |
b86a2c56 | 1141 | torture = 1; |
43f9b25a | 1142 | } else { |
7b6c32da | 1143 | err = prot_queue_del(ubi, e->pnum); |
43f9b25a AB |
1144 | if (err) { |
1145 | ubi_err("PEB %d not found", pnum); | |
1146 | ubi_ro_mode(ubi); | |
1147 | spin_unlock(&ubi->wl_lock); | |
1148 | return err; | |
1149 | } | |
1150 | } | |
801c135c AB |
1151 | } |
1152 | spin_unlock(&ubi->wl_lock); | |
1153 | ||
1154 | err = schedule_erase(ubi, e, torture); | |
1155 | if (err) { | |
1156 | spin_lock(&ubi->wl_lock); | |
5abde384 | 1157 | wl_tree_add(e, &ubi->used); |
801c135c AB |
1158 | spin_unlock(&ubi->wl_lock); |
1159 | } | |
1160 | ||
1161 | return err; | |
1162 | } | |
1163 | ||
1164 | /** | |
1165 | * ubi_wl_scrub_peb - schedule a physical eraseblock for scrubbing. | |
1166 | * @ubi: UBI device description object | |
1167 | * @pnum: the physical eraseblock to schedule | |
1168 | * | |
1169 | * If a bit-flip in a physical eraseblock is detected, this physical eraseblock | |
1170 | * needs scrubbing. This function schedules a physical eraseblock for | |
1171 | * scrubbing which is done in background. This function returns zero in case of | |
1172 | * success and a negative error code in case of failure. | |
1173 | */ | |
1174 | int ubi_wl_scrub_peb(struct ubi_device *ubi, int pnum) | |
1175 | { | |
1176 | struct ubi_wl_entry *e; | |
1177 | ||
8c1e6ee1 | 1178 | dbg_msg("schedule PEB %d for scrubbing", pnum); |
801c135c AB |
1179 | |
1180 | retry: | |
1181 | spin_lock(&ubi->wl_lock); | |
1182 | e = ubi->lookuptbl[pnum]; | |
d3f6e6c6 AB |
1183 | if (e == ubi->move_from || in_wl_tree(e, &ubi->scrub) || |
1184 | in_wl_tree(e, &ubi->erroneous)) { | |
801c135c AB |
1185 | spin_unlock(&ubi->wl_lock); |
1186 | return 0; | |
1187 | } | |
1188 | ||
1189 | if (e == ubi->move_to) { | |
1190 | /* | |
1191 | * This physical eraseblock was used to move data to. The data | |
1192 | * was moved but the PEB was not yet inserted to the proper | |
1193 | * tree. We should just wait a little and let the WL worker | |
1194 | * proceed. | |
1195 | */ | |
1196 | spin_unlock(&ubi->wl_lock); | |
1197 | dbg_wl("the PEB %d is not in proper tree, retry", pnum); | |
1198 | yield(); | |
1199 | goto retry; | |
1200 | } | |
1201 | ||
5abde384 | 1202 | if (in_wl_tree(e, &ubi->used)) { |
7bf523ae | 1203 | self_check_in_wl_tree(ubi, e, &ubi->used); |
23553b2c | 1204 | rb_erase(&e->u.rb, &ubi->used); |
43f9b25a AB |
1205 | } else { |
1206 | int err; | |
1207 | ||
7b6c32da | 1208 | err = prot_queue_del(ubi, e->pnum); |
43f9b25a AB |
1209 | if (err) { |
1210 | ubi_err("PEB %d not found", pnum); | |
1211 | ubi_ro_mode(ubi); | |
1212 | spin_unlock(&ubi->wl_lock); | |
1213 | return err; | |
1214 | } | |
1215 | } | |
801c135c | 1216 | |
5abde384 | 1217 | wl_tree_add(e, &ubi->scrub); |
801c135c AB |
1218 | spin_unlock(&ubi->wl_lock); |
1219 | ||
1220 | /* | |
1221 | * Technically scrubbing is the same as wear-leveling, so it is done | |
1222 | * by the WL worker. | |
1223 | */ | |
1224 | return ensure_wear_leveling(ubi); | |
1225 | } | |
1226 | ||
1227 | /** | |
1228 | * ubi_wl_flush - flush all pending works. | |
1229 | * @ubi: UBI device description object | |
1230 | * | |
1231 | * This function returns zero in case of success and a negative error code in | |
1232 | * case of failure. | |
1233 | */ | |
1234 | int ubi_wl_flush(struct ubi_device *ubi) | |
1235 | { | |
593dd33c | 1236 | int err; |
801c135c AB |
1237 | |
1238 | /* | |
7b6c32da | 1239 | * Erase while the pending works queue is not empty, but not more than |
801c135c AB |
1240 | * the number of currently pending works. |
1241 | */ | |
593dd33c AB |
1242 | dbg_wl("flush (%d pending works)", ubi->works_count); |
1243 | while (ubi->works_count) { | |
1244 | err = do_work(ubi); | |
1245 | if (err) | |
1246 | return err; | |
1247 | } | |
1248 | ||
1249 | /* | |
1250 | * Make sure all the works which have been done in parallel are | |
1251 | * finished. | |
1252 | */ | |
1253 | down_write(&ubi->work_sem); | |
1254 | up_write(&ubi->work_sem); | |
1255 | ||
1256 | /* | |
6fa6f5bb | 1257 | * And in case last was the WL worker and it canceled the LEB |
593dd33c AB |
1258 | * movement, flush again. |
1259 | */ | |
1260 | while (ubi->works_count) { | |
1261 | dbg_wl("flush more (%d pending works)", ubi->works_count); | |
801c135c AB |
1262 | err = do_work(ubi); |
1263 | if (err) | |
1264 | return err; | |
1265 | } | |
1266 | ||
1267 | return 0; | |
1268 | } | |
1269 | ||
1270 | /** | |
1271 | * tree_destroy - destroy an RB-tree. | |
1272 | * @root: the root of the tree to destroy | |
1273 | */ | |
1274 | static void tree_destroy(struct rb_root *root) | |
1275 | { | |
1276 | struct rb_node *rb; | |
1277 | struct ubi_wl_entry *e; | |
1278 | ||
1279 | rb = root->rb_node; | |
1280 | while (rb) { | |
1281 | if (rb->rb_left) | |
1282 | rb = rb->rb_left; | |
1283 | else if (rb->rb_right) | |
1284 | rb = rb->rb_right; | |
1285 | else { | |
23553b2c | 1286 | e = rb_entry(rb, struct ubi_wl_entry, u.rb); |
801c135c AB |
1287 | |
1288 | rb = rb_parent(rb); | |
1289 | if (rb) { | |
23553b2c | 1290 | if (rb->rb_left == &e->u.rb) |
801c135c AB |
1291 | rb->rb_left = NULL; |
1292 | else | |
1293 | rb->rb_right = NULL; | |
1294 | } | |
1295 | ||
06b68ba1 | 1296 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1297 | } |
1298 | } | |
1299 | } | |
1300 | ||
1301 | /** | |
1302 | * ubi_thread - UBI background thread. | |
1303 | * @u: the UBI device description object pointer | |
1304 | */ | |
cdfa788a | 1305 | int ubi_thread(void *u) |
801c135c AB |
1306 | { |
1307 | int failures = 0; | |
1308 | struct ubi_device *ubi = u; | |
1309 | ||
1310 | ubi_msg("background thread \"%s\" started, PID %d", | |
ba25f9dc | 1311 | ubi->bgt_name, task_pid_nr(current)); |
801c135c | 1312 | |
83144186 | 1313 | set_freezable(); |
801c135c AB |
1314 | for (;;) { |
1315 | int err; | |
1316 | ||
1317 | if (kthread_should_stop()) | |
cadb40cc | 1318 | break; |
801c135c AB |
1319 | |
1320 | if (try_to_freeze()) | |
1321 | continue; | |
1322 | ||
1323 | spin_lock(&ubi->wl_lock); | |
1324 | if (list_empty(&ubi->works) || ubi->ro_mode || | |
27a0f2a3 | 1325 | !ubi->thread_enabled || ubi_dbg_is_bgt_disabled(ubi)) { |
801c135c AB |
1326 | set_current_state(TASK_INTERRUPTIBLE); |
1327 | spin_unlock(&ubi->wl_lock); | |
1328 | schedule(); | |
1329 | continue; | |
1330 | } | |
1331 | spin_unlock(&ubi->wl_lock); | |
1332 | ||
1333 | err = do_work(ubi); | |
1334 | if (err) { | |
1335 | ubi_err("%s: work failed with error code %d", | |
1336 | ubi->bgt_name, err); | |
1337 | if (failures++ > WL_MAX_FAILURES) { | |
1338 | /* | |
1339 | * Too many failures, disable the thread and | |
1340 | * switch to read-only mode. | |
1341 | */ | |
1342 | ubi_msg("%s: %d consecutive failures", | |
1343 | ubi->bgt_name, WL_MAX_FAILURES); | |
1344 | ubi_ro_mode(ubi); | |
2ad49887 VG |
1345 | ubi->thread_enabled = 0; |
1346 | continue; | |
801c135c AB |
1347 | } |
1348 | } else | |
1349 | failures = 0; | |
1350 | ||
1351 | cond_resched(); | |
1352 | } | |
1353 | ||
801c135c AB |
1354 | dbg_wl("background thread \"%s\" is killed", ubi->bgt_name); |
1355 | return 0; | |
1356 | } | |
1357 | ||
1358 | /** | |
1359 | * cancel_pending - cancel all pending works. | |
1360 | * @ubi: UBI device description object | |
1361 | */ | |
1362 | static void cancel_pending(struct ubi_device *ubi) | |
1363 | { | |
1364 | while (!list_empty(&ubi->works)) { | |
1365 | struct ubi_work *wrk; | |
1366 | ||
1367 | wrk = list_entry(ubi->works.next, struct ubi_work, list); | |
1368 | list_del(&wrk->list); | |
1369 | wrk->func(ubi, wrk, 1); | |
1370 | ubi->works_count -= 1; | |
1371 | ubi_assert(ubi->works_count >= 0); | |
1372 | } | |
1373 | } | |
1374 | ||
1375 | /** | |
a4e6042f | 1376 | * ubi_wl_init_scan - initialize the WL sub-system using attaching information. |
801c135c | 1377 | * @ubi: UBI device description object |
a4e6042f | 1378 | * @ai: attaching information |
801c135c AB |
1379 | * |
1380 | * This function returns zero in case of success, and a negative error code in | |
1381 | * case of failure. | |
1382 | */ | |
a4e6042f | 1383 | int ubi_wl_init_scan(struct ubi_device *ubi, struct ubi_attach_info *ai) |
801c135c | 1384 | { |
7b6c32da | 1385 | int err, i; |
801c135c | 1386 | struct rb_node *rb1, *rb2; |
517af48c | 1387 | struct ubi_ainf_volume *av; |
2c5ec5ce | 1388 | struct ubi_ainf_peb *aeb, *tmp; |
801c135c AB |
1389 | struct ubi_wl_entry *e; |
1390 | ||
b86a2c56 | 1391 | ubi->used = ubi->erroneous = ubi->free = ubi->scrub = RB_ROOT; |
801c135c | 1392 | spin_lock_init(&ubi->wl_lock); |
43f9b25a | 1393 | mutex_init(&ubi->move_mutex); |
593dd33c | 1394 | init_rwsem(&ubi->work_sem); |
a4e6042f | 1395 | ubi->max_ec = ai->max_ec; |
801c135c AB |
1396 | INIT_LIST_HEAD(&ubi->works); |
1397 | ||
1398 | sprintf(ubi->bgt_name, UBI_BGT_NAME_PATTERN, ubi->ubi_num); | |
1399 | ||
801c135c AB |
1400 | err = -ENOMEM; |
1401 | ubi->lookuptbl = kzalloc(ubi->peb_count * sizeof(void *), GFP_KERNEL); | |
1402 | if (!ubi->lookuptbl) | |
cdfa788a | 1403 | return err; |
801c135c | 1404 | |
7b6c32da XX |
1405 | for (i = 0; i < UBI_PROT_QUEUE_LEN; i++) |
1406 | INIT_LIST_HEAD(&ubi->pq[i]); | |
1407 | ubi->pq_head = 0; | |
1408 | ||
a4e6042f | 1409 | list_for_each_entry_safe(aeb, tmp, &ai->erase, u.list) { |
801c135c AB |
1410 | cond_resched(); |
1411 | ||
06b68ba1 | 1412 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1413 | if (!e) |
1414 | goto out_free; | |
1415 | ||
2c5ec5ce AB |
1416 | e->pnum = aeb->pnum; |
1417 | e->ec = aeb->ec; | |
801c135c AB |
1418 | ubi->lookuptbl[e->pnum] = e; |
1419 | if (schedule_erase(ubi, e, 0)) { | |
06b68ba1 | 1420 | kmem_cache_free(ubi_wl_entry_slab, e); |
801c135c AB |
1421 | goto out_free; |
1422 | } | |
1423 | } | |
1424 | ||
a4e6042f | 1425 | list_for_each_entry(aeb, &ai->free, u.list) { |
801c135c AB |
1426 | cond_resched(); |
1427 | ||
06b68ba1 | 1428 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1429 | if (!e) |
1430 | goto out_free; | |
1431 | ||
2c5ec5ce AB |
1432 | e->pnum = aeb->pnum; |
1433 | e->ec = aeb->ec; | |
801c135c | 1434 | ubi_assert(e->ec >= 0); |
5abde384 | 1435 | wl_tree_add(e, &ubi->free); |
801c135c AB |
1436 | ubi->lookuptbl[e->pnum] = e; |
1437 | } | |
1438 | ||
517af48c AB |
1439 | ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb) { |
1440 | ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb) { | |
801c135c AB |
1441 | cond_resched(); |
1442 | ||
06b68ba1 | 1443 | e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL); |
801c135c AB |
1444 | if (!e) |
1445 | goto out_free; | |
1446 | ||
2c5ec5ce AB |
1447 | e->pnum = aeb->pnum; |
1448 | e->ec = aeb->ec; | |
801c135c | 1449 | ubi->lookuptbl[e->pnum] = e; |
2c5ec5ce | 1450 | if (!aeb->scrub) { |
801c135c AB |
1451 | dbg_wl("add PEB %d EC %d to the used tree", |
1452 | e->pnum, e->ec); | |
5abde384 | 1453 | wl_tree_add(e, &ubi->used); |
801c135c AB |
1454 | } else { |
1455 | dbg_wl("add PEB %d EC %d to the scrub tree", | |
1456 | e->pnum, e->ec); | |
5abde384 | 1457 | wl_tree_add(e, &ubi->scrub); |
801c135c AB |
1458 | } |
1459 | } | |
1460 | } | |
1461 | ||
5abde384 | 1462 | if (ubi->avail_pebs < WL_RESERVED_PEBS) { |
801c135c AB |
1463 | ubi_err("no enough physical eraseblocks (%d, need %d)", |
1464 | ubi->avail_pebs, WL_RESERVED_PEBS); | |
5fc01ab6 AB |
1465 | if (ubi->corr_peb_count) |
1466 | ubi_err("%d PEBs are corrupted and not used", | |
1467 | ubi->corr_peb_count); | |
801c135c AB |
1468 | goto out_free; |
1469 | } | |
1470 | ubi->avail_pebs -= WL_RESERVED_PEBS; | |
1471 | ubi->rsvd_pebs += WL_RESERVED_PEBS; | |
1472 | ||
1473 | /* Schedule wear-leveling if needed */ | |
1474 | err = ensure_wear_leveling(ubi); | |
1475 | if (err) | |
1476 | goto out_free; | |
1477 | ||
1478 | return 0; | |
1479 | ||
1480 | out_free: | |
1481 | cancel_pending(ubi); | |
1482 | tree_destroy(&ubi->used); | |
1483 | tree_destroy(&ubi->free); | |
1484 | tree_destroy(&ubi->scrub); | |
1485 | kfree(ubi->lookuptbl); | |
801c135c AB |
1486 | return err; |
1487 | } | |
1488 | ||
1489 | /** | |
7b6c32da | 1490 | * protection_queue_destroy - destroy the protection queue. |
801c135c AB |
1491 | * @ubi: UBI device description object |
1492 | */ | |
7b6c32da | 1493 | static void protection_queue_destroy(struct ubi_device *ubi) |
801c135c | 1494 | { |
7b6c32da XX |
1495 | int i; |
1496 | struct ubi_wl_entry *e, *tmp; | |
801c135c | 1497 | |
7b6c32da XX |
1498 | for (i = 0; i < UBI_PROT_QUEUE_LEN; ++i) { |
1499 | list_for_each_entry_safe(e, tmp, &ubi->pq[i], u.list) { | |
1500 | list_del(&e->u.list); | |
1501 | kmem_cache_free(ubi_wl_entry_slab, e); | |
801c135c AB |
1502 | } |
1503 | } | |
1504 | } | |
1505 | ||
1506 | /** | |
85c6e6e2 | 1507 | * ubi_wl_close - close the wear-leveling sub-system. |
801c135c AB |
1508 | * @ubi: UBI device description object |
1509 | */ | |
1510 | void ubi_wl_close(struct ubi_device *ubi) | |
1511 | { | |
85c6e6e2 | 1512 | dbg_wl("close the WL sub-system"); |
801c135c | 1513 | cancel_pending(ubi); |
7b6c32da | 1514 | protection_queue_destroy(ubi); |
801c135c | 1515 | tree_destroy(&ubi->used); |
b86a2c56 | 1516 | tree_destroy(&ubi->erroneous); |
801c135c AB |
1517 | tree_destroy(&ubi->free); |
1518 | tree_destroy(&ubi->scrub); | |
1519 | kfree(ubi->lookuptbl); | |
801c135c AB |
1520 | } |
1521 | ||
801c135c | 1522 | /** |
7bf523ae | 1523 | * self_check_ec - make sure that the erase counter of a PEB is correct. |
801c135c AB |
1524 | * @ubi: UBI device description object |
1525 | * @pnum: the physical eraseblock number to check | |
1526 | * @ec: the erase counter to check | |
1527 | * | |
1528 | * This function returns zero if the erase counter of physical eraseblock @pnum | |
feddbb34 AB |
1529 | * is equivalent to @ec, and a negative error code if not or if an error |
1530 | * occurred. | |
801c135c | 1531 | */ |
7bf523ae | 1532 | static int self_check_ec(struct ubi_device *ubi, int pnum, int ec) |
801c135c AB |
1533 | { |
1534 | int err; | |
1535 | long long read_ec; | |
1536 | struct ubi_ec_hdr *ec_hdr; | |
1537 | ||
2a734bb8 | 1538 | if (!ubi->dbg->chk_gen) |
92d124f5 AB |
1539 | return 0; |
1540 | ||
33818bbb | 1541 | ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_NOFS); |
801c135c AB |
1542 | if (!ec_hdr) |
1543 | return -ENOMEM; | |
1544 | ||
1545 | err = ubi_io_read_ec_hdr(ubi, pnum, ec_hdr, 0); | |
1546 | if (err && err != UBI_IO_BITFLIPS) { | |
1547 | /* The header does not have to exist */ | |
1548 | err = 0; | |
1549 | goto out_free; | |
1550 | } | |
1551 | ||
3261ebd7 | 1552 | read_ec = be64_to_cpu(ec_hdr->ec); |
801c135c | 1553 | if (ec != read_ec) { |
7bf523ae | 1554 | ubi_err("self-check failed for PEB %d", pnum); |
801c135c | 1555 | ubi_err("read EC is %lld, should be %d", read_ec, ec); |
25886a36 | 1556 | dump_stack(); |
801c135c AB |
1557 | err = 1; |
1558 | } else | |
1559 | err = 0; | |
1560 | ||
1561 | out_free: | |
1562 | kfree(ec_hdr); | |
1563 | return err; | |
1564 | } | |
1565 | ||
1566 | /** | |
7bf523ae | 1567 | * self_check_in_wl_tree - check that wear-leveling entry is in WL RB-tree. |
d99383b0 | 1568 | * @ubi: UBI device description object |
801c135c AB |
1569 | * @e: the wear-leveling entry to check |
1570 | * @root: the root of the tree | |
1571 | * | |
adbf05e3 AB |
1572 | * This function returns zero if @e is in the @root RB-tree and %-EINVAL if it |
1573 | * is not. | |
801c135c | 1574 | */ |
7bf523ae AB |
1575 | static int self_check_in_wl_tree(const struct ubi_device *ubi, |
1576 | struct ubi_wl_entry *e, struct rb_root *root) | |
801c135c | 1577 | { |
2a734bb8 | 1578 | if (!ubi->dbg->chk_gen) |
92d124f5 AB |
1579 | return 0; |
1580 | ||
801c135c AB |
1581 | if (in_wl_tree(e, root)) |
1582 | return 0; | |
1583 | ||
7bf523ae | 1584 | ubi_err("self-check failed for PEB %d, EC %d, RB-tree %p ", |
801c135c | 1585 | e->pnum, e->ec, root); |
25886a36 | 1586 | dump_stack(); |
adbf05e3 | 1587 | return -EINVAL; |
801c135c AB |
1588 | } |
1589 | ||
7b6c32da | 1590 | /** |
7bf523ae | 1591 | * self_check_in_pq - check if wear-leveling entry is in the protection |
7b6c32da XX |
1592 | * queue. |
1593 | * @ubi: UBI device description object | |
1594 | * @e: the wear-leveling entry to check | |
1595 | * | |
adbf05e3 | 1596 | * This function returns zero if @e is in @ubi->pq and %-EINVAL if it is not. |
7b6c32da | 1597 | */ |
7bf523ae AB |
1598 | static int self_check_in_pq(const struct ubi_device *ubi, |
1599 | struct ubi_wl_entry *e) | |
7b6c32da XX |
1600 | { |
1601 | struct ubi_wl_entry *p; | |
1602 | int i; | |
1603 | ||
2a734bb8 | 1604 | if (!ubi->dbg->chk_gen) |
92d124f5 AB |
1605 | return 0; |
1606 | ||
7b6c32da XX |
1607 | for (i = 0; i < UBI_PROT_QUEUE_LEN; ++i) |
1608 | list_for_each_entry(p, &ubi->pq[i], u.list) | |
1609 | if (p == e) | |
1610 | return 0; | |
1611 | ||
7bf523ae | 1612 | ubi_err("self-check failed for PEB %d, EC %d, Protect queue", |
7b6c32da | 1613 | e->pnum, e->ec); |
25886a36 | 1614 | dump_stack(); |
adbf05e3 | 1615 | return -EINVAL; |
7b6c32da | 1616 | } |