Running Unix v6 in SIMH

From Computer History Wiki
Jump to: navigation, search

The following was taken from http://wwwlehre.ba-stuttgart.de/~helbig/os/script/chapt1.1

I need to verify it still works, and include some configuration notes..

Installing Unix V6

The program product was the last one assembled manually. Unix was used to develop Unix so we'll use Unix to study Unix. For this, you first need to install a bootable binary disk from the distribution tape.

The paper Setting up UNIX - Sixth Edition explains usage of the secondary boot program. (See Installing UNIX Sixth Edition for details on what is actually happening here.) Use it to load "tmrk". "tmrk" copies blocks from tape to disk. You can use ltap as the primary boot program, which loads the secondary boot from the distribution tape. Before booting Unix from disk, set the SR to 173030. This will boot into single user mode, which is enough right now. After all, we only have a single terminal.

SIMH: A version of the distribution tape suitable for SIMH is filed at v6/dist.tap.

SIMH: Use "boot rk0" to load the first block from disk. You don't need to key in a primary boot program to boot from disk.

SIMH: Use "set tto 7b" to set the output to seven bits per character. The SIMH setting defaults to eight bits per character and doesn't work, because the TTY driver sets bit 7 to a parity bit.

The shell prompts you with "#" to enter commands. To use unix, you need to read the "Unix Programmer's Manual", see v6/doc/index.html.

The TTY driver assumes an upper case only console. Use "stty -lcase" to teach it better. Another source of confusion might be the "change directory" command, which reads "chdir" in Unix V6 and not "cd" as in its successors.

Notation: foo(number) references the manual page "foo" in man.number.ps For example stty(I), tty(IV) and chdir(I). The manual pages are provided as html pages as well, e. g., you find the stty page at v6/doc/I/stty.html

You want to verify the installation by running two file system checks, namely icheck(VIII) and dcheck(VIII). Icheck checks that every block in a filesystem is either allocated to exactly one file or on the list of free blocks. Dcheck verifies that every file has at least one name, i.e., an entry in a directory, and that the number of names of a file equals the link count stored with the file.

Icheck and dcheck require a device name argument. Device names are directory entries much like names of regular files. They bind a name to a device, i.e., memory, terminal, tape or disk. Unlike regular file names, device names are created with mknod(VIII). It takes four arguments: the name of the device, conventionally an entry in the /dev directory; the type of the device (c for character, b for block); a major number, which identifies the device driver, and a minor number, which identifies one of the eight possibly attached disk drives.

Device files are called "special files" as opposed to "regular files".

The major number is an index to the driver's entry in either the "character device switch" (cdevsw) or the "block device switch" (bdevsw). Both tables are written into the file /usr/sys/conf/c.c by a program called "mkconf". "Mkconf" itself is distributed as a source file /usr/sys/conf/mkconf.c. For the commands needed to build mkconf and c.c, consult the shell script "/usr/sys/run". It configures and builds the distribution kernels for a variety of hardware configurations.

Besides c.c, mkconf creates the assembler program l.s, which defines the low memory part of the kernel, namely the trap and interrupt vectors, which is the yellow area. In other words, mkconf selects the drivers and creates the corresponding interrupt vectors as needed by a particular hardware configuration. When the kernel is linked, only drivers that are referred to in c.c, will be included.

You only need to specify rk and tm to mkconf. Other devices, like the console driver or the memory, are included automatically by mkconf.

Here is a typescript that creates mkconf, c.c and l.s:

	# chdir /usr/sys/conf
	# cc mkconf.c
	# mv a.out mkconf
	# mkconf
	rk
	tm
	done
	#

From the c.c file we learn the assignement of major numbers to drivers: block devices rk=0, tm=3, character devices: kl=0, mem=8, rk=9, tm=12

You are now ready to name the rk0 character device. Since character device in this case means the kernel does not do any "blocking", the name traditionally starts with an extra "r" for raw device.

The typescript continued:

# /etc/mknod /dev/rrk0 c 9 0

And finally the filesystem checks:

# icheck /dev/rrk0
/dev/rrk0:
spcl       5
files    292
large     95
direc     24
indir     95
used    2902
free    1011
# dcheck /dev/rrk0
/dev/rrk0:

If your output looks similar to the above, you are ready for the completion of the installation:

Create block and character device names for rk1 and tm0. The names of the tape device files are "/dev/mt0" and "/dev/rmt0", not "/dev/tm0".

Copy the source disk (blocks [4100, 8100) on tape) to /dev/rrk1. To copy blocks, use dd(I): dd if=/dev/rmt0 of=/dev/rrk1 skip=4100 count=4000

SIMH: Remember to attach rk1 before running the dd command.

Now you are ready to mount(VII) the file system: /etc/mount /dev/rk1 /usr/source

Note that /etc/mount wants you to specifiy a block device.