Files
seaweedfs/weed/s3api/lance/server.go
T
Chris LuandClaude Opus 5.5 b9bc42c217 s3vectors: vend write credentials to index data only to the vector engine and Lance worker (#12316)
* security: share the filer admin token check

CheckGrpcAdminPeer verifies an admin Bearer token inline. Move the check
into IsFilerAdminToken so an HTTP endpoint can accept the same token.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* lance: recognize the cluster's own components by a signed token

A request whose X-Seaweedfs-Cluster-Token header carries an admin token
signed with jwt.filer_signing.key comes from a component that holds the
cluster key: the vector engine or the Lance maintenance worker. Mark such
a request as trusted. It still authenticates as an identity, which stays
the principal for vended credentials. A dedicated header leaves the
Authorization header to Lance OAuth clients. A gateway with no key
trusts no one.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* lance: vend write credentials to a vector bucket's indexes only to trusted callers

An index in a vector bucket's vector namespace is also a Lance table, so
a caller granted s3tables:UpdateTable could describe it, or declare it
again, and get credentials that write the index data without passing the
vector bucket policy. In that namespace both now vend read-only
credentials unless the caller is trusted, and a declare from anyone else
no longer writes the dataset markers.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* lance worker: prove the cluster key to the Lance namespace

The gateway now vends write credentials to a vector bucket's indexes only
to callers that send a token signed with jwt.filer_signing.key. The
worker already takes that key for the admin stream; its namespace client
now mints a token from it for every catalog request.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* weed-vector: describe the index the gateway created and prove the cluster key

CreateIndex declared a table the gateway had just created, which only
worked because declare is idempotent. Describe it instead. The Lance
namespace vends write credentials to a vector bucket's indexes only to
callers that send a token signed with jwt.filer_signing.key, so take the
key with --filer-signing-key, the same flag the Lance worker has.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* lance worker: keep the cluster key out of Debug output

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* lance: answer a declare of an existing table with its own location

CreateTable succeeds for a name that already exists without looking at
the location requested, and declare kept using that location for the
dataset markers and the vended credentials. Declaring an existing table
with another table's location, such as a vector index, got credentials
to that table's data. Use the location the catalog holds, which
CreateTable now also reports for a new table.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* lance: sign cluster tokens with a key derived for the catalog

A cluster token was a filer admin token signed with jwt.filer_signing.key
and, by default, never expired. Sent to the Lance namespace on every
request, often over plain HTTP, a captured one would also authorize the
filer's IAM service. Sign it with a key derived from the cluster key for
this one use, which the admin services do not accept, and mint it with a
five minute lifetime. weed-vector drops its token lifetime flag.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* lance: follow a reloaded jwt.filer_signing.key

The Lance namespace copied the cluster key at startup. A SIGHUP reload
refreshes the gateway's filer guard but not that copy, so after a key
rotation the restarted vector engine and worker were no longer trusted.
Read the key from the guard on every check.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* iceberg: keep a vector bucket's indexes read-only

An index in a vector bucket is also a catalog table the Iceberg REST
catalog loads, and LoadTable vended credentials that could write it to
any caller granted s3tables:UpdateTable. A commit staged its metadata
file in the index directory on the same grant. Apply the vector
namespace rule to both, after the caller's own UpdateTable check.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* iceberg: confine a created table's location to its namespace before writing

Stage-create skips the registration that confines a location to the
table's namespace, yet vends write credentials for that location. A
caller allowed to create a table in any namespace of a bucket could name
another namespace's prefix, a vector index included, and get write
credentials to it. The non-staged create also wrote its first metadata
file there before the registration refused the location.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* lance: a declare that finds an existing table answers like describe

Creating a name that already exists is idempotent, so a caller allowed
only to create tables could declare someone else's table and get write
credentials to it, and clear its deregistered marker. Describe would have
vended that caller read-only credentials. A declare now looks the name up
first; when it only finds the table, writing needs UpdateTable, as it does
for describe.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-01 18:49:01 -08:00

310 lines
12 KiB
Go

package lance
import (
"context"
"crypto/hmac"
"crypto/sha256"
"net/http"
"strings"
"time"
"github.com/gorilla/mux"
"github.com/seaweedfs/seaweedfs/weed/glog"
"github.com/seaweedfs/seaweedfs/weed/pb/filer_pb"
"github.com/seaweedfs/seaweedfs/weed/s3api/s3_constants"
"github.com/seaweedfs/seaweedfs/weed/s3api/s3err"
"github.com/seaweedfs/seaweedfs/weed/s3api/s3tables"
"github.com/seaweedfs/seaweedfs/weed/security"
)
// FilerClient provides access to the filer for storage operations.
type FilerClient interface {
WithFilerClient(streamingMode bool, fn func(client filer_pb.SeaweedFilerClient) error) error
}
type S3Authenticator interface {
AuthenticateRequest(r *http.Request) (string, interface{}, s3err.ErrorCode)
DefaultAllow() bool
}
// CredentialValidator validates S3 access key / secret key pairs
// and provides credential lookup for OAuth token verification.
type CredentialValidator interface {
// ValidateS3Credential checks if the access key and secret key are valid.
// Returns the identity name and identity object on success.
ValidateS3Credential(accessKey, secretKey string) (identityName string, identity interface{}, err error)
// GetCredentialByAccessKey looks up a credential by access key.
// Returns the identity name, identity object, and secret key.
// Used for verifying Bearer tokens signed with a specific credential.
GetCredentialByAccessKey(accessKey string) (identityName string, identity interface{}, secretKey string, err error)
}
// VendedCredentials are short-lived S3 credentials scoped to one table.
type VendedCredentials struct {
AccessKeyID string
SecretAccessKey string
SessionToken string
Expiration time.Time
}
// CredentialVendor mints credentials limited to a single table's prefix for a
// caller the catalog has already authenticated and authorized, read-only when
// readOnly is set. A nil result with no error means the deployment has vending
// switched off.
type CredentialVendor interface {
VendTableCredentials(ctx context.Context, principal, bucket, prefix string, readOnly bool) (*VendedCredentials, error)
}
// Server implements the Lance Namespace REST spec.
type Server struct {
filerClient FilerClient
tablesManager *s3tables.Manager
authenticator S3Authenticator
credentialValidator CredentialValidator
credentialVendor CredentialVendor
s3Endpoint string
s3Region string
signingKey func() security.SigningKey
}
// NewServer creates a Lance namespace server over the given filer.
func NewServer(filerClient FilerClient, authenticator S3Authenticator) *Server {
manager := s3tables.NewManager()
// Mirror the S3 port: fall open by default only when the gateway itself is
// open, so an authenticated caller still passes the normal permission check.
if authenticator != nil {
manager.SetDefaultAllow(authenticator.DefaultAllow())
}
return &Server{
filerClient: filerClient,
tablesManager: manager,
authenticator: authenticator,
}
}
// SetCredentialValidator sets the credential validator for OAuth token support.
func (s *Server) SetCredentialValidator(cv CredentialValidator) {
s.credentialValidator = cv
}
// SetCredentialVendor enables storage_options credential vending for clients
// that ask for it with vend_credentials.
func (s *Server) SetCredentialVendor(vendor CredentialVendor) {
s.credentialVendor = vendor
}
// SetSigningKey sets where to read the cluster key (jwt.filer_signing.key)
// that a trusted caller's clusterTokenHeader is checked against. It is read on
// every check, so a reloaded key takes effect at once.
func (s *Server) SetSigningKey(signingKey func() security.SigningKey) {
s.signingKey = signingKey
}
// SetS3Endpoint configures the S3 endpoint advertised in storage_options so a
// client can reach the dataset without separately discovering the S3 address.
func (s *Server) SetS3Endpoint(endpoint string) {
s.s3Endpoint = endpoint
}
// SetS3Region configures the region advertised in storage_options.
func (s *Server) SetS3Region(region string) {
s.s3Region = region
}
// RegisterRoutes registers the Lance Namespace REST routes.
//
// The spec puts the identifier in the path rather than the body so a reverse
// proxy can route and authorize without deserializing the request.
func (s *Server) RegisterRoutes(router *mux.Router) {
router.Use(loggingMiddleware)
// OAuth2 token endpoint - no auth needed (this IS the auth endpoint)
router.HandleFunc("/oauth/token", s.handleOAuthTokens).Methods(http.MethodPost)
router.HandleFunc("/v1/namespace/{id}/create", s.Auth(s.handleCreateNamespace)).Methods(http.MethodPost)
router.HandleFunc("/v1/namespace/{id}/list", s.Auth(s.handleListNamespaces)).Methods(http.MethodGet)
router.HandleFunc("/v1/namespace/{id}/describe", s.Auth(s.handleDescribeNamespace)).Methods(http.MethodPost)
router.HandleFunc("/v1/namespace/{id}/drop", s.Auth(s.handleDropNamespace)).Methods(http.MethodPost)
router.HandleFunc("/v1/namespace/{id}/exists", s.Auth(s.handleNamespaceExists)).Methods(http.MethodPost)
router.HandleFunc("/v1/namespace/{id}/table/list", s.Auth(s.handleListTables)).Methods(http.MethodGet)
router.HandleFunc("/v1/table", s.Auth(s.handleListAllTables)).Methods(http.MethodGet)
router.HandleFunc("/v1/table/{id}/declare", s.Auth(s.handleDeclareTable)).Methods(http.MethodPost)
router.HandleFunc("/v1/table/{id}/describe", s.Auth(s.handleDescribeTable)).Methods(http.MethodPost)
router.HandleFunc("/v1/table/{id}/exists", s.Auth(s.handleTableExists)).Methods(http.MethodPost)
router.HandleFunc("/v1/table/{id}/register", s.Auth(s.handleRegisterTable)).Methods(http.MethodPost)
router.HandleFunc("/v1/table/{id}/deregister", s.Auth(s.handleDeregisterTable)).Methods(http.MethodPost)
router.HandleFunc("/v1/table/{id}/drop", s.Auth(s.handleDropTable)).Methods(http.MethodPost)
router.HandleFunc("/v1/table/{id}/rename", s.Auth(s.handleRenameTable)).Methods(http.MethodPost)
// The data plane needs Lance format support that does not exist in Go. Say
// so with the spec's own code instead of returning a bare 404.
for _, action := range []string{"create", "insert", "merge_insert", "update", "delete",
"query", "count_rows", "explain_plan", "analyze_plan", "restore",
"add_columns", "alter_columns", "drop_columns", "backfill_column",
"create_index", "create_scalar_index", "stats", "schema_metadata/update"} {
router.HandleFunc("/v1/table/{id}/"+action, s.Auth(s.handleUnsupported)).Methods(http.MethodPost)
}
// Version ops exist in the spec for stores that cannot order commits
// themselves. Ours can: a Lance commit is a put-if-not-exists, and this S3
// evaluates that precondition at the object's owner filer under a per-path
// lock, so the dataset keeps its own version history and the catalog stays
// out of the commit path.
for _, action := range []string{
"version/create", "version/list", "version/describe", "version/delete",
"index/list", "tags/list", "tags/version", "tags/create", "tags/delete", "tags/update",
"branches/list", "branches/create", "branches/delete"} {
router.HandleFunc("/v1/table/{id}/"+action, s.Auth(s.handleUnsupported)).Methods(http.MethodPost)
}
router.PathPrefix("/").HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
glog.V(2).Infof("lance: no route for %s %s", r.Method, r.URL.EscapedPath())
writeError(w, r, http.StatusNotFound, codeUnsupported, "no such operation")
})
glog.V(2).Infof("Registered Lance Namespace routes")
}
func (s *Server) handleUnsupported(w http.ResponseWriter, r *http.Request) {
writeError(w, r, http.StatusNotImplemented, codeUnsupported,
"this namespace records table metadata only; run data operations through a Lance client against the table location")
}
func loggingMiddleware(next http.Handler) http.Handler {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
glog.V(2).Infof("lance request: %s %s from %s", r.Method, r.URL.EscapedPath(), r.RemoteAddr)
next.ServeHTTP(w, r)
})
}
// Auth authenticates the caller and puts the identity in the request context.
// Lance clients authenticate the catalog with a Bearer token or an x-api-key
// header; the S3 authenticator stays for callers that can SigV4-sign.
func (s *Server) Auth(handler http.HandlerFunc) http.HandlerFunc {
return func(w http.ResponseWriter, r *http.Request) {
if s.hasClusterToken(r) {
r = r.WithContext(withTrustedCaller(r.Context()))
}
// A request carrying a Bearer token is a Lance REST client. An invalid
// or expired token gets 401 immediately; falling through to the S3
// authenticator would parse the header as SigV4 and fail with a
// different error that clients cannot act on.
// The auth scheme is case-insensitive (RFC 7235).
if strings.HasPrefix(strings.ToLower(r.Header.Get("Authorization")), "bearer ") {
if identityName, identity, accessKey, secretKey, ok := s.authenticateBearerKey(r); ok {
ctx := r.Context()
ctx = s3_constants.SetIdentityNameInContext(ctx, identityName)
ctx = s3_constants.SetAccessKeyIdInContext(ctx, accessKey)
ctx = s3_constants.SetStoredAccessKeyIdInContext(ctx, accessKey, secretKey)
if identity != nil {
ctx = s3_constants.SetIdentityInContext(ctx, identity)
}
r = r.WithContext(ctx)
handler(w, r)
return
}
w.Header().Set("WWW-Authenticate", "Bearer")
writeError(w, r, http.StatusUnauthorized, codeUnauthenticated, "Bearer token is invalid or expired")
return
}
if apiKey := r.Header.Get("x-api-key"); apiKey != "" {
if identityName, identity, ok := s.authenticateApiKey(apiKey); ok {
accessKey, secretKey, _ := strings.Cut(apiKey, ":")
ctx := r.Context()
ctx = s3_constants.SetIdentityNameInContext(ctx, identityName)
ctx = s3_constants.SetAccessKeyIdInContext(ctx, accessKey)
ctx = s3_constants.SetStoredAccessKeyIdInContext(ctx, accessKey, secretKey)
if identity != nil {
ctx = s3_constants.SetIdentityInContext(ctx, identity)
}
r = r.WithContext(ctx)
handler(w, r)
return
}
writeError(w, r, http.StatusUnauthorized, codeUnauthenticated, "invalid x-api-key")
return
}
if s.authenticator == nil {
writeError(w, r, http.StatusUnauthorized, codeUnauthenticated, "authentication required")
return
}
// The verifier records the key of the credential that matched into
// this holder; the vended credentials name it.
r = s3_constants.EnsureIdentityHolder(r)
identityName, identity, errCode := s.authenticator.AuthenticateRequest(r)
if errCode != s3err.ErrNone {
apiErr := s3err.GetAPIError(errCode)
code := codeInternal
switch apiErr.HTTPStatusCode {
case http.StatusForbidden:
code = codePermissionDenied
case http.StatusUnauthorized:
code = codeUnauthenticated
case http.StatusBadRequest:
code = codeInvalidInput
}
writeError(w, r, apiErr.HTTPStatusCode, code, apiErr.Description)
return
}
if identityName != "" || identity != nil {
ctx := r.Context()
if identityName != "" {
ctx = s3_constants.SetIdentityNameInContext(ctx, identityName)
}
if identity != nil {
ctx = s3_constants.SetIdentityInContext(ctx, identity)
}
r = r.WithContext(ctx)
}
handler(w, r)
}
}
// clusterTokenHeader carries a token signed with clusterTokenKey. The vector
// engine and the Lance maintenance worker send it next to their identity's
// credentials, and only they may write index data in a vector bucket.
const clusterTokenHeader = "X-Seaweedfs-Cluster-Token"
// clusterTokenKey derives the key cluster tokens are signed with from the
// cluster key, so a token captured on its way here does not authorize the
// admin calls that take the cluster key itself.
func clusterTokenKey(key security.SigningKey) security.SigningKey {
mac := hmac.New(sha256.New, key)
mac.Write([]byte("seaweedfs-lance-cluster-token"))
return mac.Sum(nil)
}
type trustedCallerKey struct{}
func withTrustedCaller(ctx context.Context) context.Context {
return context.WithValue(ctx, trustedCallerKey{}, true)
}
func isTrustedCaller(ctx context.Context) bool {
trusted, _ := ctx.Value(trustedCallerKey{}).(bool)
return trusted
}
// hasClusterToken reports whether the request proves it holds the cluster key.
// Without a configured key no caller can.
func (s *Server) hasClusterToken(r *http.Request) bool {
token := r.Header.Get(clusterTokenHeader)
if token == "" || s.signingKey == nil {
return false
}
key := s.signingKey()
if len(key) == 0 {
return false
}
if security.IsFilerAdminToken(clusterTokenKey(key), security.EncodedJwt(token)) {
return true
}
glog.V(1).Infof("lance: %s from %s is not signed with a key derived from jwt.filer_signing.key", clusterTokenHeader, r.RemoteAddr)
return false
}