NimD now checks if another instance is running before starting up

.gitignore

@@ -5,7 +5,6 @@ main
 *.iso
 nimd
 main
-boot.sh
 rootfs
 packervm
 test
@@ -14,7 +13,5 @@ initrd*
 vmlinuz-linux
 vmlinuz*
 debian*
-start.sh
-rebuild.sh
 vm.qcow2
 *.tar.*

README.md

@@ -116,4 +116,10 @@ workers              = 1                     # Number of worker processes to use
 restartDelay         = 10                    # Delay (seconds) that nimd will wait before restarting itself after crashes
 sigtermDelay         = 90                    # Delay (seconds) that nimd will wait before terminating child processes with
                                              # SIGKILL after sending a more gentle SIGTERM upon shutdown or exit              
-```
+```
+
+## Testing NimD
+
+NimD is not quite ready for production yet, but in the `scripts` folder you can find a few simple bash scripts to test NimD
+in a minimal Alpine Linux VM using QEMU. Note that due to weirdness in how stdout is handled on the VGA port, the VM will use
+the serial port (ttyS0) as output by default (you can change this in the kernel parameters)

scripts/boot.sh

@@ -0,0 +1,115 @@
+#!/usr/bin/env bash
+
+set -Eeuo pipefail
+trap cleanup SIGINT SIGTERM ERR EXIT
+
+script_dir=$(cd "$(dirname "${BASH_SOURCE[0]}")" &>/dev/null && pwd -P)
+
+usage() {
+  cat << EOF 
+Usage: $(basename "${BASH_SOURCE[0]}") [-h] [-v] [-m] [-b] -k kernel -i initrd -r rootfs
+
+Alpinemini vm
+
+Available options:
+
+-h, --help      Print this help and exit
+-v, --verbose   Print script debug info
+-k, --kernel 	Specify kernel file
+-i, --initrd 	Specify initrd file
+-m, --memory    Set maximum vm memory
+-b, --build 	Build a new disk (and then boot)
+EOF
+  exit
+}
+
+cleanup() {
+  trap - SIGINT SIGTERM ERR EXIT
+}
+
+setup_colors() {
+  if [[ -t 2 ]] && [[ -z "${NO_COLOR-}" ]] && [[ "${TERM-}" != "dumb" ]]; then
+    NOFORMAT='\033[0m' RED='\033[0;31m' GREEN='\033[0;32m' ORANGE='\033[0;33m' BLUE='\033[0;34m' PURPLE='\033[0;35m' CYAN='\033[0;36m' YELLOW='\033[1;33m'
+  else
+    NOFORMAT='' RED='' GREEN='' ORANGE='' BLUE='' PURPLE='' CYAN='' YELLOW=''
+  fi
+}
+
+msg() {
+  echo >&2 -e "${1-}"
+}
+
+die() {
+  local msg=$1
+  local code=${2-1}
+  msg "$msg"
+  exit "$code"
+}
+
+parse_params() {
+  build=0
+  rootfs=''
+  kernel=''
+  initrd=''
+  memory=''
+
+  while :; do
+    case "${1-}" in
+    -h | --help) usage ;;
+    -v | --verbose) set -x ;;
+    --no-color) NO_COLOR=1 ;;
+    -b | --build) build=1 ;; # build disk
+    -k | --kernel) # kernel file
+      kernel="${2-}"
+      shift
+      ;;
+    -i | --initrd) # initrd file
+      initrd="${2-}"
+      shift
+      ;;
+    -m | --memory) # max memory
+      memory="${2-}"
+      shift
+      ;;
+    -?*) die "Unknown option: $1" ;;
+    *) break ;;
+    esac
+    shift
+  done
+
+  args=("$@")
+
+  # check required params and arguments
+  [[ -z "${kernel-}" ]] && die "${RED}Missing required parameter:${NOFORMAT} kernel"
+  [[ -z "${initrd-}" ]] && die "${RED}Missing required parameter:${NOFORMAT} initrd"
+
+  return 0
+}
+setup_colors
+parse_params "$@"
+
+if ! [ -x "$(command -v qemu-system-x86_64)" ]; then
+  echo '${RED}Error:${NOFORMAT} unable to find qemu-system-x86_64, please install it first.' >&2
+  exit 1
+fi
+
+if ! [ -x "$(command -v qemu-img)" ]; then
+  echo '${RED}Error:${NOFORMAT} unable to find qemu-img, please install it first.' >&2
+  exit 1
+fi
+
+
+if [[ $build -eq 1 ]]
+then
+ msg "${CYAN}Building disk... Please wait${NOFORMAT}"
+ qemu-img create -f qcow2 vm.qcow2 800M
+ qemu-system-x86_64 -m 256M -smp 1 -drive file=packervm/packer.qcow2,if=virtio,readonly=on -drive file=vm.qcow2,if=virtio -enable-kvm -fsdev local,id=rootfs_dev,path=rootfs,security_model=none -device virtio-9p-pci,fsdev=rootfs_dev,mount_tag=rootfs -display none
+fi
+
+qemumem=''
+if ! [[ $memory -eq "" ]]
+ then
+   qemumem="-m ${memory}"
+fi
+
+qemu-system-x86_64 -net nic,model=virtio,netdev=user.0 -netdev user,id=user.0 -drive file=vm.qcow2,if=virtio -enable-kvm -kernel ${kernel} -initrd ${initrd} ${qemumem}  -append 'root=/dev/vda1 rw quiet modules=ext4 console=ttyS0,9600 console=ttyS0 init=/bin/nimd'

scripts/rebuild.sh

@@ -0,0 +1,10 @@
+# Build the environment
+mkdir -p rootfs/etc/nimd
+cp nimd.conf rootfs/etc/nimd/nimd.conf
+nim -o:rootfs/bin/nimd      -d:release --gc:orc --opt:size --passL:"-static" compile src/main.nim
+nim -o:rootfs/bin/halt      -d:release --gc:orc --opt:size --passL:"-static" compile src/programs/halt.nim
+nim -o:rootfs/bin/reboot    -d:release --gc:orc --opt:size --passL:"-static" compile src/programs/reboot.nim
+nim -o:rootfs/bin/poweroff  -d:release --gc:orc --opt:size --passL:"-static" compile src/programs/poweroff.nim
+
+# Start the VM
+./scripts/boot.sh --kernel vmlinuz-linux --initrd initrd-linux.img --memory 1G --build

scripts/start.sh

@@ -0,0 +1 @@
+./scripts/boot.sh --kernel vmlinuz-linux --initrd initrd-linux.img --memory 1G

src/core/mainloop.nim

@@ -28,8 +28,8 @@ proc mainLoop*(logger: Logger, config: NimDConfig, startServices: bool = true) =
     if startServices:
         logger.info("Processing default runlevel")
         startServices(logger, workers=config.workers, level=Default)
-        logger.debug(&"Unblocking signals")
-        unblockSignals(logger)
+    logger.debug(&"Unblocking signals")
+    unblockSignals(logger)
     logger.info("System initialization complete, idling on control socket")
     var opType: string
     try:
@@ -52,6 +52,7 @@ proc mainLoop*(logger: Logger, config: NimDConfig, startServices: bool = true) =
             # - 'h' -> halt
             # - 's' -> Services-related operations (start, stop, get status, etc.)
             # - 'l' -> Reload in-memory configuration
+            # - 'c' -> Check NimD status (returns "1" if up)
             case opType:
                 of "p":
                     logger.info("Received shutdown request")
@@ -63,11 +64,14 @@ proc mainLoop*(logger: Logger, config: NimDConfig, startServices: bool = true) =
                     logger.info("Received halt request")
                     halt(logger)
                 of "s":
-                    logger.info("Received service request")
+                    logger.info("Received service-related request")
                     # TODO: Operate on services
                 of "l":
                     logger.info("Received reload request")
                     mainLoop(logger, parseConfig(logger, "/etc/nimd/nimd.conf"), startServices=false)
+                of "c":
+                    logger.info("Received check request, responding")
+                    clientSocket.send("1")
                 else:
                     logger.warning(&"Received unknown operation type '{opType}' via control socket, ignoring it")
                     discard

src/core/shutdown.nim

@@ -32,16 +32,14 @@ type ShutdownHandler* = ref object
 
 
 const reboot_codes = {"poweroff": 0x4321fedc'i64, "reboot": 0x01234567'i64, "halt": 0xcdef0123}.toTable()
+var shutdownHandlers: seq[ShutdownHandler] = @[]
+var sigTermDelay: float = 90
 
 
 proc newShutdownHandler*(body: proc (logger: Logger, code: int)): ShutdownHandler =
     result = ShutdownHandler(body: body)
 
 
-var shutdownHandlers: seq[ShutdownHandler] = @[]
-var sigTermDelay: float = 90
-
-
 proc setSigTermDelay*(delay: int = 90) =
     # Sets the sigtermDelay variable
     sigTermDelay = float(delay)
@@ -97,8 +95,8 @@ proc nimDExit*(logger: Logger, code: int, emerg: bool = true) =
         # We're in emergency mode: do not crash the kernel, spawn a shell and exit
         logger.fatal("NimD has entered emergency mode and cannot continue. You will be now (hopefully) dropped in a root shell: you're on your own. May the force be with you")
         logger.info("Terminating child processes with SIGKILL")
-        discard execCmd(os.getEnv("SHELL", "/bin/sh"))  # TODO: Is this fine? maybe use execProcess
         discard posix.kill(-1, SIGKILL)
+        discard execCmd(os.getEnv("SHELL", "/bin/sh"))  # TODO: Is this fine? maybe use execProcess
         quit(-1)
     logger.warning("The system is shutting down")
     logger.info("Processing shutdown runlevel")

src/main.nim

@@ -14,6 +14,7 @@
 import parseopt
 import strformat
 import posix
+import net
 import os
 
 # NimD's own stuff
@@ -68,10 +69,51 @@ proc addStuff =
     addService(shell)
 
 
+proc checkControlSocket(logger: Logger, config: NimDConfig): bool =
+    ## Performs some startup checks on nim's control
+    ## socket
+    result = true
+    var stat_result: Stat
+    if posix.stat(cstring(config.sock), stat_result) == -1:
+        logger.warning(&"Could not stat() {config.sock}, assuming NimD instance isn't running")
+    # I stole this from /usr/lib/python3.10/stat.py
+    elif (int(stat_result.st_mode) and 0o170000) != 0o140000:
+        logger.fatal(&"{config.sock} exists and is not a socket")
+        result = false
+    elif dirExists(config.sock):
+        logger.info("Control socket path is a directory, appending nimd.sock to it")
+        config.sock = config.sock.joinPath("nimd.sock")
+    else:
+        logger.debug("Trying to reach current NimD instance")
+        var sock = newSocket(Domain.AF_UNIX, SockType.SOCK_STREAM, Protocol.IPPROTO_IP)
+        try:
+            sock.connectUnix(config.sock)
+            logger.info("Control socket already exists, trying to reach current NimD instance")
+        except OSError:
+            logger.warning(&"Could not connect to control socket at {config.sock} ({getCurrentExceptionMsg()}), assuming NimD instance isn't running")
+            try:
+                removeFile(config.sock)
+            except OSError:
+                logger.warning(&"Could not delete dangling control socket at {config.sock} ({getCurrentExceptionMsg()})")
+        if sock.trySend("c"):
+            try:
+                if sock.recv(1, timeout=5) == "1":
+                    logger.error("Another NimD instance is running! Exiting")
+                    result = false
+            except OSError:
+                logger.warning(&"Could not read from control socket at {config.sock} ({getCurrentExceptionMsg()}), assuming NimD instance isn't running")
+            except TimeoutError:
+                logger.warning(&"Could not read from control socket at {config.sock} ({getCurrentExceptionMsg()}), assuming NimD instance isn't running")
+        else:
+            logger.fatal(&"Could not write on control socket at {config.sock}")
+            result = false
+
 
 proc main(logger: Logger, config: NimDConfig) =
     ## NimD's entry point and setup
     ## function
+    if not checkControlSocket(logger, config):
+        return
     logger.debug(&"Setting log file to '{config.logFile}'")
     setLogFile(file=config.logFile)
     logger.debug("Starting NimD: A minimal, self-contained, dependency-based Linux init system written in Nim")
@@ -173,8 +215,7 @@ when isMainModule:
                         quit(EINVAL) # EINVAL - Invalid argument
             else:
                 echo "Usage: nimd [options]"
-                quit(EINVAL) # EINVAL - Invalid argument
-
+                quit(EINVAL) # EINVAL - Invalid argument        
     setStdIoUnbuffered()   # Colors don't work otherwise!
     try:
         main(logger, parseConfig(logger, "/etc/nimd/nimd.conf"))