1673 lines
39 KiB
C
1673 lines
39 KiB
C
/*
|
|
* sys-linux.c - System-dependent procedures for setting up
|
|
* PPP interfaces on Linux systems
|
|
*
|
|
* Fairly patched version for isdn4linux
|
|
* copyright (c) 1995,1996,1997 of all patches by Michael Hipp
|
|
* still no warranties (see disclaimer)
|
|
*
|
|
* Copyright (c) 1989 Carnegie Mellon University.
|
|
* All rights reserved.
|
|
*
|
|
* 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 Carnegie Mellon University. The name of the
|
|
* University 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.
|
|
*/
|
|
|
|
char sys_rcsid[] = "$Id: sys-linux.c,v 1.15 1998/10/29 17:28:49 hipp Exp $";
|
|
|
|
#define _LINUX_STRING_H_
|
|
|
|
#include <sys/types.h>
|
|
#include <sys/ioctl.h>
|
|
#include <sys/stat.h>
|
|
#include <sys/socket.h>
|
|
#include <sys/time.h>
|
|
#include <sys/errno.h>
|
|
#include <sys/stat.h>
|
|
#include <sys/file.h>
|
|
#include <sys/utsname.h>
|
|
|
|
#include <stdio.h>
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
#include <syslog.h>
|
|
#include <time.h>
|
|
#include <memory.h>
|
|
#include <utmp.h>
|
|
#include <mntent.h>
|
|
#include <signal.h>
|
|
#include <fcntl.h>
|
|
#include <ctype.h>
|
|
#include <unistd.h>
|
|
#include <netinet/in.h>
|
|
#include <arpa/inet.h>
|
|
#include </usr/include/net/if.h>
|
|
#include </usr/include/net/if_arp.h>
|
|
#include </usr/include/net/route.h>
|
|
#if defined __GLIBC__ && __GLIBC__ >= 2
|
|
# include </usr/include/net/ppp_defs.h>
|
|
# include </usr/include/net/if_ppp.h>
|
|
# include </usr/include/net/ethernet.h>
|
|
# include "route.h"
|
|
#else
|
|
# include <linux/ppp_defs.h>
|
|
# include <linux/if_ppp.h>
|
|
# include <linux/if_ether.h>
|
|
#endif
|
|
|
|
#include <linux/isdn_ppp.h>
|
|
|
|
#include "fsm.h"
|
|
#include "ipppd.h"
|
|
#include "ipcp.h"
|
|
#include "ipxcp.h"
|
|
#include "ccp.h"
|
|
#include "lcp.h"
|
|
|
|
extern int force_driver;
|
|
|
|
static int prev_kdebugflag = 0;
|
|
static int has_default_route = 0;
|
|
static int driver_version = 0;
|
|
static int driver_modification = 0;
|
|
static int driver_patch = 0;
|
|
static int devroute = -1; /* 0 for Linux >= 2.1.x */
|
|
static int last_net_mask = 0; /* Ugly! */
|
|
|
|
int get_ether_addr (u_int32_t ipaddr, struct sockaddr *hwaddr, char *name);
|
|
static void decode_version (char *buf, int *version,
|
|
int *modification, int *patch);
|
|
int sockfd; /* socket for doing interface ioctls */
|
|
|
|
static char *lock_file;
|
|
|
|
#define MAX_IFS 5000
|
|
|
|
#define FLAGS_GOOD (IFF_UP | IFF_BROADCAST)
|
|
#define FLAGS_MASK (IFF_UP | IFF_BROADCAST | \
|
|
IFF_POINTOPOINT | IFF_LOOPBACK | IFF_NOARP)
|
|
|
|
/*
|
|
* SET_SA_FAMILY - set the sa_family field of a struct sockaddr,
|
|
* if it exists.
|
|
*/
|
|
|
|
#define SET_SA_FAMILY(addr, family) \
|
|
memset ((char *) &(addr), '\0', sizeof(addr)); \
|
|
addr.sa_family = (family);
|
|
|
|
/*
|
|
* Determine if the PPP connection should still be present.
|
|
*/
|
|
|
|
#define still_ppp(linkunit) (lns[linkunit].hungup == 0)
|
|
|
|
/*
|
|
*
|
|
*/
|
|
|
|
void enable_mp(int linkunit,int flags)
|
|
{
|
|
int mpflags,s;
|
|
s = ioctl(lns[linkunit].fd,PPPIOCGMPFLAGS,(caddr_t) &mpflags);
|
|
if(s < 0)
|
|
syslog(LOG_ERR,"ipppd: Can't get MP-Flags");
|
|
else
|
|
{
|
|
mpflags |= SC_MP_PROT;
|
|
mpflags &= ~SC_REJ_MP_PROT;
|
|
mpflags |= flags;
|
|
s = ioctl(lns[linkunit].fd,PPPIOCSMPFLAGS,(caddr_t) &mpflags);
|
|
if(s < 0)
|
|
syslog(LOG_ERR,"ipppd: Can't set MP-Flags");
|
|
}
|
|
}
|
|
|
|
|
|
/*
|
|
* Functions to read and set the flags value in the device driver
|
|
*/
|
|
|
|
static int get_flags (int tu,int *error)
|
|
{
|
|
int flags;
|
|
|
|
*error = 0;
|
|
|
|
if(lns[tu].fd < 0) {
|
|
*error = 1;
|
|
return 0;
|
|
}
|
|
|
|
if (ioctl(lns[tu].fd, PPPIOCGFLAGS, (caddr_t) &flags) < 0) {
|
|
syslog(LOG_ERR, "ioctl(PPPIOCGFLAGS): %m");
|
|
*error = 1;
|
|
return 0;
|
|
}
|
|
|
|
MAINDEBUG ((LOG_DEBUG, "get flags = %x\n", flags));
|
|
return flags;
|
|
}
|
|
|
|
static void set_flags (int flags,int tu)
|
|
{
|
|
MAINDEBUG ((LOG_DEBUG, "set flags = %x\n", flags));
|
|
|
|
if(lns[tu].fd < 0)
|
|
return;
|
|
|
|
if (ioctl(lns[tu].fd, PPPIOCSFLAGS, (caddr_t) &flags) < 0) {
|
|
syslog(LOG_ERR, "ioctl(PPPIOCSFLAGS, %x): %m", flags);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* sys_init - System-dependent initialization.
|
|
*/
|
|
void sys_init(void)
|
|
{
|
|
struct utsname uts;
|
|
int maj, min, pat;
|
|
|
|
openlog("ipppd", LOG_PID | LOG_NDELAY, LOG_PPP);
|
|
setlogmask(LOG_UPTO(LOG_INFO));
|
|
if (debug)
|
|
setlogmask(LOG_UPTO(LOG_DEBUG));
|
|
|
|
/* Get an internet socket for doing socket ioctls. */
|
|
sockfd = socket(AF_INET, SOCK_DGRAM, 0);
|
|
if (sockfd < 0) {
|
|
syslog(LOG_ERR, "Couldn't create IP socket: %m");
|
|
die(1);
|
|
}
|
|
if (devroute < 0) {
|
|
uname(&uts);
|
|
maj = min = pat = 0;
|
|
decode_version(uts.release, &maj, &min, &pat);
|
|
if (maj >= 2 && min >= 1) /* Linux >= 2.1.x */
|
|
devroute = 0;
|
|
else
|
|
devroute = 1;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* note_debug_level - note a change in the debug level.
|
|
*/
|
|
void note_debug_level (void)
|
|
{
|
|
if (debug) {
|
|
MAINDEBUG ((LOG_INFO, "Debug turned ON, Level %d", debug));
|
|
setlogmask(LOG_UPTO(LOG_DEBUG));
|
|
}
|
|
else
|
|
setlogmask(LOG_UPTO(LOG_WARNING));
|
|
}
|
|
|
|
/*
|
|
* set_kdebugflag - Define the debugging level for the kernel
|
|
*/
|
|
int set_kdebugflag (int requested_level,int tu)
|
|
{
|
|
if (ioctl(lns[tu].fd, PPPIOCGDEBUG, &prev_kdebugflag) < 0) {
|
|
syslog(LOG_ERR, "ioctl(PPPIOCGDEBUG): %m");
|
|
return 0;
|
|
}
|
|
|
|
if (prev_kdebugflag != requested_level) {
|
|
if (ioctl(lns[tu].fd, PPPIOCSDEBUG, &requested_level) < 0) {
|
|
syslog (LOG_ERR, "ioctl(PPPIOCSDEBUG): %m");
|
|
return 0;
|
|
}
|
|
MAINDEBUG ((LOG_INFO, "set kernel debugging level to %d", requested_level));
|
|
}
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* establish_ppp - get UNIT, callinfo, IF-MTU
|
|
*/
|
|
void establish_ppp (int linkunit)
|
|
{
|
|
if (ioctl(lns[linkunit].fd, PPPIOCGUNIT, &lns[linkunit].ifunit) < 0) {
|
|
syslog(LOG_ERR, "ioctl(PPPIOCGUNIT): %m");
|
|
lns[linkunit].ifunit = -1;
|
|
return;
|
|
}
|
|
lns[linkunit].master = -1;
|
|
sprintf(lns[linkunit].ifname,"%s%d","ippp",lns[linkunit].ifunit);
|
|
|
|
if( ioctl(lns[linkunit].fd, PPPIOCGCALLINFO, &lns[linkunit].pci) == 0) {
|
|
struct pppcallinfo *pci = &lns[linkunit].pci;
|
|
syslog(LOG_NOTICE, "Local number: %s, Remote number: %s, Type: %s",
|
|
pci->local_num,pci->remote_num,pci->calltype & CALLTYPE_INCOMING ? "incoming" : "outgoing" );
|
|
#ifdef RADIUS
|
|
strncpy ( lns[linkunit].remote_number, pci->remote_num, sizeof( lns[linkunit].remote_number ) ) ;
|
|
#endif
|
|
}
|
|
|
|
if(useifmtu) {
|
|
struct ifreq ifr;
|
|
|
|
memset (&ifr, '\0', sizeof (ifr));
|
|
strncpy(ifr.ifr_name, lns[linkunit].ifname, sizeof (ifr.ifr_name));
|
|
if (ioctl(sockfd, SIOCGIFMTU, (caddr_t) &ifr) < 0)
|
|
syslog(LOG_NOTICE , "Can't get MTU from device %s",lns[linkunit].ifname);
|
|
else
|
|
lcp_allowoptions[lns[linkunit].lcp_unit].mru = ifr.ifr_mtu;
|
|
}
|
|
|
|
#if 0
|
|
set_kdebugflag (kdebugflag,linkunit);
|
|
#endif
|
|
|
|
MAINDEBUG ((LOG_NOTICE, "Using version %d.%d.%d of PPP driver",
|
|
driver_version, driver_modification, driver_patch));
|
|
}
|
|
|
|
/*
|
|
* output - Output PPP packet.
|
|
*/
|
|
void output_ppp (int linkunit, unsigned char *p, int len)
|
|
{
|
|
if (debug)
|
|
log_packet(p, len, "sent ",linkunit);
|
|
|
|
if (write(lns[linkunit].fd, p, len) < 0) {
|
|
syslog(LOG_ERR, "write, unit: %d fd: %d: %m",linkunit,lns[linkunit].fd);
|
|
die(1);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* read_packet - get a PPP packet from the serial device.
|
|
*/
|
|
int read_packet (unsigned char *buf,int linkunit)
|
|
{
|
|
int len;
|
|
|
|
len = read(lns[linkunit].fd, buf, PPP_MTU + PPP_HDRLEN);
|
|
if (len < 0) {
|
|
if (errno == EWOULDBLOCK)
|
|
return -1;
|
|
syslog(LOG_ERR, "read(fd): %m");
|
|
die(1);
|
|
}
|
|
return len;
|
|
}
|
|
|
|
/*
|
|
* ppp_send_config - configure the transmit characteristics of
|
|
* the ppp interface.
|
|
*/
|
|
void ppp_send_config (int unit,int mtu,u_int32_t asyncmap,int pcomp,int accomp)
|
|
{
|
|
u_int x;
|
|
struct ifreq ifr;
|
|
int err;
|
|
|
|
MAINDEBUG ((LOG_DEBUG, "send_config: mtu = %d\n", mtu));
|
|
|
|
/*
|
|
* Ensure that the link is still up.
|
|
*/
|
|
if (still_ppp(unit)) {
|
|
/*
|
|
* Set the MTU and other parameters for the ppp device
|
|
*/
|
|
if(lns[unit].master < 0) {
|
|
memset (&ifr, '\0', sizeof (ifr));
|
|
strncpy(ifr.ifr_name, lns[unit].ifname, sizeof (ifr.ifr_name));
|
|
ifr.ifr_mtu = mtu;
|
|
|
|
if (ioctl(sockfd, SIOCSIFMTU, (caddr_t) &ifr) < 0) {
|
|
syslog(LOG_ERR, "ioctl(SIOCSIFMTU): %m, %d %s %d.",sockfd,ifr.ifr_name,ifr.ifr_mtu);
|
|
}
|
|
}
|
|
|
|
x = get_flags(unit,&err);
|
|
if(err)
|
|
return;
|
|
x = pcomp ? x | SC_COMP_PROT : x & ~SC_COMP_PROT;
|
|
x = accomp ? x | SC_COMP_AC : x & ~SC_COMP_AC;
|
|
set_flags(x,unit);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* ppp_mp_send_config, ppp_mp_recv_config
|
|
*/
|
|
void ppp_mp_send_config(int unit,int mtu)
|
|
{
|
|
int r;
|
|
r = ioctl(lns[unit].fd,PPPIOCSMPMTU,&mtu);
|
|
syslog(LOG_DEBUG,"mp_mtu: %d",r);
|
|
}
|
|
|
|
void ppp_mp_recv_config(int unit,int mru)
|
|
{
|
|
int r;
|
|
r = ioctl(lns[unit].fd,PPPIOCSMPMRU,&mru);
|
|
syslog(LOG_DEBUG,"mp_mru: %d",r);
|
|
}
|
|
|
|
/*
|
|
* ppp_set_xaccm - set the extended transmit ACCM for the interface.
|
|
*/
|
|
void ppp_set_xaccm (int unit, ext_accm accm)
|
|
{
|
|
#if 0
|
|
MAINDEBUG ((LOG_DEBUG, "set_xaccm: %08lx %08lx %08lx %08lx\n",
|
|
accm[0], accm[1], accm[2], accm[3]));
|
|
|
|
if (ioctl(lns[unit].fd, PPPIOCSXASYNCMAP, accm) < 0 && errno != ENOTTY)
|
|
syslog(LOG_WARNING, "ioctl(set extended ACCM): %m");
|
|
#endif
|
|
}
|
|
|
|
/*
|
|
* ppp_recv_config - configure the receive-side characteristics of
|
|
* the ppp interface.
|
|
*/
|
|
void ppp_recv_config (int unit,int mru,u_int32_t asyncmap,int pcomp,int accomp)
|
|
{
|
|
u_int x;
|
|
int err;
|
|
|
|
/*
|
|
* If we were called because the link has gone down then there is nothing
|
|
* which may be done. Just return without incident.
|
|
*/
|
|
if (!still_ppp(unit))
|
|
return;
|
|
|
|
/*
|
|
* Set the receiver parameters
|
|
*/
|
|
if (ioctl(lns[unit].fd, PPPIOCSMRU, (caddr_t) &mru) < 0)
|
|
syslog(LOG_ERR, "ioctl(PPPIOCSMRU): %m");
|
|
|
|
x = get_flags(unit,&err);
|
|
if(err)
|
|
return;
|
|
x = accomp ? x & ~SC_REJ_COMP_AC : x | SC_REJ_COMP_AC;
|
|
set_flags (x,unit);
|
|
}
|
|
|
|
/*
|
|
* ccp_test - ask kernel whether a given compression method
|
|
* is acceptable for use.
|
|
*/
|
|
int ccp_test (int ccp_unit, u_char *opt_ptr, int opt_len, int for_transmit)
|
|
{
|
|
#ifdef NEW_VERS
|
|
struct isdn_ppp_comp_data data;
|
|
int linkunit = ccp_fsm[ccp_unit].unit;
|
|
|
|
memset (&data, '\0', sizeof (data));
|
|
data.num = opt_ptr[0];
|
|
data.optlen = opt_len - 2;
|
|
if(data.optlen > ISDN_PPP_COMP_MAX_OPTIONS) {
|
|
syslog(LOG_NOTICE, "ccp_test: options field too long!\n");
|
|
return -1;
|
|
}
|
|
memcpy(data.options,opt_ptr+2,data.optlen);
|
|
|
|
data.flags = 0;
|
|
if(for_transmit)
|
|
data.flags |= IPPP_COMP_FLAG_XMIT;
|
|
if(ccp_fsm[ccp_unit].protocol == PPP_LINK_CCP)
|
|
data.flags |= IPPP_COMP_FLAG_LINK;
|
|
|
|
if (ioctl(lns[linkunit].fd, PPPIOCSCOMPRESSOR, (caddr_t) &data) >= 0)
|
|
return 1;
|
|
return (errno == ENOBUFS)? 0: -1;
|
|
#else
|
|
return -1;
|
|
#endif
|
|
|
|
}
|
|
|
|
int ccp_get_compressors(int ccp_unit,unsigned long *protos)
|
|
{
|
|
int linkunit = ccp_fsm[ccp_unit].unit;
|
|
|
|
if (ioctl(lns[linkunit].fd, PPPIOCGCOMPRESSORS, protos) >= 0)
|
|
return 0;
|
|
return -1;
|
|
}
|
|
|
|
|
|
/*
|
|
* ccp_flags_set - inform kernel about the current state of CCP.
|
|
*/
|
|
void ccp_flags_set (int ccp_unit, int isopen, int isup)
|
|
{
|
|
int linkunit = ccp_fsm[ccp_unit].unit;
|
|
int err;
|
|
|
|
if (still_ppp(linkunit)) {
|
|
int x = get_flags(linkunit,&err);
|
|
if(err)
|
|
return;
|
|
x = isopen? x | SC_CCP_OPEN : x &~ SC_CCP_OPEN;
|
|
x = isup? x | SC_CCP_UP : x &~ SC_CCP_UP;
|
|
set_flags (x,linkunit);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* ccp_fatal_error - returns 1 if decompression was disabled as a
|
|
* result of an error detected after decompression of a packet,
|
|
* 0 otherwise. This is necessary because of patent nonsense.
|
|
*/
|
|
int ccp_fatal_error (int ccp_unit)
|
|
{
|
|
int linkunit = ccp_fsm[ccp_unit].unit;
|
|
int err;
|
|
int x = get_flags(linkunit,&err);
|
|
|
|
if(err)
|
|
return 0;
|
|
|
|
return x & SC_DC_FERROR;
|
|
}
|
|
|
|
/*
|
|
* sifvjcomp - config tcp header compression
|
|
*/
|
|
int sifvjcomp (int unit, int vjcomp, int cidcomp, int maxcid)
|
|
{
|
|
int err;
|
|
u_int x = get_flags(unit,&err);
|
|
|
|
if(err)
|
|
return 0;
|
|
|
|
if (vjcomp) {
|
|
if (ioctl (lns[unit].fd, PPPIOCSMAXCID, (caddr_t) &maxcid) < 0) {
|
|
syslog (LOG_ERR, "ioctl(PPPIOCSFLAGS): %m");
|
|
vjcomp = 0;
|
|
}
|
|
}
|
|
|
|
x = vjcomp ? x | SC_COMP_TCP : x & ~SC_COMP_TCP;
|
|
x = cidcomp ? x & ~SC_NO_TCP_CCID : x | SC_NO_TCP_CCID;
|
|
set_flags (x,unit);
|
|
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* sifup - Config the interface up and enable IP packets to pass.
|
|
*/
|
|
int sifup (int u)
|
|
{
|
|
struct ifreq ifr;
|
|
u_int x;
|
|
|
|
memset (&ifr, '\0', sizeof (ifr));
|
|
strncpy(ifr.ifr_name, lns[u].ifname, sizeof (ifr.ifr_name));
|
|
if (ioctl(sockfd, SIOCGIFFLAGS, (caddr_t) &ifr) < 0) {
|
|
syslog(LOG_ERR, "ioctl (SIOCGIFFLAGS): %m");
|
|
return 0;
|
|
}
|
|
|
|
ifr.ifr_flags |= (IFF_UP | IFF_POINTOPOINT);
|
|
if (ioctl(sockfd, SIOCSIFFLAGS, (caddr_t) &ifr) < 0) {
|
|
syslog(LOG_ERR, "ioctl(SIOCSIFFLAGS): %m");
|
|
return 0;
|
|
}
|
|
if( ioctl(lns[u].fd,PPPIOCGFLAGS,(caddr_t) &x) < 0) {
|
|
syslog(LOG_ERR,"ioctl(PPPIOCGFLAGS): %m");
|
|
return 0;
|
|
}
|
|
x |= SC_ENABLE_IP;
|
|
if( ioctl(lns[u].fd,PPPIOCSFLAGS,(caddr_t) &x) < 0) {
|
|
syslog(LOG_ERR,"ioctl(PPPIOCSFLAGS): %m");
|
|
return 0;
|
|
}
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* sifdown - Config the interface down and disable IP.
|
|
*/
|
|
int sifdown (int u)
|
|
{
|
|
struct ifreq ifr;
|
|
|
|
memset (&ifr, '\0', sizeof (ifr));
|
|
strncpy(ifr.ifr_name, lns[u].ifname, sizeof (ifr.ifr_name));
|
|
if (ioctl(sockfd, SIOCGIFFLAGS, (caddr_t) &ifr) < 0) {
|
|
syslog(LOG_ERR, "ioctl (SIOCGIFFLAGS): %m");
|
|
return 0;
|
|
}
|
|
|
|
ifr.ifr_flags &= ~IFF_UP;
|
|
ifr.ifr_flags |= IFF_POINTOPOINT;
|
|
if (ioctl(sockfd, SIOCSIFFLAGS, (caddr_t) &ifr) < 0) {
|
|
syslog(LOG_ERR, "ioctl(SIOCSIFFLAGS): %m");
|
|
return 0;
|
|
}
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* sifbundle
|
|
*/
|
|
int sifbundle (int linkunit0,int linkunit1)
|
|
{
|
|
return ioctl(lns[linkunit0].fd, PPPIOCBUNDLE, &lns[linkunit1].ifunit );
|
|
}
|
|
|
|
/*
|
|
* sifaddr - Config the interface IP addresses and netmask.
|
|
*/
|
|
int sifaddr (int unit, int our_adr, int his_adr, int net_mask)
|
|
{
|
|
struct ifreq ifr;
|
|
struct rtentry rt;
|
|
|
|
memset (&ifr, '\0', sizeof (ifr));
|
|
memset (&rt, '\0', sizeof (rt));
|
|
|
|
SET_SA_FAMILY (ifr.ifr_addr, AF_INET);
|
|
SET_SA_FAMILY (ifr.ifr_dstaddr, AF_INET);
|
|
SET_SA_FAMILY (ifr.ifr_netmask, AF_INET);
|
|
|
|
strncpy (ifr.ifr_name, lns[unit].ifname, sizeof (ifr.ifr_name));
|
|
|
|
/*
|
|
* Set our IP address
|
|
*/
|
|
((struct sockaddr_in *) &ifr.ifr_addr)->sin_addr.s_addr = our_adr;
|
|
if (ioctl(sockfd, SIOCSIFADDR, (caddr_t) &ifr) < 0) {
|
|
if (errno != EEXIST) {
|
|
syslog (LOG_ERR, "ioctl(SIOCAIFADDR): %m");
|
|
} else {
|
|
syslog (LOG_WARNING, "ioctl(SIOCAIFADDR): Address already exists");
|
|
}
|
|
return (0);
|
|
}
|
|
|
|
/*
|
|
* Set the gateway address
|
|
*/
|
|
((struct sockaddr_in *) &ifr.ifr_dstaddr)->sin_addr.s_addr = his_adr;
|
|
if (ioctl(sockfd, SIOCSIFDSTADDR, (caddr_t) &ifr) < 0) {
|
|
syslog (LOG_ERR, "ioctl(SIOCSIFDSTADDR): %m");
|
|
return (0);
|
|
}
|
|
|
|
/*
|
|
* Set the netmask
|
|
*/
|
|
if (net_mask != 0) {
|
|
((struct sockaddr_in *) &ifr.ifr_netmask)->sin_addr.s_addr = net_mask;
|
|
if (ioctl(sockfd, SIOCSIFNETMASK, (caddr_t) &ifr) < 0) {
|
|
syslog (LOG_ERR, "ioctl(SIOCSIFNETMASK): %m");
|
|
return (0);
|
|
}
|
|
}
|
|
last_net_mask = net_mask;
|
|
|
|
/*
|
|
* Add the device route
|
|
*/
|
|
if (!devroute)
|
|
return 1;
|
|
|
|
if (hostroute) {
|
|
SET_SA_FAMILY (rt.rt_dst, AF_INET);
|
|
SET_SA_FAMILY (rt.rt_gateway, AF_INET);
|
|
SET_SA_FAMILY (rt.rt_genmask, AF_INET);
|
|
rt.rt_dev = lns[unit].ifname; /* MJC */
|
|
|
|
if (net_mask)
|
|
his_adr &= net_mask;
|
|
|
|
((struct sockaddr_in *) &rt.rt_gateway)->sin_addr.s_addr = 0L;
|
|
((struct sockaddr_in *) &rt.rt_dst)->sin_addr.s_addr = his_adr;
|
|
((struct sockaddr_in *) &rt.rt_genmask)->sin_addr.s_addr = net_mask;
|
|
rt.rt_flags = RTF_UP;
|
|
if (net_mask == 0)
|
|
rt.rt_flags |= RTF_HOST;
|
|
|
|
if (ioctl(sockfd, SIOCADDRT, &rt) < 0) {
|
|
syslog (LOG_ERR, "ioctl(SIOCADDRT) device route (%s/%s/%08x): %m",
|
|
rt.rt_dev,
|
|
inet_ntoa(((struct sockaddr_in *)&rt.rt_dst)->sin_addr),
|
|
ntohl(net_mask));
|
|
return (0);
|
|
}
|
|
|
|
}
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* cifaddr - Clear the interface IP addresses, and delete routes
|
|
* through the interface if possible.
|
|
*/
|
|
|
|
int cifaddr (int unit, int our_adr, int his_adr)
|
|
{
|
|
struct rtentry rt;
|
|
|
|
/*
|
|
* Delete the route through the device
|
|
*/
|
|
|
|
if (!devroute)
|
|
return 1;
|
|
|
|
memset (&rt, '\0', sizeof (rt));
|
|
|
|
SET_SA_FAMILY (rt.rt_dst, AF_INET);
|
|
SET_SA_FAMILY (rt.rt_gateway, AF_INET);
|
|
SET_SA_FAMILY (rt.rt_genmask, AF_INET);
|
|
rt.rt_dev = lns[unit].ifname; /* MJC */
|
|
|
|
if (last_net_mask)
|
|
his_adr &= last_net_mask;
|
|
((struct sockaddr_in *) &rt.rt_gateway)->sin_addr.s_addr = 0;
|
|
((struct sockaddr_in *) &rt.rt_dst)->sin_addr.s_addr = his_adr;
|
|
((struct sockaddr_in *) &rt.rt_genmask)->sin_addr.s_addr = last_net_mask;
|
|
rt.rt_flags = RTF_UP;
|
|
if (last_net_mask == 0)
|
|
rt.rt_flags |= RTF_HOST;
|
|
|
|
if (ioctl(sockfd, SIOCDELRT, &rt) < 0 && errno != ESRCH) {
|
|
if (still_ppp(unit)) {
|
|
syslog (LOG_ERR, "ioctl(SIOCDELRT) device route: %m");
|
|
return (0);
|
|
}
|
|
}
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* path_to_proc - determine the path to the proc file system data
|
|
*/
|
|
|
|
FILE *route_fd = (FILE *) 0;
|
|
static char route_buffer [512];
|
|
|
|
static char *path_to_proc (void);
|
|
static int open_route_table (void);
|
|
static void close_route_table (void);
|
|
static int read_route_table (struct rtentry *rt);
|
|
static int defaultroute_exists (void);
|
|
|
|
/*
|
|
* path_to_proc - find the path to the route tables in the proc file system
|
|
*/
|
|
|
|
static char *path_to_proc (void)
|
|
{
|
|
struct mntent *mntent;
|
|
FILE *fp;
|
|
|
|
fp = fopen (MOUNTED, "r");
|
|
if (fp != 0)
|
|
{
|
|
mntent = getmntent (fp);
|
|
while (mntent != (struct mntent *) 0)
|
|
{
|
|
if (strcmp (mntent->mnt_type, MNTTYPE_IGNORE) != 0)
|
|
{
|
|
if (strcmp (mntent->mnt_type, "proc") == 0)
|
|
{
|
|
strncpy (route_buffer, mntent->mnt_dir,
|
|
sizeof (route_buffer)-10);
|
|
route_buffer [sizeof (route_buffer)-10] = '\0';
|
|
fclose (fp);
|
|
return (route_buffer);
|
|
}
|
|
}
|
|
mntent = getmntent (fp);
|
|
}
|
|
fclose (fp);
|
|
}
|
|
|
|
syslog (LOG_ERR, "proc file system not mounted");
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* close_route_table - close the interface to the route table
|
|
*/
|
|
|
|
static void close_route_table (void)
|
|
{
|
|
if (route_fd != (FILE *) 0) {
|
|
fclose (route_fd);
|
|
route_fd = (FILE *) 0;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* open_route_table - open the interface to the route table
|
|
*/
|
|
|
|
static int open_route_table (void)
|
|
{
|
|
char *path;
|
|
|
|
close_route_table();
|
|
|
|
path = path_to_proc();
|
|
if (path == NULL)
|
|
{
|
|
return 0;
|
|
}
|
|
|
|
strcat (path, "/net/route");
|
|
route_fd = fopen (path, "r");
|
|
if (route_fd == (FILE *) 0)
|
|
{
|
|
syslog (LOG_ERR, "can not open %s: %m", path);
|
|
return 0;
|
|
}
|
|
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* read_route_table - read the next entry from the route table
|
|
*/
|
|
|
|
static int read_route_table (struct rtentry *rt)
|
|
{
|
|
static char delims[] = " \t\n";
|
|
char *dev_ptr, *dst_ptr, *gw_ptr, *flag_ptr;
|
|
|
|
memset (rt, '\0', sizeof (struct rtentry));
|
|
|
|
for (;;)
|
|
{
|
|
if (fgets (route_buffer, sizeof (route_buffer), route_fd) ==
|
|
(char *) 0)
|
|
{
|
|
return 0;
|
|
}
|
|
|
|
dev_ptr = strtok (route_buffer, delims); /* interface name */
|
|
dst_ptr = strtok (NULL, delims); /* destination address */
|
|
gw_ptr = strtok (NULL, delims); /* gateway */
|
|
flag_ptr = strtok (NULL, delims); /* flags */
|
|
|
|
if (flag_ptr == (char *) 0) /* assume that we failed, somewhere. */
|
|
{
|
|
return 0;
|
|
}
|
|
|
|
/* Discard that stupid header line which should never
|
|
* have been there in the first place !! */
|
|
if (isxdigit (*dst_ptr) && isxdigit (*gw_ptr) && isxdigit (*flag_ptr))
|
|
{
|
|
break;
|
|
}
|
|
}
|
|
|
|
((struct sockaddr_in *) &rt->rt_dst)->sin_addr.s_addr =
|
|
strtoul (dst_ptr, NULL, 16);
|
|
|
|
((struct sockaddr_in *) &rt->rt_gateway)->sin_addr.s_addr =
|
|
strtoul (gw_ptr, NULL, 16);
|
|
|
|
rt->rt_flags = (short) strtoul (flag_ptr, NULL, 16);
|
|
rt->rt_dev = dev_ptr;
|
|
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* defaultroute_exists - determine if there is a default route
|
|
*/
|
|
|
|
static int defaultroute_exists (void)
|
|
{
|
|
struct rtentry rt;
|
|
int result = 0;
|
|
|
|
if (!open_route_table())
|
|
{
|
|
return 0;
|
|
}
|
|
|
|
while (read_route_table(&rt) != 0)
|
|
{
|
|
if ((rt.rt_flags & RTF_UP) == 0)
|
|
{
|
|
continue;
|
|
}
|
|
|
|
if (((struct sockaddr_in *) &rt.rt_dst)->sin_addr.s_addr == 0L)
|
|
{
|
|
struct in_addr ina;
|
|
ina.s_addr = ((struct sockaddr_in *) &rt.rt_gateway)->sin_addr.s_addr;
|
|
syslog (LOG_ERR,
|
|
"ppp not replacing existing default route to %s[%s]",
|
|
rt.rt_dev, inet_ntoa (ina) );
|
|
result = 1;
|
|
break;
|
|
}
|
|
}
|
|
|
|
close_route_table();
|
|
return result;
|
|
}
|
|
|
|
/*
|
|
* sifdefaultroute - assign a default route through the address given.
|
|
*/
|
|
|
|
int sifdefaultroute (int unit, int gateway)
|
|
{
|
|
struct rtentry rt;
|
|
|
|
if (has_default_route == 0) {
|
|
if (defaultroute_exists())
|
|
return 0;
|
|
|
|
memset (&rt, '\0', sizeof (rt));
|
|
SET_SA_FAMILY (rt.rt_dst, AF_INET);
|
|
SET_SA_FAMILY (rt.rt_gateway, AF_INET);
|
|
((struct sockaddr_in *) &rt.rt_gateway)->sin_addr.s_addr = gateway;
|
|
|
|
rt.rt_flags = RTF_UP | RTF_GATEWAY;
|
|
if (ioctl(sockfd, SIOCADDRT, &rt) < 0) {
|
|
syslog (LOG_ERR, "default route ioctl(SIOCADDRT): %m");
|
|
return 0;
|
|
}
|
|
}
|
|
has_default_route = 1;
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* cifdefaultroute - delete a default route through the address given.
|
|
*/
|
|
int cifdefaultroute (int unit, int gateway)
|
|
{
|
|
struct rtentry rt;
|
|
|
|
if (has_default_route) {
|
|
memset (&rt, '\0', sizeof (rt));
|
|
SET_SA_FAMILY (rt.rt_dst, AF_INET);
|
|
SET_SA_FAMILY (rt.rt_gateway, AF_INET);
|
|
((struct sockaddr_in *) &rt.rt_gateway)->sin_addr.s_addr = gateway;
|
|
|
|
rt.rt_flags = RTF_UP | RTF_GATEWAY;
|
|
if (ioctl(sockfd, SIOCDELRT, &rt) < 0 && errno != ESRCH) {
|
|
if (still_ppp(unit)) {
|
|
syslog (LOG_ERR, "default route ioctl(SIOCDELRT): %m");
|
|
return 0;
|
|
}
|
|
}
|
|
}
|
|
has_default_route = 0;
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* sifproxyarp - Make a proxy ARP entry for the peer.
|
|
*/
|
|
int sifproxyarp (int linkunit, u_int32_t his_adr)
|
|
{
|
|
struct arpreq arpreq;
|
|
|
|
if (lns[linkunit].has_proxy_arp == 0) {
|
|
memset (&arpreq, '\0', sizeof(arpreq));
|
|
/*
|
|
* Get the hardware address of an interface on the same subnet
|
|
* as our local address.
|
|
*/
|
|
if (!get_ether_addr(his_adr, &arpreq.arp_ha, arpreq.arp_dev)) {
|
|
syslog(LOG_ERR, "Cannot determine ethernet address for proxy ARP");
|
|
return 0;
|
|
}
|
|
|
|
SET_SA_FAMILY(arpreq.arp_pa, AF_INET);
|
|
((struct sockaddr_in *) &arpreq.arp_pa)->sin_addr.s_addr = his_adr;
|
|
arpreq.arp_flags = ATF_PERM | ATF_PUBL;
|
|
|
|
if (ioctl(sockfd, SIOCSARP, (caddr_t)&arpreq) < 0) {
|
|
syslog(LOG_ERR, "ioctl(SIOCSARP): %m");
|
|
return 0;
|
|
}
|
|
}
|
|
lns[linkunit].has_proxy_arp = 1;
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* cifproxyarp - Delete the proxy ARP entry for the peer.
|
|
*/
|
|
int cifproxyarp (int linkunit, u_int32_t his_adr)
|
|
{
|
|
struct arpreq arpreq;
|
|
|
|
if (lns[linkunit].has_proxy_arp == 1) {
|
|
memset (&arpreq, '\0', sizeof(arpreq));
|
|
SET_SA_FAMILY(arpreq.arp_pa, AF_INET);
|
|
arpreq.arp_flags = ATF_PERM | ATF_PUBL;
|
|
|
|
((struct sockaddr_in *) &arpreq.arp_pa)->sin_addr.s_addr = his_adr;
|
|
if (ioctl(sockfd, SIOCDARP, (caddr_t)&arpreq) < 0) {
|
|
syslog(LOG_WARNING, "ioctl(SIOCDARP): %m");
|
|
return 0;
|
|
}
|
|
}
|
|
lns[linkunit].has_proxy_arp = 0;
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* get_ether_addr - get the hardware address of an interface on the
|
|
* the same subnet as ipaddr.
|
|
*/
|
|
static int local_get_ether_addr (u_int32_t ipaddr, struct sockaddr *hwaddr,
|
|
char *name, struct ifreq *ifs, int ifs_len)
|
|
{
|
|
struct ifreq *ifr, *ifend;
|
|
u_int32_t ina, mask;
|
|
struct ifreq ifreq;
|
|
struct ifconf ifc;
|
|
/*
|
|
* Request the total list of all devices configured on your system.
|
|
*/
|
|
ifc.ifc_len = ifs_len;
|
|
ifc.ifc_req = ifs;
|
|
if (ioctl(sockfd, SIOCGIFCONF, &ifc) < 0) {
|
|
syslog(LOG_ERR, "ioctl(SIOCGIFCONF): %m");
|
|
return 0;
|
|
}
|
|
|
|
MAINDEBUG ((LOG_DEBUG, "proxy arp: scanning %d interfaces for IP %s",
|
|
ifc.ifc_len / sizeof(struct ifreq), ip_ntoa(ipaddr)));
|
|
/*
|
|
* Scan through looking for an interface with an Internet
|
|
* address on the same subnet as `ipaddr'.
|
|
*/
|
|
ifend = ifs + (ifc.ifc_len / sizeof(struct ifreq));
|
|
for (ifr = ifc.ifc_req; ifr < ifend; ifr++) {
|
|
if (ifr->ifr_addr.sa_family == AF_INET) {
|
|
ina = ((struct sockaddr_in *) &ifr->ifr_addr)->sin_addr.s_addr;
|
|
strncpy(ifreq.ifr_name, ifr->ifr_name, sizeof(ifreq.ifr_name));
|
|
MAINDEBUG ((LOG_DEBUG, "proxy arp: examining interface %s",
|
|
ifreq.ifr_name));
|
|
/*
|
|
* Check that the interface is up, and not point-to-point
|
|
* nor loopback.
|
|
*/
|
|
if (ioctl(sockfd, SIOCGIFFLAGS, &ifreq) < 0)
|
|
continue;
|
|
if (((ifreq.ifr_flags ^ FLAGS_GOOD) & FLAGS_MASK) != 0)
|
|
continue;
|
|
/*
|
|
* Get its netmask and check that it's on the right subnet.
|
|
*/
|
|
if (ioctl(sockfd, SIOCGIFNETMASK, &ifreq) < 0)
|
|
continue;
|
|
|
|
mask = ((struct sockaddr_in *) &ifreq.ifr_addr)->sin_addr.s_addr;
|
|
MAINDEBUG ((LOG_DEBUG, "proxy arp: interface addr %s mask %lx",
|
|
ip_ntoa(ina), ntohl(mask)));
|
|
|
|
if (((ipaddr ^ ina) & mask) != 0)
|
|
continue;
|
|
break;
|
|
}
|
|
}
|
|
|
|
if(ifr >= ifend)
|
|
return 0;
|
|
|
|
memcpy (name, ifreq.ifr_name, sizeof(ifreq.ifr_name));
|
|
syslog(LOG_INFO, "found interface %s for proxy arp", name);
|
|
/*
|
|
* Now get the hardware address.
|
|
*/
|
|
memset (&ifreq.ifr_hwaddr, 0, sizeof (struct sockaddr));
|
|
if (ioctl (sockfd, SIOCGIFHWADDR, &ifreq) < 0) {
|
|
syslog(LOG_ERR, "SIOCGIFHWADDR(%s): %m", ifreq.ifr_name);
|
|
return 0;
|
|
}
|
|
|
|
memcpy (hwaddr, &ifreq.ifr_hwaddr, sizeof (struct sockaddr));
|
|
|
|
MAINDEBUG ((LOG_DEBUG,
|
|
"proxy arp: found hwaddr %02x:%02x:%02x:%02x:%02x:%02x:%02x:%02x",
|
|
(int) ((unsigned char *) &hwaddr->sa_data)[0],
|
|
(int) ((unsigned char *) &hwaddr->sa_data)[1],
|
|
(int) ((unsigned char *) &hwaddr->sa_data)[2],
|
|
(int) ((unsigned char *) &hwaddr->sa_data)[3],
|
|
(int) ((unsigned char *) &hwaddr->sa_data)[4],
|
|
(int) ((unsigned char *) &hwaddr->sa_data)[5],
|
|
(int) ((unsigned char *) &hwaddr->sa_data)[6],
|
|
(int) ((unsigned char *) &hwaddr->sa_data)[7]));
|
|
return 1;
|
|
}
|
|
|
|
int get_ether_addr (u_int32_t ipaddr, struct sockaddr *hwaddr, char *name)
|
|
{
|
|
int ifs_len;
|
|
int answer;
|
|
void *base_addr;
|
|
/*
|
|
* Allocate memory to hold the request.
|
|
*/
|
|
ifs_len = MAX_IFS * sizeof (struct ifreq);
|
|
base_addr = (void *) malloc (ifs_len);
|
|
if (!base_addr) {
|
|
syslog(LOG_ERR, "malloc(%d) failed to return memory", ifs_len);
|
|
return 0;
|
|
}
|
|
/*
|
|
* Find the hardware address associated with the controller
|
|
*/
|
|
answer = local_get_ether_addr (ipaddr, hwaddr, name,
|
|
(struct ifreq *) base_addr, ifs_len);
|
|
|
|
free (base_addr);
|
|
return answer;
|
|
}
|
|
|
|
/*
|
|
* Return user specified netmask, modified by any mask we might determine
|
|
* for address `addr' (in network byte order).
|
|
* Here we scan through the system's list of interfaces, looking for
|
|
* any non-point-to-point interfaces which might appear to be on the same
|
|
* network as `addr'. If we find any, we OR in their netmask to the
|
|
* user-specified netmask.
|
|
*
|
|
* If our address happens to be in the same network/subnet as one of
|
|
* the other interfaces, set the netmask to 255.255.255.255
|
|
*
|
|
* FIXME: This stuff is OLD and should be rewritten (we now use CIDR remember).
|
|
*/
|
|
|
|
static u_int32_t local_GetMask (u_int32_t addr, struct ifreq *ifs, int ifs_len)
|
|
{
|
|
u_int32_t mask, nmask, ina;
|
|
struct ifreq *ifr, *ifend, ifreq;
|
|
struct ifconf ifc;
|
|
|
|
ina = ntohl(addr);
|
|
|
|
if (IN_CLASSA(ina)) { /* determine network mask for address class */
|
|
nmask = IN_CLASSA_NET;
|
|
} else {
|
|
if (IN_CLASSB(ina)) {
|
|
nmask = IN_CLASSB_NET;
|
|
} else {
|
|
nmask = IN_CLASSC_NET;
|
|
}
|
|
}
|
|
nmask = htonl(nmask);
|
|
|
|
/* class D nets are disallowed by bad_ip_adrs */
|
|
mask = netmask | nmask;
|
|
|
|
if (ifs == (void *) 0) {
|
|
return mask;
|
|
}
|
|
/*
|
|
* Scan through the system's network interfaces.
|
|
*/
|
|
ifc.ifc_len = ifs_len;
|
|
ifc.ifc_req = ifs;
|
|
if (ioctl(sockfd, SIOCGIFCONF, &ifc) < 0) {
|
|
syslog(LOG_WARNING, "ioctl(SIOCGIFCONF): %m");
|
|
return mask;
|
|
}
|
|
|
|
ifend = (struct ifreq *) (ifc.ifc_buf + ifc.ifc_len);
|
|
for (ifr = ifc.ifc_req; ifr < ifend; ifr++) {
|
|
if (ifr->ifr_addr.sa_family != AF_INET) {
|
|
continue;
|
|
}
|
|
/*
|
|
* Check that the interface is up, and not point-to-point nor loopback.
|
|
*/
|
|
strncpy(ifreq.ifr_name, ifr->ifr_name, sizeof(ifreq.ifr_name));
|
|
if (ioctl(sockfd, SIOCGIFFLAGS, &ifreq) < 0)
|
|
continue;
|
|
|
|
if (((ifreq.ifr_flags ^ FLAGS_GOOD) & FLAGS_MASK) != 0)
|
|
continue;
|
|
/*
|
|
* See if our IP address is part of this subnet.
|
|
*/
|
|
ina = ((struct sockaddr_in *) &ifr->ifr_addr)->sin_addr.s_addr;
|
|
if (ina == addr)
|
|
continue;
|
|
if ((ina & nmask) == (addr & nmask)) {
|
|
mask = 0xFFFFFFFF;
|
|
break;
|
|
}
|
|
/*
|
|
* Check the interface's internet address.
|
|
*/
|
|
if (((ina ^ addr) & nmask) != 0)
|
|
continue;
|
|
/*
|
|
* Get its netmask and OR it into our mask.
|
|
*/
|
|
if (ioctl(sockfd, SIOCGIFNETMASK, &ifreq) < 0)
|
|
continue;
|
|
|
|
mask |= ((struct sockaddr_in *)&ifreq.ifr_addr)->sin_addr.s_addr;
|
|
break;
|
|
}
|
|
|
|
return mask;
|
|
}
|
|
|
|
u_int32_t GetMask (u_int32_t addr)
|
|
{
|
|
int ifs_len;
|
|
u_int32_t answer;
|
|
void *base_addr;
|
|
|
|
/*
|
|
* If user set netmask assume he knows what he's doing.
|
|
*/
|
|
if (netmask) return netmask;
|
|
|
|
/*
|
|
* Allocate memory to hold the request.
|
|
*/
|
|
ifs_len = MAX_IFS * sizeof (struct ifreq);
|
|
base_addr = (void *) malloc (ifs_len);
|
|
if (base_addr == (void *) 0)
|
|
{
|
|
syslog(LOG_ERR, "malloc(%d) failed to return memory", ifs_len);
|
|
}
|
|
/*
|
|
* Find the netmask used on the same network.
|
|
*/
|
|
answer = local_GetMask (addr, (struct ifreq *) base_addr, ifs_len);
|
|
if (base_addr != (void *) 0)
|
|
{
|
|
free (base_addr);
|
|
}
|
|
return answer;
|
|
}
|
|
|
|
/*
|
|
* Internal routine to decode the version.modification.patch level
|
|
*/
|
|
|
|
static void decode_version (char *buf, int *version,
|
|
int *modification, int *patch)
|
|
{
|
|
*version = (int) strtoul (buf, &buf, 10);
|
|
*modification = 0;
|
|
*patch = 0;
|
|
|
|
if (*buf == '.')
|
|
{
|
|
++buf;
|
|
*modification = (int) strtoul (buf, &buf, 10);
|
|
if (*buf == '.')
|
|
{
|
|
++buf;
|
|
*patch = (int) strtoul (buf, &buf, 10);
|
|
}
|
|
}
|
|
|
|
if (*buf != '\0')
|
|
{
|
|
*version =
|
|
*modification =
|
|
*patch = 0;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* ppp_available - check whether the system has any ppp interfaces
|
|
* (in fact we check whether we can do an ioctl on ppp0).
|
|
*/
|
|
int ppp_available(void)
|
|
{
|
|
int s;
|
|
struct ifreq ifr;
|
|
char abBuffer [1024];
|
|
int my_version, my_modification, my_patch;
|
|
|
|
/*
|
|
* Open a socket for doing the ioctl operations.
|
|
*/
|
|
if( (s = socket(AF_INET, SOCK_DGRAM, 0)) < 0)
|
|
return 0;
|
|
|
|
strncpy (ifr.ifr_name, "ippp0", sizeof (ifr.ifr_name));
|
|
if(ioctl(s, SIOCGIFFLAGS, (caddr_t) &ifr) < 0)
|
|
return 0;
|
|
|
|
/*
|
|
* This is the PPP device. Validate the version of the driver at this
|
|
* point to ensure that this program will work with the driver.
|
|
*/
|
|
ifr.ifr_data = abBuffer;
|
|
if (ioctl (s, SIOCGPPPVER, (caddr_t) &ifr) >= 0)
|
|
decode_version (abBuffer, &driver_version, &driver_modification, &driver_patch);
|
|
else
|
|
driver_version = driver_modification = driver_patch = 0;
|
|
|
|
close(s);
|
|
|
|
/*
|
|
* Validate the version of the driver against the version that we used.
|
|
*/
|
|
decode_version (IPPP_VERSION, &my_version, &my_modification, &my_patch);
|
|
|
|
/* The version numbers must match and the modification levels must be legal */
|
|
if ( (driver_version != my_version || driver_modification < my_modification) && !force_driver)
|
|
{
|
|
extern char *no_ppp_msg;
|
|
no_ppp_msg = route_buffer;
|
|
|
|
sprintf(no_ppp_msg,
|
|
"Sorry - isdnPPP driver version %d.%d.%d is out of date.\n"
|
|
"Maybe ippp0 has no 'syncppp' encapsulation?\n",
|
|
driver_version, driver_modification, driver_patch);
|
|
return 0;
|
|
}
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* Update the wtmp file with the appropriate user name and tty device.
|
|
*/
|
|
int logwtmputmp (int unit,char *line, char *name, char *host)
|
|
{
|
|
struct utmp ut, *utp;
|
|
pid_t mypid = getpid();
|
|
|
|
/*
|
|
* Update the signon database for users.
|
|
* Christoph Lameter: Copied from poeigl-1.36 Jan 3, 1996
|
|
*/
|
|
utmpname(_PATH_UTMP);
|
|
setutent();
|
|
while( (utp = getutent()) && (utp->ut_pid != mypid) )
|
|
;
|
|
|
|
/*
|
|
* Is this call really necessary? There is another one after the 'put'
|
|
*/
|
|
endutent();
|
|
|
|
if (utp)
|
|
memcpy(&ut, utp, sizeof(ut));
|
|
else {
|
|
/* some gettys/telnetds don't initialize utmp... */
|
|
memset(&ut, 0, sizeof(ut));
|
|
}
|
|
|
|
if (ut.ut_id[0] == 0)
|
|
strncpy(ut.ut_id, line + 3, sizeof(ut.ut_id));
|
|
|
|
strncpy(ut.ut_user, name, sizeof(ut.ut_user));
|
|
strncpy(ut.ut_line, line, sizeof(ut.ut_line));
|
|
|
|
ut.ut_time = time((void *)0);
|
|
|
|
ut.ut_type = USER_PROCESS;
|
|
ut.ut_pid = mypid;
|
|
|
|
/*
|
|
* Insert the host name if one is supplied
|
|
*/
|
|
if (*host)
|
|
strncpy (ut.ut_host, host, sizeof(ut.ut_host));
|
|
|
|
/*
|
|
* Insert the IP address of the remote system if IP is enabled
|
|
*/
|
|
if (ipcp_hisoptions[unit].neg_addr)
|
|
memcpy (&ut.ut_addr, (char *) &ipcp_hisoptions[unit].hisaddr,
|
|
sizeof(ut.ut_addr));
|
|
|
|
/*
|
|
* CL: Makes sure that the logout works
|
|
*/
|
|
if (*host == 0 && *name==0)
|
|
ut.ut_host[0]=0;
|
|
|
|
pututline(&ut);
|
|
endutent();
|
|
|
|
/*
|
|
* Update the wtmp file.
|
|
*/
|
|
#if (defined __GLIBC__ && __GLIBC__ >= 2)
|
|
updwtmp (_PATH_WTMP, &ut);
|
|
#else
|
|
{
|
|
int wtmp = open(_PATH_WTMP, O_APPEND|O_WRONLY);
|
|
if (wtmp >= 0) {
|
|
flock(wtmp, LOCK_EX);
|
|
/*
|
|
* we really should check for error on
|
|
* the write for a full disk!
|
|
*/
|
|
write (wtmp, (char *)&ut, sizeof(ut));
|
|
close (wtmp);
|
|
flock(wtmp, LOCK_UN);
|
|
}
|
|
}
|
|
#endif
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* Code for locking/unlocking the serial device.
|
|
* This code is derived from chat.c.
|
|
*/
|
|
|
|
#ifndef LOCK_PREFIX
|
|
#define LOCK_PREFIX "/var/lock/LCK.."
|
|
#endif
|
|
|
|
/*
|
|
* lock - create a lock file for the named device
|
|
*/
|
|
|
|
int lock (char *dev)
|
|
{
|
|
char hdb_lock_buffer[12];
|
|
int lfd, pid, n;
|
|
char *p;
|
|
|
|
p = strrchr(dev, '/');
|
|
if (p != NULL)
|
|
{
|
|
dev = ++p;
|
|
}
|
|
|
|
lock_file = malloc(strlen(LOCK_PREFIX) + strlen(dev) + 1);
|
|
if (lock_file == NULL)
|
|
{
|
|
novm("lock file name");
|
|
}
|
|
|
|
strcpy (lock_file, LOCK_PREFIX);
|
|
strcat (lock_file, dev);
|
|
/*
|
|
* Attempt to create the lock file at this point.
|
|
*/
|
|
while (1)
|
|
{
|
|
lfd = open(lock_file, O_EXCL | O_CREAT | O_RDWR, 0644);
|
|
if (lfd >= 0)
|
|
{
|
|
pid = getpid();
|
|
#ifndef PID_BINARY
|
|
sprintf (hdb_lock_buffer, "%010d\n", pid);
|
|
write (lfd, hdb_lock_buffer, 11);
|
|
#else
|
|
write(lfd, &pid, sizeof (pid));
|
|
#endif
|
|
close(lfd);
|
|
return 0;
|
|
}
|
|
/*
|
|
* If the file exists then check to see if the pid is stale
|
|
*/
|
|
if (errno == EEXIST)
|
|
{
|
|
lfd = open(lock_file, O_RDONLY, 0);
|
|
if (lfd < 0)
|
|
{
|
|
if (errno == ENOENT) /* This is just a timing problem. */
|
|
{
|
|
continue;
|
|
}
|
|
break;
|
|
}
|
|
|
|
/* Read the lock file to find out who has the device locked */
|
|
n = read (lfd, hdb_lock_buffer, 11);
|
|
close (lfd);
|
|
if (n < 0)
|
|
{
|
|
syslog(LOG_ERR, "Can't read pid from lock file %s", lock_file);
|
|
break;
|
|
}
|
|
|
|
/* See the process still exists. */
|
|
if (n > 0)
|
|
{
|
|
#ifndef PID_BINARY
|
|
hdb_lock_buffer[n] = '\0';
|
|
sscanf (hdb_lock_buffer, " %d", &pid);
|
|
#else
|
|
pid = ((int *) hdb_lock_buffer)[0];
|
|
#endif
|
|
if (pid == 0 || (kill(pid, 0) == -1 && errno == ESRCH))
|
|
{
|
|
n = 0;
|
|
}
|
|
}
|
|
|
|
/* If the process does not exist then try to remove the lock */
|
|
if (n == 0 && unlink (lock_file) == 0)
|
|
{
|
|
syslog (LOG_NOTICE, "Removed stale lock on %s (pid %d)",
|
|
dev, pid);
|
|
continue;
|
|
}
|
|
|
|
syslog (LOG_NOTICE, "Device %s is locked by pid %d", dev, pid);
|
|
break;
|
|
}
|
|
|
|
syslog(LOG_ERR, "Can't create lock file %s: %m", lock_file);
|
|
break;
|
|
}
|
|
|
|
free(lock_file);
|
|
lock_file = NULL;
|
|
return -1;
|
|
}
|
|
|
|
/*
|
|
* unlock - remove our lockfile
|
|
*/
|
|
|
|
void unlock(void)
|
|
{
|
|
if (lock_file)
|
|
{
|
|
unlink(lock_file);
|
|
free(lock_file);
|
|
lock_file = NULL;
|
|
}
|
|
}
|
|
|
|
void setifip(int ipcp_unit)
|
|
{
|
|
u_int32_t ouraddr,hisaddr;
|
|
struct ifreq ifr;
|
|
extern ipcp_options ipcp_wantoptions[NUM_PPP];
|
|
|
|
ipcp_options *wo = &ipcp_wantoptions[ipcp_unit];
|
|
|
|
memset (&ifr, '\0', sizeof (ifr));
|
|
strncpy(ifr.ifr_name,lns[ipcp_fsm[ipcp_unit].unit].ifname,sizeof(ifr.ifr_name));
|
|
|
|
if(ioctl(sockfd,SIOCGIFADDR,(caddr_t) &ifr) < 0)
|
|
{
|
|
syslog(LOG_ERR,"ioctl(SIOCGIFADDR): %m");
|
|
return;
|
|
}
|
|
|
|
ouraddr = ((struct sockaddr_in *) &ifr.ifr_addr)->sin_addr.s_addr;
|
|
|
|
if(debug)
|
|
syslog (LOG_DEBUG, "got if-src: %08lx\n",(unsigned long) ouraddr);
|
|
|
|
if(ouraddr)
|
|
wo->ouraddr = ouraddr;
|
|
|
|
if(ioctl(sockfd,SIOCGIFDSTADDR,(caddr_t) &ifr) < 0)
|
|
{
|
|
syslog(LOG_ERR,"ioctl(SIOCGIFDSTADDR): %m");
|
|
return;
|
|
}
|
|
|
|
hisaddr = ((struct sockaddr_in *) &ifr.ifr_dstaddr)->sin_addr.s_addr;
|
|
|
|
if(debug)
|
|
syslog (LOG_DEBUG, "got if-dst: %08lx\n",(unsigned long) hisaddr);
|
|
|
|
if(hisaddr)
|
|
wo->hisaddr = hisaddr;
|
|
}
|
|
|
|
|
|
/************************ IPX SUPPORT *********************************/
|
|
#if defined(__GLIBC__) && (__GLIBC__ > 1)
|
|
/* <linux/ipx.h> includes <linux/socket.h>, which breaks glibc 2.x support. Prevent that... */
|
|
# define _LINUX_SOCKET_H
|
|
#endif
|
|
|
|
#include <linux/ipx.h>
|
|
|
|
/*
|
|
* sipxfaddr - Config the interface IPX networknumber
|
|
*/
|
|
|
|
int sipxfaddr (int unit, u_int32_t network, unsigned char * node )
|
|
{
|
|
int skfd;
|
|
int result = 1;
|
|
/* struct sockaddr_ipx ipx_addr; */
|
|
struct ifreq ifr;
|
|
struct sockaddr_ipx *sipx = (struct sockaddr_ipx *) &ifr.ifr_addr;
|
|
|
|
skfd = socket (AF_IPX, SOCK_DGRAM, 0);
|
|
if (skfd < 0)
|
|
{
|
|
syslog (LOG_DEBUG, "socket(AF_IPX): %m");
|
|
result = 0;
|
|
}
|
|
else
|
|
{
|
|
memset (&ifr, '\0', sizeof (ifr));
|
|
strcpy (ifr.ifr_name, lns[unit].ifname);
|
|
|
|
memcpy (sipx->sipx_node, node, IPX_NODE_LEN);
|
|
sipx->sipx_family = AF_IPX;
|
|
sipx->sipx_port = 0;
|
|
sipx->sipx_network = htonl (network);
|
|
sipx->sipx_type = IPX_FRAME_ETHERII;
|
|
sipx->sipx_action = IPX_CRTITF;
|
|
/*
|
|
* Set the IPX device
|
|
*/
|
|
if (ioctl(skfd, SIOCSIFADDR, (caddr_t) &ifr) < 0)
|
|
{
|
|
result = 0;
|
|
if (errno != EEXIST)
|
|
{
|
|
syslog (LOG_DEBUG, "ioctl(SIOCAIFADDR, CRTITF): %m");
|
|
}
|
|
else
|
|
{
|
|
syslog (LOG_WARNING,
|
|
"ioctl(SIOCAIFADDR, CRTITF): Address already exists");
|
|
}
|
|
}
|
|
close (skfd);
|
|
}
|
|
return result;
|
|
}
|
|
|
|
/*
|
|
* cipxfaddr - Clear the information for the IPX network. The IPX routes
|
|
* are removed and the device is no longer able to pass IPX
|
|
* frames.
|
|
*/
|
|
|
|
int cipxfaddr (int linkunit)
|
|
{
|
|
int skfd;
|
|
int result = 1;
|
|
/* struct sockaddr_ipx ipx_addr; */
|
|
struct ifreq ifr;
|
|
struct sockaddr_ipx *sipx = (struct sockaddr_ipx *) &ifr.ifr_addr;
|
|
|
|
skfd = socket (AF_IPX, SOCK_DGRAM, 0);
|
|
if (skfd < 0)
|
|
{
|
|
syslog (LOG_DEBUG, "socket(AF_IPX): %m");
|
|
result = 0;
|
|
}
|
|
else
|
|
{
|
|
memset (&ifr, '\0', sizeof (ifr));
|
|
strcpy (ifr.ifr_name, lns[linkunit].ifname);
|
|
|
|
sipx->sipx_type = IPX_FRAME_ETHERII;
|
|
sipx->sipx_action = IPX_DLTITF;
|
|
sipx->sipx_family = AF_IPX;
|
|
/*
|
|
* Set the IPX device
|
|
*/
|
|
if (ioctl(skfd, SIOCSIFADDR, (caddr_t) &ifr) < 0)
|
|
{
|
|
syslog (LOG_INFO, "ioctl(SIOCAIFADDR, IPX_DLTITF) %d: %m", errno);
|
|
result = 0;
|
|
}
|
|
close (skfd);
|
|
}
|
|
return result;
|
|
}
|
|
|
|
/*
|
|
* Turn off any option which is not supported by this implementation.
|
|
*
|
|
* This procedure is called after all of the options have been processed.
|
|
* It gives the implementation a chance to alter the configurtion options
|
|
* based upon the current support by the operating system.
|
|
*/
|
|
|
|
void remove_sys_options(void)
|
|
{
|
|
struct stat stat_buf;
|
|
/*
|
|
* Disable the IPX protocol if the support is not present in the kernel.
|
|
* If we disable it then ensure that IP support is enabled.
|
|
*/
|
|
while (ipxcp_protent.enabled_flag) {
|
|
char *path = path_to_proc();
|
|
if (path != NULL) {
|
|
strcat (path, "/net/ipx_interface");
|
|
if (lstat (path, &stat_buf) >= 0)
|
|
break;
|
|
}
|
|
syslog (LOG_ERR, "IPX support is not present in the kernel\n");
|
|
ipxcp_protent.enabled_flag = 0;
|
|
ipcp_protent.enabled_flag = 1;
|
|
break;
|
|
}
|
|
}
|
|
|
|
|