2 * Copyright (C) STRATO AG 2012. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
18 #include <linux/sched.h>
19 #include <linux/bio.h>
20 #include <linux/slab.h>
21 #include <linux/buffer_head.h>
22 #include <linux/blkdev.h>
23 #include <linux/random.h>
24 #include <linux/iocontext.h>
25 #include <linux/capability.h>
26 #include <linux/kthread.h>
27 #include <linux/math64.h>
28 #include <asm/div64.h>
30 #include "extent_map.h"
32 #include "transaction.h"
33 #include "print-tree.h"
35 #include "async-thread.h"
36 #include "check-integrity.h"
37 #include "rcu-string.h"
38 #include "dev-replace.h"
41 static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
43 static void btrfs_dev_replace_update_device_in_mapping_tree(
44 struct btrfs_fs_info *fs_info,
45 struct btrfs_device *srcdev,
46 struct btrfs_device *tgtdev);
47 static int btrfs_dev_replace_find_srcdev(struct btrfs_root *root, u64 srcdevid,
49 struct btrfs_device **device);
50 static u64 __btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info);
51 static int btrfs_dev_replace_kthread(void *data);
52 static int btrfs_dev_replace_continue_on_mount(struct btrfs_fs_info *fs_info);
55 int btrfs_init_dev_replace(struct btrfs_fs_info *fs_info)
58 struct btrfs_root *dev_root = fs_info->dev_root;
59 struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
60 struct extent_buffer *eb;
63 struct btrfs_path *path = NULL;
65 struct btrfs_dev_replace_item *ptr;
68 path = btrfs_alloc_path();
75 key.type = BTRFS_DEV_REPLACE_KEY;
77 ret = btrfs_search_slot(NULL, dev_root, &key, path, 0, 0);
79 no_valid_dev_replace_entry_found:
81 dev_replace->replace_state =
82 BTRFS_DEV_REPLACE_ITEM_STATE_NEVER_STARTED;
83 dev_replace->cont_reading_from_srcdev_mode =
84 BTRFS_DEV_REPLACE_ITEM_CONT_READING_FROM_SRCDEV_MODE_ALWAYS;
85 dev_replace->replace_state = 0;
86 dev_replace->time_started = 0;
87 dev_replace->time_stopped = 0;
88 atomic64_set(&dev_replace->num_write_errors, 0);
89 atomic64_set(&dev_replace->num_uncorrectable_read_errors, 0);
90 dev_replace->cursor_left = 0;
91 dev_replace->committed_cursor_left = 0;
92 dev_replace->cursor_left_last_write_of_item = 0;
93 dev_replace->cursor_right = 0;
94 dev_replace->srcdev = NULL;
95 dev_replace->tgtdev = NULL;
96 dev_replace->is_valid = 0;
97 dev_replace->item_needs_writeback = 0;
100 slot = path->slots[0];
102 item_size = btrfs_item_size_nr(eb, slot);
103 ptr = btrfs_item_ptr(eb, slot, struct btrfs_dev_replace_item);
105 if (item_size != sizeof(struct btrfs_dev_replace_item)) {
107 "dev_replace entry found has unexpected size, ignore entry");
108 goto no_valid_dev_replace_entry_found;
111 src_devid = btrfs_dev_replace_src_devid(eb, ptr);
112 dev_replace->cont_reading_from_srcdev_mode =
113 btrfs_dev_replace_cont_reading_from_srcdev_mode(eb, ptr);
114 dev_replace->replace_state = btrfs_dev_replace_replace_state(eb, ptr);
115 dev_replace->time_started = btrfs_dev_replace_time_started(eb, ptr);
116 dev_replace->time_stopped =
117 btrfs_dev_replace_time_stopped(eb, ptr);
118 atomic64_set(&dev_replace->num_write_errors,
119 btrfs_dev_replace_num_write_errors(eb, ptr));
120 atomic64_set(&dev_replace->num_uncorrectable_read_errors,
121 btrfs_dev_replace_num_uncorrectable_read_errors(eb, ptr));
122 dev_replace->cursor_left = btrfs_dev_replace_cursor_left(eb, ptr);
123 dev_replace->committed_cursor_left = dev_replace->cursor_left;
124 dev_replace->cursor_left_last_write_of_item = dev_replace->cursor_left;
125 dev_replace->cursor_right = btrfs_dev_replace_cursor_right(eb, ptr);
126 dev_replace->is_valid = 1;
128 dev_replace->item_needs_writeback = 0;
129 switch (dev_replace->replace_state) {
130 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
131 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
132 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
133 dev_replace->srcdev = NULL;
134 dev_replace->tgtdev = NULL;
136 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
137 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
138 dev_replace->srcdev = btrfs_find_device(fs_info, src_devid,
140 dev_replace->tgtdev = btrfs_find_device(fs_info,
141 BTRFS_DEV_REPLACE_DEVID,
144 * allow 'btrfs dev replace_cancel' if src/tgt device is
147 if (!dev_replace->srcdev &&
148 !btrfs_test_opt(dev_root, DEGRADED)) {
151 "cannot mount because device replace operation is ongoing and");
153 "srcdev (devid %llu) is missing, need to run 'btrfs dev scan'?",
156 if (!dev_replace->tgtdev &&
157 !btrfs_test_opt(dev_root, DEGRADED)) {
160 "cannot mount because device replace operation is ongoing and");
162 "tgtdev (devid %llu) is missing, need to run 'btrfs dev scan'?",
163 BTRFS_DEV_REPLACE_DEVID);
165 if (dev_replace->tgtdev) {
166 if (dev_replace->srcdev) {
167 dev_replace->tgtdev->total_bytes =
168 dev_replace->srcdev->total_bytes;
169 dev_replace->tgtdev->disk_total_bytes =
170 dev_replace->srcdev->disk_total_bytes;
171 dev_replace->tgtdev->commit_total_bytes =
172 dev_replace->srcdev->commit_total_bytes;
173 dev_replace->tgtdev->bytes_used =
174 dev_replace->srcdev->bytes_used;
175 dev_replace->tgtdev->commit_bytes_used =
176 dev_replace->srcdev->commit_bytes_used;
178 dev_replace->tgtdev->is_tgtdev_for_dev_replace = 1;
179 btrfs_init_dev_replace_tgtdev_for_resume(fs_info,
180 dev_replace->tgtdev);
187 btrfs_free_path(path);
192 * called from commit_transaction. Writes changed device replace state to
195 int btrfs_run_dev_replace(struct btrfs_trans_handle *trans,
196 struct btrfs_fs_info *fs_info)
199 struct btrfs_root *dev_root = fs_info->dev_root;
200 struct btrfs_path *path;
201 struct btrfs_key key;
202 struct extent_buffer *eb;
203 struct btrfs_dev_replace_item *ptr;
204 struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
206 btrfs_dev_replace_lock(dev_replace);
207 if (!dev_replace->is_valid ||
208 !dev_replace->item_needs_writeback) {
209 btrfs_dev_replace_unlock(dev_replace);
212 btrfs_dev_replace_unlock(dev_replace);
215 key.type = BTRFS_DEV_REPLACE_KEY;
218 path = btrfs_alloc_path();
223 ret = btrfs_search_slot(trans, dev_root, &key, path, -1, 1);
225 btrfs_warn(fs_info, "error %d while searching for dev_replace item!",
231 btrfs_item_size_nr(path->nodes[0], path->slots[0]) < sizeof(*ptr)) {
233 * need to delete old one and insert a new one.
234 * Since no attempt is made to recover any old state, if the
235 * dev_replace state is 'running', the data on the target
237 * It would be possible to recover the state: just make sure
238 * that the beginning of the item is never changed and always
239 * contains all the essential information. Then read this
240 * minimal set of information and use it as a base for the
243 ret = btrfs_del_item(trans, dev_root, path);
245 btrfs_warn(fs_info, "delete too small dev_replace item failed %d!",
253 /* need to insert a new item */
254 btrfs_release_path(path);
255 ret = btrfs_insert_empty_item(trans, dev_root, path,
258 btrfs_warn(fs_info, "insert dev_replace item failed %d!",
265 ptr = btrfs_item_ptr(eb, path->slots[0],
266 struct btrfs_dev_replace_item);
268 btrfs_dev_replace_lock(dev_replace);
269 if (dev_replace->srcdev)
270 btrfs_set_dev_replace_src_devid(eb, ptr,
271 dev_replace->srcdev->devid);
273 btrfs_set_dev_replace_src_devid(eb, ptr, (u64)-1);
274 btrfs_set_dev_replace_cont_reading_from_srcdev_mode(eb, ptr,
275 dev_replace->cont_reading_from_srcdev_mode);
276 btrfs_set_dev_replace_replace_state(eb, ptr,
277 dev_replace->replace_state);
278 btrfs_set_dev_replace_time_started(eb, ptr, dev_replace->time_started);
279 btrfs_set_dev_replace_time_stopped(eb, ptr, dev_replace->time_stopped);
280 btrfs_set_dev_replace_num_write_errors(eb, ptr,
281 atomic64_read(&dev_replace->num_write_errors));
282 btrfs_set_dev_replace_num_uncorrectable_read_errors(eb, ptr,
283 atomic64_read(&dev_replace->num_uncorrectable_read_errors));
284 dev_replace->cursor_left_last_write_of_item =
285 dev_replace->cursor_left;
286 btrfs_set_dev_replace_cursor_left(eb, ptr,
287 dev_replace->cursor_left_last_write_of_item);
288 btrfs_set_dev_replace_cursor_right(eb, ptr,
289 dev_replace->cursor_right);
290 dev_replace->item_needs_writeback = 0;
291 btrfs_dev_replace_unlock(dev_replace);
293 btrfs_mark_buffer_dirty(eb);
296 btrfs_free_path(path);
301 void btrfs_after_dev_replace_commit(struct btrfs_fs_info *fs_info)
303 struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
305 dev_replace->committed_cursor_left =
306 dev_replace->cursor_left_last_write_of_item;
309 int btrfs_dev_replace_start(struct btrfs_root *root,
310 struct btrfs_ioctl_dev_replace_args *args)
312 struct btrfs_trans_handle *trans;
313 struct btrfs_fs_info *fs_info = root->fs_info;
314 struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
316 struct btrfs_device *tgt_device = NULL;
317 struct btrfs_device *src_device = NULL;
319 if (btrfs_fs_incompat(fs_info, RAID56)) {
320 btrfs_warn(fs_info, "dev_replace cannot yet handle RAID5/RAID6");
324 switch (args->start.cont_reading_from_srcdev_mode) {
325 case BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_ALWAYS:
326 case BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_AVOID:
332 if ((args->start.srcdevid == 0 && args->start.srcdev_name[0] == '\0') ||
333 args->start.tgtdev_name[0] == '\0')
337 * Here we commit the transaction to make sure commit_total_bytes
338 * of all the devices are updated.
340 trans = btrfs_attach_transaction(root);
341 if (!IS_ERR(trans)) {
342 ret = btrfs_commit_transaction(trans, root);
345 } else if (PTR_ERR(trans) != -ENOENT) {
346 return PTR_ERR(trans);
349 /* the disk copy procedure reuses the scrub code */
350 mutex_lock(&fs_info->volume_mutex);
351 ret = btrfs_dev_replace_find_srcdev(root, args->start.srcdevid,
352 args->start.srcdev_name,
355 mutex_unlock(&fs_info->volume_mutex);
359 ret = btrfs_init_dev_replace_tgtdev(root, args->start.tgtdev_name,
360 src_device, &tgt_device);
361 mutex_unlock(&fs_info->volume_mutex);
365 btrfs_dev_replace_lock(dev_replace);
366 switch (dev_replace->replace_state) {
367 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
368 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
369 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
371 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
372 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
373 args->result = BTRFS_IOCTL_DEV_REPLACE_RESULT_ALREADY_STARTED;
377 dev_replace->cont_reading_from_srcdev_mode =
378 args->start.cont_reading_from_srcdev_mode;
379 WARN_ON(!src_device);
380 dev_replace->srcdev = src_device;
381 WARN_ON(!tgt_device);
382 dev_replace->tgtdev = tgt_device;
384 printk_in_rcu(KERN_INFO
385 "BTRFS: dev_replace from %s (devid %llu) to %s started\n",
386 src_device->missing ? "<missing disk>" :
387 rcu_str_deref(src_device->name),
389 rcu_str_deref(tgt_device->name));
392 * from now on, the writes to the srcdev are all duplicated to
393 * go to the tgtdev as well (refer to btrfs_map_block()).
395 dev_replace->replace_state = BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED;
396 dev_replace->time_started = get_seconds();
397 dev_replace->cursor_left = 0;
398 dev_replace->committed_cursor_left = 0;
399 dev_replace->cursor_left_last_write_of_item = 0;
400 dev_replace->cursor_right = 0;
401 dev_replace->is_valid = 1;
402 dev_replace->item_needs_writeback = 1;
403 args->result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR;
404 btrfs_dev_replace_unlock(dev_replace);
406 btrfs_wait_ordered_roots(root->fs_info, -1);
408 /* force writing the updated state information to disk */
409 trans = btrfs_start_transaction(root, 0);
411 ret = PTR_ERR(trans);
412 btrfs_dev_replace_lock(dev_replace);
416 ret = btrfs_commit_transaction(trans, root);
419 /* the disk copy procedure reuses the scrub code */
420 ret = btrfs_scrub_dev(fs_info, src_device->devid, 0,
421 btrfs_device_get_total_bytes(src_device),
422 &dev_replace->scrub_progress, 0, 1);
424 ret = btrfs_dev_replace_finishing(root->fs_info, ret);
430 dev_replace->srcdev = NULL;
431 dev_replace->tgtdev = NULL;
432 btrfs_dev_replace_unlock(dev_replace);
433 btrfs_destroy_dev_replace_tgtdev(fs_info, tgt_device);
438 * blocked until all flighting bios are finished.
440 static void btrfs_rm_dev_replace_blocked(struct btrfs_fs_info *fs_info)
445 set_bit(BTRFS_FS_STATE_DEV_REPLACING, &fs_info->fs_state);
447 prepare_to_wait(&fs_info->replace_wait, &wait,
448 TASK_UNINTERRUPTIBLE);
449 writers = percpu_counter_sum(&fs_info->bio_counter);
452 finish_wait(&fs_info->replace_wait, &wait);
457 * we have removed target device, it is safe to allow new bios request.
459 static void btrfs_rm_dev_replace_unblocked(struct btrfs_fs_info *fs_info)
461 clear_bit(BTRFS_FS_STATE_DEV_REPLACING, &fs_info->fs_state);
462 if (waitqueue_active(&fs_info->replace_wait))
463 wake_up(&fs_info->replace_wait);
466 static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
469 struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
470 struct btrfs_device *tgt_device;
471 struct btrfs_device *src_device;
472 struct btrfs_root *root = fs_info->tree_root;
473 u8 uuid_tmp[BTRFS_UUID_SIZE];
474 struct btrfs_trans_handle *trans;
477 /* don't allow cancel or unmount to disturb the finishing procedure */
478 mutex_lock(&dev_replace->lock_finishing_cancel_unmount);
480 btrfs_dev_replace_lock(dev_replace);
481 /* was the operation canceled, or is it finished? */
482 if (dev_replace->replace_state !=
483 BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED) {
484 btrfs_dev_replace_unlock(dev_replace);
485 mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
489 tgt_device = dev_replace->tgtdev;
490 src_device = dev_replace->srcdev;
491 btrfs_dev_replace_unlock(dev_replace);
494 * flush all outstanding I/O and inode extent mappings before the
495 * copy operation is declared as being finished
497 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
499 mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
502 btrfs_wait_ordered_roots(root->fs_info, -1);
504 trans = btrfs_start_transaction(root, 0);
506 mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
507 return PTR_ERR(trans);
509 ret = btrfs_commit_transaction(trans, root);
512 mutex_lock(&uuid_mutex);
513 /* keep away write_all_supers() during the finishing procedure */
514 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
515 mutex_lock(&root->fs_info->chunk_mutex);
516 btrfs_dev_replace_lock(dev_replace);
517 dev_replace->replace_state =
518 scrub_ret ? BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
519 : BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED;
520 dev_replace->tgtdev = NULL;
521 dev_replace->srcdev = NULL;
522 dev_replace->time_stopped = get_seconds();
523 dev_replace->item_needs_writeback = 1;
525 /* replace old device with new one in mapping tree */
527 btrfs_dev_replace_update_device_in_mapping_tree(fs_info,
531 printk_in_rcu(KERN_ERR
532 "BTRFS: btrfs_scrub_dev(%s, %llu, %s) failed %d\n",
533 src_device->missing ? "<missing disk>" :
534 rcu_str_deref(src_device->name),
536 rcu_str_deref(tgt_device->name), scrub_ret);
537 btrfs_dev_replace_unlock(dev_replace);
538 mutex_unlock(&root->fs_info->chunk_mutex);
539 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
540 mutex_unlock(&uuid_mutex);
542 btrfs_destroy_dev_replace_tgtdev(fs_info, tgt_device);
543 mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
548 printk_in_rcu(KERN_INFO
549 "BTRFS: dev_replace from %s (devid %llu) to %s finished\n",
550 src_device->missing ? "<missing disk>" :
551 rcu_str_deref(src_device->name),
553 rcu_str_deref(tgt_device->name));
554 tgt_device->is_tgtdev_for_dev_replace = 0;
555 tgt_device->devid = src_device->devid;
556 src_device->devid = BTRFS_DEV_REPLACE_DEVID;
557 memcpy(uuid_tmp, tgt_device->uuid, sizeof(uuid_tmp));
558 memcpy(tgt_device->uuid, src_device->uuid, sizeof(tgt_device->uuid));
559 memcpy(src_device->uuid, uuid_tmp, sizeof(src_device->uuid));
560 btrfs_device_set_total_bytes(tgt_device, src_device->total_bytes);
561 btrfs_device_set_disk_total_bytes(tgt_device,
562 src_device->disk_total_bytes);
563 btrfs_device_set_bytes_used(tgt_device, src_device->bytes_used);
564 ASSERT(list_empty(&src_device->resized_list));
565 tgt_device->commit_total_bytes = src_device->commit_total_bytes;
566 tgt_device->commit_bytes_used = src_device->bytes_used;
567 if (fs_info->sb->s_bdev == src_device->bdev)
568 fs_info->sb->s_bdev = tgt_device->bdev;
569 if (fs_info->fs_devices->latest_bdev == src_device->bdev)
570 fs_info->fs_devices->latest_bdev = tgt_device->bdev;
571 list_add(&tgt_device->dev_alloc_list, &fs_info->fs_devices->alloc_list);
572 fs_info->fs_devices->rw_devices++;
574 /* replace the sysfs entry */
575 btrfs_kobj_rm_device(fs_info, src_device);
576 btrfs_kobj_add_device(fs_info, tgt_device);
578 btrfs_dev_replace_unlock(dev_replace);
580 btrfs_rm_dev_replace_blocked(fs_info);
582 btrfs_rm_dev_replace_srcdev(fs_info, src_device);
584 btrfs_rm_dev_replace_unblocked(fs_info);
587 * this is again a consistent state where no dev_replace procedure
588 * is running, the target device is part of the filesystem, the
589 * source device is not part of the filesystem anymore and its 1st
590 * superblock is scratched out so that it is no longer marked to
591 * belong to this filesystem.
593 mutex_unlock(&root->fs_info->chunk_mutex);
594 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
595 mutex_unlock(&uuid_mutex);
597 /* write back the superblocks */
598 trans = btrfs_start_transaction(root, 0);
600 btrfs_commit_transaction(trans, root);
602 mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
607 static void btrfs_dev_replace_update_device_in_mapping_tree(
608 struct btrfs_fs_info *fs_info,
609 struct btrfs_device *srcdev,
610 struct btrfs_device *tgtdev)
612 struct extent_map_tree *em_tree = &fs_info->mapping_tree.map_tree;
613 struct extent_map *em;
614 struct map_lookup *map;
618 write_lock(&em_tree->lock);
620 em = lookup_extent_mapping(em_tree, start, (u64)-1);
623 map = (struct map_lookup *)em->bdev;
624 for (i = 0; i < map->num_stripes; i++)
625 if (srcdev == map->stripes[i].dev)
626 map->stripes[i].dev = tgtdev;
627 start = em->start + em->len;
630 write_unlock(&em_tree->lock);
633 static int btrfs_dev_replace_find_srcdev(struct btrfs_root *root, u64 srcdevid,
635 struct btrfs_device **device)
641 *device = btrfs_find_device(root->fs_info, srcdevid, NULL,
646 ret = btrfs_find_device_missing_or_by_path(root, srcdev_name,
652 void btrfs_dev_replace_status(struct btrfs_fs_info *fs_info,
653 struct btrfs_ioctl_dev_replace_args *args)
655 struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
656 struct btrfs_device *srcdev;
658 btrfs_dev_replace_lock(dev_replace);
659 /* even if !dev_replace_is_valid, the values are good enough for
660 * the replace_status ioctl */
661 args->result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR;
662 args->status.replace_state = dev_replace->replace_state;
663 args->status.time_started = dev_replace->time_started;
664 args->status.time_stopped = dev_replace->time_stopped;
665 args->status.num_write_errors =
666 atomic64_read(&dev_replace->num_write_errors);
667 args->status.num_uncorrectable_read_errors =
668 atomic64_read(&dev_replace->num_uncorrectable_read_errors);
669 switch (dev_replace->replace_state) {
670 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
671 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
672 args->status.progress_1000 = 0;
674 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
675 args->status.progress_1000 = 1000;
677 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
678 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
679 srcdev = dev_replace->srcdev;
680 args->status.progress_1000 = div64_u64(dev_replace->cursor_left,
681 div64_u64(btrfs_device_get_total_bytes(srcdev), 1000));
684 btrfs_dev_replace_unlock(dev_replace);
687 int btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info,
688 struct btrfs_ioctl_dev_replace_args *args)
690 args->result = __btrfs_dev_replace_cancel(fs_info);
694 static u64 __btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info)
696 struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
697 struct btrfs_device *tgt_device = NULL;
698 struct btrfs_trans_handle *trans;
699 struct btrfs_root *root = fs_info->tree_root;
703 if (fs_info->sb->s_flags & MS_RDONLY)
706 mutex_lock(&dev_replace->lock_finishing_cancel_unmount);
707 btrfs_dev_replace_lock(dev_replace);
708 switch (dev_replace->replace_state) {
709 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
710 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
711 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
712 result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NOT_STARTED;
713 btrfs_dev_replace_unlock(dev_replace);
715 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
716 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
717 result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR;
718 tgt_device = dev_replace->tgtdev;
719 dev_replace->tgtdev = NULL;
720 dev_replace->srcdev = NULL;
723 dev_replace->replace_state = BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED;
724 dev_replace->time_stopped = get_seconds();
725 dev_replace->item_needs_writeback = 1;
726 btrfs_dev_replace_unlock(dev_replace);
727 btrfs_scrub_cancel(fs_info);
729 trans = btrfs_start_transaction(root, 0);
731 mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
732 return PTR_ERR(trans);
734 ret = btrfs_commit_transaction(trans, root);
737 btrfs_destroy_dev_replace_tgtdev(fs_info, tgt_device);
740 mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
744 void btrfs_dev_replace_suspend_for_unmount(struct btrfs_fs_info *fs_info)
746 struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
748 mutex_lock(&dev_replace->lock_finishing_cancel_unmount);
749 btrfs_dev_replace_lock(dev_replace);
750 switch (dev_replace->replace_state) {
751 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
752 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
753 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
754 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
756 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
757 dev_replace->replace_state =
758 BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED;
759 dev_replace->time_stopped = get_seconds();
760 dev_replace->item_needs_writeback = 1;
761 btrfs_info(fs_info, "suspending dev_replace for unmount");
765 btrfs_dev_replace_unlock(dev_replace);
766 mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
769 /* resume dev_replace procedure that was interrupted by unmount */
770 int btrfs_resume_dev_replace_async(struct btrfs_fs_info *fs_info)
772 struct task_struct *task;
773 struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
775 btrfs_dev_replace_lock(dev_replace);
776 switch (dev_replace->replace_state) {
777 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
778 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
779 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
780 btrfs_dev_replace_unlock(dev_replace);
782 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
784 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
785 dev_replace->replace_state =
786 BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED;
789 if (!dev_replace->tgtdev || !dev_replace->tgtdev->bdev) {
790 btrfs_info(fs_info, "cannot continue dev_replace, tgtdev is missing");
792 "you may cancel the operation after 'mount -o degraded'");
793 btrfs_dev_replace_unlock(dev_replace);
796 btrfs_dev_replace_unlock(dev_replace);
799 &fs_info->mutually_exclusive_operation_running, 1));
800 task = kthread_run(btrfs_dev_replace_kthread, fs_info, "btrfs-devrepl");
801 return PTR_ERR_OR_ZERO(task);
804 static int btrfs_dev_replace_kthread(void *data)
806 struct btrfs_fs_info *fs_info = data;
807 struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
808 struct btrfs_ioctl_dev_replace_args *status_args;
811 status_args = kzalloc(sizeof(*status_args), GFP_NOFS);
813 btrfs_dev_replace_status(fs_info, status_args);
814 progress = status_args->status.progress_1000;
816 do_div(progress, 10);
817 printk_in_rcu(KERN_INFO
818 "BTRFS: continuing dev_replace from %s (devid %llu) to %s @%u%%\n",
819 dev_replace->srcdev->missing ? "<missing disk>" :
820 rcu_str_deref(dev_replace->srcdev->name),
821 dev_replace->srcdev->devid,
822 dev_replace->tgtdev ?
823 rcu_str_deref(dev_replace->tgtdev->name) :
824 "<missing target disk>",
825 (unsigned int)progress);
827 btrfs_dev_replace_continue_on_mount(fs_info);
828 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
833 static int btrfs_dev_replace_continue_on_mount(struct btrfs_fs_info *fs_info)
835 struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
838 ret = btrfs_scrub_dev(fs_info, dev_replace->srcdev->devid,
839 dev_replace->committed_cursor_left,
840 btrfs_device_get_total_bytes(dev_replace->srcdev),
841 &dev_replace->scrub_progress, 0, 1);
842 ret = btrfs_dev_replace_finishing(fs_info, ret);
847 int btrfs_dev_replace_is_ongoing(struct btrfs_dev_replace *dev_replace)
849 if (!dev_replace->is_valid)
852 switch (dev_replace->replace_state) {
853 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
854 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
855 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
857 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
858 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
860 * return true even if tgtdev is missing (this is
861 * something that can happen if the dev_replace
862 * procedure is suspended by an umount and then
863 * the tgtdev is missing (or "btrfs dev scan") was
864 * not called and the the filesystem is remounted
865 * in degraded state. This does not stop the
866 * dev_replace procedure. It needs to be canceled
867 * manually if the cancelation is wanted.
874 void btrfs_dev_replace_lock(struct btrfs_dev_replace *dev_replace)
876 /* the beginning is just an optimization for the typical case */
877 if (atomic_read(&dev_replace->nesting_level) == 0) {
879 /* this is not a nested case where the same thread
880 * is trying to acqurire the same lock twice */
881 mutex_lock(&dev_replace->lock);
882 mutex_lock(&dev_replace->lock_management_lock);
883 dev_replace->lock_owner = current->pid;
884 atomic_inc(&dev_replace->nesting_level);
885 mutex_unlock(&dev_replace->lock_management_lock);
889 mutex_lock(&dev_replace->lock_management_lock);
890 if (atomic_read(&dev_replace->nesting_level) > 0 &&
891 dev_replace->lock_owner == current->pid) {
892 WARN_ON(!mutex_is_locked(&dev_replace->lock));
893 atomic_inc(&dev_replace->nesting_level);
894 mutex_unlock(&dev_replace->lock_management_lock);
898 mutex_unlock(&dev_replace->lock_management_lock);
902 void btrfs_dev_replace_unlock(struct btrfs_dev_replace *dev_replace)
904 WARN_ON(!mutex_is_locked(&dev_replace->lock));
905 mutex_lock(&dev_replace->lock_management_lock);
906 WARN_ON(atomic_read(&dev_replace->nesting_level) < 1);
907 WARN_ON(dev_replace->lock_owner != current->pid);
908 atomic_dec(&dev_replace->nesting_level);
909 if (atomic_read(&dev_replace->nesting_level) == 0) {
910 dev_replace->lock_owner = 0;
911 mutex_unlock(&dev_replace->lock_management_lock);
912 mutex_unlock(&dev_replace->lock);
914 mutex_unlock(&dev_replace->lock_management_lock);
918 void btrfs_bio_counter_inc_noblocked(struct btrfs_fs_info *fs_info)
920 percpu_counter_inc(&fs_info->bio_counter);
923 void btrfs_bio_counter_dec(struct btrfs_fs_info *fs_info)
925 percpu_counter_dec(&fs_info->bio_counter);
927 if (waitqueue_active(&fs_info->replace_wait))
928 wake_up(&fs_info->replace_wait);
931 void btrfs_bio_counter_inc_blocked(struct btrfs_fs_info *fs_info)
935 percpu_counter_inc(&fs_info->bio_counter);
936 if (test_bit(BTRFS_FS_STATE_DEV_REPLACING, &fs_info->fs_state)) {
937 btrfs_bio_counter_dec(fs_info);
938 wait_event(fs_info->replace_wait,
939 !test_bit(BTRFS_FS_STATE_DEV_REPLACING,
940 &fs_info->fs_state));