Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -402,6 +402,10 @@ node bin/service.js call --data-json '{"id":"123"}'

The SDK also reads the same variable from `.env` in the current working directory. It reads only that key and does not inject other `.env` variables. This variable does not affect the daemon's `--runtime serve` or `--runtime invoke` protocol. When the daemon manages instances, it continues to pass config/secret through files and file descriptors.

## Secret storage

Instance secrets are encrypted at rest with AES-256-GCM. Production deployments should provide `OCTOBUS_SECRET_ENCRYPTION_KEY` through a KMS/Vault-managed environment or provide `OCTOBUS_SECRET_ENCRYPTION_KEY_FILE` pointing to a separately protected 32-byte key file. If neither is configured, OctoBus creates `<database>.secret-key` with mode `0600`; this local fallback protects the database from casual inspection but does not protect against an attacker who can copy the entire data directory. Losing or changing the key is intentionally treated as a startup error when encrypted secrets exist.

## Development

### Architecture
Expand Down
240 changes: 240 additions & 0 deletions internal/store/secret_crypto.go
Original file line number Diff line number Diff line change
@@ -0,0 +1,240 @@
package store

import (
"context"
"crypto/aes"
"crypto/cipher"
"crypto/rand"
"database/sql"
"encoding/base64"
"errors"
"fmt"
"io"
"os"
"path/filepath"
"strings"
"time"
)

const (
encryptedSecretPrefix = "octobus-secret-v1:"
secretKeyEnv = "OCTOBUS_SECRET_ENCRYPTION_KEY"
secretKeyFileEnv = "OCTOBUS_SECRET_ENCRYPTION_KEY_FILE"
secretKeyBytes = 32
)

func loadSecretKey(dbPath string, hasEncryptedSecrets func() (bool, error)) ([]byte, error) {
if encoded := os.Getenv(secretKeyEnv); encoded != "" {
key, err := base64.StdEncoding.DecodeString(encoded)
if err != nil || len(key) != secretKeyBytes {
return nil, fmt.Errorf("%s must be base64-encoded %d-byte key", secretKeyEnv, secretKeyBytes)
}
return key, nil
}
if dbPath == ":memory:" {
return randomSecretKey()
}

keyPath := os.Getenv(secretKeyFileEnv)

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

OCTOBUS_SECRET_ENCRYPTION_KEY_FILE 指向只读/非属主的外部密钥文件时,readSecretKeyFile 的 os.Chmod 会使启动失败

本次新增的 OCTOBUS_SECRET_ENCRYPTION_KEY_FILE(第 38-41 行)让密钥路径可以指向外部受管位置(README 也建议 KMS/Vault/Secret 管理的生产部署)。但读取逻辑 readSecretKeyFile 在成功读取并校验 32 字节长度后,无条件执行 os.Chmod(path, 0o600);任何 chmod 失败都会作为非 os.ErrNotExist 错误被 loadSecretKey 原样返回,导致启动失败。外部密钥文件的典型部署形态——K8s Secret 卷/投影、Vault Agent 模板文件、只读挂载、root 属主而进程以非 root 运行——会因 EROFS/EPERM 使 chmod 失败,即便文件存在、可读且恰好 32 字节。也就是说,该特性在最常见的安全部署方式下不可用。改动前 chmod 只作用于 daemon 自己创建、位于可写数据目录的默认 .secret-key,几乎不会失败;本次变更新增了指向外部文件的入口,把这一既有行为暴露为新的故障路径(启动即报错,特性形同虚设)。

Problem code:

Changed code at internal/store/secret_crypto.go:38

Recommendation:
对已存在且可读的外部密钥文件,应把 chmod 降级为尽力而为(失败仅记录而不阻断),权限收紧只对 daemon 自己创建的文件负责(writeSecretKeyAtomically 已保证 0600);或在文档中明确要求密钥文件必须可被进程 chmod 为 0600,并在失败时给出清晰可操作的错误提示。

Suggested diff:

diff --git a/internal/store/secret_crypto.go b/internal/store/secret_crypto.go
--- a/internal/store/secret_crypto.go
+++ b/internal/store/secret_crypto.go
@@ -87,7 +87,8 @@ func readSecretKeyFile(path string) ([]byte, error) {
 	if len(key) != secretKeyBytes {
 		return nil, fmt.Errorf("%w: %q", errInvalidSecretKeyLength, path)
 	}
-	if err := os.Chmod(path, 0o600); err != nil {
-		return nil, err
-	}
+	// 只读挂载/KMS 注入等外部密钥文件可能无法 chmod;文件已可读且长度合法即应接受。
+	// 本进程创建的密钥文件已由 writeSecretKeyAtomically 置为 0600。
+	_ = os.Chmod(path, 0o600)
 	return key, nil
 }

if keyPath == "" {
keyPath = dbPath + ".secret-key"
}

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

KEY_FILE 显式配置但目标文件缺失且库中尚无加密 secret 时,会静默生成新密钥并写入该路径,掩盖配置错误并可能导致后续数据不可解密

当 OCTOBUS_SECRET_ENCRYPTION_KEY_FILE 被显式配置但目标文件不存在、且数据库尚无任何加密 secret(或 hasEncryptedSecrets 返回 false)时,loadSecretKey 会落入 randomSecretKey + writeSecretKeyAtomically 分支,在该路径自动生成并写入一个全新的随机密钥(必要时还会用 MkdirAll 创建父目录)。这与 README 的语义不一致:README 将 KEY_FILE 描述为操作者“提供”的密钥文件,并称只有“两者都未配置”时 OctoBus 才自动创建 .secret-key。在显式配置外部密钥的生产部署中,文件缺失通常意味着路径拼写错误、Secret 尚未挂载或编排顺序问题;此时静默生成密钥会掩盖该错误——应用将使用本地随机密钥加密后续写入的 secrets,一旦操作者随后修复路径或挂载真正的受管密钥并重启,这些已加密数据会因密钥不一致而无法解密(本地随机密钥已被覆盖/丢弃),同时该随机密钥也被写入了操作者本不打算用作密钥存储的位置。

Problem code:

Changed code at internal/store/secret_crypto.go:41

Recommendation:
当 secretKeyFileEnv 被显式设置且其指向的文件不存在时,应直接返回清晰错误(提示检查路径/挂载/密钥文件是否已就绪),而不是自动生成并写入新密钥;仅在未配置 KEY_FILE、走默认 .secret-key 路径时保留现有的自动创建引导逻辑。

Suggested diff:

diff --git a/internal/store/secret_crypto.go b/internal/store/secret_crypto.go
--- a/internal/store/secret_crypto.go
+++ b/internal/store/secret_crypto.go
@@ -42,6 +42,9 @@ func loadSecretKey(dbPath string, hasEncryptedSecrets func() (bool, error)) ([]b
 	key, err := readSecretKeyFile(keyPath)
 	if err == nil {
 		return key, nil
 	}
 	if !errors.Is(err, os.ErrNotExist) {
 		return nil, err
 	}
+	if os.Getenv(secretKeyFileEnv) != "" {
+		return nil, fmt.Errorf("secret key file %q from %s does not exist", keyPath, secretKeyFileEnv)
+	}
 	if hasEncryptedSecrets != nil {

key, err := readSecretKeyFile(keyPath)
if err == nil {
return key, nil
}
if !errors.Is(err, os.ErrNotExist) {
return nil, err
}
if hasEncryptedSecrets != nil {
encrypted, checkErr := hasEncryptedSecrets()
if checkErr != nil {
return nil, checkErr
}
if encrypted {
return nil, fmt.Errorf("secret key file %q is missing while encrypted instance secrets exist", keyPath)
}
}
key, err = randomSecretKey()
Comment thread
monkeyscan[bot] marked this conversation as resolved.
if err != nil {
return nil, err
}
if err := writeSecretKeyAtomically(keyPath, key); err == nil {
return key, nil
Comment thread
monkeyscan[bot] marked this conversation as resolved.
} else if !errors.Is(err, os.ErrExist) {
return nil, err
}
for attempt := 0; attempt < 5; attempt++ {
key, err = readSecretKeyFile(keyPath)
if err == nil {
return key, nil
}
if !errors.Is(err, errInvalidSecretKeyLength) {
return nil, err
}
time.Sleep(10 * time.Millisecond)
}
return nil, fmt.Errorf("read secret key file %q: %w", keyPath, errInvalidSecretKeyLength)
}

var errInvalidSecretKeyLength = errors.New("secret key has invalid length")

func readSecretKeyFile(path string) ([]byte, error) {
key, err := os.ReadFile(path)
if err != nil {
return nil, err
}
if len(key) != secretKeyBytes {
return nil, fmt.Errorf("%w: %q", errInvalidSecretKeyLength, path)
}
if err := os.Chmod(path, 0o600); err != nil {
return nil, err
}
return key, nil
}

func writeSecretKeyAtomically(path string, key []byte) error {
if err := os.MkdirAll(filepath.Dir(path), 0o700); err != nil {
return err
}
tmp, err := os.CreateTemp(filepath.Dir(path), ".secret-key-")
if err != nil {
return err
}
tmpPath := tmp.Name()
defer os.Remove(tmpPath)
if err := tmp.Chmod(0o600); err != nil {
_ = tmp.Close()
return err
}
if _, err := tmp.Write(key); err != nil {
_ = tmp.Close()
return err
}
if err := tmp.Sync(); err != nil {
_ = tmp.Close()
return err
}
if err := tmp.Close(); err != nil {
return err
}
if err := os.Link(tmpPath, path); err != nil {
return err
}
return nil
}

func (s *Store) validateSecretKey(ctx context.Context) error {
var encoded string
err := s.db.QueryRowContext(ctx, `SELECT secret_json FROM instances WHERE substr(secret_json, 1, ?) = ? LIMIT 1`, len(encryptedSecretPrefix), encryptedSecretPrefix).Scan(&encoded)
if errors.Is(err, sql.ErrNoRows) || strings.Contains(errString(err), "no such table") {
return nil
}
if err != nil {
return err
}
if _, err := decryptSecret(s.secretKey, encoded); err != nil {
return fmt.Errorf("secret encryption key does not decrypt stored instance secrets: %w", err)
}
return nil
}

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

validateSecretKey 仅校验单条加密记录,密钥不匹配/混合密钥时校验结果不确定且可能漏检

本次变更新增 validateSecretKey 的目标是「打开即校验,避免密钥与库内密文不匹配时静默解密失败」。但实现使用 SELECT secret_json FROM instances WHERE substr(secret_json,1,?)=? LIMIT 1 只取一条带前缀密文做解密验证:1) 当库内混合了由不同密钥加密的历史记录时(密钥文件曾被替换/切换过 OCTOBUS_SECRET_ENCRYPTION_KEY 后又切回,旧行未用新密钥重加密),若被选中的第一行恰巧能用当前密钥解密,校验通过、Open 成功,但其余旧密钥行在后续 GetInstance/List 读取时仍会以 gcm 校验失败告终——即本变更试图消除的静默失败路径依然存在;2) 查询没有 ORDER BY,命中哪一行由 SQLite 内部顺序决定,因此同一数据库的校验结果是不确定的:某些行序下能正常打开(漏检),换一种数据分布/插入顺序则直接启动失败(单行损坏也会拖垮整个 Open)。这是对历史 finding「密钥不匹配导致密文不可解密」的修复不完整。

Problem code:

Changed code at internal/store/secret_crypto.go:123-136

Recommendation:
改为遍历所有带前缀的实例密文逐一尝试解密(按 id 取全部匹配行),任一行解密失败即返回明确错误并在错误中附带 instance id,既避免漏检也避免仅凭单行做全局判定。示例:SELECT id, secret_json FROM instances WHERE substr(secret_json,1,?)=? 循环 Scan 后对每行调用 decryptSecret,失败时返回 fmt.Errorf("secret encryption key does not decrypt stored instance secret for instance %q: %w", id, err)。

Suggested diff:

- err := s.db.QueryRowContext(ctx, `SELECT secret_json FROM instances WHERE substr(secret_json, 1, ?) = ? LIMIT 1`, len(encryptedSecretPrefix), encryptedSecretPrefix).Scan(&encoded)
- if errors.Is(err, sql.ErrNoRows) || strings.Contains(errString(err), "no such table") {
- 	return nil
- }
- if err != nil {
- 	return err
- }
- if _, err := decryptSecret(s.secretKey, encoded); err != nil {
- 	return fmt.Errorf("secret encryption key does not decrypt stored instance secrets: %w", err)
- }
- return nil
+ rows, err := s.db.QueryContext(ctx, `SELECT id, secret_json FROM instances WHERE substr(secret_json, 1, ?) = ?`, len(encryptedSecretPrefix), encryptedSecretPrefix)
+ if err != nil {
+ 	if strings.Contains(errString(err), "no such table") {
+ 		return nil
+ 	}
+ 	return err
+ }
+ defer rows.Close()
+ for rows.Next() {
+ 	var id, encoded string
+ 	if err := rows.Scan(&id, &encoded); err != nil {
+ 		return err
+ 	}
+ 	if _, err := decryptSecret(s.secretKey, encoded); err != nil {
+ 		return fmt.Errorf("secret encryption key does not decrypt stored instance secret for instance %q: %w", id, err)
+ 	}
+ }
+ return rows.Err()


func errString(err error) string {
if err == nil {
return ""
}
return err.Error()
}

func randomSecretKey() ([]byte, error) {
key := make([]byte, secretKeyBytes)
if _, err := io.ReadFull(rand.Reader, key); err != nil {
return nil, err
}
return key, nil
}

func encryptSecret(key, plaintext []byte) (string, error) {
block, err := aes.NewCipher(key)
if err != nil {
return "", err
}
gcm, err := cipher.NewGCM(block)
if err != nil {
return "", err
}
nonce := make([]byte, gcm.NonceSize())
if _, err := io.ReadFull(rand.Reader, nonce); err != nil {
return "", err
}
ciphertext := gcm.Seal(nonce, nonce, plaintext, nil)
return encryptedSecretPrefix + base64.RawStdEncoding.EncodeToString(ciphertext), nil
}

func decryptSecret(key []byte, encoded string) ([]byte, error) {
if len(encoded) < len(encryptedSecretPrefix) || encoded[:len(encryptedSecretPrefix)] != encryptedSecretPrefix {
return []byte(encoded), nil
}
payload, err := base64.RawStdEncoding.DecodeString(encoded[len(encryptedSecretPrefix):])
if err != nil {
return nil, fmt.Errorf("decode encrypted instance secret: %w", err)
}
block, err := aes.NewCipher(key)
if err != nil {
return nil, err
}
gcm, err := cipher.NewGCM(block)
if err != nil {
return nil, err
}
if len(payload) < gcm.NonceSize() {
return nil, errors.New("encrypted instance secret is truncated")
}
return gcm.Open(nil, payload[:gcm.NonceSize()], payload[gcm.NonceSize():], nil)
}

func (s *Store) encryptLegacySecrets(ctx context.Context) error {
rows, err := s.db.QueryContext(ctx, `SELECT id, secret_json FROM instances WHERE secret_json <> '' AND substr(secret_json, 1, ?) <> ?`, len(encryptedSecretPrefix), encryptedSecretPrefix)
if err != nil {
return err
}
type legacySecret struct {
id string
data string
}
var legacy []legacySecret
for rows.Next() {
var item legacySecret
if err := rows.Scan(&item.id, &item.data); err != nil {
_ = rows.Close()
return err
}
legacy = append(legacy, item)
}
if err := rows.Err(); err != nil {
_ = rows.Close()
return err
}
if err := rows.Close(); err != nil {
return err
}
if len(legacy) == 0 {
return nil
}

tx, err := s.db.BeginTx(ctx, nil)
if err != nil {
return err
}
defer tx.Rollback()
for _, item := range legacy {
encrypted, err := encryptSecret(s.secretKey, []byte(item.data))
if err != nil {
return err
}
if _, err := tx.ExecContext(ctx, `UPDATE instances SET secret_json = ? WHERE id = ?`, encrypted, item.id); err != nil {
return err
}
}
return tx.Commit()
}
115 changes: 115 additions & 0 deletions internal/store/secret_crypto_test.go
Original file line number Diff line number Diff line change
@@ -0,0 +1,115 @@
package store

import (
"context"
"os"
"strings"
"testing"

"octobus/internal/domain"
)

func TestInstanceSecretsAreEncryptedAtRestAndReadableAcrossReopen(t *testing.T) {
dbPath := t.TempDir() + "/octobus.db"
ctx := context.Background()
st, err := Open(dbPath)
if err != nil {
t.Fatal(err)
}
if err := st.UpsertService(ctx, domain.Service{ID: "secret-service", Name: "Secret Service"}); err != nil {
t.Fatal(err)
}
plain := []byte(`{"apiToken":"do-not-store-plaintext"}`)
if err := st.UpsertInstance(ctx, domain.Instance{ID: "secret-instance", ServiceID: "secret-service", Name: "Instance", SecretJSON: plain}); err != nil {
t.Fatal(err)
}
var stored string
if err := st.DB().QueryRowContext(ctx, `SELECT secret_json FROM instances WHERE id = ?`, "secret-instance").Scan(&stored); err != nil {
t.Fatal(err)
}
if strings.Contains(stored, "do-not-store-plaintext") || !strings.HasPrefix(stored, encryptedSecretPrefix) {
t.Fatalf("secret at rest is not encrypted: %q", stored)
}
if err := st.Close(); err != nil {
t.Fatal(err)
}

st, err = Open(dbPath)
if err != nil {
t.Fatal(err)
}
defer st.Close()
got, err := st.GetInstance(ctx, "secret-instance")
if err != nil {
t.Fatal(err)
}
if string(got.SecretJSON) != string(plain) {
t.Fatalf("decrypted secret = %s", got.SecretJSON)
}
}

func TestOpenFailsWhenSecretKeyIsMissingForEncryptedData(t *testing.T) {
t.Setenv(secretKeyEnv, "")
dbPath := t.TempDir() + "/octobus.db"
ctx := context.Background()
st, err := Open(dbPath)
if err != nil {
t.Fatal(err)
}
if err := st.UpsertService(ctx, domain.Service{ID: "secret-service", Name: "Secret Service"}); err != nil {
t.Fatal(err)
}
if err := st.UpsertInstance(ctx, domain.Instance{ID: "secret-instance", ServiceID: "secret-service", Name: "Instance", SecretJSON: []byte(`{"token":"value"}`)}); err != nil {
t.Fatal(err)
}
if err := st.Close(); err != nil {
t.Fatal(err)
}
if err := os.Remove(dbPath + ".secret-key"); err != nil {
t.Fatal(err)
}
if _, err := Open(dbPath); err == nil || !strings.Contains(err.Error(), "encrypted instance secrets exist") {
t.Fatalf("missing secret key error = %v", err)
}
}

func TestLegacyInstanceSecretsAreEncryptedDuringMigration(t *testing.T) {
dbPath := t.TempDir() + "/octobus.db"
ctx := context.Background()
st, err := Open(dbPath)
if err != nil {
t.Fatal(err)
}
if err := st.UpsertService(ctx, domain.Service{ID: "legacy-service", Name: "Legacy Service"}); err != nil {
t.Fatal(err)
}
if err := st.UpsertInstance(ctx, domain.Instance{ID: "legacy-instance", ServiceID: "legacy-service", Name: "Instance", SecretJSON: []byte(`{}`)}); err != nil {
t.Fatal(err)
}
if _, err := st.DB().ExecContext(ctx, `UPDATE instances SET secret_json = ? WHERE id = ?`, `{"legacy":"secret"}`, "legacy-instance"); err != nil {
t.Fatal(err)
}
if err := st.Close(); err != nil {
t.Fatal(err)
}

st, err = Open(dbPath)
if err != nil {
t.Fatal(err)
}
defer st.Close()
var stored string
if err := st.DB().QueryRowContext(ctx, `SELECT secret_json FROM instances WHERE id = ?`, "legacy-instance").Scan(&stored); err != nil {
t.Fatal(err)
}
if strings.Contains(stored, "legacy") || !strings.HasPrefix(stored, encryptedSecretPrefix) {
t.Fatalf("legacy secret was not migrated: %q", stored)
}
got, err := st.GetInstance(ctx, "legacy-instance")
if err != nil {
t.Fatal(err)
}
if string(got.SecretJSON) != `{"legacy":"secret"}` {
t.Fatalf("migrated secret = %s", got.SecretJSON)
}
}
Loading
Loading