/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ /*** This file is part of systemd. Copyright 2013 Tom Gundersen systemd is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2.1 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 Lesser General Public License for more details. You should have received a copy of the GNU Lesser General Public License along with systemd; If not, see . ***/ #include #include #include #include #include "util.h" #include "refcnt.h" #include "sd-rtnl.h" #include "rtnl-internal.h" struct sd_rtnl_message { RefCount n_ref; struct nlmsghdr *hdr; struct rtattr *next_rta; size_t remaining_size; bool sealed:1; }; static int message_new(sd_rtnl_message **ret, size_t initial_size) { sd_rtnl_message *m; assert_return(ret, -EINVAL); assert_return(initial_size >= sizeof(struct nlmsghdr), -EINVAL); m = new0(sd_rtnl_message, 1); if (!m) return -ENOMEM; m->hdr = malloc0(initial_size); if (!m->hdr) { free(m); return -ENOMEM; } m->n_ref = REFCNT_INIT; m->hdr->nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK; m->sealed = false; *ret = m; return 0; } int sd_rtnl_message_link_new(uint16_t nlmsg_type, int index, unsigned int type, unsigned int flags, sd_rtnl_message **ret) { struct ifinfomsg *ifi; int r; assert_return(nlmsg_type == RTM_NEWLINK || nlmsg_type == RTM_DELLINK || nlmsg_type == RTM_GETLINK, -EINVAL); assert_return(index > 0, -EINVAL); assert_return(ret, -EINVAL); r = message_new(ret, NLMSG_SPACE(sizeof(struct ifinfomsg))); if (r < 0) return r; (*ret)->hdr->nlmsg_len = NLMSG_LENGTH(sizeof(struct ifinfomsg)); (*ret)->hdr->nlmsg_type = nlmsg_type; ifi = NLMSG_DATA((*ret)->hdr); ifi->ifi_family = AF_UNSPEC; ifi->ifi_index = index; ifi->ifi_type = type; ifi->ifi_flags = flags; ifi->ifi_change = 0xffffffff; return 0; } int sd_rtnl_message_addr_new(uint16_t nlmsg_type, int index, unsigned char family, unsigned char prefixlen, unsigned char flags, unsigned char scope, sd_rtnl_message **ret) { struct ifaddrmsg *ifa; int r; assert_return(nlmsg_type == RTM_NEWADDR || nlmsg_type == RTM_DELADDR || nlmsg_type == RTM_GETADDR, -EINVAL); assert_return(index > 0, -EINVAL); assert_return(ret, -EINVAL); r = message_new(ret, NLMSG_SPACE(sizeof(struct ifaddrmsg))); if (r < 0) return r; (*ret)->hdr->nlmsg_len = NLMSG_LENGTH(sizeof(struct ifaddrmsg)); (*ret)->hdr->nlmsg_type = nlmsg_type; ifa = NLMSG_DATA((*ret)->hdr); ifa->ifa_family = family; ifa->ifa_prefixlen = prefixlen; ifa->ifa_flags = flags; ifa->ifa_scope = scope; ifa->ifa_index = index; return 0; } sd_rtnl_message *sd_rtnl_message_ref(sd_rtnl_message *m) { if (m) assert_se(REFCNT_INC(m->n_ref) >= 2); return m; } sd_rtnl_message *sd_rtnl_message_unref(sd_rtnl_message *m) { if (m && REFCNT_DEC(m->n_ref) <= 0) { free(m->hdr); free(m); } return NULL; } int sd_rtnl_message_get_type(sd_rtnl_message *m, uint16_t *type) { assert_return(m, -EINVAL); assert_return(type, -EINVAL); *type = m->hdr->nlmsg_type; return 0; } /* If successful the updated message will be correctly aligned, if unsuccessful the old message is untouched */ static int add_rtattr(sd_rtnl_message *m, unsigned short type, const void *data, size_t data_length) { uint32_t rta_length, message_length; struct nlmsghdr *new_hdr; struct rtattr *rta; assert_return(m, -EINVAL); assert_return(m->hdr, -EINVAL); assert_return(NLMSG_ALIGN(m->hdr->nlmsg_len) == m->hdr->nlmsg_len, -EINVAL); assert_return(data, -EINVAL); assert_return(data_length > 0, -EINVAL); /* get the size of the new rta attribute (without padding at the end) */ rta_length = RTA_LENGTH(data_length); /* get the new message size (with padding between the old message and the new attrib, * but no padding after) */ message_length = m->hdr->nlmsg_len + RTA_ALIGN(rta_length); /* realloc to fit the new attribute */ new_hdr = realloc(m->hdr, message_length); if (!new_hdr) return -ENOMEM; m->hdr = new_hdr; /* get pointer to the attribute we are about to add */ rta = (struct rtattr *) ((uint8_t *) m->hdr + m->hdr->nlmsg_len); /* update message size */ m->hdr->nlmsg_len = message_length; /* fill in the attribute */ rta->rta_type = type; rta->rta_len = rta_length; /* we don't deal with the case where the user lies about the type and gives us * too little data (so don't do that) */ memcpy(RTA_DATA(rta), data, data_length); return 0; } int sd_rtnl_message_append(sd_rtnl_message *m, unsigned short type, const void *data) { uint16_t rtm_type; struct ifaddrmsg *ifa; assert_return(m, -EINVAL); assert_return(data, -EINVAL); sd_rtnl_message_get_type(m, &rtm_type); switch (rtm_type) { case RTM_NEWLINK: case RTM_DELLINK: case RTM_GETLINK: switch (type) { case IFLA_IFNAME: case IFLA_QDISC: return add_rtattr(m, type, data, strlen(data) + 1); case IFLA_MTU: return add_rtattr(m, type, data, sizeof(unsigned int)); case IFLA_LINK: return add_rtattr(m, type, data, sizeof(int)); case IFLA_STATS: return add_rtattr(m, type, data, sizeof(struct rtnl_link_stats)); case IFLA_ADDRESS: case IFLA_BROADCAST: return add_rtattr(m, type, data, ETH_ALEN); default: return -ENOTSUP; } case RTM_NEWADDR: case RTM_DELADDR: case RTM_GETADDR: switch (type) { case IFA_LABEL: return add_rtattr(m, type, data, strlen(data) + 1); case IFA_ADDRESS: case IFA_LOCAL: case IFA_BROADCAST: case IFA_ANYCAST: ifa = NLMSG_DATA(m->hdr); switch (ifa->ifa_family) { case AF_INET: return add_rtattr(m, type, data, sizeof(struct in_addr)); case AF_INET6: return add_rtattr(m, type, data, sizeof(struct in6_addr)); default: return -EINVAL; } default: return -ENOTSUP; } default: return -ENOTSUP; } } static int message_read(sd_rtnl_message *m, unsigned short *type, void **data) { assert_return(m, -EINVAL); assert_return(data, -EINVAL); if (!RTA_OK(m->next_rta, m->remaining_size)) return 0; *data = RTA_DATA(m->next_rta); *type = m->next_rta->rta_type; m->next_rta = RTA_NEXT(m->next_rta, m->remaining_size); return 1; } int sd_rtnl_message_read(sd_rtnl_message *m, unsigned short *type, void **data) { uint16_t rtm_type; assert_return(m, -EINVAL); assert_return(data, -EINVAL); sd_rtnl_message_get_type(m, &rtm_type); switch (rtm_type) { case RTM_NEWLINK: case RTM_DELLINK: case RTM_GETLINK: if (!m->next_rta) { struct ifinfomsg *ifi = NLMSG_DATA(m->hdr); m->next_rta = IFLA_RTA(ifi); m->remaining_size = IFLA_PAYLOAD(m->hdr); } break;; case RTM_NEWADDR: case RTM_DELADDR: case RTM_GETADDR: if (!m->next_rta) { struct ifaddrmsg *ifa = NLMSG_DATA(m->hdr); m->next_rta = IFA_RTA(ifa); m->remaining_size = IFLA_PAYLOAD(m->hdr); } break;; default: return -ENOTSUP; } return message_read(m, type, data); } int message_get_serial(sd_rtnl_message *m) { assert(m); return m->hdr->nlmsg_seq; } int message_get_errno(sd_rtnl_message *m) { struct nlmsgerr *err; assert(m); if (m->hdr->nlmsg_type != NLMSG_ERROR) return 0; err = NLMSG_DATA(m->hdr); return err->error; } int message_seal(sd_rtnl *nl, sd_rtnl_message *m) { if (m->sealed) return -EPERM; m->hdr->nlmsg_seq = nl->serial++; m->sealed = true; return 0; } static int message_receive_need(sd_rtnl *rtnl, size_t *need) { assert_return(rtnl, -EINVAL); assert_return(need, -EINVAL); /* ioctl(rtnl->fd, FIONREAD, &need) Does not appear to work on netlink sockets. libnl uses MSG_PEEK instead. I don't know if that is worth the extra roundtrip. For now we simply use the maximum message size the kernel may use (NLMSG_GOODSIZE), and then realloc to the actual size after reading the message (hence avoiding huge memory usage in case many small messages are kept around) */ *need = page_size(); if (*need > 8192UL) *need = 8192UL; return 0; } /* returns the number of bytes sent, or a negative error code */ int socket_write_message(sd_rtnl *nl, sd_rtnl_message *m) { ssize_t k; assert_return(nl, -EINVAL); assert_return(m, -EINVAL); k = sendto(nl->fd, m->hdr, m->hdr->nlmsg_len, 0, &nl->sockaddr.sa, sizeof(nl->sockaddr)); if (k < 0) return (errno == EAGAIN) ? 0 : -errno; return k; } /* On success, the number of bytes received is returned and *ret points to the received message * which has a valid header and the correct size. * If nothing useful was received 0 is returned. * On failure, a negative error code is returned. */ int socket_read_message(sd_rtnl *nl, sd_rtnl_message **ret) { sd_rtnl_message *m; socklen_t addr_len = sizeof(nl->sockaddr); int r; ssize_t k; size_t need; assert_return(nl, -EINVAL); assert_return(ret, -EINVAL); r = message_receive_need(nl, &need); if (r < 0) return r; r = message_new(&m, need); if (r < 0) return r; k = recvfrom(nl->fd, m->hdr, need, 0, &nl->sockaddr.sa, &addr_len); if (k < 0) k = (errno == EAGAIN) ? 0 : -errno; /* no data */ else if (k == 0) k = -ECONNRESET; /* connection was closed by the kernel */ else if (addr_len != sizeof(nl->sockaddr.nl) || nl->sockaddr.nl.nl_family != AF_NETLINK) k = -EIO; /* not a netlink message */ else if (nl->sockaddr.nl.nl_pid != 0) k = 0; /* not from the kernel */ else if ((size_t) k < sizeof(struct nlmsghdr) || (size_t) k < m->hdr->nlmsg_len) k = -EIO; /* too small (we do accept too big though) */ else if ((pid_t) m->hdr->nlmsg_pid != getpid()) k = 0; /* not for us */ if (k > 0) switch (m->hdr->nlmsg_type) { /* check that the size matches the message type */ case NLMSG_ERROR: if (m->hdr->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr))) k = -EIO; break;; case RTM_NEWLINK: case RTM_DELLINK: case RTM_GETLINK: if (m->hdr->nlmsg_len < NLMSG_LENGTH(sizeof(struct ifinfomsg))) k = -EIO; break;; case RTM_NEWADDR: case RTM_DELADDR: case RTM_GETADDR: if (m->hdr->nlmsg_len < NLMSG_LENGTH(sizeof(struct ifaddrmsg))) k = -EIO; break;; case NLMSG_NOOP: k = 0; break;; default: k = 0; /* ignoring message of unknown type */ } if (k <= 0) sd_rtnl_message_unref(m); else { /* we probably allocated way too much memory, give it back */ m->hdr = realloc(m->hdr, m->hdr->nlmsg_len); *ret = m; } return k; }