Btrfs: Add support for online device removal
[deliverable/linux.git] / fs / btrfs / volumes.c
index b38187573108e4263c02552c62746a790a2308a3..55da5f0c56e3dddd9a14d7ac633b46d95135fef6 100644 (file)
@@ -45,6 +45,16 @@ struct map_lookup {
 static DEFINE_MUTEX(uuid_mutex);
 static LIST_HEAD(fs_uuids);
 
+void btrfs_lock_volumes(void)
+{
+       mutex_lock(&uuid_mutex);
+}
+
+void btrfs_unlock_volumes(void)
+{
+       mutex_unlock(&uuid_mutex);
+}
+
 int btrfs_cleanup_fs_uuids(void)
 {
        struct btrfs_fs_devices *fs_devices;
@@ -193,12 +203,14 @@ int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
                        ret = PTR_ERR(bdev);
                        goto fail;
                }
+               set_blocksize(bdev, 4096);
                if (device->devid == fs_devices->latest_devid)
                        fs_devices->latest_bdev = bdev;
                if (device->devid == fs_devices->lowest_devid) {
                        fs_devices->lowest_bdev = bdev;
                }
                device->bdev = bdev;
+
        }
        mutex_unlock(&uuid_mutex);
        return 0;
@@ -393,6 +405,9 @@ int btrfs_free_dev_extent(struct btrfs_trans_handle *trans,
        struct btrfs_path *path;
        struct btrfs_root *root = device->dev_root;
        struct btrfs_key key;
+       struct btrfs_key found_key;
+       struct extent_buffer *leaf = NULL;
+       struct btrfs_dev_extent *extent = NULL;
 
        path = btrfs_alloc_path();
        if (!path)
@@ -403,8 +418,25 @@ int btrfs_free_dev_extent(struct btrfs_trans_handle *trans,
        key.type = BTRFS_DEV_EXTENT_KEY;
 
        ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
+       if (ret > 0) {
+               ret = btrfs_previous_item(root, path, key.objectid,
+                                         BTRFS_DEV_EXTENT_KEY);
+               BUG_ON(ret);
+               leaf = path->nodes[0];
+               btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+               extent = btrfs_item_ptr(leaf, path->slots[0],
+                                       struct btrfs_dev_extent);
+               BUG_ON(found_key.offset > start || found_key.offset +
+                      btrfs_dev_extent_length(leaf, extent) < start);
+               ret = 0;
+       } else if (ret == 0) {
+               leaf = path->nodes[0];
+               extent = btrfs_item_ptr(leaf, path->slots[0],
+                                       struct btrfs_dev_extent);
+       }
        BUG_ON(ret);
 
+       device->bytes_used -= btrfs_dev_extent_length(leaf, extent);
        ret = btrfs_del_item(trans, root, path);
        BUG_ON(ret);
 
@@ -593,6 +625,170 @@ out:
        return ret;
 }
 
+static int btrfs_rm_dev_item(struct btrfs_root *root,
+                            struct btrfs_device *device)
+{
+       int ret;
+       struct btrfs_path *path;
+       struct block_device *bdev = device->bdev;
+       struct btrfs_device *next_dev;
+       struct btrfs_key key;
+       u64 total_bytes;
+       struct btrfs_fs_devices *fs_devices;
+       struct btrfs_trans_handle *trans;
+
+       root = root->fs_info->chunk_root;
+
+       path = btrfs_alloc_path();
+       if (!path)
+               return -ENOMEM;
+
+       trans = btrfs_start_transaction(root, 1);
+       key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
+       key.type = BTRFS_DEV_ITEM_KEY;
+       key.offset = device->devid;
+
+       ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
+       if (ret < 0)
+               goto out;
+
+       if (ret > 0) {
+               ret = -ENOENT;
+               goto out;
+       }
+
+       ret = btrfs_del_item(trans, root, path);
+       if (ret)
+               goto out;
+
+       /*
+        * at this point, the device is zero sized.  We want to
+        * remove it from the devices list and zero out the old super
+        */
+       list_del_init(&device->dev_list);
+       list_del_init(&device->dev_alloc_list);
+       fs_devices = root->fs_info->fs_devices;
+
+       next_dev = list_entry(fs_devices->devices.next, struct btrfs_device,
+                             dev_list);
+       if (bdev == fs_devices->lowest_bdev)
+               fs_devices->lowest_bdev = next_dev->bdev;
+       if (bdev == root->fs_info->sb->s_bdev)
+               root->fs_info->sb->s_bdev = next_dev->bdev;
+       if (bdev == fs_devices->latest_bdev)
+               fs_devices->latest_bdev = next_dev->bdev;
+
+       total_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
+       btrfs_set_super_total_bytes(&root->fs_info->super_copy,
+                                   total_bytes - device->total_bytes);
+
+       total_bytes = btrfs_super_num_devices(&root->fs_info->super_copy);
+       btrfs_set_super_num_devices(&root->fs_info->super_copy,
+                                   total_bytes - 1);
+out:
+       btrfs_free_path(path);
+       btrfs_commit_transaction(trans, root);
+       return ret;
+}
+
+int btrfs_rm_device(struct btrfs_root *root, char *device_path)
+{
+       struct btrfs_device *device;
+       struct block_device *bdev;
+       struct buffer_head *bh;
+       struct btrfs_super_block *disk_super;
+       u64 all_avail;
+       u64 devid;
+       int ret = 0;
+
+       mutex_lock(&root->fs_info->fs_mutex);
+       mutex_lock(&uuid_mutex);
+
+       all_avail = root->fs_info->avail_data_alloc_bits |
+               root->fs_info->avail_system_alloc_bits |
+               root->fs_info->avail_metadata_alloc_bits;
+
+       if ((all_avail & BTRFS_BLOCK_GROUP_RAID10) &&
+           root->fs_info->fs_devices->num_devices <= 4) {
+               printk("btrfs: unable to go below four devices on raid10\n");
+               ret = -EINVAL;
+               goto out;
+       }
+
+       if ((all_avail & BTRFS_BLOCK_GROUP_RAID1) &&
+           root->fs_info->fs_devices->num_devices <= 2) {
+               printk("btrfs: unable to go below two devices on raid1\n");
+               ret = -EINVAL;
+               goto out;
+       }
+
+       bdev = open_bdev_excl(device_path, 0, root->fs_info->bdev_holder);
+       if (IS_ERR(bdev)) {
+               ret = PTR_ERR(bdev);
+               goto out;
+       }
+
+       bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
+       if (!bh) {
+               ret = -EIO;
+               goto error_close;
+       }
+       disk_super = (struct btrfs_super_block *)bh->b_data;
+       if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
+           sizeof(disk_super->magic))) {
+               ret = -ENOENT;
+               goto error_brelse;
+       }
+       if (memcmp(disk_super->fsid, root->fs_info->fsid, BTRFS_FSID_SIZE)) {
+               ret = -ENOENT;
+               goto error_brelse;
+       }
+       devid = le64_to_cpu(disk_super->dev_item.devid);
+       device = btrfs_find_device(root, devid, NULL);
+       if (!device) {
+               ret = -ENOENT;
+               goto error_brelse;
+       }
+
+       root->fs_info->fs_devices->num_devices--;
+
+       ret = btrfs_shrink_device(device, 0);
+       if (ret)
+               goto error_brelse;
+
+
+       ret = btrfs_rm_dev_item(root->fs_info->chunk_root, device);
+       if (ret)
+               goto error_brelse;
+
+       /* make sure this device isn't detected as part of the FS anymore */
+       memset(&disk_super->magic, 0, sizeof(disk_super->magic));
+       set_buffer_dirty(bh);
+       sync_dirty_buffer(bh);
+
+       brelse(bh);
+
+       /* one close for the device struct or super_block */
+       close_bdev_excl(device->bdev);
+
+       /* one close for us */
+       close_bdev_excl(device->bdev);
+
+       kfree(device->name);
+       kfree(device);
+       ret = 0;
+       goto out;
+
+error_brelse:
+       brelse(bh);
+error_close:
+       close_bdev_excl(bdev);
+out:
+       mutex_unlock(&uuid_mutex);
+       mutex_unlock(&root->fs_info->fs_mutex);
+       return ret;
+}
+
 int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
 {
        struct btrfs_trans_handle *trans;
@@ -831,13 +1027,17 @@ int btrfs_relocate_chunk(struct btrfs_root *root,
        em = lookup_extent_mapping(em_tree, chunk_offset, 1);
        spin_unlock(&em_tree->lock);
 
-       BUG_ON(em->start > chunk_offset || em->start + em->len < chunk_offset);
+       BUG_ON(em->start > chunk_offset ||
+              em->start + em->len < chunk_offset);
        map = (struct map_lookup *)em->bdev;
 
        for (i = 0; i < map->num_stripes; i++) {
                ret = btrfs_free_dev_extent(trans, map->stripes[i].dev,
                                            map->stripes[i].physical);
                BUG_ON(ret);
+
+               ret = btrfs_update_device(trans, map->stripes[i].dev);
+               BUG_ON(ret);
        }
        ret = btrfs_free_chunk(trans, root, chunk_tree, chunk_objectid,
                               chunk_offset);
@@ -847,11 +1047,8 @@ int btrfs_relocate_chunk(struct btrfs_root *root,
        if (map->type & BTRFS_BLOCK_GROUP_SYSTEM) {
                ret = btrfs_del_sys_chunk(root, chunk_objectid, chunk_offset);
                BUG_ON(ret);
-               goto out;
        }
 
-
-
        spin_lock(&em_tree->lock);
        remove_extent_mapping(em_tree, em);
        kfree(map);
@@ -861,7 +1058,6 @@ int btrfs_relocate_chunk(struct btrfs_root *root,
        free_extent_map(em);
        spin_unlock(&em_tree->lock);
 
-out:
        /* once for us */
        free_extent_map(em);
 
@@ -1449,7 +1645,7 @@ again:
                return 0;
 
        if (!em) {
-               printk("unable to find logical %Lu\n", logical);
+               printk("unable to find logical %Lu len %Lu\n", logical, *length);
                BUG();
        }
 
@@ -1712,6 +1908,7 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
 
        logical = key->offset;
        length = btrfs_chunk_length(leaf, chunk);
+
        spin_lock(&map_tree->map_tree.lock);
        em = lookup_extent_mapping(&map_tree->map_tree, logical, 1);
        spin_unlock(&map_tree->map_tree.lock);
@@ -1845,7 +2042,7 @@ int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf)
 int btrfs_read_sys_array(struct btrfs_root *root)
 {
        struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
-       struct extent_buffer *sb = root->fs_info->sb_buffer;
+       struct extent_buffer *sb;
        struct btrfs_disk_key *disk_key;
        struct btrfs_chunk *chunk;
        u8 *ptr;
@@ -1857,6 +2054,12 @@ int btrfs_read_sys_array(struct btrfs_root *root)
        u32 cur;
        struct btrfs_key key;
 
+       sb = btrfs_find_create_tree_block(root, BTRFS_SUPER_INFO_OFFSET,
+                                         BTRFS_SUPER_INFO_SIZE);
+       if (!sb)
+               return -ENOMEM;
+       btrfs_set_buffer_uptodate(sb);
+       write_extent_buffer(sb, super_copy, 0, BTRFS_SUPER_INFO_SIZE);
        array_size = btrfs_super_sys_array_size(super_copy);
 
        ptr = super_copy->sys_chunk_array;
@@ -1867,8 +2070,7 @@ int btrfs_read_sys_array(struct btrfs_root *root)
                disk_key = (struct btrfs_disk_key *)ptr;
                btrfs_disk_key_to_cpu(&key, disk_key);
 
-               len = sizeof(*disk_key);
-               ptr += len;
+               len = sizeof(*disk_key); ptr += len;
                sb_ptr += len;
                cur += len;
 
@@ -1887,6 +2089,7 @@ int btrfs_read_sys_array(struct btrfs_root *root)
                sb_ptr += len;
                cur += len;
        }
+       free_extent_buffer(sb);
        return ret;
 }
 
This page took 0.105065 seconds and 5 git commands to generate.