dev-replace.c 29.1 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
/*
 * Copyright (C) STRATO AG 2012.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */
#include <linux/sched.h>
#include <linux/bio.h>
#include <linux/slab.h>
#include <linux/buffer_head.h>
#include <linux/blkdev.h>
#include <linux/random.h>
#include <linux/iocontext.h>
#include <linux/capability.h>
#include <linux/kthread.h>
#include <linux/math64.h>
#include <asm/div64.h>
#include "ctree.h"
#include "extent_map.h"
#include "disk-io.h"
#include "transaction.h"
#include "print-tree.h"
#include "volumes.h"
#include "async-thread.h"
#include "check-integrity.h"
#include "rcu-string.h"
#include "dev-replace.h"
39
#include "sysfs.h"
40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102

static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
				       int scrub_ret);
static void btrfs_dev_replace_update_device_in_mapping_tree(
						struct btrfs_fs_info *fs_info,
						struct btrfs_device *srcdev,
						struct btrfs_device *tgtdev);
static u64 __btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info);
static int btrfs_dev_replace_kthread(void *data);
static int btrfs_dev_replace_continue_on_mount(struct btrfs_fs_info *fs_info);


int btrfs_init_dev_replace(struct btrfs_fs_info *fs_info)
{
	struct btrfs_key key;
	struct btrfs_root *dev_root = fs_info->dev_root;
	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
	struct extent_buffer *eb;
	int slot;
	int ret = 0;
	struct btrfs_path *path = NULL;
	int item_size;
	struct btrfs_dev_replace_item *ptr;
	u64 src_devid;

	path = btrfs_alloc_path();
	if (!path) {
		ret = -ENOMEM;
		goto out;
	}

	key.objectid = 0;
	key.type = BTRFS_DEV_REPLACE_KEY;
	key.offset = 0;
	ret = btrfs_search_slot(NULL, dev_root, &key, path, 0, 0);
	if (ret) {
no_valid_dev_replace_entry_found:
		ret = 0;
		dev_replace->replace_state =
			BTRFS_DEV_REPLACE_ITEM_STATE_NEVER_STARTED;
		dev_replace->cont_reading_from_srcdev_mode =
		    BTRFS_DEV_REPLACE_ITEM_CONT_READING_FROM_SRCDEV_MODE_ALWAYS;
		dev_replace->replace_state = 0;
		dev_replace->time_started = 0;
		dev_replace->time_stopped = 0;
		atomic64_set(&dev_replace->num_write_errors, 0);
		atomic64_set(&dev_replace->num_uncorrectable_read_errors, 0);
		dev_replace->cursor_left = 0;
		dev_replace->committed_cursor_left = 0;
		dev_replace->cursor_left_last_write_of_item = 0;
		dev_replace->cursor_right = 0;
		dev_replace->srcdev = NULL;
		dev_replace->tgtdev = NULL;
		dev_replace->is_valid = 0;
		dev_replace->item_needs_writeback = 0;
		goto out;
	}
	slot = path->slots[0];
	eb = path->nodes[0];
	item_size = btrfs_item_size_nr(eb, slot);
	ptr = btrfs_item_ptr(eb, slot, struct btrfs_dev_replace_item);

	if (item_size != sizeof(struct btrfs_dev_replace_item)) {
103 104
		btrfs_warn(fs_info,
			"dev_replace entry found has unexpected size, ignore entry");
105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144
		goto no_valid_dev_replace_entry_found;
	}

	src_devid = btrfs_dev_replace_src_devid(eb, ptr);
	dev_replace->cont_reading_from_srcdev_mode =
		btrfs_dev_replace_cont_reading_from_srcdev_mode(eb, ptr);
	dev_replace->replace_state = btrfs_dev_replace_replace_state(eb, ptr);
	dev_replace->time_started = btrfs_dev_replace_time_started(eb, ptr);
	dev_replace->time_stopped =
		btrfs_dev_replace_time_stopped(eb, ptr);
	atomic64_set(&dev_replace->num_write_errors,
		     btrfs_dev_replace_num_write_errors(eb, ptr));
	atomic64_set(&dev_replace->num_uncorrectable_read_errors,
		     btrfs_dev_replace_num_uncorrectable_read_errors(eb, ptr));
	dev_replace->cursor_left = btrfs_dev_replace_cursor_left(eb, ptr);
	dev_replace->committed_cursor_left = dev_replace->cursor_left;
	dev_replace->cursor_left_last_write_of_item = dev_replace->cursor_left;
	dev_replace->cursor_right = btrfs_dev_replace_cursor_right(eb, ptr);
	dev_replace->is_valid = 1;

	dev_replace->item_needs_writeback = 0;
	switch (dev_replace->replace_state) {
	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
		dev_replace->srcdev = NULL;
		dev_replace->tgtdev = NULL;
		break;
	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
		dev_replace->srcdev = btrfs_find_device(fs_info, src_devid,
							NULL, NULL);
		dev_replace->tgtdev = btrfs_find_device(fs_info,
							BTRFS_DEV_REPLACE_DEVID,
							NULL, NULL);
		/*
		 * allow 'btrfs dev replace_cancel' if src/tgt device is
		 * missing
		 */
		if (!dev_replace->srcdev &&
145
		    !btrfs_test_opt(fs_info, DEGRADED)) {
146
			ret = -EIO;
147 148 149 150 151
			btrfs_warn(fs_info,
			   "cannot mount because device replace operation is ongoing and");
			btrfs_warn(fs_info,
			   "srcdev (devid %llu) is missing, need to run 'btrfs dev scan'?",
			   src_devid);
152 153
		}
		if (!dev_replace->tgtdev &&
154
		    !btrfs_test_opt(fs_info, DEGRADED)) {
155
			ret = -EIO;
156 157 158 159
			btrfs_warn(fs_info,
			   "cannot mount because device replace operation is ongoing and");
			btrfs_warn(fs_info,
			   "tgtdev (devid %llu) is missing, need to run 'btrfs dev scan'?",
160
				BTRFS_DEV_REPLACE_DEVID);
161 162 163 164 165 166 167
		}
		if (dev_replace->tgtdev) {
			if (dev_replace->srcdev) {
				dev_replace->tgtdev->total_bytes =
					dev_replace->srcdev->total_bytes;
				dev_replace->tgtdev->disk_total_bytes =
					dev_replace->srcdev->disk_total_bytes;
168 169
				dev_replace->tgtdev->commit_total_bytes =
					dev_replace->srcdev->commit_total_bytes;
170 171
				dev_replace->tgtdev->bytes_used =
					dev_replace->srcdev->bytes_used;
172 173
				dev_replace->tgtdev->commit_bytes_used =
					dev_replace->srcdev->commit_bytes_used;
174 175 176 177 178 179 180 181 182
			}
			dev_replace->tgtdev->is_tgtdev_for_dev_replace = 1;
			btrfs_init_dev_replace_tgtdev_for_resume(fs_info,
				dev_replace->tgtdev);
		}
		break;
	}

out:
183
	btrfs_free_path(path);
184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201
	return ret;
}

/*
 * called from commit_transaction. Writes changed device replace state to
 * disk.
 */
int btrfs_run_dev_replace(struct btrfs_trans_handle *trans,
			  struct btrfs_fs_info *fs_info)
{
	int ret;
	struct btrfs_root *dev_root = fs_info->dev_root;
	struct btrfs_path *path;
	struct btrfs_key key;
	struct extent_buffer *eb;
	struct btrfs_dev_replace_item *ptr;
	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;

202
	btrfs_dev_replace_lock(dev_replace, 0);
203 204
	if (!dev_replace->is_valid ||
	    !dev_replace->item_needs_writeback) {
205
		btrfs_dev_replace_unlock(dev_replace, 0);
206 207
		return 0;
	}
208
	btrfs_dev_replace_unlock(dev_replace, 0);
209 210 211 212 213 214 215 216 217 218 219 220

	key.objectid = 0;
	key.type = BTRFS_DEV_REPLACE_KEY;
	key.offset = 0;

	path = btrfs_alloc_path();
	if (!path) {
		ret = -ENOMEM;
		goto out;
	}
	ret = btrfs_search_slot(trans, dev_root, &key, path, -1, 1);
	if (ret < 0) {
J
Jeff Mahoney 已提交
221 222 223
		btrfs_warn(fs_info,
			   "error %d while searching for dev_replace item!",
			   ret);
224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241
		goto out;
	}

	if (ret == 0 &&
	    btrfs_item_size_nr(path->nodes[0], path->slots[0]) < sizeof(*ptr)) {
		/*
		 * need to delete old one and insert a new one.
		 * Since no attempt is made to recover any old state, if the
		 * dev_replace state is 'running', the data on the target
		 * drive is lost.
		 * It would be possible to recover the state: just make sure
		 * that the beginning of the item is never changed and always
		 * contains all the essential information. Then read this
		 * minimal set of information and use it as a base for the
		 * new state.
		 */
		ret = btrfs_del_item(trans, dev_root, path);
		if (ret != 0) {
J
Jeff Mahoney 已提交
242 243 244
			btrfs_warn(fs_info,
				   "delete too small dev_replace item failed %d!",
				   ret);
245 246 247 248 249 250 251 252 253 254 255
			goto out;
		}
		ret = 1;
	}

	if (ret == 1) {
		/* need to insert a new item */
		btrfs_release_path(path);
		ret = btrfs_insert_empty_item(trans, dev_root, path,
					      &key, sizeof(*ptr));
		if (ret < 0) {
J
Jeff Mahoney 已提交
256 257
			btrfs_warn(fs_info,
				   "insert dev_replace item failed %d!", ret);
258 259 260 261 262 263 264 265
			goto out;
		}
	}

	eb = path->nodes[0];
	ptr = btrfs_item_ptr(eb, path->slots[0],
			     struct btrfs_dev_replace_item);

266
	btrfs_dev_replace_lock(dev_replace, 1);
267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288
	if (dev_replace->srcdev)
		btrfs_set_dev_replace_src_devid(eb, ptr,
			dev_replace->srcdev->devid);
	else
		btrfs_set_dev_replace_src_devid(eb, ptr, (u64)-1);
	btrfs_set_dev_replace_cont_reading_from_srcdev_mode(eb, ptr,
		dev_replace->cont_reading_from_srcdev_mode);
	btrfs_set_dev_replace_replace_state(eb, ptr,
		dev_replace->replace_state);
	btrfs_set_dev_replace_time_started(eb, ptr, dev_replace->time_started);
	btrfs_set_dev_replace_time_stopped(eb, ptr, dev_replace->time_stopped);
	btrfs_set_dev_replace_num_write_errors(eb, ptr,
		atomic64_read(&dev_replace->num_write_errors));
	btrfs_set_dev_replace_num_uncorrectable_read_errors(eb, ptr,
		atomic64_read(&dev_replace->num_uncorrectable_read_errors));
	dev_replace->cursor_left_last_write_of_item =
		dev_replace->cursor_left;
	btrfs_set_dev_replace_cursor_left(eb, ptr,
		dev_replace->cursor_left_last_write_of_item);
	btrfs_set_dev_replace_cursor_right(eb, ptr,
		dev_replace->cursor_right);
	dev_replace->item_needs_writeback = 0;
289
	btrfs_dev_replace_unlock(dev_replace, 1);
290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306

	btrfs_mark_buffer_dirty(eb);

out:
	btrfs_free_path(path);

	return ret;
}

void btrfs_after_dev_replace_commit(struct btrfs_fs_info *fs_info)
{
	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;

	dev_replace->committed_cursor_left =
		dev_replace->cursor_left_last_write_of_item;
}

307 308 309
int btrfs_dev_replace_start(struct btrfs_fs_info *fs_info,
		const char *tgtdev_name, u64 srcdevid, const char *srcdev_name,
		int read_src)
310
{
311
	struct btrfs_root *root = fs_info->dev_root;
312 313 314 315 316 317
	struct btrfs_trans_handle *trans;
	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
	int ret;
	struct btrfs_device *tgt_device = NULL;
	struct btrfs_device *src_device = NULL;

318
	/* the disk copy procedure reuses the scrub code */
319
	mutex_lock(&fs_info->volume_mutex);
320
	ret = btrfs_find_device_by_devspec(fs_info, srcdevid,
321
					    srcdev_name, &src_device);
322
	if (ret) {
323 324
		mutex_unlock(&fs_info->volume_mutex);
		return ret;
325 326
	}

327
	ret = btrfs_init_dev_replace_tgtdev(fs_info, tgtdev_name,
328 329 330 331
					    src_device, &tgt_device);
	mutex_unlock(&fs_info->volume_mutex);
	if (ret)
		return ret;
332

333 334 335 336 337 338
	/*
	 * Here we commit the transaction to make sure commit_total_bytes
	 * of all the devices are updated.
	 */
	trans = btrfs_attach_transaction(root);
	if (!IS_ERR(trans)) {
339
		ret = btrfs_commit_transaction(trans);
340 341 342 343 344 345
		if (ret)
			return ret;
	} else if (PTR_ERR(trans) != -ENOENT) {
		return PTR_ERR(trans);
	}

346
	btrfs_dev_replace_lock(dev_replace, 1);
347 348 349 350 351 352 353
	switch (dev_replace->replace_state) {
	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
		break;
	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
354
		ret = BTRFS_IOCTL_DEV_REPLACE_RESULT_ALREADY_STARTED;
355 356 357
		goto leave;
	}

358
	dev_replace->cont_reading_from_srcdev_mode = read_src;
359 360 361 362 363
	WARN_ON(!src_device);
	dev_replace->srcdev = src_device;
	WARN_ON(!tgt_device);
	dev_replace->tgtdev = tgt_device;

A
Anand Jain 已提交
364
	btrfs_info_in_rcu(fs_info,
365
		      "dev_replace from %s (devid %llu) to %s started",
366 367 368 369 370 371 372 373 374 375
		      src_device->missing ? "<missing disk>" :
		        rcu_str_deref(src_device->name),
		      src_device->devid,
		      rcu_str_deref(tgt_device->name));

	/*
	 * from now on, the writes to the srcdev are all duplicated to
	 * go to the tgtdev as well (refer to btrfs_map_block()).
	 */
	dev_replace->replace_state = BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED;
376
	dev_replace->time_started = get_seconds();
377 378 379 380 381 382
	dev_replace->cursor_left = 0;
	dev_replace->committed_cursor_left = 0;
	dev_replace->cursor_left_last_write_of_item = 0;
	dev_replace->cursor_right = 0;
	dev_replace->is_valid = 1;
	dev_replace->item_needs_writeback = 1;
383 384
	atomic64_set(&dev_replace->num_write_errors, 0);
	atomic64_set(&dev_replace->num_uncorrectable_read_errors, 0);
385
	btrfs_dev_replace_unlock(dev_replace, 1);
386

387 388
	ret = btrfs_sysfs_add_device_link(tgt_device->fs_devices, tgt_device);
	if (ret)
389
		btrfs_err(fs_info, "kobj add dev failed %d", ret);
390

391
	btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
392 393 394 395 396

	/* force writing the updated state information to disk */
	trans = btrfs_start_transaction(root, 0);
	if (IS_ERR(trans)) {
		ret = PTR_ERR(trans);
397
		btrfs_dev_replace_lock(dev_replace, 1);
398 399 400
		goto leave;
	}

401
	ret = btrfs_commit_transaction(trans);
402 403 404 405
	WARN_ON(ret);

	/* the disk copy procedure reuses the scrub code */
	ret = btrfs_scrub_dev(fs_info, src_device->devid, 0,
406
			      btrfs_device_get_total_bytes(src_device),
407 408
			      &dev_replace->scrub_progress, 0, 1);

A
Anand Jain 已提交
409
	ret = btrfs_dev_replace_finishing(fs_info, ret);
410
	if (ret == -EINPROGRESS) {
411
		ret = BTRFS_IOCTL_DEV_REPLACE_RESULT_SCRUB_INPROGRESS;
412 413 414
	} else {
		WARN_ON(ret);
	}
415

416
	return ret;
417 418 419 420

leave:
	dev_replace->srcdev = NULL;
	dev_replace->tgtdev = NULL;
421
	btrfs_dev_replace_unlock(dev_replace, 1);
422
	btrfs_destroy_dev_replace_tgtdev(fs_info, tgt_device);
423 424 425
	return ret;
}

426
int btrfs_dev_replace_by_ioctl(struct btrfs_fs_info *fs_info,
427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442
			    struct btrfs_ioctl_dev_replace_args *args)
{
	int ret;

	switch (args->start.cont_reading_from_srcdev_mode) {
	case BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_ALWAYS:
	case BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_AVOID:
		break;
	default:
		return -EINVAL;
	}

	if ((args->start.srcdevid == 0 && args->start.srcdev_name[0] == '\0') ||
	    args->start.tgtdev_name[0] == '\0')
		return -EINVAL;

443
	ret = btrfs_dev_replace_start(fs_info, args->start.tgtdev_name,
444 445 446 447 448 449 450 451 452 453 454
					args->start.srcdevid,
					args->start.srcdev_name,
					args->start.cont_reading_from_srcdev_mode);
	args->result = ret;
	/* don't warn if EINPROGRESS, someone else might be running scrub */
	if (ret == BTRFS_IOCTL_DEV_REPLACE_RESULT_SCRUB_INPROGRESS)
		ret = 0;

	return ret;
}

455
/*
456
 * blocked until all in-flight bios operations are finished.
457 458 459 460
 */
static void btrfs_rm_dev_replace_blocked(struct btrfs_fs_info *fs_info)
{
	set_bit(BTRFS_FS_STATE_DEV_REPLACING, &fs_info->fs_state);
461 462
	wait_event(fs_info->replace_wait, !percpu_counter_sum(
		   &fs_info->bio_counter));
463 464 465 466 467 468 469 470
}

/*
 * we have removed target device, it is safe to allow new bios request.
 */
static void btrfs_rm_dev_replace_unblocked(struct btrfs_fs_info *fs_info)
{
	clear_bit(BTRFS_FS_STATE_DEV_REPLACING, &fs_info->fs_state);
471
	wake_up(&fs_info->replace_wait);
472 473
}

474 475 476 477 478 479 480 481 482 483 484 485 486 487
static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
				       int scrub_ret)
{
	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
	struct btrfs_device *tgt_device;
	struct btrfs_device *src_device;
	struct btrfs_root *root = fs_info->tree_root;
	u8 uuid_tmp[BTRFS_UUID_SIZE];
	struct btrfs_trans_handle *trans;
	int ret = 0;

	/* don't allow cancel or unmount to disturb the finishing procedure */
	mutex_lock(&dev_replace->lock_finishing_cancel_unmount);

488
	btrfs_dev_replace_lock(dev_replace, 0);
489 490 491
	/* was the operation canceled, or is it finished? */
	if (dev_replace->replace_state !=
	    BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED) {
492
		btrfs_dev_replace_unlock(dev_replace, 0);
493 494 495 496 497 498
		mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
		return 0;
	}

	tgt_device = dev_replace->tgtdev;
	src_device = dev_replace->srcdev;
499
	btrfs_dev_replace_unlock(dev_replace, 0);
500 501 502 503 504

	/*
	 * flush all outstanding I/O and inode extent mappings before the
	 * copy operation is declared as being finished
	 */
505
	ret = btrfs_start_delalloc_roots(fs_info, 0, -1);
506 507 508 509
	if (ret) {
		mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
		return ret;
	}
510
	btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
511 512 513 514 515 516

	trans = btrfs_start_transaction(root, 0);
	if (IS_ERR(trans)) {
		mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
		return PTR_ERR(trans);
	}
517
	ret = btrfs_commit_transaction(trans);
518 519
	WARN_ON(ret);

520
	mutex_lock(&uuid_mutex);
521
	/* keep away write_all_supers() during the finishing procedure */
522 523
	mutex_lock(&fs_info->fs_devices->device_list_mutex);
	mutex_lock(&fs_info->chunk_mutex);
524
	btrfs_dev_replace_lock(dev_replace, 1);
525 526 527 528 529
	dev_replace->replace_state =
		scrub_ret ? BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
			  : BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED;
	dev_replace->tgtdev = NULL;
	dev_replace->srcdev = NULL;
530
	dev_replace->time_stopped = get_seconds();
531 532
	dev_replace->item_needs_writeback = 1;

533 534 535 536 537 538
	/* replace old device with new one in mapping tree */
	if (!scrub_ret) {
		btrfs_dev_replace_update_device_in_mapping_tree(fs_info,
								src_device,
								tgt_device);
	} else {
539 540 541 542 543 544
		btrfs_err_in_rcu(fs_info,
				 "btrfs_scrub_dev(%s, %llu, %s) failed %d",
				 src_device->missing ? "<missing disk>" :
				 rcu_str_deref(src_device->name),
				 src_device->devid,
				 rcu_str_deref(tgt_device->name), scrub_ret);
545
		btrfs_dev_replace_unlock(dev_replace, 1);
546 547
		mutex_unlock(&fs_info->chunk_mutex);
		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
548
		mutex_unlock(&uuid_mutex);
549
		btrfs_rm_dev_replace_blocked(fs_info);
550 551
		if (tgt_device)
			btrfs_destroy_dev_replace_tgtdev(fs_info, tgt_device);
552
		btrfs_rm_dev_replace_unblocked(fs_info);
553 554
		mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);

555
		return scrub_ret;
556 557
	}

558 559 560 561 562 563
	btrfs_info_in_rcu(fs_info,
			  "dev_replace from %s (devid %llu) to %s finished",
			  src_device->missing ? "<missing disk>" :
			  rcu_str_deref(src_device->name),
			  src_device->devid,
			  rcu_str_deref(tgt_device->name));
564 565 566 567 568 569
	tgt_device->is_tgtdev_for_dev_replace = 0;
	tgt_device->devid = src_device->devid;
	src_device->devid = BTRFS_DEV_REPLACE_DEVID;
	memcpy(uuid_tmp, tgt_device->uuid, sizeof(uuid_tmp));
	memcpy(tgt_device->uuid, src_device->uuid, sizeof(tgt_device->uuid));
	memcpy(src_device->uuid, uuid_tmp, sizeof(src_device->uuid));
570 571 572 573
	btrfs_device_set_total_bytes(tgt_device, src_device->total_bytes);
	btrfs_device_set_disk_total_bytes(tgt_device,
					  src_device->disk_total_bytes);
	btrfs_device_set_bytes_used(tgt_device, src_device->bytes_used);
574 575
	ASSERT(list_empty(&src_device->resized_list));
	tgt_device->commit_total_bytes = src_device->commit_total_bytes;
576
	tgt_device->commit_bytes_used = src_device->bytes_used;
577 578 579

	btrfs_assign_next_active_device(fs_info, src_device, tgt_device);

580
	list_add(&tgt_device->dev_alloc_list, &fs_info->fs_devices->alloc_list);
581
	fs_info->fs_devices->rw_devices++;
582

583
	btrfs_dev_replace_unlock(dev_replace, 1);
584

585 586
	btrfs_rm_dev_replace_blocked(fs_info);

587
	btrfs_rm_dev_replace_remove_srcdev(fs_info, src_device);
588

589 590
	btrfs_rm_dev_replace_unblocked(fs_info);

591 592 593 594 595 596 597
	/*
	 * this is again a consistent state where no dev_replace procedure
	 * is running, the target device is part of the filesystem, the
	 * source device is not part of the filesystem anymore and its 1st
	 * superblock is scratched out so that it is no longer marked to
	 * belong to this filesystem.
	 */
598 599
	mutex_unlock(&fs_info->chunk_mutex);
	mutex_unlock(&fs_info->fs_devices->device_list_mutex);
600
	mutex_unlock(&uuid_mutex);
601

602
	/* replace the sysfs entry */
603
	btrfs_sysfs_rm_device_link(fs_info->fs_devices, src_device);
604 605
	btrfs_rm_dev_replace_free_srcdev(fs_info, src_device);

606 607 608
	/* write back the superblocks */
	trans = btrfs_start_transaction(root, 0);
	if (!IS_ERR(trans))
609
		btrfs_commit_transaction(trans);
610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631

	mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);

	return 0;
}

static void btrfs_dev_replace_update_device_in_mapping_tree(
						struct btrfs_fs_info *fs_info,
						struct btrfs_device *srcdev,
						struct btrfs_device *tgtdev)
{
	struct extent_map_tree *em_tree = &fs_info->mapping_tree.map_tree;
	struct extent_map *em;
	struct map_lookup *map;
	u64 start = 0;
	int i;

	write_lock(&em_tree->lock);
	do {
		em = lookup_extent_mapping(em_tree, start, (u64)-1);
		if (!em)
			break;
632
		map = em->map_lookup;
633 634 635 636 637 638 639 640 641 642 643 644 645
		for (i = 0; i < map->num_stripes; i++)
			if (srcdev == map->stripes[i].dev)
				map->stripes[i].dev = tgtdev;
		start = em->start + em->len;
		free_extent_map(em);
	} while (start);
	write_unlock(&em_tree->lock);
}

void btrfs_dev_replace_status(struct btrfs_fs_info *fs_info,
			      struct btrfs_ioctl_dev_replace_args *args)
{
	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
646
	struct btrfs_device *srcdev;
647

648
	btrfs_dev_replace_lock(dev_replace, 0);
649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668
	/* even if !dev_replace_is_valid, the values are good enough for
	 * the replace_status ioctl */
	args->result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR;
	args->status.replace_state = dev_replace->replace_state;
	args->status.time_started = dev_replace->time_started;
	args->status.time_stopped = dev_replace->time_stopped;
	args->status.num_write_errors =
		atomic64_read(&dev_replace->num_write_errors);
	args->status.num_uncorrectable_read_errors =
		atomic64_read(&dev_replace->num_uncorrectable_read_errors);
	switch (dev_replace->replace_state) {
	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
		args->status.progress_1000 = 0;
		break;
	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
		args->status.progress_1000 = 1000;
		break;
	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
669
		srcdev = dev_replace->srcdev;
670
		args->status.progress_1000 = div64_u64(dev_replace->cursor_left,
671
			div_u64(btrfs_device_get_total_bytes(srcdev), 1000));
672 673
		break;
	}
674
	btrfs_dev_replace_unlock(dev_replace, 0);
675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692
}

int btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info,
			     struct btrfs_ioctl_dev_replace_args *args)
{
	args->result = __btrfs_dev_replace_cancel(fs_info);
	return 0;
}

static u64 __btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info)
{
	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
	struct btrfs_device *tgt_device = NULL;
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = fs_info->tree_root;
	u64 result;
	int ret;

693 694 695
	if (fs_info->sb->s_flags & MS_RDONLY)
		return -EROFS;

696
	mutex_lock(&dev_replace->lock_finishing_cancel_unmount);
697
	btrfs_dev_replace_lock(dev_replace, 1);
698 699 700 701 702
	switch (dev_replace->replace_state) {
	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
		result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NOT_STARTED;
703
		btrfs_dev_replace_unlock(dev_replace, 1);
704 705 706 707 708 709 710 711 712 713
		goto leave;
	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
		result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR;
		tgt_device = dev_replace->tgtdev;
		dev_replace->tgtdev = NULL;
		dev_replace->srcdev = NULL;
		break;
	}
	dev_replace->replace_state = BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED;
714
	dev_replace->time_stopped = get_seconds();
715
	dev_replace->item_needs_writeback = 1;
716
	btrfs_dev_replace_unlock(dev_replace, 1);
717 718 719 720 721 722 723
	btrfs_scrub_cancel(fs_info);

	trans = btrfs_start_transaction(root, 0);
	if (IS_ERR(trans)) {
		mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
		return PTR_ERR(trans);
	}
724
	ret = btrfs_commit_transaction(trans);
725 726 727 728 729 730 731 732 733 734 735 736 737 738
	WARN_ON(ret);
	if (tgt_device)
		btrfs_destroy_dev_replace_tgtdev(fs_info, tgt_device);

leave:
	mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
	return result;
}

void btrfs_dev_replace_suspend_for_unmount(struct btrfs_fs_info *fs_info)
{
	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;

	mutex_lock(&dev_replace->lock_finishing_cancel_unmount);
739
	btrfs_dev_replace_lock(dev_replace, 1);
740 741 742 743 744 745 746 747 748
	switch (dev_replace->replace_state) {
	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
		break;
	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
		dev_replace->replace_state =
			BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED;
749
		dev_replace->time_stopped = get_seconds();
750
		dev_replace->item_needs_writeback = 1;
751
		btrfs_info(fs_info, "suspending dev_replace for unmount");
752 753 754
		break;
	}

755
	btrfs_dev_replace_unlock(dev_replace, 1);
756 757 758 759 760 761 762 763 764
	mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
}

/* resume dev_replace procedure that was interrupted by unmount */
int btrfs_resume_dev_replace_async(struct btrfs_fs_info *fs_info)
{
	struct task_struct *task;
	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;

765
	btrfs_dev_replace_lock(dev_replace, 1);
766 767 768 769
	switch (dev_replace->replace_state) {
	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
770
		btrfs_dev_replace_unlock(dev_replace, 1);
771 772 773 774 775 776 777 778 779
		return 0;
	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
		break;
	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
		dev_replace->replace_state =
			BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED;
		break;
	}
	if (!dev_replace->tgtdev || !dev_replace->tgtdev->bdev) {
780
		btrfs_info(fs_info,
J
Jeff Mahoney 已提交
781 782 783
			   "cannot continue dev_replace, tgtdev is missing");
		btrfs_info(fs_info,
			   "you may cancel the operation after 'mount -o degraded'");
784
		btrfs_dev_replace_unlock(dev_replace, 1);
785 786
		return 0;
	}
787
	btrfs_dev_replace_unlock(dev_replace, 1);
788

789
	WARN_ON(test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
790
	task = kthread_run(btrfs_dev_replace_kthread, fs_info, "btrfs-devrepl");
791
	return PTR_ERR_OR_ZERO(task);
792 793 794 795 796 797 798 799 800
}

static int btrfs_dev_replace_kthread(void *data)
{
	struct btrfs_fs_info *fs_info = data;
	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
	struct btrfs_ioctl_dev_replace_args *status_args;
	u64 progress;

801
	status_args = kzalloc(sizeof(*status_args), GFP_KERNEL);
802 803 804 805
	if (status_args) {
		btrfs_dev_replace_status(fs_info, status_args);
		progress = status_args->status.progress_1000;
		kfree(status_args);
806
		progress = div_u64(progress, 10);
807 808
		btrfs_info_in_rcu(fs_info,
			"continuing dev_replace from %s (devid %llu) to %s @%u%%",
809 810 811 812 813 814 815
			dev_replace->srcdev->missing ? "<missing disk>" :
			rcu_str_deref(dev_replace->srcdev->name),
			dev_replace->srcdev->devid,
			dev_replace->tgtdev ?
			rcu_str_deref(dev_replace->tgtdev->name) :
			"<missing target disk>",
			(unsigned int)progress);
816 817
	}
	btrfs_dev_replace_continue_on_mount(fs_info);
818
	clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
819 820 821 822 823 824 825 826 827 828 829

	return 0;
}

static int btrfs_dev_replace_continue_on_mount(struct btrfs_fs_info *fs_info)
{
	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
	int ret;

	ret = btrfs_scrub_dev(fs_info, dev_replace->srcdev->devid,
			      dev_replace->committed_cursor_left,
830
			      btrfs_device_get_total_bytes(dev_replace->srcdev),
831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856
			      &dev_replace->scrub_progress, 0, 1);
	ret = btrfs_dev_replace_finishing(fs_info, ret);
	WARN_ON(ret);
	return 0;
}

int btrfs_dev_replace_is_ongoing(struct btrfs_dev_replace *dev_replace)
{
	if (!dev_replace->is_valid)
		return 0;

	switch (dev_replace->replace_state) {
	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
		return 0;
	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
		/*
		 * return true even if tgtdev is missing (this is
		 * something that can happen if the dev_replace
		 * procedure is suspended by an umount and then
		 * the tgtdev is missing (or "btrfs dev scan") was
		 * not called and the the filesystem is remounted
		 * in degraded state. This does not stop the
		 * dev_replace procedure. It needs to be canceled
857
		 * manually if the cancellation is wanted.
858 859 860 861 862 863
		 */
		break;
	}
	return 1;
}

864
void btrfs_dev_replace_lock(struct btrfs_dev_replace *dev_replace, int rw)
865
{
866 867 868 869 870 871 872 873 874 875 876 877 878
	if (rw == 1) {
		/* write */
again:
		wait_event(dev_replace->read_lock_wq,
			   atomic_read(&dev_replace->blocking_readers) == 0);
		write_lock(&dev_replace->lock);
		if (atomic_read(&dev_replace->blocking_readers)) {
			write_unlock(&dev_replace->lock);
			goto again;
		}
	} else {
		read_lock(&dev_replace->lock);
		atomic_inc(&dev_replace->read_locks);
879
	}
880
}
881

882 883 884 885 886 887 888 889 890 891
void btrfs_dev_replace_unlock(struct btrfs_dev_replace *dev_replace, int rw)
{
	if (rw == 1) {
		/* write */
		ASSERT(atomic_read(&dev_replace->blocking_readers) == 0);
		write_unlock(&dev_replace->lock);
	} else {
		ASSERT(atomic_read(&dev_replace->read_locks) > 0);
		atomic_dec(&dev_replace->read_locks);
		read_unlock(&dev_replace->lock);
892
	}
893
}
894

895 896 897 898 899 900 901 902
/* inc blocking cnt and release read lock */
void btrfs_dev_replace_set_lock_blocking(
					struct btrfs_dev_replace *dev_replace)
{
	/* only set blocking for read lock */
	ASSERT(atomic_read(&dev_replace->read_locks) > 0);
	atomic_inc(&dev_replace->blocking_readers);
	read_unlock(&dev_replace->lock);
903 904
}

905 906 907
/* acquire read lock and dec blocking cnt */
void btrfs_dev_replace_clear_lock_blocking(
					struct btrfs_dev_replace *dev_replace)
908
{
909 910 911 912 913 914 915
	/* only set blocking for read lock */
	ASSERT(atomic_read(&dev_replace->read_locks) > 0);
	ASSERT(atomic_read(&dev_replace->blocking_readers) > 0);
	read_lock(&dev_replace->lock);
	if (atomic_dec_and_test(&dev_replace->blocking_readers) &&
	    waitqueue_active(&dev_replace->read_lock_wq))
		wake_up(&dev_replace->read_lock_wq);
916
}
917 918 919 920 921 922

void btrfs_bio_counter_inc_noblocked(struct btrfs_fs_info *fs_info)
{
	percpu_counter_inc(&fs_info->bio_counter);
}

923
void btrfs_bio_counter_sub(struct btrfs_fs_info *fs_info, s64 amount)
924
{
925
	percpu_counter_sub(&fs_info->bio_counter, amount);
926 927 928 929 930 931 932

	if (waitqueue_active(&fs_info->replace_wait))
		wake_up(&fs_info->replace_wait);
}

void btrfs_bio_counter_inc_blocked(struct btrfs_fs_info *fs_info)
{
933 934 935 936 937 938
	while (1) {
		percpu_counter_inc(&fs_info->bio_counter);
		if (likely(!test_bit(BTRFS_FS_STATE_DEV_REPLACING,
				     &fs_info->fs_state)))
			break;

939 940 941 942 943 944
		btrfs_bio_counter_dec(fs_info);
		wait_event(fs_info->replace_wait,
			   !test_bit(BTRFS_FS_STATE_DEV_REPLACING,
				     &fs_info->fs_state));
	}
}