btrfs-progs: Remove unused parameter
[platform/upstream/btrfs-progs.git] / volumes.c
index 990300f..9ebe52a 100644 (file)
--- a/volumes.c
+++ b/volumes.c
 #include "utils.h"
 #include "kernel-lib/raid56.h"
 
+const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES] = {
+       [BTRFS_RAID_RAID10] = {
+               .sub_stripes    = 2,
+               .dev_stripes    = 1,
+               .devs_max       = 0,    /* 0 == as many as possible */
+               .devs_min       = 4,
+               .tolerated_failures = 1,
+               .devs_increment = 2,
+               .ncopies        = 2,
+       },
+       [BTRFS_RAID_RAID1] = {
+               .sub_stripes    = 1,
+               .dev_stripes    = 1,
+               .devs_max       = 2,
+               .devs_min       = 2,
+               .tolerated_failures = 1,
+               .devs_increment = 2,
+               .ncopies        = 2,
+       },
+       [BTRFS_RAID_DUP] = {
+               .sub_stripes    = 1,
+               .dev_stripes    = 2,
+               .devs_max       = 1,
+               .devs_min       = 1,
+               .tolerated_failures = 0,
+               .devs_increment = 1,
+               .ncopies        = 2,
+       },
+       [BTRFS_RAID_RAID0] = {
+               .sub_stripes    = 1,
+               .dev_stripes    = 1,
+               .devs_max       = 0,
+               .devs_min       = 2,
+               .tolerated_failures = 0,
+               .devs_increment = 1,
+               .ncopies        = 1,
+       },
+       [BTRFS_RAID_SINGLE] = {
+               .sub_stripes    = 1,
+               .dev_stripes    = 1,
+               .devs_max       = 1,
+               .devs_min       = 1,
+               .tolerated_failures = 0,
+               .devs_increment = 1,
+               .ncopies        = 1,
+       },
+       [BTRFS_RAID_RAID5] = {
+               .sub_stripes    = 1,
+               .dev_stripes    = 1,
+               .devs_max       = 0,
+               .devs_min       = 2,
+               .tolerated_failures = 1,
+               .devs_increment = 1,
+               .ncopies        = 2,
+       },
+       [BTRFS_RAID_RAID6] = {
+               .sub_stripes    = 1,
+               .dev_stripes    = 1,
+               .devs_max       = 0,
+               .devs_min       = 3,
+               .tolerated_failures = 2,
+               .devs_increment = 1,
+               .ncopies        = 3,
+       },
+};
+
 struct stripe {
        struct btrfs_device *dev;
        u64 physical;
@@ -58,10 +124,8 @@ static struct btrfs_device *__find_device(struct list_head *head, u64 devid,
                                          u8 *uuid)
 {
        struct btrfs_device *dev;
-       struct list_head *cur;
 
-       list_for_each(cur, head) {
-               dev = list_entry(cur, struct btrfs_device, dev_list);
+       list_for_each_entry(dev, head, dev_list) {
                if (dev->devid == devid &&
                    !memcmp(dev->uuid, uuid, BTRFS_UUID_SIZE)) {
                        return dev;
@@ -72,11 +136,9 @@ static struct btrfs_device *__find_device(struct list_head *head, u64 devid,
 
 static struct btrfs_fs_devices *find_fsid(u8 *fsid)
 {
-       struct list_head *cur;
        struct btrfs_fs_devices *fs_devices;
 
-       list_for_each(cur, &fs_uuids) {
-               fs_devices = list_entry(cur, struct btrfs_fs_devices, list);
+       list_for_each_entry(fs_devices, &fs_uuids, list) {
                if (memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE) == 0)
                        return fs_devices;
        }
@@ -138,7 +200,21 @@ static int device_list_add(const char *path,
                list_add(&device->dev_list, &fs_devices->devices);
                device->fs_devices = fs_devices;
        } else if (!device->name || strcmp(device->name, path)) {
-               char *name = strdup(path);
+               char *name;
+
+               /*
+                * The existing device has newer generation, so this one could
+                * be a stale one, don't add it.
+                */
+               if (found_transid < device->generation) {
+                       warning(
+       "adding device %s gen %llu but found an existing device %s gen %llu",
+                               path, found_transid, device->name,
+                               device->generation);
+                       return -EEXIST;
+               }
+
+               name = strdup(path);
                 if (!name)
                         return -ENOMEM;
                 kfree(device->name);
@@ -171,8 +247,8 @@ again:
                                    struct btrfs_device, dev_list);
                if (device->fd != -1) {
                        if (fsync(device->fd) == -1) {
-                               warning("fsync on device %llu failed: %s",
-                                       device->devid, strerror(errno));
+                               warning("fsync on device %llu failed: %m",
+                                       device->devid);
                                ret = -errno;
                        }
                        if (posix_fadvise(device->fd, 0, 0, POSIX_FADV_DONTNEED))
@@ -220,13 +296,10 @@ void btrfs_close_all_devices(void)
 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, int flags)
 {
        int fd;
-       struct list_head *head = &fs_devices->devices;
-       struct list_head *cur;
        struct btrfs_device *device;
        int ret;
 
-       list_for_each(cur, head) {
-               device = list_entry(cur, struct btrfs_device, dev_list);
+       list_for_each_entry(device, &fs_devices->devices, dev_list) {
                if (!device->name) {
                        printk("no name for device %llu, skip it now\n", device->devid);
                        continue;
@@ -235,8 +308,7 @@ int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, int flags)
                fd = open(device->name, flags);
                if (fd < 0) {
                        ret = -errno;
-                       error("cannot open device '%s': %s", device->name,
-                                       strerror(errno));
+                       error("cannot open device '%s': %m", device->name);
                        goto fail;
                }
 
@@ -302,9 +374,9 @@ int btrfs_scan_one_device(int fd, const char *path,
  * But if we don't find suitable free space, it is used to store the size of
  * the max free space.
  */
-static int find_free_dev_extent_start(struct btrfs_trans_handle *trans,
-                              struct btrfs_device *device, u64 num_bytes,
-                              u64 search_start, u64 *start, u64 *len)
+static int find_free_dev_extent_start(struct btrfs_device *device,
+                                     u64 num_bytes, u64 search_start,
+                                     u64 *start, u64 *len)
 {
        struct btrfs_key key;
        struct btrfs_root *root = device->dev_root;
@@ -443,20 +515,17 @@ out:
        return ret;
 }
 
-int find_free_dev_extent(struct btrfs_trans_handle *trans,
-                        struct btrfs_device *device, u64 num_bytes,
-                        u64 *start)
+static int find_free_dev_extent(struct btrfs_device *device, u64 num_bytes,
+                               u64 *start, u64 *len)
 {
        /* FIXME use last free of some kind */
-       return find_free_dev_extent_start(trans, device,
-                                         num_bytes, 0, start, NULL);
+       return find_free_dev_extent_start(device, num_bytes, 0, start, len);
 }
 
 static int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
                                  struct btrfs_device *device,
-                                 u64 chunk_tree, u64 chunk_objectid,
-                                 u64 chunk_offset,
-                                 u64 num_bytes, u64 *start, int convert)
+                                 u64 chunk_offset, u64 num_bytes, u64 *start,
+                                 int convert)
 {
        int ret;
        struct btrfs_path *path;
@@ -474,8 +543,7 @@ static int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
         * is responsible to make sure it's free.
         */
        if (!convert) {
-               ret = find_free_dev_extent(trans, device, num_bytes,
-                                          start);
+               ret = find_free_dev_extent(device, num_bytes, start, NULL);
                if (ret)
                        goto err;
        }
@@ -490,8 +558,9 @@ static int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
        leaf = path->nodes[0];
        extent = btrfs_item_ptr(leaf, path->slots[0],
                                struct btrfs_dev_extent);
-       btrfs_set_dev_extent_chunk_tree(leaf, extent, chunk_tree);
-       btrfs_set_dev_extent_chunk_objectid(leaf, extent, chunk_objectid);
+       btrfs_set_dev_extent_chunk_tree(leaf, extent, BTRFS_CHUNK_TREE_OBJECTID);
+       btrfs_set_dev_extent_chunk_objectid(leaf, extent,
+                                           BTRFS_FIRST_CHUNK_TREE_OBJECTID);
        btrfs_set_dev_extent_chunk_offset(leaf, extent, chunk_offset);
 
        write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
@@ -505,8 +574,9 @@ err:
        return ret;
 }
 
-static int find_next_chunk(struct btrfs_root *root, u64 objectid, u64 *offset)
+static int find_next_chunk(struct btrfs_fs_info *fs_info, u64 *offset)
 {
+       struct btrfs_root *root = fs_info->chunk_root;
        struct btrfs_path *path;
        int ret;
        struct btrfs_key key;
@@ -517,7 +587,7 @@ static int find_next_chunk(struct btrfs_root *root, u64 objectid, u64 *offset)
        if (!path)
                return -ENOMEM;
 
-       key.objectid = objectid;
+       key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID;
        key.offset = (u64)-1;
        key.type = BTRFS_CHUNK_ITEM_KEY;
 
@@ -533,7 +603,7 @@ static int find_next_chunk(struct btrfs_root *root, u64 objectid, u64 *offset)
        } else {
                btrfs_item_key_to_cpu(path->nodes[0], &found_key,
                                      path->slots[0]);
-               if (found_key.objectid != objectid)
+               if (found_key.objectid != BTRFS_FIRST_CHUNK_TREE_OBJECTID)
                        *offset = 0;
                else {
                        chunk = btrfs_item_ptr(path->nodes[0], path->slots[0],
@@ -585,7 +655,7 @@ error:
  * the btrfs_device struct should be fully filled in
  */
 int btrfs_add_device(struct btrfs_trans_handle *trans,
-                    struct btrfs_root *root,
+                    struct btrfs_fs_info *fs_info,
                     struct btrfs_device *device)
 {
        int ret;
@@ -593,11 +663,10 @@ int btrfs_add_device(struct btrfs_trans_handle *trans,
        struct btrfs_dev_item *dev_item;
        struct extent_buffer *leaf;
        struct btrfs_key key;
+       struct btrfs_root *root = fs_info->chunk_root;
        unsigned long ptr;
        u64 free_devid = 0;
 
-       root = root->fs_info->chunk_root;
-
        path = btrfs_alloc_path();
        if (!path)
                return -ENOMEM;
@@ -635,7 +704,7 @@ int btrfs_add_device(struct btrfs_trans_handle *trans,
        ptr = (unsigned long)btrfs_device_uuid(dev_item);
        write_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE);
        ptr = (unsigned long)btrfs_device_fsid(dev_item);
-       write_extent_buffer(leaf, root->fs_info->fsid, ptr, BTRFS_UUID_SIZE);
+       write_extent_buffer(leaf, fs_info->fsid, ptr, BTRFS_UUID_SIZE);
        btrfs_mark_buffer_dirty(leaf);
        ret = 0;
 
@@ -690,11 +759,10 @@ out:
        return ret;
 }
 
-int btrfs_add_system_chunk(struct btrfs_root *root,
-                          struct btrfs_key *key,
+int btrfs_add_system_chunk(struct btrfs_fs_info *fs_info, struct btrfs_key *key,
                           struct btrfs_chunk *chunk, int item_size)
 {
-       struct btrfs_super_block *super_copy = root->fs_info->super_copy;
+       struct btrfs_super_block *super_copy = fs_info->super_copy;
        struct btrfs_disk_key disk_key;
        u32 array_size;
        u8 *ptr;
@@ -824,7 +892,7 @@ error:
        return ret;
 }
 
-#define BTRFS_MAX_DEVS(r) ((BTRFS_LEAF_DATA_SIZE(r)            \
+#define BTRFS_MAX_DEVS(info) ((BTRFS_LEAF_DATA_SIZE(info)      \
                        - sizeof(struct btrfs_item)             \
                        - sizeof(struct btrfs_chunk))           \
                        / sizeof(struct btrfs_stripe) + 1)
@@ -835,11 +903,11 @@ error:
                                / sizeof(struct btrfs_stripe) + 1)
 
 int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
-                     struct btrfs_root *extent_root, u64 *start,
+                     struct btrfs_fs_info *info, u64 *start,
                      u64 *num_bytes, u64 type)
 {
        u64 dev_offset;
-       struct btrfs_fs_info *info = extent_root->fs_info;
+       struct btrfs_root *extent_root = info->extent_root;
        struct btrfs_root *chunk_root = info->chunk_root;
        struct btrfs_stripe *stripes;
        struct btrfs_device *device = NULL;
@@ -870,10 +938,7 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
                return -ENOSPC;
        }
 
-       if (type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
-                   BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6 |
-                   BTRFS_BLOCK_GROUP_RAID10 |
-                   BTRFS_BLOCK_GROUP_DUP)) {
+       if (type & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
                if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
                        calc_size = SZ_8M;
                        max_chunk_size = calc_size * 2;
@@ -883,12 +948,12 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
                        calc_size = SZ_1G;
                        max_chunk_size = 10 * calc_size;
                        min_stripe_size = SZ_64M;
-                       max_stripes = BTRFS_MAX_DEVS(chunk_root);
+                       max_stripes = BTRFS_MAX_DEVS(info);
                } else if (type & BTRFS_BLOCK_GROUP_METADATA) {
                        calc_size = SZ_1G;
                        max_chunk_size = 4 * calc_size;
                        min_stripe_size = SZ_32M;
-                       max_stripes = BTRFS_MAX_DEVS(chunk_root);
+                       max_stripes = BTRFS_MAX_DEVS(info);
                }
        }
        if (type & BTRFS_BLOCK_GROUP_RAID1) {
@@ -1000,8 +1065,7 @@ again:
                }
                return -ENOSPC;
        }
-       ret = find_next_chunk(chunk_root, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
-                             &offset);
+       ret = find_next_chunk(info, &offset);
        if (ret)
                return ret;
        key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID;
@@ -1033,15 +1097,15 @@ again:
                    (index == num_stripes - 1))
                        list_move_tail(&device->dev_list, dev_list);
 
-               ret = btrfs_alloc_dev_extent(trans, device,
-                            info->chunk_root->root_key.objectid,
-                            BTRFS_FIRST_CHUNK_TREE_OBJECTID, key.offset,
+               ret = btrfs_alloc_dev_extent(trans, device, key.offset,
                             calc_size, &dev_offset, 0);
-               BUG_ON(ret);
+               if (ret < 0)
+                       goto out_chunk_map;
 
                device->bytes_used += calc_size;
                ret = btrfs_update_device(trans, device);
-               BUG_ON(ret);
+               if (ret < 0)
+                       goto out_chunk_map;
 
                map->stripes[index].dev = device;
                map->stripes[index].physical = dev_offset;
@@ -1080,16 +1144,24 @@ again:
        map->ce.size = *num_bytes;
 
        ret = insert_cache_extent(&info->mapping_tree.cache_tree, &map->ce);
-       BUG_ON(ret);
+       if (ret < 0)
+               goto out_chunk_map;
 
        if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
-               ret = btrfs_add_system_chunk(chunk_root, &key,
+               ret = btrfs_add_system_chunk(info, &key,
                                    chunk, btrfs_chunk_item_size(num_stripes));
-               BUG_ON(ret);
+               if (ret < 0)
+                       goto out_chunk;
        }
 
        kfree(chunk);
        return ret;
+
+out_chunk_map:
+       kfree(map);
+out_chunk:
+       kfree(chunk);
+       return ret;
 }
 
 /*
@@ -1101,11 +1173,11 @@ again:
  * occupied.
  */
 int btrfs_alloc_data_chunk(struct btrfs_trans_handle *trans,
-                          struct btrfs_root *extent_root, u64 *start,
+                          struct btrfs_fs_info *info, u64 *start,
                           u64 num_bytes, u64 type, int convert)
 {
        u64 dev_offset;
-       struct btrfs_fs_info *info = extent_root->fs_info;
+       struct btrfs_root *extent_root = info->extent_root;
        struct btrfs_root *chunk_root = info->chunk_root;
        struct btrfs_stripe *stripes;
        struct btrfs_device *device = NULL;
@@ -1134,9 +1206,7 @@ int btrfs_alloc_data_chunk(struct btrfs_trans_handle *trans,
        } else {
                u64 tmp;
 
-               ret = find_next_chunk(chunk_root,
-                                     BTRFS_FIRST_CHUNK_TREE_OBJECTID,
-                                     &tmp);
+               ret = find_next_chunk(info, &tmp);
                key.offset = tmp;
                if (ret)
                        return ret;
@@ -1162,9 +1232,7 @@ int btrfs_alloc_data_chunk(struct btrfs_trans_handle *trans,
        while (index < num_stripes) {
                struct btrfs_stripe *stripe;
 
-               ret = btrfs_alloc_dev_extent(trans, device,
-                            info->chunk_root->root_key.objectid,
-                            BTRFS_FIRST_CHUNK_TREE_OBJECTID, key.offset,
+               ret = btrfs_alloc_dev_extent(trans, device, key.offset,
                             calc_size, &dev_offset, convert);
                BUG_ON(ret);
 
@@ -1215,8 +1283,9 @@ int btrfs_alloc_data_chunk(struct btrfs_trans_handle *trans,
        return ret;
 }
 
-int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len)
+int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len)
 {
+       struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
        struct cache_extent *ce;
        struct map_lookup *map;
        int ret;
@@ -1251,9 +1320,10 @@ int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len)
        return ret;
 }
 
-int btrfs_next_bg(struct btrfs_mapping_tree *map_tree, u64 *logical,
-                    u64 *size, u64 type)
+int btrfs_next_bg(struct btrfs_fs_info *fs_info, u64 *logical,
+                 u64 *size, u64 type)
 {
+       struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
        struct cache_extent *ce;
        struct map_lookup *map;
        u64 cur = *logical;
@@ -1279,15 +1349,18 @@ int btrfs_next_bg(struct btrfs_mapping_tree *map_tree, u64 *logical,
                        *size = ce->size;
                        return 0;
                }
+               if (!cur)
+                       ce = next_cache_extent(ce);
        }
 
        return -ENOENT;
 }
 
-int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
+int btrfs_rmap_block(struct btrfs_fs_info *fs_info,
                     u64 chunk_start, u64 physical, u64 devid,
                     u64 **logical, int *naddrs, int *stripe_len)
 {
+       struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
        struct cache_extent *ce;
        struct map_lookup *map;
        u64 *buf;
@@ -1379,20 +1452,21 @@ static void sort_parity_stripes(struct btrfs_multi_bio *bbio, u64 *raid_map)
        }
 }
 
-int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
+int btrfs_map_block(struct btrfs_fs_info *fs_info, int rw,
                    u64 logical, u64 *length,
                    struct btrfs_multi_bio **multi_ret, int mirror_num,
                    u64 **raid_map_ret)
 {
-       return __btrfs_map_block(map_tree, rw, logical, length, NULL,
+       return __btrfs_map_block(fs_info, rw, logical, length, NULL,
                                 multi_ret, mirror_num, raid_map_ret);
 }
 
-int __btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
-                   u64 logical, u64 *length, u64 *type,
-                   struct btrfs_multi_bio **multi_ret, int mirror_num,
-                   u64 **raid_map_ret)
+int __btrfs_map_block(struct btrfs_fs_info *fs_info, int rw,
+                     u64 logical, u64 *length, u64 *type,
+                     struct btrfs_multi_bio **multi_ret, int mirror_num,
+                     u64 **raid_map_ret)
 {
+       struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
        struct cache_extent *ce;
        struct map_lookup *map;
        u64 offset;
@@ -1597,17 +1671,17 @@ out:
        return 0;
 }
 
-struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid,
+struct btrfs_device *btrfs_find_device(struct btrfs_fs_info *fs_info, u64 devid,
                                       u8 *uuid, u8 *fsid)
 {
        struct btrfs_device *device;
        struct btrfs_fs_devices *cur_devices;
 
-       cur_devices = root->fs_info->fs_devices;
+       cur_devices = fs_info->fs_devices;
        while (cur_devices) {
                if (!fsid ||
                    (!memcmp(cur_devices->fsid, fsid, BTRFS_UUID_SIZE) ||
-                    root->fs_info->ignore_fsid_mismatch)) {
+                    fs_info->ignore_fsid_mismatch)) {
                        device = __find_device(&cur_devices->devices,
                                               devid, uuid);
                        if (device)
@@ -1633,11 +1707,11 @@ btrfs_find_device_by_devid(struct btrfs_fs_devices *fs_devices,
        return NULL;
 }
 
-int btrfs_chunk_readonly(struct btrfs_root *root, u64 chunk_offset)
+int btrfs_chunk_readonly(struct btrfs_fs_info *fs_info, u64 chunk_offset)
 {
        struct cache_extent *ce;
        struct map_lookup *map;
-       struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree;
+       struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
        int readonly = 0;
        int i;
 
@@ -1646,7 +1720,7 @@ int btrfs_chunk_readonly(struct btrfs_root *root, u64 chunk_offset)
         * corresponding chunk, we will rebuild it later
         */
        ce = search_cache_extent(&map_tree->cache_tree, chunk_offset);
-       if (!root->fs_info->is_chunk_recover)
+       if (!fs_info->is_chunk_recover)
                BUG_ON(!ce);
        else
                return 0;
@@ -1676,7 +1750,7 @@ static struct btrfs_device *fill_missing_device(u64 devid)
  * slot == -1: SYSTEM chunk
  * return -EIO on error, otherwise return 0
  */
-int btrfs_check_chunk_valid(struct btrfs_root *root,
+int btrfs_check_chunk_valid(struct btrfs_fs_info *fs_info,
                            struct extent_buffer *leaf,
                            struct btrfs_chunk *chunk,
                            int slot, u64 logical)
@@ -1687,7 +1761,7 @@ int btrfs_check_chunk_valid(struct btrfs_root *root,
        u16 sub_stripes;
        u64 type;
        u32 chunk_ondisk_size;
-       u32 sectorsize = root->fs_info->sectorsize;
+       u32 sectorsize = fs_info->sectorsize;
 
        length = btrfs_chunk_length(leaf, chunk);
        stripe_len = btrfs_chunk_stripe_len(leaf, chunk);
@@ -1703,7 +1777,7 @@ int btrfs_check_chunk_valid(struct btrfs_root *root,
                return -EIO;
        }
        if (btrfs_chunk_sector_size(leaf, chunk) != sectorsize) {
-               error("invalid chunk sectorsize %llu", 
+               error("invalid chunk sectorsize %llu",
                      (unsigned long long)btrfs_chunk_sector_size(leaf, chunk));
                return -EIO;
        }
@@ -1779,11 +1853,11 @@ int btrfs_check_chunk_valid(struct btrfs_root *root,
  *
  * For sys chunk in superblock, pass -1 to indicate sys chunk.
  */
-static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
+static int read_one_chunk(struct btrfs_fs_info *fs_info, struct btrfs_key *key,
                          struct extent_buffer *leaf,
                          struct btrfs_chunk *chunk, int slot)
 {
-       struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree;
+       struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
        struct map_lookup *map;
        struct cache_extent *ce;
        u64 logical;
@@ -1798,7 +1872,7 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
        length = btrfs_chunk_length(leaf, chunk);
        num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
        /* Validation check */
-       ret = btrfs_check_chunk_valid(root, leaf, chunk, slot, logical);
+       ret = btrfs_check_chunk_valid(fs_info, leaf, chunk, slot, logical);
        if (ret) {
                error("%s checksums match, but it has an invalid chunk, %s",
                      (slot == -1) ? "Superblock" : "Metadata",
@@ -1834,14 +1908,14 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
                read_extent_buffer(leaf, uuid, (unsigned long)
                                   btrfs_stripe_dev_uuid_nr(chunk, i),
                                   BTRFS_UUID_SIZE);
-               map->stripes[i].dev = btrfs_find_device(root, devid, uuid,
+               map->stripes[i].dev = btrfs_find_device(fs_info, devid, uuid,
                                                        NULL);
                if (!map->stripes[i].dev) {
                        map->stripes[i].dev = fill_missing_device(devid);
                        printf("warning, device %llu is missing\n",
                               (unsigned long long)devid);
                        list_add(&map->stripes[i].dev->dev_list,
-                                &root->fs_info->fs_devices->devices);
+                                &fs_info->fs_devices->devices);
                }
 
        }
@@ -1871,12 +1945,12 @@ static int fill_device_from_item(struct extent_buffer *leaf,
        return 0;
 }
 
-static int open_seed_devices(struct btrfs_root *root, u8 *fsid)
+static int open_seed_devices(struct btrfs_fs_info *fs_info, u8 *fsid)
 {
        struct btrfs_fs_devices *fs_devices;
        int ret;
 
-       fs_devices = root->fs_info->fs_devices->seed;
+       fs_devices = fs_info->fs_devices->seed;
        while (fs_devices) {
                if (!memcmp(fs_devices->fsid, fsid, BTRFS_UUID_SIZE)) {
                        ret = 0;
@@ -1902,13 +1976,13 @@ static int open_seed_devices(struct btrfs_root *root, u8 *fsid)
        if (ret)
                goto out;
 
-       fs_devices->seed = root->fs_info->fs_devices->seed;
-       root->fs_info->fs_devices->seed = fs_devices;
+       fs_devices->seed = fs_info->fs_devices->seed;
+       fs_info->fs_devices->seed = fs_devices;
 out:
        return ret;
 }
 
-static int read_one_dev(struct btrfs_root *root,
+static int read_one_dev(struct btrfs_fs_info *fs_info,
                        struct extent_buffer *leaf,
                        struct btrfs_dev_item *dev_item)
 {
@@ -1926,30 +2000,30 @@ static int read_one_dev(struct btrfs_root *root,
                           (unsigned long)btrfs_device_fsid(dev_item),
                           BTRFS_UUID_SIZE);
 
-       if (memcmp(fs_uuid, root->fs_info->fsid, BTRFS_UUID_SIZE)) {
-               ret = open_seed_devices(root, fs_uuid);
+       if (memcmp(fs_uuid, fs_info->fsid, BTRFS_UUID_SIZE)) {
+               ret = open_seed_devices(fs_info, fs_uuid);
                if (ret)
                        return ret;
        }
 
-       device = btrfs_find_device(root, devid, dev_uuid, fs_uuid);
+       device = btrfs_find_device(fs_info, devid, dev_uuid, fs_uuid);
        if (!device) {
                device = kzalloc(sizeof(*device), GFP_NOFS);
                if (!device)
                        return -ENOMEM;
                device->fd = -1;
                list_add(&device->dev_list,
-                        &root->fs_info->fs_devices->devices);
+                        &fs_info->fs_devices->devices);
        }
 
        fill_device_from_item(leaf, dev_item, device);
-       device->dev_root = root->fs_info->dev_root;
+       device->dev_root = fs_info->dev_root;
        return ret;
 }
 
-int btrfs_read_sys_array(struct btrfs_root *root)
+int btrfs_read_sys_array(struct btrfs_fs_info *fs_info)
 {
-       struct btrfs_super_block *super_copy = root->fs_info->super_copy;
+       struct btrfs_super_block *super_copy = fs_info->super_copy;
        struct extent_buffer *sb;
        struct btrfs_disk_key *disk_key;
        struct btrfs_chunk *chunk;
@@ -1962,9 +2036,12 @@ int btrfs_read_sys_array(struct btrfs_root *root)
        u32 cur_offset;
        struct btrfs_key key;
 
-       sb = btrfs_find_create_tree_block(root->fs_info,
-                                         BTRFS_SUPER_INFO_OFFSET,
-                                         BTRFS_SUPER_INFO_SIZE);
+       if (fs_info->nodesize < BTRFS_SUPER_INFO_SIZE) {
+               printf("ERROR: nodesize %u too small to read superblock\n",
+                               fs_info->nodesize);
+               return -EINVAL;
+       }
+       sb = btrfs_find_create_tree_block(fs_info, BTRFS_SUPER_INFO_OFFSET);
        if (!sb)
                return -ENOMEM;
        btrfs_set_buffer_uptodate(sb);
@@ -2010,7 +2087,7 @@ int btrfs_read_sys_array(struct btrfs_root *root)
                        if (cur_offset + len > array_size)
                                goto out_short_read;
 
-                       ret = read_one_chunk(root, &key, sb, chunk, -1);
+                       ret = read_one_chunk(fs_info, &key, sb, chunk, -1);
                        if (ret)
                                break;
                } else {
@@ -2034,17 +2111,16 @@ out_short_read:
        return -EIO;
 }
 
-int btrfs_read_chunk_tree(struct btrfs_root *root)
+int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info)
 {
        struct btrfs_path *path;
        struct extent_buffer *leaf;
        struct btrfs_key key;
        struct btrfs_key found_key;
+       struct btrfs_root *root = fs_info->chunk_root;
        int ret;
        int slot;
 
-       root = root->fs_info->chunk_root;
-
        path = btrfs_alloc_path();
        if (!path)
                return -ENOMEM;
@@ -2077,12 +2153,12 @@ int btrfs_read_chunk_tree(struct btrfs_root *root)
                        struct btrfs_dev_item *dev_item;
                        dev_item = btrfs_item_ptr(leaf, slot,
                                                  struct btrfs_dev_item);
-                       ret = read_one_dev(root, leaf, dev_item);
+                       ret = read_one_dev(fs_info, leaf, dev_item);
                        BUG_ON(ret);
                } else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) {
                        struct btrfs_chunk *chunk;
                        chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
-                       ret = read_one_chunk(root, &found_key, leaf, chunk,
+                       ret = read_one_chunk(fs_info, &found_key, leaf, chunk,
                                             slot);
                        BUG_ON(ret);
                }
@@ -2331,3 +2407,181 @@ u64 btrfs_stripe_length(struct btrfs_fs_info *fs_info,
        }
        return stripe_len;
 }
+
+/*
+ * Return 0 if size of @device is already good
+ * Return >0 if size of @device is not aligned but fixed without problems
+ * Return <0 if something wrong happened when aligning the size of @device
+ */
+int btrfs_fix_device_size(struct btrfs_fs_info *fs_info,
+                         struct btrfs_device *device)
+{
+       struct btrfs_trans_handle *trans;
+       struct btrfs_key key;
+       struct btrfs_path path;
+       struct btrfs_root *chunk_root = fs_info->chunk_root;
+       struct btrfs_dev_item *di;
+       u64 old_bytes = device->total_bytes;
+       int ret;
+
+       if (IS_ALIGNED(old_bytes, fs_info->sectorsize))
+               return 0;
+
+       /* Align the in-memory total_bytes first, and use it as correct size */
+       device->total_bytes = round_down(device->total_bytes,
+                                        fs_info->sectorsize);
+
+       key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
+       key.type = BTRFS_DEV_ITEM_KEY;
+       key.offset = device->devid;
+
+       trans = btrfs_start_transaction(chunk_root, 1);
+       if (IS_ERR(trans)) {
+               ret = PTR_ERR(trans);
+               error("error starting transaction: %d (%s)",
+                     ret, strerror(-ret));
+               return ret;
+       }
+
+       btrfs_init_path(&path);
+       ret = btrfs_search_slot(trans, chunk_root, &key, &path, 0, 1);
+       if (ret > 0) {
+               error("failed to find DEV_ITEM for devid %llu", device->devid);
+               ret = -ENOENT;
+               goto err;
+       }
+       if (ret < 0) {
+               error("failed to search chunk root: %d (%s)",
+                       ret, strerror(-ret));
+               goto err;
+       }
+       di = btrfs_item_ptr(path.nodes[0], path.slots[0], struct btrfs_dev_item);
+       btrfs_set_device_total_bytes(path.nodes[0], di, device->total_bytes);
+       btrfs_mark_buffer_dirty(path.nodes[0]);
+       ret = btrfs_commit_transaction(trans, chunk_root);
+       if (ret < 0) {
+               error("failed to commit current transaction: %d (%s)",
+                       ret, strerror(-ret));
+               btrfs_release_path(&path);
+               return ret;
+       }
+       btrfs_release_path(&path);
+       printf("Fixed device size for devid %llu, old size: %llu new size: %llu\n",
+               device->devid, old_bytes, device->total_bytes);
+       return 1;
+
+err:
+       /* We haven't modified anything, it's OK to commit current trans */
+       btrfs_commit_transaction(trans, chunk_root);
+       btrfs_release_path(&path);
+       return ret;
+}
+
+/*
+ * Return 0 if super block total_bytes matches all devices' total_bytes
+ * Return >0 if super block total_bytes mismatch but fixed without problem
+ * Return <0 if we failed to fix super block total_bytes
+ */
+int btrfs_fix_super_size(struct btrfs_fs_info *fs_info)
+{
+       struct btrfs_trans_handle *trans;
+       struct btrfs_device *device;
+       struct list_head *dev_list = &fs_info->fs_devices->devices;
+       u64 total_bytes = 0;
+       u64 old_bytes = btrfs_super_total_bytes(fs_info->super_copy);
+       int ret;
+
+       list_for_each_entry(device, dev_list, dev_list) {
+               /*
+                * Caller should ensure this function is called after aligning
+                * all devices' total_bytes.
+                */
+               if (!IS_ALIGNED(device->total_bytes, fs_info->sectorsize)) {
+                       error("device %llu total_bytes %llu not aligned to %u",
+                               device->devid, device->total_bytes,
+                               fs_info->sectorsize);
+                       return -EUCLEAN;
+               }
+               total_bytes += device->total_bytes;
+       }
+
+       if (total_bytes == old_bytes)
+               return 0;
+
+       btrfs_set_super_total_bytes(fs_info->super_copy, total_bytes);
+
+       /* Commit transaction to update all super blocks */
+       trans = btrfs_start_transaction(fs_info->tree_root, 1);
+       if (IS_ERR(trans)) {
+               ret = PTR_ERR(trans);
+               error("error starting transaction:  %d (%s)",
+                     ret, strerror(-ret));
+               return ret;
+       }
+       ret = btrfs_commit_transaction(trans, fs_info->tree_root);
+       if (ret < 0) {
+               error("failed to commit current transaction: %d (%s)",
+                       ret, strerror(-ret));
+               return ret;
+       }
+       printf("Fixed super total bytes, old size: %llu new size: %llu\n",
+               old_bytes, total_bytes);
+       return 1;
+}
+
+/*
+ * Return 0 if all devices and super block sizes are good
+ * Return >0 if any device/super size problem was found, but fixed
+ * Return <0 if something wrong happened during fixing
+ */
+int btrfs_fix_device_and_super_size(struct btrfs_fs_info *fs_info)
+{
+       struct btrfs_device *device;
+       struct list_head *dev_list = &fs_info->fs_devices->devices;
+       bool have_bad_value = false;
+       int ret;
+
+       /* Seed device is not supported yet */
+       if (fs_info->fs_devices->seed) {
+               error("fixing device size with seed device is not supported yet");
+               return -EOPNOTSUPP;
+       }
+
+       /* All devices must be set up before repairing */
+       if (list_empty(dev_list)) {
+               error("no device found");
+               return -ENODEV;
+       }
+       list_for_each_entry(device, dev_list, dev_list) {
+               if (device->fd == -1 || !device->writeable) {
+                       error("devid %llu is missing or not writeable",
+                             device->devid);
+                       error(
+       "fixing device size needs all device(s) to be present and writeable");
+                       return -ENODEV;
+               }
+       }
+
+       /* Repair total_bytes of each device */
+       list_for_each_entry(device, dev_list, dev_list) {
+               ret = btrfs_fix_device_size(fs_info, device);
+               if (ret < 0)
+                       return ret;
+               if (ret > 0)
+                       have_bad_value = true;
+       }
+
+       /* Repair super total_byte */
+       ret = btrfs_fix_super_size(fs_info);
+       if (ret > 0)
+               have_bad_value = true;
+       if (have_bad_value) {
+               printf(
+       "Fixed unaligned/mismatched total_bytes for super block and device items\n");
+               ret = 1;
+       } else {
+               printf("No device size related problem found\n");
+               ret = 0;
+       }
+       return ret;
+}