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>
31 #include "extent_map.h"
33 #include "transaction.h"
34 #include "print-tree.h"
36 #include "async-thread.h"
37 #include "check-integrity.h"
38 #include "rcu-string.h"
39 #include "dev-replace.h"
41 static u64
btrfs_get_seconds_since_1970(void);
42 static int btrfs_dev_replace_finishing(struct btrfs_fs_info
*fs_info
,
44 static void btrfs_dev_replace_update_device_in_mapping_tree(
45 struct btrfs_fs_info
*fs_info
,
46 struct btrfs_device
*srcdev
,
47 struct btrfs_device
*tgtdev
);
48 static int btrfs_dev_replace_find_srcdev(struct btrfs_root
*root
, u64 srcdevid
,
50 struct btrfs_device
**device
);
51 static u64
__btrfs_dev_replace_cancel(struct btrfs_fs_info
*fs_info
);
52 static int btrfs_dev_replace_kthread(void *data
);
53 static int btrfs_dev_replace_continue_on_mount(struct btrfs_fs_info
*fs_info
);
56 int btrfs_init_dev_replace(struct btrfs_fs_info
*fs_info
)
59 struct btrfs_root
*dev_root
= fs_info
->dev_root
;
60 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
61 struct extent_buffer
*eb
;
64 struct btrfs_path
*path
= NULL
;
66 struct btrfs_dev_replace_item
*ptr
;
69 path
= btrfs_alloc_path();
76 key
.type
= BTRFS_DEV_REPLACE_KEY
;
78 ret
= btrfs_search_slot(NULL
, dev_root
, &key
, path
, 0, 0);
80 no_valid_dev_replace_entry_found
:
82 dev_replace
->replace_state
=
83 BTRFS_DEV_REPLACE_ITEM_STATE_NEVER_STARTED
;
84 dev_replace
->cont_reading_from_srcdev_mode
=
85 BTRFS_DEV_REPLACE_ITEM_CONT_READING_FROM_SRCDEV_MODE_ALWAYS
;
86 dev_replace
->replace_state
= 0;
87 dev_replace
->time_started
= 0;
88 dev_replace
->time_stopped
= 0;
89 atomic64_set(&dev_replace
->num_write_errors
, 0);
90 atomic64_set(&dev_replace
->num_uncorrectable_read_errors
, 0);
91 dev_replace
->cursor_left
= 0;
92 dev_replace
->committed_cursor_left
= 0;
93 dev_replace
->cursor_left_last_write_of_item
= 0;
94 dev_replace
->cursor_right
= 0;
95 dev_replace
->srcdev
= NULL
;
96 dev_replace
->tgtdev
= NULL
;
97 dev_replace
->is_valid
= 0;
98 dev_replace
->item_needs_writeback
= 0;
101 slot
= path
->slots
[0];
103 item_size
= btrfs_item_size_nr(eb
, slot
);
104 ptr
= btrfs_item_ptr(eb
, slot
, struct btrfs_dev_replace_item
);
106 if (item_size
!= sizeof(struct btrfs_dev_replace_item
)) {
107 pr_warn("btrfs: dev_replace entry found has unexpected size, ignore entry\n");
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
)) {
150 pr_warn("btrfs: cannot mount because device replace operation is ongoing and\n" "srcdev (devid %llu) is missing, need to run 'btrfs dev scan'?\n",
151 (unsigned long long)src_devid
);
153 if (!dev_replace
->tgtdev
&&
154 !btrfs_test_opt(dev_root
, DEGRADED
)) {
156 pr_warn("btrfs: cannot mount because device replace operation is ongoing and\n" "tgtdev (devid %llu) is missing, need to run btrfs dev scan?\n",
157 (unsigned long long)BTRFS_DEV_REPLACE_DEVID
);
159 if (dev_replace
->tgtdev
) {
160 if (dev_replace
->srcdev
) {
161 dev_replace
->tgtdev
->total_bytes
=
162 dev_replace
->srcdev
->total_bytes
;
163 dev_replace
->tgtdev
->disk_total_bytes
=
164 dev_replace
->srcdev
->disk_total_bytes
;
165 dev_replace
->tgtdev
->bytes_used
=
166 dev_replace
->srcdev
->bytes_used
;
168 dev_replace
->tgtdev
->is_tgtdev_for_dev_replace
= 1;
169 btrfs_init_dev_replace_tgtdev_for_resume(fs_info
,
170 dev_replace
->tgtdev
);
177 btrfs_free_path(path
);
182 * called from commit_transaction. Writes changed device replace state to
185 int btrfs_run_dev_replace(struct btrfs_trans_handle
*trans
,
186 struct btrfs_fs_info
*fs_info
)
189 struct btrfs_root
*dev_root
= fs_info
->dev_root
;
190 struct btrfs_path
*path
;
191 struct btrfs_key key
;
192 struct extent_buffer
*eb
;
193 struct btrfs_dev_replace_item
*ptr
;
194 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
196 btrfs_dev_replace_lock(dev_replace
);
197 if (!dev_replace
->is_valid
||
198 !dev_replace
->item_needs_writeback
) {
199 btrfs_dev_replace_unlock(dev_replace
);
202 btrfs_dev_replace_unlock(dev_replace
);
205 key
.type
= BTRFS_DEV_REPLACE_KEY
;
208 path
= btrfs_alloc_path();
213 ret
= btrfs_search_slot(trans
, dev_root
, &key
, path
, -1, 1);
215 pr_warn("btrfs: error %d while searching for dev_replace item!\n",
221 btrfs_item_size_nr(path
->nodes
[0], path
->slots
[0]) < sizeof(*ptr
)) {
223 * need to delete old one and insert a new one.
224 * Since no attempt is made to recover any old state, if the
225 * dev_replace state is 'running', the data on the target
227 * It would be possible to recover the state: just make sure
228 * that the beginning of the item is never changed and always
229 * contains all the essential information. Then read this
230 * minimal set of information and use it as a base for the
233 ret
= btrfs_del_item(trans
, dev_root
, path
);
235 pr_warn("btrfs: delete too small dev_replace item failed %d!\n",
243 /* need to insert a new item */
244 btrfs_release_path(path
);
245 ret
= btrfs_insert_empty_item(trans
, dev_root
, path
,
248 pr_warn("btrfs: insert dev_replace item failed %d!\n",
255 ptr
= btrfs_item_ptr(eb
, path
->slots
[0],
256 struct btrfs_dev_replace_item
);
258 btrfs_dev_replace_lock(dev_replace
);
259 if (dev_replace
->srcdev
)
260 btrfs_set_dev_replace_src_devid(eb
, ptr
,
261 dev_replace
->srcdev
->devid
);
263 btrfs_set_dev_replace_src_devid(eb
, ptr
, (u64
)-1);
264 btrfs_set_dev_replace_cont_reading_from_srcdev_mode(eb
, ptr
,
265 dev_replace
->cont_reading_from_srcdev_mode
);
266 btrfs_set_dev_replace_replace_state(eb
, ptr
,
267 dev_replace
->replace_state
);
268 btrfs_set_dev_replace_time_started(eb
, ptr
, dev_replace
->time_started
);
269 btrfs_set_dev_replace_time_stopped(eb
, ptr
, dev_replace
->time_stopped
);
270 btrfs_set_dev_replace_num_write_errors(eb
, ptr
,
271 atomic64_read(&dev_replace
->num_write_errors
));
272 btrfs_set_dev_replace_num_uncorrectable_read_errors(eb
, ptr
,
273 atomic64_read(&dev_replace
->num_uncorrectable_read_errors
));
274 dev_replace
->cursor_left_last_write_of_item
=
275 dev_replace
->cursor_left
;
276 btrfs_set_dev_replace_cursor_left(eb
, ptr
,
277 dev_replace
->cursor_left_last_write_of_item
);
278 btrfs_set_dev_replace_cursor_right(eb
, ptr
,
279 dev_replace
->cursor_right
);
280 dev_replace
->item_needs_writeback
= 0;
281 btrfs_dev_replace_unlock(dev_replace
);
283 btrfs_mark_buffer_dirty(eb
);
286 btrfs_free_path(path
);
291 void btrfs_after_dev_replace_commit(struct btrfs_fs_info
*fs_info
)
293 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
295 dev_replace
->committed_cursor_left
=
296 dev_replace
->cursor_left_last_write_of_item
;
299 static u64
btrfs_get_seconds_since_1970(void)
301 struct timespec t
= CURRENT_TIME_SEC
;
306 int btrfs_dev_replace_start(struct btrfs_root
*root
,
307 struct btrfs_ioctl_dev_replace_args
*args
)
309 struct btrfs_trans_handle
*trans
;
310 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
311 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
313 struct btrfs_device
*tgt_device
= NULL
;
314 struct btrfs_device
*src_device
= NULL
;
316 switch (args
->start
.cont_reading_from_srcdev_mode
) {
317 case BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_ALWAYS
:
318 case BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_AVOID
:
324 if ((args
->start
.srcdevid
== 0 && args
->start
.srcdev_name
[0] == '\0') ||
325 args
->start
.tgtdev_name
[0] == '\0')
328 mutex_lock(&fs_info
->volume_mutex
);
329 ret
= btrfs_init_dev_replace_tgtdev(root
, args
->start
.tgtdev_name
,
332 pr_err("btrfs: target device %s is invalid!\n",
333 args
->start
.tgtdev_name
);
334 mutex_unlock(&fs_info
->volume_mutex
);
338 ret
= btrfs_dev_replace_find_srcdev(root
, args
->start
.srcdevid
,
339 args
->start
.srcdev_name
,
341 mutex_unlock(&fs_info
->volume_mutex
);
347 if (tgt_device
->total_bytes
< src_device
->total_bytes
) {
348 pr_err("btrfs: target device is smaller than source device!\n");
353 btrfs_dev_replace_lock(dev_replace
);
354 switch (dev_replace
->replace_state
) {
355 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
356 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
357 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
359 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
360 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
361 args
->result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_ALREADY_STARTED
;
365 dev_replace
->cont_reading_from_srcdev_mode
=
366 args
->start
.cont_reading_from_srcdev_mode
;
367 WARN_ON(!src_device
);
368 dev_replace
->srcdev
= src_device
;
369 WARN_ON(!tgt_device
);
370 dev_replace
->tgtdev
= tgt_device
;
372 printk_in_rcu(KERN_INFO
373 "btrfs: dev_replace from %s (devid %llu) to %s) started\n",
374 src_device
->missing
? "<missing disk>" :
375 rcu_str_deref(src_device
->name
),
377 rcu_str_deref(tgt_device
->name
));
379 tgt_device
->total_bytes
= src_device
->total_bytes
;
380 tgt_device
->disk_total_bytes
= src_device
->disk_total_bytes
;
381 tgt_device
->bytes_used
= src_device
->bytes_used
;
384 * from now on, the writes to the srcdev are all duplicated to
385 * go to the tgtdev as well (refer to btrfs_map_block()).
387 dev_replace
->replace_state
= BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
;
388 dev_replace
->time_started
= btrfs_get_seconds_since_1970();
389 dev_replace
->cursor_left
= 0;
390 dev_replace
->committed_cursor_left
= 0;
391 dev_replace
->cursor_left_last_write_of_item
= 0;
392 dev_replace
->cursor_right
= 0;
393 dev_replace
->is_valid
= 1;
394 dev_replace
->item_needs_writeback
= 1;
395 args
->result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR
;
396 btrfs_dev_replace_unlock(dev_replace
);
398 btrfs_wait_ordered_extents(root
, 0);
400 /* force writing the updated state information to disk */
401 trans
= btrfs_start_transaction(root
, 0);
403 ret
= PTR_ERR(trans
);
404 btrfs_dev_replace_lock(dev_replace
);
408 ret
= btrfs_commit_transaction(trans
, root
);
411 /* the disk copy procedure reuses the scrub code */
412 ret
= btrfs_scrub_dev(fs_info
, src_device
->devid
, 0,
413 src_device
->total_bytes
,
414 &dev_replace
->scrub_progress
, 0, 1);
416 ret
= btrfs_dev_replace_finishing(root
->fs_info
, ret
);
422 dev_replace
->srcdev
= NULL
;
423 dev_replace
->tgtdev
= NULL
;
424 btrfs_dev_replace_unlock(dev_replace
);
427 btrfs_destroy_dev_replace_tgtdev(fs_info
, tgt_device
);
431 static int btrfs_dev_replace_finishing(struct btrfs_fs_info
*fs_info
,
434 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
435 struct btrfs_device
*tgt_device
;
436 struct btrfs_device
*src_device
;
437 struct btrfs_root
*root
= fs_info
->tree_root
;
438 u8 uuid_tmp
[BTRFS_UUID_SIZE
];
439 struct btrfs_trans_handle
*trans
;
442 /* don't allow cancel or unmount to disturb the finishing procedure */
443 mutex_lock(&dev_replace
->lock_finishing_cancel_unmount
);
445 btrfs_dev_replace_lock(dev_replace
);
446 /* was the operation canceled, or is it finished? */
447 if (dev_replace
->replace_state
!=
448 BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
) {
449 btrfs_dev_replace_unlock(dev_replace
);
450 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
454 tgt_device
= dev_replace
->tgtdev
;
455 src_device
= dev_replace
->srcdev
;
456 btrfs_dev_replace_unlock(dev_replace
);
458 /* replace old device with new one in mapping tree */
460 btrfs_dev_replace_update_device_in_mapping_tree(fs_info
,
465 * flush all outstanding I/O and inode extent mappings before the
466 * copy operation is declared as being finished
468 btrfs_start_delalloc_inodes(root
, 0);
469 btrfs_wait_ordered_extents(root
, 0);
471 trans
= btrfs_start_transaction(root
, 0);
473 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
474 return PTR_ERR(trans
);
476 ret
= btrfs_commit_transaction(trans
, root
);
479 /* keep away write_all_supers() during the finishing procedure */
480 mutex_lock(&root
->fs_info
->fs_devices
->device_list_mutex
);
481 btrfs_dev_replace_lock(dev_replace
);
482 dev_replace
->replace_state
=
483 scrub_ret
? BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
484 : BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
;
485 dev_replace
->tgtdev
= NULL
;
486 dev_replace
->srcdev
= NULL
;
487 dev_replace
->time_stopped
= btrfs_get_seconds_since_1970();
488 dev_replace
->item_needs_writeback
= 1;
491 printk_in_rcu(KERN_ERR
492 "btrfs: btrfs_scrub_dev(%s, %llu, %s) failed %d\n",
493 src_device
->missing
? "<missing disk>" :
494 rcu_str_deref(src_device
->name
),
496 rcu_str_deref(tgt_device
->name
), scrub_ret
);
497 btrfs_dev_replace_unlock(dev_replace
);
498 mutex_unlock(&root
->fs_info
->fs_devices
->device_list_mutex
);
500 btrfs_destroy_dev_replace_tgtdev(fs_info
, tgt_device
);
501 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
506 printk_in_rcu(KERN_INFO
507 "btrfs: dev_replace from %s (devid %llu) to %s) finished\n",
508 src_device
->missing
? "<missing disk>" :
509 rcu_str_deref(src_device
->name
),
511 rcu_str_deref(tgt_device
->name
));
512 tgt_device
->is_tgtdev_for_dev_replace
= 0;
513 tgt_device
->devid
= src_device
->devid
;
514 src_device
->devid
= BTRFS_DEV_REPLACE_DEVID
;
515 tgt_device
->bytes_used
= src_device
->bytes_used
;
516 memcpy(uuid_tmp
, tgt_device
->uuid
, sizeof(uuid_tmp
));
517 memcpy(tgt_device
->uuid
, src_device
->uuid
, sizeof(tgt_device
->uuid
));
518 memcpy(src_device
->uuid
, uuid_tmp
, sizeof(src_device
->uuid
));
519 tgt_device
->total_bytes
= src_device
->total_bytes
;
520 tgt_device
->disk_total_bytes
= src_device
->disk_total_bytes
;
521 tgt_device
->bytes_used
= src_device
->bytes_used
;
522 if (fs_info
->sb
->s_bdev
== src_device
->bdev
)
523 fs_info
->sb
->s_bdev
= tgt_device
->bdev
;
524 if (fs_info
->fs_devices
->latest_bdev
== src_device
->bdev
)
525 fs_info
->fs_devices
->latest_bdev
= tgt_device
->bdev
;
526 list_add(&tgt_device
->dev_alloc_list
, &fs_info
->fs_devices
->alloc_list
);
528 btrfs_rm_dev_replace_srcdev(fs_info
, src_device
);
529 if (src_device
->bdev
) {
530 /* zero out the old super */
531 btrfs_scratch_superblock(src_device
);
534 * this is again a consistent state where no dev_replace procedure
535 * is running, the target device is part of the filesystem, the
536 * source device is not part of the filesystem anymore and its 1st
537 * superblock is scratched out so that it is no longer marked to
538 * belong to this filesystem.
540 btrfs_dev_replace_unlock(dev_replace
);
541 mutex_unlock(&root
->fs_info
->fs_devices
->device_list_mutex
);
543 /* write back the superblocks */
544 trans
= btrfs_start_transaction(root
, 0);
546 btrfs_commit_transaction(trans
, root
);
548 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
553 static void btrfs_dev_replace_update_device_in_mapping_tree(
554 struct btrfs_fs_info
*fs_info
,
555 struct btrfs_device
*srcdev
,
556 struct btrfs_device
*tgtdev
)
558 struct extent_map_tree
*em_tree
= &fs_info
->mapping_tree
.map_tree
;
559 struct extent_map
*em
;
560 struct map_lookup
*map
;
564 write_lock(&em_tree
->lock
);
566 em
= lookup_extent_mapping(em_tree
, start
, (u64
)-1);
569 map
= (struct map_lookup
*)em
->bdev
;
570 for (i
= 0; i
< map
->num_stripes
; i
++)
571 if (srcdev
== map
->stripes
[i
].dev
)
572 map
->stripes
[i
].dev
= tgtdev
;
573 start
= em
->start
+ em
->len
;
576 write_unlock(&em_tree
->lock
);
579 static int btrfs_dev_replace_find_srcdev(struct btrfs_root
*root
, u64 srcdevid
,
581 struct btrfs_device
**device
)
587 *device
= btrfs_find_device(root
->fs_info
, srcdevid
, NULL
,
592 ret
= btrfs_find_device_missing_or_by_path(root
, srcdev_name
,
598 void btrfs_dev_replace_status(struct btrfs_fs_info
*fs_info
,
599 struct btrfs_ioctl_dev_replace_args
*args
)
601 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
603 btrfs_dev_replace_lock(dev_replace
);
604 /* even if !dev_replace_is_valid, the values are good enough for
605 * the replace_status ioctl */
606 args
->result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR
;
607 args
->status
.replace_state
= dev_replace
->replace_state
;
608 args
->status
.time_started
= dev_replace
->time_started
;
609 args
->status
.time_stopped
= dev_replace
->time_stopped
;
610 args
->status
.num_write_errors
=
611 atomic64_read(&dev_replace
->num_write_errors
);
612 args
->status
.num_uncorrectable_read_errors
=
613 atomic64_read(&dev_replace
->num_uncorrectable_read_errors
);
614 switch (dev_replace
->replace_state
) {
615 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
616 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
617 args
->status
.progress_1000
= 0;
619 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
620 args
->status
.progress_1000
= 1000;
622 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
623 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
624 args
->status
.progress_1000
= div64_u64(dev_replace
->cursor_left
,
625 div64_u64(dev_replace
->srcdev
->total_bytes
, 1000));
628 btrfs_dev_replace_unlock(dev_replace
);
631 int btrfs_dev_replace_cancel(struct btrfs_fs_info
*fs_info
,
632 struct btrfs_ioctl_dev_replace_args
*args
)
634 args
->result
= __btrfs_dev_replace_cancel(fs_info
);
638 static u64
__btrfs_dev_replace_cancel(struct btrfs_fs_info
*fs_info
)
640 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
641 struct btrfs_device
*tgt_device
= NULL
;
642 struct btrfs_trans_handle
*trans
;
643 struct btrfs_root
*root
= fs_info
->tree_root
;
647 mutex_lock(&dev_replace
->lock_finishing_cancel_unmount
);
648 btrfs_dev_replace_lock(dev_replace
);
649 switch (dev_replace
->replace_state
) {
650 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
651 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
652 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
653 result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_NOT_STARTED
;
654 btrfs_dev_replace_unlock(dev_replace
);
656 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
657 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
658 result
= BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR
;
659 tgt_device
= dev_replace
->tgtdev
;
660 dev_replace
->tgtdev
= NULL
;
661 dev_replace
->srcdev
= NULL
;
664 dev_replace
->replace_state
= BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
;
665 dev_replace
->time_stopped
= btrfs_get_seconds_since_1970();
666 dev_replace
->item_needs_writeback
= 1;
667 btrfs_dev_replace_unlock(dev_replace
);
668 btrfs_scrub_cancel(fs_info
);
670 trans
= btrfs_start_transaction(root
, 0);
672 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
673 return PTR_ERR(trans
);
675 ret
= btrfs_commit_transaction(trans
, root
);
678 btrfs_destroy_dev_replace_tgtdev(fs_info
, tgt_device
);
681 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
685 void btrfs_dev_replace_suspend_for_unmount(struct btrfs_fs_info
*fs_info
)
687 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
689 mutex_lock(&dev_replace
->lock_finishing_cancel_unmount
);
690 btrfs_dev_replace_lock(dev_replace
);
691 switch (dev_replace
->replace_state
) {
692 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
693 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
694 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
695 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
697 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
698 dev_replace
->replace_state
=
699 BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
;
700 dev_replace
->time_stopped
= btrfs_get_seconds_since_1970();
701 dev_replace
->item_needs_writeback
= 1;
702 pr_info("btrfs: suspending dev_replace for unmount\n");
706 btrfs_dev_replace_unlock(dev_replace
);
707 mutex_unlock(&dev_replace
->lock_finishing_cancel_unmount
);
710 /* resume dev_replace procedure that was interrupted by unmount */
711 int btrfs_resume_dev_replace_async(struct btrfs_fs_info
*fs_info
)
713 struct task_struct
*task
;
714 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
716 btrfs_dev_replace_lock(dev_replace
);
717 switch (dev_replace
->replace_state
) {
718 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
719 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
720 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
721 btrfs_dev_replace_unlock(dev_replace
);
723 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
725 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
726 dev_replace
->replace_state
=
727 BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
;
730 if (!dev_replace
->tgtdev
|| !dev_replace
->tgtdev
->bdev
) {
731 pr_info("btrfs: cannot continue dev_replace, tgtdev is missing\n"
732 "btrfs: you may cancel the operation after 'mount -o degraded'\n");
733 btrfs_dev_replace_unlock(dev_replace
);
736 btrfs_dev_replace_unlock(dev_replace
);
739 &fs_info
->mutually_exclusive_operation_running
, 1));
740 task
= kthread_run(btrfs_dev_replace_kthread
, fs_info
, "btrfs-devrepl");
741 return PTR_RET(task
);
744 static int btrfs_dev_replace_kthread(void *data
)
746 struct btrfs_fs_info
*fs_info
= data
;
747 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
748 struct btrfs_ioctl_dev_replace_args
*status_args
;
751 status_args
= kzalloc(sizeof(*status_args
), GFP_NOFS
);
753 btrfs_dev_replace_status(fs_info
, status_args
);
754 progress
= status_args
->status
.progress_1000
;
756 do_div(progress
, 10);
757 printk_in_rcu(KERN_INFO
758 "btrfs: continuing dev_replace from %s (devid %llu) to %s @%u%%\n",
759 dev_replace
->srcdev
->missing
? "<missing disk>" :
760 rcu_str_deref(dev_replace
->srcdev
->name
),
761 dev_replace
->srcdev
->devid
,
762 dev_replace
->tgtdev
?
763 rcu_str_deref(dev_replace
->tgtdev
->name
) :
764 "<missing target disk>",
765 (unsigned int)progress
);
767 btrfs_dev_replace_continue_on_mount(fs_info
);
768 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
773 static int btrfs_dev_replace_continue_on_mount(struct btrfs_fs_info
*fs_info
)
775 struct btrfs_dev_replace
*dev_replace
= &fs_info
->dev_replace
;
778 ret
= btrfs_scrub_dev(fs_info
, dev_replace
->srcdev
->devid
,
779 dev_replace
->committed_cursor_left
,
780 dev_replace
->srcdev
->total_bytes
,
781 &dev_replace
->scrub_progress
, 0, 1);
782 ret
= btrfs_dev_replace_finishing(fs_info
, ret
);
787 int btrfs_dev_replace_is_ongoing(struct btrfs_dev_replace
*dev_replace
)
789 if (!dev_replace
->is_valid
)
792 switch (dev_replace
->replace_state
) {
793 case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED
:
794 case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED
:
795 case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
:
797 case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED
:
798 case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED
:
800 * return true even if tgtdev is missing (this is
801 * something that can happen if the dev_replace
802 * procedure is suspended by an umount and then
803 * the tgtdev is missing (or "btrfs dev scan") was
804 * not called and the the filesystem is remounted
805 * in degraded state. This does not stop the
806 * dev_replace procedure. It needs to be canceled
807 * manually if the cancelation is wanted.
814 void btrfs_dev_replace_lock(struct btrfs_dev_replace
*dev_replace
)
816 /* the beginning is just an optimization for the typical case */
817 if (atomic_read(&dev_replace
->nesting_level
) == 0) {
819 /* this is not a nested case where the same thread
820 * is trying to acqurire the same lock twice */
821 mutex_lock(&dev_replace
->lock
);
822 mutex_lock(&dev_replace
->lock_management_lock
);
823 dev_replace
->lock_owner
= current
->pid
;
824 atomic_inc(&dev_replace
->nesting_level
);
825 mutex_unlock(&dev_replace
->lock_management_lock
);
829 mutex_lock(&dev_replace
->lock_management_lock
);
830 if (atomic_read(&dev_replace
->nesting_level
) > 0 &&
831 dev_replace
->lock_owner
== current
->pid
) {
832 WARN_ON(!mutex_is_locked(&dev_replace
->lock
));
833 atomic_inc(&dev_replace
->nesting_level
);
834 mutex_unlock(&dev_replace
->lock_management_lock
);
838 mutex_unlock(&dev_replace
->lock_management_lock
);
842 void btrfs_dev_replace_unlock(struct btrfs_dev_replace
*dev_replace
)
844 WARN_ON(!mutex_is_locked(&dev_replace
->lock
));
845 mutex_lock(&dev_replace
->lock_management_lock
);
846 WARN_ON(atomic_read(&dev_replace
->nesting_level
) < 1);
847 WARN_ON(dev_replace
->lock_owner
!= current
->pid
);
848 atomic_dec(&dev_replace
->nesting_level
);
849 if (atomic_read(&dev_replace
->nesting_level
) == 0) {
850 dev_replace
->lock_owner
= 0;
851 mutex_unlock(&dev_replace
->lock_management_lock
);
852 mutex_unlock(&dev_replace
->lock
);
854 mutex_unlock(&dev_replace
->lock_management_lock
);