env/openbsd-amd64,cmd/bootstrapswarm: get openbsd working with LUCI

Teach bootstrapswarm to read the hostname from GCE when sending a VM
token, since that's all that'll work.

Update openbsd's make.bash:
- Do all work in a temp directory so that we don't have to manage
  individual files.
- Pass -I to pkg_add to avoid prompts.
- Increase ulimits on the worker user.
- Install python and sudo.
- Use rc.firsttime rather than reinventing it with /firstboot.
- Decide whether to start the buildlet or Swarming bot based on the
  GCE project as usual.
- Package bootstrapswarm into the image.

Change-Id: I61bbc1903f8eab596bf0d19a316993a92f6bd599
Reviewed-on: https://go-review.googlesource.com/c/build/+/526618
LUCI-TryBot-Result: Go LUCI <[email protected]>
Reviewed-by: Carlos Amedee <[email protected]>
Auto-Submit: Heschi Kreinick <[email protected]>

Author: heschi

cmd/bootstrapswarm/bootstrapswarm.go

@@ -41,13 +41,14 @@ import (
 	"os/exec"
 	"path/filepath"
 	"runtime"
+	"strings"
 
 	"cloud.google.com/go/compute/metadata"
 )
 
 var (
 	tokenFilePath = flag.String("token-file-path", defaultTokenLocation(), "Path to the token file (used when not on GCE)")
-	hostname      = flag.String("hostname", os.Getenv("HOSTNAME"), "Hostname of machine to bootstrap (required)")
+	hostname      = flag.String("hostname", os.Getenv("HOSTNAME"), "Hostname of machine to bootstrap")
 	swarming      = flag.String("swarming", "chromium-swarm.appspot.com", "Swarming server to connect to")
 )
 
@@ -70,7 +71,7 @@ func main() {
 var httpClient = http.DefaultClient
 
 func bootstrap(ctx context.Context, hostname, tokenPath string) error {
-	httpHeaders := map[string]string{"X-Luci-Swarming-Bot-ID": hostname}
+	httpHeaders := map[string]string{}
 	if metadata.OnGCE() {
 		log.Println("Bootstrapping the swarming bot with GCE authentication")
 		log.Println("retrieving the GCE VM token")
@@ -79,6 +80,14 @@ func bootstrap(ctx context.Context, hostname, tokenPath string) error {
 			return fmt.Errorf("unable to retrieve GCE Machine Token: %w", err)
 		}
 		httpHeaders["X-Luci-Gce-Vm-Token"] = token
+
+		// Override the hostname flag with the GCE hostname. This is a hard
+		// requirement for LUCI, so there's no point in trying anything else.
+		fullHost, err := metadata.Hostname()
+		if err != nil {
+			return fmt.Errorf("retrieving hostname: %w", err)
+		}
+		hostname = strings.Split(fullHost, ".")[0]
 	} else {
 		log.Println("Bootstrapping the swarming bot with certificate authentication")
 		log.Println("retrieving the luci-machine-token from the token file")
@@ -98,6 +107,7 @@ func bootstrap(ctx context.Context, hostname, tokenPath string) error {
 		}
 		httpHeaders["X-Luci-Machine-Token"] = tok.LuciMachineToken
 	}
+	httpHeaders["X-Luci-Swarming-Bot-ID"] = hostname
 	log.Println("Downloading the swarming bot")
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, "https://"+*swarming+"/bot_code", nil)
 	if err != nil {

env/openbsd-amd64/make.bash

@@ -14,6 +14,9 @@ readonly SNAPSHOT=false
 readonly ARCH="${ARCH:-amd64}"
 readonly MIRROR="${MIRROR:-cdn.openbsd.org}"
 
+readonly WORK="$(mktemp -d)"
+readonly SITE="${WORK}/site"
+
 if [[ "${ARCH}" != "amd64" && "${ARCH}" != "i386" ]]; then
   echo "ARCH must be amd64 or i386"
   exit 1
@@ -31,45 +34,48 @@ if [[ ! -f "${ISO}" ]]; then
 fi
 
 function cleanup() {
-	rm -f "${ISO_PATCHED}"
-	rm -f auto_install.conf
-	rm -f boot.conf
-	rm -f disk.raw
-	rm -f disklabel.template
-	rm -f etc/{installurl,rc.local,sysctl.conf}
-	rm -f install.site
-	rm -f random.seed
-	rm -f site${RELNO}.tgz
-	rmdir etc
+	rm -rf "${WORK}"
 }
 
 trap cleanup EXIT INT
 
 # Create custom siteXX.tgz set.
-PKG_ADD_OPTIONS=""
+PKG_ADD_OPTIONS="-I"
 if [[ "$SNAPSHOT" = true ]]; then
-  PKG_ADD_OPTIONS="-D snap"
+  PKG_ADD_OPTIONS="-I -D snap"
 fi
-mkdir -p etc
-cat >install.site <<EOF
+mkdir -p ${SITE}/etc
+cat >${SITE}/install.site <<EOF
 #!/bin/sh
 touch /firstboot
 echo 'set tty com0' > boot.conf
 EOF
 
-cat >etc/installurl <<EOF
+cat >${SITE}/etc/installurl <<EOF
 https://${MIRROR}/pub/OpenBSD
 EOF
-cat >etc/rc.local <<EOF
-if [[ -f /firstboot ]]; then
-  syspatch
-  # Run syspatch twice in case syspatch itself needs patching (this is the case with OpenBSD
-  # 7.1: https://www.openbsd.org/errata71.html )
-  syspatch
-  pkg_add -iv ${PKG_ADD_OPTIONS} bash curl git
-  rm -f /firstboot
-fi
+cat >${SITE}/etc/rc.firsttime <<EOF
+set -x
+cat > /etc/login.conf.d/moreres <<'EOLOGIN'
+moreres:\
+  :maxproc-max=1024: \
+  :maxproc-cur=1024: \
+  :openfiles-max=4096: \
+  :openfiles-cur=4096: \
+  :tc=default:
+EOLOGIN
+cap_mkdb
+usermod -L moreres swarming
+syspatch
+# Run syspatch twice in case syspatch itself needs patching (this is the case with OpenBSD
+# 7.1: https://www.openbsd.org/errata71.html )
+syspatch
+pkg_add -iv ${PKG_ADD_OPTIONS} bash curl git python3 sudo--gettext
+chown root:wheel /etc/sudoers
+halt -p
+EOF
 
+cat >${SITE}/etc/rc.local <<EOF
 (
   set -x
 
@@ -83,37 +89,52 @@ fi
   (
     set -e
     export PATH="\$PATH:/usr/local/bin"
-    /usr/local/bin/curl -o /buildlet \$(/usr/local/bin/curl --fail -H "Metadata-Flavor: Google" http://metadata.google.internal/computeMetadata/v1/instance/attributes/buildlet-binary-url)
-    chmod +x /buildlet
-    exec /buildlet
+    project=\$(curl -s -H "Metadata-Flavor: Google" http://metadata.google.internal/computeMetadata/v1/project/project-id)
+    case "\$project" in
+      *luci*)
+        gcehost=\$(curl -s -H "Metadata-Flavor: Google" http://metadata.google.internal/computeMetadata/v1/instance/hostname | cut -d . -f 1)
+        su -l swarming -c "/usr/local/bin/bootstrapswarm --hostname \$gcehost"
+      ;;
+      *)
+        /usr/local/bin/curl -o /buildlet \$(/usr/local/bin/curl --fail -H "Metadata-Flavor: Google" http://metadata.google.internal/computeMetadata/v1/instance/attributes/buildlet-binary-url)
+        chmod +x /buildlet
+        exec /buildlet
+      ;;
+    esac
   )
   echo "giving up"
   (
-    sleep 10
+    sleep 60
     halt -p
   )&
 )
 EOF
-cat >etc/sysctl.conf <<EOF
+cat >${SITE}/etc/sysctl.conf <<EOF
 hw.smt=1
 kern.timecounter.hardware=tsc
 EOF
-chmod +x install.site
-tar -zcvf site${RELNO}.tgz install.site etc/{installurl,rc.local,sysctl.conf}
+cat >${SITE}/etc/sudoers <<EOF
+root ALL=(ALL:ALL) ALL
+swarming ALL=NOPASSWD:/sbin/shutdown -r now
+EOF
+chmod +x ${SITE}/install.site
+mkdir -p ${SITE}/usr/local/bin
+CGO_ENABLED=0 GOOS=openbsd GOARCH=amd64 go1.21.0 build -o ${SITE}/usr/local/bin/bootstrapswarm golang.org/x/build/cmd/bootstrapswarm
+tar -C ${SITE} -zcf ${WORK}/site${RELNO}.tgz .
 
 # Autoinstall script.
-cat >auto_install.conf <<EOF
-System hostname = buildlet
+cat >${WORK}/auto_install.conf <<EOF
+System hostname = openbsd-amd64
 Which network interface = vio0
 IPv4 address for vio0 = dhcp
 IPv6 address for vio0 = none
 Password for root account = root
 Do you expect to run the X Window System = no
 Change the default console to com0 = yes
 Which speed should com0 use = 115200
-Setup a user = gopher
-Full name for user gopher = Gopher Gopherson
-Password for user gopher = gopher
+Setup a user = swarming
+Full name for user swarming = Swarming Gopher Gopherson
+Password for user swarming = swarming
 Allow root ssh login = no
 What timezone = US/Pacific
 Which disk = sd0
@@ -126,32 +147,32 @@ Directory does not contain SHA256.sig. Continue without verification = yes
 EOF
 
 # Disklabel template.
-cat >disklabel.template <<EOF
+cat >${WORK}/disklabel.template <<EOF
 /	5G-*	95%
 swap	1G
 EOF
 
 # Hack install CD a bit.
-echo 'set tty com0' > boot.conf
-dd if=/dev/urandom of=random.seed bs=4096 count=1
+echo 'set tty com0' > ${WORK}/boot.conf
+dd if=/dev/urandom of=${WORK}/random.seed bs=4096 count=1
 cp "${ISO}" "${ISO_PATCHED}"
 growisofs -M "${ISO_PATCHED}" -l -R -graft-points \
-  /${VERSION}/${ARCH}/site${RELNO}.tgz=site${RELNO}.tgz \
-  /auto_install.conf=auto_install.conf \
-  /disklabel.template=disklabel.template \
-  /etc/boot.conf=boot.conf \
-  /etc/random.seed=random.seed
+  /${VERSION}/${ARCH}/site${RELNO}.tgz=${WORK}/site${RELNO}.tgz \
+  /auto_install.conf=${WORK}/auto_install.conf \
+  /disklabel.template=${WORK}/disklabel.template \
+  /etc/boot.conf=${WORK}/boot.conf \
+  /etc/random.seed=${WORK}/random.seed
 
 # Initialize disk image.
-rm -f disk.raw
-qemu-img create -f raw disk.raw 10G
+rm -f ${WORK}/disk.raw
+qemu-img create -f raw ${WORK}/disk.raw 30G
 
 # Run the installer to create the disk image.
 expect <<EOF
 set timeout 1800
 
 spawn qemu-system-x86_64 -nographic -smp 2 \
-  -drive if=virtio,file=disk.raw,format=raw -cdrom "${ISO_PATCHED}" \
+  -drive if=virtio,file=${WORK}/disk.raw,format=raw -cdrom "${ISO_PATCHED}" \
   -net nic,model=virtio -net user -boot once=d
 
 expect timeout { exit 1 } "boot>"
@@ -174,6 +195,6 @@ EOF
 
 # Create Compute Engine disk image.
 echo "Archiving disk.raw... (this may take a while)"
-tar -Szcf "openbsd-${VERSION}-${ARCH}-gce.tar.gz" disk.raw
+tar -C ${WORK} -Szcf "openbsd-${VERSION}-${ARCH}-gce.tar.gz" disk.raw
 
 echo "Done. GCE image is openbsd-${VERSION}-${ARCH}-gce.tar.gz."