feat/HTTPRouteFilter: host rewrite from path

api/v1alpha1/httproutefilter_types.go

@@ -127,6 +127,10 @@ const (
 	// BackendHTTPHostnameModifier indicates that the Host header value would be replaced by the DNS name of the backend if it exists.
 	// https://www.envoyproxy.io/docs/envoy/latest/api-v3/config/route/v3/route_components.proto#envoy-v3-api-field-config-route-v3-routeaction-auto-host-rewrite
 	BackendHTTPHostnameModifier HTTPHostnameModifierType = "Backend"
+	// PathRegexHTTPHostnameModifier indicates that the Host header value would be rewritten by applying a regex
+	// match and substitution to the request path.
+	// https://www.envoyproxy.io/docs/envoy/latest/api-v3/config/route/v3/route_components.proto#envoy-v3-api-field-config-route-v3-routeaction-host-rewrite-path-regex
+	PathRegexHTTPHostnameModifier HTTPHostnameModifierType = "PathRegex"
 )
 
 type ReplaceRegexMatch struct {
@@ -139,6 +143,26 @@ type ReplaceRegexMatch struct {
 	Substitution string `json:"substitution"`
 }
 
+// HostnamePathRegexRewrite defines a hostname rewrite computed from the request path using regex.
+type HostnamePathRegexRewrite struct {
+	// Pattern matches a regular expression against the value of the HTTP Path. The regex string must
+	// adhere to the syntax documented in https://github.com/google/re2/wiki/Syntax.
+	// +kubebuilder:validation:MinLength=1
+	Pattern string `json:"pattern"`
+	// Substitution is an expression that replaces the matched portion. The expression may include numbered
+	// capture groups that adhere to syntax documented in https://github.com/google/re2/wiki/Syntax.
+	//
+	// The resulting value is used as the upstream Host header and should be constrained to a valid
+	// DNS hostname by using explicit regex capture groups in Pattern.
+	//
+	// The NUL, CR, and LF characters are not allowed: they are invalid in an HTTP header value and are
+	// rejected by the Envoy proto (well_known_regex HTTP_HEADER_VALUE), which would otherwise cause the
+	// generated configuration to be rejected by the data plane.
+	// +kubebuilder:validation:MinLength=1
+	// +kubebuilder:validation:Pattern=`^[^\r\n\x00]*$`
+	Substitution string `json:"substitution"`
+}
+
 // +kubebuilder:validation:XValidation:rule="self.type == 'ReplaceRegexMatch' ? has(self.replaceRegexMatch) : !has(self.replaceRegexMatch)",message="If HTTPPathModifier type is ReplaceRegexMatch, replaceRegexMatch field needs to be set."
 type HTTPPathModifier struct {
 	// +kubebuilder:validation:Enum=ReplaceRegexMatch
@@ -169,14 +193,28 @@ type HTTPPathModifier struct {
 
 // +kubebuilder:validation:XValidation:message="header must be nil if the type is not Header",rule="!(has(self.header) && self.type != 'Header')"
 // +kubebuilder:validation:XValidation:message="header must be specified for Header type",rule="!(!has(self.header) && self.type == 'Header')"
+// +kubebuilder:validation:XValidation:message="pathRegex must be nil if the type is not PathRegex",rule="!(has(self.pathRegex) && self.type != 'PathRegex')"
+// +kubebuilder:validation:XValidation:message="pathRegex must be specified for PathRegex type",rule="!(!has(self.pathRegex) && self.type == 'PathRegex')"
 type HTTPHostnameModifier struct {
-	// +kubebuilder:validation:Enum=Header;Backend
+	// +kubebuilder:validation:Enum=Header;Backend;PathRegex
 	// +kubebuilder:validation:Required
 	Type HTTPHostnameModifierType `json:"type"`
 
 	// Header is the name of the header whose value would be used to rewrite the Host header
 	// +optional
 	Header *string `json:"header,omitempty"`
+
+	// PathRegex defines a regex match and substitution applied to the request path to compute
+	// the rewritten Host header.
+	// For example, with:
+	// pathRegex:
+	//   pattern: "^/tenant/([a-z0-9-]+)/.*"
+	//   substitution: "\\1.example.internal"
+	// a request to "http://foo.bar.com/tenant/tenant1/api/v1" has its upstream Host header rewritten
+	// to "tenant1.example.internal" (the request path "/tenant/tenant1/api/v1" is preserved).
+	//
+	// +optional
+	PathRegex *HostnamePathRegexRewrite `json:"pathRegex,omitempty"`
 }
 
 // HTTPCredentialInjectionFilter defines the configuration to inject credentials into the request.

charts/gateway-crds-helm/templates/generated/gateway.envoyproxy.io_httproutefilters.yaml

@@ -424,12 +424,48 @@ spec:
                         description: Header is the name of the header whose value
                           would be used to rewrite the Host header
                         type: string
+                      pathRegex:
+                        description: |-
+                          PathRegex defines a regex match and substitution applied to the request path to compute
+                          the rewritten Host header.
+                          For example, with:
+                          pathRegex:
+                            pattern: "^/tenant/([a-z0-9-]+)/.*"
+                            substitution: "\\1.example.internal"
+                          a request to "http://foo.bar.com/tenant/tenant1/api/v1" has its upstream Host header rewritten
+                          to "tenant1.example.internal" (the request path "/tenant/tenant1/api/v1" is preserved).
+                        properties:
+                          pattern:
+                            description: |-
+                              Pattern matches a regular expression against the value of the HTTP Path. The regex string must
+                              adhere to the syntax documented in https://github.com/google/re2/wiki/Syntax.
+                            minLength: 1
+                            type: string
+                          substitution:
+                            description: |-
+                              Substitution is an expression that replaces the matched portion. The expression may include numbered
+                              capture groups that adhere to syntax documented in https://github.com/google/re2/wiki/Syntax.
+
+                              The resulting value is used as the upstream Host header and should be constrained to a valid
+                              DNS hostname by using explicit regex capture groups in Pattern.
+
+                              The NUL, CR, and LF characters are not allowed: they are invalid in an HTTP header value and are
+                              rejected by the Envoy proto (well_known_regex HTTP_HEADER_VALUE), which would otherwise cause the
+                              generated configuration to be rejected by the data plane.
+                            minLength: 1
+                            pattern: ^[^\r\n\x00]*$
+                            type: string
+                        required:
+                        - pattern
+                        - substitution
+                        type: object
                       type:
                         description: HTTPPathModifierType defines the type of Hostname
                           rewrite.
                         enum:
                         - Header
                         - Backend
+                        - PathRegex
                         type: string
                     required:
                     - type
@@ -439,6 +475,10 @@ spec:
                       rule: '!(has(self.header) && self.type != ''Header'')'
                     - message: header must be specified for Header type
                       rule: '!(!has(self.header) && self.type == ''Header'')'
+                    - message: pathRegex must be nil if the type is not PathRegex
+                      rule: '!(has(self.pathRegex) && self.type != ''PathRegex'')'
+                    - message: pathRegex must be specified for PathRegex type
+                      rule: '!(!has(self.pathRegex) && self.type == ''PathRegex'')'
                   path:
                     description: Path defines a path rewrite.
                     properties:

charts/gateway-helm/charts/crds/crds/generated/gateway.envoyproxy.io_httproutefilters.yaml

@@ -423,12 +423,48 @@ spec:
                         description: Header is the name of the header whose value
                           would be used to rewrite the Host header
                         type: string
+                      pathRegex:
+                        description: |-
+                          PathRegex defines a regex match and substitution applied to the request path to compute
+                          the rewritten Host header.
+                          For example, with:
+                          pathRegex:
+                            pattern: "^/tenant/([a-z0-9-]+)/.*"
+                            substitution: "\\1.example.internal"
+                          a request to "http://foo.bar.com/tenant/tenant1/api/v1" has its upstream Host header rewritten
+                          to "tenant1.example.internal" (the request path "/tenant/tenant1/api/v1" is preserved).
+                        properties:
+                          pattern:
+                            description: |-
+                              Pattern matches a regular expression against the value of the HTTP Path. The regex string must
+                              adhere to the syntax documented in https://github.com/google/re2/wiki/Syntax.
+                            minLength: 1
+                            type: string
+                          substitution:
+                            description: |-
+                              Substitution is an expression that replaces the matched portion. The expression may include numbered
+                              capture groups that adhere to syntax documented in https://github.com/google/re2/wiki/Syntax.
+
+                              The resulting value is used as the upstream Host header and should be constrained to a valid
+                              DNS hostname by using explicit regex capture groups in Pattern.
+
+                              The NUL, CR, and LF characters are not allowed: they are invalid in an HTTP header value and are
+                              rejected by the Envoy proto (well_known_regex HTTP_HEADER_VALUE), which would otherwise cause the
+                              generated configuration to be rejected by the data plane.
+                            minLength: 1
+                            pattern: ^[^\r\n\x00]*$
+                            type: string
+                        required:
+                        - pattern
+                        - substitution
+                        type: object
                       type:
                         description: HTTPPathModifierType defines the type of Hostname
                           rewrite.
                         enum:
                         - Header
                         - Backend
+                        - PathRegex
                         type: string
                     required:
                     - type
@@ -438,6 +474,10 @@ spec:
                       rule: '!(has(self.header) && self.type != ''Header'')'
                     - message: header must be specified for Header type
                       rule: '!(!has(self.header) && self.type == ''Header'')'
+                    - message: pathRegex must be nil if the type is not PathRegex
+                      rule: '!(has(self.pathRegex) && self.type != ''PathRegex'')'
+                    - message: pathRegex must be specified for PathRegex type
+                      rule: '!(!has(self.pathRegex) && self.type == ''PathRegex'')'
                   path:
                     description: Path defines a path rewrite.
                     properties:

internal/gatewayapi/filters.go

@@ -250,7 +250,8 @@ func hasMultipleCoreRewrites(rewrite *gwapiv1.HTTPURLRewriteFilter, contextRewri
 // Checks if the context and the rewrite both contain a envoy-gateway extended HTTP URL rewrite
 func hasMultipleExtensionRewrites(rewrite *egv1a1.HTTPURLRewriteFilter, contextRewrite *ir.URLRewrite) bool {
 	contextHasExtensionRewrites := (contextRewrite.Path != nil && contextRewrite.Path.RegexMatchReplace != nil) ||
-		(contextRewrite.Host != nil && (contextRewrite.Host.Header != nil || contextRewrite.Host.Backend != nil))
+		(contextRewrite.Host != nil && (contextRewrite.Host.Header != nil || contextRewrite.Host.Backend != nil ||
+			contextRewrite.Host.PathRegex != nil))
 
 	return contextHasExtensionRewrites && (rewrite.Hostname != nil || rewrite.Path != nil)
 }
@@ -259,7 +260,7 @@ func hasMultipleExtensionRewrites(rewrite *egv1a1.HTTPURLRewriteFilter, contextR
 func hasConflictingCoreAndExtensionRewrites(rewrite *gwapiv1.HTTPURLRewriteFilter, contextRewrite *ir.URLRewrite) bool {
 	contextHasExtensionPathRewrites := contextRewrite.Path != nil && contextRewrite.Path.RegexMatchReplace != nil
 	contextHasExtensionHostRewrites := contextRewrite.Host != nil && (contextRewrite.Host.Header != nil ||
-		contextRewrite.Host.Backend != nil)
+		contextRewrite.Host.Backend != nil || contextRewrite.Host.PathRegex != nil)
 	return (rewrite.Hostname != nil && contextHasExtensionHostRewrites) || (rewrite.Path != nil && contextHasExtensionPathRewrites)
 }
 
@@ -902,6 +903,28 @@ func (t *Translator) processExtensionRefHTTPFilter(extFilter *gwapiv1.LocalObjec
 							hm = &ir.HTTPHostModifier{
 								Backend: new(true),
 							}
+						case egv1a1.PathRegexHTTPHostnameModifier:
+							if hrf.Spec.URLRewrite.Hostname.PathRegex == nil ||
+								hrf.Spec.URLRewrite.Hostname.PathRegex.Pattern == "" ||
+								hrf.Spec.URLRewrite.Hostname.PathRegex.Substitution == "" {
+								return status.NewRouteStatusError(
+									errors.New("PathRegex Pattern and Substitution must be set when rewrite hostname type is \"PathRegex\""),
+									gwapiv1.RouteReasonUnsupportedValue,
+								).WithType(gwapiv1.RouteConditionAccepted)
+							} else if _, err := regexp.Compile(hrf.Spec.URLRewrite.Hostname.PathRegex.Pattern); err != nil {
+								// Avoid envoy NACKs due to invalid regex.
+								// Go's regexp syntax is RE2: https://pkg.go.dev/regexp/syntax
+								return status.NewRouteStatusError(
+									errors.New("PathRegex must be a valid RE2 regular expression"),
+									gwapiv1.RouteReasonUnsupportedValue,
+								).WithType(gwapiv1.RouteConditionAccepted)
+							}
+							hm = &ir.HTTPHostModifier{
+								PathRegex: &ir.RegexMatchReplace{
+									Pattern:      hrf.Spec.URLRewrite.Hostname.PathRegex.Pattern,
+									Substitution: hrf.Spec.URLRewrite.Hostname.PathRegex.Substitution,
+								},
+							}
 						}
 
 						if filterContext.URLRewrite != nil {