2 * Copyright (c) International Business Machines Corp., 2006
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.
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.
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
18 * Author: Artem Bityutskiy (Битюцкий Артём)
22 * The UBI Eraseblock Association (EBA) sub-system.
24 * This sub-system is responsible for I/O to/from logical eraseblock.
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.
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.
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.
44 #include <linux/slab.h>
45 #include <linux/crc32.h>
46 #include <linux/err.h>
49 /* Number of physical eraseblocks reserved for atomic LEB change operation */
50 #define EBA_RESERVED_PEBS 1
53 * next_sqnum - get next sequence number.
54 * @ubi: UBI device description object
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
60 static unsigned long long next_sqnum(struct ubi_device *ubi)
62 unsigned long long sqnum;
64 spin_lock(&ubi->ltree_lock);
65 sqnum = ubi->global_sqnum++;
66 spin_unlock(&ubi->ltree_lock);
72 * ubi_get_compat - get compatibility flags of a volume.
73 * @ubi: UBI device description object
76 * This function returns compatibility flags for an internal volume. User
77 * volumes have no compatibility flags, so %0 is returned.
79 static int ubi_get_compat(const struct ubi_device *ubi, int vol_id)
81 if (vol_id == UBI_LAYOUT_VOLUME_ID)
82 return UBI_LAYOUT_VOLUME_COMPAT;
87 * ltree_lookup - look up the lock tree.
88 * @ubi: UBI device description object
90 * @lnum: logical eraseblock number
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.
96 static struct ubi_ltree_entry *ltree_lookup(struct ubi_device *ubi, int vol_id,
101 p = ubi->ltree.rb_node;
103 struct ubi_ltree_entry *le;
105 le = rb_entry(p, struct ubi_ltree_entry, rb);
107 if (vol_id < le->vol_id)
109 else if (vol_id > le->vol_id)
114 else if (lnum > le->lnum)
125 * ltree_add_entry - add new entry to the lock tree.
126 * @ubi: UBI device description object
128 * @lnum: logical eraseblock number
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
135 static struct ubi_ltree_entry *ltree_add_entry(struct ubi_device *ubi,
136 int vol_id, int lnum)
138 struct ubi_ltree_entry *le, *le1, *le_free;
140 le = kmalloc(sizeof(struct ubi_ltree_entry), GFP_NOFS);
142 return ERR_PTR(-ENOMEM);
145 init_rwsem(&le->mutex);
149 spin_lock(&ubi->ltree_lock);
150 le1 = ltree_lookup(ubi, vol_id, lnum);
154 * This logical eraseblock is already locked. The newly
155 * allocated lock entry is not needed.
160 struct rb_node **p, *parent = NULL;
163 * No lock entry, add the newly allocated one to the
164 * @ubi->ltree RB-tree.
168 p = &ubi->ltree.rb_node;
171 le1 = rb_entry(parent, struct ubi_ltree_entry, rb);
173 if (vol_id < le1->vol_id)
175 else if (vol_id > le1->vol_id)
178 ubi_assert(lnum != le1->lnum);
179 if (lnum < le1->lnum)
186 rb_link_node(&le->rb, parent, p);
187 rb_insert_color(&le->rb, &ubi->ltree);
190 spin_unlock(&ubi->ltree_lock);
197 * leb_read_lock - lock logical eraseblock for reading.
198 * @ubi: UBI device description object
200 * @lnum: logical eraseblock number
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.
205 static int leb_read_lock(struct ubi_device *ubi, int vol_id, int lnum)
207 struct ubi_ltree_entry *le;
209 le = ltree_add_entry(ubi, vol_id, lnum);
212 down_read(&le->mutex);
217 * leb_read_unlock - unlock logical eraseblock.
218 * @ubi: UBI device description object
220 * @lnum: logical eraseblock number
222 static void leb_read_unlock(struct ubi_device *ubi, int vol_id, int lnum)
224 struct ubi_ltree_entry *le;
226 spin_lock(&ubi->ltree_lock);
227 le = ltree_lookup(ubi, vol_id, lnum);
229 ubi_assert(le->users >= 0);
231 if (le->users == 0) {
232 rb_erase(&le->rb, &ubi->ltree);
235 spin_unlock(&ubi->ltree_lock);
239 * leb_write_lock - lock logical eraseblock for writing.
240 * @ubi: UBI device description object
242 * @lnum: logical eraseblock number
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.
247 static int leb_write_lock(struct ubi_device *ubi, int vol_id, int lnum)
249 struct ubi_ltree_entry *le;
251 le = ltree_add_entry(ubi, vol_id, lnum);
254 down_write(&le->mutex);
259 * leb_write_lock - lock logical eraseblock for writing.
260 * @ubi: UBI device description object
262 * @lnum: logical eraseblock number
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
269 static int leb_write_trylock(struct ubi_device *ubi, int vol_id, int lnum)
271 struct ubi_ltree_entry *le;
273 le = ltree_add_entry(ubi, vol_id, lnum);
276 if (down_write_trylock(&le->mutex))
279 /* Contention, cancel */
280 spin_lock(&ubi->ltree_lock);
282 ubi_assert(le->users >= 0);
283 if (le->users == 0) {
284 rb_erase(&le->rb, &ubi->ltree);
287 spin_unlock(&ubi->ltree_lock);
293 * leb_write_unlock - unlock logical eraseblock.
294 * @ubi: UBI device description object
296 * @lnum: logical eraseblock number
298 static void leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
300 struct ubi_ltree_entry *le;
302 spin_lock(&ubi->ltree_lock);
303 le = ltree_lookup(ubi, vol_id, lnum);
305 ubi_assert(le->users >= 0);
306 up_write(&le->mutex);
307 if (le->users == 0) {
308 rb_erase(&le->rb, &ubi->ltree);
311 spin_unlock(&ubi->ltree_lock);
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
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.
324 int ubi_eba_unmap_leb(struct ubi_device *ubi, struct ubi_volume *vol,
327 int err, pnum, vol_id = vol->vol_id;
332 err = leb_write_lock(ubi, vol_id, lnum);
336 pnum = vol->eba_tbl[lnum];
338 /* This logical eraseblock is already unmapped */
341 dbg_eba("erase LEB %d:%d, PEB %d", vol_id, lnum, pnum);
343 vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED;
344 err = ubi_wl_put_peb(ubi, pnum, 0);
347 leb_write_unlock(ubi, vol_id, lnum);
352 * ubi_eba_read_leb - read data.
353 * @ubi: UBI device description object
354 * @vol: volume description object
355 * @lnum: logical eraseblock number
356 * @buf: buffer to store the read data
357 * @offset: offset from where to read
358 * @len: how many bytes to read
359 * @check: data CRC check flag
361 * If the logical eraseblock @lnum is unmapped, @buf is filled with 0xFF
362 * bytes. The @check flag only makes sense for static volumes and forces
363 * eraseblock data CRC checking.
365 * In case of success this function returns zero. In case of a static volume,
366 * if data CRC mismatches - %-EBADMSG is returned. %-EBADMSG may also be
367 * returned for any volume type if an ECC error was detected by the MTD device
368 * driver. Other negative error cored may be returned in case of other errors.
370 int ubi_eba_read_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
371 void *buf, int offset, int len, int check)
373 int err, pnum, scrub = 0, vol_id = vol->vol_id;
374 struct ubi_vid_hdr *vid_hdr;
375 uint32_t uninitialized_var(crc);
377 err = leb_read_lock(ubi, vol_id, lnum);
381 pnum = vol->eba_tbl[lnum];
384 * The logical eraseblock is not mapped, fill the whole buffer
385 * with 0xFF bytes. The exception is static volumes for which
386 * it is an error to read unmapped logical eraseblocks.
388 dbg_eba("read %d bytes from offset %d of LEB %d:%d (unmapped)",
389 len, offset, vol_id, lnum);
390 leb_read_unlock(ubi, vol_id, lnum);
391 ubi_assert(vol->vol_type != UBI_STATIC_VOLUME);
392 memset(buf, 0xFF, len);
396 dbg_eba("read %d bytes from offset %d of LEB %d:%d, PEB %d",
397 len, offset, vol_id, lnum, pnum);
399 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
404 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
410 err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
411 if (err && err != UBI_IO_BITFLIPS) {
414 * The header is either absent or corrupted.
415 * The former case means there is a bug -
416 * switch to read-only mode just in case.
417 * The latter case means a real corruption - we
418 * may try to recover data. FIXME: but this is
421 if (err == UBI_IO_BAD_VID_HDR) {
422 ubi_warn("bad VID header at PEB %d, LEB"
423 "%d:%d", pnum, vol_id, lnum);
429 } else if (err == UBI_IO_BITFLIPS)
432 ubi_assert(lnum < be32_to_cpu(vid_hdr->used_ebs));
433 ubi_assert(len == be32_to_cpu(vid_hdr->data_size));
435 crc = be32_to_cpu(vid_hdr->data_crc);
436 ubi_free_vid_hdr(ubi, vid_hdr);
439 err = ubi_io_read_data(ubi, buf, pnum, offset, len);
441 if (err == UBI_IO_BITFLIPS) {
444 } else if (err == -EBADMSG) {
445 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
449 ubi_msg("force data checking");
458 uint32_t crc1 = crc32(UBI_CRC32_INIT, buf, len);
460 ubi_warn("CRC error: calculated %#08x, must be %#08x",
468 err = ubi_wl_scrub_peb(ubi, pnum);
470 leb_read_unlock(ubi, vol_id, lnum);
474 ubi_free_vid_hdr(ubi, vid_hdr);
476 leb_read_unlock(ubi, vol_id, lnum);
481 * recover_peb - recover from write failure.
482 * @ubi: UBI device description object
483 * @pnum: the physical eraseblock to recover
485 * @lnum: logical eraseblock number
486 * @buf: data which was not written because of the write failure
487 * @offset: offset of the failed write
488 * @len: how many bytes should have been written
490 * This function is called in case of a write failure and moves all good data
491 * from the potentially bad physical eraseblock to a good physical eraseblock.
492 * This function also writes the data which was not written due to the failure.
493 * Returns new physical eraseblock number in case of success, and a negative
494 * error code in case of failure.
496 static int recover_peb(struct ubi_device *ubi, int pnum, int vol_id, int lnum,
497 const void *buf, int offset, int len)
499 int err, idx = vol_id2idx(ubi, vol_id), new_pnum, data_size, tries = 0;
500 struct ubi_volume *vol = ubi->volumes[idx];
501 struct ubi_vid_hdr *vid_hdr;
503 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
507 mutex_lock(&ubi->buf_mutex);
510 new_pnum = ubi_wl_get_peb(ubi, UBI_UNKNOWN);
512 mutex_unlock(&ubi->buf_mutex);
513 ubi_free_vid_hdr(ubi, vid_hdr);
517 ubi_msg("recover PEB %d, move data to PEB %d", pnum, new_pnum);
519 err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
520 if (err && err != UBI_IO_BITFLIPS) {
526 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
527 err = ubi_io_write_vid_hdr(ubi, new_pnum, vid_hdr);
531 data_size = offset + len;
532 memset(ubi->peb_buf1 + offset, 0xFF, len);
534 /* Read everything before the area where the write failure happened */
536 err = ubi_io_read_data(ubi, ubi->peb_buf1, pnum, 0, offset);
537 if (err && err != UBI_IO_BITFLIPS)
541 memcpy(ubi->peb_buf1 + offset, buf, len);
543 err = ubi_io_write_data(ubi, ubi->peb_buf1, new_pnum, 0, data_size);
547 mutex_unlock(&ubi->buf_mutex);
548 ubi_free_vid_hdr(ubi, vid_hdr);
550 vol->eba_tbl[lnum] = new_pnum;
551 ubi_wl_put_peb(ubi, pnum, 1);
553 ubi_msg("data was successfully recovered");
557 mutex_unlock(&ubi->buf_mutex);
558 ubi_wl_put_peb(ubi, new_pnum, 1);
559 ubi_free_vid_hdr(ubi, vid_hdr);
564 * Bad luck? This physical eraseblock is bad too? Crud. Let's try to
567 ubi_warn("failed to write to PEB %d", new_pnum);
568 ubi_wl_put_peb(ubi, new_pnum, 1);
569 if (++tries > UBI_IO_RETRIES) {
570 mutex_unlock(&ubi->buf_mutex);
571 ubi_free_vid_hdr(ubi, vid_hdr);
574 ubi_msg("try again");
579 * ubi_eba_write_leb - write data to dynamic volume.
580 * @ubi: UBI device description object
581 * @vol: volume description object
582 * @lnum: logical eraseblock number
583 * @buf: the data to write
584 * @offset: offset within the logical eraseblock where to write
585 * @len: how many bytes to write
588 * This function writes data to logical eraseblock @lnum of a dynamic volume
589 * @vol. Returns zero in case of success and a negative error code in case
590 * of failure. In case of error, it is possible that something was still
591 * written to the flash media, but may be some garbage.
593 int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
594 const void *buf, int offset, int len, int dtype)
596 int err, pnum, tries = 0, vol_id = vol->vol_id;
597 struct ubi_vid_hdr *vid_hdr;
602 err = leb_write_lock(ubi, vol_id, lnum);
606 pnum = vol->eba_tbl[lnum];
608 dbg_eba("write %d bytes at offset %d of LEB %d:%d, PEB %d",
609 len, offset, vol_id, lnum, pnum);
611 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
613 ubi_warn("failed to write data to PEB %d", pnum);
614 if (err == -EIO && ubi->bad_allowed)
615 err = recover_peb(ubi, pnum, vol_id, lnum, buf,
620 leb_write_unlock(ubi, vol_id, lnum);
625 * The logical eraseblock is not mapped. We have to get a free physical
626 * eraseblock and write the volume identifier header there first.
628 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
630 leb_write_unlock(ubi, vol_id, lnum);
634 vid_hdr->vol_type = UBI_VID_DYNAMIC;
635 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
636 vid_hdr->vol_id = cpu_to_be32(vol_id);
637 vid_hdr->lnum = cpu_to_be32(lnum);
638 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
639 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
642 pnum = ubi_wl_get_peb(ubi, dtype);
644 ubi_free_vid_hdr(ubi, vid_hdr);
645 leb_write_unlock(ubi, vol_id, lnum);
649 dbg_eba("write VID hdr and %d bytes at offset %d of LEB %d:%d, PEB %d",
650 len, offset, vol_id, lnum, pnum);
652 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
654 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
660 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
662 ubi_warn("failed to write %d bytes at offset %d of "
663 "LEB %d:%d, PEB %d", len, offset, vol_id,
669 vol->eba_tbl[lnum] = pnum;
671 leb_write_unlock(ubi, vol_id, lnum);
672 ubi_free_vid_hdr(ubi, vid_hdr);
676 if (err != -EIO || !ubi->bad_allowed) {
678 leb_write_unlock(ubi, vol_id, lnum);
679 ubi_free_vid_hdr(ubi, vid_hdr);
684 * Fortunately, this is the first write operation to this physical
685 * eraseblock, so just put it and request a new one. We assume that if
686 * this physical eraseblock went bad, the erase code will handle that.
688 err = ubi_wl_put_peb(ubi, pnum, 1);
689 if (err || ++tries > UBI_IO_RETRIES) {
691 leb_write_unlock(ubi, vol_id, lnum);
692 ubi_free_vid_hdr(ubi, vid_hdr);
696 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
697 ubi_msg("try another PEB");
702 * ubi_eba_write_leb_st - write data to static volume.
703 * @ubi: UBI device description object
704 * @vol: volume description object
705 * @lnum: logical eraseblock number
706 * @buf: data to write
707 * @len: how many bytes to write
709 * @used_ebs: how many logical eraseblocks will this volume contain
711 * This function writes data to logical eraseblock @lnum of static volume
712 * @vol. The @used_ebs argument should contain total number of logical
713 * eraseblock in this static volume.
715 * When writing to the last logical eraseblock, the @len argument doesn't have
716 * to be aligned to the minimal I/O unit size. Instead, it has to be equivalent
717 * to the real data size, although the @buf buffer has to contain the
718 * alignment. In all other cases, @len has to be aligned.
720 * It is prohibited to write more then once to logical eraseblocks of static
721 * volumes. This function returns zero in case of success and a negative error
722 * code in case of failure.
724 int ubi_eba_write_leb_st(struct ubi_device *ubi, struct ubi_volume *vol,
725 int lnum, const void *buf, int len, int dtype,
728 int err, pnum, tries = 0, data_size = len, vol_id = vol->vol_id;
729 struct ubi_vid_hdr *vid_hdr;
735 if (lnum == used_ebs - 1)
736 /* If this is the last LEB @len may be unaligned */
737 len = ALIGN(data_size, ubi->min_io_size);
739 ubi_assert(!(len & (ubi->min_io_size - 1)));
741 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
745 err = leb_write_lock(ubi, vol_id, lnum);
747 ubi_free_vid_hdr(ubi, vid_hdr);
751 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
752 vid_hdr->vol_id = cpu_to_be32(vol_id);
753 vid_hdr->lnum = cpu_to_be32(lnum);
754 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
755 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
757 crc = crc32(UBI_CRC32_INIT, buf, data_size);
758 vid_hdr->vol_type = UBI_VID_STATIC;
759 vid_hdr->data_size = cpu_to_be32(data_size);
760 vid_hdr->used_ebs = cpu_to_be32(used_ebs);
761 vid_hdr->data_crc = cpu_to_be32(crc);
764 pnum = ubi_wl_get_peb(ubi, dtype);
766 ubi_free_vid_hdr(ubi, vid_hdr);
767 leb_write_unlock(ubi, vol_id, lnum);
771 dbg_eba("write VID hdr and %d bytes at LEB %d:%d, PEB %d, used_ebs %d",
772 len, vol_id, lnum, pnum, used_ebs);
774 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
776 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
781 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
783 ubi_warn("failed to write %d bytes of data to PEB %d",
788 ubi_assert(vol->eba_tbl[lnum] < 0);
789 vol->eba_tbl[lnum] = pnum;
791 leb_write_unlock(ubi, vol_id, lnum);
792 ubi_free_vid_hdr(ubi, vid_hdr);
796 if (err != -EIO || !ubi->bad_allowed) {
798 * This flash device does not admit of bad eraseblocks or
799 * something nasty and unexpected happened. Switch to read-only
803 leb_write_unlock(ubi, vol_id, lnum);
804 ubi_free_vid_hdr(ubi, vid_hdr);
808 err = ubi_wl_put_peb(ubi, pnum, 1);
809 if (err || ++tries > UBI_IO_RETRIES) {
811 leb_write_unlock(ubi, vol_id, lnum);
812 ubi_free_vid_hdr(ubi, vid_hdr);
816 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
817 ubi_msg("try another PEB");
822 * ubi_eba_atomic_leb_change - change logical eraseblock atomically.
823 * @ubi: UBI device description object
824 * @vol: volume description object
825 * @lnum: logical eraseblock number
826 * @buf: data to write
827 * @len: how many bytes to write
830 * This function changes the contents of a logical eraseblock atomically. @buf
831 * has to contain new logical eraseblock data, and @len - the length of the
832 * data, which has to be aligned. This function guarantees that in case of an
833 * unclean reboot the old contents is preserved. Returns zero in case of
834 * success and a negative error code in case of failure.
836 * UBI reserves one LEB for the "atomic LEB change" operation, so only one
837 * LEB change may be done at a time. This is ensured by @ubi->alc_mutex.
839 int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
840 int lnum, const void *buf, int len, int dtype)
842 int err, pnum, tries = 0, vol_id = vol->vol_id;
843 struct ubi_vid_hdr *vid_hdr;
851 * Special case when data length is zero. In this case the LEB
852 * has to be unmapped and mapped somewhere else.
854 err = ubi_eba_unmap_leb(ubi, vol, lnum);
857 return ubi_eba_write_leb(ubi, vol, lnum, NULL, 0, 0, dtype);
860 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
864 mutex_lock(&ubi->alc_mutex);
865 err = leb_write_lock(ubi, vol_id, lnum);
869 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
870 vid_hdr->vol_id = cpu_to_be32(vol_id);
871 vid_hdr->lnum = cpu_to_be32(lnum);
872 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
873 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
875 crc = crc32(UBI_CRC32_INIT, buf, len);
876 vid_hdr->vol_type = UBI_VID_DYNAMIC;
877 vid_hdr->data_size = cpu_to_be32(len);
878 vid_hdr->copy_flag = 1;
879 vid_hdr->data_crc = cpu_to_be32(crc);
882 pnum = ubi_wl_get_peb(ubi, dtype);
888 dbg_eba("change LEB %d:%d, PEB %d, write VID hdr to PEB %d",
889 vol_id, lnum, vol->eba_tbl[lnum], pnum);
891 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
893 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
898 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
900 ubi_warn("failed to write %d bytes of data to PEB %d",
905 if (vol->eba_tbl[lnum] >= 0) {
906 err = ubi_wl_put_peb(ubi, vol->eba_tbl[lnum], 0);
911 vol->eba_tbl[lnum] = pnum;
914 leb_write_unlock(ubi, vol_id, lnum);
916 mutex_unlock(&ubi->alc_mutex);
917 ubi_free_vid_hdr(ubi, vid_hdr);
921 if (err != -EIO || !ubi->bad_allowed) {
923 * This flash device does not admit of bad eraseblocks or
924 * something nasty and unexpected happened. Switch to read-only
931 err = ubi_wl_put_peb(ubi, pnum, 1);
932 if (err || ++tries > UBI_IO_RETRIES) {
937 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
938 ubi_msg("try another PEB");
943 * ubi_eba_copy_leb - copy logical eraseblock.
944 * @ubi: UBI device description object
945 * @from: physical eraseblock number from where to copy
946 * @to: physical eraseblock number where to copy
947 * @vid_hdr: VID header of the @from physical eraseblock
949 * This function copies logical eraseblock from physical eraseblock @from to
950 * physical eraseblock @to. The @vid_hdr buffer may be changed by this
952 * o %0 in case of success;
953 * o %1 if the operation was canceled and should be tried later (e.g.,
954 * because a bit-flip was detected at the target PEB);
955 * o %2 if the volume is being deleted and this LEB should not be moved.
957 int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
958 struct ubi_vid_hdr *vid_hdr)
960 int err, vol_id, lnum, data_size, aldata_size, idx;
961 struct ubi_volume *vol;
964 vol_id = be32_to_cpu(vid_hdr->vol_id);
965 lnum = be32_to_cpu(vid_hdr->lnum);
967 dbg_eba("copy LEB %d:%d, PEB %d to PEB %d", vol_id, lnum, from, to);
969 if (vid_hdr->vol_type == UBI_VID_STATIC) {
970 data_size = be32_to_cpu(vid_hdr->data_size);
971 aldata_size = ALIGN(data_size, ubi->min_io_size);
973 data_size = aldata_size =
974 ubi->leb_size - be32_to_cpu(vid_hdr->data_pad);
976 idx = vol_id2idx(ubi, vol_id);
977 spin_lock(&ubi->volumes_lock);
979 * Note, we may race with volume deletion, which means that the volume
980 * this logical eraseblock belongs to might be being deleted. Since the
981 * volume deletion unmaps all the volume's logical eraseblocks, it will
982 * be locked in 'ubi_wl_put_peb()' and wait for the WL worker to finish.
984 vol = ubi->volumes[idx];
986 /* No need to do further work, cancel */
987 dbg_eba("volume %d is being removed, cancel", vol_id);
988 spin_unlock(&ubi->volumes_lock);
991 spin_unlock(&ubi->volumes_lock);
994 * We do not want anybody to write to this logical eraseblock while we
995 * are moving it, so lock it.
997 * Note, we are using non-waiting locking here, because we cannot sleep
998 * on the LEB, since it may cause deadlocks. Indeed, imagine a task is
999 * unmapping the LEB which is mapped to the PEB we are going to move
1000 * (@from). This task locks the LEB and goes sleep in the
1001 * 'ubi_wl_put_peb()' function on the @ubi->move_mutex. In turn, we are
1002 * holding @ubi->move_mutex and go sleep on the LEB lock. So, if the
1003 * LEB is already locked, we just do not move it and return %1.
1005 err = leb_write_trylock(ubi, vol_id, lnum);
1007 dbg_eba("contention on LEB %d:%d, cancel", vol_id, lnum);
1012 * The LEB might have been put meanwhile, and the task which put it is
1013 * probably waiting on @ubi->move_mutex. No need to continue the work,
1016 if (vol->eba_tbl[lnum] != from) {
1017 dbg_eba("LEB %d:%d is no longer mapped to PEB %d, mapped to "
1018 "PEB %d, cancel", vol_id, lnum, from,
1019 vol->eba_tbl[lnum]);
1021 goto out_unlock_leb;
1025 * OK, now the LEB is locked and we can safely start moving it. Since
1026 * this function utilizes thie @ubi->peb1_buf buffer which is shared
1027 * with some other functions, so lock the buffer by taking the
1030 mutex_lock(&ubi->buf_mutex);
1031 dbg_eba("read %d bytes of data", aldata_size);
1032 err = ubi_io_read_data(ubi, ubi->peb_buf1, from, 0, aldata_size);
1033 if (err && err != UBI_IO_BITFLIPS) {
1034 ubi_warn("error %d while reading data from PEB %d",
1036 goto out_unlock_buf;
1040 * Now we have got to calculate how much data we have to to copy. In
1041 * case of a static volume it is fairly easy - the VID header contains
1042 * the data size. In case of a dynamic volume it is more difficult - we
1043 * have to read the contents, cut 0xFF bytes from the end and copy only
1044 * the first part. We must do this to avoid writing 0xFF bytes as it
1045 * may have some side-effects. And not only this. It is important not
1046 * to include those 0xFFs to CRC because later the they may be filled
1049 if (vid_hdr->vol_type == UBI_VID_DYNAMIC)
1050 aldata_size = data_size =
1051 ubi_calc_data_len(ubi, ubi->peb_buf1, data_size);
1054 crc = crc32(UBI_CRC32_INIT, ubi->peb_buf1, data_size);
1058 * It may turn out to me that the whole @from physical eraseblock
1059 * contains only 0xFF bytes. Then we have to only write the VID header
1060 * and do not write any data. This also means we should not set
1061 * @vid_hdr->copy_flag, @vid_hdr->data_size, and @vid_hdr->data_crc.
1063 if (data_size > 0) {
1064 vid_hdr->copy_flag = 1;
1065 vid_hdr->data_size = cpu_to_be32(data_size);
1066 vid_hdr->data_crc = cpu_to_be32(crc);
1068 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
1070 err = ubi_io_write_vid_hdr(ubi, to, vid_hdr);
1072 goto out_unlock_buf;
1076 /* Read the VID header back and check if it was written correctly */
1077 err = ubi_io_read_vid_hdr(ubi, to, vid_hdr, 1);
1079 if (err != UBI_IO_BITFLIPS)
1080 ubi_warn("cannot read VID header back from PEB %d", to);
1083 goto out_unlock_buf;
1086 if (data_size > 0) {
1087 err = ubi_io_write_data(ubi, ubi->peb_buf1, to, 0, aldata_size);
1089 goto out_unlock_buf;
1094 * We've written the data and are going to read it back to make
1095 * sure it was written correctly.
1098 err = ubi_io_read_data(ubi, ubi->peb_buf2, to, 0, aldata_size);
1100 if (err != UBI_IO_BITFLIPS)
1101 ubi_warn("cannot read data back from PEB %d",
1105 goto out_unlock_buf;
1110 if (memcmp(ubi->peb_buf1, ubi->peb_buf2, aldata_size)) {
1111 ubi_warn("read data back from PEB %d - it is different",
1113 goto out_unlock_buf;
1117 ubi_assert(vol->eba_tbl[lnum] == from);
1118 vol->eba_tbl[lnum] = to;
1121 mutex_unlock(&ubi->buf_mutex);
1123 leb_write_unlock(ubi, vol_id, lnum);
1128 * ubi_eba_init_scan - initialize the EBA sub-system using scanning information.
1129 * @ubi: UBI device description object
1130 * @si: scanning information
1132 * This function returns zero in case of success and a negative error code in
1135 int ubi_eba_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
1137 int i, j, err, num_volumes;
1138 struct ubi_scan_volume *sv;
1139 struct ubi_volume *vol;
1140 struct ubi_scan_leb *seb;
1143 dbg_eba("initialize EBA sub-system");
1145 spin_lock_init(&ubi->ltree_lock);
1146 mutex_init(&ubi->alc_mutex);
1147 ubi->ltree = RB_ROOT;
1149 ubi->global_sqnum = si->max_sqnum + 1;
1150 num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
1152 for (i = 0; i < num_volumes; i++) {
1153 vol = ubi->volumes[i];
1159 vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int),
1161 if (!vol->eba_tbl) {
1166 for (j = 0; j < vol->reserved_pebs; j++)
1167 vol->eba_tbl[j] = UBI_LEB_UNMAPPED;
1169 sv = ubi_scan_find_sv(si, idx2vol_id(ubi, i));
1173 ubi_rb_for_each_entry(rb, seb, &sv->root, u.rb) {
1174 if (seb->lnum >= vol->reserved_pebs)
1176 * This may happen in case of an unclean reboot
1179 ubi_scan_move_to_list(sv, seb, &si->erase);
1180 vol->eba_tbl[seb->lnum] = seb->pnum;
1184 if (ubi->avail_pebs < EBA_RESERVED_PEBS) {
1185 ubi_err("no enough physical eraseblocks (%d, need %d)",
1186 ubi->avail_pebs, EBA_RESERVED_PEBS);
1190 ubi->avail_pebs -= EBA_RESERVED_PEBS;
1191 ubi->rsvd_pebs += EBA_RESERVED_PEBS;
1193 if (ubi->bad_allowed) {
1194 ubi_calculate_reserved(ubi);
1196 if (ubi->avail_pebs < ubi->beb_rsvd_level) {
1197 /* No enough free physical eraseblocks */
1198 ubi->beb_rsvd_pebs = ubi->avail_pebs;
1199 ubi_warn("cannot reserve enough PEBs for bad PEB "
1200 "handling, reserved %d, need %d",
1201 ubi->beb_rsvd_pebs, ubi->beb_rsvd_level);
1203 ubi->beb_rsvd_pebs = ubi->beb_rsvd_level;
1205 ubi->avail_pebs -= ubi->beb_rsvd_pebs;
1206 ubi->rsvd_pebs += ubi->beb_rsvd_pebs;
1209 dbg_eba("EBA sub-system is initialized");
1213 for (i = 0; i < num_volumes; i++) {
1214 if (!ubi->volumes[i])
1216 kfree(ubi->volumes[i]->eba_tbl);