feat(ntp): add delay between sync attempts and support NTP servers from DHCP (#162)

* feat(ntp): use ntp server from dhcp info

* feat(ntp): use ntp server from dhcp info

* feat(ntp): add delay between time sync attempts

* chore(ntp): more logging

go.mod

@@ -14,6 +14,7 @@ require (
 	github.com/google/uuid v1.6.0
 	github.com/gwatts/rootcerts v0.0.0-20240401182218-3ab9db955caf
 	github.com/hanwen/go-fuse/v2 v2.5.1
+	github.com/hashicorp/go-envparse v0.1.0
 	github.com/openstadia/go-usb-gadget v0.0.0-20231115171102-aebd56bbb965
 	github.com/pion/logging v0.2.2
 	github.com/pion/mdns/v2 v2.0.7

go.sum

@@ -49,6 +49,8 @@ github.com/gwatts/rootcerts v0.0.0-20240401182218-3ab9db955caf h1:JO6ISZIvEUitto
 github.com/gwatts/rootcerts v0.0.0-20240401182218-3ab9db955caf/go.mod h1:5Kt9XkWvkGi2OHOq0QsGxebHmhCcqJ8KCbNg/a6+n+g=
 github.com/hanwen/go-fuse/v2 v2.5.1 h1:OQBE8zVemSocRxA4OaFJbjJ5hlpCmIWbGr7r0M4uoQQ=
 github.com/hanwen/go-fuse/v2 v2.5.1/go.mod h1:xKwi1cF7nXAOBCXujD5ie0ZKsxc8GGSA1rlMJc+8IJs=
+github.com/hashicorp/go-envparse v0.1.0 h1:bE++6bhIsNCPLvgDZkYqo3nA+/PFI51pkrHdmPSDFPY=
+github.com/hashicorp/go-envparse v0.1.0/go.mod h1:OHheN1GoygLlAkTlXLXvAdnXdZxy8JUweQ1rAXx1xnc=
 github.com/json-iterator/go v1.1.12 h1:PV8peI4a0ysnczrg+LtxykD8LfKY9ML6u2jnxaEnrnM=
 github.com/json-iterator/go v1.1.12/go.mod h1:e30LSqwooZae/UwlEbR2852Gd8hjQvJoHmT4TnhNGBo=
 github.com/klauspost/cpuid/v2 v2.0.9/go.mod h1:FInQzS24/EEf25PyTYn52gqo7WaD8xa0213Md/qVLRg=

network.go

@@ -1,13 +1,19 @@
 package kvm
 
 import (
+	"bytes"
 	"fmt"
+	"net"
+	"os"
+	"strings"
+	"time"
+
+	"os/exec"
+
+	"github.com/hashicorp/go-envparse"
 	"github.com/pion/mdns/v2"
 	"golang.org/x/net/ipv4"
 	"golang.org/x/net/ipv6"
-	"net"
-	"os/exec"
-	"time"
 
 	"github.com/vishvananda/netlink"
 	"github.com/vishvananda/netlink/nl"
@@ -15,11 +21,15 @@ import (
 
 var mDNSConn *mdns.Conn
 
-var networkState struct {
+var networkState NetworkState
+
+type NetworkState struct {
 	Up   bool
 	IPv4 string
 	IPv6 string
 	MAC  string
+
+	checked bool
 }
 
 type LocalIpInfo struct {
@@ -28,43 +38,45 @@ type LocalIpInfo struct {
 	MAC  string
 }
 
+const (
+	NetIfName     = "eth0"
+	DHCPLeaseFile = "/run/udhcpc.%s.info"
+)
+
 // setDhcpClientState sends signals to udhcpc to change it's current mode
 // of operation. Setting active to true will force udhcpc to renew the DHCP lease.
 // Setting active to false will put udhcpc into idle mode.
 func setDhcpClientState(active bool) {
-	var signal string;
+	var signal string
 	if active {
 		signal = "-SIGUSR1"
 	} else {
 		signal = "-SIGUSR2"
 	}
 
-	cmd := exec.Command("/usr/bin/killall", signal, "udhcpc");
+	cmd := exec.Command("/usr/bin/killall", signal, "udhcpc")
 	if err := cmd.Run(); err != nil {
 		fmt.Printf("network: setDhcpClientState: failed to change udhcpc state: %s\n", err)
 	}
 }
 
 func checkNetworkState() {
-	iface, err := netlink.LinkByName("eth0")
+	iface, err := netlink.LinkByName(NetIfName)
 	if err != nil {
-		fmt.Printf("failed to get eth0 interface: %v\n", err)
+		fmt.Printf("failed to get [%s] interface: %v\n", NetIfName, err)
 		return
 	}
 
-	newState := struct {
-		Up   bool
-		IPv4 string
-		IPv6 string
-		MAC  string
-	}{
+	newState := NetworkState{
 		Up:  iface.Attrs().OperState == netlink.OperUp,
 		MAC: iface.Attrs().HardwareAddr.String(),
+
+		checked: true,
 	}
 
 	addrs, err := netlink.AddrList(iface, nl.FAMILY_ALL)
 	if err != nil {
-		fmt.Printf("failed to get addresses for eth0: %v\n", err)
+		fmt.Printf("failed to get addresses for [%s]: %v\n", NetIfName, err)
 	}
 
 	// If the link is going down, put udhcpc into idle mode.
@@ -144,6 +156,39 @@ func startMDNS() error {
 	return nil
 }
 
+func getNTPServersFromDHCPInfo() ([]string, error) {
+	buf, err := os.ReadFile(fmt.Sprintf(DHCPLeaseFile, NetIfName))
+	if err != nil {
+		// do not return error if file does not exist
+		if os.IsNotExist(err) {
+			return nil, nil
+		}
+		return nil, fmt.Errorf("failed to load udhcpc info: %w", err)
+	}
+
+	// parse udhcpc info
+	env, err := envparse.Parse(bytes.NewReader(buf))
+	if err != nil {
+		return nil, fmt.Errorf("failed to parse udhcpc info: %w", err)
+	}
+
+	val, ok := env["ntpsrv"]
+	if !ok {
+		return nil, nil
+	}
+
+	var servers []string
+
+	for _, server := range strings.Fields(val) {
+		if net.ParseIP(server) == nil {
+			fmt.Printf("invalid NTP server IP: %s, ignoring ... \n", server)
+		}
+		servers = append(servers, server)
+	}
+
+	return servers, nil
+}
+
 func init() {
 	updates := make(chan netlink.LinkUpdate)
 	done := make(chan struct{})
@@ -162,7 +207,7 @@ func init() {
 		for {
 			select {
 			case update := <-updates:
-				if update.Link.Attrs().Name == "eth0" {
+				if update.Link.Attrs().Name == NetIfName {
 					fmt.Printf("link update: %+v\n", update)
 					checkNetworkState()
 				}

ntp.go

@@ -11,20 +11,56 @@ import (
 	"github.com/beevik/ntp"
 )
 
-var timeSynced = false
+const (
+	timeSyncRetryStep     = 5 * time.Second
+	timeSyncRetryMaxInt   = 1 * time.Minute
+	timeSyncWaitNetChkInt = 100 * time.Millisecond
+	timeSyncWaitNetUpInt  = 3 * time.Second
+	timeSyncInterval      = 1 * time.Hour
+	timeSyncTimeout       = 2 * time.Second
+)
+
+var (
+	timeSynced            = false
+	timeSyncRetryInterval = 0 * time.Second
+	defaultNTPServers     = []string{
+		"time.cloudflare.com",
+		"time.apple.com",
+	}
+)
 
 func TimeSyncLoop() {
 	for {
-		fmt.Println("Syncing system time")
+		if !networkState.checked {
+			time.Sleep(timeSyncWaitNetChkInt)
+			continue
+		}
+
+		if !networkState.Up {
+			log.Printf("Waiting for network to come up")
+			time.Sleep(timeSyncWaitNetUpInt)
+			continue
+		}
+
+		log.Printf("Syncing system time")
 		start := time.Now()
 		err := SyncSystemTime()
 		if err != nil {
 			log.Printf("Failed to sync system time: %v", err)
+
+			// retry after a delay
+			timeSyncRetryInterval += timeSyncRetryStep
+			time.Sleep(timeSyncRetryInterval)
+			// reset the retry interval if it exceeds the max interval
+			if timeSyncRetryInterval > timeSyncRetryMaxInt {
+				timeSyncRetryInterval = 0
+			}
+
 			continue
 		}
 		log.Printf("Time sync successful, now is: %v, time taken: %v", time.Now(), time.Since(start))
 		timeSynced = true
-		time.Sleep(1 * time.Hour) //once the first sync is done, sync every hour
+		time.Sleep(timeSyncInterval) // after the first sync is done
 	}
 }
 
@@ -41,13 +77,22 @@ func SyncSystemTime() (err error) {
 }
 
 func queryNetworkTime() (*time.Time, error) {
-	ntpServers := []string{
-		"time.cloudflare.com",
-		"time.apple.com",
+	ntpServers, err := getNTPServersFromDHCPInfo()
+	if err != nil {
+		log.Printf("failed to get NTP servers from DHCP info: %v\n", err)
 	}
+
+	if ntpServers == nil {
+		ntpServers = defaultNTPServers
+		log.Printf("Using default NTP servers: %v\n", ntpServers)
+	} else {
+		log.Printf("Using NTP servers from DHCP: %v\n", ntpServers)
+	}
+
 	for _, server := range ntpServers {
-		now, err := queryNtpServer(server, 2*time.Second)
+		now, err := queryNtpServer(server, timeSyncTimeout)
 		if err == nil {
+			log.Printf("NTP server [%s] returned time: %v\n", server, now)
 			return now, nil
 		}
 	}
@@ -56,7 +101,7 @@ func queryNetworkTime() (*time.Time, error) {
 		"http://cloudflare.com",
 	}
 	for _, url := range httpUrls {
-		now, err := queryHttpTime(url, 2*time.Second)
+		now, err := queryHttpTime(url, timeSyncTimeout)
 		if err == nil {
 			return now, nil
 		}