hostagent: avoid requiring fuser and ss in the guest ; CI: update colima (v0.5.5→v0.6.5)

.github/workflows/test.yml

@@ -263,11 +263,13 @@ jobs:
     timeout-minutes: 120
     strategy:
       matrix:
-        colima-version: ["v0.5.5"]
+        colima-version: ["v0.6.5"]
     steps:
     - uses: actions/checkout@v4
       with:
-        fetch-depth: 1
+        # fetch-depth is set to 0 to let `limactl --version` print semver-ish version
+        fetch-depth: 0
+        ref: ${{ github.event.pull_request.head.sha }}
     - uses: actions/setup-go@v4
       with:
         go-version: 1.20.x

pkg/hostagent/hostagent.go

@@ -57,6 +57,9 @@ type HostAgent struct {
 
 	clientMu sync.RWMutex
 	client   guestagentclient.GuestAgentClient
+
+	guestAgentAliveCh     chan struct{} // closed on establishing the connection
+	guestAgentAliveChOnce sync.Once
 }
 
 type options struct {
@@ -160,19 +163,20 @@ func New(instName string, stdout io.Writer, sigintCh chan os.Signal, opts ...Opt
 	})
 
 	a := &HostAgent{
-		y:               y,
-		sshLocalPort:    sshLocalPort,
-		udpDNSLocalPort: udpDNSLocalPort,
-		tcpDNSLocalPort: tcpDNSLocalPort,
-		instDir:         inst.Dir,
-		instName:        instName,
-		instSSHAddress:  inst.SSHAddress,
-		sshConfig:       sshConfig,
-		portForwarder:   newPortForwarder(sshConfig, sshLocalPort, rules, inst.VMType),
-		driver:          limaDriver,
-		sigintCh:        sigintCh,
-		eventEnc:        json.NewEncoder(stdout),
-		vSockPort:       vSockPort,
+		y:                 y,
+		sshLocalPort:      sshLocalPort,
+		udpDNSLocalPort:   udpDNSLocalPort,
+		tcpDNSLocalPort:   tcpDNSLocalPort,
+		instDir:           inst.Dir,
+		instName:          instName,
+		instSSHAddress:    inst.SSHAddress,
+		sshConfig:         sshConfig,
+		portForwarder:     newPortForwarder(sshConfig, sshLocalPort, rules, inst.VMType),
+		driver:            limaDriver,
+		sigintCh:          sigintCh,
+		eventEnc:          json.NewEncoder(stdout),
+		vSockPort:         vSockPort,
+		guestAgentAliveCh: make(chan struct{}),
 	}
 	return a, nil
 }
@@ -492,6 +496,21 @@ sudo chown -R "${USER}" /run/host-services`
 	if err := a.waitForRequirements("optional", a.optionalRequirements()); err != nil {
 		errs = append(errs, err)
 	}
+	if !*a.y.Plain {
+		logrus.Info("Waiting for the guest agent to be running")
+		select {
+		case <-a.guestAgentAliveCh:
+			// NOP
+		case <-time.After(time.Minute):
+			err := errors.New("guest agent does not seem to be running; port forwards will not work")
+			if *a.y.VMType == limayaml.WSL2 {
+				// geustagent is currently not available for WSL2: https://github.com/lima-vm/lima/issues/2025
+				logrus.Warn(err)
+			} else {
+				errs = append(errs, err)
+			}
+		}
+	}
 	if err := a.waitForRequirements("final", a.finalRequirements()); err != nil {
 		errs = append(errs, err)
 	}
@@ -605,6 +624,10 @@ func (a *HostAgent) processGuestAgentEvents(ctx context.Context, client guestage
 	if err != nil {
 		return err
 	}
+	logrus.Info("Guest agent is running")
+	a.guestAgentAliveChOnce.Do(func() {
+		close(a.guestAgentAliveCh)
+	})
 
 	logrus.Debugf("guest agent info: %+v", info)
 

pkg/hostagent/requirements.go

@@ -6,7 +6,6 @@ import (
 	"time"
 
 	"github.com/lima-vm/lima/pkg/limayaml"
-	"github.com/lima-vm/lima/pkg/store/filenames"
 	"github.com/lima-vm/sshocker/pkg/ssh"
 	"github.com/sirupsen/logrus"
 )
@@ -120,40 +119,6 @@ fi
 			debugHint: `Append "user_allow_other" to /etc/fuse.conf (/etc/fuse3.conf) in the guest`,
 		})
 	}
-	if a.vSockPort != 0 {
-		req = append(req, requirement{
-			description: "the guest agent to be running",
-			script: fmt.Sprintf(`#!/bin/bash
-set -eux -o pipefail
-if ! timeout 30s bash -c "until ss -a -n --vsock --listen | grep -q ':%d'; do sleep 3; done"; then
-	echo >&2 "lima-guestagent is not installed yet"
-	exit 1
-fi
-`, a.vSockPort),
-			debugHint: fmt.Sprintf(`The guest agent with vsockPort %d does not seem running.
-Make sure that you are using an officially supported image.
-Also see "/var/log/cloud-init-output.log" in the guest.
-A possible workaround is to run "lima-guestagent install-systemd" in the guest.
-`, a.vSockPort),
-		})
-	} else {
-		req = append(req, requirement{
-			description: "the guest agent to be running",
-			script: fmt.Sprintf(`#!/bin/bash
-set -eux -o pipefail
-sock="/dev/virtio-ports/%s"
-if ! timeout 30s bash -c "until sudo fuser \"${sock}\" || sudo lsof \"${sock}\"; do sleep 3; done"; then
-	echo >&2 "lima-guestagent is not installed yet"
-	exit 1
-fi
-`, filenames.VirtioPort),
-			debugHint: fmt.Sprintf(`The guest agent with serialport /dev/virtio-ports/%s does not seem running.
-Make sure that you are using an officially supported image.
-Also see "/var/log/cloud-init-output.log" in the guest.
-A possible workaround is to run "lima-guestagent install-systemd" in the guest.
-`, filenames.VirtioPort),
-		})
-	}
 	return req
 }