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