Btrfs: add new sources for device replace code
authorStefan Behrens <sbehrens@giantdisaster.de>
Mon, 5 Nov 2012 16:33:06 +0000 (17:33 +0100)
committerJosef Bacik <jbacik@fusionio.com>
Wed, 12 Dec 2012 22:15:41 +0000 (17:15 -0500)
This adds a new file to the sources together with the header file
and the changes to ioctl.h and ctree.h that are required by the
new C source file. Additionally, 4 new functions are added to
volume.c that deal with device creation and destruction.

Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
fs/btrfs/Makefile
fs/btrfs/ctree.h
fs/btrfs/dev-replace.c [new file with mode: 0644]
fs/btrfs/dev-replace.h
fs/btrfs/ioctl.h
fs/btrfs/volumes.c
fs/btrfs/volumes.h

index d7fcdba141a2f2508999003344008458c2cf48ea..7df3e0f0ee512b7d6e7082c2c288e739ed1f6505 100644 (file)
@@ -8,7 +8,7 @@ btrfs-y += super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
           extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o \
           export.o tree-log.o free-space-cache.o zlib.o lzo.o \
           compression.o delayed-ref.o relocation.o delayed-inode.o scrub.o \
-          reada.o backref.o ulist.o qgroup.o send.o
+          reada.o backref.o ulist.o qgroup.o send.o dev-replace.o
 
 btrfs-$(CONFIG_BTRFS_FS_POSIX_ACL) += acl.o
 btrfs-$(CONFIG_BTRFS_FS_CHECK_INTEGRITY) += check-integrity.o
index ded7caa0d3045e940814330899f9cd989afe57ab..45e7f752b64ac170bf735c86358b4da12cff61b3 100644 (file)
@@ -142,6 +142,8 @@ struct btrfs_ordered_sum;
 
 #define BTRFS_EMPTY_SUBVOL_DIR_OBJECTID 2
 
+#define BTRFS_DEV_REPLACE_DEVID 0
+
 /*
  * the max metadata block size.  This limit is somewhat artificial,
  * but the memmove costs go through the roof for larger blocks.
diff --git a/fs/btrfs/dev-replace.c b/fs/btrfs/dev-replace.c
new file mode 100644 (file)
index 0000000..66dbc8d
--- /dev/null
@@ -0,0 +1,856 @@
+/*
+ * 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 "compat.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"
+
+static u64 btrfs_get_seconds_since_1970(void);
+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 int btrfs_dev_replace_find_srcdev(struct btrfs_root *root, u64 srcdevid,
+                                        char *srcdev_name,
+                                        struct btrfs_device **device);
+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)) {
+               pr_warn("btrfs: dev_replace entry found has unexpected size, ignore entry\n");
+               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 &&
+                   !btrfs_test_opt(dev_root, DEGRADED)) {
+                       ret = -EIO;
+                       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",
+                               (unsigned long long)src_devid);
+               }
+               if (!dev_replace->tgtdev &&
+                   !btrfs_test_opt(dev_root, DEGRADED)) {
+                       ret = -EIO;
+                       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",
+                               (unsigned long long)BTRFS_DEV_REPLACE_DEVID);
+               }
+               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;
+                               dev_replace->tgtdev->bytes_used =
+                                       dev_replace->srcdev->bytes_used;
+                       }
+                       dev_replace->tgtdev->is_tgtdev_for_dev_replace = 1;
+                       btrfs_init_dev_replace_tgtdev_for_resume(fs_info,
+                               dev_replace->tgtdev);
+               }
+               break;
+       }
+
+out:
+       if (path)
+               btrfs_free_path(path);
+       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;
+
+       btrfs_dev_replace_lock(dev_replace);
+       if (!dev_replace->is_valid ||
+           !dev_replace->item_needs_writeback) {
+               btrfs_dev_replace_unlock(dev_replace);
+               return 0;
+       }
+       btrfs_dev_replace_unlock(dev_replace);
+
+       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) {
+               pr_warn("btrfs: error %d while searching for dev_replace item!\n",
+                       ret);
+               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) {
+                       pr_warn("btrfs: delete too small dev_replace item failed %d!\n",
+                               ret);
+                       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) {
+                       pr_warn("btrfs: insert dev_replace item failed %d!\n",
+                               ret);
+                       goto out;
+               }
+       }
+
+       eb = path->nodes[0];
+       ptr = btrfs_item_ptr(eb, path->slots[0],
+                            struct btrfs_dev_replace_item);
+
+       btrfs_dev_replace_lock(dev_replace);
+       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;
+       btrfs_dev_replace_unlock(dev_replace);
+
+       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;
+}
+
+static u64 btrfs_get_seconds_since_1970(void)
+{
+       struct timespec t = CURRENT_TIME_SEC;
+
+       return t.tv_sec;
+}
+
+int btrfs_dev_replace_start(struct btrfs_root *root,
+                           struct btrfs_ioctl_dev_replace_args *args)
+{
+       struct btrfs_trans_handle *trans;
+       struct btrfs_fs_info *fs_info = root->fs_info;
+       struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
+       int ret;
+       struct btrfs_device *tgt_device = NULL;
+       struct btrfs_device *src_device = NULL;
+
+       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;
+
+       mutex_lock(&fs_info->volume_mutex);
+       ret = btrfs_init_dev_replace_tgtdev(root, args->start.tgtdev_name,
+                                           &tgt_device);
+       if (ret) {
+               pr_err("btrfs: target device %s is invalid!\n",
+                      args->start.tgtdev_name);
+               mutex_unlock(&fs_info->volume_mutex);
+               return -EINVAL;
+       }
+
+       ret = btrfs_dev_replace_find_srcdev(root, args->start.srcdevid,
+                                           args->start.srcdev_name,
+                                           &src_device);
+       mutex_unlock(&fs_info->volume_mutex);
+       if (ret) {
+               ret = -EINVAL;
+               goto leave_no_lock;
+       }
+
+       if (tgt_device->total_bytes < src_device->total_bytes) {
+               pr_err("btrfs: target device is smaller than source device!\n");
+               ret = -EINVAL;
+               goto leave_no_lock;
+       }
+
+       btrfs_dev_replace_lock(dev_replace);
+       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:
+               args->result = BTRFS_IOCTL_DEV_REPLACE_RESULT_ALREADY_STARTED;
+               goto leave;
+       }
+
+       dev_replace->cont_reading_from_srcdev_mode =
+               args->start.cont_reading_from_srcdev_mode;
+       WARN_ON(!src_device);
+       dev_replace->srcdev = src_device;
+       WARN_ON(!tgt_device);
+       dev_replace->tgtdev = tgt_device;
+
+       printk_in_rcu(KERN_INFO
+                     "btrfs: dev_replace from %s (devid %llu) to %s) started\n",
+                     src_device->missing ? "<missing disk>" :
+                       rcu_str_deref(src_device->name),
+                     src_device->devid,
+                     rcu_str_deref(tgt_device->name));
+
+       tgt_device->total_bytes = src_device->total_bytes;
+       tgt_device->disk_total_bytes = src_device->disk_total_bytes;
+       tgt_device->bytes_used = src_device->bytes_used;
+
+       /*
+        * 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;
+       dev_replace->time_started = btrfs_get_seconds_since_1970();
+       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;
+       args->result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR;
+       btrfs_dev_replace_unlock(dev_replace);
+
+       btrfs_wait_ordered_extents(root, 0);
+
+       /* force writing the updated state information to disk */
+       trans = btrfs_start_transaction(root, 0);
+       if (IS_ERR(trans)) {
+               ret = PTR_ERR(trans);
+               btrfs_dev_replace_lock(dev_replace);
+               goto leave;
+       }
+
+       ret = btrfs_commit_transaction(trans, root);
+       WARN_ON(ret);
+
+       /* the disk copy procedure reuses the scrub code */
+       ret = btrfs_scrub_dev(fs_info, src_device->devid, 0,
+                             src_device->total_bytes,
+                             &dev_replace->scrub_progress, 0, 1);
+
+       ret = btrfs_dev_replace_finishing(root->fs_info, ret);
+       WARN_ON(ret);
+
+       return 0;
+
+leave:
+       dev_replace->srcdev = NULL;
+       dev_replace->tgtdev = NULL;
+       btrfs_dev_replace_unlock(dev_replace);
+leave_no_lock:
+       if (tgt_device)
+               btrfs_destroy_dev_replace_tgtdev(fs_info, tgt_device);
+       return ret;
+}
+
+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);
+
+       btrfs_dev_replace_lock(dev_replace);
+       /* was the operation canceled, or is it finished? */
+       if (dev_replace->replace_state !=
+           BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED) {
+               btrfs_dev_replace_unlock(dev_replace);
+               mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
+               return 0;
+       }
+
+       tgt_device = dev_replace->tgtdev;
+       src_device = dev_replace->srcdev;
+       btrfs_dev_replace_unlock(dev_replace);
+
+       /* 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);
+
+       /*
+        * flush all outstanding I/O and inode extent mappings before the
+        * copy operation is declared as being finished
+        */
+       btrfs_start_delalloc_inodes(root, 0);
+       btrfs_wait_ordered_extents(root, 0);
+
+       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, root);
+       WARN_ON(ret);
+
+       /* keep away write_all_supers() during the finishing procedure */
+       mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
+       btrfs_dev_replace_lock(dev_replace);
+       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;
+       dev_replace->time_stopped = btrfs_get_seconds_since_1970();
+       dev_replace->item_needs_writeback = 1;
+
+       if (scrub_ret) {
+               printk_in_rcu(KERN_ERR
+                             "btrfs: btrfs_scrub_dev(%s, %llu, %s) failed %d\n",
+                             src_device->missing ? "<missing disk>" :
+                               rcu_str_deref(src_device->name),
+                             src_device->devid,
+                             rcu_str_deref(tgt_device->name), scrub_ret);
+               btrfs_dev_replace_unlock(dev_replace);
+               mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
+               if (tgt_device)
+                       btrfs_destroy_dev_replace_tgtdev(fs_info, tgt_device);
+               mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
+
+               return 0;
+       }
+
+       printk_in_rcu(KERN_INFO
+                     "btrfs: dev_replace from %s (devid %llu) to %s) finished\n",
+                     src_device->missing ? "<missing disk>" :
+                       rcu_str_deref(src_device->name),
+                     src_device->devid,
+                     rcu_str_deref(tgt_device->name));
+       tgt_device->is_tgtdev_for_dev_replace = 0;
+       tgt_device->devid = src_device->devid;
+       src_device->devid = BTRFS_DEV_REPLACE_DEVID;
+       tgt_device->bytes_used = src_device->bytes_used;
+       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));
+       tgt_device->total_bytes = src_device->total_bytes;
+       tgt_device->disk_total_bytes = src_device->disk_total_bytes;
+       tgt_device->bytes_used = src_device->bytes_used;
+       if (fs_info->sb->s_bdev == src_device->bdev)
+               fs_info->sb->s_bdev = tgt_device->bdev;
+       if (fs_info->fs_devices->latest_bdev == src_device->bdev)
+               fs_info->fs_devices->latest_bdev = tgt_device->bdev;
+       list_add(&tgt_device->dev_alloc_list, &fs_info->fs_devices->alloc_list);
+
+       btrfs_rm_dev_replace_srcdev(fs_info, src_device);
+       if (src_device->bdev) {
+               /* zero out the old super */
+               btrfs_scratch_superblock(src_device);
+       }
+       /*
+        * 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.
+        */
+       btrfs_dev_replace_unlock(dev_replace);
+       mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
+
+       /* write back the superblocks */
+       trans = btrfs_start_transaction(root, 0);
+       if (!IS_ERR(trans))
+               btrfs_commit_transaction(trans, root);
+
+       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;
+               map = (struct map_lookup *)em->bdev;
+               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);
+}
+
+static int btrfs_dev_replace_find_srcdev(struct btrfs_root *root, u64 srcdevid,
+                                        char *srcdev_name,
+                                        struct btrfs_device **device)
+{
+       int ret;
+
+       if (srcdevid) {
+               ret = 0;
+               *device = btrfs_find_device(root->fs_info, srcdevid, NULL,
+                                           NULL);
+               if (!*device)
+                       ret = -ENOENT;
+       } else {
+               ret = btrfs_find_device_missing_or_by_path(root, srcdev_name,
+                                                          device);
+       }
+       return ret;
+}
+
+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;
+
+       btrfs_dev_replace_lock(dev_replace);
+       /* 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:
+               args->status.progress_1000 = div64_u64(dev_replace->cursor_left,
+                       div64_u64(dev_replace->srcdev->total_bytes, 1000));
+               break;
+       }
+       btrfs_dev_replace_unlock(dev_replace);
+}
+
+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;
+
+       mutex_lock(&dev_replace->lock_finishing_cancel_unmount);
+       btrfs_dev_replace_lock(dev_replace);
+       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_unlock(dev_replace);
+               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;
+       dev_replace->time_stopped = btrfs_get_seconds_since_1970();
+       dev_replace->item_needs_writeback = 1;
+       btrfs_dev_replace_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, root);
+       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);
+       btrfs_dev_replace_lock(dev_replace);
+       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;
+               dev_replace->time_stopped = btrfs_get_seconds_since_1970();
+               dev_replace->item_needs_writeback = 1;
+               pr_info("btrfs: suspending dev_replace for unmount\n");
+               break;
+       }
+
+       btrfs_dev_replace_unlock(dev_replace);
+       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;
+
+       btrfs_dev_replace_lock(dev_replace);
+       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_unlock(dev_replace);
+               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) {
+               pr_info("btrfs: cannot continue dev_replace, tgtdev is missing\n"
+                       "btrfs: you may cancel the operation after 'mount -o degraded'\n");
+               btrfs_dev_replace_unlock(dev_replace);
+               return 0;
+       }
+       btrfs_dev_replace_unlock(dev_replace);
+
+       WARN_ON(atomic_xchg(
+               &fs_info->mutually_exclusive_operation_running, 1));
+       task = kthread_run(btrfs_dev_replace_kthread, fs_info, "btrfs-devrepl");
+       return PTR_RET(task);
+}
+
+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;
+
+       status_args = kzalloc(sizeof(*status_args), GFP_NOFS);
+       if (status_args) {
+               btrfs_dev_replace_status(fs_info, status_args);
+               progress = status_args->status.progress_1000;
+               kfree(status_args);
+               do_div(progress, 10);
+               printk_in_rcu(KERN_INFO
+                             "btrfs: continuing dev_replace from %s (devid %llu) to %s @%u%%\n",
+                             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);
+       }
+       btrfs_dev_replace_continue_on_mount(fs_info);
+       atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
+
+       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,
+                             dev_replace->srcdev->total_bytes,
+                             &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
+                * manually if the cancelation is wanted.
+                */
+               break;
+       }
+       return 1;
+}
+
+void btrfs_dev_replace_lock(struct btrfs_dev_replace *dev_replace)
+{
+       /* the beginning is just an optimization for the typical case */
+       if (atomic_read(&dev_replace->nesting_level) == 0) {
+acquire_lock:
+               /* this is not a nested case where the same thread
+                * is trying to acqurire the same lock twice */
+               mutex_lock(&dev_replace->lock);
+               mutex_lock(&dev_replace->lock_management_lock);
+               dev_replace->lock_owner = current->pid;
+               atomic_inc(&dev_replace->nesting_level);
+               mutex_unlock(&dev_replace->lock_management_lock);
+               return;
+       }
+
+       mutex_lock(&dev_replace->lock_management_lock);
+       if (atomic_read(&dev_replace->nesting_level) > 0 &&
+           dev_replace->lock_owner == current->pid) {
+               WARN_ON(!mutex_is_locked(&dev_replace->lock));
+               atomic_inc(&dev_replace->nesting_level);
+               mutex_unlock(&dev_replace->lock_management_lock);
+               return;
+       }
+
+       mutex_unlock(&dev_replace->lock_management_lock);
+       goto acquire_lock;
+}
+
+void btrfs_dev_replace_unlock(struct btrfs_dev_replace *dev_replace)
+{
+       WARN_ON(!mutex_is_locked(&dev_replace->lock));
+       mutex_lock(&dev_replace->lock_management_lock);
+       WARN_ON(atomic_read(&dev_replace->nesting_level) < 1);
+       WARN_ON(dev_replace->lock_owner != current->pid);
+       atomic_dec(&dev_replace->nesting_level);
+       if (atomic_read(&dev_replace->nesting_level) == 0) {
+               dev_replace->lock_owner = 0;
+               mutex_unlock(&dev_replace->lock_management_lock);
+               mutex_unlock(&dev_replace->lock);
+       } else {
+               mutex_unlock(&dev_replace->lock_management_lock);
+       }
+}
index 1fb5c89037eeb00d3ac804c6dcd013318e38ef21..20035cbbf021917f2e1c7cf7f1f786f9925de807 100644 (file)
 #if !defined(__BTRFS_DEV_REPLACE__)
 #define __BTRFS_DEV_REPLACE__
 
+struct btrfs_ioctl_dev_replace_args;
+
+int btrfs_init_dev_replace(struct btrfs_fs_info *fs_info);
+int btrfs_run_dev_replace(struct btrfs_trans_handle *trans,
+                         struct btrfs_fs_info *fs_info);
+void btrfs_after_dev_replace_commit(struct btrfs_fs_info *fs_info);
+int btrfs_dev_replace_start(struct btrfs_root *root,
+                           struct btrfs_ioctl_dev_replace_args *args);
+void btrfs_dev_replace_status(struct btrfs_fs_info *fs_info,
+                             struct btrfs_ioctl_dev_replace_args *args);
+int btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info,
+                            struct btrfs_ioctl_dev_replace_args *args);
+void btrfs_dev_replace_suspend_for_unmount(struct btrfs_fs_info *fs_info);
+int btrfs_resume_dev_replace_async(struct btrfs_fs_info *fs_info);
+int btrfs_dev_replace_is_ongoing(struct btrfs_dev_replace *dev_replace);
+void btrfs_dev_replace_lock(struct btrfs_dev_replace *dev_replace);
+void btrfs_dev_replace_unlock(struct btrfs_dev_replace *dev_replace);
+
 static inline void btrfs_dev_replace_stats_inc(atomic64_t *stat_value)
 {
        atomic64_inc(stat_value);
index 731e2875ab93900b0f623797642f371b2f042e2f..62006ba0271931bef31637e8b0059d0f3da2d51c 100644 (file)
@@ -123,6 +123,48 @@ struct btrfs_ioctl_scrub_args {
        __u64 unused[(1024-32-sizeof(struct btrfs_scrub_progress))/8];
 };
 
+#define BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_ALWAYS   0
+#define BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_AVOID    1
+struct btrfs_ioctl_dev_replace_start_params {
+       __u64 srcdevid; /* in, if 0, use srcdev_name instead */
+       __u8 srcdev_name[BTRFS_PATH_NAME_MAX + 1];      /* in */
+       __u8 tgtdev_name[BTRFS_PATH_NAME_MAX + 1];      /* in */
+       __u64 cont_reading_from_srcdev_mode;    /* in, see #define
+                                                * above */
+};
+
+#define BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED    0
+#define BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED          1
+#define BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED         2
+#define BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED         3
+#define BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED                4
+struct btrfs_ioctl_dev_replace_status_params {
+       __u64 replace_state;    /* out, see #define above */
+       __u64 progress_1000;    /* out, 0 <= x <= 1000 */
+       __u64 time_started;     /* out, seconds since 1-Jan-1970 */
+       __u64 time_stopped;     /* out, seconds since 1-Jan-1970 */
+       __u64 num_write_errors; /* out */
+       __u64 num_uncorrectable_read_errors;    /* out */
+};
+
+#define BTRFS_IOCTL_DEV_REPLACE_CMD_START                      0
+#define BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS                     1
+#define BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL                     2
+#define BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR                        0
+#define BTRFS_IOCTL_DEV_REPLACE_RESULT_NOT_STARTED             1
+#define BTRFS_IOCTL_DEV_REPLACE_RESULT_ALREADY_STARTED         2
+struct btrfs_ioctl_dev_replace_args {
+       __u64 cmd;      /* in */
+       __u64 result;   /* out */
+
+       union {
+               struct btrfs_ioctl_dev_replace_start_params start;
+               struct btrfs_ioctl_dev_replace_status_params status;
+       };      /* in/out */
+
+       __u64 spare[64];
+};
+
 #define BTRFS_DEVICE_PATH_NAME_MAX 1024
 struct btrfs_ioctl_dev_info_args {
        __u64 devid;                            /* in/out */
@@ -453,4 +495,7 @@ struct btrfs_ioctl_send_args {
                               struct btrfs_ioctl_qgroup_limit_args)
 #define BTRFS_IOC_GET_DEV_STATS _IOWR(BTRFS_IOCTL_MAGIC, 52, \
                                      struct btrfs_ioctl_get_dev_stats)
+#define BTRFS_IOC_DEV_REPLACE _IOWR(BTRFS_IOCTL_MAGIC, 53, \
+                                   struct btrfs_ioctl_dev_replace_args)
+
 #endif
index 415862885b670daf31c3b57332604e8f8b210494..5777e6a9aab17f3c0e9b2d5c620ed4817dd3dac6 100644 (file)
@@ -1537,6 +1537,53 @@ error_undo:
        goto error_brelse;
 }
 
+void btrfs_rm_dev_replace_srcdev(struct btrfs_fs_info *fs_info,
+                                struct btrfs_device *srcdev)
+{
+       WARN_ON(!mutex_is_locked(&fs_info->fs_devices->device_list_mutex));
+       list_del_rcu(&srcdev->dev_list);
+       list_del_rcu(&srcdev->dev_alloc_list);
+       fs_info->fs_devices->num_devices--;
+       if (srcdev->missing) {
+               fs_info->fs_devices->missing_devices--;
+               fs_info->fs_devices->rw_devices++;
+       }
+       if (srcdev->can_discard)
+               fs_info->fs_devices->num_can_discard--;
+       if (srcdev->bdev)
+               fs_info->fs_devices->open_devices--;
+
+       call_rcu(&srcdev->rcu, free_device);
+}
+
+void btrfs_destroy_dev_replace_tgtdev(struct btrfs_fs_info *fs_info,
+                                     struct btrfs_device *tgtdev)
+{
+       struct btrfs_device *next_device;
+
+       WARN_ON(!tgtdev);
+       mutex_lock(&fs_info->fs_devices->device_list_mutex);
+       if (tgtdev->bdev) {
+               btrfs_scratch_superblock(tgtdev);
+               fs_info->fs_devices->open_devices--;
+       }
+       fs_info->fs_devices->num_devices--;
+       if (tgtdev->can_discard)
+               fs_info->fs_devices->num_can_discard++;
+
+       next_device = list_entry(fs_info->fs_devices->devices.next,
+                                struct btrfs_device, dev_list);
+       if (tgtdev->bdev == fs_info->sb->s_bdev)
+               fs_info->sb->s_bdev = next_device->bdev;
+       if (tgtdev->bdev == fs_info->fs_devices->latest_bdev)
+               fs_info->fs_devices->latest_bdev = next_device->bdev;
+       list_del_rcu(&tgtdev->dev_list);
+
+       call_rcu(&tgtdev->rcu, free_device);
+
+       mutex_unlock(&fs_info->fs_devices->device_list_mutex);
+}
+
 int btrfs_find_device_by_path(struct btrfs_root *root, char *device_path,
                              struct btrfs_device **device)
 {
@@ -1931,6 +1978,98 @@ error:
        return ret;
 }
 
+int btrfs_init_dev_replace_tgtdev(struct btrfs_root *root, char *device_path,
+                                 struct btrfs_device **device_out)
+{
+       struct request_queue *q;
+       struct btrfs_device *device;
+       struct block_device *bdev;
+       struct btrfs_fs_info *fs_info = root->fs_info;
+       struct list_head *devices;
+       struct rcu_string *name;
+       int ret = 0;
+
+       *device_out = NULL;
+       if (fs_info->fs_devices->seeding)
+               return -EINVAL;
+
+       bdev = blkdev_get_by_path(device_path, FMODE_WRITE | FMODE_EXCL,
+                                 fs_info->bdev_holder);
+       if (IS_ERR(bdev))
+               return PTR_ERR(bdev);
+
+       filemap_write_and_wait(bdev->bd_inode->i_mapping);
+
+       devices = &fs_info->fs_devices->devices;
+       list_for_each_entry(device, devices, dev_list) {
+               if (device->bdev == bdev) {
+                       ret = -EEXIST;
+                       goto error;
+               }
+       }
+
+       device = kzalloc(sizeof(*device), GFP_NOFS);
+       if (!device) {
+               ret = -ENOMEM;
+               goto error;
+       }
+
+       name = rcu_string_strdup(device_path, GFP_NOFS);
+       if (!name) {
+               kfree(device);
+               ret = -ENOMEM;
+               goto error;
+       }
+       rcu_assign_pointer(device->name, name);
+
+       q = bdev_get_queue(bdev);
+       if (blk_queue_discard(q))
+               device->can_discard = 1;
+       mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
+       device->writeable = 1;
+       device->work.func = pending_bios_fn;
+       generate_random_uuid(device->uuid);
+       device->devid = BTRFS_DEV_REPLACE_DEVID;
+       spin_lock_init(&device->io_lock);
+       device->generation = 0;
+       device->io_width = root->sectorsize;
+       device->io_align = root->sectorsize;
+       device->sector_size = root->sectorsize;
+       device->total_bytes = i_size_read(bdev->bd_inode);
+       device->disk_total_bytes = device->total_bytes;
+       device->dev_root = fs_info->dev_root;
+       device->bdev = bdev;
+       device->in_fs_metadata = 1;
+       device->is_tgtdev_for_dev_replace = 1;
+       device->mode = FMODE_EXCL;
+       set_blocksize(device->bdev, 4096);
+       device->fs_devices = fs_info->fs_devices;
+       list_add(&device->dev_list, &fs_info->fs_devices->devices);
+       fs_info->fs_devices->num_devices++;
+       fs_info->fs_devices->open_devices++;
+       if (device->can_discard)
+               fs_info->fs_devices->num_can_discard++;
+       mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
+
+       *device_out = device;
+       return ret;
+
+error:
+       blkdev_put(bdev, FMODE_EXCL);
+       return ret;
+}
+
+void btrfs_init_dev_replace_tgtdev_for_resume(struct btrfs_fs_info *fs_info,
+                                             struct btrfs_device *tgtdev)
+{
+       WARN_ON(fs_info->fs_devices->rw_devices == 0);
+       tgtdev->io_width = fs_info->dev_root->sectorsize;
+       tgtdev->io_align = fs_info->dev_root->sectorsize;
+       tgtdev->sector_size = fs_info->dev_root->sectorsize;
+       tgtdev->dev_root = fs_info->dev_root;
+       tgtdev->in_fs_metadata = 1;
+}
+
 static noinline int btrfs_update_device(struct btrfs_trans_handle *trans,
                                        struct btrfs_device *device)
 {
index 8fd5a4d8acc8683992ffdbe63d4ff710c411155d..58d79375deaf02a6caaa9a04ceccb48c7b358a03 100644 (file)
@@ -286,6 +286,8 @@ struct btrfs_device *btrfs_find_device(struct btrfs_fs_info *fs_info, u64 devid,
                                       u8 *uuid, u8 *fsid);
 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
 int btrfs_init_new_device(struct btrfs_root *root, char *path);
+int btrfs_init_dev_replace_tgtdev(struct btrfs_root *root, char *device_path,
+                                 struct btrfs_device **device_out);
 int btrfs_balance(struct btrfs_balance_control *bctl,
                  struct btrfs_ioctl_balance_args *bargs);
 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
@@ -302,6 +304,12 @@ int btrfs_get_dev_stats(struct btrfs_root *root,
 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans,
                        struct btrfs_fs_info *fs_info);
+void btrfs_rm_dev_replace_srcdev(struct btrfs_fs_info *fs_info,
+                                struct btrfs_device *srcdev);
+void btrfs_destroy_dev_replace_tgtdev(struct btrfs_fs_info *fs_info,
+                                     struct btrfs_device *tgtdev);
+void btrfs_init_dev_replace_tgtdev_for_resume(struct btrfs_fs_info *fs_info,
+                                             struct btrfs_device *tgtdev);
 int btrfs_scratch_superblock(struct btrfs_device *device);
 
 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,