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
);
186 btrfs_free_path(path
);
191 * called from commit_transaction. Writes changed device replace state to
194 int btrfs_run_dev_replace(struct btrfs_trans_handle
*trans
,
195 struct btrfs_fs_info
*fs_info
)
198 struct btrfs_root
*dev_root
= fs_info
->dev_root
;
199 struct btrfs_path
*path
;
200 struct btrfs_key key
;
201 struct extent_buffer
*eb
;
202 struct btrfs_dev_replace_item
*ptr
;
203 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
205 btrfs_dev_replace_lock(dev_replace
);
206 if (!dev_replace
->is_valid
||
207 !dev_replace
->item_needs_writeback
) {
208 btrfs_dev_replace_unlock(dev_replace
);
211 btrfs_dev_replace_unlock(dev_replace
);
214 key
.type
= BTRFS_DEV_REPLACE_KEY
;
217 path
= btrfs_alloc_path();
222 ret
= btrfs_search_slot(trans
, dev_root
, &key
, path
, -1, 1);
224 btrfs_warn(fs_info
, "error %d while searching for dev_replace item!",
230 btrfs_item_size_nr(path
->nodes
[0], path
->slots
[0]) < sizeof(*ptr
)) {
232 * need to delete old one and insert a new one.
233 * Since no attempt is made to recover any old state, if the
234 * dev_replace state is 'running', the data on the target
236 * It would be possible to recover the state: just make sure
237 * that the beginning of the item is never changed and always
238 * contains all the essential information. Then read this
239 * minimal set of information and use it as a base for the
242 ret
= btrfs_del_item(trans
, dev_root
, path
);
244 btrfs_warn(fs_info
, "delete too small dev_replace item failed %d!",
252 /* need to insert a new item */
253 btrfs_release_path(path
);
254 ret
= btrfs_insert_empty_item(trans
, dev_root
, path
,
257 btrfs_warn(fs_info
, "insert dev_replace item failed %d!",
264 ptr
= btrfs_item_ptr(eb
, path
->slots
[0],
265 struct btrfs_dev_replace_item
);
267 btrfs_dev_replace_lock(dev_replace
);
268 if (dev_replace
->srcdev
)
269 btrfs_set_dev_replace_src_devid(eb
, ptr
,
270 dev_replace
->srcdev
->devid
);
272 btrfs_set_dev_replace_src_devid(eb
, ptr
, (u64
)-1);
273 btrfs_set_dev_replace_cont_reading_from_srcdev_mode(eb
, ptr
,
274 dev_replace
->cont_reading_from_srcdev_mode
);
275 btrfs_set_dev_replace_replace_state(eb
, ptr
,
276 dev_replace
->replace_state
);
277 btrfs_set_dev_replace_time_started(eb
, ptr
, dev_replace
->time_started
);
278 btrfs_set_dev_replace_time_stopped(eb
, ptr
, dev_replace
->time_stopped
);
279 btrfs_set_dev_replace_num_write_errors(eb
, ptr
,
280 atomic64_read(&dev_replace
->num_write_errors
));
281 btrfs_set_dev_replace_num_uncorrectable_read_errors(eb
, ptr
,
282 atomic64_read(&dev_replace
->num_uncorrectable_read_errors
));
283 dev_replace
->cursor_left_last_write_of_item
=
284 dev_replace
->cursor_left
;
285 btrfs_set_dev_replace_cursor_left(eb
, ptr
,
286 dev_replace
->cursor_left_last_write_of_item
);
287 btrfs_set_dev_replace_cursor_right(eb
, ptr
,
288 dev_replace
->cursor_right
);
289 dev_replace
->item_needs_writeback
= 0;
290 btrfs_dev_replace_unlock(dev_replace
);
292 btrfs_mark_buffer_dirty(eb
);
295 btrfs_free_path(path
);
300 void btrfs_after_dev_replace_commit(struct btrfs_fs_info
*fs_info
)
302 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
304 dev_replace
->committed_cursor_left
=
305 dev_replace
->cursor_left_last_write_of_item
;
308 int btrfs_dev_replace_start(struct btrfs_root
*root
,
309 struct btrfs_ioctl_dev_replace_args
*args
)
311 struct btrfs_trans_handle
*trans
;
312 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
313 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
315 struct btrfs_device
*tgt_device
= NULL
;
316 struct btrfs_device
*src_device
= NULL
;
318 switch (args
->start
.cont_reading_from_srcdev_mode
) {
319 case BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_ALWAYS
:
320 case BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_AVOID
:
326 if ((args
->start
.srcdevid
== 0 && args
->start
.srcdev_name
[0] == '\0') ||
327 args
->start
.tgtdev_name
[0] == '\0')
331 * Here we commit the transaction to make sure commit_total_bytes
332 * of all the devices are updated.
334 trans
= btrfs_attach_transaction(root
);
335 if (!IS_ERR(trans
)) {
336 ret
= btrfs_commit_transaction(trans
, root
);
339 } else if (PTR_ERR(trans
) != -ENOENT
) {
340 return PTR_ERR(trans
);
343 /* the disk copy procedure reuses the scrub code */
344 mutex_lock(&fs_info
->volume_mutex
);
345 ret
= btrfs_dev_replace_find_srcdev(root
, args
->start
.srcdevid
,
346 args
->start
.srcdev_name
,
349 mutex_unlock(&fs_info
->volume_mutex
);
353 ret
= btrfs_init_dev_replace_tgtdev(root
, args
->start
.tgtdev_name
,
354 src_device
, &tgt_device
);
355 mutex_unlock(&fs_info
->volume_mutex
);
359 btrfs_dev_replace_lock(dev_replace
);
360 switch (dev_replace
->replace_state
) {
361 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
362 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
363 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
365 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
366 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
367 args
->result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_ALREADY_STARTED
;
371 dev_replace
->cont_reading_from_srcdev_mode
=
372 args
->start
.cont_reading_from_srcdev_mode
;
373 WARN_ON(!src_device
);
374 dev_replace
->srcdev
= src_device
;
375 WARN_ON(!tgt_device
);
376 dev_replace
->tgtdev
= tgt_device
;
378 ret
= btrfs_kobj_add_device(tgt_device
->fs_devices
, tgt_device
);
380 btrfs_err(root
->fs_info
, "kobj add dev failed %d\n", ret
);
382 printk_in_rcu(KERN_INFO
383 "BTRFS: dev_replace from %s (devid %llu) to %s started\n",
384 src_device
->missing
? "<missing disk>" :
385 rcu_str_deref(src_device
->name
),
387 rcu_str_deref(tgt_device
->name
));
390 * from now on, the writes to the srcdev are all duplicated to
391 * go to the tgtdev as well (refer to btrfs_map_block()).
393 dev_replace
->replace_state
= BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
;
394 dev_replace
->time_started
= get_seconds();
395 dev_replace
->cursor_left
= 0;
396 dev_replace
->committed_cursor_left
= 0;
397 dev_replace
->cursor_left_last_write_of_item
= 0;
398 dev_replace
->cursor_right
= 0;
399 dev_replace
->is_valid
= 1;
400 dev_replace
->item_needs_writeback
= 1;
401 args
->result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR
;
402 btrfs_dev_replace_unlock(dev_replace
);
404 btrfs_wait_ordered_roots(root
->fs_info
, -1);
406 /* force writing the updated state information to disk */
407 trans
= btrfs_start_transaction(root
, 0);
409 ret
= PTR_ERR(trans
);
410 btrfs_dev_replace_lock(dev_replace
);
414 ret
= btrfs_commit_transaction(trans
, root
);
417 /* the disk copy procedure reuses the scrub code */
418 ret
= btrfs_scrub_dev(fs_info
, src_device
->devid
, 0,
419 btrfs_device_get_total_bytes(src_device
),
420 &dev_replace
->scrub_progress
, 0, 1);
422 ret
= btrfs_dev_replace_finishing(root
->fs_info
, ret
);
423 /* don't warn if EINPROGRESS, someone else might be running scrub */
424 if (ret
== -EINPROGRESS
) {
425 args
->result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_SCRUB_INPROGRESS
;
434 dev_replace
->srcdev
= NULL
;
435 dev_replace
->tgtdev
= NULL
;
436 btrfs_dev_replace_unlock(dev_replace
);
437 btrfs_destroy_dev_replace_tgtdev(fs_info
, tgt_device
);
442 * blocked until all flighting bios are finished.
444 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
);
447 wait_event(fs_info
->replace_wait
, !percpu_counter_sum(
448 &fs_info
->bio_counter
));
452 * we have removed target device, it is safe to allow new bios request.
454 static void btrfs_rm_dev_replace_unblocked(struct btrfs_fs_info
*fs_info
)
456 clear_bit(BTRFS_FS_STATE_DEV_REPLACING
, &fs_info
->fs_state
);
457 if (waitqueue_active(&fs_info
->replace_wait
))
458 wake_up(&fs_info
->replace_wait
);
461 static int btrfs_dev_replace_finishing(struct btrfs_fs_info
*fs_info
,
464 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
465 struct btrfs_device
*tgt_device
;
466 struct btrfs_device
*src_device
;
467 struct btrfs_root
*root
= fs_info
->tree_root
;
468 u8 uuid_tmp
[BTRFS_UUID_SIZE
];
469 struct btrfs_trans_handle
*trans
;
472 /* don't allow cancel or unmount to disturb the finishing procedure */
473 mutex_lock(&dev_replace
->lock_finishing_cancel_unmount
);
475 btrfs_dev_replace_lock(dev_replace
);
476 /* was the operation canceled, or is it finished? */
477 if (dev_replace
->replace_state
!=
478 BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
) {
479 btrfs_dev_replace_unlock(dev_replace
);
480 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
484 tgt_device
= dev_replace
->tgtdev
;
485 src_device
= dev_replace
->srcdev
;
486 btrfs_dev_replace_unlock(dev_replace
);
489 * flush all outstanding I/O and inode extent mappings before the
490 * copy operation is declared as being finished
492 ret
= btrfs_start_delalloc_roots(root
->fs_info
, 0, -1);
494 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
497 btrfs_wait_ordered_roots(root
->fs_info
, -1);
499 trans
= btrfs_start_transaction(root
, 0);
501 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
502 return PTR_ERR(trans
);
504 ret
= btrfs_commit_transaction(trans
, root
);
507 mutex_lock(&uuid_mutex
);
508 /* keep away write_all_supers() during the finishing procedure */
509 mutex_lock(&root
->fs_info
->fs_devices
->device_list_mutex
);
510 mutex_lock(&root
->fs_info
->chunk_mutex
);
511 btrfs_dev_replace_lock(dev_replace
);
512 dev_replace
->replace_state
=
513 scrub_ret
? BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
514 : BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
;
515 dev_replace
->tgtdev
= NULL
;
516 dev_replace
->srcdev
= NULL
;
517 dev_replace
->time_stopped
= get_seconds();
518 dev_replace
->item_needs_writeback
= 1;
520 /* replace old device with new one in mapping tree */
522 btrfs_dev_replace_update_device_in_mapping_tree(fs_info
,
526 printk_in_rcu(KERN_ERR
527 "BTRFS: btrfs_scrub_dev(%s, %llu, %s) failed %d\n",
528 src_device
->missing
? "<missing disk>" :
529 rcu_str_deref(src_device
->name
),
531 rcu_str_deref(tgt_device
->name
), scrub_ret
);
532 btrfs_dev_replace_unlock(dev_replace
);
533 mutex_unlock(&root
->fs_info
->chunk_mutex
);
534 mutex_unlock(&root
->fs_info
->fs_devices
->device_list_mutex
);
535 mutex_unlock(&uuid_mutex
);
537 btrfs_destroy_dev_replace_tgtdev(fs_info
, tgt_device
);
538 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
543 printk_in_rcu(KERN_INFO
544 "BTRFS: dev_replace from %s (devid %llu) to %s finished\n",
545 src_device
->missing
? "<missing disk>" :
546 rcu_str_deref(src_device
->name
),
548 rcu_str_deref(tgt_device
->name
));
549 tgt_device
->is_tgtdev_for_dev_replace
= 0;
550 tgt_device
->devid
= src_device
->devid
;
551 src_device
->devid
= BTRFS_DEV_REPLACE_DEVID
;
552 memcpy(uuid_tmp
, tgt_device
->uuid
, sizeof(uuid_tmp
));
553 memcpy(tgt_device
->uuid
, src_device
->uuid
, sizeof(tgt_device
->uuid
));
554 memcpy(src_device
->uuid
, uuid_tmp
, sizeof(src_device
->uuid
));
555 btrfs_device_set_total_bytes(tgt_device
, src_device
->total_bytes
);
556 btrfs_device_set_disk_total_bytes(tgt_device
,
557 src_device
->disk_total_bytes
);
558 btrfs_device_set_bytes_used(tgt_device
, src_device
->bytes_used
);
559 ASSERT(list_empty(&src_device
->resized_list
));
560 tgt_device
->commit_total_bytes
= src_device
->commit_total_bytes
;
561 tgt_device
->commit_bytes_used
= src_device
->bytes_used
;
562 if (fs_info
->sb
->s_bdev
== src_device
->bdev
)
563 fs_info
->sb
->s_bdev
= tgt_device
->bdev
;
564 if (fs_info
->fs_devices
->latest_bdev
== src_device
->bdev
)
565 fs_info
->fs_devices
->latest_bdev
= tgt_device
->bdev
;
566 list_add(&tgt_device
->dev_alloc_list
, &fs_info
->fs_devices
->alloc_list
);
567 fs_info
->fs_devices
->rw_devices
++;
569 btrfs_dev_replace_unlock(dev_replace
);
571 btrfs_rm_dev_replace_blocked(fs_info
);
573 btrfs_rm_dev_replace_remove_srcdev(fs_info
, src_device
);
575 btrfs_rm_dev_replace_unblocked(fs_info
);
578 * this is again a consistent state where no dev_replace procedure
579 * is running, the target device is part of the filesystem, the
580 * source device is not part of the filesystem anymore and its 1st
581 * superblock is scratched out so that it is no longer marked to
582 * belong to this filesystem.
584 mutex_unlock(&root
->fs_info
->chunk_mutex
);
585 mutex_unlock(&root
->fs_info
->fs_devices
->device_list_mutex
);
586 mutex_unlock(&uuid_mutex
);
588 /* replace the sysfs entry */
589 btrfs_kobj_rm_device(fs_info
->fs_devices
, src_device
);
590 btrfs_rm_dev_replace_free_srcdev(fs_info
, src_device
);
592 /* write back the superblocks */
593 trans
= btrfs_start_transaction(root
, 0);
595 btrfs_commit_transaction(trans
, root
);
597 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
602 static void btrfs_dev_replace_update_device_in_mapping_tree(
603 struct btrfs_fs_info
*fs_info
,
604 struct btrfs_device
*srcdev
,
605 struct btrfs_device
*tgtdev
)
607 struct extent_map_tree
*em_tree
= &fs_info
->mapping_tree
.map_tree
;
608 struct extent_map
*em
;
609 struct map_lookup
*map
;
613 write_lock(&em_tree
->lock
);
615 em
= lookup_extent_mapping(em_tree
, start
, (u64
)-1);
618 map
= (struct map_lookup
*)em
->bdev
;
619 for (i
= 0; i
< map
->num_stripes
; i
++)
620 if (srcdev
== map
->stripes
[i
].dev
)
621 map
->stripes
[i
].dev
= tgtdev
;
622 start
= em
->start
+ em
->len
;
625 write_unlock(&em_tree
->lock
);
628 static int btrfs_dev_replace_find_srcdev(struct btrfs_root
*root
, u64 srcdevid
,
630 struct btrfs_device
**device
)
636 *device
= btrfs_find_device(root
->fs_info
, srcdevid
, NULL
,
641 ret
= btrfs_find_device_missing_or_by_path(root
, srcdev_name
,
647 void btrfs_dev_replace_status(struct btrfs_fs_info
*fs_info
,
648 struct btrfs_ioctl_dev_replace_args
*args
)
650 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
651 struct btrfs_device
*srcdev
;
653 btrfs_dev_replace_lock(dev_replace
);
654 /* even if !dev_replace_is_valid, the values are good enough for
655 * the replace_status ioctl */
656 args
->result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR
;
657 args
->status
.replace_state
= dev_replace
->replace_state
;
658 args
->status
.time_started
= dev_replace
->time_started
;
659 args
->status
.time_stopped
= dev_replace
->time_stopped
;
660 args
->status
.num_write_errors
=
661 atomic64_read(&dev_replace
->num_write_errors
);
662 args
->status
.num_uncorrectable_read_errors
=
663 atomic64_read(&dev_replace
->num_uncorrectable_read_errors
);
664 switch (dev_replace
->replace_state
) {
665 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
666 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
667 args
->status
.progress_1000
= 0;
669 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
670 args
->status
.progress_1000
= 1000;
672 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
673 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
674 srcdev
= dev_replace
->srcdev
;
675 args
->status
.progress_1000
= div_u64(dev_replace
->cursor_left
,
676 div_u64(btrfs_device_get_total_bytes(srcdev
), 1000));
679 btrfs_dev_replace_unlock(dev_replace
);
682 int btrfs_dev_replace_cancel(struct btrfs_fs_info
*fs_info
,
683 struct btrfs_ioctl_dev_replace_args
*args
)
685 args
->result
= __btrfs_dev_replace_cancel(fs_info
);
689 static u64
__btrfs_dev_replace_cancel(struct btrfs_fs_info
*fs_info
)
691 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
692 struct btrfs_device
*tgt_device
= NULL
;
693 struct btrfs_trans_handle
*trans
;
694 struct btrfs_root
*root
= fs_info
->tree_root
;
698 if (fs_info
->sb
->s_flags
& MS_RDONLY
)
701 mutex_lock(&dev_replace
->lock_finishing_cancel_unmount
);
702 btrfs_dev_replace_lock(dev_replace
);
703 switch (dev_replace
->replace_state
) {
704 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
705 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
706 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
707 result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_NOT_STARTED
;
708 btrfs_dev_replace_unlock(dev_replace
);
710 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
711 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
712 result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR
;
713 tgt_device
= dev_replace
->tgtdev
;
714 dev_replace
->tgtdev
= NULL
;
715 dev_replace
->srcdev
= NULL
;
718 dev_replace
->replace_state
= BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
;
719 dev_replace
->time_stopped
= get_seconds();
720 dev_replace
->item_needs_writeback
= 1;
721 btrfs_dev_replace_unlock(dev_replace
);
722 btrfs_scrub_cancel(fs_info
);
724 trans
= btrfs_start_transaction(root
, 0);
726 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
727 return PTR_ERR(trans
);
729 ret
= btrfs_commit_transaction(trans
, root
);
732 btrfs_destroy_dev_replace_tgtdev(fs_info
, tgt_device
);
735 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
739 void btrfs_dev_replace_suspend_for_unmount(struct btrfs_fs_info
*fs_info
)
741 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
743 mutex_lock(&dev_replace
->lock_finishing_cancel_unmount
);
744 btrfs_dev_replace_lock(dev_replace
);
745 switch (dev_replace
->replace_state
) {
746 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
747 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
748 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
749 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
751 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
752 dev_replace
->replace_state
=
753 BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
;
754 dev_replace
->time_stopped
= get_seconds();
755 dev_replace
->item_needs_writeback
= 1;
756 btrfs_info(fs_info
, "suspending dev_replace for unmount");
760 btrfs_dev_replace_unlock(dev_replace
);
761 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
764 /* resume dev_replace procedure that was interrupted by unmount */
765 int btrfs_resume_dev_replace_async(struct btrfs_fs_info
*fs_info
)
767 struct task_struct
*task
;
768 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
770 btrfs_dev_replace_lock(dev_replace
);
771 switch (dev_replace
->replace_state
) {
772 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
773 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
774 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
775 btrfs_dev_replace_unlock(dev_replace
);
777 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
779 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
780 dev_replace
->replace_state
=
781 BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
;
784 if (!dev_replace
->tgtdev
|| !dev_replace
->tgtdev
->bdev
) {
785 btrfs_info(fs_info
, "cannot continue dev_replace, tgtdev is missing");
787 "you may cancel the operation after 'mount -o degraded'");
788 btrfs_dev_replace_unlock(dev_replace
);
791 btrfs_dev_replace_unlock(dev_replace
);
794 &fs_info
->mutually_exclusive_operation_running
, 1));
795 task
= kthread_run(btrfs_dev_replace_kthread
, fs_info
, "btrfs-devrepl");
796 return PTR_ERR_OR_ZERO(task
);
799 static int btrfs_dev_replace_kthread(void *data
)
801 struct btrfs_fs_info
*fs_info
= data
;
802 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
803 struct btrfs_ioctl_dev_replace_args
*status_args
;
806 status_args
= kzalloc(sizeof(*status_args
), GFP_NOFS
);
808 btrfs_dev_replace_status(fs_info
, status_args
);
809 progress
= status_args
->status
.progress_1000
;
811 progress
= div_u64(progress
, 10);
812 printk_in_rcu(KERN_INFO
813 "BTRFS: continuing dev_replace from %s (devid %llu) to %s @%u%%\n",
814 dev_replace
->srcdev
->missing
? "<missing disk>" :
815 rcu_str_deref(dev_replace
->srcdev
->name
),
816 dev_replace
->srcdev
->devid
,
817 dev_replace
->tgtdev
?
818 rcu_str_deref(dev_replace
->tgtdev
->name
) :
819 "<missing target disk>",
820 (unsigned int)progress
);
822 btrfs_dev_replace_continue_on_mount(fs_info
);
823 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
828 static int btrfs_dev_replace_continue_on_mount(struct btrfs_fs_info
*fs_info
)
830 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
833 ret
= btrfs_scrub_dev(fs_info
, dev_replace
->srcdev
->devid
,
834 dev_replace
->committed_cursor_left
,
835 btrfs_device_get_total_bytes(dev_replace
->srcdev
),
836 &dev_replace
->scrub_progress
, 0, 1);
837 ret
= btrfs_dev_replace_finishing(fs_info
, ret
);
842 int btrfs_dev_replace_is_ongoing(struct btrfs_dev_replace
*dev_replace
)
844 if (!dev_replace
->is_valid
)
847 switch (dev_replace
->replace_state
) {
848 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
849 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
850 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
852 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
853 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
855 * return true even if tgtdev is missing (this is
856 * something that can happen if the dev_replace
857 * procedure is suspended by an umount and then
858 * the tgtdev is missing (or "btrfs dev scan") was
859 * not called and the the filesystem is remounted
860 * in degraded state. This does not stop the
861 * dev_replace procedure. It needs to be canceled
862 * manually if the cancelation is wanted.
869 void btrfs_dev_replace_lock(struct btrfs_dev_replace
*dev_replace
)
871 /* the beginning is just an optimization for the typical case */
872 if (atomic_read(&dev_replace
->nesting_level
) == 0) {
874 /* this is not a nested case where the same thread
875 * is trying to acqurire the same lock twice */
876 mutex_lock(&dev_replace
->lock
);
877 mutex_lock(&dev_replace
->lock_management_lock
);
878 dev_replace
->lock_owner
= current
->pid
;
879 atomic_inc(&dev_replace
->nesting_level
);
880 mutex_unlock(&dev_replace
->lock_management_lock
);
884 mutex_lock(&dev_replace
->lock_management_lock
);
885 if (atomic_read(&dev_replace
->nesting_level
) > 0 &&
886 dev_replace
->lock_owner
== current
->pid
) {
887 WARN_ON(!mutex_is_locked(&dev_replace
->lock
));
888 atomic_inc(&dev_replace
->nesting_level
);
889 mutex_unlock(&dev_replace
->lock_management_lock
);
893 mutex_unlock(&dev_replace
->lock_management_lock
);
897 void btrfs_dev_replace_unlock(struct btrfs_dev_replace
*dev_replace
)
899 WARN_ON(!mutex_is_locked(&dev_replace
->lock
));
900 mutex_lock(&dev_replace
->lock_management_lock
);
901 WARN_ON(atomic_read(&dev_replace
->nesting_level
) < 1);
902 WARN_ON(dev_replace
->lock_owner
!= current
->pid
);
903 atomic_dec(&dev_replace
->nesting_level
);
904 if (atomic_read(&dev_replace
->nesting_level
) == 0) {
905 dev_replace
->lock_owner
= 0;
906 mutex_unlock(&dev_replace
->lock_management_lock
);
907 mutex_unlock(&dev_replace
->lock
);
909 mutex_unlock(&dev_replace
->lock_management_lock
);
913 void btrfs_bio_counter_inc_noblocked(struct btrfs_fs_info
*fs_info
)
915 percpu_counter_inc(&fs_info
->bio_counter
);
918 void btrfs_bio_counter_sub(struct btrfs_fs_info
*fs_info
, s64 amount
)
920 percpu_counter_sub(&fs_info
->bio_counter
, amount
);
922 if (waitqueue_active(&fs_info
->replace_wait
))
923 wake_up(&fs_info
->replace_wait
);
926 void btrfs_bio_counter_inc_blocked(struct btrfs_fs_info
*fs_info
)
929 percpu_counter_inc(&fs_info
->bio_counter
);
930 if (likely(!test_bit(BTRFS_FS_STATE_DEV_REPLACING
,
931 &fs_info
->fs_state
)))
934 btrfs_bio_counter_dec(fs_info
);
935 wait_event(fs_info
->replace_wait
,
936 !test_bit(BTRFS_FS_STATE_DEV_REPLACING
,
937 &fs_info
->fs_state
));