netbird/client/firewall/iptables/acl_linux.go
Viktor Liu ad9f044aad
[client] Add stateful userspace firewall and remove egress filters (#3093)
- Add stateful firewall functionality for UDP/TCP/ICMP in userspace firewalll
- Removes all egress drop rules/filters, still needs refactoring so we don't add output rules to any chains/filters.
- on Linux, if the OUTPUT policy is DROP  then we don't do anything about it (no extra allow rules). This is up to the user, if they don't want anything leaving their machine they'll have to manage these rules explicitly.
2024-12-23 18:22:17 +01:00

452 lines
12 KiB
Go

package iptables
import (
"fmt"
"net"
"strconv"
"github.com/coreos/go-iptables/iptables"
"github.com/google/uuid"
"github.com/nadoo/ipset"
log "github.com/sirupsen/logrus"
firewall "github.com/netbirdio/netbird/client/firewall/manager"
"github.com/netbirdio/netbird/client/internal/statemanager"
nbnet "github.com/netbirdio/netbird/util/net"
)
const (
tableName = "filter"
// rules chains contains the effective ACL rules
chainNameInputRules = "NETBIRD-ACL-INPUT"
chainNameOutputRules = "NETBIRD-ACL-OUTPUT"
)
type aclEntries map[string][][]string
type entry struct {
spec []string
position int
}
type aclManager struct {
iptablesClient *iptables.IPTables
wgIface iFaceMapper
routingFwChainName string
entries aclEntries
optionalEntries map[string][]entry
ipsetStore *ipsetStore
stateManager *statemanager.Manager
}
func newAclManager(iptablesClient *iptables.IPTables, wgIface iFaceMapper, routingFwChainName string) (*aclManager, error) {
m := &aclManager{
iptablesClient: iptablesClient,
wgIface: wgIface,
routingFwChainName: routingFwChainName,
entries: make(map[string][][]string),
optionalEntries: make(map[string][]entry),
ipsetStore: newIpsetStore(),
}
if err := ipset.Init(); err != nil {
return nil, fmt.Errorf("init ipset: %w", err)
}
return m, nil
}
func (m *aclManager) init(stateManager *statemanager.Manager) error {
m.stateManager = stateManager
m.seedInitialEntries()
m.seedInitialOptionalEntries()
if err := m.cleanChains(); err != nil {
return fmt.Errorf("clean chains: %w", err)
}
if err := m.createDefaultChains(); err != nil {
return fmt.Errorf("create default chains: %w", err)
}
m.updateState()
return nil
}
func (m *aclManager) AddPeerFiltering(
ip net.IP,
protocol firewall.Protocol,
sPort *firewall.Port,
dPort *firewall.Port,
direction firewall.RuleDirection,
action firewall.Action,
ipsetName string,
) ([]firewall.Rule, error) {
var dPortVal, sPortVal string
if dPort != nil && dPort.Values != nil {
// TODO: we support only one port per rule in current implementation of ACLs
dPortVal = strconv.Itoa(dPort.Values[0])
}
if sPort != nil && sPort.Values != nil {
sPortVal = strconv.Itoa(sPort.Values[0])
}
var chain string
if direction == firewall.RuleDirectionOUT {
chain = chainNameOutputRules
} else {
chain = chainNameInputRules
}
ipsetName = transformIPsetName(ipsetName, sPortVal, dPortVal)
specs := filterRuleSpecs(ip, string(protocol), sPortVal, dPortVal, direction, action, ipsetName)
if ipsetName != "" {
if ipList, ipsetExists := m.ipsetStore.ipset(ipsetName); ipsetExists {
if err := ipset.Add(ipsetName, ip.String()); err != nil {
return nil, fmt.Errorf("failed to add IP to ipset: %w", err)
}
// if ruleset already exists it means we already have the firewall rule
// so we need to update IPs in the ruleset and return new fw.Rule object for ACL manager.
ipList.addIP(ip.String())
return []firewall.Rule{&Rule{
ruleID: uuid.New().String(),
ipsetName: ipsetName,
ip: ip.String(),
chain: chain,
specs: specs,
}}, nil
}
if err := ipset.Flush(ipsetName); err != nil {
log.Errorf("flush ipset %s before use it: %s", ipsetName, err)
}
if err := ipset.Create(ipsetName); err != nil {
return nil, fmt.Errorf("failed to create ipset: %w", err)
}
if err := ipset.Add(ipsetName, ip.String()); err != nil {
return nil, fmt.Errorf("failed to add IP to ipset: %w", err)
}
ipList := newIpList(ip.String())
m.ipsetStore.addIpList(ipsetName, ipList)
}
ok, err := m.iptablesClient.Exists("filter", chain, specs...)
if err != nil {
return nil, fmt.Errorf("failed to check rule: %w", err)
}
if ok {
return nil, fmt.Errorf("rule already exists")
}
if err := m.iptablesClient.Append("filter", chain, specs...); err != nil {
return nil, err
}
rule := &Rule{
ruleID: uuid.New().String(),
specs: specs,
ipsetName: ipsetName,
ip: ip.String(),
chain: chain,
}
m.updateState()
return []firewall.Rule{rule}, nil
}
// DeletePeerRule from the firewall by rule definition
func (m *aclManager) DeletePeerRule(rule firewall.Rule) error {
r, ok := rule.(*Rule)
if !ok {
return fmt.Errorf("invalid rule type")
}
if ipsetList, ok := m.ipsetStore.ipset(r.ipsetName); ok {
// delete IP from ruleset IPs list and ipset
if _, ok := ipsetList.ips[r.ip]; ok {
if err := ipset.Del(r.ipsetName, r.ip); err != nil {
return fmt.Errorf("failed to delete ip from ipset: %w", err)
}
delete(ipsetList.ips, r.ip)
}
// if after delete, set still contains other IPs,
// no need to delete firewall rule and we should exit here
if len(ipsetList.ips) != 0 {
return nil
}
// we delete last IP from the set, that means we need to delete
// set itself and associated firewall rule too
m.ipsetStore.deleteIpset(r.ipsetName)
if err := ipset.Destroy(r.ipsetName); err != nil {
log.Errorf("delete empty ipset: %v", err)
}
}
if err := m.iptablesClient.Delete(tableName, r.chain, r.specs...); err != nil {
return fmt.Errorf("failed to delete rule: %s, %v: %w", r.chain, r.specs, err)
}
m.updateState()
return nil
}
func (m *aclManager) Reset() error {
if err := m.cleanChains(); err != nil {
return fmt.Errorf("clean chains: %w", err)
}
m.updateState()
return nil
}
// todo write less destructive cleanup mechanism
func (m *aclManager) cleanChains() error {
ok, err := m.iptablesClient.ChainExists(tableName, chainNameOutputRules)
if err != nil {
log.Debugf("failed to list chains: %s", err)
return err
}
if ok {
rules := m.entries["OUTPUT"]
for _, rule := range rules {
err := m.iptablesClient.DeleteIfExists(tableName, "OUTPUT", rule...)
if err != nil {
log.Errorf("failed to delete rule: %v, %s", rule, err)
}
}
err = m.iptablesClient.ClearAndDeleteChain(tableName, chainNameOutputRules)
if err != nil {
log.Debugf("failed to clear and delete %s chain: %s", chainNameOutputRules, err)
return err
}
}
ok, err = m.iptablesClient.ChainExists(tableName, chainNameInputRules)
if err != nil {
log.Debugf("failed to list chains: %s", err)
return err
}
if ok {
for _, rule := range m.entries["INPUT"] {
err := m.iptablesClient.DeleteIfExists(tableName, "INPUT", rule...)
if err != nil {
log.Errorf("failed to delete rule: %v, %s", rule, err)
}
}
for _, rule := range m.entries["FORWARD"] {
err := m.iptablesClient.DeleteIfExists(tableName, "FORWARD", rule...)
if err != nil {
log.Errorf("failed to delete rule: %v, %s", rule, err)
}
}
err = m.iptablesClient.ClearAndDeleteChain(tableName, chainNameInputRules)
if err != nil {
log.Debugf("failed to clear and delete %s chain: %s", chainNameInputRules, err)
return err
}
}
ok, err = m.iptablesClient.ChainExists("mangle", "PREROUTING")
if err != nil {
return fmt.Errorf("list chains: %w", err)
}
if ok {
for _, rule := range m.entries["PREROUTING"] {
err := m.iptablesClient.DeleteIfExists("mangle", "PREROUTING", rule...)
if err != nil {
log.Errorf("failed to delete rule: %v, %s", rule, err)
}
}
}
for _, ipsetName := range m.ipsetStore.ipsetNames() {
if err := ipset.Flush(ipsetName); err != nil {
log.Errorf("flush ipset %q during reset: %v", ipsetName, err)
}
if err := ipset.Destroy(ipsetName); err != nil {
log.Errorf("delete ipset %q during reset: %v", ipsetName, err)
}
m.ipsetStore.deleteIpset(ipsetName)
}
return nil
}
func (m *aclManager) createDefaultChains() error {
// chain netbird-acl-input-rules
if err := m.iptablesClient.NewChain(tableName, chainNameInputRules); err != nil {
log.Debugf("failed to create '%s' chain: %s", chainNameInputRules, err)
return err
}
// chain netbird-acl-output-rules
if err := m.iptablesClient.NewChain(tableName, chainNameOutputRules); err != nil {
log.Debugf("failed to create '%s' chain: %s", chainNameOutputRules, err)
return err
}
for chainName, rules := range m.entries {
for _, rule := range rules {
if err := m.iptablesClient.InsertUnique(tableName, chainName, 1, rule...); err != nil {
log.Debugf("failed to create input chain jump rule: %s", err)
return err
}
}
}
for chainName, entries := range m.optionalEntries {
for _, entry := range entries {
if err := m.iptablesClient.InsertUnique(tableName, chainName, entry.position, entry.spec...); err != nil {
log.Errorf("failed to insert optional entry %v: %v", entry.spec, err)
continue
}
m.entries[chainName] = append(m.entries[chainName], entry.spec)
}
}
clear(m.optionalEntries)
return nil
}
// seedInitialEntries adds default rules to the entries map, rules are inserted on pos 1, hence the order is reversed.
// We want to make sure our traffic is not dropped by existing rules.
// The existing FORWARD rules/policies decide outbound traffic towards our interface.
// In case the FORWARD policy is set to "drop", we add an established/related rule to allow return traffic for the inbound rule.
// The OUTPUT chain gets an extra rule to allow traffic to any set up routes, the return traffic is handled by the INPUT related/established rule.
func (m *aclManager) seedInitialEntries() {
established := getConntrackEstablished()
m.appendToEntries("INPUT", []string{"-i", m.wgIface.Name(), "-j", "DROP"})
m.appendToEntries("INPUT", []string{"-i", m.wgIface.Name(), "-j", chainNameInputRules})
m.appendToEntries("INPUT", append([]string{"-i", m.wgIface.Name()}, established...))
m.appendToEntries("FORWARD", []string{"-i", m.wgIface.Name(), "-j", "DROP"})
m.appendToEntries("FORWARD", []string{"-i", m.wgIface.Name(), "-j", m.routingFwChainName})
m.appendToEntries("FORWARD", append([]string{"-o", m.wgIface.Name()}, established...))
}
func (m *aclManager) seedInitialOptionalEntries() {
m.optionalEntries["FORWARD"] = []entry{
{
spec: []string{"-m", "mark", "--mark", fmt.Sprintf("%#x", nbnet.PreroutingFwmarkRedirected), "-j", chainNameInputRules},
position: 2,
},
}
m.optionalEntries["PREROUTING"] = []entry{
{
spec: []string{"-t", "mangle", "-i", m.wgIface.Name(), "-m", "addrtype", "--dst-type", "LOCAL", "-j", "MARK", "--set-mark", fmt.Sprintf("%#x", nbnet.PreroutingFwmarkRedirected)},
position: 1,
},
}
}
func (m *aclManager) appendToEntries(chainName string, spec []string) {
m.entries[chainName] = append(m.entries[chainName], spec)
}
func (m *aclManager) updateState() {
if m.stateManager == nil {
return
}
var currentState *ShutdownState
if existing := m.stateManager.GetState(currentState); existing != nil {
if existingState, ok := existing.(*ShutdownState); ok {
currentState = existingState
}
}
if currentState == nil {
currentState = &ShutdownState{}
}
currentState.Lock()
defer currentState.Unlock()
currentState.ACLEntries = m.entries
currentState.ACLIPsetStore = m.ipsetStore
if err := m.stateManager.UpdateState(currentState); err != nil {
log.Errorf("failed to update state: %v", err)
}
}
// filterRuleSpecs returns the specs of a filtering rule
func filterRuleSpecs(
ip net.IP, protocol string, sPort, dPort string, direction firewall.RuleDirection, action firewall.Action, ipsetName string,
) (specs []string) {
matchByIP := true
// don't use IP matching if IP is ip 0.0.0.0
if ip.String() == "0.0.0.0" {
matchByIP = false
}
switch direction {
case firewall.RuleDirectionIN:
if matchByIP {
if ipsetName != "" {
specs = append(specs, "-m", "set", "--set", ipsetName, "src")
} else {
specs = append(specs, "-s", ip.String())
}
}
case firewall.RuleDirectionOUT:
if matchByIP {
if ipsetName != "" {
specs = append(specs, "-m", "set", "--set", ipsetName, "dst")
} else {
specs = append(specs, "-d", ip.String())
}
}
}
if protocol != "all" {
specs = append(specs, "-p", protocol)
}
if sPort != "" {
specs = append(specs, "--sport", sPort)
}
if dPort != "" {
specs = append(specs, "--dport", dPort)
}
return append(specs, "-j", actionToStr(action))
}
func actionToStr(action firewall.Action) string {
if action == firewall.ActionAccept {
return "ACCEPT"
}
return "DROP"
}
func transformIPsetName(ipsetName string, sPort, dPort string) string {
switch {
case ipsetName == "":
return ""
case sPort != "" && dPort != "":
return ipsetName + "-sport-dport"
case sPort != "":
return ipsetName + "-sport"
case dPort != "":
return ipsetName + "-dport"
default:
return ipsetName
}
}