fix: stop the job's docker socket from becoming a directory (#1215)

Fixes https://gitea.com/gitea/runner/issues/1213

Fix the DooD regression that mounts `/var/run/docker.sock` as a directory. Keep the Docker proxy available through job and post steps. Clean stale resources before opening it, then remove containers before their networks and volumes during teardown.

Use a unique filesystem probe and preserve socket ownership. Fall back to direct access when proxying is unsupported. Preserve exec output and clean up active streams and failed starts.

Add a real Docker job test for mounted socket access, post steps and resource cleanup.

---------

Co-authored-by: silverwind <[email protected]>
Reviewed-on: https://gitea.com/gitea/runner/pulls/1215
Reviewed-by: silverwind <[email protected]>
Co-authored-by: Zettat123 <[email protected]>
This commit is contained in:
Zettat123andsilverwind authored and bircni committed 2026-09-08 04:16:02 +00:00
1 parent ba4d3c5b4f
commit ff9965e940
18 files changed
+935 -295

No files matched your search

+247 -48
View File
@@ -10,18 +10,23 @@ import (
"bufio"
"bytes"
"context"
"encoding/json"
"errors"
"fmt"
"io"
"net"
"net/http"
"os"
"path/filepath"
"runtime"
"strconv"
"strings"
"testing"
"time"
cerrdefs "github.com/containerd/errdefs"
"github.com/moby/moby/api/types/container"
"github.com/moby/moby/api/types/network"
"github.com/moby/moby/api/types/image"
"github.com/moby/moby/api/types/mount"
"github.com/moby/moby/api/types/volume"
mobyclient "github.com/moby/moby/client"
"github.com/stretchr/testify/assert"
@@ -56,7 +61,12 @@ func daemonSocketPath(t testing.TB, cli mobyclient.APIClient) string {
}
func TestDockerProxy(t *testing.T) {
if runtime.GOOS == "windows" {
t.Skip("Unix socket ownership is unavailable on Windows")
}
bodies := make(chan []byte, 8)
createStarted, releaseCreate := make(chan struct{}), make(chan struct{})
streamsDone := make(chan struct{}, 2)
daemonSocket := filepath.Join(shortTempDir(t), "d.sock")
listener, err := net.Listen("unix", daemonSocket)
require.NoError(t, err)
@@ -64,16 +74,33 @@ func TestDockerProxy(t *testing.T) {
switch {
case strings.HasSuffix(r.URL.Path, "/create"):
body, _ := io.ReadAll(r.Body)
if r.URL.RawQuery == "wait" {
close(createStarted)
<-releaseCreate
}
bodies <- body
w.WriteHeader(http.StatusCreated)
case strings.HasSuffix(r.URL.Path, "/start"):
case r.URL.Path == "/exec/detached/start" || r.URL.Path == "/exec/error/start":
if r.URL.Path == "/exec/error/start" {
w.Header().Set("Content-Type", "application/vnd.docker.raw-stream")
w.WriteHeader(http.StatusBadRequest)
}
_, _ = w.Write([]byte("ordinary"))
case rawStreamPath.MatchString(r.URL.Path) || r.URL.Path == "/session":
_, _ = io.Copy(io.Discard, r.Body)
conn, buffered, err := w.(http.Hijacker).Hijack()
if err != nil {
return
}
defer conn.Close()
_, _ = conn.Write([]byte("HTTP/1.1 200 OK\r\nContent-Type: application/vnd.docker.raw-stream\r\n\r\n"))
_, _ = io.Copy(conn, buffered)
if r.Header.Get("Upgrade") != "" {
_, _ = conn.Write([]byte("HTTP/1.1 101 Switching Protocols\r\nConnection: Upgrade\r\nUpgrade: tcp\r\n\r\nready\n"))
} else {
_, _ = conn.Write([]byte("HTTP/1.1 200 OK\r\nContent-Type: application/vnd.docker.raw-stream\r\n\r\nready\n"))
}
input, _ := io.ReadAll(buffered)
_, _ = conn.Write(append([]byte("final:"), input...))
streamsDone <- struct{}{}
default:
w.Header().Set("Api-Version", "1.47")
_, _ = w.Write([]byte("OK " + r.Method + " " + r.URL.Path))
@@ -83,34 +110,69 @@ func TestDockerProxy(t *testing.T) {
t.Cleanup(func() { _ = daemon.Close() })
proxy, err := StartDockerProxy(daemonSocket, shortTempDir(t), "job-1")
require.NoError(t, err)
client := &http.Client{Transport: &http.Transport{DialContext: func(ctx context.Context, _, _ string) (net.Conn, error) {
t.Cleanup(func() { _ = proxy.Close(context.Background()) })
client := &http.Client{Timeout: 5 * time.Second, Transport: &http.Transport{DialContext: func(ctx context.Context, _, _ string) (net.Conn, error) {
return (&net.Dialer{}).DialContext(ctx, "unix", proxy.Socket)
}}}
t.Cleanup(client.CloseIdleConnections)
dialProxy := func(t *testing.T) *net.UnixConn {
t.Helper()
conn, err := net.DialUnix("unix", nil, &net.UnixAddr{Name: proxy.Socket, Net: "unix"})
require.NoError(t, err)
t.Cleanup(func() { _ = conn.Close() })
require.NoError(t, conn.SetDeadline(time.Now().Add(5*time.Second)))
return conn
}
startStream := func(t *testing.T, path, upgrade string) (*net.UnixConn, *bufio.Reader) {
t.Helper()
conn := dialProxy(t)
_, err := fmt.Fprintf(conn, "POST %s HTTP/1.1\r\nHost: docker\r\n%sContent-Length: 2\r\n\r\n{}input", path, upgrade)
require.NoError(t, err)
reader := bufio.NewReader(conn)
resp, err := http.ReadResponse(reader, nil)
require.NoError(t, err)
t.Cleanup(func() {
_ = conn.Close()
_ = resp.Body.Close()
})
prefix, err := reader.ReadString('\n')
require.NoError(t, err)
require.Equal(t, "ready\n", prefix)
return conn, reader
}
t.Run("labels creates", func(t *testing.T) {
for path, body := range map[string]string{
"/v1.47/containers/create": `{"Image":"alpine","Labels":{"own":"1"}}`,
"/networks/create": `{"Name":"n"}`,
"/volumes/create": `{"Name":"v","labels":null}`,
for _, testCase := range []struct{ path, body, want string }{
{
"/v1.47/containers/create", `{"Image":"alpine","Unknown":{"enabled":true},"Labels":{"own":"1","com.gitea.runner.job":"other"}}`,
`{"Image":"alpine","Unknown":{"enabled":true},"Labels":{"own":"1","com.gitea.runner.job":"job-1"}}`,
},
{
"/networks/create", `{"Name":"n"}`,
`{"Name":"n","Labels":{"com.gitea.runner.job":"job-1"}}`,
},
{
"/volumes/create", `{"Name":"v","labels":null}`,
`{"Name":"v","Labels":{"com.gitea.runner.job":"job-1"}}`,
},
{
"/volumes/create", "",
`{"Labels":{"com.gitea.runner.job":"job-1"}}`,
},
{
"/volumes/create", "null",
`{"Labels":{"com.gitea.runner.job":"job-1"}}`,
},
{
"/volumes/create", `{"Labels":{"discard":"1"},"labels":null,"LABELS":{"own":"1"},"LABELS":{"extra":"2"}}`,
`{"Labels":{"own":"1","extra":"2","com.gitea.runner.job":"job-1"}}`,
},
} {
resp, err := client.Post("http://docker"+path, "application/json", strings.NewReader(body))
resp, err := client.Post("http://docker"+testCase.path, "application/json", strings.NewReader(testCase.body))
require.NoError(t, err)
resp.Body.Close()
assert.Equal(t, http.StatusCreated, resp.StatusCode)
var got map[string]json.RawMessage
require.NoError(t, json.Unmarshal(<-bodies, &got))
key := "Labels"
if _, ok := got["labels"]; ok {
key = "labels"
}
var labels map[string]string
require.NoError(t, json.Unmarshal(got[key], &labels))
assert.Equal(t, "job-1", labels[jobLabel], path)
if strings.Contains(body, "own") {
assert.Equal(t, "1", labels["own"])
assert.JSONEq(t, `"alpine"`, string(got["Image"]))
}
require.Equal(t, http.StatusCreated, resp.StatusCode)
assert.JSONEq(t, testCase.want, string(<-bodies), testCase.body)
}
})
@@ -123,40 +185,99 @@ func TestDockerProxy(t *testing.T) {
assert.Equal(t, "OK GET /v1.47/_ping", string(body))
})
t.Run("tunnels raw streams", func(t *testing.T) {
conn, err := net.Dial("unix", proxy.Socket)
require.NoError(t, err)
defer conn.Close()
_, err = conn.Write([]byte("POST /v1.47/exec/abc/start HTTP/1.1\r\nHost: docker\r\nContent-Length: 2\r\n\r\n{}ping\n"))
require.NoError(t, err)
t.Run("tunnels raw streams through stdin EOF", func(t *testing.T) {
for _, stream := range []struct{ path, upgrade string }{
{path: "/v1.47/exec/abc/start"},
{path: "/containers/abc/attach", upgrade: "Connection: Upgrade\r\nUpgrade: tcp\r\n"},
} {
conn, reader := startStream(t, stream.path, stream.upgrade)
require.NoError(t, conn.CloseWrite())
output, err := io.ReadAll(reader)
require.NoError(t, err)
assert.Equal(t, "final:input", string(output))
<-streamsDone
}
})
t.Run("detached and error responses retain keepalive label injection", func(t *testing.T) {
conn := dialProxy(t)
reader := bufio.NewReader(conn)
resp, err := http.ReadResponse(reader, nil)
for endpoint, status := range map[string]int{"detached": http.StatusOK, "error": http.StatusBadRequest} {
_, err := fmt.Fprintf(conn, "POST /exec/%s/start HTTP/1.1\r\nHost: docker\r\nContent-Length: 2\r\n\r\n{}", endpoint)
require.NoError(t, err)
resp, err := http.ReadResponse(reader, nil)
require.NoError(t, err)
body, err := io.ReadAll(resp.Body)
require.NoError(t, err)
require.NoError(t, resp.Body.Close())
assert.Equal(t, status, resp.StatusCode)
assert.Equal(t, "ordinary", string(body))
_, err = io.WriteString(conn, "POST /volumes/create HTTP/1.1\r\nHost: docker\r\nContent-Length: 0\r\n\r\n")
require.NoError(t, err)
resp, err = http.ReadResponse(reader, nil)
require.NoError(t, err)
require.NoError(t, resp.Body.Close())
require.Equal(t, http.StatusCreated, resp.StatusCode)
assert.JSONEq(t, `{"Labels":{"com.gitea.runner.job":"job-1"}}`, string(<-bodies))
}
})
t.Run("close joins streams and admitted create", func(t *testing.T) {
t.Cleanup(func() { close(releaseCreate) })
_, raw := startStream(t, "/exec/live/start", "")
_, upgraded := startStream(t, "/session", "Connection: Upgrade\r\nUpgrade: tcp\r\n")
creator := dialProxy(t)
_, err := io.WriteString(creator, "POST /volumes/create?wait HTTP/1.1\r\nHost: docker\r\nContent-Length: 2\r\n\r\n{}")
require.NoError(t, err)
defer resp.Body.Close()
assert.Equal(t, "application/vnd.docker.raw-stream", resp.Header.Get("Content-Type"))
echoed, err := reader.ReadString('\n')
<-createStarted
require.NoError(t, creator.CloseWrite())
closed := make(chan error, 1)
ctx, cancel := context.WithTimeout(t.Context(), 5*time.Second)
defer cancel()
go func() { closed <- proxy.Close(ctx) }()
for _, reader := range []*bufio.Reader{raw, upgraded} {
_, err := reader.ReadByte()
require.ErrorIs(t, err, io.EOF)
<-streamsDone
}
select {
case err := <-closed:
t.Fatalf("Close returned before admitted create settled: %v", err)
default:
}
releaseCreate <- struct{}{}
resp, err := http.ReadResponse(bufio.NewReader(creator), nil)
require.NoError(t, err)
assert.Equal(t, "{}ping\n", echoed)
require.NoError(t, resp.Body.Close())
require.Equal(t, http.StatusCreated, resp.StatusCode)
assert.JSONEq(t, `{"Labels":{"com.gitea.runner.job":"job-1"}}`, string(<-bodies))
require.NoError(t, <-closed)
})
}
func TestRemoveLabelledRemovesContainersNetworksAndVolumes(t *testing.T) {
containerFailure := errors.New("container removal failed")
listFailure := errors.New("listing failed")
volumeFailure := errors.New("volume removal failed")
ctx := context.Background()
filters := make(mobyclient.Filters).Add("label", jobLabel+"=job-1")
cli := &mockDockerClient{}
cli.On("ContainerList", ctx, mobyclient.ContainerListOptions{All: true, Filters: filters}).
Return(mobyclient.ContainerListResult{Items: []container.Summary{{ID: "c1", Names: []string{"/app"}}}}, nil)
cli.On("ContainerKill", ctx, "c1", mock.Anything).Return(mobyclient.ContainerKillResult{}, nil)
Return(mobyclient.ContainerListResult{Items: []container.Summary{{ID: "c1", Names: []string{"/app"}}}}, nil).Once()
cli.On("ContainerKill", ctx, "c1", mock.Anything).Return(mobyclient.ContainerKillResult{}, nil).Once()
cli.On("ContainerRemove", ctx, "c1", mobyclient.ContainerRemoveOptions{RemoveVolumes: true, Force: true}).
Return(mobyclient.ContainerRemoveResult{}, nil)
cli.On("NetworkList", ctx, mobyclient.NetworkListOptions{Filters: filters}).
Return(mobyclient.NetworkListResult{Items: []network.Summary{{ID: "n1", Name: "app_default"}}}, nil)
cli.On("NetworkRemove", ctx, "n1", mock.Anything).Return(mobyclient.NetworkRemoveResult{}, nil)
Return(mobyclient.ContainerRemoveResult{}, containerFailure).Once()
cli.On("NetworkList", ctx, mobyclient.NetworkListOptions{Filters: filters}).Return(mobyclient.NetworkListResult{}, listFailure).Once()
cli.On("VolumeList", ctx, mobyclient.VolumeListOptions{Filters: filters}).
Return(mobyclient.VolumeListResult{Items: []volume.Volume{{Name: "app_data"}}}, nil)
cli.On("VolumeRemove", ctx, "app_data", mobyclient.VolumeRemoveOptions{}).Return(mobyclient.VolumeRemoveResult{}, nil)
Return(mobyclient.VolumeListResult{Items: []volume.Volume{{Name: "app_data"}}}, nil).Once()
cli.On("VolumeRemove", ctx, "app_data", mobyclient.VolumeRemoveOptions{}).Return(mobyclient.VolumeRemoveResult{}, volumeFailure).Once()
require.NoError(t, removeLabelled(ctx, cli, "job-1"))
err := removeLabelled(ctx, cli, "job-1")
require.ErrorIs(t, err, containerFailure)
require.ErrorIs(t, err, listFailure)
require.ErrorIs(t, err, volumeFailure)
require.ErrorContains(t, err, "failed to remove container c1")
require.ErrorContains(t, err, "failed to remove volume app_data")
cli.AssertExpectations(t)
}
@@ -168,13 +289,14 @@ func TestDockerProxyWithDaemon(t *testing.T) {
require.NoError(t, err)
defer direct.Close()
dir := shortTempDir(t)
seen, err := daemonSeesDir(ctx, dir)
seen, err := daemonSeesDir(ctx, direct, dir)
require.NoError(t, err)
t.Logf("daemon sees the runner's filesystem: %v", seen)
job := "proxy-test-" + t.Name()
job := "proxy-test-" + strconv.FormatInt(time.Now().UnixNano(), 36)
proxy, err := StartDockerProxy(daemonSocketPath(t, direct), dir, job)
require.NoError(t, err)
t.Cleanup(func() { _ = proxy.Close(context.Background()) })
viaProxy, err := mobyclient.New(mobyclient.WithHost("unix://" + proxy.Socket))
require.NoError(t, err)
defer viaProxy.Close()
@@ -212,6 +334,7 @@ func TestDockerProxyWithDaemon(t *testing.T) {
attached.Close()
require.NoError(t, proxy.Close(ctx))
require.NoError(t, RemoveDockerJobResources(ctx, job))
_, err = direct.ContainerInspect(ctx, created.ID, mobyclient.ContainerInspectOptions{})
assert.True(t, cerrdefs.IsNotFound(err))
_, err = direct.NetworkInspect(ctx, net.ID, mobyclient.NetworkInspectOptions{})
@@ -268,3 +391,79 @@ func BenchmarkDockerProxy(b *testing.B) {
})
}
}
type probeClient struct {
mobyclient.APIClient
create func(mobyclient.ContainerCreateOptions) (mobyclient.ContainerCreateResult, error)
remove func(context.Context, string, mobyclient.ContainerRemoveOptions) (mobyclient.ContainerRemoveResult, error)
}
func (c *probeClient) ImageList(context.Context, mobyclient.ImageListOptions) (mobyclient.ImageListResult, error) {
return mobyclient.ImageListResult{Items: []image.Summary{{ID: "probe-image"}}}, nil
}
func (c *probeClient) ContainerCreate(_ context.Context, opts mobyclient.ContainerCreateOptions) (mobyclient.ContainerCreateResult, error) {
return c.create(opts)
}
func (c *probeClient) ContainerRemove(ctx context.Context, id string, opts mobyclient.ContainerRemoveOptions) (mobyclient.ContainerRemoveResult, error) {
return c.remove(ctx, id, opts)
}
func TestDaemonSeesDir(t *testing.T) {
dir := t.TempDir()
markers := make(map[string]bool)
for _, testCase := range []struct {
name string
private bool
removeErr error
}{
{name: "cleanup after cancellation"},
{name: "private filesystem with stale marker", private: true},
{name: "cleanup error after cancellation", removeErr: errors.New("cleanup failed")},
} {
t.Run(testCase.name, func(t *testing.T) {
ctx, cancel := context.WithCancel(t.Context())
defer cancel()
daemonDir := dir
if testCase.private {
daemonDir = t.TempDir()
require.NoError(t, os.WriteFile(filepath.Join(daemonDir, "probe"), []byte("stale"), 0o600))
}
removed := false
cli := &probeClient{
create: func(opts mobyclient.ContainerCreateOptions) (mobyclient.ContainerCreateResult, error) {
require.Len(t, opts.HostConfig.Mounts, 1)
marker := opts.HostConfig.Mounts[0].Source
assert.Equal(t, mount.Mount{Type: mount.TypeBind, Source: marker, Target: "/gitea-runner-probe", ReadOnly: true}, opts.HostConfig.Mounts[0])
assert.False(t, markers[marker])
markers[marker] = true
info, err := os.Stat(marker)
require.NoError(t, err)
assert.True(t, info.Mode().IsRegular())
cancel()
if _, err := os.Stat(filepath.Join(daemonDir, filepath.Base(marker))); errors.Is(err, os.ErrNotExist) {
return mobyclient.ContainerCreateResult{}, cerrdefs.ErrInvalidArgument
}
return mobyclient.ContainerCreateResult{ID: "probe"}, nil
},
remove: func(ctx context.Context, id string, opts mobyclient.ContainerRemoveOptions) (mobyclient.ContainerRemoveResult, error) {
removed = true
assert.Equal(t, "probe", id)
assert.Equal(t, mobyclient.ContainerRemoveOptions{Force: true, RemoveVolumes: true}, opts)
require.NoError(t, ctx.Err())
_, bounded := ctx.Deadline()
assert.True(t, bounded)
return mobyclient.ContainerRemoveResult{}, testCase.removeErr
},
}
seen, err := daemonSeesDir(ctx, cli, dir)
require.ErrorIs(t, err, testCase.removeErr)
assert.Equal(t, !testCase.private && testCase.removeErr == nil, seen)
assert.Equal(t, !testCase.private, removed)
entries, err := os.ReadDir(dir)
require.NoError(t, err)
assert.Empty(t, entries)
})
}
}