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.
44 #include <linux/blk-integrity.h>
65 #include <linux/percpu-refcount.h>
70 #include "md-bitmap.h"
71 #include "md-cluster.h"
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;
315 * a device node in /dev and to open it. This causes races with device-close.
352 * We hold a refcount over the call to ->make_request. By the time that
354 * and so is visible to ->quiesce(), so we don't need the refcount any more.
362 if (mddev->suspend_lo >= mddev->suspend_hi)
364 if (bio->bi_iter.bi_sector >= mddev->suspend_hi)
366 if (bio_end_sector(bio) < mddev->suspend_lo)
377 if (bio->bi_opf & REQ_NOWAIT) {
382 prepare_to_wait(&mddev->sb_wait, &__wait,
388 finish_wait(&mddev->sb_wait, &__wait);
390 if (!percpu_ref_tryget_live(&mddev->active_io))
393 if (!mddev->pers->make_request(mddev, bio)) {
394 percpu_ref_put(&mddev->active_io);
398 percpu_ref_put(&mddev->active_io);
405 struct mddev *mddev = bio->bi_bdev->bd_disk->private_data;
407 if (mddev == NULL || mddev->pers == NULL) {
412 if (unlikely(test_bit(MD_BROKEN, &mddev->flags)) && (rw == WRITE)) {
421 if (mddev->ro == MD_RDONLY && unlikely(rw == WRITE)) {
423 bio->bi_status = BLK_STS_IOERR;
429 bio->bi_opf &= ~REQ_NOMERGE;
442 struct md_thread *thread = rcu_dereference_protected(mddev->thread,
443 lockdep_is_held(&mddev->reconfig_mutex));
445 WARN_ON_ONCE(thread && current == thread->tsk);
446 if (mddev->suspended++)
448 wake_up(&mddev->sb_wait);
449 set_bit(MD_ALLOW_SB_UPDATE, &mddev->flags);
450 percpu_ref_kill(&mddev->active_io);
452 if (mddev->pers && mddev->pers->prepare_suspend)
453 mddev->pers->prepare_suspend(mddev);
455 wait_event(mddev->sb_wait, percpu_ref_is_zero(&mddev->active_io));
456 clear_bit_unlock(MD_ALLOW_SB_UPDATE, &mddev->flags);
457 wait_event(mddev->sb_wait, !test_bit(MD_UPDATING_SB, &mddev->flags));
459 del_timer_sync(&mddev->safemode_timer);
461 mddev->noio_flag = memalloc_noio_save();
467 lockdep_assert_held(&mddev->reconfig_mutex);
468 if (--mddev->suspended)
472 memalloc_noio_restore(mddev->noio_flag);
474 percpu_ref_resurrect(&mddev->active_io);
475 wake_up(&mddev->sb_wait);
477 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
478 md_wakeup_thread(mddev->thread);
479 md_wakeup_thread(mddev->sync_thread); /* possibly kick off a reshape */
489 struct md_rdev *rdev = bio->bi_private;
490 struct mddev *mddev = rdev->mddev;
496 if (atomic_dec_and_test(&mddev->flush_pending)) {
498 percpu_ref_put(&mddev->active_io);
500 /* The pre-request flush has finished */
501 queue_work(md_wq, &mddev->flush_work);
512 mddev->start_flush = ktime_get_boottime();
513 INIT_WORK(&mddev->flush_work, md_submit_flush_data);
514 atomic_set(&mddev->flush_pending, 1);
517 if (rdev->raid_disk >= 0 &&
518 !test_bit(Faulty, &rdev->flags)) {
521 atomic_inc(&rdev->nr_pending);
523 bi = bio_alloc_bioset(rdev->bdev, 0,
525 GFP_NOIO, &mddev->bio_set);
526 bi->bi_end_io = md_end_flush;
527 bi->bi_private = rdev;
528 atomic_inc(&mddev->flush_pending);
533 if (atomic_dec_and_test(&mddev->flush_pending)) {
535 percpu_ref_put(&mddev->active_io);
537 queue_work(md_wq, &mddev->flush_work);
544 struct bio *bio = mddev->flush_bio;
552 spin_lock_irq(&mddev->lock);
553 mddev->prev_flush_start = mddev->start_flush;
554 mddev->flush_bio = NULL;
555 spin_unlock_irq(&mddev->lock);
556 wake_up(&mddev->sb_wait);
558 if (bio->bi_iter.bi_size == 0) {
559 /* an empty barrier - all done */
562 bio->bi_opf &= ~REQ_PREFLUSH;
576 spin_lock_irq(&mddev->lock);
580 wait_event_lock_irq(mddev->sb_wait,
581 !mddev->flush_bio ||
582 ktime_before(req_start, mddev->prev_flush_start),
583 mddev->lock);
585 if (ktime_after(req_start, mddev->prev_flush_start)) {
586 WARN_ON(mddev->flush_bio);
597 WARN_ON(percpu_ref_is_zero(&mddev->active_io));
598 percpu_ref_get(&mddev->active_io);
599 mddev->flush_bio = bio;
602 spin_unlock_irq(&mddev->lock);
605 INIT_WORK(&mddev->flush_work, submit_flushes);
606 queue_work(md_wq, &mddev->flush_work);
609 if (bio->bi_iter.bi_size == 0)
610 /* an empty barrier - all done */
613 bio->bi_opf &= ~REQ_PREFLUSH;
625 if (test_bit(MD_DELETED, &mddev->flags))
627 atomic_inc(&mddev->active);
635 if (!atomic_dec_and_lock(&mddev->active, &all_mddevs_lock))
637 if (!mddev->raid_disks && list_empty(&mddev->disks) &&
638 mddev->ctime == 0 && !mddev->hold_active) {
641 set_bit(MD_DELETED, &mddev->flags);
648 INIT_WORK(&mddev->del_work, mddev_delayed_delete);
649 queue_work(md_misc_wq, &mddev->del_work);
658 mutex_init(&mddev->open_mutex);
659 mutex_init(&mddev->reconfig_mutex);
660 mutex_init(&mddev->sync_mutex);
661 mutex_init(&mddev->bitmap_info.mutex);
662 INIT_LIST_HEAD(&mddev->disks);
663 INIT_LIST_HEAD(&mddev->all_mddevs);
664 INIT_LIST_HEAD(&mddev->deleting);
665 timer_setup(&mddev->safemode_timer, md_safemode_timeout, 0);
666 atomic_set(&mddev->active, 1);
667 atomic_set(&mddev->openers, 0);
668 atomic_set(&mddev->sync_seq, 0);
669 spin_lock_init(&mddev->lock);
670 atomic_set(&mddev->flush_pending, 0);
671 init_waitqueue_head(&mddev->sb_wait);
672 init_waitqueue_head(&mddev->recovery_wait);
673 mddev->reshape_position = MaxSector;
674 mddev->reshape_backwards = 0;
675 mddev->last_sync_action = "none";
676 mddev->resync_min = 0;
677 mddev->resync_max = MaxSector;
678 mddev->level = LEVEL_NONE;
687 if (mddev->unit == unit)
720 unit &= ~((1 << MdpMinorShift) - 1);
724 return ERR_PTR(-ENOMEM);
729 error = -EEXIST;
732 new->unit = unit;
734 new->md_minor = MINOR(unit);
736 new->md_minor = MINOR(unit) >> MdpMinorShift;
737 new->hold_active = UNTIL_IOCTL;
739 error = -ENODEV;
740 new->unit = mddev_alloc_unit();
741 if (!new->unit)
743 new->md_minor = MINOR(new->unit);
744 new->hold_active = UNTIL_STOP;
747 list_add(&new->all_mddevs, &all_mddevs);
759 list_del(&mddev->all_mddevs);
773 if (!list_empty(&mddev->deleting))
774 list_splice_init(&mddev->deleting, &delete);
776 if (mddev->to_remove) {
782 * and anything else which might set ->to_remove or my
784 * -EBUSY if sysfs_active is still set.
789 const struct attribute_group *to_remove = mddev->to_remove;
790 mddev->to_remove = NULL;
791 mddev->sysfs_active = 1;
792 mutex_unlock(&mddev->reconfig_mutex);
794 if (mddev->kobj.sd) {
796 sysfs_remove_group(&mddev->kobj, to_remove);
797 if (mddev->pers == NULL ||
798 mddev->pers->sync_request == NULL) {
799 sysfs_remove_group(&mddev->kobj, &md_redundancy_group);
800 if (mddev->sysfs_action)
801 sysfs_put(mddev->sysfs_action);
802 if (mddev->sysfs_completed)
803 sysfs_put(mddev->sysfs_completed);
804 if (mddev->sysfs_degraded)
805 sysfs_put(mddev->sysfs_degraded);
806 mddev->sysfs_action = NULL;
807 mddev->sysfs_completed = NULL;
808 mddev->sysfs_degraded = NULL;
811 mddev->sysfs_active = 0;
813 mutex_unlock(&mddev->reconfig_mutex);
815 md_wakeup_thread(mddev->thread);
816 wake_up(&mddev->sb_wait);
819 list_del_init(&rdev->same_set);
820 kobject_del(&rdev->kobj);
831 if (rdev->desc_nr == nr)
843 if (rdev->bdev->bd_dev == dev)
854 if (rdev->bdev->bd_dev == dev)
865 if (level != LEVEL_NONE && pers->level == level)
867 if (strcmp(pers->name, clevel)==0)
876 return MD_NEW_SIZE_SECTORS(bdev_nr_sectors(rdev->bdev));
881 rdev->sb_page = alloc_page(GFP_KERNEL);
882 if (!rdev->sb_page)
883 return -ENOMEM;
889 if (rdev->sb_page) {
890 put_page(rdev->sb_page);
891 rdev->sb_loaded = 0;
892 rdev->sb_page = NULL;
893 rdev->sb_start = 0;
894 rdev->sectors = 0;
896 if (rdev->bb_page) {
897 put_page(rdev->bb_page);
898 rdev->bb_page = NULL;
900 badblocks_exit(&rdev->badblocks);
906 struct md_rdev *rdev = bio->bi_private;
907 struct mddev *mddev = rdev->mddev;
909 if (bio->bi_status) {
911 blk_status_to_errno(bio->bi_status));
913 if (!test_bit(Faulty, &rdev->flags)
914 && (bio->bi_opf & MD_FAILFAST)) {
915 set_bit(MD_SB_NEED_REWRITE, &mddev->sb_flags);
916 set_bit(LastDev, &rdev->flags);
919 clear_bit(LastDev, &rdev->flags);
925 if (atomic_dec_and_test(&mddev->pending_writes))
926 wake_up(&mddev->sb_wait);
933 * Increment mddev->pending_writes before returning
943 if (test_bit(Faulty, &rdev->flags))
946 bio = bio_alloc_bioset(rdev->meta_bdev ? rdev->meta_bdev : rdev->bdev,
950 GFP_NOIO, &mddev->sync_set);
952 atomic_inc(&rdev->nr_pending);
954 bio->bi_iter.bi_sector = sector;
956 bio->bi_private = rdev;
957 bio->bi_end_io = super_written;
959 if (test_bit(MD_FAILFAST_SUPPORTED, &mddev->flags) &&
960 test_bit(FailFast, &rdev->flags) &&
961 !test_bit(LastDev, &rdev->flags))
962 bio->bi_opf |= MD_FAILFAST;
964 atomic_inc(&mddev->pending_writes);
971 wait_event(mddev->sb_wait, atomic_read(&mddev->pending_writes)==0);
972 if (test_and_clear_bit(MD_SB_NEED_REWRITE, &mddev->sb_flags))
973 return -EAGAIN;
983 if (metadata_op && rdev->meta_bdev)
984 bio_init(&bio, rdev->meta_bdev, &bvec, 1, opf);
986 bio_init(&bio, rdev->bdev, &bvec, 1, opf);
989 bio.bi_iter.bi_sector = sector + rdev->sb_start;
990 else if (rdev->mddev->reshape_position != MaxSector &&
991 (rdev->mddev->reshape_backwards ==
992 (sector >= rdev->mddev->reshape_position)))
993 bio.bi_iter.bi_sector = sector + rdev->new_data_offset;
995 bio.bi_iter.bi_sector = sector + rdev->data_offset;
1006 if (rdev->sb_loaded)
1009 if (!sync_page_io(rdev, 0, size, rdev->sb_page, REQ_OP_READ, true))
1011 rdev->sb_loaded = 1;
1016 rdev->bdev);
1017 return -EINVAL;
1022 return sb1->set_uuid0 == sb2->set_uuid0 &&
1023 sb1->set_uuid1 == sb2->set_uuid1 &&
1024 sb1->set_uuid2 == sb2->set_uuid2 &&
1025 sb1->set_uuid3 == sb2->set_uuid3;
1047 tmp1->nr_disks = 0;
1048 tmp2->nr_disks = 0;
1070 disk_csum = sb->sb_csum;
1071 sb->sb_csum = 0;
1086 sb->sb_csum = md_csum_fold(disk_csum);
1088 sb->sb_csum = disk_csum;
1098 * We rely on user-space to write the initial superblock, and support
1105 * 0 - dev has a superblock that is compatible with refdev
1106 * 1 - dev has a superblock that is compatible and newer than refdev
1108 * -EINVAL superblock incompatible or invalid
1109 * -othererror e.g. -EIO
1113 * The first time, mddev->raid_disks will be 0, and data from
1115 * is new enough. Return 0 or -EINVAL
1144 * support bitmaps. It prints an error message and returns non-zero if mddev
1150 if (!mddev->bitmap_info.file && !mddev->bitmap_info.offset)
1153 mdname(mddev), mddev->pers->name);
1173 rdev->sb_start = calc_dev_sboffset(rdev);
1179 ret = -EINVAL;
1181 sb = page_address(rdev->sb_page);
1183 if (sb->md_magic != MD_SB_MAGIC) {
1185 rdev->bdev);
1189 if (sb->major_version != 0 ||
1190 sb->minor_version < 90 ||
1191 sb->minor_version > 91) {
1193 sb->major_version, sb->minor_version, rdev->bdev);
1197 if (sb->raid_disks <= 0)
1200 if (md_csum_fold(calc_sb_csum(sb)) != md_csum_fold(sb->sb_csum)) {
1201 pr_warn("md: invalid superblock checksum on %pg\n", rdev->bdev);
1205 rdev->preferred_minor = sb->md_minor;
1206 rdev->data_offset = 0;
1207 rdev->new_data_offset = 0;
1208 rdev->sb_size = MD_SB_BYTES;
1209 rdev->badblocks.shift = -1;
1211 if (sb->level == LEVEL_MULTIPATH)
1212 rdev->desc_nr = -1;
1214 rdev->desc_nr = sb->this_disk.number;
1217 if (sb->level == LEVEL_MULTIPATH ||
1218 (rdev->desc_nr >= 0 &&
1219 rdev->desc_nr < MD_SB_DISKS &&
1220 sb->disks[rdev->desc_nr].state &
1231 mdp_super_t *refsb = page_address(refdev->sb_page);
1234 rdev->bdev, refdev->bdev);
1239 rdev->bdev, refdev->bdev);
1250 rdev->sectors = rdev->sb_start;
1255 if ((u64)rdev->sectors >= (2ULL << 32) && sb->level >= 1)
1256 rdev->sectors = (sector_t)(2ULL << 32) - 2;
1258 if (rdev->sectors < ((sector_t)sb->size) * 2 && sb->level >= 1)
1260 ret = -EINVAL;
1273 mdp_super_t *sb = page_address(rdev->sb_page);
1276 rdev->raid_disk = -1;
1277 clear_bit(Faulty, &rdev->flags);
1278 clear_bit(In_sync, &rdev->flags);
1279 clear_bit(Bitmap_sync, &rdev->flags);
1280 clear_bit(WriteMostly, &rdev->flags);
1282 if (mddev->raid_disks == 0) {
1283 mddev->major_version = 0;
1284 mddev->minor_version = sb->minor_version;
1285 mddev->patch_version = sb->patch_version;
1286 mddev->external = 0;
1287 mddev->chunk_sectors = sb->chunk_size >> 9;
1288 mddev->ctime = sb->ctime;
1289 mddev->utime = sb->utime;
1290 mddev->level = sb->level;
1291 mddev->clevel[0] = 0;
1292 mddev->layout = sb->layout;
1293 mddev->raid_disks = sb->raid_disks;
1294 mddev->dev_sectors = ((sector_t)sb->size) * 2;
1295 mddev->events = ev1;
1296 mddev->bitmap_info.offset = 0;
1297 mddev->bitmap_info.space = 0;
1299 mddev->bitmap_info.default_offset = MD_SB_BYTES >> 9;
1300 mddev->bitmap_info.default_space = 64*2 - (MD_SB_BYTES >> 9);
1301 mddev->reshape_backwards = 0;
1303 if (mddev->minor_version >= 91) {
1304 mddev->reshape_position = sb->reshape_position;
1305 mddev->delta_disks = sb->delta_disks;
1306 mddev->new_level = sb->new_level;
1307 mddev->new_layout = sb->new_layout;
1308 mddev->new_chunk_sectors = sb->new_chunk >> 9;
1309 if (mddev->delta_disks < 0)
1310 mddev->reshape_backwards = 1;
1312 mddev->reshape_position = MaxSector;
1313 mddev->delta_disks = 0;
1314 mddev->new_level = mddev->level;
1315 mddev->new_layout = mddev->layout;
1316 mddev->new_chunk_sectors = mddev->chunk_sectors;
1318 if (mddev->level == 0)
1319 mddev->layout = -1;
1321 if (sb->state & (1<<MD_SB_CLEAN))
1322 mddev->recovery_cp = MaxSector;
1324 if (sb->events_hi == sb->cp_events_hi &&
1325 sb->events_lo == sb->cp_events_lo) {
1326 mddev->recovery_cp = sb->recovery_cp;
1328 mddev->recovery_cp = 0;
1331 memcpy(mddev->uuid+0, &sb->set_uuid0, 4);
1332 memcpy(mddev->uuid+4, &sb->set_uuid1, 4);
1333 memcpy(mddev->uuid+8, &sb->set_uuid2, 4);
1334 memcpy(mddev->uuid+12,&sb->set_uuid3, 4);
1336 mddev->max_disks = MD_SB_DISKS;
1338 if (sb->state & (1<<MD_SB_BITMAP_PRESENT) &&
1339 mddev->bitmap_info.file == NULL) {
1340 mddev->bitmap_info.offset =
1341 mddev->bitmap_info.default_offset;
1342 mddev->bitmap_info.space =
1343 mddev->bitmap_info.default_space;
1346 } else if (mddev->pers == NULL) {
1350 if (sb->disks[rdev->desc_nr].state & (
1352 if (ev1 < mddev->events)
1353 return -EINVAL;
1354 } else if (mddev->bitmap) {
1358 if (ev1 < mddev->bitmap->events_cleared)
1360 if (ev1 < mddev->events)
1361 set_bit(Bitmap_sync, &rdev->flags);
1363 if (ev1 < mddev->events)
1364 /* just a hot-add of a new device, leave raid_disk at -1 */
1368 if (mddev->level != LEVEL_MULTIPATH) {
1369 desc = sb->disks + rdev->desc_nr;
1371 if (desc->state & (1<<MD_DISK_FAULTY))
1372 set_bit(Faulty, &rdev->flags);
1373 else if (desc->state & (1<<MD_DISK_SYNC) /* &&
1374 desc->raid_disk < mddev->raid_disks */) {
1375 set_bit(In_sync, &rdev->flags);
1376 rdev->raid_disk = desc->raid_disk;
1377 rdev->saved_raid_disk = desc->raid_disk;
1378 } else if (desc->state & (1<<MD_DISK_ACTIVE)) {
1382 if (mddev->minor_version >= 91) {
1383 rdev->recovery_offset = 0;
1384 rdev->raid_disk = desc->raid_disk;
1387 if (desc->state & (1<<MD_DISK_WRITEMOSTLY))
1388 set_bit(WriteMostly, &rdev->flags);
1389 if (desc->state & (1<<MD_DISK_FAILFAST))
1390 set_bit(FailFast, &rdev->flags);
1392 set_bit(In_sync, &rdev->flags);
1403 int next_spare = mddev->raid_disks;
1405 /* make rdev->sb match mddev data..
1418 rdev->sb_size = MD_SB_BYTES;
1420 sb = page_address(rdev->sb_page);
1424 sb->md_magic = MD_SB_MAGIC;
1425 sb->major_version = mddev->major_version;
1426 sb->patch_version = mddev->patch_version;
1427 sb->gvalid_words = 0; /* ignored */
1428 memcpy(&sb->set_uuid0, mddev->uuid+0, 4);
1429 memcpy(&sb->set_uuid1, mddev->uuid+4, 4);
1430 memcpy(&sb->set_uuid2, mddev->uuid+8, 4);
1431 memcpy(&sb->set_uuid3, mddev->uuid+12,4);
1433 sb->ctime = clamp_t(time64_t, mddev->ctime, 0, U32_MAX);
1434 sb->level = mddev->level;
1435 sb->size = mddev->dev_sectors / 2;
1436 sb->raid_disks = mddev->raid_disks;
1437 sb->md_minor = mddev->md_minor;
1438 sb->not_persistent = 0;
1439 sb->utime = clamp_t(time64_t, mddev->utime, 0, U32_MAX);
1440 sb->state = 0;
1441 sb->events_hi = (mddev->events>>32);
1442 sb->events_lo = (u32)mddev->events;
1444 if (mddev->reshape_position == MaxSector)
1445 sb->minor_version = 90;
1447 sb->minor_version = 91;
1448 sb->reshape_position = mddev->reshape_position;
1449 sb->new_level = mddev->new_level;
1450 sb->delta_disks = mddev->delta_disks;
1451 sb->new_layout = mddev->new_layout;
1452 sb->new_chunk = mddev->new_chunk_sectors << 9;
1454 mddev->minor_version = sb->minor_version;
1455 if (mddev->in_sync)
1457 sb->recovery_cp = mddev->recovery_cp;
1458 sb->cp_events_hi = (mddev->events>>32);
1459 sb->cp_events_lo = (u32)mddev->events;
1460 if (mddev->recovery_cp == MaxSector)
1461 sb->state = (1<< MD_SB_CLEAN);
1463 sb->recovery_cp = 0;
1465 sb->layout = mddev->layout;
1466 sb->chunk_size = mddev->chunk_sectors << 9;
1468 if (mddev->bitmap && mddev->bitmap_info.file == NULL)
1469 sb->state |= (1<<MD_SB_BITMAP_PRESENT);
1471 sb->disks[0].state = (1<<MD_DISK_REMOVED);
1475 int is_active = test_bit(In_sync, &rdev2->flags);
1477 if (rdev2->raid_disk >= 0 &&
1478 sb->minor_version >= 91)
1481 * we can piggy-back on that.
1484 if (rdev2->raid_disk < 0 ||
1485 test_bit(Faulty, &rdev2->flags))
1488 desc_nr = rdev2->raid_disk;
1491 rdev2->desc_nr = desc_nr;
1492 d = &sb->disks[rdev2->desc_nr];
1494 d->number = rdev2->desc_nr;
1495 d->major = MAJOR(rdev2->bdev->bd_dev);
1496 d->minor = MINOR(rdev2->bdev->bd_dev);
1498 d->raid_disk = rdev2->raid_disk;
1500 d->raid_disk = rdev2->desc_nr; /* compatibility */
1501 if (test_bit(Faulty, &rdev2->flags))
1502 d->state = (1<<MD_DISK_FAULTY);
1504 d->state = (1<<MD_DISK_ACTIVE);
1505 if (test_bit(In_sync, &rdev2->flags))
1506 d->state |= (1<<MD_DISK_SYNC);
1510 d->state = 0;
1514 if (test_bit(WriteMostly, &rdev2->flags))
1515 d->state |= (1<<MD_DISK_WRITEMOSTLY);
1516 if (test_bit(FailFast, &rdev2->flags))
1517 d->state |= (1<<MD_DISK_FAILFAST);
1520 for (i=0 ; i < mddev->raid_disks ; i++) {
1521 mdp_disk_t *d = &sb->disks[i];
1522 if (d->state == 0 && d->number == 0) {
1523 d->number = i;
1524 d->raid_disk = i;
1525 d->state = (1<<MD_DISK_REMOVED);
1526 d->state |= (1<<MD_DISK_FAULTY);
1530 sb->nr_disks = nr_disks;
1531 sb->active_disks = active;
1532 sb->working_disks = working;
1533 sb->failed_disks = failed;
1534 sb->spare_disks = spare;
1536 sb->this_disk = sb->disks[rdev->desc_nr];
1537 sb->sb_csum = calc_sb_csum(sb);
1546 if (num_sectors && num_sectors < rdev->mddev->dev_sectors)
1548 if (rdev->mddev->bitmap_info.offset)
1550 rdev->sb_start = calc_dev_sboffset(rdev);
1551 if (!num_sectors || num_sectors > rdev->sb_start)
1552 num_sectors = rdev->sb_start;
1556 if ((u64)num_sectors >= (2ULL << 32) && rdev->mddev->level >= 1)
1557 num_sectors = (sector_t)(2ULL << 32) - 2;
1559 md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
1560 rdev->sb_page);
1561 } while (md_super_wait(rdev->mddev) < 0);
1568 /* non-zero offset changes not possible with v0.90 */
1581 int size = 256 + le32_to_cpu(sb->max_dev)*2;
1584 disk_csum = sb->sb_csum;
1585 sb->sb_csum = 0;
1587 for (; size >= 4; size -= 4)
1594 sb->sb_csum = disk_csum;
1617 sb_start = bdev_nr_sectors(rdev->bdev) - 8 * 2;
1618 sb_start &= ~(sector_t)(4*2-1);
1627 return -EINVAL;
1629 rdev->sb_start = sb_start;
1637 sb = page_address(rdev->sb_page);
1639 if (sb->magic != cpu_to_le32(MD_SB_MAGIC) ||
1640 sb->major_version != cpu_to_le32(1) ||
1641 le32_to_cpu(sb->max_dev) > (4096-256)/2 ||
1642 le64_to_cpu(sb->super_offset) != rdev->sb_start ||
1643 (le32_to_cpu(sb->feature_map) & ~MD_FEATURE_ALL) != 0)
1644 return -EINVAL;
1646 if (calc_sb_1_csum(sb) != sb->sb_csum) {
1648 rdev->bdev);
1649 return -EINVAL;
1651 if (le64_to_cpu(sb->data_size) < 10) {
1653 rdev->bdev);
1654 return -EINVAL;
1656 if (sb->pad0 ||
1657 sb->pad3[0] ||
1658 memcmp(sb->pad3, sb->pad3+1, sizeof(sb->pad3) - sizeof(sb->pad3[1])))
1659 /* Some padding is non-zero, might be a new feature */
1660 return -EINVAL;
1662 rdev->preferred_minor = 0xffff;
1663 rdev->data_offset = le64_to_cpu(sb->data_offset);
1664 rdev->new_data_offset = rdev->data_offset;
1665 if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_RESHAPE_ACTIVE) &&
1666 (le32_to_cpu(sb->feature_map) & MD_FEATURE_NEW_OFFSET))
1667 rdev->new_data_offset += (s32)le32_to_cpu(sb->new_offset);
1668 atomic_set(&rdev->corrected_errors, le32_to_cpu(sb->cnt_corrected_read));
1670 rdev->sb_size = le32_to_cpu(sb->max_dev) * 2 + 256;
1671 bmask = queue_logical_block_size(rdev->bdev->bd_disk->queue)-1;
1672 if (rdev->sb_size & bmask)
1673 rdev->sb_size = (rdev->sb_size | bmask) + 1;
1676 && rdev->data_offset < sb_start + (rdev->sb_size/512))
1677 return -EINVAL;
1679 && rdev->new_data_offset < sb_start + (rdev->sb_size/512))
1680 return -EINVAL;
1682 if (sb->level == cpu_to_le32(LEVEL_MULTIPATH))
1683 rdev->desc_nr = -1;
1685 rdev->desc_nr = le32_to_cpu(sb->dev_number);
1687 if (!rdev->bb_page) {
1688 rdev->bb_page = alloc_page(GFP_KERNEL);
1689 if (!rdev->bb_page)
1690 return -ENOMEM;
1692 if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_BAD_BLOCKS) &&
1693 rdev->badblocks.count == 0) {
1701 int sectors = le16_to_cpu(sb->bblog_size);
1703 return -EINVAL;
1704 offset = le32_to_cpu(sb->bblog_offset);
1706 return -EINVAL;
1709 rdev->bb_page, REQ_OP_READ, true))
1710 return -EIO;
1711 bbp = (__le64 *)page_address(rdev->bb_page);
1712 rdev->badblocks.shift = sb->bblog_shift;
1713 for (i = 0 ; i < (sectors << (9-3)) ; i++, bbp++) {
1717 sector <<= sb->bblog_shift;
1718 count <<= sb->bblog_shift;
1721 if (badblocks_set(&rdev->badblocks, sector, count, 1))
1722 return -EINVAL;
1724 } else if (sb->bblog_offset != 0)
1725 rdev->badblocks.shift = 0;
1727 if ((le32_to_cpu(sb->feature_map) &
1729 rdev->ppl.offset = (__s16)le16_to_cpu(sb->ppl.offset);
1730 rdev->ppl.size = le16_to_cpu(sb->ppl.size);
1731 rdev->ppl.sector = rdev->sb_start + rdev->ppl.offset;
1734 if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_RAID0_LAYOUT) &&
1735 sb->level != 0)
1736 return -EINVAL;
1739 if (sb->level == cpu_to_le32(LEVEL_MULTIPATH) ||
1740 (rdev->desc_nr >= 0 &&
1741 rdev->desc_nr < le32_to_cpu(sb->max_dev) &&
1742 (le16_to_cpu(sb->dev_roles[rdev->desc_nr]) < MD_DISK_ROLE_MAX ||
1743 le16_to_cpu(sb->dev_roles[rdev->desc_nr]) == MD_DISK_ROLE_JOURNAL)))
1753 struct mdp_superblock_1 *refsb = page_address(refdev->sb_page);
1755 if (memcmp(sb->set_uuid, refsb->set_uuid, 16) != 0 ||
1756 sb->level != refsb->level ||
1757 sb->layout != refsb->layout ||
1758 sb->chunksize != refsb->chunksize) {
1760 rdev->bdev,
1761 refdev->bdev);
1762 return -EINVAL;
1764 ev1 = le64_to_cpu(sb->events);
1765 ev2 = le64_to_cpu(refsb->events);
1773 sectors = bdev_nr_sectors(rdev->bdev) - rdev->data_offset;
1775 sectors = rdev->sb_start;
1776 if (sectors < le64_to_cpu(sb->data_size))
1777 return -EINVAL;
1778 rdev->sectors = le64_to_cpu(sb->data_size);
1784 struct mdp_superblock_1 *sb = page_address(rdev->sb_page);
1785 __u64 ev1 = le64_to_cpu(sb->events);
1787 rdev->raid_disk = -1;
1788 clear_bit(Faulty, &rdev->flags);
1789 clear_bit(In_sync, &rdev->flags);
1790 clear_bit(Bitmap_sync, &rdev->flags);
1791 clear_bit(WriteMostly, &rdev->flags);
1793 if (mddev->raid_disks == 0) {
1794 mddev->major_version = 1;
1795 mddev->patch_version = 0;
1796 mddev->external = 0;
1797 mddev->chunk_sectors = le32_to_cpu(sb->chunksize);
1798 mddev->ctime = le64_to_cpu(sb->ctime);
1799 mddev->utime = le64_to_cpu(sb->utime);
1800 mddev->level = le32_to_cpu(sb->level);
1801 mddev->clevel[0] = 0;
1802 mddev->layout = le32_to_cpu(sb->layout);
1803 mddev->raid_disks = le32_to_cpu(sb->raid_disks);
1804 mddev->dev_sectors = le64_to_cpu(sb->size);
1805 mddev->events = ev1;
1806 mddev->bitmap_info.offset = 0;
1807 mddev->bitmap_info.space = 0;
1809 * using 3K - total of 4K
1811 mddev->bitmap_info.default_offset = 1024 >> 9;
1812 mddev->bitmap_info.default_space = (4096-1024) >> 9;
1813 mddev->reshape_backwards = 0;
1815 mddev->recovery_cp = le64_to_cpu(sb->resync_offset);
1816 memcpy(mddev->uuid, sb->set_uuid, 16);
1818 mddev->max_disks = (4096-256)/2;
1820 if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_BITMAP_OFFSET) &&
1821 mddev->bitmap_info.file == NULL) {
1822 mddev->bitmap_info.offset =
1823 (__s32)le32_to_cpu(sb->bitmap_offset);
1829 if (mddev->minor_version > 0)
1830 mddev->bitmap_info.space = 0;
1831 else if (mddev->bitmap_info.offset > 0)
1832 mddev->bitmap_info.space =
1833 8 - mddev->bitmap_info.offset;
1835 mddev->bitmap_info.space =
1836 -mddev->bitmap_info.offset;
1839 if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_RESHAPE_ACTIVE)) {
1840 mddev->reshape_position = le64_to_cpu(sb->reshape_position);
1841 mddev->delta_disks = le32_to_cpu(sb->delta_disks);
1842 mddev->new_level = le32_to_cpu(sb->new_level);
1843 mddev->new_layout = le32_to_cpu(sb->new_layout);
1844 mddev->new_chunk_sectors = le32_to_cpu(sb->new_chunk);
1845 if (mddev->delta_disks < 0 ||
1846 (mddev->delta_disks == 0 &&
1847 (le32_to_cpu(sb->feature_map)
1849 mddev->reshape_backwards = 1;
1851 mddev->reshape_position = MaxSector;
1852 mddev->delta_disks = 0;
1853 mddev->new_level = mddev->level;
1854 mddev->new_layout = mddev->layout;
1855 mddev->new_chunk_sectors = mddev->chunk_sectors;
1858 if (mddev->level == 0 &&
1859 !(le32_to_cpu(sb->feature_map) & MD_FEATURE_RAID0_LAYOUT))
1860 mddev->layout = -1;
1862 if (le32_to_cpu(sb->feature_map) & MD_FEATURE_JOURNAL)
1863 set_bit(MD_HAS_JOURNAL, &mddev->flags);
1865 if (le32_to_cpu(sb->feature_map) &
1867 if (le32_to_cpu(sb->feature_map) &
1869 return -EINVAL;
1870 if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_PPL) &&
1871 (le32_to_cpu(sb->feature_map) &
1873 return -EINVAL;
1874 set_bit(MD_HAS_PPL, &mddev->flags);
1876 } else if (mddev->pers == NULL) {
1882 if (rdev->desc_nr >= 0 &&
1883 rdev->desc_nr < le32_to_cpu(sb->max_dev) &&
1884 (le16_to_cpu(sb->dev_roles[rdev->desc_nr]) < MD_DISK_ROLE_MAX ||
1885 le16_to_cpu(sb->dev_roles[rdev->desc_nr]) == MD_DISK_ROLE_JOURNAL))
1886 if (ev1 + 1 < mddev->events)
1887 return -EINVAL;
1888 } else if (mddev->bitmap) {
1892 if (ev1 < mddev->bitmap->events_cleared)
1894 if (ev1 < mddev->events)
1895 set_bit(Bitmap_sync, &rdev->flags);
1897 if (ev1 < mddev->events)
1898 /* just a hot-add of a new device, leave raid_disk at -1 */
1901 if (mddev->level != LEVEL_MULTIPATH) {
1903 if (rdev->desc_nr < 0 ||
1904 rdev->desc_nr >= le32_to_cpu(sb->max_dev)) {
1906 rdev->desc_nr = -1;
1907 } else if (mddev->pers == NULL && freshest && ev1 < mddev->events) {
1922 struct mdp_superblock_1 *freshest_sb = page_address(freshest->sb_page);
1923 u32 freshest_max_dev = le32_to_cpu(freshest_sb->max_dev);
1925 if (rdev->desc_nr >= freshest_max_dev) {
1927 pr_warn("md: %s: rdev[%pg]: desc_nr(%d) >= freshest(%pg)->sb->max_dev(%u)\n",
1928 mdname(mddev), rdev->bdev, rdev->desc_nr,
1929 freshest->bdev, freshest_max_dev);
1930 return -EUCLEAN;
1933 role = le16_to_cpu(freshest_sb->dev_roles[rdev->desc_nr]);
1935 mdname(mddev), rdev->bdev, role, role, freshest->bdev);
1937 role = le16_to_cpu(sb->dev_roles[rdev->desc_nr]);
1943 set_bit(Faulty, &rdev->flags);
1946 if (!(le32_to_cpu(sb->feature_map) & MD_FEATURE_JOURNAL)) {
1949 return -EINVAL;
1951 set_bit(Journal, &rdev->flags);
1952 rdev->journal_tail = le64_to_cpu(sb->journal_tail);
1953 rdev->raid_disk = 0;
1956 rdev->saved_raid_disk = role;
1957 if ((le32_to_cpu(sb->feature_map) &
1959 rdev->recovery_offset = le64_to_cpu(sb->recovery_offset);
1960 if (!(le32_to_cpu(sb->feature_map) &
1962 rdev->saved_raid_disk = -1;
1969 &mddev->recovery))
1970 set_bit(In_sync, &rdev->flags);
1972 rdev->raid_disk = role;
1975 if (sb->devflags & WriteMostly1)
1976 set_bit(WriteMostly, &rdev->flags);
1977 if (sb->devflags & FailFast1)
1978 set_bit(FailFast, &rdev->flags);
1979 if (le32_to_cpu(sb->feature_map) & MD_FEATURE_REPLACEMENT)
1980 set_bit(Replacement, &rdev->flags);
1982 set_bit(In_sync, &rdev->flags);
1992 /* make rdev->sb match mddev and rdev data. */
1994 sb = page_address(rdev->sb_page);
1996 sb->feature_map = 0;
1997 sb->pad0 = 0;
1998 sb->recovery_offset = cpu_to_le64(0);
1999 memset(sb->pad3, 0, sizeof(sb->pad3));
2001 sb->utime = cpu_to_le64((__u64)mddev->utime);
2002 sb->events = cpu_to_le64(mddev->events);
2003 if (mddev->in_sync)
2004 sb->resync_offset = cpu_to_le64(mddev->recovery_cp);
2005 else if (test_bit(MD_JOURNAL_CLEAN, &mddev->flags))
2006 sb->resync_offset = cpu_to_le64(MaxSector);
2008 sb->resync_offset = cpu_to_le64(0);
2010 sb->cnt_corrected_read = cpu_to_le32(atomic_read(&rdev->corrected_errors));
2012 sb->raid_disks = cpu_to_le32(mddev->raid_disks);
2013 sb->size = cpu_to_le64(mddev->dev_sectors);
2014 sb->chunksize = cpu_to_le32(mddev->chunk_sectors);
2015 sb->level = cpu_to_le32(mddev->level);
2016 sb->layout = cpu_to_le32(mddev->layout);
2017 if (test_bit(FailFast, &rdev->flags))
2018 sb->devflags |= FailFast1;
2020 sb->devflags &= ~FailFast1;
2022 if (test_bit(WriteMostly, &rdev->flags))
2023 sb->devflags |= WriteMostly1;
2025 sb->devflags &= ~WriteMostly1;
2026 sb->data_offset = cpu_to_le64(rdev->data_offset);
2027 sb->data_size = cpu_to_le64(rdev->sectors);
2029 if (mddev->bitmap && mddev->bitmap_info.file == NULL) {
2030 sb->bitmap_offset = cpu_to_le32((__u32)mddev->bitmap_info.offset);
2031 sb->feature_map = cpu_to_le32(MD_FEATURE_BITMAP_OFFSET);
2034 if (rdev->raid_disk >= 0 && !test_bit(Journal, &rdev->flags) &&
2035 !test_bit(In_sync, &rdev->flags)) {
2036 sb->feature_map |=
2038 sb->recovery_offset =
2039 cpu_to_le64(rdev->recovery_offset);
2040 if (rdev->saved_raid_disk >= 0 && mddev->bitmap)
2041 sb->feature_map |=
2045 if (test_bit(Journal, &rdev->flags))
2046 sb->journal_tail = cpu_to_le64(rdev->journal_tail);
2047 if (test_bit(Replacement, &rdev->flags))
2048 sb->feature_map |=
2051 if (mddev->reshape_position != MaxSector) {
2052 sb->feature_map |= cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE);
2053 sb->reshape_position = cpu_to_le64(mddev->reshape_position);
2054 sb->new_layout = cpu_to_le32(mddev->new_layout);
2055 sb->delta_disks = cpu_to_le32(mddev->delta_disks);
2056 sb->new_level = cpu_to_le32(mddev->new_level);
2057 sb->new_chunk = cpu_to_le32(mddev->new_chunk_sectors);
2058 if (mddev->delta_disks == 0 &&
2059 mddev->reshape_backwards)
2060 sb->feature_map
2062 if (rdev->new_data_offset != rdev->data_offset) {
2063 sb->feature_map
2065 sb->new_offset = cpu_to_le32((__u32)(rdev->new_data_offset
2066 - rdev->data_offset));
2071 sb->feature_map |= cpu_to_le32(MD_FEATURE_CLUSTERED);
2073 if (rdev->badblocks.count == 0)
2075 else if (sb->bblog_offset == 0)
2079 struct badblocks *bb = &rdev->badblocks;
2080 __le64 *bbp = (__le64 *)page_address(rdev->bb_page);
2081 u64 *p = bb->page;
2082 sb->feature_map |= cpu_to_le32(MD_FEATURE_BAD_BLOCKS);
2083 if (bb->changed) {
2087 seq = read_seqbegin(&bb->lock);
2091 for (i = 0 ; i < bb->count ; i++) {
2097 bb->changed = 0;
2098 if (read_seqretry(&bb->lock, seq))
2101 bb->sector = (rdev->sb_start +
2102 (int)le32_to_cpu(sb->bblog_offset));
2103 bb->size = le16_to_cpu(sb->bblog_size);
2109 if (rdev2->desc_nr+1 > max_dev)
2110 max_dev = rdev2->desc_nr+1;
2112 if (max_dev > le32_to_cpu(sb->max_dev)) {
2114 sb->max_dev = cpu_to_le32(max_dev);
2115 rdev->sb_size = max_dev * 2 + 256;
2116 bmask = queue_logical_block_size(rdev->bdev->bd_disk->queue)-1;
2117 if (rdev->sb_size & bmask)
2118 rdev->sb_size = (rdev->sb_size | bmask) + 1;
2120 max_dev = le32_to_cpu(sb->max_dev);
2123 sb->dev_roles[i] = cpu_to_le16(MD_DISK_ROLE_SPARE);
2125 if (test_bit(MD_HAS_JOURNAL, &mddev->flags))
2126 sb->feature_map |= cpu_to_le32(MD_FEATURE_JOURNAL);
2128 if (test_bit(MD_HAS_PPL, &mddev->flags)) {
2129 if (test_bit(MD_HAS_MULTIPLE_PPLS, &mddev->flags))
2130 sb->feature_map |=
2133 sb->feature_map |= cpu_to_le32(MD_FEATURE_PPL);
2134 sb->ppl.offset = cpu_to_le16(rdev->ppl.offset);
2135 sb->ppl.size = cpu_to_le16(rdev->ppl.size);
2139 i = rdev2->desc_nr;
2140 if (test_bit(Faulty, &rdev2->flags))
2141 sb->dev_roles[i] = cpu_to_le16(MD_DISK_ROLE_FAULTY);
2142 else if (test_bit(In_sync, &rdev2->flags))
2143 sb->dev_roles[i] = cpu_to_le16(rdev2->raid_disk);
2144 else if (test_bit(Journal, &rdev2->flags))
2145 sb->dev_roles[i] = cpu_to_le16(MD_DISK_ROLE_JOURNAL);
2146 else if (rdev2->raid_disk >= 0)
2147 sb->dev_roles[i] = cpu_to_le16(rdev2->raid_disk);
2149 sb->dev_roles[i] = cpu_to_le16(MD_DISK_ROLE_SPARE);
2152 sb->sb_csum = calc_sb_1_csum(sb);
2164 else if (dev_size - 64*2 >= 200*1024*1024*2)
2166 else if (dev_size - 4*2 > 8*1024*1024*2)
2178 if (num_sectors && num_sectors < rdev->mddev->dev_sectors)
2180 if (rdev->data_offset != rdev->new_data_offset)
2182 if (rdev->sb_start < rdev->data_offset) {
2184 max_sectors = bdev_nr_sectors(rdev->bdev) - rdev->data_offset;
2187 } else if (rdev->mddev->bitmap_info.offset) {
2193 sector_t dev_size = bdev_nr_sectors(rdev->bdev);
2196 sb_start = dev_size - 8*2;
2197 sb_start &= ~(sector_t)(4*2 - 1);
2204 max_sectors = sb_start - bm_space - 4*2;
2208 rdev->sb_start = sb_start;
2210 sb = page_address(rdev->sb_page);
2211 sb->data_size = cpu_to_le64(num_sectors);
2212 sb->super_offset = cpu_to_le64(rdev->sb_start);
2213 sb->sb_csum = calc_sb_1_csum(sb);
2215 md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
2216 rdev->sb_page);
2217 } while (md_super_wait(rdev->mddev) < 0);
2228 if (new_offset >= rdev->data_offset)
2233 if (rdev->mddev->minor_version == 0)
2240 * beyond write-intent bitmap
2242 if (rdev->sb_start + (32+4)*2 > new_offset)
2244 bitmap = rdev->mddev->bitmap;
2245 if (bitmap && !rdev->mddev->bitmap_info.file &&
2246 rdev->sb_start + rdev->mddev->bitmap_info.offset +
2247 bitmap->storage.file_pages * (PAGE_SIZE>>9) > new_offset)
2249 if (rdev->badblocks.sector + rdev->badblocks.size > new_offset)
2266 .name = "md-1",
2278 if (mddev->sync_super) {
2279 mddev->sync_super(mddev, rdev);
2283 BUG_ON(mddev->major_version >= ARRAY_SIZE(super_types));
2285 super_types[mddev->major_version].sync_super(mddev, rdev);
2294 if (test_bit(Faulty, &rdev->flags) ||
2295 test_bit(Journal, &rdev->flags) ||
2296 rdev->raid_disk == -1)
2299 if (test_bit(Faulty, &rdev2->flags) ||
2300 test_bit(Journal, &rdev2->flags) ||
2301 rdev2->raid_disk == -1)
2303 if (rdev->bdev->bd_disk == rdev2->bdev->bd_disk) {
2326 if (list_empty(&mddev->disks))
2328 if (!mddev->gendisk || blk_get_integrity(mddev->gendisk))
2331 /* skip spares and non-functional disks */
2332 if (test_bit(Faulty, &rdev->flags))
2334 if (rdev->raid_disk < 0)
2342 if (blk_integrity_compare(reference->bdev->bd_disk,
2343 rdev->bdev->bd_disk) < 0)
2344 return -EINVAL;
2346 if (!reference || !bdev_get_integrity(reference->bdev))
2352 blk_integrity_register(mddev->gendisk,
2353 bdev_get_integrity(reference->bdev));
2356 if (bioset_integrity_create(&mddev->bio_set, BIO_POOL_SIZE) ||
2357 (mddev->level != 1 && mddev->level != 10 &&
2358 bioset_integrity_create(&mddev->io_clone_set, BIO_POOL_SIZE))) {
2361 * because the function is called by md_run() -> pers->run(),
2362 * md_run calls bioset_exit -> bioset_integrity_free in case
2367 return -EINVAL;
2381 if (!mddev->gendisk)
2384 bi_mddev = blk_get_integrity(mddev->gendisk);
2389 if (blk_integrity_compare(mddev->gendisk, rdev->bdev->bd_disk) != 0) {
2391 mdname(mddev), rdev->bdev);
2392 return -ENXIO;
2401 return bdev_read_only(rdev->bdev) ||
2402 (rdev->meta_bdev && bdev_read_only(rdev->meta_bdev));
2411 if (find_rdev(mddev, rdev->bdev->bd_dev))
2412 return -EEXIST;
2414 if (rdev_read_only(rdev) && 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 snprintf(b, sizeof(b), "%pg", rdev->bdev);
2460 rdev->mddev = mddev;
2463 if (mddev->raid_disks)
2466 if ((err = kobject_add(&rdev->kobj, &mddev->kobj, "dev-%s", b)))
2470 err = sysfs_create_link(&rdev->kobj, bdev_kobj(rdev->bdev), "block");
2471 rdev->sysfs_state = sysfs_get_dirent_safe(rdev->kobj.sd, "state");
2472 rdev->sysfs_unack_badblocks =
2473 sysfs_get_dirent_safe(rdev->kobj.sd, "unacknowledged_bad_blocks");
2474 rdev->sysfs_badblocks =
2475 sysfs_get_dirent_safe(rdev->kobj.sd, "bad_blocks");
2477 list_add_rcu(&rdev->same_set, &mddev->disks);
2478 bd_link_disk_holder(rdev->bdev, mddev->gendisk);
2481 mddev->recovery_disabled++;
2486 pr_warn("md: failed to register dev-%s for %s\n",
2498 pr_debug("md: export_rdev(%pg)\n", rdev->bdev);
2501 if (test_bit(AutoDetected, &rdev->flags))
2502 md_autodetect_dev(rdev->bdev->bd_dev);
2504 blkdev_put(rdev->bdev,
2505 test_bit(Holder, &rdev->flags) ? rdev : &claim_rdev);
2506 rdev->bdev = NULL;
2507 kobject_put(&rdev->kobj);
2512 struct mddev *mddev = rdev->mddev;
2514 bd_unlink_disk_holder(rdev->bdev, rdev->mddev->gendisk);
2515 list_del_rcu(&rdev->same_set);
2516 pr_debug("md: unbind<%pg>\n", rdev->bdev);
2517 mddev_destroy_serial_pool(rdev->mddev, rdev, false);
2518 rdev->mddev = NULL;
2519 sysfs_remove_link(&rdev->kobj, "block");
2520 sysfs_put(rdev->sysfs_state);
2521 sysfs_put(rdev->sysfs_unack_badblocks);
2522 sysfs_put(rdev->sysfs_badblocks);
2523 rdev->sysfs_state = NULL;
2524 rdev->sysfs_unack_badblocks = NULL;
2525 rdev->sysfs_badblocks = NULL;
2526 rdev->badblocks.count = 0;
2535 list_add(&rdev->same_set, &mddev->deleting);
2542 while (!list_empty(&mddev->disks)) {
2543 rdev = list_first_entry(&mddev->disks, struct md_rdev,
2547 mddev->raid_disks = 0;
2548 mddev->major_version = 0;
2553 lockdep_assert_held(&mddev->lock);
2554 if (!mddev->in_sync) {
2555 mddev->sync_checkers++;
2556 spin_unlock(&mddev->lock);
2557 percpu_ref_switch_to_atomic_sync(&mddev->writes_pending);
2558 spin_lock(&mddev->lock);
2559 if (!mddev->in_sync &&
2560 percpu_ref_is_zero(&mddev->writes_pending)) {
2561 mddev->in_sync = 1;
2563 * Ensure ->in_sync is visible before we clear
2564 * ->sync_checkers.
2567 set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
2568 sysfs_notify_dirent_safe(mddev->sysfs_state);
2570 if (--mddev->sync_checkers == 0)
2571 percpu_ref_switch_to_percpu(&mddev->writes_pending);
2573 if (mddev->safemode == 1)
2574 mddev->safemode = 0;
2575 return mddev->in_sync;
2580 /* Update each superblock (in-memory image), but
2588 if (rdev->sb_events == mddev->events ||
2590 rdev->raid_disk < 0 &&
2591 rdev->sb_events+1 == mddev->events)) {
2593 rdev->sb_loaded = 2;
2596 rdev->sb_loaded = 1;
2609 if ((iter->raid_disk >= 0) && !test_bit(Faulty, &iter->flags)) {
2618 sb = page_address(rdev->sb_page);
2621 role = le16_to_cpu(sb->dev_roles[rdev->desc_nr]);
2623 if (role == MD_DISK_ROLE_SPARE && rdev->raid_disk >= 0 &&
2624 !test_bit(Faulty, &rdev->flags))
2627 if (test_bit(Faulty, &rdev->flags) && (role < MD_DISK_ROLE_MAX))
2632 if ((mddev->dev_sectors != le64_to_cpu(sb->size)) ||
2633 (mddev->reshape_position != le64_to_cpu(sb->reshape_position)) ||
2634 (mddev->layout != le32_to_cpu(sb->layout)) ||
2635 (mddev->raid_disks != le32_to_cpu(sb->raid_disks)) ||
2636 (mddev->chunk_sectors != le32_to_cpu(sb->chunksize)))
2648 int ret = -1;
2652 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
2658 if (test_and_clear_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags))
2660 if (test_and_clear_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags))
2662 ret = md_cluster_ops->metadata_update_start(mddev);
2666 md_cluster_ops->metadata_update_cancel(mddev);
2667 bit_clear_unless(&mddev->sb_flags, BIT(MD_SB_CHANGE_PENDING),
2677 * During reshape/resync it might use array-addresses rather
2681 if (rdev->raid_disk >= 0 &&
2682 mddev->delta_disks >= 0 &&
2683 test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) &&
2684 test_bit(MD_RECOVERY_RECOVER, &mddev->recovery) &&
2685 !test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
2686 !test_bit(Journal, &rdev->flags) &&
2687 !test_bit(In_sync, &rdev->flags) &&
2688 mddev->curr_resync_completed > rdev->recovery_offset)
2689 rdev->recovery_offset = mddev->curr_resync_completed;
2692 if (!mddev->persistent) {
2693 clear_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
2694 clear_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
2695 if (!mddev->external) {
2696 clear_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
2698 if (rdev->badblocks.changed) {
2699 rdev->badblocks.changed = 0;
2700 ack_all_badblocks(&rdev->badblocks);
2703 clear_bit(Blocked, &rdev->flags);
2704 clear_bit(BlockedBadBlocks, &rdev->flags);
2705 wake_up(&rdev->blocked_wait);
2708 wake_up(&mddev->sb_wait);
2712 spin_lock(&mddev->lock);
2714 mddev->utime = ktime_get_real_seconds();
2716 if (test_and_clear_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags))
2718 if (test_and_clear_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags))
2719 /* just a clean<-> dirty transition, possibly leave spares alone,
2726 if (mddev->degraded)
2730 * might have a event_count that still looks up-to-date,
2731 * so it can be re-added without a resync.
2738 sync_req = mddev->in_sync;
2740 /* If this is just a dirty<->clean transition, and the array is clean
2743 && (mddev->in_sync && mddev->recovery_cp == MaxSector)
2744 && mddev->can_decrease_events
2745 && mddev->events != 1) {
2746 mddev->events--;
2747 mddev->can_decrease_events = 0;
2750 mddev->events ++;
2751 mddev->can_decrease_events = nospares;
2755 * This 64-bit counter should never wrap.
2759 WARN_ON(mddev->events == 0);
2762 if (rdev->badblocks.changed)
2764 if (test_bit(Faulty, &rdev->flags))
2765 set_bit(FaultRecorded, &rdev->flags);
2769 spin_unlock(&mddev->lock);
2772 mdname(mddev), mddev->in_sync);
2774 if (mddev->queue)
2775 blk_add_trace_msg(mddev->queue, "md md_update_sb");
2777 md_bitmap_update_sb(mddev->bitmap);
2779 if (rdev->sb_loaded != 1)
2782 if (!test_bit(Faulty, &rdev->flags)) {
2784 rdev->sb_start, rdev->sb_size,
2785 rdev->sb_page);
2787 rdev->bdev,
2788 (unsigned long long)rdev->sb_start);
2789 rdev->sb_events = mddev->events;
2790 if (rdev->badblocks.size) {
2792 rdev->badblocks.sector,
2793 rdev->badblocks.size << 9,
2794 rdev->bb_page);
2795 rdev->badblocks.size = 0;
2800 rdev->bdev);
2802 if (mddev->level == LEVEL_MULTIPATH)
2808 /* if there was a failure, MD_SB_CHANGE_DEVS was set, and we re-write super */
2811 md_cluster_ops->metadata_update_finish(mddev);
2813 if (mddev->in_sync != sync_req ||
2814 !bit_clear_unless(&mddev->sb_flags, BIT(MD_SB_CHANGE_PENDING),
2818 wake_up(&mddev->sb_wait);
2819 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
2820 sysfs_notify_dirent_safe(mddev->sysfs_completed);
2823 if (test_and_clear_bit(FaultRecorded, &rdev->flags))
2824 clear_bit(Blocked, &rdev->flags);
2827 ack_all_badblocks(&rdev->badblocks);
2828 clear_bit(BlockedBadBlocks, &rdev->flags);
2829 wake_up(&rdev->blocked_wait);
2836 struct mddev *mddev = rdev->mddev;
2838 bool add_journal = test_bit(Journal, &rdev->flags);
2840 if (!mddev->pers->hot_remove_disk || add_journal) {
2845 super_types[mddev->major_version].
2849 err = mddev->pers->hot_add_disk(mddev, rdev);
2857 sysfs_notify_dirent_safe(rdev->sysfs_state);
2859 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
2860 if (mddev->degraded)
2861 set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
2862 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
2864 md_wakeup_thread(mddev->thread);
2899 unsigned long flags = READ_ONCE(rdev->flags);
2903 rdev->badblocks.unacked_exist))
2912 (rdev->badblocks.unacked_exist
2931 len -= strlen(sep);
2940 * faulty - simulates an error
2941 * remove - disconnects the device
2942 * writemostly - sets write_mostly
2943 * -writemostly - clears write_mostly
2944 * blocked - sets the Blocked flags
2945 * -blocked - clears the Blocked and possibly simulates an error
2946 * insync - sets Insync providing device isn't active
2947 * -insync - clear Insync for a device with a slot assigned,
2949 * write_error - sets WriteErrorSeen
2950 * -write_error - clears WriteErrorSeen
2951 * {,-}failfast - set/clear FailFast
2954 struct mddev *mddev = rdev->mddev;
2955 int err = -EINVAL;
2958 if (cmd_match(buf, "faulty") && rdev->mddev->pers) {
2959 md_error(rdev->mddev, rdev);
2961 if (test_bit(MD_BROKEN, &rdev->mddev->flags))
2962 err = -EBUSY;
2966 if (rdev->mddev->pers) {
2967 clear_bit(Blocked, &rdev->flags);
2968 remove_and_add_spares(rdev->mddev, rdev);
2970 if (rdev->raid_disk >= 0)
2971 err = -EBUSY;
2975 err = md_cluster_ops->remove_disk(mddev, rdev);
2979 if (mddev->pers) {
2980 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
2981 md_wakeup_thread(mddev->thread);
2987 set_bit(WriteMostly, &rdev->flags);
2988 mddev_create_serial_pool(rdev->mddev, rdev, false);
2991 } else if (cmd_match(buf, "-writemostly")) {
2992 mddev_destroy_serial_pool(rdev->mddev, rdev, false);
2993 clear_bit(WriteMostly, &rdev->flags);
2997 set_bit(Blocked, &rdev->flags);
2999 } else if (cmd_match(buf, "-blocked")) {
3000 if (!test_bit(Faulty, &rdev->flags) &&
3001 !test_bit(ExternalBbl, &rdev->flags) &&
3002 rdev->badblocks.unacked_exist) {
3006 md_error(rdev->mddev, rdev);
3008 clear_bit(Blocked, &rdev->flags);
3009 clear_bit(BlockedBadBlocks, &rdev->flags);
3010 wake_up(&rdev->blocked_wait);
3011 set_bit(MD_RECOVERY_NEEDED, &rdev->mddev->recovery);
3012 md_wakeup_thread(rdev->mddev->thread);
3015 } else if (cmd_match(buf, "insync") && rdev->raid_disk == -1) {
3016 set_bit(In_sync, &rdev->flags);
3019 set_bit(FailFast, &rdev->flags);
3022 } else if (cmd_match(buf, "-failfast")) {
3023 clear_bit(FailFast, &rdev->flags);
3026 } else if (cmd_match(buf, "-insync") && rdev->raid_disk >= 0 &&
3027 !test_bit(Journal, &rdev->flags)) {
3028 if (rdev->mddev->pers == NULL) {
3029 clear_bit(In_sync, &rdev->flags);
3030 rdev->saved_raid_disk = rdev->raid_disk;
3031 rdev->raid_disk = -1;
3035 set_bit(WriteErrorSeen, &rdev->flags);
3037 } else if (cmd_match(buf, "-write_error")) {
3038 clear_bit(WriteErrorSeen, &rdev->flags);
3041 /* Any non-spare device that is not a replacement can
3045 if (rdev->raid_disk >= 0 &&
3046 !test_bit(Journal, &rdev->flags) &&
3047 !test_bit(Replacement, &rdev->flags))
3048 set_bit(WantReplacement, &rdev->flags);
3049 set_bit(MD_RECOVERY_NEEDED, &rdev->mddev->recovery);
3050 md_wakeup_thread(rdev->mddev->thread);
3052 } else if (cmd_match(buf, "-want_replacement")) {
3057 clear_bit(WantReplacement, &rdev->flags);
3063 if (rdev->mddev->pers)
3064 err = -EBUSY;
3066 set_bit(Replacement, &rdev->flags);
3069 } else if (cmd_match(buf, "-replacement")) {
3071 if (rdev->mddev->pers)
3072 err = -EBUSY;
3074 clear_bit(Replacement, &rdev->flags);
3077 } else if (cmd_match(buf, "re-add")) {
3078 if (!rdev->mddev->pers)
3079 err = -EINVAL;
3080 else if (test_bit(Faulty, &rdev->flags) && (rdev->raid_disk == -1) &&
3081 rdev->saved_raid_disk >= 0) {
3088 if (!mddev_is_clustered(rdev->mddev) ||
3089 (err = md_cluster_ops->gather_bitmaps(rdev)) == 0) {
3090 clear_bit(Faulty, &rdev->flags);
3094 err = -EBUSY;
3095 } else if (cmd_match(buf, "external_bbl") && (rdev->mddev->external)) {
3096 set_bit(ExternalBbl, &rdev->flags);
3097 rdev->badblocks.shift = 0;
3099 } else if (cmd_match(buf, "-external_bbl") && (rdev->mddev->external)) {
3100 clear_bit(ExternalBbl, &rdev->flags);
3106 sysfs_notify_dirent_safe(rdev->sysfs_state);
3115 return sprintf(page, "%d\n", atomic_read(&rdev->corrected_errors));
3127 atomic_set(&rdev->corrected_errors, n);
3136 if (test_bit(Journal, &rdev->flags))
3138 else if (rdev->raid_disk < 0)
3141 return sprintf(page, "%d\n", rdev->raid_disk);
3150 if (test_bit(Journal, &rdev->flags))
3151 return -EBUSY;
3153 slot = -1;
3160 return -ENOSPC;
3162 if (rdev->mddev->pers && slot == -1) {
3165 * with the personality with ->hot_*_disk.
3170 if (rdev->raid_disk == -1)
3171 return -EEXIST;
3173 if (rdev->mddev->pers->hot_remove_disk == NULL)
3174 return -EINVAL;
3175 clear_bit(Blocked, &rdev->flags);
3176 remove_and_add_spares(rdev->mddev, rdev);
3177 if (rdev->raid_disk >= 0)
3178 return -EBUSY;
3179 set_bit(MD_RECOVERY_NEEDED, &rdev->mddev->recovery);
3180 md_wakeup_thread(rdev->mddev->thread);
3181 } else if (rdev->mddev->pers) {
3187 if (rdev->raid_disk != -1)
3188 return -EBUSY;
3190 if (test_bit(MD_RECOVERY_RUNNING, &rdev->mddev->recovery))
3191 return -EBUSY;
3193 if (rdev->mddev->pers->hot_add_disk == NULL)
3194 return -EINVAL;
3196 if (slot >= rdev->mddev->raid_disks &&
3197 slot >= rdev->mddev->raid_disks + rdev->mddev->delta_disks)
3198 return -ENOSPC;
3200 rdev->raid_disk = slot;
3201 if (test_bit(In_sync, &rdev->flags))
3202 rdev->saved_raid_disk = slot;
3204 rdev->saved_raid_disk = -1;
3205 clear_bit(In_sync, &rdev->flags);
3206 clear_bit(Bitmap_sync, &rdev->flags);
3207 err = rdev->mddev->pers->hot_add_disk(rdev->mddev, rdev);
3209 rdev->raid_disk = -1;
3212 sysfs_notify_dirent_safe(rdev->sysfs_state);
3214 sysfs_link_rdev(rdev->mddev, rdev);
3217 if (slot >= rdev->mddev->raid_disks &&
3218 slot >= rdev->mddev->raid_disks + rdev->mddev->delta_disks)
3219 return -ENOSPC;
3220 rdev->raid_disk = slot;
3222 clear_bit(Faulty, &rdev->flags);
3223 clear_bit(WriteMostly, &rdev->flags);
3224 set_bit(In_sync, &rdev->flags);
3225 sysfs_notify_dirent_safe(rdev->sysfs_state);
3236 return sprintf(page, "%llu\n", (unsigned long long)rdev->data_offset);
3244 return -EINVAL;
3245 if (rdev->mddev->pers && rdev->raid_disk >= 0)
3246 return -EBUSY;
3247 if (rdev->sectors && rdev->mddev->external)
3250 return -EBUSY;
3251 rdev->data_offset = offset;
3252 rdev->new_data_offset = offset;
3262 (unsigned long long)rdev->new_data_offset);
3269 struct mddev *mddev = rdev->mddev;
3272 return -EINVAL;
3274 if (mddev->sync_thread ||
3275 test_bit(MD_RECOVERY_RUNNING,&mddev->recovery))
3276 return -EBUSY;
3277 if (new_offset == rdev->data_offset)
3280 else if (new_offset > rdev->data_offset) {
3282 if (new_offset - rdev->data_offset
3283 + mddev->dev_sectors > rdev->sectors)
3284 return -E2BIG;
3291 if (new_offset < rdev->data_offset &&
3292 mddev->reshape_backwards)
3293 return -EINVAL;
3298 if (new_offset > rdev->data_offset &&
3299 !mddev->reshape_backwards)
3300 return -EINVAL;
3302 if (mddev->pers && mddev->persistent &&
3303 !super_types[mddev->major_version]
3305 return -E2BIG;
3306 rdev->new_data_offset = new_offset;
3307 if (new_offset > rdev->data_offset)
3308 mddev->reshape_backwards = 1;
3309 else if (new_offset < rdev->data_offset)
3310 mddev->reshape_backwards = 0;
3320 return sprintf(page, "%llu\n", (unsigned long long)rdev->sectors / 2);
3326 if (a->data_offset + a->sectors <= b->data_offset)
3328 if (b->data_offset + b->sectors <= a->data_offset)
3340 if (test_bit(MD_DELETED, &mddev->flags))
3343 if (rdev != rdev2 && rdev->bdev == rdev2->bdev &&
3360 return -EINVAL;
3362 if (blocks & 1ULL << (8 * sizeof(blocks) - 1))
3363 return -EINVAL; /* sector conversion overflow */
3367 return -EINVAL; /* unsigned long long to sector_t overflow */
3376 struct mddev *my_mddev = rdev->mddev;
3377 sector_t oldsectors = rdev->sectors;
3380 if (test_bit(Journal, &rdev->flags))
3381 return -EBUSY;
3383 return -EINVAL;
3384 if (rdev->data_offset != rdev->new_data_offset)
3385 return -EINVAL; /* too confusing */
3386 if (my_mddev->pers && rdev->raid_disk >= 0) {
3387 if (my_mddev->persistent) {
3388 sectors = super_types[my_mddev->major_version].
3391 return -EBUSY;
3393 sectors = bdev_nr_sectors(rdev->bdev) -
3394 rdev->data_offset;
3395 if (!my_mddev->pers->resize)
3397 return -EINVAL;
3399 if (sectors < my_mddev->dev_sectors)
3400 return -EINVAL; /* component must fit device */
3402 rdev->sectors = sectors;
3409 if (sectors > oldsectors && my_mddev->external &&
3416 rdev->sectors = oldsectors;
3417 return -EBUSY;
3427 unsigned long long recovery_start = rdev->recovery_offset;
3429 if (test_bit(In_sync, &rdev->flags) ||
3443 return -EINVAL;
3445 if (rdev->mddev->pers &&
3446 rdev->raid_disk >= 0)
3447 return -EBUSY;
3449 rdev->recovery_offset = recovery_start;
3451 set_bit(In_sync, &rdev->flags);
3453 clear_bit(In_sync, &rdev->flags);
3460 /* sysfs access to bad-blocks list.
3462 * 'bad-blocks' lists sector numbers and lengths of ranges that
3464 * the one-page limit of sysfs.
3467 * 'unacknowledged-bad-blocks' lists bad blocks that have not yet
3473 return badblocks_show(&rdev->badblocks, page, 0);
3477 int rv = badblocks_store(&rdev->badblocks, page, len, 0);
3479 if (test_and_clear_bit(BlockedBadBlocks, &rdev->flags))
3480 wake_up(&rdev->blocked_wait);
3488 return badblocks_show(&rdev->badblocks, page, 1);
3492 return badblocks_store(&rdev->badblocks, page, len, 1);
3500 return sprintf(page, "%llu\n", (unsigned long long)rdev->ppl.sector);
3509 return -EINVAL;
3511 return -EINVAL;
3513 if (rdev->mddev->pers && test_bit(MD_HAS_PPL, &rdev->mddev->flags) &&
3514 rdev->raid_disk >= 0)
3515 return -EBUSY;
3517 if (rdev->mddev->persistent) {
3518 if (rdev->mddev->major_version == 0)
3519 return -EINVAL;
3520 if ((sector > rdev->sb_start &&
3521 sector - rdev->sb_start > S16_MAX) ||
3522 (sector < rdev->sb_start &&
3523 rdev->sb_start - sector > -S16_MIN))
3524 return -EINVAL;
3525 rdev->ppl.offset = sector - rdev->sb_start;
3526 } else if (!rdev->mddev->external) {
3527 return -EBUSY;
3529 rdev->ppl.sector = sector;
3539 return sprintf(page, "%u\n", rdev->ppl.size);
3548 return -EINVAL;
3550 if (rdev->mddev->pers && test_bit(MD_HAS_PPL, &rdev->mddev->flags) &&
3551 rdev->raid_disk >= 0)
3552 return -EBUSY;
3554 if (rdev->mddev->persistent) {
3555 if (rdev->mddev->major_version == 0)
3556 return -EINVAL;
3558 return -EINVAL;
3559 } else if (!rdev->mddev->external) {
3560 return -EBUSY;
3562 rdev->ppl.size = size;
3590 if (!entry->show)
3591 return -EIO;
3592 if (!rdev->mddev)
3593 return -ENODEV;
3594 return entry->show(rdev, page);
3605 struct mddev *mddev = rdev->mddev;
3607 if (!entry->store)
3608 return -EIO;
3610 return -EACCES;
3612 if (entry->store == state_store && cmd_match(page, "remove"))
3615 rv = mddev ? mddev_lock(mddev) : -ENODEV;
3617 if (rdev->mddev == NULL)
3618 rv = -ENODEV;
3620 rv = entry->store(rdev, page, length);
3647 rdev->desc_nr = -1;
3648 rdev->saved_raid_disk = -1;
3649 rdev->raid_disk = -1;
3650 rdev->flags = 0;
3651 rdev->data_offset = 0;
3652 rdev->new_data_offset = 0;
3653 rdev->sb_events = 0;
3654 rdev->last_read_error = 0;
3655 rdev->sb_loaded = 0;
3656 rdev->bb_page = NULL;
3657 atomic_set(&rdev->nr_pending, 0);
3658 atomic_set(&rdev->read_errors, 0);
3659 atomic_set(&rdev->corrected_errors, 0);
3661 INIT_LIST_HEAD(&rdev->same_set);
3662 init_waitqueue_head(&rdev->blocked_wait);
3666 * be used - I wonder if that matters
3668 return badblocks_init(&rdev->badblocks, 0);
3677 * - the device is nonexistent (zero size)
3678 * - the device has no valid superblock
3680 * a faulty rdev _never_ has rdev->sb set.
3691 return ERR_PTR(-ENOMEM);
3700 if (super_format == -2) {
3704 set_bit(Holder, &rdev->flags);
3707 rdev->bdev = blkdev_get_by_dev(newdev, BLK_OPEN_READ | BLK_OPEN_WRITE,
3709 if (IS_ERR(rdev->bdev)) {
3710 pr_warn("md: could not open device unknown-block(%u,%u).\n",
3712 err = PTR_ERR(rdev->bdev);
3716 kobject_init(&rdev->kobj, &rdev_ktype);
3718 size = bdev_nr_bytes(rdev->bdev) >> BLOCK_SIZE_BITS;
3721 rdev->bdev);
3722 err = -EINVAL;
3729 if (err == -EINVAL) {
3731 rdev->bdev,
3737 rdev->bdev);
3745 blkdev_put(rdev->bdev, holder);
3764 switch (super_types[mddev->major_version].
3765 load_super(rdev, freshest, mddev->minor_version)) {
3772 pr_warn("md: fatal superblock inconsistency in %pg -- removing from array\n",
3773 rdev->bdev);
3780 return -EINVAL;
3783 super_types[mddev->major_version].
3788 if (mddev->max_disks &&
3789 (rdev->desc_nr >= mddev->max_disks ||
3790 i > mddev->max_disks)) {
3792 mdname(mddev), rdev->bdev,
3793 mddev->max_disks);
3798 if (super_types[mddev->major_version].
3800 pr_warn("md: kicking non-fresh %pg from array!\n",
3801 rdev->bdev);
3806 if (mddev->level == LEVEL_MULTIPATH) {
3807 rdev->desc_nr = i++;
3808 rdev->raid_disk = rdev->desc_nr;
3809 set_bit(In_sync, &rdev->flags);
3810 } else if (rdev->raid_disk >=
3811 (mddev->raid_disks - min(0, mddev->delta_disks)) &&
3812 !test_bit(Journal, &rdev->flags)) {
3813 rdev->raid_disk = -1;
3814 clear_bit(In_sync, &rdev->flags);
3821 /* Read a fixed-point number.
3829 * all without any floating-point arithmetic.
3834 long decimals = -1;
3840 value = *cp - '0';
3850 return -EINVAL;
3853 *res = result * int_pow(10, scale - decimals);
3860 unsigned int msec = ((unsigned long)mddev->safemode_delay*1000)/HZ;
3871 return -EINVAL;
3875 return -EINVAL;
3877 mddev->safemode_delay = 0;
3879 unsigned long old_delay = mddev->safemode_delay;
3884 mddev->safemode_delay = new_delay;
3886 mod_timer(&mddev->safemode_timer, jiffies+1);
3898 spin_lock(&mddev->lock);
3899 p = mddev->pers;
3901 ret = sprintf(page, "%s\n", p->name);
3902 else if (mddev->clevel[0])
3903 ret = sprintf(page, "%s\n", mddev->clevel);
3904 else if (mddev->level != LEVEL_NONE)
3905 ret = sprintf(page, "%d\n", mddev->level);
3908 spin_unlock(&mddev->lock);
3924 return -EINVAL;
3930 if (mddev->pers == NULL) {
3931 strncpy(mddev->clevel, buf, slen);
3932 if (mddev->clevel[slen-1] == '\n')
3933 slen--;
3934 mddev->clevel[slen] = 0;
3935 mddev->level = LEVEL_NONE;
3939 rv = -EROFS;
3944 * - array is not engaged in resync/recovery/reshape
3945 * - old personality can be suspended
3946 * - new personality will access other array.
3949 rv = -EBUSY;
3950 if (mddev->sync_thread ||
3951 test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
3952 mddev->reshape_position != MaxSector ||
3953 mddev->sysfs_active)
3956 rv = -EINVAL;
3957 if (!mddev->pers->quiesce) {
3959 mdname(mddev), mddev->pers->name);
3965 if (clevel[slen-1] == '\n')
3966 slen--;
3971 if (request_module("md-%s", clevel) != 0)
3972 request_module("md-level-%s", clevel);
3975 if (!pers || !try_module_get(pers->owner)) {
3978 rv = -EINVAL;
3983 if (pers == mddev->pers) {
3985 module_put(pers->owner);
3989 if (!pers->takeover) {
3990 module_put(pers->owner);
3993 rv = -EINVAL;
3998 rdev->new_raid_disk = rdev->raid_disk;
4000 /* ->takeover must set new_* and/or delta_disks
4003 priv = pers->takeover(mddev);
4005 mddev->new_level = mddev->level;
4006 mddev->new_layout = mddev->layout;
4007 mddev->new_chunk_sectors = mddev->chunk_sectors;
4008 mddev->raid_disks -= mddev->delta_disks;
4009 mddev->delta_disks = 0;
4010 mddev->reshape_backwards = 0;
4011 module_put(pers->owner);
4022 spin_lock(&mddev->lock);
4023 oldpers = mddev->pers;
4024 oldpriv = mddev->private;
4025 mddev->pers = pers;
4026 mddev->private = priv;
4027 strscpy(mddev->clevel, pers->name, sizeof(mddev->clevel));
4028 mddev->level = mddev->new_level;
4029 mddev->layout = mddev->new_layout;
4030 mddev->chunk_sectors = mddev->new_chunk_sectors;
4031 mddev->delta_disks = 0;
4032 mddev->reshape_backwards = 0;
4033 mddev->degraded = 0;
4034 spin_unlock(&mddev->lock);
4036 if (oldpers->sync_request == NULL &&
4037 mddev->external) {
4038 /* We are converting from a no-redundancy array
4042 * clean->dirty
4045 mddev->in_sync = 0;
4046 mddev->safemode_delay = 0;
4047 mddev->safemode = 0;
4050 oldpers->free(mddev, oldpriv);
4052 if (oldpers->sync_request == NULL &&
4053 pers->sync_request != NULL) {
4055 if (sysfs_create_group(&mddev->kobj, &md_redundancy_group))
4058 mddev->sysfs_action = sysfs_get_dirent(mddev->kobj.sd, "sync_action");
4059 mddev->sysfs_completed = sysfs_get_dirent_safe(mddev->kobj.sd, "sync_completed");
4060 mddev->sysfs_degraded = sysfs_get_dirent_safe(mddev->kobj.sd, "degraded");
4062 if (oldpers->sync_request != NULL &&
4063 pers->sync_request == NULL) {
4065 if (mddev->to_remove == NULL)
4066 mddev->to_remove = &md_redundancy_group;
4069 module_put(oldpers->owner);
4072 if (rdev->raid_disk < 0)
4074 if (rdev->new_raid_disk >= mddev->raid_disks)
4075 rdev->new_raid_disk = -1;
4076 if (rdev->new_raid_disk == rdev->raid_disk)
4081 if (rdev->raid_disk < 0)
4083 if (rdev->new_raid_disk == rdev->raid_disk)
4085 rdev->raid_disk = rdev->new_raid_disk;
4086 if (rdev->raid_disk < 0)
4087 clear_bit(In_sync, &rdev->flags);
4091 rdev->raid_disk, mdname(mddev));
4095 if (pers->sync_request == NULL) {
4099 mddev->in_sync = 1;
4100 del_timer_sync(&mddev->safemode_timer);
4102 blk_set_stacking_limits(&mddev->queue->limits);
4103 pers->run(mddev);
4104 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
4106 if (!mddev->thread)
4108 sysfs_notify_dirent_safe(mddev->sysfs_level);
4123 if (mddev->reshape_position != MaxSector &&
4124 mddev->layout != mddev->new_layout)
4126 mddev->new_layout, mddev->layout);
4127 return sprintf(page, "%d\n", mddev->layout);
4143 if (mddev->pers) {
4144 if (mddev->pers->check_reshape == NULL)
4145 err = -EBUSY;
4147 err = -EROFS;
4149 mddev->new_layout = n;
4150 err = mddev->pers->check_reshape(mddev);
4152 mddev->new_layout = mddev->layout;
4155 mddev->new_layout = n;
4156 if (mddev->reshape_position == MaxSector)
4157 mddev->layout = n;
4168 if (mddev->raid_disks == 0)
4170 if (mddev->reshape_position != MaxSector &&
4171 mddev->delta_disks != 0)
4172 return sprintf(page, "%d (%d)\n", mddev->raid_disks,
4173 mddev->raid_disks - mddev->delta_disks);
4174 return sprintf(page, "%d\n", mddev->raid_disks);
4192 if (mddev->pers)
4194 else if (mddev->reshape_position != MaxSector) {
4196 int olddisks = mddev->raid_disks - mddev->delta_disks;
4198 err = -EINVAL;
4201 rdev->data_offset < rdev->new_data_offset)
4204 rdev->data_offset > rdev->new_data_offset)
4208 mddev->delta_disks = n - olddisks;
4209 mddev->raid_disks = n;
4210 mddev->reshape_backwards = (mddev->delta_disks < 0);
4212 mddev->raid_disks = n;
4223 return sprintf(page, "%pU\n", mddev->uuid);
4231 if (mddev->reshape_position != MaxSector &&
4232 mddev->chunk_sectors != mddev->new_chunk_sectors)
4234 mddev->new_chunk_sectors << 9,
4235 mddev->chunk_sectors << 9);
4236 return sprintf(page, "%d\n", mddev->chunk_sectors << 9);
4252 if (mddev->pers) {
4253 if (mddev->pers->check_reshape == NULL)
4254 err = -EBUSY;
4256 err = -EROFS;
4258 mddev->new_chunk_sectors = n >> 9;
4259 err = mddev->pers->check_reshape(mddev);
4261 mddev->new_chunk_sectors = mddev->chunk_sectors;
4264 mddev->new_chunk_sectors = n >> 9;
4265 if (mddev->reshape_position == MaxSector)
4266 mddev->chunk_sectors = n >> 9;
4277 if (mddev->recovery_cp == MaxSector)
4279 return sprintf(page, "%llu\n", (unsigned long long)mddev->recovery_cp);
4295 return -EINVAL;
4301 if (mddev->pers && !test_bit(MD_RECOVERY_FROZEN, &mddev->recovery))
4302 err = -EBUSY;
4305 mddev->recovery_cp = n;
4306 if (mddev->pers)
4307 set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
4332 * read-auto
4335 * clean - no pending writes, but otherwise active.
4339 * if not known, block and switch to write-pending
4345 * write-pending
4348 * active-idle
4352 * Array is failed. It's useful because mounted-arrays aren't stopped
4359 "clear", "inactive", "suspended", "readonly", "read-auto", "clean", "active",
4360 "write-pending", "active-idle", "broken", NULL };
4376 if (mddev->pers && !test_bit(MD_NOT_READY, &mddev->flags)) {
4377 switch(mddev->ro) {
4385 spin_lock(&mddev->lock);
4386 if (test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags))
4388 else if (mddev->in_sync)
4390 else if (mddev->safemode)
4394 spin_unlock(&mddev->lock);
4397 if (test_bit(MD_BROKEN, &mddev->flags) && st == clean)
4400 if (list_empty(&mddev->disks) &&
4401 mddev->raid_disks == 0 &&
4402 mddev->dev_sectors == 0)
4420 if (mddev->pers && (st == active || st == clean) &&
4421 mddev->ro != MD_RDONLY) {
4425 spin_lock(&mddev->lock);
4428 clear_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
4429 md_wakeup_thread(mddev->thread);
4430 wake_up(&mddev->sb_wait);
4434 err = -EBUSY;
4437 sysfs_notify_dirent_safe(mddev->sysfs_state);
4438 spin_unlock(&mddev->lock);
4444 err = -EINVAL;
4454 if (mddev->pers)
4462 if (mddev->pers)
4465 mddev->ro = MD_RDONLY;
4466 set_disk_ro(mddev->gendisk, 1);
4471 if (mddev->pers) {
4474 else if (mddev->ro == MD_RDONLY)
4477 mddev->ro = MD_AUTO_READ;
4478 set_disk_ro(mddev->gendisk, 0);
4481 mddev->ro = MD_AUTO_READ;
4486 if (mddev->pers) {
4490 spin_lock(&mddev->lock);
4492 err = -EBUSY;
4493 spin_unlock(&mddev->lock);
4495 err = -EINVAL;
4498 if (mddev->pers) {
4502 clear_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
4503 wake_up(&mddev->sb_wait);
4506 mddev->ro = MD_RDWR;
4507 set_disk_ro(mddev->gendisk, 0);
4519 if (mddev->hold_active == UNTIL_IOCTL)
4520 mddev->hold_active = 0;
4521 sysfs_notify_dirent_safe(mddev->sysfs_state);
4532 atomic_read(&mddev->max_corr_read_errors));
4545 return -EINVAL;
4546 atomic_set(&mddev->max_corr_read_errors, n);
4557 return -EINVAL;
4578 return -EINVAL;
4581 return -EINVAL;
4585 return -EOVERFLOW;
4590 if (mddev->persistent) {
4591 rdev = md_import_device(dev, mddev->major_version,
4592 mddev->minor_version);
4593 if (!IS_ERR(rdev) && !list_empty(&mddev->disks)) {
4595 = list_entry(mddev->disks.next,
4597 err = super_types[mddev->major_version]
4598 .load_super(rdev, rdev0, mddev->minor_version);
4602 } else if (mddev->external)
4603 rdev = md_import_device(dev, -2, -1);
4605 rdev = md_import_device(dev, -1, -1);
4634 if (!mddev->bitmap)
4636 /* buf should be <chunk> <chunk> ... or <chunk>-<chunk> ... (range) */
4640 if (*end == '-') { /* range */
4646 md_bitmap_dirty_bits(mddev->bitmap, chunk, end_chunk);
4649 md_bitmap_unplug(mddev->bitmap); /* flush the bits to disk */
4662 (unsigned long long)mddev->dev_sectors / 2);
4672 * If array is active, we can try an on-line resize
4682 if (mddev->pers) {
4687 if (mddev->dev_sectors == 0 ||
4688 mddev->dev_sectors > sectors)
4689 mddev->dev_sectors = sectors;
4691 err = -ENOSPC;
4709 if (mddev->persistent)
4711 mddev->major_version, mddev->minor_version);
4712 else if (mddev->external)
4713 return sprintf(page, "external:%s\n", mddev->metadata_type);
4732 err = -EBUSY;
4733 if (mddev->external && strncmp(buf, "external:", 9) == 0)
4735 else if (!list_empty(&mddev->disks))
4740 mddev->persistent = 0;
4741 mddev->external = 0;
4742 mddev->major_version = 0;
4743 mddev->minor_version = 90;
4747 size_t namelen = len-9;
4748 if (namelen >= sizeof(mddev->metadata_type))
4749 namelen = sizeof(mddev->metadata_type)-1;
4750 strncpy(mddev->metadata_type, buf+9, namelen);
4751 mddev->metadata_type[namelen] = 0;
4752 if (namelen && mddev->metadata_type[namelen-1] == '\n')
4753 mddev->metadata_type[--namelen] = 0;
4754 mddev->persistent = 0;
4755 mddev->external = 1;
4756 mddev->major_version = 0;
4757 mddev->minor_version = 90;
4761 err = -EINVAL;
4768 err = -ENOENT;
4771 mddev->major_version = major;
4772 mddev->minor_version = minor;
4773 mddev->persistent = 1;
4774 mddev->external = 0;
4788 unsigned long recovery = mddev->recovery;
4804 else if (mddev->reshape_position != MaxSector)
4812 if (!test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
4822 if (!test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) {
4827 if (work_pending(&mddev->del_work))
4830 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
4835 md_wakeup_thread_directly(mddev->sync_thread);
4842 int sync_seq = atomic_read(&mddev->sync_seq);
4844 mutex_lock(&mddev->sync_mutex);
4845 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4848 wait_event(resync_wait, sync_seq != atomic_read(&mddev->sync_seq) ||
4849 !test_bit(MD_RECOVERY_RUNNING, &mddev->recovery));
4851 mutex_unlock(&mddev->sync_mutex);
4856 mutex_lock(&mddev->sync_mutex);
4857 set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4860 wait_event(resync_wait, mddev->sync_thread == NULL &&
4861 !test_bit(MD_RECOVERY_RUNNING, &mddev->recovery));
4863 mutex_unlock(&mddev->sync_mutex);
4869 if (!mddev->pers || !mddev->pers->sync_request)
4870 return -EINVAL;
4877 else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
4878 return -EBUSY;
4880 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4882 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4883 set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
4886 if (mddev->pers->start_reshape == NULL)
4887 return -EINVAL;
4890 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) {
4891 err = -EBUSY;
4892 } else if (mddev->reshape_position == MaxSector ||
4893 mddev->pers->check_reshape == NULL ||
4894 mddev->pers->check_reshape(mddev)) {
4895 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4896 err = mddev->pers->start_reshape(mddev);
4904 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4910 sysfs_notify_dirent_safe(mddev->sysfs_degraded);
4913 set_bit(MD_RECOVERY_CHECK, &mddev->recovery);
4915 return -EINVAL;
4916 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4917 set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
4918 set_bit(MD_RECOVERY_SYNC, &mddev->recovery);
4920 if (mddev->ro == MD_AUTO_READ) {
4922 * canceling read-auto mode
4924 mddev->ro = MD_RDWR;
4925 md_wakeup_thread(mddev->sync_thread);
4927 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
4928 md_wakeup_thread(mddev->thread);
4929 sysfs_notify_dirent_safe(mddev->sysfs_action);
4939 return sprintf(page, "%s\n", mddev->last_sync_action);
4949 atomic64_read(&mddev->resync_mismatches));
4958 mddev->sync_speed_min ? "local": "system");
4974 return -EINVAL;
4976 mddev->sync_speed_min = min;
4987 mddev->sync_speed_max ? "local": "system");
5003 return -EINVAL;
5005 mddev->sync_speed_max = max;
5015 return sprintf(page, "%d\n", mddev->degraded);
5022 return sprintf(page, "%d\n", mddev->parallel_resync);
5031 return -EINVAL;
5034 return -EINVAL;
5036 mddev->parallel_resync = n;
5038 if (mddev->sync_thread)
5053 if (mddev->curr_resync == MD_RESYNC_NONE)
5055 resync = mddev->curr_mark_cnt - atomic_read(&mddev->recovery_active);
5056 dt = (jiffies - mddev->resync_mark) / HZ;
5058 db = resync - mddev->resync_mark_cnt;
5069 if (!test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
5072 if (mddev->curr_resync == MD_RESYNC_YIELDED ||
5073 mddev->curr_resync == MD_RESYNC_DELAYED)
5076 if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ||
5077 test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
5078 max_sectors = mddev->resync_max_sectors;
5080 max_sectors = mddev->dev_sectors;
5082 resync = mddev->curr_resync_completed;
5093 (unsigned long long)mddev->resync_min);
5102 return -EINVAL;
5104 spin_lock(&mddev->lock);
5105 err = -EINVAL;
5106 if (min > mddev->resync_max)
5109 err = -EBUSY;
5110 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
5114 mddev->resync_min = round_down(min, 8);
5118 spin_unlock(&mddev->lock);
5128 if (mddev->resync_max == MaxSector)
5132 (unsigned long long)mddev->resync_max);
5138 spin_lock(&mddev->lock);
5140 mddev->resync_max = MaxSector;
5145 err = -EINVAL;
5148 if (max < mddev->resync_min)
5151 err = -EBUSY;
5152 if (max < mddev->resync_max && md_is_rdwr(mddev) &&
5153 test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
5157 chunk = mddev->chunk_sectors;
5161 err = -EINVAL;
5165 mddev->resync_max = max;
5167 wake_up(&mddev->recovery_wait);
5170 spin_unlock(&mddev->lock);
5180 return sprintf(page, "%llu\n", (unsigned long long)mddev->suspend_lo);
5193 return -EINVAL;
5198 err = -EINVAL;
5199 if (mddev->pers == NULL ||
5200 mddev->pers->quiesce == NULL)
5203 mddev->suspend_lo = new;
5217 return sprintf(page, "%llu\n", (unsigned long long)mddev->suspend_hi);
5230 return -EINVAL;
5235 err = -EINVAL;
5236 if (mddev->pers == NULL)
5240 mddev->suspend_hi = new;
5254 if (mddev->reshape_position != MaxSector)
5256 (unsigned long long)mddev->reshape_position);
5272 return -EINVAL;
5276 err = -EBUSY;
5277 if (mddev->pers)
5279 mddev->reshape_position = new;
5280 mddev->delta_disks = 0;
5281 mddev->reshape_backwards = 0;
5282 mddev->new_level = mddev->level;
5283 mddev->new_layout = mddev->layout;
5284 mddev->new_chunk_sectors = mddev->chunk_sectors;
5286 rdev->new_data_offset = rdev->data_offset;
5301 mddev->reshape_backwards ? "backwards" : "forwards");
5315 return -EINVAL;
5316 if (mddev->reshape_backwards == backwards)
5323 if (mddev->delta_disks)
5324 err = -EBUSY;
5325 else if (mddev->persistent &&
5326 mddev->major_version == 0)
5327 err = -EINVAL;
5329 mddev->reshape_backwards = backwards;
5341 if (mddev->external_size)
5343 (unsigned long long)mddev->array_sectors/2);
5361 return -EINVAL;
5365 if (mddev->pers)
5366 sectors = mddev->pers->size(mddev, 0, 0);
5368 sectors = mddev->array_sectors;
5370 mddev->external_size = 0;
5373 err = -EINVAL;
5374 else if (mddev->pers && mddev->pers->size(mddev, 0, 0) < sectors)
5375 err = -E2BIG;
5377 mddev->external_size = 1;
5381 mddev->array_sectors = sectors;
5382 if (mddev->pers)
5383 set_capacity_and_notify(mddev->gendisk,
5384 mddev->array_sectors);
5399 if (test_bit(MD_HAS_JOURNAL, &mddev->flags)) {
5401 } else if (test_bit(MD_HAS_PPL, &mddev->flags)) {
5403 } else if (mddev->bitmap) {
5405 } else if (mddev->pers) {
5406 if (mddev->pers->sync_request)
5422 if (mddev->pers) {
5423 if (mddev->pers->change_consistency_policy)
5424 err = mddev->pers->change_consistency_policy(mddev, buf);
5426 err = -EBUSY;
5427 } else if (mddev->external && strncmp(buf, "ppl", 3) == 0) {
5428 set_bit(MD_HAS_PPL, &mddev->flags);
5430 err = -EINVAL;
5442 return sprintf(page, "%d\n", mddev->fail_last_dev);
5459 if (value != mddev->fail_last_dev)
5460 mddev->fail_last_dev = value;
5470 if (mddev->pers == NULL || (mddev->pers->level != 1))
5473 return sprintf(page, "%d\n", mddev->serialize_policy);
5490 if (value == mddev->serialize_policy)
5496 if (mddev->pers == NULL || (mddev->pers->level != 1)) {
5498 err = -EINVAL;
5507 mddev->serialize_policy = value;
5580 if (!entry->show)
5581 return -EIO;
5585 return -EBUSY;
5589 rv = entry->show(mddev, page);
5602 if (!entry->store)
5603 return -EIO;
5605 return -EACCES;
5609 return -EBUSY;
5612 rv = entry->store(mddev, page, length);
5621 if (mddev->sysfs_state)
5622 sysfs_put(mddev->sysfs_state);
5623 if (mddev->sysfs_level)
5624 sysfs_put(mddev->sysfs_level);
5626 del_gendisk(mddev->gendisk);
5627 put_disk(mddev->gendisk);
5646 kobject_put(&mddev->kobj);
5653 if (mddev->writes_pending.percpu_count_ptr)
5655 if (percpu_ref_init(&mddev->writes_pending, no_op,
5657 return -ENOMEM;
5659 percpu_ref_put(&mddev->writes_pending);
5669 * If dev is non-zero it must be a device number with a MAJOR of
5696 partitioned = (MAJOR(mddev->unit) != MD_MAJOR);
5698 unit = MINOR(mddev->unit) >> shift;
5707 if (mddev2->gendisk &&
5708 strcmp(mddev2->gendisk->disk_name, name) == 0) {
5710 error = -EEXIST;
5719 mddev->hold_active = UNTIL_STOP;
5721 error = -ENOMEM;
5726 disk->major = MAJOR(mddev->unit);
5727 disk->first_minor = unit << shift;
5728 disk->minors = 1 << shift;
5730 strcpy(disk->disk_name, name);
5732 sprintf(disk->disk_name, "md_d%d", unit);
5734 sprintf(disk->disk_name, "md%d", unit);
5735 disk->fops = &md_fops;
5736 disk->private_data = mddev;
5738 mddev->queue = disk->queue;
5739 blk_set_stacking_limits(&mddev->queue->limits);
5740 blk_queue_write_cache(mddev->queue, true, true);
5741 disk->events |= DISK_EVENT_MEDIA_CHANGE;
5742 mddev->gendisk = disk;
5747 kobject_init(&mddev->kobj, &md_ktype);
5748 error = kobject_add(&mddev->kobj, &disk_to_dev(disk)->kobj, "%s", "md");
5755 mddev->hold_active = 0;
5761 kobject_uevent(&mddev->kobj, KOBJ_ADD);
5762 mddev->sysfs_state = sysfs_get_dirent_safe(mddev->kobj.sd, "array_state");
5763 mddev->sysfs_level = sysfs_get_dirent_safe(mddev->kobj.sd, "level");
5807 while (len && val[len-1] == '\n')
5808 len--;
5810 return -E2BIG;
5820 return -EINVAL;
5827 mddev->safemode = 1;
5828 if (mddev->external)
5829 sysfs_notify_dirent_safe(mddev->sysfs_state);
5831 md_wakeup_thread(mddev->thread);
5839 wake_up(&mddev->sb_wait);
5849 if (list_empty(&mddev->disks))
5851 return -EINVAL;
5853 if (mddev->pers)
5854 return -EBUSY;
5856 if (mddev->sysfs_active)
5857 return -EBUSY;
5862 if (!mddev->raid_disks) {
5863 if (!mddev->persistent)
5864 return -EINVAL;
5867 return -EINVAL;
5870 if (mddev->level != LEVEL_NONE)
5871 request_module("md-level-%d", mddev->level);
5872 else if (mddev->clevel[0])
5873 request_module("md-%s", mddev->clevel);
5880 mddev->has_superblocks = false;
5882 if (test_bit(Faulty, &rdev->flags))
5884 sync_blockdev(rdev->bdev);
5885 invalidate_bdev(rdev->bdev);
5886 if (mddev->ro != MD_RDONLY && rdev_read_only(rdev)) {
5887 mddev->ro = MD_RDONLY;
5888 if (mddev->gendisk)
5889 set_disk_ro(mddev->gendisk, 1);
5892 if (rdev->sb_page)
5893 mddev->has_superblocks = true;
5899 if (rdev->meta_bdev) {
5901 } else if (rdev->data_offset < rdev->sb_start) {
5902 if (mddev->dev_sectors &&
5903 rdev->data_offset + mddev->dev_sectors
5904 > rdev->sb_start) {
5907 return -EINVAL;
5910 if (rdev->sb_start + rdev->sb_size/512
5911 > rdev->data_offset) {
5914 return -EINVAL;
5917 sysfs_notify_dirent_safe(rdev->sysfs_state);
5918 nowait = nowait && bdev_nowait(rdev->bdev);
5921 err = percpu_ref_init(&mddev->active_io, active_io_release,
5926 if (!bioset_initialized(&mddev->bio_set)) {
5927 err = bioset_init(&mddev->bio_set, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS);
5931 if (!bioset_initialized(&mddev->sync_set)) {
5932 err = bioset_init(&mddev->sync_set, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS);
5937 if (!bioset_initialized(&mddev->io_clone_set)) {
5938 err = bioset_init(&mddev->io_clone_set, BIO_POOL_SIZE,
5945 pers = find_pers(mddev->level, mddev->clevel);
5946 if (!pers || !try_module_get(pers->owner)) {
5948 if (mddev->level != LEVEL_NONE)
5950 mddev->level);
5953 mddev->clevel);
5954 err = -EINVAL;
5958 if (mddev->level != pers->level) {
5959 mddev->level = pers->level;
5960 mddev->new_level = pers->level;
5962 strscpy(mddev->clevel, pers->name, sizeof(mddev->clevel));
5964 if (mddev->reshape_position != MaxSector &&
5965 pers->start_reshape == NULL) {
5967 module_put(pers->owner);
5968 err = -EINVAL;
5972 if (pers->sync_request) {
5982 rdev->bdev->bd_disk ==
5983 rdev2->bdev->bd_disk) {
5986 rdev->bdev,
5987 rdev2->bdev);
5993 pr_warn("True protection against single-disk failure might be compromised.\n");
5996 mddev->recovery = 0;
5997 /* may be over-ridden by personality */
5998 mddev->resync_max_sectors = mddev->dev_sectors;
6000 mddev->ok_start_degraded = start_dirty_degraded;
6003 mddev->ro = MD_AUTO_READ; /* read-only, but switch on first write */
6005 err = pers->run(mddev);
6007 pr_warn("md: pers->run() failed ...\n");
6008 else if (pers->size(mddev, 0, 0) < mddev->array_sectors) {
6009 WARN_ONCE(!mddev->external_size,
6013 (unsigned long long)mddev->array_sectors / 2,
6014 (unsigned long long)pers->size(mddev, 0, 0) / 2);
6015 err = -EINVAL;
6017 if (err == 0 && pers->sync_request &&
6018 (mddev->bitmap_info.file || mddev->bitmap_info.offset)) {
6021 bitmap = md_bitmap_create(mddev, -1);
6027 mddev->bitmap = bitmap;
6033 if (mddev->bitmap_info.max_write_behind > 0) {
6037 if (test_bit(WriteMostly, &rdev->flags) &&
6041 if (create_pool && mddev->serial_info_pool == NULL) {
6042 mddev->serial_info_pool =
6045 if (!mddev->serial_info_pool) {
6046 err = -ENOMEM;
6052 if (mddev->queue) {
6056 if (rdev->raid_disk >= 0 && !bdev_nonrot(rdev->bdev)) {
6061 if (mddev->degraded)
6064 blk_queue_flag_set(QUEUE_FLAG_NONROT, mddev->queue);
6066 blk_queue_flag_clear(QUEUE_FLAG_NONROT, mddev->queue);
6067 blk_queue_flag_set(QUEUE_FLAG_IO_STAT, mddev->queue);
6071 blk_queue_flag_set(QUEUE_FLAG_NOWAIT, mddev->queue);
6073 if (pers->sync_request) {
6074 if (mddev->kobj.sd &&
6075 sysfs_create_group(&mddev->kobj, &md_redundancy_group))
6078 mddev->sysfs_action = sysfs_get_dirent_safe(mddev->kobj.sd, "sync_action");
6079 mddev->sysfs_completed = sysfs_get_dirent_safe(mddev->kobj.sd, "sync_completed");
6080 mddev->sysfs_degraded = sysfs_get_dirent_safe(mddev->kobj.sd, "degraded");
6081 } else if (mddev->ro == MD_AUTO_READ)
6082 mddev->ro = MD_RDWR;
6084 atomic_set(&mddev->max_corr_read_errors,
6086 mddev->safemode = 0;
6088 mddev->safemode_delay = 0;
6090 mddev->safemode_delay = DEFAULT_SAFEMODE_DELAY;
6091 mddev->in_sync = 1;
6093 spin_lock(&mddev->lock);
6094 mddev->pers = pers;
6095 spin_unlock(&mddev->lock);
6097 if (rdev->raid_disk >= 0)
6100 if (mddev->degraded && md_is_rdwr(mddev))
6102 * via sysfs - until a lack of spares is confirmed.
6104 set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
6105 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
6107 if (mddev->sb_flags)
6115 if (mddev->private)
6116 pers->free(mddev, mddev->private);
6117 mddev->private = NULL;
6118 module_put(pers->owner);
6121 bioset_exit(&mddev->io_clone_set);
6123 bioset_exit(&mddev->sync_set);
6125 bioset_exit(&mddev->bio_set);
6127 percpu_ref_exit(&mddev->active_io);
6136 set_bit(MD_NOT_READY, &mddev->flags);
6152 md_wakeup_thread(mddev->thread);
6153 md_wakeup_thread(mddev->sync_thread); /* possibly kick off a reshape */
6155 set_capacity_and_notify(mddev->gendisk, mddev->array_sectors);
6156 clear_bit(MD_NOT_READY, &mddev->flags);
6157 mddev->changed = 1;
6158 kobject_uevent(&disk_to_dev(mddev->gendisk)->kobj, KOBJ_CHANGE);
6159 sysfs_notify_dirent_safe(mddev->sysfs_state);
6160 sysfs_notify_dirent_safe(mddev->sysfs_action);
6161 sysfs_notify_dirent_safe(mddev->sysfs_degraded);
6163 clear_bit(MD_NOT_READY, &mddev->flags);
6171 if (mddev->pers->start) {
6172 set_bit(MD_RECOVERY_WAIT, &mddev->recovery);
6173 md_wakeup_thread(mddev->thread);
6174 ret = mddev->pers->start(mddev);
6175 clear_bit(MD_RECOVERY_WAIT, &mddev->recovery);
6176 md_wakeup_thread(mddev->sync_thread);
6184 struct gendisk *disk = mddev->gendisk;
6190 if (list_empty(&mddev->disks))
6191 return -ENXIO;
6192 if (!mddev->pers)
6193 return -EINVAL;
6195 return -EBUSY;
6199 if (test_bit(Journal, &rdev->flags) &&
6200 !test_bit(Faulty, &rdev->flags))
6206 if (test_bit(MD_HAS_JOURNAL, &mddev->flags) && !has_journal)
6208 return -EINVAL;
6210 return -EROFS;
6212 mddev->safemode = 0;
6213 mddev->ro = MD_RDWR;
6215 pr_debug("md: %s switched to read-write mode.\n", mdname(mddev));
6217 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
6218 md_wakeup_thread(mddev->thread);
6219 md_wakeup_thread(mddev->sync_thread);
6220 sysfs_notify_dirent_safe(mddev->sysfs_state);
6226 mddev->array_sectors = 0;
6227 mddev->external_size = 0;
6228 mddev->dev_sectors = 0;
6229 mddev->raid_disks = 0;
6230 mddev->recovery_cp = 0;
6231 mddev->resync_min = 0;
6232 mddev->resync_max = MaxSector;
6233 mddev->reshape_position = MaxSector;
6234 /* we still need mddev->external in export_rdev, do not clear it yet */
6235 mddev->persistent = 0;
6236 mddev->level = LEVEL_NONE;
6237 mddev->clevel[0] = 0;
6243 if (mddev->hold_active)
6244 mddev->flags = 0;
6246 mddev->flags &= BIT_ULL_MASK(MD_CLOSING);
6247 mddev->sb_flags = 0;
6248 mddev->ro = MD_RDWR;
6249 mddev->metadata_type[0] = 0;
6250 mddev->chunk_sectors = 0;
6251 mddev->ctime = mddev->utime = 0;
6252 mddev->layout = 0;
6253 mddev->max_disks = 0;
6254 mddev->events = 0;
6255 mddev->can_decrease_events = 0;
6256 mddev->delta_disks = 0;
6257 mddev->reshape_backwards = 0;
6258 mddev->new_level = LEVEL_NONE;
6259 mddev->new_layout = 0;
6260 mddev->new_chunk_sectors = 0;
6261 mddev->curr_resync = MD_RESYNC_NONE;
6262 atomic64_set(&mddev->resync_mismatches, 0);
6263 mddev->suspend_lo = mddev->suspend_hi = 0;
6264 mddev->sync_speed_min = mddev->sync_speed_max = 0;
6265 mddev->recovery = 0;
6266 mddev->in_sync = 0;
6267 mddev->changed = 0;
6268 mddev->degraded = 0;
6269 mddev->safemode = 0;
6270 mddev->private = NULL;
6271 mddev->cluster_info = NULL;
6272 mddev->bitmap_info.offset = 0;
6273 mddev->bitmap_info.default_offset = 0;
6274 mddev->bitmap_info.default_space = 0;
6275 mddev->bitmap_info.chunksize = 0;
6276 mddev->bitmap_info.daemon_sleep = 0;
6277 mddev->bitmap_info.max_write_behind = 0;
6278 mddev->bitmap_info.nodes = 0;
6283 set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
6284 if (work_pending(&mddev->del_work))
6286 if (mddev->sync_thread) {
6287 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
6291 del_timer_sync(&mddev->safemode_timer);
6293 if (mddev->pers && mddev->pers->quiesce) {
6294 mddev->pers->quiesce(mddev, 1);
6295 mddev->pers->quiesce(mddev, 0);
6300 ((!mddev->in_sync && !mddev_is_clustered(mddev)) ||
6301 mddev->sb_flags)) {
6304 mddev->in_sync = 1;
6308 mddev->serialize_policy = 0;
6323 if (mddev->pers && mddev->pers->quiesce && !is_md_suspended(mddev)) {
6324 mddev->pers->quiesce(mddev, 1);
6325 mddev->pers->quiesce(mddev, 0);
6327 md_unregister_thread(mddev, &mddev->thread);
6328 if (mddev->queue)
6329 blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
6334 struct md_personality *pers = mddev->pers;
6337 /* Ensure ->event_work is done */
6338 if (mddev->event_work.func)
6340 spin_lock(&mddev->lock);
6341 mddev->pers = NULL;
6342 spin_unlock(&mddev->lock);
6343 if (mddev->private)
6344 pers->free(mddev, mddev->private);
6345 mddev->private = NULL;
6346 if (pers->sync_request && mddev->to_remove == NULL)
6347 mddev->to_remove = &md_redundancy_group;
6348 module_put(pers->owner);
6349 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
6351 percpu_ref_exit(&mddev->active_io);
6352 bioset_exit(&mddev->bio_set);
6353 bioset_exit(&mddev->sync_set);
6354 bioset_exit(&mddev->io_clone_set);
6359 lockdep_assert_held(&mddev->reconfig_mutex);
6362 * This is called from dm-raid
6366 percpu_ref_exit(&mddev->writes_pending);
6376 if (mddev->external && test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags))
6377 return -EBUSY;
6379 if (!test_bit(MD_RECOVERY_FROZEN, &mddev->recovery)) {
6381 set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
6382 md_wakeup_thread(mddev->thread);
6384 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
6385 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
6391 md_wakeup_thread_directly(mddev->sync_thread);
6395 &mddev->recovery));
6396 wait_event(mddev->sb_wait,
6397 !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags));
6400 mutex_lock(&mddev->open_mutex);
6401 if ((mddev->pers && atomic_read(&mddev->openers) > !!bdev) ||
6402 mddev->sync_thread ||
6403 test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) {
6405 err = -EBUSY;
6409 if (mddev->pers) {
6412 if (mddev->ro == MD_RDONLY) {
6413 err = -ENXIO;
6417 mddev->ro = MD_RDONLY;
6418 set_disk_ro(mddev->gendisk, 1);
6422 if ((mddev->pers && !err) || did_freeze) {
6423 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
6424 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
6425 md_wakeup_thread(mddev->thread);
6426 sysfs_notify_dirent_safe(mddev->sysfs_state);
6429 mutex_unlock(&mddev->open_mutex);
6434 * 0 - completely stop and dis-assemble array
6435 * 2 - stop but do not disassemble array
6440 struct gendisk *disk = mddev->gendisk;
6444 if (!test_bit(MD_RECOVERY_FROZEN, &mddev->recovery)) {
6446 set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
6447 md_wakeup_thread(mddev->thread);
6449 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
6450 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
6456 md_wakeup_thread_directly(mddev->sync_thread);
6459 wait_event(resync_wait, (mddev->sync_thread == NULL &&
6461 &mddev->recovery)));
6464 mutex_lock(&mddev->open_mutex);
6465 if ((mddev->pers && atomic_read(&mddev->openers) > !!bdev) ||
6466 mddev->sysfs_active ||
6467 mddev->sync_thread ||
6468 test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) {
6470 mutex_unlock(&mddev->open_mutex);
6472 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
6473 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
6474 md_wakeup_thread(mddev->thread);
6476 return -EBUSY;
6478 if (mddev->pers) {
6486 sysfs_notify_dirent_safe(mddev->sysfs_state);
6489 if (rdev->raid_disk >= 0)
6493 mutex_unlock(&mddev->open_mutex);
6494 mddev->changed = 1;
6497 mddev->ro = MD_RDWR;
6499 mutex_unlock(&mddev->open_mutex);
6506 if (mddev->bitmap_info.file) {
6507 struct file *f = mddev->bitmap_info.file;
6508 spin_lock(&mddev->lock);
6509 mddev->bitmap_info.file = NULL;
6510 spin_unlock(&mddev->lock);
6513 mddev->bitmap_info.offset = 0;
6518 if (mddev->hold_active == UNTIL_STOP)
6519 mddev->hold_active = 0;
6522 sysfs_notify_dirent_safe(mddev->sysfs_state);
6532 if (list_empty(&mddev->disks))
6538 pr_cont("<%pg>", rdev->bdev);
6574 pr_debug("md: considering %pg ...\n", rdev0->bdev);
6579 rdev->bdev);
6580 list_move(&rdev->same_set, &candidates);
6589 rdev0->preferred_minor << MdpMinorShift);
6592 dev = MKDEV(MD_MAJOR, rdev0->preferred_minor);
6595 if (rdev0->preferred_minor != unit) {
6597 rdev0->bdev, rdev0->preferred_minor);
6607 else if (mddev->raid_disks || mddev->major_version
6608 || !list_empty(&mddev->disks)) {
6610 mdname(mddev), rdev0->bdev);
6614 mddev->persistent = 1;
6616 list_del_init(&rdev->same_set);
6627 list_del_init(&rdev->same_set);
6645 return -EFAULT;
6660 if (test_bit(Faulty, &rdev->flags))
6664 if (test_bit(In_sync, &rdev->flags))
6666 else if (test_bit(Journal, &rdev->flags))
6675 info.major_version = mddev->major_version;
6676 info.minor_version = mddev->minor_version;
6678 info.ctime = clamp_t(time64_t, mddev->ctime, 0, U32_MAX);
6679 info.level = mddev->level;
6680 info.size = mddev->dev_sectors / 2;
6681 if (info.size != mddev->dev_sectors / 2) /* overflow */
6682 info.size = -1;
6684 info.raid_disks = mddev->raid_disks;
6685 info.md_minor = mddev->md_minor;
6686 info.not_persistent= !mddev->persistent;
6688 info.utime = clamp_t(time64_t, mddev->utime, 0, U32_MAX);
6690 if (mddev->in_sync)
6692 if (mddev->bitmap && mddev->bitmap_info.offset)
6701 info.layout = mddev->layout;
6702 info.chunk_size = mddev->chunk_sectors << 9;
6705 return -EFAULT;
6718 return -ENOMEM;
6721 spin_lock(&mddev->lock);
6723 if (mddev->bitmap_info.file) {
6724 ptr = file_path(mddev->bitmap_info.file, file->pathname,
6725 sizeof(file->pathname));
6729 memmove(file->pathname, ptr,
6730 sizeof(file->pathname)-(ptr-file->pathname));
6732 spin_unlock(&mddev->lock);
6736 err = -EFAULT;
6748 return -EFAULT;
6753 info.major = MAJOR(rdev->bdev->bd_dev);
6754 info.minor = MINOR(rdev->bdev->bd_dev);
6755 info.raid_disk = rdev->raid_disk;
6757 if (test_bit(Faulty, &rdev->flags))
6759 else if (test_bit(In_sync, &rdev->flags)) {
6763 if (test_bit(Journal, &rdev->flags))
6765 if (test_bit(WriteMostly, &rdev->flags))
6767 if (test_bit(FailFast, &rdev->flags))
6771 info.raid_disk = -1;
6777 return -EFAULT;
6785 dev_t dev = MKDEV(info->major,info->minor);
6788 !(info->state & ((1 << MD_DISK_CLUSTER_ADD) | (1 << MD_DISK_CANDIDATE)))) {
6791 return -EINVAL;
6794 if (info->major != MAJOR(dev) || info->minor != MINOR(dev))
6795 return -EOVERFLOW;
6797 if (!mddev->raid_disks) {
6800 rdev = md_import_device(dev, mddev->major_version, mddev->minor_version);
6806 if (!list_empty(&mddev->disks)) {
6808 = list_entry(mddev->disks.next,
6810 err = super_types[mddev->major_version]
6811 .load_super(rdev, rdev0, mddev->minor_version);
6814 rdev->bdev,
6815 rdev0->bdev);
6817 return -EINVAL;
6831 if (mddev->pers) {
6833 if (!mddev->pers->hot_add_disk) {
6836 return -EINVAL;
6838 if (mddev->persistent)
6839 rdev = md_import_device(dev, mddev->major_version,
6840 mddev->minor_version);
6842 rdev = md_import_device(dev, -1, -1);
6849 if (!mddev->persistent) {
6850 if (info->state & (1<<MD_DISK_SYNC) &&
6851 info->raid_disk < mddev->raid_disks) {
6852 rdev->raid_disk = info->raid_disk;
6853 clear_bit(Bitmap_sync, &rdev->flags);
6855 rdev->raid_disk = -1;
6856 rdev->saved_raid_disk = rdev->raid_disk;
6858 super_types[mddev->major_version].
6860 if ((info->state & (1<<MD_DISK_SYNC)) &&
6861 rdev->raid_disk != info->raid_disk) {
6862 /* This was a hot-add request, but events doesn't
6866 return -EINVAL;
6869 clear_bit(In_sync, &rdev->flags); /* just to be sure */
6870 if (info->state & (1<<MD_DISK_WRITEMOSTLY))
6871 set_bit(WriteMostly, &rdev->flags);
6873 clear_bit(WriteMostly, &rdev->flags);
6874 if (info->state & (1<<MD_DISK_FAILFAST))
6875 set_bit(FailFast, &rdev->flags);
6877 clear_bit(FailFast, &rdev->flags);
6879 if (info->state & (1<<MD_DISK_JOURNAL)) {
6885 if (test_bit(Journal, &rdev2->flags)) {
6890 if (has_journal || mddev->bitmap) {
6892 return -EBUSY;
6894 set_bit(Journal, &rdev->flags);
6900 if (info->state & (1 << MD_DISK_CANDIDATE))
6901 set_bit(Candidate, &rdev->flags);
6902 else if (info->state & (1 << MD_DISK_CLUSTER_ADD)) {
6903 /* --add initiated by this node */
6904 err = md_cluster_ops->add_new_disk(mddev, rdev);
6912 rdev->raid_disk = -1;
6919 if (info->state & (1 << MD_DISK_CANDIDATE)) {
6921 err = md_cluster_ops->new_disk_ack(mddev,
6928 md_cluster_ops->add_new_disk_cancel(mddev);
6942 if (mddev->major_version != 0) {
6944 return -EINVAL;
6947 if (!(info->state & (1<<MD_DISK_FAULTY))) {
6949 rdev = md_import_device(dev, -1, 0);
6955 rdev->desc_nr = info->number;
6956 if (info->raid_disk < mddev->raid_disks)
6957 rdev->raid_disk = info->raid_disk;
6959 rdev->raid_disk = -1;
6961 if (rdev->raid_disk < mddev->raid_disks)
6962 if (info->state & (1<<MD_DISK_SYNC))
6963 set_bit(In_sync, &rdev->flags);
6965 if (info->state & (1<<MD_DISK_WRITEMOSTLY))
6966 set_bit(WriteMostly, &rdev->flags);
6967 if (info->state & (1<<MD_DISK_FAILFAST))
6968 set_bit(FailFast, &rdev->flags);
6970 if (!mddev->persistent) {
6972 rdev->sb_start = bdev_nr_sectors(rdev->bdev);
6974 rdev->sb_start = calc_dev_sboffset(rdev);
6975 rdev->sectors = rdev->sb_start;
6991 if (!mddev->pers)
6992 return -ENODEV;
6996 return -ENXIO;
6998 if (rdev->raid_disk < 0)
7001 clear_bit(Blocked, &rdev->flags);
7004 if (rdev->raid_disk >= 0)
7009 if (md_cluster_ops->remove_disk(mddev, rdev))
7014 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
7015 if (mddev->thread)
7016 md_wakeup_thread(mddev->thread);
7024 rdev->bdev, mdname(mddev));
7025 return -EBUSY;
7033 if (!mddev->pers)
7034 return -ENODEV;
7036 if (mddev->major_version != 0) {
7037 pr_warn("%s: HOT_ADD may only be used with version-0 superblocks.\n",
7039 return -EINVAL;
7041 if (!mddev->pers->hot_add_disk) {
7044 return -EINVAL;
7047 rdev = md_import_device(dev, -1, 0);
7051 return -EINVAL;
7054 if (mddev->persistent)
7055 rdev->sb_start = calc_dev_sboffset(rdev);
7057 rdev->sb_start = bdev_nr_sectors(rdev->bdev);
7059 rdev->sectors = rdev->sb_start;
7061 if (test_bit(Faulty, &rdev->flags)) {
7062 pr_warn("md: can not hot-add faulty %pg disk to %s!\n",
7063 rdev->bdev, mdname(mddev));
7064 err = -EINVAL;
7068 clear_bit(In_sync, &rdev->flags);
7069 rdev->desc_nr = -1;
7070 rdev->saved_raid_disk = -1;
7080 rdev->raid_disk = -1;
7082 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
7083 if (!mddev->thread)
7089 if (!bdev_nowait(rdev->bdev)) {
7091 mdname(mddev), rdev->bdev);
7092 blk_queue_flag_clear(QUEUE_FLAG_NOWAIT, mddev->queue);
7098 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
7099 md_wakeup_thread(mddev->thread);
7112 if (mddev->pers) {
7113 if (!mddev->pers->quiesce || !mddev->thread)
7114 return -EBUSY;
7115 if (mddev->recovery || mddev->sync_thread)
7116 return -EBUSY;
7124 if (mddev->bitmap || mddev->bitmap_info.file)
7125 return -EEXIST; /* cannot add when bitmap is present */
7130 return -EINVAL;
7140 return -EBADF;
7143 inode = f->f_mapping->host;
7144 if (!S_ISREG(inode->i_mode)) {
7147 err = -EBADF;
7148 } else if (!(f->f_mode & FMODE_WRITE)) {
7151 err = -EBADF;
7152 } else if (atomic_read(&inode->i_writecount) != 1) {
7155 err = -EBUSY;
7161 mddev->bitmap_info.file = f;
7162 mddev->bitmap_info.offset = 0; /* file overrides offset */
7163 } else if (mddev->bitmap == NULL)
7164 return -ENOENT; /* cannot remove what isn't there */
7166 if (mddev->pers) {
7170 bitmap = md_bitmap_create(mddev, -1);
7173 mddev->bitmap = bitmap;
7179 fd = -1;
7189 struct file *f = mddev->bitmap_info.file;
7191 spin_lock(&mddev->lock);
7192 mddev->bitmap_info.file = NULL;
7193 spin_unlock(&mddev->lock);
7207 * This will always create an array with a type-0.90.0 superblock.
7210 * use to determine which style super-blocks are to be found on the devices.
7216 if (info->raid_disks == 0) {
7218 if (info->major_version < 0 ||
7219 info->major_version >= ARRAY_SIZE(super_types) ||
7220 super_types[info->major_version].name == NULL) {
7221 /* maybe try to auto-load a module? */
7223 info->major_version);
7224 return -EINVAL;
7226 mddev->major_version = info->major_version;
7227 mddev->minor_version = info->minor_version;
7228 mddev->patch_version = info->patch_version;
7229 mddev->persistent = !info->not_persistent;
7233 mddev->ctime = ktime_get_real_seconds();
7236 mddev->major_version = MD_MAJOR_VERSION;
7237 mddev->minor_version = MD_MINOR_VERSION;
7238 mddev->patch_version = MD_PATCHLEVEL_VERSION;
7239 mddev->ctime = ktime_get_real_seconds();
7241 mddev->level = info->level;
7242 mddev->clevel[0] = 0;
7243 mddev->dev_sectors = 2 * (sector_t)info->size;
7244 mddev->raid_disks = info->raid_disks;
7248 if (info->state & (1<<MD_SB_CLEAN))
7249 mddev->recovery_cp = MaxSector;
7251 mddev->recovery_cp = 0;
7252 mddev->persistent = ! info->not_persistent;
7253 mddev->external = 0;
7255 mddev->layout = info->layout;
7256 if (mddev->level == 0)
7258 mddev->layout = -1;
7259 mddev->chunk_sectors = info->chunk_size >> 9;
7261 if (mddev->persistent) {
7262 mddev->max_disks = MD_SB_DISKS;
7263 mddev->flags = 0;
7264 mddev->sb_flags = 0;
7266 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
7268 mddev->bitmap_info.default_offset = MD_SB_BYTES >> 9;
7269 mddev->bitmap_info.default_space = 64*2 - (MD_SB_BYTES >> 9);
7270 mddev->bitmap_info.offset = 0;
7272 mddev->reshape_position = MaxSector;
7277 get_random_bytes(mddev->uuid, 16);
7279 mddev->new_level = mddev->level;
7280 mddev->new_chunk_sectors = mddev->chunk_sectors;
7281 mddev->new_layout = mddev->layout;
7282 mddev->delta_disks = 0;
7283 mddev->reshape_backwards = 0;
7290 lockdep_assert_held(&mddev->reconfig_mutex);
7292 if (mddev->external_size)
7295 mddev->array_sectors = array_sectors;
7304 sector_t old_dev_sectors = mddev->dev_sectors;
7306 if (mddev->pers->resize == NULL)
7307 return -EINVAL;
7317 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
7318 mddev->sync_thread)
7319 return -EBUSY;
7321 return -EROFS;
7324 sector_t avail = rdev->sectors;
7329 return -ENOSPC;
7331 rv = mddev->pers->resize(mddev, num_sectors);
7334 md_cluster_ops->update_size(mddev, old_dev_sectors);
7335 else if (mddev->queue) {
7336 set_capacity_and_notify(mddev->gendisk,
7337 mddev->array_sectors);
7348 if (mddev->pers->check_reshape == NULL)
7349 return -EINVAL;
7351 return -EROFS;
7353 (mddev->max_disks && raid_disks >= mddev->max_disks))
7354 return -EINVAL;
7355 if (mddev->sync_thread ||
7356 test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
7357 test_bit(MD_RESYNCING_REMOTE, &mddev->recovery) ||
7358 mddev->reshape_position != MaxSector)
7359 return -EBUSY;
7362 if (mddev->raid_disks < raid_disks &&
7363 rdev->data_offset < rdev->new_data_offset)
7364 return -EINVAL;
7365 if (mddev->raid_disks > raid_disks &&
7366 rdev->data_offset > rdev->new_data_offset)
7367 return -EINVAL;
7370 mddev->delta_disks = raid_disks - mddev->raid_disks;
7371 if (mddev->delta_disks < 0)
7372 mddev->reshape_backwards = 1;
7373 else if (mddev->delta_disks > 0)
7374 mddev->reshape_backwards = 0;
7376 rv = mddev->pers->check_reshape(mddev);
7378 mddev->delta_disks = 0;
7379 mddev->reshape_backwards = 0;
7386 * on-line array.
7399 if (mddev->bitmap && mddev->bitmap_info.offset)
7402 if (mddev->major_version != info->major_version ||
7403 mddev->minor_version != info->minor_version ||
7404 /* mddev->patch_version != info->patch_version || */
7405 mddev->ctime != info->ctime ||
7406 mddev->level != info->level ||
7407 /* mddev->layout != info->layout || */
7408 mddev->persistent != !info->not_persistent ||
7409 mddev->chunk_sectors != info->chunk_size >> 9 ||
7411 ((state^info->state) & 0xfffffe00)
7413 return -EINVAL;
7415 if (info->size >= 0 && mddev->dev_sectors / 2 != info->size)
7417 if (mddev->raid_disks != info->raid_disks)
7419 if (mddev->layout != info->layout)
7421 if ((state ^ info->state) & (1<<MD_SB_BITMAP_PRESENT))
7426 return -EINVAL;
7428 if (mddev->layout != info->layout) {
7433 if (mddev->pers->check_reshape == NULL)
7434 return -EINVAL;
7436 mddev->new_layout = info->layout;
7437 rv = mddev->pers->check_reshape(mddev);
7439 mddev->new_layout = mddev->layout;
7443 if (info->size >= 0 && mddev->dev_sectors / 2 != info->size)
7444 rv = update_size(mddev, (sector_t)info->size * 2);
7446 if (mddev->raid_disks != info->raid_disks)
7447 rv = update_raid_disks(mddev, info->raid_disks);
7449 if ((state ^ info->state) & (1<<MD_SB_BITMAP_PRESENT)) {
7450 if (mddev->pers->quiesce == NULL || mddev->thread == NULL) {
7451 rv = -EINVAL;
7454 if (mddev->recovery || mddev->sync_thread) {
7455 rv = -EBUSY;
7458 if (info->state & (1<<MD_SB_BITMAP_PRESENT)) {
7461 if (mddev->bitmap) {
7462 rv = -EEXIST;
7465 if (mddev->bitmap_info.default_offset == 0) {
7466 rv = -EINVAL;
7469 mddev->bitmap_info.offset =
7470 mddev->bitmap_info.default_offset;
7471 mddev->bitmap_info.space =
7472 mddev->bitmap_info.default_space;
7473 bitmap = md_bitmap_create(mddev, -1);
7476 mddev->bitmap = bitmap;
7485 if (!mddev->bitmap) {
7486 rv = -ENOENT;
7489 if (mddev->bitmap->storage.file) {
7490 rv = -EINVAL;
7493 if (mddev->bitmap_info.nodes) {
7495 if (md_cluster_ops->lock_all_bitmaps(mddev) <= 0) {
7497 rv = -EPERM;
7498 md_cluster_ops->unlock_all_bitmaps(mddev);
7502 mddev->bitmap_info.nodes = 0;
7503 md_cluster_ops->leave(mddev);
7505 mddev->safemode_delay = DEFAULT_SAFEMODE_DELAY;
7510 mddev->bitmap_info.offset = 0;
7524 if (mddev->pers == NULL)
7525 return -ENODEV;
7530 err = -ENODEV;
7533 if (test_bit(MD_BROKEN, &mddev->flags))
7534 err = -EBUSY;
7544 * dosfs just mad... ;-)
7548 struct mddev *mddev = bdev->bd_disk->private_data;
7550 geo->heads = 2;
7551 geo->sectors = 4;
7552 geo->cylinders = mddev->array_sectors / 8;
7588 return -EFAULT;
7590 if (mddev->pers) {
7597 if (!list_empty(&mddev->disks)) {
7599 return -EBUSY;
7602 if (mddev->raid_disks) {
7604 return -EBUSY;
7622 return -ENOTTY;
7631 return -EACCES;
7649 mddev = bdev->bd_disk->private_data;
7659 if (!mddev->raid_disks && !mddev->external)
7660 err = -ENODEV;
7666 if (!mddev->raid_disks && !mddev->external)
7667 err = -ENODEV;
7684 wait_event_interruptible_timeout(mddev->sb_wait,
7686 &mddev->recovery),
7689 /* Need to flush page cache, and ensure no-one else opens
7692 mutex_lock(&mddev->open_mutex);
7693 if (mddev->pers && atomic_read(&mddev->openers) > 1) {
7694 mutex_unlock(&mddev->open_mutex);
7695 err = -EBUSY;
7698 if (test_and_set_bit(MD_CLOSING, &mddev->flags)) {
7699 mutex_unlock(&mddev->open_mutex);
7700 err = -EBUSY;
7703 mutex_unlock(&mddev->open_mutex);
7723 if ((!mddev->raid_disks && !mddev->external)
7727 err = -ENODEV;
7732 * Commands even a read-only array can execute:
7752 /* We can support ADD_NEW_DISK on read-only arrays
7753 * only if we are re-adding a preexisting device.
7754 * So require mddev->pers and MD_DISK_SYNC.
7756 if (mddev->pers) {
7759 err = -EFAULT;
7761 /* Need to clear read-only for this */
7772 * superblock, so we do not allow them on read-only arrays.
7774 if (!md_is_rdwr(mddev) && mddev->pers) {
7775 if (mddev->ro != MD_AUTO_READ) {
7776 err = -EROFS;
7779 mddev->ro = MD_RDWR;
7780 sysfs_notify_dirent_safe(mddev->sysfs_state);
7781 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
7783 /* If a device failed while we were read-only, we
7786 if (test_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags)) {
7788 wait_event(mddev->sb_wait,
7789 !test_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags) &&
7790 !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags));
7800 err = -EFAULT;
7808 md_cluster_ops->new_disk_ack(mddev, false);
7810 err = -EINVAL;
7826 err = -EINVAL;
7831 if (mddev->hold_active == UNTIL_IOCTL &&
7832 err != -EINVAL)
7833 mddev->hold_active = 0;
7837 clear_bit(MD_CLOSING, &mddev->flags);
7862 struct mddev *mddev = bdev->bd_disk->private_data;
7869 if (!mddev->raid_disks && !mddev->external) {
7870 err = -ENODEV;
7875 * Transitioning to read-auto need only happen for arrays that call
7878 if (!ro && mddev->ro == MD_RDONLY && mddev->pers) {
7882 mddev->ro = MD_AUTO_READ;
7896 mddev = mddev_get(disk->private_data);
7899 return -ENODEV;
7901 err = mutex_lock_interruptible(&mddev->open_mutex);
7905 err = -ENODEV;
7906 if (test_bit(MD_CLOSING, &mddev->flags))
7909 atomic_inc(&mddev->openers);
7910 mutex_unlock(&mddev->open_mutex);
7916 mutex_unlock(&mddev->open_mutex);
7924 struct mddev *mddev = disk->private_data;
7927 atomic_dec(&mddev->openers);
7933 struct mddev *mddev = disk->private_data;
7936 if (mddev->changed)
7938 mddev->changed = 0;
7944 struct mddev *mddev = disk->private_data;
7946 percpu_ref_exit(&mddev->writes_pending);
7971 * md_thread is a 'system-thread', it's priority should be very
7986 * we don't add to the load-average.
7994 (thread->wqueue,
7995 test_bit(THREAD_WAKEUP, &thread->flags)
7997 thread->timeout);
7999 clear_bit(THREAD_WAKEUP, &thread->flags);
8003 thread->run(thread);
8016 wake_up_process(t->tsk);
8027 pr_debug("md: waking up MD thread %s.\n", t->tsk->comm);
8028 set_bit(THREAD_WAKEUP, &t->flags);
8029 wake_up(&t->wqueue);
8044 init_waitqueue_head(&thread->wqueue);
8046 thread->run = run;
8047 thread->mddev = mddev;
8048 thread->timeout = MAX_SCHEDULE_TIMEOUT;
8049 thread->tsk = kthread_run(md_thread, thread,
8051 mdname(thread->mddev),
8053 if (IS_ERR(thread->tsk)) {
8064 lockdep_is_held(&mddev->reconfig_mutex));
8072 pr_debug("interrupting MD-thread pid %d\n", task_pid_nr(thread->tsk));
8073 kthread_stop(thread->tsk);
8080 if (!rdev || test_bit(Faulty, &rdev->flags))
8083 if (!mddev->pers || !mddev->pers->error_handler)
8085 mddev->pers->error_handler(mddev, rdev);
8087 if (mddev->pers->level == 0 || mddev->pers->level == LEVEL_LINEAR)
8090 if (mddev->degraded && !test_bit(MD_BROKEN, &mddev->flags))
8091 set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
8092 sysfs_notify_dirent_safe(rdev->sysfs_state);
8093 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
8094 if (!test_bit(MD_BROKEN, &mddev->flags)) {
8095 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
8096 md_wakeup_thread(mddev->thread);
8098 if (mddev->event_work.func)
8099 queue_work(md_misc_wq, &mddev->event_work);
8115 seq_printf(seq, "%pg ", rdev->bdev);
8131 if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ||
8132 test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
8133 max_sectors = mddev->resync_max_sectors;
8135 max_sectors = mddev->dev_sectors;
8137 resync = mddev->curr_resync;
8139 if (test_bit(MD_RECOVERY_DONE, &mddev->recovery))
8145 res = atomic_read(&mddev->recovery_active);
8151 if (resync < res || resync - res < MD_RESYNC_ACTIVE)
8154 resync -= res;
8158 if (test_bit(MD_RESYNCING_REMOTE, &mddev->recovery)) {
8162 if (rdev->raid_disk >= 0 &&
8163 !test_bit(Faulty, &rdev->flags) &&
8164 rdev->recovery_offset != MaxSector &&
8165 rdev->recovery_offset) {
8169 if (mddev->reshape_position != MaxSector)
8175 if (mddev->recovery_cp < MaxSector) {
8202 int i, x = per_milli/50, y = 20-x;
8212 (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)?
8214 (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)?
8216 (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ?
8239 dt = ((jiffies - mddev->resync_mark) / HZ);
8242 curr_mark_cnt = mddev->curr_mark_cnt;
8243 recovery_active = atomic_read(&mddev->recovery_active);
8244 resync_mark_cnt = mddev->resync_mark_cnt;
8247 db = curr_mark_cnt - (recovery_active + resync_mark_cnt);
8249 rt = max_sectors - resync; /* number of remaining sectors */
8273 if (!l--)
8279 if (!l--) {
8287 if (!l--)
8307 tmp = mddev->all_mddevs.next;
8320 tmp = mddev->all_mddevs.next;
8349 seq_printf(seq, "[%s] ", pers->name);
8353 seq->poll_event = atomic_read(&md_event_count);
8361 spin_lock(&mddev->lock);
8362 if (mddev->pers || mddev->raid_disks || !list_empty(&mddev->disks)) {
8364 mddev->pers ? "" : "in");
8365 if (mddev->pers) {
8366 if (mddev->ro == MD_RDONLY)
8367 seq_printf(seq, " (read-only)");
8368 if (mddev->ro == MD_AUTO_READ)
8369 seq_printf(seq, " (auto-read-only)");
8370 seq_printf(seq, " %s", mddev->pers->name);
8376 seq_printf(seq, " %pg[%d]", rdev->bdev, rdev->desc_nr);
8378 if (test_bit(WriteMostly, &rdev->flags))
8380 if (test_bit(Journal, &rdev->flags))
8382 if (test_bit(Faulty, &rdev->flags)) {
8386 if (rdev->raid_disk < 0)
8388 if (test_bit(Replacement, &rdev->flags))
8390 sectors += rdev->sectors;
8394 if (!list_empty(&mddev->disks)) {
8395 if (mddev->pers)
8398 mddev->array_sectors / 2);
8403 if (mddev->persistent) {
8404 if (mddev->major_version != 0 ||
8405 mddev->minor_version != 90) {
8407 mddev->major_version,
8408 mddev->minor_version);
8410 } else if (mddev->external)
8412 mddev->metadata_type);
8414 seq_printf(seq, " super non-persistent");
8416 if (mddev->pers) {
8417 mddev->pers->status(seq, mddev);
8419 if (mddev->pers->sync_request) {
8426 md_bitmap_status(seq, mddev->bitmap);
8430 spin_unlock(&mddev->lock);
8451 seq = file->private_data;
8452 seq->poll_event = atomic_read(&md_event_count);
8459 struct seq_file *seq = filp->private_data;
8469 if (seq->poll_event != atomic_read(&md_event_count))
8485 p->name, p->level);
8487 list_add_tail(&p->list, &pers_list);
8495 pr_debug("md: %s personality unregistered\n", p->name);
8497 list_del_init(&p->list);
8509 ret = -EALREADY;
8532 request_module("md-cluster");
8536 pr_warn("can't find md-cluster module or get its reference.\n");
8538 return -ENOENT;
8542 ret = md_cluster_ops->join(mddev, nodes);
8544 mddev->safemode_delay = 0;
8552 md_cluster_ops->leave(mddev);
8565 struct gendisk *disk = rdev->bdev->bd_disk;
8566 curr_events = (int)part_stat_read_accum(disk->part0, sectors) -
8567 atomic_read(&disk->sync_io);
8573 * non-sync IO will cause disk_stat to increase without
8577 * the array to appear non-idle, and resync will slow
8581 * completing might cause the array to appear non-idle
8583 * not have been non-resync activity. This will only
8590 if (init || curr_events - rdev->last_events > 64) {
8591 rdev->last_events = curr_events;
8602 atomic_sub(blocks, &mddev->recovery_active);
8603 wake_up(&mddev->recovery_wait);
8605 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
8606 set_bit(MD_RECOVERY_ERROR, &mddev->recovery);
8607 md_wakeup_thread(mddev->thread);
8627 BUG_ON(mddev->ro == MD_RDONLY);
8628 if (mddev->ro == MD_AUTO_READ) {
8630 mddev->ro = MD_RDWR;
8631 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
8632 md_wakeup_thread(mddev->thread);
8633 md_wakeup_thread(mddev->sync_thread);
8637 percpu_ref_get(&mddev->writes_pending);
8639 if (mddev->safemode == 1)
8640 mddev->safemode = 0;
8641 /* sync_checkers is always 0 when writes_pending is in per-cpu mode */
8642 if (mddev->in_sync || mddev->sync_checkers) {
8643 spin_lock(&mddev->lock);
8644 if (mddev->in_sync) {
8645 mddev->in_sync = 0;
8646 set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
8647 set_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
8648 md_wakeup_thread(mddev->thread);
8651 spin_unlock(&mddev->lock);
8655 sysfs_notify_dirent_safe(mddev->sysfs_state);
8656 if (!mddev->has_superblocks)
8658 wait_event(mddev->sb_wait,
8659 !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags) ||
8661 if (test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)) {
8662 percpu_ref_put(&mddev->writes_pending);
8681 WARN_ON_ONCE(mddev->in_sync || !md_is_rdwr(mddev));
8682 percpu_ref_get(&mddev->writes_pending);
8688 percpu_ref_put(&mddev->writes_pending);
8690 if (mddev->safemode == 2)
8691 md_wakeup_thread(mddev->thread);
8692 else if (mddev->safemode_delay)
8694 * every ->safemode_delay jiffies
8696 mod_timer(&mddev->safemode_timer,
8697 roundup(jiffies, mddev->safemode_delay) +
8698 mddev->safemode_delay);
8709 if (__blkdev_issue_discard(rdev->bdev, start, size, GFP_NOIO,
8715 if (mddev->gendisk)
8717 disk_devt(mddev->gendisk),
8718 bio->bi_iter.bi_sector);
8725 struct md_io_clone *md_io_clone = bio->bi_private;
8726 struct bio *orig_bio = md_io_clone->orig_bio;
8727 struct mddev *mddev = md_io_clone->mddev;
8729 if (bio->bi_status && !orig_bio->bi_status)
8730 orig_bio->bi_status = bio->bi_status;
8732 if (md_io_clone->start_time)
8733 bio_end_io_acct(orig_bio, md_io_clone->start_time);
8737 percpu_ref_put(&mddev->active_io);
8742 struct block_device *bdev = (*bio)->bi_bdev;
8745 bio_alloc_clone(bdev, *bio, GFP_NOIO, &mddev->io_clone_set);
8748 md_io_clone->orig_bio = *bio;
8749 md_io_clone->mddev = mddev;
8750 if (blk_queue_io_stat(bdev->bd_disk->queue))
8751 md_io_clone->start_time = bio_start_io_acct(*bio);
8753 clone->bi_end_io = md_end_clone_io;
8754 clone->bi_private = md_io_clone;
8760 percpu_ref_get(&mddev->active_io);
8773 if (!mddev->pers)
8777 if (!mddev->pers->sync_request)
8780 spin_lock(&mddev->lock);
8781 if (mddev->in_sync) {
8782 mddev->in_sync = 0;
8783 set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
8784 set_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
8785 if (mddev->safemode_delay &&
8786 mddev->safemode == 0)
8787 mddev->safemode = 1;
8788 spin_unlock(&mddev->lock);
8790 sysfs_notify_dirent_safe(mddev->sysfs_state);
8792 wait_event(mddev->sb_wait,
8793 !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags));
8795 spin_unlock(&mddev->lock);
8804 struct mddev *mddev = thread->mddev;
8820 if (test_bit(MD_RECOVERY_DONE, &mddev->recovery) ||
8821 test_bit(MD_RECOVERY_WAIT, &mddev->recovery))
8823 if (!md_is_rdwr(mddev)) {/* never try to sync a read-only array */
8824 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
8829 ret = md_cluster_ops->resync_start(mddev);
8833 set_bit(MD_CLUSTER_RESYNC_LOCKED, &mddev->flags);
8834 if (!(test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ||
8835 test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) ||
8836 test_bit(MD_RECOVERY_RECOVER, &mddev->recovery))
8837 && ((unsigned long long)mddev->curr_resync_completed
8838 < (unsigned long long)mddev->resync_max_sectors))
8842 if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
8843 if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)) {
8844 desc = "data-check";
8846 } else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
8847 desc = "requested-resync";
8851 } else if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
8856 mddev->last_sync_action = action ?: desc;
8869 int mddev2_minor = -1;
8870 mddev->curr_resync = MD_RESYNC_DELAYED;
8873 if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
8877 if (test_bit(MD_DELETED, &mddev2->flags))
8881 if (!mddev->parallel_resync
8882 && mddev2->curr_resync
8886 mddev->curr_resync == MD_RESYNC_DELAYED) {
8888 mddev->curr_resync = MD_RESYNC_YIELDED;
8892 mddev->curr_resync == MD_RESYNC_YIELDED)
8902 if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
8903 mddev2->curr_resync >= mddev->curr_resync) {
8904 if (mddev2_minor != mddev2->md_minor) {
8905 mddev2_minor = mddev2->md_minor;
8922 } while (mddev->curr_resync < MD_RESYNC_DELAYED);
8925 if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
8929 max_sectors = mddev->resync_max_sectors;
8930 atomic64_set(&mddev->resync_mismatches, 0);
8932 if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
8933 j = mddev->resync_min;
8934 else if (!mddev->bitmap)
8935 j = mddev->recovery_cp;
8937 } else if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) {
8938 max_sectors = mddev->resync_max_sectors;
8945 mddev->reshape_position != MaxSector)
8946 j = mddev->reshape_position;
8949 max_sectors = mddev->dev_sectors;
8953 if (rdev->raid_disk >= 0 &&
8954 !test_bit(Journal, &rdev->flags) &&
8955 !test_bit(Faulty, &rdev->flags) &&
8956 !test_bit(In_sync, &rdev->flags) &&
8957 rdev->recovery_offset < j)
8958 j = rdev->recovery_offset;
8969 if (mddev->bitmap) {
8970 mddev->pers->quiesce(mddev, 1);
8971 mddev->pers->quiesce(mddev, 0);
8988 mddev->resync_mark = mark[last_mark];
8989 mddev->resync_mark_cnt = mark_cnt[last_mark];
8998 atomic_set(&mddev->recovery_active, 0);
9004 mddev->curr_resync = j;
9006 mddev->curr_resync = MD_RESYNC_ACTIVE; /* no longer delayed */
9007 mddev->curr_resync_completed = j;
9008 sysfs_notify_dirent_safe(mddev->sysfs_completed);
9018 if (!test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
9019 ((mddev->curr_resync > mddev->curr_resync_completed &&
9020 (mddev->curr_resync - mddev->curr_resync_completed)
9023 (j - mddev->curr_resync_completed)*2
9024 >= mddev->resync_max - mddev->curr_resync_completed ||
9025 mddev->curr_resync_completed > mddev->resync_max
9028 wait_event(mddev->recovery_wait,
9029 atomic_read(&mddev->recovery_active) == 0);
9030 mddev->curr_resync_completed = j;
9031 if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) &&
9032 j > mddev->recovery_cp)
9033 mddev->recovery_cp = j;
9035 set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
9036 sysfs_notify_dirent_safe(mddev->sysfs_completed);
9039 while (j >= mddev->resync_max &&
9040 !test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
9041 /* As this condition is controlled by user-space,
9046 wait_event_interruptible(mddev->recovery_wait,
9047 mddev->resync_max > j
9049 &mddev->recovery));
9052 if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
9055 sectors = mddev->pers->sync_request(mddev, j, &skipped);
9057 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
9063 atomic_add(sectors, &mddev->recovery_active);
9066 if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
9074 mddev->curr_resync = j;
9075 mddev->curr_mark_cnt = io_sectors;
9091 mddev->resync_mark = mark[next];
9092 mddev->resync_mark_cnt = mark_cnt[next];
9094 mark_cnt[next] = io_sectors - atomic_read(&mddev->recovery_active);
9098 if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
9103 * the 'hard' speed limit, or the system was IO-idle for
9105 * the system might be non-idle CPU-wise, but we only care
9111 recovery_done = io_sectors - atomic_read(&mddev->recovery_active);
9112 currspeed = ((unsigned long)(recovery_done - mddev->resync_mark_cnt))/2
9113 /((jiffies-mddev->resync_mark)/HZ +1) +1;
9125 wait_event(mddev->recovery_wait,
9126 !atomic_read(&mddev->recovery_active));
9131 test_bit(MD_RECOVERY_INTR, &mddev->recovery)
9137 wait_event(mddev->recovery_wait, !atomic_read(&mddev->recovery_active));
9139 if (!test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
9140 !test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
9141 mddev->curr_resync >= MD_RESYNC_ACTIVE) {
9142 mddev->curr_resync_completed = mddev->curr_resync;
9143 sysfs_notify_dirent_safe(mddev->sysfs_completed);
9145 mddev->pers->sync_request(mddev, max_sectors, &skipped);
9147 if (!test_bit(MD_RECOVERY_CHECK, &mddev->recovery) &&
9148 mddev->curr_resync > MD_RESYNC_ACTIVE) {
9149 if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
9150 if (test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
9151 if (mddev->curr_resync >= mddev->recovery_cp) {
9155 &mddev->recovery))
9156 mddev->recovery_cp =
9157 mddev->curr_resync_completed;
9159 mddev->recovery_cp =
9160 mddev->curr_resync;
9163 mddev->recovery_cp = MaxSector;
9165 if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery))
9166 mddev->curr_resync = MaxSector;
9167 if (!test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
9168 test_bit(MD_RECOVERY_RECOVER, &mddev->recovery)) {
9171 if (rdev->raid_disk >= 0 &&
9172 mddev->delta_disks >= 0 &&
9173 !test_bit(Journal, &rdev->flags) &&
9174 !test_bit(Faulty, &rdev->flags) &&
9175 !test_bit(In_sync, &rdev->flags) &&
9176 rdev->recovery_offset < mddev->curr_resync)
9177 rdev->recovery_offset = mddev->curr_resync;
9186 set_mask_bits(&mddev->sb_flags, 0,
9189 if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
9190 !test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
9191 mddev->delta_disks > 0 &&
9192 mddev->pers->finish_reshape &&
9193 mddev->pers->size &&
9194 mddev->queue) {
9196 md_set_array_sectors(mddev, mddev->pers->size(mddev, 0, 0));
9199 set_capacity_and_notify(mddev->gendisk,
9200 mddev->array_sectors);
9203 spin_lock(&mddev->lock);
9204 if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
9206 if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
9207 mddev->resync_min = 0;
9208 mddev->resync_max = MaxSector;
9209 } else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
9210 mddev->resync_min = mddev->curr_resync_completed;
9211 set_bit(MD_RECOVERY_DONE, &mddev->recovery);
9212 mddev->curr_resync = MD_RESYNC_NONE;
9213 spin_unlock(&mddev->lock);
9216 wake_up(&mddev->sb_wait);
9217 md_wakeup_thread(mddev->thread);
9230 if (this && test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
9236 rdev->raid_disk >= 0 &&
9237 !test_bit(Blocked, &rdev->flags) &&
9238 test_bit(Faulty, &rdev->flags) &&
9239 atomic_read(&rdev->nr_pending)==0) {
9240 /* Faulty non-Blocked devices with nr_pending == 0
9246 set_bit(RemoveSynchronized, &rdev->flags);
9254 rdev->raid_disk >= 0 &&
9255 !test_bit(Blocked, &rdev->flags) &&
9256 ((test_bit(RemoveSynchronized, &rdev->flags) ||
9257 (!test_bit(In_sync, &rdev->flags) &&
9258 !test_bit(Journal, &rdev->flags))) &&
9259 atomic_read(&rdev->nr_pending)==0)) {
9260 if (mddev->pers->hot_remove_disk(
9263 rdev->saved_raid_disk = rdev->raid_disk;
9264 rdev->raid_disk = -1;
9268 if (remove_some && test_bit(RemoveSynchronized, &rdev->flags))
9269 clear_bit(RemoveSynchronized, &rdev->flags);
9272 if (removed && mddev->kobj.sd)
9273 sysfs_notify_dirent_safe(mddev->sysfs_degraded);
9281 if (test_bit(Candidate, &rdev->flags))
9283 if (rdev->raid_disk >= 0 &&
9284 !test_bit(In_sync, &rdev->flags) &&
9285 !test_bit(Journal, &rdev->flags) &&
9286 !test_bit(Faulty, &rdev->flags))
9288 if (rdev->raid_disk >= 0)
9290 if (test_bit(Faulty, &rdev->flags))
9292 if (!test_bit(Journal, &rdev->flags)) {
9294 !(rdev->saved_raid_disk >= 0 &&
9295 !test_bit(Bitmap_sync, &rdev->flags)))
9298 rdev->recovery_offset = 0;
9300 if (mddev->pers->hot_add_disk(mddev, rdev) == 0) {
9303 if (!test_bit(Journal, &rdev->flags))
9306 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
9311 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
9319 rcu_assign_pointer(mddev->sync_thread,
9321 if (!mddev->sync_thread) {
9325 clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
9326 clear_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
9327 clear_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
9328 clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
9329 clear_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
9332 &mddev->recovery))
9333 if (mddev->sysfs_action)
9334 sysfs_notify_dirent_safe(mddev->sysfs_action);
9336 md_wakeup_thread(mddev->sync_thread);
9337 sysfs_notify_dirent_safe(mddev->sysfs_action);
9342 * This routine is regularly called by all per-raid-array threads to
9343 * deal with generic issues like resync and super-block update.
9350 * "->recovery" and create a thread at ->sync_thread.
9361 * 6/ If array has spares or is not in-sync, start a resync thread.
9365 if (test_bit(MD_ALLOW_SB_UPDATE, &mddev->flags) && mddev->sb_flags) {
9366 /* Write superblock - thread that called mddev_suspend()
9369 set_bit(MD_UPDATING_SB, &mddev->flags);
9371 if (test_bit(MD_ALLOW_SB_UPDATE, &mddev->flags))
9373 clear_bit_unlock(MD_UPDATING_SB, &mddev->flags);
9374 wake_up(&mddev->sb_wait);
9380 if (mddev->bitmap)
9384 if (mddev->pers->sync_request && !mddev->external) {
9387 mddev->safemode = 2;
9393 !test_bit(MD_RECOVERY_NEEDED, &mddev->recovery))
9396 (mddev->sb_flags & ~ (1<<MD_SB_CHANGE_PENDING)) ||
9397 test_bit(MD_RECOVERY_NEEDED, &mddev->recovery) ||
9398 test_bit(MD_RECOVERY_DONE, &mddev->recovery) ||
9399 (mddev->external == 0 && mddev->safemode == 1) ||
9400 (mddev->safemode == 2
9401 && !mddev->in_sync && mddev->recovery_cp == MaxSector)
9407 bool try_set_sync = mddev->safemode != 0;
9409 if (!mddev->external && mddev->safemode == 1)
9410 mddev->safemode = 0;
9414 if (!mddev->external && mddev->in_sync)
9421 clear_bit(Blocked, &rdev->flags);
9422 /* On a read-only array we can:
9423 * - remove failed devices
9424 * - add already-in_sync devices if the array itself
9425 * is in-sync.
9426 * As we only add devices that are already in-sync,
9431 * ->spare_active and clear saved_raid_disk
9433 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
9435 clear_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
9436 clear_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
9437 clear_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
9447 if (test_and_clear_bit(ClusterRemove, &rdev->flags) &&
9448 rdev->raid_disk < 0)
9453 if (try_set_sync && !mddev->external && !mddev->in_sync) {
9454 spin_lock(&mddev->lock);
9456 spin_unlock(&mddev->lock);
9459 if (mddev->sb_flags)
9466 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) {
9467 if (!test_bit(MD_RECOVERY_DONE, &mddev->recovery)) {
9469 clear_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
9473 if (WARN_ON_ONCE(!mddev->sync_thread))
9483 mddev->curr_resync_completed = 0;
9484 spin_lock(&mddev->lock);
9485 set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
9486 spin_unlock(&mddev->lock);
9490 clear_bit(MD_RECOVERY_INTR, &mddev->recovery);
9491 clear_bit(MD_RECOVERY_DONE, &mddev->recovery);
9493 if (!test_and_clear_bit(MD_RECOVERY_NEEDED, &mddev->recovery) ||
9494 test_bit(MD_RECOVERY_FROZEN, &mddev->recovery))
9499 * Spares are also removed and re-added, to allow
9500 * the personality to fail the re-add.
9503 if (mddev->reshape_position != MaxSector) {
9504 if (mddev->pers->check_reshape == NULL ||
9505 mddev->pers->check_reshape(mddev) != 0)
9508 set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
9509 clear_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
9511 clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
9512 clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
9513 clear_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
9514 set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
9515 } else if (mddev->recovery_cp < MaxSector) {
9516 set_bit(MD_RECOVERY_SYNC, &mddev->recovery);
9517 clear_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
9518 } else if (!test_bit(MD_RECOVERY_SYNC, &mddev->recovery))
9522 if (mddev->pers->sync_request) {
9528 md_bitmap_write_all(mddev->bitmap);
9530 INIT_WORK(&mddev->del_work, md_start_sync);
9531 queue_work(md_misc_wq, &mddev->del_work);
9535 if (!mddev->sync_thread) {
9536 clear_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
9539 &mddev->recovery))
9540 if (mddev->sysfs_action)
9541 sysfs_notify_dirent_safe(mddev->sysfs_action);
9544 wake_up(&mddev->sb_wait);
9553 sector_t old_dev_sectors = mddev->dev_sectors;
9557 md_unregister_thread(mddev, &mddev->sync_thread);
9558 atomic_inc(&mddev->sync_seq);
9560 if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
9561 !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
9562 mddev->degraded != mddev->raid_disks) {
9565 if (mddev->pers->spare_active(mddev)) {
9566 sysfs_notify_dirent_safe(mddev->sysfs_degraded);
9567 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
9570 if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
9571 mddev->pers->finish_reshape) {
9572 mddev->pers->finish_reshape(mddev);
9577 /* If array is no-longer degraded, then any saved_raid_disk
9580 if (!mddev->degraded)
9582 rdev->saved_raid_disk = -1;
9588 if (test_and_clear_bit(MD_CLUSTER_RESYNC_LOCKED, &mddev->flags))
9589 md_cluster_ops->resync_finish(mddev);
9590 clear_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
9591 clear_bit(MD_RECOVERY_DONE, &mddev->recovery);
9592 clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
9593 clear_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
9594 clear_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
9595 clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
9597 * We call md_cluster_ops->update_size here because sync_size could
9602 && !test_bit(MD_CLOSING, &mddev->flags))
9603 md_cluster_ops->update_size(mddev, old_dev_sectors);
9605 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
9606 sysfs_notify_dirent_safe(mddev->sysfs_completed);
9607 sysfs_notify_dirent_safe(mddev->sysfs_action);
9609 if (mddev->event_work.func)
9610 queue_work(md_misc_wq, &mddev->event_work);
9617 sysfs_notify_dirent_safe(rdev->sysfs_state);
9618 wait_event_timeout(rdev->blocked_wait,
9619 !test_bit(Blocked, &rdev->flags) &&
9620 !test_bit(BlockedBadBlocks, &rdev->flags),
9632 if (rdev->data_offset > rdev->new_data_offset)
9633 rdev->sectors += rdev->data_offset - rdev->new_data_offset;
9635 rdev->sectors -= rdev->new_data_offset - rdev->data_offset;
9636 rdev->data_offset = rdev->new_data_offset;
9647 struct mddev *mddev = rdev->mddev;
9650 s += rdev->new_data_offset;
9652 s += rdev->data_offset;
9653 rv = badblocks_set(&rdev->badblocks, s, sectors, 0);
9656 if (test_bit(ExternalBbl, &rdev->flags))
9657 sysfs_notify_dirent_safe(rdev->sysfs_unack_badblocks);
9658 sysfs_notify_dirent_safe(rdev->sysfs_state);
9659 set_mask_bits(&mddev->sb_flags, 0,
9661 md_wakeup_thread(rdev->mddev->thread);
9673 s += rdev->new_data_offset;
9675 s += rdev->data_offset;
9676 rv = badblocks_clear(&rdev->badblocks, s, sectors);
9677 if ((rv == 0) && test_bit(ExternalBbl, &rdev->flags))
9678 sysfs_notify_dirent_safe(rdev->sysfs_badblocks);
9695 if (mddev->pers)
9697 if (mddev->persistent)
9698 mddev->safemode = 2;
9734 int ret = -ENOMEM;
9778 struct mdp_superblock_1 *sb = page_address(rdev->sb_page);
9786 if (mddev->dev_sectors != le64_to_cpu(sb->size)) {
9787 ret = mddev->pers->resize(mddev, le64_to_cpu(sb->size));
9789 pr_info("md-cluster: resize failed\n");
9791 md_bitmap_update_sb(mddev->bitmap);
9796 if (test_bit(Faulty, &rdev2->flags))
9800 role = le16_to_cpu(sb->dev_roles[rdev2->desc_nr]);
9802 if (test_bit(Candidate, &rdev2->flags)) {
9805 rdev2->bdev);
9810 clear_bit(Candidate, &rdev2->flags);
9813 if (role != rdev2->raid_disk) {
9817 if (rdev2->raid_disk == -1 && role != MD_DISK_ROLE_SPARE &&
9818 !(le32_to_cpu(sb->feature_map) &
9820 rdev2->saved_raid_disk = role;
9823 rdev2->bdev);
9824 /* wakeup mddev->thread here, so array could
9826 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
9827 md_wakeup_thread(mddev->thread);
9837 clear_bit(Blocked, &rdev2->flags);
9842 if (mddev->raid_disks != le32_to_cpu(sb->raid_disks)) {
9843 ret = update_raid_disks(mddev, le32_to_cpu(sb->raid_disks));
9849 * Since mddev->delta_disks has already updated in update_raid_disks,
9852 if (test_bit(MD_RESYNCING_REMOTE, &mddev->recovery) &&
9853 (le32_to_cpu(sb->feature_map) & MD_FEATURE_RESHAPE_ACTIVE)) {
9858 mddev->reshape_position = le64_to_cpu(sb->reshape_position);
9859 if (mddev->pers->update_reshape_pos)
9860 mddev->pers->update_reshape_pos(mddev);
9861 if (mddev->pers->start_reshape)
9862 mddev->pers->start_reshape(mddev);
9863 } else if (test_bit(MD_RESYNCING_REMOTE, &mddev->recovery) &&
9864 mddev->reshape_position != MaxSector &&
9865 !(le32_to_cpu(sb->feature_map) & MD_FEATURE_RESHAPE_ACTIVE)) {
9867 mddev->reshape_position = MaxSector;
9868 if (mddev->pers->update_reshape_pos)
9869 mddev->pers->update_reshape_pos(mddev);
9873 mddev->events = le64_to_cpu(sb->events);
9879 struct page *swapout = rdev->sb_page;
9885 rdev->sb_page = NULL;
9888 ClearPageUptodate(rdev->sb_page);
9889 rdev->sb_loaded = 0;
9890 err = super_types[mddev->major_version].
9891 load_super(rdev, NULL, mddev->minor_version);
9895 __func__, __LINE__, rdev->desc_nr, err);
9896 if (rdev->sb_page)
9897 put_page(rdev->sb_page);
9898 rdev->sb_page = swapout;
9899 rdev->sb_loaded = 1;
9903 sb = page_address(rdev->sb_page);
9908 if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_RECOVERY_OFFSET))
9909 rdev->recovery_offset = le64_to_cpu(sb->recovery_offset);
9912 * device In_sync and mddev->degraded
9914 if (rdev->recovery_offset == MaxSector &&
9915 !test_bit(In_sync, &rdev->flags) &&
9916 mddev->pers->spare_active(mddev))
9917 sysfs_notify_dirent_safe(mddev->sysfs_degraded);
9930 if (iter->desc_nr == nr) {
9949 if (!test_bit(Faulty, &rdev->flags))
9975 node_detected_dev->dev = dev;
9977 list_add_tail(&node_detected_dev->list, &all_detected_devices);
9999 list_del(&node_detected_dev->list);
10000 dev = node_detected_dev->dev;
10008 if (test_bit(Faulty, &rdev->flags))
10011 set_bit(AutoDetected, &rdev->flags);
10012 list_add(&rdev->same_set, &pending_raid_disks);
10035 * waiting for us in select() or poll() - wake them up
10052 mddev->ctime = 0;
10053 mddev->hold_active = 0;