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_KERNEL
);
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
);
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
;
498 unsigned char *new_buf
;
500 vid_hdr
= ubi_zalloc_vid_hdr(ubi
);
506 new_pnum
= ubi_wl_get_peb(ubi
, UBI_UNKNOWN
);
508 ubi_free_vid_hdr(ubi
, vid_hdr
);
512 ubi_msg("recover PEB %d, move data to PEB %d", pnum
, new_pnum
);
514 err
= ubi_io_read_vid_hdr(ubi
, pnum
, vid_hdr
, 1);
515 if (err
&& err
!= UBI_IO_BITFLIPS
) {
521 vid_hdr
->sqnum
= cpu_to_be64(next_sqnum(ubi
));
522 err
= ubi_io_write_vid_hdr(ubi
, new_pnum
, vid_hdr
);
526 data_size
= offset
+ len
;
527 new_buf
= vmalloc(data_size
);
532 memset(new_buf
+ offset
, 0xFF, len
);
534 /* Read everything before the area where the write failure happened */
536 err
= ubi_io_read_data(ubi
, new_buf
, pnum
, 0, offset
);
537 if (err
&& err
!= UBI_IO_BITFLIPS
) {
543 memcpy(new_buf
+ offset
, buf
, len
);
545 err
= ubi_io_write_data(ubi
, new_buf
, new_pnum
, 0, data_size
);
552 ubi_free_vid_hdr(ubi
, vid_hdr
);
554 vol
->eba_tbl
[lnum
] = new_pnum
;
555 ubi_wl_put_peb(ubi
, pnum
, 1);
557 ubi_msg("data was successfully recovered");
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 ubi_free_vid_hdr(ubi
, vid_hdr
);
576 ubi_msg("try again");
581 * ubi_eba_write_leb - write data to dynamic volume.
582 * @ubi: UBI device description object
584 * @lnum: logical eraseblock number
585 * @buf: the data to write
586 * @offset: offset within the logical eraseblock where to write
587 * @len: how many bytes to write
590 * This function writes data to logical eraseblock @lnum of a dynamic volume
591 * @vol_id. Returns zero in case of success and a negative error code in case
592 * of failure. In case of error, it is possible that something was still
593 * written to the flash media, but may be some garbage.
595 int ubi_eba_write_leb(struct ubi_device
*ubi
, int vol_id
, int lnum
,
596 const void *buf
, int offset
, int len
, int dtype
)
598 int idx
= vol_id2idx(ubi
, vol_id
), err
, pnum
, tries
= 0;
599 struct ubi_volume
*vol
= ubi
->volumes
[idx
];
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
, offset
, len
);
622 leb_write_unlock(ubi
, vol_id
, lnum
);
627 * The logical eraseblock is not mapped. We have to get a free physical
628 * eraseblock and write the volume identifier header there first.
630 vid_hdr
= ubi_zalloc_vid_hdr(ubi
);
632 leb_write_unlock(ubi
, vol_id
, lnum
);
636 vid_hdr
->vol_type
= UBI_VID_DYNAMIC
;
637 vid_hdr
->sqnum
= cpu_to_be64(next_sqnum(ubi
));
638 vid_hdr
->vol_id
= cpu_to_be32(vol_id
);
639 vid_hdr
->lnum
= cpu_to_be32(lnum
);
640 vid_hdr
->compat
= ubi_get_compat(ubi
, vol_id
);
641 vid_hdr
->data_pad
= cpu_to_be32(vol
->data_pad
);
644 pnum
= ubi_wl_get_peb(ubi
, dtype
);
646 ubi_free_vid_hdr(ubi
, vid_hdr
);
647 leb_write_unlock(ubi
, vol_id
, lnum
);
651 dbg_eba("write VID hdr and %d bytes at offset %d of LEB %d:%d, PEB %d",
652 len
, offset
, vol_id
, lnum
, pnum
);
654 err
= ubi_io_write_vid_hdr(ubi
, pnum
, vid_hdr
);
656 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
661 err
= ubi_io_write_data(ubi
, buf
, pnum
, offset
, len
);
663 ubi_warn("failed to write %d bytes at offset %d of LEB %d:%d, "
664 "PEB %d", len
, offset
, vol_id
, lnum
, pnum
);
668 vol
->eba_tbl
[lnum
] = pnum
;
670 leb_write_unlock(ubi
, vol_id
, lnum
);
671 ubi_free_vid_hdr(ubi
, vid_hdr
);
675 if (err
!= -EIO
|| !ubi
->bad_allowed
) {
677 leb_write_unlock(ubi
, vol_id
, lnum
);
678 ubi_free_vid_hdr(ubi
, vid_hdr
);
683 * Fortunately, this is the first write operation to this physical
684 * eraseblock, so just put it and request a new one. We assume that if
685 * this physical eraseblock went bad, the erase code will handle that.
687 err
= ubi_wl_put_peb(ubi
, pnum
, 1);
688 if (err
|| ++tries
> UBI_IO_RETRIES
) {
690 leb_write_unlock(ubi
, vol_id
, lnum
);
691 ubi_free_vid_hdr(ubi
, vid_hdr
);
695 vid_hdr
->sqnum
= cpu_to_be64(next_sqnum(ubi
));
696 ubi_msg("try another PEB");
701 * ubi_eba_write_leb_st - write data to static volume.
702 * @ubi: UBI device description object
704 * @lnum: logical eraseblock number
705 * @buf: data to write
706 * @len: how many bytes to write
708 * @used_ebs: how many logical eraseblocks will this volume contain
710 * This function writes data to logical eraseblock @lnum of static volume
711 * @vol_id. The @used_ebs argument should contain total number of logical
712 * eraseblock in this static volume.
714 * When writing to the last logical eraseblock, the @len argument doesn't have
715 * to be aligned to the minimal I/O unit size. Instead, it has to be equivalent
716 * to the real data size, although the @buf buffer has to contain the
717 * alignment. In all other cases, @len has to be aligned.
719 * It is prohibited to write more then once to logical eraseblocks of static
720 * volumes. This function returns zero in case of success and a negative error
721 * code in case of failure.
723 int ubi_eba_write_leb_st(struct ubi_device
*ubi
, int vol_id
, int lnum
,
724 const void *buf
, int len
, int dtype
, int used_ebs
)
726 int err
, pnum
, tries
= 0, data_size
= len
;
727 int idx
= vol_id2idx(ubi
, vol_id
);
728 struct ubi_volume
*vol
= ubi
->volumes
[idx
];
729 struct ubi_vid_hdr
*vid_hdr
;
735 if (lnum
== used_ebs
- 1)
736 /* If this is the last LEB @len may be unaligned */
737 len
= ALIGN(data_size
, ubi
->min_io_size
);
739 ubi_assert(len
% ubi
->min_io_size
== 0);
741 vid_hdr
= ubi_zalloc_vid_hdr(ubi
);
745 err
= leb_write_lock(ubi
, vol_id
, lnum
);
747 ubi_free_vid_hdr(ubi
, vid_hdr
);
751 vid_hdr
->sqnum
= cpu_to_be64(next_sqnum(ubi
));
752 vid_hdr
->vol_id
= cpu_to_be32(vol_id
);
753 vid_hdr
->lnum
= cpu_to_be32(lnum
);
754 vid_hdr
->compat
= ubi_get_compat(ubi
, vol_id
);
755 vid_hdr
->data_pad
= cpu_to_be32(vol
->data_pad
);
757 crc
= crc32(UBI_CRC32_INIT
, buf
, data_size
);
758 vid_hdr
->vol_type
= UBI_VID_STATIC
;
759 vid_hdr
->data_size
= cpu_to_be32(data_size
);
760 vid_hdr
->used_ebs
= cpu_to_be32(used_ebs
);
761 vid_hdr
->data_crc
= cpu_to_be32(crc
);
764 pnum
= ubi_wl_get_peb(ubi
, dtype
);
766 ubi_free_vid_hdr(ubi
, vid_hdr
);
767 leb_write_unlock(ubi
, vol_id
, lnum
);
771 dbg_eba("write VID hdr and %d bytes at LEB %d:%d, PEB %d, used_ebs %d",
772 len
, vol_id
, lnum
, pnum
, used_ebs
);
774 err
= ubi_io_write_vid_hdr(ubi
, pnum
, vid_hdr
);
776 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
781 err
= ubi_io_write_data(ubi
, buf
, pnum
, 0, len
);
783 ubi_warn("failed to write %d bytes of data to PEB %d",
788 ubi_assert(vol
->eba_tbl
[lnum
] < 0);
789 vol
->eba_tbl
[lnum
] = pnum
;
791 leb_write_unlock(ubi
, vol_id
, lnum
);
792 ubi_free_vid_hdr(ubi
, vid_hdr
);
796 if (err
!= -EIO
|| !ubi
->bad_allowed
) {
798 * This flash device does not admit of bad eraseblocks or
799 * something nasty and unexpected happened. Switch to read-only
803 leb_write_unlock(ubi
, vol_id
, lnum
);
804 ubi_free_vid_hdr(ubi
, vid_hdr
);
808 err
= ubi_wl_put_peb(ubi
, pnum
, 1);
809 if (err
|| ++tries
> UBI_IO_RETRIES
) {
811 leb_write_unlock(ubi
, vol_id
, lnum
);
812 ubi_free_vid_hdr(ubi
, vid_hdr
);
816 vid_hdr
->sqnum
= cpu_to_be64(next_sqnum(ubi
));
817 ubi_msg("try another PEB");
822 * ubi_eba_atomic_leb_change - change logical eraseblock atomically.
823 * @ubi: UBI device description object
825 * @lnum: logical eraseblock number
826 * @buf: data to write
827 * @len: how many bytes to write
830 * This function changes the contents of a logical eraseblock atomically. @buf
831 * has to contain new logical eraseblock data, and @len - the length of the
832 * data, which has to be aligned. This function guarantees that in case of an
833 * unclean reboot the old contents is preserved. Returns zero in case of
834 * success and a negative error code in case of failure.
836 int ubi_eba_atomic_leb_change(struct ubi_device
*ubi
, int vol_id
, int lnum
,
837 const void *buf
, int len
, int dtype
)
839 int err
, pnum
, tries
= 0, idx
= vol_id2idx(ubi
, vol_id
);
840 struct ubi_volume
*vol
= ubi
->volumes
[idx
];
841 struct ubi_vid_hdr
*vid_hdr
;
847 vid_hdr
= ubi_zalloc_vid_hdr(ubi
);
851 err
= leb_write_lock(ubi
, vol_id
, lnum
);
853 ubi_free_vid_hdr(ubi
, vid_hdr
);
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
);
872 ubi_free_vid_hdr(ubi
, vid_hdr
);
873 leb_write_unlock(ubi
, vol_id
, lnum
);
877 dbg_eba("change LEB %d:%d, PEB %d, write VID hdr to PEB %d",
878 vol_id
, lnum
, vol
->eba_tbl
[lnum
], pnum
);
880 err
= ubi_io_write_vid_hdr(ubi
, pnum
, vid_hdr
);
882 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
887 err
= ubi_io_write_data(ubi
, buf
, pnum
, 0, len
);
889 ubi_warn("failed to write %d bytes of data to PEB %d",
894 if (vol
->eba_tbl
[lnum
] >= 0) {
895 err
= ubi_wl_put_peb(ubi
, vol
->eba_tbl
[lnum
], 1);
897 ubi_free_vid_hdr(ubi
, vid_hdr
);
898 leb_write_unlock(ubi
, vol_id
, lnum
);
903 vol
->eba_tbl
[lnum
] = pnum
;
904 leb_write_unlock(ubi
, vol_id
, lnum
);
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
916 leb_write_unlock(ubi
, vol_id
, lnum
);
917 ubi_free_vid_hdr(ubi
, vid_hdr
);
921 err
= ubi_wl_put_peb(ubi
, pnum
, 1);
922 if (err
|| ++tries
> UBI_IO_RETRIES
) {
924 leb_write_unlock(ubi
, vol_id
, lnum
);
925 ubi_free_vid_hdr(ubi
, vid_hdr
);
929 vid_hdr
->sqnum
= cpu_to_be64(next_sqnum(ubi
));
930 ubi_msg("try another PEB");
935 * ltree_entry_ctor - lock tree entries slab cache constructor.
936 * @obj: the lock-tree entry to construct
937 * @cache: the lock tree entry slab cache
938 * @flags: constructor flags
940 static void ltree_entry_ctor(void *obj
, struct kmem_cache
*cache
,
943 struct ltree_entry
*le
= obj
;
946 init_rwsem(&le
->mutex
);
950 * ubi_eba_copy_leb - copy logical eraseblock.
951 * @ubi: UBI device description object
952 * @from: physical eraseblock number from where to copy
953 * @to: physical eraseblock number where to copy
954 * @vid_hdr: VID header of the @from physical eraseblock
956 * This function copies logical eraseblock from physical eraseblock @from to
957 * physical eraseblock @to. The @vid_hdr buffer may be changed by this
958 * function. Returns zero in case of success, %UBI_IO_BITFLIPS if the operation
959 * was canceled because bit-flips were detected at the target PEB, and a
960 * negative error code in case of failure.
962 int ubi_eba_copy_leb(struct ubi_device
*ubi
, int from
, int to
,
963 struct ubi_vid_hdr
*vid_hdr
)
965 int err
, vol_id
, lnum
, data_size
, aldata_size
, pnum
, idx
;
966 struct ubi_volume
*vol
;
968 void *buf
, *buf1
= NULL
;
970 vol_id
= be32_to_cpu(vid_hdr
->vol_id
);
971 lnum
= be32_to_cpu(vid_hdr
->lnum
);
973 dbg_eba("copy LEB %d:%d, PEB %d to PEB %d", vol_id
, lnum
, from
, to
);
975 if (vid_hdr
->vol_type
== UBI_VID_STATIC
) {
976 data_size
= be32_to_cpu(vid_hdr
->data_size
);
977 aldata_size
= ALIGN(data_size
, ubi
->min_io_size
);
979 data_size
= aldata_size
=
980 ubi
->leb_size
- be32_to_cpu(vid_hdr
->data_pad
);
982 buf
= vmalloc(aldata_size
);
987 * We do not want anybody to write to this logical eraseblock while we
988 * are moving it, so we lock it.
990 err
= leb_write_lock(ubi
, vol_id
, lnum
);
997 * But the logical eraseblock might have been put by this time.
998 * Cancel if it is true.
1000 idx
= vol_id2idx(ubi
, vol_id
);
1003 * We may race with volume deletion/re-size, so we have to hold
1004 * @ubi->volumes_lock.
1006 spin_lock(&ubi
->volumes_lock
);
1007 vol
= ubi
->volumes
[idx
];
1009 dbg_eba("volume %d was removed meanwhile", vol_id
);
1010 spin_unlock(&ubi
->volumes_lock
);
1014 pnum
= vol
->eba_tbl
[lnum
];
1016 dbg_eba("LEB %d:%d is no longer mapped to PEB %d, mapped to "
1017 "PEB %d, cancel", vol_id
, lnum
, from
, pnum
);
1018 spin_unlock(&ubi
->volumes_lock
);
1021 spin_unlock(&ubi
->volumes_lock
);
1023 /* OK, now the LEB is locked and we can safely start moving it */
1025 dbg_eba("read %d bytes of data", aldata_size
);
1026 err
= ubi_io_read_data(ubi
, buf
, from
, 0, aldata_size
);
1027 if (err
&& err
!= UBI_IO_BITFLIPS
) {
1028 ubi_warn("error %d while reading data from PEB %d",
1034 * Now we have got to calculate how much data we have to to copy. In
1035 * case of a static volume it is fairly easy - the VID header contains
1036 * the data size. In case of a dynamic volume it is more difficult - we
1037 * have to read the contents, cut 0xFF bytes from the end and copy only
1038 * the first part. We must do this to avoid writing 0xFF bytes as it
1039 * may have some side-effects. And not only this. It is important not
1040 * to include those 0xFFs to CRC because later the they may be filled
1043 if (vid_hdr
->vol_type
== UBI_VID_DYNAMIC
)
1044 aldata_size
= data_size
=
1045 ubi_calc_data_len(ubi
, buf
, data_size
);
1048 crc
= crc32(UBI_CRC32_INIT
, buf
, data_size
);
1052 * It may turn out to me that the whole @from physical eraseblock
1053 * contains only 0xFF bytes. Then we have to only write the VID header
1054 * and do not write any data. This also means we should not set
1055 * @vid_hdr->copy_flag, @vid_hdr->data_size, and @vid_hdr->data_crc.
1057 if (data_size
> 0) {
1058 vid_hdr
->copy_flag
= 1;
1059 vid_hdr
->data_size
= cpu_to_be32(data_size
);
1060 vid_hdr
->data_crc
= cpu_to_be32(crc
);
1062 vid_hdr
->sqnum
= cpu_to_be64(next_sqnum(ubi
));
1064 err
= ubi_io_write_vid_hdr(ubi
, to
, vid_hdr
);
1070 /* Read the VID header back and check if it was written correctly */
1071 err
= ubi_io_read_vid_hdr(ubi
, to
, vid_hdr
, 1);
1073 if (err
!= UBI_IO_BITFLIPS
)
1074 ubi_warn("cannot read VID header back from PEB %d", to
);
1078 if (data_size
> 0) {
1079 err
= ubi_io_write_data(ubi
, buf
, to
, 0, aldata_size
);
1084 * We've written the data and are going to read it back to make
1085 * sure it was written correctly.
1087 buf1
= vmalloc(aldata_size
);
1095 err
= ubi_io_read_data(ubi
, buf1
, to
, 0, aldata_size
);
1097 if (err
!= UBI_IO_BITFLIPS
)
1098 ubi_warn("cannot read data back from PEB %d",
1105 if (memcmp(buf
, buf1
, aldata_size
)) {
1106 ubi_warn("read data back from PEB %d - it is different",
1112 ubi_assert(vol
->eba_tbl
[lnum
] == from
);
1113 vol
->eba_tbl
[lnum
] = to
;
1115 leb_write_unlock(ubi
, vol_id
, lnum
);
1122 leb_write_unlock(ubi
, vol_id
, lnum
);
1129 * ubi_eba_init_scan - initialize the EBA unit using scanning information.
1130 * @ubi: UBI device description object
1131 * @si: scanning information
1133 * This function returns zero in case of success and a negative error code in
1136 int ubi_eba_init_scan(struct ubi_device
*ubi
, struct ubi_scan_info
*si
)
1138 int i
, j
, err
, num_volumes
;
1139 struct ubi_scan_volume
*sv
;
1140 struct ubi_volume
*vol
;
1141 struct ubi_scan_leb
*seb
;
1144 dbg_eba("initialize EBA unit");
1146 spin_lock_init(&ubi
->ltree_lock
);
1147 ubi
->ltree
= RB_ROOT
;
1149 if (ubi_devices_cnt
== 0) {
1150 ltree_slab
= kmem_cache_create("ubi_ltree_slab",
1151 sizeof(struct ltree_entry
), 0,
1152 0, <ree_entry_ctor
, NULL
);
1157 ubi
->global_sqnum
= si
->max_sqnum
+ 1;
1158 num_volumes
= ubi
->vtbl_slots
+ UBI_INT_VOL_COUNT
;
1160 for (i
= 0; i
< num_volumes
; i
++) {
1161 vol
= ubi
->volumes
[i
];
1167 vol
->eba_tbl
= kmalloc(vol
->reserved_pebs
* sizeof(int),
1169 if (!vol
->eba_tbl
) {
1174 for (j
= 0; j
< vol
->reserved_pebs
; j
++)
1175 vol
->eba_tbl
[j
] = UBI_LEB_UNMAPPED
;
1177 sv
= ubi_scan_find_sv(si
, idx2vol_id(ubi
, i
));
1181 ubi_rb_for_each_entry(rb
, seb
, &sv
->root
, u
.rb
) {
1182 if (seb
->lnum
>= vol
->reserved_pebs
)
1184 * This may happen in case of an unclean reboot
1187 ubi_scan_move_to_list(sv
, seb
, &si
->erase
);
1188 vol
->eba_tbl
[seb
->lnum
] = seb
->pnum
;
1192 if (ubi
->bad_allowed
) {
1193 ubi_calculate_reserved(ubi
);
1195 if (ubi
->avail_pebs
< ubi
->beb_rsvd_level
) {
1196 /* No enough free physical eraseblocks */
1197 ubi
->beb_rsvd_pebs
= ubi
->avail_pebs
;
1198 ubi_warn("cannot reserve enough PEBs for bad PEB "
1199 "handling, reserved %d, need %d",
1200 ubi
->beb_rsvd_pebs
, ubi
->beb_rsvd_level
);
1202 ubi
->beb_rsvd_pebs
= ubi
->beb_rsvd_level
;
1204 ubi
->avail_pebs
-= ubi
->beb_rsvd_pebs
;
1205 ubi
->rsvd_pebs
+= ubi
->beb_rsvd_pebs
;
1208 dbg_eba("EBA unit is initialized");
1212 for (i
= 0; i
< num_volumes
; i
++) {
1213 if (!ubi
->volumes
[i
])
1215 kfree(ubi
->volumes
[i
]->eba_tbl
);
1217 if (ubi_devices_cnt
== 0)
1218 kmem_cache_destroy(ltree_slab
);
1223 * ubi_eba_close - close EBA unit.
1224 * @ubi: UBI device description object
1226 void ubi_eba_close(const struct ubi_device
*ubi
)
1228 int i
, num_volumes
= ubi
->vtbl_slots
+ UBI_INT_VOL_COUNT
;
1230 dbg_eba("close EBA unit");
1232 for (i
= 0; i
< num_volumes
; i
++) {
1233 if (!ubi
->volumes
[i
])
1235 kfree(ubi
->volumes
[i
]->eba_tbl
);
1237 if (ubi_devices_cnt
== 1)
1238 kmem_cache_destroy(ltree_slab
);