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) unit.
24 * This unit 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 unit implements per-logical eraseblock locking. Before accessing a
31 * logical eraseblock it is locked for reading or writing. The per-logical
32 * eraseblock locking is implemented by means of the lock tree. The lock tree
33 * is an RB-tree which refers all the currently locked logical eraseblocks. The
34 * lock tree elements are &struct ubi_ltree_entry objects. They are indexed by
35 * (@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);
199 * leb_read_lock - lock logical eraseblock for reading.
200 * @ubi: UBI device description object
202 * @lnum: logical eraseblock number
204 * This function locks a logical eraseblock for reading. Returns zero in case
205 * of success and a negative error code in case of failure.
207 static int leb_read_lock(struct ubi_device *ubi, int vol_id, int lnum)
209 struct ubi_ltree_entry *le;
211 le = ltree_add_entry(ubi, vol_id, lnum);
214 down_read(&le->mutex);
219 * leb_read_unlock - unlock logical eraseblock.
220 * @ubi: UBI device description object
222 * @lnum: logical eraseblock number
224 static void leb_read_unlock(struct ubi_device *ubi, int vol_id, int lnum)
226 struct ubi_ltree_entry *le;
228 spin_lock(&ubi->ltree_lock);
229 le = ltree_lookup(ubi, vol_id, lnum);
231 ubi_assert(le->users >= 0);
233 if (le->users == 0) {
234 rb_erase(&le->rb, &ubi->ltree);
237 spin_unlock(&ubi->ltree_lock);
241 * leb_write_lock - lock logical eraseblock for writing.
242 * @ubi: UBI device description object
244 * @lnum: logical eraseblock number
246 * This function locks a logical eraseblock for writing. Returns zero in case
247 * of success and a negative error code in case of failure.
249 static int leb_write_lock(struct ubi_device *ubi, int vol_id, int lnum)
251 struct ubi_ltree_entry *le;
253 le = ltree_add_entry(ubi, vol_id, lnum);
256 down_write(&le->mutex);
261 * leb_write_lock - lock logical eraseblock for writing.
262 * @ubi: UBI device description object
264 * @lnum: logical eraseblock number
266 * This function locks a logical eraseblock for writing if there is no
267 * contention and does nothing if there is contention. Returns %0 in case of
268 * success, %1 in case of contention, and and a negative error code in case of
271 static int leb_write_trylock(struct ubi_device *ubi, int vol_id, int lnum)
273 struct ubi_ltree_entry *le;
275 le = ltree_add_entry(ubi, vol_id, lnum);
278 if (down_write_trylock(&le->mutex))
281 /* Contention, cancel */
282 spin_lock(&ubi->ltree_lock);
284 ubi_assert(le->users >= 0);
285 if (le->users == 0) {
286 rb_erase(&le->rb, &ubi->ltree);
289 spin_unlock(&ubi->ltree_lock);
295 * leb_write_unlock - unlock logical eraseblock.
296 * @ubi: UBI device description object
298 * @lnum: logical eraseblock number
300 static void leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
302 struct ubi_ltree_entry *le;
304 spin_lock(&ubi->ltree_lock);
305 le = ltree_lookup(ubi, vol_id, lnum);
307 ubi_assert(le->users >= 0);
308 up_write(&le->mutex);
309 if (le->users == 0) {
310 rb_erase(&le->rb, &ubi->ltree);
313 spin_unlock(&ubi->ltree_lock);
317 * ubi_eba_unmap_leb - un-map logical eraseblock.
318 * @ubi: UBI device description object
319 * @vol: volume description object
320 * @lnum: logical eraseblock number
322 * This function un-maps logical eraseblock @lnum and schedules corresponding
323 * physical eraseblock for erasure. Returns zero in case of success and a
324 * negative error code in case of failure.
326 int ubi_eba_unmap_leb(struct ubi_device *ubi, struct ubi_volume *vol,
329 int err, pnum, vol_id = vol->vol_id;
334 err = leb_write_lock(ubi, vol_id, lnum);
338 pnum = vol->eba_tbl[lnum];
340 /* This logical eraseblock is already unmapped */
343 dbg_eba("erase LEB %d:%d, PEB %d", vol_id, lnum, pnum);
345 vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED;
346 err = ubi_wl_put_peb(ubi, pnum, 0);
349 leb_write_unlock(ubi, vol_id, lnum);
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
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.
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.
372 int ubi_eba_read_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
373 void *buf, int offset, int len, int check)
375 int err, pnum, scrub = 0, vol_id = vol->vol_id;
376 struct ubi_vid_hdr *vid_hdr;
377 uint32_t uninitialized_var(crc);
379 err = leb_read_lock(ubi, vol_id, lnum);
383 pnum = vol->eba_tbl[lnum];
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.
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);
398 dbg_eba("read %d bytes from offset %d of LEB %d:%d, PEB %d",
399 len, offset, vol_id, lnum, pnum);
401 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
406 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
412 err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
413 if (err && err != UBI_IO_BITFLIPS) {
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
423 if (err == UBI_IO_BAD_VID_HDR) {
424 ubi_warn("bad VID header at PEB %d, LEB"
425 "%d:%d", pnum, vol_id, lnum);
431 } else if (err == UBI_IO_BITFLIPS)
434 ubi_assert(lnum < be32_to_cpu(vid_hdr->used_ebs));
435 ubi_assert(len == be32_to_cpu(vid_hdr->data_size));
437 crc = be32_to_cpu(vid_hdr->data_crc);
438 ubi_free_vid_hdr(ubi, vid_hdr);
441 err = ubi_io_read_data(ubi, buf, pnum, offset, len);
443 if (err == UBI_IO_BITFLIPS) {
446 } else if (err == -EBADMSG) {
447 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
451 ubi_msg("force data checking");
460 uint32_t crc1 = crc32(UBI_CRC32_INIT, buf, len);
462 ubi_warn("CRC error: calculated %#08x, must be %#08x",
470 err = ubi_wl_scrub_peb(ubi, pnum);
472 leb_read_unlock(ubi, vol_id, lnum);
476 ubi_free_vid_hdr(ubi, vid_hdr);
478 leb_read_unlock(ubi, vol_id, lnum);
483 * recover_peb - recover from write failure.
484 * @ubi: UBI device description object
485 * @pnum: the physical eraseblock to recover
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
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.
498 static int recover_peb(struct ubi_device *ubi, int pnum, int vol_id, int lnum,
499 const void *buf, int offset, int len)
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;
505 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
510 mutex_lock(&ubi->buf_mutex);
513 new_pnum = ubi_wl_get_peb(ubi, UBI_UNKNOWN);
515 mutex_unlock(&ubi->buf_mutex);
516 ubi_free_vid_hdr(ubi, vid_hdr);
520 ubi_msg("recover PEB %d, move data to PEB %d", pnum, new_pnum);
522 err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
523 if (err && err != UBI_IO_BITFLIPS) {
529 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
530 err = ubi_io_write_vid_hdr(ubi, new_pnum, vid_hdr);
534 data_size = offset + len;
535 memset(ubi->peb_buf1 + offset, 0xFF, len);
537 /* Read everything before the area where the write failure happened */
539 err = ubi_io_read_data(ubi, ubi->peb_buf1, pnum, 0, offset);
540 if (err && err != UBI_IO_BITFLIPS)
544 memcpy(ubi->peb_buf1 + offset, buf, len);
546 err = ubi_io_write_data(ubi, ubi->peb_buf1, new_pnum, 0, data_size);
550 mutex_unlock(&ubi->buf_mutex);
551 ubi_free_vid_hdr(ubi, vid_hdr);
553 vol->eba_tbl[lnum] = new_pnum;
554 ubi_wl_put_peb(ubi, pnum, 1);
556 ubi_msg("data was successfully recovered");
560 mutex_unlock(&ubi->buf_mutex);
561 ubi_wl_put_peb(ubi, new_pnum, 1);
562 ubi_free_vid_hdr(ubi, vid_hdr);
567 * Bad luck? This physical eraseblock is bad too? Crud. Let's try to
570 ubi_warn("failed to write to PEB %d", new_pnum);
571 ubi_wl_put_peb(ubi, new_pnum, 1);
572 if (++tries > UBI_IO_RETRIES) {
573 mutex_unlock(&ubi->buf_mutex);
574 ubi_free_vid_hdr(ubi, vid_hdr);
577 ubi_msg("try again");
582 * ubi_eba_write_leb - write data to dynamic volume.
583 * @ubi: UBI device description object
584 * @vol: volume description object
585 * @lnum: logical eraseblock number
586 * @buf: the data to write
587 * @offset: offset within the logical eraseblock where to write
588 * @len: how many bytes to write
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.
596 int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
597 const void *buf, int offset, int len, int dtype)
599 int err, pnum, tries = 0, vol_id = vol->vol_id;
600 struct ubi_vid_hdr *vid_hdr;
605 err = leb_write_lock(ubi, vol_id, lnum);
609 pnum = vol->eba_tbl[lnum];
611 dbg_eba("write %d bytes at offset %d of LEB %d:%d, PEB %d",
612 len, offset, vol_id, lnum, pnum);
614 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
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,
623 leb_write_unlock(ubi, vol_id, lnum);
628 * The logical eraseblock is not mapped. We have to get a free physical
629 * eraseblock and write the volume identifier header there first.
631 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
633 leb_write_unlock(ubi, vol_id, lnum);
637 vid_hdr->vol_type = UBI_VID_DYNAMIC;
638 vid_hdr->sqnum = cpu_to_be64(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);
645 pnum = ubi_wl_get_peb(ubi, dtype);
647 ubi_free_vid_hdr(ubi, vid_hdr);
648 leb_write_unlock(ubi, vol_id, lnum);
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);
655 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
657 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
663 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
665 ubi_warn("failed to write %d bytes at offset %d of "
666 "LEB %d:%d, PEB %d", len, offset, vol_id,
672 vol->eba_tbl[lnum] = pnum;
674 leb_write_unlock(ubi, vol_id, lnum);
675 ubi_free_vid_hdr(ubi, vid_hdr);
679 if (err != -EIO || !ubi->bad_allowed) {
681 leb_write_unlock(ubi, vol_id, lnum);
682 ubi_free_vid_hdr(ubi, vid_hdr);
687 * Fortunately, this is the first write operation to this physical
688 * eraseblock, so just put it and request a new one. We assume that if
689 * this physical eraseblock went bad, the erase code will handle that.
691 err = ubi_wl_put_peb(ubi, pnum, 1);
692 if (err || ++tries > UBI_IO_RETRIES) {
694 leb_write_unlock(ubi, vol_id, lnum);
695 ubi_free_vid_hdr(ubi, vid_hdr);
699 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
700 ubi_msg("try another PEB");
705 * ubi_eba_write_leb_st - write data to static volume.
706 * @ubi: UBI device description object
707 * @vol: volume description object
708 * @lnum: logical eraseblock number
709 * @buf: data to write
710 * @len: how many bytes to write
712 * @used_ebs: how many logical eraseblocks will this volume contain
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.
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.
723 * It is prohibited to write more then 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.
727 int ubi_eba_write_leb_st(struct ubi_device *ubi, struct ubi_volume *vol,
728 int lnum, const void *buf, int len, int dtype,
731 int err, pnum, tries = 0, data_size = len, vol_id = vol->vol_id;
732 struct ubi_vid_hdr *vid_hdr;
738 if (lnum == used_ebs - 1)
739 /* If this is the last LEB @len may be unaligned */
740 len = ALIGN(data_size, ubi->min_io_size);
742 ubi_assert(!(len & (ubi->min_io_size - 1)));
744 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
748 err = leb_write_lock(ubi, vol_id, lnum);
750 ubi_free_vid_hdr(ubi, vid_hdr);
754 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
755 vid_hdr->vol_id = cpu_to_be32(vol_id);
756 vid_hdr->lnum = cpu_to_be32(lnum);
757 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
758 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
760 crc = crc32(UBI_CRC32_INIT, buf, data_size);
761 vid_hdr->vol_type = UBI_VID_STATIC;
762 vid_hdr->data_size = cpu_to_be32(data_size);
763 vid_hdr->used_ebs = cpu_to_be32(used_ebs);
764 vid_hdr->data_crc = cpu_to_be32(crc);
767 pnum = ubi_wl_get_peb(ubi, dtype);
769 ubi_free_vid_hdr(ubi, vid_hdr);
770 leb_write_unlock(ubi, vol_id, lnum);
774 dbg_eba("write VID hdr and %d bytes at LEB %d:%d, PEB %d, used_ebs %d",
775 len, vol_id, lnum, pnum, used_ebs);
777 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
779 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
784 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
786 ubi_warn("failed to write %d bytes of data to PEB %d",
791 ubi_assert(vol->eba_tbl[lnum] < 0);
792 vol->eba_tbl[lnum] = pnum;
794 leb_write_unlock(ubi, vol_id, lnum);
795 ubi_free_vid_hdr(ubi, vid_hdr);
799 if (err != -EIO || !ubi->bad_allowed) {
801 * This flash device does not admit of bad eraseblocks or
802 * something nasty and unexpected happened. Switch to read-only
806 leb_write_unlock(ubi, vol_id, lnum);
807 ubi_free_vid_hdr(ubi, vid_hdr);
811 err = ubi_wl_put_peb(ubi, pnum, 1);
812 if (err || ++tries > UBI_IO_RETRIES) {
814 leb_write_unlock(ubi, vol_id, lnum);
815 ubi_free_vid_hdr(ubi, vid_hdr);
819 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
820 ubi_msg("try another PEB");
825 * ubi_eba_atomic_leb_change - change logical eraseblock atomically.
826 * @ubi: UBI device description object
827 * @vol: volume description object
828 * @lnum: logical eraseblock number
829 * @buf: data to write
830 * @len: how many bytes to write
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.
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.
842 int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
843 int lnum, const void *buf, int len, int dtype)
845 int err, pnum, tries = 0, vol_id = vol->vol_id;
846 struct ubi_vid_hdr *vid_hdr;
854 * Special case when data length is zero. In this case the LEB
855 * has to be unmapped and mapped somewhere else.
857 err = ubi_eba_unmap_leb(ubi, vol, lnum);
860 return ubi_eba_write_leb(ubi, vol, lnum, NULL, 0, 0, dtype);
863 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
867 mutex_lock(&ubi->alc_mutex);
868 err = leb_write_lock(ubi, vol_id, lnum);
872 vid_hdr->sqnum = cpu_to_be64(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);
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);
885 pnum = ubi_wl_get_peb(ubi, dtype);
891 dbg_eba("change LEB %d:%d, PEB %d, write VID hdr to PEB %d",
892 vol_id, lnum, vol->eba_tbl[lnum], pnum);
894 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
896 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
901 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
903 ubi_warn("failed to write %d bytes of data to PEB %d",
908 if (vol->eba_tbl[lnum] >= 0) {
909 err = ubi_wl_put_peb(ubi, vol->eba_tbl[lnum], 1);
914 vol->eba_tbl[lnum] = pnum;
917 leb_write_unlock(ubi, vol_id, lnum);
919 mutex_unlock(&ubi->alc_mutex);
920 ubi_free_vid_hdr(ubi, vid_hdr);
924 if (err != -EIO || !ubi->bad_allowed) {
926 * This flash device does not admit of bad eraseblocks or
927 * something nasty and unexpected happened. Switch to read-only
934 err = ubi_wl_put_peb(ubi, pnum, 1);
935 if (err || ++tries > UBI_IO_RETRIES) {
940 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
941 ubi_msg("try another PEB");
946 * ubi_eba_copy_leb - copy logical eraseblock.
947 * @ubi: UBI device description object
948 * @from: physical eraseblock number from where to copy
949 * @to: physical eraseblock number where to copy
950 * @vid_hdr: VID header of the @from physical eraseblock
952 * This function copies logical eraseblock from physical eraseblock @from to
953 * physical eraseblock @to. The @vid_hdr buffer may be changed by this
955 * o %0 in case of success;
956 * o %1 if the operation was canceled and should be tried later (e.g.,
957 * because a bit-flip was detected at the target PEB);
958 * o %2 if the volume is being deleted and this LEB should not be moved.
960 int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
961 struct ubi_vid_hdr *vid_hdr)
963 int err, vol_id, lnum, data_size, aldata_size, idx;
964 struct ubi_volume *vol;
967 vol_id = be32_to_cpu(vid_hdr->vol_id);
968 lnum = be32_to_cpu(vid_hdr->lnum);
970 dbg_eba("copy LEB %d:%d, PEB %d to PEB %d", vol_id, lnum, from, to);
972 if (vid_hdr->vol_type == UBI_VID_STATIC) {
973 data_size = be32_to_cpu(vid_hdr->data_size);
974 aldata_size = ALIGN(data_size, ubi->min_io_size);
976 data_size = aldata_size =
977 ubi->leb_size - be32_to_cpu(vid_hdr->data_pad);
979 idx = vol_id2idx(ubi, vol_id);
980 spin_lock(&ubi->volumes_lock);
982 * Note, we may race with volume deletion, which means that the volume
983 * this logical eraseblock belongs to might be being deleted. Since the
984 * volume deletion unmaps all the volume's logical eraseblocks, it will
985 * be locked in 'ubi_wl_put_peb()' and wait for the WL worker to finish.
987 vol = ubi->volumes[idx];
989 /* No need to do further work, cancel */
990 dbg_eba("volume %d is being removed, cancel", vol_id);
991 spin_unlock(&ubi->volumes_lock);
994 spin_unlock(&ubi->volumes_lock);
997 * We do not want anybody to write to this logical eraseblock while we
998 * are moving it, so lock it.
1000 * Note, we are using non-waiting locking here, because we cannot sleep
1001 * on the LEB, since it may cause deadlocks. Indeed, imagine a task is
1002 * unmapping the LEB which is mapped to the PEB we are going to move
1003 * (@from). This task locks the LEB and goes sleep in the
1004 * 'ubi_wl_put_peb()' function on the @ubi->move_mutex. In turn, we are
1005 * holding @ubi->move_mutex and go sleep on the LEB lock. So, if the
1006 * LEB is already locked, we just do not move it and return %1.
1008 err = leb_write_trylock(ubi, vol_id, lnum);
1010 dbg_eba("contention on LEB %d:%d, cancel", vol_id, lnum);
1015 * The LEB might have been put meanwhile, and the task which put it is
1016 * probably waiting on @ubi->move_mutex. No need to continue the work,
1019 if (vol->eba_tbl[lnum] != from) {
1020 dbg_eba("LEB %d:%d is no longer mapped to PEB %d, mapped to "
1021 "PEB %d, cancel", vol_id, lnum, from,
1022 vol->eba_tbl[lnum]);
1024 goto out_unlock_leb;
1028 * OK, now the LEB is locked and we can safely start moving iy. Since
1029 * this function utilizes thie @ubi->peb1_buf buffer which is shared
1030 * with some other functions, so lock the buffer by taking the
1033 mutex_lock(&ubi->buf_mutex);
1034 dbg_eba("read %d bytes of data", aldata_size);
1035 err = ubi_io_read_data(ubi, ubi->peb_buf1, from, 0, aldata_size);
1036 if (err && err != UBI_IO_BITFLIPS) {
1037 ubi_warn("error %d while reading data from PEB %d",
1039 goto out_unlock_buf;
1043 * Now we have got to calculate how much data we have to to copy. In
1044 * case of a static volume it is fairly easy - the VID header contains
1045 * the data size. In case of a dynamic volume it is more difficult - we
1046 * have to read the contents, cut 0xFF bytes from the end and copy only
1047 * the first part. We must do this to avoid writing 0xFF bytes as it
1048 * may have some side-effects. And not only this. It is important not
1049 * to include those 0xFFs to CRC because later the they may be filled
1052 if (vid_hdr->vol_type == UBI_VID_DYNAMIC)
1053 aldata_size = data_size =
1054 ubi_calc_data_len(ubi, ubi->peb_buf1, data_size);
1057 crc = crc32(UBI_CRC32_INIT, ubi->peb_buf1, data_size);
1061 * It may turn out to me that the whole @from physical eraseblock
1062 * contains only 0xFF bytes. Then we have to only write the VID header
1063 * and do not write any data. This also means we should not set
1064 * @vid_hdr->copy_flag, @vid_hdr->data_size, and @vid_hdr->data_crc.
1066 if (data_size > 0) {
1067 vid_hdr->copy_flag = 1;
1068 vid_hdr->data_size = cpu_to_be32(data_size);
1069 vid_hdr->data_crc = cpu_to_be32(crc);
1071 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
1073 err = ubi_io_write_vid_hdr(ubi, to, vid_hdr);
1075 goto out_unlock_buf;
1079 /* Read the VID header back and check if it was written correctly */
1080 err = ubi_io_read_vid_hdr(ubi, to, vid_hdr, 1);
1082 if (err != UBI_IO_BITFLIPS)
1083 ubi_warn("cannot read VID header back from PEB %d", to);
1086 goto out_unlock_buf;
1089 if (data_size > 0) {
1090 err = ubi_io_write_data(ubi, ubi->peb_buf1, to, 0, aldata_size);
1092 goto out_unlock_buf;
1097 * We've written the data and are going to read it back to make
1098 * sure it was written correctly.
1101 err = ubi_io_read_data(ubi, ubi->peb_buf2, to, 0, aldata_size);
1103 if (err != UBI_IO_BITFLIPS)
1104 ubi_warn("cannot read data back from PEB %d",
1108 goto out_unlock_buf;
1113 if (memcmp(ubi->peb_buf1, ubi->peb_buf2, aldata_size)) {
1114 ubi_warn("read data back from PEB %d - it is different",
1116 goto out_unlock_buf;
1120 ubi_assert(vol->eba_tbl[lnum] == from);
1121 vol->eba_tbl[lnum] = to;
1124 mutex_unlock(&ubi->buf_mutex);
1126 leb_write_unlock(ubi, vol_id, lnum);
1131 * ubi_eba_init_scan - initialize the EBA unit using scanning information.
1132 * @ubi: UBI device description object
1133 * @si: scanning information
1135 * This function returns zero in case of success and a negative error code in
1138 int ubi_eba_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
1140 int i, j, err, num_volumes;
1141 struct ubi_scan_volume *sv;
1142 struct ubi_volume *vol;
1143 struct ubi_scan_leb *seb;
1146 dbg_eba("initialize EBA unit");
1148 spin_lock_init(&ubi->ltree_lock);
1149 mutex_init(&ubi->alc_mutex);
1150 ubi->ltree = RB_ROOT;
1152 ubi->global_sqnum = si->max_sqnum + 1;
1153 num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
1155 for (i = 0; i < num_volumes; i++) {
1156 vol = ubi->volumes[i];
1162 vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int),
1164 if (!vol->eba_tbl) {
1169 for (j = 0; j < vol->reserved_pebs; j++)
1170 vol->eba_tbl[j] = UBI_LEB_UNMAPPED;
1172 sv = ubi_scan_find_sv(si, idx2vol_id(ubi, i));
1176 ubi_rb_for_each_entry(rb, seb, &sv->root, u.rb) {
1177 if (seb->lnum >= vol->reserved_pebs)
1179 * This may happen in case of an unclean reboot
1182 ubi_scan_move_to_list(sv, seb, &si->erase);
1183 vol->eba_tbl[seb->lnum] = seb->pnum;
1187 if (ubi->avail_pebs < EBA_RESERVED_PEBS) {
1188 ubi_err("no enough physical eraseblocks (%d, need %d)",
1189 ubi->avail_pebs, EBA_RESERVED_PEBS);
1193 ubi->avail_pebs -= EBA_RESERVED_PEBS;
1194 ubi->rsvd_pebs += EBA_RESERVED_PEBS;
1196 if (ubi->bad_allowed) {
1197 ubi_calculate_reserved(ubi);
1199 if (ubi->avail_pebs < ubi->beb_rsvd_level) {
1200 /* No enough free physical eraseblocks */
1201 ubi->beb_rsvd_pebs = ubi->avail_pebs;
1202 ubi_warn("cannot reserve enough PEBs for bad PEB "
1203 "handling, reserved %d, need %d",
1204 ubi->beb_rsvd_pebs, ubi->beb_rsvd_level);
1206 ubi->beb_rsvd_pebs = ubi->beb_rsvd_level;
1208 ubi->avail_pebs -= ubi->beb_rsvd_pebs;
1209 ubi->rsvd_pebs += ubi->beb_rsvd_pebs;
1212 dbg_eba("EBA unit is initialized");
1216 for (i = 0; i < num_volumes; i++) {
1217 if (!ubi->volumes[i])
1219 kfree(ubi->volumes[i]->eba_tbl);