1
0
mirror of https://git.FreeBSD.org/src.git synced 2024-12-19 10:53:58 +00:00
freebsd/usr.sbin/ppp/ipcp.c
Brian Somers ab2de065b2 o Obsolete the undocumented ``set weight'' command.
o If we're using RADIUS and the RADIUS mtu is less than our
  peers mru/mrru, reduce our mtu to this value for NetBSD too.
o Make struct throughput's sample period dynamic and tweak the ppp
  version number to reflect the extra stuff being passed through
  the local domain socket as a result (MP mode).
o Measure the current throughput based on the number of samples actually
  taken rather than on the full sample period.
o Keep the throughput statisics persistent while being passed to
  another ppp invocation through the local domain socket.
o When showing throughput statistics after the timer has stopped, use
  the stopped time for overall calculations, not the current time.
  Also show the stopped time and how long the current throughput has
  been sampled for.
o Use time() consistently in throughput.c
o Tighten up the ``show bundle'' output.
o Introduce the ``set bandwidth'' command.
o Rewrite the ``set autoload'' command.  It now takes three arguments
  and works based on a rolling bundle throughput average compared against
  the theoretical bundle bandwidth over a given period (read: it's now
  functional).
1999-08-05 10:32:16 +00:00

1232 lines
35 KiB
C

/*
* PPP IP Control Protocol (IPCP) Module
*
* Written by Toshiharu OHNO (tony-o@iij.ad.jp)
*
* Copyright (C) 1993, Internet Initiative Japan, Inc. All rights reserverd.
*
* Redistribution and use in source and binary forms are permitted
* provided that the above copyright notice and this paragraph are
* duplicated in all such forms and that any documentation,
* advertising materials, and other materials related to such
* distribution and use acknowledge that the software was developed
* by the Internet Initiative Japan, Inc. The name of the
* IIJ may not be used to endorse or promote products derived
* from this software without specific prior written permission.
* THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
* WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE.
*
* $Id: ipcp.c,v 1.81 1999/06/08 20:12:06 brian Exp $
*
* TODO:
* o Support IPADDRS properly
* o Validate the length in IpcpDecodeConfig
*/
#include <sys/param.h>
#include <netinet/in_systm.h>
#include <netinet/in.h>
#include <netinet/ip.h>
#include <arpa/inet.h>
#include <sys/socket.h>
#include <net/route.h>
#include <netdb.h>
#include <sys/un.h>
#include <errno.h>
#include <fcntl.h>
#include <resolv.h>
#include <stdlib.h>
#include <string.h>
#include <termios.h>
#include <unistd.h>
#ifndef NOALIAS
#ifdef __FreeBSD__
#include <alias.h>
#else
#include "alias.h"
#endif
#endif
#include "layer.h"
#include "ua.h"
#include "defs.h"
#include "command.h"
#include "mbuf.h"
#include "log.h"
#include "timer.h"
#include "fsm.h"
#include "proto.h"
#include "lcp.h"
#include "iplist.h"
#include "throughput.h"
#include "slcompress.h"
#include "lqr.h"
#include "hdlc.h"
#include "ipcp.h"
#include "filter.h"
#include "descriptor.h"
#include "vjcomp.h"
#include "async.h"
#include "ccp.h"
#include "link.h"
#include "physical.h"
#include "mp.h"
#ifndef NORADIUS
#include "radius.h"
#endif
#include "bundle.h"
#include "id.h"
#include "arp.h"
#include "systems.h"
#include "prompt.h"
#include "route.h"
#include "iface.h"
#undef REJECTED
#define REJECTED(p, x) ((p)->peer_reject & (1<<(x)))
#define issep(ch) ((ch) == ' ' || (ch) == '\t')
#define isip(ch) (((ch) >= '0' && (ch) <= '9') || (ch) == '.')
struct compreq {
u_short proto;
u_char slots;
u_char compcid;
};
static int IpcpLayerUp(struct fsm *);
static void IpcpLayerDown(struct fsm *);
static void IpcpLayerStart(struct fsm *);
static void IpcpLayerFinish(struct fsm *);
static void IpcpInitRestartCounter(struct fsm *, int);
static void IpcpSendConfigReq(struct fsm *);
static void IpcpSentTerminateReq(struct fsm *);
static void IpcpSendTerminateAck(struct fsm *, u_char);
static void IpcpDecodeConfig(struct fsm *, u_char *, int, int,
struct fsm_decode *);
static struct fsm_callbacks ipcp_Callbacks = {
IpcpLayerUp,
IpcpLayerDown,
IpcpLayerStart,
IpcpLayerFinish,
IpcpInitRestartCounter,
IpcpSendConfigReq,
IpcpSentTerminateReq,
IpcpSendTerminateAck,
IpcpDecodeConfig,
fsm_NullRecvResetReq,
fsm_NullRecvResetAck
};
static const char *cftypes[] = {
/* Check out the latest ``Assigned numbers'' rfc (rfc1700.txt) */
"???",
"IPADDRS", /* 1: IP-Addresses */ /* deprecated */
"COMPPROTO", /* 2: IP-Compression-Protocol */
"IPADDR", /* 3: IP-Address */
};
#define NCFTYPES (sizeof cftypes/sizeof cftypes[0])
static const char *cftypes128[] = {
/* Check out the latest ``Assigned numbers'' rfc (rfc1700.txt) */
"???",
"PRIDNS", /* 129: Primary DNS Server Address */
"PRINBNS", /* 130: Primary NBNS Server Address */
"SECDNS", /* 131: Secondary DNS Server Address */
"SECNBNS", /* 132: Secondary NBNS Server Address */
};
#define NCFTYPES128 (sizeof cftypes128/sizeof cftypes128[0])
void
ipcp_AddInOctets(struct ipcp *ipcp, int n)
{
throughput_addin(&ipcp->throughput, n);
}
void
ipcp_AddOutOctets(struct ipcp *ipcp, int n)
{
throughput_addout(&ipcp->throughput, n);
}
static void
getdns(struct ipcp *ipcp, struct in_addr addr[2])
{
FILE *fp;
addr[0].s_addr = addr[1].s_addr = INADDR_ANY;
if ((fp = fopen(_PATH_RESCONF, "r")) != NULL) {
char buf[LINE_LEN], *cp, *end;
int n;
n = 0;
buf[sizeof buf - 1] = '\0';
while (fgets(buf, sizeof buf - 1, fp)) {
if (!strncmp(buf, "nameserver", 10) && issep(buf[10])) {
for (cp = buf + 11; issep(*cp); cp++)
;
for (end = cp; isip(*end); end++)
;
*end = '\0';
if (inet_aton(cp, addr+n) && ++n == 2)
break;
}
}
if (n == 1)
addr[1] = addr[0];
fclose(fp);
}
}
static int
setdns(struct ipcp *ipcp, struct in_addr addr[2])
{
FILE *fp;
char wbuf[LINE_LEN + 54];
int wlen;
if (addr[0].s_addr == INADDR_ANY || addr[1].s_addr == INADDR_ANY) {
struct in_addr old[2];
getdns(ipcp, old);
if (addr[0].s_addr == INADDR_ANY)
addr[0] = old[0];
if (addr[1].s_addr == INADDR_ANY)
addr[1] = old[1];
}
if (addr[0].s_addr == INADDR_ANY && addr[1].s_addr == INADDR_ANY) {
log_Printf(LogWARN, "%s not modified: All nameservers NAKd\n",
_PATH_RESCONF);
return 0;
}
wlen = 0;
if ((fp = fopen(_PATH_RESCONF, "r")) != NULL) {
char buf[LINE_LEN];
int len;
buf[sizeof buf - 1] = '\0';
while (fgets(buf, sizeof buf - 1, fp)) {
if (strncmp(buf, "nameserver", 10) || !issep(buf[10])) {
len = strlen(buf);
if (len > sizeof wbuf - wlen) {
log_Printf(LogWARN, "%s: Can only cope with max file size %d\n",
_PATH_RESCONF, LINE_LEN);
fclose(fp);
return 0;
}
memcpy(wbuf + wlen, buf, len);
wlen += len;
}
}
fclose(fp);
}
if (addr[0].s_addr != INADDR_ANY) {
snprintf(wbuf + wlen, sizeof wbuf - wlen, "nameserver %s\n",
inet_ntoa(addr[0]));
log_Printf(LogIPCP, "Primary nameserver set to %s", wbuf + wlen + 11);
wlen += strlen(wbuf + wlen);
}
if (addr[1].s_addr != INADDR_ANY && addr[1].s_addr != addr[0].s_addr) {
snprintf(wbuf + wlen, sizeof wbuf - wlen, "nameserver %s\n",
inet_ntoa(addr[1]));
log_Printf(LogIPCP, "Secondary nameserver set to %s", wbuf + wlen + 11);
wlen += strlen(wbuf + wlen);
}
if (wlen) {
int fd;
if ((fd = ID0open(_PATH_RESCONF, O_WRONLY|O_CREAT, 0644)) != -1) {
if (write(fd, wbuf, wlen) != wlen) {
log_Printf(LogERROR, "setdns: write(): %s\n", strerror(errno));
close(fd);
return 0;
}
if (ftruncate(fd, wlen) == -1) {
log_Printf(LogERROR, "setdns: truncate(): %s\n", strerror(errno));
close(fd);
return 0;
}
close(fd);
} else {
log_Printf(LogERROR, "setdns: open(): %s\n", strerror(errno));
return 0;
}
}
return 1;
}
int
ipcp_Show(struct cmdargs const *arg)
{
struct ipcp *ipcp = &arg->bundle->ncp.ipcp;
prompt_Printf(arg->prompt, "%s [%s]\n", ipcp->fsm.name,
State2Nam(ipcp->fsm.state));
if (ipcp->fsm.state == ST_OPENED) {
prompt_Printf(arg->prompt, " His side: %s, %s\n",
inet_ntoa(ipcp->peer_ip), vj2asc(ipcp->peer_compproto));
prompt_Printf(arg->prompt, " My side: %s, %s\n",
inet_ntoa(ipcp->my_ip), vj2asc(ipcp->my_compproto));
}
if (ipcp->route) {
prompt_Printf(arg->prompt, "\n");
route_ShowSticky(arg->prompt, ipcp->route, "Sticky routes", 1);
}
prompt_Printf(arg->prompt, "\nDefaults:\n");
prompt_Printf(arg->prompt, " FSM retry = %us, max %u Config"
" REQ%s, %u Term REQ%s\n", ipcp->cfg.fsm.timeout,
ipcp->cfg.fsm.maxreq, ipcp->cfg.fsm.maxreq == 1 ? "" : "s",
ipcp->cfg.fsm.maxtrm, ipcp->cfg.fsm.maxtrm == 1 ? "" : "s");
prompt_Printf(arg->prompt, " My Address: %s/%d",
inet_ntoa(ipcp->cfg.my_range.ipaddr), ipcp->cfg.my_range.width);
prompt_Printf(arg->prompt, ", netmask %s\n", inet_ntoa(ipcp->cfg.netmask));
if (ipcp->cfg.HaveTriggerAddress)
prompt_Printf(arg->prompt, " Trigger address: %s\n",
inet_ntoa(ipcp->cfg.TriggerAddress));
prompt_Printf(arg->prompt, " VJ compression: %s (%d slots %s slot "
"compression)\n", command_ShowNegval(ipcp->cfg.vj.neg),
ipcp->cfg.vj.slots, ipcp->cfg.vj.slotcomp ? "with" : "without");
if (iplist_isvalid(&ipcp->cfg.peer_list))
prompt_Printf(arg->prompt, " His Address: %s\n",
ipcp->cfg.peer_list.src);
else
prompt_Printf(arg->prompt, " His Address: %s/%d\n",
inet_ntoa(ipcp->cfg.peer_range.ipaddr),
ipcp->cfg.peer_range.width);
prompt_Printf(arg->prompt, " DNS: %s, ",
inet_ntoa(ipcp->cfg.ns.dns[0]));
prompt_Printf(arg->prompt, "%s, %s\n", inet_ntoa(ipcp->cfg.ns.dns[1]),
command_ShowNegval(ipcp->cfg.ns.dns_neg));
prompt_Printf(arg->prompt, " NetBIOS NS: %s, ",
inet_ntoa(ipcp->cfg.ns.nbns[0]));
prompt_Printf(arg->prompt, "%s\n", inet_ntoa(ipcp->cfg.ns.nbns[1]));
prompt_Printf(arg->prompt, "\n");
throughput_disp(&ipcp->throughput, arg->prompt);
return 0;
}
int
ipcp_vjset(struct cmdargs const *arg)
{
if (arg->argc != arg->argn+2)
return -1;
if (!strcasecmp(arg->argv[arg->argn], "slots")) {
int slots;
slots = atoi(arg->argv[arg->argn+1]);
if (slots < 4 || slots > 16)
return 1;
arg->bundle->ncp.ipcp.cfg.vj.slots = slots;
return 0;
} else if (!strcasecmp(arg->argv[arg->argn], "slotcomp")) {
if (!strcasecmp(arg->argv[arg->argn+1], "on"))
arg->bundle->ncp.ipcp.cfg.vj.slotcomp = 1;
else if (!strcasecmp(arg->argv[arg->argn+1], "off"))
arg->bundle->ncp.ipcp.cfg.vj.slotcomp = 0;
else
return 2;
return 0;
}
return -1;
}
void
ipcp_Init(struct ipcp *ipcp, struct bundle *bundle, struct link *l,
const struct fsm_parent *parent)
{
struct hostent *hp;
char name[MAXHOSTNAMELEN];
static const char *timer_names[] =
{"IPCP restart", "IPCP openmode", "IPCP stopped"};
fsm_Init(&ipcp->fsm, "IPCP", PROTO_IPCP, 1, IPCP_MAXCODE, LogIPCP,
bundle, l, parent, &ipcp_Callbacks, timer_names);
ipcp->route = NULL;
ipcp->cfg.vj.slots = DEF_VJ_STATES;
ipcp->cfg.vj.slotcomp = 1;
memset(&ipcp->cfg.my_range, '\0', sizeof ipcp->cfg.my_range);
if (gethostname(name, sizeof name) == 0) {
hp = gethostbyname(name);
if (hp && hp->h_addrtype == AF_INET)
memcpy(&ipcp->cfg.my_range.ipaddr.s_addr, hp->h_addr, hp->h_length);
}
ipcp->cfg.netmask.s_addr = INADDR_ANY;
memset(&ipcp->cfg.peer_range, '\0', sizeof ipcp->cfg.peer_range);
iplist_setsrc(&ipcp->cfg.peer_list, "");
ipcp->cfg.HaveTriggerAddress = 0;
ipcp->cfg.ns.dns[0].s_addr = INADDR_ANY;
ipcp->cfg.ns.dns[1].s_addr = INADDR_ANY;
ipcp->cfg.ns.dns_neg = 0;
ipcp->cfg.ns.nbns[0].s_addr = INADDR_ANY;
ipcp->cfg.ns.nbns[1].s_addr = INADDR_ANY;
ipcp->cfg.fsm.timeout = DEF_FSMRETRY;
ipcp->cfg.fsm.maxreq = DEF_FSMTRIES;
ipcp->cfg.fsm.maxtrm = DEF_FSMTRIES;
ipcp->cfg.vj.neg = NEG_ENABLED|NEG_ACCEPTED;
memset(&ipcp->vj, '\0', sizeof ipcp->vj);
throughput_init(&ipcp->throughput, SAMPLE_PERIOD);
memset(ipcp->Queue, '\0', sizeof ipcp->Queue);
ipcp_Setup(ipcp, INADDR_NONE);
}
void
ipcp_SetLink(struct ipcp *ipcp, struct link *l)
{
ipcp->fsm.link = l;
}
void
ipcp_Setup(struct ipcp *ipcp, u_int32_t mask)
{
struct iface *iface = ipcp->fsm.bundle->iface;
int pos, n;
ipcp->fsm.open_mode = 0;
ipcp->ifmask.s_addr = mask == INADDR_NONE ? ipcp->cfg.netmask.s_addr : mask;
if (iplist_isvalid(&ipcp->cfg.peer_list)) {
/* Try to give the peer a previously configured IP address */
for (n = 0; n < iface->in_addrs; n++) {
pos = iplist_ip2pos(&ipcp->cfg.peer_list, iface->in_addr[n].brd);
if (pos != -1) {
ipcp->cfg.peer_range.ipaddr =
iplist_setcurpos(&ipcp->cfg.peer_list, pos);
break;
}
}
if (n == iface->in_addrs)
/* Ok, so none of 'em fit.... pick a random one */
ipcp->cfg.peer_range.ipaddr = iplist_setrandpos(&ipcp->cfg.peer_list);
ipcp->cfg.peer_range.mask.s_addr = INADDR_BROADCAST;
ipcp->cfg.peer_range.width = 32;
}
ipcp->heis1172 = 0;
ipcp->peer_ip = ipcp->cfg.peer_range.ipaddr;
ipcp->peer_compproto = 0;
if (ipcp->cfg.HaveTriggerAddress) {
/*
* Some implementations of PPP require that we send a
* *special* value as our address, even though the rfc specifies
* full negotiation (e.g. "0.0.0.0" or Not "0.0.0.0").
*/
ipcp->my_ip = ipcp->cfg.TriggerAddress;
log_Printf(LogIPCP, "Using trigger address %s\n",
inet_ntoa(ipcp->cfg.TriggerAddress));
} else {
/*
* Otherwise, if we've used an IP number before and it's still within
* the network specified on the ``set ifaddr'' line, we really
* want to keep that IP number so that we can keep any existing
* connections that are bound to that IP (assuming we're not
* ``iface-alias''ing).
*/
for (n = 0; n < iface->in_addrs; n++)
if ((iface->in_addr[n].ifa.s_addr & ipcp->cfg.my_range.mask.s_addr) ==
(ipcp->cfg.my_range.ipaddr.s_addr & ipcp->cfg.my_range.mask.s_addr)) {
ipcp->my_ip = iface->in_addr[n].ifa;
break;
}
if (n == iface->in_addrs)
ipcp->my_ip = ipcp->cfg.my_range.ipaddr;
}
if (IsEnabled(ipcp->cfg.vj.neg)
#ifndef NORADIUS
|| (ipcp->fsm.bundle->radius.valid && ipcp->fsm.bundle->radius.vj)
#endif
)
ipcp->my_compproto = (PROTO_VJCOMP << 16) +
((ipcp->cfg.vj.slots - 1) << 8) +
ipcp->cfg.vj.slotcomp;
else
ipcp->my_compproto = 0;
sl_compress_init(&ipcp->vj.cslc, ipcp->cfg.vj.slots - 1);
ipcp->peer_reject = 0;
ipcp->my_reject = 0;
}
static int
ipcp_doproxyall(struct bundle *bundle,
int (*proxyfun)(struct bundle *, struct in_addr, int), int s)
{
int n, ret;
struct sticky_route *rp;
struct in_addr addr;
struct ipcp *ipcp;
ipcp = &bundle->ncp.ipcp;
for (rp = ipcp->route; rp != NULL; rp = rp->next) {
if (rp->mask.s_addr == INADDR_BROADCAST)
continue;
n = ntohl(INADDR_BROADCAST) - ntohl(rp->mask.s_addr) - 1;
if (n > 0 && n <= 254 && rp->dst.s_addr != INADDR_ANY) {
addr = rp->dst;
while (n--) {
addr.s_addr = htonl(ntohl(addr.s_addr) + 1);
log_Printf(LogDEBUG, "ipcp_doproxyall: %s\n", inet_ntoa(addr));
ret = (*proxyfun)(bundle, addr, s);
if (!ret)
return ret;
}
}
}
return 0;
}
static int
ipcp_SetIPaddress(struct bundle *bundle, struct in_addr myaddr,
struct in_addr hisaddr, int silent)
{
struct in_addr mask, oaddr, none = { INADDR_ANY };
mask = addr2mask(myaddr);
if (bundle->ncp.ipcp.ifmask.s_addr != INADDR_ANY &&
(bundle->ncp.ipcp.ifmask.s_addr & mask.s_addr) == mask.s_addr)
mask.s_addr = bundle->ncp.ipcp.ifmask.s_addr;
oaddr.s_addr = bundle->iface->in_addrs ?
bundle->iface->in_addr[0].ifa.s_addr : INADDR_ANY;
if (!iface_inAdd(bundle->iface, myaddr, mask, hisaddr,
IFACE_ADD_FIRST|IFACE_FORCE_ADD))
return -1;
if (!Enabled(bundle, OPT_IFACEALIAS) && bundle->iface->in_addrs > 1
&& myaddr.s_addr != oaddr.s_addr)
/* Nuke the old one */
iface_inDelete(bundle->iface, oaddr);
if (bundle->ncp.ipcp.cfg.sendpipe > 0 || bundle->ncp.ipcp.cfg.recvpipe > 0)
bundle_SetRoute(bundle, RTM_CHANGE, hisaddr, myaddr, none, 0, 0);
if (Enabled(bundle, OPT_SROUTES))
route_Change(bundle, bundle->ncp.ipcp.route, myaddr, hisaddr);
#ifndef NORADIUS
if (bundle->radius.valid)
route_Change(bundle, bundle->radius.routes, myaddr, hisaddr);
#endif
if (Enabled(bundle, OPT_PROXY) || Enabled(bundle, OPT_PROXYALL)) {
int s = ID0socket(AF_INET, SOCK_DGRAM, 0);
if (s < 0)
log_Printf(LogERROR, "ipcp_SetIPaddress: socket(): %s\n",
strerror(errno));
else {
if (Enabled(bundle, OPT_PROXYALL))
ipcp_doproxyall(bundle, arp_SetProxy, s);
else if (Enabled(bundle, OPT_PROXY))
arp_SetProxy(bundle, hisaddr, s);
close(s);
}
}
return 0;
}
static struct in_addr
ChooseHisAddr(struct bundle *bundle, struct in_addr gw)
{
struct in_addr try;
u_long f;
for (f = 0; f < bundle->ncp.ipcp.cfg.peer_list.nItems; f++) {
try = iplist_next(&bundle->ncp.ipcp.cfg.peer_list);
log_Printf(LogDEBUG, "ChooseHisAddr: Check item %ld (%s)\n",
f, inet_ntoa(try));
if (ipcp_SetIPaddress(bundle, gw, try, 1) == 0) {
log_Printf(LogIPCP, "Selected IP address %s\n", inet_ntoa(try));
break;
}
}
if (f == bundle->ncp.ipcp.cfg.peer_list.nItems) {
log_Printf(LogDEBUG, "ChooseHisAddr: All addresses in use !\n");
try.s_addr = INADDR_ANY;
}
return try;
}
static void
IpcpInitRestartCounter(struct fsm *fp, int what)
{
/* Set fsm timer load */
struct ipcp *ipcp = fsm2ipcp(fp);
fp->FsmTimer.load = ipcp->cfg.fsm.timeout * SECTICKS;
switch (what) {
case FSM_REQ_TIMER:
fp->restart = ipcp->cfg.fsm.maxreq;
break;
case FSM_TRM_TIMER:
fp->restart = ipcp->cfg.fsm.maxtrm;
break;
default:
fp->restart = 1;
break;
}
}
static void
IpcpSendConfigReq(struct fsm *fp)
{
/* Send config REQ please */
struct physical *p = link2physical(fp->link);
struct ipcp *ipcp = fsm2ipcp(fp);
u_char buff[24];
struct lcp_opt *o;
o = (struct lcp_opt *)buff;
if ((p && !physical_IsSync(p)) || !REJECTED(ipcp, TY_IPADDR)) {
memcpy(o->data, &ipcp->my_ip.s_addr, 4);
INC_LCP_OPT(TY_IPADDR, 6, o);
}
if (ipcp->my_compproto && !REJECTED(ipcp, TY_COMPPROTO)) {
if (ipcp->heis1172) {
u_int16_t proto = PROTO_VJCOMP;
ua_htons(&proto, o->data);
INC_LCP_OPT(TY_COMPPROTO, 4, o);
} else {
struct compreq req;
req.proto = htons(ipcp->my_compproto >> 16);
req.slots = (ipcp->my_compproto >> 8) & 255;
req.compcid = ipcp->my_compproto & 1;
memcpy(o->data, &req, 4);
INC_LCP_OPT(TY_COMPPROTO, 6, o);
}
}
if (IsEnabled(ipcp->cfg.ns.dns_neg) &&
!REJECTED(ipcp, TY_PRIMARY_DNS - TY_ADJUST_NS) &&
!REJECTED(ipcp, TY_SECONDARY_DNS - TY_ADJUST_NS)) {
struct in_addr dns[2];
getdns(ipcp, dns);
memcpy(o->data, &dns[0].s_addr, 4);
INC_LCP_OPT(TY_PRIMARY_DNS, 6, o);
memcpy(o->data, &dns[1].s_addr, 4);
INC_LCP_OPT(TY_SECONDARY_DNS, 6, o);
}
fsm_Output(fp, CODE_CONFIGREQ, fp->reqid, buff, (u_char *)o - buff,
MB_IPCPOUT);
}
static void
IpcpSentTerminateReq(struct fsm *fp)
{
/* Term REQ just sent by FSM */
}
static void
IpcpSendTerminateAck(struct fsm *fp, u_char id)
{
/* Send Term ACK please */
fsm_Output(fp, CODE_TERMACK, id, NULL, 0, MB_IPCPOUT);
}
static void
IpcpLayerStart(struct fsm *fp)
{
/* We're about to start up ! */
struct ipcp *ipcp = fsm2ipcp(fp);
log_Printf(LogIPCP, "%s: LayerStart.\n", fp->link->name);
throughput_start(&ipcp->throughput, "IPCP throughput",
Enabled(fp->bundle, OPT_THROUGHPUT));
fp->more.reqs = fp->more.naks = fp->more.rejs = ipcp->cfg.fsm.maxreq * 3;
}
static void
IpcpLayerFinish(struct fsm *fp)
{
/* We're now down */
struct ipcp *ipcp = fsm2ipcp(fp);
log_Printf(LogIPCP, "%s: LayerFinish.\n", fp->link->name);
throughput_stop(&ipcp->throughput);
throughput_log(&ipcp->throughput, LogIPCP, NULL);
}
void
ipcp_CleanInterface(struct ipcp *ipcp)
{
struct iface *iface = ipcp->fsm.bundle->iface;
route_Clean(ipcp->fsm.bundle, ipcp->route);
if (iface->in_addrs && (Enabled(ipcp->fsm.bundle, OPT_PROXY) ||
Enabled(ipcp->fsm.bundle, OPT_PROXYALL))) {
int s = ID0socket(AF_INET, SOCK_DGRAM, 0);
if (s < 0)
log_Printf(LogERROR, "ipcp_CleanInterface: socket: %s\n",
strerror(errno));
else {
if (Enabled(ipcp->fsm.bundle, OPT_PROXYALL))
ipcp_doproxyall(ipcp->fsm.bundle, arp_ClearProxy, s);
else if (Enabled(ipcp->fsm.bundle, OPT_PROXY))
arp_ClearProxy(ipcp->fsm.bundle, iface->in_addr[0].brd, s);
close(s);
}
}
iface_inClear(ipcp->fsm.bundle->iface, IFACE_CLEAR_ALL);
}
static void
IpcpLayerDown(struct fsm *fp)
{
/* About to come down */
static int recursing;
struct ipcp *ipcp = fsm2ipcp(fp);
const char *s;
if (!recursing++) {
if (ipcp->fsm.bundle->iface->in_addrs)
s = inet_ntoa(ipcp->fsm.bundle->iface->in_addr[0].ifa);
else
s = "Interface configuration error !";
log_Printf(LogIPCP, "%s: LayerDown: %s\n", fp->link->name, s);
/*
* XXX this stuff should really live in the FSM. Our config should
* associate executable sections in files with events.
*/
if (system_Select(fp->bundle, s, LINKDOWNFILE, NULL, NULL) < 0) {
if (bundle_GetLabel(fp->bundle)) {
if (system_Select(fp->bundle, bundle_GetLabel(fp->bundle),
LINKDOWNFILE, NULL, NULL) < 0)
system_Select(fp->bundle, "MYADDR", LINKDOWNFILE, NULL, NULL);
} else
system_Select(fp->bundle, "MYADDR", LINKDOWNFILE, NULL, NULL);
}
ipcp_Setup(ipcp, INADDR_NONE);
}
recursing--;
}
int
ipcp_InterfaceUp(struct ipcp *ipcp)
{
if (ipcp_SetIPaddress(ipcp->fsm.bundle, ipcp->my_ip, ipcp->peer_ip, 0) < 0) {
log_Printf(LogERROR, "ipcp_InterfaceUp: unable to set ip address\n");
return 0;
}
#ifndef NOALIAS
if (ipcp->fsm.bundle->AliasEnabled)
PacketAliasSetAddress(ipcp->my_ip);
#endif
return 1;
}
static int
IpcpLayerUp(struct fsm *fp)
{
/* We're now up */
struct ipcp *ipcp = fsm2ipcp(fp);
char tbuff[16];
log_Printf(LogIPCP, "%s: LayerUp.\n", fp->link->name);
snprintf(tbuff, sizeof tbuff, "%s", inet_ntoa(ipcp->my_ip));
log_Printf(LogIPCP, "myaddr %s hisaddr = %s\n",
tbuff, inet_ntoa(ipcp->peer_ip));
if (ipcp->peer_compproto >> 16 == PROTO_VJCOMP)
sl_compress_init(&ipcp->vj.cslc, (ipcp->peer_compproto >> 8) & 255);
if (!ipcp_InterfaceUp(ipcp))
return 0;
/*
* XXX this stuff should really live in the FSM. Our config should
* associate executable sections in files with events.
*/
if (system_Select(fp->bundle, tbuff, LINKUPFILE, NULL, NULL) < 0) {
if (bundle_GetLabel(fp->bundle)) {
if (system_Select(fp->bundle, bundle_GetLabel(fp->bundle),
LINKUPFILE, NULL, NULL) < 0)
system_Select(fp->bundle, "MYADDR", LINKUPFILE, NULL, NULL);
} else
system_Select(fp->bundle, "MYADDR", LINKUPFILE, NULL, NULL);
}
fp->more.reqs = fp->more.naks = fp->more.rejs = ipcp->cfg.fsm.maxreq * 3;
log_DisplayPrompts();
return 1;
}
static int
AcceptableAddr(const struct in_range *prange, struct in_addr ipaddr)
{
/* Is the given IP in the given range ? */
return (prange->ipaddr.s_addr & prange->mask.s_addr) ==
(ipaddr.s_addr & prange->mask.s_addr) && ipaddr.s_addr;
}
static void
IpcpDecodeConfig(struct fsm *fp, u_char *cp, int plen, int mode_type,
struct fsm_decode *dec)
{
/* Deal with incoming PROTO_IPCP */
struct iface *iface = fp->bundle->iface;
struct ipcp *ipcp = fsm2ipcp(fp);
int type, length, gotdns, gotdnsnak, n;
u_int32_t compproto;
struct compreq *pcomp;
struct in_addr ipaddr, dstipaddr, have_ip, dns[2], dnsnak[2];
char tbuff[100], tbuff2[100];
gotdns = 0;
gotdnsnak = 0;
dnsnak[0].s_addr = dnsnak[1].s_addr = INADDR_ANY;
while (plen >= sizeof(struct fsmconfig)) {
type = *cp;
length = cp[1];
if (length == 0) {
log_Printf(LogIPCP, "%s: IPCP size zero\n", fp->link->name);
break;
}
if (type < NCFTYPES)
snprintf(tbuff, sizeof tbuff, " %s[%d] ", cftypes[type], length);
else if (type > 128 && type < 128 + NCFTYPES128)
snprintf(tbuff, sizeof tbuff, " %s[%d] ", cftypes128[type-128], length);
else
snprintf(tbuff, sizeof tbuff, " <%d>[%d] ", type, length);
switch (type) {
case TY_IPADDR: /* RFC1332 */
memcpy(&ipaddr.s_addr, cp + 2, 4);
log_Printf(LogIPCP, "%s %s\n", tbuff, inet_ntoa(ipaddr));
switch (mode_type) {
case MODE_REQ:
if (iplist_isvalid(&ipcp->cfg.peer_list)) {
if (ipaddr.s_addr == INADDR_ANY ||
iplist_ip2pos(&ipcp->cfg.peer_list, ipaddr) < 0 ||
ipcp_SetIPaddress(fp->bundle, ipcp->cfg.my_range.ipaddr,
ipaddr, 1)) {
log_Printf(LogIPCP, "%s: Address invalid or already in use\n",
inet_ntoa(ipaddr));
/*
* If we've already had a valid address configured for the peer,
* try NAKing with that so that we don't have to upset things
* too much.
*/
for (n = 0; n < iface->in_addrs; n++)
if (iplist_ip2pos(&ipcp->cfg.peer_list, iface->in_addr[n].brd)
>=0) {
ipcp->peer_ip = iface->in_addr[n].brd;
break;
}
if (n == iface->in_addrs)
/* Just pick an IP number from our list */
ipcp->peer_ip = ChooseHisAddr
(fp->bundle, ipcp->cfg.my_range.ipaddr);
if (ipcp->peer_ip.s_addr == INADDR_ANY) {
memcpy(dec->rejend, cp, length);
dec->rejend += length;
} else {
memcpy(dec->nakend, cp, 2);
memcpy(dec->nakend + 2, &ipcp->peer_ip.s_addr, length - 2);
dec->nakend += length;
}
break;
}
} else if (!AcceptableAddr(&ipcp->cfg.peer_range, ipaddr)) {
/*
* If destination address is not acceptable, NAK with what we
* want to use.
*/
memcpy(dec->nakend, cp, 2);
for (n = 0; n < iface->in_addrs; n++)
if ((iface->in_addr[n].brd.s_addr &
ipcp->cfg.peer_range.mask.s_addr)
== (ipcp->cfg.peer_range.ipaddr.s_addr &
ipcp->cfg.peer_range.mask.s_addr)) {
/* We prefer the already-configured address */
memcpy(dec->nakend + 2, &iface->in_addr[n].brd.s_addr,
length - 2);
break;
}
if (n == iface->in_addrs)
memcpy(dec->nakend + 2, &ipcp->peer_ip.s_addr, length - 2);
dec->nakend += length;
break;
}
ipcp->peer_ip = ipaddr;
memcpy(dec->ackend, cp, length);
dec->ackend += length;
break;
case MODE_NAK:
if (AcceptableAddr(&ipcp->cfg.my_range, ipaddr)) {
/* Use address suggested by peer */
snprintf(tbuff2, sizeof tbuff2, "%s changing address: %s ", tbuff,
inet_ntoa(ipcp->my_ip));
log_Printf(LogIPCP, "%s --> %s\n", tbuff2, inet_ntoa(ipaddr));
ipcp->my_ip = ipaddr;
bundle_AdjustFilters(fp->bundle, &ipcp->my_ip, NULL);
} else {
log_Printf(log_IsKept(LogIPCP) ? LogIPCP : LogPHASE,
"%s: Unacceptable address!\n", inet_ntoa(ipaddr));
fsm_Close(&ipcp->fsm);
}
break;
case MODE_REJ:
ipcp->peer_reject |= (1 << type);
break;
}
break;
case TY_COMPPROTO:
pcomp = (struct compreq *)(cp + 2);
compproto = (ntohs(pcomp->proto) << 16) + (pcomp->slots << 8) +
pcomp->compcid;
log_Printf(LogIPCP, "%s %s\n", tbuff, vj2asc(compproto));
switch (mode_type) {
case MODE_REQ:
if (!IsAccepted(ipcp->cfg.vj.neg)) {
memcpy(dec->rejend, cp, length);
dec->rejend += length;
} else {
switch (length) {
case 4: /* RFC1172 */
if (ntohs(pcomp->proto) == PROTO_VJCOMP) {
log_Printf(LogWARN, "Peer is speaking RFC1172 compression "
"protocol !\n");
ipcp->heis1172 = 1;
ipcp->peer_compproto = compproto;
memcpy(dec->ackend, cp, length);
dec->ackend += length;
} else {
memcpy(dec->nakend, cp, 2);
pcomp->proto = htons(PROTO_VJCOMP);
memcpy(dec->nakend+2, &pcomp, 2);
dec->nakend += length;
}
break;
case 6: /* RFC1332 */
if (ntohs(pcomp->proto) == PROTO_VJCOMP) {
if (pcomp->slots <= MAX_VJ_STATES
&& pcomp->slots >= MIN_VJ_STATES) {
/* Ok, we can do that */
ipcp->peer_compproto = compproto;
ipcp->heis1172 = 0;
memcpy(dec->ackend, cp, length);
dec->ackend += length;
} else {
/* Get as close as we can to what he wants */
ipcp->heis1172 = 0;
memcpy(dec->nakend, cp, 2);
pcomp->slots = pcomp->slots < MIN_VJ_STATES ?
MIN_VJ_STATES : MAX_VJ_STATES;
memcpy(dec->nakend+2, &pcomp, sizeof pcomp);
dec->nakend += length;
}
} else {
/* What we really want */
memcpy(dec->nakend, cp, 2);
pcomp->proto = htons(PROTO_VJCOMP);
pcomp->slots = DEF_VJ_STATES;
pcomp->compcid = 1;
memcpy(dec->nakend+2, &pcomp, sizeof pcomp);
dec->nakend += length;
}
break;
default:
memcpy(dec->rejend, cp, length);
dec->rejend += length;
break;
}
}
break;
case MODE_NAK:
if (ntohs(pcomp->proto) == PROTO_VJCOMP) {
if (pcomp->slots > MAX_VJ_STATES)
pcomp->slots = MAX_VJ_STATES;
else if (pcomp->slots < MIN_VJ_STATES)
pcomp->slots = MIN_VJ_STATES;
compproto = (ntohs(pcomp->proto) << 16) + (pcomp->slots << 8) +
pcomp->compcid;
} else
compproto = 0;
log_Printf(LogIPCP, "%s changing compproto: %08x --> %08x\n",
tbuff, ipcp->my_compproto, compproto);
ipcp->my_compproto = compproto;
break;
case MODE_REJ:
ipcp->peer_reject |= (1 << type);
break;
}
break;
case TY_IPADDRS: /* RFC1172 */
memcpy(&ipaddr.s_addr, cp + 2, 4);
memcpy(&dstipaddr.s_addr, cp + 6, 4);
snprintf(tbuff2, sizeof tbuff2, "%s %s,", tbuff, inet_ntoa(ipaddr));
log_Printf(LogIPCP, "%s %s\n", tbuff2, inet_ntoa(dstipaddr));
switch (mode_type) {
case MODE_REQ:
memcpy(dec->rejend, cp, length);
dec->rejend += length;
break;
case MODE_NAK:
case MODE_REJ:
break;
}
break;
case TY_PRIMARY_DNS: /* DNS negotiation (rfc1877) */
case TY_SECONDARY_DNS:
memcpy(&ipaddr.s_addr, cp + 2, 4);
log_Printf(LogIPCP, "%s %s\n", tbuff, inet_ntoa(ipaddr));
switch (mode_type) {
case MODE_REQ:
if (!IsAccepted(ipcp->cfg.ns.dns_neg)) {
ipcp->my_reject |= (1 << (type - TY_ADJUST_NS));
memcpy(dec->rejend, cp, length);
dec->rejend += length;
break;
}
if (!gotdns) {
dns[0] = ipcp->cfg.ns.dns[0];
dns[1] = ipcp->cfg.ns.dns[1];
if (dns[0].s_addr == INADDR_ANY && dns[1].s_addr == INADDR_ANY)
getdns(ipcp, dns);
gotdns = 1;
}
have_ip = dns[type == TY_PRIMARY_DNS ? 0 : 1];
if (ipaddr.s_addr != have_ip.s_addr) {
/*
* The client has got the DNS stuff wrong (first request) so
* we'll tell 'em how it is
*/
memcpy(dec->nakend, cp, 2); /* copy first two (type/length) */
memcpy(dec->nakend + 2, &have_ip.s_addr, length - 2);
dec->nakend += length;
} else {
/*
* Otherwise they have it right (this time) so we send a ack packet
* back confirming it... end of story
*/
memcpy(dec->ackend, cp, length);
dec->ackend += length;
}
break;
case MODE_NAK: /* what does this mean?? */
if (IsEnabled(ipcp->cfg.ns.dns_neg)) {
gotdnsnak = 1;
memcpy(&dnsnak[type == TY_PRIMARY_DNS ? 0 : 1].s_addr, cp + 2, 4);
}
break;
case MODE_REJ: /* Can't do much, stop asking */
ipcp->peer_reject |= (1 << (type - TY_ADJUST_NS));
break;
}
break;
case TY_PRIMARY_NBNS: /* M$ NetBIOS nameserver hack (rfc1877) */
case TY_SECONDARY_NBNS:
memcpy(&ipaddr.s_addr, cp + 2, 4);
log_Printf(LogIPCP, "%s %s\n", tbuff, inet_ntoa(ipaddr));
switch (mode_type) {
case MODE_REQ:
have_ip.s_addr =
ipcp->cfg.ns.nbns[type == TY_PRIMARY_NBNS ? 0 : 1].s_addr;
if (have_ip.s_addr == INADDR_ANY) {
log_Printf(LogIPCP, "NBNS REQ - rejected - nbns not set\n");
ipcp->my_reject |= (1 << (type - TY_ADJUST_NS));
memcpy(dec->rejend, cp, length);
dec->rejend += length;
break;
}
if (ipaddr.s_addr != have_ip.s_addr) {
memcpy(dec->nakend, cp, 2);
memcpy(dec->nakend+2, &have_ip.s_addr, length);
dec->nakend += length;
} else {
memcpy(dec->ackend, cp, length);
dec->ackend += length;
}
break;
case MODE_NAK:
log_Printf(LogIPCP, "MS NBNS req %d - NAK??\n", type);
break;
case MODE_REJ:
log_Printf(LogIPCP, "MS NBNS req %d - REJ??\n", type);
break;
}
break;
default:
if (mode_type != MODE_NOP) {
ipcp->my_reject |= (1 << type);
memcpy(dec->rejend, cp, length);
dec->rejend += length;
}
break;
}
plen -= length;
cp += length;
}
if (gotdnsnak)
if (!setdns(ipcp, dnsnak)) {
ipcp->peer_reject |= (1 << (TY_PRIMARY_DNS - TY_ADJUST_NS));
ipcp->peer_reject |= (1 << (TY_SECONDARY_DNS - TY_ADJUST_NS));
}
if (mode_type != MODE_NOP) {
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;
}
}
extern struct mbuf *
ipcp_Input(struct bundle *bundle, struct link *l, struct mbuf *bp)
{
/* Got PROTO_IPCP from link */
mbuf_SetType(bp, MB_IPCPIN);
if (bundle_Phase(bundle) == PHASE_NETWORK)
fsm_Input(&bundle->ncp.ipcp.fsm, bp);
else {
if (bundle_Phase(bundle) < PHASE_NETWORK)
log_Printf(LogIPCP, "%s: Error: Unexpected IPCP in phase %s (ignored)\n",
l->name, bundle_PhaseName(bundle));
mbuf_Free(bp);
}
return NULL;
}
int
ipcp_UseHisIPaddr(struct bundle *bundle, struct in_addr hisaddr)
{
struct ipcp *ipcp = &bundle->ncp.ipcp;
memset(&ipcp->cfg.peer_range, '\0', sizeof ipcp->cfg.peer_range);
iplist_reset(&ipcp->cfg.peer_list);
ipcp->peer_ip = ipcp->cfg.peer_range.ipaddr = hisaddr;
ipcp->cfg.peer_range.mask.s_addr = INADDR_BROADCAST;
ipcp->cfg.peer_range.width = 32;
if (ipcp_SetIPaddress(bundle, ipcp->cfg.my_range.ipaddr, hisaddr, 0) < 0)
return 0;
return 1; /* Ok */
}
int
ipcp_UseHisaddr(struct bundle *bundle, const char *hisaddr, int setaddr)
{
struct ipcp *ipcp = &bundle->ncp.ipcp;
/* Use `hisaddr' for the peers address (set iface if `setaddr') */
memset(&ipcp->cfg.peer_range, '\0', sizeof ipcp->cfg.peer_range);
iplist_reset(&ipcp->cfg.peer_list);
if (strpbrk(hisaddr, ",-")) {
iplist_setsrc(&ipcp->cfg.peer_list, hisaddr);
if (iplist_isvalid(&ipcp->cfg.peer_list)) {
iplist_setrandpos(&ipcp->cfg.peer_list);
ipcp->peer_ip = ChooseHisAddr(bundle, ipcp->my_ip);
if (ipcp->peer_ip.s_addr == INADDR_ANY) {
log_Printf(LogWARN, "%s: None available !\n", ipcp->cfg.peer_list.src);
return 0;
}
ipcp->cfg.peer_range.ipaddr.s_addr = ipcp->peer_ip.s_addr;
ipcp->cfg.peer_range.mask.s_addr = INADDR_BROADCAST;
ipcp->cfg.peer_range.width = 32;
} else {
log_Printf(LogWARN, "%s: Invalid range !\n", hisaddr);
return 0;
}
} else if (ParseAddr(ipcp, hisaddr, &ipcp->cfg.peer_range.ipaddr,
&ipcp->cfg.peer_range.mask,
&ipcp->cfg.peer_range.width) != 0) {
ipcp->peer_ip.s_addr = ipcp->cfg.peer_range.ipaddr.s_addr;
if (setaddr && ipcp_SetIPaddress(bundle, ipcp->cfg.my_range.ipaddr,
ipcp->cfg.peer_range.ipaddr, 0) < 0)
return 0;
} else
return 0;
bundle_AdjustFilters(bundle, NULL, &ipcp->peer_ip);
return 1; /* Ok */
}
struct in_addr
addr2mask(struct in_addr addr)
{
u_int32_t haddr = ntohl(addr.s_addr);
haddr = IN_CLASSA(haddr) ? IN_CLASSA_NET :
IN_CLASSB(haddr) ? IN_CLASSB_NET :
IN_CLASSC_NET;
addr.s_addr = htonl(haddr);
return addr;
}