Writing my own init with Go - Part 2

February 9, 2016

Earlier: Writing my own init with Go - Part 1

Next: Writing my own init with Go - Part 3, Packaging/Configs

I am working on myinit.go. Init process is no good if it cannot spawn other processes. I will use the other processes for critical functionality, i.e getting an IP address for future interactivity over SSH or maybe HTTP, or possibly Web Sockets. Because, why not?

Anyways, I modified myinit.go to following and put it in the /init folder to have a nicer structure. It basically spawns a new process, and prints it output. The modified init code is as follows:

package main 

import (

func execute(filepath string) error {
    // Prepare running a command!
    cmd := exec.Command(filepath)    
    sout, _ := cmd.StdoutPipe()
    serr, _ := cmd.StderrPipe()
    // Start and get the output
    err := cmd.Start()
    if err != nil {
        log.Println("Error starting the program", err)
    go io.Copy(os.Stdout, sout)
    go io.Copy(os.Stdout, serr)    
    err = cmd.Wait()
    if err != nil {
        log.Println("Error waiting program")
    log.Println("Your program exited!")
    return nil

func main(){
    log.Println("Hello world from Go!")
    log.Println("Gopher is alive!")
    log.Println("Executing some binary!")
    // Wait forever
    for {

And my /init/test/mybinary.go is just a println to stderr.

package main 

import (

func main(){
    log.Println("Hello world, stderr")
    log.Println("Hello world, stdout")

I upgraded to Ubuntu 16.04 having 4.4 Kernel, so outputs might change a little. For instance, in 3.19 I could not write to stdout, but now I can write to both stdout & stderr as well.

The output is following:

[    6.441931] EXT4-fs (sda1): mounted filesystem with ordered data mode. Opts: (null)
Begin: Running /scripts/local-bottom ... done.
Begin: Running /scripts/init-bottom ... done.
2016/02/08 21:48:50 Hello world from Go!
2016/02/08 21:48:50 Gopher is alive!
2016/02/08 21:48:50 Executing some binary!
2016/02/08 21:48:50 Hello world, stderr
2016/02/08 21:48:50 Hello world, stdout
2016/02/08 21:48:50 Your program exited!

Copying the stdout and stderr with go is less then ideal, we should be able to just create files for each and pass their file descriptiors. At least that’s what I think.

Sometimes sync does not sync, I might be using it wrong. After a little research, I found out that sudo blockdev --flushbufs /dev/nbd0p1 works better.

I modified the code so that the stdout and stderr is redirected to files. Also, the kernel paremeters must have the rw flag or you have to remount the /dev/sda1 as rw, but changing kernel parameters are much more easier.

$ kvm -m 1G -nographic -kernel vmlinuz-4.4.0-2-generic -initrd initrd.img-4.4.0-2-generic -append "console=ttyS0 root=/dev/sda1 rw init=/init/myinit" -hda mydisk.img 
func listFiles(filepath string) {
    files, _ := ioutil.ReadDir(filepath)
    for _, f := range files {

func execute(filepath string) error {
    cmd := exec.Command(filepath)   
    fout, err := os.Create("/init/test/stdout")
    if err != nil {
    ferr, err := os.Create("/init/test/stderr")
    if err != nil {
    cmd.Stdout = fout
    cmd.Stderr = ferr
    defer fout.Close()
    defer ferr.Close()

func main(){
    sout, _ := ioutil.ReadFile("/init/test/stdout")
    serr, _ := ioutil.ReadFile("/init/test/stderr")
    log.Println("Stdout of program", string(sout))
    log.Println("Stderr of program", string(serr))

And then the output becomes:

2016/02/08 22:07:58 Hello world from Go!
2016/02/08 22:07:58 Gopher is alive!
2016/02/08 22:07:58 Executing some binary!
2016/02/08 22:07:58 Your program exited!
2016/02/08 22:07:58 mybinary
2016/02/08 22:07:58 mybinary.go
2016/02/08 22:07:58 stderr
2016/02/08 22:07:58 stdout
2016/02/08 22:07:58 Stdout of program 2016/02/08 22:07:58 Hello world, stdout

2016/02/08 22:07:58 Stderr of program 2016/02/08 22:07:58 Hello world, stderr

This has been a fun ride, I can spawn processes, see and save their outputs. However, I have no idea what will happen for long running commands. Because, unlike using dup2, I do not duplicate the actual file descriptors, but it uses io.Copy inside, and one or more Go routine is probably spawned in the background because docs for os.Exec states:

	// Stdout and Stderr specify the process's standard output and error.
	// If either is nil, Run connects the corresponding file descriptor
	// to the null device (os.DevNull).
	// If Stdout and Stderr are the same writer, at most one
	// goroutine at a time will call Write.
	Stdout io.Writer
	Stderr io.Writer

and I see several Go routines for io.Copy in stdin() and writerDescriptor() functions inside it. Therefore, better solution would be to avoid Go’s higher level functions such as os.Command() and use the ForkExec(argv0 string, argv []string, attr *ProcAttr) (pid int, err error) that takes the following structs as a parameter besides program name and parameters:

type ProcAttr struct {
        Dir   string    // Current working directory.
        Env   []string  // Environment.
        Files []uintptr // File descriptors.
        Sys   *SysProcAttr

type SysProcAttr struct {
        Chroot      string         // Chroot.
        Credential  *Credential    // Credential.
        Ptrace      bool           // Enable tracing.
        Setsid      bool           // Create session.
        Setpgid     bool           // Set process group ID to Pgid, or, if Pgid == 0, to new pid.
        Setctty     bool           // Set controlling terminal to fd Ctty (only meaningful if Setsid is set)
        Noctty      bool           // Detach fd 0 from controlling terminal
        Ctty        int            // Controlling TTY fd
        Foreground  bool           // Place child's process group in foreground. (Implies Setpgid. Uses Ctty as fd of controlling TTY)
        Pgid        int            // Child's process group ID if Setpgid.
        Pdeathsig   Signal         // Signal that the process will get when its parent dies (Linux only)
        Cloneflags  uintptr        // Flags for clone calls (Linux only)
        UidMappings []SysProcIDMap // User ID mappings for user namespaces.
        GidMappings []SysProcIDMap // Group ID mappings for user namespaces.
        // GidMappingsEnableSetgroups enabling setgroups syscall.
        // If false, then setgroups syscall will be disabled for the child process.
        // This parameter is no-op if GidMappings == nil. Otherwise for unprivileged
        // users this should be set to false for mappings work.
        GidMappingsEnableSetgroups bool

As you see this struct has many features from users to tty’s settings and even uid-gid mappings! I am still reading how to handle tty stuff properly. This was just an experiment to see how can we spawn dumb processes that spawns and dies. As I mentioned earlier, I will use those processes to offload some functionality from init, such as getting an IP from DCHP and setting it on interfaces. Therefore, my init can be simpler.

comments powered by Disqus