fix(traffic): fix flow counter inflation from TOCTOU race in agent traffic reporter

Replace read-then-subtract pattern in collectAndReport with atomic
swap-to-zero to eliminate race where AddTraffic increments counters
between snapshot and clearReportedTraffic, causing residual traffic
to accumulate indefinitely and inflate user flow counters.

Also add defensive check in processFlowItem to skip AddFlow when
forward no longer exists, and send DeleteService to clean up orphaned
agent services.
This commit is contained in:
sagitchu
2026-04-21 14:32:08 +08:00
parent eaf16bf17b
commit 30e1473f06
3 changed files with 47 additions and 70 deletions
@@ -43,16 +43,19 @@ func (h *Handler) processFlowItem(nodeID int64, item flowItem) {
forwardID, userID, userTunnelID, ok := parseFlowServiceIDs(serviceName)
if ok {
inFlow, outFlow := h.scaleFlowByTunnel(forwardID, item.D, item.U)
_ = h.repo.AddFlow(forwardID, userID, userTunnelID, inFlow, outFlow)
if quota, quotaErr := h.repo.AddUserQuotaUsage(userID, inFlow+outFlow, time.Now()); quotaErr == nil {
h.enforceUserQuotaIfNeeded(userID, quota)
if h.forwardExists(forwardID) {
inFlow, outFlow := h.scaleFlowByTunnel(forwardID, item.D, item.U)
_ = h.repo.AddFlow(forwardID, userID, userTunnelID, inFlow, outFlow)
if quota, quotaErr := h.repo.AddUserQuotaUsage(userID, inFlow+outFlow, time.Now()); quotaErr == nil {
h.enforceUserQuotaIfNeeded(userID, quota)
}
if userTunnelID > 0 {
h.enforceFlowPolicies(userID, userTunnelID)
}
} else if nodeID > 0 {
h.sendDeleteOrphanedForwardService(nodeID, serviceName)
}
h.processPeerShareFlowFromForward(forwardID, nodeID, serviceName, item)
if userTunnelID > 0 {
h.enforceFlowPolicies(userID, userTunnelID)
}
return
}
@@ -628,6 +631,21 @@ func (h *Handler) forwardExists(forwardID int64) bool {
return ok
}
func (h *Handler) sendDeleteOrphanedForwardService(nodeID int64, serviceName string) {
parts := strings.Split(serviceName, "_")
if len(parts) < 3 {
return
}
forwardID, err := strconv.ParseInt(parts[0], 10, 64)
if err != nil || forwardID <= 0 {
return
}
base := parts[0] + "_" + parts[1] + "_" + parts[2]
_, _ = h.sendNodeCommand(nodeID, "DeleteService", map[string]interface{}{
"services": []string{base + "_tcp", base + "_udp"},
}, false, true)
}
func (h *Handler) speedLimiterExists(name string) bool {
if name == "" {
return false
+20 -61
View File
@@ -88,56 +88,45 @@ func (m *GlobalTrafficManager) startReporting() {
// collectAndReport 收集所有服务流量并合并上报
func (m *GlobalTrafficManager) collectAndReport() {
m.mu.Lock()
// 如果没有流量,直接返回
if len(m.serviceTraffic) == 0 {
m.mu.Unlock()
return
}
// 复制当前所有流量数据(避免长时间持锁)
trafficSnapshot := make(map[string]*ServiceTraffic)
reportData := make(map[string]struct {
up int64
down int64
})
reportItems := make([]TrafficReportItem, 0, len(m.serviceTraffic))
for name, traffic := range m.serviceTraffic {
traffic.mu.Lock()
if traffic.UpBytes > 0 || traffic.DownBytes > 0 {
trafficSnapshot[name] = traffic
reportData[name] = struct {
up int64
down int64
}{
up: traffic.UpBytes,
down: traffic.DownBytes,
}
up := traffic.UpBytes
down := traffic.DownBytes
if up > 0 || down > 0 {
traffic.UpBytes = 0
traffic.DownBytes = 0
}
traffic.mu.Unlock()
if up > 0 || down > 0 {
reportItems = append(reportItems, TrafficReportItem{
N: name,
U: up,
D: down,
})
}
}
m.mu.Unlock()
// 如果没有需要上报的流量,返回
if len(reportData) == 0 {
if len(reportItems) == 0 {
return
}
// 构建上报数据数组(保持每个服务独立)
reportItems := make([]TrafficReportItem, 0, len(reportData))
var totalUp, totalDown int64
for serviceName, data := range reportData {
reportItems = append(reportItems, TrafficReportItem{
N: serviceName, // 保持服务名不变
U: data.up,
D: data.down,
})
totalUp += data.up
totalDown += data.down
for _, item := range reportItems {
totalUp += item.U
totalDown += item.D
}
// 批量发送上报请求(一次HTTP请求包含所有服务)
success, err := sendBatchTrafficReport(m.ctx, reportItems)
if err != nil {
fmt.Printf("❌ 全局流量上报失败: %v (总流量: ↑%d ↓%d, %d个服务)\n", err, totalUp, totalDown, len(reportItems))
@@ -146,36 +135,6 @@ func (m *GlobalTrafficManager) collectAndReport() {
if !success {
fmt.Printf("⚠️ 全局流量上报未成功 (总流量: ↑%d ↓%d, %d个服务)\n", totalUp, totalDown, len(reportItems))
return
}
// 上报成功,清空已上报的流量
m.clearReportedTraffic(reportData)
}
// clearReportedTraffic 清空已成功上报的流量
func (m *GlobalTrafficManager) clearReportedTraffic(reportedData map[string]struct {
up int64
down int64
}) {
m.mu.Lock()
defer m.mu.Unlock()
for serviceName, reported := range reportedData {
if traffic, exists := m.serviceTraffic[serviceName]; exists {
traffic.mu.Lock()
// 减去已上报的流量
traffic.UpBytes -= reported.up
traffic.DownBytes -= reported.down
// 如果流量归零,从map中删除该服务记录(避免内存泄漏)
if traffic.UpBytes <= 0 && traffic.DownBytes <= 0 {
traffic.mu.Unlock()
delete(m.serviceTraffic, serviceName)
} else {
traffic.mu.Unlock()
}
}
}
}
@@ -85,7 +85,7 @@ function DropdownMenuSubContent({
return (
<DropdownMenuPrimitive.SubContent
className={cn(
"z-50 min-w-32 overflow-y-auto rounded-md border border-default-200 bg-white p-1.5 text-foreground shadow-lg max-h-[--radix-dropdown-menu-content-available-height] [&::-webkit-scrollbar]:hidden [-ms-overflow-style:none] [scrollbar-width:none] data-[state=open]:animate-in data-[state=closed]:animate-out data-[state=closed]:fade-out-0 data-[state=open]:fade-in-0 data-[state=closed]:zoom-out-95 data-[state=open]:zoom-in-95 data-[side=top]:slide-in-from-bottom-2 data-[side=bottom]:slide-in-from-top-2 dark:bg-default-50",
"z-50 min-w-32 overflow-y-auto rounded-md border border-default-200 bg-white p-1.5 text-foreground shadow-lg max-h-[--radix-dropdown-menu-content-available-height] [&::-webkit-scrollbar]:hidden [-ms-overflow-style:none] [scrollbar-width:none] data-[state=open]:animate-in data-[state=closed]:animate-out data-[state=closed]:fade-out-0 data-[state=open]:fade-in-0 data-[state=closed]:zoom-out-95 data-[state=open]:zoom-in-95 data-[side=top]:slide-in-from-bottom-2 data-[side=bottom]:slide-in-from-top-2 dark:bg-default-50",
className,
)}
data-slot="dropdown-menu-sub-content"