[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

etch-m68k in a linux container

Hi guys,

as you seem ready to play with simulators to build debian m68k packages,
I think it should be a good idea to play with the linux-user mode of
qemu. Used with the linux containers, you can run real m68k users binary
on an x86 kernel (it means gigabytes of memory, several CPU cores,
Terabytes of SATA disks, gigabits ethernet, hundreds of virtual m68k
machines on one host).

qemu linux-user mode traps guest (m68k) syscalls to translate them to
native ones. It's not perfect. For instance, the clone() syscall seems
to not work correctly (at least in one case, gnome-terminal), but you
will, at least, be able to rebuild some packages. I don't think it is
really usable for a debian buildd, but it can be used to develop and
correct build issues. For instance, you can build glibc, kernel and gcc
concurrently on one machine in less than a day ;-)

The attached script will help you to test this. This script :
- checks all needed stuffs are there (I'm not sure of the full list)
- clones my qemu-m68k fork and compiles it
- debootstrap and configures an etch-m68k root filesystem
- configure binfmt to use qemu-m68k to load m68k ELF binaries
- creates an LXC container for it

run it with "sudo ./create-m68k-lxc.sh".

I have tested it on an ubuntu 12.10 x86_64 system, and it seems to work

Default installation paths are:
- /containers/m68k for the m68k root filesystem
- $HOME/qemu-m68k for qemu sources
- $HOME/lxc-m68k.conf for the LXC configuration file

The name of the LXC container is virtm68k, but you can edit variables at
the beginning of the script.

Then you can start the container with "sudo lxc-start -n virtm68k", the
system console is on the stdio. You can access the linux virtual
consoles by running several "sudo lxc-console -n virtm68k".

By default, on an Ubuntu 12.10 system, when lxc is installed, there is
an interface "lxcbridge" allowing the virtual machine to access to
network. By default the bridge is and the virtm68k machine is
configured with an eth0 on "openssh-server" is not installed
by default, but you can add it from the console with apt-get (don't
forget to add a password for root).

If you don't like this, you can destroy the container with "sudo
lxc-destroy -n virtm68k", but, WARNING, it will also remove the
directory !

Have fun,

Note1: "sudo" will not work inside the container as "qemu-m68k" has not
the SUID bit. Just log on on an alternate console as root.

Note2: you can mount your host home directory using "mount
-b /home /container/m68k/home", but don't forget to create your user

Note3: and for those that want to know the real power of their new m68k
machine, I've added in attachement bogomips.c to compute the bogomips
and the equivalent 040 cpu frequency. Mine (on a Q6600 a 2.4 Ghz) is:
$ ./bogomips 
Clocking:	132
BogoMips:	104.00
Calibration:	524288

Attachment: create-m68k-lxc.sh
Description: application/shellscript

#include <stdio.h>
#include <time.h>

 * This is the number of bits of precision for the loops_per_jiffy.  Each
 * bit takes on average 1.5/CLOCKS_PER_SEC seconds.  This (like the original) is a little
 * better than 1%

#define LPS_PREC 8

static inline void __delay(unsigned long loops)
	__asm__ __volatile__ ("1: subql #1,%0; jcc 1b"
		: "=d" (loops) : "0" (loops));

int main(void)
	unsigned long ticks, loopbit;
	int lps_precision = LPS_PREC;
	unsigned long loops_per_jiffy;
	unsigned long bogomips;

	loops_per_jiffy = (1<<12);

	while ((loops_per_jiffy <<= 1) != 0) {
		ticks = clock();
		while (ticks == clock()) ;
		ticks = clock();
		ticks = clock() - ticks;
		if (ticks)

	/* Round the value and print it */
	bogomips = (loops_per_jiffy * 2) / ticks * CLOCKS_PER_SEC;
	printf("Clocking:\t%1g\n", (double)bogomips * 33 / 26 / 1000000 );
		bogomips / 1000000, (bogomips / 10000 ) % 100);
	printf("Calibration:\t%ld\n", loops_per_jiffy);

	return 0;

Reply to: