mirror of
https://github.com/juanfont/headscale.git
synced 2024-11-25 02:55:29 +00:00
ac8491efec
this commit changes and streamlines the dns_config into a new key, dns. It removes a combination of outdates and incompatible configuration options that made it easy to confuse what headscale could and could not do, or what to expect from ones configuration. Signed-off-by: Kristoffer Dalby <kristoffer@tailscale.com>
561 lines
14 KiB
Go
561 lines
14 KiB
Go
package types
|
|
|
|
import (
|
|
"database/sql"
|
|
"encoding/json"
|
|
"errors"
|
|
"fmt"
|
|
"net/netip"
|
|
"strconv"
|
|
"strings"
|
|
"time"
|
|
|
|
v1 "github.com/juanfont/headscale/gen/go/headscale/v1"
|
|
"github.com/juanfont/headscale/hscontrol/policy/matcher"
|
|
"github.com/juanfont/headscale/hscontrol/util"
|
|
"go4.org/netipx"
|
|
"google.golang.org/protobuf/types/known/timestamppb"
|
|
"gorm.io/gorm"
|
|
"tailscale.com/tailcfg"
|
|
"tailscale.com/types/key"
|
|
)
|
|
|
|
var (
|
|
ErrNodeAddressesInvalid = errors.New("failed to parse node addresses")
|
|
ErrHostnameTooLong = errors.New("hostname too long, cannot except 255 ASCII chars")
|
|
ErrNodeHasNoGivenName = errors.New("node has no given name")
|
|
ErrNodeUserHasNoName = errors.New("node user has no name")
|
|
)
|
|
|
|
type NodeID uint64
|
|
|
|
// type NodeConnectedMap *xsync.MapOf[NodeID, bool]
|
|
|
|
func (id NodeID) StableID() tailcfg.StableNodeID {
|
|
return tailcfg.StableNodeID(strconv.FormatUint(uint64(id), util.Base10))
|
|
}
|
|
|
|
func (id NodeID) NodeID() tailcfg.NodeID {
|
|
return tailcfg.NodeID(id)
|
|
}
|
|
|
|
func (id NodeID) Uint64() uint64 {
|
|
return uint64(id)
|
|
}
|
|
|
|
func (id NodeID) String() string {
|
|
return strconv.FormatUint(id.Uint64(), util.Base10)
|
|
}
|
|
|
|
// Node is a Headscale client.
|
|
type Node struct {
|
|
ID NodeID `gorm:"primary_key"`
|
|
|
|
// MachineKeyDatabaseField is the string representation of MachineKey
|
|
// it is _only_ used for reading and writing the key to the
|
|
// database and should not be used.
|
|
// Use MachineKey instead.
|
|
MachineKeyDatabaseField string `gorm:"column:machine_key;unique_index"`
|
|
MachineKey key.MachinePublic `gorm:"-"`
|
|
|
|
// NodeKeyDatabaseField is the string representation of NodeKey
|
|
// it is _only_ used for reading and writing the key to the
|
|
// database and should not be used.
|
|
// Use NodeKey instead.
|
|
NodeKeyDatabaseField string `gorm:"column:node_key"`
|
|
NodeKey key.NodePublic `gorm:"-"`
|
|
|
|
// DiscoKeyDatabaseField is the string representation of DiscoKey
|
|
// it is _only_ used for reading and writing the key to the
|
|
// database and should not be used.
|
|
// Use DiscoKey instead.
|
|
DiscoKeyDatabaseField string `gorm:"column:disco_key"`
|
|
DiscoKey key.DiscoPublic `gorm:"-"`
|
|
|
|
// EndpointsDatabaseField is the string list representation of Endpoints
|
|
// it is _only_ used for reading and writing the key to the
|
|
// database and should not be used.
|
|
// Use Endpoints instead.
|
|
EndpointsDatabaseField StringList `gorm:"column:endpoints"`
|
|
Endpoints []netip.AddrPort `gorm:"-"`
|
|
|
|
// EndpointsDatabaseField is the string list representation of Endpoints
|
|
// it is _only_ used for reading and writing the key to the
|
|
// database and should not be used.
|
|
// Use Endpoints instead.
|
|
HostinfoDatabaseField string `gorm:"column:host_info"`
|
|
Hostinfo *tailcfg.Hostinfo `gorm:"-"`
|
|
|
|
// IPv4DatabaseField is the string representation of v4 address,
|
|
// it is _only_ used for reading and writing the key to the
|
|
// database and should not be used.
|
|
// Use V4 instead.
|
|
IPv4DatabaseField sql.NullString `gorm:"column:ipv4"`
|
|
IPv4 *netip.Addr `gorm:"-"`
|
|
|
|
// IPv6DatabaseField is the string representation of v4 address,
|
|
// it is _only_ used for reading and writing the key to the
|
|
// database and should not be used.
|
|
// Use V6 instead.
|
|
IPv6DatabaseField sql.NullString `gorm:"column:ipv6"`
|
|
IPv6 *netip.Addr `gorm:"-"`
|
|
|
|
// Hostname represents the name given by the Tailscale
|
|
// client during registration
|
|
Hostname string
|
|
|
|
// Givenname represents either:
|
|
// a DNS normalized version of Hostname
|
|
// a valid name set by the User
|
|
//
|
|
// GivenName is the name used in all DNS related
|
|
// parts of headscale.
|
|
GivenName string `gorm:"type:varchar(63);unique_index"`
|
|
UserID uint
|
|
User User `gorm:"constraint:OnDelete:CASCADE;"`
|
|
|
|
RegisterMethod string
|
|
|
|
ForcedTags StringList
|
|
|
|
// TODO(kradalby): This seems like irrelevant information?
|
|
AuthKeyID *uint64 `sql:"DEFAULT:NULL"`
|
|
AuthKey *PreAuthKey `gorm:"constraint:OnDelete:SET NULL;"`
|
|
|
|
LastSeen *time.Time
|
|
Expiry *time.Time
|
|
|
|
Routes []Route `gorm:"constraint:OnDelete:CASCADE;"`
|
|
|
|
CreatedAt time.Time
|
|
UpdatedAt time.Time
|
|
DeletedAt *time.Time
|
|
|
|
IsOnline *bool `gorm:"-"`
|
|
}
|
|
|
|
type (
|
|
Nodes []*Node
|
|
)
|
|
|
|
// IsExpired returns whether the node registration has expired.
|
|
func (node Node) IsExpired() bool {
|
|
// If Expiry is not set, the client has not indicated that
|
|
// it wants an expiry time, it is therefor considered
|
|
// to mean "not expired"
|
|
if node.Expiry == nil || node.Expiry.IsZero() {
|
|
return false
|
|
}
|
|
|
|
return time.Since(*node.Expiry) > 0
|
|
}
|
|
|
|
// IsEphemeral returns if the node is registered as an Ephemeral node.
|
|
// https://tailscale.com/kb/1111/ephemeral-nodes/
|
|
func (node *Node) IsEphemeral() bool {
|
|
return node.AuthKey != nil && node.AuthKey.Ephemeral
|
|
}
|
|
|
|
func (node *Node) IPs() []netip.Addr {
|
|
var ret []netip.Addr
|
|
|
|
if node.IPv4 != nil {
|
|
ret = append(ret, *node.IPv4)
|
|
}
|
|
|
|
if node.IPv6 != nil {
|
|
ret = append(ret, *node.IPv6)
|
|
}
|
|
|
|
return ret
|
|
}
|
|
|
|
func (node *Node) Prefixes() []netip.Prefix {
|
|
addrs := []netip.Prefix{}
|
|
for _, nodeAddress := range node.IPs() {
|
|
ip := netip.PrefixFrom(nodeAddress, nodeAddress.BitLen())
|
|
addrs = append(addrs, ip)
|
|
}
|
|
|
|
return addrs
|
|
}
|
|
|
|
func (node *Node) IPsAsString() []string {
|
|
var ret []string
|
|
|
|
if node.IPv4 != nil {
|
|
ret = append(ret, node.IPv4.String())
|
|
}
|
|
|
|
if node.IPv6 != nil {
|
|
ret = append(ret, node.IPv6.String())
|
|
}
|
|
|
|
return ret
|
|
}
|
|
|
|
func (node *Node) InIPSet(set *netipx.IPSet) bool {
|
|
for _, nodeAddr := range node.IPs() {
|
|
if set.Contains(nodeAddr) {
|
|
return true
|
|
}
|
|
}
|
|
|
|
return false
|
|
}
|
|
|
|
// AppendToIPSet adds the individual ips in NodeAddresses to a
|
|
// given netipx.IPSetBuilder.
|
|
func (node *Node) AppendToIPSet(build *netipx.IPSetBuilder) {
|
|
for _, ip := range node.IPs() {
|
|
build.Add(ip)
|
|
}
|
|
}
|
|
|
|
func (node *Node) CanAccess(filter []tailcfg.FilterRule, node2 *Node) bool {
|
|
src := node.IPs()
|
|
allowedIPs := node2.IPs()
|
|
|
|
for _, route := range node2.Routes {
|
|
if route.Enabled {
|
|
allowedIPs = append(allowedIPs, netip.Prefix(route.Prefix).Addr())
|
|
}
|
|
}
|
|
|
|
for _, rule := range filter {
|
|
// TODO(kradalby): Cache or pregen this
|
|
matcher := matcher.MatchFromFilterRule(rule)
|
|
|
|
if !matcher.SrcsContainsIPs(src) {
|
|
continue
|
|
}
|
|
|
|
if matcher.DestsContainsIP(allowedIPs) {
|
|
return true
|
|
}
|
|
}
|
|
|
|
return false
|
|
}
|
|
|
|
func (nodes Nodes) FilterByIP(ip netip.Addr) Nodes {
|
|
var found Nodes
|
|
|
|
for _, node := range nodes {
|
|
if node.IPv4 != nil && ip == *node.IPv4 {
|
|
found = append(found, node)
|
|
continue
|
|
}
|
|
|
|
if node.IPv6 != nil && ip == *node.IPv6 {
|
|
found = append(found, node)
|
|
}
|
|
}
|
|
|
|
return found
|
|
}
|
|
|
|
// BeforeSave is a hook that ensures that some values that
|
|
// cannot be directly marshalled into database values are stored
|
|
// correctly in the database.
|
|
// This currently means storing the keys as strings.
|
|
func (node *Node) BeforeSave(tx *gorm.DB) error {
|
|
node.MachineKeyDatabaseField = node.MachineKey.String()
|
|
node.NodeKeyDatabaseField = node.NodeKey.String()
|
|
node.DiscoKeyDatabaseField = node.DiscoKey.String()
|
|
|
|
var endpoints StringList
|
|
for _, addrPort := range node.Endpoints {
|
|
endpoints = append(endpoints, addrPort.String())
|
|
}
|
|
|
|
node.EndpointsDatabaseField = endpoints
|
|
|
|
hi, err := json.Marshal(node.Hostinfo)
|
|
if err != nil {
|
|
return fmt.Errorf("marshalling Hostinfo to store in db: %w", err)
|
|
}
|
|
node.HostinfoDatabaseField = string(hi)
|
|
|
|
if node.IPv4 != nil {
|
|
node.IPv4DatabaseField.String, node.IPv4DatabaseField.Valid = node.IPv4.String(), true
|
|
} else {
|
|
node.IPv4DatabaseField.String, node.IPv4DatabaseField.Valid = "", false
|
|
}
|
|
|
|
if node.IPv6 != nil {
|
|
node.IPv6DatabaseField.String, node.IPv6DatabaseField.Valid = node.IPv6.String(), true
|
|
} else {
|
|
node.IPv6DatabaseField.String, node.IPv6DatabaseField.Valid = "", false
|
|
}
|
|
|
|
return nil
|
|
}
|
|
|
|
// AfterFind is a hook that ensures that Node objects fields that
|
|
// has a different type in the database is unwrapped and populated
|
|
// correctly.
|
|
// This currently unmarshals all the keys, stored as strings, into
|
|
// the proper types.
|
|
func (node *Node) AfterFind(tx *gorm.DB) error {
|
|
var machineKey key.MachinePublic
|
|
if err := machineKey.UnmarshalText([]byte(node.MachineKeyDatabaseField)); err != nil {
|
|
return fmt.Errorf("unmarshalling machine key from db: %w", err)
|
|
}
|
|
node.MachineKey = machineKey
|
|
|
|
var nodeKey key.NodePublic
|
|
if err := nodeKey.UnmarshalText([]byte(node.NodeKeyDatabaseField)); err != nil {
|
|
return fmt.Errorf("unmarshalling node key from db: %w", err)
|
|
}
|
|
node.NodeKey = nodeKey
|
|
|
|
// DiscoKey might be empty if a node has not sent it to headscale.
|
|
// This means that this might fail if the disco key is empty.
|
|
if node.DiscoKeyDatabaseField != "" {
|
|
var discoKey key.DiscoPublic
|
|
if err := discoKey.UnmarshalText([]byte(node.DiscoKeyDatabaseField)); err != nil {
|
|
return fmt.Errorf("unmarshalling disco key from db: %w", err)
|
|
}
|
|
node.DiscoKey = discoKey
|
|
}
|
|
|
|
endpoints := make([]netip.AddrPort, len(node.EndpointsDatabaseField))
|
|
for idx, ep := range node.EndpointsDatabaseField {
|
|
addrPort, err := netip.ParseAddrPort(ep)
|
|
if err != nil {
|
|
return fmt.Errorf("parsing endpoint from db: %w", err)
|
|
}
|
|
|
|
endpoints[idx] = addrPort
|
|
}
|
|
node.Endpoints = endpoints
|
|
|
|
var hi tailcfg.Hostinfo
|
|
if err := json.Unmarshal([]byte(node.HostinfoDatabaseField), &hi); err != nil {
|
|
return fmt.Errorf("unmarshalling hostinfo from database: %w", err)
|
|
}
|
|
node.Hostinfo = &hi
|
|
|
|
if node.IPv4DatabaseField.Valid {
|
|
ip, err := netip.ParseAddr(node.IPv4DatabaseField.String)
|
|
if err != nil {
|
|
return fmt.Errorf("parsing IPv4 from database: %w", err)
|
|
}
|
|
|
|
node.IPv4 = &ip
|
|
}
|
|
|
|
if node.IPv6DatabaseField.Valid {
|
|
ip, err := netip.ParseAddr(node.IPv6DatabaseField.String)
|
|
if err != nil {
|
|
return fmt.Errorf("parsing IPv6 from database: %w", err)
|
|
}
|
|
|
|
node.IPv6 = &ip
|
|
}
|
|
|
|
return nil
|
|
}
|
|
|
|
func (node *Node) Proto() *v1.Node {
|
|
nodeProto := &v1.Node{
|
|
Id: uint64(node.ID),
|
|
MachineKey: node.MachineKey.String(),
|
|
|
|
NodeKey: node.NodeKey.String(),
|
|
DiscoKey: node.DiscoKey.String(),
|
|
|
|
// TODO(kradalby): replace list with v4, v6 field?
|
|
IpAddresses: node.IPsAsString(),
|
|
Name: node.Hostname,
|
|
GivenName: node.GivenName,
|
|
User: node.User.Proto(),
|
|
ForcedTags: node.ForcedTags,
|
|
|
|
RegisterMethod: node.RegisterMethodToV1Enum(),
|
|
|
|
CreatedAt: timestamppb.New(node.CreatedAt),
|
|
}
|
|
|
|
if node.AuthKey != nil {
|
|
nodeProto.PreAuthKey = node.AuthKey.Proto()
|
|
}
|
|
|
|
if node.LastSeen != nil {
|
|
nodeProto.LastSeen = timestamppb.New(*node.LastSeen)
|
|
}
|
|
|
|
if node.Expiry != nil {
|
|
nodeProto.Expiry = timestamppb.New(*node.Expiry)
|
|
}
|
|
|
|
return nodeProto
|
|
}
|
|
|
|
func (node *Node) GetFQDN(cfg *Config, baseDomain string) (string, error) {
|
|
if node.GivenName == "" {
|
|
return "", fmt.Errorf("failed to create valid FQDN: %w", ErrNodeHasNoGivenName)
|
|
}
|
|
|
|
hostname := node.GivenName
|
|
|
|
if baseDomain != "" {
|
|
hostname = fmt.Sprintf(
|
|
"%s.%s",
|
|
node.GivenName,
|
|
baseDomain,
|
|
)
|
|
}
|
|
|
|
if cfg.DNSUserNameInMagicDNS {
|
|
if node.User.Name == "" {
|
|
return "", fmt.Errorf("failed to create valid FQDN: %w", ErrNodeUserHasNoName)
|
|
}
|
|
|
|
hostname = fmt.Sprintf(
|
|
"%s.%s.%s",
|
|
node.GivenName,
|
|
node.User.Name,
|
|
baseDomain,
|
|
)
|
|
}
|
|
|
|
if len(hostname) > MaxHostnameLength {
|
|
return "", fmt.Errorf(
|
|
"failed to create valid FQDN (%s): %w",
|
|
hostname,
|
|
ErrHostnameTooLong,
|
|
)
|
|
}
|
|
|
|
return hostname, nil
|
|
}
|
|
|
|
// func (node *Node) String() string {
|
|
// return node.Hostname
|
|
// }
|
|
|
|
// PeerChangeFromMapRequest takes a MapRequest and compares it to the node
|
|
// to produce a PeerChange struct that can be used to updated the node and
|
|
// inform peers about smaller changes to the node.
|
|
// When a field is added to this function, remember to also add it to:
|
|
// - node.ApplyPeerChange
|
|
// - logTracePeerChange in poll.go.
|
|
func (node *Node) PeerChangeFromMapRequest(req tailcfg.MapRequest) tailcfg.PeerChange {
|
|
ret := tailcfg.PeerChange{
|
|
NodeID: tailcfg.NodeID(node.ID),
|
|
}
|
|
|
|
if node.NodeKey.String() != req.NodeKey.String() {
|
|
ret.Key = &req.NodeKey
|
|
}
|
|
|
|
if node.DiscoKey.String() != req.DiscoKey.String() {
|
|
ret.DiscoKey = &req.DiscoKey
|
|
}
|
|
|
|
if node.Hostinfo != nil &&
|
|
node.Hostinfo.NetInfo != nil &&
|
|
req.Hostinfo != nil &&
|
|
req.Hostinfo.NetInfo != nil &&
|
|
node.Hostinfo.NetInfo.PreferredDERP != req.Hostinfo.NetInfo.PreferredDERP {
|
|
ret.DERPRegion = req.Hostinfo.NetInfo.PreferredDERP
|
|
}
|
|
|
|
if req.Hostinfo != nil && req.Hostinfo.NetInfo != nil {
|
|
// If there is no stored Hostinfo or NetInfo, use
|
|
// the new PreferredDERP.
|
|
if node.Hostinfo == nil {
|
|
ret.DERPRegion = req.Hostinfo.NetInfo.PreferredDERP
|
|
} else if node.Hostinfo.NetInfo == nil {
|
|
ret.DERPRegion = req.Hostinfo.NetInfo.PreferredDERP
|
|
} else {
|
|
// If there is a PreferredDERP check if it has changed.
|
|
if node.Hostinfo.NetInfo.PreferredDERP != req.Hostinfo.NetInfo.PreferredDERP {
|
|
ret.DERPRegion = req.Hostinfo.NetInfo.PreferredDERP
|
|
}
|
|
}
|
|
}
|
|
|
|
// TODO(kradalby): Find a good way to compare updates
|
|
ret.Endpoints = req.Endpoints
|
|
|
|
now := time.Now()
|
|
ret.LastSeen = &now
|
|
|
|
return ret
|
|
}
|
|
|
|
func (node *Node) RegisterMethodToV1Enum() v1.RegisterMethod {
|
|
switch node.RegisterMethod {
|
|
case "authkey":
|
|
return v1.RegisterMethod_REGISTER_METHOD_AUTH_KEY
|
|
case "oidc":
|
|
return v1.RegisterMethod_REGISTER_METHOD_OIDC
|
|
case "cli":
|
|
return v1.RegisterMethod_REGISTER_METHOD_CLI
|
|
default:
|
|
return v1.RegisterMethod_REGISTER_METHOD_UNSPECIFIED
|
|
}
|
|
}
|
|
|
|
// ApplyPeerChange takes a PeerChange struct and updates the node.
|
|
func (node *Node) ApplyPeerChange(change *tailcfg.PeerChange) {
|
|
if change.Key != nil {
|
|
node.NodeKey = *change.Key
|
|
}
|
|
|
|
if change.DiscoKey != nil {
|
|
node.DiscoKey = *change.DiscoKey
|
|
}
|
|
|
|
if change.Online != nil {
|
|
node.IsOnline = change.Online
|
|
}
|
|
|
|
if change.Endpoints != nil {
|
|
node.Endpoints = change.Endpoints
|
|
}
|
|
|
|
// This might technically not be useful as we replace
|
|
// the whole hostinfo blob when it has changed.
|
|
if change.DERPRegion != 0 {
|
|
if node.Hostinfo == nil {
|
|
node.Hostinfo = &tailcfg.Hostinfo{
|
|
NetInfo: &tailcfg.NetInfo{
|
|
PreferredDERP: change.DERPRegion,
|
|
},
|
|
}
|
|
} else if node.Hostinfo.NetInfo == nil {
|
|
node.Hostinfo.NetInfo = &tailcfg.NetInfo{
|
|
PreferredDERP: change.DERPRegion,
|
|
}
|
|
} else {
|
|
node.Hostinfo.NetInfo.PreferredDERP = change.DERPRegion
|
|
}
|
|
}
|
|
|
|
node.LastSeen = change.LastSeen
|
|
}
|
|
|
|
func (nodes Nodes) String() string {
|
|
temp := make([]string, len(nodes))
|
|
|
|
for index, node := range nodes {
|
|
temp[index] = node.Hostname
|
|
}
|
|
|
|
return fmt.Sprintf("[ %s ](%d)", strings.Join(temp, ", "), len(temp))
|
|
}
|
|
|
|
func (nodes Nodes) IDMap() map[NodeID]*Node {
|
|
ret := map[NodeID]*Node{}
|
|
|
|
for _, node := range nodes {
|
|
ret[node.ID] = node
|
|
}
|
|
|
|
return ret
|
|
}
|