/* SPDX-License-Identifier: LGPL-2.1-or-later */ #include #include #include #include #include #include #include #include #include "sd-bus.h" #include "sd-daemon.h" #include "sd-event.h" #include "sd-id128.h" #include "bus-common-errors.h" #include "bus-internal.h" #include "bus-label.h" #include "bus-util.h" #include "path-util.h" #include "socket-util.h" #include "stdio-util.h" static int name_owner_change_callback(sd_bus_message *m, void *userdata, sd_bus_error *ret_error) { sd_event *e = userdata; assert(m); assert(e); sd_bus_close(sd_bus_message_get_bus(m)); sd_event_exit(e, 0); return 1; } int bus_log_address_error(int r, BusTransport transport) { bool hint = transport == BUS_TRANSPORT_LOCAL && r == -ENOMEDIUM; return log_error_errno(r, hint ? "Failed to set bus address: $DBUS_SESSION_BUS_ADDRESS and $XDG_RUNTIME_DIR not defined (consider using --machine=@.host --user to connect to bus of other user)" : "Failed to set bus address: %m"); } int bus_log_connect_error(int r, BusTransport transport) { bool hint_vars = transport == BUS_TRANSPORT_LOCAL && r == -ENOMEDIUM, hint_addr = transport == BUS_TRANSPORT_LOCAL && ERRNO_IS_PRIVILEGE(r); return log_error_errno(r, r == hint_vars ? "Failed to connect to bus: $DBUS_SESSION_BUS_ADDRESS and $XDG_RUNTIME_DIR not defined (consider using --machine=@.host --user to connect to bus of other user)" : r == hint_addr ? "Failed to connect to bus: Operation not permitted (consider using --machine=@.host --user to connect to bus of other user)" : "Failed to connect to bus: %m"); } int bus_async_unregister_and_exit(sd_event *e, sd_bus *bus, const char *name) { const char *match; const char *unique; int r; assert(e); assert(bus); assert(name); /* We unregister the name here and then wait for the * NameOwnerChanged signal for this event to arrive before we * quit. We do this in order to make sure that any queued * requests are still processed before we really exit. */ r = sd_bus_get_unique_name(bus, &unique); if (r < 0) return r; match = strjoina( "sender='org.freedesktop.DBus'," "type='signal'," "interface='org.freedesktop.DBus'," "member='NameOwnerChanged'," "path='/org/freedesktop/DBus'," "arg0='", name, "',", "arg1='", unique, "',", "arg2=''"); r = sd_bus_add_match_async(bus, NULL, match, name_owner_change_callback, NULL, e); if (r < 0) return r; r = sd_bus_release_name_async(bus, NULL, name, NULL, NULL); if (r < 0) return r; return 0; } int bus_event_loop_with_idle( sd_event *e, sd_bus *bus, const char *name, usec_t timeout, check_idle_t check_idle, void *userdata) { bool exiting = false; int r, code; assert(e); assert(bus); assert(name); for (;;) { bool idle; r = sd_event_get_state(e); if (r < 0) return r; if (r == SD_EVENT_FINISHED) break; if (check_idle) idle = check_idle(userdata); else idle = true; r = sd_event_run(e, exiting || !idle ? UINT64_MAX : timeout); if (r < 0) return r; if (r == 0 && !exiting && idle) { /* Inform the service manager that we are going down, so that it will queue all * further start requests, instead of assuming we are already running. */ sd_notify(false, "STOPPING=1"); r = bus_async_unregister_and_exit(e, bus, name); if (r < 0) return r; exiting = true; continue; } } r = sd_event_get_exit_code(e, &code); if (r < 0) return r; return code; } int bus_name_has_owner(sd_bus *c, const char *name, sd_bus_error *error) { _cleanup_(sd_bus_message_unrefp) sd_bus_message *rep = NULL; int r, has_owner = 0; assert(c); assert(name); r = sd_bus_call_method(c, "org.freedesktop.DBus", "/org/freedesktop/dbus", "org.freedesktop.DBus", "NameHasOwner", error, &rep, "s", name); if (r < 0) return r; r = sd_bus_message_read_basic(rep, 'b', &has_owner); if (r < 0) return sd_bus_error_set_errno(error, r); return has_owner; } bool bus_error_is_unknown_service(const sd_bus_error *error) { return sd_bus_error_has_names(error, SD_BUS_ERROR_SERVICE_UNKNOWN, SD_BUS_ERROR_NAME_HAS_NO_OWNER, BUS_ERROR_NO_SUCH_UNIT); } int bus_check_peercred(sd_bus *c) { struct ucred ucred; int fd, r; assert(c); fd = sd_bus_get_fd(c); if (fd < 0) return fd; r = getpeercred(fd, &ucred); if (r < 0) return r; if (ucred.uid != 0 && ucred.uid != geteuid()) return -EPERM; return 1; } int bus_connect_system_systemd(sd_bus **_bus) { _cleanup_(sd_bus_close_unrefp) sd_bus *bus = NULL; int r; assert(_bus); if (geteuid() != 0) return sd_bus_default_system(_bus); /* If we are root then let's talk directly to the system * instance, instead of going via the bus */ r = sd_bus_new(&bus); if (r < 0) return r; r = sd_bus_set_address(bus, "unix:path=/run/systemd/private"); if (r < 0) return r; r = sd_bus_start(bus); if (r < 0) return sd_bus_default_system(_bus); r = bus_check_peercred(bus); if (r < 0) return r; *_bus = TAKE_PTR(bus); return 0; } int bus_connect_user_systemd(sd_bus **_bus) { _cleanup_(sd_bus_close_unrefp) sd_bus *bus = NULL; _cleanup_free_ char *ee = NULL; const char *e; int r; assert(_bus); e = secure_getenv("XDG_RUNTIME_DIR"); if (!e) return sd_bus_default_user(_bus); ee = bus_address_escape(e); if (!ee) return -ENOMEM; r = sd_bus_new(&bus); if (r < 0) return r; bus->address = strjoin("unix:path=", ee, "/systemd/private"); if (!bus->address) return -ENOMEM; r = sd_bus_start(bus); if (r < 0) return sd_bus_default_user(_bus); r = bus_check_peercred(bus); if (r < 0) return r; *_bus = TAKE_PTR(bus); return 0; } int bus_connect_transport( BusTransport transport, const char *host, bool user, sd_bus **ret) { _cleanup_(sd_bus_close_unrefp) sd_bus *bus = NULL; int r; assert(transport >= 0); assert(transport < _BUS_TRANSPORT_MAX); assert(ret); assert_return((transport == BUS_TRANSPORT_LOCAL) == !host, -EINVAL); assert_return(transport != BUS_TRANSPORT_REMOTE || !user, -EOPNOTSUPP); switch (transport) { case BUS_TRANSPORT_LOCAL: if (user) r = sd_bus_default_user(&bus); else { if (sd_booted() <= 0) /* Print a friendly message when the local system is actually not running systemd as PID 1. */ return log_error_errno(SYNTHETIC_ERRNO(EHOSTDOWN), "System has not been booted with systemd as init system (PID 1). Can't operate."); r = sd_bus_default_system(&bus); } break; case BUS_TRANSPORT_REMOTE: r = sd_bus_open_system_remote(&bus, host); break; case BUS_TRANSPORT_MACHINE: if (user) r = sd_bus_open_user_machine(&bus, host); else r = sd_bus_open_system_machine(&bus, host); break; default: assert_not_reached(); } if (r < 0) return r; r = sd_bus_set_exit_on_disconnect(bus, true); if (r < 0) return r; *ret = TAKE_PTR(bus); return 0; } int bus_connect_transport_systemd(BusTransport transport, const char *host, bool user, sd_bus **bus) { int r; assert(transport >= 0); assert(transport < _BUS_TRANSPORT_MAX); assert(bus); assert_return((transport == BUS_TRANSPORT_LOCAL) == !host, -EINVAL); assert_return(transport == BUS_TRANSPORT_LOCAL || !user, -EOPNOTSUPP); switch (transport) { case BUS_TRANSPORT_LOCAL: if (user) r = bus_connect_user_systemd(bus); else { if (sd_booted() <= 0) /* Print a friendly message when the local system is actually not running systemd as PID 1. */ return log_error_errno(SYNTHETIC_ERRNO(EHOSTDOWN), "System has not been booted with systemd as init system (PID 1). Can't operate."); r = bus_connect_system_systemd(bus); } break; case BUS_TRANSPORT_REMOTE: r = sd_bus_open_system_remote(bus, host); break; case BUS_TRANSPORT_MACHINE: r = sd_bus_open_system_machine(bus, host); break; default: assert_not_reached(); } return r; } /** * bus_path_encode_unique() - encode unique object path * @b: bus connection or NULL * @prefix: object path prefix * @sender_id: unique-name of client, or NULL * @external_id: external ID to be chosen by client, or NULL * @ret_path: storage for encoded object path pointer * * Whenever we provide a bus API that allows clients to create and manage * server-side objects, we need to provide a unique name for these objects. If * we let the server choose the name, we suffer from a race condition: If a * client creates an object asynchronously, it cannot destroy that object until * it received the method reply. It cannot know the name of the new object, * thus, it cannot destroy it. Furthermore, it enforces a round-trip. * * Therefore, many APIs allow the client to choose the unique name for newly * created objects. There're two problems to solve, though: * 1) Object names are usually defined via dbus object paths, which are * usually globally namespaced. Therefore, multiple clients must be able * to choose unique object names without interference. * 2) If multiple libraries share the same bus connection, they must be * able to choose unique object names without interference. * The first problem is solved easily by prefixing a name with the * unique-bus-name of a connection. The server side must enforce this and * reject any other name. The second problem is solved by providing unique * suffixes from within sd-bus. * * This helper allows clients to create unique object-paths. It uses the * template '/prefix/sender_id/external_id' and returns the new path in * @ret_path (must be freed by the caller). * If @sender_id is NULL, the unique-name of @b is used. If @external_id is * NULL, this function allocates a unique suffix via @b (by requesting a new * cookie). If both @sender_id and @external_id are given, @b can be passed as * NULL. * * Returns: 0 on success, negative error code on failure. */ int bus_path_encode_unique(sd_bus *b, const char *prefix, const char *sender_id, const char *external_id, char **ret_path) { _cleanup_free_ char *sender_label = NULL, *external_label = NULL; char external_buf[DECIMAL_STR_MAX(uint64_t)], *p; int r; assert_return(b || (sender_id && external_id), -EINVAL); assert_return(sd_bus_object_path_is_valid(prefix), -EINVAL); assert_return(ret_path, -EINVAL); if (!sender_id) { r = sd_bus_get_unique_name(b, &sender_id); if (r < 0) return r; } if (!external_id) { xsprintf(external_buf, "%"PRIu64, ++b->cookie); external_id = external_buf; } sender_label = bus_label_escape(sender_id); if (!sender_label) return -ENOMEM; external_label = bus_label_escape(external_id); if (!external_label) return -ENOMEM; p = path_join(prefix, sender_label, external_label); if (!p) return -ENOMEM; *ret_path = p; return 0; } /** * bus_path_decode_unique() - decode unique object path * @path: object path to decode * @prefix: object path prefix * @ret_sender: output parameter for sender-id label * @ret_external: output parameter for external-id label * * This does the reverse of bus_path_encode_unique() (see its description for * details). Both trailing labels, sender-id and external-id, are unescaped and * returned in the given output parameters (the caller must free them). * * Note that this function returns 0 if the path does not match the template * (see bus_path_encode_unique()), 1 if it matched. * * Returns: Negative error code on failure, 0 if the given object path does not * match the template (return parameters are set to NULL), 1 if it was * parsed successfully (return parameters contain allocated labels). */ int bus_path_decode_unique(const char *path, const char *prefix, char **ret_sender, char **ret_external) { const char *p, *q; char *sender, *external; assert(sd_bus_object_path_is_valid(path)); assert(sd_bus_object_path_is_valid(prefix)); assert(ret_sender); assert(ret_external); p = object_path_startswith(path, prefix); if (!p) { *ret_sender = NULL; *ret_external = NULL; return 0; } q = strchr(p, '/'); if (!q) { *ret_sender = NULL; *ret_external = NULL; return 0; } sender = bus_label_unescape_n(p, q - p); external = bus_label_unescape(q + 1); if (!sender || !external) { free(sender); free(external); return -ENOMEM; } *ret_sender = sender; *ret_external = external; return 1; } int bus_track_add_name_many(sd_bus_track *t, char **l) { int r = 0; char **i; assert(t); /* Continues adding after failure, and returns the first failure. */ STRV_FOREACH(i, l) { int k; k = sd_bus_track_add_name(t, *i); if (k < 0 && r >= 0) r = k; } return r; } int bus_open_system_watch_bind_with_description(sd_bus **ret, const char *description) { _cleanup_(sd_bus_close_unrefp) sd_bus *bus = NULL; const char *e; int r; assert(ret); /* Match like sd_bus_open_system(), but with the "watch_bind" feature and the Connected() signal * turned on. */ r = sd_bus_new(&bus); if (r < 0) return r; if (description) { r = sd_bus_set_description(bus, description); if (r < 0) return r; } e = secure_getenv("DBUS_SYSTEM_BUS_ADDRESS"); if (!e) e = DEFAULT_SYSTEM_BUS_ADDRESS; r = sd_bus_set_address(bus, e); if (r < 0) return r; r = sd_bus_set_bus_client(bus, true); if (r < 0) return r; r = sd_bus_negotiate_creds(bus, true, SD_BUS_CREDS_UID|SD_BUS_CREDS_EUID|SD_BUS_CREDS_EFFECTIVE_CAPS); if (r < 0) return r; r = sd_bus_set_watch_bind(bus, true); if (r < 0) return r; r = sd_bus_set_connected_signal(bus, true); if (r < 0) return r; r = sd_bus_start(bus); if (r < 0) return r; *ret = TAKE_PTR(bus); return 0; } int bus_reply_pair_array(sd_bus_message *m, char **l) { _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; char **k, **v; int r; assert(m); /* Reply to the specified message with a message containing a dictionary put together from the * specified strv */ r = sd_bus_message_new_method_return(m, &reply); if (r < 0) return r; r = sd_bus_message_open_container(reply, 'a', "{ss}"); if (r < 0) return r; STRV_FOREACH_PAIR(k, v, l) { r = sd_bus_message_append(reply, "{ss}", *k, *v); if (r < 0) return r; } r = sd_bus_message_close_container(reply); if (r < 0) return r; return sd_bus_send(NULL, reply, NULL); } static void bus_message_unref_wrapper(void *m) { sd_bus_message_unref(m); } const struct hash_ops bus_message_hash_ops = { .hash = trivial_hash_func, .compare = trivial_compare_func, .free_value = bus_message_unref_wrapper, };