1. Off-line Direct Disk Access System
CSE 237A
Alan Del Tredici

2. Motivation Number of portable digital devices increasing
Audio recorders
Cameras
Currently, data transferred to PC for
Editing
Storage / media transfer
Network portal
Requires
1. Boot PC
2. Transfer the data
3. Shut down PC

3. Proposed Solution Direct transfer to disk without booting PC
Offline Access
Module

4. Project Goals Demonstrate off-line disk access procedure
Write data to disk
Without standard file I/O calls
Without corrupting existing data
Retrieve data via standard file I/O
Proof-of-concept stage is software only

5. Development Choices Chose to use Linux as native OS
Widely used, source code available, and free!
Use “Second Extended File System” (Ext2)
Unix-style file system
De facto standard for Linux
Basis from which other file systems extend

6. Development Choices (cont.)
Make use of disk partitions
Make 1 large disk look like several small
Provides efficiency, modularity
In Linux, floppy-disk accessed just like a hard-disk partition
Use floppy disk as test target
No change in concept
Safer for development!

7. Off-line Access Strategy:
1. Directly access disk to determine existing file system configuration
2. Add data, maintaining file system structure.
3. When booted, Linux will find the data

8. Development Approaches
Write code to directly access disk
Linux has strong protection – user can’t access hardware!
Linux provides raw block read/write calls…
Want access to device in terms of file-system components
This is provided by libext2fs.
Basic functions to access file system structures
Still requires us to know how to use the pieces

9. Ext2 – Partition Components
Boot sector
1 or more Groups
Super block
Group descriptor
Block usage map
I-node usage map
I-node table
Data blocks
Redundant copies
Disk is an array of “blocks” – organized by these conventions

10. Ext2 – Boot Sector Boot sector 1 or more Groups Boot code
Super block
Group descriptor
Block usage map
I-node usage map
I-node table
Data blocks
Boot code
May be unused
Partition tables
List start of partitions

11. Ext2 – Super Block Boot sector 1 or more Groups Basic parameters
Group descriptor
Block usage map
I-node usage map
I-node table
Data blocks
Basic parameters
Block size (1 kB typ)
Block counts
Version info

12. Ext2 – Group Descriptor Boot sector 1 or more Groups
Super block
Group descriptor
Block usage map
I-node usage map
I-node table
Data blocks
Map & table locations
Handy counters

13. Ext2 – Usage Maps Boot sector 1 or more Groups
Super block
Group descriptor
Block usage map
I-node usage map
I-node table
Data blocks
Bit-maps where each bit indicates used/unused item
A 1 kB block holds 8k bits
Supports 8,192 entries

14. Ext2 – Index nodes (i-nodes)
Boot sector
1 or more Groups
Super block
Group descriptor
Block usage map
I-node usage map
I-node table
Data blocks
Ext2 i-node = 128 bytes
struct ext2_inode {
U32 size
U32 Block count
U32 Time stamps :
U32 Block pointer[0]
U32 Block pointer[1]
U32 Block pointer[11]
U32 Indirect block pointer
U32 2x-indirect block pointer
U32 3x-indirect block pointer }

15. Ext2 – I-node Block Pointers

16. Ext2 – Data Blocks Boot sector 1 or more Groups
Super block
Group descriptor
Block usage map
I-node usage map
I-node table
Data blocks
All remaining blocks in the group

17. Off-line Access Example
Initialized Disk:
I-node table: 23 blocks
Root directory:
Directory is a type of ‘file’
Index to other i-nodes by name
Root = i-node #2
“lost+found” reserve: 12 blocks
Block map
I-node map
Group data
Super block
Boot block : 2 3 1 4 5 27 28 40 41 29
1439
I-node table
Root directory
“lost+found”
1st free data block

18. Off-line Access Steps Read configuration information Super block
Block size (1 kbyte)
I-node count (184 on fd)
Block count (1440 on fd)
Group block
Location of maps
Location of i-node table
Block map
I-node map
Group data
Super block
Boot block : 2 3 1 4 5 27 28 40 41 29
1439
I-node table
Root directory
“lost+found”
1st free data block

19. File Creation Read the i-node and block maps
Find the next available i-node and data block
Block map
I-node map
Group data
Super block
Boot block : 2 3 1 4 5 27 28 40 41 29
1439
I-node table
Root directory
“lost+found”
1st free data block

20. File Creation (continued)
Add entry to specified directory
File name
File type
I-node number
Perform i-node “accounting”
Mark bit in i-node map
Block map
I-node map
Group data
Super block
Boot block : 2 3 1 4 5 27 28 40 41 29
1439
I-node table
Root directory
“lost+found”
1st free data block

21. File Creation (continued)
Write data to data blocks
Update i-node with block numbers
Allocate indirect blocks from data block area as needed
Block map
I-node map
Group data
Super block
Boot block : 2 3 1 4 5 27 28 40 41 29
1439
I-node table
Root directory
“lost+found”
Our data!

22. File Creation (continued)
Finish configuring the i-node
Set permissions
Size of data written
Time stamp
When Linux reads this file system, our data will appear!
Block map
I-node map
Group data
Super block
Boot block : 2 3 1 4 5 27 28 40 41 29
1439
I-node table
Root directory
“lost+found”
Our data!

23. Accomplishments Met the technical objectives
Data stored to hard disk without file I/O calls
Linux can later retrieve data
Documented pertinent Ext2 operation
Extracted details from source code & by testing
Created working proof-of-concept application
Available on web site:

24. Future Work Test on hard disk with large files
Implement on stand-alone hardware
Atmel AT89C5132:
A mass storage device
controlling data exchange
between various Flash modules, HDD and CD-ROM.

25. Off-line Disk Access
Questions?

26. Off-line Disk Access
Thank you!