Mercurial > hg > dhcpcd
view src/if.c @ 5567:4fe5c2a71254 draft
hooks: add NOCARRIER_ROAMING reason
This is given when the OS supports the concept of wireless roaming
or the IP setup can be persisted when the carrier drops.
When this happens, routes are moved to a higher metric (if supported)
to support non preferred but non roaming routes.
The `interface_order` hook variable will now order the interfaces
according to priority and move roaming interfaces to the back of the
list.
If resolvconf is present then it is called with the -C option
to deprecate DNS and if carrier comes back it is called again with the
-c option to activate it once more.
As part of this change, default route metrics have been changed to
support a larger number of interfaces.
base metric 1000 (was 200)
wireless offset 2000 (was 100)
IPv4LL offset 1000000 (was 10000)
roaming offset 2000000
| author | Roy Marples <roy@marples.name> |
|---|---|
| date | Sun, 27 Dec 2020 19:53:31 +0000 |
| parents | 4cdb6a13f1b7 |
| children |
line wrap: on
line source
/* SPDX-License-Identifier: BSD-2-Clause */ /* * dhcpcd - DHCP client daemon * Copyright (c) 2006-2020 Roy Marples <roy@marples.name> * All rights reserved * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include <sys/param.h> #include <sys/types.h> #include <sys/ioctl.h> #include <sys/socket.h> #include <fcntl.h> /* Needs to be here for old Linux */ #include "config.h" #include <net/if.h> #include <net/if_arp.h> #include <netinet/in.h> #ifdef AF_LINK # include <net/if_dl.h> # include <net/if_types.h> # include <netinet/in_var.h> # undef AF_PACKET /* Newer Illumos defines this */ #endif #ifdef AF_PACKET # include <netpacket/packet.h> #endif #ifdef SIOCGIFMEDIA # include <net/if_media.h> #endif #include <net/route.h> #include <ctype.h> #include <errno.h> #include <ifaddrs.h> #include <inttypes.h> #include <fnmatch.h> #include <stddef.h> #include <stdio.h> #include <stdlib.h> #include <string.h> #include <syslog.h> #include <unistd.h> #define ELOOP_QUEUE ELOOP_IF #include "common.h" #include "eloop.h" #include "dev.h" #include "dhcp.h" #include "dhcp6.h" #include "if.h" #include "if-options.h" #include "ipv4.h" #include "ipv4ll.h" #include "ipv6nd.h" #include "logerr.h" #include "privsep.h" void if_free(struct interface *ifp) { if (ifp == NULL) return; #ifdef IPV4LL ipv4ll_free(ifp); #endif #ifdef INET dhcp_free(ifp); ipv4_free(ifp); #endif #ifdef DHCP6 dhcp6_free(ifp); #endif #ifdef INET6 ipv6nd_free(ifp); ipv6_free(ifp); #endif rt_freeif(ifp); free_options(ifp->ctx, ifp->options); free(ifp); } int if_opensockets(struct dhcpcd_ctx *ctx) { if (if_opensockets_os(ctx) == -1) return -1; #ifdef IFLR_ACTIVE ctx->pf_link_fd = xsocket(PF_LINK, SOCK_DGRAM | SOCK_CLOEXEC, 0); if (ctx->pf_link_fd == -1) return -1; #ifdef HAVE_CAPSICUM if (ps_rights_limit_ioctl(ctx->pf_link_fd) == -1) return -1; #endif #endif /* We use this socket for some operations without INET. */ ctx->pf_inet_fd = xsocket(PF_INET, SOCK_DGRAM | SOCK_CLOEXEC, 0); if (ctx->pf_inet_fd == -1) return -1; return 0; } void if_closesockets(struct dhcpcd_ctx *ctx) { if (ctx->pf_inet_fd != -1) close(ctx->pf_inet_fd); #ifdef PF_LINK if (ctx->pf_link_fd != -1) close(ctx->pf_link_fd); #endif if (ctx->priv) { if_closesockets_os(ctx); free(ctx->priv); } } int if_ioctl(struct dhcpcd_ctx *ctx, ioctl_request_t req, void *data, size_t len) { #ifdef PRIVSEP if (ctx->options & DHCPCD_PRIVSEP) return (int)ps_root_ioctl(ctx, req, data, len); #endif return ioctl(ctx->pf_inet_fd, req, data, len); } int if_getflags(struct interface *ifp) { struct ifreq ifr = { .ifr_flags = 0 }; strlcpy(ifr.ifr_name, ifp->name, sizeof(ifr.ifr_name)); if (ioctl(ifp->ctx->pf_inet_fd, SIOCGIFFLAGS, &ifr) == -1) return -1; ifp->flags = (unsigned int)ifr.ifr_flags; return 0; } int if_setflag(struct interface *ifp, short setflag, short unsetflag) { struct ifreq ifr = { .ifr_flags = 0 }; short oflags; strlcpy(ifr.ifr_name, ifp->name, sizeof(ifr.ifr_name)); if (ioctl(ifp->ctx->pf_inet_fd, SIOCGIFFLAGS, &ifr) == -1) return -1; oflags = ifr.ifr_flags; ifr.ifr_flags |= setflag; ifr.ifr_flags &= (short)~unsetflag; if (ifr.ifr_flags != oflags && if_ioctl(ifp->ctx, SIOCSIFFLAGS, &ifr, sizeof(ifr)) == -1) return -1; /* * Do NOT set ifp->flags here. * We need to listen for flag updates from the kernel as they * need to sync with carrier. */ return 0; } bool if_is_link_up(const struct interface *ifp) { return ifp->flags & IFF_UP && (ifp->carrier != LINK_DOWN || (ifp->options != NULL && !(ifp->options->options & DHCPCD_LINK))); } int if_randomisemac(struct interface *ifp) { uint32_t randnum; size_t hwlen = ifp->hwlen, rlen = 0; uint8_t buf[HWADDR_LEN], *bp = buf, *rp = (uint8_t *)&randnum; char sbuf[HWADDR_LEN * 3]; int retval; if (hwlen == 0) { errno = ENOTSUP; return -1; } if (hwlen > sizeof(buf)) { errno = ENOBUFS; return -1; } for (; hwlen != 0; hwlen--) { if (rlen == 0) { randnum = arc4random(); rp = (uint8_t *)&randnum; rlen = sizeof(randnum); } *bp++ = *rp++; rlen--; } /* Unicast address and locally administered. */ buf[0] &= 0xFC; buf[0] |= 0x02; logdebugx("%s: hardware address randomised to %s", ifp->name, hwaddr_ntoa(buf, ifp->hwlen, sbuf, sizeof(sbuf))); retval = if_setmac(ifp, buf, ifp->hwlen); if (retval == 0) memcpy(ifp->hwaddr, buf, ifp->hwlen); return retval; } static int if_hasconf(struct dhcpcd_ctx *ctx, const char *ifname) { int i; for (i = 0; i < ctx->ifcc; i++) { if (strcmp(ctx->ifcv[i], ifname) == 0) return 1; } return 0; } void if_markaddrsstale(struct if_head *ifs) { struct interface *ifp; TAILQ_FOREACH(ifp, ifs, next) { #ifdef INET ipv4_markaddrsstale(ifp); #endif #ifdef INET6 ipv6_markaddrsstale(ifp, 0); #endif } } void if_learnaddrs(struct dhcpcd_ctx *ctx, struct if_head *ifs, struct ifaddrs **ifaddrs) { struct ifaddrs *ifa; struct interface *ifp; #ifdef INET const struct sockaddr_in *addr, *net, *brd; #endif #ifdef INET6 struct sockaddr_in6 *sin6, *net6; #endif int addrflags; for (ifa = *ifaddrs; ifa; ifa = ifa->ifa_next) { if (ifa->ifa_addr == NULL) continue; if ((ifp = if_find(ifs, ifa->ifa_name)) == NULL) continue; #ifdef HAVE_IFADDRS_ADDRFLAGS addrflags = (int)ifa->ifa_addrflags; #endif switch(ifa->ifa_addr->sa_family) { #ifdef INET case AF_INET: addr = (void *)ifa->ifa_addr; net = (void *)ifa->ifa_netmask; if (ifa->ifa_flags & IFF_POINTOPOINT) brd = (void *)ifa->ifa_dstaddr; else brd = (void *)ifa->ifa_broadaddr; #ifndef HAVE_IFADDRS_ADDRFLAGS addrflags = if_addrflags(ifp, &addr->sin_addr, ifa->ifa_name); if (addrflags == -1) { if (errno != EEXIST && errno != EADDRNOTAVAIL) { char dbuf[INET_ADDRSTRLEN]; const char *dbp; dbp = inet_ntop(AF_INET, &addr->sin_addr, dbuf, sizeof(dbuf)); logerr("%s: if_addrflags: %s%%%s", __func__, dbp, ifp->name); } continue; } #endif ipv4_handleifa(ctx, RTM_NEWADDR, ifs, ifa->ifa_name, &addr->sin_addr, &net->sin_addr, brd ? &brd->sin_addr : NULL, addrflags, 0); break; #endif #ifdef INET6 case AF_INET6: sin6 = (void *)ifa->ifa_addr; net6 = (void *)ifa->ifa_netmask; #ifdef __KAME__ if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) /* Remove the scope from the address */ sin6->sin6_addr.s6_addr[2] = sin6->sin6_addr.s6_addr[3] = '\0'; #endif #ifndef HAVE_IFADDRS_ADDRFLAGS addrflags = if_addrflags6(ifp, &sin6->sin6_addr, ifa->ifa_name); if (addrflags == -1) { if (errno != EEXIST && errno != EADDRNOTAVAIL) { char dbuf[INET6_ADDRSTRLEN]; const char *dbp; dbp = inet_ntop(AF_INET6, &sin6->sin6_addr, dbuf, sizeof(dbuf)); logerr("%s: if_addrflags6: %s%%%s", __func__, dbp, ifp->name); } continue; } #endif ipv6_handleifa(ctx, RTM_NEWADDR, ifs, ifa->ifa_name, &sin6->sin6_addr, ipv6_prefixlen(&net6->sin6_addr), addrflags, 0); break; #endif } } #ifdef PRIVSEP_GETIFADDRS if (IN_PRIVSEP(ctx)) free(*ifaddrs); else #endif freeifaddrs(*ifaddrs); *ifaddrs = NULL; } void if_deletestaleaddrs(struct if_head *ifs) { struct interface *ifp; TAILQ_FOREACH(ifp, ifs, next) { #ifdef INET ipv4_deletestaleaddrs(ifp); #endif #ifdef INET6 ipv6_deletestaleaddrs(ifp); #endif } } bool if_valid_hwaddr(const uint8_t *hwaddr, size_t hwlen) { size_t i; bool all_zeros, all_ones; all_zeros = all_ones = true; for (i = 0; i < hwlen; i++) { if (hwaddr[i] != 0x00) all_zeros = false; if (hwaddr[i] != 0xff) all_ones = false; if (!all_zeros && !all_ones) return true; } return false; } #if defined(AF_PACKET) && !defined(AF_LINK) static unsigned int if_check_arphrd(struct interface *ifp, unsigned int active, bool if_noconf) { switch(ifp->hwtype) { case ARPHRD_ETHER: /* FALLTHROUGH */ case ARPHRD_IEEE1394: /* FALLTHROUGH */ case ARPHRD_INFINIBAND: /* FALLTHROUGH */ case ARPHRD_NONE: /* FALLTHROUGH */ break; case ARPHRD_LOOPBACK: case ARPHRD_PPP: if (if_noconf && active) { logdebugx("%s: ignoring due to interface type and" " no config", ifp->name); active = IF_INACTIVE; } break; default: if (active) { int i; if (if_noconf) active = IF_INACTIVE; i = active ? LOG_WARNING : LOG_DEBUG; logmessage(i, "%s: unsupported" " interface type 0x%.2x", ifp->name, ifp->hwtype); } break; } return active; } #endif struct if_head * if_discover(struct dhcpcd_ctx *ctx, struct ifaddrs **ifaddrs, int argc, char * const *argv) { struct ifaddrs *ifa; int i; unsigned int active; struct if_head *ifs; struct interface *ifp; struct if_spec spec; bool if_noconf; #ifdef AF_LINK const struct sockaddr_dl *sdl; #ifdef IFLR_ACTIVE struct if_laddrreq iflr = { .flags = IFLR_PREFIX }; #endif #elif defined(AF_PACKET) const struct sockaddr_ll *sll; #endif #if defined(SIOCGIFPRIORITY) struct ifreq ifr; #endif if ((ifs = malloc(sizeof(*ifs))) == NULL) { logerr(__func__); return NULL; } TAILQ_INIT(ifs); #ifdef PRIVSEP_GETIFADDRS if (ctx->options & DHCPCD_PRIVSEP) { if (ps_root_getifaddrs(ctx, ifaddrs) == -1) { logerr("ps_root_getifaddrs"); free(ifs); return NULL; } } else #endif if (getifaddrs(ifaddrs) == -1) { logerr("getifaddrs"); free(ifs); return NULL; } for (ifa = *ifaddrs; ifa; ifa = ifa->ifa_next) { if (ifa->ifa_addr != NULL) { #ifdef AF_LINK if (ifa->ifa_addr->sa_family != AF_LINK) continue; #elif defined(AF_PACKET) if (ifa->ifa_addr->sa_family != AF_PACKET) continue; #endif } if (if_nametospec(ifa->ifa_name, &spec) != 0) continue; /* It's possible for an interface to have >1 AF_LINK. * For our purposes, we use the first one. */ TAILQ_FOREACH(ifp, ifs, next) { if (strcmp(ifp->name, spec.devname) == 0) break; } if (ifp) continue; if (argc > 0) { for (i = 0; i < argc; i++) { if (strcmp(argv[i], spec.devname) == 0) break; } active = (i == argc) ? IF_INACTIVE : IF_ACTIVE_USER; } else { /* -1 means we're discovering against a specific * interface, but we still need the below rules * to apply. */ if (argc == -1 && strcmp(argv[0], spec.devname) != 0) continue; active = ctx->options & DHCPCD_INACTIVE ? IF_INACTIVE: IF_ACTIVE_USER; } for (i = 0; i < ctx->ifdc; i++) if (fnmatch(ctx->ifdv[i], spec.devname, 0) == 0) break; if (i < ctx->ifdc) active = IF_INACTIVE; for (i = 0; i < ctx->ifc; i++) if (fnmatch(ctx->ifv[i], spec.devname, 0) == 0) break; if (ctx->ifc && i == ctx->ifc) active = IF_INACTIVE; for (i = 0; i < ctx->ifac; i++) if (fnmatch(ctx->ifav[i], spec.devname, 0) == 0) break; if (ctx->ifac && i == ctx->ifac) active = IF_INACTIVE; #ifdef PLUGIN_DEV /* Ensure that the interface name has settled */ if (!dev_initialised(ctx, spec.devname)) { logdebugx("%s: waiting for interface to initialise", spec.devname); continue; } #endif if (if_vimaster(ctx, spec.devname) == 1) { int loglevel = argc != 0 ? LOG_ERR : LOG_DEBUG; logmessage(loglevel, "%s: is a Virtual Interface Master, skipping", spec.devname); continue; } if_noconf = ((argc == 0 || argc == -1) && ctx->ifac == 0 && !if_hasconf(ctx, spec.devname)); /* Don't allow some reserved interface names unless explicit. */ if (if_noconf && if_ignore(ctx, spec.devname)) { logdebugx("%s: ignoring due to interface type and" " no config", spec.devname); active = IF_INACTIVE; } ifp = calloc(1, sizeof(*ifp)); if (ifp == NULL) { logerr(__func__); break; } ifp->ctx = ctx; strlcpy(ifp->name, spec.devname, sizeof(ifp->name)); ifp->flags = ifa->ifa_flags; if (ifa->ifa_addr != NULL) { #ifdef AF_LINK sdl = (const void *)ifa->ifa_addr; #ifdef IFLR_ACTIVE /* We need to check for active address */ strlcpy(iflr.iflr_name, ifp->name, sizeof(iflr.iflr_name)); memcpy(&iflr.addr, ifa->ifa_addr, MIN(ifa->ifa_addr->sa_len, sizeof(iflr.addr))); iflr.flags = IFLR_PREFIX; iflr.prefixlen = (unsigned int)sdl->sdl_alen * NBBY; if (ioctl(ctx->pf_link_fd, SIOCGLIFADDR, &iflr) == -1 || !(iflr.flags & IFLR_ACTIVE)) { if_free(ifp); continue; } #endif ifp->index = sdl->sdl_index; switch(sdl->sdl_type) { #ifdef IFT_BRIDGE case IFT_BRIDGE: /* FALLTHROUGH */ #endif #ifdef IFT_PROPVIRTUAL case IFT_PROPVIRTUAL: /* FALLTHROUGH */ #endif #ifdef IFT_TUNNEL case IFT_TUNNEL: /* FALLTHROUGH */ #endif case IFT_LOOP: /* FALLTHROUGH */ case IFT_PPP: /* Don't allow unless explicit */ if (if_noconf && active) { logdebugx("%s: ignoring due to" " interface type and" " no config", ifp->name); active = IF_INACTIVE; } __fallthrough; /* appease gcc */ /* FALLTHROUGH */ #ifdef IFT_L2VLAN case IFT_L2VLAN: /* FALLTHROUGH */ #endif #ifdef IFT_L3IPVLAN case IFT_L3IPVLAN: /* FALLTHROUGH */ #endif case IFT_ETHER: ifp->hwtype = ARPHRD_ETHER; break; #ifdef IFT_IEEE1394 case IFT_IEEE1394: ifp->hwtype = ARPHRD_IEEE1394; break; #endif #ifdef IFT_INFINIBAND case IFT_INFINIBAND: ifp->hwtype = ARPHRD_INFINIBAND; break; #endif default: /* Don't allow unless explicit */ if (active) { if (if_noconf) active = IF_INACTIVE; i = active ? LOG_WARNING : LOG_DEBUG; logmessage(i, "%s: unsupported" " interface type 0x%.2x", ifp->name, sdl->sdl_type); } /* Pretend it's ethernet */ ifp->hwtype = ARPHRD_ETHER; break; } ifp->hwlen = sdl->sdl_alen; memcpy(ifp->hwaddr, CLLADDR(sdl), ifp->hwlen); #elif defined(AF_PACKET) sll = (const void *)ifa->ifa_addr; ifp->index = (unsigned int)sll->sll_ifindex; ifp->hwtype = sll->sll_hatype; ifp->hwlen = sll->sll_halen; if (ifp->hwlen != 0) memcpy(ifp->hwaddr, sll->sll_addr, ifp->hwlen); active = if_check_arphrd(ifp, active, if_noconf); #endif } #ifdef __linux__ else { struct ifreq ifr = { .ifr_flags = 0 }; /* This is a huge bug in getifaddrs(3) as there * is no reason why this can't be returned in * ifa_addr. */ strlcpy(ifr.ifr_name, ifa->ifa_name, sizeof(ifr.ifr_name)); if (ioctl(ctx->pf_inet_fd, SIOCGIFHWADDR, &ifr) == -1) logerr("%s: SIOCGIFHWADDR", ifa->ifa_name); ifp->hwtype = ifr.ifr_hwaddr.sa_family; if (ioctl(ctx->pf_inet_fd, SIOCGIFINDEX, &ifr) == -1) logerr("%s: SIOCGIFINDEX", ifa->ifa_name); ifp->index = (unsigned int)ifr.ifr_ifindex; if_check_arphrd(ifp, active, if_noconf); } #endif if (!(ctx->options & (DHCPCD_DUMPLEASE | DHCPCD_TEST))) { /* Handle any platform init for the interface */ if (active != IF_INACTIVE && if_init(ifp) == -1) { logerr("%s: if_init", ifp->name); if_free(ifp); continue; } } ifp->vlanid = if_vlanid(ifp); #ifdef SIOCGIFPRIORITY /* Respect the interface priority */ memset(&ifr, 0, sizeof(ifr)); strlcpy(ifr.ifr_name, ifp->name, sizeof(ifr.ifr_name)); if (pioctl(ctx, SIOCGIFPRIORITY, &ifr, sizeof(ifr)) == 0) ifp->metric = (unsigned int)ifr.ifr_metric; if_getssid(ifp); #else /* Leave a low portion for user config */ ifp->metric = RTMETRIC_BASE + ifp->index; if (if_getssid(ifp) != -1) { ifp->wireless = true; ifp->metric += RTMETRIC_WIRELESS; } #endif ifp->active = active; ifp->carrier = if_carrier(ifp, ifa->ifa_data); TAILQ_INSERT_TAIL(ifs, ifp, next); } return ifs; } /* * eth0.100:2 OR eth0i100:2 (seems to be NetBSD xvif(4) only) * * drvname == eth * devname == eth0.100 OR eth0i100 * ppa = 0 * lun = 2 */ int if_nametospec(const char *ifname, struct if_spec *spec) { char *ep, *pp; int e; if (ifname == NULL || *ifname == '\0' || strlcpy(spec->ifname, ifname, sizeof(spec->ifname)) >= sizeof(spec->ifname) || strlcpy(spec->drvname, ifname, sizeof(spec->drvname)) >= sizeof(spec->drvname)) { errno = EINVAL; return -1; } /* :N is an alias */ ep = strchr(spec->drvname, ':'); if (ep) { spec->lun = (int)strtoi(ep + 1, NULL, 10, 0, INT_MAX, &e); if (e != 0) { errno = e; return -1; } *ep = '\0'; #ifdef __sun ep--; #endif } else { spec->lun = -1; #ifdef __sun ep = spec->drvname + strlen(spec->drvname) - 1; #endif } strlcpy(spec->devname, spec->drvname, sizeof(spec->devname)); #ifdef __sun /* Solaris has numbers in the driver name, such as e1000g */ while (ep > spec->drvname && isdigit((int)*ep)) ep--; if (*ep++ == ':') { errno = EINVAL; return -1; } #else /* BSD and Linux no not have numbers in the driver name */ for (ep = spec->drvname; *ep != '\0' && !isdigit((int)*ep); ep++) { if (*ep == ':') { errno = EINVAL; return -1; } } #endif spec->ppa = (int)strtoi(ep, &pp, 10, 0, INT_MAX, &e); *ep = '\0'; #ifndef __sun /* * . is used for VLAN style names * i is used on NetBSD for xvif interfaces */ if (pp != NULL && (*pp == '.' || *pp == 'i')) { spec->vlid = (int)strtoi(pp + 1, NULL, 10, 0, INT_MAX, &e); if (e) spec->vlid = -1; } else #endif spec->vlid = -1; return 0; } static struct interface * if_findindexname(struct if_head *ifaces, unsigned int idx, const char *name) { if (ifaces != NULL) { struct if_spec spec; struct interface *ifp; if (name && if_nametospec(name, &spec) == -1) return NULL; TAILQ_FOREACH(ifp, ifaces, next) { if ((name && strcmp(ifp->name, spec.devname) == 0) || (!name && ifp->index == idx)) return ifp; } } errno = ENXIO; return NULL; } struct interface * if_find(struct if_head *ifaces, const char *name) { return if_findindexname(ifaces, 0, name); } struct interface * if_findindex(struct if_head *ifaces, unsigned int idx) { return if_findindexname(ifaces, idx, NULL); } struct interface * if_loopback(struct dhcpcd_ctx *ctx) { struct interface *ifp; TAILQ_FOREACH(ifp, ctx->ifaces, next) { if (ifp->flags & IFF_LOOPBACK) return ifp; } return NULL; } int if_domtu(const struct interface *ifp, short int mtu) { int r; struct ifreq ifr; #ifdef __sun if (mtu == 0) return if_mtu_os(ifp); #endif memset(&ifr, 0, sizeof(ifr)); strlcpy(ifr.ifr_name, ifp->name, sizeof(ifr.ifr_name)); ifr.ifr_mtu = mtu; if (mtu != 0) r = if_ioctl(ifp->ctx, SIOCSIFMTU, &ifr, sizeof(ifr)); else r = pioctl(ifp->ctx, SIOCGIFMTU, &ifr, sizeof(ifr)); if (r == -1) return -1; return ifr.ifr_mtu; } #ifdef ALIAS_ADDR int if_makealias(char *alias, size_t alias_len, const char *ifname, int lun) { if (lun == 0) return strlcpy(alias, ifname, alias_len); return snprintf(alias, alias_len, "%s:%u", ifname, lun); } #endif struct interface * if_findifpfromcmsg(struct dhcpcd_ctx *ctx, struct msghdr *msg, int *hoplimit) { struct cmsghdr *cm; unsigned int ifindex = 0; struct interface *ifp; #ifdef INET #ifdef IP_RECVIF struct sockaddr_dl sdl; #else struct in_pktinfo ipi; #endif #endif #ifdef INET6 struct in6_pktinfo ipi6; #else UNUSED(hoplimit); #endif for (cm = (struct cmsghdr *)CMSG_FIRSTHDR(msg); cm; cm = (struct cmsghdr *)CMSG_NXTHDR(msg, cm)) { #ifdef INET if (cm->cmsg_level == IPPROTO_IP) { switch(cm->cmsg_type) { #ifdef IP_RECVIF case IP_RECVIF: if (cm->cmsg_len < offsetof(struct sockaddr_dl, sdl_index) + sizeof(sdl.sdl_index)) continue; memcpy(&sdl, CMSG_DATA(cm), MIN(sizeof(sdl), cm->cmsg_len)); ifindex = sdl.sdl_index; break; #else case IP_PKTINFO: if (cm->cmsg_len != CMSG_LEN(sizeof(ipi))) continue; memcpy(&ipi, CMSG_DATA(cm), sizeof(ipi)); ifindex = (unsigned int)ipi.ipi_ifindex; break; #endif } } #endif #ifdef INET6 if (cm->cmsg_level == IPPROTO_IPV6) { switch(cm->cmsg_type) { case IPV6_PKTINFO: if (cm->cmsg_len != CMSG_LEN(sizeof(ipi6))) continue; memcpy(&ipi6, CMSG_DATA(cm), sizeof(ipi6)); ifindex = (unsigned int)ipi6.ipi6_ifindex; break; case IPV6_HOPLIMIT: if (cm->cmsg_len != CMSG_LEN(sizeof(int))) continue; if (hoplimit == NULL) break; memcpy(hoplimit, CMSG_DATA(cm), sizeof(int)); break; } } #endif } /* Find the receiving interface */ TAILQ_FOREACH(ifp, ctx->ifaces, next) { if (ifp->index == ifindex) break; } if (ifp == NULL) errno = ESRCH; return ifp; } int xsocket(int domain, int type, int protocol) { int s; #if !defined(HAVE_SOCK_CLOEXEC) || !defined(HAVE_SOCK_NONBLOCK) int xflags, xtype = type; #endif #ifndef HAVE_SOCK_CLOEXEC if (xtype & SOCK_CLOEXEC) type &= ~SOCK_CLOEXEC; #endif #ifndef HAVE_SOCK_NONBLOCK if (xtype & SOCK_NONBLOCK) type &= ~SOCK_NONBLOCK; #endif if ((s = socket(domain, type, protocol)) == -1) return -1; #ifndef HAVE_SOCK_CLOEXEC if ((xtype & SOCK_CLOEXEC) && ((xflags = fcntl(s, F_GETFD)) == -1 || fcntl(s, F_SETFD, xflags | FD_CLOEXEC) == -1)) goto out; #endif #ifndef HAVE_SOCK_NONBLOCK if ((xtype & SOCK_NONBLOCK) && ((xflags = fcntl(s, F_GETFL)) == -1 || fcntl(s, F_SETFL, xflags | O_NONBLOCK) == -1)) goto out; #endif return s; #if !defined(HAVE_SOCK_CLOEXEC) || !defined(HAVE_SOCK_NONBLOCK) out: close(s); return -1; #endif } int xsocketpair(int domain, int type, int protocol, int fd[2]) { int s; #if !defined(HAVE_SOCK_CLOEXEC) || !defined(HAVE_SOCK_NONBLOCK) int xflags, xtype = type; #endif #ifndef HAVE_SOCK_CLOEXEC if (xtype & SOCK_CLOEXEC) type &= ~SOCK_CLOEXEC; #endif #ifndef HAVE_SOCK_NONBLOCK if (xtype & SOCK_NONBLOCK) type &= ~SOCK_NONBLOCK; #endif if ((s = socketpair(domain, type, protocol, fd)) == -1) return -1; #ifndef HAVE_SOCK_CLOEXEC if ((xtype & SOCK_CLOEXEC) && ((xflags = fcntl(fd[0], F_GETFD)) == -1 || fcntl(fd[0], F_SETFD, xflags | FD_CLOEXEC) == -1)) goto out; if ((xtype & SOCK_CLOEXEC) && ((xflags = fcntl(fd[1], F_GETFD)) == -1 || fcntl(fd[1], F_SETFD, xflags | FD_CLOEXEC) == -1)) goto out; #endif #ifndef HAVE_SOCK_NONBLOCK if ((xtype & SOCK_NONBLOCK) && ((xflags = fcntl(fd[0], F_GETFL)) == -1 || fcntl(fd[0], F_SETFL, xflags | O_NONBLOCK) == -1)) goto out; if ((xtype & SOCK_NONBLOCK) && ((xflags = fcntl(fd[1], F_GETFL)) == -1 || fcntl(fd[1], F_SETFL, xflags | O_NONBLOCK) == -1)) goto out; #endif return s; #if !defined(HAVE_SOCK_CLOEXEC) || !defined(HAVE_SOCK_NONBLOCK) out: close(fd[0]); close(fd[1]); return -1; #endif }
