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>
50 * struct ltree_entry - an entry in the lock tree.
51 * @rb: links RB-tree nodes
52 * @vol_id: volume ID of the locked logical eraseblock
53 * @lnum: locked logical eraseblock number
54 * @users: how many tasks are using this logical eraseblock or wait for it
55 * @mutex: read/write mutex to implement read/write access serialization to
56 * the (@vol_id, @lnum) logical eraseblock
58 * When a logical eraseblock is being locked - corresponding &struct ltree_entry
59 * object is inserted to the lock tree (@ubi->ltree).
66 struct rw_semaphore mutex;
69 /* Slab cache for lock-tree entries */
70 static struct kmem_cache *ltree_slab;
73 * next_sqnum - get next sequence number.
74 * @ubi: UBI device description object
76 * This function returns next sequence number to use, which is just the current
77 * global sequence counter value. It also increases the global sequence
80 static unsigned long long next_sqnum(struct ubi_device *ubi)
82 unsigned long long sqnum;
84 spin_lock(&ubi->ltree_lock);
85 sqnum = ubi->global_sqnum++;
86 spin_unlock(&ubi->ltree_lock);
92 * ubi_get_compat - get compatibility flags of a volume.
93 * @ubi: UBI device description object
96 * This function returns compatibility flags for an internal volume. User
97 * volumes have no compatibility flags, so %0 is returned.
99 static int ubi_get_compat(const struct ubi_device *ubi, int vol_id)
101 if (vol_id == UBI_LAYOUT_VOL_ID)
102 return UBI_LAYOUT_VOLUME_COMPAT;
107 * ltree_lookup - look up the lock tree.
108 * @ubi: UBI device description object
110 * @lnum: logical eraseblock number
112 * This function returns a pointer to the corresponding &struct ltree_entry
113 * object if the logical eraseblock is locked and %NULL if it is not.
114 * @ubi->ltree_lock has to be locked.
116 static struct ltree_entry *ltree_lookup(struct ubi_device *ubi, int vol_id,
121 p = ubi->ltree.rb_node;
123 struct ltree_entry *le;
125 le = rb_entry(p, struct ltree_entry, rb);
127 if (vol_id < le->vol_id)
129 else if (vol_id > le->vol_id)
134 else if (lnum > le->lnum)
145 * ltree_add_entry - add new entry to the lock tree.
146 * @ubi: UBI device description object
148 * @lnum: logical eraseblock number
150 * This function adds new entry for logical eraseblock (@vol_id, @lnum) to the
151 * lock tree. If such entry is already there, its usage counter is increased.
152 * Returns pointer to the lock tree entry or %-ENOMEM if memory allocation
155 static struct ltree_entry *ltree_add_entry(struct ubi_device *ubi, int vol_id,
158 struct ltree_entry *le, *le1, *le_free;
160 le = kmem_cache_alloc(ltree_slab, GFP_NOFS);
162 return ERR_PTR(-ENOMEM);
167 spin_lock(&ubi->ltree_lock);
168 le1 = ltree_lookup(ubi, vol_id, lnum);
172 * This logical eraseblock is already locked. The newly
173 * allocated lock entry is not needed.
178 struct rb_node **p, *parent = NULL;
181 * No lock entry, add the newly allocated one to the
182 * @ubi->ltree RB-tree.
186 p = &ubi->ltree.rb_node;
189 le1 = rb_entry(parent, struct ltree_entry, rb);
191 if (vol_id < le1->vol_id)
193 else if (vol_id > le1->vol_id)
196 ubi_assert(lnum != le1->lnum);
197 if (lnum < le1->lnum)
204 rb_link_node(&le->rb, parent, p);
205 rb_insert_color(&le->rb, &ubi->ltree);
208 spin_unlock(&ubi->ltree_lock);
211 kmem_cache_free(ltree_slab, le_free);
217 * leb_read_lock - lock logical eraseblock for reading.
218 * @ubi: UBI device description object
220 * @lnum: logical eraseblock number
222 * This function locks a logical eraseblock for reading. Returns zero in case
223 * of success and a negative error code in case of failure.
225 static int leb_read_lock(struct ubi_device *ubi, int vol_id, int lnum)
227 struct ltree_entry *le;
229 le = ltree_add_entry(ubi, vol_id, lnum);
232 down_read(&le->mutex);
237 * leb_read_unlock - unlock logical eraseblock.
238 * @ubi: UBI device description object
240 * @lnum: logical eraseblock number
242 static void leb_read_unlock(struct ubi_device *ubi, int vol_id, int lnum)
245 struct ltree_entry *le;
247 spin_lock(&ubi->ltree_lock);
248 le = ltree_lookup(ubi, vol_id, lnum);
250 ubi_assert(le->users >= 0);
251 if (le->users == 0) {
252 rb_erase(&le->rb, &ubi->ltree);
255 spin_unlock(&ubi->ltree_lock);
259 kmem_cache_free(ltree_slab, le);
263 * leb_write_lock - lock logical eraseblock for writing.
264 * @ubi: UBI device description object
266 * @lnum: logical eraseblock number
268 * This function locks a logical eraseblock for writing. Returns zero in case
269 * of success and a negative error code in case of failure.
271 static int leb_write_lock(struct ubi_device *ubi, int vol_id, int lnum)
273 struct ltree_entry *le;
275 le = ltree_add_entry(ubi, vol_id, lnum);
278 down_write(&le->mutex);
283 * leb_write_unlock - unlock logical eraseblock.
284 * @ubi: UBI device description object
286 * @lnum: logical eraseblock number
288 static void leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
291 struct ltree_entry *le;
293 spin_lock(&ubi->ltree_lock);
294 le = ltree_lookup(ubi, vol_id, lnum);
296 ubi_assert(le->users >= 0);
297 if (le->users == 0) {
298 rb_erase(&le->rb, &ubi->ltree);
302 spin_unlock(&ubi->ltree_lock);
304 up_write(&le->mutex);
306 kmem_cache_free(ltree_slab, le);
310 * ubi_eba_unmap_leb - un-map logical eraseblock.
311 * @ubi: UBI device description object
313 * @lnum: logical eraseblock number
315 * This function un-maps logical eraseblock @lnum and schedules corresponding
316 * physical eraseblock for erasure. Returns zero in case of success and a
317 * negative error code in case of failure.
319 int ubi_eba_unmap_leb(struct ubi_device *ubi, int vol_id, int lnum)
321 int idx = vol_id2idx(ubi, vol_id), err, pnum;
322 struct ubi_volume *vol = ubi->volumes[idx];
327 err = leb_write_lock(ubi, vol_id, lnum);
331 pnum = vol->eba_tbl[lnum];
333 /* This logical eraseblock is already unmapped */
336 dbg_eba("erase LEB %d:%d, PEB %d", vol_id, lnum, pnum);
338 vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED;
339 err = ubi_wl_put_peb(ubi, pnum, 0);
342 leb_write_unlock(ubi, vol_id, lnum);
347 * ubi_eba_read_leb - read data.
348 * @ubi: UBI device description object
350 * @lnum: logical eraseblock number
351 * @buf: buffer to store the read data
352 * @offset: offset from where to read
353 * @len: how many bytes to read
354 * @check: data CRC check flag
356 * If the logical eraseblock @lnum is unmapped, @buf is filled with 0xFF
357 * bytes. The @check flag only makes sense for static volumes and forces
358 * eraseblock data CRC checking.
360 * In case of success this function returns zero. In case of a static volume,
361 * if data CRC mismatches - %-EBADMSG is returned. %-EBADMSG may also be
362 * returned for any volume type if an ECC error was detected by the MTD device
363 * driver. Other negative error cored may be returned in case of other errors.
365 int ubi_eba_read_leb(struct ubi_device *ubi, int vol_id, int lnum, void *buf,
366 int offset, int len, int check)
368 int err, pnum, scrub = 0, idx = vol_id2idx(ubi, vol_id);
369 struct ubi_vid_hdr *vid_hdr;
370 struct ubi_volume *vol = ubi->volumes[idx];
371 uint32_t uninitialized_var(crc);
373 err = leb_read_lock(ubi, vol_id, lnum);
377 pnum = vol->eba_tbl[lnum];
380 * The logical eraseblock is not mapped, fill the whole buffer
381 * with 0xFF bytes. The exception is static volumes for which
382 * it is an error to read unmapped logical eraseblocks.
384 dbg_eba("read %d bytes from offset %d of LEB %d:%d (unmapped)",
385 len, offset, vol_id, lnum);
386 leb_read_unlock(ubi, vol_id, lnum);
387 ubi_assert(vol->vol_type != UBI_STATIC_VOLUME);
388 memset(buf, 0xFF, len);
392 dbg_eba("read %d bytes from offset %d of LEB %d:%d, PEB %d",
393 len, offset, vol_id, lnum, pnum);
395 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
400 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
406 err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
407 if (err && err != UBI_IO_BITFLIPS) {
410 * The header is either absent or corrupted.
411 * The former case means there is a bug -
412 * switch to read-only mode just in case.
413 * The latter case means a real corruption - we
414 * may try to recover data. FIXME: but this is
417 if (err == UBI_IO_BAD_VID_HDR) {
418 ubi_warn("bad VID header at PEB %d, LEB"
419 "%d:%d", pnum, vol_id, lnum);
425 } else if (err == UBI_IO_BITFLIPS)
428 ubi_assert(lnum < be32_to_cpu(vid_hdr->used_ebs));
429 ubi_assert(len == be32_to_cpu(vid_hdr->data_size));
431 crc = be32_to_cpu(vid_hdr->data_crc);
432 ubi_free_vid_hdr(ubi, vid_hdr);
435 err = ubi_io_read_data(ubi, buf, pnum, offset, len);
437 if (err == UBI_IO_BITFLIPS) {
440 } else if (err == -EBADMSG) {
441 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
445 ubi_msg("force data checking");
454 uint32_t crc1 = crc32(UBI_CRC32_INIT, buf, len);
456 ubi_warn("CRC error: calculated %#08x, must be %#08x",
464 err = ubi_wl_scrub_peb(ubi, pnum);
466 leb_read_unlock(ubi, vol_id, lnum);
470 ubi_free_vid_hdr(ubi, vid_hdr);
472 leb_read_unlock(ubi, vol_id, lnum);
477 * recover_peb - recover from write failure.
478 * @ubi: UBI device description object
479 * @pnum: the physical eraseblock to recover
481 * @lnum: logical eraseblock number
482 * @buf: data which was not written because of the write failure
483 * @offset: offset of the failed write
484 * @len: how many bytes should have been written
486 * This function is called in case of a write failure and moves all good data
487 * from the potentially bad physical eraseblock to a good physical eraseblock.
488 * This function also writes the data which was not written due to the failure.
489 * Returns new physical eraseblock number in case of success, and a negative
490 * error code in case of failure.
492 static int recover_peb(struct ubi_device *ubi, int pnum, int vol_id, int lnum,
493 const void *buf, int offset, int len)
495 int err, idx = vol_id2idx(ubi, vol_id), new_pnum, data_size, tries = 0;
496 struct ubi_volume *vol = ubi->volumes[idx];
497 struct ubi_vid_hdr *vid_hdr;
499 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
504 mutex_lock(&ubi->buf_mutex);
507 new_pnum = ubi_wl_get_peb(ubi, UBI_UNKNOWN);
509 mutex_unlock(&ubi->buf_mutex);
510 ubi_free_vid_hdr(ubi, vid_hdr);
514 ubi_msg("recover PEB %d, move data to PEB %d", pnum, new_pnum);
516 err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
517 if (err && err != UBI_IO_BITFLIPS) {
523 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
524 err = ubi_io_write_vid_hdr(ubi, new_pnum, vid_hdr);
528 data_size = offset + len;
529 memset(ubi->peb_buf1 + offset, 0xFF, len);
531 /* Read everything before the area where the write failure happened */
533 err = ubi_io_read_data(ubi, ubi->peb_buf1, pnum, 0, offset);
534 if (err && err != UBI_IO_BITFLIPS)
538 memcpy(ubi->peb_buf1 + offset, buf, len);
540 err = ubi_io_write_data(ubi, ubi->peb_buf1, new_pnum, 0, data_size);
544 mutex_unlock(&ubi->buf_mutex);
545 ubi_free_vid_hdr(ubi, vid_hdr);
547 vol->eba_tbl[lnum] = new_pnum;
548 ubi_wl_put_peb(ubi, pnum, 1);
550 ubi_msg("data was successfully recovered");
554 mutex_unlock(&ubi->buf_mutex);
555 ubi_wl_put_peb(ubi, new_pnum, 1);
556 ubi_free_vid_hdr(ubi, vid_hdr);
561 * Bad luck? This physical eraseblock is bad too? Crud. Let's try to
564 ubi_warn("failed to write to PEB %d", new_pnum);
565 ubi_wl_put_peb(ubi, new_pnum, 1);
566 if (++tries > UBI_IO_RETRIES) {
567 mutex_unlock(&ubi->buf_mutex);
568 ubi_free_vid_hdr(ubi, vid_hdr);
571 ubi_msg("try again");
576 * ubi_eba_write_leb - write data to dynamic volume.
577 * @ubi: UBI device description object
579 * @lnum: logical eraseblock number
580 * @buf: the data to write
581 * @offset: offset within the logical eraseblock where to write
582 * @len: how many bytes to write
585 * This function writes data to logical eraseblock @lnum of a dynamic volume
586 * @vol_id. Returns zero in case of success and a negative error code in case
587 * of failure. In case of error, it is possible that something was still
588 * written to the flash media, but may be some garbage.
590 int ubi_eba_write_leb(struct ubi_device *ubi, int vol_id, int lnum,
591 const void *buf, int offset, int len, int dtype)
593 int idx = vol_id2idx(ubi, vol_id), err, pnum, tries = 0;
594 struct ubi_volume *vol = ubi->volumes[idx];
595 struct ubi_vid_hdr *vid_hdr;
600 err = leb_write_lock(ubi, vol_id, lnum);
604 pnum = vol->eba_tbl[lnum];
606 dbg_eba("write %d bytes at offset %d of LEB %d:%d, PEB %d",
607 len, offset, vol_id, lnum, pnum);
609 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
611 ubi_warn("failed to write data to PEB %d", pnum);
612 if (err == -EIO && ubi->bad_allowed)
613 err = recover_peb(ubi, pnum, vol_id, lnum, buf, offset, len);
617 leb_write_unlock(ubi, vol_id, lnum);
622 * The logical eraseblock is not mapped. We have to get a free physical
623 * eraseblock and write the volume identifier header there first.
625 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
627 leb_write_unlock(ubi, vol_id, lnum);
631 vid_hdr->vol_type = UBI_VID_DYNAMIC;
632 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
633 vid_hdr->vol_id = cpu_to_be32(vol_id);
634 vid_hdr->lnum = cpu_to_be32(lnum);
635 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
636 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
639 pnum = ubi_wl_get_peb(ubi, dtype);
641 ubi_free_vid_hdr(ubi, vid_hdr);
642 leb_write_unlock(ubi, vol_id, lnum);
646 dbg_eba("write VID hdr and %d bytes at offset %d of LEB %d:%d, PEB %d",
647 len, offset, vol_id, lnum, pnum);
649 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
651 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
656 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
658 ubi_warn("failed to write %d bytes at offset %d of LEB %d:%d, "
659 "PEB %d", len, offset, vol_id, lnum, pnum);
663 vol->eba_tbl[lnum] = pnum;
665 leb_write_unlock(ubi, vol_id, lnum);
666 ubi_free_vid_hdr(ubi, vid_hdr);
670 if (err != -EIO || !ubi->bad_allowed) {
672 leb_write_unlock(ubi, vol_id, lnum);
673 ubi_free_vid_hdr(ubi, vid_hdr);
678 * Fortunately, this is the first write operation to this physical
679 * eraseblock, so just put it and request a new one. We assume that if
680 * this physical eraseblock went bad, the erase code will handle that.
682 err = ubi_wl_put_peb(ubi, pnum, 1);
683 if (err || ++tries > UBI_IO_RETRIES) {
685 leb_write_unlock(ubi, vol_id, lnum);
686 ubi_free_vid_hdr(ubi, vid_hdr);
690 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
691 ubi_msg("try another PEB");
696 * ubi_eba_write_leb_st - write data to static volume.
697 * @ubi: UBI device description object
699 * @lnum: logical eraseblock number
700 * @buf: data to write
701 * @len: how many bytes to write
703 * @used_ebs: how many logical eraseblocks will this volume contain
705 * This function writes data to logical eraseblock @lnum of static volume
706 * @vol_id. The @used_ebs argument should contain total number of logical
707 * eraseblock in this static volume.
709 * When writing to the last logical eraseblock, the @len argument doesn't have
710 * to be aligned to the minimal I/O unit size. Instead, it has to be equivalent
711 * to the real data size, although the @buf buffer has to contain the
712 * alignment. In all other cases, @len has to be aligned.
714 * It is prohibited to write more then once to logical eraseblocks of static
715 * volumes. This function returns zero in case of success and a negative error
716 * code in case of failure.
718 int ubi_eba_write_leb_st(struct ubi_device *ubi, int vol_id, int lnum,
719 const void *buf, int len, int dtype, int used_ebs)
721 int err, pnum, tries = 0, data_size = len;
722 int idx = vol_id2idx(ubi, vol_id);
723 struct ubi_volume *vol = ubi->volumes[idx];
724 struct ubi_vid_hdr *vid_hdr;
730 if (lnum == used_ebs - 1)
731 /* If this is the last LEB @len may be unaligned */
732 len = ALIGN(data_size, ubi->min_io_size);
734 ubi_assert(len % ubi->min_io_size == 0);
736 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
740 err = leb_write_lock(ubi, vol_id, lnum);
742 ubi_free_vid_hdr(ubi, vid_hdr);
746 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
747 vid_hdr->vol_id = cpu_to_be32(vol_id);
748 vid_hdr->lnum = cpu_to_be32(lnum);
749 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
750 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
752 crc = crc32(UBI_CRC32_INIT, buf, data_size);
753 vid_hdr->vol_type = UBI_VID_STATIC;
754 vid_hdr->data_size = cpu_to_be32(data_size);
755 vid_hdr->used_ebs = cpu_to_be32(used_ebs);
756 vid_hdr->data_crc = cpu_to_be32(crc);
759 pnum = ubi_wl_get_peb(ubi, dtype);
761 ubi_free_vid_hdr(ubi, vid_hdr);
762 leb_write_unlock(ubi, vol_id, lnum);
766 dbg_eba("write VID hdr and %d bytes at LEB %d:%d, PEB %d, used_ebs %d",
767 len, vol_id, lnum, pnum, used_ebs);
769 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
771 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
776 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
778 ubi_warn("failed to write %d bytes of data to PEB %d",
783 ubi_assert(vol->eba_tbl[lnum] < 0);
784 vol->eba_tbl[lnum] = pnum;
786 leb_write_unlock(ubi, vol_id, lnum);
787 ubi_free_vid_hdr(ubi, vid_hdr);
791 if (err != -EIO || !ubi->bad_allowed) {
793 * This flash device does not admit of bad eraseblocks or
794 * something nasty and unexpected happened. Switch to read-only
798 leb_write_unlock(ubi, vol_id, lnum);
799 ubi_free_vid_hdr(ubi, vid_hdr);
803 err = ubi_wl_put_peb(ubi, pnum, 1);
804 if (err || ++tries > UBI_IO_RETRIES) {
806 leb_write_unlock(ubi, vol_id, lnum);
807 ubi_free_vid_hdr(ubi, vid_hdr);
811 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
812 ubi_msg("try another PEB");
817 * ubi_eba_atomic_leb_change - change logical eraseblock atomically.
818 * @ubi: UBI device description object
820 * @lnum: logical eraseblock number
821 * @buf: data to write
822 * @len: how many bytes to write
825 * This function changes the contents of a logical eraseblock atomically. @buf
826 * has to contain new logical eraseblock data, and @len - the length of the
827 * data, which has to be aligned. This function guarantees that in case of an
828 * unclean reboot the old contents is preserved. Returns zero in case of
829 * success and a negative error code in case of failure.
831 int ubi_eba_atomic_leb_change(struct ubi_device *ubi, int vol_id, int lnum,
832 const void *buf, int len, int dtype)
834 int err, pnum, tries = 0, idx = vol_id2idx(ubi, vol_id);
835 struct ubi_volume *vol = ubi->volumes[idx];
836 struct ubi_vid_hdr *vid_hdr;
842 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
846 err = leb_write_lock(ubi, vol_id, lnum);
848 ubi_free_vid_hdr(ubi, vid_hdr);
852 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
853 vid_hdr->vol_id = cpu_to_be32(vol_id);
854 vid_hdr->lnum = cpu_to_be32(lnum);
855 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
856 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
858 crc = crc32(UBI_CRC32_INIT, buf, len);
859 vid_hdr->vol_type = UBI_VID_DYNAMIC;
860 vid_hdr->data_size = cpu_to_be32(len);
861 vid_hdr->copy_flag = 1;
862 vid_hdr->data_crc = cpu_to_be32(crc);
865 pnum = ubi_wl_get_peb(ubi, dtype);
867 ubi_free_vid_hdr(ubi, vid_hdr);
868 leb_write_unlock(ubi, vol_id, lnum);
872 dbg_eba("change LEB %d:%d, PEB %d, write VID hdr to PEB %d",
873 vol_id, lnum, vol->eba_tbl[lnum], pnum);
875 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
877 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
882 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
884 ubi_warn("failed to write %d bytes of data to PEB %d",
889 if (vol->eba_tbl[lnum] >= 0) {
890 err = ubi_wl_put_peb(ubi, vol->eba_tbl[lnum], 1);
892 ubi_free_vid_hdr(ubi, vid_hdr);
893 leb_write_unlock(ubi, vol_id, lnum);
898 vol->eba_tbl[lnum] = pnum;
899 leb_write_unlock(ubi, vol_id, lnum);
900 ubi_free_vid_hdr(ubi, vid_hdr);
904 if (err != -EIO || !ubi->bad_allowed) {
906 * This flash device does not admit of bad eraseblocks or
907 * something nasty and unexpected happened. Switch to read-only
911 leb_write_unlock(ubi, vol_id, lnum);
912 ubi_free_vid_hdr(ubi, vid_hdr);
916 err = ubi_wl_put_peb(ubi, pnum, 1);
917 if (err || ++tries > UBI_IO_RETRIES) {
919 leb_write_unlock(ubi, vol_id, lnum);
920 ubi_free_vid_hdr(ubi, vid_hdr);
924 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
925 ubi_msg("try another PEB");
930 * ltree_entry_ctor - lock tree entries slab cache constructor.
931 * @obj: the lock-tree entry to construct
932 * @cache: the lock tree entry slab cache
933 * @flags: constructor flags
935 static void ltree_entry_ctor(void *obj, struct kmem_cache *cache,
938 struct ltree_entry *le = obj;
941 init_rwsem(&le->mutex);
945 * ubi_eba_copy_leb - copy logical eraseblock.
946 * @ubi: UBI device description object
947 * @from: physical eraseblock number from where to copy
948 * @to: physical eraseblock number where to copy
949 * @vid_hdr: VID header of the @from physical eraseblock
951 * This function copies logical eraseblock from physical eraseblock @from to
952 * physical eraseblock @to. The @vid_hdr buffer may be changed by this
953 * function. Returns zero in case of success, %UBI_IO_BITFLIPS if the operation
954 * was canceled because bit-flips were detected at the target PEB, and a
955 * negative error code in case of failure.
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, pnum, 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);
977 * We do not want anybody to write to this logical eraseblock while we
978 * are moving it, so we lock it.
980 err = leb_write_lock(ubi, vol_id, lnum);
984 mutex_lock(&ubi->buf_mutex);
987 * But the logical eraseblock might have been put by this time.
988 * Cancel if it is true.
990 idx = vol_id2idx(ubi, vol_id);
993 * We may race with volume deletion/re-size, so we have to hold
994 * @ubi->volumes_lock.
996 spin_lock(&ubi->volumes_lock);
997 vol = ubi->volumes[idx];
999 dbg_eba("volume %d was removed meanwhile", vol_id);
1000 spin_unlock(&ubi->volumes_lock);
1004 pnum = vol->eba_tbl[lnum];
1006 dbg_eba("LEB %d:%d is no longer mapped to PEB %d, mapped to "
1007 "PEB %d, cancel", vol_id, lnum, from, pnum);
1008 spin_unlock(&ubi->volumes_lock);
1011 spin_unlock(&ubi->volumes_lock);
1013 /* OK, now the LEB is locked and we can safely start moving it */
1015 dbg_eba("read %d bytes of data", aldata_size);
1016 err = ubi_io_read_data(ubi, ubi->peb_buf1, from, 0, aldata_size);
1017 if (err && err != UBI_IO_BITFLIPS) {
1018 ubi_warn("error %d while reading data from PEB %d",
1024 * Now we have got to calculate how much data we have to to copy. In
1025 * case of a static volume it is fairly easy - the VID header contains
1026 * the data size. In case of a dynamic volume it is more difficult - we
1027 * have to read the contents, cut 0xFF bytes from the end and copy only
1028 * the first part. We must do this to avoid writing 0xFF bytes as it
1029 * may have some side-effects. And not only this. It is important not
1030 * to include those 0xFFs to CRC because later the they may be filled
1033 if (vid_hdr->vol_type == UBI_VID_DYNAMIC)
1034 aldata_size = data_size =
1035 ubi_calc_data_len(ubi, ubi->peb_buf1, data_size);
1038 crc = crc32(UBI_CRC32_INIT, ubi->peb_buf1, data_size);
1042 * It may turn out to me that the whole @from physical eraseblock
1043 * contains only 0xFF bytes. Then we have to only write the VID header
1044 * and do not write any data. This also means we should not set
1045 * @vid_hdr->copy_flag, @vid_hdr->data_size, and @vid_hdr->data_crc.
1047 if (data_size > 0) {
1048 vid_hdr->copy_flag = 1;
1049 vid_hdr->data_size = cpu_to_be32(data_size);
1050 vid_hdr->data_crc = cpu_to_be32(crc);
1052 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
1054 err = ubi_io_write_vid_hdr(ubi, to, vid_hdr);
1060 /* Read the VID header back and check if it was written correctly */
1061 err = ubi_io_read_vid_hdr(ubi, to, vid_hdr, 1);
1063 if (err != UBI_IO_BITFLIPS)
1064 ubi_warn("cannot read VID header back from PEB %d", to);
1068 if (data_size > 0) {
1069 err = ubi_io_write_data(ubi, ubi->peb_buf1, to, 0, aldata_size);
1076 * We've written the data and are going to read it back to make
1077 * sure it was written correctly.
1080 err = ubi_io_read_data(ubi, ubi->peb_buf2, to, 0, aldata_size);
1082 if (err != UBI_IO_BITFLIPS)
1083 ubi_warn("cannot read data back from PEB %d",
1090 if (memcmp(ubi->peb_buf1, ubi->peb_buf2, aldata_size)) {
1091 ubi_warn("read data back from PEB %d - it is different",
1097 ubi_assert(vol->eba_tbl[lnum] == from);
1098 vol->eba_tbl[lnum] = to;
1101 mutex_unlock(&ubi->buf_mutex);
1102 leb_write_unlock(ubi, vol_id, lnum);
1107 * ubi_eba_init_scan - initialize the EBA unit using scanning information.
1108 * @ubi: UBI device description object
1109 * @si: scanning information
1111 * This function returns zero in case of success and a negative error code in
1114 int ubi_eba_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
1116 int i, j, err, num_volumes;
1117 struct ubi_scan_volume *sv;
1118 struct ubi_volume *vol;
1119 struct ubi_scan_leb *seb;
1122 dbg_eba("initialize EBA unit");
1124 spin_lock_init(&ubi->ltree_lock);
1125 ubi->ltree = RB_ROOT;
1127 if (ubi_devices_cnt == 0) {
1128 ltree_slab = kmem_cache_create("ubi_ltree_slab",
1129 sizeof(struct ltree_entry), 0,
1130 0, <ree_entry_ctor);
1135 ubi->global_sqnum = si->max_sqnum + 1;
1136 num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
1138 for (i = 0; i < num_volumes; i++) {
1139 vol = ubi->volumes[i];
1145 vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int),
1147 if (!vol->eba_tbl) {
1152 for (j = 0; j < vol->reserved_pebs; j++)
1153 vol->eba_tbl[j] = UBI_LEB_UNMAPPED;
1155 sv = ubi_scan_find_sv(si, idx2vol_id(ubi, i));
1159 ubi_rb_for_each_entry(rb, seb, &sv->root, u.rb) {
1160 if (seb->lnum >= vol->reserved_pebs)
1162 * This may happen in case of an unclean reboot
1165 ubi_scan_move_to_list(sv, seb, &si->erase);
1166 vol->eba_tbl[seb->lnum] = seb->pnum;
1170 if (ubi->bad_allowed) {
1171 ubi_calculate_reserved(ubi);
1173 if (ubi->avail_pebs < ubi->beb_rsvd_level) {
1174 /* No enough free physical eraseblocks */
1175 ubi->beb_rsvd_pebs = ubi->avail_pebs;
1176 ubi_warn("cannot reserve enough PEBs for bad PEB "
1177 "handling, reserved %d, need %d",
1178 ubi->beb_rsvd_pebs, ubi->beb_rsvd_level);
1180 ubi->beb_rsvd_pebs = ubi->beb_rsvd_level;
1182 ubi->avail_pebs -= ubi->beb_rsvd_pebs;
1183 ubi->rsvd_pebs += ubi->beb_rsvd_pebs;
1186 dbg_eba("EBA unit is initialized");
1190 for (i = 0; i < num_volumes; i++) {
1191 if (!ubi->volumes[i])
1193 kfree(ubi->volumes[i]->eba_tbl);
1195 if (ubi_devices_cnt == 0)
1196 kmem_cache_destroy(ltree_slab);
1201 * ubi_eba_close - close EBA unit.
1202 * @ubi: UBI device description object
1204 void ubi_eba_close(const struct ubi_device *ubi)
1206 int i, num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
1208 dbg_eba("close EBA unit");
1210 for (i = 0; i < num_volumes; i++) {
1211 if (!ubi->volumes[i])
1213 kfree(ubi->volumes[i]->eba_tbl);
1215 if (ubi_devices_cnt == 1)
1216 kmem_cache_destroy(ltree_slab);