529 lines
16 KiB
Go
529 lines
16 KiB
Go
package services
|
||
|
||
import (
|
||
"context"
|
||
"crypto"
|
||
"crypto/ecdsa"
|
||
"crypto/elliptic"
|
||
"crypto/rand"
|
||
"crypto/x509"
|
||
"crypto/x509/pkix"
|
||
"encoding/pem"
|
||
"fmt"
|
||
"os"
|
||
"path/filepath"
|
||
"regexp"
|
||
"strings"
|
||
"sync"
|
||
"time"
|
||
|
||
"server/models"
|
||
|
||
"github.com/beego/beego/v2/core/logs"
|
||
beego "github.com/beego/beego/v2/server/web"
|
||
"golang.org/x/crypto/acme"
|
||
)
|
||
|
||
const (
|
||
acmeAccountDir = "_account"
|
||
acmeAccountKeyFile = "key.pem"
|
||
acmeChallengeDir = "_challenges"
|
||
certFileName = "fullchain.pem"
|
||
keyFileName = "privkey.pem"
|
||
|
||
// 签发失败后的冷却时间。Let's Encrypt 对同一账号的验证失败次数有频控
|
||
// (每小时 5 次),冷却可避免租户反复点击把额度打满。
|
||
issueFailureCooldown = 30 * time.Minute
|
||
|
||
// 单次签发的整体超时。HTTP-01 需要等 CA 回访,给足时间但不能无限等。
|
||
issueTimeout = 3 * time.Minute
|
||
|
||
// 证书到期前多少天开始续期
|
||
renewBeforeDays = 30
|
||
)
|
||
|
||
var (
|
||
challengeMu sync.RWMutex
|
||
challengeStore = make(map[string]string) // token -> keyAuthorization
|
||
|
||
issueMu sync.Mutex
|
||
issuing = make(map[string]bool) // host -> 是否正在签发
|
||
lastFailureAt = make(map[string]time.Time) // host -> 最近一次签发失败时间
|
||
)
|
||
|
||
// ===== challenge 存取(供 /.well-known/acme-challenge/:token 路由使用)=====
|
||
|
||
// challengeTokenRe ACME token 的合法字符集(base64url)。
|
||
// token 会被拼进文件名,必须先校验,避免路径穿越。
|
||
var challengeTokenRe = regexp.MustCompile(`^[A-Za-z0-9_-]{16,128}$`)
|
||
|
||
func challengeFilePath(token string) string {
|
||
return filepath.Join(SSLCertDir(), acmeChallengeDir, token)
|
||
}
|
||
|
||
// GetHTTP01Challenge 返回 token 对应的 keyAuthorization。
|
||
//
|
||
// 先查内存,再查磁盘。磁盘这层是为拆进程部署准备的:签发是由 backend/platform
|
||
// 接口触发的,而 CA 的回访只会落到 Nginx 扑底站点指向的那个进程,两者可能不是
|
||
// 同一个进程,只靠内存会验证失败。
|
||
func GetHTTP01Challenge(token string) (string, bool) {
|
||
challengeMu.RLock()
|
||
v, ok := challengeStore[token]
|
||
challengeMu.RUnlock()
|
||
if ok {
|
||
return v, true
|
||
}
|
||
|
||
if !challengeTokenRe.MatchString(token) {
|
||
return "", false
|
||
}
|
||
raw, err := os.ReadFile(challengeFilePath(token))
|
||
if err != nil {
|
||
return "", false
|
||
}
|
||
return strings.TrimSpace(string(raw)), true
|
||
}
|
||
|
||
func putHTTP01Challenge(token, keyAuth string) {
|
||
challengeMu.Lock()
|
||
challengeStore[token] = keyAuth
|
||
challengeMu.Unlock()
|
||
|
||
if !challengeTokenRe.MatchString(token) {
|
||
return
|
||
}
|
||
dir := filepath.Join(SSLCertDir(), acmeChallengeDir)
|
||
if err := os.MkdirAll(dir, 0o755); err != nil {
|
||
logs.Warn("[acme] 创建 challenge 目录失败(将只依赖内存): %v", err)
|
||
return
|
||
}
|
||
if err := os.WriteFile(challengeFilePath(token), []byte(keyAuth), 0o644); err != nil {
|
||
logs.Warn("[acme] 写入 challenge 文件失败(将只依赖内存): %v", err)
|
||
}
|
||
}
|
||
|
||
func deleteHTTP01Challenge(token string) {
|
||
challengeMu.Lock()
|
||
delete(challengeStore, token)
|
||
challengeMu.Unlock()
|
||
|
||
if challengeTokenRe.MatchString(token) {
|
||
_ = os.Remove(challengeFilePath(token))
|
||
}
|
||
}
|
||
|
||
// ===== 配置 =====
|
||
|
||
// SSLCertDir 证书落盘根目录
|
||
func SSLCertDir() string {
|
||
dir, _ := beego.AppConfig.String("ssl_cert_dir")
|
||
dir = strings.TrimSpace(dir)
|
||
if dir == "" {
|
||
dir = "/www/wwwroot/ssl-certs"
|
||
}
|
||
return dir
|
||
}
|
||
|
||
func acmeDirectoryURL() string {
|
||
u, _ := beego.AppConfig.String("acme_directory")
|
||
return strings.TrimSpace(u)
|
||
}
|
||
|
||
func acmeContactEmail() string {
|
||
e, _ := beego.AppConfig.String("acme_email")
|
||
return strings.TrimSpace(e)
|
||
}
|
||
|
||
// CertPaths 返回某域名的证书与私钥落盘路径。
|
||
// host 会先做归一化:域名来自数据库,归一化可避免脏数据里的 ../ 之类写法拼出目录外的路径。
|
||
func CertPaths(host string) (certPath, keyPath string) {
|
||
h := NormalizeHost(host)
|
||
if h == "" {
|
||
h = "_invalid"
|
||
}
|
||
base := filepath.Join(SSLCertDir(), h)
|
||
return filepath.Join(base, certFileName), filepath.Join(base, keyFileName)
|
||
}
|
||
|
||
// ===== ACME 账号 =====
|
||
|
||
// loadOrCreateAccountKey 读取(或首次生成)ACME 账号私钥。
|
||
// 账号私钥必须持久化:换了 key 等于换账号,历史签发记录和频控额度都会重置。
|
||
func loadOrCreateAccountKey() (crypto.Signer, error) {
|
||
dir := filepath.Join(SSLCertDir(), acmeAccountDir)
|
||
path := filepath.Join(dir, acmeAccountKeyFile)
|
||
|
||
if raw, err := os.ReadFile(path); err == nil {
|
||
block, _ := pem.Decode(raw)
|
||
if block == nil {
|
||
return nil, fmt.Errorf("账号私钥格式错误: %s", path)
|
||
}
|
||
key, err := x509.ParseECPrivateKey(block.Bytes)
|
||
if err != nil {
|
||
return nil, fmt.Errorf("解析账号私钥失败: %v", err)
|
||
}
|
||
return key, nil
|
||
}
|
||
|
||
if err := os.MkdirAll(dir, 0o700); err != nil {
|
||
return nil, fmt.Errorf("创建账号目录失败: %v", err)
|
||
}
|
||
key, err := ecdsa.GenerateKey(elliptic.P256(), rand.Reader)
|
||
if err != nil {
|
||
return nil, fmt.Errorf("生成账号私钥失败: %v", err)
|
||
}
|
||
der, err := x509.MarshalECPrivateKey(key)
|
||
if err != nil {
|
||
return nil, fmt.Errorf("序列化账号私钥失败: %v", err)
|
||
}
|
||
pemBytes := pem.EncodeToMemory(&pem.Block{Type: "EC PRIVATE KEY", Bytes: der})
|
||
if err := os.WriteFile(path, pemBytes, 0o600); err != nil {
|
||
return nil, fmt.Errorf("写入账号私钥失败: %v", err)
|
||
}
|
||
return key, nil
|
||
}
|
||
|
||
// newACMEClient 构造 ACME 客户端并确保账号已注册
|
||
func newACMEClient(ctx context.Context) (*acme.Client, error) {
|
||
key, err := loadOrCreateAccountKey()
|
||
if err != nil {
|
||
return nil, err
|
||
}
|
||
client := &acme.Client{
|
||
Key: key,
|
||
DirectoryURL: acmeDirectoryURL(), // 留空则 acme 包默认使用 Let's Encrypt 正式环境
|
||
}
|
||
|
||
acct := &acme.Account{}
|
||
if email := acmeContactEmail(); email != "" {
|
||
acct.Contact = []string{"mailto:" + email}
|
||
}
|
||
// 已注册过会返回 ErrAccountAlreadyExists,属于正常情况
|
||
if _, err := client.Register(ctx, acct, acme.AcceptTOS); err != nil && err != acme.ErrAccountAlreadyExists {
|
||
return nil, fmt.Errorf("注册 ACME 账号失败: %v", err)
|
||
}
|
||
return client, nil
|
||
}
|
||
|
||
// ===== 证书签发 =====
|
||
|
||
// obtainCertificate 走完整的 HTTP-01 流程拿到证书并落盘,返回证书到期时间。
|
||
func obtainCertificate(ctx context.Context, host string) (time.Time, error) {
|
||
client, err := newACMEClient(ctx)
|
||
if err != nil {
|
||
return time.Time{}, err
|
||
}
|
||
|
||
order, err := client.AuthorizeOrder(ctx, acme.DomainIDs(host))
|
||
if err != nil {
|
||
return time.Time{}, fmt.Errorf("创建证书订单失败: %v", err)
|
||
}
|
||
|
||
// 逐个完成 pending 的授权
|
||
for _, authzURL := range order.AuthzURLs {
|
||
authz, err := client.GetAuthorization(ctx, authzURL)
|
||
if err != nil {
|
||
return time.Time{}, fmt.Errorf("获取授权信息失败: %v", err)
|
||
}
|
||
if authz.Status != acme.StatusPending {
|
||
continue
|
||
}
|
||
|
||
var chal *acme.Challenge
|
||
for _, c := range authz.Challenges {
|
||
if c.Type == "http-01" {
|
||
chal = c
|
||
break
|
||
}
|
||
}
|
||
if chal == nil {
|
||
return time.Time{}, fmt.Errorf("证书颁发机构未提供 http-01 验证方式")
|
||
}
|
||
|
||
keyAuth, err := client.HTTP01ChallengeResponse(chal.Token)
|
||
if err != nil {
|
||
return time.Time{}, fmt.Errorf("生成验证响应失败: %v", err)
|
||
}
|
||
putHTTP01Challenge(chal.Token, keyAuth)
|
||
|
||
if _, err := client.Accept(ctx, chal); err != nil {
|
||
deleteHTTP01Challenge(chal.Token)
|
||
return time.Time{}, fmt.Errorf("提交验证失败: %v", err)
|
||
}
|
||
_, err = client.WaitAuthorization(ctx, authz.URI)
|
||
deleteHTTP01Challenge(chal.Token)
|
||
if err != nil {
|
||
return time.Time{}, fmt.Errorf("域名验证未通过(请确认 80 端口可从公网访问且未强制跳转 HTTPS): %v", err)
|
||
}
|
||
}
|
||
|
||
if _, err := client.WaitOrder(ctx, order.URI); err != nil {
|
||
return time.Time{}, fmt.Errorf("订单未就绪: %v", err)
|
||
}
|
||
|
||
// 生成证书私钥与 CSR
|
||
certKey, err := ecdsa.GenerateKey(elliptic.P256(), rand.Reader)
|
||
if err != nil {
|
||
return time.Time{}, fmt.Errorf("生成证书私钥失败: %v", err)
|
||
}
|
||
csrDER, err := x509.CreateCertificateRequest(rand.Reader, &x509.CertificateRequest{
|
||
Subject: pkix.Name{CommonName: host},
|
||
DNSNames: []string{host},
|
||
}, certKey)
|
||
if err != nil {
|
||
return time.Time{}, fmt.Errorf("生成 CSR 失败: %v", err)
|
||
}
|
||
|
||
chainDER, _, err := client.CreateOrderCert(ctx, order.FinalizeURL, csrDER, true)
|
||
if err != nil {
|
||
return time.Time{}, fmt.Errorf("签发证书失败: %v", err)
|
||
}
|
||
if len(chainDER) == 0 {
|
||
return time.Time{}, fmt.Errorf("证书颁发机构返回了空证书链")
|
||
}
|
||
|
||
leaf, err := x509.ParseCertificate(chainDER[0])
|
||
if err != nil {
|
||
return time.Time{}, fmt.Errorf("解析证书失败: %v", err)
|
||
}
|
||
|
||
if err := writeCertFiles(host, chainDER, certKey); err != nil {
|
||
return time.Time{}, err
|
||
}
|
||
return leaf.NotAfter, nil
|
||
}
|
||
|
||
// writeCertFiles 把证书链与私钥写入 {ssl_cert_dir}/{host}/。
|
||
// 先写临时文件再 rename,避免 nginx 读到写了一半的文件。
|
||
func writeCertFiles(host string, chainDER [][]byte, key *ecdsa.PrivateKey) error {
|
||
certPath, keyPath := CertPaths(host)
|
||
dir := filepath.Dir(certPath)
|
||
if err := os.MkdirAll(dir, 0o755); err != nil {
|
||
return fmt.Errorf("创建证书目录失败: %v", err)
|
||
}
|
||
|
||
var chainPEM strings.Builder
|
||
for _, der := range chainDER {
|
||
if err := pem.Encode(&chainPEM, &pem.Block{Type: "CERTIFICATE", Bytes: der}); err != nil {
|
||
return fmt.Errorf("编码证书失败: %v", err)
|
||
}
|
||
}
|
||
if err := writeFileAtomic(certPath, []byte(chainPEM.String()), 0o644); err != nil {
|
||
return fmt.Errorf("写入证书失败: %v", err)
|
||
}
|
||
|
||
keyDER, err := x509.MarshalECPrivateKey(key)
|
||
if err != nil {
|
||
return fmt.Errorf("序列化证书私钥失败: %v", err)
|
||
}
|
||
keyPEM := pem.EncodeToMemory(&pem.Block{Type: "EC PRIVATE KEY", Bytes: keyDER})
|
||
if err := writeFileAtomic(keyPath, keyPEM, 0o600); err != nil {
|
||
return fmt.Errorf("写入证书私钥失败: %v", err)
|
||
}
|
||
return nil
|
||
}
|
||
|
||
func writeFileAtomic(path string, data []byte, perm os.FileMode) error {
|
||
tmp := path + ".tmp"
|
||
if err := os.WriteFile(tmp, data, perm); err != nil {
|
||
return err
|
||
}
|
||
// Linux 上 rename 覆盖同名文件是原子的,nginx 不会读到半个文件。
|
||
// Windows 上 rename 到已存在文件会失败,才退回「先删再改名」。
|
||
if err := os.Rename(tmp, path); err != nil {
|
||
_ = os.Remove(path)
|
||
if err2 := os.Rename(tmp, path); err2 != nil {
|
||
_ = os.Remove(tmp)
|
||
return err2
|
||
}
|
||
}
|
||
return nil
|
||
}
|
||
|
||
// ===== 与域名记录联动 =====
|
||
|
||
// tryLockIssue 抢占某域名的签发权。返回 false 表示正在签发或处于失败冷却期。
|
||
func tryLockIssue(host string) (bool, string) {
|
||
issueMu.Lock()
|
||
defer issueMu.Unlock()
|
||
|
||
if issuing[host] {
|
||
return false, "该域名正在签发证书,请稍后查看结果"
|
||
}
|
||
if at, ok := lastFailureAt[host]; ok {
|
||
if wait := issueFailureCooldown - time.Since(at); wait > 0 {
|
||
return false, fmt.Sprintf("上次签发失败,请在 %d 分钟后重试", int(wait.Minutes())+1)
|
||
}
|
||
}
|
||
issuing[host] = true
|
||
return true, ""
|
||
}
|
||
|
||
func unlockIssue(host string, failed bool) {
|
||
issueMu.Lock()
|
||
defer issueMu.Unlock()
|
||
delete(issuing, host)
|
||
if failed {
|
||
lastFailureAt[host] = time.Now()
|
||
} else {
|
||
delete(lastFailureAt, host)
|
||
}
|
||
}
|
||
|
||
// IssueCertificateAsync 异步为域名记录签发证书。
|
||
// 立即返回;成功后把记录置为 ssl_status=2 且 status=1(已生效)。
|
||
// 返回 false 表示当前不可发起(正在签发或冷却中),第二个返回值为原因。
|
||
func IssueCertificateAsync(domainID uint64, host string) (bool, string) {
|
||
host = NormalizeHost(host)
|
||
if host == "" {
|
||
return false, "域名格式不正确"
|
||
}
|
||
ok, reason := tryLockIssue(host)
|
||
if !ok {
|
||
return false, reason
|
||
}
|
||
|
||
updateSSLState(domainID, models.SSLStatusIssuing, nil, nil)
|
||
|
||
go func() {
|
||
ctx, cancel := context.WithTimeout(context.Background(), issueTimeout)
|
||
defer cancel()
|
||
|
||
expireAt, err := obtainCertificate(ctx, host)
|
||
if err != nil {
|
||
logs.Error("[acme] 签发证书失败 host=%s: %v", host, err)
|
||
msg := err.Error()
|
||
updateSSLState(domainID, models.SSLStatusFailed, nil, &msg)
|
||
unlockIssue(host, true)
|
||
return
|
||
}
|
||
|
||
logs.Info("[acme] 签发证书成功 host=%s 到期=%s", host, expireAt.Format("2006-01-02"))
|
||
updateSSLState(domainID, models.SSLStatusIssued, &expireAt, nil)
|
||
// 证书就绪,域名正式生效。
|
||
// 只把「待生效(0)」推到「已生效(1)」:被平台禁用(2)的记录不能靠续期自己复活。
|
||
now := time.Now()
|
||
if _, err := models.Orm.QueryTable(new(models.SystemTenantDomain)).
|
||
Filter("id", domainID).
|
||
Filter("status", 0).
|
||
Filter("delete_time__isnull", true).
|
||
Update(map[string]interface{}{"status": 1, "update_time": now}); err != nil {
|
||
logs.Error("[acme] 更新域名生效状态失败 id=%d: %v", domainID, err)
|
||
}
|
||
unlockIssue(host, false)
|
||
}()
|
||
|
||
return true, ""
|
||
}
|
||
|
||
// updateSSLState 更新证书状态字段。
|
||
// 注意:beego ORM 的 Update 会对参数做 reflect 解引用,直接塞 nil 的 *string 会 panic,
|
||
// 因此这里统一转成空字符串 / 显式值。
|
||
func updateSSLState(domainID uint64, sslStatus int, expireAt *time.Time, errMsg *string) {
|
||
msg := ""
|
||
if errMsg != nil {
|
||
msg = *errMsg
|
||
// ssl_error 字段 varchar(500),超长会被 MySQL 严格模式拒绝
|
||
if len(msg) > 480 {
|
||
msg = msg[:480]
|
||
}
|
||
}
|
||
values := map[string]interface{}{
|
||
"ssl_status": sslStatus,
|
||
"ssl_error": msg,
|
||
"update_time": time.Now(),
|
||
}
|
||
if expireAt != nil {
|
||
values["ssl_expire_at"] = *expireAt
|
||
}
|
||
if _, err := models.Orm.QueryTable(new(models.SystemTenantDomain)).
|
||
Filter("id", domainID).
|
||
Filter("delete_time__isnull", true).
|
||
Update(values); err != nil {
|
||
logs.Error("[acme] 更新证书状态失败 id=%d: %v", domainID, err)
|
||
}
|
||
}
|
||
|
||
// ===== 续期调度 =====
|
||
|
||
// StartCertRenewScheduler 启动证书续期调度:启动 1 分钟后跑一次,之后每 12 小时扫一次。
|
||
func StartCertRenewScheduler(stopChan chan struct{}) {
|
||
go func() {
|
||
// 启动稍等,避开数据库初始化与首屏请求高峰
|
||
firstRun := time.NewTimer(1 * time.Minute)
|
||
defer firstRun.Stop()
|
||
select {
|
||
case <-firstRun.C:
|
||
renewExpiringCertificates()
|
||
case <-stopChan:
|
||
return
|
||
}
|
||
|
||
ticker := time.NewTicker(12 * time.Hour)
|
||
defer ticker.Stop()
|
||
for {
|
||
select {
|
||
case <-ticker.C:
|
||
renewExpiringCertificates()
|
||
case <-stopChan:
|
||
return
|
||
}
|
||
}
|
||
}()
|
||
}
|
||
|
||
// renewExpiringCertificates 扫描自有域名,为「即将到期」「证书文件缺失」「之前失败」的记录重新签发。
|
||
// 已生效域名续期失败不会下线站点,只写 ssl_error 等下一轮重试。
|
||
func renewExpiringCertificates() {
|
||
var rows []models.SystemTenantDomain
|
||
_, err := models.Orm.QueryTable(new(models.SystemTenantDomain)).
|
||
Filter("domain_type", models.DomainTypeCustom).
|
||
Filter("delete_time__isnull", true).
|
||
Exclude("status", 2). // 已禁用的不续
|
||
All(&rows)
|
||
if err != nil {
|
||
logs.Error("[acme] 扫描待续期域名失败: %v", err)
|
||
return
|
||
}
|
||
|
||
deadline := time.Now().AddDate(0, 0, renewBeforeDays)
|
||
for i := range rows {
|
||
row := rows[i]
|
||
if row.FullDomain == nil || *row.FullDomain == "" {
|
||
continue
|
||
}
|
||
// DNS 还没通过的不去打 CA,避免浪费频控额度
|
||
if row.VerifyStatus != models.VerifyStatusPassed {
|
||
continue
|
||
}
|
||
if !needsRenew(row, deadline) {
|
||
continue
|
||
}
|
||
if ok, reason := IssueCertificateAsync(row.ID, *row.FullDomain); !ok {
|
||
logs.Info("[acme] 跳过续期 host=%s: %s", *row.FullDomain, reason)
|
||
}
|
||
}
|
||
}
|
||
|
||
// needsRenew 判断一条记录是否需要(重新)签发
|
||
func needsRenew(row models.SystemTenantDomain, deadline time.Time) bool {
|
||
if row.SSLStatus == models.SSLStatusIssuing {
|
||
return false
|
||
}
|
||
if row.SSLStatus != models.SSLStatusIssued {
|
||
// 待处理或曾失败:重试(失败冷却由 tryLockIssue 兜住)
|
||
return true
|
||
}
|
||
// 证书文件被误删也要补签
|
||
certPath, keyPath := CertPaths(*row.FullDomain)
|
||
if !fileExists(certPath) || !fileExists(keyPath) {
|
||
return true
|
||
}
|
||
return row.SSLExpireAt == nil || row.SSLExpireAt.Before(deadline)
|
||
}
|
||
|
||
func fileExists(path string) bool {
|
||
info, err := os.Stat(path)
|
||
return err == nil && !info.IsDir()
|
||
}
|