distribution/docs/registry.go

250 lines
7.2 KiB
Go
Raw Normal View History

// Package registry contains client primitives to interact with a remote Docker registry.
2013-05-15 01:41:39 +00:00
package registry
import (
Add support for client certificates for registries This lets you specify custom client TLS certificates and CA root for a specific registry hostname. Docker will then verify the registry against the CA and present the client cert when talking to that registry. This allows the registry to verify that the client has a proper key, indicating that the client is allowed to access the images. A custom cert is configured by creating a directory in /etc/docker/certs.d with the same name as the registry hostname. Inside this directory all *.crt files are added as CA Roots (if none exists, the system default is used) and pair of files <filename>.key and <filename>.cert indicate a custom certificate to present to the registry. If there are multiple certificates each one will be tried in alphabetical order, proceeding to the next if we get a 403 of 5xx response. So, an example setup would be: /etc/docker/certs.d/ └── localhost ├── client.cert ├── client.key └── localhost.crt A simple way to test this setup is to use an apache server to host a registry. Just copy a registry tree into the apache root, here is an example one containing the busybox image: http://people.gnome.org/~alexl/v1.tar.gz Then add this conf file as /etc/httpd/conf.d/registry.conf: # This must be in the root context, otherwise it causes a re-negotiation # which is not supported by the tls implementation in go SSLVerifyClient optional_no_ca <Location /v1> Action cert-protected /cgi-bin/cert.cgi SetHandler cert-protected Header set x-docker-registry-version "0.6.2" SetEnvIf Host (.*) custom_host=$1 Header set X-Docker-Endpoints "%{custom_host}e" </Location> And this as /var/www/cgi-bin/cert.cgi #!/bin/bash if [ "$HTTPS" != "on" ]; then echo "Status: 403 Not using SSL" echo "x-docker-registry-version: 0.6.2" echo exit 0 fi if [ "$SSL_CLIENT_VERIFY" == "NONE" ]; then echo "Status: 403 Client certificate invalid" echo "x-docker-registry-version: 0.6.2" echo exit 0 fi echo "Content-length: $(stat --printf='%s' $PATH_TRANSLATED)" echo "x-docker-registry-version: 0.6.2" echo "X-Docker-Endpoints: $SERVER_NAME" echo "X-Docker-Size: 0" echo cat $PATH_TRANSLATED This will return 403 for all accessed to /v1 unless *any* client cert is presented. Obviously a real implementation would verify more details about the certificate. Example client certs can be generated with: openssl genrsa -out client.key 1024 openssl req -new -x509 -text -key client.key -out client.cert Docker-DCO-1.1-Signed-off-by: Alexander Larsson <alexl@redhat.com> (github: alexlarsson)
2013-12-04 14:03:51 +00:00
"crypto/tls"
"crypto/x509"
2013-05-15 20:22:57 +00:00
"errors"
"fmt"
"io/ioutil"
"net"
2013-05-15 01:41:39 +00:00
"net/http"
Add support for client certificates for registries This lets you specify custom client TLS certificates and CA root for a specific registry hostname. Docker will then verify the registry against the CA and present the client cert when talking to that registry. This allows the registry to verify that the client has a proper key, indicating that the client is allowed to access the images. A custom cert is configured by creating a directory in /etc/docker/certs.d with the same name as the registry hostname. Inside this directory all *.crt files are added as CA Roots (if none exists, the system default is used) and pair of files <filename>.key and <filename>.cert indicate a custom certificate to present to the registry. If there are multiple certificates each one will be tried in alphabetical order, proceeding to the next if we get a 403 of 5xx response. So, an example setup would be: /etc/docker/certs.d/ └── localhost ├── client.cert ├── client.key └── localhost.crt A simple way to test this setup is to use an apache server to host a registry. Just copy a registry tree into the apache root, here is an example one containing the busybox image: http://people.gnome.org/~alexl/v1.tar.gz Then add this conf file as /etc/httpd/conf.d/registry.conf: # This must be in the root context, otherwise it causes a re-negotiation # which is not supported by the tls implementation in go SSLVerifyClient optional_no_ca <Location /v1> Action cert-protected /cgi-bin/cert.cgi SetHandler cert-protected Header set x-docker-registry-version "0.6.2" SetEnvIf Host (.*) custom_host=$1 Header set X-Docker-Endpoints "%{custom_host}e" </Location> And this as /var/www/cgi-bin/cert.cgi #!/bin/bash if [ "$HTTPS" != "on" ]; then echo "Status: 403 Not using SSL" echo "x-docker-registry-version: 0.6.2" echo exit 0 fi if [ "$SSL_CLIENT_VERIFY" == "NONE" ]; then echo "Status: 403 Client certificate invalid" echo "x-docker-registry-version: 0.6.2" echo exit 0 fi echo "Content-length: $(stat --printf='%s' $PATH_TRANSLATED)" echo "x-docker-registry-version: 0.6.2" echo "X-Docker-Endpoints: $SERVER_NAME" echo "X-Docker-Size: 0" echo cat $PATH_TRANSLATED This will return 403 for all accessed to /v1 unless *any* client cert is presented. Obviously a real implementation would verify more details about the certificate. Example client certs can be generated with: openssl genrsa -out client.key 1024 openssl req -new -x509 -text -key client.key -out client.cert Docker-DCO-1.1-Signed-off-by: Alexander Larsson <alexl@redhat.com> (github: alexlarsson)
2013-12-04 14:03:51 +00:00
"os"
"path/filepath"
"runtime"
2013-05-15 01:41:39 +00:00
"strings"
"time"
"github.com/Sirupsen/logrus"
"github.com/docker/distribution/registry/api/errcode"
"github.com/docker/distribution/registry/api/v2"
"github.com/docker/distribution/registry/client"
"github.com/docker/distribution/registry/client/transport"
"github.com/docker/docker/autogen/dockerversion"
"github.com/docker/docker/pkg/parsers/kernel"
"github.com/docker/docker/pkg/tlsconfig"
"github.com/docker/docker/pkg/useragent"
2013-05-15 01:41:39 +00:00
)
var (
// ErrAlreadyExists is an error returned if an image being pushed
// already exists on the remote side
ErrAlreadyExists = errors.New("Image already exists")
errLoginRequired = errors.New("Authentication is required.")
)
2013-05-15 20:22:57 +00:00
// dockerUserAgent is the User-Agent the Docker client uses to identify itself.
// It is populated on init(), comprising version information of different components.
var dockerUserAgent string
func init() {
httpVersion := make([]useragent.VersionInfo, 0, 6)
httpVersion = append(httpVersion, useragent.VersionInfo{"docker", dockerversion.VERSION})
httpVersion = append(httpVersion, useragent.VersionInfo{"go", runtime.Version()})
httpVersion = append(httpVersion, useragent.VersionInfo{"git-commit", dockerversion.GITCOMMIT})
if kernelVersion, err := kernel.GetKernelVersion(); err == nil {
httpVersion = append(httpVersion, useragent.VersionInfo{"kernel", kernelVersion.String()})
}
httpVersion = append(httpVersion, useragent.VersionInfo{"os", runtime.GOOS})
httpVersion = append(httpVersion, useragent.VersionInfo{"arch", runtime.GOARCH})
dockerUserAgent = useragent.AppendVersions("", httpVersion...)
if runtime.GOOS != "linux" {
V2Only = true
}
Add support for client certificates for registries This lets you specify custom client TLS certificates and CA root for a specific registry hostname. Docker will then verify the registry against the CA and present the client cert when talking to that registry. This allows the registry to verify that the client has a proper key, indicating that the client is allowed to access the images. A custom cert is configured by creating a directory in /etc/docker/certs.d with the same name as the registry hostname. Inside this directory all *.crt files are added as CA Roots (if none exists, the system default is used) and pair of files <filename>.key and <filename>.cert indicate a custom certificate to present to the registry. If there are multiple certificates each one will be tried in alphabetical order, proceeding to the next if we get a 403 of 5xx response. So, an example setup would be: /etc/docker/certs.d/ └── localhost ├── client.cert ├── client.key └── localhost.crt A simple way to test this setup is to use an apache server to host a registry. Just copy a registry tree into the apache root, here is an example one containing the busybox image: http://people.gnome.org/~alexl/v1.tar.gz Then add this conf file as /etc/httpd/conf.d/registry.conf: # This must be in the root context, otherwise it causes a re-negotiation # which is not supported by the tls implementation in go SSLVerifyClient optional_no_ca <Location /v1> Action cert-protected /cgi-bin/cert.cgi SetHandler cert-protected Header set x-docker-registry-version "0.6.2" SetEnvIf Host (.*) custom_host=$1 Header set X-Docker-Endpoints "%{custom_host}e" </Location> And this as /var/www/cgi-bin/cert.cgi #!/bin/bash if [ "$HTTPS" != "on" ]; then echo "Status: 403 Not using SSL" echo "x-docker-registry-version: 0.6.2" echo exit 0 fi if [ "$SSL_CLIENT_VERIFY" == "NONE" ]; then echo "Status: 403 Client certificate invalid" echo "x-docker-registry-version: 0.6.2" echo exit 0 fi echo "Content-length: $(stat --printf='%s' $PATH_TRANSLATED)" echo "x-docker-registry-version: 0.6.2" echo "X-Docker-Endpoints: $SERVER_NAME" echo "X-Docker-Size: 0" echo cat $PATH_TRANSLATED This will return 403 for all accessed to /v1 unless *any* client cert is presented. Obviously a real implementation would verify more details about the certificate. Example client certs can be generated with: openssl genrsa -out client.key 1024 openssl req -new -x509 -text -key client.key -out client.cert Docker-DCO-1.1-Signed-off-by: Alexander Larsson <alexl@redhat.com> (github: alexlarsson)
2013-12-04 14:03:51 +00:00
}
func newTLSConfig(hostname string, isSecure bool) (*tls.Config, error) {
// PreferredServerCipherSuites should have no effect
tlsConfig := tlsconfig.ServerDefault
tlsConfig.InsecureSkipVerify = !isSecure
if isSecure {
hostDir := filepath.Join(CertsDir, cleanPath(hostname))
logrus.Debugf("hostDir: %s", hostDir)
if err := ReadCertsDirectory(&tlsConfig, hostDir); err != nil {
return nil, err
}
}
return &tlsConfig, nil
}
func hasFile(files []os.FileInfo, name string) bool {
for _, f := range files {
if f.Name() == name {
return true
Add support for client certificates for registries This lets you specify custom client TLS certificates and CA root for a specific registry hostname. Docker will then verify the registry against the CA and present the client cert when talking to that registry. This allows the registry to verify that the client has a proper key, indicating that the client is allowed to access the images. A custom cert is configured by creating a directory in /etc/docker/certs.d with the same name as the registry hostname. Inside this directory all *.crt files are added as CA Roots (if none exists, the system default is used) and pair of files <filename>.key and <filename>.cert indicate a custom certificate to present to the registry. If there are multiple certificates each one will be tried in alphabetical order, proceeding to the next if we get a 403 of 5xx response. So, an example setup would be: /etc/docker/certs.d/ └── localhost ├── client.cert ├── client.key └── localhost.crt A simple way to test this setup is to use an apache server to host a registry. Just copy a registry tree into the apache root, here is an example one containing the busybox image: http://people.gnome.org/~alexl/v1.tar.gz Then add this conf file as /etc/httpd/conf.d/registry.conf: # This must be in the root context, otherwise it causes a re-negotiation # which is not supported by the tls implementation in go SSLVerifyClient optional_no_ca <Location /v1> Action cert-protected /cgi-bin/cert.cgi SetHandler cert-protected Header set x-docker-registry-version "0.6.2" SetEnvIf Host (.*) custom_host=$1 Header set X-Docker-Endpoints "%{custom_host}e" </Location> And this as /var/www/cgi-bin/cert.cgi #!/bin/bash if [ "$HTTPS" != "on" ]; then echo "Status: 403 Not using SSL" echo "x-docker-registry-version: 0.6.2" echo exit 0 fi if [ "$SSL_CLIENT_VERIFY" == "NONE" ]; then echo "Status: 403 Client certificate invalid" echo "x-docker-registry-version: 0.6.2" echo exit 0 fi echo "Content-length: $(stat --printf='%s' $PATH_TRANSLATED)" echo "x-docker-registry-version: 0.6.2" echo "X-Docker-Endpoints: $SERVER_NAME" echo "X-Docker-Size: 0" echo cat $PATH_TRANSLATED This will return 403 for all accessed to /v1 unless *any* client cert is presented. Obviously a real implementation would verify more details about the certificate. Example client certs can be generated with: openssl genrsa -out client.key 1024 openssl req -new -x509 -text -key client.key -out client.cert Docker-DCO-1.1-Signed-off-by: Alexander Larsson <alexl@redhat.com> (github: alexlarsson)
2013-12-04 14:03:51 +00:00
}
}
return false
}
// ReadCertsDirectory reads the directory for TLS certificates
// including roots and certificate pairs and updates the
// provided TLS configuration.
func ReadCertsDirectory(tlsConfig *tls.Config, directory string) error {
fs, err := ioutil.ReadDir(directory)
if err != nil && !os.IsNotExist(err) {
return err
}
for _, f := range fs {
if strings.HasSuffix(f.Name(), ".crt") {
if tlsConfig.RootCAs == nil {
// TODO(dmcgowan): Copy system pool
tlsConfig.RootCAs = x509.NewCertPool()
}
logrus.Debugf("crt: %s", filepath.Join(directory, f.Name()))
data, err := ioutil.ReadFile(filepath.Join(directory, f.Name()))
if err != nil {
return err
}
tlsConfig.RootCAs.AppendCertsFromPEM(data)
}
if strings.HasSuffix(f.Name(), ".cert") {
certName := f.Name()
keyName := certName[:len(certName)-5] + ".key"
logrus.Debugf("cert: %s", filepath.Join(directory, f.Name()))
if !hasFile(fs, keyName) {
return fmt.Errorf("Missing key %s for certificate %s", keyName, certName)
}
cert, err := tls.LoadX509KeyPair(filepath.Join(directory, certName), filepath.Join(directory, keyName))
if err != nil {
return err
}
tlsConfig.Certificates = append(tlsConfig.Certificates, cert)
}
if strings.HasSuffix(f.Name(), ".key") {
keyName := f.Name()
certName := keyName[:len(keyName)-4] + ".cert"
logrus.Debugf("key: %s", filepath.Join(directory, f.Name()))
if !hasFile(fs, certName) {
return fmt.Errorf("Missing certificate %s for key %s", certName, keyName)
}
}
}
return nil
}
// DockerHeaders returns request modifiers that ensure requests have
// the User-Agent header set to dockerUserAgent and that metaHeaders
// are added.
func DockerHeaders(metaHeaders http.Header) []transport.RequestModifier {
modifiers := []transport.RequestModifier{
transport.NewHeaderRequestModifier(http.Header{"User-Agent": []string{dockerUserAgent}}),
}
if metaHeaders != nil {
modifiers = append(modifiers, transport.NewHeaderRequestModifier(metaHeaders))
}
return modifiers
}
// HTTPClient returns a HTTP client structure which uses the given transport
// and contains the necessary headers for redirected requests
func HTTPClient(transport http.RoundTripper) *http.Client {
return &http.Client{
Transport: transport,
CheckRedirect: addRequiredHeadersToRedirectedRequests,
}
Add support for client certificates for registries This lets you specify custom client TLS certificates and CA root for a specific registry hostname. Docker will then verify the registry against the CA and present the client cert when talking to that registry. This allows the registry to verify that the client has a proper key, indicating that the client is allowed to access the images. A custom cert is configured by creating a directory in /etc/docker/certs.d with the same name as the registry hostname. Inside this directory all *.crt files are added as CA Roots (if none exists, the system default is used) and pair of files <filename>.key and <filename>.cert indicate a custom certificate to present to the registry. If there are multiple certificates each one will be tried in alphabetical order, proceeding to the next if we get a 403 of 5xx response. So, an example setup would be: /etc/docker/certs.d/ └── localhost ├── client.cert ├── client.key └── localhost.crt A simple way to test this setup is to use an apache server to host a registry. Just copy a registry tree into the apache root, here is an example one containing the busybox image: http://people.gnome.org/~alexl/v1.tar.gz Then add this conf file as /etc/httpd/conf.d/registry.conf: # This must be in the root context, otherwise it causes a re-negotiation # which is not supported by the tls implementation in go SSLVerifyClient optional_no_ca <Location /v1> Action cert-protected /cgi-bin/cert.cgi SetHandler cert-protected Header set x-docker-registry-version "0.6.2" SetEnvIf Host (.*) custom_host=$1 Header set X-Docker-Endpoints "%{custom_host}e" </Location> And this as /var/www/cgi-bin/cert.cgi #!/bin/bash if [ "$HTTPS" != "on" ]; then echo "Status: 403 Not using SSL" echo "x-docker-registry-version: 0.6.2" echo exit 0 fi if [ "$SSL_CLIENT_VERIFY" == "NONE" ]; then echo "Status: 403 Client certificate invalid" echo "x-docker-registry-version: 0.6.2" echo exit 0 fi echo "Content-length: $(stat --printf='%s' $PATH_TRANSLATED)" echo "x-docker-registry-version: 0.6.2" echo "X-Docker-Endpoints: $SERVER_NAME" echo "X-Docker-Size: 0" echo cat $PATH_TRANSLATED This will return 403 for all accessed to /v1 unless *any* client cert is presented. Obviously a real implementation would verify more details about the certificate. Example client certs can be generated with: openssl genrsa -out client.key 1024 openssl req -new -x509 -text -key client.key -out client.cert Docker-DCO-1.1-Signed-off-by: Alexander Larsson <alexl@redhat.com> (github: alexlarsson)
2013-12-04 14:03:51 +00:00
}
func trustedLocation(req *http.Request) bool {
var (
trusteds = []string{"docker.com", "docker.io"}
hostname = strings.SplitN(req.Host, ":", 2)[0]
)
if req.URL.Scheme != "https" {
return false
}
for _, trusted := range trusteds {
if hostname == trusted || strings.HasSuffix(hostname, "."+trusted) {
return true
}
}
return false
}
// addRequiredHeadersToRedirectedRequests adds the necessary redirection headers
// for redirected requests
func addRequiredHeadersToRedirectedRequests(req *http.Request, via []*http.Request) error {
if via != nil && via[0] != nil {
if trustedLocation(req) && trustedLocation(via[0]) {
req.Header = via[0].Header
return nil
}
for k, v := range via[0].Header {
if k != "Authorization" {
for _, vv := range v {
req.Header.Add(k, vv)
}
}
}
}
return nil
}
func shouldV2Fallback(err errcode.Error) bool {
logrus.Debugf("v2 error: %T %v", err, err)
switch err.Code {
case errcode.ErrorCodeUnauthorized, v2.ErrorCodeManifestUnknown:
return true
}
return false
}
// ErrNoSupport is an error type used for errors indicating that an operation
// is not supported. It encapsulates a more specific error.
type ErrNoSupport struct{ Err error }
func (e ErrNoSupport) Error() string {
if e.Err == nil {
return "not supported"
}
return e.Err.Error()
}
// ContinueOnError returns true if we should fallback to the next endpoint
// as a result of this error.
func ContinueOnError(err error) bool {
switch v := err.(type) {
case errcode.Errors:
return ContinueOnError(v[0])
case ErrNoSupport:
return ContinueOnError(v.Err)
case errcode.Error:
return shouldV2Fallback(v)
case *client.UnexpectedHTTPResponseError:
return true
}
// let's be nice and fallback if the error is a completely
// unexpected one.
// If new errors have to be handled in some way, please
// add them to the switch above.
return true
}
// NewTransport returns a new HTTP transport. If tlsConfig is nil, it uses the
// default TLS configuration.
func NewTransport(tlsConfig *tls.Config) *http.Transport {
if tlsConfig == nil {
var cfg = tlsconfig.ServerDefault
tlsConfig = &cfg
}
return &http.Transport{
Proxy: http.ProxyFromEnvironment,
Dial: (&net.Dialer{
Timeout: 30 * time.Second,
KeepAlive: 30 * time.Second,
DualStack: true,
}).Dial,
TLSHandshakeTimeout: 10 * time.Second,
TLSClientConfig: tlsConfig,
// TODO(dmcgowan): Call close idle connections when complete and use keep alive
DisableKeepAlives: true,
}
}