diff --git a/fs/btrfs/dev-replace.c b/fs/btrfs/dev-replace.c
index 9814347..4889082 100644
--- a/fs/btrfs/dev-replace.c
+++ b/fs/btrfs/dev-replace.c
@@ -9,6 +9,7 @@
 #include <linux/blkdev.h>
 #include <linux/kthread.h>
 #include <linux/math64.h>
+#include "misc.h"
 #include "ctree.h"
 #include "extent_map.h"
 #include "disk-io.h"
@@ -56,10 +57,9 @@
 no_valid_dev_replace_entry_found:
 		ret = 0;
 		dev_replace->replace_state =
-			BTRFS_DEV_REPLACE_ITEM_STATE_NEVER_STARTED;
+			BTRFS_IOCTL_DEV_REPLACE_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);
@@ -112,11 +112,11 @@
 		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,
+		dev_replace->srcdev = btrfs_find_device(fs_info->fs_devices,
+						src_devid, NULL, NULL, true);
+		dev_replace->tgtdev = btrfs_find_device(fs_info->fs_devices,
 							BTRFS_DEV_REPLACE_DEVID,
-							NULL, NULL);
+							NULL, NULL, true);
 		/*
 		 * allow 'btrfs dev replace_cancel' if src/tgt device is
 		 * missing
@@ -202,7 +202,7 @@
 		return PTR_ERR(bdev);
 	}
 
-	filemap_write_and_wait(bdev->bd_inode->i_mapping);
+	sync_blockdev(bdev);
 
 	devices = &fs_info->fs_devices->devices;
 	list_for_each_entry(device, devices, dev_list) {
@@ -238,7 +238,6 @@
 	}
 	rcu_assign_pointer(device->name, name);
 
-	mutex_lock(&fs_info->fs_devices->device_list_mutex);
 	set_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state);
 	device->generation = 0;
 	device->io_width = fs_info->sectorsize;
@@ -257,6 +256,8 @@
 	device->dev_stats_valid = 1;
 	set_blocksize(device->bdev, BTRFS_BDEV_BLOCKSIZE);
 	device->fs_devices = fs_info->fs_devices;
+
+	mutex_lock(&fs_info->fs_devices->device_list_mutex);
 	list_add(&device->dev_list, &fs_info->fs_devices->devices);
 	fs_info->fs_devices->num_devices++;
 	fs_info->fs_devices->open_devices++;
@@ -274,9 +275,9 @@
  * 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 btrfs_run_dev_replace(struct btrfs_trans_handle *trans)
 {
+	struct btrfs_fs_info *fs_info = trans->fs_info;
 	int ret;
 	struct btrfs_root *dev_root = fs_info->dev_root;
 	struct btrfs_path *path;
@@ -285,13 +286,13 @@
 	struct btrfs_dev_replace_item *ptr;
 	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
 
-	btrfs_dev_replace_read_lock(dev_replace);
+	down_read(&dev_replace->rwsem);
 	if (!dev_replace->is_valid ||
 	    !dev_replace->item_needs_writeback) {
-		btrfs_dev_replace_read_unlock(dev_replace);
+		up_read(&dev_replace->rwsem);
 		return 0;
 	}
-	btrfs_dev_replace_read_unlock(dev_replace);
+	up_read(&dev_replace->rwsem);
 
 	key.objectid = 0;
 	key.type = BTRFS_DEV_REPLACE_KEY;
@@ -349,7 +350,7 @@
 	ptr = btrfs_item_ptr(eb, path->slots[0],
 			     struct btrfs_dev_replace_item);
 
-	btrfs_dev_replace_write_lock(dev_replace);
+	down_write(&dev_replace->rwsem);
 	if (dev_replace->srcdev)
 		btrfs_set_dev_replace_src_devid(eb, ptr,
 			dev_replace->srcdev->devid);
@@ -372,7 +373,7 @@
 	btrfs_set_dev_replace_cursor_right(eb, ptr,
 		dev_replace->cursor_right);
 	dev_replace->item_needs_writeback = 0;
-	btrfs_dev_replace_write_unlock(dev_replace);
+	up_write(&dev_replace->rwsem);
 
 	btrfs_mark_buffer_dirty(eb);
 
@@ -382,14 +383,6 @@
 	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;
-}
-
 static char* btrfs_dev_name(struct btrfs_device *device)
 {
 	if (!device || test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state))
@@ -398,7 +391,7 @@
 		return rcu_str_deref(device->name);
 }
 
-int btrfs_dev_replace_start(struct btrfs_fs_info *fs_info,
+static int btrfs_dev_replace_start(struct btrfs_fs_info *fs_info,
 		const char *tgtdev_name, u64 srcdevid, const char *srcdev_name,
 		int read_src)
 {
@@ -409,15 +402,17 @@
 	struct btrfs_device *tgt_device = NULL;
 	struct btrfs_device *src_device = NULL;
 
-	ret = btrfs_find_device_by_devspec(fs_info, srcdevid,
-					    srcdev_name, &src_device);
-	if (ret)
-		return ret;
+	src_device = btrfs_find_device_by_devspec(fs_info, srcdevid,
+						  srcdev_name);
+	if (IS_ERR(src_device))
+		return PTR_ERR(src_device);
 
-	ret = btrfs_init_dev_replace_tgtdev(fs_info, tgtdev_name,
-					    src_device, &tgt_device);
-	if (ret)
-		return ret;
+	if (btrfs_pinned_by_swapfile(fs_info, src_device)) {
+		btrfs_warn_in_rcu(fs_info,
+	  "cannot replace device %s (devid %llu) due to active swapfile",
+			btrfs_dev_name(src_device), src_device->devid);
+		return -ETXTBSY;
+	}
 
 	/*
 	 * Here we commit the transaction to make sure commit_total_bytes
@@ -432,7 +427,12 @@
 		return PTR_ERR(trans);
 	}
 
-	btrfs_dev_replace_write_lock(dev_replace);
+	ret = btrfs_init_dev_replace_tgtdev(fs_info, tgtdev_name,
+					    src_device, &tgt_device);
+	if (ret)
+		return ret;
+
+	down_write(&dev_replace->rwsem);
 	switch (dev_replace->replace_state) {
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
@@ -442,11 +442,11 @@
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
 		ASSERT(0);
 		ret = BTRFS_IOCTL_DEV_REPLACE_RESULT_ALREADY_STARTED;
+		up_write(&dev_replace->rwsem);
 		goto leave;
 	}
 
 	dev_replace->cont_reading_from_srcdev_mode = read_src;
-	WARN_ON(!src_device);
 	dev_replace->srcdev = src_device;
 	dev_replace->tgtdev = tgt_device;
 
@@ -470,7 +470,7 @@
 	dev_replace->item_needs_writeback = 1;
 	atomic64_set(&dev_replace->num_write_errors, 0);
 	atomic64_set(&dev_replace->num_uncorrectable_read_errors, 0);
-	btrfs_dev_replace_write_unlock(dev_replace);
+	up_write(&dev_replace->rwsem);
 
 	ret = btrfs_sysfs_add_device_link(tgt_device->fs_devices, tgt_device);
 	if (ret)
@@ -478,15 +478,16 @@
 
 	btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
 
-	/* force writing the updated state information to disk */
-	trans = btrfs_start_transaction(root, 0);
+	/* Commit dev_replace state and reserve 1 item for it. */
+	trans = btrfs_start_transaction(root, 1);
 	if (IS_ERR(trans)) {
 		ret = PTR_ERR(trans);
-		btrfs_dev_replace_write_lock(dev_replace);
+		down_write(&dev_replace->rwsem);
 		dev_replace->replace_state =
 			BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED;
 		dev_replace->srcdev = NULL;
 		dev_replace->tgtdev = NULL;
+		up_write(&dev_replace->rwsem);
 		goto leave;
 	}
 
@@ -501,14 +502,13 @@
 	ret = btrfs_dev_replace_finishing(fs_info, ret);
 	if (ret == -EINPROGRESS) {
 		ret = BTRFS_IOCTL_DEV_REPLACE_RESULT_SCRUB_INPROGRESS;
-	} else {
+	} else if (ret != -ECANCELED) {
 		WARN_ON(ret);
 	}
 
 	return ret;
 
 leave:
-	btrfs_dev_replace_write_unlock(dev_replace);
 	btrfs_destroy_dev_replace_tgtdev(tgt_device);
 	return ret;
 }
@@ -536,8 +536,9 @@
 					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;
+	if (ret == BTRFS_IOCTL_DEV_REPLACE_RESULT_SCRUB_INPROGRESS ||
+	    ret == BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR)
+		return 0;
 
 	return ret;
 }
@@ -548,8 +549,8 @@
 static void btrfs_rm_dev_replace_blocked(struct btrfs_fs_info *fs_info)
 {
 	set_bit(BTRFS_FS_STATE_DEV_REPLACING, &fs_info->fs_state);
-	wait_event(fs_info->replace_wait, !percpu_counter_sum(
-		   &fs_info->bio_counter));
+	wait_event(fs_info->dev_replace.replace_wait, !percpu_counter_sum(
+		   &fs_info->dev_replace.bio_counter));
 }
 
 /*
@@ -558,7 +559,7 @@
 static void btrfs_rm_dev_replace_unblocked(struct btrfs_fs_info *fs_info)
 {
 	clear_bit(BTRFS_FS_STATE_DEV_REPLACING, &fs_info->fs_state);
-	wake_up(&fs_info->replace_wait);
+	wake_up(&fs_info->dev_replace.replace_wait);
 }
 
 static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
@@ -575,18 +576,18 @@
 	/* don't allow cancel or unmount to disturb the finishing procedure */
 	mutex_lock(&dev_replace->lock_finishing_cancel_unmount);
 
-	btrfs_dev_replace_read_lock(dev_replace);
+	down_read(&dev_replace->rwsem);
 	/* was the operation canceled, or is it finished? */
 	if (dev_replace->replace_state !=
 	    BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED) {
-		btrfs_dev_replace_read_unlock(dev_replace);
+		up_read(&dev_replace->rwsem);
 		mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
 		return 0;
 	}
 
 	tgt_device = dev_replace->tgtdev;
 	src_device = dev_replace->srcdev;
-	btrfs_dev_replace_read_unlock(dev_replace);
+	up_read(&dev_replace->rwsem);
 
 	/*
 	 * flush all outstanding I/O and inode extent mappings before the
@@ -599,18 +600,34 @@
 	}
 	btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
 
-	trans = btrfs_start_transaction(root, 0);
-	if (IS_ERR(trans)) {
-		mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
-		return PTR_ERR(trans);
-	}
-	ret = btrfs_commit_transaction(trans);
-	WARN_ON(ret);
+	/*
+	 * We have to use this loop approach because at this point src_device
+	 * has to be available for transaction commit to complete, yet new
+	 * chunks shouldn't be allocated on the device.
+	 */
+	while (1) {
+		trans = btrfs_start_transaction(root, 0);
+		if (IS_ERR(trans)) {
+			mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
+			return PTR_ERR(trans);
+		}
+		ret = btrfs_commit_transaction(trans);
+		WARN_ON(ret);
 
-	/* keep away write_all_supers() during the finishing procedure */
-	mutex_lock(&fs_info->fs_devices->device_list_mutex);
-	mutex_lock(&fs_info->chunk_mutex);
-	btrfs_dev_replace_write_lock(dev_replace);
+		/* Prevent write_all_supers() during the finishing procedure */
+		mutex_lock(&fs_info->fs_devices->device_list_mutex);
+		/* Prevent new chunks being allocated on the source device */
+		mutex_lock(&fs_info->chunk_mutex);
+
+		if (!list_empty(&src_device->post_commit_list)) {
+			mutex_unlock(&fs_info->fs_devices->device_list_mutex);
+			mutex_unlock(&fs_info->chunk_mutex);
+		} else {
+			break;
+		}
+	}
+
+	down_write(&dev_replace->rwsem);
 	dev_replace->replace_state =
 		scrub_ret ? BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
 			  : BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED;
@@ -625,12 +642,13 @@
 								src_device,
 								tgt_device);
 	} else {
-		btrfs_err_in_rcu(fs_info,
+		if (scrub_ret != -ECANCELED)
+			btrfs_err_in_rcu(fs_info,
 				 "btrfs_scrub_dev(%s, %llu, %s) failed %d",
 				 btrfs_dev_name(src_device),
 				 src_device->devid,
 				 rcu_str_deref(tgt_device->name), scrub_ret);
-		btrfs_dev_replace_write_unlock(dev_replace);
+		up_write(&dev_replace->rwsem);
 		mutex_unlock(&fs_info->chunk_mutex);
 		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
 		btrfs_rm_dev_replace_blocked(fs_info);
@@ -657,8 +675,6 @@
 	btrfs_device_set_disk_total_bytes(tgt_device,
 					  src_device->disk_total_bytes);
 	btrfs_device_set_bytes_used(tgt_device, src_device->bytes_used);
-	ASSERT(list_empty(&src_device->resized_list));
-	tgt_device->commit_total_bytes = src_device->commit_total_bytes;
 	tgt_device->commit_bytes_used = src_device->bytes_used;
 
 	btrfs_assign_next_active_device(src_device, tgt_device);
@@ -666,8 +682,7 @@
 	list_add(&tgt_device->dev_alloc_list, &fs_info->fs_devices->alloc_list);
 	fs_info->fs_devices->rw_devices++;
 
-	btrfs_dev_replace_write_unlock(dev_replace);
-
+	up_write(&dev_replace->rwsem);
 	btrfs_rm_dev_replace_blocked(fs_info);
 
 	btrfs_rm_dev_replace_remove_srcdev(src_device);
@@ -692,7 +707,7 @@
 
 	/* replace the sysfs entry */
 	btrfs_sysfs_rm_device_link(fs_info->fs_devices, src_device);
-	btrfs_rm_dev_replace_free_srcdev(fs_info, src_device);
+	btrfs_rm_dev_replace_free_srcdev(src_device);
 
 	/* write back the superblocks */
 	trans = btrfs_start_transaction(root, 0);
@@ -709,7 +724,7 @@
 						struct btrfs_device *srcdev,
 						struct btrfs_device *tgtdev)
 {
-	struct extent_map_tree *em_tree = &fs_info->mapping_tree.map_tree;
+	struct extent_map_tree *em_tree = &fs_info->mapping_tree;
 	struct extent_map *em;
 	struct map_lookup *map;
 	u64 start = 0;
@@ -764,7 +779,7 @@
 {
 	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
 
-	btrfs_dev_replace_read_lock(dev_replace);
+	down_read(&dev_replace->rwsem);
 	/* 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;
@@ -776,7 +791,7 @@
 	args->status.num_uncorrectable_read_errors =
 		atomic64_read(&dev_replace->num_uncorrectable_read_errors);
 	args->status.progress_1000 = btrfs_dev_replace_progress(fs_info);
-	btrfs_dev_replace_read_unlock(dev_replace);
+	up_read(&dev_replace->rwsem);
 }
 
 int btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info)
@@ -793,46 +808,75 @@
 		return -EROFS;
 
 	mutex_lock(&dev_replace->lock_finishing_cancel_unmount);
-	btrfs_dev_replace_write_lock(dev_replace);
+	down_write(&dev_replace->rwsem);
 	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;
-		btrfs_dev_replace_write_unlock(dev_replace);
-		goto leave;
+		up_write(&dev_replace->rwsem);
+		break;
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
+		tgt_device = dev_replace->tgtdev;
+		src_device = dev_replace->srcdev;
+		up_write(&dev_replace->rwsem);
+		ret = btrfs_scrub_cancel(fs_info);
+		if (ret < 0) {
+			result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NOT_STARTED;
+		} else {
+			result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR;
+			/*
+			 * btrfs_dev_replace_finishing() will handle the
+			 * cleanup part
+			 */
+			btrfs_info_in_rcu(fs_info,
+				"dev_replace from %s (devid %llu) to %s canceled",
+				btrfs_dev_name(src_device), src_device->devid,
+				btrfs_dev_name(tgt_device));
+		}
+		break;
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
+		/*
+		 * Scrub doing the replace isn't running so we need to do the
+		 * cleanup step of btrfs_dev_replace_finishing() here
+		 */
 		result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR;
 		tgt_device = dev_replace->tgtdev;
 		src_device = dev_replace->srcdev;
 		dev_replace->tgtdev = NULL;
 		dev_replace->srcdev = NULL;
+		dev_replace->replace_state =
+				BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED;
+		dev_replace->time_stopped = ktime_get_real_seconds();
+		dev_replace->item_needs_writeback = 1;
+
+		up_write(&dev_replace->rwsem);
+
+		/* Scrub for replace must not be running in suspended state */
+		ret = btrfs_scrub_cancel(fs_info);
+		ASSERT(ret != -ENOTCONN);
+
+		trans = btrfs_start_transaction(root, 0);
+		if (IS_ERR(trans)) {
+			mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
+			return PTR_ERR(trans);
+		}
+		ret = btrfs_commit_transaction(trans);
+		WARN_ON(ret);
+
+		btrfs_info_in_rcu(fs_info,
+		"suspended dev_replace from %s (devid %llu) to %s canceled",
+			btrfs_dev_name(src_device), src_device->devid,
+			btrfs_dev_name(tgt_device));
+
+		if (tgt_device)
+			btrfs_destroy_dev_replace_tgtdev(tgt_device);
 		break;
+	default:
+		up_write(&dev_replace->rwsem);
+		result = -EINVAL;
 	}
-	dev_replace->replace_state = BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED;
-	dev_replace->time_stopped = ktime_get_real_seconds();
-	dev_replace->item_needs_writeback = 1;
-	btrfs_dev_replace_write_unlock(dev_replace);
-	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);
-	}
-	ret = btrfs_commit_transaction(trans);
-	WARN_ON(ret);
-
-	btrfs_info_in_rcu(fs_info,
-		"dev_replace from %s (devid %llu) to %s canceled",
-		btrfs_dev_name(src_device), src_device->devid,
-		btrfs_dev_name(tgt_device));
-
-	if (tgt_device)
-		btrfs_destroy_dev_replace_tgtdev(tgt_device);
-
-leave:
 	mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
 	return result;
 }
@@ -842,7 +886,8 @@
 	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
 
 	mutex_lock(&dev_replace->lock_finishing_cancel_unmount);
-	btrfs_dev_replace_write_lock(dev_replace);
+	down_write(&dev_replace->rwsem);
+
 	switch (dev_replace->replace_state) {
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
@@ -858,7 +903,7 @@
 		break;
 	}
 
-	btrfs_dev_replace_write_unlock(dev_replace);
+	up_write(&dev_replace->rwsem);
 	mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
 }
 
@@ -868,12 +913,13 @@
 	struct task_struct *task;
 	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
 
-	btrfs_dev_replace_write_lock(dev_replace);
+	down_write(&dev_replace->rwsem);
+
 	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:
-		btrfs_dev_replace_write_unlock(dev_replace);
+		up_write(&dev_replace->rwsem);
 		return 0;
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
 		break;
@@ -887,10 +933,12 @@
 			   "cannot continue dev_replace, tgtdev is missing");
 		btrfs_info(fs_info,
 			   "you may cancel the operation after 'mount -o degraded'");
-		btrfs_dev_replace_write_unlock(dev_replace);
+		dev_replace->replace_state =
+					BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED;
+		up_write(&dev_replace->rwsem);
 		return 0;
 	}
-	btrfs_dev_replace_write_unlock(dev_replace);
+	up_write(&dev_replace->rwsem);
 
 	/*
 	 * This could collide with a paused balance, but the exclusive op logic
@@ -898,6 +946,10 @@
 	 * dev-replace to start anyway.
 	 */
 	if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
+		down_write(&dev_replace->rwsem);
+		dev_replace->replace_state =
+					BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED;
+		up_write(&dev_replace->rwsem);
 		btrfs_info(fs_info,
 		"cannot resume dev-replace, other exclusive operation running");
 		return 0;
@@ -928,7 +980,7 @@
 			      btrfs_device_get_total_bytes(dev_replace->srcdev),
 			      &dev_replace->scrub_progress, 0, 1);
 	ret = btrfs_dev_replace_finishing(fs_info, ret);
-	WARN_ON(ret);
+	WARN_ON(ret && ret != -ECANCELED);
 
 	clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
 	return 0;
@@ -951,7 +1003,7 @@
 		 * 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
+		 * not called and the filesystem is remounted
 		 * in degraded state. This does not stop the
 		 * dev_replace procedure. It needs to be canceled
 		 * manually if the cancellation is wanted.
@@ -961,81 +1013,27 @@
 	return 1;
 }
 
-void btrfs_dev_replace_read_lock(struct btrfs_dev_replace *dev_replace)
-{
-	read_lock(&dev_replace->lock);
-	atomic_inc(&dev_replace->read_locks);
-}
-
-void btrfs_dev_replace_read_unlock(struct btrfs_dev_replace *dev_replace)
-{
-	ASSERT(atomic_read(&dev_replace->read_locks) > 0);
-	atomic_dec(&dev_replace->read_locks);
-	read_unlock(&dev_replace->lock);
-}
-
-void btrfs_dev_replace_write_lock(struct btrfs_dev_replace *dev_replace)
-{
-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;
-	}
-}
-
-void btrfs_dev_replace_write_unlock(struct btrfs_dev_replace *dev_replace)
-{
-	ASSERT(atomic_read(&dev_replace->blocking_readers) == 0);
-	write_unlock(&dev_replace->lock);
-}
-
-/* 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);
-}
-
-/* acquire read lock and dec blocking cnt */
-void btrfs_dev_replace_clear_lock_blocking(
-					struct btrfs_dev_replace *dev_replace)
-{
-	/* 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);
-	/* Barrier implied by atomic_dec_and_test */
-	if (atomic_dec_and_test(&dev_replace->blocking_readers))
-		cond_wake_up_nomb(&dev_replace->read_lock_wq);
-}
-
 void btrfs_bio_counter_inc_noblocked(struct btrfs_fs_info *fs_info)
 {
-	percpu_counter_inc(&fs_info->bio_counter);
+	percpu_counter_inc(&fs_info->dev_replace.bio_counter);
 }
 
 void btrfs_bio_counter_sub(struct btrfs_fs_info *fs_info, s64 amount)
 {
-	percpu_counter_sub(&fs_info->bio_counter, amount);
-	cond_wake_up_nomb(&fs_info->replace_wait);
+	percpu_counter_sub(&fs_info->dev_replace.bio_counter, amount);
+	cond_wake_up_nomb(&fs_info->dev_replace.replace_wait);
 }
 
 void btrfs_bio_counter_inc_blocked(struct btrfs_fs_info *fs_info)
 {
 	while (1) {
-		percpu_counter_inc(&fs_info->bio_counter);
+		percpu_counter_inc(&fs_info->dev_replace.bio_counter);
 		if (likely(!test_bit(BTRFS_FS_STATE_DEV_REPLACING,
 				     &fs_info->fs_state)))
 			break;
 
 		btrfs_bio_counter_dec(fs_info);
-		wait_event(fs_info->replace_wait,
+		wait_event(fs_info->dev_replace.replace_wait,
 			   !test_bit(BTRFS_FS_STATE_DEV_REPLACING,
 				     &fs_info->fs_state));
 	}