Chapter 40File System

ImplementationChien-Chung Shen

CIS, UDcshen@cis.udel.edu

Learning by Example

• A simple file system implementation– vsfs (Very Simple File System)– a simplified version of Unix file system– features include on-disk structures, access methods, and

various policies • File system is pure software so that no special

hardware are added for better performance; however, need to understand and take advantage of device characteristics to make sure file systems work well

• e.g., AFS -> ZFS with different data structures and features (pros and cons): learning via case studies

Mental Model of FS

• Mental models are what you are really trying to develop when learning about systems

• Mental model of file systems include answers to questions like– what on-disk structures store the file system’s data and

metadata? – what happens when a process opens a file? – which on-disk structures are accessed during a read or

write?

The Way to Think

• Two aspects: object-oriented view– data structures, e.g., what types of

on-disk structures are utilized by file system to organize its data and metadata (e.g., array of blocks or tree-based structure)

– access methods, e.g., how system calls are mapped onto the data structures (e.g., which structures are read during the execution of a particular system call)

Overall Organization (1)

• Overall on-disk organization of the data structures of the vsfs file system – disk is divided into a serious of equal-sized blocks– e.g., a small disk is divided into 64 “4KB blocks”

– 56-block “data region” for user data

Overall Organization (2)

• Metadata [stored in inode] tracks each file (its data blocks, owner, access rights, modify time, etc.)

• inode table on disk (5*16 “256-byte inodes”)• Need to track whether inodes and data blocks are

free or allocated (allocation structures)– free list or bitmap (data bitmap and inode bitmap)

Overall Organization (3)

• Superblock: contains information about this file system, including how many inodes and data blocks are in the file system (80 and 56, respectively), where the inode table begins (block 3), etc.

• When mounting file system, OS reads the superblock first to initialize various parameters

File Organization: inode (1)

• “Index node” holding a file’s metadata

• Inode table of 5 “4KB blocks” • To read inode number 32, calculate offset into

inode region (32*sizeof(inode) = 8192), add it to the start address of the inode table on disk (12KB) to get 20KB

File Organization: inode (2)

• Disks are not byte addressable, but rather consist of many addressable sectors, usually 512 bytes.

• Thus, to fetch the block of inodes that contains inode 32, file system issues a read to sector 40 (20*1024/512) to fetch the desired inode block

blk = (inumber * sizeof(inode_t)) / blockSize; sector = ((blk * blockSize) + inodeStartAddr) / sectorSize;

Where Data Blocks Are

• One option: have one or more direct pointers (disk addresses) inside the inode, each refers to one disk block

• Limited file size• Multi-level index

– indirect pointers: instead of pointing to a block that contains user data, it points to a block that contains more pointers, each of which point to user data

– e.g., 4 KB block and 4-byte disk addresses (12+1024)*4K

– double indirect pointers– triple indirect pointers

Multi-Level Index