mesh-controller/internal/link/handover.go

package link

import (
	"context"
	"crypto/rand"
	"encoding/hex"
	"encoding/json"
	"errors"
	"fmt"
	"time"

	"github.com/nats-io/nats.go"

	"github.com/novox/mesh-controller/internal/broker"
)

// A hand-over asked of a machine's node-engine (novox/hq issue 356, issue 339).
//
// The operator hands a directory the node-engine uses as found to the mesh at the controller's terminal:
// `nox node hand-over <node> <path>` on the control-node (ADR 0272). The controller asks that machine's
// engine on its own subject, a request on core NATS the engine answers once; the engine judges every
// value and records the hand-over, or refuses and records nothing. The engine holds the same two shapes
// in its own link code (mesh-host internal/link HandOverAsk, HandOverAnswer); a test on each side holds
// the field names.

// HandOverAsk is what the controller asks: the node it is for, the directory's absolute path as the engine states
// it, who asked in the controller's words, when the ask stops being good, and a nonce the engine takes once.
type HandOverAsk struct {
	Node    string    `json:"node"`
	Path    string    `json:"path"`
	By      string    `json:"by"`
	Expires time.Time `json:"expires"`
	Nonce   string    `json:"nonce"`
}

// SignedHandOver is the ask as it travels: its bytes exactly as signed, and the signature.
//
// **Signed, because the subject proves nothing** (review of issue 356). Only the controller may publish
// `mesh.node.<node>.ask.hand-over`, but the bus lets any principal allowed to answer reply to a message it
// received, on whatever reply subject that message named — so a tool server asked on its own subject with that
// reply could hand the engine an ask the controller never made. The engine verifies this signature, with the
// key it verifies declarations with, before it reads anything out of the ask.
type SignedHandOver struct {
	Ask       []byte `json:"ask"`
	Signature []byte `json:"signature"`
}

// HandOverContext is prefixed to an ask's bytes before signing, so a hand-over's signature is never a
// declaration's: the same key signs both, and a declaration is signed over its bytes alone.
const HandOverContext = "novox-mesh hand-over v1\n"

// HandOverGood is how long a signed ask is good for: the engine refuses one past it, and one further ahead.
const HandOverGood = time.Minute

// HandOverAnswer is the engine's answer: what it recorded, or why it refused.
type HandOverAnswer struct {
	Said    string `json:"said,omitempty"`
	Refused string `json:"refused,omitempty"`
}

// HandOverWithin is how long the controller waits for the engine's answer: a file write, on a machine that is
// up; a machine that is down is said as not answering.
const HandOverWithin = 30 * time.Second

// AskHandOver asks one machine's node-engine to hand a directory used as found to the mesh, and reads its
// answer. An error is the ask not reaching an engine, or an answer that is not one; a refusal is the engine's
// and comes back in the answer.
func AskHandOver(ctx context.Context, conn *nats.Conn, signer Signer, node, path, by string,
	timeout time.Duration) (HandOverAnswer, error) {
	if conn == nil {
		return HandOverAnswer{}, errors.New("this controller is not on the bus")
	}
	body, err := SignHandOver(ctx, signer, HandOverAsk{Node: node, Path: path, By: by})
	if err != nil {
		return HandOverAnswer{}, err
	}
	return askSigned(ctx, conn, broker.AskHandOverSubject(node), body, node, timeout, askWords{
		what: "a hand-over", nothing: "nothing was handed over", issue: "issue 356",
		unknown: "whether it was recorded is not known — the module's condition says whether the directory is " +
			"still used as found", record: "a record"})
}

// askWords are what a signed ask's failures say of it.
type askWords struct{ what, nothing, issue, unknown, record string }

// askSigned sends one signed ask on subject and reads the engine's answer.
func askSigned(ctx context.Context, conn *nats.Conn, subject string, body []byte, node string, timeout time.Duration,
	w askWords) (HandOverAnswer, error) {
	asking, cancel := context.WithTimeout(ctx, timeout)
	defer cancel()
	refused, stop := refusalsOf(conn, subject)
	defer stop()
	type replied struct {
		msg *nats.Msg
		err error
	}
	done := make(chan replied, 1)
	go func() {
		msg, err := conn.RequestWithContext(asking, subject, body)
		done <- replied{msg, err}
	}()
	var reply *nats.Msg
	var err error
	select {
	case r := <-done:
		reply, err = r.msg, r.err
	case why := <-refused:
		cancel()
		return HandOverAnswer{}, fmt.Errorf("the bus refused the controller asking %s for %s: %v", node, w.what, why)
	}
	switch {
	case errors.Is(err, nats.ErrNoResponders):
		return HandOverAnswer{}, fmt.Errorf("nothing on %s answers %s: its node-engine is not running, is not "+
			"on the bus, or is older than this ask (novox/hq %s); %s", node, w.what, w.issue, w.nothing)
	case errors.Is(err, context.DeadlineExceeded), errors.Is(err, nats.ErrTimeout):
		return HandOverAnswer{}, fmt.Errorf("%s did not answer %s within %s; %s", node, w.what, timeout, w.unknown)
	case err != nil:
		return HandOverAnswer{}, err
	}
	var answer HandOverAnswer
	if err := json.Unmarshal(reply.Data, &answer); err != nil {
		return HandOverAnswer{}, fmt.Errorf("%s answered %s with something unreadable: %w", node, w.what, err)
	}
	if answer.Said == "" && answer.Refused == "" {
		return HandOverAnswer{}, fmt.Errorf("%s answered %s with neither %s nor a refusal", node, w.what, w.record)
	}
	return answer, nil
}

// SetuidSearchContext is prefixed to a setuid search ask's bytes before signing (novox/hq issue 361): never a
// hand-over's signature, nor a declaration's. The engine holds the same words.
const SetuidSearchContext = "novox-mesh setuid-search v1\n"

// AskSetuidSearch asks one machine's node-engine to throw its last search for setuid programs away and start a
// full one (novox/hq issue 361): the ask a hand-over is, naming no path, signed under SetuidSearchContext.
func AskSetuidSearch(ctx context.Context, conn *nats.Conn, signer Signer, node, by string,
	timeout time.Duration) (HandOverAnswer, error) {
	if conn == nil {
		return HandOverAnswer{}, errors.New("this controller is not on the bus")
	}
	body, err := signAsk(ctx, signer, SetuidSearchContext, HandOverAsk{Node: node, By: by})
	if err != nil {
		return HandOverAnswer{}, err
	}
	return askSigned(ctx, conn, broker.AskSetuidSearchSubject(node), body, node, timeout, askWords{
		what: "a setuid search", nothing: "no search was started", issue: "issue 361",
		unknown: "whether it started is not known — the controller's root-free verb says whether a search runs " +
			"there", record: "a start"})
}

// SignHandOver fills the ask's expiry and nonce and signs it with the mesh's key, over HandOverContext and the
// ask's bytes exactly as they travel.
func SignHandOver(ctx context.Context, signer Signer, ask HandOverAsk) ([]byte, error) {
	return signAsk(ctx, signer, HandOverContext, ask)
}

// signAsk fills an ask's expiry and nonce and signs it over prefix (its signing context) and its bytes.
func signAsk(ctx context.Context, signer Signer, prefix string, ask HandOverAsk) ([]byte, error) {
	if signer == nil {
		return nil, errors.New("no signing key, so nothing can be asked of an engine")
	}
	nonce := make([]byte, 16)
	if _, err := rand.Read(nonce); err != nil {
		return nil, err
	}
	ask.Nonce = hex.EncodeToString(nonce)
	ask.Expires = time.Now().UTC().Add(HandOverGood)
	raw, err := json.Marshal(ask)
	if err != nil {
		return nil, err
	}
	signature, err := signer.Sign(ctx, append([]byte(prefix), raw...))
	if err != nil {
		return nil, fmt.Errorf("cannot sign the ask: %w", err)
	}
	return json.Marshal(SignedHandOver{Ask: raw, Signature: signature})
}