Lines Matching defs:mddev
88 static int remove_and_add_spares(struct mddev *mddev,
90 static void mddev_detach(struct mddev *mddev);
91 static void export_rdev(struct md_rdev *rdev, struct mddev *mddev);
117 static inline int speed_min(struct mddev *mddev)
119 return mddev->sync_speed_min ?
120 mddev->sync_speed_min : sysctl_speed_limit_min;
123 static inline int speed_max(struct mddev *mddev)
125 return mddev->sync_speed_max ?
126 mddev->sync_speed_max : sysctl_speed_limit_max;
138 static void rdevs_uninit_serial(struct mddev *mddev)
142 rdev_for_each(rdev, mddev)
174 static int rdevs_init_serial(struct mddev *mddev)
179 rdev_for_each(rdev, mddev) {
186 if (ret && !mddev->serial_info_pool)
187 rdevs_uninit_serial(mddev);
199 return (rdev && rdev->mddev->bitmap_info.max_write_behind > 0 &&
209 void mddev_create_serial_pool(struct mddev *mddev, struct md_rdev *rdev,
219 mddev_suspend(mddev);
222 ret = rdevs_init_serial(mddev);
228 if (mddev->serial_info_pool == NULL) {
233 mddev->serial_info_pool =
236 if (!mddev->serial_info_pool) {
237 rdevs_uninit_serial(mddev);
244 mddev_resume(mddev);
253 void mddev_destroy_serial_pool(struct mddev *mddev, struct md_rdev *rdev,
259 if (mddev->serial_info_pool) {
264 mddev_suspend(mddev);
265 rdev_for_each(temp, mddev) {
267 if (!mddev->serialize_policy ||
283 mempool_destroy(mddev->serial_info_pool);
284 mddev->serial_info_pool = NULL;
287 mddev_resume(mddev);
356 static bool is_suspended(struct mddev *mddev, struct bio *bio)
358 if (is_md_suspended(mddev))
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)
371 void md_handle_request(struct mddev *mddev, struct bio *bio)
374 if (is_suspended(mddev, bio)) {
382 prepare_to_wait(&mddev->sb_wait, &__wait,
384 if (!is_suspended(mddev, bio))
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)) {
431 md_handle_request(mddev, bio);
440 void mddev_suspend(struct mddev *mddev)
442 struct md_thread *thread = rcu_dereference_protected(mddev->thread,
443 lockdep_is_held(&mddev->reconfig_mutex));
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();
465 void mddev_resume(struct mddev *mddev)
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 */
490 struct mddev *mddev = rdev->mddev;
494 rdev_dec_pending(rdev, mddev);
496 if (atomic_dec_and_test(&mddev->flush_pending)) {
498 percpu_ref_put(&mddev->active_io);
501 queue_work(md_wq, &mddev->flush_work);
509 struct mddev *mddev = container_of(ws, struct 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);
516 rdev_for_each_rcu(rdev, mddev)
525 GFP_NOIO, &mddev->bio_set);
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);
543 struct mddev *mddev = container_of(ws, struct 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);
563 md_handle_request(mddev, bio);
573 bool md_flush_request(struct mddev *mddev, struct bio *bio)
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);
621 static inline struct mddev *mddev_get(struct mddev *mddev)
625 if (test_bit(MD_DELETED, &mddev->flags))
627 atomic_inc(&mddev->active);
628 return mddev;
633 void mddev_put(struct mddev *mddev)
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);
656 void mddev_init(struct mddev *mddev)
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;
682 static struct mddev *mddev_find_locked(dev_t unit)
684 struct mddev *mddev;
686 list_for_each_entry(mddev, &all_mddevs, all_mddevs)
687 if (mddev->unit == unit)
688 return mddev;
714 static struct mddev *mddev_alloc(dev_t unit)
716 struct mddev *new;
756 static void mddev_free(struct mddev *mddev)
759 list_del(&mddev->all_mddevs);
762 kfree(mddev);
767 void mddev_unlock(struct mddev *mddev)
773 if (!list_empty(&mddev->deleting))
774 list_splice_init(&mddev->deleting, &delete);
776 if (mddev->to_remove) {
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);
821 export_rdev(rdev, mddev);
826 struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr)
830 rdev_for_each_rcu(rdev, mddev)
838 static struct md_rdev *find_rdev(struct mddev *mddev, dev_t dev)
842 rdev_for_each(rdev, mddev)
849 struct md_rdev *md_find_rdev_rcu(struct mddev *mddev, dev_t dev)
853 rdev_for_each_rcu(rdev, mddev)
907 struct mddev *mddev = rdev->mddev;
912 md_error(mddev, rdev);
915 set_bit(MD_SB_NEED_REWRITE, &mddev->sb_flags);
923 rdev_dec_pending(rdev, mddev);
925 if (atomic_dec_and_test(&mddev->pending_writes))
926 wake_up(&mddev->sb_wait);
929 void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
933 * Increment mddev->pending_writes before returning
950 GFP_NOIO, &mddev->sync_set);
959 if (test_bit(MD_FAILFAST_SUPPORTED, &mddev->flags) &&
964 atomic_inc(&mddev->pending_writes);
968 int md_super_wait(struct mddev *mddev)
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))
990 else if (rdev->mddev->reshape_position != MaxSector &&
991 (rdev->mddev->reshape_backwards ==
992 (sector >= rdev->mddev->reshape_position)))
1111 * int validate_super(struct mddev *mddev, struct md_rdev *dev)
1112 * Verify that dev is acceptable into mddev.
1113 * The first time, mddev->raid_disks will be 0, and data from
1117 * void sync_super(struct mddev *mddev, struct md_rdev *dev)
1118 * Update the superblock for rdev with data in mddev
1129 int (*validate_super)(struct mddev *mddev,
1132 void (*sync_super)(struct mddev *mddev,
1141 * Check that the given mddev has no bitmap.
1144 * support bitmaps. It prints an error message and returns non-zero if mddev
1148 int md_check_no_bitmap(struct mddev *mddev)
1150 if (!mddev->bitmap_info.file && !mddev->bitmap_info.offset)
1153 mdname(mddev), mddev->pers->name);
1270 static int super_90_validate(struct mddev *mddev, struct md_rdev *freshest, struct md_rdev *rdev)
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;
1322 mddev->recovery_cp = MaxSector;
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;
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) {
1352 if (ev1 < mddev->events)
1354 } else if (mddev->bitmap) {
1358 if (ev1 < mddev->bitmap->events_cleared)
1360 if (ev1 < mddev->events)
1363 if (ev1 < mddev->events)
1368 if (mddev->level != LEVEL_MULTIPATH) {
1374 desc->raid_disk < mddev->raid_disks */) {
1382 if (mddev->minor_version >= 91) {
1399 static void super_90_sync(struct mddev *mddev, struct md_rdev *rdev)
1403 int next_spare = mddev->raid_disks;
1405 /* make rdev->sb match mddev data..
1425 sb->major_version = mddev->major_version;
1426 sb->patch_version = mddev->patch_version;
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;
1439 sb->utime = clamp_t(time64_t, mddev->utime, 0, U32_MAX);
1441 sb->events_hi = (mddev->events>>32);
1442 sb->events_lo = (u32)mddev->events;
1444 if (mddev->reshape_position == MaxSector)
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)
1465 sb->layout = mddev->layout;
1466 sb->chunk_size = mddev->chunk_sectors << 9;
1468 if (mddev->bitmap && mddev->bitmap_info.file == NULL)
1472 rdev_for_each(rdev2, mddev) {
1520 for (i=0 ; i < mddev->raid_disks ; i++) {
1546 if (num_sectors && num_sectors < rdev->mddev->dev_sectors)
1548 if (rdev->mddev->bitmap_info.offset)
1556 if ((u64)num_sectors >= (2ULL << 32) && rdev->mddev->level >= 1)
1559 md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
1561 } while (md_super_wait(rdev->mddev) < 0);
1782 static int super_1_validate(struct mddev *mddev, struct md_rdev *freshest, struct md_rdev *rdev)
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;
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;
1821 mddev->bitmap_info.file == NULL) {
1822 mddev->bitmap_info.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;
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 &&
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 &&
1860 mddev->layout = -1;
1863 set_bit(MD_HAS_JOURNAL, &mddev->flags);
1874 set_bit(MD_HAS_PPL, &mddev->flags);
1876 } else if (mddev->pers == NULL) {
1886 if (ev1 + 1 < mddev->events)
1888 } else if (mddev->bitmap) {
1892 if (ev1 < mddev->bitmap->events_cleared)
1894 if (ev1 < mddev->events)
1897 if (ev1 < mddev->events)
1901 if (mddev->level != LEVEL_MULTIPATH) {
1907 } else if (mddev->pers == NULL && freshest && ev1 < mddev->events) {
1928 mdname(mddev), rdev->bdev, rdev->desc_nr,
1935 mdname(mddev), rdev->bdev, role, role, freshest->bdev);
1969 &mddev->recovery))
1987 static void super_1_sync(struct mddev *mddev, struct md_rdev *rdev)
1992 /* make rdev->sb match mddev and rdev data. */
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))
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);
2029 if (mddev->bitmap && mddev->bitmap_info.file == NULL) {
2030 sb->bitmap_offset = cpu_to_le32((__u32)mddev->bitmap_info.offset);
2040 if (rdev->saved_raid_disk >= 0 && mddev->bitmap)
2051 if (mddev->reshape_position != MaxSector) {
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)
2070 if (mddev_is_clustered(mddev))
2077 md_error(mddev, rdev);
2108 rdev_for_each(rdev2, mddev)
2125 if (test_bit(MD_HAS_JOURNAL, &mddev->flags))
2128 if (test_bit(MD_HAS_PPL, &mddev->flags)) {
2129 if (test_bit(MD_HAS_MULTIPLE_PPLS, &mddev->flags))
2138 rdev_for_each(rdev2, mddev) {
2178 if (num_sectors && num_sectors < rdev->mddev->dev_sectors)
2187 } else if (rdev->mddev->bitmap_info.offset) {
2215 md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
2217 } while (md_super_wait(rdev->mddev) < 0);
2233 if (rdev->mddev->minor_version == 0)
2244 bitmap = rdev->mddev->bitmap;
2245 if (bitmap && !rdev->mddev->bitmap_info.file &&
2246 rdev->sb_start + rdev->mddev->bitmap_info.offset +
2276 static void sync_super(struct mddev *mddev, struct md_rdev *rdev)
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);
2288 static int match_mddev_units(struct mddev *mddev1, struct mddev *mddev2)
2316 * Try to register data integrity profile for an mddev
2322 int md_integrity_register(struct mddev *mddev)
2326 if (list_empty(&mddev->disks))
2328 if (!mddev->gendisk || blk_get_integrity(mddev->gendisk))
2330 rdev_for_each(rdev, mddev) {
2352 blk_integrity_register(mddev->gendisk,
2355 pr_debug("md: data integrity enabled on %s\n", mdname(mddev));
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))) {
2366 mdname(mddev));
2377 int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev)
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);
2405 static int bind_rdev_to_array(struct md_rdev *rdev, struct mddev *mddev)
2411 if (find_rdev(mddev, rdev->bdev->bd_dev))
2414 if (rdev_read_only(rdev) && mddev->pers)
2417 /* make sure rdev->sectors exceeds mddev->dev_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)
2429 mddev->dev_sectors = rdev->sectors;
2439 if (mddev->pers)
2440 choice = mddev->raid_disks;
2441 while (md_find_rdev_nr_rcu(mddev, choice))
2445 if (md_find_rdev_nr_rcu(mddev, rdev->desc_nr)) {
2452 mddev->max_disks && rdev->desc_nr >= mddev->max_disks) {
2454 mdname(mddev), mddev->max_disks);
2460 rdev->mddev = mddev;
2463 if (mddev->raid_disks)
2464 mddev_create_serial_pool(mddev, rdev, false);
2466 if ((err = kobject_add(&rdev->kobj, &mddev->kobj, "dev-%s", b)))
2477 list_add_rcu(&rdev->same_set, &mddev->disks);
2478 bd_link_disk_holder(rdev->bdev, mddev->gendisk);
2481 mddev->recovery_disabled++;
2487 b, mdname(mddev));
2496 static void export_rdev(struct md_rdev *rdev, struct mddev *mddev)
2512 struct mddev *mddev = rdev->mddev;
2514 bd_unlink_disk_holder(rdev->bdev, rdev->mddev->gendisk);
2517 mddev_destroy_serial_pool(rdev->mddev, rdev, false);
2518 rdev->mddev = NULL;
2535 list_add(&rdev->same_set, &mddev->deleting);
2538 static void export_array(struct mddev *mddev)
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;
2551 static bool set_in_sync(struct mddev *mddev)
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;
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;
2578 static void sync_sbs(struct mddev *mddev, int nospares)
2587 rdev_for_each(rdev, mddev) {
2588 if (rdev->sb_events == mddev->events ||
2591 rdev->sb_events+1 == mddev->events)) {
2595 sync_super(mddev, rdev);
2601 static bool does_sb_need_changing(struct mddev *mddev)
2608 rdev_for_each(iter, mddev)
2620 rdev_for_each(rdev, mddev) {
2631 /* Check if any mddev parameters have changed */
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)))
2642 void md_update_sb(struct mddev *mddev, int force_change)
2650 if (!md_is_rdwr(mddev)) {
2652 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
2657 if (mddev_is_clustered(mddev)) {
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);
2664 if (!does_sb_need_changing(mddev)) {
2666 md_cluster_ops->metadata_update_cancel(mddev);
2667 bit_clear_unless(&mddev->sb_flags, BIT(MD_SB_CHANGE_PENDING),
2680 rdev_for_each(rdev, mddev) {
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) &&
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);
2697 rdev_for_each(rdev, mddev) {
2701 md_error(mddev, rdev);
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))
2726 if (mddev->degraded)
2738 sync_req = mddev->in_sync;
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;
2759 WARN_ON(mddev->events == 0);
2761 rdev_for_each(rdev, mddev) {
2768 sync_sbs(mddev, nospares);
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);
2778 rdev_for_each(rdev, mddev) {
2783 md_super_write(mddev,rdev,
2789 rdev->sb_events = mddev->events;
2791 md_super_write(mddev, rdev,
2802 if (mddev->level == LEVEL_MULTIPATH)
2806 if (md_super_wait(mddev) < 0)
2810 if (mddev_is_clustered(mddev) && ret == 0)
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);
2822 rdev_for_each(rdev, mddev) {
2836 struct mddev *mddev = rdev->mddev;
2840 if (!mddev->pers->hot_remove_disk || add_journal) {
2845 super_types[mddev->major_version].
2846 validate_super(mddev, NULL/*freshest*/, rdev);
2848 mddev_suspend(mddev);
2849 err = mddev->pers->hot_add_disk(mddev, rdev);
2851 mddev_resume(mddev);
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);
2954 struct mddev *mddev = rdev->mddev;
2958 if (cmd_match(buf, "faulty") && rdev->mddev->pers) {
2959 md_error(rdev->mddev, rdev);
2961 if (test_bit(MD_BROKEN, &rdev->mddev->flags))
2966 if (rdev->mddev->pers) {
2968 remove_and_add_spares(rdev->mddev, rdev);
2974 if (mddev_is_clustered(mddev))
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);
2988 mddev_create_serial_pool(rdev->mddev, rdev, false);
2992 mddev_destroy_serial_pool(rdev->mddev, rdev, false);
3006 md_error(rdev->mddev, rdev);
3011 set_bit(MD_RECOVERY_NEEDED, &rdev->mddev->recovery);
3012 md_wakeup_thread(rdev->mddev->thread);
3028 if (rdev->mddev->pers == NULL) {
3049 set_bit(MD_RECOVERY_NEEDED, &rdev->mddev->recovery);
3050 md_wakeup_thread(rdev->mddev->thread);
3063 if (rdev->mddev->pers)
3071 if (rdev->mddev->pers)
3078 if (!rdev->mddev->pers)
3088 if (!mddev_is_clustered(rdev->mddev) ||
3095 } else if (cmd_match(buf, "external_bbl") && (rdev->mddev->external)) {
3099 } else if (cmd_match(buf, "-external_bbl") && (rdev->mddev->external)) {
3104 md_update_sb(mddev, 1);
3162 if (rdev->mddev->pers && slot == -1) {
3173 if (rdev->mddev->pers->hot_remove_disk == NULL)
3176 remove_and_add_spares(rdev->mddev, rdev);
3179 set_bit(MD_RECOVERY_NEEDED, &rdev->mddev->recovery);
3180 md_wakeup_thread(rdev->mddev->thread);
3181 } else if (rdev->mddev->pers) {
3190 if (test_bit(MD_RECOVERY_RUNNING, &rdev->mddev->recovery))
3193 if (rdev->mddev->pers->hot_add_disk == NULL)
3196 if (slot >= rdev->mddev->raid_disks &&
3197 slot >= rdev->mddev->raid_disks + rdev->mddev->delta_disks)
3207 err = rdev->mddev->pers->hot_add_disk(rdev->mddev, rdev);
3214 sysfs_link_rdev(rdev->mddev, rdev);
3217 if (slot >= rdev->mddev->raid_disks &&
3218 slot >= rdev->mddev->raid_disks + rdev->mddev->delta_disks)
3245 if (rdev->mddev->pers && rdev->raid_disk >= 0)
3247 if (rdev->sectors && rdev->mddev->external)
3269 struct mddev *mddev = rdev->mddev;
3274 if (mddev->sync_thread ||
3275 test_bit(MD_RECOVERY_RUNNING,&mddev->recovery))
3283 + mddev->dev_sectors > rdev->sectors)
3292 mddev->reshape_backwards)
3299 !mddev->reshape_backwards)
3302 if (mddev->pers && mddev->persistent &&
3303 !super_types[mddev->major_version]
3308 mddev->reshape_backwards = 1;
3310 mddev->reshape_backwards = 0;
3335 struct mddev *mddev;
3339 list_for_each_entry(mddev, &all_mddevs, all_mddevs) {
3340 if (test_bit(MD_DELETED, &mddev->flags))
3342 rdev_for_each(rdev2, mddev) {
3376 struct mddev *my_mddev = rdev->mddev;
3445 if (rdev->mddev->pers &&
3513 if (rdev->mddev->pers && test_bit(MD_HAS_PPL, &rdev->mddev->flags) &&
3517 if (rdev->mddev->persistent) {
3518 if (rdev->mddev->major_version == 0)
3526 } else if (!rdev->mddev->external) {
3550 if (rdev->mddev->pers && test_bit(MD_HAS_PPL, &rdev->mddev->flags) &&
3554 if (rdev->mddev->persistent) {
3555 if (rdev->mddev->major_version == 0)
3559 } else if (!rdev->mddev->external) {
3592 if (!rdev->mddev)
3605 struct mddev *mddev = rdev->mddev;
3615 rv = mddev ? mddev_lock(mddev) : -ENODEV;
3617 if (rdev->mddev == NULL)
3621 mddev_unlock(mddev);
3757 static int analyze_sbs(struct mddev *mddev)
3763 rdev_for_each_safe(rdev, tmp, mddev)
3764 switch (super_types[mddev->major_version].
3765 load_super(rdev, freshest, mddev->minor_version)) {
3783 super_types[mddev->major_version].
3784 validate_super(mddev, NULL/*freshest*/, freshest);
3787 rdev_for_each_safe(rdev, tmp, mddev) {
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].
3799 validate_super(mddev, freshest, rdev)) {
3806 if (mddev->level == LEVEL_MULTIPATH) {
3811 (mddev->raid_disks - min(0, mddev->delta_disks)) &&
3858 safe_delay_show(struct mddev *mddev, char *page)
3860 unsigned int msec = ((unsigned long)mddev->safemode_delay*1000)/HZ;
3865 safe_delay_store(struct mddev *mddev, const char *cbuf, size_t len)
3869 if (mddev_is_clustered(mddev)) {
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);
3894 level_show(struct mddev *mddev, char *page)
3898 spin_lock(&mddev->lock);
3899 p = mddev->pers;
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);
3913 level_store(struct mddev *mddev, const char *buf, size_t len)
3926 rv = mddev_lock(mddev);
3930 if (mddev->pers == NULL) {
3931 strncpy(mddev->clevel, buf, slen);
3932 if (mddev->clevel[slen-1] == '\n')
3934 mddev->clevel[slen] = 0;
3935 mddev->level = LEVEL_NONE;
3940 if (!md_is_rdwr(mddev))
3950 if (mddev->sync_thread ||
3951 test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
3952 mddev->reshape_position != MaxSector ||
3953 mddev->sysfs_active)
3957 if (!mddev->pers->quiesce) {
3959 mdname(mddev), mddev->pers->name);
3983 if (pers == mddev->pers) {
3992 mdname(mddev), clevel);
3997 rdev_for_each(rdev, mddev)
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;
4013 mdname(mddev), clevel);
4019 mddev_suspend(mddev);
4020 mddev_detach(mddev);
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);
4037 mddev->external) {
4045 mddev->in_sync = 0;
4046 mddev->safemode_delay = 0;
4047 mddev->safemode = 0;
4050 oldpers->free(mddev, oldpriv);
4055 if (sysfs_create_group(&mddev->kobj, &md_redundancy_group))
4057 mdname(mddev));
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");
4065 if (mddev->to_remove == NULL)
4066 mddev->to_remove = &md_redundancy_group;
4071 rdev_for_each(rdev, mddev) {
4074 if (rdev->new_raid_disk >= mddev->raid_disks)
4078 sysfs_unlink_rdev(mddev, rdev);
4080 rdev_for_each(rdev, mddev) {
4089 if (sysfs_link_rdev(mddev, rdev))
4091 rdev->raid_disk, mdname(mddev));
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);
4105 mddev_resume(mddev);
4106 if (!mddev->thread)
4107 md_update_sb(mddev, 1);
4108 sysfs_notify_dirent_safe(mddev->sysfs_level);
4112 mddev_unlock(mddev);
4120 layout_show(struct mddev *mddev, char *page)
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);
4131 layout_store(struct mddev *mddev, const char *buf, size_t len)
4139 err = mddev_lock(mddev);
4143 if (mddev->pers) {
4144 if (mddev->pers->check_reshape == NULL)
4146 else if (!md_is_rdwr(mddev))
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;
4159 mddev_unlock(mddev);
4166 raid_disks_show(struct mddev *mddev, char *page)
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);
4177 static int update_raid_disks(struct mddev *mddev, int raid_disks);
4180 raid_disks_store(struct mddev *mddev, const char *buf, size_t len)
4189 err = mddev_lock(mddev);
4192 if (mddev->pers)
4193 err = update_raid_disks(mddev, n);
4194 else if (mddev->reshape_position != MaxSector) {
4196 int olddisks = mddev->raid_disks - mddev->delta_disks;
4199 rdev_for_each(rdev, mddev) {
4208 mddev->delta_disks = n - olddisks;
4209 mddev->raid_disks = n;
4210 mddev->reshape_backwards = (mddev->delta_disks < 0);
4212 mddev->raid_disks = n;
4214 mddev_unlock(mddev);
4221 uuid_show(struct mddev *mddev, char *page)
4223 return sprintf(page, "%pU\n", mddev->uuid);
4229 chunk_size_show(struct mddev *mddev, char *page)
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);
4240 chunk_size_store(struct mddev *mddev, const char *buf, size_t len)
4249 err = mddev_lock(mddev);
4252 if (mddev->pers) {
4253 if (mddev->pers->check_reshape == NULL)
4255 else if (!md_is_rdwr(mddev))
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;
4268 mddev_unlock(mddev);
4275 resync_start_show(struct mddev *mddev, char *page)
4277 if (mddev->recovery_cp == MaxSector)
4279 return sprintf(page, "%llu\n", (unsigned long long)mddev->recovery_cp);
4283 resync_start_store(struct mddev *mddev, const char *buf, size_t len)
4298 err = mddev_lock(mddev);
4301 if (mddev->pers && !test_bit(MD_RECOVERY_FROZEN, &mddev->recovery))
4305 mddev->recovery_cp = n;
4306 if (mddev->pers)
4307 set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
4309 mddev_unlock(mddev);
4372 array_state_show(struct mddev *mddev, char *page)
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)
4410 static int do_md_stop(struct mddev *mddev, int ro, struct block_device *bdev);
4411 static int md_set_readonly(struct mddev *mddev, struct block_device *bdev);
4412 static int restart_array(struct mddev *mddev);
4415 array_state_store(struct mddev *mddev, const char *buf, size_t len)
4420 if (mddev->pers && (st == active || st == clean) &&
4421 mddev->ro != MD_RDONLY) {
4425 spin_lock(&mddev->lock);
4427 restart_array(mddev);
4428 clear_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
4429 md_wakeup_thread(mddev->thread);
4430 wake_up(&mddev->sb_wait);
4432 restart_array(mddev);
4433 if (!set_in_sync(mddev))
4437 sysfs_notify_dirent_safe(mddev->sysfs_state);
4438 spin_unlock(&mddev->lock);
4441 err = mddev_lock(mddev);
4450 err = do_md_stop(mddev, 0, NULL);
4454 if (mddev->pers)
4455 err = do_md_stop(mddev, 2, NULL);
4462 if (mddev->pers)
4463 err = md_set_readonly(mddev, NULL);
4465 mddev->ro = MD_RDONLY;
4466 set_disk_ro(mddev->gendisk, 1);
4467 err = do_md_run(mddev);
4471 if (mddev->pers) {
4472 if (md_is_rdwr(mddev))
4473 err = md_set_readonly(mddev, NULL);
4474 else if (mddev->ro == MD_RDONLY)
4475 err = restart_array(mddev);
4477 mddev->ro = MD_AUTO_READ;
4478 set_disk_ro(mddev->gendisk, 0);
4481 mddev->ro = MD_AUTO_READ;
4482 err = do_md_run(mddev);
4486 if (mddev->pers) {
4487 err = restart_array(mddev);
4490 spin_lock(&mddev->lock);
4491 if (!set_in_sync(mddev))
4493 spin_unlock(&mddev->lock);
4498 if (mddev->pers) {
4499 err = restart_array(mddev);
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);
4508 err = do_md_run(mddev);
4519 if (mddev->hold_active == UNTIL_IOCTL)
4520 mddev->hold_active = 0;
4521 sysfs_notify_dirent_safe(mddev->sysfs_state);
4523 mddev_unlock(mddev);
4530 max_corrected_read_errors_show(struct mddev *mddev, char *page) {
4532 atomic_read(&mddev->max_corr_read_errors));
4536 max_corrected_read_errors_store(struct mddev *mddev, const char *buf, size_t len)
4546 atomic_set(&mddev->max_corr_read_errors, n);
4555 null_show(struct mddev *mddev, char *page)
4561 new_dev_store(struct mddev *mddev, const char *buf, size_t len)
4587 err = mddev_lock(mddev);
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)
4608 mddev_unlock(mddev);
4611 err = bind_rdev_to_array(rdev, mddev);
4614 export_rdev(rdev, mddev);
4615 mddev_unlock(mddev);
4625 bitmap_store(struct mddev *mddev, const char *buf, size_t len)
4631 err = mddev_lock(mddev);
4634 if (!mddev->bitmap)
4646 md_bitmap_dirty_bits(mddev->bitmap, chunk, end_chunk);
4649 md_bitmap_unplug(mddev->bitmap); /* flush the bits to disk */
4651 mddev_unlock(mddev);
4659 size_show(struct mddev *mddev, char *page)
4662 (unsigned long long)mddev->dev_sectors / 2);
4665 static int update_size(struct mddev *mddev, sector_t num_sectors);
4668 size_store(struct mddev *mddev, const char *buf, size_t len)
4679 err = mddev_lock(mddev);
4682 if (mddev->pers) {
4683 err = update_size(mddev, sectors);
4685 md_update_sb(mddev, 1);
4687 if (mddev->dev_sectors == 0 ||
4688 mddev->dev_sectors > sectors)
4689 mddev->dev_sectors = sectors;
4693 mddev_unlock(mddev);
4707 metadata_show(struct mddev *mddev, char *page)
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);
4719 metadata_store(struct mddev *mddev, const char *buf, size_t len)
4729 err = mddev_lock(mddev);
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;
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;
4771 mddev->major_version = major;
4772 mddev->minor_version = minor;
4773 mddev->persistent = 1;
4774 mddev->external = 0;
4777 mddev_unlock(mddev);
4785 action_show(struct mddev *mddev, char *page)
4788 unsigned long recovery = mddev->recovery;
4792 (md_is_rdwr(mddev) && test_bit(MD_RECOVERY_NEEDED, &recovery))) {
4804 else if (mddev->reshape_position != MaxSector)
4810 static void stop_sync_thread(struct mddev *mddev)
4812 if (!test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
4815 if (mddev_lock(mddev))
4822 if (!test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) {
4823 mddev_unlock(mddev);
4827 if (work_pending(&mddev->del_work))
4830 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
4835 md_wakeup_thread_directly(mddev->sync_thread);
4837 mddev_unlock(mddev);
4840 static void idle_sync_thread(struct mddev *mddev)
4842 int sync_seq = atomic_read(&mddev->sync_seq);
4844 mutex_lock(&mddev->sync_mutex);
4845 clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4846 stop_sync_thread(mddev);
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);
4854 static void frozen_sync_thread(struct mddev *mddev)
4856 mutex_lock(&mddev->sync_mutex);
4857 set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
4858 stop_sync_thread(mddev);
4860 wait_event(resync_wait, mddev->sync_thread == NULL &&
4861 !test_bit(MD_RECOVERY_RUNNING, &mddev->recovery));
4863 mutex_unlock(&mddev->sync_mutex);
4867 action_store(struct mddev *mddev, const char *page, size_t len)
4869 if (!mddev->pers || !mddev->pers->sync_request)
4874 idle_sync_thread(mddev);
4876 frozen_sync_thread(mddev);
4877 else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
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)
4888 err = mddev_lock(mddev);
4890 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) {
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);
4906 mddev_unlock(mddev);
4910 sysfs_notify_dirent_safe(mddev->sysfs_degraded);
4913 set_bit(MD_RECOVERY_CHECK, &mddev->recovery);
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) {
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);
4937 last_sync_action_show(struct mddev *mddev, char *page)
4939 return sprintf(page, "%s\n", mddev->last_sync_action);
4945 mismatch_cnt_show(struct mddev *mddev, char *page)
4949 atomic64_read(&mddev->resync_mismatches));
4955 sync_min_show(struct mddev *mddev, char *page)
4957 return sprintf(page, "%d (%s)\n", speed_min(mddev),
4958 mddev->sync_speed_min ? "local": "system");
4962 sync_min_store(struct mddev *mddev, const char *buf, size_t len)
4976 mddev->sync_speed_min = min;
4984 sync_max_show(struct mddev *mddev, char *page)
4986 return sprintf(page, "%d (%s)\n", speed_max(mddev),
4987 mddev->sync_speed_max ? "local": "system");
4991 sync_max_store(struct mddev *mddev, const char *buf, size_t len)
5005 mddev->sync_speed_max = max;
5013 degraded_show(struct mddev *mddev, char *page)
5015 return sprintf(page, "%d\n", mddev->degraded);
5020 sync_force_parallel_show(struct mddev *mddev, char *page)
5022 return sprintf(page, "%d\n", mddev->parallel_resync);
5026 sync_force_parallel_store(struct mddev *mddev, const char *buf, size_t len)
5036 mddev->parallel_resync = n;
5038 if (mddev->sync_thread)
5050 sync_speed_show(struct mddev *mddev, char *page)
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;
5065 sync_completed_show(struct mddev *mddev, char *page)
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;
5090 min_sync_show(struct mddev *mddev, char *page)
5093 (unsigned long long)mddev->resync_min);
5096 min_sync_store(struct mddev *mddev, const char *buf, size_t len)
5104 spin_lock(&mddev->lock);
5106 if (min > mddev->resync_max)
5110 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
5114 mddev->resync_min = round_down(min, 8);
5118 spin_unlock(&mddev->lock);
5126 max_sync_show(struct mddev *mddev, char *page)
5128 if (mddev->resync_max == MaxSector)
5132 (unsigned long long)mddev->resync_max);
5135 max_sync_store(struct mddev *mddev, const char *buf, size_t len)
5138 spin_lock(&mddev->lock);
5140 mddev->resync_max = MaxSector;
5148 if (max < mddev->resync_min)
5152 if (max < mddev->resync_max && md_is_rdwr(mddev) &&
5153 test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
5157 chunk = mddev->chunk_sectors;
5165 mddev->resync_max = max;
5167 wake_up(&mddev->recovery_wait);
5170 spin_unlock(&mddev->lock);
5178 suspend_lo_show(struct mddev *mddev, char *page)
5180 return sprintf(page, "%llu\n", (unsigned long long)mddev->suspend_lo);
5184 suspend_lo_store(struct mddev *mddev, const char *buf, size_t len)
5195 err = mddev_lock(mddev);
5199 if (mddev->pers == NULL ||
5200 mddev->pers->quiesce == NULL)
5202 mddev_suspend(mddev);
5203 mddev->suspend_lo = new;
5204 mddev_resume(mddev);
5208 mddev_unlock(mddev);
5215 suspend_hi_show(struct mddev *mddev, char *page)
5217 return sprintf(page, "%llu\n", (unsigned long long)mddev->suspend_hi);
5221 suspend_hi_store(struct mddev *mddev, const char *buf, size_t len)
5232 err = mddev_lock(mddev);
5236 if (mddev->pers == NULL)
5239 mddev_suspend(mddev);
5240 mddev->suspend_hi = new;
5241 mddev_resume(mddev);
5245 mddev_unlock(mddev);
5252 reshape_position_show(struct mddev *mddev, char *page)
5254 if (mddev->reshape_position != MaxSector)
5256 (unsigned long long)mddev->reshape_position);
5262 reshape_position_store(struct mddev *mddev, const char *buf, size_t len)
5273 err = mddev_lock(mddev);
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;
5285 rdev_for_each(rdev, mddev)
5289 mddev_unlock(mddev);
5298 reshape_direction_show(struct mddev *mddev, char *page)
5301 mddev->reshape_backwards ? "backwards" : "forwards");
5305 reshape_direction_store(struct mddev *mddev, const char *buf, size_t len)
5316 if (mddev->reshape_backwards == backwards)
5319 err = mddev_lock(mddev);
5323 if (mddev->delta_disks)
5325 else if (mddev->persistent &&
5326 mddev->major_version == 0)
5329 mddev->reshape_backwards = backwards;
5330 mddev_unlock(mddev);
5339 array_size_show(struct mddev *mddev, char *page)
5341 if (mddev->external_size)
5343 (unsigned long long)mddev->array_sectors/2);
5349 array_size_store(struct mddev *mddev, const char *buf, size_t len)
5354 err = mddev_lock(mddev);
5359 if (mddev_is_clustered(mddev)) {
5360 mddev_unlock(mddev);
5365 if (mddev->pers)
5366 sectors = mddev->pers->size(mddev, 0, 0);
5368 sectors = mddev->array_sectors;
5370 mddev->external_size = 0;
5374 else if (mddev->pers && mddev->pers->size(mddev, 0, 0) < sectors)
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);
5386 mddev_unlock(mddev);
5395 consistency_policy_show(struct mddev *mddev, char *page)
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)
5418 consistency_policy_store(struct mddev *mddev, const char *buf, size_t len)
5422 if (mddev->pers) {
5423 if (mddev->pers->change_consistency_policy)
5424 err = mddev->pers->change_consistency_policy(mddev, buf);
5427 } else if (mddev->external && strncmp(buf, "ppl", 3) == 0) {
5428 set_bit(MD_HAS_PPL, &mddev->flags);
5440 static ssize_t fail_last_dev_show(struct mddev *mddev, char *page)
5442 return sprintf(page, "%d\n", mddev->fail_last_dev);
5450 fail_last_dev_store(struct mddev *mddev, const char *buf, size_t len)
5459 if (value != mddev->fail_last_dev)
5460 mddev->fail_last_dev = value;
5468 static ssize_t serialize_policy_show(struct mddev *mddev, char *page)
5470 if (mddev->pers == NULL || (mddev->pers->level != 1))
5473 return sprintf(page, "%d\n", mddev->serialize_policy);
5481 serialize_policy_store(struct mddev *mddev, const char *buf, size_t len)
5490 if (value == mddev->serialize_policy)
5493 err = mddev_lock(mddev);
5496 if (mddev->pers == NULL || (mddev->pers->level != 1)) {
5502 mddev_suspend(mddev);
5504 mddev_create_serial_pool(mddev, NULL, true);
5506 mddev_destroy_serial_pool(mddev, NULL, true);
5507 mddev->serialize_policy = value;
5508 mddev_resume(mddev);
5510 mddev_unlock(mddev);
5577 struct mddev *mddev = container_of(kobj, struct mddev, kobj);
5583 if (!mddev_get(mddev)) {
5589 rv = entry->show(mddev, page);
5590 mddev_put(mddev);
5599 struct mddev *mddev = container_of(kobj, struct mddev, kobj);
5607 if (!mddev_get(mddev)) {
5612 rv = entry->store(mddev, page, length);
5613 mddev_put(mddev);
5619 struct mddev *mddev = container_of(ko, struct mddev, kobj);
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);
5644 struct mddev *mddev = container_of(ws, struct mddev, del_work);
5646 kobject_put(&mddev->kobj);
5651 int mddev_init_writes_pending(struct mddev *mddev)
5653 if (mddev->writes_pending.percpu_count_ptr)
5655 if (percpu_ref_init(&mddev->writes_pending, no_op,
5659 percpu_ref_put(&mddev->writes_pending);
5664 struct mddev *md_alloc(dev_t dev, char *name)
5676 struct mddev *mddev;
5690 mddev = mddev_alloc(dev);
5691 if (IS_ERR(mddev)) {
5692 error = PTR_ERR(mddev);
5696 partitioned = (MAJOR(mddev->unit) != MD_MAJOR);
5698 unit = MINOR(mddev->unit) >> shift;
5703 struct mddev *mddev2;
5719 mddev->hold_active = UNTIL_STOP;
5726 disk->major = MAJOR(mddev->unit);
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);
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;
5757 mddev_put(mddev);
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");
5765 return mddev;
5770 mddev_free(mddev);
5778 struct mddev *mddev = md_alloc(dev, name);
5780 if (IS_ERR(mddev))
5781 return PTR_ERR(mddev);
5782 mddev_put(mddev);
5825 struct mddev *mddev = from_timer(mddev, t, safemode_timer);
5827 mddev->safemode = 1;
5828 if (mddev->external)
5829 sysfs_notify_dirent_safe(mddev->sysfs_state);
5831 md_wakeup_thread(mddev->thread);
5837 struct mddev *mddev = container_of(ref, struct mddev, active_io);
5839 wake_up(&mddev->sb_wait);
5842 int md_run(struct mddev *mddev)
5849 if (list_empty(&mddev->disks))
5853 if (mddev->pers)
5856 if (mddev->sysfs_active)
5862 if (!mddev->raid_disks) {
5863 if (!mddev->persistent)
5865 err = analyze_sbs(mddev);
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;
5881 rdev_for_each(rdev, mddev) {
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);
5893 mddev->has_superblocks = true;
5902 if (mddev->dev_sectors &&
5903 rdev->data_offset + mddev->dev_sectors
5906 mdname(mddev));
5913 mdname(mddev));
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);
5948 if (mddev->level != LEVEL_NONE)
5950 mddev->level);
5953 mddev->clevel);
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 &&
5979 rdev_for_each(rdev, mddev)
5980 rdev_for_each(rdev2, mddev) {
5985 mdname(mddev),
5996 mddev->recovery = 0;
5998 mddev->resync_max_sectors = mddev->dev_sectors;
6000 mddev->ok_start_degraded = start_dirty_degraded;
6002 if (start_readonly && md_is_rdwr(mddev))
6003 mddev->ro = MD_AUTO_READ; /* read-only, but switch on first write */
6005 err = pers->run(mddev);
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);
6018 (mddev->bitmap_info.file || mddev->bitmap_info.offset)) {
6021 bitmap = md_bitmap_create(mddev, -1);
6025 mdname(mddev), err);
6027 mddev->bitmap = bitmap;
6033 if (mddev->bitmap_info.max_write_behind > 0) {
6036 rdev_for_each(rdev, mddev) {
6041 if (create_pool && mddev->serial_info_pool == NULL) {
6042 mddev->serial_info_pool =
6045 if (!mddev->serial_info_pool) {
6052 if (mddev->queue) {
6055 rdev_for_each(rdev, mddev) {
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);
6074 if (mddev->kobj.sd &&
6075 sysfs_create_group(&mddev->kobj, &md_redundancy_group))
6077 mdname(mddev));
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;
6087 if (mddev_is_clustered(mddev))
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);
6096 rdev_for_each(rdev, mddev)
6098 sysfs_link_rdev(mddev, rdev); /* failure here is OK */
6100 if (mddev->degraded && md_is_rdwr(mddev))
6104 set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
6105 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
6107 if (mddev->sb_flags)
6108 md_update_sb(mddev, 0);
6114 mddev_detach(mddev);
6115 if (mddev->private)
6116 pers->free(mddev, mddev->private);
6117 mddev->private = NULL;
6119 md_bitmap_destroy(mddev);
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);
6132 int do_md_run(struct mddev *mddev)
6136 set_bit(MD_NOT_READY, &mddev->flags);
6137 err = md_run(mddev);
6140 err = md_bitmap_load(mddev);
6142 md_bitmap_destroy(mddev);
6146 if (mddev_is_clustered(mddev))
6147 md_allow_write(mddev);
6150 md_start(mddev);
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);
6167 int md_start(struct mddev *mddev)
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);
6182 static int restart_array(struct mddev *mddev)
6184 struct gendisk *disk = mddev->gendisk;
6190 if (list_empty(&mddev->disks))
6192 if (!mddev->pers)
6194 if (md_is_rdwr(mddev))
6198 rdev_for_each_rcu(rdev, mddev) {
6206 if (test_bit(MD_HAS_JOURNAL, &mddev->flags) && !has_journal)
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);
6224 static void md_clean(struct mddev *mddev)
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;
6239 * Don't clear MD_CLOSING, or mddev can be opened again.
6240 * 'hold_active != 0' means mddev is still in the creation
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;
6281 static void __md_stop_writes(struct mddev *mddev)
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);
6288 md_reap_sync_thread(mddev);
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);
6297 md_bitmap_flush(mddev);
6299 if (md_is_rdwr(mddev) &&
6300 ((!mddev->in_sync && !mddev_is_clustered(mddev)) ||
6301 mddev->sb_flags)) {
6303 if (!mddev_is_clustered(mddev))
6304 mddev->in_sync = 1;
6305 md_update_sb(mddev, 1);
6308 mddev->serialize_policy = 0;
6309 mddev_destroy_serial_pool(mddev, NULL, true);
6312 void md_stop_writes(struct mddev *mddev)
6314 mddev_lock_nointr(mddev);
6315 __md_stop_writes(mddev);
6316 mddev_unlock(mddev);
6320 static void mddev_detach(struct mddev *mddev)
6322 md_bitmap_wait_behind_writes(mddev);
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'*/
6332 static void __md_stop(struct mddev *mddev)
6334 struct md_personality *pers = mddev->pers;
6335 md_bitmap_destroy(mddev);
6336 mddev_detach(mddev);
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;
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);
6357 void md_stop(struct mddev *mddev)
6359 lockdep_assert_held(&mddev->reconfig_mutex);
6364 __md_stop_writes(mddev);
6365 __md_stop(mddev);
6366 percpu_ref_exit(&mddev->writes_pending);
6371 static int md_set_readonly(struct mddev *mddev, struct block_device *bdev)
6376 if (mddev->external && test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags))
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);
6393 mddev_unlock(mddev);
6395 &mddev->recovery));
6396 wait_event(mddev->sb_wait,
6397 !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags));
6398 mddev_lock_nointr(mddev);
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)) {
6404 pr_warn("md: %s still in use.\n",mdname(mddev));
6409 if (mddev->pers) {
6410 __md_stop_writes(mddev);
6412 if (mddev->ro == MD_RDONLY) {
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);
6437 static int do_md_stop(struct mddev *mddev, int mode,
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);
6458 mddev_unlock(mddev);
6459 wait_event(resync_wait, (mddev->sync_thread == NULL &&
6461 &mddev->recovery)));
6462 mddev_lock_nointr(mddev);
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)) {
6469 pr_warn("md: %s still in use.\n",mdname(mddev));
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);
6478 if (mddev->pers) {
6479 if (!md_is_rdwr(mddev))
6482 __md_stop_writes(mddev);
6483 __md_stop(mddev);
6486 sysfs_notify_dirent_safe(mddev->sysfs_state);
6488 rdev_for_each(rdev, mddev)
6490 sysfs_unlink_rdev(mddev, rdev);
6493 mutex_unlock(&mddev->open_mutex);
6494 mddev->changed = 1;
6496 if (!md_is_rdwr(mddev))
6497 mddev->ro = MD_RDWR;
6499 mutex_unlock(&mddev->open_mutex);
6504 pr_info("md: %s stopped.\n", mdname(mddev));
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;
6515 export_array(mddev);
6517 md_clean(mddev);
6518 if (mddev->hold_active == UNTIL_STOP)
6519 mddev->hold_active = 0;
6522 sysfs_notify_dirent_safe(mddev->sysfs_state);
6527 static void autorun_array(struct mddev *mddev)
6532 if (list_empty(&mddev->disks))
6537 rdev_for_each(rdev, mddev) {
6542 err = do_md_run(mddev);
6545 do_md_stop(mddev, 0, NULL);
6564 struct mddev *mddev;
6585 * mddev.
6601 mddev = md_alloc(dev, NULL);
6602 if (IS_ERR(mddev))
6605 if (mddev_lock(mddev))
6606 pr_warn("md: %s locked, cannot run\n", mdname(mddev));
6607 else if (mddev->raid_disks || mddev->major_version
6608 || !list_empty(&mddev->disks)) {
6610 mdname(mddev), rdev0->bdev);
6611 mddev_unlock(mddev);
6613 pr_debug("md: created %s\n", mdname(mddev));
6614 mddev->persistent = 1;
6617 if (bind_rdev_to_array(rdev, mddev))
6618 export_rdev(rdev, mddev);
6620 autorun_array(mddev);
6621 mddev_unlock(mddev);
6628 export_rdev(rdev, mddev);
6630 mddev_put(mddev);
6650 static int get_array_info(struct mddev *mddev, void __user *arg)
6658 rdev_for_each_rcu(rdev, mddev) {
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 */
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)
6694 if (mddev_is_clustered(mddev))
6701 info.layout = mddev->layout;
6702 info.chunk_size = mddev->chunk_sectors << 9;
6710 static int get_bitmap_file(struct mddev *mddev, void __user * arg)
6721 spin_lock(&mddev->lock);
6723 if (mddev->bitmap_info.file) {
6724 ptr = file_path(mddev->bitmap_info.file, file->pathname,
6732 spin_unlock(&mddev->lock);
6742 static int get_disk_info(struct mddev *mddev, void __user * arg)
6751 rdev = md_find_rdev_nr_rcu(mddev, info.number);
6782 int md_add_new_disk(struct mddev *mddev, struct mdu_disk_info_s *info)
6787 if (mddev_is_clustered(mddev) &&
6789 pr_warn("%s: Cannot add to clustered mddev.\n",
6790 mdname(mddev));
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);
6816 export_rdev(rdev, mddev);
6820 err = bind_rdev_to_array(rdev, mddev);
6822 export_rdev(rdev, mddev);
6831 if (mddev->pers) {
6833 if (!mddev->pers->hot_add_disk) {
6835 mdname(mddev));
6838 if (mddev->persistent)
6839 rdev = md_import_device(dev, mddev->major_version,
6840 mddev->minor_version);
6849 if (!mddev->persistent) {
6851 info->raid_disk < mddev->raid_disks) {
6858 super_types[mddev->major_version].
6859 validate_super(mddev, NULL/*freshest*/, rdev);
6865 export_rdev(rdev, mddev);
6884 rdev_for_each(rdev2, mddev) {
6890 if (has_journal || mddev->bitmap) {
6891 export_rdev(rdev, mddev);
6899 if (mddev_is_clustered(mddev)) {
6904 err = md_cluster_ops->add_new_disk(mddev, rdev);
6906 export_rdev(rdev, mddev);
6913 err = bind_rdev_to_array(rdev, mddev);
6916 export_rdev(rdev, mddev);
6918 if (mddev_is_clustered(mddev)) {
6921 err = md_cluster_ops->new_disk_ack(mddev,
6928 md_cluster_ops->add_new_disk_cancel(mddev);
6942 if (mddev->major_version != 0) {
6943 pr_warn("%s: ADD_NEW_DISK not supported\n", mdname(mddev));
6956 if (info->raid_disk < mddev->raid_disks)
6961 if (rdev->raid_disk < mddev->raid_disks)
6970 if (!mddev->persistent) {
6977 err = bind_rdev_to_array(rdev, mddev);
6979 export_rdev(rdev, mddev);
6987 static int hot_remove_disk(struct mddev *mddev, dev_t dev)
6991 if (!mddev->pers)
6994 rdev = find_rdev(mddev, dev);
7002 remove_and_add_spares(mddev, rdev);
7008 if (mddev_is_clustered(mddev)) {
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);
7018 md_update_sb(mddev, 1);
7024 rdev->bdev, mdname(mddev));
7028 static int hot_add_disk(struct mddev *mddev, dev_t dev)
7033 if (!mddev->pers)
7036 if (mddev->major_version != 0) {
7038 mdname(mddev));
7041 if (!mddev->pers->hot_add_disk) {
7043 mdname(mddev));
7054 if (mddev->persistent)
7063 rdev->bdev, mdname(mddev));
7071 err = bind_rdev_to_array(rdev, mddev);
7082 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
7083 if (!mddev->thread)
7084 md_update_sb(mddev, 1);
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);
7104 export_rdev(rdev, mddev);
7108 static int set_bitmap_file(struct mddev *mddev, int fd)
7112 if (mddev->pers) {
7113 if (!mddev->pers->quiesce || !mddev->thread)
7115 if (mddev->recovery || mddev->sync_thread)
7124 if (mddev->bitmap || mddev->bitmap_info.file)
7129 mdname(mddev));
7133 mdname(mddev));
7139 mdname(mddev));
7146 mdname(mddev));
7150 mdname(mddev));
7154 mdname(mddev));
7161 mddev->bitmap_info.file = f;
7162 mddev->bitmap_info.offset = 0; /* file overrides offset */
7163 } else if (mddev->bitmap == NULL)
7166 if (mddev->pers) {
7170 bitmap = md_bitmap_create(mddev, -1);
7171 mddev_suspend(mddev);
7173 mddev->bitmap = bitmap;
7174 err = md_bitmap_load(mddev);
7178 md_bitmap_destroy(mddev);
7181 mddev_resume(mddev);
7183 mddev_suspend(mddev);
7184 md_bitmap_destroy(mddev);
7185 mddev_resume(mddev);
7189 struct file *f = mddev->bitmap_info.file;
7191 spin_lock(&mddev->lock);
7192 mddev->bitmap_info.file = NULL;
7193 spin_unlock(&mddev->lock);
7214 int md_set_array_info(struct mddev *mddev, struct mdu_array_info_s *info)
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;
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;
7288 void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors)
7290 lockdep_assert_held(&mddev->reconfig_mutex);
7292 if (mddev->external_size)
7295 mddev->array_sectors = array_sectors;
7299 static int update_size(struct mddev *mddev, sector_t num_sectors)
7304 sector_t old_dev_sectors = mddev->dev_sectors;
7306 if (mddev->pers->resize == NULL)
7317 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
7318 mddev->sync_thread)
7320 if (!md_is_rdwr(mddev))
7323 rdev_for_each(rdev, mddev) {
7331 rv = mddev->pers->resize(mddev, num_sectors);
7333 if (mddev_is_clustered(mddev))
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);
7343 static int update_raid_disks(struct mddev *mddev, int raid_disks)
7348 if (mddev->pers->check_reshape == NULL)
7350 if (!md_is_rdwr(mddev))
7353 (mddev->max_disks && raid_disks >= mddev->max_disks))
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)
7361 rdev_for_each(rdev, mddev) {
7362 if (mddev->raid_disks < raid_disks &&
7365 if (mddev->raid_disks > raid_disks &&
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;
7392 static int update_array_info(struct mddev *mddev, mdu_array_info_t *info)
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 ||
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)
7428 if (mddev->layout != info->layout) {
7433 if (mddev->pers->check_reshape == NULL)
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);
7450 if (mddev->pers->quiesce == NULL || mddev->thread == NULL) {
7454 if (mddev->recovery || mddev->sync_thread) {
7461 if (mddev->bitmap) {
7465 if (mddev->bitmap_info.default_offset == 0) {
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);
7474 mddev_suspend(mddev);
7476 mddev->bitmap = bitmap;
7477 rv = md_bitmap_load(mddev);
7481 md_bitmap_destroy(mddev);
7482 mddev_resume(mddev);
7485 if (!mddev->bitmap) {
7489 if (mddev->bitmap->storage.file) {
7493 if (mddev->bitmap_info.nodes) {
7495 if (md_cluster_ops->lock_all_bitmaps(mddev) <= 0) {
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;
7507 mddev_suspend(mddev);
7508 md_bitmap_destroy(mddev);
7509 mddev_resume(mddev);
7510 mddev->bitmap_info.offset = 0;
7513 md_update_sb(mddev, 1);
7519 static int set_disk_faulty(struct mddev *mddev, dev_t dev)
7524 if (mddev->pers == NULL)
7528 rdev = md_find_rdev_rcu(mddev, dev);
7532 md_error(mddev, rdev);
7533 if (test_bit(MD_BROKEN, &mddev->flags))
7548 struct mddev *mddev = bdev->bd_disk->private_data;
7552 geo->cylinders = mddev->array_sectors / 8;
7580 static int __md_set_array_info(struct mddev *mddev, void __user *argp)
7590 if (mddev->pers) {
7591 err = update_array_info(mddev, &info);
7597 if (!list_empty(&mddev->disks)) {
7598 pr_warn("md: array %s already has disks!\n", mdname(mddev));
7602 if (mddev->raid_disks) {
7603 pr_warn("md: array %s already initialised!\n", mdname(mddev));
7607 err = md_set_array_info(mddev, &info);
7619 struct mddev *mddev = NULL;
7649 mddev = bdev->bd_disk->private_data;
7651 if (!mddev) {
7659 if (!mddev->raid_disks && !mddev->external)
7662 err = get_array_info(mddev, argp);
7666 if (!mddev->raid_disks && !mddev->external)
7669 err = get_disk_info(mddev, argp);
7673 err = set_disk_faulty(mddev, new_decode_dev(arg));
7677 err = get_bitmap_file(mddev, argp);
7684 wait_event_interruptible_timeout(mddev->sb_wait,
7686 &mddev->recovery),
7692 mutex_lock(&mddev->open_mutex);
7693 if (mddev->pers && atomic_read(&mddev->openers) > 1) {
7694 mutex_unlock(&mddev->open_mutex);
7698 if (test_and_set_bit(MD_CLOSING, &mddev->flags)) {
7699 mutex_unlock(&mddev->open_mutex);
7703 mutex_unlock(&mddev->open_mutex);
7706 err = mddev_lock(mddev);
7714 err = __md_set_array_info(mddev, argp);
7723 if ((!mddev->raid_disks && !mddev->external)
7736 err = restart_array(mddev);
7740 err = do_md_stop(mddev, 0, bdev);
7744 err = md_set_readonly(mddev, bdev);
7748 err = hot_remove_disk(mddev, new_decode_dev(arg));
7754 * So require mddev->pers and MD_DISK_SYNC.
7756 if (mddev->pers) {
7764 err = md_add_new_disk(mddev, &info);
7774 if (!md_is_rdwr(mddev) && mddev->pers) {
7775 if (mddev->ro != MD_AUTO_READ) {
7779 mddev->ro = MD_RDWR;
7780 sysfs_notify_dirent_safe(mddev->sysfs_state);
7781 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
7786 if (test_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags)) {
7787 mddev_unlock(mddev);
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));
7791 mddev_lock_nointr(mddev);
7802 err = md_add_new_disk(mddev, &info);
7807 if (mddev_is_clustered(mddev))
7808 md_cluster_ops->new_disk_ack(mddev, false);
7814 err = hot_add_disk(mddev, new_decode_dev(arg));
7818 err = do_md_run(mddev);
7822 err = set_bitmap_file(mddev, (int)arg);
7831 if (mddev->hold_active == UNTIL_IOCTL &&
7833 mddev->hold_active = 0;
7834 mddev_unlock(mddev);
7837 clear_bit(MD_CLOSING, &mddev->flags);
7862 struct mddev *mddev = bdev->bd_disk->private_data;
7865 err = mddev_lock(mddev);
7869 if (!mddev->raid_disks && !mddev->external) {
7878 if (!ro && mddev->ro == MD_RDONLY && mddev->pers) {
7879 err = restart_array(mddev);
7882 mddev->ro = MD_AUTO_READ;
7886 mddev_unlock(mddev);
7892 struct mddev *mddev;
7896 mddev = mddev_get(disk->private_data);
7898 if (!mddev)
7901 err = mutex_lock_interruptible(&mddev->open_mutex);
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);
7918 mddev_put(mddev);
7924 struct mddev *mddev = disk->private_data;
7926 BUG_ON(!mddev);
7927 atomic_dec(&mddev->openers);
7928 mddev_put(mddev);
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);
7947 mddev_free(mddev);
8036 struct mddev *mddev, const char *name)
8047 thread->mddev = mddev;
8051 mdname(thread->mddev),
8061 void md_unregister_thread(struct mddev *mddev, struct md_thread __rcu **threadp)
8064 lockdep_is_held(&mddev->reconfig_mutex));
8078 void md_error(struct mddev *mddev, struct md_rdev *rdev)
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);
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);
8123 static int status_resync(struct seq_file *seq, struct mddev *mddev)
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);
8158 if (test_bit(MD_RESYNCING_REMOTE, &mddev->recovery)) {
8161 rdev_for_each(rdev, mddev)
8169 if (mddev->reshape_position != MaxSector)
8175 if (mddev->recovery_cp < MaxSector) {
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;
8265 struct mddev *mddev;
8280 mddev = list_entry(tmp, struct mddev, all_mddevs);
8281 if (!mddev_get(mddev))
8284 return mddev;
8295 struct mddev *next_mddev, *mddev = v;
8296 struct mddev *to_put = NULL;
8306 to_put = mddev;
8307 tmp = mddev->all_mddevs.next;
8316 next_mddev = list_entry(tmp, struct mddev, all_mddevs);
8319 mddev = next_mddev;
8320 tmp = mddev->all_mddevs.next;
8332 struct mddev *mddev = v;
8334 if (mddev && v != (void*)1 && v != (void*)2)
8335 mddev_put(mddev);
8340 struct mddev *mddev = v;
8361 spin_lock(&mddev->lock);
8362 if (mddev->pers || mddev->raid_disks || !list_empty(&mddev->disks)) {
8363 seq_printf(seq, "%s : %sactive", mdname(mddev),
8364 mddev->pers ? "" : "in");
8365 if (mddev->pers) {
8366 if (mddev->ro == MD_RDONLY)
8368 if (mddev->ro == MD_AUTO_READ)
8370 seq_printf(seq, " %s", mddev->pers->name);
8375 rdev_for_each_rcu(rdev, mddev) {
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);
8416 if (mddev->pers) {
8417 mddev->pers->status(seq, mddev);
8419 if (mddev->pers->sync_request) {
8420 if (status_resync(seq, mddev))
8426 md_bitmap_status(seq, mddev->bitmap);
8430 spin_unlock(&mddev->lock);
8528 int md_setup_cluster(struct mddev *mddev, int nodes)
8542 ret = md_cluster_ops->join(mddev, nodes);
8544 mddev->safemode_delay = 0;
8548 void md_cluster_stop(struct mddev *mddev)
8552 md_cluster_ops->leave(mddev);
8556 static int is_mddev_idle(struct mddev *mddev, int init)
8564 rdev_for_each_rcu(rdev, mddev) {
8599 void md_done_sync(struct mddev *mddev, int blocks, int ok)
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);
8613 /* md_write_start(mddev, bi)
8620 bool md_write_start(struct mddev *mddev, struct bio *bi)
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;
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) ||
8660 is_md_suspended(mddev));
8661 if (test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)) {
8662 percpu_ref_put(&mddev->writes_pending);
8677 void md_write_inc(struct mddev *mddev, struct bio *bi)
8681 WARN_ON_ONCE(mddev->in_sync || !md_is_rdwr(mddev));
8682 percpu_ref_get(&mddev->writes_pending);
8686 void md_write_end(struct mddev *mddev)
8688 percpu_ref_put(&mddev->writes_pending);
8690 if (mddev->safemode == 2)
8691 md_wakeup_thread(mddev->thread);
8692 else if (mddev->safemode_delay)
8696 mod_timer(&mddev->safemode_timer,
8697 roundup(jiffies, mddev->safemode_delay) +
8698 mddev->safemode_delay);
8704 void md_submit_discard_bio(struct mddev *mddev, struct md_rdev *rdev,
8715 if (mddev->gendisk)
8717 disk_devt(mddev->gendisk),
8727 struct mddev *mddev = md_io_clone->mddev;
8737 percpu_ref_put(&mddev->active_io);
8740 static void md_clone_bio(struct mddev *mddev, struct bio **bio)
8745 bio_alloc_clone(bdev, *bio, GFP_NOIO, &mddev->io_clone_set);
8749 md_io_clone->mddev = mddev;
8758 void md_account_bio(struct mddev *mddev, struct bio **bio)
8760 percpu_ref_get(&mddev->active_io);
8761 md_clone_bio(mddev, bio);
8765 /* md_allow_write(mddev)
8768 * attempting a GFP_KERNEL allocation while holding the mddev lock.
8771 void md_allow_write(struct mddev *mddev)
8773 if (!mddev->pers)
8775 if (!md_is_rdwr(mddev))
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);
8789 md_update_sb(mddev, 0);
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;
8805 struct mddev *mddev2;
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);
8828 if (mddev_is_clustered(mddev)) {
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)) {
8846 } else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
8851 } else if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
8856 mddev->last_sync_action = action ?: desc;
8863 * to 1 if we choose to yield (based arbitrarily on address of mddev structure).
8870 mddev->curr_resync = MD_RESYNC_DELAYED;
8873 if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
8879 if (mddev2 == mddev)
8881 if (!mddev->parallel_resync
8883 && match_mddev_units(mddev, mddev2)) {
8885 if (mddev < mddev2 &&
8886 mddev->curr_resync == MD_RESYNC_DELAYED) {
8888 mddev->curr_resync = MD_RESYNC_YIELDED;
8891 if (mddev > mddev2 &&
8892 mddev->curr_resync == MD_RESYNC_YIELDED)
8902 if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
8903 mddev2->curr_resync >= mddev->curr_resync) {
8907 desc, mdname(mddev),
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;
8944 if (mddev_is_clustered(mddev) &&
8945 mddev->reshape_position != MaxSector)
8946 j = mddev->reshape_position;
8949 max_sectors = mddev->dev_sectors;
8952 rdev_for_each_rcu(rdev, mddev)
8969 if (mddev->bitmap) {
8970 mddev->pers->quiesce(mddev, 1);
8971 mddev->pers->quiesce(mddev, 0);
8975 pr_info("md: %s of RAID array %s\n", desc, mdname(mddev));
8976 pr_debug("md: minimum _guaranteed_ speed: %d KB/sec/disk.\n", speed_min(mddev));
8978 speed_max(mddev), desc);
8980 is_mddev_idle(mddev, 1); /* this initializes IO event counters */
8988 mddev->resync_mark = mark[last_mark];
8989 mddev->resync_mark_cnt = mark_cnt[last_mark];
8998 atomic_set(&mddev->recovery_active, 0);
9003 desc, mdname(mddev));
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)) {
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))
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;
9115 if (currspeed > speed_min(mddev)) {
9116 if (currspeed > speed_max(mddev)) {
9120 if (!is_mddev_idle(mddev, 0)) {
9125 wait_event(mddev->recovery_wait,
9126 !atomic_read(&mddev->recovery_active));
9130 pr_info("md: %s: %s %s.\n",mdname(mddev), desc,
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) {
9153 desc, mdname(mddev));
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)) {
9170 rdev_for_each_rcu(rdev, mddev)
9172 mddev->delta_disks >= 0 &&
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) {
9195 mddev_lock_nointr(mddev);
9196 md_set_array_sectors(mddev, mddev->pers->size(mddev, 0, 0));
9197 mddev_unlock(mddev);
9198 if (!mddev_is_clustered(mddev))
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);
9222 static int remove_and_add_spares(struct mddev *mddev,
9230 if (this && test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
9234 rdev_for_each(rdev, mddev) {
9252 rdev_for_each(rdev, mddev) {
9260 if (mddev->pers->hot_remove_disk(
9261 mddev, rdev) == 0) {
9262 sysfs_unlink_rdev(mddev, rdev);
9272 if (removed && mddev->kobj.sd)
9273 sysfs_notify_dirent_safe(mddev->sysfs_degraded);
9278 rdev_for_each(rdev, mddev) {
9293 if (!md_is_rdwr(mddev) &&
9300 if (mddev->pers->hot_add_disk(mddev, rdev) == 0) {
9302 sysfs_link_rdev(mddev, rdev);
9306 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
9311 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
9317 struct mddev *mddev = container_of(ws, struct mddev, del_work);
9319 rcu_assign_pointer(mddev->sync_thread,
9320 md_register_thread(md_do_sync, mddev, "resync"));
9321 if (!mddev->sync_thread) {
9323 mdname(mddev));
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);
9363 void md_check_recovery(struct mddev *mddev)
9365 if (test_bit(MD_ALLOW_SB_UPDATE, &mddev->flags) && mddev->sb_flags) {
9369 set_bit(MD_UPDATING_SB, &mddev->flags);
9371 if (test_bit(MD_ALLOW_SB_UPDATE, &mddev->flags))
9372 md_update_sb(mddev, 0);
9373 clear_bit_unlock(MD_UPDATING_SB, &mddev->flags);
9374 wake_up(&mddev->sb_wait);
9377 if (is_md_suspended(mddev))
9380 if (mddev->bitmap)
9381 md_bitmap_daemon_work(mddev);
9384 if (mddev->pers->sync_request && !mddev->external) {
9386 mdname(mddev));
9387 mddev->safemode = 2;
9392 if (!md_is_rdwr(mddev) &&
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)
9405 if (mddev_trylock(mddev)) {
9407 bool try_set_sync = mddev->safemode != 0;
9409 if (!mddev->external && mddev->safemode == 1)
9410 mddev->safemode = 0;
9412 if (!md_is_rdwr(mddev)) {
9414 if (!mddev->external && mddev->in_sync)
9420 rdev_for_each(rdev, mddev)
9429 remove_and_add_spares(mddev, NULL);
9433 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
9434 md_reap_sync_thread(mddev);
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);
9441 if (mddev_is_clustered(mddev)) {
9446 rdev_for_each_safe(rdev, tmp, mddev) {
9453 if (try_set_sync && !mddev->external && !mddev->in_sync) {
9454 spin_lock(&mddev->lock);
9455 set_in_sync(mddev);
9456 spin_unlock(&mddev->lock);
9459 if (mddev->sb_flags)
9460 md_update_sb(mddev, 0);
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))
9476 md_reap_sync_thread(mddev);
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))
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);
9510 } else if ((spares = remove_and_add_spares(mddev, NULL))) {
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);
9545 mddev_unlock(mddev);
9550 void md_reap_sync_thread(struct mddev *mddev)
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);
9573 if (mddev_is_clustered(mddev))
9580 if (!mddev->degraded)
9581 rdev_for_each(rdev, mddev)
9584 md_update_sb(mddev, 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);
9601 if (mddev_is_clustered(mddev) && is_reshaped
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);
9615 void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev)
9622 rdev_dec_pending(rdev, mddev);
9626 void md_finish_reshape(struct mddev *mddev)
9631 rdev_for_each(rdev, mddev) {
9647 struct mddev *mddev = rdev->mddev;
9659 set_mask_bits(&mddev->sb_flags, 0,
9661 md_wakeup_thread(rdev->mddev->thread);
9686 struct mddev *mddev, *n;
9690 list_for_each_entry_safe(mddev, n, &all_mddevs, all_mddevs) {
9691 if (!mddev_get(mddev))
9694 if (mddev_trylock(mddev)) {
9695 if (mddev->pers)
9696 __md_stop_writes(mddev);
9697 if (mddev->persistent)
9698 mddev->safemode = 2;
9699 mddev_unlock(mddev);
9702 mddev_put(mddev);
9776 static void check_sb_changes(struct mddev *mddev, struct md_rdev *rdev)
9786 if (mddev->dev_sectors != le64_to_cpu(sb->size)) {
9787 ret = mddev->pers->resize(mddev, le64_to_cpu(sb->size));
9791 md_bitmap_update_sb(mddev->bitmap);
9795 rdev_for_each_safe(rdev2, tmp, mddev) {
9821 ret = remove_and_add_spares(mddev, rdev2);
9824 /* wakeup mddev->thread here, so array could
9826 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
9827 md_wakeup_thread(mddev->thread);
9836 md_error(mddev, rdev2);
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) &&
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 &&
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);
9876 static int read_rdev(struct mddev *mddev, struct md_rdev *rdev)
9890 err = super_types[mddev->major_version].
9891 load_super(rdev, NULL, mddev->minor_version);
9912 * device In_sync and mddev->degraded
9916 mddev->pers->spare_active(mddev))
9917 sysfs_notify_dirent_safe(mddev->sysfs_degraded);
9923 void md_reload_sb(struct mddev *mddev, int nr)
9929 rdev_for_each_rcu(iter, mddev) {
9941 err = read_rdev(mddev, rdev);
9945 check_sb_changes(mddev, rdev);
9948 rdev_for_each_rcu(rdev, mddev) {
9950 read_rdev(mddev, rdev);
10026 struct mddev *mddev, *n;
10047 list_for_each_entry_safe(mddev, n, &all_mddevs, all_mddevs) {
10048 if (!mddev_get(mddev))
10051 export_array(mddev);
10052 mddev->ctime = 0;
10053 mddev->hold_active = 0;
10055 * As the mddev is now fully clear, mddev_put will schedule
10056 * the mddev for destruction by a workqueue, and the
10059 mddev_put(mddev);