/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ /*** This file is part of systemd. Copyright 2010 Lennart Poettering systemd is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. systemd is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ #include <errno.h> #include <sys/socket.h> #include <net/if.h> #include <asm/types.h> #include <netinet/in.h> #include <string.h> #include <stdlib.h> #include <unistd.h> #include <linux/netlink.h> #include <linux/rtnetlink.h> #include "util.h" #include "macro.h" #include "loopback-setup.h" #include "socket-util.h" #define NLMSG_TAIL(nmsg) \ ((struct rtattr *) (((uint8_t*) (nmsg)) + NLMSG_ALIGN((nmsg)->nlmsg_len))) static int add_rtattr(struct nlmsghdr *n, size_t max_length, int type, const void *data, size_t data_length) { size_t length; struct rtattr *rta; length = RTA_LENGTH(data_length); if (NLMSG_ALIGN(n->nlmsg_len) + RTA_ALIGN(length) > max_length) return -E2BIG; rta = NLMSG_TAIL(n); rta->rta_type = type; rta->rta_len = length; memcpy(RTA_DATA(rta), data, data_length); n->nlmsg_len = NLMSG_ALIGN(n->nlmsg_len) + RTA_ALIGN(length); return 0; } static ssize_t sendto_loop(int fd, const void *buf, size_t buf_len, int flags, const struct sockaddr *sa, socklen_t sa_len) { for (;;) { ssize_t l; if ((l = sendto(fd, buf, buf_len, flags, sa, sa_len)) >= 0) return l; if (errno != EINTR) return -errno; } } static ssize_t recvfrom_loop(int fd, void *buf, size_t buf_len, int flags, struct sockaddr *sa, socklen_t *sa_len) { for (;;) { ssize_t l; if ((l = recvfrom(fd, buf, buf_len, flags, sa, sa_len)) >= 0) return l; if (errno != EINTR) return -errno; } } static int add_adresses(int fd, int if_loopback, unsigned *requests) { union { struct sockaddr sa; struct sockaddr_nl nl; } sa; union { struct nlmsghdr header; uint8_t buf[NLMSG_ALIGN(sizeof(struct nlmsghdr)) + NLMSG_ALIGN(sizeof(struct ifaddrmsg)) + RTA_LENGTH(sizeof(struct in6_addr))]; } request; struct ifaddrmsg *ifaddrmsg; uint32_t ipv4_address = htonl(INADDR_LOOPBACK); int r; zero(request); request.header.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifaddrmsg)); request.header.nlmsg_type = RTM_NEWADDR; request.header.nlmsg_flags = NLM_F_REQUEST|NLM_F_CREATE|NLM_F_ACK; request.header.nlmsg_seq = *requests + 1; ifaddrmsg = NLMSG_DATA(&request.header); ifaddrmsg->ifa_family = AF_INET; ifaddrmsg->ifa_prefixlen = 8; ifaddrmsg->ifa_flags = IFA_F_PERMANENT; ifaddrmsg->ifa_scope = RT_SCOPE_HOST; ifaddrmsg->ifa_index = if_loopback; if ((r = add_rtattr(&request.header, sizeof(request), IFA_LOCAL, &ipv4_address, sizeof(ipv4_address))) < 0) return r; zero(sa); sa.nl.nl_family = AF_NETLINK; if (sendto_loop(fd, &request, request.header.nlmsg_len, 0, &sa.sa, sizeof(sa)) < 0) return -errno; (*requests)++; if (!socket_ipv6_is_supported()) return 0; request.header.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifaddrmsg)); request.header.nlmsg_seq = *requests + 1; ifaddrmsg->ifa_family = AF_INET6; ifaddrmsg->ifa_prefixlen = 128; if ((r = add_rtattr(&request.header, sizeof(request), IFA_LOCAL, &in6addr_loopback, sizeof(in6addr_loopback))) < 0) return r; if (sendto_loop(fd, &request, request.header.nlmsg_len, 0, &sa.sa, sizeof(sa)) < 0) return -errno; (*requests)++; return 0; } static int start_interface(int fd, int if_loopback, unsigned *requests) { union { struct sockaddr sa; struct sockaddr_nl nl; } sa; union { struct nlmsghdr header; uint8_t buf[NLMSG_ALIGN(sizeof(struct nlmsghdr)) + NLMSG_ALIGN(sizeof(struct ifinfomsg))]; } request; struct ifinfomsg *ifinfomsg; zero(request); request.header.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifinfomsg)); request.header.nlmsg_type = RTM_NEWLINK; request.header.nlmsg_flags = NLM_F_REQUEST|NLM_F_ACK; request.header.nlmsg_seq = *requests + 1; ifinfomsg = NLMSG_DATA(&request.header); ifinfomsg->ifi_family = AF_UNSPEC; ifinfomsg->ifi_index = if_loopback; ifinfomsg->ifi_flags = IFF_UP; ifinfomsg->ifi_change = IFF_UP; zero(sa); sa.nl.nl_family = AF_NETLINK; if (sendto_loop(fd, &request, request.header.nlmsg_len, 0, &sa.sa, sizeof(sa)) < 0) return -errno; (*requests)++; return 0; } static int read_response(int fd, unsigned requests_max) { union { struct sockaddr sa; struct sockaddr_nl nl; } sa; union { struct nlmsghdr header; uint8_t buf[NLMSG_ALIGN(sizeof(struct nlmsghdr)) + NLMSG_ALIGN(sizeof(struct nlmsgerr))]; } response; ssize_t l; socklen_t sa_len = sizeof(sa); struct nlmsgerr *nlmsgerr; if ((l = recvfrom_loop(fd, &response, sizeof(response), 0, &sa.sa, &sa_len)) < 0) return -errno; if (sa_len != sizeof(sa.nl) || sa.nl.nl_family != AF_NETLINK) return -EIO; if (sa.nl.nl_pid != 0) return 0; if ((size_t) l < sizeof(struct nlmsghdr)) return -EIO; if (response.header.nlmsg_type != NLMSG_ERROR || (pid_t) response.header.nlmsg_pid != getpid() || response.header.nlmsg_seq >= requests_max) return 0; if ((size_t) l < NLMSG_LENGTH(sizeof(struct nlmsgerr)) || response.header.nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr))) return -EIO; nlmsgerr = NLMSG_DATA(&response.header); if (nlmsgerr->error < 0 && nlmsgerr->error != -EEXIST) { log_warning("Netlink failure for request %i: %s", response.header.nlmsg_seq, strerror(-nlmsgerr->error)); return nlmsgerr->error; } return response.header.nlmsg_seq; } int loopback_setup(void) { int r, if_loopback; union { struct sockaddr sa; struct sockaddr_nl nl; struct sockaddr_storage storage; } sa; unsigned requests = 0, i; int fd; errno = 0; if ((if_loopback = (int) if_nametoindex("lo")) <= 0) return errno ? -errno : -ENODEV; if ((fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE)) < 0) return -errno; zero(sa); sa.nl.nl_family = AF_NETLINK; if (bind(fd, &sa.sa, sizeof(sa)) < 0) { r = -errno; goto finish; } if ((r = add_adresses(fd, if_loopback, &requests)) < 0) goto finish; if ((r = start_interface(fd, if_loopback, &requests)) < 0) goto finish; for (i = 0; i < requests; i++) { if ((r = read_response(fd, requests)) < 0) goto finish; } r = 0; finish: if (r < 0) log_error("Failed to configure loopback device: %s", strerror(-r)); if (fd >= 0) close_nointr_nofail(fd); return r; }