2 * Copyright (c) International Business Machines Corp., 2006
3 * Copyright (c) Nokia Corporation, 2006, 2007
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
13 * the GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19 * Author: Artem Bityutskiy (Битюцкий Артём)
23 * This file includes volume table manipulation code. The volume table is an
24 * on-flash table containing volume meta-data like name, number of reserved
25 * physical eraseblocks, type, etc. The volume table is stored in the so-called
28 * The layout volume is an internal volume which is organized as follows. It
29 * consists of two logical eraseblocks - LEB 0 and LEB 1. Each logical
30 * eraseblock stores one volume table copy, i.e. LEB 0 and LEB 1 duplicate each
31 * other. This redundancy guarantees robustness to unclean reboots. The volume
32 * table is basically an array of volume table records. Each record contains
33 * full information about the volume and protected by a CRC checksum.
35 * The volume table is changed, it is first changed in RAM. Then LEB 0 is
36 * erased, and the updated volume table is written back to LEB 0. Then same for
37 * LEB 1. This scheme guarantees recoverability from unclean reboots.
39 * In this UBI implementation the on-flash volume table does not contain any
40 * information about how many data static volumes contain. This information may
41 * be found from the scanning data.
43 * But it would still be beneficial to store this information in the volume
44 * table. For example, suppose we have a static volume X, and all its physical
45 * eraseblocks became bad for some reasons. Suppose we are attaching the
46 * corresponding MTD device, the scanning has found no logical eraseblocks
47 * corresponding to the volume X. According to the volume table volume X does
48 * exist. So we don't know whether it is just empty or all its physical
49 * eraseblocks went bad. So we cannot alarm the user about this corruption.
51 * The volume table also stores so-called "update marker", which is used for
52 * volume updates. Before updating the volume, the update marker is set, and
53 * after the update operation is finished, the update marker is cleared. So if
54 * the update operation was interrupted (e.g. by an unclean reboot) - the
55 * update marker is still there and we know that the volume's contents is
59 #include <linux/crc32.h>
60 #include <linux/err.h>
61 #include <asm/div64.h>
64 #ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
65 static void paranoid_vtbl_check(const struct ubi_device *ubi);
67 #define paranoid_vtbl_check(ubi)
70 /* Empty volume table record */
71 static struct ubi_vtbl_record empty_vtbl_record;
74 * ubi_change_vtbl_record - change volume table record.
75 * @ubi: UBI device description object
76 * @idx: table index to change
77 * @vtbl_rec: new volume table record
79 * This function changes volume table record @idx. If @vtbl_rec is %NULL, empty
80 * volume table record is written. The caller does not have to calculate CRC of
81 * the record as it is done by this function. Returns zero in case of success
82 * and a negative error code in case of failure.
84 int ubi_change_vtbl_record(struct ubi_device *ubi, int idx,
85 struct ubi_vtbl_record *vtbl_rec)
90 ubi_assert(idx >= 0 && idx < ubi->vtbl_slots);
93 vtbl_rec = &empty_vtbl_record;
95 crc = crc32(UBI_CRC32_INIT, vtbl_rec, UBI_VTBL_RECORD_SIZE_CRC);
96 vtbl_rec->crc = cpu_to_be32(crc);
99 mutex_lock(&ubi->vtbl_mutex);
100 memcpy(&ubi->vtbl[idx], vtbl_rec, sizeof(struct ubi_vtbl_record));
101 for (i = 0; i < UBI_LAYOUT_VOLUME_EBS; i++) {
102 err = ubi_eba_unmap_leb(ubi, UBI_LAYOUT_VOL_ID, i);
104 mutex_unlock(&ubi->vtbl_mutex);
107 err = ubi_eba_write_leb(ubi, UBI_LAYOUT_VOL_ID, i, ubi->vtbl, 0,
108 ubi->vtbl_size, UBI_LONGTERM);
110 mutex_unlock(&ubi->vtbl_mutex);
115 paranoid_vtbl_check(ubi);
116 mutex_unlock(&ubi->vtbl_mutex);
117 return ubi_wl_flush(ubi);
121 * vol_til_check - check if volume table is not corrupted and contains sensible
124 * @ubi: UBI device description object
125 * @vtbl: volume table
127 * This function returns zero if @vtbl is all right, %1 if CRC is incorrect,
128 * and %-EINVAL if it contains inconsistent data.
130 static int vtbl_check(const struct ubi_device *ubi,
131 const struct ubi_vtbl_record *vtbl)
133 int i, n, reserved_pebs, alignment, data_pad, vol_type, name_len;
138 for (i = 0; i < ubi->vtbl_slots; i++) {
141 reserved_pebs = be32_to_cpu(vtbl[i].reserved_pebs);
142 alignment = be32_to_cpu(vtbl[i].alignment);
143 data_pad = be32_to_cpu(vtbl[i].data_pad);
144 upd_marker = vtbl[i].upd_marker;
145 vol_type = vtbl[i].vol_type;
146 name_len = be16_to_cpu(vtbl[i].name_len);
147 name = &vtbl[i].name[0];
149 crc = crc32(UBI_CRC32_INIT, &vtbl[i], UBI_VTBL_RECORD_SIZE_CRC);
150 if (be32_to_cpu(vtbl[i].crc) != crc) {
151 ubi_err("bad CRC at record %u: %#08x, not %#08x",
152 i, crc, be32_to_cpu(vtbl[i].crc));
153 ubi_dbg_dump_vtbl_record(&vtbl[i], i);
157 if (reserved_pebs == 0) {
158 if (memcmp(&vtbl[i], &empty_vtbl_record,
159 UBI_VTBL_RECORD_SIZE)) {
160 dbg_err("bad empty record");
166 if (reserved_pebs < 0 || alignment < 0 || data_pad < 0 ||
168 dbg_err("negative values");
172 if (alignment > ubi->leb_size || alignment == 0) {
173 dbg_err("bad alignment");
177 n = alignment % ubi->min_io_size;
178 if (alignment != 1 && n) {
179 dbg_err("alignment is not multiple of min I/O unit");
183 n = ubi->leb_size % alignment;
185 dbg_err("bad data_pad, has to be %d", n);
189 if (vol_type != UBI_VID_DYNAMIC && vol_type != UBI_VID_STATIC) {
190 dbg_err("bad vol_type");
194 if (upd_marker != 0 && upd_marker != 1) {
195 dbg_err("bad upd_marker");
199 if (reserved_pebs > ubi->good_peb_count) {
200 dbg_err("too large reserved_pebs, good PEBs %d",
201 ubi->good_peb_count);
205 if (name_len > UBI_VOL_NAME_MAX) {
206 dbg_err("too long volume name, max %d",
211 if (name[0] == '\0') {
212 dbg_err("NULL volume name");
216 if (name_len != strnlen(name, name_len + 1)) {
217 dbg_err("bad name_len");
222 /* Checks that all names are unique */
223 for (i = 0; i < ubi->vtbl_slots - 1; i++) {
224 for (n = i + 1; n < ubi->vtbl_slots; n++) {
225 int len1 = be16_to_cpu(vtbl[i].name_len);
226 int len2 = be16_to_cpu(vtbl[n].name_len);
228 if (len1 > 0 && len1 == len2 &&
229 !strncmp(vtbl[i].name, vtbl[n].name, len1)) {
230 ubi_err("volumes %d and %d have the same name"
231 " \"%s\"", i, n, vtbl[i].name);
232 ubi_dbg_dump_vtbl_record(&vtbl[i], i);
233 ubi_dbg_dump_vtbl_record(&vtbl[n], n);
242 ubi_err("volume table check failed, record %d", i);
243 ubi_dbg_dump_vtbl_record(&vtbl[i], i);
248 * create_vtbl - create a copy of volume table.
249 * @ubi: UBI device description object
250 * @si: scanning information
251 * @copy: number of the volume table copy
252 * @vtbl: contents of the volume table
254 * This function returns zero in case of success and a negative error code in
257 static int create_vtbl(const struct ubi_device *ubi, struct ubi_scan_info *si,
258 int copy, void *vtbl)
261 static struct ubi_vid_hdr *vid_hdr;
262 struct ubi_scan_volume *sv;
263 struct ubi_scan_leb *new_seb, *old_seb = NULL;
265 ubi_msg("create volume table (copy #%d)", copy + 1);
267 vid_hdr = ubi_zalloc_vid_hdr(ubi);
272 * Check if there is a logical eraseblock which would have to contain
273 * this volume table copy was found during scanning. It has to be wiped
276 sv = ubi_scan_find_sv(si, UBI_LAYOUT_VOL_ID);
278 old_seb = ubi_scan_find_seb(sv, copy);
281 new_seb = ubi_scan_get_free_peb(ubi, si);
282 if (IS_ERR(new_seb)) {
283 err = PTR_ERR(new_seb);
287 vid_hdr->vol_type = UBI_VID_DYNAMIC;
288 vid_hdr->vol_id = cpu_to_be32(UBI_LAYOUT_VOL_ID);
289 vid_hdr->compat = UBI_LAYOUT_VOLUME_COMPAT;
290 vid_hdr->data_size = vid_hdr->used_ebs =
291 vid_hdr->data_pad = cpu_to_be32(0);
292 vid_hdr->lnum = cpu_to_be32(copy);
293 vid_hdr->sqnum = cpu_to_be64(++si->max_sqnum);
294 vid_hdr->leb_ver = cpu_to_be32(old_seb ? old_seb->leb_ver + 1: 0);
296 /* The EC header is already there, write the VID header */
297 err = ubi_io_write_vid_hdr(ubi, new_seb->pnum, vid_hdr);
301 /* Write the layout volume contents */
302 err = ubi_io_write_data(ubi, vtbl, new_seb->pnum, 0, ubi->vtbl_size);
307 * And add it to the scanning information. Don't delete the old
308 * @old_seb as it will be deleted and freed in 'ubi_scan_add_used()'.
310 err = ubi_scan_add_used(ubi, si, new_seb->pnum, new_seb->ec,
313 ubi_free_vid_hdr(ubi, vid_hdr);
317 if (err == -EIO && ++tries <= 5) {
319 * Probably this physical eraseblock went bad, try to pick
322 list_add_tail(&new_seb->u.list, &si->corr);
327 ubi_free_vid_hdr(ubi, vid_hdr);
333 * process_lvol - process the layout volume.
334 * @ubi: UBI device description object
335 * @si: scanning information
336 * @sv: layout volume scanning information
338 * This function is responsible for reading the layout volume, ensuring it is
339 * not corrupted, and recovering from corruptions if needed. Returns volume
340 * table in case of success and a negative error code in case of failure.
342 static struct ubi_vtbl_record *process_lvol(const struct ubi_device *ubi,
343 struct ubi_scan_info *si,
344 struct ubi_scan_volume *sv)
348 struct ubi_scan_leb *seb;
349 struct ubi_vtbl_record *leb[UBI_LAYOUT_VOLUME_EBS] = { NULL, NULL };
350 int leb_corrupted[UBI_LAYOUT_VOLUME_EBS] = {1, 1};
353 * UBI goes through the following steps when it changes the layout
356 * b. write new data to LEB 0;
358 * d. write new data to LEB 1.
360 * Before the change, both LEBs contain the same data.
362 * Due to unclean reboots, the contents of LEB 0 may be lost, but there
363 * should LEB 1. So it is OK if LEB 0 is corrupted while LEB 1 is not.
364 * Similarly, LEB 1 may be lost, but there should be LEB 0. And
365 * finally, unclean reboots may result in a situation when neither LEB
366 * 0 nor LEB 1 are corrupted, but they are different. In this case, LEB
367 * 0 contains more recent information.
369 * So the plan is to first check LEB 0. Then
370 * a. if LEB 0 is OK, it must be containing the most resent data; then
371 * we compare it with LEB 1, and if they are different, we copy LEB
373 * b. if LEB 0 is corrupted, but LEB 1 has to be OK, and we copy LEB 1
377 dbg_msg("check layout volume");
379 /* Read both LEB 0 and LEB 1 into memory */
380 ubi_rb_for_each_entry(rb, seb, &sv->root, u.rb) {
381 leb[seb->lnum] = vmalloc(ubi->vtbl_size);
382 if (!leb[seb->lnum]) {
386 memset(leb[seb->lnum], 0, ubi->vtbl_size);
388 err = ubi_io_read_data(ubi, leb[seb->lnum], seb->pnum, 0,
390 if (err == UBI_IO_BITFLIPS || err == -EBADMSG)
391 /* Scrub the PEB later */
399 leb_corrupted[0] = vtbl_check(ubi, leb[0]);
400 if (leb_corrupted[0] < 0)
404 if (!leb_corrupted[0]) {
407 leb_corrupted[1] = memcmp(leb[0], leb[1], ubi->vtbl_size);
408 if (leb_corrupted[1]) {
409 ubi_warn("volume table copy #2 is corrupted");
410 err = create_vtbl(ubi, si, 1, leb[0]);
413 ubi_msg("volume table was restored");
416 /* Both LEB 1 and LEB 2 are OK and consistent */
420 /* LEB 0 is corrupted or does not exist */
422 leb_corrupted[1] = vtbl_check(ubi, leb[1]);
423 if (leb_corrupted[1] < 0)
426 if (leb_corrupted[1]) {
427 /* Both LEB 0 and LEB 1 are corrupted */
428 ubi_err("both volume tables are corrupted");
432 ubi_warn("volume table copy #1 is corrupted");
433 err = create_vtbl(ubi, si, 0, leb[1]);
436 ubi_msg("volume table was restored");
449 * create_empty_lvol - create empty layout volume.
450 * @ubi: UBI device description object
451 * @si: scanning information
453 * This function returns volume table contents in case of success and a
454 * negative error code in case of failure.
456 static struct ubi_vtbl_record *create_empty_lvol(const struct ubi_device *ubi,
457 struct ubi_scan_info *si)
460 struct ubi_vtbl_record *vtbl;
462 vtbl = vmalloc(ubi->vtbl_size);
464 return ERR_PTR(-ENOMEM);
465 memset(vtbl, 0, ubi->vtbl_size);
467 for (i = 0; i < ubi->vtbl_slots; i++)
468 memcpy(&vtbl[i], &empty_vtbl_record, UBI_VTBL_RECORD_SIZE);
470 for (i = 0; i < UBI_LAYOUT_VOLUME_EBS; i++) {
473 err = create_vtbl(ubi, si, i, vtbl);
484 * init_volumes - initialize volume information for existing volumes.
485 * @ubi: UBI device description object
486 * @si: scanning information
487 * @vtbl: volume table
489 * This function allocates volume description objects for existing volumes.
490 * Returns zero in case of success and a negative error code in case of
493 static int init_volumes(struct ubi_device *ubi, const struct ubi_scan_info *si,
494 const struct ubi_vtbl_record *vtbl)
496 int i, reserved_pebs = 0;
497 struct ubi_scan_volume *sv;
498 struct ubi_volume *vol;
500 for (i = 0; i < ubi->vtbl_slots; i++) {
503 if (be32_to_cpu(vtbl[i].reserved_pebs) == 0)
504 continue; /* Empty record */
506 vol = kzalloc(sizeof(struct ubi_volume), GFP_KERNEL);
510 vol->reserved_pebs = be32_to_cpu(vtbl[i].reserved_pebs);
511 vol->alignment = be32_to_cpu(vtbl[i].alignment);
512 vol->data_pad = be32_to_cpu(vtbl[i].data_pad);
513 vol->vol_type = vtbl[i].vol_type == UBI_VID_DYNAMIC ?
514 UBI_DYNAMIC_VOLUME : UBI_STATIC_VOLUME;
515 vol->name_len = be16_to_cpu(vtbl[i].name_len);
516 vol->usable_leb_size = ubi->leb_size - vol->data_pad;
517 memcpy(vol->name, vtbl[i].name, vol->name_len);
518 vol->name[vol->name_len] = '\0';
521 ubi_assert(!ubi->volumes[i]);
522 ubi->volumes[i] = vol;
525 reserved_pebs += vol->reserved_pebs;
528 * In case of dynamic volume UBI knows nothing about how many
529 * data is stored there. So assume the whole volume is used.
531 if (vol->vol_type == UBI_DYNAMIC_VOLUME) {
532 vol->used_ebs = vol->reserved_pebs;
533 vol->last_eb_bytes = vol->usable_leb_size;
535 (long long)vol->used_ebs * vol->usable_leb_size;
539 /* Static volumes only */
540 sv = ubi_scan_find_sv(si, i);
543 * No eraseblocks belonging to this volume found. We
544 * don't actually know whether this static volume is
545 * completely corrupted or just contains no data. And
546 * we cannot know this as long as data size is not
547 * stored on flash. So we just assume the volume is
548 * empty. FIXME: this should be handled.
553 if (sv->leb_count != sv->used_ebs) {
555 * We found a static volume which misses several
556 * eraseblocks. Treat it as corrupted.
558 ubi_warn("static volume %d misses %d LEBs - corrupted",
559 sv->vol_id, sv->used_ebs - sv->leb_count);
564 vol->used_ebs = sv->used_ebs;
566 (long long)(vol->used_ebs - 1) * vol->usable_leb_size;
567 vol->used_bytes += sv->last_data_size;
568 vol->last_eb_bytes = sv->last_data_size;
571 vol = kzalloc(sizeof(struct ubi_volume), GFP_KERNEL);
575 vol->reserved_pebs = UBI_LAYOUT_VOLUME_EBS;
577 vol->vol_type = UBI_DYNAMIC_VOLUME;
578 vol->name_len = sizeof(UBI_LAYOUT_VOLUME_NAME) - 1;
579 memcpy(vol->name, UBI_LAYOUT_VOLUME_NAME, vol->name_len + 1);
580 vol->usable_leb_size = ubi->leb_size;
581 vol->used_ebs = vol->reserved_pebs;
582 vol->last_eb_bytes = vol->reserved_pebs;
584 (long long)vol->used_ebs * (ubi->leb_size - vol->data_pad);
585 vol->vol_id = UBI_LAYOUT_VOL_ID;
587 ubi_assert(!ubi->volumes[i]);
588 ubi->volumes[vol_id2idx(ubi, vol->vol_id)] = vol;
589 reserved_pebs += vol->reserved_pebs;
593 if (reserved_pebs > ubi->avail_pebs)
594 ubi_err("not enough PEBs, required %d, available %d",
595 reserved_pebs, ubi->avail_pebs);
596 ubi->rsvd_pebs += reserved_pebs;
597 ubi->avail_pebs -= reserved_pebs;
603 * check_sv - check volume scanning information.
604 * @vol: UBI volume description object
605 * @sv: volume scanning information
607 * This function returns zero if the volume scanning information is consistent
608 * to the data read from the volume tabla, and %-EINVAL if not.
610 static int check_sv(const struct ubi_volume *vol,
611 const struct ubi_scan_volume *sv)
613 if (sv->highest_lnum >= vol->reserved_pebs) {
614 dbg_err("bad highest_lnum");
617 if (sv->leb_count > vol->reserved_pebs) {
618 dbg_err("bad leb_count");
621 if (sv->vol_type != vol->vol_type) {
622 dbg_err("bad vol_type");
625 if (sv->used_ebs > vol->reserved_pebs) {
626 dbg_err("bad used_ebs");
629 if (sv->data_pad != vol->data_pad) {
630 dbg_err("bad data_pad");
636 ubi_err("bad scanning information");
638 ubi_dbg_dump_vol_info(vol);
643 * check_scanning_info - check that scanning information.
644 * @ubi: UBI device description object
645 * @si: scanning information
647 * Even though we protect on-flash data by CRC checksums, we still don't trust
648 * the media. This function ensures that scanning information is consistent to
649 * the information read from the volume table. Returns zero if the scanning
650 * information is OK and %-EINVAL if it is not.
652 static int check_scanning_info(const struct ubi_device *ubi,
653 struct ubi_scan_info *si)
656 struct ubi_scan_volume *sv;
657 struct ubi_volume *vol;
659 if (si->vols_found > UBI_INT_VOL_COUNT + ubi->vtbl_slots) {
660 ubi_err("scanning found %d volumes, maximum is %d + %d",
661 si->vols_found, UBI_INT_VOL_COUNT, ubi->vtbl_slots);
665 if (si->highest_vol_id >= ubi->vtbl_slots + UBI_INT_VOL_COUNT&&
666 si->highest_vol_id < UBI_INTERNAL_VOL_START) {
667 ubi_err("too large volume ID %d found by scanning",
673 for (i = 0; i < ubi->vtbl_slots + UBI_INT_VOL_COUNT; i++) {
676 sv = ubi_scan_find_sv(si, i);
677 vol = ubi->volumes[i];
680 ubi_scan_rm_volume(si, sv);
684 if (vol->reserved_pebs == 0) {
685 ubi_assert(i < ubi->vtbl_slots);
691 * During scanning we found a volume which does not
692 * exist according to the information in the volume
693 * table. This must have happened due to an unclean
694 * reboot while the volume was being removed. Discard
697 ubi_msg("finish volume %d removal", sv->vol_id);
698 ubi_scan_rm_volume(si, sv);
700 err = check_sv(vol, sv);
710 * ubi_read_volume_table - read volume table.
712 * @ubi: UBI device description object
713 * @si: scanning information
715 * This function reads volume table, checks it, recover from errors if needed,
716 * or creates it if needed. Returns zero in case of success and a negative
717 * error code in case of failure.
719 int ubi_read_volume_table(struct ubi_device *ubi, struct ubi_scan_info *si)
722 struct ubi_scan_volume *sv;
724 empty_vtbl_record.crc = cpu_to_be32(0xf116c36b);
727 * The number of supported volumes is limited by the eraseblock size
728 * and by the UBI_MAX_VOLUMES constant.
730 ubi->vtbl_slots = ubi->leb_size / UBI_VTBL_RECORD_SIZE;
731 if (ubi->vtbl_slots > UBI_MAX_VOLUMES)
732 ubi->vtbl_slots = UBI_MAX_VOLUMES;
734 ubi->vtbl_size = ubi->vtbl_slots * UBI_VTBL_RECORD_SIZE;
735 ubi->vtbl_size = ALIGN(ubi->vtbl_size, ubi->min_io_size);
737 sv = ubi_scan_find_sv(si, UBI_LAYOUT_VOL_ID);
740 * No logical eraseblocks belonging to the layout volume were
741 * found. This could mean that the flash is just empty. In
742 * this case we create empty layout volume.
744 * But if flash is not empty this must be a corruption or the
745 * MTD device just contains garbage.
748 ubi->vtbl = create_empty_lvol(ubi, si);
749 if (IS_ERR(ubi->vtbl))
750 return PTR_ERR(ubi->vtbl);
752 ubi_err("the layout volume was not found");
756 if (sv->leb_count > UBI_LAYOUT_VOLUME_EBS) {
757 /* This must not happen with proper UBI images */
758 dbg_err("too many LEBs (%d) in layout volume",
763 ubi->vtbl = process_lvol(ubi, si, sv);
764 if (IS_ERR(ubi->vtbl))
765 return PTR_ERR(ubi->vtbl);
768 ubi->avail_pebs = ubi->good_peb_count;
771 * The layout volume is OK, initialize the corresponding in-RAM data
774 err = init_volumes(ubi, si, ubi->vtbl);
779 * Get sure that the scanning information is consistent to the
780 * information stored in the volume table.
782 err = check_scanning_info(ubi, si);
790 for (i = 0; i < ubi->vtbl_slots + UBI_INT_VOL_COUNT; i++)
791 if (ubi->volumes[i]) {
792 kfree(ubi->volumes[i]);
793 ubi->volumes[i] = NULL;
798 #ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
801 * paranoid_vtbl_check - check volume table.
802 * @ubi: UBI device description object
804 static void paranoid_vtbl_check(const struct ubi_device *ubi)
806 if (vtbl_check(ubi, ubi->vtbl)) {
807 ubi_err("paranoid check failed");
812 #endif /* CONFIG_MTD_UBI_DEBUG_PARANOID */