Adds a flag to the create and run command, `--use-api-socket`, that can be used to start a container with the correctly configured parameters to ensure that accessing the docker socket will work with out managing bind mounts and authentication injection. The implementation in this PR resolves the tokens for the current credential set in the client and then copies it into a container at the well know location of /run/secrets/docker/config.json, setting DOCKER_CONFIG to ensure it is resolved by existing tooling. We use a compose-compatible secret location with the hope that the CLI and compose can work together seamlessly. The bind mount for the socket is resolved from the current context, erroring out if the flag is set and the provided socket is not a unix socket. There are a few drawbacks to this approach but it resolves a long standing pain point. We'll continue to develop this as we understand more use cases but it is marked as experimental for now. Signed-off-by: Stephen Day <stephen.day@docker.com>
452 lines
14 KiB
Go
452 lines
14 KiB
Go
package container
|
|
|
|
import (
|
|
"archive/tar"
|
|
"bytes"
|
|
"context"
|
|
"fmt"
|
|
"io"
|
|
"net/netip"
|
|
"os"
|
|
"path"
|
|
"strings"
|
|
|
|
"github.com/containerd/platforms"
|
|
"github.com/distribution/reference"
|
|
"github.com/docker/cli/cli"
|
|
"github.com/docker/cli/cli/command"
|
|
"github.com/docker/cli/cli/command/completion"
|
|
"github.com/docker/cli/cli/command/image"
|
|
"github.com/docker/cli/cli/config/configfile"
|
|
"github.com/docker/cli/cli/config/types"
|
|
"github.com/docker/cli/cli/internal/jsonstream"
|
|
"github.com/docker/cli/cli/streams"
|
|
"github.com/docker/cli/cli/trust"
|
|
"github.com/docker/cli/opts"
|
|
"github.com/docker/docker/api/types/container"
|
|
imagetypes "github.com/docker/docker/api/types/image"
|
|
"github.com/docker/docker/api/types/mount"
|
|
"github.com/docker/docker/api/types/versions"
|
|
"github.com/docker/docker/client"
|
|
"github.com/docker/docker/errdefs"
|
|
specs "github.com/opencontainers/image-spec/specs-go/v1"
|
|
"github.com/pkg/errors"
|
|
"github.com/spf13/cobra"
|
|
"github.com/spf13/pflag"
|
|
)
|
|
|
|
// Pull constants
|
|
const (
|
|
PullImageAlways = "always"
|
|
PullImageMissing = "missing" // Default (matches previous behavior)
|
|
PullImageNever = "never"
|
|
)
|
|
|
|
type createOptions struct {
|
|
name string
|
|
platform string
|
|
untrusted bool
|
|
pull string // always, missing, never
|
|
quiet bool
|
|
useAPISocket bool
|
|
}
|
|
|
|
// NewCreateCommand creates a new cobra.Command for `docker create`
|
|
func NewCreateCommand(dockerCli command.Cli) *cobra.Command {
|
|
var options createOptions
|
|
var copts *containerOptions
|
|
|
|
cmd := &cobra.Command{
|
|
Use: "create [OPTIONS] IMAGE [COMMAND] [ARG...]",
|
|
Short: "Create a new container",
|
|
Args: cli.RequiresMinArgs(1),
|
|
RunE: func(cmd *cobra.Command, args []string) error {
|
|
copts.Image = args[0]
|
|
if len(args) > 1 {
|
|
copts.Args = args[1:]
|
|
}
|
|
return runCreate(cmd.Context(), dockerCli, cmd.Flags(), &options, copts)
|
|
},
|
|
Annotations: map[string]string{
|
|
"aliases": "docker container create, docker create",
|
|
},
|
|
ValidArgsFunction: completion.ImageNames(dockerCli, -1),
|
|
}
|
|
|
|
flags := cmd.Flags()
|
|
flags.SetInterspersed(false)
|
|
|
|
flags.StringVar(&options.name, "name", "", "Assign a name to the container")
|
|
flags.StringVar(&options.pull, "pull", PullImageMissing, `Pull image before creating ("`+PullImageAlways+`", "|`+PullImageMissing+`", "`+PullImageNever+`")`)
|
|
flags.BoolVarP(&options.quiet, "quiet", "q", false, "Suppress the pull output")
|
|
flags.BoolVarP(&options.useAPISocket, "use-api-socket", "", false, "Bind mount Docker API socket and required auth")
|
|
flags.SetAnnotation("use-api-socket", "experimentalCLI", nil) // Marks flag as experimental for now.
|
|
|
|
// Add an explicit help that doesn't have a `-h` to prevent the conflict
|
|
// with hostname
|
|
flags.Bool("help", false, "Print usage")
|
|
|
|
command.AddPlatformFlag(flags, &options.platform)
|
|
command.AddTrustVerificationFlags(flags, &options.untrusted, dockerCli.ContentTrustEnabled())
|
|
copts = addFlags(flags)
|
|
|
|
addCompletions(cmd, dockerCli)
|
|
|
|
flags.VisitAll(func(flag *pflag.Flag) {
|
|
// Set a default completion function if none was set. We don't look
|
|
// up if it does already have one set, because Cobra does this for
|
|
// us, and returns an error (which we ignore for this reason).
|
|
_ = cmd.RegisterFlagCompletionFunc(flag.Name, completion.NoComplete)
|
|
})
|
|
|
|
return cmd
|
|
}
|
|
|
|
func runCreate(ctx context.Context, dockerCli command.Cli, flags *pflag.FlagSet, options *createOptions, copts *containerOptions) error {
|
|
if err := validatePullOpt(options.pull); err != nil {
|
|
return cli.StatusError{
|
|
Status: withHelp(err, "create").Error(),
|
|
StatusCode: 125,
|
|
}
|
|
}
|
|
proxyConfig := dockerCli.ConfigFile().ParseProxyConfig(dockerCli.Client().DaemonHost(), opts.ConvertKVStringsToMapWithNil(copts.env.GetAll()))
|
|
newEnv := []string{}
|
|
for k, v := range proxyConfig {
|
|
if v == nil {
|
|
newEnv = append(newEnv, k)
|
|
} else {
|
|
newEnv = append(newEnv, k+"="+*v)
|
|
}
|
|
}
|
|
copts.env = *opts.NewListOptsRef(&newEnv, nil)
|
|
containerCfg, err := parse(flags, copts, dockerCli.ServerInfo().OSType)
|
|
if err != nil {
|
|
return cli.StatusError{
|
|
Status: withHelp(err, "create").Error(),
|
|
StatusCode: 125,
|
|
}
|
|
}
|
|
id, err := createContainer(ctx, dockerCli, containerCfg, options)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
_, _ = fmt.Fprintln(dockerCli.Out(), id)
|
|
return nil
|
|
}
|
|
|
|
// FIXME(thaJeztah): this is the only code-path that uses APIClient.ImageCreate. Rewrite this to use the regular "pull" code (or vice-versa).
|
|
func pullImage(ctx context.Context, dockerCli command.Cli, img string, options *createOptions) error {
|
|
encodedAuth, err := command.RetrieveAuthTokenFromImage(dockerCli.ConfigFile(), img)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
|
|
responseBody, err := dockerCli.Client().ImageCreate(ctx, img, imagetypes.CreateOptions{
|
|
RegistryAuth: encodedAuth,
|
|
Platform: options.platform,
|
|
})
|
|
if err != nil {
|
|
return err
|
|
}
|
|
defer responseBody.Close()
|
|
|
|
out := dockerCli.Err()
|
|
if options.quiet {
|
|
out = streams.NewOut(io.Discard)
|
|
}
|
|
return jsonstream.Display(ctx, responseBody, out)
|
|
}
|
|
|
|
type cidFile struct {
|
|
path string
|
|
file *os.File
|
|
written bool
|
|
}
|
|
|
|
func (cid *cidFile) Close() error {
|
|
if cid.file == nil {
|
|
return nil
|
|
}
|
|
cid.file.Close()
|
|
|
|
if cid.written {
|
|
return nil
|
|
}
|
|
if err := os.Remove(cid.path); err != nil {
|
|
return errors.Wrapf(err, "failed to remove the CID file '%s'", cid.path)
|
|
}
|
|
|
|
return nil
|
|
}
|
|
|
|
func (cid *cidFile) Write(id string) error {
|
|
if cid.file == nil {
|
|
return nil
|
|
}
|
|
if _, err := cid.file.Write([]byte(id)); err != nil {
|
|
return errors.Wrap(err, "failed to write the container ID to the file")
|
|
}
|
|
cid.written = true
|
|
return nil
|
|
}
|
|
|
|
func newCIDFile(cidPath string) (*cidFile, error) {
|
|
if cidPath == "" {
|
|
return &cidFile{}, nil
|
|
}
|
|
if _, err := os.Stat(cidPath); err == nil {
|
|
return nil, errors.Errorf("container ID file found, make sure the other container isn't running or delete %s", cidPath)
|
|
}
|
|
|
|
f, err := os.Create(cidPath)
|
|
if err != nil {
|
|
return nil, errors.Wrap(err, "failed to create the container ID file")
|
|
}
|
|
|
|
return &cidFile{path: cidPath, file: f}, nil
|
|
}
|
|
|
|
//nolint:gocyclo
|
|
func createContainer(ctx context.Context, dockerCli command.Cli, containerCfg *containerConfig, options *createOptions) (containerID string, err error) {
|
|
config := containerCfg.Config
|
|
hostConfig := containerCfg.HostConfig
|
|
networkingConfig := containerCfg.NetworkingConfig
|
|
|
|
var (
|
|
trustedRef reference.Canonical
|
|
namedRef reference.Named
|
|
)
|
|
|
|
containerIDFile, err := newCIDFile(hostConfig.ContainerIDFile)
|
|
if err != nil {
|
|
return "", err
|
|
}
|
|
defer containerIDFile.Close()
|
|
|
|
ref, err := reference.ParseAnyReference(config.Image)
|
|
if err != nil {
|
|
return "", err
|
|
}
|
|
if named, ok := ref.(reference.Named); ok {
|
|
namedRef = reference.TagNameOnly(named)
|
|
|
|
if taggedRef, ok := namedRef.(reference.NamedTagged); ok && !options.untrusted {
|
|
var err error
|
|
trustedRef, err = image.TrustedReference(ctx, dockerCli, taggedRef)
|
|
if err != nil {
|
|
return "", err
|
|
}
|
|
config.Image = reference.FamiliarString(trustedRef)
|
|
}
|
|
}
|
|
|
|
pullAndTagImage := func() error {
|
|
if err := pullImage(ctx, dockerCli, config.Image, options); err != nil {
|
|
return err
|
|
}
|
|
if taggedRef, ok := namedRef.(reference.NamedTagged); ok && trustedRef != nil {
|
|
return trust.TagTrusted(ctx, dockerCli.Client(), dockerCli.Err(), trustedRef, taggedRef)
|
|
}
|
|
return nil
|
|
}
|
|
|
|
const dockerConfigPathInContainer = "/run/secrets/docker/config.json"
|
|
var apiSocketCreds map[string]types.AuthConfig
|
|
|
|
if options.useAPISocket {
|
|
// We'll create two new mounts to handle this flag:
|
|
// 1. Mount the actual docker socket.
|
|
// 2. A synthezised ~/.docker/config.json with resolved tokens.
|
|
|
|
socket := dockerCli.DockerEndpoint().Host
|
|
if !strings.HasPrefix(socket, "unix://") {
|
|
return "", fmt.Errorf("flag --use-api-socket can only be used with unix sockets: docker endpoint %s incompatible", socket)
|
|
}
|
|
socket = strings.TrimPrefix(socket, "unix://") // should we confirm absolute path?
|
|
|
|
containerCfg.HostConfig.Mounts = append(containerCfg.HostConfig.Mounts, mount.Mount{
|
|
Type: mount.TypeBind,
|
|
Source: socket,
|
|
Target: "/var/run/docker.sock",
|
|
BindOptions: &mount.BindOptions{},
|
|
})
|
|
|
|
/*
|
|
|
|
Ideally, we'd like to copy the config into a tmpfs but unfortunately,
|
|
the mounts won't be in place until we start the container. This can
|
|
leave around the config if the container doesn't get deleted.
|
|
|
|
We are using the most compose-secret-compatible approach,
|
|
which is implemented at
|
|
https://github.com/docker/compose/blob/main/pkg/compose/convergence.go#L737
|
|
|
|
// Prepare a tmpfs mount for our credentials so they go away after the
|
|
// container exits. We'll copy into this mount after the container is
|
|
// created.
|
|
containerCfg.HostConfig.Mounts = append(containerCfg.HostConfig.Mounts, mount.Mount{
|
|
Type: mount.TypeTmpfs,
|
|
Target: "/docker/",
|
|
TmpfsOptions: &mount.TmpfsOptions{
|
|
SizeBytes: 1 << 20, // only need a small partition
|
|
Mode: 0o600,
|
|
},
|
|
})
|
|
*/
|
|
|
|
var envvarPresent bool
|
|
for _, envvar := range containerCfg.Config.Env {
|
|
if strings.HasPrefix(envvar, "DOCKER_CONFIG=") {
|
|
envvarPresent = true
|
|
}
|
|
}
|
|
|
|
// If the DOCKER_CONFIG env var is already present, we assume the client knows
|
|
// what they're doing and don't inject the creds.
|
|
if !envvarPresent {
|
|
// Set our special little location for the config file.
|
|
containerCfg.Config.Env = append(containerCfg.Config.Env,
|
|
"DOCKER_CONFIG="+path.Dir(dockerConfigPathInContainer))
|
|
|
|
// Resolve this here for later, ensuring we error our before we create the container.
|
|
creds, err := dockerCli.ConfigFile().GetAllCredentials()
|
|
if err != nil {
|
|
return "", fmt.Errorf("resolving credentials failed: %w", err)
|
|
}
|
|
apiSocketCreds = creds // inject these after container creation.
|
|
}
|
|
}
|
|
|
|
var platform *specs.Platform
|
|
// Engine API version 1.41 first introduced the option to specify platform on
|
|
// create. It will produce an error if you try to set a platform on older API
|
|
// versions, so check the API version here to maintain backwards
|
|
// compatibility for CLI users.
|
|
if options.platform != "" && versions.GreaterThanOrEqualTo(dockerCli.Client().ClientVersion(), "1.41") {
|
|
p, err := platforms.Parse(options.platform)
|
|
if err != nil {
|
|
return "", errors.Wrap(errdefs.InvalidParameter(err), "error parsing specified platform")
|
|
}
|
|
platform = &p
|
|
}
|
|
|
|
if options.pull == PullImageAlways {
|
|
if err := pullAndTagImage(); err != nil {
|
|
return "", err
|
|
}
|
|
}
|
|
|
|
hostConfig.ConsoleSize[0], hostConfig.ConsoleSize[1] = dockerCli.Out().GetTtySize()
|
|
|
|
response, err := dockerCli.Client().ContainerCreate(ctx, config, hostConfig, networkingConfig, platform, options.name)
|
|
if err != nil {
|
|
// Pull image if it does not exist locally and we have the PullImageMissing option. Default behavior.
|
|
if errdefs.IsNotFound(err) && namedRef != nil && options.pull == PullImageMissing {
|
|
if !options.quiet {
|
|
// we don't want to write to stdout anything apart from container.ID
|
|
_, _ = fmt.Fprintf(dockerCli.Err(), "Unable to find image '%s' locally\n", reference.FamiliarString(namedRef))
|
|
}
|
|
|
|
if err := pullAndTagImage(); err != nil {
|
|
return "", err
|
|
}
|
|
|
|
var retryErr error
|
|
response, retryErr = dockerCli.Client().ContainerCreate(ctx, config, hostConfig, networkingConfig, platform, options.name)
|
|
if retryErr != nil {
|
|
return "", retryErr
|
|
}
|
|
} else {
|
|
return "", err
|
|
}
|
|
}
|
|
|
|
if warn := localhostDNSWarning(*hostConfig); warn != "" {
|
|
response.Warnings = append(response.Warnings, warn)
|
|
}
|
|
|
|
containerID = response.ID
|
|
for _, w := range response.Warnings {
|
|
_, _ = fmt.Fprintln(dockerCli.Err(), "WARNING:", w)
|
|
}
|
|
err = containerIDFile.Write(containerID)
|
|
|
|
if options.useAPISocket && apiSocketCreds != nil {
|
|
// Create a new config file with just the auth.
|
|
newConfig := &configfile.ConfigFile{
|
|
AuthConfigs: apiSocketCreds,
|
|
}
|
|
|
|
if err := copyDockerConfigIntoContainer(ctx, dockerCli.Client(), containerID, dockerConfigPathInContainer, newConfig); err != nil {
|
|
return "", fmt.Errorf("injecting docker config.json into container failed: %w", err)
|
|
}
|
|
}
|
|
|
|
return containerID, err
|
|
}
|
|
|
|
// check the DNS settings passed via --dns against localhost regexp to warn if
|
|
// they are trying to set a DNS to a localhost address.
|
|
//
|
|
// TODO(thaJeztah): move this to the daemon, which can make a better call if it will work or not (depending on networking mode).
|
|
func localhostDNSWarning(hostConfig container.HostConfig) string {
|
|
for _, dnsIP := range hostConfig.DNS {
|
|
if addr, err := netip.ParseAddr(dnsIP); err == nil && addr.IsLoopback() {
|
|
return fmt.Sprintf("Localhost DNS (%s) may fail in containers.", addr)
|
|
}
|
|
}
|
|
return ""
|
|
}
|
|
|
|
func validatePullOpt(val string) error {
|
|
switch val {
|
|
case PullImageAlways, PullImageMissing, PullImageNever, "":
|
|
// valid option, but nothing to do yet
|
|
return nil
|
|
default:
|
|
return fmt.Errorf(
|
|
"invalid pull option: '%s': must be one of %q, %q or %q",
|
|
val,
|
|
PullImageAlways,
|
|
PullImageMissing,
|
|
PullImageNever,
|
|
)
|
|
}
|
|
}
|
|
|
|
// copyDockerConfigIntoContainer takes the client configuration and copies it
|
|
// into the container.
|
|
//
|
|
// The path should be an absolute path in the container, commonly
|
|
// /root/.docker/config.json.
|
|
func copyDockerConfigIntoContainer(ctx context.Context, dockerAPI client.APIClient, containerID string, configPath string, config *configfile.ConfigFile) error {
|
|
var configBuf bytes.Buffer
|
|
if err := config.SaveToWriter(&configBuf); err != nil {
|
|
return fmt.Errorf("saving creds: %w", err)
|
|
}
|
|
|
|
// We don't need to get super fancy with the tar creation.
|
|
var tarBuf bytes.Buffer
|
|
tarWriter := tar.NewWriter(&tarBuf)
|
|
tarWriter.WriteHeader(&tar.Header{
|
|
Name: configPath,
|
|
Size: int64(configBuf.Len()),
|
|
Mode: 0o600,
|
|
})
|
|
|
|
if _, err := io.Copy(tarWriter, &configBuf); err != nil {
|
|
return fmt.Errorf("writing config to tar file for config copy: %w", err)
|
|
}
|
|
|
|
if err := tarWriter.Close(); err != nil {
|
|
return fmt.Errorf("closing tar for config copy failed: %w", err)
|
|
}
|
|
|
|
if err := dockerAPI.CopyToContainer(ctx, containerID, "/",
|
|
&tarBuf, container.CopyToContainerOptions{}); err != nil {
|
|
return fmt.Errorf("copying config.json into container failed: %w", err)
|
|
}
|
|
|
|
return nil
|
|
}
|