btrfs-progs: extent-tree: Introduce functions to free in-memory block group cache
[platform/upstream/btrfs-progs.git] / mkfs.c
diff --git a/mkfs.c b/mkfs.c
index 7cd0c23..b60fc5a 100644 (file)
--- a/mkfs.c
+++ b/mkfs.c
  * Boston, MA 021110-1307, USA.
  */
 
-#define _XOPEN_SOURCE 500
-#define _GNU_SOURCE
-
 #include "kerncompat.h"
 
-#ifndef __CHECKER__
 #include <sys/ioctl.h>
 #include <sys/mount.h>
 #include "ioctl.h"
-#endif
-
 #include <stdio.h>
 #include <stdlib.h>
 #include <sys/types.h>
 #include <unistd.h>
 #include <getopt.h>
 #include <uuid/uuid.h>
-#include <linux/fs.h>
 #include <ctype.h>
-#include <attr/xattr.h>
+#include <sys/xattr.h>
+#include <limits.h>
+#include <linux/limits.h>
+#include <blkid/blkid.h>
+#include <ftw.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "volumes.h"
 #include "transaction.h"
 #include "utils.h"
-#include "version.h"
 
 static u64 index_cnt = 2;
+static int verbose = 1;
 
 struct directory_name_entry {
        char *dir_name;
@@ -55,23 +52,33 @@ struct directory_name_entry {
        struct list_head list;
 };
 
-static int make_root_dir(struct btrfs_root *root, int mixed)
+struct mkfs_allocation {
+       u64 data;
+       u64 metadata;
+       u64 mixed;
+       u64 system;
+};
+
+static int create_metadata_block_groups(struct btrfs_root *root,
+               u64 metadata_profile, int mixed,
+               struct mkfs_allocation *allocation)
 {
        struct btrfs_trans_handle *trans;
-       struct btrfs_key location;
        u64 bytes_used;
        u64 chunk_start = 0;
        u64 chunk_size = 0;
        int ret;
 
        trans = btrfs_start_transaction(root, 1);
-       bytes_used = btrfs_super_bytes_used(&root->fs_info->super_copy);
+       bytes_used = btrfs_super_bytes_used(root->fs_info->super_copy);
 
        root->fs_info->system_allocs = 1;
        ret = btrfs_make_block_group(trans, root, bytes_used,
+                                    metadata_profile |
                                     BTRFS_BLOCK_GROUP_SYSTEM,
                                     BTRFS_FIRST_CHUNK_TREE_OBJECTID,
                                     0, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
+       allocation->system += BTRFS_MKFS_SYSTEM_GROUP_SIZE;
        BUG_ON(ret);
 
        if (mixed) {
@@ -79,43 +86,81 @@ static int make_root_dir(struct btrfs_root *root, int mixed)
                                        &chunk_start, &chunk_size,
                                        BTRFS_BLOCK_GROUP_METADATA |
                                        BTRFS_BLOCK_GROUP_DATA);
+               if (ret == -ENOSPC) {
+                       fprintf(stderr,
+                               "no space to alloc data/metadata chunk\n");
+                       goto err;
+               }
                BUG_ON(ret);
                ret = btrfs_make_block_group(trans, root, 0,
+                                            metadata_profile |
                                             BTRFS_BLOCK_GROUP_METADATA |
                                             BTRFS_BLOCK_GROUP_DATA,
                                             BTRFS_FIRST_CHUNK_TREE_OBJECTID,
                                             chunk_start, chunk_size);
                BUG_ON(ret);
-               printf("Created a data/metadata chunk of size %llu\n", chunk_size);
+               allocation->mixed += chunk_size;
        } else {
                ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
                                        &chunk_start, &chunk_size,
                                        BTRFS_BLOCK_GROUP_METADATA);
+               if (ret == -ENOSPC) {
+                       fprintf(stderr, "no space to alloc metadata chunk\n");
+                       goto err;
+               }
                BUG_ON(ret);
                ret = btrfs_make_block_group(trans, root, 0,
+                                            metadata_profile |
                                             BTRFS_BLOCK_GROUP_METADATA,
                                             BTRFS_FIRST_CHUNK_TREE_OBJECTID,
                                             chunk_start, chunk_size);
+               allocation->metadata += chunk_size;
                BUG_ON(ret);
        }
 
        root->fs_info->system_allocs = 0;
        btrfs_commit_transaction(trans, root);
-       trans = btrfs_start_transaction(root, 1);
-       BUG_ON(!trans);
+
+err:
+       return ret;
+}
+
+static int create_data_block_groups(struct btrfs_trans_handle *trans,
+               struct btrfs_root *root, u64 data_profile, int mixed,
+               struct mkfs_allocation *allocation)
+{
+       u64 chunk_start = 0;
+       u64 chunk_size = 0;
+       int ret = 0;
 
        if (!mixed) {
                ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
                                        &chunk_start, &chunk_size,
                                        BTRFS_BLOCK_GROUP_DATA);
+               if (ret == -ENOSPC) {
+                       fprintf(stderr, "no space to alloc data chunk\n");
+                       goto err;
+               }
                BUG_ON(ret);
                ret = btrfs_make_block_group(trans, root, 0,
+                                            data_profile |
                                             BTRFS_BLOCK_GROUP_DATA,
                                             BTRFS_FIRST_CHUNK_TREE_OBJECTID,
                                             chunk_start, chunk_size);
+               allocation->data += chunk_size;
                BUG_ON(ret);
        }
 
+err:
+       return ret;
+}
+
+static int make_root_dir(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+               int mixed, struct mkfs_allocation *allocation)
+{
+       struct btrfs_key location;
+       int ret;
+
        ret = btrfs_make_root_dir(trans, root->fs_info->tree_root,
                              BTRFS_ROOT_TREE_DIR_OBJECTID);
        if (ret)
@@ -127,7 +172,7 @@ static int make_root_dir(struct btrfs_root *root, int mixed)
        location.offset = (u64)-1;
        ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
                        "default", 7,
-                       btrfs_super_root_dir(&root->fs_info->super_copy),
+                       btrfs_super_root_dir(root->fs_info->super_copy),
                        &location, BTRFS_FT_DIR, 0);
        if (ret)
                goto err;
@@ -138,54 +183,42 @@ static int make_root_dir(struct btrfs_root *root, int mixed)
        if (ret)
                goto err;
 
-       btrfs_commit_transaction(trans, root);
 err:
        return ret;
 }
 
-static int recow_roots(struct btrfs_trans_handle *trans,
-                      struct btrfs_root *root)
+static void __recow_root(struct btrfs_trans_handle *trans,
+                        struct btrfs_root *root)
 {
        int ret;
        struct extent_buffer *tmp;
-       struct btrfs_fs_info *info = root->fs_info;
-
-       ret = __btrfs_cow_block(trans, info->fs_root, info->fs_root->node,
-                               NULL, 0, &tmp, 0, 0);
-       BUG_ON(ret);
-       free_extent_buffer(tmp);
-
-       ret = __btrfs_cow_block(trans, info->tree_root, info->tree_root->node,
-                               NULL, 0, &tmp, 0, 0);
-       BUG_ON(ret);
-       free_extent_buffer(tmp);
-
-       ret = __btrfs_cow_block(trans, info->extent_root,
-                               info->extent_root->node, NULL, 0, &tmp, 0, 0);
-       BUG_ON(ret);
-       free_extent_buffer(tmp);
-
-       ret = __btrfs_cow_block(trans, info->chunk_root, info->chunk_root->node,
-                               NULL, 0, &tmp, 0, 0);
-       BUG_ON(ret);
-       free_extent_buffer(tmp);
-
 
-       ret = __btrfs_cow_block(trans, info->dev_root, info->dev_root->node,
-                               NULL, 0, &tmp, 0, 0);
-       BUG_ON(ret);
-       free_extent_buffer(tmp);
+       if (trans->transid != btrfs_root_generation(&root->root_item)) {
+               extent_buffer_get(root->node);
+               ret = __btrfs_cow_block(trans, root, root->node,
+                                       NULL, 0, &tmp, 0, 0);
+               BUG_ON(ret);
+               free_extent_buffer(tmp);
+       }
+}
 
-       ret = __btrfs_cow_block(trans, info->csum_root, info->csum_root->node,
-                               NULL, 0, &tmp, 0, 0);
-       BUG_ON(ret);
-       free_extent_buffer(tmp);
+static void recow_roots(struct btrfs_trans_handle *trans,
+                      struct btrfs_root *root)
+{
+       struct btrfs_fs_info *info = root->fs_info;
 
-       return 0;
+       __recow_root(trans, info->fs_root);
+       __recow_root(trans, info->tree_root);
+       __recow_root(trans, info->extent_root);
+       __recow_root(trans, info->chunk_root);
+       __recow_root(trans, info->dev_root);
+       __recow_root(trans, info->csum_root);
 }
 
 static int create_one_raid_group(struct btrfs_trans_handle *trans,
-                             struct btrfs_root *root, u64 type)
+                             struct btrfs_root *root, u64 type,
+                             struct mkfs_allocation *allocation)
+
 {
        u64 chunk_start;
        u64 chunk_size;
@@ -193,97 +226,62 @@ static int create_one_raid_group(struct btrfs_trans_handle *trans,
 
        ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
                                &chunk_start, &chunk_size, type);
+       if (ret == -ENOSPC) {
+               fprintf(stderr, "not enough free space\n");
+               exit(1);
+       }
        BUG_ON(ret);
        ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
                                     type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
                                     chunk_start, chunk_size);
+       if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == BTRFS_BLOCK_GROUP_DATA)
+               allocation->data += chunk_size;
+       else if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == BTRFS_BLOCK_GROUP_METADATA)
+               allocation->metadata += chunk_size;
+       else if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == BTRFS_BLOCK_GROUP_SYSTEM)
+               allocation->system += chunk_size;
+       else if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) ==
+                       (BTRFS_BLOCK_GROUP_METADATA|BTRFS_BLOCK_GROUP_DATA))
+               allocation->mixed += chunk_size;
+       else
+               BUG_ON(1);
+
        BUG_ON(ret);
        return ret;
 }
 
 static int create_raid_groups(struct btrfs_trans_handle *trans,
                              struct btrfs_root *root, u64 data_profile,
-                             int data_profile_opt, u64 metadata_profile,
-                             int metadata_profile_opt, int mixed)
+                             u64 metadata_profile, int mixed,
+                             struct mkfs_allocation *allocation)
 {
-       u64 num_devices = btrfs_super_num_devices(&root->fs_info->super_copy);
-       u64 allowed = 0;
-       u64 devices_for_raid = num_devices;
+       u64 num_devices = btrfs_super_num_devices(root->fs_info->super_copy);
        int ret;
 
-       /*
-        * Set default profiles according to number of added devices.
-        * For mixed groups defaults are single/single.
-        */
-       if (!metadata_profile_opt && !mixed) {
-               metadata_profile = (num_devices > 1) ?
-                       BTRFS_BLOCK_GROUP_RAID1 : BTRFS_BLOCK_GROUP_DUP;
-       }
-       if (!data_profile_opt && !mixed) {
-               data_profile = (num_devices > 1) ?
-                       BTRFS_BLOCK_GROUP_RAID0 : 0; /* raid0 or single */
-       }
-
-       if (devices_for_raid > 4)
-               devices_for_raid = 4;
-
-       switch (devices_for_raid) {
-       default:
-       case 4:
-               allowed |= BTRFS_BLOCK_GROUP_RAID10;
-       case 3:
-               allowed |= BTRFS_BLOCK_GROUP_RAID6;
-       case 2:
-               allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
-                       BTRFS_BLOCK_GROUP_RAID5;
-               break;
-       case 1:
-               allowed |= BTRFS_BLOCK_GROUP_DUP;
-       }
-
-       if (metadata_profile & ~allowed) {
-               fprintf(stderr, "unable to create FS with metadata "
-                       "profile %llu (have %llu devices)\n", metadata_profile,
-                       num_devices);
-               exit(1);
-       }
-       if (data_profile & ~allowed) {
-               fprintf(stderr, "unable to create FS with data "
-                       "profile %llu (have %llu devices)\n", data_profile,
-                       num_devices);
-               exit(1);
-       }
-
-       /* allow dup'ed data chunks only in mixed mode */
-       if (!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP)) {
-               fprintf(stderr, "dup for data is allowed only in mixed mode\n");
-               exit(1);
-       }
-
-       if (allowed & metadata_profile) {
+       if (metadata_profile) {
                u64 meta_flags = BTRFS_BLOCK_GROUP_METADATA;
 
                ret = create_one_raid_group(trans, root,
                                            BTRFS_BLOCK_GROUP_SYSTEM |
-                                           (allowed & metadata_profile));
+                                           metadata_profile, allocation);
                BUG_ON(ret);
 
                if (mixed)
                        meta_flags |= BTRFS_BLOCK_GROUP_DATA;
 
                ret = create_one_raid_group(trans, root, meta_flags |
-                                           (allowed & metadata_profile));
+                                           metadata_profile, allocation);
                BUG_ON(ret);
 
-               ret = recow_roots(trans, root);
-               BUG_ON(ret);
        }
-       if (!mixed && num_devices > 1 && (allowed & data_profile)) {
+       if (!mixed && num_devices > 1 && data_profile) {
                ret = create_one_raid_group(trans, root,
                                            BTRFS_BLOCK_GROUP_DATA |
-                                           (allowed & data_profile));
+                                           data_profile, allocation);
                BUG_ON(ret);
        }
+       recow_roots(trans, root);
+
        return 0;
 }
 
@@ -314,29 +312,34 @@ static int create_data_reloc_tree(struct btrfs_trans_handle *trans,
        return 0;
 }
 
-static void print_usage(void)
+static void print_usage(int ret)
 {
        fprintf(stderr, "usage: mkfs.btrfs [options] dev [ dev ... ]\n");
        fprintf(stderr, "options:\n");
-       fprintf(stderr, "\t -A --alloc-start the offset to start the FS\n");
-       fprintf(stderr, "\t -b --byte-count total number of bytes in the FS\n");
-       fprintf(stderr, "\t -d --data data profile, raid0, raid1, raid10, dup or single\n");
-       fprintf(stderr, "\t -l --leafsize size of btree leaves\n");
-       fprintf(stderr, "\t -L --label set a label\n");
-       fprintf(stderr, "\t -m --metadata metadata profile, values like data profile\n");
-       fprintf(stderr, "\t -M --mixed mix metadata and data together\n");
-       fprintf(stderr, "\t -n --nodesize size of btree nodes\n");
-       fprintf(stderr, "\t -s --sectorsize min block allocation\n");
-       fprintf(stderr, "\t -r --rootdir the source directory\n");
-       fprintf(stderr, "\t -K --nodiscard do not perform whole device TRIM\n");
-       fprintf(stderr, "\t -V --version print the mkfs.btrfs version and exit\n");
-       fprintf(stderr, "%s\n", BTRFS_BUILD_VERSION);
-       exit(1);
+       fprintf(stderr, "\t-A|--alloc-start START  the offset to start the FS\n");
+       fprintf(stderr, "\t-b|--byte-count SIZE    total number of bytes in the FS\n");
+       fprintf(stderr, "\t-d|--data PROFILE       data profile, raid0, raid1, raid5, raid6, raid10, dup or single\n");
+       fprintf(stderr, "\t-f|--force              force overwrite of existing filesystem\n");
+       fprintf(stderr, "\t-l|--leafsize SIZE      deprecated, alias for nodesize\n");
+       fprintf(stderr, "\t-L|--label LABEL        set a label\n");
+       fprintf(stderr, "\t-m|--metadata PROFILE   metadata profile, values like data profile\n");
+       fprintf(stderr, "\t-M|--mixed              mix metadata and data together\n");
+       fprintf(stderr, "\t-n|--nodesize SIZE      size of btree nodes\n");
+       fprintf(stderr, "\t-s|--sectorsize SIZE    min block allocation (may not mountable by current kernel)\n");
+       fprintf(stderr, "\t-r|--rootdir DIR        the source directory\n");
+       fprintf(stderr, "\t-K|--nodiscard          do not perform whole device TRIM\n");
+       fprintf(stderr, "\t-O|--features LIST      comma separated list of filesystem features, use '-O list-all' to list features\n");
+       fprintf(stderr, "\t-U|--uuid UUID          specify the filesystem UUID\n");
+       fprintf(stderr, "\t-q|--quiet              no messages except errors\n");
+       fprintf(stderr, "\t-V|--version            print the mkfs.btrfs version and exit\n");
+       fprintf(stderr, "%s\n", PACKAGE_STRING);
+       exit(ret);
 }
 
+static void print_version(void) __attribute__((noreturn));
 static void print_version(void)
 {
-       fprintf(stderr, "mkfs.btrfs, part of %s\n", BTRFS_BUILD_VERSION);
+       fprintf(stderr, "mkfs.btrfs, part of %s\n", PACKAGE_STRING);
        exit(0);
 }
 
@@ -344,21 +347,20 @@ static u64 parse_profile(char *s)
 {
        if (strcmp(s, "raid0") == 0) {
                return BTRFS_BLOCK_GROUP_RAID0;
-       } else if (strcmp(s, "raid1") == 0) {
+       } else if (strcasecmp(s, "raid1") == 0) {
                return BTRFS_BLOCK_GROUP_RAID1;
-       } else if (strcmp(s, "raid5") == 0) {
+       } else if (strcasecmp(s, "raid5") == 0) {
                return BTRFS_BLOCK_GROUP_RAID5;
-       } else if (strcmp(s, "raid6") == 0) {
+       } else if (strcasecmp(s, "raid6") == 0) {
                return BTRFS_BLOCK_GROUP_RAID6;
-       } else if (strcmp(s, "raid10") == 0) {
+       } else if (strcasecmp(s, "raid10") == 0) {
                return BTRFS_BLOCK_GROUP_RAID10;
-       } else if (strcmp(s, "dup") == 0) {
+       } else if (strcasecmp(s, "dup") == 0) {
                return BTRFS_BLOCK_GROUP_DUP;
-       } else if (strcmp(s, "single") == 0) {
+       } else if (strcasecmp(s, "single") == 0) {
                return 0;
        } else {
                fprintf(stderr, "Unknown profile %s\n", s);
-               print_usage();
        }
        /* not reached */
        return 0;
@@ -366,7 +368,6 @@ static u64 parse_profile(char *s)
 
 static char *parse_label(char *input)
 {
-       int i;
        int len = strlen(input);
 
        if (len >= BTRFS_LABEL_SIZE) {
@@ -374,31 +375,9 @@ static char *parse_label(char *input)
                        BTRFS_LABEL_SIZE - 1);
                exit(1);
        }
-       for (i = 0; i < len; i++) {
-               if (input[i] == '/' || input[i] == '\\') {
-                       fprintf(stderr, "invalid label %s\n", input);
-                       exit(1);
-               }
-       }
        return strdup(input);
 }
 
-static struct option long_options[] = {
-       { "alloc-start", 1, NULL, 'A'},
-       { "byte-count", 1, NULL, 'b' },
-       { "leafsize", 1, NULL, 'l' },
-       { "label", 1, NULL, 'L'},
-       { "metadata", 1, NULL, 'm' },
-       { "mixed", 0, NULL, 'M' },
-       { "nodesize", 1, NULL, 'n' },
-       { "sectorsize", 1, NULL, 's' },
-       { "data", 1, NULL, 'd' },
-       { "version", 0, NULL, 'V' },
-       { "rootdir", 1, NULL, 'r' },
-       { "nodiscard", 0, NULL, 'K' },
-       { 0, 0, 0, 0}
-};
-
 static int add_directory_items(struct btrfs_trans_handle *trans,
                               struct btrfs_root *root, u64 objectid,
                               ino_t parent_inum, const char *name,
@@ -425,7 +404,10 @@ static int add_directory_items(struct btrfs_trans_handle *trans,
        ret = btrfs_insert_dir_item(trans, root, name, name_len,
                                    parent_inum, &location,
                                    filetype, index_cnt);
-
+       if (ret)
+               return ret;
+       ret = btrfs_insert_inode_ref(trans, root, name, name_len,
+                                    objectid, parent_inum, index_cnt);
        *dir_index_cnt = index_cnt;
        index_cnt++;
 
@@ -538,9 +520,7 @@ static int add_inode_items(struct btrfs_trans_handle *trans,
        struct btrfs_inode_item btrfs_inode;
        u64 objectid;
        u64 inode_size = 0;
-       int name_len;
 
-       name_len = strlen(name);
        fill_inode_item(trans, root, &btrfs_inode, st);
        objectid = self_objectid;
 
@@ -554,16 +534,8 @@ static int add_inode_items(struct btrfs_trans_handle *trans,
        btrfs_set_key_type(&inode_key, BTRFS_INODE_ITEM_KEY);
 
        ret = btrfs_insert_inode(trans, root, objectid, &btrfs_inode);
-       if (ret)
-               goto fail;
-
-       ret = btrfs_insert_inode_ref(trans, root, name, name_len,
-                                    objectid, parent_inum, dir_index_cnt);
-       if (ret)
-               goto fail;
 
        *inode_ret = btrfs_inode;
-fail:
        return ret;
 }
 
@@ -617,137 +589,6 @@ static int add_xattr_item(struct btrfs_trans_handle *trans,
 
        return ret;
 }
-static int custom_alloc_extent(struct btrfs_root *root, u64 num_bytes,
-                              u64 hint_byte, struct btrfs_key *ins)
-{
-       u64 start;
-       u64 end;
-       u64 last = hint_byte;
-       int ret;
-       int wrapped = 0;
-       struct btrfs_block_group_cache *cache;
-
-       while (1) {
-               ret = find_first_extent_bit(&root->fs_info->free_space_cache,
-                                           last, &start, &end, EXTENT_DIRTY);
-               if (ret) {
-                       if (wrapped++ == 0) {
-                               last = 0;
-                               continue;
-                       } else {
-                               goto fail;
-                       }
-               }
-
-               start = max(last, start);
-               last = end + 1;
-               if (last - start < num_bytes)
-                       continue;
-
-               last = start + num_bytes;
-               if (test_range_bit(&root->fs_info->pinned_extents,
-                                  start, last - 1, EXTENT_DIRTY, 0))
-                       continue;
-
-               cache = btrfs_lookup_block_group(root->fs_info, start);
-               BUG_ON(!cache);
-               if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM ||
-                   last > cache->key.objectid + cache->key.offset) {
-                       last = cache->key.objectid + cache->key.offset;
-                       continue;
-               }
-
-               if (cache->flags & (BTRFS_BLOCK_GROUP_SYSTEM |
-                           BTRFS_BLOCK_GROUP_METADATA)) {
-                       last = cache->key.objectid + cache->key.offset;
-                       continue;
-               }
-
-               clear_extent_dirty(&root->fs_info->free_space_cache,
-                                  start, start + num_bytes - 1, 0);
-
-               ins->objectid = start;
-               ins->offset = num_bytes;
-               ins->type = BTRFS_EXTENT_ITEM_KEY;
-               return 0;
-       }
-fail:
-       fprintf(stderr, "not enough free space\n");
-       return -ENOSPC;
-}
-
-static int record_file_extent(struct btrfs_trans_handle *trans,
-                             struct btrfs_root *root, u64 objectid,
-                             struct btrfs_inode_item *inode,
-                             u64 file_pos, u64 disk_bytenr,
-                             u64 num_bytes)
-{
-       int ret;
-       struct btrfs_fs_info *info = root->fs_info;
-       struct btrfs_root *extent_root = info->extent_root;
-       struct extent_buffer *leaf;
-       struct btrfs_file_extent_item *fi;
-       struct btrfs_key ins_key;
-       struct btrfs_path path;
-       struct btrfs_extent_item *ei;
-
-       btrfs_init_path(&path);
-
-       ins_key.objectid = objectid;
-       ins_key.offset = 0;
-       btrfs_set_key_type(&ins_key, BTRFS_EXTENT_DATA_KEY);
-       ret = btrfs_insert_empty_item(trans, root, &path, &ins_key,
-                                     sizeof(*fi));
-       if (ret)
-               goto fail;
-       leaf = path.nodes[0];
-       fi = btrfs_item_ptr(leaf, path.slots[0],
-                           struct btrfs_file_extent_item);
-       btrfs_set_file_extent_generation(leaf, fi, trans->transid);
-       btrfs_set_file_extent_type(leaf, fi, BTRFS_FILE_EXTENT_REG);
-       btrfs_set_file_extent_disk_bytenr(leaf, fi, disk_bytenr);
-       btrfs_set_file_extent_disk_num_bytes(leaf, fi, num_bytes);
-       btrfs_set_file_extent_offset(leaf, fi, 0);
-       btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
-       btrfs_set_file_extent_ram_bytes(leaf, fi, num_bytes);
-       btrfs_set_file_extent_compression(leaf, fi, 0);
-       btrfs_set_file_extent_encryption(leaf, fi, 0);
-       btrfs_set_file_extent_other_encoding(leaf, fi, 0);
-       btrfs_mark_buffer_dirty(leaf);
-
-       btrfs_release_path(root, &path);
-
-       ins_key.objectid = disk_bytenr;
-       ins_key.offset = num_bytes;
-       ins_key.type = BTRFS_EXTENT_ITEM_KEY;
-
-       ret = btrfs_insert_empty_item(trans, extent_root, &path,
-                               &ins_key, sizeof(*ei));
-       if (ret == 0) {
-               leaf = path.nodes[0];
-               ei = btrfs_item_ptr(leaf, path.slots[0],
-                                   struct btrfs_extent_item);
-
-               btrfs_set_extent_refs(leaf, ei, 0);
-               btrfs_set_extent_generation(leaf, ei, trans->transid);
-               btrfs_set_extent_flags(leaf, ei, BTRFS_EXTENT_FLAG_DATA);
-
-               btrfs_mark_buffer_dirty(leaf);
-               ret = btrfs_update_block_group(trans, root, disk_bytenr,
-                                              num_bytes, 1, 0);
-               if (ret)
-                       goto fail;
-       } else if (ret != -EEXIST) {
-               goto fail;
-       }
-
-       ret = btrfs_inc_extent_ref(trans, root, disk_bytenr, num_bytes, 0,
-                                  root->root_key.objectid,
-                                  objectid, 0);
-fail:
-       btrfs_release_path(root, &path);
-       return ret;
-}
 
 static int add_symbolic_link(struct btrfs_trans_handle *trans,
                             struct btrfs_root *root,
@@ -785,18 +626,23 @@ static int add_file_items(struct btrfs_trans_handle *trans,
        int ret = -1;
        ssize_t ret_read;
        u64 bytes_read = 0;
-       char *buffer = NULL;
        struct btrfs_key key;
        int blocks;
        u32 sectorsize = root->sectorsize;
        u64 first_block = 0;
-       u64 num_blocks = 0;
+       u64 file_pos = 0;
+       u64 cur_bytes;
+       u64 total_bytes;
+       struct extent_buffer *eb = NULL;
        int fd;
 
+       if (st->st_size == 0)
+               return 0;
+
        fd = open(path_name, O_RDONLY);
        if (fd == -1) {
                fprintf(stderr, "%s open failed\n", path_name);
-               goto end;
+               return ret;
        }
 
        blocks = st->st_size / sectorsize;
@@ -804,66 +650,98 @@ static int add_file_items(struct btrfs_trans_handle *trans,
                blocks += 1;
 
        if (st->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
-               buffer = malloc(st->st_size);
+               char *buffer = malloc(st->st_size);
                ret_read = pread64(fd, buffer, st->st_size, bytes_read);
                if (ret_read == -1) {
                        fprintf(stderr, "%s read failed\n", path_name);
+                       free(buffer);
                        goto end;
                }
 
                ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
                                                 buffer, st->st_size);
+               free(buffer);
                goto end;
        }
 
-       ret = custom_alloc_extent(root, blocks * sectorsize, 0, &key);
+       /* round up our st_size to the FS blocksize */
+       total_bytes = (u64)blocks * sectorsize;
+
+       /*
+        * do our IO in extent buffers so it can work
+        * against any raid type
+        */
+       eb = malloc(sizeof(*eb) + sectorsize);
+       if (!eb) {
+               ret = -ENOMEM;
+               goto end;
+       }
+       memset(eb, 0, sizeof(*eb) + sectorsize);
+
+again:
+
+       /*
+        * keep our extent size at 1MB max, this makes it easier to work inside
+        * the tiny block groups created during mkfs
+        */
+       cur_bytes = min(total_bytes, 1024ULL * 1024);
+       ret = btrfs_reserve_extent(trans, root, cur_bytes, 0, 0, (u64)-1,
+                                  &key, 1);
        if (ret)
                goto end;
 
        first_block = key.objectid;
        bytes_read = 0;
-       buffer = malloc(sectorsize);
 
-       do {
-               memset(buffer, 0, sectorsize);
-               ret_read = pread64(fd, buffer, sectorsize, bytes_read);
+       while (bytes_read < cur_bytes) {
+
+               memset(eb->data, 0, sectorsize);
+
+               ret_read = pread64(fd, eb->data, sectorsize, file_pos + bytes_read);
                if (ret_read == -1) {
                        fprintf(stderr, "%s read failed\n", path_name);
                        goto end;
                }
 
-               ret = pwrite64(out_fd, buffer, sectorsize,
-                              first_block + bytes_read);
-               if (ret != sectorsize) {
-                       fprintf(stderr, "output file write failed\n");
-                       goto end;
-               }
+               eb->start = first_block + bytes_read;
+               eb->len = sectorsize;
 
-               /* checksum for file data */
+               /*
+                * we're doing the csum before we record the extent, but
+                * that's ok
+                */
                ret = btrfs_csum_file_block(trans, root->fs_info->csum_root,
-                               first_block + (blocks * sectorsize),
-                               first_block + bytes_read,
-                               buffer, sectorsize);
+                                           first_block + bytes_read + sectorsize,
+                                           first_block + bytes_read,
+                                           eb->data, sectorsize);
+               if (ret)
+                       goto end;
+
+               ret = write_and_map_eb(trans, root, eb);
                if (ret) {
-                       fprintf(stderr, "%s checksum failed\n", path_name);
+                       fprintf(stderr, "output file write failed\n");
                        goto end;
                }
 
-               bytes_read += ret_read;
-               num_blocks++;
-       } while (ret_read == sectorsize);
+               bytes_read += sectorsize;
+       }
 
-       if (num_blocks > 0) {
-               ret = record_file_extent(trans, root, objectid, btrfs_inode,
-                                        first_block, first_block,
-                                        blocks * sectorsize);
+       if (bytes_read) {
+               ret = btrfs_record_file_extent(trans, root, objectid, btrfs_inode,
+                                              file_pos, first_block, cur_bytes);
                if (ret)
                        goto end;
+
        }
 
+       file_pos += cur_bytes;
+       total_bytes -= cur_bytes;
+
+       if (total_bytes)
+               goto again;
+
 end:
-       if (buffer)
-               free(buffer);
+       free(eb);
        close(fd);
        return ret;
 }
@@ -898,6 +776,7 @@ static int traverse_directory(struct btrfs_trans_handle *trans,
        ino_t parent_inum, cur_inum;
        ino_t highest_inum = 0;
        char *parent_dir_name;
+       char real_path[PATH_MAX];
        struct btrfs_path path;
        struct extent_buffer *leaf;
        struct btrfs_key root_dir_key;
@@ -906,7 +785,12 @@ static int traverse_directory(struct btrfs_trans_handle *trans,
        /* Add list for source directory */
        dir_entry = malloc(sizeof(struct directory_name_entry));
        dir_entry->dir_name = dir_name;
-       dir_entry->path = strdup(dir_name);
+       dir_entry->path = realpath(dir_name, real_path);
+       if (!dir_entry->path) {
+               fprintf(stderr, "get directory real path error\n");
+               ret = -1;
+               goto fail_no_dir;
+       }
 
        parent_inum = highest_inum + BTRFS_FIRST_FREE_OBJECTID;
        dir_entry->inum = parent_inum;
@@ -920,7 +804,7 @@ static int traverse_directory(struct btrfs_trans_handle *trans,
        ret = btrfs_lookup_inode(trans, root, &path, &root_dir_key, 1);
        if (ret) {
                fprintf(stderr, "root dir lookup error\n");
-               return -1;
+               goto fail_no_dir;
        }
 
        leaf = path.nodes[0];
@@ -931,7 +815,7 @@ static int traverse_directory(struct btrfs_trans_handle *trans,
        btrfs_set_inode_size(leaf, inode_item, root_dir_inode_size);
        btrfs_mark_buffer_dirty(leaf);
 
-       btrfs_release_path(root, &path);
+       btrfs_release_path(&path);
 
        do {
                parent_dir_entry = list_entry(dir_head->list.next,
@@ -944,6 +828,7 @@ static int traverse_directory(struct btrfs_trans_handle *trans,
                if (chdir(parent_dir_entry->path)) {
                        fprintf(stderr, "chdir error for %s\n",
                                parent_dir_name);
+                       ret = -1;
                        goto fail_no_files;
                }
 
@@ -953,6 +838,7 @@ static int traverse_directory(struct btrfs_trans_handle *trans,
                {
                        fprintf(stderr, "scandir for %s failed: %s\n",
                                parent_dir_name, strerror (errno));
+                       ret = -1;
                        goto fail;
                }
 
@@ -962,10 +848,11 @@ static int traverse_directory(struct btrfs_trans_handle *trans,
                        if (lstat(cur_file->d_name, &st) == -1) {
                                fprintf(stderr, "lstat failed for file %s\n",
                                        cur_file->d_name);
+                               ret = -1;
                                goto fail;
                        }
 
-                       cur_inum = ++highest_inum + BTRFS_FIRST_FREE_OBJECTID;
+                       cur_inum = st.st_ino;
                        ret = add_directory_items(trans, root,
                                                  cur_inum, parent_inum,
                                                  cur_file->d_name,
@@ -979,6 +866,10 @@ static int traverse_directory(struct btrfs_trans_handle *trans,
                                              cur_file->d_name, cur_inum,
                                              parent_inum, dir_index_cnt,
                                              &cur_inode);
+                       if (ret == -EEXIST) {
+                               BUG_ON(st.st_nlink <= 1);
+                               continue;
+                       }
                        if (ret) {
                                fprintf(stderr, "add_inode_items failed\n");
                                goto fail;
@@ -1018,20 +909,22 @@ static int traverse_directory(struct btrfs_trans_handle *trans,
                }
 
                free_namelist(files, count);
-               free(parent_dir_entry->path);
                free(parent_dir_entry);
 
                index_cnt = 2;
 
        } while (!list_empty(&dir_head->list));
 
-       return 0;
+out:
+       return !!ret;
 fail:
        free_namelist(files, count);
 fail_no_files:
-       free(parent_dir_entry->path);
        free(parent_dir_entry);
-       return -1;
+       goto out;
+fail_no_dir:
+       free(dir_entry);
+       goto out;
 }
 
 static int open_target(char *output_name)
@@ -1045,7 +938,8 @@ static int open_target(char *output_name)
 
 static int create_chunks(struct btrfs_trans_handle *trans,
                         struct btrfs_root *root, u64 num_of_meta_chunks,
-                        u64 size_of_data)
+                        u64 size_of_data,
+                        struct mkfs_allocation *allocation)
 {
        u64 chunk_start;
        u64 chunk_size;
@@ -1062,6 +956,7 @@ static int create_chunks(struct btrfs_trans_handle *trans,
                ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
                                             meta_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
                                             chunk_start, chunk_size);
+               allocation->metadata += chunk_size;
                BUG_ON(ret);
                set_extent_dirty(&root->fs_info->free_space_cache,
                                 chunk_start, chunk_start + chunk_size - 1, 0);
@@ -1069,12 +964,14 @@ static int create_chunks(struct btrfs_trans_handle *trans,
 
        if (size_of_data < minimum_data_chunk_size)
                size_of_data = minimum_data_chunk_size;
+
        ret = btrfs_alloc_data_chunk(trans, root->fs_info->extent_root,
                                     &chunk_start, size_of_data, data_type);
        BUG_ON(ret);
        ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
                                     data_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
                                     chunk_start, size_of_data);
+       allocation->data += size_of_data;
        BUG_ON(ret);
        set_extent_dirty(&root->fs_info->free_space_cache,
                         chunk_start, chunk_start + size_of_data - 1, 0);
@@ -1090,10 +987,12 @@ static int make_image(char *source_dir, struct btrfs_root *root, int out_fd)
 
        struct directory_name_entry dir_head;
 
+       struct directory_name_entry *dir_entry = NULL;
+
        ret = lstat(source_dir, &root_st);
        if (ret) {
                fprintf(stderr, "unable to lstat the %s\n", source_dir);
-               goto fail;
+               goto out;
        }
 
        INIT_LIST_HEAD(&dir_head.list);
@@ -1106,47 +1005,64 @@ static int make_image(char *source_dir, struct btrfs_root *root, int out_fd)
        }
        btrfs_commit_transaction(trans, root);
 
-       printf("Making image is completed.\n");
+       if (verbose)
+               printf("Making image is completed.\n");
        return 0;
 fail:
+       while (!list_empty(&dir_head.list)) {
+               dir_entry = list_entry(dir_head.list.next,
+                                      struct directory_name_entry, list);
+               list_del(&dir_entry->list);
+               free(dir_entry);
+       }
+out:
        fprintf(stderr, "Making image is aborted.\n");
        return -1;
 }
 
+/*
+ * This ignores symlinks with unreadable targets and subdirs that can't
+ * be read.  It's a best-effort to give a rough estimate of the size of
+ * a subdir.  It doesn't guarantee that prepopulating btrfs from this
+ * tree won't still run out of space. 
+ *
+ * The rounding up to 4096 is questionable.  Previous code used du -B 4096.
+ */
+static u64 global_total_size;
+static int ftw_add_entry_size(const char *fpath, const struct stat *st,
+                             int type)
+{
+       if (type == FTW_F || type == FTW_D)
+               global_total_size += round_up(st->st_size, 4096);
+
+       return 0;
+}
+
 static u64 size_sourcedir(char *dir_name, u64 sectorsize,
                          u64 *num_of_meta_chunks_ret, u64 *size_of_data_ret)
 {
        u64 dir_size = 0;
        u64 total_size = 0;
        int ret;
-       char command[1024];
-       char path[512];
-       char *file_name = "temp_file";
-       FILE *file;
        u64 default_chunk_size = 8 * 1024 * 1024;       /* 8MB */
        u64 allocated_meta_size = 8 * 1024 * 1024;      /* 8MB */
        u64 allocated_total_size = 20 * 1024 * 1024;    /* 20MB */
        u64 num_of_meta_chunks = 0;
+       u64 num_of_data_chunks = 0;
        u64 num_of_allocated_meta_chunks =
                        allocated_meta_size / default_chunk_size;
 
-       ret = sprintf(command, "du -B 4096 -s ");
+       global_total_size = 0;
+       ret = ftw(dir_name, ftw_add_entry_size, 10);
+       dir_size = global_total_size;
        if (ret < 0) {
-               fprintf(stderr, "error executing sprintf for du command\n");
-               return -1;
+               fprintf(stderr, "ftw subdir walk of '%s' failed: %s\n",
+                       dir_name, strerror(errno));
+               exit(1);
        }
-       strcat(command, dir_name);
-       strcat(command, " > ");
-       strcat(command, file_name);
-       ret = system(command);
-
-       file = fopen(file_name, "r");
-       ret = fscanf(file, "%lld %s\n", &dir_size, path);
-       fclose(file);
-       remove(file_name);
 
-       dir_size *= sectorsize;
-       *size_of_data_ret = dir_size;
+       num_of_data_chunks = (dir_size + default_chunk_size - 1) /
+               default_chunk_size;
 
        num_of_meta_chunks = (dir_size / 2) / default_chunk_size;
        if (((dir_size / 2) % default_chunk_size) != 0)
@@ -1156,11 +1072,12 @@ static u64 size_sourcedir(char *dir_name, u64 sectorsize,
        else
                num_of_meta_chunks -= num_of_allocated_meta_chunks;
 
-       total_size = allocated_total_size + dir_size +
+       total_size = allocated_total_size +
+                    (num_of_data_chunks * default_chunk_size) +
                     (num_of_meta_chunks * default_chunk_size);
 
        *num_of_meta_chunks_ret = num_of_meta_chunks;
-
+       *size_of_data_ret = num_of_data_chunks * default_chunk_size;
        return total_size;
 }
 
@@ -1186,25 +1103,83 @@ static int zero_output_file(int out_fd, u64 size, u32 sectorsize)
        return ret;
 }
 
-static int check_leaf_or_node_size(u32 size, u32 sectorsize)
+static int is_ssd(const char *file)
 {
-       if (size < sectorsize) {
-               fprintf(stderr,
-                       "Illegal leafsize (or nodesize) %u (smaller than %u)\n",
-                       size, sectorsize);
-               return -1;
-       } else if (size > BTRFS_MAX_METADATA_BLOCKSIZE) {
-               fprintf(stderr,
-                       "Illegal leafsize (or nodesize) %u (larger than %u)\n",
-                       size, BTRFS_MAX_METADATA_BLOCKSIZE);
-               return -1;
-       } else if (size & (sectorsize - 1)) {
-               fprintf(stderr,
-                       "Illegal leafsize (or nodesize) %u (not align to %u)\n",
-                       size, sectorsize);
-               return -1;
+       blkid_probe probe;
+       char wholedisk[32];
+       char sysfs_path[PATH_MAX];
+       dev_t devno;
+       int fd;
+       char rotational;
+       int ret;
+
+       probe = blkid_new_probe_from_filename(file);
+       if (!probe)
+               return 0;
+
+       /* Device number of this disk (possibly a partition) */
+       devno = blkid_probe_get_devno(probe);
+       if (!devno) {
+               blkid_free_probe(probe);
+               return 0;
        }
-       return 0;
+
+       /* Get whole disk name (not full path) for this devno */
+       ret = blkid_devno_to_wholedisk(devno,
+                       wholedisk, sizeof(wholedisk), NULL);
+       if (ret) {
+               blkid_free_probe(probe);
+               return 0;
+       }
+
+       snprintf(sysfs_path, PATH_MAX, "/sys/block/%s/queue/rotational",
+                wholedisk);
+
+       blkid_free_probe(probe);
+
+       fd = open(sysfs_path, O_RDONLY);
+       if (fd < 0) {
+               return 0;
+       }
+
+       if (read(fd, &rotational, sizeof(char)) < sizeof(char)) {
+               close(fd);
+               return 0;
+       }
+       close(fd);
+
+       return !atoi((const char *)&rotational);
+}
+
+static void list_all_devices(struct btrfs_root *root)
+{
+       struct btrfs_fs_devices *fs_devices;
+       struct btrfs_device *device;
+       int number_of_devices = 0;
+       u64 total_block_count = 0;
+
+       fs_devices = root->fs_info->fs_devices;
+
+       list_for_each_entry(device, &fs_devices->devices, dev_list)
+               number_of_devices++;
+
+       printf("Number of devices:  %d\n", number_of_devices);
+       /* printf("Total devices size: %10s\n", */
+               /* pretty_size(total_block_count)); */
+       printf("Devices:\n");
+       printf("   ID        SIZE  PATH\n");
+       list_for_each_entry_reverse(device, &fs_devices->devices, dev_list) {
+               char dev_uuid[BTRFS_UUID_UNPARSED_SIZE];
+
+               uuid_unparse(device->uuid, dev_uuid);
+               printf("  %3llu  %10s  %s\n",
+                       device->devid,
+                       pretty_size(device->total_bytes),
+                       device->name);
+               total_block_count += device->total_bytes;
+       }
+
+       printf("\n");
 }
 
 int main(int ac, char **av)
@@ -1213,52 +1188,85 @@ int main(int ac, char **av)
        struct btrfs_root *root;
        struct btrfs_trans_handle *trans;
        char *label = NULL;
-       char *first_file;
        u64 block_count = 0;
        u64 dev_block_count = 0;
        u64 blocks[7];
        u64 alloc_start = 0;
        u64 metadata_profile = 0;
        u64 data_profile = 0;
-       u32 leafsize = getpagesize();
+       u64 default_metadata_profile = 0;
+       u64 default_data_profile = 0;
+       u32 nodesize = max_t(u32, sysconf(_SC_PAGESIZE),
+                       BTRFS_MKFS_DEFAULT_NODE_SIZE);
        u32 sectorsize = 4096;
-       u32 nodesize = leafsize;
        u32 stripesize = 4096;
        int zero_end = 1;
-       int option_index = 0;
        int fd;
        int ret;
        int i;
        int mixed = 0;
+       int nodesize_forced = 0;
        int data_profile_opt = 0;
        int metadata_profile_opt = 0;
-       int nodiscard = 0;
-
+       int discard = 1;
+       int ssd = 0;
+       int force_overwrite = 0;
        char *source_dir = NULL;
        int source_dir_set = 0;
        u64 num_of_meta_chunks = 0;
        u64 size_of_data = 0;
        u64 source_dir_size = 0;
-       char *pretty_buf;
+       int dev_cnt = 0;
+       int saved_optind;
+       char fs_uuid[BTRFS_UUID_UNPARSED_SIZE] = { 0 };
+       u64 features = BTRFS_MKFS_DEFAULT_FEATURES;
+       struct mkfs_allocation allocation = { 0 };
+       struct btrfs_mkfs_config mkfs_cfg;
 
        while(1) {
                int c;
-               c = getopt_long(ac, av, "A:b:l:n:s:m:d:L:r:VMK", long_options,
-                               &option_index);
+               static const struct option long_options[] = {
+                       { "alloc-start", required_argument, NULL, 'A'},
+                       { "byte-count", required_argument, NULL, 'b' },
+                       { "force", no_argument, NULL, 'f' },
+                       { "leafsize", required_argument, NULL, 'l' },
+                       { "label", required_argument, NULL, 'L'},
+                       { "metadata", required_argument, NULL, 'm' },
+                       { "mixed", no_argument, NULL, 'M' },
+                       { "nodesize", required_argument, NULL, 'n' },
+                       { "sectorsize", required_argument, NULL, 's' },
+                       { "data", required_argument, NULL, 'd' },
+                       { "version", no_argument, NULL, 'V' },
+                       { "rootdir", required_argument, NULL, 'r' },
+                       { "nodiscard", no_argument, NULL, 'K' },
+                       { "features", required_argument, NULL, 'O' },
+                       { "uuid", required_argument, NULL, 'U' },
+                       { "quiet", 0, NULL, 'q' },
+                       { "help", no_argument, NULL, GETOPT_VAL_HELP },
+                       { NULL, 0, NULL, 0}
+               };
+
+               c = getopt_long(ac, av, "A:b:fl:n:s:m:d:L:O:r:U:VMKq",
+                               long_options, NULL);
                if (c < 0)
                        break;
                switch(c) {
                        case 'A':
                                alloc_start = parse_size(optarg);
                                break;
+                       case 'f':
+                               force_overwrite = 1;
+                               break;
                        case 'd':
                                data_profile = parse_profile(optarg);
                                data_profile_opt = 1;
                                break;
                        case 'l':
+                               fprintf(stderr,
+                       "WARNING: --leafsize is deprecated, use --nodesize\n");
                        case 'n':
                                nodesize = parse_size(optarg);
-                               leafsize = parse_size(optarg);
+                               nodesize_forced = 1;
                                break;
                        case 'L':
                                label = parse_label(optarg);
@@ -1270,16 +1278,32 @@ int main(int ac, char **av)
                        case 'M':
                                mixed = 1;
                                break;
+                       case 'O': {
+                               char *orig = strdup(optarg);
+                               char *tmp = orig;
+
+                               tmp = btrfs_parse_fs_features(tmp, &features);
+                               if (tmp) {
+                                       fprintf(stderr,
+                                               "Unrecognized filesystem feature '%s'\n",
+                                                       tmp);
+                                       free(orig);
+                                       exit(1);
+                               }
+                               free(orig);
+                               if (features & BTRFS_FEATURE_LIST_ALL) {
+                                       btrfs_list_all_fs_features(0);
+                                       exit(0);
+                               }
+                               break;
+                               }
                        case 's':
                                sectorsize = parse_size(optarg);
                                break;
                        case 'b':
                                block_count = parse_size(optarg);
-                               if (block_count <= 1024*1024*1024) {
-                                       printf("SMALL VOLUME: forcing mixed "
-                                              "metadata/data groups\n");
+                               if (block_count <= BTRFS_MKFS_SMALL_VOLUME_SIZE)
                                        mixed = 1;
-                               }
                                zero_end = 0;
                                break;
                        case 'V':
@@ -1289,59 +1313,181 @@ int main(int ac, char **av)
                                source_dir = optarg;
                                source_dir_set = 1;
                                break;
+                       case 'U':
+                               strncpy(fs_uuid, optarg,
+                                       BTRFS_UUID_UNPARSED_SIZE - 1);
+                               break;
                        case 'K':
-                               nodiscard=1;
+                               discard = 0;
                                break;
+                       case 'q':
+                               verbose = 0;
+                               break;
+                       case GETOPT_VAL_HELP:
                        default:
-                               print_usage();
+                               print_usage(c != GETOPT_VAL_HELP);
                }
        }
-       sectorsize = max(sectorsize, (u32)getpagesize());
-       if (check_leaf_or_node_size(leafsize, sectorsize))
+       sectorsize = max(sectorsize, (u32)sysconf(_SC_PAGESIZE));
+       if (btrfs_check_nodesize(nodesize, sectorsize))
                exit(1);
-       if (check_leaf_or_node_size(nodesize, sectorsize))
+       saved_optind = optind;
+       dev_cnt = ac - optind;
+       if (dev_cnt == 0)
+               print_usage(1);
+
+       if (source_dir_set && dev_cnt > 1) {
+               fprintf(stderr,
+                       "The -r option is limited to a single device\n");
                exit(1);
-       ac = ac - optind;
-       if (ac == 0)
-               print_usage();
+       }
 
-       printf("\nWARNING! - %s IS EXPERIMENTAL\n", BTRFS_BUILD_VERSION);
-       printf("WARNING! - see http://btrfs.wiki.kernel.org before using\n\n");
+       if (*fs_uuid) {
+               uuid_t dummy_uuid;
 
-       if (source_dir == 0) {
+               if (uuid_parse(fs_uuid, dummy_uuid) != 0) {
+                       fprintf(stderr, "could not parse UUID: %s\n", fs_uuid);
+                       exit(1);
+               }
+               if (!test_uuid_unique(fs_uuid)) {
+                       fprintf(stderr, "non-unique UUID: %s\n", fs_uuid);
+                       exit(1);
+               }
+       }
+       
+       while (dev_cnt-- > 0) {
                file = av[optind++];
-               ret = check_mounted(file);
-               if (ret < 0) {
-                       fprintf(stderr, "error checking %s mount status\n", file);
+               if (is_block_device(file))
+                       if (test_dev_for_mkfs(file, force_overwrite))
+                               exit(1);
+       }
+
+       optind = saved_optind;
+       dev_cnt = ac - optind;
+
+       file = av[optind++];
+       ssd = is_ssd(file);
+
+       if (is_vol_small(file) || mixed) {
+               if (verbose)
+                       printf("SMALL VOLUME: forcing mixed metadata/data groups\n");
+               mixed = 1;
+       }
+
+       /*
+       * Set default profiles according to number of added devices.
+       * For mixed groups defaults are single/single.
+       */
+       if (!mixed) {
+               if (!metadata_profile_opt) {
+                       if (dev_cnt == 1 && ssd && verbose)
+                               printf("Detected a SSD, turning off metadata "
+                               "duplication.  Mkfs with -m dup if you want to "
+                               "force metadata duplication.\n");
+
+                       metadata_profile = (dev_cnt > 1) ?
+                                       BTRFS_BLOCK_GROUP_RAID1 : (ssd) ?
+                                       0: BTRFS_BLOCK_GROUP_DUP;
+               }
+               if (!data_profile_opt) {
+                       data_profile = (dev_cnt > 1) ?
+                               BTRFS_BLOCK_GROUP_RAID0 : 0; /* raid0 or single */
+               }
+       } else {
+               u32 best_nodesize = max_t(u32, sysconf(_SC_PAGESIZE), sectorsize);
+
+               if (metadata_profile_opt || data_profile_opt) {
+                       if (metadata_profile != data_profile) {
+                               fprintf(stderr,
+       "ERROR: With mixed block groups data and metadata profiles must be the same\n");
+                               exit(1);
+                       }
+               }
+
+               if (!nodesize_forced) {
+                       nodesize = best_nodesize;
+                       if (btrfs_check_nodesize(nodesize, sectorsize))
+                               exit(1);
+               }
+               if (nodesize != sectorsize) {
+                       fprintf(stderr, "Error: mixed metadata/data block groups "
+                               "require metadata blocksizes equal to the sectorsize\n");
                        exit(1);
                }
-               if (ret == 1) {
-                       fprintf(stderr, "%s is mounted\n", file);
+       }
+
+       /* Check device/block_count after the nodesize is determined */
+       if (block_count && block_count < btrfs_min_dev_size(nodesize)) {
+               fprintf(stderr,
+                       "Size '%llu' is too small to make a usable filesystem\n",
+                       block_count);
+               fprintf(stderr,
+                       "Minimum size for btrfs filesystem is %llu\n",
+                       btrfs_min_dev_size(nodesize));
+               exit(1);
+       }
+       for (i = saved_optind; i < saved_optind + dev_cnt; i++) {
+               char *path;
+
+               path = av[i];
+               ret = test_minimum_size(path, nodesize);
+               if (ret < 0) {
+                       fprintf(stderr, "Failed to check size for '%s': %s\n",
+                               path, strerror(-ret));
+                       exit (1);
+               }
+               if (ret > 0) {
+                       fprintf(stderr,
+                               "'%s' is too small to make a usable filesystem\n",
+                               path);
+                       fprintf(stderr,
+                               "Minimum size for each btrfs device is %llu.\n",
+                               btrfs_min_dev_size(nodesize));
                        exit(1);
                }
-               ac--;
+       }
+       ret = test_num_disk_vs_raid(metadata_profile, data_profile,
+                       dev_cnt, mixed);
+       if (ret)
+               exit(1);
+
+       /* if we are here that means all devs are good to btrfsify */
+       if (verbose) {
+               printf("%s\n", PACKAGE_STRING);
+               printf("See %s for more information.\n\n", PACKAGE_URL);
+       }
+
+       dev_cnt--;
+
+       if (!source_dir_set) {
+               /*
+                * open without O_EXCL so that the problem should not
+                * occur by the following processing.
+                * (btrfs_register_one_device() fails if O_EXCL is on)
+                */
                fd = open(file, O_RDWR);
                if (fd < 0) {
-                       fprintf(stderr, "unable to open %s\n", file);
+                       fprintf(stderr, "unable to open %s: %s\n", file,
+                               strerror(errno));
                        exit(1);
                }
-               first_file = file;
                ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
-                                          block_count, &mixed, nodiscard);
+                                          block_count, &mixed, discard);
+               if (ret) {
+                       close(fd);
+                       exit(1);
+               }
                if (block_count && block_count > dev_block_count) {
                        fprintf(stderr, "%s is smaller than requested size\n", file);
                        exit(1);
                }
        } else {
-               ac = 0;
-               file = av[optind++];
                fd = open_target(file);
                if (fd < 0) {
                        fprintf(stderr, "unable to open the %s\n", file);
                        exit(1);
                }
 
-               first_file = file;
                source_dir_size = size_sourcedir(source_dir, sectorsize,
                                             &num_of_meta_chunks, &size_of_data);
                if(block_count < source_dir_size)
@@ -1351,67 +1497,121 @@ int main(int ac, char **av)
                        fprintf(stderr, "unable to zero the output file\n");
                        exit(1);
                }
+               /* our "device" is the new image file */
+               dev_block_count = block_count;
        }
-       if (mixed) {
-               if (metadata_profile != data_profile) {
-                       fprintf(stderr, "With mixed block groups data and metadata "
-                               "profiles must be the same\n");
-                       exit(1);
-               }
+
+       /* To create the first block group and chunk 0 in make_btrfs */
+       if (dev_block_count < BTRFS_MKFS_SYSTEM_GROUP_SIZE) {
+               fprintf(stderr, "device is too small to make filesystem\n");
+               exit(1);
        }
 
        blocks[0] = BTRFS_SUPER_INFO_OFFSET;
        for (i = 1; i < 7; i++) {
                blocks[i] = BTRFS_SUPER_INFO_OFFSET + 1024 * 1024 +
-                       leafsize * i;
+                       nodesize * i;
        }
 
-       ret = make_btrfs(fd, file, label, blocks, dev_block_count,
-                        nodesize, leafsize,
-                        sectorsize, stripesize);
+       if (group_profile_max_safe_loss(metadata_profile) <
+               group_profile_max_safe_loss(data_profile)){
+               fprintf(stderr,
+                       "WARNING: metatdata has lower redundancy than data!\n\n");
+       }
+
+       /*
+        * FS features that can be set by other means than -O
+        * just set the bit here
+        */
+       if (mixed)
+               features |= BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS;
+
+       if ((data_profile | metadata_profile) &
+           (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) {
+               features |= BTRFS_FEATURE_INCOMPAT_RAID56;
+       }
+
+       mkfs_cfg.label = label;
+       mkfs_cfg.fs_uuid = fs_uuid;
+       memcpy(mkfs_cfg.blocks, blocks, sizeof(blocks));
+       mkfs_cfg.num_bytes = dev_block_count;
+       mkfs_cfg.nodesize = nodesize;
+       mkfs_cfg.sectorsize = sectorsize;
+       mkfs_cfg.stripesize = stripesize;
+       mkfs_cfg.features = features;
+
+       ret = make_btrfs(fd, &mkfs_cfg);
        if (ret) {
-               fprintf(stderr, "error during mkfs %d\n", ret);
+               fprintf(stderr, "error during mkfs: %s\n", strerror(-ret));
                exit(1);
        }
 
-       root = open_ctree(file, 0, O_RDWR);
+       root = open_ctree(file, 0, OPEN_CTREE_WRITES);
        if (!root) {
-               fprintf(stderr, "ctree init failed\n");
+               fprintf(stderr, "Open ctree failed\n");
+               close(fd);
                exit(1);
        }
        root->fs_info->alloc_start = alloc_start;
 
-       ret = make_root_dir(root, mixed);
+       if (dev_cnt == 0) {
+               default_metadata_profile = metadata_profile;
+               default_data_profile = data_profile;
+       } else {
+               /*
+                * Temporary groups to store new device entries
+                */
+               default_metadata_profile = 0;
+               default_data_profile = 0;
+       }
+
+       ret = create_metadata_block_groups(root, default_metadata_profile,
+                       mixed, &allocation);
+       if (ret) {
+               fprintf(stderr, "failed to create default block groups\n");
+               exit(1);
+       }
+
+       trans = btrfs_start_transaction(root, 1);
+       BUG_ON(!trans);
+
+       ret = create_data_block_groups(trans, root, default_data_profile,
+                       mixed, &allocation);
+       if (ret) {
+               fprintf(stderr, "failed to create default data block groups\n");
+               exit(1);
+       }
+
+       ret = make_root_dir(trans, root, mixed, &allocation);
        if (ret) {
                fprintf(stderr, "failed to setup the root directory\n");
                exit(1);
        }
 
+       btrfs_commit_transaction(trans, root);
+
        trans = btrfs_start_transaction(root, 1);
 
-       if (ac == 0)
-               goto raid_groups;
+       if (is_block_device(file))
+               btrfs_register_one_device(file);
 
-       btrfs_register_one_device(file);
+       if (dev_cnt == 0)
+               goto skip_multidev;
 
-       zero_end = 1;
-       while(ac-- > 0) {
+       while (dev_cnt-- > 0) {
                int old_mixed = mixed;
 
                file = av[optind++];
-               ret = check_mounted(file);
-               if (ret < 0) {
-                       fprintf(stderr, "error checking %s mount status\n",
-                               file);
-                       exit(1);
-               }
-               if (ret == 1) {
-                       fprintf(stderr, "%s is mounted\n", file);
-                       exit(1);
-               }
+
+               /*
+                * open without O_EXCL so that the problem should not
+                * occur by the following processing.
+                * (btrfs_register_one_device() fails if O_EXCL is on)
+                */
                fd = open(file, O_RDWR);
                if (fd < 0) {
-                       fprintf(stderr, "unable to open %s\n", file);
+                       fprintf(stderr, "unable to open %s: %s\n", file,
+                               strerror(errno));
                        exit(1);
                }
                ret = btrfs_device_already_in_root(root, fd,
@@ -1423,58 +1623,46 @@ int main(int ac, char **av)
                        continue;
                }
                ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
-                                          block_count, &mixed, nodiscard);
+                                          block_count, &mixed, discard);
+               if (ret) {
+                       close(fd);
+                       exit(1);
+               }
                mixed = old_mixed;
-               BUG_ON(ret);
 
                ret = btrfs_add_to_fsid(trans, root, fd, file, dev_block_count,
                                        sectorsize, sectorsize, sectorsize);
                BUG_ON(ret);
-               btrfs_register_one_device(file);
+               if (verbose >= 2) {
+                       struct btrfs_device *device;
+
+                       device = container_of(root->fs_info->fs_devices->devices.next,
+                                       struct btrfs_device, dev_list);
+                       printf("adding device %s id %llu\n", file,
+                               (unsigned long long)device->devid);
+               }
+
+               if (is_block_device(file))
+                       btrfs_register_one_device(file);
        }
 
-raid_groups:
        if (!source_dir_set) {
                ret = create_raid_groups(trans, root, data_profile,
-                                data_profile_opt, metadata_profile,
-                                metadata_profile_opt, mixed);
+                                metadata_profile, mixed, &allocation);
                BUG_ON(ret);
        }
+skip_multidev:
 
        ret = create_data_reloc_tree(trans, root);
        BUG_ON(ret);
 
-       if (mixed) {
-               struct btrfs_super_block *super = &root->fs_info->super_copy;
-               u64 flags = btrfs_super_incompat_flags(super);
-
-               flags |= BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS;
-               btrfs_set_super_incompat_flags(super, flags);
-       }
-
-       if ((data_profile | metadata_profile) &
-           (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) {
-               struct btrfs_super_block *super = &root->fs_info->super_copy;
-               u64 flags = btrfs_super_incompat_flags(super);
-
-               flags |= BTRFS_FEATURE_INCOMPAT_RAID56;
-               btrfs_set_super_incompat_flags(super, flags);
-               printf("Setting RAID5/6 feature flag\n");
-       }
-
-       printf("fs created label %s on %s\n\tnodesize %u leafsize %u "
-           "sectorsize %u size %s\n",
-           label, first_file, nodesize, leafsize, sectorsize,
-           pretty_buf = pretty_sizes(btrfs_super_total_bytes(&root->fs_info->super_copy)));
-       free(pretty_buf);
-
-       printf("%s\n", BTRFS_BUILD_VERSION);
        btrfs_commit_transaction(trans, root);
 
        if (source_dir_set) {
                trans = btrfs_start_transaction(root, 1);
                ret = create_chunks(trans, root,
-                                   num_of_meta_chunks, size_of_data);
+                                   num_of_meta_chunks, size_of_data,
+                                   &allocation);
                BUG_ON(ret);
                btrfs_commit_transaction(trans, root);
 
@@ -1482,10 +1670,41 @@ raid_groups:
                BUG_ON(ret);
        }
 
+       if (verbose) {
+               char features_buf[64];
+
+               printf("Label:              %s\n", label);
+               printf("UUID:               %s\n", fs_uuid);
+               printf("Node size:          %u\n", nodesize);
+               printf("Sector size:        %u\n", sectorsize);
+               printf("Filesystem size:    %s\n",
+                       pretty_size(btrfs_super_total_bytes(root->fs_info->super_copy)));
+               printf("Block group profiles:\n");
+               if (allocation.data)
+                       printf("  Data:             %-8s %16s\n",
+                               btrfs_group_profile_str(data_profile),
+                               pretty_size(allocation.data));
+               if (allocation.metadata)
+                       printf("  Metadata:         %-8s %16s\n",
+                               btrfs_group_profile_str(metadata_profile),
+                               pretty_size(allocation.metadata));
+               if (allocation.mixed)
+                       printf("  Data+Metadata:    %-8s %16s\n",
+                               btrfs_group_profile_str(data_profile),
+                               pretty_size(allocation.mixed));
+               printf("  System:           %-8s %16s\n",
+                       btrfs_group_profile_str(metadata_profile),
+                       pretty_size(allocation.system));
+               printf("SSD detected:       %s\n", ssd ? "yes" : "no");
+               btrfs_parse_features_to_string(features_buf, features);
+               printf("Incompat features:  %s", features_buf);
+               printf("\n");
+
+               list_all_devices(root);
+       }
+
        ret = close_ctree(root);
        BUG_ON(ret);
-
        free(label);
        return 0;
 }
-