From b750101eb236130cf056c675997decbac904cc49 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Sun, 7 Apr 2024 17:35:18 +0200 Subject: Adding upstream version 252.22. Signed-off-by: Daniel Baumann --- src/network/networkd-manager.c | 1075 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 1075 insertions(+) create mode 100644 src/network/networkd-manager.c (limited to 'src/network/networkd-manager.c') diff --git a/src/network/networkd-manager.c b/src/network/networkd-manager.c new file mode 100644 index 0000000..79ef6d4 --- /dev/null +++ b/src/network/networkd-manager.c @@ -0,0 +1,1075 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include +#include +#include +#include +#include +#include +#include + +#include "sd-netlink.h" + +#include "alloc-util.h" +#include "bus-error.h" +#include "bus-locator.h" +#include "bus-log-control-api.h" +#include "bus-polkit.h" +#include "bus-util.h" +#include "conf-parser.h" +#include "daemon-util.h" +#include "def.h" +#include "device-private.h" +#include "device-util.h" +#include "dns-domain.h" +#include "fd-util.h" +#include "fileio.h" +#include "firewall-util.h" +#include "fs-util.h" +#include "local-addresses.h" +#include "netlink-util.h" +#include "network-internal.h" +#include "networkd-address-pool.h" +#include "networkd-address.h" +#include "networkd-dhcp-server-bus.h" +#include "networkd-dhcp6.h" +#include "networkd-link-bus.h" +#include "networkd-manager-bus.h" +#include "networkd-manager.h" +#include "networkd-neighbor.h" +#include "networkd-network-bus.h" +#include "networkd-nexthop.h" +#include "networkd-queue.h" +#include "networkd-route.h" +#include "networkd-routing-policy-rule.h" +#include "networkd-speed-meter.h" +#include "networkd-state-file.h" +#include "networkd-wifi.h" +#include "networkd-wiphy.h" +#include "ordered-set.h" +#include "path-lookup.h" +#include "path-util.h" +#include "qdisc.h" +#include "selinux-util.h" +#include "set.h" +#include "signal-util.h" +#include "stat-util.h" +#include "strv.h" +#include "sysctl-util.h" +#include "tclass.h" +#include "tmpfile-util.h" +#include "tuntap.h" +#include "udev-util.h" + +/* use 128 MB for receive socket kernel queue. */ +#define RCVBUF_SIZE (128*1024*1024) + +static int manager_reset_all(Manager *m) { + Link *link; + int r; + + assert(m); + + HASHMAP_FOREACH(link, m->links_by_index) { + r = link_reconfigure_after_sleep(link); + if (r < 0) { + log_link_warning_errno(link, r, "Failed to reconfigure interface: %m"); + link_enter_failed(link); + } + } + + return 0; +} + +static int match_prepare_for_sleep(sd_bus_message *message, void *userdata, sd_bus_error *ret_error) { + Manager *m = ASSERT_PTR(userdata); + int b, r; + + assert(message); + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) { + bus_log_parse_error(r); + return 0; + } + + if (b) + return 0; + + log_debug("Coming back from suspend, resetting all connections..."); + + (void) manager_reset_all(m); + + return 0; +} + +static int on_connected(sd_bus_message *message, void *userdata, sd_bus_error *ret_error) { + Manager *m = ASSERT_PTR(userdata); + + assert(message); + + /* Did we get a timezone or transient hostname from DHCP while D-Bus wasn't up yet? */ + if (m->dynamic_hostname) + (void) manager_set_hostname(m, m->dynamic_hostname); + if (m->dynamic_timezone) + (void) manager_set_timezone(m, m->dynamic_timezone); + if (m->product_uuid_requested) + (void) manager_request_product_uuid(m); + + return 0; +} + +static int manager_connect_bus(Manager *m) { + int r; + + assert(m); + assert(!m->bus); + + r = bus_open_system_watch_bind_with_description(&m->bus, "bus-api-network"); + if (r < 0) + return log_error_errno(r, "Failed to connect to bus: %m"); + + r = bus_add_implementation(m->bus, &manager_object, m); + if (r < 0) + return r; + + r = bus_log_control_api_register(m->bus); + if (r < 0) + return r; + + r = sd_bus_request_name_async(m->bus, NULL, "org.freedesktop.network1", 0, NULL, NULL); + if (r < 0) + return log_error_errno(r, "Failed to request name: %m"); + + r = sd_bus_attach_event(m->bus, m->event, 0); + if (r < 0) + return log_error_errno(r, "Failed to attach bus to event loop: %m"); + + r = sd_bus_match_signal_async( + m->bus, + NULL, + "org.freedesktop.DBus.Local", + NULL, + "org.freedesktop.DBus.Local", + "Connected", + on_connected, NULL, m); + if (r < 0) + return log_error_errno(r, "Failed to request match on Connected signal: %m"); + + r = sd_bus_match_signal_async( + m->bus, + NULL, + "org.freedesktop.login1", + "/org/freedesktop/login1", + "org.freedesktop.login1.Manager", + "PrepareForSleep", + match_prepare_for_sleep, NULL, m); + if (r < 0) + log_warning_errno(r, "Failed to request match for PrepareForSleep, ignoring: %m"); + + return 0; +} + +static int manager_process_uevent(sd_device_monitor *monitor, sd_device *device, void *userdata) { + Manager *m = ASSERT_PTR(userdata); + sd_device_action_t action; + const char *s; + int r; + + assert(device); + + r = sd_device_get_action(device, &action); + if (r < 0) + return log_device_warning_errno(device, r, "Failed to get udev action, ignoring: %m"); + + r = sd_device_get_subsystem(device, &s); + if (r < 0) + return log_device_warning_errno(device, r, "Failed to get subsystem, ignoring: %m"); + + if (streq(s, "net")) + r = manager_udev_process_link(m, device, action); + else if (streq(s, "ieee80211")) + r = manager_udev_process_wiphy(m, device, action); + else if (streq(s, "rfkill")) + r = manager_udev_process_rfkill(m, device, action); + else { + log_device_debug(device, "Received device with unexpected subsystem \"%s\", ignoring.", s); + return 0; + } + if (r < 0) + log_device_warning_errno(device, r, "Failed to process \"%s\" uevent, ignoring: %m", + device_action_to_string(action)); + + return 0; +} + +static int manager_connect_udev(Manager *m) { + int r; + + /* udev does not initialize devices inside containers, so we rely on them being already + * initialized before entering the container. */ + if (!udev_available()) + return 0; + + r = sd_device_monitor_new(&m->device_monitor); + if (r < 0) + return log_error_errno(r, "Failed to initialize device monitor: %m"); + + r = sd_device_monitor_filter_add_match_subsystem_devtype(m->device_monitor, "net", NULL); + if (r < 0) + return log_error_errno(r, "Could not add device monitor filter for net subsystem: %m"); + + r = sd_device_monitor_filter_add_match_subsystem_devtype(m->device_monitor, "ieee80211", NULL); + if (r < 0) + return log_error_errno(r, "Could not add device monitor filter for ieee80211 subsystem: %m"); + + r = sd_device_monitor_filter_add_match_subsystem_devtype(m->device_monitor, "rfkill", NULL); + if (r < 0) + return log_error_errno(r, "Could not add device monitor filter for rfkill subsystem: %m"); + + r = sd_device_monitor_attach_event(m->device_monitor, m->event); + if (r < 0) + return log_error_errno(r, "Failed to attach event to device monitor: %m"); + + r = sd_device_monitor_start(m->device_monitor, manager_process_uevent, m); + if (r < 0) + return log_error_errno(r, "Failed to start device monitor: %m"); + + return 0; +} + +static int manager_listen_fds(Manager *m, int *ret_rtnl_fd) { + _cleanup_strv_free_ char **names = NULL; + int n, rtnl_fd = -1; + + assert(m); + assert(ret_rtnl_fd); + + n = sd_listen_fds_with_names(/* unset_environment = */ true, &names); + if (n < 0) + return n; + + if (strv_length(names) != (size_t) n) + return -EINVAL; + + for (int i = 0; i < n; i++) { + int fd = i + SD_LISTEN_FDS_START; + + if (sd_is_socket(fd, AF_NETLINK, SOCK_RAW, -1) > 0) { + if (rtnl_fd >= 0) { + log_debug("Received multiple netlink socket, ignoring."); + safe_close(fd); + continue; + } + + rtnl_fd = fd; + continue; + } + + if (manager_add_tuntap_fd(m, fd, names[i]) >= 0) + continue; + + if (m->test_mode) + safe_close(fd); + else + close_and_notify_warn(fd, names[i]); + } + + *ret_rtnl_fd = rtnl_fd; + return 0; +} + +static int manager_connect_genl(Manager *m) { + int r; + + assert(m); + + r = sd_genl_socket_open(&m->genl); + if (r < 0) + return r; + + r = sd_netlink_increase_rxbuf(m->genl, RCVBUF_SIZE); + if (r < 0) + log_warning_errno(r, "Failed to increase receive buffer size for general netlink socket, ignoring: %m"); + + r = sd_netlink_attach_event(m->genl, m->event, 0); + if (r < 0) + return r; + + r = genl_add_match(m->genl, NULL, NL80211_GENL_NAME, NL80211_MULTICAST_GROUP_CONFIG, 0, + &manager_genl_process_nl80211_config, NULL, m, "network-genl_process_nl80211_config"); + if (r < 0 && r != -EOPNOTSUPP) + return r; + + r = genl_add_match(m->genl, NULL, NL80211_GENL_NAME, NL80211_MULTICAST_GROUP_MLME, 0, + &manager_genl_process_nl80211_mlme, NULL, m, "network-genl_process_nl80211_mlme"); + if (r < 0 && r != -EOPNOTSUPP) + return r; + + return 0; +} + +static int manager_setup_rtnl_filter(Manager *manager) { + struct sock_filter filter[] = { + /* Check the packet length. */ + BPF_STMT(BPF_LD + BPF_W + BPF_LEN, 0), /* A <- packet length */ + BPF_JUMP(BPF_JMP + BPF_JGE + BPF_K, sizeof(struct nlmsghdr), 1, 0), /* A (packet length) >= sizeof(struct nlmsghdr) ? */ + BPF_STMT(BPF_RET + BPF_K, 0), /* reject */ + /* Always accept multipart message. */ + BPF_STMT(BPF_LD + BPF_H + BPF_ABS, offsetof(struct nlmsghdr, nlmsg_flags)), /* A <- message flags */ + BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, htobe16(NLM_F_MULTI), 0, 1), /* message flags has NLM_F_MULTI ? */ + BPF_STMT(BPF_RET + BPF_K, UINT32_MAX), /* accept */ + /* Accept all message types except for RTM_NEWNEIGH or RTM_DELNEIGH. */ + BPF_STMT(BPF_LD + BPF_H + BPF_ABS, offsetof(struct nlmsghdr, nlmsg_type)), /* A <- message type */ + BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, htobe16(RTM_NEWNEIGH), 2, 0), /* message type == RTM_NEWNEIGH ? */ + BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, htobe16(RTM_DELNEIGH), 1, 0), /* message type == RTM_DELNEIGH ? */ + BPF_STMT(BPF_RET + BPF_K, UINT32_MAX), /* accept */ + /* Check the packet length. */ + BPF_STMT(BPF_LD + BPF_W + BPF_LEN, 0), /* A <- packet length */ + BPF_JUMP(BPF_JMP + BPF_JGE + BPF_K, sizeof(struct nlmsghdr) + sizeof(struct ndmsg), 1, 0), + /* packet length >= sizeof(struct nlmsghdr) + sizeof(struct ndmsg) ? */ + BPF_STMT(BPF_RET + BPF_K, 0), /* reject */ + /* Reject the message when the neighbor state does not have NUD_PERMANENT flag. */ + BPF_STMT(BPF_LD + BPF_H + BPF_ABS, sizeof(struct nlmsghdr) + offsetof(struct ndmsg, ndm_state)), + /* A <- neighbor state */ + BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, htobe16(NUD_PERMANENT), 1, 0), /* neighbor state has NUD_PERMANENT ? */ + BPF_STMT(BPF_RET + BPF_K, 0), /* reject */ + BPF_STMT(BPF_RET + BPF_K, UINT32_MAX), /* accept */ + }; + + assert(manager); + assert(manager->rtnl); + + return sd_netlink_attach_filter(manager->rtnl, ELEMENTSOF(filter), filter); +} + +static int manager_connect_rtnl(Manager *m, int fd) { + _unused_ _cleanup_close_ int fd_close = fd; + int r; + + assert(m); + + /* This takes input fd. */ + + if (fd < 0) + r = sd_netlink_open(&m->rtnl); + else + r = sd_netlink_open_fd(&m->rtnl, fd); + if (r < 0) + return r; + TAKE_FD(fd_close); + + /* Bump receiver buffer, but only if we are not called via socket activation, as in that + * case systemd sets the receive buffer size for us, and the value in the .socket unit + * should take full effect. */ + if (fd < 0) { + r = sd_netlink_increase_rxbuf(m->rtnl, RCVBUF_SIZE); + if (r < 0) + log_warning_errno(r, "Failed to increase receive buffer size for rtnl socket, ignoring: %m"); + } + + r = sd_netlink_attach_event(m->rtnl, m->event, 0); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_NEWLINK, &manager_rtnl_process_link, NULL, m, "network-rtnl_process_link"); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_DELLINK, &manager_rtnl_process_link, NULL, m, "network-rtnl_process_link"); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_NEWQDISC, &manager_rtnl_process_qdisc, NULL, m, "network-rtnl_process_qdisc"); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_DELQDISC, &manager_rtnl_process_qdisc, NULL, m, "network-rtnl_process_qdisc"); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_NEWTCLASS, &manager_rtnl_process_tclass, NULL, m, "network-rtnl_process_tclass"); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_DELTCLASS, &manager_rtnl_process_tclass, NULL, m, "network-rtnl_process_tclass"); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_NEWADDR, &manager_rtnl_process_address, NULL, m, "network-rtnl_process_address"); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_DELADDR, &manager_rtnl_process_address, NULL, m, "network-rtnl_process_address"); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_NEWNEIGH, &manager_rtnl_process_neighbor, NULL, m, "network-rtnl_process_neighbor"); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_DELNEIGH, &manager_rtnl_process_neighbor, NULL, m, "network-rtnl_process_neighbor"); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_NEWROUTE, &manager_rtnl_process_route, NULL, m, "network-rtnl_process_route"); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_DELROUTE, &manager_rtnl_process_route, NULL, m, "network-rtnl_process_route"); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_NEWRULE, &manager_rtnl_process_rule, NULL, m, "network-rtnl_process_rule"); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_DELRULE, &manager_rtnl_process_rule, NULL, m, "network-rtnl_process_rule"); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_NEWNEXTHOP, &manager_rtnl_process_nexthop, NULL, m, "network-rtnl_process_nexthop"); + if (r < 0) + return r; + + r = netlink_add_match(m->rtnl, NULL, RTM_DELNEXTHOP, &manager_rtnl_process_nexthop, NULL, m, "network-rtnl_process_nexthop"); + if (r < 0) + return r; + + return manager_setup_rtnl_filter(m); +} + +static int manager_dirty_handler(sd_event_source *s, void *userdata) { + Manager *m = ASSERT_PTR(userdata); + Link *link; + int r; + + if (m->dirty) { + r = manager_save(m); + if (r < 0) + log_warning_errno(r, "Failed to update state file %s, ignoring: %m", m->state_file); + } + + SET_FOREACH(link, m->dirty_links) { + r = link_save_and_clean(link); + if (r < 0) + log_link_warning_errno(link, r, "Failed to update link state file %s, ignoring: %m", link->state_file); + } + + return 1; +} + +static int signal_terminate_callback(sd_event_source *s, const struct signalfd_siginfo *si, void *userdata) { + Manager *m = ASSERT_PTR(userdata); + + m->restarting = false; + + log_debug("Terminate operation initiated."); + + return sd_event_exit(sd_event_source_get_event(s), 0); +} + +static int signal_restart_callback(sd_event_source *s, const struct signalfd_siginfo *si, void *userdata) { + Manager *m = ASSERT_PTR(userdata); + + m->restarting = true; + + log_debug("Restart operation initiated."); + + return sd_event_exit(sd_event_source_get_event(s), 0); +} + +static int manager_set_keep_configuration(Manager *m) { + int r; + + assert(m); + + if (in_initrd()) { + log_debug("Running in initrd, keep DHCPv4 addresses on stopping networkd by default."); + m->keep_configuration = KEEP_CONFIGURATION_DHCP_ON_STOP; + return 0; + } + + r = path_is_network_fs("/"); + if (r < 0) + return log_error_errno(r, "Failed to detect if root is network filesystem: %m"); + if (r == 0) { + m->keep_configuration = _KEEP_CONFIGURATION_INVALID; + return 0; + } + + log_debug("Running on network filesystem, enabling KeepConfiguration= by default."); + m->keep_configuration = KEEP_CONFIGURATION_YES; + return 0; +} + +int manager_setup(Manager *m) { + _cleanup_close_ int rtnl_fd = -1; + int r; + + assert(m); + + r = sd_event_default(&m->event); + if (r < 0) + return r; + + assert_se(sigprocmask_many(SIG_SETMASK, NULL, SIGINT, SIGTERM, SIGUSR2, -1) >= 0); + + (void) sd_event_set_watchdog(m->event, true); + (void) sd_event_add_signal(m->event, NULL, SIGTERM, signal_terminate_callback, m); + (void) sd_event_add_signal(m->event, NULL, SIGINT, signal_terminate_callback, m); + (void) sd_event_add_signal(m->event, NULL, SIGUSR2, signal_restart_callback, m); + + r = sd_event_add_post(m->event, NULL, manager_dirty_handler, m); + if (r < 0) + return r; + + r = sd_event_add_post(m->event, NULL, manager_process_requests, m); + if (r < 0) + return r; + + r = manager_listen_fds(m, &rtnl_fd); + if (r < 0) + return r; + + r = manager_connect_rtnl(m, TAKE_FD(rtnl_fd)); + if (r < 0) + return r; + + r = manager_connect_genl(m); + if (r < 0) + return r; + + if (m->test_mode) + return 0; + + r = manager_connect_bus(m); + if (r < 0) + return r; + + r = manager_connect_udev(m); + if (r < 0) + return r; + + r = sd_resolve_default(&m->resolve); + if (r < 0) + return r; + + r = sd_resolve_attach_event(m->resolve, m->event, 0); + if (r < 0) + return r; + + r = address_pool_setup_default(m); + if (r < 0) + return r; + + r = manager_set_keep_configuration(m); + if (r < 0) + return r; + + m->state_file = strdup("/run/systemd/netif/state"); + if (!m->state_file) + return -ENOMEM; + + return 0; +} + +int manager_new(Manager **ret, bool test_mode) { + _cleanup_(manager_freep) Manager *m = NULL; + + m = new(Manager, 1); + if (!m) + return -ENOMEM; + + *m = (Manager) { + .keep_configuration = _KEEP_CONFIGURATION_INVALID, + .test_mode = test_mode, + .speed_meter_interval_usec = SPEED_METER_DEFAULT_TIME_INTERVAL, + .online_state = _LINK_ONLINE_STATE_INVALID, + .manage_foreign_routes = true, + .manage_foreign_rules = true, + .ethtool_fd = -1, + .dhcp_duid.type = DUID_TYPE_EN, + .dhcp6_duid.type = DUID_TYPE_EN, + .duid_product_uuid.type = DUID_TYPE_UUID, + }; + + *ret = TAKE_PTR(m); + return 0; +} + +Manager* manager_free(Manager *m) { + Link *link; + + if (!m) + return NULL; + + free(m->state_file); + + HASHMAP_FOREACH(link, m->links_by_index) + (void) link_stop_engines(link, true); + + m->request_queue = ordered_set_free(m->request_queue); + + m->dirty_links = set_free_with_destructor(m->dirty_links, link_unref); + m->links_by_name = hashmap_free(m->links_by_name); + m->links_by_hw_addr = hashmap_free(m->links_by_hw_addr); + m->links_by_dhcp_pd_subnet_prefix = hashmap_free(m->links_by_dhcp_pd_subnet_prefix); + m->links_by_index = hashmap_free_with_destructor(m->links_by_index, link_unref); + + m->dhcp_pd_subnet_ids = set_free(m->dhcp_pd_subnet_ids); + m->networks = ordered_hashmap_free_with_destructor(m->networks, network_unref); + + m->netdevs = hashmap_free_with_destructor(m->netdevs, netdev_unref); + + m->tuntap_fds_by_name = hashmap_free(m->tuntap_fds_by_name); + + m->wiphy_by_name = hashmap_free(m->wiphy_by_name); + m->wiphy_by_index = hashmap_free_with_destructor(m->wiphy_by_index, wiphy_free); + + ordered_set_free_free(m->address_pools); + + hashmap_free(m->route_table_names_by_number); + hashmap_free(m->route_table_numbers_by_name); + + set_free(m->rules); + + sd_netlink_unref(m->rtnl); + sd_netlink_unref(m->genl); + sd_resolve_unref(m->resolve); + + /* reject (e.g. unreachable) type routes are managed by Manager, but may be referenced by a + * link. E.g., DHCP6 with prefix delegation creates unreachable routes, and they are referenced + * by the upstream link. And the links may be referenced by netlink slots. Hence, two + * set_free() must be called after the above sd_netlink_unref(). */ + m->routes = set_free(m->routes); + + m->nexthops = set_free(m->nexthops); + m->nexthops_by_id = hashmap_free(m->nexthops_by_id); + + sd_event_source_unref(m->speed_meter_event_source); + sd_event_unref(m->event); + + sd_device_monitor_unref(m->device_monitor); + + bus_verify_polkit_async_registry_free(m->polkit_registry); + sd_bus_flush_close_unref(m->bus); + + free(m->dynamic_timezone); + free(m->dynamic_hostname); + + safe_close(m->ethtool_fd); + + m->fw_ctx = fw_ctx_free(m->fw_ctx); + + return mfree(m); +} + +int manager_start(Manager *m) { + Link *link; + int r; + + assert(m); + + r = manager_start_speed_meter(m); + if (r < 0) + return log_error_errno(r, "Failed to initialize speed meter: %m"); + + /* The dirty handler will deal with future serialization, but the first one + must be done explicitly. */ + + r = manager_save(m); + if (r < 0) + log_warning_errno(r, "Failed to update state file %s, ignoring: %m", m->state_file); + + HASHMAP_FOREACH(link, m->links_by_index) { + r = link_save(link); + if (r < 0) + log_link_warning_errno(link, r, "Failed to update link state file %s, ignoring: %m", link->state_file); + } + + return 0; +} + +int manager_load_config(Manager *m) { + int r; + + r = netdev_load(m, false); + if (r < 0) + return r; + + manager_clear_unmanaged_tuntap_fds(m); + + r = network_load(m, &m->networks); + if (r < 0) + return r; + + return manager_build_dhcp_pd_subnet_ids(m); +} + +static int manager_enumerate_internal( + Manager *m, + sd_netlink *nl, + sd_netlink_message *req, + int (*process)(sd_netlink *, sd_netlink_message *, Manager *)) { + + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *reply = NULL; + int k, r; + + assert(m); + assert(nl); + assert(req); + assert(process); + + r = sd_netlink_message_set_request_dump(req, true); + if (r < 0) + return r; + + r = sd_netlink_call(nl, req, 0, &reply); + if (r < 0) + return r; + + m->enumerating = true; + for (sd_netlink_message *reply_one = reply; reply_one; reply_one = sd_netlink_message_next(reply_one)) { + k = process(nl, reply_one, m); + if (k < 0 && r >= 0) + r = k; + } + m->enumerating = false; + + return r; +} + +static int manager_enumerate_links(Manager *m) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL; + int r; + + assert(m); + assert(m->rtnl); + + r = sd_rtnl_message_new_link(m->rtnl, &req, RTM_GETLINK, 0); + if (r < 0) + return r; + + return manager_enumerate_internal(m, m->rtnl, req, manager_rtnl_process_link); +} + +static int manager_enumerate_qdisc(Manager *m) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL; + int r; + + assert(m); + assert(m->rtnl); + + r = sd_rtnl_message_new_traffic_control(m->rtnl, &req, RTM_GETQDISC, 0, 0, 0); + if (r < 0) + return r; + + return manager_enumerate_internal(m, m->rtnl, req, manager_rtnl_process_qdisc); +} + +static int manager_enumerate_tclass(Manager *m) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL; + int r; + + assert(m); + assert(m->rtnl); + + r = sd_rtnl_message_new_traffic_control(m->rtnl, &req, RTM_GETTCLASS, 0, 0, 0); + if (r < 0) + return r; + + return manager_enumerate_internal(m, m->rtnl, req, manager_rtnl_process_tclass); +} + +static int manager_enumerate_addresses(Manager *m) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL; + int r; + + assert(m); + assert(m->rtnl); + + r = sd_rtnl_message_new_addr(m->rtnl, &req, RTM_GETADDR, 0, 0); + if (r < 0) + return r; + + return manager_enumerate_internal(m, m->rtnl, req, manager_rtnl_process_address); +} + +static int manager_enumerate_neighbors(Manager *m) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL; + int r; + + assert(m); + assert(m->rtnl); + + r = sd_rtnl_message_new_neigh(m->rtnl, &req, RTM_GETNEIGH, 0, AF_UNSPEC); + if (r < 0) + return r; + + return manager_enumerate_internal(m, m->rtnl, req, manager_rtnl_process_neighbor); +} + +static int manager_enumerate_routes(Manager *m) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL; + int r; + + assert(m); + assert(m->rtnl); + + if (!m->manage_foreign_routes) + return 0; + + r = sd_rtnl_message_new_route(m->rtnl, &req, RTM_GETROUTE, 0, 0); + if (r < 0) + return r; + + return manager_enumerate_internal(m, m->rtnl, req, manager_rtnl_process_route); +} + +static int manager_enumerate_rules(Manager *m) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL; + int r; + + assert(m); + assert(m->rtnl); + + if (!m->manage_foreign_rules) + return 0; + + r = sd_rtnl_message_new_routing_policy_rule(m->rtnl, &req, RTM_GETRULE, 0); + if (r < 0) + return r; + + return manager_enumerate_internal(m, m->rtnl, req, manager_rtnl_process_rule); +} + +static int manager_enumerate_nexthop(Manager *m) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL; + int r; + + assert(m); + assert(m->rtnl); + + r = sd_rtnl_message_new_nexthop(m->rtnl, &req, RTM_GETNEXTHOP, 0, 0); + if (r < 0) + return r; + + return manager_enumerate_internal(m, m->rtnl, req, manager_rtnl_process_nexthop); +} + +static int manager_enumerate_nl80211_wiphy(Manager *m) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL; + int r; + + assert(m); + assert(m->genl); + + r = sd_genl_message_new(m->genl, NL80211_GENL_NAME, NL80211_CMD_GET_WIPHY, &req); + if (r < 0) + return r; + + return manager_enumerate_internal(m, m->genl, req, manager_genl_process_nl80211_wiphy); +} + +static int manager_enumerate_nl80211_config(Manager *m) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL; + int r; + + assert(m); + assert(m->genl); + + r = sd_genl_message_new(m->genl, NL80211_GENL_NAME, NL80211_CMD_GET_INTERFACE, &req); + if (r < 0) + return r; + + return manager_enumerate_internal(m, m->genl, req, manager_genl_process_nl80211_config); +} + +static int manager_enumerate_nl80211_mlme(Manager *m) { + Link *link; + int r; + + assert(m); + assert(m->genl); + + HASHMAP_FOREACH(link, m->links_by_index) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL; + + if (link->wlan_iftype != NL80211_IFTYPE_STATION) + continue; + + r = sd_genl_message_new(m->genl, NL80211_GENL_NAME, NL80211_CMD_GET_STATION, &req); + if (r < 0) + return r; + + r = sd_netlink_message_append_u32(req, NL80211_ATTR_IFINDEX, link->ifindex); + if (r < 0) + return r; + + r = manager_enumerate_internal(m, m->genl, req, manager_genl_process_nl80211_mlme); + if (r < 0) + return r; + } + + return 0; +} + +int manager_enumerate(Manager *m) { + int r; + + r = manager_enumerate_links(m); + if (r < 0) + return log_error_errno(r, "Could not enumerate links: %m"); + + r = manager_enumerate_qdisc(m); + if (r == -EOPNOTSUPP) + log_debug_errno(r, "Could not enumerate QDiscs, ignoring: %m"); + else if (r < 0) + return log_error_errno(r, "Could not enumerate QDisc: %m"); + + r = manager_enumerate_tclass(m); + if (r == -EOPNOTSUPP) + log_debug_errno(r, "Could not enumerate TClasses, ignoring: %m"); + else if (r < 0) + return log_error_errno(r, "Could not enumerate TClass: %m"); + + r = manager_enumerate_addresses(m); + if (r < 0) + return log_error_errno(r, "Could not enumerate addresses: %m"); + + r = manager_enumerate_neighbors(m); + if (r < 0) + return log_error_errno(r, "Could not enumerate neighbors: %m"); + + /* NextHop support is added in kernel v5.3 (65ee00a9409f751188a8cdc0988167858eb4a536), + * and older kernels return -EOPNOTSUPP, or -EINVAL if SELinux is enabled. */ + r = manager_enumerate_nexthop(m); + if (r == -EOPNOTSUPP || (r == -EINVAL && mac_selinux_enforcing())) + log_debug_errno(r, "Could not enumerate nexthops, ignoring: %m"); + else if (r < 0) + return log_error_errno(r, "Could not enumerate nexthops: %m"); + + r = manager_enumerate_routes(m); + if (r < 0) + return log_error_errno(r, "Could not enumerate routes: %m"); + + /* If kernel is built with CONFIG_FIB_RULES=n, it returns -EOPNOTSUPP. */ + r = manager_enumerate_rules(m); + if (r == -EOPNOTSUPP) + log_debug_errno(r, "Could not enumerate routing policy rules, ignoring: %m"); + else if (r < 0) + return log_error_errno(r, "Could not enumerate routing policy rules: %m"); + + r = manager_enumerate_nl80211_wiphy(m); + if (r == -EOPNOTSUPP) + log_debug_errno(r, "Could not enumerate wireless LAN phy, ignoring: %m"); + else if (r < 0) + return log_error_errno(r, "Could not enumerate wireless LAN phy: %m"); + + r = manager_enumerate_nl80211_config(m); + if (r == -EOPNOTSUPP) + log_debug_errno(r, "Could not enumerate wireless LAN interfaces, ignoring: %m"); + else if (r < 0) + return log_error_errno(r, "Could not enumerate wireless LAN interfaces: %m"); + + r = manager_enumerate_nl80211_mlme(m); + if (r == -EOPNOTSUPP) + log_debug_errno(r, "Could not enumerate wireless LAN stations, ignoring: %m"); + else if (r < 0) + return log_error_errno(r, "Could not enumerate wireless LAN stations: %m"); + + return 0; +} + +static int set_hostname_handler(sd_bus_message *m, void *userdata, sd_bus_error *ret_error) { + const sd_bus_error *e; + int r; + + assert(m); + + e = sd_bus_message_get_error(m); + if (e) { + r = sd_bus_error_get_errno(e); + log_warning_errno(r, "Could not set hostname: %s", bus_error_message(e, r)); + } + + return 1; +} + +int manager_set_hostname(Manager *m, const char *hostname) { + int r; + + log_debug("Setting transient hostname: '%s'", strna(hostname)); + + r = free_and_strdup_warn(&m->dynamic_hostname, hostname); + if (r < 0) + return r; + + if (sd_bus_is_ready(m->bus) <= 0) { + log_debug("Not connected to system bus, setting system hostname later."); + return 0; + } + + r = bus_call_method_async( + m->bus, + NULL, + bus_hostname, + "SetHostname", + set_hostname_handler, + m, + "sb", + hostname, + false); + if (r < 0) + return log_error_errno(r, "Could not set transient hostname: %m"); + + return 0; +} + +static int set_timezone_handler(sd_bus_message *m, void *userdata, sd_bus_error *ret_error) { + const sd_bus_error *e; + int r; + + assert(m); + + e = sd_bus_message_get_error(m); + if (e) { + r = sd_bus_error_get_errno(e); + log_warning_errno(r, "Could not set timezone: %s", bus_error_message(e, r)); + } + + return 1; +} + +int manager_set_timezone(Manager *m, const char *tz) { + int r; + + assert(m); + assert(tz); + + log_debug("Setting system timezone: '%s'", tz); + r = free_and_strdup_warn(&m->dynamic_timezone, tz); + if (r < 0) + return r; + + if (sd_bus_is_ready(m->bus) <= 0) { + log_debug("Not connected to system bus, setting system timezone later."); + return 0; + } + + r = bus_call_method_async( + m->bus, + NULL, + bus_timedate, + "SetTimezone", + set_timezone_handler, + m, + "sb", + tz, + false); + if (r < 0) + return log_error_errno(r, "Could not set timezone: %m"); + + return 0; +} -- cgit v1.2.3