I/O system

I/O hardware

• Port: a connection point between I/O devices and the host.
• E.g.: user ports.
• Bus: a set of wires and a well-defined protocol that specifies messages sent over the wires.
• E.g.: PCI bus.
• Controller: a collection of electronics that can operate a port, a bus, or a device.
• A controller could have its own processor and memory. Etc. (e.g.: SCSI controller).

Basic I/O Method (Port-mapped I/O)

• Each I/O port (device) is identified by the unique port address.
• Each I/O port consists of four registers (1~4bytes).
• Data-in register: read by the host to get input
• Data-out register: written by the host to send output
• Status register: read by the host to check I/O status
• Control register: written by the host to control the device
• The program interacts with an I/O port through special I/O instructions (different from mem. access).
• X86: IN, OUT

I/O methods Categorization

• Depending on how to address a device:
• Port-mapped I/O.
• Use different address spaces for memory.
• Access by special I/O instruction (e.g. IN, OUT).
• Memory-mapped I/O.
• Reserve specific memory space for the device.
• Access by standard data-transfer instruction (e.g. MOV).
• Good: 
• More efficient for large memory I/O (e.g. graphic card).
• Bad:
• Vulnerable to accident modification error.
• Depending on how to interact with the device:
• Poll(busy-waiting): processor periodically checks the status register of a device
• Interrupt: device notifies the processor of its completion
• Depending on who controls the transfer:
• Programmed I/O: transfer controlled by CPU.
• Direct memory access(DMA) I/O: controlled by a DMA controller (a special purpose controller).
• Design for large data transfer.
• Commonly used with memory-mapped I/O and interrupt.

I/O subsystem

• I/O Scheduling - improve system performance by ordering the jobs in the I/O queue.
• E.g. disk I/O order scheduling.
• Buffering - store data in memory while transferring between I/O devices.
• Speed mismatch between I/O devices.
• Devices with different data-transfer sizes.
• Support copy semantics.
• Caching - fast memory that holds copies of data.
• Always just a copy.
• Key to performance.
• Spooling - holds output for a device.
• E.g. printing (cannot accept interleaved files).
• Error handling - when an I/O error happens.
• E.g. SCSI devices return error information.
• I/O protection
• Privilege instructions.

Blocking and Nonblocking I/O

• Blocking: process suspended until I/O completed
• Easy to use and understand
• Insufficient for some needs 
• Use for synchronous communication & I/O
• Nonblocking
• Implemented via multi-threading
• Returns quickly with the count of bytes read or written
• Use for asynchronous communication & I/O

Performance

• I/O is a major in system performance.
• It places heavy demands on the CPU to execute device driver code.
• The resulting context switches stress the CPU and its hardware caches.
• I/O loads down the memory bus during data copy between controllers and physical memory.
• Interrupt handling is a relatively expensive task.
• Busy waiting could be more efficient than interrupt-driven if I/O time is small.

Improving performance

• Reduce the number of context switches
• Reduce data copying
• Reduce interrupts by using large transfers, smart controllers, polling
• Use DMA
• Balance CPU, memory, bus, and I/O performance for highest throughput

reference:

https://www.amazon.com/-/zh_TW/Operating-System-Concepts-Abraham-Silberschatz/dp/1119800366/ref=sr_1_1?keywords=Operating-System-Concepts&qid=1669538704&s=books&sr=1-1