QMCS 360 Class Notes # 7	Dr. Rick Smith Quantitative Methods and Computer Science
Dr. Smith Home | Research | Classes | Blackboard | Cryptosmith | QMCS Home | UST A-Z | UST Home last update: Thursday, October 06, 2005

Homework

Make it due Thursday

Quiz comments

1. Why block vs suspend

Only a couple people mentioned memory shortage

2. Supervisor call versus an interrupt

Here we're interrupting one program with another.

The process might not change, so the process state (running vs blocked) might not change.

3. Mode switch vs process switch

Some OSes run kernel mode in the user's process context, so they aren't the same.

4. Storing the PC

PC = Program Counter - without it, the process can't start up again where it left off executing

5. Preemption

Taking the processor away from one process to run another one. Typically this involves a timer interrupt and the scheduler deciding that the preempted process has run long enough, though it could also happen due to user input (an I/O interrupt) that causes a change in the importance of a running process.

Notes from my notebook

Kernels and microkernels

What do OSes do again?

Microkernel features (break 7 in book down to 3 categories)

Minimal ("Micro" kernel)

Limited functions and API

Easier to validate since it's smaller, thus more reliable

Easier to port to new hardware (thus also flexible)

Flexible

Can add OS functions because the kernel functions are designed to allow OS functions to exist outside the kernel

Easier to port (feature of being minimal)

Distributed computing support

Can schedule separate OS processes on separate processors

Message passing allows you to distribute processes remotely - even outside the physical bounds of a computer's RAM

Major functions

Basic process management

Once a separate scheduling task has chosen a process to run, the microkernel makes it execute

Enough basic process management to get a more sophisticated scheduling process running

Inter-process communication (IPC)

Instead of doing API calls on the stack, the arguments are stored in RAM and passed between processes as data in "messages"

Allows work to be done by separate processes - on-the-stack API forces the requester to block until done

I/O and interrupt handling

Microkernel software fields interrupts and converts them into OS-level "signal" events that get passed to device driver processes

Device drivers get interrupts indirectly

Memory control

Basic functions, while real work is done by a separate memory management process

Grant - give space to a process

Map - share space with a process

Flush - retrieve space previously given to a process

Memory management "owns" all RAM

Grants subsets of RAM to running processes

Processes may grant or map their RAM with other processes

A process may flush its own RAM, but not others'

Diagrams of processes/threads in Windows and Unixes