Minipython is a Xen-based unikernel to run Python scripts. It consists of MicroPython (https://github.com/micropython/micropython) running on top of the Xen paravirtualized Mini-OS operating system (https://wiki.xenproject.org/wiki/Mini-OS).
Clone the Xen sources, the toolstack, Mini-OS and Minipython:
$ git clone git://xenbits.xen.org/xen.git
$ git clone https://github.com/sysml/mini-os.git
$ git clone https://github.com/sysml/toolchain.git
$ git clone https://github.com/sysml/minipython.git
Check out a stable version of the Xen sources:
$ cd xen
$ git checkout stable-4.9
$ cd ..
Set up environment variables:
$ export XEN_ROOT=$(pwd)/xen
$ export TOOLCHAIN_ROOT=$(pwd)/toolchain
$ export MINIOS_ROOT=$(pwd)/mini-os
Prepare to build Minipython:
$ cd minipython/micropython
$ git submodule init
$ git submodule update
$ cd ..
Next, edit minipython's Makefile (found in minios/Makefile) and set the paths to the Xen sources, the Xen toolchain sources and the Mini-OS sources:
XEN_ROOT ?= $(realpath ../../xen)
TOOLCHAIN_ROOT ?= $(realpath ../../toolchain)
MINIOS_ROOT ?= $(realpath ../../mini-os)
There are a number of options at the top of that Makefile that you can also set. For instance, by default, networking via lwip is enabled.
To build minipython:
$ cd minios
$ make
To specify a Python script edit the run_script() function in minios/main.c . You have two main options: do_str(), which allows you to specify a string for your script directly in main.c, and do_file() which retrieves and executes a file from your filesystem.
The minipython VM can be simply run with
$ cd minios
$ xl create -c minipython.xen
Before this will work though we need to set up a filesystem and networking (the latter is only needed if the Makefile has networking enabled; by default it is). Please see instructions below.
Minipython uses FAT as its default filesystem type. To get you started, you can use the demo filesystem in these sources (filesystems/minipython-demo-fatfs.img) which contains a few basic scripts. First uncompress it with:
$ gzip -d minipython-demo-fatfs.img.gz
Then mount it:
$ losetup /dev/loop0 filesystems/minipython-demo-fatfs.img
$ kpartx -a /dev/loop0
$ mkdir -p /mnt/fat
$ mount /dev/mapper/loop0p1 /mnt/fat/
Finally specify a disk in minios/minipython.xen (the version of this file in these sources already has this line):
disk = ['phy:/dev/loop0,xvda,w']
[Use this only if you know what you're doing!]
First, set CONFIG_SHFS=y in minios/Makefile (make distclean; make if you had previously built the image). Then add a disk to minios/minipython.xen:
disk = [ 'file:/path/to/your/volume/minipython-vol.shfs,xvda,w']
To create minipython-vol.shfs, in the minios/shfs/shfs-tools/ subdirectory run make, then:
./mkwebfs /path/to/your/python/files minipython-vol.shfs
Get the signature of the file you want to run with:
./shfs_admin -l minipython-vol.shfs
Set the default file flag with
/shfs_admin -d [signature] minipython-vol.shfs
First, create a bridge with brctl and bring it up:
$ brctl addbr minipythonbr
$ ifconfig minipythonbr 192.168.0.1 netmask 255.255.255.0 up
Then, add a vif to minios/minipython.xen:
vif = [ 'mac=00:11:22:33:44:55,bridge=minipythonbr,ip=192.168.0.100' ]
To test it out you can run the HTTP server script provided with the minipython sources:
$ cp minios/examples/http_server.py /mnt/fat
Add the line to main.c's run_script() function:
do_file("http_server.py");
Run make and xl create -c minipython.xen. You can test that it is working by running the following command on the host (i.e., dom0):
$ wget 192.168.0.100:8080 --no-proxy
To add https://github.com/micropython/micropython-lib to minipython, edit minos/tools/libimport.py and set the MICROPY_LIB_DIR and MINIPYTHON_TARGET_DIR variables, then
$ python minios/tools/libimport.py
The program will print which modules it actually copied (it ignores placeholder libraries, i.e., those with empty .py files).
The output under "Added libs" (a Python array) can be copied into minios/examples/test_tryexcept.py to see which modules will actually run under minipython.