open-vault/vault/router.go
Jeff Mitchell 5f7321dcc7 Fix a case where mounts could be duplicated (#6771)
When unmounting, the router entry would be tainted, preventing routing.
However, we would then unmount the router before clearing storage, so if
an error occurred the router would have forgotten the path. For auth
mounts this isn't a problem since they had a secondary check, but
regular mounts didn't (not sure why, but this is true back to at least
0.2.0). This meant you could then create a duplicate mount using the
same path which would then not conflict in the router until postUnseal.

This adds the extra check to regular mounts, and also moves the location
of the router unmount.

This also ensures that on the next router.Mount, tainted is set to the
mount entry's tainted status.

Fixes #6769
2019-06-04 10:33:36 -07:00

844 lines
24 KiB
Go

package vault
import (
"context"
"fmt"
"strings"
"sync"
"sync/atomic"
"time"
metrics "github.com/armon/go-metrics"
radix "github.com/armon/go-radix"
"github.com/hashicorp/vault/helper/namespace"
"github.com/hashicorp/vault/sdk/helper/consts"
"github.com/hashicorp/vault/sdk/helper/salt"
"github.com/hashicorp/vault/sdk/helper/strutil"
"github.com/hashicorp/vault/sdk/logical"
)
var (
deniedPassthroughRequestHeaders = []string{
consts.AuthHeaderName,
}
)
// Router is used to do prefix based routing of a request to a logical backend
type Router struct {
l sync.RWMutex
root *radix.Tree
mountUUIDCache *radix.Tree
mountAccessorCache *radix.Tree
tokenStoreSaltFunc func(context.Context) (*salt.Salt, error)
// storagePrefix maps the prefix used for storage (ala the BarrierView)
// to the backend. This is used to map a key back into the backend that owns it.
// For example, logical/uuid1/foobar -> secrets/ (kv backend) + foobar
storagePrefix *radix.Tree
}
// NewRouter returns a new router
func NewRouter() *Router {
r := &Router{
root: radix.New(),
storagePrefix: radix.New(),
mountUUIDCache: radix.New(),
mountAccessorCache: radix.New(),
}
return r
}
// routeEntry is used to represent a mount point in the router
type routeEntry struct {
tainted bool
backend logical.Backend
mountEntry *MountEntry
storageView logical.Storage
storagePrefix string
rootPaths atomic.Value
loginPaths atomic.Value
l sync.RWMutex
}
type validateMountResponse struct {
MountType string `json:"mount_type" structs:"mount_type" mapstructure:"mount_type"`
MountAccessor string `json:"mount_accessor" structs:"mount_accessor" mapstructure:"mount_accessor"`
MountPath string `json:"mount_path" structs:"mount_path" mapstructure:"mount_path"`
MountLocal bool `json:"mount_local" structs:"mount_local" mapstructure:"mount_local"`
}
// validateMountByAccessor returns the mount type and ID for a given mount
// accessor
func (r *Router) validateMountByAccessor(accessor string) *validateMountResponse {
if accessor == "" {
return nil
}
mountEntry := r.MatchingMountByAccessor(accessor)
if mountEntry == nil {
return nil
}
mountPath := mountEntry.Path
if mountEntry.Table == credentialTableType {
mountPath = credentialRoutePrefix + mountPath
}
return &validateMountResponse{
MountAccessor: mountEntry.Accessor,
MountType: mountEntry.Type,
MountPath: mountPath,
MountLocal: mountEntry.Local,
}
}
// SaltID is used to apply a salt and hash to an ID to make sure its not reversible
func (re *routeEntry) SaltID(id string) string {
return salt.SaltID(re.mountEntry.UUID, id, salt.SHA1Hash)
}
// Mount is used to expose a logical backend at a given prefix, using a unique salt,
// and the barrier view for that path.
func (r *Router) Mount(backend logical.Backend, prefix string, mountEntry *MountEntry, storageView *BarrierView) error {
r.l.Lock()
defer r.l.Unlock()
// prepend namespace
prefix = mountEntry.Namespace().Path + prefix
// Check if this is a nested mount
if existing, _, ok := r.root.LongestPrefix(prefix); ok && existing != "" {
return fmt.Errorf("cannot mount under existing mount %q", existing)
}
// Build the paths
paths := new(logical.Paths)
if backend != nil {
specialPaths := backend.SpecialPaths()
if specialPaths != nil {
paths = specialPaths
}
}
// Create a mount entry
re := &routeEntry{
tainted: mountEntry.Tainted,
backend: backend,
mountEntry: mountEntry,
storagePrefix: storageView.Prefix(),
storageView: storageView,
}
re.rootPaths.Store(pathsToRadix(paths.Root))
re.loginPaths.Store(pathsToRadix(paths.Unauthenticated))
switch {
case prefix == "":
return fmt.Errorf("missing prefix to be used for router entry; mount_path: %q, mount_type: %q", re.mountEntry.Path, re.mountEntry.Type)
case re.storagePrefix == "":
return fmt.Errorf("missing storage view prefix; mount_path: %q, mount_type: %q", re.mountEntry.Path, re.mountEntry.Type)
case re.mountEntry.UUID == "":
return fmt.Errorf("missing mount identifier; mount_path: %q, mount_type: %q", re.mountEntry.Path, re.mountEntry.Type)
case re.mountEntry.Accessor == "":
return fmt.Errorf("missing mount accessor; mount_path: %q, mount_type: %q", re.mountEntry.Path, re.mountEntry.Type)
}
r.root.Insert(prefix, re)
r.storagePrefix.Insert(re.storagePrefix, re)
r.mountUUIDCache.Insert(re.mountEntry.UUID, re.mountEntry)
r.mountAccessorCache.Insert(re.mountEntry.Accessor, re.mountEntry)
return nil
}
// Unmount is used to remove a logical backend from a given prefix
func (r *Router) Unmount(ctx context.Context, prefix string) error {
ns, err := namespace.FromContext(ctx)
if err != nil {
return err
}
prefix = ns.Path + prefix
r.l.Lock()
defer r.l.Unlock()
// Fast-path out if the backend doesn't exist
raw, ok := r.root.Get(prefix)
if !ok {
return nil
}
// Call backend's Cleanup routine
re := raw.(*routeEntry)
if re.backend != nil {
re.backend.Cleanup(ctx)
}
// Purge from the radix trees
r.root.Delete(prefix)
r.storagePrefix.Delete(re.storagePrefix)
r.mountUUIDCache.Delete(re.mountEntry.UUID)
r.mountAccessorCache.Delete(re.mountEntry.Accessor)
return nil
}
// Remount is used to change the mount location of a logical backend
func (r *Router) Remount(ctx context.Context, src, dst string) error {
ns, err := namespace.FromContext(ctx)
if err != nil {
return err
}
src = ns.Path + src
dst = ns.Path + dst
r.l.Lock()
defer r.l.Unlock()
// Check for existing mount
raw, ok := r.root.Get(src)
if !ok {
return fmt.Errorf("no mount at %q", src)
}
// Update the mount point
r.root.Delete(src)
r.root.Insert(dst, raw)
return nil
}
// Taint is used to mark a path as tainted. This means only RollbackOperation
// RevokeOperation requests are allowed to proceed
func (r *Router) Taint(ctx context.Context, path string) error {
ns, err := namespace.FromContext(ctx)
if err != nil {
return err
}
path = ns.Path + path
r.l.Lock()
defer r.l.Unlock()
_, raw, ok := r.root.LongestPrefix(path)
if ok {
raw.(*routeEntry).tainted = true
}
return nil
}
// Untaint is used to unmark a path as tainted.
func (r *Router) Untaint(ctx context.Context, path string) error {
ns, err := namespace.FromContext(ctx)
if err != nil {
return err
}
path = ns.Path + path
r.l.Lock()
defer r.l.Unlock()
_, raw, ok := r.root.LongestPrefix(path)
if ok {
raw.(*routeEntry).tainted = false
}
return nil
}
func (r *Router) MatchingMountByUUID(mountID string) *MountEntry {
if mountID == "" {
return nil
}
r.l.RLock()
_, raw, ok := r.mountUUIDCache.LongestPrefix(mountID)
if !ok {
r.l.RUnlock()
return nil
}
r.l.RUnlock()
return raw.(*MountEntry)
}
// MatchingMountByAccessor returns the MountEntry by accessor lookup
func (r *Router) MatchingMountByAccessor(mountAccessor string) *MountEntry {
if mountAccessor == "" {
return nil
}
r.l.RLock()
_, raw, ok := r.mountAccessorCache.LongestPrefix(mountAccessor)
if !ok {
r.l.RUnlock()
return nil
}
r.l.RUnlock()
return raw.(*MountEntry)
}
// MatchingMount returns the mount prefix that would be used for a path
func (r *Router) MatchingMount(ctx context.Context, path string) string {
r.l.RLock()
mount := r.matchingMountInternal(ctx, path)
r.l.RUnlock()
return mount
}
func (r *Router) matchingMountInternal(ctx context.Context, path string) string {
ns, err := namespace.FromContext(ctx)
if err != nil {
return ""
}
path = ns.Path + path
mount, _, ok := r.root.LongestPrefix(path)
if !ok {
return ""
}
return mount
}
// matchingPrefixInternal returns a mount prefix that a path may be a part of
func (r *Router) matchingPrefixInternal(ctx context.Context, path string) string {
ns, err := namespace.FromContext(ctx)
if err != nil {
return ""
}
path = ns.Path + path
var existing string
fn := func(existingPath string, v interface{}) bool {
if strings.HasPrefix(existingPath, path) {
existing = existingPath
return true
}
return false
}
r.root.WalkPrefix(path, fn)
return existing
}
// MountConflict determines if there are potential path conflicts
func (r *Router) MountConflict(ctx context.Context, path string) string {
r.l.RLock()
defer r.l.RUnlock()
if exactMatch := r.matchingMountInternal(ctx, path); exactMatch != "" {
return exactMatch
}
if prefixMatch := r.matchingPrefixInternal(ctx, path); prefixMatch != "" {
return prefixMatch
}
return ""
}
// MatchingStorageByAPIPath/StoragePath returns the storage used for
// API/Storage paths respectively
func (r *Router) MatchingStorageByAPIPath(ctx context.Context, path string) logical.Storage {
return r.matchingStorage(ctx, path, true)
}
func (r *Router) MatchingStorageByStoragePath(ctx context.Context, path string) logical.Storage {
return r.matchingStorage(ctx, path, false)
}
func (r *Router) matchingStorage(ctx context.Context, path string, apiPath bool) logical.Storage {
ns, err := namespace.FromContext(ctx)
if err != nil {
return nil
}
path = ns.Path + path
var raw interface{}
var ok bool
r.l.RLock()
if apiPath {
_, raw, ok = r.root.LongestPrefix(path)
} else {
_, raw, ok = r.storagePrefix.LongestPrefix(path)
}
r.l.RUnlock()
if !ok {
return nil
}
return raw.(*routeEntry).storageView
}
// MatchingMountEntry returns the MountEntry used for a path
func (r *Router) MatchingMountEntry(ctx context.Context, path string) *MountEntry {
ns, err := namespace.FromContext(ctx)
if err != nil {
return nil
}
path = ns.Path + path
r.l.RLock()
_, raw, ok := r.root.LongestPrefix(path)
r.l.RUnlock()
if !ok {
return nil
}
return raw.(*routeEntry).mountEntry
}
// MatchingBackend returns the backend used for a path
func (r *Router) MatchingBackend(ctx context.Context, path string) logical.Backend {
ns, err := namespace.FromContext(ctx)
if err != nil {
return nil
}
path = ns.Path + path
r.l.RLock()
_, raw, ok := r.root.LongestPrefix(path)
r.l.RUnlock()
if !ok {
return nil
}
return raw.(*routeEntry).backend
}
// MatchingSystemView returns the SystemView used for a path
func (r *Router) MatchingSystemView(ctx context.Context, path string) logical.SystemView {
ns, err := namespace.FromContext(ctx)
if err != nil {
return nil
}
path = ns.Path + path
r.l.RLock()
_, raw, ok := r.root.LongestPrefix(path)
r.l.RUnlock()
if !ok {
return nil
}
return raw.(*routeEntry).backend.System()
}
// MatchingStoragePrefixByAPIPath the storage prefix for the given api path
func (r *Router) MatchingStoragePrefixByAPIPath(ctx context.Context, path string) (string, bool) {
ns, err := namespace.FromContext(ctx)
if err != nil {
return "", false
}
path = ns.Path + path
_, prefix, found := r.matchingMountEntryByPath(ctx, path, true)
return prefix, found
}
// MatchingAPIPrefixByStoragePath the api path information for the given storage path
func (r *Router) MatchingAPIPrefixByStoragePath(ctx context.Context, path string) (*namespace.Namespace, string, string, bool) {
me, prefix, found := r.matchingMountEntryByPath(ctx, path, false)
if !found {
return nil, "", "", found
}
mountPath := me.Path
// Add back the prefix for credential backends
if strings.HasPrefix(path, credentialBarrierPrefix) {
mountPath = credentialRoutePrefix + mountPath
}
return me.Namespace(), mountPath, prefix, found
}
func (r *Router) matchingMountEntryByPath(ctx context.Context, path string, apiPath bool) (*MountEntry, string, bool) {
var raw interface{}
var ok bool
r.l.RLock()
if apiPath {
_, raw, ok = r.root.LongestPrefix(path)
} else {
_, raw, ok = r.storagePrefix.LongestPrefix(path)
}
r.l.RUnlock()
if !ok {
return nil, "", false
}
// Extract the mount path and storage prefix
re := raw.(*routeEntry)
prefix := re.storagePrefix
return re.mountEntry, prefix, true
}
// Route is used to route a given request
func (r *Router) Route(ctx context.Context, req *logical.Request) (*logical.Response, error) {
resp, _, _, err := r.routeCommon(ctx, req, false)
return resp, err
}
// RouteExistenceCheck is used to route a given existence check request
func (r *Router) RouteExistenceCheck(ctx context.Context, req *logical.Request) (*logical.Response, bool, bool, error) {
resp, ok, exists, err := r.routeCommon(ctx, req, true)
return resp, ok, exists, err
}
func (r *Router) routeCommon(ctx context.Context, req *logical.Request, existenceCheck bool) (*logical.Response, bool, bool, error) {
ns, err := namespace.FromContext(ctx)
if err != nil {
return nil, false, false, err
}
// Find the mount point
r.l.RLock()
adjustedPath := req.Path
mount, raw, ok := r.root.LongestPrefix(ns.Path + adjustedPath)
if !ok && !strings.HasSuffix(adjustedPath, "/") {
// Re-check for a backend by appending a slash. This lets "foo" mean
// "foo/" at the root level which is almost always what we want.
adjustedPath += "/"
mount, raw, ok = r.root.LongestPrefix(ns.Path + adjustedPath)
}
r.l.RUnlock()
if !ok {
return logical.ErrorResponse(fmt.Sprintf("no handler for route '%s'", req.Path)), false, false, logical.ErrUnsupportedPath
}
req.Path = adjustedPath
defer metrics.MeasureSince([]string{"route", string(req.Operation),
strings.Replace(mount, "/", "-", -1)}, time.Now())
re := raw.(*routeEntry)
// Grab a read lock on the route entry, this protects against the backend
// being reloaded during a request. The exception is a renew request on the
// token store; such a request will have already been routed through the
// token store -> exp manager -> here so we need to not grab the lock again
// or we'll be recursively grabbing it.
if !(req.Operation == logical.RenewOperation && strings.HasPrefix(req.Path, "auth/token/")) {
re.l.RLock()
defer re.l.RUnlock()
}
// Filtered mounts will have a nil backend
if re.backend == nil {
return logical.ErrorResponse(fmt.Sprintf("no handler for route '%s'", req.Path)), false, false, logical.ErrUnsupportedPath
}
// If the path is tainted, we reject any operation except for
// Rollback and Revoke
if re.tainted {
switch req.Operation {
case logical.RevokeOperation, logical.RollbackOperation:
default:
return logical.ErrorResponse(fmt.Sprintf("no handler for route '%s'", req.Path)), false, false, logical.ErrUnsupportedPath
}
}
// Adjust the path to exclude the routing prefix
originalPath := req.Path
req.Path = strings.TrimPrefix(ns.Path+req.Path, mount)
req.MountPoint = mount
req.MountType = re.mountEntry.Type
if req.Path == "/" {
req.Path = ""
}
// Attach the storage view for the request
req.Storage = re.storageView
originalEntityID := req.EntityID
// Hash the request token unless the request is being routed to the token
// or system backend.
clientToken := req.ClientToken
switch {
case strings.HasPrefix(originalPath, "auth/token/"):
case strings.HasPrefix(originalPath, "sys/"):
case strings.HasPrefix(originalPath, cubbyholeMountPath):
if req.Operation == logical.RollbackOperation {
// Backend doesn't support this and it can't properly look up a
// cubbyhole ID so just return here
return nil, false, false, nil
}
te := req.TokenEntry()
if te == nil {
return nil, false, false, fmt.Errorf("nil token entry")
}
if te.Type != logical.TokenTypeService {
return logical.ErrorResponse(`cubbyhole operations are only supported by "service" type tokens`), false, false, nil
}
switch {
case te.NamespaceID == namespace.RootNamespaceID && !strings.HasPrefix(req.ClientToken, "s."):
// In order for the token store to revoke later, we need to have the same
// salted ID, so we double-salt what's going to the cubbyhole backend
salt, err := r.tokenStoreSaltFunc(ctx)
if err != nil {
return nil, false, false, err
}
req.ClientToken = re.SaltID(salt.SaltID(req.ClientToken))
default:
if te.CubbyholeID == "" {
return nil, false, false, fmt.Errorf("empty cubbyhole id")
}
req.ClientToken = te.CubbyholeID
}
default:
req.ClientToken = re.SaltID(req.ClientToken)
}
// Cache the pointer to the original connection object
originalConn := req.Connection
// Cache the identifier of the request
originalReqID := req.ID
// Cache the client token's number of uses in the request
originalClientTokenRemainingUses := req.ClientTokenRemainingUses
req.ClientTokenRemainingUses = 0
originalMFACreds := req.MFACreds
req.MFACreds = nil
originalControlGroup := req.ControlGroup
req.ControlGroup = nil
// Cache the headers
headers := req.Headers
req.Headers = nil
// Filter and add passthrough headers to the backend
var passthroughRequestHeaders []string
if rawVal, ok := re.mountEntry.synthesizedConfigCache.Load("passthrough_request_headers"); ok {
passthroughRequestHeaders = rawVal.([]string)
}
var allowedResponseHeaders []string
if rawVal, ok := re.mountEntry.synthesizedConfigCache.Load("allowed_response_headers"); ok {
allowedResponseHeaders = rawVal.([]string)
}
if len(passthroughRequestHeaders) > 0 {
req.Headers = filteredHeaders(headers, passthroughRequestHeaders, deniedPassthroughRequestHeaders)
}
// Cache the wrap info of the request
var wrapInfo *logical.RequestWrapInfo
if req.WrapInfo != nil {
wrapInfo = &logical.RequestWrapInfo{
TTL: req.WrapInfo.TTL,
Format: req.WrapInfo.Format,
SealWrap: req.WrapInfo.SealWrap,
}
}
originalPolicyOverride := req.PolicyOverride
reqTokenEntry := req.TokenEntry()
req.SetTokenEntry(nil)
// Reset the request before returning
defer func() {
req.Path = originalPath
req.MountPoint = mount
req.MountType = re.mountEntry.Type
req.Connection = originalConn
req.ID = originalReqID
req.Storage = nil
req.ClientToken = clientToken
req.ClientTokenRemainingUses = originalClientTokenRemainingUses
req.WrapInfo = wrapInfo
req.Headers = headers
req.PolicyOverride = originalPolicyOverride
// This is only set in one place, after routing, so should never be set
// by a backend
req.SetLastRemoteWAL(0)
// This will be used for attaching the mount accessor for the identities
// returned by the authentication backends
req.MountAccessor = re.mountEntry.Accessor
req.EntityID = originalEntityID
req.MFACreds = originalMFACreds
req.SetTokenEntry(reqTokenEntry)
req.ControlGroup = originalControlGroup
}()
// Invoke the backend
if existenceCheck {
ok, exists, err := re.backend.HandleExistenceCheck(ctx, req)
return nil, ok, exists, err
} else {
resp, err := re.backend.HandleRequest(ctx, req)
if resp != nil {
if len(allowedResponseHeaders) > 0 {
resp.Headers = filteredHeaders(resp.Headers, allowedResponseHeaders, nil)
} else {
resp.Headers = nil
}
if resp.Auth != nil {
// When a token gets renewed, the request hits this path and
// reaches token store. Token store delegates the renewal to the
// expiration manager. Expiration manager in-turn creates a
// different logical request and forwards the request to the auth
// backend that had initially authenticated the login request. The
// forwarding to auth backend will make this code path hit for the
// second time for the same renewal request. The accessors in the
// Alias structs should be of the auth backend and not of the token
// store. Therefore, avoiding the overwriting of accessors by
// having a check for path prefix having "renew". This gets applied
// for "renew" and "renew-self" requests.
if !strings.HasPrefix(req.Path, "renew") {
if resp.Auth.Alias != nil {
resp.Auth.Alias.MountAccessor = re.mountEntry.Accessor
}
for _, alias := range resp.Auth.GroupAliases {
alias.MountAccessor = re.mountEntry.Accessor
}
}
switch re.mountEntry.Type {
case "token", "ns_token":
// Nothing; we respect what the token store is telling us and
// we don't allow tuning
default:
switch re.mountEntry.Config.TokenType {
case logical.TokenTypeService, logical.TokenTypeBatch:
resp.Auth.TokenType = re.mountEntry.Config.TokenType
case logical.TokenTypeDefault, logical.TokenTypeDefaultService:
if resp.Auth.TokenType == logical.TokenTypeDefault {
resp.Auth.TokenType = logical.TokenTypeService
}
case logical.TokenTypeDefaultBatch:
if resp.Auth.TokenType == logical.TokenTypeDefault {
resp.Auth.TokenType = logical.TokenTypeBatch
}
}
}
}
}
return resp, false, false, err
}
}
// RootPath checks if the given path requires root privileges
func (r *Router) RootPath(ctx context.Context, path string) bool {
ns, err := namespace.FromContext(ctx)
if err != nil {
return false
}
adjustedPath := ns.Path + path
r.l.RLock()
mount, raw, ok := r.root.LongestPrefix(adjustedPath)
r.l.RUnlock()
if !ok {
return false
}
re := raw.(*routeEntry)
// Trim to get remaining path
remain := strings.TrimPrefix(adjustedPath, mount)
// Check the rootPaths of this backend
rootPaths := re.rootPaths.Load().(*radix.Tree)
match, raw, ok := rootPaths.LongestPrefix(remain)
if !ok {
return false
}
prefixMatch := raw.(bool)
// Handle the prefix match case
if prefixMatch {
return strings.HasPrefix(remain, match)
}
// Handle the exact match case
return match == remain
}
// LoginPath checks if the given path is used for logins
func (r *Router) LoginPath(ctx context.Context, path string) bool {
ns, err := namespace.FromContext(ctx)
if err != nil {
return false
}
adjustedPath := ns.Path + path
r.l.RLock()
mount, raw, ok := r.root.LongestPrefix(adjustedPath)
r.l.RUnlock()
if !ok {
return false
}
re := raw.(*routeEntry)
// Trim to get remaining path
remain := strings.TrimPrefix(adjustedPath, mount)
// Check the loginPaths of this backend
loginPaths := re.loginPaths.Load().(*radix.Tree)
match, raw, ok := loginPaths.LongestPrefix(remain)
if !ok {
return false
}
prefixMatch := raw.(bool)
// Handle the prefix match case
if prefixMatch {
return strings.HasPrefix(remain, match)
}
// Handle the exact match case
return match == remain
}
// pathsToRadix converts a list of special paths to a radix tree.
func pathsToRadix(paths []string) *radix.Tree {
tree := radix.New()
for _, path := range paths {
// Check if this is a prefix or exact match
prefixMatch := len(path) >= 1 && path[len(path)-1] == '*'
if prefixMatch {
path = path[:len(path)-1]
}
tree.Insert(path, prefixMatch)
}
return tree
}
// filteredHeaders returns a headers map[string][]string that
// contains the filtered values contained in candidateHeaders. Filtering of
// candidateHeaders from the origHeaders is done is a case-insensitive manner.
// Headers that match values from deniedHeaders will be ignored.
func filteredHeaders(origHeaders map[string][]string, candidateHeaders, deniedHeaders []string) map[string][]string {
// Short-circuit if there's nothing to filter
if len(candidateHeaders) == 0 {
return nil
}
retHeaders := make(map[string][]string, len(origHeaders))
// Filter candidateHeaders values through deniedHeaders first. Returns the
// lowercased complement set. We call even if no denied headers to get the
// values lowercased.
allowedCandidateHeaders := strutil.Difference(candidateHeaders, deniedHeaders, true)
// Create a map that uses lowercased header values as the key and the original
// header naming as the value for comparison down below.
lowerOrigHeaderKeys := make(map[string]string, len(origHeaders))
for key := range origHeaders {
lowerOrigHeaderKeys[strings.ToLower(key)] = key
}
// Case-insensitive compare of passthrough headers against originating
// headers. The returned headers will be the same casing as the originating
// header name.
for _, ch := range allowedCandidateHeaders {
if header, ok := lowerOrigHeaderKeys[ch]; ok {
retHeaders[header] = origHeaders[header]
}
}
return retHeaders
}