mirror of
https://git.FreeBSD.org/src.git
synced 2024-12-24 11:29:10 +00:00
5cbe92e7e8
Not spotted by: FreeBSD's gcc version 2.95.3 20010315 (release) Spotted by: OpenBSD's gcc version 2.95.3 20010125 (prerelease)
619 lines
16 KiB
C
619 lines
16 KiB
C
/*-
|
|
* Copyright (c) 2001 Brian Somers <brian@Awfulhak.org>
|
|
* All rights reserved.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
* documentation and/or other materials provided with the distribution.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
|
|
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
|
|
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
|
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
|
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
|
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
|
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
|
* SUCH DAMAGE.
|
|
*
|
|
* $FreeBSD$
|
|
*/
|
|
|
|
#include <sys/param.h>
|
|
#include <netinet/in_systm.h>
|
|
#include <netinet/in.h>
|
|
#include <netinet/ip.h>
|
|
#include <sys/socket.h>
|
|
#include <net/route.h>
|
|
#include <net/if.h>
|
|
#include <sys/un.h>
|
|
|
|
#include <stdio.h>
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
#include <termios.h>
|
|
|
|
#include "layer.h"
|
|
#include "defs.h"
|
|
#include "mbuf.h"
|
|
#include "timer.h"
|
|
#include "fsm.h"
|
|
#include "iplist.h"
|
|
#include "throughput.h"
|
|
#include "slcompress.h"
|
|
#include "lqr.h"
|
|
#include "hdlc.h"
|
|
#include "lcp.h"
|
|
#include "ncpaddr.h"
|
|
#include "ip.h"
|
|
#include "ipcp.h"
|
|
#include "ipv6cp.h"
|
|
#include "filter.h"
|
|
#include "descriptor.h"
|
|
#include "ccp.h"
|
|
#include "link.h"
|
|
#include "mp.h"
|
|
#ifndef NORADIUS
|
|
#include "radius.h"
|
|
#endif
|
|
#include "ncp.h"
|
|
#include "bundle.h"
|
|
#include "route.h"
|
|
#include "iface.h"
|
|
#include "log.h"
|
|
#include "proto.h"
|
|
#include "command.h"
|
|
#include "prompt.h"
|
|
#include "async.h"
|
|
#include "physical.h"
|
|
#include "probe.h"
|
|
|
|
|
|
#ifndef NOINET6
|
|
static int ipv6cp_LayerUp(struct fsm *);
|
|
static void ipv6cp_LayerDown(struct fsm *);
|
|
static void ipv6cp_LayerStart(struct fsm *);
|
|
static void ipv6cp_LayerFinish(struct fsm *);
|
|
static void ipv6cp_InitRestartCounter(struct fsm *, int);
|
|
static void ipv6cp_SendConfigReq(struct fsm *);
|
|
static void ipv6cp_SentTerminateReq(struct fsm *);
|
|
static void ipv6cp_SendTerminateAck(struct fsm *, u_char);
|
|
static void ipv6cp_DecodeConfig(struct fsm *, u_char *, int, int,
|
|
struct fsm_decode *);
|
|
|
|
static struct fsm_callbacks ipv6cp_Callbacks = {
|
|
ipv6cp_LayerUp,
|
|
ipv6cp_LayerDown,
|
|
ipv6cp_LayerStart,
|
|
ipv6cp_LayerFinish,
|
|
ipv6cp_InitRestartCounter,
|
|
ipv6cp_SendConfigReq,
|
|
ipv6cp_SentTerminateReq,
|
|
ipv6cp_SendTerminateAck,
|
|
ipv6cp_DecodeConfig,
|
|
fsm_NullRecvResetReq,
|
|
fsm_NullRecvResetAck
|
|
};
|
|
|
|
static u_int32_t
|
|
GenerateToken(void)
|
|
{
|
|
/* Generate random number which will be used as negotiation token */
|
|
randinit();
|
|
|
|
return random() + 1;
|
|
}
|
|
|
|
static int
|
|
ipcp_SetIPv6address(struct ipv6cp *ipv6cp, u_int32_t mytok, u_int32_t histok)
|
|
{
|
|
struct bundle *bundle = ipv6cp->fsm.bundle;
|
|
struct in6_addr myaddr, hisaddr;
|
|
struct ncprange myrange;
|
|
struct sockaddr_storage ssdst, ssgw, ssmask;
|
|
struct sockaddr *sadst, *sagw, *samask;
|
|
|
|
sadst = (struct sockaddr *)&ssdst;
|
|
sagw = (struct sockaddr *)&ssgw;
|
|
samask = (struct sockaddr *)&ssmask;
|
|
|
|
memset(&myaddr, '\0', sizeof myaddr);
|
|
memset(&hisaddr, '\0', sizeof hisaddr);
|
|
|
|
myaddr.s6_addr[0] = 0xfe;
|
|
myaddr.s6_addr[1] = 0x80;
|
|
*(u_int32_t *)(myaddr.s6_addr + 12) = htonl(mytok);
|
|
|
|
hisaddr.s6_addr[0] = 0xfe;
|
|
hisaddr.s6_addr[1] = 0x80;
|
|
*(u_int32_t *)(hisaddr.s6_addr + 12) = htonl(histok);
|
|
|
|
ncpaddr_setip6(&ipv6cp->myaddr, &myaddr);
|
|
ncpaddr_setip6(&ipv6cp->hisaddr, &hisaddr);
|
|
ncprange_sethost(&myrange, &ipv6cp->myaddr);
|
|
|
|
if (!iface_Add(bundle->iface, &bundle->ncp, &myrange, &ipv6cp->hisaddr,
|
|
IFACE_ADD_FIRST|IFACE_FORCE_ADD|IFACE_SYSTEM))
|
|
return 0;
|
|
|
|
if (!Enabled(bundle, OPT_IFACEALIAS))
|
|
iface_Clear(bundle->iface, &bundle->ncp, AF_INET6,
|
|
IFACE_CLEAR_ALIASES|IFACE_SYSTEM);
|
|
|
|
if (bundle->ncp.cfg.sendpipe > 0 || bundle->ncp.cfg.recvpipe > 0) {
|
|
ncprange_getsa(&myrange, &ssgw, &ssmask);
|
|
if (ncpaddr_isset(&ipv6cp->hisaddr))
|
|
ncpaddr_getsa(&ipv6cp->hisaddr, &ssdst);
|
|
else
|
|
sadst = NULL;
|
|
rt_Update(bundle, sadst, sagw, samask);
|
|
}
|
|
|
|
if (Enabled(bundle, OPT_SROUTES))
|
|
route_Change(bundle, bundle->ncp.route, &ipv6cp->myaddr, &ipv6cp->hisaddr);
|
|
|
|
#ifndef NORADIUS
|
|
if (bundle->radius.valid)
|
|
route_Change(bundle, bundle->radius.routes, &ipv6cp->myaddr,
|
|
&ipv6cp->hisaddr);
|
|
#endif
|
|
|
|
return 1; /* Ok */
|
|
}
|
|
|
|
void
|
|
ipv6cp_Init(struct ipv6cp *ipv6cp, struct bundle *bundle, struct link *l,
|
|
const struct fsm_parent *parent)
|
|
{
|
|
static const char * const timer_names[] =
|
|
{"IPV6CP restart", "IPV6CP openmode", "IPV6CP stopped"};
|
|
int n;
|
|
|
|
fsm_Init(&ipv6cp->fsm, "IPV6CP", PROTO_IPV6CP, 1, IPV6CP_MAXCODE, LogIPV6CP,
|
|
bundle, l, parent, &ipv6cp_Callbacks, timer_names);
|
|
|
|
ipv6cp->cfg.fsm.timeout = DEF_FSMRETRY;
|
|
ipv6cp->cfg.fsm.maxreq = DEF_FSMTRIES;
|
|
ipv6cp->cfg.fsm.maxtrm = DEF_FSMTRIES;
|
|
|
|
ipv6cp->my_token = GenerateToken();
|
|
while ((ipv6cp->peer_token = GenerateToken()) == ipv6cp->my_token)
|
|
;
|
|
|
|
if (probe.ipv6_available) {
|
|
n = 100;
|
|
while (n &&
|
|
!ipcp_SetIPv6address(ipv6cp, ipv6cp->my_token, ipv6cp->peer_token)) {
|
|
n--;
|
|
while (n && (ipv6cp->my_token = GenerateToken()) == ipv6cp->peer_token)
|
|
n--;
|
|
}
|
|
}
|
|
|
|
throughput_init(&ipv6cp->throughput, SAMPLE_PERIOD);
|
|
memset(ipv6cp->Queue, '\0', sizeof ipv6cp->Queue);
|
|
ipv6cp_Setup(ipv6cp);
|
|
}
|
|
|
|
void
|
|
ipv6cp_Destroy(struct ipv6cp *ipv6cp)
|
|
{
|
|
throughput_destroy(&ipv6cp->throughput);
|
|
}
|
|
|
|
void
|
|
ipv6cp_Setup(struct ipv6cp *ipv6cp)
|
|
{
|
|
ncpaddr_init(&ipv6cp->myaddr);
|
|
ncpaddr_init(&ipv6cp->hisaddr);
|
|
|
|
ipv6cp->his_reject = 0;
|
|
ipv6cp->my_reject = 0;
|
|
}
|
|
|
|
void
|
|
ipv6cp_SetLink(struct ipv6cp *ipv6cp, struct link *l)
|
|
{
|
|
ipv6cp->fsm.link = l;
|
|
}
|
|
|
|
int
|
|
ipv6cp_Show(struct cmdargs const *arg)
|
|
{
|
|
struct ipv6cp *ipv6cp = &arg->bundle->ncp.ipv6cp;
|
|
|
|
prompt_Printf(arg->prompt, "%s [%s]\n", ipv6cp->fsm.name,
|
|
State2Nam(ipv6cp->fsm.state));
|
|
if (ipv6cp->fsm.state == ST_OPENED) {
|
|
prompt_Printf(arg->prompt, " His side: %s\n",
|
|
ncpaddr_ntoa(&ipv6cp->hisaddr));
|
|
prompt_Printf(arg->prompt, " My side: %s\n",
|
|
ncpaddr_ntoa(&ipv6cp->myaddr));
|
|
prompt_Printf(arg->prompt, " Queued packets: %lu\n",
|
|
(unsigned long)ipv6cp_QueueLen(ipv6cp));
|
|
}
|
|
|
|
prompt_Printf(arg->prompt, "\nDefaults:\n");
|
|
prompt_Printf(arg->prompt, " FSM retry = %us, max %u Config"
|
|
" REQ%s, %u Term REQ%s\n\n", ipv6cp->cfg.fsm.timeout,
|
|
ipv6cp->cfg.fsm.maxreq, ipv6cp->cfg.fsm.maxreq == 1 ? "" : "s",
|
|
ipv6cp->cfg.fsm.maxtrm, ipv6cp->cfg.fsm.maxtrm == 1 ? "" : "s");
|
|
|
|
throughput_disp(&ipv6cp->throughput, arg->prompt);
|
|
|
|
return 0;
|
|
}
|
|
|
|
struct mbuf *
|
|
ipv6cp_Input(struct bundle *bundle, struct link *l, struct mbuf *bp)
|
|
{
|
|
/* Got PROTO_IPV6CP from link */
|
|
m_settype(bp, MB_IPV6CPIN);
|
|
if (bundle_Phase(bundle) == PHASE_NETWORK)
|
|
fsm_Input(&bundle->ncp.ipv6cp.fsm, bp);
|
|
else {
|
|
if (bundle_Phase(bundle) < PHASE_NETWORK)
|
|
log_Printf(LogIPV6CP, "%s: Error: Unexpected IPV6CP in phase %s"
|
|
" (ignored)\n", l->name, bundle_PhaseName(bundle));
|
|
m_freem(bp);
|
|
}
|
|
return NULL;
|
|
}
|
|
|
|
void
|
|
ipv6cp_AddInOctets(struct ipv6cp *ipv6cp, int n)
|
|
{
|
|
throughput_addin(&ipv6cp->throughput, n);
|
|
}
|
|
|
|
void
|
|
ipv6cp_AddOutOctets(struct ipv6cp *ipv6cp, int n)
|
|
{
|
|
throughput_addout(&ipv6cp->throughput, n);
|
|
}
|
|
|
|
void
|
|
ipv6cp_IfaceAddrAdded(struct ipv6cp *ipv6cp, const struct iface_addr *addr)
|
|
{
|
|
}
|
|
|
|
void
|
|
ipv6cp_IfaceAddrDeleted(struct ipv6cp *ipv6cp, const struct iface_addr *addr)
|
|
{
|
|
}
|
|
|
|
int
|
|
ipv6cp_InterfaceUp(struct ipv6cp *ipv6cp)
|
|
{
|
|
if (!ipcp_SetIPv6address(ipv6cp, ipv6cp->my_token, ipv6cp->peer_token)) {
|
|
log_Printf(LogERROR, "ipv6cp_InterfaceUp: unable to set ipv6 address\n");
|
|
return 0;
|
|
}
|
|
|
|
if (!iface_SetFlags(ipv6cp->fsm.bundle->iface->name, IFF_UP)) {
|
|
log_Printf(LogERROR, "ipv6cp_InterfaceUp: Can't set the IFF_UP"
|
|
" flag on %s\n", ipv6cp->fsm.bundle->iface->name);
|
|
return 0;
|
|
}
|
|
|
|
return 1;
|
|
}
|
|
|
|
size_t
|
|
ipv6cp_QueueLen(struct ipv6cp *ipv6cp)
|
|
{
|
|
struct mqueue *q;
|
|
size_t result;
|
|
|
|
result = 0;
|
|
for (q = ipv6cp->Queue; q < ipv6cp->Queue + IPV6CP_QUEUES(ipv6cp); q++)
|
|
result += q->len;
|
|
|
|
return result;
|
|
}
|
|
|
|
int
|
|
ipv6cp_PushPacket(struct ipv6cp *ipv6cp, struct link *l)
|
|
{
|
|
struct bundle *bundle = ipv6cp->fsm.bundle;
|
|
struct mqueue *queue;
|
|
struct mbuf *bp;
|
|
int m_len;
|
|
u_int32_t secs = 0;
|
|
unsigned alivesecs = 0;
|
|
|
|
if (ipv6cp->fsm.state != ST_OPENED)
|
|
return 0;
|
|
|
|
/*
|
|
* If ccp is not open but is required, do nothing.
|
|
*/
|
|
if (l->ccp.fsm.state != ST_OPENED && ccp_Required(&l->ccp)) {
|
|
log_Printf(LogPHASE, "%s: Not transmitting... waiting for CCP\n", l->name);
|
|
return 0;
|
|
}
|
|
|
|
queue = ipv6cp->Queue + IPV6CP_QUEUES(ipv6cp) - 1;
|
|
do {
|
|
if (queue->top) {
|
|
bp = m_dequeue(queue);
|
|
bp = mbuf_Read(bp, &secs, sizeof secs);
|
|
bp = m_pullup(bp);
|
|
m_len = m_length(bp);
|
|
if (!FilterCheck(MBUF_CTOP(bp), AF_INET6, &bundle->filter.alive,
|
|
&alivesecs)) {
|
|
if (secs == 0)
|
|
secs = alivesecs;
|
|
bundle_StartIdleTimer(bundle, secs);
|
|
}
|
|
link_PushPacket(l, bp, bundle, 0, PROTO_IPV6);
|
|
ipv6cp_AddOutOctets(ipv6cp, m_len);
|
|
return 1;
|
|
}
|
|
} while (queue-- != ipv6cp->Queue);
|
|
|
|
return 0;
|
|
}
|
|
|
|
static int
|
|
ipv6cp_LayerUp(struct fsm *fp)
|
|
{
|
|
/* We're now up */
|
|
struct ipv6cp *ipv6cp = fsm2ipv6cp(fp);
|
|
char tbuff[40];
|
|
|
|
log_Printf(LogIPV6CP, "%s: LayerUp.\n", fp->link->name);
|
|
if (!ipv6cp_InterfaceUp(ipv6cp))
|
|
return 0;
|
|
|
|
snprintf(tbuff, sizeof tbuff, "%s", ncpaddr_ntoa(&ipv6cp->myaddr));
|
|
log_Printf(LogIPV6CP, "myaddr %s hisaddr = %s\n",
|
|
tbuff, ncpaddr_ntoa(&ipv6cp->hisaddr));
|
|
|
|
/* XXX: Call radius_Account() and system_Select() */
|
|
|
|
fp->more.reqs = fp->more.naks = fp->more.rejs = ipv6cp->cfg.fsm.maxreq * 3;
|
|
log_DisplayPrompts();
|
|
|
|
return 1;
|
|
}
|
|
|
|
static void
|
|
ipv6cp_LayerDown(struct fsm *fp)
|
|
{
|
|
/* About to come down */
|
|
struct ipv6cp *ipv6cp = fsm2ipv6cp(fp);
|
|
static int recursing;
|
|
char addr[40];
|
|
|
|
if (!recursing++) {
|
|
snprintf(addr, sizeof addr, "%s", ncpaddr_ntoa(&ipv6cp->myaddr));
|
|
log_Printf(LogIPV6CP, "%s: LayerDown: %s\n", fp->link->name, addr);
|
|
|
|
/* XXX: Call radius_Account() and system_Select() */
|
|
|
|
ipv6cp_Setup(ipv6cp);
|
|
}
|
|
recursing--;
|
|
}
|
|
|
|
static void
|
|
ipv6cp_LayerStart(struct fsm *fp)
|
|
{
|
|
/* We're about to start up ! */
|
|
struct ipv6cp *ipv6cp = fsm2ipv6cp(fp);
|
|
|
|
log_Printf(LogIPV6CP, "%s: LayerStart.\n", fp->link->name);
|
|
throughput_start(&ipv6cp->throughput, "IPV6CP throughput",
|
|
Enabled(fp->bundle, OPT_THROUGHPUT));
|
|
fp->more.reqs = fp->more.naks = fp->more.rejs = ipv6cp->cfg.fsm.maxreq * 3;
|
|
ipv6cp->peer_tokenreq = 0;
|
|
}
|
|
|
|
static void
|
|
ipv6cp_LayerFinish(struct fsm *fp)
|
|
{
|
|
/* We're now down */
|
|
struct ipv6cp *ipv6cp = fsm2ipv6cp(fp);
|
|
|
|
log_Printf(LogIPV6CP, "%s: LayerFinish.\n", fp->link->name);
|
|
throughput_stop(&ipv6cp->throughput);
|
|
throughput_log(&ipv6cp->throughput, LogIPV6CP, NULL);
|
|
}
|
|
|
|
static void
|
|
ipv6cp_InitRestartCounter(struct fsm *fp, int what)
|
|
{
|
|
/* Set fsm timer load */
|
|
struct ipv6cp *ipv6cp = fsm2ipv6cp(fp);
|
|
|
|
fp->FsmTimer.load = ipv6cp->cfg.fsm.timeout * SECTICKS;
|
|
switch (what) {
|
|
case FSM_REQ_TIMER:
|
|
fp->restart = ipv6cp->cfg.fsm.maxreq;
|
|
break;
|
|
case FSM_TRM_TIMER:
|
|
fp->restart = ipv6cp->cfg.fsm.maxtrm;
|
|
break;
|
|
default:
|
|
fp->restart = 1;
|
|
break;
|
|
}
|
|
}
|
|
|
|
static void
|
|
ipv6cp_SendConfigReq(struct fsm *fp)
|
|
{
|
|
/* Send config REQ please */
|
|
struct physical *p = link2physical(fp->link);
|
|
struct ipv6cp *ipv6cp = fsm2ipv6cp(fp);
|
|
u_char buff[6];
|
|
struct lcp_opt *o;
|
|
|
|
o = (struct lcp_opt *)buff;
|
|
|
|
if ((p && !physical_IsSync(p)) || !REJECTED(ipv6cp, TY_TOKEN)) {
|
|
memcpy(o->data, &ipv6cp->my_token, 4);
|
|
INC_LCP_OPT(TY_TOKEN, 6, o);
|
|
}
|
|
|
|
fsm_Output(fp, CODE_CONFIGREQ, fp->reqid, buff, (u_char *)o - buff,
|
|
MB_IPV6CPOUT);
|
|
}
|
|
|
|
static void
|
|
ipv6cp_SentTerminateReq(struct fsm *fp)
|
|
{
|
|
/* Term REQ just sent by FSM */
|
|
}
|
|
|
|
static void
|
|
ipv6cp_SendTerminateAck(struct fsm *fp, u_char id)
|
|
{
|
|
/* Send Term ACK please */
|
|
fsm_Output(fp, CODE_TERMACK, id, NULL, 0, MB_IPV6CPOUT);
|
|
}
|
|
|
|
static const char *
|
|
protoname(int proto)
|
|
{
|
|
static const char *cftypes[] = { "TOKEN", "COMPPROTO" };
|
|
|
|
if (proto > 0 && proto <= sizeof cftypes / sizeof *cftypes)
|
|
return cftypes[proto - 1];
|
|
|
|
return NumStr(proto, NULL, 0);
|
|
}
|
|
|
|
static void
|
|
ipv6cp_ValidateToken(struct ipv6cp *ipv6cp, u_int32_t token,
|
|
struct fsm_decode *dec)
|
|
{
|
|
if (token != 0 && token != ipv6cp->my_token)
|
|
ipv6cp->peer_token = token;
|
|
|
|
if (token == ipv6cp->peer_token) {
|
|
*dec->ackend++ = TY_TOKEN;
|
|
*dec->ackend++ = 6;
|
|
memcpy(dec->ackend, &ipv6cp->peer_token, 4);
|
|
dec->ackend += 4;
|
|
} else {
|
|
*dec->nakend++ = TY_TOKEN;
|
|
*dec->nakend++ = 6;
|
|
memcpy(dec->nakend, &ipv6cp->peer_token, 4);
|
|
dec->nakend += 4;
|
|
}
|
|
}
|
|
|
|
static void
|
|
ipv6cp_DecodeConfig(struct fsm *fp, u_char *cp, int plen, int mode_type,
|
|
struct fsm_decode *dec)
|
|
{
|
|
/* Deal with incoming PROTO_IPV6CP */
|
|
struct ipv6cp *ipv6cp = fsm2ipv6cp(fp);
|
|
int type, length, n;
|
|
char tbuff[100];
|
|
u_int32_t token;
|
|
|
|
while (plen >= sizeof(struct fsmconfig)) {
|
|
type = *cp;
|
|
length = cp[1];
|
|
|
|
if (length == 0) {
|
|
log_Printf(LogIPV6CP, "%s: IPV6CP size zero\n", fp->link->name);
|
|
break;
|
|
}
|
|
|
|
snprintf(tbuff, sizeof tbuff, " %s[%d] ", protoname(type), length);
|
|
|
|
switch (type) {
|
|
case TY_TOKEN:
|
|
memcpy(&token, cp + 2, 4);
|
|
log_Printf(LogIPV6CP, "%s 0x%08lx\n", tbuff, (unsigned long)token);
|
|
|
|
switch (mode_type) {
|
|
case MODE_REQ:
|
|
ipv6cp->peer_tokenreq = 1;
|
|
ipv6cp_ValidateToken(ipv6cp, token, dec);
|
|
break;
|
|
|
|
case MODE_NAK:
|
|
if (token == 0) {
|
|
log_Printf(log_IsKept(LogIPV6CP) ? LogIPV6CP : LogPHASE,
|
|
"0x00000000: Unacceptable token!\n");
|
|
fsm_Close(&ipv6cp->fsm);
|
|
} else if (token == ipv6cp->peer_token)
|
|
log_Printf(log_IsKept(LogIPV6CP) ? LogIPV6CP : LogPHASE,
|
|
"0x%08lx: Unacceptable token!\n", (unsigned long)token);
|
|
else if (token != ipv6cp->my_token) {
|
|
n = 100;
|
|
while (n && !ipcp_SetIPv6address(ipv6cp, token, ipv6cp->peer_token)) {
|
|
n--;
|
|
while (n && (token = GenerateToken()) == ipv6cp->peer_token)
|
|
n--;
|
|
}
|
|
|
|
if (n == 0) {
|
|
log_Printf(log_IsKept(LogIPV6CP) ? LogIPV6CP : LogPHASE,
|
|
"0x00000000: Unacceptable token!\n");
|
|
fsm_Close(&ipv6cp->fsm);
|
|
} else {
|
|
log_Printf(LogIPV6CP, "%s changing token: 0x%08lx --> 0x%08lx\n",
|
|
tbuff, (unsigned long)ipv6cp->my_token,
|
|
(unsigned long)token);
|
|
ipv6cp->my_token = token;
|
|
bundle_AdjustFilters(fp->bundle, &ipv6cp->myaddr, NULL);
|
|
}
|
|
}
|
|
break;
|
|
|
|
case MODE_REJ:
|
|
ipv6cp->his_reject |= (1 << type);
|
|
break;
|
|
}
|
|
break;
|
|
|
|
default:
|
|
if (mode_type != MODE_NOP) {
|
|
ipv6cp->my_reject |= (1 << type);
|
|
memcpy(dec->rejend, cp, length);
|
|
dec->rejend += length;
|
|
}
|
|
break;
|
|
}
|
|
plen -= length;
|
|
cp += length;
|
|
}
|
|
|
|
if (mode_type != MODE_NOP) {
|
|
if (mode_type == MODE_REQ && !ipv6cp->peer_tokenreq) {
|
|
if (dec->rejend == dec->rej && dec->nakend == dec->nak) {
|
|
/*
|
|
* Pretend the peer has requested a TOKEN.
|
|
* We do this to ensure that we only send one NAK if the only
|
|
* reason for the NAK is because the peer isn't sending a
|
|
* TY_TOKEN REQ. This stops us from repeatedly trying to tell
|
|
* the peer that we have to have an IP address on their end.
|
|
*/
|
|
ipv6cp->peer_tokenreq = 1;
|
|
}
|
|
ipv6cp_ValidateToken(ipv6cp, 0, dec);
|
|
}
|
|
if (dec->rejend != dec->rej) {
|
|
/* rejects are preferred */
|
|
dec->ackend = dec->ack;
|
|
dec->nakend = dec->nak;
|
|
} else if (dec->nakend != dec->nak)
|
|
/* then NAKs */
|
|
dec->ackend = dec->ack;
|
|
}
|
|
}
|
|
#endif
|