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_VOL_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 = kmem_cache_alloc(ubi_ltree_slab, GFP_NOFS);
142 return ERR_PTR(-ENOMEM);
147 spin_lock(&ubi->ltree_lock);
148 le1 = ltree_lookup(ubi, vol_id, lnum);
152 * This logical eraseblock is already locked. The newly
153 * allocated lock entry is not needed.
158 struct rb_node **p, *parent = NULL;
161 * No lock entry, add the newly allocated one to the
162 * @ubi->ltree RB-tree.
166 p = &ubi->ltree.rb_node;
169 le1 = rb_entry(parent, struct ubi_ltree_entry, rb);
171 if (vol_id < le1->vol_id)
173 else if (vol_id > le1->vol_id)
176 ubi_assert(lnum != le1->lnum);
177 if (lnum < le1->lnum)
184 rb_link_node(&le->rb, parent, p);
185 rb_insert_color(&le->rb, &ubi->ltree);
188 spin_unlock(&ubi->ltree_lock);
191 kmem_cache_free(ubi_ltree_slab, le_free);
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)
225 struct ubi_ltree_entry *le;
227 spin_lock(&ubi->ltree_lock);
228 le = ltree_lookup(ubi, vol_id, lnum);
230 ubi_assert(le->users >= 0);
231 if (le->users == 0) {
232 rb_erase(&le->rb, &ubi->ltree);
235 spin_unlock(&ubi->ltree_lock);
239 kmem_cache_free(ubi_ltree_slab, le);
243 * leb_write_lock - lock logical eraseblock for writing.
244 * @ubi: UBI device description object
246 * @lnum: logical eraseblock number
248 * This function locks a logical eraseblock for writing. Returns zero in case
249 * of success and a negative error code in case of failure.
251 static int leb_write_lock(struct ubi_device *ubi, int vol_id, int lnum)
253 struct ubi_ltree_entry *le;
255 le = ltree_add_entry(ubi, vol_id, lnum);
258 down_write(&le->mutex);
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 if there is no
269 * contention and does nothing if there is contention. Returns %0 in case of
270 * success, %1 in case of contention, and and a negative error code in case of
273 static int leb_write_trylock(struct ubi_device *ubi, int vol_id, int lnum)
276 struct ubi_ltree_entry *le;
278 le = ltree_add_entry(ubi, vol_id, lnum);
281 if (down_write_trylock(&le->mutex))
284 /* Contention, cancel */
285 spin_lock(&ubi->ltree_lock);
287 ubi_assert(le->users >= 0);
288 if (le->users == 0) {
289 rb_erase(&le->rb, &ubi->ltree);
293 spin_unlock(&ubi->ltree_lock);
295 kmem_cache_free(ubi_ltree_slab, le);
301 * leb_write_unlock - unlock logical eraseblock.
302 * @ubi: UBI device description object
304 * @lnum: logical eraseblock number
306 static void leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
309 struct ubi_ltree_entry *le;
311 spin_lock(&ubi->ltree_lock);
312 le = ltree_lookup(ubi, vol_id, lnum);
314 ubi_assert(le->users >= 0);
315 if (le->users == 0) {
316 rb_erase(&le->rb, &ubi->ltree);
320 spin_unlock(&ubi->ltree_lock);
322 up_write(&le->mutex);
324 kmem_cache_free(ubi_ltree_slab, le);
328 * ubi_eba_unmap_leb - un-map logical eraseblock.
329 * @ubi: UBI device description object
330 * @vol: volume description object
331 * @lnum: logical eraseblock number
333 * This function un-maps logical eraseblock @lnum and schedules corresponding
334 * physical eraseblock for erasure. Returns zero in case of success and a
335 * negative error code in case of failure.
337 int ubi_eba_unmap_leb(struct ubi_device *ubi, struct ubi_volume *vol,
340 int err, pnum, vol_id = vol->vol_id;
342 ubi_assert(ubi->ref_count > 0);
343 ubi_assert(vol->ref_count > 0);
348 err = leb_write_lock(ubi, vol_id, lnum);
352 pnum = vol->eba_tbl[lnum];
354 /* This logical eraseblock is already unmapped */
357 dbg_eba("erase LEB %d:%d, PEB %d", vol_id, lnum, pnum);
359 vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED;
360 err = ubi_wl_put_peb(ubi, pnum, 0);
363 leb_write_unlock(ubi, vol_id, lnum);
368 * ubi_eba_read_leb - read data.
369 * @ubi: UBI device description object
370 * @vol: volume description object
371 * @lnum: logical eraseblock number
372 * @buf: buffer to store the read data
373 * @offset: offset from where to read
374 * @len: how many bytes to read
375 * @check: data CRC check flag
377 * If the logical eraseblock @lnum is unmapped, @buf is filled with 0xFF
378 * bytes. The @check flag only makes sense for static volumes and forces
379 * eraseblock data CRC checking.
381 * In case of success this function returns zero. In case of a static volume,
382 * if data CRC mismatches - %-EBADMSG is returned. %-EBADMSG may also be
383 * returned for any volume type if an ECC error was detected by the MTD device
384 * driver. Other negative error cored may be returned in case of other errors.
386 int ubi_eba_read_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
387 void *buf, int offset, int len, int check)
389 int err, pnum, scrub = 0, vol_id = vol->vol_id;
390 struct ubi_vid_hdr *vid_hdr;
391 uint32_t uninitialized_var(crc);
393 ubi_assert(ubi->ref_count > 0);
394 ubi_assert(vol->ref_count > 0);
396 err = leb_read_lock(ubi, vol_id, lnum);
400 pnum = vol->eba_tbl[lnum];
403 * The logical eraseblock is not mapped, fill the whole buffer
404 * with 0xFF bytes. The exception is static volumes for which
405 * it is an error to read unmapped logical eraseblocks.
407 dbg_eba("read %d bytes from offset %d of LEB %d:%d (unmapped)",
408 len, offset, vol_id, lnum);
409 leb_read_unlock(ubi, vol_id, lnum);
410 ubi_assert(vol->vol_type != UBI_STATIC_VOLUME);
411 memset(buf, 0xFF, len);
415 dbg_eba("read %d bytes from offset %d of LEB %d:%d, PEB %d",
416 len, offset, vol_id, lnum, pnum);
418 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
423 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
429 err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
430 if (err && err != UBI_IO_BITFLIPS) {
433 * The header is either absent or corrupted.
434 * The former case means there is a bug -
435 * switch to read-only mode just in case.
436 * The latter case means a real corruption - we
437 * may try to recover data. FIXME: but this is
440 if (err == UBI_IO_BAD_VID_HDR) {
441 ubi_warn("bad VID header at PEB %d, LEB"
442 "%d:%d", pnum, vol_id, lnum);
448 } else if (err == UBI_IO_BITFLIPS)
451 ubi_assert(lnum < be32_to_cpu(vid_hdr->used_ebs));
452 ubi_assert(len == be32_to_cpu(vid_hdr->data_size));
454 crc = be32_to_cpu(vid_hdr->data_crc);
455 ubi_free_vid_hdr(ubi, vid_hdr);
458 err = ubi_io_read_data(ubi, buf, pnum, offset, len);
460 if (err == UBI_IO_BITFLIPS) {
463 } else if (err == -EBADMSG) {
464 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
468 ubi_msg("force data checking");
477 uint32_t crc1 = crc32(UBI_CRC32_INIT, buf, len);
479 ubi_warn("CRC error: calculated %#08x, must be %#08x",
487 err = ubi_wl_scrub_peb(ubi, pnum);
489 leb_read_unlock(ubi, vol_id, lnum);
493 ubi_free_vid_hdr(ubi, vid_hdr);
495 leb_read_unlock(ubi, vol_id, lnum);
500 * recover_peb - recover from write failure.
501 * @ubi: UBI device description object
502 * @pnum: the physical eraseblock to recover
504 * @lnum: logical eraseblock number
505 * @buf: data which was not written because of the write failure
506 * @offset: offset of the failed write
507 * @len: how many bytes should have been written
509 * This function is called in case of a write failure and moves all good data
510 * from the potentially bad physical eraseblock to a good physical eraseblock.
511 * This function also writes the data which was not written due to the failure.
512 * Returns new physical eraseblock number in case of success, and a negative
513 * error code in case of failure.
515 static int recover_peb(struct ubi_device *ubi, int pnum, int vol_id, int lnum,
516 const void *buf, int offset, int len)
518 int err, idx = vol_id2idx(ubi, vol_id), new_pnum, data_size, tries = 0;
519 struct ubi_volume *vol = ubi->volumes[idx];
520 struct ubi_vid_hdr *vid_hdr;
522 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
527 mutex_lock(&ubi->buf_mutex);
530 new_pnum = ubi_wl_get_peb(ubi, UBI_UNKNOWN);
532 mutex_unlock(&ubi->buf_mutex);
533 ubi_free_vid_hdr(ubi, vid_hdr);
537 ubi_msg("recover PEB %d, move data to PEB %d", pnum, new_pnum);
539 err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
540 if (err && err != UBI_IO_BITFLIPS) {
546 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
547 err = ubi_io_write_vid_hdr(ubi, new_pnum, vid_hdr);
551 data_size = offset + len;
552 memset(ubi->peb_buf1 + offset, 0xFF, len);
554 /* Read everything before the area where the write failure happened */
556 err = ubi_io_read_data(ubi, ubi->peb_buf1, pnum, 0, offset);
557 if (err && err != UBI_IO_BITFLIPS)
561 memcpy(ubi->peb_buf1 + offset, buf, len);
563 err = ubi_io_write_data(ubi, ubi->peb_buf1, new_pnum, 0, data_size);
567 mutex_unlock(&ubi->buf_mutex);
568 ubi_free_vid_hdr(ubi, vid_hdr);
570 vol->eba_tbl[lnum] = new_pnum;
571 ubi_wl_put_peb(ubi, pnum, 1);
573 ubi_msg("data was successfully recovered");
577 mutex_unlock(&ubi->buf_mutex);
578 ubi_wl_put_peb(ubi, new_pnum, 1);
579 ubi_free_vid_hdr(ubi, vid_hdr);
584 * Bad luck? This physical eraseblock is bad too? Crud. Let's try to
587 ubi_warn("failed to write to PEB %d", new_pnum);
588 ubi_wl_put_peb(ubi, new_pnum, 1);
589 if (++tries > UBI_IO_RETRIES) {
590 mutex_unlock(&ubi->buf_mutex);
591 ubi_free_vid_hdr(ubi, vid_hdr);
594 ubi_msg("try again");
599 * ubi_eba_write_leb - write data to dynamic volume.
600 * @ubi: UBI device description object
601 * @vol: volume description object
602 * @lnum: logical eraseblock number
603 * @buf: the data to write
604 * @offset: offset within the logical eraseblock where to write
605 * @len: how many bytes to write
608 * This function writes data to logical eraseblock @lnum of a dynamic volume
609 * @vol. Returns zero in case of success and a negative error code in case
610 * of failure. In case of error, it is possible that something was still
611 * written to the flash media, but may be some garbage.
613 int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
614 const void *buf, int offset, int len, int dtype)
616 int err, pnum, tries = 0, vol_id = vol->vol_id;
617 struct ubi_vid_hdr *vid_hdr;
619 ubi_assert(ubi->ref_count > 0);
620 ubi_assert(vol->ref_count > 0);
625 err = leb_write_lock(ubi, vol_id, lnum);
629 pnum = vol->eba_tbl[lnum];
631 dbg_eba("write %d bytes at offset %d of LEB %d:%d, PEB %d",
632 len, offset, vol_id, lnum, pnum);
634 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
636 ubi_warn("failed to write data to PEB %d", pnum);
637 if (err == -EIO && ubi->bad_allowed)
638 err = recover_peb(ubi, pnum, vol_id, lnum, buf,
643 leb_write_unlock(ubi, vol_id, lnum);
648 * The logical eraseblock is not mapped. We have to get a free physical
649 * eraseblock and write the volume identifier header there first.
651 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
653 leb_write_unlock(ubi, vol_id, lnum);
657 vid_hdr->vol_type = UBI_VID_DYNAMIC;
658 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
659 vid_hdr->vol_id = cpu_to_be32(vol_id);
660 vid_hdr->lnum = cpu_to_be32(lnum);
661 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
662 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
665 pnum = ubi_wl_get_peb(ubi, dtype);
667 ubi_free_vid_hdr(ubi, vid_hdr);
668 leb_write_unlock(ubi, vol_id, lnum);
672 dbg_eba("write VID hdr and %d bytes at offset %d of LEB %d:%d, PEB %d",
673 len, offset, vol_id, lnum, pnum);
675 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
677 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
683 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
685 ubi_warn("failed to write %d bytes at offset %d of "
686 "LEB %d:%d, PEB %d", len, offset, vol_id,
692 vol->eba_tbl[lnum] = pnum;
694 leb_write_unlock(ubi, vol_id, lnum);
695 ubi_free_vid_hdr(ubi, vid_hdr);
699 if (err != -EIO || !ubi->bad_allowed) {
701 leb_write_unlock(ubi, vol_id, lnum);
702 ubi_free_vid_hdr(ubi, vid_hdr);
707 * Fortunately, this is the first write operation to this physical
708 * eraseblock, so just put it and request a new one. We assume that if
709 * this physical eraseblock went bad, the erase code will handle that.
711 err = ubi_wl_put_peb(ubi, pnum, 1);
712 if (err || ++tries > UBI_IO_RETRIES) {
714 leb_write_unlock(ubi, vol_id, lnum);
715 ubi_free_vid_hdr(ubi, vid_hdr);
719 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
720 ubi_msg("try another PEB");
725 * ubi_eba_write_leb_st - write data to static volume.
726 * @ubi: UBI device description object
727 * @vol: volume description object
728 * @lnum: logical eraseblock number
729 * @buf: data to write
730 * @len: how many bytes to write
732 * @used_ebs: how many logical eraseblocks will this volume contain
734 * This function writes data to logical eraseblock @lnum of static volume
735 * @vol. The @used_ebs argument should contain total number of logical
736 * eraseblock in this static volume.
738 * When writing to the last logical eraseblock, the @len argument doesn't have
739 * to be aligned to the minimal I/O unit size. Instead, it has to be equivalent
740 * to the real data size, although the @buf buffer has to contain the
741 * alignment. In all other cases, @len has to be aligned.
743 * It is prohibited to write more then once to logical eraseblocks of static
744 * volumes. This function returns zero in case of success and a negative error
745 * code in case of failure.
747 int ubi_eba_write_leb_st(struct ubi_device *ubi, struct ubi_volume *vol,
748 int lnum, const void *buf, int len, int dtype,
751 int err, pnum, tries = 0, data_size = len, vol_id = vol->vol_id;
752 struct ubi_vid_hdr *vid_hdr;
755 ubi_assert(ubi->ref_count > 0);
756 ubi_assert(vol->ref_count > 0);
761 if (lnum == used_ebs - 1)
762 /* If this is the last LEB @len may be unaligned */
763 len = ALIGN(data_size, ubi->min_io_size);
765 ubi_assert(len % ubi->min_io_size == 0);
767 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
771 err = leb_write_lock(ubi, vol_id, lnum);
773 ubi_free_vid_hdr(ubi, vid_hdr);
777 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
778 vid_hdr->vol_id = cpu_to_be32(vol_id);
779 vid_hdr->lnum = cpu_to_be32(lnum);
780 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
781 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
783 crc = crc32(UBI_CRC32_INIT, buf, data_size);
784 vid_hdr->vol_type = UBI_VID_STATIC;
785 vid_hdr->data_size = cpu_to_be32(data_size);
786 vid_hdr->used_ebs = cpu_to_be32(used_ebs);
787 vid_hdr->data_crc = cpu_to_be32(crc);
790 pnum = ubi_wl_get_peb(ubi, dtype);
792 ubi_free_vid_hdr(ubi, vid_hdr);
793 leb_write_unlock(ubi, vol_id, lnum);
797 dbg_eba("write VID hdr and %d bytes at LEB %d:%d, PEB %d, used_ebs %d",
798 len, vol_id, lnum, pnum, used_ebs);
800 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
802 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
807 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
809 ubi_warn("failed to write %d bytes of data to PEB %d",
814 ubi_assert(vol->eba_tbl[lnum] < 0);
815 vol->eba_tbl[lnum] = pnum;
817 leb_write_unlock(ubi, vol_id, lnum);
818 ubi_free_vid_hdr(ubi, vid_hdr);
822 if (err != -EIO || !ubi->bad_allowed) {
824 * This flash device does not admit of bad eraseblocks or
825 * something nasty and unexpected happened. Switch to read-only
829 leb_write_unlock(ubi, vol_id, lnum);
830 ubi_free_vid_hdr(ubi, vid_hdr);
834 err = ubi_wl_put_peb(ubi, pnum, 1);
835 if (err || ++tries > UBI_IO_RETRIES) {
837 leb_write_unlock(ubi, vol_id, lnum);
838 ubi_free_vid_hdr(ubi, vid_hdr);
842 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
843 ubi_msg("try another PEB");
848 * ubi_eba_atomic_leb_change - change logical eraseblock atomically.
849 * @ubi: UBI device description object
850 * @vol: volume description object
851 * @lnum: logical eraseblock number
852 * @buf: data to write
853 * @len: how many bytes to write
856 * This function changes the contents of a logical eraseblock atomically. @buf
857 * has to contain new logical eraseblock data, and @len - the length of the
858 * data, which has to be aligned. This function guarantees that in case of an
859 * unclean reboot the old contents is preserved. Returns zero in case of
860 * success and a negative error code in case of failure.
862 * UBI reserves one LEB for the "atomic LEB change" operation, so only one
863 * LEB change may be done at a time. This is ensured by @ubi->alc_mutex.
865 int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
866 int lnum, const void *buf, int len, int dtype)
868 int err, pnum, tries = 0, vol_id = vol->vol_id;
869 struct ubi_vid_hdr *vid_hdr;
872 ubi_assert(ubi->ref_count > 0);
873 ubi_assert(vol->ref_count > 0);
878 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
882 mutex_lock(&ubi->alc_mutex);
883 err = leb_write_lock(ubi, vol_id, lnum);
887 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
888 vid_hdr->vol_id = cpu_to_be32(vol_id);
889 vid_hdr->lnum = cpu_to_be32(lnum);
890 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
891 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
893 crc = crc32(UBI_CRC32_INIT, buf, len);
894 vid_hdr->vol_type = UBI_VID_DYNAMIC;
895 vid_hdr->data_size = cpu_to_be32(len);
896 vid_hdr->copy_flag = 1;
897 vid_hdr->data_crc = cpu_to_be32(crc);
900 pnum = ubi_wl_get_peb(ubi, dtype);
906 dbg_eba("change LEB %d:%d, PEB %d, write VID hdr to PEB %d",
907 vol_id, lnum, vol->eba_tbl[lnum], pnum);
909 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
911 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
916 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
918 ubi_warn("failed to write %d bytes of data to PEB %d",
923 if (vol->eba_tbl[lnum] >= 0) {
924 err = ubi_wl_put_peb(ubi, vol->eba_tbl[lnum], 1);
929 vol->eba_tbl[lnum] = pnum;
932 leb_write_unlock(ubi, vol_id, lnum);
934 mutex_unlock(&ubi->alc_mutex);
935 ubi_free_vid_hdr(ubi, vid_hdr);
939 if (err != -EIO || !ubi->bad_allowed) {
941 * This flash device does not admit of bad eraseblocks or
942 * something nasty and unexpected happened. Switch to read-only
949 err = ubi_wl_put_peb(ubi, pnum, 1);
950 if (err || ++tries > UBI_IO_RETRIES) {
955 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
956 ubi_msg("try another PEB");
961 * ubi_eba_copy_leb - copy logical eraseblock.
962 * @ubi: UBI device description object
963 * @from: physical eraseblock number from where to copy
964 * @to: physical eraseblock number where to copy
965 * @vid_hdr: VID header of the @from physical eraseblock
967 * This function copies logical eraseblock from physical eraseblock @from to
968 * physical eraseblock @to. The @vid_hdr buffer may be changed by this
970 * o %0 in case of success;
971 * o %1 if the operation was canceled and should be tried later (e.g.,
972 * because a bit-flip was detected at the target PEB);
973 * o %2 if the volume is being deleted and this LEB should not be moved.
975 int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
976 struct ubi_vid_hdr *vid_hdr)
978 int err, vol_id, lnum, data_size, aldata_size, idx;
979 struct ubi_volume *vol;
982 vol_id = be32_to_cpu(vid_hdr->vol_id);
983 lnum = be32_to_cpu(vid_hdr->lnum);
985 dbg_eba("copy LEB %d:%d, PEB %d to PEB %d", vol_id, lnum, from, to);
987 if (vid_hdr->vol_type == UBI_VID_STATIC) {
988 data_size = be32_to_cpu(vid_hdr->data_size);
989 aldata_size = ALIGN(data_size, ubi->min_io_size);
991 data_size = aldata_size =
992 ubi->leb_size - be32_to_cpu(vid_hdr->data_pad);
994 idx = vol_id2idx(ubi, vol_id);
995 spin_lock(&ubi->volumes_lock);
997 * Note, we may race with volume deletion, which means that the volume
998 * this logical eraseblock belongs to might be being deleted. Since the
999 * volume deletion unmaps all the volume's logical eraseblocks, it will
1000 * be locked in 'ubi_wl_put_peb()' and wait for the WL worker to finish.
1002 vol = ubi->volumes[idx];
1004 /* No need to do further work, cancel */
1005 dbg_eba("volume %d is being removed, cancel", vol_id);
1006 spin_unlock(&ubi->volumes_lock);
1009 spin_unlock(&ubi->volumes_lock);
1012 * We do not want anybody to write to this logical eraseblock while we
1013 * are moving it, so lock it.
1015 * Note, we are using non-waiting locking here, because we cannot sleep
1016 * on the LEB, since it may cause deadlocks. Indeed, imagine a task is
1017 * unmapping the LEB which is mapped to the PEB we are going to move
1018 * (@from). This task locks the LEB and goes sleep in the
1019 * 'ubi_wl_put_peb()' function on the @ubi->move_mutex. In turn, we are
1020 * holding @ubi->move_mutex and go sleep on the LEB lock. So, if the
1021 * LEB is already locked, we just do not move it and return %1.
1023 err = leb_write_trylock(ubi, vol_id, lnum);
1025 dbg_eba("contention on LEB %d:%d, cancel", vol_id, lnum);
1030 * The LEB might have been put meanwhile, and the task which put it is
1031 * probably waiting on @ubi->move_mutex. No need to continue the work,
1034 if (vol->eba_tbl[lnum] != from) {
1035 dbg_eba("LEB %d:%d is no longer mapped to PEB %d, mapped to "
1036 "PEB %d, cancel", vol_id, lnum, from,
1037 vol->eba_tbl[lnum]);
1039 goto out_unlock_leb;
1043 * OK, now the LEB is locked and we can safely start moving iy. Since
1044 * this function utilizes thie @ubi->peb1_buf buffer which is shared
1045 * with some other functions, so lock the buffer by taking the
1048 mutex_lock(&ubi->buf_mutex);
1049 dbg_eba("read %d bytes of data", aldata_size);
1050 err = ubi_io_read_data(ubi, ubi->peb_buf1, from, 0, aldata_size);
1051 if (err && err != UBI_IO_BITFLIPS) {
1052 ubi_warn("error %d while reading data from PEB %d",
1054 goto out_unlock_buf;
1058 * Now we have got to calculate how much data we have to to copy. In
1059 * case of a static volume it is fairly easy - the VID header contains
1060 * the data size. In case of a dynamic volume it is more difficult - we
1061 * have to read the contents, cut 0xFF bytes from the end and copy only
1062 * the first part. We must do this to avoid writing 0xFF bytes as it
1063 * may have some side-effects. And not only this. It is important not
1064 * to include those 0xFFs to CRC because later the they may be filled
1067 if (vid_hdr->vol_type == UBI_VID_DYNAMIC)
1068 aldata_size = data_size =
1069 ubi_calc_data_len(ubi, ubi->peb_buf1, data_size);
1072 crc = crc32(UBI_CRC32_INIT, ubi->peb_buf1, data_size);
1076 * It may turn out to me that the whole @from physical eraseblock
1077 * contains only 0xFF bytes. Then we have to only write the VID header
1078 * and do not write any data. This also means we should not set
1079 * @vid_hdr->copy_flag, @vid_hdr->data_size, and @vid_hdr->data_crc.
1081 if (data_size > 0) {
1082 vid_hdr->copy_flag = 1;
1083 vid_hdr->data_size = cpu_to_be32(data_size);
1084 vid_hdr->data_crc = cpu_to_be32(crc);
1086 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
1088 err = ubi_io_write_vid_hdr(ubi, to, vid_hdr);
1090 goto out_unlock_buf;
1094 /* Read the VID header back and check if it was written correctly */
1095 err = ubi_io_read_vid_hdr(ubi, to, vid_hdr, 1);
1097 if (err != UBI_IO_BITFLIPS)
1098 ubi_warn("cannot read VID header back from PEB %d", to);
1101 goto out_unlock_buf;
1104 if (data_size > 0) {
1105 err = ubi_io_write_data(ubi, ubi->peb_buf1, to, 0, aldata_size);
1107 goto out_unlock_buf;
1112 * We've written the data and are going to read it back to make
1113 * sure it was written correctly.
1116 err = ubi_io_read_data(ubi, ubi->peb_buf2, to, 0, aldata_size);
1118 if (err != UBI_IO_BITFLIPS)
1119 ubi_warn("cannot read data back from PEB %d",
1123 goto out_unlock_buf;
1128 if (memcmp(ubi->peb_buf1, ubi->peb_buf2, aldata_size)) {
1129 ubi_warn("read data back from PEB %d - it is different",
1131 goto out_unlock_buf;
1135 ubi_assert(vol->eba_tbl[lnum] == from);
1136 vol->eba_tbl[lnum] = to;
1139 mutex_unlock(&ubi->buf_mutex);
1141 leb_write_unlock(ubi, vol_id, lnum);
1146 * ubi_eba_init_scan - initialize the EBA unit using scanning information.
1147 * @ubi: UBI device description object
1148 * @si: scanning information
1150 * This function returns zero in case of success and a negative error code in
1153 int ubi_eba_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
1155 int i, j, err, num_volumes;
1156 struct ubi_scan_volume *sv;
1157 struct ubi_volume *vol;
1158 struct ubi_scan_leb *seb;
1161 dbg_eba("initialize EBA unit");
1163 spin_lock_init(&ubi->ltree_lock);
1164 mutex_init(&ubi->alc_mutex);
1165 ubi->ltree = RB_ROOT;
1167 ubi->global_sqnum = si->max_sqnum + 1;
1168 num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
1170 for (i = 0; i < num_volumes; i++) {
1171 vol = ubi->volumes[i];
1177 vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int),
1179 if (!vol->eba_tbl) {
1184 for (j = 0; j < vol->reserved_pebs; j++)
1185 vol->eba_tbl[j] = UBI_LEB_UNMAPPED;
1187 sv = ubi_scan_find_sv(si, idx2vol_id(ubi, i));
1191 ubi_rb_for_each_entry(rb, seb, &sv->root, u.rb) {
1192 if (seb->lnum >= vol->reserved_pebs)
1194 * This may happen in case of an unclean reboot
1197 ubi_scan_move_to_list(sv, seb, &si->erase);
1198 vol->eba_tbl[seb->lnum] = seb->pnum;
1202 if (ubi->avail_pebs < EBA_RESERVED_PEBS) {
1203 ubi_err("no enough physical eraseblocks (%d, need %d)",
1204 ubi->avail_pebs, EBA_RESERVED_PEBS);
1208 ubi->avail_pebs -= EBA_RESERVED_PEBS;
1209 ubi->rsvd_pebs += EBA_RESERVED_PEBS;
1211 if (ubi->bad_allowed) {
1212 ubi_calculate_reserved(ubi);
1214 if (ubi->avail_pebs < ubi->beb_rsvd_level) {
1215 /* No enough free physical eraseblocks */
1216 ubi->beb_rsvd_pebs = ubi->avail_pebs;
1217 ubi_warn("cannot reserve enough PEBs for bad PEB "
1218 "handling, reserved %d, need %d",
1219 ubi->beb_rsvd_pebs, ubi->beb_rsvd_level);
1221 ubi->beb_rsvd_pebs = ubi->beb_rsvd_level;
1223 ubi->avail_pebs -= ubi->beb_rsvd_pebs;
1224 ubi->rsvd_pebs += ubi->beb_rsvd_pebs;
1227 dbg_eba("EBA unit is initialized");
1231 for (i = 0; i < num_volumes; i++) {
1232 if (!ubi->volumes[i])
1234 kfree(ubi->volumes[i]->eba_tbl);
1240 * ubi_eba_close - close EBA unit.
1241 * @ubi: UBI device description object
1243 void ubi_eba_close(const struct ubi_device *ubi)
1245 int i, num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
1247 dbg_eba("close EBA unit");
1249 for (i = 0; i < num_volumes; i++) {
1250 if (!ubi->volumes[i])
1252 kfree(ubi->volumes[i]->eba_tbl);