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 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 * struct ltree_entry - an entry in the lock tree.
54 * @rb: links RB-tree nodes
55 * @vol_id: volume ID of the locked logical eraseblock
56 * @lnum: locked logical eraseblock number
57 * @users: how many tasks are using this logical eraseblock or wait for it
58 * @mutex: read/write mutex to implement read/write access serialization to
59 * the (@vol_id, @lnum) logical eraseblock
61 * When a logical eraseblock is being locked - corresponding &struct ltree_entry
62 * object is inserted to the lock tree (@ubi->ltree).
69 struct rw_semaphore mutex;
72 /* Slab cache for lock-tree entries */
73 static struct kmem_cache *ltree_slab;
76 * next_sqnum - get next sequence number.
77 * @ubi: UBI device description object
79 * This function returns next sequence number to use, which is just the current
80 * global sequence counter value. It also increases the global sequence
83 static unsigned long long next_sqnum(struct ubi_device *ubi)
85 unsigned long long sqnum;
87 spin_lock(&ubi->ltree_lock);
88 sqnum = ubi->global_sqnum++;
89 spin_unlock(&ubi->ltree_lock);
95 * ubi_get_compat - get compatibility flags of a volume.
96 * @ubi: UBI device description object
99 * This function returns compatibility flags for an internal volume. User
100 * volumes have no compatibility flags, so %0 is returned.
102 static int ubi_get_compat(const struct ubi_device *ubi, int vol_id)
104 if (vol_id == UBI_LAYOUT_VOL_ID)
105 return UBI_LAYOUT_VOLUME_COMPAT;
110 * ltree_lookup - look up the lock tree.
111 * @ubi: UBI device description object
113 * @lnum: logical eraseblock number
115 * This function returns a pointer to the corresponding &struct ltree_entry
116 * object if the logical eraseblock is locked and %NULL if it is not.
117 * @ubi->ltree_lock has to be locked.
119 static struct ltree_entry *ltree_lookup(struct ubi_device *ubi, int vol_id,
124 p = ubi->ltree.rb_node;
126 struct ltree_entry *le;
128 le = rb_entry(p, struct ltree_entry, rb);
130 if (vol_id < le->vol_id)
132 else if (vol_id > le->vol_id)
137 else if (lnum > le->lnum)
148 * ltree_add_entry - add new entry to the lock tree.
149 * @ubi: UBI device description object
151 * @lnum: logical eraseblock number
153 * This function adds new entry for logical eraseblock (@vol_id, @lnum) to the
154 * lock tree. If such entry is already there, its usage counter is increased.
155 * Returns pointer to the lock tree entry or %-ENOMEM if memory allocation
158 static struct ltree_entry *ltree_add_entry(struct ubi_device *ubi, int vol_id,
161 struct ltree_entry *le, *le1, *le_free;
163 le = kmem_cache_alloc(ltree_slab, GFP_NOFS);
165 return ERR_PTR(-ENOMEM);
170 spin_lock(&ubi->ltree_lock);
171 le1 = ltree_lookup(ubi, vol_id, lnum);
175 * This logical eraseblock is already locked. The newly
176 * allocated lock entry is not needed.
181 struct rb_node **p, *parent = NULL;
184 * No lock entry, add the newly allocated one to the
185 * @ubi->ltree RB-tree.
189 p = &ubi->ltree.rb_node;
192 le1 = rb_entry(parent, struct ltree_entry, rb);
194 if (vol_id < le1->vol_id)
196 else if (vol_id > le1->vol_id)
199 ubi_assert(lnum != le1->lnum);
200 if (lnum < le1->lnum)
207 rb_link_node(&le->rb, parent, p);
208 rb_insert_color(&le->rb, &ubi->ltree);
211 spin_unlock(&ubi->ltree_lock);
214 kmem_cache_free(ltree_slab, le_free);
220 * leb_read_lock - lock logical eraseblock for reading.
221 * @ubi: UBI device description object
223 * @lnum: logical eraseblock number
225 * This function locks a logical eraseblock for reading. Returns zero in case
226 * of success and a negative error code in case of failure.
228 static int leb_read_lock(struct ubi_device *ubi, int vol_id, int lnum)
230 struct ltree_entry *le;
232 le = ltree_add_entry(ubi, vol_id, lnum);
235 down_read(&le->mutex);
240 * leb_read_unlock - unlock logical eraseblock.
241 * @ubi: UBI device description object
243 * @lnum: logical eraseblock number
245 static void leb_read_unlock(struct ubi_device *ubi, int vol_id, int lnum)
248 struct ltree_entry *le;
250 spin_lock(&ubi->ltree_lock);
251 le = ltree_lookup(ubi, vol_id, lnum);
253 ubi_assert(le->users >= 0);
254 if (le->users == 0) {
255 rb_erase(&le->rb, &ubi->ltree);
258 spin_unlock(&ubi->ltree_lock);
262 kmem_cache_free(ltree_slab, le);
266 * leb_write_lock - lock logical eraseblock for writing.
267 * @ubi: UBI device description object
269 * @lnum: logical eraseblock number
271 * This function locks a logical eraseblock for writing. Returns zero in case
272 * of success and a negative error code in case of failure.
274 static int leb_write_lock(struct ubi_device *ubi, int vol_id, int lnum)
276 struct ltree_entry *le;
278 le = ltree_add_entry(ubi, vol_id, lnum);
281 down_write(&le->mutex);
286 * leb_write_unlock - unlock logical eraseblock.
287 * @ubi: UBI device description object
289 * @lnum: logical eraseblock number
291 static void leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
294 struct ltree_entry *le;
296 spin_lock(&ubi->ltree_lock);
297 le = ltree_lookup(ubi, vol_id, lnum);
299 ubi_assert(le->users >= 0);
300 if (le->users == 0) {
301 rb_erase(&le->rb, &ubi->ltree);
305 spin_unlock(&ubi->ltree_lock);
307 up_write(&le->mutex);
309 kmem_cache_free(ltree_slab, le);
313 * ubi_eba_unmap_leb - un-map logical eraseblock.
314 * @ubi: UBI device description object
316 * @lnum: logical eraseblock number
318 * This function un-maps logical eraseblock @lnum and schedules corresponding
319 * physical eraseblock for erasure. Returns zero in case of success and a
320 * negative error code in case of failure.
322 int ubi_eba_unmap_leb(struct ubi_device *ubi, int vol_id, int lnum)
324 int idx = vol_id2idx(ubi, vol_id), err, pnum;
325 struct ubi_volume *vol = ubi->volumes[idx];
330 err = leb_write_lock(ubi, vol_id, lnum);
334 pnum = vol->eba_tbl[lnum];
336 /* This logical eraseblock is already unmapped */
339 dbg_eba("erase LEB %d:%d, PEB %d", vol_id, lnum, pnum);
341 vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED;
342 err = ubi_wl_put_peb(ubi, pnum, 0);
345 leb_write_unlock(ubi, vol_id, lnum);
350 * ubi_eba_read_leb - read data.
351 * @ubi: UBI device description object
353 * @lnum: logical eraseblock number
354 * @buf: buffer to store the read data
355 * @offset: offset from where to read
356 * @len: how many bytes to read
357 * @check: data CRC check flag
359 * If the logical eraseblock @lnum is unmapped, @buf is filled with 0xFF
360 * bytes. The @check flag only makes sense for static volumes and forces
361 * eraseblock data CRC checking.
363 * In case of success this function returns zero. In case of a static volume,
364 * if data CRC mismatches - %-EBADMSG is returned. %-EBADMSG may also be
365 * returned for any volume type if an ECC error was detected by the MTD device
366 * driver. Other negative error cored may be returned in case of other errors.
368 int ubi_eba_read_leb(struct ubi_device *ubi, int vol_id, int lnum, void *buf,
369 int offset, int len, int check)
371 int err, pnum, scrub = 0, idx = vol_id2idx(ubi, vol_id);
372 struct ubi_vid_hdr *vid_hdr;
373 struct ubi_volume *vol = ubi->volumes[idx];
374 uint32_t uninitialized_var(crc);
376 err = leb_read_lock(ubi, vol_id, lnum);
380 pnum = vol->eba_tbl[lnum];
383 * The logical eraseblock is not mapped, fill the whole buffer
384 * with 0xFF bytes. The exception is static volumes for which
385 * it is an error to read unmapped logical eraseblocks.
387 dbg_eba("read %d bytes from offset %d of LEB %d:%d (unmapped)",
388 len, offset, vol_id, lnum);
389 leb_read_unlock(ubi, vol_id, lnum);
390 ubi_assert(vol->vol_type != UBI_STATIC_VOLUME);
391 memset(buf, 0xFF, len);
395 dbg_eba("read %d bytes from offset %d of LEB %d:%d, PEB %d",
396 len, offset, vol_id, lnum, pnum);
398 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
403 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
409 err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
410 if (err && err != UBI_IO_BITFLIPS) {
413 * The header is either absent or corrupted.
414 * The former case means there is a bug -
415 * switch to read-only mode just in case.
416 * The latter case means a real corruption - we
417 * may try to recover data. FIXME: but this is
420 if (err == UBI_IO_BAD_VID_HDR) {
421 ubi_warn("bad VID header at PEB %d, LEB"
422 "%d:%d", pnum, vol_id, lnum);
428 } else if (err == UBI_IO_BITFLIPS)
431 ubi_assert(lnum < be32_to_cpu(vid_hdr->used_ebs));
432 ubi_assert(len == be32_to_cpu(vid_hdr->data_size));
434 crc = be32_to_cpu(vid_hdr->data_crc);
435 ubi_free_vid_hdr(ubi, vid_hdr);
438 err = ubi_io_read_data(ubi, buf, pnum, offset, len);
440 if (err == UBI_IO_BITFLIPS) {
443 } else if (err == -EBADMSG) {
444 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
448 ubi_msg("force data checking");
457 uint32_t crc1 = crc32(UBI_CRC32_INIT, buf, len);
459 ubi_warn("CRC error: calculated %#08x, must be %#08x",
467 err = ubi_wl_scrub_peb(ubi, pnum);
469 leb_read_unlock(ubi, vol_id, lnum);
473 ubi_free_vid_hdr(ubi, vid_hdr);
475 leb_read_unlock(ubi, vol_id, lnum);
480 * recover_peb - recover from write failure.
481 * @ubi: UBI device description object
482 * @pnum: the physical eraseblock to recover
484 * @lnum: logical eraseblock number
485 * @buf: data which was not written because of the write failure
486 * @offset: offset of the failed write
487 * @len: how many bytes should have been written
489 * This function is called in case of a write failure and moves all good data
490 * from the potentially bad physical eraseblock to a good physical eraseblock.
491 * This function also writes the data which was not written due to the failure.
492 * Returns new physical eraseblock number in case of success, and a negative
493 * error code in case of failure.
495 static int recover_peb(struct ubi_device *ubi, int pnum, int vol_id, int lnum,
496 const void *buf, int offset, int len)
498 int err, idx = vol_id2idx(ubi, vol_id), new_pnum, data_size, tries = 0;
499 struct ubi_volume *vol = ubi->volumes[idx];
500 struct ubi_vid_hdr *vid_hdr;
502 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
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_id. 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, int vol_id, int lnum,
594 const void *buf, int offset, int len, int dtype)
596 int idx = vol_id2idx(ubi, vol_id), err, pnum, tries = 0;
597 struct ubi_volume *vol = ubi->volumes[idx];
598 struct ubi_vid_hdr *vid_hdr;
603 err = leb_write_lock(ubi, vol_id, lnum);
607 pnum = vol->eba_tbl[lnum];
609 dbg_eba("write %d bytes at offset %d of LEB %d:%d, PEB %d",
610 len, offset, vol_id, lnum, pnum);
612 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
614 ubi_warn("failed to write data to PEB %d", pnum);
615 if (err == -EIO && ubi->bad_allowed)
616 err = recover_peb(ubi, pnum, vol_id, lnum, buf, offset, len);
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",
659 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
661 ubi_warn("failed to write %d bytes at offset %d of LEB %d:%d, "
662 "PEB %d", len, offset, vol_id, lnum, pnum);
666 vol->eba_tbl[lnum] = pnum;
668 leb_write_unlock(ubi, vol_id, lnum);
669 ubi_free_vid_hdr(ubi, vid_hdr);
673 if (err != -EIO || !ubi->bad_allowed) {
675 leb_write_unlock(ubi, vol_id, lnum);
676 ubi_free_vid_hdr(ubi, vid_hdr);
681 * Fortunately, this is the first write operation to this physical
682 * eraseblock, so just put it and request a new one. We assume that if
683 * this physical eraseblock went bad, the erase code will handle that.
685 err = ubi_wl_put_peb(ubi, pnum, 1);
686 if (err || ++tries > UBI_IO_RETRIES) {
688 leb_write_unlock(ubi, vol_id, lnum);
689 ubi_free_vid_hdr(ubi, vid_hdr);
693 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
694 ubi_msg("try another PEB");
699 * ubi_eba_write_leb_st - write data to static volume.
700 * @ubi: UBI device description object
702 * @lnum: logical eraseblock number
703 * @buf: data to write
704 * @len: how many bytes to write
706 * @used_ebs: how many logical eraseblocks will this volume contain
708 * This function writes data to logical eraseblock @lnum of static volume
709 * @vol_id. The @used_ebs argument should contain total number of logical
710 * eraseblock in this static volume.
712 * When writing to the last logical eraseblock, the @len argument doesn't have
713 * to be aligned to the minimal I/O unit size. Instead, it has to be equivalent
714 * to the real data size, although the @buf buffer has to contain the
715 * alignment. In all other cases, @len has to be aligned.
717 * It is prohibited to write more then once to logical eraseblocks of static
718 * volumes. This function returns zero in case of success and a negative error
719 * code in case of failure.
721 int ubi_eba_write_leb_st(struct ubi_device *ubi, int vol_id, int lnum,
722 const void *buf, int len, int dtype, int used_ebs)
724 int err, pnum, tries = 0, data_size = len;
725 int idx = vol_id2idx(ubi, vol_id);
726 struct ubi_volume *vol = ubi->volumes[idx];
727 struct ubi_vid_hdr *vid_hdr;
733 if (lnum == used_ebs - 1)
734 /* If this is the last LEB @len may be unaligned */
735 len = ALIGN(data_size, ubi->min_io_size);
737 ubi_assert(len % ubi->min_io_size == 0);
739 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
743 err = leb_write_lock(ubi, vol_id, lnum);
745 ubi_free_vid_hdr(ubi, vid_hdr);
749 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
750 vid_hdr->vol_id = cpu_to_be32(vol_id);
751 vid_hdr->lnum = cpu_to_be32(lnum);
752 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
753 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
755 crc = crc32(UBI_CRC32_INIT, buf, data_size);
756 vid_hdr->vol_type = UBI_VID_STATIC;
757 vid_hdr->data_size = cpu_to_be32(data_size);
758 vid_hdr->used_ebs = cpu_to_be32(used_ebs);
759 vid_hdr->data_crc = cpu_to_be32(crc);
762 pnum = ubi_wl_get_peb(ubi, dtype);
764 ubi_free_vid_hdr(ubi, vid_hdr);
765 leb_write_unlock(ubi, vol_id, lnum);
769 dbg_eba("write VID hdr and %d bytes at LEB %d:%d, PEB %d, used_ebs %d",
770 len, vol_id, lnum, pnum, used_ebs);
772 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
774 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
779 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
781 ubi_warn("failed to write %d bytes of data to PEB %d",
786 ubi_assert(vol->eba_tbl[lnum] < 0);
787 vol->eba_tbl[lnum] = pnum;
789 leb_write_unlock(ubi, vol_id, lnum);
790 ubi_free_vid_hdr(ubi, vid_hdr);
794 if (err != -EIO || !ubi->bad_allowed) {
796 * This flash device does not admit of bad eraseblocks or
797 * something nasty and unexpected happened. Switch to read-only
801 leb_write_unlock(ubi, vol_id, lnum);
802 ubi_free_vid_hdr(ubi, vid_hdr);
806 err = ubi_wl_put_peb(ubi, pnum, 1);
807 if (err || ++tries > UBI_IO_RETRIES) {
809 leb_write_unlock(ubi, vol_id, lnum);
810 ubi_free_vid_hdr(ubi, vid_hdr);
814 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
815 ubi_msg("try another PEB");
820 * ubi_eba_atomic_leb_change - change logical eraseblock atomically.
821 * @ubi: UBI device description object
823 * @lnum: logical eraseblock number
824 * @buf: data to write
825 * @len: how many bytes to write
828 * This function changes the contents of a logical eraseblock atomically. @buf
829 * has to contain new logical eraseblock data, and @len - the length of the
830 * data, which has to be aligned. This function guarantees that in case of an
831 * unclean reboot the old contents is preserved. Returns zero in case of
832 * success and a negative error code in case of failure.
834 * UBI reserves one LEB for the "atomic LEB change" operation, so only one
835 * LEB change may be done at a time. This is ensured by @ubi->alc_mutex.
837 int ubi_eba_atomic_leb_change(struct ubi_device *ubi, int vol_id, int lnum,
838 const void *buf, int len, int dtype)
840 int err, pnum, tries = 0, idx = vol_id2idx(ubi, vol_id);
841 struct ubi_volume *vol = ubi->volumes[idx];
842 struct ubi_vid_hdr *vid_hdr;
848 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
852 mutex_lock(&ubi->alc_mutex);
853 err = leb_write_lock(ubi, vol_id, lnum);
857 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
858 vid_hdr->vol_id = cpu_to_be32(vol_id);
859 vid_hdr->lnum = cpu_to_be32(lnum);
860 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
861 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
863 crc = crc32(UBI_CRC32_INIT, buf, len);
864 vid_hdr->vol_type = UBI_VID_DYNAMIC;
865 vid_hdr->data_size = cpu_to_be32(len);
866 vid_hdr->copy_flag = 1;
867 vid_hdr->data_crc = cpu_to_be32(crc);
870 pnum = ubi_wl_get_peb(ubi, dtype);
876 dbg_eba("change LEB %d:%d, PEB %d, write VID hdr to PEB %d",
877 vol_id, lnum, vol->eba_tbl[lnum], pnum);
879 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
881 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
886 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
888 ubi_warn("failed to write %d bytes of data to PEB %d",
893 if (vol->eba_tbl[lnum] >= 0) {
894 err = ubi_wl_put_peb(ubi, vol->eba_tbl[lnum], 1);
899 vol->eba_tbl[lnum] = pnum;
902 leb_write_unlock(ubi, vol_id, lnum);
904 mutex_unlock(&ubi->alc_mutex);
905 ubi_free_vid_hdr(ubi, vid_hdr);
909 if (err != -EIO || !ubi->bad_allowed) {
911 * This flash device does not admit of bad eraseblocks or
912 * something nasty and unexpected happened. Switch to read-only
919 err = ubi_wl_put_peb(ubi, pnum, 1);
920 if (err || ++tries > UBI_IO_RETRIES) {
925 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
926 ubi_msg("try another PEB");
931 * ltree_entry_ctor - lock tree entries slab cache constructor.
932 * @obj: the lock-tree entry to construct
933 * @cache: the lock tree entry slab cache
934 * @flags: constructor flags
936 static void ltree_entry_ctor(void *obj, struct kmem_cache *cache,
939 struct ltree_entry *le = obj;
942 init_rwsem(&le->mutex);
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
954 * function. Returns zero in case of success, %UBI_IO_BITFLIPS if the operation
955 * was canceled because bit-flips were detected at the target PEB, and a
956 * negative error code in case of failure.
958 int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
959 struct ubi_vid_hdr *vid_hdr)
961 int err, vol_id, lnum, data_size, aldata_size, pnum, idx;
962 struct ubi_volume *vol;
965 vol_id = be32_to_cpu(vid_hdr->vol_id);
966 lnum = be32_to_cpu(vid_hdr->lnum);
968 dbg_eba("copy LEB %d:%d, PEB %d to PEB %d", vol_id, lnum, from, to);
970 if (vid_hdr->vol_type == UBI_VID_STATIC) {
971 data_size = be32_to_cpu(vid_hdr->data_size);
972 aldata_size = ALIGN(data_size, ubi->min_io_size);
974 data_size = aldata_size =
975 ubi->leb_size - be32_to_cpu(vid_hdr->data_pad);
978 * We do not want anybody to write to this logical eraseblock while we
979 * are moving it, so we lock it.
981 err = leb_write_lock(ubi, vol_id, lnum);
985 mutex_lock(&ubi->buf_mutex);
988 * But the logical eraseblock might have been put by this time.
989 * Cancel if it is true.
991 idx = vol_id2idx(ubi, vol_id);
994 * We may race with volume deletion/re-size, so we have to hold
995 * @ubi->volumes_lock.
997 spin_lock(&ubi->volumes_lock);
998 vol = ubi->volumes[idx];
1000 dbg_eba("volume %d was removed meanwhile", vol_id);
1001 spin_unlock(&ubi->volumes_lock);
1005 pnum = vol->eba_tbl[lnum];
1007 dbg_eba("LEB %d:%d is no longer mapped to PEB %d, mapped to "
1008 "PEB %d, cancel", vol_id, lnum, from, pnum);
1009 spin_unlock(&ubi->volumes_lock);
1012 spin_unlock(&ubi->volumes_lock);
1014 /* OK, now the LEB is locked and we can safely start moving it */
1016 dbg_eba("read %d bytes of data", aldata_size);
1017 err = ubi_io_read_data(ubi, ubi->peb_buf1, from, 0, aldata_size);
1018 if (err && err != UBI_IO_BITFLIPS) {
1019 ubi_warn("error %d while reading data from PEB %d",
1025 * Now we have got to calculate how much data we have to to copy. In
1026 * case of a static volume it is fairly easy - the VID header contains
1027 * the data size. In case of a dynamic volume it is more difficult - we
1028 * have to read the contents, cut 0xFF bytes from the end and copy only
1029 * the first part. We must do this to avoid writing 0xFF bytes as it
1030 * may have some side-effects. And not only this. It is important not
1031 * to include those 0xFFs to CRC because later the they may be filled
1034 if (vid_hdr->vol_type == UBI_VID_DYNAMIC)
1035 aldata_size = data_size =
1036 ubi_calc_data_len(ubi, ubi->peb_buf1, data_size);
1039 crc = crc32(UBI_CRC32_INIT, ubi->peb_buf1, data_size);
1043 * It may turn out to me that the whole @from physical eraseblock
1044 * contains only 0xFF bytes. Then we have to only write the VID header
1045 * and do not write any data. This also means we should not set
1046 * @vid_hdr->copy_flag, @vid_hdr->data_size, and @vid_hdr->data_crc.
1048 if (data_size > 0) {
1049 vid_hdr->copy_flag = 1;
1050 vid_hdr->data_size = cpu_to_be32(data_size);
1051 vid_hdr->data_crc = cpu_to_be32(crc);
1053 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
1055 err = ubi_io_write_vid_hdr(ubi, to, vid_hdr);
1061 /* Read the VID header back and check if it was written correctly */
1062 err = ubi_io_read_vid_hdr(ubi, to, vid_hdr, 1);
1064 if (err != UBI_IO_BITFLIPS)
1065 ubi_warn("cannot read VID header back from PEB %d", to);
1069 if (data_size > 0) {
1070 err = ubi_io_write_data(ubi, ubi->peb_buf1, to, 0, aldata_size);
1077 * We've written the data and are going to read it back to make
1078 * sure it was written correctly.
1081 err = ubi_io_read_data(ubi, ubi->peb_buf2, to, 0, aldata_size);
1083 if (err != UBI_IO_BITFLIPS)
1084 ubi_warn("cannot read data back from PEB %d",
1091 if (memcmp(ubi->peb_buf1, ubi->peb_buf2, aldata_size)) {
1092 ubi_warn("read data back from PEB %d - it is different",
1098 ubi_assert(vol->eba_tbl[lnum] == from);
1099 vol->eba_tbl[lnum] = to;
1102 mutex_unlock(&ubi->buf_mutex);
1103 leb_write_unlock(ubi, vol_id, lnum);
1108 * ubi_eba_init_scan - initialize the EBA unit using scanning information.
1109 * @ubi: UBI device description object
1110 * @si: scanning information
1112 * This function returns zero in case of success and a negative error code in
1115 int ubi_eba_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
1117 int i, j, err, num_volumes;
1118 struct ubi_scan_volume *sv;
1119 struct ubi_volume *vol;
1120 struct ubi_scan_leb *seb;
1123 dbg_eba("initialize EBA unit");
1125 spin_lock_init(&ubi->ltree_lock);
1126 mutex_init(&ubi->alc_mutex);
1127 ubi->ltree = RB_ROOT;
1129 if (ubi_devices_cnt == 0) {
1130 ltree_slab = kmem_cache_create("ubi_ltree_slab",
1131 sizeof(struct ltree_entry), 0,
1132 0, <ree_entry_ctor);
1137 ubi->global_sqnum = si->max_sqnum + 1;
1138 num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
1140 for (i = 0; i < num_volumes; i++) {
1141 vol = ubi->volumes[i];
1147 vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int),
1149 if (!vol->eba_tbl) {
1154 for (j = 0; j < vol->reserved_pebs; j++)
1155 vol->eba_tbl[j] = UBI_LEB_UNMAPPED;
1157 sv = ubi_scan_find_sv(si, idx2vol_id(ubi, i));
1161 ubi_rb_for_each_entry(rb, seb, &sv->root, u.rb) {
1162 if (seb->lnum >= vol->reserved_pebs)
1164 * This may happen in case of an unclean reboot
1167 ubi_scan_move_to_list(sv, seb, &si->erase);
1168 vol->eba_tbl[seb->lnum] = seb->pnum;
1172 if (ubi->bad_allowed) {
1173 ubi_calculate_reserved(ubi);
1175 if (ubi->avail_pebs < ubi->beb_rsvd_level) {
1176 /* No enough free physical eraseblocks */
1177 ubi->beb_rsvd_pebs = ubi->avail_pebs;
1178 ubi_warn("cannot reserve enough PEBs for bad PEB "
1179 "handling, reserved %d, need %d",
1180 ubi->beb_rsvd_pebs, ubi->beb_rsvd_level);
1182 ubi->beb_rsvd_pebs = ubi->beb_rsvd_level;
1184 ubi->avail_pebs -= ubi->beb_rsvd_pebs;
1185 ubi->rsvd_pebs += ubi->beb_rsvd_pebs;
1188 if (ubi->avail_pebs < EBA_RESERVED_PEBS) {
1189 ubi_err("no enough physical eraseblocks (%d, need %d)",
1190 ubi->avail_pebs, EBA_RESERVED_PEBS);
1194 ubi->avail_pebs -= EBA_RESERVED_PEBS;
1195 ubi->rsvd_pebs += EBA_RESERVED_PEBS;
1197 dbg_eba("EBA unit is initialized");
1201 for (i = 0; i < num_volumes; i++) {
1202 if (!ubi->volumes[i])
1204 kfree(ubi->volumes[i]->eba_tbl);
1206 if (ubi_devices_cnt == 0)
1207 kmem_cache_destroy(ltree_slab);
1212 * ubi_eba_close - close EBA unit.
1213 * @ubi: UBI device description object
1215 void ubi_eba_close(const struct ubi_device *ubi)
1217 int i, num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
1219 dbg_eba("close EBA unit");
1221 for (i = 0; i < num_volumes; i++) {
1222 if (!ubi->volumes[i])
1224 kfree(ubi->volumes[i]->eba_tbl);
1226 if (ubi_devices_cnt == 1)
1227 kmem_cache_destroy(ltree_slab);