Assignment: Synchronous Serial Driver

From Embedded Xinu
Revision as of 07:25, 12 December 2008 by Akoehler (talk | contribs) (New page: Category:Student Built Xinu = Overview = This assignment is intended to develop the student's proficiency for programming in C. This assignment is part of the Student Built Xinu t...)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Overview

This assignment is intended to develop the student's proficiency for programming in C. This assignment is part of the Student Built Xinu track for professors that are Teaching With Xinu. This assignment should be completed in teams of two.

Preparation

You will have to familiarize yourself with several common UNIX tools for this assignment. The first of these is tar, a utility originally devised to create tape archives for the purpose of backing files up onto computer tapes.

While tar is still used to create tape backups of file systems, it has become far more common to use tar to group files and/or directories together into a single entity, typically called a "tar-ball." (So common is the use of tar that it has been verbed in computer science terminology: We speak of "tarring" files, or files that have been "tarred up.") Tar syntax is somewhat arcane, as tar came into existence before modern standards for command-line options.

Change to your working directory and execute the following command. This untars the files into your working directory, if the tar-ball was created properly all the files should go into a subdirectory.

tar xvzf <tar-ball location> 

For more information on tar, please see the UNIX man pages.

Building

While the gcc command-line options provide a great deal of flexibility when compiling programs, things quickly become unmanageable when the number of source files exceeds what you can conveniently type in a few seconds.

The make utility can be thought of as a companion to the compiler infrastructure (preprocessor, compiler, assembler, and linker) that allows the build rules for large projects to be explicitly encoded in Makefiles. A Makefile typically consists of common definitions, (such as, which compiler to use), and a set of rules. Each rule has a target, such as the file that is to be built, and can be followed by a list of dependencies and a sequence of steps to perform in order to build that target. In addition, make has quite a few common rules built into it.

You will not have to write your own Makefiles for this course, but you will have to use and possibly modify some for all of our remaining assignments. The Makefile is always human-readable, so feel free to open them up and look around.

To build the Xinu operating system, perform the following steps:

  • Change directory into the top level produced by the tar-ball.
  • Change directory into the subdirectory "compile". This directory contains the XINU project Makefile, and is where all of the compiled ".o" files will go.
  • Execute the the following command:
  make clean 
  • By standard convention, almost all Makefiles include a target called "clean" that removes everything except the source code. The tar-ball you unpacked already should be clean, but it never hurts to make sure that you are starting from a clean slate. You may find yourself using this command often.
  • Execute the the following command:
  make

This should produce about a page of output as each source file is compiled, and the resulting object files are linked together to form the operating system, a simple set of library functions, and the boot loader. If all goes as it should, you should find the directory full of .o files from all of the source code in the other subdirectories, and most importantly, a newly compiled operating system image called "xinu.boot."

For more information on make, please see the UNIX man pages.

Running

Your Xinu image is now ready to be run on a backend machine. To transfer it there, we have a special utility called mips-console. Execute mips-console in the compile directory where your xinu.boot file resides. Mips-console will connect your terminal to the first available backend machine, and you should see a message like:

connection 'xinurouter', class 'mips', host 'xinuserver-hostname'

This will be immediately followed by a stream of automated commands as the embedded target system boots, configures its network settings, and uploads your xinu.boot kernel.

The most important thing to remember about mips-console is that it is modal, like vi. You start out in direct connection mode, in which your terminal connects directly through special hardware to the serial console on your backend machine. To get out of mips-console, hit Control-Space, followed by the 'q' key.

Embedded Xinu Source

The source tar-ball we are starting with contains only a few files for the operating system proper, in the subdirectory system. We will be adding files into this directory in every subsequent assignment. The other files in the XINU subdirectories break down as follows:

  • compile/ contains the compilation files for XINU.
  • include/ contains all of our local .h header files used throughout the rest of the source code files. At compile time, these are treated as being in a standard location, so they can be included with "#include<...>" rather than "#include"..."". The header file particularly important for this assignment is include/uart.h, which contains the structure and constants for the serial port hardware.
  • lib/ contains a small library of standard C functions we can rely upon in the operating system, like strcmp() and atoi(), etc. Remember -- the UNIX system libraries are not available to our operating system running on the backend!
  • loader/ contains the files for the Mips Xinu boot loader, which clears the processor data caches, identifies the processor type, and then passes control to the operating system startup code in the system subdirectory.

Synchronous Serial Driver

Your task for this assignment is to write a simple synchronous serial driver for the embedded operating system, so that you can see what you are doing in all subsequent assignments.

The driver is "synchronous" because it waits for the slow I/O device to do its work, rather than using interrupts to communicate with the hardware. We will write the interrupt-driven, "asynchronous" version of the driver later in the term.

The driver is "serial" because it sends characters one at a time down an RS-232 serial port interface, like the one found on most modern PC's.

The driver is a "driver" because it provides the software interface necessary for the operating system to talk to a hardware I/O device.

This platform's serial port, or UART (Universal Asynchronous Receiver / Transmitter) is a member of the venerable 16550 family of UARTs, documented in the File:Uartspec.pdf. Of particular interest to us is section 8 of the specification, starting on page 14, which describes the registers accessible to programmers. The UART control and status registers are memory-mapped, starting with base address 0xB8000300.

The file system/kprintf.c has the skeleton code for four I/O-related functions: kputc(), (puts a single character to the serial port,) kgetc(), (gets a single character from the serial port) kungetc(), (puts "back" a single character) and kcheckc() (checks whether a character is available). Each function contains a "TODO" comment where you should add code. The actual kprintf() is already complete, and will begin working as soon as you complete the kputc() function it relies upon.