feat: add boundary log forwarding from agent to coderd (#21345)

Add agent forwarding of boundary audit logs from workspaces to coderd
via agent API, and re-emission of boundary logs to coderd stderr. This
change adds a server to the workspace agent that always listens on a
unix socket for boundary to connect and send audit logs.

coderd log format example:
```
[API] 2025-12-23 18:31:46.755 [info] coderd.agentrpc: boundary_request owner=.. workspace_name=.. agent_name=.. decision=.. workspace_id=.. http_method=.. http_url=.. event_time=.. request_id=..
```

Corresponding boundary PR: https://github.com/coder/boundary/pull/124
RFC:
https://www.notion.so/coderhq/Agent-Boundary-Logs-2afd579be59280f29629fc9823ac41ba
https://github.com/coder/coder/issues/21280
This commit is contained in:
Zach
2025-12-31 16:38:19 -07:00
committed by GitHub
parent 21241abc4e
commit 07924037e7
11 changed files with 1339 additions and 15 deletions
+42 -4
View File
@@ -43,6 +43,7 @@ import (
"github.com/coder/coder/v2/agent/agentscripts"
"github.com/coder/coder/v2/agent/agentsocket"
"github.com/coder/coder/v2/agent/agentssh"
"github.com/coder/coder/v2/agent/boundarylogproxy"
"github.com/coder/coder/v2/agent/proto"
"github.com/coder/coder/v2/agent/proto/resourcesmonitor"
"github.com/coder/coder/v2/agent/reconnectingpty"
@@ -102,6 +103,7 @@ type Options struct {
Clock quartz.Clock
SocketServerEnabled bool
SocketPath string // Path for the agent socket server socket
BoundaryLogProxySocketPath string
}
type Client interface {
@@ -205,10 +207,11 @@ func New(options Options) Agent {
metrics: newAgentMetrics(prometheusRegistry),
execer: options.Execer,
devcontainers: options.Devcontainers,
containerAPIOptions: options.DevcontainerAPIOptions,
socketPath: options.SocketPath,
socketServerEnabled: options.SocketServerEnabled,
devcontainers: options.Devcontainers,
containerAPIOptions: options.DevcontainerAPIOptions,
socketPath: options.SocketPath,
socketServerEnabled: options.SocketServerEnabled,
boundaryLogProxySocketPath: options.BoundaryLogProxySocketPath,
}
// Initially, we have a closed channel, reflecting the fact that we are not initially connected.
// Each time we connect we replace the channel (while holding the closeMutex) with a new one
@@ -277,6 +280,11 @@ type agent struct {
logSender *agentsdk.LogSender
// boundaryLogProxy is a socket server that forwards boundary audit logs to coderd.
// It may be nil if there is a problem starting the server.
boundaryLogProxy *boundarylogproxy.Server
boundaryLogProxySocketPath string
prometheusRegistry *prometheus.Registry
// metrics are prometheus registered metrics that will be collected and
// labeled in Coder with the agent + workspace.
@@ -371,6 +379,7 @@ func (a *agent) init() {
)
a.initSocketServer()
a.startBoundaryLogProxyServer()
go a.runLoop()
}
@@ -395,6 +404,19 @@ func (a *agent) initSocketServer() {
a.logger.Debug(a.hardCtx, "socket server started", slog.F("path", a.socketPath))
}
// startBoundaryLogProxyServer starts the boundary log proxy socket server.
func (a *agent) startBoundaryLogProxyServer() {
proxy := boundarylogproxy.NewServer(a.logger, a.boundaryLogProxySocketPath)
if err := proxy.Start(); err != nil {
a.logger.Warn(a.hardCtx, "failed to start boundary log proxy", slog.Error(err))
return
}
a.boundaryLogProxy = proxy
a.logger.Info(a.hardCtx, "boundary log proxy server started",
slog.F("socket_path", a.boundaryLogProxySocketPath))
}
// runLoop attempts to start the agent in a retry loop.
// Coder may be offline temporarily, a connection issue
// may be happening, but regardless after the intermittent
@@ -1012,6 +1034,15 @@ func (a *agent) run() (retErr error) {
return err
})
// Forward boundary audit logs to coderd if boundary log forwarding is enabled.
// These are audit logs so they should continue during graceful shutdown.
if a.boundaryLogProxy != nil {
proxyFunc := func(ctx context.Context, aAPI proto.DRPCAgentClient27) error {
return a.boundaryLogProxy.RunForwarder(ctx, aAPI)
}
connMan.startAgentAPI("boundary log proxy", gracefulShutdownBehaviorRemain, proxyFunc)
}
// part of graceful shut down is reporting the final lifecycle states, e.g "ShuttingDown" so the
// lifecycle reporting has to be via gracefulShutdownBehaviorRemain
connMan.startAgentAPI("report lifecycle", gracefulShutdownBehaviorRemain, a.reportLifecycle)
@@ -1982,6 +2013,13 @@ func (a *agent) Close() error {
a.logger.Error(a.hardCtx, "container API close", slog.Error(err))
}
if a.boundaryLogProxy != nil {
err = a.boundaryLogProxy.Close()
if err != nil {
a.logger.Warn(context.Background(), "close boundary log proxy", slog.Error(err))
}
}
// Wait for the graceful shutdown to complete, but don't wait forever so
// that we don't break user expectations.
go func() {
+165
View File
@@ -0,0 +1,165 @@
//go:build linux || darwin
package agent_test
import (
"context"
"net"
"path/filepath"
"sync"
"testing"
"github.com/google/uuid"
"github.com/stretchr/testify/require"
"google.golang.org/protobuf/proto"
"google.golang.org/protobuf/types/known/timestamppb"
"cdr.dev/slog"
"github.com/coder/coder/v2/agent/boundarylogproxy"
"github.com/coder/coder/v2/agent/boundarylogproxy/codec"
agentproto "github.com/coder/coder/v2/agent/proto"
"github.com/coder/coder/v2/coderd/agentapi"
"github.com/coder/coder/v2/testutil"
)
// logSink captures structured log entries for testing.
type logSink struct {
mu sync.Mutex
entries []slog.SinkEntry
}
func (s *logSink) LogEntry(_ context.Context, e slog.SinkEntry) {
s.mu.Lock()
defer s.mu.Unlock()
s.entries = append(s.entries, e)
}
func (*logSink) Sync() {}
func (s *logSink) getEntries() []slog.SinkEntry {
s.mu.Lock()
defer s.mu.Unlock()
return append([]slog.SinkEntry{}, s.entries...)
}
// getField returns the value of a field by name from a slog.Map.
func getField(fields slog.Map, name string) interface{} {
for _, f := range fields {
if f.Name == name {
return f.Value
}
}
return nil
}
func sendBoundaryLogsRequest(t *testing.T, conn net.Conn, req *agentproto.ReportBoundaryLogsRequest) {
t.Helper()
data, err := proto.Marshal(req)
require.NoError(t, err)
err = codec.WriteFrame(conn, codec.TagV1, data)
require.NoError(t, err)
}
// TestBoundaryLogs_EndToEnd is an end-to-end test that sends a protobuf
// message over the agent's unix socket (as boundary would) and verifies
// it is ultimately logged by coderd with the correct structured fields.
func TestBoundaryLogs_EndToEnd(t *testing.T) {
t.Parallel()
socketPath := filepath.Join(testutil.TempDirUnixSocket(t), "boundary.sock")
srv := boundarylogproxy.NewServer(testutil.Logger(t), socketPath)
err := srv.Start()
require.NoError(t, err)
t.Cleanup(func() { require.NoError(t, srv.Close()) })
sink := &logSink{}
logger := slog.Make(sink)
workspaceID := uuid.New()
reporter := &agentapi.BoundaryLogsAPI{
Log: logger,
WorkspaceID: workspaceID,
}
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
forwarderDone := make(chan error, 1)
go func() {
forwarderDone <- srv.RunForwarder(ctx, reporter)
}()
conn, err := net.Dial("unix", socketPath)
require.NoError(t, err)
defer conn.Close()
// Allowed HTTP request.
req := &agentproto.ReportBoundaryLogsRequest{
Logs: []*agentproto.BoundaryLog{
{
Allowed: true,
Time: timestamppb.Now(),
Resource: &agentproto.BoundaryLog_HttpRequest_{
HttpRequest: &agentproto.BoundaryLog_HttpRequest{
Method: "GET",
Url: "https://example.com/allowed",
MatchedRule: "*.example.com",
},
},
},
},
}
sendBoundaryLogsRequest(t, conn, req)
require.Eventually(t, func() bool {
return len(sink.getEntries()) >= 1
}, testutil.WaitShort, testutil.IntervalFast)
entries := sink.getEntries()
require.Len(t, entries, 1)
entry := entries[0]
require.Equal(t, slog.LevelInfo, entry.Level)
require.Equal(t, "boundary_request", entry.Message)
require.Equal(t, "allow", getField(entry.Fields, "decision"))
require.Equal(t, workspaceID.String(), getField(entry.Fields, "workspace_id"))
require.Equal(t, "GET", getField(entry.Fields, "http_method"))
require.Equal(t, "https://example.com/allowed", getField(entry.Fields, "http_url"))
require.Equal(t, "*.example.com", getField(entry.Fields, "matched_rule"))
// Denied HTTP request.
req2 := &agentproto.ReportBoundaryLogsRequest{
Logs: []*agentproto.BoundaryLog{
{
Allowed: false,
Time: timestamppb.Now(),
Resource: &agentproto.BoundaryLog_HttpRequest_{
HttpRequest: &agentproto.BoundaryLog_HttpRequest{
Method: "POST",
Url: "https://blocked.com/denied",
},
},
},
},
}
sendBoundaryLogsRequest(t, conn, req2)
require.Eventually(t, func() bool {
return len(sink.getEntries()) >= 2
}, testutil.WaitShort, testutil.IntervalFast)
entries = sink.getEntries()
entry = entries[1]
require.Len(t, entries, 2)
require.Equal(t, slog.LevelInfo, entry.Level)
require.Equal(t, "boundary_request", entry.Message)
require.Equal(t, "deny", getField(entry.Fields, "decision"))
require.Equal(t, workspaceID.String(), getField(entry.Fields, "workspace_id"))
require.Equal(t, "POST", getField(entry.Fields, "http_method"))
require.Equal(t, "https://blocked.com/denied", getField(entry.Fields, "http_url"))
require.Equal(t, nil, getField(entry.Fields, "matched_rule"))
cancel()
<-forwarderDone
}
+127
View File
@@ -0,0 +1,127 @@
// Package codec implements the wire format for agent <-> boundary communication.
//
// Wire Format:
// - 8 bits: big-endian tag
// - 24 bits: big-endian length of the protobuf data (bit usage depends on tag)
// - length bytes: encoded protobuf data
//
// Note that while there are 24 bits available for the length, the actual maximum
// length depends on the tag. For TagV1, only 15 bits are used (MaxMessageSizeV1).
package codec
import (
"encoding/binary"
"io"
"golang.org/x/xerrors"
)
type Tag uint8
const (
// TagV1 identifies the first revision of the protocol. This version has a maximum
// data length of MaxMessageSizeV1.
TagV1 Tag = 1
)
const (
// DataLength is the number of bits used for the length of encoded protobuf data.
DataLength = 24
// tagLength is the number of bits used for the tag.
tagLength = 8
// MaxMessageSizeV1 is the maximum size of the encoded protobuf messages sent
// over the wire for the TagV1 tag. While the wire format allows 24 bits for
// length, TagV1 only uses 15 bits.
MaxMessageSizeV1 uint32 = 1 << 15
)
var (
// ErrMessageTooLarge is returned when the message exceeds the maximum size
// allowed for the tag.
ErrMessageTooLarge = xerrors.New("message too large")
// ErrUnsupportedTag is returned when an unrecognized tag is encountered.
ErrUnsupportedTag = xerrors.New("unsupported tag")
)
// WriteFrame writes a framed message with the given tag and data. The data
// must not exceed 2^DataLength in length.
func WriteFrame(w io.Writer, tag Tag, data []byte) error {
var maxSize uint32
switch tag {
case TagV1:
maxSize = MaxMessageSizeV1
default:
return xerrors.Errorf("%w: %d", ErrUnsupportedTag, tag)
}
if len(data) > int(maxSize) {
return xerrors.Errorf("%w for tag %d: %d > %d", ErrMessageTooLarge, tag, len(data), maxSize)
}
var header uint32
//nolint:gosec // The length check above ensures there's no overflow.
header |= uint32(len(data))
header |= uint32(tag) << DataLength
if err := binary.Write(w, binary.BigEndian, header); err != nil {
return xerrors.Errorf("write header error: %w", err)
}
if _, err := w.Write(data); err != nil {
return xerrors.Errorf("write data error: %w", err)
}
return nil
}
// ReadFrame reads a framed message, returning the decoded tag and data. If the
// message size exceeds MaxMessageSizeV1, ErrMessageTooLarge is returned. The
// provided buf is used if it has sufficient capacity; otherwise a new buffer is
// allocated. To reuse the buffer across calls, pass in the returned data slice:
//
// buf := make([]byte, initialSize)
// for {
// _, buf, _ = ReadFrame(r, buf)
// }
func ReadFrame(r io.Reader, buf []byte) (Tag, []byte, error) {
var header uint32
if err := binary.Read(r, binary.BigEndian, &header); err != nil {
return 0, nil, xerrors.Errorf("read header error: %w", err)
}
const lengthMask = (1 << DataLength) - 1
length := header & lengthMask
const tagMask = (1 << tagLength) - 1 // 0xFF
shifted := (header >> DataLength) & tagMask
if shifted > tagMask {
// This is really only here to satisfy the gosec linter. We know from above that
// shifted <= tagMask.
return 0, nil, xerrors.Errorf("invalid tag: %d", shifted)
}
tag := Tag(shifted)
var maxSize uint32
switch tag {
case TagV1:
maxSize = MaxMessageSizeV1
default:
return 0, nil, xerrors.Errorf("%w: %d", ErrUnsupportedTag, tag)
}
if length > maxSize {
return 0, nil, ErrMessageTooLarge
}
if cap(buf) < int(length) {
buf = make([]byte, length)
} else {
buf = buf[:length:cap(buf)]
}
if _, err := io.ReadFull(r, buf[:length]); err != nil {
return 0, nil, xerrors.Errorf("read full error: %w", err)
}
return tag, buf[:length], nil
}
+145
View File
@@ -0,0 +1,145 @@
package codec_test
import (
"bytes"
"encoding/binary"
"io"
"testing"
"github.com/stretchr/testify/require"
"github.com/coder/coder/v2/agent/boundarylogproxy/codec"
)
func TestRoundTrip(t *testing.T) {
t.Parallel()
tests := []struct {
name string
tag codec.Tag
data []byte
}{
{
name: "empty data",
tag: codec.TagV1,
data: []byte{},
},
{
name: "simple data",
tag: codec.TagV1,
data: []byte("hello world"),
},
{
name: "binary data",
tag: codec.TagV1,
data: []byte{0x00, 0x01, 0x02, 0xff, 0xfe},
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
t.Parallel()
var buf bytes.Buffer
err := codec.WriteFrame(&buf, tt.tag, tt.data)
require.NoError(t, err)
readBuf := make([]byte, codec.MaxMessageSizeV1)
tag, data, err := codec.ReadFrame(&buf, readBuf)
require.NoError(t, err)
require.Equal(t, tt.tag, tag)
require.Equal(t, tt.data, data)
})
}
}
func TestReadFrameTooLarge(t *testing.T) {
t.Parallel()
// Hand construct a header that indicates the message size exceeds the maximum
// message size for codec.TagV1 by one. We just write the header to buf because
// we expect codec.ReadFrame to bail out when reading the invalid length.
header := uint32(codec.TagV1)<<codec.DataLength | (codec.MaxMessageSizeV1 + 1)
data := make([]byte, 4)
binary.BigEndian.PutUint32(data, header)
var buf bytes.Buffer
_, err := buf.Write(data)
require.NoError(t, err)
readBuf := make([]byte, 1)
_, _, err = codec.ReadFrame(&buf, readBuf)
require.ErrorIs(t, err, codec.ErrMessageTooLarge)
}
func TestReadFrameEmptyReader(t *testing.T) {
t.Parallel()
var buf bytes.Buffer
readBuf := make([]byte, codec.MaxMessageSizeV1)
_, _, err := codec.ReadFrame(&buf, readBuf)
require.ErrorIs(t, err, io.EOF)
}
func TestReadFrameInvalidTag(t *testing.T) {
t.Parallel()
// Hand construct a header that indicates a tag we don't know about. We just
// write the header to buf because we expect codec.ReadFrame to bail out when
// reading the invalid tag.
const (
dataLength uint32 = 10
bogusTag uint32 = 2
)
header := bogusTag<<codec.DataLength | dataLength
data := make([]byte, 4)
binary.BigEndian.PutUint32(data, header)
var buf bytes.Buffer
_, err := buf.Write(data)
require.NoError(t, err)
readBuf := make([]byte, 1)
_, _, err = codec.ReadFrame(&buf, readBuf)
require.ErrorIs(t, err, codec.ErrUnsupportedTag)
}
func TestReadFrameAllocatesWhenNeeded(t *testing.T) {
t.Parallel()
var buf bytes.Buffer
data := []byte("this message is longer than the buffer")
err := codec.WriteFrame(&buf, codec.TagV1, data)
require.NoError(t, err)
// Buffer with insufficient capacity triggers allocation.
readBuf := make([]byte, 4)
tag, got, err := codec.ReadFrame(&buf, readBuf)
require.NoError(t, err)
require.Equal(t, codec.TagV1, tag)
require.Equal(t, data, got)
}
func TestWriteFrameDataSize(t *testing.T) {
t.Parallel()
var buf bytes.Buffer
data := make([]byte, codec.MaxMessageSizeV1)
err := codec.WriteFrame(&buf, codec.TagV1, data)
require.NoError(t, err)
//nolint: makezero // This intentionally increases the slice length.
data = append(data, 0) // One byte over the maximum
err = codec.WriteFrame(&buf, codec.TagV1, data)
require.ErrorIs(t, err, codec.ErrMessageTooLarge)
}
func TestWriteFrameInvalidTag(t *testing.T) {
t.Parallel()
var buf bytes.Buffer
data := make([]byte, 1)
const bogusTag = 2
err := codec.WriteFrame(&buf, codec.Tag(bogusTag), data)
require.ErrorIs(t, err, codec.ErrUnsupportedTag)
}
+201 -3
View File
@@ -2,6 +2,204 @@
// audit logs and forwards them to coderd via the agent API.
package boundarylogproxy
// Server a placeholder for the server that will listen on a Unix socket for
// boundary logs to be forwarded.
type Server struct{}
import (
"context"
"errors"
"io"
"net"
"os"
"path/filepath"
"sync"
"golang.org/x/xerrors"
"google.golang.org/protobuf/proto"
"cdr.dev/slog"
"github.com/coder/coder/v2/agent/boundarylogproxy/codec"
agentproto "github.com/coder/coder/v2/agent/proto"
)
const (
// logBufferSize is the size of the channel buffer for incoming log requests
// from workspaces. This buffer size is intended to handle short bursts of workspaces
// forwarding batches of logs in parallel.
logBufferSize = 100
)
// DefaultSocketPath returns the default path for the boundary audit log socket.
func DefaultSocketPath() string {
return filepath.Join(os.TempDir(), "boundary-audit.sock")
}
// Reporter reports boundary logs from workspaces.
type Reporter interface {
ReportBoundaryLogs(ctx context.Context, req *agentproto.ReportBoundaryLogsRequest) (*agentproto.ReportBoundaryLogsResponse, error)
}
// Server listens on a Unix socket for boundary log messages and buffers them
// for forwarding to coderd. The socket server and the forwarder are decoupled:
// - Start() creates the socket and accepts a connection from boundary
// - RunForwarder() drains the buffer and sends logs to coderd via AgentAPI
type Server struct {
logger slog.Logger
socketPath string
listener net.Listener
cancel context.CancelFunc
wg sync.WaitGroup
// logs buffers incoming log requests for the forwarder to drain.
logs chan *agentproto.ReportBoundaryLogsRequest
}
// NewServer creates a new boundary log proxy server.
func NewServer(logger slog.Logger, socketPath string) *Server {
return &Server{
logger: logger.Named("boundary-log-proxy"),
socketPath: socketPath,
logs: make(chan *agentproto.ReportBoundaryLogsRequest, logBufferSize),
}
}
// Start begins listening for connections on the Unix socket, and handles new
// connections in a separate goroutine. Incoming logs from connections are
// buffered until RunForwarder drains them.
func (s *Server) Start() error {
if err := os.Remove(s.socketPath); err != nil && !os.IsNotExist(err) {
return xerrors.Errorf("remove existing socket: %w", err)
}
listener, err := net.Listen("unix", s.socketPath)
if err != nil {
return xerrors.Errorf("listen on socket: %w", err)
}
s.listener = listener
ctx, cancel := context.WithCancel(context.Background())
s.cancel = cancel
s.wg.Add(1)
go s.acceptLoop(ctx)
s.logger.Info(ctx, "boundary log proxy started", slog.F("socket_path", s.socketPath))
return nil
}
// RunForwarder drains the log buffer and forwards logs to coderd.
// It blocks until ctx is canceled.
func (s *Server) RunForwarder(ctx context.Context, sender Reporter) error {
s.logger.Debug(ctx, "boundary log forwarder started")
for {
select {
case <-ctx.Done():
return ctx.Err()
case req := <-s.logs:
_, err := sender.ReportBoundaryLogs(ctx, req)
if err != nil {
s.logger.Warn(ctx, "failed to forward boundary logs",
slog.Error(err),
slog.F("log_count", len(req.Logs)))
// Continue forwarding other logs. The current batch is lost,
// but the socket stays alive.
}
}
}
}
func (s *Server) acceptLoop(ctx context.Context) {
defer s.wg.Done()
for {
conn, err := s.listener.Accept()
if err != nil {
if ctx.Err() != nil {
s.logger.Warn(ctx, "accept loop terminated", slog.Error(ctx.Err()))
return
}
s.logger.Warn(ctx, "socket accept error", slog.Error(err))
continue
}
s.wg.Add(1)
go s.handleConnection(ctx, conn)
}
}
func (s *Server) handleConnection(ctx context.Context, conn net.Conn) {
defer s.wg.Done()
ctx, cancel := context.WithCancel(ctx)
defer cancel()
s.wg.Add(1)
go func() {
defer s.wg.Done()
<-ctx.Done()
_ = conn.Close()
}()
// This is intended to be a sane starting point for the read buffer size. It may be
// grown by codec.ReadFrame if necessary.
const initBufSize = 1 << 10
buf := make([]byte, initBufSize)
for {
select {
case <-ctx.Done():
return
default:
}
var (
tag codec.Tag
err error
)
tag, buf, err = codec.ReadFrame(conn, buf)
switch {
case errors.Is(err, io.EOF) || errors.Is(err, net.ErrClosed):
return
case err != nil:
s.logger.Warn(ctx, "read frame error", slog.Error(err))
return
}
if tag != codec.TagV1 {
s.logger.Warn(ctx, "invalid tag value", slog.F("tag", tag))
return
}
var req agentproto.ReportBoundaryLogsRequest
if err := proto.Unmarshal(buf, &req); err != nil {
s.logger.Warn(ctx, "proto unmarshal error", slog.Error(err))
continue
}
select {
case s.logs <- &req:
default:
s.logger.Warn(ctx, "dropping boundary logs, buffer full",
slog.F("log_count", len(req.Logs)))
}
}
}
// Close stops the server and blocks until resources have been cleaned up.
// It must be called after Start.
func (s *Server) Close() error {
if s.cancel != nil {
s.cancel()
}
if s.listener != nil {
_ = s.listener.Close()
}
s.wg.Wait()
err := os.Remove(s.socketPath)
if err != nil && !errors.Is(err, os.ErrNotExist) {
return err
}
return nil
}
+578
View File
@@ -0,0 +1,578 @@
//go:build linux || darwin
package boundarylogproxy_test
import (
"context"
"encoding/binary"
"net"
"path/filepath"
"sync"
"testing"
"time"
"github.com/stretchr/testify/require"
"google.golang.org/protobuf/proto"
"google.golang.org/protobuf/types/known/timestamppb"
"github.com/coder/coder/v2/agent/boundarylogproxy"
"github.com/coder/coder/v2/agent/boundarylogproxy/codec"
agentproto "github.com/coder/coder/v2/agent/proto"
"github.com/coder/coder/v2/testutil"
)
// sendMessage writes a framed protobuf message to the connection.
func sendMessage(t *testing.T, conn net.Conn, req *agentproto.ReportBoundaryLogsRequest) {
t.Helper()
data, err := proto.Marshal(req)
if err != nil {
//nolint:gocritic // In tests we're not worried about conn being nil.
t.Errorf("%s marshal req: %s", conn.LocalAddr().String(), err)
}
err = codec.WriteFrame(conn, codec.TagV1, data)
if err != nil {
//nolint:gocritic // In tests we're not worried about conn being nil.
t.Errorf("%s write frame: %s", conn.LocalAddr().String(), err)
}
}
// fakeReporter implements boundarylogproxy.Reporter for testing.
type fakeReporter struct {
mu sync.Mutex
logs []*agentproto.ReportBoundaryLogsRequest
err error
errOnce bool // only error once, then succeed
// reportCb is called when a ReportBoundaryLogsRequest is processed. It must not
// block.
reportCb func()
}
func (f *fakeReporter) ReportBoundaryLogs(_ context.Context, req *agentproto.ReportBoundaryLogsRequest) (*agentproto.ReportBoundaryLogsResponse, error) {
f.mu.Lock()
defer f.mu.Unlock()
if f.reportCb != nil {
f.reportCb()
}
if f.err != nil {
if f.errOnce {
err := f.err
f.err = nil
return nil, err
}
return nil, f.err
}
f.logs = append(f.logs, req)
return &agentproto.ReportBoundaryLogsResponse{}, nil
}
func (f *fakeReporter) getLogs() []*agentproto.ReportBoundaryLogsRequest {
f.mu.Lock()
defer f.mu.Unlock()
return append([]*agentproto.ReportBoundaryLogsRequest{}, f.logs...)
}
func TestServer_StartAndClose(t *testing.T) {
t.Parallel()
socketPath := filepath.Join(testutil.TempDirUnixSocket(t), "boundary.sock")
srv := boundarylogproxy.NewServer(testutil.Logger(t), socketPath)
err := srv.Start()
require.NoError(t, err)
// Verify socket exists and is connectable.
conn, err := net.Dial("unix", socketPath)
require.NoError(t, err)
err = conn.Close()
require.NoError(t, err)
err = srv.Close()
require.NoError(t, err)
}
func TestServer_ReceiveAndForwardLogs(t *testing.T) {
t.Parallel()
socketPath := filepath.Join(testutil.TempDirUnixSocket(t), "boundary.sock")
srv := boundarylogproxy.NewServer(testutil.Logger(t), socketPath)
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
err := srv.Start()
require.NoError(t, err)
t.Cleanup(func() { require.NoError(t, srv.Close()) })
reporter := &fakeReporter{}
// Start forwarder in background.
forwarderDone := make(chan error, 1)
go func() {
forwarderDone <- srv.RunForwarder(ctx, reporter)
}()
// Connect and send a log message.
conn, err := net.Dial("unix", socketPath)
require.NoError(t, err)
defer conn.Close()
req := &agentproto.ReportBoundaryLogsRequest{
Logs: []*agentproto.BoundaryLog{
{
Allowed: true,
Time: timestamppb.Now(),
Resource: &agentproto.BoundaryLog_HttpRequest_{
HttpRequest: &agentproto.BoundaryLog_HttpRequest{
Method: "GET",
Url: "https://example.com",
},
},
},
},
}
sendMessage(t, conn, req)
// Wait for the reporter to receive the log.
require.Eventually(t, func() bool {
logs := reporter.getLogs()
return len(logs) == 1
}, testutil.WaitShort, testutil.IntervalFast)
logs := reporter.getLogs()
require.Len(t, logs, 1)
require.Len(t, logs[0].Logs, 1)
require.True(t, logs[0].Logs[0].Allowed)
require.Equal(t, "GET", logs[0].Logs[0].GetHttpRequest().Method)
require.Equal(t, "https://example.com", logs[0].Logs[0].GetHttpRequest().Url)
cancel()
<-forwarderDone
}
func TestServer_MultipleMessages(t *testing.T) {
t.Parallel()
socketPath := filepath.Join(testutil.TempDirUnixSocket(t), "boundary.sock")
srv := boundarylogproxy.NewServer(testutil.Logger(t), socketPath)
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
err := srv.Start()
require.NoError(t, err)
defer srv.Close()
reporter := &fakeReporter{}
forwarderDone := make(chan error, 1)
go func() {
forwarderDone <- srv.RunForwarder(ctx, reporter)
}()
conn, err := net.Dial("unix", socketPath)
require.NoError(t, err)
defer conn.Close()
// Send multiple messages and verify they are all received.
for range 5 {
req := &agentproto.ReportBoundaryLogsRequest{
Logs: []*agentproto.BoundaryLog{
{
Allowed: true,
Time: timestamppb.Now(),
Resource: &agentproto.BoundaryLog_HttpRequest_{
HttpRequest: &agentproto.BoundaryLog_HttpRequest{
Method: "POST",
Url: "https://example.com/api",
},
},
},
},
}
sendMessage(t, conn, req)
}
require.Eventually(t, func() bool {
logs := reporter.getLogs()
return len(logs) == 5
}, testutil.WaitShort, testutil.IntervalFast)
cancel()
<-forwarderDone
}
func TestServer_MultipleConnections(t *testing.T) {
t.Parallel()
socketPath := filepath.Join(testutil.TempDirUnixSocket(t), "boundary.sock")
srv := boundarylogproxy.NewServer(testutil.Logger(t), socketPath)
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
err := srv.Start()
require.NoError(t, err)
t.Cleanup(func() { require.NoError(t, srv.Close()) })
reporter := &fakeReporter{}
forwarderDone := make(chan error, 1)
go func() {
forwarderDone <- srv.RunForwarder(ctx, reporter)
}()
// Create multiple connections and send from each.
const numConns = 3
var wg sync.WaitGroup
wg.Add(numConns)
for i := range numConns {
go func(connID int) {
defer wg.Done()
conn, err := net.Dial("unix", socketPath)
if err != nil {
t.Errorf("conn %d dial: %s", connID, err)
}
defer conn.Close()
req := &agentproto.ReportBoundaryLogsRequest{
Logs: []*agentproto.BoundaryLog{
{
Allowed: true,
Time: timestamppb.Now(),
Resource: &agentproto.BoundaryLog_HttpRequest_{
HttpRequest: &agentproto.BoundaryLog_HttpRequest{
Method: "GET",
Url: "https://example.com",
},
},
},
},
}
sendMessage(t, conn, req)
}(i)
}
wg.Wait()
require.Eventually(t, func() bool {
logs := reporter.getLogs()
return len(logs) == numConns
}, testutil.WaitShort, testutil.IntervalFast)
cancel()
<-forwarderDone
}
func TestServer_MessageTooLarge(t *testing.T) {
t.Parallel()
socketPath := filepath.Join(testutil.TempDirUnixSocket(t), "boundary.sock")
srv := boundarylogproxy.NewServer(testutil.Logger(t), socketPath)
err := srv.Start()
require.NoError(t, err)
t.Cleanup(func() { require.NoError(t, srv.Close()) })
conn, err := net.Dial("unix", socketPath)
require.NoError(t, err)
defer conn.Close()
// Send a message claiming to be larger than the max message size.
var length uint32 = codec.MaxMessageSizeV1 + 1
err = binary.Write(conn, binary.BigEndian, length)
require.NoError(t, err)
// The server should close the connection after receiving an oversized
// message length.
buf := make([]byte, 1)
err = conn.SetReadDeadline(time.Now().Add(time.Second))
require.NoError(t, err)
_, err = conn.Read(buf)
require.Error(t, err) // Should get EOF or closed connection.
}
func TestServer_ForwarderContinuesAfterError(t *testing.T) {
t.Parallel()
socketPath := filepath.Join(testutil.TempDirUnixSocket(t), "boundary.sock")
srv := boundarylogproxy.NewServer(testutil.Logger(t), socketPath)
err := srv.Start()
require.NoError(t, err)
t.Cleanup(func() { require.NoError(t, srv.Close()) })
reportNotify := make(chan struct{}, 1)
reporter := &fakeReporter{
// Simulate an error on the first call.
err: context.DeadlineExceeded,
errOnce: true,
reportCb: func() {
reportNotify <- struct{}{}
},
}
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
forwarderDone := make(chan error, 1)
go func() {
forwarderDone <- srv.RunForwarder(ctx, reporter)
}()
conn, err := net.Dial("unix", socketPath)
require.NoError(t, err)
defer conn.Close()
// Send the first message to be processed and wait for failure.
req1 := &agentproto.ReportBoundaryLogsRequest{
Logs: []*agentproto.BoundaryLog{
{
Allowed: true,
Time: timestamppb.Now(),
Resource: &agentproto.BoundaryLog_HttpRequest_{
HttpRequest: &agentproto.BoundaryLog_HttpRequest{
Method: "GET",
Url: "https://example.com/first",
},
},
},
},
}
sendMessage(t, conn, req1)
select {
case <-reportNotify:
case <-time.After(testutil.WaitShort):
t.Fatal("timed out waiting for first message to be processed")
}
// Send the second message, which should succeed.
req2 := &agentproto.ReportBoundaryLogsRequest{
Logs: []*agentproto.BoundaryLog{
{
Allowed: false,
Time: timestamppb.Now(),
Resource: &agentproto.BoundaryLog_HttpRequest_{
HttpRequest: &agentproto.BoundaryLog_HttpRequest{
Method: "POST",
Url: "https://example.com/second",
},
},
},
},
}
sendMessage(t, conn, req2)
// Only the second message should be recorded.
require.Eventually(t, func() bool {
logs := reporter.getLogs()
return len(logs) == 1
}, testutil.WaitShort, testutil.IntervalFast)
logs := reporter.getLogs()
require.Len(t, logs, 1)
require.Equal(t, "https://example.com/second", logs[0].Logs[0].GetHttpRequest().Url)
cancel()
<-forwarderDone
}
func TestServer_CloseStopsForwarder(t *testing.T) {
t.Parallel()
socketPath := filepath.Join(testutil.TempDirUnixSocket(t), "boundary.sock")
srv := boundarylogproxy.NewServer(testutil.Logger(t), socketPath)
err := srv.Start()
require.NoError(t, err)
t.Cleanup(func() { require.NoError(t, srv.Close()) })
reporter := &fakeReporter{}
forwarderCtx, forwarderCancel := context.WithCancel(context.Background())
forwarderDone := make(chan error, 1)
go func() {
forwarderDone <- srv.RunForwarder(forwarderCtx, reporter)
}()
// Cancel the forwarder context and verify it stops.
forwarderCancel()
select {
case err := <-forwarderDone:
require.ErrorIs(t, err, context.Canceled)
case <-time.After(testutil.WaitShort):
t.Fatal("forwarder did not stop")
}
}
func TestServer_InvalidProtobuf(t *testing.T) {
t.Parallel()
socketPath := filepath.Join(testutil.TempDirUnixSocket(t), "boundary.sock")
srv := boundarylogproxy.NewServer(testutil.Logger(t), socketPath)
err := srv.Start()
require.NoError(t, err)
t.Cleanup(func() { require.NoError(t, srv.Close()) })
reporter := &fakeReporter{}
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
forwarderDone := make(chan error, 1)
go func() {
forwarderDone <- srv.RunForwarder(ctx, reporter)
}()
conn, err := net.Dial("unix", socketPath)
require.NoError(t, err)
defer conn.Close()
// Send a valid header with garbage protobuf data.
// The server should log an unmarshal error but continue processing.
invalidProto := []byte{0xFF, 0xFF, 0xFF, 0xFF, 0xFF}
//nolint: gosec // codec.DataLength is always less than the size of the header.
header := (uint32(codec.TagV1) << codec.DataLength) | uint32(len(invalidProto))
err = binary.Write(conn, binary.BigEndian, header)
require.NoError(t, err)
_, err = conn.Write(invalidProto)
require.NoError(t, err)
// Now send a valid message. The server should continue processing.
req := &agentproto.ReportBoundaryLogsRequest{
Logs: []*agentproto.BoundaryLog{
{
Allowed: true,
Time: timestamppb.Now(),
Resource: &agentproto.BoundaryLog_HttpRequest_{
HttpRequest: &agentproto.BoundaryLog_HttpRequest{
Method: "GET",
Url: "https://example.com/valid",
},
},
},
},
}
sendMessage(t, conn, req)
require.Eventually(t, func() bool {
logs := reporter.getLogs()
return len(logs) == 1
}, testutil.WaitShort, testutil.IntervalFast)
cancel()
<-forwarderDone
}
func TestServer_InvalidHeader(t *testing.T) {
t.Parallel()
socketPath := filepath.Join(testutil.TempDirUnixSocket(t), "boundary.sock")
srv := boundarylogproxy.NewServer(testutil.Logger(t), socketPath)
err := srv.Start()
require.NoError(t, err)
t.Cleanup(func() { require.NoError(t, srv.Close()) })
reporter := &fakeReporter{}
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
forwarderDone := make(chan error, 1)
go func() {
forwarderDone <- srv.RunForwarder(ctx, reporter)
}()
// sendInvalidHeader sends a header and verifies the server closes the
// connection.
sendInvalidHeader := func(t *testing.T, name string, header uint32) {
t.Helper()
conn, err := net.Dial("unix", socketPath)
require.NoError(t, err)
defer conn.Close()
err = binary.Write(conn, binary.BigEndian, header)
require.NoError(t, err, name)
// The server closes the connection on invalid header, so the next
// write should fail with a broken pipe error.
require.Eventually(t, func() bool {
_, err := conn.Write([]byte{0x00})
return err != nil
}, testutil.WaitShort, testutil.IntervalFast, name)
}
// TagV1 with length exceeding MaxMessageSizeV1.
sendInvalidHeader(t, "v1 too large", (uint32(codec.TagV1)<<codec.DataLength)|(codec.MaxMessageSizeV1+1))
// Unknown tag.
const bogusTag = 0xFF
sendInvalidHeader(t, "unknown tag too large", (bogusTag<<codec.DataLength)|(codec.MaxMessageSizeV1+1))
cancel()
<-forwarderDone
}
func TestServer_AllowRequest(t *testing.T) {
t.Parallel()
socketPath := filepath.Join(testutil.TempDirUnixSocket(t), "boundary.sock")
srv := boundarylogproxy.NewServer(testutil.Logger(t), socketPath)
err := srv.Start()
require.NoError(t, err)
t.Cleanup(func() { require.NoError(t, srv.Close()) })
reporter := &fakeReporter{}
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
forwarderDone := make(chan error, 1)
go func() {
forwarderDone <- srv.RunForwarder(ctx, reporter)
}()
conn, err := net.Dial("unix", socketPath)
require.NoError(t, err)
defer conn.Close()
// Send an allowed request with a matched rule.
logTime := timestamppb.Now()
req := &agentproto.ReportBoundaryLogsRequest{
Logs: []*agentproto.BoundaryLog{
{
Allowed: true,
Time: logTime,
Resource: &agentproto.BoundaryLog_HttpRequest_{
HttpRequest: &agentproto.BoundaryLog_HttpRequest{
Method: "GET",
Url: "https://malicious.com/attack",
MatchedRule: "*.malicious.com",
},
},
},
},
}
sendMessage(t, conn, req)
require.Eventually(t, func() bool {
logs := reporter.getLogs()
return len(logs) == 1
}, testutil.WaitShort, testutil.IntervalFast)
logs := reporter.getLogs()
require.Len(t, logs, 1)
require.True(t, logs[0].Logs[0].Allowed)
require.Equal(t, logTime.Seconds, logs[0].Logs[0].Time.Seconds)
require.Equal(t, logTime.Nanos, logs[0].Logs[0].Time.Nanos)
require.Equal(t, "*.malicious.com", logs[0].Logs[0].GetHttpRequest().MatchedRule)
cancel()
<-forwarderDone
}
+12 -2
View File
@@ -31,6 +31,7 @@ import (
"github.com/coder/coder/v2/agent/agentcontainers"
"github.com/coder/coder/v2/agent/agentexec"
"github.com/coder/coder/v2/agent/agentssh"
"github.com/coder/coder/v2/agent/boundarylogproxy"
"github.com/coder/coder/v2/agent/reaper"
"github.com/coder/coder/v2/buildinfo"
"github.com/coder/coder/v2/cli/clilog"
@@ -59,6 +60,7 @@ func workspaceAgent() *serpent.Command {
devcontainerDiscoveryAutostart bool
socketServerEnabled bool
socketPath string
boundaryLogProxySocketPath string
)
agentAuth := &AgentAuth{}
cmd := &serpent.Command{
@@ -319,8 +321,9 @@ func workspaceAgent() *serpent.Command {
agentcontainers.WithProjectDiscovery(devcontainerProjectDiscovery),
agentcontainers.WithDiscoveryAutostart(devcontainerDiscoveryAutostart),
},
SocketPath: socketPath,
SocketServerEnabled: socketServerEnabled,
SocketPath: socketPath,
SocketServerEnabled: socketServerEnabled,
BoundaryLogProxySocketPath: boundaryLogProxySocketPath,
})
if debugAddress != "" {
@@ -494,6 +497,13 @@ func workspaceAgent() *serpent.Command {
Description: "Specify the path for the agent socket.",
Value: serpent.StringOf(&socketPath),
},
{
Flag: "boundary-log-proxy-socket-path",
Default: boundarylogproxy.DefaultSocketPath(),
Env: "CODER_AGENT_BOUNDARY_LOG_PROXY_SOCKET_PATH",
Description: "The path for the boundary log proxy server Unix socket. Boundary should write audit logs to this socket.",
Value: serpent.StringOf(&boundaryLogProxySocketPath),
},
}
agentAuth.AttachOptions(cmd, false)
return cmd
+11 -1
View File
@@ -138,6 +138,17 @@ func normalizeGoldenFile(t *testing.T, byt []byte) []byte {
// The home directory changes depending on the test environment.
byt = bytes.ReplaceAll(byt, []byte(homeDir), []byte("~"))
// Normalize the temp directory. os.TempDir() may include a trailing slash
// (macOS) or not (Linux/Windows), and the temp directory may be followed by
// more filepath elements with an OS-specific separator. We handle all cases
// by replacing tempdir+separator first, then tempdir alone.
tempDir := filepath.Clean(os.TempDir())
byt = bytes.ReplaceAll(byt, []byte(tempDir+string(filepath.Separator)), []byte("/tmp/"))
byt = bytes.ReplaceAll(byt, []byte(tempDir), []byte("/tmp"))
// Clean up trailing slash when temp dir is used standalone (e.g., "/tmp/)" -> "/tmp)").
byt = bytes.ReplaceAll(byt, []byte("/tmp/)"), []byte("/tmp)"))
for _, r := range []struct {
old string
new string
@@ -145,7 +156,6 @@ func normalizeGoldenFile(t *testing.T, byt []byte) []byte {
{"\r\n", "\n"},
{`~\.cache\coder`, "~/.cache/coder"},
{`C:\Users\RUNNER~1\AppData\Local\Temp`, "/tmp"},
{os.TempDir(), "/tmp"},
} {
byt = bytes.ReplaceAll(byt, []byte(r.old), []byte(r.new))
}
+4
View File
@@ -39,6 +39,10 @@ OPTIONS:
--block-file-transfer bool, $CODER_AGENT_BLOCK_FILE_TRANSFER (default: false)
Block file transfer using known applications: nc,rsync,scp,sftp.
--boundary-log-proxy-socket-path string, $CODER_AGENT_BOUNDARY_LOG_PROXY_SOCKET_PATH (default: /tmp/boundary-audit.sock)
The path for the boundary log proxy server Unix socket. Boundary
should write audit logs to this socket.
--debug-address string, $CODER_AGENT_DEBUG_ADDRESS (default: 127.0.0.1:2113)
The bind address to serve a debug HTTP server.
+4 -1
View File
@@ -220,7 +220,10 @@ func New(opts Options, workspace database.Workspace) *API {
Database: opts.Database,
}
api.BoundaryLogsAPI = &BoundaryLogsAPI{}
api.BoundaryLogsAPI = &BoundaryLogsAPI{
Log: opts.Log,
WorkspaceID: opts.WorkspaceID,
}
// Start background cache refresh loop to handle workspace changes
// like prebuild claims where owner_id and other fields may be modified in the DB.
+50 -4
View File
@@ -2,14 +2,60 @@ package agentapi
import (
"context"
"time"
"golang.org/x/xerrors"
"github.com/google/uuid"
"cdr.dev/slog"
agentproto "github.com/coder/coder/v2/agent/proto"
)
type BoundaryLogsAPI struct{}
type BoundaryLogsAPI struct {
Log slog.Logger
WorkspaceID uuid.UUID
}
func (*BoundaryLogsAPI) ReportBoundaryLogs(context.Context, *agentproto.ReportBoundaryLogsRequest) (*agentproto.ReportBoundaryLogsResponse, error) {
return nil, xerrors.New("not implemented")
func (a *BoundaryLogsAPI) ReportBoundaryLogs(ctx context.Context, req *agentproto.ReportBoundaryLogsRequest) (*agentproto.ReportBoundaryLogsResponse, error) {
for _, l := range req.Logs {
var logTime time.Time
if l.Time != nil {
logTime = l.Time.AsTime()
}
switch r := l.Resource.(type) {
case *agentproto.BoundaryLog_HttpRequest_:
if r.HttpRequest == nil {
a.Log.Warn(ctx, "empty http request resource",
slog.F("workspace_id", a.WorkspaceID.String()))
continue
}
fields := []slog.Field{
slog.F("decision", allowBoolToString(l.Allowed)),
slog.F("workspace_id", a.WorkspaceID.String()),
slog.F("http_method", r.HttpRequest.Method),
slog.F("http_url", r.HttpRequest.Url),
slog.F("event_time", logTime.Format(time.RFC3339Nano)),
}
if l.Allowed {
fields = append(fields, slog.F("matched_rule", r.HttpRequest.MatchedRule))
}
a.Log.With(fields...).Info(ctx, "boundary_request")
default:
a.Log.Warn(ctx, "unknown resource type",
slog.F("workspace_id", a.WorkspaceID.String()))
}
}
return &agentproto.ReportBoundaryLogsResponse{}, nil
}
//nolint:revive // This stringifies the boolean argument.
func allowBoolToString(b bool) string {
if b {
return "allow"
}
return "deny"
}