[go: nahoru, domu]

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 * Author: Artem Bityutskiy (Битюцкий Артём)
19 */
20
21/*
22 * The UBI Eraseblock Association (EBA) sub-system.
23 *
24 * This sub-system is responsible for I/O to/from logical eraseblock.
25 *
26 * Although in this implementation the EBA table is fully kept and managed in
27 * RAM, which assumes poor scalability, it might be (partially) maintained on
28 * flash in future implementations.
29 *
30 * The EBA sub-system implements per-logical eraseblock locking. Before
31 * accessing a logical eraseblock it is locked for reading or writing. The
32 * per-logical eraseblock locking is implemented by means of the lock tree. The
33 * lock tree is an RB-tree which refers all the currently locked logical
34 * eraseblocks. The lock tree elements are &struct ubi_ltree_entry objects.
35 * They are indexed by (@vol_id, @lnum) pairs.
36 *
37 * EBA also maintains the global sequence counter which is incremented each
38 * time a logical eraseblock is mapped to a physical eraseblock and it is
39 * stored in the volume identifier header. This means that each VID header has
40 * a unique sequence number. The sequence number is only increased an we assume
41 * 64 bits is enough to never overflow.
42 */
43
44#include <linux/slab.h>
45#include <linux/crc32.h>
46#include <linux/err.h>
47#include "ubi.h"
48
49/* Number of physical eraseblocks reserved for atomic LEB change operation */
50#define EBA_RESERVED_PEBS 1
51
52/**
53 * next_sqnum - get next sequence number.
54 * @ubi: UBI device description object
55 *
56 * This function returns next sequence number to use, which is just the current
57 * global sequence counter value. It also increases the global sequence
58 * counter.
59 */
60unsigned long long ubi_next_sqnum(struct ubi_device *ubi)
61{
62	unsigned long long sqnum;
63
64	spin_lock(&ubi->ltree_lock);
65	sqnum = ubi->global_sqnum++;
66	spin_unlock(&ubi->ltree_lock);
67
68	return sqnum;
69}
70
71/**
72 * ubi_get_compat - get compatibility flags of a volume.
73 * @ubi: UBI device description object
74 * @vol_id: volume ID
75 *
76 * This function returns compatibility flags for an internal volume. User
77 * volumes have no compatibility flags, so %0 is returned.
78 */
79static int ubi_get_compat(const struct ubi_device *ubi, int vol_id)
80{
81	if (vol_id == UBI_LAYOUT_VOLUME_ID)
82		return UBI_LAYOUT_VOLUME_COMPAT;
83	return 0;
84}
85
86/**
87 * ltree_lookup - look up the lock tree.
88 * @ubi: UBI device description object
89 * @vol_id: volume ID
90 * @lnum: logical eraseblock number
91 *
92 * This function returns a pointer to the corresponding &struct ubi_ltree_entry
93 * object if the logical eraseblock is locked and %NULL if it is not.
94 * @ubi->ltree_lock has to be locked.
95 */
96static struct ubi_ltree_entry *ltree_lookup(struct ubi_device *ubi, int vol_id,
97					    int lnum)
98{
99	struct rb_node *p;
100
101	p = ubi->ltree.rb_node;
102	while (p) {
103		struct ubi_ltree_entry *le;
104
105		le = rb_entry(p, struct ubi_ltree_entry, rb);
106
107		if (vol_id < le->vol_id)
108			p = p->rb_left;
109		else if (vol_id > le->vol_id)
110			p = p->rb_right;
111		else {
112			if (lnum < le->lnum)
113				p = p->rb_left;
114			else if (lnum > le->lnum)
115				p = p->rb_right;
116			else
117				return le;
118		}
119	}
120
121	return NULL;
122}
123
124/**
125 * ltree_add_entry - add new entry to the lock tree.
126 * @ubi: UBI device description object
127 * @vol_id: volume ID
128 * @lnum: logical eraseblock number
129 *
130 * This function adds new entry for logical eraseblock (@vol_id, @lnum) to the
131 * lock tree. If such entry is already there, its usage counter is increased.
132 * Returns pointer to the lock tree entry or %-ENOMEM if memory allocation
133 * failed.
134 */
135static struct ubi_ltree_entry *ltree_add_entry(struct ubi_device *ubi,
136					       int vol_id, int lnum)
137{
138	struct ubi_ltree_entry *le, *le1, *le_free;
139
140	le = kmalloc(sizeof(struct ubi_ltree_entry), GFP_NOFS);
141	if (!le)
142		return ERR_PTR(-ENOMEM);
143
144	le->users = 0;
145	init_rwsem(&le->mutex);
146	le->vol_id = vol_id;
147	le->lnum = lnum;
148
149	spin_lock(&ubi->ltree_lock);
150	le1 = ltree_lookup(ubi, vol_id, lnum);
151
152	if (le1) {
153		/*
154		 * This logical eraseblock is already locked. The newly
155		 * allocated lock entry is not needed.
156		 */
157		le_free = le;
158		le = le1;
159	} else {
160		struct rb_node **p, *parent = NULL;
161
162		/*
163		 * No lock entry, add the newly allocated one to the
164		 * @ubi->ltree RB-tree.
165		 */
166		le_free = NULL;
167
168		p = &ubi->ltree.rb_node;
169		while (*p) {
170			parent = *p;
171			le1 = rb_entry(parent, struct ubi_ltree_entry, rb);
172
173			if (vol_id < le1->vol_id)
174				p = &(*p)->rb_left;
175			else if (vol_id > le1->vol_id)
176				p = &(*p)->rb_right;
177			else {
178				ubi_assert(lnum != le1->lnum);
179				if (lnum < le1->lnum)
180					p = &(*p)->rb_left;
181				else
182					p = &(*p)->rb_right;
183			}
184		}
185
186		rb_link_node(&le->rb, parent, p);
187		rb_insert_color(&le->rb, &ubi->ltree);
188	}
189	le->users += 1;
190	spin_unlock(&ubi->ltree_lock);
191
192	kfree(le_free);
193	return le;
194}
195
196/**
197 * leb_read_lock - lock logical eraseblock for reading.
198 * @ubi: UBI device description object
199 * @vol_id: volume ID
200 * @lnum: logical eraseblock number
201 *
202 * This function locks a logical eraseblock for reading. Returns zero in case
203 * of success and a negative error code in case of failure.
204 */
205static int leb_read_lock(struct ubi_device *ubi, int vol_id, int lnum)
206{
207	struct ubi_ltree_entry *le;
208
209	le = ltree_add_entry(ubi, vol_id, lnum);
210	if (IS_ERR(le))
211		return PTR_ERR(le);
212	down_read(&le->mutex);
213	return 0;
214}
215
216/**
217 * leb_read_unlock - unlock logical eraseblock.
218 * @ubi: UBI device description object
219 * @vol_id: volume ID
220 * @lnum: logical eraseblock number
221 */
222static void leb_read_unlock(struct ubi_device *ubi, int vol_id, int lnum)
223{
224	struct ubi_ltree_entry *le;
225
226	spin_lock(&ubi->ltree_lock);
227	le = ltree_lookup(ubi, vol_id, lnum);
228	le->users -= 1;
229	ubi_assert(le->users >= 0);
230	up_read(&le->mutex);
231	if (le->users == 0) {
232		rb_erase(&le->rb, &ubi->ltree);
233		kfree(le);
234	}
235	spin_unlock(&ubi->ltree_lock);
236}
237
238/**
239 * leb_write_lock - lock logical eraseblock for writing.
240 * @ubi: UBI device description object
241 * @vol_id: volume ID
242 * @lnum: logical eraseblock number
243 *
244 * This function locks a logical eraseblock for writing. Returns zero in case
245 * of success and a negative error code in case of failure.
246 */
247static int leb_write_lock(struct ubi_device *ubi, int vol_id, int lnum)
248{
249	struct ubi_ltree_entry *le;
250
251	le = ltree_add_entry(ubi, vol_id, lnum);
252	if (IS_ERR(le))
253		return PTR_ERR(le);
254	down_write(&le->mutex);
255	return 0;
256}
257
258/**
259 * leb_write_lock - lock logical eraseblock for writing.
260 * @ubi: UBI device description object
261 * @vol_id: volume ID
262 * @lnum: logical eraseblock number
263 *
264 * This function locks a logical eraseblock for writing if there is no
265 * contention and does nothing if there is contention. Returns %0 in case of
266 * success, %1 in case of contention, and and a negative error code in case of
267 * failure.
268 */
269static int leb_write_trylock(struct ubi_device *ubi, int vol_id, int lnum)
270{
271	struct ubi_ltree_entry *le;
272
273	le = ltree_add_entry(ubi, vol_id, lnum);
274	if (IS_ERR(le))
275		return PTR_ERR(le);
276	if (down_write_trylock(&le->mutex))
277		return 0;
278
279	/* Contention, cancel */
280	spin_lock(&ubi->ltree_lock);
281	le->users -= 1;
282	ubi_assert(le->users >= 0);
283	if (le->users == 0) {
284		rb_erase(&le->rb, &ubi->ltree);
285		kfree(le);
286	}
287	spin_unlock(&ubi->ltree_lock);
288
289	return 1;
290}
291
292/**
293 * leb_write_unlock - unlock logical eraseblock.
294 * @ubi: UBI device description object
295 * @vol_id: volume ID
296 * @lnum: logical eraseblock number
297 */
298static void leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
299{
300	struct ubi_ltree_entry *le;
301
302	spin_lock(&ubi->ltree_lock);
303	le = ltree_lookup(ubi, vol_id, lnum);
304	le->users -= 1;
305	ubi_assert(le->users >= 0);
306	up_write(&le->mutex);
307	if (le->users == 0) {
308		rb_erase(&le->rb, &ubi->ltree);
309		kfree(le);
310	}
311	spin_unlock(&ubi->ltree_lock);
312}
313
314/**
315 * ubi_eba_unmap_leb - un-map logical eraseblock.
316 * @ubi: UBI device description object
317 * @vol: volume description object
318 * @lnum: logical eraseblock number
319 *
320 * This function un-maps logical eraseblock @lnum and schedules corresponding
321 * physical eraseblock for erasure. Returns zero in case of success and a
322 * negative error code in case of failure.
323 */
324int ubi_eba_unmap_leb(struct ubi_device *ubi, struct ubi_volume *vol,
325		      int lnum)
326{
327	int err, pnum, vol_id = vol->vol_id;
328
329	if (ubi->ro_mode)
330		return -EROFS;
331
332	err = leb_write_lock(ubi, vol_id, lnum);
333	if (err)
334		return err;
335
336	pnum = vol->eba_tbl[lnum];
337	if (pnum < 0)
338		/* This logical eraseblock is already unmapped */
339		goto out_unlock;
340
341	dbg_eba("erase LEB %d:%d, PEB %d", vol_id, lnum, pnum);
342
343	down_read(&ubi->fm_sem);
344	vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED;
345	up_read(&ubi->fm_sem);
346	err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 0);
347
348out_unlock:
349	leb_write_unlock(ubi, vol_id, lnum);
350	return err;
351}
352
353/**
354 * ubi_eba_read_leb - read data.
355 * @ubi: UBI device description object
356 * @vol: volume description object
357 * @lnum: logical eraseblock number
358 * @buf: buffer to store the read data
359 * @offset: offset from where to read
360 * @len: how many bytes to read
361 * @check: data CRC check flag
362 *
363 * If the logical eraseblock @lnum is unmapped, @buf is filled with 0xFF
364 * bytes. The @check flag only makes sense for static volumes and forces
365 * eraseblock data CRC checking.
366 *
367 * In case of success this function returns zero. In case of a static volume,
368 * if data CRC mismatches - %-EBADMSG is returned. %-EBADMSG may also be
369 * returned for any volume type if an ECC error was detected by the MTD device
370 * driver. Other negative error cored may be returned in case of other errors.
371 */
372int ubi_eba_read_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
373		     void *buf, int offset, int len, int check)
374{
375	int err, pnum, scrub = 0, vol_id = vol->vol_id;
376	struct ubi_vid_hdr *vid_hdr;
377	uint32_t uninitialized_var(crc);
378
379	err = leb_read_lock(ubi, vol_id, lnum);
380	if (err)
381		return err;
382
383	pnum = vol->eba_tbl[lnum];
384	if (pnum < 0) {
385		/*
386		 * The logical eraseblock is not mapped, fill the whole buffer
387		 * with 0xFF bytes. The exception is static volumes for which
388		 * it is an error to read unmapped logical eraseblocks.
389		 */
390		dbg_eba("read %d bytes from offset %d of LEB %d:%d (unmapped)",
391			len, offset, vol_id, lnum);
392		leb_read_unlock(ubi, vol_id, lnum);
393		ubi_assert(vol->vol_type != UBI_STATIC_VOLUME);
394		memset(buf, 0xFF, len);
395		return 0;
396	}
397
398	dbg_eba("read %d bytes from offset %d of LEB %d:%d, PEB %d",
399		len, offset, vol_id, lnum, pnum);
400
401	if (vol->vol_type == UBI_DYNAMIC_VOLUME)
402		check = 0;
403
404retry:
405	if (check) {
406		vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
407		if (!vid_hdr) {
408			err = -ENOMEM;
409			goto out_unlock;
410		}
411
412		err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
413		if (err && err != UBI_IO_BITFLIPS) {
414			if (err > 0) {
415				/*
416				 * The header is either absent or corrupted.
417				 * The former case means there is a bug -
418				 * switch to read-only mode just in case.
419				 * The latter case means a real corruption - we
420				 * may try to recover data. FIXME: but this is
421				 * not implemented.
422				 */
423				if (err == UBI_IO_BAD_HDR_EBADMSG ||
424				    err == UBI_IO_BAD_HDR) {
425					ubi_warn("corrupted VID header at PEB %d, LEB %d:%d",
426						 pnum, vol_id, lnum);
427					err = -EBADMSG;
428				} else
429					ubi_ro_mode(ubi);
430			}
431			goto out_free;
432		} else if (err == UBI_IO_BITFLIPS)
433			scrub = 1;
434
435		ubi_assert(lnum < be32_to_cpu(vid_hdr->used_ebs));
436		ubi_assert(len == be32_to_cpu(vid_hdr->data_size));
437
438		crc = be32_to_cpu(vid_hdr->data_crc);
439		ubi_free_vid_hdr(ubi, vid_hdr);
440	}
441
442	err = ubi_io_read_data(ubi, buf, pnum, offset, len);
443	if (err) {
444		if (err == UBI_IO_BITFLIPS)
445			scrub = 1;
446		else if (mtd_is_eccerr(err)) {
447			if (vol->vol_type == UBI_DYNAMIC_VOLUME)
448				goto out_unlock;
449			scrub = 1;
450			if (!check) {
451				ubi_msg("force data checking");
452				check = 1;
453				goto retry;
454			}
455		} else
456			goto out_unlock;
457	}
458
459	if (check) {
460		uint32_t crc1 = crc32(UBI_CRC32_INIT, buf, len);
461		if (crc1 != crc) {
462			ubi_warn("CRC error: calculated %#08x, must be %#08x",
463				 crc1, crc);
464			err = -EBADMSG;
465			goto out_unlock;
466		}
467	}
468
469	if (scrub)
470		err = ubi_wl_scrub_peb(ubi, pnum);
471
472	leb_read_unlock(ubi, vol_id, lnum);
473	return err;
474
475out_free:
476	ubi_free_vid_hdr(ubi, vid_hdr);
477out_unlock:
478	leb_read_unlock(ubi, vol_id, lnum);
479	return err;
480}
481
482/**
483 * recover_peb - recover from write failure.
484 * @ubi: UBI device description object
485 * @pnum: the physical eraseblock to recover
486 * @vol_id: volume ID
487 * @lnum: logical eraseblock number
488 * @buf: data which was not written because of the write failure
489 * @offset: offset of the failed write
490 * @len: how many bytes should have been written
491 *
492 * This function is called in case of a write failure and moves all good data
493 * from the potentially bad physical eraseblock to a good physical eraseblock.
494 * This function also writes the data which was not written due to the failure.
495 * Returns new physical eraseblock number in case of success, and a negative
496 * error code in case of failure.
497 */
498static int recover_peb(struct ubi_device *ubi, int pnum, int vol_id, int lnum,
499		       const void *buf, int offset, int len)
500{
501	int err, idx = vol_id2idx(ubi, vol_id), new_pnum, data_size, tries = 0;
502	struct ubi_volume *vol = ubi->volumes[idx];
503	struct ubi_vid_hdr *vid_hdr;
504
505	vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
506	if (!vid_hdr)
507		return -ENOMEM;
508
509retry:
510	new_pnum = ubi_wl_get_peb(ubi);
511	if (new_pnum < 0) {
512		ubi_free_vid_hdr(ubi, vid_hdr);
513		return new_pnum;
514	}
515
516	ubi_msg("recover PEB %d, move data to PEB %d", pnum, new_pnum);
517
518	err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
519	if (err && err != UBI_IO_BITFLIPS) {
520		if (err > 0)
521			err = -EIO;
522		goto out_put;
523	}
524
525	vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
526	err = ubi_io_write_vid_hdr(ubi, new_pnum, vid_hdr);
527	if (err)
528		goto write_error;
529
530	data_size = offset + len;
531	mutex_lock(&ubi->buf_mutex);
532	memset(ubi->peb_buf + offset, 0xFF, len);
533
534	/* Read everything before the area where the write failure happened */
535	if (offset > 0) {
536		err = ubi_io_read_data(ubi, ubi->peb_buf, pnum, 0, offset);
537		if (err && err != UBI_IO_BITFLIPS)
538			goto out_unlock;
539	}
540
541	memcpy(ubi->peb_buf + offset, buf, len);
542
543	err = ubi_io_write_data(ubi, ubi->peb_buf, new_pnum, 0, data_size);
544	if (err) {
545		mutex_unlock(&ubi->buf_mutex);
546		goto write_error;
547	}
548
549	mutex_unlock(&ubi->buf_mutex);
550	ubi_free_vid_hdr(ubi, vid_hdr);
551
552	down_read(&ubi->fm_sem);
553	vol->eba_tbl[lnum] = new_pnum;
554	up_read(&ubi->fm_sem);
555	ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
556
557	ubi_msg("data was successfully recovered");
558	return 0;
559
560out_unlock:
561	mutex_unlock(&ubi->buf_mutex);
562out_put:
563	ubi_wl_put_peb(ubi, vol_id, lnum, new_pnum, 1);
564	ubi_free_vid_hdr(ubi, vid_hdr);
565	return err;
566
567write_error:
568	/*
569	 * Bad luck? This physical eraseblock is bad too? Crud. Let's try to
570	 * get another one.
571	 */
572	ubi_warn("failed to write to PEB %d", new_pnum);
573	ubi_wl_put_peb(ubi, vol_id, lnum, new_pnum, 1);
574	if (++tries > UBI_IO_RETRIES) {
575		ubi_free_vid_hdr(ubi, vid_hdr);
576		return err;
577	}
578	ubi_msg("try again");
579	goto retry;
580}
581
582/**
583 * ubi_eba_write_leb - write data to dynamic volume.
584 * @ubi: UBI device description object
585 * @vol: volume description object
586 * @lnum: logical eraseblock number
587 * @buf: the data to write
588 * @offset: offset within the logical eraseblock where to write
589 * @len: how many bytes to write
590 *
591 * This function writes data to logical eraseblock @lnum of a dynamic volume
592 * @vol. Returns zero in case of success and a negative error code in case
593 * of failure. In case of error, it is possible that something was still
594 * written to the flash media, but may be some garbage.
595 */
596int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
597		      const void *buf, int offset, int len)
598{
599	int err, pnum, tries = 0, vol_id = vol->vol_id;
600	struct ubi_vid_hdr *vid_hdr;
601
602	if (ubi->ro_mode)
603		return -EROFS;
604
605	err = leb_write_lock(ubi, vol_id, lnum);
606	if (err)
607		return err;
608
609	pnum = vol->eba_tbl[lnum];
610	if (pnum >= 0) {
611		dbg_eba("write %d bytes at offset %d of LEB %d:%d, PEB %d",
612			len, offset, vol_id, lnum, pnum);
613
614		err = ubi_io_write_data(ubi, buf, pnum, offset, len);
615		if (err) {
616			ubi_warn("failed to write data to PEB %d", pnum);
617			if (err == -EIO && ubi->bad_allowed)
618				err = recover_peb(ubi, pnum, vol_id, lnum, buf,
619						  offset, len);
620			if (err)
621				ubi_ro_mode(ubi);
622		}
623		leb_write_unlock(ubi, vol_id, lnum);
624		return err;
625	}
626
627	/*
628	 * The logical eraseblock is not mapped. We have to get a free physical
629	 * eraseblock and write the volume identifier header there first.
630	 */
631	vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
632	if (!vid_hdr) {
633		leb_write_unlock(ubi, vol_id, lnum);
634		return -ENOMEM;
635	}
636
637	vid_hdr->vol_type = UBI_VID_DYNAMIC;
638	vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
639	vid_hdr->vol_id = cpu_to_be32(vol_id);
640	vid_hdr->lnum = cpu_to_be32(lnum);
641	vid_hdr->compat = ubi_get_compat(ubi, vol_id);
642	vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
643
644retry:
645	pnum = ubi_wl_get_peb(ubi);
646	if (pnum < 0) {
647		ubi_free_vid_hdr(ubi, vid_hdr);
648		leb_write_unlock(ubi, vol_id, lnum);
649		return pnum;
650	}
651
652	dbg_eba("write VID hdr and %d bytes at offset %d of LEB %d:%d, PEB %d",
653		len, offset, vol_id, lnum, pnum);
654
655	err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
656	if (err) {
657		ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
658			 vol_id, lnum, pnum);
659		goto write_error;
660	}
661
662	if (len) {
663		err = ubi_io_write_data(ubi, buf, pnum, offset, len);
664		if (err) {
665			ubi_warn("failed to write %d bytes at offset %d of LEB %d:%d, PEB %d",
666				 len, offset, vol_id, lnum, pnum);
667			goto write_error;
668		}
669	}
670
671	down_read(&ubi->fm_sem);
672	vol->eba_tbl[lnum] = pnum;
673	up_read(&ubi->fm_sem);
674
675	leb_write_unlock(ubi, vol_id, lnum);
676	ubi_free_vid_hdr(ubi, vid_hdr);
677	return 0;
678
679write_error:
680	if (err != -EIO || !ubi->bad_allowed) {
681		ubi_ro_mode(ubi);
682		leb_write_unlock(ubi, vol_id, lnum);
683		ubi_free_vid_hdr(ubi, vid_hdr);
684		return err;
685	}
686
687	/*
688	 * Fortunately, this is the first write operation to this physical
689	 * eraseblock, so just put it and request a new one. We assume that if
690	 * this physical eraseblock went bad, the erase code will handle that.
691	 */
692	err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
693	if (err || ++tries > UBI_IO_RETRIES) {
694		ubi_ro_mode(ubi);
695		leb_write_unlock(ubi, vol_id, lnum);
696		ubi_free_vid_hdr(ubi, vid_hdr);
697		return err;
698	}
699
700	vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
701	ubi_msg("try another PEB");
702	goto retry;
703}
704
705/**
706 * ubi_eba_write_leb_st - write data to static volume.
707 * @ubi: UBI device description object
708 * @vol: volume description object
709 * @lnum: logical eraseblock number
710 * @buf: data to write
711 * @len: how many bytes to write
712 * @used_ebs: how many logical eraseblocks will this volume contain
713 *
714 * This function writes data to logical eraseblock @lnum of static volume
715 * @vol. The @used_ebs argument should contain total number of logical
716 * eraseblock in this static volume.
717 *
718 * When writing to the last logical eraseblock, the @len argument doesn't have
719 * to be aligned to the minimal I/O unit size. Instead, it has to be equivalent
720 * to the real data size, although the @buf buffer has to contain the
721 * alignment. In all other cases, @len has to be aligned.
722 *
723 * It is prohibited to write more than once to logical eraseblocks of static
724 * volumes. This function returns zero in case of success and a negative error
725 * code in case of failure.
726 */
727int ubi_eba_write_leb_st(struct ubi_device *ubi, struct ubi_volume *vol,
728			 int lnum, const void *buf, int len, int used_ebs)
729{
730	int err, pnum, tries = 0, data_size = len, vol_id = vol->vol_id;
731	struct ubi_vid_hdr *vid_hdr;
732	uint32_t crc;
733
734	if (ubi->ro_mode)
735		return -EROFS;
736
737	if (lnum == used_ebs - 1)
738		/* If this is the last LEB @len may be unaligned */
739		len = ALIGN(data_size, ubi->min_io_size);
740	else
741		ubi_assert(!(len & (ubi->min_io_size - 1)));
742
743	vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
744	if (!vid_hdr)
745		return -ENOMEM;
746
747	err = leb_write_lock(ubi, vol_id, lnum);
748	if (err) {
749		ubi_free_vid_hdr(ubi, vid_hdr);
750		return err;
751	}
752
753	vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
754	vid_hdr->vol_id = cpu_to_be32(vol_id);
755	vid_hdr->lnum = cpu_to_be32(lnum);
756	vid_hdr->compat = ubi_get_compat(ubi, vol_id);
757	vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
758
759	crc = crc32(UBI_CRC32_INIT, buf, data_size);
760	vid_hdr->vol_type = UBI_VID_STATIC;
761	vid_hdr->data_size = cpu_to_be32(data_size);
762	vid_hdr->used_ebs = cpu_to_be32(used_ebs);
763	vid_hdr->data_crc = cpu_to_be32(crc);
764
765retry:
766	pnum = ubi_wl_get_peb(ubi);
767	if (pnum < 0) {
768		ubi_free_vid_hdr(ubi, vid_hdr);
769		leb_write_unlock(ubi, vol_id, lnum);
770		return pnum;
771	}
772
773	dbg_eba("write VID hdr and %d bytes at LEB %d:%d, PEB %d, used_ebs %d",
774		len, vol_id, lnum, pnum, used_ebs);
775
776	err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
777	if (err) {
778		ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
779			 vol_id, lnum, pnum);
780		goto write_error;
781	}
782
783	err = ubi_io_write_data(ubi, buf, pnum, 0, len);
784	if (err) {
785		ubi_warn("failed to write %d bytes of data to PEB %d",
786			 len, pnum);
787		goto write_error;
788	}
789
790	ubi_assert(vol->eba_tbl[lnum] < 0);
791	down_read(&ubi->fm_sem);
792	vol->eba_tbl[lnum] = pnum;
793	up_read(&ubi->fm_sem);
794
795	leb_write_unlock(ubi, vol_id, lnum);
796	ubi_free_vid_hdr(ubi, vid_hdr);
797	return 0;
798
799write_error:
800	if (err != -EIO || !ubi->bad_allowed) {
801		/*
802		 * This flash device does not admit of bad eraseblocks or
803		 * something nasty and unexpected happened. Switch to read-only
804		 * mode just in case.
805		 */
806		ubi_ro_mode(ubi);
807		leb_write_unlock(ubi, vol_id, lnum);
808		ubi_free_vid_hdr(ubi, vid_hdr);
809		return err;
810	}
811
812	err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
813	if (err || ++tries > UBI_IO_RETRIES) {
814		ubi_ro_mode(ubi);
815		leb_write_unlock(ubi, vol_id, lnum);
816		ubi_free_vid_hdr(ubi, vid_hdr);
817		return err;
818	}
819
820	vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
821	ubi_msg("try another PEB");
822	goto retry;
823}
824
825/*
826 * ubi_eba_atomic_leb_change - change logical eraseblock atomically.
827 * @ubi: UBI device description object
828 * @vol: volume description object
829 * @lnum: logical eraseblock number
830 * @buf: data to write
831 * @len: how many bytes to write
832 *
833 * This function changes the contents of a logical eraseblock atomically. @buf
834 * has to contain new logical eraseblock data, and @len - the length of the
835 * data, which has to be aligned. This function guarantees that in case of an
836 * unclean reboot the old contents is preserved. Returns zero in case of
837 * success and a negative error code in case of failure.
838 *
839 * UBI reserves one LEB for the "atomic LEB change" operation, so only one
840 * LEB change may be done at a time. This is ensured by @ubi->alc_mutex.
841 */
842int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
843			      int lnum, const void *buf, int len)
844{
845	int err, pnum, tries = 0, vol_id = vol->vol_id;
846	struct ubi_vid_hdr *vid_hdr;
847	uint32_t crc;
848
849	if (ubi->ro_mode)
850		return -EROFS;
851
852	if (len == 0) {
853		/*
854		 * Special case when data length is zero. In this case the LEB
855		 * has to be unmapped and mapped somewhere else.
856		 */
857		err = ubi_eba_unmap_leb(ubi, vol, lnum);
858		if (err)
859			return err;
860		return ubi_eba_write_leb(ubi, vol, lnum, NULL, 0, 0);
861	}
862
863	vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
864	if (!vid_hdr)
865		return -ENOMEM;
866
867	mutex_lock(&ubi->alc_mutex);
868	err = leb_write_lock(ubi, vol_id, lnum);
869	if (err)
870		goto out_mutex;
871
872	vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
873	vid_hdr->vol_id = cpu_to_be32(vol_id);
874	vid_hdr->lnum = cpu_to_be32(lnum);
875	vid_hdr->compat = ubi_get_compat(ubi, vol_id);
876	vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
877
878	crc = crc32(UBI_CRC32_INIT, buf, len);
879	vid_hdr->vol_type = UBI_VID_DYNAMIC;
880	vid_hdr->data_size = cpu_to_be32(len);
881	vid_hdr->copy_flag = 1;
882	vid_hdr->data_crc = cpu_to_be32(crc);
883
884retry:
885	pnum = ubi_wl_get_peb(ubi);
886	if (pnum < 0) {
887		err = pnum;
888		goto out_leb_unlock;
889	}
890
891	dbg_eba("change LEB %d:%d, PEB %d, write VID hdr to PEB %d",
892		vol_id, lnum, vol->eba_tbl[lnum], pnum);
893
894	err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
895	if (err) {
896		ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
897			 vol_id, lnum, pnum);
898		goto write_error;
899	}
900
901	err = ubi_io_write_data(ubi, buf, pnum, 0, len);
902	if (err) {
903		ubi_warn("failed to write %d bytes of data to PEB %d",
904			 len, pnum);
905		goto write_error;
906	}
907
908	if (vol->eba_tbl[lnum] >= 0) {
909		err = ubi_wl_put_peb(ubi, vol_id, lnum, vol->eba_tbl[lnum], 0);
910		if (err)
911			goto out_leb_unlock;
912	}
913
914	down_read(&ubi->fm_sem);
915	vol->eba_tbl[lnum] = pnum;
916	up_read(&ubi->fm_sem);
917
918out_leb_unlock:
919	leb_write_unlock(ubi, vol_id, lnum);
920out_mutex:
921	mutex_unlock(&ubi->alc_mutex);
922	ubi_free_vid_hdr(ubi, vid_hdr);
923	return err;
924
925write_error:
926	if (err != -EIO || !ubi->bad_allowed) {
927		/*
928		 * This flash device does not admit of bad eraseblocks or
929		 * something nasty and unexpected happened. Switch to read-only
930		 * mode just in case.
931		 */
932		ubi_ro_mode(ubi);
933		goto out_leb_unlock;
934	}
935
936	err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
937	if (err || ++tries > UBI_IO_RETRIES) {
938		ubi_ro_mode(ubi);
939		goto out_leb_unlock;
940	}
941
942	vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
943	ubi_msg("try another PEB");
944	goto retry;
945}
946
947/**
948 * is_error_sane - check whether a read error is sane.
949 * @err: code of the error happened during reading
950 *
951 * This is a helper function for 'ubi_eba_copy_leb()' which is called when we
952 * cannot read data from the target PEB (an error @err happened). If the error
953 * code is sane, then we treat this error as non-fatal. Otherwise the error is
954 * fatal and UBI will be switched to R/O mode later.
955 *
956 * The idea is that we try not to switch to R/O mode if the read error is
957 * something which suggests there was a real read problem. E.g., %-EIO. Or a
958 * memory allocation failed (-%ENOMEM). Otherwise, it is safer to switch to R/O
959 * mode, simply because we do not know what happened at the MTD level, and we
960 * cannot handle this. E.g., the underlying driver may have become crazy, and
961 * it is safer to switch to R/O mode to preserve the data.
962 *
963 * And bear in mind, this is about reading from the target PEB, i.e. the PEB
964 * which we have just written.
965 */
966static int is_error_sane(int err)
967{
968	if (err == -EIO || err == -ENOMEM || err == UBI_IO_BAD_HDR ||
969	    err == UBI_IO_BAD_HDR_EBADMSG || err == -ETIMEDOUT)
970		return 0;
971	return 1;
972}
973
974/**
975 * ubi_eba_copy_leb - copy logical eraseblock.
976 * @ubi: UBI device description object
977 * @from: physical eraseblock number from where to copy
978 * @to: physical eraseblock number where to copy
979 * @vid_hdr: VID header of the @from physical eraseblock
980 *
981 * This function copies logical eraseblock from physical eraseblock @from to
982 * physical eraseblock @to. The @vid_hdr buffer may be changed by this
983 * function. Returns:
984 *   o %0 in case of success;
985 *   o %MOVE_CANCEL_RACE, %MOVE_TARGET_WR_ERR, %MOVE_TARGET_BITFLIPS, etc;
986 *   o a negative error code in case of failure.
987 */
988int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
989		     struct ubi_vid_hdr *vid_hdr)
990{
991	int err, vol_id, lnum, data_size, aldata_size, idx;
992	struct ubi_volume *vol;
993	uint32_t crc;
994
995	vol_id = be32_to_cpu(vid_hdr->vol_id);
996	lnum = be32_to_cpu(vid_hdr->lnum);
997
998	dbg_wl("copy LEB %d:%d, PEB %d to PEB %d", vol_id, lnum, from, to);
999
1000	if (vid_hdr->vol_type == UBI_VID_STATIC) {
1001		data_size = be32_to_cpu(vid_hdr->data_size);
1002		aldata_size = ALIGN(data_size, ubi->min_io_size);
1003	} else
1004		data_size = aldata_size =
1005			    ubi->leb_size - be32_to_cpu(vid_hdr->data_pad);
1006
1007	idx = vol_id2idx(ubi, vol_id);
1008	spin_lock(&ubi->volumes_lock);
1009	/*
1010	 * Note, we may race with volume deletion, which means that the volume
1011	 * this logical eraseblock belongs to might be being deleted. Since the
1012	 * volume deletion un-maps all the volume's logical eraseblocks, it will
1013	 * be locked in 'ubi_wl_put_peb()' and wait for the WL worker to finish.
1014	 */
1015	vol = ubi->volumes[idx];
1016	spin_unlock(&ubi->volumes_lock);
1017	if (!vol) {
1018		/* No need to do further work, cancel */
1019		dbg_wl("volume %d is being removed, cancel", vol_id);
1020		return MOVE_CANCEL_RACE;
1021	}
1022
1023	/*
1024	 * We do not want anybody to write to this logical eraseblock while we
1025	 * are moving it, so lock it.
1026	 *
1027	 * Note, we are using non-waiting locking here, because we cannot sleep
1028	 * on the LEB, since it may cause deadlocks. Indeed, imagine a task is
1029	 * unmapping the LEB which is mapped to the PEB we are going to move
1030	 * (@from). This task locks the LEB and goes sleep in the
1031	 * 'ubi_wl_put_peb()' function on the @ubi->move_mutex. In turn, we are
1032	 * holding @ubi->move_mutex and go sleep on the LEB lock. So, if the
1033	 * LEB is already locked, we just do not move it and return
1034	 * %MOVE_RETRY. Note, we do not return %MOVE_CANCEL_RACE here because
1035	 * we do not know the reasons of the contention - it may be just a
1036	 * normal I/O on this LEB, so we want to re-try.
1037	 */
1038	err = leb_write_trylock(ubi, vol_id, lnum);
1039	if (err) {
1040		dbg_wl("contention on LEB %d:%d, cancel", vol_id, lnum);
1041		return MOVE_RETRY;
1042	}
1043
1044	/*
1045	 * The LEB might have been put meanwhile, and the task which put it is
1046	 * probably waiting on @ubi->move_mutex. No need to continue the work,
1047	 * cancel it.
1048	 */
1049	if (vol->eba_tbl[lnum] != from) {
1050		dbg_wl("LEB %d:%d is no longer mapped to PEB %d, mapped to PEB %d, cancel",
1051		       vol_id, lnum, from, vol->eba_tbl[lnum]);
1052		err = MOVE_CANCEL_RACE;
1053		goto out_unlock_leb;
1054	}
1055
1056	/*
1057	 * OK, now the LEB is locked and we can safely start moving it. Since
1058	 * this function utilizes the @ubi->peb_buf buffer which is shared
1059	 * with some other functions - we lock the buffer by taking the
1060	 * @ubi->buf_mutex.
1061	 */
1062	mutex_lock(&ubi->buf_mutex);
1063	dbg_wl("read %d bytes of data", aldata_size);
1064	err = ubi_io_read_data(ubi, ubi->peb_buf, from, 0, aldata_size);
1065	if (err && err != UBI_IO_BITFLIPS) {
1066		ubi_warn("error %d while reading data from PEB %d",
1067			 err, from);
1068		err = MOVE_SOURCE_RD_ERR;
1069		goto out_unlock_buf;
1070	}
1071
1072	/*
1073	 * Now we have got to calculate how much data we have to copy. In
1074	 * case of a static volume it is fairly easy - the VID header contains
1075	 * the data size. In case of a dynamic volume it is more difficult - we
1076	 * have to read the contents, cut 0xFF bytes from the end and copy only
1077	 * the first part. We must do this to avoid writing 0xFF bytes as it
1078	 * may have some side-effects. And not only this. It is important not
1079	 * to include those 0xFFs to CRC because later the they may be filled
1080	 * by data.
1081	 */
1082	if (vid_hdr->vol_type == UBI_VID_DYNAMIC)
1083		aldata_size = data_size =
1084			ubi_calc_data_len(ubi, ubi->peb_buf, data_size);
1085
1086	cond_resched();
1087	crc = crc32(UBI_CRC32_INIT, ubi->peb_buf, data_size);
1088	cond_resched();
1089
1090	/*
1091	 * It may turn out to be that the whole @from physical eraseblock
1092	 * contains only 0xFF bytes. Then we have to only write the VID header
1093	 * and do not write any data. This also means we should not set
1094	 * @vid_hdr->copy_flag, @vid_hdr->data_size, and @vid_hdr->data_crc.
1095	 */
1096	if (data_size > 0) {
1097		vid_hdr->copy_flag = 1;
1098		vid_hdr->data_size = cpu_to_be32(data_size);
1099		vid_hdr->data_crc = cpu_to_be32(crc);
1100	}
1101	vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
1102
1103	err = ubi_io_write_vid_hdr(ubi, to, vid_hdr);
1104	if (err) {
1105		if (err == -EIO)
1106			err = MOVE_TARGET_WR_ERR;
1107		goto out_unlock_buf;
1108	}
1109
1110	cond_resched();
1111
1112	/* Read the VID header back and check if it was written correctly */
1113	err = ubi_io_read_vid_hdr(ubi, to, vid_hdr, 1);
1114	if (err) {
1115		if (err != UBI_IO_BITFLIPS) {
1116			ubi_warn("error %d while reading VID header back from PEB %d",
1117				 err, to);
1118			if (is_error_sane(err))
1119				err = MOVE_TARGET_RD_ERR;
1120		} else
1121			err = MOVE_TARGET_BITFLIPS;
1122		goto out_unlock_buf;
1123	}
1124
1125	if (data_size > 0) {
1126		err = ubi_io_write_data(ubi, ubi->peb_buf, to, 0, aldata_size);
1127		if (err) {
1128			if (err == -EIO)
1129				err = MOVE_TARGET_WR_ERR;
1130			goto out_unlock_buf;
1131		}
1132
1133		cond_resched();
1134
1135		/*
1136		 * We've written the data and are going to read it back to make
1137		 * sure it was written correctly.
1138		 */
1139		memset(ubi->peb_buf, 0xFF, aldata_size);
1140		err = ubi_io_read_data(ubi, ubi->peb_buf, to, 0, aldata_size);
1141		if (err) {
1142			if (err != UBI_IO_BITFLIPS) {
1143				ubi_warn("error %d while reading data back from PEB %d",
1144					 err, to);
1145				if (is_error_sane(err))
1146					err = MOVE_TARGET_RD_ERR;
1147			} else
1148				err = MOVE_TARGET_BITFLIPS;
1149			goto out_unlock_buf;
1150		}
1151
1152		cond_resched();
1153
1154		if (crc != crc32(UBI_CRC32_INIT, ubi->peb_buf, data_size)) {
1155			ubi_warn("read data back from PEB %d and it is different",
1156				 to);
1157			err = -EINVAL;
1158			goto out_unlock_buf;
1159		}
1160	}
1161
1162	ubi_assert(vol->eba_tbl[lnum] == from);
1163	down_read(&ubi->fm_sem);
1164	vol->eba_tbl[lnum] = to;
1165	up_read(&ubi->fm_sem);
1166
1167out_unlock_buf:
1168	mutex_unlock(&ubi->buf_mutex);
1169out_unlock_leb:
1170	leb_write_unlock(ubi, vol_id, lnum);
1171	return err;
1172}
1173
1174/**
1175 * print_rsvd_warning - warn about not having enough reserved PEBs.
1176 * @ubi: UBI device description object
1177 *
1178 * This is a helper function for 'ubi_eba_init()' which is called when UBI
1179 * cannot reserve enough PEBs for bad block handling. This function makes a
1180 * decision whether we have to print a warning or not. The algorithm is as
1181 * follows:
1182 *   o if this is a new UBI image, then just print the warning
1183 *   o if this is an UBI image which has already been used for some time, print
1184 *     a warning only if we can reserve less than 10% of the expected amount of
1185 *     the reserved PEB.
1186 *
1187 * The idea is that when UBI is used, PEBs become bad, and the reserved pool
1188 * of PEBs becomes smaller, which is normal and we do not want to scare users
1189 * with a warning every time they attach the MTD device. This was an issue
1190 * reported by real users.
1191 */
1192static void print_rsvd_warning(struct ubi_device *ubi,
1193			       struct ubi_attach_info *ai)
1194{
1195	/*
1196	 * The 1 << 18 (256KiB) number is picked randomly, just a reasonably
1197	 * large number to distinguish between newly flashed and used images.
1198	 */
1199	if (ai->max_sqnum > (1 << 18)) {
1200		int min = ubi->beb_rsvd_level / 10;
1201
1202		if (!min)
1203			min = 1;
1204		if (ubi->beb_rsvd_pebs > min)
1205			return;
1206	}
1207
1208	ubi_warn("cannot reserve enough PEBs for bad PEB handling, reserved %d, need %d",
1209		 ubi->beb_rsvd_pebs, ubi->beb_rsvd_level);
1210	if (ubi->corr_peb_count)
1211		ubi_warn("%d PEBs are corrupted and not used",
1212			 ubi->corr_peb_count);
1213}
1214
1215/**
1216 * self_check_eba - run a self check on the EBA table constructed by fastmap.
1217 * @ubi: UBI device description object
1218 * @ai_fastmap: UBI attach info object created by fastmap
1219 * @ai_scan: UBI attach info object created by scanning
1220 *
1221 * Returns < 0 in case of an internal error, 0 otherwise.
1222 * If a bad EBA table entry was found it will be printed out and
1223 * ubi_assert() triggers.
1224 */
1225int self_check_eba(struct ubi_device *ubi, struct ubi_attach_info *ai_fastmap,
1226		   struct ubi_attach_info *ai_scan)
1227{
1228	int i, j, num_volumes, ret = 0;
1229	int **scan_eba, **fm_eba;
1230	struct ubi_ainf_volume *av;
1231	struct ubi_volume *vol;
1232	struct ubi_ainf_peb *aeb;
1233	struct rb_node *rb;
1234
1235	num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
1236
1237	scan_eba = kmalloc(sizeof(*scan_eba) * num_volumes, GFP_KERNEL);
1238	if (!scan_eba)
1239		return -ENOMEM;
1240
1241	fm_eba = kmalloc(sizeof(*fm_eba) * num_volumes, GFP_KERNEL);
1242	if (!fm_eba) {
1243		kfree(scan_eba);
1244		return -ENOMEM;
1245	}
1246
1247	for (i = 0; i < num_volumes; i++) {
1248		vol = ubi->volumes[i];
1249		if (!vol)
1250			continue;
1251
1252		scan_eba[i] = kmalloc(vol->reserved_pebs * sizeof(**scan_eba),
1253				      GFP_KERNEL);
1254		if (!scan_eba[i]) {
1255			ret = -ENOMEM;
1256			goto out_free;
1257		}
1258
1259		fm_eba[i] = kmalloc(vol->reserved_pebs * sizeof(**fm_eba),
1260				    GFP_KERNEL);
1261		if (!fm_eba[i]) {
1262			ret = -ENOMEM;
1263			goto out_free;
1264		}
1265
1266		for (j = 0; j < vol->reserved_pebs; j++)
1267			scan_eba[i][j] = fm_eba[i][j] = UBI_LEB_UNMAPPED;
1268
1269		av = ubi_find_av(ai_scan, idx2vol_id(ubi, i));
1270		if (!av)
1271			continue;
1272
1273		ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb)
1274			scan_eba[i][aeb->lnum] = aeb->pnum;
1275
1276		av = ubi_find_av(ai_fastmap, idx2vol_id(ubi, i));
1277		if (!av)
1278			continue;
1279
1280		ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb)
1281			fm_eba[i][aeb->lnum] = aeb->pnum;
1282
1283		for (j = 0; j < vol->reserved_pebs; j++) {
1284			if (scan_eba[i][j] != fm_eba[i][j]) {
1285				if (scan_eba[i][j] == UBI_LEB_UNMAPPED ||
1286					fm_eba[i][j] == UBI_LEB_UNMAPPED)
1287					continue;
1288
1289				ubi_err("LEB:%i:%i is PEB:%i instead of %i!",
1290					vol->vol_id, i, fm_eba[i][j],
1291					scan_eba[i][j]);
1292				ubi_assert(0);
1293			}
1294		}
1295	}
1296
1297out_free:
1298	for (i = 0; i < num_volumes; i++) {
1299		if (!ubi->volumes[i])
1300			continue;
1301
1302		kfree(scan_eba[i]);
1303		kfree(fm_eba[i]);
1304	}
1305
1306	kfree(scan_eba);
1307	kfree(fm_eba);
1308	return ret;
1309}
1310
1311/**
1312 * ubi_eba_init - initialize the EBA sub-system using attaching information.
1313 * @ubi: UBI device description object
1314 * @ai: attaching information
1315 *
1316 * This function returns zero in case of success and a negative error code in
1317 * case of failure.
1318 */
1319int ubi_eba_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
1320{
1321	int i, j, err, num_volumes;
1322	struct ubi_ainf_volume *av;
1323	struct ubi_volume *vol;
1324	struct ubi_ainf_peb *aeb;
1325	struct rb_node *rb;
1326
1327	dbg_eba("initialize EBA sub-system");
1328
1329	spin_lock_init(&ubi->ltree_lock);
1330	mutex_init(&ubi->alc_mutex);
1331	ubi->ltree = RB_ROOT;
1332
1333	ubi->global_sqnum = ai->max_sqnum + 1;
1334	num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
1335
1336	for (i = 0; i < num_volumes; i++) {
1337		vol = ubi->volumes[i];
1338		if (!vol)
1339			continue;
1340
1341		cond_resched();
1342
1343		vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int),
1344				       GFP_KERNEL);
1345		if (!vol->eba_tbl) {
1346			err = -ENOMEM;
1347			goto out_free;
1348		}
1349
1350		for (j = 0; j < vol->reserved_pebs; j++)
1351			vol->eba_tbl[j] = UBI_LEB_UNMAPPED;
1352
1353		av = ubi_find_av(ai, idx2vol_id(ubi, i));
1354		if (!av)
1355			continue;
1356
1357		ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb) {
1358			if (aeb->lnum >= vol->reserved_pebs)
1359				/*
1360				 * This may happen in case of an unclean reboot
1361				 * during re-size.
1362				 */
1363				ubi_move_aeb_to_list(av, aeb, &ai->erase);
1364			vol->eba_tbl[aeb->lnum] = aeb->pnum;
1365		}
1366	}
1367
1368	if (ubi->avail_pebs < EBA_RESERVED_PEBS) {
1369		ubi_err("no enough physical eraseblocks (%d, need %d)",
1370			ubi->avail_pebs, EBA_RESERVED_PEBS);
1371		if (ubi->corr_peb_count)
1372			ubi_err("%d PEBs are corrupted and not used",
1373				ubi->corr_peb_count);
1374		err = -ENOSPC;
1375		goto out_free;
1376	}
1377	ubi->avail_pebs -= EBA_RESERVED_PEBS;
1378	ubi->rsvd_pebs += EBA_RESERVED_PEBS;
1379
1380	if (ubi->bad_allowed) {
1381		ubi_calculate_reserved(ubi);
1382
1383		if (ubi->avail_pebs < ubi->beb_rsvd_level) {
1384			/* No enough free physical eraseblocks */
1385			ubi->beb_rsvd_pebs = ubi->avail_pebs;
1386			print_rsvd_warning(ubi, ai);
1387		} else
1388			ubi->beb_rsvd_pebs = ubi->beb_rsvd_level;
1389
1390		ubi->avail_pebs -= ubi->beb_rsvd_pebs;
1391		ubi->rsvd_pebs  += ubi->beb_rsvd_pebs;
1392	}
1393
1394	dbg_eba("EBA sub-system is initialized");
1395	return 0;
1396
1397out_free:
1398	for (i = 0; i < num_volumes; i++) {
1399		if (!ubi->volumes[i])
1400			continue;
1401		kfree(ubi->volumes[i]->eba_tbl);
1402		ubi->volumes[i]->eba_tbl = NULL;
1403	}
1404	return err;
1405}
1406