2189 lines
54 KiB
Text
2189 lines
54 KiB
Text
$NetBSD: patch-az,v 1.4 2008/12/26 21:41:40 hasso Exp $
|
|
|
|
--- pppd/sys-bsd.c.orig 2008-12-26 23:11:29 +0200
|
|
+++ pppd/sys-bsd.c 2008-12-26 23:12:45 +0200
|
|
@@ -0,0 +1,2184 @@
|
|
+/* NetBSD: sys-bsd.c,v 1.50 2004/12/05 04:16:22 christos Exp */
|
|
+
|
|
+/*
|
|
+ * sys-bsd.c - System-dependent procedures for setting up
|
|
+ * PPP interfaces on bsd-4.4-ish systems (including 386BSD, NetBSD, etc.)
|
|
+ *
|
|
+ * Copyright (c) 1984-2000 Carnegie Mellon University. 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.
|
|
+ *
|
|
+ * 3. The name "Carnegie Mellon University" must not be used to
|
|
+ * endorse or promote products derived from this software without
|
|
+ * prior written permission. For permission or any legal
|
|
+ * details, please contact
|
|
+ * Office of Technology Transfer
|
|
+ * Carnegie Mellon University
|
|
+ * 5000 Forbes Avenue
|
|
+ * Pittsburgh, PA 15213-3890
|
|
+ * (412) 268-4387, fax: (412) 268-7395
|
|
+ * tech-transfer@andrew.cmu.edu
|
|
+ *
|
|
+ * 4. Redistributions of any form whatsoever must retain the following
|
|
+ * acknowledgment:
|
|
+ * "This product includes software developed by Computing Services
|
|
+ * at Carnegie Mellon University (http://www.cmu.edu/computing/)."
|
|
+ *
|
|
+ * CARNEGIE MELLON UNIVERSITY DISCLAIMS ALL WARRANTIES WITH REGARD TO
|
|
+ * THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY
|
|
+ * AND FITNESS, IN NO EVENT SHALL CARNEGIE MELLON UNIVERSITY BE LIABLE
|
|
+ * FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
|
|
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN
|
|
+ * AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING
|
|
+ * OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
|
|
+ *
|
|
+ * Copyright (c) 1989-2002 Paul Mackerras. 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.
|
|
+ *
|
|
+ * 3. The name(s) of the authors of this software must not be used to
|
|
+ * endorse or promote products derived from this software without
|
|
+ * prior written permission.
|
|
+ *
|
|
+ * 4. Redistributions of any form whatsoever must retain the following
|
|
+ * acknowledgment:
|
|
+ * "This product includes software developed by Paul Mackerras
|
|
+ * <paulus@samba.org>".
|
|
+ *
|
|
+ * THE AUTHORS OF THIS SOFTWARE DISCLAIM ALL WARRANTIES WITH REGARD TO
|
|
+ * THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY
|
|
+ * AND FITNESS, IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY
|
|
+ * SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
|
|
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN
|
|
+ * AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING
|
|
+ * OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
|
|
+ */
|
|
+
|
|
+#include <sys/cdefs.h>
|
|
+#ifndef lint
|
|
+#if 0
|
|
+#define RCSID "Id: sys-bsd.c,v 1.47 2000/04/13 12:04:23 paulus Exp "
|
|
+#else
|
|
+__RCSID("NetBSD: sys-bsd.c,v 1.50 2004/12/05 04:16:22 christos Exp ");
|
|
+#endif
|
|
+#endif
|
|
+
|
|
+/*
|
|
+ * TODO:
|
|
+ */
|
|
+
|
|
+#include <stdio.h>
|
|
+#include <string.h>
|
|
+#include <stdlib.h>
|
|
+#include <unistd.h>
|
|
+#include <errno.h>
|
|
+#include <fcntl.h>
|
|
+#include <termios.h>
|
|
+#include <signal.h>
|
|
+#include <vis.h>
|
|
+#include <sys/ioctl.h>
|
|
+#include <sys/types.h>
|
|
+#include <sys/socket.h>
|
|
+#include <sys/time.h>
|
|
+#include <sys/stat.h>
|
|
+#include <sys/param.h>
|
|
+#if defined(NetBSD1_2) || defined(__NetBSD_Version__)
|
|
+#include <util.h>
|
|
+#endif
|
|
+#ifdef PPP_FILTER
|
|
+#include <net/bpf.h>
|
|
+#endif
|
|
+
|
|
+#include <net/if.h>
|
|
+#include <net/ppp_defs.h>
|
|
+#include <net/if_ppp.h>
|
|
+#include <net/route.h>
|
|
+#include <net/if_dl.h>
|
|
+#include <netinet/in.h>
|
|
+#ifdef __KAME__
|
|
+#include <netinet6/in6_var.h>
|
|
+#include <netinet6/nd6.h>
|
|
+#endif
|
|
+#include <ifaddrs.h>
|
|
+
|
|
+#if RTM_VERSION >= 3
|
|
+#include <sys/param.h>
|
|
+#if defined(NetBSD) && (NetBSD >= 199703)
|
|
+#include <netinet/if_inarp.h>
|
|
+#else /* NetBSD 1.2D or later */
|
|
+#if defined(__FreeBSD__) || defined(__DragonFly__)
|
|
+#include <netinet/if_ether.h>
|
|
+#else
|
|
+#include <net/if_ether.h>
|
|
+#endif
|
|
+#endif
|
|
+#endif
|
|
+
|
|
+#include "pppd.h"
|
|
+#include "fsm.h"
|
|
+#include "ipcp.h"
|
|
+
|
|
+#ifdef RCSID
|
|
+static const char rcsid[] = RCSID;
|
|
+#endif
|
|
+
|
|
+static int initdisc = -1; /* Initial TTY discipline for ppp_fd */
|
|
+static int initfdflags = -1; /* Initial file descriptor flags for ppp_fd */
|
|
+static int ppp_fd = -1; /* fd which is set to PPP discipline */
|
|
+static int rtm_seq;
|
|
+
|
|
+static int restore_term; /* 1 => we've munged the terminal */
|
|
+static struct termios inittermios; /* Initial TTY termios */
|
|
+static struct winsize wsinfo; /* Initial window size info */
|
|
+
|
|
+static int loop_slave = -1;
|
|
+static int loop_master;
|
|
+static char loop_name[20];
|
|
+
|
|
+static unsigned char inbuf[512]; /* buffer for chars read from loopback */
|
|
+
|
|
+static int sock_fd; /* socket for doing interface ioctls */
|
|
+#ifdef INET6
|
|
+static int sock6_fd = -1; /* socket for doing ipv6 interface ioctls */
|
|
+#endif /* INET6 */
|
|
+static int ttyfd = -1; /* the file descriptor of the tty */
|
|
+
|
|
+static fd_set in_fds; /* set of fds that wait_input waits for */
|
|
+static int max_in_fd; /* highest fd set in in_fds */
|
|
+
|
|
+static int if_is_up; /* the interface is currently up */
|
|
+static u_int32_t ifaddrs[2]; /* local and remote addresses we set */
|
|
+static u_int32_t default_route_gateway; /* gateway addr for default route */
|
|
+static u_int32_t proxy_arp_addr; /* remote addr for proxy arp */
|
|
+
|
|
+/* Prototypes for procedures local to this file. */
|
|
+static int get_flags __P((int));
|
|
+static void set_flags __P((int, int));
|
|
+static int dodefaultroute __P((u_int32_t, int));
|
|
+static int get_ether_addr __P((u_int32_t, struct sockaddr_dl *));
|
|
+static void restore_loop(void); /* Transfer ppp unit back to loopback */
|
|
+
|
|
+
|
|
+/********************************************************************
|
|
+ *
|
|
+ * Functions to read and set the flags value in the device driver
|
|
+ */
|
|
+
|
|
+static int
|
|
+get_flags(fd)
|
|
+ int fd;
|
|
+{
|
|
+ int flags;
|
|
+
|
|
+ if (ioctl(fd, PPPIOCGFLAGS, (caddr_t) &flags) == -1)
|
|
+ fatal("ioctl(PPPIOCGFLAGS): %m");
|
|
+
|
|
+ SYSDEBUG((LOG_DEBUG, "get flags = %x\n", flags));
|
|
+ return flags;
|
|
+}
|
|
+
|
|
+/********************************************************************/
|
|
+
|
|
+static void
|
|
+set_flags(fd, flags)
|
|
+ int fd, flags;
|
|
+{
|
|
+ SYSDEBUG((LOG_DEBUG, "set flags = %x\n", flags));
|
|
+
|
|
+ if (ioctl(fd, PPPIOCSFLAGS, (caddr_t) &flags) == -1)
|
|
+ fatal("ioctl(PPPIOCSFLAGS, %x): %m", flags, errno);
|
|
+}
|
|
+
|
|
+/*
|
|
+ * sys_init - System-dependent initialization.
|
|
+ */
|
|
+void
|
|
+sys_init()
|
|
+{
|
|
+ /* Get an internet socket for doing socket ioctl's on. */
|
|
+ if ((sock_fd = socket(AF_INET, SOCK_DGRAM, 0)) < 0)
|
|
+ fatal("Couldn't create IP socket: %m");
|
|
+
|
|
+#ifdef INET6
|
|
+ if ((sock6_fd = socket(AF_INET6, SOCK_DGRAM, 0)) < 0) {
|
|
+ /* check it at runtime */
|
|
+ sock6_fd = -1;
|
|
+ }
|
|
+#endif
|
|
+
|
|
+ FD_ZERO(&in_fds);
|
|
+ max_in_fd = 0;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * sys_cleanup - restore any system state we modified before exiting:
|
|
+ * mark the interface down, delete default route and/or proxy arp entry.
|
|
+ * This should call die() because it's called from die().
|
|
+ */
|
|
+void
|
|
+sys_cleanup()
|
|
+{
|
|
+ struct ifreq ifr;
|
|
+
|
|
+ if (if_is_up) {
|
|
+ strlcpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
|
|
+ if (ioctl(sock_fd, SIOCGIFFLAGS, &ifr) >= 0
|
|
+ && ((ifr.ifr_flags & IFF_UP) != 0)) {
|
|
+ ifr.ifr_flags &= ~IFF_UP;
|
|
+ ioctl(sock_fd, SIOCSIFFLAGS, &ifr);
|
|
+ }
|
|
+ }
|
|
+ if (ifaddrs[0] != 0)
|
|
+ cifaddr(0, ifaddrs[0], ifaddrs[1]);
|
|
+ if (default_route_gateway)
|
|
+ cifdefaultroute(0, 0, default_route_gateway);
|
|
+ if (proxy_arp_addr)
|
|
+ cifproxyarp(0, proxy_arp_addr);
|
|
+}
|
|
+
|
|
+/*
|
|
+ * sys_close - Clean up in a child process before execing.
|
|
+ */
|
|
+void
|
|
+sys_close()
|
|
+{
|
|
+ close(sock_fd);
|
|
+ if (loop_slave >= 0) {
|
|
+ close(loop_slave);
|
|
+ close(loop_master);
|
|
+ }
|
|
+}
|
|
+
|
|
+/*
|
|
+ * sys_check_options - check the options that the user specified
|
|
+ */
|
|
+int
|
|
+sys_check_options()
|
|
+{
|
|
+#ifndef CDTRCTS
|
|
+ if (crtscts == 2) {
|
|
+ warn("DTR/CTS flow control is not supported on this system");
|
|
+ return 0;
|
|
+ }
|
|
+#endif
|
|
+ return 1;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * 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()
|
|
+{
|
|
+ struct if_clonereq ifcr;
|
|
+ char *cp, *buf;
|
|
+ int idx, s, ok;
|
|
+ extern char *no_ppp_msg;
|
|
+
|
|
+ (void)memset(&ifcr, 0, sizeof(ifcr));
|
|
+
|
|
+ if ((s = socket(AF_INET, SOCK_DGRAM, 0)) < 0)
|
|
+ fatal("socket: %m");
|
|
+
|
|
+ if (ioctl(s, SIOCIFGCLONERS, &ifcr) == -1)
|
|
+ fatal("ioctl(get cloners): %m");
|
|
+
|
|
+ buf = malloc(ifcr.ifcr_total * IFNAMSIZ);
|
|
+ if (buf == NULL)
|
|
+ fatal("Unable to allocate cloner name buffer: %m");
|
|
+
|
|
+ ifcr.ifcr_count = ifcr.ifcr_total;
|
|
+ ifcr.ifcr_buffer = buf;
|
|
+
|
|
+ if (ioctl(s, SIOCIFGCLONERS, &ifcr) == -1)
|
|
+ fatal("ioctl(get cloners): %m");
|
|
+
|
|
+ /*
|
|
+ * In case some disappeared in the mean time, clamp it down.
|
|
+ */
|
|
+ if (ifcr.ifcr_count > ifcr.ifcr_total)
|
|
+ ifcr.ifcr_count = ifcr.ifcr_total;
|
|
+
|
|
+ for (cp = buf, idx = 0; idx < ifcr.ifcr_count; idx++, cp += IFNAMSIZ) {
|
|
+ if (strcmp(cp, "ppp") == 0)
|
|
+ break;
|
|
+ }
|
|
+ free(buf);
|
|
+
|
|
+ if (idx == ifcr.ifcr_count) {
|
|
+ /*
|
|
+ * Try ppp0 directly, as before ppp was clonable.
|
|
+ */
|
|
+ struct ifreq ifr;
|
|
+ (void)memset(&ifr, 0, sizeof(ifr));
|
|
+ strlcpy(ifr.ifr_name, "ppp0", sizeof (ifr.ifr_name));
|
|
+ ok = ioctl(s, SIOCGIFFLAGS, (caddr_t) &ifr) != -1;
|
|
+ } else
|
|
+ ok = 1;
|
|
+
|
|
+ (void)close(s);
|
|
+
|
|
+#ifdef __NetBSD__
|
|
+ no_ppp_msg = "\
|
|
+This system lacks kernel support for PPP. To include PPP support\n\
|
|
+in the kernel, please read the ppp(4) manual page.\n";
|
|
+#else
|
|
+ no_ppp_msg = "\
|
|
+This system lacks kernel support for PPP. To include PPP support\n\
|
|
+in the kernel, please follow the steps detailed in the README.bsd\n\
|
|
+file in the ppp-2.2 distribution.\n";
|
|
+#endif
|
|
+ return ok;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * tty_establish_ppp - Turn the serial port into a ppp interface.
|
|
+ */
|
|
+int
|
|
+tty_establish_ppp(fd)
|
|
+ int fd;
|
|
+{
|
|
+ int pppdisc = PPPDISC;
|
|
+ int x;
|
|
+ ttyfd = fd;
|
|
+
|
|
+ if (demand) {
|
|
+ /*
|
|
+ * Demand mode - prime the old ppp device to relinquish the unit.
|
|
+ */
|
|
+ if (ioctl(ppp_fd, PPPIOCXFERUNIT, 0) < 0)
|
|
+ fatal("ioctl(transfer ppp unit): %m");
|
|
+ }
|
|
+
|
|
+ /*
|
|
+ * Save the old line discipline of fd, and set it to PPP.
|
|
+ */
|
|
+ if (ioctl(fd, TIOCGETD, &initdisc) < 0)
|
|
+ fatal("ioctl(TIOCGETD): %m");
|
|
+ if (ioctl(fd, TIOCSETD, &pppdisc) < 0)
|
|
+ fatal("ioctl(TIOCSETD): %m");
|
|
+
|
|
+ if (!demand) {
|
|
+ /*
|
|
+ * Find out which interface we were given.
|
|
+ */
|
|
+ if (ioctl(fd, PPPIOCGUNIT, &ifunit) < 0)
|
|
+ fatal("ioctl(PPPIOCGUNIT): %m");
|
|
+ } else {
|
|
+ /*
|
|
+ * Check that we got the same unit again.
|
|
+ */
|
|
+ if (ioctl(fd, PPPIOCGUNIT, &x) < 0)
|
|
+ fatal("ioctl(PPPIOCGUNIT): %m");
|
|
+ if (x != ifunit)
|
|
+ fatal("transfer_ppp failed: wanted unit %d, got %d", ifunit, x);
|
|
+ x = TTYDISC;
|
|
+ ioctl(loop_slave, TIOCSETD, &x);
|
|
+ }
|
|
+
|
|
+ ppp_fd = fd;
|
|
+
|
|
+ /*
|
|
+ * Enable debug in the driver if requested.
|
|
+ */
|
|
+ if (kdebugflag) {
|
|
+ x = get_flags(fd);
|
|
+ x |= (kdebugflag & 0xFF) * SC_DEBUG;
|
|
+ set_flags(fd, x);
|
|
+ }
|
|
+
|
|
+ /*
|
|
+ * Set device for non-blocking reads.
|
|
+ */
|
|
+ if ((initfdflags = fcntl(fd, F_GETFL)) == -1
|
|
+ || fcntl(fd, F_SETFL, initfdflags | O_NONBLOCK) == -1) {
|
|
+ warn("Couldn't set device to non-blocking mode: %m");
|
|
+ }
|
|
+
|
|
+ return fd;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * restore_loop - reattach the ppp unit to the loopback.
|
|
+ */
|
|
+static void
|
|
+restore_loop()
|
|
+{
|
|
+ int x;
|
|
+
|
|
+ /*
|
|
+ * Transfer the ppp interface back to the loopback.
|
|
+ */
|
|
+ if (ioctl(ppp_fd, PPPIOCXFERUNIT, 0) < 0)
|
|
+ fatal("ioctl(transfer ppp unit): %m");
|
|
+ x = PPPDISC;
|
|
+ if (ioctl(loop_slave, TIOCSETD, &x) < 0)
|
|
+ fatal("ioctl(TIOCSETD): %m");
|
|
+
|
|
+ /*
|
|
+ * Check that we got the same unit again.
|
|
+ */
|
|
+ if (ioctl(loop_slave, PPPIOCGUNIT, &x) < 0)
|
|
+ fatal("ioctl(PPPIOCGUNIT): %m");
|
|
+ if (x != ifunit)
|
|
+ fatal("transfer_ppp failed: wanted unit %d, got %d", ifunit, x);
|
|
+ ppp_fd = loop_slave;
|
|
+}
|
|
+
|
|
+
|
|
+/*
|
|
+ * Determine if the PPP connection should still be present.
|
|
+ */
|
|
+extern int hungup;
|
|
+
|
|
+/*
|
|
+ * tty_disestablish_ppp - Restore the serial port to normal operation.
|
|
+ * and reconnect the ppp unit to the loopback if in demand mode.
|
|
+ * This shouldn't call die() because it's called from die().
|
|
+ */
|
|
+void
|
|
+tty_disestablish_ppp(fd)
|
|
+ int fd;
|
|
+{
|
|
+ if (demand)
|
|
+ restore_loop();
|
|
+
|
|
+ if (!hungup) {
|
|
+
|
|
+
|
|
+ /* Flush the tty output buffer so that the TIOCSETD doesn't hang. */
|
|
+ if (tcflush(fd, TCIOFLUSH) < 0)
|
|
+ warn("tcflush failed: %m");
|
|
+
|
|
+ /* Restore old line discipline. */
|
|
+ if (initdisc >= 0 && ioctl(fd, TIOCSETD, &initdisc) < 0)
|
|
+ error("ioctl(TIOCSETD): %m");
|
|
+ initdisc = -1;
|
|
+
|
|
+ /* Reset non-blocking mode on fd. */
|
|
+ if (initfdflags != -1 && fcntl(fd, F_SETFL, initfdflags) < 0)
|
|
+ warn("Couldn't restore device fd flags: %m");
|
|
+ }
|
|
+ initfdflags = -1;
|
|
+
|
|
+ if (fd == ppp_fd)
|
|
+ ppp_fd = -1;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * cfg_bundle - configure the existing bundle.
|
|
+ * Used in demand mode.
|
|
+ */
|
|
+void
|
|
+cfg_bundle(mrru, mtru, rssn, tssn)
|
|
+ int mrru, mtru, rssn, tssn;
|
|
+{
|
|
+ abort();
|
|
+#ifdef notyet
|
|
+ int flags;
|
|
+ struct ifreq ifr;
|
|
+
|
|
+ if (!new_style_driver)
|
|
+ return;
|
|
+
|
|
+ /* set the mrru, mtu and flags */
|
|
+ if (ioctl(ppp_dev_fd, PPPIOCSMRRU, &mrru) < 0)
|
|
+ error("Couldn't set MRRU: %m");
|
|
+ flags = get_flags(ppp_dev_fd);
|
|
+ flags &= ~(SC_MP_SHORTSEQ | SC_MP_XSHORTSEQ);
|
|
+ flags |= (rssn? SC_MP_SHORTSEQ: 0) | (tssn? SC_MP_XSHORTSEQ: 0)
|
|
+ | (mrru? SC_MULTILINK: 0);
|
|
+
|
|
+ set_flags(ppp_dev_fd, flags);
|
|
+
|
|
+ /* connect up the channel */
|
|
+ if (ioctl(ppp_fd, PPPIOCCONNECT, &ifunit) < 0)
|
|
+ fatal("Couldn't attach to PPP unit %d: %m", ifunit);
|
|
+ add_fd(ppp_dev_fd);
|
|
+#endif
|
|
+}
|
|
+
|
|
+/*
|
|
+ * make_new_bundle - create a new PPP unit (i.e. a bundle)
|
|
+ * and connect our channel to it. This should only get called
|
|
+ * if `multilink' was set at the time establish_ppp was called.
|
|
+ * In demand mode this uses our existing bundle instead of making
|
|
+ * a new one.
|
|
+ */
|
|
+void
|
|
+make_new_bundle(mrru, mtru, rssn, tssn)
|
|
+ int mrru, mtru, rssn, tssn;
|
|
+{
|
|
+ abort();
|
|
+#ifdef notyet
|
|
+ if (!new_style_driver)
|
|
+ return;
|
|
+
|
|
+ /* make us a ppp unit */
|
|
+ if (make_ppp_unit() < 0)
|
|
+ die(1);
|
|
+
|
|
+ /* set the mrru, mtu and flags */
|
|
+ cfg_bundle(mrru, mtru, rssn, tssn);
|
|
+#endif
|
|
+}
|
|
+
|
|
+/*
|
|
+ * bundle_attach - attach our link to a given PPP unit.
|
|
+ * We assume the unit is controlled by another pppd.
|
|
+ */
|
|
+int
|
|
+bundle_attach(ifnum)
|
|
+ int ifnum;
|
|
+{
|
|
+ abort();
|
|
+#ifdef notyet
|
|
+ if (!new_style_driver)
|
|
+ return -1;
|
|
+
|
|
+ if (ioctl(ppp_dev_fd, PPPIOCATTACH, &ifnum) < 0) {
|
|
+ if (errno == ENXIO)
|
|
+ return 0; /* doesn't still exist */
|
|
+ fatal("Couldn't attach to interface unit %d: %m\n", ifnum);
|
|
+ }
|
|
+ if (ioctl(ppp_fd, PPPIOCCONNECT, &ifnum) < 0)
|
|
+ fatal("Couldn't connect to interface unit %d: %m", ifnum);
|
|
+ set_flags(ppp_dev_fd, get_flags(ppp_dev_fd) | SC_MULTILINK);
|
|
+
|
|
+ ifunit = ifnum;
|
|
+#endif
|
|
+ return 1;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * destroy_bundle - tell the driver to destroy our bundle.
|
|
+ */
|
|
+void destroy_bundle(void)
|
|
+{
|
|
+#if notyet
|
|
+ if (ppp_dev_fd >= 0) {
|
|
+ close(ppp_dev_fd);
|
|
+ remove_fd(ppp_dev_fd);
|
|
+ ppp_dev_fd = -1;
|
|
+ }
|
|
+#endif
|
|
+}
|
|
+
|
|
+
|
|
+/*
|
|
+ * Check whether the link seems not to be 8-bit clean.
|
|
+ */
|
|
+void
|
|
+clean_check()
|
|
+{
|
|
+ int x;
|
|
+ char *s;
|
|
+
|
|
+ if (ioctl(ppp_fd, PPPIOCGFLAGS, (caddr_t) &x) == 0) {
|
|
+ s = NULL;
|
|
+ switch (~x & (SC_RCV_B7_0|SC_RCV_B7_1|SC_RCV_EVNP|SC_RCV_ODDP)) {
|
|
+ case SC_RCV_B7_0:
|
|
+ s = "bit 7 set to 1";
|
|
+ break;
|
|
+ case SC_RCV_B7_1:
|
|
+ s = "bit 7 set to 0";
|
|
+ break;
|
|
+ case SC_RCV_EVNP:
|
|
+ s = "odd parity";
|
|
+ break;
|
|
+ case SC_RCV_ODDP:
|
|
+ s = "even parity";
|
|
+ break;
|
|
+ }
|
|
+ if (s != NULL) {
|
|
+ struct ppp_rawin win;
|
|
+ char buf[4 * sizeof(win.buf) + 1];
|
|
+ int i;
|
|
+ warn("Serial link is not 8-bit clean:");
|
|
+ warn("All received characters had %s", s);
|
|
+ if (ioctl(ppp_fd, PPPIOCGRAWIN, &win) == -1) {
|
|
+ warn("ioctl(PPPIOCGRAWIN): %s", strerror(errno));
|
|
+ return;
|
|
+ }
|
|
+ for (i = 0; i < sizeof(win.buf); i++)
|
|
+ win.buf[i] = win.buf[i] & 0x7f;
|
|
+ strvisx(buf, (char *)win.buf, win.count, VIS_CSTYLE);
|
|
+ warn("Last %d characters were: %s", (int)win.count, buf);
|
|
+ }
|
|
+ }
|
|
+}
|
|
+
|
|
+/*
|
|
+ * set_up_tty: Set up the serial port on `fd' for 8 bits, no parity,
|
|
+ * at the requested speed, etc. If `local' is true, set CLOCAL
|
|
+ * regardless of whether the modem option was specified.
|
|
+ *
|
|
+ * For *BSD, we assume that speed_t values numerically equal bits/second.
|
|
+ */
|
|
+void
|
|
+set_up_tty(fd, local)
|
|
+ int fd, local;
|
|
+{
|
|
+ struct termios tios;
|
|
+
|
|
+ if (tcgetattr(fd, &tios) < 0)
|
|
+ fatal("tcgetattr: %m");
|
|
+
|
|
+ if (!restore_term) {
|
|
+ inittermios = tios;
|
|
+ ioctl(fd, TIOCGWINSZ, &wsinfo);
|
|
+ }
|
|
+
|
|
+ tios.c_cflag &= ~(CSIZE | CSTOPB | PARENB | CLOCAL);
|
|
+ if (crtscts > 0 && !local) {
|
|
+ if (crtscts == 2) {
|
|
+#ifdef CDTRCTS
|
|
+ tios.c_cflag |= CDTRCTS;
|
|
+#endif
|
|
+ } else
|
|
+ tios.c_cflag |= CRTSCTS;
|
|
+ } else if (crtscts < 0) {
|
|
+ tios.c_cflag &= ~CRTSCTS;
|
|
+#ifdef CDTRCTS
|
|
+ tios.c_cflag &= ~CDTRCTS;
|
|
+#endif
|
|
+ }
|
|
+
|
|
+ tios.c_cflag |= CS8 | CREAD | HUPCL;
|
|
+ if (local || !modem)
|
|
+ tios.c_cflag |= CLOCAL;
|
|
+ tios.c_iflag = IGNBRK | IGNPAR;
|
|
+ tios.c_oflag = 0;
|
|
+ tios.c_lflag = 0;
|
|
+ tios.c_cc[VMIN] = 1;
|
|
+ tios.c_cc[VTIME] = 0;
|
|
+
|
|
+ if (crtscts == -2) {
|
|
+ tios.c_iflag |= IXON | IXOFF;
|
|
+ tios.c_cc[VSTOP] = 0x13; /* DC3 = XOFF = ^S */
|
|
+ tios.c_cc[VSTART] = 0x11; /* DC1 = XON = ^Q */
|
|
+ }
|
|
+
|
|
+ if (inspeed) {
|
|
+ cfsetospeed(&tios, inspeed);
|
|
+ cfsetispeed(&tios, inspeed);
|
|
+ } else {
|
|
+ inspeed = cfgetospeed(&tios);
|
|
+ /*
|
|
+ * We can't proceed if the serial port speed is 0,
|
|
+ * since that implies that the serial port is disabled.
|
|
+ */
|
|
+ if (inspeed == 0)
|
|
+ fatal("Baud rate for %s is 0; need explicit baud rate", devnam);
|
|
+ }
|
|
+ baud_rate = inspeed;
|
|
+
|
|
+ if (tcsetattr(fd, TCSAFLUSH, &tios) < 0)
|
|
+ fatal("tcsetattr: %m");
|
|
+
|
|
+ restore_term = 1;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * restore_tty - restore the terminal to the saved settings.
|
|
+ */
|
|
+void
|
|
+restore_tty(fd)
|
|
+ int fd;
|
|
+{
|
|
+ if (restore_term) {
|
|
+ if (!default_device) {
|
|
+ /*
|
|
+ * Turn off echoing, because otherwise we can get into
|
|
+ * a loop with the tty and the modem echoing to each other.
|
|
+ * We presume we are the sole user of this tty device, so
|
|
+ * when we close it, it will revert to its defaults anyway.
|
|
+ */
|
|
+ inittermios.c_lflag &= ~(ECHO | ECHONL);
|
|
+ }
|
|
+ if (tcsetattr(fd, TCSAFLUSH, &inittermios) < 0)
|
|
+ if (errno != ENXIO)
|
|
+ warn("tcsetattr: %m");
|
|
+ ioctl(fd, TIOCSWINSZ, &wsinfo);
|
|
+ restore_term = 0;
|
|
+ }
|
|
+}
|
|
+
|
|
+/*
|
|
+ * setdtr - control the DTR line on the serial port.
|
|
+ * This is called from die(), so it shouldn't call die().
|
|
+ */
|
|
+void
|
|
+setdtr(fd, on)
|
|
+int fd, on;
|
|
+{
|
|
+ int modembits = TIOCM_DTR;
|
|
+
|
|
+ ioctl(fd, (on? TIOCMBIS: TIOCMBIC), &modembits);
|
|
+}
|
|
+
|
|
+#ifdef INET6
|
|
+/*
|
|
+ * sif6addr - Config the interface with an IPv6 link-local address
|
|
+ */
|
|
+int
|
|
+sif6addr(unit, our_eui64, his_eui64)
|
|
+ int unit;
|
|
+ eui64_t our_eui64, his_eui64;
|
|
+{
|
|
+#ifdef __KAME__
|
|
+ int ifindex;
|
|
+ struct in6_aliasreq addreq6;
|
|
+
|
|
+ if (sock6_fd < 0) {
|
|
+ fatal("No IPv6 socket available");
|
|
+ /*NOTREACHED*/
|
|
+ }
|
|
+
|
|
+ /* actually, this part is not kame local - RFC2553 conformant */
|
|
+ ifindex = if_nametoindex(ifname);
|
|
+ if (ifindex == 0) {
|
|
+ error("sifaddr6: no interface %s", ifname);
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ memset(&addreq6, 0, sizeof(addreq6));
|
|
+ strlcpy(addreq6.ifra_name, ifname, sizeof(addreq6.ifra_name));
|
|
+
|
|
+ /* my addr */
|
|
+ addreq6.ifra_addr.sin6_family = AF_INET6;
|
|
+ addreq6.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
|
|
+ addreq6.ifra_addr.sin6_addr.s6_addr[0] = 0xfe;
|
|
+ addreq6.ifra_addr.sin6_addr.s6_addr[1] = 0x80;
|
|
+ memcpy(&addreq6.ifra_addr.sin6_addr.s6_addr[8], &our_eui64,
|
|
+ sizeof(our_eui64));
|
|
+ /* KAME ifindex hack */
|
|
+ *(u_int16_t *)&addreq6.ifra_addr.sin6_addr.s6_addr[2] = htons(ifindex);
|
|
+
|
|
+ /* his addr */
|
|
+ addreq6.ifra_dstaddr.sin6_family = AF_INET6;
|
|
+ addreq6.ifra_dstaddr.sin6_len = sizeof(struct sockaddr_in6);
|
|
+ addreq6.ifra_dstaddr.sin6_addr.s6_addr[0] = 0xfe;
|
|
+ addreq6.ifra_dstaddr.sin6_addr.s6_addr[1] = 0x80;
|
|
+ memcpy(&addreq6.ifra_dstaddr.sin6_addr.s6_addr[8], &his_eui64,
|
|
+ sizeof(our_eui64));
|
|
+ /* KAME ifindex hack */
|
|
+ *(u_int16_t *)&addreq6.ifra_dstaddr.sin6_addr.s6_addr[2] = htons(ifindex);
|
|
+
|
|
+ /* prefix mask: 128bit */
|
|
+ addreq6.ifra_prefixmask.sin6_family = AF_INET6;
|
|
+ addreq6.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
|
|
+ memset(&addreq6.ifra_prefixmask.sin6_addr, 0xff,
|
|
+ sizeof(addreq6.ifra_prefixmask.sin6_addr));
|
|
+
|
|
+ /* address lifetime (infty) */
|
|
+ addreq6.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
|
|
+ addreq6.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
|
|
+
|
|
+ if (ioctl(sock6_fd, SIOCAIFADDR_IN6, &addreq6) < 0) {
|
|
+ error("sif6addr: ioctl(SIOCAIFADDR_IN6): %m");
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ return 1;
|
|
+#else
|
|
+ struct in6_ifreq ifr6;
|
|
+ struct ifreq ifr;
|
|
+ struct in6_rtmsg rt6;
|
|
+
|
|
+ if (sock6_fd < 0) {
|
|
+ fatal("No IPv6 socket available");
|
|
+ /*NOTREACHED*/
|
|
+ }
|
|
+
|
|
+ memset(&ifr, 0, sizeof (ifr));
|
|
+ strlcpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
|
|
+ if (ioctl(sock6_fd, SIOCGIFINDEX, (caddr_t) &ifr) < 0) {
|
|
+ error("sif6addr: ioctl(SIOCGIFINDEX): %m");
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ /* Local interface */
|
|
+ memset(&ifr6, 0, sizeof(ifr6));
|
|
+ IN6_LLADDR_FROM_EUI64(ifr6.ifr6_addr, our_eui64);
|
|
+ ifr6.ifr6_ifindex = ifindex;
|
|
+ ifr6.ifr6_prefixlen = 10;
|
|
+
|
|
+ if (ioctl(sock6_fd, SIOCSIFADDR, &ifr6) < 0) {
|
|
+ error("sif6addr: ioctl(SIOCSIFADDR): %m");
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ /* Route to remote host */
|
|
+ memset(&rt6, 0, sizeof(rt6));
|
|
+ IN6_LLADDR_FROM_EUI64(rt6.rtmsg_dst, his_eui64);
|
|
+ rt6.rtmsg_flags = RTF_UP | RTF_HOST;
|
|
+ rt6.rtmsg_dst_len = 128;
|
|
+ rt6.rtmsg_ifindex = ifr.ifr_ifindex;
|
|
+ rt6.rtmsg_metric = 1;
|
|
+
|
|
+ if (ioctl(sock6_fd, SIOCADDRT, &rt6) < 0) {
|
|
+ error("sif6addr: ioctl(SIOCADDRT): %m");
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ return 1;
|
|
+#endif
|
|
+}
|
|
+
|
|
+
|
|
+/*
|
|
+ * cif6addr - Remove IPv6 address from interface
|
|
+ */
|
|
+int
|
|
+cif6addr(unit, our_eui64, his_eui64)
|
|
+ int unit;
|
|
+ eui64_t our_eui64, his_eui64;
|
|
+{
|
|
+#ifdef __KAME__
|
|
+ int ifindex;
|
|
+ struct in6_ifreq delreq6;
|
|
+
|
|
+ if (sock6_fd < 0) {
|
|
+ fatal("No IPv6 socket available");
|
|
+ /*NOTREACHED*/
|
|
+ }
|
|
+
|
|
+ /* actually, this part is not kame local - RFC2553 conformant */
|
|
+ ifindex = if_nametoindex(ifname);
|
|
+ if (ifindex == 0) {
|
|
+ error("cifaddr6: no interface %s", ifname);
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ memset(&delreq6, 0, sizeof(delreq6));
|
|
+ strlcpy(delreq6.ifr_name, ifname, sizeof(delreq6.ifr_name));
|
|
+
|
|
+ /* my addr */
|
|
+ delreq6.ifr_ifru.ifru_addr.sin6_family = AF_INET6;
|
|
+ delreq6.ifr_ifru.ifru_addr.sin6_len = sizeof(struct sockaddr_in6);
|
|
+ delreq6.ifr_ifru.ifru_addr.sin6_addr.s6_addr[0] = 0xfe;
|
|
+ delreq6.ifr_ifru.ifru_addr.sin6_addr.s6_addr[1] = 0x80;
|
|
+ memcpy(&delreq6.ifr_ifru.ifru_addr.sin6_addr.s6_addr[8], &our_eui64,
|
|
+ sizeof(our_eui64));
|
|
+ /* KAME ifindex hack */
|
|
+ *(u_int16_t *)&delreq6.ifr_ifru.ifru_addr.sin6_addr.s6_addr[2] =
|
|
+ htons(ifindex);
|
|
+
|
|
+ if (ioctl(sock6_fd, SIOCDIFADDR_IN6, &delreq6) < 0) {
|
|
+ error("cif6addr: ioctl(SIOCDIFADDR_IN6): %m");
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ return 1;
|
|
+#else
|
|
+ struct ifreq ifr;
|
|
+ struct in6_ifreq ifr6;
|
|
+
|
|
+ if (sock6_fd < 0) {
|
|
+ fatal("No IPv6 socket available");
|
|
+ /*NOTREACHED*/
|
|
+ }
|
|
+
|
|
+ memset(&ifr, 0, sizeof(ifr));
|
|
+ strlcpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
|
|
+ if (ioctl(sock6_fd, SIOCGIFINDEX, (caddr_t) &ifr) < 0) {
|
|
+ error("cif6addr: ioctl(SIOCGIFINDEX): %m");
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ memset(&ifr6, 0, sizeof(ifr6));
|
|
+ IN6_LLADDR_FROM_EUI64(ifr6.ifr6_addr, our_eui64);
|
|
+ ifr6.ifr6_ifindex = ifr.ifr_ifindex;
|
|
+ ifr6.ifr6_prefixlen = 10;
|
|
+
|
|
+ if (ioctl(sock6_fd, SIOCDIFADDR, &ifr6) < 0) {
|
|
+ if (errno != EADDRNOTAVAIL) {
|
|
+ if (! ok_error (errno))
|
|
+ error("cif6addr: ioctl(SIOCDIFADDR): %m");
|
|
+ }
|
|
+ else {
|
|
+ warn("cif6addr: ioctl(SIOCDIFADDR): No such address");
|
|
+ }
|
|
+ return (0);
|
|
+ }
|
|
+ return 1;
|
|
+#endif
|
|
+}
|
|
+#endif /* INET6 */
|
|
+
|
|
+/*
|
|
+ * get_pty - get a pty master/slave pair and chown the slave side
|
|
+ * to the uid given. Assumes slave_name points to >= 12 bytes of space.
|
|
+ */
|
|
+int
|
|
+get_pty(master_fdp, slave_fdp, slave_name, uid)
|
|
+ int *master_fdp;
|
|
+ int *slave_fdp;
|
|
+ char *slave_name;
|
|
+ int uid;
|
|
+{
|
|
+ struct termios tios;
|
|
+
|
|
+ if (openpty(master_fdp, slave_fdp, slave_name, NULL, NULL) < 0)
|
|
+ return 0;
|
|
+
|
|
+ fchown(*slave_fdp, uid, -1);
|
|
+ fchmod(*slave_fdp, S_IRUSR | S_IWUSR);
|
|
+ if (tcgetattr(*slave_fdp, &tios) == 0) {
|
|
+ tios.c_cflag &= ~(CSIZE | CSTOPB | PARENB);
|
|
+ tios.c_cflag |= CS8 | CREAD;
|
|
+ tios.c_iflag = IGNPAR | CLOCAL;
|
|
+ tios.c_oflag = 0;
|
|
+ tios.c_lflag = 0;
|
|
+ if (tcsetattr(*slave_fdp, TCSAFLUSH, &tios) < 0)
|
|
+ warn("couldn't set attributes on pty: %m");
|
|
+ } else
|
|
+ warn("couldn't get attributes on pty: %m");
|
|
+
|
|
+ return 1;
|
|
+}
|
|
+
|
|
+
|
|
+/*
|
|
+ * open_ppp_loopback - open the device we use for getting
|
|
+ * packets in demand mode, and connect it to a ppp interface.
|
|
+ * Here we use a pty.
|
|
+ */
|
|
+int
|
|
+open_ppp_loopback()
|
|
+{
|
|
+ int flags;
|
|
+ struct termios tios;
|
|
+ int pppdisc = PPPDISC;
|
|
+
|
|
+ if (openpty(&loop_master, &loop_slave, loop_name, NULL, NULL) < 0)
|
|
+ fatal("No free pty for loopback");
|
|
+ SYSDEBUG(("using %s for loopback", loop_name));
|
|
+
|
|
+ if (tcgetattr(loop_slave, &tios) == 0) {
|
|
+ tios.c_cflag &= ~(CSIZE | CSTOPB | PARENB);
|
|
+ tios.c_cflag |= CS8 | CREAD;
|
|
+ tios.c_iflag = IGNPAR;
|
|
+ tios.c_oflag = 0;
|
|
+ tios.c_lflag = 0;
|
|
+ if (tcsetattr(loop_slave, TCSAFLUSH, &tios) < 0)
|
|
+ warn("couldn't set attributes on loopback: %m");
|
|
+ }
|
|
+
|
|
+ if ((flags = fcntl(loop_master, F_GETFL)) != -1)
|
|
+ if (fcntl(loop_master, F_SETFL, flags | O_NONBLOCK) == -1)
|
|
+ warn("couldn't set loopback to nonblock: %m");
|
|
+
|
|
+ ppp_fd = loop_slave;
|
|
+ if (ioctl(ppp_fd, TIOCSETD, &pppdisc) < 0)
|
|
+ fatal("ioctl(TIOCSETD): %m");
|
|
+
|
|
+ /*
|
|
+ * Find out which interface we were given.
|
|
+ */
|
|
+ if (ioctl(ppp_fd, PPPIOCGUNIT, &ifunit) < 0)
|
|
+ fatal("ioctl(PPPIOCGUNIT): %m");
|
|
+
|
|
+ /*
|
|
+ * Enable debug in the driver if requested.
|
|
+ */
|
|
+ if (kdebugflag) {
|
|
+ flags = get_flags(ppp_fd);
|
|
+ flags |= (kdebugflag & 0xFF) * SC_DEBUG;
|
|
+ set_flags(ppp_fd, flags);
|
|
+ }
|
|
+
|
|
+ return loop_master;
|
|
+}
|
|
+
|
|
+
|
|
+/*
|
|
+ * output - Output PPP packet.
|
|
+ */
|
|
+void
|
|
+output(unit, p, len)
|
|
+ int unit;
|
|
+ u_char *p;
|
|
+ int len;
|
|
+{
|
|
+ if (debug)
|
|
+ dbglog("sent %P", p, len);
|
|
+
|
|
+ if (write(ttyfd, p, len) < 0) {
|
|
+ if (errno != EIO)
|
|
+ error("write: %m");
|
|
+ }
|
|
+}
|
|
+
|
|
+
|
|
+/*
|
|
+ * wait_input - wait until there is data available,
|
|
+ * for the length of time specified by *timo (indefinite
|
|
+ * if timo is NULL).
|
|
+ */
|
|
+void
|
|
+wait_input(timo)
|
|
+ struct timeval *timo;
|
|
+{
|
|
+ fd_set ready;
|
|
+ int n;
|
|
+
|
|
+ ready = in_fds;
|
|
+ n = select(max_in_fd + 1, &ready, NULL, &ready, timo);
|
|
+ if (n < 0 && errno != EINTR)
|
|
+ fatal("select: %m");
|
|
+}
|
|
+
|
|
+
|
|
+/*
|
|
+ * add_fd - add an fd to the set that wait_input waits for.
|
|
+ */
|
|
+void add_fd(fd)
|
|
+ int fd;
|
|
+{
|
|
+ if (fd >= FD_SETSIZE)
|
|
+ fatal("descriptor too big");
|
|
+ FD_SET(fd, &in_fds);
|
|
+ if (fd > max_in_fd)
|
|
+ max_in_fd = fd;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * remove_fd - remove an fd from the set that wait_input waits for.
|
|
+ */
|
|
+void remove_fd(fd)
|
|
+ int fd;
|
|
+{
|
|
+ FD_CLR(fd, &in_fds);
|
|
+}
|
|
+
|
|
+#if 0
|
|
+/*
|
|
+ * wait_loop_output - wait until there is data available on the
|
|
+ * loopback, for the length of time specified by *timo (indefinite
|
|
+ * if timo is NULL).
|
|
+ */
|
|
+void
|
|
+wait_loop_output(timo)
|
|
+ struct timeval *timo;
|
|
+{
|
|
+ fd_set ready;
|
|
+ int n;
|
|
+
|
|
+ FD_ZERO(&ready);
|
|
+ if (loop_master >= FD_SETSIZE)
|
|
+ fatal("descriptor too big");
|
|
+ FD_SET(loop_master, &ready);
|
|
+ n = select(loop_master + 1, &ready, NULL, &ready, timo);
|
|
+ if (n < 0 && errno != EINTR)
|
|
+ fatal("select: %m");
|
|
+}
|
|
+
|
|
+
|
|
+/*
|
|
+ * wait_time - wait for a given length of time or until a
|
|
+ * signal is received.
|
|
+ */
|
|
+void
|
|
+wait_time(timo)
|
|
+ struct timeval *timo;
|
|
+{
|
|
+ int n;
|
|
+
|
|
+ n = select(0, NULL, NULL, NULL, timo);
|
|
+ if (n < 0 && errno != EINTR)
|
|
+ fatal("select: %m");
|
|
+}
|
|
+#endif
|
|
+
|
|
+
|
|
+/*
|
|
+ * read_packet - get a PPP packet from the serial device.
|
|
+ */
|
|
+int
|
|
+read_packet(buf)
|
|
+ u_char *buf;
|
|
+{
|
|
+ int len;
|
|
+
|
|
+ if ((len = read(ttyfd, buf, PPP_MTU + PPP_HDRLEN)) < 0) {
|
|
+ if (errno == EWOULDBLOCK || errno == EINTR)
|
|
+ return -1;
|
|
+ fatal("read: %m");
|
|
+ }
|
|
+ return len;
|
|
+}
|
|
+
|
|
+
|
|
+/*
|
|
+ * get_loop_output - read characters from the loopback, form them
|
|
+ * into frames, and detect when we want to bring the real link up.
|
|
+ * Return value is 1 if we need to bring up the link, 0 otherwise.
|
|
+ */
|
|
+int
|
|
+get_loop_output()
|
|
+{
|
|
+ int rv = 0;
|
|
+ int n;
|
|
+
|
|
+ while ((n = read(loop_master, inbuf, sizeof(inbuf))) >= 0) {
|
|
+ if (loop_chars(inbuf, n))
|
|
+ rv = 1;
|
|
+ }
|
|
+
|
|
+ if (n == 0)
|
|
+ fatal("eof on loopback");
|
|
+ if (errno != EWOULDBLOCK)
|
|
+ fatal("read from loopback: %m");
|
|
+
|
|
+ return rv;
|
|
+}
|
|
+
|
|
+
|
|
+/*
|
|
+ * netif_set_mtu - set the MTU on the PPP network interface.
|
|
+ */
|
|
+void
|
|
+netif_set_mtu(unit, mtu)
|
|
+ int unit, mtu;
|
|
+{
|
|
+ struct ifreq ifr;
|
|
+
|
|
+ SYSDEBUG((LOG_DEBUG, "netif_set_mtu: mtu = %d\n", mtu));
|
|
+
|
|
+ memset(&ifr, '\0', sizeof (ifr));
|
|
+ strlcpy(ifr.ifr_name, ifname, sizeof (ifr.ifr_name));
|
|
+ ifr.ifr_mtu = mtu;
|
|
+
|
|
+ if (ifunit >= 0 && ioctl(sock_fd, SIOCSIFMTU, (caddr_t) &ifr) < 0)
|
|
+ fatal("ioctl(SIOCSIFMTU): %m");
|
|
+}
|
|
+
|
|
+/*
|
|
+ * netif_get_mtu - get the MTU on the PPP network interface.
|
|
+ */
|
|
+int
|
|
+netif_get_mtu(int unit)
|
|
+{
|
|
+ struct ifreq ifr;
|
|
+
|
|
+ memset (&ifr, '\0', sizeof (ifr));
|
|
+ strlcpy(ifr.ifr_name, ifname, sizeof (ifr.ifr_name));
|
|
+
|
|
+ if (ifunit >= 0 && ioctl(sock_fd, SIOCGIFMTU, (caddr_t) &ifr) < 0) {
|
|
+ error("ioctl(SIOCGIFMTU): %m (line %d)", __LINE__);
|
|
+ return 0;
|
|
+ }
|
|
+ return ifr.ifr_mtu;
|
|
+}
|
|
+
|
|
+
|
|
+/*
|
|
+ * tty_send_config - configure the transmit characteristics of
|
|
+ * the ppp interface.
|
|
+ */
|
|
+void
|
|
+tty_send_config(mtu, asyncmap, pcomp, accomp)
|
|
+ int mtu;
|
|
+ u_int32_t asyncmap;
|
|
+ int pcomp, accomp;
|
|
+{
|
|
+ u_int x;
|
|
+#if 0
|
|
+ /* Linux code does not do anything with the mtu here */
|
|
+ ifnet_set_mtu(-1, mtu);
|
|
+#endif
|
|
+
|
|
+ if (ioctl(ppp_fd, PPPIOCSASYNCMAP, (caddr_t) &asyncmap) < 0)
|
|
+ fatal("ioctl(PPPIOCSASYNCMAP): %m");
|
|
+
|
|
+ x = get_flags(ppp_fd);
|
|
+ x = pcomp? x | SC_COMP_PROT: x &~ SC_COMP_PROT;
|
|
+ x = accomp? x | SC_COMP_AC: x &~ SC_COMP_AC;
|
|
+ x = sync_serial ? x | SC_SYNC : x & ~SC_SYNC;
|
|
+ set_flags(ppp_fd, x);
|
|
+}
|
|
+
|
|
+
|
|
+/*
|
|
+ * ppp_set_xaccm - set the extended transmit ACCM for the interface.
|
|
+ */
|
|
+void
|
|
+tty_set_xaccm(accm)
|
|
+ ext_accm accm;
|
|
+{
|
|
+ if (ioctl(ppp_fd, PPPIOCSXASYNCMAP, accm) < 0 && errno != ENOTTY)
|
|
+ warn("ioctl(set extended ACCM): %m");
|
|
+}
|
|
+
|
|
+
|
|
+/*
|
|
+ * ppp_recv_config - configure the receive-side characteristics of
|
|
+ * the ppp interface.
|
|
+ */
|
|
+void
|
|
+tty_recv_config(mru, asyncmap, pcomp, accomp)
|
|
+ int mru;
|
|
+ u_int32_t asyncmap;
|
|
+ int pcomp, accomp;
|
|
+{
|
|
+ int x;
|
|
+
|
|
+ if (ioctl(ppp_fd, PPPIOCSMRU, (caddr_t) &mru) < 0)
|
|
+ fatal("ioctl(PPPIOCSMRU): %m");
|
|
+ if (ioctl(ppp_fd, PPPIOCSRASYNCMAP, (caddr_t) &asyncmap) < 0)
|
|
+ fatal("ioctl(PPPIOCSRASYNCMAP): %m");
|
|
+ x = get_flags(ppp_fd);
|
|
+ x = !accomp? x | SC_REJ_COMP_AC: x &~ SC_REJ_COMP_AC;
|
|
+ set_flags(ppp_fd, x);
|
|
+}
|
|
+
|
|
+/*
|
|
+ * ccp_test - ask kernel whether a given compression method
|
|
+ * is acceptable for use. Returns 1 if the method and parameters
|
|
+ * are OK, 0 if the method is known but the parameters are not OK
|
|
+ * (e.g. code size should be reduced), or -1 if the method is unknown.
|
|
+ */
|
|
+int
|
|
+ccp_test(unit, opt_ptr, opt_len, for_transmit)
|
|
+ int unit, opt_len, for_transmit;
|
|
+ u_char *opt_ptr;
|
|
+{
|
|
+ struct ppp_option_data data;
|
|
+
|
|
+ data.ptr = opt_ptr;
|
|
+ data.length = opt_len;
|
|
+ data.transmit = for_transmit;
|
|
+ if (ioctl(ttyfd, PPPIOCSCOMPRESS, (caddr_t) &data) >= 0)
|
|
+ return 1;
|
|
+ return (errno == ENOBUFS)? 0: -1;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * ccp_flags_set - inform kernel about the current state of CCP.
|
|
+ */
|
|
+void
|
|
+ccp_flags_set(unit, isopen, isup)
|
|
+ int unit, isopen, isup;
|
|
+{
|
|
+ int x;
|
|
+
|
|
+ x = get_flags(ppp_fd);
|
|
+ x = isopen? x | SC_CCP_OPEN: x &~ SC_CCP_OPEN;
|
|
+ x = isup? x | SC_CCP_UP: x &~ SC_CCP_UP;
|
|
+ set_flags(ppp_fd, x);
|
|
+}
|
|
+
|
|
+/*
|
|
+ * 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(unit)
|
|
+ int unit;
|
|
+{
|
|
+ int x;
|
|
+
|
|
+ x = get_flags(ppp_fd);
|
|
+ return x & SC_DC_FERROR;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * get_idle_time - return how long the link has been idle.
|
|
+ */
|
|
+int
|
|
+get_idle_time(u, ip)
|
|
+ int u;
|
|
+ struct ppp_idle *ip;
|
|
+{
|
|
+ return ioctl(ppp_fd, PPPIOCGIDLE, ip) >= 0;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * get_ppp_stats - return statistics for the link.
|
|
+ */
|
|
+int
|
|
+get_ppp_stats(u, stats)
|
|
+ int u;
|
|
+ struct pppd_stats *stats;
|
|
+{
|
|
+ struct ifpppstatsreq req;
|
|
+
|
|
+ memset (&req, 0, sizeof (req));
|
|
+ strlcpy(req.ifr_name, ifname, sizeof(req.ifr_name));
|
|
+ if (ioctl(sock_fd, SIOCGPPPSTATS, &req) < 0) {
|
|
+ error("Couldn't get PPP statistics: %m");
|
|
+ return 0;
|
|
+ }
|
|
+ stats->bytes_in = req.stats.p.ppp_ibytes;
|
|
+ stats->bytes_out = req.stats.p.ppp_obytes;
|
|
+ return 1;
|
|
+}
|
|
+
|
|
+
|
|
+#ifdef PPP_FILTER
|
|
+/*
|
|
+ * set_filters - transfer the pass and active filters to the kernel.
|
|
+ */
|
|
+int
|
|
+set_filters(pass_in, pass_out, active_in, active_out)
|
|
+ struct bpf_program *pass_in, *pass_out, *active_in, *active_out;
|
|
+{
|
|
+ int ret = 1;
|
|
+
|
|
+ if (pass_in->bf_len > 0) {
|
|
+ if (ioctl(ppp_fd, PPPIOCSIPASS, pass_in) < 0) {
|
|
+ error("Couldn't set pass-filter-in in kernel: %m");
|
|
+ ret = 0;
|
|
+ }
|
|
+ }
|
|
+
|
|
+ if (pass_out->bf_len > 0) {
|
|
+ if (ioctl(ppp_fd, PPPIOCSOPASS, pass_out) < 0) {
|
|
+ error("Couldn't set pass-filter-out in kernel: %m");
|
|
+ ret = 0;
|
|
+ }
|
|
+ }
|
|
+
|
|
+ if (active_in->bf_len > 0) {
|
|
+ if (ioctl(ppp_fd, PPPIOCSIACTIVE, active_in) < 0) {
|
|
+ error("Couldn't set active-filter-in in kernel: %m");
|
|
+ ret = 0;
|
|
+ }
|
|
+ }
|
|
+
|
|
+ if (active_out->bf_len > 0) {
|
|
+ if (ioctl(ppp_fd, PPPIOCSOACTIVE, active_out) < 0) {
|
|
+ error("Couldn't set active-filter-out in kernel: %m");
|
|
+ ret = 0;
|
|
+ }
|
|
+ }
|
|
+
|
|
+ return ret;
|
|
+}
|
|
+#endif
|
|
+
|
|
+/*
|
|
+ * sifvjcomp - config tcp header compression
|
|
+ */
|
|
+int
|
|
+sifvjcomp(u, vjcomp, cidcomp, maxcid)
|
|
+ int u, vjcomp, cidcomp, maxcid;
|
|
+{
|
|
+ u_int x;
|
|
+
|
|
+ x = get_flags(ppp_fd);
|
|
+ x = vjcomp ? x | SC_COMP_TCP: x &~ SC_COMP_TCP;
|
|
+ x = cidcomp? x & ~SC_NO_TCP_CCID: x | SC_NO_TCP_CCID;
|
|
+ set_flags(ppp_fd, x);
|
|
+ if (vjcomp && ioctl(ppp_fd, PPPIOCSMAXCID, (caddr_t) &maxcid) < 0) {
|
|
+ error("ioctl(PPPIOCSMAXCID): %m");
|
|
+ return 0;
|
|
+ }
|
|
+ return 1;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * sifup - Config the interface up and enable IP packets to pass.
|
|
+ */
|
|
+int
|
|
+sifup(u)
|
|
+ int u;
|
|
+{
|
|
+ struct ifreq ifr;
|
|
+
|
|
+ strlcpy(ifr.ifr_name, ifname, sizeof (ifr.ifr_name));
|
|
+ if (ioctl(sock_fd, SIOCGIFFLAGS, (caddr_t) &ifr) < 0) {
|
|
+ error("ioctl (SIOCGIFFLAGS): %m");
|
|
+ return 0;
|
|
+ }
|
|
+ ifr.ifr_flags |= IFF_UP;
|
|
+ if (ioctl(sock_fd, SIOCSIFFLAGS, (caddr_t) &ifr) < 0) {
|
|
+ error("ioctl(SIOCSIFFLAGS): %m");
|
|
+ return 0;
|
|
+ }
|
|
+ if_is_up = 1;
|
|
+ return 1;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * sifnpmode - Set the mode for handling packets for a given NP.
|
|
+ */
|
|
+int
|
|
+sifnpmode(u, proto, mode)
|
|
+ int u;
|
|
+ int proto;
|
|
+ enum NPmode mode;
|
|
+{
|
|
+ struct npioctl npi;
|
|
+
|
|
+ npi.protocol = proto;
|
|
+ npi.mode = mode;
|
|
+ if (ioctl(ppp_fd, PPPIOCSNPMODE, &npi) < 0) {
|
|
+ error("ioctl(set NP %d mode to %d): %m", proto, mode);
|
|
+ return 0;
|
|
+ }
|
|
+ return 1;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * sifdown - Config the interface down and disable IP.
|
|
+ */
|
|
+int
|
|
+sifdown(u)
|
|
+ int u;
|
|
+{
|
|
+ struct ifreq ifr;
|
|
+ int rv;
|
|
+ struct npioctl npi;
|
|
+
|
|
+ rv = 1;
|
|
+ npi.protocol = PPP_IP;
|
|
+ npi.mode = NPMODE_ERROR;
|
|
+ ioctl(ppp_fd, PPPIOCSNPMODE, (caddr_t) &npi);
|
|
+ /* ignore errors, because ppp_fd might have been closed by now. */
|
|
+
|
|
+ strlcpy(ifr.ifr_name, ifname, sizeof (ifr.ifr_name));
|
|
+ if (ioctl(sock_fd, SIOCGIFFLAGS, (caddr_t) &ifr) < 0) {
|
|
+ error("ioctl (SIOCGIFFLAGS): %m");
|
|
+ rv = 0;
|
|
+ } else {
|
|
+ ifr.ifr_flags &= ~IFF_UP;
|
|
+ if (ioctl(sock_fd, SIOCSIFFLAGS, (caddr_t) &ifr) < 0) {
|
|
+ error("ioctl(SIOCSIFFLAGS): %m");
|
|
+ rv = 0;
|
|
+ } else
|
|
+ if_is_up = 0;
|
|
+ }
|
|
+ return rv;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * SET_SA_FAMILY - set the sa_family field of a struct sockaddr,
|
|
+ * if it exists.
|
|
+ */
|
|
+#define SET_SA_FAMILY(addr, family) \
|
|
+ BZERO((char *) &(addr), sizeof(addr)); \
|
|
+ addr.sa_family = (family); \
|
|
+ addr.sa_len = sizeof(addr);
|
|
+
|
|
+/*
|
|
+ * sifaddr - Config the interface IP addresses and netmask.
|
|
+ */
|
|
+int
|
|
+sifaddr(u, o, h, m)
|
|
+ int u;
|
|
+ u_int32_t o, h, m;
|
|
+{
|
|
+ struct ifaliasreq ifra;
|
|
+ struct ifreq ifr;
|
|
+
|
|
+ strlcpy(ifra.ifra_name, ifname, sizeof(ifra.ifra_name));
|
|
+ SET_SA_FAMILY(ifra.ifra_addr, AF_INET);
|
|
+ ((struct sockaddr_in *) &ifra.ifra_addr)->sin_addr.s_addr = o;
|
|
+ SET_SA_FAMILY(ifra.ifra_broadaddr, AF_INET);
|
|
+ ((struct sockaddr_in *) &ifra.ifra_broadaddr)->sin_addr.s_addr = h;
|
|
+ if (m != 0) {
|
|
+ SET_SA_FAMILY(ifra.ifra_mask, AF_INET);
|
|
+ ((struct sockaddr_in *) &ifra.ifra_mask)->sin_addr.s_addr = m;
|
|
+ } else
|
|
+ BZERO(&ifra.ifra_mask, sizeof(ifra.ifra_mask));
|
|
+ BZERO(&ifr, sizeof(ifr));
|
|
+ strlcpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
|
|
+ if (ioctl(sock_fd, SIOCDIFADDR, (caddr_t) &ifr) < 0) {
|
|
+ if (errno != EADDRNOTAVAIL)
|
|
+ warn("Couldn't remove interface address: %m");
|
|
+ }
|
|
+ if (ioctl(sock_fd, SIOCAIFADDR, (caddr_t) &ifra) < 0) {
|
|
+ if (errno != EEXIST) {
|
|
+ error("Couldn't set interface address: %m");
|
|
+ return 0;
|
|
+ }
|
|
+ warn("Couldn't set interface address: Address %I already exists", o);
|
|
+ }
|
|
+ ifaddrs[0] = o;
|
|
+ ifaddrs[1] = h;
|
|
+ return 1;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * cifaddr - Clear the interface IP addresses, and delete routes
|
|
+ * through the interface if possible.
|
|
+ */
|
|
+int
|
|
+cifaddr(u, o, h)
|
|
+ int u;
|
|
+ u_int32_t o, h;
|
|
+{
|
|
+ struct ifaliasreq ifra;
|
|
+
|
|
+ ifaddrs[0] = 0;
|
|
+ strlcpy(ifra.ifra_name, ifname, sizeof(ifra.ifra_name));
|
|
+ SET_SA_FAMILY(ifra.ifra_addr, AF_INET);
|
|
+ ((struct sockaddr_in *) &ifra.ifra_addr)->sin_addr.s_addr = o;
|
|
+ SET_SA_FAMILY(ifra.ifra_broadaddr, AF_INET);
|
|
+ ((struct sockaddr_in *) &ifra.ifra_broadaddr)->sin_addr.s_addr = h;
|
|
+ BZERO(&ifra.ifra_mask, sizeof(ifra.ifra_mask));
|
|
+ if (ioctl(sock_fd, SIOCDIFADDR, (caddr_t) &ifra) < 0) {
|
|
+ if (errno != EADDRNOTAVAIL)
|
|
+ warn("Couldn't delete interface address: %m");
|
|
+ return 0;
|
|
+ }
|
|
+ return 1;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * sifdefaultroute - assign a default route through the address given.
|
|
+ */
|
|
+int
|
|
+sifdefaultroute(u, l, g)
|
|
+ int u;
|
|
+ u_int32_t l, g;
|
|
+{
|
|
+ return dodefaultroute(g, 's');
|
|
+}
|
|
+
|
|
+/*
|
|
+ * cifdefaultroute - delete a default route through the address given.
|
|
+ */
|
|
+int
|
|
+cifdefaultroute(u, l, g)
|
|
+ int u;
|
|
+ u_int32_t l, g;
|
|
+{
|
|
+ return dodefaultroute(g, 'c');
|
|
+}
|
|
+
|
|
+/********************************************************************
|
|
+ *
|
|
+ * sipxfaddr - Config the interface IPX networknumber
|
|
+ */
|
|
+
|
|
+int sipxfaddr (int unit, unsigned long int network, unsigned char * node )
|
|
+{
|
|
+ int result = 1;
|
|
+
|
|
+#ifdef IPX_CHANGE
|
|
+ int skfd;
|
|
+ struct ifreq ifr;
|
|
+ struct sockaddr_ipx *sipx = (struct sockaddr_ipx *) &ifr.ifr_addr;
|
|
+
|
|
+ skfd = socket (AF_IPX, SOCK_DGRAM, 0);
|
|
+ if (skfd < 0) {
|
|
+ if (! ok_error (errno))
|
|
+ dbglog("socket(AF_IPX): %m (line %d)", __LINE__);
|
|
+ result = 0;
|
|
+ }
|
|
+ else {
|
|
+ memset (&ifr, '\0', sizeof (ifr));
|
|
+ strlcpy (ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
|
|
+
|
|
+ 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) {
|
|
+ if (! ok_error (errno))
|
|
+ dbglog("ioctl(SIOCSIFADDR, CRTITF): %m (line %d)", __LINE__);
|
|
+ }
|
|
+ else {
|
|
+ warn("ioctl(SIOCSIFADDR, CRTITF): Address already exists");
|
|
+ }
|
|
+ }
|
|
+ close (skfd);
|
|
+ }
|
|
+#endif
|
|
+ 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 unit)
|
|
+{
|
|
+ int result = 1;
|
|
+
|
|
+#ifdef IPX_CHANGE
|
|
+ int skfd;
|
|
+ struct ifreq ifr;
|
|
+ struct sockaddr_ipx *sipx = (struct sockaddr_ipx *) &ifr.ifr_addr;
|
|
+
|
|
+ skfd = socket (AF_IPX, SOCK_DGRAM, 0);
|
|
+ if (skfd < 0) {
|
|
+ if (! ok_error (errno))
|
|
+ dbglog("socket(AF_IPX): %m (line %d)", __LINE__);
|
|
+ result = 0;
|
|
+ }
|
|
+ else {
|
|
+ memset (&ifr, '\0', sizeof (ifr));
|
|
+ strlcpy (ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
|
|
+
|
|
+ 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) {
|
|
+ if (! ok_error (errno))
|
|
+ info("ioctl(SIOCSIFADDR, IPX_DLTITF): %m (line %d)", __LINE__);
|
|
+ result = 0;
|
|
+ }
|
|
+ close (skfd);
|
|
+ }
|
|
+#endif
|
|
+ return result;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * dodefaultroute - talk to a routing socket to add/delete a default route.
|
|
+ */
|
|
+static int
|
|
+dodefaultroute(g, cmd)
|
|
+ u_int32_t g;
|
|
+ int cmd;
|
|
+{
|
|
+ int routes;
|
|
+ struct {
|
|
+ struct rt_msghdr hdr;
|
|
+ struct sockaddr_in dst;
|
|
+ struct sockaddr_in gway;
|
|
+ struct sockaddr_in netmask;
|
|
+ struct sockaddr_in genmask;
|
|
+ struct sockaddr_dl ifp;
|
|
+ } rtmsg;
|
|
+
|
|
+ if ((routes = socket(PF_ROUTE, SOCK_RAW, AF_INET)) < 0) {
|
|
+ error("Couldn't %s default route: socket: %m",
|
|
+ cmd == 's' ? "add" : "delete");
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ memset(&rtmsg, 0, sizeof(rtmsg));
|
|
+
|
|
+ rtmsg.hdr.rtm_type = cmd == 's' ? RTM_ADD : RTM_DELETE;
|
|
+ rtmsg.hdr.rtm_flags = RTF_UP | RTF_GATEWAY | RTF_STATIC;
|
|
+ rtmsg.hdr.rtm_version = RTM_VERSION;
|
|
+ rtmsg.hdr.rtm_seq = ++rtm_seq;
|
|
+ rtmsg.hdr.rtm_addrs =
|
|
+ RTA_DST | RTA_GATEWAY | RTA_NETMASK | RTA_GENMASK | RTA_IFP;
|
|
+
|
|
+ rtmsg.dst.sin_len = sizeof(rtmsg.dst);
|
|
+ rtmsg.dst.sin_family = AF_INET;
|
|
+ rtmsg.dst.sin_addr.s_addr = 0;
|
|
+
|
|
+ rtmsg.gway.sin_len = sizeof(rtmsg.gway);
|
|
+ rtmsg.gway.sin_family = AF_INET;
|
|
+ rtmsg.gway.sin_addr.s_addr = g;
|
|
+
|
|
+ rtmsg.netmask.sin_len = sizeof(rtmsg.netmask);
|
|
+ rtmsg.netmask.sin_family = AF_INET;
|
|
+ rtmsg.netmask.sin_addr.s_addr = 0;
|
|
+
|
|
+ rtmsg.genmask.sin_len = sizeof(rtmsg.genmask);
|
|
+ rtmsg.genmask.sin_family = AF_INET;
|
|
+ rtmsg.genmask.sin_addr.s_addr = 0;
|
|
+
|
|
+ rtmsg.ifp.sdl_family = AF_LINK;
|
|
+ rtmsg.ifp.sdl_len = sizeof(rtmsg.ifp);
|
|
+ link_addr(ifname, &rtmsg.ifp);
|
|
+
|
|
+ rtmsg.hdr.rtm_msglen = sizeof(rtmsg);
|
|
+
|
|
+ if (write(routes, &rtmsg, sizeof(rtmsg)) < 0) {
|
|
+ error("Couldn't %s default route: %m",
|
|
+ cmd == 's' ? "add" : "delete");
|
|
+ close(routes);
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ close(routes);
|
|
+ default_route_gateway = (cmd == 's') ? g : 0;
|
|
+ return 1;
|
|
+}
|
|
+
|
|
+#if RTM_VERSION >= 3
|
|
+
|
|
+/*
|
|
+ * sifproxyarp - Make a proxy ARP entry for the peer.
|
|
+ */
|
|
+static struct {
|
|
+ struct rt_msghdr hdr;
|
|
+ struct sockaddr_inarp dst;
|
|
+ struct sockaddr_dl hwa;
|
|
+ char extra[128];
|
|
+} arpmsg;
|
|
+
|
|
+static int arpmsg_valid;
|
|
+
|
|
+int
|
|
+sifproxyarp(unit, hisaddr)
|
|
+ int unit;
|
|
+ u_int32_t hisaddr;
|
|
+{
|
|
+ int routes;
|
|
+
|
|
+ /*
|
|
+ * Get the hardware address of an interface on the same subnet
|
|
+ * as our local address.
|
|
+ */
|
|
+ memset(&arpmsg, 0, sizeof(arpmsg));
|
|
+ if (!get_ether_addr(hisaddr, &arpmsg.hwa)) {
|
|
+ error("Cannot determine ethernet address for proxy ARP");
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ if ((routes = socket(PF_ROUTE, SOCK_RAW, AF_INET)) < 0) {
|
|
+ error("Couldn't add proxy arp entry: socket: %m");
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ arpmsg.hdr.rtm_type = RTM_ADD;
|
|
+ arpmsg.hdr.rtm_flags = RTF_ANNOUNCE | RTF_HOST | RTF_STATIC;
|
|
+ arpmsg.hdr.rtm_version = RTM_VERSION;
|
|
+ arpmsg.hdr.rtm_seq = ++rtm_seq;
|
|
+ arpmsg.hdr.rtm_addrs = RTA_DST | RTA_GATEWAY;
|
|
+ arpmsg.hdr.rtm_inits = RTV_EXPIRE;
|
|
+ arpmsg.dst.sin_len = sizeof(struct sockaddr_inarp);
|
|
+ arpmsg.dst.sin_family = AF_INET;
|
|
+ arpmsg.dst.sin_addr.s_addr = hisaddr;
|
|
+ arpmsg.dst.sin_other = SIN_PROXY;
|
|
+
|
|
+ arpmsg.hdr.rtm_msglen = (char *) &arpmsg.hwa - (char *) &arpmsg
|
|
+ + arpmsg.hwa.sdl_len;
|
|
+ if (write(routes, &arpmsg, arpmsg.hdr.rtm_msglen) < 0) {
|
|
+ error("Couldn't add proxy arp entry: %m");
|
|
+ close(routes);
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ close(routes);
|
|
+ arpmsg_valid = 1;
|
|
+ proxy_arp_addr = hisaddr;
|
|
+ return 1;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * cifproxyarp - Delete the proxy ARP entry for the peer.
|
|
+ */
|
|
+int
|
|
+cifproxyarp(unit, hisaddr)
|
|
+ int unit;
|
|
+ u_int32_t hisaddr;
|
|
+{
|
|
+ int routes;
|
|
+
|
|
+ if (!arpmsg_valid)
|
|
+ return 0;
|
|
+ arpmsg_valid = 0;
|
|
+
|
|
+ arpmsg.hdr.rtm_type = RTM_DELETE;
|
|
+ arpmsg.hdr.rtm_seq = ++rtm_seq;
|
|
+
|
|
+ if ((routes = socket(PF_ROUTE, SOCK_RAW, AF_INET)) < 0) {
|
|
+ error("Couldn't delete proxy arp entry: socket: %m");
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ if (write(routes, &arpmsg, arpmsg.hdr.rtm_msglen) < 0) {
|
|
+ error("Couldn't delete proxy arp entry: %m");
|
|
+ close(routes);
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ close(routes);
|
|
+ proxy_arp_addr = 0;
|
|
+ return 1;
|
|
+}
|
|
+
|
|
+#else /* RTM_VERSION */
|
|
+
|
|
+/*
|
|
+ * sifproxyarp - Make a proxy ARP entry for the peer.
|
|
+ */
|
|
+int
|
|
+sifproxyarp(unit, hisaddr)
|
|
+ int unit;
|
|
+ u_int32_t hisaddr;
|
|
+{
|
|
+ struct arpreq arpreq;
|
|
+ struct {
|
|
+ struct sockaddr_dl sdl;
|
|
+ char space[128];
|
|
+ } dls;
|
|
+
|
|
+ BZERO(&arpreq, sizeof(arpreq));
|
|
+
|
|
+ /*
|
|
+ * Get the hardware address of an interface on the same subnet
|
|
+ * as our local address.
|
|
+ */
|
|
+ if (!get_ether_addr(hisaddr, &dls.sdl)) {
|
|
+ error("Cannot determine ethernet address for proxy ARP");
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ arpreq.arp_ha.sa_len = sizeof(struct sockaddr);
|
|
+ arpreq.arp_ha.sa_family = AF_UNSPEC;
|
|
+ BCOPY(LLADDR(&dls.sdl), arpreq.arp_ha.sa_data, dls.sdl.sdl_alen);
|
|
+ SET_SA_FAMILY(arpreq.arp_pa, AF_INET);
|
|
+ ((struct sockaddr_in *) &arpreq.arp_pa)->sin_addr.s_addr = hisaddr;
|
|
+ arpreq.arp_flags = ATF_PERM | ATF_PUBL;
|
|
+ if (ioctl(sock_fd, SIOCSARP, (caddr_t)&arpreq) < 0) {
|
|
+ error("Couldn't add proxy arp entry: %m");
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ proxy_arp_addr = hisaddr;
|
|
+ return 1;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * cifproxyarp - Delete the proxy ARP entry for the peer.
|
|
+ */
|
|
+int
|
|
+cifproxyarp(unit, hisaddr)
|
|
+ int unit;
|
|
+ u_int32_t hisaddr;
|
|
+{
|
|
+ struct arpreq arpreq;
|
|
+
|
|
+ BZERO(&arpreq, sizeof(arpreq));
|
|
+ SET_SA_FAMILY(arpreq.arp_pa, AF_INET);
|
|
+ ((struct sockaddr_in *) &arpreq.arp_pa)->sin_addr.s_addr = hisaddr;
|
|
+ if (ioctl(sock_fd, SIOCDARP, (caddr_t)&arpreq) < 0) {
|
|
+ warn("Couldn't delete proxy arp entry: %m");
|
|
+ return 0;
|
|
+ }
|
|
+ proxy_arp_addr = 0;
|
|
+ return 1;
|
|
+}
|
|
+#endif /* RTM_VERSION */
|
|
+
|
|
+
|
|
+/*
|
|
+ * get_ether_addr - get the hardware address of an interface on the
|
|
+ * the same subnet as ipaddr.
|
|
+ */
|
|
+static int
|
|
+get_ether_addr(ipaddr, hwaddr)
|
|
+ u_int32_t ipaddr;
|
|
+ struct sockaddr_dl *hwaddr;
|
|
+{
|
|
+ u_int32_t ina, mask;
|
|
+ struct sockaddr_dl *dla;
|
|
+ struct ifaddrs *ifap, *ifa, *ifp;
|
|
+
|
|
+ /*
|
|
+ * Scan through looking for an interface with an Internet
|
|
+ * address on the same subnet as `ipaddr'.
|
|
+ */
|
|
+ if (getifaddrs(&ifap) != 0) {
|
|
+ error("getifaddrs: %m");
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
|
|
+ if (ifa->ifa_addr->sa_family != AF_INET)
|
|
+ continue;
|
|
+ ina = ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr.s_addr;
|
|
+ /*
|
|
+ * Check that the interface is up, and not point-to-point
|
|
+ * or loopback.
|
|
+ */
|
|
+ if ((ifa->ifa_flags &
|
|
+ (IFF_UP|IFF_BROADCAST|IFF_POINTOPOINT|IFF_LOOPBACK|IFF_NOARP))
|
|
+ != (IFF_UP|IFF_BROADCAST))
|
|
+ continue;
|
|
+ /*
|
|
+ * Get its netmask and check that it's on the right subnet.
|
|
+ */
|
|
+ mask = ((struct sockaddr_in *) ifa->ifa_netmask)->sin_addr.s_addr;
|
|
+ if ((ipaddr & mask) != (ina & mask))
|
|
+ continue;
|
|
+ break;
|
|
+ }
|
|
+
|
|
+ if (!ifa) {
|
|
+ freeifaddrs(ifap);
|
|
+ return 0;
|
|
+ }
|
|
+ info("found interface %s for proxy arp", ifa->ifa_name);
|
|
+
|
|
+ ifp = ifa;
|
|
+
|
|
+ /*
|
|
+ * Now scan through again looking for a link-level address
|
|
+ * for this interface.
|
|
+ */
|
|
+ for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
|
|
+ if (strcmp(ifp->ifa_name, ifa->ifa_name) != 0)
|
|
+ continue;
|
|
+ if (ifa->ifa_addr->sa_family != AF_LINK)
|
|
+ continue;
|
|
+ /*
|
|
+ * Found the link-level address - copy it out
|
|
+ */
|
|
+ dla = (struct sockaddr_dl *) ifa->ifa_addr;
|
|
+ BCOPY(dla, hwaddr, dla->sdl_len);
|
|
+ freeifaddrs(ifap);
|
|
+ return 1;
|
|
+ }
|
|
+
|
|
+ freeifaddrs(ifap);
|
|
+ return 0;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * get_if_hwaddr - get the hardware address for the specified
|
|
+ * network interface device.
|
|
+ */
|
|
+int
|
|
+get_if_hwaddr(addr, name)
|
|
+ u_char *addr;
|
|
+ char *name;
|
|
+{
|
|
+ struct ifreq ifreq;
|
|
+ struct sockaddr_dl *sdl = (struct sockaddr_dl *) &ifreq.ifr_addr;
|
|
+ int fd;
|
|
+
|
|
+ if ((fd = socket(AF_INET, SOCK_DGRAM, 0)) == -1)
|
|
+ return 0;
|
|
+ (void)memset(sdl, 0, sizeof(*sdl));
|
|
+ sdl->sdl_family = AF_LINK;
|
|
+ (void)strlcpy(ifreq.ifr_name, name, sizeof(ifreq.ifr_name));
|
|
+ if (ioctl(fd, SIOCGIFADDR, &ifreq) == -1) {
|
|
+ (void)close(fd);
|
|
+ return 0;
|
|
+ }
|
|
+ (void)close(fd);
|
|
+ (void)memcpy(addr, LLADDR(sdl), sdl->sdl_alen);
|
|
+ return sdl->sdl_nlen;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * get_first_ethernet - return the name of the first ethernet-style
|
|
+ * interface on this system.
|
|
+ */
|
|
+char *
|
|
+get_first_ethernet()
|
|
+{
|
|
+ static char ifname[IFNAMSIZ];
|
|
+ struct ifaddrs *ifap, *ifa;
|
|
+
|
|
+ /*
|
|
+ * Scan through the system's network interfaces.
|
|
+ */
|
|
+ if (getifaddrs(&ifap) != 0) {
|
|
+ warn("getifaddrs: %m");
|
|
+ return NULL;
|
|
+ }
|
|
+ for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
|
|
+ /*
|
|
+ * Check the interface's internet address.
|
|
+ */
|
|
+ if (ifa->ifa_addr->sa_family != AF_INET)
|
|
+ continue;
|
|
+ /*
|
|
+ * Check that the interface is up, and not point-to-point or loopback.
|
|
+ */
|
|
+ if ((ifa->ifa_flags & (IFF_UP|IFF_POINTOPOINT|IFF_LOOPBACK))
|
|
+ != IFF_UP) {
|
|
+ strlcpy(ifname, ifa->ifa_name, sizeof(ifname));
|
|
+ freeifaddrs(ifap);
|
|
+ return ifname;
|
|
+ }
|
|
+ }
|
|
+ freeifaddrs(ifap);
|
|
+ return NULL;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * 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.
|
|
+ */
|
|
+u_int32_t
|
|
+GetMask(addr)
|
|
+ u_int32_t addr;
|
|
+{
|
|
+ u_int32_t mask, nmask, ina;
|
|
+ struct ifaddrs *ifap, *ifa;
|
|
+
|
|
+ addr = ntohl(addr);
|
|
+ if (IN_CLASSA(addr)) /* determine network mask for address class */
|
|
+ nmask = IN_CLASSA_NET;
|
|
+ else if (IN_CLASSB(addr))
|
|
+ nmask = IN_CLASSB_NET;
|
|
+ else
|
|
+ nmask = IN_CLASSC_NET;
|
|
+ /* class D nets are disallowed by bad_ip_adrs */
|
|
+ mask = netmask | htonl(nmask);
|
|
+
|
|
+ /*
|
|
+ * Scan through the system's network interfaces.
|
|
+ */
|
|
+ if (getifaddrs(&ifap) != 0) {
|
|
+ warn("getifaddrs: %m");
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
|
|
+ /*
|
|
+ * Check the interface's internet address.
|
|
+ */
|
|
+ if (ifa->ifa_addr->sa_family != AF_INET)
|
|
+ continue;
|
|
+ ina = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr;
|
|
+ if ((ntohl(ina) & nmask) != (addr & nmask))
|
|
+ continue;
|
|
+ /*
|
|
+ * Check that the interface is up, and not point-to-point or loopback.
|
|
+ */
|
|
+ if ((ifa->ifa_flags & (IFF_UP|IFF_POINTOPOINT|IFF_LOOPBACK)) != IFF_UP)
|
|
+ continue;
|
|
+ /*
|
|
+ * Get its netmask and OR it into our mask.
|
|
+ */
|
|
+ mask |= ((struct sockaddr_in *)&ifa->ifa_netmask)->sin_addr.s_addr;
|
|
+ }
|
|
+
|
|
+ freeifaddrs(ifap);
|
|
+ return mask;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * have_route_to - determine if the system has any route to
|
|
+ * a given IP address.
|
|
+ * For demand mode to work properly, we have to ignore routes
|
|
+ * through our own interface.
|
|
+ */
|
|
+int have_route_to(u_int32_t addr)
|
|
+{
|
|
+ return -1;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * Use the hostid as part of the random number seed.
|
|
+ */
|
|
+int
|
|
+get_host_seed()
|
|
+{
|
|
+ return gethostid();
|
|
+}
|
|
+
|
|
+#if 0
|
|
+/*
|
|
+ * lock - create a lock file for the named lock device
|
|
+ */
|
|
+#define LOCK_PREFIX "/var/spool/lock/LCK.."
|
|
+
|
|
+static char *lock_file; /* name of lock file created */
|
|
+
|
|
+int
|
|
+lock(dev)
|
|
+ char *dev;
|
|
+{
|
|
+ char hdb_lock_buffer[12];
|
|
+ int fd, pid, n;
|
|
+ char *p;
|
|
+ size_t l;
|
|
+
|
|
+ if ((p = strrchr(dev, '/')) != NULL)
|
|
+ dev = p + 1;
|
|
+ l = strlen(LOCK_PREFIX) + strlen(dev) + 1;
|
|
+ lock_file = malloc(l);
|
|
+ if (lock_file == NULL)
|
|
+ novm("lock file name");
|
|
+ slprintf(lock_file, l, "%s%s", LOCK_PREFIX, dev);
|
|
+
|
|
+ while ((fd = open(lock_file, O_EXCL | O_CREAT | O_RDWR, 0644)) < 0) {
|
|
+ if (errno == EEXIST
|
|
+ && (fd = open(lock_file, O_RDONLY, 0)) >= 0) {
|
|
+ /* Read the lock file to find out who has the device locked */
|
|
+ n = read(fd, hdb_lock_buffer, 11);
|
|
+ if (n <= 0) {
|
|
+ error("Can't read pid from lock file %s", lock_file);
|
|
+ close(fd);
|
|
+ } else {
|
|
+ hdb_lock_buffer[n] = 0;
|
|
+ pid = atoi(hdb_lock_buffer);
|
|
+ if (kill(pid, 0) == -1 && errno == ESRCH) {
|
|
+ /* pid no longer exists - remove the lock file */
|
|
+ if (unlink(lock_file) == 0) {
|
|
+ close(fd);
|
|
+ notice("Removed stale lock on %s (pid %d)",
|
|
+ dev, pid);
|
|
+ continue;
|
|
+ } else
|
|
+ warn("Couldn't remove stale lock on %s",
|
|
+ dev);
|
|
+ } else
|
|
+ notice("Device %s is locked by pid %d",
|
|
+ dev, pid);
|
|
+ }
|
|
+ close(fd);
|
|
+ } else
|
|
+ error("Can't create lock file %s: %m", lock_file);
|
|
+ free(lock_file);
|
|
+ lock_file = NULL;
|
|
+ return -1;
|
|
+ }
|
|
+
|
|
+ slprintf(hdb_lock_buffer, sizeof(hdb_lock_buffer), "%10d\n", getpid());
|
|
+ write(fd, hdb_lock_buffer, 11);
|
|
+
|
|
+ close(fd);
|
|
+ return 0;
|
|
+}
|
|
+
|
|
+/*
|
|
+ * unlock - remove our lockfile
|
|
+ */
|
|
+void
|
|
+unlock()
|
|
+{
|
|
+ if (lock_file) {
|
|
+ unlink(lock_file);
|
|
+ free(lock_file);
|
|
+ lock_file = NULL;
|
|
+ }
|
|
+}
|
|
+#endif
|