1. Device Management

2. Device Drivers
The part of OS that implements device management is called collectively as device manager, which includes device drivers and device driver infrastructure
Device drivers can be complex
However, the drivers for different devices implement a similar interface so that a higher-level component has a uniform interface for different devices

3. Device Driver Interface
Device Interface …
write(…);
Terminal
Driver
Printer
Disk
Controller

4. Device Management Organization
Application
Process
File
Manager
Device Controller
Command
Status
Data
Hardware Interface
System Interface
Device-Independent
Device-Dependent

5. System Call Interface Functions available to application programs
Abstract all devices (and files) to a few interfaces
Make interfaces as similar as possible
Block vs character
Sequential vs direct access
Device driver implements functions (one entry point per API function)

6. Example: BSD UNIX Driver
open Prepare dev for operation
close No longer using the device
ioctl Character dev specific info
read Character dev input op
write Character dev output op
strategy Block dev input/output ops
select Character dev check for data
stop Discontinue a stream output op

7. Overlapping the Operation of a Device and the CPU
. . .
startRead(dev_I, “%d”, x);
While(stillReading()) ;
y = f(x)
. . .
read(dev_I, “%d”, x);
y = f(x)
Data on device
Variable x
Register
Device dev_I
Memory
CPU

8. Overlapping CPU-Controller Operations in a Process
I/O Ctlr t1 t2 t3 t4 t5 t6 t7 t8 t9

9. Overlapping Processing and I/O
I/O Ctlr t1 t2 t3 t4
Dashed line is App 1’s activity
Solid line is App 2’s activity

10. Device Management
Responsible for directly manipulating the hardware devices
Several strategies
Direct I/O: CPU stays in control
Use device polling
Use interrupts
Memory-mapped I/O
Direct memory access

11. I/O Strategies Direct I/O Direct Memory Access (DMA)
The CPU is responsible for determining when the I/O operation has completed and then for transferring the data between the primary memory and the device controller data registers
Polling or interrupt-driven I/O
This seems not effective for devices with large transfers, such as a disk drive
Direct Memory Access (DMA)

12. Direct-Memory Access Direct memory access controllers
Can read/write data directly from/to primary memory addresses with no CPU intervention after the driver has started the I/O operation

13. Direct-Memory Access – cont.

14. Direct-Memory Access – cont.

15. I/O Strategies – cont. Direct I/O or DMA for data transferring
Polling or interrupt-driven
I/O strategies
Direct I/O with polling
DMA I/O with polling
Not supported in general
Direct I/O with interrupts
DMA I/O with interrupts

16. Direct I/O with Polling
read(device, …);
Data
Device Controller
Command
Status
read function
write function 1 2 3 4 5
Hardware Interface
System Interface

17. Device Controllers – cont.
Command
Status
Data 0
Data 1
Data n-1
Logic
busy
done
Error code
. . .
busy done
idle
finished
working
(undefined)

18. Polling I/O busy done Software Hardware … // Start the device
While(busy == 1)
wait();
// Device I/O complete
done = 0;
while((busy == 0) && (done == 1))
// Do the I/O operation
busy = 1;
busy
done
Software
Hardware

19. Interrupt-Driven I/O System Interface Hardware Interface
read(device, …);
Data
Device Controller
Command
Status
read driver
write driver 1 2 3 4 5
Hardware Interface
System Interface
Device Status Table
Device
Handler
Interrupt 6 7 8a 8b 9

20. Review: Interrupt Handling
At the end of each instruction cycle, we check if an interrupt has occurred

21. Interrupt-Driven Direct-Memory Access – cont.

22. Interrupt-Driven Versus Polling
In general, an interrupt-driven system will have a higher CPU utilization than a polling-based system
The CPU time on polling by one process may be utilized by another process to perform computation

23. Comparison of Polling and Interrupt-Driven I/O Performance
Polling is normally superior from the viewpoint of a single process
Because overhead of polling is less
Interrupt-driven I/O approach gives better overall system performance

24. I/O-bound and Compute-bound Processes
I/O-bound process
A process’s overall execution time is dominated by the time to perform I/O operations
Compute-bound process
A process’s time on I/O is small compared to the amount of time spent using the CPU
Many processes have I/O-bound and compute-bound phases

25. Phases of a Process
Compute-bound
I/O-bound
Time

26. Device Manager Design Device drivers
Application programming interface
Coordination among application processes, drivers, and device controllers
Performance optimization

27. The Kernel Interface

28. Device-Independent Function Call
funci(…)
Trap Table
dev_func_i(devID, …) {
// Processing common to all devices …
switch(devID) {
case dev0: dev0_func_i(…);
break;
case dev1: dev1_func_i(…);
case devM: devM_func_i(…); }; }

29. Re-configurable Device Drivers
Other
Kernel
services
Entry Points for Device j
open(){…}
read(){…}
etc.
System call interface
Driver for Device j

30. Handling Interrupts Device driver J Device interrupt handler J
Device status table
int read(…) {
// Prepare for I/O
save_state(J);
out dev#
// Done (no return) }
void dev_handler(…) {
get_state(J);
//Cleanup after op
signal(dev[j]);
return_from_sys_call(); } J
Interrupt Handler
Device Controller

31. Handling Interrupts – cont.
Device driver J
Device interrupt handler J
int read(…) { …
out dev#
// Return after interrupt
wait(dev[J});
return_from_sys_call(); }
void dev_handler(…) {
//Cleanup after op
signal(dev[j]); }
Interrupt Handler
Device Controller

32. CPU-device Interactions

33. Buffering
Buffering
A technique to keep slower I/O devices busy during times when a process is not requiring I/O operations
Input buffering
Output buffering

34. The Pure Cycle Water Company
Customer Office
Water Company
Returning the Empties
Water Producer
Water Consumers
Delivering Water

35. Buffering – cont. Unbuffered Process reads bi-1 Controller reads bi
Data
Device B A
Unbuffered
Process reads bi-1
Controller reads bi
Process reads bi
Controller reads bi+1

36. Double Buffering
Process
Controller B
Device A
Hardware
Driver

37. Circular Buffers - Producer-consumer problem To data consumer Buffer i
From data producer
To data consumer
Buffer i
Buffer j
- Producer-consumer problem

38. Spooling
A spool is a buffer that holds output for a device that cannot accept interleaved data streams
Printer is an example

39. Device Characteristics
Character devices vs. block devices
Communication devices
Sequentially accessed storage devices
Randomly accessed storage devices

40. Communication Devices
Generic
Controller
Local
Device
Communications
Cabling connecting the
controller to the device
Printer
Modem
Network
Bus

41. Serial Port
CPU
Memory
Serial
Device
Printer
Terminal
Modem
Mouse
etc.

42. Serial Port RS-232 Interface Serial Device (UART) UART API
9-pin connector
4-wires
bit transmit/receive
...
Serial Device
(UART)
UART API
parity
bits per byte
etc.
Device Driver
Set UART parms
read/write ops
Interrupt hander
Software on the CPU
Device Driver API
Bus Interface

43. Logical Communication
Data Networks
Technology focus includes protocols and software
(more on this later … Chapter 15 and beyond ...)
Network
Device
Memory
CPU
Data Network
Logical Communication

44. Randomly Accessed Devices
Cylinder
(set of tracks)
Track (Cylinder)
Sector
(a) Multi-surface Disk
(b) Disk Surface
(b) Cylinders

45. Optimizing Access to Rotating Devices
Access time has three major components
Seek time – the time for the disk arm to move the heads to the cylinder containing the desired sector
Rotational latency – the additional time waiting for the disk to rotate the desired sector to the disk head
Transfer Time - time to copy bits from disk surface to memory

46. Optimize Access Times
Several techniques used to try to optimize the time required to service an I/O request
FCFS – first come, first served
Requests are serviced in order received
No optimization
SSTF – shortest seek time first
Find track “closest” to current track
Could prevent distant requests from being serviced

47. Optimize Access Times – cont.
Scan/Look
Starts at track 0, moving towards last track, servicing requests for tracks as it passes them
Reverses direction when it reaches last track
Look is variant; ceases scan in a direction once last track request in that direction has been handled
Circular Scan/Look
Similar to scan/look but always moves in same direction; doesn’t reverse
Relies on existence of special homing command in device hardware

48. Optimizing Seek Time
Multiprogramming on I/O-bound programs => set of processes waiting for disk
Seek time dominates access time => minimize seek time across the set
Example:
Tracks 0:299; Head at track 76, requests for 124, 17, 269, 201, 29, 137, 12

49. Optimizing Access to Rotating Devices
First-Come-First-Served
(76), 124, 17, 269, 201, 29, 137, 12
=>
= 880 steps

50. Optimizing Access to Rotating Devices – cont.
Shortest-Seek-Time-First
(76), 29, 17, 12, 124, 137, 201, 269
=>
= 321 steps

51. Optimizing Access to Rotating Devices – cont.
Scan
(76), 124, 137, 201, 269, 299, 29, 17, 12
=>
= 510 steps
Look
(76), 124, 137, 201, 269, 29, 17, 12
=> = 450 steps

52. Optimizing Access to Rotating Devices – cont.
Circular Scan/
Circular Look
Circular Look: (76), 124, 137, 201, 269, 12, 17, 29
home = home
Circular Scan: (76), 124, 137, 201, 269, 299, 12, 17, 29
home = home

53. Device Management in Linux
While it builds on the same basic principles, it is more complex and with more details
There are a lot of device drivers included and you can also develop your own device drivers for specific purpose devices
Linux divides into char-oriented devices and block oriented devices
fs/blk_dev.c
fs/char_dev.c
See pp

54. NT Driver Organization

55. NT Device Drivers API model is the same as for a file
Extend device management by adding modules to the stream
Device driver is invoked via an Interrupt Request Packet (IRP)
IRP can come from another stream module
IRP can come from the OS
Driver must respond to minimum set of IRPs