X-Git-Url: https://git.archive.openwrt.org/?p=project%2Fnetifd.git;a=blobdiff_plain;f=system-linux.c;h=7cec649fdae76f5dff94d1695d65454401956281;hp=76740f7d6078c64c94cf58b842e3579c320b9a3f;hb=50640800da52239eb6bff87a32fb5253c3ef5824;hpb=97e748de49c02f36a1e75bd9f9552311fb4051b4 diff --git a/system-linux.c b/system-linux.c index 76740f7..7cec649 100644 --- a/system-linux.c +++ b/system-linux.c @@ -29,6 +29,7 @@ #include #include #include +#include #include #include #include @@ -36,7 +37,10 @@ #include #include -#include +#ifndef RTN_FAILED_POLICY +#define RTN_FAILED_POLICY 12 +#endif + #include #include #include @@ -54,7 +58,7 @@ struct event_socket { struct uloop_fd uloop; struct nl_sock *sock; - struct nl_cb *cb; + int bufsize; }; static int sock_ioctl = -1; @@ -65,13 +69,42 @@ static void handle_hotplug_event(struct uloop_fd *u, unsigned int events); static char dev_buf[256]; -static bool iprules_flushed = false; - static void handler_nl_event(struct uloop_fd *u, unsigned int events) { struct event_socket *ev = container_of(u, struct event_socket, uloop); - nl_recvmsgs(ev->sock, ev->cb); + int err; + socklen_t errlen = sizeof(err); + + if (!u->error) { + nl_recvmsgs_default(ev->sock); + return; + } + + if (getsockopt(u->fd, SOL_SOCKET, SO_ERROR, (void *)&err, &errlen)) + goto abort; + + switch(err) { + case ENOBUFS: + // Increase rx buffer size on netlink socket + ev->bufsize *= 2; + if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0)) + goto abort; + + // Request full dump since some info got dropped + struct rtgenmsg msg = { .rtgen_family = AF_UNSPEC }; + nl_send_simple(ev->sock, RTM_GETLINK, NLM_F_DUMP, &msg, sizeof(msg)); + break; + + default: + goto abort; + } + u->error = false; + return; + +abort: + uloop_fd_delete(&ev->uloop); + return; } static struct nl_sock * @@ -94,7 +127,7 @@ create_socket(int protocol, int groups) static bool create_raw_event_socket(struct event_socket *ev, int protocol, int groups, - uloop_fd_handler cb) + uloop_fd_handler cb, int flags) { ev->sock = create_socket(protocol, groups); if (!ev->sock) @@ -102,7 +135,9 @@ create_raw_event_socket(struct event_socket *ev, int protocol, int groups, ev->uloop.fd = nl_socket_get_fd(ev->sock); ev->uloop.cb = cb; - uloop_fd_add(&ev->uloop, ULOOP_READ | ULOOP_EDGE_TRIGGER); + if (uloop_fd_add(&ev->uloop, ULOOP_READ|flags)) + return false; + return true; } @@ -110,14 +145,21 @@ static bool create_event_socket(struct event_socket *ev, int protocol, int (*cb)(struct nl_msg *msg, void *arg)) { - // Prepare socket for link events - ev->cb = nl_cb_alloc(NL_CB_DEFAULT); - if (!ev->cb) + if (!create_raw_event_socket(ev, protocol, 0, handler_nl_event, ULOOP_ERROR_CB)) return false; - nl_cb_set(ev->cb, NL_CB_VALID, NL_CB_CUSTOM, cb, NULL); + // Install the valid custom callback handler + nl_socket_modify_cb(ev->sock, NL_CB_VALID, NL_CB_CUSTOM, cb, NULL); + + // Disable sequence number checking on event sockets + nl_socket_disable_seq_check(ev->sock); - return create_raw_event_socket(ev, protocol, 0, handler_nl_event); + // Increase rx buffer size to 65K on event sockets + ev->bufsize = 65535; + if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0)) + return false; + + return true; } int system_init(void) @@ -126,7 +168,7 @@ int system_init(void) static struct event_socket hotplug_event; sock_ioctl = socket(AF_LOCAL, SOCK_DGRAM, 0); - fcntl(sock_ioctl, F_SETFD, fcntl(sock_ioctl, F_GETFD) | FD_CLOEXEC); + system_fd_set_cloexec(sock_ioctl); // Prepare socket for routing / address control sock_rtnl = create_socket(NETLINK_ROUTE, 0); @@ -137,7 +179,7 @@ int system_init(void) return -1; if (!create_raw_event_socket(&hotplug_event, NETLINK_KOBJECT_UEVENT, 1, - handle_hotplug_event)) + handle_hotplug_event, 0)) return -1; // Receive network link events form kernel @@ -154,7 +196,7 @@ static void system_set_sysctl(const char *path, const char *val) if (fd < 0) return; - write(fd, val, strlen(val)); + if (write(fd, val, strlen(val))) {} close(fd); } @@ -169,6 +211,10 @@ static void system_set_disable_ipv6(struct device *dev, const char *val) system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6", dev->ifname, val); } +#ifndef IFF_LOWER_UP +#define IFF_LOWER_UP 0x10000 +#endif + // Evaluate netlink messages static int cb_rtnl_event(struct nl_msg *msg, void *arg) { @@ -176,19 +222,19 @@ static int cb_rtnl_event(struct nl_msg *msg, void *arg) struct ifinfomsg *ifi = NLMSG_DATA(nh); struct nlattr *nla[__IFLA_MAX]; - if (nh->nlmsg_type != RTM_DELLINK && nh->nlmsg_type != RTM_NEWLINK) + if (nh->nlmsg_type != RTM_NEWLINK) goto out; nlmsg_parse(nh, sizeof(*ifi), nla, __IFLA_MAX - 1, NULL); if (!nla[IFLA_IFNAME]) goto out; - struct device *dev = device_get(RTA_DATA(nla[IFLA_IFNAME]), false); + struct device *dev = device_get(nla_data(nla[IFLA_IFNAME]), false); if (!dev) goto out; - dev->ifindex = ifi->ifi_index; - /* TODO: parse link status */ + device_set_ifindex(dev, ifi->ifi_index); + device_set_link(dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false); out: return 0; @@ -454,8 +500,9 @@ static int cb_clear_event(struct nl_msg *msg, void *arg) hdr->nlmsg_type = type; hdr->nlmsg_flags = NLM_F_REQUEST; - if (!nl_send_auto_complete(sock_rtnl, clr->msg)) - nl_wait_for_ack(sock_rtnl); + nl_socket_disable_auto_ack(sock_rtnl); + nl_send_auto_complete(sock_rtnl, clr->msg); + nl_socket_enable_auto_ack(sock_rtnl); return NL_SKIP; } @@ -535,7 +582,7 @@ void system_if_clear_state(struct device *dev) if (dev->external) return; - dev->ifindex = system_if_resolve(dev); + device_set_ifindex(dev, system_if_resolve(dev)); if (!dev->ifindex) return; @@ -609,6 +656,93 @@ int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg) return 0; } +int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg) +{ + struct nl_msg *msg; + struct nlattr *linkinfo, *data; + struct ifinfomsg iim = { .ifi_family = AF_INET }; + int ifindex = system_if_resolve(dev); + int i, rv; + static const struct { + const char *name; + enum macvlan_mode val; + } modes[] = { + { "private", MACVLAN_MODE_PRIVATE }, + { "vepa", MACVLAN_MODE_VEPA }, + { "bridge", MACVLAN_MODE_BRIDGE }, + { "passthru", MACVLAN_MODE_PASSTHRU }, + }; + + if (ifindex == 0) + return -ENOENT; + + msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL); + + if (!msg) + return -1; + + nlmsg_append(msg, &iim, sizeof(iim), 0); + + if (cfg->flags & MACVLAN_OPT_MACADDR) + nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr); + nla_put(msg, IFLA_IFNAME, IFNAMSIZ, macvlan->ifname); + nla_put_u32(msg, IFLA_LINK, ifindex); + + if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO))) + goto nla_put_failure; + + nla_put(msg, IFLA_INFO_KIND, strlen("macvlan"), "macvlan"); + + if (!(data = nla_nest_start(msg, IFLA_INFO_DATA))) + goto nla_put_failure; + + if (cfg->mode) { + for (i = 0; i < ARRAY_SIZE(modes); i++) { + if (strcmp(cfg->mode, modes[i].name) != 0) + continue; + + nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val); + break; + } + } + + nla_nest_end(msg, data); + nla_nest_end(msg, linkinfo); + + rv = system_rtnl_call(msg); + if (rv) + D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv); + + return rv; + +nla_put_failure: + nlmsg_free(msg); + return -ENOMEM; +} + +int system_macvlan_del(struct device *macvlan) +{ + struct nl_msg *msg; + struct ifinfomsg iim; + + iim.ifi_family = AF_INET; + iim.ifi_index = 0; + + msg = nlmsg_alloc_simple(RTM_DELLINK, 0); + + if (!msg) + return -1; + + nlmsg_append(msg, &iim, sizeof(iim), 0); + + nla_put(msg, IFLA_INFO_KIND, strlen("macvlan"), "macvlan"); + nla_put(msg, IFLA_IFNAME, sizeof(macvlan->ifname), macvlan->ifname); + + system_rtnl_call(msg); + + return 0; +} + static int system_vlan(struct device *dev, int id) { struct vlan_ioctl_args ifr = { @@ -664,23 +798,26 @@ system_if_get_settings(struct device *dev, struct device_settings *s) } void -system_if_apply_settings(struct device *dev, struct device_settings *s) +system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask) { struct ifreq ifr; + if (!apply_mask) + return; + memset(&ifr, 0, sizeof(ifr)); strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name)); - if (s->flags & DEV_OPT_MTU) { + if (s->flags & DEV_OPT_MTU & apply_mask) { ifr.ifr_mtu = s->mtu; if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0) s->flags &= ~DEV_OPT_MTU; } - if (s->flags & DEV_OPT_TXQUEUELEN) { + if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) { ifr.ifr_qlen = s->txqueuelen; if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0) s->flags &= ~DEV_OPT_TXQUEUELEN; } - if ((s->flags & DEV_OPT_MACADDR) && !dev->external) { + if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) { ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER; memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr)); if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0) @@ -691,8 +828,8 @@ system_if_apply_settings(struct device *dev, struct device_settings *s) int system_if_up(struct device *dev) { system_if_get_settings(dev, &dev->orig_settings); - system_if_apply_settings(dev, &dev->settings); - dev->ifindex = system_if_resolve(dev); + system_if_apply_settings(dev, &dev->settings, dev->settings.flags); + device_set_ifindex(dev, system_if_resolve(dev)); return system_if_flags(dev->ifname, IFF_UP, 0); } @@ -700,14 +837,82 @@ int system_if_down(struct device *dev) { int ret = system_if_flags(dev->ifname, 0, IFF_UP); dev->orig_settings.flags &= dev->settings.flags; - system_if_apply_settings(dev, &dev->orig_settings); + system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags); return ret; } +struct if_check_data { + struct device *dev; + int pending; + int ret; +}; + +static int cb_if_check_valid(struct nl_msg *msg, void *arg) +{ + struct nlmsghdr *nh = nlmsg_hdr(msg); + struct ifinfomsg *ifi = NLMSG_DATA(nh); + struct if_check_data *chk = (struct if_check_data *)arg; + + if (nh->nlmsg_type != RTM_NEWLINK) + return NL_SKIP; + + device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false); + device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false); + + return NL_OK; +} + +static int cb_if_check_ack(struct nl_msg *msg, void *arg) +{ + struct if_check_data *chk = (struct if_check_data *)arg; + chk->pending = 0; + return NL_STOP; +} + +static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg) +{ + struct if_check_data *chk = (struct if_check_data *)arg; + + device_set_present(chk->dev, false); + device_set_link(chk->dev, false); + chk->pending = err->error; + + return NL_STOP; +} + int system_if_check(struct device *dev) { - device_set_present(dev, (system_if_resolve(dev) > 0)); - return 0; + struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT); + struct nl_msg *msg; + struct ifinfomsg ifi = { + .ifi_family = AF_UNSPEC, + .ifi_index = 0, + }; + struct if_check_data chk = { + .dev = dev, + .pending = 1, + }; + int ret = 1; + + msg = nlmsg_alloc_simple(RTM_GETLINK, 0); + if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) || + nla_put_string(msg, IFLA_IFNAME, dev->ifname)) + goto out; + + nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk); + nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk); + nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk); + + nl_send_auto_complete(sock_rtnl, msg); + while (chk.pending > 0) + nl_recvmsgs(sock_rtnl, cb); + + nlmsg_free(msg); + ret = chk.pending; + +out: + nl_cb_put(cb); + return ret; } struct device * @@ -825,14 +1030,10 @@ system_if_dump_info(struct device *dev, struct blob_buf *b) char buf[64], *s; void *c; int dir_fd; - uint64_t val = 0; snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname); dir_fd = open(buf, O_DIRECTORY); - if (read_uint64_file(dir_fd, "carrier", &val)) - blobmsg_add_u8(b, "link", !!val); - memset(&ecmd, 0, sizeof(ecmd)); memset(&ifr, 0, sizeof(ifr)); strcpy(ifr.ifr_name, dev->ifname); @@ -920,17 +1121,21 @@ static int system_addr(struct device *dev, struct device_addr *addr, int cmd) struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0}; if (addr->preferred_until) { - int preferred = addr->preferred_until - now; + int64_t preferred = addr->preferred_until - now; if (preferred < 0) preferred = 0; + else if (preferred > UINT32_MAX) + preferred = UINT32_MAX; cinfo.ifa_prefered = preferred; } if (addr->valid_until) { - int valid = addr->valid_until - now; + int64_t valid = addr->valid_until - now; if (valid <= 0) return -1; + else if (valid > UINT32_MAX) + valid = UINT32_MAX; cinfo.ifa_valid = valid; } @@ -974,6 +1179,7 @@ static int system_rt(struct device *dev, struct device_route *route, int cmd) struct rtmsg rtm = { .rtm_family = (alen == 4) ? AF_INET : AF_INET6, .rtm_dst_len = route->mask, + .rtm_src_len = route->sourcemask, .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC, .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC, .rtm_scope = scope, @@ -999,6 +1205,9 @@ static int system_rt(struct device *dev, struct device_route *route, int cmd) if (route->mask) nla_put(msg, RTA_DST, alen, &route->addr); + if (route->sourcemask) + nla_put(msg, RTA_SRC, alen, &route->source); + if (route->metric > 0) nla_put_u32(msg, RTA_PRIORITY, route->metric); @@ -1037,7 +1246,7 @@ int system_flush_routes(void) if (fd < 0) continue; - write(fd, "-1", 2); + if (write(fd, "-1", 2)) {} close(fd); } return 0; @@ -1179,13 +1388,6 @@ static int system_iprule(struct iprule *rule, int cmd) int system_add_iprule(struct iprule *rule) { - /* trigger flush of existing rules when adding first rule the first time */ - if (!iprules_flushed) - { - system_flush_iprules(); - iprules_flushed = true; - } - return system_iprule(rule, RTM_NEWRULE); } @@ -1260,6 +1462,8 @@ bool system_resolve_iprule_action(const char *action, unsigned int *id) n = RTN_UNICAST; else if (!strcmp(action, "throw")) n = RTN_THROW; + else if (!strcmp(action, "failed_policy")) + n = RTN_FAILED_POLICY; else { n = strtoul(action, &e, 0); if (!e || *e || e == action || n > 255) @@ -1330,11 +1534,11 @@ out: return ret; } - int system_add_ip_tunnel(const char *name, struct blob_attr *attr) { struct blob_attr *tb[__TUNNEL_ATTR_MAX]; struct blob_attr *cur; + bool set_df = true; const char *str; system_del_ip_tunnel(name); @@ -1346,9 +1550,15 @@ int system_add_ip_tunnel(const char *name, struct blob_attr *attr) return -EINVAL; str = blobmsg_data(cur); + if ((cur = tb[TUNNEL_ATTR_DF])) + set_df = blobmsg_get_bool(cur); + unsigned int ttl = 0; - if ((cur = tb[TUNNEL_ATTR_TTL]) && (ttl = blobmsg_get_u32(cur)) > 255) - return -EINVAL; + if ((cur = tb[TUNNEL_ATTR_TTL])) { + ttl = blobmsg_get_u32(cur); + if (ttl > 255 || (!set_df && ttl)) + return -EINVAL; + } unsigned int link = 0; if ((cur = tb[TUNNEL_ATTR_LINK])) { @@ -1360,14 +1570,13 @@ int system_add_ip_tunnel(const char *name, struct blob_attr *attr) link = iface->l3_dev.dev->ifindex; } - if (!strcmp(str, "sit")) { struct ip_tunnel_parm p = { .link = link, .iph = { .version = 4, .ihl = 5, - .frag_off = htons(IP_DF), + .frag_off = set_df ? htons(IP_DF) : 0, .protocol = IPPROTO_IPV6, .ttl = ttl }