Lines Matching +full:conf +full:- +full:pu
1 // SPDX-License-Identifier: GPL-2.0-or-later
10 - RAID-1/RAID-5 extensions by Miguel de Icaza, Gadi Oxman, Ingo Molnar
11 - RAID-6 extensions by H. Peter Anvin <hpa@zytor.com>
12 - boot support for linear and striped mode by Harald Hoyer <HarryH@Royal.Net>
13 - kerneld support by Boris Tobotras <boris@xtalk.msk.su>
14 - kmod support by: Cyrus Durgin
15 - RAID0 bugfixes: Mark Anthony Lisher <markal@iname.com>
16 - Devfs support by Richard Gooch <rgooch@atnf.csiro.au>
18 - lots of fixes and improvements to the RAID1/RAID5 and generic
23 - persistent bitmap code
24 Copyright (C) 2003-2004, Paul Clements, SteelEye Technology, Inc.
63 #include <linux/percpu-refcount.h>
68 #include "md-bitmap.h"
69 #include "md-cluster.h"
74 * mddev->thread when the mutex cannot be held.
103 * Current RAID-1,4,5 parallel reconstruction 'guaranteed speed limit'
108 * speed limit - in case reconstruction slows down your system despite
119 return mddev->sync_speed_min ?
120 mddev->sync_speed_min : sysctl_speed_limit_min;
125 return mddev->sync_speed_max ?
126 mddev->sync_speed_max : sysctl_speed_limit_max;
131 if (!test_and_clear_bit(CollisionCheck, &rdev->flags))
134 kvfree(rdev->serial);
135 rdev->serial = NULL;
149 int i, serial_nums = 1 << ((PAGE_SHIFT - ilog2(sizeof(atomic_t))));
152 if (test_bit(CollisionCheck, &rdev->flags))
158 return -ENOMEM;
163 spin_lock_init(&serial_tmp->serial_lock);
164 serial_tmp->serial_rb = RB_ROOT_CACHED;
165 init_waitqueue_head(&serial_tmp->serial_io_wait);
168 rdev->serial = serial;
169 set_bit(CollisionCheck, &rdev->flags);
186 if (ret && !mddev->serial_info_pool)
194 * 1. it is multi-queue device flaged with writemostly.
195 * 2. the write-behind mode is enabled.
199 return (rdev && rdev->mddev->bitmap_info.max_write_behind > 0 &&
200 rdev->bdev->bd_disk->queue->nr_hw_queues != 1 &&
201 test_bit(WriteMostly, &rdev->flags));
215 !test_bit(CollisionCheck, &rdev->flags))
228 if (mddev->serial_info_pool == NULL) {
233 mddev->serial_info_pool =
236 if (!mddev->serial_info_pool) {
256 if (rdev && !test_bit(CollisionCheck, &rdev->flags))
259 if (mddev->serial_info_pool) {
267 if (!mddev->serialize_policy ||
273 test_bit(CollisionCheck, &temp->flags))
283 mempool_destroy(mddev->serial_info_pool);
284 mddev->serial_info_pool = NULL;
335 * a device node in /dev and to open it. This causes races with device-close.
346 if (!mddev || !bioset_initialized(&mddev->bio_set))
349 return bio_alloc_bioset(gfp_mask, nr_iovecs, &mddev->bio_set);
355 if (!mddev || !bioset_initialized(&mddev->sync_set))
358 return bio_alloc_bioset(GFP_NOIO, 1, &mddev->sync_set);
406 _tmp = _tmp->next;}) \
412 * We hold a refcount over the call to ->make_request. By the time that
414 * and so is visible to ->quiesce(), so we don't need the refcount any more.
418 if (mddev->suspended)
422 if (mddev->suspend_lo >= mddev->suspend_hi)
424 if (bio->bi_iter.bi_sector >= mddev->suspend_hi)
426 if (bio_end_sector(bio) < mddev->suspend_lo)
438 prepare_to_wait(&mddev->sb_wait, &__wait,
446 finish_wait(&mddev->sb_wait, &__wait);
448 atomic_inc(&mddev->active_io);
451 if (!mddev->pers->make_request(mddev, bio)) {
452 atomic_dec(&mddev->active_io);
453 wake_up(&mddev->sb_wait);
457 if (atomic_dec_and_test(&mddev->active_io) && mddev->suspended)
458 wake_up(&mddev->sb_wait);
465 struct mddev *mddev = bio->bi_disk->private_data;
467 if (mddev == NULL || mddev->pers == NULL) {
472 if (unlikely(test_bit(MD_BROKEN, &mddev->flags)) && (rw == WRITE)) {
479 if (mddev->ro == 1 && unlikely(rw == WRITE)) {
481 bio->bi_status = BLK_STS_IOERR;
487 bio->bi_opf &= ~REQ_NOMERGE;
502 WARN_ON_ONCE(mddev->thread && current == mddev->thread->tsk);
503 lockdep_assert_held(&mddev->reconfig_mutex);
504 if (mddev->suspended++)
507 wake_up(&mddev->sb_wait);
508 set_bit(MD_ALLOW_SB_UPDATE, &mddev->flags);
510 wait_event(mddev->sb_wait, atomic_read(&mddev->active_io) == 0);
511 mddev->pers->quiesce(mddev, 1);
512 clear_bit_unlock(MD_ALLOW_SB_UPDATE, &mddev->flags);
513 wait_event(mddev->sb_wait, !test_bit(MD_UPDATING_SB, &mddev->flags));
515 del_timer_sync(&mddev->safemode_timer);
517 mddev->noio_flag = memalloc_noio_save();
524 memalloc_noio_restore(mddev->noio_flag);
525 lockdep_assert_held(&mddev->reconfig_mutex);
526 if (--mddev->suspended)
528 wake_up(&mddev->sb_wait);
529 mddev->pers->quiesce(mddev, 0);
531 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
532 md_wakeup_thread(mddev->thread);
533 md_wakeup_thread(mddev->sync_thread); /* possibly kick off a reshape */
543 struct md_rdev *rdev = bio->bi_private;
544 struct mddev *mddev = rdev->mddev;
550 if (atomic_dec_and_test(&mddev->flush_pending)) {
551 /* The pre-request flush has finished */
552 queue_work(md_wq, &mddev->flush_work);
563 mddev->start_flush = ktime_get_boottime();
564 INIT_WORK(&mddev->flush_work, md_submit_flush_data);
565 atomic_set(&mddev->flush_pending, 1);
568 if (rdev->raid_disk >= 0 &&
569 !test_bit(Faulty, &rdev->flags)) {
575 atomic_inc(&rdev->nr_pending);
576 atomic_inc(&rdev->nr_pending);
579 bi->bi_end_io = md_end_flush;
580 bi->bi_private = rdev;
581 bio_set_dev(bi, rdev->bdev);
582 bi->bi_opf = REQ_OP_WRITE | REQ_PREFLUSH;
583 atomic_inc(&mddev->flush_pending);
589 if (atomic_dec_and_test(&mddev->flush_pending))
590 queue_work(md_wq, &mddev->flush_work);
596 struct bio *bio = mddev->flush_bio;
604 spin_lock_irq(&mddev->lock);
605 mddev->last_flush = mddev->start_flush;
606 mddev->flush_bio = NULL;
607 spin_unlock_irq(&mddev->lock);
608 wake_up(&mddev->sb_wait);
610 if (bio->bi_iter.bi_size == 0) {
611 /* an empty barrier - all done */
614 bio->bi_opf &= ~REQ_PREFLUSH;
628 spin_lock_irq(&mddev->lock);
629 wait_event_lock_irq(mddev->sb_wait,
630 !mddev->flush_bio ||
631 ktime_after(mddev->last_flush, start),
632 mddev->lock);
633 if (!ktime_after(mddev->last_flush, start)) {
634 WARN_ON(mddev->flush_bio);
635 mddev->flush_bio = bio;
638 spin_unlock_irq(&mddev->lock);
641 INIT_WORK(&mddev->flush_work, submit_flushes);
642 queue_work(md_wq, &mddev->flush_work);
645 if (bio->bi_iter.bi_size == 0)
646 /* an empty barrier - all done */
649 bio->bi_opf &= ~REQ_PREFLUSH;
659 atomic_inc(&mddev->active);
667 if (!atomic_dec_and_lock(&mddev->active, &all_mddevs_lock))
669 if (!mddev->raid_disks && list_empty(&mddev->disks) &&
670 mddev->ctime == 0 && !mddev->hold_active) {
673 list_del_init(&mddev->all_mddevs);
680 INIT_WORK(&mddev->del_work, mddev_delayed_delete);
681 queue_work(md_misc_wq, &mddev->del_work);
690 kobject_init(&mddev->kobj, &md_ktype);
691 mutex_init(&mddev->open_mutex);
692 mutex_init(&mddev->reconfig_mutex);
693 mutex_init(&mddev->bitmap_info.mutex);
694 INIT_LIST_HEAD(&mddev->disks);
695 INIT_LIST_HEAD(&mddev->all_mddevs);
696 timer_setup(&mddev->safemode_timer, md_safemode_timeout, 0);
697 atomic_set(&mddev->active, 1);
698 atomic_set(&mddev->openers, 0);
699 atomic_set(&mddev->active_io, 0);
700 spin_lock_init(&mddev->lock);
701 atomic_set(&mddev->flush_pending, 0);
702 init_waitqueue_head(&mddev->sb_wait);
703 init_waitqueue_head(&mddev->recovery_wait);
704 mddev->reshape_position = MaxSector;
705 mddev->reshape_backwards = 0;
706 mddev->last_sync_action = "none";
707 mddev->resync_min = 0;
708 mddev->resync_max = MaxSector;
709 mddev->level = LEVEL_NONE;
718 if (mddev->unit == unit)
729 unit &= ~((1 << MdpMinorShift) - 1);
745 unit &= ~((1<<MdpMinorShift)-1);
760 list_add(&new->all_mddevs, &all_mddevs);
762 new->hold_active = UNTIL_IOCTL;
785 new->unit = dev;
786 new->md_minor = MINOR(dev);
787 new->hold_active = UNTIL_STOP;
788 list_add(&new->all_mddevs, &all_mddevs);
798 new->unit = unit;
800 new->md_minor = MINOR(unit);
802 new->md_minor = MINOR(unit) >> MdpMinorShift;
813 if (mddev->to_remove) {
819 * and anything else which might set ->to_remove or my
821 * -EBUSY if sysfs_active is still set.
826 struct attribute_group *to_remove = mddev->to_remove;
827 mddev->to_remove = NULL;
828 mddev->sysfs_active = 1;
829 mutex_unlock(&mddev->reconfig_mutex);
831 if (mddev->kobj.sd) {
833 sysfs_remove_group(&mddev->kobj, to_remove);
834 if (mddev->pers == NULL ||
835 mddev->pers->sync_request == NULL) {
836 sysfs_remove_group(&mddev->kobj, &md_redundancy_group);
837 if (mddev->sysfs_action)
838 sysfs_put(mddev->sysfs_action);
839 if (mddev->sysfs_completed)
840 sysfs_put(mddev->sysfs_completed);
841 if (mddev->sysfs_degraded)
842 sysfs_put(mddev->sysfs_degraded);
843 mddev->sysfs_action = NULL;
844 mddev->sysfs_completed = NULL;
845 mddev->sysfs_degraded = NULL;
848 mddev->sysfs_active = 0;
850 mutex_unlock(&mddev->reconfig_mutex);
856 md_wakeup_thread(mddev->thread);
857 wake_up(&mddev->sb_wait);
867 if (rdev->desc_nr == nr)
879 if (rdev->bdev->bd_dev == dev)
890 if (rdev->bdev->bd_dev == dev)
901 if (level != LEVEL_NONE && pers->level == level)
903 if (strcmp(pers->name, clevel)==0)
912 sector_t num_sectors = i_size_read(rdev->bdev->bd_inode) / 512;
918 rdev->sb_page = alloc_page(GFP_KERNEL);
919 if (!rdev->sb_page)
920 return -ENOMEM;
926 if (rdev->sb_page) {
927 put_page(rdev->sb_page);
928 rdev->sb_loaded = 0;
929 rdev->sb_page = NULL;
930 rdev->sb_start = 0;
931 rdev->sectors = 0;
933 if (rdev->bb_page) {
934 put_page(rdev->bb_page);
935 rdev->bb_page = NULL;
937 badblocks_exit(&rdev->badblocks);
943 struct md_rdev *rdev = bio->bi_private;
944 struct mddev *mddev = rdev->mddev;
946 if (bio->bi_status) {
948 blk_status_to_errno(bio->bi_status));
950 if (!test_bit(Faulty, &rdev->flags)
951 && (bio->bi_opf & MD_FAILFAST)) {
952 set_bit(MD_SB_NEED_REWRITE, &mddev->sb_flags);
953 set_bit(LastDev, &rdev->flags);
956 clear_bit(LastDev, &rdev->flags);
962 if (atomic_dec_and_test(&mddev->pending_writes))
963 wake_up(&mddev->sb_wait);
970 * Increment mddev->pending_writes before returning
981 if (test_bit(Faulty, &rdev->flags))
986 atomic_inc(&rdev->nr_pending);
988 bio_set_dev(bio, rdev->meta_bdev ? rdev->meta_bdev : rdev->bdev);
989 bio->bi_iter.bi_sector = sector;
991 bio->bi_private = rdev;
992 bio->bi_end_io = super_written;
994 if (test_bit(MD_FAILFAST_SUPPORTED, &mddev->flags) &&
995 test_bit(FailFast, &rdev->flags) &&
996 !test_bit(LastDev, &rdev->flags))
998 bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH | REQ_FUA | ff;
1000 atomic_inc(&mddev->pending_writes);
1007 wait_event(mddev->sb_wait, atomic_read(&mddev->pending_writes)==0);
1008 if (test_and_clear_bit(MD_SB_NEED_REWRITE, &mddev->sb_flags))
1009 return -EAGAIN;
1016 struct bio *bio = md_bio_alloc_sync(rdev->mddev);
1019 if (metadata_op && rdev->meta_bdev)
1020 bio_set_dev(bio, rdev->meta_bdev);
1022 bio_set_dev(bio, rdev->bdev);
1025 bio->bi_iter.bi_sector = sector + rdev->sb_start;
1026 else if (rdev->mddev->reshape_position != MaxSector &&
1027 (rdev->mddev->reshape_backwards ==
1028 (sector >= rdev->mddev->reshape_position)))
1029 bio->bi_iter.bi_sector = sector + rdev->new_data_offset;
1031 bio->bi_iter.bi_sector = sector + rdev->data_offset;
1036 ret = !bio->bi_status;
1046 if (rdev->sb_loaded)
1049 if (!sync_page_io(rdev, 0, size, rdev->sb_page, REQ_OP_READ, 0, true))
1051 rdev->sb_loaded = 1;
1056 bdevname(rdev->bdev,b));
1057 return -EINVAL;
1062 return sb1->set_uuid0 == sb2->set_uuid0 &&
1063 sb1->set_uuid1 == sb2->set_uuid1 &&
1064 sb1->set_uuid2 == sb2->set_uuid2 &&
1065 sb1->set_uuid3 == sb2->set_uuid3;
1087 tmp1->nr_disks = 0;
1088 tmp2->nr_disks = 0;
1110 disk_csum = sb->sb_csum;
1111 sb->sb_csum = 0;
1126 sb->sb_csum = md_csum_fold(disk_csum);
1128 sb->sb_csum = disk_csum;
1138 * We rely on user-space to write the initial superblock, and support
1145 * 0 - dev has a superblock that is compatible with refdev
1146 * 1 - dev has a superblock that is compatible and newer than refdev
1148 * -EINVAL superblock incompatible or invalid
1149 * -othererror e.g. -EIO
1153 * The first time, mddev->raid_disks will be 0, and data from
1155 * is new enough. Return 0 or -EINVAL
1183 * support bitmaps. It prints an error message and returns non-zero if mddev
1189 if (!mddev->bitmap_info.file && !mddev->bitmap_info.offset)
1192 mdname(mddev), mddev->pers->name);
1213 rdev->sb_start = calc_dev_sboffset(rdev);
1219 ret = -EINVAL;
1221 bdevname(rdev->bdev, b);
1222 sb = page_address(rdev->sb_page);
1224 if (sb->md_magic != MD_SB_MAGIC) {
1229 if (sb->major_version != 0 ||
1230 sb->minor_version < 90 ||
1231 sb->minor_version > 91) {
1233 sb->major_version, sb->minor_version, b);
1237 if (sb->raid_disks <= 0)
1240 if (md_csum_fold(calc_sb_csum(sb)) != md_csum_fold(sb->sb_csum)) {
1245 rdev->preferred_minor = sb->md_minor;
1246 rdev->data_offset = 0;
1247 rdev->new_data_offset = 0;
1248 rdev->sb_size = MD_SB_BYTES;
1249 rdev->badblocks.shift = -1;
1251 if (sb->level == LEVEL_MULTIPATH)
1252 rdev->desc_nr = -1;
1254 rdev->desc_nr = sb->this_disk.number;
1257 if (sb->level == LEVEL_MULTIPATH ||
1258 (rdev->desc_nr >= 0 &&
1259 rdev->desc_nr < MD_SB_DISKS &&
1260 sb->disks[rdev->desc_nr].state &
1271 mdp_super_t *refsb = page_address(refdev->sb_page);
1274 b, bdevname(refdev->bdev,b2));
1279 b, bdevname(refdev->bdev, b2));
1290 rdev->sectors = rdev->sb_start;
1295 if ((u64)rdev->sectors >= (2ULL << 32) && sb->level >= 1)
1296 rdev->sectors = (sector_t)(2ULL << 32) - 2;
1298 if (rdev->sectors < ((sector_t)sb->size) * 2 && sb->level >= 1)
1300 ret = -EINVAL;
1312 mdp_super_t *sb = page_address(rdev->sb_page);
1315 rdev->raid_disk = -1;
1316 clear_bit(Faulty, &rdev->flags);
1317 clear_bit(In_sync, &rdev->flags);
1318 clear_bit(Bitmap_sync, &rdev->flags);
1319 clear_bit(WriteMostly, &rdev->flags);
1321 if (mddev->raid_disks == 0) {
1322 mddev->major_version = 0;
1323 mddev->minor_version = sb->minor_version;
1324 mddev->patch_version = sb->patch_version;
1325 mddev->external = 0;
1326 mddev->chunk_sectors = sb->chunk_size >> 9;
1327 mddev->ctime = sb->ctime;
1328 mddev->utime = sb->utime;
1329 mddev->level = sb->level;
1330 mddev->clevel[0] = 0;
1331 mddev->layout = sb->layout;
1332 mddev->raid_disks = sb->raid_disks;
1333 mddev->dev_sectors = ((sector_t)sb->size) * 2;
1334 mddev->events = ev1;
1335 mddev->bitmap_info.offset = 0;
1336 mddev->bitmap_info.space = 0;
1338 mddev->bitmap_info.default_offset = MD_SB_BYTES >> 9;
1339 mddev->bitmap_info.default_space = 64*2 - (MD_SB_BYTES >> 9);
1340 mddev->reshape_backwards = 0;
1342 if (mddev->minor_version >= 91) {
1343 mddev->reshape_position = sb->reshape_position;
1344 mddev->delta_disks = sb->delta_disks;
1345 mddev->new_level = sb->new_level;
1346 mddev->new_layout = sb->new_layout;
1347 mddev->new_chunk_sectors = sb->new_chunk >> 9;
1348 if (mddev->delta_disks < 0)
1349 mddev->reshape_backwards = 1;
1351 mddev->reshape_position = MaxSector;
1352 mddev->delta_disks = 0;
1353 mddev->new_level = mddev->level;
1354 mddev->new_layout = mddev->layout;
1355 mddev->new_chunk_sectors = mddev->chunk_sectors;
1357 if (mddev->level == 0)
1358 mddev->layout = -1;
1360 if (sb->state & (1<<MD_SB_CLEAN))
1361 mddev->recovery_cp = MaxSector;
1363 if (sb->events_hi == sb->cp_events_hi &&
1364 sb->events_lo == sb->cp_events_lo) {
1365 mddev->recovery_cp = sb->recovery_cp;
1367 mddev->recovery_cp = 0;
1370 memcpy(mddev->uuid+0, &sb->set_uuid0, 4);
1371 memcpy(mddev->uuid+4, &sb->set_uuid1, 4);
1372 memcpy(mddev->uuid+8, &sb->set_uuid2, 4);
1373 memcpy(mddev->uuid+12,&sb->set_uuid3, 4);
1375 mddev->max_disks = MD_SB_DISKS;
1377 if (sb->state & (1<<MD_SB_BITMAP_PRESENT) &&
1378 mddev->bitmap_info.file == NULL) {
1379 mddev->bitmap_info.offset =
1380 mddev->bitmap_info.default_offset;
1381 mddev->bitmap_info.space =
1382 mddev->bitmap_info.default_space;
1385 } else if (mddev->pers == NULL) {
1389 if (sb->disks[rdev->desc_nr].state & (
1391 if (ev1 < mddev->events)
1392 return -EINVAL;
1393 } else if (mddev->bitmap) {
1397 if (ev1 < mddev->bitmap->events_cleared)
1399 if (ev1 < mddev->events)
1400 set_bit(Bitmap_sync, &rdev->flags);
1402 if (ev1 < mddev->events)
1403 /* just a hot-add of a new device, leave raid_disk at -1 */
1407 if (mddev->level != LEVEL_MULTIPATH) {
1408 desc = sb->disks + rdev->desc_nr;
1410 if (desc->state & (1<<MD_DISK_FAULTY))
1411 set_bit(Faulty, &rdev->flags);
1412 else if (desc->state & (1<<MD_DISK_SYNC) /* &&
1413 desc->raid_disk < mddev->raid_disks */) {
1414 set_bit(In_sync, &rdev->flags);
1415 rdev->raid_disk = desc->raid_disk;
1416 rdev->saved_raid_disk = desc->raid_disk;
1417 } else if (desc->state & (1<<MD_DISK_ACTIVE)) {
1421 if (mddev->minor_version >= 91) {
1422 rdev->recovery_offset = 0;
1423 rdev->raid_disk = desc->raid_disk;
1426 if (desc->state & (1<<MD_DISK_WRITEMOSTLY))
1427 set_bit(WriteMostly, &rdev->flags);
1428 if (desc->state & (1<<MD_DISK_FAILFAST))
1429 set_bit(FailFast, &rdev->flags);
1431 set_bit(In_sync, &rdev->flags);
1442 int next_spare = mddev->raid_disks;
1444 /* make rdev->sb match mddev data..
1457 rdev->sb_size = MD_SB_BYTES;
1459 sb = page_address(rdev->sb_page);
1463 sb->md_magic = MD_SB_MAGIC;
1464 sb->major_version = mddev->major_version;
1465 sb->patch_version = mddev->patch_version;
1466 sb->gvalid_words = 0; /* ignored */
1467 memcpy(&sb->set_uuid0, mddev->uuid+0, 4);
1468 memcpy(&sb->set_uuid1, mddev->uuid+4, 4);
1469 memcpy(&sb->set_uuid2, mddev->uuid+8, 4);
1470 memcpy(&sb->set_uuid3, mddev->uuid+12,4);
1472 sb->ctime = clamp_t(time64_t, mddev->ctime, 0, U32_MAX);
1473 sb->level = mddev->level;
1474 sb->size = mddev->dev_sectors / 2;
1475 sb->raid_disks = mddev->raid_disks;
1476 sb->md_minor = mddev->md_minor;
1477 sb->not_persistent = 0;
1478 sb->utime = clamp_t(time64_t, mddev->utime, 0, U32_MAX);
1479 sb->state = 0;
1480 sb->events_hi = (mddev->events>>32);
1481 sb->events_lo = (u32)mddev->events;
1483 if (mddev->reshape_position == MaxSector)
1484 sb->minor_version = 90;
1486 sb->minor_version = 91;
1487 sb->reshape_position = mddev->reshape_position;
1488 sb->new_level = mddev->new_level;
1489 sb->delta_disks = mddev->delta_disks;
1490 sb->new_layout = mddev->new_layout;
1491 sb->new_chunk = mddev->new_chunk_sectors << 9;
1493 mddev->minor_version = sb->minor_version;
1494 if (mddev->in_sync)
1496 sb->recovery_cp = mddev->recovery_cp;
1497 sb->cp_events_hi = (mddev->events>>32);
1498 sb->cp_events_lo = (u32)mddev->events;
1499 if (mddev->recovery_cp == MaxSector)
1500 sb->state = (1<< MD_SB_CLEAN);
1502 sb->recovery_cp = 0;
1504 sb->layout = mddev->layout;
1505 sb->chunk_size = mddev->chunk_sectors << 9;
1507 if (mddev->bitmap && mddev->bitmap_info.file == NULL)
1508 sb->state |= (1<<MD_SB_BITMAP_PRESENT);
1510 sb->disks[0].state = (1<<MD_DISK_REMOVED);
1514 int is_active = test_bit(In_sync, &rdev2->flags);
1516 if (rdev2->raid_disk >= 0 &&
1517 sb->minor_version >= 91)
1520 * we can piggy-back on that.
1523 if (rdev2->raid_disk < 0 ||
1524 test_bit(Faulty, &rdev2->flags))
1527 desc_nr = rdev2->raid_disk;
1530 rdev2->desc_nr = desc_nr;
1531 d = &sb->disks[rdev2->desc_nr];
1533 d->number = rdev2->desc_nr;
1534 d->major = MAJOR(rdev2->bdev->bd_dev);
1535 d->minor = MINOR(rdev2->bdev->bd_dev);
1537 d->raid_disk = rdev2->raid_disk;
1539 d->raid_disk = rdev2->desc_nr; /* compatibility */
1540 if (test_bit(Faulty, &rdev2->flags))
1541 d->state = (1<<MD_DISK_FAULTY);
1543 d->state = (1<<MD_DISK_ACTIVE);
1544 if (test_bit(In_sync, &rdev2->flags))
1545 d->state |= (1<<MD_DISK_SYNC);
1549 d->state = 0;
1553 if (test_bit(WriteMostly, &rdev2->flags))
1554 d->state |= (1<<MD_DISK_WRITEMOSTLY);
1555 if (test_bit(FailFast, &rdev2->flags))
1556 d->state |= (1<<MD_DISK_FAILFAST);
1559 for (i=0 ; i < mddev->raid_disks ; i++) {
1560 mdp_disk_t *d = &sb->disks[i];
1561 if (d->state == 0 && d->number == 0) {
1562 d->number = i;
1563 d->raid_disk = i;
1564 d->state = (1<<MD_DISK_REMOVED);
1565 d->state |= (1<<MD_DISK_FAULTY);
1569 sb->nr_disks = nr_disks;
1570 sb->active_disks = active;
1571 sb->working_disks = working;
1572 sb->failed_disks = failed;
1573 sb->spare_disks = spare;
1575 sb->this_disk = sb->disks[rdev->desc_nr];
1576 sb->sb_csum = calc_sb_csum(sb);
1585 if (num_sectors && num_sectors < rdev->mddev->dev_sectors)
1587 if (rdev->mddev->bitmap_info.offset)
1589 rdev->sb_start = calc_dev_sboffset(rdev);
1590 if (!num_sectors || num_sectors > rdev->sb_start)
1591 num_sectors = rdev->sb_start;
1595 if ((u64)num_sectors >= (2ULL << 32) && rdev->mddev->level >= 1)
1596 num_sectors = (sector_t)(2ULL << 32) - 2;
1598 md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
1599 rdev->sb_page);
1600 } while (md_super_wait(rdev->mddev) < 0);
1607 /* non-zero offset changes not possible with v0.90 */
1620 int size = 256 + le32_to_cpu(sb->max_dev)*2;
1623 disk_csum = sb->sb_csum;
1624 sb->sb_csum = 0;
1626 for (; size >= 4; size -= 4)
1633 sb->sb_csum = disk_csum;
1657 sb_start = i_size_read(rdev->bdev->bd_inode) >> 9;
1658 sb_start -= 8*2;
1659 sb_start &= ~(sector_t)(4*2-1);
1668 return -EINVAL;
1670 rdev->sb_start = sb_start;
1678 sb = page_address(rdev->sb_page);
1680 if (sb->magic != cpu_to_le32(MD_SB_MAGIC) ||
1681 sb->major_version != cpu_to_le32(1) ||
1682 le32_to_cpu(sb->max_dev) > (4096-256)/2 ||
1683 le64_to_cpu(sb->super_offset) != rdev->sb_start ||
1684 (le32_to_cpu(sb->feature_map) & ~MD_FEATURE_ALL) != 0)
1685 return -EINVAL;
1687 if (calc_sb_1_csum(sb) != sb->sb_csum) {
1689 bdevname(rdev->bdev,b));
1690 return -EINVAL;
1692 if (le64_to_cpu(sb->data_size) < 10) {
1694 bdevname(rdev->bdev,b));
1695 return -EINVAL;
1697 if (sb->pad0 ||
1698 sb->pad3[0] ||
1699 memcmp(sb->pad3, sb->pad3+1, sizeof(sb->pad3) - sizeof(sb->pad3[1])))
1700 /* Some padding is non-zero, might be a new feature */
1701 return -EINVAL;
1703 rdev->preferred_minor = 0xffff;
1704 rdev->data_offset = le64_to_cpu(sb->data_offset);
1705 rdev->new_data_offset = rdev->data_offset;
1706 if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_RESHAPE_ACTIVE) &&
1707 (le32_to_cpu(sb->feature_map) & MD_FEATURE_NEW_OFFSET))
1708 rdev->new_data_offset += (s32)le32_to_cpu(sb->new_offset);
1709 atomic_set(&rdev->corrected_errors, le32_to_cpu(sb->cnt_corrected_read));
1711 rdev->sb_size = le32_to_cpu(sb->max_dev) * 2 + 256;
1712 bmask = queue_logical_block_size(rdev->bdev->bd_disk->queue)-1;
1713 if (rdev->sb_size & bmask)
1714 rdev->sb_size = (rdev->sb_size | bmask) + 1;
1717 && rdev->data_offset < sb_start + (rdev->sb_size/512))
1718 return -EINVAL;
1720 && rdev->new_data_offset < sb_start + (rdev->sb_size/512))
1721 return -EINVAL;
1723 if (sb->level == cpu_to_le32(LEVEL_MULTIPATH))
1724 rdev->desc_nr = -1;
1726 rdev->desc_nr = le32_to_cpu(sb->dev_number);
1728 if (!rdev->bb_page) {
1729 rdev->bb_page = alloc_page(GFP_KERNEL);
1730 if (!rdev->bb_page)
1731 return -ENOMEM;
1733 if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_BAD_BLOCKS) &&
1734 rdev->badblocks.count == 0) {
1742 int sectors = le16_to_cpu(sb->bblog_size);
1744 return -EINVAL;
1745 offset = le32_to_cpu(sb->bblog_offset);
1747 return -EINVAL;
1750 rdev->bb_page, REQ_OP_READ, 0, true))
1751 return -EIO;
1752 bbp = (__le64 *)page_address(rdev->bb_page);
1753 rdev->badblocks.shift = sb->bblog_shift;
1754 for (i = 0 ; i < (sectors << (9-3)) ; i++, bbp++) {
1758 sector <<= sb->bblog_shift;
1759 count <<= sb->bblog_shift;
1762 if (badblocks_set(&rdev->badblocks, sector, count, 1))
1763 return -EINVAL;
1765 } else if (sb->bblog_offset != 0)
1766 rdev->badblocks.shift = 0;
1768 if ((le32_to_cpu(sb->feature_map) &
1770 rdev->ppl.offset = (__s16)le16_to_cpu(sb->ppl.offset);
1771 rdev->ppl.size = le16_to_cpu(sb->ppl.size);
1772 rdev->ppl.sector = rdev->sb_start + rdev->ppl.offset;
1775 if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_RAID0_LAYOUT) &&
1776 sb->level != 0)
1777 return -EINVAL;
1780 if (sb->level == cpu_to_le32(LEVEL_MULTIPATH) ||
1781 (rdev->desc_nr >= 0 &&
1782 rdev->desc_nr < le32_to_cpu(sb->max_dev) &&
1783 (le16_to_cpu(sb->dev_roles[rdev->desc_nr]) < MD_DISK_ROLE_MAX ||
1784 le16_to_cpu(sb->dev_roles[rdev->desc_nr]) == MD_DISK_ROLE_JOURNAL)))
1794 struct mdp_superblock_1 *refsb = page_address(refdev->sb_page);
1796 if (memcmp(sb->set_uuid, refsb->set_uuid, 16) != 0 ||
1797 sb->level != refsb->level ||
1798 sb->layout != refsb->layout ||
1799 sb->chunksize != refsb->chunksize) {
1801 bdevname(rdev->bdev,b),
1802 bdevname(refdev->bdev,b2));
1803 return -EINVAL;
1805 ev1 = le64_to_cpu(sb->events);
1806 ev2 = le64_to_cpu(refsb->events);
1814 sectors = (i_size_read(rdev->bdev->bd_inode) >> 9);
1815 sectors -= rdev->data_offset;
1817 sectors = rdev->sb_start;
1818 if (sectors < le64_to_cpu(sb->data_size))
1819 return -EINVAL;
1820 rdev->sectors = le64_to_cpu(sb->data_size);
1826 struct mdp_superblock_1 *sb = page_address(rdev->sb_page);
1827 __u64 ev1 = le64_to_cpu(sb->events);
1829 rdev->raid_disk = -1;
1830 clear_bit(Faulty, &rdev->flags);
1831 clear_bit(In_sync, &rdev->flags);
1832 clear_bit(Bitmap_sync, &rdev->flags);
1833 clear_bit(WriteMostly, &rdev->flags);
1835 if (mddev->raid_disks == 0) {
1836 mddev->major_version = 1;
1837 mddev->patch_version = 0;
1838 mddev->external = 0;
1839 mddev->chunk_sectors = le32_to_cpu(sb->chunksize);
1840 mddev->ctime = le64_to_cpu(sb->ctime);
1841 mddev->utime = le64_to_cpu(sb->utime);
1842 mddev->level = le32_to_cpu(sb->level);
1843 mddev->clevel[0] = 0;
1844 mddev->layout = le32_to_cpu(sb->layout);
1845 mddev->raid_disks = le32_to_cpu(sb->raid_disks);
1846 mddev->dev_sectors = le64_to_cpu(sb->size);
1847 mddev->events = ev1;
1848 mddev->bitmap_info.offset = 0;
1849 mddev->bitmap_info.space = 0;
1851 * using 3K - total of 4K
1853 mddev->bitmap_info.default_offset = 1024 >> 9;
1854 mddev->bitmap_info.default_space = (4096-1024) >> 9;
1855 mddev->reshape_backwards = 0;
1857 mddev->recovery_cp = le64_to_cpu(sb->resync_offset);
1858 memcpy(mddev->uuid, sb->set_uuid, 16);
1860 mddev->max_disks = (4096-256)/2;
1862 if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_BITMAP_OFFSET) &&
1863 mddev->bitmap_info.file == NULL) {
1864 mddev->bitmap_info.offset =
1865 (__s32)le32_to_cpu(sb->bitmap_offset);
1871 if (mddev->minor_version > 0)
1872 mddev->bitmap_info.space = 0;
1873 else if (mddev->bitmap_info.offset > 0)
1874 mddev->bitmap_info.space =
1875 8 - mddev->bitmap_info.offset;
1877 mddev->bitmap_info.space =
1878 -mddev->bitmap_info.offset;
1881 if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_RESHAPE_ACTIVE)) {
1882 mddev->reshape_position = le64_to_cpu(sb->reshape_position);
1883 mddev->delta_disks = le32_to_cpu(sb->delta_disks);
1884 mddev->new_level = le32_to_cpu(sb->new_level);
1885 mddev->new_layout = le32_to_cpu(sb->new_layout);
1886 mddev->new_chunk_sectors = le32_to_cpu(sb->new_chunk);
1887 if (mddev->delta_disks < 0 ||
1888 (mddev->delta_disks == 0 &&
1889 (le32_to_cpu(sb->feature_map)
1891 mddev->reshape_backwards = 1;
1893 mddev->reshape_position = MaxSector;
1894 mddev->delta_disks = 0;
1895 mddev->new_level = mddev->level;
1896 mddev->new_layout = mddev->layout;
1897 mddev->new_chunk_sectors = mddev->chunk_sectors;
1900 if (mddev->level == 0 &&
1901 !(le32_to_cpu(sb->feature_map) & MD_FEATURE_RAID0_LAYOUT))
1902 mddev->layout = -1;
1904 if (le32_to_cpu(sb->feature_map) & MD_FEATURE_JOURNAL)
1905 set_bit(MD_HAS_JOURNAL, &mddev->flags);
1907 if (le32_to_cpu(sb->feature_map) &
1909 if (le32_to_cpu(sb->feature_map) &
1911 return -EINVAL;
1912 if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_PPL) &&
1913 (le32_to_cpu(sb->feature_map) &
1915 return -EINVAL;
1916 set_bit(MD_HAS_PPL, &mddev->flags);
1918 } else if (mddev->pers == NULL) {
1922 if (rdev->desc_nr >= 0 &&
1923 rdev->desc_nr < le32_to_cpu(sb->max_dev) &&
1924 (le16_to_cpu(sb->dev_roles[rdev->desc_nr]) < MD_DISK_ROLE_MAX ||
1925 le16_to_cpu(sb->dev_roles[rdev->desc_nr]) == MD_DISK_ROLE_JOURNAL))
1926 if (ev1 < mddev->events)
1927 return -EINVAL;
1928 } else if (mddev->bitmap) {
1932 if (ev1 < mddev->bitmap->events_cleared)
1934 if (ev1 < mddev->events)
1935 set_bit(Bitmap_sync, &rdev->flags);
1937 if (ev1 < mddev->events)
1938 /* just a hot-add of a new device, leave raid_disk at -1 */
1941 if (mddev->level != LEVEL_MULTIPATH) {
1943 if (rdev->desc_nr < 0 ||
1944 rdev->desc_nr >= le32_to_cpu(sb->max_dev)) {
1946 rdev->desc_nr = -1;
1948 role = le16_to_cpu(sb->dev_roles[rdev->desc_nr]);
1953 set_bit(Faulty, &rdev->flags);
1956 if (!(le32_to_cpu(sb->feature_map) & MD_FEATURE_JOURNAL)) {
1959 return -EINVAL;
1961 set_bit(Journal, &rdev->flags);
1962 rdev->journal_tail = le64_to_cpu(sb->journal_tail);
1963 rdev->raid_disk = 0;
1966 rdev->saved_raid_disk = role;
1967 if ((le32_to_cpu(sb->feature_map) &
1969 rdev->recovery_offset = le64_to_cpu(sb->recovery_offset);
1970 if (!(le32_to_cpu(sb->feature_map) &
1972 rdev->saved_raid_disk = -1;
1979 &mddev->recovery))
1980 set_bit(In_sync, &rdev->flags);
1982 rdev->raid_disk = role;
1985 if (sb->devflags & WriteMostly1)
1986 set_bit(WriteMostly, &rdev->flags);
1987 if (sb->devflags & FailFast1)
1988 set_bit(FailFast, &rdev->flags);
1989 if (le32_to_cpu(sb->feature_map) & MD_FEATURE_REPLACEMENT)
1990 set_bit(Replacement, &rdev->flags);
1992 set_bit(In_sync, &rdev->flags);
2002 /* make rdev->sb match mddev and rdev data. */
2004 sb = page_address(rdev->sb_page);
2006 sb->feature_map = 0;
2007 sb->pad0 = 0;
2008 sb->recovery_offset = cpu_to_le64(0);
2009 memset(sb->pad3, 0, sizeof(sb->pad3));
2011 sb->utime = cpu_to_le64((__u64)mddev->utime);
2012 sb->events = cpu_to_le64(mddev->events);
2013 if (mddev->in_sync)
2014 sb->resync_offset = cpu_to_le64(mddev->recovery_cp);
2015 else if (test_bit(MD_JOURNAL_CLEAN, &mddev->flags))
2016 sb->resync_offset = cpu_to_le64(MaxSector);
2018 sb->resync_offset = cpu_to_le64(0);
2020 sb->cnt_corrected_read = cpu_to_le32(atomic_read(&rdev->corrected_errors));
2022 sb->raid_disks = cpu_to_le32(mddev->raid_disks);
2023 sb->size = cpu_to_le64(mddev->dev_sectors);
2024 sb->chunksize = cpu_to_le32(mddev->chunk_sectors);
2025 sb->level = cpu_to_le32(mddev->level);
2026 sb->layout = cpu_to_le32(mddev->layout);
2027 if (test_bit(FailFast, &rdev->flags))
2028 sb->devflags |= FailFast1;
2030 sb->devflags &= ~FailFast1;
2032 if (test_bit(WriteMostly, &rdev->flags))
2033 sb->devflags |= WriteMostly1;
2035 sb->devflags &= ~WriteMostly1;
2036 sb->data_offset = cpu_to_le64(rdev->data_offset);
2037 sb->data_size = cpu_to_le64(rdev->sectors);
2039 if (mddev->bitmap && mddev->bitmap_info.file == NULL) {
2040 sb->bitmap_offset = cpu_to_le32((__u32)mddev->bitmap_info.offset);
2041 sb->feature_map = cpu_to_le32(MD_FEATURE_BITMAP_OFFSET);
2044 if (rdev->raid_disk >= 0 && !test_bit(Journal, &rdev->flags) &&
2045 !test_bit(In_sync, &rdev->flags)) {
2046 sb->feature_map |=
2048 sb->recovery_offset =
2049 cpu_to_le64(rdev->recovery_offset);
2050 if (rdev->saved_raid_disk >= 0 && mddev->bitmap)
2051 sb->feature_map |=
2055 if (test_bit(Journal, &rdev->flags))
2056 sb->journal_tail = cpu_to_le64(rdev->journal_tail);
2057 if (test_bit(Replacement, &rdev->flags))
2058 sb->feature_map |=
2061 if (mddev->reshape_position != MaxSector) {
2062 sb->feature_map |= cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE);
2063 sb->reshape_position = cpu_to_le64(mddev->reshape_position);
2064 sb->new_layout = cpu_to_le32(mddev->new_layout);
2065 sb->delta_disks = cpu_to_le32(mddev->delta_disks);
2066 sb->new_level = cpu_to_le32(mddev->new_level);
2067 sb->new_chunk = cpu_to_le32(mddev->new_chunk_sectors);
2068 if (mddev->delta_disks == 0 &&
2069 mddev->reshape_backwards)
2070 sb->feature_map
2072 if (rdev->new_data_offset != rdev->data_offset) {
2073 sb->feature_map
2075 sb->new_offset = cpu_to_le32((__u32)(rdev->new_data_offset
2076 - rdev->data_offset));
2081 sb->feature_map |= cpu_to_le32(MD_FEATURE_CLUSTERED);
2083 if (rdev->badblocks.count == 0)
2085 else if (sb->bblog_offset == 0)
2089 struct badblocks *bb = &rdev->badblocks;
2090 __le64 *bbp = (__le64 *)page_address(rdev->bb_page);
2091 u64 *p = bb->page;
2092 sb->feature_map |= cpu_to_le32(MD_FEATURE_BAD_BLOCKS);
2093 if (bb->changed) {
2097 seq = read_seqbegin(&bb->lock);
2101 for (i = 0 ; i < bb->count ; i++) {
2107 bb->changed = 0;
2108 if (read_seqretry(&bb->lock, seq))
2111 bb->sector = (rdev->sb_start +
2112 (int)le32_to_cpu(sb->bblog_offset));
2113 bb->size = le16_to_cpu(sb->bblog_size);
2119 if (rdev2->desc_nr+1 > max_dev)
2120 max_dev = rdev2->desc_nr+1;
2122 if (max_dev > le32_to_cpu(sb->max_dev)) {
2124 sb->max_dev = cpu_to_le32(max_dev);
2125 rdev->sb_size = max_dev * 2 + 256;
2126 bmask = queue_logical_block_size(rdev->bdev->bd_disk->queue)-1;
2127 if (rdev->sb_size & bmask)
2128 rdev->sb_size = (rdev->sb_size | bmask) + 1;
2130 max_dev = le32_to_cpu(sb->max_dev);
2133 sb->dev_roles[i] = cpu_to_le16(MD_DISK_ROLE_SPARE);
2135 if (test_bit(MD_HAS_JOURNAL, &mddev->flags))
2136 sb->feature_map |= cpu_to_le32(MD_FEATURE_JOURNAL);
2138 if (test_bit(MD_HAS_PPL, &mddev->flags)) {
2139 if (test_bit(MD_HAS_MULTIPLE_PPLS, &mddev->flags))
2140 sb->feature_map |=
2143 sb->feature_map |= cpu_to_le32(MD_FEATURE_PPL);
2144 sb->ppl.offset = cpu_to_le16(rdev->ppl.offset);
2145 sb->ppl.size = cpu_to_le16(rdev->ppl.size);
2149 i = rdev2->desc_nr;
2150 if (test_bit(Faulty, &rdev2->flags))
2151 sb->dev_roles[i] = cpu_to_le16(MD_DISK_ROLE_FAULTY);
2152 else if (test_bit(In_sync, &rdev2->flags))
2153 sb->dev_roles[i] = cpu_to_le16(rdev2->raid_disk);
2154 else if (test_bit(Journal, &rdev2->flags))
2155 sb->dev_roles[i] = cpu_to_le16(MD_DISK_ROLE_JOURNAL);
2156 else if (rdev2->raid_disk >= 0)
2157 sb->dev_roles[i] = cpu_to_le16(rdev2->raid_disk);
2159 sb->dev_roles[i] = cpu_to_le16(MD_DISK_ROLE_SPARE);
2162 sb->sb_csum = calc_sb_1_csum(sb);
2174 else if (dev_size - 64*2 >= 200*1024*1024*2)
2176 else if (dev_size - 4*2 > 8*1024*1024*2)
2188 if (num_sectors && num_sectors < rdev->mddev->dev_sectors)
2190 if (rdev->data_offset != rdev->new_data_offset)
2192 if (rdev->sb_start < rdev->data_offset) {
2194 max_sectors = i_size_read(rdev->bdev->bd_inode) >> 9;
2195 max_sectors -= rdev->data_offset;
2198 } else if (rdev->mddev->bitmap_info.offset) {
2204 sector_t dev_size = i_size_read(rdev->bdev->bd_inode) >> 9;
2207 sb_start = dev_size - 8*2;
2208 sb_start &= ~(sector_t)(4*2 - 1);
2215 max_sectors = sb_start - bm_space - 4*2;
2219 rdev->sb_start = sb_start;
2221 sb = page_address(rdev->sb_page);
2222 sb->data_size = cpu_to_le64(num_sectors);
2223 sb->super_offset = cpu_to_le64(rdev->sb_start);
2224 sb->sb_csum = calc_sb_1_csum(sb);
2226 md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
2227 rdev->sb_page);
2228 } while (md_super_wait(rdev->mddev) < 0);
2239 if (new_offset >= rdev->data_offset)
2244 if (rdev->mddev->minor_version == 0)
2251 * beyond write-intent bitmap
2253 if (rdev->sb_start + (32+4)*2 > new_offset)
2255 bitmap = rdev->mddev->bitmap;
2256 if (bitmap && !rdev->mddev->bitmap_info.file &&
2257 rdev->sb_start + rdev->mddev->bitmap_info.offset +
2258 bitmap->storage.file_pages * (PAGE_SIZE>>9) > new_offset)
2260 if (rdev->badblocks.sector + rdev->badblocks.size > new_offset)
2277 .name = "md-1",
2289 if (mddev->sync_super) {
2290 mddev->sync_super(mddev, rdev);
2294 BUG_ON(mddev->major_version >= ARRAY_SIZE(super_types));
2296 super_types[mddev->major_version].sync_super(mddev, rdev);
2305 if (test_bit(Faulty, &rdev->flags) ||
2306 test_bit(Journal, &rdev->flags) ||
2307 rdev->raid_disk == -1)
2310 if (test_bit(Faulty, &rdev2->flags) ||
2311 test_bit(Journal, &rdev2->flags) ||
2312 rdev2->raid_disk == -1)
2314 if (rdev->bdev->bd_disk == rdev2->bdev->bd_disk) {
2337 if (list_empty(&mddev->disks))
2339 if (!mddev->gendisk || blk_get_integrity(mddev->gendisk))
2342 /* skip spares and non-functional disks */
2343 if (test_bit(Faulty, &rdev->flags))
2345 if (rdev->raid_disk < 0)
2353 if (blk_integrity_compare(reference->bdev->bd_disk,
2354 rdev->bdev->bd_disk) < 0)
2355 return -EINVAL;
2357 if (!reference || !bdev_get_integrity(reference->bdev))
2363 blk_integrity_register(mddev->gendisk,
2364 bdev_get_integrity(reference->bdev));
2367 if (bioset_integrity_create(&mddev->bio_set, BIO_POOL_SIZE)) {
2370 return -EINVAL;
2385 if (!mddev->gendisk)
2388 bi_mddev = blk_get_integrity(mddev->gendisk);
2393 if (blk_integrity_compare(mddev->gendisk, rdev->bdev->bd_disk) != 0) {
2395 mdname(mddev), bdevname(rdev->bdev, name));
2396 return -ENXIO;
2410 if (find_rdev(mddev, rdev->bdev->bd_dev))
2411 return -EEXIST;
2413 if ((bdev_read_only(rdev->bdev) || bdev_read_only(rdev->meta_bdev)) &&
2414 mddev->pers)
2415 return -EROFS;
2417 /* make sure rdev->sectors exceeds mddev->dev_sectors */
2418 if (!test_bit(Journal, &rdev->flags) &&
2419 rdev->sectors &&
2420 (mddev->dev_sectors == 0 || rdev->sectors < mddev->dev_sectors)) {
2421 if (mddev->pers) {
2423 * If mddev->level <= 0, then we don't care
2426 if (mddev->level > 0)
2427 return -ENOSPC;
2429 mddev->dev_sectors = rdev->sectors;
2432 /* Verify rdev->desc_nr is unique.
2433 * If it is -1, assign a free number, else
2437 if (rdev->desc_nr < 0) {
2439 if (mddev->pers)
2440 choice = mddev->raid_disks;
2443 rdev->desc_nr = choice;
2445 if (md_find_rdev_nr_rcu(mddev, rdev->desc_nr)) {
2447 return -EBUSY;
2451 if (!test_bit(Journal, &rdev->flags) &&
2452 mddev->max_disks && rdev->desc_nr >= mddev->max_disks) {
2454 mdname(mddev), mddev->max_disks);
2455 return -EBUSY;
2457 bdevname(rdev->bdev,b);
2460 rdev->mddev = mddev;
2463 if (mddev->raid_disks)
2466 if ((err = kobject_add(&rdev->kobj, &mddev->kobj, "dev-%s", b)))
2469 ko = &part_to_dev(rdev->bdev->bd_part)->kobj;
2471 err = sysfs_create_link(&rdev->kobj, ko, "block");
2472 rdev->sysfs_state = sysfs_get_dirent_safe(rdev->kobj.sd, "state");
2473 rdev->sysfs_unack_badblocks =
2474 sysfs_get_dirent_safe(rdev->kobj.sd, "unacknowledged_bad_blocks");
2475 rdev->sysfs_badblocks =
2476 sysfs_get_dirent_safe(rdev->kobj.sd, "bad_blocks");
2478 list_add_rcu(&rdev->same_set, &mddev->disks);
2479 bd_link_disk_holder(rdev->bdev, mddev->gendisk);
2482 mddev->recovery_disabled++;
2487 pr_warn("md: failed to register dev-%s for %s\n",
2496 kobject_del(&rdev->kobj);
2497 kobject_put(&rdev->kobj);
2504 bd_unlink_disk_holder(rdev->bdev, rdev->mddev->gendisk);
2505 list_del_rcu(&rdev->same_set);
2506 pr_debug("md: unbind<%s>\n", bdevname(rdev->bdev,b));
2507 mddev_destroy_serial_pool(rdev->mddev, rdev, false);
2508 rdev->mddev = NULL;
2509 sysfs_remove_link(&rdev->kobj, "block");
2510 sysfs_put(rdev->sysfs_state);
2511 sysfs_put(rdev->sysfs_unack_badblocks);
2512 sysfs_put(rdev->sysfs_badblocks);
2513 rdev->sysfs_state = NULL;
2514 rdev->sysfs_unack_badblocks = NULL;
2515 rdev->sysfs_badblocks = NULL;
2516 rdev->badblocks.count = 0;
2522 INIT_WORK(&rdev->del_work, rdev_delayed_delete);
2523 kobject_get(&rdev->kobj);
2524 queue_work(md_rdev_misc_wq, &rdev->del_work);
2540 pr_warn("md: could not open device unknown-block(%u,%u).\n",
2544 rdev->bdev = bdev;
2550 struct block_device *bdev = rdev->bdev;
2551 rdev->bdev = NULL;
2561 pr_debug("md: export_rdev(%s)\n", bdevname(rdev->bdev,b));
2564 if (test_bit(AutoDetected, &rdev->flags))
2565 md_autodetect_dev(rdev->bdev->bd_dev);
2568 kobject_put(&rdev->kobj);
2582 while (!list_empty(&mddev->disks)) {
2583 rdev = list_first_entry(&mddev->disks, struct md_rdev,
2587 mddev->raid_disks = 0;
2588 mddev->major_version = 0;
2593 lockdep_assert_held(&mddev->lock);
2594 if (!mddev->in_sync) {
2595 mddev->sync_checkers++;
2596 spin_unlock(&mddev->lock);
2597 percpu_ref_switch_to_atomic_sync(&mddev->writes_pending);
2598 spin_lock(&mddev->lock);
2599 if (!mddev->in_sync &&
2600 percpu_ref_is_zero(&mddev->writes_pending)) {
2601 mddev->in_sync = 1;
2603 * Ensure ->in_sync is visible before we clear
2604 * ->sync_checkers.
2607 set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
2608 sysfs_notify_dirent_safe(mddev->sysfs_state);
2610 if (--mddev->sync_checkers == 0)
2611 percpu_ref_switch_to_percpu(&mddev->writes_pending);
2613 if (mddev->safemode == 1)
2614 mddev->safemode = 0;
2615 return mddev->in_sync;
2620 /* Update each superblock (in-memory image), but
2628 if (rdev->sb_events == mddev->events ||
2630 rdev->raid_disk < 0 &&
2631 rdev->sb_events+1 == mddev->events)) {
2633 rdev->sb_loaded = 2;
2636 rdev->sb_loaded = 1;
2649 if ((iter->raid_disk >= 0) && !test_bit(Faulty, &iter->flags)) {
2658 sb = page_address(rdev->sb_page);
2661 role = le16_to_cpu(sb->dev_roles[rdev->desc_nr]);
2663 if (role == 0xffff && rdev->raid_disk >=0 &&
2664 !test_bit(Faulty, &rdev->flags))
2667 if (test_bit(Faulty, &rdev->flags) && (role < 0xfffd))
2672 if ((mddev->dev_sectors != le64_to_cpu(sb->size)) ||
2673 (mddev->reshape_position != le64_to_cpu(sb->reshape_position)) ||
2674 (mddev->layout != le32_to_cpu(sb->layout)) ||
2675 (mddev->raid_disks != le32_to_cpu(sb->raid_disks)) ||
2676 (mddev->chunk_sectors != le32_to_cpu(sb->chunksize)))
2688 int ret = -1;
2690 if (mddev->ro) {
2692 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
2698 if (test_and_clear_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags))
2700 if (test_and_clear_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags))
2702 ret = md_cluster_ops->metadata_update_start(mddev);
2706 md_cluster_ops->metadata_update_cancel(mddev);
2707 bit_clear_unless(&mddev->sb_flags, BIT(MD_SB_CHANGE_PENDING),
2717 * During reshape/resync it might use array-addresses rather
2721 if (rdev->raid_disk >= 0 &&
2722 mddev->delta_disks >= 0 &&
2723 test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) &&
2724 test_bit(MD_RECOVERY_RECOVER, &mddev->recovery) &&
2725 !test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
2726 !test_bit(Journal, &rdev->flags) &&
2727 !test_bit(In_sync, &rdev->flags) &&
2728 mddev->curr_resync_completed > rdev->recovery_offset)
2729 rdev->recovery_offset = mddev->curr_resync_completed;
2732 if (!mddev->persistent) {
2733 clear_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
2734 clear_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
2735 if (!mddev->external) {
2736 clear_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
2738 if (rdev->badblocks.changed) {
2739 rdev->badblocks.changed = 0;
2740 ack_all_badblocks(&rdev->badblocks);
2743 clear_bit(Blocked, &rdev->flags);
2744 clear_bit(BlockedBadBlocks, &rdev->flags);
2745 wake_up(&rdev->blocked_wait);
2748 wake_up(&mddev->sb_wait);
2752 spin_lock(&mddev->lock);
2754 mddev->utime = ktime_get_real_seconds();
2756 if (test_and_clear_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags))
2758 if (test_and_clear_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags))
2759 /* just a clean<-> dirty transition, possibly leave spares alone,
2766 if (mddev->degraded)
2770 * might have a event_count that still looks up-to-date,
2771 * so it can be re-added without a resync.
2778 sync_req = mddev->in_sync;
2780 /* If this is just a dirty<->clean transition, and the array is clean
2783 && (mddev->in_sync && mddev->recovery_cp == MaxSector)
2784 && mddev->can_decrease_events
2785 && mddev->events != 1) {
2786 mddev->events--;
2787 mddev->can_decrease_events = 0;
2790 mddev->events ++;
2791 mddev->can_decrease_events = nospares;
2795 * This 64-bit counter should never wrap.
2799 WARN_ON(mddev->events == 0);
2802 if (rdev->badblocks.changed)
2804 if (test_bit(Faulty, &rdev->flags))
2805 set_bit(FaultRecorded, &rdev->flags);
2809 spin_unlock(&mddev->lock);
2812 mdname(mddev), mddev->in_sync);
2814 if (mddev->queue)
2815 blk_add_trace_msg(mddev->queue, "md md_update_sb");
2817 md_bitmap_update_sb(mddev->bitmap);
2821 if (rdev->sb_loaded != 1)
2824 if (!test_bit(Faulty, &rdev->flags)) {
2826 rdev->sb_start, rdev->sb_size,
2827 rdev->sb_page);
2829 bdevname(rdev->bdev, b),
2830 (unsigned long long)rdev->sb_start);
2831 rdev->sb_events = mddev->events;
2832 if (rdev->badblocks.size) {
2834 rdev->badblocks.sector,
2835 rdev->badblocks.size << 9,
2836 rdev->bb_page);
2837 rdev->badblocks.size = 0;
2842 bdevname(rdev->bdev, b));
2844 if (mddev->level == LEVEL_MULTIPATH)
2850 /* if there was a failure, MD_SB_CHANGE_DEVS was set, and we re-write super */
2853 md_cluster_ops->metadata_update_finish(mddev);
2855 if (mddev->in_sync != sync_req ||
2856 !bit_clear_unless(&mddev->sb_flags, BIT(MD_SB_CHANGE_PENDING),
2860 wake_up(&mddev->sb_wait);
2861 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
2862 sysfs_notify_dirent_safe(mddev->sysfs_completed);
2865 if (test_and_clear_bit(FaultRecorded, &rdev->flags))
2866 clear_bit(Blocked, &rdev->flags);
2869 ack_all_badblocks(&rdev->badblocks);
2870 clear_bit(BlockedBadBlocks, &rdev->flags);
2871 wake_up(&rdev->blocked_wait);
2878 struct mddev *mddev = rdev->mddev;
2880 bool add_journal = test_bit(Journal, &rdev->flags);
2882 if (!mddev->pers->hot_remove_disk || add_journal) {
2887 super_types[mddev->major_version].
2891 err = mddev->pers->hot_add_disk(mddev, rdev);
2899 sysfs_notify_dirent_safe(rdev->sysfs_state);
2901 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
2902 if (mddev->degraded)
2903 set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
2904 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
2906 md_wakeup_thread(mddev->thread);
2941 unsigned long flags = READ_ONCE(rdev->flags);
2945 rdev->badblocks.unacked_exist))
2954 (rdev->badblocks.unacked_exist
2973 len -= strlen(sep);
2982 * faulty - simulates an error
2983 * remove - disconnects the device
2984 * writemostly - sets write_mostly
2985 * -writemostly - clears write_mostly
2986 * blocked - sets the Blocked flags
2987 * -blocked - clears the Blocked and possibly simulates an error
2988 * insync - sets Insync providing device isn't active
2989 * -insync - clear Insync for a device with a slot assigned,
2991 * write_error - sets WriteErrorSeen
2992 * -write_error - clears WriteErrorSeen
2993 * {,-}failfast - set/clear FailFast
2996 struct mddev *mddev = rdev->mddev;
2997 int err = -EINVAL;
3000 if (cmd_match(buf, "faulty") && rdev->mddev->pers) {
3001 md_error(rdev->mddev, rdev);
3002 if (test_bit(Faulty, &rdev->flags))
3005 err = -EBUSY;
3007 if (rdev->mddev->pers) {
3008 clear_bit(Blocked, &rdev->flags);
3009 remove_and_add_spares(rdev->mddev, rdev);
3011 if (rdev->raid_disk >= 0)
3012 err = -EBUSY;
3016 err = md_cluster_ops->remove_disk(mddev, rdev);
3020 if (mddev->pers) {
3021 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
3022 md_wakeup_thread(mddev->thread);
3028 set_bit(WriteMostly, &rdev->flags);
3029 mddev_create_serial_pool(rdev->mddev, rdev, false);
3032 } else if (cmd_match(buf, "-writemostly")) {
3033 mddev_destroy_serial_pool(rdev->mddev, rdev, false);
3034 clear_bit(WriteMostly, &rdev->flags);
3038 set_bit(Blocked, &rdev->flags);
3040 } else if (cmd_match(buf, "-blocked")) {
3041 if (!test_bit(Faulty, &rdev->flags) &&
3042 !test_bit(ExternalBbl, &rdev->flags) &&
3043 rdev->badblocks.unacked_exist) {
3047 md_error(rdev->mddev, rdev);
3049 clear_bit(Blocked, &rdev->flags);
3050 clear_bit(BlockedBadBlocks, &rdev->flags);
3051 wake_up(&rdev->blocked_wait);
3052 set_bit(MD_RECOVERY_NEEDED, &rdev->mddev->recovery);
3053 md_wakeup_thread(rdev->mddev->thread);
3056 } else if (cmd_match(buf, "insync") && rdev->raid_disk == -1) {
3057 set_bit(In_sync, &rdev->flags);
3060 set_bit(FailFast, &rdev->flags);
3063 } else if (cmd_match(buf, "-failfast")) {
3064 clear_bit(FailFast, &rdev->flags);
3067 } else if (cmd_match(buf, "-insync") && rdev->raid_disk >= 0 &&
3068 !test_bit(Journal, &rdev->flags)) {
3069 if (rdev->mddev->pers == NULL) {
3070 clear_bit(In_sync, &rdev->flags);
3071 rdev->saved_raid_disk = rdev->raid_disk;
3072 rdev->raid_disk = -1;
3076 set_bit(WriteErrorSeen, &rdev->flags);
3078 } else if (cmd_match(buf, "-write_error")) {
3079 clear_bit(WriteErrorSeen, &rdev->flags);
3082 /* Any non-spare device that is not a replacement can
3086 if (rdev->raid_disk >= 0 &&
3087 !test_bit(Journal, &rdev->flags) &&
3088 !test_bit(Replacement, &rdev->flags))
3089 set_bit(WantReplacement, &rdev->flags);
3090 set_bit(MD_RECOVERY_NEEDED, &rdev->mddev->recovery);
3091 md_wakeup_thread(rdev->mddev->thread);
3093 } else if (cmd_match(buf, "-want_replacement")) {
3098 clear_bit(WantReplacement, &rdev->flags);
3104 if (rdev->mddev->pers)
3105 err = -EBUSY;
3107 set_bit(Replacement, &rdev->flags);
3110 } else if (cmd_match(buf, "-replacement")) {
3112 if (rdev->mddev->pers)
3113 err = -EBUSY;
3115 clear_bit(Replacement, &rdev->flags);
3118 } else if (cmd_match(buf, "re-add")) {
3119 if (!rdev->mddev->pers)
3120 err = -EINVAL;
3121 else if (test_bit(Faulty, &rdev->flags) && (rdev->raid_disk == -1) &&
3122 rdev->saved_raid_disk >= 0) {
3129 if (!mddev_is_clustered(rdev->mddev) ||
3130 (err = md_cluster_ops->gather_bitmaps(rdev)) == 0) {
3131 clear_bit(Faulty, &rdev->flags);
3135 err = -EBUSY;
3136 } else if (cmd_match(buf, "external_bbl") && (rdev->mddev->external)) {
3137 set_bit(ExternalBbl, &rdev->flags);
3138 rdev->badblocks.shift = 0;
3140 } else if (cmd_match(buf, "-external_bbl") && (rdev->mddev->external)) {
3141 clear_bit(ExternalBbl, &rdev->flags);
3147 sysfs_notify_dirent_safe(rdev->sysfs_state);
3156 return sprintf(page, "%d\n", atomic_read(&rdev->corrected_errors));
3168 atomic_set(&rdev->corrected_errors, n);
3177 if (test_bit(Journal, &rdev->flags))
3179 else if (rdev->raid_disk < 0)
3182 return sprintf(page, "%d\n", rdev->raid_disk);
3191 if (test_bit(Journal, &rdev->flags))
3192 return -EBUSY;
3194 slot = -1;
3201 return -ENOSPC;
3203 if (rdev->mddev->pers && slot == -1) {
3206 * with the personality with ->hot_*_disk.
3211 if (rdev->raid_disk == -1)
3212 return -EEXIST;
3214 if (rdev->mddev->pers->hot_remove_disk == NULL)
3215 return -EINVAL;
3216 clear_bit(Blocked, &rdev->flags);
3217 remove_and_add_spares(rdev->mddev, rdev);
3218 if (rdev->raid_disk >= 0)
3219 return -EBUSY;
3220 set_bit(MD_RECOVERY_NEEDED, &rdev->mddev->recovery);
3221 md_wakeup_thread(rdev->mddev->thread);
3222 } else if (rdev->mddev->pers) {
3228 if (rdev->raid_disk != -1)
3229 return -EBUSY;
3231 if (test_bit(MD_RECOVERY_RUNNING, &rdev->mddev->recovery))
3232 return -EBUSY;
3234 if (rdev->mddev->pers->hot_add_disk == NULL)
3235 return -EINVAL;
3237 if (slot >= rdev->mddev->raid_disks &&
3238 slot >= rdev->mddev->raid_disks + rdev->mddev->delta_disks)
3239 return -ENOSPC;
3241 rdev->raid_disk = slot;
3242 if (test_bit(In_sync, &rdev->flags))
3243 rdev->saved_raid_disk = slot;
3245 rdev->saved_raid_disk = -1;
3246 clear_bit(In_sync, &rdev->flags);
3247 clear_bit(Bitmap_sync, &rdev->flags);
3248 err = rdev->mddev->pers->hot_add_disk(rdev->mddev, rdev);
3250 rdev->raid_disk = -1;
3253 sysfs_notify_dirent_safe(rdev->sysfs_state);
3255 sysfs_link_rdev(rdev->mddev, rdev);
3258 if (slot >= rdev->mddev->raid_disks &&
3259 slot >= rdev->mddev->raid_disks + rdev->mddev->delta_disks)
3260 return -ENOSPC;
3261 rdev->raid_disk = slot;
3263 clear_bit(Faulty, &rdev->flags);
3264 clear_bit(WriteMostly, &rdev->flags);
3265 set_bit(In_sync, &rdev->flags);
3266 sysfs_notify_dirent_safe(rdev->sysfs_state);
3277 return sprintf(page, "%llu\n", (unsigned long long)rdev->data_offset);
3285 return -EINVAL;
3286 if (rdev->mddev->pers && rdev->raid_disk >= 0)
3287 return -EBUSY;
3288 if (rdev->sectors && rdev->mddev->external)
3291 return -EBUSY;
3292 rdev->data_offset = offset;
3293 rdev->new_data_offset = offset;
3303 (unsigned long long)rdev->new_data_offset);
3310 struct mddev *mddev = rdev->mddev;
3313 return -EINVAL;
3315 if (mddev->sync_thread ||
3316 test_bit(MD_RECOVERY_RUNNING,&mddev->recovery))
3317 return -EBUSY;
3318 if (new_offset == rdev->data_offset)
3321 else if (new_offset > rdev->data_offset) {
3323 if (new_offset - rdev->data_offset
3324 + mddev->dev_sectors > rdev->sectors)
3325 return -E2BIG;
3332 if (new_offset < rdev->data_offset &&
3333 mddev->reshape_backwards)
3334 return -EINVAL;
3339 if (new_offset > rdev->data_offset &&
3340 !mddev->reshape_backwards)
3341 return -EINVAL;
3343 if (mddev->pers && mddev->persistent &&
3344 !super_types[mddev->major_version]
3346 return -E2BIG;
3347 rdev->new_data_offset = new_offset;
3348 if (new_offset > rdev->data_offset)
3349 mddev->reshape_backwards = 1;
3350 else if (new_offset < rdev->data_offset)
3351 mddev->reshape_backwards = 0;
3361 return sprintf(page, "%llu\n", (unsigned long long)rdev->sectors / 2);
3380 return -EINVAL;
3382 if (blocks & 1ULL << (8 * sizeof(blocks) - 1))
3383 return -EINVAL; /* sector conversion overflow */
3387 return -EINVAL; /* unsigned long long to sector_t overflow */
3396 struct mddev *my_mddev = rdev->mddev;
3397 sector_t oldsectors = rdev->sectors;
3400 if (test_bit(Journal, &rdev->flags))
3401 return -EBUSY;
3403 return -EINVAL;
3404 if (rdev->data_offset != rdev->new_data_offset)
3405 return -EINVAL; /* too confusing */
3406 if (my_mddev->pers && rdev->raid_disk >= 0) {
3407 if (my_mddev->persistent) {
3408 sectors = super_types[my_mddev->major_version].
3411 return -EBUSY;
3413 sectors = (i_size_read(rdev->bdev->bd_inode) >> 9) -
3414 rdev->data_offset;
3415 if (!my_mddev->pers->resize)
3417 return -EINVAL;
3419 if (sectors < my_mddev->dev_sectors)
3420 return -EINVAL; /* component must fit device */
3422 rdev->sectors = sectors;
3423 if (sectors > oldsectors && my_mddev->external) {
3425 * ->bdev do not overlap. 'rcu' is sufficient to walk
3439 if (rdev->bdev == rdev2->bdev &&
3441 overlaps(rdev->data_offset, rdev->sectors,
3442 rdev2->data_offset,
3443 rdev2->sectors)) {
3460 rdev->sectors = oldsectors;
3461 return -EBUSY;
3472 unsigned long long recovery_start = rdev->recovery_offset;
3474 if (test_bit(In_sync, &rdev->flags) ||
3488 return -EINVAL;
3490 if (rdev->mddev->pers &&
3491 rdev->raid_disk >= 0)
3492 return -EBUSY;
3494 rdev->recovery_offset = recovery_start;
3496 set_bit(In_sync, &rdev->flags);
3498 clear_bit(In_sync, &rdev->flags);
3505 /* sysfs access to bad-blocks list.
3507 * 'bad-blocks' lists sector numbers and lengths of ranges that
3509 * the one-page limit of sysfs.
3512 * 'unacknowledged-bad-blocks' lists bad blocks that have not yet
3518 return badblocks_show(&rdev->badblocks, page, 0);
3522 int rv = badblocks_store(&rdev->badblocks, page, len, 0);
3524 if (test_and_clear_bit(BlockedBadBlocks, &rdev->flags))
3525 wake_up(&rdev->blocked_wait);
3533 return badblocks_show(&rdev->badblocks, page, 1);
3537 return badblocks_store(&rdev->badblocks, page, len, 1);
3545 return sprintf(page, "%llu\n", (unsigned long long)rdev->ppl.sector);
3554 return -EINVAL;
3556 return -EINVAL;
3558 if (rdev->mddev->pers && test_bit(MD_HAS_PPL, &rdev->mddev->flags) &&
3559 rdev->raid_disk >= 0)
3560 return -EBUSY;
3562 if (rdev->mddev->persistent) {
3563 if (rdev->mddev->major_version == 0)
3564 return -EINVAL;
3565 if ((sector > rdev->sb_start &&
3566 sector - rdev->sb_start > S16_MAX) ||
3567 (sector < rdev->sb_start &&
3568 rdev->sb_start - sector > -S16_MIN))
3569 return -EINVAL;
3570 rdev->ppl.offset = sector - rdev->sb_start;
3571 } else if (!rdev->mddev->external) {
3572 return -EBUSY;
3574 rdev->ppl.sector = sector;
3584 return sprintf(page, "%u\n", rdev->ppl.size);
3593 return -EINVAL;
3595 if (rdev->mddev->pers && test_bit(MD_HAS_PPL, &rdev->mddev->flags) &&
3596 rdev->raid_disk >= 0)
3597 return -EBUSY;
3599 if (rdev->mddev->persistent) {
3600 if (rdev->mddev->major_version == 0)
3601 return -EINVAL;
3603 return -EINVAL;
3604 } else if (!rdev->mddev->external) {
3605 return -EBUSY;
3607 rdev->ppl.size = size;
3634 if (!entry->show)
3635 return -EIO;
3636 if (!rdev->mddev)
3637 return -ENODEV;
3638 return entry->show(rdev, page);
3648 struct mddev *mddev = rdev->mddev;
3650 if (!entry->store)
3651 return -EIO;
3653 return -EACCES;
3654 rv = mddev ? mddev_lock(mddev) : -ENODEV;
3656 if (rdev->mddev == NULL)
3657 rv = -ENODEV;
3659 rv = entry->store(rdev, page, length);
3682 rdev->desc_nr = -1;
3683 rdev->saved_raid_disk = -1;
3684 rdev->raid_disk = -1;
3685 rdev->flags = 0;
3686 rdev->data_offset = 0;
3687 rdev->new_data_offset = 0;
3688 rdev->sb_events = 0;
3689 rdev->last_read_error = 0;
3690 rdev->sb_loaded = 0;
3691 rdev->bb_page = NULL;
3692 atomic_set(&rdev->nr_pending, 0);
3693 atomic_set(&rdev->read_errors, 0);
3694 atomic_set(&rdev->corrected_errors, 0);
3696 INIT_LIST_HEAD(&rdev->same_set);
3697 init_waitqueue_head(&rdev->blocked_wait);
3701 * be used - I wonder if that matters
3703 return badblocks_init(&rdev->badblocks, 0);
3711 * - the device is nonexistent (zero size)
3712 * - the device has no valid superblock
3714 * a faulty rdev _never_ has rdev->sb set.
3725 return ERR_PTR(-ENOMEM);
3734 err = lock_rdev(rdev, newdev, super_format == -2);
3738 kobject_init(&rdev->kobj, &rdev_ktype);
3740 size = i_size_read(rdev->bdev->bd_inode) >> BLOCK_SIZE_BITS;
3743 bdevname(rdev->bdev,b));
3744 err = -EINVAL;
3751 if (err == -EINVAL) {
3753 bdevname(rdev->bdev,b),
3759 bdevname(rdev->bdev,b));
3767 if (rdev->bdev)
3786 switch (super_types[mddev->major_version].
3787 load_super(rdev, freshest, mddev->minor_version)) {
3794 pr_warn("md: fatal superblock inconsistency in %s -- removing from array\n",
3795 bdevname(rdev->bdev,b));
3802 return -EINVAL;
3805 super_types[mddev->major_version].
3810 if (mddev->max_disks &&
3811 (rdev->desc_nr >= mddev->max_disks ||
3812 i > mddev->max_disks)) {
3814 mdname(mddev), bdevname(rdev->bdev, b),
3815 mddev->max_disks);
3820 if (super_types[mddev->major_version].
3822 pr_warn("md: kicking non-fresh %s from array!\n",
3823 bdevname(rdev->bdev,b));
3828 if (mddev->level == LEVEL_MULTIPATH) {
3829 rdev->desc_nr = i++;
3830 rdev->raid_disk = rdev->desc_nr;
3831 set_bit(In_sync, &rdev->flags);
3832 } else if (rdev->raid_disk >=
3833 (mddev->raid_disks - min(0, mddev->delta_disks)) &&
3834 !test_bit(Journal, &rdev->flags)) {
3835 rdev->raid_disk = -1;
3836 clear_bit(In_sync, &rdev->flags);
3843 /* Read a fixed-point number.
3851 * all without any floating-point arithmetic.
3856 long decimals = -1;
3862 value = *cp - '0';
3872 return -EINVAL;
3875 *res = result * int_pow(10, scale - decimals);
3882 unsigned int msec = ((unsigned long)mddev->safemode_delay*1000)/HZ;
3893 return -EINVAL;
3897 return -EINVAL;
3899 mddev->safemode_delay = 0;
3901 unsigned long old_delay = mddev->safemode_delay;
3906 mddev->safemode_delay = new_delay;
3908 mod_timer(&mddev->safemode_timer, jiffies+1);
3920 spin_lock(&mddev->lock);
3921 p = mddev->pers;
3923 ret = sprintf(page, "%s\n", p->name);
3924 else if (mddev->clevel[0])
3925 ret = sprintf(page, "%s\n", mddev->clevel);
3926 else if (mddev->level != LEVEL_NONE)
3927 ret = sprintf(page, "%d\n", mddev->level);
3930 spin_unlock(&mddev->lock);
3946 return -EINVAL;
3952 if (mddev->pers == NULL) {
3953 strncpy(mddev->clevel, buf, slen);
3954 if (mddev->clevel[slen-1] == '\n')
3955 slen--;
3956 mddev->clevel[slen] = 0;
3957 mddev->level = LEVEL_NONE;
3961 rv = -EROFS;
3962 if (mddev->ro)
3966 * - array is not engaged in resync/recovery/reshape
3967 * - old personality can be suspended
3968 * - new personality will access other array.
3971 rv = -EBUSY;
3972 if (mddev->sync_thread ||
3973 test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
3974 mddev->reshape_position != MaxSector ||
3975 mddev->sysfs_active)
3978 rv = -EINVAL;
3979 if (!mddev->pers->quiesce) {
3981 mdname(mddev), mddev->pers->name);
3987 if (clevel[slen-1] == '\n')
3988 slen--;
3993 if (request_module("md-%s", clevel) != 0)
3994 request_module("md-level-%s", clevel);
3997 if (!pers || !try_module_get(pers->owner)) {
4000 rv = -EINVAL;
4005 if (pers == mddev->pers) {
4007 module_put(pers->owner);
4011 if (!pers->takeover) {
4012 module_put(pers->owner);
4015 rv = -EINVAL;
4020 rdev->new_raid_disk = rdev->raid_disk;
4022 /* ->takeover must set new_* and/or delta_disks
4025 priv = pers->takeover(mddev);
4027 mddev->new_level = mddev->level;
4028 mddev->new_layout = mddev->layout;
4029 mddev->new_chunk_sectors = mddev->chunk_sectors;
4030 mddev->raid_disks -= mddev->delta_disks;
4031 mddev->delta_disks = 0;
4032 mddev->reshape_backwards = 0;
4033 module_put(pers->owner);
4044 spin_lock(&mddev->lock);
4045 oldpers = mddev->pers;
4046 oldpriv = mddev->private;
4047 mddev->pers = pers;
4048 mddev->private = priv;
4049 strlcpy(mddev->clevel, pers->name, sizeof(mddev->clevel));
4050 mddev->level = mddev->new_level;
4051 mddev->layout = mddev->new_layout;
4052 mddev->chunk_sectors = mddev->new_chunk_sectors;
4053 mddev->delta_disks = 0;
4054 mddev->reshape_backwards = 0;
4055 mddev->degraded = 0;
4056 spin_unlock(&mddev->lock);
4058 if (oldpers->sync_request == NULL &&
4059 mddev->external) {
4060 /* We are converting from a no-redundancy array
4064 * clean->dirty
4067 mddev->in_sync = 0;
4068 mddev->safemode_delay = 0;
4069 mddev->safemode = 0;
4072 oldpers->free(mddev, oldpriv);
4074 if (oldpers->sync_request == NULL &&
4075 pers->sync_request != NULL) {
4077 if (sysfs_create_group(&mddev->kobj, &md_redundancy_group))
4080 mddev->sysfs_action = sysfs_get_dirent(mddev->kobj.sd, "sync_action");
4081 mddev->sysfs_completed = sysfs_get_dirent_safe(mddev->kobj.sd, "sync_completed");
4082 mddev->sysfs_degraded = sysfs_get_dirent_safe(mddev->kobj.sd, "degraded");
4084 if (oldpers->sync_request != NULL &&
4085 pers->sync_request == NULL) {
4087 if (mddev->to_remove == NULL)
4088 mddev->to_remove = &md_redundancy_group;
4091 module_put(oldpers->owner);
4094 if (rdev->raid_disk < 0)
4096 if (rdev->new_raid_disk >= mddev->raid_disks)
4097 rdev->new_raid_disk = -1;
4098 if (rdev->new_raid_disk == rdev->raid_disk)
4103 if (rdev->raid_disk < 0)
4105 if (rdev->new_raid_disk == rdev->raid_disk)
4107 rdev->raid_disk = rdev->new_raid_disk;
4108 if (rdev->raid_disk < 0)
4109 clear_bit(In_sync, &rdev->flags);
4113 rdev->raid_disk, mdname(mddev));
4117 if (pers->sync_request == NULL) {
4121 mddev->in_sync = 1;
4122 del_timer_sync(&mddev->safemode_timer);
4124 blk_set_stacking_limits(&mddev->queue->limits);
4125 pers->run(mddev);
4126 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
4128 if (!mddev->thread)
4130 sysfs_notify_dirent_safe(mddev->sysfs_level);
4145 if (mddev->reshape_position != MaxSector &&
4146 mddev->layout != mddev->new_layout)
4148 mddev->new_layout, mddev->layout);
4149 return sprintf(page, "%d\n", mddev->layout);
4165 if (mddev->pers) {
4166 if (mddev->pers->check_reshape == NULL)
4167 err = -EBUSY;
4168 else if (mddev->ro)
4169 err = -EROFS;
4171 mddev->new_layout = n;
4172 err = mddev->pers->check_reshape(mddev);
4174 mddev->new_layout = mddev->layout;
4177 mddev->new_layout = n;
4178 if (mddev->reshape_position == MaxSector)
4179 mddev->layout = n;
4190 if (mddev->raid_disks == 0)
4192 if (mddev->reshape_position != MaxSector &&
4193 mddev->delta_disks != 0)
4194 return sprintf(page, "%d (%d)\n", mddev->raid_disks,
4195 mddev->raid_disks - mddev->delta_disks);
4196 return sprintf(page, "%d\n", mddev->raid_disks);
4214 if (mddev->pers)
4216 else if (mddev->reshape_position != MaxSector) {
4218 int olddisks = mddev->raid_disks - mddev->delta_disks;
4220 err = -EINVAL;
4223 rdev->data_offset < rdev->new_data_offset)
4226 rdev->data_offset > rdev->new_data_offset)
4230 mddev->delta_disks = n - olddisks;
4231 mddev->raid_disks = n;
4232 mddev->reshape_backwards = (mddev->delta_disks < 0);
4234 mddev->raid_disks = n;
4245 return sprintf(page, "%pU\n", mddev->uuid);
4253 if (mddev->reshape_position != MaxSector &&
4254 mddev->chunk_sectors != mddev->new_chunk_sectors)
4256 mddev->new_chunk_sectors << 9,
4257 mddev->chunk_sectors << 9);
4258 return sprintf(page, "%d\n", mddev->chunk_sectors << 9);
4274 if (mddev->pers) {
4275 if (mddev->pers->check_reshape == NULL)
4276 err = -EBUSY;
4277 else if (mddev->ro)
4278 err = -EROFS;
4280 mddev->new_chunk_sectors = n >> 9;
4281 err = mddev->pers->check_reshape(mddev);
4283 mddev->new_chunk_sectors = mddev->chunk_sectors;
4286 mddev->new_chunk_sectors = n >> 9;
4287 if (mddev->reshape_position == MaxSector)
4288 mddev->chunk_sectors = n >> 9;
4299 if (mddev->recovery_cp == MaxSector)
4301 return sprintf(page, "%llu\n", (unsigned long long)mddev->recovery_cp);
4317 return -EINVAL;
4323 if (mddev->pers && !test_bit(MD_RECOVERY_FROZEN, &mddev->recovery))
4324 err = -EBUSY;
4327 mddev->recovery_cp = n;
4328 if (mddev->pers)
4329 set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
4354 * read-auto
4357 * clean - no pending writes, but otherwise active.
4361 * if not known, block and switch to write-pending
4367 * write-pending
4370 * active-idle
4374 * RAID0/LINEAR-only: same as clean, but array is missing a member.
4375 * It's useful because RAID0/LINEAR mounted-arrays aren't stopped
4382 "clear", "inactive", "suspended", "readonly", "read-auto", "clean", "active",
4383 "write-pending", "active-idle", "broken", NULL };
4399 if (mddev->pers && !test_bit(MD_NOT_READY, &mddev->flags)) {
4400 switch(mddev->ro) {
4408 spin_lock(&mddev->lock);
4409 if (test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags))
4411 else if (mddev->in_sync)
4413 else if (mddev->safemode)
4417 spin_unlock(&mddev->lock);
4420 if (test_bit(MD_BROKEN, &mddev->flags) && st == clean)
4423 if (list_empty(&mddev->disks) &&
4424 mddev->raid_disks == 0 &&
4425 mddev->dev_sectors == 0)
4443 if (mddev->pers && (st == active || st == clean) && mddev->ro != 1) {
4447 spin_lock(&mddev->lock);
4450 clear_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
4451 md_wakeup_thread(mddev->thread);
4452 wake_up(&mddev->sb_wait);
4456 err = -EBUSY;
4459 sysfs_notify_dirent_safe(mddev->sysfs_state);
4460 spin_unlock(&mddev->lock);
4466 err = -EINVAL;
4476 if (mddev->pers)
4484 if (mddev->pers)
4487 mddev->ro = 1;
4488 set_disk_ro(mddev->gendisk, 1);
4493 if (mddev->pers) {
4494 if (mddev->ro == 0)
4496 else if (mddev->ro == 1)
4499 mddev->ro = 2;
4500 set_disk_ro(mddev->gendisk, 0);
4503 mddev->ro = 2;
4508 if (mddev->pers) {
4512 spin_lock(&mddev->lock);
4514 err = -EBUSY;
4515 spin_unlock(&mddev->lock);
4517 err = -EINVAL;
4520 if (mddev->pers) {
4524 clear_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
4525 wake_up(&mddev->sb_wait);
4528 mddev->ro = 0;
4529 set_disk_ro(mddev->gendisk, 0);
4541 if (mddev->hold_active == UNTIL_IOCTL)
4542 mddev->hold_active = 0;
4543 sysfs_notify_dirent_safe(mddev->sysfs_state);
4554 atomic_read(&mddev->max_corr_read_errors));
4567 return -EINVAL;
4568 atomic_set(&mddev->max_corr_read_errors, n);
4579 return -EINVAL;
4589 if (work_pending(&rdev->del_work)) {
4614 return -EINVAL;
4617 return -EINVAL;
4621 return -EOVERFLOW;
4627 if (mddev->persistent) {
4628 rdev = md_import_device(dev, mddev->major_version,
4629 mddev->minor_version);
4630 if (!IS_ERR(rdev) && !list_empty(&mddev->disks)) {
4632 = list_entry(mddev->disks.next,
4634 err = super_types[mddev->major_version]
4635 .load_super(rdev, rdev0, mddev->minor_version);
4639 } else if (mddev->external)
4640 rdev = md_import_device(dev, -2, -1);
4642 rdev = md_import_device(dev, -1, -1);
4671 if (!mddev->bitmap)
4673 /* buf should be <chunk> <chunk> ... or <chunk>-<chunk> ... (range) */
4677 if (*end == '-') { /* range */
4683 md_bitmap_dirty_bits(mddev->bitmap, chunk, end_chunk);
4686 md_bitmap_unplug(mddev->bitmap); /* flush the bits to disk */
4699 (unsigned long long)mddev->dev_sectors / 2);
4709 * If array is active, we can try an on-line resize
4719 if (mddev->pers) {
4724 if (mddev->dev_sectors == 0 ||
4725 mddev->dev_sectors > sectors)
4726 mddev->dev_sectors = sectors;
4728 err = -ENOSPC;
4746 if (mddev->persistent)
4748 mddev->major_version, mddev->minor_version);
4749 else if (mddev->external)
4750 return sprintf(page, "external:%s\n", mddev->metadata_type);
4769 err = -EBUSY;
4770 if (mddev->external && strncmp(buf, "external:", 9) == 0)
4772 else if (!list_empty(&mddev->disks))
4777 mddev->persistent = 0;
4778 mddev->external = 0;
4779 mddev->major_version = 0;
4780 mddev->minor_version = 90;
4784 size_t namelen = len-9;
4785 if (namelen >= sizeof(mddev->metadata_type))
4786 namelen = sizeof(mddev->metadata_type)-1;
4787 strncpy(mddev->metadata_type, buf+9, namelen);
4788 mddev->metadata_type[namelen] = 0;
4789 if (namelen && mddev->metadata_type[namelen-1] == '\n')
4790 mddev->metadata_type[--namelen] = 0;
4791 mddev->persistent = 0;
4792 mddev->external = 1;
4793 mddev->major_version = 0;
4794 mddev->minor_version = 90;
4798 err = -EINVAL;
4805 err = -ENOENT;
4808 mddev->major_version = major;
4809 mddev->minor_version = minor;
4810 mddev->persistent = 1;
4811 mddev->external = 0;
4825 unsigned long recovery = mddev->recovery;
4829 (!mddev->ro && test_bit(MD_RECOVERY_NEEDED, &recovery))) {
4841 else if (mddev->reshape_position != MaxSector)
4850 if (!mddev->pers || !mddev->pers->sync_request)
4851 return -EINVAL;
4856 set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4858 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4859 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) &&
4861 if (work_pending(&mddev->del_work))
4863 if (mddev->sync_thread) {
4864 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
4869 } else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
4870 return -EBUSY;
4872 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4874 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4875 set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
4878 if (mddev->pers->start_reshape == NULL)
4879 return -EINVAL;
4882 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) {
4883 err = -EBUSY;
4884 } else if (mddev->reshape_position == MaxSector ||
4885 mddev->pers->check_reshape == NULL ||
4886 mddev->pers->check_reshape(mddev)) {
4887 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4888 err = mddev->pers->start_reshape(mddev);
4896 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4902 sysfs_notify_dirent_safe(mddev->sysfs_degraded);
4905 set_bit(MD_RECOVERY_CHECK, &mddev->recovery);
4907 return -EINVAL;
4908 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4909 set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
4910 set_bit(MD_RECOVERY_SYNC, &mddev->recovery);
4912 if (mddev->ro == 2) {
4914 * canceling read-auto mode
4916 mddev->ro = 0;
4917 md_wakeup_thread(mddev->sync_thread);
4919 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
4920 md_wakeup_thread(mddev->thread);
4921 sysfs_notify_dirent_safe(mddev->sysfs_action);
4931 return sprintf(page, "%s\n", mddev->last_sync_action);
4941 atomic64_read(&mddev->resync_mismatches));
4950 mddev->sync_speed_min ? "local": "system");
4966 return -EINVAL;
4968 mddev->sync_speed_min = min;
4979 mddev->sync_speed_max ? "local": "system");
4995 return -EINVAL;
4997 mddev->sync_speed_max = max;
5007 return sprintf(page, "%d\n", mddev->degraded);
5014 return sprintf(page, "%d\n", mddev->parallel_resync);
5023 return -EINVAL;
5026 return -EINVAL;
5028 mddev->parallel_resync = n;
5030 if (mddev->sync_thread)
5045 if (mddev->curr_resync == 0)
5047 resync = mddev->curr_mark_cnt - atomic_read(&mddev->recovery_active);
5048 dt = (jiffies - mddev->resync_mark) / HZ;
5050 db = resync - mddev->resync_mark_cnt;
5061 if (!test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
5064 if (mddev->curr_resync == 1 ||
5065 mddev->curr_resync == 2)
5068 if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ||
5069 test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
5070 max_sectors = mddev->resync_max_sectors;
5072 max_sectors = mddev->dev_sectors;
5074 resync = mddev->curr_resync_completed;
5085 (unsigned long long)mddev->resync_min);
5094 return -EINVAL;
5096 spin_lock(&mddev->lock);
5097 err = -EINVAL;
5098 if (min > mddev->resync_max)
5101 err = -EBUSY;
5102 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
5106 mddev->resync_min = round_down(min, 8);
5110 spin_unlock(&mddev->lock);
5120 if (mddev->resync_max == MaxSector)
5124 (unsigned long long)mddev->resync_max);
5130 spin_lock(&mddev->lock);
5132 mddev->resync_max = MaxSector;
5137 err = -EINVAL;
5140 if (max < mddev->resync_min)
5143 err = -EBUSY;
5144 if (max < mddev->resync_max &&
5145 mddev->ro == 0 &&
5146 test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
5150 chunk = mddev->chunk_sectors;
5154 err = -EINVAL;
5158 mddev->resync_max = max;
5160 wake_up(&mddev->recovery_wait);
5163 spin_unlock(&mddev->lock);
5173 return sprintf(page, "%llu\n", (unsigned long long)mddev->suspend_lo);
5186 return -EINVAL;
5191 err = -EINVAL;
5192 if (mddev->pers == NULL ||
5193 mddev->pers->quiesce == NULL)
5196 mddev->suspend_lo = new;
5210 return sprintf(page, "%llu\n", (unsigned long long)mddev->suspend_hi);
5223 return -EINVAL;
5228 err = -EINVAL;
5229 if (mddev->pers == NULL)
5233 mddev->suspend_hi = new;
5247 if (mddev->reshape_position != MaxSector)
5249 (unsigned long long)mddev->reshape_position);
5265 return -EINVAL;
5269 err = -EBUSY;
5270 if (mddev->pers)
5272 mddev->reshape_position = new;
5273 mddev->delta_disks = 0;
5274 mddev->reshape_backwards = 0;
5275 mddev->new_level = mddev->level;
5276 mddev->new_layout = mddev->layout;
5277 mddev->new_chunk_sectors = mddev->chunk_sectors;
5279 rdev->new_data_offset = rdev->data_offset;
5294 mddev->reshape_backwards ? "backwards" : "forwards");
5308 return -EINVAL;
5309 if (mddev->reshape_backwards == backwards)
5316 if (mddev->delta_disks)
5317 err = -EBUSY;
5318 else if (mddev->persistent &&
5319 mddev->major_version == 0)
5320 err = -EINVAL;
5322 mddev->reshape_backwards = backwards;
5334 if (mddev->external_size)
5336 (unsigned long long)mddev->array_sectors/2);
5354 return -EINVAL;
5358 if (mddev->pers)
5359 sectors = mddev->pers->size(mddev, 0, 0);
5361 sectors = mddev->array_sectors;
5363 mddev->external_size = 0;
5366 err = -EINVAL;
5367 else if (mddev->pers && mddev->pers->size(mddev, 0, 0) < sectors)
5368 err = -E2BIG;
5370 mddev->external_size = 1;
5374 mddev->array_sectors = sectors;
5375 if (mddev->pers) {
5376 set_capacity(mddev->gendisk, mddev->array_sectors);
5377 revalidate_disk_size(mddev->gendisk, true);
5393 if (test_bit(MD_HAS_JOURNAL, &mddev->flags)) {
5395 } else if (test_bit(MD_HAS_PPL, &mddev->flags)) {
5397 } else if (mddev->bitmap) {
5399 } else if (mddev->pers) {
5400 if (mddev->pers->sync_request)
5416 if (mddev->pers) {
5417 if (mddev->pers->change_consistency_policy)
5418 err = mddev->pers->change_consistency_policy(mddev, buf);
5420 err = -EBUSY;
5421 } else if (mddev->external && strncmp(buf, "ppl", 3) == 0) {
5422 set_bit(MD_HAS_PPL, &mddev->flags);
5424 err = -EINVAL;
5436 return sprintf(page, "%d\n", mddev->fail_last_dev);
5453 if (value != mddev->fail_last_dev)
5454 mddev->fail_last_dev = value;
5464 if (mddev->pers == NULL || (mddev->pers->level != 1))
5467 return sprintf(page, "%d\n", mddev->serialize_policy);
5484 if (value == mddev->serialize_policy)
5490 if (mddev->pers == NULL || (mddev->pers->level != 1)) {
5492 err = -EINVAL;
5501 mddev->serialize_policy = value;
5564 if (!entry->show)
5565 return -EIO;
5567 if (list_empty(&mddev->all_mddevs)) {
5569 return -EBUSY;
5574 rv = entry->show(mddev, page);
5587 if (!entry->store)
5588 return -EIO;
5590 return -EACCES;
5592 if (list_empty(&mddev->all_mddevs)) {
5594 return -EBUSY;
5598 rv = entry->store(mddev, page, length);
5607 if (mddev->sysfs_state)
5608 sysfs_put(mddev->sysfs_state);
5609 if (mddev->sysfs_level)
5610 sysfs_put(mddev->sysfs_level);
5612 if (mddev->gendisk)
5613 del_gendisk(mddev->gendisk);
5614 if (mddev->queue)
5615 blk_cleanup_queue(mddev->queue);
5616 if (mddev->gendisk)
5617 put_disk(mddev->gendisk);
5618 percpu_ref_exit(&mddev->writes_pending);
5620 bioset_exit(&mddev->bio_set);
5621 bioset_exit(&mddev->sync_set);
5641 sysfs_remove_group(&mddev->kobj, &md_bitmap_group);
5642 kobject_del(&mddev->kobj);
5643 kobject_put(&mddev->kobj);
5650 if (mddev->writes_pending.percpu_count_ptr)
5652 if (percpu_ref_init(&mddev->writes_pending, no_op,
5654 return -ENOMEM;
5656 percpu_ref_put(&mddev->writes_pending);
5666 * If dev is non-zero it must be a device number with a MAJOR of
5681 return -ENODEV;
5683 partitioned = (MAJOR(mddev->unit) != MD_MAJOR);
5685 unit = MINOR(mddev->unit) >> shift;
5694 error = -EEXIST;
5695 if (mddev->gendisk)
5705 if (mddev2->gendisk &&
5706 strcmp(mddev2->gendisk->disk_name, name) == 0) {
5716 mddev->hold_active = UNTIL_STOP;
5718 error = -ENOMEM;
5719 mddev->queue = blk_alloc_queue(NUMA_NO_NODE);
5720 if (!mddev->queue)
5723 blk_set_stacking_limits(&mddev->queue->limits);
5727 blk_cleanup_queue(mddev->queue);
5728 mddev->queue = NULL;
5731 disk->major = MAJOR(mddev->unit);
5732 disk->first_minor = unit << shift;
5734 strcpy(disk->disk_name, name);
5736 sprintf(disk->disk_name, "md_d%d", unit);
5738 sprintf(disk->disk_name, "md%d", unit);
5739 disk->fops = &md_fops;
5740 disk->private_data = mddev;
5741 disk->queue = mddev->queue;
5742 blk_queue_write_cache(mddev->queue, true, true);
5747 disk->flags |= GENHD_FL_EXT_DEVT;
5748 disk->events |= DISK_EVENT_MEDIA_CHANGE;
5749 mddev->gendisk = disk;
5752 error = kobject_add(&mddev->kobj, &disk_to_dev(disk)->kobj, "%s", "md");
5757 pr_debug("md: cannot register %s/md - name in use\n",
5758 disk->disk_name);
5761 if (mddev->kobj.sd &&
5762 sysfs_create_group(&mddev->kobj, &md_bitmap_group))
5766 if (!error && mddev->kobj.sd) {
5767 kobject_uevent(&mddev->kobj, KOBJ_ADD);
5768 mddev->sysfs_state = sysfs_get_dirent_safe(mddev->kobj.sd, "array_state");
5769 mddev->sysfs_level = sysfs_get_dirent_safe(mddev->kobj.sd, "level");
5795 while (len && val[len-1] == '\n')
5796 len--;
5798 return -E2BIG;
5808 return -EINVAL;
5815 mddev->safemode = 1;
5816 if (mddev->external)
5817 sysfs_notify_dirent_safe(mddev->sysfs_state);
5819 md_wakeup_thread(mddev->thread);
5830 if (list_empty(&mddev->disks))
5832 return -EINVAL;
5834 if (mddev->pers)
5835 return -EBUSY;
5837 if (mddev->sysfs_active)
5838 return -EBUSY;
5843 if (!mddev->raid_disks) {
5844 if (!mddev->persistent)
5845 return -EINVAL;
5848 return -EINVAL;
5851 if (mddev->level != LEVEL_NONE)
5852 request_module("md-level-%d", mddev->level);
5853 else if (mddev->clevel[0])
5854 request_module("md-%s", mddev->clevel);
5861 mddev->has_superblocks = false;
5863 if (test_bit(Faulty, &rdev->flags))
5865 sync_blockdev(rdev->bdev);
5866 invalidate_bdev(rdev->bdev);
5867 if (mddev->ro != 1 &&
5868 (bdev_read_only(rdev->bdev) ||
5869 bdev_read_only(rdev->meta_bdev))) {
5870 mddev->ro = 1;
5871 if (mddev->gendisk)
5872 set_disk_ro(mddev->gendisk, 1);
5875 if (rdev->sb_page)
5876 mddev->has_superblocks = true;
5882 if (rdev->meta_bdev) {
5884 } else if (rdev->data_offset < rdev->sb_start) {
5885 if (mddev->dev_sectors &&
5886 rdev->data_offset + mddev->dev_sectors
5887 > rdev->sb_start) {
5890 return -EINVAL;
5893 if (rdev->sb_start + rdev->sb_size/512
5894 > rdev->data_offset) {
5897 return -EINVAL;
5900 sysfs_notify_dirent_safe(rdev->sysfs_state);
5903 if (!bioset_initialized(&mddev->bio_set)) {
5904 err = bioset_init(&mddev->bio_set, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS);
5908 if (!bioset_initialized(&mddev->sync_set)) {
5909 err = bioset_init(&mddev->sync_set, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS);
5915 pers = find_pers(mddev->level, mddev->clevel);
5916 if (!pers || !try_module_get(pers->owner)) {
5918 if (mddev->level != LEVEL_NONE)
5920 mddev->level);
5923 mddev->clevel);
5924 err = -EINVAL;
5928 if (mddev->level != pers->level) {
5929 mddev->level = pers->level;
5930 mddev->new_level = pers->level;
5932 strlcpy(mddev->clevel, pers->name, sizeof(mddev->clevel));
5934 if (mddev->reshape_position != MaxSector &&
5935 pers->start_reshape == NULL) {
5937 module_put(pers->owner);
5938 err = -EINVAL;
5942 if (pers->sync_request) {
5953 rdev->bdev->bd_disk ==
5954 rdev2->bdev->bd_disk) {
5957 bdevname(rdev->bdev,b),
5958 bdevname(rdev2->bdev,b2));
5964 pr_warn("True protection against single-disk failure might be compromised.\n");
5967 mddev->recovery = 0;
5968 /* may be over-ridden by personality */
5969 mddev->resync_max_sectors = mddev->dev_sectors;
5971 mddev->ok_start_degraded = start_dirty_degraded;
5973 if (start_readonly && mddev->ro == 0)
5974 mddev->ro = 2; /* read-only, but switch on first write */
5976 err = pers->run(mddev);
5978 pr_warn("md: pers->run() failed ...\n");
5979 else if (pers->size(mddev, 0, 0) < mddev->array_sectors) {
5980 WARN_ONCE(!mddev->external_size,
5984 (unsigned long long)mddev->array_sectors / 2,
5985 (unsigned long long)pers->size(mddev, 0, 0) / 2);
5986 err = -EINVAL;
5988 if (err == 0 && pers->sync_request &&
5989 (mddev->bitmap_info.file || mddev->bitmap_info.offset)) {
5992 bitmap = md_bitmap_create(mddev, -1);
5998 mddev->bitmap = bitmap;
6004 if (mddev->bitmap_info.max_write_behind > 0) {
6008 if (test_bit(WriteMostly, &rdev->flags) &&
6012 if (create_pool && mddev->serial_info_pool == NULL) {
6013 mddev->serial_info_pool =
6016 if (!mddev->serial_info_pool) {
6017 err = -ENOMEM;
6023 if (mddev->queue) {
6027 if (rdev->raid_disk >= 0 &&
6028 !blk_queue_nonrot(bdev_get_queue(rdev->bdev))) {
6033 if (mddev->degraded)
6036 blk_queue_flag_set(QUEUE_FLAG_NONROT, mddev->queue);
6038 blk_queue_flag_clear(QUEUE_FLAG_NONROT, mddev->queue);
6040 if (pers->sync_request) {
6041 if (mddev->kobj.sd &&
6042 sysfs_create_group(&mddev->kobj, &md_redundancy_group))
6045 mddev->sysfs_action = sysfs_get_dirent_safe(mddev->kobj.sd, "sync_action");
6046 mddev->sysfs_completed = sysfs_get_dirent_safe(mddev->kobj.sd, "sync_completed");
6047 mddev->sysfs_degraded = sysfs_get_dirent_safe(mddev->kobj.sd, "degraded");
6048 } else if (mddev->ro == 2) /* auto-readonly not meaningful */
6049 mddev->ro = 0;
6051 atomic_set(&mddev->max_corr_read_errors,
6053 mddev->safemode = 0;
6055 mddev->safemode_delay = 0;
6057 mddev->safemode_delay = DEFAULT_SAFEMODE_DELAY;
6058 mddev->in_sync = 1;
6060 spin_lock(&mddev->lock);
6061 mddev->pers = pers;
6062 spin_unlock(&mddev->lock);
6064 if (rdev->raid_disk >= 0)
6067 if (mddev->degraded && !mddev->ro)
6069 * via sysfs - until a lack of spares is confirmed.
6071 set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
6072 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
6074 if (mddev->sb_flags)
6082 if (mddev->private)
6083 pers->free(mddev, mddev->private);
6084 mddev->private = NULL;
6085 module_put(pers->owner);
6088 bioset_exit(&mddev->bio_set);
6089 bioset_exit(&mddev->sync_set);
6098 set_bit(MD_NOT_READY, &mddev->flags);
6114 md_wakeup_thread(mddev->thread);
6115 md_wakeup_thread(mddev->sync_thread); /* possibly kick off a reshape */
6117 set_capacity(mddev->gendisk, mddev->array_sectors);
6118 revalidate_disk_size(mddev->gendisk, true);
6119 clear_bit(MD_NOT_READY, &mddev->flags);
6120 mddev->changed = 1;
6121 kobject_uevent(&disk_to_dev(mddev->gendisk)->kobj, KOBJ_CHANGE);
6122 sysfs_notify_dirent_safe(mddev->sysfs_state);
6123 sysfs_notify_dirent_safe(mddev->sysfs_action);
6124 sysfs_notify_dirent_safe(mddev->sysfs_degraded);
6126 clear_bit(MD_NOT_READY, &mddev->flags);
6134 if (mddev->pers->start) {
6135 set_bit(MD_RECOVERY_WAIT, &mddev->recovery);
6136 md_wakeup_thread(mddev->thread);
6137 ret = mddev->pers->start(mddev);
6138 clear_bit(MD_RECOVERY_WAIT, &mddev->recovery);
6139 md_wakeup_thread(mddev->sync_thread);
6147 struct gendisk *disk = mddev->gendisk;
6153 if (list_empty(&mddev->disks))
6154 return -ENXIO;
6155 if (!mddev->pers)
6156 return -EINVAL;
6157 if (!mddev->ro)
6158 return -EBUSY;
6162 if (test_bit(Journal, &rdev->flags) &&
6163 !test_bit(Faulty, &rdev->flags))
6165 if (bdev_read_only(rdev->bdev))
6169 if (test_bit(MD_HAS_JOURNAL, &mddev->flags) && !has_journal)
6171 return -EINVAL;
6173 return -EROFS;
6175 mddev->safemode = 0;
6176 mddev->ro = 0;
6178 pr_debug("md: %s switched to read-write mode.\n", mdname(mddev));
6180 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
6181 md_wakeup_thread(mddev->thread);
6182 md_wakeup_thread(mddev->sync_thread);
6183 sysfs_notify_dirent_safe(mddev->sysfs_state);
6189 mddev->array_sectors = 0;
6190 mddev->external_size = 0;
6191 mddev->dev_sectors = 0;
6192 mddev->raid_disks = 0;
6193 mddev->recovery_cp = 0;
6194 mddev->resync_min = 0;
6195 mddev->resync_max = MaxSector;
6196 mddev->reshape_position = MaxSector;
6197 mddev->external = 0;
6198 mddev->persistent = 0;
6199 mddev->level = LEVEL_NONE;
6200 mddev->clevel[0] = 0;
6201 mddev->flags = 0;
6202 mddev->sb_flags = 0;
6203 mddev->ro = 0;
6204 mddev->metadata_type[0] = 0;
6205 mddev->chunk_sectors = 0;
6206 mddev->ctime = mddev->utime = 0;
6207 mddev->layout = 0;
6208 mddev->max_disks = 0;
6209 mddev->events = 0;
6210 mddev->can_decrease_events = 0;
6211 mddev->delta_disks = 0;
6212 mddev->reshape_backwards = 0;
6213 mddev->new_level = LEVEL_NONE;
6214 mddev->new_layout = 0;
6215 mddev->new_chunk_sectors = 0;
6216 mddev->curr_resync = 0;
6217 atomic64_set(&mddev->resync_mismatches, 0);
6218 mddev->suspend_lo = mddev->suspend_hi = 0;
6219 mddev->sync_speed_min = mddev->sync_speed_max = 0;
6220 mddev->recovery = 0;
6221 mddev->in_sync = 0;
6222 mddev->changed = 0;
6223 mddev->degraded = 0;
6224 mddev->safemode = 0;
6225 mddev->private = NULL;
6226 mddev->cluster_info = NULL;
6227 mddev->bitmap_info.offset = 0;
6228 mddev->bitmap_info.default_offset = 0;
6229 mddev->bitmap_info.default_space = 0;
6230 mddev->bitmap_info.chunksize = 0;
6231 mddev->bitmap_info.daemon_sleep = 0;
6232 mddev->bitmap_info.max_write_behind = 0;
6233 mddev->bitmap_info.nodes = 0;
6238 set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
6239 if (work_pending(&mddev->del_work))
6241 if (mddev->sync_thread) {
6242 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
6246 del_timer_sync(&mddev->safemode_timer);
6248 if (mddev->pers && mddev->pers->quiesce) {
6249 mddev->pers->quiesce(mddev, 1);
6250 mddev->pers->quiesce(mddev, 0);
6254 if (mddev->ro == 0 &&
6255 ((!mddev->in_sync && !mddev_is_clustered(mddev)) ||
6256 mddev->sb_flags)) {
6259 mddev->in_sync = 1;
6263 mddev->serialize_policy = 0;
6278 if (mddev->pers && mddev->pers->quiesce && !mddev->suspended) {
6279 mddev->pers->quiesce(mddev, 1);
6280 mddev->pers->quiesce(mddev, 0);
6282 md_unregister_thread(&mddev->thread);
6283 if (mddev->queue)
6284 blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
6289 struct md_personality *pers = mddev->pers;
6292 /* Ensure ->event_work is done */
6293 if (mddev->event_work.func)
6295 spin_lock(&mddev->lock);
6296 mddev->pers = NULL;
6297 spin_unlock(&mddev->lock);
6298 pers->free(mddev, mddev->private);
6299 mddev->private = NULL;
6300 if (pers->sync_request && mddev->to_remove == NULL)
6301 mddev->to_remove = &md_redundancy_group;
6302 module_put(pers->owner);
6303 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
6308 lockdep_assert_held(&mddev->reconfig_mutex);
6311 * This is called from dm-raid
6315 bioset_exit(&mddev->bio_set);
6316 bioset_exit(&mddev->sync_set);
6326 if (!test_bit(MD_RECOVERY_FROZEN, &mddev->recovery)) {
6328 set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
6329 md_wakeup_thread(mddev->thread);
6331 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
6332 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
6333 if (mddev->sync_thread)
6336 wake_up_process(mddev->sync_thread->tsk);
6338 if (mddev->external && test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags))
6339 return -EBUSY;
6342 &mddev->recovery));
6343 wait_event(mddev->sb_wait,
6344 !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags));
6347 mutex_lock(&mddev->open_mutex);
6348 if ((mddev->pers && atomic_read(&mddev->openers) > !!bdev) ||
6349 mddev->sync_thread ||
6350 test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) {
6353 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
6354 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
6355 md_wakeup_thread(mddev->thread);
6357 err = -EBUSY;
6360 if (mddev->pers) {
6363 err = -ENXIO;
6364 if (mddev->ro==1)
6366 mddev->ro = 1;
6367 set_disk_ro(mddev->gendisk, 1);
6368 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
6369 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
6370 md_wakeup_thread(mddev->thread);
6371 sysfs_notify_dirent_safe(mddev->sysfs_state);
6375 mutex_unlock(&mddev->open_mutex);
6380 * 0 - completely stop and dis-assemble array
6381 * 2 - stop but do not disassemble array
6386 struct gendisk *disk = mddev->gendisk;
6390 if (!test_bit(MD_RECOVERY_FROZEN, &mddev->recovery)) {
6392 set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
6393 md_wakeup_thread(mddev->thread);
6395 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
6396 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
6397 if (mddev->sync_thread)
6400 wake_up_process(mddev->sync_thread->tsk);
6403 wait_event(resync_wait, (mddev->sync_thread == NULL &&
6405 &mddev->recovery)));
6408 mutex_lock(&mddev->open_mutex);
6409 if ((mddev->pers && atomic_read(&mddev->openers) > !!bdev) ||
6410 mddev->sysfs_active ||
6411 mddev->sync_thread ||
6412 test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) {
6414 mutex_unlock(&mddev->open_mutex);
6416 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
6417 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
6418 md_wakeup_thread(mddev->thread);
6420 return -EBUSY;
6422 if (mddev->pers) {
6423 if (mddev->ro)
6430 sysfs_notify_dirent_safe(mddev->sysfs_state);
6433 if (rdev->raid_disk >= 0)
6437 mutex_unlock(&mddev->open_mutex);
6438 mddev->changed = 1;
6441 if (mddev->ro)
6442 mddev->ro = 0;
6444 mutex_unlock(&mddev->open_mutex);
6451 if (mddev->bitmap_info.file) {
6452 struct file *f = mddev->bitmap_info.file;
6453 spin_lock(&mddev->lock);
6454 mddev->bitmap_info.file = NULL;
6455 spin_unlock(&mddev->lock);
6458 mddev->bitmap_info.offset = 0;
6463 if (mddev->hold_active == UNTIL_STOP)
6464 mddev->hold_active = 0;
6467 sysfs_notify_dirent_safe(mddev->sysfs_state);
6477 if (list_empty(&mddev->disks))
6484 pr_cont("<%s>", bdevname(rdev->bdev,b));
6521 pr_debug("md: considering %s ...\n", bdevname(rdev0->bdev,b));
6526 bdevname(rdev->bdev,b));
6527 list_move(&rdev->same_set, &candidates);
6536 rdev0->preferred_minor << MdpMinorShift);
6539 dev = MKDEV(MD_MAJOR, rdev0->preferred_minor);
6542 if (rdev0->preferred_minor != unit) {
6544 bdevname(rdev0->bdev, b), rdev0->preferred_minor);
6555 else if (mddev->raid_disks || mddev->major_version
6556 || !list_empty(&mddev->disks)) {
6558 mdname(mddev), bdevname(rdev0->bdev,b));
6562 mddev->persistent = 1;
6564 list_del_init(&rdev->same_set);
6575 list_del_init(&rdev->same_set);
6593 return -EFAULT;
6608 if (test_bit(Faulty, &rdev->flags))
6612 if (test_bit(In_sync, &rdev->flags))
6614 else if (test_bit(Journal, &rdev->flags))
6623 info.major_version = mddev->major_version;
6624 info.minor_version = mddev->minor_version;
6626 info.ctime = clamp_t(time64_t, mddev->ctime, 0, U32_MAX);
6627 info.level = mddev->level;
6628 info.size = mddev->dev_sectors / 2;
6629 if (info.size != mddev->dev_sectors / 2) /* overflow */
6630 info.size = -1;
6632 info.raid_disks = mddev->raid_disks;
6633 info.md_minor = mddev->md_minor;
6634 info.not_persistent= !mddev->persistent;
6636 info.utime = clamp_t(time64_t, mddev->utime, 0, U32_MAX);
6638 if (mddev->in_sync)
6640 if (mddev->bitmap && mddev->bitmap_info.offset)
6649 info.layout = mddev->layout;
6650 info.chunk_size = mddev->chunk_sectors << 9;
6653 return -EFAULT;
6666 return -ENOMEM;
6669 spin_lock(&mddev->lock);
6671 if (mddev->bitmap_info.file) {
6672 ptr = file_path(mddev->bitmap_info.file, file->pathname,
6673 sizeof(file->pathname));
6677 memmove(file->pathname, ptr,
6678 sizeof(file->pathname)-(ptr-file->pathname));
6680 spin_unlock(&mddev->lock);
6684 err = -EFAULT;
6696 return -EFAULT;
6701 info.major = MAJOR(rdev->bdev->bd_dev);
6702 info.minor = MINOR(rdev->bdev->bd_dev);
6703 info.raid_disk = rdev->raid_disk;
6705 if (test_bit(Faulty, &rdev->flags))
6707 else if (test_bit(In_sync, &rdev->flags)) {
6711 if (test_bit(Journal, &rdev->flags))
6713 if (test_bit(WriteMostly, &rdev->flags))
6715 if (test_bit(FailFast, &rdev->flags))
6719 info.raid_disk = -1;
6725 return -EFAULT;
6734 dev_t dev = MKDEV(info->major,info->minor);
6737 !(info->state & ((1 << MD_DISK_CLUSTER_ADD) | (1 << MD_DISK_CANDIDATE)))) {
6740 return -EINVAL;
6743 if (info->major != MAJOR(dev) || info->minor != MINOR(dev))
6744 return -EOVERFLOW;
6746 if (!mddev->raid_disks) {
6749 rdev = md_import_device(dev, mddev->major_version, mddev->minor_version);
6755 if (!list_empty(&mddev->disks)) {
6757 = list_entry(mddev->disks.next,
6759 err = super_types[mddev->major_version]
6760 .load_super(rdev, rdev0, mddev->minor_version);
6763 bdevname(rdev->bdev,b),
6764 bdevname(rdev0->bdev,b2));
6766 return -EINVAL;
6780 if (mddev->pers) {
6782 if (!mddev->pers->hot_add_disk) {
6785 return -EINVAL;
6787 if (mddev->persistent)
6788 rdev = md_import_device(dev, mddev->major_version,
6789 mddev->minor_version);
6791 rdev = md_import_device(dev, -1, -1);
6798 if (!mddev->persistent) {
6799 if (info->state & (1<<MD_DISK_SYNC) &&
6800 info->raid_disk < mddev->raid_disks) {
6801 rdev->raid_disk = info->raid_disk;
6802 set_bit(In_sync, &rdev->flags);
6803 clear_bit(Bitmap_sync, &rdev->flags);
6805 rdev->raid_disk = -1;
6806 rdev->saved_raid_disk = rdev->raid_disk;
6808 super_types[mddev->major_version].
6810 if ((info->state & (1<<MD_DISK_SYNC)) &&
6811 rdev->raid_disk != info->raid_disk) {
6812 /* This was a hot-add request, but events doesn't
6816 return -EINVAL;
6819 clear_bit(In_sync, &rdev->flags); /* just to be sure */
6820 if (info->state & (1<<MD_DISK_WRITEMOSTLY))
6821 set_bit(WriteMostly, &rdev->flags);
6823 clear_bit(WriteMostly, &rdev->flags);
6824 if (info->state & (1<<MD_DISK_FAILFAST))
6825 set_bit(FailFast, &rdev->flags);
6827 clear_bit(FailFast, &rdev->flags);
6829 if (info->state & (1<<MD_DISK_JOURNAL)) {
6835 if (test_bit(Journal, &rdev2->flags)) {
6840 if (has_journal || mddev->bitmap) {
6842 return -EBUSY;
6844 set_bit(Journal, &rdev->flags);
6850 if (info->state & (1 << MD_DISK_CANDIDATE))
6851 set_bit(Candidate, &rdev->flags);
6852 else if (info->state & (1 << MD_DISK_CLUSTER_ADD)) {
6853 /* --add initiated by this node */
6854 err = md_cluster_ops->add_new_disk(mddev, rdev);
6862 rdev->raid_disk = -1;
6869 if (info->state & (1 << MD_DISK_CANDIDATE)) {
6871 err = md_cluster_ops->new_disk_ack(mddev,
6878 md_cluster_ops->add_new_disk_cancel(mddev);
6892 if (mddev->major_version != 0) {
6894 return -EINVAL;
6897 if (!(info->state & (1<<MD_DISK_FAULTY))) {
6899 rdev = md_import_device(dev, -1, 0);
6905 rdev->desc_nr = info->number;
6906 if (info->raid_disk < mddev->raid_disks)
6907 rdev->raid_disk = info->raid_disk;
6909 rdev->raid_disk = -1;
6911 if (rdev->raid_disk < mddev->raid_disks)
6912 if (info->state & (1<<MD_DISK_SYNC))
6913 set_bit(In_sync, &rdev->flags);
6915 if (info->state & (1<<MD_DISK_WRITEMOSTLY))
6916 set_bit(WriteMostly, &rdev->flags);
6917 if (info->state & (1<<MD_DISK_FAILFAST))
6918 set_bit(FailFast, &rdev->flags);
6920 if (!mddev->persistent) {
6922 rdev->sb_start = i_size_read(rdev->bdev->bd_inode) / 512;
6924 rdev->sb_start = calc_dev_sboffset(rdev);
6925 rdev->sectors = rdev->sb_start;
6942 if (!mddev->pers)
6943 return -ENODEV;
6947 return -ENXIO;
6949 if (rdev->raid_disk < 0)
6952 clear_bit(Blocked, &rdev->flags);
6955 if (rdev->raid_disk >= 0)
6960 if (md_cluster_ops->remove_disk(mddev, rdev))
6965 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
6966 if (mddev->thread)
6967 md_wakeup_thread(mddev->thread);
6975 bdevname(rdev->bdev,b), mdname(mddev));
6976 return -EBUSY;
6985 if (!mddev->pers)
6986 return -ENODEV;
6988 if (mddev->major_version != 0) {
6989 pr_warn("%s: HOT_ADD may only be used with version-0 superblocks.\n",
6991 return -EINVAL;
6993 if (!mddev->pers->hot_add_disk) {
6996 return -EINVAL;
6999 rdev = md_import_device(dev, -1, 0);
7003 return -EINVAL;
7006 if (mddev->persistent)
7007 rdev->sb_start = calc_dev_sboffset(rdev);
7009 rdev->sb_start = i_size_read(rdev->bdev->bd_inode) / 512;
7011 rdev->sectors = rdev->sb_start;
7013 if (test_bit(Faulty, &rdev->flags)) {
7014 pr_warn("md: can not hot-add faulty %s disk to %s!\n",
7015 bdevname(rdev->bdev,b), mdname(mddev));
7016 err = -EINVAL;
7020 clear_bit(In_sync, &rdev->flags);
7021 rdev->desc_nr = -1;
7022 rdev->saved_raid_disk = -1;
7032 rdev->raid_disk = -1;
7034 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
7035 if (!mddev->thread)
7041 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
7042 md_wakeup_thread(mddev->thread);
7055 if (mddev->pers) {
7056 if (!mddev->pers->quiesce || !mddev->thread)
7057 return -EBUSY;
7058 if (mddev->recovery || mddev->sync_thread)
7059 return -EBUSY;
7067 if (mddev->bitmap || mddev->bitmap_info.file)
7068 return -EEXIST; /* cannot add when bitmap is present */
7074 return -EBADF;
7077 inode = f->f_mapping->host;
7078 if (!S_ISREG(inode->i_mode)) {
7081 err = -EBADF;
7082 } else if (!(f->f_mode & FMODE_WRITE)) {
7085 err = -EBADF;
7086 } else if (atomic_read(&inode->i_writecount) != 1) {
7089 err = -EBUSY;
7095 mddev->bitmap_info.file = f;
7096 mddev->bitmap_info.offset = 0; /* file overrides offset */
7097 } else if (mddev->bitmap == NULL)
7098 return -ENOENT; /* cannot remove what isn't there */
7100 if (mddev->pers) {
7104 bitmap = md_bitmap_create(mddev, -1);
7107 mddev->bitmap = bitmap;
7113 fd = -1;
7123 struct file *f = mddev->bitmap_info.file;
7125 spin_lock(&mddev->lock);
7126 mddev->bitmap_info.file = NULL;
7127 spin_unlock(&mddev->lock);
7141 * This will always create an array with a type-0.90.0 superblock.
7144 * use to determine which style super-blocks are to be found on the devices.
7150 if (info->raid_disks == 0) {
7152 if (info->major_version < 0 ||
7153 info->major_version >= ARRAY_SIZE(super_types) ||
7154 super_types[info->major_version].name == NULL) {
7155 /* maybe try to auto-load a module? */
7157 info->major_version);
7158 return -EINVAL;
7160 mddev->major_version = info->major_version;
7161 mddev->minor_version = info->minor_version;
7162 mddev->patch_version = info->patch_version;
7163 mddev->persistent = !info->not_persistent;
7167 mddev->ctime = ktime_get_real_seconds();
7170 mddev->major_version = MD_MAJOR_VERSION;
7171 mddev->minor_version = MD_MINOR_VERSION;
7172 mddev->patch_version = MD_PATCHLEVEL_VERSION;
7173 mddev->ctime = ktime_get_real_seconds();
7175 mddev->level = info->level;
7176 mddev->clevel[0] = 0;
7177 mddev->dev_sectors = 2 * (sector_t)info->size;
7178 mddev->raid_disks = info->raid_disks;
7182 if (info->state & (1<<MD_SB_CLEAN))
7183 mddev->recovery_cp = MaxSector;
7185 mddev->recovery_cp = 0;
7186 mddev->persistent = ! info->not_persistent;
7187 mddev->external = 0;
7189 mddev->layout = info->layout;
7190 if (mddev->level == 0)
7192 mddev->layout = -1;
7193 mddev->chunk_sectors = info->chunk_size >> 9;
7195 if (mddev->persistent) {
7196 mddev->max_disks = MD_SB_DISKS;
7197 mddev->flags = 0;
7198 mddev->sb_flags = 0;
7200 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
7202 mddev->bitmap_info.default_offset = MD_SB_BYTES >> 9;
7203 mddev->bitmap_info.default_space = 64*2 - (MD_SB_BYTES >> 9);
7204 mddev->bitmap_info.offset = 0;
7206 mddev->reshape_position = MaxSector;
7211 get_random_bytes(mddev->uuid, 16);
7213 mddev->new_level = mddev->level;
7214 mddev->new_chunk_sectors = mddev->chunk_sectors;
7215 mddev->new_layout = mddev->layout;
7216 mddev->delta_disks = 0;
7217 mddev->reshape_backwards = 0;
7224 lockdep_assert_held(&mddev->reconfig_mutex);
7226 if (mddev->external_size)
7229 mddev->array_sectors = array_sectors;
7238 sector_t old_dev_sectors = mddev->dev_sectors;
7240 if (mddev->pers->resize == NULL)
7241 return -EINVAL;
7251 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
7252 mddev->sync_thread)
7253 return -EBUSY;
7254 if (mddev->ro)
7255 return -EROFS;
7258 sector_t avail = rdev->sectors;
7263 return -ENOSPC;
7265 rv = mddev->pers->resize(mddev, num_sectors);
7268 md_cluster_ops->update_size(mddev, old_dev_sectors);
7269 else if (mddev->queue) {
7270 set_capacity(mddev->gendisk, mddev->array_sectors);
7271 revalidate_disk_size(mddev->gendisk, true);
7282 if (mddev->pers->check_reshape == NULL)
7283 return -EINVAL;
7284 if (mddev->ro)
7285 return -EROFS;
7287 (mddev->max_disks && raid_disks >= mddev->max_disks))
7288 return -EINVAL;
7289 if (mddev->sync_thread ||
7290 test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
7291 test_bit(MD_RESYNCING_REMOTE, &mddev->recovery) ||
7292 mddev->reshape_position != MaxSector)
7293 return -EBUSY;
7296 if (mddev->raid_disks < raid_disks &&
7297 rdev->data_offset < rdev->new_data_offset)
7298 return -EINVAL;
7299 if (mddev->raid_disks > raid_disks &&
7300 rdev->data_offset > rdev->new_data_offset)
7301 return -EINVAL;
7304 mddev->delta_disks = raid_disks - mddev->raid_disks;
7305 if (mddev->delta_disks < 0)
7306 mddev->reshape_backwards = 1;
7307 else if (mddev->delta_disks > 0)
7308 mddev->reshape_backwards = 0;
7310 rv = mddev->pers->check_reshape(mddev);
7312 mddev->delta_disks = 0;
7313 mddev->reshape_backwards = 0;
7320 * on-line array.
7333 if (mddev->bitmap && mddev->bitmap_info.offset)
7336 if (mddev->major_version != info->major_version ||
7337 mddev->minor_version != info->minor_version ||
7338 /* mddev->patch_version != info->patch_version || */
7339 mddev->ctime != info->ctime ||
7340 mddev->level != info->level ||
7341 /* mddev->layout != info->layout || */
7342 mddev->persistent != !info->not_persistent ||
7343 mddev->chunk_sectors != info->chunk_size >> 9 ||
7345 ((state^info->state) & 0xfffffe00)
7347 return -EINVAL;
7349 if (info->size >= 0 && mddev->dev_sectors / 2 != info->size)
7351 if (mddev->raid_disks != info->raid_disks)
7353 if (mddev->layout != info->layout)
7355 if ((state ^ info->state) & (1<<MD_SB_BITMAP_PRESENT))
7360 return -EINVAL;
7362 if (mddev->layout != info->layout) {
7367 if (mddev->pers->check_reshape == NULL)
7368 return -EINVAL;
7370 mddev->new_layout = info->layout;
7371 rv = mddev->pers->check_reshape(mddev);
7373 mddev->new_layout = mddev->layout;
7377 if (info->size >= 0 && mddev->dev_sectors / 2 != info->size)
7378 rv = update_size(mddev, (sector_t)info->size * 2);
7380 if (mddev->raid_disks != info->raid_disks)
7381 rv = update_raid_disks(mddev, info->raid_disks);
7383 if ((state ^ info->state) & (1<<MD_SB_BITMAP_PRESENT)) {
7384 if (mddev->pers->quiesce == NULL || mddev->thread == NULL) {
7385 rv = -EINVAL;
7388 if (mddev->recovery || mddev->sync_thread) {
7389 rv = -EBUSY;
7392 if (info->state & (1<<MD_SB_BITMAP_PRESENT)) {
7395 if (mddev->bitmap) {
7396 rv = -EEXIST;
7399 if (mddev->bitmap_info.default_offset == 0) {
7400 rv = -EINVAL;
7403 mddev->bitmap_info.offset =
7404 mddev->bitmap_info.default_offset;
7405 mddev->bitmap_info.space =
7406 mddev->bitmap_info.default_space;
7407 bitmap = md_bitmap_create(mddev, -1);
7410 mddev->bitmap = bitmap;
7419 if (!mddev->bitmap) {
7420 rv = -ENOENT;
7423 if (mddev->bitmap->storage.file) {
7424 rv = -EINVAL;
7427 if (mddev->bitmap_info.nodes) {
7429 if (md_cluster_ops->lock_all_bitmaps(mddev) <= 0) {
7431 rv = -EPERM;
7432 md_cluster_ops->unlock_all_bitmaps(mddev);
7436 mddev->bitmap_info.nodes = 0;
7437 md_cluster_ops->leave(mddev);
7439 mddev->safemode_delay = DEFAULT_SAFEMODE_DELAY;
7444 mddev->bitmap_info.offset = 0;
7458 if (mddev->pers == NULL)
7459 return -ENODEV;
7464 err = -ENODEV;
7467 if (!test_bit(Faulty, &rdev->flags))
7468 err = -EBUSY;
7478 * dosfs just mad... ;-)
7482 struct mddev *mddev = bdev->bd_disk->private_data;
7484 geo->heads = 2;
7485 geo->sectors = 4;
7486 geo->cylinders = mddev->array_sectors / 8;
7525 return -ENOTTY;
7534 return -EACCES;
7552 mddev = bdev->bd_disk->private_data;
7562 if (!mddev->raid_disks && !mddev->external)
7563 err = -ENODEV;
7569 if (!mddev->raid_disks && !mddev->external)
7570 err = -ENODEV;
7590 wait_event_interruptible_timeout(mddev->sb_wait,
7592 &mddev->recovery),
7595 /* Need to flush page cache, and ensure no-one else opens
7598 mutex_lock(&mddev->open_mutex);
7599 if (mddev->pers && atomic_read(&mddev->openers) > 1) {
7600 mutex_unlock(&mddev->open_mutex);
7601 err = -EBUSY;
7604 if (test_and_set_bit(MD_CLOSING, &mddev->flags)) {
7605 mutex_unlock(&mddev->open_mutex);
7606 err = -EBUSY;
7610 mutex_unlock(&mddev->open_mutex);
7625 err = -EFAULT;
7628 if (mddev->pers) {
7636 if (!list_empty(&mddev->disks)) {
7638 err = -EBUSY;
7641 if (mddev->raid_disks) {
7643 err = -EBUSY;
7659 if ((!mddev->raid_disks && !mddev->external)
7663 err = -ENODEV;
7668 * Commands even a read-only array can execute:
7688 /* We can support ADD_NEW_DISK on read-only arrays
7689 * only if we are re-adding a preexisting device.
7690 * So require mddev->pers and MD_DISK_SYNC.
7692 if (mddev->pers) {
7695 err = -EFAULT;
7697 /* Need to clear read-only for this */
7707 err = -EFAULT;
7710 err = -EINVAL;
7712 /* if the bdev is going readonly the value of mddev->ro
7719 if (mddev->ro != 1)
7725 if (mddev->pers) {
7728 mddev->ro = 2;
7729 set_disk_ro(mddev->gendisk, 0);
7737 * superblock, so we do not allow them on read-only arrays.
7739 if (mddev->ro && mddev->pers) {
7740 if (mddev->ro == 2) {
7741 mddev->ro = 0;
7742 sysfs_notify_dirent_safe(mddev->sysfs_state);
7743 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
7745 /* If a device failed while we were read-only, we
7748 if (test_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags)) {
7750 wait_event(mddev->sb_wait,
7751 !test_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags) &&
7752 !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags));
7756 err = -EROFS;
7766 err = -EFAULT;
7774 md_cluster_ops->new_disk_ack(mddev, false);
7776 err = -EINVAL;
7792 err = -EINVAL;
7797 if (mddev->hold_active == UNTIL_IOCTL &&
7798 err != -EINVAL)
7799 mddev->hold_active = 0;
7803 clear_bit(MD_CLOSING, &mddev->flags);
7832 struct mddev *mddev = mddev_find(bdev->bd_dev);
7836 return -ENODEV;
7838 if (mddev->gendisk != bdev->bd_disk) {
7843 /* Wait until bdev->bd_disk is definitely gone */
7844 if (work_pending(&mddev->del_work))
7846 return -EBUSY;
7848 BUG_ON(mddev != bdev->bd_disk->private_data);
7850 if ((err = mutex_lock_interruptible(&mddev->open_mutex)))
7853 if (test_bit(MD_CLOSING, &mddev->flags)) {
7854 mutex_unlock(&mddev->open_mutex);
7855 err = -ENODEV;
7860 atomic_inc(&mddev->openers);
7861 mutex_unlock(&mddev->open_mutex);
7872 struct mddev *mddev = disk->private_data;
7875 atomic_dec(&mddev->openers);
7881 struct mddev *mddev = disk->private_data;
7884 if (mddev->changed)
7886 mddev->changed = 0;
7909 * md_thread is a 'system-thread', it's priority should be very
7924 * we don't add to the load-average.
7932 (thread->wqueue,
7933 test_bit(THREAD_WAKEUP, &thread->flags)
7935 thread->timeout);
7937 clear_bit(THREAD_WAKEUP, &thread->flags);
7941 thread->run(thread);
7950 pr_debug("md: waking up MD thread %s.\n", thread->tsk->comm);
7951 set_bit(THREAD_WAKEUP, &thread->flags);
7952 wake_up(&thread->wqueue);
7966 init_waitqueue_head(&thread->wqueue);
7968 thread->run = run;
7969 thread->mddev = mddev;
7970 thread->timeout = MAX_SCHEDULE_TIMEOUT;
7971 thread->tsk = kthread_run(md_thread, thread,
7973 mdname(thread->mddev),
7975 if (IS_ERR(thread->tsk)) {
7989 * non-existent thread
8000 pr_debug("interrupting MD-thread pid %d\n", task_pid_nr(thread->tsk));
8001 kthread_stop(thread->tsk);
8008 if (!rdev || test_bit(Faulty, &rdev->flags))
8011 if (!mddev->pers || !mddev->pers->error_handler)
8013 mddev->pers->error_handler(mddev,rdev);
8014 if (mddev->degraded)
8015 set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
8016 sysfs_notify_dirent_safe(rdev->sysfs_state);
8017 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
8018 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
8019 md_wakeup_thread(mddev->thread);
8020 if (mddev->event_work.func)
8021 queue_work(md_misc_wq, &mddev->event_work);
8039 bdevname(rdev->bdev,b));
8055 if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ||
8056 test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
8057 max_sectors = mddev->resync_max_sectors;
8059 max_sectors = mddev->dev_sectors;
8061 resync = mddev->curr_resync;
8063 if (test_bit(MD_RECOVERY_DONE, &mddev->recovery))
8069 resync -= atomic_read(&mddev->recovery_active);
8072 if (test_bit(MD_RESYNCING_REMOTE, &mddev->recovery)) {
8076 if (rdev->raid_disk >= 0 &&
8077 !test_bit(Faulty, &rdev->flags) &&
8078 rdev->recovery_offset != MaxSector &&
8079 rdev->recovery_offset) {
8083 if (mddev->reshape_position != MaxSector)
8089 if (mddev->recovery_cp < MaxSector) {
8116 int i, x = per_milli/50, y = 20-x;
8126 (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)?
8128 (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)?
8130 (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ?
8153 dt = ((jiffies - mddev->resync_mark) / HZ);
8156 curr_mark_cnt = mddev->curr_mark_cnt;
8157 recovery_active = atomic_read(&mddev->recovery_active);
8158 resync_mark_cnt = mddev->resync_mark_cnt;
8161 db = curr_mark_cnt - (recovery_active + resync_mark_cnt);
8163 rt = max_sectors - resync; /* number of remaining sectors */
8187 if (!l--)
8193 if (!l--) {
8200 if (!l--)
8218 tmp = mddev->all_mddevs.next;
8252 seq_printf(seq, "[%s] ", pers->name);
8256 seq->poll_event = atomic_read(&md_event_count);
8264 spin_lock(&mddev->lock);
8265 if (mddev->pers || mddev->raid_disks || !list_empty(&mddev->disks)) {
8267 mddev->pers ? "" : "in");
8268 if (mddev->pers) {
8269 if (mddev->ro==1)
8270 seq_printf(seq, " (read-only)");
8271 if (mddev->ro==2)
8272 seq_printf(seq, " (auto-read-only)");
8273 seq_printf(seq, " %s", mddev->pers->name);
8281 bdevname(rdev->bdev,b), rdev->desc_nr);
8282 if (test_bit(WriteMostly, &rdev->flags))
8284 if (test_bit(Journal, &rdev->flags))
8286 if (test_bit(Faulty, &rdev->flags)) {
8290 if (rdev->raid_disk < 0)
8292 if (test_bit(Replacement, &rdev->flags))
8294 sectors += rdev->sectors;
8298 if (!list_empty(&mddev->disks)) {
8299 if (mddev->pers)
8302 mddev->array_sectors / 2);
8307 if (mddev->persistent) {
8308 if (mddev->major_version != 0 ||
8309 mddev->minor_version != 90) {
8311 mddev->major_version,
8312 mddev->minor_version);
8314 } else if (mddev->external)
8316 mddev->metadata_type);
8318 seq_printf(seq, " super non-persistent");
8320 if (mddev->pers) {
8321 mddev->pers->status(seq, mddev);
8323 if (mddev->pers->sync_request) {
8330 md_bitmap_status(seq, mddev->bitmap);
8334 spin_unlock(&mddev->lock);
8355 seq = file->private_data;
8356 seq->poll_event = atomic_read(&md_event_count);
8363 struct seq_file *seq = filp->private_data;
8373 if (seq->poll_event != atomic_read(&md_event_count))
8389 p->name, p->level);
8391 list_add_tail(&p->list, &pers_list);
8399 pr_debug("md: %s personality unregistered\n", p->name);
8401 list_del_init(&p->list);
8413 ret = -EALREADY;
8436 request_module("md-cluster");
8440 pr_warn("can't find md-cluster module or get it's reference.\n");
8442 return -ENOENT;
8446 ret = md_cluster_ops->join(mddev, nodes);
8448 mddev->safemode_delay = 0;
8456 md_cluster_ops->leave(mddev);
8469 struct gendisk *disk = rdev->bdev->bd_disk;
8470 curr_events = (int)part_stat_read_accum(&disk->part0, sectors) -
8471 atomic_read(&disk->sync_io);
8477 * non-sync IO will cause disk_stat to increase without
8481 * the array to appear non-idle, and resync will slow
8485 * completing might cause the array to appear non-idle
8487 * not have been non-resync activity. This will only
8494 if (init || curr_events - rdev->last_events > 64) {
8495 rdev->last_events = curr_events;
8506 atomic_sub(blocks, &mddev->recovery_active);
8507 wake_up(&mddev->recovery_wait);
8509 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
8510 set_bit(MD_RECOVERY_ERROR, &mddev->recovery);
8511 md_wakeup_thread(mddev->thread);
8531 BUG_ON(mddev->ro == 1);
8532 if (mddev->ro == 2) {
8534 mddev->ro = 0;
8535 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
8536 md_wakeup_thread(mddev->thread);
8537 md_wakeup_thread(mddev->sync_thread);
8541 percpu_ref_get(&mddev->writes_pending);
8543 if (mddev->safemode == 1)
8544 mddev->safemode = 0;
8545 /* sync_checkers is always 0 when writes_pending is in per-cpu mode */
8546 if (mddev->in_sync || mddev->sync_checkers) {
8547 spin_lock(&mddev->lock);
8548 if (mddev->in_sync) {
8549 mddev->in_sync = 0;
8550 set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
8551 set_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
8552 md_wakeup_thread(mddev->thread);
8555 spin_unlock(&mddev->lock);
8559 sysfs_notify_dirent_safe(mddev->sysfs_state);
8560 if (!mddev->has_superblocks)
8562 wait_event(mddev->sb_wait,
8563 !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags) ||
8564 mddev->suspended);
8565 if (test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)) {
8566 percpu_ref_put(&mddev->writes_pending);
8585 WARN_ON_ONCE(mddev->in_sync || mddev->ro);
8586 percpu_ref_get(&mddev->writes_pending);
8592 percpu_ref_put(&mddev->writes_pending);
8594 if (mddev->safemode == 2)
8595 md_wakeup_thread(mddev->thread);
8596 else if (mddev->safemode_delay)
8598 * every ->safemode_delay jiffies
8600 mod_timer(&mddev->safemode_timer,
8601 roundup(jiffies, mddev->safemode_delay) +
8602 mddev->safemode_delay);
8615 if (!mddev->pers)
8617 if (mddev->ro)
8619 if (!mddev->pers->sync_request)
8622 spin_lock(&mddev->lock);
8623 if (mddev->in_sync) {
8624 mddev->in_sync = 0;
8625 set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
8626 set_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
8627 if (mddev->safemode_delay &&
8628 mddev->safemode == 0)
8629 mddev->safemode = 1;
8630 spin_unlock(&mddev->lock);
8632 sysfs_notify_dirent_safe(mddev->sysfs_state);
8634 wait_event(mddev->sb_wait,
8635 !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags));
8637 spin_unlock(&mddev->lock);
8646 struct mddev *mddev = thread->mddev;
8663 if (test_bit(MD_RECOVERY_DONE, &mddev->recovery) ||
8664 test_bit(MD_RECOVERY_WAIT, &mddev->recovery))
8666 if (mddev->ro) {/* never try to sync a read-only array */
8667 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
8672 ret = md_cluster_ops->resync_start(mddev);
8676 set_bit(MD_CLUSTER_RESYNC_LOCKED, &mddev->flags);
8677 if (!(test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ||
8678 test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) ||
8679 test_bit(MD_RECOVERY_RECOVER, &mddev->recovery))
8680 && ((unsigned long long)mddev->curr_resync_completed
8681 < (unsigned long long)mddev->resync_max_sectors))
8685 if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
8686 if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)) {
8687 desc = "data-check";
8689 } else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
8690 desc = "requested-resync";
8694 } else if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
8699 mddev->last_sync_action = action ?: desc;
8706 * other == active in resync - this many blocks
8718 int mddev2_minor = -1;
8719 mddev->curr_resync = 2;
8722 if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
8727 if (!mddev->parallel_resync
8728 && mddev2->curr_resync
8731 if (mddev < mddev2 && mddev->curr_resync == 2) {
8733 mddev->curr_resync = 1;
8736 if (mddev > mddev2 && mddev->curr_resync == 1)
8746 if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
8747 mddev2->curr_resync >= mddev->curr_resync) {
8748 if (mddev2_minor != mddev2->md_minor) {
8749 mddev2_minor = mddev2->md_minor;
8764 } while (mddev->curr_resync < 2);
8767 if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
8771 max_sectors = mddev->resync_max_sectors;
8772 atomic64_set(&mddev->resync_mismatches, 0);
8774 if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
8775 j = mddev->resync_min;
8776 else if (!mddev->bitmap)
8777 j = mddev->recovery_cp;
8779 } else if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) {
8780 max_sectors = mddev->resync_max_sectors;
8787 mddev->reshape_position != MaxSector)
8788 j = mddev->reshape_position;
8791 max_sectors = mddev->dev_sectors;
8795 if (rdev->raid_disk >= 0 &&
8796 !test_bit(Journal, &rdev->flags) &&
8797 !test_bit(Faulty, &rdev->flags) &&
8798 !test_bit(In_sync, &rdev->flags) &&
8799 rdev->recovery_offset < j)
8800 j = rdev->recovery_offset;
8811 if (mddev->bitmap) {
8812 mddev->pers->quiesce(mddev, 1);
8813 mddev->pers->quiesce(mddev, 0);
8830 mddev->resync_mark = mark[last_mark];
8831 mddev->resync_mark_cnt = mark_cnt[last_mark];
8840 atomic_set(&mddev->recovery_active, 0);
8846 mddev->curr_resync = j;
8848 mddev->curr_resync = 3; /* no longer delayed */
8849 mddev->curr_resync_completed = j;
8850 sysfs_notify_dirent_safe(mddev->sysfs_completed);
8860 if (!test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
8861 ((mddev->curr_resync > mddev->curr_resync_completed &&
8862 (mddev->curr_resync - mddev->curr_resync_completed)
8865 (j - mddev->curr_resync_completed)*2
8866 >= mddev->resync_max - mddev->curr_resync_completed ||
8867 mddev->curr_resync_completed > mddev->resync_max
8870 wait_event(mddev->recovery_wait,
8871 atomic_read(&mddev->recovery_active) == 0);
8872 mddev->curr_resync_completed = j;
8873 if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) &&
8874 j > mddev->recovery_cp)
8875 mddev->recovery_cp = j;
8877 set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
8878 sysfs_notify_dirent_safe(mddev->sysfs_completed);
8881 while (j >= mddev->resync_max &&
8882 !test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
8883 /* As this condition is controlled by user-space,
8888 wait_event_interruptible(mddev->recovery_wait,
8889 mddev->resync_max > j
8891 &mddev->recovery));
8894 if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
8897 sectors = mddev->pers->sync_request(mddev, j, &skipped);
8899 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
8905 atomic_add(sectors, &mddev->recovery_active);
8908 if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
8916 mddev->curr_resync = j;
8917 mddev->curr_mark_cnt = io_sectors;
8933 mddev->resync_mark = mark[next];
8934 mddev->resync_mark_cnt = mark_cnt[next];
8936 mark_cnt[next] = io_sectors - atomic_read(&mddev->recovery_active);
8940 if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
8945 * the 'hard' speed limit, or the system was IO-idle for
8947 * the system might be non-idle CPU-wise, but we only care
8953 recovery_done = io_sectors - atomic_read(&mddev->recovery_active);
8954 currspeed = ((unsigned long)(recovery_done - mddev->resync_mark_cnt))/2
8955 /((jiffies-mddev->resync_mark)/HZ +1) +1;
8967 wait_event(mddev->recovery_wait,
8968 !atomic_read(&mddev->recovery_active));
8973 test_bit(MD_RECOVERY_INTR, &mddev->recovery)
8979 wait_event(mddev->recovery_wait, !atomic_read(&mddev->recovery_active));
8981 if (!test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
8982 !test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
8983 mddev->curr_resync > 3) {
8984 mddev->curr_resync_completed = mddev->curr_resync;
8985 sysfs_notify_dirent_safe(mddev->sysfs_completed);
8987 mddev->pers->sync_request(mddev, max_sectors, &skipped);
8989 if (!test_bit(MD_RECOVERY_CHECK, &mddev->recovery) &&
8990 mddev->curr_resync > 3) {
8991 if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
8992 if (test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
8993 if (mddev->curr_resync >= mddev->recovery_cp) {
8997 &mddev->recovery))
8998 mddev->recovery_cp =
8999 mddev->curr_resync_completed;
9001 mddev->recovery_cp =
9002 mddev->curr_resync;
9005 mddev->recovery_cp = MaxSector;
9007 if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery))
9008 mddev->curr_resync = MaxSector;
9009 if (!test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
9010 test_bit(MD_RECOVERY_RECOVER, &mddev->recovery)) {
9013 if (rdev->raid_disk >= 0 &&
9014 mddev->delta_disks >= 0 &&
9015 !test_bit(Journal, &rdev->flags) &&
9016 !test_bit(Faulty, &rdev->flags) &&
9017 !test_bit(In_sync, &rdev->flags) &&
9018 rdev->recovery_offset < mddev->curr_resync)
9019 rdev->recovery_offset = mddev->curr_resync;
9028 set_mask_bits(&mddev->sb_flags, 0,
9031 if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
9032 !test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
9033 mddev->delta_disks > 0 &&
9034 mddev->pers->finish_reshape &&
9035 mddev->pers->size &&
9036 mddev->queue) {
9038 md_set_array_sectors(mddev, mddev->pers->size(mddev, 0, 0));
9041 set_capacity(mddev->gendisk, mddev->array_sectors);
9042 revalidate_disk_size(mddev->gendisk, true);
9046 spin_lock(&mddev->lock);
9047 if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
9049 if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
9050 mddev->resync_min = 0;
9051 mddev->resync_max = MaxSector;
9052 } else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
9053 mddev->resync_min = mddev->curr_resync_completed;
9054 set_bit(MD_RECOVERY_DONE, &mddev->recovery);
9055 mddev->curr_resync = 0;
9056 spin_unlock(&mddev->lock);
9059 md_wakeup_thread(mddev->thread);
9072 if (this && test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
9078 rdev->raid_disk >= 0 &&
9079 !test_bit(Blocked, &rdev->flags) &&
9080 test_bit(Faulty, &rdev->flags) &&
9081 atomic_read(&rdev->nr_pending)==0) {
9082 /* Faulty non-Blocked devices with nr_pending == 0
9088 set_bit(RemoveSynchronized, &rdev->flags);
9096 rdev->raid_disk >= 0 &&
9097 !test_bit(Blocked, &rdev->flags) &&
9098 ((test_bit(RemoveSynchronized, &rdev->flags) ||
9099 (!test_bit(In_sync, &rdev->flags) &&
9100 !test_bit(Journal, &rdev->flags))) &&
9101 atomic_read(&rdev->nr_pending)==0)) {
9102 if (mddev->pers->hot_remove_disk(
9105 rdev->saved_raid_disk = rdev->raid_disk;
9106 rdev->raid_disk = -1;
9110 if (remove_some && test_bit(RemoveSynchronized, &rdev->flags))
9111 clear_bit(RemoveSynchronized, &rdev->flags);
9114 if (removed && mddev->kobj.sd)
9115 sysfs_notify_dirent_safe(mddev->sysfs_degraded);
9123 if (test_bit(Candidate, &rdev->flags))
9125 if (rdev->raid_disk >= 0 &&
9126 !test_bit(In_sync, &rdev->flags) &&
9127 !test_bit(Journal, &rdev->flags) &&
9128 !test_bit(Faulty, &rdev->flags))
9130 if (rdev->raid_disk >= 0)
9132 if (test_bit(Faulty, &rdev->flags))
9134 if (!test_bit(Journal, &rdev->flags)) {
9135 if (mddev->ro &&
9136 ! (rdev->saved_raid_disk >= 0 &&
9137 !test_bit(Bitmap_sync, &rdev->flags)))
9140 rdev->recovery_offset = 0;
9142 if (mddev->pers->hot_add_disk(mddev, rdev) == 0) {
9145 if (!test_bit(Journal, &rdev->flags))
9148 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
9153 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
9161 mddev->sync_thread = md_register_thread(md_do_sync,
9164 if (!mddev->sync_thread) {
9168 clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
9169 clear_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
9170 clear_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
9171 clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
9172 clear_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
9175 &mddev->recovery))
9176 if (mddev->sysfs_action)
9177 sysfs_notify_dirent_safe(mddev->sysfs_action);
9179 md_wakeup_thread(mddev->sync_thread);
9180 sysfs_notify_dirent_safe(mddev->sysfs_action);
9185 * This routine is regularly called by all per-raid-array threads to
9186 * deal with generic issues like resync and super-block update.
9193 * "->recovery" and create a thread at ->sync_thread.
9204 * 6/ If array has spares or is not in-sync, start a resync thread.
9208 if (test_bit(MD_ALLOW_SB_UPDATE, &mddev->flags) && mddev->sb_flags) {
9209 /* Write superblock - thread that called mddev_suspend()
9212 set_bit(MD_UPDATING_SB, &mddev->flags);
9214 if (test_bit(MD_ALLOW_SB_UPDATE, &mddev->flags))
9216 clear_bit_unlock(MD_UPDATING_SB, &mddev->flags);
9217 wake_up(&mddev->sb_wait);
9220 if (mddev->suspended)
9223 if (mddev->bitmap)
9227 if (mddev->pers->sync_request && !mddev->external) {
9230 mddev->safemode = 2;
9235 if (mddev->ro && !test_bit(MD_RECOVERY_NEEDED, &mddev->recovery))
9238 (mddev->sb_flags & ~ (1<<MD_SB_CHANGE_PENDING)) ||
9239 test_bit(MD_RECOVERY_NEEDED, &mddev->recovery) ||
9240 test_bit(MD_RECOVERY_DONE, &mddev->recovery) ||
9241 (mddev->external == 0 && mddev->safemode == 1) ||
9242 (mddev->safemode == 2
9243 && !mddev->in_sync && mddev->recovery_cp == MaxSector)
9249 bool try_set_sync = mddev->safemode != 0;
9251 if (!mddev->external && mddev->safemode == 1)
9252 mddev->safemode = 0;
9254 if (mddev->ro) {
9256 if (!mddev->external && mddev->in_sync)
9263 clear_bit(Blocked, &rdev->flags);
9264 /* On a read-only array we can:
9265 * - remove failed devices
9266 * - add already-in_sync devices if the array itself
9267 * is in-sync.
9268 * As we only add devices that are already in-sync,
9273 * ->spare_active and clear saved_raid_disk
9275 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
9277 clear_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
9278 clear_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
9279 clear_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
9289 if (test_and_clear_bit(ClusterRemove, &rdev->flags) &&
9290 rdev->raid_disk < 0)
9295 if (try_set_sync && !mddev->external && !mddev->in_sync) {
9296 spin_lock(&mddev->lock);
9298 spin_unlock(&mddev->lock);
9301 if (mddev->sb_flags)
9304 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) &&
9305 !test_bit(MD_RECOVERY_DONE, &mddev->recovery)) {
9307 clear_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
9310 if (mddev->sync_thread) {
9317 mddev->curr_resync_completed = 0;
9318 spin_lock(&mddev->lock);
9319 set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
9320 spin_unlock(&mddev->lock);
9324 clear_bit(MD_RECOVERY_INTR, &mddev->recovery);
9325 clear_bit(MD_RECOVERY_DONE, &mddev->recovery);
9327 if (!test_and_clear_bit(MD_RECOVERY_NEEDED, &mddev->recovery) ||
9328 test_bit(MD_RECOVERY_FROZEN, &mddev->recovery))
9333 * Spares are also removed and re-added, to allow
9334 * the personality to fail the re-add.
9337 if (mddev->reshape_position != MaxSector) {
9338 if (mddev->pers->check_reshape == NULL ||
9339 mddev->pers->check_reshape(mddev) != 0)
9342 set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
9343 clear_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
9345 clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
9346 clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
9347 clear_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
9348 set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
9349 } else if (mddev->recovery_cp < MaxSector) {
9350 set_bit(MD_RECOVERY_SYNC, &mddev->recovery);
9351 clear_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
9352 } else if (!test_bit(MD_RECOVERY_SYNC, &mddev->recovery))
9356 if (mddev->pers->sync_request) {
9362 md_bitmap_write_all(mddev->bitmap);
9364 INIT_WORK(&mddev->del_work, md_start_sync);
9365 queue_work(md_misc_wq, &mddev->del_work);
9369 if (!mddev->sync_thread) {
9370 clear_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
9373 &mddev->recovery))
9374 if (mddev->sysfs_action)
9375 sysfs_notify_dirent_safe(mddev->sysfs_action);
9378 wake_up(&mddev->sb_wait);
9387 sector_t old_dev_sectors = mddev->dev_sectors;
9391 md_unregister_thread(&mddev->sync_thread);
9392 if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
9393 !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
9394 mddev->degraded != mddev->raid_disks) {
9397 if (mddev->pers->spare_active(mddev)) {
9398 sysfs_notify_dirent_safe(mddev->sysfs_degraded);
9399 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
9402 if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
9403 mddev->pers->finish_reshape) {
9404 mddev->pers->finish_reshape(mddev);
9409 /* If array is no-longer degraded, then any saved_raid_disk
9412 if (!mddev->degraded)
9414 rdev->saved_raid_disk = -1;
9420 if (test_and_clear_bit(MD_CLUSTER_RESYNC_LOCKED, &mddev->flags))
9421 md_cluster_ops->resync_finish(mddev);
9422 clear_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
9423 clear_bit(MD_RECOVERY_DONE, &mddev->recovery);
9424 clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
9425 clear_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
9426 clear_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
9427 clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
9429 * We call md_cluster_ops->update_size here because sync_size could
9434 && !test_bit(MD_CLOSING, &mddev->flags))
9435 md_cluster_ops->update_size(mddev, old_dev_sectors);
9438 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
9439 sysfs_notify_dirent_safe(mddev->sysfs_completed);
9440 sysfs_notify_dirent_safe(mddev->sysfs_action);
9442 if (mddev->event_work.func)
9443 queue_work(md_misc_wq, &mddev->event_work);
9449 sysfs_notify_dirent_safe(rdev->sysfs_state);
9450 wait_event_timeout(rdev->blocked_wait,
9451 !test_bit(Blocked, &rdev->flags) &&
9452 !test_bit(BlockedBadBlocks, &rdev->flags),
9464 if (rdev->data_offset > rdev->new_data_offset)
9465 rdev->sectors += rdev->data_offset - rdev->new_data_offset;
9467 rdev->sectors -= rdev->new_data_offset - rdev->data_offset;
9468 rdev->data_offset = rdev->new_data_offset;
9479 struct mddev *mddev = rdev->mddev;
9482 s += rdev->new_data_offset;
9484 s += rdev->data_offset;
9485 rv = badblocks_set(&rdev->badblocks, s, sectors, 0);
9488 if (test_bit(ExternalBbl, &rdev->flags))
9489 sysfs_notify_dirent_safe(rdev->sysfs_unack_badblocks);
9490 sysfs_notify_dirent_safe(rdev->sysfs_state);
9491 set_mask_bits(&mddev->sb_flags, 0,
9493 md_wakeup_thread(rdev->mddev->thread);
9505 s += rdev->new_data_offset;
9507 s += rdev->data_offset;
9508 rv = badblocks_clear(&rdev->badblocks, s, sectors);
9509 if ((rv == 0) && test_bit(ExternalBbl, &rdev->flags))
9510 sysfs_notify_dirent_safe(rdev->sysfs_badblocks);
9524 if (mddev->pers)
9526 if (mddev->persistent)
9527 mddev->safemode = 2;
9559 int ret = -ENOMEM;
9605 struct mdp_superblock_1 *sb = page_address(rdev->sb_page);
9614 if (mddev->dev_sectors != le64_to_cpu(sb->size)) {
9615 ret = mddev->pers->resize(mddev, le64_to_cpu(sb->size));
9617 pr_info("md-cluster: resize failed\n");
9619 md_bitmap_update_sb(mddev->bitmap);
9624 if (test_bit(Faulty, &rdev2->flags))
9628 role = le16_to_cpu(sb->dev_roles[rdev2->desc_nr]);
9630 if (test_bit(Candidate, &rdev2->flags)) {
9632 pr_info("md: Removing Candidate device %s because add failed\n", bdevname(rdev2->bdev,b));
9637 clear_bit(Candidate, &rdev2->flags);
9640 if (role != rdev2->raid_disk) {
9644 if (rdev2->raid_disk == -1 && role != 0xffff &&
9645 !(le32_to_cpu(sb->feature_map) &
9647 rdev2->saved_raid_disk = role;
9650 bdevname(rdev2->bdev,b));
9651 /* wakeup mddev->thread here, so array could
9653 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
9654 md_wakeup_thread(mddev->thread);
9663 clear_bit(Blocked, &rdev2->flags);
9668 if (mddev->raid_disks != le32_to_cpu(sb->raid_disks)) {
9669 ret = update_raid_disks(mddev, le32_to_cpu(sb->raid_disks));
9675 * Since mddev->delta_disks has already updated in update_raid_disks,
9678 if (test_bit(MD_RESYNCING_REMOTE, &mddev->recovery) &&
9679 (le32_to_cpu(sb->feature_map) & MD_FEATURE_RESHAPE_ACTIVE)) {
9684 mddev->reshape_position = le64_to_cpu(sb->reshape_position);
9685 if (mddev->pers->update_reshape_pos)
9686 mddev->pers->update_reshape_pos(mddev);
9687 if (mddev->pers->start_reshape)
9688 mddev->pers->start_reshape(mddev);
9689 } else if (test_bit(MD_RESYNCING_REMOTE, &mddev->recovery) &&
9690 mddev->reshape_position != MaxSector &&
9691 !(le32_to_cpu(sb->feature_map) & MD_FEATURE_RESHAPE_ACTIVE)) {
9693 mddev->reshape_position = MaxSector;
9694 if (mddev->pers->update_reshape_pos)
9695 mddev->pers->update_reshape_pos(mddev);
9699 mddev->events = le64_to_cpu(sb->events);
9705 struct page *swapout = rdev->sb_page;
9711 rdev->sb_page = NULL;
9714 ClearPageUptodate(rdev->sb_page);
9715 rdev->sb_loaded = 0;
9716 err = super_types[mddev->major_version].
9717 load_super(rdev, NULL, mddev->minor_version);
9721 __func__, __LINE__, rdev->desc_nr, err);
9722 if (rdev->sb_page)
9723 put_page(rdev->sb_page);
9724 rdev->sb_page = swapout;
9725 rdev->sb_loaded = 1;
9729 sb = page_address(rdev->sb_page);
9734 if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_RECOVERY_OFFSET))
9735 rdev->recovery_offset = le64_to_cpu(sb->recovery_offset);
9738 * device In_sync and mddev->degraded
9740 if (rdev->recovery_offset == MaxSector &&
9741 !test_bit(In_sync, &rdev->flags) &&
9742 mddev->pers->spare_active(mddev))
9743 sysfs_notify_dirent_safe(mddev->sysfs_degraded);
9756 if (iter->desc_nr == nr) {
9775 if (!test_bit(Faulty, &rdev->flags))
9801 node_detected_dev->dev = dev;
9803 list_add_tail(&node_detected_dev->list, &all_detected_devices);
9825 list_del(&node_detected_dev->list);
9826 dev = node_detected_dev->dev;
9834 if (test_bit(Faulty, &rdev->flags))
9837 set_bit(AutoDetected, &rdev->flags);
9838 list_add(&rdev->same_set, &pending_raid_disks);
9865 * waiting for us in select() or poll() - wake them up
9878 mddev->ctime = 0;
9879 mddev->hold_active = 0;