Lines Matching +full:rs +full:-
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
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
218 #define DISKS_ARRAY_ELEMS ((MAX_RAID_DEVICES + (sizeof(uint64_t) * 8 - 1)) / sizeof(uint64_t) / 8)
260 static void rs_config_backup(struct raid_set *rs, struct rs_layout *l) in rs_config_backup() argument
262 struct mddev *mddev = &rs->md; in rs_config_backup()
264 l->new_level = mddev->new_level; in rs_config_backup()
265 l->new_layout = mddev->new_layout; in rs_config_backup()
266 l->new_chunk_sectors = mddev->new_chunk_sectors; in rs_config_backup()
269 static void rs_config_restore(struct raid_set *rs, struct rs_layout *l) in rs_config_restore() argument
271 struct mddev *mddev = &rs->md; in rs_config_restore()
273 mddev->new_level = l->new_level; in rs_config_restore()
274 mddev->new_layout = l->new_layout; in rs_config_restore()
275 mddev->new_chunk_sectors = l->new_chunk_sectors; in rs_config_restore()
351 while (anf-- > __arg_name_flags) in dm_raid_arg_name_by_flag()
352 if (flag & anf->flag) in dm_raid_arg_name_by_flag()
353 return anf->name; in dm_raid_arg_name_by_flag()
361 /* Define correlation of raid456 journal cache modes and dm-raid target line parameters */
375 while (m--) in dm_raid_journal_mode_to_md()
379 return -EINVAL; in dm_raid_journal_mode_to_md()
382 /* Return dm-raid raid4/5/6 journal mode string for @mode */
387 while (m--) in md_journal_mode_to_dm_raid()
399 /* Return true, if raid set in @rs is raid0 */
400 static bool rs_is_raid0(struct raid_set *rs) in rs_is_raid0() argument
402 return !rs->md.level; in rs_is_raid0()
405 /* Return true, if raid set in @rs is raid1 */
406 static bool rs_is_raid1(struct raid_set *rs) in rs_is_raid1() argument
408 return rs->md.level == 1; in rs_is_raid1()
411 /* Return true, if raid set in @rs is raid10 */
412 static bool rs_is_raid10(struct raid_set *rs) in rs_is_raid10() argument
414 return rs->md.level == 10; in rs_is_raid10()
417 /* Return true, if raid set in @rs is level 6 */
418 static bool rs_is_raid6(struct raid_set *rs) in rs_is_raid6() argument
420 return rs->md.level == 6; in rs_is_raid6()
423 /* Return true, if raid set in @rs is level 4, 5 or 6 */
424 static bool rs_is_raid456(struct raid_set *rs) in rs_is_raid456() argument
426 return __within_range(rs->md.level, 4, 6); in rs_is_raid456()
429 /* Return true, if raid set in @rs is reshapable */
431 static bool rs_is_reshapable(struct raid_set *rs) in rs_is_reshapable() argument
433 return rs_is_raid456(rs) || in rs_is_reshapable()
434 (rs_is_raid10(rs) && !__is_raid10_far(rs->md.new_layout)); in rs_is_reshapable()
437 /* Return true, if raid set in @rs is recovering */
438 static bool rs_is_recovering(struct raid_set *rs) in rs_is_recovering() argument
440 return rs->md.recovery_cp < rs->md.dev_sectors; in rs_is_recovering()
443 /* Return true, if raid set in @rs is reshaping */
444 static bool rs_is_reshaping(struct raid_set *rs) in rs_is_reshaping() argument
446 return rs->md.reshape_position != MaxSector; in rs_is_reshaping()
456 return !rt->level; in rt_is_raid0()
462 return rt->level == 1; in rt_is_raid1()
468 return rt->level == 10; in rt_is_raid10()
474 return __within_range(rt->level, 4, 5); in rt_is_raid45()
480 return rt->level == 6; in rt_is_raid6()
486 return __within_range(rt->level, 4, 6); in rt_is_raid456()
490 /* Return valid ctr flags for the raid level of @rs */
491 static unsigned long __valid_flags(struct raid_set *rs) in __valid_flags() argument
493 if (rt_is_raid0(rs->raid_type)) in __valid_flags()
495 else if (rt_is_raid1(rs->raid_type)) in __valid_flags()
497 else if (rt_is_raid10(rs->raid_type)) in __valid_flags()
499 else if (rt_is_raid45(rs->raid_type)) in __valid_flags()
501 else if (rt_is_raid6(rs->raid_type)) in __valid_flags()
508 * Check for valid flags set on @rs
512 static int rs_check_for_valid_flags(struct raid_set *rs) in rs_check_for_valid_flags() argument
514 if (rs->ctr_flags & ~__valid_flags(rs)) { in rs_check_for_valid_flags()
515 rs->ti->error = "Invalid flags combination"; in rs_check_for_valid_flags()
516 return -EINVAL; in rs_check_for_valid_flags()
589 return -EINVAL; in raid10_name_to_format()
599 static int raid10_format_to_md_layout(struct raid_set *rs, in raid10_format_to_md_layout() argument
611 * -> disable it for now in the definition above in raid10_format_to_md_layout()
620 if (!test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) in raid10_format_to_md_layout()
625 if (!test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) in raid10_format_to_md_layout()
629 return -EINVAL; in raid10_format_to_md_layout()
638 if (rtp->level == 10) { in __got_raid10()
639 switch (rtp->algorithm) { in __got_raid10()
660 while (rtp-- > raid_types) in get_raid_type()
661 if (!strcasecmp(rtp->name, name)) in get_raid_type()
672 while (rtp-- > raid_types) { in get_raid_type_by_ll()
674 if (rtp->level == level && in get_raid_type_by_ll()
675 (__got_raid10(rtp, layout) || rtp->algorithm == layout)) in get_raid_type_by_ll()
683 static void rs_set_rdev_sectors(struct raid_set *rs) in rs_set_rdev_sectors() argument
685 struct mddev *mddev = &rs->md; in rs_set_rdev_sectors()
689 * raid10 sets rdev->sector to the device size, which in rs_set_rdev_sectors()
690 * is unintended in case of out-of-place reshaping in rs_set_rdev_sectors()
693 if (!test_bit(Journal, &rdev->flags)) in rs_set_rdev_sectors()
694 rdev->sectors = mddev->dev_sectors; in rs_set_rdev_sectors()
698 * Change bdev capacity of @rs in case of a disk add/remove reshape
700 static void rs_set_capacity(struct raid_set *rs) in rs_set_capacity() argument
702 struct gendisk *gendisk = dm_disk(dm_table_get_md(rs->ti->table)); in rs_set_capacity()
704 set_capacity_and_notify(gendisk, rs->md.array_sectors); in rs_set_capacity()
708 * Set the mddev properties in @rs to the current
711 static void rs_set_cur(struct raid_set *rs) in rs_set_cur() argument
713 struct mddev *mddev = &rs->md; in rs_set_cur()
715 mddev->new_level = mddev->level; in rs_set_cur()
716 mddev->new_layout = mddev->layout; in rs_set_cur()
717 mddev->new_chunk_sectors = mddev->chunk_sectors; in rs_set_cur()
721 * Set the mddev properties in @rs to the new
724 static void rs_set_new(struct raid_set *rs) in rs_set_new() argument
726 struct mddev *mddev = &rs->md; in rs_set_new()
728 mddev->level = mddev->new_level; in rs_set_new()
729 mddev->layout = mddev->new_layout; in rs_set_new()
730 mddev->chunk_sectors = mddev->new_chunk_sectors; in rs_set_new()
731 mddev->raid_disks = rs->raid_disks; in rs_set_new()
732 mddev->delta_disks = 0; in rs_set_new()
739 struct raid_set *rs; in raid_set_alloc() local
741 if (raid_devs <= raid_type->parity_devs) { in raid_set_alloc()
742 ti->error = "Insufficient number of devices"; in raid_set_alloc()
743 return ERR_PTR(-EINVAL); in raid_set_alloc()
746 rs = kzalloc(struct_size(rs, dev, raid_devs), GFP_KERNEL); in raid_set_alloc()
747 if (!rs) { in raid_set_alloc()
748 ti->error = "Cannot allocate raid context"; in raid_set_alloc()
749 return ERR_PTR(-ENOMEM); in raid_set_alloc()
752 mddev_init(&rs->md); in raid_set_alloc()
754 rs->raid_disks = raid_devs; in raid_set_alloc()
755 rs->delta_disks = 0; in raid_set_alloc()
757 rs->ti = ti; in raid_set_alloc()
758 rs->raid_type = raid_type; in raid_set_alloc()
759 rs->stripe_cache_entries = 256; in raid_set_alloc()
760 rs->md.raid_disks = raid_devs; in raid_set_alloc()
761 rs->md.level = raid_type->level; in raid_set_alloc()
762 rs->md.new_level = rs->md.level; in raid_set_alloc()
763 rs->md.layout = raid_type->algorithm; in raid_set_alloc()
764 rs->md.new_layout = rs->md.layout; in raid_set_alloc()
765 rs->md.delta_disks = 0; in raid_set_alloc()
766 rs->md.recovery_cp = MaxSector; in raid_set_alloc()
769 md_rdev_init(&rs->dev[i].rdev); in raid_set_alloc()
773 * rs->md.persistent in raid_set_alloc()
774 * rs->md.external in raid_set_alloc()
775 * rs->md.chunk_sectors in raid_set_alloc()
776 * rs->md.new_chunk_sectors in raid_set_alloc()
777 * rs->md.dev_sectors in raid_set_alloc()
780 return rs; in raid_set_alloc()
783 /* Free all @rs allocations */
784 static void raid_set_free(struct raid_set *rs) in raid_set_free() argument
788 if (rs->journal_dev.dev) { in raid_set_free()
789 md_rdev_clear(&rs->journal_dev.rdev); in raid_set_free()
790 dm_put_device(rs->ti, rs->journal_dev.dev); in raid_set_free()
793 for (i = 0; i < rs->raid_disks; i++) { in raid_set_free()
794 if (rs->dev[i].meta_dev) in raid_set_free()
795 dm_put_device(rs->ti, rs->dev[i].meta_dev); in raid_set_free()
796 md_rdev_clear(&rs->dev[i].rdev); in raid_set_free()
797 if (rs->dev[i].data_dev) in raid_set_free()
798 dm_put_device(rs->ti, rs->dev[i].data_dev); in raid_set_free()
801 kfree(rs); in raid_set_free()
806 * <meta_dev>: meta device name or '-' if missing
807 * <data_dev>: data device name or '-' if missing
810 * - -
811 * - <data_dev>
815 * <meta_dev> -
820 static int parse_dev_params(struct raid_set *rs, struct dm_arg_set *as) in parse_dev_params() argument
831 return -EINVAL; in parse_dev_params()
833 for (i = 0; i < rs->raid_disks; i++) { in parse_dev_params()
834 rs->dev[i].rdev.raid_disk = i; in parse_dev_params()
836 rs->dev[i].meta_dev = NULL; in parse_dev_params()
837 rs->dev[i].data_dev = NULL; in parse_dev_params()
843 rs->dev[i].rdev.data_offset = 0; in parse_dev_params()
844 rs->dev[i].rdev.new_data_offset = 0; in parse_dev_params()
845 rs->dev[i].rdev.mddev = &rs->md; in parse_dev_params()
849 return -EINVAL; in parse_dev_params()
851 if (strcmp(arg, "-")) { in parse_dev_params()
852 r = dm_get_device(rs->ti, arg, dm_table_get_mode(rs->ti->table), in parse_dev_params()
853 &rs->dev[i].meta_dev); in parse_dev_params()
855 rs->ti->error = "RAID metadata device lookup failure"; in parse_dev_params()
859 rs->dev[i].rdev.sb_page = alloc_page(GFP_KERNEL); in parse_dev_params()
860 if (!rs->dev[i].rdev.sb_page) { in parse_dev_params()
861 rs->ti->error = "Failed to allocate superblock page"; in parse_dev_params()
862 return -ENOMEM; in parse_dev_params()
868 return -EINVAL; in parse_dev_params()
870 if (!strcmp(arg, "-")) { in parse_dev_params()
871 if (!test_bit(In_sync, &rs->dev[i].rdev.flags) && in parse_dev_params()
872 (!rs->dev[i].rdev.recovery_offset)) { in parse_dev_params()
873 rs->ti->error = "Drive designated for rebuild not specified"; in parse_dev_params()
874 return -EINVAL; in parse_dev_params()
877 if (rs->dev[i].meta_dev) { in parse_dev_params()
878 rs->ti->error = "No data device supplied with metadata device"; in parse_dev_params()
879 return -EINVAL; in parse_dev_params()
885 r = dm_get_device(rs->ti, arg, dm_table_get_mode(rs->ti->table), in parse_dev_params()
886 &rs->dev[i].data_dev); in parse_dev_params()
888 rs->ti->error = "RAID device lookup failure"; in parse_dev_params()
892 if (rs->dev[i].meta_dev) { in parse_dev_params()
894 rs->dev[i].rdev.meta_bdev = rs->dev[i].meta_dev->bdev; in parse_dev_params()
896 rs->dev[i].rdev.bdev = rs->dev[i].data_dev->bdev; in parse_dev_params()
897 list_add_tail(&rs->dev[i].rdev.same_set, &rs->md.disks); in parse_dev_params()
898 if (!test_bit(In_sync, &rs->dev[i].rdev.flags)) in parse_dev_params()
902 if (rs->journal_dev.dev) in parse_dev_params()
903 list_add_tail(&rs->journal_dev.rdev.same_set, &rs->md.disks); in parse_dev_params()
906 rs->md.external = 0; in parse_dev_params()
907 rs->md.persistent = 1; in parse_dev_params()
908 rs->md.major_version = 2; in parse_dev_params()
909 } else if (rebuild && !rs->md.recovery_cp) { in parse_dev_params()
912 * is in-sync or not - we must assume it is not. Therefore, in parse_dev_params()
915 * Even if there is metadata, the on-disk information may in parse_dev_params()
916 * indicate that the array is not in-sync and it will then in parse_dev_params()
921 rs->ti->error = "Unable to rebuild drive while array is not in-sync"; in parse_dev_params()
922 return -EINVAL; in parse_dev_params()
930 * @rs
933 * Set rs->md.bitmap_info.chunksize (which really refers to 'region size').
934 * Ensure that (ti->len/region_size < 2^21) - required by MD bitmap.
936 * Returns: 0 on success, -EINVAL on failure.
938 static int validate_region_size(struct raid_set *rs, unsigned long region_size) in validate_region_size() argument
940 unsigned long min_region_size = rs->ti->len / (1 << 21); in validate_region_size()
942 if (rs_is_raid0(rs)) in validate_region_size()
960 * Validate user-supplied value. in validate_region_size()
962 if (region_size > rs->ti->len) { in validate_region_size()
963 rs->ti->error = "Supplied region size is too large"; in validate_region_size()
964 return -EINVAL; in validate_region_size()
970 rs->ti->error = "Supplied region size is too small"; in validate_region_size()
971 return -EINVAL; in validate_region_size()
975 rs->ti->error = "Region size is not a power of 2"; in validate_region_size()
976 return -EINVAL; in validate_region_size()
979 if (region_size < rs->md.chunk_sectors) { in validate_region_size()
980 rs->ti->error = "Region size is smaller than the chunk size"; in validate_region_size()
981 return -EINVAL; in validate_region_size()
988 rs->md.bitmap_info.chunksize = to_bytes(region_size); in validate_region_size()
995 * @rs
1000 * Returns: 0 on success, -EINVAL on failure.
1002 static int validate_raid_redundancy(struct raid_set *rs) in validate_raid_redundancy() argument
1008 for (i = 0; i < rs->raid_disks; i++) in validate_raid_redundancy()
1009 if (!test_bit(FirstUse, &rs->dev[i].rdev.flags) && in validate_raid_redundancy()
1010 ((!test_bit(In_sync, &rs->dev[i].rdev.flags) || in validate_raid_redundancy()
1011 !rs->dev[i].rdev.sb_page))) in validate_raid_redundancy()
1014 switch (rs->md.level) { in validate_raid_redundancy()
1018 if (rebuild_cnt >= rs->md.raid_disks) in validate_raid_redundancy()
1024 if (rebuild_cnt > rs->raid_type->parity_devs) in validate_raid_redundancy()
1028 copies = raid10_md_layout_to_copies(rs->md.new_layout); in validate_raid_redundancy()
1031 return -EINVAL; in validate_raid_redundancy()
1051 raid_disks = min(rs->raid_disks, rs->md.raid_disks); in validate_raid_redundancy()
1052 if (__is_raid10_near(rs->md.new_layout)) { in validate_raid_redundancy()
1056 if ((!rs->dev[i].rdev.sb_page || in validate_raid_redundancy()
1057 !test_bit(In_sync, &rs->dev[i].rdev.flags)) && in validate_raid_redundancy()
1077 last_group_start = (raid_disks / group_size) - 1; in validate_raid_redundancy()
1082 if ((!rs->dev[i].rdev.sb_page || in validate_raid_redundancy()
1083 !test_bit(In_sync, &rs->dev[i].rdev.flags)) && in validate_raid_redundancy()
1090 return -EINVAL; in validate_raid_redundancy()
1096 return -EINVAL; in validate_raid_redundancy()
1114 * [max_write_behind <sectors>] See '-write-behind=' (man mdadm)
1120 * RAID10-only options:
1124 static int parse_raid_params(struct raid_set *rs, struct dm_arg_set *as, in parse_raid_params() argument
1134 struct raid_type *rt = rs->raid_type; in parse_raid_params()
1137 num_raid_params--; /* Account for chunk_size argument */ in parse_raid_params()
1140 rs->ti->error = "Bad numerical argument given for chunk_size"; in parse_raid_params()
1141 return -EINVAL; in parse_raid_params()
1145 * First, parse the in-order required arguments in parse_raid_params()
1153 rs->ti->error = "Chunk size must be a power of 2"; in parse_raid_params()
1154 return -EINVAL; in parse_raid_params()
1156 rs->ti->error = "Chunk size value is too small"; in parse_raid_params()
1157 return -EINVAL; in parse_raid_params()
1160 rs->md.new_chunk_sectors = rs->md.chunk_sectors = value; in parse_raid_params()
1168 * - Device is reset when param is read. in parse_raid_params()
1170 * - No matching superblock found, resets device. in parse_raid_params()
1172 * - Failure noticed, resets device for bitmap replay. in parse_raid_params()
1174 * - Superblock is read and overrides recovery_offset. in parse_raid_params()
1179 for (i = 0; i < rs->raid_disks; i++) { in parse_raid_params()
1180 set_bit(In_sync, &rs->dev[i].rdev.flags); in parse_raid_params()
1181 rs->dev[i].rdev.recovery_offset = MaxSector; in parse_raid_params()
1190 rs->ti->error = "Not enough raid parameters given"; in parse_raid_params()
1191 return -EINVAL; in parse_raid_params()
1195 if (test_and_set_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) { in parse_raid_params()
1196 rs->ti->error = "Only one 'nosync' argument allowed"; in parse_raid_params()
1197 return -EINVAL; in parse_raid_params()
1202 if (test_and_set_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) { in parse_raid_params()
1203 rs->ti->error = "Only one 'sync' argument allowed"; in parse_raid_params()
1204 return -EINVAL; in parse_raid_params()
1209 if (test_and_set_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) { in parse_raid_params()
1210 rs->ti->error = "Only one 'raid10_use_new_sets' argument allowed"; in parse_raid_params()
1211 return -EINVAL; in parse_raid_params()
1219 rs->ti->error = "Wrong number of raid parameters given"; in parse_raid_params()
1220 return -EINVAL; in parse_raid_params()
1228 if (test_and_set_bit(__CTR_FLAG_RAID10_FORMAT, &rs->ctr_flags)) { in parse_raid_params()
1229 rs->ti->error = "Only one 'raid10_format' argument pair allowed"; in parse_raid_params()
1230 return -EINVAL; in parse_raid_params()
1233 rs->ti->error = "'raid10_format' is an invalid parameter for this RAID type"; in parse_raid_params()
1234 return -EINVAL; in parse_raid_params()
1238 rs->ti->error = "Invalid 'raid10_format' value given"; in parse_raid_params()
1249 if (test_and_set_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags)) { in parse_raid_params()
1250 rs->ti->error = "Only one raid4/5/6 set journaling device allowed"; in parse_raid_params()
1251 return -EINVAL; in parse_raid_params()
1254 rs->ti->error = "'journal_dev' is an invalid parameter for this RAID type"; in parse_raid_params()
1255 return -EINVAL; in parse_raid_params()
1257 r = dm_get_device(rs->ti, arg, dm_table_get_mode(rs->ti->table), in parse_raid_params()
1258 &rs->journal_dev.dev); in parse_raid_params()
1260 rs->ti->error = "raid4/5/6 journal device lookup failure"; in parse_raid_params()
1263 jdev = &rs->journal_dev.rdev; in parse_raid_params()
1265 jdev->mddev = &rs->md; in parse_raid_params()
1266 jdev->bdev = rs->journal_dev.dev->bdev; in parse_raid_params()
1267 jdev->sectors = bdev_nr_sectors(jdev->bdev); in parse_raid_params()
1268 if (jdev->sectors < MIN_RAID456_JOURNAL_SPACE) { in parse_raid_params()
1269 rs->ti->error = "No space for raid4/5/6 journal"; in parse_raid_params()
1270 return -ENOSPC; in parse_raid_params()
1272 rs->journal_dev.mode = R5C_JOURNAL_MODE_WRITE_THROUGH; in parse_raid_params()
1273 set_bit(Journal, &jdev->flags); in parse_raid_params()
1281 if (!test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags)) { in parse_raid_params()
1282 rs->ti->error = "raid4/5/6 'journal_mode' is invalid without 'journal_dev'"; in parse_raid_params()
1283 return -EINVAL; in parse_raid_params()
1285 if (test_and_set_bit(__CTR_FLAG_JOURNAL_MODE, &rs->ctr_flags)) { in parse_raid_params()
1286 rs->ti->error = "Only one raid4/5/6 'journal_mode' argument allowed"; in parse_raid_params()
1287 return -EINVAL; in parse_raid_params()
1291 rs->ti->error = "Invalid 'journal_mode' argument"; in parse_raid_params()
1294 rs->journal_dev.mode = r; in parse_raid_params()
1302 rs->ti->error = "Bad numerical argument given in raid params"; in parse_raid_params()
1303 return -EINVAL; in parse_raid_params()
1312 if (!__within_range(value, 0, rs->raid_disks - 1)) { in parse_raid_params()
1313 rs->ti->error = "Invalid rebuild index given"; in parse_raid_params()
1314 return -EINVAL; in parse_raid_params()
1317 if (test_and_set_bit(value, (void *) rs->rebuild_disks)) { in parse_raid_params()
1318 rs->ti->error = "rebuild for this index already given"; in parse_raid_params()
1319 return -EINVAL; in parse_raid_params()
1322 rd = rs->dev + value; in parse_raid_params()
1323 clear_bit(In_sync, &rd->rdev.flags); in parse_raid_params()
1324 clear_bit(Faulty, &rd->rdev.flags); in parse_raid_params()
1325 rd->rdev.recovery_offset = 0; in parse_raid_params()
1326 set_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags); in parse_raid_params()
1329 rs->ti->error = "write_mostly option is only valid for RAID1"; in parse_raid_params()
1330 return -EINVAL; in parse_raid_params()
1333 if (!__within_range(value, 0, rs->md.raid_disks - 1)) { in parse_raid_params()
1334 rs->ti->error = "Invalid write_mostly index given"; in parse_raid_params()
1335 return -EINVAL; in parse_raid_params()
1339 set_bit(WriteMostly, &rs->dev[value].rdev.flags); in parse_raid_params()
1340 set_bit(__CTR_FLAG_WRITE_MOSTLY, &rs->ctr_flags); in parse_raid_params()
1343 rs->ti->error = "max_write_behind option is only valid for RAID1"; in parse_raid_params()
1344 return -EINVAL; in parse_raid_params()
1347 if (test_and_set_bit(__CTR_FLAG_MAX_WRITE_BEHIND, &rs->ctr_flags)) { in parse_raid_params()
1348 rs->ti->error = "Only one max_write_behind argument pair allowed"; in parse_raid_params()
1349 return -EINVAL; in parse_raid_params()
1353 * In device-mapper, we specify things in sectors, but in parse_raid_params()
1357 rs->ti->error = "Max write-behind limit out of range"; in parse_raid_params()
1358 return -EINVAL; in parse_raid_params()
1361 rs->md.bitmap_info.max_write_behind = value / 2; in parse_raid_params()
1363 if (test_and_set_bit(__CTR_FLAG_DAEMON_SLEEP, &rs->ctr_flags)) { in parse_raid_params()
1364 rs->ti->error = "Only one daemon_sleep argument pair allowed"; in parse_raid_params()
1365 return -EINVAL; in parse_raid_params()
1368 rs->ti->error = "daemon sleep period out of range"; in parse_raid_params()
1369 return -EINVAL; in parse_raid_params()
1371 rs->md.bitmap_info.daemon_sleep = value; in parse_raid_params()
1374 if (test_and_set_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) { in parse_raid_params()
1375 rs->ti->error = "Only one data_offset argument pair allowed"; in parse_raid_params()
1376 return -EINVAL; in parse_raid_params()
1381 rs->ti->error = "Bogus data_offset value"; in parse_raid_params()
1382 return -EINVAL; in parse_raid_params()
1384 rs->data_offset = value; in parse_raid_params()
1386 /* Define the +/-# of disks to add to/remove from the given raid set */ in parse_raid_params()
1387 if (test_and_set_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) { in parse_raid_params()
1388 rs->ti->error = "Only one delta_disks argument pair allowed"; in parse_raid_params()
1389 return -EINVAL; in parse_raid_params()
1392 if (!__within_range(abs(value), 1, MAX_RAID_DEVICES - rt->minimal_devs)) { in parse_raid_params()
1393 rs->ti->error = "Too many delta_disk requested"; in parse_raid_params()
1394 return -EINVAL; in parse_raid_params()
1397 rs->delta_disks = value; in parse_raid_params()
1399 if (test_and_set_bit(__CTR_FLAG_STRIPE_CACHE, &rs->ctr_flags)) { in parse_raid_params()
1400 rs->ti->error = "Only one stripe_cache argument pair allowed"; in parse_raid_params()
1401 return -EINVAL; in parse_raid_params()
1405 rs->ti->error = "Inappropriate argument: stripe_cache"; in parse_raid_params()
1406 return -EINVAL; in parse_raid_params()
1410 rs->ti->error = "Bogus stripe cache entries value"; in parse_raid_params()
1411 return -EINVAL; in parse_raid_params()
1413 rs->stripe_cache_entries = value; in parse_raid_params()
1415 if (test_and_set_bit(__CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags)) { in parse_raid_params()
1416 rs->ti->error = "Only one min_recovery_rate argument pair allowed"; in parse_raid_params()
1417 return -EINVAL; in parse_raid_params()
1421 rs->ti->error = "min_recovery_rate out of range"; in parse_raid_params()
1422 return -EINVAL; in parse_raid_params()
1424 rs->md.sync_speed_min = value; in parse_raid_params()
1426 if (test_and_set_bit(__CTR_FLAG_MAX_RECOVERY_RATE, &rs->ctr_flags)) { in parse_raid_params()
1427 rs->ti->error = "Only one max_recovery_rate argument pair allowed"; in parse_raid_params()
1428 return -EINVAL; in parse_raid_params()
1432 rs->ti->error = "max_recovery_rate out of range"; in parse_raid_params()
1433 return -EINVAL; in parse_raid_params()
1435 rs->md.sync_speed_max = value; in parse_raid_params()
1437 if (test_and_set_bit(__CTR_FLAG_REGION_SIZE, &rs->ctr_flags)) { in parse_raid_params()
1438 rs->ti->error = "Only one region_size argument pair allowed"; in parse_raid_params()
1439 return -EINVAL; in parse_raid_params()
1443 rs->requested_bitmap_chunk_sectors = value; in parse_raid_params()
1445 if (test_and_set_bit(__CTR_FLAG_RAID10_COPIES, &rs->ctr_flags)) { in parse_raid_params()
1446 rs->ti->error = "Only one raid10_copies argument pair allowed"; in parse_raid_params()
1447 return -EINVAL; in parse_raid_params()
1450 if (!__within_range(value, 2, rs->md.raid_disks)) { in parse_raid_params()
1451 rs->ti->error = "Bad value for 'raid10_copies'"; in parse_raid_params()
1452 return -EINVAL; in parse_raid_params()
1458 rs->ti->error = "Unable to parse RAID parameter"; in parse_raid_params()
1459 return -EINVAL; in parse_raid_params()
1463 if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags) && in parse_raid_params()
1464 test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) { in parse_raid_params()
1465 rs->ti->error = "sync and nosync are mutually exclusive"; in parse_raid_params()
1466 return -EINVAL; in parse_raid_params()
1469 if (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags) && in parse_raid_params()
1470 (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags) || in parse_raid_params()
1471 test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags))) { in parse_raid_params()
1472 rs->ti->error = "sync/nosync and rebuild are mutually exclusive"; in parse_raid_params()
1473 return -EINVAL; in parse_raid_params()
1476 if (write_mostly >= rs->md.raid_disks) { in parse_raid_params()
1477 rs->ti->error = "Can't set all raid1 devices to write_mostly"; in parse_raid_params()
1478 return -EINVAL; in parse_raid_params()
1481 if (rs->md.sync_speed_max && in parse_raid_params()
1482 rs->md.sync_speed_min > rs->md.sync_speed_max) { in parse_raid_params()
1483 rs->ti->error = "Bogus recovery rates"; in parse_raid_params()
1484 return -EINVAL; in parse_raid_params()
1487 if (validate_region_size(rs, region_size)) in parse_raid_params()
1488 return -EINVAL; in parse_raid_params()
1490 if (rs->md.chunk_sectors) in parse_raid_params()
1491 max_io_len = rs->md.chunk_sectors; in parse_raid_params()
1495 if (dm_set_target_max_io_len(rs->ti, max_io_len)) in parse_raid_params()
1496 return -EINVAL; in parse_raid_params()
1499 if (raid10_copies > rs->md.raid_disks) { in parse_raid_params()
1500 rs->ti->error = "Not enough devices to satisfy specification"; in parse_raid_params()
1501 return -EINVAL; in parse_raid_params()
1504 rs->md.new_layout = raid10_format_to_md_layout(rs, raid10_format, raid10_copies); in parse_raid_params()
1505 if (rs->md.new_layout < 0) { in parse_raid_params()
1506 rs->ti->error = "Error getting raid10 format"; in parse_raid_params()
1507 return rs->md.new_layout; in parse_raid_params()
1510 rt = get_raid_type_by_ll(10, rs->md.new_layout); in parse_raid_params()
1512 rs->ti->error = "Failed to recognize new raid10 layout"; in parse_raid_params()
1513 return -EINVAL; in parse_raid_params()
1516 if ((rt->algorithm == ALGORITHM_RAID10_DEFAULT || in parse_raid_params()
1517 rt->algorithm == ALGORITHM_RAID10_NEAR) && in parse_raid_params()
1518 test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) { in parse_raid_params()
1519 rs->ti->error = "RAID10 format 'near' and 'raid10_use_near_sets' are incompatible"; in parse_raid_params()
1520 return -EINVAL; in parse_raid_params()
1524 rs->raid10_copies = raid10_copies; in parse_raid_params()
1527 rs->md.persistent = 0; in parse_raid_params()
1528 rs->md.external = 1; in parse_raid_params()
1531 return rs_check_for_valid_flags(rs); in parse_raid_params()
1535 static int rs_set_raid456_stripe_cache(struct raid_set *rs) in rs_set_raid456_stripe_cache() argument
1539 struct mddev *mddev = &rs->md; in rs_set_raid456_stripe_cache()
1540 uint32_t min_stripes = max(mddev->chunk_sectors, mddev->new_chunk_sectors) / 2; in rs_set_raid456_stripe_cache()
1541 uint32_t nr_stripes = rs->stripe_cache_entries; in rs_set_raid456_stripe_cache()
1543 if (!rt_is_raid456(rs->raid_type)) { in rs_set_raid456_stripe_cache()
1544 rs->ti->error = "Inappropriate raid level; cannot change stripe_cache size"; in rs_set_raid456_stripe_cache()
1545 return -EINVAL; in rs_set_raid456_stripe_cache()
1554 conf = mddev->private; in rs_set_raid456_stripe_cache()
1556 rs->ti->error = "Cannot change stripe_cache size on inactive RAID set"; in rs_set_raid456_stripe_cache()
1557 return -EINVAL; in rs_set_raid456_stripe_cache()
1561 if (conf->min_nr_stripes != nr_stripes) { in rs_set_raid456_stripe_cache()
1564 rs->ti->error = "Failed to set raid4/5/6 stripe cache size"; in rs_set_raid456_stripe_cache()
1574 /* Return # of data stripes as kept in mddev as of @rs (i.e. as of superblock) */
1575 static unsigned int mddev_data_stripes(struct raid_set *rs) in mddev_data_stripes() argument
1577 return rs->md.raid_disks - rs->raid_type->parity_devs; in mddev_data_stripes()
1580 /* Return # of data stripes of @rs (i.e. as of ctr) */
1581 static unsigned int rs_data_stripes(struct raid_set *rs) in rs_data_stripes() argument
1583 return rs->raid_disks - rs->raid_type->parity_devs; in rs_data_stripes()
1587 * Retrieve rdev->sectors from any valid raid device of @rs
1588 * to allow userpace to pass in arbitray "- -" device tupples.
1590 static sector_t __rdev_sectors(struct raid_set *rs) in __rdev_sectors() argument
1594 for (i = 0; i < rs->raid_disks; i++) { in __rdev_sectors()
1595 struct md_rdev *rdev = &rs->dev[i].rdev; in __rdev_sectors()
1597 if (!test_bit(Journal, &rdev->flags) && in __rdev_sectors()
1598 rdev->bdev && rdev->sectors) in __rdev_sectors()
1599 return rdev->sectors; in __rdev_sectors()
1606 static int _check_data_dev_sectors(struct raid_set *rs) in _check_data_dev_sectors() argument
1611 rdev_for_each(rdev, &rs->md) in _check_data_dev_sectors()
1612 if (!test_bit(Journal, &rdev->flags) && rdev->bdev) { in _check_data_dev_sectors()
1613 ds = min(ds, bdev_nr_sectors(rdev->bdev)); in _check_data_dev_sectors()
1614 if (ds < rs->md.dev_sectors) { in _check_data_dev_sectors()
1615 rs->ti->error = "Component device(s) too small"; in _check_data_dev_sectors()
1616 return -EINVAL; in _check_data_dev_sectors()
1623 /* Calculate the sectors per device and per array used for @rs */
1624 static int rs_set_dev_and_array_sectors(struct raid_set *rs, sector_t sectors, bool use_mddev) in rs_set_dev_and_array_sectors() argument
1629 struct mddev *mddev = &rs->md; in rs_set_dev_and_array_sectors()
1632 delta_disks = mddev->delta_disks; in rs_set_dev_and_array_sectors()
1633 data_stripes = mddev_data_stripes(rs); in rs_set_dev_and_array_sectors()
1635 delta_disks = rs->delta_disks; in rs_set_dev_and_array_sectors()
1636 data_stripes = rs_data_stripes(rs); in rs_set_dev_and_array_sectors()
1640 if (rt_is_raid1(rs->raid_type)) in rs_set_dev_and_array_sectors()
1642 else if (rt_is_raid10(rs->raid_type)) { in rs_set_dev_and_array_sectors()
1643 if (rs->raid10_copies < 2 || in rs_set_dev_and_array_sectors()
1645 rs->ti->error = "Bogus raid10 data copies or delta disks"; in rs_set_dev_and_array_sectors()
1646 return -EINVAL; in rs_set_dev_and_array_sectors()
1649 dev_sectors *= rs->raid10_copies; in rs_set_dev_and_array_sectors()
1654 if (sector_div(array_sectors, rs->raid10_copies)) in rs_set_dev_and_array_sectors()
1664 mddev->array_sectors = array_sectors; in rs_set_dev_and_array_sectors()
1665 mddev->dev_sectors = dev_sectors; in rs_set_dev_and_array_sectors()
1666 rs_set_rdev_sectors(rs); in rs_set_dev_and_array_sectors()
1668 return _check_data_dev_sectors(rs); in rs_set_dev_and_array_sectors()
1670 rs->ti->error = "Target length not divisible by number of data devices"; in rs_set_dev_and_array_sectors()
1671 return -EINVAL; in rs_set_dev_and_array_sectors()
1674 /* Setup recovery on @rs */
1675 static void rs_setup_recovery(struct raid_set *rs, sector_t dev_sectors) in rs_setup_recovery() argument
1678 if (rs_is_raid0(rs)) in rs_setup_recovery()
1679 rs->md.recovery_cp = MaxSector; in rs_setup_recovery()
1683 * ensure proper parity and Q-Syndrome in rs_setup_recovery()
1685 else if (rs_is_raid6(rs)) in rs_setup_recovery()
1686 rs->md.recovery_cp = dev_sectors; in rs_setup_recovery()
1692 rs->md.recovery_cp = test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags) in rs_setup_recovery()
1698 struct raid_set *rs = container_of(ws, struct raid_set, md.event_work); in do_table_event() local
1701 if (!rs_is_reshaping(rs)) { in do_table_event()
1702 if (rs_is_raid10(rs)) in do_table_event()
1703 rs_set_rdev_sectors(rs); in do_table_event()
1704 rs_set_capacity(rs); in do_table_event()
1706 dm_table_event(rs->ti->table); in do_table_event()
1710 * Make sure a valid takover (level switch) is being requested on @rs
1715 static int rs_check_takeover(struct raid_set *rs) in rs_check_takeover() argument
1717 struct mddev *mddev = &rs->md; in rs_check_takeover()
1720 if (rs->md.degraded) { in rs_check_takeover()
1721 rs->ti->error = "Can't takeover degraded raid set"; in rs_check_takeover()
1722 return -EPERM; in rs_check_takeover()
1725 if (rs_is_reshaping(rs)) { in rs_check_takeover()
1726 rs->ti->error = "Can't takeover reshaping raid set"; in rs_check_takeover()
1727 return -EPERM; in rs_check_takeover()
1730 switch (mddev->level) { in rs_check_takeover()
1732 /* raid0 -> raid1/5 with one disk */ in rs_check_takeover()
1733 if ((mddev->new_level == 1 || mddev->new_level == 5) && in rs_check_takeover()
1734 mddev->raid_disks == 1) in rs_check_takeover()
1737 /* raid0 -> raid10 */ in rs_check_takeover()
1738 if (mddev->new_level == 10 && in rs_check_takeover()
1739 !(rs->raid_disks % mddev->raid_disks)) in rs_check_takeover()
1742 /* raid0 with multiple disks -> raid4/5/6 */ in rs_check_takeover()
1743 if (__within_range(mddev->new_level, 4, 6) && in rs_check_takeover()
1744 mddev->new_layout == ALGORITHM_PARITY_N && in rs_check_takeover()
1745 mddev->raid_disks > 1) in rs_check_takeover()
1752 if (__is_raid10_offset(mddev->layout)) in rs_check_takeover()
1755 near_copies = __raid10_near_copies(mddev->layout); in rs_check_takeover()
1757 /* raid10* -> raid0 */ in rs_check_takeover()
1758 if (mddev->new_level == 0) { in rs_check_takeover()
1761 !(mddev->raid_disks % near_copies)) { in rs_check_takeover()
1762 mddev->raid_disks /= near_copies; in rs_check_takeover()
1763 mddev->delta_disks = mddev->raid_disks; in rs_check_takeover()
1769 __raid10_far_copies(mddev->layout) > 1) in rs_check_takeover()
1775 /* raid10_{near,far} -> raid1 */ in rs_check_takeover()
1776 if (mddev->new_level == 1 && in rs_check_takeover()
1777 max(near_copies, __raid10_far_copies(mddev->layout)) == mddev->raid_disks) in rs_check_takeover()
1780 /* raid10_{near,far} with 2 disks -> raid4/5 */ in rs_check_takeover()
1781 if (__within_range(mddev->new_level, 4, 5) && in rs_check_takeover()
1782 mddev->raid_disks == 2) in rs_check_takeover()
1787 /* raid1 with 2 disks -> raid4/5 */ in rs_check_takeover()
1788 if (__within_range(mddev->new_level, 4, 5) && in rs_check_takeover()
1789 mddev->raid_disks == 2) { in rs_check_takeover()
1790 mddev->degraded = 1; in rs_check_takeover()
1794 /* raid1 -> raid0 */ in rs_check_takeover()
1795 if (mddev->new_level == 0 && in rs_check_takeover()
1796 mddev->raid_disks == 1) in rs_check_takeover()
1799 /* raid1 -> raid10 */ in rs_check_takeover()
1800 if (mddev->new_level == 10) in rs_check_takeover()
1805 /* raid4 -> raid0 */ in rs_check_takeover()
1806 if (mddev->new_level == 0) in rs_check_takeover()
1809 /* raid4 -> raid1/5 with 2 disks */ in rs_check_takeover()
1810 if ((mddev->new_level == 1 || mddev->new_level == 5) && in rs_check_takeover()
1811 mddev->raid_disks == 2) in rs_check_takeover()
1814 /* raid4 -> raid5/6 with parity N */ in rs_check_takeover()
1815 if (__within_range(mddev->new_level, 5, 6) && in rs_check_takeover()
1816 mddev->layout == ALGORITHM_PARITY_N) in rs_check_takeover()
1821 /* raid5 with parity N -> raid0 */ in rs_check_takeover()
1822 if (mddev->new_level == 0 && in rs_check_takeover()
1823 mddev->layout == ALGORITHM_PARITY_N) in rs_check_takeover()
1826 /* raid5 with parity N -> raid4 */ in rs_check_takeover()
1827 if (mddev->new_level == 4 && in rs_check_takeover()
1828 mddev->layout == ALGORITHM_PARITY_N) in rs_check_takeover()
1831 /* raid5 with 2 disks -> raid1/4/10 */ in rs_check_takeover()
1832 if ((mddev->new_level == 1 || mddev->new_level == 4 || mddev->new_level == 10) && in rs_check_takeover()
1833 mddev->raid_disks == 2) in rs_check_takeover()
1836 /* raid5_* -> raid6_*_6 with Q-Syndrome N (e.g. raid5_ra -> raid6_ra_6 */ in rs_check_takeover()
1837 if (mddev->new_level == 6 && in rs_check_takeover()
1838 ((mddev->layout == ALGORITHM_PARITY_N && mddev->new_layout == ALGORITHM_PARITY_N) || in rs_check_takeover()
1839 … __within_range(mddev->new_layout, ALGORITHM_LEFT_ASYMMETRIC_6, ALGORITHM_RIGHT_SYMMETRIC_6))) in rs_check_takeover()
1844 /* raid6 with parity N -> raid0 */ in rs_check_takeover()
1845 if (mddev->new_level == 0 && in rs_check_takeover()
1846 mddev->layout == ALGORITHM_PARITY_N) in rs_check_takeover()
1849 /* raid6 with parity N -> raid4 */ in rs_check_takeover()
1850 if (mddev->new_level == 4 && in rs_check_takeover()
1851 mddev->layout == ALGORITHM_PARITY_N) in rs_check_takeover()
1854 /* raid6_*_n with Q-Syndrome N -> raid5_* */ in rs_check_takeover()
1855 if (mddev->new_level == 5 && in rs_check_takeover()
1856 ((mddev->layout == ALGORITHM_PARITY_N && mddev->new_layout == ALGORITHM_PARITY_N) || in rs_check_takeover()
1857 __within_range(mddev->new_layout, ALGORITHM_LEFT_ASYMMETRIC, ALGORITHM_RIGHT_SYMMETRIC))) in rs_check_takeover()
1865 rs->ti->error = "takeover not possible"; in rs_check_takeover()
1866 return -EINVAL; in rs_check_takeover()
1869 /* True if @rs requested to be taken over */
1870 static bool rs_takeover_requested(struct raid_set *rs) in rs_takeover_requested() argument
1872 return rs->md.new_level != rs->md.level; in rs_takeover_requested()
1876 static bool rs_is_layout_change(struct raid_set *rs, bool use_mddev) in rs_is_layout_change() argument
1878 return (use_mddev ? rs->md.delta_disks : rs->delta_disks) || in rs_is_layout_change()
1879 rs->md.new_layout != rs->md.layout || in rs_is_layout_change()
1880 rs->md.new_chunk_sectors != rs->md.chunk_sectors; in rs_is_layout_change()
1883 /* True if @rs is requested to reshape by ctr */
1884 static bool rs_reshape_requested(struct raid_set *rs) in rs_reshape_requested() argument
1887 struct mddev *mddev = &rs->md; in rs_reshape_requested()
1889 if (rs_takeover_requested(rs)) in rs_reshape_requested()
1892 if (rs_is_raid0(rs)) in rs_reshape_requested()
1895 change = rs_is_layout_change(rs, false); in rs_reshape_requested()
1898 if (rs_is_raid1(rs)) { in rs_reshape_requested()
1899 if (rs->delta_disks) in rs_reshape_requested()
1900 return !!rs->delta_disks; in rs_reshape_requested()
1903 mddev->raid_disks != rs->raid_disks; in rs_reshape_requested()
1906 if (rs_is_raid10(rs)) in rs_reshape_requested()
1908 !__is_raid10_far(mddev->new_layout) && in rs_reshape_requested()
1909 rs->delta_disks >= 0; in rs_reshape_requested()
1917 /* State flags for sb->flags */
1923 * Devices with this superblock should only ever be accessed via device-mapper.
1993 * up to 256 devices with the 1.9.0 on-disk metadata format
1995 __le64 extended_failed_devices[DISKS_ARRAY_ELEMS - 1];
2003 * Check for reshape constraints on raid set @rs:
2005 * - reshape function non-existent
2006 * - degraded set
2007 * - ongoing recovery
2008 * - ongoing reshape
2010 * Returns 0 if none or -EPERM if given constraint
2013 static int rs_check_reshape(struct raid_set *rs) in rs_check_reshape() argument
2015 struct mddev *mddev = &rs->md; in rs_check_reshape()
2017 if (!mddev->pers || !mddev->pers->check_reshape) in rs_check_reshape()
2018 rs->ti->error = "Reshape not supported"; in rs_check_reshape()
2019 else if (mddev->degraded) in rs_check_reshape()
2020 rs->ti->error = "Can't reshape degraded raid set"; in rs_check_reshape()
2021 else if (rs_is_recovering(rs)) in rs_check_reshape()
2022 rs->ti->error = "Convert request on recovering raid set prohibited"; in rs_check_reshape()
2023 else if (rs_is_reshaping(rs)) in rs_check_reshape()
2024 rs->ti->error = "raid set already reshaping!"; in rs_check_reshape()
2025 else if (!(rs_is_raid1(rs) || rs_is_raid10(rs) || rs_is_raid456(rs))) in rs_check_reshape()
2026 rs->ti->error = "Reshaping only supported for raid1/4/5/6/10"; in rs_check_reshape()
2030 return -EPERM; in rs_check_reshape()
2035 BUG_ON(!rdev->sb_page); in read_disk_sb()
2037 if (rdev->sb_loaded && !force_reload) in read_disk_sb()
2040 rdev->sb_loaded = 0; in read_disk_sb()
2042 if (!sync_page_io(rdev, 0, size, rdev->sb_page, REQ_OP_READ, true)) { in read_disk_sb()
2044 rdev->raid_disk); in read_disk_sb()
2045 md_error(rdev->mddev, rdev); in read_disk_sb()
2046 set_bit(Faulty, &rdev->flags); in read_disk_sb()
2047 return -EIO; in read_disk_sb()
2050 rdev->sb_loaded = 1; in read_disk_sb()
2057 failed_devices[0] = le64_to_cpu(sb->failed_devices); in sb_retrieve_failed_devices()
2058 memset(failed_devices + 1, 0, sizeof(sb->extended_failed_devices)); in sb_retrieve_failed_devices()
2060 if (le32_to_cpu(sb->compat_features) & FEATURE_FLAG_SUPPORTS_V190) { in sb_retrieve_failed_devices()
2061 int i = ARRAY_SIZE(sb->extended_failed_devices); in sb_retrieve_failed_devices()
2063 while (i--) in sb_retrieve_failed_devices()
2064 failed_devices[i+1] = le64_to_cpu(sb->extended_failed_devices[i]); in sb_retrieve_failed_devices()
2070 int i = ARRAY_SIZE(sb->extended_failed_devices); in sb_update_failed_devices()
2072 sb->failed_devices = cpu_to_le64(failed_devices[0]); in sb_update_failed_devices()
2073 while (i--) in sb_update_failed_devices()
2074 sb->extended_failed_devices[i] = cpu_to_le64(failed_devices[i+1]); in sb_update_failed_devices()
2088 struct raid_set *rs = container_of(mddev, struct raid_set, md); in super_sync() local
2091 if (!rdev->meta_bdev) in super_sync()
2094 BUG_ON(!rdev->sb_page); in super_sync()
2096 sb = page_address(rdev->sb_page); in super_sync()
2100 for (i = 0; i < rs->raid_disks; i++) in super_sync()
2101 if (!rs->dev[i].data_dev || test_bit(Faulty, &rs->dev[i].rdev.flags)) { in super_sync()
2109 sb->magic = cpu_to_le32(DM_RAID_MAGIC); in super_sync()
2110 sb->compat_features = cpu_to_le32(FEATURE_FLAG_SUPPORTS_V190); in super_sync()
2112 sb->num_devices = cpu_to_le32(mddev->raid_disks); in super_sync()
2113 sb->array_position = cpu_to_le32(rdev->raid_disk); in super_sync()
2115 sb->events = cpu_to_le64(mddev->events); in super_sync()
2117 sb->disk_recovery_offset = cpu_to_le64(rdev->recovery_offset); in super_sync()
2118 sb->array_resync_offset = cpu_to_le64(mddev->recovery_cp); in super_sync()
2120 sb->level = cpu_to_le32(mddev->level); in super_sync()
2121 sb->layout = cpu_to_le32(mddev->layout); in super_sync()
2122 sb->stripe_sectors = cpu_to_le32(mddev->chunk_sectors); in super_sync()
2129 sb->new_level = cpu_to_le32(mddev->new_level); in super_sync()
2130 sb->new_layout = cpu_to_le32(mddev->new_layout); in super_sync()
2131 sb->new_stripe_sectors = cpu_to_le32(mddev->new_chunk_sectors); in super_sync()
2133 sb->delta_disks = cpu_to_le32(mddev->delta_disks); in super_sync()
2136 sb->reshape_position = cpu_to_le64(mddev->reshape_position); in super_sync()
2137 if (le64_to_cpu(sb->reshape_position) != MaxSector) { in super_sync()
2139 sb->flags |= cpu_to_le32(SB_FLAG_RESHAPE_ACTIVE); in super_sync()
2141 if (mddev->delta_disks < 0 || mddev->reshape_backwards) in super_sync()
2142 sb->flags |= cpu_to_le32(SB_FLAG_RESHAPE_BACKWARDS); in super_sync()
2145 sb->flags &= ~(cpu_to_le32(SB_FLAG_RESHAPE_ACTIVE|SB_FLAG_RESHAPE_BACKWARDS)); in super_sync()
2148 sb->array_sectors = cpu_to_le64(mddev->array_sectors); in super_sync()
2149 sb->data_offset = cpu_to_le64(rdev->data_offset); in super_sync()
2150 sb->new_data_offset = cpu_to_le64(rdev->new_data_offset); in super_sync()
2151 sb->sectors = cpu_to_le64(rdev->sectors); in super_sync()
2152 sb->incompat_features = cpu_to_le32(0); in super_sync()
2155 memset(sb + 1, 0, rdev->sb_size - sizeof(*sb)); in super_sync()
2164 * Return: 1 if use rdev, 0 if use refdev, -Exxx otherwise
2173 r = read_disk_sb(rdev, rdev->sb_size, false); in super_load()
2177 sb = page_address(rdev->sb_page); in super_load()
2184 if ((sb->magic != cpu_to_le32(DM_RAID_MAGIC)) || in super_load()
2185 (!test_bit(In_sync, &rdev->flags) && !rdev->recovery_offset)) { in super_load()
2186 super_sync(rdev->mddev, rdev); in super_load()
2188 set_bit(FirstUse, &rdev->flags); in super_load()
2189 sb->compat_features = cpu_to_le32(FEATURE_FLAG_SUPPORTS_V190); in super_load()
2192 set_bit(MD_SB_CHANGE_DEVS, &rdev->mddev->sb_flags); in super_load()
2201 events_sb = le64_to_cpu(sb->events); in super_load()
2203 refsb = page_address(refdev->sb_page); in super_load()
2204 events_refsb = le64_to_cpu(refsb->events); in super_load()
2209 static int super_init_validation(struct raid_set *rs, struct md_rdev *rdev) in super_init_validation() argument
2212 struct mddev *mddev = &rs->md; in super_init_validation()
2220 sb = page_address(rdev->sb_page); in super_init_validation()
2221 events_sb = le64_to_cpu(sb->events); in super_init_validation()
2226 mddev->events = events_sb ? : 1; in super_init_validation()
2228 mddev->reshape_position = MaxSector; in super_init_validation()
2230 mddev->raid_disks = le32_to_cpu(sb->num_devices); in super_init_validation()
2231 mddev->level = le32_to_cpu(sb->level); in super_init_validation()
2232 mddev->layout = le32_to_cpu(sb->layout); in super_init_validation()
2233 mddev->chunk_sectors = le32_to_cpu(sb->stripe_sectors); in super_init_validation()
2239 if (le32_to_cpu(sb->compat_features) & FEATURE_FLAG_SUPPORTS_V190) { in super_init_validation()
2241 mddev->new_level = le32_to_cpu(sb->new_level); in super_init_validation()
2242 mddev->new_layout = le32_to_cpu(sb->new_layout); in super_init_validation()
2243 mddev->new_chunk_sectors = le32_to_cpu(sb->new_stripe_sectors); in super_init_validation()
2244 mddev->delta_disks = le32_to_cpu(sb->delta_disks); in super_init_validation()
2245 mddev->array_sectors = le64_to_cpu(sb->array_sectors); in super_init_validation()
2248 if (le32_to_cpu(sb->flags) & SB_FLAG_RESHAPE_ACTIVE) { in super_init_validation()
2249 if (test_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) { in super_init_validation()
2251 return -EINVAL; in super_init_validation()
2254 if (mddev->delta_disks < 0 || in super_init_validation()
2255 (!mddev->delta_disks && (le32_to_cpu(sb->flags) & SB_FLAG_RESHAPE_BACKWARDS))) in super_init_validation()
2256 mddev->reshape_backwards = 1; in super_init_validation()
2258 mddev->reshape_backwards = 0; in super_init_validation()
2260 mddev->reshape_position = le64_to_cpu(sb->reshape_position); in super_init_validation()
2261 rs->raid_type = get_raid_type_by_ll(mddev->level, mddev->layout); in super_init_validation()
2268 struct raid_type *rt_cur = get_raid_type_by_ll(mddev->level, mddev->layout); in super_init_validation()
2269 struct raid_type *rt_new = get_raid_type_by_ll(mddev->new_level, mddev->new_layout); in super_init_validation()
2271 if (rs_takeover_requested(rs)) { in super_init_validation()
2274 rt_cur->name, rt_new->name); in super_init_validation()
2277 return -EINVAL; in super_init_validation()
2278 } else if (rs_reshape_requested(rs)) { in super_init_validation()
2280 if (mddev->layout != mddev->new_layout) { in super_init_validation()
2283 rt_cur->name, rt_new->name); in super_init_validation()
2286 le32_to_cpu(sb->layout), mddev->new_layout); in super_init_validation()
2288 if (mddev->chunk_sectors != mddev->new_chunk_sectors) in super_init_validation()
2290 mddev->chunk_sectors, mddev->new_chunk_sectors); in super_init_validation()
2291 if (rs->delta_disks) in super_init_validation()
2293 mddev->raid_disks, mddev->raid_disks + rs->delta_disks); in super_init_validation()
2294 if (rs_is_raid10(rs)) { in super_init_validation()
2296 raid10_md_layout_to_format(mddev->layout), in super_init_validation()
2297 raid10_md_layout_to_copies(mddev->layout)); in super_init_validation()
2299 raid10_md_layout_to_format(mddev->new_layout), in super_init_validation()
2300 raid10_md_layout_to_copies(mddev->new_layout)); in super_init_validation()
2302 return -EINVAL; in super_init_validation()
2308 if (!test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) in super_init_validation()
2309 mddev->recovery_cp = le64_to_cpu(sb->array_resync_offset); in super_init_validation()
2314 * 1) The raid set is brand new - in which case, all of the in super_init_validation()
2318 * and the new device needs to be rebuilt - in which in super_init_validation()
2327 if (test_bit(Journal, &rdev->flags)) in super_init_validation()
2330 if (test_bit(FirstUse, &r->flags)) in super_init_validation()
2333 if (!test_bit(In_sync, &r->flags)) { in super_init_validation()
2335 r->raid_disk); in super_init_validation()
2338 if (test_bit(FirstUse, &r->flags)) in super_init_validation()
2343 if (new_devs == rs->raid_disks || !rebuilds) { in super_init_validation()
2345 if (new_devs == rs->raid_disks) { in super_init_validation()
2347 set_bit(MD_ARRAY_FIRST_USE, &mddev->flags); in super_init_validation()
2349 new_devs != rs->delta_disks) { in super_init_validation()
2352 return -EINVAL; in super_init_validation()
2356 " a raid set with %u other first-time devices", in super_init_validation()
2358 return -EINVAL; in super_init_validation()
2363 return -EINVAL; in super_init_validation()
2364 } else if (!test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags) && rs_is_recovering(rs)) { in super_init_validation()
2365 DMERR("'rebuild' specified while raid set is not in-sync (recovery_cp=%llu)", in super_init_validation()
2366 (unsigned long long) mddev->recovery_cp); in super_init_validation()
2367 return -EINVAL; in super_init_validation()
2368 } else if (rs_is_reshaping(rs)) { in super_init_validation()
2370 (unsigned long long) mddev->reshape_position); in super_init_validation()
2371 return -EINVAL; in super_init_validation()
2381 if (test_bit(Journal, &rdev->flags) || in super_init_validation()
2382 !r->sb_page) in super_init_validation()
2384 sb2 = page_address(r->sb_page); in super_init_validation()
2385 sb2->failed_devices = 0; in super_init_validation()
2386 memset(sb2->extended_failed_devices, 0, sizeof(sb2->extended_failed_devices)); in super_init_validation()
2389 * Check for any device re-ordering. in super_init_validation()
2391 if (!test_bit(FirstUse, &r->flags) && (r->raid_disk >= 0)) { in super_init_validation()
2392 role = le32_to_cpu(sb2->array_position); in super_init_validation()
2396 if (role != r->raid_disk) { in super_init_validation()
2397 if (rs_is_raid10(rs) && __is_raid10_near(mddev->layout)) { in super_init_validation()
2398 if (mddev->raid_disks % __raid10_near_copies(mddev->layout) || in super_init_validation()
2399 rs->raid_disks % rs->raid10_copies) { in super_init_validation()
2400 rs->ti->error = in super_init_validation()
2402 return -EINVAL; in super_init_validation()
2405 sb2->array_position = cpu_to_le32(r->raid_disk); in super_init_validation()
2407 } else if (!(rs_is_raid10(rs) && rt_is_raid0(rs->raid_type)) && in super_init_validation()
2408 !(rs_is_raid0(rs) && rt_is_raid10(rs->raid_type)) && in super_init_validation()
2409 !rt_is_raid1(rs->raid_type)) { in super_init_validation()
2410 rs->ti->error = "Cannot change device positions in raid set"; in super_init_validation()
2411 return -EINVAL; in super_init_validation()
2414 DMINFO("raid device #%d now at position #%d", role, r->raid_disk); in super_init_validation()
2422 set_bit(Faulty, &r->flags); in super_init_validation()
2429 static int super_validate(struct raid_set *rs, struct md_rdev *rdev) in super_validate() argument
2431 struct mddev *mddev = &rs->md; in super_validate()
2434 if (rs_is_raid0(rs) || !rdev->sb_page || rdev->raid_disk < 0) in super_validate()
2437 sb = page_address(rdev->sb_page); in super_validate()
2440 * If mddev->events is not set, we know we have not yet initialized in super_validate()
2443 if (!mddev->events && super_init_validation(rs, rdev)) in super_validate()
2444 return -EINVAL; in super_validate()
2446 if (le32_to_cpu(sb->compat_features) && in super_validate()
2447 le32_to_cpu(sb->compat_features) != FEATURE_FLAG_SUPPORTS_V190) { in super_validate()
2448 rs->ti->error = "Unable to assemble array: Unknown flag(s) in compatible feature flags"; in super_validate()
2449 return -EINVAL; in super_validate()
2452 if (sb->incompat_features) { in super_validate()
2453 rs->ti->error = "Unable to assemble array: No incompatible feature flags supported yet"; in super_validate()
2454 return -EINVAL; in super_validate()
2457 /* Enable bitmap creation on @rs unless no metadevs or raid0 or journaled raid4/5/6 set. */ in super_validate()
2458 …mddev->bitmap_info.offset = (rt_is_raid0(rs->raid_type) || rs->journal_dev.dev) ? 0 : to_sector(40… in super_validate()
2459 mddev->bitmap_info.default_offset = mddev->bitmap_info.offset; in super_validate()
2461 if (!test_and_clear_bit(FirstUse, &rdev->flags)) { in super_validate()
2467 if (le32_to_cpu(sb->compat_features) & FEATURE_FLAG_SUPPORTS_V190) in super_validate()
2468 rdev->sectors = le64_to_cpu(sb->sectors); in super_validate()
2470 rdev->recovery_offset = le64_to_cpu(sb->disk_recovery_offset); in super_validate()
2471 if (rdev->recovery_offset == MaxSector) in super_validate()
2472 set_bit(In_sync, &rdev->flags); in super_validate()
2474 * If no reshape in progress -> we're recovering single in super_validate()
2475 * disk(s) and have to set the device(s) to out-of-sync in super_validate()
2477 else if (!rs_is_reshaping(rs)) in super_validate()
2478 clear_bit(In_sync, &rdev->flags); /* Mandatory for recovery */ in super_validate()
2484 if (test_and_clear_bit(Faulty, &rdev->flags)) { in super_validate()
2485 rdev->recovery_offset = 0; in super_validate()
2486 clear_bit(In_sync, &rdev->flags); in super_validate()
2487 rdev->saved_raid_disk = rdev->raid_disk; in super_validate()
2490 /* Reshape support -> restore repective data offsets */ in super_validate()
2491 rdev->data_offset = le64_to_cpu(sb->data_offset); in super_validate()
2492 rdev->new_data_offset = le64_to_cpu(sb->new_data_offset); in super_validate()
2500 static int analyse_superblocks(struct dm_target *ti, struct raid_set *rs) in analyse_superblocks() argument
2504 struct mddev *mddev = &rs->md; in analyse_superblocks()
2508 if (test_bit(Journal, &rdev->flags)) in analyse_superblocks()
2511 if (!rdev->meta_bdev) in analyse_superblocks()
2515 rdev->sb_start = 0; in analyse_superblocks()
2516 rdev->sb_size = bdev_logical_block_size(rdev->meta_bdev); in analyse_superblocks()
2517 if (rdev->sb_size < sizeof(struct dm_raid_superblock) || rdev->sb_size > PAGE_SIZE) { in analyse_superblocks()
2519 return -EINVAL; in analyse_superblocks()
2531 if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) in analyse_superblocks()
2548 if (rs_is_raid0(rs)) in analyse_superblocks()
2554 * mistakenly acting as if '- -' got passed into the constructor). in analyse_superblocks()
2559 rdev->raid_disk = rdev->saved_raid_disk = -1; in analyse_superblocks()
2571 rs->ti->error = "Unable to assemble array: Invalid superblocks"; in analyse_superblocks()
2572 if (super_validate(rs, freshest)) in analyse_superblocks()
2573 return -EINVAL; in analyse_superblocks()
2575 if (validate_raid_redundancy(rs)) { in analyse_superblocks()
2576 rs->ti->error = "Insufficient redundancy to activate array"; in analyse_superblocks()
2577 return -EINVAL; in analyse_superblocks()
2581 if (!test_bit(Journal, &rdev->flags) && in analyse_superblocks()
2583 super_validate(rs, rdev)) in analyse_superblocks()
2584 return -EINVAL; in analyse_superblocks()
2589 * Adjust data_offset and new_data_offset on all disk members of @rs
2596 static int rs_adjust_data_offsets(struct raid_set *rs) in rs_adjust_data_offsets() argument
2602 if (!test_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) { in rs_adjust_data_offsets()
2603 if (!rs_is_reshapable(rs)) in rs_adjust_data_offsets()
2610 rdev = &rs->dev[0].rdev; in rs_adjust_data_offsets()
2612 if (rs->delta_disks < 0) { in rs_adjust_data_offsets()
2616 * - before reshape: data is at offset 0 and free space in rs_adjust_data_offsets()
2619 * - after reshape: data is at offset rs->data_offset != 0 on each component LV in rs_adjust_data_offsets()
2622 new_data_offset = rs->data_offset; in rs_adjust_data_offsets()
2624 } else if (rs->delta_disks > 0) { in rs_adjust_data_offsets()
2628 * - before reshape: data is at offset rs->data_offset != 0 and in rs_adjust_data_offsets()
2631 * - after reshape: data is at offset 0 on each component LV in rs_adjust_data_offsets()
2633 data_offset = rs->data_offset; in rs_adjust_data_offsets()
2640 * - or - (data offset != 0) in rs_adjust_data_offsets()
2642 * Changing RAID layout or chunk size -> toggle offsets in rs_adjust_data_offsets()
2644 * - before reshape: data is at offset rs->data_offset 0 and in rs_adjust_data_offsets()
2646 * -or- in rs_adjust_data_offsets()
2647 * data is at offset rs->data_offset != 0 and in rs_adjust_data_offsets()
2650 * - after reshape: data is at offset 0 if it was at offset != 0 in rs_adjust_data_offsets()
2655 data_offset = rs->data_offset ? rdev->data_offset : 0; in rs_adjust_data_offsets()
2656 new_data_offset = data_offset ? 0 : rs->data_offset; in rs_adjust_data_offsets()
2657 set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); in rs_adjust_data_offsets()
2663 if (rs->data_offset && in rs_adjust_data_offsets()
2664 bdev_nr_sectors(rdev->bdev) - rs->md.dev_sectors < MIN_FREE_RESHAPE_SPACE) { in rs_adjust_data_offsets()
2665 rs->ti->error = data_offset ? "No space for forward reshape" : in rs_adjust_data_offsets()
2667 return -ENOSPC; in rs_adjust_data_offsets()
2674 if (rs->md.recovery_cp < rs->md.dev_sectors) in rs_adjust_data_offsets()
2675 rs->md.recovery_cp += rs->dev[0].rdev.data_offset; in rs_adjust_data_offsets()
2678 rdev_for_each(rdev, &rs->md) { in rs_adjust_data_offsets()
2679 if (!test_bit(Journal, &rdev->flags)) { in rs_adjust_data_offsets()
2680 rdev->data_offset = data_offset; in rs_adjust_data_offsets()
2681 rdev->new_data_offset = new_data_offset; in rs_adjust_data_offsets()
2688 /* Userpace reordered disks -> adjust raid_disk indexes in @rs */
2689 static void __reorder_raid_disk_indexes(struct raid_set *rs) in __reorder_raid_disk_indexes() argument
2694 rdev_for_each(rdev, &rs->md) { in __reorder_raid_disk_indexes()
2695 if (!test_bit(Journal, &rdev->flags)) { in __reorder_raid_disk_indexes()
2696 rdev->raid_disk = i++; in __reorder_raid_disk_indexes()
2697 rdev->saved_raid_disk = rdev->new_raid_disk = -1; in __reorder_raid_disk_indexes()
2703 * Setup @rs for takeover by a different raid level
2705 static int rs_setup_takeover(struct raid_set *rs) in rs_setup_takeover() argument
2707 struct mddev *mddev = &rs->md; in rs_setup_takeover()
2709 unsigned int d = mddev->raid_disks = rs->raid_disks; in rs_setup_takeover()
2710 sector_t new_data_offset = rs->dev[0].rdev.data_offset ? 0 : rs->data_offset; in rs_setup_takeover()
2712 if (rt_is_raid10(rs->raid_type)) { in rs_setup_takeover()
2713 if (rs_is_raid0(rs)) { in rs_setup_takeover()
2714 /* Userpace reordered disks -> adjust raid_disk indexes */ in rs_setup_takeover()
2715 __reorder_raid_disk_indexes(rs); in rs_setup_takeover()
2717 /* raid0 -> raid10_far layout */ in rs_setup_takeover()
2718 mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_FAR, in rs_setup_takeover()
2719 rs->raid10_copies); in rs_setup_takeover()
2720 } else if (rs_is_raid1(rs)) in rs_setup_takeover()
2721 /* raid1 -> raid10_near layout */ in rs_setup_takeover()
2722 mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_NEAR, in rs_setup_takeover()
2723 rs->raid_disks); in rs_setup_takeover()
2725 return -EINVAL; in rs_setup_takeover()
2729 clear_bit(MD_ARRAY_FIRST_USE, &mddev->flags); in rs_setup_takeover()
2730 mddev->recovery_cp = MaxSector; in rs_setup_takeover()
2732 while (d--) { in rs_setup_takeover()
2733 rdev = &rs->dev[d].rdev; in rs_setup_takeover()
2735 if (test_bit(d, (void *) rs->rebuild_disks)) { in rs_setup_takeover()
2736 clear_bit(In_sync, &rdev->flags); in rs_setup_takeover()
2737 clear_bit(Faulty, &rdev->flags); in rs_setup_takeover()
2738 mddev->recovery_cp = rdev->recovery_offset = 0; in rs_setup_takeover()
2740 set_bit(MD_ARRAY_FIRST_USE, &mddev->flags); in rs_setup_takeover()
2743 rdev->new_data_offset = new_data_offset; in rs_setup_takeover()
2749 /* Prepare @rs for reshape */
2750 static int rs_prepare_reshape(struct raid_set *rs) in rs_prepare_reshape() argument
2753 struct mddev *mddev = &rs->md; in rs_prepare_reshape()
2755 if (rs_is_raid10(rs)) { in rs_prepare_reshape()
2756 if (rs->raid_disks != mddev->raid_disks && in rs_prepare_reshape()
2757 __is_raid10_near(mddev->layout) && in rs_prepare_reshape()
2758 rs->raid10_copies && in rs_prepare_reshape()
2759 rs->raid10_copies != __raid10_near_copies(mddev->layout)) { in rs_prepare_reshape()
2766 if (rs->raid_disks % rs->raid10_copies) { in rs_prepare_reshape()
2767 rs->ti->error = "Can't reshape raid10 mirror groups"; in rs_prepare_reshape()
2768 return -EINVAL; in rs_prepare_reshape()
2771 /* Userpace reordered disks to add/remove mirrors -> adjust raid_disk indexes */ in rs_prepare_reshape()
2772 __reorder_raid_disk_indexes(rs); in rs_prepare_reshape()
2773 mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_NEAR, in rs_prepare_reshape()
2774 rs->raid10_copies); in rs_prepare_reshape()
2775 mddev->new_layout = mddev->layout; in rs_prepare_reshape()
2780 } else if (rs_is_raid456(rs)) in rs_prepare_reshape()
2783 else if (rs_is_raid1(rs)) { in rs_prepare_reshape()
2784 if (rs->delta_disks) { in rs_prepare_reshape()
2786 mddev->degraded = rs->delta_disks < 0 ? -rs->delta_disks : rs->delta_disks; in rs_prepare_reshape()
2790 mddev->raid_disks = rs->raid_disks; in rs_prepare_reshape()
2794 rs->ti->error = "Called with bogus raid type"; in rs_prepare_reshape()
2795 return -EINVAL; in rs_prepare_reshape()
2799 set_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags); in rs_prepare_reshape()
2800 set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); in rs_prepare_reshape()
2801 } else if (mddev->raid_disks < rs->raid_disks) in rs_prepare_reshape()
2803 set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); in rs_prepare_reshape()
2809 static sector_t _get_reshape_sectors(struct raid_set *rs) in _get_reshape_sectors() argument
2814 rdev_for_each(rdev, &rs->md) in _get_reshape_sectors()
2815 if (!test_bit(Journal, &rdev->flags)) { in _get_reshape_sectors()
2816 reshape_sectors = (rdev->data_offset > rdev->new_data_offset) ? in _get_reshape_sectors()
2817 rdev->data_offset - rdev->new_data_offset : in _get_reshape_sectors()
2818 rdev->new_data_offset - rdev->data_offset; in _get_reshape_sectors()
2822 return max(reshape_sectors, (sector_t) rs->data_offset); in _get_reshape_sectors()
2827 * - change raid layout
2828 * - change chunk size
2829 * - add disks
2830 * - remove disks
2832 static int rs_setup_reshape(struct raid_set *rs) in rs_setup_reshape() argument
2836 sector_t reshape_sectors = _get_reshape_sectors(rs); in rs_setup_reshape()
2837 struct mddev *mddev = &rs->md; in rs_setup_reshape()
2840 mddev->delta_disks = rs->delta_disks; in rs_setup_reshape()
2841 cur_raid_devs = mddev->raid_disks; in rs_setup_reshape()
2844 if (mddev->delta_disks && in rs_setup_reshape()
2845 mddev->layout != mddev->new_layout) { in rs_setup_reshape()
2846 DMINFO("Ignoring invalid layout change with delta_disks=%d", rs->delta_disks); in rs_setup_reshape()
2847 mddev->new_layout = mddev->layout; in rs_setup_reshape()
2853 * - in case of adding disk(s), array size has in rs_setup_reshape()
2859 * - in case of removing disk(s), array size in rs_setup_reshape()
2865 * - data_offset and new_data_offset are in rs_setup_reshape()
2873 if (rs->delta_disks > 0) { in rs_setup_reshape()
2875 for (d = cur_raid_devs; d < rs->raid_disks; d++) { in rs_setup_reshape()
2876 rdev = &rs->dev[d].rdev; in rs_setup_reshape()
2877 clear_bit(In_sync, &rdev->flags); in rs_setup_reshape()
2880 * save_raid_disk needs to be -1, or recovery_offset will be set to 0 in rs_setup_reshape()
2883 rdev->saved_raid_disk = -1; in rs_setup_reshape()
2884 rdev->raid_disk = d; in rs_setup_reshape()
2886 rdev->sectors = mddev->dev_sectors; in rs_setup_reshape()
2887 rdev->recovery_offset = rs_is_raid1(rs) ? 0 : MaxSector; in rs_setup_reshape()
2890 mddev->reshape_backwards = 0; /* adding disk(s) -> forward reshape */ in rs_setup_reshape()
2893 } else if (rs->delta_disks < 0) { in rs_setup_reshape()
2894 r = rs_set_dev_and_array_sectors(rs, rs->ti->len, true); in rs_setup_reshape()
2895 mddev->reshape_backwards = 1; /* removing disk(s) -> backward reshape */ in rs_setup_reshape()
2900 * Reshape layout (e.g. raid5_ls -> raid5_n) and/or chunk size: in rs_setup_reshape()
2902 * keeping number of disks and do layout change -> in rs_setup_reshape()
2906 * - free space upfront -> reshape forward in rs_setup_reshape()
2908 * - free space at the end -> reshape backward in rs_setup_reshape()
2917 * add disk(s) -> begin in rs_setup_reshape()
2918 * remove disk(s)-> end in rs_setup_reshape()
2920 mddev->reshape_backwards = rs->dev[0].rdev.data_offset ? 0 : 1; in rs_setup_reshape()
2927 if (!mddev->reshape_backwards) in rs_setup_reshape()
2928 rdev_for_each(rdev, &rs->md) in rs_setup_reshape()
2929 if (!test_bit(Journal, &rdev->flags)) in rs_setup_reshape()
2930 rdev->sectors += reshape_sectors; in rs_setup_reshape()
2938 * changes -> reset the latter.
2940 static void rs_reset_inconclusive_reshape(struct raid_set *rs) in rs_reset_inconclusive_reshape() argument
2942 if (!rs_is_reshaping(rs) && rs_is_layout_change(rs, true)) { in rs_reset_inconclusive_reshape()
2943 rs_set_cur(rs); in rs_reset_inconclusive_reshape()
2944 rs->md.delta_disks = 0; in rs_reset_inconclusive_reshape()
2945 rs->md.reshape_backwards = 0; in rs_reset_inconclusive_reshape()
2953 static void configure_discard_support(struct raid_set *rs) in configure_discard_support() argument
2957 struct dm_target *ti = rs->ti; in configure_discard_support()
2962 raid456 = rs_is_raid456(rs); in configure_discard_support()
2964 for (i = 0; i < rs->raid_disks; i++) { in configure_discard_support()
2965 if (!rs->dev[i].rdev.bdev || in configure_discard_support()
2966 !bdev_max_discard_sectors(rs->dev[i].rdev.bdev)) in configure_discard_support()
2972 DMERR("Set dm-raid.devices_handle_discard_safely=Y to override."); in configure_discard_support()
2978 ti->num_discard_bios = 1; in configure_discard_support()
3001 struct raid_set *rs = NULL; in raid_ctr() local
3012 ti->error = "No arguments"; in raid_ctr()
3013 return -EINVAL; in raid_ctr()
3018 ti->error = "Unrecognised raid_type"; in raid_ctr()
3019 return -EINVAL; in raid_ctr()
3023 if (dm_read_arg_group(_args, &as, &num_raid_params, &ti->error)) in raid_ctr()
3024 return -EINVAL; in raid_ctr()
3029 _args[1].max = (as_nrd.argc - 1) / 2; in raid_ctr()
3030 if (dm_read_arg(_args + 1, &as_nrd, &num_raid_devs, &ti->error)) in raid_ctr()
3031 return -EINVAL; in raid_ctr()
3034 ti->error = "Invalid number of supplied raid devices"; in raid_ctr()
3035 return -EINVAL; in raid_ctr()
3038 rs = raid_set_alloc(ti, rt, num_raid_devs); in raid_ctr()
3039 if (IS_ERR(rs)) in raid_ctr()
3040 return PTR_ERR(rs); in raid_ctr()
3042 r = parse_raid_params(rs, &as, num_raid_params); in raid_ctr()
3046 r = parse_dev_params(rs, &as); in raid_ctr()
3050 rs->md.sync_super = super_sync; in raid_ctr()
3058 r = rs_set_dev_and_array_sectors(rs, rs->ti->len, false); in raid_ctr()
3063 rs->array_sectors = rs->md.array_sectors; in raid_ctr()
3064 rs->dev_sectors = rs->md.dev_sectors; in raid_ctr()
3071 rs_config_backup(rs, &rs_layout); in raid_ctr()
3073 r = analyse_superblocks(ti, rs); in raid_ctr()
3077 /* All in-core metadata now as of current superblocks after calling analyse_superblocks() */ in raid_ctr()
3078 sb_array_sectors = rs->md.array_sectors; in raid_ctr()
3079 rdev_sectors = __rdev_sectors(rs); in raid_ctr()
3081 ti->error = "Invalid rdev size"; in raid_ctr()
3082 r = -EINVAL; in raid_ctr()
3087 reshape_sectors = _get_reshape_sectors(rs); in raid_ctr()
3088 if (rs->dev_sectors != rdev_sectors) { in raid_ctr()
3089 resize = (rs->dev_sectors != rdev_sectors - reshape_sectors); in raid_ctr()
3090 if (rs->dev_sectors > rdev_sectors - reshape_sectors) in raid_ctr()
3091 set_bit(RT_FLAG_RS_GROW, &rs->runtime_flags); in raid_ctr()
3094 INIT_WORK(&rs->md.event_work, do_table_event); in raid_ctr()
3095 ti->private = rs; in raid_ctr()
3096 ti->num_flush_bios = 1; in raid_ctr()
3097 ti->needs_bio_set_dev = true; in raid_ctr()
3100 rs_config_restore(rs, &rs_layout); in raid_ctr()
3108 if (test_bit(MD_ARRAY_FIRST_USE, &rs->md.flags)) { in raid_ctr()
3109 /* A new raid6 set has to be recovered to ensure proper parity and Q-Syndrome */ in raid_ctr()
3110 if (rs_is_raid6(rs) && in raid_ctr()
3111 test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) { in raid_ctr()
3112 ti->error = "'nosync' not allowed for new raid6 set"; in raid_ctr()
3113 r = -EINVAL; in raid_ctr()
3116 rs_setup_recovery(rs, 0); in raid_ctr()
3117 set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); in raid_ctr()
3118 rs_set_new(rs); in raid_ctr()
3119 } else if (rs_is_recovering(rs)) { in raid_ctr()
3122 } else if (rs_is_reshaping(rs)) { in raid_ctr()
3125 ti->error = "Can't resize a reshaping raid set"; in raid_ctr()
3126 r = -EPERM; in raid_ctr()
3129 /* skip setup rs */ in raid_ctr()
3130 } else if (rs_takeover_requested(rs)) { in raid_ctr()
3131 if (rs_is_reshaping(rs)) { in raid_ctr()
3132 ti->error = "Can't takeover a reshaping raid set"; in raid_ctr()
3133 r = -EPERM; in raid_ctr()
3138 if (test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags)) { in raid_ctr()
3139 ti->error = "Can't takeover a journaled raid4/5/6 set"; in raid_ctr()
3140 r = -EPERM; in raid_ctr()
3152 r = rs_check_takeover(rs); in raid_ctr()
3156 r = rs_setup_takeover(rs); in raid_ctr()
3160 set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); in raid_ctr()
3162 rs_setup_recovery(rs, MaxSector); in raid_ctr()
3163 rs_set_new(rs); in raid_ctr()
3164 } else if (rs_reshape_requested(rs)) { in raid_ctr()
3166 clear_bit(RT_FLAG_RS_GROW, &rs->runtime_flags); in raid_ctr()
3174 if (test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags)) { in raid_ctr()
3175 ti->error = "Can't reshape a journaled raid4/5/6 set"; in raid_ctr()
3176 r = -EPERM; in raid_ctr()
3180 /* Out-of-place space has to be available to allow for a reshape unless raid1! */ in raid_ctr()
3181 if (reshape_sectors || rs_is_raid1(rs)) { in raid_ctr()
3189 r = rs_prepare_reshape(rs); in raid_ctr()
3194 rs_setup_recovery(rs, MaxSector); in raid_ctr()
3196 rs_set_cur(rs); in raid_ctr()
3200 if (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags)) { in raid_ctr()
3201 clear_bit(RT_FLAG_RS_GROW, &rs->runtime_flags); in raid_ctr()
3202 set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); in raid_ctr()
3203 rs_setup_recovery(rs, MaxSector); in raid_ctr()
3204 } else if (test_bit(RT_FLAG_RS_GROW, &rs->runtime_flags)) { in raid_ctr()
3209 r = rs_set_dev_and_array_sectors(rs, sb_array_sectors, false); in raid_ctr()
3213 …rs_setup_recovery(rs, rs->md.recovery_cp < rs->md.dev_sectors ? rs->md.recovery_cp : rs->md.dev_se… in raid_ctr()
3216 r = rs_set_dev_and_array_sectors(rs, rs->ti->len, false); in raid_ctr()
3220 if (sb_array_sectors > rs->array_sectors) in raid_ctr()
3221 set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); in raid_ctr()
3223 rs_set_cur(rs); in raid_ctr()
3227 r = rs_adjust_data_offsets(rs); in raid_ctr()
3232 rs_reset_inconclusive_reshape(rs); in raid_ctr()
3234 /* Start raid set read-only and assumed clean to change in raid_resume() */ in raid_ctr()
3235 rs->md.ro = 1; in raid_ctr()
3236 rs->md.in_sync = 1; in raid_ctr()
3239 set_bit(MD_RECOVERY_FROZEN, &rs->md.recovery); in raid_ctr()
3242 mddev_lock_nointr(&rs->md); in raid_ctr()
3243 r = md_run(&rs->md); in raid_ctr()
3244 rs->md.in_sync = 0; /* Assume already marked dirty */ in raid_ctr()
3246 ti->error = "Failed to run raid array"; in raid_ctr()
3247 mddev_unlock(&rs->md); in raid_ctr()
3251 r = md_start(&rs->md); in raid_ctr()
3253 ti->error = "Failed to start raid array"; in raid_ctr()
3257 /* If raid4/5/6 journal mode explicitly requested (only possible with journal dev) -> set it */ in raid_ctr()
3258 if (test_bit(__CTR_FLAG_JOURNAL_MODE, &rs->ctr_flags)) { in raid_ctr()
3259 r = r5c_journal_mode_set(&rs->md, rs->journal_dev.mode); in raid_ctr()
3261 ti->error = "Failed to set raid4/5/6 journal mode"; in raid_ctr()
3266 mddev_suspend(&rs->md); in raid_ctr()
3267 set_bit(RT_FLAG_RS_SUSPENDED, &rs->runtime_flags); in raid_ctr()
3270 if (rs_is_raid456(rs)) { in raid_ctr()
3271 r = rs_set_raid456_stripe_cache(rs); in raid_ctr()
3277 if (test_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags)) { in raid_ctr()
3278 r = rs_check_reshape(rs); in raid_ctr()
3283 rs_config_restore(rs, &rs_layout); in raid_ctr()
3285 if (rs->md.pers->start_reshape) { in raid_ctr()
3286 r = rs->md.pers->check_reshape(&rs->md); in raid_ctr()
3288 ti->error = "Reshape check failed"; in raid_ctr()
3295 configure_discard_support(rs); in raid_ctr()
3297 mddev_unlock(&rs->md); in raid_ctr()
3301 md_stop(&rs->md); in raid_ctr()
3302 mddev_unlock(&rs->md); in raid_ctr()
3304 raid_set_free(rs); in raid_ctr()
3311 struct raid_set *rs = ti->private; in raid_dtr() local
3313 mddev_lock_nointr(&rs->md); in raid_dtr()
3314 md_stop(&rs->md); in raid_dtr()
3315 mddev_unlock(&rs->md); in raid_dtr()
3316 raid_set_free(rs); in raid_dtr()
3321 struct raid_set *rs = ti->private; in raid_map() local
3322 struct mddev *mddev = &rs->md; in raid_map()
3325 * If we're reshaping to add disk(s), ti->len and in raid_map()
3326 * mddev->array_sectors will differ during the process in raid_map()
3327 * (ti->len > mddev->array_sectors), so we have to requeue in raid_map()
3328 * bios with addresses > mddev->array_sectors here or in raid_map()
3332 if (unlikely(bio_has_data(bio) && bio_end_sector(bio) > mddev->array_sectors)) in raid_map()
3355 return __within_range(state, 0, ARRAY_SIZE(sync_strs) - 1) ? sync_strs[state] : "undef"; in sync_str()
3367 (!mddev->ro && test_bit(MD_RECOVERY_NEEDED, &recovery)))) { in decipher_sync_action()
3382 if (mddev->reshape_position != MaxSector) in decipher_sync_action()
3395 * 'a' = Alive but not in-sync raid set component _or_ alive raid4/5/6 'write_back' journal device
3396 * 'A' = Alive and in-sync raid set component _or_ alive raid4/5/6 'write_through' journal device
3397 * '-' = Non-existing device (i.e. uspace passed '- -' into the ctr)
3399 static const char *__raid_dev_status(struct raid_set *rs, struct md_rdev *rdev) in __raid_dev_status() argument
3401 if (!rdev->bdev) in __raid_dev_status()
3402 return "-"; in __raid_dev_status()
3403 else if (test_bit(Faulty, &rdev->flags)) in __raid_dev_status()
3405 else if (test_bit(Journal, &rdev->flags)) in __raid_dev_status()
3406 return (rs->journal_dev.mode == R5C_JOURNAL_MODE_WRITE_THROUGH) ? "A" : "a"; in __raid_dev_status()
3407 else if (test_bit(RT_FLAG_RS_RESYNCING, &rs->runtime_flags) || in __raid_dev_status()
3408 (!test_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags) && in __raid_dev_status()
3409 !test_bit(In_sync, &rdev->flags))) in __raid_dev_status()
3415 /* Helper to return resync/reshape progress for @rs and runtime flags for raid set in sync / resync…
3416 static sector_t rs_get_progress(struct raid_set *rs, unsigned long recovery, in rs_get_progress() argument
3420 struct mddev *mddev = &rs->md; in rs_get_progress()
3422 clear_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags); in rs_get_progress()
3423 clear_bit(RT_FLAG_RS_RESYNCING, &rs->runtime_flags); in rs_get_progress()
3425 if (rs_is_raid0(rs)) { in rs_get_progress()
3427 set_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags); in rs_get_progress()
3431 r = mddev->recovery_cp; in rs_get_progress()
3433 r = mddev->curr_resync_completed; in rs_get_progress()
3441 set_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags); in rs_get_progress()
3458 set_bit(RT_FLAG_RS_RESYNCING, &rs->runtime_flags); in rs_get_progress()
3466 set_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags); in rs_get_progress()
3471 * caused by components still set to in-sync by constructor. in rs_get_progress()
3473 set_bit(RT_FLAG_RS_RESYNCING, &rs->runtime_flags); in rs_get_progress()
3484 set_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags); in rs_get_progress()
3486 if (!test_bit(Journal, &rdev->flags) && in rs_get_progress()
3487 !test_bit(In_sync, &rdev->flags)) { in rs_get_progress()
3488 clear_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags); in rs_get_progress()
3497 /* Helper to return @dev name or "-" if !@dev */
3500 return dev ? dev->name : "-"; in __get_dev_name()
3506 struct raid_set *rs = ti->private; in raid_status() local
3507 struct mddev *mddev = &rs->md; in raid_status()
3508 struct r5conf *conf = rs_is_raid456(rs) ? mddev->private : NULL; in raid_status()
3509 int i, max_nr_stripes = conf ? conf->max_nr_stripes : 0; in raid_status()
3521 rt = get_raid_type_by_ll(mddev->new_level, mddev->new_layout); in raid_status()
3525 DMEMIT("%s %d ", rt->name, mddev->raid_disks); in raid_status()
3530 resync_max_sectors = test_bit(RT_FLAG_RS_PRERESUMED, &rs->runtime_flags) ? in raid_status()
3531 mddev->resync_max_sectors : mddev->dev_sectors; in raid_status()
3532 recovery = rs->md.recovery; in raid_status()
3534 progress = rs_get_progress(rs, recovery, state, resync_max_sectors); in raid_status()
3535 resync_mismatches = (mddev->last_sync_action && !strcasecmp(mddev->last_sync_action, "check")) ? in raid_status()
3536 atomic64_read(&mddev->resync_mismatches) : 0; in raid_status()
3538 /* HM FIXME: do we want another state char for raid0? It shows 'D'/'A'/'-' now */ in raid_status()
3539 for (i = 0; i < rs->raid_disks; i++) in raid_status()
3540 DMEMIT(__raid_dev_status(rs, &rs->dev[i].rdev)); in raid_status()
3543 * In-sync/Reshape ratio: in raid_status()
3544 * The in-sync ratio shows the progress of: in raid_status()
3545 * - Initializing the raid set in raid_status()
3546 * - Rebuilding a subset of devices of the raid set in raid_status()
3561 * See Documentation/admin-guide/device-mapper/dm-raid.rst for in raid_status()
3585 DMEMIT(" %llu", (unsigned long long) rs->dev[0].rdev.data_offset); in raid_status()
3590 DMEMIT(" %s", test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags) ? in raid_status()
3591 __raid_dev_status(rs, &rs->journal_dev.rdev) : "-"); in raid_status()
3601 for (i = 0; i < rs->raid_disks; i++) { in raid_status()
3602 rebuild_writemostly_count += (test_bit(i, (void *) rs->rebuild_disks) ? 2 : 0) + in raid_status()
3603 (test_bit(WriteMostly, &rs->dev[i].rdev.flags) ? 2 : 0); in raid_status()
3605 rebuild_writemostly_count -= (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags) ? 2 : 0) + in raid_status()
3606 (test_bit(__CTR_FLAG_WRITE_MOSTLY, &rs->ctr_flags) ? 2 : 0); in raid_status()
3609 hweight32(rs->ctr_flags & CTR_FLAG_OPTIONS_NO_ARGS) + in raid_status()
3610 hweight32(rs->ctr_flags & CTR_FLAG_OPTIONS_ONE_ARG) * 2; in raid_status()
3613 DMEMIT("%s %u %u", rs->raid_type->name, raid_param_cnt, mddev->new_chunk_sectors); in raid_status()
3614 if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) in raid_status()
3616 if (test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) in raid_status()
3618 if (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags)) in raid_status()
3619 for (i = 0; i < rs->raid_disks; i++) in raid_status()
3620 if (test_bit(i, (void *) rs->rebuild_disks)) in raid_status()
3622 if (test_bit(__CTR_FLAG_DAEMON_SLEEP, &rs->ctr_flags)) in raid_status()
3624 mddev->bitmap_info.daemon_sleep); in raid_status()
3625 if (test_bit(__CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags)) in raid_status()
3627 mddev->sync_speed_min); in raid_status()
3628 if (test_bit(__CTR_FLAG_MAX_RECOVERY_RATE, &rs->ctr_flags)) in raid_status()
3630 mddev->sync_speed_max); in raid_status()
3631 if (test_bit(__CTR_FLAG_WRITE_MOSTLY, &rs->ctr_flags)) in raid_status()
3632 for (i = 0; i < rs->raid_disks; i++) in raid_status()
3633 if (test_bit(WriteMostly, &rs->dev[i].rdev.flags)) in raid_status()
3635 rs->dev[i].rdev.raid_disk); in raid_status()
3636 if (test_bit(__CTR_FLAG_MAX_WRITE_BEHIND, &rs->ctr_flags)) in raid_status()
3638 mddev->bitmap_info.max_write_behind); in raid_status()
3639 if (test_bit(__CTR_FLAG_STRIPE_CACHE, &rs->ctr_flags)) in raid_status()
3642 if (test_bit(__CTR_FLAG_REGION_SIZE, &rs->ctr_flags)) in raid_status()
3644 (unsigned long long) to_sector(mddev->bitmap_info.chunksize)); in raid_status()
3645 if (test_bit(__CTR_FLAG_RAID10_COPIES, &rs->ctr_flags)) in raid_status()
3647 raid10_md_layout_to_copies(mddev->layout)); in raid_status()
3648 if (test_bit(__CTR_FLAG_RAID10_FORMAT, &rs->ctr_flags)) in raid_status()
3650 raid10_md_layout_to_format(mddev->layout)); in raid_status()
3651 if (test_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) in raid_status()
3653 max(rs->delta_disks, mddev->delta_disks)); in raid_status()
3654 if (test_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) in raid_status()
3656 (unsigned long long) rs->data_offset); in raid_status()
3657 if (test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags)) in raid_status()
3659 __get_dev_name(rs->journal_dev.dev)); in raid_status()
3660 if (test_bit(__CTR_FLAG_JOURNAL_MODE, &rs->ctr_flags)) in raid_status()
3662 md_journal_mode_to_dm_raid(rs->journal_dev.mode)); in raid_status()
3663 DMEMIT(" %d", rs->raid_disks); in raid_status()
3664 for (i = 0; i < rs->raid_disks; i++) in raid_status()
3665 DMEMIT(" %s %s", __get_dev_name(rs->dev[i].meta_dev), in raid_status()
3666 __get_dev_name(rs->dev[i].data_dev)); in raid_status()
3670 rt = get_raid_type_by_ll(mddev->new_level, mddev->new_layout); in raid_status()
3674 DMEMIT_TARGET_NAME_VERSION(ti->type); in raid_status()
3675 DMEMIT(",raid_type=%s,raid_disks=%d", rt->name, mddev->raid_disks); in raid_status()
3679 recovery = rs->md.recovery; in raid_status()
3683 for (i = 0; i < rs->raid_disks; i++) { in raid_status()
3685 DMEMIT(__raid_dev_status(rs, &rs->dev[i].rdev)); in raid_status()
3690 switch (rs->journal_dev.mode) { in raid_status()
3712 struct raid_set *rs = ti->private; in raid_message() local
3713 struct mddev *mddev = &rs->md; in raid_message()
3715 if (!mddev->pers || !mddev->pers->sync_request) in raid_message()
3716 return -EINVAL; in raid_message()
3719 set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); in raid_message()
3721 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); in raid_message()
3724 if (mddev->sync_thread) { in raid_message()
3725 set_bit(MD_RECOVERY_INTR, &mddev->recovery); in raid_message()
3728 } else if (decipher_sync_action(mddev, mddev->recovery) != st_idle) in raid_message()
3729 return -EBUSY; in raid_message()
3733 set_bit(MD_RECOVERY_RECOVER, &mddev->recovery); in raid_message()
3736 set_bit(MD_RECOVERY_CHECK, &mddev->recovery); in raid_message()
3737 set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery); in raid_message()
3738 set_bit(MD_RECOVERY_SYNC, &mddev->recovery); in raid_message()
3740 set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery); in raid_message()
3741 set_bit(MD_RECOVERY_SYNC, &mddev->recovery); in raid_message()
3743 return -EINVAL; in raid_message()
3745 if (mddev->ro == 2) { in raid_message()
3747 * canceling read-auto mode in raid_message()
3749 mddev->ro = 0; in raid_message()
3750 if (!mddev->suspended) in raid_message()
3751 md_wakeup_thread(mddev->sync_thread); in raid_message()
3753 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); in raid_message()
3754 if (!mddev->suspended) in raid_message()
3755 md_wakeup_thread(mddev->thread); in raid_message()
3763 struct raid_set *rs = ti->private; in raid_iterate_devices() local
3767 for (i = 0; !r && i < rs->raid_disks; i++) { in raid_iterate_devices()
3768 if (rs->dev[i].data_dev) { in raid_iterate_devices()
3769 r = fn(ti, rs->dev[i].data_dev, in raid_iterate_devices()
3771 rs->md.dev_sectors, data); in raid_iterate_devices()
3780 struct raid_set *rs = ti->private; in raid_io_hints() local
3781 unsigned int chunk_size_bytes = to_bytes(rs->md.chunk_sectors); in raid_io_hints()
3784 blk_limits_io_opt(limits, chunk_size_bytes * mddev_data_stripes(rs)); in raid_io_hints()
3789 struct raid_set *rs = ti->private; in raid_postsuspend() local
3791 if (!test_and_set_bit(RT_FLAG_RS_SUSPENDED, &rs->runtime_flags)) { in raid_postsuspend()
3793 if (!test_bit(MD_RECOVERY_FROZEN, &rs->md.recovery)) in raid_postsuspend()
3794 md_stop_writes(&rs->md); in raid_postsuspend()
3796 mddev_lock_nointr(&rs->md); in raid_postsuspend()
3797 mddev_suspend(&rs->md); in raid_postsuspend()
3798 mddev_unlock(&rs->md); in raid_postsuspend()
3802 static void attempt_restore_of_faulty_devices(struct raid_set *rs) in attempt_restore_of_faulty_devices() argument
3809 struct mddev *mddev = &rs->md; in attempt_restore_of_faulty_devices()
3813 if (!mddev->pers || !mddev->pers->hot_add_disk || !mddev->pers->hot_remove_disk) in attempt_restore_of_faulty_devices()
3818 for (i = 0; i < rs->raid_disks; i++) { in attempt_restore_of_faulty_devices()
3819 r = &rs->dev[i].rdev; in attempt_restore_of_faulty_devices()
3821 if (test_bit(Journal, &r->flags)) in attempt_restore_of_faulty_devices()
3824 if (test_bit(Faulty, &r->flags) && in attempt_restore_of_faulty_devices()
3825 r->meta_bdev && !read_disk_sb(r, r->sb_size, true)) { in attempt_restore_of_faulty_devices()
3828 rs->raid_type->name, i); in attempt_restore_of_faulty_devices()
3836 * '>= 0' - meaning we must call this function in attempt_restore_of_faulty_devices()
3839 flags = r->flags; in attempt_restore_of_faulty_devices()
3840 clear_bit(In_sync, &r->flags); /* Mandatory for hot remove. */ in attempt_restore_of_faulty_devices()
3841 if (r->raid_disk >= 0) { in attempt_restore_of_faulty_devices()
3842 if (mddev->pers->hot_remove_disk(mddev, r)) { in attempt_restore_of_faulty_devices()
3844 r->flags = flags; in attempt_restore_of_faulty_devices()
3848 r->raid_disk = r->saved_raid_disk = i; in attempt_restore_of_faulty_devices()
3850 clear_bit(Faulty, &r->flags); in attempt_restore_of_faulty_devices()
3851 clear_bit(WriteErrorSeen, &r->flags); in attempt_restore_of_faulty_devices()
3853 if (mddev->pers->hot_add_disk(mddev, r)) { in attempt_restore_of_faulty_devices()
3855 r->raid_disk = r->saved_raid_disk = -1; in attempt_restore_of_faulty_devices()
3856 r->flags = flags; in attempt_restore_of_faulty_devices()
3858 clear_bit(In_sync, &r->flags); in attempt_restore_of_faulty_devices()
3859 r->recovery_offset = 0; in attempt_restore_of_faulty_devices()
3870 rdev_for_each(r, &rs->md) { in attempt_restore_of_faulty_devices()
3871 if (test_bit(Journal, &r->flags)) in attempt_restore_of_faulty_devices()
3874 sb = page_address(r->sb_page); in attempt_restore_of_faulty_devices()
3885 static int __load_dirty_region_bitmap(struct raid_set *rs) in __load_dirty_region_bitmap() argument
3890 if (!rs_is_raid0(rs) && in __load_dirty_region_bitmap()
3891 !test_and_set_bit(RT_FLAG_RS_BITMAP_LOADED, &rs->runtime_flags)) { in __load_dirty_region_bitmap()
3892 r = md_bitmap_load(&rs->md); in __load_dirty_region_bitmap()
3901 static void rs_update_sbs(struct raid_set *rs) in rs_update_sbs() argument
3903 struct mddev *mddev = &rs->md; in rs_update_sbs()
3904 int ro = mddev->ro; in rs_update_sbs()
3906 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags); in rs_update_sbs()
3907 mddev->ro = 0; in rs_update_sbs()
3909 mddev->ro = ro; in rs_update_sbs()
3913 * Reshape changes raid algorithm of @rs to new one within personality
3914 * (e.g. raid6_zr -> raid6_nc), changes stripe size, adds/removes
3919 static int rs_start_reshape(struct raid_set *rs) in rs_start_reshape() argument
3922 struct mddev *mddev = &rs->md; in rs_start_reshape()
3923 struct md_personality *pers = mddev->pers; in rs_start_reshape()
3926 set_bit(MD_RECOVERY_WAIT, &mddev->recovery); in rs_start_reshape()
3928 r = rs_setup_reshape(rs); in rs_start_reshape()
3935 * May as well already kick the reshape off so that * pers->start_reshape() becomes optional. in rs_start_reshape()
3937 r = pers->check_reshape(mddev); in rs_start_reshape()
3939 rs->ti->error = "pers->check_reshape() failed"; in rs_start_reshape()
3947 if (pers->start_reshape) { in rs_start_reshape()
3948 r = pers->start_reshape(mddev); in rs_start_reshape()
3950 rs->ti->error = "pers->start_reshape() failed"; in rs_start_reshape()
3960 rs_update_sbs(rs); in rs_start_reshape()
3968 struct raid_set *rs = ti->private; in raid_preresume() local
3969 struct mddev *mddev = &rs->md; in raid_preresume()
3971 /* This is a resume after a suspend of the set -> it's already started. */ in raid_preresume()
3972 if (test_and_set_bit(RT_FLAG_RS_PRERESUMED, &rs->runtime_flags)) in raid_preresume()
3981 if (test_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags)) in raid_preresume()
3982 rs_update_sbs(rs); in raid_preresume()
3985 r = __load_dirty_region_bitmap(rs); in raid_preresume()
3990 if (test_bit(RT_FLAG_RS_GROW, &rs->runtime_flags)) { in raid_preresume()
3991 mddev->array_sectors = rs->array_sectors; in raid_preresume()
3992 mddev->dev_sectors = rs->dev_sectors; in raid_preresume()
3993 rs_set_rdev_sectors(rs); in raid_preresume()
3994 rs_set_capacity(rs); in raid_preresume()
3998 if (test_bit(RT_FLAG_RS_BITMAP_LOADED, &rs->runtime_flags) && mddev->bitmap && in raid_preresume()
3999 (test_bit(RT_FLAG_RS_GROW, &rs->runtime_flags) || in raid_preresume()
4000 (rs->requested_bitmap_chunk_sectors && in raid_preresume()
4001 mddev->bitmap_info.chunksize != to_bytes(rs->requested_bitmap_chunk_sectors)))) { in raid_preresume()
4002 int chunksize = to_bytes(rs->requested_bitmap_chunk_sectors) ?: mddev->bitmap_info.chunksize; in raid_preresume()
4004 r = md_bitmap_resize(mddev->bitmap, mddev->dev_sectors, chunksize, 0); in raid_preresume()
4009 /* Check for any resize/reshape on @rs and adjust/initiate */ in raid_preresume()
4011 set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); in raid_preresume()
4012 if (mddev->recovery_cp && mddev->recovery_cp < MaxSector) { in raid_preresume()
4013 set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery); in raid_preresume()
4014 mddev->resync_min = mddev->recovery_cp; in raid_preresume()
4015 if (test_bit(RT_FLAG_RS_GROW, &rs->runtime_flags)) in raid_preresume()
4016 mddev->resync_max_sectors = mddev->dev_sectors; in raid_preresume()
4020 if (test_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags)) { in raid_preresume()
4022 rs_set_rdev_sectors(rs); in raid_preresume()
4024 r = rs_start_reshape(rs); in raid_preresume()
4036 struct raid_set *rs = ti->private; in raid_resume() local
4037 struct mddev *mddev = &rs->md; in raid_resume()
4039 if (test_and_set_bit(RT_FLAG_RS_RESUMED, &rs->runtime_flags)) { in raid_resume()
4046 attempt_restore_of_faulty_devices(rs); in raid_resume()
4050 if (test_and_clear_bit(RT_FLAG_RS_SUSPENDED, &rs->runtime_flags)) { in raid_resume()
4052 if (mddev->delta_disks < 0) in raid_resume()
4053 rs_set_capacity(rs); in raid_resume()
4056 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); in raid_resume()
4057 mddev->ro = 0; in raid_resume()
4058 mddev->in_sync = 0; in raid_resume()
4086 MODULE_ALIAS("dm-raid0");
4087 MODULE_ALIAS("dm-raid1");
4088 MODULE_ALIAS("dm-raid10");
4089 MODULE_ALIAS("dm-raid4");
4090 MODULE_ALIAS("dm-raid5");
4091 MODULE_ALIAS("dm-raid6");
4092 MODULE_AUTHOR("Neil Brown <dm-devel@redhat.com>");
4093 MODULE_AUTHOR("Heinz Mauelshagen <dm-devel@redhat.com>");