I/O Management

I/O which means input or output is mediated by the OS also known as operating system. I/O resources are shared by multiple programs therefore protection and scheduling is needed. I/O can causes asynchronous interrupts but there are exception for the same mechanism. I/O programming is fiddly as OS provides abstractions to programs.

I/O Commands

I/O devices are managed by I/O controller hardware and data are transferring to and fro device. It also synchronizes operation with software. Command registers will cause device to do something while status registers indicate what the devices is doing and occurrence of errors. As for data register, write simply means  transfer data to a device while read means the other way round.

Technique of I/O

There are 5 types of I/O techniques:

1)      Programmed I/O: The CPU issues a command and wait for I/O operations to be complete. The CPU is faster than the I/O module therefore it is wasteful.

2)      Interrupt Driven I/O: The CPU will issues commands then proceed to its normal work until interrupted by I/O devices on the work completion.

3)      DMA(Direct Memory Access) : In this technique CPU and I/O exchange date without CPU involved.

4)      Memory Mapped I/O: Memory and I/O are treated as memory only which means no signal like IO/M.

5)      Isolated I/O: Address space of memory and I/O is isolated and it uses IO/M signal.

Programmed I/O

Also called polling. It check I/O status register periodically and if device is ready operation will be execute while take action if error occurs. It is common in small or low-performance real-time embedded systems because of its predictable timing and low hardware cost while in other system, CPU time is wasted. The code in the OS for Programmed I/O is more like:

keyboard_wait:               ; for get_ch

      test Keyboard_Status, 80000000h

      jz keyboard_wait

      mov eax, Keyboard_Data

and

   display_wait:                   ; for put_ch

      test Display_Status, 80000000h

      jz display_wait

      mov Display_Data, eax

    This scheme is known as BUSY WAITING, or SPIN WAITING. The little loop is called a SPIN WAIT LOOP.

The problems with Programmed I/O is much time is wasted spin waiting. If it takes 100 instructions to program this, and each instruction takes 20ns to execute, then it takes

   100 * 20nsec = 2000nsec = 2 usec to execute

If a device takes 2msec (=2000usec) to deal with one character, then the percent of time spent waiting

   time waiting / total time = 2000us / 2000us +2us =99.9%

However we will likely have a solution that actually spent less time “doing nothing”.

Interrupt Driven I/O

When a device is ready or an error occurs, controller will interrupts CPU. Interrupt is like an exception but not synchronized to instruction execution and can invoke handler between instructions therefore causes information often identifies the interrupting devices. Priority interrupts are devices needing more urgent attention thus getting higher priority and can interrupt handlers for a lower priority interrupt.

  

Above is the Interrupt Processing and below is the Interrupt Driven I/O Circle.

I/O Data Transfer

For Polling and Interrupt-Driven I/O, CPU transfer data between memory and I/O data registers and its time consuming for high speed devices. Direct Memory Access (DMA) is provided a starting address in memory by the OS and I/O controller transfer to and fro memory autonomously but interrupt upon completion or error.  

Yap Chi Hian B031210334