/*
* ZeroTier One - Network Virtualization Everywhere
* Copyright (C) 2011-2017 ZeroTier, Inc. https://www.zerotier.com/
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see .
*
* --
*
* You can be released from the requirements of the license by purchasing
* a commercial license. Buying such a license is mandatory as soon as you
* develop commercial closed-source software that incorporates or links
* directly against ZeroTier software without disclosing the source code
* of your own application.
*/
#include
#include
#include
#include
#include "Constants.hpp"
#include "../version.h"
#include "Network.hpp"
#include "RuntimeEnvironment.hpp"
#include "MAC.hpp"
#include "Address.hpp"
#include "InetAddress.hpp"
#include "Switch.hpp"
#include "Buffer.hpp"
#include "Packet.hpp"
#include "NetworkController.hpp"
#include "Node.hpp"
#include "Peer.hpp"
#include "Trace.hpp"
namespace ZeroTier {
namespace {
// Returns true if packet appears valid; pos and proto will be set
static bool _ipv6GetPayload(const uint8_t *frameData,unsigned int frameLen,unsigned int &pos,unsigned int &proto)
{
if (frameLen < 40)
return false;
pos = 40;
proto = frameData[6];
while (pos <= frameLen) {
switch(proto) {
case 0: // hop-by-hop options
case 43: // routing
case 60: // destination options
case 135: // mobility options
if ((pos + 8) > frameLen)
return false; // invalid!
proto = frameData[pos];
pos += ((unsigned int)frameData[pos + 1] * 8) + 8;
break;
//case 44: // fragment -- we currently can't parse these and they are deprecated in IPv6 anyway
//case 50:
//case 51: // IPSec ESP and AH -- we have to stop here since this is encrypted stuff
default:
return true;
}
}
return false; // overflow == invalid
}
enum _doZtFilterResult
{
DOZTFILTER_NO_MATCH,
DOZTFILTER_DROP,
DOZTFILTER_REDIRECT,
DOZTFILTER_ACCEPT,
DOZTFILTER_SUPER_ACCEPT
};
static _doZtFilterResult _doZtFilter(
const RuntimeEnvironment *RR,
Trace::RuleResultLog &rrl,
const NetworkConfig &nconf,
const Membership *membership, // can be NULL
const bool inbound,
const Address &ztSource,
Address &ztDest, // MUTABLE -- is changed on REDIRECT actions
const MAC &macSource,
const MAC &macDest,
const uint8_t *const frameData,
const unsigned int frameLen,
const unsigned int etherType,
const unsigned int vlanId,
const ZT_VirtualNetworkRule *rules, // cannot be NULL
const unsigned int ruleCount,
Address &cc, // MUTABLE -- set to TEE destination if TEE action is taken or left alone otherwise
unsigned int &ccLength, // MUTABLE -- set to length of packet payload to TEE
bool &ccWatch) // MUTABLE -- set to true for WATCH target as opposed to normal TEE
{
// Set to true if we are a TEE/REDIRECT/WATCH target
bool superAccept = false;
// The default match state for each set of entries starts as 'true' since an
// ACTION with no MATCH entries preceding it is always taken.
uint8_t thisSetMatches = 1;
rrl.clear();
for(unsigned int rn=0;rnidentity.address()) {
if (inbound) {
return DOZTFILTER_SUPER_ACCEPT;
} else {
}
} else if (fwdAddr == ztDest) {
} else {
if (rt == ZT_NETWORK_RULE_ACTION_REDIRECT) {
ztDest = fwdAddr;
return DOZTFILTER_REDIRECT;
} else {
cc = fwdAddr;
ccLength = (rules[rn].v.fwd.length != 0) ? ((frameLen < (unsigned int)rules[rn].v.fwd.length) ? frameLen : (unsigned int)rules[rn].v.fwd.length) : frameLen;
ccWatch = (rt == ZT_NETWORK_RULE_ACTION_WATCH);
}
}
} continue;
case ZT_NETWORK_RULE_ACTION_BREAK:
return DOZTFILTER_NO_MATCH;
// Unrecognized ACTIONs are ignored as no-ops
default:
continue;
}
} else {
// If this is an incoming packet and we are a TEE or REDIRECT target, we should
// super-accept if we accept at all. This will cause us to accept redirected or
// tee'd packets in spite of MAC and ZT addressing checks.
if (inbound) {
switch(rt) {
case ZT_NETWORK_RULE_ACTION_TEE:
case ZT_NETWORK_RULE_ACTION_WATCH:
case ZT_NETWORK_RULE_ACTION_REDIRECT:
if (RR->identity.address() == rules[rn].v.fwd.address)
superAccept = true;
break;
default:
break;
}
}
thisSetMatches = 1; // reset to default true for next batch of entries
continue;
}
}
// Circuit breaker: no need to evaluate an AND if the set's match state
// is currently false since anything AND false is false.
if ((!thisSetMatches)&&(!(rules[rn].t & 0x40))) {
rrl.logSkipped(rn,thisSetMatches);
continue;
}
// If this was not an ACTION evaluate next MATCH and update thisSetMatches with (AND [result])
uint8_t thisRuleMatches = 0;
uint64_t ownershipVerificationMask = 1; // this magic value means it hasn't been computed yet -- this is done lazily the first time it's needed
switch(rt) {
case ZT_NETWORK_RULE_MATCH_SOURCE_ZEROTIER_ADDRESS:
thisRuleMatches = (uint8_t)(rules[rn].v.zt == ztSource.toInt());
break;
case ZT_NETWORK_RULE_MATCH_DEST_ZEROTIER_ADDRESS:
thisRuleMatches = (uint8_t)(rules[rn].v.zt == ztDest.toInt());
break;
case ZT_NETWORK_RULE_MATCH_VLAN_ID:
thisRuleMatches = (uint8_t)(rules[rn].v.vlanId == (uint16_t)vlanId);
break;
case ZT_NETWORK_RULE_MATCH_VLAN_PCP:
// NOT SUPPORTED YET
thisRuleMatches = (uint8_t)(rules[rn].v.vlanPcp == 0);
break;
case ZT_NETWORK_RULE_MATCH_VLAN_DEI:
// NOT SUPPORTED YET
thisRuleMatches = (uint8_t)(rules[rn].v.vlanDei == 0);
break;
case ZT_NETWORK_RULE_MATCH_MAC_SOURCE:
thisRuleMatches = (uint8_t)(MAC(rules[rn].v.mac,6) == macSource);
break;
case ZT_NETWORK_RULE_MATCH_MAC_DEST:
thisRuleMatches = (uint8_t)(MAC(rules[rn].v.mac,6) == macDest);
break;
case ZT_NETWORK_RULE_MATCH_IPV4_SOURCE:
if ((etherType == ZT_ETHERTYPE_IPV4)&&(frameLen >= 20)) {
thisRuleMatches = (uint8_t)(InetAddress((const void *)&(rules[rn].v.ipv4.ip),4,rules[rn].v.ipv4.mask).containsAddress(InetAddress((const void *)(frameData + 12),4,0)));
} else {
thisRuleMatches = 0;
}
break;
case ZT_NETWORK_RULE_MATCH_IPV4_DEST:
if ((etherType == ZT_ETHERTYPE_IPV4)&&(frameLen >= 20)) {
thisRuleMatches = (uint8_t)(InetAddress((const void *)&(rules[rn].v.ipv4.ip),4,rules[rn].v.ipv4.mask).containsAddress(InetAddress((const void *)(frameData + 16),4,0)));
} else {
thisRuleMatches = 0;
}
break;
case ZT_NETWORK_RULE_MATCH_IPV6_SOURCE:
if ((etherType == ZT_ETHERTYPE_IPV6)&&(frameLen >= 40)) {
thisRuleMatches = (uint8_t)(InetAddress((const void *)rules[rn].v.ipv6.ip,16,rules[rn].v.ipv6.mask).containsAddress(InetAddress((const void *)(frameData + 8),16,0)));
} else {
thisRuleMatches = 0;
}
break;
case ZT_NETWORK_RULE_MATCH_IPV6_DEST:
if ((etherType == ZT_ETHERTYPE_IPV6)&&(frameLen >= 40)) {
thisRuleMatches = (uint8_t)(InetAddress((const void *)rules[rn].v.ipv6.ip,16,rules[rn].v.ipv6.mask).containsAddress(InetAddress((const void *)(frameData + 24),16,0)));
} else {
thisRuleMatches = 0;
}
break;
case ZT_NETWORK_RULE_MATCH_IP_TOS:
if ((etherType == ZT_ETHERTYPE_IPV4)&&(frameLen >= 20)) {
const uint8_t tosMasked = frameData[1] & rules[rn].v.ipTos.mask;
thisRuleMatches = (uint8_t)((tosMasked >= rules[rn].v.ipTos.value[0])&&(tosMasked <= rules[rn].v.ipTos.value[1]));
} else if ((etherType == ZT_ETHERTYPE_IPV6)&&(frameLen >= 40)) {
const uint8_t tosMasked = (((frameData[0] << 4) & 0xf0) | ((frameData[1] >> 4) & 0x0f)) & rules[rn].v.ipTos.mask;
thisRuleMatches = (uint8_t)((tosMasked >= rules[rn].v.ipTos.value[0])&&(tosMasked <= rules[rn].v.ipTos.value[1]));
} else {
thisRuleMatches = 0;
}
break;
case ZT_NETWORK_RULE_MATCH_IP_PROTOCOL:
if ((etherType == ZT_ETHERTYPE_IPV4)&&(frameLen >= 20)) {
thisRuleMatches = (uint8_t)(rules[rn].v.ipProtocol == frameData[9]);
} else if (etherType == ZT_ETHERTYPE_IPV6) {
unsigned int pos = 0,proto = 0;
if (_ipv6GetPayload(frameData,frameLen,pos,proto)) {
thisRuleMatches = (uint8_t)(rules[rn].v.ipProtocol == (uint8_t)proto);
} else {
thisRuleMatches = 0;
}
} else {
thisRuleMatches = 0;
}
break;
case ZT_NETWORK_RULE_MATCH_ETHERTYPE:
thisRuleMatches = (uint8_t)(rules[rn].v.etherType == (uint16_t)etherType);
break;
case ZT_NETWORK_RULE_MATCH_ICMP:
if ((etherType == ZT_ETHERTYPE_IPV4)&&(frameLen >= 20)) {
if (frameData[9] == 0x01) { // IP protocol == ICMP
const unsigned int ihl = (frameData[0] & 0xf) * 4;
if (frameLen >= (ihl + 2)) {
if (rules[rn].v.icmp.type == frameData[ihl]) {
if ((rules[rn].v.icmp.flags & 0x01) != 0) {
thisRuleMatches = (uint8_t)(frameData[ihl+1] == rules[rn].v.icmp.code);
} else {
thisRuleMatches = 1;
}
} else {
thisRuleMatches = 0;
}
} else {
thisRuleMatches = 0;
}
} else {
thisRuleMatches = 0;
}
} else if (etherType == ZT_ETHERTYPE_IPV6) {
unsigned int pos = 0,proto = 0;
if (_ipv6GetPayload(frameData,frameLen,pos,proto)) {
if ((proto == 0x3a)&&(frameLen >= (pos+2))) {
if (rules[rn].v.icmp.type == frameData[pos]) {
if ((rules[rn].v.icmp.flags & 0x01) != 0) {
thisRuleMatches = (uint8_t)(frameData[pos+1] == rules[rn].v.icmp.code);
} else {
thisRuleMatches = 1;
}
} else {
thisRuleMatches = 0;
}
} else {
thisRuleMatches = 0;
}
} else {
thisRuleMatches = 0;
}
} else {
thisRuleMatches = 0;
}
break;
case ZT_NETWORK_RULE_MATCH_IP_SOURCE_PORT_RANGE:
case ZT_NETWORK_RULE_MATCH_IP_DEST_PORT_RANGE:
if ((etherType == ZT_ETHERTYPE_IPV4)&&(frameLen >= 20)) {
const unsigned int headerLen = 4 * (frameData[0] & 0xf);
int p = -1;
switch(frameData[9]) { // IP protocol number
// All these start with 16-bit source and destination port in that order
case 0x06: // TCP
case 0x11: // UDP
case 0x84: // SCTP
case 0x88: // UDPLite
if (frameLen > (headerLen + 4)) {
unsigned int pos = headerLen + ((rt == ZT_NETWORK_RULE_MATCH_IP_DEST_PORT_RANGE) ? 2 : 0);
p = (int)frameData[pos++] << 8;
p |= (int)frameData[pos];
}
break;
}
thisRuleMatches = (p >= 0) ? (uint8_t)((p >= (int)rules[rn].v.port[0])&&(p <= (int)rules[rn].v.port[1])) : (uint8_t)0;
} else if (etherType == ZT_ETHERTYPE_IPV6) {
unsigned int pos = 0,proto = 0;
if (_ipv6GetPayload(frameData,frameLen,pos,proto)) {
int p = -1;
switch(proto) { // IP protocol number
// All these start with 16-bit source and destination port in that order
case 0x06: // TCP
case 0x11: // UDP
case 0x84: // SCTP
case 0x88: // UDPLite
if (frameLen > (pos + 4)) {
if (rt == ZT_NETWORK_RULE_MATCH_IP_DEST_PORT_RANGE) pos += 2;
p = (int)frameData[pos++] << 8;
p |= (int)frameData[pos];
}
break;
}
thisRuleMatches = (p > 0) ? (uint8_t)((p >= (int)rules[rn].v.port[0])&&(p <= (int)rules[rn].v.port[1])) : (uint8_t)0;
} else {
thisRuleMatches = 0;
}
} else {
thisRuleMatches = 0;
}
break;
case ZT_NETWORK_RULE_MATCH_CHARACTERISTICS: {
uint64_t cf = (inbound) ? ZT_RULE_PACKET_CHARACTERISTICS_INBOUND : 0ULL;
if (macDest.isMulticast()) cf |= ZT_RULE_PACKET_CHARACTERISTICS_MULTICAST;
if (macDest.isBroadcast()) cf |= ZT_RULE_PACKET_CHARACTERISTICS_BROADCAST;
if (ownershipVerificationMask == 1) {
ownershipVerificationMask = 0;
InetAddress src;
if ((etherType == ZT_ETHERTYPE_IPV4)&&(frameLen >= 20)) {
src.set((const void *)(frameData + 12),4,0);
} else if ((etherType == ZT_ETHERTYPE_IPV6)&&(frameLen >= 40)) {
// IPv6 NDP requires special handling, since the src and dest IPs in the packet are empty or link-local.
if ( (frameLen >= (40 + 8 + 16)) && (frameData[6] == 0x3a) && ((frameData[40] == 0x87)||(frameData[40] == 0x88)) ) {
if (frameData[40] == 0x87) {
// Neighbor solicitations contain no reliable source address, so we implement a small
// hack by considering them authenticated. Otherwise you would pretty much have to do
// this manually in the rule set for IPv6 to work at all.
ownershipVerificationMask |= ZT_RULE_PACKET_CHARACTERISTICS_SENDER_IP_AUTHENTICATED;
} else {
// Neighbor advertisements on the other hand can absolutely be authenticated.
src.set((const void *)(frameData + 40 + 8),16,0);
}
} else {
// Other IPv6 packets can be handled normally
src.set((const void *)(frameData + 8),16,0);
}
} else if ((etherType == ZT_ETHERTYPE_ARP)&&(frameLen >= 28)) {
src.set((const void *)(frameData + 14),4,0);
}
if (inbound) {
if (membership) {
if ((src)&&(membership->hasCertificateOfOwnershipFor(nconf,src)))
ownershipVerificationMask |= ZT_RULE_PACKET_CHARACTERISTICS_SENDER_IP_AUTHENTICATED;
if (membership->hasCertificateOfOwnershipFor(nconf,macSource))
ownershipVerificationMask |= ZT_RULE_PACKET_CHARACTERISTICS_SENDER_MAC_AUTHENTICATED;
}
} else {
for(unsigned int i=0;i= 20)&&(frameData[9] == 0x06)) {
const unsigned int headerLen = 4 * (frameData[0] & 0xf);
cf |= (uint64_t)frameData[headerLen + 13];
cf |= (((uint64_t)(frameData[headerLen + 12] & 0x0f)) << 8);
} else if (etherType == ZT_ETHERTYPE_IPV6) {
unsigned int pos = 0,proto = 0;
if (_ipv6GetPayload(frameData,frameLen,pos,proto)) {
if ((proto == 0x06)&&(frameLen > (pos + 14))) {
cf |= (uint64_t)frameData[pos + 13];
cf |= (((uint64_t)(frameData[pos + 12] & 0x0f)) << 8);
}
}
}
thisRuleMatches = (uint8_t)((cf & rules[rn].v.characteristics) != 0);
} break;
case ZT_NETWORK_RULE_MATCH_FRAME_SIZE_RANGE:
thisRuleMatches = (uint8_t)((frameLen >= (unsigned int)rules[rn].v.frameSize[0])&&(frameLen <= (unsigned int)rules[rn].v.frameSize[1]));
break;
case ZT_NETWORK_RULE_MATCH_RANDOM:
thisRuleMatches = (uint8_t)((uint32_t)(RR->node->prng() & 0xffffffffULL) <= rules[rn].v.randomProbability);
break;
case ZT_NETWORK_RULE_MATCH_TAGS_DIFFERENCE:
case ZT_NETWORK_RULE_MATCH_TAGS_BITWISE_AND:
case ZT_NETWORK_RULE_MATCH_TAGS_BITWISE_OR:
case ZT_NETWORK_RULE_MATCH_TAGS_BITWISE_XOR:
case ZT_NETWORK_RULE_MATCH_TAGS_EQUAL: {
const Tag *const localTag = std::lower_bound(&(nconf.tags[0]),&(nconf.tags[nconf.tagCount]),rules[rn].v.tag.id,Tag::IdComparePredicate());
if ((localTag != &(nconf.tags[nconf.tagCount]))&&(localTag->id() == rules[rn].v.tag.id)) {
const Tag *const remoteTag = ((membership) ? membership->getTag(nconf,rules[rn].v.tag.id) : (const Tag *)0);
if (remoteTag) {
const uint32_t ltv = localTag->value();
const uint32_t rtv = remoteTag->value();
if (rt == ZT_NETWORK_RULE_MATCH_TAGS_DIFFERENCE) {
const uint32_t diff = (ltv > rtv) ? (ltv - rtv) : (rtv - ltv);
thisRuleMatches = (uint8_t)(diff <= rules[rn].v.tag.value);
} else if (rt == ZT_NETWORK_RULE_MATCH_TAGS_BITWISE_AND) {
thisRuleMatches = (uint8_t)((ltv & rtv) == rules[rn].v.tag.value);
} else if (rt == ZT_NETWORK_RULE_MATCH_TAGS_BITWISE_OR) {
thisRuleMatches = (uint8_t)((ltv | rtv) == rules[rn].v.tag.value);
} else if (rt == ZT_NETWORK_RULE_MATCH_TAGS_BITWISE_XOR) {
thisRuleMatches = (uint8_t)((ltv ^ rtv) == rules[rn].v.tag.value);
} else if (rt == ZT_NETWORK_RULE_MATCH_TAGS_EQUAL) {
thisRuleMatches = (uint8_t)((ltv == rules[rn].v.tag.value)&&(rtv == rules[rn].v.tag.value));
} else { // sanity check, can't really happen
thisRuleMatches = 0;
}
} else {
if ((inbound)&&(!superAccept)) {
thisRuleMatches = 0;
} else {
// Outbound side is not strict since if we have to match both tags and
// we are sending a first packet to a recipient, we probably do not know
// about their tags yet. They will filter on inbound and we will filter
// once we get their tag. If we are a tee/redirect target we are also
// not strict since we likely do not have these tags.
thisRuleMatches = 1;
}
}
} else {
thisRuleMatches = 0;
}
} break;
case ZT_NETWORK_RULE_MATCH_TAG_SENDER:
case ZT_NETWORK_RULE_MATCH_TAG_RECEIVER: {
if (superAccept) {
thisRuleMatches = 1;
} else if ( ((rt == ZT_NETWORK_RULE_MATCH_TAG_SENDER)&&(inbound)) || ((rt == ZT_NETWORK_RULE_MATCH_TAG_RECEIVER)&&(!inbound)) ) {
const Tag *const remoteTag = ((membership) ? membership->getTag(nconf,rules[rn].v.tag.id) : (const Tag *)0);
if (remoteTag) {
thisRuleMatches = (uint8_t)(remoteTag->value() == rules[rn].v.tag.value);
} else {
if (rt == ZT_NETWORK_RULE_MATCH_TAG_RECEIVER) {
// If we are checking the receiver and this is an outbound packet, we
// can't be strict since we may not yet know the receiver's tag.
thisRuleMatches = 1;
} else {
thisRuleMatches = 0;
}
}
} else { // sender and outbound or receiver and inbound
const Tag *const localTag = std::lower_bound(&(nconf.tags[0]),&(nconf.tags[nconf.tagCount]),rules[rn].v.tag.id,Tag::IdComparePredicate());
if ((localTag != &(nconf.tags[nconf.tagCount]))&&(localTag->id() == rules[rn].v.tag.id)) {
thisRuleMatches = (uint8_t)(localTag->value() == rules[rn].v.tag.value);
} else {
thisRuleMatches = 0;
}
}
} break;
// The result of an unsupported MATCH is configurable at the network
// level via a flag.
default:
thisRuleMatches = (uint8_t)((nconf.flags & ZT_NETWORKCONFIG_FLAG_RULES_RESULT_OF_UNSUPPORTED_MATCH) != 0);
break;
}
rrl.log(rn,thisRuleMatches,thisSetMatches);
if ((rules[rn].t & 0x40))
thisSetMatches |= (thisRuleMatches ^ ((rules[rn].t >> 7) & 1));
else thisSetMatches &= (thisRuleMatches ^ ((rules[rn].t >> 7) & 1));
}
return DOZTFILTER_NO_MATCH;
}
} // anonymous namespace
const ZeroTier::MulticastGroup Network::BROADCAST(ZeroTier::MAC(0xffffffffffffULL),0);
Network::Network(const RuntimeEnvironment *renv,void *tPtr,uint64_t nwid,void *uptr,const NetworkConfig *nconf) :
RR(renv),
_uPtr(uptr),
_id(nwid),
_lastAnnouncedMulticastGroupsUpstream(0),
_mac(renv->identity.address(),nwid),
_portInitialized(false),
_lastConfigUpdate(0),
_destroyed(false),
_netconfFailure(NETCONF_FAILURE_NONE),
_portError(0)
{
for(int i=0;isetConfiguration(tPtr,*nconf,false);
_lastConfigUpdate = 0; // still want to re-request since it's likely outdated
} else {
uint64_t tmp[2];
tmp[0] = nwid; tmp[1] = 0;
bool got = false;
Dictionary *dict = new Dictionary();
try {
int n = RR->node->stateObjectGet(tPtr,ZT_STATE_OBJECT_NETWORK_CONFIG,tmp,dict->unsafeData(),ZT_NETWORKCONFIG_DICT_CAPACITY - 1);
if (n > 1) {
NetworkConfig *nconf = new NetworkConfig();
try {
if (nconf->fromDictionary(*dict)) {
this->setConfiguration(tPtr,*nconf,false);
_lastConfigUpdate = 0; // still want to re-request an update since it's likely outdated
got = true;
}
} catch ( ... ) {}
delete nconf;
}
} catch ( ... ) {}
delete dict;
if (!got)
RR->node->stateObjectPut(tPtr,ZT_STATE_OBJECT_NETWORK_CONFIG,tmp,"\n",1);
}
if (!_portInitialized) {
ZT_VirtualNetworkConfig ctmp;
_externalConfig(&ctmp);
_portError = RR->node->configureVirtualNetworkPort(tPtr,_id,&_uPtr,ZT_VIRTUAL_NETWORK_CONFIG_OPERATION_UP,&ctmp);
_portInitialized = true;
}
}
Network::~Network()
{
ZT_VirtualNetworkConfig ctmp;
_externalConfig(&ctmp);
if (_destroyed) {
// This is done in Node::leave() so we can pass tPtr properly
//RR->node->configureVirtualNetworkPort((void *)0,_id,&_uPtr,ZT_VIRTUAL_NETWORK_CONFIG_OPERATION_DESTROY,&ctmp);
} else {
RR->node->configureVirtualNetworkPort((void *)0,_id,&_uPtr,ZT_VIRTUAL_NETWORK_CONFIG_OPERATION_DOWN,&ctmp);
}
}
bool Network::filterOutgoingPacket(
void *tPtr,
const bool noTee,
const Address &ztSource,
const Address &ztDest,
const MAC &macSource,
const MAC &macDest,
const uint8_t *frameData,
const unsigned int frameLen,
const unsigned int etherType,
const unsigned int vlanId)
{
const uint64_t now = RR->node->now();
Address ztFinalDest(ztDest);
int localCapabilityIndex = -1;
int accept = 0;
Trace::RuleResultLog rrl,crrl;
Address cc;
unsigned int ccLength = 0;
bool ccWatch = false;
Mutex::Lock _l(_lock);
Membership *const membership = (ztDest) ? _memberships.get(ztDest) : (Membership *)0;
switch(_doZtFilter(RR,rrl,_config,membership,false,ztSource,ztFinalDest,macSource,macDest,frameData,frameLen,etherType,vlanId,_config.rules,_config.ruleCount,cc,ccLength,ccWatch)) {
case DOZTFILTER_NO_MATCH: {
for(unsigned int c=0;c<_config.capabilityCount;++c) {
ztFinalDest = ztDest; // sanity check, shouldn't be possible if there was no match
Address cc2;
unsigned int ccLength2 = 0;
bool ccWatch2 = false;
switch (_doZtFilter(RR,crrl,_config,membership,false,ztSource,ztFinalDest,macSource,macDest,frameData,frameLen,etherType,vlanId,_config.capabilities[c].rules(),_config.capabilities[c].ruleCount(),cc2,ccLength2,ccWatch2)) {
case DOZTFILTER_NO_MATCH:
case DOZTFILTER_DROP: // explicit DROP in a capability just terminates its evaluation and is an anti-pattern
break;
case DOZTFILTER_REDIRECT: // interpreted as ACCEPT but ztFinalDest will have been changed in _doZtFilter()
case DOZTFILTER_ACCEPT:
case DOZTFILTER_SUPER_ACCEPT: // no difference in behavior on outbound side in capabilities
localCapabilityIndex = (int)c;
accept = 1;
if ((!noTee)&&(cc2)) {
Membership &m2 = _membership(cc2);
m2.pushCredentials(RR,tPtr,now,cc2,_config,localCapabilityIndex,false);
Packet outp(cc2,RR->identity.address(),Packet::VERB_EXT_FRAME);
outp.append(_id);
outp.append((uint8_t)(ccWatch2 ? 0x16 : 0x02));
macDest.appendTo(outp);
macSource.appendTo(outp);
outp.append((uint16_t)etherType);
outp.append(frameData,ccLength2);
outp.compress();
RR->sw->send(tPtr,outp,true);
}
break;
}
if (accept)
break;
}
} break;
case DOZTFILTER_DROP:
if (_config.remoteTraceTarget)
RR->t->networkFilter(tPtr,*this,rrl,(Trace::RuleResultLog *)0,(Capability *)0,ztSource,ztDest,macSource,macDest,frameData,frameLen,etherType,vlanId,noTee,false,0);
return false;
case DOZTFILTER_REDIRECT: // interpreted as ACCEPT but ztFinalDest will have been changed in _doZtFilter()
case DOZTFILTER_ACCEPT:
accept = 1;
break;
case DOZTFILTER_SUPER_ACCEPT:
accept = 2;
break;
}
if (accept) {
if (membership)
membership->pushCredentials(RR,tPtr,now,ztDest,_config,localCapabilityIndex,false);
if ((!noTee)&&(cc)) {
Membership &m2 = _membership(cc);
m2.pushCredentials(RR,tPtr,now,cc,_config,localCapabilityIndex,false);
Packet outp(cc,RR->identity.address(),Packet::VERB_EXT_FRAME);
outp.append(_id);
outp.append((uint8_t)(ccWatch ? 0x16 : 0x02));
macDest.appendTo(outp);
macSource.appendTo(outp);
outp.append((uint16_t)etherType);
outp.append(frameData,ccLength);
outp.compress();
RR->sw->send(tPtr,outp,true);
}
if ((ztDest != ztFinalDest)&&(ztFinalDest)) {
Membership &m2 = _membership(ztFinalDest);
m2.pushCredentials(RR,tPtr,now,ztFinalDest,_config,localCapabilityIndex,false);
Packet outp(ztFinalDest,RR->identity.address(),Packet::VERB_EXT_FRAME);
outp.append(_id);
outp.append((uint8_t)0x04);
macDest.appendTo(outp);
macSource.appendTo(outp);
outp.append((uint16_t)etherType);
outp.append(frameData,frameLen);
outp.compress();
RR->sw->send(tPtr,outp,true);
if (_config.remoteTraceTarget)
RR->t->networkFilter(tPtr,*this,rrl,(localCapabilityIndex >= 0) ? &crrl : (Trace::RuleResultLog *)0,(localCapabilityIndex >= 0) ? &(_config.capabilities[localCapabilityIndex]) : (Capability *)0,ztSource,ztDest,macSource,macDest,frameData,frameLen,etherType,vlanId,noTee,false,0);
return false; // DROP locally, since we redirected
} else {
if (_config.remoteTraceTarget)
RR->t->networkFilter(tPtr,*this,rrl,(localCapabilityIndex >= 0) ? &crrl : (Trace::RuleResultLog *)0,(localCapabilityIndex >= 0) ? &(_config.capabilities[localCapabilityIndex]) : (Capability *)0,ztSource,ztDest,macSource,macDest,frameData,frameLen,etherType,vlanId,noTee,false,1);
return true;
}
} else {
if (_config.remoteTraceTarget)
RR->t->networkFilter(tPtr,*this,rrl,(localCapabilityIndex >= 0) ? &crrl : (Trace::RuleResultLog *)0,(localCapabilityIndex >= 0) ? &(_config.capabilities[localCapabilityIndex]) : (Capability *)0,ztSource,ztDest,macSource,macDest,frameData,frameLen,etherType,vlanId,noTee,false,0);
return false;
}
}
int Network::filterIncomingPacket(
void *tPtr,
const SharedPtr &sourcePeer,
const Address &ztDest,
const MAC &macSource,
const MAC &macDest,
const uint8_t *frameData,
const unsigned int frameLen,
const unsigned int etherType,
const unsigned int vlanId)
{
Address ztFinalDest(ztDest);
Trace::RuleResultLog rrl,crrl;
int accept = 0;
Address cc;
unsigned int ccLength = 0;
bool ccWatch = false;
const Capability *c = (Capability *)0;
Mutex::Lock _l(_lock);
Membership &membership = _membership(sourcePeer->address());
switch (_doZtFilter(RR,rrl,_config,&membership,true,sourcePeer->address(),ztFinalDest,macSource,macDest,frameData,frameLen,etherType,vlanId,_config.rules,_config.ruleCount,cc,ccLength,ccWatch)) {
case DOZTFILTER_NO_MATCH: {
Membership::CapabilityIterator mci(membership,_config);
while ((c = mci.next())) {
ztFinalDest = ztDest; // sanity check, should be unmodified if there was no match
Address cc2;
unsigned int ccLength2 = 0;
bool ccWatch2 = false;
switch(_doZtFilter(RR,crrl,_config,&membership,true,sourcePeer->address(),ztFinalDest,macSource,macDest,frameData,frameLen,etherType,vlanId,c->rules(),c->ruleCount(),cc2,ccLength2,ccWatch2)) {
case DOZTFILTER_NO_MATCH:
case DOZTFILTER_DROP: // explicit DROP in a capability just terminates its evaluation and is an anti-pattern
break;
case DOZTFILTER_REDIRECT: // interpreted as ACCEPT but ztDest will have been changed in _doZtFilter()
case DOZTFILTER_ACCEPT:
accept = 1; // ACCEPT
break;
case DOZTFILTER_SUPER_ACCEPT:
accept = 2; // super-ACCEPT
break;
}
if (accept) {
if (cc2) {
_membership(cc2).pushCredentials(RR,tPtr,RR->node->now(),cc2,_config,-1,false);
Packet outp(cc2,RR->identity.address(),Packet::VERB_EXT_FRAME);
outp.append(_id);
outp.append((uint8_t)(ccWatch2 ? 0x1c : 0x08));
macDest.appendTo(outp);
macSource.appendTo(outp);
outp.append((uint16_t)etherType);
outp.append(frameData,ccLength2);
outp.compress();
RR->sw->send(tPtr,outp,true);
}
break;
}
}
} break;
case DOZTFILTER_DROP:
if (_config.remoteTraceTarget)
RR->t->networkFilter(tPtr,*this,rrl,(Trace::RuleResultLog *)0,(Capability *)0,sourcePeer->address(),ztDest,macSource,macDest,frameData,frameLen,etherType,vlanId,false,true,0);
return 0; // DROP
case DOZTFILTER_REDIRECT: // interpreted as ACCEPT but ztFinalDest will have been changed in _doZtFilter()
case DOZTFILTER_ACCEPT:
accept = 1; // ACCEPT
break;
case DOZTFILTER_SUPER_ACCEPT:
accept = 2; // super-ACCEPT
break;
}
if (accept) {
if (cc) {
_membership(cc).pushCredentials(RR,tPtr,RR->node->now(),cc,_config,-1,false);
Packet outp(cc,RR->identity.address(),Packet::VERB_EXT_FRAME);
outp.append(_id);
outp.append((uint8_t)(ccWatch ? 0x1c : 0x08));
macDest.appendTo(outp);
macSource.appendTo(outp);
outp.append((uint16_t)etherType);
outp.append(frameData,ccLength);
outp.compress();
RR->sw->send(tPtr,outp,true);
}
if ((ztDest != ztFinalDest)&&(ztFinalDest)) {
_membership(ztFinalDest).pushCredentials(RR,tPtr,RR->node->now(),ztFinalDest,_config,-1,false);
Packet outp(ztFinalDest,RR->identity.address(),Packet::VERB_EXT_FRAME);
outp.append(_id);
outp.append((uint8_t)0x0a);
macDest.appendTo(outp);
macSource.appendTo(outp);
outp.append((uint16_t)etherType);
outp.append(frameData,frameLen);
outp.compress();
RR->sw->send(tPtr,outp,true);
if (_config.remoteTraceTarget)
RR->t->networkFilter(tPtr,*this,rrl,(c) ? &crrl : (Trace::RuleResultLog *)0,c,sourcePeer->address(),ztDest,macSource,macDest,frameData,frameLen,etherType,vlanId,false,true,0);
return 0; // DROP locally, since we redirected
}
}
if (_config.remoteTraceTarget)
RR->t->networkFilter(tPtr,*this,rrl,(c) ? &crrl : (Trace::RuleResultLog *)0,c,sourcePeer->address(),ztDest,macSource,macDest,frameData,frameLen,etherType,vlanId,false,true,accept);
return accept;
}
bool Network::subscribedToMulticastGroup(const MulticastGroup &mg,bool includeBridgedGroups) const
{
Mutex::Lock _l(_lock);
if (std::binary_search(_myMulticastGroups.begin(),_myMulticastGroups.end(),mg))
return true;
else if (includeBridgedGroups)
return _multicastGroupsBehindMe.contains(mg);
return false;
}
void Network::multicastSubscribe(void *tPtr,const MulticastGroup &mg)
{
Mutex::Lock _l(_lock);
if (!std::binary_search(_myMulticastGroups.begin(),_myMulticastGroups.end(),mg)) {
_myMulticastGroups.insert(std::upper_bound(_myMulticastGroups.begin(),_myMulticastGroups.end(),mg),mg);
_sendUpdatesToMembers(tPtr,&mg);
}
}
void Network::multicastUnsubscribe(const MulticastGroup &mg)
{
Mutex::Lock _l(_lock);
std::vector::iterator i(std::lower_bound(_myMulticastGroups.begin(),_myMulticastGroups.end(),mg));
if ( (i != _myMulticastGroups.end()) && (*i == mg) )
_myMulticastGroups.erase(i);
}
uint64_t Network::handleConfigChunk(void *tPtr,const uint64_t packetId,const Address &source,const Buffer &chunk,unsigned int ptr)
{
if (_destroyed)
return 0;
const unsigned int start = ptr;
ptr += 8; // skip network ID, which is already obviously known
const unsigned int chunkLen = chunk.at(ptr); ptr += 2;
const void *chunkData = chunk.field(ptr,chunkLen); ptr += chunkLen;
NetworkConfig *nc = (NetworkConfig *)0;
uint64_t configUpdateId;
{
Mutex::Lock _l(_lock);
_IncomingConfigChunk *c = (_IncomingConfigChunk *)0;
uint64_t chunkId = 0;
unsigned long totalLength,chunkIndex;
if (ptr < chunk.size()) {
const bool fastPropagate = ((chunk[ptr++] & 0x01) != 0);
configUpdateId = chunk.at(ptr); ptr += 8;
totalLength = chunk.at(ptr); ptr += 4;
chunkIndex = chunk.at(ptr); ptr += 4;
if (((chunkIndex + chunkLen) > totalLength)||(totalLength >= ZT_NETWORKCONFIG_DICT_CAPACITY)) // >= since we need room for a null at the end
return 0;
if ((chunk[ptr] != 1)||(chunk.at(ptr + 1) != ZT_C25519_SIGNATURE_LEN))
return 0;
const uint8_t *sig = reinterpret_cast(chunk.field(ptr + 3,ZT_C25519_SIGNATURE_LEN));
// We can use the signature, which is unique per chunk, to get a per-chunk ID for local deduplication use
for(unsigned int i=0;i<16;++i)
reinterpret_cast(&chunkId)[i & 7] ^= sig[i];
// Find existing or new slot for this update and check if this is a duplicate chunk
for(int i=0;ihaveChunks;++j) {
if (c->haveChunkIds[j] == chunkId)
return 0;
}
break;
} else if ((!c)||(_incomingConfigChunks[i].ts < c->ts)) {
c = &(_incomingConfigChunks[i]);
}
}
// If it's not a duplicate, check chunk signature
const Identity controllerId(RR->topology->getIdentity(tPtr,controller()));
if (!controllerId) // we should always have the controller identity by now, otherwise how would we have queried it the first time?
return 0;
if (!controllerId.verify(chunk.field(start,ptr - start),ptr - start,sig,ZT_C25519_SIGNATURE_LEN))
return 0;
// New properly verified chunks can be flooded "virally" through the network
if (fastPropagate) {
Address *a = (Address *)0;
Membership *m = (Membership *)0;
Hashtable::Iterator i(_memberships);
while (i.next(a,m)) {
if ((*a != source)&&(*a != controller())) {
Packet outp(*a,RR->identity.address(),Packet::VERB_NETWORK_CONFIG);
outp.append(reinterpret_cast(chunk.data()) + start,chunk.size() - start);
RR->sw->send(tPtr,outp,true);
}
}
}
} else if ((source == controller())||(!source)) { // since old chunks aren't signed, only accept from controller itself (or via cluster backplane)
// Legacy support for OK(NETWORK_CONFIG_REQUEST) from older controllers
chunkId = packetId;
configUpdateId = chunkId;
totalLength = chunkLen;
chunkIndex = 0;
if (totalLength >= ZT_NETWORKCONFIG_DICT_CAPACITY)
return 0;
for(int i=0;its))
c = &(_incomingConfigChunks[i]);
}
} else {
// Single-chunk unsigned legacy configs are only allowed from the controller itself
return 0;
}
++c->ts; // newer is higher, that's all we need
if (c->updateId != configUpdateId) {
c->updateId = configUpdateId;
c->haveChunks = 0;
c->haveBytes = 0;
}
if (c->haveChunks >= ZT_NETWORK_MAX_UPDATE_CHUNKS)
return false;
c->haveChunkIds[c->haveChunks++] = chunkId;
memcpy(c->data.unsafeData() + chunkIndex,chunkData,chunkLen);
c->haveBytes += chunkLen;
if (c->haveBytes == totalLength) {
c->data.unsafeData()[c->haveBytes] = (char)0; // ensure null terminated
nc = new NetworkConfig();
try {
if (!nc->fromDictionary(c->data)) {
delete nc;
nc = (NetworkConfig *)0;
}
} catch ( ... ) {
delete nc;
nc = (NetworkConfig *)0;
}
}
}
if (nc) {
this->setConfiguration(tPtr,*nc,true);
delete nc;
return configUpdateId;
} else {
return 0;
}
return 0;
}
int Network::setConfiguration(void *tPtr,const NetworkConfig &nconf,bool saveToDisk)
{
if (_destroyed)
return 0;
// _lock is NOT locked when this is called
try {
if ((nconf.issuedTo != RR->identity.address())||(nconf.networkId != _id))
return 0; // invalid config that is not for us or not for this network
if (_config == nconf)
return 1; // OK config, but duplicate of what we already have
ZT_VirtualNetworkConfig ctmp;
bool oldPortInitialized;
{ // do things that require lock here, but unlock before calling callbacks
Mutex::Lock _l(_lock);
_config = nconf;
_lastConfigUpdate = RR->node->now();
_netconfFailure = NETCONF_FAILURE_NONE;
oldPortInitialized = _portInitialized;
_portInitialized = true;
_externalConfig(&ctmp);
Address *a = (Address *)0;
Membership *m = (Membership *)0;
Hashtable::Iterator i(_memberships);
while (i.next(a,m))
m->resetPushState();
}
_portError = RR->node->configureVirtualNetworkPort(tPtr,_id,&_uPtr,(oldPortInitialized) ? ZT_VIRTUAL_NETWORK_CONFIG_OPERATION_CONFIG_UPDATE : ZT_VIRTUAL_NETWORK_CONFIG_OPERATION_UP,&ctmp);
if (saveToDisk) {
Dictionary *d = new Dictionary();
try {
if (nconf.toDictionary(*d,false)) {
uint64_t tmp[2];
tmp[0] = _id; tmp[1] = 0;
RR->node->stateObjectPut(tPtr,ZT_STATE_OBJECT_NETWORK_CONFIG,tmp,d->data(),d->sizeBytes());
}
} catch ( ... ) {}
delete d;
}
return 2; // OK and configuration has changed
} catch ( ... ) {} // ignore invalid configs
return 0;
}
void Network::requestConfiguration(void *tPtr)
{
if (_destroyed)
return;
/* ZeroTier addresses can't begin with 0xff, so this is used to mark controllerless
* network IDs. Controllerless network IDs only support unicast IPv6 using the 6plane
* addressing scheme and have the following format: 0xffSSSSEEEE000000 where SSSS
* is the 16-bit starting IP port range allowed and EEEE is the 16-bit ending IP port
* range allowed. Remaining digits are reserved for future use and must be zero. */
if ((_id >> 56) == 0xff) {
const uint16_t startPortRange = (uint16_t)((_id >> 40) & 0xffff);
const uint16_t endPortRange = (uint16_t)((_id >> 24) & 0xffff);
if (((_id & 0xffffff) == 0)&&(endPortRange >= startPortRange)) {
NetworkConfig *const nconf = new NetworkConfig();
nconf->networkId = _id;
nconf->timestamp = RR->node->now();
nconf->credentialTimeMaxDelta = ZT_NETWORKCONFIG_DEFAULT_CREDENTIAL_TIME_MAX_MAX_DELTA;
nconf->revision = 1;
nconf->issuedTo = RR->identity.address();
nconf->flags = ZT_NETWORKCONFIG_FLAG_ENABLE_IPV6_NDP_EMULATION;
nconf->mtu = ZT_DEFAULT_MTU;
nconf->multicastLimit = 0;
nconf->staticIpCount = 1;
nconf->ruleCount = 14;
nconf->staticIps[0] = InetAddress::makeIpv66plane(_id,RR->identity.address().toInt());
// Drop everything but IPv6
nconf->rules[0].t = (uint8_t)ZT_NETWORK_RULE_MATCH_ETHERTYPE | 0x80; // NOT
nconf->rules[0].v.etherType = 0x86dd; // IPv6
nconf->rules[1].t = (uint8_t)ZT_NETWORK_RULE_ACTION_DROP;
// Allow ICMPv6
nconf->rules[2].t = (uint8_t)ZT_NETWORK_RULE_MATCH_IP_PROTOCOL;
nconf->rules[2].v.ipProtocol = 0x3a; // ICMPv6
nconf->rules[3].t = (uint8_t)ZT_NETWORK_RULE_ACTION_ACCEPT;
// Allow destination ports within range
nconf->rules[4].t = (uint8_t)ZT_NETWORK_RULE_MATCH_IP_PROTOCOL;
nconf->rules[4].v.ipProtocol = 0x11; // UDP
nconf->rules[5].t = (uint8_t)ZT_NETWORK_RULE_MATCH_IP_PROTOCOL | 0x40; // OR
nconf->rules[5].v.ipProtocol = 0x06; // TCP
nconf->rules[6].t = (uint8_t)ZT_NETWORK_RULE_MATCH_IP_DEST_PORT_RANGE;
nconf->rules[6].v.port[0] = startPortRange;
nconf->rules[6].v.port[1] = endPortRange;
nconf->rules[7].t = (uint8_t)ZT_NETWORK_RULE_ACTION_ACCEPT;
// Allow non-SYN TCP packets to permit non-connection-initiating traffic
nconf->rules[8].t = (uint8_t)ZT_NETWORK_RULE_MATCH_CHARACTERISTICS | 0x80; // NOT
nconf->rules[8].v.characteristics = ZT_RULE_PACKET_CHARACTERISTICS_TCP_SYN;
nconf->rules[9].t = (uint8_t)ZT_NETWORK_RULE_ACTION_ACCEPT;
// Also allow SYN+ACK which are replies to SYN
nconf->rules[10].t = (uint8_t)ZT_NETWORK_RULE_MATCH_CHARACTERISTICS;
nconf->rules[10].v.characteristics = ZT_RULE_PACKET_CHARACTERISTICS_TCP_SYN;
nconf->rules[11].t = (uint8_t)ZT_NETWORK_RULE_MATCH_CHARACTERISTICS;
nconf->rules[11].v.characteristics = ZT_RULE_PACKET_CHARACTERISTICS_TCP_ACK;
nconf->rules[12].t = (uint8_t)ZT_NETWORK_RULE_ACTION_ACCEPT;
nconf->rules[13].t = (uint8_t)ZT_NETWORK_RULE_ACTION_DROP;
nconf->type = ZT_NETWORK_TYPE_PUBLIC;
nconf->name[0] = 'a';
nconf->name[1] = 'd';
nconf->name[2] = 'h';
nconf->name[3] = 'o';
nconf->name[4] = 'c';
nconf->name[5] = '-';
Utils::hex((uint16_t)startPortRange,nconf->name + 6);
nconf->name[10] = '-';
Utils::hex((uint16_t)endPortRange,nconf->name + 11);
nconf->name[15] = (char)0;
this->setConfiguration(tPtr,*nconf,false);
delete nconf;
} else {
this->setNotFound();
}
return;
}
const Address ctrl(controller());
Dictionary rmd;
rmd.add(ZT_NETWORKCONFIG_REQUEST_METADATA_KEY_VERSION,(uint64_t)ZT_NETWORKCONFIG_VERSION);
rmd.add(ZT_NETWORKCONFIG_REQUEST_METADATA_KEY_NODE_VENDOR,(uint64_t)ZT_VENDOR_ZEROTIER);
rmd.add(ZT_NETWORKCONFIG_REQUEST_METADATA_KEY_PROTOCOL_VERSION,(uint64_t)ZT_PROTO_VERSION);
rmd.add(ZT_NETWORKCONFIG_REQUEST_METADATA_KEY_NODE_MAJOR_VERSION,(uint64_t)ZEROTIER_ONE_VERSION_MAJOR);
rmd.add(ZT_NETWORKCONFIG_REQUEST_METADATA_KEY_NODE_MINOR_VERSION,(uint64_t)ZEROTIER_ONE_VERSION_MINOR);
rmd.add(ZT_NETWORKCONFIG_REQUEST_METADATA_KEY_NODE_REVISION,(uint64_t)ZEROTIER_ONE_VERSION_REVISION);
rmd.add(ZT_NETWORKCONFIG_REQUEST_METADATA_KEY_MAX_NETWORK_RULES,(uint64_t)ZT_MAX_NETWORK_RULES);
rmd.add(ZT_NETWORKCONFIG_REQUEST_METADATA_KEY_MAX_NETWORK_CAPABILITIES,(uint64_t)ZT_MAX_NETWORK_CAPABILITIES);
rmd.add(ZT_NETWORKCONFIG_REQUEST_METADATA_KEY_MAX_CAPABILITY_RULES,(uint64_t)ZT_MAX_CAPABILITY_RULES);
rmd.add(ZT_NETWORKCONFIG_REQUEST_METADATA_KEY_MAX_NETWORK_TAGS,(uint64_t)ZT_MAX_NETWORK_TAGS);
rmd.add(ZT_NETWORKCONFIG_REQUEST_METADATA_KEY_FLAGS,(uint64_t)0);
rmd.add(ZT_NETWORKCONFIG_REQUEST_METADATA_KEY_RULES_ENGINE_REV,(uint64_t)ZT_RULES_ENGINE_REVISION);
RR->t->networkConfigRequestSent(tPtr,*this,ctrl);
if (ctrl == RR->identity.address()) {
if (RR->localNetworkController) {
RR->localNetworkController->request(_id,InetAddress(),0xffffffffffffffffULL,RR->identity,rmd);
} else {
this->setNotFound();
}
return;
}
Packet outp(ctrl,RR->identity.address(),Packet::VERB_NETWORK_CONFIG_REQUEST);
outp.append((uint64_t)_id);
const unsigned int rmdSize = rmd.sizeBytes();
outp.append((uint16_t)rmdSize);
outp.append((const void *)rmd.data(),rmdSize);
if (_config) {
outp.append((uint64_t)_config.revision);
outp.append((uint64_t)_config.timestamp);
} else {
outp.append((unsigned char)0,16);
}
outp.compress();
RR->node->expectReplyTo(outp.packetId());
RR->sw->send(tPtr,outp,true);
}
bool Network::gate(void *tPtr,const SharedPtr &peer)
{
const uint64_t now = RR->node->now();
Mutex::Lock _l(_lock);
try {
if (_config) {
Membership *m = _memberships.get(peer->address());
if ( (_config.isPublic()) || ((m)&&(m->isAllowedOnNetwork(_config))) ) {
if (!m)
m = &(_membership(peer->address()));
if (m->multicastLikeGate(now)) {
m->pushCredentials(RR,tPtr,now,peer->address(),_config,-1,false);
_announceMulticastGroupsTo(tPtr,peer->address(),_allMulticastGroups());
}
return true;
}
}
} catch ( ... ) {}
return false;
}
bool Network::recentlyAssociatedWith(const Address &addr)
{
Mutex::Lock _l(_lock);
const Membership *m = _memberships.get(addr);
return ((m)&&(m->recentlyAssociated(RR->node->now())));
}
void Network::clean()
{
const uint64_t now = RR->node->now();
Mutex::Lock _l(_lock);
if (_destroyed)
return;
{
Hashtable< MulticastGroup,uint64_t >::Iterator i(_multicastGroupsBehindMe);
MulticastGroup *mg = (MulticastGroup *)0;
uint64_t *ts = (uint64_t *)0;
while (i.next(mg,ts)) {
if ((now - *ts) > (ZT_MULTICAST_LIKE_EXPIRE * 2))
_multicastGroupsBehindMe.erase(*mg);
}
}
{
Address *a = (Address *)0;
Membership *m = (Membership *)0;
Hashtable::Iterator i(_memberships);
while (i.next(a,m)) {
if (!RR->topology->getPeerNoCache(*a))
_memberships.erase(*a);
else m->clean(now,_config);
}
}
}
void Network::learnBridgeRoute(const MAC &mac,const Address &addr)
{
Mutex::Lock _l(_lock);
_remoteBridgeRoutes[mac] = addr;
// Anti-DOS circuit breaker to prevent nodes from spamming us with absurd numbers of bridge routes
while (_remoteBridgeRoutes.size() > ZT_MAX_BRIDGE_ROUTES) {
Hashtable< Address,unsigned long > counts;
Address maxAddr;
unsigned long maxCount = 0;
MAC *m = (MAC *)0;
Address *a = (Address *)0;
// Find the address responsible for the most entries
{
Hashtable::Iterator i(_remoteBridgeRoutes);
while (i.next(m,a)) {
const unsigned long c = ++counts[*a];
if (c > maxCount) {
maxCount = c;
maxAddr = *a;
}
}
}
// Kill this address from our table, since it's most likely spamming us
{
Hashtable::Iterator i(_remoteBridgeRoutes);
while (i.next(m,a)) {
if (*a == maxAddr)
_remoteBridgeRoutes.erase(*m);
}
}
}
}
void Network::learnBridgedMulticastGroup(void *tPtr,const MulticastGroup &mg,uint64_t now)
{
Mutex::Lock _l(_lock);
const unsigned long tmp = (unsigned long)_multicastGroupsBehindMe.size();
_multicastGroupsBehindMe.set(mg,now);
if (tmp != _multicastGroupsBehindMe.size())
_sendUpdatesToMembers(tPtr,&mg);
}
Membership::AddCredentialResult Network::addCredential(void *tPtr,const CertificateOfMembership &com)
{
if (com.networkId() != _id)
return Membership::ADD_REJECTED;
const Address a(com.issuedTo());
Mutex::Lock _l(_lock);
Membership &m = _membership(a);
const Membership::AddCredentialResult result = m.addCredential(RR,tPtr,_config,com);
if ((result == Membership::ADD_ACCEPTED_NEW)||(result == Membership::ADD_ACCEPTED_REDUNDANT)) {
m.pushCredentials(RR,tPtr,RR->node->now(),a,_config,-1,false);
RR->mc->addCredential(tPtr,com,true);
}
return result;
}
Membership::AddCredentialResult Network::addCredential(void *tPtr,const Address &sentFrom,const Revocation &rev)
{
if (rev.networkId() != _id)
return Membership::ADD_REJECTED;
Mutex::Lock _l(_lock);
Membership &m = _membership(rev.target());
const Membership::AddCredentialResult result = m.addCredential(RR,tPtr,_config,rev);
if ((result == Membership::ADD_ACCEPTED_NEW)&&(rev.fastPropagate())) {
Address *a = (Address *)0;
Membership *m = (Membership *)0;
Hashtable::Iterator i(_memberships);
while (i.next(a,m)) {
if ((*a != sentFrom)&&(*a != rev.signer())) {
Packet outp(*a,RR->identity.address(),Packet::VERB_NETWORK_CREDENTIALS);
outp.append((uint8_t)0x00); // no COM
outp.append((uint16_t)0); // no capabilities
outp.append((uint16_t)0); // no tags
outp.append((uint16_t)1); // one revocation!
rev.serialize(outp);
outp.append((uint16_t)0); // no certificates of ownership
RR->sw->send(tPtr,outp,true);
}
}
}
return result;
}
void Network::destroy()
{
Mutex::Lock _l(_lock);
_destroyed = true;
}
ZT_VirtualNetworkStatus Network::_status() const
{
// assumes _lock is locked
if (_portError)
return ZT_NETWORK_STATUS_PORT_ERROR;
switch(_netconfFailure) {
case NETCONF_FAILURE_ACCESS_DENIED:
return ZT_NETWORK_STATUS_ACCESS_DENIED;
case NETCONF_FAILURE_NOT_FOUND:
return ZT_NETWORK_STATUS_NOT_FOUND;
case NETCONF_FAILURE_NONE:
return ((_config) ? ZT_NETWORK_STATUS_OK : ZT_NETWORK_STATUS_REQUESTING_CONFIGURATION);
default:
return ZT_NETWORK_STATUS_PORT_ERROR;
}
}
void Network::_externalConfig(ZT_VirtualNetworkConfig *ec) const
{
// assumes _lock is locked
ec->nwid = _id;
ec->mac = _mac.toInt();
if (_config)
Utils::scopy(ec->name,sizeof(ec->name),_config.name);
else ec->name[0] = (char)0;
ec->status = _status();
ec->type = (_config) ? (_config.isPrivate() ? ZT_NETWORK_TYPE_PRIVATE : ZT_NETWORK_TYPE_PUBLIC) : ZT_NETWORK_TYPE_PRIVATE;
ec->mtu = (_config) ? _config.mtu : ZT_DEFAULT_MTU;
ec->dhcp = 0;
std::vector ab(_config.activeBridges());
ec->bridge = ((_config.allowPassiveBridging())||(std::find(ab.begin(),ab.end(),RR->identity.address()) != ab.end())) ? 1 : 0;
ec->broadcastEnabled = (_config) ? (_config.enableBroadcast() ? 1 : 0) : 0;
ec->portError = _portError;
ec->netconfRevision = (_config) ? (unsigned long)_config.revision : 0;
ec->assignedAddressCount = 0;
for(unsigned int i=0;iassignedAddresses[i]),&(_config.staticIps[i]),sizeof(struct sockaddr_storage));
++ec->assignedAddressCount;
} else {
memset(&(ec->assignedAddresses[i]),0,sizeof(struct sockaddr_storage));
}
}
ec->routeCount = 0;
for(unsigned int i=0;iroutes[i]),&(_config.routes[i]),sizeof(ZT_VirtualNetworkRoute));
++ec->routeCount;
} else {
memset(&(ec->routes[i]),0,sizeof(ZT_VirtualNetworkRoute));
}
}
}
void Network::_sendUpdatesToMembers(void *tPtr,const MulticastGroup *const newMulticastGroup)
{
// Assumes _lock is locked
const uint64_t now = RR->node->now();
std::vector groups;
if (newMulticastGroup)
groups.push_back(*newMulticastGroup);
else groups = _allMulticastGroups();
if ((newMulticastGroup)||((now - _lastAnnouncedMulticastGroupsUpstream) >= ZT_MULTICAST_ANNOUNCE_PERIOD)) {
if (!newMulticastGroup)
_lastAnnouncedMulticastGroupsUpstream = now;
// Announce multicast groups to upstream peers (roots, etc.) and also send
// them our COM so that MULTICAST_GATHER can be authenticated properly.
const std::vector upstreams(RR->topology->upstreamAddresses());
for(std::vector::const_iterator a(upstreams.begin());a!=upstreams.end();++a) {
if (_config.com) {
Packet outp(*a,RR->identity.address(),Packet::VERB_NETWORK_CREDENTIALS);
_config.com.serialize(outp);
outp.append((uint8_t)0x00);
outp.append((uint16_t)0); // no capabilities
outp.append((uint16_t)0); // no tags
outp.append((uint16_t)0); // no revocations
outp.append((uint16_t)0); // no certificates of ownership
RR->sw->send(tPtr,outp,true);
}
_announceMulticastGroupsTo(tPtr,*a,groups);
}
// Also announce to controller, and send COM to simplify and generalize behavior even though in theory it does not need it
const Address c(controller());
if ( (std::find(upstreams.begin(),upstreams.end(),c) == upstreams.end()) && (!_memberships.contains(c)) ) {
if (_config.com) {
Packet outp(c,RR->identity.address(),Packet::VERB_NETWORK_CREDENTIALS);
_config.com.serialize(outp);
outp.append((uint8_t)0x00);
outp.append((uint16_t)0); // no capabilities
outp.append((uint16_t)0); // no tags
outp.append((uint16_t)0); // no revocations
outp.append((uint16_t)0); // no certificates of ownership
RR->sw->send(tPtr,outp,true);
}
_announceMulticastGroupsTo(tPtr,c,groups);
}
}
// Make sure that all "network anchors" have Membership records so we will
// push multicasts to them.
const std::vector anchors(_config.anchors());
for(std::vector::const_iterator a(anchors.begin());a!=anchors.end();++a)
_membership(*a);
// Send credentials and multicast LIKEs to members, upstreams, and controller
{
Address *a = (Address *)0;
Membership *m = (Membership *)0;
Hashtable::Iterator i(_memberships);
while (i.next(a,m)) {
m->pushCredentials(RR,tPtr,now,*a,_config,-1,false);
if ( ( m->multicastLikeGate(now) || (newMulticastGroup) ) && (m->isAllowedOnNetwork(_config)) )
_announceMulticastGroupsTo(tPtr,*a,groups);
}
}
}
void Network::_announceMulticastGroupsTo(void *tPtr,const Address &peer,const std::vector &allMulticastGroups)
{
// Assumes _lock is locked
Packet outp(peer,RR->identity.address(),Packet::VERB_MULTICAST_LIKE);
for(std::vector::const_iterator mg(allMulticastGroups.begin());mg!=allMulticastGroups.end();++mg) {
if ((outp.size() + 24) >= ZT_PROTO_MAX_PACKET_LENGTH) {
outp.compress();
RR->sw->send(tPtr,outp,true);
outp.reset(peer,RR->identity.address(),Packet::VERB_MULTICAST_LIKE);
}
// network ID, MAC, ADI
outp.append((uint64_t)_id);
mg->mac().appendTo(outp);
outp.append((uint32_t)mg->adi());
}
if (outp.size() > ZT_PROTO_MIN_PACKET_LENGTH) {
outp.compress();
RR->sw->send(tPtr,outp,true);
}
}
std::vector Network::_allMulticastGroups() const
{
// Assumes _lock is locked
std::vector mgs;
mgs.reserve(_myMulticastGroups.size() + _multicastGroupsBehindMe.size() + 1);
mgs.insert(mgs.end(),_myMulticastGroups.begin(),_myMulticastGroups.end());
_multicastGroupsBehindMe.appendKeys(mgs);
if ((_config)&&(_config.enableBroadcast()))
mgs.push_back(Network::BROADCAST);
std::sort(mgs.begin(),mgs.end());
mgs.erase(std::unique(mgs.begin(),mgs.end()),mgs.end());
return mgs;
}
Membership &Network::_membership(const Address &a)
{
// assumes _lock is locked
return _memberships[a];
}
} // namespace ZeroTier