|
In addition to the "working" guest operating systems listed above, you might find the following information interesting:
Although it is possible to let a guest OS access real hardware, such as harddisks, it is much more flexible and attractive to simulate harddisks using files residing in the host's filesystem. On Unix-like systems, files may contain holes, which makes this really simple. To the guest operating system, the harddisk image looks and acts like a real disk.
To install NetBSD/pmax onto a harddisk image in the emulator, follow these instructions:
$ dd if=/dev/zero of=nbsd_pmax.img bs=1 count=512 seek=1900000000
From this point, there are two separate ways to continue the installation. You can either download a CD-ROM iso image (and let the installation program copy files from the CD-ROM image to the harddisk image), or you can install via ftp. For an installation from a CD-ROM image, follow these steps:
ftp://ftp.netbsd.org/pub/NetBSD/iso/1.6.2/pmaxcd.iso or ftp://ftp.netbsd.org/pub/NetBSD/iso/2.0.2/pmaxcd.iso
$ gxemul -X -E dec -e 3max -d nbsd_pmax.img -d bc:pmaxcd.isoand proceed like you would do if you were installing NetBSD on a real DECstation.
For an ftp install, substitute steps 2 and 3 above with these:
ftp://ftp.netbsd.org/pub/NetBSD/NetBSD-1.6.2/pmax/binary/kernel/netbsd-INSTALL.gz or ftp://ftp.netbsd.org/pub/NetBSD/NetBSD-2.0.2/pmax/binary/kernel/netbsd-INSTALL.gz
$ gxemul -X -E dec -e 3max -d nbsd_pmax.img -O netbsd-INSTALL.gzand proceed like you would do if you were installing NetBSD on a real DECstation. Suitable networking parameters are as follows:
Which device shall I use? [le0]: le0 .. Your DNS domain: mydomain.com Your host name: foo Your IPv4 number: 10.0.0.1 IPv4 Netmask [0xff000000]: 0xff000000 IPv4 gateway: 10.0.0.254 IPv4 name server: 10.0.0.254(If using 10.0.0.254 as the nameserver fails, then try entering the IP number of a real-world nameserver instead.)
(If you don't want to use a graphical framebuffer during the install, you can remove -X from the command line, but then make sure you choose "vt100" when prompted with which terminal type to use, and not "rcons". If you want to use X, but think that the default framebuffer window is too large, try adding -Y2 to the command line.)
When the installation is completed, the following command should start NetBSD from the harddisk image:
$ gxemul -X -M64 -E dec -e 3max -d nbsd_pmax.img
Use startx to start X windows.
NOTE: For some reason, NetBSD 2.0.2 doesn't work with X out-of-the-box on pmax. It seems that this has to do with a switch to WSCONS. For now, if you want X, then try NetBSD 1.6.2.
If you want to run without the X framebuffer, use this instead:
$ gxemul -E dec -e 3max -d nbsd_pmax.img
To install NetBSD/arc from a CDROM image onto an emulated harddisk image, follow these instructions:
$ dd if=/dev/zero of=nbsd_arc.img bs=1024 count=1 seek=900000
ftp://ftp.netbsd.org/pub/NetBSD/iso/1.6.2/arccd.iso
$ gxemul -E arc -e pica -x -d nbsd_arc.img -d bc:arccd.iso \ -j arc/binary/kernel/netbsd.RAMDISK.gz(Try removing -x if you have problems with the xterm.)
$ mount /dev/cd0a /mnt2 $ disklabel -i -I sd0 (for example 'a', '4.2BSD', '1c', '700M', 'b', 'swap', '701M', '$', 'P', 'W', 'y', and 'Q') $ newfs /dev/sd0a $ mount /dev/sd0a /mnt $ cd /mnt $ for a in /mnt2/arc/binary/sets/*.tgz; do echo $a; tar xzpf $a; done $ cd dev; sh MAKEDEV all $ cd ../etc; echo "rc_configured=YES" >> rc.conf $ cat > /mnt/etc/fstab /dev/sd0a / ffs rw 1 1 /dev/sd0b none swap sw 0 0 (ctrl-d) $ cd /; umount /mnt; umount /mnt2 $ halt
ftp://ftp.netbsd.org/pub/NetBSD/NetBSD-1.6.2/arc/binary/kernel/netbsd-GENERIC.gz
You can now use the generic NetBSD/arc kernel to boot from the harddisk image, using the following command:
$ gxemul -E arc -e pica -x -d nbsd_arc.img netbsd-GENERIC.gz
When asked for "root device: ", enter sd0.
| Model: | Framebuffer size/depth: | Framebuffer address: | |||
| MobilePro 770 ( |
640 x 240, 16 bits | 0xa000000 | |||
| MobilePro 780 | 640 x 240, 16 bits | 0xa180100 ( |
|||
| MobilePro 800 | 800 x 600, 16 bits | 0xa000000 | |||
| MobilePro 880 | 800 x 600, 16 bits | 0xa0ea600 ( |
(
(
These instructions show an example of how to install NetBSD/hpcmips on an emulated MobilePro 800:
$ dd if=/dev/zero of=nbsd_hpcmips.img bs=1024 count=1 seek=1990000
ftp://ftp.se.netbsd.org/pub/NetBSD/iso/2.0.2/hpcmipscd.iso(You may want to choose a mirror closer to you, if .se is slow.)
$ gxemul -E hpc -e mobilepro800 -X -d nbsd_hpcmips.img \ -d b:hpcmipscd.iso -j hpcmips/installation/netbsd.gzand proceed like you would do if you were installing NetBSD on a real MobilePro 800. (Install onto wd0, choose "Use entire disk" when doing the MBR partitioning, and choose wd1d (not cd0c) as the CDROM device to install from.)
If everything worked, NetBSD should now be installed on the disk image. GXemul does not (yet) support reading the kernel directly from the disk image, so you need to download a generic kernel separately:
ftp://ftp.netbsd.org/pub/NetBSD/NetBSD-2.0.2/hpcmips/binary/kernel/netbsd-GENERIC.gz
The installation is now complete. Use the following command line to boot the emulated hpcmips machine:
$ gxemul -E hpc -e mobilepro800 -X -d nbsd_hpcmips.img netbsd-GENERIC.gz
If you change your mind at this point regarding which machine type to emulate, you might for example prefer a MobilePro 770, then you can change that at any time. NetBSD is designed to be able to boot on many types, without any need to change the kernel.
When you have logged in as root, you can use startx to start X Windows. (Note: There is no mouse support yet; you can only use keyboard input.)
The following instructions will let you install NetBSD/cobalt onto a disk image, from an emulated DECstation 3MAX machine:
$ dd if=/dev/zero of=nbsd_cobalt.img bs=1024 count=1 seek=1999000
ftp://ftp.se.netbsd.org/pub/NetBSD/NetBSD-2.0.2/cobalt/binary/kernel/netbsd-GENERIC.gz ftp://ftp.se.netbsd.org/pub/NetBSD/iso/2.0.2/cobaltcd.iso(You may want to choose a mirror closer to you, if .se is slow.)
$ gxemul -Edec -e3max -d nbsd_pmax.img -d cobaltcd.iso -d nbsd_cobalt.img
newfs /dev/sd1c mount /dev/cd0c /mnt mkdir /mnt2; mount /dev/sd1c /mnt2 cd /mnt2; sh for a in /mnt/cobalt/binary/sets/*.tgz; do echo $a; tar zxfp $a; done exit cd dev; sh ./MAKEDEV all; cd ../etc echo rc_configured=YES >> rc.conf echo "/dev/wd0d / ffs rw 1 1" > fstab cd /; umount /mnt; umount /mnt2; halt
You should now be able to boot NetBSD/cobalt like this:
$ gxemul -M128 -E cobalt -d nbsd_cobalt.img netbsd-GENERIC.gzNote that the installation instructions above create a filesystem without a disklabel, so there is only one ffs partition and no swap. You will need to enter the following things when booting with the generic kernel:
root device (default wd0a): wd0d dump device (default wd0b): none file system (default generic): ffs init path (default /sbin/init): (just press enter here)
It is tricky to install, because there is (as far as I know) no INSTALL kernel. One way to install the NetBSD/evbmips distribution onto a disk image is to install the files is to do it using another (emulated) machine.
The following instructions will let you install NetBSD/evbmips onto a disk image, from an emulated DECstation 3MAX machine:
$ dd if=/dev/zero of=nbsd_malta.img bs=1024 count=1 seek=999000
ftp://ftp.se.netbsd.org/pub/NetBSD/NetBSD-2.0.2/evbmips-mipsel/binary/kernel/netbsd-MALTA.gz ftp://ftp.se.netbsd.org/pub/NetBSD/iso/2.0.2/evbmips-mipselcd.iso(You may want to choose a mirror closer to you, if .se is slow.)
$ gxemul -Edec -e3max -d nbsd_pmax.img -d nbsd_malta.img -d evbmips-mipselcd.iso
newfs /dev/sd1c mount /dev/cd0c /mnt mkdir /mnt2; mount /dev/sd1c /mnt2 cd /mnt2; sh for a in /mnt/evbmips-mipsel/binary/sets/*.tgz; do echo $a; tar zxfp $a; done exit cd dev; sh ./MAKEDEV all; cd ../etc echo rc_configured=YES >> rc.conf echo "/dev/wd0c / ffs rw 1 1" > fstab cd /; umount /mnt; umount /mnt2; halt
You should now be able to boot NetBSD/evbmips like this:
$ gxemul -Eevbmips -emalta -d nbsd_malta.img netbsd-MALTA.gz
Note 1: NetBSD detects a very fast CPU although the emulation isn't really very fast, so delays take very long. Even on a multi-GHz host, you will need a lot of patience.
Note 2: To select a 4Kc (MIPS32) CPU instead of the default 5Kc (MIPS64) CPU, add -C 4Kc to the command line. With NetBSD 2.0.2, however, there will be little or no difference in functionality. (NetBSD still runs in 32-bit mode on 64-bit MIPS CPUs.)
Note 3: The installation instructions above create a filesystem without a disklabel, so there is only one ffs partition and no swap. You will need to enter the following things when booting with the generic kernel:
root device (default wd0a): wd0c dump device (default wd0b): none file system (default generic): ffs init path (default /sbin/init): (just press enter here)
NetBSD/sgimips can run in GXemul on an emulated O2 (SGI-IP32). However, GXemul does not yet emulate the AHC PCI SCSI controller in the O2. (I have mailed Adaptec several times, asking for documentation, but never received any reply.) NetBSD can still run, as long as it doesn't use SCSI.
For a simple test with the 2.0.2 ramdisk (install) kernel, try dowloading
ftp://ftp.NetBSD.org/pub/NetBSD/NetBSD-2.0.2/sgimips/binary/kernel/netbsd-INSTALL32_IP3x.gzand run gxemul -E sgi -e o2 netbsd-INSTALL32_IP3x.gz.
It is possible to set up an environment for netbooting the emulated SGI machine off of another emulated machine. Performing this setup is quite time consuming, but necessary:
$ gxemul -M64 -Edec -e3max -d nbsd_pmax.imgand enter the following commands as root inside the emulator:
echo hostname=server >> /etc/rc.conf
echo ifconfig_le0=\"inet 10.0.0.2\" >> /etc/rc.conf
echo nameserver 10.0.0.254 >> /etc/resolv.conf
echo 10.0.0.254 > /etc/mygate
echo /tftpboot -maproot=root 10.0.0.1 > /etc/exports
echo rpcbind=YES >> /etc/rc.conf
echo nfs_server=YES >> /etc/rc.conf
echo mountd=YES >> /etc/rc.conf
echo bootparamd=YES >> /etc/rc.conf
printf "client root=10.0.0.2:/tftpboot \\\n swap=10.0.0.2:/tftpboot/swap\n" > /etc/bootparams
echo "bootps dgram udp wait root /usr/sbin/bootpd bootpd -d 4 -h 10.0.0.2" >> /etc/inetd.conf
cat >> /etc/bootptab
client:\
:ht=ether:\
:ha=102030000010:\
:sm=255.0.0.0:\
:lg=10.0.0.254:\
:ip=10.0.0.1:\
:rp=/tftpboot:
(press CTRL-D)
echo "10:20:30:00:00:10 client" > /etc/ethers
echo 10.0.0.1 client > /etc/hosts
reboot
|
cd /tftpboot; ftp -i ftp.se.netbsd.org (log in as anonymous...) cd /pub/NetBSD/NetBSD-2.0.2/sgimips/binary/sets mget base.tgz comp.tgz etc.tgz games.tgz man.tgz misc.tgz text.tgz quit sh for a in *.tgz; do echo $a; tar zxfp $a; done echo 10.0.0.2:/tftpboot / nfs rw 0 0 > /tftpboot/etc/fstab echo rc_configured=YES >> /tftpboot/etc/rc.conf dd if=/dev/zero of=swap bs=1024 count=32768 halt |
ftp://ftp.NetBSD.org/pub/NetBSD/NetBSD-2.0.2/sgimips/binary/kernel/netbsd-GENERIC32_IP3x.gz ftp://ftp.NetBSD.org/pub/NetBSD/NetBSD-2.0.2/sgimips/binary/kernel/netbsd-INSTALL32_IP3x.gz
!!gxemul
!
! Configuration file for running NetBSD/sgimips diskless with
! a NetBSD/pmax machine as the nfs server.
emul(
net(
add_remote("127.0.0.1:12444") ! the server
local_port(12445) ! the client
)
machine(
name("client machine")
serial_nr(1)
type("sgi")
subtype("o2")
load("netbsd-INSTALL32_IP3x.gz")
! load("netbsd-GENERIC32_IP3x.gz")
)
)
|
!!gxemul
emul(
net(
local_port(12444) ! the server
add_remote("127.0.0.1:12445") ! the client
)
machine(
name("nfs server")
serial_nr(2)
type("dec")
subtype("5000/200")
disk("nbsd_pmax.img")
)
)
|
in one xterm: $ gxemul @config_server and then, in another xterm: $ gxemul @config_client
ifconfig mec0 10.0.0.1; route add default 10.0.0.254 mount -v 10.0.0.2:/tftpboot /mnt cd /mnt/dev; ./MAKEDEV all; cd /; umount /mnt haltThen log in as root on the server machine and type reboot.
You might want to log in as root on the server machine, and run tcpdump -lnvv or similar, to see that what the client machine actually does on the network.
It should now be possible to boot NetBSD/sgimips using the NetBSD/pmax nfs server, using the following commands: (NOTE! Execute these in two separate xterms!)
$ gxemul @config_server $ gxemul @config_client
When asked for "root device:" etc. on the clientmachine, enter the following values:
root device: mec0 dump device: (leave blank) file system (default generic): (leave blank) .. init path (default /sbin/init): (leave blank) Enter pathname of shell or RETURN for /bin/sh: (leave blank) Terminal type? [unknown] xterm .. # exit (to leave the single-user shell)
Note: Netbooting like this is very slow, so you need a lot of patience. For example, when NetBSD says "nfs_boot: trying DHCP/BOOTP", there will be a long pause, even on a very fast host machine. The reason for this is mostly because the emulator doesn't deal with timing issues very well, but also because NetBSD tries IPv6 first, before falling back to IPv4.
OpenBSD/pmax died at release 2.8 of OpenBSD, so you should be aware of the fact that this will not give you an up-to-date OpenBSD system.
Following these instructions might work. If not, then use common sense and imagination to modify them as you see fit.
$ dd if=/dev/zero of=obsd_pmax.img bs=1 count=512 seek=900000000
$ wget -r ftp://ftp.se.openbsd.org/pub/OpenBSD/2.8/pmax/
$ mv ftp.se.openbsd.org/pub/OpenBSD/2.8/pmax/simpleroot28.fs.gz . $ gunzip simpleroot28.fs.gz $ chmod +w simpleroot28.fs <--- make sure
$ mkisofs -o openbsd_pmax_2.8.iso ftp.se.openbsd.org/pub/OpenBSD/2.8/pmax
$ gxemul -E dec -e 3max -d obsd_pmax.img -d b:simpleroot28.fs -j bsd -d c:openbsd_pmax_2.8.iso(If you add -X, you will run with the graphical framebuffer. This is REALLY slow because the console has to scroll a lot during the install. I don't recommend it.)
$ fsck /dev/rz1a (and mark the filesystem as clean) $ mount /dev/rz1a / $ mount -t kernfs kern kern $ ./installand proceed with the install. Good luck. :-)
$ gxemul -E dec -e 3max -d obsd_pmax.img -d 2c:openbsd_pmax_2.8.iso -j bsd -o '-s'While you are at it, you might want to extract the X11 install sets as well, as the installer seems to ignore them too. (Perhaps due to a bug in the installer, perhaps because of the way I used mkisofs.)
Execute the following commands in the emulator:
fsck /dev/rz0a mount / passwd root cd /; mount -t cd9660 /dev/rz2c /mnt; sh for a in /mnt/[xX]*; do tar zxvf $a; done ln -s /usr/X11R6/bin/Xcfbpmax /usr/X11R6/bin/X ln -s /dev/fb0 /dev/mouse echo /usr/X11R6/lib >> /etc/ld.so.conf ldconfig sync halt
NOTE: It is also possible to install via ftp instead of using a CDROM image. This is not much less awkward, you still need the simpleroot filesystem image, and you still have to manually add the X11 install sets and set the root password, and so on.
Once you have completed the installation procedure, the following command will let you boot from the new rootdisk image:
$ gxemul -E dec -e 3max -X -M64 -o '-aN' -d obsd_pmax.img -j bsd
(Normally, you would be asked about which root device to use (rz0), but using -o '-aN' supresses that.)
When asked for which terminal type to use, when logging in as root,
enter rcons if you are using the graphical framebuffer,
vt100 for text-mode.
Use startx to start X windows.
(You should be aware of the fact that OpenBSD for the ARC platform died at release 2.3, so this will not give you an up-to-date OpenBSD system. See http://www.openbsd.org/arc.html for more information.)
To install OpenBSD/arc onto an emulated harddisk image, follow these instructions:
$ dd if=/dev/zero of=obsd_arc.img bs=1024 count=1 seek=700000
$ wget -np -l 0 -r ftp://ftp.se.openbsd.org/pub/OpenBSD/2.3/arc/
$ mkisofs -o openbsd_arc_2.3.iso ftp.se.openbsd.org/pub/OpenBSD/
$ gxemul -X -E arc -e pica -d obsd_arc.img -d b:openbsd_arc_2.3.iso -j 2.3/arc/bsd.rdand proceed like you would do if you were installing OpenBSD on a real Acer PICA-61. (Answer 'no' when asked if you want to configure networking, and then install from CD-ROM.)
Once the install has finished, the following command should let you boot from the harddisk image:
$ gxemul -X -E arc -e pica -d obsd_arc.img ftp.se.openbsd.org/pub/OpenBSD/2.3/arc/bsd
The following instructions should let you install Ultrix onto a disk image:
$ dd if=/dev/zero of=rootdisk.img bs=1024 count=1 seek=800000
$ gxemul -X -M64 -E dec -e 3max -d rootdisk.img -d bc:/dev/cd0c -j vmunix
$ gxemul -X -M64 -E dec -e 3max -d rootdisk.img -d c:/dev/cd0c -j vmunix
When the installation is completed, the following command should start Ultrix from the harddisk image:
$ gxemul -X -M64 -E dec -e 3max -j vmunix -d rootdisk.img
Ultrix mostly seems to work with dynamic binary translation (which can be disabled by the -B command line option). If you have a very fast host machine, and use bintrans, you might experience a weird timer related bug, which makes it impossible to logon to the system. It is triggered when the emulation goes faster than any real DECstation machine was capable of running. A temporary workaround is to add -I33000000 to fix the emulated clock speed to 33 million instructions per emulated second. (When using -CR4400, -I16000000 should be used instead.)
You can experiment with adding -Z2 (for emulating a dual-headed workstation) or even -Z3 (tripple-headed), and also the -Y2 option for scaling down the framebuffer windows by a factor 2x2. There is also a -z option for supplying names of X11 displays to use. The following example starts Ultrix on an emulated tripple-headed workstation, on three different displays (remote1:0.0, localhost:0.0, and remote2:0.0), using no scaledown:
$ gxemul -M64 -N -E dec -e 3max -jgenvmunix -d rootdisk.img \
-XZ3 -z remote1:0.0 -z localhost:0.0 -z remote2:0.0
The photo below shows a single Ultrix session running tripple-headed in GXemul on an Alpha 21164PC, with displays on a Sun Ultra1 (to the left), on the Alpha itself (in the middle), and on an HP700/RX X-terminal (8-bit color depth, running off the Alpha) to the right.
The X11 displays may differ in bit depth and endianness. Unfortunately, there is no way yet to set the scaledown factor on a per-window basis, so the scaledown factor affects all windows.
(If you didn't use -Zn during the installation, and compiled your own /vmunix, then it will not contain support for multiple graphics cards. To overcome this problem, use the generic kernel, -j genvmunix, whenever you are running the emulator with a different setup than the one you used when Ultrix was installed.)
A note for the historically interested: OSF/1 for MIPS was quite similar to Ultrix, so that is possible to run as well. If you are unsuccessful in installing Ultrix or OSF/1 directly in the emulator, you can always install it on your real machine onto a real SCSI disk, and then copy the contents of that SCSI disk into a file (using dd(1)), and use that file as a disk image file in the emulator.
The following instructions should let you run Sprite in the emulator:
ftp://ftp.es.embnet.org/pub/misc/TUHS/other_os/Sprite/ds5000.bt 83973120 bytes, MD5 = ec84eeeb20fe77b758370d5e312e4a5e
$ gxemul -X -E dec -e 3max -M128 -d ds5000.bt -j vmsprite -o ''
The first time you boot up with the disk image, you will be asked a number of questions regarding network settings. If you feel like entering correct values, then you should use the following:
Your machine's Ethernet address: 10:20:30:00:00:10
Your machine's IP: 10.0.0.1
Subnet mask: 0xff000000
Gateway's Ethernet address: 60:50:40:30:20:10
Gateway's IP: 10.0.0.254
Unfortunately, at the end of ftp://ftp.es.embnet.org/pub/misc/TUHS/other_os/Sprite/boot.txt, the following sad statement can be found:
The bootable Sprite image is meant to be a demonstration of Sprite, not
a robust Sprite system. There are several missing things, such as
floating point and network support.
Once you are logged in as root, running xinit will start the X11 environment.
I think this has to do with interrupts from the serial controller. Hopefully using the -U command line option will reduce the risk for such crashes. (I haven't had time to come up with a clean solution to this yet; it feels like a buffer overflow in Linux' serial driver for the 5000/200, but it is also likely that it is a bug in GXemul.)
Everything runs extremely slow. Even if you have a very fast host machine, an install attempt can still take several hours!
The following steps should let you install Debian GNU/Linux for DECstation onto a harddisk image:
$ dd if=/dev/zero of=debian.img bs=1024 count=1 seek=2000000
http://ftp.egr.msu.edu/debian/dists/stable/main/installer-mipsel/current/images/r3k-kn02/boot.img
$ gxemul -E dec -e 3max -U -M64 -o 'console=ttyS3' -d debian.img -O boot.img(If you want to, you can try -X instead of -o 'console=ttyS3' on the command line. This will cause Linux to use the graphical framebuffer. Unfortunately, Linux does not seem to have a driver for the DZ11 keyboard controller yet, so you cannot interact with the system. You will see the penguin in the upper lefthand corner while booting, and nicely rendered Unicode characters, but that's about it.)
You need to enter some values during the installation procedure, for example network settings. The following should work:
DHCP: No, choose "Configure network manually" IP address: 10.0.0.1 Netmask: 255.0.0.0 Gateway: 10.0.0.254 Name server addresses: 10.0.0.254
$ gxemul -E dec -e 3max -U -M64 -o 'console=ttyS3' -d debian.imgNote: All these steps take a lot of time, so you will have plenty of time to drink lots of cups of coffee.
http://ftp.egr.msu.edu/debian/dists/stable/main/installer-mipsel/current/images/cdrom/vmlinux-2.4.27-r3k-kn02and boot Debian using the following command line:
$ gxemul -E dec -e 3max -U -M64 -o \ 'console=ttyS3 root=/dev/sda1 rw init=/bin/sh' \ -d debian.img vmlinux-2.4.27-r3k-kn02You'll enter single-user mode. You need to add a line to /etc/inittab, to enable logins via serial console.
sh-2.05b# echo 'T3:23:respawn:/sbin/getty -L ttyS3 9600 vt100' >> /etc/inittab sh-2.05b# echo 'ttyS3' >> /etc/securetty sh-2.05b# sync; umount / sh-2.05b# halt
The system should now be ready for everyday use.
Use this command to boot from the completely installed disk image:
$ gxemul -E dec -e 3max -U -M64 -o 'console=ttyS3' -d debian.img
[ UPDATE 2005-01-19: Kaj-Michael Lang noticed that the current CVS-version of linux-mips has support for keyboards now, on DECstation 5000/200, so it is possible to run Debian GNU/Linux with framebuffer/keyboard. (Add -X (or -XY2) and remove the console=ttyS3 option.) He has made a kernel available here: http://home.tal.org/~milang/o2/kernels/vmlinux-2.4.29-rc2-r3k-mipsel-decstation It has other problems (ethernet doesn't seem to work, for example), but at least it doesn't Oops that often. ]
The following steps should let you run Redhat Linux for DECstation in GXemul:
http://ftp.egr.msu.edu/debian/dists/stable/main/installer-mipsel/current/images/cdrom/vmlinux-2.4.27-r3k-kn02
ftp://ftp.uni-wuppertal.de/pub/linux/mips/ mipsel-linux/root/mipsel-root-20011216.tgz 19486676 bytes, md5 = 5bcb725c90209479cd7ead8ad0c4a414
/dev/root / nfs defaults 1 1 #/dev/sdc1 / ext2 defaults 1 1 none /proc proc defaults 0 0 none /dev/pts devpts mode=0622 0 0to
#/dev/root / nfs defaults 1 1 /dev/sda1 / ext2 defaults 1 1 none /proc proc defaults 0 0 none /dev/pts devpts mode=0622 0 0(Note sda1 instead of sdc1.)
$ gxemul -E dec -e 3max -U -M128 -o \ "console=ttyS3 root=/dev/sda1 ro" -d redhat.img vmlinux-2.4.27-r3k-kn02
If you need to boot into single user mode, try the following:
$ gxemul -E dec -e 3max -U -o "console=ttyS3 root=/dev/sda1 rw init=/bin/sh" \ -d redhat.img vmlinux-2.4.27-r3k-kn02
Redhat Linux on DECstation in R3000 mode should work fine with dynamic binary translation, but if things are buggy, it can be disabled by using the -B command line option.
NOTE: You can add -X and remove console=ttyS3 from the command line, if you want to use a graphical framebuffer. Unfortunately, Linux doesn't have support for keyboards on DECstation 5000/200 yet, so you cannot actually interact with the sytem. :-(
[ UPDATE 2005-01-22: Read the 2005-01-19 update in the Debian section above, and then, if you do not need ethernet support, try Kaj-Michael Lang's kernel compiled from linux-mips' CVS. http://home.tal.org/~milang/o2/patches/vmlinux-2.4.29-rc2-r3k-mipsel-decstation It should work with framebuffer/keyboard. ]
NOTE: Mach for DECstation requires some files (called 'startup' and 'emulator') which I haven't been able to find on the web. Without these, Mach will not get very far. These installation instructions are preliminary.
The following steps should let you experiment with running Mach for DECstation in the emulator:
$ ./configure --enable-caches; make
http://lost-contact.mit.edu/afs/athena/user/d/a/daveg/Info/Links/Mach/src/release/pmax.tar.Z
7263343 bytes, md5 = f9d76c240a6e169921a1df99ad560cc0
$ tar xfvz pmax.tar.Z pmax_mach/special/mach.boot.MK83.STD+ANY
$ dd if=/dev/zero of=disk.img bs=1 count=512 seek=400000000
$ gxemul -E dec -e 3max -X -d disk.img \ pmax_mach/special/mach.boot.MK83.STD+ANY
NOTE: I haven't succeeded all the way with this yet, and this shows/triggers many bugs in the emulator, but some of it works.
GXemul does not yet emulate the AHC PCI SCSI controller in the O2. (I have mailed Adaptec several times, asking for documentation, but never received any reply.) OpenBSD/sgi can still run, as long as it doesn't use SCSI. For a simple test with the ramdisk (install) kernel, try dowloading
ftp://ftp.se.openbsd.org/pub/OpenBSD/3.7/sgi/bsd.rdand run gxemul -E sgi -e o2 bsd.rd.
It might also be possible to netboot. Another emulated machine must then be used as the nfs root server, and the emulated O2 machine must boot as a diskless client. Performing this setup is quite time consuming, but necessary:
$ gxemul -M64 -Edec -e3max -d nbsd_pmax.imgand enter the following commands as root inside the emulator:
echo hostname=server >> /etc/rc.conf echo ifconfig_le0=\"inet 10.0.0.2\" >> /etc/rc.conf echo nameserver 10.0.0.254 >> /etc/resolv.conf echo 10.0.0.254 > /etc/mygate echo /tftpboot -maproot=root 10.0.0.1 > /etc/exports echo rpcbind=YES >> /etc/rc.conf echo nfs_server=YES >> /etc/rc.conf echo mountd=YES >> /etc/rc.conf echo bootparamd=YES >> /etc/rc.conf printf "client root=10.0.0.2:/tftpboot \\\n swap=10.0.0.2:/tftpboot/swap\n" > /etc/bootparams echo "10:20:30:00:00:10 client" > /etc/ethers echo 10.0.0.1 client > /etc/hosts reboot |
cd /tftpboot; ftp -i ftp.se.openbsd.org (log in as anonymous...) cd pub/OpenBSD/3.7/sgi mget b* c* e* g* m* quit sh for a in *.tgz; do echo $a; tar zxfp $a; done echo 10.0.0.2:/tftpboot / nfs rw 0 0 > /tftpboot/etc/fstab dd if=/dev/zero of=swap bs=1024 count=32768 halt |
ftp://ftp.se.openbsd.org/pub/OpenBSD/3.7/sgi/bsd MD5 (bsd) = f16eaf3dcbd51876db7c25f70e6d8a08 ftp://ftp.se.openbsd.org/pub/OpenBSD/3.7/sgi/bsd.rd MD5 (bsd.rd) = 4843e6139d8dd04b03d5f0e33e9a4f7b
!!gxemul
!
! Configuration file for running OpenBSD/sgi diskless with
! a NetBSD/pmax machine as the nfs server.
!
! This config file is for the client.
emul(
net(
add_remote("127.0.0.1:12444") ! the server
local_port(12445) ! the client
)
machine(
name("client machine")
serial_nr(1)
type("sgi")
subtype("o2")
! load("bsd")
load("bsd.rd")
)
)
|
!!gxemul
emul(
net(
local_port(12444) ! the server
add_remote("127.0.0.1:12445") ! the client
)
machine(
name("nfs server")
serial_nr(2)
type("dec")
subtype("5000/200")
disk("nbsd_pmax.img")
)
)
|
in one xterm: $ gxemul @config_server and then, in another xterm: $ gxemul @config_client
ifconfig mec0 10.0.0.1; route add default 10.0.0.254 mount -v 10.0.0.2:/tftpboot /mnt cd /mnt/dev; ./MAKEDEV all; cd /; umount /mnt halt |
You might want to log in as root on the server machine, and run tcpdump -lnvv or similar, to see that what the client machine actually does on the network. The MAKEDEV script takes almost forever, so be patient.
NOTE: Everything up to this point should work. However, the next step (in gray) doesn't actually work:
Once everything has been set up correctly, change bsd.rd in config_client to just bsd (the GENERIC kernel). It should now be possible to boot OpenBSD/sgi using the NetBSD/pmax nfs server. (When asked for "root device :" on the OpenBSD machine, enter mec0.)
But it doesn't work. Probably because GXemul's implementation of the mec (ethernet card used in the O2) is too much of a quick hack. For now, use the bsd.rd kernel, and (at every boot) type:
s (for Shell) ifconfig mec0 10.0.0.1; route add default 10.0.0.254 mount -v 10.0.0.2:/tftpboot /mnt cd /mnt; usr/sbin/chroot . sh etc/rc
This will not cause OpenBSD to be booted normally, but at least a few basic things will work. By the way, the emulator performs so poorly, that you will have time to fetch several cups of coffee for each of the steps above.