Lines Matching +full:mode +full:- +full:recovery
1 // SPDX-License-Identifier: GPL-2.0-only
3 * Copyright (C) 2010-2011 Neil Brown
4 * Copyright (C) 2010-2018 Red Hat, Inc. All rights reserved.
16 #include "md-bitmap.h"
18 #include <linux/device-mapper.h>
21 #define MAX_RAID_DEVICES 253 /* md-raid kernel limit */
36 * The following flags are used by dm-raid to set up the array state.
45 * ti->len and give more flexibility in altering size and
51 * |--------- Physical Device ---------|
52 * |- meta_dev -|------ data_dev ------|
60 * Bits for establishing rs->ctr_flags
86 #define __CTR_FLAG_JOURNAL_MODE 16 /* 2 */ /* Only with raid4/5/6 (journal mode)! */
89 * Flags for rs->ctr_flags field.
171 * that both parity and q-syndrome are being written properly with
200 * Flags for rs->runtime_flags field
219 #define DISKS_ARRAY_ELEMS ((MAX_RAID_DEVICES + (sizeof(uint64_t) * 8 - 1)) / sizeof(uint64_t) / 8)
255 int mode;
263 struct mddev *mddev = &rs->md;
265 l->new_level = mddev->new_level;
266 l->new_layout = mddev->new_layout;
267 l->new_chunk_sectors = mddev->new_chunk_sectors;
272 struct mddev *mddev = &rs->md;
274 mddev->new_level = l->new_level;
275 mddev->new_layout = l->new_layout;
276 mddev->new_chunk_sectors = l->new_chunk_sectors;
352 while (anf-- > __arg_name_flags)
353 if (flag & anf->flag)
354 return anf->name;
362 /* Define correlation of raid456 journal cache modes and dm-raid target line parameters */
364 const int mode;
371 /* Return MD raid4/5/6 journal mode for dm @journal_mode one */
372 static int dm_raid_journal_mode_to_md(const char *mode)
376 while (m--)
377 if (!strcasecmp(mode, _raid456_journal_mode[m].param))
378 return _raid456_journal_mode[m].mode;
380 return -EINVAL;
383 /* Return dm-raid raid4/5/6 journal mode string for @mode */
384 static const char *md_journal_mode_to_dm_raid(const int mode)
388 while (m--)
389 if (mode == _raid456_journal_mode[m].mode)
403 return !rs->md.level;
409 return rs->md.level == 1;
415 return rs->md.level == 10;
421 return rs->md.level == 6;
427 return __within_range(rs->md.level, 4, 6);
435 (rs_is_raid10(rs) && !__is_raid10_far(rs->md.new_layout));
441 return rs->md.recovery_cp < rs->md.dev_sectors;
447 return rs->md.reshape_position != MaxSector;
457 return !rt->level;
463 return rt->level == 1;
469 return rt->level == 10;
475 return __within_range(rt->level, 4, 5);
481 return rt->level == 6;
487 return __within_range(rt->level, 4, 6);
494 if (rt_is_raid0(rs->raid_type))
496 else if (rt_is_raid1(rs->raid_type))
498 else if (rt_is_raid10(rs->raid_type))
500 else if (rt_is_raid45(rs->raid_type))
502 else if (rt_is_raid6(rs->raid_type))
515 if (rs->ctr_flags & ~__valid_flags(rs)) {
516 rs->ti->error = "Invalid flags combination";
517 return -EINVAL;
590 return -EINVAL;
612 * -> disable it for now in the definition above
621 if (!test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags))
626 if (!test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags))
630 return -EINVAL;
639 if (rtp->level == 10) {
640 switch (rtp->algorithm) {
661 while (rtp-- > raid_types)
662 if (!strcasecmp(rtp->name, name))
673 while (rtp-- > raid_types) {
675 if (rtp->level == level &&
676 (__got_raid10(rtp, layout) || rtp->algorithm == layout))
686 struct mddev *mddev = &rs->md;
690 * raid10 sets rdev->sector to the device size, which
691 * is unintended in case of out-of-place reshaping
694 if (!test_bit(Journal, &rdev->flags))
695 rdev->sectors = mddev->dev_sectors;
703 struct gendisk *gendisk = dm_disk(dm_table_get_md(rs->ti->table));
705 set_capacity_and_notify(gendisk, rs->md.array_sectors);
714 struct mddev *mddev = &rs->md;
716 mddev->new_level = mddev->level;
717 mddev->new_layout = mddev->layout;
718 mddev->new_chunk_sectors = mddev->chunk_sectors;
727 struct mddev *mddev = &rs->md;
729 mddev->level = mddev->new_level;
730 mddev->layout = mddev->new_layout;
731 mddev->chunk_sectors = mddev->new_chunk_sectors;
732 mddev->raid_disks = rs->raid_disks;
733 mddev->delta_disks = 0;
742 if (raid_devs <= raid_type->parity_devs) {
743 ti->error = "Insufficient number of devices";
744 return ERR_PTR(-EINVAL);
749 ti->error = "Cannot allocate raid context";
750 return ERR_PTR(-ENOMEM);
753 if (mddev_init(&rs->md)) {
755 ti->error = "Cannot initialize raid context";
756 return ERR_PTR(-ENOMEM);
759 rs->raid_disks = raid_devs;
760 rs->delta_disks = 0;
762 rs->ti = ti;
763 rs->raid_type = raid_type;
764 rs->stripe_cache_entries = 256;
765 rs->md.raid_disks = raid_devs;
766 rs->md.level = raid_type->level;
767 rs->md.new_level = rs->md.level;
768 rs->md.layout = raid_type->algorithm;
769 rs->md.new_layout = rs->md.layout;
770 rs->md.delta_disks = 0;
771 rs->md.recovery_cp = MaxSector;
774 md_rdev_init(&rs->dev[i].rdev);
778 * rs->md.persistent
779 * rs->md.external
780 * rs->md.chunk_sectors
781 * rs->md.new_chunk_sectors
782 * rs->md.dev_sectors
793 if (rs->journal_dev.dev) {
794 md_rdev_clear(&rs->journal_dev.rdev);
795 dm_put_device(rs->ti, rs->journal_dev.dev);
798 for (i = 0; i < rs->raid_disks; i++) {
799 if (rs->dev[i].meta_dev)
800 dm_put_device(rs->ti, rs->dev[i].meta_dev);
801 md_rdev_clear(&rs->dev[i].rdev);
802 if (rs->dev[i].data_dev)
803 dm_put_device(rs->ti, rs->dev[i].data_dev);
806 mddev_destroy(&rs->md);
812 * <meta_dev>: meta device name or '-' if missing
813 * <data_dev>: data device name or '-' if missing
816 * - -
817 * - <data_dev>
821 * <meta_dev> -
837 return -EINVAL;
839 for (i = 0; i < rs->raid_disks; i++) {
840 rs->dev[i].rdev.raid_disk = i;
842 rs->dev[i].meta_dev = NULL;
843 rs->dev[i].data_dev = NULL;
849 rs->dev[i].rdev.data_offset = 0;
850 rs->dev[i].rdev.new_data_offset = 0;
851 rs->dev[i].rdev.mddev = &rs->md;
855 return -EINVAL;
857 if (strcmp(arg, "-")) {
858 r = dm_get_device(rs->ti, arg, dm_table_get_mode(rs->ti->table),
859 &rs->dev[i].meta_dev);
861 rs->ti->error = "RAID metadata device lookup failure";
865 rs->dev[i].rdev.sb_page = alloc_page(GFP_KERNEL);
866 if (!rs->dev[i].rdev.sb_page) {
867 rs->ti->error = "Failed to allocate superblock page";
868 return -ENOMEM;
874 return -EINVAL;
876 if (!strcmp(arg, "-")) {
877 if (!test_bit(In_sync, &rs->dev[i].rdev.flags) &&
878 (!rs->dev[i].rdev.recovery_offset)) {
879 rs->ti->error = "Drive designated for rebuild not specified";
880 return -EINVAL;
883 if (rs->dev[i].meta_dev) {
884 rs->ti->error = "No data device supplied with metadata device";
885 return -EINVAL;
891 r = dm_get_device(rs->ti, arg, dm_table_get_mode(rs->ti->table),
892 &rs->dev[i].data_dev);
894 rs->ti->error = "RAID device lookup failure";
898 if (rs->dev[i].meta_dev) {
900 rs->dev[i].rdev.meta_bdev = rs->dev[i].meta_dev->bdev;
902 rs->dev[i].rdev.bdev = rs->dev[i].data_dev->bdev;
903 list_add_tail(&rs->dev[i].rdev.same_set, &rs->md.disks);
904 if (!test_bit(In_sync, &rs->dev[i].rdev.flags))
908 if (rs->journal_dev.dev)
909 list_add_tail(&rs->journal_dev.rdev.same_set, &rs->md.disks);
912 rs->md.external = 0;
913 rs->md.persistent = 1;
914 rs->md.major_version = 2;
915 } else if (rebuild && !rs->md.recovery_cp) {
918 * is in-sync or not - we must assume it is not. Therefore,
921 * Even if there is metadata, the on-disk information may
922 * indicate that the array is not in-sync and it will then
927 rs->ti->error = "Unable to rebuild drive while array is not in-sync";
928 return -EINVAL;
939 * Set rs->md.bitmap_info.chunksize (which really refers to 'region size').
940 * Ensure that (ti->len/region_size < 2^21) - required by MD bitmap.
942 * Returns: 0 on success, -EINVAL on failure.
946 unsigned long min_region_size = rs->ti->len / (1 << 21);
966 * Validate user-supplied value.
968 if (region_size > rs->ti->len) {
969 rs->ti->error = "Supplied region size is too large";
970 return -EINVAL;
976 rs->ti->error = "Supplied region size is too small";
977 return -EINVAL;
981 rs->ti->error = "Region size is not a power of 2";
982 return -EINVAL;
985 if (region_size < rs->md.chunk_sectors) {
986 rs->ti->error = "Region size is smaller than the chunk size";
987 return -EINVAL;
994 rs->md.bitmap_info.chunksize = to_bytes(region_size);
1006 * Returns: 0 on success, -EINVAL on failure.
1014 for (i = 0; i < rs->raid_disks; i++)
1015 if (!test_bit(FirstUse, &rs->dev[i].rdev.flags) &&
1016 ((!test_bit(In_sync, &rs->dev[i].rdev.flags) ||
1017 !rs->dev[i].rdev.sb_page)))
1020 switch (rs->md.level) {
1024 if (rebuild_cnt >= rs->md.raid_disks)
1030 if (rebuild_cnt > rs->raid_type->parity_devs)
1034 copies = raid10_md_layout_to_copies(rs->md.new_layout);
1037 return -EINVAL;
1057 raid_disks = min(rs->raid_disks, rs->md.raid_disks);
1058 if (__is_raid10_near(rs->md.new_layout)) {
1062 if ((!rs->dev[i].rdev.sb_page ||
1063 !test_bit(In_sync, &rs->dev[i].rdev.flags)) &&
1083 last_group_start = (raid_disks / group_size) - 1;
1088 if ((!rs->dev[i].rdev.sb_page ||
1089 !test_bit(In_sync, &rs->dev[i].rdev.flags)) &&
1096 return -EINVAL;
1102 return -EINVAL;
1112 * [[no]sync] Force or prevent recovery of the
1120 * [max_write_behind <sectors>] See '-write-behind=' (man mdadm)
1126 * RAID10-only options:
1140 struct raid_type *rt = rs->raid_type;
1143 num_raid_params--; /* Account for chunk_size argument */
1146 rs->ti->error = "Bad numerical argument given for chunk_size";
1147 return -EINVAL;
1151 * First, parse the in-order required arguments
1159 rs->ti->error = "Chunk size must be a power of 2";
1160 return -EINVAL;
1162 rs->ti->error = "Chunk size value is too small";
1163 return -EINVAL;
1166 rs->md.new_chunk_sectors = rs->md.chunk_sectors = value;
1174 * - Device is reset when param is read.
1176 * - No matching superblock found, resets device.
1178 * - Failure noticed, resets device for bitmap replay.
1179 * 4) Device hadn't completed recovery after previous failure.
1180 * - Superblock is read and overrides recovery_offset.
1185 for (i = 0; i < rs->raid_disks; i++) {
1186 set_bit(In_sync, &rs->dev[i].rdev.flags);
1187 rs->dev[i].rdev.recovery_offset = MaxSector;
1196 rs->ti->error = "Not enough raid parameters given";
1197 return -EINVAL;
1201 if (test_and_set_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) {
1202 rs->ti->error = "Only one 'nosync' argument allowed";
1203 return -EINVAL;
1208 if (test_and_set_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) {
1209 rs->ti->error = "Only one 'sync' argument allowed";
1210 return -EINVAL;
1215 if (test_and_set_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) {
1216 rs->ti->error = "Only one 'raid10_use_new_sets' argument allowed";
1217 return -EINVAL;
1225 rs->ti->error = "Wrong number of raid parameters given";
1226 return -EINVAL;
1234 if (test_and_set_bit(__CTR_FLAG_RAID10_FORMAT, &rs->ctr_flags)) {
1235 rs->ti->error = "Only one 'raid10_format' argument pair allowed";
1236 return -EINVAL;
1239 rs->ti->error = "'raid10_format' is an invalid parameter for this RAID type";
1240 return -EINVAL;
1244 rs->ti->error = "Invalid 'raid10_format' value given";
1255 if (test_and_set_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags)) {
1256 rs->ti->error = "Only one raid4/5/6 set journaling device allowed";
1257 return -EINVAL;
1260 rs->ti->error = "'journal_dev' is an invalid parameter for this RAID type";
1261 return -EINVAL;
1263 r = dm_get_device(rs->ti, arg, dm_table_get_mode(rs->ti->table),
1264 &rs->journal_dev.dev);
1266 rs->ti->error = "raid4/5/6 journal device lookup failure";
1269 jdev = &rs->journal_dev.rdev;
1271 jdev->mddev = &rs->md;
1272 jdev->bdev = rs->journal_dev.dev->bdev;
1273 jdev->sectors = bdev_nr_sectors(jdev->bdev);
1274 if (jdev->sectors < MIN_RAID456_JOURNAL_SPACE) {
1275 rs->ti->error = "No space for raid4/5/6 journal";
1276 return -ENOSPC;
1278 rs->journal_dev.mode = R5C_JOURNAL_MODE_WRITE_THROUGH;
1279 set_bit(Journal, &jdev->flags);
1283 /* "journal_mode <mode>" ("journal_dev" mandatory!) */
1287 if (!test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags)) {
1288 rs->ti->error = "raid4/5/6 'journal_mode' is invalid without 'journal_dev'";
1289 return -EINVAL;
1291 if (test_and_set_bit(__CTR_FLAG_JOURNAL_MODE, &rs->ctr_flags)) {
1292 rs->ti->error = "Only one raid4/5/6 'journal_mode' argument allowed";
1293 return -EINVAL;
1297 rs->ti->error = "Invalid 'journal_mode' argument";
1300 rs->journal_dev.mode = r;
1308 rs->ti->error = "Bad numerical argument given in raid params";
1309 return -EINVAL;
1318 if (!__within_range(value, 0, rs->raid_disks - 1)) {
1319 rs->ti->error = "Invalid rebuild index given";
1320 return -EINVAL;
1323 if (test_and_set_bit(value, (void *) rs->rebuild_disks)) {
1324 rs->ti->error = "rebuild for this index already given";
1325 return -EINVAL;
1328 rd = rs->dev + value;
1329 clear_bit(In_sync, &rd->rdev.flags);
1330 clear_bit(Faulty, &rd->rdev.flags);
1331 rd->rdev.recovery_offset = 0;
1332 set_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags);
1335 rs->ti->error = "write_mostly option is only valid for RAID1";
1336 return -EINVAL;
1339 if (!__within_range(value, 0, rs->md.raid_disks - 1)) {
1340 rs->ti->error = "Invalid write_mostly index given";
1341 return -EINVAL;
1345 set_bit(WriteMostly, &rs->dev[value].rdev.flags);
1346 set_bit(__CTR_FLAG_WRITE_MOSTLY, &rs->ctr_flags);
1349 rs->ti->error = "max_write_behind option is only valid for RAID1";
1350 return -EINVAL;
1353 if (test_and_set_bit(__CTR_FLAG_MAX_WRITE_BEHIND, &rs->ctr_flags)) {
1354 rs->ti->error = "Only one max_write_behind argument pair allowed";
1355 return -EINVAL;
1359 * In device-mapper, we specify things in sectors, but
1363 rs->ti->error = "Max write-behind limit out of range";
1364 return -EINVAL;
1367 rs->md.bitmap_info.max_write_behind = value / 2;
1369 if (test_and_set_bit(__CTR_FLAG_DAEMON_SLEEP, &rs->ctr_flags)) {
1370 rs->ti->error = "Only one daemon_sleep argument pair allowed";
1371 return -EINVAL;
1374 rs->ti->error = "daemon sleep period out of range";
1375 return -EINVAL;
1377 rs->md.bitmap_info.daemon_sleep = value;
1380 if (test_and_set_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) {
1381 rs->ti->error = "Only one data_offset argument pair allowed";
1382 return -EINVAL;
1387 rs->ti->error = "Bogus data_offset value";
1388 return -EINVAL;
1390 rs->data_offset = value;
1392 /* Define the +/-# of disks to add to/remove from the given raid set */
1393 if (test_and_set_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) {
1394 rs->ti->error = "Only one delta_disks argument pair allowed";
1395 return -EINVAL;
1398 if (!__within_range(abs(value), 1, MAX_RAID_DEVICES - rt->minimal_devs)) {
1399 rs->ti->error = "Too many delta_disk requested";
1400 return -EINVAL;
1403 rs->delta_disks = value;
1405 if (test_and_set_bit(__CTR_FLAG_STRIPE_CACHE, &rs->ctr_flags)) {
1406 rs->ti->error = "Only one stripe_cache argument pair allowed";
1407 return -EINVAL;
1411 rs->ti->error = "Inappropriate argument: stripe_cache";
1412 return -EINVAL;
1416 rs->ti->error = "Bogus stripe cache entries value";
1417 return -EINVAL;
1419 rs->stripe_cache_entries = value;
1421 if (test_and_set_bit(__CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags)) {
1422 rs->ti->error = "Only one min_recovery_rate argument pair allowed";
1423 return -EINVAL;
1427 rs->ti->error = "min_recovery_rate out of range";
1428 return -EINVAL;
1430 rs->md.sync_speed_min = value;
1432 if (test_and_set_bit(__CTR_FLAG_MAX_RECOVERY_RATE, &rs->ctr_flags)) {
1433 rs->ti->error = "Only one max_recovery_rate argument pair allowed";
1434 return -EINVAL;
1438 rs->ti->error = "max_recovery_rate out of range";
1439 return -EINVAL;
1441 rs->md.sync_speed_max = value;
1443 if (test_and_set_bit(__CTR_FLAG_REGION_SIZE, &rs->ctr_flags)) {
1444 rs->ti->error = "Only one region_size argument pair allowed";
1445 return -EINVAL;
1449 rs->requested_bitmap_chunk_sectors = value;
1451 if (test_and_set_bit(__CTR_FLAG_RAID10_COPIES, &rs->ctr_flags)) {
1452 rs->ti->error = "Only one raid10_copies argument pair allowed";
1453 return -EINVAL;
1456 if (!__within_range(value, 2, rs->md.raid_disks)) {
1457 rs->ti->error = "Bad value for 'raid10_copies'";
1458 return -EINVAL;
1464 rs->ti->error = "Unable to parse RAID parameter";
1465 return -EINVAL;
1469 if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags) &&
1470 test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) {
1471 rs->ti->error = "sync and nosync are mutually exclusive";
1472 return -EINVAL;
1475 if (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags) &&
1476 (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags) ||
1477 test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags))) {
1478 rs->ti->error = "sync/nosync and rebuild are mutually exclusive";
1479 return -EINVAL;
1482 if (write_mostly >= rs->md.raid_disks) {
1483 rs->ti->error = "Can't set all raid1 devices to write_mostly";
1484 return -EINVAL;
1487 if (rs->md.sync_speed_max &&
1488 rs->md.sync_speed_min > rs->md.sync_speed_max) {
1489 rs->ti->error = "Bogus recovery rates";
1490 return -EINVAL;
1494 return -EINVAL;
1496 if (rs->md.chunk_sectors)
1497 max_io_len = rs->md.chunk_sectors;
1501 if (dm_set_target_max_io_len(rs->ti, max_io_len))
1502 return -EINVAL;
1505 if (raid10_copies > rs->md.raid_disks) {
1506 rs->ti->error = "Not enough devices to satisfy specification";
1507 return -EINVAL;
1510 rs->md.new_layout = raid10_format_to_md_layout(rs, raid10_format, raid10_copies);
1511 if (rs->md.new_layout < 0) {
1512 rs->ti->error = "Error getting raid10 format";
1513 return rs->md.new_layout;
1516 rt = get_raid_type_by_ll(10, rs->md.new_layout);
1518 rs->ti->error = "Failed to recognize new raid10 layout";
1519 return -EINVAL;
1522 if ((rt->algorithm == ALGORITHM_RAID10_DEFAULT ||
1523 rt->algorithm == ALGORITHM_RAID10_NEAR) &&
1524 test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) {
1525 rs->ti->error = "RAID10 format 'near' and 'raid10_use_near_sets' are incompatible";
1526 return -EINVAL;
1530 rs->raid10_copies = raid10_copies;
1533 rs->md.persistent = 0;
1534 rs->md.external = 1;
1545 struct mddev *mddev = &rs->md;
1546 uint32_t min_stripes = max(mddev->chunk_sectors, mddev->new_chunk_sectors) / 2;
1547 uint32_t nr_stripes = rs->stripe_cache_entries;
1549 if (!rt_is_raid456(rs->raid_type)) {
1550 rs->ti->error = "Inappropriate raid level; cannot change stripe_cache size";
1551 return -EINVAL;
1560 conf = mddev->private;
1562 rs->ti->error = "Cannot change stripe_cache size on inactive RAID set";
1563 return -EINVAL;
1567 if (conf->min_nr_stripes != nr_stripes) {
1570 rs->ti->error = "Failed to set raid4/5/6 stripe cache size";
1583 return rs->md.raid_disks - rs->raid_type->parity_devs;
1589 return rs->raid_disks - rs->raid_type->parity_devs;
1593 * Retrieve rdev->sectors from any valid raid device of @rs
1594 * to allow userpace to pass in arbitray "- -" device tupples.
1600 for (i = 0; i < rs->raid_disks; i++) {
1601 struct md_rdev *rdev = &rs->dev[i].rdev;
1603 if (!test_bit(Journal, &rdev->flags) &&
1604 rdev->bdev && rdev->sectors)
1605 return rdev->sectors;
1617 rdev_for_each(rdev, &rs->md)
1618 if (!test_bit(Journal, &rdev->flags) && rdev->bdev) {
1619 ds = min(ds, bdev_nr_sectors(rdev->bdev));
1620 if (ds < rs->md.dev_sectors) {
1621 rs->ti->error = "Component device(s) too small";
1622 return -EINVAL;
1635 rdev_for_each(rdev, &rs->md)
1636 if (!test_bit(Journal, &rdev->flags)) {
1637 reshape_sectors = (rdev->data_offset > rdev->new_data_offset) ?
1638 rdev->data_offset - rdev->new_data_offset :
1639 rdev->new_data_offset - rdev->data_offset;
1643 return max(reshape_sectors, (sector_t) rs->data_offset);
1652 struct mddev *mddev = &rs->md;
1655 delta_disks = mddev->delta_disks;
1658 delta_disks = rs->delta_disks;
1663 if (rt_is_raid1(rs->raid_type))
1665 else if (rt_is_raid10(rs->raid_type)) {
1666 if (rs->raid10_copies < 2 ||
1668 rs->ti->error = "Bogus raid10 data copies or delta disks";
1669 return -EINVAL;
1672 dev_sectors *= rs->raid10_copies;
1676 array_sectors = (data_stripes + delta_disks) * (dev_sectors - _get_reshape_sectors(rs));
1677 if (sector_div(array_sectors, rs->raid10_copies))
1685 array_sectors = (data_stripes + delta_disks) * (dev_sectors - _get_reshape_sectors(rs));
1687 mddev->array_sectors = array_sectors;
1688 mddev->dev_sectors = dev_sectors;
1693 rs->ti->error = "Target length not divisible by number of data devices";
1694 return -EINVAL;
1697 /* Setup recovery on @rs */
1702 rs->md.recovery_cp = MaxSector;
1706 * ensure proper parity and Q-Syndrome
1709 rs->md.recovery_cp = dev_sectors;
1711 * Other raid set types may skip recovery
1715 rs->md.recovery_cp = test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)
1727 rs->array_sectors > rs->md.array_sectors &&
1728 !rs->md.delta_disks &&
1729 rs->md.raid_disks == rs->raid_disks) {
1730 /* The raid10 personality doesn't provide proper device sizes -> correct. */
1734 rs->md.array_sectors = rs->array_sectors;
1738 dm_table_event(rs->ti->table);
1749 struct mddev *mddev = &rs->md;
1752 if (rs->md.degraded) {
1753 rs->ti->error = "Can't takeover degraded raid set";
1754 return -EPERM;
1758 rs->ti->error = "Can't takeover reshaping raid set";
1759 return -EPERM;
1762 switch (mddev->level) {
1764 /* raid0 -> raid1/5 with one disk */
1765 if ((mddev->new_level == 1 || mddev->new_level == 5) &&
1766 mddev->raid_disks == 1)
1769 /* raid0 -> raid10 */
1770 if (mddev->new_level == 10 &&
1771 !(rs->raid_disks % mddev->raid_disks))
1774 /* raid0 with multiple disks -> raid4/5/6 */
1775 if (__within_range(mddev->new_level, 4, 6) &&
1776 mddev->new_layout == ALGORITHM_PARITY_N &&
1777 mddev->raid_disks > 1)
1784 if (__is_raid10_offset(mddev->layout))
1787 near_copies = __raid10_near_copies(mddev->layout);
1789 /* raid10* -> raid0 */
1790 if (mddev->new_level == 0) {
1793 !(mddev->raid_disks % near_copies)) {
1794 mddev->raid_disks /= near_copies;
1795 mddev->delta_disks = mddev->raid_disks;
1801 __raid10_far_copies(mddev->layout) > 1)
1807 /* raid10_{near,far} -> raid1 */
1808 if (mddev->new_level == 1 &&
1809 max(near_copies, __raid10_far_copies(mddev->layout)) == mddev->raid_disks)
1812 /* raid10_{near,far} with 2 disks -> raid4/5 */
1813 if (__within_range(mddev->new_level, 4, 5) &&
1814 mddev->raid_disks == 2)
1819 /* raid1 with 2 disks -> raid4/5 */
1820 if (__within_range(mddev->new_level, 4, 5) &&
1821 mddev->raid_disks == 2) {
1822 mddev->degraded = 1;
1826 /* raid1 -> raid0 */
1827 if (mddev->new_level == 0 &&
1828 mddev->raid_disks == 1)
1831 /* raid1 -> raid10 */
1832 if (mddev->new_level == 10)
1837 /* raid4 -> raid0 */
1838 if (mddev->new_level == 0)
1841 /* raid4 -> raid1/5 with 2 disks */
1842 if ((mddev->new_level == 1 || mddev->new_level == 5) &&
1843 mddev->raid_disks == 2)
1846 /* raid4 -> raid5/6 with parity N */
1847 if (__within_range(mddev->new_level, 5, 6) &&
1848 mddev->layout == ALGORITHM_PARITY_N)
1853 /* raid5 with parity N -> raid0 */
1854 if (mddev->new_level == 0 &&
1855 mddev->layout == ALGORITHM_PARITY_N)
1858 /* raid5 with parity N -> raid4 */
1859 if (mddev->new_level == 4 &&
1860 mddev->layout == ALGORITHM_PARITY_N)
1863 /* raid5 with 2 disks -> raid1/4/10 */
1864 if ((mddev->new_level == 1 || mddev->new_level == 4 || mddev->new_level == 10) &&
1865 mddev->raid_disks == 2)
1868 /* raid5_* -> raid6_*_6 with Q-Syndrome N (e.g. raid5_ra -> raid6_ra_6 */
1869 if (mddev->new_level == 6 &&
1870 ((mddev->layout == ALGORITHM_PARITY_N && mddev->new_layout == ALGORITHM_PARITY_N) ||
1871 __within_range(mddev->new_layout, ALGORITHM_LEFT_ASYMMETRIC_6, ALGORITHM_RIGHT_SYMMETRIC_6)))
1876 /* raid6 with parity N -> raid0 */
1877 if (mddev->new_level == 0 &&
1878 mddev->layout == ALGORITHM_PARITY_N)
1881 /* raid6 with parity N -> raid4 */
1882 if (mddev->new_level == 4 &&
1883 mddev->layout == ALGORITHM_PARITY_N)
1886 /* raid6_*_n with Q-Syndrome N -> raid5_* */
1887 if (mddev->new_level == 5 &&
1888 ((mddev->layout == ALGORITHM_PARITY_N && mddev->new_layout == ALGORITHM_PARITY_N) ||
1889 __within_range(mddev->new_layout, ALGORITHM_LEFT_ASYMMETRIC, ALGORITHM_RIGHT_SYMMETRIC)))
1897 rs->ti->error = "takeover not possible";
1898 return -EINVAL;
1904 return rs->md.new_level != rs->md.level;
1910 return (use_mddev ? rs->md.delta_disks : rs->delta_disks) ||
1911 rs->md.new_layout != rs->md.layout ||
1912 rs->md.new_chunk_sectors != rs->md.chunk_sectors;
1919 struct mddev *mddev = &rs->md;
1931 if (rs->delta_disks)
1932 return !!rs->delta_disks;
1935 mddev->raid_disks != rs->raid_disks;
1940 !__is_raid10_far(mddev->new_layout) &&
1941 rs->delta_disks >= 0;
1949 /* State flags for sb->flags */
1955 * Devices with this superblock should only ever be accessed via device-mapper.
2025 * up to 256 devices with the 1.9.0 on-disk metadata format
2027 __le64 extended_failed_devices[DISKS_ARRAY_ELEMS - 1];
2037 * - reshape function non-existent
2038 * - degraded set
2039 * - ongoing recovery
2040 * - ongoing reshape
2042 * Returns 0 if none or -EPERM if given constraint
2047 struct mddev *mddev = &rs->md;
2049 if (!mddev->pers || !mddev->pers->check_reshape)
2050 rs->ti->error = "Reshape not supported";
2051 else if (mddev->degraded)
2052 rs->ti->error = "Can't reshape degraded raid set";
2054 rs->ti->error = "Convert request on recovering raid set prohibited";
2056 rs->ti->error = "raid set already reshaping!";
2058 rs->ti->error = "Reshaping only supported for raid1/4/5/6/10";
2062 return -EPERM;
2067 BUG_ON(!rdev->sb_page);
2069 if (rdev->sb_loaded && !force_reload)
2072 rdev->sb_loaded = 0;
2074 if (!sync_page_io(rdev, 0, size, rdev->sb_page, REQ_OP_READ, true)) {
2076 rdev->raid_disk);
2077 md_error(rdev->mddev, rdev);
2078 set_bit(Faulty, &rdev->flags);
2079 return -EIO;
2082 rdev->sb_loaded = 1;
2089 failed_devices[0] = le64_to_cpu(sb->failed_devices);
2090 memset(failed_devices + 1, 0, sizeof(sb->extended_failed_devices));
2092 if (le32_to_cpu(sb->compat_features) & FEATURE_FLAG_SUPPORTS_V190) {
2093 int i = ARRAY_SIZE(sb->extended_failed_devices);
2095 while (i--)
2096 failed_devices[i+1] = le64_to_cpu(sb->extended_failed_devices[i]);
2102 int i = ARRAY_SIZE(sb->extended_failed_devices);
2104 sb->failed_devices = cpu_to_le64(failed_devices[0]);
2105 while (i--)
2106 sb->extended_failed_devices[i] = cpu_to_le64(failed_devices[i+1]);
2123 if (!rdev->meta_bdev)
2126 BUG_ON(!rdev->sb_page);
2128 sb = page_address(rdev->sb_page);
2132 for (i = 0; i < rs->raid_disks; i++)
2133 if (!rs->dev[i].data_dev || test_bit(Faulty, &rs->dev[i].rdev.flags)) {
2141 sb->magic = cpu_to_le32(DM_RAID_MAGIC);
2142 sb->compat_features = cpu_to_le32(FEATURE_FLAG_SUPPORTS_V190);
2144 sb->num_devices = cpu_to_le32(mddev->raid_disks);
2145 sb->array_position = cpu_to_le32(rdev->raid_disk);
2147 sb->events = cpu_to_le64(mddev->events);
2149 sb->disk_recovery_offset = cpu_to_le64(rdev->recovery_offset);
2150 sb->array_resync_offset = cpu_to_le64(mddev->recovery_cp);
2152 sb->level = cpu_to_le32(mddev->level);
2153 sb->layout = cpu_to_le32(mddev->layout);
2154 sb->stripe_sectors = cpu_to_le32(mddev->chunk_sectors);
2161 sb->new_level = cpu_to_le32(mddev->new_level);
2162 sb->new_layout = cpu_to_le32(mddev->new_layout);
2163 sb->new_stripe_sectors = cpu_to_le32(mddev->new_chunk_sectors);
2165 sb->delta_disks = cpu_to_le32(mddev->delta_disks);
2168 sb->reshape_position = cpu_to_le64(mddev->reshape_position);
2169 if (le64_to_cpu(sb->reshape_position) != MaxSector) {
2171 sb->flags |= cpu_to_le32(SB_FLAG_RESHAPE_ACTIVE);
2173 if (mddev->delta_disks < 0 || mddev->reshape_backwards)
2174 sb->flags |= cpu_to_le32(SB_FLAG_RESHAPE_BACKWARDS);
2177 sb->flags &= ~(cpu_to_le32(SB_FLAG_RESHAPE_ACTIVE|SB_FLAG_RESHAPE_BACKWARDS));
2180 sb->array_sectors = cpu_to_le64(mddev->array_sectors);
2181 sb->data_offset = cpu_to_le64(rdev->data_offset);
2182 sb->new_data_offset = cpu_to_le64(rdev->new_data_offset);
2183 sb->sectors = cpu_to_le64(rdev->sectors);
2184 sb->incompat_features = cpu_to_le32(0);
2187 memset(sb + 1, 0, rdev->sb_size - sizeof(*sb));
2196 * Return: 1 if use rdev, 0 if use refdev, -Exxx otherwise
2205 r = read_disk_sb(rdev, rdev->sb_size, false);
2209 sb = page_address(rdev->sb_page);
2216 if ((sb->magic != cpu_to_le32(DM_RAID_MAGIC)) ||
2217 (!test_bit(In_sync, &rdev->flags) && !rdev->recovery_offset)) {
2218 super_sync(rdev->mddev, rdev);
2220 set_bit(FirstUse, &rdev->flags);
2221 sb->compat_features = cpu_to_le32(FEATURE_FLAG_SUPPORTS_V190);
2224 set_bit(MD_SB_CHANGE_DEVS, &rdev->mddev->sb_flags);
2233 events_sb = le64_to_cpu(sb->events);
2235 refsb = page_address(refdev->sb_page);
2236 events_refsb = le64_to_cpu(refsb->events);
2244 struct mddev *mddev = &rs->md;
2252 sb = page_address(rdev->sb_page);
2253 events_sb = le64_to_cpu(sb->events);
2258 mddev->events = events_sb ? : 1;
2260 mddev->reshape_position = MaxSector;
2262 mddev->raid_disks = le32_to_cpu(sb->num_devices);
2263 mddev->level = le32_to_cpu(sb->level);
2264 mddev->layout = le32_to_cpu(sb->layout);
2265 mddev->chunk_sectors = le32_to_cpu(sb->stripe_sectors);
2271 if (le32_to_cpu(sb->compat_features) & FEATURE_FLAG_SUPPORTS_V190) {
2273 mddev->new_level = le32_to_cpu(sb->new_level);
2274 mddev->new_layout = le32_to_cpu(sb->new_layout);
2275 mddev->new_chunk_sectors = le32_to_cpu(sb->new_stripe_sectors);
2276 mddev->delta_disks = le32_to_cpu(sb->delta_disks);
2277 mddev->array_sectors = le64_to_cpu(sb->array_sectors);
2280 if (le32_to_cpu(sb->flags) & SB_FLAG_RESHAPE_ACTIVE) {
2281 if (test_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) {
2283 return -EINVAL;
2286 if (mddev->delta_disks < 0 ||
2287 (!mddev->delta_disks && (le32_to_cpu(sb->flags) & SB_FLAG_RESHAPE_BACKWARDS)))
2288 mddev->reshape_backwards = 1;
2290 mddev->reshape_backwards = 0;
2292 mddev->reshape_position = le64_to_cpu(sb->reshape_position);
2293 rs->raid_type = get_raid_type_by_ll(mddev->level, mddev->layout);
2300 struct raid_type *rt_cur = get_raid_type_by_ll(mddev->level, mddev->layout);
2301 struct raid_type *rt_new = get_raid_type_by_ll(mddev->new_level, mddev->new_layout);
2306 rt_cur->name, rt_new->name);
2309 return -EINVAL;
2312 if (mddev->layout != mddev->new_layout) {
2315 rt_cur->name, rt_new->name);
2318 le32_to_cpu(sb->layout), mddev->new_layout);
2320 if (mddev->chunk_sectors != mddev->new_chunk_sectors)
2322 mddev->chunk_sectors, mddev->new_chunk_sectors);
2323 if (rs->delta_disks)
2325 mddev->raid_disks, mddev->raid_disks + rs->delta_disks);
2328 raid10_md_layout_to_format(mddev->layout),
2329 raid10_md_layout_to_copies(mddev->layout));
2331 raid10_md_layout_to_format(mddev->new_layout),
2332 raid10_md_layout_to_copies(mddev->new_layout));
2334 return -EINVAL;
2340 if (!test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags))
2341 mddev->recovery_cp = le64_to_cpu(sb->array_resync_offset);
2346 * 1) The raid set is brand new - in which case, all of the
2350 * and the new device needs to be rebuilt - in which
2359 if (test_bit(Journal, &rdev->flags))
2362 if (test_bit(FirstUse, &r->flags))
2365 if (!test_bit(In_sync, &r->flags)) {
2367 r->raid_disk);
2370 if (test_bit(FirstUse, &r->flags))
2375 if (new_devs == rs->raid_disks || !rebuilds) {
2377 if (new_devs == rs->raid_disks) {
2379 set_bit(MD_ARRAY_FIRST_USE, &mddev->flags);
2381 new_devs != rs->delta_disks) {
2384 return -EINVAL;
2388 " a raid set with %u other first-time devices",
2390 return -EINVAL;
2395 return -EINVAL;
2396 } else if (!test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags) && rs_is_recovering(rs)) {
2397 DMERR("'rebuild' specified while raid set is not in-sync (recovery_cp=%llu)",
2398 (unsigned long long) mddev->recovery_cp);
2399 return -EINVAL;
2402 (unsigned long long) mddev->reshape_position);
2403 return -EINVAL;
2413 if (test_bit(Journal, &rdev->flags) ||
2414 !r->sb_page)
2416 sb2 = page_address(r->sb_page);
2417 sb2->failed_devices = 0;
2418 memset(sb2->extended_failed_devices, 0, sizeof(sb2->extended_failed_devices));
2421 * Check for any device re-ordering.
2423 if (!test_bit(FirstUse, &r->flags) && (r->raid_disk >= 0)) {
2424 role = le32_to_cpu(sb2->array_position);
2428 if (role != r->raid_disk) {
2429 if (rs_is_raid10(rs) && __is_raid10_near(mddev->layout)) {
2430 if (mddev->raid_disks % __raid10_near_copies(mddev->layout) ||
2431 rs->raid_disks % rs->raid10_copies) {
2432 rs->ti->error =
2434 return -EINVAL;
2437 sb2->array_position = cpu_to_le32(r->raid_disk);
2439 } else if (!(rs_is_raid10(rs) && rt_is_raid0(rs->raid_type)) &&
2440 !(rs_is_raid0(rs) && rt_is_raid10(rs->raid_type)) &&
2441 !rt_is_raid1(rs->raid_type)) {
2442 rs->ti->error = "Cannot change device positions in raid set";
2443 return -EINVAL;
2446 DMINFO("raid device #%d now at position #%d", role, r->raid_disk);
2450 * Partial recovery is performed on
2454 set_bit(Faulty, &r->flags);
2463 struct mddev *mddev = &rs->md;
2466 if (rs_is_raid0(rs) || !rdev->sb_page || rdev->raid_disk < 0)
2469 sb = page_address(rdev->sb_page);
2472 * If mddev->events is not set, we know we have not yet initialized
2475 if (!mddev->events && super_init_validation(rs, rdev))
2476 return -EINVAL;
2478 if (le32_to_cpu(sb->compat_features) &&
2479 le32_to_cpu(sb->compat_features) != FEATURE_FLAG_SUPPORTS_V190) {
2480 rs->ti->error = "Unable to assemble array: Unknown flag(s) in compatible feature flags";
2481 return -EINVAL;
2484 if (sb->incompat_features) {
2485 rs->ti->error = "Unable to assemble array: No incompatible feature flags supported yet";
2486 return -EINVAL;
2490 mddev->bitmap_info.offset = (rt_is_raid0(rs->raid_type) || rs->journal_dev.dev) ? 0 : to_sector(4096);
2491 mddev->bitmap_info.default_offset = mddev->bitmap_info.offset;
2493 if (!test_and_clear_bit(FirstUse, &rdev->flags)) {
2499 if (le32_to_cpu(sb->compat_features) & FEATURE_FLAG_SUPPORTS_V190)
2500 rdev->sectors = le64_to_cpu(sb->sectors);
2502 rdev->recovery_offset = le64_to_cpu(sb->disk_recovery_offset);
2503 if (rdev->recovery_offset == MaxSector)
2504 set_bit(In_sync, &rdev->flags);
2506 * If no reshape in progress -> we're recovering single
2507 * disk(s) and have to set the device(s) to out-of-sync
2510 clear_bit(In_sync, &rdev->flags); /* Mandatory for recovery */
2516 if (test_and_clear_bit(Faulty, &rdev->flags)) {
2517 rdev->recovery_offset = 0;
2518 clear_bit(In_sync, &rdev->flags);
2519 rdev->saved_raid_disk = rdev->raid_disk;
2522 /* Reshape support -> restore respective data offsets */
2523 rdev->data_offset = le64_to_cpu(sb->data_offset);
2524 rdev->new_data_offset = le64_to_cpu(sb->new_data_offset);
2536 struct mddev *mddev = &rs->md;
2540 if (test_bit(Journal, &rdev->flags))
2543 if (!rdev->meta_bdev)
2547 rdev->sb_start = 0;
2548 rdev->sb_size = bdev_logical_block_size(rdev->meta_bdev);
2549 if (rdev->sb_size < sizeof(struct dm_raid_superblock) || rdev->sb_size > PAGE_SIZE) {
2551 return -EINVAL;
2563 if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags))
2586 * mistakenly acting as if '- -' got passed into the constructor).
2591 rdev->raid_disk = rdev->saved_raid_disk = -1;
2603 rs->ti->error = "Unable to assemble array: Invalid superblocks";
2605 return -EINVAL;
2608 rs->ti->error = "Insufficient redundancy to activate array";
2609 return -EINVAL;
2613 if (!test_bit(Journal, &rdev->flags) &&
2616 return -EINVAL;
2634 if (!test_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) {
2642 rdev = &rs->dev[0].rdev;
2644 if (rs->delta_disks < 0) {
2648 * - before reshape: data is at offset 0 and free space
2651 * - after reshape: data is at offset rs->data_offset != 0 on each component LV
2654 new_data_offset = rs->data_offset;
2656 } else if (rs->delta_disks > 0) {
2660 * - before reshape: data is at offset rs->data_offset != 0 and
2663 * - after reshape: data is at offset 0 on each component LV
2665 data_offset = rs->data_offset;
2672 * - or - (data offset != 0)
2674 * Changing RAID layout or chunk size -> toggle offsets
2676 * - before reshape: data is at offset rs->data_offset 0 and
2678 * -or-
2679 * data is at offset rs->data_offset != 0 and
2682 * - after reshape: data is at offset 0 if it was at offset != 0
2687 data_offset = rs->data_offset ? rdev->data_offset : 0;
2688 new_data_offset = data_offset ? 0 : rs->data_offset;
2689 set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags);
2695 if (rs->data_offset &&
2696 bdev_nr_sectors(rdev->bdev) - rs->md.dev_sectors < MIN_FREE_RESHAPE_SPACE) {
2697 rs->ti->error = data_offset ? "No space for forward reshape" :
2699 return -ENOSPC;
2704 * avoid false recovery positives in the constructor.
2706 if (rs->md.recovery_cp < rs->md.dev_sectors)
2707 rs->md.recovery_cp += rs->dev[0].rdev.data_offset;
2710 rdev_for_each(rdev, &rs->md) {
2711 if (!test_bit(Journal, &rdev->flags)) {
2712 rdev->data_offset = data_offset;
2713 rdev->new_data_offset = new_data_offset;
2720 /* Userpace reordered disks -> adjust raid_disk indexes in @rs */
2726 rdev_for_each(rdev, &rs->md) {
2727 if (!test_bit(Journal, &rdev->flags)) {
2728 rdev->raid_disk = i++;
2729 rdev->saved_raid_disk = rdev->new_raid_disk = -1;
2739 struct mddev *mddev = &rs->md;
2741 unsigned int d = mddev->raid_disks = rs->raid_disks;
2742 sector_t new_data_offset = rs->dev[0].rdev.data_offset ? 0 : rs->data_offset;
2744 if (rt_is_raid10(rs->raid_type)) {
2746 /* Userpace reordered disks -> adjust raid_disk indexes */
2749 /* raid0 -> raid10_far layout */
2750 mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_FAR,
2751 rs->raid10_copies);
2753 /* raid1 -> raid10_near layout */
2754 mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_NEAR,
2755 rs->raid_disks);
2757 return -EINVAL;
2761 clear_bit(MD_ARRAY_FIRST_USE, &mddev->flags);
2762 mddev->recovery_cp = MaxSector;
2764 while (d--) {
2765 rdev = &rs->dev[d].rdev;
2767 if (test_bit(d, (void *) rs->rebuild_disks)) {
2768 clear_bit(In_sync, &rdev->flags);
2769 clear_bit(Faulty, &rdev->flags);
2770 mddev->recovery_cp = rdev->recovery_offset = 0;
2772 set_bit(MD_ARRAY_FIRST_USE, &mddev->flags);
2775 rdev->new_data_offset = new_data_offset;
2785 struct mddev *mddev = &rs->md;
2788 if (rs->raid_disks != mddev->raid_disks &&
2789 __is_raid10_near(mddev->layout) &&
2790 rs->raid10_copies &&
2791 rs->raid10_copies != __raid10_near_copies(mddev->layout)) {
2798 if (rs->raid_disks % rs->raid10_copies) {
2799 rs->ti->error = "Can't reshape raid10 mirror groups";
2800 return -EINVAL;
2803 /* Userpace reordered disks to add/remove mirrors -> adjust raid_disk indexes */
2805 mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_NEAR,
2806 rs->raid10_copies);
2807 mddev->new_layout = mddev->layout;
2816 if (rs->delta_disks) {
2818 mddev->degraded = rs->delta_disks < 0 ? -rs->delta_disks : rs->delta_disks;
2822 mddev->raid_disks = rs->raid_disks;
2826 rs->ti->error = "Called with bogus raid type";
2827 return -EINVAL;
2831 set_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags);
2832 set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags);
2833 } else if (mddev->raid_disks < rs->raid_disks)
2835 set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags);
2842 * - change raid layout
2843 * - change chunk size
2844 * - add disks
2845 * - remove disks
2852 struct mddev *mddev = &rs->md;
2855 mddev->delta_disks = rs->delta_disks;
2856 cur_raid_devs = mddev->raid_disks;
2859 if (mddev->delta_disks &&
2860 mddev->layout != mddev->new_layout) {
2861 DMINFO("Ignoring invalid layout change with delta_disks=%d", rs->delta_disks);
2862 mddev->new_layout = mddev->layout;
2868 * - in case of adding disk(s), array size has
2874 * - in case of removing disk(s), array size
2880 * - data_offset and new_data_offset are
2888 if (rs->delta_disks > 0) {
2890 for (d = cur_raid_devs; d < rs->raid_disks; d++) {
2891 rdev = &rs->dev[d].rdev;
2892 clear_bit(In_sync, &rdev->flags);
2895 * save_raid_disk needs to be -1, or recovery_offset will be set to 0
2898 rdev->saved_raid_disk = -1;
2899 rdev->raid_disk = d;
2901 rdev->sectors = mddev->dev_sectors;
2902 rdev->recovery_offset = rs_is_raid1(rs) ? 0 : MaxSector;
2905 mddev->reshape_backwards = 0; /* adding disk(s) -> forward reshape */
2908 } else if (rs->delta_disks < 0) {
2909 r = rs_set_dev_and_array_sectors(rs, rs->ti->len, true);
2910 mddev->reshape_backwards = 1; /* removing disk(s) -> backward reshape */
2915 * Reshape layout (e.g. raid5_ls -> raid5_n) and/or chunk size:
2917 * keeping number of disks and do layout change ->
2921 * - free space upfront -> reshape forward
2923 * - free space at the end -> reshape backward
2932 * add disk(s) -> begin
2933 * remove disk(s)-> end
2935 mddev->reshape_backwards = rs->dev[0].rdev.data_offset ? 0 : 1;
2942 if (!mddev->reshape_backwards)
2943 rdev_for_each(rdev, &rs->md)
2944 if (!test_bit(Journal, &rdev->flags))
2945 rdev->sectors += reshape_sectors;
2953 * changes -> reset the latter.
2959 rs->md.delta_disks = 0;
2960 rs->md.reshape_backwards = 0;
2972 struct dm_target *ti = rs->ti;
2979 for (i = 0; i < rs->raid_disks; i++) {
2980 if (!rs->dev[i].rdev.bdev ||
2981 !bdev_max_discard_sectors(rs->dev[i].rdev.bdev))
2987 DMERR("Set dm-raid.devices_handle_discard_safely=Y to override.");
2993 ti->num_discard_bios = 1;
3027 ti->error = "No arguments";
3028 return -EINVAL;
3033 ti->error = "Unrecognised raid_type";
3034 return -EINVAL;
3038 if (dm_read_arg_group(_args, &as, &num_raid_params, &ti->error))
3039 return -EINVAL;
3044 _args[1].max = (as_nrd.argc - 1) / 2;
3045 if (dm_read_arg(_args + 1, &as_nrd, &num_raid_devs, &ti->error))
3046 return -EINVAL;
3049 ti->error = "Invalid number of supplied raid devices";
3050 return -EINVAL;
3065 rs->md.sync_super = super_sync;
3073 r = rs_set_dev_and_array_sectors(rs, rs->ti->len, false);
3078 rs->array_sectors = rs->md.array_sectors;
3079 rs->dev_sectors = rs->md.dev_sectors;
3092 /* All in-core metadata now as of current superblocks after calling analyse_superblocks() */
3093 sb_array_sectors = rs->md.array_sectors;
3096 ti->error = "Invalid rdev size";
3097 r = -EINVAL;
3103 if (rs->dev_sectors != rdev_sectors) {
3104 resize = (rs->dev_sectors != rdev_sectors - reshape_sectors);
3105 if (rs->dev_sectors > rdev_sectors - reshape_sectors)
3106 set_bit(RT_FLAG_RS_GROW, &rs->runtime_flags);
3109 INIT_WORK(&rs->md.event_work, do_table_event);
3110 ti->private = rs;
3111 ti->num_flush_bios = 1;
3112 ti->needs_bio_set_dev = true;
3123 if (test_bit(MD_ARRAY_FIRST_USE, &rs->md.flags)) {
3124 /* A new raid6 set has to be recovered to ensure proper parity and Q-Syndrome */
3126 test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) {
3127 ti->error = "'nosync' not allowed for new raid6 set";
3128 r = -EINVAL;
3132 set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags);
3140 ti->error = "Can't resize a reshaping raid set";
3141 r = -EPERM;
3147 ti->error = "Can't takeover a reshaping raid set";
3148 r = -EPERM;
3153 if (test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags)) {
3154 ti->error = "Can't takeover a journaled raid4/5/6 set";
3155 r = -EPERM;
3164 * one, prohibit requesting recovery, allow the raid
3175 set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags);
3176 /* Takeover ain't recovery, so disable recovery */
3181 clear_bit(RT_FLAG_RS_GROW, &rs->runtime_flags);
3189 if (test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags)) {
3190 ti->error = "Can't reshape a journaled raid4/5/6 set";
3191 r = -EPERM;
3195 /* Out-of-place space has to be available to allow for a reshape unless raid1! */
3208 /* Reshaping ain't recovery, so disable recovery */
3214 /* May not set recovery when a device rebuild is requested */
3215 if (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags)) {
3216 clear_bit(RT_FLAG_RS_GROW, &rs->runtime_flags);
3217 set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags);
3219 } else if (test_bit(RT_FLAG_RS_GROW, &rs->runtime_flags)) {
3228 rs_setup_recovery(rs, rs->md.recovery_cp < rs->md.dev_sectors ? rs->md.recovery_cp : rs->md.dev_sectors);
3231 r = rs_set_dev_and_array_sectors(rs, rs->ti->len, false);
3235 if (sb_array_sectors > rs->array_sectors)
3236 set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags);
3249 /* Start raid set read-only and assumed clean to change in raid_resume() */
3250 rs->md.ro = 1;
3251 rs->md.in_sync = 1;
3254 mddev_suspend_and_lock_nointr(&rs->md);
3257 md_frozen_sync_thread(&rs->md);
3259 r = md_run(&rs->md);
3260 rs->md.in_sync = 0; /* Assume already marked dirty */
3262 ti->error = "Failed to run raid array";
3263 mddev_unlock(&rs->md);
3267 r = md_start(&rs->md);
3269 ti->error = "Failed to start raid array";
3273 /* If raid4/5/6 journal mode explicitly requested (only possible with journal dev) -> set it */
3274 if (test_bit(__CTR_FLAG_JOURNAL_MODE, &rs->ctr_flags)) {
3275 r = r5c_journal_mode_set(&rs->md, rs->journal_dev.mode);
3277 ti->error = "Failed to set raid4/5/6 journal mode";
3282 set_bit(RT_FLAG_RS_SUSPENDED, &rs->runtime_flags);
3292 if (test_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags)) {
3300 if (rs->md.pers->start_reshape) {
3301 r = rs->md.pers->check_reshape(&rs->md);
3303 ti->error = "Reshape check failed";
3312 mddev_unlock(&rs->md);
3316 md_stop(&rs->md);
3317 mddev_unlock(&rs->md);
3326 struct raid_set *rs = ti->private;
3328 mddev_lock_nointr(&rs->md);
3329 md_stop(&rs->md);
3330 mddev_unlock(&rs->md);
3332 if (work_pending(&rs->md.event_work))
3333 flush_work(&rs->md.event_work);
3339 struct raid_set *rs = ti->private;
3340 struct mddev *mddev = &rs->md;
3343 * If we're reshaping to add disk(s), ti->len and
3344 * mddev->array_sectors will differ during the process
3345 * (ti->len > mddev->array_sectors), so we have to requeue
3346 * bios with addresses > mddev->array_sectors here or
3350 if (unlikely(bio_has_data(bio) && bio_end_sector(bio) > mddev->array_sectors))
3374 return __within_range(state, 0, ARRAY_SIZE(sync_strs) - 1) ? sync_strs[state] : "undef";
3377 /* Return enum sync_state for @mddev derived from @recovery flags */
3378 static enum sync_state decipher_sync_action(struct mddev *mddev, unsigned long recovery)
3380 if (test_bit(MD_RECOVERY_FROZEN, &recovery))
3384 if (!test_bit(MD_RECOVERY_DONE, &recovery) &&
3385 (test_bit(MD_RECOVERY_RUNNING, &recovery) ||
3386 (!mddev->ro && test_bit(MD_RECOVERY_NEEDED, &recovery)))) {
3387 if (test_bit(MD_RECOVERY_RESHAPE, &recovery))
3390 if (test_bit(MD_RECOVERY_SYNC, &recovery)) {
3391 if (!test_bit(MD_RECOVERY_REQUESTED, &recovery))
3393 if (test_bit(MD_RECOVERY_CHECK, &recovery))
3398 if (test_bit(MD_RECOVERY_RECOVER, &recovery))
3401 if (mddev->reshape_position != MaxSector)
3414 * 'a' = Alive but not in-sync raid set component _or_ alive raid4/5/6 'write_back' journal device
3415 * 'A' = Alive and in-sync raid set component _or_ alive raid4/5/6 'write_through' journal device
3416 * '-' = Non-existing device (i.e. uspace passed '- -' into the ctr)
3420 if (!rdev->bdev)
3421 return "-";
3422 else if (test_bit(Faulty, &rdev->flags))
3424 else if (test_bit(Journal, &rdev->flags))
3425 return (rs->journal_dev.mode == R5C_JOURNAL_MODE_WRITE_THROUGH) ? "A" : "a";
3426 else if (test_bit(RT_FLAG_RS_RESYNCING, &rs->runtime_flags) ||
3427 (!test_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags) &&
3428 !test_bit(In_sync, &rdev->flags)))
3435 static sector_t rs_get_progress(struct raid_set *rs, unsigned long recovery,
3439 struct mddev *mddev = &rs->md;
3441 clear_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags);
3442 clear_bit(RT_FLAG_RS_RESYNCING, &rs->runtime_flags);
3446 set_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags);
3449 if (state == st_idle && !test_bit(MD_RECOVERY_INTR, &recovery))
3450 r = mddev->recovery_cp;
3452 r = mddev->curr_resync_completed;
3459 if (test_bit(MD_RECOVERY_RECOVER, &recovery))
3460 set_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags);
3467 * Already retrieved recovery offset from curr_resync_completed above.
3477 set_bit(RT_FLAG_RS_RESYNCING, &rs->runtime_flags);
3485 set_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags);
3487 else if (test_bit(MD_RECOVERY_NEEDED, &recovery))
3489 * We are idle and recovery is needed, prevent 'A' chars race
3490 * caused by components still set to in-sync by constructor.
3492 set_bit(RT_FLAG_RS_RESYNCING, &rs->runtime_flags);
3503 set_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags);
3505 if (!test_bit(Journal, &rdev->flags) &&
3506 !test_bit(In_sync, &rdev->flags)) {
3507 clear_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags);
3516 /* Helper to return @dev name or "-" if !@dev */
3519 return dev ? dev->name : "-";
3525 struct raid_set *rs = ti->private;
3526 struct mddev *mddev = &rs->md;
3527 struct r5conf *conf = rs_is_raid456(rs) ? mddev->private : NULL;
3528 int i, max_nr_stripes = conf ? conf->max_nr_stripes : 0;
3529 unsigned long recovery;
3540 rt = get_raid_type_by_ll(mddev->new_level, mddev->new_layout);
3544 DMEMIT("%s %d ", rt->name, mddev->raid_disks);
3549 resync_max_sectors = test_bit(RT_FLAG_RS_PRERESUMED, &rs->runtime_flags) ?
3550 mddev->resync_max_sectors : mddev->dev_sectors;
3551 recovery = rs->md.recovery;
3552 state = decipher_sync_action(mddev, recovery);
3553 progress = rs_get_progress(rs, recovery, state, resync_max_sectors);
3554 resync_mismatches = mddev->last_sync_action == ACTION_CHECK ?
3555 atomic64_read(&mddev->resync_mismatches) : 0;
3557 /* HM FIXME: do we want another state char for raid0? It shows 'D'/'A'/'-' now */
3558 for (i = 0; i < rs->raid_disks; i++)
3559 DMEMIT(__raid_dev_status(rs, &rs->dev[i].rdev));
3562 * In-sync/Reshape ratio:
3563 * The in-sync ratio shows the progress of:
3564 * - Initializing the raid set
3565 * - Rebuilding a subset of devices of the raid set
3580 * See Documentation/admin-guide/device-mapper/dm-raid.rst for
3604 DMEMIT(" %llu", (unsigned long long) rs->dev[0].rdev.data_offset);
3609 DMEMIT(" %s", test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags) ?
3610 __raid_dev_status(rs, &rs->journal_dev.rdev) : "-");
3620 for (i = 0; i < rs->raid_disks; i++) {
3621 rebuild_writemostly_count += (test_bit(i, (void *) rs->rebuild_disks) ? 2 : 0) +
3622 (test_bit(WriteMostly, &rs->dev[i].rdev.flags) ? 2 : 0);
3624 rebuild_writemostly_count -= (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags) ? 2 : 0) +
3625 (test_bit(__CTR_FLAG_WRITE_MOSTLY, &rs->ctr_flags) ? 2 : 0);
3628 hweight32(rs->ctr_flags & CTR_FLAG_OPTIONS_NO_ARGS) +
3629 hweight32(rs->ctr_flags & CTR_FLAG_OPTIONS_ONE_ARG) * 2;
3632 DMEMIT("%s %u %u", rs->raid_type->name, raid_param_cnt, mddev->new_chunk_sectors);
3633 if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags))
3635 if (test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags))
3637 if (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags))
3638 for (i = 0; i < rs->raid_disks; i++)
3639 if (test_bit(i, (void *) rs->rebuild_disks))
3641 if (test_bit(__CTR_FLAG_DAEMON_SLEEP, &rs->ctr_flags))
3643 mddev->bitmap_info.daemon_sleep);
3644 if (test_bit(__CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags))
3646 mddev->sync_speed_min);
3647 if (test_bit(__CTR_FLAG_MAX_RECOVERY_RATE, &rs->ctr_flags))
3649 mddev->sync_speed_max);
3650 if (test_bit(__CTR_FLAG_WRITE_MOSTLY, &rs->ctr_flags))
3651 for (i = 0; i < rs->raid_disks; i++)
3652 if (test_bit(WriteMostly, &rs->dev[i].rdev.flags))
3654 rs->dev[i].rdev.raid_disk);
3655 if (test_bit(__CTR_FLAG_MAX_WRITE_BEHIND, &rs->ctr_flags))
3657 mddev->bitmap_info.max_write_behind);
3658 if (test_bit(__CTR_FLAG_STRIPE_CACHE, &rs->ctr_flags))
3661 if (test_bit(__CTR_FLAG_REGION_SIZE, &rs->ctr_flags))
3663 (unsigned long long) to_sector(mddev->bitmap_info.chunksize));
3664 if (test_bit(__CTR_FLAG_RAID10_COPIES, &rs->ctr_flags))
3666 raid10_md_layout_to_copies(mddev->layout));
3667 if (test_bit(__CTR_FLAG_RAID10_FORMAT, &rs->ctr_flags))
3669 raid10_md_layout_to_format(mddev->layout));
3670 if (test_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags))
3672 max(rs->delta_disks, mddev->delta_disks));
3673 if (test_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags))
3675 (unsigned long long) rs->data_offset);
3676 if (test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags))
3678 __get_dev_name(rs->journal_dev.dev));
3679 if (test_bit(__CTR_FLAG_JOURNAL_MODE, &rs->ctr_flags))
3681 md_journal_mode_to_dm_raid(rs->journal_dev.mode));
3682 DMEMIT(" %d", rs->raid_disks);
3683 for (i = 0; i < rs->raid_disks; i++)
3684 DMEMIT(" %s %s", __get_dev_name(rs->dev[i].meta_dev),
3685 __get_dev_name(rs->dev[i].data_dev));
3689 rt = get_raid_type_by_ll(mddev->new_level, mddev->new_layout);
3693 DMEMIT_TARGET_NAME_VERSION(ti->type);
3694 DMEMIT(",raid_type=%s,raid_disks=%d", rt->name, mddev->raid_disks);
3698 recovery = rs->md.recovery;
3699 state = decipher_sync_action(mddev, recovery);
3702 for (i = 0; i < rs->raid_disks; i++) {
3704 DMEMIT(__raid_dev_status(rs, &rs->dev[i].rdev));
3709 switch (rs->journal_dev.mode) {
3731 struct raid_set *rs = ti->private;
3732 struct mddev *mddev = &rs->md;
3735 if (!mddev->pers || !mddev->pers->sync_request)
3736 return -EINVAL;
3738 if (test_bit(RT_FLAG_RS_SUSPENDED, &rs->runtime_flags) ||
3739 test_bit(RT_FLAG_RS_FROZEN, &rs->runtime_flags))
3740 return -EBUSY;
3758 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
3759 if (decipher_sync_action(mddev, mddev->recovery) != st_idle)
3760 return -EBUSY;
3764 set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
3767 set_bit(MD_RECOVERY_CHECK, &mddev->recovery);
3768 set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
3769 set_bit(MD_RECOVERY_SYNC, &mddev->recovery);
3771 set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
3772 set_bit(MD_RECOVERY_SYNC, &mddev->recovery);
3774 return -EINVAL;
3776 if (mddev->ro == 2) {
3778 * canceling read-auto mode
3780 mddev->ro = 0;
3781 if (!mddev->suspended)
3782 md_wakeup_thread(mddev->sync_thread);
3784 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
3785 if (!mddev->suspended)
3786 md_wakeup_thread(mddev->thread);
3794 struct raid_set *rs = ti->private;
3798 for (i = 0; !r && i < rs->raid_disks; i++) {
3799 if (rs->dev[i].data_dev) {
3800 r = fn(ti, rs->dev[i].data_dev,
3802 rs->md.dev_sectors, data);
3811 struct raid_set *rs = ti->private;
3812 unsigned int chunk_size_bytes = to_bytes(rs->md.chunk_sectors);
3814 limits->io_min = chunk_size_bytes;
3815 limits->io_opt = chunk_size_bytes * mddev_data_stripes(rs);
3820 struct raid_set *rs = ti->private;
3821 struct mddev *mddev = &rs->md;
3827 set_bit(RT_FLAG_RS_FROZEN, &rs->runtime_flags);
3837 if (mddev->pers && mddev->pers->prepare_suspend)
3838 mddev->pers->prepare_suspend(mddev);
3843 struct raid_set *rs = ti->private;
3845 clear_bit(RT_FLAG_RS_FROZEN, &rs->runtime_flags);
3850 struct raid_set *rs = ti->private;
3852 if (!test_and_set_bit(RT_FLAG_RS_SUSPENDED, &rs->runtime_flags)) {
3857 md_stop_writes(&rs->md);
3858 mddev_suspend(&rs->md, false);
3869 struct mddev *mddev = &rs->md;
3873 if (!mddev->pers || !mddev->pers->hot_add_disk || !mddev->pers->hot_remove_disk)
3878 for (i = 0; i < rs->raid_disks; i++) {
3879 r = &rs->dev[i].rdev;
3880 /* HM FIXME: enhance journal device recovery processing */
3881 if (test_bit(Journal, &r->flags))
3884 if (test_bit(Faulty, &r->flags) &&
3885 r->meta_bdev && !read_disk_sb(r, r->sb_size, true)) {
3888 rs->raid_type->name, i);
3896 * '>= 0' - meaning we must call this function
3899 flags = r->flags;
3900 clear_bit(In_sync, &r->flags); /* Mandatory for hot remove. */
3901 if (r->raid_disk >= 0) {
3902 if (mddev->pers->hot_remove_disk(mddev, r)) {
3904 r->flags = flags;
3908 r->raid_disk = r->saved_raid_disk = i;
3910 clear_bit(Faulty, &r->flags);
3911 clear_bit(WriteErrorSeen, &r->flags);
3913 if (mddev->pers->hot_add_disk(mddev, r)) {
3915 r->raid_disk = r->saved_raid_disk = -1;
3916 r->flags = flags;
3918 clear_bit(In_sync, &r->flags);
3919 r->recovery_offset = 0;
3930 rdev_for_each(r, &rs->md) {
3931 if (test_bit(Journal, &r->flags))
3934 sb = page_address(r->sb_page);
3951 !test_and_set_bit(RT_FLAG_RS_BITMAP_LOADED, &rs->runtime_flags)) {
3952 struct mddev *mddev = &rs->md;
3954 r = mddev->bitmap_ops->load(mddev);
3965 struct mddev *mddev = &rs->md;
3966 int ro = mddev->ro;
3968 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
3969 mddev->ro = 0;
3971 mddev->ro = ro;
3976 * (e.g. raid6_zr -> raid6_nc), changes stripe size, adds/removes
3984 struct mddev *mddev = &rs->md;
3985 struct md_personality *pers = mddev->pers;
3988 set_bit(MD_RECOVERY_WAIT, &mddev->recovery);
3997 * May as well already kick the reshape off so that * pers->start_reshape() becomes optional.
3999 r = pers->check_reshape(mddev);
4001 rs->ti->error = "pers->check_reshape() failed";
4009 if (pers->start_reshape) {
4010 r = pers->start_reshape(mddev);
4012 rs->ti->error = "pers->start_reshape() failed";
4030 struct raid_set *rs = ti->private;
4031 struct mddev *mddev = &rs->md;
4033 /* This is a resume after a suspend of the set -> it's already started. */
4034 if (test_and_set_bit(RT_FLAG_RS_PRERESUMED, &rs->runtime_flags))
4038 if (!test_bit(RT_FLAG_RS_GROW, &rs->runtime_flags) &&
4039 rs->array_sectors != mddev->array_sectors)
4048 if (test_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags))
4057 if (test_bit(RT_FLAG_RS_GROW, &rs->runtime_flags)) {
4058 mddev->array_sectors = rs->array_sectors;
4059 mddev->dev_sectors = rs->dev_sectors;
4065 if (test_bit(RT_FLAG_RS_BITMAP_LOADED, &rs->runtime_flags) && mddev->bitmap &&
4066 (test_bit(RT_FLAG_RS_GROW, &rs->runtime_flags) ||
4067 (rs->requested_bitmap_chunk_sectors &&
4068 mddev->bitmap_info.chunksize != to_bytes(rs->requested_bitmap_chunk_sectors)))) {
4069 int chunksize = to_bytes(rs->requested_bitmap_chunk_sectors) ?: mddev->bitmap_info.chunksize;
4071 r = mddev->bitmap_ops->resize(mddev, mddev->dev_sectors,
4078 if (mddev->recovery_cp && mddev->recovery_cp < MaxSector) {
4079 set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
4080 mddev->resync_min = mddev->recovery_cp;
4081 if (test_bit(RT_FLAG_RS_GROW, &rs->runtime_flags))
4082 mddev->resync_max_sectors = mddev->dev_sectors;
4086 if (test_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags)) {
4102 struct raid_set *rs = ti->private;
4103 struct mddev *mddev = &rs->md;
4105 if (test_and_set_bit(RT_FLAG_RS_RESUMED, &rs->runtime_flags)) {
4116 if (test_and_clear_bit(RT_FLAG_RS_SUSPENDED, &rs->runtime_flags)) {
4118 if (mddev->delta_disks < 0)
4122 WARN_ON_ONCE(!test_bit(MD_RECOVERY_FROZEN, &mddev->recovery));
4123 WARN_ON_ONCE(rcu_dereference_protected(mddev->sync_thread,
4124 lockdep_is_held(&mddev->reconfig_mutex)));
4125 clear_bit(RT_FLAG_RS_FROZEN, &rs->runtime_flags);
4126 mddev->ro = 0;
4127 mddev->in_sync = 0;
4157 MODULE_ALIAS("dm-raid0");
4158 MODULE_ALIAS("dm-raid1");
4159 MODULE_ALIAS("dm-raid10");
4160 MODULE_ALIAS("dm-raid4");
4161 MODULE_ALIAS("dm-raid5");
4162 MODULE_ALIAS("dm-raid6");
4163 MODULE_AUTHOR("Neil Brown <dm-devel@lists.linux.dev>");
4164 MODULE_AUTHOR("Heinz Mauelshagen <dm-devel@lists.linux.dev>");