diff --git a/src/main.c b/src/main.c index 5704ff9..0b5a083 100644 --- a/src/main.c +++ b/src/main.c @@ -1,1330 +1,1330 @@ /* * Copyright (C) 2011 Jiaju Zhang * Copyright (C) 2013-2014 Philipp Marek * * This program 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 2.1 of the License, or (at your option) any later version. * * This software 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 this library; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "log.h" #include "booth.h" #include "config.h" #include "transport.h" #include "inline-fn.h" #include "pacemaker.h" #include "ticket.h" #define RELEASE_VERSION "1.0" #define CLIENT_NALLOC 32 int daemonize = 0; /** Structure for "clients". * Filehandles with incoming data get registered here (and in pollfds), * along with their callbacks. * Because these can be reallocated with every new fd, addressing * happens _only_ by their numeric index. */ struct client *clients = NULL; struct pollfd *pollfds = NULL; static int client_maxi; static int client_size = 0; static const struct booth_site _no_leader = { .addr_string = "none", .site_id = NO_ONE, }; struct booth_site *no_leader = (struct booth_site*)& _no_leader; typedef enum { BOOTHD_STARTED=0, BOOTHD_STARTING } BOOTH_DAEMON_STATE; int poll_timeout = POLL_TIMEOUT; typedef enum { OP_LIST = 1, OP_GRANT, OP_REVOKE, } operation_t; struct booth_config *booth_conf; struct command_line cl; int do_read(int fd, void *buf, size_t count) { int rv, off = 0; while (off < count) { rv = read(fd, (char *)buf + off, count - off); if (rv == 0) return -1; if (rv == -1 && errno == EINTR) continue; if (rv == -1) return -1; off += rv; } return 0; } int do_write(int fd, void *buf, size_t count) { int rv, off = 0; retry: rv = write(fd, (char *)buf + off, count); if (rv == -1 && errno == EINTR) goto retry; /* If we cannot write _any_ data, we'd be in an (potential) loop. */ if (rv <= 0) { log_error("write failed: %s (%d)", strerror(errno), errno); return rv; } if (rv != count) { count -= rv; off += rv; goto retry; } return 0; } static void client_alloc(void) { int i; if (!clients) { clients = malloc(CLIENT_NALLOC * sizeof(struct client)); pollfds = malloc(CLIENT_NALLOC * sizeof(struct pollfd)); } else { clients = realloc(clients, (client_size + CLIENT_NALLOC) * sizeof(struct client)); pollfds = realloc(pollfds, (client_size + CLIENT_NALLOC) * sizeof(struct pollfd)); } if (!clients || !pollfds) { log_error("can't alloc for client array"); exit(1); } for (i = client_size; i < client_size + CLIENT_NALLOC; i++) { clients[i].workfn = NULL; clients[i].deadfn = NULL; clients[i].fd = -1; pollfds[i].fd = -1; pollfds[i].revents = 0; } client_size += CLIENT_NALLOC; } static void client_dead(int ci) { if (clients[ci].fd != -1) close(clients[ci].fd); clients[ci].fd = -1; clients[ci].workfn = NULL; pollfds[ci].fd = -1; } int client_add(int fd, const struct booth_transport *tpt, void (*workfn)(int ci), void (*deadfn)(int ci)) { int i; struct client *c; if (client_size + 2 >= client_maxi ) { client_alloc(); } for (i = 0; i < client_size; i++) { c = clients + i; if (c->fd != -1) continue; c->workfn = workfn; if (deadfn) c->deadfn = deadfn; else c->deadfn = client_dead; c->transport = tpt; c->fd = fd; pollfds[i].fd = fd; pollfds[i].events = POLLIN; if (i > client_maxi) client_maxi = i; return i; } assert(!"no client"); } /* Only used for client requests, TCP ???*/ void process_connection(int ci) { struct boothc_ticket_msg msg; int rv, len, expr, fd; void (*deadfn) (int ci); fd = clients[ci].fd; rv = do_read(fd, &msg.header, sizeof(msg.header)); if (rv < 0) { if (errno == ECONNRESET) log_debug("client %d connection reset for fd %d", ci, clients[ci].fd); goto kill; } if (check_boothc_header(&msg.header, -1) < 0) goto kill; /* Basic sanity checks already done. */ len = ntohl(msg.header.length); if (len) { if (len != sizeof(msg)) { bad_len: log_error("got wrong length %u", len); return; } expr = len - sizeof(msg.header); rv = do_read(clients[ci].fd, msg.header.data, expr); if (rv < 0) { log_error("connection %d read data error %d, wanted %d", ci, rv, expr); goto kill; } } /* For CMD_GRANT and CMD_REVOKE: * Don't close connection immediately, but send * result a second later? */ switch (ntohl(msg.header.cmd)) { case CMD_LIST: ticket_answer_list(fd, &msg); goto kill; case CMD_GRANT: /* Expect boothc_ticket_site_msg. */ if (len != sizeof(msg)) goto bad_len; ticket_answer_grant(fd, &msg); goto kill; case CMD_REVOKE: /* Expect boothc_ticket_site_msg. */ if (len != sizeof(msg)) goto bad_len; ticket_answer_revoke(fd, &msg); goto kill; default: log_error("connection %d cmd %x unknown", ci, ntohl(msg.header.cmd)); init_header(&msg.header,CMR_GENERAL, RLT_INVALID_ARG, sizeof(msg.header)); send_header_only(fd, &msg.header); goto kill; } assert(0); return; kill: deadfn = clients[ci].deadfn; if(deadfn) { deadfn(ci); } return; } /** Callback function for the listening TCP socket. */ static void process_listener(int ci) { int fd, i; fd = accept(clients[ci].fd, NULL, NULL); if (fd < 0) { log_error("process_listener: accept error for fd %d: %s (%d)", clients[ci].fd, strerror(errno), errno); if (clients[ci].deadfn) clients[ci].deadfn(ci); return; } i = client_add(fd, clients[ci].transport, process_connection, NULL); log_debug("add client connection %d fd %d", i, fd); } static int setup_config(int type) { int rv; rv = read_config(cl.configfile); if (rv < 0) goto out; /* Set "local" pointer, ignoring errors. */ if (cl.type == DAEMON && cl.site[0]) { if (!find_site_by_name(cl.site, &local, 1)) { log_error("Cannot find \"%s\" in the configuration.", cl.site); return -EINVAL; } local->local = 1; } else find_myself(NULL, type == CLIENT); rv = check_config(type); if (rv < 0) goto out; /* Per default the PID file name is derived from the * configuration name. */ if (!cl.lockfile[0]) { snprintf(cl.lockfile, sizeof(cl.lockfile)-1, "%s/%s.pid", BOOTH_RUN_DIR, booth_conf->name); } out: return rv; } static int setup_transport(void) { int rv; rv = transport()->init(message_recv); if (rv < 0) { log_error("failed to init booth_transport %s", transport()->name); goto out; } rv = booth_transport[TCP].init(NULL); if (rv < 0) { log_error("failed to init booth_transport[TCP]"); goto out; } out: return rv; } static int write_daemon_state(int fd, int state) { char buffer[1024]; int rv, size; size = sizeof(buffer) - 1; rv = snprintf(buffer, size, "booth_pid=%d " "booth_state=%s " "booth_type=%s " "booth_cfg_name='%s' " "booth_addr_string='%s' " "booth_port=%d\n", getpid(), ( state == BOOTHD_STARTED ? "started" : state == BOOTHD_STARTING ? "starting" : "invalid"), type_to_string(local->type), booth_conf->name, local->addr_string, booth_conf->port); if (rv < 0 || rv == size) { log_error("Buffer filled up in write_daemon_state()."); return -1; } size = rv; rv = ftruncate(fd, 0); if (rv < 0) { log_error("lockfile %s truncate error %d: %s", cl.lockfile, errno, strerror(errno)); return rv; } rv = lseek(fd, 0, SEEK_SET); if (rv < 0) { log_error("lseek set fd(%d) offset to 0 error, return(%d), message(%s)", fd, rv, strerror(errno)); rv = -1; return rv; } rv = write(fd, buffer, size); if (rv != size) { log_error("write to fd(%d, %d) returned %d, errno %d, message(%s)", fd, size, rv, errno, strerror(errno)); return -1; } return 0; } static int loop(int fd) { void (*workfn) (int ci); void (*deadfn) (int ci); int rv, i; rv = setup_transport(); if (rv < 0) goto fail; rv = setup_ticket(); if (rv < 0) goto fail; client_add(local->tcp_fd, booth_transport + TCP, process_listener, NULL); rv = write_daemon_state(fd, BOOTHD_STARTED); if (rv != 0) { log_error("write daemon state %d to lockfile error %s: %s", BOOTHD_STARTED, cl.lockfile, strerror(errno)); goto fail; } if (cl.type == ARBITRATOR) log_info("BOOTH arbitrator daemon started"); else if (cl.type == SITE) log_info("BOOTH cluster site daemon started"); while (1) { rv = poll(pollfds, client_maxi + 1, poll_timeout); if (rv == -1 && errno == EINTR) continue; if (rv < 0) { log_error("poll failed: %s (%d)", strerror(errno), errno); goto fail; } for (i = 0; i <= client_maxi; i++) { if (clients[i].fd < 0) continue; if (pollfds[i].revents & POLLIN) { workfn = clients[i].workfn; if (workfn) workfn(i); } if (pollfds[i].revents & (POLLERR | POLLHUP | POLLNVAL)) { deadfn = clients[i].deadfn; if (deadfn) deadfn(i); } } process_tickets(); } return 0; fail: return -1; } static int query_get_string_answer(cmd_request_t cmd) { struct booth_site *site; struct boothc_header reply; char *data; int data_len; int rv; struct booth_transport const *tpt; data = NULL; init_header(&cl.msg.header, cmd, 0, sizeof(cl.msg)); if (!*cl.site) site = local; else if (!find_site_by_name(cl.site, &site, 1)) { log_error("cannot find site \"%s\"", cl.site); rv = ENOENT; goto out; } tpt = booth_transport + TCP; rv = tpt->open(site); if (rv < 0) goto out_free; rv = tpt->send(site, &cl.msg, sizeof(cl.msg)); if (rv < 0) goto out_free; rv = tpt->recv(site, &reply, sizeof(reply)); if (rv < 0) goto out_free; data_len = ntohl(reply.length) - sizeof(reply); data = malloc(data_len); if (!data) { rv = -ENOMEM; goto out_free; } rv = tpt->recv(site, data, data_len); if (rv < 0) goto out_free; do_write(STDOUT_FILENO, data, data_len); rv = 0; out_free: free(data); tpt->close(site); out: return rv; } static int test_reply(int reply_code, cmd_request_t cmd) { int rv = 0; switch (reply_code) { case RLT_OVERGRANT: log_info("You're granting a granted ticket. " "If you wanted to migrate a ticket, " "use revoke first, then use grant."); rv = -1; break; case RLT_ASYNC: if (cmd == CMD_GRANT) log_info("grant command sent, result will be returned " "asynchronously, you can get the result from " "the log files"); else if (cmd == CMD_REVOKE) log_info("revoke command sent, result will be returned " "asynchronously, you can get the result from " "the log files."); else log_error("internal error reading reply result!"); rv = 0; break; case RLT_SYNC_SUCC: case RLT_SUCCESS: if (cmd == CMD_GRANT) log_info("grant succeeded!"); else if (cmd == CMD_REVOKE) log_info("revoke succeeded!"); rv = 0; break; case RLT_SYNC_FAIL: if (cmd == CMD_GRANT) log_info("grant failed!"); else if (cmd == CMD_REVOKE) log_info("revoke failed!"); rv = -1; break; case RLT_INVALID_ARG: log_error("ticket \"%s\" does not exist", cl.msg.ticket.id); break; case RLT_REDIRECT: /* talk to another site */ rv = 1; break; default: log_error("got an error code: %x", rv); rv = -1; } return rv; } static int do_command(cmd_request_t cmd) { struct booth_site *site; struct boothc_ticket_msg reply; struct booth_transport const *tpt; uint32_t leader_id; int rv; rv = 0; site = NULL; if (!*cl.site) site = local; else { if (!find_site_by_name(cl.site, &site, 1)) { log_error("Site \"%s\" not configured.", cl.site); goto out_close; } if (site->type == ARBITRATOR) { log_error("Site \"%s\" is an arbitrator, cannot grant ticket there.", cl.site); goto out_close; } assert(site->type == SITE); } /* We don't check for existence of ticket, so that asking can be * done without local configuration, too. * Although, that means that the UDP port has to be specified, too. */ if (!cl.msg.ticket.id[0]) { /* If the loaded configuration has only a single ticket defined, use that. */ if (booth_conf->ticket_count == 1) { strcpy(cl.msg.ticket.id, booth_conf->ticket[0].name); } else { log_error("No ticket given."); goto out_close; } } redirect: init_header(&cl.msg.header, cmd, 0, sizeof(cl.msg)); /* Always use TCP for client - at least for now. */ tpt = booth_transport + TCP; rv = tpt->open(site); if (rv < 0) goto out_close; rv = tpt->send(site, &cl.msg, sizeof(cl.msg)); if (rv < 0) goto out_close; rv = tpt->recv(site, &reply, sizeof(reply)); if (rv < 0) goto out_close; rv = test_reply(ntohl(reply.header.result), cmd); if (rv == 1) { local_transport->close(site); leader_id = ntohl(reply.ticket.leader); if (!find_site_by_id(leader_id, &site)) { log_error("Message with unknown redirect site %x received", leader_id); return rv; } goto redirect; } out_close: if (site) local_transport->close(site); return rv; } static int do_grant(void) { return do_command(CMD_GRANT); } static int do_revoke(void) { return do_command(CMD_REVOKE); } static int _lockfile(int mode, int *fdp, pid_t *locked_by) { struct flock lock; int fd, rv; /* After reboot the directory may not yet exist. * Try to create it, but ignore errors. */ if (strncmp(cl.lockfile, BOOTH_RUN_DIR, strlen(BOOTH_RUN_DIR)) == 0) mkdir(BOOTH_RUN_DIR, 0775); if (locked_by) *locked_by = 0; *fdp = -1; fd = open(cl.lockfile, mode, 0664); if (fd < 0) return errno; *fdp = fd; lock.l_type = F_WRLCK; lock.l_start = 0; lock.l_whence = SEEK_SET; lock.l_len = 0; lock.l_pid = 0; if (fcntl(fd, F_SETLK, &lock) == 0) return 0; rv = errno; if (locked_by) if (fcntl(fd, F_GETLK, &lock) == 0) *locked_by = lock.l_pid; return rv; } static inline int is_root(void) { /* TODO: getuid()? Better way to check? */ return geteuid() == 0; } static int create_lockfile(void) { int rv, fd; fd = -1; rv = _lockfile(O_CREAT | O_WRONLY, &fd, NULL); if (fd == -1) { log_error("lockfile %s open error %d: %s", cl.lockfile, rv, strerror(rv)); return -1; } if (rv < 0) { log_error("lockfile %s setlk error %d: %s", cl.lockfile, rv, strerror(rv)); goto fail; } rv = write_daemon_state(fd, BOOTHD_STARTING); if (rv != 0) { log_error("write daemon state %d to lockfile error %s: %s", BOOTHD_STARTING, cl.lockfile, strerror(errno)); goto fail; } if (is_root()) { if (fchown(fd, booth_conf->uid, booth_conf->gid) < 0) log_error("fchown() on lockfile said %d: %s", errno, strerror(errno)); } return fd; fail: close(fd); return -1; } static void unlink_lockfile(int fd) { unlink(cl.lockfile); close(fd); } static void print_usage(void) { printf("Usages:\n"); printf(" booth daemon [-c config] [-D]\n"); printf(" booth [client] {list|grant|revoke} [options]\n"); printf(" booth status [-c config] [-D]\n"); printf("\n"); printf("Client operations:\n"); printf(" list: List all the tickets\n"); printf(" grant: Grant ticket to site\n"); printf(" revoke: Revoke ticket from site\n"); printf("\n"); printf("Options:\n"); printf(" -c FILE Specify config file [default " BOOTH_DEFAULT_CONF "]\n"); printf(" Can be a path or a name without \".conf\" suffix\n"); printf(" -D Enable debugging to stderr and don't fork\n"); printf(" -S Systemd mode (no forking)\n"); printf(" -t ticket name\n"); printf(" -s site name\n"); printf(" -l LOCKFILE Specify lock file path (daemon only)\n"); printf(" -h Print this help, then exit\n"); printf("\n"); printf("Please see the man page for details.\n"); } #define OPTION_STRING "c:Dl:t:s:hS" void safe_copy(char *dest, char *value, size_t buflen, const char *description) { int content_len = buflen - 1; if (strlen(value) >= content_len) { fprintf(stderr, "'%s' exceeds maximum %s length of %d\n", value, description, content_len); exit(EXIT_FAILURE); } strncpy(dest, value, content_len); dest[content_len] = 0; } static int host_convert(char *hostname, char *ip_str, size_t ip_size) { struct addrinfo *result = NULL, hints = {0}; int re = -1; memset(&hints, 0, sizeof(hints)); hints.ai_family = BOOTH_PROTO_FAMILY; hints.ai_socktype = SOCK_DGRAM; re = getaddrinfo(hostname, NULL, &hints, &result); if (re == 0) { struct in_addr addr = ((struct sockaddr_in *)result->ai_addr)->sin_addr; const char *re_ntop = inet_ntop(BOOTH_PROTO_FAMILY, &addr, ip_str, ip_size); if (re_ntop == NULL) { re = -1; } } freeaddrinfo(result); return re; } static int read_arguments(int argc, char **argv) { int optchar; char *arg1 = argv[1]; char *op = NULL; char *cp; char site_arg[INET_ADDRSTRLEN] = {0}; int left; if (argc < 2 || !strcmp(arg1, "help") || !strcmp(arg1, "--help") || !strcmp(arg1, "-h")) { print_usage(); exit(EXIT_SUCCESS); } if (!strcmp(arg1, "version") || !strcmp(arg1, "--version") || !strcmp(arg1, "-V")) { printf("%s %s (built %s %s)\n", argv[0], RELEASE_VERSION, __DATE__, __TIME__); exit(EXIT_SUCCESS); } if (strcmp(arg1, "arbitrator") == 0 || strcmp(arg1, "site") == 0 || strcmp(arg1, "start") == 0 || strcmp(arg1, "daemon") == 0) { cl.type = DAEMON; optind = 2; } else if (strcmp(arg1, "status") == 0) { cl.type = STATUS; optind = 2; } else if (strcmp(arg1, "client") == 0) { cl.type = CLIENT; if (argc < 3) { print_usage(); exit(EXIT_FAILURE); } op = argv[2]; optind = 3; } else { cl.type = CLIENT; op = argv[1]; optind = 2; } if (cl.type == CLIENT) { if (!strcmp(op, "list")) cl.op = OP_LIST; else if (!strcmp(op, "grant")) cl.op = OP_GRANT; else if (!strcmp(op, "revoke")) cl.op = OP_REVOKE; else { fprintf(stderr, "client operation \"%s\" is unknown\n", op); exit(EXIT_FAILURE); } } while (optind < argc) { optchar = getopt(argc, argv, OPTION_STRING); switch (optchar) { case 'c': if (strchr(optarg, '/')) { safe_copy(cl.configfile, optarg, sizeof(cl.configfile), "config file"); } else { /* If no "/" in there, use with default directory. */ strcpy(cl.configfile, BOOTH_DEFAULT_CONF_DIR); cp = cl.configfile + strlen(BOOTH_DEFAULT_CONF_DIR); assert(cp > cl.configfile); assert(*(cp-1) == '/'); /* Write at the \0, ie. after the "/" */ safe_copy(cp, optarg, (sizeof(cl.configfile) - (cp - cl.configfile) - strlen(BOOTH_DEFAULT_CONF_EXT)), "config name"); /* If no extension, append ".conf". * Space is available, see -strlen() above. */ if (!strchr(cp, '.')) strcat(cp, BOOTH_DEFAULT_CONF_EXT); } break; case 'D': debug_level++; /* Fall through */ case 'S': daemonize = 1; break; case 'l': safe_copy(cl.lockfile, optarg, sizeof(cl.lockfile), "lock file"); break; case 't': if (cl.op == OP_GRANT || cl.op == OP_REVOKE) { safe_copy(cl.msg.ticket.id, optarg, sizeof(cl.msg.ticket.id), "ticket name"); } else { print_usage(); exit(EXIT_FAILURE); } break; case 's': /* For testing and debugging: allow "-s site" also for * daemon start, so that the address that should be used * can be set manually. * This makes it easier to start multiple processes * on one machine. */ if (cl.type == CLIENT || (cl.type == DAEMON && debug_level)) { int re = host_convert(optarg, site_arg, INET_ADDRSTRLEN); if (re == 0) { safe_copy(cl.site, site_arg, sizeof(cl.site), "site name"); } else { safe_copy(cl.site, optarg, sizeof(cl.site), "site name"); } } else { log_error("\"-s\" not allowed in daemon mode."); exit(EXIT_FAILURE); } break; case 'h': print_usage(); exit(EXIT_SUCCESS); break; case ':': case '?': fprintf(stderr, "Please use '-h' for usage.\n"); exit(EXIT_FAILURE); break; case -1: /* No more parameters on cmdline, only arguments. */ goto extra_args; default: goto unknown; }; } return 0; extra_args: if (cl.type == CLIENT && !cl.msg.ticket.id[0]) { /* Use additional argument as ticket name. */ safe_copy(cl.msg.ticket.id, argv[optind], sizeof(cl.msg.ticket.id), "ticket name"); optind++; } if (optind == argc) return 0; left = argc - optind; fprintf(stderr, "Superfluous argument%s: %s%s\n", left == 1 ? "" : "s", argv[optind], left == 1 ? "" : "..."); exit(EXIT_FAILURE); unknown: fprintf(stderr, "unknown option: %s\n", argv[optind]); exit(EXIT_FAILURE); } static void set_scheduler(void) { struct sched_param sched_param; struct rlimit rlimit; int rv; rlimit.rlim_cur = RLIM_INFINITY; rlimit.rlim_max = RLIM_INFINITY; setrlimit(RLIMIT_MEMLOCK, &rlimit); rv = mlockall(MCL_CURRENT | MCL_FUTURE); if (rv < 0) { log_error("mlockall failed"); } rv = sched_get_priority_max(SCHED_RR); if (rv != -1) { sched_param.sched_priority = rv; rv = sched_setscheduler(0, SCHED_RR, &sched_param); if (rv == -1) log_error("could not set SCHED_RR priority %d: %s (%d)", sched_param.sched_priority, strerror(errno), errno); } else { log_error("could not get maximum scheduler priority err %d", errno); } } static void set_oom_adj(int val) { FILE *fp; fp = fopen("/proc/self/oom_adj", "w"); if (!fp) return; fprintf(fp, "%i", val); fclose(fp); } static int do_status(int type) { pid_t pid; int rv, lock_fd, ret; const char *reason = NULL; char lockfile_data[1024], *cp; ret = PCMK_OCF_NOT_RUNNING; /* TODO: query all, and return quit only if it's _cleanly_ not * running, ie. _neither_ of port/lockfile/process is available? * * Currently a single failure says "not running", even if "only" the * lockfile has been removed. */ rv = setup_config(type); if (rv) { reason = "Error reading configuration."; ret = PCMK_OCF_UNKNOWN_ERROR; goto quit; } if (!local) { reason = "No Service IP active here."; goto quit; } rv = _lockfile(O_RDWR, &lock_fd, &pid); if (rv == 0) { reason = "PID file not locked."; goto quit; } if (lock_fd == -1) { reason = "No PID file."; goto quit; } if (pid) { fprintf(stdout, "booth_lockpid=%d ", pid); fflush(stdout); } rv = read(lock_fd, lockfile_data, sizeof(lockfile_data) - 1); if (rv < 4) { reason = "Cannot read lockfile data."; ret = PCMK_LSB_UNKNOWN_ERROR; goto quit; } lockfile_data[rv] = 0; if (lock_fd != -1) close(lock_fd); /* Make sure it's only a single line */ cp = strchr(lockfile_data, '\r'); if (cp) *cp = 0; cp = strchr(lockfile_data, '\n'); if (cp) *cp = 0; - rv = setup_udp_server(1); + rv = setup_tcp_listener(1); if (rv == 0) { - reason = "UDP port not in use."; + reason = "TCP port not in use."; goto quit; } fprintf(stdout, "booth_lockfile='%s' %s\n", cl.lockfile, lockfile_data); if (daemonize) fprintf(stderr, "Booth at %s port %d seems to be running.\n", local->addr_string, booth_conf->port); return 0; quit: log_debug("not running: %s", reason); /* Ie. "DEBUG" */ if (daemonize) fprintf(stderr, "not running: %s\n", reason); return ret; } static int limit_this_process(void) { int rv; if (!is_root()) return 0; if (setregid(booth_conf->gid, booth_conf->gid) < 0) { rv = errno; log_error("setregid() didn't work: %s", strerror(rv)); return rv; } if (setreuid(booth_conf->uid, booth_conf->uid) < 0) { rv = errno; log_error("setreuid() didn't work: %s", strerror(rv)); return rv; } /* TODO: ulimits? But that would restrict crm_ticket and handler * scripts, too! */ return 0; } static int do_server(int type) { int lock_fd = -1; int rv = -1; static char log_ent[128] = DAEMON_NAME "-"; rv = setup_config(type); if (rv < 0) goto out; if (!local) { log_error("Cannot find myself in the configuration."); exit(EXIT_FAILURE); } if (!daemonize) { if (daemon(0, 0) < 0) { perror("daemon error"); exit(EXIT_FAILURE); } } /* The lockfile must be written to _after_ the call to daemon(), so * that the lockfile contains the pid of the daemon, not the parent. */ lock_fd = create_lockfile(); if (lock_fd < 0) return lock_fd; strcat(log_ent, type_to_string(local->type)); cl_log_set_entity(log_ent); cl_log_enable_stderr(debug_level ? TRUE : FALSE); cl_log_set_facility(HA_LOG_FACILITY); cl_inherit_logging_environment(0); log_info("BOOTH %s daemon is starting, node id is 0x%08X (%d).", type_to_string(local->type), local->site_id, local->site_id); signal(SIGUSR1, (__sighandler_t)tickets_log_info); set_scheduler(); set_oom_adj(-16); set_proc_title("%s %s for [%s]:%d", DAEMON_NAME, type_to_string(local->type), local->addr_string, booth_conf->port); rv = limit_this_process(); if (rv) return rv; rv = loop(lock_fd); out: if (lock_fd >= 0) { /* We might not be able to delete it, but at least * make it empty. */ rv = ftruncate(lock_fd, 0); (void)rv; unlink_lockfile(lock_fd); } return rv; } static int do_client(void) { int rv = -1; rv = setup_config(CLIENT); if (rv < 0) { log_error("cannot read config"); goto out; } switch (cl.op) { case OP_LIST: rv = query_get_string_answer(CMD_LIST); break; case OP_GRANT: rv = do_grant(); break; case OP_REVOKE: rv = do_revoke(); break; } out: return rv; } int main(int argc, char *argv[], char *envp[]) { int rv; init_set_proc_title(argc, argv, envp); memset(&cl, 0, sizeof(cl)); strncpy(cl.configfile, BOOTH_DEFAULT_CONF, BOOTH_PATH_LEN - 1); cl.lockfile[0] = 0; debug_level = 0; cl_log_set_entity("booth"); cl_log_enable_stderr(TRUE); cl_log_set_facility(0); rv = read_arguments(argc, argv); if (rv < 0) goto out; switch (cl.type) { case STATUS: rv = do_status(cl.type); break; case ARBITRATOR: case DAEMON: case SITE: rv = do_server(cl.type); break; case CLIENT: rv = do_client(); break; } out: /* Normalize values. 0x100 would be seen as "OK" by waitpid(). */ return (rv >= 0 && rv < 0x70) ? rv : 1; } diff --git a/src/transport.c b/src/transport.c index cb4b3a5..fe96f34 100644 --- a/src/transport.c +++ b/src/transport.c @@ -1,732 +1,733 @@ /* * Copyright (C) 2011 Jiaju Zhang * Copyright (C) 2013-2014 Philipp Marek * * This program 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 2.1 of the License, or (at your option) any later version. * * This software 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 this library; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include "booth.h" #include "inline-fn.h" #include "log.h" #include "config.h" #include "ticket.h" #include "transport.h" #define BOOTH_IPADDR_LEN (sizeof(struct in6_addr)) #define NETLINK_BUFSIZE 16384 #define SOCKET_BUFFER_SIZE 160000 #define FRAME_SIZE_MAX 10000 struct booth_site *local = NULL; static int (*deliver_fn) (void *msg, int msglen); static void parse_rtattr(struct rtattr *tb[], int max, struct rtattr *rta, int len) { while (RTA_OK(rta, len)) { if (rta->rta_type <= max) tb[rta->rta_type] = rta; rta = RTA_NEXT(rta,len); } } static int find_address(unsigned char ipaddr[BOOTH_IPADDR_LEN], int family, int prefixlen, int fuzzy_allowed, struct booth_site **me, int *address_bits_matched) { int i; struct booth_site *node; int bytes, bits_left, mask; unsigned char node_bits, ip_bits; uint8_t *n_a; int matched, did_match; bytes = prefixlen / 8; bits_left = prefixlen % 8; /* One bit left to check means ignore 7 lowest bits. */ mask = ~( (1 << (8 - bits_left)) -1); did_match = 0; for (i = 0; i < booth_conf->site_count; i++) { node = booth_conf->site + i; if (family != node->family) continue; n_a = node_to_addr_pointer(node); for(matched = 0; matched < node->addrlen; matched++) if (ipaddr[matched] != n_a[matched]) break; if (matched == node->addrlen) { /* Full match. */ *address_bits_matched = matched * 8; found: *me = node; did_match = 1; continue; } if (!fuzzy_allowed) continue; /* Check prefix, whole bytes */ if (matched < bytes) continue; if (matched * 8 < *address_bits_matched) continue; if (!bits_left) goto found; node_bits = n_a[bytes]; ip_bits = ipaddr[bytes]; if (((node_bits ^ ip_bits) & mask) == 0) { /* _At_least_ prefixlen bits matched. */ *address_bits_matched = prefixlen; goto found; } } return did_match; } int _find_myself(int family, struct booth_site **mep, int fuzzy_allowed); int _find_myself(int family, struct booth_site **mep, int fuzzy_allowed) { int fd; struct sockaddr_nl nladdr; struct booth_site *me; unsigned char ipaddr[BOOTH_IPADDR_LEN]; static char rcvbuf[NETLINK_BUFSIZE]; struct { struct nlmsghdr nlh; struct rtgenmsg g; } req; int address_bits_matched; if (local) goto found; me = NULL; address_bits_matched = 0; if (mep) *mep = NULL; fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE); if (fd < 0) { log_error("failed to create netlink socket"); return 0; } setsockopt(fd, SOL_SOCKET, SO_RCVBUF, &rcvbuf, sizeof(rcvbuf)); memset(&nladdr, 0, sizeof(nladdr)); nladdr.nl_family = AF_NETLINK; memset(&req, 0, sizeof(req)); req.nlh.nlmsg_len = sizeof(req); req.nlh.nlmsg_type = RTM_GETADDR; req.nlh.nlmsg_flags = NLM_F_ROOT|NLM_F_MATCH|NLM_F_REQUEST; req.nlh.nlmsg_pid = 0; req.nlh.nlmsg_seq = 1; req.g.rtgen_family = family; if (sendto(fd, (void *)&req, sizeof(req), 0, (struct sockaddr*)&nladdr, sizeof(nladdr)) < 0) { close(fd); log_error("failed to send data to netlink socket"); return 0; } while (1) { int status; struct nlmsghdr *h; struct iovec iov = { rcvbuf, sizeof(rcvbuf) }; struct msghdr msg = { (void *)&nladdr, sizeof(nladdr), &iov, 1, NULL, 0, 0 }; status = recvmsg(fd, &msg, 0); if (!status) { close(fd); log_error("failed to recvmsg from netlink socket"); return 0; } h = (struct nlmsghdr *)rcvbuf; if (h->nlmsg_type == NLMSG_DONE) break; if (h->nlmsg_type == NLMSG_ERROR) { close(fd); log_error("netlink socket recvmsg error"); return 0; } while (NLMSG_OK(h, status)) { if (h->nlmsg_type == RTM_NEWADDR) { struct ifaddrmsg *ifa = NLMSG_DATA(h); struct rtattr *tb[IFA_MAX+1]; int len = h->nlmsg_len - NLMSG_LENGTH(sizeof(*ifa)); memset(tb, 0, sizeof(tb)); parse_rtattr(tb, IFA_MAX, IFA_RTA(ifa), len); memset(ipaddr, 0, BOOTH_IPADDR_LEN); /* prefer IFA_LOCAL if it exists, for p-t-p * interfaces, otherwise use IFA_ADDRESS */ if (tb[IFA_LOCAL]) { memcpy(ipaddr, RTA_DATA(tb[IFA_LOCAL]), BOOTH_IPADDR_LEN); } else { memcpy(ipaddr, RTA_DATA(tb[IFA_ADDRESS]), BOOTH_IPADDR_LEN); } /* First try with exact addresses, then optionally with subnet matching. */ if (ifa->ifa_prefixlen > address_bits_matched) find_address(ipaddr, ifa->ifa_family, ifa->ifa_prefixlen, fuzzy_allowed, &me, &address_bits_matched); } h = NLMSG_NEXT(h, status); } } close(fd); if (!me) return 0; me->local = 1; local = me; found: if (mep) *mep = local; return 1; } int find_myself(struct booth_site **mep, int fuzzy_allowed) { return _find_myself(AF_INET6, mep, fuzzy_allowed) || _find_myself(AF_INET, mep, fuzzy_allowed); } /** Checks the header fields for validity. * cf. init_header(). * For @len_incl_data < 0 the length is not checked. * Return <0 if error, else bytes read. */ int check_boothc_header(struct boothc_header *h, int len_incl_data) { int l; if (h->magic != htonl(BOOTHC_MAGIC)) { log_error("magic error %x", ntohl(h->magic)); return -EINVAL; } if (h->version != htonl(BOOTHC_VERSION)) { log_error("version error %x", ntohl(h->version)); return -EINVAL; } l = ntohl(h->length); if (l < sizeof(*h)) { log_error("length %d out of range", l); return -EINVAL; } if (len_incl_data < 0) return 0; if (l != len_incl_data) { log_error("length error - got %d, wanted %d", l, len_incl_data); return -EINVAL; } return len_incl_data; } static void process_tcp_listener(int ci) { int fd, i, one = 1; socklen_t addrlen = sizeof(struct sockaddr); struct sockaddr addr; fd = accept(clients[ci].fd, &addr, &addrlen); if (fd < 0) { log_error("process_tcp_listener: accept error %d %d", fd, errno); return; } setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, (char *)&one, sizeof(one)); i = client_add(fd, clients[ci].transport, process_connection, NULL); log_debug("client connection %d fd %d", i, fd); } -static int setup_tcp_listener(void) +int setup_tcp_listener(int test_only) { int s, rv; int one = 1; s = socket(local->family, SOCK_STREAM, 0); if (s == -1) { log_error("failed to create tcp socket %s", strerror(errno)); return s; } rv = setsockopt(s, SOL_SOCKET, SO_REUSEADDR, (char *)&one, sizeof(one)); if (rv == -1) { log_error("failed to set the SO_REUSEADDR option"); return rv; } rv = bind(s, &local->sa6, local->saddrlen); + if (test_only) { + rv = (rv == -1) ? errno : 0; + close(s); + return rv; + } + if (rv == -1) { log_error("failed to bind socket %s", strerror(errno)); return rv; } rv = listen(s, 5); if (rv == -1) { log_error("failed to listen on socket %s", strerror(errno)); return rv; } return s; } static int booth_tcp_init(void * unused __attribute__((unused))) { int rv; if (get_local_id() < 0) return -1; - rv = setup_tcp_listener(); + rv = setup_tcp_listener(0); if (rv < 0) return rv; client_add(rv, booth_transport + TCP, process_tcp_listener, NULL); return 0; } static int connect_nonb(int sockfd, const struct sockaddr *saptr, socklen_t salen, int sec) { int flags, n, error; socklen_t len; fd_set rset, wset; struct timeval tval; flags = fcntl(sockfd, F_GETFL, 0); fcntl(sockfd, F_SETFL, flags | O_NONBLOCK); error = 0; if ( (n = connect(sockfd, saptr, salen)) < 0) if (errno != EINPROGRESS) return -1; if (n == 0) goto done; /* connect completed immediately */ FD_ZERO(&rset); FD_SET(sockfd, &rset); wset = rset; tval.tv_sec = sec; tval.tv_usec = 0; if ((n = select(sockfd + 1, &rset, &wset, NULL, sec ? &tval : NULL)) == 0) { /* leave outside function to close */ /* timeout */ /* close(sockfd); */ errno = ETIMEDOUT; return -1; } if (FD_ISSET(sockfd, &rset) || FD_ISSET(sockfd, &wset)) { len = sizeof(error); if (getsockopt(sockfd, SOL_SOCKET, SO_ERROR, &error, &len) < 0) return -1; /* Solaris pending error */ } else { log_error("select error: sockfd not set"); return -1; } done: fcntl(sockfd, F_SETFL, flags); /* restore file status flags */ if (error) { /* leave outside function to close */ /* close(sockfd); */ errno = error; return -1; } return 0; } int booth_tcp_open(struct booth_site *to) { int s, rv; if (to->tcp_fd >= STDERR_FILENO) goto found; s = socket(to->family, SOCK_STREAM, 0); if (s == -1) { log_error("cannot create socket of family %d", to->family); return -1; } rv = connect_nonb(s, (struct sockaddr *)&to->sa6, to->saddrlen, 10); if (rv == -1) { if( errno == ETIMEDOUT) log_error("connect to \"%s\" got a timeout", to->addr_string); else log_error("connect to \"%s\" got an error: %s", to->addr_string, strerror(errno)); goto error; } to->tcp_fd = s; found: return 1; error: if (s >= 0) close(s); return -1; } int booth_tcp_send(struct booth_site *to, void *buf, int len) { return do_write(to->tcp_fd, buf, len); } static int booth_tcp_recv(struct booth_site *from, void *buf, int len) { int got; /* Needs timeouts! */ got = do_read(from->tcp_fd, buf, len); if (got < 0) return got; return len; } static int booth_tcp_close(struct booth_site *to) { if (to) { if (to->tcp_fd > STDERR_FILENO) close(to->tcp_fd); to->tcp_fd = -1; } return 0; } static int booth_tcp_exit(void) { return 0; } -int setup_udp_server(int try_only) +static int setup_udp_server(void) { int rv, fd; int one = 1; unsigned int recvbuf_size; fd = socket(local->family, SOCK_DGRAM, 0); if (fd == -1) { log_error("failed to create UDP socket %s", strerror(errno)); goto ex; } rv = fcntl(fd, F_SETFL, O_NONBLOCK); if (rv == -1) { log_error("failed to set non-blocking operation " "on UDP socket: %s", strerror(errno)); goto ex; } rv = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, (char *)&one, sizeof(one)); if (rv == -1) { log_error("failed to set the SO_REUSEADDR option"); goto ex; } rv = bind(fd, (struct sockaddr *)&local->sa6, local->saddrlen); - if (try_only) { - rv = (rv == -1) ? errno : 0; - close(fd); - return rv; - } if (rv == -1) { log_error("failed to bind UDP socket to [%s]:%d: %s", local->addr_string, booth_conf->port, strerror(errno)); goto ex; } recvbuf_size = SOCKET_BUFFER_SIZE; rv = setsockopt(fd, SOL_SOCKET, SO_RCVBUF, &recvbuf_size, sizeof(recvbuf_size)); if (rv == -1) { log_error("failed to set recvbuf size"); goto ex; } local->udp_fd = fd; return 0; ex: if (fd >= 0) close(fd); return -1; } /* Receive/process callback for UDP */ static void process_recv(int ci) { struct sockaddr_storage sa; int rv; socklen_t sa_len; char buffer[256]; /* Used for unit tests */ struct boothc_ticket_msg *msg; sa_len = sizeof(sa); msg = (void*)buffer; rv = recvfrom(clients[ci].fd, buffer, sizeof(buffer), MSG_NOSIGNAL | MSG_DONTWAIT, (struct sockaddr *)&sa, &sa_len); if (rv == -1) return; deliver_fn(msg, rv); } static int booth_udp_init(void *f) { int rv; - rv = setup_udp_server(0); + rv = setup_udp_server(); if (rv < 0) return rv; deliver_fn = f; client_add(local->udp_fd, booth_transport + UDP, process_recv, NULL); return 0; } int booth_udp_send(struct booth_site *to, void *buf, int len) { int rv; rv = sendto(local->udp_fd, buf, len, MSG_NOSIGNAL, (struct sockaddr *)&to->sa6, to->saddrlen); if (rv == len) { rv = 0; } else if (rv < 0) { rv = errno; log_error("Cannot send to \"%s\": %d %s", to->addr_string, errno, strerror(errno)); } else { rv = EBUSY; log_error("Packet sent to \"%s\" got truncated", to->addr_string); } return rv; } static int booth_udp_broadcast(void *buf, int len) { int i, rv, rvs; struct booth_site *site; if (!booth_conf || !booth_conf->site_count) return -1; rvs = 0; foreach_node(i, site) { if (site != local) { rv = booth_udp_send(site, buf, len); if (!rvs) rvs = rv; } } return rvs; } static int booth_udp_exit(void) { return 0; } /* SCTP transport layer has not been developed yet */ static int booth_sctp_init(void *f __attribute__((unused))) { return 0; } static int booth_sctp_send(struct booth_site * to __attribute__((unused)), void *buf __attribute__((unused)), int len __attribute__((unused))) { return 0; } static int booth_sctp_broadcast(void *buf __attribute__((unused)), int len __attribute__((unused))) { return 0; } static int return_0_booth_site(struct booth_site *v __attribute((unused))) { return 0; } static int return_0(void) { return 0; } const struct booth_transport booth_transport[TRANSPORT_ENTRIES] = { [TCP] = { .name = "TCP", .init = booth_tcp_init, .open = booth_tcp_open, .send = booth_tcp_send, .recv = booth_tcp_recv, .close = booth_tcp_close, .exit = booth_tcp_exit }, [UDP] = { .name = "UDP", .init = booth_udp_init, .open = return_0_booth_site, .send = booth_udp_send, .close = return_0_booth_site, .broadcast = booth_udp_broadcast, .exit = booth_udp_exit }, [SCTP] = { .name = "SCTP", .init = booth_sctp_init, .open = return_0_booth_site, .send = booth_sctp_send, .broadcast = booth_sctp_broadcast, .exit = return_0, } }; const struct booth_transport *local_transport = booth_transport+TCP; int send_header_only(int fd, struct boothc_header *hdr) { int rv; rv = do_write(fd, hdr, sizeof(*hdr)); return rv; } int send_ticket_msg(int fd, struct boothc_ticket_msg *msg) { int rv; rv = do_write(fd, msg, sizeof(*msg)); return rv; } int send_header_plus(int fd, struct boothc_header *hdr, void *data, int len) { int rv; int l; if (data == hdr->data) { l = sizeof(*hdr) + len; assert(l == ntohl(hdr->length)); /* One struct */ rv = do_write(fd, hdr, l); } else { /* Header and data in two locations */ rv = send_header_only(fd, hdr); if (rv >= 0 && len) rv = do_write(fd, data, len); } return rv; } diff --git a/src/transport.h b/src/transport.h index 9130bab..a5ae0ea 100644 --- a/src/transport.h +++ b/src/transport.h @@ -1,77 +1,77 @@ /* * Copyright (C) 2011 Jiaju Zhang * Copyright (C) 2013 Philipp Marek * * This program 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 2.1 of the License, or (at your option) any later version. * * This software 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 this library; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ #ifndef _TRANSPORT_H #define _TRANSPORT_H #include "booth.h" typedef enum { TCP = 1, UDP, SCTP, TRANSPORT_ENTRIES, } transport_layer_t; typedef enum { ARBITRATOR = 0x50, SITE, CLIENT, DAEMON, STATUS, } action_t; struct booth_transport { const char *name; int (*init) (void *); int (*open) (struct booth_site *); int (*send) (struct booth_site *, void *, int); int (*recv) (struct booth_site *, void *, int); int (*broadcast) (void *, int); int (*close) (struct booth_site *); int (*exit) (void); }; extern const struct booth_transport booth_transport[TRANSPORT_ENTRIES]; int find_myself(struct booth_site **me, int fuzzy_allowed); int check_boothc_header(struct boothc_header *data, int len_incl_data); -int setup_udp_server(int try_only); +int setup_tcp_listener(int test_only); int booth_udp_send(struct booth_site *to, void *buf, int len); int booth_tcp_open(struct booth_site *to); int booth_tcp_send(struct booth_site *to, void *buf, int len); inline static void * node_to_addr_pointer(struct booth_site *node) { switch (node->family) { case AF_INET: return &node->sa4.sin_addr; case AF_INET6: return &node->sa6.sin6_addr; } return NULL; } extern const struct booth_transport *local_transport; int send_header_only(int fd, struct boothc_header *hdr); int send_header_plus(int fd, struct boothc_header *hdr, void *data, int len); int send_ticket_msg(int fd, struct boothc_ticket_msg *msg); #endif /* _TRANSPORT_H */