hub: S1 wgsync (pinned-SSH push + declarative reconciler) + /admin/wg API + env wiring

internal/wgsync: x/crypto/ssh client with ssh.FixedHostKey pin (no insecure
fallback), forced-command exec, ok/applied response contract; Reconciler pushes
the FULL peer list on Trigger or 5-min tick (drift repair by construction).
internal/api/wg.go: PUT/GET /admin/wg/endpoint + POST/DELETE/GET /admin/wg/peers,
global-key-only, pubkey in body (base64 vs URL), sync ok|deferred|disabled.
main.go: WG_ENDPOINT_SSH_* env wiring, disabled-with-INFO when unconfigured.
Groups B/C/D tests incl. in-process SSH server; red-proofs b/c/d run + reverted.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PSK5g6qYLknKj8u3QAFEr6
This commit is contained in:
2026-07-03 23:40:22 +02:00
parent b18f6aee1b
commit fbeeacb124
8 changed files with 1190 additions and 0 deletions
+16
View File
@@ -49,6 +49,10 @@ type Handler struct {
mailSender mailrelay.Sender
mailLimiter *mailRateLimiter
mailFromAllow map[string]bool
// S1 offsite connectivity: the wgsync reconciler seam (internal/api/wg.go). nil = peer-sync
// disabled — mutations still persist, responses carry sync:"disabled".
wgSyncer WGSyncer
}
// SetLatestVersionProvider wires the registry version checker so the controller report ACK can
@@ -187,6 +191,18 @@ func (h *Handler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
case r.Method == http.MethodPost && strings.HasPrefix(path, "/admin/hosts/") && strings.HasSuffix(path, "/jobs"):
hostID := strings.TrimSuffix(strings.TrimPrefix(path, "/admin/hosts/"), "/jobs")
h.handleAdminEnqueueJob(w, r, hostID)
// S1 offsite connectivity — WG endpoint record + peer registry (global key only, api/wg.go).
// DELETE carries the pubkey in the body: base64 '/'+'+' keep pubkeys out of URL paths.
case r.Method == http.MethodPut && path == "/admin/wg/endpoint":
h.handleAdminSetWGEndpoint(w, r)
case r.Method == http.MethodGet && path == "/admin/wg/endpoint":
h.handleAdminGetWGEndpoint(w, r)
case r.Method == http.MethodPost && path == "/admin/wg/peers":
h.handleAdminAddWGPeer(w, r)
case r.Method == http.MethodDelete && path == "/admin/wg/peers":
h.handleAdminDeleteWGPeer(w, r)
case r.Method == http.MethodGet && path == "/admin/wg/peers":
h.handleAdminListWGPeers(w, r)
case r.Method == http.MethodPost && path == "/event":
h.handleEvent(w, r)
case r.Method == http.MethodPost && path == "/mail":
+264
View File
@@ -0,0 +1,264 @@
package api
// S1 offsite connectivity (doc 06 §3.2/§5): the operator admin surface for the WG endpoint
// record + peer registry. GLOBAL key ONLY on every route (the handleAdminSetDesiredState gate) —
// a per-host key must never author the peer list; the box-facing registration path is S2.
// DELETE takes the pubkey in the JSON body: WG pubkeys are std base64 ('/' and '+'), so a pubkey
// NEVER appears in a URL path — and no, URL-escaping is not the fix (see the S1 spec §8).
import (
"context"
"database/sql"
"encoding/base64"
"encoding/json"
"fmt"
"io"
"net/http"
"net/netip"
"time"
"gitea.dooplex.hu/admin/felhom-hub/internal/store"
)
// WGSyncer is the reconciler seam (satisfied by *wgsync.Reconciler; tests inject a fake). nil =
// peer-sync disabled: mutations still hit the DB (the source of truth) and report sync:"disabled".
type WGSyncer interface {
SyncNow(ctx context.Context) error
Trigger()
}
// SetWGSyncer wires the wgsync reconciler (mirror of SetLatestVersionProvider; nil-safe).
func (h *Handler) SetWGSyncer(s WGSyncer) {
h.wgSyncer = s
}
// validateWGPubkey enforces the exact WG public-key shape: 44 chars of std base64 decoding to
// 32 bytes. Anything else is rejected before any allocation.
func validateWGPubkey(pk string) error {
if len(pk) != 44 {
return fmt.Errorf("pubkey must be 44 base64 chars, got %d", len(pk))
}
raw, err := base64.StdEncoding.DecodeString(pk)
if err != nil {
return fmt.Errorf("pubkey is not valid base64: %v", err)
}
if len(raw) != 32 {
return fmt.Errorf("pubkey must decode to 32 bytes, got %d", len(raw))
}
return nil
}
// syncAfterMutation runs an inline sync after a peer mutation. The DB write already happened —
// it is the source of truth — so a push failure is REPORTED, not rolled back: the reconciler's
// next tick converges the endpoint (Scenario D).
func (h *Handler) syncAfterMutation(ctx context.Context) string {
if h.wgSyncer == nil {
return "disabled"
}
syncCtx, cancel := context.WithTimeout(ctx, 10*time.Second)
defer cancel()
if err := h.wgSyncer.SyncNow(syncCtx); err != nil {
h.logger.Printf("[ERROR] wgsync: inline push after mutation failed: %v (reconciler will retry)", err)
h.wgSyncer.Trigger()
return "deferred: " + err.Error()
}
return "ok"
}
// handleAdminSetWGEndpoint — PUT /admin/wg/endpoint. Upserts the endpoint record.
func (h *Handler) handleAdminSetWGEndpoint(w http.ResponseWriter, r *http.Request) {
_, _, isGlobal, ok := h.checkAuthHost(r)
if !ok || !isGlobal {
http.Error(w, "Forbidden: global key required", http.StatusForbidden)
return
}
body, err := io.ReadAll(io.LimitReader(r.Body, 1<<20))
if err != nil {
http.Error(w, "Bad request", http.StatusBadRequest)
return
}
var req struct {
EndpointID string `json:"endpoint_id"`
DNSName string `json:"dns_name"`
WGPort int `json:"wg_port"`
ServerPubkey string `json:"server_pubkey"`
TunnelSubnet string `json:"tunnel_subnet"`
PBSTunnelIP string `json:"pbs_tunnel_ip"`
}
if err := json.Unmarshal(body, &req); err != nil {
http.Error(w, "Invalid payload: body must be JSON", http.StatusBadRequest)
return
}
if req.DNSName == "" {
http.Error(w, "Invalid payload: dns_name required", http.StatusBadRequest)
return
}
if req.WGPort < 1 || req.WGPort > 65535 {
http.Error(w, "Invalid payload: wg_port must be 1-65535", http.StatusBadRequest)
return
}
if err := validateWGPubkey(req.ServerPubkey); err != nil {
http.Error(w, "Invalid payload: server_pubkey: "+err.Error(), http.StatusBadRequest)
return
}
prefix, err := netip.ParsePrefix(req.TunnelSubnet)
if err != nil {
http.Error(w, "Invalid payload: tunnel_subnet must be CIDR", http.StatusBadRequest)
return
}
pbsAddr, err := netip.ParseAddr(req.PBSTunnelIP)
if err != nil || !prefix.Contains(pbsAddr) {
http.Error(w, "Invalid payload: pbs_tunnel_ip must be an address inside tunnel_subnet", http.StatusBadRequest)
return
}
if err := h.store.SetWGEndpoint(&store.WGEndpoint{
EndpointID: req.EndpointID, DNSName: req.DNSName, WGPort: req.WGPort,
ServerPubkey: req.ServerPubkey, TunnelSubnet: req.TunnelSubnet, PBSTunnelIP: req.PBSTunnelIP,
}); err != nil {
h.logger.Printf("[ERROR] set wg endpoint: %v", err)
http.Error(w, "Internal error", http.StatusInternalServerError)
return
}
h.logger.Printf("[INFO] wg endpoint set: %s (%s:%d, subnet %s)", req.DNSName, req.DNSName, req.WGPort, req.TunnelSubnet)
w.Header().Set("Content-Type", "application/json")
w.WriteHeader(http.StatusOK)
w.Write([]byte(`{"status":"ok"}`))
}
// handleAdminGetWGEndpoint — GET /admin/wg/endpoint.
func (h *Handler) handleAdminGetWGEndpoint(w http.ResponseWriter, r *http.Request) {
_, _, isGlobal, ok := h.checkAuthHost(r)
if !ok || !isGlobal {
http.Error(w, "Forbidden: global key required", http.StatusForbidden)
return
}
e, err := h.store.GetWGEndpoint()
if err == sql.ErrNoRows {
http.Error(w, "wg endpoint not configured", http.StatusNotFound)
return
}
if err != nil {
h.logger.Printf("[ERROR] get wg endpoint: %v", err)
http.Error(w, "Internal error", http.StatusInternalServerError)
return
}
w.Header().Set("Content-Type", "application/json")
json.NewEncoder(w).Encode(map[string]interface{}{
"endpoint_id": e.EndpointID, "dns_name": e.DNSName, "wg_port": e.WGPort,
"server_pubkey": e.ServerPubkey, "tunnel_subnet": e.TunnelSubnet, "pbs_tunnel_ip": e.PBSTunnelIP,
})
}
// handleAdminAddWGPeer — POST /admin/wg/peers. Allocates a /32 (idempotent on pubkey) and
// pushes the full list inline (sync semantics: ok | deferred | disabled).
func (h *Handler) handleAdminAddWGPeer(w http.ResponseWriter, r *http.Request) {
_, _, isGlobal, ok := h.checkAuthHost(r)
if !ok || !isGlobal {
http.Error(w, "Forbidden: global key required", http.StatusForbidden)
return
}
body, err := io.ReadAll(io.LimitReader(r.Body, 1<<20))
if err != nil {
http.Error(w, "Bad request", http.StatusBadRequest)
return
}
var req struct {
Pubkey string `json:"pubkey"`
HostID string `json:"host_id"`
Note string `json:"note"`
}
if err := json.Unmarshal(body, &req); err != nil {
http.Error(w, "Invalid payload: body must be JSON", http.StatusBadRequest)
return
}
if err := validateWGPubkey(req.Pubkey); err != nil {
http.Error(w, "Invalid payload: "+err.Error(), http.StatusBadRequest)
return
}
ip, existed, err := h.store.AddWGPeer(req.Pubkey, req.HostID, req.Note)
if err == store.ErrWGEndpointUnset {
http.Error(w, "wg endpoint not configured", http.StatusConflict)
return
}
if err == store.ErrWGSubnetExhausted {
http.Error(w, "tunnel subnet exhausted", http.StatusConflict)
return
}
if err != nil {
h.logger.Printf("[ERROR] add wg peer: %v", err)
http.Error(w, "Internal error", http.StatusInternalServerError)
return
}
syncStatus := h.syncAfterMutation(r.Context())
h.logger.Printf("[INFO] wg peer added: %s -> %s/32 (existed=%v, sync=%s)", req.Pubkey, ip, existed, syncStatus)
w.Header().Set("Content-Type", "application/json")
w.WriteHeader(http.StatusOK)
json.NewEncoder(w).Encode(map[string]interface{}{
"pubkey": req.Pubkey, "assigned_ip": ip + "/32", "existed": existed, "sync": syncStatus,
})
}
// handleAdminDeleteWGPeer — DELETE /admin/wg/peers, pubkey in the JSON body (never the URL).
// Unknown pubkey → 404 with NO sync (nothing changed). Known → delete + inline push: the pushed
// full list no longer contains the peer, so revocation lands with the push (Scenario B).
func (h *Handler) handleAdminDeleteWGPeer(w http.ResponseWriter, r *http.Request) {
_, _, isGlobal, ok := h.checkAuthHost(r)
if !ok || !isGlobal {
http.Error(w, "Forbidden: global key required", http.StatusForbidden)
return
}
body, err := io.ReadAll(io.LimitReader(r.Body, 1<<20))
if err != nil {
http.Error(w, "Bad request", http.StatusBadRequest)
return
}
var req struct {
Pubkey string `json:"pubkey"`
}
if err := json.Unmarshal(body, &req); err != nil {
http.Error(w, "Invalid payload: body must be JSON", http.StatusBadRequest)
return
}
if err := validateWGPubkey(req.Pubkey); err != nil {
http.Error(w, "Invalid payload: "+err.Error(), http.StatusBadRequest)
return
}
err = h.store.RemoveWGPeer(req.Pubkey)
if err == sql.ErrNoRows {
http.Error(w, "Unknown pubkey", http.StatusNotFound)
return
}
if err != nil {
h.logger.Printf("[ERROR] remove wg peer: %v", err)
http.Error(w, "Internal error", http.StatusInternalServerError)
return
}
syncStatus := h.syncAfterMutation(r.Context())
h.logger.Printf("[INFO] wg peer removed: %s (sync=%s)", req.Pubkey, syncStatus)
w.Header().Set("Content-Type", "application/json")
w.WriteHeader(http.StatusOK)
json.NewEncoder(w).Encode(map[string]interface{}{"status": "ok", "sync": syncStatus})
}
// handleAdminListWGPeers — GET /admin/wg/peers. The verification surface (S2 builds UI on top).
func (h *Handler) handleAdminListWGPeers(w http.ResponseWriter, r *http.Request) {
_, _, isGlobal, ok := h.checkAuthHost(r)
if !ok || !isGlobal {
http.Error(w, "Forbidden: global key required", http.StatusForbidden)
return
}
peers, err := h.store.ListWGPeers()
if err != nil {
h.logger.Printf("[ERROR] list wg peers: %v", err)
http.Error(w, "Internal error", http.StatusInternalServerError)
return
}
out := make([]map[string]interface{}, 0, len(peers))
for _, p := range peers {
out = append(out, map[string]interface{}{
"pubkey": p.Pubkey, "assigned_ip": p.AssignedIP + "/32", "host_id": p.HostID, "note": p.Note,
})
}
w.Header().Set("Content-Type", "application/json")
json.NewEncoder(w).Encode(map[string]interface{}{"peers": out})
}
+212
View File
@@ -0,0 +1,212 @@
package api
// Group B — WG admin API auth + validation (Scenario C). Non-hollow: asserts store effects and
// fake-syncer call counts, not just statuses.
import (
"bytes"
"context"
"encoding/base64"
"encoding/json"
"net/http"
"testing"
"gitea.dooplex.hu/admin/felhom-hub/internal/store"
)
// fakeWGSyncer counts SyncNow/Trigger calls; err scripts the SyncNow result.
type fakeWGSyncer struct {
syncCalls int
triggerCalls int
err error
}
func (f *fakeWGSyncer) SyncNow(ctx context.Context) error { f.syncCalls++; return f.err }
func (f *fakeWGSyncer) Trigger() { f.triggerCalls++ }
// testPK returns a VALID WG-shaped pubkey (44 std-base64 chars, 32 bytes) unique per fill byte.
func testPK(fill byte) string {
return base64.StdEncoding.EncodeToString(bytes.Repeat([]byte{fill}, 32))
}
func putTestEndpoint(t *testing.T, h *Handler) {
t.Helper()
body := `{"dns_name":"ep0.example","wg_port":443,"server_pubkey":"` + testPK(9) + `",` +
`"tunnel_subnet":"10.77.0.0/24","pbs_tunnel_ip":"10.77.0.1"}`
rr := do(h, http.MethodPut, "/admin/wg/endpoint", globalKey, body)
if rr.Code != http.StatusOK {
t.Fatalf("PUT endpoint = %d: %s", rr.Code, rr.Body.String())
}
}
func TestWGPeers_PerHostKeyForbidden(t *testing.T) {
h, st, _ := newTestHandler(t)
st.UpsertHost(&store.Host{HostID: "h1", CustomerID: "c1", APIKey: "HKEY"})
putTestEndpoint(t, h)
fake := &fakeWGSyncer{}
h.SetWGSyncer(fake)
for _, m := range []string{http.MethodPost, http.MethodDelete, http.MethodGet} {
rr := do(h, m, "/admin/wg/peers", "HKEY", `{"pubkey":"`+testPK(1)+`"}`)
if rr.Code != http.StatusForbidden {
t.Errorf("%s with per-host key = %d, want 403", m, rr.Code)
}
}
rr := do(h, http.MethodPut, "/admin/wg/endpoint", "HKEY", `{}`)
if rr.Code != http.StatusForbidden {
t.Errorf("PUT endpoint with per-host key = %d, want 403", rr.Code)
}
peers, _ := st.ListWGPeers()
if len(peers) != 0 {
t.Errorf("rows created despite 403: %d", len(peers))
}
if fake.syncCalls != 0 {
t.Errorf("sync ran despite 403: %d calls", fake.syncCalls)
}
}
func TestWGPeers_AddHappyPath(t *testing.T) {
h, st, _ := newTestHandler(t)
putTestEndpoint(t, h)
fake := &fakeWGSyncer{}
h.SetWGSyncer(fake)
rr := do(h, http.MethodPost, "/admin/wg/peers", globalKey, `{"pubkey":"`+testPK(1)+`","note":"test"}`)
if rr.Code != http.StatusOK {
t.Fatalf("POST = %d: %s", rr.Code, rr.Body.String())
}
var resp struct {
Pubkey string `json:"pubkey"`
AssignedIP string `json:"assigned_ip"`
Existed bool `json:"existed"`
Sync string `json:"sync"`
}
json.Unmarshal(rr.Body.Bytes(), &resp)
if resp.AssignedIP != "10.77.0.2/32" || resp.Existed || resp.Sync != "ok" {
t.Errorf("resp = %+v, want .2/32 existed=false sync=ok", resp)
}
if fake.syncCalls != 1 {
t.Errorf("sync calls = %d, want 1", fake.syncCalls)
}
peers, _ := st.ListWGPeers()
if len(peers) != 1 || peers[0].AssignedIP != "10.77.0.2" || peers[0].Note != "test" {
t.Errorf("stored peers = %+v", peers)
}
}
func TestWGPeers_SyncDeferredOnPushFailure(t *testing.T) {
h, _, _ := newTestHandler(t)
putTestEndpoint(t, h)
fake := &fakeWGSyncer{err: context.DeadlineExceeded}
h.SetWGSyncer(fake)
rr := do(h, http.MethodPost, "/admin/wg/peers", globalKey, `{"pubkey":"`+testPK(1)+`"}`)
if rr.Code != http.StatusOK {
t.Fatalf("POST = %d (DB write is source of truth; push failure must not fail the request)", rr.Code)
}
var resp struct {
Sync string `json:"sync"`
}
json.Unmarshal(rr.Body.Bytes(), &resp)
if len(resp.Sync) < 8 || resp.Sync[:8] != "deferred" {
t.Errorf("sync = %q, want deferred:...", resp.Sync)
}
if fake.triggerCalls != 1 {
t.Errorf("Trigger calls = %d, want 1 (reconciler retry requested)", fake.triggerCalls)
}
}
func TestWGPeers_SyncDisabledWhenUnwired(t *testing.T) {
h, _, _ := newTestHandler(t)
putTestEndpoint(t, h)
// no SetWGSyncer — nil seam
rr := do(h, http.MethodPost, "/admin/wg/peers", globalKey, `{"pubkey":"`+testPK(1)+`"}`)
if rr.Code != http.StatusOK {
t.Fatalf("POST = %d", rr.Code)
}
var resp struct {
Sync string `json:"sync"`
}
json.Unmarshal(rr.Body.Bytes(), &resp)
if resp.Sync != "disabled" {
t.Errorf("sync = %q, want disabled", resp.Sync)
}
}
func TestWGPeers_BadPubkeyRejected(t *testing.T) {
h, st, _ := newTestHandler(t)
putTestEndpoint(t, h)
bad := []string{
"not-base64",
base64.StdEncoding.EncodeToString(bytes.Repeat([]byte{1}, 16)), // 24 chars
"!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!", // 44 chars, not base64
base64.StdEncoding.EncodeToString(bytes.Repeat([]byte{1}, 33)), // 44 chars but 33 bytes
}
for _, pk := range bad {
rr := do(h, http.MethodPost, "/admin/wg/peers", globalKey, `{"pubkey":"`+pk+`"}`)
if rr.Code != http.StatusBadRequest {
t.Errorf("pubkey %q = %d, want 400", pk, rr.Code)
}
}
peers, _ := st.ListWGPeers()
if len(peers) != 0 {
t.Errorf("allocation happened for bad pubkey: %+v", peers)
}
}
func TestWGPeers_NoEndpointIs409(t *testing.T) {
h, _, _ := newTestHandler(t)
rr := do(h, http.MethodPost, "/admin/wg/peers", globalKey, `{"pubkey":"`+testPK(1)+`"}`)
if rr.Code != http.StatusConflict {
t.Errorf("POST without endpoint = %d, want 409", rr.Code)
}
}
func TestWGEndpoint_PBSOutsideSubnetRejected(t *testing.T) {
h, _, _ := newTestHandler(t)
body := `{"dns_name":"ep0.example","wg_port":443,"server_pubkey":"` + testPK(9) + `",` +
`"tunnel_subnet":"10.77.0.0/24","pbs_tunnel_ip":"10.88.0.1"}`
rr := do(h, http.MethodPut, "/admin/wg/endpoint", globalKey, body)
if rr.Code != http.StatusBadRequest {
t.Errorf("PUT with pbs outside subnet = %d, want 400", rr.Code)
}
rr = do(h, http.MethodGet, "/admin/wg/endpoint", globalKey, "")
if rr.Code != http.StatusNotFound {
t.Errorf("GET after rejected PUT = %d, want 404 (nothing stored)", rr.Code)
}
}
func TestWGPeers_DeleteUnknown404NoSync(t *testing.T) {
h, _, _ := newTestHandler(t)
putTestEndpoint(t, h)
fake := &fakeWGSyncer{}
h.SetWGSyncer(fake)
rr := do(h, http.MethodDelete, "/admin/wg/peers", globalKey, `{"pubkey":"`+testPK(7)+`"}`)
if rr.Code != http.StatusNotFound {
t.Errorf("DELETE unknown = %d, want 404", rr.Code)
}
if fake.syncCalls != 0 {
t.Errorf("sync ran on 404 delete: %d", fake.syncCalls)
}
}
func TestWGPeers_DeleteKnownRemovesAndSyncs(t *testing.T) {
h, st, _ := newTestHandler(t)
putTestEndpoint(t, h)
fake := &fakeWGSyncer{}
h.SetWGSyncer(fake)
do(h, http.MethodPost, "/admin/wg/peers", globalKey, `{"pubkey":"`+testPK(1)+`"}`)
do(h, http.MethodPost, "/admin/wg/peers", globalKey, `{"pubkey":"`+testPK(2)+`"}`)
rr := do(h, http.MethodDelete, "/admin/wg/peers", globalKey, `{"pubkey":"`+testPK(1)+`"}`)
if rr.Code != http.StatusOK {
t.Fatalf("DELETE = %d: %s", rr.Code, rr.Body.String())
}
peers, _ := st.ListWGPeers()
if len(peers) != 1 || peers[0].Pubkey != testPK(2) {
t.Errorf("peers after delete = %+v, want only p2", peers)
}
if fake.syncCalls != 3 { // 2 adds + 1 delete
t.Errorf("sync calls = %d, want 3", fake.syncCalls)
}
}