XFS

Filesystems
Virtual Filesystems
VFS
Disk Filesystems
FAT 12/16/32, VFAT, ExFAT Ext 2/3/4 LEAN HPFS NTFS HFS HFS+ MFS ReiserFS FFS (Amiga) FFS (BSD)/UFS BeFS BFS XFS SFS ZDSFS ZFS USTAR
CD/DVD Filesystems
ISO 9660 Joliet UDF
Network Filesystems
NFS RFS AFS
Flash Filesystems
JFFS2 YAFFS

XFS is a high-performance journaling file system created to replace EFS on IRIX. It has since been ported to Linux and received additional development there.

Basic Specifications

XFS is a 64-bit file system. It supports a maximum file system size of 8 EiB, though many operating systems have implementations that limit this. IRIX is known to support up to 8 TiB volumes.^[1]

Directory Structure	B+ Trees
File Contents	B+ Trees
Max Volume Size	8 Terabytes (IRIX), 8 Exabytes (GNU/Linux)
Max File Size	4 Terabytes
Max Filename Length	255 Bytes

On-disk Structure

On disk Header Info

struct xfs_ondisk_hdr {
__be32 magic; /* magic number */
__be32 crc; /* CRC, not logged */
uuid_t uuid; /* filesystem identifier */
__be64 owner; /* parent object */
__be64 blkno; /* location on disk */
__be64 lsn; /* last modification in log, not logged */
};

• magic: This is the unique ID telling the OS that the filesystem is an XFS filesystem. The magic humber for XFS is 58 46 53 42 which in ASCII is "XFSB" standing for XFS Block.
• crc: This is a checksum for the header to ensure data integrity. It’s not logged, meaning it doesn’t get saved in the XFS transaction log (journal).
• uuid: This is the filesystem identifier. It’s a unique 128-bit identifier (UUID) for the XFS filesystem, used to distinguish between different XFS filesystems.
• owner: This represents the parent object of the filesystem (could be a device or logical unit), and it’s typically used in a system where objects can be nested. It helps identify who owns the filesystem.
• blkno: This is the block number or the location of the XFS filesystem's header on the disk. It specifies where this specific structure is physically located.

• lsn: The log sequence number (LSN) represents the last modification timestamp in the log. It indicates when the filesystem was last modified but is not logged itself, meaning it’s not part of the transaction log.

Allocation Group B+ Tree Blocks

struct xfs_btree_sblock {
__be32 bb_magic;
__be16 bb_level;
__be16 bb_numrecs;
__be32 bb_leftsib;
__be32 bb_rightsib;
/* version 5 extras start here */
__be64 bb_blkno;
__be64 bb_lsn;
uuid_t bb_uuid;
__be32 bb_owner;
__le32 bb_crc;
};

Short blocks use 32-bit pointers for the filesystem blocks.

• bb_magic: Specifies the magic number for the per-allocation group B+tree block.
• bb_level:The level of the tree in which this block is found. If this value is 0, this is a leaf block and contains records; otherwise, it is a node block and contains keys and pointers. Level values increase towards the root.
• bb_numrecs Number of records in this block.
• bb_leftsib allocation group block number of the left sibling of this B+tree node.
• bb_rightsib AG block number of the right sibling of this B+tree node.
• bb_blkno FS block number of this B+tree block.
• bb_lsn Log sequence number of the last write to this block.
• bb_uuid The UUID of this block, which must match either sb_uuid or sb_meta_uuid depending on which features are set.
• bb_owner The AG number that this B+tree block ought to be in.
• bb_crc Checksum of the B+tree block.

^[2]

struct xfs_btree_lblock {
__be32 bb_magic;
__be16 bb_level;
__be16 bb_numrecs;
__be64 bb_leftsib;
__be64 bb_rightsib;
/* version 5 additions start here */
__be64 bb_blkno;
__be64 bb_lsn;
uuid_t bb_uuid;
__be64 bb_owner;
__le32 bb_crc;
__be32 bb_pad;
};

Long format B+trees are given 64-bit filesystem block numbers instead of 32-bit allocation group block numbers.

• bb_magic Specifies the magic number for the B+tree block.
• bb_level The level of the tree in which this block is found. If this value is 0, this is a leaf block and contains records; otherwise, it is a node block and contains keys and pointers.
• bb_numrecs Number of records in this block.
• bb_leftsib Filesystem block number of the left sibling of this B+tree node.
• bb_rightsib Filesystem block number of the right sibling of this B+tree node.
• bb_blkno Filesystem block number of this B+tree block.
• bb_lsn Log sequence number of the last write to this block.
• bb_uuid The UUID of this block, which must match either sb_uuid or sb_meta_uuid depending on which features are set.
• bb_owner The allocation group number that this B+tree block is located in.
• bb_crc Checksum of the B+tree block.
• bb_pad Pads the structure to 64 bytes.

Block Headers

struct xfs_da3_blkinfo {
/* these values are shared with xfs_da_blkinfo */
__be32 forw;
__be32 back;
__be16 magic;
__be16 pad;
/* These values are unique to da3 blkinfo */
__be32 crc;
__be64 blkno;
__be64 lsn;
uuid_t uuid;
__be64 owner;
};

• forw Logical block offset of the previous B+tree block at this level.
• back Logical block offset of the next B+tree block at this level.
• magic Magic number for this directory/attribute block.
• pad Padding to maintain alignment.
• crc Checksum of the directory/attribute block.
• blkno Block number of this directory/attribute block.
• lsn Log sequence number of the last write to this block.
• uuid The UUID of this block, which must match either sb_uuid or sb_meta_uuid depending on which features are set.
• owner The inode number that this directory/attribute block belongs to.

xfsprogs

xfsprogs is the open source set of xfs utilities used on GNU/Linux to perform filesystem tasks. It includes:

• mkfs.xfs - used to make new xfs filesystems
• fsck.xfs - used to ensure filesystem consistency
• xfs_freeze - suspends access to filesystem to allow dumping or growing the filesystem.
• xfs_info - prints geometry info about xfs.
• xfs_repair - repairs corrupted xfs volumes
• xfs_scrub - used to rebuild filesystem metadata.

Other programs are included as well. These are the notable ones.^[3]

OS Support

Linux kernel has supported XFS since around 2001, in the 2.4.x branch.

IRIX began using XFS starting with major version 6.2. Special 5.3 version was offered with XFS as well.

Shrinking XFS Volumes

As of 2025 neither IRIX nor Linux support shrinking XFS volumes for several reasons:^[4]

• Filesystem has no way to mark allocation groups as unavailable (to prevent use during shrink operation)
• Inodes need to be removed from unavailable allocation groups.
• Somehow move data out of unavailable groups easily.
• xfsprogs lacks ability to move journal structures safely

Further Reading

• XFS Algorithms and Data Structures - Contains useful specifications for the XFS file system, but is from 2006 and is likely outdated.

References