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 switch (args
->start
.cont_reading_from_srcdev_mode
) {
320 case BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_ALWAYS
:
321 case BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_AVOID
:
327 if ((args
->start
.srcdevid
== 0 && args
->start
.srcdev_name
[0] == '\0') ||
328 args
->start
.tgtdev_name
[0] == '\0')
332 * Here we commit the transaction to make sure commit_total_bytes
333 * of all the devices are updated.
335 trans
= btrfs_attach_transaction(root
);
336 if (!IS_ERR(trans
)) {
337 ret
= btrfs_commit_transaction(trans
, root
);
340 } else if (PTR_ERR(trans
) != -ENOENT
) {
341 return PTR_ERR(trans
);
344 /* the disk copy procedure reuses the scrub code */
345 mutex_lock(&fs_info
->volume_mutex
);
346 ret
= btrfs_dev_replace_find_srcdev(root
, args
->start
.srcdevid
,
347 args
->start
.srcdev_name
,
350 mutex_unlock(&fs_info
->volume_mutex
);
354 ret
= btrfs_init_dev_replace_tgtdev(root
, args
->start
.tgtdev_name
,
355 src_device
, &tgt_device
);
356 mutex_unlock(&fs_info
->volume_mutex
);
360 btrfs_dev_replace_lock(dev_replace
);
361 switch (dev_replace
->replace_state
) {
362 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
363 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
364 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
366 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
367 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
368 args
->result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_ALREADY_STARTED
;
372 dev_replace
->cont_reading_from_srcdev_mode
=
373 args
->start
.cont_reading_from_srcdev_mode
;
374 WARN_ON(!src_device
);
375 dev_replace
->srcdev
= src_device
;
376 WARN_ON(!tgt_device
);
377 dev_replace
->tgtdev
= tgt_device
;
379 printk_in_rcu(KERN_INFO
380 "BTRFS: dev_replace from %s (devid %llu) to %s started\n",
381 src_device
->missing
? "<missing disk>" :
382 rcu_str_deref(src_device
->name
),
384 rcu_str_deref(tgt_device
->name
));
387 * from now on, the writes to the srcdev are all duplicated to
388 * go to the tgtdev as well (refer to btrfs_map_block()).
390 dev_replace
->replace_state
= BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
;
391 dev_replace
->time_started
= get_seconds();
392 dev_replace
->cursor_left
= 0;
393 dev_replace
->committed_cursor_left
= 0;
394 dev_replace
->cursor_left_last_write_of_item
= 0;
395 dev_replace
->cursor_right
= 0;
396 dev_replace
->is_valid
= 1;
397 dev_replace
->item_needs_writeback
= 1;
398 args
->result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR
;
399 btrfs_dev_replace_unlock(dev_replace
);
401 btrfs_wait_ordered_roots(root
->fs_info
, -1);
403 /* force writing the updated state information to disk */
404 trans
= btrfs_start_transaction(root
, 0);
406 ret
= PTR_ERR(trans
);
407 btrfs_dev_replace_lock(dev_replace
);
411 ret
= btrfs_commit_transaction(trans
, root
);
414 /* the disk copy procedure reuses the scrub code */
415 ret
= btrfs_scrub_dev(fs_info
, src_device
->devid
, 0,
416 btrfs_device_get_total_bytes(src_device
),
417 &dev_replace
->scrub_progress
, 0, 1);
419 ret
= btrfs_dev_replace_finishing(root
->fs_info
, ret
);
420 /* don't warn if EINPROGRESS, someone else might be running scrub */
421 if (ret
== -EINPROGRESS
) {
422 args
->result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_SCRUB_INPROGRESS
;
431 dev_replace
->srcdev
= NULL
;
432 dev_replace
->tgtdev
= NULL
;
433 btrfs_dev_replace_unlock(dev_replace
);
434 btrfs_destroy_dev_replace_tgtdev(fs_info
, tgt_device
);
439 * blocked until all flighting bios are finished.
441 static void btrfs_rm_dev_replace_blocked(struct btrfs_fs_info
*fs_info
)
446 set_bit(BTRFS_FS_STATE_DEV_REPLACING
, &fs_info
->fs_state
);
448 prepare_to_wait(&fs_info
->replace_wait
, &wait
,
449 TASK_UNINTERRUPTIBLE
);
450 writers
= percpu_counter_sum(&fs_info
->bio_counter
);
453 finish_wait(&fs_info
->replace_wait
, &wait
);
458 * we have removed target device, it is safe to allow new bios request.
460 static void btrfs_rm_dev_replace_unblocked(struct btrfs_fs_info
*fs_info
)
462 clear_bit(BTRFS_FS_STATE_DEV_REPLACING
, &fs_info
->fs_state
);
463 if (waitqueue_active(&fs_info
->replace_wait
))
464 wake_up(&fs_info
->replace_wait
);
467 static int btrfs_dev_replace_finishing(struct btrfs_fs_info
*fs_info
,
470 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
471 struct btrfs_device
*tgt_device
;
472 struct btrfs_device
*src_device
;
473 struct btrfs_root
*root
= fs_info
->tree_root
;
474 u8 uuid_tmp
[BTRFS_UUID_SIZE
];
475 struct btrfs_trans_handle
*trans
;
478 /* don't allow cancel or unmount to disturb the finishing procedure */
479 mutex_lock(&dev_replace
->lock_finishing_cancel_unmount
);
481 btrfs_dev_replace_lock(dev_replace
);
482 /* was the operation canceled, or is it finished? */
483 if (dev_replace
->replace_state
!=
484 BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
) {
485 btrfs_dev_replace_unlock(dev_replace
);
486 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
490 tgt_device
= dev_replace
->tgtdev
;
491 src_device
= dev_replace
->srcdev
;
492 btrfs_dev_replace_unlock(dev_replace
);
495 * flush all outstanding I/O and inode extent mappings before the
496 * copy operation is declared as being finished
498 ret
= btrfs_start_delalloc_roots(root
->fs_info
, 0, -1);
500 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
503 btrfs_wait_ordered_roots(root
->fs_info
, -1);
505 trans
= btrfs_start_transaction(root
, 0);
507 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
508 return PTR_ERR(trans
);
510 ret
= btrfs_commit_transaction(trans
, root
);
513 mutex_lock(&uuid_mutex
);
514 /* keep away write_all_supers() during the finishing procedure */
515 mutex_lock(&root
->fs_info
->fs_devices
->device_list_mutex
);
516 mutex_lock(&root
->fs_info
->chunk_mutex
);
517 btrfs_dev_replace_lock(dev_replace
);
518 dev_replace
->replace_state
=
519 scrub_ret
? BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
520 : BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
;
521 dev_replace
->tgtdev
= NULL
;
522 dev_replace
->srcdev
= NULL
;
523 dev_replace
->time_stopped
= get_seconds();
524 dev_replace
->item_needs_writeback
= 1;
526 /* replace old device with new one in mapping tree */
528 btrfs_dev_replace_update_device_in_mapping_tree(fs_info
,
532 printk_in_rcu(KERN_ERR
533 "BTRFS: btrfs_scrub_dev(%s, %llu, %s) failed %d\n",
534 src_device
->missing
? "<missing disk>" :
535 rcu_str_deref(src_device
->name
),
537 rcu_str_deref(tgt_device
->name
), scrub_ret
);
538 btrfs_dev_replace_unlock(dev_replace
);
539 mutex_unlock(&root
->fs_info
->chunk_mutex
);
540 mutex_unlock(&root
->fs_info
->fs_devices
->device_list_mutex
);
541 mutex_unlock(&uuid_mutex
);
543 btrfs_destroy_dev_replace_tgtdev(fs_info
, tgt_device
);
544 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
549 printk_in_rcu(KERN_INFO
550 "BTRFS: dev_replace from %s (devid %llu) to %s finished\n",
551 src_device
->missing
? "<missing disk>" :
552 rcu_str_deref(src_device
->name
),
554 rcu_str_deref(tgt_device
->name
));
555 tgt_device
->is_tgtdev_for_dev_replace
= 0;
556 tgt_device
->devid
= src_device
->devid
;
557 src_device
->devid
= BTRFS_DEV_REPLACE_DEVID
;
558 memcpy(uuid_tmp
, tgt_device
->uuid
, sizeof(uuid_tmp
));
559 memcpy(tgt_device
->uuid
, src_device
->uuid
, sizeof(tgt_device
->uuid
));
560 memcpy(src_device
->uuid
, uuid_tmp
, sizeof(src_device
->uuid
));
561 btrfs_device_set_total_bytes(tgt_device
, src_device
->total_bytes
);
562 btrfs_device_set_disk_total_bytes(tgt_device
,
563 src_device
->disk_total_bytes
);
564 btrfs_device_set_bytes_used(tgt_device
, src_device
->bytes_used
);
565 ASSERT(list_empty(&src_device
->resized_list
));
566 tgt_device
->commit_total_bytes
= src_device
->commit_total_bytes
;
567 tgt_device
->commit_bytes_used
= src_device
->bytes_used
;
568 if (fs_info
->sb
->s_bdev
== src_device
->bdev
)
569 fs_info
->sb
->s_bdev
= tgt_device
->bdev
;
570 if (fs_info
->fs_devices
->latest_bdev
== src_device
->bdev
)
571 fs_info
->fs_devices
->latest_bdev
= tgt_device
->bdev
;
572 list_add(&tgt_device
->dev_alloc_list
, &fs_info
->fs_devices
->alloc_list
);
573 fs_info
->fs_devices
->rw_devices
++;
575 btrfs_dev_replace_unlock(dev_replace
);
577 btrfs_rm_dev_replace_blocked(fs_info
);
579 btrfs_rm_dev_replace_remove_srcdev(fs_info
, src_device
);
581 btrfs_rm_dev_replace_unblocked(fs_info
);
584 * this is again a consistent state where no dev_replace procedure
585 * is running, the target device is part of the filesystem, the
586 * source device is not part of the filesystem anymore and its 1st
587 * superblock is scratched out so that it is no longer marked to
588 * belong to this filesystem.
590 mutex_unlock(&root
->fs_info
->chunk_mutex
);
591 mutex_unlock(&root
->fs_info
->fs_devices
->device_list_mutex
);
592 mutex_unlock(&uuid_mutex
);
594 /* replace the sysfs entry */
595 btrfs_kobj_rm_device(fs_info
, src_device
);
596 btrfs_kobj_add_device(fs_info
, tgt_device
);
597 btrfs_rm_dev_replace_free_srcdev(fs_info
, src_device
);
599 /* write back the superblocks */
600 trans
= btrfs_start_transaction(root
, 0);
602 btrfs_commit_transaction(trans
, root
);
604 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
609 static void btrfs_dev_replace_update_device_in_mapping_tree(
610 struct btrfs_fs_info
*fs_info
,
611 struct btrfs_device
*srcdev
,
612 struct btrfs_device
*tgtdev
)
614 struct extent_map_tree
*em_tree
= &fs_info
->mapping_tree
.map_tree
;
615 struct extent_map
*em
;
616 struct map_lookup
*map
;
620 write_lock(&em_tree
->lock
);
622 em
= lookup_extent_mapping(em_tree
, start
, (u64
)-1);
625 map
= (struct map_lookup
*)em
->bdev
;
626 for (i
= 0; i
< map
->num_stripes
; i
++)
627 if (srcdev
== map
->stripes
[i
].dev
)
628 map
->stripes
[i
].dev
= tgtdev
;
629 start
= em
->start
+ em
->len
;
632 write_unlock(&em_tree
->lock
);
635 static int btrfs_dev_replace_find_srcdev(struct btrfs_root
*root
, u64 srcdevid
,
637 struct btrfs_device
**device
)
643 *device
= btrfs_find_device(root
->fs_info
, srcdevid
, NULL
,
648 ret
= btrfs_find_device_missing_or_by_path(root
, srcdev_name
,
654 void btrfs_dev_replace_status(struct btrfs_fs_info
*fs_info
,
655 struct btrfs_ioctl_dev_replace_args
*args
)
657 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
658 struct btrfs_device
*srcdev
;
660 btrfs_dev_replace_lock(dev_replace
);
661 /* even if !dev_replace_is_valid, the values are good enough for
662 * the replace_status ioctl */
663 args
->result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR
;
664 args
->status
.replace_state
= dev_replace
->replace_state
;
665 args
->status
.time_started
= dev_replace
->time_started
;
666 args
->status
.time_stopped
= dev_replace
->time_stopped
;
667 args
->status
.num_write_errors
=
668 atomic64_read(&dev_replace
->num_write_errors
);
669 args
->status
.num_uncorrectable_read_errors
=
670 atomic64_read(&dev_replace
->num_uncorrectable_read_errors
);
671 switch (dev_replace
->replace_state
) {
672 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
673 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
674 args
->status
.progress_1000
= 0;
676 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
677 args
->status
.progress_1000
= 1000;
679 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
680 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
681 srcdev
= dev_replace
->srcdev
;
682 args
->status
.progress_1000
= div64_u64(dev_replace
->cursor_left
,
683 div64_u64(btrfs_device_get_total_bytes(srcdev
), 1000));
686 btrfs_dev_replace_unlock(dev_replace
);
689 int btrfs_dev_replace_cancel(struct btrfs_fs_info
*fs_info
,
690 struct btrfs_ioctl_dev_replace_args
*args
)
692 args
->result
= __btrfs_dev_replace_cancel(fs_info
);
696 static u64
__btrfs_dev_replace_cancel(struct btrfs_fs_info
*fs_info
)
698 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
699 struct btrfs_device
*tgt_device
= NULL
;
700 struct btrfs_trans_handle
*trans
;
701 struct btrfs_root
*root
= fs_info
->tree_root
;
705 if (fs_info
->sb
->s_flags
& MS_RDONLY
)
708 mutex_lock(&dev_replace
->lock_finishing_cancel_unmount
);
709 btrfs_dev_replace_lock(dev_replace
);
710 switch (dev_replace
->replace_state
) {
711 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
712 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
713 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
714 result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_NOT_STARTED
;
715 btrfs_dev_replace_unlock(dev_replace
);
717 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
718 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
719 result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR
;
720 tgt_device
= dev_replace
->tgtdev
;
721 dev_replace
->tgtdev
= NULL
;
722 dev_replace
->srcdev
= NULL
;
725 dev_replace
->replace_state
= BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
;
726 dev_replace
->time_stopped
= get_seconds();
727 dev_replace
->item_needs_writeback
= 1;
728 btrfs_dev_replace_unlock(dev_replace
);
729 btrfs_scrub_cancel(fs_info
);
731 trans
= btrfs_start_transaction(root
, 0);
733 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
734 return PTR_ERR(trans
);
736 ret
= btrfs_commit_transaction(trans
, root
);
739 btrfs_destroy_dev_replace_tgtdev(fs_info
, tgt_device
);
742 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
746 void btrfs_dev_replace_suspend_for_unmount(struct btrfs_fs_info
*fs_info
)
748 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
750 mutex_lock(&dev_replace
->lock_finishing_cancel_unmount
);
751 btrfs_dev_replace_lock(dev_replace
);
752 switch (dev_replace
->replace_state
) {
753 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
754 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
755 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
756 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
758 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
759 dev_replace
->replace_state
=
760 BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
;
761 dev_replace
->time_stopped
= get_seconds();
762 dev_replace
->item_needs_writeback
= 1;
763 btrfs_info(fs_info
, "suspending dev_replace for unmount");
767 btrfs_dev_replace_unlock(dev_replace
);
768 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
771 /* resume dev_replace procedure that was interrupted by unmount */
772 int btrfs_resume_dev_replace_async(struct btrfs_fs_info
*fs_info
)
774 struct task_struct
*task
;
775 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
777 btrfs_dev_replace_lock(dev_replace
);
778 switch (dev_replace
->replace_state
) {
779 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
780 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
781 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
782 btrfs_dev_replace_unlock(dev_replace
);
784 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
786 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
787 dev_replace
->replace_state
=
788 BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
;
791 if (!dev_replace
->tgtdev
|| !dev_replace
->tgtdev
->bdev
) {
792 btrfs_info(fs_info
, "cannot continue dev_replace, tgtdev is missing");
794 "you may cancel the operation after 'mount -o degraded'");
795 btrfs_dev_replace_unlock(dev_replace
);
798 btrfs_dev_replace_unlock(dev_replace
);
801 &fs_info
->mutually_exclusive_operation_running
, 1));
802 task
= kthread_run(btrfs_dev_replace_kthread
, fs_info
, "btrfs-devrepl");
803 return PTR_ERR_OR_ZERO(task
);
806 static int btrfs_dev_replace_kthread(void *data
)
808 struct btrfs_fs_info
*fs_info
= data
;
809 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
810 struct btrfs_ioctl_dev_replace_args
*status_args
;
813 status_args
= kzalloc(sizeof(*status_args
), GFP_NOFS
);
815 btrfs_dev_replace_status(fs_info
, status_args
);
816 progress
= status_args
->status
.progress_1000
;
818 do_div(progress
, 10);
819 printk_in_rcu(KERN_INFO
820 "BTRFS: continuing dev_replace from %s (devid %llu) to %s @%u%%\n",
821 dev_replace
->srcdev
->missing
? "<missing disk>" :
822 rcu_str_deref(dev_replace
->srcdev
->name
),
823 dev_replace
->srcdev
->devid
,
824 dev_replace
->tgtdev
?
825 rcu_str_deref(dev_replace
->tgtdev
->name
) :
826 "<missing target disk>",
827 (unsigned int)progress
);
829 btrfs_dev_replace_continue_on_mount(fs_info
);
830 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
835 static int btrfs_dev_replace_continue_on_mount(struct btrfs_fs_info
*fs_info
)
837 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
840 ret
= btrfs_scrub_dev(fs_info
, dev_replace
->srcdev
->devid
,
841 dev_replace
->committed_cursor_left
,
842 btrfs_device_get_total_bytes(dev_replace
->srcdev
),
843 &dev_replace
->scrub_progress
, 0, 1);
844 ret
= btrfs_dev_replace_finishing(fs_info
, ret
);
849 int btrfs_dev_replace_is_ongoing(struct btrfs_dev_replace
*dev_replace
)
851 if (!dev_replace
->is_valid
)
854 switch (dev_replace
->replace_state
) {
855 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
856 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
857 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
859 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
860 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
862 * return true even if tgtdev is missing (this is
863 * something that can happen if the dev_replace
864 * procedure is suspended by an umount and then
865 * the tgtdev is missing (or "btrfs dev scan") was
866 * not called and the the filesystem is remounted
867 * in degraded state. This does not stop the
868 * dev_replace procedure. It needs to be canceled
869 * manually if the cancelation is wanted.
876 void btrfs_dev_replace_lock(struct btrfs_dev_replace
*dev_replace
)
878 /* the beginning is just an optimization for the typical case */
879 if (atomic_read(&dev_replace
->nesting_level
) == 0) {
881 /* this is not a nested case where the same thread
882 * is trying to acqurire the same lock twice */
883 mutex_lock(&dev_replace
->lock
);
884 mutex_lock(&dev_replace
->lock_management_lock
);
885 dev_replace
->lock_owner
= current
->pid
;
886 atomic_inc(&dev_replace
->nesting_level
);
887 mutex_unlock(&dev_replace
->lock_management_lock
);
891 mutex_lock(&dev_replace
->lock_management_lock
);
892 if (atomic_read(&dev_replace
->nesting_level
) > 0 &&
893 dev_replace
->lock_owner
== current
->pid
) {
894 WARN_ON(!mutex_is_locked(&dev_replace
->lock
));
895 atomic_inc(&dev_replace
->nesting_level
);
896 mutex_unlock(&dev_replace
->lock_management_lock
);
900 mutex_unlock(&dev_replace
->lock_management_lock
);
904 void btrfs_dev_replace_unlock(struct btrfs_dev_replace
*dev_replace
)
906 WARN_ON(!mutex_is_locked(&dev_replace
->lock
));
907 mutex_lock(&dev_replace
->lock_management_lock
);
908 WARN_ON(atomic_read(&dev_replace
->nesting_level
) < 1);
909 WARN_ON(dev_replace
->lock_owner
!= current
->pid
);
910 atomic_dec(&dev_replace
->nesting_level
);
911 if (atomic_read(&dev_replace
->nesting_level
) == 0) {
912 dev_replace
->lock_owner
= 0;
913 mutex_unlock(&dev_replace
->lock_management_lock
);
914 mutex_unlock(&dev_replace
->lock
);
916 mutex_unlock(&dev_replace
->lock_management_lock
);
920 void btrfs_bio_counter_inc_noblocked(struct btrfs_fs_info
*fs_info
)
922 percpu_counter_inc(&fs_info
->bio_counter
);
925 void btrfs_bio_counter_sub(struct btrfs_fs_info
*fs_info
, s64 amount
)
927 percpu_counter_sub(&fs_info
->bio_counter
, amount
);
929 if (waitqueue_active(&fs_info
->replace_wait
))
930 wake_up(&fs_info
->replace_wait
);
933 void btrfs_bio_counter_inc_blocked(struct btrfs_fs_info
*fs_info
)
937 percpu_counter_inc(&fs_info
->bio_counter
);
938 if (test_bit(BTRFS_FS_STATE_DEV_REPLACING
, &fs_info
->fs_state
)) {
939 btrfs_bio_counter_dec(fs_info
);
940 wait_event(fs_info
->replace_wait
,
941 !test_bit(BTRFS_FS_STATE_DEV_REPLACING
,
942 &fs_info
->fs_state
));