Address review comments

Author: sjmiller609

lib/hypervisor/cloudhypervisor/process.go

@@ -3,10 +3,12 @@ package cloudhypervisor
 import (
 	"context"
 	"fmt"
+	"syscall"
 
 	"github.com/onkernel/hypeman/lib/hypervisor"
 	"github.com/onkernel/hypeman/lib/paths"
 	"github.com/onkernel/hypeman/lib/vmm"
+	"gvisor.dev/gvisor/pkg/cleanup"
 )
 
 func init() {
@@ -53,6 +55,12 @@ func (s *Starter) StartVM(ctx context.Context, p *paths.Paths, version string, s
 		return 0, nil, fmt.Errorf("start process: %w", err)
 	}
 
+	// Setup cleanup to kill the process if subsequent steps fail
+	cu := cleanup.Make(func() {
+		syscall.Kill(pid, syscall.SIGKILL)
+	})
+	defer cu.Clean()
+
 	// 2. Create the HTTP client
 	hv, err := New(socketPath)
 	if err != nil {
@@ -78,6 +86,8 @@ func (s *Starter) StartVM(ctx context.Context, p *paths.Paths, version string, s
 		return 0, nil, fmt.Errorf("boot vm failed with status %d: %s", bootResp.StatusCode(), string(bootResp.Body))
 	}
 
+	// Success - release cleanup to prevent killing the process
+	cu.Release()
 	return pid, hv, nil
 }
 
@@ -96,6 +106,12 @@ func (s *Starter) RestoreVM(ctx context.Context, p *paths.Paths, version string,
 		return 0, nil, fmt.Errorf("start process: %w", err)
 	}
 
+	// Setup cleanup to kill the process if subsequent steps fail
+	cu := cleanup.Make(func() {
+		syscall.Kill(pid, syscall.SIGKILL)
+	})
+	defer cu.Clean()
+
 	// 2. Create the HTTP client
 	hv, err := New(socketPath)
 	if err != nil {
@@ -110,14 +126,14 @@ func (s *Starter) RestoreVM(ctx context.Context, p *paths.Paths, version string,
 	}
 	resp, err := hv.client.PutVmRestoreWithResponse(ctx, restoreConfig)
 	if err != nil {
-		hv.Shutdown(ctx) // Cleanup on failure
 		return 0, nil, fmt.Errorf("restore: %w", err)
 	}
 	if resp.StatusCode() != 204 {
-		hv.Shutdown(ctx) // Cleanup on failure
 		return 0, nil, fmt.Errorf("restore failed with status %d: %s", resp.StatusCode(), string(resp.Body))
 	}
 
+	// Success - release cleanup to prevent killing the process
+	cu.Release()
 	return pid, hv, nil
 }
 

lib/hypervisor/qemu/process.go

@@ -14,6 +14,7 @@ import (
 
 	"github.com/onkernel/hypeman/lib/hypervisor"
 	"github.com/onkernel/hypeman/lib/paths"
+	"gvisor.dev/gvisor/pkg/cleanup"
 )
 
 func init() {
@@ -121,6 +122,12 @@ func (s *Starter) StartVM(ctx context.Context, p *paths.Paths, version string, s
 
 	pid := cmd.Process.Pid
 
+	// Setup cleanup to kill the process if subsequent steps fail
+	cu := cleanup.Make(func() {
+		syscall.Kill(pid, syscall.SIGKILL)
+	})
+	defer cu.Clean()
+
 	// Wait for socket to be ready
 	if err := waitForSocket(socketPath, 10*time.Second); err != nil {
 		vmmLogPath := filepath.Join(logsDir, "vmm.log")
@@ -136,6 +143,8 @@ func (s *Starter) StartVM(ctx context.Context, p *paths.Paths, version string, s
 		return 0, nil, fmt.Errorf("create client: %w", err)
 	}
 
+	// Success - release cleanup to prevent killing the process
+	cu.Release()
 	return pid, hv, nil
 }
 

lib/hypervisor/qemu/vsock.go

@@ -172,7 +172,16 @@ func (c *vsockConn) Read(b []byte) (int, error) {
 }
 
 func (c *vsockConn) Write(b []byte) (int, error) {
-	return unix.Write(c.fd, b)
+	n, err := unix.Write(c.fd, b)
+	// Ensure we never return negative n (violates io.Writer contract)
+	// This can happen when the vsock fd becomes invalid (VM died)
+	if n < 0 {
+		if err == nil {
+			err = io.ErrClosedPipe
+		}
+		return 0, err
+	}
+	return n, err
 }
 
 func (c *vsockConn) Close() error {