Skip to content
Open
Show file tree
Hide file tree
Changes from 2 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
@@ -0,0 +1,61 @@
---
id: TASK-12141
title: Fix VZ Linux real smoke exec output over VSock
status: Done
assignee: []
created_date: ''
updated_date: '2026-08-27 02:02'
labels:
- sandbox
- vz-linux
- real-smoke
- bugfix
dependencies: []
---

## Description

<!-- SECTION:DESCRIPTION:BEGIN -->
Real Apple Virtualization.framework smoke reaches guest readiness but exec output propagation is unreliable: direct helper exec can time out and the host smoke can complete with empty stdout. Investigate and fix the guest/helper VSock request-response path with minimal code and tests.
<!-- SECTION:DESCRIPTION:END -->

## Acceptance Criteria
<!-- AC:BEGIN -->
- [x] #1 Guest VSock client preserves host bytes buffered immediately after ready ACK
- [x] #2 Regression test covers ready ACK plus exec request in one host write
- [x] #3 Real host VZ smoke passes with rebuilt guest bundle
<!-- AC:END -->

## Implementation Notes

<!-- SECTION:NOTES:BEGIN -->
<!-- SECTION:IMPLEMENTATION_NOTES:BEGIN -->
['Observed on local Apple Silicon host with signed macos-vz-helper and Debian arm64 bundle.', 'Helper daemon bundle smoke passed. Real smoke selected tests: 2 passed, 1 failed on missing stdout token.', 'Manual helper probe reached guest readiness then direct exec_guest timed out at guest_transport_timeout.', 'Leading hypothesis: guest VSock client primes handshake/ready with bufio.Reader, then ServeStream reads from raw conn; bytes buffered after ready can be lost from ServeStream.']
<!-- SECTION:IMPLEMENTATION_NOTES:END -->

Root cause confirmed with failing test: primeConnection used a bufio.Reader for handshake/ready, then Run passed the raw conn into ServeStream. If the helper's ready ACK and first exec request arrived together, the exec bytes could remain buffered in the discarded reader.

Fix: return the priming bufio.Reader from primeConnection and pass it to ServeStream so buffered exec bytes remain visible.

Verification: GOCACHE=/private/tmp/tldw-go-build-cache go test ./internal/guest; GOCACHE=/private/tmp/tldw-go-build-cache go test ./... in tools/tldw-agent; rebuilt Debian arm64 bundle; real host smoke final_exit_code=0 at /private/tmp/tvz-e2e.4iJ3wt/evidence/host-smoke-evidence.json.

Bandit: not applicable to touched production code; this change is Go-only under tools/tldw-agent.

PR review follow-up: updated TestGuestVSockClientRunPreservesExecBufferedAfterReady to wait for and validate the exec response before closing the helper-side pipe, removing the race noted in review.
<!-- SECTION:NOTES:END -->

## Final Summary

<!-- SECTION:FINAL_SUMMARY:BEGIN -->
Fixed VZ Linux guest VSock exec delivery by preserving the buffered reader used during handshake/readiness when entering ServeStream. Added a regression test for ready ACK plus exec request arriving in one host write, then verified unit tests and real Apple VZ smoke against a rebuilt Debian arm64 bundle.
<!-- SECTION:FINAL_SUMMARY:END -->

## Definition of Done
<!-- DOD:BEGIN -->
- [x] #1 Acceptance criteria completed
- [x] #2 Tests or verification recorded
- [x] #3 Documentation updated when relevant
- [x] #4 Bandit run for touched code when applicable or document non-code/environment skip
- [x] #5 Final summary added
- [x] #6 Known skips or blockers documented
<!-- DOD:END -->
16 changes: 8 additions & 8 deletions tools/tldw-agent/internal/guest/vsock_client.go
Original file line number Diff line number Diff line change
Expand Up @@ -68,13 +68,13 @@ func (c *VSockClient) Run(ctx context.Context, server *Server) error {
return err
}

err = c.primeConnection(ctx, conn, server, reconnect)
reader, err := c.primeConnection(ctx, conn, server, reconnect)
if err != nil {
_ = conn.Close()
return err
}

err = server.ServeStream(conn, conn)
err = server.ServeStream(reader, conn)
_ = conn.Close()
if err != nil {
return err
Expand All @@ -97,24 +97,24 @@ func (c *VSockClient) primeConnection(
conn io.ReadWriteCloser,
server *Server,
reconnect bool,
) error {
) (*bufio.Reader, error) {
reader := bufio.NewReader(conn)
if reconnect {
if err := c.sendReconnect(conn, reader); err != nil {
return err
return nil, err
}
} else {
if err := c.sendHandshake(conn, reader); err != nil {
return err
return nil, err
}
}
if err := c.sendReady(conn, reader, server); err != nil {
return err
return nil, err
}
if ctx.Err() != nil {
return ctx.Err()
return nil, ctx.Err()
}
return nil
return reader, nil
}

func (c *VSockClient) sendHandshake(conn io.Writer, reader *bufio.Reader) error {
Expand Down
103 changes: 100 additions & 3 deletions tools/tldw-agent/internal/guest/vsock_client_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -3,10 +3,12 @@ package guest
import (
"bufio"
"context"
"encoding/json"
"io"
"net"
"reflect"
"testing"
"time"
)

type recordingDialer struct {
Expand Down Expand Up @@ -104,7 +106,7 @@ func TestGuestVSockClientSendsHandshakeAndReady(t *testing.T) {
done <- nil
}()

if err := client.primeConnection(context.Background(), guestConn, server, false); err != nil {
if _, err := client.primeConnection(context.Background(), guestConn, server, false); err != nil {
t.Fatalf("primeConnection() error = %v", err)
}
if err := <-done; err != nil {
Expand Down Expand Up @@ -204,21 +206,116 @@ func TestGuestVSockClientReconnectsWithSameVMIDAndToken(t *testing.T) {
go assertHello(firstHelperConn, "handshake", firstDone)
go assertHello(secondHelperConn, "reconnect", secondDone)

if err := client.primeConnection(context.Background(), firstGuestConn, server, false); err != nil {
if _, err := client.primeConnection(context.Background(), firstGuestConn, server, false); err != nil {
t.Fatalf("primeConnection(first) error = %v", err)
}
if err := <-firstDone; err != nil {
t.Fatalf("first helper side error = %v", err)
}

if err := client.primeConnection(context.Background(), secondGuestConn, server, true); err != nil {
if _, err := client.primeConnection(context.Background(), secondGuestConn, server, true); err != nil {
t.Fatalf("primeConnection(second) error = %v", err)
}
if err := <-secondDone; err != nil {
t.Fatalf("second helper side error = %v", err)
}
}

func TestGuestVSockClientRunPreservesExecBufferedAfterReady(t *testing.T) {
root := t.TempDir()
server, err := NewServer(root)
if err != nil {
t.Fatalf("NewServer() error = %v", err)
}

guestConn, helperConn := net.Pipe()
client := &VSockClient{
cfg: VSockClientConfig{
VMID: "vm-buffered-exec",
ConnectionToken: "token-buffered-exec",
HostPort: 4242,
WorkspaceRoot: root,
GuestVersion: "1.0.0",
},
dialer: &recordingDialer{conns: []io.ReadWriteCloser{guestConn}},
}

ctx, cancel := context.WithCancel(context.Background())
defer cancel()
helperDone := make(chan error, 1)
go func() {
defer helperConn.Close()
reader := bufio.NewReader(helperConn)

var handshake HandshakeRequest
if err := decodeLine(reader, &handshake); err != nil {
helperDone <- err
return
}
if err := writeJSONLine(helperConn, HandshakeAck{
ProtocolVersion: ProtocolVersion,
RequestID: handshake.RequestID,
Type: "handshake_ack",
Status: "accepted",
VMID: handshake.VMID,
}); err != nil {
helperDone <- err
return
}

var ready ReadyRequest
if err := decodeLine(reader, &ready); err != nil {
helperDone <- err
return
}
readyPayload, err := json.Marshal(ReadyResponse{
ProtocolVersion: ProtocolVersion,
RequestID: ready.RequestID,
Status: "ready",
WorkspaceRoot: root,
})
if err != nil {
helperDone <- err
return
}

execPayload := []byte(`{"protocol_version":"1","request_id":"req-buffered-exec","type":"exec","argv":["/bin/echo","preserved"],"cwd":"."}`)
payload := append(append(readyPayload, '\n'), execPayload...)
payload = append(payload, '\n')
if _, err = helperConn.Write(payload); err != nil {
helperDone <- err
return
}

if err = helperConn.SetReadDeadline(time.Now().Add(time.Second)); err != nil {
Comment thread
rmusser01 marked this conversation as resolved.
Outdated
helperDone <- err
return
}
var execResp ExecResponse
if err = decodeLine(reader, &execResp); err != nil {
helperDone <- err
return
}
if execResp.RequestID != "req-buffered-exec" {
helperDone <- errUnexpectedValue("exec request id", execResp.RequestID)
return
}
if execResp.Stdout != "preserved\n" {
helperDone <- errUnexpectedValue("exec stdout", execResp.Stdout)
return
}
cancel()
helperDone <- nil
}()

if err := client.Run(ctx, server); err != nil {
t.Fatalf("Run() error = %v", err)
}
if err := <-helperDone; err != nil {
t.Fatalf("helper side error = %v", err)
}
}

type unexpectedValueError struct {
name string
value any
Expand Down
Loading