Lines Matching +full:update +full:- +full:fc +full:- +full:fixup
1 // SPDX-License-Identifier: GPL-2.0-or-later
5 * Copyright (C) 2000-2004 Neil Brown
7 * RAID-10 support for md.
24 #include "md-bitmap.h"
78 do { if ((md)->queue) blk_add_trace_msg((md)->queue, "raid10 " fmt, ##args); } while (0)
80 #include "raid1-10.c"
83 * for resync bio, r10bio pointer can be retrieved from the per-bio
88 return get_resync_pages(bio)->raid_bio; in get_resync_r10bio()
94 int size = offsetof(struct r10bio, devs[conf->copies]); in r10bio_pool_alloc()
129 if (test_bit(MD_RECOVERY_SYNC, &conf->mddev->recovery) || in r10buf_pool_alloc()
130 test_bit(MD_RECOVERY_RESHAPE, &conf->mddev->recovery)) in r10buf_pool_alloc()
131 nalloc = conf->copies; /* resync */ in r10buf_pool_alloc()
136 if (!conf->have_replacement) in r10buf_pool_alloc()
147 for (j = nalloc ; j-- ; ) { in r10buf_pool_alloc()
151 r10_bio->devs[j].bio = bio; in r10buf_pool_alloc()
152 if (!conf->have_replacement) in r10buf_pool_alloc()
157 r10_bio->devs[j].repl_bio = bio; in r10buf_pool_alloc()
164 struct bio *rbio = r10_bio->devs[j].repl_bio; in r10buf_pool_alloc()
171 bio = r10_bio->devs[j].bio; in r10buf_pool_alloc()
174 &conf->mddev->recovery)) { in r10buf_pool_alloc()
182 rp->raid_bio = r10_bio; in r10buf_pool_alloc()
183 bio->bi_private = rp; in r10buf_pool_alloc()
186 rbio->bi_private = rp_repl; in r10buf_pool_alloc()
193 while (--j >= 0) in r10buf_pool_alloc()
199 if (r10_bio->devs[j].bio) in r10buf_pool_alloc()
200 bio_put(r10_bio->devs[j].bio); in r10buf_pool_alloc()
201 if (r10_bio->devs[j].repl_bio) in r10buf_pool_alloc()
202 bio_put(r10_bio->devs[j].repl_bio); in r10buf_pool_alloc()
217 for (j = conf->copies; j--; ) { in r10buf_pool_free()
218 struct bio *bio = r10bio->devs[j].bio; in r10buf_pool_free()
226 bio = r10bio->devs[j].repl_bio; in r10buf_pool_free()
241 for (i = 0; i < conf->copies; i++) { in put_all_bios()
242 struct bio **bio = & r10_bio->devs[i].bio; in put_all_bios()
246 bio = &r10_bio->devs[i].repl_bio; in put_all_bios()
247 if (r10_bio->read_slot < 0 && !BIO_SPECIAL(*bio)) in put_all_bios()
255 struct r10conf *conf = r10_bio->mddev->private; in free_r10bio()
258 mempool_free(r10_bio, &conf->r10bio_pool); in free_r10bio()
263 struct r10conf *conf = r10_bio->mddev->private; in put_buf()
265 mempool_free(r10_bio, &conf->r10buf_pool); in put_buf()
273 struct mddev *mddev = r10_bio->mddev; in reschedule_retry()
274 struct r10conf *conf = mddev->private; in reschedule_retry()
276 spin_lock_irqsave(&conf->device_lock, flags); in reschedule_retry()
277 list_add(&r10_bio->retry_list, &conf->retry_list); in reschedule_retry()
278 conf->nr_queued ++; in reschedule_retry()
279 spin_unlock_irqrestore(&conf->device_lock, flags); in reschedule_retry()
282 wake_up(&conf->wait_barrier); in reschedule_retry()
284 md_wakeup_thread(mddev->thread); in reschedule_retry()
294 struct bio *bio = r10_bio->master_bio; in raid_end_bio_io()
295 struct r10conf *conf = r10_bio->mddev->private; in raid_end_bio_io()
297 if (!test_bit(R10BIO_Uptodate, &r10_bio->state)) in raid_end_bio_io()
298 bio->bi_status = BLK_STS_IOERR; in raid_end_bio_io()
311 * Update disk head position estimator based on IRQ completion info.
315 struct r10conf *conf = r10_bio->mddev->private; in update_head_pos()
317 conf->mirrors[r10_bio->devs[slot].devnum].head_position = in update_head_pos()
318 r10_bio->devs[slot].addr + (r10_bio->sectors); in update_head_pos()
330 for (slot = 0; slot < conf->copies; slot++) { in find_bio_disk()
331 if (r10_bio->devs[slot].bio == bio) in find_bio_disk()
333 if (r10_bio->devs[slot].repl_bio == bio) { in find_bio_disk()
339 BUG_ON(slot == conf->copies); in find_bio_disk()
346 return r10_bio->devs[slot].devnum; in find_bio_disk()
351 int uptodate = !bio->bi_status; in raid10_end_read_request()
352 struct r10bio *r10_bio = bio->bi_private; in raid10_end_read_request()
355 struct r10conf *conf = r10_bio->mddev->private; in raid10_end_read_request()
357 slot = r10_bio->read_slot; in raid10_end_read_request()
358 rdev = r10_bio->devs[slot].rdev; in raid10_end_read_request()
371 * user-side. So if something waits for IO, then it will in raid10_end_read_request()
374 set_bit(R10BIO_Uptodate, &r10_bio->state); in raid10_end_read_request()
381 if (!_enough(conf, test_bit(R10BIO_Previous, &r10_bio->state), in raid10_end_read_request()
382 rdev->raid_disk)) in raid10_end_read_request()
387 rdev_dec_pending(rdev, conf->mddev); in raid10_end_read_request()
390 * oops, read error - keep the refcount on the rdev in raid10_end_read_request()
394 mdname(conf->mddev), in raid10_end_read_request()
395 bdevname(rdev->bdev, b), in raid10_end_read_request()
396 (unsigned long long)r10_bio->sector); in raid10_end_read_request()
397 set_bit(R10BIO_ReadError, &r10_bio->state); in raid10_end_read_request()
405 md_bitmap_endwrite(r10_bio->mddev->bitmap, r10_bio->sector, in close_write()
406 r10_bio->sectors, in close_write()
407 !test_bit(R10BIO_Degraded, &r10_bio->state), in close_write()
409 md_write_end(r10_bio->mddev); in close_write()
414 if (atomic_dec_and_test(&r10_bio->remaining)) { in one_write_done()
415 if (test_bit(R10BIO_WriteError, &r10_bio->state)) in one_write_done()
419 if (test_bit(R10BIO_MadeGood, &r10_bio->state)) in one_write_done()
429 struct r10bio *r10_bio = bio->bi_private; in raid10_end_write_request()
432 struct r10conf *conf = r10_bio->mddev->private; in raid10_end_write_request()
438 discard_error = bio->bi_status && bio_op(bio) == REQ_OP_DISCARD; in raid10_end_write_request()
443 rdev = conf->mirrors[dev].replacement; in raid10_end_write_request()
447 rdev = conf->mirrors[dev].rdev; in raid10_end_write_request()
452 if (bio->bi_status && !discard_error) { in raid10_end_write_request()
457 md_error(rdev->mddev, rdev); in raid10_end_write_request()
459 set_bit(WriteErrorSeen, &rdev->flags); in raid10_end_write_request()
460 if (!test_and_set_bit(WantReplacement, &rdev->flags)) in raid10_end_write_request()
462 &rdev->mddev->recovery); in raid10_end_write_request()
465 if (test_bit(FailFast, &rdev->flags) && in raid10_end_write_request()
466 (bio->bi_opf & MD_FAILFAST)) { in raid10_end_write_request()
467 md_error(rdev->mddev, rdev); in raid10_end_write_request()
474 if (!test_bit(Faulty, &rdev->flags)) in raid10_end_write_request()
475 set_bit(R10BIO_WriteError, &r10_bio->state); in raid10_end_write_request()
478 set_bit(R10BIO_Degraded, &r10_bio->state); in raid10_end_write_request()
479 r10_bio->devs[slot].bio = NULL; in raid10_end_write_request()
491 * user-side. So if something waits for IO, then it will in raid10_end_write_request()
502 * before rdev->recovery_offset, but for simplicity we don't in raid10_end_write_request()
505 if (test_bit(In_sync, &rdev->flags) && in raid10_end_write_request()
506 !test_bit(Faulty, &rdev->flags)) in raid10_end_write_request()
507 set_bit(R10BIO_Uptodate, &r10_bio->state); in raid10_end_write_request()
511 r10_bio->devs[slot].addr, in raid10_end_write_request()
512 r10_bio->sectors, in raid10_end_write_request()
516 r10_bio->devs[slot].repl_bio = IO_MADE_GOOD; in raid10_end_write_request()
518 r10_bio->devs[slot].bio = IO_MADE_GOOD; in raid10_end_write_request()
520 set_bit(R10BIO_MadeGood, &r10_bio->state); in raid10_end_write_request()
531 rdev_dec_pending(rdev, conf->mddev); in raid10_end_write_request()
571 last_far_set_start = (geo->raid_disks / geo->far_set_size) - 1; in __raid10_find_phys()
572 last_far_set_start *= geo->far_set_size; in __raid10_find_phys()
574 last_far_set_size = geo->far_set_size; in __raid10_find_phys()
575 last_far_set_size += (geo->raid_disks % geo->far_set_size); in __raid10_find_phys()
578 chunk = r10bio->sector >> geo->chunk_shift; in __raid10_find_phys()
579 sector = r10bio->sector & geo->chunk_mask; in __raid10_find_phys()
581 chunk *= geo->near_copies; in __raid10_find_phys()
583 dev = sector_div(stripe, geo->raid_disks); in __raid10_find_phys()
584 if (geo->far_offset) in __raid10_find_phys()
585 stripe *= geo->far_copies; in __raid10_find_phys()
587 sector += stripe << geo->chunk_shift; in __raid10_find_phys()
590 for (n = 0; n < geo->near_copies; n++) { in __raid10_find_phys()
594 r10bio->devs[slot].devnum = d; in __raid10_find_phys()
595 r10bio->devs[slot].addr = s; in __raid10_find_phys()
598 for (f = 1; f < geo->far_copies; f++) { in __raid10_find_phys()
599 set = d / geo->far_set_size; in __raid10_find_phys()
600 d += geo->near_copies; in __raid10_find_phys()
602 if ((geo->raid_disks % geo->far_set_size) && in __raid10_find_phys()
604 d -= last_far_set_start; in __raid10_find_phys()
608 d %= geo->far_set_size; in __raid10_find_phys()
609 d += geo->far_set_size * set; in __raid10_find_phys()
611 s += geo->stride; in __raid10_find_phys()
612 r10bio->devs[slot].devnum = d; in __raid10_find_phys()
613 r10bio->devs[slot].addr = s; in __raid10_find_phys()
617 if (dev >= geo->raid_disks) { in __raid10_find_phys()
619 sector += (geo->chunk_mask + 1); in __raid10_find_phys()
626 struct geom *geo = &conf->geo; in raid10_find_phys()
628 if (conf->reshape_progress != MaxSector && in raid10_find_phys()
629 ((r10bio->sector >= conf->reshape_progress) != in raid10_find_phys()
630 conf->mddev->reshape_backwards)) { in raid10_find_phys()
631 set_bit(R10BIO_Previous, &r10bio->state); in raid10_find_phys()
632 geo = &conf->prev; in raid10_find_phys()
634 clear_bit(R10BIO_Previous, &r10bio->state); in raid10_find_phys()
642 /* Never use conf->prev as this is only called during resync in raid10_find_virt()
645 struct geom *geo = &conf->geo; in raid10_find_virt()
646 int far_set_start = (dev / geo->far_set_size) * geo->far_set_size; in raid10_find_virt()
647 int far_set_size = geo->far_set_size; in raid10_find_virt()
650 if (geo->raid_disks % geo->far_set_size) { in raid10_find_virt()
651 last_far_set_start = (geo->raid_disks / geo->far_set_size) - 1; in raid10_find_virt()
652 last_far_set_start *= geo->far_set_size; in raid10_find_virt()
655 far_set_size = geo->far_set_size; in raid10_find_virt()
656 far_set_size += (geo->raid_disks % geo->far_set_size); in raid10_find_virt()
661 offset = sector & geo->chunk_mask; in raid10_find_virt()
662 if (geo->far_offset) { in raid10_find_virt()
663 int fc; in raid10_find_virt() local
664 chunk = sector >> geo->chunk_shift; in raid10_find_virt()
665 fc = sector_div(chunk, geo->far_copies); in raid10_find_virt()
666 dev -= fc * geo->near_copies; in raid10_find_virt()
670 while (sector >= geo->stride) { in raid10_find_virt()
671 sector -= geo->stride; in raid10_find_virt()
672 if (dev < (geo->near_copies + far_set_start)) in raid10_find_virt()
673 dev += far_set_size - geo->near_copies; in raid10_find_virt()
675 dev -= geo->near_copies; in raid10_find_virt()
677 chunk = sector >> geo->chunk_shift; in raid10_find_virt()
679 vchunk = chunk * geo->raid_disks + dev; in raid10_find_virt()
680 sector_div(vchunk, geo->near_copies); in raid10_find_virt()
681 return (vchunk << geo->chunk_shift) + offset; in raid10_find_virt()
686 * be done. There is a per-array 'next expected sequential IO' sector
687 * number - if this matches on the next IO then we use the last disk.
688 * There is also a per-disk 'last know head position' sector that is
690 * completion handlers update this position correctly. If there is no
707 const sector_t this_sector = r10_bio->sector; in read_balance()
709 int sectors = r10_bio->sectors; in read_balance()
717 struct geom *geo = &conf->geo; in read_balance()
721 best_dist_slot = -1; in read_balance()
728 clear_bit(R10BIO_FailFast, &r10_bio->state); in read_balance()
735 if ((conf->mddev->recovery_cp < MaxSector in read_balance()
736 && (this_sector + sectors >= conf->next_resync)) || in read_balance()
737 (mddev_is_clustered(conf->mddev) && in read_balance()
738 md_cluster_ops->area_resyncing(conf->mddev, READ, this_sector, in read_balance()
742 for (slot = 0; slot < conf->copies ; slot++) { in read_balance()
749 if (r10_bio->devs[slot].bio == IO_BLOCKED) in read_balance()
751 disk = r10_bio->devs[slot].devnum; in read_balance()
752 rdev = rcu_dereference(conf->mirrors[disk].replacement); in read_balance()
753 if (rdev == NULL || test_bit(Faulty, &rdev->flags) || in read_balance()
754 r10_bio->devs[slot].addr + sectors > in read_balance()
755 rdev->recovery_offset) { in read_balance()
762 rdev = rcu_dereference(conf->mirrors[disk].rdev); in read_balance()
765 test_bit(Faulty, &rdev->flags)) in read_balance()
767 if (!test_bit(In_sync, &rdev->flags) && in read_balance()
768 r10_bio->devs[slot].addr + sectors > rdev->recovery_offset) in read_balance()
771 dev_sector = r10_bio->devs[slot].addr; in read_balance()
782 bad_sectors -= (dev_sector - first_bad); in read_balance()
789 first_bad - dev_sector; in read_balance()
806 nonrot = blk_queue_nonrot(bdev_get_queue(rdev->bdev)); in read_balance()
808 pending = atomic_read(&rdev->nr_pending); in read_balance()
817 set_bit(R10BIO_FailFast, &r10_bio->state); in read_balance()
822 if (geo->near_copies > 1 && !pending) in read_balance()
826 else if (geo->far_copies > 1) in read_balance()
827 new_distance = r10_bio->devs[slot].addr; in read_balance()
829 new_distance = abs(r10_bio->devs[slot].addr - in read_balance()
830 conf->mirrors[disk].head_position); in read_balance()
838 if (slot >= conf->copies) { in read_balance()
849 atomic_inc(&rdev->nr_pending); in read_balance()
850 r10_bio->read_slot = slot; in read_balance()
864 spin_lock_irq(&conf->device_lock); in flush_pending_writes()
866 if (conf->pending_bio_list.head) { in flush_pending_writes()
870 bio = bio_list_get(&conf->pending_bio_list); in flush_pending_writes()
871 conf->pending_count = 0; in flush_pending_writes()
872 spin_unlock_irq(&conf->device_lock); in flush_pending_writes()
876 * current->state might be TASK_UNINTERRUPTIBLE which will in flush_pending_writes()
880 * is a false-positive. Silence the warning by resetting in flush_pending_writes()
888 md_bitmap_unplug(conf->mddev->bitmap); in flush_pending_writes()
889 wake_up(&conf->wait_barrier); in flush_pending_writes()
892 struct bio *next = bio->bi_next; in flush_pending_writes()
893 struct md_rdev *rdev = (void*)bio->bi_disk; in flush_pending_writes()
894 bio->bi_next = NULL; in flush_pending_writes()
895 bio_set_dev(bio, rdev->bdev); in flush_pending_writes()
896 if (test_bit(Faulty, &rdev->flags)) { in flush_pending_writes()
899 !blk_queue_discard(bio->bi_disk->queue))) in flush_pending_writes()
909 spin_unlock_irq(&conf->device_lock); in flush_pending_writes()
921 * We choose only to raise the barrier if no-one is waiting for the
936 BUG_ON(force && !conf->barrier); in raise_barrier()
937 spin_lock_irq(&conf->resync_lock); in raise_barrier()
940 wait_event_lock_irq(conf->wait_barrier, force || !conf->nr_waiting, in raise_barrier()
941 conf->resync_lock); in raise_barrier()
944 conf->barrier++; in raise_barrier()
947 wait_event_lock_irq(conf->wait_barrier, in raise_barrier()
948 !atomic_read(&conf->nr_pending) && conf->barrier < RESYNC_DEPTH, in raise_barrier()
949 conf->resync_lock); in raise_barrier()
951 spin_unlock_irq(&conf->resync_lock); in raise_barrier()
957 spin_lock_irqsave(&conf->resync_lock, flags); in lower_barrier()
958 conf->barrier--; in lower_barrier()
959 spin_unlock_irqrestore(&conf->resync_lock, flags); in lower_barrier()
960 wake_up(&conf->wait_barrier); in lower_barrier()
965 spin_lock_irq(&conf->resync_lock); in wait_barrier()
966 if (conf->barrier) { in wait_barrier()
967 struct bio_list *bio_list = current->bio_list; in wait_barrier()
968 conf->nr_waiting++; in wait_barrier()
973 * pre-process bio queue isn't empty, in wait_barrier()
978 raid10_log(conf->mddev, "wait barrier"); in wait_barrier()
979 wait_event_lock_irq(conf->wait_barrier, in wait_barrier()
980 !conf->barrier || in wait_barrier()
981 (atomic_read(&conf->nr_pending) && in wait_barrier()
988 (conf->mddev->thread->tsk == current && in wait_barrier()
990 &conf->mddev->recovery) && in wait_barrier()
991 conf->nr_queued > 0), in wait_barrier()
992 conf->resync_lock); in wait_barrier()
993 conf->nr_waiting--; in wait_barrier()
994 if (!conf->nr_waiting) in wait_barrier()
995 wake_up(&conf->wait_barrier); in wait_barrier()
997 atomic_inc(&conf->nr_pending); in wait_barrier()
998 spin_unlock_irq(&conf->resync_lock); in wait_barrier()
1003 if ((atomic_dec_and_test(&conf->nr_pending)) || in allow_barrier()
1004 (conf->array_freeze_pending)) in allow_barrier()
1005 wake_up(&conf->wait_barrier); in allow_barrier()
1017 * pending IO requests to complete or be queued for re-try. in freeze_array()
1022 spin_lock_irq(&conf->resync_lock); in freeze_array()
1023 conf->array_freeze_pending++; in freeze_array()
1024 conf->barrier++; in freeze_array()
1025 conf->nr_waiting++; in freeze_array()
1026 wait_event_lock_irq_cmd(conf->wait_barrier, in freeze_array()
1027 atomic_read(&conf->nr_pending) == conf->nr_queued+extra, in freeze_array()
1028 conf->resync_lock, in freeze_array()
1031 conf->array_freeze_pending--; in freeze_array()
1032 spin_unlock_irq(&conf->resync_lock); in freeze_array()
1038 spin_lock_irq(&conf->resync_lock); in unfreeze_array()
1039 conf->barrier--; in unfreeze_array()
1040 conf->nr_waiting--; in unfreeze_array()
1041 wake_up(&conf->wait_barrier); in unfreeze_array()
1042 spin_unlock_irq(&conf->resync_lock); in unfreeze_array()
1048 if (!test_bit(MD_RECOVERY_RESHAPE, &rdev->mddev->recovery) || in choose_data_offset()
1049 test_bit(R10BIO_Previous, &r10_bio->state)) in choose_data_offset()
1050 return rdev->data_offset; in choose_data_offset()
1052 return rdev->new_data_offset; in choose_data_offset()
1065 struct mddev *mddev = plug->cb.data; in raid10_unplug()
1066 struct r10conf *conf = mddev->private; in raid10_unplug()
1069 if (from_schedule || current->bio_list) { in raid10_unplug()
1070 spin_lock_irq(&conf->device_lock); in raid10_unplug()
1071 bio_list_merge(&conf->pending_bio_list, &plug->pending); in raid10_unplug()
1072 conf->pending_count += plug->pending_cnt; in raid10_unplug()
1073 spin_unlock_irq(&conf->device_lock); in raid10_unplug()
1074 wake_up(&conf->wait_barrier); in raid10_unplug()
1075 md_wakeup_thread(mddev->thread); in raid10_unplug()
1080 /* we aren't scheduling, so we can do the write-out directly. */ in raid10_unplug()
1081 bio = bio_list_get(&plug->pending); in raid10_unplug()
1082 md_bitmap_unplug(mddev->bitmap); in raid10_unplug()
1083 wake_up(&conf->wait_barrier); in raid10_unplug()
1086 struct bio *next = bio->bi_next; in raid10_unplug()
1087 struct md_rdev *rdev = (void*)bio->bi_disk; in raid10_unplug()
1088 bio->bi_next = NULL; in raid10_unplug()
1089 bio_set_dev(bio, rdev->bdev); in raid10_unplug()
1090 if (test_bit(Faulty, &rdev->flags)) { in raid10_unplug()
1093 !blk_queue_discard(bio->bi_disk->queue))) in raid10_unplug()
1114 while (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) && in regular_request_wait()
1115 bio->bi_iter.bi_sector < conf->reshape_progress && in regular_request_wait()
1116 bio->bi_iter.bi_sector + sectors > conf->reshape_progress) { in regular_request_wait()
1117 raid10_log(conf->mddev, "wait reshape"); in regular_request_wait()
1119 wait_event(conf->wait_barrier, in regular_request_wait()
1120 conf->reshape_progress <= bio->bi_iter.bi_sector || in regular_request_wait()
1121 conf->reshape_progress >= bio->bi_iter.bi_sector + in regular_request_wait()
1130 struct r10conf *conf = mddev->private; in raid10_read_request()
1133 const unsigned long do_sync = (bio->bi_opf & REQ_SYNC); in raid10_read_request()
1137 int slot = r10_bio->read_slot; in raid10_read_request()
1141 if (slot >= 0 && r10_bio->devs[slot].rdev) { in raid10_read_request()
1157 disk = r10_bio->devs[slot].devnum; in raid10_read_request()
1158 err_rdev = rcu_dereference(conf->mirrors[disk].rdev); in raid10_read_request()
1160 bdevname(err_rdev->bdev, b); in raid10_read_request()
1164 err_rdev = r10_bio->devs[slot].rdev; in raid10_read_request()
1169 regular_request_wait(mddev, conf, bio, r10_bio->sectors); in raid10_read_request()
1175 (unsigned long long)r10_bio->sector); in raid10_read_request()
1183 bdevname(rdev->bdev, b), in raid10_read_request()
1184 (unsigned long long)r10_bio->sector); in raid10_read_request()
1187 gfp, &conf->bio_split); in raid10_read_request()
1193 r10_bio->master_bio = bio; in raid10_read_request()
1194 r10_bio->sectors = max_sectors; in raid10_read_request()
1196 slot = r10_bio->read_slot; in raid10_read_request()
1198 read_bio = bio_clone_fast(bio, gfp, &mddev->bio_set); in raid10_read_request()
1200 r10_bio->devs[slot].bio = read_bio; in raid10_read_request()
1201 r10_bio->devs[slot].rdev = rdev; in raid10_read_request()
1203 read_bio->bi_iter.bi_sector = r10_bio->devs[slot].addr + in raid10_read_request()
1205 bio_set_dev(read_bio, rdev->bdev); in raid10_read_request()
1206 read_bio->bi_end_io = raid10_end_read_request; in raid10_read_request()
1208 if (test_bit(FailFast, &rdev->flags) && in raid10_read_request()
1209 test_bit(R10BIO_FailFast, &r10_bio->state)) in raid10_read_request()
1210 read_bio->bi_opf |= MD_FAILFAST; in raid10_read_request()
1211 read_bio->bi_private = r10_bio; in raid10_read_request()
1213 if (mddev->gendisk) in raid10_read_request()
1214 trace_block_bio_remap(read_bio->bi_disk->queue, in raid10_read_request()
1215 read_bio, disk_devt(mddev->gendisk), in raid10_read_request()
1216 r10_bio->sector); in raid10_read_request()
1226 const unsigned long do_sync = (bio->bi_opf & REQ_SYNC); in raid10_write_one_disk()
1227 const unsigned long do_fua = (bio->bi_opf & REQ_FUA); in raid10_write_one_disk()
1231 struct r10conf *conf = mddev->private; in raid10_write_one_disk()
1233 int devnum = r10_bio->devs[n_copy].devnum; in raid10_write_one_disk()
1237 rdev = conf->mirrors[devnum].replacement; in raid10_write_one_disk()
1241 rdev = conf->mirrors[devnum].rdev; in raid10_write_one_disk()
1244 rdev = conf->mirrors[devnum].rdev; in raid10_write_one_disk()
1246 mbio = bio_clone_fast(bio, GFP_NOIO, &mddev->bio_set); in raid10_write_one_disk()
1248 r10_bio->devs[n_copy].repl_bio = mbio; in raid10_write_one_disk()
1250 r10_bio->devs[n_copy].bio = mbio; in raid10_write_one_disk()
1252 mbio->bi_iter.bi_sector = (r10_bio->devs[n_copy].addr + in raid10_write_one_disk()
1254 bio_set_dev(mbio, rdev->bdev); in raid10_write_one_disk()
1255 mbio->bi_end_io = raid10_end_write_request; in raid10_write_one_disk()
1258 &conf->mirrors[devnum].rdev->flags) in raid10_write_one_disk()
1260 mbio->bi_opf |= MD_FAILFAST; in raid10_write_one_disk()
1261 mbio->bi_private = r10_bio; in raid10_write_one_disk()
1263 if (conf->mddev->gendisk) in raid10_write_one_disk()
1264 trace_block_bio_remap(mbio->bi_disk->queue, in raid10_write_one_disk()
1265 mbio, disk_devt(conf->mddev->gendisk), in raid10_write_one_disk()
1266 r10_bio->sector); in raid10_write_one_disk()
1268 mbio->bi_disk = (void *)rdev; in raid10_write_one_disk()
1270 atomic_inc(&r10_bio->remaining); in raid10_write_one_disk()
1278 bio_list_add(&plug->pending, mbio); in raid10_write_one_disk()
1279 plug->pending_cnt++; in raid10_write_one_disk()
1281 spin_lock_irqsave(&conf->device_lock, flags); in raid10_write_one_disk()
1282 bio_list_add(&conf->pending_bio_list, mbio); in raid10_write_one_disk()
1283 conf->pending_count++; in raid10_write_one_disk()
1284 spin_unlock_irqrestore(&conf->device_lock, flags); in raid10_write_one_disk()
1285 md_wakeup_thread(mddev->thread); in raid10_write_one_disk()
1292 struct r10conf *conf = mddev->private; in raid10_write_request()
1299 md_cluster_ops->area_resyncing(mddev, WRITE, in raid10_write_request()
1300 bio->bi_iter.bi_sector, in raid10_write_request()
1304 prepare_to_wait(&conf->wait_barrier, in raid10_write_request()
1306 if (!md_cluster_ops->area_resyncing(mddev, WRITE, in raid10_write_request()
1307 bio->bi_iter.bi_sector, bio_end_sector(bio))) in raid10_write_request()
1311 finish_wait(&conf->wait_barrier, &w); in raid10_write_request()
1314 sectors = r10_bio->sectors; in raid10_write_request()
1316 if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) && in raid10_write_request()
1317 (mddev->reshape_backwards in raid10_write_request()
1318 ? (bio->bi_iter.bi_sector < conf->reshape_safe && in raid10_write_request()
1319 bio->bi_iter.bi_sector + sectors > conf->reshape_progress) in raid10_write_request()
1320 : (bio->bi_iter.bi_sector + sectors > conf->reshape_safe && in raid10_write_request()
1321 bio->bi_iter.bi_sector < conf->reshape_progress))) { in raid10_write_request()
1322 /* Need to update reshape_position in metadata */ in raid10_write_request()
1323 mddev->reshape_position = conf->reshape_progress; in raid10_write_request()
1324 set_mask_bits(&mddev->sb_flags, 0, in raid10_write_request()
1326 md_wakeup_thread(mddev->thread); in raid10_write_request()
1327 raid10_log(conf->mddev, "wait reshape metadata"); in raid10_write_request()
1328 wait_event(mddev->sb_wait, in raid10_write_request()
1329 !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)); in raid10_write_request()
1331 conf->reshape_safe = mddev->reshape_position; in raid10_write_request()
1334 if (conf->pending_count >= max_queued_requests) { in raid10_write_request()
1335 md_wakeup_thread(mddev->thread); in raid10_write_request()
1337 wait_event(conf->wait_barrier, in raid10_write_request()
1338 conf->pending_count < max_queued_requests); in raid10_write_request()
1350 r10_bio->read_slot = -1; /* make sure repl_bio gets freed */ in raid10_write_request()
1355 max_sectors = r10_bio->sectors; in raid10_write_request()
1357 for (i = 0; i < conf->copies; i++) { in raid10_write_request()
1358 int d = r10_bio->devs[i].devnum; in raid10_write_request()
1361 rrdev = rcu_dereference(conf->mirrors[d].replacement); in raid10_write_request()
1367 rdev = rcu_dereference(conf->mirrors[d].rdev); in raid10_write_request()
1370 if (rdev && unlikely(test_bit(Blocked, &rdev->flags))) { in raid10_write_request()
1371 atomic_inc(&rdev->nr_pending); in raid10_write_request()
1375 if (rrdev && unlikely(test_bit(Blocked, &rrdev->flags))) { in raid10_write_request()
1376 atomic_inc(&rrdev->nr_pending); in raid10_write_request()
1380 if (rdev && (test_bit(Faulty, &rdev->flags))) in raid10_write_request()
1382 if (rrdev && (test_bit(Faulty, &rrdev->flags))) in raid10_write_request()
1385 r10_bio->devs[i].bio = NULL; in raid10_write_request()
1386 r10_bio->devs[i].repl_bio = NULL; in raid10_write_request()
1389 set_bit(R10BIO_Degraded, &r10_bio->state); in raid10_write_request()
1392 if (rdev && test_bit(WriteErrorSeen, &rdev->flags)) { in raid10_write_request()
1394 sector_t dev_sector = r10_bio->devs[i].addr; in raid10_write_request()
1404 atomic_inc(&rdev->nr_pending); in raid10_write_request()
1405 set_bit(BlockedBadBlocks, &rdev->flags); in raid10_write_request()
1411 bad_sectors -= (dev_sector - first_bad); in raid10_write_request()
1419 * so it might be re-added, and we want to in raid10_write_request()
1422 * fact that this chunk is not in-sync is in raid10_write_request()
1428 int good_sectors = first_bad - dev_sector; in raid10_write_request()
1434 r10_bio->devs[i].bio = bio; in raid10_write_request()
1435 atomic_inc(&rdev->nr_pending); in raid10_write_request()
1438 r10_bio->devs[i].repl_bio = bio; in raid10_write_request()
1439 atomic_inc(&rrdev->nr_pending); in raid10_write_request()
1450 if (r10_bio->devs[j].bio) { in raid10_write_request()
1451 d = r10_bio->devs[j].devnum; in raid10_write_request()
1452 rdev_dec_pending(conf->mirrors[d].rdev, mddev); in raid10_write_request()
1454 if (r10_bio->devs[j].repl_bio) { in raid10_write_request()
1456 d = r10_bio->devs[j].devnum; in raid10_write_request()
1457 rdev = conf->mirrors[d].replacement; in raid10_write_request()
1461 rdev = conf->mirrors[d].rdev; in raid10_write_request()
1467 raid10_log(conf->mddev, "wait rdev %d blocked", blocked_rdev->raid_disk); in raid10_write_request()
1473 if (max_sectors < r10_bio->sectors) in raid10_write_request()
1474 r10_bio->sectors = max_sectors; in raid10_write_request()
1476 if (r10_bio->sectors < bio_sectors(bio)) { in raid10_write_request()
1477 struct bio *split = bio_split(bio, r10_bio->sectors, in raid10_write_request()
1478 GFP_NOIO, &conf->bio_split); in raid10_write_request()
1484 r10_bio->master_bio = bio; in raid10_write_request()
1487 atomic_set(&r10_bio->remaining, 1); in raid10_write_request()
1488 md_bitmap_startwrite(mddev->bitmap, r10_bio->sector, r10_bio->sectors, 0); in raid10_write_request()
1490 for (i = 0; i < conf->copies; i++) { in raid10_write_request()
1491 if (r10_bio->devs[i].bio) in raid10_write_request()
1493 if (r10_bio->devs[i].repl_bio) in raid10_write_request()
1501 struct r10conf *conf = mddev->private; in __make_request()
1504 r10_bio = mempool_alloc(&conf->r10bio_pool, GFP_NOIO); in __make_request()
1506 r10_bio->master_bio = bio; in __make_request()
1507 r10_bio->sectors = sectors; in __make_request()
1509 r10_bio->mddev = mddev; in __make_request()
1510 r10_bio->sector = bio->bi_iter.bi_sector; in __make_request()
1511 r10_bio->state = 0; in __make_request()
1512 r10_bio->read_slot = -1; in __make_request()
1513 memset(r10_bio->devs, 0, sizeof(r10_bio->devs[0]) * conf->copies); in __make_request()
1523 struct r10conf *conf = mddev->private; in raid10_make_request()
1524 sector_t chunk_mask = (conf->geo.chunk_mask & conf->prev.chunk_mask); in raid10_make_request()
1528 if (unlikely(bio->bi_opf & REQ_PREFLUSH) in raid10_make_request()
1539 if (unlikely((bio->bi_iter.bi_sector & chunk_mask) + in raid10_make_request()
1541 && (conf->geo.near_copies < conf->geo.raid_disks in raid10_make_request()
1542 || conf->prev.near_copies < in raid10_make_request()
1543 conf->prev.raid_disks))) in raid10_make_request()
1544 sectors = chunk_sects - in raid10_make_request()
1545 (bio->bi_iter.bi_sector & in raid10_make_request()
1546 (chunk_sects - 1)); in raid10_make_request()
1550 wake_up(&conf->wait_barrier); in raid10_make_request()
1556 struct r10conf *conf = mddev->private; in raid10_status()
1559 if (conf->geo.near_copies < conf->geo.raid_disks) in raid10_status()
1560 seq_printf(seq, " %dK chunks", mddev->chunk_sectors / 2); in raid10_status()
1561 if (conf->geo.near_copies > 1) in raid10_status()
1562 seq_printf(seq, " %d near-copies", conf->geo.near_copies); in raid10_status()
1563 if (conf->geo.far_copies > 1) { in raid10_status()
1564 if (conf->geo.far_offset) in raid10_status()
1565 seq_printf(seq, " %d offset-copies", conf->geo.far_copies); in raid10_status()
1567 seq_printf(seq, " %d far-copies", conf->geo.far_copies); in raid10_status()
1568 if (conf->geo.far_set_size != conf->geo.raid_disks) in raid10_status()
1569 seq_printf(seq, " %d devices per set", conf->geo.far_set_size); in raid10_status()
1571 seq_printf(seq, " [%d/%d] [", conf->geo.raid_disks, in raid10_status()
1572 conf->geo.raid_disks - mddev->degraded); in raid10_status()
1574 for (i = 0; i < conf->geo.raid_disks; i++) { in raid10_status()
1575 struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev); in raid10_status()
1576 seq_printf(seq, "%s", rdev && test_bit(In_sync, &rdev->flags) ? "U" : "_"); in raid10_status()
1593 disks = conf->prev.raid_disks; in _enough()
1594 ncopies = conf->prev.near_copies; in _enough()
1596 disks = conf->geo.raid_disks; in _enough()
1597 ncopies = conf->geo.near_copies; in _enough()
1602 int n = conf->copies; in _enough()
1605 while (n--) { in _enough()
1608 (rdev = rcu_dereference(conf->mirrors[this].rdev)) && in _enough()
1609 test_bit(In_sync, &rdev->flags)) in _enough()
1627 * This is ensured if ->reconfig_mutex or ->device_lock in enough()
1637 struct r10conf *conf = mddev->private; in raid10_error()
1646 spin_lock_irqsave(&conf->device_lock, flags); in raid10_error()
1647 if (test_bit(In_sync, &rdev->flags) && !mddev->fail_last_dev in raid10_error()
1648 && !enough(conf, rdev->raid_disk)) { in raid10_error()
1652 spin_unlock_irqrestore(&conf->device_lock, flags); in raid10_error()
1655 if (test_and_clear_bit(In_sync, &rdev->flags)) in raid10_error()
1656 mddev->degraded++; in raid10_error()
1660 set_bit(MD_RECOVERY_INTR, &mddev->recovery); in raid10_error()
1661 set_bit(Blocked, &rdev->flags); in raid10_error()
1662 set_bit(Faulty, &rdev->flags); in raid10_error()
1663 set_mask_bits(&mddev->sb_flags, 0, in raid10_error()
1665 spin_unlock_irqrestore(&conf->device_lock, flags); in raid10_error()
1668 mdname(mddev), bdevname(rdev->bdev, b), in raid10_error()
1669 mdname(mddev), conf->geo.raid_disks - mddev->degraded); in raid10_error()
1682 pr_debug(" --- wd:%d rd:%d\n", conf->geo.raid_disks - conf->mddev->degraded, in print_conf()
1683 conf->geo.raid_disks); in print_conf()
1685 /* This is only called with ->reconfix_mutex held, so in print_conf()
1687 for (i = 0; i < conf->geo.raid_disks; i++) { in print_conf()
1689 rdev = conf->mirrors[i].rdev; in print_conf()
1692 i, !test_bit(In_sync, &rdev->flags), in print_conf()
1693 !test_bit(Faulty, &rdev->flags), in print_conf()
1694 bdevname(rdev->bdev,b)); in print_conf()
1703 mempool_exit(&conf->r10buf_pool); in close_sync()
1709 struct r10conf *conf = mddev->private; in raid10_spare_active()
1715 * Find all non-in_sync disks within the RAID10 configuration in raid10_spare_active()
1718 for (i = 0; i < conf->geo.raid_disks; i++) { in raid10_spare_active()
1719 tmp = conf->mirrors + i; in raid10_spare_active()
1720 if (tmp->replacement in raid10_spare_active()
1721 && tmp->replacement->recovery_offset == MaxSector in raid10_spare_active()
1722 && !test_bit(Faulty, &tmp->replacement->flags) in raid10_spare_active()
1723 && !test_and_set_bit(In_sync, &tmp->replacement->flags)) { in raid10_spare_active()
1725 if (!tmp->rdev in raid10_spare_active()
1726 || !test_and_clear_bit(In_sync, &tmp->rdev->flags)) in raid10_spare_active()
1728 if (tmp->rdev) { in raid10_spare_active()
1731 * and never re-added. in raid10_spare_active()
1733 set_bit(Faulty, &tmp->rdev->flags); in raid10_spare_active()
1735 tmp->rdev->sysfs_state); in raid10_spare_active()
1737 sysfs_notify_dirent_safe(tmp->replacement->sysfs_state); in raid10_spare_active()
1738 } else if (tmp->rdev in raid10_spare_active()
1739 && tmp->rdev->recovery_offset == MaxSector in raid10_spare_active()
1740 && !test_bit(Faulty, &tmp->rdev->flags) in raid10_spare_active()
1741 && !test_and_set_bit(In_sync, &tmp->rdev->flags)) { in raid10_spare_active()
1743 sysfs_notify_dirent_safe(tmp->rdev->sysfs_state); in raid10_spare_active()
1746 spin_lock_irqsave(&conf->device_lock, flags); in raid10_spare_active()
1747 mddev->degraded -= count; in raid10_spare_active()
1748 spin_unlock_irqrestore(&conf->device_lock, flags); in raid10_spare_active()
1756 struct r10conf *conf = mddev->private; in raid10_add_disk()
1757 int err = -EEXIST; in raid10_add_disk()
1760 int last = conf->geo.raid_disks - 1; in raid10_add_disk()
1762 if (mddev->recovery_cp < MaxSector) in raid10_add_disk()
1763 /* only hot-add to in-sync arrays, as recovery is in raid10_add_disk()
1766 return -EBUSY; in raid10_add_disk()
1767 if (rdev->saved_raid_disk < 0 && !_enough(conf, 1, -1)) in raid10_add_disk()
1768 return -EINVAL; in raid10_add_disk()
1771 return -ENXIO; in raid10_add_disk()
1773 if (rdev->raid_disk >= 0) in raid10_add_disk()
1774 first = last = rdev->raid_disk; in raid10_add_disk()
1776 if (rdev->saved_raid_disk >= first && in raid10_add_disk()
1777 rdev->saved_raid_disk < conf->geo.raid_disks && in raid10_add_disk()
1778 conf->mirrors[rdev->saved_raid_disk].rdev == NULL) in raid10_add_disk()
1779 mirror = rdev->saved_raid_disk; in raid10_add_disk()
1783 struct raid10_info *p = &conf->mirrors[mirror]; in raid10_add_disk()
1784 if (p->recovery_disabled == mddev->recovery_disabled) in raid10_add_disk()
1786 if (p->rdev) { in raid10_add_disk()
1787 if (!test_bit(WantReplacement, &p->rdev->flags) || in raid10_add_disk()
1788 p->replacement != NULL) in raid10_add_disk()
1790 clear_bit(In_sync, &rdev->flags); in raid10_add_disk()
1791 set_bit(Replacement, &rdev->flags); in raid10_add_disk()
1792 rdev->raid_disk = mirror; in raid10_add_disk()
1794 if (mddev->gendisk) in raid10_add_disk()
1795 disk_stack_limits(mddev->gendisk, rdev->bdev, in raid10_add_disk()
1796 rdev->data_offset << 9); in raid10_add_disk()
1797 conf->fullsync = 1; in raid10_add_disk()
1798 rcu_assign_pointer(p->replacement, rdev); in raid10_add_disk()
1802 if (mddev->gendisk) in raid10_add_disk()
1803 disk_stack_limits(mddev->gendisk, rdev->bdev, in raid10_add_disk()
1804 rdev->data_offset << 9); in raid10_add_disk()
1806 p->head_position = 0; in raid10_add_disk()
1807 p->recovery_disabled = mddev->recovery_disabled - 1; in raid10_add_disk()
1808 rdev->raid_disk = mirror; in raid10_add_disk()
1810 if (rdev->saved_raid_disk != mirror) in raid10_add_disk()
1811 conf->fullsync = 1; in raid10_add_disk()
1812 rcu_assign_pointer(p->rdev, rdev); in raid10_add_disk()
1815 if (mddev->queue && blk_queue_discard(bdev_get_queue(rdev->bdev))) in raid10_add_disk()
1816 blk_queue_flag_set(QUEUE_FLAG_DISCARD, mddev->queue); in raid10_add_disk()
1824 struct r10conf *conf = mddev->private; in raid10_remove_disk()
1826 int number = rdev->raid_disk; in raid10_remove_disk()
1831 if (unlikely(number >= mddev->raid_disks)) in raid10_remove_disk()
1833 p = conf->mirrors + number; in raid10_remove_disk()
1834 if (rdev == p->rdev) in raid10_remove_disk()
1835 rdevp = &p->rdev; in raid10_remove_disk()
1836 else if (rdev == p->replacement) in raid10_remove_disk()
1837 rdevp = &p->replacement; in raid10_remove_disk()
1841 if (test_bit(In_sync, &rdev->flags) || in raid10_remove_disk()
1842 atomic_read(&rdev->nr_pending)) { in raid10_remove_disk()
1843 err = -EBUSY; in raid10_remove_disk()
1846 /* Only remove non-faulty devices if recovery in raid10_remove_disk()
1849 if (!test_bit(Faulty, &rdev->flags) && in raid10_remove_disk()
1850 mddev->recovery_disabled != p->recovery_disabled && in raid10_remove_disk()
1851 (!p->replacement || p->replacement == rdev) && in raid10_remove_disk()
1852 number < conf->geo.raid_disks && in raid10_remove_disk()
1853 enough(conf, -1)) { in raid10_remove_disk()
1854 err = -EBUSY; in raid10_remove_disk()
1858 if (!test_bit(RemoveSynchronized, &rdev->flags)) { in raid10_remove_disk()
1860 if (atomic_read(&rdev->nr_pending)) { in raid10_remove_disk()
1862 err = -EBUSY; in raid10_remove_disk()
1867 if (p->replacement) { in raid10_remove_disk()
1869 p->rdev = p->replacement; in raid10_remove_disk()
1870 clear_bit(Replacement, &p->replacement->flags); in raid10_remove_disk()
1872 * but will never see neither -- if they are careful. in raid10_remove_disk()
1874 p->replacement = NULL; in raid10_remove_disk()
1877 clear_bit(WantReplacement, &rdev->flags); in raid10_remove_disk()
1888 struct r10conf *conf = r10_bio->mddev->private; in __end_sync_read()
1890 if (!bio->bi_status) in __end_sync_read()
1891 set_bit(R10BIO_Uptodate, &r10_bio->state); in __end_sync_read()
1896 atomic_add(r10_bio->sectors, in __end_sync_read()
1897 &conf->mirrors[d].rdev->corrected_errors); in __end_sync_read()
1902 rdev_dec_pending(conf->mirrors[d].rdev, conf->mddev); in __end_sync_read()
1903 if (test_bit(R10BIO_IsRecover, &r10_bio->state) || in __end_sync_read()
1904 atomic_dec_and_test(&r10_bio->remaining)) { in __end_sync_read()
1915 struct r10conf *conf = r10_bio->mddev->private; in end_sync_read()
1924 struct r10bio *r10_bio = bio->bi_private; in end_reshape_read()
1926 __end_sync_read(r10_bio, bio, r10_bio->read_slot); in end_reshape_read()
1931 struct mddev *mddev = r10_bio->mddev; in end_sync_request()
1933 while (atomic_dec_and_test(&r10_bio->remaining)) { in end_sync_request()
1934 if (r10_bio->master_bio == NULL) { in end_sync_request()
1936 sector_t s = r10_bio->sectors; in end_sync_request()
1937 if (test_bit(R10BIO_MadeGood, &r10_bio->state) || in end_sync_request()
1938 test_bit(R10BIO_WriteError, &r10_bio->state)) in end_sync_request()
1945 struct r10bio *r10_bio2 = (struct r10bio *)r10_bio->master_bio; in end_sync_request()
1946 if (test_bit(R10BIO_MadeGood, &r10_bio->state) || in end_sync_request()
1947 test_bit(R10BIO_WriteError, &r10_bio->state)) in end_sync_request()
1959 struct mddev *mddev = r10_bio->mddev; in end_sync_write()
1960 struct r10conf *conf = mddev->private; in end_sync_write()
1970 rdev = conf->mirrors[d].replacement; in end_sync_write()
1972 rdev = conf->mirrors[d].rdev; in end_sync_write()
1974 if (bio->bi_status) { in end_sync_write()
1978 set_bit(WriteErrorSeen, &rdev->flags); in end_sync_write()
1979 if (!test_and_set_bit(WantReplacement, &rdev->flags)) in end_sync_write()
1981 &rdev->mddev->recovery); in end_sync_write()
1982 set_bit(R10BIO_WriteError, &r10_bio->state); in end_sync_write()
1985 r10_bio->devs[slot].addr, in end_sync_write()
1986 r10_bio->sectors, in end_sync_write()
1988 set_bit(R10BIO_MadeGood, &r10_bio->state); in end_sync_write()
2008 * We check if all blocks are in-sync and only write to blocks that
2013 struct r10conf *conf = mddev->private; in sync_request_write()
2019 atomic_set(&r10_bio->remaining, 1); in sync_request_write()
2022 for (i=0; i<conf->copies; i++) in sync_request_write()
2023 if (!r10_bio->devs[i].bio->bi_status) in sync_request_write()
2026 if (i == conf->copies) in sync_request_write()
2030 fbio = r10_bio->devs[i].bio; in sync_request_write()
2031 fbio->bi_iter.bi_size = r10_bio->sectors << 9; in sync_request_write()
2032 fbio->bi_iter.bi_idx = 0; in sync_request_write()
2033 fpages = get_resync_pages(fbio)->pages; in sync_request_write()
2035 vcnt = (r10_bio->sectors + (PAGE_SIZE >> 9) - 1) >> (PAGE_SHIFT - 9); in sync_request_write()
2037 for (i=0 ; i < conf->copies ; i++) { in sync_request_write()
2042 tbio = r10_bio->devs[i].bio; in sync_request_write()
2044 if (tbio->bi_end_io != end_sync_read) in sync_request_write()
2049 tpages = get_resync_pages(tbio)->pages; in sync_request_write()
2050 d = r10_bio->devs[i].devnum; in sync_request_write()
2051 rdev = conf->mirrors[d].rdev; in sync_request_write()
2052 if (!r10_bio->devs[i].bio->bi_status) { in sync_request_write()
2057 int sectors = r10_bio->sectors; in sync_request_write()
2066 sectors -= len/512; in sync_request_write()
2070 atomic64_add(r10_bio->sectors, &mddev->resync_mismatches); in sync_request_write()
2071 if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)) in sync_request_write()
2074 } else if (test_bit(FailFast, &rdev->flags)) { in sync_request_write()
2076 md_error(rdev->mddev, rdev); in sync_request_write()
2081 * First we need to fixup bv_offset, bv_len and in sync_request_write()
2087 md_bio_reset_resync_pages(tbio, rp, fbio->bi_iter.bi_size); in sync_request_write()
2089 rp->raid_bio = r10_bio; in sync_request_write()
2090 tbio->bi_private = rp; in sync_request_write()
2091 tbio->bi_iter.bi_sector = r10_bio->devs[i].addr; in sync_request_write()
2092 tbio->bi_end_io = end_sync_write; in sync_request_write()
2097 atomic_inc(&conf->mirrors[d].rdev->nr_pending); in sync_request_write()
2098 atomic_inc(&r10_bio->remaining); in sync_request_write()
2099 md_sync_acct(conf->mirrors[d].rdev->bdev, bio_sectors(tbio)); in sync_request_write()
2101 if (test_bit(FailFast, &conf->mirrors[d].rdev->flags)) in sync_request_write()
2102 tbio->bi_opf |= MD_FAILFAST; in sync_request_write()
2103 tbio->bi_iter.bi_sector += conf->mirrors[d].rdev->data_offset; in sync_request_write()
2104 bio_set_dev(tbio, conf->mirrors[d].rdev->bdev); in sync_request_write()
2111 for (i = 0; i < conf->copies; i++) { in sync_request_write()
2114 tbio = r10_bio->devs[i].repl_bio; in sync_request_write()
2115 if (!tbio || !tbio->bi_end_io) in sync_request_write()
2117 if (r10_bio->devs[i].bio->bi_end_io != end_sync_write in sync_request_write()
2118 && r10_bio->devs[i].bio != fbio) in sync_request_write()
2120 d = r10_bio->devs[i].devnum; in sync_request_write()
2121 atomic_inc(&r10_bio->remaining); in sync_request_write()
2122 md_sync_acct(conf->mirrors[d].replacement->bdev, in sync_request_write()
2128 if (atomic_dec_and_test(&r10_bio->remaining)) { in sync_request_write()
2129 md_done_sync(mddev, r10_bio->sectors, 1); in sync_request_write()
2137 * We recover all non-is_sync drives by finding the virtual address of
2139 * There is a separate r10_bio for each non-in_sync drive.
2147 * We repeat the read in smaller page-sized sections. in fix_recovery_read_error()
2153 struct mddev *mddev = r10_bio->mddev; in fix_recovery_read_error()
2154 struct r10conf *conf = mddev->private; in fix_recovery_read_error()
2155 struct bio *bio = r10_bio->devs[0].bio; in fix_recovery_read_error()
2157 int sectors = r10_bio->sectors; in fix_recovery_read_error()
2159 int dr = r10_bio->devs[0].devnum; in fix_recovery_read_error()
2160 int dw = r10_bio->devs[1].devnum; in fix_recovery_read_error()
2161 struct page **pages = get_resync_pages(bio)->pages; in fix_recovery_read_error()
2172 rdev = conf->mirrors[dr].rdev; in fix_recovery_read_error()
2173 addr = r10_bio->devs[0].addr + sect, in fix_recovery_read_error()
2180 rdev = conf->mirrors[dw].rdev; in fix_recovery_read_error()
2181 addr = r10_bio->devs[1].addr + sect; in fix_recovery_read_error()
2188 set_bit(WriteErrorSeen, &rdev->flags); in fix_recovery_read_error()
2190 &rdev->flags)) in fix_recovery_read_error()
2192 &rdev->mddev->recovery); in fix_recovery_read_error()
2196 /* We don't worry if we cannot set a bad block - in fix_recovery_read_error()
2202 if (rdev != conf->mirrors[dw].rdev) { in fix_recovery_read_error()
2204 struct md_rdev *rdev2 = conf->mirrors[dw].rdev; in fix_recovery_read_error()
2205 addr = r10_bio->devs[1].addr + sect; in fix_recovery_read_error()
2212 conf->mirrors[dw].recovery_disabled in fix_recovery_read_error()
2213 = mddev->recovery_disabled; in fix_recovery_read_error()
2215 &mddev->recovery); in fix_recovery_read_error()
2221 sectors -= s; in fix_recovery_read_error()
2229 struct r10conf *conf = mddev->private; in recovery_request_write()
2231 struct bio *wbio = r10_bio->devs[1].bio; in recovery_request_write()
2232 struct bio *wbio2 = r10_bio->devs[1].repl_bio; in recovery_request_write()
2234 /* Need to test wbio2->bi_end_io before we call in recovery_request_write()
2238 if (wbio2 && !wbio2->bi_end_io) in recovery_request_write()
2241 if (!test_bit(R10BIO_Uptodate, &r10_bio->state)) { in recovery_request_write()
2243 if (wbio->bi_end_io) in recovery_request_write()
2254 d = r10_bio->devs[1].devnum; in recovery_request_write()
2255 if (wbio->bi_end_io) { in recovery_request_write()
2256 atomic_inc(&conf->mirrors[d].rdev->nr_pending); in recovery_request_write()
2257 md_sync_acct(conf->mirrors[d].rdev->bdev, bio_sectors(wbio)); in recovery_request_write()
2261 atomic_inc(&conf->mirrors[d].replacement->nr_pending); in recovery_request_write()
2262 md_sync_acct(conf->mirrors[d].replacement->bdev, in recovery_request_write()
2278 unsigned int read_errors = atomic_read(&rdev->read_errors); in check_decay_read_errors()
2282 if (rdev->last_read_error == 0) { in check_decay_read_errors()
2284 rdev->last_read_error = cur_time_mon; in check_decay_read_errors()
2288 hours_since_last = (long)(cur_time_mon - in check_decay_read_errors()
2289 rdev->last_read_error) / 3600; in check_decay_read_errors()
2291 rdev->last_read_error = cur_time_mon; in check_decay_read_errors()
2299 atomic_set(&rdev->read_errors, 0); in check_decay_read_errors()
2301 atomic_set(&rdev->read_errors, read_errors >> hours_since_last); in check_decay_read_errors()
2311 && (rw == READ || test_bit(WriteErrorSeen, &rdev->flags))) in r10_sync_page_io()
2312 return -1; in r10_sync_page_io()
2317 set_bit(WriteErrorSeen, &rdev->flags); in r10_sync_page_io()
2318 if (!test_and_set_bit(WantReplacement, &rdev->flags)) in r10_sync_page_io()
2320 &rdev->mddev->recovery); in r10_sync_page_io()
2322 /* need to record an error - either for the block or the device */ in r10_sync_page_io()
2324 md_error(rdev->mddev, rdev); in r10_sync_page_io()
2338 int sect = 0; /* Offset from r10_bio->sector */ in fix_read_error()
2339 int sectors = r10_bio->sectors; in fix_read_error()
2341 int max_read_errors = atomic_read(&mddev->max_corr_read_errors); in fix_read_error()
2342 int d = r10_bio->devs[r10_bio->read_slot].devnum; in fix_read_error()
2347 rdev = conf->mirrors[d].rdev; in fix_read_error()
2349 if (test_bit(Faulty, &rdev->flags)) in fix_read_error()
2355 atomic_inc(&rdev->read_errors); in fix_read_error()
2356 if (atomic_read(&rdev->read_errors) > max_read_errors) { in fix_read_error()
2358 bdevname(rdev->bdev, b); in fix_read_error()
2362 atomic_read(&rdev->read_errors), max_read_errors); in fix_read_error()
2366 r10_bio->devs[r10_bio->read_slot].bio = IO_BLOCKED; in fix_read_error()
2372 int sl = r10_bio->read_slot; in fix_read_error()
2384 d = r10_bio->devs[sl].devnum; in fix_read_error()
2385 rdev = rcu_dereference(conf->mirrors[d].rdev); in fix_read_error()
2387 test_bit(In_sync, &rdev->flags) && in fix_read_error()
2388 !test_bit(Faulty, &rdev->flags) && in fix_read_error()
2389 is_badblock(rdev, r10_bio->devs[sl].addr + sect, s, in fix_read_error()
2391 atomic_inc(&rdev->nr_pending); in fix_read_error()
2394 r10_bio->devs[sl].addr + in fix_read_error()
2397 conf->tmppage, in fix_read_error()
2405 if (sl == conf->copies) in fix_read_error()
2407 } while (!success && sl != r10_bio->read_slot); in fix_read_error()
2415 int dn = r10_bio->devs[r10_bio->read_slot].devnum; in fix_read_error()
2416 rdev = conf->mirrors[dn].rdev; in fix_read_error()
2420 r10_bio->devs[r10_bio->read_slot].addr in fix_read_error()
2424 r10_bio->devs[r10_bio->read_slot].bio in fix_read_error()
2431 /* write it back and re-read */ in fix_read_error()
2433 while (sl != r10_bio->read_slot) { in fix_read_error()
2437 sl = conf->copies; in fix_read_error()
2438 sl--; in fix_read_error()
2439 d = r10_bio->devs[sl].devnum; in fix_read_error()
2440 rdev = rcu_dereference(conf->mirrors[d].rdev); in fix_read_error()
2442 test_bit(Faulty, &rdev->flags) || in fix_read_error()
2443 !test_bit(In_sync, &rdev->flags)) in fix_read_error()
2446 atomic_inc(&rdev->nr_pending); in fix_read_error()
2449 r10_bio->devs[sl].addr + in fix_read_error()
2451 s, conf->tmppage, WRITE) in fix_read_error()
2460 bdevname(rdev->bdev, b)); in fix_read_error()
2463 bdevname(rdev->bdev, b)); in fix_read_error()
2469 while (sl != r10_bio->read_slot) { in fix_read_error()
2473 sl = conf->copies; in fix_read_error()
2474 sl--; in fix_read_error()
2475 d = r10_bio->devs[sl].devnum; in fix_read_error()
2476 rdev = rcu_dereference(conf->mirrors[d].rdev); in fix_read_error()
2478 test_bit(Faulty, &rdev->flags) || in fix_read_error()
2479 !test_bit(In_sync, &rdev->flags)) in fix_read_error()
2482 atomic_inc(&rdev->nr_pending); in fix_read_error()
2485 r10_bio->devs[sl].addr + in fix_read_error()
2487 s, conf->tmppage, in fix_read_error()
2496 bdevname(rdev->bdev, b)); in fix_read_error()
2499 bdevname(rdev->bdev, b)); in fix_read_error()
2507 bdevname(rdev->bdev, b)); in fix_read_error()
2508 atomic_add(s, &rdev->corrected_errors); in fix_read_error()
2516 sectors -= s; in fix_read_error()
2523 struct bio *bio = r10_bio->master_bio; in narrow_write_error()
2524 struct mddev *mddev = r10_bio->mddev; in narrow_write_error()
2525 struct r10conf *conf = mddev->private; in narrow_write_error()
2526 struct md_rdev *rdev = conf->mirrors[r10_bio->devs[i].devnum].rdev; in narrow_write_error()
2541 int sect_to_write = r10_bio->sectors; in narrow_write_error()
2544 if (rdev->badblocks.shift < 0) in narrow_write_error()
2547 block_sectors = roundup(1 << rdev->badblocks.shift, in narrow_write_error()
2548 bdev_logical_block_size(rdev->bdev) >> 9); in narrow_write_error()
2549 sector = r10_bio->sector; in narrow_write_error()
2550 sectors = ((r10_bio->sector + block_sectors) in narrow_write_error()
2551 & ~(sector_t)(block_sectors - 1)) in narrow_write_error()
2552 - sector; in narrow_write_error()
2560 wbio = bio_clone_fast(bio, GFP_NOIO, &mddev->bio_set); in narrow_write_error()
2561 bio_trim(wbio, sector - bio->bi_iter.bi_sector, sectors); in narrow_write_error()
2562 wsector = r10_bio->devs[i].addr + (sector - r10_bio->sector); in narrow_write_error()
2563 wbio->bi_iter.bi_sector = wsector + in narrow_write_error()
2565 bio_set_dev(wbio, rdev->bdev); in narrow_write_error()
2575 sect_to_write -= sectors; in narrow_write_error()
2584 int slot = r10_bio->read_slot; in handle_read_error()
2586 struct r10conf *conf = mddev->private; in handle_read_error()
2587 struct md_rdev *rdev = r10_bio->devs[slot].rdev; in handle_read_error()
2592 * other devices. When we find one, we re-write in handle_read_error()
2597 bio = r10_bio->devs[slot].bio; in handle_read_error()
2599 r10_bio->devs[slot].bio = NULL; in handle_read_error()
2601 if (mddev->ro) in handle_read_error()
2602 r10_bio->devs[slot].bio = IO_BLOCKED; in handle_read_error()
2603 else if (!test_bit(FailFast, &rdev->flags)) { in handle_read_error()
2612 r10_bio->state = 0; in handle_read_error()
2613 raid10_read_request(mddev, r10_bio->master_bio, r10_bio); in handle_read_error()
2627 if (test_bit(R10BIO_IsSync, &r10_bio->state) || in handle_write_completed()
2628 test_bit(R10BIO_IsRecover, &r10_bio->state)) { in handle_write_completed()
2629 for (m = 0; m < conf->copies; m++) { in handle_write_completed()
2630 int dev = r10_bio->devs[m].devnum; in handle_write_completed()
2631 rdev = conf->mirrors[dev].rdev; in handle_write_completed()
2632 if (r10_bio->devs[m].bio == NULL || in handle_write_completed()
2633 r10_bio->devs[m].bio->bi_end_io == NULL) in handle_write_completed()
2635 if (!r10_bio->devs[m].bio->bi_status) { in handle_write_completed()
2638 r10_bio->devs[m].addr, in handle_write_completed()
2639 r10_bio->sectors, 0); in handle_write_completed()
2643 r10_bio->devs[m].addr, in handle_write_completed()
2644 r10_bio->sectors, 0)) in handle_write_completed()
2645 md_error(conf->mddev, rdev); in handle_write_completed()
2647 rdev = conf->mirrors[dev].replacement; in handle_write_completed()
2648 if (r10_bio->devs[m].repl_bio == NULL || in handle_write_completed()
2649 r10_bio->devs[m].repl_bio->bi_end_io == NULL) in handle_write_completed()
2652 if (!r10_bio->devs[m].repl_bio->bi_status) { in handle_write_completed()
2655 r10_bio->devs[m].addr, in handle_write_completed()
2656 r10_bio->sectors, 0); in handle_write_completed()
2660 r10_bio->devs[m].addr, in handle_write_completed()
2661 r10_bio->sectors, 0)) in handle_write_completed()
2662 md_error(conf->mddev, rdev); in handle_write_completed()
2668 for (m = 0; m < conf->copies; m++) { in handle_write_completed()
2669 int dev = r10_bio->devs[m].devnum; in handle_write_completed()
2670 struct bio *bio = r10_bio->devs[m].bio; in handle_write_completed()
2671 rdev = conf->mirrors[dev].rdev; in handle_write_completed()
2675 r10_bio->devs[m].addr, in handle_write_completed()
2676 r10_bio->sectors, 0); in handle_write_completed()
2677 rdev_dec_pending(rdev, conf->mddev); in handle_write_completed()
2678 } else if (bio != NULL && bio->bi_status) { in handle_write_completed()
2681 md_error(conf->mddev, rdev); in handle_write_completed()
2683 &r10_bio->state); in handle_write_completed()
2685 rdev_dec_pending(rdev, conf->mddev); in handle_write_completed()
2687 bio = r10_bio->devs[m].repl_bio; in handle_write_completed()
2688 rdev = conf->mirrors[dev].replacement; in handle_write_completed()
2692 r10_bio->devs[m].addr, in handle_write_completed()
2693 r10_bio->sectors, 0); in handle_write_completed()
2694 rdev_dec_pending(rdev, conf->mddev); in handle_write_completed()
2698 spin_lock_irq(&conf->device_lock); in handle_write_completed()
2699 list_add(&r10_bio->retry_list, &conf->bio_end_io_list); in handle_write_completed()
2700 conf->nr_queued++; in handle_write_completed()
2701 spin_unlock_irq(&conf->device_lock); in handle_write_completed()
2706 wake_up(&conf->wait_barrier); in handle_write_completed()
2707 md_wakeup_thread(conf->mddev->thread); in handle_write_completed()
2710 &r10_bio->state)) in handle_write_completed()
2719 struct mddev *mddev = thread->mddev; in raid10d()
2722 struct r10conf *conf = mddev->private; in raid10d()
2723 struct list_head *head = &conf->retry_list; in raid10d()
2728 if (!list_empty_careful(&conf->bio_end_io_list) && in raid10d()
2729 !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)) { in raid10d()
2731 spin_lock_irqsave(&conf->device_lock, flags); in raid10d()
2732 if (!test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)) { in raid10d()
2733 while (!list_empty(&conf->bio_end_io_list)) { in raid10d()
2734 list_move(conf->bio_end_io_list.prev, &tmp); in raid10d()
2735 conf->nr_queued--; in raid10d()
2738 spin_unlock_irqrestore(&conf->device_lock, flags); in raid10d()
2742 list_del(&r10_bio->retry_list); in raid10d()
2743 if (mddev->degraded) in raid10d()
2744 set_bit(R10BIO_Degraded, &r10_bio->state); in raid10d()
2747 &r10_bio->state)) in raid10d()
2758 spin_lock_irqsave(&conf->device_lock, flags); in raid10d()
2760 spin_unlock_irqrestore(&conf->device_lock, flags); in raid10d()
2763 r10_bio = list_entry(head->prev, struct r10bio, retry_list); in raid10d()
2764 list_del(head->prev); in raid10d()
2765 conf->nr_queued--; in raid10d()
2766 spin_unlock_irqrestore(&conf->device_lock, flags); in raid10d()
2768 mddev = r10_bio->mddev; in raid10d()
2769 conf = mddev->private; in raid10d()
2770 if (test_bit(R10BIO_MadeGood, &r10_bio->state) || in raid10d()
2771 test_bit(R10BIO_WriteError, &r10_bio->state)) in raid10d()
2773 else if (test_bit(R10BIO_IsReshape, &r10_bio->state)) in raid10d()
2775 else if (test_bit(R10BIO_IsSync, &r10_bio->state)) in raid10d()
2777 else if (test_bit(R10BIO_IsRecover, &r10_bio->state)) in raid10d()
2779 else if (test_bit(R10BIO_ReadError, &r10_bio->state)) in raid10d()
2785 if (mddev->sb_flags & ~(1<<MD_SB_CHANGE_PENDING)) in raid10d()
2796 BUG_ON(mempool_initialized(&conf->r10buf_pool)); in init_resync()
2797 conf->have_replacement = 0; in init_resync()
2798 for (i = 0; i < conf->geo.raid_disks; i++) in init_resync()
2799 if (conf->mirrors[i].replacement) in init_resync()
2800 conf->have_replacement = 1; in init_resync()
2801 ret = mempool_init(&conf->r10buf_pool, buffs, in init_resync()
2805 conf->next_resync = 0; in init_resync()
2811 struct r10bio *r10bio = mempool_alloc(&conf->r10buf_pool, GFP_NOIO); in raid10_alloc_init_r10buf()
2817 if (test_bit(MD_RECOVERY_SYNC, &conf->mddev->recovery) || in raid10_alloc_init_r10buf()
2818 test_bit(MD_RECOVERY_RESHAPE, &conf->mddev->recovery)) in raid10_alloc_init_r10buf()
2819 nalloc = conf->copies; /* resync */ in raid10_alloc_init_r10buf()
2824 bio = r10bio->devs[i].bio; in raid10_alloc_init_r10buf()
2825 rp = bio->bi_private; in raid10_alloc_init_r10buf()
2827 bio->bi_private = rp; in raid10_alloc_init_r10buf()
2828 bio = r10bio->devs[i].repl_bio; in raid10_alloc_init_r10buf()
2830 rp = bio->bi_private; in raid10_alloc_init_r10buf()
2832 bio->bi_private = rp; in raid10_alloc_init_r10buf()
2859 chunks = conf->geo.raid_disks / conf->geo.near_copies; in raid10_set_cluster_sync_high()
2860 if (conf->geo.raid_disks % conf->geo.near_copies == 0) in raid10_set_cluster_sync_high()
2864 window_size = (chunks + extra_chunk) * conf->mddev->chunk_sectors; in raid10_set_cluster_sync_high()
2872 conf->cluster_sync_high = conf->cluster_sync_low + window_size; in raid10_set_cluster_sync_high()
2878 * We need to make sure that no normal I/O request - particularly write
2879 * requests - conflict with active sync requests.
2885 * We differentiate by looking at MD_RECOVERY_SYNC in mddev->recovery.
2888 * and update if there are differences. If only one copy is live,
2891 * value for each non-in_sync drive, and over-write.
2894 * given address, one for each out-of-sync device. We model this by allocating
2895 * a number of r10_bio structures, one for each out-of-sync device.
2901 * This link is counted in ->remaining. When the r10_bio that points to NULL
2910 struct r10conf *conf = mddev->private; in raid10_sync_request()
2919 sector_t chunk_mask = conf->geo.chunk_mask; in raid10_sync_request()
2926 if (mddev->bitmap == NULL && in raid10_sync_request()
2927 mddev->recovery_cp == MaxSector && in raid10_sync_request()
2928 mddev->reshape_position == MaxSector && in raid10_sync_request()
2929 !test_bit(MD_RECOVERY_SYNC, &mddev->recovery) && in raid10_sync_request()
2930 !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) && in raid10_sync_request()
2931 !test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) && in raid10_sync_request()
2932 conf->fullsync == 0) { in raid10_sync_request()
2934 return mddev->dev_sectors - sector_nr; in raid10_sync_request()
2937 if (!mempool_initialized(&conf->r10buf_pool)) in raid10_sync_request()
2942 max_sector = mddev->dev_sectors; in raid10_sync_request()
2943 if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) || in raid10_sync_request()
2944 test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) in raid10_sync_request()
2945 max_sector = mddev->resync_max_sectors; in raid10_sync_request()
2947 conf->cluster_sync_low = 0; in raid10_sync_request()
2948 conf->cluster_sync_high = 0; in raid10_sync_request()
2955 * mddev->curr_resync, but for recovery, in raid10_sync_request()
2959 if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) { in raid10_sync_request()
2965 if (mddev->curr_resync < max_sector) { /* aborted */ in raid10_sync_request()
2966 if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) in raid10_sync_request()
2967 md_bitmap_end_sync(mddev->bitmap, mddev->curr_resync, in raid10_sync_request()
2969 else for (i = 0; i < conf->geo.raid_disks; i++) { in raid10_sync_request()
2971 raid10_find_virt(conf, mddev->curr_resync, i); in raid10_sync_request()
2972 md_bitmap_end_sync(mddev->bitmap, sect, in raid10_sync_request()
2977 if ((!mddev->bitmap || conf->fullsync) in raid10_sync_request()
2978 && conf->have_replacement in raid10_sync_request()
2979 && test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) { in raid10_sync_request()
2984 for (i = 0; i < conf->geo.raid_disks; i++) { in raid10_sync_request()
2986 rcu_dereference(conf->mirrors[i].replacement); in raid10_sync_request()
2988 rdev->recovery_offset = MaxSector; in raid10_sync_request()
2992 conf->fullsync = 0; in raid10_sync_request()
2994 md_bitmap_close_sync(mddev->bitmap); in raid10_sync_request()
3000 if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) in raid10_sync_request()
3003 if (chunks_skipped >= conf->geo.raid_disks) { in raid10_sync_request()
3008 return (max_sector - sector_nr) + sectors_skipped; in raid10_sync_request()
3011 if (max_sector > mddev->resync_max) in raid10_sync_request()
3012 max_sector = mddev->resync_max; /* Don't do IO beyond here */ in raid10_sync_request()
3014 /* make sure whole request will fit in a chunk - if chunks in raid10_sync_request()
3017 if (conf->geo.near_copies < conf->geo.raid_disks && in raid10_sync_request()
3022 * If there is non-resync activity waiting for a turn, then let it in raid10_sync_request()
3025 if (conf->nr_waiting) in raid10_sync_request()
3038 /* First, we decide what to do and set ->bi_end_io in raid10_sync_request()
3043 max_sync = RESYNC_PAGES << (PAGE_SHIFT-9); in raid10_sync_request()
3044 if (!test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) { in raid10_sync_request()
3049 for (i = 0 ; i < conf->geo.raid_disks; i++) { in raid10_sync_request()
3056 struct raid10_info *mirror = &conf->mirrors[i]; in raid10_sync_request()
3060 mrdev = rcu_dereference(mirror->rdev); in raid10_sync_request()
3061 mreplace = rcu_dereference(mirror->replacement); in raid10_sync_request()
3064 !test_bit(Faulty, &mrdev->flags) && in raid10_sync_request()
3065 !test_bit(In_sync, &mrdev->flags)) in raid10_sync_request()
3067 if (mreplace && test_bit(Faulty, &mreplace->flags)) in raid10_sync_request()
3079 if (sect >= mddev->resync_max_sectors) { in raid10_sync_request()
3080 /* last stripe is not complete - don't in raid10_sync_request()
3090 must_sync = md_bitmap_start_sync(mddev->bitmap, sect, in raid10_sync_request()
3096 !conf->fullsync) { in raid10_sync_request()
3100 chunks_skipped = -1; in raid10_sync_request()
3104 atomic_inc(&mrdev->nr_pending); in raid10_sync_request()
3106 atomic_inc(&mreplace->nr_pending); in raid10_sync_request()
3110 r10_bio->state = 0; in raid10_sync_request()
3112 atomic_set(&r10_bio->remaining, 0); in raid10_sync_request()
3114 r10_bio->master_bio = (struct bio*)rb2; in raid10_sync_request()
3116 atomic_inc(&rb2->remaining); in raid10_sync_request()
3117 r10_bio->mddev = mddev; in raid10_sync_request()
3118 set_bit(R10BIO_IsRecover, &r10_bio->state); in raid10_sync_request()
3119 r10_bio->sector = sect; in raid10_sync_request()
3127 for (j = 0; j < conf->geo.raid_disks; j++) { in raid10_sync_request()
3129 conf->mirrors[j].rdev); in raid10_sync_request()
3130 if (rdev == NULL || test_bit(Faulty, &rdev->flags)) { in raid10_sync_request()
3136 must_sync = md_bitmap_start_sync(mddev->bitmap, sect, in raid10_sync_request()
3140 for (j=0; j<conf->copies;j++) { in raid10_sync_request()
3142 int d = r10_bio->devs[j].devnum; in raid10_sync_request()
3145 rcu_dereference(conf->mirrors[d].rdev); in raid10_sync_request()
3149 !test_bit(In_sync, &rdev->flags)) in raid10_sync_request()
3153 sector = r10_bio->devs[j].addr; in raid10_sync_request()
3158 max_sync = first_bad - sector; in raid10_sync_request()
3160 bad_sectors -= (sector in raid10_sync_request()
3161 - first_bad); in raid10_sync_request()
3167 bio = r10_bio->devs[0].bio; in raid10_sync_request()
3168 bio->bi_next = biolist; in raid10_sync_request()
3170 bio->bi_end_io = end_sync_read; in raid10_sync_request()
3172 if (test_bit(FailFast, &rdev->flags)) in raid10_sync_request()
3173 bio->bi_opf |= MD_FAILFAST; in raid10_sync_request()
3174 from_addr = r10_bio->devs[j].addr; in raid10_sync_request()
3175 bio->bi_iter.bi_sector = from_addr + in raid10_sync_request()
3176 rdev->data_offset; in raid10_sync_request()
3177 bio_set_dev(bio, rdev->bdev); in raid10_sync_request()
3178 atomic_inc(&rdev->nr_pending); in raid10_sync_request()
3181 for (k=0; k<conf->copies; k++) in raid10_sync_request()
3182 if (r10_bio->devs[k].devnum == i) in raid10_sync_request()
3184 BUG_ON(k == conf->copies); in raid10_sync_request()
3185 to_addr = r10_bio->devs[k].addr; in raid10_sync_request()
3186 r10_bio->devs[0].devnum = d; in raid10_sync_request()
3187 r10_bio->devs[0].addr = from_addr; in raid10_sync_request()
3188 r10_bio->devs[1].devnum = i; in raid10_sync_request()
3189 r10_bio->devs[1].addr = to_addr; in raid10_sync_request()
3192 bio = r10_bio->devs[1].bio; in raid10_sync_request()
3193 bio->bi_next = biolist; in raid10_sync_request()
3195 bio->bi_end_io = end_sync_write; in raid10_sync_request()
3197 bio->bi_iter.bi_sector = to_addr in raid10_sync_request()
3198 + mrdev->data_offset; in raid10_sync_request()
3199 bio_set_dev(bio, mrdev->bdev); in raid10_sync_request()
3200 atomic_inc(&r10_bio->remaining); in raid10_sync_request()
3202 r10_bio->devs[1].bio->bi_end_io = NULL; in raid10_sync_request()
3205 bio = r10_bio->devs[1].repl_bio; in raid10_sync_request()
3207 bio->bi_end_io = NULL; in raid10_sync_request()
3214 bio->bi_next = biolist; in raid10_sync_request()
3216 bio->bi_end_io = end_sync_write; in raid10_sync_request()
3218 bio->bi_iter.bi_sector = to_addr + in raid10_sync_request()
3219 mreplace->data_offset; in raid10_sync_request()
3220 bio_set_dev(bio, mreplace->bdev); in raid10_sync_request()
3221 atomic_inc(&r10_bio->remaining); in raid10_sync_request()
3225 if (j == conf->copies) { in raid10_sync_request()
3233 for (k = 0; k < conf->copies; k++) in raid10_sync_request()
3234 if (r10_bio->devs[k].devnum == i) in raid10_sync_request()
3237 &mrdev->flags) in raid10_sync_request()
3240 r10_bio->devs[k].addr, in raid10_sync_request()
3246 r10_bio->devs[k].addr, in raid10_sync_request()
3252 &mddev->recovery)) in raid10_sync_request()
3255 mirror->recovery_disabled in raid10_sync_request()
3256 = mddev->recovery_disabled; in raid10_sync_request()
3260 atomic_dec(&rb2->remaining); in raid10_sync_request()
3270 if (r10_bio->devs[0].bio->bi_opf & MD_FAILFAST) { in raid10_sync_request()
3276 for (; j < conf->copies; j++) { in raid10_sync_request()
3277 int d = r10_bio->devs[j].devnum; in raid10_sync_request()
3278 if (conf->mirrors[d].rdev && in raid10_sync_request()
3280 &conf->mirrors[d].rdev->flags)) in raid10_sync_request()
3284 r10_bio->devs[0].bio->bi_opf in raid10_sync_request()
3291 r10_bio = (struct r10bio*) rb2->master_bio; in raid10_sync_request()
3292 rb2->master_bio = NULL; in raid10_sync_request()
3302 * Since curr_resync_completed could probably not update in in raid10_sync_request()
3308 md_bitmap_cond_end_sync(mddev->bitmap, sector_nr, in raid10_sync_request()
3310 (sector_nr + 2 * RESYNC_SECTORS > conf->cluster_sync_high)); in raid10_sync_request()
3312 if (!md_bitmap_start_sync(mddev->bitmap, sector_nr, in raid10_sync_request()
3313 &sync_blocks, mddev->degraded) && in raid10_sync_request()
3314 !conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED, in raid10_sync_request()
3315 &mddev->recovery)) { in raid10_sync_request()
3323 r10_bio->state = 0; in raid10_sync_request()
3325 r10_bio->mddev = mddev; in raid10_sync_request()
3326 atomic_set(&r10_bio->remaining, 0); in raid10_sync_request()
3328 conf->next_resync = sector_nr; in raid10_sync_request()
3330 r10_bio->master_bio = NULL; in raid10_sync_request()
3331 r10_bio->sector = sector_nr; in raid10_sync_request()
3332 set_bit(R10BIO_IsSync, &r10_bio->state); in raid10_sync_request()
3334 r10_bio->sectors = (sector_nr | chunk_mask) - sector_nr + 1; in raid10_sync_request()
3336 for (i = 0; i < conf->copies; i++) { in raid10_sync_request()
3337 int d = r10_bio->devs[i].devnum; in raid10_sync_request()
3342 if (r10_bio->devs[i].repl_bio) in raid10_sync_request()
3343 r10_bio->devs[i].repl_bio->bi_end_io = NULL; in raid10_sync_request()
3345 bio = r10_bio->devs[i].bio; in raid10_sync_request()
3346 bio->bi_status = BLK_STS_IOERR; in raid10_sync_request()
3348 rdev = rcu_dereference(conf->mirrors[d].rdev); in raid10_sync_request()
3349 if (rdev == NULL || test_bit(Faulty, &rdev->flags)) { in raid10_sync_request()
3353 sector = r10_bio->devs[i].addr; in raid10_sync_request()
3357 max_sync = first_bad - sector; in raid10_sync_request()
3359 bad_sectors -= (sector - first_bad); in raid10_sync_request()
3366 atomic_inc(&rdev->nr_pending); in raid10_sync_request()
3367 atomic_inc(&r10_bio->remaining); in raid10_sync_request()
3368 bio->bi_next = biolist; in raid10_sync_request()
3370 bio->bi_end_io = end_sync_read; in raid10_sync_request()
3372 if (test_bit(FailFast, &rdev->flags)) in raid10_sync_request()
3373 bio->bi_opf |= MD_FAILFAST; in raid10_sync_request()
3374 bio->bi_iter.bi_sector = sector + rdev->data_offset; in raid10_sync_request()
3375 bio_set_dev(bio, rdev->bdev); in raid10_sync_request()
3378 rdev = rcu_dereference(conf->mirrors[d].replacement); in raid10_sync_request()
3379 if (rdev == NULL || test_bit(Faulty, &rdev->flags)) { in raid10_sync_request()
3383 atomic_inc(&rdev->nr_pending); in raid10_sync_request()
3386 bio = r10_bio->devs[i].repl_bio; in raid10_sync_request()
3387 bio->bi_status = BLK_STS_IOERR; in raid10_sync_request()
3389 sector = r10_bio->devs[i].addr; in raid10_sync_request()
3390 bio->bi_next = biolist; in raid10_sync_request()
3392 bio->bi_end_io = end_sync_write; in raid10_sync_request()
3394 if (test_bit(FailFast, &rdev->flags)) in raid10_sync_request()
3395 bio->bi_opf |= MD_FAILFAST; in raid10_sync_request()
3396 bio->bi_iter.bi_sector = sector + rdev->data_offset; in raid10_sync_request()
3397 bio_set_dev(bio, rdev->bdev); in raid10_sync_request()
3403 for (i=0; i<conf->copies; i++) { in raid10_sync_request()
3404 int d = r10_bio->devs[i].devnum; in raid10_sync_request()
3405 if (r10_bio->devs[i].bio->bi_end_io) in raid10_sync_request()
3406 rdev_dec_pending(conf->mirrors[d].rdev, in raid10_sync_request()
3408 if (r10_bio->devs[i].repl_bio && in raid10_sync_request()
3409 r10_bio->devs[i].repl_bio->bi_end_io) in raid10_sync_request()
3411 conf->mirrors[d].replacement, in raid10_sync_request()
3427 len = (max_sector - sector_nr) << 9; in raid10_sync_request()
3430 for (bio= biolist ; bio ; bio=bio->bi_next) { in raid10_sync_request()
3442 r10_bio->sectors = nr_sectors; in raid10_sync_request()
3445 test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) { in raid10_sync_request()
3447 if (conf->cluster_sync_high < sector_nr + nr_sectors) { in raid10_sync_request()
3448 conf->cluster_sync_low = mddev->curr_resync_completed; in raid10_sync_request()
3451 md_cluster_ops->resync_info_update(mddev, in raid10_sync_request()
3452 conf->cluster_sync_low, in raid10_sync_request()
3453 conf->cluster_sync_high); in raid10_sync_request()
3460 for (i = 0; i < conf->geo.raid_disks; i++) { in raid10_sync_request()
3468 if (conf->cluster_sync_high < sect_va1 + nr_sectors) { in raid10_sync_request()
3475 mddev->curr_resync_completed, i); in raid10_sync_request()
3477 if (conf->cluster_sync_low == 0 || in raid10_sync_request()
3478 conf->cluster_sync_low > sect_va2) in raid10_sync_request()
3479 conf->cluster_sync_low = sect_va2; in raid10_sync_request()
3484 md_cluster_ops->resync_info_update(mddev, in raid10_sync_request()
3485 conf->cluster_sync_low, in raid10_sync_request()
3486 conf->cluster_sync_high); in raid10_sync_request()
3492 biolist = biolist->bi_next; in raid10_sync_request()
3494 bio->bi_next = NULL; in raid10_sync_request()
3496 r10_bio->sectors = nr_sectors; in raid10_sync_request()
3498 if (bio->bi_end_io == end_sync_read) { in raid10_sync_request()
3500 bio->bi_status = 0; in raid10_sync_request()
3513 /* There is nowhere to write, so all non-sync in raid10_sync_request()
3520 sectors_skipped += (max_sector - sector_nr); in raid10_sync_request()
3530 struct r10conf *conf = mddev->private; in raid10_size()
3533 raid_disks = min(conf->geo.raid_disks, in raid10_size()
3534 conf->prev.raid_disks); in raid10_size()
3536 sectors = conf->dev_sectors; in raid10_size()
3538 size = sectors >> conf->geo.chunk_shift; in raid10_size()
3539 sector_div(size, conf->geo.far_copies); in raid10_size()
3541 sector_div(size, conf->geo.near_copies); in raid10_size()
3543 return size << conf->geo.chunk_shift; in raid10_size()
3548 /* Calculate the number of sectors-per-device that will in calc_sectors()
3549 * actually be used, and set conf->dev_sectors and in calc_sectors()
3550 * conf->stride in calc_sectors()
3553 size = size >> conf->geo.chunk_shift; in calc_sectors()
3554 sector_div(size, conf->geo.far_copies); in calc_sectors()
3555 size = size * conf->geo.raid_disks; in calc_sectors()
3556 sector_div(size, conf->geo.near_copies); in calc_sectors()
3559 size = size * conf->copies; in calc_sectors()
3564 size = DIV_ROUND_UP_SECTOR_T(size, conf->geo.raid_disks); in calc_sectors()
3566 conf->dev_sectors = size << conf->geo.chunk_shift; in calc_sectors()
3568 if (conf->geo.far_offset) in calc_sectors()
3569 conf->geo.stride = 1 << conf->geo.chunk_shift; in calc_sectors()
3571 sector_div(size, conf->geo.far_copies); in calc_sectors()
3572 conf->geo.stride = size << conf->geo.chunk_shift; in calc_sectors()
3579 int nc, fc, fo; in setup_geo() local
3583 layout = mddev->layout; in setup_geo()
3584 chunk = mddev->chunk_sectors; in setup_geo()
3585 disks = mddev->raid_disks - mddev->delta_disks; in setup_geo()
3588 layout = mddev->new_layout; in setup_geo()
3589 chunk = mddev->new_chunk_sectors; in setup_geo()
3590 disks = mddev->raid_disks; in setup_geo()
3593 case geo_start: /* new when starting reshape - raid_disks not in setup_geo()
3595 layout = mddev->new_layout; in setup_geo()
3596 chunk = mddev->new_chunk_sectors; in setup_geo()
3597 disks = mddev->raid_disks + mddev->delta_disks; in setup_geo()
3601 return -1; in setup_geo()
3604 return -2; in setup_geo()
3606 fc = (layout >> 8) & 255; in setup_geo()
3608 geo->raid_disks = disks; in setup_geo()
3609 geo->near_copies = nc; in setup_geo()
3610 geo->far_copies = fc; in setup_geo()
3611 geo->far_offset = fo; in setup_geo()
3614 geo->far_set_size = disks; in setup_geo()
3616 case 1: /* "improved" layout which was buggy. Hopefully no-one is in setup_geo()
3618 geo->far_set_size = disks/fc; in setup_geo()
3619 WARN(geo->far_set_size < fc, in setup_geo()
3620 "This RAID10 layout does not provide data safety - please backup and create new array\n"); in setup_geo()
3623 geo->far_set_size = fc * nc; in setup_geo()
3626 return -1; in setup_geo()
3628 geo->chunk_mask = chunk - 1; in setup_geo()
3629 geo->chunk_shift = ffz(~chunk); in setup_geo()
3630 return nc*fc; in setup_geo()
3638 mempool_exit(&conf->r10bio_pool); in raid10_free_conf()
3639 kfree(conf->mirrors); in raid10_free_conf()
3640 kfree(conf->mirrors_old); in raid10_free_conf()
3641 kfree(conf->mirrors_new); in raid10_free_conf()
3642 safe_put_page(conf->tmppage); in raid10_free_conf()
3643 bioset_exit(&conf->bio_split); in raid10_free_conf()
3650 int err = -EINVAL; in setup_conf()
3656 if (copies == -2) { in setup_conf()
3662 if (copies < 2 || copies > mddev->raid_disks) { in setup_conf()
3664 mdname(mddev), mddev->new_layout); in setup_conf()
3668 err = -ENOMEM; in setup_conf()
3674 conf->mirrors = kcalloc(mddev->raid_disks + max(0, -mddev->delta_disks), in setup_conf()
3677 if (!conf->mirrors) in setup_conf()
3680 conf->tmppage = alloc_page(GFP_KERNEL); in setup_conf()
3681 if (!conf->tmppage) in setup_conf()
3684 conf->geo = geo; in setup_conf()
3685 conf->copies = copies; in setup_conf()
3686 err = mempool_init(&conf->r10bio_pool, NR_RAID_BIOS, r10bio_pool_alloc, in setup_conf()
3691 err = bioset_init(&conf->bio_split, BIO_POOL_SIZE, 0, 0); in setup_conf()
3695 calc_sectors(conf, mddev->dev_sectors); in setup_conf()
3696 if (mddev->reshape_position == MaxSector) { in setup_conf()
3697 conf->prev = conf->geo; in setup_conf()
3698 conf->reshape_progress = MaxSector; in setup_conf()
3700 if (setup_geo(&conf->prev, mddev, geo_old) != conf->copies) { in setup_conf()
3701 err = -EINVAL; in setup_conf()
3704 conf->reshape_progress = mddev->reshape_position; in setup_conf()
3705 if (conf->prev.far_offset) in setup_conf()
3706 conf->prev.stride = 1 << conf->prev.chunk_shift; in setup_conf()
3709 conf->prev.stride = conf->dev_sectors; in setup_conf()
3711 conf->reshape_safe = conf->reshape_progress; in setup_conf()
3712 spin_lock_init(&conf->device_lock); in setup_conf()
3713 INIT_LIST_HEAD(&conf->retry_list); in setup_conf()
3714 INIT_LIST_HEAD(&conf->bio_end_io_list); in setup_conf()
3716 spin_lock_init(&conf->resync_lock); in setup_conf()
3717 init_waitqueue_head(&conf->wait_barrier); in setup_conf()
3718 atomic_set(&conf->nr_pending, 0); in setup_conf()
3720 err = -ENOMEM; in setup_conf()
3721 conf->thread = md_register_thread(raid10d, mddev, "raid10"); in setup_conf()
3722 if (!conf->thread) in setup_conf()
3725 conf->mddev = mddev; in setup_conf()
3735 int raid_disks = conf->geo.raid_disks; in raid10_set_io_opt()
3737 if (!(conf->geo.raid_disks % conf->geo.near_copies)) in raid10_set_io_opt()
3738 raid_disks /= conf->geo.near_copies; in raid10_set_io_opt()
3739 blk_queue_io_opt(conf->mddev->queue, (conf->mddev->chunk_sectors << 9) * in raid10_set_io_opt()
3755 return -ENOMEM; in raid10_run()
3757 if (mddev->private == NULL) { in raid10_run()
3761 mddev->private = conf; in raid10_run()
3763 conf = mddev->private; in raid10_run()
3767 mddev->thread = conf->thread; in raid10_run()
3768 conf->thread = NULL; in raid10_run()
3770 if (mddev_is_clustered(conf->mddev)) { in raid10_run()
3771 int fc, fo; in raid10_run() local
3773 fc = (mddev->layout >> 8) & 255; in raid10_run()
3774 fo = mddev->layout & (1<<16); in raid10_run()
3775 if (fc > 1 || fo > 0) { in raid10_run()
3782 if (mddev->queue) { in raid10_run()
3783 blk_queue_max_discard_sectors(mddev->queue, in raid10_run()
3784 mddev->chunk_sectors); in raid10_run()
3785 blk_queue_max_write_same_sectors(mddev->queue, 0); in raid10_run()
3786 blk_queue_max_write_zeroes_sectors(mddev->queue, 0); in raid10_run()
3787 blk_queue_io_min(mddev->queue, mddev->chunk_sectors << 9); in raid10_run()
3794 disk_idx = rdev->raid_disk; in raid10_run()
3797 if (disk_idx >= conf->geo.raid_disks && in raid10_run()
3798 disk_idx >= conf->prev.raid_disks) in raid10_run()
3800 disk = conf->mirrors + disk_idx; in raid10_run()
3802 if (test_bit(Replacement, &rdev->flags)) { in raid10_run()
3803 if (disk->replacement) in raid10_run()
3805 disk->replacement = rdev; in raid10_run()
3807 if (disk->rdev) in raid10_run()
3809 disk->rdev = rdev; in raid10_run()
3811 diff = (rdev->new_data_offset - rdev->data_offset); in raid10_run()
3812 if (!mddev->reshape_backwards) in raid10_run()
3813 diff = -diff; in raid10_run()
3819 if (mddev->gendisk) in raid10_run()
3820 disk_stack_limits(mddev->gendisk, rdev->bdev, in raid10_run()
3821 rdev->data_offset << 9); in raid10_run()
3823 disk->head_position = 0; in raid10_run()
3825 if (blk_queue_discard(bdev_get_queue(rdev->bdev))) in raid10_run()
3830 if (mddev->queue) { in raid10_run()
3833 mddev->queue); in raid10_run()
3836 mddev->queue); in raid10_run()
3839 if (!enough(conf, -1)) { in raid10_run()
3845 if (conf->reshape_progress != MaxSector) { in raid10_run()
3847 if (conf->geo.far_copies != 1 && in raid10_run()
3848 conf->geo.far_offset == 0) in raid10_run()
3850 if (conf->prev.far_copies != 1 && in raid10_run()
3851 conf->prev.far_offset == 0) in raid10_run()
3855 mddev->degraded = 0; in raid10_run()
3857 i < conf->geo.raid_disks in raid10_run()
3858 || i < conf->prev.raid_disks; in raid10_run()
3861 disk = conf->mirrors + i; in raid10_run()
3863 if (!disk->rdev && disk->replacement) { in raid10_run()
3864 /* The replacement is all we have - use it */ in raid10_run()
3865 disk->rdev = disk->replacement; in raid10_run()
3866 disk->replacement = NULL; in raid10_run()
3867 clear_bit(Replacement, &disk->rdev->flags); in raid10_run()
3870 if (!disk->rdev || in raid10_run()
3871 !test_bit(In_sync, &disk->rdev->flags)) { in raid10_run()
3872 disk->head_position = 0; in raid10_run()
3873 mddev->degraded++; in raid10_run()
3874 if (disk->rdev && in raid10_run()
3875 disk->rdev->saved_raid_disk < 0) in raid10_run()
3876 conf->fullsync = 1; in raid10_run()
3879 if (disk->replacement && in raid10_run()
3880 !test_bit(In_sync, &disk->replacement->flags) && in raid10_run()
3881 disk->replacement->saved_raid_disk < 0) { in raid10_run()
3882 conf->fullsync = 1; in raid10_run()
3885 disk->recovery_disabled = mddev->recovery_disabled - 1; in raid10_run()
3888 if (mddev->recovery_cp != MaxSector) in raid10_run()
3889 pr_notice("md/raid10:%s: not clean -- starting background reconstruction\n", in raid10_run()
3892 mdname(mddev), conf->geo.raid_disks - mddev->degraded, in raid10_run()
3893 conf->geo.raid_disks); in raid10_run()
3897 mddev->dev_sectors = conf->dev_sectors; in raid10_run()
3900 mddev->resync_max_sectors = size; in raid10_run()
3901 set_bit(MD_FAILFAST_SUPPORTED, &mddev->flags); in raid10_run()
3906 if (conf->reshape_progress != MaxSector) { in raid10_run()
3909 before_length = ((1 << conf->prev.chunk_shift) * in raid10_run()
3910 conf->prev.far_copies); in raid10_run()
3911 after_length = ((1 << conf->geo.chunk_shift) * in raid10_run()
3912 conf->geo.far_copies); in raid10_run()
3919 conf->offset_diff = min_offset_diff; in raid10_run()
3921 clear_bit(MD_RECOVERY_SYNC, &mddev->recovery); in raid10_run()
3922 clear_bit(MD_RECOVERY_CHECK, &mddev->recovery); in raid10_run()
3923 set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery); in raid10_run()
3924 set_bit(MD_RECOVERY_RUNNING, &mddev->recovery); in raid10_run()
3925 mddev->sync_thread = md_register_thread(md_do_sync, mddev, in raid10_run()
3927 if (!mddev->sync_thread) in raid10_run()
3934 md_unregister_thread(&mddev->thread); in raid10_run()
3936 mddev->private = NULL; in raid10_run()
3938 return -EIO; in raid10_run()
3948 struct r10conf *conf = mddev->private; in raid10_quiesce()
3970 struct r10conf *conf = mddev->private; in raid10_resize()
3973 if (mddev->reshape_position != MaxSector) in raid10_resize()
3974 return -EBUSY; in raid10_resize()
3976 if (conf->geo.far_copies > 1 && !conf->geo.far_offset) in raid10_resize()
3977 return -EINVAL; in raid10_resize()
3981 if (mddev->external_size && in raid10_resize()
3982 mddev->array_sectors > size) in raid10_resize()
3983 return -EINVAL; in raid10_resize()
3984 if (mddev->bitmap) { in raid10_resize()
3985 int ret = md_bitmap_resize(mddev->bitmap, size, 0, 0); in raid10_resize()
3990 if (sectors > mddev->dev_sectors && in raid10_resize()
3991 mddev->recovery_cp > oldsize) { in raid10_resize()
3992 mddev->recovery_cp = oldsize; in raid10_resize()
3993 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); in raid10_resize()
3996 mddev->dev_sectors = conf->dev_sectors; in raid10_resize()
3997 mddev->resync_max_sectors = size; in raid10_resize()
4006 if (mddev->degraded > 0) { in raid10_takeover_raid0()
4009 return ERR_PTR(-EINVAL); in raid10_takeover_raid0()
4014 mddev->new_level = 10; in raid10_takeover_raid0()
4016 mddev->new_layout = (1<<8) + 2; in raid10_takeover_raid0()
4017 mddev->new_chunk_sectors = mddev->chunk_sectors; in raid10_takeover_raid0()
4018 mddev->delta_disks = mddev->raid_disks; in raid10_takeover_raid0()
4019 mddev->raid_disks *= 2; in raid10_takeover_raid0()
4021 mddev->recovery_cp = MaxSector; in raid10_takeover_raid0()
4022 mddev->dev_sectors = size; in raid10_takeover_raid0()
4027 if (rdev->raid_disk >= 0) { in raid10_takeover_raid0()
4028 rdev->new_raid_disk = rdev->raid_disk * 2; in raid10_takeover_raid0()
4029 rdev->sectors = size; in raid10_takeover_raid0()
4031 conf->barrier = 1; in raid10_takeover_raid0()
4042 * raid0 - providing it has only two drives in raid10_takeover()
4044 if (mddev->level == 0) { in raid10_takeover()
4046 raid0_conf = mddev->private; in raid10_takeover()
4047 if (raid0_conf->nr_strip_zones > 1) { in raid10_takeover()
4050 return ERR_PTR(-EINVAL); in raid10_takeover()
4053 raid0_conf->strip_zone->zone_end, in raid10_takeover()
4054 raid0_conf->strip_zone->nb_dev); in raid10_takeover()
4056 return ERR_PTR(-EINVAL); in raid10_takeover()
4062 * - layout (to ->new_layout) in raid10_check_reshape()
4063 * - chunk size (to ->new_chunk_sectors) in raid10_check_reshape()
4064 * - raid_disks (by delta_disks) in raid10_check_reshape()
4075 struct r10conf *conf = mddev->private; in raid10_check_reshape()
4078 if (conf->geo.far_copies != 1 && !conf->geo.far_offset) in raid10_check_reshape()
4079 return -EINVAL; in raid10_check_reshape()
4081 if (setup_geo(&geo, mddev, geo_start) != conf->copies) in raid10_check_reshape()
4083 return -EINVAL; in raid10_check_reshape()
4086 return -EINVAL; in raid10_check_reshape()
4088 if (mddev->array_sectors & geo.chunk_mask) in raid10_check_reshape()
4090 return -EINVAL; in raid10_check_reshape()
4092 if (!enough(conf, -1)) in raid10_check_reshape()
4093 return -EINVAL; in raid10_check_reshape()
4095 kfree(conf->mirrors_new); in raid10_check_reshape()
4096 conf->mirrors_new = NULL; in raid10_check_reshape()
4097 if (mddev->delta_disks > 0) { in raid10_check_reshape()
4099 conf->mirrors_new = in raid10_check_reshape()
4100 kcalloc(mddev->raid_disks + mddev->delta_disks, in raid10_check_reshape()
4103 if (!conf->mirrors_new) in raid10_check_reshape()
4104 return -ENOMEM; in raid10_check_reshape()
4111 * - start an array
4112 * - remove non-faulty devices
4113 * - add a spare
4114 * - allow a reshape
4119 * of the two sections, and some non-in_sync devices may
4130 for (i = 0; i < conf->prev.raid_disks; i++) { in calc_degraded()
4131 struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev); in calc_degraded()
4132 if (!rdev || test_bit(Faulty, &rdev->flags)) in calc_degraded()
4134 else if (!test_bit(In_sync, &rdev->flags)) in calc_degraded()
4142 if (conf->geo.raid_disks == conf->prev.raid_disks) in calc_degraded()
4146 for (i = 0; i < conf->geo.raid_disks; i++) { in calc_degraded()
4147 struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev); in calc_degraded()
4148 if (!rdev || test_bit(Faulty, &rdev->flags)) in calc_degraded()
4150 else if (!test_bit(In_sync, &rdev->flags)) { in calc_degraded()
4156 if (conf->geo.raid_disks <= conf->prev.raid_disks) in calc_degraded()
4173 * array non-degraded. We also require that the difference in raid10_start_reshape()
4174 * between old and new data_offset - on each device - is in raid10_start_reshape()
4175 * enough that we never risk over-writing. in raid10_start_reshape()
4182 struct r10conf *conf = mddev->private; in raid10_start_reshape()
4187 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) in raid10_start_reshape()
4188 return -EBUSY; in raid10_start_reshape()
4190 if (setup_geo(&new, mddev, geo_start) != conf->copies) in raid10_start_reshape()
4191 return -EINVAL; in raid10_start_reshape()
4193 before_length = ((1 << conf->prev.chunk_shift) * in raid10_start_reshape()
4194 conf->prev.far_copies); in raid10_start_reshape()
4195 after_length = ((1 << conf->geo.chunk_shift) * in raid10_start_reshape()
4196 conf->geo.far_copies); in raid10_start_reshape()
4199 if (!test_bit(In_sync, &rdev->flags) in raid10_start_reshape()
4200 && !test_bit(Faulty, &rdev->flags)) in raid10_start_reshape()
4202 if (rdev->raid_disk >= 0) { in raid10_start_reshape()
4203 long long diff = (rdev->new_data_offset in raid10_start_reshape()
4204 - rdev->data_offset); in raid10_start_reshape()
4205 if (!mddev->reshape_backwards) in raid10_start_reshape()
4206 diff = -diff; in raid10_start_reshape()
4216 return -EINVAL; in raid10_start_reshape()
4218 if (spares < mddev->delta_disks) in raid10_start_reshape()
4219 return -EINVAL; in raid10_start_reshape()
4221 conf->offset_diff = min_offset_diff; in raid10_start_reshape()
4222 spin_lock_irq(&conf->device_lock); in raid10_start_reshape()
4223 if (conf->mirrors_new) { in raid10_start_reshape()
4224 memcpy(conf->mirrors_new, conf->mirrors, in raid10_start_reshape()
4225 sizeof(struct raid10_info)*conf->prev.raid_disks); in raid10_start_reshape()
4227 kfree(conf->mirrors_old); in raid10_start_reshape()
4228 conf->mirrors_old = conf->mirrors; in raid10_start_reshape()
4229 conf->mirrors = conf->mirrors_new; in raid10_start_reshape()
4230 conf->mirrors_new = NULL; in raid10_start_reshape()
4232 setup_geo(&conf->geo, mddev, geo_start); in raid10_start_reshape()
4234 if (mddev->reshape_backwards) { in raid10_start_reshape()
4236 if (size < mddev->array_sectors) { in raid10_start_reshape()
4237 spin_unlock_irq(&conf->device_lock); in raid10_start_reshape()
4240 return -EINVAL; in raid10_start_reshape()
4242 mddev->resync_max_sectors = size; in raid10_start_reshape()
4243 conf->reshape_progress = size; in raid10_start_reshape()
4245 conf->reshape_progress = 0; in raid10_start_reshape()
4246 conf->reshape_safe = conf->reshape_progress; in raid10_start_reshape()
4247 spin_unlock_irq(&conf->device_lock); in raid10_start_reshape()
4249 if (mddev->delta_disks && mddev->bitmap) { in raid10_start_reshape()
4254 newsize = raid10_size(mddev, 0, conf->geo.raid_disks); in raid10_start_reshape()
4257 ret = md_bitmap_resize(mddev->bitmap, newsize, 0, 0); in raid10_start_reshape()
4265 if (rdev->raid_disk > -1 && in raid10_start_reshape()
4266 !test_bit(Faulty, &rdev->flags)) in raid10_start_reshape()
4267 sb = page_address(rdev->sb_page); in raid10_start_reshape()
4275 if ((sb && (le32_to_cpu(sb->feature_map) & in raid10_start_reshape()
4279 ret = md_bitmap_resize(mddev->bitmap, newsize, 0, 0); in raid10_start_reshape()
4283 ret = md_cluster_ops->resize_bitmaps(mddev, newsize, oldsize); in raid10_start_reshape()
4285 md_bitmap_resize(mddev->bitmap, oldsize, 0, 0); in raid10_start_reshape()
4290 if (mddev->delta_disks > 0) { in raid10_start_reshape()
4292 if (rdev->raid_disk < 0 && in raid10_start_reshape()
4293 !test_bit(Faulty, &rdev->flags)) { in raid10_start_reshape()
4295 if (rdev->raid_disk >= in raid10_start_reshape()
4296 conf->prev.raid_disks) in raid10_start_reshape()
4297 set_bit(In_sync, &rdev->flags); in raid10_start_reshape()
4299 rdev->recovery_offset = 0; in raid10_start_reshape()
4304 } else if (rdev->raid_disk >= conf->prev.raid_disks in raid10_start_reshape()
4305 && !test_bit(Faulty, &rdev->flags)) { in raid10_start_reshape()
4307 set_bit(In_sync, &rdev->flags); in raid10_start_reshape()
4311 * ->degraded is measured against the larger of the in raid10_start_reshape()
4314 spin_lock_irq(&conf->device_lock); in raid10_start_reshape()
4315 mddev->degraded = calc_degraded(conf); in raid10_start_reshape()
4316 spin_unlock_irq(&conf->device_lock); in raid10_start_reshape()
4317 mddev->raid_disks = conf->geo.raid_disks; in raid10_start_reshape()
4318 mddev->reshape_position = conf->reshape_progress; in raid10_start_reshape()
4319 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags); in raid10_start_reshape()
4321 clear_bit(MD_RECOVERY_SYNC, &mddev->recovery); in raid10_start_reshape()
4322 clear_bit(MD_RECOVERY_CHECK, &mddev->recovery); in raid10_start_reshape()
4323 clear_bit(MD_RECOVERY_DONE, &mddev->recovery); in raid10_start_reshape()
4324 set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery); in raid10_start_reshape()
4325 set_bit(MD_RECOVERY_RUNNING, &mddev->recovery); in raid10_start_reshape()
4327 mddev->sync_thread = md_register_thread(md_do_sync, mddev, in raid10_start_reshape()
4329 if (!mddev->sync_thread) { in raid10_start_reshape()
4330 ret = -EAGAIN; in raid10_start_reshape()
4333 conf->reshape_checkpoint = jiffies; in raid10_start_reshape()
4334 md_wakeup_thread(mddev->sync_thread); in raid10_start_reshape()
4339 mddev->recovery = 0; in raid10_start_reshape()
4340 spin_lock_irq(&conf->device_lock); in raid10_start_reshape()
4341 conf->geo = conf->prev; in raid10_start_reshape()
4342 mddev->raid_disks = conf->geo.raid_disks; in raid10_start_reshape()
4344 rdev->new_data_offset = rdev->data_offset; in raid10_start_reshape()
4346 conf->reshape_progress = MaxSector; in raid10_start_reshape()
4347 conf->reshape_safe = MaxSector; in raid10_start_reshape()
4348 mddev->reshape_position = MaxSector; in raid10_start_reshape()
4349 spin_unlock_irq(&conf->device_lock); in raid10_start_reshape()
4353 /* Calculate the last device-address that could contain
4354 * any block from the chunk that includes the array-address 's'
4356 * i.e. the address returned will be chunk-aligned and after
4361 s = (s | geo->chunk_mask) + 1; in last_dev_address()
4362 s >>= geo->chunk_shift; in last_dev_address()
4363 s *= geo->near_copies; in last_dev_address()
4364 s = DIV_ROUND_UP_SECTOR_T(s, geo->raid_disks); in last_dev_address()
4365 s *= geo->far_copies; in last_dev_address()
4366 s <<= geo->chunk_shift; in last_dev_address()
4370 /* Calculate the first device-address that could contain
4371 * any block from the chunk that includes the array-address 's'.
4376 s >>= geo->chunk_shift; in first_dev_address()
4377 s *= geo->near_copies; in first_dev_address()
4378 sector_div(s, geo->raid_disks); in first_dev_address()
4379 s *= geo->far_copies; in first_dev_address()
4380 s <<= geo->chunk_shift; in first_dev_address()
4391 * a chunk, but that is not a problem as mddev->reshape_position in reshape_request()
4396 * we need to flush all reshape requests and update the metadata. in reshape_request()
4407 * we must update the metadata to increase the 'safe' position. in reshape_request()
4414 * (conf->offset_diff - always positive) allows a bit of slack, in reshape_request()
4419 * The means one for each copy for write-out and an extra one for in reshape_request()
4420 * read-in. in reshape_request()
4421 * We store the read-in bio in ->master_bio and the others in in reshape_request()
4422 * ->devs[x].bio and ->devs[x].repl_bio. in reshape_request()
4424 struct r10conf *conf = mddev->private; in reshape_request()
4439 if (mddev->reshape_backwards && in reshape_request()
4440 conf->reshape_progress < raid10_size(mddev, 0, 0)) { in reshape_request()
4442 - conf->reshape_progress); in reshape_request()
4443 } else if (!mddev->reshape_backwards && in reshape_request()
4444 conf->reshape_progress > 0) in reshape_request()
4445 sector_nr = conf->reshape_progress; in reshape_request()
4447 mddev->curr_resync_completed = sector_nr; in reshape_request()
4448 sysfs_notify_dirent_safe(mddev->sysfs_completed); in reshape_request()
4455 * as that doesn't work well for ->reshape_backwards. in reshape_request()
4456 * So just use ->reshape_progress. in reshape_request()
4458 if (mddev->reshape_backwards) { in reshape_request()
4462 next = first_dev_address(conf->reshape_progress - 1, in reshape_request()
4463 &conf->geo); in reshape_request()
4468 safe = last_dev_address(conf->reshape_safe - 1, in reshape_request()
4469 &conf->prev); in reshape_request()
4471 if (next + conf->offset_diff < safe) in reshape_request()
4474 last = conf->reshape_progress - 1; in reshape_request()
4475 sector_nr = last & ~(sector_t)(conf->geo.chunk_mask in reshape_request()
4476 & conf->prev.chunk_mask); in reshape_request()
4478 sector_nr = last + 1 - RESYNC_SECTORS; in reshape_request()
4483 next = last_dev_address(conf->reshape_progress, &conf->geo); in reshape_request()
4488 safe = first_dev_address(conf->reshape_safe, &conf->prev); in reshape_request()
4490 /* Need to update metadata if 'next' might be beyond 'safe' in reshape_request()
4493 if (next > safe + conf->offset_diff) in reshape_request()
4496 sector_nr = conf->reshape_progress; in reshape_request()
4497 last = sector_nr | (conf->geo.chunk_mask in reshape_request()
4498 & conf->prev.chunk_mask); in reshape_request()
4501 last = sector_nr + RESYNC_SECTORS - 1; in reshape_request()
4505 time_after(jiffies, conf->reshape_checkpoint + 10*HZ)) { in reshape_request()
4506 /* Need to update reshape_position in metadata */ in reshape_request()
4508 mddev->reshape_position = conf->reshape_progress; in reshape_request()
4509 if (mddev->reshape_backwards) in reshape_request()
4510 mddev->curr_resync_completed = raid10_size(mddev, 0, 0) in reshape_request()
4511 - conf->reshape_progress; in reshape_request()
4513 mddev->curr_resync_completed = conf->reshape_progress; in reshape_request()
4514 conf->reshape_checkpoint = jiffies; in reshape_request()
4515 set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags); in reshape_request()
4516 md_wakeup_thread(mddev->thread); in reshape_request()
4517 wait_event(mddev->sb_wait, mddev->sb_flags == 0 || in reshape_request()
4518 test_bit(MD_RECOVERY_INTR, &mddev->recovery)); in reshape_request()
4519 if (test_bit(MD_RECOVERY_INTR, &mddev->recovery)) { in reshape_request()
4523 conf->reshape_safe = mddev->reshape_position; in reshape_request()
4531 r10_bio->state = 0; in reshape_request()
4533 atomic_set(&r10_bio->remaining, 0); in reshape_request()
4534 r10_bio->mddev = mddev; in reshape_request()
4535 r10_bio->sector = sector_nr; in reshape_request()
4536 set_bit(R10BIO_IsReshape, &r10_bio->state); in reshape_request()
4537 r10_bio->sectors = last - sector_nr + 1; in reshape_request()
4539 BUG_ON(!test_bit(R10BIO_Previous, &r10_bio->state)); in reshape_request()
4546 mempool_free(r10_bio, &conf->r10buf_pool); in reshape_request()
4547 set_bit(MD_RECOVERY_INTR, &mddev->recovery); in reshape_request()
4553 bio_set_dev(read_bio, rdev->bdev); in reshape_request()
4554 read_bio->bi_iter.bi_sector = (r10_bio->devs[r10_bio->read_slot].addr in reshape_request()
4555 + rdev->data_offset); in reshape_request()
4556 read_bio->bi_private = r10_bio; in reshape_request()
4557 read_bio->bi_end_io = end_reshape_read; in reshape_request()
4559 read_bio->bi_flags &= (~0UL << BIO_RESET_BITS); in reshape_request()
4560 read_bio->bi_status = 0; in reshape_request()
4561 read_bio->bi_vcnt = 0; in reshape_request()
4562 read_bio->bi_iter.bi_size = 0; in reshape_request()
4563 r10_bio->master_bio = read_bio; in reshape_request()
4564 r10_bio->read_slot = r10_bio->devs[r10_bio->read_slot].devnum; in reshape_request()
4570 if (mddev_is_clustered(mddev) && conf->cluster_sync_high <= sector_nr) { in reshape_request()
4574 conf->cluster_sync_low = sector_nr; in reshape_request()
4575 conf->cluster_sync_high = sector_nr + CLUSTER_RESYNC_WINDOW_SECTORS; in reshape_request()
4576 sb = page_address(rdev->sb_page); in reshape_request()
4578 sb_reshape_pos = le64_to_cpu(sb->reshape_position); in reshape_request()
4582 * update cluster_sync_low until it has finished reshape. in reshape_request()
4584 if (sb_reshape_pos < conf->cluster_sync_low) in reshape_request()
4585 conf->cluster_sync_low = sb_reshape_pos; in reshape_request()
4588 md_cluster_ops->resync_info_update(mddev, conf->cluster_sync_low, in reshape_request()
4589 conf->cluster_sync_high); in reshape_request()
4593 __raid10_find_phys(&conf->geo, r10_bio); in reshape_request()
4596 read_bio->bi_next = NULL; in reshape_request()
4599 for (s = 0; s < conf->copies*2; s++) { in reshape_request()
4601 int d = r10_bio->devs[s/2].devnum; in reshape_request()
4604 rdev2 = rcu_dereference(conf->mirrors[d].replacement); in reshape_request()
4605 b = r10_bio->devs[s/2].repl_bio; in reshape_request()
4607 rdev2 = rcu_dereference(conf->mirrors[d].rdev); in reshape_request()
4608 b = r10_bio->devs[s/2].bio; in reshape_request()
4610 if (!rdev2 || test_bit(Faulty, &rdev2->flags)) in reshape_request()
4613 bio_set_dev(b, rdev2->bdev); in reshape_request()
4614 b->bi_iter.bi_sector = r10_bio->devs[s/2].addr + in reshape_request()
4615 rdev2->new_data_offset; in reshape_request()
4616 b->bi_end_io = end_reshape_write; in reshape_request()
4618 b->bi_next = blist; in reshape_request()
4625 pages = get_resync_pages(r10_bio->devs[0].bio)->pages; in reshape_request()
4628 int len = (max_sectors - s) << 9; in reshape_request()
4631 for (bio = blist; bio ; bio = bio->bi_next) { in reshape_request()
4642 r10_bio->sectors = nr_sectors; in reshape_request()
4645 md_sync_acct_bio(read_bio, r10_bio->sectors); in reshape_request()
4646 atomic_inc(&r10_bio->remaining); in reshape_request()
4647 read_bio->bi_next = NULL; in reshape_request()
4656 * update reshape_progress in reshape_request()
4658 if (mddev->reshape_backwards) in reshape_request()
4659 conf->reshape_progress -= sectors_done; in reshape_request()
4661 conf->reshape_progress += sectors_done; in reshape_request()
4673 * If we got a read error then we do sync 1-page reads from in reshape_request_write()
4674 * elsewhere until we find the data - or give up. in reshape_request_write()
4676 struct r10conf *conf = mddev->private; in reshape_request_write()
4679 if (!test_bit(R10BIO_Uptodate, &r10_bio->state)) in reshape_request_write()
4682 md_done_sync(mddev, r10_bio->sectors, 0); in reshape_request_write()
4689 atomic_set(&r10_bio->remaining, 1); in reshape_request_write()
4690 for (s = 0; s < conf->copies*2; s++) { in reshape_request_write()
4692 int d = r10_bio->devs[s/2].devnum; in reshape_request_write()
4696 rdev = rcu_dereference(conf->mirrors[d].replacement); in reshape_request_write()
4697 b = r10_bio->devs[s/2].repl_bio; in reshape_request_write()
4699 rdev = rcu_dereference(conf->mirrors[d].rdev); in reshape_request_write()
4700 b = r10_bio->devs[s/2].bio; in reshape_request_write()
4702 if (!rdev || test_bit(Faulty, &rdev->flags)) { in reshape_request_write()
4706 atomic_inc(&rdev->nr_pending); in reshape_request_write()
4708 md_sync_acct_bio(b, r10_bio->sectors); in reshape_request_write()
4709 atomic_inc(&r10_bio->remaining); in reshape_request_write()
4710 b->bi_next = NULL; in reshape_request_write()
4718 if (test_bit(MD_RECOVERY_INTR, &conf->mddev->recovery)) in end_reshape()
4721 spin_lock_irq(&conf->device_lock); in end_reshape()
4722 conf->prev = conf->geo; in end_reshape()
4723 md_finish_reshape(conf->mddev); in end_reshape()
4725 conf->reshape_progress = MaxSector; in end_reshape()
4726 conf->reshape_safe = MaxSector; in end_reshape()
4727 spin_unlock_irq(&conf->device_lock); in end_reshape()
4729 if (conf->mddev->queue) in end_reshape()
4731 conf->fullsync = 0; in end_reshape()
4736 struct r10conf *conf = mddev->private; in raid10_update_reshape_pos()
4739 md_cluster_ops->resync_info_get(mddev, &lo, &hi); in raid10_update_reshape_pos()
4740 if (((mddev->reshape_position <= hi) && (mddev->reshape_position >= lo)) in raid10_update_reshape_pos()
4741 || mddev->reshape_position == MaxSector) in raid10_update_reshape_pos()
4742 conf->reshape_progress = mddev->reshape_position; in raid10_update_reshape_pos()
4751 int sectors = r10_bio->sectors; in handle_reshape_read_error()
4752 struct r10conf *conf = mddev->private; in handle_reshape_read_error()
4758 r10b = kmalloc(struct_size(r10b, devs, conf->copies), GFP_NOIO); in handle_reshape_read_error()
4760 set_bit(MD_RECOVERY_INTR, &mddev->recovery); in handle_reshape_read_error()
4761 return -ENOMEM; in handle_reshape_read_error()
4765 pages = get_resync_pages(r10_bio->devs[0].bio)->pages; in handle_reshape_read_error()
4767 r10b->sector = r10_bio->sector; in handle_reshape_read_error()
4768 __raid10_find_phys(&conf->prev, r10b); in handle_reshape_read_error()
4780 int d = r10b->devs[slot].devnum; in handle_reshape_read_error()
4781 struct md_rdev *rdev = rcu_dereference(conf->mirrors[d].rdev); in handle_reshape_read_error()
4784 test_bit(Faulty, &rdev->flags) || in handle_reshape_read_error()
4785 !test_bit(In_sync, &rdev->flags)) in handle_reshape_read_error()
4788 addr = r10b->devs[slot].addr + idx * PAGE_SIZE; in handle_reshape_read_error()
4789 atomic_inc(&rdev->nr_pending); in handle_reshape_read_error()
4802 if (slot >= conf->copies) in handle_reshape_read_error()
4811 &mddev->recovery); in handle_reshape_read_error()
4813 return -EIO; in handle_reshape_read_error()
4815 sectors -= s; in handle_reshape_read_error()
4825 struct mddev *mddev = r10_bio->mddev; in end_reshape_write()
4826 struct r10conf *conf = mddev->private; in end_reshape_write()
4834 rdev = conf->mirrors[d].replacement; in end_reshape_write()
4837 rdev = conf->mirrors[d].rdev; in end_reshape_write()
4840 if (bio->bi_status) { in end_reshape_write()
4851 if (!atomic_dec_and_test(&r10_bio->remaining)) in end_reshape_request()
4853 md_done_sync(r10_bio->mddev, r10_bio->sectors, 1); in end_reshape_request()
4854 bio_put(r10_bio->master_bio); in end_reshape_request()
4860 struct r10conf *conf = mddev->private; in raid10_finish_reshape()
4862 if (test_bit(MD_RECOVERY_INTR, &mddev->recovery)) in raid10_finish_reshape()
4865 if (mddev->delta_disks > 0) { in raid10_finish_reshape()
4866 if (mddev->recovery_cp > mddev->resync_max_sectors) { in raid10_finish_reshape()
4867 mddev->recovery_cp = mddev->resync_max_sectors; in raid10_finish_reshape()
4868 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); in raid10_finish_reshape()
4870 mddev->resync_max_sectors = mddev->array_sectors; in raid10_finish_reshape()
4874 for (d = conf->geo.raid_disks ; in raid10_finish_reshape()
4875 d < conf->geo.raid_disks - mddev->delta_disks; in raid10_finish_reshape()
4877 struct md_rdev *rdev = rcu_dereference(conf->mirrors[d].rdev); in raid10_finish_reshape()
4879 clear_bit(In_sync, &rdev->flags); in raid10_finish_reshape()
4880 rdev = rcu_dereference(conf->mirrors[d].replacement); in raid10_finish_reshape()
4882 clear_bit(In_sync, &rdev->flags); in raid10_finish_reshape()
4886 mddev->layout = mddev->new_layout; in raid10_finish_reshape()
4887 mddev->chunk_sectors = 1 << conf->geo.chunk_shift; in raid10_finish_reshape()
4888 mddev->reshape_position = MaxSector; in raid10_finish_reshape()
4889 mddev->delta_disks = 0; in raid10_finish_reshape()
4890 mddev->reshape_backwards = 0; in raid10_finish_reshape()
4931 MODULE_ALIAS("md-personality-9"); /* RAID10 */
4932 MODULE_ALIAS("md-raid10");
4933 MODULE_ALIAS("md-level-10");