Apple File System Reference

 Developer
Contents
About Apple File System 6

General-Purpose Types 8
paddr_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
prange_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
uuid_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

Objects 9
obj_phys_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Supporting Data Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Object Identifier Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Object Type Masks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Object Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Object Type Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

EFI Jumpstart 20
Booting from an Apple File System Partition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
nx_efi_jumpstart_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Partition UUIDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

Container 24
Mounting an Apple File System Partition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
nx_superblock_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Container Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Optional Container Feature Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Read-Only Compatible Container Feature Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Incompatible Container Feature Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Block and Container Sizes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
nx_counter_id_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
checkpoint_mapping_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
checkpoint_map_phys_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Checkpoint Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
evict_mapping_val_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

Object Map 41
omap_phys_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
omap_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
omap_val_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
omap_snapshot_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Object Map Value Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Snapshot Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Object Map Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Object Map Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Object Map Reaper Phases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

2
Volume 48
apfs_superblock_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
apfs_modified_by_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Volume Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Optional Volume Feature Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Read-Only Compatible Volume Feature Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Incompatible Volume Feature Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

File-System Objects 61
j_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
j_inode_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
j_inode_val_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
j_drec_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
j_drec_hashed_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
j_drec_val_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
j_dir_stats_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
j_dir_stats_val_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
j_xattr_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
j_xattr_val_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

File-System Constants 73
j_obj_types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
j_obj_kinds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
j_inode_flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
j_xattr_flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
dir_rec_flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Inode Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Extended Attributes Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
File-System Object Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
File Extent Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
File Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Directory Entry File Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

Data Streams 88
j_phys_ext_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
j_phys_ext_val_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
j_file_extent_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
j_file_extent_val_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
j_dstream_id_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
j_dstream_id_val_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
j_xattr_dstream_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
j_dstream_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

Extended Fields 94
xf_blob_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
x_field_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Extended-Field Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Extended-Field Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

3
Siblings 100
j_sibling_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
j_sibling_val_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
j_sibling_map_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
j_sibling_map_val_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

Snapshot Metadata 102

j_snap_metadata_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
j_snap_metadata_val_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
j_snap_name_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
j_snap_name_val_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
snap_meta_flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

B-Tree 105
btree_node_phys_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
btree_info_fixed_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
btree_info_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
nloc_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
kvloc_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
kvoff_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
B-Tree Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
B-Tree Table of Contents Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
B-Tree Node Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
B-Tree Node Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

Space Manager 116

chunk_info_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
chunk_info_block . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
cib_addr_block . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
spaceman_free_queue_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
spaceman_free_queue_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
spaceman_device_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
spaceman_phys_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
sfq . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
smdev . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Chunk Info Block Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Internal-Pool Bitmap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

Reaper 120
nx_reaper_phys_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
nx_reap_list_phys_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
nx_reap_list_entry_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
Volume Reaper States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Reaper Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Reaper List Entry Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Reaper List Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
omap_reap_state_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
omap_cleanup_state_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
apfs_reap_state_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

4
Encryption 125
j_crypto_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
j_crypto_val_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
crypto_flags_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
cp_key_class_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
cp_key_os_version_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
cp_key_revision_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
cpx_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
wrapped_crypto_state_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
wrapped_meta_crypto_state_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
Protection Class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Encryption Identifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

Keybag 129
kb_locker_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
keybag_entry_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
media_keybag_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
mk_obj_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Keybag Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Keybag Tags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

Encryption Rolling 131

er_state_phys_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
er_phase_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
gbitmap_block_phys_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
gbitmap_phys_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Encryption-Rolling Checksum Block Sizes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Encryption Rolling Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Encryption-Rolling Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

Fusion 134
fusion_wbc_phys_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
fusion_wbc_list_entry_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
fusion_wbc_list_phys_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
Address Markers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
fusion_mt_key_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
fusion_mt_val_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Fusion Middle-Tree Flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135

Symbol Index 136

Revision History 142

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

5
About Apple File System
Apple File System is the default file format used on Apple platforms. Apple File System is the successor to HFS Plus, so
some aspects of its design intentionally follow HFS Plus to enable data migration from HFS Plus to Apple File System.
Other aspects of its design address limitations with HFS Plus and enable features such as cloning files, snapshots,
encryption, and sharing free space between volumes.

Most apps interact with the file system using high-level interfaces provided by Foundation, which means most de-
velopers donʼt need to read this document. This document is for developers of software that interacts with the file
system directly, without using any frameworks or the operating system—for example, a disk recovery utility or an im-
plementation of Apple File System on another platform. The on-disk data structures described in this document make
up the file system; software that interacts with them defines corresponding in-memory data structures.

Note

If you need to boot from an Apple File System volume, but donʼt need to mount the volume or interact with the
file system directly, read Booting from an Apple File System Partition.

Layered Design

The Apple File System is conceptually divided into two layers, the container layer and the file-system layer. The
container layer organizes file-system layer information and stores higher level information, such as volume metadata,
snapshots of the volume, and encryption state. The file-system layer is made up of the data structures that store
information, such as directory structures, file metadata, and file content. Many types are prefixed with nx_ or j_,
which indicates that theyʼre part of the container layer or the file-system layer, respectively. The abbreviated prefixes
donʼt have a meaningful long form; theyʼre an artifact of how Appleʼs implementation was developed.

There are several design differences between the layers. Container objects are larger, with a typical size measured
in blocks, and contain padding fields that keep data aligned on 64-bit boundaries, to avoid the performance penalty
of unaligned memory access. File-system objects are smaller, with a typical size measured in bytes, and are almost
always packed to minimize space used.

Numbers in both layers are stored on disk in little-endian order. Objects in both layers begin with a common header
that enables object-oriented design patterns in implementations of Apple File System, although the layers have dif-
ferent headers. Container layer objects begin with an instance of obj_phys_t and file-system objects begin with an
instance of j_key_t,

Container Layer

Container objects have an object identifier that you use to locate the object; the steps vary depending on how the
object is stored:

• Physical objects are stored on disk at a particular physical block address.

• Ephemeral objects are stored in memory while the container is mounted and in a checkpoint when the container
isnʼt mounted.
• Virtual objects are stored on disk at a location that you look up in an object map (an instance of omap_phys_t).

The object map includes a B-tree whose keys contain a transaction identifier and an object identifier and whose values
contain a physical block address where the object is stored.

6
About Apple File System

An Apple File System partition has a single container, which provides space management and crash protection. A
container can contain multiple volumes (also known as file systems), each of which contains a directory structure for
files and folders. For example, the figure below shows a storage device that has one Apple File System partition, and
it shows the major divisions of the space inside that container.

Although thereʼs only one container, there are several copies of the container superblock (an instance of nx_super
block_t) stored on disk. These copies hold the state of the container at past points in time. Block zero contains
a copy of the container superblock thatʼs used as part of the mounting process to find the checkpoints. Block zero
is typically a copy of the latest container superblock, assuming the device was properly unmounted and was last
modified by a correct Apple File System implementation. However, in practice, you use the block zero copy only to
find the checkpoints and use the latest version from the checkpoint for everything else.

Within a container, the checkpoint mechanism and the copy-on-write approach to modifying objects enable crash pro-
tection. In-memory state is periodically written to disk in checkpoints, followed by a copy of the container superblock
at that point in time. Checkpoint information is stored in two regions: The checkpoint descriptor area contains in-
stances of checkpoint_map_phys_t and nx_superblock_t, and the checkpoint data area contains ephemeral
objects that represent the in-memory state at the point in time when the checkpoint was written to disk.

When mounting a device, you use the most recent checkpoint information thatʼs valid, as discussed in Mounting
an Apple File System Partition. If the process of writing a checkpoint is interrupted, that checkpoint is invalid and
therefore is ignored the next time the device is mounted, rolling the file system back to the last valid state. Because the
checkpoint stores in-memory state, mounting an Apple File System partition includes reading the ephemeral objects
from the checkpoint back into memory, re-creating that state in memory.

File-System Layer

File-system objects are made up of several records, and each record is stored as a key and value in a B-tree (an
instance of btree_node_phys_t). For example, a typical directory object is made up of an inode record, several
directory entry records, and an extended attributes record. A record contains an object identifier thatʼs used to find it
within the B-tree that contains it.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

7
General-Purpose Types
Basic types that are used in a variety of contexts, and arenʼt associated with any particular functionality.

paddr_t
A physical address of an on-disk block.

typedef int64_t paddr_t;

Negative numbers arenʼt valid addresses. This value is modeled as a signed integer to match IOKit.

prange_t
A range of physical addresses.

struct prange {
paddr_t pr_start_paddr;
uint64_t pr_block_count;
};
typedef struct prange prange_t;

pr_start_paddr

The first block in the range.

paddr_t pr_start_paddr;

pr_block_count

The number of blocks in the range.

uint64_t pr_block_count;

uuid_t
A universally unique identifier.

typedef unsigned char uuid_t[16];

8
Objects
Depending on how theyʼre stored, objects have some differences, the most important of which is the way you use an
object identifier to find an object. At the container level, there are three storage methods for objects:

• Ephemeral objects are stored in memory for a mounted container, and are persisted across unmounts in a
checkpoint. Ephemeral objects for a mounted partition can be modified in place while theyʼre in memory, but
theyʼre always written back to disk as part of a new checkpoint. Theyʼre used for information thatʼs frequently
updated because of the performance benefits of in-place, in-memory changes.

• Physical objects are stored at a known block address on the disk, and are modified by writing the copy to a new
location on disk. Because the object identifier for a physical object is its physical address, this copy-on-write
behavior means that the modified copy has a different object identifier.

• Virtual objects are stored on disk at a block address that you look up using an object map. Virtual objects are
also copied when they are modified; however, both the original and the modified copy have the same object
identifier. When you look up a virtual object in an object map, you use a transaction identifier, in addition to the
object identifier, to specify the point in time that you want.

Regardless of their storage, objects on disk are never modified in place, and modified copies of an object are always
written to a new location on disk. To access an object, you need to know its storage and its identifier. For virtual objects,
you also need a transaction identifier. The storage for an object is almost always implicit from the context in which
that identifier appears. For example, the object identifier for the space manager is stored in the nx_spaceman_oid
field of nx_superblock_t, and the documentation for that field says that the space manager is always an ephemeral
object.

Object identifiers are unique inside the entire container, within their storage method. For example, no two virtual
objects can have the same identifier—even when stored in different object maps—because their storage methods are
the same. However, a virtual object and a physical object can have the same identifier because their storage methods
are different.

For information about determining the identifier for a new object, see oid_t.

obj_phys_t
A header used at the beginning of all objects.

struct obj_phys {
uint8_t o_cksum[MAX_CKSUM_SIZE];
oid_t o_oid;
xid_t o_xid;
uint32_t o_type;
uint32_t o_subtype;
};
typedef struct obj_phys obj_phys_t;

#define MAX_CKSUM_SIZE 8

9
Objects
Supporting Data Types

o_cksum

The Fletcher 64 checksum of the object.

uint8_t o_cksum[MAX_CKSUM_SIZE];

o_oid

The objectʼs identifier.

oid_t o_oid;

o_xid

The identifier of the most recent transaction that this object was modified in.

xid_t o_xid;

o_type

The objectʼs type and flags.

uint32_t o_type;

An object type is a 32-bit value: The low 16 bits indicate the type using the values listed in Object Types, and the high
16 bits are flags using the values listed in Object Type Flags.

o_subtype

The objectʼs subtype.

uint32_t o_subtype;

For the values used in this field, see Object Types.

Subtypes indicate the type of data stored in a data structure such as a B-tree. For example, a node in a B-tree that
contains volume records has a type of OBJECT_TYPE_BTREE_NODE and a subtype of OBJECT_TYPE_FS.

MAX_CKSUM_SIZE

The number of bytes used for an object checksum.

#define MAX_CKSUM_SIZE 8

Supporting Data Types

Types used as unique identifiers within an object.

typedef uint64_t oid_t;

typedef uint64_t xid_t;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

10
Objects
Object Identifier Constants

oid_t

An object identifier.

typedef uint64_t oid_t;

Objects are identified by this number as follows:

• For a physical object, its identifier is the logical block address on disk where the object is stored.

• For an ephemeral object, its identifier is a number.

• For a virtual object, its identifier is a number.

For more information about physical, ephemeral, or virtual objects, see Objects.

To determine the identifier for a new physical object, find a free block using the space manager, and use that blockʼs
address. To determine the identifier for a new ephemeral or virtual object, check the value of nx_superblock_t.
nx_next_oid. New ephemeral and virtual object identifiers must be monotonically increasing.

Note

Although both ephemeral and virtual objects use nx_next_oid field of nx_superblock_t in Appleʼs imple-
mentation, this isnʼt guaranteed or required. Ephemeral and virtual objects are stored in different places, so itʼs
valid to encounter (or create) an ephemeral object and a virtual object that have the same identifier.

xid_t

A transaction identifier.

typedef uint64_t xid_t;

Transactions are uniquely identified by a monotonically increasing number.

The number zero isnʼt a valid transaction identifier. Implementations of Apple File System can use it as a sentinel value
in memory — for example, to refer to the current transaction—but must not let it appear on disk.

This data type is sufficiently large that you arenʼt expected to ever run out of transaction identifiers. For example,
if you created 1,000,000 transactions per second, it would take more than 5,000 centuries to exhaust the available
transaction identifiers.

If a new transaction identifier isnʼt available, thatʼs an unrecoverable error. Identifiers arenʼt allowed to restart from
one or to be reused.

Object Identifier Constants

Constants used for virtual objects that always have a given identifier.

#define OID_NX_SUPERBLOCK 1

#define OID_INVALID 0ULL

#define OID_RESERVED_COUNT 1024

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

11
Objects
Object Type Masks

OID_NX_SUPERBLOCK

The ephemeral object identifier for the container superblock.

#define OID_NX_SUPERBLOCK 1

Although the container superblock is stored in memory like other ephemeral objects, it isnʼt saved on disk in the same
area. For details, see Mounting an Apple File System Partition.

OID_INVALID

An invalid object identifier.

#define OID_INVALID 0ULL

OID_RESERVED_COUNT

The number of object identifiers that are reserved for objects with a fixed object identifier.

#define OID_RESERVED_COUNT 1024

This range of identifiers is reserved for physical, virtual, and ephemeral objects.

Currently, the only object with a reserved identifier is the container superblock, as described in OID_NX_SUPERBLOCK.
All other object identifiers less than OID_RESERVED_COUNT are reserved by Apple.

Object Type Masks

Bit masks used to access specific portions of an object type.

#define OBJECT_TYPE_MASK 0x0000ffff

#define OBJECT_TYPE_FLAGS_MASK 0xffff0000

#define OBJ_STORAGETYPE_MASK 0xc0000000

#define OBJECT_TYPE_FLAGS_DEFINED_MASK 0xf8000000

OBJECT_TYPE_MASK

The bit mask used to access the type.

#define OBJECT_TYPE_MASK 0x0000ffff

For the values that appear in this bit field, see Object Types.

OBJECT_TYPE_FLAGS_MASK

The bit mask used to access the flags.

#define OBJECT_TYPE_FLAGS_MASK 0xffff0000

For the values that appear in this bit field, see Object Type Flags.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

12
Objects
Object Types

OBJ_STORAGETYPE_MASK

The bit mask used to access the storage portion of the object type.

#define OBJ_STORAGETYPE_MASK 0xc0000000

For the values that appear in this bit field, see Object Type Flags.

OBJECT_TYPE_FLAGS_DEFINED_MASK

A bit mask of all bits for which flags are defined.

#define OBJECT_TYPE_FLAGS_DEFINED_MASK 0xf8000000

Object Types
Values used as types and subtypes by the obj_phys_t structure.

#define OBJECT_TYPE_NX_SUPERBLOCK 0x00000001

#define OBJECT_TYPE_BTREE 0x00000002

#define OBJECT_TYPE_BTREE_NODE 0x00000003

#define OBJECT_TYPE_SPACEMAN 0x00000005

#define OBJECT_TYPE_SPACEMAN_CAB 0x00000006
#define OBJECT_TYPE_SPACEMAN_CIB 0x00000007
#define OBJECT_TYPE_SPACEMAN_BITMAP 0x00000008
#define OBJECT_TYPE_SPACEMAN_FREE_QUEUE 0x00000009

#define OBJECT_TYPE_EXTENT_LIST_TREE 0x0000000a

#define OBJECT_TYPE_OMAP 0x0000000b
#define OBJECT_TYPE_CHECKPOINT_MAP 0x0000000c

#define OBJECT_TYPE_FS 0x0000000d

#define OBJECT_TYPE_FSTREE 0x0000000e
#define OBJECT_TYPE_BLOCKREFTREE 0x0000000f
#define OBJECT_TYPE_SNAPMETATREE 0x00000010

#define OBJECT_TYPE_NX_REAPER 0x00000011

#define OBJECT_TYPE_NX_REAP_LIST 0x00000012
#define OBJECT_TYPE_OMAP_SNAPSHOT 0x00000013
#define OBJECT_TYPE_EFI_JUMPSTART 0x00000014
#define OBJECT_TYPE_FUSION_MIDDLE_TREE 0x00000015
#define OBJECT_TYPE_NX_FUSION_WBC 0x00000016
#define OBJECT_TYPE_NX_FUSION_WBC_LIST 0x00000017
#define OBJECT_TYPE_ER_STATE 0x00000018

#define OBJECT_TYPE_GBITMAP 0x00000019

#define OBJECT_TYPE_GBITMAP_TREE 0x0000001a
#define OBJECT_TYPE_GBITMAP_BLOCK 0x0000001b

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

13
Objects
Object Types

#define OBJECT_TYPE_INVALID 0x00000000

#define OBJECT_TYPE_TEST 0x000000ff

The value of obj_phys_t.o_type & OBJECT_TYPE_MASK is one of these constants.

OBJECT_TYPE_NX_SUPERBLOCK

A container superblock (nx_superblock_t).

#define OBJECT_TYPE_NX_SUPERBLOCK 0x00000001

OBJECT_TYPE_BTREE

A B-tree root node (btree_node_phys_t).

#define OBJECT_TYPE_BTREE 0x00000002

OBJECT_TYPE_BTREE_NODE

A B-tree node (btree_node_phys_t).

#define OBJECT_TYPE_BTREE_NODE 0x00000003

OBJECT_TYPE_SPACEMAN

A space manager (spaceman_phys_t).

#define OBJECT_TYPE_SPACEMAN 0x00000005

OBJECT_TYPE_SPACEMAN_CAB

A chunk-info address block (cib_addr_block) used by the space manager.

#define OBJECT_TYPE_SPACEMAN_CAB 0x00000006

OBJECT_TYPE_SPACEMAN_CIB

A chunk-info block (chunk_info_block) used by the space manager.

#define OBJECT_TYPE_SPACEMAN_CIB 0x00000007

OBJECT_TYPE_SPACEMAN_BITMAP

A free-space bitmap used by the space manager.

#define OBJECT_TYPE_SPACEMAN_BITMAP 0x00000008

OBJECT_TYPE_SPACEMAN_FREE_QUEUE

A free-space queue (a mapping from spaceman_free_queue_key_t to spaceman_free_queue_t), used by the

space manager.

#define OBJECT_TYPE_SPACEMAN_FREE_QUEUE 0x00000009

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

14
Objects
Object Types

This type is used only as a subtype of a tree.

OBJECT_TYPE_EXTENT_LIST_TREE

An extents-list tree (a mapping from paddr_t to prange_t).

#define OBJECT_TYPE_EXTENT_LIST_TREE 0x0000000a

The keys are an offset into the logical start of the extent, and the value is the physical location where that data is
stored.

This type is used only as a subtype of a tree.

OBJECT_TYPE_OMAP

As a type, an object map (omap_phys_t); as a subtype, a tree that stores the records of an object map (a mapping
from omap_key_t to omap_val_t).

#define OBJECT_TYPE_OMAP 0x0000000b

OBJECT_TYPE_CHECKPOINT_MAP

A checkpoint map (checkpoint_map_phys_t).

#define OBJECT_TYPE_CHECKPOINT_MAP 0x0000000c

OBJECT_TYPE_FS

A volume (apfs_superblock_t).

#define OBJECT_TYPE_FS 0x0000000d

OBJECT_TYPE_FSTREE

A tree containing file-system records.

#define OBJECT_TYPE_FSTREE 0x0000000e

This type is used only as a subtype of a tree.

The keys and values stored in the tree vary. Each key begins with j_key_t, which contains a field that indicates the
type of that key and its value.

OBJECT_TYPE_BLOCKREFTREE

A tree containing extent references (a mapping from j_phys_ext_key_t to j_phys_ext_val_t).

#define OBJECT_TYPE_BLOCKREFTREE 0x0000000f

This type is used only as a subtype of a tree.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

15
Objects
Object Types

OBJECT_TYPE_SNAPMETATREE

A tree containing snapshot metadata for a volume (a mapping from j_snap_metadata_key_t to j_snap_
metadata_val_t).

#define OBJECT_TYPE_SNAPMETATREE 0x00000010

This type is used only as a subtype of a tree.

OBJECT_TYPE_NX_REAPER

A reaper (nx_reaper_phys_t).

#define OBJECT_TYPE_NX_REAPER 0x00000011

OBJECT_TYPE_NX_REAP_LIST

A reaper list (nx_reap_list_phys_t).

#define OBJECT_TYPE_NX_REAP_LIST 0x00000012

OBJECT_TYPE_OMAP_SNAPSHOT

A tree containing information about snapshots of an object map (a mapping from xid_t to omap_snapshot_t).

#define OBJECT_TYPE_OMAP_SNAPSHOT 0x00000013

This type is used only as a subtype of a tree.

OBJECT_TYPE_EFI_JUMPSTART

EFI information used for booting (nx_efi_jumpstart_t).

#define OBJECT_TYPE_EFI_JUMPSTART 0x00000014

OBJECT_TYPE_FUSION_MIDDLE_TREE

A tree used for Fusion devices to track blocks from the hard drive that are cached on the solid-state drive (a mapping
from fusion_mt_key_t to fusion_mt_val_t).

#define OBJECT_TYPE_FUSION_MIDDLE_TREE 0x00000015

This type is used only as a subtype of a tree.

OBJECT_TYPE_NX_FUSION_WBC

A write-back cache state (fusion_wbc_phys_t) used for Fusion devices.

#define OBJECT_TYPE_NX_FUSION_WBC 0x00000016

OBJECT_TYPE_NX_FUSION_WBC_LIST

A write-back cache list (fusion_wbc_list_phys_t) used for Fusion devices.

#define OBJECT_TYPE_NX_FUSION_WBC_LIST 0x00000017

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

16
Objects
Object Type Flags

OBJECT_TYPE_ER_STATE

An encryption-rolling state (er_state_phys_t).

#define OBJECT_TYPE_ER_STATE 0x00000018

OBJECT_TYPE_GBITMAP

A general-purpose bitmap (gbitmap_phys_t).

#define OBJECT_TYPE_GBITMAP 0x00000019

OBJECT_TYPE_GBITMAP_TREE

A B-tree of general-purpose bitmaps (a mapping from uint64_t to uint64_t).

#define OBJECT_TYPE_GBITMAP_TREE 0x0000001a

This type is used only as a subtype of a tree.

OBJECT_TYPE_GBITMAP_BLOCK

A block containing a general-purpose bitmap (gbitmap_block_phys_t).

#define OBJECT_TYPE_GBITMAP_BLOCK 0x0000001b

OBJECT_TYPE_INVALID

As a type, an invalid object; as a subtype, an object with no subtype.

#define OBJECT_TYPE_INVALID 0x00000000

OBJECT_TYPE_TEST

Reserved for testing.

#define OBJECT_TYPE_TEST 0x000000ff

Donʼt create objects of this type on disk. If you find an object of this type in production, file a bug against the Apple
File System implementation.

This type isnʼt reserved by Apple; non-Apple implementations of Apple File System can use it during testing.

Object Type Flags

The flags used in the object type to provide additional information.

#define OBJ_VIRTUAL 0x00000000

#define OBJ_EPHEMERAL 0x80000000
#define OBJ_PHYSICAL 0x40000000

#define OBJ_NOHEADER 0x20000000

#define OBJ_ENCRYPTED 0x10000000
#define OBJ_NONPERSISTENT 0x08000000

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

17
Objects
Object Type Flags

The value of obj_phys_t.o_type & OBJECT_TYPE_FLAGS_MASK uses these constants. The value of obj_
phys_t.o_type & OBJ_STORAGETYPE_MASK uses only OBJ_VIRTUAL, OBJ_EPHEMERAL, and OBJ_PHYSICAL.

The flags on an objectʼs type must indicate whether the object is virtual, ephemeral, or physical by setting either the
OBJ_EPHEMERAL or OBJ_PHYSICAL flag, or setting neither flag. An object type that contains both flags is invalid.

The absence of both flags indicates a virtual object. The OBJ_VIRTUAL constant is defined to allow code that tests
for virtual objects to match code testing for physical or ephemeral objects, even though thereʼs no corresponding bit
set in the objectʼs type. For example:

obj_phys_t obj = /* assume this exists */

if ((obj.o_type & OBJ_STORAGETYPE_MASK) == OBJ_VIRTUAL) { ... }
elif ((obj.o_type & OBJ_STORAGETYPE_MASK) == OBJ_EPHEMERAL) { ... }
elif ((obj.o_type & OBJ_STORAGETYPE_MASK) == OBJ_PHYSICAL) { ... }
else { /* error */ }

OBJ_VIRTUAL

A virtual object.

#define OBJ_VIRTUAL 0x00000000

OBJ_EPHEMERAL

An ephemeral object.

#define OBJ_EPHEMERAL 0x80000000

OBJ_PHYSICAL

A physical object.

#define OBJ_PHYSICAL 0x40000000

OBJ_NOHEADER

An object stored without an obj_phys_t header.

#define OBJ_NOHEADER 0x20000000

This flag is used, for example, by the space managerʼs bitmap.

OBJ_ENCRYPTED

An encrypted object.

#define OBJ_ENCRYPTED 0x10000000

OBJ_NONPERSISTENT

An ephemeral object that isnʼt persisted across unmounting.

#define OBJ_NONPERSISTENT 0x08000000

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

18
Objects
Object Type Flags

Objects with this flag never appear on disk. If you find an object of this type in production, file a bug against the Apple
File System implementation.

This flag isnʼt reserved by Apple; non-Apple implementations of Apple File System can mark their runtime-only data
structures with OBJ_NONPERSISTENT | OBJ_EPHEMERAL.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

19
EFI Jumpstart
A partition formatted using the Apple File System contains an embedded EFI driver thatʼs used to boot a machine from
that partition.

Booting from an Apple File System Partition

You can locate the EFI driver by reading a few data structures, starting at a known physical address on disk. You donʼt
need any support for reading or mounting Apple File System to locate the EFI driver. This design intentionally simplifies
the steps needed to boot, which means the code needed to boot a piece of hardware or virtualization software can
likewise be simpler. To boot using the embedded EFI driver, do the following:

1. Read physical block zero from the partition. This block contains a copy of the container superblock, which is
an instance of nx_superblock_t.

2. Read the nx_o field of the superblock, which is an instance of obj_phys_t. Then read the o_cksum field
of the nx_o field of the superblock, which contains the Fletcher 64 checksum of the object. Verify that the
checksum is correct.

3. Read the nx_magic field of the superblock. Verify that the fieldʼs value is NX_MAGIC (the four-character code
'BSXN').

4. Read the nx_efi_jumpstart field of the superblock. This field contains the physical block address
(also referred to as the physical object identifier) for the EFI jumpstart information, which is an instance of
nx_efi_jumpstart_t.

5. Read the nej_magic field of the EFI jumpstart information. Verify that the fieldʼs value is NX_EFI_JUMP
START_MAGIC (the four-character code 'RDSJ').

6. Read the nej_o field of the EFI jumpstart information, which is an instance of obj_phys_t. Then read the
o_cksum field of the nej_o field of the jumpstart information, which contains the Fletcher 64 checksum of the
object. Verify that the checksum is correct.

7. Read the nej_version field of the EFI jumpstart information. This field contains the EFI jumpstart version
number. Verify that the fieldʼs value is NX_EFI_JUMPSTART_VERSION (the number 1).

8. Read the nej_efi_file_len field of the jumpstart information. This field contains the length, in bytes, of
the embedded EFI driver. Allocate a contiguous block of memory of at least that size, which youʼll later use to
store the EFI driver.

9. Read the nej_num_extents field of the jumpstart information, and then read that number of prange_t
records from the nej_rec_extents field.

10. Read each extent of the EFI driver into memory, contiguously, in the order theyʼre listed.

11. Load the EFI driver and start executing it.

Implementation Outline

The code listing below shows one way to boot using the embedded EFI driver, assuming the functions listed at the
beginning are defined.

nx_superblock_t* read_superblock(int address) {

20
EFI Jumpstart
Booting from an Apple File System Partition

// Read the given physical block from disk

// and return its contents as a pointer to an nx_superblock_t.
}

nx_efi_jumpstart_t* read_jumpstart(int address) {

// Read the given physical block from disk
// and return its contents as a pointer to an nx_efi_jumpstart_t.
}

void* read_block(int address) {

// Read the given physical block from disk
// and return a pointer to its contents.
}

uint8_t* fletcher64_checksum(void* object) {

// Calculate and return a Fletcher 64 checksum.
}

void load_and_execute(void* address) {

// Load the EFI driver at the specified address
// and then start executing it.
}

int main() {
nx_superblock_t* superblock = read_superblock(0);
assert(superblock->nx_o.o_cksum == fletcher64_checksum(&superblock));
assert(superblock->nx_magic == 'BSXN');

paddr_t jumpstart_address = superblock->nx_efi_jumpstart;

nx_efi_jumpstart_t* jumpstart = read_jumpstart(jumpstart_address);

assert(jumpstart->nej_o.o_cksum == fletcher64_checksum(&jumpstart));
assert(jumpstart->nej_version == 1);

void* efi_driver = malloc(jumpstart->nej_efi_file_len);

void* efi_driver_cursor = efi_driver;

for (int i = 0; i < jumpstart->nej_num_extents; i++) {

prange_t efi_extent_address = jumpstart->nej_rec_extents[i];
for (int j = 0; j < efi_extent_address.pr_block_count; j++) {
void* efi_block = read_block(efi_extent_address.pr_start_paddr + j);
memcpy(efi_driver_cursor, efi_block, superblock->nx_block_size);
efi_driver_cursor += superblock->nx_block_size;
}
}

load_and_execute(efi_driver);

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

21
EFI Jumpstart
nx_efi_jumpstart_t

return 0;
}

nx_efi_jumpstart_t
Information about the embedded EFI driver thatʼs used to boot from an Apple File System partition.

typedef struct nx_efi_jumpstart {

obj_phys_t nej_o;
uint32_t nej_magic;
uint32_t nej_version;
uint32_t nej_efi_file_len;
uint32_t nej_num_extents;
uint64_t nej_reserved[16];
prange_t nej_rec_extents[];
} nx_efi_jumpstart_t;
#define NX_EFI_JUMPSTART_MAGIC 'RDSJ'
#define NX_EFI_JUMPSTART_VERSION 1

nej_o

The objectʼs header.

obj_phys_t nej_o;

nej_magic

A number that can be used to verify that youʼre reading an instance of nx_efi_jumpstart_t.

uint32_t nej_magic;

The value of this field is always NX_EFI_JUMPSTART_MAGIC.

nej_version

The version of this data structure.

uint32_t nej_version;

The value of this field is always NX_EFI_JUMPSTART_VERSION.

nej_efi_file_len

The length, in bytes, of the embedded EFI driver.

uint32_t nej_efi_file_len;

nej_num_extents

The number of extents in the array.

uint32_t nej_num_extents;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

22
EFI Jumpstart
Partition UUIDs

nej_reserved

Reserved.

uint64_t nej_reserved[16];

Populate this field with 0 when you create a new instance, and preserve its value when you modify an existing instance.

nej_rec_extents

The array of extents where the EFI driver is stored.

prange_t nej_rec_extents[];

NX_EFI_JUMPSTART_MAGIC

The value of the nej_magic field.

#define NX_EFI_JUMPSTART_MAGIC 'RDSJ'

This magic number was chosen because in hex dumps it appears as “JSDR”, which is an abbreviated form of jumpstart
driver record.

NX_EFI_JUMPSTART_VERSION

The version number for the EFI jumpstart.

#define NX_EFI_JUMPSTART_VERSION 1

Partition UUIDs
Partition types used in GUID partition table entries.

#define APFS_GPT_PARTITION_UUID ”7C3457EF-0000-11AA-AA11-00306543ECAC”

APFS_GPT_PARTITION_UUID

The partition type for a partition that contains an Apple File System container.

#define APFS_GPT_PARTITION_UUID ”7C3457EF-0000-11AA-AA11-00306543ECAC”

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

23
Container
The container includes several top-level objects that are shared by all of the containerʼs volumes:

• Checkpoint description and data areas store ephemeral objects in a way that provides crash protection. At the
end of each transaction, new state is saved by writing a checkpoint.

• The space manager keeps track of available space within the container and is used to allocate and free blocks
that store objects and file data.

• The reaper manages the deletion of objects that are too large to be deleted in the time between transactions.
It keeps track of the deletion state so these objects can be deleted across multiple transactions.

The container superblock describes the location of all of these objects.

Because a single container can have multiple volumes, configurations that would require multiple partitions under
other file systems can usually share a single partition with Apple File System. For example, a drive can be configured
with two bootable volumes—one with a shipping version of macOS and one with a beta version—as well as a user data
volume. All three of these volumes share free space, meaning you donʼt have to decide ahead of time how to divide
space between them.

Mounting an Apple File System Partition

To mount the volumes of a partition thatʼs formatted using the Apple File System, do the following:

1. Read block zero of the partition. This block contains a copy of the container superblock (an instance of
nx_superblock_t). It might be a copy of the latest version or an old version, depending on whether the
drive was unmounted cleanly.

2. Use the block-zero copy of the container superblock to locate the checkpoint descriptor area by reading the
nx_xp_desc_base field.

3. Read the entries in the checkpoint descriptor area, which are instances of checkpoint_map_phys_t or
nx_superblock_t.

4. Find the container superblock that has the largest transaction identifier and isnʼt malformed. For example, con-
firm that its magic number and checksum are valid. That superblock and its checkpoint-mapping blocks com-
prise the latest valid checkpoint. The superblockʼs fields, such as nx_xp_desc_blocks and nx_data_len,
indicate which checkpoint-mapping blocks belong to that superblock.

Note

The checkpoint description area is a ring buffer stored as an array. Walking backward from the latest valid
superblock to read all of its checkpoint-mapping blocks sometimes requires wrapping around from the
first block to the last block.

5. Read the ephemeral objects listed in the checkpoint from the checkpoint data area into memory. If any of the
ephemeral objects is malformed, the checkpoint that contains that object is malformed; go back to the previous
step and mount from an older checkpoint.

24
Container
nx_superblock_t

The details of this step vary. For example, if youʼre mounting the partition read-only and performance isnʼt a
consideration, you can skip this step and read from the checkpoint every time you need to access an ephemeral
object.

6. Locate the container object map using the nx_omap_oid field of the container superblock.

7. Read the list of volumes from the nx_fs_oid field of the container superblock. If youʼre mounting only a
particular volume, you can ignore the virtual object identifiers for the other volumes.

8. For each volume, look up the specified virtual object identifier in the container object map to locate the volume
superblock (an instance of apfs_superblock_t). If youʼre mounting only a particular volume, you can skip
this step for the other volumes.

9. For each volume, read the root file system treeʼs virtual object identifier from the apfs_root_tree_oid field,
and then look it up in the volume object map indicated by the apfs_omap_oid field. If youʼre mounting only a
particular volume, you can skip this step for the other volumes.

10. Walk the root file system tree as needed by your implementation to mount the file system.

nx_superblock_t
A container superblock.

struct nx_superblock {
obj_phys_t nx_o;
uint32_t nx_magic;
uint32_t nx_block_size;
uint64_t nx_block_count;

uint64_t nx_features;
uint64_t nx_readonly_compatible_features;
uint64_t nx_incompatible_features;

uuid_t nx_uuid;

oid_t nx_next_oid;
xid_t nx_next_xid;

uint32_t nx_xp_desc_blocks;
uint32_t nx_xp_data_blocks;
paddr_t nx_xp_desc_base;
paddr_t nx_xp_data_base;
uint32_t nx_xp_desc_next;
uint32_t nx_xp_data_next;
uint32_t nx_xp_desc_index;
uint32_t nx_xp_desc_len;
uint32_t nx_xp_data_index;
uint32_t nx_xp_data_len;

oid_t nx_spaceman_oid;
oid_t nx_omap_oid;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

25
Container
nx_superblock_t

oid_t nx_reaper_oid;

uint32_t nx_test_type;

uint32_t nx_max_file_systems;
oid_t nx_fs_oid[NX_MAX_FILE_SYSTEMS];
uint64_t nx_counters[NX_NUM_COUNTERS];
prange_t nx_blocked_out_prange;
oid_t nx_evict_mapping_tree_oid;
uint64_t nx_flags;
paddr_t nx_efi_jumpstart;
uuid_t nx_fusion_uuid;
prange_t nx_keylocker;
uint64_t nx_ephemeral_info[NX_EPH_INFO_COUNT];

oid_t nx_test_oid;

oid_t nx_fusion_mt_oid;
oid_t nx_fusion_wbc_oid;
prange_t nx_fusion_wbc;
}
typedef struct nx_superblock nx_superblock_t;

#define NX_MAGIC 'BSXN'

#define NX_MAX_FILE_SYSTEMS 100

#define NX_EPH_INFO_COUNT 4
#define NX_EPH_MIN_BLOCK_COUNT 8
#define NX_MAX_FILE_SYSTEM_EPH_STRUCTS 4
#define NX_TX_MIN_CHECKPOINT_COUNT 4
#define NX_EPH_INFO_VERSION_1 1

Note that all fields are 64-bit aligned.

nx_o

The objectʼs header.

obj_phys_t nx_o;

nx_magic

A number that can be used to verify that youʼre reading an instance of nx_superblock_t.

uint32_t nx_magic;

The value of this field is always NX_MAGIC.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

26
Container
nx_superblock_t

nx_block_size

The logical block size used in the Apple File System container.

uint32_t nx_block_size;

This size is often the same as the block size used by the underlying storage device, but it can also be an integer
multiple of the deviceʼs block size.

nx_block_count

The total number of logical blocks available in the container.

uint64_t nx_block_count;

nx_features

A bit field of the optional features being used by this container.

uint64_t nx_features;

For the values used in this bit field, see Optional Container Feature Flags.

If your implementation doesnʼt implement an optional feature thatʼs in use, ignore that feature in this list and mount
the containerʼs volumes as usual.

nx_readonly_compatible_features

A bit field of the read-only compatible features being used by this container.

uint64_t nx_readonly_compatible_features;

For the values used in this bit field, see Read-Only Compatible Container Feature Flags.

If your implementation doesnʼt implement a read-only compatible feature thatʼs in use, mount the containerʼs volumes
as read-only.

nx_incompatible_features

A bit field of the backward-incompatible features being used by this container.

uint64_t nx_incompatible_features;

For the values used in this bit field, see Incompatible Container Feature Flags.

If your implementation doesnʼt implement a read-only feature thatʼs in use, it must not mount the containerʼs volumes.

nx_uuid

The universally unique identifier of this container.

uuid_t nx_uuid;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

27
Container
nx_superblock_t

nx_next_oid

The next object identifier to be used for a new ephemeral or virtual object.

oid_t nx_next_oid;

nx_next_xid

The next transaction to be used.

xid_t nx_next_xid;

nx_xp_desc_blocks

The number of blocks used by the checkpoint descriptor area.

uint32_t nx_xp_desc_blocks;

The highest bit of this number is used as a flag, as discussed in nx_xp_desc_base. Ignore that bit when accessing
this field as a count.

nx_xp_data_blocks

The number of blocks used by the checkpoint data area.

uint32_t nx_xp_data_blocks;

The highest bit of this number is used as a flag, as discussed in nx_xp_data_base. Ignore that bit when accessing
this field as a count.

nx_xp_desc_base

Either the base address of the checkpoint descriptor area or the physical object identifier of a tree that contains the
address information.

paddr_t nx_xp_desc_base;

If the highest bit of nx_xp_desc_blocks is 0, the checkpoint descriptor area is contiguous and this field contains
the address of the first block. Otherwise, the checkpoint descriptor area isnʼt contiguous and this field contains the
physical object identifier of a B-tree. The treeʼs keys are block offsets into the checkpoint descriptor area, and its
values are instances of prange_t that contain the fragmentʼs size and location.

nx_xp_data_base

Either the base address of the checkpoint data area or the physical object identifier of a tree that contains the address
information.

paddr_t nx_xp_data_base;

If the highest bit of nx_xp_data_blocks is 0, the checkpoint data area is contiguous and this field contains the
address of the first block. Otherwise, the checkpoint data area isnʼt contiguous and this field contains the object
identifier of a B-tree. The treeʼs keys are block offsets into the checkpoint data area, and its values are instances of
prange_t that contain the fragmentʼs size and location.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

28
Container
nx_superblock_t

nx_xp_desc_next

The next index to use in the checkpoint descriptor area.

uint32_t nx_xp_desc_next;

If the superblock is part of a checkpoint, this field must have a value. Otherwise, ignore the value of this field when
reading, and use 0 as the value when creating a new instance. For example, this field has no meaning for the copy of
the superblock thatʼs stored in block zero.

nx_xp_data_next

The next index to use in the checkpoint data area.

uint32_t nx_xp_data_next;

If the superblock is part of a checkpoint, this field must have a value. Otherwise, ignore the value of this field when
reading, and use 0 as the value when creating a new instance. For example, this field has no meaning for the copy of
the superblock thatʼs stored in block zero.

nx_xp_desc_index

The index of the first valid item in the checkpoint descriptor area.

uint32_t nx_xp_desc_index;

If the superblock is part of a checkpoint, this field must have a value. Otherwise, ignore the value of this field when
reading, and use 0 as the value when creating a new instance. For example, this field has no meaning for the copy of
the superblock thatʼs stored in block zero.

nx_xp_desc_len

The number of blocks in the checkpoint descriptor area used by the checkpoint that this superblock belongs to.

uint32_t nx_xp_desc_len;

If the superblock is part of a checkpoint, this field must have a value. Otherwise, ignore the value of this field when
reading, and use 0 as the value when creating a new instance. For example, this field has no meaning for the copy of
the superblock thatʼs stored in block zero.

nx_xp_data_index

The index of the first valid item in the checkpoint data area.

uint32_t nx_xp_data_index;

If the superblock is part of a checkpoint, this field must have a value. Otherwise, ignore the value of this field when
reading, and use 0 as the value when creating a new instance. For example, this field has no meaning for the copy of
the superblock thatʼs stored in block zero.

nx_xp_data_len

The number of blocks in the checkpoint data area used by the checkpoint that this superblock belongs to.

uint32_t nx_xp_data_len;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

29
Container
nx_superblock_t

If the superblock is part of a checkpoint, this field must have a value. Otherwise, ignore the value of this field when
reading, and use 0 as the value when creating a new instance. For example, this field has no meaning for the copy of
the superblock thatʼs stored in block zero.

nx_spaceman_oid

The ephemeral object identifier for the space manager.

oid_t nx_spaceman_oid;

nx_omap_oid

The physical object identifier for the containerʼs object map.

oid_t nx_omap_oid;

nx_reaper_oid

The ephemeral object identifier for the reaper.

oid_t nx_reaper_oid;

nx_test_type

Reserved for testing.

uint32_t nx_test_type;

This field never has a value other than 0 on disk. If you find another value in production, file a bug against the Apple
File System implementation.

This field isnʼt reserved by Apple; non-Apple implementations of Apple File System can use it to store an object type
during testing.

nx_max_file_systems

The maximum number of volumes that can be stored in this container.

uint32_t nx_max_file_systems;

To calculate this value, divide the size of the container by 512 MiB and round up. For example, a container with 1.3 GiB
of space can contain three volumes. This value must not be larger than the value of NX_MAX_FILE_SYSTEMS.

nx_fs_oid

An array of virtual object identifiers for volumes.

oid_t nx_fs_oid[NX_MAX_FILE_SYSTEMS];

nx_counters

An array of counters that store information about the container.

uint64_t nx_counters[NX_NUM_COUNTERS];

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

30
Container
nx_superblock_t

These counters are primarily intended to help during development and debugging of Apple File System implementa-
tions. For the meaning of these counters, see nx_counter_id_t.

nx_blocked_out_prange

The physical range of blocks where space will not be allocated.

prange_t nx_blocked_out_prange;

This field is used with nx_evict_mapping_tree_oid while shrinking a partition. If nothing is currently blocked
out, the value of nx_blocked_out_prange.pr_block_count is 0 and the value of nx_blocked_out_prange.
pr_start_paddr is ignored.

nx_evict_mapping_tree_oid

The physical object identifier of a tree used to keep track of objects that must be moved out of blocked-out storage.

oid_t nx_evict_mapping_tree_oid;

The keys in this tree are physical addresses of blocks that must be moved, and the values are instances of
evict_mapping_val_t that describe where the blocks are being moved to.

This identifier is valid only while shrinking a partition. First, the blocks to be removed from the partition are added to the
nx_blocked_out_prange field. Next, every object thatʼs stored in a blocked-out range is added to this tree. Finally,
every object in this tree has space allocated and is moved into the new space. Because the space manager honors
the blocked-out range, data is never moved from one blocked-out address to another address thatʼs also blocked out.
After all data has been removed from the blocked-out range and this tree is empty, the partition shrinks and the block
count of nx_blocked_out_prange is set to 0, which clears the field.

nx_flags

Other container flags.

uint64_t nx_flags;

For the values used in this bit field, see Container Flags.

nx_efi_jumpstart

The physical object identifier of the object that contains EFI driver data extents.

paddr_t nx_efi_jumpstart;

The object is an instance of nx_efi_jumpstart_t.

nx_fusion_uuid

The universally unique identifier of the containerʼs Fusion set, or 0 for non-Fusion containers.

uuid_t nx_fusion_uuid;

The hard drive and the solid-state drive each have a partition, which combine to make a single container. Each partition
has its own copy of the container superblock at block zero, and each copy has the same value for the low 127 bits of
this field. The highest bit is 1 for the Fusion setʼs main device and 0 for the second-tier device.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

31
Container
nx_superblock_t

nx_keylocker

The range of blocks where cryptographic key information is stored.

prange_t nx_keylocker;

nx_ephemeral_info

An array of fields used in the management of ephemeral data.

uint64_t nx_ephemeral_info[NX_EPH_INFO_COUNT];

The first array entry records information about how the checkpoint data areaʼs size was chosen as follows:

nx_ephemeral_info[0] = (min_block_count << 32)

| ((NX_MAX_FILE_SYSTEM_EPH_STRUCTS & 0xFFFF) << 16)
| NX_EPH_INFO_VERSION_1;

The value of min_block_count depends on the size of the container. If the container is larger than 128 MiB, it takes
the value of NX_EPH_MIN_BLOCK_COUNT. Otherwise, it takes the value of spaceman_phys_t.sm_fq[SFQ_MAIN]
.sfq_tree_node_limit from the space manager.

nx_test_oid

Reserved for testing.

oid_t nx_test_oid;

This field never has a value other than 0 on disk. If you find another value in production, file a bug against the Apple
File System implementation.

This field isnʼt reserved by Apple; non-Apple implementations of Apple File System can use it to store an object iden-
tifier during testing.

nx_fusion_mt_oid

The physical object identifier of the Fusion middle tree (a B-tree mapping fusion_mt_key_t to fusion_mt_val_t),
or 0 if for non-Fusion drives.

oid_t nx_fusion_mt_oid;

nx_fusion_wbc_oid

The ephemeral object identifier of the Fusion write-back cache state (fusion_wbc_phys_t), or 0 for non-Fusion
drives.

oid_t nx_fusion_wbc_oid;

nx_fusion_wbc

The blocks used for the Fusion write-back cache area, or 0 for non-Fusion drives.

prange_t nx_fusion_wbc;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

32
Container
nx_superblock_t

NX_MAGIC

The value of the nx_magic field.

#define NX_MAGIC 'BSXN'

This magic number was chosen because in hex dumps it appears as “NXSB”, which is an abbreviated form of NX
superblock.

NX_MAX_FILE_SYSTEMS

The maximum number of volumes that can be in a single container.

#define NX_MAX_FILE_SYSTEMS 100

NX_EPH_INFO_COUNT

The length of the array in the nx_ephemeral_info field.

#define NX_EPH_INFO_COUNT 4

NX_EPH_MIN_BLOCK_COUNT

The default minimum size, in blocks, for structures that contain ephemeral data.

#define NX_EPH_MIN_BLOCK_COUNT 8

This value is used when choosing the size for a new containerʼs checkpoint data area, and the value used is recorded
in the nx_ephemeral_info field.

NX_MAX_FILE_SYSTEM_EPH_STRUCTS

The number of structures that contain ephemeral data that a volume can have.

#define NX_MAX_FILE_SYSTEM_EPH_STRUCTS 4

This value is used when choosing the size for a new containerʼs checkpoint data area, and the value used is recorded
in the nx_ephemeral_info field.

NX_TX_MIN_CHECKPOINT_COUNT

The minimum number of checkpoints that can fit in the checkpoint data area.

#define NX_TX_MIN_CHECKPOINT_COUNT 4

This value is used when choosing the size for a new containerʼs checkpoint data area.

NX_EPH_INFO_VERSION_1

The version number for structures that contain ephemeral data.

#define NX_EPH_INFO_VERSION_1 1

This value is recorded in the nx_ephemeral_info field.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

33
Container
Container Flags

Container Flags
The flags used for general information about a container.

#define NX_RESERVED_1 0x00000001LL

#define NX_RESERVED_2 0x00000002LL
#define NX_CRYPTO_SW 0x00000004LL

These flags are used by the nx_flags field of nx_superblock_t.

NX_RESERVED_1

Reserved.

#define NX_RESERVED_1 0x00000001LL

Donʼt set this flag, but preserve it if itʼs already set.

NX_RESERVED_2

Reserved.

#define NX_RESERVED_2 0x00000002LL

Donʼt add this flag to a container. If this flag is already set, remove it when you modify the container, and preserve it
otherwise.

NX_CRYPTO_SW

The container uses software cryptography.

#define NX_CRYPTO_SW 0x00000004LL

See also CRYPTO_SW_ID.

Optional Container Feature Flags

The flags used to describe optional features of an Apple File System container.

#define NX_FEATURE_DEFRAG 0x0000000000000001ULL

#define NX_FEATURE_LCFD 0x0000000000000002ULL
#define NX_SUPPORTED_FEATURES_MASK (NX_FEATURE_DEFRAG | NX_FEATURE_LCFD)

These flags are used by the nx_features field of nx_superblock_t.

NX_FEATURE_DEFRAG

The volumes in this container support defragmentation.

#define NX_FEATURE_DEFRAG 0x0000000000000001ULL

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

34
Container
Read-Only Compatible Container Feature Flags

NX_FEATURE_LCFD

This container is using low-capacity Fusion Drive mode.

#define NX_FEATURE_LCFD 0x0000000000000002ULL

Low-capacity Fusion Drive mode is enabled when the solid-state drive has a smaller capacity and so the cache must
be smaller.

NX_SUPPORTED_FEATURES_MASK

A bit mask of all the optional features.

#define NX_SUPPORTED_FEATURES_MASK (NX_FEATURE_DEFRAG | NX_FEATURE_LCFD)

Read-Only Compatible Container Feature Flags

The flags used to describe read-only compatible features of an Apple File System container.

#define NX_SUPPORTED_ROCOMPAT_MASK (0x0ULL)

These flags are used by the nx_readonly_compatible_features field of nx_superblock_t. There are cur-
rently none defined.

NX_SUPPORTED_ROCOMPAT_MASK

A bit mask of all read-only compatible features.

#define NX_SUPPORTED_ROCOMPAT_MASK (0x0ULL)

Incompatible Container Feature Flags

The flags used to describe backward-incompatible features of an Apple File System container.

#define NX_INCOMPAT_VERSION1 0x0000000000000001ULL

#define NX_INCOMPAT_VERSION2 0x0000000000000002ULL
#define NX_INCOMPAT_FUSION 0x0000000000000100ULL
#define NX_SUPPORTED_INCOMPAT_MASK (NX_INCOMPAT_VERSION2 | NX_INCOMPAT_FUSION)

These flags are used by the nx_incompatible_features field of nx_superblock_t.

NX_INCOMPAT_VERSION1

The container uses version 1 of Apple File System, as implemented in macOS 10.12.

#define NX_INCOMPAT_VERSION1 0x0000000000000001ULL

Important

Version 1 of the Apple File System was a prerelease thatʼs incompatible with later versions. This document
describes only version 2 and later.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

35
Container
Block and Container Sizes

NX_INCOMPAT_VERSION2

The container uses version 2 of Apple File System, as implemented in macOS 10.13 and iOS 10.3.

#define NX_INCOMPAT_VERSION2 0x0000000000000002ULL

NX_INCOMPAT_FUSION

The container supports Fusion Drives.

#define NX_INCOMPAT_FUSION 0x0000000000000100ULL

NX_SUPPORTED_INCOMPAT_MASK

A bit mask of all the backward-incompatible features.

#define NX_SUPPORTED_INCOMPAT_MASK (NX_INCOMPAT_VERSION2 | NX_INCOMPAT_FUSION)

Block and Container Sizes

The block size for a container is defined by the nx_block_size field of nx_superblock_t. It can vary as described
below.

#define NX_MINIMUM_BLOCK_SIZE 4096

#define NX_DEFAULT_BLOCK_SIZE 4096
#define NX_MAXIMUM_BLOCK_SIZE 65536

#define NX_MINIMUM_CONTAINER_SIZE 1048576

NX_MINIMUM_BLOCK_SIZE

The smallest supported size, in bytes, for a block.

#define NX_MINIMUM_BLOCK_SIZE 4096

If you try to define a block size thatʼs too small, some data structures wonʼt be able to fit in a single block.

NX_DEFAULT_BLOCK_SIZE

The default size, in bytes, for a block.

#define NX_DEFAULT_BLOCK_SIZE 4096

NX_MAXIMUM_BLOCK_SIZE

The largest supported size, in bytes, for a block.

#define NX_MAXIMUM_BLOCK_SIZE 65536

If you try to define a block size thatʼs too large, parts of the block will be outside of the range of a 16-bit address.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

36
Container
nx_counter_id_t

NX_MINIMUM_CONTAINER_SIZE

The smallest supported size, in bytes, for a container.

#define NX_MINIMUM_CONTAINER_SIZE 1048576

This value is slightly less that the capacity of a floppy disk. For a container this size, statically allocated metadata
takes up about a third of the available space.

nx_counter_id_t
Indexes into a container superblockʼs array of counters.

typedef enum {
NX_CNTR_OBJ_CKSUM_SET = 0,
NX_CNTR_OBJ_CKSUM_FAIL = 1,

NX_NUM_COUNTERS = 32
} nx_counter_id_t;

These values are used as indexes into the array stored in the nx_counters field of nx_superblock_t.

NX_CNTR_OBJ_CKSUM_SET

The number of times a checksum has been computed while writing objects to disk.

NX_CNTR_OBJ_CKSUM_SET = 0

NX_CNTR_OBJ_CKSUM_FAIL

The number of times an objectʼs checksum was invalid when reading from disk.

NX_CNTR_OBJ_CKSUM_FAIL = 1

NX_NUM_COUNTERS

The maximum number of counters.

NX_NUM_COUNTERS = 32

checkpoint_mapping_t
A mapping from an ephemeral object identifier to its physical address in the checkpoint data area.

struct checkpoint_mapping {
uint32_t cpm_type;
uint32_t cpm_subtype;
uint32_t cpm_size;
uint32_t cpm_pad;
oid_t cpm_fs_oid;
oid_t cpm_oid;
oid_t cpm_paddr;
};

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

37
Container
checkpoint_mapping_t

typedef struct checkpoint_mapping checkpoint_mapping_t;

cpm_type

The objectʼs type.

uint32_t cpm_type;

An object type is a 32-bit value: The low 16 bits indicate the type using the values listed in Object Types, and the high
16 bits are flags using the values listed in Object Type Flags.

This field has the same meaning and behavior as the o_type field of obj_phys_t.

cpm_subtype

The objectʼs subtype.

uint32_t cpm_subtype;

One of the values listed in Object Types.

Subtypes indicate the type of data stored in a data structure such as a B-tree. For example, a leaf node in a B-tree that
contains file-system records has a type of OBJECT_TYPE_BTREE_NODE and a subtype of OBJECT_TYPE_FSTREE.

This field has the same meaning and behavior as the o_subtype field of obj_phys_t.

cpm_size

The size, in bytes, of the object.

uint32_t cpm_size;

cpm_pad

Reserved.

uint32_t cpm_pad;

Populate this field with 0 when you create a new mapping, and preserve its value when you modify an existing mapping.

This field is padding.

cpm_fs_oid

The virtual object identifier of the volume that the object is associated with.

oid_t cpm_fs_oid;

cpm_oid

The ephemeral object identifier.

oid_t cpm_oid;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

38
Container
checkpoint_map_phys_t

cpm_paddr

The address in the checkpoint data area where the object is stored.

oid_t cpm_paddr;

checkpoint_map_phys_t
A checkpoint-mapping block.

struct checkpoint_map_phys {
obj_phys_t cpm_o;
uint32_t cpm_flags;
uint32_t cpm_count;
checkpoint_mapping_t cpm_map[];
};

If a checkpoint needs to store more mappings than a single block can hold, the checkpoint has multiple checkpoint-
mapping blocks stored contiguously in the checkpoint descriptor area. The last checkpoint-mapping block is marked
with the CHECKPOINT_MAP_LAST flag.

cpm_o

The objectʼs header.

obj_phys_t cpm_o;

cpm_flags

A bit field that contains additional information about the list of checkpoint mappings.

uint32_t cpm_flags;

For the values used in this bit field, see Checkpoint Flags.

cpm_count

The number of checkpoint mappings in the array.

uint32_t cpm_count;

cpm_map

The array of checkpoint mappings.

checkpoint_mapping_t cpm_map[];

Checkpoint Flags
The flags used by a checkpoint-mapping block.

#define CHECKPOINT_MAP_LAST 0x00000001

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

39
Container
evict_mapping_val_t

CHECKPOINT_MAP_LAST

A flag marking the last checkpoint-mapping block in a given checkpoint.

#define CHECKPOINT_MAP_LAST 0x00000001

evict_mapping_val_t
A range of physical addresses that data is being moved into.

struct evict_mapping_val {
paddr_t dst_paddr;
uint64_t len;
} __attribute__((packed));
typedef struct evict_mapping_val evict_mapping_val_t;

This data type is used by the evict-mapping tree, which is accessed through the nx_evict_mapping_tree_oid
field of nx_superblock_t.

dst_paddr

The address where the destination starts.

paddr_t dst_paddr;

len

The number of blocks being moved.

uint64_t len;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

40
Object Map
An object map uses a B-tree to store a mapping from virtual object identifiers and transaction identifiers to the physical
addresses where those objects are stored. The keys in the B-tree are instances of omap_key_t and the values are
instances of paddr_t.

To access a virtual object using the object map, perform the following operations:

1. Determine which object map to use. Objects that are within a volume use that volumeʼs object map, and all
other objects use the containerʼs object map.

2. Locate the object map for the volume by reading the apfs_omap_oid field of apfs_superblock_t or the
nx_omap_oid field of nx_superblock_t.

3. Locate the B-tree for the object map by reading the om_tree_oid field of omap_phys_t.

4. Search the B-tree for a key whose object identifier is the same as the desired object identifier, and whose
transaction identifier is less than or equal to the desired transaction identifier. If there are multiple keys that
satisfy this test, use the key with the largest transaction identifier.

5. Using the table of contents entry, read the corresponding value for the key you found, which contains a physical
address.

6. Read the object from disk at that address.

For example, assume the object mapʼs B-tree contains the following mappings:

OID 588, XID 2101 -> Address 200

OID 588, XID 2202 -> Address 300
OID 588, XID 2300 -> Address 100

To access object 588 as of transaction 2300, you use the last entry—its object and transaction identifiers match
exactly—and read physical address 100.

To access object 588 as of transaction 2290, you use the second entry. Thereʼs no entry with the transaction identifier
2290, and 2202 is the largest transaction identifier in the object map thatʼs still less than 2290. That entry tells you
to read physical address 300.

omap_phys_t
An object map.

struct omap_phys {
obj_phys_t om_o;
uint32_t om_flags;
uint32_t om_snap_count;
uint32_t om_tree_type;
uint32_t om_snapshot_tree_type;
oid_t om_tree_oid;
oid_t om_snapshot_tree_oid;
xid_t om_most_recent_snap;
xid_t om_pending_revert_min;

41
Object Map
omap_phys_t

xid_t om_pending_revert_max;
};
typedef struct omap_phys omap_phys_t;

om_o

The objectʼs header.

obj_phys_t om_o;

om_flags

The object mapʼs flags.

uint32_t om_flags;

For the values used in this bit field, see Object Map Flags.

om_tree_type

The type of tree being used for object mappings.

uint32_t om_tree_type;

om_tree_oid

The virtual object identifier of the tree being used for object mappings.

oid_t om_tree_oid;

om_snapshot_tree_oid

The virtual object identifier of the tree being used to hold snapshot information.

oid_t om_snapshot_tree_oid;

om_snapshot_tree_type

The type of tree being used for snapshots.

uint32_t om_snapshot_tree_type;

om_snap_count

The number of snapshots that this object map has.

uint32_t om_snap_count;

om_most_recent_snap

The transaction identifier of the most recent snapshot thatʼs stored in this object map.

xid_t om_most_recent_snap;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

42
Object Map
omap_key_t

om_pending_revert_min

The smallest transaction identifier for an in-progress revert.

xid_t om_pending_revert_min;

om_pending_revert_max

The largest transaction identifier for an in-progress revert.

xid_t om_pending_revert_max;

omap_key_t
A key used to access an entry in the object map.

struct omap_key {
oid_t ok_oid;
xid_t ok_xid;
};
typedef struct omap_key omap_key_t;

ok_oid

The object identifier.

oid_t ok_oid;

ok_xid

The transaction identifier.

xid_t ok_xid;

omap_val_t
A value in the object map.

struct omap_val {
uint32_t ov_flags;
uint32_t ov_size;
paddr_t ov_paddr;
};
typedef struct omap_val omap_val_t;

ov_flags

A bit field of flags.

uint32_t ov_flags;

For the values used in this bit field, see Object Map Value Flags.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

43
Object Map
omap_snapshot_t

ov_size

The size, in bytes, of the object.

uint32_t ov_size;

This value must be a multiple of the containerʼs logical block size. If the object is smaller than one logical block, the
value of this field is the size of one logical block.

ov_paddr

The address of the object.

paddr_t ov_paddr;

omap_snapshot_t
Information about a snapshot of an object map.

struct omap_snapshot {
uint32_t oms_flags;
uint32_t oms_pad;
oid_t oms_oid;
};
typedef struct omap_snapshot omap_snapshot_t;

When accessing or storing a snapshot in the snapshot tree, use the transaction identifier as the key. This structure is
the value stored in a snapshot tree.

oms_flags

The snapshotʼs flags.

uint32_t oms_flags;

For the values used in this bit field, see Snapshot Flags.

oms_pad

Reserved.

uint32_t oms_pad;

Populate this field with 0 when you create a new snapshot, and preserve its value when you modify an existing
snapshot.

This field is padding.

oms_oid

Reserved.

oid_t oms_oid;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

44
Object Map
Object Map Value Flags

Populate this field with 0 when you create a new snapshot, and preserve its value when you modify an existing
snapshot.

Object Map Value Flags

The flags used by entries in the object map.

#define OMAP_VAL_DELETED 0x00000001

#define OMAP_VAL_SAVED 0x00000002
#define OMAP_VAL_ENCRYPTED 0x00000004
#define OMAP_VAL_NOHEADER 0x00000008
#define OMAP_VAL_CRYPTO_GENERATION 0x00000010

OMAP_VAL_DELETED

The object has been deleted, and this mapping is a placeholder.

#define OMAP_VAL_DELETED 0x00000001

OMAP_VAL_SAVED

This object mapping shouldnʼt be replaced when the object is updated.

#define OMAP_VAL_SAVED 0x00000002

This flag is used only on mappings in an object map thatʼs manually managed. In the current Apple implementation,
itʼs never used.

See also the OMAP_MANUALLY_MANAGED flag.

OMAP_VAL_ENCRYPTED

The object is encrypted.

#define OMAP_VAL_ENCRYPTED 0x00000004

OMAP_VAL_NOHEADER

The object is stored without an obj_phys_t header.

#define OMAP_VAL_NOHEADER 0x00000008

OMAP_VAL_CRYPTO_GENERATION

A one-bit flag that tracks encryption configuration.

#define OMAP_VAL_CRYPTO_GENERATION 0x00000010

During the transition from an old encryption configuration to a new one, not all objects have been reencrypted using
the new configuration. When the encryption configuration is changed, the object mapʼs flag is toggled. After an object
is reencrypted, the objectʼs flag is also toggled.

If this flag doesnʼt match the flag on the object map, the encryption configuration has changed, but the object hasnʼt
been reencrypted yet. Use the previous encryption configuration to decrypt the object.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

45
Object Map
Snapshot Flags

See also OMAP_CRYPTO_GENERATION, which is used by the omap_phys_t field of om_flags.

Snapshot Flags
The flags used to describe the state of a snapshot.

#define OMAP_SNAPSHOT_DELETED 0x00000001

#define OMAP_SNAPSHOT_REVERTED 0x00000002

OMAP_SNAPSHOT_DELETED

The snapshot has been deleted.

#define OMAP_SNAPSHOT_DELETED 0x00000001

OMAP_SNAPSHOT_REVERTED

The snapshot has been deleted as part of a revert.

#define OMAP_SNAPSHOT_REVERTED 0x00000002

Object Map Flags

The flags used by object maps.

#define OMAP_MANUALLY_MANAGED 0x00000001

#define OMAP_ENCRYPTING 0x00000002
#define OMAP_DECRYPTING 0x00000004
#define OMAP_KEYROLLING 0x00000008
#define OMAP_CRYPTO_GENERATION 0x00000010

OMAP_MANUALLY_MANAGED

The object map doesnʼt support snapshots.

#define OMAP_MANUALLY_MANAGED 0x00000001

This flag must be set on the containerʼs object map and is invalid on a volumeʼs object map.

OMAP_ENCRYPTING

A transition is in progress from unencrypted storage to encrypted storage.

#define OMAP_ENCRYPTING 0x00000002

OMAP_DECRYPTING

A transition is in progress from encrypted storage to unencrypted storage.

#define OMAP_DECRYPTING 0x00000004

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

46
Object Map
Object Map Constants

OMAP_KEYROLLING

A transition is in progress from encrypted storage using an old key to encrypted storage using a new key.

#define OMAP_KEYROLLING 0x00000008

OMAP_CRYPTO_GENERATION

A one-bit flag that tracks encryption configuration.

#define OMAP_CRYPTO_GENERATION 0x00000010

For information about how this flag is used to track the old and new encryption configuration, see OMAP_VAL_
CRYPTO_GENERATION, which is used by the ov_flags field of omap_val_t.

Object Map Constants

Constants that specify size constraints of an object map.

#define OMAP_MAX_SNAP_COUNT UINT32_MAX

OMAP_MAX_SNAP_COUNT

The maximum number of snapshots that can be stored in an object map.

#define OMAP_MAX_SNAP_COUNT UINT32_MAX

Object Map Reaper Phases

Phases used by the reaper when deleting objects that are stored in an object map.

#define OMAP_REAP_PHASE_MAP_TREE 1
#define OMAP_REAP_PHASE_SNAPSHOT_TREE 2

OMAP_REAP_PHASE_MAP_TREE

The reaper is deleting entries from the object mapping tree.

#define OMAP_REAP_PHASE_MAP_TREE 1

OMAP_REAP_PHASE_SNAPSHOT_TREE

The reaper is deleting entries from the snapshot tree.

#define OMAP_REAP_PHASE_SNAPSHOT_TREE 2

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

47
Volume
A volume contains a file system, the files and metadata that make up that file system, and various supporting data
structures like an object map.

apfs_superblock_t
A volume superblock.

struct apfs_superblock {
obj_phys_t apfs_o;

uint32_t apfs_magic;
uint32_t apfs_fs_index;

uint64_t apfs_features;
uint64_t apfs_readonly_compatible_features;
uint64_t apfs_incompatible_features;

uint64_t apfs_unmount_time;

uint64_t apfs_fs_reserve_block_count;
uint64_t apfs_fs_quota_block_count;
uint64_t apfs_fs_alloc_count;

wrapped_meta_crypto_state_t apfs_meta_crypto;

uint32_t apfs_root_tree_type;
uint32_t apfs_extentref_tree_type;
uint32_t apfs_snap_meta_tree_type;

oid_t apfs_omap_oid;
oid_t apfs_root_tree_oid;
oid_t apfs_extentref_tree_oid;
oid_t apfs_snap_meta_tree_oid;

xid_t apfs_revert_to_xid;
oid_t apfs_revert_to_sblock_oid;

uint64_t apfs_next_obj_id;

uint64_t apfs_num_files;
uint64_t apfs_num_directories;
uint64_t apfs_num_symlinks;
uint64_t apfs_num_other_fsobjects;
uint64_t apfs_num_snapshots;

48
Volume
apfs_superblock_t

uint64_t apfs_total_blocks_alloced;
uint64_t apfs_total_blocks_freed;

uuid_t apfs_vol_uuid;
uint64_t apfs_last_mod_time;

uint64_t apfs_fs_flags;

apfs_modified_by_t apfs_formatted_by;
apfs_modified_by_t apfs_modified_by[APFS_MAX_HIST];

uint8_t apfs_volname[APFS_VOLNAME_LEN];
uint32_t apfs_next_doc_id;

uint16_t apfs_role;
uint16_t reserved;

xid_t apfs_root_to_xid;
oid_t apfs_er_state_oid;
};

#define APFS_MAGIC 'BSPA'

#define APFS_MAX_HIST 8
#define APFS_VOLNAME_LEN 256

apfs_o

The objectʼs header.

obj_phys_t apfs_o;

apfs_magic

A number that can be used to verify that youʼre reading an instance of apfs_superblock_t.

uint32_t apfs_magic;

The value of this field is always APFS_MAGIC.

apfs_fs_index

The index of this volumeʼs object identifier in the containerʼs array of volumes.

uint32_t apfs_fs_index

The containerʼs array is stored in the nx_fs_oid field of nx_superblock_t.

When a volume is being deleted, itʼs removed from the containerʼs array of volumes before apfs_superblock_t
object is destroyed. If you read this field of a volume thatʼs being deleted, the specified entry in the array might have
already been reused for another volume.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

49
Volume
apfs_superblock_t

apfs_features

A bit field of the optional features being used by this volume.

uint64_t apfs_features

For the values used in this bit field, see Optional Volume Feature Flags.

If your implementation doesnʼt support an optional feature thatʼs in use, ignore that feature in this list and mount the
volume as usual.

apfs_readonly_compatible_features

A bit field of the read-only compatible features being used by this volume.

uint64_t apfs_readonly_compatible_features

For the values used in this bit field, see Read-Only Compatible Volume Feature Flags.

If your implementation doesnʼt support a read-only compatible feature thatʼs in use, mount the volume as read-only.

apfs_incompatible_features

A bit field of the backward-incompatible features being used by this volume.

uint64_t apfs_incompatible_features

For the values used in this bit field, see Incompatible Volume Feature Flags.

If your implementation doesnʼt support a backward-incompatible feature thatʼs in use, it must not mount the volume.

apfs_unmount_time

The time that this volume was last unmounted.

uint64_t apfs_unmount_time

This timestamp is represented as the number of nanoseconds since January 1, 1970 at 0 00 UTC, disregarding leap
seconds.

apfs_fs_reserve_block_count

The number of blocks that have been reserved for this volume to allocate.

uint64_t apfs_fs_reserve_block_count

apfs_fs_quota_block_count

The maximum number of blocks that this volume can allocate.

uint64_t apfs_fs_quota_block_count

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

50
Volume
apfs_superblock_t

apfs_fs_alloc_count

The number of blocks currently allocated for this volumeʼs file system.

uint64_t apfs_fs_alloc_count

apfs_meta_crypto

Effaceable cryptographic metadata for this volume.

wrapped_meta_crypto_state_t apfs_meta_crypto

apfs_root_tree_type

The type of the root file-system tree.

uint32_t apfs_root_tree_type

The value is typically OBJ_VIRTUAL | OBJECT_TYPE_BTREE, with a subtype of OBJECT_TYPE_FSTREE. For pos-
sible values, see Object Types.

apfs_extentref_tree_type

The type of the extent-reference tree.

uint32_t apfs_extentref_tree_type

The value is typically OBJ_PHYSICAL | OBJECT_TYPE_BTREE, with a subtype of OBJECT_TYPE_BLOCKREF. For

possible values, see Object Types.

apfs_snap_meta_tree_type

The type of the snapshot metadata tree.

uint32_t apfs_snap_meta_tree_type

The value is typically OBJ_PHYSICAL | OBJECT_TYPE_BTREE, with a subtype of OBJECT_TYPE_BLOCKREF. For

possible values, see Object Types.

apfs_omap_oid

The physical object identifier of the volumeʼs object map.

oid_t apfs_omap_oid

apfs_root_tree_oid

The virtual object identifier of the root file-system tree.

oid_t apfs_root_tree_oid

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

51
Volume
apfs_superblock_t

apfs_extentref_tree_oid

The physical object identifier of the extent-reference tree.

oid_t apfs_extentref_tree_oid

When a snapshot is created, the current extent-reference tree is moved to the snapshot. A new, empty, extent-
reference tree is created and its object identifier becomes the new value of this field.

apfs_snap_meta_tree_oid

The virtual object identifier of the snapshot metadata tree.

oid_t apfs_snap_meta_tree_oid

apfs_revert_to_xid

The transaction identifier of a snapshot that the volume will revert to.

xid_t apfs_revert_to_xid

When mounting a volume, if the value of this field isnʼt 0, revert to the specified snapshot by deleting all snapshots
after the specified transaction identifier and deleting the current state, and then setting this field to 0.

apfs_revert_to_sblock_oid

The physical object identifier of a volume superblock that the volume will revert to.

oid_t apfs_revert_to_sblock_oid

When mounting a volume, if the apfs_revert_to_xid field isnʼt 0, ignore the value of this field. Otherwise, revert
to the specified volume superblock.

apfs_next_obj_id

The next identifier that will be assigned to a file-system object in this volume.

uint64_t apfs_next_obj_id

apfs_num_files

The number of regular files in this volume.

uint64_t apfs_num_files

apfs_num_directories

The number of directories in this volume.

uint64_t apfs_num_directories

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

52
Volume
apfs_superblock_t

apfs_num_symlinks

The number of symbolic links in this volume.

uint64_t apfs_num_symlinks

apfs_num_other_fsobjects

The number of other files in this volume.

uint64_t apfs_num_other_fsobjects

The value of this field includes all files that arenʼt included in the apfs_num_symlinks, apfs_num_directories,
or apfs_num_files fields.

apfs_num_snapshots

The number of snapshots in this volume.

uint64_t apfs_num_snapshots

apfs_total_blocks_alloced

The total number of blocks that have been allocated by this volume.

uint64_t apfs_total_blocks_alloced

The value of this field increases when blocks are allocated, but isnʼt modified when theyʼre freed. If the volume doesnʼt
contain any files, the value of this field matches apfs_total_blocks_freed.

apfs_total_blocks_freed

The total number of blocks that have been freed by this volume.

uint64_t apfs_total_blocks_freed

The value of this field isnʼt modified when blocks are allocated, but increases when theyʼre freed. If the volume doesnʼt
contain any files, the value of this field matches apfs_total_blocks_alloced.

apfs_vol_uuid

The universally unique identifier for this volume.

uuid_t apfs_vol_uuid

apfs_last_mod_time

The time that this volume was last modified.

uint64_t apfs_last_mod_time

This timestamp is represented as the number of nanoseconds since January 1, 1970 at 0 00 UTC, disregarding leap
seconds.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

53
Volume
apfs_superblock_t

apfs_fs_flags

The volumeʼs flags.

uint64_t apfs_fs_flags

For the values used in this bit field, see Volume Flags.

apfs_formatted_by

Information about the software that created this volume.

apfs_modified_by_t apfs_formatted_by

This field is set only once, when the volume is created.

apfs_modified_by

Information about the software that has modified this volume.

apfs_modified_by_t apfs_modified_by[APFS_MAX_HIST]

The newest element in this array is stored at index 0. To update this field when you modify a volume, move each
element to the index thatʼs larger by one, and then write the new modification information. When you create a new
volume, fill the arrayʼs memory with zeros.

If the implementationʼs information is already present in this field, you can update the field as usual (creating a dupli-
cate), or leave the fieldʼs value unmodified. Both behaviors are permitted.

apfs_volname

The name of the volume, represented as a null-terminated UTF-8 string.

uint8_t apfs_volname[APFS_VOLNAME_LEN]

The APFS_INCOMPAT_NON_UTF8_FNAMES flag has no effect on this fieldʼs value.

apfs_next_doc_id

The next document identifier that will be assigned.

uint32_t apfs_next_doc_id

A documentʼs identifier is stored in the INO_EXT_TYPE_DOCUMENT_ID extended field of the inode.

After assigning a new document identifier, this field is incremented by one. Valid document identifiers are greater than
MIN_DOC_ID and less than UINT32_MAX - 1. If a new document identifier isnʼt available, thatʼs an unrecoverable
error. Identifiers arenʼt allowed to restart from one or to be reused.

apfs_role

The role of this volume within the container.

uint16_t apfs_role

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

54
Volume
apfs_modified_by_t

reserved

Reserved.

uint16_t reserved

Populate this field with 0 when you create a new volume, and preserve its value when you modify an existing volume.

apfs_root_to_xid

The transaction identifier of the snapshot to root from, or 0 to root normally.

xid_t apfs_root_to_xid

apfs_er_state_oid

The current state of encryption or decryption for a drive thatʼs being encrypted or decrypted, or 0 if no encryption
change is in progress.

oid_t apfs_er_state_oid

APFS_MAGIC

The value of the apfs_magic field.

#define APFS_MAGIC 'BSPA'

This magic number was chosen because in hex dumps it appears as “APSB”, which is an abbreviated form of APFS
superblock.

APFS_MAX_HIST

The number of entries stored in the apfs_modified_by field.

#define APFS_MAX_HIST 8

APFS_VOLNAME_LEN

The maximum length of the volume name stored in the apfs_volname field.

#define APFS_VOLNAME_LEN 256

apfs_modified_by_t
Information about a program that modified the volume.

struct apfs_modified_by {
uint8_t id[APFS_MODIFIED_NAMELEN];
uint64_t timestamp;
xid_t last_xid;
}
typedef struct apfs_modified_by apfs_modified_by_t;

#define APFS_MODIFIED_NAMELEN 32

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

55
Volume
Volume Flags

This structure is used by the apfs_modified_by and apfs_formatted_by fields of apfs_superblock_t.

id

A string that identifies the program and its version.

uint8_t id[APFS_MODIFIED_NAMELEN];

timestamp

The time that the program last modified this volume.

uint64_t timestamp;

This timestamp is represented as the number of nanoseconds since January 1, 1970 at 0 00 UTC, disregarding leap
seconds.

last_xid

The last transaction identifier thatʼs part of this programʼs modifications.

xid_t last_xid;

Volume Flags
The flags used to indicate volume status.

#define APFS_FS_UNENCRYPTED 0x00000001LL

#define APFS_FS_EFFACEABLE 0x00000002LL
#define APFS_FS_RESERVED_4 0x00000004LL
#define APFS_FS_ONEKEY 0x00000008LL
#define APFS_FS_SPILLEDOVER 0x00000010LL
#define APFS_FS_RUN_SPILLOVER_CLEANER 0x00000020LL

#define APFS_FS_FLAGS_VALID_MASK (APFS_FS_UNENCRYPTED \

| APFS_FS_EFFACEABLE \
| APFS_RESERVED_4 \
| APFS_FS_ONEKEY \
| APFS_FS_SPILLEDOVER \
| APFS_FS_RUN_SPILLOVER_CLEANER)

#define APFS_FS_CRYPTOFLAGS (APFS_FS_UNENCRYPTED \

| APFS_FS_EFFACEABLE \
| APFS_FS_ONEKEY)

APFS_FS_UNENCRYPTED

The volume isnʼt encrypted.

#define APFS_FS_UNENCRYPTED 0x00000001LL

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

56
Volume
Volume Flags

APFS_FS_EFFACEABLE

The volume supports effaceable storage.

#define APFS_FS_EFFACEABLE 0x00000002LL

APFS_RESERVED_4

Reserved.

#define APFS_RESERVED_4 0x00000004LL

Donʼt set this flag, but preserve it if itʼs already set.

APFS_FS_ONEKEY

The volume uses software encryption with a single key.

#define APFS_FS_ONEKEY 0x00000008LL

APFS_FS_SPILLEDOVER

The volume has run out of allocated space on the solid-state drive.

#define APFS_FS_SPILLEDOVER 0x00000010LL

See also INODE_ALLOCATION_SPILLEDOVER.

APFS_FS_RUN_SPILLOVER_CLEANER

The volume has spilled over and the spillover cleaner must be run.

#define APFS_FS_RUN_SPILLOVER_CLEANER 0x00000020LL

APFS_FS_FLAGS_VALID_MASK

A bit mask of all volume flags.

#define APFS_FS_FLAGS_VALID_MASK (APFS_FS_UNENCRYPTED \

| APFS_FS_EFFACEABLE \
| APFS_FS_RESERVED_4 \
| APFS_FS_ONEKEY \
| APFS_FS_SPILLEDOVER \
| APFS_FS_RUN_SPILLOVER_CLEANER)

APFS_FS_CRYPTOFLAGS

A bit mask of all encryption-related volume flags.

#define APFS_FS_CRYPTOFLAGS (APFS_FS_UNENCRYPTED \

| APFS_FS_EFFACEABLE \
| APFS_FS_ONEKEY)

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

57
Volume
Optional Volume Feature Flags

Optional Volume Feature Flags

The flags used to describe optional features of an Apple File System volume.

#define APFS_FEATURE_DEFRAG_PRERELEASE 0x00000001LL

#define APFS_FEATURE_HARDLINK_MAP_RECORDS 0x00000002LL
#define APFS_FEATURE_DEFRAG 0x00000004LL

#define APFS_SUPPORTED_FEATURES_MASK (APFS_FEATURE_DEFRAG \

| APFS_FEATURE_DEFRAG_PRERELEASE \
| APFS_FEATURE_HARDLINK_MAP_RECORDS)

These flags are used by the apfs_features field of apfs_superblock_t.

APFS_FEATURE_DEFRAG_PRERELEASE

Reserved.

#define APFS_FEATURE_DEFRAG_PRERELEASE 0x00000001LL

Warning

To avoid data corruption, this flag must not be set.

This flag enabled a prerelease version of the defragmentation system in macOS 10.13 versions. Itʼs ignored by
macOS 10.13.6 and later.

APFS_FEATURE_HARDLINK_MAP_RECORDS

The volume has hardlink map records.

#define APFS_FEATURE_HARDLINK_MAP_RECORDS 0x00000002LL

For details about hardlink map records, see Siblings.

APFS_FEATURE_DEFRAG

The volume supports defragmentation.

#define APFS_FEATURE_DEFRAG 0x00000004LL

This flag is ignored by versions before macOS 10.14.

APFS_SUPPORTED_FEATURES_MASK

A bit mask of all the optional volume features.

#define APFS_SUPPORTED_FEATURES_MASK (APFS_FEATURE_DEFRAG \

| APFS_FEATURE_DEFRAG_PRERELEASE \
| APFS_FEATURE_HARDLINK_MAP_RECORDS)

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

58
Volume
Read-Only Compatible Volume Feature Flags

Read-Only Compatible Volume Feature Flags

The flags used to describe read-only compatible features of an Apple File System volume.

#define APFS_SUPPORTED_ROCOMPAT_MASK (0x0ULL)

These flags are used by the apfs_readonly_compatible_features field of apfs_superblock_t. There are
currently none defined.

APFS_SUPPORTED_ROCOMPAT_MASK

A bit mask of all read-only compatible volume features.

#define APFS_SUPPORTED_ROCOMPAT_MASK (0x0ULL)

Incompatible Volume Feature Flags

The flags used to describe backward-incompatible features of an Apple File System volume.

#define APFS_INCOMPAT_CASE_INSENSITIVE 0x00000001LL

#define APFS_INCOMPAT_DATALESS_SNAPS 0x00000002LL
#define APFS_INCOMPAT_ENC_ROLLED 0x00000004LL
#define APFS_INCOMPAT_NORMALIZATION_INSENSITIVE 0x00000008LL

#define APFS_SUPPORTED_INCOMPAT_MASK (APFS_INCOMPAT_CASE_INSENSITIVE \

| APFS_INCOMPAT_DATALESS_SNAPS \
| APFS_INCOMPAT_ENC_ROLLED \
| APFS_INCOMPAT_NORMALIZATION_INSENSITIVE)

These flags are used by the apfs_incompatible_features field of apfs_superblock_t.

APFS_INCOMPAT_CASE_INSENSITIVE

Filenames on this volume are case insensitive.

#define APFS_INCOMPAT_CASE_INSENSITIVE 0x00000001LL

APFS_INCOMPAT_DATALESS_SNAPS

At least one snapshot with no data exists for this volume.

#define APFS_INCOMPAT_DATALESS_SNAPS 0x00000002LL

APFS_INCOMPAT_ENC_ROLLED

This volumeʼs encryption has changed keys at least once.

#define APFS_INCOMPAT_ENC_ROLLED 0x00000004LL

APFS_INCOMPAT_NORMALIZATION_INSENSITIVE

Filenames on this volume are normalization insensitive.

#define APFS_INCOMPAT_NORMALIZATION_INSENSITIVE 0x00000008LL

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

59
Volume
Incompatible Volume Feature Flags

Normalization insensitivity is part of hashing filenames, as described in the name_len_and_hash field of j_drec_
hashed_key_t.

APFS_SUPPORTED_INCOMPAT_MASK

A bit mask of all the backward-incompatible volume features.

#define APFS_SUPPORTED_INCOMPAT_MASK (APFS_INCOMPAT_CASE_INSENSITIVE \

| APFS_INCOMPAT_DATALESS_SNAPS \
| APFS_INCOMPAT_ENC_ROLLED \
| APFS_INCOMPAT_NORMALIZATION_INSENSITIVE)

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

60
File-System Objects
A file-system object stores information about a part of the file system, such as a directory or a file on disk. These
objects are stored as one or more records. For example, the file-system object for a directory that contains
two files is stored as three records: a record of type APFS_TYPE_INODE for the inode, and two records of type
APFS_TYPE_DIR_REC for the directory entries. This record-based method of storing file-system objects helps make
efficient use of disk space.

File-system records are stored as key/value pairs in a B-tree. The key contains information, such as the object identifier
and the record type, used to look up a record. Keys begin with an instance of j_key_t, and many records use
j_key_t as their entire key.

For sorting file-system records—for example, to keep them ordered in a B-tree—the following comparison rules are
used:

1. Compare the object identifiers numerically:

j_key_t.obj_id_and_type & OBJ_ID_MASK

2. Compare the object types numerically:

(j_key_t.obj_id_and_type & OBJ_TYPE_MASK) >> OBJ_TYPE_SHIFT

3. For extended attribute records and directory entry records, compare the names lexicographically:

j_drec_key_t.name

Because of all the records for a file-system object have the same object identifier, all of the records that make up a
single object are sorted next to each other.

The relationship between file-system objects and the records theyʼre made up from is as follows:

Files

• APFS_TYPE_INODE Required
• APFS_TYPE_CRYPTO_STATE
• APFS_TYPE_DSTREAM_ID
• APFS_TYPE_EXTENT
• APFS_TYPE_FILE_EXTENT
• APFS_TYPE_SIBLING_LINK
• APFS_TYPE_XATTR

Directories

• APFS_TYPE_INODE Required
• APFS_TYPE_CRYPTO_STATE
• APFS_TYPE_DIR_REC
• APFS_TYPE_DIR_STATS
• APFS_TYPE_XATTR

Symbolic Links

• APFS_TYPE_INODE Required
• APFS_TYPE_XATTR Required

61
File-System Objects
j_key_t

• APFS_TYPE_CRYPTO_STATE
• APFS_TYPE_DSTREAM_ID
• APFS_TYPE_EXTENT
• APFS_TYPE_FILE_EXTENT

There must be an extended attribute whose name is SYMLINK_EA_NAME and whose value is the path to the target
file.

Snapshots

• APFS_TYPE_SNAP_METADATA Required
• APFS_TYPE_SNAP_NAME Required
• APFS_TYPE_CRYPTO_STATE
• APFS_TYPE_EXTENT

Sibling Maps

• APFS_TYPE_SIBLING_MAP Required

Tip

To simplify manipulating file-system objects, define custom types that combine the key and value of a record,
and custom types that combine the objectʼs records.

j_key_t
A header used at the beginning of all file-system keys.

typedef struct j_key {

uint64_t obj_id_and_type;
} __attribute__((packed)) j_key_t;

#define OBJ_ID_MASK 0x0fffffffffffffffULL

#define OBJ_TYPE_MASK 0xf000000000000000ULL
#define OBJ_TYPE_SHIFT 60

All file-system objects have a key that begins with this information. The key for some object types have additional
fields that follow this header, and other object types use j_key_t as their entire key.

The following record types use this structure as their key without adding any additional fields:

obj_id_and_type

A bit field that contains the objectʼs identifier and its type.

uint64_t obj_id_and_type;

The objectʼs identifier is a uint64_t value accessed as obj_id_and_type & OBJ_ID_MASK. The objectʼs type
is a uint8_t value accessed as (obj_id_and_type & OBJ_TYPE_MASK) >> OBJ_TYPE_SHIFT. The objectʼs
type is one of the constants defined by j_obj_types.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

62
File-System Objects
j_inode_key_t

OBJ_ID_MASK

The bit mask used to access the object identifier.

#define OBJ_ID_MASK 0x0fffffffffffffffULL

OBJ_TYPE_MASK

The bit mask used to access the object type.

#define OBJ_TYPE_MASK 0xf000000000000000ULL

OBJ_TYPE_SHIFT

The bit shift used to access the object type.

#define OBJ_TYPE_SHIFT 60

j_inode_key_t
The key half of a directory-information record.

typedef struct j_inode_key_t {

j_key_t hdr;
} __attribute__((packed)) j_dir_stats_key_t;

hdr

The recordʼs header.

j_key_t hdr;

The object identifier in the header is the file-system objectʼs identifier, also known as its inode number. The type in
the header is always APFS_TYPE_INODE.

j_inode_val_t
The value half of an inode record.

typedef struct j_inode_val {

uint64_t parent_id;
uint64_t private_id;

uint64_t create_time;
uint64_t mod_time;
uint64_t change_time;
uint64_t access_time;

uint64_t internal_flags;

union {
int32_t nchildren;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

63
File-System Objects
j_inode_val_t

int32_t nlink;
};

cp_key_class_t default_protection_class;
uint32_t write_generation_counter;
uint32_t bsd_flags;
uid_t owner;
gid_t group;
mode_t mode;
uint16_t pad1;
uint64_t pad2;
uint8_t xfields[];
} __attribute__((packed)) j_inode_val_t;

typedef uint32_t uid_t;

typedef uint32_t gid_t;

parent_id

The identifier of the file system record for the parent directory.

uint64_t parent_id;

private_id

The unique identifier used by this fileʼs data stream.

uint64_t private_id;

This identifier appears in the owning_obj_id field of j_phys_ext_val_t records that describe the extents where
the data is stored.

For an inode that doesnʼt have data, the value of this field is the file-system objectʼs identifier.

create_time

The time that this record was created.

uint64_t create_time;

This timestamp is represented as the number of nanoseconds since January 1, 1970 at 0 00 UTC, disregarding leap
seconds.

mod_time

The time that this record was last modified.

uint64_t mod_time;

This timestamp is represented as the number of nanoseconds since January 1, 1970 at 0 00 UTC, disregarding leap
seconds.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

64
File-System Objects
j_inode_val_t

change_time

The time that this recordʼs attributes were last modified.

uint64_t change_time;

This timestamp is represented as the number of nanoseconds since January 1, 1970 at 0 00 UTC, disregarding leap
seconds.

access_time

The time that this record was last accessed.

uint64_t access_time;

This timestamp is represented as the number of nanoseconds since January 1, 1970 at 0 00 UTC, disregarding leap
seconds.

internal_flags

The inodeʼs flags.

uint64_t internal_flags;

For the values used in this bit field, see j_inode_flags.

nchildren

The number of directory entries.

int32_t nchildren;

This union field is valid only if the inode is a directory.

nlink

The number of hard links whose target is this inode.

int32_t nlink;

This union field is valid only if the inode isnʼt a directory.

Inodes with multiple hard links—as indicated by a value greater than one in this field—have additional invariants:

• The parent_id field refers to the parent directory of the primary link.

• The name field contains the name of the primary link.

• The INO_EXT_TYPE_NAME extended field contains the name of this link.

• The file-system object includes sibling-link records, as discussed in Siblings.

default_protection_class

The default protection class for this inode.

cp_key_class_t default_protection_class;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

65
File-System Objects
j_inode_val_t

write_generation_counter

A monotonically increasing counter thatʼs incremented each time this inode or its data is modified.

uint32_t write_generation_counter;

This value is allowed to overflow and restart from zero.

bsd_flags

The inodeʼs BSD flags.

uint32_t bsd_flags;

For information about these flags, see the chflags(2) man page and the <sys/stat.h> header file.

owner

The user identifier of the inodeʼs owner.

uid_t owner;

group

The group identifier of the inodeʼs group.

gid_t group;

mode

The fileʼs mode.

mode_t mode;

For possible values, see File Modes.

pad1

Reserved.

uint16_t pad1;

Populate this field with 0 when you create a new inode, and preserve its value when you modify an existing inode.

This field is padding.

pad2

Reserved.

uint64_t pad2;

Populate this field with 0 when you create a new inode, and preserve its value when you modify an existing inode.

This field is padding.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

66
File-System Objects
j_drec_key_t

xfields

The inodeʼs extended fields.

uint8_t xfields[];

This location on disk contains several pieces of data that have variable sizes. For information about reading extended
fields, see Extended Fields.

uid_t

A user identifier.

typedef uint32_t uid_t;

gid_t

A group identifier.

typedef uint32_t gid_t;

j_drec_key_t
The key half of a directory entry record.

typedef struct j_drec_key {

j_key_t hdr;
uint16_t name_len;
uint8_t name[0];
} __attribute__((packed)) j_drec_key_t;

hdr

The recordʼs header.

j_key_t hdr;

The object identifier in the header is the file-system objectʼs identifier. The type in the header is always APFS_TYPE_
DIR_REC.

name_len_and_hash

The length of the name, including the final null character (U+0000).

uint32_t name_len_and_hash;

name

The name, represented as a null-terminated UTF-8 string.

uint8_t name[0];

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

67
File-System Objects
j_drec_hashed_key_t

j_drec_hashed_key_t
The key half of a directory entry record, including a precomputed hash of its name.

typedef struct j_drec_hashed_key {

j_key_t hdr;
uint32_t name_len_and_hash;
uint8_t name[0];
} __attribute__((packed)) j_drec_hashed_key_t;

#define J_DREC_LEN_MASK 0x000003ff

#define J_DREC_HASH_MASK 0xfffff400
#define J_DREC_HASH_SHIFT 10

hdr

The recordʼs header.

j_key_t hdr;

name_len_and_hash

The hash and length of the name.

uint32_t name_len_and_hash;

The length is a 10-bit unsigned integer, accessed as name_len_and_hash & J_DREC_LEN_MASK. The length in-
cludes the final null character (U+0000).

The hash is an unsigned 22-bit integer, accessed as (name_len_and_hash & J_DREC_HASH_MASK) >>
J_DREC_HASH_SHIFT. The hash is computed as follows:

1. Start with the filename, represented as a null-terminated UTF-8 string.

2. Normalize the string using canonical decomposition (NFD).
3. Represent the normalized filename as a null-terminated UTF-32 string.
4. Compute the CRC-32c hash of the UTF-32 string.
5. Complement the bits of the hash.
6. Keep only the low 22 bits of the hash.

name

The name, represented as a null-terminated UTF-8 string.

uint8_t name[0];

J_DREC_LEN_MASK

The bit mask used to access the length of the name.

#define J_DREC_LEN_MASK 0x000003ff

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

68
File-System Objects
j_drec_val_t

J_DREC_HASH_MASK

The bit mask used to access the hash of the name.

#define J_DREC_HASH_MASK 0xfffff400

J_DREC_HASH_SHIFT

The bit shift used to access the hash of the name.

#define J_DREC_HASH_SHIFT 10

j_drec_val_t
The value half of a directory entry record.

typedef struct j_drec_val {

uint64_t file_id;
uint64_t date_added;
uint16_t flags;
uint8_t xfields[];
} __attribute__((packed)) j_drec_val_t;

file_id

The identifier of the inode that this directory entry represents.

uint64_t file_id;

date_added

The time that this directory entry was added to the directory.

uint64_t date_added;

This timestamp is represented as the number of nanoseconds since January 1, 1970 at 0 00 UTC, disregarding leap
seconds. Itʼs not updated when modifying the directory entry—for example, by renaming a file without moving it to a
different directory.

flags

The directory entryʼs flags.

uint16_t flags;

The bits that are set in DREC_TYPE_MASK store the inodeʼs file type, and the remaining bits are reserved. Populate
the reserved bits with 0 when you create a new directory entry, and preserve their value when you modify an existing
directory entry.

For possible values, see Directory Entry File Types.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

69
File-System Objects
j_dir_stats_key_t

xfields

The directory entryʼs extended fields.

uint8_t xfields[];

This location on disk contains several pieces of data that have variable sizes. For information about reading extended
fields, see Extended Fields.

j_dir_stats_key_t
The key half of a directory-information record.

typedef struct j_dir_stats_key {

j_key_t hdr;
} __attribute__((packed)) j_dir_stats_key_t;

hdr

The recordʼs header.

j_key_t hdr;

The object identifier in the header is the file-system objectʼs identifier. The type in the header is always
APFS_TYPE_DIR_REC.

j_dir_stats_val_t
The value half of a directory-information record.

typedef struct j_dir_stats_val {

uint64_t num_children;
uint64_t total_size;
uint64_t chained_key;
uint64_t gen_count;
} __attribute__((packed)) j_dir_stats_val_t;

num_children

The number of files and folders contained by the directory.

uint64_t num_children;

total_size

The total size, in bytes, of all the files stored in this directory and all of this directoryʼs descendants.

uint64_t total_size;

Hard links contribute to the total_size of every directory they appear in.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

70
File-System Objects
j_xattr_key_t

chained_key

The parent directoryʼs file system object identifier.

uint64_t chained_key;

gen_count

A monotonically increasing counter thatʼs incremented each time this inode or any of its children is modified.

uint64_t gen_count;

Modifying the contents of a file requires updating the inodeʼs modification time and write generation, which means
this counter must be incremented for the directory that contains the file.

If this counter canʼt be incremented without overflow, thatʼs an unrecoverable error.

j_xattr_key_t
The key half of an extended attribute record.

typedef struct j_xattr_key {

j_key_t hdr;
uint16_t name_len;
uint8_t name[0];
} __attribute__((packed)) j_xattr_key_t;

hdr

The recordʼs header.

j_key_t hdr;

The object identifier in the header is the file-system objectʼs identifier. The type in the header is always
APFS_TYPE_XATTR.

name_len

The length of the extended attributeʼs name, including the final null character (U+0000).

uint16_t name_len;

name

The extended attributeʼs name, represented as a null-terminated UTF-8 string.

uint8_t name[0];

j_xattr_val_t
The value half of an extended attribute record.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

71
File-System Objects
j_xattr_val_t

typedef struct j_xattr_val {

uint16_t flags;
uint16_t xdata_len;
uint8_t xdata[0];
} __attribute__((packed)) j_xattr_val_t;

flags

The extended attribute recordʼs flags.

uint16_t flags;

For the values used in this bit field, see j_xattr_flags. Either the XATTR_DATA_EMBEDDED or XATTR_DATA_
STREAM flag must be set.

xdata_len

The length of the extended attribute data.

uint16_t xdata_len;

If the XATTR_DATA_EMBEDDED flag is set, this field is the length of the data in the xdata field. Otherwise, this field
is ignored.

xdata

The extended attribute data or the identifier of a data stream that contains the data.

uint8_t xdata[0];

If the XATTR_DATA_EMBEDDED flag is set, the extended attribute data is stored directly in this field. Otherwise, this
field contains the identifier (uint64_t) for a data stream record that stores the extended attribute data. See also
j_xattr_dstream_t.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

72
File-System Constants
File-system objects use several groups of constants to define values for record types, reserved inode numbers, and
flags and bit masks used in bit fields.

j_obj_types
The type of a file-system record.

typedef enum {
APFS_TYPE_ANY = 0,

APFS_TYPE_SNAP_METADATA = 1,
APFS_TYPE_EXTENT = 2,
APFS_TYPE_INODE = 3,
APFS_TYPE_XATTR = 4,
APFS_TYPE_SIBLING_LINK = 5,
APFS_TYPE_DSTREAM_ID = 6,
APFS_TYPE_CRYPTO_STATE = 7,
APFS_TYPE_FILE_EXTENT = 8,
APFS_TYPE_DIR_REC = 9,
APFS_TYPE_DIR_STATS = 10,
APFS_TYPE_SNAP_NAME = 11,
APFS_TYPE_SIBLING_MAP = 12,

APFS_TYPE_MAX_VALID = 12,
APFS_TYPE_MAX = 15,

APFS_TYPE_INVALID = 15,
} j_obj_types;

This value is stored in the type bits of a j_key_t structureʼs obj_id_and_type field.

APFS_TYPE_ANY

A record of any type.

APFS_TYPE_ANY = 0

This enumeration case is used only in search queries and in tests when iterating over objects. Itʼs not valid as the type
of a file-system object.

APFS_TYPE_SNAP_METADATA

Metadata about a snapshot.

APFS_TYPE_SNAP_METADATA = 1

The key is an instance of j_snap_metadata_key_t and the value is an instance of j_snap_metadata_val_t.

73
File-System Constants
j_obj_types

APFS_TYPE_EXTENT

A physical extent record.

APFS_TYPE_EXTENT = 2

The key is an instance of j_phys_ext_key_t and the value is an instance of j_phys_ext_val_t.

APFS_TYPE_INODE

An inode.

APFS_TYPE_INODE = 3

The key is an instance of j_inode_key_t and the value is an instance of j_inode_val_t.

APFS_TYPE_XATTR

An extended attribute.

APFS_TYPE_XATTR = 4

The key is an instance of j_xattr_key_t and the value is an instance of j_xattr_val_t.

APFS_TYPE_SIBLING_LINK

A mapping from an inode to hard links that the inode is the target of.

APFS_TYPE_SIBLING_LINK = 5

The key is an instance of j_sibling_key_t and the value is an instance of j_sibling_val_t.

APFS_TYPE_DSTREAM_ID

A data stream.

APFS_TYPE_DSTREAM_ID = 6

The key is an instance of j_dstream_id_key_t and the value is an instance of j_dstream_id_val_t.

APFS_TYPE_CRYPTO_STATE

Information about the encryption state.

APFS_TYPE_CRYPTO_STATE = 7

The key is an instance of j_crypto_key_t and the value is an instance of j_crypto_val_t.

APFS_TYPE_FILE_EXTENT

A physical extent record for a file.

APFS_TYPE_FILE_EXTENT = 8

The key is an instance of j_file_extent_key_t and the value is an instance of j_file_extent_val_t.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

74
File-System Constants
j_obj_kinds

APFS_TYPE_DIR_REC

A directory entry.

APFS_TYPE_DIR_REC = 9

The key is an instance of j_drec_key_t and the value is an instance of j_drec_val_t.

APFS_TYPE_DIR_STATS

Information about a directory.

APFS_TYPE_DIR_STATS = 10

The key is an instance of j_dir_stats_key_t and the value is an instance of j_drec_val_t.

APFS_TYPE_SNAP_NAME

The name of a snapshot.

APFS_TYPE_SNAP_NAME = 11

The key is an instance of j_snap_name_key_t and the value is an instance of j_snap_name_val_t.

APFS_TYPE_SIBLING_MAP

A mapping from a hard link to its target inode.

APFS_TYPE_SIBLING_MAP = 12

The key is an instance of j_sibling_map_key_t and the value is an instance of j_sibling_map_val_t.

APFS_TYPE_MAX_VALID

The largest valid value for a file-system objectʼs type.

APFS_TYPE_MAX_VALID = 12

APFS_TYPE_MAX

The largest value for a file-system objectʼs type.

APFS_TYPE_MAX = 15

APFS_TYPE_INVALID

An invalid object type.

APFS_TYPE_INVALID = 15

j_obj_kinds
The kind of a file-system record.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

75
File-System Constants
j_obj_kinds

typedef enum {
APFS_KIND_ANY = 0,
APFS_KIND_NEW = 1,
APFS_KIND_UPDATE = 2,
APFS_KIND_DEAD = 3,
APFS_KIND_UPDATE_REFCNT = 4,

APFS_KIND_INVALID = 255
} j_obj_kinds;

This value is stored in the kind bits of a j_phys_ext_val_t structureʼs len_and_kind field.

APFS_KIND_ANY

A record of any kind.

APFS_KIND_ANY = 0

This value isnʼt valid as the kind of a file-system record on disk. However, implementations of Apple File System can
use it internally—for example, in search queries and in tests when iterating over objects.

APFS_KIND_NEW

A new record.

APFS_KIND_NEW = 1

This record adds data that isnʼt part of any snapshots.

APFS_KIND_UPDATE

An updated record.

APFS_KIND_UPDATE = 2

This record changes data thatʼs part of an existing snapshot.

APFS_KIND_DEAD

A record thatʼs being deleted.

APFS_KIND_DEAD = 3

This value isnʼt valid as the kind of a file-system record on disk. However, implementations of Apple File System can
use it internally.

APFS_KIND_UPDATE_REFCNT

An update to the reference count of a record.

APFS_KIND_UPDATE_REFCNT = 4

This value isnʼt valid as the kind of a file-system record on disk. However, implementations of Apple File System can
use it internally.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

76
File-System Constants
j_inode_flags

APFS_KIND_INVALID

An invalid record kind.

APFS_KIND_INVALID = 255

j_inode_flags
The flags used by inodes.

typedef enum {
INODE_IS_APFS_PRIVATE = 0x00000001,
INODE_MAINTAIN_DIR_STATS = 0x00000002,
INODE_DIR_STATS_ORIGIN = 0x00000004,
INODE_PROT_CLASS_EXPLICIT = 0x00000008,
INODE_WAS_CLONED = 0x00000010,
INODE_FLAG_UNUSED = 0x00000020,
INODE_HAS_SECURITY_EA = 0x00000040,
INODE_BEING_TRUNCATED = 0x00000080,
INODE_HAS_FINDER_INFO = 0x00000100,
INODE_IS_SPARSE = 0x00000200,
INODE_WAS_EVER_CLONED = 0x00000400,
INODE_ACTIVE_FILE_TRIMMED = 0x00000800,
INODE_PINNED_TO_MAIN = 0x00001000,
INODE_PINNED_TO_TIER2 = 0x00002000,
INODE_HAS_RSRC_FORK = 0x00004000,
INODE_NO_RSRC_FORK = 0x00008000,
INODE_ALLOCATION_SPILLEDOVER = 0x00010000,

INODE_INHERITED_INTERNAL_FLAGS = (INODE_MAINTAIN_DIR_STATS),
INODE_CLONED_INTERNAL_FLAGS = (INODE_HAS_RSRC_FORK \
| INODE_NO_RSRC_FORK \
| INODE_HAS_FINDER_INFO),
} j_inode_flags;

#define APFS_VALID_INTERNAL_INODE_FLAGS (INODE_IS_APFS_PRIVATE \

| INODE_MAINTAIN_DIR_STATS \
| INODE_DIR_STATS_ORIGIN \
| INODE_PROT_CLASS_EXPLICIT \
| INODE_WAS_CLONED \
| INODE_HAS_SECURITY_EA \
| INODE_BEING_TRUNCATED \
| INODE_HAS_FINDER_INFO \
| INODE_IS_SPARSE \
| INODE_WAS_EVER_CLONED \
| INODE_ACTIVE_FILE_TRIMMED \
| INODE_PINNED_TO_MAIN \
| INODE_PINNED_TO_TIER2 \
| INODE_HAS_RSRC_FORK \

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

77
File-System Constants
j_inode_flags

| INODE_NO_RSRC_FORK \
| INODE_ALLOCATION_SPILLEDOVER)

#define APFS_INODE_PINNED_MASK (INODE_PINNED_TO_MAIN | INODE_PINNED_TO_TIER2)

INODE_IS_APFS_PRIVATE

The inode is used internally by an implementation of Apple File System.

INODE_IS_APFS_PRIVATE = 0x00000001

Inodes with this flag set arenʼt considered part of the volume. They canʼt be cloned, renamed, or deleted. Theyʼre
ignored by operations such as counting the number of files on disk, and theyʼre hidden from the user during operations
like listing the files of a directory.

This flag isnʼt reserved by Apple; implementations of the Apple File System must set this flag on any inodes they create
for their own record keeping. However, to prevent implementations from interfering with each other, an implementation
modifies inodes with this flag only if the implementation created that inode.

Appleʼs implementation uses this flag for temporary files.

See also PRIV_DIR_INO_NUM.

INODE_MAINTAIN_DIR_STATS

The inode tracks the size of all of its children.

INODE_MAINTAIN_DIR_STATS = 0x00000002

This flag is only valid on a directory, and must also be set on the directoryʼs subdirectories.

When removing the INODE_MAINTAIN_DIR_STATS flag from a directory, walk its subdirectories and remove it from
any directories that inherited it from this directory. Directories that have the INODE_DIR_STATS_ORIGIN flag set,
and subdirectories of those directories, continue to have the INODE_MAINTAIN_DIR_STATS flag set, because they
donʼt inherit it from this directory.

INODE_DIR_STATS_ORIGIN

The inode has the INODE_MAINTAIN_DIR_STATS flag set explicitly, not due to inheritance.

INODE_DIR_STATS_ORIGIN = 0x00000004

More than one directory in a hierarchy can have this flag set.

INODE_PROT_CLASS_EXPLICIT

The inodeʼs data protection class was set explicitly when the inode was created.

INODE_PROT_CLASS_EXPLICIT = 0x00000008

INODE_WAS_CLONED

The inode was created by cloning another inode.

INODE_WAS_CLONED = 0x00000010

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

78
File-System Constants
j_inode_flags

INODE_FLAG_UNUSED

Reserved.

INODE_FLAG_UNUSED = 0x00000020

Leave this flag unset when you create a new inode, and preserve its value when you modify an existing inode.

INODE_HAS_SECURITY_EA

The inode has an access control list.

INODE_HAS_SECURITY_EA = 0x00000040

INODE_BEING_TRUNCATED

The inode was truncated.

INODE_BEING_TRUNCATED = 0x00000080

This flag is used as follows to allow the truncation operation to complete after a crash:

1. The system is asked to truncate an inode

2. This flag is set on the inode
3. The system starts truncating the file
4. A crash occurs
5. In the post-crash recovery process, this flag is detected
6. The system finishes truncating the inode

Note that after a crash, the truncation operation might not resume until the next time the inode is accessed.

INODE_HAS_FINDER_INFO

The inode has a Finder info extended field.

INODE_HAS_FINDER_INFO = 0x00000100

See also INO_EXT_TYPE_FINDER_INFO.

INODE_IS_SPARSE

The inode has a sparse byte count extended field.

INODE_IS_SPARSE = 0x00000200

See also INO_EXT_TYPE_SPARSE_BYTES.

INODE_WAS_EVER_CLONED

The inode has been cloned at least once.

INODE_WAS_EVER_CLONED = 0x00000400

If this flag is set, the blocks on disk that store this inode might also be in use with another inode. For example, when
deleting this inode, you need to check reference counts before deallocating storage.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

79
File-System Constants
j_inode_flags

INODE_ACTIVE_FILE_TRIMMED

The inode is an overprovisioning file that has been trimmed.

INODE_ACTIVE_FILE_TRIMMED = 0x00000800

This file type is used only on devices running iOS. By allocating space for the file, but never writing to that space, extra
blocks are set aside for overprovisioning thatʼs performed by the underlying NAND storage.

INODE_PINNED_TO_MAIN

The inodeʼs file content is always on the main storage device.

INODE_PINNED_TO_MAIN = 0x00001000

This flag is only valid for Fusion systems. The main storage is a solid-state drive.

INODE_PINNED_TO_TIER2

The inodeʼs file content is always on the secondary storage device.

INODE_PINNED_TO_TIER2 = 0x00002000

This flag is only valid for Fusion systems. The secondary storage is a hard drive.

INODE_HAS_RSRC_FORK

The inode has a resource fork.

INODE_HAS_RSRC_FORK = 0x00004000

If this flag is set, INODE_NO_RSRC_FORK must not be set. Itʼs also valid for neither flag to be set, which implicitly
indicates that the inode doesnʼt have a resource fork.

INODE_NO_RSRC_FORK

The inode doesnʼt have a resource fork.

INODE_NO_RSRC_FORK = 0x00008000

If this flag is set, INODE_HAS_RSRC_FORK must not be set. Itʼs also valid for neither flag to be set, which implicitly
indicates that the inode doesnʼt have a resource fork.

INODE_ALLOCATION_SPILLEDOVER

The inodeʼs file content has some space allocated outside of the preferred storage tier for that file.

INODE_ALLOCATION_SPILLEDOVER = 0x00010000

See also APFS_FS_SPILLEDOVER.

INODE_INHERITED_INTERNAL_FLAGS

A bit mask of the flags that are inherited by the files and subdirectories in a directory.

INODE_INHERITED_INTERNAL_FLAGS = (INODE_MAINTAIN_DIR_STATS)

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

80
File-System Constants
j_xattr_flags

INODE_CLONED_INTERNAL_FLAGS

A bit mask of the flags that are preserved when cloning.

INODE_CLONED_INTERNAL_FLAGS = (INODE_HAS_RSRC_FORK
| INODE_NO_RSRC_FORK \
| INODE_HAS_FINDER_INFO)

APFS_VALID_INTERNAL_INODE_FLAGS

A bit mask of all valid flags.

#define APFS_VALID_INTERNAL_INODE_FLAGS (INODE_IS_APFS_PRIVATE \

| INODE_MAINTAIN_DIR_STATS \
| INODE_DIR_STATS_ORIGIN \
| INODE_PROT_CLASS_EXPLICIT \
| INODE_WAS_CLONED \
| INODE_HAS_SECURITY_EA \
| INODE_BEING_TRUNCATED \
| INODE_HAS_FINDER_INFO \
| INODE_IS_SPARSE \
| INODE_WAS_EVER_CLONED \
| INODE_ACTIVE_FILE_TRIMMED \
| INODE_PINNED_TO_MAIN \
| INODE_PINNED_TO_TIER2 \
| INODE_HAS_RSRC_FORK \
| INODE_NO_RSRC_FORK \
| INODE_ALLOCATION_SPILLEDOVER)

APFS_INODE_PINNED_MASK

A bit mask of the flags that are related to pinning.

#define APFS_INODE_PINNED_MASK (INODE_PINNED_TO_MAIN | INODE_PINNED_TO_TIER2)

j_xattr_flags
The flags used in an extended attribute record to provide additional information.

typedef enum {
XATTR_DATA_STREAM = 0x00000001,
XATTR_DATA_EMBEDDED = 0x00000002,
XATTR_FILE_SYSTEM_OWNED = 0x00000004,
XATTR_RESERVED_8 = 0x00000008,
} j_xattr_flags;

XATTR_DATA_STREAM

The attribute data is stored in a data stream.

XATTR_DATA_STREAM = 0x00000001

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

81
File-System Constants
dir_rec_flags

If this flag is set, XATTR_DATA_EMBEDDED must not be set.

XATTR_DATA_EMBEDDED

The attribute data is stored directly in the record.

XATTR_DATA_EMBEDDED = 0x00000002

If this flag is set, the size of the value be smaller than XATTR_MAX_EMBEDDED_SIZE, and XATTR_DATA_STREAM
must not be set.

XATTR_FILE_SYSTEM_OWNED

The extended attribute record is owned by the file system.

XATTR_FILE_SYSTEM_OWNED = 0x00000004

For example, this flag is used on symbolic links. The links have an extended attribute whose name is SYMLINK_EA_
NAME, and this flag is set on that attribute.

XATTR_RESERVED_8

Reserved.

XATTR_RESERVED_8 = 0x00000008

Donʼt add this flag to an extended attribute record, but preserve it if it already exists.

dir_rec_flags
The flags used by directory records.

typedef enum {
DREC_TYPE_MASK = 0x000f,
RESERVED_10 = 0x0010
} dir_rec_flags;

DREC_TYPE_MASK

The bit mask used to access the type.

DREC_TYPE_MASK = 0x000f

This bit mask is used with the flags field of j_drec_val_t.

RESERVED_10

Reserved.

RESERVED_10 = 0x0010

Donʼt set this flag. If you find a directory record with this flag set in production, file a bug against the Apple File System
implementation.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

82
File-System Constants
Inode Numbers

Inode Numbers
Inodes whose number is always the same.

#define INVALID_INO_NUM 0

#define ROOT_DIR_PARENT 1
#define ROOT_DIR_INO_NUM 2
#define PRIV_DIR_INO_NUM 3
#define SNAP_DIR_INO_NUM 6

#define MIN_USER_INO_NUM 16

INVALID_INO_NUM

An invalid inode number.

#define INVALID_INO_NUM 0

ROOT_DIR_PARENT

The inode number for the root directoryʼs parent.

#define ROOT_DIR_PARENT 1

This is a sentinel value; thereʼs no inode on disk with this inode number.

ROOT_DIR_INO_NUM

The inode number for the root directory of the volume.

#define ROOT_DIR_INO_NUM 2

PRIV_DIR_INO_NUM

The inode number for the private directory.

#define PRIV_DIR_INO_NUM 3

The private directoryʼs filename is “private-dir”. When creating a new volume, you must create a directory with this
name and inode number.

This directory isnʼt reserved by Apple; implementations of the Apple File System can use it to store their own record-
keeping information. However, to prevent implementations from interfering with each other, an implementation modi-
fies files in the private directory only if the implementation created the files.

See also INODE_IS_APFS_PRIVATE.

SNAP_DIR_INO_NUM

The inode number for the directory where snapshot metadata is stored.

#define SNAP_DIR_INO_NUM 6

Snapshot inodes are stored in the snapshot metedata tree.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

83
File-System Constants
Extended Attributes Constants

MIN_USER_INO_NUM

The smallest inode number available for user content.

#define MIN_USER_INO_NUM 16

All inode numbers less than this value are reserved.

Extended Attributes Constants

Constants used with extended attributes.

#define XATTR_MAX_EMBEDDED_SIZE 3804

#define SYMLINK_EA_NAME ”com.apple.fs.symlink”

XATTR_MAX_EMBEDDED_SIZE

The largest size, in bytes, of an extended attribute whose value is stored directly in the record.

#define XATTR_MAX_EMBEDDED_SIZE 3804

For information about embedded values, see j_xattr_val_t.

SYMLINK_EA_NAME

The name of an extended attribute for a symbolic link whose value is the target file.

#define SYMLINK_EA_NAME ”com.apple.fs.symlink”

File-System Object Constants

No overview available.

#define OWNING_OBJ_ID_INVALID ~0ULL

#define OWNING_OBJ_ID_UNKNOWN ~1ULL

#define JOBJ_MAX_KEY_SIZE 832

#define JOBJ_MAX_VALUE_SIZE 3808

#define MIN_DOC_ID 3

MIN_DOC_ID

The smallest document identifier available for user content.

#define MIN_DOC_ID 3

All document identifiers less than this value are reserved.

File Extent Constants

No overview available.

#define FEXT_CRYPTO_ID_IS_TWEAK 0x01

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

84
File-System Constants
File Modes

File Modes
The values used by the mode field of j_inode_val_t to indicate a fileʼs mode.

typedef uint16_t mode_t;

#define S_IFMT 0170000

#define S_IFIFO 0010000

#define S_IFCHR 0020000
#define S_IFDIR 0040000
#define S_IFBLK 0060000
#define S_IFREG 0100000
#define S_IFLNK 0120000
#define S_IFSOCK 0140000
#define S_IFWHT 0160000

The names, values, and meanings of these constants are the same as the constants provided by <sys/stat.h>.
These values are the same as the values defined in Directory Entry File Types, except for a bit shift.

mode_t

A file mode.

typedef uint16_t mode_t;

S_IFMT

The bit mask used to access the file type.

#define S_IFMT 0170000

S_IFIFO

A named pipe.

#define S_IFIFO 0010000

S_IFCHR

A character-special file.

#define S_IFCHR 0020000

S_IFDIR

A directory.

#define S_IFDIR 0040000

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

85
File-System Constants
Directory Entry File Types

S_IFBLK

A block-special file.

#define S_IFBLK 0060000

S_IFREG

A regular file.

#define S_IFREG 0100000

S_IFLNK

A symbolic link.

#define S_IFLNK 0120000

S_IFSOCK

A socket.

#define S_IFSOCK 0140000

S_IFWHT

A whiteout.

#define S_IFWHT 0160000

Directory Entry File Types

Values used by the flags field of j_drec_val_t to indicate a directory entryʼs type.

#define DT_UNKNOWN 0
#define DT_FIFO 1
#define DT_CHR 2
#define DT_DIR 4
#define DT_BLK 6
#define DT_REG 8
#define DT_LNK 10
#define DT_SOCK 12
#define DT_WHT 14

These values are the same as the values defined in File Modes, except for a bit shift.

DT_UNKNOWN

An unknown directory entry.

#define DT_UNKNOWN 0

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

86
File-System Constants
Directory Entry File Types

DT_FIFO

A named pipe

#define DT_FIFO 1

DT_CHR

A character-special file.

#define DT_CHR 2

DT_DIR

A directory.

#define DT_DIR 4

DT_BLK

A block-special file.

#define DT_BLK 6

DT_REG

A regular file.

#define DT_REG 8

DT_LNK

A symbolic link.

#define DT_LNK 10

DT_SOCK

A socket.

#define DT_SOCK 12

DT_WHT

A whiteout.

#define DT_WHT 14

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

87
Data Streams
Short pieces of information like a fileʼs name are stored inside the data structures that contain metadata. Data thatʼs
too large to store inline is stored separately, in a data stream. This includes the contents of files, and the value of some
attributes.

j_phys_ext_key_t
The key half of a physical extent record.

typedef struct j_phys_ext_key {

j_key_t hdr;
} __attribute__((packed)) j_phys_ext_key_t;

hdr

The recordʼs header.

j_key_t hdr;

The object identifier in the header is the physical block address of the start of the extent. The type in the header is
always APFS_TYPE_EXTENT.

j_phys_ext_val_t
The value half of a physical extent record.

typedef struct j_phys_ext_val {

uint64_t len_and_kind;
uint64_t owning_obj_id;
int32_t refcnt;
} __attribute__((packed)) j_phys_ext_val_t;

#define PEXT_LEN_MASK 0x0fffffffffffffffULL

#define PEXT_KIND_MASK 0xf000000000000000ULL
#define PEXT_KIND_SHIFT 60

len_and_kind

A bit field that contains the length of the extent and its kind.

uint64_t len_and_kind;

The extentʼs length is a uint64_t value, accessed as len_and_kind & PEXT_LEN_MASK, and measured in
blocks. The extentʼs kind is a j_obj_kinds value, accessed as (len_and_kind & PEXT_KIND_MASK) >>
PEXT_KIND_SHIFT.

For a volume that has no snapshots, the kind is always APFS_KIND_NEW.

88
Data Streams
j_file_extent_key_t

owning_obj_id

The identifier of the file system record thatʼs using this extent.

uint64_t owning_obj_id;

If the owning record is an inode, this field contains the inodeʼs private identifier (the private_id field of
j_inode_val_t). If the owning record is an extended attribute, this field contains the extended attributeʼs
record identifier (the identifier from the hdr field of j_xattr_key_t).

refcnt

The reference count.

int32_t refcnt;

The extent can be garbage collected when its reference count reaches zero.

PEXT_LEN_MASK

The bit mask used to access the extent length.

#define PEXT_LEN_MASK 0x0fffffffffffffffULL

PEXT_KIND_MASK

The bit mask used to access the extent kind.

#define PEXT_KIND_MASK 0xf000000000000000ULL

PEXT_KIND_SHIFT

The bit shift used to access the extent kind.

#define PEXT_LEN_SHIFT 60

j_file_extent_key_t
The key half of a file extent record.

typedef struct j_file_extent_key {

j_key_t hdr;
uint64_t logical_addr;
} __attribute__((packed)) j_file_extent_key_t;

hdr

The recordʼs header.

j_key_t hdr;

The object identifier in the header is the file-system objectʼs identifier. The type in the header is always
APFS_TYPE_FILE_EXTENT.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

89
Data Streams
j_file_extent_val_t

logical_addr

The offset within the fileʼs data, in bytes, for the data stored in this extent.

uint64_t logical_addr;

j_file_extent_val_t
The value half of a file extent record.

typedef struct j_file_extent_val {

uint64_t len_and_flags;
uint64_t phys_block_num;
uint64_t crypto_id;
} __attribute__((packed)) j_file_extent_val_t;

#define J_FILE_EXTENT_LEN_MASK 0x00ffffffffffffffULL

#define J_FILE_EXTENT_FLAG_MASK 0xff00000000000000ULL
#define J_FILE_EXTENT_FLAG_SHIFT 56

len_and_flags

A bit field that contains the length of the extent and its flags.

uint64_t len_and_flags;

The extentʼs length is a uint64_t value, accessed as len_and_kind & J_FILE_EXTENT_LEN_MASK, and
measured in bytes. The length must be a multiple of the block size defined by the nx_block_size field of
nx_superblock_t. The extentʼs flags are accessed as (len_and_kind & J_FILE_EXTENT_FLAG_MASK) >>
J_FILE_EXTENT_FLAG_SHIFT.

There are currently no flags defined.

phys_block_num

The physical block address that the extent starts at.

uint64_t phys_block_num;

crypto_id

The encryption used in this extent.

uint64_t crypto_id;

This value matches the obj_id field in the j_key_t key that corresponds to a j_crypto_val_t value.

The default value for this field is the value of the default_crypto_id field of the j_dstream_t for the data stream
that this extent is part of.

J_FILE_EXTENT_LEN_MASK

The bit mask used to access the extent length.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

90
Data Streams
j_dstream_id_key_t

#define J_FILE_EXTENT_LEN_MASK 0x00ffffffffffffffULL

J_FILE_EXTENT_FLAG_MASK

The bit mask used to access the flags.

#define J_FILE_EXTENT_FLAG_MASK 0xff00000000000000ULL

J_FILE_EXTENT_FLAG_SHIFT

The bit shift used to access the flags.

#define J_FILE_EXTENT_FLAG_SHIFT 56

j_dstream_id_key_t
The key half of a directory-information record.

typedef struct j_dstream_id_key_t {

j_key_t hdr;
} __attribute__((packed)) j_dir_stats_key_t;

hdr

The recordʼs header.

j_key_t hdr;

The object identifier in the header is the file-system objectʼs identifier. The type in the header is always
APFS_TYPE_DSTREAM_ID.

j_dstream_id_val_t
The value half of a data stream record.

typedef struct j_dstream_id_val {

uint32_t refcnt;
} __attribute__((packed)) j_dstream_id_val_t;

refcnt

The reference count.

uint32_t refcnt;

The data stream record can be garbage collected when its reference count reaches zero.

j_xattr_dstream_t
A data stream for extended attributes.

typedef struct j_xattr_dstream {

uint64_t xattr_obj_id;
j_dstream_t dstream;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

91
Data Streams
j_dstream_t

} j_xattr_dstream_t;

To access the data in the stream, read the object identifier and then find the corresponding extents.

xattr_obj_id

The identifier for the data stream.

uint64_t xattr_obj_id;

This field contains the record identifier of the data stream that owns this record.

dstream

Information about the data stream.

j_dstream_t dstream;

j_dstream_t
Information about a data stream.

typedef struct j_dstream {

uint64_t size;
uint64_t alloced_size;
uint64_t default_crypto_id;
uint64_t total_bytes_written;
uint64_t total_bytes_read;
} __attribute__((aligned(8), packed)) j_dstream_t;

This structure is used inside j_xattr_dstream_t.

size

The size, in bytes, of the data.

uint64_t size;

alloced_size

The total space allocated for the data stream, including any unused space.

uint64_t alloced_size;

default_crypto_id

The default encryption used in this data stream.

uint64_t default_crypto_id;

This value matches the obj_id field in the j_key_t key that corresponds to a j_crypto_val_t value.

This value is used as the default value by file extents (j_file_extent_val_t) that make up this data stream.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

92
Data Streams
j_dstream_t

total_bytes_written

The total number of bytes that have been written to this data stream.

uint64_t total_bytes_written;

The value of this field increases every time a write operation occurs. This value is allowed to overflow and restart from
zero.

total_bytes_read

The total number of bytes that have been read from this data stream.

uint64_t total_bytes_read;

The value of this field increases every time a read operation occurs. This value is allowed to overflow and restart from
zero.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

93
Extended Fields
Directory entries and inodes use extended fields to store a dynamically extensible set of member fields.

To determine whether a directory entry or an inode has any extended fields, find the table of contents entry for the
file-system record, and then compare the recorded size to the size of the structure. For example:

kvloc_t toc_entry = /* assume this exists */

if (toc_entry.v.len == sizeof(j_drec_val_t)) {
// no extended fields
} else {
// at least one extended field
}

Both j_drec_val_t and j_inode_val_t have an xfields field that contains several kinds of data, stored one
after another, ordered as follows:

1. An instance of xf_blob_t, which tells you how many extended fields there are, and how many bytes they take
up on disk.

2. An array of instances of x_field_t, one for each extended field, which tells you the fieldʼs type and size.

3. An array of extended-field data, aligned to eight-byte boundaries.

The arrays of extended-field metadata (x_field_t) and extended-field data are stored in the same order. The
extended-field dataʼs type depends on the field. For a list of field types, see Extended-Field Types.

xf_blob_t
A collection of extended attributes.

typedef struct xf_blob {

uint16_t xf_num_exts;
uint16_t xf_used_data;
uint8_t xf_data[];
} xf_blob_t;

Directory entries (j_drec_val_t) and inodes (j_inode_val_t) use this data type to store their extended fields.

xf_num_exts

The number of extended attributes.

uint16_t xf_num_exts;

xf_used_data

The amount of space, in bytes, used to store the extended attributes.

uint16_t xf_used_data;

This total includes both the space used to store metadata, as instances of x_field_t, and values.

94
Extended Fields
x_field_t

xf_data[]

The extended fields.

uint8_t xf_data[];

This field contains an array of instances of x_field_t, followed by the extended field data.

x_field_t
An extended fieldʼs metadata.

typedef struct x_field {

uint8_t x_type;
uint8_t x_flags;
uint16_t x_size;
} x_field_t;

This type is used by xf_blob_t to store an array of extended fields. Within the array, each extended field must have
a unique type.

The extended fieldʼs data is stored outside of this structure, as part of the space set aside by the directory entry or
inode.

x_type

The extended fieldʼs data type.

uint8_t x_type;

For possible values, see Extended-Field Types.

x_flags

The extended fieldʼs flags.

uint8_t x_flags;

For the values used in this bit field, see Extended-Field Flags.

x_size

The size, in bytes, of the data stored in the extended field.

uint16_t x_size;

Extended-Field Types
Values used by the x_type field of x_field_t to indicate an extended fieldʼs type.

#define DREC_EXT_TYPE_SIBLING_ID 1

#define INO_EXT_TYPE_SNAP_XID 1
#define INO_EXT_TYPE_DELTA_TREE_OID 2

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

95
Extended Fields
Extended-Field Types

#define INO_EXT_TYPE_DOCUMENT_ID 3
#define INO_EXT_TYPE_NAME 4
#define INO_EXT_TYPE_PREV_FSIZE 5
#define INO_EXT_TYPE_RESERVED_6 6
#define INO_EXT_TYPE_FINDER_INFO 7
#define INO_EXT_TYPE_DSTREAM 8
#define INO_EXT_TYPE_RESERVED_9 9
#define INO_EXT_TYPE_DIR_STATS_KEY 10
#define INO_EXT_TYPE_FS_UUID 11
#define INO_EXT_TYPE_RESERVED_12 12
#define INO_EXT_TYPE_SPARSE_BYTES 13
#define INO_EXT_TYPE_RDEV 14

DREC_EXT_TYPE_SIBLING_ID

The sibling identifier for a directory record (uint64_t).

#define DREC_EXT_TYPE_SIBLING_ID 1

The corresponding sibling-link record has the same identifier in the sibling_id field of j_sibling_key_t.

This extended field is used only for hard links.

INO_EXT_TYPE_SNAP_XID

The transaction identifier for a snapshot (xid_t).

#define INO_EXT_TYPE_SNAP_XID 1

INO_EXT_TYPE_DELTA_TREE_OID

The virtual object identifier of the file-system tree that corresponds to a snapshotʼs extent delta list (oid_t).

#define INO_EXT_TYPE_DELTA_TREE_OID 2

The tree objectʼs subtype is always OBJECT_TYPE_FSTREE.

INO_EXT_TYPE_DOCUMENT_ID

The fileʼs document identifier (uint32_t).

#define INO_EXT_TYPE_DOCUMENT_ID 3

The document identifier lets applications keep track of the document during operations such as atomic save, where
one folder replaces another. The document identifier remains associated with the full path, not just with the inode
thatʼs currently at that path. Implementations of Apple File System must preserve the document identifier when the
inode at that path is replaced.

Both documents that are stored as a bundle and documents that are stored as a single file can have a document
identifier assigned.

Valid document identifiers are greater than MIN_DOC_ID and less than UINT32_MAX - 1. For the next document
identifier that will be assigned, see the apfs_next_doc_id field of apfs_superblock_t.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

96
Extended Fields
Extended-Field Types

INO_EXT_TYPE_NAME

The name of the file, represented as a null-terminated UTF-8 string.

#define INO_EXT_TYPE_NAME 4

This extended field is used only for hard links: The name stored in the inode is the name of the primary link to the file,
and the name of the hard link is stored in this extended field.

INO_EXT_TYPE_PREV_FSIZE

The fileʼs previous size (uint64_t).

#define INO_EXT_TYPE_PREV_FSIZE 5

This extended field is used for recovering after a crash. If itʼs set on an inode, truncate the file back to the size contained
in this field.

INO_EXT_TYPE_RESERVED_6

Reserved.

#define INO_EXT_TYPE_RESERVED_6 6

Donʼt create extended fields of this type in your own code. Preserve the value of any extended fields of this type.

INO_EXT_TYPE_FINDER_INFO

Opaque data stored and used by Finder (32 bytes).

#define INO_EXT_TYPE_FINDER_INFO 7

INO_EXT_TYPE_DSTREAM

A data stream (j_dstream_t).

#define INO_EXT_TYPE_DSTREAM 8

INO_EXT_TYPE_RESERVED_9

Reserved.

#define INO_EXT_TYPE_RESERVED_9 9

Donʼt create extended fields of this type. When you modify an existing volume, preserve the contents of any extended
fields of this type.

INO_EXT_TYPE_DIR_STATS_KEY

Statistics about a directory (j_dir_stats_val_t).

#define INO_EXT_TYPE_DIR_STATS_KEY 10

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

97
Extended Fields
Extended-Field Flags

INO_EXT_TYPE_FS_UUID

The UUID of a file system thatʼs automatically mounted in this directory (uuid_t).

#define INO_EXT_TYPE_FS_UUID 11

This value matches the value of the apfs_vol_uuid field of apfs_superblock_t.

INO_EXT_TYPE_RESERVED_12

Reserved.

#define INO_EXT_TYPE_RESERVED_12 12

Donʼt create extended fields of this type. If you find an object of this type in production, file a bug against the Apple
File System implementation.

INO_EXT_TYPE_SPARSE_BYTES

The number of sparse bytes in the data stream (uint64_t).

#define INO_EXT_TYPE_SPARSE_BYTES 13

INO_EXT_TYPE_RDEV

The device identifier for a block- or character-special device (uint32_t).

#define INO_EXT_TYPE_RDEV 14

This extended field stores the same information as the st_rdev field of the stat structure defined in <sys/stat.h>.

Extended-Field Flags
Flags used by the x_flags field of x_field_t.

#define XF_DATA_DEPENDENT 0x0001

#define XF_DO_NOT_COPY 0x0002
#define XF_RESERVED_4 0x0004
#define XF_CHILDREN_INHERIT 0x0008
#define XF_USER_FIELD 0x0010
#define XF_SYSTEM_FIELD 0x0020
#define XF_RESERVED_40 0x0040
#define XF_RESERVED_80 0x0080

XF_DATA_DEPENDENT

The data in this extended field depends on the fileʼs data.

#define XF_DATA_DEPENDENT 0x0001

When the file data changes, this extended field must be updated to match the new data. If itʼs not possible to update
the field—for example, because the Apple File System implementation doesnʼt recognize the fieldʼs type—the field
must be removed.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

98
Extended Fields
Extended-Field Flags

XF_DO_NOT_COPY

When copying this file, omit this extended field from the copy.

#define XF_DO_NOT_COPY 0x0002

XF_RESERVED_4

Reserved.

#define XF_RESERVED_4 0x0004

Donʼt set this flag, but preserve it if itʼs already set.

XF_CHILDREN_INHERIT

When creating a new entry in this directory, copy this extended field to the new directory entry.

#define XF_CHILDREN_INHERIT 0x0008

XF_USER_FIELD

This extended field was added by a user-space program.

#define XF_USER_FIELD 0x0010

XF_SYSTEM_FIELD

This extended field was added by the kernel, by the implementation of Apple File System, or by another system com-
ponent.

#define XF_SYSTEM_FIELD 0x0020

Extended fields with this flag set canʼt be removed or modified by a program running in user space.

XF_RESERVED_40

When sharing this file, omit this extended field from the shared copy.

#define XF_RESERVED_40 0x0040

Donʼt set this flag, but preserve it if itʼs already set.

XF_RESERVED_80

Reserved.

#define XF_RESERVED_80 0x0080

Donʼt set this flag, but preserve it if itʼs already set.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

99
Siblings
Hard links that all refer to the same inode are called siblings. Each sibling has its own identifier thatʼs used instead
of the shared inode number when siblings need to be distinguished. For example, some Carbon APIs on macOS use
sibling identifiers.

The sibling whose identifier is the lowest number is called the primary link. The other siblings copy various properties
of the primary link, as discussed in j_inode_val_t.

You use sibling links and sibling maps to convert between sibling identifiers and inode numbers. Sibling-link records
let you find all the hard links whose target is a given inode. Sibling-map records let you find the target inode of a given
hard link.

j_sibling_key_t
The key half of a sibling-link record.

typedef struct j_sibling_key {

j_key_t hdr;
uint64_t sibling_id;
} __attribute__((packed)) j_sibling_key_t;

hdr

The recordʼs header.

j_key_t hdr;

The object identifier in the header is the file-system objectʼs identifier, that is, its inode number. The type in the header
is always APFS_TYPE_SIBLING_LINK.

sibling_id

The siblingʼs unique identifier.

uint64_t sibling_id;

This value matches the object identifier for the sibling map record (j_sibling_key_t).

j_sibling_val_t
The value half of a sibling-link record.

typedef struct j_sibling_val {

uint64_t parent_id;
uint16_t name_len;
uint8_t name[0];
} __attribute__((packed)) j_sibling_val_t;

100
Siblings
j_sibling_map_key_t

parent_id

The object identifier for the inode thatʼs the parent directory.

uint64_t parent_id;

name_len

The length of the name, including the final null character (U+0000).

uint16_t name_len;

name

The name, represented as a null-terminated UTF-8 string.

uint8_t name[0];

j_sibling_map_key_t
The key half of a sibling-map record.

typedef struct j_sibling_map_key {

j_key_t hdr;
} __attribute__((packed)) j_sibling_map_key_t;

hdr

The recordʼs header.

j_key_t hdr;

The object identifier in the header is the siblingʼs unique identifier, which matches the sibling_id field of
j_sibling_key_t. The type in the header is always APFS_TYPE_SIBLING_MAP.

j_sibling_map_val_t
The value half of a sibling-map record.

typedef struct j_sibling_map_val {

uint64_t file_id;
} __attribute__((packed)) j_sibling_map_val_t;

file_id

The inode number of the underlying file.

uint64_t file_id;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

101
Snapshot Metadata
Snapshots let you get a stable, read-only copy of the filesystem at a given point in time—for example, while updating a
backup of the entire drive. Snapshots are designed to be fast and inexpensive to create; however, deleting a snapshot
involves more work.

j_snap_metadata_key_t
The key half of a record containing metadata about a snapshot.

typedef struct j_snap_metadata_key {

j_key_t hdr;
} __attribute__((packed)) j_snap_metadata_key_t;

hdr

The recordʼs header.

j_key_t hdr;

The object identifier in the header is the snapshotʼs transaction identifier. The type in the header is always
APFS_TYPE_SNAP_METADATA.

j_snap_metadata_val_t
The value half of a record containing metadata about a snapshot.

typedef struct j_snap_metadata_val {

oid_t extentref_tree_oid;
oid_t sblock_oid;
uint64_t create_time;
uint64_t change_time;
uint64_t inum;
uint32_t extentref_tree_type;
uint32_t flags;
uint16_t name_len;
uint8_t name[0];
} __attribute__((packed)) j_snap_metadata_val_t;

extentref_tree_oid

The virtual object identifier of the B-tree that stores extents information.

oid_t extentref_tree_oid;

sblock_oid

The virtual object identifier of the volume superblock.

oid_t sblock_oid;

102
Snapshot Metadata
j_snap_name_key_t

create_time

The time that this snapshot was created.

uint64_t create_time;

This timestamp is represented as the number of nanoseconds since January 1, 1970 at 0 00 UTC, disregarding leap
seconds.

change_time

The time that this snapshot was last modified.

uint64_t change_time;

This timestamp is represented as the number of nanoseconds since January 1, 1970 at 0 00 UTC, disregarding leap
seconds.

inum

uint64_t inum;

extentref_tree_type

The type of the B-tree that stores extents information.

uint32_t extentref_tree_type;

flags

A bit field that contains additional information about a snapshot metadata record.

uint32_t flags;

For the values used in this bit field, see snap_meta_flags.

name_len

The length of the snapshotʼs name, including the final null character (U+0000).

uint16_t name_len;

name

The snapshotʼs name, represented as a null-terminated UTF-8 string.

uint8_t name[0];

j_snap_name_key_t
The key half of a snapshot name record.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

103
Snapshot Metadata
j_snap_name_val_t

typedef struct j_snap_name_key {

j_key_t hdr;
uint16_t name_len;
uint8_t name[0];
} __attribute__((packed)) j_snap_name_key_t;

hdr

The recordʼs header.

j_key_t hdr;

The object identifier in the header is always ~0ULL. The type in the header is always APFS_TYPE_SNAP_NAME.

name_len

The length of the extended attributeʼs name, including the final null character (U+0000).

uint16_t name_len;

name

The extended attributeʼs name, represented as a null-terminated UTF-8 string.

uint8_t name[0];

j_snap_name_val_t
The value half of a snapshot name record.

typedef struct j_snap_name_val {

xid_t snap_xid;
} __attribute__((packed)) j_snap_name_val_t;

snap_xid

The last transaction identifier included in the snapshot.

xid_t snap_xid;

snap_meta_flags
No overview available.

typedef enum {
SNAP_META_PENDING_DATALESS = 0x00000001,
} snap_meta_flags;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

104
B-Tree
The B-trees used in Apple File System are implemented using the btree_node_phys_t structure to represent a
node. The same structure is used for all nodes in a tree. Within a node, storage is divided into several areas:

• Information about the node

• The table of contents, which lists the location of keys and values
• Storage for the keys
• Storage for the values
• Information about the entire tree

The figure below shows the storage areas of a typical root node.

The instance of btree_node_phys_t stores information about this B-tree node, such as its flags and the location of
its keys, and is always located at the beginning of the block. For a root node, an instance of btree_info_t is located
at the end of the block, and contains information such as the sizes of keys and values, the total number of keys in the
tree, and the number of nodes in the tree. Nonroot nodes omit btree_info_t. The rest of the block (the btn_data
field of btree_node_phys_t) is organized dynamically.

Compared to other B-tree implementations, this data structure has some unique characteristics. Traversal is always
done from the root node because nodes donʼt have parent or sibling pointers. All values are stored in leaf nodes, which
makes these B+ trees, and the values in nonleaf nodes are object identifiers of child nodes. The keys, values, or both
can be of variable size; if the keys and values of a node are both fixed in size, some optimizations for the table of
contents are possible.

Keys and Values

The keys and values are stored starting at opposite ends of the B-tree nodeʼs storage area, with free space thatʼs
available for new keys or values in the available portion of the storage area between them. The key and value areas
grow toward each other into their shared free space. Free space within the key area and within the value area is
organized using a free list. For example, free space appears outside the shared free space when an entry is removed
from a B-tree. The figure below shows free space for keys and values in a typical nonroot node.

The locations of keys and values are stored as offsets, which uses less on-disk space than storing the full location.
The offset to a key is counted from the beginning of the key area to the beginning of the key. The offset to a value is
counted from the end of the value area to the beginning of the value.

105
B-Tree
btree_node_phys_t

Keys and value are normally aligned to eight-byte boundaries when stored. The length recorded for a key or value in
the table of contents omits any padding needed for alignment. If the BTREE_KV_NONALIGNED flag is set, keys and
values are stored without padding.

If the BTREE_ALLOW_GHOSTS flag is set on the B-tree, the tree can contain keys that have no value.

Table of Contents

The table of contents stores the location of each key and value that form a key-value pair.

If the BTNODE_FIXED_KV_SIZE flag is set, the table of contents stores only the offsets for keys and values. Other-
wise, it stores both their offsets and lengths.

Free space within the table of contents is located at the end. If thereʼs no free space remaining, but a new entry
is needed, the table of contents area must be expanded. The entire key area is shifted to make space available,
using some of the shared free space for key space, and some space from the beginning of the key space for the
table of contents. Because the offset to a key is counted relative to the beginning of the key area, moving the en-
tire key area doesnʼt invalidate any of these offsets. Likewise, when the table of contents has too much unused
space, it shrinks, and the key area is shifted into the space from the table of contents. Appleʼs implementation uses
BTREE_TOC_ENTRY_INCREMENT and BTREE_TOC_ENTRY_MAX_UNUSED to determine when to expand or shrink the
table of contents.

Note

When the BTNODE_FIXED_KV_SIZE flag is set, Appleʼs implementation allocates enough space for the table of
contents to avoid the need to expand it later. This is possible because the maximum number of entries is known,
as well as the size of an entry. However, if the BTREE_ALLOW_GHOSTS flag is also set, the table of contents
might still need to expand.

Key Comparison

The entries in the table of contents are sorted by key. The comparison function used for sorting depends on the keyʼs
type. Object map B-trees are sorted by object identifier and then by transaction identifier. Free queue B-trees are
sorted by transaction identifier and then by physical address. File-system records are sorted according to the rules
listed in File-System Objects.

btree_node_phys_t
A B-tree node.

struct btree_node_phys {
obj_phys_t btn_o;
uint16_t btn_flags;
uint16_t btn_level;
uint32_t btn_nkeys;
nloc_t btn_table_space;
nloc_t btn_free_space;
nloc_t btn_key_free_list;
nloc_t btn_val_free_list;
uint64_t btn_data[];

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

106
B-Tree
btree_node_phys_t

};
typedef struct btree_node_phys btree_node_phys_t;

The locations of the key and value areas arenʼt stored explicitly. The key area begins after the end of the table of
contents and ends before the start of the shared free space. The value area begins after the end of shared free space
and ends at the end of the B-tree node (for nonroot nodes) or before the instance of btree_info_t thatʼs at the end
of a root node.

btn_o

The objectʼs header.

obj_phys_t btn_o;

btn_flags

The B-tree nodeʼs flags.

uint16_t btn_flags;

For the values used in this bit field, see B-Tree Node Flags.

btn_level

The number of child levels below this node.

uint16_t btn_level;

For example, the value of this field is 0 for a leaf node and 1 for the immediate parent of a leaf node. Likewise, the
height of a tree is one plus the value of this field on the treeʼs root node.

btn_nkeys

The number of keys stored in this node.

uint32_t btn_nkeys;

btn_table_space

The location of the table of contents.

nloc_t btn_table_space;

The offset for the table of contents is counted from the beginning of the nodeʼs btn_data field to the beginning of
the table of contents.

If the BTNODE_FIXED_KV_SIZE flag is set, the table of contents is an array of instances of kvoff_t; otherwise, itʼs
an array of instances of kvloc_t.

btn_free_space

The location of the shared free space for keys and values.

nloc_t btn_free_space;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

107
B-Tree
btree_info_fixed_t

The locationʼs offset is counted from the beginning of the key area to the beginning of the free space.

btn_key_free_list

A linked list that tracks free key space.

nloc_t btn_key_free_list;

The offset from the beginning of the key area to the first available space for a key is stored in the off field, and the
total amount of free key space is stored in the len field. Each free space stores an instance of nloc_t whose len
field indicates the size of that free space and whose off field contains the location of the next free space.

btn_val_free_list

A linked list that tracks free value space.

nloc_t btn_val_free_list;

The offset from the end of the value area to the first available space for a value is stored in the off field, and the total
amount of free value space is stored in the len field. Each free space stores an instance of nloc_t whose len field
indicates the size of that free space and whose off field contains the location of the next free space.

btn_data

The nodeʼs storage area.

uint64_t btn_data[];

This area contains the table of contents, keys, free space, and values. A root node also has as an instance of
btree_info_t at the end of its storage area. For more information, see B-tree.

btree_info_fixed_t
Static information about a B-tree.

struct btree_info_fixed {
uint32_t bt_flags;
uint32_t bt_node_size;
uint32_t bt_key_size;
uint32_t bt_val_size;
};
typedef struct btree_info_fixed btree_info_fixed_t;

This information doesnʼt change over time as the B-tree is modified. Itʼs stored separately from the rest of the infor-
mation in btree_info_t, which does change, to enable this information to be cached more easily.

bt_flags

The B-treeʼs flags.

uint32_t bt_flags;

For the values used in this bit field, see B-Tree Flags.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

108
B-Tree
btree_info_t

bt_node_size

The on-disk size, in bytes, of a node in this B-tree.

uint32_t bt_node_size;

Leaf nodes, nonleaf nodes, and the root node are all the same size.

bt_key_size

The size of a key, or 0 if the keys have variable size.

uint32_t bt_key_size;

If this field has a value of 0, the btn_flags field of instances of btree_node_phys_t in this tree must not include
BTNODE_FIXED_KV_SIZE.

bt_val_size

The size of a value, or 0 if the values have variable size.

uint32_t bt_val_size;

If this field has a value of 0, the btn_flags field of instances of btree_node_phys_t for leaf nodes in this
tree must not include BTNODE_FIXED_KV_SIZE. Nonleaf nodes in a tree with variable-size values include
BTNODE_FIXED_KV_SIZE, because the values stored in those nodes are the object identifiers of their child nodes,
and object identifiers have a fixed size.

btree_info_t
Information about a B-tree.

struct btree_info {
btree_info_fixed_t bt_fixed;
uint32_t bt_longest_key;
uint32_t bt_longest_val;
uint64_t bt_key_count;
uint64_t bt_node_count;
};
typedef struct btree_info btree_info_t;

This information appears only in a root node, stored at the end of the node.

btree_info_fixed_t

Information about the B-tree that doesnʼt change over time.

btree_info_fixed_t bt_fixed;

bt_longest_key

The length, in bytes, of the longest key that has ever been stored in the B-tree.

uint32_t bt_longest_key;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

109
B-Tree
nloc_t

bt_longest_val

The length, in bytes, of the longest value that has ever been stored in the B-tree.

uint32_t bt_longest_val;

bt_key_count

The number of keys stored in the B-tree.

uint64_t bt_key_count;

bt_node_count

The number of nodes stored in the B-tree.

uint64_t bt_node_count;

nloc_t
A location within a B-tree node.

struct nloc {
uint16_t off;
uint16_t len;
};
typedef struct nloc nloc_t;

#define BTOFF_INVALID 0xffff

off

The offset, in bytes.

uint16_t off;

Depending on the data type that contains this location, the offset is either implicitly positive or negative, and is counted
starting at different points in the B-tree node.

len

The length, in bytes.

uint16_t len;

BTOFF_INVALID

An invalid offset.

#define BTOFF_INVALID 0xffff

This value is stored in the off field of nloc_t to indicate that thereʼs no offset. For example, the last entry in a free
list has no entry after it, so it uses this value for its off field.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

110
B-Tree
kvloc_t

kvloc_t
The location, within a B-tree node, of a key and value.

struct kvloc {
nloc_t k;
nloc_t v;
};
typedef struct kvloc kvloc_t;

The B-tree nodeʼs table of contents uses this structure when the keys and values are not both fixed in size.

nloc_t

The location of the key.

nloc_t k;

nloc_t

The location of the value.

nloc_t v;

kvoff_t
The location, within a B-tree node, of a fixed-size key and value.

struct kvoff {
uint16_t k;
uint16_t v;
};
typedef struct kvoff kvoff_t;

The B-tree nodeʼs table of contents uses this structure when the keys and values are both fixed in size. The meaning
of the offsets stored in this structureʼs k and v fields is the same as the meaning of the off field in an instance of
nloc_t. This structure doesnʼt have a field thatʼs equivalent to the len field of nloc_t—the key and value lengths
are always the same, and omitting them from the table of contents saves space.

The offset of the key.

uint16_t k;

The offset of the value.

uint16_t v;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

111
B-Tree
B-Tree Flags

B-Tree Flags
The flags used to describe configuration options for a B-tree.

#define BTREE_UINT64_KEYS 0x00000001

#define BTREE_SEQUENTIAL_INSERT 0x00000002
#define BTREE_ALLOW_GHOSTS 0x00000004
#define BTREE_EPHEMERAL 0x00000008
#define BTREE_PHYSICAL 0x00000010
#define BTREE_NONPERSISTENT 0x00000020
#define BTREE_KV_NONALIGNED 0x00000040

BTREE_UINT64_KEYS

Code that works with the B-tree should enable optimizations to make comparison of keys fast.

#define BTREE_UINT64_KEYS 0x00000001

This is a hint used by Appleʼs implementation.

BTREE_SEQUENTIAL_INSERT

Code that works with the B-tree should enable optimizations to keep the B-tree compact during sequential insertion
of entries.

#define BTREE_SEQUENTIAL_INSERT 0x00000002

This is a hint used by Appleʼs implementation.

Normally, nodes are split in half when they become almost full. With this flag set, a new node is added to provide the
needed space, instead of splitting a node thatʼs almost full. This yields a tree with nodes that are almost full instead
of nodes that are about half full.

BTREE_ALLOW_GHOSTS

The table of contents is allowed to contain keys that have no corresponding value.

#define BTREE_ALLOW_GHOSTS 0x00000004

In the table of contents, a ghost is indicated by a value whose location offset is BTOFF_INVALID.

The meaning of a ghost depends on context—it can indicate a key that has been deleted and should be ignored, or
a key whose value is implicit from context. For example, in the space managerʼs free queue, a ghost indicates a free
extent thatʼs one block long.

Using ghosts to store an implicit value allows more entries to be stored in some circumstances because no space in
the value area is used by the ghost.

BTREE_EPHEMERAL

The nodes in the B-tree use ephemeral object identifiers to link to child nodes.

#define BTREE_EPHEMERAL 0x00000008

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

112
B-Tree
B-Tree Table of Contents Constants

If this flag is set, BTREE_PHYSICAL must not be set. If neither flag is set, nodes in the B-tree use virtual object
identifiers to link to their child nodes.

BTREE_PHYSICAL

The nodes in the B-tree use physical object identifiers to link to child nodes.

#define BTREE_PHYSICAL 0x00000010

If this flag is set, BTREE_EPHEMERAL must not be set. If neither flag is set, nodes in the B-tree use virtual object
identifiers to link to their child nodes.

BTREE_NONPERSISTENT

The B-tree isnʼt persisted across unmounting.

#define BTREE_NONPERSISTENT 0x00000020

This flag is valid only when ‘BTREE_EPHEMERAL is also set, and only on in-memory B-trees.

BTREE_KV_NONALIGNED

The keys and values in the B-tree arenʼt required to be aligned to eight-byte boundaries.

#define BTREE_KV_NONALIGNED 0x00000040

Aligning to eight-byte boundaries avoids unaligned reads on 64-bit platforms, which improves performance, but
wastes space on disk for structures whose size isnʼt a multiple of eight bytes.

B-Tree Table of Contents Constants

Constants used in managing the size of the table of contents in a B-tree node.

#define BTREE_TOC_ENTRY_INCREMENT 8
#define BTREE_TOC_ENTRY_MAX_UNUSED (2 * BTREE_TOC_ENTRY_INCREMENT)

These values are used by Appleʼs implementation; other implementations can choose different values. If you donʼt
use these values, profile your implementation to determine the performance impact of your chosen values.

BTREE_TOC_ENTRY_INCREMENT

The number of entries that are added or removed when changing the size of the table of contents.

#define BTREE_TOC_ENTRY_INCREMENT 8

BTREE_TOC_ENTRY_MAX_UNUSED

The maximum allowed number of unused entries in the table of contents.

#define BTREE_TOC_ENTRY_MAX_UNUSED (2 * BTREE_TOC_ENTRY_INCREMENT)

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

113
B-Tree
B-Tree Node Flags

B-Tree Node Flags

The flags used with a B-tree node.

#define BTNODE_ROOT 0x0001

#define BTNODE_LEAF 0x0002

#define BTNODE_FIXED_KV_SIZE 0x0004

#define BTNODE_CHECK_KOFF_INVAL 0x8000

BTNODE_ROOT

The B-tree node is a root node.

#define BTNODE_ROOT 0x0001

If this flag is set, the nodeʼs object type is OBJECT_TYPE_BTREE. If this is the treeʼs only node, both BTNODE_ROOT
and BTNODE_LEAF are set. Otherwise, the BTNODE_LEAF flag must not be set.

BTNODE_LEAF

The B-tree node is a leaf node.

#define BTNODE_LEAF 0x0002

If this is the treeʼs only node, the node objectʼs type is OBJECT_TYPE_BTREE, and both BTNODE_ROOT and
BTNODE_LEAF are set. Otherwise, the nodeʼs object type is OBJECT_TYPE_BTREE_NODE, and the BTNODE_ROOT
flag must not be set.

BTNODE_FIXED_KV_SIZE

The B-tree node has keys and values of a fixed size, and the table of contents omits their lengths.

#define BTNODE_FIXED_KV_SIZE 0x0004

If the keys and values both have a fixed size, this flag must be set.

Within the same B-tree, itʼs valid to have a mix of nodes that have this flag set and nodes that donʼt. For example,
consider a B-tree with fixed-sized keys and variable-sized values. Leaf nodes in that tree donʼt have this flag set
because of the variable-sized values. However, nonleaf nodes in in the same tree do have this flag set. The values
stored in nonleaf nodes are object identifiers, which are fixed-sized values; therefore, this flag can be applied to
nonleaf nodes of any tree with fixed-size keys.

BTNODE_CHECK_KOFF_INVAL

The B-tree node is in a transient state.

#define BTNODE_CHECK_KOFF_INVAL 0x8000

Objects with this flag never appear on disk. If you find an object of this type in production, file a bug against the Apple
File System implementation.

This flag isnʼt reserved by Apple; non-Apple implementations of Apple File System can set it on B-tree nodes in mem-
ory.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

114
B-Tree
B-Tree Node Constants

B-Tree Node Constants

Constants used to determine the size of a B-tree node.

#define BTREE_NODE_SIZE_DEFAULT 4096

#define BTREE_NODE_MIN_ENTRY_COUNT 4

A node is almost always one logical block in size. Smaller nodes waste space, and larger nodes can experience allo-
cation issues when space is fragmented. For example, a five-block node requires five adjacent blocks to all be free,
but on a fragmented disk such a large free space might not exist.

BTREE_NODE_SIZE_DEFAULT

The default size, in bytes, of a B-tree node.

#define BTREE_NODE_SIZE_DEFAULT 4096

BTREE_NODE_MIN_ENTRY_COUNT

The minimum number of entries that must be able to fit in a nonleaf B-tree node.

#define BTREE_NODE_MIN_ENTRY_COUNT 4

To satisfy this requirement, reduce the size of the keys stored in the node. The maximum key size is computed as
follows:

uint32_t btree_key_max_size(uint32_t nodesize) {

uint32_t dataspace, esize, count, kvspace;

dataspace = nodesize - offsetof(btree_node_phys_t, btn_data)

- sizeof(btree_info_t);
esize = sizeof(kvloc_t);
count = BTREE_TOC_ENTRY_INCREMENT;
kvspace = dataspace - (count * esize);
return ((kvspace / BTREE_NODE_MIN_ENTRY_COUNT) - sizeof(oid_t));
}

Note

This requirement comes from logic in Appleʼs implementation that performs proactive splitting of B-tree nodes.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

115
Space Manager
The space manager allocates and frees blocks where objects and file data can be stored. Thereʼs exactly one instance
of this structure in a container.

chunk_info_t
No overview available.

struct chunk_info {
uint64_t ci_xid;
uint64_t ci_addr;
uint32_t ci_block_count;
uint32_t ci_free_count;
paddr_t ci_bitmap_addr;
};
typedef struct chunk_info chunk_info_t;

chunk_info_block
A block that contains an array of chunk-info structures.

struct chunk_info_block {
obj_phys_t cib_o;
uint32_t cib_index;
uint32_t cib_chunk_info_count;
chunk_info_t cib_chunk_info[];
};
typedef struct chunk_info_block chunk_info_block_t;

No overview available.

cib_addr_block
A block that contains an array of chunk-info block addresses.

struct cib_addr_block {
obj_phys_t cab_o;
uint32_t cab_index;
uint32_t cab_cib_count;
paddr_t cab_cib_addr[];
};
typedef struct cib_addr_block cib_addr_block_t;

No overview available.

spaceman_free_queue_key_t
No overview available.

116
Space Manager
spaceman_free_queue_t

struct spaceman_free_queue_key {
xid_t sfqk_xid;
paddr_t sfqk_paddr;
};
typedef struct spaceman_free_queue_key spaceman_free_queue_key_t;

spaceman_free_queue_t
No overview available.

struct spaceman_free_queue {
uint64_t sfq_count;
oid_t sfq_tree_oid;
xid_t sfq_oldest_xid;
uint16_t sfq_tree_node_limit;
uint16_t sfq_pad16;
uint32_t sfq_pad32;
uint64_t sfq_reserved;
};
typedef struct spaceman_free_queue spaceman_free_queue_t;

spaceman_device_t
No overview available.

struct spaceman_device {
uint64_t sm_block_count;
uint64_t sm_chunk_count;
uint32_t sm_cib_count;
uint32_t sm_cab_count;
uint64_t sm_free_count;
uint32_t sm_addr_offset;
uint32_t sm_reserved;
uint64_t sm_reserved2;
};
typedef struct spaceman_device spaceman_device_t;

spaceman_phys_t
No overview available.

struct spaceman_phys {
obj_phys_t sm_o;
uint32_t sm_block_size;
uint32_t sm_blocks_per_chunk;
uint32_t sm_chunks_per_cib;
uint32_t sm_cibs_per_cab;
spaceman_device_t sm_dev[SD_COUNT];
uint32_t sm_flags;
uint32_t sm_ip_bm_tx_multiplier;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

117
Space Manager
sfq

uint64_t sm_ip_block_count;
uint32_t sm_ip_bm_size_in_blocks;
uint32_t sm_ip_bm_block_count;
paddr_t sm_ip_bm_base;
paddr_t sm_ip_base;
uint64_t sm_fs_reserve_block_count;
uint64_t sm_fs_reserve_alloc_count;
spaceman_free_queue_t sm_fq[SFQ_COUNT];
uint16_t sm_ip_bm_free_head;
uint16_t sm_ip_bm_free_tail;
uint32_t sm_ip_bm_xid_offset;
uint32_t sm_ip_bitmap_offset;
uint32_t sm_ip_bm_free_next_offset;
uint32_t sm_version;
uint32_t sm_struct_size;
spaceman_datazone_info_phys_t sm_datazone;
};
typedef struct spaceman_phys spaceman_phys_t;

#define SM_FLAG_VERSIONED 0x00000001

sfq
No overview available.

enum sfq {
SFQ_IP = 0,
SFQ_MAIN = 1,
SFQ_TIER2 = 2,
SFQ_COUNT = 3
};

smdev
No overview available.

enum smdev {
SD_MAIN = 0,
SD_TIER2 = 1,
SD_COUNT = 2
};

Chunk Info Block Constants

No overview available.

#define CI_COUNT_MASK 0x000fffff

#define CI_COUNT_RESERVED_MASK 0xfff00000

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

118
Space Manager
Internal-Pool Bitmap

Internal-Pool Bitmap
No overview available.

#define SPACEMAN_IP_BM_TX_MULTIPLIER 16
#define SPACEMAN_IP_BM_INDEX_INVALID 0xffff
#define SPACEMAN_IP_BM_BLOCK_COUNT_MAX 0xfffe

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

119
Reaper
The reaper is a mechanism that allows large objects to be deleted over a period spanning multiple transactions. Thereʼs
exactly one instance of this structure in a container.

nx_reaper_phys_t
No overview available.

struct nx_reaper_phys {
obj_phys_t nr_o;
uint64_t nr_next_reap_id;
uint64_t nr_completed_id;
oid_t nr_head;
oid_t nr_tail;
uint32_t nr_flags;
uint32_t nr_rlcount;
uint32_t nr_type;
uint32_t nr_size;
oid_t nr_fs_oid;
oid_t nr_oid;
xid_t nr_xid;
uint32_t nr_nrle_flags;
uint32_t nr_state_buffer_size;
uint8_t nr_state_buffer[];
};
typedef struct nx_reaper_phys nx_reaper_phys_t;

nx_reap_list_phys_t
No overview available.

struct nx_reap_list_phys {
obj_phys_t nrl_o;
oid_t nrl_next;
uint32_t nrl_flags;
uint32_t nrl_max;
uint32_t nrl_count;
uint32_t nrl_first;
uint32_t nrl_last;
uint32_t nrl_free;
nx_reap_list_entry_t nrl_entries[];
};
typedef struct nx_reap_list_phys nx_reap_list_phys_t;

nx_reap_list_entry_t
No overview available.

120
Reaper
Volume Reaper States

struct nx_reap_list_entry {
uint32_t nrle_next;
uint32_t nrle_flags;
uint32_t nrle_type;
uint32_t nrle_size;
oid_t nrle_fs_oid;
oid_t nrle_oid;
xid_t nrle_xid;
};
typedef struct nx_reap_list_entry nx_reap_list_entry_t;

Volume Reaper States

No overview available.

enum {
APFS_REAP_PHASE_START = 0,
APFS_REAP_PHASE_SNAPSHOTS = 1,
APFS_REAP_PHASE_ACTIVE_FS = 2,
APFS_REAP_PHASE_DESTROY_OMAP = 3,
APFS_REAP_PHASE_DONE = 4
};

Reaper Flags
The flags used for general information about a reaper.

#define NR_BHM_FLAG 0x00000001

#define NR_CONTINUE 0x00000002

These flags are used by the nr_flags field of nx_reaper_phys_t.

NR_BHM_FLAG

Reserved.

#define NR_BHM_FLAG 0x00000001

This flag must always be set.

NR_CONTINUE

The current object is being reaped.

#define NR_CONTINUE 0x00000002

Reaper List Entry Flags

No overview available.

#define NRLE_VALID 0x00000001

#define NRLE_REAP_ID_RECORD 0x00000002

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

121
Reaper
Reaper List Flags

#define NRLE_CALL 0x00000004

#define NRLE_COMPLETION 0x00000008
#define NRLE_CLEANUP 0x00000010

Reaper List Flags

No overview available.

#define NRL_INDEX_INVALID 0xffffffff

omap_reap_state_t
State used when reaping an object map.

struct omap_reap_state {
uint32_t omr_phase;
omap_key_t omr_ok;
};
typedef struct omap_reap_state omap_reap_state_t;

The reaper uses the state thatʼs stored in this structure to resume after an interruption.

omr_phase

The current reaping phase.

uint32_t omr_phase;

For the values used in this field, see Object Map Reaper Phases.

omr_ok

The key of the most recently freed entry in the object map.

omap_key_t omr_ok;

This field allows the reaper to resume after the last entry it processed.

omap_cleanup_state_t
State used when reaping to clean up deleted snapshots.

struct omap_cleanup_state {
uint32_t omc_cleaning;
uint32_t omc_omsflags;
xid_t omc_sxidprev;
xid_t omc_sxidstart;
xid_t omc_sxidend;
xid_t omc_sxidnext;
omap_key_t omc_curkey;
};
typedef struct omap_cleanup_state omap_cleanup_state_t;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

122
Reaper
apfs_reap_state_t

omc_cleaning

A flag that indicates whether the structure has valid data in it.

uint32_t omc_cleaning;

If the value of this field is 0, the structure has been allocated and zeroed, but doesnʼt yet contain valid data. If the
value is anything other than 0, the structure is valid.

omc_omsflags

The flags for the snapshot being deleted.

uint32_t omc_omsflags;

The value for this field is the same as the value of the snapshotʼs omap_snapshot_t.oms_flags field.

omc_sxidprev

The transaction identifier of the snapshot prior to the snapshots being deleted.

xid_t omc_sxidprev;

omc_sxidstart

The transaction identifier of the first snapshot being deleted.

xid_t omc_sxidstart;

omc_sxidend

The transaction identifier of the last snapshot being deleted.

xid_t omc_sxidend;

omc_sxidnext

The transaction identifier of the snapshot after the snapshots being deleted.

xid_t omc_sxidnext;

omc_curkey

The key of the next object mapping to consider for deletion.

omap_key_t omc_curkey;

apfs_reap_state_t
No overview available.

struct apfs_reap_state {
uint64_t last_pbn;
xid_t cur_snap_xid;
uint32_t phase;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

123
Reaper
apfs_reap_state_t

} __attribute__((packed));
typedef struct apfs_reap_state apfs_reap_state_t;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

124
Encryption
No overview available.

j_crypto_key_t
The key half of an encryption state record.

typedef struct j_crypto_key_t {

j_key_t hdr;
} __attribute__((packed)) j_dir_stats_key_t;

hdr

The recordʼs header.

j_key_t hdr;

The object identifier in the header is the file-system objectʼs identifier. The type in the header is always
APFS_TYPE_CRYPTO_STATE.

j_crypto_val_t
The value half of an encryption state record.

typedef struct j_crypto_val {

uint32_t refcnt;
wrapped_crypto_state_t state;
} __attribute__((aligned(4),packed)) j_crypto_val_t;

refcnt

The reference count.

int32_t refcnt;

The encryption state record can be garbage collected when its reference count reaches zero.

state

The encryption state information.

wrapped_crypto_state_t state;

crypto_flags_t
No overview available.

typedef uint32_t crypto_flags_t;

125
Encryption
cp_key_class_t

cp_key_class_t
No overview available.

typedef uint32_t cp_key_class_t;

cp_key_os_version_t
No overview available.

typedef uint32_t cp_key_os_version_t;

cp_key_revision_t
No overview available.

typedef uint16_t cp_key_revision_t;

cpx_t
No overerview available.

typedef struct cpx* cpx_t;

wrapped_crypto_state_t
No overview available.

typedef struct wrapped_crypto_state {

uint16_t major_version;
uint16_t minor_version;
crypto_flags_t cpflags;
cp_key_class_t persistent_class;
cp_key_os_version_t key_os_version;
cp_key_revision_t key_revision;
uint16_t key_len;
uint8_t persistent_key[0];
} __attribute__((aligned(2), packed)) wrapped_crypto_state_t;

This structure is used inside of j_crypto_val_t.

wrapped_meta_crypto_state_t
No overview available.

typedef struct wrapped_meta_crypto_state {

uint16_t major_version;
uint16_t minor_version;
crypto_flags_t cpflags;
cp_key_class_t persistent_class;
cp_key_os_version_t key_os_version;
cp_key_revision_t key_revision;

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

126
Encryption
Protection Class

uint16_t unused;
} __attribute__((aligned(2), packed)) wrapped_meta_crypto_state_t;

Protection Class
No overview available.

#define CP_EFFECTIVE_CLASSMASK 0x0000001f

#define CP_CLASS(x) ((cp_key_class_t)(CP_EFFECTIVE_CLASSMASK & (x)))

#define CP_IV_KEYSIZE 16
#define CP_MAX_KEYSIZE 32
#define CP_MAX_CACHEBUFLEN 64

#define CP_INITIAL_WRAPPEDKEYSIZE 40
#define CP_V2_WRAPPEDKEYSIZE 40
#define CP_V4_RESERVEDBYTES 16

#define CP_MAX_WRAPPEDKEYSIZE 128

#define CP_CURRENT_VERS CP_VERS_5

#define CP_VERS_5 5
#define CP_VERS_4 4
#define CP_MINOR_VERS 0

#define PROTECTION_CLASS_DIR_NONE 0
#define PROTECTION_CLASS_A 1
#define PROTECTION_CLASS_B 2
#define PROTECTION_CLASS_C 3
#define PROTECTION_CLASS_D 4
#define PROTECTION_CLASS_F 6

Encryption Identifiers
Identifiers used to indicate special encryption information.

#define CRYPTO_SW_ID 4
#define CRYPTO_VOLKEY_ID 5

These identifiers are used instead of the identifier of an encryption record (j_crypto_val_t).

CRYPTO_SW_ID

The volume-level encryption is performed in software.

#define CRYPTO_SW_ID 4

See also NX_CRYPTO_SW.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

127
Encryption
Encryption Identifiers

CRYPTO_VOLKEY_ID

The volume-level encryption is performed in hardware.

#define CRYPTO_VOLKEY_ID 5

This encryption identifier is valid only on macOS.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

128
Keybag
No overview available.

kb_locker_t
No overview available.

typedef struct kb_locker {

uint16_t kl_version;
uint16_t kl_nkeys;
uint32_t kl_nbytes;
uint8_t _padding_[8];
keybag_entry_t kl_entries[];
} kb_locker_t;

keybag_entry_t
No overview available.

typedef struct keybag_entry {

uuid_t ke_uuid;
uint16_t ke_tag;
uint16_t ke_keylen;
uint8_t _padding_[4];
uint8_t ke_keydata[];
} keybag_entry_t;

media_keybag_t
No overview available.

typedef struct media_keybag {

mk_obj_t mk_obj;
kb_locker_t mk_locker;
} media_keybag_t;

mk_obj_t
No overview available.

typedef struct mk_obj {

uint8_t o_cksum[MAX_CKSUM_SIZE];
oid_t o_oid;
xid_t o_xid;
uint32_t o_type;
uint32_t o_subtype;
} mk_obj_t;

129
Keybag
Keybag Constants

Keybag Constants
No overview available.

#define APFS_KEYBAG_VERSION 2

#define APFS_KEYBAG_OBJ 'keys'

#define APFS_VOL_KEYBAG_OBJ 'recs'

#define APFS_VOL_KEYBAG_ENTRY_MAX_SIZE 512

#define APFS_ENTRY_ALIGN 16

Keybag Tags
No overview available.

enum {
KB_TAG_UNKNOWN = 0,
KB_TAG_WRAPPING_KEY,
KB_TAG_VOLUME_KEY,
KB_TAG_VOLUME_UNLOCK_RECORDS,
KB_TAG_VOLUME_PASSPHRASE_HINT,
KB_TAG_USER_PAYLOAD = 0xF8
};

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

130
Encryption Rolling
No overview available.

er_state_phys_t
No overview available.

struct er_state_phys {
obj_phys_t ersb_o;
uint32_t ersb_magic;
uint32_t ersb_version;
uint64_t ersb_flags;
uint64_t ersb_snap_xid;
uint64_t ersb_current_fext_obj_id;
uint64_t ersb_file_offset;
uint64_t ersb_fext_pbn;
uint64_t ersb_paddr;
uint64_t ersb_progress;
uint64_t ersb_total_blk_to_encrypt;
uint64_t ersb_blockmap_oid;
uint32_t ersb_checksum_count;
uint32_t ersb_reserved;
uint64_t ersb_fext_cid;
uint8_t ersb_checksum[0];
};
typedef struct er_state_phys er_state_phys_t;

er_phase_t
No overview available.

enum er_phase_enum {
ER_PHASE_OMAP_ROLL = 1,
ER_PHASE_DATA_ROLL = 2,
ER_PHASE_SNAP_ROLL = 3,
};
typedef struct er_phase_enum er_phase_t;

gbitmap_block_phys_t
No overview available.

struct gbitmap_block_phys {
obj_phys_t bmb_o;
uint64_t bmb_field[0];
};
typedef struct gbitmap_block_phys gbitmap_block_phys_t;

131
Encryption Rolling
gbitmap_phys_t

gbitmap_phys_t
No overview available.

struct gbitmap_phys {
obj_phys_t bm_o;
oid_t bm_tree_oid;
uint64_t bm_bit_count;
uint64_t bm_flags;
};
typedef struct gbitmap_phys gbitmap_phys_t;

Encryption-Rolling Checksum Block Sizes

No overview available.

enum {
ER_512B_BLOCKSIZE = 0,
ER_2KiB_BLOCKSIZE = 1,
ER_4KiB_BLOCKSIZE = 2,
ER_8KiB_BLOCKSIZE = 3,
ER_16KiB_BLOCKSIZE = 4,
ER_32KiB_BLOCKSIZE = 5,
ER_64KiB_BLOCKSIZE = 6,
};

Encryption Rolling Flags

No overview available.

#define ERSB_FLAG_ENCRYPTING 0x00000001

#define ERSB_FLAG_DECRYPTING 0x00000002
#define ERSB_FLAG_KEYROLLING 0x00000004
#define ERSB_FLAG_PAUSED 0x00000008
#define ERSB_FLAG_FAILED 0x00000010
#define ERSB_FLAG_CID_IS_TWEAK 0x00000020
#define ERSB_FLAG_FREE_1 0x00000040
#define ERSB_FLAG_FREE_2 0x00000080

#define ERSB_FLAG_CM_BLOCK_SIZE_MASK 0x00000F00

#define ERSB_FLAG_CM_BLOCK_SIZE_SHIFT 8

#define ERSB_FLAG_ER_PHASE_MASK 0x00003000

#define ERSB_FLAG_ER_PHASE_SHIFT 12
#define ERSB_FLAG_FROM_ONEKEY 0x00004000

Encryption-Rolling Constants
No overview available.

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

132
Encryption Rolling
Encryption-Rolling Constants

#define ER_CHECKSUM_LENGTH 8
#define ER_MAGIC 'FLAB'
#define ER_VERSION 1

#define ER_MAX_CHECKSUM_COUNT_SHIFT 16
#define ER_CUR_CHECKSUM_COUNT_MASK 0x0000FFFF

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

133
Fusion
No overview available.

fusion_wbc_phys_t
No overview available.

typedef struct {
obj_phys_t fwp_objHdr;
uint64_t fwp_version;
oid_t fwp_listHeadOid;
oid_t fwp_listTailOid;
uint64_t fwp_stableHeadOffset;
uint64_t fwp_stableTailOffset;
uint32_t fwp_listBlocksCount;
uint32_t fwp_reserved;
uint64_t fwp_usedByRC;
prange_t fwp_rcStash;
} fusion_wbc_phys_t;

fusion_wbc_list_entry_t
No overview available.

typedef struct {
paddr_t fwle_wbcLba;
paddr_t fwle_targetLba;
uint64_t fwle_length;
} fusion_wbc_list_entry_t;

fusion_wbc_list_phys_t
No overview available.

typedef struct {
obj_phys_t fwlp_objHdr;
uint64_t fwlp_version;
uint64_t fwlp_tailOffset;
uint32_t fwlp_indexBegin;
uint32_t fwlp_indexEnd;
uint32_t fwlp_indexMax;
uint32_t fwlp_reserved;
fusion_wbc_list_entry_t fwlp_listEntries[];
} fusion_wbc_list_phys_t;

This mapping keeps track of data from the hard drive thatʼs cached on the solid-state drive. For read caching, the
same data is stored on both the hard drive and the solid-state drive. For write caching, the data is stored on the solid-

134
Fusion
Address Markers

state drive, but space for the data has been allocated on the hard drive, and the data will eventually be copied to that
space.

Address Markers
No overview available.

#define FUSION_TIER2_DEVICE_BYTE_ADDR 0x4000000000000000ULL

#define FUSION_TIER2_DEVICE_BLOCK_ADDR(_blksize) \
(FUSION_TIER2_DEVICE_BYTE_ADDR >> __builtin_ctzl(_blksize))

#define FUSION_BLKNO(_fusion_tier2, _blkno, _blksize) \

((_fusion_tier2) \
? (FUSION_TIER2_DEVICE_BLOCK_ADDR(_blksize) | (_blkno)) \
: (_blkno))

fusion_mt_key_t
No overview available.

typedef paddr_t fusion_mt_key_t;

fusion_mt_val_t
No overview available.

typedef struct {
paddr_t fmv_lba;
uint32_t fmv_length;
uint32_t fmv_flags;
} fusion_mt_val_t;

Fusion Middle-Tree Flags

No overview available.

#define FUSION_MT_DIRTY (1 << 0)

#define FUSION_MT_TENANT (1 << 1)

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

135
Symbol Index
APFS_FEATURE_DEFRAG 58 APFS_SUPPORTED_INCOMPAT_MASK 60

APFS_FEATURE_DEFRAG_PRERELEASE 58 APFS_SUPPORTED_ROCOMPAT_MASK 59

APFS_FEATURE_HARDLINK_MAP_RECORDS 58 APFS_TYPE_ANY 73

APFS_FS_CRYPTOFLAGS 57 APFS_TYPE_CRYPTO_STATE 74

APFS_FS_EFFACEABLE 57 APFS_TYPE_DIR_REC 75

APFS_FS_FLAGS_VALID_MASK 57 APFS_TYPE_DIR_STATS 75

APFS_FS_ONEKEY 57 APFS_TYPE_DSTREAM_ID 74

APFS_FS_RUN_SPILLOVER_CLEANER 57 APFS_TYPE_EXTENT 74

APFS_FS_SPILLEDOVER 57 APFS_TYPE_FILE_EXTENT 74

APFS_FS_UNENCRYPTED 56 APFS_TYPE_INODE 74

APFS_GPT_PARTITION_UUID 23 APFS_TYPE_INVALID 75

APFS_INCOMPAT_CASE_INSENSITIVE 59 APFS_TYPE_MAX 75

APFS_INCOMPAT_DATALESS_SNAPS 59 APFS_TYPE_MAX_VALID 75

APFS_INCOMPAT_ENC_ROLLED 59 APFS_TYPE_SIBLING_LINK 74

APFS_INCOMPAT_NORMALIZATION_INSENSITIVE 59 APFS_TYPE_SIBLING_MAP 75

APFS_INODE_PINNED_MASK 81 APFS_TYPE_SNAP_METADATA 73

APFS_KIND_ANY 76 APFS_TYPE_SNAP_NAME 75

APFS_KIND_DEAD 76 APFS_TYPE_XATTR 74

APFS_KIND_INVALID 77 APFS_VALID_INTERNAL_INODE_FLAGS 81

APFS_KIND_NEW 76 APFS_VOLNAME_LEN 55

APFS_KIND_UPDATE 76 BTNODE_CHECK_KOFF_INVAL 114

APFS_KIND_UPDATE_REFCNT 76 BTNODE_FIXED_KV_SIZE 114

APFS_MAGIC 55 BTNODE_LEAF 114

APFS_MAX_HIST 55 BTNODE_ROOT 114

apfs_modified_by_t 55 BTOFF_INVALID 110

apfs_reap_state_t 123 BTREE_ALLOW_GHOSTS 112

APFS_RESERVED_4 57 BTREE_EPHEMERAL 112

apfs_superblock_t 48 btree_info_fixed_t 108

APFS_SUPPORTED_FEATURES_MASK 58 btree_info_t 109

136
Symbol Index

BTREE_KV_NONALIGNED 113 DT_SOCK 87

BTREE_NODE_MIN_ENTRY_COUNT 115 DT_UNKNOWN 86

btree_node_phys_t 106 DT_WHT 87

BTREE_NODE_SIZE_DEFAULT 115 er_phase_t 131

BTREE_NONPERSISTENT 113 er_state_phys_t 131

BTREE_PHYSICAL 113 evict_mapping_val_t 40

BTREE_SEQUENTIAL_INSERT 112 fusion_mt_key_t 135

BTREE_TOC_ENTRY_INCREMENT 113 fusion_mt_val_t 135

BTREE_TOC_ENTRY_MAX_UNUSED 113 fusion_wbc_list_entry_t 134

BTREE_UINT64_KEYS 112 fusion_wbc_list_phys_t 134

CHECKPOINT_MAP_LAST 40 fusion_wbc_phys_t 134

checkpoint_map_phys_t 39 gbitmap_block_phys_t 131

checkpoint_mapping_t 37 gbitmap_phys_t 132

chunk_info_block 116 INO_EXT_TYPE_DELTA_TREE_OID 96

chunk_info_t 116 INO_EXT_TYPE_DIR_STATS_KEY 97

cib_addr_block 116 INO_EXT_TYPE_DOCUMENT_ID 96

cp_key_class_t 126 INO_EXT_TYPE_DSTREAM 97

cp_key_os_version_t 126 INO_EXT_TYPE_FINDER_INFO 97

cp_key_revision_t 126 INO_EXT_TYPE_FS_UUID 98

cpx_t 126 INO_EXT_TYPE_NAME 97

crypto_flags_t 125 INO_EXT_TYPE_PREV_FSIZE 97

CRYPTO_SW_ID 127 INO_EXT_TYPE_RDEV 98

CRYPTO_VOLKEY_ID 128 INO_EXT_TYPE_RESERVED_12 98

dir_rec_flags 82 INO_EXT_TYPE_RESERVED_6 97

DREC_EXT_TYPE_SIBLING_ID 96 INO_EXT_TYPE_RESERVED_9 97

DREC_TYPE_MASK 82 INO_EXT_TYPE_SNAP_XID 96

DT_BLK 87 INO_EXT_TYPE_SPARSE_BYTES 98

DT_CHR 87 INODE_ACTIVE_FILE_TRIMMED 80

DT_DIR 87 INODE_ALLOCATION_SPILLEDOVER 80

DT_FIFO 87 INODE_BEING_TRUNCATED 79

DT_LNK 87 INODE_CLONED_INTERNAL_FLAGS 81

DT_REG 87 INODE_DIR_STATS_ORIGIN 78

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

137
Symbol Index

INODE_FLAG_UNUSED 79 j_file_extent_val_t 90

INODE_HAS_FINDER_INFO 79 j_inode_flags 77

INODE_HAS_RSRC_FORK 80 j_inode_key_t 63

INODE_HAS_SECURITY_EA 79 j_inode_val_t 63

INODE_INHERITED_INTERNAL_FLAGS 80 j_key_t 62

INODE_IS_APFS_PRIVATE 78 j_obj_kinds 75

INODE_IS_SPARSE 79 j_obj_types 73

INODE_MAINTAIN_DIR_STATS 78 j_phys_ext_key_t 88

INODE_NO_RSRC_FORK 80 j_phys_ext_val_t 88

INODE_PINNED_TO_MAIN 80 j_sibling_key_t 100

INODE_PINNED_TO_TIER2 80 j_sibling_map_key_t 101

INODE_PROT_CLASS_EXPLICIT 78 j_sibling_map_val_t 101

INODE_WAS_CLONED 78 j_sibling_val_t 100

INODE_WAS_EVER_CLONED 79 j_snap_metadata_key_t 102

INVALID_INO_NUM 83 j_snap_metadata_val_t 102

j_crypto_key_t 125 j_snap_name_key_t 103

j_crypto_val_t 125 j_snap_name_val_t 104

j_dir_stats_key_t 70 j_xattr_dstream_t 91

j_dir_stats_val_t 70 j_xattr_flags 81

J_DREC_HASH_MASK 69 j_xattr_key_t 71

J_DREC_HASH_SHIFT 69 j_xattr_val_t 71

j_drec_hashed_key_t 68 kb_locker_t 129

j_drec_key_t 67 keybag_entry_t 129

J_DREC_LEN_MASK 68 kvloc_t 111

j_drec_val_t 69 kvoff_t 111

j_dstream_id_key_t 91 MAX_CKSUM_SIZE 10

j_dstream_id_val_t 91 media_keybag_t 129

j_dstream_t 92 MIN_DOC_ID 84

J_FILE_EXTENT_FLAG_MASK 91 MIN_USER_INO_NUM 84

J_FILE_EXTENT_FLAG_SHIFT 91 mk_obj_t 129

j_file_extent_key_t 89 nloc_t 110

J_FILE_EXTENT_LEN_MASK 90 NR_BHM_FLAG 121

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

138
Symbol Index

NR_CONTINUE 121 NX_SUPPORTED_ROCOMPAT_MASK 35

NX_CNTR_OBJ_CKSUM_FAIL 37 NX_TX_MIN_CHECKPOINT_COUNT 33

NX_CNTR_OBJ_CKSUM_SET 37 OBJ_ENCRYPTED 18

nx_counter_id_t 37 OBJ_EPHEMERAL 18

NX_CRYPTO_SW 34 OBJ_ID_MASK 63

NX_DEFAULT_BLOCK_SIZE 36 OBJ_NOHEADER 18

NX_EFI_JUMPSTART_MAGIC 23 OBJ_NONPERSISTENT 18

nx_efi_jumpstart_t 22 obj_phys_t 9

NX_EFI_JUMPSTART_VERSION 23 OBJ_PHYSICAL 18

NX_EPH_INFO_COUNT 33 OBJ_STORAGETYPE_MASK 13

NX_EPH_INFO_VERSION_1 33 OBJ_TYPE_MASK 63

NX_EPH_MIN_BLOCK_COUNT 33 OBJ_TYPE_SHIFT 63

NX_FEATURE_DEFRAG 34 OBJ_VIRTUAL 18

NX_FEATURE_LCFD 35 OBJECT_TYPE_BLOCKREFTREE 15

NX_INCOMPAT_FUSION 36 OBJECT_TYPE_BTREE 14

NX_INCOMPAT_VERSION1 35 OBJECT_TYPE_BTREE_NODE 14

NX_INCOMPAT_VERSION2 36 OBJECT_TYPE_CHECKPOINT_MAP 15

NX_MAGIC 33 OBJECT_TYPE_EFI_JUMPSTART 16

NX_MAX_FILE_SYSTEM_EPH_STRUCTS 33 OBJECT_TYPE_ER_STATE 17

NX_MAX_FILE_SYSTEMS 33 OBJECT_TYPE_EXTENT_LIST_TREE 15

NX_MAXIMUM_BLOCK_SIZE 36 OBJECT_TYPE_FLAGS_DEFINED_MASK 13

NX_MINIMUM_BLOCK_SIZE 36 OBJECT_TYPE_FLAGS_MASK 12

NX_MINIMUM_CONTAINER_SIZE 37 OBJECT_TYPE_FS 15

NX_NUM_COUNTERS 37 OBJECT_TYPE_FSTREE 15

nx_reap_list_entry_t 120 OBJECT_TYPE_FUSION_MIDDLE_TREE 16

nx_reap_list_phys_t 120 OBJECT_TYPE_GBITMAP 17

nx_reaper_phys_t 120 OBJECT_TYPE_GBITMAP_BLOCK 17

NX_RESERVED_1 34 OBJECT_TYPE_GBITMAP_TREE 17

NX_RESERVED_2 34 OBJECT_TYPE_INVALID 17

nx_superblock_t 25 OBJECT_TYPE_MASK 12

NX_SUPPORTED_FEATURES_MASK 35 OBJECT_TYPE_NX_FUSION_WBC 16

NX_SUPPORTED_INCOMPAT_MASK 36 OBJECT_TYPE_NX_FUSION_WBC_LIST 16

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

139
Symbol Index

OBJECT_TYPE_NX_REAP_LIST 16 OMAP_VAL_ENCRYPTED 45

OBJECT_TYPE_NX_REAPER 16 OMAP_VAL_NOHEADER 45

OBJECT_TYPE_NX_SUPERBLOCK 14 OMAP_VAL_SAVED 45

OBJECT_TYPE_OMAP 15 omap_val_t 43

OBJECT_TYPE_OMAP_SNAPSHOT 16 paddr_t 8

OBJECT_TYPE_SNAPMETATREE 16 PEXT_KIND_MASK 89

OBJECT_TYPE_SPACEMAN 14 PEXT_KIND_SHIFT 89

OBJECT_TYPE_SPACEMAN_BITMAP 14 PEXT_LEN_MASK 89

OBJECT_TYPE_SPACEMAN_CAB 14 prange_t 8

OBJECT_TYPE_SPACEMAN_CIB 14 PRIV_DIR_INO_NUM 83

OBJECT_TYPE_SPACEMAN_FREE_QUEUE 14 RESERVED_10 82

OBJECT_TYPE_TEST 17 ROOT_DIR_INO_NUM 83

OID_INVALID 12 ROOT_DIR_PARENT 83

OID_NX_SUPERBLOCK 12 S_IFBLK 86

OID_RESERVED_COUNT 12 S_IFCHR 85

omap_cleanup_state_t 122 S_IFDIR 85

OMAP_CRYPTO_GENERATION 47 S_IFIFO 85

OMAP_DECRYPTING 46 S_IFLNK 86

OMAP_ENCRYPTING 46 S_IFMT 85

omap_key_t 43 S_IFREG 86

OMAP_KEYROLLING 47 S_IFSOCK 86

OMAP_MANUALLY_MANAGED 46 S_IFWHT 86

OMAP_MAX_SNAP_COUNT 47 sfq 118

omap_phys_t 41 smdev 118

OMAP_REAP_PHASE_MAP_TREE 47 SNAP_DIR_INO_NUM 83

OMAP_REAP_PHASE_SNAPSHOT_TREE 47 snap_meta_flags 104

omap_reap_state_t 122 spaceman_device_t 117

OMAP_SNAPSHOT_DELETED 46 spaceman_free_queue_key_t 116

OMAP_SNAPSHOT_REVERTED 46 spaceman_free_queue_t 117

omap_snapshot_t 44 spaceman_phys_t 117

OMAP_VAL_CRYPTO_GENERATION 45 SYMLINK_EA_NAME 84

OMAP_VAL_DELETED 45 uuid_t 8

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

140
Symbol Index

wrapped_crypto_state_t 126 XF_CHILDREN_INHERIT 99

wrapped_meta_crypto_state_t 126 XF_DATA_DEPENDENT 98

x_field_t 95
XF_DO_NOT_COPY 99
XATTR_DATA_EMBEDDED 82
XF_RESERVED_4 99
XATTR_DATA_STREAM 81
XF_RESERVED_40 99
XATTR_FILE_SYSTEM_OWNED 82
XF_RESERVED_80 99
XATTR_MAX_EMBEDDED_SIZE 84

XATTR_RESERVED_8 82 XF_SYSTEM_FIELD 99

xf_blob_t 94 XF_USER_FIELD 99

2018-09-17 | Copyright © 2018 Apple Inc. All Rights Reserved.

141
Revision History
2018-09-17

New document that describes the data structures used for read-only access to Apple File System on unencrypted,
non-Fusion storage.

142
Copyright and Notices

Apple Inc.
Copyright © 2018 Apple Inc.
All rights reserved.

No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, mechanical, electronic, photocopying,
recording, or otherwise, without prior written permission of Apple Inc., with the following exceptions: Any person is hereby authorized to store documentation
on a single computer or device for personal use only and to print copies of documentation for personal use provided that the documentation contains Appleʼs
copyright notice.

No licenses, express or implied, are granted with respect to any of the technology described in this document. Apple retains all intellectual property rights
associated with the technology described in this document. This document is intended to assist application developers to develop applications only for Apple-
branded products.

Apple Inc.
One Apple Park Way
Cupertino, CA 95014
USA
408-996-1010

Apple is a trademark of Apple Inc., registered in the U.S. and other countries.

APPLE MAKES NO WARRANTY OR REPRESENTATION, EITHER EXPRESS OR IMPLIED, WITH RESPECT TO THIS DOCUMENT, ITS QUALITY, ACCURACY,
MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE. AS A RESULT, THIS DOCUMENT IS PROVIDED ”AS IS,” AND YOU, THE READER, ARE
ASSUMING THE ENTIRE RISK AS TO ITS QUALITY AND ACCURACY.

IN NO EVENT WILL APPLE BE LIABLE FOR DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES RESULTING FROM ANY DEFECT,
ERROR OR INACCURACY IN THIS DOCUMENT, even if advised of the possibility of such damages.

Some jurisdictions do not allow the exclusion of implied warranties or liability, so the above exclusion may not apply to you.

143