Documentation and guides

Hello MirageOS World

First make sure you have followed the installation instructions to get a working MirageOS installation. The examples below are in the mirage-skeleton repository. Begin by cloning and changing directory to it:

$ git clone git://github.com/mirage/mirage-skeleton.git
$ cd mirage-skeleton

The mirage-skeleton repository classifies its examples into three groups. This document refers to unikernels in the tutorial directory.

Note: Before we begin, if you aren't familiar with the Lwt library (and the >>= operator it provides), you may want to read at least the start of the Lwt tutorial first.

Additional note: Throughout the tutorial, we'll use mirage configure -t unix to demonstrate building MirageOS applications. If you're using a Mac OS X machine, you should use mirage configure -t macosx instead.

Step 0: Doing Nothing!

Before we try and do anything complicated, let's do nothing briefly. That is, let's build a unikernel that simply starts and then exits -- nothing else. The code for this is, as you might hope, fairly short. First the unikernel itself:

$ cat tutorial/noop/unikernel.ml
let start =
  Lwt.return_unit

So this is a unikernel whose entry point (start) does nothing other than return an Lwt thread that will evaluate to unit. Easy.

Before we can build even our noop unikernel, we must define its configuration. That is, we need to tell Mirage what OCaml module contains the start entry point. We do this by writing a config.ml file that sits next to our unikernel.ml file:

$ cat tutorial/noop/config.ml
open Mirage

let main =
  foreign "Unikernel" job

let () =
  register "noop" [main]

There's a little more going on here than in unikernel.ml. First we open the Mirage module to save on typing. Next, we define a value main (named so by convention because, at heart, some of us are still C programmers -- feel free to call it something else if you wish!) which calls the foreign function passing two parameters. The first is a string declaring the module name that contains our entry point-- in this case, standard OCaml compilation behaviour means that the unikernel.ml file produces a module named Unikernel. Again, there's nothing special about this name -- if you want to sue something else here, simply rename unikernel.ml accordingly.

The second parameter, job, is a bit more interesting. This declares the type of our unikernel in terms of the devices (that is, things such as network interfaces, network stacks, filesystems and so on) it requires to operate. As this is a unikernel that does nothing, it needs no devices and so is simply job.

Finally, we declare the entry point to OCaml in the usual way (let () = ...), registering our unikernel entry point (main) with a name ("noop" in this case) to be used when we build our unikernel.

To build our unikernel is then simply a matter of evaluating its configuration:

$ cd tutorial/noop
/Users/mort/research/projects/mirage/src/mirage-skeleton/tutorial/noop
$ mirage configure -t unix

...installing dependencies:

$ make depend
opam pin add --no-action --yes mirage-unikernel-noop-unix .
[NOTE] Package mirage-unikernel-noop-unix is already path-pinned to /Users/mort/research/projects/mirage/src/mirage-skeleton/noop.
       This will erase any previous custom definition.
Proceed ? [Y/n] y

[mirage-unikernel-noop-unix] /Users/mort/research/projects/mirage/src/mirage-skeleton/noop/ synchronized
[mirage-unikernel-noop-unix] Installing new package description from /Users/mort/research/projects/mirage/src/mirage-skeleton/noop

opam depext --yes mirage-unikernel-noop-unix
# Detecting depexts using flags: x86_64 osx homebrew
# The following system packages are needed:
#  - camlp4
# All required OS packages found.
opam install --yes --deps-only mirage-unikernel-noop-unix

=-=- Synchronising pinned packages =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=  🐫
[mirage-unikernel-noop-unix] /Users/mort/research/projects/mirage/src/mirage-skeleton/noop/ already up-to-date
opam pin remove --no-action mirage-unikernel-noop-unix
mirage-unikernel-noop-unix is now unpinned from path /Users/mort/research/projects/mirage/src/mirage-skeleton/noop

...and compiling:

$ make
mirage build
ocamlfind ocamldep -package mirage-unix -package mirage-types-lwt -package mirage-types -package mirage-runtime -package mirage-logs -package mirage-clock-unix -package lwt -package functoria-runtime -predicates mirage_unix -modules main.ml > main.ml.depends
ocamlfind ocamldep -package mirage-unix -package mirage-types-lwt -package mirage-types -package mirage-runtime -package mirage-logs -package mirage-clock-unix -package lwt -package functoria-runtime -predicates mirage_unix -modules key_gen.ml > key_gen.ml.depends
ocamlfind ocamldep -package mirage-unix -package mirage-types-lwt -package mirage-types -package mirage-runtime -package mirage-logs -package mirage-clock-unix -package lwt -package functoria-runtime -predicates mirage_unix -modules unikernel.ml > unikernel.ml.depends
ocamlfind ocamlc -c -g -g -bin-annot -safe-string -principal -strict-sequence -package mirage-unix -package mirage-types-lwt -package mirage-types -package mirage-runtime -package mirage-logs -package mirage-clock-unix -package lwt -package functoria-runtime -predicates mirage_unix -w A-4-41-42-44 -color always -o key_gen.cmo key_gen.ml
ocamlfind ocamlc -c -g -g -bin-annot -safe-string -principal -strict-sequence -package mirage-unix -package mirage-types-lwt -package mirage-types -package mirage-runtime -package mirage-logs -package mirage-clock-unix -package lwt -package functoria-runtime -predicates mirage_unix -w A-4-41-42-44 -color always -o unikernel.cmo unikernel.ml
ocamlfind ocamlc -c -g -g -bin-annot -safe-string -principal -strict-sequence -package mirage-unix -package mirage-types-lwt -package mirage-types -package mirage-runtime -package mirage-logs -package mirage-clock-unix -package lwt -package functoria-runtime -predicates mirage_unix -w A-4-41-42-44 -color always -o main.cmo main.ml
ocamlfind ocamlopt -c -g -g -bin-annot -safe-string -principal -strict-sequence -package mirage-unix -package mirage-types-lwt -package mirage-types -package mirage-runtime -package mirage-logs -package mirage-clock-unix -package lwt -package functoria-runtime -predicates mirage_unix -w A-4-41-42-44 -color always -o key_gen.cmx key_gen.ml
ocamlfind ocamlopt -c -g -g -bin-annot -safe-string -principal -strict-sequence -package mirage-unix -package mirage-types-lwt -package mirage-types -package mirage-runtime -package mirage-logs -package mirage-clock-unix -package lwt -package functoria-runtime -predicates mirage_unix -w A-4-41-42-44 -color always -o unikernel.cmx unikernel.ml
ocamlfind ocamlopt -c -g -g -bin-annot -safe-string -principal -strict-sequence -package mirage-unix -package mirage-types-lwt -package mirage-types -package mirage-runtime -package mirage-logs -package mirage-clock-unix -package lwt -package functoria-runtime -predicates mirage_unix -w A-4-41-42-44 -color always -o main.cmx main.ml
ocamlfind ocamlopt -g -linkpkg -g -package mirage-unix -package mirage-types-lwt -package mirage-types -package mirage-runtime -package mirage-logs -package mirage-clock-unix -package lwt -package functoria-runtime -predicates mirage_unix key_gen.cmx unikernel.cmx main.cmx -o main.native

As we configured for UNIX (the -t unix argument to the mirage configure command), the result is a standard UNIX ELF binary that can simply be executed:

$ ls -l noop
lrwxrwxr-x  1 mort  staff  18 Jan 12 12:11 noop@ -> _build/main.native
$ ls -l _build/main.native
-rwxrwxr-x  1 mort  staff  2690564 Jan 12 12:11 _build/main.native*
$ ./noop
$ echo $?
0

And that's it -- you've just built and run your very first unikernel!

Aside: Doing Nothing with Functors!

Functors are one of those OCaml things that can seem a bit intimidating at first (traditionally, "monads" get the same sort of reaction). However, as they're used fairly widely throughout Mirage, a very brief introduction to the commonest way we use them is needed. For a better introduction as to how to actually make use of them, what they are, and so on see Real World OCaml, Ch.9 (and probably also Ch.10, First-Class Modules.

In short, in Mirage, they're used as a way to abstract over the target environment for the unikernel. Functors are, roughly, functions from modules to modules, and they allow us to pass modules into a unikernel so that the code inside unikernel can interact with its environment (read files, send packets, etc) without needing to care whether its been built to target UNIX, Xen, KVM, or something else entirely. The modules that are passed into the unikernel in this way are required to conform to type signatures that are specified when the unikernel job value is created in the config.ml file.

We'll see several examples of this below but, for now, we need to wrap up our noop unikernel in a module inside the Unikernel module so that we can use it as a functor. This is actually quite straightforward -- we simply wrap the start function in unikernel.ml inside some module. For example,

$ cat tutorial/noop-functor/unikernel.ml
module Main = struct

  let start =
    Lwt.return_unit

end

The use of the name Main is purely convention -- again, call it something completely different if you wish to put C programming firmly behind you!

The only other change is to the corresponding invocation in config.ml:

$ cat tutorial/noop-functor/config.ml
open Mirage

let main =
  foreign "Unikernel.Main" job

let () =
  register "noop" [main]

Note that the string passed to foreign is now "Unikernel.Main" as we must refer to the Main module inside the Unikernel module. Everything else stays the same-- go ahead and try that out by building the unikernel inside noop-functor.

...and now, onwards to unikernels that actually do something!

Step 1: Hello World!

As a first step, let's build and run the MirageOS "Hello World" unikernel -- this will print a log message with the word hello 4 times before terminating:

2017-02-08 09:54:44 -01:00: INF [application] hello
2017-02-08 09:54:45 -01:00: INF [application] hello
2017-02-08 09:54:46 -01:00: INF [application] hello
2017-02-08 09:54:47 -01:00: INF [application] hello

First, let's look at the code:

$ cat hello/unikernel.ml
open Lwt.Infix

module Hello (Time : Mirage_time_lwt.S) = struct

  let start _time =

    let rec loop = function
      | 0 -> Lwt.return_unit
      | n ->
        Logs.info (fun f -> f "hello");
        Time.sleep_ns (Duration.of_sec 1) >>= fun () ->
        loop (n-1)
    in
    loop 4

end

To veteran OCaml programmers among you, this might look a little odd: we have a Main module parameterised a module (Time, of type Mirage_time_lwt.S) that contains a method start taking an ignored parameter _time (an instance of a time). This is the basic structure required to make this a MirageOS unikernel rather than a standard OCaml POSIX application.

The module type for our Time module, Mirage_time_lwt.S, is defined in an external package mirage-time. The name S for "the module type of things like this" is a common OCaml convention (comparable to naming the most-used type in a module t). There are many packages defining module types for use in Mirage. For ease of discovery, a list of the module types that Mirage knows about is maintained in the types/ directory of the main MirageOS repository. The Mirage_types module gives abstract definitions that leave some important primitives unspecified; the Mirage_types_lwt module contains more concrete definitions for use in programs. Since you'll find yourself referring back to these quite often when building MirageOS applications, it's worth bookmarking the documentation for this module.

The concrete implementation of Time will be supplied at compile-time, depending on the target that you are compiling for. This configuration is stored in config.ml, so let's take a look:

$ cat tutorial/hello/config.ml

open Mirage

let main =
  foreign
    ~packages:[package "duration"]
    "Unikernel.Hello" (time @-> job)

let () =
  register "hello" [main $ default_time]

The configuration file is a normal OCaml module that calls register to create one or more jobs, each of which represent a process (with a start/stop lifecycle). Each job most likely depends on some device drivers; all the available device drivers are defined in the Mirage module (see the Mirage module documentation).

In this case, the main variable declares that the entry point of the process is the Main module from the file unikernel.ml. The @-> combinator is used to add a device driver to the list of functor arguments in the job definition (see unikernel.ml), and the final value of using this combinator should always be a job if you intend to register it.

The foreign function also takes some additional arguments: ~keys, the list of configuration keys we want to allow the user to specify at configuration or build time, and packages, a list of additional opam packages that should be included in the list of build dependencies for the project. We'll talk more about configuration keys in the next example.

Notice that we refer to the module name as a string ("Unikernel.Main") when calling foreign, instead of directly as an OCaml value. The mirage command-line tool evaluates this configuration file at build-time and outputs a main.ml that has the concrete values filled in for you, with the exact modules varying by which backend you selected (e.g. Unix or Xen).

MirageOS mirrors the unikernel model on UNIX as far as possible: your application is built as a unikernel which needs to be instantiated and run whether on UNIX or on Xen. When your unikernel is run, it starts much as a VM on Xen does -- and so must be passed references to devices such as the console, network interfaces and block devices on startup.

In this case, this simple hello world example requires some notion of time, so we register a single Job consisting of the Hello.Main module (and, implicitly its start function) and passing it references to a timer.

Building a Unix binary

We invoke all this by configuring, building and finally running the resulting unikernel under Unix first.

$ cd tutorial/hello
$ mirage configure -t unix

mirage configure generates a Makefile with all the build rules included from evaluating the configuration file, a main.ml that represents the entry point of your unikernel, and an opam file with a list of the packages necessary to build the unikernel.

$ make depend

In order to automatically install the dependencies discovered by mirage configure in your current opam switch, execute make depend.

$ make

This builds a UNIX binary called console that contains the simple console application. If you are on a multicore machine and want to do parallel builds, export OPAMJOBS=4 (or some other value equal to the number of cores) will do the trick.

Finally to run your application, as it is a standard Unix binary, simply run it directly and observe the exciting log messages that our for loop is generating:

$ ./hello

Building for Another Backend

Note: The following sections of this tutorial use the ukvm backend as an example - this backend is specific to Linux host systems with direct access to KVM via /dev/kvm. If KVM isn't available on your machine, or you are using FreeBSD, you may be interested in building and running unikernels via the virtio target which provides support for more hypervisors such as KVM/QEMU, bhyve and Google Compute Engine.

To make a unikernel that will use Solo5 to run on KVM, re-run mirage configure and ask for the ukvm target instead of unix.

$ mirage configure -t ukvm
$ make depend
$ make

Everything else remains the same! The set of dependencies required, the main.ml, and the Makefile differ significantly, but since the source code of your application was parameterised over the Time type, it doesn't matter-- you do not need to make any changes for your code to run when linked against the Solo5 console driver instead of Unix.

When you build the ukvm version, you'll see some new artifacts: a ukvm-bin binary and a file called hello.ukvm. hello.ukvm is the unikernel, and ukvm-bin is a dynamically-generated monitor program specialised to your unikernel. To try running hello.ukvm, pass it as an argument to ukvm-bin:

$ ./ukvm-bin hello.ukvm
            |      ___|
  __|  _ \  |  _ \ __ \
\__ \ (   | | (   |  ) |
____/\___/ _|\___/____/
Solo5: Memory map: 512 MB addressable:
Solo5:     unused @ (0x0 - 0xfffff)
Solo5:       text @ (0x100000 - 0x1d8fff)
Solo5:     rodata @ (0x1d9000 - 0x20bfff)
Solo5:       data @ (0x20c000 - 0x2b3fff)
Solo5:       heap >= 0x2b4000 < stack < 0x20000000
Solo5: Clock source: KVM paravirtualized clock
Solo5: new bindings
STUB: getenv() called
2017-02-08 23:58:20 -00:00: INF [application] hello
2017-02-08 23:58:21 -00:00: INF [application] hello
2017-02-08 23:58:22 -00:00: INF [application] hello
2017-02-08 23:58:23 -00:00: INF [application] hello
Solo5: solo5_app_main() returned with 0

We get some additional output from the initialization of the unikernel and its successful boot, then we see our expected output, and Solo5's report of the application's successful completion.

Configuration Keys

It's very common to pass additional runtime information to a program via command-line options or arguments. But a unikernel doesn't have access to a command line, so how can we pass it runtime information?

Mirage provides a nice abstraction for this in the form of configuration keys. The Mirage module provides a module Key, which contains functions for creating and using configuration keys. For an example, let's have a look at hello-key:

$ cd tutorial/hello-key
$ cat config.ml
open Mirage

let key =
  let doc = Key.Arg.info ~doc:"How to say hello." ["hello"] in
  Key.(create "hello" Arg.(opt string "Hello World!" doc))

let main =
  foreign
    ~keys:[Key.abstract key]
    ~packages:[package "duration"]
    "Unikernel.Hello" (time @-> job)

let () =
  register "hello" [main $ default_time]

We create a key with Key.create which is an optional bit of configuration. It will default to "Hello World!" if unspecified. This particular key happens to be of type string, so no conversion will be required, but it's possible to ask for more exotic types in the call to Arg -- see the Functoria Key.Arg module documentation for more details.

Once we've created our configuration key, we specify that we'd like it available in the unikernel by passing it to foreign in the keys parameter.

We can then read the value corresponding to configuration key using the generated function Key_gen.hello as shown below.

$ cat unikernel.ml
open Lwt.Infix

module Hello (Time : Mirage_time_lwt.S) = struct

  let start _time =

    let hello = Key_gen.hello () in

    let rec loop = function
      | 0 -> Lwt.return_unit
      | n ->
        Logs.info (fun f -> f "%s" hello);
        Time.sleep_ns (Duration.of_sec 1) >>= fun () ->
        loop (n-1)
    in
    loop 4

end

Let's configure the example for UNIX and build it:

$ mirage configure -t unix
$ make depend
$ make

When the target is Unix, Mirage will use an implementation for configuration keys that looks at the contents of OS.Env.argv -- in other words, it looks directly at the command line that was used to invoke the program. If we call hello with no arguments, the default value is used:

./hello
2017-02-08 18:18:23 -03:00: INF [application] Hello World!
2017-02-08 18:18:24 -03:00: INF [application] Hello World!
2017-02-08 18:18:25 -03:00: INF [application] Hello World!
2017-02-08 18:18:26 -03:00: INF [application] Hello World!

but we can ask for something else:

./hello --hello="Bonjour!"
$ ./hello --hello="Bonjour!"
2017-02-08 18:20:46 +09:00: INF [application] Bonjour!
2017-02-08 18:20:47 +09:00: INF [application] Bonjour!
2017-02-08 18:20:48 +09:00: INF [application] Bonjour!
2017-02-08 18:20:49 +09:00: INF [application] Bonjour!

When the target is Unix, it's also possible to get useful hints by calling the generated program with --help.

Many configuration keys can be specified either at configuration time or at run time. mirage configure will allow us to change the default value for hello, while retaining the ability to override it at runtime:

$ mirage configure -t unix --hello="Hola!"
$ make depend
$ make
$ ./hello
2017-02-08 18:30:30 +06:00: INF [application] Hola!
2017-02-08 18:30:31 +06:00: INF [application] Hola!
2017-02-08 18:30:32 +06:00: INF [application] Hola!
2017-02-08 18:30:33 +06:00: INF [application] Hola!
$ ./hello --hello="Hi!"
2017-02-08 18:30:54 +06:00: INF [application] Hi!
2017-02-08 18:30:55 +06:00: INF [application] Hi!
2017-02-08 18:30:56 +06:00: INF [application] Hi!
2017-02-08 18:30:57 +06:00: INF [application] Hi!

When configured for non-Unix backends, other mechanisms are used to pass the runtime information to the unikernel. ukvm-bin, which we used to run hello.ukvm in the non-keyed example, will pass keys specified on the command line to the unikernel when invoked:

$ cd tutorial/hello-key
$ mirage configure -t ukvm
$ make depend
$ make
$ ./ukvm-bin hello.ukvm --hello="Hola!"
            |      ___|
  __|  _ \  |  _ \ __ \
\__ \ (   | | (   |  ) |
____/\___/ _|\___/____/
Solo5: Memory map: 512 MB addressable:
Solo5:     unused @ (0x0 - 0xfffff)
Solo5:       text @ (0x100000 - 0x1d8fff)
Solo5:     rodata @ (0x1d9000 - 0x20bfff)
Solo5:       data @ (0x20c000 - 0x2b3fff)
Solo5:       heap >= 0x2b4000 < stack < 0x20000000
Solo5: Clock source: KVM paravirtualized clock
Solo5: new bindings
STUB: getenv() called
2017-02-09 00:26:00 -00:00: INF [application] Hola!
2017-02-09 00:26:01 -00:00: INF [application] Hola!
2017-02-09 00:26:02 -00:00: INF [application] Hola!
2017-02-09 00:26:03 -00:00: INF [application] Hola!
Solo5: solo5_app_main() returned with 0

Step 2: Getting a block device

Most useful unikernels will need to obtain data from the outside world, so we'll explain this subsystem next.

Sector-addressible block devices

The device-usage/block/ directory in mirage-skeleton contains an example of attaching a raw block device to your unikernel. The Mirage_block_lwt interface signature contains the operations that are possible on a block device: primarily reading and writing aligned buffers to a 64-bit offset within the device.

On Unix, the development workflow to handle block devices is by mapping them onto local files. The config.ml for the block example contains some logic for automatically creating a disk image file (and removing it when mirage clean is called), in addition to a more familiar-looking set of calls to foreign and register:

open Mirage

type shellconfig = ShellConfig
let shellconfig = Type ShellConfig

let config_shell = impl @@ object
    inherit base_configurable

    method build _i =
      Bos.OS.Cmd.run Bos.Cmd.(v "dd" % "if=/dev/zero" % "of=disk.img" % "count=100000")

    method clean _i =
      Bos.OS.File.delete (Fpath.v "disk.img")

    method module_name = "Functoria_runtime"
    method name = "shell_config"
    method ty = shellconfig
end


let main =
  let packages = [ package "io-page"; package "duration"; package ~build:true "bos"; package ~build:true "fpath" ] in
  foreign
    ~packages
    ~deps:[abstract config_shell] "Unikernel.Main" (time @-> block @-> job)

let img = block_of_file "disk.img"

let () =
  register "block_test" [main $ default_time $ img]

The main binding looks much like the earlier hello example, except for the addition of a block device in the list. When we register the job, we supply a block device from a local file via block_of_file.


As an aside, if you have your editor configured with OCaml mode, you should be able to see the inferred types for some of the variables in the configuration file. The @-> and $ combinators are designed such that any mismatches in the declared device driver types and the concrete registered implementations will result in a type error at configuration time.

Build this on Unix in the same way as the previous examples:

$ cd device-usage/block
$ mirage configure -t unix
$ make depend
$ make
$ ./block_test

block_test will write a series of patterns to the block device and read them back to check that they are the same (the logic for this is in unikernel.ml within the Block_test module).

We can build this example for another backend too:

$ mirage configure -t ukvm
$ make depend
$ make

Now we just need to boot the unikernel with ukvm-bin as before, and we should see the same output (after the VM boot preamble) -- but now MirageOS is linked against the Solo5 block device driver and is mapping the unikernel's block requests directly through to it, rather than relying on the host OS (the Linux or FreeBSD kernel).

If we tell ukvm-bin where the disk image is, it will provide that disk image to the unikernel:



$ ./ukvm-bin --disk=disk.img block_test.ukvm
            |      ___|
  __|  _ \  |  _ \ __ \
\__ \ (   | | (   |  ) |
____/\___/ _|\___/____/
Solo5: Memory map: 512 MB addressable:
Solo5:     unused @ (0x0 - 0xfffff)
Solo5:       text @ (0x100000 - 0x1e1fff)
Solo5:     rodata @ (0x1e2000 - 0x216fff)
Solo5:       data @ (0x217000 - 0x2c6fff)
Solo5:       heap >= 0x2c7000 < stack < 0x20000000
Solo5: Clock source: KVM paravirtualized clock
Solo5: new bindings
STUB: getenv() called
2017-02-09 00:48:14 -00:00: INF [block] sectors = 100000
read_write=true
sector_size=512

2017-02-09 00:48:14 -00:00: ERR [block] Expecting error output from the following operation...
blk write failed... 512 to sector=100000
2017-02-09 00:48:14 -00:00: ERR [block] Expecting error output from the following operation...
blk write failed... 512 to sector=100000
2017-02-09 00:48:14 -00:00: ERR [block] Expecting error output from the following operation...
virtio read failed... 512 from sector=100000
2017-02-09 00:48:14 -00:00: ERR [block] Expecting error output from the following operation...
virtio read failed... 512 from sector=100000
2017-02-09 00:48:14 -00:00: INF [block] Test sequence finished

2017-02-09 00:48:14 -00:00: INF [block] Total tests started: 10

2017-02-09 00:48:14 -00:00: INF [block] Total tests passed:  10

2017-02-09 00:48:14 -00:00: INF [block] Total tests failed:  0

Solo5: solo5_app_main() returned with 0

Step 3: Key/value stores

The earlier block device example shows how very low-level access can work. Now let's move up to a more familiar abstraction: a key/value store that can retrieve buffers from string keys. This is essential for many common uses such as retrieving configuration data or website HTML and images.

The device-usage/kv_ro directory in mirage-skeleton contains a simple key/value store example. The subdirectory t/ contains a few files, one of which the unikernel will compare against a known constant.

The config.ml might look familiar after the earlier block and console examples:

open Mirage

let disk1 = generic_kv_ro "t"

let main =
  foreign
    "Unikernel.Main" (kv_ro @-> job)

let () =
  register "kv_ro" [main $ disk]

We construct the kv_ro device disk by using the generic_kv_ro function. This takes a single directory as its argument, and will do its best to provide the content of that directory to the unikernel by whatever means make sense given the target provided at configuration time -- this might be an implementation that calls functions from OCaml's Unix module, or a function that transforms an entire directory into a static ML file that can respond with the file contents directly. This removes the need to have an external block device entirely and is very convenient indeed for small files.

Using generic_kv_ro in your config.ml causes to Mirage to automatically create a configuration key, kv_ro, which you can use to request a specific implementation of the key-value store's implementation. To see documentation, try:

$ cd device-usage/kv_ro
$ mirage help configure

Under the "UNIKERNEL PARAMETERS" section, you should see:

       --kv_ro=KV_RO (absent=crunch)
           Use a fat, archive, crunch or direct pass-through implementation
           for the unikernel.

More documentation is available at the Mirage module documentation for generic_kv_ro.

Let's try a few different kinds of key-value implementations. First, we'll build a Unix version. If we don't specify which kind of kv_ro we want, we'll get a crunch implementation, the contents of which we can see at static1.ml:

$ cd device-usage/kv_ro
$ mirage configure -t unix
$ make depend
$ make
$ less static1.ml # the generated filesystem
$ ./kv_ro

We can use the direct implementation with the Unix target as well:

$ cd device-usage/kv_ro
$ mirage configure -t unix --kv_ro=direct
$ make depend
$ make
$ ./kv_ro

You may have noticed that, unlike with our hello_key example, the kv_ro key can't be specified at runtime -- it's only understood as an argument to mirage configure. This is because the kv_ro implementation we choose influences the set of dependencies that are assembled and baked into the final product. If we choose direct, we'll get a different set of software than if we choose crunch -- in either case, no code that isn't required will be included in the final product.

You should now be seeing the power of the MirageOS configuration tool: we have built several applications that use fairly complex concepts such as filesystems and block devices that are independent of the implementations (by virtue of our application logic being a functor), and then are able to assemble several combinations of unikernels via relatively simple configuration files and options passed at compile-time and runtime.

Step 4: Networking

There are several ways that we might want to configure our network for a Mirage application:

All of this can be manipulated via command-line arguments or environment variables, just as we configured the key-value store in the previous example. The example in the device-usage/network directory of mirage-skeleton is illustrative:

open Mirage

let port =
  let doc = Key.Arg.info ~doc:"The TCP port on which to listen for incoming connections." ["port"] in
  Key.(create "port" Arg.(opt int 8080 doc))

let main = foreign ~keys:[Key.abstract port] "Unikernel.Main" (stackv4 @-> job)

let stack = generic_stackv4 default_network

let () =
  register "network" [
    main $ stack
  ]

We have a custom configuration key defining which TCP port to listen for connections on. The network device is derived from default_network, a function provided by Mirage which will choose a reasonable default based on the target the user chooses to pass to mirage configure - just like the reasonable default provided by generic_kv_ro in the previous example.

generic_stackv4 attempts to build a sensible network stack on top of the physical interface given by default_network. There are quite a few configuration keys exposed when generic_stackv4 is given related to networking configuration -- for a full list, try mirage help configure in the device-usage/network directory.

Unix / Socket networking

Let's get the network stack compiling using the standard Unix sockets APIs first.

$ cd device-usage/network
$ mirage configure -t unix --net socket
$ make depend
$ make
$ ./network

This Unix application is now listening on TCP port 8080, and will print to the console information about data received. Let's try talking to it using the commonly available netcat nc(1) utility. From a different console execute:

$ echo -n hello tcp world | nc -nw1 127.0.0.1 8080

You should see log messages documenting your connection from 127.0.0.1 in the console running ./network. You may have noticed that some information that you may have expected to see after looking at unikernel.ml isn't being output. That's because we haven't specified the log level for ./network, and it defaults to info. Some of the output for this application is sent with the log level set to debug, so to see it, we need to run ./network with a higher log level for all logs:

$ ./network -l "*:debug"

The program will then output the debug-level logs, which include the content of any messages it reads. Here's an example of what you might see:

$ ./network -l "*:debug"
2017-02-10 17:23:24 +02:00: INF [tcpip-stack-socket] Manager: connect
2017-02-10 17:23:24 +02:00: INF [tcpip-stack-socket] Manager: configuring
2017-02-10 17:23:27 +02:00: INF [application] new tcp connection from IP 127.0.0.1 on port 36358
2017-02-10 17:23:27 +02:00: DBG [application] read: 15 bytes:
hello tcp world

Unix / MirageOS Stack with DHCP

Next, let's try using the direct MirageOS network stack. It will be necessary to run these programs with sudo or as the root user, as they need direct access to a network device. We won't be able to contact them via the loopback interface on 127.0.0.1 either -- the stack will need to either obtain IP address information via DHCP, or it can be configured directly via the --ipv4 configuration key.

To configure via DHCP:

$ cd device-usage/network 
$ mirage configure -t unix --dhcp true --net direct
$ make depend
$ make
$ sudo ./network

Hopefully, the application will successfully receive its network configuration. Once the program has completed the lease transaction, it will log the configuration information, and you'll be able to contact it as before via its own IP.

Unix / MirageOS Stack with static IP addresses

By default, if we do not use DHCP with a direct network stack, Mirage will configure the stack to use an address of 10.0.0.2. You can specify a different address with the --ipv4 configuration key. Depending on whether you've configured with -t macosx or -t unix, the logic for contacting the application from another terminal will be different.

For unix: Verify that you have an existing tap0 interface by reviewing $ sudo ip link show; if you do not, load the tuntap kernel module ($ sudo modprobe tun) and create a tap0 interface owned by you ($ sudo tunctl -u $USER -t tap0). Bring tap0 up using $ sudo ifconfig tap0 10.0.0.1 up, then:

$ cd device-usage/network
$ mirage configure -t unix --dhcp false --net direct
$ make depend
$ make
$ sudo ./network

For macosx:

(* TODO! *)

Now you should be able to ping the unikernel's interface:

$ ping 10.0.0.2
PING 10.0.0.2 (10.0.0.2) 56(84) bytes of data.
64 bytes from 10.0.0.2: icmp_seq=1 ttl=38 time=0.527 ms
64 bytes from 10.0.0.2: icmp_seq=2 ttl=38 time=0.367 ms
64 bytes from 10.0.0.2: icmp_seq=3 ttl=38 time=0.291 ms
^C
--- 10.0.0.2 ping statistics ---
3 packets transmitted, 3 received, 0% packet loss, time 2005ms
rtt min/avg/max/mdev = 0.291/0.395/0.527/0.098 ms

Finally, you can then execute the same nc(1) commands as before (modulo the target IP address of course!) to interact with the running unikernel:

$ echo -n hello tcp world | nc -nw1 10.0.0.2 8080

And you will see the same output in the unikernel's terminal:

read: 15 "hello tcp world"

Ukvm

Let's make a network-enabled unikernel! The IP configuration should be similar to what you've set up in the previous examples, but instead of -t unix or -t macosx, build with a ukvm target. If you need to specify a static IP address, remember that it should go at the end of the command in which you invoke ukvm-bin, just like the argument to hello in the hello-key example.

$ cd device-usage/network
$ mirage configure -t ukvm --dhcp true # for environments where DHCP works
$ make depend
$ make
$ ./ukvm-bin network.ukvm

What's Next?

There are a number of other examples in device-usage/ which show some simple invocations of various devices like consoles and clocks. You may also be interested in the applications/ directory of the mirage-skeleton repository, which contains examples that use multiple devices to build nontrivial applications, like DNS, DHCP, and HTTPS servers.

The real MirageOS website (which is itself a unikernel) may also be of interest to you! Documentation is available at mirage-www, and the source code is published in a public GitHub repository.